1 2021-12-21 Jiang Haochen <haochen.jiang@intel.com>
3 * config/i386/bmiintrin.h (_tzcnt_u16): New intrinsic.
7 2021-12-21 Xionghu Luo <luoxhu@linux.ibm.com>
9 * tree-ssa-loop-im.c (bb_colder_than_loop_preheader): New
11 (get_coldest_out_loop): New function.
12 (determine_max_movement): Use get_coldest_out_loop.
13 (move_computations_worker): Adjust and fix iteration udpate.
14 (class ref_in_loop_hot_body): New functor.
15 (ref_in_loop_hot_body::operator): New.
16 (can_sm_ref_p): Use for_all_locs_in_loop.
17 (fill_coldest_and_hotter_out_loop): New.
18 (tree_ssa_lim_finalize): Free coldest_outermost_loop and
19 hotter_than_inner_loop.
20 (loop_invariant_motion_in_fun): Call fill_coldest_and_hotter_out_loop.
22 2021-12-21 Xionghu Luo <luoxhu@linux.ibm.com>
24 * tree-ssa-loop-split.c (split_loop): Fix incorrect
25 profile_count and probability.
26 (do_split_loop_on_cond): Likewise.
28 2021-12-21 Xionghu Luo <luoxhu@linux.ibm.com>
31 * predict.c (predict_extra_loop_exits): Add loop parameter.
32 (predict_loops): Call with loop argument.
34 2021-12-21 Xionghu Luo <luoxhu@linux.ibm.com>
36 * config/rs6000/altivec.md (altivec_vaddu<VI_char>s): Replace
37 UNSPEC_VADDU with us_plus.
38 (altivec_vadds<VI_char>s): Replace UNSPEC_VADDS with ss_plus.
39 (altivec_vsubu<VI_char>s): Replace UNSPEC_VSUBU with us_minus.
40 (altivec_vsubs<VI_char>s): Replace UNSPEC_VSUBS with ss_minus.
41 (altivec_abss_<mode>): Likewise.
43 2021-12-20 Uroš Bizjak <ubizjak@gmail.com>
46 * config/i386/sse.md (<sse2p4_1>_pinsr<ssemodesuffix>): Add
47 earlyclobber to (x,x,x,i) alternative.
48 (<sse2p4_1>_pinsr<ssemodesuffix> peephole2): Remove.
49 (<sse2p4_1>_pinsr<ssemodesuffix> splitter): Use output
50 operand as a temporary register. Split after reload_completed.
52 2021-12-20 Roger Sayle <roger@nextmovesoftware.com>
53 Uroš Bizjak <ubizjak@gmail.com>
55 * config/i386/i386.md (any_mul_highpart): New code iterator.
56 (sgnprefix, s): Add attribute support for [su]mul_highpart.
57 (<s>mul<mode>3_highpart): Delete expander.
58 (<s>mul<mode>3_highpart, <s>mulsi32_highpart_zext):
59 New define_insn patterns.
60 (define_peephole2): Tweak the register allocation for the above
61 instructions after reload.
63 2021-12-20 Joel Sherrill <joel@rtems.org>
65 * config.gcc: Obsolete m32c-*-rtems* target.
67 2021-12-20 Martin Liska <mliska@suse.cz>
69 * opts.c (default_options_optimization): Support -Oz in -Ox option hints.
71 2021-12-20 Jan Hubicka <hubicka@ucw.cz>
74 * ipa-modref.c (modref_eaf_analysis::analyze_ssa_name): Add deferred
76 (modref_eaf_analysis::propagate): Use it.
78 2021-12-20 liuhongt <hongtao.liu@intel.com>
81 * config/i386/sse.md (*bit_and_float_vector_all_ones): New
84 2021-12-19 Jan Hubicka <hubicka@ucw.cz>
87 * ipa-modref.c (modref_merge_call_site_flags): Fix early exit condition
89 2021-12-19 Patrick Palka <ppalka@redhat.com>
91 * print-tree.c (print_node) <case tcc_declaration>: Dump
94 2021-12-19 Andrew Pinski <apinski@marvell.com>
96 * tree-ssa-phiopt.c (gimple_simplify_phiopt): Annotate the
97 new sequence with the location of the conditional statement.
99 2021-12-18 Roger Sayle <roger@nextmovesoftware.com>
100 Uroš Bizjak <ubizjak@gmail.com>
103 * config/i386/i386.md (any_or_plus): New code iterator.
104 (define_split): Split (HI<<32)|zext(LO) into piece-wise
105 move instructions on !TARGET_64BIT.
106 * config/i386/sse.md (*vec_extractv4si_0_zext_sse4):
107 Restrict to TARGET_64BIT.
109 2021-12-18 Roger Sayle <roger@nextmovesoftware.com>
112 * common.opt (Oz): New command line option.
113 * doc/invoke.texi: Document the new -Oz option.
114 * lto-wrapper.c (merge_and_complain, append_compiler_options):
115 Treat OPT_Oz as synonymous with OPT_Os.
116 * optc-save-gen.awk: Increase maximum value of optimize_size to 2.
117 * opts.c (default_options_optimization) [OPT_Oz]: Handle OPT_Oz
118 just like OPT_Os, except set opt->x_optimize_size to 2.
119 (common_handle_option): Skip OPT_Oz just like OPT_Os.
120 * config/i386/i386.md (*movdi_internal): Use a push/pop sequence
121 for suitable SImode TYPE_IMOV moves when optimize_size > 1.
122 (*movsi_internal): Likewise.
124 2021-12-18 Siddhesh Poyarekar <siddhesh@gotplt.org>
126 PR tree-optimization/103759
127 * tree-object-size.c (unknown, initval): Remove functions.
128 (size_unknown, size_initval, size_unknown_p): Operate directly
131 2021-12-18 François-Xavier Coudert <fxcoudert@gcc.gnu.org>
133 * config/darwin-driver.c: Make version code more future-proof.
134 * config.gcc: Homogeneize darwin versions.
135 * configure.ac: Homogeneize darwin versions.
136 * configure: Regenerate.
138 2021-12-17 Marek Polacek <polacek@redhat.com>
141 * attribs.c (handle_ignored_attributes_option): Create the fake
142 attribute with max_length == -2.
143 (attribute_ignored_p): New overloads.
144 * attribs.h (attribute_ignored_p): Declare them.
145 * tree-core.h (struct attribute_spec): Document that max_length
148 2021-12-17 Segher Boessenkool <segher@kernel.crashing.org>
151 * config/rs6000/rs6000-builtins.def (__builtin_darn): Expand to
152 darn_64_di. Add {32bit} attribute. Return long.
153 (__builtin_darn_32): Expand to darn_32_di. Add {32bit} attribute.
155 (__builtin_darn_raw): Expand to darn_raw_di. Add {32bit} attribute.
157 * config/rs6000/rs6000-call.c (rs6000_expand_builtin): Expand the darn
158 builtins to the _si variants for -m32.
159 * config/rs6000/rs6000.md (UNSPECV_DARN_32, UNSPECV_DARN_RAW): Delete.
160 (UNSPECV_DARN): Update comment.
161 (darn_32, darn_raw, darn): Delete.
162 (darn_32_<mode>, darn_64_<mode>, darn_raw_<mode> for GPR): New.
163 (@darn<mode> for GPR): New.
165 2021-12-17 Bill Schmidt <wschmidt@linux.ibm.com>
167 * config/rs6000/rs6000-overload.def (__builtin_vec_promote): Add second
170 2021-12-17 Richard Sandiford <richard.sandiford@arm.com>
172 PR tree-optimization/103744
173 * tree-vect-stmts.c (vectorizable_load): Handle multi-vector
176 2021-12-17 Martin Liska <mliska@suse.cz>
178 * doc/invoke.texi: Rename to -fstack-protector.
180 2021-12-17 Martin Liska <mliska@suse.cz>
182 * doc/invoke.texi: Fix spelling issues.
184 2021-12-17 Tamar Christina <tamar.christina@arm.com>
186 PR tree-optimization/103741
187 * tree-vect-stmts.c (vectorizable_operation): Check for boolean.
189 2021-12-17 Iain Sandoe <iain@sandoe.co.uk>
191 * config/darwin-driver.c (darwin_driver_init): Exit from the
192 option handling early if the command line is definitely enpty.
193 * config/darwin.h (SUBTARGET_DRIVER_SELF_SPECS): Remove
194 setting for the default content of weak_reference_mismatches.
196 2021-12-17 Iain Sandoe <iain@sandoe.co.uk>
198 * config/rs6000/darwin.h: Drop trailing _x from the
199 builtin_decls array name.
201 2021-12-17 Haochen Jiang <haochen.jiang@intel.com>
204 * config/i386/sse.md (*avx_cmp<mode>3_lt, *avx_cmp<mode>3_ltint):
205 Remove MEM_P restriction and add force_reg for operands[2].
206 (*avx_cmp<mode>3_ltint_not): Add new define_insn_and_split.
208 2021-12-17 Siddhesh Poyarekar <siddhesh@gotplt.org>
210 * builtins.def (BUILT_IN_DYNAMIC_OBJECT_SIZE): New builtin.
211 * tree-object-size.h: Move object size type bits enum from
212 tree-object-size.c and add new value OST_DYNAMIC.
213 * builtins.c (expand_builtin, fold_builtin_2): Handle it.
214 (fold_builtin_object_size): Handle new builtin and adjust for
215 change to compute_builtin_object_size.
216 * tree-object-size.c: Include builtins.h.
217 (compute_builtin_object_size): Adjust.
218 (early_object_sizes_execute_one,
219 dynamic_object_sizes_execute_one): New functions.
220 (object_sizes_execute): Rename insert_min_max_p argument to
221 early. Handle BUILT_IN_DYNAMIC_OBJECT_SIZE and call the new
223 * doc/extend.texi (__builtin_dynamic_object_size): Document new
226 2021-12-17 Siddhesh Poyarekar <siddhesh@gotplt.org>
228 * tree-object-size.h (compute_builtin_object_size): Return tree
229 instead of HOST_WIDE_INT.
230 * builtins.c (fold_builtin_object_size): Adjust.
231 * gimple-fold.c (gimple_fold_builtin_strncat): Likewise.
232 * ubsan.c (instrument_object_size): Likewise.
233 * tree-object-size.c (object_size): New structure.
234 (object_sizes): Change type to vec<object_size>.
235 (initval): New function.
237 (size_unknown_p, size_initval, size_unknown): New functions.
238 (object_sizes_unknown_p): Use it.
239 (object_sizes_get): Return tree.
240 (object_sizes_initialize): Rename from object_sizes_set_force
241 and set VAL parameter type as tree. Add new parameter WHOLEVAL.
242 (object_sizes_set): Set VAL parameter type as tree and adjust
243 implementation. Add new parameter WHOLEVAL.
244 (size_for_offset): New function.
245 (decl_init_size): Adjust comment.
246 (addr_object_size): Change PSIZE parameter to tree and adjust
247 implementation. Add new parameter PWHOLESIZE.
248 (alloc_object_size): Return tree.
249 (compute_builtin_object_size): Return tree in PSIZE.
250 (expr_object_size, call_object_size, unknown_object_size):
251 Adjust for object_sizes_set change.
252 (merge_object_sizes): Drop OFFSET parameter and adjust
253 implementation for tree change.
254 (plus_stmt_object_size): Call collect_object_sizes_for directly
255 instead of merge_object_size and call size_for_offset to get net
257 (cond_expr_object_size, collect_object_sizes_for,
258 object_sizes_execute): Adjust for change of type from
259 HOST_WIDE_INT to tree.
260 (check_for_plus_in_loops_1): Likewise and skip non-positive
263 2021-12-17 Jason Merrill <jason@redhat.com>
266 * common.opt (fabi-version): Add v17.
268 2021-12-16 Uroš Bizjak <ubizjak@gmail.com>
271 * config/i386/i386.h (VALID_AVX256_REG_MODE): Add V16HFmode.
272 (VALID_AVX256_REG_OR_OI_VHF_MODE): Replace with ...
273 (VALID_AVX256_REG_OR_OI_MODE): ... this. Remove V16HFmode.
274 (VALID_AVX512F_SCALAR_MODE): Remove HImode and HFmode.
275 (VALID_AVX512FP16_SCALAR_MODE): New.
276 (VALID_AVX512F_REG_MODE): Add V32HFmode.
277 (VALID_SSE2_REG_MODE): Add V8HFmode, V4HFmode and V2HFmode.
278 (VALID_SSE2_REG_VHF_MODE): Remove.
279 (VALID_INT_MODE_P): Add V2HFmode.
280 * config/i386/i386.c (function_arg_advance_64):
281 Remove explicit mention of V16HFmode and V32HFmode.
282 (ix86_hard_regno_mode_ok): Remove explicit mention of XImode
283 and V32HFmode, use VALID_AVX512F_REG_OR_XI_MODE instead.
284 Use VALID_AVX512FP_SCALAR_MODE for TARGET_aVX512FP16.
285 Use VALID_AVX256_REG_OR_OI_MODE instead of
286 VALID_AVX256_REG_OR_OI_VHF_MODE and VALID_SSE2_REG_MODE instead
287 of VALID_SSE2_REG_VHF_MODE.
288 (ix86_set_reg_reg_cost): Remove usge of VALID_AVX512FP16_REG_MODE.
289 (ix86_vector_mode_supported): Ditto.
291 2021-12-16 Martin Liska <mliska@suse.cz>
294 * attribs.c (decl_attributes): Check if
295 target_option_current_node is changed.
297 2021-12-16 Martin Liska <mliska@suse.cz>
299 * doc/invoke.texi: Add missing dash.
301 2021-12-16 Bill Schmidt <wschmidt@linux.ibm.com>
303 * config/rs6000/rs6000-c.c (resolution): New enum.
304 (resolve_vec_mul): New function.
305 (resolve_vec_cmpne): Likewise.
306 (resolve_vec_adde_sube): Likewise.
307 (resolve_vec_addec_subec): Likewise.
308 (resolve_vec_splats): Likewise.
309 (resolve_vec_extract): Likewise.
310 (resolve_vec_insert): Likewise.
311 (resolve_vec_step): Likewise.
312 (find_instance): Likewise.
313 (altivec_resolve_overloaded_builtin): Many cleanups. Call factored-out
314 functions. Move variable declarations closer to uses. Add commentary.
315 Remove unnecessary levels of braces. Avoid use of gotos. Change
316 misleading variable names. Use switches over if-else-if chains.
318 2021-12-16 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
321 * config/aarch64/aarch64-simd.md (aarch64_movv8di): Allow big endian
322 targets to move V8DI.
324 2021-12-16 Tobias Burnus <tobias@codesourcery.com>
327 * gcc.c (check_offload_target_name): Add 'default' and 'disable'
328 to the candidate list.
330 2021-12-15 Iain Sandoe <iain@sandoe.co.uk>
332 * configure: Regenerate.
334 2021-12-15 Roger Sayle <roger@nextmovesoftware.com>
336 * config/nvptx/nvptx-opts.h (ptx_isa): PTX_ISA_SM75 and PTX_ISA_SM80
338 * config/nvptx/nvptx.opt: Add sm_75 and sm_80 to -misa.
339 * config/nvptx/nvptx.h (TARGET_SM75, TARGET_SM80):
340 New helper macros to conditionalize functionality on target ISA.
341 * config/nvptx/nvptx-c.c (nvptx_cpu_cpp_builtins): Add __PTX_SM__
342 support for the new ISA levels.
343 * config/nvptx/nvptx.c (nvptx_file_start): Add support for TARGET_SM75
345 * config/nvptx/nvptx.md (define_c_enum "unspec"): New UNSPEC_TANH.
346 (define_mode_iterator HSFM): New iterator for HFmode and SFmode.
347 (exp2hf2): New define_insn controlled by TARGET_SM75.
348 (tanh<mode>2): New define_insn controlled by TARGET_SM75.
349 (sminhf3, smaxhf3): New define_isnns controlled by TARGET_SM80.
351 2021-12-15 Tom de Vries <tdevries@suse.de>
353 * config/nvptx/nvptx-opts.h (enum ptx_version): Add PTX_VERSION_7_0.
354 * config/nvptx/nvptx.c (nvptx_file_start): Handle TARGET_PTX_7_0.
355 * config/nvptx/nvptx.h (TARGET_PTX_7_0): New macro.
356 * config/nvptx/nvptx.opt (ptx_version): Add 7.0.
358 2021-12-15 Richard Sandiford <richard.sandiford@arm.com>
359 Tamar Christina <tamar.christina@arm.com>
362 * config/aarch64/aarch64.c (aarch64_short_vector_p): Return false
363 for structure modes, rather than ignoring the type in that case.
365 2021-12-15 Tamar Christina <tamar.christina@arm.com>
367 PR rtl-optimization/103350
368 * ree.c (add_removable_extension): Don't stop at first definition but
371 2021-12-15 Jakub Jelinek <jakub@redhat.com>
374 * dwarf2cfi.c (dwf_cfa_reg): Remove gcc_assert.
375 (operator==, operator!=): New overloaded operators.
376 (dwarf2out_frame_debug_adjust_cfa, dwarf2out_frame_debug_cfa_offset,
377 dwarf2out_frame_debug_expr): Compare vars with cfa_reg type directly
378 with REG rtxes rather than with dwf_cfa_reg results on those REGs.
379 (create_cie_data): Use stack_pointer_rtx instead of
380 gen_rtx_REG (Pmode, STACK_POINTER_REGNUM).
381 (execute_dwarf2_frame): Use hard_frame_pointer_rtx instead of
382 gen_rtx_REG (Pmode, HARD_FRAME_POINTER_REGNUM).
384 2021-12-15 Martin Liska <mliska@suse.cz>
387 * config/i386/i386-builtins.c (fold_builtin_cpu): Compare to 0
388 as API expects that non-zero values are returned (do that
390 For "avx512vbmi2" argument, we return now 1 << 31, which is a
391 negative integer value.
393 2021-12-15 Haochen Jiang <haochen.jiang@intel.com>
396 * config/i386/predicates.md (const_vector_operand):
398 * config/i386/sse.md(<insn><mode>3<mask_name>):
399 Add new define_split below.
401 2021-12-15 Michael Meissner <meissner@the-meissners.org>
403 * config/rs6000/rs6000.md (UNSPEC_XXSPLTIDP_CONST): New unspec.
404 (UNSPEC_XXSPLTIW_CONST): New unspec.
405 (movsf_hardfloat): Add support for generating XXSPLTIDP.
406 (mov<mode>_hardfloat32): Likewise.
407 (mov<mode>_hardfloat64): Likewise.
408 (xxspltidp_<mode>_internal): New insns.
409 (xxspltiw_<mode>_internal): New insns.
410 (splitters for SF/DFmode): Add new splitters for XXSPLTIDP.
412 2021-12-15 Michael Meissner <meissner@the-meissners.org>
414 * config/rs6000/predicates.md (easy_fp_constant): Add support for
415 generating XXSPLTIDP.
416 (vsx_prefixed_constant): Likewise.
417 (easy_vector_constant): Likewise.
418 * config/rs6000/rs6000-protos.h (constant_generates_xxspltidp):
420 * config/rs6000/rs6000.c (output_vec_const_move): Add support for
421 generating XXSPLTIDP.
422 (prefixed_xxsplti_p): Likewise.
423 (constant_generates_xxspltidp): New function.
424 * config/rs6000/rs6000.opt (-msplat-float-constant): New debug option.
426 2021-12-15 Michael Meissner <meissner@linux.ibm.com>
428 * config/rs6000/constraints.md (eP): Update comment.
429 * config/rs6000/predicates.md (easy_fp_constant): Add support for
431 (vsx_prefixed_constant): New predicate.
432 (easy_vector_constant): Add support for
434 * config/rs6000/rs6000-protos.h (prefixed_xxsplti_p): New
436 (constant_generates_xxspltiw): Likewise.
437 * config/rs6000/rs6000.c (xxspltib_constant_p): Generate XXSPLTIW
438 if possible instead of XXSPLTIB and sign extending the constant.
439 (output_vec_const_move): Add support for XXSPLTIW.
440 (prefixed_xxsplti_p): New function.
441 (constant_generates_xxspltiw): New function.
442 * config/rs6000/rs6000.md (prefixed attribute): Add support to
443 mark XXSPLTI* instructions as being prefixed.
444 * config/rs6000/rs6000.opt (-msplat-word-constant): New debug
446 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Add support for
447 generating XXSPLTIW or XXSPLTIDP.
448 (vsx_mov<mode>_32bit): Likewise.
449 * doc/md.texi (PowerPC and IBM RS6000 constraints): Document the
452 2021-12-15 Michael Meissner <meissner@the-meissners.org>
454 * config/rs6000/constraints.md (eQ): New constraint.
455 * config/rs6000/predicates.md (easy_fp_constant): Add support for
456 generating the LXVKQ instruction.
457 (easy_vector_constant_ieee128): New predicate.
458 (easy_vector_constant): Add support for generating the LXVKQ
460 * config/rs6000/rs6000-protos.h (constant_generates_lxvkq): New
462 * config/rs6000/rs6000.c (output_vec_const_move): Add support for
464 (constant_generates_lxvkq): New function.
465 * config/rs6000/rs6000.opt (-mieee128-constant): New debug
467 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Add support for
469 (vsx_mov<mode>_32bit): Likewise.
470 * doc/md.texi (PowerPC and IBM RS6000 constraints): Document the
473 2021-12-15 Michael Meissner <meissner@the-meissners.org>
475 * config/rs6000/rs6000-protos.h (VECTOR_128BIT_BITS): New macro.
476 (VECTOR_128BIT_BYTES): Likewise.
477 (VECTOR_128BIT_HALF_WORDS): Likewise.
478 (VECTOR_128BIT_WORDS): Likewise.
479 (VECTOR_128BIT_DOUBLE_WORDS): Likewise.
480 (vec_const_128bit_type): New structure type.
481 (vec_const_128bit_to_bytes): New declaration.
482 * config/rs6000/rs6000.c (constant_int_to_128bit_vector): New
484 (constant_fp_to_128bit_vector): New helper function.
485 (vec_const_128bit_to_bytes): New function.
487 2021-12-15 Alexandre Oliva <oliva@adacore.com>
490 * builtins.c (try_store_by_multiple_pieces): Drop address
491 conversion to ptr_mode.
493 2021-12-15 Alexandre Oliva <oliva@adacore.com>
496 * builtins.c (try_store_by_multiple_pieces): Fail if min_len
497 is greater than max_len.
499 2021-12-14 liuhongt <hongtao.liu@intel.com>
502 * tree-ssa-ccp.c (optimize_atomic_bit_test_and): Check
503 is_gimple_assign before gimple_assign_rhs_code.
505 2021-12-14 Sören Tempel <soeren@soeren-tempel.net>
507 * ginclude/stddef.h (__DEFINED_ptrdiff_t): Add support for musl
508 libc typedef macro guard.
509 (__DEFINED_size_t): Ditto.
510 (__DEFINED_wchar_t): Ditto.
512 2021-12-14 JoJo R <rjiejie@linux.alibaba.com>
514 * regrename.c (find_rename_reg): Return satisfied regno
515 if instruction is noop move.
517 2021-12-14 Bill Schmidt <wschmidt@linux.ibm.com>
519 * config/rs6000/rs6000-c.c (altivec_build_resolved_builtin): Rename
520 rs6000_builtin_decls_x to rs6000_builtin_decls.
521 (altivec_resolve_overloaded_builtin): Likewise. Also rename
522 rs6000_builtin_info_x to rs6000_builtin_info.
523 * config/rs6000/rs6000-call.c (rs6000_invalid_builtin): Rename
524 rs6000_builtin_info_x to rs6000_builtin_info.
525 (rs6000_builtin_is_supported): Likewise.
526 (rs6000_gimple_fold_mma_builtin): Likewise. Also rename
527 rs6000_builtin_decls_x to rs6000_builtin_decls.
528 (rs6000_gimple_fold_builtin): Rename rs6000_builtin_info_x to
530 (cpu_expand_builtin): Likewise.
531 (rs6000_expand_builtin): Likewise.
532 (rs6000_init_builtins): Likewise. Also rename rs6000_builtin_decls_x
533 to rs6000_builtin_decls.
534 (rs6000_builtin_decl): Rename rs6000_builtin_decls_x to
535 rs6000_builtin_decls.
536 * config/rs6000/rs6000-gen-builtins.c (write_decls): In generated code,
537 rename rs6000_builtin_decls_x to rs6000_builtin_decls, and rename
538 rs6000_builtin_info_x to rs6000_builtin_info.
539 (write_bif_static_init): In generated code, rename
540 rs6000_builtin_info_x to rs6000_builtin_info.
541 (write_init_bif_table): In generated code, rename
542 rs6000_builtin_decls_x to rs6000_builtin_decls, and rename
543 rs6000_builtin_info_x to rs6000_builtin_info.
544 (write_init_ovld_table): In generated code, rename
545 rs6000_builtin_decls_x to rs6000_builtin_decls.
546 (write_init_file): Likewise.
547 * config/rs6000/rs6000.c (rs6000_builtin_vectorized_function):
549 (rs6000_builtin_md_vectorized_function): Likewise.
550 (rs6000_builtin_reciprocal): Likewise.
551 (add_condition_to_bb): Likewise.
552 (rs6000_atomic_assign_expand_fenv): Likewise.
554 2021-12-14 Bill Schmidt <wschmidt@linux.ibm.com>
556 * config/rs6000/rs6000-c.c (altivec_resolve_new_overloaded_builtin):
557 Remove forward declaration.
558 (rs6000_new_builtin_type_compatible): Rename to
559 rs6000_builtin_type_compatible.
560 (rs6000_builtin_type_compatible): Remove.
561 (altivec_resolve_overloaded_builtin): Remove.
562 (altivec_build_new_resolved_builtin): Rename to
563 altivec_build_resolved_builtin.
564 (altivec_resolve_new_overloaded_builtin): Rename to
565 altivec_resolve_overloaded_builtin. Remove static keyword. Adjust
566 called function names.
567 * config/rs6000/rs6000-call.c (rs6000_expand_new_builtin): Remove
569 (rs6000_gimple_fold_new_builtin): Likewise.
570 (rs6000_invalid_new_builtin): Rename to rs6000_invalid_builtin.
571 (rs6000_gimple_fold_builtin): Remove.
572 (rs6000_new_builtin_valid_without_lhs): Rename to
573 rs6000_builtin_valid_without_lhs.
574 (rs6000_new_builtin_is_supported): Rename to
575 rs6000_builtin_is_supported.
576 (rs6000_gimple_fold_new_mma_builtin): Rename to
577 rs6000_gimple_fold_mma_builtin.
578 (rs6000_gimple_fold_new_builtin): Rename to
579 rs6000_gimple_fold_builtin. Remove static keyword. Adjust called
581 (rs6000_expand_builtin): Remove.
582 (new_cpu_expand_builtin): Rename to cpu_expand_builtin.
583 (new_mma_expand_builtin): Rename to mma_expand_builtin.
584 (new_htm_spr_num): Rename to htm_spr_num.
585 (new_htm_expand_builtin): Rename to htm_expand_builtin. Change name
587 (rs6000_expand_new_builtin): Rename to rs6000_expand_builtin. Remove
588 static keyword. Adjust called function names.
589 (rs6000_new_builtin_decl): Rename to rs6000_builtin_decl. Remove
591 (rs6000_builtin_decl): Remove.
592 * config/rs6000/rs6000-gen-builtins.c (write_decls): In gnerated code,
593 rename rs6000_new_builtin_is_supported to rs6000_builtin_is_supported.
594 * config/rs6000/rs6000-internal.h (rs6000_invalid_new_builtin): Rename
595 to rs6000_invalid_builtin.
596 * config/rs6000/rs6000.c (rs6000_new_builtin_vectorized_function):
597 Rename to rs6000_builtin_vectorized_function.
598 (rs6000_new_builtin_md_vectorized_function): Rename to
599 rs6000_builtin_md_vectorized_function.
600 (rs6000_builtin_vectorized_function): Remove.
601 (rs6000_builtin_md_vectorized_function): Remove.
603 2021-12-14 Bill Schmidt <wschmidt@linux.ibm.com>
605 * config/rs6000/rs6000-builtin.def: Delete.
606 * config/rs6000/rs6000-call.c (builtin_compatibility): Delete.
607 (builtin_description): Delete.
608 (builtin_hash_struct): Delete.
609 (builtin_hasher): Delete.
610 (builtin_hash_table): Delete.
611 (builtin_hasher::hash): Delete.
612 (builtin_hasher::equal): Delete.
613 (rs6000_builtin_info_type): Delete.
614 (rs6000_builtin_info): Delete.
615 (bdesc_compat): Delete.
616 (bdesc_3arg): Delete.
617 (bdesc_4arg): Delete.
619 (bdesc_2arg): Delete.
620 (bdesc_altivec_preds): Delete.
622 (bdesc_1arg): Delete.
623 (bdesc_0arg): Delete.
626 (rs6000_overloaded_builtin_p): Delete.
627 (rs6000_overloaded_builtin_name): Delete.
628 (htm_spr_num): Delete.
629 (rs6000_builtin_is_supported_p): Delete.
630 (rs6000_gimple_fold_mma_builtin): Delete.
631 (gt-rs6000-call.h): Remove include directive.
632 * config/rs6000/rs6000-protos.h (rs6000_overloaded_builtin_p): Delete.
633 (rs6000_builtin_is_supported_p): Delete.
634 (rs6000_overloaded_builtin_name): Delete.
635 * config/rs6000/rs6000.c (rs6000_builtin_decls): Delete.
636 (rs6000_debug_reg_global): Remove reference to RS6000_BUILTIN_COUNT.
637 * config/rs6000/rs6000.h (rs6000_builtins): Delete.
638 (altivec_builtin_types): Delete.
639 (rs6000_builtin_decls): Delete.
640 * config/rs6000/t-rs6000 (TM_H): Don't add rs6000-builtin.def.
642 2021-12-14 Bill Schmidt <wschmidt@linux.ibm.com>
644 * config/rs6000/rs6000-builtin-new.def: Rename to...
645 * config/rs6000/rs6000-builtins.def: ...this.
646 * config/rs6000/rs6000-gen-builtins.c: Adjust header commentary.
647 * config/rs6000/t-rs6000 (EXTRA_GTYPE_DEPS): Rename
648 rs6000-builtin-new.def to rs6000-builtins.def.
649 (rs6000-builtins.c): Likewise.
651 2021-12-14 Bill Schmidt <wschmidt@linux.ibm.com>
653 * config/rs6000/rs6000-call.c (altivec_overloaded_builtins): Remove.
654 * config/rs6000/rs6000.h (altivec_overloaded_builtins): Remove.
656 2021-12-14 Peter Bergner <bergner@linux.ibm.com>
659 * config/rs6000/mma.md (UNSPEC_MMA_ASSEMBLE): Rename unspec from this...
660 (UNSPEC_VSX_ASSEMBLE): ...to this.
661 (UNSPECV_MMA_ASSEMBLE): New unspecv.
662 (vsx_assemble_pair): Use UNSPEC_VSX_ASSEMBLE.
663 (*vsx_assemble_pair): Likewise.
664 (mma_assemble_acc): Use UNSPECV_MMA_ASSEMBLE.
665 (*mma_assemble_acc): Likewise.
666 * config/rs6000/rs6000.c (rs6000_split_multireg_move): Handle
667 UNSPEC_VOLATILE. Use UNSPEC_VSX_ASSEMBLE and UNSPECV_MMA_ASSEMBLE.
669 2021-12-14 Uroš Bizjak <ubizjak@gmail.com>
672 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate)
673 <case E_V8HFmode>: Implement for TARGET_SSE2.
674 <case E_V16HFmode>: Implement for TARGET_AVX.
675 <case E_V32HFmode>: Implement for TARGET_AVX512F.
676 (ix86_expand_vector_set_var): Handle V32HFmode
677 without TARGET_AVX512BW.
678 (ix86_expand_vector_extract)
679 <case E_V8HFmode>: Implement for TARGET_SSE2.
680 <case E_V16HFmode>: Implement for TARGET_AVX.
681 <case E_V32HFmode>: Implement for TARGET_AVX512BW.
682 (expand_vec_perm_broadcast_1) <case E_V8HFmode>: New.
683 * config/i386/sse.md (VI12HF_AVX512VL): Remove
684 TARGET_AVX512FP16 condition.
687 (avx_vbroadcastf128_<mode>): Use V_256H mode iterator.
689 2021-12-14 Bill Schmidt <wschmidt@linux.ibm.com>
691 * config/rs6000/darwin.h (SUBTARGET_INIT_BUILTINS): Remove
692 test for new_builtins_are_live and simplify.
693 * config/rs6000/rs6000-c.c (altivec_build_resolved_builtin): Remove
695 (altivec_resolve_overloaded_builtin): Remove test for
696 new_builtins_are_live and simplify.
697 * config/rs6000/rs6000-call.c (altivec_init_builtins): Remove forward
699 (builtin_function_type): Likewise.
700 (rs6000_common_init_builtins): Likewise.
701 (htm_init_builtins): Likewise.
702 (mma_init_builtins): Likewise.
703 (def_builtin): Remove dead function.
704 (rs6000_expand_zeroop_builtin): Likewise.
705 (rs6000_expand_mtfsf_builtin): Likewise.
706 (rs6000_expand_mtfsb_builtin): Likewise.
707 (rs6000_expand_set_fpscr_rn_builtin): Likewise.
708 (rs6000_expand_set_fpscr_drn_builtin): Likewise.
709 (rs6000_expand_unop_builtin): Likewise.
710 (altivec_expand_abs_builtin): Likewise.
711 (rs6000_expand_binop_builtin): Likewise.
712 (altivec_expand_lxvr_builtin): Likewise.
713 (altivec_expand_lv_builtin): Likewise.
714 (altivec_expand_stxvl_builtin): Likewise.
715 (altivec_expand_stv_builtin): Likewise.
716 (mma_expand_builtin): Likewise.
717 (htm_expand_builtin): Likewise.
718 (cpu_expand_builtin): Likewise.
719 (rs6000_expand_quaternop_builtin): Likewise.
720 (rs6000_expand_ternop_builtin): Likewise.
721 (altivec_expand_dst_builtin): Likewise.
722 (altivec_expand_vec_sel_builtin): Likewise.
723 (altivec_expand_builtin): Likewise.
724 (rs6000_invalid_builtin): Likewise.
725 (rs6000_builtin_valid_without_lhs): Likewise.
726 (rs6000_gimple_fold_builtin): Remove test for new_builtins_are_live and
728 (rs6000_expand_builtin): Likewise.
729 (rs6000_init_builtins): Remove tests for new_builtins_are_live and
731 (rs6000_builtin_decl): Likewise.
732 (altivec_init_builtins): Remove dead function.
733 (mma_init_builtins): Likewise.
734 (htm_init_builtins): Likewise.
735 (builtin_quaternary_function_type): Likewise.
736 (builtin_function_type): Likewise.
737 (rs6000_common_init_builtins): Likewise.
738 * config/rs6000/rs6000-gen-builtins.c (write_header_file): Don't
739 declare new_builtins_are_live.
740 (write_init_bif_table): In generated code, remove test for
741 new_builtins_are_live and simplify.
742 (write_init_ovld_table): Likewise.
743 (write_init_file): Don't initialize new_builtins_are_live.
744 * config/rs6000/rs6000.c (rs6000_builtin_vectorized_function): Remove
745 test for new_builtins_are_live and simplify.
746 (rs6000_builtin_md_vectorized_function): Likewise.
747 (rs6000_builtin_reciprocal): Likewise.
748 (add_condition_to_bb): Likewise.
749 (rs6000_atomic_assign_expand_fenv): Likewise.
751 2021-12-14 Bill Schmidt <wschmidt@linux.ibm.com>
754 * config/rs6000/rs6000-builtin-new.def (__builtin_altivec_vcmpequd):
755 Move to power8-vector stanza.
756 (__builtin_altivec_vcmpequd_p): Likewise.
757 (__builtin_altivec_vcmpgtsd): Likewise.
758 (__builtin_altivec_vcmpgtsd_p): Likewise.
759 (__builtin_altivec_vcmpgtud): Likewise.
760 (__builtin_altivec_vcmpgtud_p): Likewise.
762 2021-12-14 Bill Schmidt <wschmidt@linux.ibm.com>
765 * config/rs6000/rs6000-builtin-new.def (__builtin_pack_longdouble): Add
767 (__builtin_unpack_longdouble): Likewise.
768 * config/rs6000/rs6000-call.c (rs6000_expand_new_builtin): Add special
769 handling for ibmld attribute.
770 * config/rs6000/rs6000-gen-builtins.c (attrinfo): Add isibmld.
771 (parse_bif_attrs): Handle ibmld.
772 (write_decls): Likewise.
773 (write_bif_static_init): Likewise.
775 2021-12-14 Jan Hubicka <hubicka@ucw.cz>
778 * ipa-modref-tree.c (modref_access_node::range_info_useful_p): Handle
779 MODREF_GLOBAL_MEMORY_PARM.
780 (modref_access_node::dump): Likewise.
781 (modref_access_node::get_call_arg): Likewise.
782 * ipa-modref-tree.h (enum modref_special_parms): Add
783 MODREF_GLOBAL_MEMORY_PARM.
784 (modref_access_node::useful_for_kill): Handle
785 MODREF_GLOBAL_MEMORY_PARM.
786 (modref:tree::merge): Add promote_unknown_to_global.
787 * ipa-modref.c (verify_arg):New function.
788 (may_access_nonescaping_parm_p): New function.
789 (modref_access_analysis::record_global_memory_load): New member
791 (modref_access_analysis::record_global_memory_store): Likewise.
792 (modref_access_analysis::process_fnspec): Distingush global and local
794 (modref_access_analysis::analyze_call): Likewise.
795 * tree-ssa-alias.c (ref_may_access_global_memory_p): New function.
796 (modref_may_conflict): Use it.
798 2021-12-14 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
800 * config/aarch64/aarch64-builtins.c (enum aarch64_builtins):
801 Define AARCH64_LS64_BUILTIN_LD64B, AARCH64_LS64_BUILTIN_ST64B,
802 AARCH64_LS64_BUILTIN_ST64BV, AARCH64_LS64_BUILTIN_ST64BV0.
803 (aarch64_init_ls64_builtin_decl): Helper function.
804 (aarch64_init_ls64_builtins): Helper function.
805 (aarch64_init_ls64_builtins_types): Helper function.
806 (aarch64_general_init_builtins): Init LS64 intrisics for
808 (aarch64_expand_builtin_ls64): LS64 intrinsics expander.
809 (aarch64_general_expand_builtin): Handle aarch64_expand_builtin_ls64.
810 (ls64_builtins_data): New helper struct.
811 (v8di_UP): New define.
812 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Define
814 * config/aarch64/aarch64.c (aarch64_classify_address): Enforce the
815 V8DI range (7-bit signed scaled) for both ends of the range.
816 * config/aarch64/aarch64-simd.md (movv8di): New pattern.
817 (aarch64_movv8di): New pattern.
818 * config/aarch64/aarch64.h (AARCH64_ISA_LS64): New define.
819 (TARGET_LS64): New define.
820 * config/aarch64/aarch64.md: Add UNSPEC_LD64B, UNSPEC_ST64B,
821 UNSPEC_ST64BV and UNSPEC_ST64BV0.
822 (ld64b): New define_insn.
823 (st64b): New define_insn.
824 (st64bv): New define_insn.
825 (st64bv0): New define_insn.
826 * config/aarch64/arm_acle.h (data512_t): New type derived from
828 (__arm_data512_t): New internal type.
829 (__arm_ld64b): New intrinsic.
830 (__arm_st64b): New intrinsic.
831 (__arm_st64bv): New intrinsic.
832 (__arm_st64bv0): New intrinsic.
833 * config/arm/types.md: Add new type ls64.
835 2021-12-14 Olivier Hainque <hainque@adacore.com>
837 * config/i386/t-vxworks: Drop the fPIC multilibs.
839 2021-12-14 Fred Konrad <konrad@adacore.com>
841 * config/rs6000/t-vxworks: Drop the fPIC multilib.
843 2021-12-13 Vladimir N. Makarov <vmakarov@redhat.com>
846 * ira-costs.c (record_operand_costs): Do not take pseudo class
847 calculated on the 1st iteration into account when processing move
848 involving the pseudo and a hard register.
850 2021-12-13 Roger Sayle <roger@nextmovesoftware.com>
852 * config/i386/i386.md (define_split any_or:SWI248 -> orb %?h):
853 Optimize the case where the integer constant operand is zero.
855 2021-12-13 Doug Rupp <rupp@adacore.com>
856 Olivier Hainque <hainque@adacore.com>
858 * config/vxworks.h (VXWORKS_LINK_OS_SPEC): New spec.
859 (VXWORKS_BASE_LINK_SPEC): New spec, using the former.
860 (VXWORKS_EXTRA_LINK_SPEC): New spec for old and new VxWorks.
861 (VXWORKS_LINK_SPEC): Combo of BASE and EXTRA specs.
862 * config/rs6000/vxworks.h (VXWORKS_LINK_OS_SPEC): Empty.
863 (LINK_OS_EXTRA_SPEC32): Use VXWORKS_LINK_SPEC.
864 (LINK_OS_EXTRA_SPEC64): Likewise.
866 2021-12-13 Fred Konrad <konrad@adacore.com>
867 Olivier Hainque <hainque@adacore.com>
869 * config/vxworks.h (VXWORKS_BASE_LIBS_RTP): Guard -lc_internal
870 on !shared+!non-static and document.
871 (VXWORKS_LIB_SPEC): Remove the bits intended to drag the
872 init/fini functions from libc_internal in the shared lib case.
873 (VX_CRTBEGIN_SPEC/VX_CRTEND_SPEC): Use vxcrtstuff objects also in
874 configurations with shared lib and INITFINI_ARRAY support.
876 2021-12-13 Fred Konrad <konrad@adacore.com>
877 Olivier Hainque <hainque@adacore.com>
879 * config/vx-common.h: Define REAL_LIBGCC_SPEC since the
880 '-non-static' option is not standard.
881 * config/vxworks.h (VXWORKS_LIBGCC_SPEC): Implement the LIBGCC_SPEC
882 since REAL_LIBGCC_SPEC is used now.
883 (STARTFILE_PREFIX_SPEC): Use the PIC VSB when building shared libraries
884 or non-static binaries.
886 2021-12-13 Jan Hubicka <hubicka@ucw.cz>
888 * common.opt: Add -fipa-strict-aliasing.
889 * doc/invoke.texi: Document -fipa-strict-aliasing.
890 * ipa-modref.c (modref_access_analysis::record_access): Honor
891 -fipa-strict-aliasing.
892 (modref_access_analysis::record_access_lto): Likewise.
894 2021-12-13 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
896 * config/aarch64/aarch64-arches.def (armv8.8-a): Define.
897 * config/aarch64/aarch64.h (AARCH64_FL_V8_8): Define.
898 (AARCH64_FL_FOR_ARCH8_8): Define.
899 * doc/invoke.texi: Document -march=armv8.8-a.
901 2021-12-13 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
903 * config/aarch64/aarch64.c (aarch64_expand_setmem_mops): Define.
904 (aarch64_expand_setmem): Adjust for TARGET_MOPS.
905 * config/aarch64/aarch64.h (CLEAR_RATIO): Adjust for TARGET_MOPS.
906 (SET_RATIO): Likewise.
907 * config/aarch64/aarch64.md ("unspec"): Add UNSPEC_SETMEM.
908 (aarch64_setmemdi): Define.
909 (setmemdi): Adjust for TARGET_MOPS.
910 * config/aarch64/aarch64.opt (aarch64-mops-memset-size-threshold):
913 2021-12-13 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
915 * config/aarch64/aarch64.md (aarch64_movmemdi): Define.
917 (unspec): Add UNSPEC_MOVMEM.
918 * config/aarch64/aarch64.opt (aarch64-mops-memmove-size-threshold):
921 2021-12-13 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
923 * config/aarch64/aarch64-option-extensions.def (mops): Define.
924 * config/aarch64/aarch64.c (aarch64_expand_cpymem_mops): Define.
925 (aarch64_expand_cpymem): Define.
926 * config/aarch64/aarch64.h (AARCH64_FL_MOPS): Define.
927 (AARCH64_ISA_MOPS): Define.
928 (TARGET_MOPS): Define.
929 (MOVE_RATIO): Adjust for TARGET_MOPS.
930 * config/aarch64/aarch64.md ("unspec"): Add UNSPEC_CPYMEM.
931 (aarch64_cpymemdi): New pattern.
932 (cpymemdi): Adjust for TARGET_MOPS.
933 * config/aarch64/aarch64.opt (aarch64-mops-memcpy-size-threshol):
935 * doc/invoke.texi (AArch64 Options): Document +mops.
937 2021-12-13 Martin Liska <mliska@suse.cz>
940 * ipa-inline.c (can_inline_edge_p): Move logic checking
941 no_profile_instrument_function logic to ...
942 (can_early_inline_edge_p): ... here.
944 2021-12-13 Olivier Hainque <hainque@adacore.com>
946 * config/vxworks/_yvals.h: #include yvals.h also if
949 2021-12-13 Olivier Hainque <hainque@adacore.com>
951 * config/vxworks.h (VXWORKS_OS_CPP_BUILTINS): Define
954 2021-12-13 Olivier Hainque <hainque@adacore.com>
956 * config/t-vxworks: Clear NATIVE_SYSTEM_HEADER_DIR.
957 * config/vxworks.h (SYSROOT_HEADERS_SUFFIX_SPEC): Define, for
958 VxWorks 7 and earlier.
959 (VXWORKS_ADDITIONAL_CPP_SPEC): Simplify accordingly.
960 (STARTFILE_PREFIX_SPEC): Adjust accordingly.
961 * config/rs6000/vxworks.h (STARTFILE_PREFIX_SPEC): Adjust.
963 2021-12-13 Martin Liska <mliska@suse.cz>
965 * doc/extend.texi: Use @item for the first @itemx entry.
967 2021-12-13 Jakub Jelinek <jakub@redhat.com>
969 * machmode.h (gt_pch_nx): Use gt_pointer_operator as type of second
970 argument instead of equivalent void (*) (void *, void *, void *).
971 * poly-int.h (gt_pch_nx): Likewise.
972 * wide-int.h (gt_pch_nx): Likewise.
973 * config/aarch64/aarch64-sve-builtins.cc (gt_pch_nx): Likewise.
975 2021-12-13 Jan Hubicka <hubicka@ucw.cz>
978 * ipa-fnsummary.c (evaluate_conditions_for_known_args): Do not ICE
979 on ternary expression.
981 2021-12-13 Kewen Lin <linkw@linux.ibm.com>
984 * attribs.c (decl_attributes): Check if target options change and
985 create one node if so.
987 2021-12-12 Jonathan Wakely <jwakely@redhat.com>
989 * Makefile.in: Remove unique-ptr-tests.o.
990 * selftest-run-tests.c (selftest::run_tests): Remove
991 unique_ptr_tests_cc_tests.
992 * selftest.h (unique_ptr_tests_cc_tests): Remove.
993 * system.h: Check INCLUDE_MEMORY instead of INCLUDE_UNIQUE_PTR
994 and include <memory> instead of "unique-ptr.h".
995 * unique-ptr-tests.cc: Removed.
997 2021-12-12 Roger Sayle <roger@nextmovesoftware.com>
999 * config/nvptx/nvptx-opts.h (ptx_isa): Add PTX_ISA_SM53 ISA level
1001 * config/nvptx/nvptx.opt: Add sm_53 to -misa.
1002 * config/nvptx/nvptx-modes.def: Add support for HFmode.
1003 * config/nvptx/nvptx.h (TARGET_SM53):
1004 New helper macro to conditionalize functionality on target ISA.
1005 * config/nvptx/nvptx-c.c (nvptx_cpu_cpp_builtins): Add __PTX_SM__
1006 support for the new ISA levels.
1007 * config/nvptx/nvptx.c (nvtx_ptx_type_from_mode): Support new HFmode
1008 with the ".f16" suffix/qualifier.
1009 (nvptx_file_start): Add support for TARGET_SM53.
1010 (nvptx_omp_device_kind_arch_isa): Add support for TARGET_SM53
1011 and tweak TARGET_SM35.
1012 (nvptx_scalar_mode_supported_p): Target hook with conditional
1013 HFmode support on TARGET_SM53 and higher.
1014 (nvptx_libgcc_floating_mode_supported_p): Likewise.
1015 (TARGET_SCALAR_MODE_SUPPORTED_P): Use nvptx_scalar_mode_supported_p.
1016 (TARGET_LIBGCC_FLOATING_MODE_SUPPORTED_P): Likewise, use new hook.
1017 * config/nvptx/nvptx.md (*movhf_insn): New define_insn.
1018 (movhf): New define_expand for HFmode moves.
1019 (addhf3, subhf3, mulhf, extendhf<mode>2, trunc<mode>hf2): New
1020 instructions conditional on TARGET_SM53 (i.e. -misa=sm_53).
1022 2021-12-12 Jan Hubicka <hubicka@ucw.cz>
1025 * ipa-modref.c (modref_access_analysis::analyze): Terminate BB
1026 analysis on NULL memory access.
1027 * ipa-pure-const.c (analyze_function): Likewise.
1029 2021-12-11 Jan Hubicka <hubicka@ucw.cz>
1031 * ipa-profile.c (ipa_profile): Do not update hot bb threshold.
1033 2021-12-11 Jan Hubicka <hubicka@ucw.cz>
1035 * ipa-modref.c (get_modref_function_summary): Use ultimate_alias_target.
1036 (ignore_edge): Likewise.
1037 (compute_parm_map): Likewise.
1038 (modref_propagate_in_scc): Likewise.
1039 (modref_propagate_flags_in_scc): Likewise.
1041 2021-12-10 Jason Merrill <jason@redhat.com>
1043 * symtab.c (symtab_node::equal_address_to): Fix comment typo.
1045 2021-12-10 Doug Rupp <rupp@adacore.com>
1047 * config/vxworks.h (LINK_SPEC): Remove %(link_target).
1048 Change %{v:-v} to %{v:-V}.
1050 2021-12-10 Olivier Hainque <hainque@adacore.com>
1052 * config/t-vxworks: Remove assignment to STMP_FIXINC.
1054 2021-12-10 Martin Liska <mliska@suse.cz>
1056 * params.opt: Add missing dot.
1058 2021-12-10 Roger Sayle <roger@nextmovesoftware.com>
1061 * ipa-modref-tree.h (useful_for_kill_p): Zero width accesses aren't
1062 useful for kill tracking.
1064 2021-12-10 Andrew Stubbs <ams@codesourcery.com>
1066 * config/gcn/mkoffload.c (process_asm): Process the variable table
1067 completely differently.
1068 (process_obj): Encode the varaible data differently.
1070 2021-12-10 Joel Hutton <joel.hutton@arm.com>
1072 PR tree-optimization/103523
1073 * tree-vect-loop.c (vectorizable_induction): Check for
1074 PLUS_EXPR/MINUS_EXPR support.
1076 2021-12-10 Cui,Lili <lili.cui@intel.com>
1078 * config/i386/i386.c (ix86_vector_costs::add_stmt_cost): Remove Tremont.
1080 2021-12-09 Jan Hubicka <hubicka@ucw.cz>
1082 * doc/invoke.texi (max-inline-functions-called-once-loop-depth,
1083 max-inline-functions-called-once-insns): New parameters.
1084 * ipa-inline.c (check_callers): Handle
1085 param_inline_functions_called_once_loop_depth and
1086 param_inline_functions_called_once_insns.
1087 (edge_badness): Fix linebreaks.
1088 * params.opt (param=max-inline-functions-called-once-loop-depth,
1089 param=max-inline-functions-called-once-insn): New params.
1091 2021-12-09 Martin Sebor <msebor@redhat.com>
1093 PR tree-optimization/103215
1094 * pointer-query.cc (access_ref::merge_ref): Extend the offset and
1095 size of the merged object instead of using the larger.
1097 2021-12-09 Martin Sebor <msebor@redhat.com>
1099 PR middle-end/101751
1100 * doc/extend.texi (attribute access): Adjust.
1101 * gimple-ssa-warn-access.cc (pass_waccess::maybe_check_access_sizes):
1102 Treat access mode none on a void* argument as expecting as few as
1105 2021-12-09 Jakub Jelinek <jakub@redhat.com>
1108 * config/aarch64/aarch64-sve-builtins.cc (gt_pch_nx): Change type of
1109 second argument from function with 2 pointer arguments to function
1110 with 3 pointer arguments.
1112 2021-12-09 Olivier Hainque <hainque@adacore.com>
1114 * config/aarch64/aarch64-vxworks.h (TARGET_OS_CPP_BUILTINS):
1115 Use VX_CPU_PREFIX in CPU definitions.
1117 2021-12-09 Martin Sebor <msebor@redhat.com>
1119 * pointer-query.cc (access_ref::dump): Define new function
1120 (pointer_query::dump): Call it.
1121 * pointer-query.h (access_ref::dump): Declare new function.
1123 2021-12-09 Martin Sebor <msebor@redhat.com>
1125 * pointer-query.cc (compute_objsize_r): Add an argument.
1126 (gimple_call_return_array): Pass a new argument to compute_objsize_r.
1127 (access_ref::merge_ref): Same.
1128 (access_ref::inform_access): Add an argument and use it.
1129 (access_data::access_data): Initialize new member.
1130 (handle_min_max_size): Pass a new argument to compute_objsize_r.
1131 (handle_decl): New function.
1132 (handle_array_ref): Pass a new argument to compute_objsize_r.
1133 Avoid incrementing deref.
1134 (set_component_ref_size): New function.
1135 (handle_component_ref): New function.
1136 (handle_mem_ref): Pass a new argument to compute_objsize_r.
1137 Only increment deref after successfully computing object size.
1138 (handle_ssa_name): New function.
1139 (compute_objsize_r): Move code into helpers and call them.
1140 (compute_objsize): Pass a new argument to compute_objsize_r.
1141 * pointer-query.h (access_ref::inform_access): Add an argument.
1142 (access_data::ostype): New member.
1144 2021-12-09 Martin Sebor <msebor@redhat.com>
1146 * pointer-query.cc (access_ref::merge_ref): Define new function.
1147 (access_ref::get_ref): Move code into merge_ref and call it.
1148 * pointer-query.h (access_ref::merge_ref): Declare new function.
1150 2021-12-09 Martin Sebor <msebor@redhat.com>
1152 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Pass
1153 GIMPLE statement to compute_objsize.
1154 * pointer-query.cc (compute_objsize): Add a statement argument.
1155 * pointer-query.h (compute_objsize): Define a new overload.
1157 2021-12-09 Martin Sebor <msebor@redhat.com>
1159 * gimple-ssa-warn-access.cc (check_access): Adjust to member name
1161 (pass_waccess::check_strncmp): Same.
1162 * pointer-query.cc (access_ref::access_ref): Remove arguments.
1164 (access_data::access_data): Define new ctors.
1165 (access_data::set_bound): Define new member function.
1166 (compute_objsize_r): Remove unnecessary code.
1167 * pointer-query.h (struct access_ref): Remove ctor arguments.
1168 (struct access_data): Declare ctor overloads.
1169 (access_data::dst_bndrng): New member.
1170 (access_data::src_bndrng): New member.
1172 2021-12-09 Martin Sebor <msebor@redhat.com>
1174 PR middle-end/103143
1175 * pointer-query.cc (gimple_call_return_array): Call compute_objsize_r.
1177 2021-12-09 Olivier Hainque <hainque@adacore.com>
1178 Rasmus Villemoes <rv@rasmusvillemoes.dk>
1180 * Makefile.in (T_STDINT_GCC_H): New variable, path to
1181 stdint-gcc.h that a target configuration may override when
1182 use_gcc_stdint is "provide".
1183 (stmp-int-hdrs): Depend on it and copy that for
1184 USE_GCC_INT=provide.
1185 * config.gcc (vxworks): Revert to use_gcc_stdint=provide.
1186 * config/t-vxworks (T_STDINT_GCC_H): Define, as vxw-stdint-gcc.h.
1187 (vxw-stdint-gcc.h): New target, produced from the original
1189 (vxw-glimits.h): Use an automatic variable to designate the
1190 first and only prerequisite.
1191 * config/vxworks/stdint.h: Remove.
1193 2021-12-09 Iain Sandoe <iain@sandoe.co.uk>
1196 * config/host-darwin.c (SAFE_ALLOC_SIZE): Remove.
1197 (darwin_gt_pch_get_address): Rework for relocatable PCH.
1198 (darwin_gt_pch_use_address): Likewise.
1200 2021-12-09 Jakub Jelinek <jakub@redhat.com>
1203 * config/host-darwin.c (darwin_gt_pch_use_address): When reading
1204 manually the file into mapped area, update mapped_addr as
1205 an automatic variable rather than addr which is a reference parameter.
1206 * config/host-hpux.c (hpux_gt_pch_use_address): When reading
1207 manually the file into mapped area, update addr as
1208 an automatic variable rather than base which is a reference parameter.
1210 2021-12-09 Jakub Jelinek <jakub@redhat.com>
1213 * coretypes.h (gt_pointer_operator): Use 3 pointer arguments instead
1215 * gengtype.c (struct walk_type_data): Add in_nested_ptr argument.
1216 (walk_type): Temporarily set d->in_nested_ptr around nested_ptr
1218 (write_types_local_user_process_field): Pass a new middle pointer
1219 to gt_pointer_operator op calls, if d->in_nested_ptr pass there
1220 address of d->prev_val[2], otherwise NULL.
1221 (write_types_local_process_field): Likewise.
1222 * ggc-common.c (relocate_ptrs): Add real_ptr_p argument. If equal
1223 to ptr_p, do nothing, otherwise if NULL remember ptr_p's
1224 or if non-NULL real_ptr_p's corresponding new address in
1226 (reloc_addrs_vec): New variable.
1227 (compare_ptr, read_uleb128, write_uleb128): New functions.
1228 (gt_pch_save): When iterating over objects through relocate_ptrs,
1229 save current i into state.ptrs_i. Sort reloc_addrs_vec and emit
1230 it as uleb128 of differences between pointer addresses into the
1232 (gt_pch_restore): Allow restoring of PCH to a different address
1233 than the preferred one, in that case adjust global pointers by bias
1234 and also adjust by bias addresses read from the relocation table
1235 as uleb128 differences. Otherwise fseek over it. Perform
1236 gt_pch_restore_stringpool only after adjusting callbacks and for
1237 callback adjustments also take into account the bias.
1238 (default_gt_pch_use_address): Change type of first argument from
1240 (mmap_gt_pch_use_address): Likewise.
1241 * ggc-tests.c (gt_pch_nx): Pass NULL as new middle argument to op.
1242 * hash-map.h (hash_map::pch_nx_helper): Likewise.
1243 (gt_pch_nx): Likewise.
1244 * hash-set.h (gt_pch_nx): Likewise.
1245 * hash-table.h (gt_pch_nx): Likewise.
1246 * hash-traits.h (ggc_remove::pch_nx): Likewise.
1247 * hosthooks-def.h (default_gt_pch_use_address): Change type of first
1248 argument from void * to void *&.
1249 (mmap_gt_pch_use_address): Likewise.
1250 * hosthooks.h (struct host_hooks): Change type of first argument of
1251 gt_pch_use_address hook from void * to void *&.
1252 * machmode.h (gt_pch_nx): Expect a callback with 3 pointers instead of
1253 two in the middle argument.
1254 * poly-int.h (gt_pch_nx): Likewise.
1255 * stringpool.c (gt_pch_nx): Pass NULL as new middle argument to op.
1256 * tree-cfg.c (gt_pch_nx): Likewise, except for LOCATION_BLOCK pass
1257 the same &(block) twice.
1258 * value-range.h (gt_pch_nx): Pass NULL as new middle argument to op.
1259 * vec.h (gt_pch_nx): Likewise.
1260 * wide-int.h (gt_pch_nx): Likewise.
1261 * config/host-darwin.c (darwin_gt_pch_use_address): Change type of
1262 first argument from void * to void *&.
1263 * config/host-darwin.h (darwin_gt_pch_use_address): Likewise.
1264 * config/host-hpux.c (hpux_gt_pch_use_address): Likewise.
1265 * config/host-linux.c (linux_gt_pch_use_address): Likewise. If
1266 it couldn't succeed to mmap at the preferred location, set base
1267 to the actual one. Update addr in the manual reading loop instead of
1269 * config/host-netbsd.c (netbsd_gt_pch_use_address): Change type of
1270 first argument from void * to void *&.
1271 * config/host-openbsd.c (openbsd_gt_pch_use_address): Likewise.
1272 * config/host-solaris.c (sol_gt_pch_use_address): Likewise.
1273 * config/i386/host-mingw32.c (mingw32_gt_pch_use_address): Likewise.
1274 * config/rs6000/rs6000-gen-builtins.c (write_init_file): Pass NULL
1275 as new middle argument to op in the generated code.
1276 * doc/gty.texi: Adjust samples for the addition of middle pointer
1277 to gt_pointer_operator callback.
1279 2021-12-09 Alexandre Oliva <oliva@adacore.com>
1282 * reg-stack.c (convert_regs_1): Move any_malformed_asm
1284 (reg_to_stack): ... here.
1286 2021-12-09 Alexandre Oliva <oliva@adacore.com>
1289 * expr.c (emit_move_multi_word): Skip clobber during lra.
1291 2021-12-09 Alexandre Oliva <oliva@adacore.com>
1293 PR tree-optimization/103024
1294 PR middle-end/103530
1295 * gimple-harden-conditionals.cc (non_eh_succ_edge): New.
1296 (pass_harden_compares::execute): Accept 1-bit integral types,
1297 and cope with throwing compares.
1299 2021-12-08 Iain Sandoe <iain@sandoe.co.uk>
1301 * config/darwin.h (DARWIN_PIE_SPEC): Add -no_pie when
1302 linking mdynamic-no-pic code on macOS > 10.7.
1304 2021-12-08 Dimitar Dimitrov <dimitar@dinux.eu>
1306 * config/pru/pru.c (pru_section_type_flags): New function.
1307 (TARGET_SECTION_TYPE_FLAGS): Wire it.
1309 2021-12-08 David Faust <david.faust@oracle.com>
1311 * config/bpf/bpf.c (handle_attr_preserve): Avoid calling
1312 is_gimple_assign with a NULL pointer.
1314 2021-12-08 Martin Liska <mliska@suse.cz>
1316 * profile-count.c (profile_count::dump): Add function
1317 that can dump to a provided buffer.
1318 (profile_probability::dump): Likewise.
1319 * profile-count.h: Likewise.
1320 * tree-ssa-loop-unswitch.c (tree_unswitch_single_loop):
1321 Use dump_printf_loc infrastructure.
1322 (tree_unswitch_outer_loop): Likewise.
1323 (find_loop_guard): Likewise.
1324 (hoist_guard): Likewise.
1326 2021-12-08 Chung-Lin Tang <cltang@codesourcery.com>
1328 * gimplify.c (extract_base_bit_offset): Add 'tree *offsetp' parameter,
1329 accomodate case where 'offset' return of get_inner_reference is
1331 (is_or_contains_p): Further robustify conditions.
1332 (omp_target_reorder_clauses): In alloc/to/from sorting phase, also
1333 move following GOMP_MAP_ALWAYS_POINTER maps along. Add new sorting
1334 phase where we make sure pointers with an attach/detach map are ordered
1336 (gimplify_scan_omp_clauses): Add modifications to avoid creating
1337 GOMP_MAP_STRUCT and associated alloc map for attach/detach maps.
1339 2021-12-08 Roger Sayle <roger@nextmovesoftware.com>
1341 * config/nvptx/nvptx.md (*extend_trunc_<mode>2_qi,
1342 *extend_trunc_<mode>2_hi, *extend_trunc_di2_si): New insns.
1343 Use cvt to perform sign-extension of truncation in one step.
1345 2021-12-08 Chung-Lin Tang <cltang@codesourcery.com>
1348 * gimplify.c ("tree-hash-traits.h"): Add include.
1349 (gimplify_scan_omp_clauses): Change struct_map_to_clause to type
1350 hash_map<tree_operand, tree> *. Adjust struct map handling to handle
1351 cases of *A and A->B expressions. Under !DECL_P case of
1352 GOMP_CLAUSE_MAP handling, add STRIP_NOPS for indir_p case, add to
1353 struct_deref_set for map(*ptr_to_struct) cases. Add MEM_REF case when
1354 handling component_ref_p case. Add unshare_expr and gimplification
1355 when created GOMP_MAP_STRUCT is not a DECL. Add code to add
1356 firstprivate pointer for *pointer-to-struct case.
1357 (gimplify_adjust_omp_clauses): Move GOMP_MAP_STRUCT removal code for
1358 exit data directives code to earlier position.
1359 * omp-low.c (lower_omp_target):
1360 Handle GOMP_MAP_ATTACH_ZERO_LENGTH_ARRAY_SECTION, and
1361 GOMP_MAP_POINTER_TO_ZERO_LENGTH_ARRAY_SECTION map kinds.
1362 * tree-pretty-print.c (dump_omp_clause): Likewise.
1364 2021-12-08 Andrew Stubbs <ams@codesourcery.com>
1365 Hafiz Abid Qadeer <abidh@codesourcery.com>
1367 * dwarf2cfi.c (dw_stack_pointer_regnum): Change type to struct cfa_reg.
1368 (dw_frame_pointer_regnum): Likewise.
1369 (new_cfi_row): Use set_by_dwreg.
1370 (get_cfa_from_loc_descr): Use set_by_dwreg. Support register spans.
1371 handle DW_OP_bregx with DW_OP_breg{0-31}. Support DW_OP_lit*,
1372 DW_OP_const*, DW_OP_minus, DW_OP_shl and DW_OP_plus.
1373 (lookup_cfa_1): Use set_by_dwreg.
1374 (def_cfa_0): Update for cfa_reg and support register spans.
1375 (reg_save): Change sreg parameter to struct cfa_reg. Support register
1377 (dwf_cfa_reg): New function.
1378 (dwarf2out_flush_queued_reg_saves): Use dwf_cfa_reg instead of
1380 (dwarf2out_frame_debug_def_cfa): Likewise.
1381 (dwarf2out_frame_debug_adjust_cfa): Likewise.
1382 (dwarf2out_frame_debug_cfa_offset): Likewise. Update reg_save usage.
1383 (dwarf2out_frame_debug_cfa_register): Likewise.
1384 (dwarf2out_frame_debug_expr): Likewise.
1385 (create_pseudo_cfg): Use set_by_dwreg.
1386 (initial_return_save): Use set_by_dwreg and dwf_cfa_reg,
1387 (create_cie_data): Use dwf_cfa_reg.
1388 (execute_dwarf2_frame): Use dwf_cfa_reg.
1389 (dump_cfi_row): Use set_by_dwreg.
1390 * dwarf2out.c (build_span_loc, build_breg_loc): New function.
1391 (build_cfa_loc): Support register spans.
1392 (build_cfa_aligned_loc): Update cfa_reg usage.
1393 (convert_cfa_to_fb_loc_list): Use set_by_dwreg.
1394 * dwarf2out.h (struct cfa_reg): New type.
1395 (struct dw_cfa_location): Use struct cfa_reg.
1396 (build_span_loc): New prototype.
1398 2021-12-08 Haochen Jiang <haochen.jiang@intel.com>
1401 * config/i386/sse.md
1402 (*<sse4_1>_blendv<ssefltmodesuffix><avxsizesuffix>_not_ltint):
1403 Add new define_insn_and_split.
1405 2021-12-08 Alexandre Oliva <oliva@adacore.com>
1407 PR middle-end/103149
1408 * gimple-harden-conditionals.cc (detach_value): Use memory if
1409 general regs won't do.
1411 2021-12-07 Martin Liska <mliska@suse.cz>
1413 PR middle-end/103438
1414 * config/s390/s390.c (s390_valid_target_attribute_inner_p):
1415 Use new enum CLVC_INTEGER.
1416 * opt-functions.awk: Use new CLVC_INTEGER.
1417 * opts-common.c (set_option): Likewise.
1418 (option_enabled): Return -1,0,1 for CLVC_INTEGER.
1419 (get_option_state): Use new CLVC_INTEGER.
1420 (control_warning_option): Likewise.
1421 * opts.h (enum cl_var_type): Likewise.
1423 2021-12-07 H.J. Lu <hjl.tools@gmail.com>
1426 * config/i386/i386.c (ix86_call_use_plt_p): Check FUNCTION_DECL
1427 before calling cgraph_node::get.
1429 2021-12-07 Richard Biener <rguenther@suse.de>
1431 PR tree-optimization/103596
1432 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_stmt):
1433 Note we are not propagating into a PHI argument to may_propagate_copy.
1434 * tree-ssa-propagate.h (may_propagate_copy): Add
1435 argument specifying whether we propagate into a PHI arg.
1436 * tree-ssa-propagate.c (may_propagate_copy): Likewise.
1437 When not doing so we can replace an abnormal with
1439 (may_propagate_into_stmt): Update may_propagate_copy calls.
1440 (replace_exp_1): Move propagation checking code to
1441 propagate_value and rename to ...
1442 (replace_exp): ... this and elide previous wrapper.
1443 (propagate_value): Perform checking with adjusted
1444 may_propagate_copy call and dispatch to replace_exp.
1446 2021-12-07 Matthias Kretz <m.kretz@gsi.de>
1448 * hash-map.h (hash_map::traverse): Let both overloads behave the
1450 * predict.c (assert_is_empty): Return true, thus not changing
1453 2021-12-07 YunQiang Su <yunqiang.su@cipunited.com>
1455 * config/mips/mips.h (ISA_HAS_UNALIGNED_ACCESS, STRICT_ALIGNMENT):
1456 R6 can unaligned access.
1457 * config/mips/mips.md (movmisalign<mode>): Likewise.
1458 * config/mips/mips.opt: add -m(no-)unaligned-access
1459 * doc/invoke.texi: Likewise.
1461 2021-12-07 Eugene Rozenfeld <erozen@microsoft.com>
1463 * auto-profile.c (afdo_propagate_edge): Improve count propagation algorithm.
1465 2021-12-06 Paul A. Clarke <pc@us.ibm.com>
1468 * config/rs6000/xmmintrin.h (_mm_movemask_ps): Replace "vector" with
1471 2021-12-06 Jose E. Marchesi <jose.marchesi@oracle.com>
1473 * config/bpf/bpf.c (bpf_handle_preserve_access_index_attribute):
1474 Mark arguments `args' and flags' as unused.
1475 (bpf_core_newdecl): Remove unused local `newdecl'.
1476 (bpf_core_newdecl): Remove unused argument `loc'.
1477 (ctfc_debuginfo_early_finish_p): Remove unused function.
1478 (TARGET_CTFC_DEBUGINFO_EARLY_FINISH_P): Remove definition.
1479 (bpf_core_walk): Do not pass a location to bpf_core_newdecl.
1481 2021-12-06 Richard Sandiford <richard.sandiford@arm.com>
1483 * gimple-range-edge.cc (gimple_outgoing_range::edge_range_p): Add
1484 a shortcut for blocks with single successors.
1485 * gimple-range-gori.cc (gori_map::calculate_gori): Likewise.
1487 2021-12-06 Richard Sandiford <richard.sandiford@arm.com>
1489 * value-range.cc (irange::irange_union): Use quick_push rather
1490 than safe_push. Use widest_int rather than wide_int. Avoid
1491 assigning wi::to_* results to wide*_int temporaries.
1493 2021-12-06 Andrew MacLeod <amacleod@redhat.com>
1495 * gimple-range-cache.cc (ranger_cache::fill_block_cache): Check for
1496 a range from dominators before filling the cache.
1497 (ranger_cache::range_from_dom): New.
1498 * gimple-range-cache.h (ranger_cache::range_from_dom): Add prototype.
1500 2021-12-06 Andrew MacLeod <amacleod@redhat.com>
1502 * gimple-range-gori.h (class gori_compute):: Add prototypes.
1503 * gimple-range-gori.cc (gori_compute::has_edge_range_p): Add alternate
1504 API for basic block. Call for edge alterantive.
1505 (gori_compute::may_recompute_p): Ditto.
1507 2021-12-06 Richard Biener <rguenther@suse.de>
1509 PR tree-optimization/103581
1510 * tree-vect-stmts.c (vect_build_gather_load_calls): Properly
1511 guard all the AVX512 mask cases.
1513 2021-12-06 Richard Biener <rguenther@suse.de>
1515 PR tree-optimization/103544
1516 * tree-vect-slp.c (vect_analyze_slp): Only add a SLP reduction
1517 opportunity if the stmt in question is the reduction root.
1518 (dot_slp_tree): Add missing check for NULL child.
1520 2021-12-06 Jakub Jelinek <jakub@redhat.com>
1523 * config/avr/avr.c (avr_output_data_section_asm_op,
1524 avr_output_bss_section_asm_op): Change argument type from const void *
1527 2021-12-06 Tamar Christina <tamar.christina@arm.com>
1529 PR rtl-optimization/103404
1530 * cse.c (find_sets_in_insn): Don't select elements out of a V1 mode
1533 2021-12-06 Hongtao Liu <Hongtao.liu@intel.com>
1534 Uroš Bizjak <ubizjak@gmail.com>
1537 * config/i386/i386.c (ix86_preferred_reload_class): Allow
1538 integer regs when moves between register units are cheap.
1539 * config/i386/i386.h (INT_SSE_CLASS_P): New.
1541 2021-12-05 David Edelsohn <dje.gcc@gmail.com>
1543 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Move
1544 AIX math builtin initialization before new_builtins_are_live.
1546 2021-12-04 Jakub Jelinek <jakub@redhat.com>
1548 * config/i386/x86-tune.def (X86_TUNE_PARTIAL_REG_DEPENDENCY): Fix
1549 comment typo, Preffer -> prefer.
1550 * ipa-modref-tree.c (modref_access_node::closer_pair_p): Likewise.
1552 2021-12-04 Alexandre Oliva <oliva@adacore.com>
1554 PR rtl-optimization/103028
1555 * ifcvt.c (find_cond_trap): Validate new insns more strictly
1558 2021-12-03 Martin Liska <mliska@suse.cz>
1559 Peter Bergner <bergner@linux.ibm.com>
1562 * config/rs6000/rs6000.c (rs6000_option_override_internal): Move the
1563 disabling of shrink-wrapping when using -mrop-protect from here...
1564 (rs6000_override_options_after_change): ...to here.
1566 2021-12-03 H.J. Lu <hjl.tools@gmail.com>
1569 * config/i386/i386-expand.c (ix86_expand_builtin): Pass PVW_NONE
1570 and PVW_NONE to ix86_target_string.
1571 * config/i386/i386-options.c (ix86_target_string): Add arguments
1572 for move_max and store_max.
1573 (ix86_target_string::add_vector_width): New lambda.
1574 (ix86_debug_options): Pass ix86_move_max and ix86_store_max to
1576 (ix86_function_specific_print): Pass ptr->x_ix86_move_max and
1577 ptr->x_ix86_store_max to ix86_target_string.
1578 (ix86_valid_target_attribute_tree): Handle x_ix86_move_max and
1580 (ix86_option_override_internal): Set the default x_ix86_move_max
1581 and x_ix86_store_max.
1582 * config/i386/i386-options.h (ix86_target_string): Add
1583 prefer_vector_width and prefer_vector_width.
1584 * config/i386/i386.h (TARGET_AVX256_MOVE_BY_PIECES): Removed.
1585 (TARGET_AVX256_STORE_BY_PIECES): Likewise.
1586 (MOVE_MAX): Use 64 if ix86_move_max or ix86_store_max ==
1587 PVW_AVX512. Use 32 if ix86_move_max or ix86_store_max >=
1589 (STORE_MAX_PIECES): Use 64 if ix86_store_max == PVW_AVX512.
1590 Use 32 if ix86_store_max >= PVW_AVX256.
1591 * config/i386/i386.opt: Add -mmove-max=bits and -mstore-max=bits.
1592 * config/i386/x86-tune.def (X86_TUNE_AVX512_MOVE_BY_PIECES): New.
1593 (X86_TUNE_AVX512_STORE_BY_PIECES): Likewise.
1594 * doc/invoke.texi: Document -mmove-max=bits and -mstore-max=bits.
1596 2021-12-03 Bill Schmidt <wschmidt@linux.ibm.com>
1598 * config/rs6000/rs6000.c (rs6000_builtin_reciprocal): Fix builtin
1601 2021-12-03 SiYu Wu <siyu@isrc.iscas.ac.cn>
1603 * common/config/riscv/riscv-common.c (riscv_implied_info):
1604 Add K-ext related entry.
1605 (riscv_supported_std_ext): Add 'k'.
1606 * config/riscv/arch-canonicalize (CANONICAL_ORDER): Add 'k'.
1607 (IMPLIED_EXT): Add K-ext related entry.
1609 2021-12-03 SiYu Wu <siyu@isrc.iscas.ac.cn>
1611 * common/config/riscv/riscv-common.c (riscv_ext_version_table):
1613 * config/riscv/riscv-opts.h (MASK_ZBKB): New.
1620 (MASK_ZKSED): Ditto.
1623 (TARGET_ZBKB): Ditto.
1624 (TARGET_ZBKC): Ditto.
1625 (TARGET_ZBKX): Ditto.
1626 (TARGET_ZKNE): Ditto.
1627 (TARGET_ZKND): Ditto.
1628 (TARGET_ZKNH): Ditto.
1629 (TARGET_ZKR): Ditto.
1630 (TARGET_ZKSED): Ditto.
1631 (TARGET_ZKSH): Ditto.
1632 (TARGET_ZKT): Ditto.
1633 * config/riscv/riscv.opt (riscv_zk_subext): New.
1635 2021-12-03 Tamar Christina <tamar.christina@arm.com>
1637 * tree-vect-stmts.c (prepare_load_store_mask): Rename to...
1638 (prepare_vec_mask): ...This and record operations that have already been
1640 (vectorizable_call): Use it.
1641 (vectorizable_operation): Likewise.
1642 (vectorizable_store): Likewise.
1643 (vectorizable_load): Likewise.
1644 * tree-vectorizer.h (class _loop_vec_info): Add vec_cond_masked_set.
1645 (vec_cond_masked_set_type, tree_cond_mask_hash): New.
1647 2021-12-03 H.J. Lu <hjl.tools@gmail.com>
1651 * target.def (ifunc_ref_local_ok): Add a target hook.
1652 * varasm.c (default_binds_local_p_3): Force indirect function
1653 resolver non-local only if targetm.ifunc_ref_local_ok returns
1655 * config/i386/i386-expand.c (ix86_expand_call): Call
1656 ix86_call_use_plt_p to check if PLT should be used.
1657 * config/i386/i386-protos.h (ix86_call_use_plt_p): New.
1658 * config/i386/i386.c (output_pic_addr_const): Call
1659 ix86_call_use_plt_p to check if "@PLT" is needed.
1660 (ix86_call_use_plt_p): New.
1661 (TARGET_IFUNC_REF_LOCAL_OK): New.
1662 * doc/tm.texi.in: Add TARGET_IFUNC_REF_LOCAL_OK.
1663 * doc/tm.texi: Regenerated.
1665 2021-12-03 Jakub Jelinek <jakub@redhat.com>
1667 * attribs.h (simple_cst_list_equal): Declare.
1668 * attribs.c (simple_cst_list_equal): No longer static.
1669 * config/i386/i386-options.c (target_attribute_cache): New variable.
1670 (ix86_valid_target_attribute_p): Cache DECL_FUNCTION_SPECIFIC_TARGET
1671 and DECL_FUNCTION_SPECIFIC_OPTIMIZATION based on args.
1673 2021-12-03 Jakub Jelinek <jakub@redhat.com>
1676 * ggc.h (gt_pch_note_callback): Declare.
1677 * gengtype.h (enum typekind): Add TYPE_CALLBACK.
1678 (callback_type): Declare.
1679 * gengtype.c (dbgprint_count_type_at): Handle TYPE_CALLBACK.
1680 (callback_type): New variable.
1681 (process_gc_options): Add CALLBACK argument, handle callback
1683 (set_gc_used_type): Adjust process_gc_options caller, if callback,
1684 set type to &callback_type.
1685 (output_mangled_typename): Handle TYPE_CALLBACK.
1686 (walk_type): Likewise. Handle callback option.
1687 (write_types_process_field): Handle TYPE_CALLBACK.
1688 (write_types_local_user_process_field): Likewise.
1689 (write_types_local_process_field): Likewise.
1690 (write_root): Likewise.
1691 (dump_typekind): Likewise.
1692 (dump_type): Likewise.
1693 * gengtype-state.c (type_lineloc): Handle TYPE_CALLBACK.
1694 (state_writer::write_state_callback_type): New method.
1695 (state_writer::write_state_type): Handle TYPE_CALLBACK.
1696 (read_state_callback_type): New function.
1697 (read_state_type): Handle TYPE_CALLBACK.
1698 * ggc-common.c (callback_vec): New variable.
1699 (gt_pch_note_callback): New function.
1700 (gt_pch_save): Stream out gt_pch_save function address and relocation
1702 (gt_pch_restore): Stream in saved gt_pch_save function address and
1703 relocation table and apply relocations if needed.
1704 * doc/gty.texi (callback): Document new GTY option.
1705 * varasm.c (get_unnamed_section): Change callback argument's type and
1706 last argument's type from const void * to const char *.
1707 (output_section_asm_op): Change argument's type from const void *
1708 to const char *, remove unnecessary cast.
1709 * tree-core.h (struct tree_translation_unit_decl): Drop GTY((skip))
1710 from language member.
1711 * output.h (unnamed_section_callback): Change argument type from
1712 const void * to const char *.
1713 (struct unnamed_section): Use GTY((callback)) instead of GTY((skip))
1714 for callback member. Change data member type from const void *
1716 (struct noswitch_section): Use GTY((callback)) instead of GTY((skip))
1717 for callback member.
1718 (get_unnamed_section): Change callback argument's type and
1719 last argument's type from const void * to const char *.
1720 (output_section_asm_op): Change argument's type from const void *
1722 * config/avr/avr.c (avr_output_progmem_section_asm_op): Likewise.
1723 Remove unneeded cast.
1724 * config/darwin.c (output_objc_section_asm_op): Change argument's type
1725 from const void * to const char *.
1726 * config/pa/pa.c (som_output_text_section_asm_op): Likewise.
1727 (som_output_comdat_data_section_asm_op): Likewise.
1728 * config/rs6000/rs6000.c (rs6000_elf_output_toc_section_asm_op):
1730 (rs6000_xcoff_output_readonly_section_asm_op): Likewise. Instead
1731 of dereferencing directive hardcode variable names and decide based on
1732 whether directive is NULL or not.
1733 (rs6000_xcoff_output_readwrite_section_asm_op): Change argument's type
1734 from const void * to const char *.
1735 (rs6000_xcoff_output_tls_section_asm_op): Likewise. Instead
1736 of dereferencing directive hardcode variable names and decide based on
1737 whether directive is NULL or not.
1738 (rs6000_xcoff_output_toc_section_asm_op): Change argument's type
1739 from const void * to const char *.
1740 (rs6000_xcoff_asm_init_sections): Adjust get_unnamed_section callers.
1742 2021-12-03 Richard Purdie <richard.purdie@linuxfoundation.org>
1744 * Makefile.in: Fix "argument list too long" from install-plugins.
1746 2021-12-03 Matt Jacobson <mhjacobson@me.com>
1748 * config.gcc: For the AVR target, populate TM_MULTILIB_CONFIG.
1749 * config/avr/genmultilib.awk: Add ability to filter generated multilib
1751 * config/avr/t-avr: Pass TM_MULTILIB_CONFIG to genmultilib.awk.
1752 * configure.ac: Update help string for --with-multilib-list.
1753 * configure: Regenerate.
1755 2021-12-02 Ilya Lipnitskiy <ilya.lipnitskiy@gmail.com>
1757 * configure: Regenerate.
1758 * configure.ac: Define TARGET_LIBC_GNUSTACK on musl.
1760 2021-12-02 Iain Sandoe <iain@sandoe.co.uk>
1763 * config/host-darwin.c (TRY_EMPTY_VM_SPACE,
1764 SAFE_ALLOC_SIZE): New.
1765 (darwin_gt_pch_get_address): Rewrite to use nominated
1766 memory segments rather than part of the compiler __DATA
1768 (darwin_gt_pch_use_address): Likewise.
1770 2021-12-02 Gerald Pfeifer <gerald@pfeifer.com>
1772 * doc/install.texi (*-*-freebsd*): Remove references to
1773 FreeBSD 1 and FreeBSD 2.
1775 2021-12-02 Vladimir N. Makarov <vmakarov@redhat.com>
1777 PR rtl-optimization/103437
1778 * ira-color.c (setup_allocno_priorities): Switch off backup code
1779 for overflow if compiler has __builtin_smul_overflow. Use <
1780 for comparison with -INT_MAX.
1782 2021-12-02 Bill Schmidt <wschmidt@linux.ibm.com>
1784 * config/rs6000/altivec.h: Delete a number of #defines that are now
1785 superfluous. Alphabetize. Include rs6000-vecdefines.h. Include some
1787 * config/rs6000/rs6000-builtin-new.def (CMPB): Flag as no32bit.
1788 (BPERMD): Flag as 32bit (needing special handling for 32-bit).
1789 (UNPACK_TD): Return unsigned long long instead of unsigned long.
1790 (GET_TEXASR): Return unsigned long instead of unsigned long long.
1791 (GET_TEXASRU): Likewise.
1792 (GET_TFHAR): Likewise.
1793 (GET_TFIAR): Likewise.
1794 (SET_TEXASR): Pass unsigned long instead of unsigned long long.
1795 (SET_TEXASRU): Likewise.
1796 (SET_TFHAR): Likewise.
1797 (SET_TFIAR): Likewise.
1798 (TABORTDC): Likewise.
1799 (TABORTDCI): Likewise.
1800 * config/rs6000/rs6000-call.c (rs6000_expand_new_builtin): Fix error
1801 handling for no32bit. Add 32bit handling for RS6000_BIF_BPERMD.
1802 * config/rs6000/rs6000-gen-builtins.c (write_init_file): Initialize
1803 new_builtins_are_live to 1.
1805 2021-12-02 Vladimir N. Makarov <vmakarov@redhat.com>
1807 PR rtl-optimization/103437
1808 * ira-color.c (setup_allocno_priorities): Use long long
1809 multiplication as backup for overflow processing.
1811 2021-12-02 Tamar Christina <tamar.christina@arm.com>
1813 * config/aarch64/aarch64-protos.h (aarch64_gen_shareable_zero): New.
1814 * config/aarch64/aarch64-simd.md (aarch64_rshrn<mode>,
1815 aarch64_rshrn2<mode>): Generate rounding half-ing add when appropriate.
1816 * config/aarch64/aarch64.c (aarch64_gen_shareable_zero): New.
1818 2021-12-02 Vladimir N. Makarov <vmakarov@redhat.com>
1820 PR rtl-optimization/103437
1821 * ira-color.c (setup_allocno_priorities): Process multiplication
1824 2021-12-02 Martin Liska <mliska@suse.cz>
1826 PR gcov-profile/96092
1827 * common.opt: New option.
1828 * coverage.c (coverage_begin_function): Emit filename with
1829 remap_profile_filename.
1830 * doc/invoke.texi: Document the new option.
1831 * file-prefix-map.c (add_profile_prefix_map): New.
1832 (remap_profile_filename): Likewise.
1833 * file-prefix-map.h (add_profile_prefix_map): Likewise.
1834 (remap_profile_filename): Likewise.
1835 * lto-opts.c (lto_write_options): Handle
1836 OPT_fprofile_prefix_map_.
1837 * opts-global.c (handle_common_deferred_options): Likewise.
1838 * opts.c (common_handle_option): Likewise.
1839 (gen_command_line_string): Likewise.
1840 * profile.c (output_location): Emit filename with
1841 remap_profile_filename.
1843 2021-12-02 Richard Biener <rguenther@suse.de>
1845 PR middle-end/103271
1846 * internal-fn.c (expand_DEFERRED_INIT): When the base
1847 of the LHS is a decl with matching constant size use
1848 that as the initialization target instead of an
1849 eventual VLA typed one.
1851 2021-12-02 Richard Biener <rguenther@suse.de>
1853 PR tree-optimization/103527
1854 * tree-vect-stmts.c (vect_build_gather_load_calls): Always
1855 use a truth type for building the vector mask.
1857 2021-12-02 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
1859 * config/vxworks.h (LTO_PLUGIN_SPEC): Adapt to corrected
1860 spelling of -linker-output-auto-nolto-rel.
1862 2021-12-01 Uroš Bizjak <ubizjak@gmail.com>
1865 * config/i386/sse.md (VI2F): Remove mode iterator.
1866 (VI2F_256_512): New mode iterator.
1867 (vec_set<V8_128:mode>_0): New insn pattern.
1868 (vec_set<VI2F_256_512:mode>_0>): Rename from vec_set<VI2F:mode>mode.
1869 Use VI2F_256_512 mode iterator instead of VI2F.
1870 (*axv512fp16_movsh): Remove.
1871 (<sse2p4_1>_pinsr<ssemodesuffix>): Add (x,x,x) AVX2 alternative.
1872 Do not disable V8HF mode insn on AVX2 targets.
1873 (pinsrw -> pbroadcast + pblendw peephole2): New peephole.
1874 (pinsrw -> pbroadcast + pblendw splitter): New post-reload splitter.
1875 * config/i386/i386.md (extendhfsf): Call gen_vec_setv8hf_0.
1876 * config/i386/i386-expand.c (ix86_expand_vector_set)
1877 <case E_V8HFmode>: Use vec_merge path for TARGET_AVX2.
1879 2021-12-01 Roger Sayle <roger@nextmovesoftware.com>
1880 Richard Biener <rguenther@suse.de>
1882 * tree-ssa-loop-niter.c (number_of_iterations_until_wrap):
1883 Check if simplify_using_initial_conditions allows us to
1884 simplify the expression for may_be_zero.
1885 * match.pd (X != C ? -X : -C -> -X): New transform.
1886 (X != C ? ~X : ~C -> ~X): Likewise.
1887 ((X+1) > Y ? -X : 1 -> X >= Y ? -X : 1): Likewise.
1889 2021-12-01 Peter Bergner <bergner@linux.ibm.com>
1891 PR middle-end/103127
1892 * gimplify.c (is_var_need_auto_init): Handle opaque types.
1894 2021-12-01 Iain Buclaw <ibuclaw@gdcproject.org>
1896 * common.opt (static-libphobos): Add option.
1897 * config/darwin.h (LINK_SPEC): Substitute -lgphobos with libgphobos.a
1898 when linking statically.
1899 * gcc.c (driver_handle_option): Set -static-libphobos as always valid.
1901 2021-12-01 Jason Merrill <jason@redhat.com>
1904 * fold-const.c (maybe_nonzero_address): Use get_create or get
1905 depending on folding_initializer.
1906 (fold_binary_initializer_loc): New.
1907 * fold-const.h (fold_binary_initializer_loc): Declare.
1908 * varasm.c (mark_weak): Don't use the decl location.
1909 * doc/invoke.texi: Document -fconstexpr-fp-except.
1911 2021-12-01 Bill Schmidt <wschmidt@linux.ibm.com>
1914 * config/rs6000/rs6000-c.c (rs6000-builtins.h): Stop including.
1915 (rs6000-internal.h): Include.
1916 (altivec_resolve_new_overloaded_builtin): Move call to
1917 rs6000_invalid_new_builtin here from rs6000_new_builtin_decl.
1918 * config/rs6000/rs6000-call.c (rs6000-builtins.h): Stop including.
1919 (rs6000_invalid_new_builtin): Remove static qualifier.
1920 (rs6000_new_builtin_decl): Remove test for supported builtin.
1921 * config/rs6000/rs6000-internal.h (rs6000-builtins.h): Include.
1922 (rs6000_invalid_new_builtin): Declare.
1923 * config/rs6000/rs6000.c (rs6000-builtins.h): Don't include.
1925 2021-12-01 Aldy Hernandez <aldyh@redhat.com>
1927 PR tree-optimization/103409
1928 * gimple-range-path.cc (path_range_query::compute_ranges_in_phis):
1929 Do all the work with just one ssa_global_cache.
1930 * gimple-range-path.h: Remove m_tmp_phi_cache.
1932 2021-12-01 Richard Sandiford <richard.sandiford@arm.com>
1934 PR tree-optimization/103517
1935 * tree-vect-slp.c (vect_build_slp_tree_1): When allowing two
1936 different component references, check the codes of both them,
1937 rather than just the first.
1939 2021-12-01 Andrew MacLeod <amacleod@redhat.com>
1941 PR tree-optimization/103464
1942 * gimple-range.cc (gimple_ranger::prefill_name): Process phis also.
1943 (gimple_ranger::prefill_stmt_dependencies): Ditto.
1945 2021-12-01 Iain Buclaw <ibuclaw@gdcproject.org>
1947 * doc/install.texi (Prerequisites): Add note that D front end now
1948 requires GDC installed in order to bootstrap.
1949 (Building): Add D compiler section, referencing prerequisites.
1951 2021-12-01 Jakub Jelinek <jakub@redhat.com>
1953 * config/i386/x86-tune.def (X86_TUNE_SCHEDULE,
1954 X86_TUNE_SSE_PARTIAL_REG_FP_CONVERTS_DEPENDENCY,
1955 X86_TUNE_MISALIGNED_MOVE_STRING_PRO_EPILOGUES): Formatting fixes.
1956 (X86_TUNE_USE_GATHER): Put m_GENERIC last for consistency.
1958 2021-12-01 Jakub Jelinek <jakub@redhat.com>
1960 PR rtl-optimization/102356
1961 * rtl.h (simplify_context): Add assoc_count member and
1962 max_assoc_count static member.
1963 * simplify-rtx.c (simplify_associative_operation): Don't reassociate
1964 more than max_assoc_count times within one outermost simplify_* call.
1965 * dwarf2out.c (mem_loc_descriptor): Optimize binary operation
1966 with both operands the same using DW_OP_dup.
1968 2021-12-01 Tamar Christina <tamar.christina@arm.com>
1970 * match.pd: Move below pattern that rewrites to EQ, NE.
1971 * tree.c (bitmask_inv_cst_vector_p): Correct do .. while indentation.
1973 2021-12-01 Siddhesh Poyarekar <siddhesh@gotplt.org>
1974 Martin Liška <mliska@suse.cz>
1976 PR tree-optimization/103456
1977 * tree-object-size.c (merge_object_sizes): Update osi->changed
1978 only if object_sizes_set succeeded.
1980 2021-11-30 liuhongt <hongtao.liu@intel.com>
1984 * config/i386/i386.md (*x86_64_shld_1): Set_attr
1985 length_immediate to 1.
1986 (*x86_shld_1): Ditto.
1987 (*x86_64_shrd_1): Ditto.
1988 (*x86_shrd_1): Ditto.
1990 2021-11-30 Bill Schmidt <wschmidt@linux.ibm.com>
1992 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
1994 (altivec_resolve_new_overloaded_builtin): Likewise.
1996 2021-11-30 Martin Jambor <mjambor@suse.cz>
1999 * ipa-sra.c (scan_function): Also check ECF_LOOPING_CONST_OR_PURE flag.
2001 2021-11-30 Richard Sandiford <richard.sandiford@arm.com>
2003 PR tree-optimization/103494
2004 * tree-vect-stmts.c (vect_get_gather_scatter_ops): Remove ncopies
2005 argument and calculate ncopies from gs_info->offset_vectype
2007 (vectorizable_store, vectorizable_load): Update accordingly.
2009 2021-11-30 Iain Buclaw <ibuclaw@gdcproject.org>
2011 * Makefile.in (GDC): New variable.
2012 (GDCFLAGS): New variable.
2013 * configure: Regenerate.
2014 * configure.ac: Add call to ACX_PROG_GDC. Substitute GDCFLAGS.
2016 2021-11-30 Martin Liska <mliska@suse.cz>
2017 Martin Jambor <mjambor@suse.cz>
2020 * ipa-param-manipulation.c
2021 (ipa_param_body_adjustments::prepare_debug_expressions): Be
2022 careful about hash_map reallocating itself. Simpify a return
2023 which always returns true.
2025 2021-11-30 Andrew MacLeod <amacleod@redhat.com>
2027 PR tree-optimization/103440
2028 * gimple-range-fold.cc (fold_using_range::range_of_phi): Continue
2029 normal param processing for equiv params.
2031 2021-11-30 Richard Biener <rguenther@suse.de>
2033 * config/i386/i386.c (ix86_shift_rotate_cost): Remove
2035 * tree-chrec.c (evolution_function_is_invariant_rec_p):
2037 * tree-if-conv.c (if_convertible_stmt_p): Likewise.
2038 * tree-ssa-pre.c (fully_constant_expression): Likewise.
2039 * tree-vrp.c (operand_less_p): Likewise.
2040 * reload.c (reg_overlap_mentioned_for_reload_p): Remove
2041 unreachable gcc_unreachable ().
2042 * sel-sched-ir.h (bb_next_bb): Likewise.
2043 * varasm.c (compare_constant): Likewise.
2045 2021-11-30 Richard Biener <rguenther@suse.de>
2047 PR tree-optimization/103489
2048 * tree-vect-loop.c (vectorizable_phi): Verify argument
2049 vector type compatibility to mitigate bool pattern recog
2052 2021-11-30 Thomas Schwinge <thomas@codesourcery.com>
2054 * omp-offload.c (oacc_loop_auto_partitions): Remove erroneous
2055 "Orphan reductions cannot have gang partitioning" handling.
2057 2021-11-30 Thomas Schwinge <thomas@codesourcery.com>
2059 * omp-offload.c (oacc_loop_process): Implement "gang reduction on
2060 an orphan loop" checking.
2062 2021-11-30 Cesar Philippidis <cesar@codesourcery.com>
2063 Thomas Schwinge <thomas@codesourcery.com>
2065 * omp-general.h (enum oacc_loop_flags): Add OLF_REDUCTION enum.
2066 * omp-low.c (lower_oacc_head_mark): Use it to mark OpenACC
2068 * omp-offload.c (oacc_loop_auto_partitions): Don't assign gang
2069 level parallelism to orphan reductions.
2071 2021-11-30 Aldy Hernandez <aldyh@redhat.com>
2073 PR tree-optimization/103451
2074 * range-op.cc (operator_div::wi_fold): Remove
2075 can_throw_non_call_exceptions special case.
2076 * tree-ssa-sink.c (sink_code_in_bb): Same.
2078 2021-11-30 Richard Sandiford <richard.sandiford@arm.com>
2080 * tree-vect-slp.c (arg1_arg4_map): New variable.
2081 (vect_get_operand_map): Handle IFN_MASK_GATHER_LOAD.
2082 (vect_build_slp_tree_1): Likewise.
2083 (vect_build_slp_tree_2): Likewise.
2084 * tree-vect-stmts.c (vectorizable_load): Expect the mask to be
2085 the last SLP child node rather than the first.
2087 2021-11-30 Richard Sandiford <richard.sandiford@arm.com>
2089 * tree-if-conv.c: Include tree-eh.h.
2090 (predicate_statements): Remove pe argument. Don't hoist
2092 (combine_blocks): Remove pe argument.
2093 (ifcvt_available_on_edge_p, ifcvt_can_hoist): New functions.
2094 (ifcvt_hoist_invariants): Likewise.
2095 (tree_if_conversion): Update call to combine_blocks. Call
2096 ifcvt_hoist_invariants after VN.
2098 2021-11-30 Richard Sandiford <richard.sandiford@arm.com>
2100 PR tree-optimization/102467
2101 * doc/sourcebuild.texi (vect_gather_load_ifn): Document.
2102 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
2103 Commonize safelen handling. Punt for anything involving
2104 gathers and scatters unless safelen says otherwise.
2105 * tree-vect-slp.c (arg1_map): New variable.
2106 (vect_get_operand_map): Handle IFN_GATHER_LOAD.
2107 (vect_build_slp_tree_1): Likewise.
2108 (vect_build_slp_tree_2): Likewise.
2109 (compatible_calls_p): If vect_get_operand_map returns nonnull,
2110 check that any skipped arguments are equal.
2111 (vect_slp_analyze_node_operations_1): Tighten reduction check.
2112 * tree-vect-stmts.c (check_load_store_for_partial_vectors): Take
2113 an ncopies argument.
2114 (vect_get_gather_scatter_ops): Take slp_node and ncopies arguments.
2116 (vectorizable_store, vectorizable_load): Adjust accordingly.
2118 2021-11-30 Richard Sandiford <richard.sandiford@arm.com>
2120 * gimple.h (gimple_num_args, gimple_arg, gimple_arg_ptr): New
2122 * tree-vect-slp.c (cond_expr_maps, arg2_map): New variables.
2123 (vect_get_operand_map): New function.
2124 (vect_get_and_check_slp_defs): Fix outdated comment.
2125 Use vect_get_operand_map and new gimple argument accessors.
2126 (vect_build_slp_tree_2): Likewise.
2128 2021-11-30 Richard Sandiford <richard.sandiford@arm.com>
2130 * tree-vect-slp.c (vect_build_slp_tree_1): Use code_helper
2131 to record the operations performed by statements, only using
2132 CALL_EXPR for things that don't map to built-in or internal
2133 functions. For shifts, require all shift amounts to be equal
2134 if optab_vector is not supported but optab_scalar is.
2136 2021-11-30 Richard Sandiford <richard.sandiford@arm.com>
2138 * tree-vect-data-refs.c (vect_check_gather_scatter): Continue
2139 processing conversions if the current offset is a pointer.
2141 2021-11-30 Richard Sandiford <richard.sandiford@arm.com>
2143 * tree-vectorizer.h (vect_is_reduction): Use STMT_VINFO_REDUC_IDX.
2145 2021-11-30 Richard Sandiford <richard.sandiford@arm.com>
2147 * genopinit.c (main): Turn supports_vec_gather_load and
2148 supports_vec_scatter_store into signed char arrays and remove
2149 supports_vec_gather_load_cached and supports_vec_scatter_store_cached.
2150 * optabs-query.c (supports_vec_convert_optab_p): Add a mode parameter.
2151 If the mode is not VOIDmode, test only for that mode.
2152 (supports_vec_gather_load_p): Likewise.
2153 (supports_vec_scatter_store_p): Likewise.
2154 * optabs-query.h (supports_vec_gather_load_p): Likewise.
2155 (supports_vec_scatter_store_p): Likewise.
2156 * tree-vect-data-refs.c (vect_check_gather_scatter): Pass the
2157 vector mode to supports_vec_gather_load_p and
2158 supports_vec_scatter_store_p.
2160 2021-11-30 Richard Sandiford <richard.sandiford@arm.com>
2162 * internal-fn.c (commutative_binary_fn_p): Handle IFN_ADD_OVERFLOW
2163 and IFN_MUL_OVERFLOW.
2165 2021-11-30 Richard Sandiford <richard.sandiford@arm.com>
2167 * internal-fn.c (commutative_binary_fn_p): Handle IFN_UBSAN_CHECK_ADD
2168 and IFN_UBSAN_CHECK_MUL.
2170 2021-11-30 Richard Sandiford <richard.sandiford@arm.com>
2172 * internal-fn.c (commutative_binary_fn_p): Handle IFN_COMPLEX_MUL.
2174 2021-11-30 Richard Sandiford <richard.sandiford@arm.com>
2176 * gimple-fold.c: Include internal-fn.h.
2177 (fold_stmt_1): If a function maps to an internal one, use
2178 first_commutative_argument to canonicalize the order of
2179 commutative arguments.
2180 * gimple-match-head.c (gimple_resimplify2, gimple_resimplify3)
2181 (gimple_resimplify4, gimple_resimplify5): Extend commutativity
2182 checks to functions.
2184 2021-11-30 Richard Sandiford <richard.sandiford@arm.com>
2186 * doc/md.texi (reduc_fmin_scal_@var{m}): Document.
2187 (reduc_fmax_scal_@var{m}): Likewise.
2188 * optabs.def (reduc_fmax_scal_optab): New optab.
2189 (reduc_fmin_scal_optab): Likewise
2190 * internal-fn.def (REDUC_FMAX, REDUC_FMIN): New functions.
2191 * tree-vect-loop.c (reduction_fn_for_scalar_code): Handle
2192 CASE_CFN_FMAX and CASE_CFN_FMIN.
2193 (neutral_op_for_reduction): Likewise.
2194 (needs_fold_left_reduction_p): Likewise.
2195 * config/aarch64/iterators.md (FMAXMINV): New iterator.
2196 (fmaxmin): Handle UNSPEC_FMAXNMV and UNSPEC_FMINNMV.
2197 * config/aarch64/aarch64-simd.md (reduc_<optab>_scal_<mode>): Fix
2199 (reduc_<fmaxmin>_scal_<mode>): New pattern.
2200 * config/aarch64/aarch64-sve.md (reduc_<fmaxmin>_scal_<mode>):
2203 2021-11-30 Richard Sandiford <richard.sandiford@arm.com>
2205 * builtins.h (associated_internal_fn): Declare overload that
2206 takes a (combined_cfn, return type) pair.
2207 * builtins.c (associated_internal_fn): Split new overload out
2208 of original fndecl version. Also provide an overload that takes
2209 a (combined_cfn, return type) pair.
2210 * internal-fn.h (commutative_binary_fn_p): Declare.
2211 (commutative_ternary_fn_p): Likewise.
2212 (associative_binary_fn_p): Likewise.
2213 * internal-fn.c (commutative_binary_fn_p, commutative_ternary_fn_p):
2214 New functions, split out from...
2215 (first_commutative_argument): ...here.
2216 (associative_binary_fn_p): New function.
2217 * gimple-match.h (code_helper): Add a constructor that takes
2219 (commutative_binary_op_p): Declare.
2220 (commutative_ternary_op_p): Likewise.
2221 (first_commutative_argument): Likewise.
2222 (associative_binary_op_p): Likewise.
2223 (canonicalize_code): Likewise.
2224 (directly_supported_p): Likewise.
2225 (get_conditional_internal_fn): Likewise.
2226 (gimple_build): New overloads that takes a code_helper.
2227 * gimple-fold.c (gimple_build): Likewise.
2228 * gimple-match-head.c (commutative_binary_op_p): New function.
2229 (commutative_ternary_op_p): Likewise.
2230 (first_commutative_argument): Likewise.
2231 (associative_binary_op_p): Likewise.
2232 (canonicalize_code): Likewise.
2233 (directly_supported_p): Likewise.
2234 (get_conditional_internal_fn): Likewise.
2235 * tree-vectorizer.h: Include gimple-match.h.
2236 (neutral_op_for_reduction): Take a code_helper instead of a tree_code.
2237 (needs_fold_left_reduction_p): Likewise.
2238 (reduction_fn_for_scalar_code): Likewise.
2239 (vect_can_vectorize_without_simd_p): Declare a nNew overload that takes
2241 * tree-vect-loop.c: Include case-cfn-macros.h.
2242 (fold_left_reduction_fn): Take a code_helper instead of a tree_code.
2243 (reduction_fn_for_scalar_code): Likewise.
2244 (neutral_op_for_reduction): Likewise.
2245 (needs_fold_left_reduction_p): Likewise.
2246 (use_mask_by_cond_expr_p): Likewise.
2247 (build_vect_cond_expr): Likewise.
2248 (vect_create_partial_epilog): Likewise. Use gimple_build rather
2249 than gimple_build_assign.
2250 (check_reduction_path): Handle calls and operate on code_helpers
2251 rather than tree_codes.
2252 (vect_is_simple_reduction): Likewise.
2253 (vect_model_reduction_cost): Likewise.
2254 (vect_find_reusable_accumulator): Likewise.
2255 (vect_create_epilog_for_reduction): Likewise.
2256 (vect_transform_cycle_phi): Likewise.
2257 (vectorizable_reduction): Likewise. Make more use of
2259 (vect_transform_reduction): Use gimple_extract_op but expect
2260 a tree_code for now.
2261 (vect_can_vectorize_without_simd_p): New overload that takes
2263 * tree-vect-stmts.c (vectorizable_call): Handle reductions in
2265 * tree-vect-patterns.c (vect_mark_pattern_stmts): Use
2266 gimple_extract_op when updating STMT_VINFO_REDUC_IDX.
2268 2021-11-30 Richard Sandiford <richard.sandiford@arm.com>
2270 * gimple-match.h (code_helper): Provide == and != overloads.
2271 (code_helper::operator tree_code): Make explicit.
2272 (code_helper::operator combined_fn): Likewise.
2273 * gimple-match-head.c (convert_conditional_op): Use explicit
2274 conversions where necessary.
2275 (gimple_resimplify1, gimple_resimplify2, gimple_resimplify3): Likewise.
2276 (maybe_push_res_to_seq, gimple_simplify): Likewise.
2277 * gimple-fold.c (replace_stmt_with_simplification): Likewise.
2279 2021-11-30 Richard Sandiford <richard.sandiford@arm.com>
2281 * gimple-match.h (code_helper): Add functions for querying whether
2282 the code represents an internal_fn or a built_in_function.
2283 Provide explicit conversion operators for both cases.
2284 (gimple_extract_op): Declare.
2285 * gimple-match-head.c (gimple_extract): New function, extracted from...
2286 (gimple_simplify): ...here.
2287 (gimple_extract_op): New function.
2289 2021-11-30 Eric Botcazou <ebotcazou@adacore.com>
2292 * config/i386/i386.c (ix86_output_call_insn): Beef up comment about
2293 nops emitted with SEH.
2294 * config/i386/winnt.c (i386_pe_seh_unwind_emit): When switching to
2295 the cold section, emit a nop before the directive if the previous
2296 active instruction is a call.
2298 2021-11-30 Roger Sayle <roger@nextmovesoftware.com>
2299 Uroš Bizjak <ubizjak@gmail.com>
2302 * config/i386/sse.md (define_split): New splitters to simplify
2303 not;vec_duplicate;and as vec_duplicate;andn.
2305 2021-11-30 Richard Biener <rguenther@suse.de>
2307 PR middle-end/103485
2308 * match.pd (-((int)x >> (prec - 1)) to (unsigned)x >> (prec - 1)):
2309 Use VIEW_CONVERT_EXPR for vectors.
2311 2021-11-30 Richard Biener <rguenther@suse.de>
2313 * cfgrtl.c (skip_insns_after_block): Refactor code to
2314 be more easily readable.
2315 * expr.c (op_by_pieces_d::run): Remove unreachable
2317 * sched-deps.c (sched_analyze): Remove unreachable
2319 * sel-sched-ir.c (in_same_ebb_p): Likewise.
2320 * tree-ssa-alias.c (nonoverlapping_refs_since_match_p):
2321 Remove unreachable code.
2322 * tree-vect-slp.c (vectorize_slp_instance_root_stmt):
2323 Refactor to avoid unreachable loop iteration.
2324 * tree.c (walk_tree_1): Remove unreachable break.
2325 * vec-perm-indices.c (vec_perm_indices::series_p): Remove
2328 2021-11-30 Kewen Lin <linkw@linux.ibm.com>
2331 * config/rs6000/rs6000-call.c (rs6000_builtin_decl): Remove builtin mask
2334 2021-11-30 Kewen Lin <linkw@linux.ibm.com>
2336 * config/rs6000/rs6000.c
2337 (rs6000_cost_data::update_target_cost_per_stmt): Adjust the way to
2338 compute extra penalized cost. Remove useless parameter.
2339 (rs6000_cost_data::rs6000_add_stmt_cost): Adjust the call to function
2340 update_target_cost_per_stmt.
2342 2021-11-30 Kewen Lin <linkw@linux.ibm.com>
2345 2021-11-17 Kewen Lin <linkw@linux.ibm.com>
2347 * config/visium/visium.md (*add<mode>3_insn, *addsi3_insn, *addi3_insn,
2348 *sub<mode>3_insn, *subsi3_insn, *subdi3_insn, *neg<mode>2_insn,
2349 *negdi2_insn, *and<mode>3_insn, *ior<mode>3_insn, *xor<mode>3_insn,
2350 *one_cmpl<mode>2_insn, *ashl<mode>3_insn, *ashr<mode>3_insn,
2351 *lshr<mode>3_insn, *trunchiqi2_insn, *truncsihi2_insn,
2352 *truncdisi2_insn, *extendqihi2_insn, *extendqisi2_insn,
2353 *extendhisi2_insn, *extendsidi2_insn, *zero_extendqihi2_insn,
2354 *zero_extendqisi2_insn, *zero_extendsidi2_insn): Fix split condition.
2356 2021-11-30 Andrew MacLeod <amacleod@redhat.com>
2358 PR tree-optimization/103467
2359 * gimple-range-gori.cc (range_def_chain::register_dependency): Don't
2360 use an object reference after a potential resize.
2362 2021-11-29 Uroš Bizjak <ubizjak@gmail.com>
2365 * config/i386/i386.md (*movhi_internal): Introduce (*v,C) alternative.
2366 Do not allocate non-GPR registers. Optimize xmm->xmm moves when
2367 optimizing for size. Fix vpinsrw insn template.
2368 (*movhf_internal): Fix pinsrw and pextrw insn templates for
2369 AVX targets. Use sselog1 type instead of sselog. Optimize GPR moves.
2370 Optimize xmm->xmm moves for TARGET_SSE_PARTIAL_REG_DEPENDENCY
2371 and TARGET_SSE_SPLIT_REGS targets.
2373 2021-11-29 Eric Gallager <egallager@gcc.gnu.org>
2375 * configure: Re-regenerate.
2377 2021-11-29 Eric Gallager <egallager@gcc.gnu.org>
2380 * Makefile.in: Substitute CTAGS, ETAGS, and CSCOPE
2381 variables. Use ETAGS variable in TAGS target.
2382 * configure: Regenerate.
2383 * configure.ac: Allow CTAGS, ETAGS, and CSCOPE
2384 variables to be overridden.
2386 2021-11-29 Paul A. Clarke <pc@us.ibm.com>
2388 * config/rs6000/xmmintrin.h (_mm_movemask_ps): Use vec_extractm
2390 * config/rs6000/emmintrin.h (_mm_movemask_pd): Likewise.
2391 (_mm_movemask_epi8): Likewise.
2393 2021-11-29 Richard Biener <rguenther@suse.de>
2395 * read-rtl-function.c (function_reader::read_rtx_operand):
2396 Return only after resetting m_in_call_function_usage.
2398 2021-11-29 Roger Sayle <roger@nextmovesoftware.com>
2400 * config/i386/i386-expand.c (ix86_expand_v1ti_to_ti): Perform the
2401 conversion via V2DImode using vec_extractv2didi on TARGET_SSE2.
2402 * config/i386/sse.md (rotlv1ti3, rotrv1ti3): Change constraint
2403 on QImode shift amounts from const_int_operand to general_operand.
2405 2021-11-29 Richard Biener <rguenther@suse.de>
2407 * tree.h (reverse_storage_order_for_component_p): Remove
2408 spurious gcc_unreachable.
2409 * cfganal.c (dfs_find_deadend): Likewise.
2410 * fold-const-call.c (fold_const_logb): Likewise.
2411 (fold_const_significand): Likewise.
2412 * gimple-ssa-store-merging.c (lhs_valid_for_store_merging_p):
2415 2021-11-29 Richard Biener <rguenther@suse.de>
2417 * vec.c (qsort_chk): Do not return the void return value
2418 from the noreturn qsort_chk_error.
2419 * ccmp.c (expand_ccmp_expr_1): Remove unreachable return.
2420 * df-scan.c (df_ref_equal_p): Likewise.
2421 * dwarf2out.c (is_base_type): Likewise.
2422 (add_const_value_attribute): Likewise.
2423 * fixed-value.c (fixed_arithmetic): Likewise.
2424 * gimple-fold.c (gimple_fold_builtin_fputs): Likewise.
2425 * gimple-ssa-strength-reduction.c (stmt_cost): Likewise.
2426 * graphite-isl-ast-to-gimple.c
2427 (gcc_expression_from_isl_expr_op): Likewise.
2428 (gcc_expression_from_isl_expression): Likewise.
2429 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate):
2431 * lto-streamer-in.c (lto_input_mode_table): Likewise.
2433 2021-11-29 liuhongt <hongtao.liu@intel.com>
2436 * config/i386/i386.c (inline_secondary_memory_needed): HImode
2437 move between GPR and SSE registers is supported under
2438 TARGET_SSE2 and above.
2439 * config/i386/i386.md (extendhfsf2): Optimize expander.
2440 (truncsfhf2): Ditto.
2441 * config/i386/sse.md (sse2p4_1): Adjust attr for V8HFmode to
2442 align with V8HImode.
2444 2021-11-29 liuhongt <hongtao.liu@intel.com>
2448 * config/i386/i386.c (ix86_secondary_reload): Without
2449 TARGET_SSE4_1, General register is needed to move HImode from
2450 sse register to memory.
2451 * config/i386/sse.md (*vec_extrachf): Use %vpextrw instead of
2452 pextrw in output templates.
2453 * config/i386/i386.md (movhi_internal): Ditto, also fix typo of
2454 MEM_P (operands[1]) and adjust mode/prefix/type attribute for
2455 alternatives related to sse register.
2457 2021-11-29 Richard Biener <rguenther@suse.de>
2459 PR tree-optimization/103458
2460 * tree-ssa-dce.c (make_forwarders_with_degenerate_phis): Do not
2461 create forwarders for blocks with abnormal predecessors.
2463 2021-11-29 Richard Biener <rguenther@suse.de>
2465 * gimple-predicate-analysis.cc (can_be_invalidated_p):
2466 Restore semantics to the one before the split from
2469 2021-11-28 Jan Hubicka <hubicka@ucw.cz>
2471 * profile.c: Include sreal.h
2472 (struct bb_stats): New.
2473 (cmp_stats): New function.
2474 (compute_branch_probabilities): Output bb stats.
2476 2021-11-28 Jan Hubicka <hubicka@ucw.cz>
2478 * cfghooks.c: Include sreal.h, profile.h.
2479 (profile_record_check_consistency): Fix checking of count counsistency;
2480 record also dynamic mismatches.
2481 * cfgrtl.c (rtl_account_profile_record): Similarly.
2482 * tree-cfg.c (gimple_account_profile_record): Likewise.
2483 * cfghooks.h (struct profile_record): Remove num_mismatched_freq_in,
2484 num_mismatched_freq_out, turn time to double, add
2485 dyn_mismatched_prob_out, dyn_mismatched_count_in,
2486 num_mismatched_prob_out; remove num_mismatched_count_out.
2487 * passes.c (account_profile_1): New function.
2488 (account_profile_in_list): New function.
2489 (pass_manager::dump_profile_report): Rewrite.
2490 (execute_one_ipa_transform_pass): Check profile consistency after
2492 (execute_all_ipa_transforms): Remove cfun test; record all transform
2494 (execute_one_pass): Fix collecting of profile stats.
2496 2021-11-27 Jakub Jelinek <jakub@redhat.com>
2498 PR middle-end/103431
2499 * config/i386/i386.md (x86_64_shld, x86_shld, x86_64_shrd, x86_shrd):
2500 Change insn pattern to accurately describe the instructions.
2501 (*x86_64_shld_1, *x86_shld_1, *x86_64_shrd_1, *x86_shrd_1): New
2502 define_insn patterns.
2503 (*x86_64_shld_2, *x86_shld_2, *x86_64_shrd_2, *x86_shrd_2): New
2504 define_insn_and_split patterns.
2505 (*ashl<dwi>3_doubleword_mask, *ashl<dwi>3_doubleword_mask_1,
2506 *<insn><dwi>3_doubleword_mask, *<insn><dwi>3_doubleword_mask_1,
2507 ix86_rotl<dwi>3_doubleword, ix86_rotr<dwi>3_doubleword): Adjust
2508 splitters for x86_{,64_}sh{l,r}d pattern changes.
2510 2021-11-27 Jakub Jelinek <jakub@redhat.com>
2512 PR tree-optimization/103435
2513 * gimple-ssa-store-merging.c (find_bswap_or_nop_finalize): Avoid UB if
2514 n->range - rsize == 8, just clear both *cmpnop and *cmpxchg in that
2517 2021-11-27 Martin Jambor <mjambor@suse.cz>
2520 * ipa-prop.c (ipcp_transform_function): Call
2521 delete_unreachable_blocks_update_callgraph instead of returning
2524 2021-11-26 Siddhesh Poyarekar <siddhesh@gotplt.org>
2526 * tree-object-size.c (object_sizes_grow, object_sizes_release,
2527 object_sizes_unknown_p, object_sizes_get, object_size_set_force,
2528 object_sizes_set): New functions.
2529 (addr_object_size, compute_builtin_object_size,
2530 expr_object_size, call_object_size, unknown_object_size,
2531 merge_object_sizes, plus_stmt_object_size,
2532 cond_expr_object_size, collect_object_sizes_for,
2533 check_for_plus_in_loops_1, init_object_sizes,
2534 fini_object_sizes): Adjust.
2536 2021-11-26 Siddhesh Poyarekar <siddhesh@gotplt.org>
2538 * tree-object-size.c: New enum.
2539 (object_sizes, computed, addr_object_size,
2540 compute_builtin_object_size, expr_object_size, call_object_size,
2541 merge_object_sizes, plus_stmt_object_size,
2542 collect_object_sizes_for, init_object_sizes, fini_object_sizes,
2543 object_sizes_execute): Replace magic numbers with enums.
2545 2021-11-26 Roger Sayle <roger@nextmovesoftware.com>
2547 * tree-ssa-loop-ivopts.c (cand_value_at): Take a class
2548 tree_niter_desc* argument instead of just a tree for NITER.
2549 If we require the iv candidate value at the end of the final
2550 loop iteration, try using the original loop bound as the
2551 NITER for sufficiently simple loops.
2552 (may_eliminate_iv): Update (only) call to cand_value_at.
2554 2021-11-26 Jan Hubicka <hubicka@ucw.cz>
2556 * ipa-modref.c (analyze_function): Drop parameter F and dump
2557 cgraph node name rather than cfun name.
2558 (modref_generate): Update.
2559 (modref_summaries::insert):Update.
2560 (modref_summaries_lto::insert):Update.
2561 (pass_modref::execute):Update.
2562 (ipa_merge_modref_summary_after_inlining): Improve combining of
2565 2021-11-26 Jan Hubicka <jh@suse.cz>
2568 * ipa-modref.c (update_escape_summary_1): Fix handling of min_flags.
2570 2021-11-26 konglin1 <lingling.kong@intel.com>
2573 * config/i386/i386.c (ix86_can_change_mode_class): Allow 16 bit data in XMM register
2575 * config/i386/i386.md (extendhfsf2): Add extenndhfsf2 for TARGET_F16C.
2576 (extendhfdf2): Restrict extendhfdf for TARGET_AVX512FP16 only.
2577 (*extendhf<mode>2): Rename from extendhf<mode>2.
2578 (truncsfhf2): Likewise.
2579 (truncdfhf2): Likewise.
2580 (*trunc<mode>2): Likewise.
2582 2021-11-26 liuhongt <hongtao.liu@intel.com>
2584 PR middle-end/103419
2585 * match.pd: Fix typo, use the type of second parameter, not
2588 2021-11-25 Jan Hubicka <jh@suse.cz>
2590 * ipa-cp.c (ipa_value_range_from_jfunc): Remove forgotten early return.
2592 2021-11-25 Roger Sayle <roger@nextmovesoftware.com>
2594 PR middle-end/103406
2595 * match.pd (minus @0 @0): Check tree_expr_maybe_infinite_p.
2597 2021-11-25 Martin Jambor <mjambor@suse.cz>
2600 * ipa-prop.h (ipa_get_param): New overload. Move bits of the existing
2602 * ipa-param-manipulation.h (ipa_param_adjustments): New member
2603 function get_updated_index_or_split.
2604 * ipa-param-manipulation.c
2605 (ipa_param_adjustments::get_updated_index_or_split): New function.
2606 * ipa-prop.c (adjust_agg_replacement_values): Reimplement, add
2607 capability to identify scalarized parameters and perform substitution
2609 (ipcp_transform_function): Create descriptors earlier, handle new
2610 return values of adjust_agg_replacement_values.
2612 2021-11-25 Aldy Hernandez <aldyh@redhat.com>
2614 * gimple-range-path.cc (path_range_query::compute_ranges_defined): Remove.
2615 (path_range_query::compute_ranges_in_block): Revert to bitmap order.
2616 * gimple-range-path.h: Remove compute_ranges_defined.
2618 2021-11-25 Andrew Stubbs <ams@codesourcery.com>
2621 * config/gcn/gcn.c (move_callee_saved_registers): Ensure that the
2622 number of spilled registers is counted correctly.
2624 2021-11-25 Jan Hubicka <hubicka@ucw.cz>
2626 * ipa-prop.h (ipa_node_params::ipa_node_params): Initialize
2629 2021-11-25 Andrew MacLeod <amacleod@redhat.com>
2631 PR tree-optimization/103359
2632 * gimple-range-fold.cc (fold_using_range::range_of_phi): If arg is
2633 equivalent to def, don't initially include it's range.
2635 2021-11-25 Jan Hubicka <hubicka@ucw.cz>
2637 * tree-ssa-alias.c (ref_maybe_used_by_call_p_1): Do not check
2638 gimple_static_chain.
2640 2021-11-25 Richard Biener <rguenther@suse.de>
2642 * cgraphunit.c (symbol_table::output_weakrefs): Remove
2644 (get_alias_symbol): Remove now unused function.
2646 2021-11-25 Richard Biener <rguenther@suse.de>
2648 * cfgrtl.c (rtl_verify_fallthru): Do not stop verifying
2650 (skip_insns_after_block): Remove unreachable break and continue.
2652 2021-11-25 Richard Biener <rguenther@suse.de>
2654 * cfgexpand.c (label_rtx_for_bb): Remove dead loop construct.
2656 2021-11-25 Richard Biener <rguenther@suse.de>
2658 * regset.h (REG_SET_EMPTY_P): New macro.
2659 * cfgcleanup.c (thread_jump): Use REG_SET_EMPTY_P.
2661 2021-11-25 Martin Liska <mliska@suse.cz>
2663 * doc/invoke.texi: Use @option for -Wuninitialized.
2665 2021-11-25 Aldy Hernandez <aldyh@redhat.com>
2667 PR tree-optimization/103254
2668 * gimple-range-path.cc (path_range_query::compute_ranges): Move
2669 exported boolean code...
2670 (path_range_query::compute_imports): ...here.
2672 2021-11-25 Aldy Hernandez <aldyh@redhat.com>
2674 PR tree-optimization/103254
2675 * gimple-range-path.cc (path_range_query::compute_ranges_defined): New
2676 (path_range_query::compute_ranges_in_block): Move to
2677 compute_ranges_defined.
2678 * gimple-range-path.h (compute_ranges_defined): New.
2680 2021-11-25 Jakub Jelinek <jakub@redhat.com>
2682 PR tree-optimization/103417
2683 * match.pd ((X & Y) CMP 0): Only handle eq and ne. Commonalize
2686 2021-11-25 Jakub Jelinek <jakub@redhat.com>
2688 PR tree-optimization/103376
2689 * gimple-ssa-store-merging.c (perform_symbolic_merge): For
2690 BIT_IOR_EXPR, if masked1 && masked2 && masked1 != masked2, don't
2691 punt, but set the corresponding result byte to MARKER_BYTE_UNKNOWN.
2692 For BIT_XOR_EXPR similarly and if masked1 == masked2 and the
2693 byte isn't MARKER_BYTE_UNKNOWN, set the corresponding result byte to
2696 2021-11-25 Jakub Jelinek <jakub@redhat.com>
2699 * doc/invoke.texi (-Wcomma-subscript): Document that for
2700 -std=c++20 the option isn't enabled by default with -Wno-deprecated
2701 but for -std=c++23 it is.
2703 2021-11-24 Jan Hubicka <hubicka@ucw.cz>
2705 * ipa-modref.c (implicit_eaf_flags_for_edge_and_arg): Break out from...
2706 (modref_merge_call_site_flags): ... here.
2707 (ipa_merge_modref_summary_after_inlining): Use it.
2709 2021-11-24 Thomas Schwinge <thomas@codesourcery.com>
2711 * cfgloop.c (verify_loop_structure): Reduce scope of
2712 'class loop *loop' variable.
2713 * ipa-fnsummary.c (analyze_function_body): Likewise.
2714 * loop-init.c (fix_loop_structure): Likewise.
2715 * loop-invariant.c (calculate_loop_reg_pressure): Likewise.
2716 * predict.c (predict_loops): Likewise.
2717 * tree-loop-distribution.c (loop_distribution::execute): Likewise.
2718 * tree-vectorizer.c (pass_vectorize::execute): Likewise.
2720 2021-11-24 Andrew MacLeod <amacleod@redhat.com>
2722 PR tree-optimization/103231
2723 * gimple-range.cc (gimple_ranger::gimple_ranger): Create stmt stack.
2724 (gimple_ranger::gimple_ranger): Delete stmt stack.
2725 (gimple_ranger::range_of_stmt): Process depenedencies if they have no
2727 (gimple_ranger::prefill_name): New.
2728 (gimple_ranger::prefill_stmt_dependencies): New.
2729 * gimple-range.h (class gimple_ranger): Add prototypes.
2731 2021-11-24 Andrew MacLeod <amacleod@redhat.com>
2733 * gimple-range-cache.cc (ranger_cache::get_global_range): Always
2734 return a range, return if it came from the cache or not.
2735 (get_non_stale_global_range): Rename to get_global_range, and return
2736 the temporal state in a flag.
2737 * gimple-range-cache.h (get_non_stale_global_range): Rename and adjust.
2738 * gimple-range.cc (gimple_ranger::range_of_expr): No need to query
2740 (gimple_ranger::range_of_stmt): Adjust for global cache temporal state
2743 2021-11-24 Andrew MacLeod <amacleod@redhat.com>
2745 * gimple-range.cc (gimple_ranger::range_on_edge): Call trailer when
2746 a constant is encountered to terminate the trace.
2748 2021-11-24 Maciej W. Rozycki <macro@embecosm.com>
2750 PR middle-end/103059
2751 * reload.c (find_reloads_address_1): Also accept the ASHIFT form
2752 of indexed addressing.
2753 (find_reloads): Adjust accordingly.
2755 2021-11-24 Richard Biener <rguenther@suse.de>
2756 Jan Hubicka <jh@suse.cz>
2758 PR tree-optimization/103168
2759 * ipa-modref.h (struct modref_summary): Add load_accesses.
2760 * ipa-modref.c (modref_summary::finalize): Initialize load_accesses.
2761 * tree-ssa-sccvn.c (visit_reference_op_call): Use modref
2762 info to walk the virtual use->def chain to CSE const/pure
2763 function calls possibly reading from memory.
2765 2021-11-24 Thomas Schwinge <thomas@codesourcery.com>
2767 PR middle-end/103244
2768 * gimplify.c (gimplify_adjust_omp_clauses): Restore previous
2771 2021-11-24 Richard Biener <rguenther@suse.de>
2773 PR middle-end/103193
2774 * match.pd: Avoid canonicalizing (le/ge @0 @0) to (eq @0 @0)
2775 with NaNs and -ftrapping-math.
2777 2021-11-24 Jakub Jelinek <jakub@redhat.com>
2779 PR middle-end/103384
2780 * omp-general.c (omp_context_selector_matches): For ACCEL_COMPILER,
2781 return 0 for kind(host) and continue for kind(nohost).
2783 2021-11-24 Jakub Jelinek <jakub@redhat.com>
2785 PR middle-end/103365
2786 * attribs.h (lookup_attribute): Allow attr_name to start with
2787 underscore, as long as canonicalize_attr_name returns false.
2788 (lookup_attribute_by_prefix): Don't call get_attribute_name twice.
2789 * attribs.c (extract_attribute_substring): Reimplement using
2790 canonicalize_attr_name.
2791 (register_scoped_attribute): Change gcc_assert into
2792 gcc_checking_assert, verify !canonicalize_attr_name rather than
2793 that str.str doesn't start with '_'.
2795 2021-11-24 Jakub Jelinek <jakub@redhat.com>
2797 PR tree-optimization/103376
2798 * gimple-ssa-store-merging.c (perform_symbolic_merge): Add CODE
2799 argument. If CODE is not BIT_IOR_EXPR, ensure that one of masked1
2801 (find_bswap_or_nop_1, find_bswap_or_nop,
2802 imm_store_chain_info::try_coalesce_bswap): Adjust
2803 perform_symbolic_merge callers.
2805 2021-11-24 Richard Biener <rguenther@suse.de>
2807 * tree-ssa-loop-ivopts.c (find_givs): Take loop body as
2808 argument instead of re-computing it.
2809 (find_interesting_uses): Likewise.
2810 (find_induction_variables): Pass through loop body.
2811 (tree_ssa_iv_optimize_loop): Pass down loop body.
2813 2021-11-24 Tamar Christina <tamar.christina@arm.com>
2815 * tree-ssa-phiopt.c (spaceship_replacement): Handle new canonical
2818 2021-11-24 Tamar Christina <tamar.christina@arm.com>
2820 * tree.c (bitmask_inv_cst_vector_p): New.
2821 * tree.h (bitmask_inv_cst_vector_p): New.
2822 * match.pd: Use it in new bitmask compare pattern.
2824 2021-11-24 Jason Merrill <jason@redhat.com>
2826 * timevar.h (class auto_cond_timevar): New.
2828 2021-11-24 Hongtao Liu <hongtao.liu@intel.com>
2829 H.J. Lu <hongjiu.lu@intel.com>
2831 PR tree-optimization/103194
2832 * match.pd (gimple_nop_atomic_bit_test_and_p): Extended to
2834 * tree-ssa-ccp.c (gimple_nop_convert): Declare.
2835 (optimize_atomic_bit_test_and): Enhance
2836 optimize_atomic_bit_test_and to handle truncation.
2838 2021-11-23 Martin Sebor <msebor@redhat.com>
2841 * Makefile.in (OBJS): Add gimple-warn-recursion.o.
2842 * common.opt: Add -Winfinite-recursion.
2843 * doc/invoke.texi (-Winfinite-recursion): Document.
2844 * passes.def (pass_warn_recursion): Schedule a new pass.
2845 * tree-pass.h (make_pass_warn_recursion): Declare.
2846 * gimple-warn-recursion.c: New file.
2848 2021-11-23 Bill Schmidt <wschmidt@linux.ibm.com>
2850 * config/rs6000/rs6000-call.c (rs6000_invalid_new_builtin): Change
2851 error messages for ENB_P8V and ENB_P9V.
2853 2021-11-23 Bill Schmidt <wschmidt@linux.ibm.com>
2855 * config/rs6000/rs6000-builtin-new.def: Add power6-64 stanza. Move
2856 CMPB to power6-64 stanza.
2857 * config/rs6000/rs6000-call.c (rs6000_invalid_new_builtin): Handle
2859 (rs6000_new_builtin_is_supported): Likewise.
2860 (rs6000_expand_new_builtin): Likewise. Clean up formatting.
2861 (rs6000_init_builtins): Handle ENB_P6_64 case.
2862 * config/rs6000/rs6000-gen-builtins.c (bif_stanza): Add BSTZ_P6_64.
2863 (stanza_map): Add entry mapping power6-64 to BSTZ_P6_64.
2864 (enable_string): Add "ENB_P6_64".
2865 (write_decls): Add ENB_P6_64 to bif_enable enum.
2867 2021-11-23 Jan Hubicka <jh@suse.cz>
2869 * ipa-modref-tree.h (struct modref_tree): Remove max_bases, max_refs
2871 (modref_tree::modref_tree): Remove parametr.
2872 (modref_tree::insert_base): Add max_bases parameter.
2873 (modref_tree::insert): Add max_bases, max_refs, max_accesses
2875 (modref_tree::insert): New member function.
2876 (modref_tree::merge): Add max_bases, max_refs, max_accesses
2878 (modref_tree::insert): New member function.
2879 * ipa-modref-tree.c (test_insert_search_collapse): Update.
2880 (test_merge): Update.
2881 * ipa-modref.c (dump_records): Don't dump max_refs and max_bases.
2882 (dump_lto_records): Likewise.
2883 (modref_summary::finalize): Fix whitespace.
2884 (get_modref_function_summary): Likewise.
2885 (modref_access_analysis::record_access): Update.
2886 (modref_access_analysis::record_access_lto): Update.
2887 (modref_access_analysis::process_fnspec): Update.
2888 (analyze_function): Update.
2889 (modref_summaries::duplicate): Update.
2890 (modref_summaries_lto::duplicate): Update.
2891 (write_modref_records): Update.
2892 (read_modref_records): Update.
2893 (read_section): Update.
2894 (propagate_unknown_call): Update.
2895 (modref_propagate_in_scc): Update.
2896 (ipa_merge_modref_summary_after_inlining): Update.
2898 2021-11-23 Martin Liska <mliska@suse.cz>
2900 * doc/invoke.texi: Remove 2 more duplicite param descriptions.
2902 2021-11-23 Richard Biener <rguenther@suse.de>
2904 PR tree-optimization/103361
2905 * gimple-loop-jam.c (adjust_unroll_factor): Use lambda_int
2906 for the dependence distance.
2907 * tree-data-ref.c (print_lambda_vector): Properly print a lambda_int.
2909 2021-11-23 Jakub Jelinek <jakub@redhat.com>
2911 * tree-inline.h (struct copy_body_data): Remove
2912 transform_lang_insert_block member.
2913 * tree-inline.c (remap_block): Don't call
2914 id->transform_lang_insert_block.
2915 (optimize_inline_calls, copy_gimple_seq_and_replace_locals,
2916 tree_function_versioning, maybe_inline_call_in_expr,
2917 copy_fn): Don't initialize id.transform_lang_insert_block.
2918 * gimplify.c (gimplify_omp_loop): Likewise.
2920 2021-11-23 Jan Hubicka <hubicka@ucw.cz>
2922 PR tree-optimization/103335
2923 * tree-ssa-dse.c (valid_ao_ref_for_dse): Rename to ...
2924 (valid_ao_ref_kill_for_dse): ... this; do not check that boundaries
2925 are divisible by BITS_PER_UNIT.
2926 (get_byte_aligned_range_containing_ref): New function.
2927 (get_byte_aligned_range_contained_in_ref): New function.
2928 (normalize_ref): Rename to ...
2929 (get_byte_range): ... this one; handle accesses not aligned to byte
2930 boundary; return range in bytes rater than updating ao_ref.
2931 (clear_live_bytes_for_ref): Take write ref by reference; simplify using
2933 (setup_live_bytes_from_ref): Likewise.
2934 (clear_bytes_written_by): Update.
2935 (live_bytes_read): Update.
2936 (dse_classify_store): Simplify tech before live_bytes_read checks.
2938 2021-11-23 Andrew Pinski <apinski@marvell.com>
2940 PR tree-optimization/102216
2941 * gimple-fold.c (fold_stmt_1): Add canonicalization
2942 of "&MEM[ssa_n, CST]" to "ssa_n p+ CST", note this
2943 can only be done if !in_place.
2945 2021-11-23 Jakub Jelinek <jakub@redhat.com>
2947 PR middle-end/102431
2948 * gimplify.c (replace_reduction_placeholders): Remove.
2949 (note_no_context_vars): New function.
2950 (gimplify_omp_loop): For OMP_PARALLEL's BIND_EXPR create a new
2951 BLOCK. Use copy_tree_body_r with walk_tree instead of unshare_expr
2952 and replace_reduction_placeholders for duplication of
2953 OMP_CLAUSE_REDUCTION_{INIT,MERGE} expressions. Ensure all mentioned
2954 automatic vars have DECL_CONTEXT set to non-NULL before doing so
2955 and reset it afterwards for those vars and their corresponding
2958 2021-11-23 Haochen Gui <guihaoc@gcc.gnu.org>
2961 * config/rs6000/altivec.md (altivec_vreve<mode>2 for VEC_K): Use
2962 xxbrq for v16qi, xxbrq + xxbrh for v8hi and xxbrq + xxbrw for v4si
2963 or v4sf when p9_vector is set.
2964 (altivec_vreve<mode>2 for VEC_64): Defined. Implemented by xxswapd.
2966 2021-11-23 Navid Rahimi <navidrahimi@microsoft.com>
2968 PR tree-optimization/102232
2969 * match.pd (x * (1 + y / x) - y) -> (x - y % x): New optimization.
2971 2021-11-23 Navid Rahimi <navidrahimi@microsoft.com>
2973 PR tree-optimization/96779
2974 * match.pd (-x == x) -> (x == 0): New optimization.
2976 2021-11-22 Roger Sayle <roger@nextmovesoftware.com>
2978 PR tree-optimization/98953
2979 PR tree-optimization/103345
2980 * gimple-ssa-store-merging.c (find_bswap_or_nop_1): Handle
2981 BIT_XOR_EXPR and PLUS_EXPR the same as BIT_IOR_EXPR.
2982 (pass_optimize_bswap::execute): Likewise.
2984 2021-11-22 Martin Liska <mliska@suse.cz>
2986 * doc/invoke.texi: Remove duplicate documentation for 3 params.
2988 2021-11-22 Richard Biener <rguenther@suse.de>
2990 PR tree-optimization/103351
2991 * tree-ssa-dce.c (sort_phi_args): Sort after e->dest_idx as
2994 2021-11-22 liuhongt <hongtao.liu@intel.com>
2997 * config/i386/constraints.md (Bk): New
2998 define_memory_constraint.
2999 * config/i386/i386-protos.h (ix86_gpr_tls_address_pattern_p):
3001 * config/i386/i386.c (ix86_gpr_tls_address_pattern_p): New
3003 * config/i386/i386.md (*movsi_internal): Don't allow
3004 mask/sse/mmx move in TLS code sequences.
3005 (*movdi_internal): Ditto.
3007 2021-11-22 Kewen Lin <linkw@linux.ibm.com>
3009 * config/xtensa/xtensa.md (movdi_internal, movdf_internal): Fix split
3012 2021-11-21 Jakub Jelinek <jakub@redhat.com>
3015 * config/i386/i386-options.c (ix86_valid_target_attribute_p): If
3016 fndecl already has DECL_FUNCTION_SPECIFIC_TARGET, use that as base
3017 instead of target_option_default_node.
3019 2021-11-21 Jan Hubicka <hubicka@ucw.cz>
3022 * ipa-modref.c (parm_map_for_arg): Rename to ...
3023 (parm_map_for_ptr): .. this one; handle static chain and calls to
3025 (modref_access_analysis::get_access): Use parm_map_for_ptr.
3026 (modref_access_analysis::process_fnspec): Update.
3027 (modref_access_analysis::analyze_load): Update.
3028 (modref_access_analysis::analyze_store): Update.
3030 2021-11-21 Jan Hubicka <hubicka@ucw.cz>
3032 * ipa-modref.c (ignore_nondeterminism_p): Move earlier in source
3034 (ignore_retval_p): Likewise.
3035 (ignore_stores_p): Likewise.
3036 (parm_map_for_arg): Likewise.
3037 (class modref_access_analysis): New class.
3038 (modref_access_analysis::set_side_effects): New member function.
3039 (modref_access_analysis::set_nondeterministic): New member function.
3040 (get_access): Turn to ...
3041 (modref_access_analysis::get_access): ... this one.
3042 (record_access): Turn to ...
3043 (modref_access_analysis::record_access): ... this one.
3044 (record_access_lto): Turn to ...
3045 (modref_access_analysis::record_access_lto): ... This one.
3046 (record_access_p): Turn to ...
3047 (modref_access_analysis::record_access_p): ... This one
3048 (modref_access_analysis::record_unknown_load): New member function.
3049 (modref_access_analysis::record_unknown_store): New member function.
3050 (get_access_for_fnspec): Turn to ...
3051 (modref_access_analysis::get_access_for_fnspec): ... this one.
3052 (merge_call_side_effects): Turn to ...
3053 (moderf_access_analysis::merge_call_side_effects): Turn to ...
3054 (collapse_loads): Move later in source code.
3055 (collapse_stores): Move later in source code.
3056 (process_fnspec): Turn to ...
3057 (modref_access_analysis::process_fnspec): ... this one.
3058 (analyze_call): Turn to ...
3059 (modref_access_analysis::analyze_call): ... this one.
3060 (struct summary_ptrs): Remove.
3061 (analyze_load): Turn to ...
3062 (modref_access_analysis::analyze_load): ... this one.
3063 (analyze_store): Turn to ...
3064 (modref_access_analysis::analyze_store): ... this one.
3065 (analyze_stmt): Turn to ...
3066 (modref_access_analysis::analyze_stmt): ... This one.
3067 (remove_summary): Remove.
3068 (modref_access_analysis::propagate): Break out from ...
3069 (modref_access_analysis::analyze): Break out from ...
3070 (analyze_function): ... here.
3072 2021-11-21 Roger Sayle <roger@nextmovesoftware.com>
3073 Robin Dapp <rdapp@linux.ibm.com>
3076 * tree-ssa-math-opts.c (convert_mult_to_widen): Recognize
3077 signed WIDEN_MULT_EXPR if the target supports umul_widen_optab.
3079 2021-11-20 Jan Hubicka <hubicka@ucw.cz>
3082 * ipa-modref.c (ignore_nondeterminism_p): Allow looping pure/cont.
3083 (merge_call_side_effects): Improve debug output.
3085 2021-11-20 Jan Hubicka <hubicka@ucw.cz>
3088 * ipa-pure-const.c (propagate_pure_const): Fix merging of loping flag.
3090 2021-11-20 Jeff Law <jeffreyalaw@gmail.com>
3092 PR tree-optimization/103226
3093 * config/bfin/bfin.md (doloop pattern, splitter and expander): Clobber
3096 2021-11-20 Andrew Pinski <apinski@marvell.com>
3098 PR tree-optimization/103220
3099 * match.pd ((type) X bitop CST): Don't check if CST
3100 fits into the type if only the sign changes.
3102 2021-11-20 Alexandre Oliva <oliva@adacore.com>
3104 PR tree-optimization/102988
3105 * gimple-harden-conditionals.cc (detach_value): Copy SSA_NAME
3106 without decl sharing.
3108 2021-11-19 Iain Sandoe <iain@sandoe.co.uk>
3110 * doc/invoke.texi: Remove whitespace after an @option.
3112 2021-11-19 Paul A. Clarke <pc@us.ibm.com>
3114 * config/rs6000/emmintrin.h (_mm_sad_epu8): Use vec_absd when
3115 _ARCH_PWR9, optimize vec_sum2s when LE.
3117 2021-11-19 Iain Sandoe <iain@sandoe.co.uk>
3120 * config/darwin-driver.c (darwin_driver_init): Handle exported
3121 symbols and symbol lists (suppress automatic export of the TLS
3123 * config/darwin.c (darwin_rename_builtins): Remove workaround.
3124 * config/darwin.h (LINK_GCC_C_SEQUENCE_SPEC): Likewise.
3125 (REAL_LIBGCC_SPEC): Handle revised library uses.
3126 * config/darwin.opt (nodefaultexport): New.
3127 * config/i386/darwin.h (PR80556_WORKAROUND): Remove.
3128 * config/i386/darwin32-biarch.h (PR80556_WORKAROUND): Likewise.
3129 * config/i386/darwin64-biarch.h (PR80556_WORKAROUND): Likewise.
3131 2021-11-19 Martin Jambor <mjambor@suse.cz>
3133 * opts.c (default_options_table): Switch off
3134 flag_semantic_interposition at Ofast.
3135 * doc/invoke.texi (Optimize Options): Document that Ofast switches off
3136 -fsemantic-interposition.
3138 2021-11-19 Jan Hubicka <hubicka@ucw.cz>
3140 * tree-ssa-alias.c (ref_maybe_used_by_call_p_1): Do not guard modref
3141 by !gimple_call_chain.
3143 2021-11-19 Martin Sebor <msebor@redhat.com>
3147 * doc/invoke.texi (-Waddress): Update.
3149 2021-11-19 Andrew MacLeod <amacleod@redhat.com>
3151 PR tree-optimization/103254
3152 * gimple-range-gori.cc (range_def_chain::get_def_chain): Limit the
3153 depth for all statements with multple ssa names.
3155 2021-11-19 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
3157 * config/s390/s390.md (define_peephole2): Variable insn points
3158 to the first matched insn. Use peep2_next_insn(1) to refer to
3159 the second matched insn.
3161 2021-11-19 Tamar Christina <tamar.christina@arm.com>
3163 PR tree-optimization/103311
3165 * tree-vect-slp-patterns.c (vect_validate_multiplication): Fix CONJ
3166 test to new codegen.
3167 (complex_mul_pattern::matches): Move check downwards.
3169 2021-11-19 Martin Liska <mliska@suse.cz>
3172 2021-11-19 Martin Liska <mliska@suse.cz>
3174 * cfgexpand.c (pass_expand::execute): Use option directly.
3175 * function.c (allocate_struct_function): Likewise.
3176 * gimple-low.c (lower_function_body): Likewise.
3177 (lower_stmt): Likewise.
3178 * gimple-ssa-backprop.c (backprop::prepare_change): Likewise.
3179 * ipa-param-manipulation.c (ipa_param_adjustments::modify_call): Likewise.
3180 * ipa-split.c (split_function): Likewise.
3181 * lto-streamer-in.c (input_function): Likewise.
3182 * sese.c (sese_insert_phis_for_liveouts): Likewise.
3183 * ssa-iterators.h (num_imm_uses): Likewise.
3184 * tree-cfg.c (make_blocks): Likewise.
3185 (gimple_merge_blocks): Likewise.
3186 * tree-inline.c (tree_function_versioning): Likewise.
3187 * tree-loop-distribution.c (generate_loops_for_partition): Likewise.
3188 * tree-sra.c (analyze_access_subtree): Likewise.
3189 * tree-ssa-dce.c (remove_dead_stmt): Likewise.
3190 * tree-ssa-loop-ivopts.c (remove_unused_ivs): Likewise.
3191 * tree-ssa-phiopt.c (spaceship_replacement): Likewise.
3192 * tree-ssa-reassoc.c (reassoc_remove_stmt): Likewise.
3193 * tree-ssa-tail-merge.c (tail_merge_optimize): Likewise.
3194 * tree-ssa-threadedge.c (propagate_threaded_block_debug_into): Likewise.
3195 * tree-ssa.c (gimple_replace_ssa_lhs): Likewise.
3196 (target_for_debug_bind): Likewise.
3197 (insert_debug_temp_for_var_def): Likewise.
3198 (insert_debug_temps_for_defs): Likewise.
3199 (reset_debug_uses): Likewise.
3200 * tree-ssanames.c (release_ssa_name_fn): Likewise.
3201 * tree-vect-loop-manip.c (adjust_vec_debug_stmts): Likewise.
3202 (adjust_debug_stmts): Likewise.
3203 (adjust_phi_and_debug_stmts): Likewise.
3204 (vect_do_peeling): Likewise.
3205 * tree-vect-loop.c (vect_transform_loop_stmt): Likewise.
3206 (vect_transform_loop): Likewise.
3207 * tree.h (MAY_HAVE_DEBUG_MARKER_STMTS): Remove
3208 (MAY_HAVE_DEBUG_BIND_STMTS): Remove.
3209 (MAY_HAVE_DEBUG_STMTS): Use options directly.
3211 2021-11-19 Giuliano Belinassi <gbelinassi@suse.de>
3213 * gcc.c (process_command): Skip dumpdir override if file is a
3215 * doc/invoke.texi: Update -dumpdir documentation.
3217 2021-11-19 Andrew Pinski <apinski@marvell.com>
3219 PR tree-optimization/103314
3220 * match.pd ((type) X op CST): Restrict the equal
3221 TYPE_PRECISION case to GIMPLE only.
3223 2021-11-19 Martin Liska <mliska@suse.cz>
3226 * ipa-modref-tree.h (struct modref_parm_map): Add default
3228 * ipa-modref.c (ipa_merge_modref_summary_after_inlining): Use it.
3230 2021-11-19 Richard Biener <rguenther@suse.de>
3232 PR middle-end/103248
3233 * tree-eh.c (operation_could_trap_helper_p): Properly handle
3234 fixed-point RDIV_EXPR.
3236 2021-11-19 Richard Biener <rguenther@suse.de>
3238 PR tree-optimization/102436
3239 * tree-ssa-loop-im.c (execute_sm_if_changed): Add mode
3240 to just create the if structure and return the then block.
3241 (execute_sm): Add flag to indicate the var will re-use
3243 (hoist_memory_references): Support a single conditional
3244 block with all stores as special case.
3246 2021-11-19 Andrew Pinski <apinski@marvell.com>
3248 PR tree-optimization/103317
3249 * tree-ssa-phiopt.c (minmax_replacement): For the non empty
3250 middle bb case, check to make sure it has a single predecessor.
3252 2021-11-19 Andrew Pinski <apinski@marvell.com>
3254 PR tree-optimization/103257
3256 ((m1 >/</>=/<= m2) * d -> (m1 >/</>=/<= m2) ? d : 0):
3257 Disable until !canonicalize_math_p ().
3259 2021-11-19 Marek Polacek <polacek@redhat.com>
3263 * doc/invoke.texi: Update documentation for -Wuninitialized.
3264 * tree.c (stabilize_reference): Set location.
3266 2021-11-19 liuhongt <hongtao.liu@intel.com>
3269 * config/i386/x86-tune-costs.h (skylake_cost): Reduce cost of
3270 storing 256/512-bit SSE register to be equal to cost of
3271 unaligned store to avoid odd alignment peeling.
3272 (icelake_cost): Ditto.
3274 2021-11-18 David Edelsohn <dje.gcc@gmail.com>
3276 * config/rs6000/predicates.md (current_file_function_operand):
3277 Add flag_semantic_interposition to call of decl_replaceable_p.
3279 2021-11-18 Martin Liska <mliska@suse.cz>
3281 * ipa-modref.c (analyze_function): Do not execute the code
3282 only if dump_file != NULL.
3284 2021-11-18 Martin Liska <mliska@suse.cz>
3286 * ipa-modref.c (analyze_function): Use fnode instead of repeated
3287 cgraph_node::get (current_function_decl).
3289 2021-11-18 Jan Hubicka <hubicka@ucw.cz>
3291 * cgraph.c (cgraph_node::get_availability): Update call of
3293 (cgraph_node::verify_node): Verify that semantic_interposition flag
3295 * cgraph.h: (symtab_node): Add semantic_interposition flag.
3296 * cgraphclones.c (set_new_clone_decl_and_node_flags): Clear
3297 semantic_interposition flag.
3298 * cgraphunit.c (cgraph_node::finalize_function): Set
3299 semantic_interposition flag.
3300 (cgraph_node::add_new_function): Likewise.
3301 (varpool_node::finalize_decl): Likewise.
3302 (cgraph_node::create_wrapper): Likewise.
3303 * common.opt (fsemantic-interposition): Turn to optimization node.
3304 * lto-cgraph.c (lto_output_node): Stream semantic_interposition.
3305 (lto_output_varpool_node): Likewise.
3306 (input_overwrite_node): Likewise.
3307 (input_varpool_node): Likewise.
3308 * symtab.c (symtab_node::dump_base): Dump new flag.
3309 * varasm.c (decl_replaceable_p): Add semantic_interposition_p
3311 * varasm.h (decl_replaceable_p): Update declaration.
3312 * varpool.c (varpool_node::ctor_useable_for_folding_p):
3313 Use semantic_interposition flag.
3314 (varpool_node::get_availability): Likewise.
3315 (varpool_node::create_alias): Copy semantic_interposition flag.
3317 2021-11-18 Jan Hubicka <hubicka@ucw.cz>
3320 * ipa-modref.c (modref_eaf_analysis::merge_call_lhs_flags): Unused
3321 parameter may still be returned.
3322 (modref_eaf_analysis::analyze_ssa_name): Call merge_call_lhs_flags
3323 even for unused function args.
3325 2021-11-18 Tamar Christina <tamar.christina@arm.com>
3327 PR tree-optimization/103311
3328 * tree-vect-slp-patterns.c (complex_mul_pattern::matches,
3329 complex_fms_pattern::matches): Check for multiplications.
3331 2021-11-18 H.J. Lu <hjl.tools@gmail.com>
3334 * config/i386/i386.c (ix86_output_jmp_thunk_or_indirect): Emit
3335 CS prefix for -mindirect-branch-cs-prefix.
3336 (ix86_output_indirect_branch_via_reg): Likewise.
3337 * config/i386/i386.opt: Add -mindirect-branch-cs-prefix.
3338 * doc/invoke.texi: Document -mindirect-branch-cs-prefix.
3340 2021-11-18 David Edelsohn <dje.gcc@gmail.com>
3342 * config/rs6000/driver-rs6000.c (detect_processor_aix): Add
3345 2021-11-18 Tamar Christina <tamar.christina@arm.com>
3347 PR tree-optimization/103253
3348 * tree-ssa-math-opts.c (convert_mult_to_fma): Check for LHS.
3350 2021-11-18 Matthias Kretz <m.kretz@gsi.de>
3352 * doc/extend.texi: Document __builtin_assoc_barrier.
3354 2021-11-18 Martin Liska <mliska@suse.cz>
3356 * cfgexpand.c (pass_expand::execute): Use option directly.
3357 * function.c (allocate_struct_function): Likewise.
3358 * gimple-low.c (lower_function_body): Likewise.
3359 (lower_stmt): Likewise.
3360 * gimple-ssa-backprop.c (backprop::prepare_change): Likewise.
3361 * ipa-param-manipulation.c (ipa_param_adjustments::modify_call): Likewise.
3362 * ipa-split.c (split_function): Likewise.
3363 * lto-streamer-in.c (input_function): Likewise.
3364 * sese.c (sese_insert_phis_for_liveouts): Likewise.
3365 * ssa-iterators.h (num_imm_uses): Likewise.
3366 * tree-cfg.c (make_blocks): Likewise.
3367 (gimple_merge_blocks): Likewise.
3368 * tree-inline.c (tree_function_versioning): Likewise.
3369 * tree-loop-distribution.c (generate_loops_for_partition): Likewise.
3370 * tree-sra.c (analyze_access_subtree): Likewise.
3371 * tree-ssa-dce.c (remove_dead_stmt): Likewise.
3372 * tree-ssa-loop-ivopts.c (remove_unused_ivs): Likewise.
3373 * tree-ssa-phiopt.c (spaceship_replacement): Likewise.
3374 * tree-ssa-reassoc.c (reassoc_remove_stmt): Likewise.
3375 * tree-ssa-tail-merge.c (tail_merge_optimize): Likewise.
3376 * tree-ssa-threadedge.c (propagate_threaded_block_debug_into): Likewise.
3377 * tree-ssa.c (gimple_replace_ssa_lhs): Likewise.
3378 (target_for_debug_bind): Likewise.
3379 (insert_debug_temp_for_var_def): Likewise.
3380 (insert_debug_temps_for_defs): Likewise.
3381 (reset_debug_uses): Likewise.
3382 * tree-ssanames.c (release_ssa_name_fn): Likewise.
3383 * tree-vect-loop-manip.c (adjust_vec_debug_stmts): Likewise.
3384 (adjust_debug_stmts): Likewise.
3385 (adjust_phi_and_debug_stmts): Likewise.
3386 (vect_do_peeling): Likewise.
3387 * tree-vect-loop.c (vect_transform_loop_stmt): Likewise.
3388 (vect_transform_loop): Likewise.
3389 * tree.h (MAY_HAVE_DEBUG_MARKER_STMTS): Remove
3390 (MAY_HAVE_DEBUG_BIND_STMTS): Remove.
3391 (MAY_HAVE_DEBUG_STMTS): Use options directly.
3393 2021-11-18 Richard Biener <rguenther@suse.de>
3395 PR tree-optimization/103277
3396 * tree-ssa-dse.c (need_ab_cleanup): New.
3397 (dse_optimize_redundant_stores): Adjust.
3398 (delete_dead_or_redundant_assignment): Get extra
3399 need_ab_cleanup argument and set when abnormal cleanup is
3401 (dse_optimize_call): Adjust.
3402 (dse_optimize_stmt): Likewise.
3403 (pass_dse::execute): Allocate and deallocate need_ab_cleanup.
3404 Perform abnormal cleanup.
3405 * tree-ssa-dse.h (delete_dead_or_redundant_assignment): Adjust.
3407 2021-11-18 Hongyu Wang <hongyu.wang@intel.com>
3409 * config/i386/i386-expand.c (ix86_expand_atomic_fetch_op_loop):
3410 Adjust generated cfg to avoid infinite loop.
3412 2021-11-18 konglin1 <lingling.kong@intel.com>
3414 * config/i386/avx512fp16intrin.h (_mm512_mul_pch): Add alias for _mm512_fmul_pch.
3415 (_mm512_mask_mul_pch): Likewise.
3416 (_mm512_maskz_mul_pch): Likewise.
3417 (_mm512_mul_round_pch): Likewise.
3418 (_mm512_mask_mul_round_pch): Likewise.
3419 (_mm512_maskz_mul_round_pch): Likewise.
3420 (_mm512_cmul_pch): Likewise.
3421 (_mm512_mask_cmul_pch): Likewise.
3422 (_mm512_maskz_cmul_pch): Likewise.
3423 (_mm512_cmul_round_pch): Likewise.
3424 (_mm512_mask_cmul_round_pch): Likewise.
3425 (_mm512_maskz_cmul_round_pch): Likewise.
3426 (_mm_mul_sch): Likewise.
3427 (_mm_mask_mul_sch): Likewise.
3428 (_mm_maskz_mul_sch): Likewise.
3429 (_mm_mul_round_sch): Likewise.
3430 (_mm_mask_mul_round_sch): Likewise.
3431 (_mm_maskz_mul_round_sch): Likewise.
3432 (_mm_cmul_sch): Likewise.
3433 (_mm_mask_cmul_sch): Likewise.
3434 (_mm_maskz_cmul_sch): Likewise.
3435 (_mm_cmul_round_sch): Likewise.
3436 (_mm_mask_cmul_round_sch): Likewise.
3437 (_mm_maskz_cmul_round_sch): Likewise.
3438 * config/i386/avx512fp16vlintrin.h (_mm_mul_pch): Likewise.
3439 (_mm_mask_mul_pch): Likewise.
3440 (_mm_maskz_mul_pch): Likewise.
3441 (_mm256_mul_pch): Likewise.
3442 (_mm256_mask_mul_pch): Likewise.
3443 (_mm256_maskz_mul_pch): Likewise.
3444 (_mm_cmul_pch): Likewise.
3445 (_mm_mask_cmul_pch): Likewise.
3446 (_mm_maskz_cmul_pch): Likewise.
3447 (_mm256_cmul_pch): Likewise.
3448 (_mm256_mask_cmul_pch): Likewise.
3449 (_mm256_maskz_cmul_pch): Likewise.
3451 2021-11-17 Andrew Pinski <apinski@marvell.com>
3453 PR tree-optimization/103228
3454 PR tree-optimization/55177
3455 * match.pd ((type) X bitop CST): Also do this
3456 transformation for nop conversions.
3458 2021-11-17 Martin Sebor <msebor@redhat.com>
3460 PR tree-optimization/102759
3461 * gimple-array-bounds.cc (build_printable_array_type): Move...
3462 * gimple-ssa-warn-access.cc (build_printable_array_type): Avoid
3463 pathological function redeclarations that remove a previously
3465 Improve formatting of function arguments in informational notes.
3466 * pointer-query.cc (build_printable_array_type): ...to here.
3467 * pointer-query.h (build_printable_array_type): Declared.
3469 2021-11-17 H.J. Lu <hjl.tools@gmail.com>
3472 * config/i386/i386-opts.h (harden_sls): New enum.
3473 * config/i386/i386.c (output_indirect_thunk): Mitigate against
3474 SLS for function return.
3475 (ix86_output_function_return): Likewise.
3476 (ix86_output_jmp_thunk_or_indirect): Mitigate against indirect
3478 (ix86_output_indirect_jmp): Likewise.
3479 (ix86_output_call_insn): Likewise.
3480 * config/i386/i386.opt: Add -mharden-sls=.
3481 * doc/invoke.texi: Document -mharden-sls=.
3483 2021-11-17 H.J. Lu <hjl.tools@gmail.com>
3486 * config/i386/i386.c (ix86_code_end): Remove "%!" before ret.
3487 (ix86_output_function_return): Likewise.
3488 * config/i386/i386.md (simple_return_pop_internal): Likewise.
3490 2021-11-17 Jan Hubicka <hubicka@ucw.cz>
3493 * ipa-modref.c (read_modref_records): Fix streaminig in of every_access
3496 2021-11-17 Uroš Bizjak <ubizjak@gmail.com>
3498 * config/i386/i386.c (indirect_thunks_used): Redefine as HARD_REG_SET.
3499 (ix86_code_end): Use TEST_HARD_REG_BIT on indirect_thunks_used.
3500 (ix86_output_indirect_branch_via_reg): Use SET_HARD_REG_BIT
3501 on indirect_thunks_used.
3502 (ix86_output_indirect_function_return): Ditto.
3504 2021-11-17 Jan Hubicka <hubicka@ucw.cz>
3506 * ipa-modref-tree.c: Include cgraph.h and tree-streamer.h.
3507 (modref_access_node::stream_out): New member function.
3508 (modref_access_node::stream_in): New member function.
3509 * ipa-modref-tree.h (modref_access_node::stream_out,
3510 modref_access_node::stream_in): Declare.
3511 * ipa-modref.c (modref_summary_lto::useful_p): Free useless kills.
3512 (modref_summary_lto::dump): Dump kills.
3513 (analyze_store): Record kills for LTO
3514 (analyze_stmt): Likewise.
3515 (modref_summaries_lto::duplicate): Duplicate kills.
3516 (write_modref_records): Use new stream_out member function.
3517 (read_modref_records): Likewise.
3518 (modref_write): Stream out kills.
3519 (read_section): Stream in kills
3520 (remap_kills): New function.
3521 (update_signature): Use it.
3523 2021-11-17 Uroš Bizjak <ubizjak@gmail.com>
3525 * config/i386/i386.h (LEGACY_SSE_REGNO_P): New predicate.
3526 (SSE_REGNO_P): Use LEGACY_SSE_REGNO_P predicate.
3527 * config/i386/i386.c (zero_all_vector_registers):
3528 Use LEGACY_SSE_REGNO_P predicate.
3529 (ix86_register_priority): Use REX_INT_REGNO_P, REX_SSE_REGNO_P
3530 and EXT_REG_SSE_REGNO_P predicates.
3531 (ix86_hard_regno_call_part_clobbered): Use REX_SSE_REGNO_P
3532 and LEGACY_SSE_REGNO_P predicates.
3534 2021-11-17 Jason Merrill <jason@redhat.com>
3536 * doc/invoke.texi (C++ Dialect Options): Document
3537 -fimplicit-constexpr.
3539 2021-11-17 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
3541 * config/aarch64/aarch64-modes.def (VECTOR_MODE): New V8DI mode.
3542 * config/aarch64/aarch64.c (aarch64_hard_regno_mode_ok): Handle
3544 * config/aarch64/iterators.md (define_mode_attr nunits): Add entry
3547 2021-11-17 Martin Uecker <uecker@gcc.gnu.org>
3551 * gimplify.c (gimplify_var_or_parm_decl): Update comment.
3552 (gimplify_compound_lval): Gimplify base expression first.
3553 (gimplify_target_expr): Add comment.
3555 2021-11-17 Jakub Jelinek <jakub@redhat.com>
3557 PR tree-optimization/103192
3558 * tree-ssa-loop-im.c (move_computations_worker): Use
3559 reset_flow_sensitive_info instead of manually clearing
3560 SSA_NAME_RANGE_INFO and do it for all SSA_NAMEs, not just ones
3561 with integral types.
3563 2021-11-17 Jakub Jelinek <jakub@redhat.com>
3565 PR tree-optimization/103255
3566 * gimple-range-fold.cc (fold_using_range::range_of_address): Return
3567 range_nonzero rather than unadjusted base's range. Formatting fixes.
3569 2021-11-17 Richard Sandiford <richard.sandiford@arm.com>
3571 * doc/md.texi (cond_fmin@var{mode}, cond_fmax@var{mode}): Document.
3572 * optabs.def (cond_fmin_optab, cond_fmax_optab): New optabs.
3573 * internal-fn.def (COND_FMIN, COND_FMAX): New functions.
3574 * internal-fn.c (first_commutative_argument): Handle them.
3575 (FOR_EACH_COND_FN_PAIR): Likewise.
3576 * match.pd (UNCOND_BINARY, COND_BINARY): Likewise.
3577 * config/aarch64/aarch64-sve.md (cond_<fmaxmin><mode>): New
3580 2021-11-17 Kewen Lin <linkw@linux.ibm.com>
3582 * config/i386/i386.md (*add<dwi>3_doubleword, *addv<dwi>4_doubleword,
3583 *addv<dwi>4_doubleword_1, *sub<dwi>3_doubleword,
3584 *subv<dwi>4_doubleword, *subv<dwi>4_doubleword_1,
3585 *add<dwi>3_doubleword_cc_overflow_1, *divmodsi4_const,
3586 *neg<dwi>2_doubleword, *tls_dynamic_gnu2_combine_64_<mode>): Fix split
3589 2021-11-17 Andrew Pinski <apinski@marvell.com>
3591 PR tree-optimization/103288
3592 * tree-ssa-phiopt.c (value_replacement): Return early if middle
3593 block has more than one pred.
3595 2021-11-17 Kewen Lin <linkw@linux.ibm.com>
3597 * config/visium/visium.md (*add<mode>3_insn, *addsi3_insn, *addi3_insn,
3598 *sub<mode>3_insn, *subsi3_insn, *subdi3_insn, *neg<mode>2_insn,
3599 *negdi2_insn, *and<mode>3_insn, *ior<mode>3_insn, *xor<mode>3_insn,
3600 *one_cmpl<mode>2_insn, *ashl<mode>3_insn, *ashr<mode>3_insn,
3601 *lshr<mode>3_insn, *trunchiqi2_insn, *truncsihi2_insn,
3602 *truncdisi2_insn, *extendqihi2_insn, *extendqisi2_insn,
3603 *extendhisi2_insn, *extendsidi2_insn, *zero_extendqihi2_insn,
3604 *zero_extendqisi2_insn, *zero_extendsidi2_insn): Fix split condition.
3606 2021-11-17 Marek Polacek <polacek@redhat.com>
3608 PR preprocessor/103026
3609 * doc/invoke.texi: Document -Wbidi-chars.
3611 2021-11-17 Jan Hubicka <hubicka@ucw.cz>
3614 * ipa-modref.c (ipa_merge_modref_summary_after_inlining): Fix clearing
3617 2021-11-16 Jan Hubicka <jh@suse.cz>
3619 * ipa-modref.c (get_modref_function_summary): Declare.
3620 * ipa-modref.h (get_modref_function_summary): New function.
3621 * tree-ssa-dse.c (clear_live_bytes_for_ref): Break out from ...
3622 (clear_bytes_written_by): ... here; also clear memory killed by
3625 2021-11-16 Iain Sandoe <iain@sandoe.co.uk>
3627 * ggc-common.c (gt_pch_save): If we cannot find a suitable
3628 memory segment for save, then error-out, do not try to
3630 (gt_pch_restore): Save the existing line table, and when
3631 the replacement is being read, use that when constructing
3634 2021-11-16 Peter Bergner <bergner@linux.ibm.com>
3637 * config/rs6000/mma.md (*vsx_assemble_pair): Add early-clobber for
3639 (*mma_assemble_acc): Likewise.
3641 2021-11-16 Martin Sebor <msebor@redhat.com>
3643 PR tree-optimization/102960
3644 * gimple-fold.c (get_range_strlen): Take bitmap as an argument rather
3645 than a pointer to it.
3646 (get_range_strlen_tree): Same. Remove bitmap allocation. Use
3648 (get_maxval_strlen): Use an auto_bitmap.
3649 * tree-ssa-strlen.c (get_range_strlen_dynamic): Factor out PHI
3651 (get_range_strlen_phi): ...into this function.
3652 Avoid assuming maximum string length is constant
3653 (printf_strlen_execute): Dump pointer query cache contents when
3654 details are requisted.
3656 2021-11-16 Jason Merrill <jason@redhat.com>
3658 * langhooks.h (struct lang_hooks): Adjust comment.
3659 * print-tree.c (print_node): Also call print_xnode hook for
3662 2021-11-16 Andrew Pinski <apinski@marvell.com>
3664 PR tree-optimization/103218
3665 * match.pd: New pattern for "((type)(a<0)) << SIGNBITOFA".
3667 2021-11-16 Claudiu Zissulescu <claziss@synopsys.com>
3669 * config/arc/arc.md (maddhisi4): Use a single move to accumulator.
3670 (umaddhisi4): Likewise.
3671 (machi): Update pattern.
3674 2021-11-16 Richard Biener <rguenther@suse.de>
3676 PR tree-optimization/102880
3677 * tree-ssa-dce.c (sort_phi_args): New function.
3678 (make_forwarders_with_degenerate_phis): Likewise.
3679 (perform_tree_ssa_dce): Call
3680 make_forwarders_with_degenerate_phis.
3682 2021-11-16 Richard Biener <rguenther@suse.de>
3684 PR tree-optimization/102880
3685 * tree-ssa-phiopt.c (tree_ssa_phiopt_worker): Push
3686 single_pred (bb1) condition to places that really need it.
3687 (match_simplify_replacement): Likewise.
3688 (value_replacement): Likewise.
3689 (replace_phi_edge_with_variable): Deal with extra edges
3692 2021-11-16 Martin Jambor <mjambor@suse.cz>
3694 * cfgexpand.c (expand_gimple_basic_block): Use build_debug_expr_decl,
3695 add a fixme note about the mode assignment perhaps being unnecessary.
3696 * ipa-param-manipulation.c (ipa_param_adjustments::modify_call):
3698 (ipa_param_body_adjustments::mark_dead_statements): Likewise.
3699 (ipa_param_body_adjustments::reset_debug_stmts): Likewise.
3700 * tree-inline.c (remap_ssa_name): Likewise.
3701 (tree_function_versioning): Likewise.
3702 * tree-into-ssa.c (rewrite_debug_stmt_uses): Likewise.
3703 * tree-ssa-loop-ivopts.c (remove_unused_ivs): Likewise.
3704 * tree-ssa.c (insert_debug_temp_for_var_def): Likewise.
3706 2021-11-16 Jakub Jelinek <jakub@redhat.com>
3708 PR tree-optimization/103208
3709 * omp-expand.c (expand_omp_build_cond): New function.
3710 (expand_omp_for_init_counts, expand_omp_for_init_vars,
3711 expand_omp_for_static_nochunk, expand_omp_for_static_chunk): Use it.
3713 2021-11-16 Jakub Jelinek <jakub@redhat.com>
3715 PR tree-optimization/102009
3716 * gimple-ssa-warn-access.cc (pass_waccess::check_alloc_size_call):
3717 Punt if any of alloc_size arguments is out of bounds vs. number of
3720 2021-11-16 Roger Sayle <roger@nextmovesoftware.com>
3722 * config/i386/i386.md (*bmi2_rorx<mode3>_1): Make conditional
3723 on !optimize_function_for_size_p.
3724 (*<any_rotate><mode>3_1): Add preferred_for_size attribute.
3725 (define_splits): Conditionalize on !optimize_function_for_size_p.
3726 (*bmi2_rorxsi3_1_zext): Likewise.
3727 (*<any_rotate>si2_1_zext): Add preferred_for_size attribute.
3728 (define_splits): Conditionalize on !optimize_function_for_size_p.
3730 2021-11-16 Jan Hubicka <jh@suse.cz>
3733 * ipa-modref.c (merge_call_side_effects): Fix uninitialized
3736 2021-11-16 Andrew Pinski <apinski@marvell.com>
3738 PR tree-optimization/103245
3739 * match.pd: Combine the abs pattern matching using multiplication.
3740 Adding optional nop_convert too.
3742 2021-11-16 H.J. Lu <hjl.tools@gmail.com>
3744 PR middle-end/103268
3745 * tree-ssa-ccp.c (optimize_atomic_bit_test_and): Add a missing
3748 2021-11-15 Siddhesh Poyarekar <siddhesh@gotplt.org>
3750 * gimple-fold.c (gimple_fold_builtin_strncat): Use ranges to
3751 determine if it is safe to transform to strcat.
3752 (gimple_fold_builtin_snprintf): Likewise.
3754 2021-11-15 Siddhesh Poyarekar <siddhesh@gotplt.org>
3756 * gimple-fold.c (known_lower): New function.
3757 (gimple_fold_builtin_strncat_chk,
3758 gimple_fold_builtin_memory_chk, gimple_fold_builtin_stxcpy_chk,
3759 gimple_fold_builtin_stxncpy_chk,
3760 gimple_fold_builtin_snprintf_chk,
3761 gimple_fold_builtin_sprintf_chk): Use it.
3763 2021-11-15 Siddhesh Poyarekar <siddhesh@gotplt.org>
3765 * gimple-fold.c (dump_transformation): New function.
3766 (gimple_fold_builtin_stxcpy_chk,
3767 gimple_fold_builtin_stxncpy_chk): Use it. Simplify to
3768 BUILT_IN_STRNCPY if return value is not used.
3770 2021-11-15 H.J. Lu <hjl.tools@gmail.com>
3772 PR middle-end/103184
3773 * tree-ssa-ccp.c (optimize_atomic_bit_test_and): Check optab
3774 before transforming equivalent, but slighly different cases to
3775 their canonical forms.
3777 2021-11-15 Iain Sandoe <iain@sandoe.co.uk>
3780 * config/darwin.h (TARGET_DTORS_FROM_CXA_ATEXIT): New.
3781 * doc/tm.texi: Regenerated.
3782 * doc/tm.texi.in: Add TARGET_DTORS_FROM_CXA_ATEXIT hook.
3783 * ipa.c (cgraph_build_static_cdtor_1): Return the built
3785 (build_cxa_atexit_decl): New.
3786 (build_dso_handle_decl): New.
3787 (build_cxa_dtor_registrations): New.
3788 (compare_cdtor_tu_order): New.
3789 (build_cxa_atexit_fns): New.
3790 (ipa_cdtor_merge): If dtors_from_cxa_atexit is set,
3791 process the DTORs/CTORs accordingly.
3792 (pass_ipa_cdtor_merge::gate): Also run if
3793 dtors_from_cxa_atexit is set.
3794 * target.def (dtors_from_cxa_atexit): New hook.
3796 2021-11-15 Iain Sandoe <iain@sandoe.co.uk>
3798 * config.in: Regenerate.
3799 * configure: Regenerate.
3800 * configure.ac: Test ld64 for -platform-version support.
3802 2021-11-15 Thomas Schwinge <thomas@codesourcery.com>
3804 * diagnostic-spec.h (typedef xint_hash_t)
3805 (typedef xint_hash_map_t): Replace with...
3806 (typedef nowarn_map_t): ... this.
3807 (nowarn_map): Adjust.
3808 * diagnostic-spec.c (nowarn_map, suppress_warning_at): Likewise.
3810 2021-11-15 Thomas Schwinge <thomas@codesourcery.com>
3812 * profile.c (branch_prob): Use 'location_hash' for
3815 2021-11-15 Aldy Hernandez <aldyh@redhat.com>
3817 PR tree-optimization/103207
3818 * value-range.cc (irange::set): Drop overflow.
3820 2021-11-15 Jakub Jelinek <jakub@redhat.com>
3822 * gimplify.c (optimize_target_teams): Only add OMP_CLAUSE_THREAD_LIMIT
3823 to OMP_TARGET_CLAUSES if it isn't there already.
3825 2021-11-15 Aldy Hernandez <aldyh@redhat.com>
3827 * gimple-range-path.cc (path_range_query::internal_range_of_expr):
3828 Remove useless code.
3829 (path_range_query::ssa_defined_in_bb): New.
3830 (path_range_query::ssa_range_in_phi): Avoid fold_range call that
3831 could trigger additional lookups.
3832 Do not use the cache for ARGs defined in this block.
3833 (path_range_query::compute_ranges_in_block): Use ssa_defined_in_bb.
3834 (path_range_query::maybe_register_phi_relation): Same.
3835 (path_range_query::range_of_stmt): Adjust comment.
3836 * gimple-range-path.h (ssa_defined_in_bb): New.
3838 2021-11-15 Aldy Hernandez <aldyh@redhat.com>
3840 * gimple-range-path.cc (path_range_query::internal_range_of_expr):
3841 Default to global range if nothing found.
3843 2021-11-15 Richard Biener <rguenther@suse.de>
3845 PR tree-optimization/103237
3846 * tree-vect-loop.c (vect_is_simple_reduction): Fail for
3847 double reductions with multiple inner loop LC PHI nodes.
3849 2021-11-15 Hongyu Wang <hongyu.wang@intel.com>
3852 * config/i386/i386-expand.c (ix86_expand_atomic_fetch_op_loop):
3853 New expand function.
3854 * config/i386/i386-options.c (ix86_target_string): Add
3855 -mrelax-cmpxchg-loop flag.
3856 (ix86_valid_target_attribute_inner_p): Likewise.
3857 * config/i386/i386-protos.h (ix86_expand_atomic_fetch_op_loop):
3858 New expand function prototype.
3859 * config/i386/i386.opt: Add -mrelax-cmpxchg-loop.
3860 * config/i386/sync.md (atomic_fetch_<logic><mode>): New expander
3862 (atomic_<logic>_fetch<mode>): Likewise.
3863 (atomic_fetch_nand<mode>): Likewise.
3864 (atomic_nand_fetch<mode>): Likewise.
3865 (atomic_fetch_<logic><mode>): New expander for DI,TI modes.
3866 (atomic_<logic>_fetch<mode>): Likewise.
3867 (atomic_fetch_nand<mode>): Likewise.
3868 (atomic_nand_fetch<mode>): Likewise.
3869 * doc/invoke.texi: Document -mrelax-cmpxchg-loop.
3871 2021-11-15 Richard Biener <rguenther@suse.de>
3873 PR tree-optimization/103219
3874 * gimple-loop-jam.c (tree_loop_unroll_and_jam): Use single_exit
3875 to determine the exit for the VN region.
3877 2021-11-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
3879 * tree-ssa-loop.c (pass_vectorize): Move to tree-vectorizer.c.
3880 (pass_data_vectorize): Likewise.
3881 (make_pass_vectorize): Likewise.
3882 * tree-vectorizer.c (vectorize_loops): Merge with
3883 pass_vectorize::execute and replace cfun occurences with fun param.
3884 (adjust_simduid_builtins): Add fun param, replace cfun occurences with
3885 fun, and adjust callers approrpiately.
3886 (note_simd_array_uses): Likewise.
3887 (vect_loop_dist_alias_call): Likewise.
3888 (set_uid_loop_bbs): Likewise.
3889 (vect_transform_loops): Likewise.
3890 (try_vectorize_loop_1): Likewise.
3891 (try_vectorize_loop): Likewise.
3893 2021-11-15 Jakub Jelinek <jakub@redhat.com>
3896 * config/i386/sync.md (atomic_bit_test_and_set<mode>,
3897 atomic_bit_test_and_complement<mode>,
3898 atomic_bit_test_and_reset<mode>): Use OPTAB_WIDEN instead of
3901 2021-11-15 Maciej W. Rozycki <macro@embecosm.com>
3903 * config/vax/vax.c (vax_lra_p): New prototype and function.
3904 (TARGET_LRA_P): Wire it.
3905 * config/vax/vax.opt (mlra): New option.
3906 * doc/invoke.texi (Option Summary, VAX Options): Document the
3909 2021-11-14 Jan Hubicka <jh@suse.cz>
3911 * ipa-modref.h (struct modref_summary): Add nondeterministic
3912 and calls_interposable flags.
3913 * ipa-modref.c (modref_summary::modref_summary): Initialize new flags.
3914 (modref_summary::useful_p): Check new flags.
3915 (struct modref_summary_lto): Add nondeterministic and
3916 calls_interposable flags.
3917 (modref_summary_lto::modref_summary_lto): Initialize new flags.
3918 (modref_summary_lto::useful_p): Check new flags.
3919 (modref_summary::dump): Dump new flags.
3920 (modref_summary_lto::dump): Dump new flags.
3921 (ignore_nondeterminism_p): New function.
3922 (merge_call_side_effects): Merge new flags.
3923 (process_fnspec): Likewise.
3924 (analyze_load): Volatile access is nondeterministic.
3925 (analyze_store): Liekwise.
3926 (analyze_stmt): Volatile ASM is nondeterministic.
3927 (analyze_function): Clear new flags.
3928 (modref_summaries::duplicate): Duplicate new flags.
3929 (modref_summaries_lto::duplicate): Duplicate new flags.
3930 (modref_write): Stream new flags.
3931 (read_section): Stream new flags.
3932 (propagate_unknown_call): Update new flags.
3933 (modref_propagate_in_scc): Propagate new flags.
3934 * tree-ssa-alias.c (ref_maybe_used_by_call_p_1): Check
3936 * tree-ssa-structalias.c (determine_global_memory_access):
3939 2021-11-14 Maciej W. Rozycki <macro@embecosm.com>
3941 * config/vax/vax.h (SET_RATIO): New macro.
3942 * config/vax/vax.md (UNSPEC_SETMEM_FILL): New constant.
3943 (setmemhi): New expander.
3944 (setmemhi1): New insn and splitter.
3945 (*setmemhi1): New insn.
3947 2021-11-14 Jan Hubicka <hubicka@ucw.cz>
3949 * ipa-modref-tree.c (modref_access_node::update_for_kills): New
3951 (modref_access_node::merge_for_kills): Likewise.
3952 (modref_access_node::insert_kill): Likewise.
3953 * ipa-modref-tree.h (modref_access_node::update_for_kills,
3954 modref_access_node::merge_for_kills, modref_access_node::insert_kill):
3956 (modref_access_node::useful_for_kill): New member function.
3957 * ipa-modref.c (modref_summary::useful_p): Release useless kills.
3958 (lto_modref_summary): Add kills.
3959 (modref_summary::dump): Dump kills.
3960 (record_access): Add mdoref_access_node parameter.
3961 (record_access_lto): Likewise.
3962 (merge_call_side_effects): Merge kills.
3963 (analyze_call): Add ALWAYS_EXECUTED param and pass it around.
3964 (struct summary_ptrs): Add always_executed filed.
3965 (analyze_load): Update.
3966 (analyze_store): Update; record kills.
3967 (analyze_stmt): Add always_executed; record kills in clobbers.
3968 (analyze_function): Track always_executed.
3969 (modref_summaries::duplicate): Duplicate kills.
3970 (update_signature): Release kills.
3971 * ipa-modref.h (struct modref_summary): Add kills.
3972 * tree-ssa-alias.c (alias_stats): Add kill stats.
3973 (dump_alias_stats): Dump kill stats.
3974 (store_kills_ref_p): Break out from ...
3975 (stmt_kills_ref_p): Use it; handle modref info based kills.
3977 2021-11-14 Aldy Hernandez <aldyh@redhat.com>
3979 PR tree-optimization/103229
3980 * gimple-range-cache.cc (ssa_global_cache::clear): Do not pass
3981 null value to memset.
3983 2021-11-14 Jan Hubicka <hubicka@ucw.cz>
3985 * ipa-modref-tree.c (modref_access_node::get_call_arg): New member
3987 (modref_access_node::get_ao_ref): Likewise.
3988 * ipa-modref-tree.h (modref_access_node::get_call_arg): Declare.
3989 (modref_access_node::get_ao_ref): Declare.
3990 * tree-ssa-alias.c (modref_may_conflict): Use new accessors.
3991 * tree-ssa-dse.c (dse_optimize_call): Use new accessors.
3993 2021-11-13 Jan Hubicka <jh@suse.cz>
3996 * dbgcnt.def (ipa_attr): New counters.
3997 * ipa-pure-const.c: Include dbgcnt.c
3998 (ipa_make_function_const): Use debug counter.
3999 (ipa_make_function_pure): Likewise.
4000 (propagate_pure_const): Fix bug in my previous change.
4002 2021-11-13 Jan Hubicka <hubicka@ucw.cz>
4004 * ipa-modref-tree.c (modref_access_node::range_info_useful_p):
4005 Offline from ipa-modref-tree.h.
4006 (modref_access_node::dump): Move from ipa-modref.c; make member
4008 * ipa-modref-tree.h (modref_access_node::range_info_useful_p.
4009 modref_access_node::dump): Declare.
4010 * ipa-modref.c (dump_access): Remove.
4011 (dump_records): Update.
4012 (dump_lto_records): Update.
4013 (record_access): Update.
4014 (record_access_lto): Update.
4016 2021-11-13 Jan Hubicka <hubicka@ucw.cz>
4018 * ipa-modref.c (modref_summary::modref_summary): Clear new flags.
4019 (modref_summary::dump): Dump try_dse.
4020 (modref_summary::finalize): Add FUN attribute; compute try-dse.
4021 (analyze_function): Update.
4022 (read_section): Update.
4023 (update_signature): Update.
4024 (pass_ipa_modref::execute): Update.
4025 * ipa-modref.h (struct modref_summary):
4026 * tree-ssa-alias.c (ao_ref_init_from_ptr_and_range): Export.
4027 * tree-ssa-alias.h (ao_ref_init_from_ptr_and_range): Declare.
4028 * tree-ssa-dse.c (dse_optimize_call): New function.
4029 (dse_optimize_stmt): Use it.
4031 2021-11-13 Jan Hubicka <hubicka@ucw.cz>
4033 * ipa-modref-tree.c: Move #if CHECKING_P to proper place.
4035 2021-11-13 Jan Hubicka <jh@suse.cz>
4038 (struct modref_access_node): Move longer member functions to
4040 (modref_ref_node::try_merge_with): Turn into modreef_acces_node member
4042 * ipa-modref-tree.c (modref_access_node::contains): Move here
4043 from ipa-modref-tree.h.
4044 (modref_access_node::update): Likewise.
4045 (modref_access_node::merge): Likewise.
4046 (modref_access_node::closer_pair_p): Likewise.
4047 (modref_access_node::forced_merge): Likewise.
4048 (modref_access_node::update2): Likewise.
4049 (modref_access_node::combined_offsets): Likewise.
4050 (modref_access_node::try_merge_with): Likewise.
4051 (modref_access_node::insert): Likewise.
4053 2021-11-13 Jan Hubicka <jh@suse.cz>
4055 * ipa-modref.c (modref_summary::global_memory_read_p): Remove.
4056 (modref_summary::global_memory_written_p): Remove.
4057 (modref_summary::dump): Dump new flags.
4058 (modref_summary::finalize): New member function.
4059 (analyze_function): Call it.
4060 (read_section): Call it.
4061 (update_signature): Call it.
4062 (pass_ipa_modref::execute): Call it.
4063 * ipa-modref.h (struct modref_summary): Remove
4064 global_memory_read_p and global_memory_written_p.
4065 Add global_memory_read, global_memory_written.
4066 * tree-ssa-structalias.c (determine_global_memory_access):
4069 2021-11-13 Jan Hubicka <jh@suse.cz>
4071 * ipa-fnsummary.c (compute_fn_summary): Use type_attribut_allowed_p
4072 * ipa-param-manipulation.c
4073 (ipa_param_adjustments::type_attribute_allowed_p):
4074 New member function.
4075 (drop_type_attribute_if_params_changed_p): New function.
4076 (build_adjusted_function_type): Use it.
4077 * ipa-param-manipulation.h: Add type_attribute_allowed_p.
4079 2021-11-13 David Malcolm <dmalcolm@redhat.com>
4081 * doc/invoke.texi (Static Analyzer Options): Add
4082 -Wno-analyzer-tainted-allocation-size,
4083 -Wno-analyzer-tainted-divisor, -Wno-analyzer-tainted-offset, and
4084 -Wno-analyzer-tainted-size to list. Add
4085 -Wanalyzer-tainted-allocation-size, -Wanalyzer-tainted-divisor,
4086 -Wanalyzer-tainted-offset, and -Wanalyzer-tainted-size to list
4087 of options effectively enabled by -fanalyzer.
4088 (-Wanalyzer-tainted-allocation-size): New.
4089 (-Wanalyzer-tainted-array-index): Tweak wording; add link to CWE.
4090 (-Wanalyzer-tainted-divisor): New.
4091 (-Wanalyzer-tainted-offset): New.
4092 (-Wanalyzer-tainted-size): New.
4094 2021-11-13 Jan Hubicka <jh@suse.cz>
4096 * attr-fnspec.h (attr_fnspec::arg_eaf_flags): Break out from ...
4097 * gimple.c (gimple_call_arg_flags): ... here.
4098 * ipa-modref.c (analyze_parms): Record flags known from fnspec.
4099 (modref_merge_call_site_flags): Use arg_eaf_flags.
4101 2021-11-13 Aldy Hernandez <aldyh@redhat.com>
4103 PR tree-optimization/103222
4104 * gimple-range-path.cc (path_range_query::compute_ranges_in_phis):
4106 (path_range_query::compute_ranges_in_block): Call
4107 compute_ranges_in_phis.
4108 * gimple-range-path.h (path_range_query::compute_ranges_in_phis):
4111 2021-11-13 Jan Hubicka <jh@suse.cz>
4113 * ipa-fnsummary.c (compute_fn_summary): Do not give up on signature
4114 changes on "fn spec" attribute; give up on varadic types.
4115 * ipa-param-manipulation.c: Include attribs.h.
4116 (build_adjusted_function_type): New parameter ARG_MODIFIED; if it is
4117 true remove "fn spec" attribute.
4118 (ipa_param_adjustments::build_new_function_type): Update.
4119 (ipa_param_body_adjustments::modify_formal_parameters): update.
4120 * ipa-sra.c: Include attribs.h.
4121 (ipa_sra_preliminary_function_checks): Do not check for TYPE_ATTRIBUTES.
4123 2021-11-13 Aldy Hernandez <aldyh@redhat.com>
4125 * gimple-range-path.cc (path_range_query::path_range_query): Merge
4127 (path_range_query::import_p): Move from header file.
4128 (path_range_query::~path_range_query): Adjust for combined ctors.
4129 * gimple-range-path.h: Merge ctors.
4130 (path_range_query::import_p): Move to .cc file.
4132 2021-11-13 Jan Hubicka <jh@suse.cz>
4134 PR tree-optimization/103182
4135 * ipa-modref.c (callee_to_caller_flags): Fix merging of flags.
4136 (modref_eaf_analysis::analyze_ssa_name): Fix merging of flags.
4138 2021-11-12 Stafford Horne <shorne@gmail.com>
4140 * config/or1k/or1k-protos.h (or1k_profile_hook): New function.
4141 * config/or1k/or1k.h (PROFILE_HOOK): Change macro to reference
4142 new function or1k_profile_hook.
4143 * config/or1k/or1k.c (struct machine_function): Add new field
4144 set_mcount_arg_insn.
4145 (or1k_profile_hook): New function.
4146 (or1k_init_pic_reg): Update to inject pic rtx after _mcount arg
4148 (or1k_frame_pointer_required): Frame pointer no longer needed
4151 2021-11-12 Jan Hubicka <jh@suse.cz>
4153 PR tree-optimization/103209
4154 * tree-ssa-structalias.c (find_func_aliases_for_call): Fix
4155 use of handle_rhs_call
4157 2021-11-12 Aldy Hernandez <aldyh@redhat.com>
4159 PR tree-optimization/103202
4160 * gimple-range-path.cc
4161 (path_range_query::compute_ranges_in_block): Solve PHI imports first.
4163 2021-11-12 Jan Hubicka <jh@suse.cz>
4165 * ipa-pure-const.c (propagate_pure_const): Remove redundant check;
4166 fix call of ipa_make_function_const and ipa_make_function_pure.
4168 2021-11-12 Richard Sandiford <richard.sandiford@arm.com>
4170 * config/aarch64/aarch64.c (aarch64_vector_op::n_advsimd_ops): Delete.
4171 (aarch64_vector_op::m_seen_loads): Likewise.
4172 (aarch64_vector_costs::aarch64_vector_costs): Don't push to
4174 (aarch64_vector_op::count_ops): Remove vectype and factor parameters.
4175 Remove code that tries to predict different vec_flags from the
4177 (aarch64_vector_costs::add_stmt_cost): Update accordingly.
4178 Remove m_advsimd_ops handling.
4180 2021-11-12 Richard Sandiford <richard.sandiford@arm.com>
4182 * config/aarch64/aarch64.c (aarch64_vector_costs::m_saw_sve_only_op)
4183 (aarch64_sve_only_stmt_p): Delete.
4184 (aarch64_vector_costs::prefer_unrolled_loop): New function,
4185 extracted from adjust_body_cost.
4186 (aarch64_vector_costs::better_main_loop_than_p): New function,
4187 using heuristics extracted from adjust_body_cost and
4188 adjust_body_cost_sve.
4189 (aarch64_vector_costs::adjust_body_cost_sve): Remove
4190 advsimd_cycles_per_iter and could_use_advsimd parameters.
4191 Update after changes above.
4192 (aarch64_vector_costs::adjust_body_cost): Update after changes above.
4194 2021-11-12 Richard Sandiford <richard.sandiford@arm.com>
4196 * config/aarch64/aarch64.c (aarch64_vec_op_count::m_vf_factor):
4197 New member variable.
4198 (aarch64_vec_op_count::aarch64_vec_op_count): Add a parameter for it.
4199 (aarch64_vec_op_count::vf_factor): New function.
4200 (aarch64_vector_costs::aarch64_vector_costs): When costing for
4201 neoverse-512tvb, pass a vf_factor of 2 for the Neoverse V1 version
4203 (aarch64_vector_costs::adjust_body_cost): Read the vf factor
4204 instead of hard-coding 2.
4206 2021-11-12 Richard Sandiford <richard.sandiford@arm.com>
4208 * config/aarch64/aarch64.c
4209 (aarch64_vec_op_count::rename_cycles_per_iter): New function.
4210 (aarch64_vec_op_count::min_nonpred_cycles_per_iter): Likewise.
4211 (aarch64_vec_op_count::min_pred_cycles_per_iter): Likewise.
4212 (aarch64_vec_op_count::min_cycles_per_iter): Likewise.
4213 (aarch64_vec_op_count::dump): Move earlier in file. Dump the
4214 above properties too.
4215 (aarch64_estimate_min_cycles_per_iter): Delete.
4216 (adjust_body_cost): Use aarch64_vec_op_count::min_cycles_per_iter
4217 instead of aarch64_estimate_min_cycles_per_iter. Rely on the dump
4218 routine to print CPI estimates.
4219 (adjust_body_cost_sve): Likewise. Use the other functions above
4220 instead of doing the work inline.
4222 2021-11-12 Richard Sandiford <richard.sandiford@arm.com>
4224 * config/aarch64/aarch64.c (aarch64_vec_op_count): Allow default
4226 (aarch64_vec_op_count::base_issue_info): Remove handling of null
4228 (aarch64_vec_op_count::simd_issue_info): Likewise.
4229 (aarch64_vec_op_count::sve_issue_info): Likewise.
4230 (aarch64_vector_costs::m_ops): Turn into a vector.
4231 (aarch64_vector_costs::m_advsimd_ops): Likewise.
4232 (aarch64_vector_costs::aarch64_vector_costs): Add entries to
4233 the vectors based on aarch64_tune_params.
4234 (aarch64_vector_costs::analyze_loop_vinfo): Update the pred_ops
4235 of all entries in m_ops.
4236 (aarch64_vector_costs::add_stmt_cost): Call count_ops for all
4238 (aarch64_estimate_min_cycles_per_iter): Remove issue_info
4239 parameter and get the information from the ops instead.
4240 (aarch64_vector_costs::adjust_body_cost_sve): Take a
4241 aarch64_vec_issue_info instead of a aarch64_vec_op_count.
4242 (aarch64_vector_costs::adjust_body_cost): Update call accordingly.
4243 Exit earlier if m_ops is empty for either cost structure.
4245 2021-11-12 Richard Sandiford <richard.sandiford@arm.com>
4247 * config/aarch64/aarch64.c (aarch64_vector_costs::m_scalar_ops)
4248 (aarch64_vector_costs::m_sve_ops): Replace with...
4249 (aarch64_vector_costs::m_ops): ...this.
4250 (aarch64_vector_costs::analyze_loop_vinfo): Update accordingly.
4251 (aarch64_vector_costs::adjust_body_cost_sve): Likewise.
4252 (aarch64_vector_costs::aarch64_vector_costs): Likewise.
4253 Initialize m_vec_flags here rather than in add_stmt_cost.
4254 (aarch64_vector_costs::count_ops): Test for scalar reductions too.
4255 Allow vectype to be null.
4256 (aarch64_vector_costs::add_stmt_cost): Call count_ops for scalar
4257 code too. Don't require vectype to be nonnull.
4258 (aarch64_vector_costs::adjust_body_cost): Take the loop_vec_info
4259 and scalar costs as parameters. Use the scalar costs to determine
4260 the cycles per iteration of the scalar loop, then multiply it
4261 by the estimated VF.
4262 (aarch64_vector_costs::finish_cost): Update call accordingly.
4264 2021-11-12 Richard Sandiford <richard.sandiford@arm.com>
4266 * config/aarch64/aarch64.c (aarch64_dr_type): New function.
4267 (aarch64_vector_costs::count_ops): Use it rather than the
4268 vectype to determine floatness.
4270 2021-11-12 Richard Sandiford <richard.sandiford@arm.com>
4272 * config/aarch64/aarch64.c (aarch64_sve_in_loop_reduction_latency):
4273 Remove vectype parameter and get floatness from the type of the
4275 (arch64_in_loop_reduction_latency): Likewise.
4276 (aarch64_detect_vector_stmt_subtype): Update caller.
4277 (aarch64_vector_costs::count_ops): Likewise.
4279 2021-11-12 Richard Sandiford <richard.sandiford@arm.com>
4281 * config/aarch64/aarch64.c (aarch64_sve_op_count): Fold into...
4282 (aarch64_vec_op_count): ...this. Add a constructor.
4283 (aarch64_vec_op_count::vec_flags): New function.
4284 (aarch64_vec_op_count::base_issue_info): Likewise.
4285 (aarch64_vec_op_count::simd_issue_info): Likewise.
4286 (aarch64_vec_op_count::sve_issue_info): Likewise.
4287 (aarch64_vec_op_count::m_issue_info): New member variable.
4288 (aarch64_vec_op_count::m_vec_flags): Likewise.
4289 (aarch64_vector_costs): Add a constructor.
4290 (aarch64_vector_costs::m_sve_ops): Change type to aarch64_vec_op_count.
4291 (aarch64_vector_costs::aarch64_vector_costs): New function.
4292 Initialize m_scalar_ops, m_advsimd_ops and m_sve_ops.
4293 (aarch64_vector_costs::count_ops): Remove vec_flags and
4294 issue_info parameters, using the new aarch64_vec_op_count
4296 (aarch64_vector_costs::add_stmt_cost): Update call accordingly.
4297 (aarch64_sve_op_count::dump): Fold into...
4298 (aarch64_vec_op_count::dump): ..here.
4300 2021-11-12 Richard Sandiford <richard.sandiford@arm.com>
4302 * config/aarch64/aarch64.c: Include tree-dfa.h.
4303 (aarch64_check_consecutive_mems): New function that takes MEM_EXPR
4304 and MEM_OFFSET into account.
4305 (aarch64_swap_ldrstr_operands): Use it.
4306 (aarch64_operands_ok_for_ldpstp): Likewise. Check that the
4307 address of the second memory doesn't depend on the result of
4310 2021-11-12 Jan Hubicka <hubicka@ucw.cz>
4312 * ipa-pure-const.c (ipa_make_function_pure): Fix exit condition.
4314 2021-11-12 Jan Hubicka <jh@suse.cz>
4316 PR tree-optimization/103175
4317 * ipa-modref.c (modref_lattice::merge): Add sanity check.
4318 (callee_to_caller_flags): Make flags adjustment sane.
4319 (modref_eaf_analysis::analyze_ssa_name): Likewise.
4321 2021-11-12 Jan Hubicka <jh@suse.cz>
4324 * ipa-modref.c (analyze_function, modref_propagate_in_scc): Do
4325 not mark pure/const function if there are side-effects.
4327 2021-11-12 Chung-Lin Tang <cltang@codesourcery.com>
4329 * tree.h (OMP_CLAUSE_MAP_RUNTIME_IMPLICIT_P): New access macro for
4330 'implicit' bit, using 'base.deprecated_flag' field of tree_node.
4331 * tree-pretty-print.c (dump_omp_clause): Add support for printing
4332 implicit attribute in tree dumping.
4333 * gimplify.c (gimplify_adjust_omp_clauses_1):
4334 Set OMP_CLAUSE_MAP_RUNTIME_IMPLICIT_P to 1 if map clause is implicitly
4336 (gimplify_adjust_omp_clauses): Adjust place of adding implicitly created
4337 clauses, from simple append, to starting of list, after non-map clauses.
4338 * omp-low.c (lower_omp_target): Add GOMP_MAP_IMPLICIT bits into kind
4339 values passed to libgomp for implicit maps.
4341 2021-11-12 Jakub Jelinek <jakub@redhat.com>
4343 * omp-builtins.def (BUILT_IN_GOMP_TEAMS): Remove.
4344 (BUILT_IN_GOMP_TEAMS4): New.
4345 * builtin-types.def (BT_FN_VOID_UINT_UINT): Remove.
4346 (BT_FN_BOOL_UINT_UINT_UINT_BOOL): New.
4347 * omp-low.c (lower_omp_teams): Use GOMP_teams4 instead of
4348 GOMP_teams, pass to it also num_teams lower-bound expression
4349 or a dup of upper-bound if it is missing and a flag whether
4350 it is the first call or not.
4352 2021-11-12 Martin Liska <mliska@suse.cz>
4354 PR tree-optimization/102497
4355 * gimple-predicate-analysis.cc (add_pred): Remove unused
4358 2021-11-12 Richard Biener <rguenther@suse.de>
4360 PR tree-optimization/103204
4361 * tree-ssa-sccvn.c (valueize_refs_1): Re-valueize the
4362 top operand after folding in an address.
4364 2021-11-11 Aldy Hernandez <aldyh@redhat.com>
4366 * gimple-range-path.cc (path_range_query::path_range_query): New
4367 ctor without a ranger.
4368 (path_range_query::~path_range_query): Free ranger if necessary.
4369 (path_range_query::range_on_path_entry): Adjust m_ranger for pointer.
4370 (path_range_query::ssa_range_in_phi): Same.
4371 (path_range_query::compute_ranges_in_block): Same.
4372 (path_range_query::compute_imports): Same.
4373 (path_range_query::compute_ranges): Same.
4374 (path_range_query::range_of_stmt): Same.
4375 (path_range_query::compute_outgoing_relations): Same.
4376 * gimple-range-path.h (class path_range_query): New ctor.
4377 * tree-ssa-loop-ch.c (ch_base::copy_headers): Remove gimple_ranger
4378 as path_range_query allocates one.
4379 * tree-ssa-threadbackward.c (class back_threader): Remove m_ranger.
4380 (back_threader::~back_threader): Same.
4382 2021-11-11 Aldy Hernandez <aldyh@redhat.com>
4384 * tree-ssa-threadbackward.c
4385 (back_threader_profitability::profitable_path_p): Remove loop
4386 crossing restriction.
4388 2021-11-11 Jan Hubicka <hubicka@ucw.cz>
4390 * ipa-modref.c (analyze_function): Do pure/const discovery, return
4392 (pass_modref::execute): If pure/const is discovered fixup cfg.
4393 (ignore_edge): Do not ignore pure/const edges.
4394 (modref_propagate_in_scc): Do pure/const discovery, return true if
4395 cdtor was promoted pure/const.
4396 (pass_ipa_modref::execute): If needed remove unreachable functions.
4397 * ipa-pure-const.c (warn_function_noreturn): Fix whitespace.
4398 (warn_function_cold): Likewise.
4399 (skip_function_for_local_pure_const): Move earlier.
4400 (ipa_make_function_const): Break out from ...
4401 (ipa_make_function_pure): Break out from ...
4402 (propagate_pure_const): ... here.
4403 (pass_local_pure_const::execute): Use it.
4404 * ipa-utils.h (ipa_make_function_const): Declare.
4405 (ipa_make_function_pure): Declare.
4406 * passes.def: Move early modref after pure-const.
4408 2021-11-11 David Malcolm <dmalcolm@redhat.com>
4411 * diagnostic-show-locus.c (def_policy): Use def_tabstop.
4413 2021-11-11 Jonathan Wright <jonathan.wright@arm.com>
4415 * config/aarch64/aarch64-builtins.c (TYPES_COMBINE): Delete.
4416 (TYPES_COMBINEP): Delete.
4417 * config/aarch64/aarch64-simd-builtins.def: Declare type-
4418 qualified builtins for vcombine_* intrinsics.
4419 * config/aarch64/arm_neon.h (vcombine_s8): Remove unnecessary
4421 (vcombine_s16): Likewise.
4422 (vcombine_s32): Likewise.
4423 (vcombine_f32): Likewise.
4424 (vcombine_u8): Use type-qualified builtin and remove casts.
4425 (vcombine_u16): Likewise.
4426 (vcombine_u32): Likewise.
4427 (vcombine_u64): Likewise.
4428 (vcombine_p8): Likewise.
4429 (vcombine_p16): Likewise.
4430 (vcombine_p64): Likewise.
4431 (vcombine_bf16): Remove unnecessary cast.
4432 * config/aarch64/iterators.md (VD_I): New mode iterator.
4433 (VDC_P): New mode iterator.
4435 2021-11-11 Jonathan Wright <jonathan.wright@arm.com>
4437 * config/aarch64/aarch64-builtins.c (TYPES_LOAD1_U): Define.
4438 (TYPES_LOAD1_P): Define.
4439 (TYPES_STORE1_U): Define.
4440 (TYPES_STORE1P): Rename to...
4441 (TYPES_STORE1_P): This.
4442 (get_mem_type_for_load_store): Add unsigned and poly types.
4443 (aarch64_general_gimple_fold_builtin): Add unsigned and poly
4444 type-qualified builtin declarations.
4445 * config/aarch64/aarch64-simd-builtins.def: Declare type-
4446 qualified builtins for LD1/ST1.
4447 * config/aarch64/arm_neon.h (vld1_p8): Use type-qualified
4448 builtin and remove cast.
4449 (vld1_p16): Likewise.
4450 (vld1_u8): Likewise.
4451 (vld1_u16): Likewise.
4452 (vld1_u32): Likewise.
4453 (vld1q_p8): Likewise.
4454 (vld1q_p16): Likewise.
4455 (vld1q_p64): Likewise.
4456 (vld1q_u8): Likewise.
4457 (vld1q_u16): Likewise.
4458 (vld1q_u32): Likewise.
4459 (vld1q_u64): Likewise.
4460 (vst1_p8): Likewise.
4461 (vst1_p16): Likewise.
4462 (vst1_u8): Likewise.
4463 (vst1_u16): Likewise.
4464 (vst1_u32): Likewise.
4465 (vst1q_p8): Likewise.
4466 (vst1q_p16): Likewise.
4467 (vst1q_p64): Likewise.
4468 (vst1q_u8): Likewise.
4469 (vst1q_u16): Likewise.
4470 (vst1q_u32): Likewise.
4471 (vst1q_u64): Likewise.
4472 * config/aarch64/iterators.md (VALLP_NO_DI): New iterator.
4474 2021-11-11 Jonathan Wright <jonathan.wright@arm.com>
4476 * config/aarch64/aarch64-simd-builtins.def: Declare unsigned
4477 builtins for vector reduction.
4478 * config/aarch64/arm_neon.h (vaddv_u8): Use type-qualified
4479 builtin and remove casts.
4480 (vaddv_u16): Likewise.
4481 (vaddv_u32): Likewise.
4482 (vaddvq_u8): Likewise.
4483 (vaddvq_u16): Likewise.
4484 (vaddvq_u32): Likewise.
4485 (vaddvq_u64): Likewise.
4487 2021-11-11 Jonathan Wright <jonathan.wright@arm.com>
4489 * config/aarch64/aarch64-simd-builtins.def:
4490 * config/aarch64/arm_neon.h (vpaddq_u8): Use type-qualified
4491 builtin and remove casts.
4492 (vpaddq_u16): Likewise.
4493 (vpaddq_u32): Likewise.
4494 (vpaddq_u64): Likewise.
4495 (vpadd_u8): Likewise.
4496 (vpadd_u16): Likewise.
4497 (vpadd_u32): Likewise.
4498 (vpaddd_u64): Likewise.
4500 2021-11-11 Jonathan Wright <jonathan.wright@arm.com>
4502 * config/aarch64/aarch64-simd-builtins.def: Declare unsigned
4503 builtins for [r]subhn[2].
4504 * config/aarch64/arm_neon.h (vsubhn_s16): Remove unnecessary
4506 (vsubhn_s32): Likewise.
4507 (vsubhn_s64): Likewise.
4508 (vsubhn_u16): Use type-qualified builtin and remove casts.
4509 (vsubhn_u32): Likewise.
4510 (vsubhn_u64): Likewise.
4511 (vrsubhn_s16): Remove unnecessary cast.
4512 (vrsubhn_s32): Likewise.
4513 (vrsubhn_s64): Likewise.
4514 (vrsubhn_u16): Use type-qualified builtin and remove casts.
4515 (vrsubhn_u32): Likewise.
4516 (vrsubhn_u64): Likewise.
4517 (vrsubhn_high_s16): Remove unnecessary cast.
4518 (vrsubhn_high_s32): Likewise.
4519 (vrsubhn_high_s64): Likewise.
4520 (vrsubhn_high_u16): Use type-qualified builtin and remove
4522 (vrsubhn_high_u32): Likewise.
4523 (vrsubhn_high_u64): Likewise.
4524 (vsubhn_high_s16): Remove unnecessary cast.
4525 (vsubhn_high_s32): Likewise.
4526 (vsubhn_high_s64): Likewise.
4527 (vsubhn_high_u16): Use type-qualified builtin and remove
4529 (vsubhn_high_u32): Likewise.
4530 (vsubhn_high_u64): Likewise.
4532 2021-11-11 Jonathan Wright <jonathan.wright@arm.com>
4534 * config/aarch64/aarch64-simd-builtins.def: Declare unsigned
4535 builtins for [r]addhn[2].
4536 * config/aarch64/arm_neon.h (vaddhn_s16): Remove unnecessary
4538 (vaddhn_s32): Likewise.
4539 (vaddhn_s64): Likewise.
4540 (vaddhn_u16): Use type-qualified builtin and remove casts.
4541 (vaddhn_u32): Likewise.
4542 (vaddhn_u64): Likewise.
4543 (vraddhn_s16): Remove unnecessary cast.
4544 (vraddhn_s32): Likewise.
4545 (vraddhn_s64): Likewise.
4546 (vraddhn_u16): Use type-qualified builtin and remove casts.
4547 (vraddhn_u32): Likewise.
4548 (vraddhn_u64): Likewise.
4549 (vaddhn_high_s16): Remove unnecessary cast.
4550 (vaddhn_high_s32): Likewise.
4551 (vaddhn_high_s64): Likewise.
4552 (vaddhn_high_u16): Use type-qualified builtin and remove
4554 (vaddhn_high_u32): Likewise.
4555 (vaddhn_high_u64): Likewise.
4556 (vraddhn_high_s16): Remove unnecessary cast.
4557 (vraddhn_high_s32): Likewise.
4558 (vraddhn_high_s64): Likewise.
4559 (vraddhn_high_u16): Use type-qualified builtin and remove
4561 (vraddhn_high_u32): Likewise.
4562 (vraddhn_high_u64): Likewise.
4564 2021-11-11 Jonathan Wright <jonathan.wright@arm.com>
4566 * config/aarch64/aarch64-simd-builtins.def: Use BINOPU type
4567 qualifiers in generator macros for uhsub builtins.
4568 * config/aarch64/arm_neon.h (vhsub_s8): Remove unnecessary
4570 (vhsub_s16): Likewise.
4571 (vhsub_s32): Likewise.
4572 (vhsub_u8): Use type-qualified builtin and remove casts.
4573 (vhsub_u16): Likewise.
4574 (vhsub_u32): Likewise.
4575 (vhsubq_s8): Remove unnecessary cast.
4576 (vhsubq_s16): Likewise.
4577 (vhsubq_s32): Likewise.
4578 (vhsubq_u8): Use type-qualified builtin and remove casts.
4579 (vhsubq_u16): Likewise.
4580 (vhsubq_u32): Likewise.
4582 2021-11-11 Jonathan Wright <jonathan.wright@arm.com>
4584 * config/aarch64/aarch64-simd-builtins.def: Use BINOPU type
4585 qualifiers in generator macros for u[r]hadd builtins.
4586 * config/aarch64/arm_neon.h (vhadd_s8): Remove unnecessary
4588 (vhadd_s16): Likewise.
4589 (vhadd_s32): Likewise.
4590 (vhadd_u8): Use type-qualified builtin and remove casts.
4591 (vhadd_u16): Likewise.
4592 (vhadd_u32): Likewise.
4593 (vhaddq_s8): Remove unnecessary cast.
4594 (vhaddq_s16): Likewise.
4595 (vhaddq_s32): Likewise.
4596 (vhaddq_u8): Use type-qualified builtin and remove casts.
4597 (vhaddq_u16): Likewise.
4598 (vhaddq_u32): Likewise.
4599 (vrhadd_s8): Remove unnecessary cast.
4600 (vrhadd_s16): Likewise.
4601 (vrhadd_s32): Likewise.
4602 (vrhadd_u8): Use type-qualified builtin and remove casts.
4603 (vrhadd_u16): Likewise.
4604 (vrhadd_u32): Likewise.
4605 (vrhaddq_s8): Remove unnecessary cast.
4606 (vrhaddq_s16): Likewise.
4607 (vrhaddq_s32): Likewise.
4608 (vrhaddq_u8): Use type-wualified builtin and remove casts.
4609 (vrhaddq_u16): Likewise.
4610 (vrhaddq_u32): Likewise.
4612 2021-11-11 Jonathan Wright <jonathan.wright@arm.com>
4614 * config/aarch64/aarch64-simd-builtins.def: Use BINOPU type
4615 qualifiers in generator macros for usub[lw][2] builtins.
4616 * config/aarch64/arm_neon.h (vsubl_s8): Remove unnecessary
4618 (vsubl_s16): Likewise.
4619 (vsubl_s32): Likewise.
4620 (vsubl_u8): Use type-qualified builtin and remove casts.
4621 (vsubl_u16): Likewise.
4622 (vsubl_u32): Likewise.
4623 (vsubl_high_s8): Remove unnecessary cast.
4624 (vsubl_high_s16): Likewise.
4625 (vsubl_high_s32): Likewise.
4626 (vsubl_high_u8): Use type-qualified builtin and remove casts.
4627 (vsubl_high_u16): Likewise.
4628 (vsubl_high_u32): Likewise.
4629 (vsubw_s8): Remove unnecessary casts.
4630 (vsubw_s16): Likewise.
4631 (vsubw_s32): Likewise.
4632 (vsubw_u8): Use type-qualified builtin and remove casts.
4633 (vsubw_u16): Likewise.
4634 (vsubw_u32): Likewise.
4635 (vsubw_high_s8): Remove unnecessary cast.
4636 (vsubw_high_s16): Likewise.
4637 (vsubw_high_s32): Likewise.
4638 (vsubw_high_u8): Use type-qualified builtin and remove casts.
4639 (vsubw_high_u16): Likewise.
4640 (vsubw_high_u32): Likewise.
4642 2021-11-11 Jonathan Wright <jonathan.wright@arm.com>
4644 * config/aarch64/aarch64-simd-builtins.def: Use BINOPU type
4645 qualifiers in generator macros for uadd[lw][2] builtins.
4646 * config/aarch64/arm_neon.h (vaddl_s8): Remove unnecessary
4648 (vaddl_s16): Likewise.
4649 (vaddl_s32): Likewise.
4650 (vaddl_u8): Use type-qualified builtin and remove casts.
4651 (vaddl_u16): Likewise.
4652 (vaddl_u32): Likewise.
4653 (vaddl_high_s8): Remove unnecessary cast.
4654 (vaddl_high_s16): Likewise.
4655 (vaddl_high_s32): Likewise.
4656 (vaddl_high_u8): Use type-qualified builtin and remove casts.
4657 (vaddl_high_u16): Likewise.
4658 (vaddl_high_u32): Likewise.
4659 (vaddw_s8): Remove unnecessary cast.
4660 (vaddw_s16): Likewise.
4661 (vaddw_s32): Likewise.
4662 (vaddw_u8): Use type-qualified builtin and remove casts.
4663 (vaddw_u16): Likewise.
4664 (vaddw_u32): Likewise.
4665 (vaddw_high_s8): Remove unnecessary cast.
4666 (vaddw_high_s16): Likewise.
4667 (vaddw_high_s32): Likewise.
4668 (vaddw_high_u8): Use type-qualified builtin and remove casts.
4669 (vaddw_high_u16): Likewise.
4670 (vaddw_high_u32): Likewise.
4672 2021-11-11 Jonathan Wright <jonathan.wright@arm.com>
4674 * config/aarch64/aarch64-simd-builtins.def: Declare type-
4675 qualified builtins for [R]SHRN[2].
4676 * config/aarch64/arm_neon.h (vshrn_n_u16): Use type-qualified
4677 builtin and remove casts.
4678 (vshrn_n_u32): Likewise.
4679 (vshrn_n_u64): Likewise.
4680 (vrshrn_high_n_u16): Likewise.
4681 (vrshrn_high_n_u32): Likewise.
4682 (vrshrn_high_n_u64): Likewise.
4683 (vrshrn_n_u16): Likewise.
4684 (vrshrn_n_u32): Likewise.
4685 (vrshrn_n_u64): Likewise.
4686 (vshrn_high_n_u16): Likewise.
4687 (vshrn_high_n_u32): Likewise.
4688 (vshrn_high_n_u64): Likewise.
4690 2021-11-11 Jonathan Wright <jonathan.wright@arm.com>
4692 * config/aarch64/aarch64-simd-builtins.def: Declare unsigned
4693 type-qualified builtins for XTN[2].
4694 * config/aarch64/arm_neon.h (vmovn_high_u16): Use type-
4695 qualified builtin and remove casts.
4696 (vmovn_high_u32): Likewise.
4697 (vmovn_high_u64): Likewise.
4698 (vmovn_u16): Likewise.
4699 (vmovn_u32): Likewise.
4700 (vmovn_u64): Likewise.
4702 2021-11-11 Jonathan Wright <jonathan.wright@arm.com>
4704 * config/aarch64/aarch64-simd-builtins.def: Use poly type
4705 qualifier in builtin generator macros.
4706 * config/aarch64/arm_neon.h (vmul_p8): Use type-qualified
4707 builtin and remove casts.
4708 (vmulq_p8): Likewise.
4709 (vmull_high_p8): Likewise.
4710 (vmull_p8): Likewise.
4712 2021-11-11 Jonathan Wright <jonathan.wright@arm.com>
4714 * config/aarch64/aarch64-simd-builtins.def: Declare type-
4715 qualified builtin generators for unsigned MLA/MLS intrinsics.
4716 * config/aarch64/arm_neon.h (vmla_n_u16): Use type-qualified
4718 (vmla_n_u32): Likewise.
4719 (vmla_u8): Likewise.
4720 (vmla_u16): Likewise.
4721 (vmla_u32): Likewise.
4722 (vmlaq_n_u16): Likewise.
4723 (vmlaq_n_u32): Likewise.
4724 (vmlaq_u8): Likewise.
4725 (vmlaq_u16): Likewise.
4726 (vmlaq_u32): Likewise.
4727 (vmls_n_u16): Likewise.
4728 (vmls_n_u32): Likewise.
4729 (vmls_u8): Likewise.
4730 (vmls_u16): Likewise.
4731 (vmls_u32): Likewise.
4732 (vmlsq_n_u16): Likewise.
4733 (vmlsq_n_u32): Likewise.
4734 (vmlsq_u8): Likewise.
4735 (vmlsq_u16): Likewise.
4736 (vmlsq_u32): Likewise.
4738 2021-11-11 Jan Hubicka <hubicka@ucw.cz>
4740 * ipa-modref.c (modref_summary::useful_p): Check also for side-effects
4741 with looping const/pure.
4742 (modref_summary_lto::useful_p): Likewise.
4743 (merge_call_side_effects): Merge side effects before early exit
4745 (process_fnspec): Also handle pure functions.
4746 (analyze_call): Do not early exit on looping pure const.
4747 (propagate_unknown_call): Also handle nontrivial SCC as side-effect.
4748 (modref_propagate_in_scc): Update.
4750 2021-11-11 Richard Biener <rguenther@suse.de>
4752 PR tree-optimization/103190
4753 * tree-ssa-reassoc.c (insert_stmt_after): Only assert on asm goto.
4755 2021-11-11 Aldy Hernandez <aldyh@redhat.com>
4757 * gimple-range-path.cc (path_range_query::add_copies_to_imports):
4759 (path_range_query::compute_imports): ...this. Adapt it so it can
4760 be passed the imports bitmap instead of working on m_imports.
4761 (path_range_query::compute_ranges): Call compute_imports in all
4762 cases unless an imports bitmap is passed.
4763 * gimple-range-path.h (path_range_query::compute_imports): New.
4764 (path_range_query::add_copies_to_imports): Remove.
4765 * tree-ssa-threadbackward.c (back_threader::resolve_def): Remove.
4766 (back_threader::find_paths_to_names): Inline resolve_def.
4767 (back_threader::find_paths): Call compute_imports.
4768 (back_threader::resolve_phi): Adjust comment.
4770 2021-11-11 Richard Biener <rguenther@suse.de>
4772 PR tree-optimization/103188
4773 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p):
4774 Remove query parameter, split out check for size
4776 (ch_base::m_ranger, cb_base::m_query): Remove.
4777 (ch_base::copy_headers): Split processing loop into
4778 analysis around which we allocate and use ranger and
4779 transform where we do not.
4780 (pass_ch::execute): Do not allocate/free ranger here.
4781 (pass_ch_vect::execute): Likewise.
4783 2021-11-11 Jan Hubicka <hubicka@ucw.cz>
4785 * ipa-pure-const.c (propagate_pure_const): Self recursion is
4788 2021-11-11 Jan Hubicka <hubicka@ucw.cz>
4790 * cgraph.c (set_noreturn_flag_1): New function.
4791 (cgraph_node::set_noreturn_flag): New member function
4792 * cgraph.h (cgraph_node::set_noreturn_flags): Declare.
4793 * ipa-pure-const.c (pass_local_pure_const::execute): Use it.
4795 2021-11-11 Aldy Hernandez <aldyh@redhat.com>
4796 Richard Biener <rguenther@suse.de>
4798 * tree-ssa-loop-ch.c (entry_loop_condition_is_static): Resolve
4799 statically to the edge remaining in the loop.
4801 2021-11-11 Richard Biener <rguenther@suse.de>
4803 PR middle-end/103181
4804 * tree-eh.c (operation_could_trap_helper_p): Properly
4805 check vector constants for a zero element for integer
4806 division. Separate floating point and integer division code.
4808 2021-11-11 Jakub Jelinek <jakub@redhat.com>
4811 * dwarf2out.c (field_byte_offset): Do the PCC_BITFIELD_TYPE_MATTERS
4812 handling only for DECL_BIT_FIELD_TYPE decls.
4814 2021-11-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
4817 * config/aarch64/aarch64.c (aarch64_process_target_attr): Check if
4818 token is arch extension without leading '+' and emit appropriate
4819 diagnostic for the same.
4821 2021-11-11 Jakub Jelinek <jakub@redhat.com>
4823 * tree.h (OMP_CLAUSE_NUM_TEAMS_EXPR): Rename to ...
4824 (OMP_CLAUSE_NUM_TEAMS_UPPER_EXPR): ... this.
4825 (OMP_CLAUSE_NUM_TEAMS_LOWER_EXPR): Define.
4826 * tree.c (omp_clause_num_ops): Increase num ops for
4827 OMP_CLAUSE_NUM_TEAMS to 2.
4828 * tree-pretty-print.c (dump_omp_clause): Print optional lower bound
4829 for OMP_CLAUSE_NUM_TEAMS.
4830 * gimplify.c (gimplify_scan_omp_clauses): Gimplify
4831 OMP_CLAUSE_NUM_TEAMS_LOWER_EXPR if non-NULL.
4832 (optimize_target_teams): Use OMP_CLAUSE_NUM_TEAMS_UPPER_EXPR instead
4833 of OMP_CLAUSE_NUM_TEAMS_EXPR. Handle OMP_CLAUSE_NUM_TEAMS_LOWER_EXPR.
4834 * omp-low.c (lower_omp_teams): Use OMP_CLAUSE_NUM_TEAMS_UPPER_EXPR
4835 instead of OMP_CLAUSE_NUM_TEAMS_EXPR.
4836 * omp-expand.c (expand_teams_call, get_target_arguments): Likewise.
4838 2021-11-11 Richard Biener <rguenther@suse.de>
4840 * cfganal.c (find_pdom): Remove.
4841 (control_dependences::find_control_dependence): Remove
4842 special-casing of entry block, call get_immediate_dominator
4844 * gimple-predicate-analysis.cc (find_pdom): Remove.
4845 (find_dom): Likewise.
4846 (find_control_equiv_block): Call get_immediate_dominator
4848 (compute_control_dep_chain): Likewise.
4849 (predicate::init_from_phi_def): Likewise.
4851 2021-11-11 Richard Biener <rguenther@suse.de>
4853 * cfganal.h (control_dependences::control_dependence_map):
4855 (control_dependences::m_bitmaps): New.
4856 * cfganal.c (control_dependences::set_control_dependence_map_bit):
4858 (control_dependences::clear_control_dependence_bitmap):
4860 (control_dependences::find_control_dependence): Do not
4861 find_edge for the abnormal edge test.
4862 (control_dependences::control_dependences): Instead do not
4863 add abnormal edges to the edge list. Adjust.
4864 (control_dependences::~control_dependences): Likewise.
4865 (control_dependences::get_edges_dependent_on): Likewise.
4866 * function-tests.c: Include bitmap.h.
4868 2021-11-11 Kewen Lin <linkw@linux.ibm.com>
4870 * doc/invoke.texi: Change references to "future cpu" to "power10",
4871 "-mcpu=future" to "-mcpu=power10". Adjust words for float128.
4873 2021-11-11 Cui,Lili <lili.cui@intel.com>
4875 * config/i386/i386-options.c (m_CORE_AVX2): Remove Alderlake
4877 (processor_cost_table): Use alderlake_cost for Alderlake.
4878 * config/i386/i386.c (ix86_sched_init_global): Handle Alderlake.
4879 * config/i386/x86-tune-costs.h (struct processor_costs): Add alderlake
4881 * config/i386/x86-tune-sched.c (ix86_issue_rate): Change Alderlake
4883 (ix86_adjust_cost): Handle Alderlake.
4884 * config/i386/x86-tune.def (X86_TUNE_SCHEDULE): Enable for Alderlake.
4885 (X86_TUNE_PARTIAL_REG_DEPENDENCY): Likewise.
4886 (X86_TUNE_SSE_PARTIAL_REG_DEPENDENCY): Likewise.
4887 (X86_TUNE_SSE_PARTIAL_REG_FP_CONVERTS_DEPENDENCY): Likewise.
4888 (X86_TUNE_SSE_PARTIAL_REG_CONVERTS_DEPENDENCY): Likewise.
4889 (X86_TUNE_MEMORY_MISMATCH_STALL): Likewise.
4890 (X86_TUNE_USE_LEAVE): Likewise.
4891 (X86_TUNE_PUSH_MEMORY): Likewise.
4892 (X86_TUNE_USE_INCDEC): Likewise.
4893 (X86_TUNE_INTEGER_DFMODE_MOVES): Likewise.
4894 (X86_TUNE_MISALIGNED_MOVE_STRING_PRO_EPILOGUES): Likewise.
4895 (X86_TUNE_USE_SAHF): Likewise.
4896 (X86_TUNE_USE_BT): Likewise.
4897 (X86_TUNE_AVOID_FALSE_DEP_FOR_BMI): Likewise.
4898 (X86_TUNE_ONE_IF_CONV_INSN): Likewise.
4899 (X86_TUNE_AVOID_MFENCE): Likewise.
4900 (X86_TUNE_USE_SIMODE_FIOP): Likewise.
4901 (X86_TUNE_EXT_80387_CONSTANTS): Likewise.
4902 (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL): Likewise.
4903 (X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Likewise.
4904 (X86_TUNE_SSE_TYPELESS_STORES): Likewise.
4905 (X86_TUNE_SSE_LOAD0_BY_PXOR): Likewise.
4906 (X86_TUNE_AVOID_4BYTE_PREFIXES): Likewise.
4907 (X86_TUNE_USE_GATHER): Disable for Alderlake.
4908 (X86_TUNE_AVX256_MOVE_BY_PIECES): Likewise.
4909 (X86_TUNE_AVX256_STORE_BY_PIECES): Likewise.
4911 2021-11-11 liuhongt <hongtao.liu@intel.com>
4914 * config/i386/sse.md (V_128_256): Extend to V8HF/V16HF.
4915 (avxsizesuffix): Ditto.
4917 2021-11-11 Kito Cheng <kito.cheng@sifive.com>
4919 * common/config/riscv/riscv-common.c (riscv_subset_list::to_string): Fix
4920 wrong marco checking.
4922 2021-11-10 Aldy Hernandez <aldyh@redhat.com>
4924 PR tree-optimization/102906
4925 * tree-ssa-loop-ch.c (entry_loop_condition_is_static): New.
4926 (should_duplicate_loop_header_p): Call entry_loop_condition_is_static.
4927 (class ch_base): Add m_ranger and m_query.
4928 (ch_base::copy_headers): Pass m_query to
4929 entry_loop_condition_is_static.
4930 (pass_ch::execute): Allocate and deallocate m_ranger and
4932 (pass_ch_vect::execute): Same.
4934 2021-11-10 Andrew Pinski <apinski@marvell.com>
4937 * config/aarch64/aarch64-simd.md (aarch64_simd_dup<mode>):
4938 Use vwcore iterator for the r constraint output string.
4940 2021-11-10 qing zhao <qing.zhao@oracle.com>
4942 * internal-fn.c (expand_DEFERRED_INIT): Apply pattern initialization
4943 only when have_insn_for return true for the mode. Fix a memory leak.
4945 2021-11-10 Christophe Lyon <christophe.lyon@foss.st.com>
4947 * config/arm/arm.c (cortexa9_extra_costs, cortexa8_extra_costs,
4948 cortexa5_extra_costs, cortexa7_extra_costs,
4949 cortexa12_extra_costs, cortexa15_extra_costs, v7m_extra_costs):
4950 Initialize movi, dup and extract costing fields.
4952 2021-11-10 Aldy Hernandez <aldyh@redhat.com>
4954 * gimple-range-path.cc (path_range_query::path_range_query): Do
4956 (path_range_query::dump): Change m_path uses to non-pointer.
4957 (path_range_query::defined_outside_path): Same.
4958 (path_range_query::set_path): Same.
4959 (path_range_query::add_copies_to_imports): Same.
4960 (path_range_query::range_of_stmt): Same.
4961 (path_range_query::compute_outgoing_relations): Same.
4962 (path_range_query::compute_ranges): Imports are now optional.
4963 Implement overload that takes an edge.
4964 * gimple-range-path.h (class path_range_query): Make imports
4965 optional for compute_ranges. Add compute_ranges(edge) overload.
4966 Make m_path an auto_vec instead of a pointer and adjust
4969 2021-11-10 Tamar Christina <tamar.christina@arm.com>
4971 * tree-vectorizer.h (struct scalar_cond_masked_key): Add inverted_p.
4972 (default_hash_traits<scalar_conf_masked_key>): Likewise.
4973 * tree-vect-stmts.c (vectorizable_condition): Check if inverse of mask
4975 * tree-vectorizer.c (scalar_cond_masked_key::get_cond_ops_from_tree):
4976 Register mask inverses.
4978 2021-11-10 Tamar Christina <tamar.christina@arm.com>
4980 * tree-vectorizer.c (vectorize_loops): Do local CSE through RPVN upon
4981 successful vectorization.
4983 2021-11-10 Andrew MacLeod <amacleod@redhat.com>
4985 * gimple-range-cache.cc (sbr_vector::grow): New.
4986 (sbr_vector::set_bb_range): Call grow.
4987 (sbr_vector::get_bb_range): Same.
4988 (sbr_vector::bb_range_p): Remove assert.
4990 2021-11-10 Tamar Christina <tamar.christina@arm.com>
4992 * config/aarch64/aarch64-simd.md (*aarch64_topbits_shuffle<mode>_le
4993 ,*aarch64_topbits_shuffle<mode>_be): Remove.
4995 2021-11-10 Jan Hubicka <jh@suse.cz>
4997 * ipa-modref.c: Include tree-eh.h
4998 (modref_summary::modref_summary): Initialize side_effects.
4999 (struct modref_summary_lto): New bool field side_effects.
5000 (modref_summary_lto::modref_summary_lto): Initialize side_effects.
5001 (modref_summary::dump): Dump side_effects.
5002 (modref_summary_lto::dump): Dump side_effects.
5003 (merge_call_side_effects): Merge side effects.
5004 (process_fnspec): Calls to non-const/pure or looping
5005 function is a side effect.
5006 (analyze_call): Self-recursion is a side-effect; handle
5008 (analyze_load): Watch for volatile and throwing memory.
5009 (analyze_store): Likewise.
5010 (analyze_stmt): Watch for volatitle asm.
5011 (analyze_function): Handle side_effects.
5012 (modref_summaries::duplicate): Duplicate side_effects.
5013 (modref_summaries_lto::duplicate): Likewise.
5014 (modref_write): Stream side_effects.
5015 (read_section): Likewise.
5016 (update_signature): Update.
5017 (propagate_unknown_call): Handle side_effects.
5018 (modref_propagate_in_scc): Likewise.
5019 * ipa-modref.h (struct modref_summary): Add side_effects.
5020 * ipa-pure-const.c (special_builtin_state): Rename to ...
5021 (builtin_safe_for_const_function_p): ... this one.
5022 (check_call): Update.
5023 (finite_function_p): Break out from ...
5024 (propagate_pure_const): ... here
5025 * ipa-utils.h (finite_function): Declare.
5027 2021-11-10 Lucas A. M. Magalhães <lamm@linux.ibm.com>
5029 * config.gcc (powerpc*-*-*): Remove -rpath from
5030 --with-advance-toolchain.
5032 2021-11-10 Marek Polacek <polacek@redhat.com>
5035 * attribs.c (struct scoped_attributes): Add a bool member.
5036 (lookup_scoped_attribute_spec): Forward declare.
5037 (register_scoped_attributes): New bool parameter, defaulted to
5039 (handle_ignored_attributes_option): New function.
5040 (free_attr_data): New function.
5041 (init_attributes): Call handle_ignored_attributes_option.
5042 (attr_namespace_ignored_p): New function.
5043 (decl_attributes): Check attr_namespace_ignored_p before
5045 * attribs.h (free_attr_data): Declare.
5046 (register_scoped_attributes): Adjust declaration.
5047 (handle_ignored_attributes_option): Declare.
5048 (canonicalize_attr_name): New function template.
5049 (canonicalize_attr_name): Use it.
5050 * common.opt (Wattributes=): New option with a variable.
5051 * doc/extend.texi: Document #pragma GCC diagnostic ignored_attributes.
5052 * doc/invoke.texi: Document -Wno-attributes=.
5053 * opts.c (common_handle_option) <case OPT_Wattributes_>: Handle.
5054 * plugin.h (register_scoped_attributes): Adjust declaration.
5055 * toplev.c (compile_file): Call free_attr_data.
5057 2021-11-10 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
5059 * config/arm/arm-cpus.in (cortex-a710): New CPU.
5060 * config/arm/arm-tables.opt: Regenerate.
5061 * config/arm/arm-tune.md: Regenerate.
5062 * doc/invoke.texi: Update docs.
5064 2021-11-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
5066 * config/aarch64/aarch64-builtins.c
5067 (aarch64_general_gimple_fold_builtin): Mark argument as unused.
5069 2021-11-10 Martin Liska <mliska@suse.cz>
5071 * lto-wrapper.c (merge_and_complain): Make the first argument
5074 2021-11-10 Richard Sandiford <richard.sandiford@arm.com>
5076 * config/aarch64/iterators.md (optab): Use fmax_nan instead of
5077 smax_nan and fmin_nan instead of smin_nan.
5078 (maxmin_uns): Rename to...
5079 (fmaxmin): ...this and make the same changes. Remove entries
5080 unrelated to fmax* and fmin*.
5081 * config/aarch64/aarch64.md (<maxmin_uns><mode>3): Rename to...
5082 (<fmaxmin><mode>3): ...this.
5083 * config/aarch64/aarch64-simd.md (aarch64_<maxmin_uns>p<mode>):
5085 (aarch64_<optab>p<mode>): ...this.
5086 (<maxmin_uns><mode>3): Rename to...
5087 (<fmaxmin><mode>3): ...this.
5088 (reduc_<maxmin_uns>_scal_<mode>): Rename to...
5089 (reduc_<optab>_scal_<mode>): ...this and update gen* call.
5090 (aarch64_reduc_<maxmin_uns>_internal<mode>): Rename to...
5091 (aarch64_reduc_<optab>_internal<mode>): ...this.
5092 (aarch64_reduc_<maxmin_uns>_internalv2si): Rename to...
5093 (aarch64_reduc_<optab>_internalv2si): ...this.
5094 * config/aarch64/aarch64-sve.md (<maxmin_uns><mode>3): Rename to...
5095 (<fmaxmin><mode>3): ...this.
5096 * config/aarch64/aarch64-simd-builtins.def (smax_nan, smin_nan)
5098 (fmax_nan, fmin_nan): ...this.
5099 * config/aarch64/arm_neon.h (vmax_f32, vmax_f64, vmaxq_f32, vmaxq_f64)
5100 (vmin_f32, vmin_f64, vminq_f32, vminq_f64, vmax_f16, vmaxq_f16)
5101 (vmin_f16, vminq_f16): Update accordingly.
5103 2021-11-10 Richard Sandiford <richard.sandiford@arm.com>
5105 * tree-vectorizer.h (vector_costs::finish_cost): Take the
5106 corresponding scalar costs as a parameter.
5107 (finish_cost): Likewise.
5108 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
5109 (vect_estimate_min_profitable_iters): Update accordingly.
5110 * tree-vect-slp.c (vect_bb_vectorization_profitable_p): Likewise.
5111 * tree-vectorizer.c (vector_costs::finish_cost): Likewise.
5112 * config/aarch64/aarch64.c (aarch64_vector_costs::finish_cost):
5114 * config/rs6000/rs6000.c (rs6000_cost_data::finish_cost): Likewise.
5116 2021-11-10 Richard Sandiford <richard.sandiford@arm.com>
5118 * tree-vectorizer.h (_loop_vec_info::scalar_costs): New member
5120 (_loop_vec_info::single_scalar_iteration_cost): Delete.
5121 (LOOP_VINFO_SINGLE_SCALAR_ITERATION_COST): Delete.
5122 (vector_costs::total_cost): New function.
5123 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Update
5124 after above changes.
5125 (_loop_vec_info::~_loop_vec_info): Delete scalar_costs.
5126 (vect_compute_single_scalar_iteration_cost): Store the costs
5127 in loop_vinfo->scalar_costs.
5128 (vect_estimate_min_profitable_iters): Get the scalar cost from
5129 loop_vinfo->scalar_costs.
5131 2021-11-10 Richard Sandiford <richard.sandiford@arm.com>
5133 * tree-vectorizer.h (vector_costs::better_main_loop_than_p)
5134 (vector_costs::better_epilogue_loop_than_p)
5135 (vector_costs::compare_inside_loop_cost)
5136 (vector_costs::compare_outside_loop_cost): Likewise.
5137 * tree-vectorizer.c (vector_costs::better_main_loop_than_p)
5138 (vector_costs::better_epilogue_loop_than_p)
5139 (vector_costs::compare_inside_loop_cost)
5140 (vector_costs::compare_outside_loop_cost): New functions,
5141 containing code moved from...
5142 * tree-vect-loop.c (vect_better_loop_vinfo_p): ...here.
5144 2021-11-10 Richard Sandiford <richard.sandiford@arm.com>
5146 * tree-vectorizer.h (_loop_vec_info): Remove vec_outside_cost
5147 and vec_inside_cost.
5148 (vector_costs::outside_cost): New function.
5149 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Update
5151 (vect_estimate_min_profitable_iters): Likewise.
5152 (vect_better_loop_vinfo_p): Get the inside and outside costs
5153 from the loop_vec_infos' vector_costs.
5155 2021-11-10 Richard Sandiford <richard.sandiford@arm.com>
5157 * tree-vectorizer.h (vec_info::target_cost_data): Replace with...
5158 (_loop_vec_info::vector_costs): ...this.
5159 (LOOP_VINFO_TARGET_COST_DATA): Delete.
5160 * tree-vectorizer.c (vec_info::vec_info): Remove target_cost_data
5162 (vec_info::~vec_info): Remove corresponding delete.
5163 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
5164 vector_costs to null.
5165 (_loop_vec_info::~_loop_vec_info): Delete vector_costs.
5166 (vect_analyze_loop_operations): Update after above changes.
5167 (vect_analyze_loop_2): Likewise.
5168 (vect_estimate_min_profitable_iters): Likewise.
5169 * tree-vect-slp.c (vect_slp_analyze_operations): Likewise.
5171 2021-11-10 Jan Hubicka <hubicka@ucw.cz>
5173 * tree-core.h (EAF_DIRECT): Remove.
5174 (EAF_NOCLOBBER): Remove.
5175 (EAF_UNUSED): Remove.
5176 (EAF_NOESCAPE): Remove.
5177 (EAF_NO_DIRECT_CLOBBER): New.
5178 (EAF_NO_INDIRECT_CLOBBER): New.
5179 (EAF_NODIRECTESCAPE): Remove.
5180 (EAF_NO_DIRECT_ESCAPE): New.
5181 (EAF_NO_INDIRECT_ESCAPE): New.
5182 (EAF_NOT_RETURNED): Remove.
5183 (EAF_NOT_RETURNED_INDIRECTLY): New.
5184 (EAF_NOREAD): Remove.
5185 (EAF_NO_DIRECT_READ): New.
5186 (EAF_NO_INDIRECT_READ): New.
5187 * gimple.c (gimple_call_arg_flags): Update for new flags.
5188 (gimple_call_retslot_flags): Update for new flags.
5189 * ipa-modref.c (dump_eaf_flags): Likewise.
5190 (remove_useless_eaf_flags): Likewise.
5191 (deref_flags): Likewise.
5192 (modref_lattice::init): Likewise.
5193 (modref_lattice::merge): Likewise.
5194 (modref_lattice::merge_direct_load): Likewise.
5195 (modref_lattice::merge_direct_store): Likewise.
5196 (modref_eaf_analysis::merge_call_lhs_flags): Likewise.
5197 (callee_to_caller_flags): Likewise.
5198 (modref_eaf_analysis::analyze_ssa_name): Likewise.
5199 (modref_eaf_analysis::propagate): Likewise.
5200 (modref_merge_call_site_flags): Likewise.
5201 * ipa-modref.h (interposable_eaf_flags): Likewise.
5202 * tree-ssa-alias.c: (ref_maybe_used_by_call_p_1) Likewise.
5203 * tree-ssa-structalias.c (handle_call_arg): Likewise.
5204 (handle_rhs_call): Likewise.
5205 * tree-ssa-uninit.c (maybe_warn_pass_by_reference): Likewise.
5207 2021-11-10 Tamar Christina <tamar.christina@arm.com>
5209 * tree-vect-slp-patterns.c (complex_mul_pattern::matches): Remove l1node.
5211 2021-11-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
5213 * config/aarch64/aarch64-builtins.c
5214 (aarch64_general_gimple_fold_builtin): Change pointer alignment and
5217 2021-11-10 Jan Hubicka <jh@suse.cz>
5219 * ipa-modref-tree.h (modref_tree::remap_params): Fix off-by-one error.
5221 2021-11-10 H.J. Lu <hongjiu.lu@intel.com>
5222 Hongtao Liu <hongtao.liu@intel.com>
5224 PR middle-end/102566
5225 * match.pd (nop_atomic_bit_test_and_p): New match.
5226 * tree-ssa-ccp.c (convert_atomic_bit_not): New function.
5227 (gimple_nop_atomic_bit_test_and_p): New prototype.
5228 (optimize_atomic_bit_test_and): Transform equivalent, but slighly
5229 different cases to their canonical forms.
5231 2021-11-10 liuhongt <hongtao.liu@intel.com>
5233 PR tree-optimization/103126
5234 * tree-vect-loop.c (neutral_op_for_reduction): Remove static.
5235 * tree-vectorizer.h (neutral_op_for_reduction): Declare.
5236 * tree-if-conv.c : Include tree-vectorizer.h.
5237 (is_cond_scalar_reduction): Handle
5238 BIT_XOR_EXPR/BIT_IOR_EXPR/BIT_AND_EXPR.
5239 (convert_scalar_cond_reduction): Ditto.
5241 2021-11-10 konglin1 <lingling.kong@intel.com>
5243 * config/i386/sse.md (cmul<conj_op><mode>3): add new define_expand.
5244 (cmla<conj_op><mode>4): Likewise
5246 2021-11-10 Aldy Hernandez <aldyh@redhat.com>
5248 * tree-ssa-threadedge.c: Do not include
5249 gimple-ssa-evrp-analyze.h.
5250 * value-pointer-equiv.cc: Same.
5252 2021-11-10 Aldy Hernandez <aldyh@redhat.com>
5254 * tree-ssa-threadbackward.c
5255 (back_threader::maybe_register_path_dump): Abstract path dumping...
5256 (dump_path): ...here.
5257 (back_threader::resolve_phi): Call dump_path.
5260 2021-11-10 konglin1 <lingling.kong@intel.com>
5262 * config/i386/sse.md (fma_<complexpairopname>_<mode>_pair):
5263 Add new define_insn.
5264 (fma_<mode>_fmaddc_bcst): Add new define_insn_and_split.
5265 (fma_<mode>_fcmaddc_bcst): Likewise
5267 2021-11-10 liuhongt <hongtao.liu@intel.com>
5270 * match.pd: Simplify (trunc)fmax/fmin((extend)a, (extend)b) to
5273 2021-11-10 Andrew Pinski <apinski@marvell.com>
5276 * config/aarch64/aarch64.c (aarch64_evpc_ins): Don't use target
5277 as an input, use original one.
5279 2021-11-10 Sandra Loosemore <sandra@codesourcery.com>
5281 * config/nios2/nios2.c (nios2_can_inline_p): New.
5282 (TARGET_CAN_INLINE_P): Define.
5284 2021-11-09 Bill Schmidt <wschmidt@linux.ibm.com>
5286 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_new_builtin):
5287 Disable gimple fold for RS6000_BIF_{XVMINDP,XVMINSP,VMINFP} and
5288 RS6000_BIF_{XVMAXDP,XVMAXSP,VMAXFP} when fast-math is not set.
5289 (lxvrse_expand_builtin): Modify the expansion for sign extension.
5290 All extensions are done within VSX registers.
5292 2021-11-09 Andrew Pinski <apinski@marvell.com>
5294 PR tree-optimization/10352
5295 * match.pd: Remove check of TYPE_PRECISION for
5298 2021-11-09 Andrew MacLeod <amacleod@redhat.com>
5300 * function.c (allocate_struct_function): Don't set x_range_query.
5301 * function.h (get_range_query): Move to value-query.h.
5302 * gimple-range.cc (enable_ranger): Check that query is currently NULL.
5303 (disable_ranger): Clear function current query field.
5304 * value-query.cc (get_global_range_query): Relocate to:
5305 * value-query.h (get_global_range_query): Here and inline.
5306 (get_range_query): Relocate here from function.h.
5308 2021-11-09 Aldy Hernandez <aldyh@redhat.com>
5310 * tree-ssa-threadbackward.c
5311 (back_threader::maybe_register_path_dump): New.
5312 (back_threader::maybe_register_path): Call maybe_register_path_dump.
5314 2021-11-09 Aldy Hernandez <aldyh@redhat.com>
5316 * tree-ssa-threadbackward.c (back_threader::maybe_register_path):
5317 Return NULL when unprofitable.
5319 2021-11-09 Martin Jambor <mjambor@suse.cz>
5321 * tree.h (build_debug_expr_decl): Declare.
5322 * tree.c (build_debug_expr_decl): New function.
5323 * cfgexpand.c (avoid_deep_ter_for_debug): Use build_debug_expr_decl
5324 instead of building a DEBUG_EXPR_DECL.
5325 * ipa-param-manipulation.c
5326 (ipa_param_body_adjustments::prepare_debug_expressions): Likewise.
5327 * omp-simd-clone.c (ipa_simd_modify_stmt_ops): Likewise.
5328 * tree-ssa-ccp.c (optimize_atomic_bit_test_and): Likewise.
5329 * tree-ssa-phiopt.c (spaceship_replacement): Likewise.
5330 * tree-ssa-reassoc.c (make_new_ssa_for_def): Likewise.
5332 2021-11-09 Bill Schmidt <wschmidt@linux.ibm.com>
5334 * config/rs6000/rs6000-call.c (rs6000_type_string): New function.
5335 (def_builtin): Change debug formatting for easier parsing and
5336 include more information.
5337 (rs6000_init_builtins): Add dump of autogenerated builtins.
5338 (altivec_init_builtins): Dump __builtin_altivec_mask_for_load for
5341 2021-11-09 Marek Polacek <polacek@redhat.com>
5343 * ipa-modref.c (analyze_parms): Fix a typo.
5345 2021-11-09 Kito Cheng <kito.cheng@sifive.com>
5348 * common/config/riscv/riscv-common.c (multi_letter_subset_rank): Remove
5349 assertion for Z*-ext.
5351 2021-11-09 Jan Hubicka <hubicka@ucw.cz>
5353 * ipa-modref.c (analyze_parms): Add past_flags, past_retslot_flags
5354 and past_static_chain; merge past summary with current one.
5355 (analyze_function): Update.
5357 2021-11-09 Aldy Hernandez <aldyh@redhat.com>
5359 * doc/invoke.texi (Invoking GCC): Document --param=threader-debug.
5361 2021-11-09 Martin Liska <mliska@suse.cz>
5363 * print-rtl.c (rtx_writer::rtx_writer): Make the compilation
5364 conditional based on
5365 * print-rtl.h (class rtx_writer): Likewise.
5367 2021-11-09 Thomas Schwinge <thomas@codesourcery.com>
5369 * input.h (location_hash): Use 'BUILTINS_LOCATION' as spare value
5370 for 'Deleted'. Turn into a '#define'.
5372 2021-11-09 Aldy Hernandez <aldyh@redhat.com>
5374 * gimple-range-path.cc (path_range_query::dump): Clean up.
5375 (path_range_query::compute_ranges): Same.
5376 * value-relation.cc (path_oracle::dump): Same.
5378 2021-11-09 Aldy Hernandez <aldyh@redhat.com>
5380 * dumpfile.c (dump_options): Remove TDF_THREADING entry.
5381 * dumpfile.h (enum dump_flag): Remove TDF_THREADING and adjust
5383 * flag-types.h (enum threader_debug): New.
5384 * gimple-range-path.cc (DEBUG_SOLVER): Use param_threader_debug.
5385 * params.opt: Add entry for --param=threader-debug=.
5387 2021-11-09 Thomas Schwinge <thomas@codesourcery.com>
5389 PR middle-end/101204
5391 * diagnostic-spec.h (typedef xint_hash_t): Turn into...
5392 (struct xint_hash_t): ... this.
5393 * doc/gty.texi: Update.
5395 2021-11-09 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
5397 * config/arm/arm-cpus.in (armv9): New define.
5398 (ARMv9a): New group.
5399 (armv9-a): New arch definition.
5400 * config/arm/arm-tables.opt: Regenerate.
5401 * config/arm/arm.h (BASE_ARCH_9A): New arch enum value.
5402 * config/arm/t-aprofile: Added armv9-a and armv9+simd.
5403 * config/arm/t-arm-elf: Added arm9-a, v9_fps and all_v9_archs
5404 to MULTILIB_MATCHES.
5405 * config/arm/t-multilib: Added v9_a_nosimd_variants and
5406 v9_a_simd_variants to MULTILIB_MATCHES.
5407 * doc/invoke.texi: Update docs.
5409 2021-11-09 Martin Jambor <mjambor@suse.cz>
5412 * ipa-param-manipulation.c (replace_with_mapped_expr): Early
5413 return with error_mark_mode when part of expression is mapped to
5415 (ipa_param_body_adjustments::remap_with_debug_expressions): Set
5416 mapped value to NULL if walk_tree returns error_mark_mode.
5418 2021-11-09 Aldy Hernandez <aldyh@redhat.com>
5420 * tree-ssa-strlen.c (compare_nonzero_chars): Pass statement
5422 (get_addr_stridx): Same.
5424 (get_range_strlen_dynamic): Same.
5425 (handle_builtin_strlen): Same.
5426 (handle_builtin_strchr): Same.
5427 (handle_builtin_strcpy): Same.
5428 (maybe_diag_stxncpy_trunc): Same.
5429 (handle_builtin_stxncpy_strncat): Same.
5430 (handle_builtin_memcpy): Same.
5431 (handle_builtin_strcat): Same.
5432 (handle_alloc_call): Same.
5433 (handle_builtin_memset): Same.
5434 (handle_builtin_string_cmp): Same.
5435 (handle_pointer_plus): Same.
5436 (count_nonzero_bytes_addr): Same.
5437 (count_nonzero_bytes): Same.
5438 (handle_store): Same.
5439 (fold_strstr_to_strncmp): Same.
5440 (handle_integral_assign): Same.
5441 (check_and_optimize_stmt): Same.
5442 (class strlen_dom_walker): Replace evrp with ranger.
5443 (strlen_dom_walker::before_dom_children): Remove evrp.
5444 (strlen_dom_walker::after_dom_children): Remove evrp.
5445 * gimple-ssa-warn-access.cc (maybe_check_access_sizes):
5446 Restrict sprintf output.
5448 2021-11-09 Martin Liska <mliska@suse.cz>
5450 * genconditions.c (write_one_condition): Add const qualifier
5453 2021-11-09 Jeff Law <jeffreyalaw@gmail.com>
5455 * match.pd: New pattern to simplify (1 << n) & M ==/!= 0 for M
5458 2021-11-08 Bill Schmidt <wschmidt@linux.ibm.com>
5460 * config/rs6000/rs6000-builtin-new.def (VEC_INIT_V16QI): Use
5461 escape-newline support.
5462 (VEC_INIT_V4SI): Likewise.
5463 (VEC_INIT_V8HI): Likewise.
5464 (PACK_V1TI): Likewise.
5466 (VFIRSTMISMATCHOREOSINDEX_V16QI): Likewise.
5467 (VFIRSTMISMATCHOREOSINDEX_V8HI): Likewise.
5468 (VFIRSTMISMATCHOREOSINDEX_V4SI): Likewise.
5472 (FMAF128_ODD): Likewise.
5473 (VSCEQPUO): Likewise.
5475 (VSIEQPF): Likewise.
5476 (VSTDCQP): Likewise.
5477 (PACK_TD): Likewise.
5478 (TABORTDC): Likewise.
5479 (TABORTDCI): Likewise.
5480 (SE_LXVRBX): Likewise.
5481 (SE_LXVRHX): Likewise.
5482 (SE_LXVRWX): Likewise.
5483 (SE_LXVRDX): Likewise.
5484 (VREPLACE_UN_UV2DI): Likewise.
5485 (VREPLACE_UN_UV4SI): Likewise.
5486 (VREPLACE_UN_V2DI): Likewise.
5487 (VREPLACE_ELT_UV2DI): Likewise.
5488 (VREPLACE_ELT_V2DI): Likewise.
5489 (ZE_LXVRBX): Likewise.
5490 (ZE_LXVRHX): Likewise.
5491 (ZE_LXVRWX): Likewise.
5492 (ZE_LXVRDX): Likewise.
5494 (CNTLZDM): Likewise.
5495 (CNTTZDM): Likewise.
5498 (PMXVBF16GER2): Likewise.
5499 (PMXVBF16GER2_INTERNAL): Likewise.
5500 (PMXVBF16GER2NN): Likewise.
5501 (PMXVBF16GER2NN_INTERNAL): Likewise.
5502 (PMXVBF16GER2NP): Likewise.
5503 (PMXVBF16GER2NP_INTERNAL): Likewise.
5504 (PMXVBF16GER2PN): Likewise.
5505 (PMXVBF16GER2PN_INTERNAL): Likewise.
5506 (PMXVBF16GER2PP): Likewise.
5507 (PMXVBF16GER2PP_INTERNAL): Likewise.
5508 (PMXVF16GER2): Likewise.
5509 (PMXVF16GER2_INTERNAL): Likewise.
5510 (PMXVF16GER2NN): Likewise.
5511 (PMXVF16GER2NN_INTERNAL): Likewise.
5512 (PMXVF16GER2NP): Likewise.
5513 (PMXVF16GER2NP_INTERNAL): Likewise.
5514 (PMXVF16GER2PN): Likewise.
5515 (PMXVF16GER2PN_INTERNAL): Likewise.
5516 (PMXVF16GER2PP): Likewise.
5517 (PMXVF16GER2PP_INTERNAL): Likewise.
5518 (PMXVF32GER_INTERNAL): Likewise.
5519 (PMXVF32GERNN): Likewise.
5520 (PMXVF32GERNN_INTERNAL): Likewise.
5521 (PMXVF32GERNP): Likewise.
5522 (PMXVF32GERNP_INTERNAL): Likewise.
5523 (PMXVF32GERPN): Likewise.
5524 (PMXVF32GERPN_INTERNAL): Likewise.
5525 (PMXVF32GERPP): Likewise.
5526 (PMXVF32GERPP_INTERNAL): Likewise.
5527 (PMXVF64GER): Likewise.
5528 (PMXVF64GER_INTERNAL): Likewise.
5529 (PMXVF64GERNN): Likewise.
5530 (PMXVF64GERNN_INTERNAL): Likewise.
5531 (PMXVF64GERNP): Likewise.
5532 (PMXVF64GERNP_INTERNAL): Likewise.
5533 (PMXVF64GERPN): Likewise.
5534 (PMXVF64GERPN_INTERNAL): Likewise.
5535 (PMXVF64GERPP): Likewise.
5536 (PMXVF64GERPP_INTERNAL): Likewise.
5537 (PMXVI16GER2): Likewise.
5538 (PMXVI16GER2_INTERNAL): Likewise.
5539 (PMXVI16GER2PP): Likewise.
5540 (PMXVI16GER2PP_INTERNAL): Likewise.
5541 (PMXVI16GER2S): Likewise.
5542 (PMXVI16GER2S_INTERNAL): Likewise.
5543 (PMXVI16GER2SPP): Likewise.
5544 (PMXVI16GER2SPP_INTERNAL): Likewise.
5545 (PMXVI4GER8): Likewise.
5546 (PMXVI4GER8_INTERNAL): Likewise.
5547 (PMXVI4GER8PP): Likewise.
5548 (PMXVI4GER8PP_INTERNAL): Likewise.
5549 (PMXVI8GER4): Likewise.
5550 (PMXVI8GER4_INTERNAL): Likewise.
5551 (PMXVI8GER4PP): Likewise.
5552 (PMXVI8GER4PP_INTERNAL): Likewise.
5553 (PMXVI8GER4SPP): Likewise.
5554 (PMXVI8GER4SPP_INTERNAL): Likewise.
5555 * config/rs6000/rs6000-gen-builtins.c (MAXLINES): New macro.
5556 (linebuf): Increase size.
5557 (lines): New variable.
5558 (lastline): Likewise.
5559 (real_line_pos): New function.
5560 (diag): Change signature.
5561 (bif_diag): Change signature; support escape-newline handling.
5562 (ovld_diag): Likewise.
5563 (fatal): Move earlier.
5564 (consume_whitespace): Adjust diag call.
5565 (advance_line): Add escape-newline handling; call fatal.
5566 (safe_inc_pos): Adjust diag call.
5567 (match_identifier): Likewise.
5568 (match_integer): Likewise.
5569 (match_to_right_bracket): Call fatal instead of diag; adjust diag
5571 (match_basetype): Adjust diag calls.
5572 (match_bracketed_pair): Likewise.
5573 (match_const_restriction): Likewise.
5574 (match_type): Likewise.
5575 (parse_args): Likewise.
5576 (parse_bif_attrs): Likewise.
5577 (complete_vector_type): Likewise.
5578 (complete_base_type): Likewise.
5579 (parse_prototype): Likewise.
5580 (parse_bif_entry): Likewise.
5581 (parse_bif_stanza): Likewise.
5582 (parse_ovld_entry): Likewise.
5583 (parse_ovld_stanza): Likewise.
5584 (main): Allocate buffers for lines[].
5586 2021-11-08 Eric Botcazou <ebotcazou@adacore.com>
5588 * config/i386/i386.c (legitimize_pic_address): Adjust comment and
5589 use the REG argument on the CM_LARGE_PIC code path as well.
5590 * config/i386/predicates.md (gotoff_operand): Do not treat VxWorks
5591 specially with the large code models.
5593 2021-11-08 Jan Hubicka <hubicka@ucw.cz>
5595 * gimple.c (gimple_call_static_chain_flags): Revert the workaround
5596 allowing interposition since issues with binds_to_local_def were
5599 2021-11-08 Andrew MacLeod <amacleod@redhat.com>
5601 PR tree-optimization/103122
5602 * gimple-range.cc (gimple_ranger::range_of_expr): Request the cache
5603 entry with "calulate new values" set to false.
5605 2021-11-08 Jan Hubicka <hubicka@ucw.cz>
5607 * builtins.c (is_simple_builtin): Add builitin_dwarf_cfa
5608 and builtin_return_address.
5609 (builtin_fnspec): Annotate builtin_return,
5610 bulitin_eh_pointer, builtin_eh_filter, builtin_unwind_resume,
5611 builtin_cxa_end_cleanup, builtin_eh_copy_values,
5612 builtin_frame_address, builtin_apply_args,
5613 builtin_asan_before_dynamic_init, builtin_asan_after_dynamic_init,
5614 builtin_prefetch, builtin_dwarf_cfa, builtin_return_addrss
5616 * ipa-pure-const.c (special_builtin_state): Add builtin_dwarf_cfa
5617 and builtin_return_address.
5619 2021-11-08 Jan Hubicka <hubicka@ucw.cz>
5621 PR tree-optimization/103177
5622 * passes.def: Move uncprop after pure/const and modref.
5624 2021-11-08 Martin Jambor <mjambor@suse.cz>
5628 * tree-inline.c (remap_gimple_stmt): Unshare the expression without
5629 location before invoking remap_with_debug_expressions on it.
5630 * ipa-param-manipulation.c
5631 (ipa_param_body_adjustments::prepare_debug_expressions): Likewise.
5633 2021-11-08 David Edelsohn <dje.gcc@gmail.com>
5635 * config/rs6000/vsx.md (vsx_splat_v4si_di): Revert "wa"
5638 2021-11-08 Richard Biener <rguenther@suse.de>
5640 * cfgloop.c (verify_loop_structure): Use a temporary BB flag
5641 instead of an sbitmap to cache irreducible state.
5643 2021-11-08 Aldy Hernandez <aldyh@redhat.com>
5645 PR tree-optimization/103120
5646 * gimple-range-path.cc (path_range_query::range_defined_in_block):
5647 Bail if there's a cache entry.
5649 2021-11-08 Bill Schmidt <wschmidt@linux.ibm.com>
5651 * config/rs6000/rs6000.c (rs6000_builtin_reciprocal): Use
5652 rs6000_builtin_decls_x when appropriate.
5653 (add_condition_to_bb): Likewise.
5654 (rs6000_atomic_assign_expand_fenv): Likewise.
5656 2021-11-08 Bill Schmidt <wschmidt@linux.ibm.com>
5658 * config/rs6000/rs6000-call.c (rs6000_new_builtin_decl): New function.
5659 (rs6000_builtin_decl): Call it.
5661 2021-11-08 Martin Liska <mliska@suse.cz>
5663 * cgraph.c (cgraph_node::dump): Dump it from decl.
5665 2021-11-08 Martin Liska <mliska@suse.cz>
5667 PR gcov-profile/100520
5668 * coverage.c (coverage_compute_profile_id): Strip .gk when
5669 compare debug is used.
5670 * system.h (endswith): New function.
5672 2021-11-08 Martin Liska <mliska@suse.cz>
5674 * cgraph.c (cgraph_node::dump): Dump static_chain_decl.
5676 2021-11-08 Thomas Schwinge <thomas@codesourcery.com>
5678 * config/rs6000/rbtree.c: Fix 'Copyright (C) 2020-21' into '2020-2021'
5679 * config/rs6000/rbtree.h: Likewise.
5680 * config/rs6000/rs6000-builtin-new.def: Likewise.
5681 * config/rs6000/rs6000-gen-builtins.c: Likewise.
5682 * config/rs6000/rs6000-overload.def: Likewise.
5684 2021-11-08 Richard Sandiford <richard.sandiford@arm.com>
5686 * config/aarch64/aarch64.c (aarch64_ldn_stn_vectors): New function.
5687 (aarch64_address_cost): Use it instead of testing for CImode and
5690 2021-11-08 Richard Sandiford <richard.sandiford@arm.com>
5692 * genmodes.c (emit_insn_modes_h): Define NUM_MODE_* macros.
5693 * expmed.h (NUM_MODE_INT): Delete in favor of genmodes definitions.
5694 (NUM_MODE_PARTIAL_INT, NUM_MODE_VECTOR_INT): Likewise.
5695 * real.h (real_format_for_mode): Use NUM_MODE_FLOAT and
5696 NUM_MODE_DECIMAL_FLOAT.
5697 (REAL_MODE_FORMAT): Likewise.
5699 2021-11-08 Andre Vieira <andre.simoesdiasvieira@arm.com>
5701 * tree-vect-loop.c (vect_better_loop_vinfo_p): Change how epilogue loop
5704 2021-11-08 Richard Biener <rguenther@suse.de>
5706 * tree-vectorizer.h (vect_create_loop_vinfo): Add main_loop_info
5708 * tree-vect-loop.c (vect_create_loop_vinfo): Likewise. Set
5709 LOOP_VINFO_ORIG_LOOP_INFO and conditionalize set of
5710 LOOP_VINFO_NITERS_ASSUMPTIONS.
5711 (vect_analyze_loop_1): Adjust.
5712 (vect_analyze_loop): Move loop constraint setting and
5713 SCEV/niter reset here from vect_create_loop_vinfo to perform
5715 (vect_analyze_loop_form): Move dumping of symbolic niters
5716 here from vect_create_loop_vinfo.
5718 2021-11-08 Jan Hubicka <hubicka@ucw.cz>
5720 * ipa-modref-tree.h (enum modref_special_parms): New enum.
5721 (struct modref_access_node): update for special parms.
5722 (struct modref_ref_node): Likewise.
5723 (struct modref_parm_map): Likewise.
5724 (struct modref_tree): Likewise.
5725 * ipa-modref.c (dump_access): Likewise.
5726 (get_access): Detect static chain.
5727 (parm_map_for_arg): Take tree as arg instead of
5729 (merge_call_side_effects): Compute map for static chain.
5730 (process_fnspec): Update.
5731 (struct escape_point): Remove retslot_arg and static_chain_arg.
5732 (analyze_parms): Update.
5733 (compute_parm_map): Update.
5734 (propagate_unknown_call): Update.
5735 (modref_propagate_in_scc): Update.
5736 (modref_merge_call_site_flags): Update.
5737 (ipa_merge_modref_summary_after_inlining): Update.
5738 * tree-ssa-alias.c (modref_may_conflict): Handle static chain.
5739 * ipa-modref-tree.c (test_merge): Update.
5741 2021-11-08 Haochen Gui <guihaoc@gcc.gnu.org>
5743 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_builtin): Disable
5744 gimple fold for VSX_BUILTIN_XVMINDP, ALTIVEC_BUILTIN_VMINFP,
5745 VSX_BUILTIN_XVMAXDP, ALTIVEC_BUILTIN_VMAXFP when fast-math is not
5748 2021-11-08 liuhongt <hongtao.liu@intel.com>
5750 PR tree-optimization/103077
5751 * doc/invoke.texi (Options That Control Optimization):
5752 Update documentation for -ftree-loop-vectorize and
5753 -ftree-slp-vectorize which are enabled by default at -02.
5755 2021-11-08 liuhongt <hongtao.liu@intel.com>
5758 * match.pd (Simplifcation (trunc)copysign((extend)a, (extend)b)
5759 to .COPYSIGN (a, b)): Add !HONOR_SNANS.
5761 2021-11-08 liuhongt <hongtao.liu@intel.com>
5764 * match.pd: Simplify
5765 (trunc)fma ((extend)a, (extend)b, (extend)c) to IFN_FMA (a, b,
5766 c) under flag_unsafe_math_optimizations.
5768 2021-11-07 Jan Hubicka <hubicka@ucw.cz>
5771 * params.opt (modref-max-depth): Add range.
5772 (modref-max-adjustments): Fix range.
5774 2021-11-07 Aldy Hernandez <aldyh@redhat.com>
5776 * tree-pass.h (make_pass_vrp_threader): Remove.
5777 * tree-ssa-threadbackward.c
5778 (back_threader_profitability::profitable_path_p): Remove
5779 ASSERT_EXPR references.
5780 * tree-ssa-threadedge.c (jt_state::register_equivs_stmt): Same.
5781 * tree-vrp.c (vrp_folder::simplify_casted_conds): Same.
5782 (execute_vrp): Same.
5783 (class hybrid_threader): Remove.
5784 (hybrid_threader::hybrid_threader): Remove.
5785 (hybrid_threader::~hybrid_threader): Remove.
5786 (hybrid_threader::before_dom_children): Remove.
5787 (hybrid_threader::after_dom_children): Remove.
5788 (execute_vrp_threader): Remove.
5789 (class pass_vrp_threader): Remove.
5790 (make_pass_vrp_threader): Remove.
5792 2021-11-07 Jan Hubicka <hubicka@ucw.cz>
5794 * gimple.c (gimple_call_arg_flags): Use interposable_eaf_flags.
5795 (gimple_call_retslot_flags): Likewise.
5796 (gimple_call_static_chain_flags): Likewise.
5797 * ipa-modref.c (remove_useless_eaf_flags): Do not remove everything for
5799 (modref_summary::useful_p): Likewise.
5800 (modref_summary_lto::useful_p): Likewise.
5801 (analyze_parms): Do not give up on NOVOPS.
5802 (analyze_function): When dumping report chnages in EAF flags
5803 between IPA and local pass.
5804 (modref_merge_call_site_flags): Compute implicit eaf flags
5805 based on callee ecf_flags and fnspec; if the function does not
5806 bind to current defs use interposable_eaf_flags.
5807 (modref_propagate_flags_in_scc): Update.
5808 * ipa-modref.h (interposable_eaf_flags): New function.
5810 2021-11-07 Bill Schmidt <wschmidt@linux.ibm.com>
5812 * config/rs6000/rs6000-call.c (rs6000_expand_new_builtin): New
5814 (rs6000_invalid_new_builtin): New function.
5815 (rs6000_expand_builtin): Call rs6000_expand_new_builtin.
5816 (rs6000_expand_ldst_mask): New function.
5817 (new_cpu_expand_builtin): Likewise.
5818 (elemrev_icode): Likewise.
5819 (ldv_expand_builtin): Likewise.
5820 (lxvrse_expand_builtin): Likewise.
5821 (lxvrze_expand_builtin): Likewise.
5822 (stv_expand_builtin): Likewise.
5823 (new_mma_expand_builtin): Likewise.
5824 (new_htm_spr_num): Likewise.
5825 (new_htm_expand_builtin): Likewise.
5826 (rs6000_expand_new_builtin): Likewise.
5827 (rs6000_init_builtins): Initialize altivec_builtin_mask_for_load.
5829 2021-11-07 Jan Hubicka <hubicka@ucw.cz>
5831 * ipa-modref.c (modref_lattice): Add do_dataflow,
5832 changed and propagate_to fields.
5833 (modref_lattice::release): Free propagate_to
5834 (modref_lattice::merge): Do not give up early on unknown
5836 (modref_lattice::merge_deref): Likewise.
5837 (modref_eaf_analysis): Update toplevel comment.
5838 (modref_eaf_analysis::analyze_ssa_name): Record postponned ssa names;
5839 do optimistic dataflow initialization.
5840 (modref_eaf_analysis::merge_with_ssa_name): Build dataflow graph.
5841 (modref_eaf_analysis::propagate): New member function.
5842 (analyze_parms): Update to new API of modref_eaf_analysis.
5844 2021-11-06 Jan Hubicka <hubicka@ucw.cz>
5846 * cgraph.h (cgraph_node::can_be_discarded_p): Do not
5847 return true on functions from other partition.
5849 2021-11-06 David Edelsohn <dje.gcc@gmail.com>
5851 * config/rs6000/vsx.md (vsx_splat_v4si): Change constraints to "wa".
5852 (vsx_splat_v4si_di): Change constraint to "wa".
5854 2021-11-06 Aldy Hernandez <aldyh@redhat.com>
5856 PR tree-optimization/103061
5857 * value-relation.cc (path_oracle::path_oracle): Initialize
5859 (path_oracle::killing_def): Set m_killed_defs.
5860 (path_oracle::query_relation): Do not look at the root oracle for
5862 * value-relation.h (class path_oracle): Add m_killed_defs.
5864 2021-11-06 Aldy Hernandez <aldyh@redhat.com>
5866 * tree-ssa-threadbackward.c (back_threader::find_paths_to_names):
5867 Remove gotos and other cleanups.
5869 2021-11-05 Jan Hubicka <hubicka@ucw.cz>
5872 * ipa-modref-tree.h (modref_tree::insert): Do nothing for
5873 paradoxical and zero sized accesses.
5875 2021-11-05 Jan Hubicka <hubicka@ucw.cz>
5878 * ipa-modref-tree.h (struct modref_access_node): Avoid left shift
5881 2021-11-05 Iain Sandoe <iain@sandoe.co.uk>
5883 * config/darwin.h (ASM_GENERATE_INTERNAL_LABEL): Add LTRAMP
5884 to the list of symbol prefixes that must be made linker-
5887 2021-11-05 Iain Sandoe <iain@sandoe.co.uk>
5889 * config.host: Add support for aarch64-*-darwin.
5890 * config/aarch64/host-aarch64-darwin.c: New file.
5891 * config/aarch64/x-darwin: New file.
5893 2021-11-05 Andrew MacLeod <amacleod@redhat.com>
5895 PR tree-optimization/103093
5896 * gimple-range-gori.cc (range_def_chain::get_imports): Remove assert.
5898 2021-11-05 Andrew MacLeod <amacleod@redhat.com>
5900 PR tree-optimization/102943
5901 * gimple-range-cache.cc (class update_list): New.
5902 (update_list::add): Replace add_to_update.
5903 (update_list::pop): New.
5904 (ranger_cache::ranger_cache): Adjust.
5905 (ranger_cache::~ranger_cache): Adjust.
5906 (ranger_cache::add_to_update): Delete.
5907 (ranger_cache::propagate_cache): Adjust to new class.
5908 (ranger_cache::propagate_updated_value): Ditto.
5909 (ranger_cache::fill_block_cache): Ditto.
5910 * gimple-range-cache.h (class ranger_cache): Adjust to update class.
5912 2021-11-05 Richard Biener <rguenther@suse.de>
5914 * tree-vect-loop.c (vect_analyze_loop): Remove obsolete
5915 comment and expand on another one. Combine nested if.
5917 2021-11-05 John David Anglin <danglin@gcc.gnu.org>
5920 * config/pa/pa-modes.def: Add OImode integer type.
5921 * config/pa/pa.c (pa_scalar_mode_supported_p): Allow TImode
5923 * config/pa/pa.h (MIN_UNITS_PER_WORD) Define to MIN_UNITS_PER_WORD
5924 to UNITS_PER_WORD if IN_LIBGCC2.
5925 * config/pa/pa.md (addti3, addvti3, subti3, subvti3, negti2,
5926 negvti2, ashlti3, shrpd_internal): New patterns.
5927 Change some multi instruction types to multi.
5929 2021-11-05 Jakub Jelinek <jakub@redhat.com>
5932 * config/i386/i386.h
5933 (stringop_algs::stringop_strategy::stringop_strategy): Make the ctor
5936 2021-11-05 Wilco Dijkstra <wdijkstr@arm.com>
5939 * config/aarch64/aarch64.c (aarch64_mov_operand_p): Strip the salt
5941 * config/aarch64/constraints.md: Support const in Usw.
5943 2021-11-05 John David Anglin <danglin@gcc.gnu.org>
5945 * config/pa/pa.h (PREFERRED_DEBUGGING_TYPE): Define to DWARF2_DEBUG.
5946 * config/pa/pa64-hpux.h (PREFERRED_DEBUGGING_TYPE): Remove define.
5948 2021-11-05 Richard Biener <rguenther@suse.de>
5950 * tree-vectorizer.h (vec_info_shared::n_stmts): Add.
5951 (LOOP_VINFO_N_STMTS): Likewise.
5952 (vec_info_for_bb): Remove unused function.
5953 * tree-vectorizer.c (vec_info_shared::vec_info_shared):
5954 Initialize n_stmts member.
5955 * tree-vect-loop.c: Remove INCLUDE_FUNCTIONAL.
5956 (vect_create_loop_vinfo): Do not set loop->aux.
5957 (vect_analyze_loop_2): Do not get n_stmts as argument,
5958 instead use LOOP_VINFO_N_STMTS. Set LOOP_VINFO_VECTORIZABLE_P
5960 (vect_analyze_loop_1): Remove callback, get the mode iterator
5961 and autodetected_vector_mode as argument, advancing the
5962 iterator and initializing autodetected_vector_mode here.
5963 (vect_analyze_loop): Split analysis loop into two, first
5964 processing main loops only and then epilogues.
5966 2021-11-05 Martin Jambor <mjambor@suse.cz>
5968 * ipa-prop.c (compute_complex_assign_jump_func): Remove
5969 unnecessary check for RECORD_TYPE.
5971 2021-11-05 Gerald Pfeifer <gerald@pfeifer.com>
5973 * doc/install.texi2html: Do not generate old.html any longer.
5975 2021-11-05 Martin Liska <mliska@suse.cz>
5978 * opts.c (finish_options): Reset flag_gtoggle when it is used.
5980 2021-11-05 Jakub Jelinek <jakub@redhat.com>
5983 * dwarf2out.c (add_const_value_attribute): Add MODE argument, use it
5984 in CONST_WIDE_INT handling. Adjust recursive calls.
5985 (add_location_or_const_value_attribute): Pass DECL_MODE (decl) to
5986 new add_const_value_attribute argument.
5987 (tree_add_const_value_attribute): Pass TYPE_MODE (type) to new
5988 add_const_value_attribute argument.
5990 2021-11-05 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
5992 * config/vx-common.h: Test value of TARGET_VXWORKS7 rather
5995 2021-11-05 Richard Biener <rguenther@suse.de>
5997 * tree-vectorizer.h (struct vect_loop_form_info): New.
5998 (vect_analyze_loop_form): Adjust.
5999 (vect_create_loop_vinfo): New.
6000 * tree-parloops.c (gather_scalar_reductions): Adjust for
6001 vect_analyze_loop_form API change.
6002 * tree-vect-loop.c: Include <functional>.
6003 (vect_analyze_loop_form_1): Rename to vect_analyze_loop_form,
6004 take struct vect_loop_form_info as output parameter and adjust.
6005 (vect_analyze_loop_form): Rename to vect_create_loop_vinfo and
6006 split out call to the original vect_analyze_loop_form_1.
6007 (vect_reanalyze_as_main_loop): Rename to...
6008 (vect_analyze_loop_1): ... this, factor out the call to
6009 vect_analyze_loop_form and generalize to be able to use it twice ...
6010 (vect_analyze_loop): ... here. Perform vect_analyze_loop_form
6013 2021-11-05 Xionghu Luo <luoxhu@linux.ibm.com>
6016 * config/rs6000/fusion.md: Regenerate.
6017 * config/rs6000/genfusion.pl: Fix incorrect clobber constraint.
6019 2021-11-04 Andreas Krebbel <krebbel@linux.ibm.com>
6021 * config/s390/s390.h (STACK_CHECK_MOVING_SP): New macro
6024 2021-11-04 Tamar Christina <tamar.christina@arm.com>
6026 * config/aarch64/aarch64-builtins.c
6027 (aarch64_general_gimple_fold_builtin): Add ashl, sshl, ushl, ashr,
6028 ashr_simd, lshr, lshr_simd.
6029 * config/aarch64/aarch64-simd-builtins.def (lshr): Use USHIFTIMM.
6030 * config/aarch64/arm_neon.h (vshr_n_u8, vshr_n_u16, vshr_n_u32,
6031 vshrq_n_u8, vshrq_n_u16, vshrq_n_u32, vshrq_n_u64): Fix type hack.
6033 2021-11-04 Tamar Christina <tamar.christina@arm.com>
6035 * match.pd: New negate+shift pattern.
6037 2021-11-04 Andrew MacLeod <amacleod@redhat.com>
6039 PR tree-optimization/103079
6040 * gimple-range-gori.cc (gimple_range_calc_op1): Treat undefined as
6042 (gimple_range_calc_op2): Ditto.
6044 2021-11-04 Martin Jambor <mjambor@suse.cz>
6047 * ipa-param-manipulation.h (class ipa_param_body_adjustments): New
6048 members remap_with_debug_expressions, m_dead_ssa_debug_equiv,
6049 m_dead_stmt_debug_equiv and prepare_debug_expressions. Added
6050 parameter to mark_dead_statements.
6051 * ipa-param-manipulation.c: Include tree-phinodes.h and cfgexpand.h.
6052 (ipa_param_body_adjustments::mark_dead_statements): New parameter
6053 debugstack, push into it all SSA names used in debug statements,
6054 produce m_dead_ssa_debug_equiv mapping for the removed param.
6055 (replace_with_mapped_expr): New function.
6056 (ipa_param_body_adjustments::remap_with_debug_expressions): Likewise.
6057 (ipa_param_body_adjustments::prepare_debug_expressions): Likewise.
6058 (ipa_param_body_adjustments::common_initialization): Gather and
6059 procecc SSA which will be removed but are in debug statements. Simplify.
6060 (ipa_param_body_adjustments::ipa_param_body_adjustments): Initialize
6062 * tree-inline.c (remap_gimple_stmt): Create a debug bind when possible
6063 when avoiding a copy of an unnecessary statement. Remap removed SSA
6064 names in existing debug statements.
6065 (tree_function_versioning): Do not create DEBUG_EXPR_DECL for removed
6066 parameters if we have already done so.
6068 2021-11-04 Jan Hubicka <hubicka@ucw.cz>
6071 * gimple.c (gimple_call_static_chain_flags): Handle case when
6072 nested function does not bind locally.
6074 2021-11-04 Jonathan Wright <jonathan.wright@arm.com>
6076 * config/aarch64/aarch64.c (aarch64_function_value): Generate
6077 a register rtx for Neon vector-tuple modes.
6078 (aarch64_layout_arg): Likewise.
6080 2021-11-04 Jonathan Wright <jonathan.wright@arm.com>
6082 * lower-subreg.c (simple_move): Prevent decomposition if
6083 modes are not tieable.
6085 2021-11-04 Jonathan Wright <jonathan.wright@arm.com>
6086 Richard Sandiford <richard.sandiford@arm.com>
6088 * config/aarch64/aarch64-builtins.c (v2x8qi_UP): Define.
6089 (v2x4hi_UP): Likewise.
6090 (v2x4hf_UP): Likewise.
6091 (v2x4bf_UP): Likewise.
6092 (v2x2si_UP): Likewise.
6093 (v2x2sf_UP): Likewise.
6094 (v2x1di_UP): Likewise.
6095 (v2x1df_UP): Likewise.
6096 (v2x16qi_UP): Likewise.
6097 (v2x8hi_UP): Likewise.
6098 (v2x8hf_UP): Likewise.
6099 (v2x8bf_UP): Likewise.
6100 (v2x4si_UP): Likewise.
6101 (v2x4sf_UP): Likewise.
6102 (v2x2di_UP): Likewise.
6103 (v2x2df_UP): Likewise.
6104 (v3x8qi_UP): Likewise.
6105 (v3x4hi_UP): Likewise.
6106 (v3x4hf_UP): Likewise.
6107 (v3x4bf_UP): Likewise.
6108 (v3x2si_UP): Likewise.
6109 (v3x2sf_UP): Likewise.
6110 (v3x1di_UP): Likewise.
6111 (v3x1df_UP): Likewise.
6112 (v3x16qi_UP): Likewise.
6113 (v3x8hi_UP): Likewise.
6114 (v3x8hf_UP): Likewise.
6115 (v3x8bf_UP): Likewise.
6116 (v3x4si_UP): Likewise.
6117 (v3x4sf_UP): Likewise.
6118 (v3x2di_UP): Likewise.
6119 (v3x2df_UP): Likewise.
6120 (v4x8qi_UP): Likewise.
6121 (v4x4hi_UP): Likewise.
6122 (v4x4hf_UP): Likewise.
6123 (v4x4bf_UP): Likewise.
6124 (v4x2si_UP): Likewise.
6125 (v4x2sf_UP): Likewise.
6126 (v4x1di_UP): Likewise.
6127 (v4x1df_UP): Likewise.
6128 (v4x16qi_UP): Likewise.
6129 (v4x8hi_UP): Likewise.
6130 (v4x8hf_UP): Likewise.
6131 (v4x8bf_UP): Likewise.
6132 (v4x4si_UP): Likewise.
6133 (v4x4sf_UP): Likewise.
6134 (v4x2di_UP): Likewise.
6135 (v4x2df_UP): Likewise.
6136 (TYPES_GETREGP): Delete.
6137 (TYPES_SETREGP): Likewise.
6138 (TYPES_LOADSTRUCT_U): Define.
6139 (TYPES_LOADSTRUCT_P): Likewise.
6140 (TYPES_LOADSTRUCT_LANE_U): Likewise.
6141 (TYPES_LOADSTRUCT_LANE_P): Likewise.
6142 (TYPES_STORE1P): Move for consistency.
6143 (TYPES_STORESTRUCT_U): Define.
6144 (TYPES_STORESTRUCT_P): Likewise.
6145 (TYPES_STORESTRUCT_LANE_U): Likewise.
6146 (TYPES_STORESTRUCT_LANE_P): Likewise.
6147 (aarch64_simd_tuple_types): Define.
6148 (aarch64_lookup_simd_builtin_type): Handle tuple type lookup.
6149 (aarch64_init_simd_builtin_functions): Update frontend lookup
6150 for builtin functions after handling arm_neon.h pragma.
6151 (register_tuple_type): Manually set modes of single-integer
6152 tuple types. Record tuple types.
6153 * config/aarch64/aarch64-modes.def
6154 (ADV_SIMD_D_REG_STRUCT_MODES): Define D-register tuple modes.
6155 (ADV_SIMD_Q_REG_STRUCT_MODES): Define Q-register tuple modes.
6156 (SVE_MODES): Give single-vector modes priority over vector-
6158 (VECTOR_MODES_WITH_PREFIX): Set partial-vector mode order to
6159 be after all single-vector modes.
6160 * config/aarch64/aarch64-simd-builtins.def: Update builtin
6161 generator macros to reflect modifications to the backend
6163 * config/aarch64/aarch64-simd.md (aarch64_simd_ld2<mode>):
6164 Use vector-tuple mode iterator and rename to...
6165 (aarch64_simd_ld2<vstruct_elt>): This.
6166 (aarch64_simd_ld2r<mode>): Use vector-tuple mode iterator and
6168 (aarch64_simd_ld2r<vstruct_elt>): This.
6169 (aarch64_vec_load_lanesoi_lane<mode>): Use vector-tuple mode
6170 iterator and rename to...
6171 (aarch64_vec_load_lanes<mode>_lane<vstruct_elt>): This.
6172 (vec_load_lanesoi<mode>): Use vector-tuple mode iterator and
6174 (vec_load_lanes<mode><vstruct_elt>): This.
6175 (aarch64_simd_st2<mode>): Use vector-tuple mode iterator and
6177 (aarch64_simd_st2<vstruct_elt>): This.
6178 (aarch64_vec_store_lanesoi_lane<mode>): Use vector-tuple mode
6179 iterator and rename to...
6180 (aarch64_vec_store_lanes<mode>_lane<vstruct_elt>): This.
6181 (vec_store_lanesoi<mode>): Use vector-tuple mode iterator and
6183 (vec_store_lanes<mode><vstruct_elt>): This.
6184 (aarch64_simd_ld3<mode>): Use vector-tuple mode iterator and
6186 (aarch64_simd_ld3<vstruct_elt>): This.
6187 (aarch64_simd_ld3r<mode>): Use vector-tuple mode iterator and
6189 (aarch64_simd_ld3r<vstruct_elt>): This.
6190 (aarch64_vec_load_lanesci_lane<mode>): Use vector-tuple mode
6191 iterator and rename to...
6192 (vec_load_lanesci<mode>): This.
6193 (aarch64_simd_st3<mode>): Use vector-tuple mode iterator and
6195 (aarch64_simd_st3<vstruct_elt>): This.
6196 (aarch64_vec_store_lanesci_lane<mode>): Use vector-tuple mode
6197 iterator and rename to...
6198 (vec_store_lanesci<mode>): This.
6199 (aarch64_simd_ld4<mode>): Use vector-tuple mode iterator and
6201 (aarch64_simd_ld4<vstruct_elt>): This.
6202 (aarch64_simd_ld4r<mode>): Use vector-tuple mode iterator and
6204 (aarch64_simd_ld4r<vstruct_elt>): This.
6205 (aarch64_vec_load_lanesxi_lane<mode>): Use vector-tuple mode
6206 iterator and rename to...
6207 (vec_load_lanesxi<mode>): This.
6208 (aarch64_simd_st4<mode>): Use vector-tuple mode iterator and
6210 (aarch64_simd_st4<vstruct_elt>): This.
6211 (aarch64_vec_store_lanesxi_lane<mode>): Use vector-tuple mode
6212 iterator and rename to...
6213 (vec_store_lanesxi<mode>): This.
6214 (mov<mode>): Define for Neon vector-tuple modes.
6215 (aarch64_ld1x3<VALLDIF:mode>): Use vector-tuple mode iterator
6217 (aarch64_ld1x3<vstruct_elt>): This.
6218 (aarch64_ld1_x3_<mode>): Use vector-tuple mode iterator and
6220 (aarch64_ld1_x3_<vstruct_elt>): This.
6221 (aarch64_ld1x4<VALLDIF:mode>): Use vector-tuple mode iterator
6223 (aarch64_ld1x4<vstruct_elt>): This.
6224 (aarch64_ld1_x4_<mode>): Use vector-tuple mode iterator and
6226 (aarch64_ld1_x4_<vstruct_elt>): This.
6227 (aarch64_st1x2<VALLDIF:mode>): Use vector-tuple mode iterator
6229 (aarch64_st1x2<vstruct_elt>): This.
6230 (aarch64_st1_x2_<mode>): Use vector-tuple mode iterator and
6232 (aarch64_st1_x2_<vstruct_elt>): This.
6233 (aarch64_st1x3<VALLDIF:mode>): Use vector-tuple mode iterator
6235 (aarch64_st1x3<vstruct_elt>): This.
6236 (aarch64_st1_x3_<mode>): Use vector-tuple mode iterator and
6238 (aarch64_st1_x3_<vstruct_elt>): This.
6239 (aarch64_st1x4<VALLDIF:mode>): Use vector-tuple mode iterator
6241 (aarch64_st1x4<vstruct_elt>): This.
6242 (aarch64_st1_x4_<mode>): Use vector-tuple mode iterator and
6244 (aarch64_st1_x4_<vstruct_elt>): This.
6245 (*aarch64_mov<mode>): Define for vector-tuple modes.
6246 (*aarch64_be_mov<mode>): Likewise.
6247 (aarch64_ld<VSTRUCT:nregs>r<VALLDIF:mode>): Use vector-tuple
6248 mode iterator and rename to...
6249 (aarch64_ld<nregs>r<vstruct_elt>): This.
6250 (aarch64_ld2<mode>_dreg): Use vector-tuple mode iterator and
6252 (aarch64_ld2<vstruct_elt>_dreg): This.
6253 (aarch64_ld3<mode>_dreg): Use vector-tuple mode iterator and
6255 (aarch64_ld3<vstruct_elt>_dreg): This.
6256 (aarch64_ld4<mode>_dreg): Use vector-tuple mode iterator and
6258 (aarch64_ld4<vstruct_elt>_dreg): This.
6259 (aarch64_ld<VSTRUCT:nregs><VDC:mode>): Use vector-tuple mode
6260 iterator and rename to...
6261 (aarch64_ld<nregs><vstruct_elt>): Use vector-tuple mode
6262 iterator and rename to...
6263 (aarch64_ld<VSTRUCT:nregs><VQ:mode>): Use vector-tuple mode
6264 (aarch64_ld1x2<VQ:mode>): Delete.
6265 (aarch64_ld1x2<VDC:mode>): Use vector-tuple mode iterator and
6267 (aarch64_ld1x2<vstruct_elt>): This.
6268 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use vector-
6269 tuple mode iterator and rename to...
6270 (aarch64_ld<nregs>_lane<vstruct_elt>): This.
6271 (aarch64_get_dreg<VSTRUCT:mode><VDC:mode>): Delete.
6272 (aarch64_get_qreg<VSTRUCT:mode><VQ:mode>): Likewise.
6273 (aarch64_st2<mode>_dreg): Use vector-tuple mode iterator and
6275 (aarch64_st2<vstruct_elt>_dreg): This.
6276 (aarch64_st3<mode>_dreg): Use vector-tuple mode iterator and
6278 (aarch64_st3<vstruct_elt>_dreg): This.
6279 (aarch64_st4<mode>_dreg): Use vector-tuple mode iterator and
6281 (aarch64_st4<vstruct_elt>_dreg): This.
6282 (aarch64_st<VSTRUCT:nregs><VDC:mode>): Use vector-tuple mode
6283 iterator and rename to...
6284 (aarch64_st<nregs><vstruct_elt>): This.
6285 (aarch64_st<VSTRUCT:nregs><VQ:mode>): Use vector-tuple mode
6286 iterator and rename to aarch64_st<nregs><vstruct_elt>.
6287 (aarch64_st<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use vector-
6288 tuple mode iterator and rename to...
6289 (aarch64_st<nregs>_lane<vstruct_elt>): This.
6290 (aarch64_set_qreg<VSTRUCT:mode><VQ:mode>): Delete.
6291 (aarch64_simd_ld1<mode>_x2): Use vector-tuple mode iterator
6293 (aarch64_simd_ld1<vstruct_elt>_x2): This.
6294 * config/aarch64/aarch64.c (aarch64_advsimd_struct_mode_p):
6295 Refactor to include new vector-tuple modes.
6296 (aarch64_classify_vector_mode): Add cases for new vector-
6298 (aarch64_advsimd_partial_struct_mode_p): Define.
6299 (aarch64_advsimd_full_struct_mode_p): Likewise.
6300 (aarch64_advsimd_vector_array_mode): Likewise.
6301 (aarch64_sve_data_mode): Change location in file.
6302 (aarch64_array_mode): Handle case of Neon vector-tuple modes.
6303 (aarch64_hard_regno_nregs): Handle case of partial Neon
6305 (aarch64_classify_address): Refactor to include handling of
6306 Neon vector-tuple modes.
6307 (aarch64_print_operand): Print "d" for "%R" for a partial
6308 Neon vector structure.
6309 (aarch64_expand_vec_perm_1): Use new vector-tuple mode.
6310 (aarch64_modes_tieable_p): Prevent tieing Neon partial struct
6311 modes with scalar machines modes larger than 8 bytes.
6312 (aarch64_can_change_mode_class): Don't allow changes between
6313 partial and full Neon vector-structure modes.
6314 * config/aarch64/arm_neon.h (vst2_lane_f16): Use updated
6315 builtin and remove boiler-plate code for opaque mode.
6316 (vst2_lane_f32): Likewise.
6317 (vst2_lane_f64): Likewise.
6318 (vst2_lane_p8): Likewise.
6319 (vst2_lane_p16): Likewise.
6320 (vst2_lane_p64): Likewise.
6321 (vst2_lane_s8): Likewise.
6322 (vst2_lane_s16): Likewise.
6323 (vst2_lane_s32): Likewise.
6324 (vst2_lane_s64): Likewise.
6325 (vst2_lane_u8): Likewise.
6326 (vst2_lane_u16): Likewise.
6327 (vst2_lane_u32): Likewise.
6328 (vst2_lane_u64): Likewise.
6329 (vst2q_lane_f16): Likewise.
6330 (vst2q_lane_f32): Likewise.
6331 (vst2q_lane_f64): Likewise.
6332 (vst2q_lane_p8): Likewise.
6333 (vst2q_lane_p16): Likewise.
6334 (vst2q_lane_p64): Likewise.
6335 (vst2q_lane_s8): Likewise.
6336 (vst2q_lane_s16): Likewise.
6337 (vst2q_lane_s32): Likewise.
6338 (vst2q_lane_s64): Likewise.
6339 (vst2q_lane_u8): Likewise.
6340 (vst2q_lane_u16): Likewise.
6341 (vst2q_lane_u32): Likewise.
6342 (vst2q_lane_u64): Likewise.
6343 (vst3_lane_f16): Likewise.
6344 (vst3_lane_f32): Likewise.
6345 (vst3_lane_f64): Likewise.
6346 (vst3_lane_p8): Likewise.
6347 (vst3_lane_p16): Likewise.
6348 (vst3_lane_p64): Likewise.
6349 (vst3_lane_s8): Likewise.
6350 (vst3_lane_s16): Likewise.
6351 (vst3_lane_s32): Likewise.
6352 (vst3_lane_s64): Likewise.
6353 (vst3_lane_u8): Likewise.
6354 (vst3_lane_u16): Likewise.
6355 (vst3_lane_u32): Likewise.
6356 (vst3_lane_u64): Likewise.
6357 (vst3q_lane_f16): Likewise.
6358 (vst3q_lane_f32): Likewise.
6359 (vst3q_lane_f64): Likewise.
6360 (vst3q_lane_p8): Likewise.
6361 (vst3q_lane_p16): Likewise.
6362 (vst3q_lane_p64): Likewise.
6363 (vst3q_lane_s8): Likewise.
6364 (vst3q_lane_s16): Likewise.
6365 (vst3q_lane_s32): Likewise.
6366 (vst3q_lane_s64): Likewise.
6367 (vst3q_lane_u8): Likewise.
6368 (vst3q_lane_u16): Likewise.
6369 (vst3q_lane_u32): Likewise.
6370 (vst3q_lane_u64): Likewise.
6371 (vst4_lane_f16): Likewise.
6372 (vst4_lane_f32): Likewise.
6373 (vst4_lane_f64): Likewise.
6374 (vst4_lane_p8): Likewise.
6375 (vst4_lane_p16): Likewise.
6376 (vst4_lane_p64): Likewise.
6377 (vst4_lane_s8): Likewise.
6378 (vst4_lane_s16): Likewise.
6379 (vst4_lane_s32): Likewise.
6380 (vst4_lane_s64): Likewise.
6381 (vst4_lane_u8): Likewise.
6382 (vst4_lane_u16): Likewise.
6383 (vst4_lane_u32): Likewise.
6384 (vst4_lane_u64): Likewise.
6385 (vst4q_lane_f16): Likewise.
6386 (vst4q_lane_f32): Likewise.
6387 (vst4q_lane_f64): Likewise.
6388 (vst4q_lane_p8): Likewise.
6389 (vst4q_lane_p16): Likewise.
6390 (vst4q_lane_p64): Likewise.
6391 (vst4q_lane_s8): Likewise.
6392 (vst4q_lane_s16): Likewise.
6393 (vst4q_lane_s32): Likewise.
6394 (vst4q_lane_s64): Likewise.
6395 (vst4q_lane_u8): Likewise.
6396 (vst4q_lane_u16): Likewise.
6397 (vst4q_lane_u32): Likewise.
6398 (vst4q_lane_u64): Likewise.
6399 (vtbl3_s8): Likewise.
6400 (vtbl3_u8): Likewise.
6401 (vtbl3_p8): Likewise.
6402 (vtbl4_s8): Likewise.
6403 (vtbl4_u8): Likewise.
6404 (vtbl4_p8): Likewise.
6405 (vld1_u8_x3): Likewise.
6406 (vld1_s8_x3): Likewise.
6407 (vld1_u16_x3): Likewise.
6408 (vld1_s16_x3): Likewise.
6409 (vld1_u32_x3): Likewise.
6410 (vld1_s32_x3): Likewise.
6411 (vld1_u64_x3): Likewise.
6412 (vld1_s64_x3): Likewise.
6413 (vld1_f16_x3): Likewise.
6414 (vld1_f32_x3): Likewise.
6415 (vld1_f64_x3): Likewise.
6416 (vld1_p8_x3): Likewise.
6417 (vld1_p16_x3): Likewise.
6418 (vld1_p64_x3): Likewise.
6419 (vld1q_u8_x3): Likewise.
6420 (vld1q_s8_x3): Likewise.
6421 (vld1q_u16_x3): Likewise.
6422 (vld1q_s16_x3): Likewise.
6423 (vld1q_u32_x3): Likewise.
6424 (vld1q_s32_x3): Likewise.
6425 (vld1q_u64_x3): Likewise.
6426 (vld1q_s64_x3): Likewise.
6427 (vld1q_f16_x3): Likewise.
6428 (vld1q_f32_x3): Likewise.
6429 (vld1q_f64_x3): Likewise.
6430 (vld1q_p8_x3): Likewise.
6431 (vld1q_p16_x3): Likewise.
6432 (vld1q_p64_x3): Likewise.
6433 (vld1_u8_x2): Likewise.
6434 (vld1_s8_x2): Likewise.
6435 (vld1_u16_x2): Likewise.
6436 (vld1_s16_x2): Likewise.
6437 (vld1_u32_x2): Likewise.
6438 (vld1_s32_x2): Likewise.
6439 (vld1_u64_x2): Likewise.
6440 (vld1_s64_x2): Likewise.
6441 (vld1_f16_x2): Likewise.
6442 (vld1_f32_x2): Likewise.
6443 (vld1_f64_x2): Likewise.
6444 (vld1_p8_x2): Likewise.
6445 (vld1_p16_x2): Likewise.
6446 (vld1_p64_x2): Likewise.
6447 (vld1q_u8_x2): Likewise.
6448 (vld1q_s8_x2): Likewise.
6449 (vld1q_u16_x2): Likewise.
6450 (vld1q_s16_x2): Likewise.
6451 (vld1q_u32_x2): Likewise.
6452 (vld1q_s32_x2): Likewise.
6453 (vld1q_u64_x2): Likewise.
6454 (vld1q_s64_x2): Likewise.
6455 (vld1q_f16_x2): Likewise.
6456 (vld1q_f32_x2): Likewise.
6457 (vld1q_f64_x2): Likewise.
6458 (vld1q_p8_x2): Likewise.
6459 (vld1q_p16_x2): Likewise.
6460 (vld1q_p64_x2): Likewise.
6461 (vld1_s8_x4): Likewise.
6462 (vld1q_s8_x4): Likewise.
6463 (vld1_s16_x4): Likewise.
6464 (vld1q_s16_x4): Likewise.
6465 (vld1_s32_x4): Likewise.
6466 (vld1q_s32_x4): Likewise.
6467 (vld1_u8_x4): Likewise.
6468 (vld1q_u8_x4): Likewise.
6469 (vld1_u16_x4): Likewise.
6470 (vld1q_u16_x4): Likewise.
6471 (vld1_u32_x4): Likewise.
6472 (vld1q_u32_x4): Likewise.
6473 (vld1_f16_x4): Likewise.
6474 (vld1q_f16_x4): Likewise.
6475 (vld1_f32_x4): Likewise.
6476 (vld1q_f32_x4): Likewise.
6477 (vld1_p8_x4): Likewise.
6478 (vld1q_p8_x4): Likewise.
6479 (vld1_p16_x4): Likewise.
6480 (vld1q_p16_x4): Likewise.
6481 (vld1_s64_x4): Likewise.
6482 (vld1_u64_x4): Likewise.
6483 (vld1_p64_x4): Likewise.
6484 (vld1q_s64_x4): Likewise.
6485 (vld1q_u64_x4): Likewise.
6486 (vld1q_p64_x4): Likewise.
6487 (vld1_f64_x4): Likewise.
6488 (vld1q_f64_x4): Likewise.
6489 (vld2_s64): Likewise.
6490 (vld2_u64): Likewise.
6491 (vld2_f64): Likewise.
6492 (vld2_s8): Likewise.
6493 (vld2_p8): Likewise.
6494 (vld2_p64): Likewise.
6495 (vld2_s16): Likewise.
6496 (vld2_p16): Likewise.
6497 (vld2_s32): Likewise.
6498 (vld2_u8): Likewise.
6499 (vld2_u16): Likewise.
6500 (vld2_u32): Likewise.
6501 (vld2_f16): Likewise.
6502 (vld2_f32): Likewise.
6503 (vld2q_s8): Likewise.
6504 (vld2q_p8): Likewise.
6505 (vld2q_s16): Likewise.
6506 (vld2q_p16): Likewise.
6507 (vld2q_p64): Likewise.
6508 (vld2q_s32): Likewise.
6509 (vld2q_s64): Likewise.
6510 (vld2q_u8): Likewise.
6511 (vld2q_u16): Likewise.
6512 (vld2q_u32): Likewise.
6513 (vld2q_u64): Likewise.
6514 (vld2q_f16): Likewise.
6515 (vld2q_f32): Likewise.
6516 (vld2q_f64): Likewise.
6517 (vld3_s64): Likewise.
6518 (vld3_u64): Likewise.
6519 (vld3_f64): Likewise.
6520 (vld3_s8): Likewise.
6521 (vld3_p8): Likewise.
6522 (vld3_s16): Likewise.
6523 (vld3_p16): Likewise.
6524 (vld3_s32): Likewise.
6525 (vld3_u8): Likewise.
6526 (vld3_u16): Likewise.
6527 (vld3_u32): Likewise.
6528 (vld3_f16): Likewise.
6529 (vld3_f32): Likewise.
6530 (vld3_p64): Likewise.
6531 (vld3q_s8): Likewise.
6532 (vld3q_p8): Likewise.
6533 (vld3q_s16): Likewise.
6534 (vld3q_p16): Likewise.
6535 (vld3q_s32): Likewise.
6536 (vld3q_s64): Likewise.
6537 (vld3q_u8): Likewise.
6538 (vld3q_u16): Likewise.
6539 (vld3q_u32): Likewise.
6540 (vld3q_u64): Likewise.
6541 (vld3q_f16): Likewise.
6542 (vld3q_f32): Likewise.
6543 (vld3q_f64): Likewise.
6544 (vld3q_p64): Likewise.
6545 (vld4_s64): Likewise.
6546 (vld4_u64): Likewise.
6547 (vld4_f64): Likewise.
6548 (vld4_s8): Likewise.
6549 (vld4_p8): Likewise.
6550 (vld4_s16): Likewise.
6551 (vld4_p16): Likewise.
6552 (vld4_s32): Likewise.
6553 (vld4_u8): Likewise.
6554 (vld4_u16): Likewise.
6555 (vld4_u32): Likewise.
6556 (vld4_f16): Likewise.
6557 (vld4_f32): Likewise.
6558 (vld4_p64): Likewise.
6559 (vld4q_s8): Likewise.
6560 (vld4q_p8): Likewise.
6561 (vld4q_s16): Likewise.
6562 (vld4q_p16): Likewise.
6563 (vld4q_s32): Likewise.
6564 (vld4q_s64): Likewise.
6565 (vld4q_u8): Likewise.
6566 (vld4q_u16): Likewise.
6567 (vld4q_u32): Likewise.
6568 (vld4q_u64): Likewise.
6569 (vld4q_f16): Likewise.
6570 (vld4q_f32): Likewise.
6571 (vld4q_f64): Likewise.
6572 (vld4q_p64): Likewise.
6573 (vld2_dup_s8): Likewise.
6574 (vld2_dup_s16): Likewise.
6575 (vld2_dup_s32): Likewise.
6576 (vld2_dup_f16): Likewise.
6577 (vld2_dup_f32): Likewise.
6578 (vld2_dup_f64): Likewise.
6579 (vld2_dup_u8): Likewise.
6580 (vld2_dup_u16): Likewise.
6581 (vld2_dup_u32): Likewise.
6582 (vld2_dup_p8): Likewise.
6583 (vld2_dup_p16): Likewise.
6584 (vld2_dup_p64): Likewise.
6585 (vld2_dup_s64): Likewise.
6586 (vld2_dup_u64): Likewise.
6587 (vld2q_dup_s8): Likewise.
6588 (vld2q_dup_p8): Likewise.
6589 (vld2q_dup_s16): Likewise.
6590 (vld2q_dup_p16): Likewise.
6591 (vld2q_dup_s32): Likewise.
6592 (vld2q_dup_s64): Likewise.
6593 (vld2q_dup_u8): Likewise.
6594 (vld2q_dup_u16): Likewise.
6595 (vld2q_dup_u32): Likewise.
6596 (vld2q_dup_u64): Likewise.
6597 (vld2q_dup_f16): Likewise.
6598 (vld2q_dup_f32): Likewise.
6599 (vld2q_dup_f64): Likewise.
6600 (vld2q_dup_p64): Likewise.
6601 (vld3_dup_s64): Likewise.
6602 (vld3_dup_u64): Likewise.
6603 (vld3_dup_f64): Likewise.
6604 (vld3_dup_s8): Likewise.
6605 (vld3_dup_p8): Likewise.
6606 (vld3_dup_s16): Likewise.
6607 (vld3_dup_p16): Likewise.
6608 (vld3_dup_s32): Likewise.
6609 (vld3_dup_u8): Likewise.
6610 (vld3_dup_u16): Likewise.
6611 (vld3_dup_u32): Likewise.
6612 (vld3_dup_f16): Likewise.
6613 (vld3_dup_f32): Likewise.
6614 (vld3_dup_p64): Likewise.
6615 (vld3q_dup_s8): Likewise.
6616 (vld3q_dup_p8): Likewise.
6617 (vld3q_dup_s16): Likewise.
6618 (vld3q_dup_p16): Likewise.
6619 (vld3q_dup_s32): Likewise.
6620 (vld3q_dup_s64): Likewise.
6621 (vld3q_dup_u8): Likewise.
6622 (vld3q_dup_u16): Likewise.
6623 (vld3q_dup_u32): Likewise.
6624 (vld3q_dup_u64): Likewise.
6625 (vld3q_dup_f16): Likewise.
6626 (vld3q_dup_f32): Likewise.
6627 (vld3q_dup_f64): Likewise.
6628 (vld3q_dup_p64): Likewise.
6629 (vld4_dup_s64): Likewise.
6630 (vld4_dup_u64): Likewise.
6631 (vld4_dup_f64): Likewise.
6632 (vld4_dup_s8): Likewise.
6633 (vld4_dup_p8): Likewise.
6634 (vld4_dup_s16): Likewise.
6635 (vld4_dup_p16): Likewise.
6636 (vld4_dup_s32): Likewise.
6637 (vld4_dup_u8): Likewise.
6638 (vld4_dup_u16): Likewise.
6639 (vld4_dup_u32): Likewise.
6640 (vld4_dup_f16): Likewise.
6641 (vld4_dup_f32): Likewise.
6642 (vld4_dup_p64): Likewise.
6643 (vld4q_dup_s8): Likewise.
6644 (vld4q_dup_p8): Likewise.
6645 (vld4q_dup_s16): Likewise.
6646 (vld4q_dup_p16): Likewise.
6647 (vld4q_dup_s32): Likewise.
6648 (vld4q_dup_s64): Likewise.
6649 (vld4q_dup_u8): Likewise.
6650 (vld4q_dup_u16): Likewise.
6651 (vld4q_dup_u32): Likewise.
6652 (vld4q_dup_u64): Likewise.
6653 (vld4q_dup_f16): Likewise.
6654 (vld4q_dup_f32): Likewise.
6655 (vld4q_dup_f64): Likewise.
6656 (vld4q_dup_p64): Likewise.
6657 (vld2_lane_u8): Likewise.
6658 (vld2_lane_u16): Likewise.
6659 (vld2_lane_u32): Likewise.
6660 (vld2_lane_u64): Likewise.
6661 (vld2_lane_s8): Likewise.
6662 (vld2_lane_s16): Likewise.
6663 (vld2_lane_s32): Likewise.
6664 (vld2_lane_s64): Likewise.
6665 (vld2_lane_f16): Likewise.
6666 (vld2_lane_f32): Likewise.
6667 (vld2_lane_f64): Likewise.
6668 (vld2_lane_p8): Likewise.
6669 (vld2_lane_p16): Likewise.
6670 (vld2_lane_p64): Likewise.
6671 (vld2q_lane_u8): Likewise.
6672 (vld2q_lane_u16): Likewise.
6673 (vld2q_lane_u32): Likewise.
6674 (vld2q_lane_u64): Likewise.
6675 (vld2q_lane_s8): Likewise.
6676 (vld2q_lane_s16): Likewise.
6677 (vld2q_lane_s32): Likewise.
6678 (vld2q_lane_s64): Likewise.
6679 (vld2q_lane_f16): Likewise.
6680 (vld2q_lane_f32): Likewise.
6681 (vld2q_lane_f64): Likewise.
6682 (vld2q_lane_p8): Likewise.
6683 (vld2q_lane_p16): Likewise.
6684 (vld2q_lane_p64): Likewise.
6685 (vld3_lane_u8): Likewise.
6686 (vld3_lane_u16): Likewise.
6687 (vld3_lane_u32): Likewise.
6688 (vld3_lane_u64): Likewise.
6689 (vld3_lane_s8): Likewise.
6690 (vld3_lane_s16): Likewise.
6691 (vld3_lane_s32): Likewise.
6692 (vld3_lane_s64): Likewise.
6693 (vld3_lane_f16): Likewise.
6694 (vld3_lane_f32): Likewise.
6695 (vld3_lane_f64): Likewise.
6696 (vld3_lane_p8): Likewise.
6697 (vld3_lane_p16): Likewise.
6698 (vld3_lane_p64): Likewise.
6699 (vld3q_lane_u8): Likewise.
6700 (vld3q_lane_u16): Likewise.
6701 (vld3q_lane_u32): Likewise.
6702 (vld3q_lane_u64): Likewise.
6703 (vld3q_lane_s8): Likewise.
6704 (vld3q_lane_s16): Likewise.
6705 (vld3q_lane_s32): Likewise.
6706 (vld3q_lane_s64): Likewise.
6707 (vld3q_lane_f16): Likewise.
6708 (vld3q_lane_f32): Likewise.
6709 (vld3q_lane_f64): Likewise.
6710 (vld3q_lane_p8): Likewise.
6711 (vld3q_lane_p16): Likewise.
6712 (vld3q_lane_p64): Likewise.
6713 (vld4_lane_u8): Likewise.
6714 (vld4_lane_u16): Likewise.
6715 (vld4_lane_u32): Likewise.
6716 (vld4_lane_u64): Likewise.
6717 (vld4_lane_s8): Likewise.
6718 (vld4_lane_s16): Likewise.
6719 (vld4_lane_s32): Likewise.
6720 (vld4_lane_s64): Likewise.
6721 (vld4_lane_f16): Likewise.
6722 (vld4_lane_f32): Likewise.
6723 (vld4_lane_f64): Likewise.
6724 (vld4_lane_p8): Likewise.
6725 (vld4_lane_p16): Likewise.
6726 (vld4_lane_p64): Likewise.
6727 (vld4q_lane_u8): Likewise.
6728 (vld4q_lane_u16): Likewise.
6729 (vld4q_lane_u32): Likewise.
6730 (vld4q_lane_u64): Likewise.
6731 (vld4q_lane_s8): Likewise.
6732 (vld4q_lane_s16): Likewise.
6733 (vld4q_lane_s32): Likewise.
6734 (vld4q_lane_s64): Likewise.
6735 (vld4q_lane_f16): Likewise.
6736 (vld4q_lane_f32): Likewise.
6737 (vld4q_lane_f64): Likewise.
6738 (vld4q_lane_p8): Likewise.
6739 (vld4q_lane_p16): Likewise.
6740 (vld4q_lane_p64): Likewise.
6741 (vqtbl2_s8): Likewise.
6742 (vqtbl2_u8): Likewise.
6743 (vqtbl2_p8): Likewise.
6744 (vqtbl2q_s8): Likewise.
6745 (vqtbl2q_u8): Likewise.
6746 (vqtbl2q_p8): Likewise.
6747 (vqtbl3_s8): Likewise.
6748 (vqtbl3_u8): Likewise.
6749 (vqtbl3_p8): Likewise.
6750 (vqtbl3q_s8): Likewise.
6751 (vqtbl3q_u8): Likewise.
6752 (vqtbl3q_p8): Likewise.
6753 (vqtbl4_s8): Likewise.
6754 (vqtbl4_u8): Likewise.
6755 (vqtbl4_p8): Likewise.
6756 (vqtbl4q_s8): Likewise.
6757 (vqtbl4q_u8): Likewise.
6758 (vqtbl4q_p8): Likewise.
6759 (vqtbx2_s8): Likewise.
6760 (vqtbx2_u8): Likewise.
6761 (vqtbx2_p8): Likewise.
6762 (vqtbx2q_s8): Likewise.
6763 (vqtbx2q_u8): Likewise.
6764 (vqtbx2q_p8): Likewise.
6765 (vqtbx3_s8): Likewise.
6766 (vqtbx3_u8): Likewise.
6767 (vqtbx3_p8): Likewise.
6768 (vqtbx3q_s8): Likewise.
6769 (vqtbx3q_u8): Likewise.
6770 (vqtbx3q_p8): Likewise.
6771 (vqtbx4_s8): Likewise.
6772 (vqtbx4_u8): Likewise.
6773 (vqtbx4_p8): Likewise.
6774 (vqtbx4q_s8): Likewise.
6775 (vqtbx4q_u8): Likewise.
6776 (vqtbx4q_p8): Likewise.
6777 (vst1_s64_x2): Likewise.
6778 (vst1_u64_x2): Likewise.
6779 (vst1_f64_x2): Likewise.
6780 (vst1_s8_x2): Likewise.
6781 (vst1_p8_x2): Likewise.
6782 (vst1_s16_x2): Likewise.
6783 (vst1_p16_x2): Likewise.
6784 (vst1_s32_x2): Likewise.
6785 (vst1_u8_x2): Likewise.
6786 (vst1_u16_x2): Likewise.
6787 (vst1_u32_x2): Likewise.
6788 (vst1_f16_x2): Likewise.
6789 (vst1_f32_x2): Likewise.
6790 (vst1_p64_x2): Likewise.
6791 (vst1q_s8_x2): Likewise.
6792 (vst1q_p8_x2): Likewise.
6793 (vst1q_s16_x2): Likewise.
6794 (vst1q_p16_x2): Likewise.
6795 (vst1q_s32_x2): Likewise.
6796 (vst1q_s64_x2): Likewise.
6797 (vst1q_u8_x2): Likewise.
6798 (vst1q_u16_x2): Likewise.
6799 (vst1q_u32_x2): Likewise.
6800 (vst1q_u64_x2): Likewise.
6801 (vst1q_f16_x2): Likewise.
6802 (vst1q_f32_x2): Likewise.
6803 (vst1q_f64_x2): Likewise.
6804 (vst1q_p64_x2): Likewise.
6805 (vst1_s64_x3): Likewise.
6806 (vst1_u64_x3): Likewise.
6807 (vst1_f64_x3): Likewise.
6808 (vst1_s8_x3): Likewise.
6809 (vst1_p8_x3): Likewise.
6810 (vst1_s16_x3): Likewise.
6811 (vst1_p16_x3): Likewise.
6812 (vst1_s32_x3): Likewise.
6813 (vst1_u8_x3): Likewise.
6814 (vst1_u16_x3): Likewise.
6815 (vst1_u32_x3): Likewise.
6816 (vst1_f16_x3): Likewise.
6817 (vst1_f32_x3): Likewise.
6818 (vst1_p64_x3): Likewise.
6819 (vst1q_s8_x3): Likewise.
6820 (vst1q_p8_x3): Likewise.
6821 (vst1q_s16_x3): Likewise.
6822 (vst1q_p16_x3): Likewise.
6823 (vst1q_s32_x3): Likewise.
6824 (vst1q_s64_x3): Likewise.
6825 (vst1q_u8_x3): Likewise.
6826 (vst1q_u16_x3): Likewise.
6827 (vst1q_u32_x3): Likewise.
6828 (vst1q_u64_x3): Likewise.
6829 (vst1q_f16_x3): Likewise.
6830 (vst1q_f32_x3): Likewise.
6831 (vst1q_f64_x3): Likewise.
6832 (vst1q_p64_x3): Likewise.
6833 (vst1_s8_x4): Likewise.
6834 (vst1q_s8_x4): Likewise.
6835 (vst1_s16_x4): Likewise.
6836 (vst1q_s16_x4): Likewise.
6837 (vst1_s32_x4): Likewise.
6838 (vst1q_s32_x4): Likewise.
6839 (vst1_u8_x4): Likewise.
6840 (vst1q_u8_x4): Likewise.
6841 (vst1_u16_x4): Likewise.
6842 (vst1q_u16_x4): Likewise.
6843 (vst1_u32_x4): Likewise.
6844 (vst1q_u32_x4): Likewise.
6845 (vst1_f16_x4): Likewise.
6846 (vst1q_f16_x4): Likewise.
6847 (vst1_f32_x4): Likewise.
6848 (vst1q_f32_x4): Likewise.
6849 (vst1_p8_x4): Likewise.
6850 (vst1q_p8_x4): Likewise.
6851 (vst1_p16_x4): Likewise.
6852 (vst1q_p16_x4): Likewise.
6853 (vst1_s64_x4): Likewise.
6854 (vst1_u64_x4): Likewise.
6855 (vst1_p64_x4): Likewise.
6856 (vst1q_s64_x4): Likewise.
6857 (vst1q_u64_x4): Likewise.
6858 (vst1q_p64_x4): Likewise.
6859 (vst1_f64_x4): Likewise.
6860 (vst1q_f64_x4): Likewise.
6861 (vst2_s64): Likewise.
6862 (vst2_u64): Likewise.
6863 (vst2_f64): Likewise.
6864 (vst2_s8): Likewise.
6865 (vst2_p8): Likewise.
6866 (vst2_s16): Likewise.
6867 (vst2_p16): Likewise.
6868 (vst2_s32): Likewise.
6869 (vst2_u8): Likewise.
6870 (vst2_u16): Likewise.
6871 (vst2_u32): Likewise.
6872 (vst2_f16): Likewise.
6873 (vst2_f32): Likewise.
6874 (vst2_p64): Likewise.
6875 (vst2q_s8): Likewise.
6876 (vst2q_p8): Likewise.
6877 (vst2q_s16): Likewise.
6878 (vst2q_p16): Likewise.
6879 (vst2q_s32): Likewise.
6880 (vst2q_s64): Likewise.
6881 (vst2q_u8): Likewise.
6882 (vst2q_u16): Likewise.
6883 (vst2q_u32): Likewise.
6884 (vst2q_u64): Likewise.
6885 (vst2q_f16): Likewise.
6886 (vst2q_f32): Likewise.
6887 (vst2q_f64): Likewise.
6888 (vst2q_p64): Likewise.
6889 (vst3_s64): Likewise.
6890 (vst3_u64): Likewise.
6891 (vst3_f64): Likewise.
6892 (vst3_s8): Likewise.
6893 (vst3_p8): Likewise.
6894 (vst3_s16): Likewise.
6895 (vst3_p16): Likewise.
6896 (vst3_s32): Likewise.
6897 (vst3_u8): Likewise.
6898 (vst3_u16): Likewise.
6899 (vst3_u32): Likewise.
6900 (vst3_f16): Likewise.
6901 (vst3_f32): Likewise.
6902 (vst3_p64): Likewise.
6903 (vst3q_s8): Likewise.
6904 (vst3q_p8): Likewise.
6905 (vst3q_s16): Likewise.
6906 (vst3q_p16): Likewise.
6907 (vst3q_s32): Likewise.
6908 (vst3q_s64): Likewise.
6909 (vst3q_u8): Likewise.
6910 (vst3q_u16): Likewise.
6911 (vst3q_u32): Likewise.
6912 (vst3q_u64): Likewise.
6913 (vst3q_f16): Likewise.
6914 (vst3q_f32): Likewise.
6915 (vst3q_f64): Likewise.
6916 (vst3q_p64): Likewise.
6917 (vst4_s64): Likewise.
6918 (vst4_u64): Likewise.
6919 (vst4_f64): Likewise.
6920 (vst4_s8): Likewise.
6921 (vst4_p8): Likewise.
6922 (vst4_s16): Likewise.
6923 (vst4_p16): Likewise.
6924 (vst4_s32): Likewise.
6925 (vst4_u8): Likewise.
6926 (vst4_u16): Likewise.
6927 (vst4_u32): Likewise.
6928 (vst4_f16): Likewise.
6929 (vst4_f32): Likewise.
6930 (vst4_p64): Likewise.
6931 (vst4q_s8): Likewise.
6932 (vst4q_p8): Likewise.
6933 (vst4q_s16): Likewise.
6934 (vst4q_p16): Likewise.
6935 (vst4q_s32): Likewise.
6936 (vst4q_s64): Likewise.
6937 (vst4q_u8): Likewise.
6938 (vst4q_u16): Likewise.
6939 (vst4q_u32): Likewise.
6940 (vst4q_u64): Likewise.
6941 (vst4q_f16): Likewise.
6942 (vst4q_f32): Likewise.
6943 (vst4q_f64): Likewise.
6944 (vst4q_p64): Likewise.
6945 (vtbx4_s8): Likewise.
6946 (vtbx4_u8): Likewise.
6947 (vtbx4_p8): Likewise.
6948 (vld1_bf16_x2): Likewise.
6949 (vld1q_bf16_x2): Likewise.
6950 (vld1_bf16_x3): Likewise.
6951 (vld1q_bf16_x3): Likewise.
6952 (vld1_bf16_x4): Likewise.
6953 (vld1q_bf16_x4): Likewise.
6954 (vld2_bf16): Likewise.
6955 (vld2q_bf16): Likewise.
6956 (vld2_dup_bf16): Likewise.
6957 (vld2q_dup_bf16): Likewise.
6958 (vld3_bf16): Likewise.
6959 (vld3q_bf16): Likewise.
6960 (vld3_dup_bf16): Likewise.
6961 (vld3q_dup_bf16): Likewise.
6962 (vld4_bf16): Likewise.
6963 (vld4q_bf16): Likewise.
6964 (vld4_dup_bf16): Likewise.
6965 (vld4q_dup_bf16): Likewise.
6966 (vst1_bf16_x2): Likewise.
6967 (vst1q_bf16_x2): Likewise.
6968 (vst1_bf16_x3): Likewise.
6969 (vst1q_bf16_x3): Likewise.
6970 (vst1_bf16_x4): Likewise.
6971 (vst1q_bf16_x4): Likewise.
6972 (vst2_bf16): Likewise.
6973 (vst2q_bf16): Likewise.
6974 (vst3_bf16): Likewise.
6975 (vst3q_bf16): Likewise.
6976 (vst4_bf16): Likewise.
6977 (vst4q_bf16): Likewise.
6978 (vld2_lane_bf16): Likewise.
6979 (vld2q_lane_bf16): Likewise.
6980 (vld3_lane_bf16): Likewise.
6981 (vld3q_lane_bf16): Likewise.
6982 (vld4_lane_bf16): Likewise.
6983 (vld4q_lane_bf16): Likewise.
6984 (vst2_lane_bf16): Likewise.
6985 (vst2q_lane_bf16): Likewise.
6986 (vst3_lane_bf16): Likewise.
6987 (vst3q_lane_bf16): Likewise.
6988 (vst4_lane_bf16): Likewise.
6989 (vst4q_lane_bf16): Likewise.
6990 * config/aarch64/geniterators.sh: Modify iterator regex to
6991 match new vector-tuple modes.
6992 * config/aarch64/iterators.md (insn_count): Extend mode
6993 attribute with vector-tuple type information.
6995 (Vendreg): Likewise.
6998 (VSTRUCT_2D): New mode iterator.
6999 (VSTRUCT_2DNX): Likewise.
7000 (VSTRUCT_2DX): Likewise.
7001 (VSTRUCT_2Q): Likewise.
7002 (VSTRUCT_2QD): Likewise.
7003 (VSTRUCT_3D): Likewise.
7004 (VSTRUCT_3DNX): Likewise.
7005 (VSTRUCT_3DX): Likewise.
7006 (VSTRUCT_3Q): Likewise.
7007 (VSTRUCT_3QD): Likewise.
7008 (VSTRUCT_4D): Likewise.
7009 (VSTRUCT_4DNX): Likewise.
7010 (VSTRUCT_4DX): Likewise.
7011 (VSTRUCT_4Q): Likewise.
7012 (VSTRUCT_4QD): Likewise.
7013 (VSTRUCT_D): Likewise.
7014 (VSTRUCT_Q): Likewise.
7015 (VSTRUCT_QD): Likewise.
7016 (VSTRUCT_ELT): New mode attribute.
7017 (vstruct_elt): Likewise.
7018 * genmodes.c (VECTOR_MODE): Add default prefix and order
7020 (VECTOR_MODE_WITH_PREFIX): Define.
7021 (make_vector_mode): Add mode prefix and order parameters.
7023 2021-11-04 Jonathan Wright <jonathan.wright@arm.com>
7025 * expmed.c (extract_bit_field_1): Ensure modes are tieable.
7027 2021-11-04 Jonathan Wright <jonathan.wright@arm.com>
7029 * expr.c (emit_group_load_1): Remove historic workaround.
7031 2021-11-04 Jonathan Wright <jonathan.wright@arm.com>
7033 * config/aarch64/aarch64-builtins.c (aarch64_init_simd_builtins):
7034 Factor out main loop to...
7035 (aarch64_init_simd_builtin_functions): This new function.
7036 (register_tuple_type): Define.
7037 (aarch64_scalar_builtin_type_p): Define.
7038 (handle_arm_neon_h): Define.
7039 * config/aarch64/aarch64-c.c (aarch64_pragma_aarch64): Handle
7040 pragma for arm_neon.h.
7041 * config/aarch64/aarch64-protos.h (aarch64_advsimd_struct_mode_p):
7043 (handle_arm_neon_h): Likewise.
7044 * config/aarch64/aarch64.c (aarch64_advsimd_struct_mode_p):
7045 Remove static modifier.
7046 * config/aarch64/arm_neon.h (target): Remove Neon vector
7047 structure type definitions.
7049 2021-11-04 Aldy Hernandez <aldyh@redhat.com>
7051 PR tree-optimization/102943
7052 * gimple-range-path.cc (path_range_query::range_on_path_entry):
7053 Prefer range_of_expr unless there are no statements in the BB.
7055 2021-11-04 Aldy Hernandez <aldyh@redhat.com>
7057 PR tree-optimization/102943
7058 * tree-ssa-threadbackward.c (back_threader::find_paths_to_names):
7059 Avoid duplicate calculation of paths.
7061 2021-11-04 Aldy Hernandez <aldyh@redhat.com>
7063 PR tree-optimization/102943
7064 * gimple-range-path.cc (path_range_query::compute_phi_relations):
7065 Only compute relations for SSA names in the import list.
7066 (path_range_query::compute_outgoing_relations): Same.
7067 * gimple-range-path.h (path_range_query::import_p): New.
7069 2021-11-04 Richard Biener <rguenther@suse.de>
7071 PR rtl-optimization/103075
7072 * simplify-rtx.c (exact_int_to_float_conversion_p): Return
7073 false for a VOIDmode operand.
7075 2021-11-04 Richard Sandiford <richard.sandiford@arm.com>
7077 * config/aarch64/aarch64.c (aarch64_vector_costs): Make member
7078 variables private and add "m_" to their names. Remove is_loop.
7079 (aarch64_record_potential_advsimd_unrolling): Replace with...
7080 (aarch64_vector_costs::record_potential_advsimd_unrolling): ...this.
7081 (aarch64_analyze_loop_vinfo): Replace with...
7082 (aarch64_vector_costs::analyze_loop_vinfo): ...this.
7083 Move initialization of (m_)vec_flags to add_stmt_cost.
7084 (aarch64_analyze_bb_vinfo): Delete.
7085 (aarch64_count_ops): Replace with...
7086 (aarch64_vector_costs::count_ops): ...this.
7087 (aarch64_vector_costs::add_stmt_cost): Set m_vec_flags,
7088 using m_costing_for_scalar to test whether we're costing
7089 scalar or vector code.
7090 (aarch64_adjust_body_cost_sve): Replace with...
7091 (aarch64_vector_costs::adjust_body_cost_sve): ...this.
7092 (aarch64_adjust_body_cost): Replace with...
7093 (aarch64_vector_costs::adjust_body_cost): ...this.
7094 (aarch64_vector_costs::finish_cost): Use m_vinfo instead of is_loop.
7096 2021-11-04 Richard Sandiford <richard.sandiford@arm.com>
7098 * target.def (targetm.vectorize.init_cost): Replace with...
7099 (targetm.vectorize.create_costs): ...this.
7100 (targetm.vectorize.add_stmt_cost): Delete.
7101 (targetm.vectorize.finish_cost): Likewise.
7102 (targetm.vectorize.destroy_cost_data): Likewise.
7103 * doc/tm.texi.in (TARGET_VECTORIZE_INIT_COST): Replace with...
7104 (TARGET_VECTORIZE_CREATE_COSTS): ...this.
7105 (TARGET_VECTORIZE_ADD_STMT_COST): Delete.
7106 (TARGET_VECTORIZE_FINISH_COST): Likewise.
7107 (TARGET_VECTORIZE_DESTROY_COST_DATA): Likewise.
7108 * doc/tm.texi: Regenerate.
7109 * tree-vectorizer.h (vec_info::vec_info): Remove target_cost_data
7111 (vec_info::target_cost_data): Change from a void * to a vector_costs *.
7112 (vector_costs): New class.
7113 (init_cost): Take a vec_info and return a vector_costs.
7114 (dump_stmt_cost): Remove data parameter.
7115 (add_stmt_cost): Replace vinfo and data parameters with a vector_costs.
7116 (add_stmt_costs): Likewise.
7117 (finish_cost): Replace data parameter with a vector_costs.
7118 (destroy_cost_data): Delete.
7119 * tree-vectorizer.c (dump_stmt_cost): Remove data argument and
7121 (vec_info::vec_info): Remove the target_cost_data parameter and
7122 initialize the member variable to null instead.
7123 (vec_info::~vec_info): Delete target_cost_data instead of calling
7125 (vector_costs::add_stmt_cost): New function.
7126 (vector_costs::finish_cost): Likewise.
7127 (vector_costs::record_stmt_cost): Likewise.
7128 (vector_costs::adjust_cost_for_freq): Likewise.
7129 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Update
7130 call to vec_info::vec_info.
7131 (vect_compute_single_scalar_iteration_cost): Update after above
7132 changes to costing interface.
7133 (vect_analyze_loop_operations): Likewise.
7134 (vect_estimate_min_profitable_iters): Likewise.
7135 (vect_analyze_loop_2): Initialize LOOP_VINFO_TARGET_COST_DATA
7136 at the start_over point, where it needs to be recreated after
7137 trying without slp. Update retry code accordingly.
7138 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Update call
7139 to vec_info::vec_info.
7140 (vect_slp_analyze_operation): Update after above changes to costing
7142 (vect_bb_vectorization_profitable_p): Likewise.
7143 * targhooks.h (default_init_cost): Replace with...
7144 (default_vectorize_create_costs): ...this.
7145 (default_add_stmt_cost): Delete.
7146 (default_finish_cost, default_destroy_cost_data): Likewise.
7147 * targhooks.c (default_init_cost): Replace with...
7148 (default_vectorize_create_costs): ...this.
7149 (default_add_stmt_cost): Delete, moving logic to vector_costs instead.
7150 (default_finish_cost, default_destroy_cost_data): Delete.
7151 * config/aarch64/aarch64.c (aarch64_vector_costs): Inherit from
7152 vector_costs. Add a constructor.
7153 (aarch64_init_cost): Replace with...
7154 (aarch64_vectorize_create_costs): ...this.
7155 (aarch64_add_stmt_cost): Replace with...
7156 (aarch64_vector_costs::add_stmt_cost): ...this. Use record_stmt_cost
7157 to adjust the cost for inner loops.
7158 (aarch64_finish_cost): Replace with...
7159 (aarch64_vector_costs::finish_cost): ...this.
7160 (aarch64_destroy_cost_data): Delete.
7161 (TARGET_VECTORIZE_INIT_COST): Replace with...
7162 (TARGET_VECTORIZE_CREATE_COSTS): ...this.
7163 (TARGET_VECTORIZE_ADD_STMT_COST): Delete.
7164 (TARGET_VECTORIZE_FINISH_COST): Likewise.
7165 (TARGET_VECTORIZE_DESTROY_COST_DATA): Likewise.
7166 * config/i386/i386.c (ix86_vector_costs): New structure.
7167 (ix86_init_cost): Replace with...
7168 (ix86_vectorize_create_costs): ...this.
7169 (ix86_add_stmt_cost): Replace with...
7170 (ix86_vector_costs::add_stmt_cost): ...this. Use adjust_cost_for_freq
7171 to adjust the cost for inner loops.
7172 (ix86_finish_cost, ix86_destroy_cost_data): Delete.
7173 (TARGET_VECTORIZE_INIT_COST): Replace with...
7174 (TARGET_VECTORIZE_CREATE_COSTS): ...this.
7175 (TARGET_VECTORIZE_ADD_STMT_COST): Delete.
7176 (TARGET_VECTORIZE_FINISH_COST): Likewise.
7177 (TARGET_VECTORIZE_DESTROY_COST_DATA): Likewise.
7178 * config/rs6000/rs6000.c (TARGET_VECTORIZE_INIT_COST): Replace with...
7179 (TARGET_VECTORIZE_CREATE_COSTS): ...this.
7180 (TARGET_VECTORIZE_ADD_STMT_COST): Delete.
7181 (TARGET_VECTORIZE_FINISH_COST): Likewise.
7182 (TARGET_VECTORIZE_DESTROY_COST_DATA): Likewise.
7183 (rs6000_cost_data): Inherit from vector_costs.
7184 Add a constructor. Drop loop_info, cost and costing_for_scalar
7185 in favor of the corresponding vector_costs member variables.
7186 Add "m_" to the names of the remaining member variables and
7188 (rs6000_density_test): Replace with...
7189 (rs6000_cost_data::density_test): ...this.
7190 (rs6000_init_cost): Replace with...
7191 (rs6000_vectorize_create_costs): ...this.
7192 (rs6000_update_target_cost_per_stmt): Replace with...
7193 (rs6000_cost_data::update_target_cost_per_stmt): ...this.
7194 (rs6000_add_stmt_cost): Replace with...
7195 (rs6000_cost_data::add_stmt_cost): ...this. Use adjust_cost_for_freq
7196 to adjust the cost for inner loops.
7197 (rs6000_adjust_vect_cost_per_loop): Replace with...
7198 (rs6000_cost_data::adjust_vect_cost_per_loop): ...this.
7199 (rs6000_finish_cost): Replace with...
7200 (rs6000_cost_data::finish_cost): ...this. Group loop code
7201 into a single if statement and pass the loop_vinfo down to
7203 (rs6000_destroy_cost_data): Delete.
7205 2021-11-04 Aldy Hernandez <aldyh@redhat.com>
7207 PR tree-optimization/103062
7208 PR tree-optimization/103062
7209 * value-pointer-equiv.cc (ssa_equiv_stack::ssa_equiv_stack):
7210 Increase size of allocation by 1.
7211 (ssa_equiv_stack::push_replacement): Grow as needed.
7212 (ssa_equiv_stack::get_replacement): Same.
7213 (pointer_equiv_analyzer::pointer_equiv_analyzer): Same.
7214 (pointer_equiv_analyzer::~pointer_equiv_analyzer): Remove delete.
7215 (pointer_equiv_analyzer::set_global_equiv): Grow as needed.
7216 (pointer_equiv_analyzer::get_equiv): Same.
7217 (pointer_equiv_analyzer::get_equiv_expr): Remove const.
7218 * value-pointer-equiv.h (class pointer_equiv_analyzer): Remove
7219 const markers. Use auto_vec instead of tree *.
7221 2021-11-04 Richard Biener <rguenther@suse.de>
7223 * tree-ssa-sccvn.c (vn_nary_op_insert_into): Remove always
7224 true parameter and inline valueization.
7225 (vn_nary_op_lookup_1): Inline valueization from ...
7226 (vn_nary_op_compute_hash): ... here and remove it here.
7227 * tree-ssa-pre.c (phi_translate_1): Do not valueize
7228 before vn_nary_lookup_pieces.
7229 (get_representative_for): Mark created SSA representatives
7232 2021-11-04 Richard Sandiford <richard.sandiford@arm.com>
7234 * simplify-rtx.c (simplify_context::simplify_gen_vec_select): Assert
7235 that the operand has a vector mode. Use subreg_lowpart_offset
7236 to test whether an index corresponds to the low part.
7238 2021-11-04 Richard Sandiford <richard.sandiford@arm.com>
7240 * read-rtl.c: Remove dead !GENERATOR_FILE block.
7241 * read-rtl-function.c (function_reader::consolidate_singletons):
7242 Generate canonical CONST_VECTORs.
7244 2021-11-04 liuhongt <hongtao.liu@intel.com>
7247 * config/i386/predicates.md (reg_or_notreg_operand): Rename to ..
7248 (regmem_or_bitnot_regmem_operand): .. and extend to handle
7250 * config/i386/sse.md (*<avx512>_vpternlog<mode>_1): Force_reg
7251 the operands which are required to be register_operand.
7252 (*<avx512>_vpternlog<mode>_2): Ditto.
7253 (*<avx512>_vpternlog<mode>_3): Ditto.
7254 (*<avx512>_vternlog<mode>_all): Disallow embeded broadcast for
7255 vector HFmodes since it's not a real AVX512FP16 instruction.
7257 2021-11-04 liuhongt <hongtao.liu@intel.com>
7260 * match.pd: simplify (trunc)copysign((extend)a, (extend)b) to
7261 .COPYSIGN (a,b) when a and b are same type as the truncation
7262 type and has less precision than extend type.
7264 2021-11-04 Richard Biener <rguenther@suse.de>
7266 * doc/generic.texi: Update TARGET_MEM_REF and MEM_REF
7269 2021-11-04 Hongyu Wang <hongyu.wang@intel.com>
7271 * config/i386/sse.md (VI2_AVX512VNNIBW): New mode iterator.
7272 (VI1_AVX512VNNI): Likewise.
7273 (SDOT_VPDP_SUF): New mode_attr.
7276 (sdot_prod<mode>): Use VI2_AVX512F iterator, expand to
7277 vpdpwssd when VNNI targets available.
7278 (usdot_prod<mode>): New expander for vector QImode.
7280 2021-11-04 Hongyu Wang <hongyu.wang@intel.com>
7282 * config/i386/amxtileintrin.h (_tile_loadd_internal): Add
7283 parentheses to base and stride.
7284 (_tile_stream_loadd_internal): Likewise.
7285 (_tile_stored_internal): Likewise.
7287 2021-11-03 Maciej W. Rozycki <macro@embecosm.com>
7289 * config/riscv/riscv.c (riscv_class_max_nregs): Swap the
7290 arguments to `reg_class_subset_p'.
7292 2021-11-03 Joseph Myers <joseph@codesourcery.com>
7295 * fold-const.c (fold_init): New function.
7296 * fold-const.h (fold_init): New prototype.
7298 2021-11-03 Andrew MacLeod <amacleod@redhat.com>
7300 * params.opt (param_vrp2_mode): Make ranger the default for VRP2.
7302 2021-11-03 Andrew MacLeod <amacleod@redhat.com>
7304 * gimple-range-fold.cc (fold_using_range::range_of_phi): Don't import
7305 a range from edge if arg == phidef.
7307 2021-11-03 Andrew MacLeod <amacleod@redhat.com>
7309 * gimple-range-fold.cc (fold_using_range::range_of_builtin_call): Test
7310 for constant before any other processing.
7312 2021-11-03 Andrew MacLeod <amacleod@redhat.com>
7314 * flag-types.h (RANGER_DEBUG_ALL): Fix values.
7316 2021-11-03 Andrew MacLeod <amacleod@redhat.com>
7318 * gimple-range.cc (gimple_ranger::gimple_ranger): Initialize current_bb.
7319 (gimple_ranger::range_of_expr): Pick up range_on_entry when there is
7320 no explcit context and current_bb is set.
7321 (gimple_ranger::fold_stmt): New.
7322 * gimple-range.h (current_bb, fold_stmt): New.
7323 * tree-vrp.c (rvrp_folder::fold_stmt): Call ranger's fold_stmt.
7325 2021-11-03 Richard Biener <rguenther@suse.de>
7327 PR tree-optimization/102970
7328 * tree-ssa-pre.c (phi_translate_1): Drop clique and base
7329 when translating a MEM_REF over a backedge.
7331 2021-11-03 Philipp Tomsich <philipp.tomsich@vrull.eu>
7333 * config/aarch64/aarch64-cores.def (AARCH64_CORE): New Ampere-1 core.
7334 * config/aarch64/aarch64-tune.md: Regenerate.
7335 * config/aarch64/aarch64-cost-tables.h: Add extra costs for Ampere-1.
7336 * config/aarch64/aarch64.c: Add tuning structures for Ampere-1.
7337 * doc/invoke.texi: Add documentation for Ampere-1 core.
7339 2021-11-03 Wilco Dijkstra <wdijkstr@arm.com>
7341 * config/aarch64/aarch64.md (movsi): Add alternative for GOT accesses.
7343 (ldr_got_small_<mode>): Remove pattern.
7344 (ldr_got_small_sidi): Likewise.
7345 * config/aarch64/aarch64.c (aarch64_load_symref_appropriately): Keep
7346 GOT accesses as moves.
7347 (aarch64_print_operand): Correctly print got_lo12 in L specifier.
7348 (aarch64_mov_operand_p): Make GOT accesses valid move operands.
7349 * config/aarch64/constraints.md: Add new constraint Usw for GOT access.
7351 2021-11-03 Martin Liska <mliska@suse.cz>
7353 * gcov.c (read_line): Remove dead variable.
7355 2021-11-03 Martin Liska <mliska@suse.cz>
7358 * ipa-fnsummary.c (edge_predicate_pool): Rename predicate class to ipa_predicate.
7359 (ipa_fn_summary::account_size_time): Likewise.
7360 (edge_set_predicate): Likewise.
7361 (set_hint_predicate): Likewise.
7362 (add_freqcounting_predicate): Likewise.
7363 (evaluate_conditions_for_known_args): Likewise.
7364 (evaluate_properties_for_edge): Likewise.
7365 (remap_freqcounting_preds_after_dup): Likewise.
7366 (ipa_fn_summary_t::duplicate): Likewise.
7367 (set_cond_stmt_execution_predicate): Likewise.
7368 (set_switch_stmt_execution_predicate): Likewise.
7369 (compute_bb_predicates): Likewise.
7370 (will_be_nonconstant_expr_predicate): Likewise.
7371 (will_be_nonconstant_predicate): Likewise.
7372 (phi_result_unknown_predicate): Likewise.
7373 (predicate_for_phi_result): Likewise.
7374 (analyze_function_body): Likewise.
7375 (compute_fn_summary): Likewise.
7376 (summarize_calls_size_and_time): Likewise.
7377 (estimate_calls_size_and_time): Likewise.
7378 (ipa_call_context::estimate_size_and_time): Likewise.
7379 (remap_edge_summaries): Likewise.
7380 (remap_freqcounting_predicate): Likewise.
7381 (ipa_merge_fn_summary_after_inlining): Likewise.
7382 (ipa_update_overall_fn_summary): Likewise.
7383 (read_ipa_call_summary): Likewise.
7384 (inline_read_section): Likewise.
7385 * ipa-fnsummary.h (struct ipa_freqcounting_predicate): Likewise.
7386 * ipa-predicate.c (predicate::add_clause): Likewise.
7387 (ipa_predicate::add_clause): Likewise.
7388 (predicate::or_with): Likewise.
7389 (ipa_predicate::or_with): Likewise.
7390 (predicate::evaluate): Likewise.
7391 (ipa_predicate::evaluate): Likewise.
7392 (predicate::probability): Likewise.
7393 (ipa_predicate::probability): Likewise.
7394 (dump_condition): Likewise.
7395 (dump_clause): Likewise.
7396 (predicate::dump): Likewise.
7397 (ipa_predicate::dump): Likewise.
7398 (predicate::debug): Likewise.
7399 (ipa_predicate::debug): Likewise.
7400 (predicate::remap_after_duplication): Likewise.
7401 (ipa_predicate::remap_after_duplication): Likewise.
7402 (predicate::remap_after_inlining): Likewise.
7403 (ipa_predicate::remap_after_inlining): Likewise.
7404 (predicate::stream_in): Likewise.
7405 (ipa_predicate::stream_in): Likewise.
7406 (predicate::stream_out): Likewise.
7407 (ipa_predicate::stream_out): Likewise.
7408 (add_condition): Likewise.
7409 * ipa-predicate.h (class predicate): Likewise.
7410 (class ipa_predicate): Likewise.
7411 (add_condition): Likewise.
7413 2021-11-03 Richard Biener <rguenther@suse.de>
7415 * bitmap.h (bitmap_bit_p): Change the return type to bool.
7416 * bitmap.c (bitmap_bit_p): Likewise.
7417 * sbitmap.h (bitmap_bit_p): Likewise.
7418 (bitmap_set_bit): Return whether the bit changed.
7419 (bitmap_clear_bit): Likewise.
7420 * tree-ssa.c (verify_vssa): Make use of the changed state
7421 from bitmap_set_bit.
7423 2021-11-03 Richard Biener <rguenther@suse.de>
7425 PR middle-end/103033
7426 * internal-fn.c (expand_DEFERRED_INIT): Elide the
7427 native_interpret_expr path in favor of folding the
7428 VIEW_CONVERT_EXPR generated when punning the RHS.
7430 2021-11-03 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
7432 * config/s390/s390.c (s390_loop_unroll_adjust): In case of early
7435 2021-11-03 Jan Hubicka <hubicka@ucw.cz>
7438 * ipa-modref.c (callee_to_caller_flags): New function.
7439 (modref_eaf_analysis::analyze_ssa_name): Use it.
7440 (ipa_merge_modref_summary_after_inlining): Fix whitespace.
7442 2021-11-02 Jan Hubicka <hubicka@ucw.cz>
7444 * ipa-modref.c (modref_eaf_analysis::analyze_ssa_name): Revert
7447 2021-11-02 Roger Sayle <roger@nextmovesoftware.com>
7448 Uroš Bizjak <ubizjak@gmail.com>
7450 * config/i386/i386.md (<any_rotate>ti3): Provide expansion for
7451 rotations by non-constant amounts.
7453 2021-11-02 Jan Hubicka <hubicka@ucw.cz>
7455 * ipa-modref.c: Fix anonymous namespace placement.
7456 (class modref_eaf_analysis): New class.
7457 (analyze_ssa_name_flags): Turn to ...
7458 (modref_eaf_analysis::analyze_ssa_name): ... this one.
7459 (merge_call_lhs_flags): Turn to ...
7460 (modref_eaf_analysis::merge_call_lhs_flags): .. this one
7461 (modref_eaf_analysis::merge_with_ssa_name): New member function.
7462 (record_escape_points): Turn to ...
7463 (modref_eaf_analysis::record_escape_points): ... this one.
7464 (analyze_parms): Updat
7465 (ipa_merge_modref_summary_after_inlining): Move to the end of file.
7467 2021-11-02 Jan Hubicka <hubicka@ucw.cz>
7469 * gimple.c (gimple_call_static_chain_flags): New function.
7470 * gimple.h (gimple_call_static_chain_flags): Declare
7471 * ipa-modref.c (modref_summary::modref_summary): Initialize
7473 (modref_summary_lto::modref_summary_lto): Likewise.
7474 (modref_summary::useful_p): Test static_chain_flags.
7475 (modref_summary_lto::useful_p): Likewise.
7476 (struct modref_summary_lto): Add static_chain_flags.
7477 (modref_summary::dump): Dump static_chain_flags.
7478 (modref_summary_lto::dump): Likewise.
7479 (struct escape_point): Add static_cahin_arg.
7480 (analyze_ssa_name_flags): Use gimple_call_static_chain_flags.
7481 (analyze_parms): Handle static chains.
7482 (modref_summaries::duplicate): Duplicate static_chain_flags.
7483 (modref_summaries_lto::duplicate): Likewise.
7484 (modref_write): Stream static_chain_flags.
7485 (read_section): Likewise.
7486 (modref_merge_call_site_flags): Handle static_chain_flags.
7487 * ipa-modref.h (struct modref_summary): Add static_chain_flags.
7488 * tree-ssa-structalias.c (handle_rhs_call): Use
7489 gimple_static_chain_flags.
7491 2021-11-02 Richard Biener <rguenther@suse.de>
7493 PR tree-optimization/103029
7494 * tree-vect-loop-manip.c (vect_loop_versioning): Ensure
7495 the PHI nodes in the loop maintain their original operand
7498 2021-11-02 Jan Hubicka <hubicka@ucw.cz>
7500 * tree-core.h (EAF_NOT_RETURNED_DIRECTLY): New flag.
7501 (EAF_NOREAD): Renumber.
7502 * ipa-modref.c (dump_eaf_flags): Dump EAF_NOT_RETURNED_DIRECTLY.
7503 (remove_useless_eaf_flags): Handle EAF_NOT_RETURNED_DIRECTLY
7504 (deref_flags): Likewise.
7505 (modref_lattice::init): Likewise.
7506 (modref_lattice::merge): Likewise.
7507 (merge_call_lhs_flags): Likewise.
7508 (analyze_ssa_name_flags): Likewise.
7509 (modref_merge_call_site_flags): Likewise.
7510 * tree-ssa-structalias.c (handle_call_arg): Likewise.
7512 2021-11-02 Maciej W. Rozycki <macro@embecosm.com>
7514 * config/riscv/riscv.c (riscv_rtx_costs): Correct a CONST_INT_P
7515 check and remove an unused local variable with shNadd/shNadd.uw
7518 2021-11-02 Tamar Christina <tamar.christina@arm.com>
7520 PR tree-optimization/103007
7521 * tree-vect-slp-patterns.c (complex_fms_pattern::matches): Add elem
7524 2021-11-02 Richard Biener <rguenther@suse.de>
7526 PR middle-end/103038
7527 * fold-const.c (native_interpret_expr): Handle OFFSET_TYPE.
7528 (can_native_interpret_type_p): Likewise.
7529 * internal-fn.c (expand_DEFERRED_INIT): View-convert the
7530 RHS if the LHS is an SSA name.
7532 2021-11-02 Richard Sandiford <richard.sandiford@arm.com>
7534 * langhooks.h (lang_hooks_for_types::simulate_record_decl): New hook.
7535 * langhooks-def.h (lhd_simulate_record_decl): Declare.
7536 (LANG_HOOKS_SIMULATE_RECORD_DECL): Define.
7537 (LANG_HOOKS_FOR_TYPES_INITIALIZER): Include it.
7538 * langhooks.c (lhd_simulate_record_decl): New function.
7540 2021-11-02 Jakub Jelinek <jakub@redhat.com>
7543 * config/i386/i386.c (ix86_vector_mode_supported_p): Reject vector
7544 modes with TImode inner mode if 32-bit.
7546 2021-11-02 liuhongt <hongtao.liu@intel.com>
7548 * doc/sourcebuild.texi (vect_slp_v4qi_store_unalign,
7549 vect_slp_v2hi_store_unalign, vect_slp_v4hi_store_unalign,
7550 vect_slp_v4si_store_unalign): Document efficient target.
7551 (vect_slp_v4qi_store_unalign_1, vect_slp_v8qi_store_unalign_1,
7552 vect_slp_v16qi_store_unalign_1): Ditto.
7553 (vect_slp_v2hi_store_align,vect_slp_v2qi_store_align,
7554 vect_slp_v2si_store_align, vect_slp_v4qi_store_align): Ditto.
7555 (struct_4char_block_move, struct_8char_block_move,
7556 struct_16char_block_move): Ditto.
7558 2021-11-02 Roger Sayle <roger@nextmovesoftware.com>
7559 Jakub Jelinek <jakub@redhat.com>
7562 * config/i386/i386-expand.c (ix86_expand_v1ti_to_ti,
7563 ix86_expand_ti_to_v1ti): New helper functions.
7564 (ix86_expand_v1ti_shift): Check if the amount operand is an
7565 integer constant, and expand as a TImode shift if it isn't.
7566 (ix86_expand_v1ti_rotate): Check if the amount operand is an
7567 integer constant, and expand as a TImode rotate if it isn't.
7568 (ix86_expand_v1ti_ashiftrt): New function to expand arithmetic
7569 right shifts of V1TImode quantities.
7570 * config/i386/i386-protos.h (ix86_expand_v1ti_ashift): Prototype.
7571 * config/i386/sse.md (ashlv1ti3, lshrv1ti3): Change constraints
7572 to QImode general_operand, and let the helper functions lower
7573 shifts by non-constant operands, as TImode shifts. Make
7574 conditional on TARGET_64BIT.
7575 (ashrv1ti3): New expander calling ix86_expand_v1ti_ashiftrt.
7576 (rotlv1ti3, rotrv1ti3): Change shift operand to QImode.
7577 Make conditional on TARGET_64BIT.
7579 2021-11-02 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
7581 * config/s390/s390.md ("*cc_to_int", "tabort", "*tabort_1",
7582 "*tabort_1_plus"): Remove operands non-null check.
7584 2021-11-01 Martin Liska <mliska@suse.cz>
7586 * opt-functions.awk: Add new sanity checking.
7587 * optc-gen.awk: Add new argument to integer_range_info.
7588 * params.opt: Update 2 params which have negative IntegerRange.
7590 2021-11-01 qing zhao <qing.zhao@oracle.com>
7592 * gimplify.c (gimplify_decl_expr): Do not add call to
7593 __builtin_clear_padding when a variable is a gimple register
7594 or it might not have padding.
7595 (gimplify_init_constructor): Likewise.
7597 2021-11-01 Tamar Christina <tamar.christina@arm.com>
7599 * config/arm/aarch-common-protos.h (struct vector_cost_table): Add
7600 movi, dup and extract costing fields.
7601 * config/aarch64/aarch64-cost-tables.h (qdf24xx_extra_costs,
7602 thunderx_extra_costs, thunderx2t99_extra_costs,
7603 thunderx3t110_extra_costs, tsv110_extra_costs, a64fx_extra_costs): Use
7605 * config/arm/aarch-cost-tables.h (generic_extra_costs,
7606 cortexa53_extra_costs, cortexa57_extra_costs, cortexa76_extra_costs,
7607 exynosm1_extra_costs, xgene1_extra_costs): Likewise
7608 * config/aarch64/aarch64-simd.md (aarch64_simd_dup<mode>): Add r->w dup.
7609 * config/aarch64/aarch64.c (aarch64_rtx_costs): Add extra costs.
7611 2021-11-01 Tamar Christina <tamar.christina@arm.com>
7613 * cse.c (add_to_set): New.
7614 (find_sets_in_insn): Register constants in sets.
7615 (canonicalize_insn): Use auto_vec instead.
7616 (cse_insn): Try materializing using vec_dup.
7617 * rtl.h (simplify_context::simplify_gen_vec_select,
7618 simplify_gen_vec_select): New.
7619 * simplify-rtx.c (simplify_context::simplify_gen_vec_select): New.
7621 2021-11-01 David Malcolm <dmalcolm@redhat.com>
7623 * common.opt (fdiagnostics-escape-format=): New.
7624 (diagnostics_escape_format): New enum.
7625 (DIAGNOSTICS_ESCAPE_FORMAT_UNICODE): New enum value.
7626 (DIAGNOSTICS_ESCAPE_FORMAT_BYTES): Likewise.
7627 * diagnostic-format-json.cc (json_end_diagnostic): Add
7628 "escape-source" attribute.
7629 * diagnostic-show-locus.c
7630 (exploc_with_display_col::exploc_with_display_col): Replace
7631 "tabstop" param with a cpp_char_column_policy and add an "aspect"
7632 param. Use these to compute m_display_col accordingly.
7633 (struct char_display_policy): New struct.
7634 (layout::m_policy): New field.
7635 (layout::m_escape_on_output): New field.
7636 (def_policy): New function.
7637 (make_range): Update for changes to exploc_with_display_col ctor.
7638 (default_print_decoded_ch): New.
7639 (width_per_escaped_byte): New.
7640 (escape_as_bytes_width): New.
7641 (escape_as_bytes_print): New.
7642 (escape_as_unicode_width): New.
7643 (escape_as_unicode_print): New.
7645 (layout::layout): Initialize new fields. Update m_exploc ctor
7646 call for above change to ctor.
7647 (layout::maybe_add_location_range): Update for changes to
7648 exploc_with_display_col ctor.
7649 (layout::calculate_x_offset_display): Update for change to
7651 (layout::print_source_line): Pass policy
7652 to cpp_display_width_computation. Capture cpp_decoded_char when
7653 calling process_next_codepoint. Move printing of source code to
7654 m_policy.m_print_cb.
7655 (line_label::line_label): Pass in policy rather than context.
7656 (layout::print_any_labels): Update for change to line_label ctor.
7657 (get_affected_range): Pass in policy rather than context, updating
7658 calls to location_compute_display_column accordingly.
7659 (get_printed_columns): Likewise, also for cpp_display_width.
7660 (correction::correction): Pass in policy rather than tabstop.
7661 (correction::compute_display_cols): Pass m_policy rather than
7662 m_tabstop to cpp_display_width.
7663 (correction::m_tabstop): Replace with...
7664 (correction::m_policy): ...this.
7665 (line_corrections::line_corrections): Pass in policy rather than
7667 (line_corrections::m_context): Replace with...
7668 (line_corrections::m_policy): ...this.
7669 (line_corrections::add_hint): Update to use m_policy rather than
7671 (line_corrections::add_hint): Likewise.
7672 (layout::print_trailing_fixits): Likewise.
7673 (selftest::test_display_widths): New.
7674 (selftest::test_layout_x_offset_display_utf8): Update to use
7675 policy rather than tabstop.
7676 (selftest::test_one_liner_labels_utf8): Add test of escaping
7678 (selftest::test_diagnostic_show_locus_one_liner_utf8): Update to
7679 use policy rather than tabstop.
7680 (selftest::test_overlapped_fixit_printing): Likewise.
7681 (selftest::test_overlapped_fixit_printing_utf8): Likewise.
7682 (selftest::test_overlapped_fixit_printing_2): Likewise.
7683 (selftest::test_tab_expansion): Likewise.
7684 (selftest::test_escaping_bytes_1): New.
7685 (selftest::test_escaping_bytes_2): New.
7686 (selftest::diagnostic_show_locus_c_tests): Call the new tests.
7687 * diagnostic.c (diagnostic_initialize): Initialize
7688 context->escape_format.
7689 (convert_column_unit): Update to use default character width policy.
7690 (selftest::test_diagnostic_get_location_text): Likewise.
7691 * diagnostic.h (enum diagnostics_escape_format): New enum.
7692 (diagnostic_context::escape_format): New field.
7693 * doc/invoke.texi (-fdiagnostics-escape-format=): New option.
7694 (-fdiagnostics-format=): Add "escape-source" attribute to examples
7695 of JSON output, and document it.
7696 * input.c (location_compute_display_column): Pass in "policy"
7697 rather than "tabstop", passing to
7698 cpp_byte_column_to_display_column.
7699 (selftest::test_cpp_utf8): Update to use cpp_char_column_policy.
7700 * input.h (class cpp_char_column_policy): New forward decl.
7701 (location_compute_display_column): Pass in "policy" rather than
7703 * opts.c (common_handle_option): Handle
7704 OPT_fdiagnostics_escape_format_.
7705 * selftest.c (temp_source_file::temp_source_file): New ctor
7706 overload taking a size_t.
7707 * selftest.h (temp_source_file::temp_source_file): Likewise.
7709 2021-11-01 Aldy Hernandez <aldyh@redhat.com>
7711 * dbgcnt.def: Add debug counter for back_thread[12] and
7712 back_threadfull[12].
7713 * passes.def: Pass "first" argument to each back threading pass.
7714 * tree-ssa-threadbackward.c (back_threader::back_threader): Add
7716 (back_threader::debug_counter): New.
7717 (back_threader::maybe_register_path): Call debug_counter.
7719 2021-11-01 Aldy Hernandez <aldyh@redhat.com>
7721 * tree-ssa-threadbackward.c (BT_NONE): New.
7724 (back_threader::back_threader): Add flags.
7725 Move loop initialization here.
7726 (back_threader::~back_threader): New.
7727 (back_threader::find_taken_edge_switch): Change solver and ranger
7729 (back_threader::find_taken_edge_cond): Same.
7730 (back_threader::find_paths_to_names): Same.
7731 (back_threader::find_paths): Same.
7732 (back_threader::dump): Same.
7733 (try_thread_blocks): Merge into thread_blocks.
7734 (back_threader::thread_blocks): New.
7735 (do_early_thread_jumps): Merge into thread_blocks.
7736 (do_thread_jumps): Merge into thread_blocks.
7737 (back_threader::thread_through_all_blocks): Remove.
7739 2021-11-01 Andrew MacLeod <amacleod@redhat.com>
7741 PR tree-optimization/103003
7742 * value-relation.cc (dom_oracle::register_relation): If the 2
7743 ssa names are the same, don't register any relation.
7745 2021-11-01 Dan Li <ashimida@linux.alibaba.com>
7747 * config/aarch64/aarch64.c (aarch64_expand_epilogue): Remove
7748 redundant check for calls_eh_return.
7749 * config/aarch64/aarch64.md (*do_return): Likewise.
7751 2021-11-01 Xionghu Luo <luoxhu@linux.ibm.com>
7753 * cfghooks.c (cfg_hook_duplicate_loop_to_header_edge): Rename
7754 duplicate_loop_to_header_edge to
7755 duplicate_loop_body_to_header_edge.
7756 (cfg_hook_duplicate_loop_body_to_header_edge): Likewise.
7757 * cfghooks.h (struct cfg_hooks): Likewise.
7758 (cfg_hook_duplicate_loop_body_to_header_edge): Likewise.
7759 * cfgloopmanip.c (duplicate_loop_body_to_header_edge): Likewise.
7760 (clone_loop_to_header_edge): Likewise.
7761 * cfgloopmanip.h (duplicate_loop_body_to_header_edge): Likewise.
7762 * cfgrtl.c (struct cfg_hooks): Likewise.
7763 * doc/loop.texi: Likewise.
7764 * loop-unroll.c (unroll_loop_constant_iterations): Likewise.
7765 (unroll_loop_runtime_iterations): Likewise.
7766 (unroll_loop_stupid): Likewise.
7767 (apply_opt_in_copies): Likewise.
7768 * tree-cfg.c (struct cfg_hooks): Likewise.
7769 * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Likewise.
7770 (try_peel_loop): Likewise.
7771 * tree-ssa-loop-manip.c (copy_phi_node_args): Likewise.
7772 (gimple_duplicate_loop_body_to_header_edge): Likewise.
7773 (tree_transform_and_unroll_loop): Likewise.
7774 * tree-ssa-loop-manip.h (gimple_duplicate_loop_body_to_header_edge):
7777 2021-11-01 Xionghu Luo <luoxhu@linux.ibm.com>
7779 * cfgloopmanip.c (loop_version): Refactor loopify to
7780 loop_version. Move condition generation after loopify.
7782 * cfgloopmanip.h (loopify): Delete.
7784 2021-10-31 Jan Hubicka <hubicka@ucw.cz>
7786 * ipa-fnsummary.c: Include tree-dfa.h.
7787 (points_to_local_or_readonly_memory_p): Return true on return
7789 * ipa-modref.c (analyze_ssa_name_flags): Fix handling of copy
7792 2021-10-30 Tobias Burnus <tobias@codesourcery.com>
7794 PR middle-end/102972
7795 * omp-low.c (omp_runtime_api_call): Use DECL_ASSEMBLER_NAME to get
7796 internal Fortran name; new permit_num_teams arg to permit
7797 omp_get_num_teams and omp_get_team_num.
7798 (scan_omp_1_stmt): Update call to it, add missing call for
7799 reverse offload, and check for strictly nested API calls in teams.
7801 2021-10-30 Jakub Jelinek <jakub@redhat.com>
7803 * gimplify.c (gimplify_omp_for): Diagnose threadprivate iterators.
7805 2021-10-29 David Malcolm <dmalcolm@redhat.com>
7807 * selftest.c (assert_streq): Add newlines when emitting non-equal
7810 2021-10-29 David Malcolm <dmalcolm@redhat.com>
7812 * Makefile.in: Fix syntax for reference to LIBDEPS in
7815 2021-10-29 Bernhard Reutner-Fischer <aldot@gcc.gnu.org>
7817 * doc/install.texi: Bump required minimum DejaGnu version.
7819 2021-10-29 Aldy Hernandez <aldyh@redhat.com>
7820 Andrew MacLeod <amacleod@redhat.com>
7822 * value-relation.cc (path_oracle::killing_def): Add a
7823 self-equivalence so we don't look to the root oracle.
7825 2021-10-29 Aldy Hernandez <aldyh@redhat.com>
7827 * passes.def: Replace the pass_thread_jumps before VRP* with
7828 pass_thread_jumps_full. Remove all pass_vrp_threader instances.
7829 * tree-ssa-threadbackward.c (pass_data_thread_jumps_full):
7830 Remove hyphen from "thread-full" name.
7832 2021-10-29 Andrew MacLeod <amacleod@redhat.com>
7834 PR tree-optimization/102983
7835 * gimple-range-cache.h (propagate_updated_value): Make public.
7836 * gimple-range.cc (gimple_ranger::range_of_stmt): Propagate exports
7837 when processing gcond stmts.
7839 2021-10-29 Jan Hubicka <hubicka@ucw.cz>
7841 * gimple.c (gimple_call_retslot_flags): New function.
7842 * gimple.h (gimple_call_retslot_flags): Declare.
7843 * ipa-modref.c: Include tree-cfg.h.
7844 (struct escape_entry): Turn parm_index to signed.
7845 (modref_summary_lto::modref_summary_lto): Add retslot_flags.
7846 (modref_summary::modref_summary): Initialize retslot_flags.
7847 (struct modref_summary_lto): Likewise.
7848 (modref_summary::useful_p): Check retslot_flags.
7849 (modref_summary_lto::useful_p): Likewise.
7850 (modref_summary::dump): Dump retslot_flags.
7851 (modref_summary_lto::dump): Likewise.
7852 (struct escape_point): Add hidden_args enum.
7853 (analyze_ssa_name_flags): Ignore return slot return;
7854 use gimple_call_retslot_flags.
7855 (record_escape_points): Break out from ...
7856 (analyze_parms): ... here; handle retslot_flags.
7857 (modref_summaries::duplicate): Duplicate retslot_flags.
7858 (modref_summaries_lto::duplicate): Likewise.
7859 (modref_write_escape_summary): Stream parm_index as signed.
7860 (modref_read_escape_summary): Likewise.
7861 (modref_write): Stream retslot_flags.
7862 (read_section): Likewise.
7863 (struct escape_map): Fix typo in comment.
7864 (update_escape_summary_1): Fix whitespace.
7865 (ipa_merge_modref_summary_after_inlining): Drop retslot_flags.
7866 (modref_merge_call_site_flags): Merge retslot_flags.
7867 * ipa-modref.h (struct modref_summary): Add retslot_flags.
7868 * tree-ssa-structalias.c (handle_rhs_call): Handle retslot_flags.
7870 2021-10-29 Tamar Christina <tamar.christina@arm.com>
7872 PR tree-optimization/102977
7873 * tree-vect-slp-patterns.c (vect_match_call_p): Remove.
7874 (vect_detect_pair_op): Add crosslane check.
7875 (vect_match_call_complex_mla): Remove.
7876 (class complex_mul_pattern): Update comment.
7877 (complex_mul_pattern::matches): Update detection.
7878 (class complex_fma_pattern): Remove.
7879 (complex_fma_pattern::matches): Remove.
7880 (complex_fma_pattern::recognize): Remove.
7881 (complex_fma_pattern::build): Remove.
7882 (class complex_fms_pattern): Update comment.
7883 (complex_fms_pattern::matches): Remove.
7884 (complex_operations_pattern::recognize): Remove complex_fma_pattern
7886 2021-10-29 Jakub Jelinek <jakub@redhat.com>
7888 * gimple-fold.c (gimple_fold_builtin_memset): Copy over location from
7891 2021-10-29 Xionghu Luo <luoxhu@linux.ibm.com>
7894 * config/rs6000/rs6000.c (altivec_expand_vec_perm_const): Add
7895 patterns match and emit for VSX xxpermdi.
7897 2021-10-29 liuhongt <hongtao.liu@intel.com>
7900 * config/i386/i386-builtin-types.def (V8HF_FTYPE_V8HF): New
7902 (V16HF_FTYPE_V16HF): Ditto.
7903 (V32HF_FTYPE_V32HF): Ditto.
7904 (V8HF_FTYPE_V8HF_ROUND): Ditto.
7905 (V16HF_FTYPE_V16HF_ROUND): Ditto.
7906 (V32HF_FTYPE_V32HF_ROUND): Ditto.
7907 * config/i386/i386-builtin.def ( IX86_BUILTIN_FLOORPH,
7908 IX86_BUILTIN_CEILPH, IX86_BUILTIN_TRUNCPH,
7909 IX86_BUILTIN_FLOORPH256, IX86_BUILTIN_CEILPH256,
7910 IX86_BUILTIN_TRUNCPH256, IX86_BUILTIN_FLOORPH512,
7911 IX86_BUILTIN_CEILPH512, IX86_BUILTIN_TRUNCPH512): New builtin.
7912 * config/i386/i386-builtins.c
7913 (ix86_builtin_vectorized_function): Enable vectorization for
7914 HFmode FLOOR/CEIL/TRUNC operation.
7915 * config/i386/i386-expand.c (ix86_expand_args_builtin): Handle
7917 * config/i386/sse.md (rint<mode>2, nearbyint<mode>2): Extend
7920 2021-10-28 Aldy Hernandez <aldyh@redhat.com>
7921 Andrew MacLeod <amacleod@redhat.com>
7923 * value-relation.cc (path_oracle::killing_def): Walk the
7924 equivalency list and remove SSA from any equivalencies.
7926 2021-10-28 Stafford Horne <shorne@gmail.com>
7928 * config/or1k/or1k.h (PROFILE_HOOK): Add return address argument
7931 2021-10-28 Jakub Jelinek <jakub@redhat.com>
7933 PR tree-optimization/102951
7934 * fold-const.h (address_compare): Declare.
7935 * fold-const.c (address_compare): New function.
7936 * match.pd (cmp (convert1?@2 addr@0) (convert2? addr@1)): Use
7937 address_compare helper.
7938 (minmax cmp (convert1?@2 addr@0) (convert2?@3 addr@1)): New
7941 2021-10-28 Andrew MacLeod <amacleod@redhat.com>
7943 * vr-values.c (simplify_using_ranges::fold_cond): Change fold message.
7945 2021-10-28 Andrew MacLeod <amacleod@redhat.com>
7947 PR tree-optimization/102940
7948 * tree-vrp.c (execute_ranger_vrp): Reset scev.
7950 2021-10-28 Richard Purdie <richard.purdie@linuxfoundation.org>
7952 * config/nios2/linux.h (MUSL_DYNAMIC_LINKER): Add musl linker
7954 2021-10-28 Richard Purdie <richard.purdie@linuxfoundation.org>
7956 * configure: Regenerate.
7957 * configure.ac: Use CPPFLAGS_FOR_BUILD for GMPINC
7959 2021-10-28 Eric Botcazou <ebotcazou@adacore.com>
7961 * doc/invoke.texi (%X): Remove obsolete reference to -Wl.
7963 2021-10-28 Richard Biener <rguenther@suse.de>
7966 * fold-const.c (fold_convert_const): Avoid int to float
7967 constant folding with -frounding-math and inexact result.
7968 * simplify-rtx.c (simplify_const_unary_operation): Likewise
7969 for both float and unsigned_float.
7971 2021-10-28 Aldy Hernandez <aldyh@redhat.com>
7973 * tree-ssa-threadbackward.c
7974 (back_threader::find_taken_edge_switch): Use find_case_label_range
7975 instead of find_taken_edge.
7977 2021-10-28 Aldy Hernandez <aldyh@redhat.com>
7979 * tree-ssa-threadbackward.c (class back_threader_registry):
7980 Inherit from back_jt_path_registry.
7981 (back_threader_registry::thread_through_all_blocks): Remove.
7982 (back_threader_registry::register_path): Remove
7983 m_lowlevel_registry prefix.
7985 2021-10-28 Richard Biener <rguenther@suse.de>
7988 * fold-const.c (fold_convert_const_real_from_real): Honor
7989 -frounding-math if the conversion is not exact.
7990 * simplify-rtx.c (simplify_const_unary_operation): Do not
7991 simplify FLOAT_TRUNCATE with sign dependent rounding.
7993 2021-10-28 Richard Biener <rguenther@suse.de>
7995 PR tree-optimization/102949
7996 * tree-vect-stmts.c (ensure_base_align): Look at the
7997 dr_info of a group leader and assert we are looking at
7998 one with analyzed alignment.
8000 2021-10-28 Kewen Lin <linkw@linux.ibm.com>
8003 * config/rs6000/rs6000.c (rs6000_builtin_vectorization_cost): Consider
8004 V1T1 mode for unaligned load and store.
8006 2021-10-28 Kito Cheng <kito.cheng@sifive.com>
8008 * config/riscv/riscv.md (zero_extendsidi2_internal): Allow ZBB
8011 2021-10-28 Kito Cheng <kito.cheng@sifive.com>
8013 * config/riscv/arch-canonicalize (CANONICAL_ORDER): Add `i` to
8016 2021-10-28 Alexandre Oliva <oliva@adacore.com>
8018 * common.opt (fharden-compares): New.
8019 (fharden-conditional-branches): New.
8020 * doc/invoke.texi: Document new options.
8021 * gimple-harden-conditionals.cc: New.
8022 * Makefile.in (OBJS): Build it.
8023 * passes.def: Add new passes.
8024 * tree-pass.h (make_pass_harden_compares): Declare.
8025 (make_pass_harden_conditional_branches): Declare.
8027 2021-10-28 Xionghu Luo <luoxhu@linux.ibm.com>
8030 * config/rs6000/altivec.md: Add vsx register constraints.
8031 * config/rs6000/vsx.md (vsx_xxsel<mode>): Delete.
8032 (vsx_xxsel<mode>2): Likewise.
8033 (vsx_xxsel<mode>3): Likewise.
8034 (vsx_xxsel<mode>4): Likewise.
8036 2021-10-28 Xionghu Luo <luoxhu@linux.ibm.com>
8039 * config/rs6000/altivec.md (*altivec_vsel<mode>): Change to ...
8040 (altivec_vsel<mode>): ... this and update define.
8041 (*altivec_vsel<mode>_uns): Delete.
8042 (altivec_vsel<mode>2): New define_insn.
8043 (altivec_vsel<mode>3): Likewise.
8044 (altivec_vsel<mode>4): Likewise.
8045 * config/rs6000/rs6000-call.c (altivec_expand_vec_sel_builtin): New.
8046 (altivec_expand_builtin): Call altivec_expand_vec_sel_builtin to expand
8048 * config/rs6000/rs6000.c (rs6000_emit_vector_cond_expr): Use bit-wise
8049 selection instead of per element.
8050 * config/rs6000/vector.md:
8051 * config/rs6000/vsx.md (*vsx_xxsel<mode>): Change to ...
8052 (vsx_xxsel<mode>): ... this and update define.
8053 (*vsx_xxsel<mode>_uns): Delete.
8054 (vsx_xxsel<mode>2): New define_insn.
8055 (vsx_xxsel<mode>3): Likewise.
8056 (vsx_xxsel<mode>4): Likewise.
8058 2021-10-28 Hongyu Wang <hongyu.wang@intel.com>
8060 * config/i386/i386.c (use_rsqrt_p): Add mode parameter, enable
8061 HFmode rsqrt without TARGET_SSE_MATH.
8062 (ix86_optab_supported_p): Refactor rint, adjust floor, ceil,
8063 btrunc condition to be restricted by -ftrapping-math, adjust
8064 use_rsqrt_p function call.
8065 * config/i386/i386.md (rcphf2): New define_insn.
8066 (rsqrthf2): Likewise.
8067 * config/i386/sse.md (div<mode>3): Change VF2H to VF2.
8068 (div<mode>3): New expander for HF mode.
8069 (rsqrt<mode>2): Likewise.
8070 (*avx512fp16_vmrcpv8hf2): New define_insn for rpad pass.
8071 (*avx512fp16_vmrsqrtv8hf2): Likewise.
8073 2021-10-27 Saagar Jha <saagar@saagarjha.com>
8075 * config.gcc: Adjust for Darwin21.
8076 * config/darwin-c.c (macosx_version_as_macro): Likewise.
8077 * config/darwin-driver.c (validate_macosx_version_min):
8079 (darwin_find_version_from_kernel): Likewise.
8081 2021-10-27 Aldy Hernandez <aldyh@redhat.com>
8083 * gimple-range-path.cc
8084 (path_range_query::range_defined_in_block): Call killing_def.
8086 2021-10-27 Aldy Hernandez <aldyh@redhat.com>
8088 * gimple-range-fold.cc (fold_using_range::range_of_range_op): Dump
8089 operands as well as relation.
8090 * gimple-range-path.cc
8091 (path_range_query::compute_ranges_in_block): Compute PHI relations
8092 first. Compute outgoing relations at the end.
8093 (path_range_query::compute_ranges): Remove call to compute_relations.
8094 (path_range_query::compute_relations): Remove.
8095 (path_range_query::maybe_register_phi_relation): New.
8096 (path_range_query::compute_phi_relations): Abstract out
8097 registering one PHI relation to...
8098 (path_range_query::compute_outgoing_relations): ...here.
8099 * gimple-range-path.h (class path_range_query): Remove
8101 Add maybe_register_phi_relation.
8103 2021-10-27 Aldy Hernandez <aldyh@redhat.com>
8104 Andrew MacLeod <amacleod@redhat.com>
8106 * value-relation.cc (path_oracle::killing_def): Kill second
8109 2021-10-27 Martin Jambor <mjambor@suse.cz>
8111 * ipa-cp.c (good_cloning_opportunity_p): Decide whether to use
8112 profile feedback depending on their local availability.
8114 2021-10-27 Martin Jambor <mjambor@suse.cz>
8116 * params.opt (param_ipa_cp_profile_count_base): New parameter.
8117 * doc/invoke.texi (Optimize Options): Add entry for
8118 ipa-cp-profile-count-base.
8119 * ipa-cp.c (max_count): Replace with base_count, replace all
8120 occurrences too, unless otherwise stated.
8121 (ipcp_cloning_candidate_p): identify mostly-directly called
8122 functions based on their counts, not max_count.
8123 (compare_edge_profile_counts): New function.
8124 (ipcp_propagate_stage): Instead of setting max_count, find the
8125 appropriate edge count in a sorted vector of counts of eligible
8126 edges and make it the base_count.
8128 2021-10-27 Martin Jambor <mjambor@suse.cz>
8130 * ipa-cp.c (struct caller_statistics): New fields rec_count_sum,
8131 n_nonrec_calls and itself, document all fields.
8132 (init_caller_stats): Initialize the above new fields.
8133 (gather_caller_stats): Gather self-recursive counts and calls number.
8134 (get_info_about_necessary_edges): Gather counts of self-recursive and
8135 other edges bringing in the requested value separately.
8136 (dump_profile_updates): Rework to dump info about a single node only.
8137 (lenient_count_portion_handling): New function.
8138 (struct gather_other_count_struct): New type.
8139 (gather_count_of_non_rec_edges): New function.
8140 (struct desc_incoming_count_struct): New type.
8141 (analyze_clone_icoming_counts): New function.
8142 (adjust_clone_incoming_counts): Likewise.
8143 (update_counts_for_self_gen_clones): Likewise.
8144 (update_profiling_info): Rewritten.
8145 (update_specialized_profile): Adjust call to dump_profile_updates.
8146 (create_specialized_node): Do not update profiling info.
8147 (decide_about_value): New parameter self_gen_clones, either push new
8148 clones into it or updat their profile counts. For self-recursively
8149 generated values, use a portion of the node count instead of count
8150 from self-recursive edges to estimate goodness.
8151 (decide_whether_version_node): Gather clones for self-generated values
8152 in a new vector, update their profiles at once at the end.
8154 2021-10-27 Richard Biener <rguenther@suse.de>
8156 * tree-vectorizer.c (vect_transform_loops): New function,
8158 (try_vectorize_loop_1): ... here. Simplify as epilogues
8159 are now fully handled in the split part.
8161 2021-10-27 Jakub Jelinek <jakub@redhat.com>
8163 * omp-expand.c (expand_omp_for_init_counts): Handle non-rectangular
8164 iterators with pointer types.
8165 (expand_omp_for_init_vars, extract_omp_for_update_vars): Likewise.
8167 2021-10-26 Martin Sebor <msebor@redhat.com>
8169 PR tree-optimization/102238
8170 PR tree-optimization/102919
8171 * gimple-ssa-sprintf.c (get_string_length): Add an argument.
8172 (array_elt_at_offset): Move to pointer-query.
8173 (set_aggregate_size_and_offset): New function.
8174 (field_at_offset): Move to pointer-query.
8175 (get_origin_and_offset): Rename...
8176 (get_origin_and_offset_r): this. Add an argument. Make aggregate
8177 handling more robust.
8178 (get_origin_and_offset): New.
8179 (alias_offset): Add an argument.
8180 (format_string): Use subobject size determined by get_origin_and_offset.
8181 * pointer-query.cc (field_at_offset): Move from gimple-ssa-sprintf.c.
8182 Improve/correct handling of aggregates.
8183 (array_elt_at_offset): Same.
8184 * pointer-query.h (field_at_offset): Declare.
8185 (array_elt_at_offset): Declare.
8187 2021-10-26 Martin Sebor <msebor@redhat.com>
8189 * builtins.c (check_strncat_sizes): Pass access_data ctor additional
8191 (expand_builtin_memcmp): Move code to gimple-ssa-warn-access.cc.
8192 (expand_builtin_fork_or_exec): Same.
8193 * gimple-array-bounds.cc (array_bounds_checker::check_mem_ref): Pass
8194 compute_objsize additional arguments.
8195 (inbounds_memaccess_p): Same.
8196 (array_bounds_checker::check_array_bounds): Add an assert. Stash
8197 statement in a member.
8198 (check_array_bounds_dom_walker::before_dom_children): Same.
8199 * gimple-array-bounds.h (array_bounds_checker::m_stmt): New member.
8200 * gimple-ssa-sprintf.c (get_destination_size): Add an argument.
8201 (handle_printf_call): Pass a new argument.
8202 * gimple-ssa-warn-access.cc (get_size_range): Add an argument.
8203 (check_access): Add an argument and pass it along to callees.
8204 (check_read_access): Make a member function.
8205 (pass_waccess::check_strcat): Pass access_data ctor additional
8207 (pass_waccess::check_strncat): Same.
8208 (pass_waccess::check_stxcpy): Same.
8209 (pass_waccess::check_stxncpy): Same.
8210 (pass_waccess::check_strncmp): Same.
8211 (pass_waccess::check_read_access): Same.
8212 (pass_waccess::check_builtin): Same.
8213 (pass_waccess::maybe_check_access_sizes): Same.
8214 (pass_waccess::maybe_check_dealloc_call): Same.
8215 * gimple-ssa-warn-access.h (check_read_access): Declare a new
8217 * pointer-query.cc (compute_objsize_r): Add an argument.
8218 (gimple_call_return_array): Same.
8219 (gimple_call_alloc_size): Same.
8220 (access_ref::access_ref): Same.
8221 (access_ref::get_ref): Same.
8222 (pointer_query::get_ref): Same.
8223 (handle_min_max_size): Pass an arguments to callees.
8224 (handle_array_ref): Add an argument.
8225 (handle_mem_ref): Same.
8226 (compute_objsize): Same.
8227 * pointer-query.h (struct access_ref): Adjust signatures.
8228 (struct access_data): Same.
8229 (gimple_call_alloc_size): Add an argument.
8230 (gimple_parm_array_size): Same.
8231 (compute_objsize): Same.
8232 * tree-ssa-strlen.c (strlen_pass::adjust_last_stmt): Pass an additional
8233 argument to compute_objsize.
8234 (strlen_pass::maybe_warn_overflow): Same.
8235 (maybe_diag_stxncpy_trunc): Same.
8237 2021-10-26 Martin Sebor <msebor@redhat.com>
8239 PR middle-end/102453
8240 * gimple-ssa-warn-access.cc (pass_waccess::check_atomic_builtin): New.
8241 (pass_waccess::check_atomic_builtin): Call it.
8243 2021-10-26 Vladimir N. Makarov <vmakarov@redhat.com>
8245 PR rtl-optimization/102842
8246 * lra-constraints.c (match_reload): Ignore out in checking values
8248 (curr_insn_transform): Collect outputs before doing reloads of operands.
8250 2021-10-26 Paul A. Clarke <pc@us.ibm.com>
8253 * config/rs6000/x86intrin.h: Move some included headers to new
8254 headers. Include new immintrin.h instead of those headers.
8255 * config/rs6000/immintrin.h: New.
8256 * config/rs6000/x86gprintrin.h: New.
8257 * config.gcc (powerpc*-*-*): Add new headers to extra_headers.
8259 2021-10-26 Richard Biener <rguenther@suse.de>
8261 * tree-vectorizer.h (vect_create_addr_base_for_vector_ref):
8262 Remove byte_offset parameter.
8263 (vect_create_data_ref_ptr): Likewise.
8264 * tree-vect-data-refs.c (vect_create_addr_base_for_vector_ref):
8266 (vect_create_data_ref_ptr): Likewise.
8267 * tree-vect-stmts.c (vectorizable_store): Adjust.
8268 (vectorizable_load): Likewise.
8270 2021-10-26 Richard Biener <rguenther@suse.de>
8272 PR tree-optimization/96109
8273 * tree-vectorizer.h (dr_misalignment): Add optional offset
8275 * tree-vect-data-refs.c (dr_misalignment): Likewise. Remove
8276 offset applied for negative stride accesses.
8277 (vect_enhance_data_refs_alignment): Compute negative stride
8278 access offset and pass it to dr_misalignment.
8279 * tree-vect-stmts.c (get_negative_load_store_type): Pass
8280 negative offset to dr_misalignment.
8281 (get_group_load_store_type): Likewise.
8282 (get_load_store_type): Likewise.
8283 (vectorizable_store): Remove asserts about alignment.
8284 (vectorizable_load): Likewise.
8286 2021-10-26 Kewen Lin <linkw@linux.ibm.com>
8288 PR tree-optimization/102897
8289 * tree-ssa-forwprop.c (simplify_permutation): Remove a wrong assertion.
8291 2021-10-26 Richard Biener <rguenther@suse.de>
8293 * tree-vect-data-refs.c (vect_create_addr_base_for_vector_ref):
8294 Take offset in bytes.
8295 (vect_create_data_ref_ptr): Likewise.
8296 * tree-vect-loop-manip.c (get_misalign_in_elems): Multiply
8297 offset by element size.
8298 (vect_create_cond_for_align_checks): Likewise.
8299 * tree-vect-stmts.c (get_negative_load_store_type): Likewise.
8300 (vectorizable_load): Remove duplicate leftover from merge
8303 2021-10-26 Roger Sayle <roger@nextmovesoftware.com>
8305 * config/i386/i386-expand.c (ix86_expand_v1ti_shift): New helper
8306 function to expand V1TI mode logical shifts by integer constants.
8307 (ix86_expand_v1ti_rotate): New helper function to expand V1TI
8308 mode rotations by integer constants.
8309 * config/i386/i386-protos.h (ix86_expand_v1ti_shift,
8310 ix86_expand_v1ti_rotate): Prototype new functions here.
8311 * config/i386/sse.md (ashlv1ti3, lshrv1ti3, rotlv1ti3, rotrv1ti3):
8312 New TARGET_SSE2 expanders to implement V1TI shifts and rotations.
8314 2021-10-26 Aldy Hernandez <aldyh@redhat.com>
8316 * tree-ssa-threadbackward.c (back_threader::maybe_register_path):
8317 Avoid threading circular paths.
8319 2021-10-26 Aldy Hernandez <aldyh@redhat.com>
8321 * tree-ssa-threadbackward.c (back_threader::resolve_phi):
8322 Attempt to resolve all incoming paths to a PHI.
8323 (back_threader::resolve_def): Always return true for PHIs.
8325 2021-10-26 Aldy Hernandez <aldyh@redhat.com>
8327 * tree-ssa-threadbackward.c (back_threader::find_paths_to_names):
8328 Always try to resolve path without looking back.
8329 * tree-ssa-threadupdate.c (dump_jump_thread): Indidicate whether
8330 edge is a back edge.
8332 2021-10-26 Kewen Lin <linkw@linux.ibm.com>
8334 PR tree-optimization/102789
8335 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
8336 update inits of simd_lane_access.
8338 2021-10-25 Andrew MacLeod <amacleod@redhat.com>
8340 * tree-vrp.c (vrp_simplify_cond_using_ranges): Add return type and
8341 move to vr-values.c.
8342 (simplify_casted_conds): Move to vrp_folder class.
8343 (execute_vrp): Call via vrp_folder now.
8344 * vr-values.c (simplify_cond_using_ranges_1): Call simplify_casted_cond.
8345 (simplify_using_ranges::simplify_casted_cond): Relocate from tree-vrp.c.
8346 * vr-values.h (simplify_casted_cond): Add prototype.
8348 2021-10-25 Andrew MacLeod <amacleod@redhat.com>
8350 * tree-vrp.c (rvrp_folder::fold_stmt): If simplification fails, try
8353 2021-10-25 Paul A. Clarke <pc@us.ibm.com>
8355 * config/rs6000/smmintrin.h (_mm_testz_si128): Add "extern" to
8357 (_mm_testc_si128): Likewise.
8358 (_mm_testnzc_si128): Likewise.
8359 (_mm_blend_ps): Likewise.
8360 (_mm_blendv_ps): Likewise.
8361 (_mm_blend_pd): Likewise.
8362 (_mm_blendv_pd): Likewise.
8363 (_mm_ceil_pd): Likewise.
8364 (_mm_ceil_sd): Likewise.
8365 (_mm_ceil_ps): Likewise.
8366 (_mm_ceil_ss): Likewise.
8367 (_mm_floor_pd): Likewise.
8368 (_mm_floor_sd): Likewise.
8369 (_mm_floor_ps): Likewise.
8370 (_mm_floor_ss): Likewise.
8371 (_mm_minpos_epu16): Likewise.
8372 (_mm_mul_epi32): Likewise.
8373 (_mm_cvtepi8_epi16): Likewise.
8374 (_mm_packus_epi32): Likewise.
8375 (_mm_cmpgt_epi64): Likewise.
8377 2021-10-25 Roger Sayle <roger@nextmovesoftware.com>
8379 * simplify-rtx.c (simplify_binary_operation_1) [SS_ASHIFT]: Simplify
8380 shifts of the mode's smin_value and smax_value when the bit count
8381 operand doesn't have side-effects.
8382 [US_ASHIFT]: Likewise, simplify shifts of the mode's umax_value
8383 when the bit count operand doesn't have side-effects.
8384 (simplify_const_binary_operation) [SS_ASHIFT, US_ASHIFT]: Perform
8385 compile-time evaluation of saturating left shifts with constant
8388 2021-10-25 Andrew MacLeod <amacleod@redhat.com>
8390 * gimple-range.cc (gimple_ranger::export_global_ranges): Remove check
8393 2021-10-25 Andrew MacLeod <amacleod@redhat.com>
8395 * flag-types.h (enum ranger_debug): Adjust values.
8396 * params.opt (ranger_debug): Ditto.
8398 2021-10-25 Martin Jambor <mjambor@suse.cz>
8400 PR tree-optimization/102886
8401 * tree-sra.c (totally_scalarize_subtree): Fix the out of
8404 2021-10-25 Andrew Pinski <apinski@marvell.com>
8406 * tree-ssa-dce.c (simple_dce_from_worklist):
8407 Check stmt_unremovable_because_of_non_call_eh_p also
8408 before removing the statement.
8410 2021-10-25 Richard Biener <rguenther@suse.de>
8412 PR tree-optimization/102905
8413 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment):
8414 Use vect_supportable_dr_alignment again to determine whether
8415 an access is supported when not aligned.
8417 2021-10-25 Kito Cheng <kito.cheng@sifive.com>
8419 * config/riscv/riscv.c (riscv_rtx_costs): Handle cost model
8422 2021-10-25 Jim Wilson <jimw@sifive.com>
8423 Kito Cheng <kito.cheng@sifive.com>
8424 Jia-Wei Chen <jiawei@iscas.ac.cn>
8425 Shi-Hua Liao <shihua@iscas.ac.cn>
8427 * config/riscv/bitmanip.md (shiftm1): New.
8428 (*bset<mode>): Ditto.
8429 (*bset<mode>_mask): Ditto.
8430 (*bset<mode>_1): Ditto.
8431 (*bset<mode>_1_mask): Ditto.
8432 (*bseti<mode>): Ditto.
8433 (*bclr<mode>): Ditto.
8434 (*bclri<mode>): Ditto.
8435 (*binv<mode>): Ditto.
8436 (*binvi<mode>): Ditto.
8437 (*bext<mode>): Ditto.
8439 * config/riscv/predicates.md (splittable_const_int_operand):
8441 (single_bit_mask_operand): New.
8442 (not_single_bit_mask_operand): Ditto.
8443 (const31_operand): Ditto.
8444 (const63_operand): Ditto.
8445 * config/riscv/riscv.c (riscv_build_integer_1): Handle bseti.
8446 (riscv_output_move): Ditto.
8447 (riscv_print_operand): Handle new operand type: T and S.
8448 * config/riscv/riscv.h (SINGLE_BIT_MASK_OPERAND): New.
8450 2021-10-25 Jim Wilson <jimw@sifive.com>
8452 * config/riscv/riscv.c (riscv_build_integer_1): Build integer
8455 2021-10-25 Kito Cheng <kito.cheng@sifive.com>
8457 * config/riscv/riscv.c (riscv_extend_cost): Handle cost model
8459 (riscv_rtx_costs): Ditto.
8461 2021-10-25 Jim Wilson <jimw@sifive.com>
8462 Kito Cheng <kito.cheng@sifive.com>
8463 Jia-Wei Chen <jiawei@iscas.ac.cn>
8465 * config/riscv/bitmanip.md (bitmanip_bitwise): New.
8466 (bitmanip_minmax): New.
8467 (clz_ctz_pcnt): New.
8468 (bitmanip_optab): New.
8469 (bitmanip_insn): New.
8470 (*<optab>_not<mode>): New.
8471 (*xor_not<mode>): New.
8472 (<bitmanip_optab>si2): New.
8473 (*<bitmanip_optab>disi2): New.
8474 (<bitmanip_optab>di2): New.
8475 (*zero_extendhi<GPR:mode>2_bitmanip): New.
8476 (*extend<SHORT:mode><SUPERQI:mode>2_zbb): New.
8477 (*zero_extendhi<GPR:mode>2_zbb): New.
8480 (rotrsi3_sext): New.
8483 (rotlsi3_sext): New.
8484 (bswap<mode>2): New.
8485 (<bitmanip_optab><mode>3): New.
8486 * config/riscv/riscv.md (type): Add rotate.
8487 (zero_extendhi<GPR:mode>2): Change to define_expand pattern.
8488 (*zero_extendhi<GPR:mode>2): New.
8489 (extend<SHORT:mode><SUPERQI:mode>2): Change to define_expand pattern.
8490 (*extend<SHORT:mode><SUPERQI:mode>2): New.
8492 2021-10-25 Kito Cheng <kito.cheng@sifive.com>
8494 * config/riscv/riscv.c (riscv_extend_cost): Handle cost model
8496 (riscv_rtx_costs): Ditto.
8498 2021-10-25 Jim Wilson <jimw@sifive.com>
8499 Kito Cheng <kito.cheng@sifive.com>
8500 Jia-Wei Chen <jiawei@iscas.ac.cn>
8502 * config/riscv/bitmanip.md (*zero_extendsidi2_bitmanip): New.
8507 (riscv_rtx_costs): Ditto.
8508 * config/riscv/riscv.md: Include bitmanip.md
8509 (type): Add bitmanip bype.
8510 (zero_extendsidi2): Change to define_expand pattern.
8511 (*zero_extendsidi2_internal): New.
8512 (zero_extendsidi2_shifted): Disable for ZBA.
8514 2021-10-25 Kito Cheng <kito.cheng@sifive.com>
8516 * common/config/riscv/riscv-common.c (riscv_ext_version_table):
8517 Add zba, zbb, zbc and zbs.
8518 (riscv_ext_flag_table): Ditto.
8519 * config/riscv/riscv-opts.h (MASK_ZBA): New.
8523 (TARGET_ZBA): Ditto.
8524 (TARGET_ZBB): Ditto.
8525 (TARGET_ZBC): Ditto.
8526 (TARGET_ZBS): Ditto.
8527 * config/riscv/riscv.opt (riscv_zb_subext): New.
8529 2021-10-25 liuhongt <hongtao.liu@intel.com>
8532 * match.pd: Simplify (_Float16) sqrtf((float) a) to .SQRT(a)
8533 when direct_internal_fn_supported_p, similar for sqrt/sqrtl.
8535 2021-10-25 Richard Biener <rguenther@suse.de>
8537 PR tree-optimization/102920
8538 * tree-ssa-sccvn.h (expressions_equal_p): Add argument
8539 controlling VN_TOP matching behavior.
8540 * tree-ssa-sccvn.c (expressions_equal_p): Likewise.
8541 (vn_phi_eq): Do not optimistically match VN_TOP.
8543 2021-10-25 konglin1 <lingling.kong@intel.com>
8545 * config/i386/sse.md (fma_<mode>_fadd_fmul): Add new
8546 define_insn_and_split.
8547 (fma_<mode>_fadd_fcmul):Likewise
8548 (fma_<complexopname>_<mode>_fma_zero):Likewise
8550 2021-10-24 John David Anglin <danglin@gcc.gnu.org>
8552 * config/pa/pa-d.c (pa_d_handle_target_float_abi): Don't check
8553 TARGET_DISABLE_FPREGS.
8554 * config/pa/pa.c (fix_range): Use MASK_SOFT_FLOAT instead of
8555 MASK_DISABLE_FPREGS.
8556 (hppa_rtx_costs): Don't check TARGET_DISABLE_FPREGS. Adjust
8557 cost of hardware integer multiplication.
8558 (pa_conditional_register_usage): Don't check TARGET_DISABLE_FPREGS.
8559 * config/pa/pa.h (INT14_OK_STRICT): Likewise.
8560 * config/pa/pa.md: Don't check TARGET_DISABLE_FPREGS. Check
8561 TARGET_SOFT_FLOAT in patterns that use xmpyu instruction.
8562 * config/pa/pa.opt (mdisable-fpregs): Change target mask to
8563 SOFT_FLOAT. Revise comment.
8564 (msoft-float): New option.
8566 2021-10-24 John David Anglin <danglin@gcc.gnu.org>
8568 * config/pa/pa.md: Don't use 'G' constraint in integer move patterns.
8570 2021-10-24 Gerald Pfeifer <gerald@pfeifer.com>
8572 * doc/install.texi (Specific): Remove obsolete details
8573 around GNU/Linux on Itanium.
8574 (Specific): Remove reference to Windows for Itanium.
8576 2021-10-23 Bernhard Reutner-Fischer <aldot@gcc.gnu.org>
8578 * config/i386/x86-tune-sched-bd.c (dispatch_group): Commentary
8581 2021-10-23 Jan Hubicka <hubicka@ucw.cz>
8583 * tree-ssa-structalias.c (compute_points_to_sets): Cleanup.
8585 2021-10-23 Roger Sayle <roger@nextmovesoftware.com>
8587 * config/i386/sse.md (<any_logic>v1ti3): New define_insn to
8588 implement V1TImode AND, IOR and XOR on TARGET_SSE2 (and above).
8589 (one_cmplv1ti2): New define expand.
8591 2021-10-22 Eric Gallager <egallager@gcc.gnu.org>
8594 * Makefile.in: Handle dvidir and install-dvi target.
8595 * configure: Regenerate.
8596 * configure.ac: Add install-dvi to target_list.
8598 2021-10-22 Gerald Pfeifer <gerald@pfeifer.com>
8600 * doc/install.texi (Binaries): Convert mingw-w64.org to https.
8603 2021-10-22 Richard Biener <rguenther@suse.de>
8605 PR tree-optimization/102893
8606 * tree-ssa-dce.c (find_obviously_necessary_stmts): Fix the
8607 test for an exit edge.
8609 2021-10-22 Aldy Hernandez <aldyh@redhat.com>
8610 Andrew MacLeod <amacleod@redhat.com>
8612 * gimple-range-path.cc (path_range_query::compute_phi_relations):
8613 Kill any global relations we may know before registering a new
8615 * value-relation.cc (path_oracle::killing_def): New.
8616 * value-relation.h (path_oracle::killing_def): New.
8618 2021-10-22 Richard Biener <rguenther@suse.de>
8621 * tree-ssa-sccvn.c (vn_phi_insert): For undefined SSA args
8623 (vn_phi_lookup): Likewise.
8625 2021-10-21 H.J. Lu <hjl.tools@gmail.com>
8628 * doc/invoke.texi: Document -fcf-protection requires i686 or
8631 2021-10-21 Eric Botcazou <ebotcazou@adacore.com>
8633 PR middle-end/102764
8634 * cfgexpand.c (expand_gimple_basic_block): Robustify latest change.
8636 2021-10-21 Jonathan Wright <jonathan.wright@arm.com>
8638 * config/aarch64/arm_neon.h (__STRUCTN): Delete function
8639 macro and all invocations.
8641 2021-10-21 Andrew MacLeod <amacleod@redhat.com>
8643 * doc/invoke.texi (ranger-debug): Document.
8644 * flag-types.h (enum ranger_debug): New.
8645 (enum evrp_mode): Remove debug values.
8646 * gimple-range-cache.cc (DEBUG_RANGE_CACHE): Use new debug flag.
8647 * gimple-range-gori.cc (gori_compute::gori_compute): Ditto.
8648 * gimple-range.cc (gimple_ranger::gimple_ranger): Ditto.
8649 * gimple-ssa-evrp.c (hybrid_folder::choose_value): Ditto.
8650 (execute_early_vrp): Use evrp-mode directly.
8651 * params.opt (enum evrp_mode): Remove debug values.
8652 (ranger-debug): New.
8653 (ranger-logical-depth): Relocate to be in alphabetical order.
8655 2021-10-21 Andrew MacLeod <amacleod@redhat.com>
8657 * doc/invoke.texi: (vrp1-mode, vrp2-mode): Document.
8658 * flag-types.h: (enum vrp_mode): New.
8659 * params.opt: (vrp1-mode, vrp2-mode): New.
8660 * tree-vrp.c (vrp_pass_num): New.
8661 (pass_vrp::pass_vrp): Set pass number.
8662 (pass_vrp::execute): Choose which VRP mode to execute.
8664 2021-10-21 Andrew MacLeod <amacleod@redhat.com>
8666 * gimple-ssa-evrp.c (class rvrp_folder): Move to tree-vrp.c.
8667 (execute_early_vrp): For ranger only mode, invoke ranger_vrp.
8668 * tree-vrp.c (class rvrp_folder): Relocate here.
8669 (execute_ranger_vrp): New.
8670 * tree-vrp.h (execute_ranger_vrp): Export.
8672 2021-10-21 Martin Liska <mliska@suse.cz>
8676 * opts.c (finish_options): Process flag_var_tracking* options
8677 here as they can be adjusted by optimize attribute.
8678 Process also flag_syntax_only and flag_gtoggle.
8679 * toplev.c (process_options): Remove it here.
8680 * common.opt: Make debug_nonbind_markers_p as PerFunction
8681 attribute as it depends on optimization level.
8683 2021-10-21 Martin Jambor <mjambor@suse.cz>
8685 PR tree-optimization/102505
8686 * tree-sra.c (totally_scalarize_subtree): Check that the
8687 encountered field fits within the acces we would like to put it
8690 2021-10-21 Aldy Hernandez <aldyh@redhat.com>
8692 * tree-ssa-threadbackward.c
8693 (back_threader::maybe_register_path): Remove circular paths check.
8695 2021-10-21 Richard Biener <rguenther@suse.de>
8697 * toplev.c (process_options): Move the initial debug_hooks
8699 (toplev::main): ... before the call of the post_options
8702 2021-10-21 Richard Biener <rguenther@suse.de>
8704 PR tree-optimization/102847
8705 * tree-vect-stmts.c (vect_model_load_cost): Add the scalar
8706 load cost in the prologue for VMAT_INVARIANT.
8708 2021-10-21 Richard Biener <rguenther@suse.de>
8710 PR tree-optimization/102847
8711 * tree-vect-stmts.c (vect_model_load_cost): Explicitely
8712 handle VMAT_INVARIANT as a splat in the prologue.
8714 2021-10-21 Hongyu Wang <hongyu.wang@intel.com>
8717 * config/i386/i386.c (ix86_get_ssemov): Adjust HFmode vector
8718 move to use the same logic as HImode.
8720 2021-10-21 Richard Biener <rguenther@suse.de>
8722 * tree-vect-slp.c (vect_build_slp_tree_1): Remove
8723 superfluous gimple_call_nothrow_p check.
8725 2021-10-21 Andrew Pinski <apinski@marvell.com>
8727 * tree-cfg.c (maybe_remove_writeonly_store): Add dce_ssa_names argument.
8728 Mark the ssa-name of the rhs as one to be removed.
8729 (execute_fixup_cfg): Update call to maybe_remove_writeonly_store.
8730 Call simple_dce_from_worklist at the end to a simple dce.
8732 2021-10-21 Andrew Pinski <apinski@marvell.com>
8734 * tree-cfg.c (maybe_remove_writeonly_store): New function
8735 factored out from ...
8736 (execute_fixup_cfg): Here. Call maybe_remove_writeonly_store.
8738 2021-10-21 Andrew Pinski <apinski@marvell.com>
8740 * tree-cfg.c (execute_fixup_cfg): Remove comment
8741 about standalone pass.
8743 2021-10-21 Andrew Pinski <apinski@marvell.com>
8745 * tree-cfg.c (execute_fixup_cfg): Output when the statement
8746 is removed when it is a write only var.
8748 2021-10-21 Aldy Hernandez <aldyh@redhat.com>
8750 * tree-ssa-threadbackward.c (back_threader::maybe_register_path):
8751 Avoid threading circular paths.
8753 2021-10-20 Alex Coplan <alex.coplan@arm.com>
8755 * calls.c (initialize_argument_information): Remove some dead
8756 code, remove handling for function_arg returning const_int.
8757 * doc/tm.texi: Delete documentation for unused target hooks.
8758 * doc/tm.texi.in: Likewise.
8759 * target.def (load_bounds_for_arg): Delete.
8760 (store_bounds_for_arg): Delete.
8761 (load_returned_bounds): Delete.
8762 (store_returned_bounds): Delete.
8763 * targhooks.c (default_load_bounds_for_arg): Delete.
8764 (default_store_bounds_for_arg): Delete.
8765 (default_load_returned_bounds): Delete.
8766 (default_store_returned_bounds): Delete.
8767 * targhooks.h (default_load_bounds_for_arg): Delete.
8768 (default_store_bounds_for_arg): Delete.
8769 (default_load_returned_bounds): Delete.
8770 (default_store_returned_bounds): Delete.
8772 2021-10-20 Jonathan Wakely <jwakely@redhat.com>
8774 * doc/extend.texi (Basic Asm): Clarify that asm is not an
8776 * doc/invoke.texi (-fno-asm): Fix description for C++.
8778 2021-10-20 Jonathan Wakely <jwakely@redhat.com>
8780 * doc/install.texi: Remove link to old.html
8782 2021-10-20 Tamar Christina <tamar.christina@arm.com>
8784 * config/aarch64/aarch64-simd.md (*aarch64_cmtst_same_<mode>): New.
8786 2021-10-20 Tamar Christina <tamar.christina@arm.com>
8788 * config/aarch64/aarch64-simd.md (*aarch64_narrow_trunc<mode>): New.
8790 2021-10-20 Tamar Christina <tamar.christina@arm.com>
8792 * config/aarch64/aarch64-simd.md (aarch64_simd_ashr<mode>): Add case cmp
8794 * config/aarch64/constraints.md (D1): New.
8796 2021-10-20 Tamar Christina <tamar.christina@arm.com>
8798 * config/aarch64/aarch64-simd.md
8799 (*aarch64_<srn_op>topbits_shuffle<mode>_le): New.
8800 (*aarch64_topbits_shuffle<mode>_le): New.
8801 (*aarch64_<srn_op>topbits_shuffle<mode>_be): New.
8802 (*aarch64_topbits_shuffle<mode>_be): New.
8803 * config/aarch64/predicates.md
8804 (aarch64_simd_shift_imm_vec_exact_top): New.
8806 2021-10-20 Tamar Christina <tamar.christina@arm.com>
8808 * config/aarch64/aarch64-simd.md (*aarch64_<srn_op>shrn<mode>_vect,
8809 *aarch64_<srn_op>shrn<mode>2_vect_le,
8810 *aarch64_<srn_op>shrn<mode>2_vect_be): New.
8811 * config/aarch64/iterators.md (srn_op): New.
8813 2021-10-20 Chung-Lin Tang <cltang@codesourcery.com>
8815 * omp-low.c (omp_copy_decl_2): For !ctx, use record_vars to add new copy
8817 (scan_sharing_clauses): Place copy of OMP_CLAUSE_IN_REDUCTION decl in
8818 ctx->outer instead of ctx.
8820 2021-10-20 Martin Liska <mliska@suse.cz>
8823 2021-10-19 Martin Liska <mliska@suse.cz>
8826 * config/i386/i386-options.c (ix86_valid_target_attribute_inner_p): Strip whitespaces.
8827 * system.h (strip_whilespaces): New function.
8829 2021-10-20 Martin Liska <mliska@suse.cz>
8832 2021-10-19 Martin Liska <mliska@suse.cz>
8835 * config/aarch64/aarch64.c (aarch64_process_one_target_attr):
8838 2021-10-20 Wilco Dijkstra <wdijkstr@arm.com>
8840 * config/aarch64/aarch64.c (aarch64_case_values_threshold):
8841 Change to 8 with -Os, 11 otherwise.
8843 2021-10-20 Wilco Dijkstra <wdijkstr@arm.com>
8845 * config/aarch64/aarch64.c (neoversev1_tunings):
8846 Enable AARCH64_EXTRA_TUNE_CHEAP_SHIFT_EXTEND.
8847 (neoversen2_tunings): Likewise.
8849 2021-10-20 Wilco Dijkstra <wdijkstr@arm.com>
8852 * config/aarch64/aarch64.md (frint_pattern): Update comment.
8853 * config/aarch64/aarch64-simd-builtins.def: Change frintn to roundeven.
8854 * config/aarch64/arm_fp16.h: Change frintn to roundeven.
8855 * config/aarch64/arm_neon.h: Likewise.
8856 * config/aarch64/iterators.md (frint_pattern): Use roundeven for FRINTN.
8858 2021-10-20 Martin Liska <mliska@suse.cz>
8860 * config/arm/arm.c (arm_unwind_emit_sequence): Do not declare
8861 already declared global variable.
8862 (arm_unwind_emit_set): Use out_file as function argument.
8863 (arm_unwind_emit): Likewise.
8864 * config/darwin.c (machopic_output_data_section_indirection): Likewise.
8865 (machopic_output_stub_indirection): Likewise.
8866 (machopic_output_indirection): Likewise.
8867 (machopic_finish): Likewise.
8868 * config/i386/i386.c (ix86_asm_output_function_label): Likewise.
8869 * config/i386/winnt.c (i386_pe_seh_unwind_emit): Likewise.
8870 * config/ia64/ia64.c (process_epilogue): Likewise.
8871 (process_cfa_adjust_cfa): Likewise.
8872 (process_cfa_register): Likewise.
8873 (process_cfa_offset): Likewise.
8874 (ia64_asm_unwind_emit): Likewise.
8875 * config/s390/s390.c (s390_asm_output_function_label): Likewise.
8877 2021-10-20 Andre Simoes Dias Vieira <andre.simoesdiasvieira@arm.com>
8879 * config/aarch64/aarch64-builtins.c
8880 (aarch64_general_gimple_fold_builtin):
8881 lower vld1 and vst1 variants of the neon builtins
8882 * config/aarch64/aarch64-protos.h:
8883 (aarch64_general_gimple_fold_builtin): Add gsi parameter.
8884 * config/aarch64/aarch64.c (aarch64_general_gimple_fold_builtin):
8887 2021-10-20 Andre Simoes Dias Vieira <andre.simoesdiasvieira@arm.com>
8889 * match.pd: Generate IFN_TRUNC.
8891 2021-10-20 Richard Biener <rguenther@suse.de>
8893 PR tree-optimization/102853
8894 * tree-data-ref.c (split_constant_offset_1): Bail out
8895 immediately if the expression traps on overflow.
8897 2021-10-20 Aldy Hernandez <aldyh@redhat.com>
8899 * tree-ssa-threadbackward.c (back_threader::~back_threader): Remove.
8901 2021-10-20 Aldy Hernandez <aldyh@redhat.com>
8903 * tree-ssa-threadupdate.c (back_jt_path_registry::adjust_paths_after_duplication):
8904 Remove superflous debugging message.
8905 (back_jt_path_registry::duplicate_thread_path): Same.
8907 2021-10-20 Aldy Hernandez <aldyh@redhat.com>
8909 * tree-ssa-threadbackward.c (back_threader_registry::back_threader_registry):
8911 (back_threader_registry::register_path): Remove m_threaded_paths.
8913 2021-10-20 Aldy Hernandez <aldyh@redhat.com>
8915 PR tree-optimization/102814
8916 * doc/invoke.texi: Document --param=max-fsm-thread-length.
8917 * params.opt: Add --param=max-fsm-thread-length.
8918 * tree-ssa-threadbackward.c
8919 (back_threader_profitability::profitable_path_p): Fail on paths
8920 longer than max-fsm-thread-length.
8922 2021-10-20 Eric Botcazou <ebotcazou@adacore.com>
8924 PR middle-end/102764
8925 * cfgexpand.c (expand_gimple_basic_block): Disregard a final debug
8926 statement to reset the current location for the outgoing edges.
8928 2021-10-20 Aldy Hernandez <aldyh@redhat.com>
8929 Richard Biener <rguenther@suse.de>
8931 * tree-ssa-threadupdate.c (cancel_thread): Dump threading reason
8932 on the same line as the threading cancellation.
8933 (jt_path_registry::cancel_invalid_paths): Avoid rotating loops.
8934 Avoid threading through loop headers where the path remains in the
8937 2021-10-20 Siddhesh Poyarekar <siddhesh@gotplt.org>
8939 * tree-object-size.c (unknown): Make into a function. Adjust
8941 (unknown_object_size): Simplify implementation.
8943 2021-10-20 Hongtao Liu <hongtao.liu@intel.com>
8944 Kewen Lin <linkw@linux.ibm.com>
8946 * doc/sourcebuild.texi (Effective-Target Keywords): Document
8947 vect_slp_v2qi_store, vect_slp_v4qi_store, vect_slp_v8qi_store,
8948 vect_slp_v16qi_store, vect_slp_v2hi_store,
8949 vect_slp_v4hi_store, vect_slp_v2si_store, vect_slp_v4si_store.
8951 2021-10-19 Jonathan Wakely <jwakely@redhat.com>
8953 * doc/extend.texi (Basic PowerPC Built-in Functions): Fix typo.
8955 2021-10-19 Paul A. Clarke <pc@us.ibm.com>
8959 * config/rs6000/emmintrin.h: Guard POWER8 intrinsics.
8960 * config/rs6000/pmmintrin.h: Same.
8961 * config/rs6000/smmintrin.h: Same.
8962 * config/rs6000/tmmintrin.h: Same.
8964 2021-10-19 Paul A. Clarke <pc@us.ibm.com>
8966 * config.gcc (extra_headers): Add nmmintrin.h.
8968 2021-10-19 Richard Biener <rguenther@suse.de>
8970 * tree-vectorizer.h (vect_supportable_dr_alignment): Add
8971 misalignment parameter.
8972 * tree-vect-data-refs.c (vect_get_peeling_costs_all_drs):
8973 Do not change DR_MISALIGNMENT in place, instead pass the
8974 adjusted misalignment to vect_supportable_dr_alignment.
8975 (vect_peeling_supportable): Likewise.
8976 (vect_peeling_hash_get_lowest_cost): Adjust.
8977 (vect_enhance_data_refs_alignment): Likewise.
8978 (vect_vfa_access_size): Likewise.
8979 (vect_supportable_dr_alignment): Add misalignment
8980 parameter and simplify.
8981 * tree-vect-stmts.c (get_negative_load_store_type): Adjust.
8982 (get_group_load_store_type): Likewise.
8983 (get_load_store_type): Likewise.
8985 2021-10-19 Clément Chigot <clement.chigot@atos.net>
8987 * config/rs6000/rs6000.c (rs6000_xcoff_file_end): Move
8988 __tls_get_addr reference to .text csect.
8990 2021-10-19 Martin Liska <mliska@suse.cz>
8993 * config/aarch64/aarch64.c (aarch64_process_one_target_attr):
8996 2021-10-19 Richard Biener <rguenther@suse.de>
8998 * tree-vectorizer.h (vect_get_store_cost): Adjust signature.
8999 (vect_get_load_cost): Likewise.
9000 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
9001 alignment support scheme and misalignment as arguments
9003 (vect_get_peeling_costs_all_drs): Compute that info here
9004 and note that we shouldn't need to.
9005 * tree-vect-stmts.c (vect_model_store_cost): Get
9006 alignment support scheme and misalignment as arguments.
9007 (vect_get_store_cost): Likewise.
9008 (vect_model_load_cost): Likewise.
9009 (vect_get_load_cost): Likewise.
9010 (vectorizable_store): Pass down alignment support scheme
9011 and misalignment to costing.
9012 (vectorizable_load): Likewise.
9014 2021-10-19 Richard Biener <rguenther@suse.de>
9016 * tree-vect-stmts.c (get_negative_load_store_type): Add
9017 offset output parameter and initialize it.
9018 (get_group_load_store_type): Likewise.
9019 (get_load_store_type): Likewise.
9020 (vectorizable_store): Use offset as computed by
9021 get_load_store_type.
9022 (vectorizable_load): Likewise.
9024 2021-10-19 Richard Biener <rguenther@suse.de>
9026 PR tree-optimization/102827
9027 * tree-if-conv.c (predicate_statements): Add pe parameter
9028 and use that edge to insert invariant stmts on.
9029 (combine_blocks): Pass through pe.
9030 (tree_if_conversion): Compute the edge to insert invariant
9031 stmts on and pass it along.
9033 2021-10-19 Roger Sayle <roger@nextmovesoftware.com>
9036 * config/bfin/bfin.md (addsubv2hi3, subaddv2hi3, ssaddsubv2hi3,
9037 sssubaddv2hi3): Swap the order of operators in vec_concat.
9039 2021-10-19 Xionghu Luo <luoxhu@linux.ibm.com>
9041 * config/rs6000/altivec.md (*altivec_vmrghb_internal): Delete.
9042 (altivec_vmrghb_direct): New.
9043 (*altivec_vmrghh_internal): Delete.
9044 (altivec_vmrghh_direct): New.
9045 (*altivec_vmrghw_internal): Delete.
9046 (altivec_vmrghw_direct_<mode>): New.
9047 (altivec_vmrghw_direct): Delete.
9048 (*altivec_vmrglb_internal): Delete.
9049 (altivec_vmrglb_direct): New.
9050 (*altivec_vmrglh_internal): Delete.
9051 (altivec_vmrglh_direct): New.
9052 (*altivec_vmrglw_internal): Delete.
9053 (altivec_vmrglw_direct_<mode>): New.
9054 (altivec_vmrglw_direct): Delete.
9055 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Adjust.
9056 * config/rs6000/rs6000.c (altivec_expand_vec_perm_const):
9058 * config/rs6000/vsx.md (vsx_xxmrghw_<mode>): Adjust.
9059 (vsx_xxmrglw_<mode>): Adjust.
9061 2021-10-19 Aldy Hernandez <aldyh@redhat.com>
9063 * passes.def: Change threading comment before pass_ccp pass.
9065 2021-10-19 Haochen Gui <guihaoc@gcc.gnu.org>
9067 * config/rs6000/rs6000-call.c (altivec_expand_lxvr_builtin):
9068 Modify the expansion for sign extension. All extensions are done
9069 within VSX registers.
9071 2021-10-19 Richard Biener <rguenther@suse.de>
9073 * tree-vect-stmts.c (get_group_load_store_type): Add
9074 misalignment output parameter and initialize it.
9075 (get_group_load_store_type): Likewise.
9076 (vectorizable_store): Remove now redundant queries.
9077 (vectorizable_load): Likewise.
9079 2021-10-19 Richard Biener <rguenther@suse.de>
9081 * tree-vectorizer.h (vect_supportable_dr_alignment): Remove
9082 check_aligned argument.
9083 * tree-vect-data-refs.c (vect_supportable_dr_alignment):
9085 (vect_peeling_hash_insert): Add supportable_if_not_aligned
9086 argument and do not call vect_supportable_dr_alignment here.
9087 (vect_peeling_supportable): Adjust.
9088 (vect_enhance_data_refs_alignment): Compute whether the
9089 access is supported with different alignment here and
9090 pass that down to vect_peeling_hash_insert.
9091 (vect_vfa_access_size): Adjust.
9092 * tree-vect-stmts.c (vect_get_store_cost): Likewise.
9093 (vect_get_load_cost): Likewise.
9094 (get_negative_load_store_type): Likewise.
9095 (get_group_load_store_type): Likewise.
9096 (get_load_store_type): Likewise.
9098 2021-10-19 Martin Liska <mliska@suse.cz>
9101 * config/i386/i386-options.c (ix86_valid_target_attribute_inner_p): Strip whitespaces.
9102 * system.h (strip_whilespaces): New function.
9104 2021-10-19 dianhong xu <dianhong.xu@intel.com>
9106 * config/i386/avx512fp16intrin.h:
9107 (_mm512_set1_pch): New intrinsic.
9108 * config/i386/avx512fp16vlintrin.h:
9109 (_mm256_set1_pch): New intrinsic.
9110 (_mm_set1_pch): Ditto.
9112 2021-10-18 Andrew MacLeod <amacleod@redhat.com>
9114 PR tree-optimization/102796
9115 * gimple-range.cc (gimple_ranger::range_on_edge): Process EH edges
9116 normally. Return get_tree_range for non gimple_range_ssa_p names.
9117 (gimple_ranger::range_of_stmt): Use get_tree_range for non
9118 gimple_range_ssa_p names.
9120 2021-10-18 Uroš Bizjak <ubizjak@gmail.com>
9123 * config/i386/i386.c (ix86_print_operand_address):
9124 Error out for non-address_operand asm operands.
9126 2021-10-18 Richard Biener <rguenther@suse.de>
9128 * tree-vect-data-refs.c (vect_peeling_hash_insert): Do
9129 not auto-convert dr_alignment_support to bool.
9130 (vect_peeling_supportable): Likewise.
9131 (vect_enhance_data_refs_alignment): Likewise.
9132 (vect_supportable_dr_alignment): Commonize read/write case.
9133 * tree-vect-stmts.c (vect_get_store_cost): Use
9134 dr_alignment_support, not int, for the vect_supportable_dr_alignment
9136 (vect_get_load_cost): Likewise.
9138 2021-10-18 Siddhesh Poyarekar <siddhesh@gotplt.org>
9140 * tree-object-size.c (object_sizes_execute): Consolidate LHS
9141 null check and do it early.
9143 2021-10-18 Richard Biener <rguenther@suse.de>
9145 * tree-vect-stmts.c (vectorizable_store): Use the
9146 computed alignment scheme instead of querying
9149 2021-10-18 Richard Biener <rguenther@suse.de>
9151 * tree-vect-stmts.c (vectorizable_store): Do not recompute
9152 alignment scheme already determined by get_load_store_type.
9154 2021-10-18 Aldy Hernandez <aldyh@redhat.com>
9156 * tree-ssa-threadbackward.c (class pass_thread_jumps_full):
9157 Clone corresponding pass.
9159 2021-10-18 Roger Sayle <roger@nextmovesoftware.com>
9161 * combine.c (recog_for_combine): For an unrecognized move/set of
9162 a constant, try force_const_mem to place it in the constant pool.
9163 * cse.c (constant_pool_entries_cost, constant_pool_entries_regcost):
9164 Delete global variables (that are no longer assigned a cost value).
9165 (cse_insn): Simplify logic for deciding whether to place a folded
9166 constant in the constant pool using force_const_mem.
9167 (cse_main): Remove zero initialization of constant_pool_entries_cost
9168 and constant_pool_entries_regcost.
9169 * config/i386/i386.c (ix86_rtx_costs): Make memory accesses
9170 fractionally more expensive, when optimizing for speed.
9172 2021-10-18 Martin Liska <mliska@suse.cz>
9174 PR gcov-profile/102746
9175 PR gcov-profile/102747
9176 * gcov.c (main): Return return_code.
9177 (output_gcov_file): Mark return_code when error happens.
9178 (generate_results): Likewise.
9179 (read_graph_file): Likewise.
9180 (read_count_file): Likewise.
9182 2021-10-18 Roger Sayle <roger@nextmovesoftware.com>
9184 * config/bfin/bfin.md (define_constants): Remove UNSPEC_ONES.
9185 (define_insn "ones"): Replace UNSPEC_ONES with a truncate of
9186 a popcount, allowing compile-time evaluation/simplification.
9187 (popcountsi2, popcounthi2): New expanders using a "ones" insn.
9189 2021-10-18 Richard Biener <rguenther@suse.de>
9191 PR tree-optimization/102788
9192 * tree-vect-patterns.c (vect_init_pattern_stmt): Allow
9194 (vect_pattern_recog_1): Likewise.
9195 (vect_recog_bool_pattern): Continue matching the pattern
9196 even if we do not have a vector type for a conversion
9199 2021-10-18 Roger Sayle <roger@nextmovesoftware.com>
9201 * simplify-rtx.c (simplify_const_unary_operation) [SS_NEG, SS_ABS]:
9202 Evalute SS_NEG and SS_ABS of a constant argument.
9204 2021-10-18 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
9207 * gimple-match-head.c (try_conditional_simplification): Add case for single operand.
9208 * internal-fn.def: Add entry for COND_NEG internal function.
9209 * internal-fn.c (FOR_EACH_CODE_MAPPING): Add entry for
9210 NEGATE_EXPR, COND_NEG mapping.
9211 * optabs.def: Add entry for cond_neg_optab.
9212 * match.pd (UNCOND_UNARY, COND_UNARY): New operator lists.
9213 (vec_cond COND (foo A) B) -> (IFN_COND_FOO COND A B): New pattern.
9214 (vec_cond COND B (foo A)) -> (IFN_COND_FOO ~COND A B): Likewise.
9216 2021-10-18 Aldy Hernandez <aldyh@redhat.com>
9218 * tree-ssa-strlen.c (class strlen_pass): Rename from
9220 (handle_builtin_stxncpy_strncat): Move to strlen_pass.
9221 (handle_assign): Same.
9222 (adjust_last_stmt): Same.
9223 (maybe_warn_overflow): Same.
9224 (handle_builtin_strlen): Same.
9225 (handle_builtin_strchr): Same.
9226 (handle_builtin_strcpy): Same.
9227 (handle_builtin_strncat): Same.
9228 (handle_builtin_stxncpy_strncat): Same.
9229 (handle_builtin_memcpy): Same.
9230 (handle_builtin_strcat): Same.
9231 (handle_alloc_call): Same.
9232 (handle_builtin_memset): Same.
9233 (handle_builtin_memcmp): Same.
9234 (get_len_or_size): Same.
9235 (strxcmp_eqz_result): Same.
9236 (handle_builtin_string_cmp): Same.
9237 (handle_pointer_plus): Same.
9238 (count_nonzero_bytes_addr): Same.
9239 (count_nonzero_bytes): Same.
9240 (handle_store): Same.
9241 (strlen_check_and_optimize_call): Same.
9242 (handle_integral_assign): Same.
9243 (check_and_optimize_stmt): Same.
9244 (printf_strlen_execute): Rename strlen_dom_walker to strlen_pass.
9246 2021-10-18 Richard Biener <rguenther@suse.de>
9248 PR tree-optimization/102798
9249 * tree-vect-data-refs.c (vect_create_addr_base_for_vector_ref):
9250 Only copy points-to info to newly generated SSA names.
9252 2021-10-18 Martin Liska <mliska@suse.cz>
9254 * dbgcnt.c (dbg_cnt_process_opt): Remove unused but set variable.
9255 * gcov.c (get_cycles_count): Likewise.
9256 * lto-compress.c (lto_compression_zlib): Likewise.
9257 (lto_uncompression_zlib): Likewise.
9258 * targhooks.c (default_pch_valid_p): Likewise.
9260 2021-10-17 Aldy Hernandez <aldyh@redhat.com>
9262 * tree-pass.h (make_pass_thread_jumps_full): New.
9263 * tree-ssa-threadbackward.c (pass_thread_jumps::gate): Inline.
9264 (try_thread_blocks): Add resolve and speed arguments.
9265 (pass_thread_jumps::execute): Inline.
9266 (do_early_thread_jumps): New.
9267 (do_thread_jumps): New.
9268 (make_pass_thread_jumps): Move.
9269 (pass_early_thread_jumps::gate): Inline.
9270 (pass_early_thread_jumps::execute): Inline.
9271 (class pass_thread_jumps_full): New.
9273 2021-10-16 Piotr Kubaj <pkubaj@FreeBSD.org>
9275 * configure.ac: Treat powerpc64*-*-freebsd* the same as
9276 powerpc64-*-freebsd*.
9277 * configure: Regenerate.
9279 2021-10-16 H.J. Lu <hjl.tools@gmail.com>
9281 * value-query.cc (get_ssa_name_ptr_info_nonnull): Change
9282 set_ptr_nonull to set_ptr_nonnull in comments.
9284 2021-10-16 Jan Hubicka <hubicka@ucw.cz>
9286 PR tree-optimization/102720
9287 * tree-ssa-structalias.c (compute_points_to_sets): Fix producing
9288 of call used and clobbered sets.
9290 2021-10-15 Iain Sandoe <iain@sandoe.co.uk>
9292 * config/darwin.h (LINK_COMMAND_SPEC_A): Update 'r' handling to
9293 skip gomp and itm when r or nodefaultlibs is given.
9294 (DSYMUTIL_SPEC): Do not call dsymutil for '-r' link lines.
9295 Update ordering of exclusions, remove duplicate 'v' addition
9296 (collect2 will add this from the main command line).
9298 2021-10-15 Iain Sandoe <iain@sandoe.co.uk>
9300 * config/darwin-driver.c (darwin_driver_init): Revise comments, handle
9301 filelist and framework options in specs instead of code.
9302 * config/darwin.h (SUBTARGET_DRIVER_SELF_SPECS): Update to handle link
9303 specs that are really driver ones.
9304 (DARWIN_CC1_SPEC): Likewise.
9305 (CPP_SPEC): Likewise.
9306 (SYSROOT_SPEC): Append space.
9307 (LINK_SYSROOT_SPEC): Remove most driver link specs.
9308 (STANDARD_STARTFILE_PREFIX_2): Update link-related specs.
9309 (STARTFILE_SPEC): Likewise.
9310 (ASM_MMACOSX_VERSION_MIN_SPEC): Fix line wrap.
9311 (ASM_SPEC): Update driver-related specs.
9312 (ASM_FINAL_SPEC): Likewise.
9313 * config/darwin.opt: Remove now unused option aliases.
9314 * config/i386/darwin.h (EXTRA_ASM_OPTS): Ensure space after opt.
9315 (ASM_SPEC): Update driver-related specs.
9317 2021-10-15 Roger Sayle <roger@nextmovesoftware.com>
9319 * config/i386/i386.c (ix86_hardreg_mov_ok): For vector modes,
9320 allow standard_sse_constant_p immediate constants.
9322 2021-10-15 Iain Sandoe <iain@sandoe.co.uk>
9324 * config.gcc: Add tm-dwarf2.h to tm_d-file.
9326 2021-10-15 Andrew MacLeod <amacleod@redhat.com>
9328 * gimple-range-fold.h (gimple_range_ssa_p): Don't process names
9329 that occur in abnormal phis.
9330 * gimple-range.cc (gimple_ranger::range_on_edge): Return false for
9331 abnormal and EH edges.
9332 * gimple-ssa-evrp.c (rvrp_folder::value_of_expr): Ditto.
9333 (rvrp_folder::value_on_edge): Ditto.
9334 (rvrp_folder::value_of_stmt): Ditto.
9335 (hybrid_folder::value_of_expr): Ditto for ranger queries.
9336 (hybrid_folder::value_on_edge): Ditto.
9337 (hybrid_folder::value_of_stmt): Ditto.
9338 * value-query.cc (gimple_range_global): Always return a range if
9339 the type is supported.
9341 2021-10-15 John David Anglin <danglin@gcc.gnu.org>
9343 * config/pa/pa.md: Consistently use "rG" constraint for copy
9344 instruction in move patterns.
9346 2021-10-15 Andrew Stubbs <ams@codesourcery.com>
9348 * config/gcn/gcn-hsa.h (S_FIJI): Set unconditionally.
9351 * config/gcn/gcn.c: Hard code SRAM ECC settings for old architectures.
9352 * config/gcn/mkoffload.c (ELFABIVERSION_AMDGPU_HSA): Rename to ...
9353 (ELFABIVERSION_AMDGPU_HSA_V3): ... this.
9354 (ELFABIVERSION_AMDGPU_HSA_V4): New.
9355 (SET_SRAM_ECC_UNSUPPORTED): New.
9356 (copy_early_debug_info): Create elf flags to match the other objects.
9357 (main): Just let the attribute flags pass through.
9359 2021-10-15 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
9361 * tree-loop-distribution.c (reduction_var_overflows_first):
9362 Pass the type of reduction_var as first argument as it is also
9363 done for the load type.
9364 (loop_distribution::transform_reduction_loop): Add missing
9365 TREE_TYPE while determining precission of reduction_var.
9367 2021-10-15 Richard Biener <rguenther@suse.de>
9369 * defaults.h (PREFERRED_DEBUGGING_TYPE): Choose DWARF2_DEBUG
9371 * toplev.c (process_options): Warn when STABS debugging is
9372 enabled but not the preferred format.
9373 * config/pa/som.h (PREFERRED_DEBUGGING_TYPE): Define to
9375 * config/pdp11/pdp11.h (PREFERRED_DEBUGGING_TYPE): Likewise.
9377 2021-10-15 Richard Biener <rguenther@suse.de>
9380 * tree-inline.c (copy_bb): Avoid underflowing nargs.
9382 2021-10-15 Hongyu Wang <hongyu.wang@intel.com>
9384 * config/i386/i386-expand.c (ix86_expand_vec_perm): Convert
9385 HFmode input operand to HImode.
9386 (ix86_vectorize_vec_perm_const): Likewise.
9387 * config/i386/sse.md (*avx512bw_permvar_truncv16siv16hi_1_hf):
9389 (*avx512f_permvar_truncv8siv8hi_1_hf):
9392 2021-10-15 Richard Biener <rguenther@suse.de>
9394 PR middle-end/102682
9395 * expmed.c (store_bit_field_1): Ensure a LHS subreg would
9396 not create a paradoxical subreg.
9398 2021-10-15 Hongyu Wang <hongyu.wang@intel.com>
9400 * config/i386/i386-expand.c (ix86_expand_vector_init):
9401 For half_vector concat for HFmode, handle them like HImode.
9403 2021-10-14 Aldy Hernandez <aldyh@redhat.com>
9405 * tree-ssa-threadbackward.c (class back_threader): Add m_resolve.
9406 (back_threader::back_threader): Same.
9407 (back_threader::resolve_phi): Try to solve without looking back if
9409 (back_threader::find_paths_to_names): Same.
9410 (try_thread_blocks): Pass resolve argument to back threader.
9411 (pass_early_thread_jumps::execute): Same.
9413 2021-10-14 Aldy Hernandez <aldyh@redhat.com>
9415 * doc/invoke.texi: Remove max-fsm-thread-length,
9416 max-fsm-thread-paths, and fsm-maximum-phi-arguments.
9418 * tree-ssa-threadbackward.c (back_threader::back_threader): Remove
9420 (back_threader_registry::back_threader_registry): Same.
9421 (back_threader_profitability::profitable_path_p): Remove
9422 param_max_fsm_thread-length.
9423 (back_threader_registry::register_path): Remove
9424 m_max_allowable_paths.
9426 2021-10-14 Aldy Hernandez <aldyh@redhat.com>
9428 * tree-ssa-threadbackward.c (class back_threader): Make m_imports
9430 (back_threader::~back_threader): Do not release m_path.
9432 2021-10-14 Andrew MacLeod <amacleod@redhat.com>
9434 PR tree-optimization/102738
9435 * vr-values.c (simplify_using_ranges::simplify): Handle RSHIFT_EXPR.
9437 2021-10-14 Kwok Cheung Yeung <kcy@codesourcery.com>
9439 * omp-general.c (omp_check_context_selector): Move from c-omp.c.
9440 (omp_mark_declare_variant): Move from c-omp.c.
9441 (omp_context_name_list_prop): Update for Fortran strings.
9442 * omp-general.h (omp_check_context_selector): New prototype.
9443 (omp_mark_declare_variant): New prototype.
9445 2021-10-14 Richard Sandiford <richard.sandiford@arm.com>
9447 * config/rs6000/rs6000.c (rs6000_density_test): Move early
9448 exit test further up the function.
9450 2021-10-14 Richard Sandiford <richard.sandiford@arm.com>
9452 * config/arm/arm.c (arm_add_stmt_cost): Delete.
9453 (TARGET_VECTORIZE_ADD_STMT_COST): Delete.
9455 2021-10-14 Martin Jambor <mjambor@suse.cz>
9457 * doc/invoke.texi (Optimize Options): Add entry for
9458 ipa-cp-recursive-freq-factor.
9460 2021-10-14 Tamar Christina <tamar.christina@arm.com>
9462 * match.pd: New rule.
9464 2021-10-14 Jan Hubicka <hubicka@ucw.cz>
9467 * ipa-modref-tree.h (modref_access_node::update2):
9468 Also check that parm_offset is unchanged.
9469 (modref_ref_node::insert_access): Fix updating of
9472 2021-10-14 Aldy Hernandez <aldyh@redhat.com>
9474 * tree-ssa-threadbackward.c (back_threader::resolve_phi): Add
9477 2021-10-14 Richard Biener <rguenther@suse.de>
9479 PR tree-optimization/102659
9480 * tree-if-conv.c (if_convertible_gimple_assign_stmt_p): Also
9481 rewrite pointer typed undefined overflow operations.
9482 (predicate_statements): Likewise. Make sure to emit invariant
9483 conversions in the preheader.
9484 * tree-vectorizer.c (vect_loop_vectorized_call): Look through
9485 non-empty preheaders.
9486 * tree-data-ref.c (dr_analyze_indices): Strip useless
9487 conversions to the MEM_REF base type.
9489 2021-10-14 Martin Liska <mliska@suse.cz>
9491 * common.opt: Stop using AUTODETECT_VALUE
9492 and use EnabledBy where possible.
9493 * opts.c: Enable OPT_fvar_tracking with optimize >= 1.
9494 * toplev.c (AUTODETECT_VALUE): Remove macro.
9495 (process_options): Simplify by using EnabledBy and
9496 OPT_fvar_tracking. Use OPTION_SET_P macro instead of
9499 2021-10-14 Jonathan Wright <jonathan.wright@arm.com>
9501 * config/aarch64/arm_neon.h (vld1_s8_x3): Use signed type for
9503 (vld1_s32_x3): Likewise.
9505 2021-10-14 Aldy Hernandez <aldyh@redhat.com>
9507 PR tree-optimization/102736
9508 PR tree-optimization/102736
9509 * gimple-range-path.cc (path_range_query::range_on_path_entry):
9510 Assert that the requested range is defined outside the path.
9511 (path_range_query::ssa_range_in_phi): Do not call
9512 range_on_path_entry for SSA names that are defined within the
9515 2021-10-14 Iain Sandoe <iain@sandoe.co.uk>
9517 * config/darwin-driver.c (darwin_find_version_from_kernel):
9518 Quote internal identifiers and avoid contractions in
9520 (darwin_default_min_version): Likewise.
9521 (darwin_driver_init): Likewise.
9523 2021-10-14 Martin Jambor <mjambor@suse.cz>
9525 * params.opt (ipa-cp-recursive-freq-factor): New.
9526 * ipa-cp.c (ipcp_value): Switch to inline initialization. New members
9527 scc_no, self_recursion_generated_level, same_scc and
9528 self_recursion_generated_p.
9529 (ipcp_lattice::add_value): Replaced parameter unlimited with
9530 same_lat_gen_level, usit it determine limit of values and store it to
9532 (ipcp_lattice<valtype>::print): Dump the new fileds.
9533 (allocate_and_init_ipcp_value): Take same_lat_gen_level as a new
9534 parameter and store it to the new value.
9535 (self_recursively_generated_p): Removed.
9536 (propagate_vals_across_arith_jfunc): Use self_recursion_generated_p
9537 instead of self_recursively_generated_p, store self generation level
9539 (value_topo_info<valtype>::add_val): Set scc_no.
9540 (value_topo_info<valtype>::propagate_effects): Multiply frequencies of
9541 recursively feeding values and self generated values by appropriate
9544 2021-10-14 Richard Sandiford <richard.sandiford@arm.com>
9546 * config/aarch64/aarch64.c (aarch64_add_stmt_cost): Remove
9547 redundant test for flag_vect_cost_model.
9549 2021-10-14 Aldy Hernandez <aldyh@redhat.com>
9551 * bitmap.c (debug): New overloaded function for auto_bitmaps.
9552 * bitmap.h (debug): Same.
9554 2021-10-14 Hongyu Wang <hongyu.wang@intel.com>
9556 * config/i386/avx512fp16intrin.h (_mm512_mask_fcmadd_pch):
9557 Adjust builtin call.
9558 (_mm512_mask3_fcmadd_pch): Likewise.
9559 (_mm512_mask_fmadd_pch): Likewise
9560 (_mm512_mask3_fmadd_pch): Likewise
9561 (_mm512_mask_fcmadd_round_pch): Likewise
9562 (_mm512_mask3_fcmadd_round_pch): Likewise
9563 (_mm512_mask_fmadd_round_pch): Likewise
9564 (_mm512_mask3_fmadd_round_pch): Likewise
9565 (_mm_mask_fcmadd_sch): Likewise
9566 (_mm_mask3_fcmadd_sch): Likewise
9567 (_mm_mask_fmadd_sch): Likewise
9568 (_mm_mask3_fmadd_sch): Likewise
9569 (_mm_mask_fcmadd_round_sch): Likewise
9570 (_mm_mask3_fcmadd_round_sch): Likewise
9571 (_mm_mask_fmadd_round_sch): Likewise
9572 (_mm_mask3_fmadd_round_sch): Likewise
9573 (_mm_fcmadd_round_sch): Likewise
9574 * config/i386/avx512fp16vlintrin.h (_mm_mask_fmadd_pch):
9575 Adjust builtin call.
9576 (_mm_mask3_fmadd_pch): Likewise
9577 (_mm256_mask_fmadd_pch): Likewise
9578 (_mm256_mask3_fmadd_pch): Likewise
9579 (_mm_mask_fcmadd_pch): Likewise
9580 (_mm_mask3_fcmadd_pch): Likewise
9581 (_mm256_mask_fcmadd_pch): Likewise
9582 (_mm256_mask3_fcmadd_pch): Likewise
9583 * config/i386/i386-builtin.def: Add mask3 builtin for complex
9584 fma, and adjust mask_builtin to corresponding expander.
9585 * config/i386/i386-expand.c (ix86_expand_round_builtin):
9586 Skip eraseing embedded rounding for expanders that emits
9588 * config/i386/sse.md (complexmove): New mode_attr.
9589 (<avx512>_fmaddc_<mode>_mask1<round_expand_name>): New expander.
9590 (<avx512>_fcmaddc_<mode>_mask1<round_expand_name>): Likewise.
9591 (avx512fp16_fmaddcsh_v8hf_mask1<round_expand_name>): Likewise.
9592 (avx512fp16_fcmaddcsh_v8hf_mask1<round_expand_name>): Likewise.
9593 (avx512fp16_fcmaddcsh_v8hf_mask3<round_expand_name>): Likewise.
9594 (avx512fp16_fmaddcsh_v8hf_mask3<round_expand_name>): Likewise.
9595 * config/i386/subst.md (round_embedded_complex): New subst.
9597 2021-10-14 John David Anglin <danglin@gcc.gnu.org>
9599 * config/pa/pa.md (cbranchsf4): Disable if TARGET_SOFT_FLOAT.
9600 (cbranchdf4): Likewise.
9601 Add missing move patterns for TARGET_SOFT_FLOAT.
9603 2021-10-13 Roger Sayle <roger@nextmovesoftware.com>
9605 * config/i386/i386-expand.c (ix86_expand_vector_move): Use a
9606 pseudo intermediate when moving a SUBREG into a hard register,
9607 by checking ix86_hardreg_mov_ok.
9608 (ix86_expand_vector_extract): Store zero-extended SImode
9609 intermediate in a pseudo, then set target using a SUBREG_PROMOTED
9611 * config/i386/sse.md (mov<VMOVE>_internal): Prevent CSE creating
9612 complex (SUBREG) sets of (vector) hard registers before reload, by
9613 checking ix86_hardreg_mov_ok.
9615 2021-10-13 Indu Bhagat <indu.bhagat@oracle.com>
9617 * ctfc.h (enum ctf_dtu_d_union_enum): Remove redundant comma.
9619 2021-10-13 Indu Bhagat <indu.bhagat@oracle.com>
9621 * dwarf2ctf.c (gen_ctf_array_type): Fix typo in comment.
9623 2021-10-13 Martin Sebor <msebor@redhat.com>
9625 PR middle-end/102630
9626 * pointer-query.cc (compute_objsize_r): Handle named address spaces.
9628 2021-10-13 Iain Sandoe <iain@sandoe.co.uk>
9630 * collect2.c (is_lto_object_file): Release simple-object
9631 resources, close files.
9633 2021-10-13 Andre Vieira <andre.simoesdiasvieira@arm.com>
9635 * config/arm/arm.c (thumb2_legitimate_address_p): Use VALID_MVE_MODE
9636 when checking mve addressing modes.
9637 (mve_vector_mem_operand): Fix the way we handle pre, post and offset
9639 (arm_print_operand): Fix printing of POST_ and PRE_MODIFY.
9640 * config/arm/mve.md: Use mve_memory_operand predicate everywhere where
9641 there is a single Ux constraint.
9643 2021-10-13 John David Anglin <danglin@gcc.gnu.org>
9645 * config/pa/pa.md (muldi3): Add support for inlining 64-bit
9646 multiplication on 32-bit PA 1.1 and 2.0 targets.
9648 2021-10-13 Martin Liska <mliska@suse.cz>
9650 PR gcov-profile/90364
9651 * coverage.c (build_info): Emit checksum to the global variable.
9652 (build_info_type): Add new field for checksum.
9653 (coverage_obj_finish): Pass object_checksum.
9654 (coverage_init): Use 0 as checksum for .gcno files.
9655 * gcov-dump.c (dump_gcov_file): Dump also new checksum field.
9656 * gcov.c (read_graph_file): Read also checksum.
9657 * doc/invoke.texi: Document the behaviour change.
9659 2021-10-13 Richard Biener <rguenther@suse.de>
9661 * gimple-iterator.h (gsi_iterator_update): Add GSI_LAST_NEW_STMT,
9662 start at integer value 2.
9663 * gimple-iterator.c (gsi_insert_seq_nodes_before): Update
9664 the iterator for GSI_LAST_NEW_STMT.
9665 (gsi_insert_seq_nodes_after): Likewise.
9666 * tree-if-conv.c (predicate_statements): Use GSI_LAST_NEW_STMT.
9667 * tree-ssa.c (execute_update_addresses_taken): Correct bogus
9668 arguments to gsi_replace.
9670 2021-10-13 Martin Liska <mliska@suse.cz>
9673 * common.opt: Use EnabledBy instead of detection in
9674 finish_options and process_options.
9675 * opts.c (finish_options): Remove handling of
9676 x_flag_unroll_all_loops.
9677 * toplev.c (process_options): Likewise for flag_web and
9678 flag_rename_registers.
9680 2021-10-13 Richard Biener <rguenther@suse.de>
9682 PR tree-optimization/102659
9683 * tree-if-conv.c (need_to_rewrite_undefined): New flag.
9684 (if_convertible_gimple_assign_stmt_p): Mark the loop for
9685 rewrite when stmts with undefined behavior on integer
9687 (combine_blocks): Predicate also when we need to rewrite stmts.
9688 (predicate_statements): Rewrite affected stmts to something
9689 with well-defined behavior on overflow.
9690 (tree_if_conversion): Initialize need_to_rewrite_undefined.
9692 2021-10-13 Richard Biener <rguenther@suse.de>
9695 * ipa-sra.c (ptr_parm_has_nonarg_uses): Fix volatileness
9698 2021-10-13 Jose E. Marchesi <jose.marchesi@oracle.com>
9700 * dwarf2ctf.c: Fix typo in comment.
9702 2021-10-12 Bill Schmidt <wschmidt@linux.ibm.com>
9705 * config/rs6000/altivec.h (vec_cpsgn): Swap operand order.
9706 * config/rs6000/rs6000-overload.def (VEC_COPYSIGN): Use SKIP to
9707 avoid generating an automatic #define of vec_cpsgn. Use the
9708 correct built-in for V4SFmode that doesn't depend on VSX.
9710 2021-10-12 Uroš Bizjak <ubizjak@gmail.com>
9714 * config/i386/i386.md (*add<mode>_1_slp): Rewrite as
9715 define_insn_and_split pattern. Add alternative 1 and split it
9716 post reload to insert operand 1 into the low part of operand 0.
9717 (*sub<mode>_1_slp): Ditto.
9718 (*and<mode>_1_slp): Ditto.
9719 (*<any_or:code><mode>_1_slp): Ditto.
9720 (*ashl<mode>3_1_slp): Ditto.
9721 (*<any_shiftrt:insn><mode>3_1_slp): Ditto.
9722 (*<any_rotate:insn><mode>3_1_slp): Ditto.
9723 (*neg<mode>_1_slp): New insn_and_split pattern.
9724 (*one_cmpl<mode>_1_slp): Ditto.
9726 2021-10-12 David Edelsohn <dje.gcc@gmail.com>
9728 * doc/install.texi: Update MinGW and mingw-64 Binaries
9731 2021-10-12 Daniel Le Duc Khoi Nguyen <greenrecyclebin@gmail.com>
9733 * doc/extend.texi (Common Variable Attributes): Fix typos in
9734 alloc_size documentation.
9736 2021-10-12 Richard Biener <rguenther@suse.de>
9738 PR tree-optimization/102696
9739 * tree-vect-slp.c (vect_build_slp_tree_2): Properly mark
9740 the tree fatally failed when we reject a BIT_FIELD_REF.
9742 2021-10-12 Richard Biener <rguenther@suse.de>
9744 PR tree-optimization/102572
9745 * tree-vect-stmts.c (vect_build_gather_load_calls): When
9746 gathering the vectorized defs for the mask pass in the
9747 desired mask vector type so invariants will be handled
9750 2021-10-12 Tamar Christina <tamar.christina@arm.com>
9752 * config/aarch64/aarch64-sve.md (*fcm<cmp_op><mode>_bic_combine,
9753 *fcm<cmp_op><mode>_nor_combine, *fcmuo<mode>_bic_combine,
9754 *fcmuo<mode>_nor_combine): New.
9756 2021-10-12 Eric Botcazou <ebotcazou@adacore.com>
9759 * config/sparc/sparc-modes.def (OI): New integer mode.
9761 2021-10-12 Jakub Jelinek <jakub@redhat.com>
9763 * gimple-fold.h (clear_padding_type_may_have_padding_p): Declare.
9764 * gimple-fold.c (clear_padding_type_may_have_padding_p): No longer
9767 2021-10-12 Jakub Jelinek <jakub@redhat.com>
9769 * tree-vectorizer.h (loop_cost_model): New function.
9770 (unlimited_cost_model): Use it.
9771 * tree-vect-loop.c (vect_analyze_loop_costing): Use loop_cost_model
9772 call instead of flag_vect_cost_model.
9773 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
9774 (vect_prune_runtime_alias_test_list): Likewise. Also use it instead
9775 of flag_simd_cost_model.
9777 2021-10-12 liuhongt <hongtao.liu@intel.com>
9780 * config/i386/i386-expand.c (emit_reduc_half): Handle
9782 * config/i386/mmx.md (reduc_<code>_scal_v4qi): New expander.
9783 (reduc_plus_scal_v4qi): Ditto.
9785 2021-10-12 Paul A. Clarke <pc@us.ibm.com>
9787 * config/rs6000/smmintrin.h (_mm_cmpeq_epi64, _mm_cmpgt_epi64,
9788 _mm_mullo_epi32, _mm_mul_epi32, _mm_packus_epi32): New.
9789 * config/rs6000/nmmintrin.h: Copy from i386, tweak to suit.
9791 2021-10-12 Paul A. Clarke <pc@us.ibm.com>
9793 * config/rs6000/smmintrin.h (_mm_cvtepi8_epi16, _mm_cvtepi8_epi32,
9794 _mm_cvtepi8_epi64, _mm_cvtepi16_epi32, _mm_cvtepi16_epi64,
9795 _mm_cvtepi32_epi64, _mm_cvtepu8_epi16, _mm_cvtepu8_epi32,
9796 _mm_cvtepu8_epi64, _mm_cvtepu16_epi32, _mm_cvtepu16_epi64,
9797 _mm_cvtepu32_epi64): New.
9799 2021-10-12 Paul A. Clarke <pc@us.ibm.com>
9801 * config/rs6000/smmintrin.h (_mm_test_all_zeros,
9802 _mm_test_all_ones, _mm_test_mix_ones_zeros): Rewrite as macro.
9804 2021-10-12 Paul A. Clarke <pc@us.ibm.com>
9806 * config/rs6000/smmintrin.h (_mm_min_epi8, _mm_min_epu16,
9807 _mm_min_epi32, _mm_min_epu32, _mm_max_epi8, _mm_max_epu16,
9808 _mm_max_epi32, _mm_max_epu32): New.
9810 2021-10-11 Jan Hubicka <hubicka@ucw.cz>
9812 * ipa-modref-tree.h (struct modref_access_node): Revert
9814 (struct modref_ref_node): Likewise.
9816 2021-10-11 Jan Hubicka <hubicka@ucw.cz>
9818 * ipa-modref-tree.h (modref_tree::global_access_p): New member
9821 (implicint_const_eaf_flags,implicit_pure_eaf_flags,
9822 ignore_stores_eaf_flags): Move to ipa-modref.h
9823 (remove_useless_eaf_flags): Remove early exit on NOCLOBBER.
9824 (modref_summary::global_memory_read_p): New member function.
9825 (modref_summary::global_memory_written_p): New member function.
9826 * ipa-modref.h (modref_summary::global_memory_read_p,
9827 modref_summary::global_memory_written_p): Declare.
9828 (implicint_const_eaf_flags,implicit_pure_eaf_flags,
9829 ignore_stores_eaf_flags): move here.
9830 * tree-ssa-structalias.c: Include ipa-modref-tree.h, ipa-modref.h
9832 (handle_rhs_call): Rewrite.
9833 (handle_call_arg): New function.
9834 (determine_global_memory_access): New function.
9835 (handle_const_call): Remove
9836 (handle_pure_call): Remove
9837 (find_func_aliases_for_call): Update use of handle_rhs_call.
9838 (compute_points_to_sets): Handle global memory acccesses
9841 2021-10-11 Diane Meirowitz <diane.meirowitz@oracle.com>
9843 * doc/invoke.texi: Add link to UndefinedBehaviorSanitizer
9844 documentation, mention UBSAN_OPTIONS, similar to what is done
9845 for AddressSanitizer.
9847 2021-10-11 Richard Biener <rguenther@suse.de>
9849 PR middle-end/102683
9850 * internal-fn.c (expand_DEFERRED_INIT): Check for mode
9851 availability before building an integer type for storage
9854 2021-10-11 Richard Biener <rguenther@suse.de>
9856 PR middle-end/101480
9857 * gimple.c (gimple_call_fnspec): Do not mark operator new/delete
9860 2021-10-11 Martin Liska <mliska@suse.cz>
9862 * common.opt: Remove Init(2) for some options.
9863 * toplev.c (process_options): Do not use AUTODETECT_VALUE, but
9864 use rather OPTION_SET_P.
9866 2021-10-11 Martin Liska <mliska@suse.cz>
9868 * common.opt: Remove usage of IRA_REGION_AUTODETECT.
9869 * flag-types.h (enum ira_region): Likewise.
9870 * toplev.c (process_options): Use OPTION_SET_P instead of
9871 IRA_REGION_AUTODETECT.
9873 2021-10-11 Jakub Jelinek <jakub@redhat.com>
9875 * omp-low.c (omp_runtime_api_call): Handle omp_get_max_teams,
9876 omp_[sg]et_teams_thread_limit and omp_set_num_teams.
9878 2021-10-11 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
9880 * config/s390/s390-protos.h (s390_rawmemchr): Add prototype.
9881 * config/s390/s390.c (s390_rawmemchr): New function.
9882 * config/s390/s390.md (rawmemchr<SINT:mode>): New expander.
9883 * config/s390/vector.md (@vec_vfees<mode>): Basically a copy of
9884 the pattern vfees<mode> from vx-builtins.md.
9885 * config/s390/vx-builtins.md (*vfees<mode>): Remove.
9887 2021-10-11 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
9889 * builtins.c (get_memory_rtx): Change to external linkage.
9890 * builtins.h (get_memory_rtx): Add function prototype.
9891 * doc/md.texi (rawmemchr<mode>): Document.
9892 * internal-fn.c (expand_RAWMEMCHR): Define.
9893 * internal-fn.def (RAWMEMCHR): Add.
9894 * optabs.def (rawmemchr_optab): Add.
9895 * tree-loop-distribution.c (find_single_drs): Change return code
9896 behaviour by also returning true if no single store was found
9898 (loop_distribution::classify_partition): Respect the new return
9899 code behaviour of function find_single_drs.
9900 (loop_distribution::execute): Call new function
9901 transform_reduction_loop in order to replace rawmemchr or strlen
9902 like loops by calls into builtins.
9903 (generate_reduction_builtin_1): New function.
9904 (generate_rawmemchr_builtin): New function.
9905 (generate_strlen_builtin_1): New function.
9906 (generate_strlen_builtin): New function.
9907 (generate_strlen_builtin_using_rawmemchr): New function.
9908 (reduction_var_overflows_first): New function.
9909 (determine_reduction_stmt_1): New function.
9910 (determine_reduction_stmt): New function.
9911 (loop_distribution::transform_reduction_loop): New function.
9913 2021-10-11 Martin Liska <mliska@suse.cz>
9915 * tree.c (cl_option_hasher::hash): Use cl_optimization_hash
9916 and remove legacy hashing code.
9918 2021-10-11 Kito Cheng <kito.cheng@sifive.com>
9921 * builtins.c (maybe_emit_call_builtin___clear_cache): Allow
9922 CONST_INT for BEGIN and END, and use gcc_assert rather than
9925 2021-10-10 Jakub Jelinek <jakub@redhat.com>
9928 * var-tracking.c (add_stores): For cselib_sp_derived_value_p values
9929 use MO_VAL_SET if loc is not sp.
9931 2021-10-10 Andrew Pinski <apinski@marvell.com>
9933 PR tree-optimization/102622
9934 * match.pd: Swap the order of a?pow2cst:0 and a?-1:0 transformations.
9935 Swap the order of a?0:pow2cst and a?0:-1 transformations.
9937 2021-10-09 liuhongt <hongtao.liu@intel.com>
9940 * config/i386/i386-expand.c (ix86_valid_mask_cmp_mode): Handle
9942 (ix86_use_mask_cmp_p): Ditto.
9943 (ix86_expand_sse_movcc): Ditto.
9944 * config/i386/i386.md (setcc_hf_mask): New define_insn.
9945 (movhf_mask): Ditto.
9946 (UNSPEC_MOVCC_MASK): New unspec.
9947 * config/i386/sse.md (UNSPEC_PCMP): Move to i386.md.
9949 2021-10-08 Vladimir N. Makarov <vmakarov@redhat.com>
9951 PR rtl-optimization/102627
9952 * lra-constraints.c (split_reg): Use at least natural mode of hard reg.
9954 2021-10-08 Aldy Hernandez <aldyh@redhat.com>
9956 * gimple-range-cache.cc (non_null_ref::non_null_deref_p): Grow
9959 2021-10-08 Aldy Hernandez <aldyh@redhat.com>
9961 * value-range.cc (irange::debug): New.
9962 * value-range.h (irange::debug): New.
9964 2021-10-08 Richard Sandiford <richard.sandiford@arm.com>
9966 PR tree-optimization/102385
9967 * predict.h (change_edge_frequency): Declare.
9968 * predict.c (change_edge_frequency): New function.
9969 * tree-ssa-loop-manip.h (tree_transform_and_unroll_loop): Remove
9971 (tree_unroll_loop): Likewise.
9972 * gimple-loop-jam.c (tree_loop_unroll_and_jam): Update accordingly.
9973 * tree-predcom.c (pcom_worker::tree_predictive_commoning_loop):
9975 * tree-ssa-loop-prefetch.c (loop_prefetch_arrays): Likewise.
9976 * tree-ssa-loop-manip.c (tree_unroll_loop): Likewise.
9977 (tree_transform_and_unroll_loop): Likewise. Use single_dom_exit
9978 to retrieve the exit edges. Make all the old profile update code
9979 conditional on !single_loop_p -- the case it was written for --
9980 and use a different approach for the single-loop case.
9982 2021-10-08 Martin Liska <mliska@suse.cz>
9984 * config/alpha/alpha.c (alpha_option_override): Use new macro
9986 * config/arc/arc.c (arc_override_options): Likewise.
9987 * config/arm/arm.c (arm_option_override): Likewise.
9988 * config/bfin/bfin.c (bfin_load_pic_reg): Likewise.
9989 * config/c6x/c6x.c (c6x_option_override): Likewise.
9990 * config/csky/csky.c: Likewise.
9991 * config/darwin.c (darwin_override_options): Likewise.
9992 * config/frv/frv.c (frv_option_override): Likewise.
9993 * config/i386/djgpp.h: Likewise.
9994 * config/i386/i386.c (ix86_stack_protect_guard): Likewise.
9995 (ix86_max_noce_ifcvt_seq_cost): Likewise.
9996 * config/ia64/ia64.c (ia64_option_override): Likewise.
9997 (ia64_override_options_after_change): Likewise.
9998 * config/m32c/m32c.c (m32c_option_override): Likewise.
9999 * config/m32r/m32r.c (m32r_init): Likewise.
10000 * config/m68k/m68k.c (m68k_option_override): Likewise.
10001 * config/microblaze/microblaze.c (microblaze_option_override): Likewise.
10002 * config/mips/mips.c (mips_option_override): Likewise.
10003 * config/nios2/nios2.c (nios2_option_override): Likewise.
10004 * config/nvptx/nvptx.c (nvptx_option_override): Likewise.
10005 * config/pa/pa.c (pa_option_override): Likewise.
10006 * config/riscv/riscv.c (riscv_option_override): Likewise.
10007 * config/rs6000/aix71.h: Likewise.
10008 * config/rs6000/aix72.h: Likewise.
10009 * config/rs6000/aix73.h: Likewise.
10010 * config/rs6000/rs6000.c (darwin_rs6000_override_options): Likewise.
10011 (rs6000_override_options_after_change): Likewise.
10012 (rs6000_linux64_override_options): Likewise.
10013 (glibc_supports_ieee_128bit): Likewise.
10014 (rs6000_option_override_internal): Likewise.
10015 (rs6000_file_start): Likewise.
10016 (rs6000_darwin_file_start): Likewise.
10017 * config/rs6000/rtems.h: Likewise.
10018 * config/rs6000/sysv4.h: Likewise.
10019 * config/rs6000/vxworks.h (SUB3TARGET_OVERRIDE_OPTIONS): Likewise.
10020 * config/s390/s390.c (s390_option_override): Likewise.
10021 * config/sh/linux.h: Likewise.
10022 * config/sh/netbsd-elf.h (while): Likewise.
10023 * config/sh/sh.c (sh_option_override): Likewise.
10024 * config/sol2.c (solaris_override_options): Likewise.
10025 * config/sparc/sparc.c (sparc_option_override): Likewise.
10026 * config/tilegx/tilegx.c (tilegx_option_override): Likewise.
10027 * config/visium/visium.c (visium_option_override): Likewise.
10028 * config/vxworks.c (vxworks_override_options): Likewise.
10029 * lto-opts.c (lto_write_options): Likewise.
10030 * omp-expand.c (expand_omp_simd): Likewise.
10031 * omp-general.c (omp_max_vf): Likewise.
10032 * omp-offload.c (oacc_xform_loop): Likewise.
10033 * opts.h (OPTION_SET_P): Likewise.
10034 * targhooks.c (default_max_noce_ifcvt_seq_cost): Likewise.
10035 * toplev.c (process_options): Likewise.
10036 * tree-predcom.c: Likewise.
10037 * tree-sra.c (analyze_all_variable_accesses): Likewise.
10039 2021-10-08 liuhongt <hongtao.liu@intel.com>
10042 * config/i386/i386.c (ix86_optab_supported_p):
10043 Return true for HFmode.
10044 * match.pd: Simplify (_Float16) ceil ((double) x) to
10045 __builtin_ceilf16 (a) when a is _Float16 type and
10046 direct_internal_fn_supported_p.
10048 2021-10-08 liuhongt <hongtao.liu@intel.com>
10051 * config/i386/i386-expand.c (emit_reduc_half): Hanlde V4HImode.
10052 * config/i386/mmx.md (reduc_plus_scal_v4hi): New.
10053 (reduc_<code>_scal_v4hi): New.
10055 2021-10-08 liuhongt <hongtao.liu@intel.com>
10057 * common.opt (ftree-vectorize): Add Var(flag_tree_vectorize).
10058 * doc/invoke.texi (Options That Control Optimization): Update
10060 * opts.c (default_options_table): Enable auto-vectorization at
10061 O2 with very-cheap cost model.
10062 (finish_options): Use cheap cost model for
10063 explicit -ftree{,-loop}-vectorize.
10065 2021-10-07 Indu Bhagat <indu.bhagat@oracle.com>
10067 * ctfc.c (ctfc_delete_container): Free hash table contents.
10069 2021-10-07 Indu Bhagat <indu.bhagat@oracle.com>
10071 * toplev.c (process_options): Do not warn for GNU GIMPLE.
10073 2021-10-07 Siddhesh Poyarekar <siddhesh@gotplt.org>
10075 * tree-object-size.c (addr_object_size,
10076 compute_builtin_object_size): Drop PDECL and POFF arguments.
10077 (addr_object_size): Adjust calls.
10078 * tree-object-size.h (compute_builtin_object_size): Drop PDECL
10079 and POFF arguments.
10081 2021-10-07 Roger Sayle <roger@nextmovesoftware.com>
10083 * rtl.def (SMUL_HIGHPART, UMUL_HIGHPART): New RTX codes for
10084 representing signed and unsigned high-part multiplication resp.
10085 * simplify-rtx.c (simplify_binary_operation_1) [SMUL_HIGHPART,
10086 UMUL_HIGHPART]: Simplify high-part multiplications by zero.
10087 [SS_PLUS, US_PLUS, SS_MINUS, US_MINUS, SS_MULT, US_MULT,
10088 SS_DIV, US_DIV]: Similar simplifications for saturating
10090 (simplify_const_binary_operation) [SS_PLUS, US_PLUS, SS_MINUS,
10091 US_MINUS, SS_MULT, US_MULT, SMUL_HIGHPART, UMUL_HIGHPART]:
10092 Implement compile-time evaluation for constant operands.
10093 * dwarf2out.c (mem_loc_descriptor): Skip SMUL_HIGHPART and
10095 * doc/rtl.texi (smul_highpart, umul_highpart): Document RTX codes.
10096 * doc/md.texi (smul@var{m}3_highpart, umul@var{m3}_highpart):
10097 Mention the new smul_highpart and umul_highpart RTX codes.
10098 * doc/invoke.texi: Silence @xref "compilation" warnings.
10100 2021-10-07 Martin Jambor <mjambor@suse.cz>
10103 * ipa-prop.c (ipa_edge_args_sum_t::duplicate): Also handle the
10104 case when the source reference description corresponds to a
10105 referance taken in a function src->caller is inlined to.
10107 2021-10-07 Jan Hubicka <hubicka@ucw.cz>
10110 * ipa-modref-tree.h (modref_access_node::contains_p): Handle offsets
10112 (modref_access_node::try_merge_with): Add sanity check that there
10113 are no redundant entries in the list.
10115 2021-10-07 Richard Biener <rguenther@suse.de>
10117 PR tree-optimization/102608
10118 * tree-ssa-sccvn.c (visit_stmt): Drop .DEFERRED_INIT to
10121 2021-10-07 Martin Liska <mliska@suse.cz>
10123 * toplev.c (toplev::main): Make
10124 save_opt_decoded_options a pointer type
10125 * toplev.h: Likewise.
10127 2021-10-07 Andrew Stubbs <ams@codesourcery.com>
10129 * config/gcn/gcn-valu.md (gather<mode>_insn_2offsets<exec>): Apply
10130 HAVE_GCN_ASM_GLOBAL_LOAD_FIXED.
10131 (scatter<mode>_insn_2offsets<exec_scatter>): Likewise.
10133 2021-10-07 Andrew Stubbs <ams@codesourcery.com>
10135 * config/gcn/gcn-hsa.h (SRAMOPT): Include the whole option string.
10136 Adjust for new -msram-ecc=any behaviour.
10137 (ASM_SPEC): Adjust -mxnack and -msram-ecc usage.
10138 * config/gcn/gcn.c (output_file_start): Implement -msram-ecc=any.
10139 * config/gcn/mkoffload.c (EF_AMDGPU_XNACK): Rename to ...
10140 (EF_AMDGPU_XNACK_V3): ... this.
10141 (EF_AMDGPU_SRAM_ECC): Rename to ...
10142 (EF_AMDGPU_SRAM_ECC_V3): ... this.
10143 (EF_AMDGPU_FEATURE_XNACK_V4): New.
10144 (EF_AMDGPU_FEATURE_XNACK_UNSUPPORTED_V4): New.
10145 (EF_AMDGPU_FEATURE_XNACK_ANY_V4): New.
10146 (EF_AMDGPU_FEATURE_XNACK_OFF_V4): New.
10147 (EF_AMDGPU_FEATURE_XNACK_ON_V4): New.
10148 (EF_AMDGPU_FEATURE_SRAMECC_V4): New.
10149 (EF_AMDGPU_FEATURE_SRAMECC_UNSUPPORTED_V4): New.
10150 (EF_AMDGPU_FEATURE_SRAMECC_ANY_V4): New.
10151 (EF_AMDGPU_FEATURE_SRAMECC_OFF_V4): New.
10152 (EF_AMDGPU_FEATURE_SRAMECC_ON_V4): New.
10153 (SET_XNACK_ON): New.
10154 (SET_XNACK_OFF): New.
10156 (SET_SRAM_ECC_ON): New.
10157 (SET_SRAM_ECC_ANY): New.
10158 (SET_SRAM_ECC_OFF): New.
10159 (TEST_SRAM_ECC_ANY): New.
10160 (TEST_SRAM_ECC_ON): New.
10161 (main): Implement HSACOv4 and -msram-ecc=any.
10163 2021-10-07 Andrew Stubbs <ams@codesourcery.com>
10165 * config.in: Regenerate.
10166 * config/gcn/gcn-hsa.h (X_FIJI): New macro.
10167 (X_900): New macro.
10168 (X_906): New macro.
10169 (X_908): New macro.
10170 (A_FIJI): Rename to ...
10171 (S_FIJI): ... this.
10172 (A_900): Rename to ...
10174 (A_906): Rename to ...
10176 (A_908): Rename to ...
10178 (SRAMOPT): New macro.
10179 (ASM_SPEC): Adjust xnack option usage.
10180 * config/gcn/gcn.c (output_file_start): Adjust amdgcn_target usage.
10181 * configure: Regenerate.
10182 * configure.ac: Detect LLVM assembler dialect.
10184 2021-10-07 Richard Biener <rguenther@suse.de>
10186 * tree-pretty-print.c (dump_generic_node): Do not elide
10187 printing '&' when dumping with -gimple.
10189 2021-10-06 Andrew MacLeod <amacleod@redhat.com>
10191 * gimple-range-cache.cc (non_null_ref::adjust_range): Call new
10193 * gimple-range-fold.cc (adjust_pointer_diff_expr): Ditto.
10194 (adjust_imagpart_expr): Ditto.
10195 * value-range.cc (irange::irange_intersect): Call new routine if
10196 RHS is a single pair.
10197 (irange::intersect): New wide_int version.
10198 * value-range.h (class irange): New prototype.
10200 2021-10-06 Andrew MacLeod <amacleod@redhat.com>
10202 * gimple-range-edge.cc (gimple_outgoing_range::gimple_outgoing_range):
10203 Add parameter to limit size when recognizing switches.
10204 (gimple_outgoing_range::edge_range_p): Check size limit.
10205 * gimple-range-edge.h (gimple_outgoing_range): Add size field.
10206 * gimple-range-gori.cc (gori_map::calculate_gori): Ignore switches
10207 that exceed the size limit.
10208 (gori_compute::gori_compute): Add initializer.
10209 * params.opt (evrp-switch-limit): New.
10210 * doc/invoke.texi: Update docs.
10212 2021-10-06 Andrew MacLeod <amacleod@redhat.com>
10214 * value-range.h (irange::set_varying): Use TYPE_MIN_VALUE and
10215 TYPE_MAX_VALUE instead of creating new trees when possible.
10217 2021-10-06 Andrew MacLeod <amacleod@redhat.com>
10219 * gimple-range-cache.cc (non_null_ref::adjust_range): Check for
10220 zero and non-zero more efficently.
10222 2021-10-06 Richard Biener <rguenther@suse.de>
10225 * dumpfile.h (TDF_GIMPLE_VAL): New.
10226 (dump_flag): Re-order and adjust TDF_* flags. Make
10227 the enum uint32_t. Use std::underlying_type in the
10228 operator overloads.
10229 (optgroup_flag): Likewise for the operator overloads.
10230 * tree-pretty-print.c (dump_generic_node): Wrap ADDR_EXPR
10231 in _Literal if TDF_GIMPLE_VAL.
10232 * gimple-pretty-print.c (dump_gimple_assign): Add
10233 TDF_GIMPLE_VAL to flags when dumping operands where only
10234 is_gimple_val are allowed.
10235 (dump_gimple_cond): Likewise.
10237 2021-10-06 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
10239 * gimple-isel.cc (gimple_expand_vec_cond_expr): Remove redundant if
10242 2021-10-05 qing zhao <qing.zhao@oracle.com>
10244 PR middle-end/102359
10245 * gimplify.c (gimplify_decl_expr): Not add initialization for an
10246 auto variable when it has been initialized by frontend.
10248 2021-10-05 Aldy Hernandez <aldyh@redhat.com>
10250 * tree-ssa-threadupdate.c (jt_path_registry::cancel_invalid_paths):
10251 Loosen restrictions
10253 2021-10-05 Jan-Benedict Glaw <jbglaw@lug-owl.de>
10255 * common/config/avr/avr-common.c (avr_handle_option): Mark
10256 argument as ATTRIBUTE_UNUSED.
10258 2021-10-05 Jan-Benedict Glaw <jbglaw@lug-owl.de>
10260 * config/lm32/uclinux-elf.h (LINK_GCC_C_SEQUENCE_SPEC):
10261 Undefine before redefinition.
10263 2021-10-05 Richard Biener <rguenther@suse.de>
10265 * toplev.c (no_backend): Remove global var.
10266 (process_options): Pass in no_backend, move post_options
10267 langhook call to toplev::main.
10268 (do_compile): Pass in no_backend, move process_options call
10270 (toplev::run_self_tests): Check no_backend at the caller.
10271 (toplev::main): Call post_options and process_options
10272 split out from do_compile, do self-tests only if
10273 no_backend is initialized.
10275 2021-10-05 Richard Biener <rguenther@suse.de>
10277 * tree-cfg.c (dump_function_to_file): Dump the UID of the
10278 function as part of the name when requested.
10279 * tree-pretty-print.c (dump_function_name): Dump the UID when
10280 requested and the langhook produced the actual name.
10282 2021-10-05 Richard Biener <rguenther@suse.de>
10284 PR middle-end/102587
10285 PR middle-end/102285
10286 * internal-fn.c (expand_DEFERRED_INIT): Fall back to
10287 zero-initialization as last resort, use the constant
10288 size as given by the DEFERRED_INIT argument to build
10291 2021-10-04 Marek Polacek <polacek@redhat.com>
10294 * doc/invoke.texi: Document -Warray-compare.
10296 2021-10-04 Richard Biener <rguenther@suse.de>
10298 * gimplify.c (is_var_need_auto_init): DECL_HARD_REGISTER
10299 variables are not to be initialized.
10301 2021-10-04 Richard Biener <rguenther@suse.de>
10303 * expr.h (non_mem_decl_p): Declare.
10304 (mem_ref_refers_to_non_mem_p): Likewise.
10305 * expr.c (non_mem_decl_p): Export.
10306 (mem_ref_refers_to_non_mem_p): Likewise.
10307 * internal-fn.c (expand_DEFERRED_INIT): Do not expand the LHS
10308 but check the base with mem_ref_refers_to_non_mem_p
10309 and non_mem_decl_p.
10311 2021-10-04 Richard Biener <rguenther@suse.de>
10313 PR tree-optimization/102570
10314 * tree-ssa-sccvn.h (vn_reference_op_struct): Document
10315 we are using clique for the internal function code.
10316 * tree-ssa-sccvn.c (vn_reference_op_eq): Compare the
10317 internal function code.
10318 (print_vn_reference_ops): Print the internal function code.
10319 (vn_reference_op_compute_hash): Hash it.
10320 (copy_reference_ops_from_call): Record it.
10321 (visit_stmt): Remove the restriction around internal function
10323 (fully_constant_vn_reference_p): Use fold_const_call and handle
10324 internal functions.
10325 (vn_reference_eq): Compare call return types.
10326 * tree-ssa-pre.c (create_expression_by_pieces): Handle
10327 generating calls to internal functions.
10328 (compute_avail): Remove the restriction around internal function
10331 2021-10-04 Aldy Hernandez <aldyh@redhat.com>
10333 PR tree-optimization/102560
10334 * gimple-ssa-warn-alloca.c (alloca_call_type): Remove static
10335 marker for invalid_range.
10337 2021-10-04 Richard Biener <rguenther@suse.de>
10339 PR middle-end/102587
10340 * internal-fn.c (expand_DEFERRED_INIT): Guard register
10341 initialization path an avoid initializing VLA registers
10344 2021-10-04 Eric Botcazou <ebotcazou@adacore.com>
10346 * config/rs6000/vxworks.h (TARGET_INIT_LIBFUNCS): Delete.
10348 2021-10-03 Martin Liska <mliska@suse.cz>
10350 * toplev.c (toplev::main): Check opt_index if it is a part
10353 2021-10-02 Aldy Hernandez <aldyh@redhat.com>
10355 PR tree-optimization/102563
10356 * range-op.cc (operator_lshift::op1_range): Do not clobber
10359 2021-10-02 Martin Liska <mliska@suse.cz>
10361 * toplev.c (toplev::main): save_decoded_options[0] is program
10362 name and so it should be skipped.
10364 2021-10-01 Aldy Hernandez <aldyh@redhat.com>
10366 PR tree-optimization/102546
10367 * range-op.cc (operator_lshift::op1_range): Teach range-ops that
10368 X << Y is non-zero implies X is also non-zero.
10370 2021-10-01 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
10372 * config/aarch64/aarch64-cores.def (AARCH64_CORE): New
10374 * config/aarch64/aarch64-tune.md: Regenerate.
10375 * doc/invoke.texi: Update docs.
10377 2021-10-01 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
10379 * config/aarch64/aarch64-cores.def (AARCH64_CORE): New
10381 * config/aarch64/aarch64-tune.md: Regenerate.
10382 * doc/invoke.texi: Update docs.
10384 2021-10-01 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
10386 * config/aarch64/aarch64-cores.def (AARCH64_CORE): New
10388 * config/aarch64/aarch64-tune.md: Regenerate.
10389 * doc/invoke.texi: Update docs.
10391 2021-10-01 Martin Sebor <msebor@redhat.com>
10394 * doc/invoke.texi (-Waddress): Update.
10395 * gengtype.c (write_types): Avoid -Waddress.
10396 * poly-int.h (POLY_SET_COEFF): Avoid using null.
10398 2021-10-01 John David Anglin <danglin@gcc.gnu.org>
10401 * config/pa/pa.c (pa_option_override): Default to dwarf version 4
10404 2021-10-01 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
10406 * config/aarch64/aarch64.h (AARCH64_FL_V9): Update value.
10408 2021-10-01 Aldy Hernandez <aldyh@redhat.com>
10410 * gimple-range-path.cc (path_range_query::compute_ranges): Use
10412 * gimple-range-path.h (class path_range_query): Remove shadowed
10414 (path_range_query::get_path_oracle): New.
10416 2021-10-01 Jakub Jelinek <jakub@redhat.com>
10417 Richard Biener <rguenther@suse.de>
10419 PR sanitizer/102515
10420 * doc/invoke.texi (-fsanitize=integer-divide-by-zero): Remove
10421 INT_MIN / -1 division detection from here ...
10422 (-fsanitize=signed-integer-overflow): ... and add it here.
10424 2021-10-01 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
10426 * config/aarch64/aarch64-arches.def (AARCH64_ARCH): Added
10428 * config/aarch64/aarch64.h (AARCH64_FL_V9): New.
10429 (AARCH64_FL_FOR_ARCH9): New flags for Armv9-A.
10430 (AARCH64_ISA_V9): New ISA flag.
10431 * doc/invoke.texi: Update docs.
10433 2021-10-01 Martin Liska <mliska@suse.cz>
10435 * toplev.c (toplev::main): Save decoded optimization options.
10436 * toplev.h (save_opt_decoded_options): New.
10437 * doc/extend.texi: Be more clear about optimize and target
10440 2021-10-01 Eric Botcazou <ebotcazou@adacore.com>
10442 * explow.c: Include langhooks.h.
10443 (set_stack_check_libfunc): Build a proper function type.
10445 2021-10-01 Eric Botcazou <ebotcazou@adacore.com>
10448 * config/i386/i386.c (legitimate_pic_address_disp_p): For PE-COFF do
10449 not return true for external weak function symbols in medium model.
10451 2021-10-01 Jakub Jelinek <jakub@redhat.com>
10453 * tree.h (OMP_CLAUSE_ORDER_REPRODUCIBLE): Define.
10454 * tree-pretty-print.c (dump_omp_clause) <case OMP_CLAUSE_ORDER>: Print
10455 reproducible: for OMP_CLAUSE_ORDER_REPRODUCIBLE.
10456 * omp-general.c (omp_extract_for_data): If OMP_CLAUSE_ORDER is seen
10457 without OMP_CLAUSE_ORDER_UNCONSTRAINED, overwrite sched_kind to
10458 OMP_CLAUSE_SCHEDULE_STATIC.
10460 2021-10-01 Richard Biener <rguenther@suse.de>
10462 PR middle-end/102518
10463 * tree-inline.c (setup_one_parameter): Avoid substituting
10464 an invariant into contexts where a GIMPLE register is not valid.
10466 2021-09-30 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
10468 * config/arm/arm-cpus.in: Add Cortex-R52+ CPU.
10469 * config/arm/arm-tables.opt: Regenerate.
10470 * config/arm/arm-tune.md: Regenerate.
10471 * doc/invoke.texi: Update docs.
10473 2021-09-30 Uroš Bizjak <ubizjak@gmail.com>
10476 * config/i386/i386.md
10477 (sign_extend:WIDE (any_logic:NARROW (memory, immediate)) splitters):
10480 2021-09-30 Tobias Burnus <tobias@codesourcery.com>
10482 * omp-low.c (omp_runtime_api_call): Add omp_aligned_{,c}alloc and
10483 omp_{c,re}alloc, fix omp_alloc/omp_free.
10485 2021-09-30 Martin Liska <mliska@suse.cz>
10487 * defaults.h (ASM_OUTPUT_ASCII): Do not hide global variable
10488 asm_out_file and stream directly to MYFILE.
10490 2021-09-30 Richard Biener <rguenther@suse.de>
10492 * tree-vect-data-refs.c (vect_update_misalignment_for_peel):
10493 Restore and fix condition under which we apply npeel to
10494 the DRs misalignment value.
10496 2021-09-30 Richard Biener <rguenther@suse.de>
10498 * tree-vect-data-refs.c (vect_update_misalignment_for_peel):
10499 Fix npeel check for variable amount of peeling.
10501 2021-09-30 Aldy Hernandez <aldyh@redhat.com>
10503 * lto-wrapper.c (run_gcc): Plug snprintf overflow.
10505 2021-09-30 Aldy Hernandez <aldyh@redhat.com>
10507 * gimple-range.cc (gimple_ranger::debug): New.
10508 * gimple-range.h (class gimple_ranger): Add debug.
10510 2021-09-30 Aldy Hernandez <aldyh@redhat.com>
10512 PR middle-end/102519
10513 * tree-vrp.c (hybrid_threader::~hybrid_threader): Free m_query.
10515 2021-09-29 Indu Bhagat <indu.bhagat@oracle.com>
10518 * btfout.c (GTY): Add GTY (()) albeit for cosmetic only purpose.
10519 (btf_finalize): Empty the hash_map btf_var_ids.
10521 2021-09-29 Aldy Hernandez <aldyh@redhat.com>
10523 * tree-vrp.c (thread_through_all_blocks): Return bool.
10524 (execute_vrp_threader): Return TODO_* flags.
10525 (pass_data_vrp_threader): Set todo_flags_finish to 0.
10527 2021-09-29 Aldy Hernandez <aldyh@redhat.com>
10529 * timevar.def (TV_TREE_VRP_THREADER): New.
10530 * tree-vrp.c: Use TV_TREE_VRP_THREADER for VRP threader pass.
10532 2021-09-29 David Faust <david.faust@oracle.com>
10534 * config.gcc (bpf-*-*): Do not overwrite extra_headers.
10536 2021-09-29 Jonathan Wright <jonathan.wright@arm.com>
10538 * config/aarch64/aarch64-builtins.c (TYPES_BINOP_PPU): Define
10539 new type qualifier enum.
10540 (TYPES_TERNOP_SSSU): Likewise.
10541 (TYPES_TERNOP_PPPU): Likewise.
10542 * config/aarch64/aarch64-simd-builtins.def: Define PPU, SSU,
10543 PPPU and SSSU builtin generator macros for qtbl1 and qtbx1
10545 * config/aarch64/arm_neon.h (vqtbl1_p8): Use type-qualified
10546 builtin and remove casts.
10547 (vqtbl1_s8): Likewise.
10548 (vqtbl1q_p8): Likewise.
10549 (vqtbl1q_s8): Likewise.
10550 (vqtbx1_s8): Likewise.
10551 (vqtbx1_p8): Likewise.
10552 (vqtbx1q_s8): Likewise.
10553 (vqtbx1q_p8): Likewise.
10554 (vtbl1_p8): Likewise.
10555 (vtbl2_p8): Likewise.
10556 (vtbx2_p8): Likewise.
10558 2021-09-29 Richard Biener <rguenther@suse.de>
10560 * tree-vect-data-refs.c (vect_dr_misalign_for_aligned_access):
10562 (vect_update_misalignment_for_peel): Use it to update
10563 misaligned to the value necessary for an aligned access.
10564 (vect_get_peeling_costs_all_drs): Likewise.
10565 (vect_enhance_data_refs_alignment): Likewise.
10567 2021-09-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10569 * config/aarch64/aarch64.c (aarch64_expand_cpymem): Count number of
10570 emitted operations and adjust heuristic for code size.
10572 2021-09-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10574 * config/aarch64/aarch64.c (aarch64_expand_setmem): Count number of
10575 emitted operations and adjust heuristic for code size.
10577 2021-09-29 Jakub Jelinek <jakub@redhat.com>
10579 PR middle-end/102504
10580 * gimplify.c (gimplify_scan_omp_clauses): Use omp_check_private even
10581 in OMP_SCOPE clauses, not just on worksharing construct clauses.
10583 2021-09-28 Geng Qi <gengqi@linux.alibaba.com>
10585 * config/riscv/riscv.md (mulv<mode>4): Call gen_smul<mode>3_highpart.
10586 (<u>mulditi3): Call <su>muldi3_highpart.
10587 (<u>muldi3_highpart): Rename to <su>muldi3_highpart.
10588 (<u>mulsidi3): Call <su>mulsi3_highpart.
10589 (<u>mulsi3_highpart): Rename to <su>mulsi3_highpart.
10591 2021-09-28 Iain Sandoe <iain@sandoe.co.uk>
10593 * config/darwin.h (DSYMUTIL_SPEC): Recognize D sources.
10595 2021-09-28 Iain Sandoe <iain@sandoe.co.uk>
10597 * config/rs6000/darwin.h (FIXED_R13): Add for PPC64.
10598 (FIRST_SAVED_GP_REGNO): Save from R13 even when it is one
10601 2021-09-28 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10603 * config/aarch64/aarch64.h (AARCH64_FL_LS64): Define
10604 (AARCH64_FL_V8_7): Likewise.
10605 (AARCH64_FL_FOR_ARCH8_7): Likewise.
10606 * config/aarch64/aarch64-arches.def (armv8.7-a): Define.
10607 * config/aarch64/aarch64-option-extensions.def (ls64): Define.
10608 * doc/invoke.texi: Document the above.
10610 2021-09-28 Aldy Hernandez <aldyh@redhat.com>
10612 * dbgcnt.c (dbg_cnt_counter): New.
10613 * dbgcnt.h (dbg_cnt_counter): New.
10614 * dumpfile.c (dump_options): Add entry for TDF_THREADING.
10615 * dumpfile.h (enum dump_flag): Add TDF_THREADING.
10616 * gimple-range-path.cc (DEBUG_SOLVER): Use TDF_THREADING.
10617 * tree-ssa-threadupdate.c (dump_jump_thread_path): Dump out
10620 2021-09-28 Aldy Hernandez <aldyh@redhat.com>
10622 * cfgcleanup.c (pass_jump::execute): Check
10623 flag_expensive_optimizations.
10624 (pass_jump_after_combine::gate): Same.
10625 * doc/invoke.texi (-fthread-jumps): Enable for -O1.
10626 * opts.c (default_options_table): Enable -fthread-jumps at -O1.
10627 * tree-ssa-threadupdate.c
10628 (fwd_jt_path_registry::remove_jump_threads_including): Bail unless
10631 2021-09-28 Ilya Leoshkevich <iii@linux.ibm.com>
10633 * tree-ssa-reassoc.c (biased_names): New global.
10634 (propagate_bias_p): New function.
10635 (loop_carried_phi): Remove.
10636 (propagate_rank): Propagate bias along single uses.
10637 (get_rank): Update biased_names when needed.
10639 2021-09-28 Ilya Leoshkevich <iii@linux.ibm.com>
10641 * passes.def (pass_reassoc): Rename parameter to early_p.
10642 * tree-ssa-reassoc.c (reassoc_bias_loop_carried_phi_ranks_p):
10644 (phi_rank): Don't bias loop-carried phi ranks
10645 before vectorization pass.
10646 (execute_reassoc): Add bias_loop_carried_phi_ranks_p parameter.
10647 (pass_reassoc::pass_reassoc): Add bias_loop_carried_phi_ranks_p
10649 (pass_reassoc::set_param): Set bias_loop_carried_phi_ranks_p
10651 (pass_reassoc::execute): Pass bias_loop_carried_phi_ranks_p to
10653 (pass_reassoc::bias_loop_carried_phi_ranks_p): New member.
10655 2021-09-28 Jakub Jelinek <jakub@redhat.com>
10658 * config/i386/i386.c (standard_80387_constant_p): Don't recognize
10659 special 80387 instruction XFmode constants if flag_rounding_math.
10661 2021-09-28 Richard Biener <rguenther@suse.de>
10663 PR tree-optimization/100112
10664 * tree-ssa-sccvn.c (visit_reference_op_load): Record the
10665 referece into the hashtable twice in case last_vuse is
10666 different from the original vuse on the stmt.
10668 2021-09-28 Jakub Jelinek <jakub@redhat.com>
10670 PR middle-end/102492
10671 * gimplify.c (gimplify_adjust_omp_clauses_1): Don't call the
10672 omp_finish_clause langhook on implicitly added OMP_CLAUSE_PRIVATE
10673 clauses on SIMD constructs.
10675 2021-09-28 Aldy Hernandez <aldyh@redhat.com>
10677 PR tree-optimization/102511
10678 * gimple-range-path.cc (path_range_query::range_on_path_entry):
10679 Return VARYING when nothing found.
10681 2021-09-28 Hongyu Wang <hongyu.wang@intel.com>
10684 * config/i386/i386.h (VALID_AVX512FP16_REG_MODE): Add
10686 (VALID_SSE2_REG_VHF_MODE): Add V4HFmode and V2HFmode.
10687 (VALID_MMX_REG_MODE): Add V4HFmode.
10688 (SSE_REG_MODE_P): Replace VALID_AVX512FP16_REG_MODE with
10689 vector mode condition.
10690 * config/i386/i386.c (classify_argument): Parse V4HF/V2HF
10692 (function_arg_32): Add V4HFmode.
10693 (function_arg_advance_32): Likewise.
10694 * config/i386/i386.md (mode): Add V4HF/V2HF.
10695 (MODE_SIZE): Likewise.
10696 * config/i386/mmx.md (MMXMODE): Add V4HF mode.
10697 (V_32): Add V2HF mode.
10698 (VHF_32_64): New mode iterator.
10699 (*mov<mode>_internal): Adjust sse alternatives to support
10701 (*mov<mode>_internal): Adjust sse alternatives to support
10703 (<insn><mode>3): New define_insn for add/sub/mul/div.
10705 2021-09-28 Aldy Hernandez <aldyh@redhat.com>
10707 * tree-ssa-threadbackward.c (pass_thread_jumps::gate): Check
10709 (pass_early_thread_jumps::gate): Same.
10710 * tree-ssa-threadedge.c (jump_threader::thread_outgoing_edges):
10711 Return if !flag_thread_jumps.
10712 * tree-ssa-threadupdate.c
10713 (jt_path_registry::register_jump_thread): Assert that
10714 flag_thread_jumps is true.
10716 2021-09-28 liuhongt <hongtao.liu@intel.com>
10719 (simplify_context::simplify_binary_operation_1): Relax
10720 condition of simplifying (vec_concat:M (vec_select op0
10721 index0)(vec_select op1 index1)) to allow different modes
10722 between op0 and M, but have same inner mode.
10724 2021-09-28 liuhongt <hongtao.liu@intel.com>
10726 * config/i386/i386-expand.c (emit_reduc_half): Handle
10727 V8HF/V16HF/V32HFmode.
10728 * config/i386/sse.md (REDUC_SSE_PLUS_MODE): Add V8HF.
10729 (REDUC_SSE_SMINMAX_MODE): Ditto.
10730 (REDUC_PLUS_MODE): Add V16HF and V32HF.
10731 (REDUC_SMINMAX_MODE): Ditto.
10733 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
10735 * gimple-range-path.cc
10736 (path_range_query::precompute_ranges_in_block): Rename to...
10737 (path_range_query::compute_ranges_in_block): ...this.
10738 (path_range_query::precompute_ranges): Rename to...
10739 (path_range_query::compute_ranges): ...this.
10740 (path_range_query::precompute_relations): Rename to...
10741 (path_range_query::compute_relations): ...this.
10742 (path_range_query::precompute_phi_relations): Rename to...
10743 (path_range_query::compute_phi_relations): ...this.
10744 * gimple-range-path.h: Rename precompute* to compute*.
10745 * tree-ssa-threadbackward.c
10746 (back_threader::find_taken_edge_switch): Same.
10747 (back_threader::find_taken_edge_cond): Same.
10748 * tree-ssa-threadedge.c
10749 (hybrid_jt_simplifier::compute_ranges_from_state): Same.
10750 (hybrid_jt_state::register_equivs_stmt): Inline...
10751 * tree-ssa-threadedge.h: ...here.
10753 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
10755 * tree-vrp.c (lhs_of_dominating_assert): Remove.
10756 (class vrp_jt_state): Remove.
10757 (class vrp_jt_simplifier): Remove.
10758 (vrp_jt_simplifier::simplify): Remove.
10759 (class vrp_jump_threader): Remove.
10760 (vrp_jump_threader::vrp_jump_threader): Remove.
10761 (vrp_jump_threader::~vrp_jump_threader): Remove.
10762 (vrp_jump_threader::before_dom_children): Remove.
10763 (vrp_jump_threader::after_dom_children): Remove.
10765 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
10767 * passes.def (pass_vrp_threader): New.
10768 * tree-pass.h (make_pass_vrp_threader): Add make_pass_vrp_threader.
10769 * tree-ssa-threadedge.c (hybrid_jt_state::register_equivs_stmt): New.
10770 (hybrid_jt_simplifier::hybrid_jt_simplifier): New.
10771 (hybrid_jt_simplifier::simplify): New.
10772 (hybrid_jt_simplifier::compute_ranges_from_state): New.
10773 * tree-ssa-threadedge.h (class hybrid_jt_state): New.
10774 (class hybrid_jt_simplifier): New.
10775 * tree-vrp.c (execute_vrp): Remove ASSERT_EXPR based jump
10777 (class hybrid_threader): New.
10778 (hybrid_threader::hybrid_threader): New.
10779 (hybrid_threader::~hybrid_threader): New.
10780 (hybrid_threader::before_dom_children): New.
10781 (hybrid_threader::after_dom_children): New.
10782 (execute_vrp_threader): New.
10783 (class pass_vrp_threader): New.
10784 (make_pass_vrp_threader): New.
10786 2021-09-27 Martin Liska <mliska@suse.cz>
10788 * output.h (enum section_flag): New.
10789 (SECTION_FORGET): Remove.
10790 (SECTION_ENTSIZE): Make it (1UL << 8) - 1.
10791 (SECTION_STYLE_MASK): Define it based on other enum
10793 * varasm.c (switch_to_section): Remove unused handling of
10796 2021-09-27 Martin Liska <mliska@suse.cz>
10798 * common.opt: Add new variable flag_default_complex_method.
10799 * opts.c (finish_options): Handle flags related to
10800 x_flag_complex_method.
10801 * toplev.c (process_options): Remove option handling related
10802 to flag_complex_method.
10804 2021-09-27 Richard Biener <rguenther@suse.de>
10806 PR middle-end/102450
10807 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid using
10808 type_for_size, instead use int_mode_for_size.
10810 2021-09-27 Andrew Pinski <apinski@marvell.com>
10813 * gimplify.c (gimplify_save_expr): Return early
10814 if the type of val is error_mark_node.
10816 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
10818 * tree-ssanames.c (ssa_name_has_boolean_range): Use
10821 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
10823 * gimple-ssa-evrp-analyze.h (class evrp_range_analyzer): Remove
10824 vrp_visit_cond_stmt.
10825 * tree-ssa-dom.c (cprop_operand): Convert to range_query API.
10826 (cprop_into_stmt): Same.
10827 (dom_opt_dom_walker::optimize_stmt): Same.
10829 2021-09-27 Richard Biener <rguenther@suse.de>
10831 PR tree-optimization/97351
10832 PR tree-optimization/97352
10833 PR tree-optimization/82426
10834 * tree-vectorizer.h (dr_misalignment): Add vector type
10836 (aligned_access_p): Likewise.
10837 (known_alignment_for_access_p): Likewise.
10838 (vect_supportable_dr_alignment): Likewise.
10839 (vect_known_alignment_in_bytes): Likewise. Refactor.
10840 (DR_MISALIGNMENT): Remove.
10841 (vect_update_shared_vectype): Likewise.
10842 * tree-vect-data-refs.c (dr_misalignment): Refactor, handle
10843 a vector type with larger alignment requirement and apply
10844 the negative step adjustment here.
10845 (vect_calculate_target_alignment): Remove.
10846 (vect_compute_data_ref_alignment): Get explicit vector type
10847 argument, do not apply a negative step alignment adjustment
10849 (vect_slp_analyze_node_alignment): Re-analyze alignment
10850 when we re-visit the DR with a bigger desired alignment but
10851 keep more precise results from smaller alignments.
10852 * tree-vect-slp.c (vect_update_shared_vectype): Remove.
10853 (vect_slp_analyze_node_operations_1): Do not update the
10854 shared vector type on stmts.
10855 * tree-vect-stmts.c (vect_analyze_stmt): Push/pop the
10856 vector type of an SLP node to the representative stmt-info.
10857 (vect_transform_stmt): Likewise.
10859 2021-09-27 liuhongt <hongtao.liu@intel.com>
10862 2021-09-09 liuhongt <hongtao.liu@intel.com>
10865 * config/i386/sse.md (reduc_plus_scal_<mode>): Split to ..
10866 (reduc_plus_scal_v4sf): .. this, New define_expand.
10867 (reduc_plus_scal_v2df): .. and this, New define_expand.
10869 2021-09-26 liuhongt <hongtao.liu@intel.com>
10871 * doc/extend.texi (Half-Precision): Remove storage only
10872 description for _Float16 w/o avx512fp16.
10874 2021-09-25 Dimitar Dimitrov <dimitar@dinux.eu>
10876 * config/pru/constraints.md (Rrio): New constraint.
10877 * config/pru/predicates.md (regio_operand): New predicate.
10878 * config/pru/pru-pragma.c (pru_register_pragmas): Register
10879 the __regio_symbol address space.
10880 * config/pru/pru-protos.h (pru_symref2ioregno): Declaration.
10881 * config/pru/pru.c (pru_symref2ioregno): New helper function.
10882 (pru_legitimate_address_p): Remove.
10883 (pru_addr_space_legitimate_address_p): Use the address space
10884 aware hook variant.
10885 (pru_nongeneric_pointer_addrspace): New helper function.
10886 (pru_insert_attributes): New function to validate __regio_symbol
10888 (TARGET_INSERT_ATTRIBUTES): New macro.
10889 (TARGET_LEGITIMATE_ADDRESS_P): Remove.
10890 (TARGET_ADDR_SPACE_LEGITIMATE_ADDRESS_P): New macro.
10891 * config/pru/pru.h (enum reg_class): Add REGIO_REGS class.
10892 * config/pru/pru.md (*regio_readsi): New pattern to read I/O
10894 (*regio_nozext_writesi): New pattern to write to I/O registers.
10895 (*regio_zext_write_r30<EQS0:mode>): Ditto.
10896 * doc/extend.texi: Document the new PRU Named Address Space.
10898 2021-09-24 Patrick Palka <ppalka@redhat.com>
10902 * real.c (encode_ieee_double): Avoid unwanted sign extension.
10903 (encode_ieee_quad): Likewise.
10905 2021-09-24 Vladimir Makarov <vmakarov@redhat.com>
10907 PR rtl-optimization/102147
10908 * ira-build.c (ira_conflict_vector_profitable_p): Make
10909 profitability calculation independent of host compiler pointer and
10910 IRA_INT_BITS sizes.
10912 2021-09-24 Aldy Hernandez <aldyh@redhat.com>
10914 * gimple-range-path.cc (path_range_query::path_range_query):
10915 Move debugging header...
10916 (path_range_query::precompute_ranges): ...here.
10917 (path_range_query::internal_range_of_expr): Do not call
10918 range_on_path_entry if NAME is defined in the current block.
10920 2021-09-24 Richard Biener <rguenther@suse.de>
10922 * cfghooks.c (verify_flow_info): Verify unallocated BB and
10923 edge flags are not set.
10925 2021-09-24 Aldy Hernandez <aldyh@redhat.com>
10927 * tree-ssa-threadupdate.c (jt_path_registry::cancel_invalid_paths):
10929 (jt_path_registry::register_jump_thread): Call
10930 cancel_invalid_paths.
10931 * tree-ssa-threadupdate.h (class jt_path_registry): Add
10932 cancel_invalid_paths.
10934 2021-09-24 Feng Xue <fxue@os.amperecomputing.com>
10936 PR tree-optimization/102400
10937 * tree-ssa-sccvn.c (vn_reference_insert_pieces): Initialize
10938 result_vdef to zero value.
10940 2021-09-24 Feng Xue <fxue@os.amperecomputing.com>
10942 PR tree-optimization/102451
10943 * tree-ssa-dse.c (delete_dead_or_redundant_call): Record bb of stmt
10946 2021-09-24 Hongyu Wang <hongyu.wang@intel.com>
10948 * config/i386/sse.md (cond_<insn><mode>): Extend to support
10950 (cond_mul<mode>): Likewise.
10951 (cond_div<mode>): Likewise.
10952 (cond_<code><mode>): Likewise.
10953 (cond_fma<mode>): Likewise.
10954 (cond_fms<mode>): Likewise.
10955 (cond_fnma<mode>): Likewise.
10956 (cond_fnms<mode>): Likewise.
10958 2021-09-23 Andrew MacLeod <amacleod@redhat.com>
10960 PR tree-optimization/102463
10961 * gimple-range-fold.cc (fold_using_range::relation_fold_and_or): If
10962 there is no range-ops handler, don't look for a relation.
10964 2021-09-23 Andrew MacLeod <amacleod@redhat.com>
10966 * gimple-range-cache.cc (ranger_cache::ranger_cache): Take
10967 non-executable_edge flag as parameter.
10968 * gimple-range-cache.h (ranger_cache): Adjust prototype.
10969 * gimple-range-gori.cc (gori_compute::gori_compute): Take
10970 non-executable_edge flag as parameter.
10971 (gori_compute::outgoing_edge_range_p): Check new flag.
10972 * gimple-range-gori.h (gori_compute): Adjust prototype.
10973 * gimple-range.cc (gimple_ranger::gimple_ranger): Create new flag.
10974 (gimple_ranger::range_on_edge): Check new flag.
10975 * gimple-range.h (gimple_ranger::non_executable_edge_flag): New.
10976 * gimple-ssa-evrp.c (rvrp_folder): Pass ranger flag to simplifer.
10977 (hybrid_folder::hybrid_folder): Set ranger non-executable flag value.
10978 (hybrid_folder::fold_stmt): Set flag value in the simplifer.
10979 * vr-values.c (simplify_using_ranges::set_and_propagate_unexecutable):
10980 Use not_executable flag if provided inmstead of EDGE_EXECUTABLE.
10981 (simplify_using_ranges::simplify_switch_using_ranges): Clear
10982 EDGE_EXECUTABLE like it originally did.
10983 (simplify_using_ranges::cleanup_edges_and_switches): Clear any
10984 NON_EXECUTABLE flags.
10985 (simplify_using_ranges::simplify_using_ranges): Adjust.
10986 * vr-values.h (class simplify_using_ranges): Adjust.
10987 (simplify_using_ranges::set_range_query): Add non-executable flag param.
10989 2021-09-23 Bill Schmidt <wschmidt@linux.ibm.com>
10992 * config/rs6000/rs6000-call.c (rs6000_aggregate_candidate): Detect
10993 zero-width bit fields and return indicator.
10994 (rs6000_discover_homogeneous_aggregate): Diagnose when the
10995 presence of a zero-width bit field changes parameter passing in
10998 2021-09-23 Aldy Hernandez <aldyh@redhat.com>
11000 * gimple-range-fold.cc (fold_using_range::range_of_phi):
11001 Remove dominator check.
11003 2021-09-23 Aldy Hernandez <aldyh@redhat.com>
11005 * gimple-range-path.cc (path_range_query::precompute_relations):
11006 Hoist edge calculations before using EDGE_SUCC.
11008 2021-09-23 Jonathan Wakely <jwakely@redhat.com>
11010 * configure.ac: Fix --with-multilib-list description.
11011 * configure: Regenerate.
11013 2021-09-23 Richard Biener <rguenther@suse.de>
11015 PR tree-optimization/102448
11016 * tree-vect-data-refs.c (vect_duplicate_ssa_name_ptr_info):
11017 Clear alignment info copied from DR_PTR_INFO.
11019 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
11021 * config/i386/i386-expand.c (ix86_use_mask_cmp_p): Enable
11023 * config/i386/sse.md (sseintvecmodelower): Add HF vector modes.
11024 (<avx512>_store<mode>_mask): Extend to support HF vector modes.
11025 (vec_cmp<mode><avx512fmaskmodelower>): Likewise.
11026 (vcond_mask_<mode><avx512fmaskmodelower>): Likewise.
11027 (vcond<mode><mode>): New expander.
11028 (vcond<mode><sseintvecmodelower>): Likewise.
11029 (vcond<sseintvecmodelower><mode>): Likewise.
11030 (vcondu<mode><sseintvecmodelower>): Likewise.
11032 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
11034 * config/i386/sse.md (extend<ssePHmodelower><mode>2):
11036 (extendv4hf<mode>2): Likewise.
11037 (extendv2hfv2df2): Likewise.
11038 (trunc<mode><ssePHmodelower>2): Likewise.
11039 (avx512fp16_vcvt<castmode>2ph_<mode>): Rename to ...
11040 (trunc<mode>v4hf2): ... this, and drop constraints.
11041 (avx512fp16_vcvtpd2ph_v2df): Rename to ...
11042 (truncv2dfv2hf2): ... this, and likewise.
11044 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
11046 * config/i386/sse.md (float<floatunssuffix><mode><ssePHmodelower>2):
11048 (avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>):
11050 (float<floatunssuffix><mode>v4hf2): ... this, and drop constraints.
11051 (avx512fp16_vcvt<floatsuffix>qq2ph_v2di): Rename to ...
11052 (float<floatunssuffix>v2div2hf2): ... this, and likewise.
11054 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
11056 * config/i386/i386.md (fix<fixunssuffix>_trunchf<mode>2): New expander.
11057 (fixuns_trunchfhi2): Likewise.
11058 (*fixuns_trunchfsi2zext): New define_insn.
11059 * config/i386/sse.md (ssePHmodelower): New mode_attr.
11060 (fix<fixunssuffix>_trunc<ssePHmodelower><mode>2):
11061 New expander for same element vector fix_truncate.
11062 (fix<fixunssuffix>_trunc<ssePHmodelower><mode>2):
11063 Likewise for V4HF to V4SI/V4DI fix_truncate.
11064 (fix<fixunssuffix>_truncv2hfv2di2):
11065 Likeise for V2HF to V2DI fix_truncate.
11067 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
11069 * config/i386/i386.md (<code>hf3): New expander.
11071 2021-09-23 liuhongt <hongtao.liu@intel.com>
11073 * config/i386/sse.md (FMAMODEM): extend to handle FP16.
11074 (VFH_SF_AVX512VL): Extend to handle HFmode.
11075 (VF_SF_AVX512VL): Deleted.
11077 2021-09-23 liuhongt <hongtao.liu@intel.com>
11079 * config/i386/i386.md (rinthf2): New expander.
11080 (nearbyinthf2): New expander.
11082 2021-09-23 Aldy Hernandez <aldyh@redhat.com>
11084 * tree-ssa-dom.c (class dom_jump_threader_simplifier): Rename...
11085 (class dom_jt_state): ...this and provide virtual overrides.
11086 (dom_jt_state::register_equiv): New.
11087 (class dom_jt_simplifier): Rename from
11088 dom_jump_threader_simplifier.
11089 (dom_jump_threader_simplifier::simplify): Rename...
11090 (dom_jt_simplifier::simplify): ...to this.
11091 (pass_dominator::execute): Use dom_jt_simplifier and
11093 * tree-ssa-threadedge.c (jump_threader::jump_threader):
11095 (jt_state::register_equivs_stmt): Abstract out...
11096 (jump_threader::record_temporary_equivalences_from_stmts_at_dest):
11098 (jump_threader::thread_around_empty_blocks): Update state.
11099 (jump_threader::thread_through_normal_block): Same.
11100 (jt_state::jt_state): Remove.
11101 (jt_state::push): Remove pass specific bits. Keep block vector
11103 (jt_state::append_path): New.
11104 (jt_state::pop): Remove pass specific bits.
11105 (jt_state::register_equiv): Same.
11106 (jt_state::record_ranges_from_stmt): Same.
11107 (jt_state::register_equivs_on_edge): Same. Rename...
11108 (jt_state::register_equivs_edge): ...to this.
11109 (jt_state::dump): New.
11110 (jt_state::debug): New.
11111 (jump_threader_simplifier::simplify): Remove.
11112 (jt_state::get_path): New.
11113 * tree-ssa-threadedge.h (class jt_simplifier): Make into a base
11114 class. Expose common functionality as virtual methods.
11115 (class jump_threader_simplifier): Same. Rename...
11116 (class jt_simplifier): ...to this.
11117 * tree-vrp.c (class vrp_jump_threader_simplifier): Rename...
11118 (class vrp_jt_simplifier): ...to this. Provide pass specific
11120 (class vrp_jt_state): New.
11121 (vrp_jump_threader_simplifier::simplify): Rename...
11122 (vrp_jt_simplifier::simplify): ...to this. Inline code from
11123 what used to be the base class.
11124 (vrp_jump_threader::vrp_jump_threader): Use vrp_jt_state and
11127 2021-09-22 Tobias Burnus <tobias@codesourcery.com>
11130 * doc/invoke.texi (-Wno-missing-include-dirs.): Document Fortran
11133 2021-09-22 Roger Sayle <roger@nextmovesoftware.com>
11134 Richard Biener <rguenther@suse.de>
11136 * match.pd (negation simplifications): Implement some negation
11137 folding transformations from fold-const.c's fold_negate_expr.
11138 * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Add a SIMPLIFY
11139 argument, to control whether the op should be simplified prior
11140 to looking up/assigning a value number.
11141 (vn_nary_build_or_lookup): Update call to vn_nary_build_or_lookup_1.
11142 (vn_nary_simplify): Likewise.
11143 (visit_nary_op): Likewise, but when constructing a NEGATE_EXPR
11144 now call vn_nary_build_or_lookup_1 disabling simplification.
11146 2021-09-22 Jiufu Guo <guojiufu@linux.ibm.com>
11148 PR tree-optimization/102087
11149 * tree-ssa-loop-niter.c (number_of_iterations_until_wrap):
11150 Update bound/cmp/control for niter.
11152 2021-09-22 Aldy Hernandez <aldyh@redhat.com>
11154 * gimple-range-fold.cc (fold_using_range::range_of_range_op):
11155 Move check for non-empty BB here.
11156 (fur_source::register_outgoing_edges): ...from here.
11158 2021-09-22 Aldy Hernandez <aldyh@redhat.com>
11160 * gimple-range-path.cc (path_range_query::internal_range_of_expr):
11161 Remove call to improve_range_with_equivs.
11162 (path_range_query::improve_range_with_equivs): Remove
11163 * gimple-range-path.h: Remove improve_range_with_equivs.
11165 2021-09-22 dianhong xu <dianhong.xu@intel.com>
11167 * config/i386/avx512fp16intrin.h:
11168 (_mm512_mask_blend_ph): New intrinsic.
11169 (_mm512_permutex2var_ph): Ditto.
11170 (_mm512_permutexvar_ph): Ditto.
11171 * config/i386/avx512fp16vlintrin.h:
11172 (_mm256_mask_blend_ph): New intrinsic.
11173 (_mm256_permutex2var_ph): Ditto.
11174 (_mm256_permutexvar_ph): Ditto.
11175 (_mm_mask_blend_ph): Ditto.
11176 (_mm_permutex2var_ph): Ditto.
11177 (_mm_permutexvar_ph): Ditto.
11179 2021-09-22 dianhong xu <dianhong.xu@intel.com>
11181 * config/i386/avx512fp16intrin.h: Add new intrinsics.
11182 (_mm512_conj_pch): New intrinsic.
11183 (_mm512_mask_conj_pch): Ditto.
11184 (_mm512_maskz_conj_pch): Ditto.
11185 * config/i386/avx512fp16vlintrin.h: Add new intrinsics.
11186 (_mm256_conj_pch): New intrinsic.
11187 (_mm256_mask_conj_pch): Ditto.
11188 (_mm256_maskz_conj_pch): Ditto.
11189 (_mm_conj_pch): Ditto.
11190 (_mm_mask_conj_pch): Ditto.
11191 (_mm_maskz_conj_pch): Ditto.
11193 2021-09-22 dianhong xu <dianhong.xu@intel.com>
11195 * config/i386/avx512fp16intrin.h (_MM512_REDUCE_OP): New macro
11196 (_mm512_reduce_add_ph): New intrinsic.
11197 (_mm512_reduce_mul_ph): Ditto.
11198 (_mm512_reduce_min_ph): Ditto.
11199 (_mm512_reduce_max_ph): Ditto.
11200 * config/i386/avx512fp16vlintrin.h
11201 (_MM256_REDUCE_OP/_MM_REDUCE_OP): New macro.
11202 (_mm256_reduce_add_ph): New intrinsic.
11203 (_mm256_reduce_mul_ph): Ditto.
11204 (_mm256_reduce_min_ph): Ditto.
11205 (_mm256_reduce_max_ph): Ditto.
11206 (_mm_reduce_add_ph): Ditto.
11207 (_mm_reduce_mul_ph): Ditto.
11208 (_mm_reduce_min_ph): Ditto.
11209 (_mm_reduce_max_ph): Ditto.
11211 2021-09-22 dianhong xu <dianhong.xu@intel.com>
11213 * config/i386/avx512fp16intrin.h (__m512h_u, __m256h_u,
11214 __m128h_u): New typedef.
11215 (_mm512_load_ph): New intrinsic.
11216 (_mm256_load_ph): Ditto.
11217 (_mm_load_ph): Ditto.
11218 (_mm512_loadu_ph): Ditto.
11219 (_mm256_loadu_ph): Ditto.
11220 (_mm_loadu_ph): Ditto.
11221 (_mm512_store_ph): Ditto.
11222 (_mm256_store_ph): Ditto.
11223 (_mm_store_ph): Ditto.
11224 (_mm512_storeu_ph): Ditto.
11225 (_mm256_storeu_ph): Ditto.
11226 (_mm_storeu_ph): Ditto.
11227 (_mm512_abs_ph): Ditto.
11228 * config/i386/avx512fp16vlintrin.h
11229 (_mm_abs_ph): Ditto.
11230 (_mm256_abs_ph): Ditto.
11232 2021-09-22 Andreas Krebbel <krebbel@linux.ibm.com>
11234 * config/s390/tpf.md (prologue_tpf, epilogue_tpf): Add cc clobber.
11236 2021-09-22 Andreas Krebbel <krebbel@linux.ibm.com>
11239 * config/s390/s390.c (s390_expand_insv): Emit a normal move if it
11240 is actually a full copy of the source operand into the target.
11241 Don't emit a strict low part move if source and target mode match.
11243 2021-09-22 Jakub Jelinek <jakub@redhat.com>
11245 PR middle-end/102415
11246 * omp-expand.c (expand_omp_single): If region->exit is NULL,
11247 assert region->entry is GIMPLE_OMP_SCOPE region and return.
11249 2021-09-22 Jakub Jelinek <jakub@redhat.com>
11251 * tree.h (OMP_CLAUSE_ALLOCATE_ALIGN): Define.
11252 * tree.c (omp_clause_num_ops): Change number of OMP_CLAUSE_ALLOCATE
11253 arguments from 2 to 3.
11254 * tree-pretty-print.c (dump_omp_clause): Print allocator() around
11255 allocate clause allocator and print align if present.
11256 * omp-low.c (scan_sharing_clauses): Force allocate_map entry even
11257 for omp_default_mem_alloc if align modifier is present. If align
11258 modifier is present, use TREE_LIST to encode both allocator and
11260 (lower_private_allocate, lower_rec_input_clauses, create_task_copyfn):
11261 Handle align modifier on allocator clause if present.
11263 2021-09-22 liuhongt <hongtao.liu@intel.com>
11265 * config/i386/i386.md (define_attr "isa"): Add
11267 (define_attr "enabled"): Correspond fma_or_avx512vl to
11268 TARGET_FMA || TARGET_AVX512VL.
11269 * config/i386/mmx.md (fmav2sf4): Extend to AVX512 fma.
11271 (fnmav2sf4): Ditto.
11272 (fnmsv2sf4): Ditto.
11274 2021-09-22 liuhongt <hongtao.liu@intel.com>
11276 * config/i386/i386.md (cstorehf3): New define_expand.
11278 2021-09-22 liuhongt <hongtao.liu@intel.com>
11280 * config/i386/i386.md (<rounding_insn>hf2): New expander.
11281 (sse4_1_round<mode>2): Extend from MODEF to MODEFH.
11282 * config/i386/sse.md (*sse4_1_round<ssescalarmodesuffix>):
11283 Extend from VF_128 to VFH_128.
11285 2021-09-22 liuhongt <hongtao.liu@intel.com>
11287 * config/i386/i386-features.c (i386-features.c): Handle
11289 * config/i386/i386.md (sqrthf2): New expander.
11290 (*sqrthf2): New define_insn.
11291 * config/i386/sse.md
11292 (*<sse>_vmsqrt<mode>2<mask_scalar_name><round_scalar_name>):
11295 2021-09-22 liuhongt <hongtao.liu@intel.com>
11297 * config/i386/avx512fp16intrin.h (_mm_mask_fcmadd_sch):
11299 (_mm_mask3_fcmadd_sch): Likewise.
11300 (_mm_maskz_fcmadd_sch): Likewise.
11301 (_mm_fcmadd_sch): Likewise.
11302 (_mm_mask_fmadd_sch): Likewise.
11303 (_mm_mask3_fmadd_sch): Likewise.
11304 (_mm_maskz_fmadd_sch): Likewise.
11305 (_mm_fmadd_sch): Likewise.
11306 (_mm_mask_fcmadd_round_sch): Likewise.
11307 (_mm_mask3_fcmadd_round_sch): Likewise.
11308 (_mm_maskz_fcmadd_round_sch): Likewise.
11309 (_mm_fcmadd_round_sch): Likewise.
11310 (_mm_mask_fmadd_round_sch): Likewise.
11311 (_mm_mask3_fmadd_round_sch): Likewise.
11312 (_mm_maskz_fmadd_round_sch): Likewise.
11313 (_mm_fmadd_round_sch): Likewise.
11314 (_mm_fcmul_sch): Likewise.
11315 (_mm_mask_fcmul_sch): Likewise.
11316 (_mm_maskz_fcmul_sch): Likewise.
11317 (_mm_fmul_sch): Likewise.
11318 (_mm_mask_fmul_sch): Likewise.
11319 (_mm_maskz_fmul_sch): Likewise.
11320 (_mm_fcmul_round_sch): Likewise.
11321 (_mm_mask_fcmul_round_sch): Likewise.
11322 (_mm_maskz_fcmul_round_sch): Likewise.
11323 (_mm_fmul_round_sch): Likewise.
11324 (_mm_mask_fmul_round_sch): Likewise.
11325 (_mm_maskz_fmul_round_sch): Likewise.
11326 * config/i386/i386-builtin.def: Add corresponding new builtins.
11327 * config/i386/sse.md
11328 (avx512fp16_fmaddcsh_v8hf_maskz<round_expand_name>): New expander.
11329 (avx512fp16_fcmaddcsh_v8hf_maskz<round_expand_name>): Ditto.
11330 (avx512fp16_fma_<complexopname>sh_v8hf<mask_scalarcz_name><round_scalarcz_name>):
11332 (avx512fp16_<complexopname>sh_v8hf_mask<round_name>): Ditto.
11333 (avx512fp16_<complexopname>sh_v8hf<mask_scalarc_name><round_scalarcz_name>):
11335 * config/i386/subst.md (mask_scalarcz_name): New.
11336 (mask_scalarc_name): Ditto.
11337 (mask_scalarc_operand3): Ditto.
11338 (mask_scalarcz_operand4): Ditto.
11339 (round_scalarcz_name): Ditto.
11340 (round_scalarc_mask_operand3): Ditto.
11341 (round_scalarcz_mask_operand4): Ditto.
11342 (round_scalarc_mask_op3): Ditto.
11343 (round_scalarcz_mask_op4): Ditto.
11344 (round_scalarcz_constraint): Ditto.
11345 (round_scalarcz_nimm_predicate): Ditto.
11346 (mask_scalarcz): Ditto.
11347 (mask_scalarc): Ditto.
11348 (round_scalarcz): Ditto.
11350 2021-09-22 liuhongt <hongtao.liu@intel.com>
11352 * config/i386/avx512fp16intrin.h (_mm512_fcmadd_pch):
11354 (_mm512_mask_fcmadd_pch): Likewise.
11355 (_mm512_mask3_fcmadd_pch): Likewise.
11356 (_mm512_maskz_fcmadd_pch): Likewise.
11357 (_mm512_fmadd_pch): Likewise.
11358 (_mm512_mask_fmadd_pch): Likewise.
11359 (_mm512_mask3_fmadd_pch): Likewise.
11360 (_mm512_maskz_fmadd_pch): Likewise.
11361 (_mm512_fcmadd_round_pch): Likewise.
11362 (_mm512_mask_fcmadd_round_pch): Likewise.
11363 (_mm512_mask3_fcmadd_round_pch): Likewise.
11364 (_mm512_maskz_fcmadd_round_pch): Likewise.
11365 (_mm512_fmadd_round_pch): Likewise.
11366 (_mm512_mask_fmadd_round_pch): Likewise.
11367 (_mm512_mask3_fmadd_round_pch): Likewise.
11368 (_mm512_maskz_fmadd_round_pch): Likewise.
11369 (_mm512_fcmul_pch): Likewise.
11370 (_mm512_mask_fcmul_pch): Likewise.
11371 (_mm512_maskz_fcmul_pch): Likewise.
11372 (_mm512_fmul_pch): Likewise.
11373 (_mm512_mask_fmul_pch): Likewise.
11374 (_mm512_maskz_fmul_pch): Likewise.
11375 (_mm512_fcmul_round_pch): Likewise.
11376 (_mm512_mask_fcmul_round_pch): Likewise.
11377 (_mm512_maskz_fcmul_round_pch): Likewise.
11378 (_mm512_fmul_round_pch): Likewise.
11379 (_mm512_mask_fmul_round_pch): Likewise.
11380 (_mm512_maskz_fmul_round_pch): Likewise.
11381 * config/i386/avx512fp16vlintrin.h (_mm_fmadd_pch):
11383 (_mm_mask_fmadd_pch): Likewise.
11384 (_mm_mask3_fmadd_pch): Likewise.
11385 (_mm_maskz_fmadd_pch): Likewise.
11386 (_mm256_fmadd_pch): Likewise.
11387 (_mm256_mask_fmadd_pch): Likewise.
11388 (_mm256_mask3_fmadd_pch): Likewise.
11389 (_mm256_maskz_fmadd_pch): Likewise.
11390 (_mm_fcmadd_pch): Likewise.
11391 (_mm_mask_fcmadd_pch): Likewise.
11392 (_mm_mask3_fcmadd_pch): Likewise.
11393 (_mm_maskz_fcmadd_pch): Likewise.
11394 (_mm256_fcmadd_pch): Likewise.
11395 (_mm256_mask_fcmadd_pch): Likewise.
11396 (_mm256_mask3_fcmadd_pch): Likewise.
11397 (_mm256_maskz_fcmadd_pch): Likewise.
11398 (_mm_fmul_pch): Likewise.
11399 (_mm_mask_fmul_pch): Likewise.
11400 (_mm_maskz_fmul_pch): Likewise.
11401 (_mm256_fmul_pch): Likewise.
11402 (_mm256_mask_fmul_pch): Likewise.
11403 (_mm256_maskz_fmul_pch): Likewise.
11404 (_mm_fcmul_pch): Likewise.
11405 (_mm_mask_fcmul_pch): Likewise.
11406 (_mm_maskz_fcmul_pch): Likewise.
11407 (_mm256_fcmul_pch): Likewise.
11408 (_mm256_mask_fcmul_pch): Likewise.
11409 (_mm256_maskz_fcmul_pch): Likewise.
11410 * config/i386/i386-builtin-types.def (V8HF_FTYPE_V8HF_V8HF_V8HF,
11411 V8HF_FTYPE_V16HF_V16HF_V16HF, V16HF_FTYPE_V16HF_V16HF_V16HF_UQI,
11412 V32HF_FTYPE_V32HF_V32HF_V32HF_INT,
11413 V32HF_FTYPE_V32HF_V32HF_V32HF_UHI_INT): Add new builtin types.
11414 * config/i386/i386-builtin.def: Add new builtins.
11415 * config/i386/i386-expand.c: Handle new builtin types.
11416 * config/i386/subst.md (SUBST_CV): New.
11417 (maskc_name): Ditto.
11418 (maskc_operand3): Ditto.
11420 (sdc_maskz_name): Ditto.
11421 (sdc_mask_op4): Ditto.
11422 (sdc_mask_op5): Ditto.
11423 (sdc_mask_mode512bit_condition): Ditto.
11425 (round_maskc_operand3): Ditto.
11426 (round_sdc_mask_operand4): Ditto.
11427 (round_maskc_op3): Ditto.
11428 (round_sdc_mask_op4): Ditto.
11429 (round_saeonly_sdc_mask_operand5): Ditto.
11430 * config/i386/sse.md (unspec): Add complex fma unspecs.
11431 (avx512fmaskcmode): New.
11432 (UNSPEC_COMPLEX_F_C_MA): Ditto.
11433 (UNSPEC_COMPLEX_F_C_MUL): Ditto.
11434 (complexopname): Ditto.
11435 (<avx512>_fmaddc_<mode>_maskz<round_expand_name>): New expander.
11436 (<avx512>_fcmaddc_<mode>_maskz<round_expand_name>): Ditto.
11437 (fma_<complexopname>_<mode><sdc_maskz_name><round_name>): New
11439 (<avx512>_<complexopname>_<mode>_mask<round_name>): Ditto.
11440 (<avx512>_<complexopname>_<mode><maskc_name><round_name>): Ditto.
11442 2021-09-22 Kewen Lin <linkw@linux.ibm.com>
11444 * config/rs6000/rs6000.opt (rs6000-density-pct-threshold,
11445 rs6000-density-size-threshold, rs6000-density-penalty,
11446 rs6000-density-load-pct-threshold,
11447 rs6000-density-load-num-threshold): New parameter.
11448 * config/rs6000/rs6000.c (rs6000_density_test): Adjust with
11449 corresponding parameters.
11451 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
11453 * gimple-range-path.cc (path_range_query::defined_outside_path):
11455 (path_range_query::range_on_path_entry): New.
11456 (path_range_query::internal_range_of_expr): Resolve unknowns
11458 (path_range_query::improve_range_with_equivs): New.
11459 (path_range_query::ssa_range_in_phi): Resolve unknowns with
11461 * gimple-range-path.h (class path_range_query): Add
11462 defined_outside_path, range_on_path_entry, and
11463 improve_range_with_equivs.
11465 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
11467 * gimple-range-path.cc (path_range_query::add_to_imports): New.
11468 (path_range_query::add_copies_to_imports): New.
11469 (path_range_query::precompute_ranges): Call
11470 add_copies_to_imports.
11471 * gimple-range-path.h (class path_range_query): Add prototypes
11472 for add_copies_to_imports and add_to_imports.
11474 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
11476 * gimple-range-path.cc (path_range_query::range_defined_in_block):
11477 Remove useless code.
11479 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
11481 * gimple-range-fold.h (class fur_source): Make oracle protected.
11482 * gimple-range-path.cc (path_range_query::path_range_query): Add
11483 resolve argument. Initialize oracle.
11484 (path_range_query::~path_range_query): Delete oracle.
11485 (path_range_query::range_of_stmt): Adapt to use relations.
11486 (path_range_query::precompute_ranges): Pre-compute relations.
11487 (class jt_fur_source): New
11488 (jt_fur_source::jt_fur_source): New.
11489 (jt_fur_source::register_relation): New.
11490 (jt_fur_source::query_relation): New.
11491 (path_range_query::precompute_relations): New.
11492 (path_range_query::precompute_phi_relations): New.
11493 * gimple-range-path.h (path_range_query): Add resolve argument.
11494 Add oracle, precompute_relations, precompute_phi_relations.
11495 * tree-ssa-threadbackward.c (back_threader::back_threader): Pass
11496 resolve argument to solver.
11498 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
11500 * gimple-range-fold.cc (fold_using_range::range_of_range_op):
11501 Rename postfold_gcond_edges to register_outgoing_edges and
11503 (fold_using_range::postfold_gcond_edges): Rename...
11504 (fur_source::register_outgoing_edges): ...to this.
11505 * gimple-range-fold.h (postfold_gcond_edges): Rename to
11506 register_outgoing_edges and move to fur_source.
11508 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
11510 * gimple-range-fold.cc (fold_using_range::range_of_phi): Check
11511 dom_info_available_p.
11513 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
11515 * gimple-range-cache.cc (non_null_ref::non_null_ref): Use create
11516 and quick_grow_cleared instead of safe_grow_cleared.
11518 2021-09-21 Thomas Schwinge <thomas@codesourcery.com>
11521 * omp-oacc-neuter-broadcast.cc (oacc_do_neutering): Evaluate
11522 'random ()' to '0'.
11524 2021-09-21 Richard Earnshaw <rearnsha@arm.com>
11526 * configure.ac: Detect when the assembler supports new-style
11527 architecture extensions.
11528 * common/config/arm/arm-common.c (arm_rewrite_mcpu): Return
11529 the full CPU string if the assembler can grok it.
11530 (arm_rewrite_march): Likewise but for the architecture.
11531 * config.in: Regenerate.
11532 * configure: Regenerate.
11534 2021-09-21 Richard Biener <rguenther@suse.de>
11536 PR tree-optimization/102421
11537 * tree-vect-loop.c (vect_dissolve_slp_only_groups): Copy and
11538 adjust alignment info.
11540 2021-09-21 Kewen Lin <linkw@linux.ibm.com>
11542 * ipa-fnsummary.c (ipa_fn_summary_write): Remove inconsistent
11543 bitfield stream out.
11545 2021-09-20 Andrew MacLeod <amacleod@redhat.com>
11547 * gimple-range-fold.cc (fold_using_range::range_of_phi): Ignore
11548 undefined edges, apply an equivalence if appropriate.
11549 * gimple-range-gori.cc (gori_compute::outgoing_edge_range_p): Return
11550 UNDEFINED if EDGE_EXECUTABLE is not set.
11551 * gimple-range.cc (gimple_ranger::gimple_ranger): Set all edges
11552 as EXECUTABLE upon startup.
11553 (gimple_ranger::range_on_edge): Return UNDEFINED for edges without
11554 EDGE_EXECUTABLE set.
11555 * vr-values.c (set_and_propagate_unexecutable): New.
11556 (simplify_using_ranges::fold_cond): Call set_and_propagate.
11557 (simplify_using_ranges::simplify_switch_using_ranges): Ditto.
11558 * vr-values.h: Add prototype.
11560 2021-09-20 Andrew MacLeod <amacleod@redhat.com>
11562 * value-relation.cc (equiv_oracle::register_initial_def): New.
11563 (equiv_oracle::register_relation): Call register_initial_def.
11564 (equiv_oracle::add_equiv_to_block): New. Split register_relation.
11565 (relation_oracle::register_stmt): Check def block of PHI arguments.
11566 * value-relation.h (equiv_oracle): Add new prototypes.
11568 2021-09-20 Matthias Kretz <m.kretz@gsi.de>
11570 * cppbuiltin.c (define_builtin_macros_for_compilation_flags):
11571 Define __RECIPROCAL_MATH__, __NO_SIGNED_ZEROS__,
11572 __NO_TRAPPING_MATH__, __ASSOCIATIVE_MATH__, and
11573 __ROUNDING_MATH__ according to their corresponding flags.
11574 * doc/cpp.texi: Document __RECIPROCAL_MATH__,
11575 __NO_SIGNED_ZEROS__, __NO_TRAPPING_MATH__, __ASSOCIATIVE_MATH__,
11576 and __ROUNDING_MATH__.
11578 2021-09-20 Richard Biener <rguenther@suse.de>
11580 * tree-vect-stmts.c (vectorizable_load): Use the vectype
11583 2021-09-20 Richard Biener <rguenther@suse.de>
11585 * tree-vect-data-refs.c (vect_duplicate_ssa_name_ptr_info):
11586 Do not compute alignment of the vectorized access here.
11588 2021-09-20 Richard Biener <rguenther@suse.de>
11590 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment):
11591 Store -1 for runtime alias peeling iterations.
11593 2021-09-20 Richard Biener <rguenther@suse.de>
11595 * config.gcc: Obsolete hppa[12]*-*-hpux10* and hppa[12]*-*-hpux11*.
11597 2021-09-20 Thomas Schwinge <thomas@codesourcery.com>
11599 * input.c (string_concat_db::record_string_concatenation)
11600 (string_concat_db::get_string_concatenation): Skip for
11601 'RESERVED_LOCATION_P'.
11603 2021-09-20 Richard Biener <rguenther@suse.de>
11605 PR tree-optimization/65206
11606 * tree-data-ref.h (struct data_reference): Add alt_indices,
11608 * tree-data-ref.c (free_data_ref): Release alt_indices.
11609 (dr_analyze_indices): Work on struct indices and get DR_REF as tree.
11610 (create_data_ref): Adjust.
11611 (initialize_data_dependence_relation): Split into head
11612 and tail. When the base objects fail to match up try
11613 again with pointer-based analysis of indices.
11614 * tree-vectorizer.c (vec_info_shared::check_datarefs): Do
11615 not compare the lazily computed alternate set of indices.
11617 2021-09-20 Iain Sandoe <iain@sandoe.co.uk>
11619 * gcc.c: Test for execute OK when we find the
11620 programs for assembler linker and dsymutil and those
11621 were specified at configure-time.
11623 2021-09-19 Martin Sebor <msebor@redhat.com>
11625 PR middle-end/102403
11626 * gimple-predicate-analysis.cc (predicate::init_from_control_deps):
11627 Correct a function pre/postcondition.
11629 2021-09-19 Martin Sebor <msebor@redhat.com>
11631 PR middle-end/102243
11632 * tree-ssa-strlen.c (get_range): Handle null cfun.
11634 2021-09-19 Iain Sandoe <iain@sandoe.co.uk>
11636 * config/darwin.h (LINK_COMMAND_SPEC_A): Use Darwin10
11637 unwinder shim as a convenience library.
11639 2021-09-19 Andrew Pinski <apinski@marvell.com>
11641 * doc/install.texi: Add note about
11642 binutils 2.35 is required for LTO usage.
11644 2021-09-19 Aldy Hernandez <aldyh@redhat.com>
11646 * tree-ssa-threadbackward.c
11647 (back_threader_registry::register_path): Use push_edge.
11648 * tree-ssa-threadedge.c
11649 (jump_threader::thread_around_empty_blocks): Same.
11650 (jump_threader::thread_through_normal_block): Same.
11651 (jump_threader::thread_across_edge): Same. Also, use auto_bitmap.
11653 * tree-ssa-threadupdate.c
11654 (jt_path_registry::allocate_thread_edge): Remove.
11655 (jt_path_registry::push_edge): New.
11656 (dump_jump_thread_path): Make static.
11657 * tree-ssa-threadupdate.h (allocate_thread_edge): Remove.
11660 2021-09-19 Aldy Hernandez <aldyh@redhat.com>
11662 * gimple-range-path.cc (path_range_query::path_range_query): Add
11664 (path_range_query::dump): Remove extern declaration of dump_ranger.
11665 * gimple-range-trace.cc (dump_ranger): Add DEBUG_FUNCTION marker.
11666 * gimple-range-trace.h (dump_ranger): Add prototype.
11668 2021-09-19 John Ericson <git@JohnEricson.me>
11670 * gcc.c (find_a_program): New function, factored out of...
11671 (find_a_file): Here.
11672 (execute): Use find_a_program when looking for programs rather
11675 2021-09-19 Matwey V. Kornilov <matwey.kornilov@gmail.com>
11677 * config/avr/avr-mcus.def: Add atmega324pb.
11678 * doc/avr-mmcu.texi: Corresponding changes.
11680 2021-09-19 Roger Sayle <roger@nextmovesoftware.com>
11682 PR middle-end/88173
11683 * match.pd (cmp @0 REAL_CST@1): When @0 is also REAL_CST, apply
11684 the same transformations as to @1. For comparisons against NaN,
11685 don't check HONOR_SNANS but confirm that neither operand is a
11688 2021-09-19 Benjamin Peterson <benjamin@locrian.net>
11690 * attribs.c (make_unique_name): Delete.
11691 * attribs.h (make_unique_name): Delete.
11693 2021-09-19 Andrew Pinski <apinski@marvell.com>
11695 * lra-constraints.c (check_and_process_move): Assert
11696 that dclass and sclass are greater than or equal to NO_REGS.
11698 2021-09-18 Jakub Jelinek <jakub@redhat.com>
11700 * tree.h (OMP_CLAUSE_ORDER_UNCONSTRAINED): Define.
11701 * tree-pretty-print.c (dump_omp_clause): Print unconstrained:
11702 for OMP_CLAUSE_ORDER_UNCONSTRAINED.
11704 2021-09-18 liuhongt <hongtao.liu@intel.com>
11706 * config/i386/i386-features.c (remove_partial_avx_dependency):
11707 Restrict TARGET_USE_VECTOR_FP_CONVERTS and
11708 TARGET_USE_VECTOR_CONVERTS to conversion instructions only.
11710 2021-09-18 Jakub Jelinek <jakub@redhat.com>
11712 * gimplify.c (omp_default_clause): For C/C++ default({,first}private),
11713 if file/namespace scope variable doesn't have predetermined sharing,
11714 treat it as if there was default(none).
11716 2021-09-18 liuhongt <hongtao.liu@intel.com>
11718 * config/i386/avx512fp16intrin.h (_mm_fmadd_sh):
11720 (_mm_mask_fmadd_sh): Likewise.
11721 (_mm_mask3_fmadd_sh): Likewise.
11722 (_mm_maskz_fmadd_sh): Likewise.
11723 (_mm_fmadd_round_sh): Likewise.
11724 (_mm_mask_fmadd_round_sh): Likewise.
11725 (_mm_mask3_fmadd_round_sh): Likewise.
11726 (_mm_maskz_fmadd_round_sh): Likewise.
11727 (_mm_fnmadd_sh): Likewise.
11728 (_mm_mask_fnmadd_sh): Likewise.
11729 (_mm_mask3_fnmadd_sh): Likewise.
11730 (_mm_maskz_fnmadd_sh): Likewise.
11731 (_mm_fnmadd_round_sh): Likewise.
11732 (_mm_mask_fnmadd_round_sh): Likewise.
11733 (_mm_mask3_fnmadd_round_sh): Likewise.
11734 (_mm_maskz_fnmadd_round_sh): Likewise.
11735 (_mm_fmsub_sh): Likewise.
11736 (_mm_mask_fmsub_sh): Likewise.
11737 (_mm_mask3_fmsub_sh): Likewise.
11738 (_mm_maskz_fmsub_sh): Likewise.
11739 (_mm_fmsub_round_sh): Likewise.
11740 (_mm_mask_fmsub_round_sh): Likewise.
11741 (_mm_mask3_fmsub_round_sh): Likewise.
11742 (_mm_maskz_fmsub_round_sh): Likewise.
11743 (_mm_fnmsub_sh): Likewise.
11744 (_mm_mask_fnmsub_sh): Likewise.
11745 (_mm_mask3_fnmsub_sh): Likewise.
11746 (_mm_maskz_fnmsub_sh): Likewise.
11747 (_mm_fnmsub_round_sh): Likewise.
11748 (_mm_mask_fnmsub_round_sh): Likewise.
11749 (_mm_mask3_fnmsub_round_sh): Likewise.
11750 (_mm_maskz_fnmsub_round_sh): Likewise.
11751 * config/i386/i386-builtin-types.def
11752 (V8HF_FTYPE_V8HF_V8HF_V8HF_UQI_INT): New builtin type.
11753 * config/i386/i386-builtin.def: Add new builtins.
11754 * config/i386/i386-expand.c: Handle new builtin type.
11755 * config/i386/sse.md (fmai_vmfmadd_<mode><round_name>):
11756 Ajdust to support FP16.
11757 (fmai_vmfmsub_<mode><round_name>): Ditto.
11758 (fmai_vmfnmadd_<mode><round_name>): Ditto.
11759 (fmai_vmfnmsub_<mode><round_name>): Ditto.
11760 (*fmai_fmadd_<mode>): Ditto.
11761 (*fmai_fmsub_<mode>): Ditto.
11762 (*fmai_fnmadd_<mode><round_name>): Ditto.
11763 (*fmai_fnmsub_<mode><round_name>): Ditto.
11764 (avx512f_vmfmadd_<mode>_mask<round_name>): Ditto.
11765 (avx512f_vmfmadd_<mode>_mask3<round_name>): Ditto.
11766 (avx512f_vmfmadd_<mode>_maskz<round_expand_name>): Ditto.
11767 (avx512f_vmfmadd_<mode>_maskz_1<round_name>): Ditto.
11768 (*avx512f_vmfmsub_<mode>_mask<round_name>): Ditto.
11769 (avx512f_vmfmsub_<mode>_mask3<round_name>): Ditto.
11770 (*avx512f_vmfmsub_<mode>_maskz_1<round_name>): Ditto.
11771 (*avx512f_vmfnmsub_<mode>_mask<round_name>): Ditto.
11772 (*avx512f_vmfnmsub_<mode>_mask3<round_name>): Ditto.
11773 (*avx512f_vmfnmsub_<mode>_mask<round_name>): Ditto.
11774 (*avx512f_vmfnmadd_<mode>_mask<round_name>): Renamed to ...
11775 (avx512f_vmfnmadd_<mode>_mask<round_name>) ... this, and
11776 adjust to support FP16.
11777 (avx512f_vmfnmadd_<mode>_mask3<round_name>): Ditto.
11778 (avx512f_vmfnmadd_<mode>_maskz_1<round_name>): Ditto.
11779 (avx512f_vmfnmadd_<mode>_maskz<round_expand_name>): New
11782 2021-09-18 H.J. Lu <hjl.tools@gmail.com>
11784 * config/i386/sse.md (avx512fmaskmodelower): Extend to support
11786 (maskload<mode><avx512fmaskmodelower>): Ditto.
11787 (maskstore<mode><avx512fmaskmodelower>): Ditto.
11789 2021-09-18 H.J. Lu <hjl.tools@gmail.com>
11791 * config/i386/i386-expand.c (ix86_expand_fp_absneg_operator):
11793 (ix86_expand_copysign): Ditto.
11794 (ix86_expand_xorsign): Ditto.
11795 * config/i386/i386.c (ix86_build_const_vector): Handle HF vector
11797 (ix86_build_signbit_mask): Ditto.
11798 (ix86_can_change_mode_class): Ditto.
11799 * config/i386/i386.md
11800 (SSEMODEF): Add HFmode.
11801 (ssevecmodef): Ditto.
11802 (<code>hf2): New define_expand.
11803 (*<code>hf2_1): New define_insn_and_split.
11804 (copysign<mode>): Extend to support HFmode under AVX512FP16.
11805 (xorsign<mode>): Ditto.
11806 * config/i386/sse.md (VFB): New mode iterator.
11807 (VFB_128_256): Ditto.
11809 (sseintvecmode2): Support HF vector mode.
11810 (<code><mode>2): Use new mode iterator.
11811 (*<code><mode>2): Ditto.
11812 (copysign<mode>3): Ditto.
11813 (xorsign<mode>3): Ditto.
11814 (<code><mode>3<mask_name>): Ditto.
11815 (<code><mode>3<mask_name>): Ditto.
11816 (<sse>_andnot<mode>3<mask_name>): Adjust for HF vector mode.
11817 (<sse>_andnot<mode>3<mask_name>): Ditto.
11818 (*<code><mode>3<mask_name>): Ditto.
11819 (*<code><mode>3<mask_name>): Ditto.
11821 2021-09-18 liuhongt <hongtao.liu@intel.com>
11823 * config/i386/avx512fp16intrin.h (_mm512_mask_fmadd_ph):
11825 (_mm512_mask3_fmadd_ph): Likewise.
11826 (_mm512_maskz_fmadd_ph): Likewise.
11827 (_mm512_fmadd_round_ph): Likewise.
11828 (_mm512_mask_fmadd_round_ph): Likewise.
11829 (_mm512_mask3_fmadd_round_ph): Likewise.
11830 (_mm512_maskz_fmadd_round_ph): Likewise.
11831 (_mm512_fnmadd_ph): Likewise.
11832 (_mm512_mask_fnmadd_ph): Likewise.
11833 (_mm512_mask3_fnmadd_ph): Likewise.
11834 (_mm512_maskz_fnmadd_ph): Likewise.
11835 (_mm512_fnmadd_round_ph): Likewise.
11836 (_mm512_mask_fnmadd_round_ph): Likewise.
11837 (_mm512_mask3_fnmadd_round_ph): Likewise.
11838 (_mm512_maskz_fnmadd_round_ph): Likewise.
11839 (_mm512_fmsub_ph): Likewise.
11840 (_mm512_mask_fmsub_ph): Likewise.
11841 (_mm512_mask3_fmsub_ph): Likewise.
11842 (_mm512_maskz_fmsub_ph): Likewise.
11843 (_mm512_fmsub_round_ph): Likewise.
11844 (_mm512_mask_fmsub_round_ph): Likewise.
11845 (_mm512_mask3_fmsub_round_ph): Likewise.
11846 (_mm512_maskz_fmsub_round_ph): Likewise.
11847 (_mm512_fnmsub_ph): Likewise.
11848 (_mm512_mask_fnmsub_ph): Likewise.
11849 (_mm512_mask3_fnmsub_ph): Likewise.
11850 (_mm512_maskz_fnmsub_ph): Likewise.
11851 (_mm512_fnmsub_round_ph): Likewise.
11852 (_mm512_mask_fnmsub_round_ph): Likewise.
11853 (_mm512_mask3_fnmsub_round_ph): Likewise.
11854 (_mm512_maskz_fnmsub_round_ph): Likewise.
11855 * config/i386/avx512fp16vlintrin.h (_mm256_fmadd_ph):
11857 (_mm256_mask_fmadd_ph): Likewise.
11858 (_mm256_mask3_fmadd_ph): Likewise.
11859 (_mm256_maskz_fmadd_ph): Likewise.
11860 (_mm_fmadd_ph): Likewise.
11861 (_mm_mask_fmadd_ph): Likewise.
11862 (_mm_mask3_fmadd_ph): Likewise.
11863 (_mm_maskz_fmadd_ph): Likewise.
11864 (_mm256_fnmadd_ph): Likewise.
11865 (_mm256_mask_fnmadd_ph): Likewise.
11866 (_mm256_mask3_fnmadd_ph): Likewise.
11867 (_mm256_maskz_fnmadd_ph): Likewise.
11868 (_mm_fnmadd_ph): Likewise.
11869 (_mm_mask_fnmadd_ph): Likewise.
11870 (_mm_mask3_fnmadd_ph): Likewise.
11871 (_mm_maskz_fnmadd_ph): Likewise.
11872 (_mm256_fmsub_ph): Likewise.
11873 (_mm256_mask_fmsub_ph): Likewise.
11874 (_mm256_mask3_fmsub_ph): Likewise.
11875 (_mm256_maskz_fmsub_ph): Likewise.
11876 (_mm_fmsub_ph): Likewise.
11877 (_mm_mask_fmsub_ph): Likewise.
11878 (_mm_mask3_fmsub_ph): Likewise.
11879 (_mm_maskz_fmsub_ph): Likewise.
11880 (_mm256_fnmsub_ph): Likewise.
11881 (_mm256_mask_fnmsub_ph): Likewise.
11882 (_mm256_mask3_fnmsub_ph): Likewise.
11883 (_mm256_maskz_fnmsub_ph): Likewise.
11884 (_mm_fnmsub_ph): Likewise.
11885 (_mm_mask_fnmsub_ph): Likewise.
11886 (_mm_mask3_fnmsub_ph): Likewise.
11887 (_mm_maskz_fnmsub_ph): Likewise.
11888 * config/i386/i386-builtin.def: Add corresponding new builtins.
11889 * config/i386/sse.md
11890 (<avx512>_fmadd_<mode>_maskz<round_expand_name>): Adjust to
11891 support HF vector modes.
11892 (<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name><round_name>):
11894 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1): Ditto.
11895 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_2): Ditto.
11896 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_3): Ditto.
11897 (<avx512>_fmadd_<mode>_mask<round_name>): Ditto.
11898 (<avx512>_fmadd_<mode>_mask3<round_name>): Ditto.
11899 (<avx512>_fmsub_<mode>_maskz<round_expand_name>): Ditto.
11900 (<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name><round_name>):
11902 (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1): Ditto.
11903 (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_2): Ditto.
11904 (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_3): Ditto.
11905 (<avx512>_fmsub_<mode>_mask<round_name>): Ditto.
11906 (<avx512>_fmsub_<mode>_mask3<round_name>): Ditto.
11907 (<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name><round_name>):
11909 (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1): Ditto.
11910 (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_2): Ditto.
11911 (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_3): Ditto.
11912 (<avx512>_fnmadd_<mode>_mask<round_name>): Ditto.
11913 (<avx512>_fnmadd_<mode>_mask3<round_name>): Ditto.
11914 (<avx512>_fnmsub_<mode>_maskz<round_expand_name>): Ditto.
11915 (<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name><round_name>):
11917 (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1): Ditto.
11918 (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_2): Ditto.
11919 (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_3): Ditto.
11920 (<avx512>_fnmsub_<mode>_mask<round_name>): Ditto.
11921 (<avx512>_fnmsub_<mode>_mask3<round_name>): Ditto.
11923 2021-09-18 liuhongt <hongtao.liu@intel.com>
11925 * config/i386/avx512fp16intrin.h (_mm512_fmaddsub_ph):
11927 (_mm512_mask_fmaddsub_ph): Likewise.
11928 (_mm512_mask3_fmaddsub_ph): Likewise.
11929 (_mm512_maskz_fmaddsub_ph): Likewise.
11930 (_mm512_fmaddsub_round_ph): Likewise.
11931 (_mm512_mask_fmaddsub_round_ph): Likewise.
11932 (_mm512_mask3_fmaddsub_round_ph): Likewise.
11933 (_mm512_maskz_fmaddsub_round_ph): Likewise.
11934 (_mm512_mask_fmsubadd_ph): Likewise.
11935 (_mm512_mask3_fmsubadd_ph): Likewise.
11936 (_mm512_maskz_fmsubadd_ph): Likewise.
11937 (_mm512_fmsubadd_round_ph): Likewise.
11938 (_mm512_mask_fmsubadd_round_ph): Likewise.
11939 (_mm512_mask3_fmsubadd_round_ph): Likewise.
11940 (_mm512_maskz_fmsubadd_round_ph): Likewise.
11941 * config/i386/avx512fp16vlintrin.h (_mm256_fmaddsub_ph):
11943 (_mm256_mask_fmaddsub_ph): Likewise.
11944 (_mm256_mask3_fmaddsub_ph): Likewise.
11945 (_mm256_maskz_fmaddsub_ph): Likewise.
11946 (_mm_fmaddsub_ph): Likewise.
11947 (_mm_mask_fmaddsub_ph): Likewise.
11948 (_mm_mask3_fmaddsub_ph): Likewise.
11949 (_mm_maskz_fmaddsub_ph): Likewise.
11950 (_mm256_fmsubadd_ph): Likewise.
11951 (_mm256_mask_fmsubadd_ph): Likewise.
11952 (_mm256_mask3_fmsubadd_ph): Likewise.
11953 (_mm256_maskz_fmsubadd_ph): Likewise.
11954 (_mm_fmsubadd_ph): Likewise.
11955 (_mm_mask_fmsubadd_ph): Likewise.
11956 (_mm_mask3_fmsubadd_ph): Likewise.
11957 (_mm_maskz_fmsubadd_ph): Likewise.
11958 * config/i386/i386-builtin.def: Add corresponding new builtins.
11959 * config/i386/sse.md (VFH_SF_AVX512VL): New mode iterator.
11960 * (<avx512>_fmsubadd_<mode>_maskz<round_expand_name>): New expander.
11961 * (<avx512>_fmaddsub_<mode>_maskz<round_expand_name>): Use
11963 * (<sd_mask_codefor>fma_fmaddsub_<mode><sd_maskz_name><round_name>):
11965 * (<avx512>_fmaddsub_<mode>_mask<round_name>): Ditto.
11966 * (<avx512>_fmaddsub_<mode>_mask3<round_name>): Ditto.
11967 * (<sd_mask_codefor>fma_fmsubadd_<mode><sd_maskz_name><round_name>):
11969 * (<avx512>_fmsubadd_<mode>_mask<round_name>): Ditto.
11970 * (<avx512>_fmsubadd_<mode>_mask3<round_name>): Ditto.
11972 2021-09-18 liuhongt <hongtao.liu@intel.com>
11975 * config/i386/i386.c (ix86_print_operand): Handle
11976 V8HF/V16HF/V32HFmode.
11977 * config/i386/i386.h (VALID_BCST_MODE_P): Add HFmode.
11978 * config/i386/sse.md (avx512bcst): Remove.
11980 2021-09-17 Martin Sebor <msebor@redhat.com>
11982 * Makefile.in (OBJS): Add gimple-predicate-analysis.o.
11983 * tree-ssa-uninit.c (max_phi_args): Move to gimple-predicate-analysis.
11984 (MASK_SET_BIT, MASK_TEST_BIT, MASK_EMPTY): Same.
11985 (check_defs): Add comment.
11986 (can_skip_redundant_opnd): Update comment.
11987 (compute_uninit_opnds_pos): Adjust to namespace change.
11988 (find_pdom): Move to gimple-predicate-analysis.cc.
11990 (struct uninit_undef_val_t): New.
11991 (is_non_loop_exit_postdominating): Move to gimple-predicate-analysis.cc.
11992 (find_control_equiv_block): Same.
11993 (MAX_NUM_CHAINS, MAX_CHAIN_LEN, MAX_POSTDOM_CHECK): Same.
11994 (MAX_SWITCH_CASES): Same.
11995 (compute_control_dep_chain): Same.
11996 (find_uninit_use): Use predicate analyzer.
11997 (struct pred_info): Move to gimple-predicate-analysis.
11998 (convert_control_dep_chain_into_preds): Same.
11999 (find_predicates): Same.
12000 (collect_phi_def_edges): Same.
12001 (warn_uninitialized_phi): Use predicate analyzer.
12002 (find_def_preds): Move to gimple-predicate-analysis.
12003 (dump_pred_info): Same.
12004 (dump_pred_chain): Same.
12005 (dump_predicates): Same.
12006 (destroy_predicate_vecs): Remove.
12007 (execute_late_warn_uninitialized): New.
12008 (get_cmp_code): Move to gimple-predicate-analysis.
12009 (is_value_included_in): Same.
12010 (value_sat_pred_p): Same.
12011 (find_matching_predicate_in_rest_chains): Same.
12012 (is_use_properly_guarded): Same.
12013 (prune_uninit_phi_opnds): Same.
12014 (find_var_cmp_const): Same.
12015 (use_pred_not_overlap_with_undef_path_pred): Same.
12016 (pred_equal_p): Same.
12017 (is_neq_relop_p): Same.
12018 (is_neq_zero_form_p): Same.
12019 (pred_expr_equal_p): Same.
12020 (is_pred_expr_subset_of): Same.
12021 (is_pred_chain_subset_of): Same.
12022 (is_included_in): Same.
12023 (is_superset_of): Same.
12024 (pred_neg_p): Same.
12025 (simplify_pred): Same.
12026 (simplify_preds_2): Same.
12027 (simplify_preds_3): Same.
12028 (simplify_preds_4): Same.
12029 (simplify_preds): Same.
12031 (push_to_worklist): Same.
12032 (get_pred_info_from_cmp): Same.
12033 (is_degenerated_phi): Same.
12034 (normalize_one_pred_1): Same.
12035 (normalize_one_pred): Same.
12036 (normalize_one_pred_chain): Same.
12037 (normalize_preds): Same.
12038 (can_one_predicate_be_invalidated_p): Same.
12039 (can_chain_union_be_invalidated_p): Same.
12040 (uninit_uses_cannot_happen): Same.
12041 (pass_late_warn_uninitialized::execute): Define.
12042 * gimple-predicate-analysis.cc: New file.
12043 * gimple-predicate-analysis.h: New file.
12045 2021-09-17 Julian Brown <julian@codesourcery.com>
12047 * config/gcn/gcn.c (gimple.h): Include.
12048 (gcn_fork_join): Emit barrier for worker-level joins.
12049 * omp-oacc-neuter-broadcast.cc (find_local_vars_to_propagate): Add
12050 writes_gang_private bitmap parameter. Set bit for blocks
12051 containing gang-private variable writes.
12052 (worker_single_simple): Don't emit barrier after predicated block.
12053 (worker_single_copy): Don't emit barrier if we're not broadcasting
12054 anything and the block contains no gang-private writes.
12055 (neuter_worker_single): Don't predicate blocks that only contain
12056 NOPs or internal marker functions. Pass has_gang_private_write
12057 argument to worker_single_copy.
12058 (oacc_do_neutering): Add writes_gang_private bitmap handling.
12060 2021-09-17 Julian Brown <julian@codesourcery.com>
12062 * config/gcn/gcn-protos.h
12063 (gcn_goacc_create_worker_broadcast_record): Update prototype.
12064 * config/gcn/gcn-tree.c (gcn_goacc_get_worker_red_decl): Use
12065 preallocated block of LDS memory. Do not cache/share decls for
12066 reduction temporaries between invocations.
12067 (gcn_goacc_reduction_teardown): Unshare VAR on second use.
12068 (gcn_goacc_create_worker_broadcast_record): Add OFFSET parameter
12069 and return temporary LDS space at that offset. Return pointer in
12071 * config/gcn/gcn.c (acc_lds_size, gang_private_hwm, lds_allocs):
12073 (ACC_LDS_SIZE): Define as acc_lds_size.
12074 (gcn_init_machine_status): Don't initialise lds_allocated,
12075 lds_allocs, reduc_decls fields of machine function struct.
12076 (gcn_option_override): Handle default size for gang-private
12077 variables and -mgang-private-size option.
12078 (gcn_expand_prologue): Use LDS_SIZE instead of LDS_SIZE-1 when
12079 initialising M0_REG.
12080 (gcn_shared_mem_layout): New function.
12081 (gcn_print_lds_decl): Update comment. Use global lds_allocs map and
12082 gang_private_hwm variable.
12083 (TARGET_GOACC_SHARED_MEM_LAYOUT): Define target hook.
12084 * config/gcn/gcn.h (machine_function): Remove lds_allocated,
12085 lds_allocs, reduc_decls. Add reduction_base, reduction_limit.
12086 * config/gcn/gcn.opt (gang_private_size_opt): New global.
12087 (mgang-private-size=): New option.
12088 * doc/tm.texi.in (TARGET_GOACC_SHARED_MEM_LAYOUT): Place
12089 documentation hook.
12090 * doc/tm.texi: Regenerate.
12091 * omp-oacc-neuter-broadcast.cc (targhooks.h, diagnostic-core.h):
12093 (build_sender_ref): Handle sender_decl being pointer.
12094 (worker_single_copy): Add PLACEMENT and ISOLATE_BROADCASTS
12095 parameters. Pass placement argument to
12096 create_worker_broadcast_record hook invocations. Handle
12097 sender_decl being pointer and isolate_broadcasts inserting extra
12099 (blk_offset_map_t): Add typedef.
12100 (neuter_worker_single): Add BLK_OFFSET_MAP parameter. Pass
12101 preallocated range to worker_single_copy call.
12102 (dfs_broadcast_reachable_1): New function.
12103 (idx_decl_pair_t, used_range_vec_t): New typedefs.
12104 (sort_size_descending): New function.
12105 (addr_range): New class.
12106 (splay_tree_compare_addr_range, splay_tree_free_key)
12107 (first_fit_range, merge_ranges_1, merge_ranges): New functions.
12108 (execute_omp_oacc_neuter_broadcast): Rename to...
12109 (oacc_do_neutering): ... this. Add BOUNDS_LO, BOUNDS_HI
12110 parameters. Arrange layout of shared memory for broadcast
12112 (execute_omp_oacc_neuter_broadcast): New function.
12113 (pass_omp_oacc_neuter_broadcast::gate): Remove num_workers==1
12114 handling from here. Enable pass for all OpenACC routines in order
12115 to call shared memory-layout hook.
12116 * target.def (create_worker_broadcast_record): Add OFFSET
12118 (shared_mem_layout): New hook.
12120 2021-09-17 Julian Brown <julian@codesourcery.com>
12121 Thomas Schwinge <thomas@codesourcery.com>
12123 * omp-oacc-neuter-broadcast.cc
12124 (pass_omp_oacc_neuter_broadcast::gate): Disable if num_workers is
12126 (execute_omp_oacc_neuter_broadcast): Adjust.
12128 2021-09-17 Andrew MacLeod <amacleod@redhat.com>
12130 * value-relation.cc (class equiv_chain): Move to header file.
12131 (path_oracle::path_oracle): New.
12132 (path_oracle::~path_oracle): New.
12133 (path_oracle::register_relation): New.
12134 (path_oracle::query_relation): New.
12135 (path_oracle::reset_path): New.
12136 (path_oracle::dump): New.
12137 * value-relation.h (class equiv_chain): Move to here.
12138 (class path_oracle): New.
12140 2021-09-17 Andrew MacLeod <amacleod@redhat.com>
12142 * gimple-range-cache.cc (ranger_cache::ranger_cache): Create a DOM
12144 * gimple-range-fold.cc (fur_depend::register_relation): Use
12145 register_stmt/edge routines.
12146 * value-relation.cc (equiv_chain::find): Relocate from equiv_oracle.
12147 (equiv_oracle::equiv_oracle): Create self equivalence cache.
12148 (equiv_oracle::~equiv_oracle): Release same.
12149 (equiv_oracle::equiv_set): Return entry from self equiv cache if there
12150 are no equivalences.
12151 (equiv_oracle::find_equiv_block): Move list find to equiv_chain.
12152 (equiv_oracle::register_relation): Rename from register_equiv.
12153 (relation_chain_head::find_relation): Relocate from dom_oracle.
12154 (relation_oracle::register_stmt): New.
12155 (relation_oracle::register_edge): New.
12156 (dom_oracle::*): Rename from relation_oracle.
12157 (dom_oracle::register_relation): Adjust to call equiv_oracle.
12158 (dom_oracle::set_one_relation): Split from register_relation.
12159 (dom_oracle::register_transitives): Consolidate 2 methods.
12160 (dom_oracle::find_relation_block): Move core to relation_chain.
12161 (dom_oracle::query_relation): Rename from find_relation_dom and adjust.
12162 * value-relation.h (class relation_oracle): New pure virtual base.
12163 (class equiv_oracle): Inherit from relation_oracle and adjust.
12164 (class dom_oracle): Rename from old relation_oracle and adjust.
12166 2021-09-17 Martin Sebor <msebor@redhat.com>
12168 PR middle-end/102200
12169 * pointer-query.cc (access_ref::inform_access): Handle MIN/MAX_EXPR.
12170 (handle_min_max_size): Change argument. Store original SSA_NAME for
12171 operands to potentially distinct (sub)objects.
12172 (compute_objsize_r): Adjust call to the above.
12174 2021-09-17 Bill Schmidt <wschmidt@linux.ibm.com>
12176 * config/rs6000/rs6000.c (rs6000-builtins.h): New include.
12177 (rs6000_new_builtin_vectorized_function): New function.
12178 (rs6000_new_builtin_md_vectorized_function): Likewise.
12179 (rs6000_builtin_vectorized_function): Call
12180 rs6000_new_builtin_vectorized_function.
12181 (rs6000_builtin_md_vectorized_function): Call
12182 rs6000_new_builtin_md_vectorized_function.
12184 2021-09-17 Bill Schmidt <wschmidt@linux.ibm.com>
12186 * config/rs6000/rs6000-builtin-new.def (ASSEMBLE_ACC): Add mmaint flag.
12187 (ASSEMBLE_PAIR): Likewise.
12188 (BUILD_ACC): Likewise.
12189 (DISASSEMBLE_ACC): Likewise.
12190 (DISASSEMBLE_PAIR): Likewise.
12191 (PMXVBF16GER2): Likewise.
12192 (PMXVBF16GER2NN): Likewise.
12193 (PMXVBF16GER2NP): Likewise.
12194 (PMXVBF16GER2PN): Likewise.
12195 (PMXVBF16GER2PP): Likewise.
12196 (PMXVF16GER2): Likewise.
12197 (PMXVF16GER2NN): Likewise.
12198 (PMXVF16GER2NP): Likewise.
12199 (PMXVF16GER2PN): Likewise.
12200 (PMXVF16GER2PP): Likewise.
12201 (PMXVF32GER): Likewise.
12202 (PMXVF32GERNN): Likewise.
12203 (PMXVF32GERNP): Likewise.
12204 (PMXVF32GERPN): Likewise.
12205 (PMXVF32GERPP): Likewise.
12206 (PMXVF64GER): Likewise.
12207 (PMXVF64GERNN): Likewise.
12208 (PMXVF64GERNP): Likewise.
12209 (PMXVF64GERPN): Likewise.
12210 (PMXVF64GERPP): Likewise.
12211 (PMXVI16GER2): Likewise.
12212 (PMXVI16GER2PP): Likewise.
12213 (PMXVI16GER2S): Likewise.
12214 (PMXVI16GER2SPP): Likewise.
12215 (PMXVI4GER8): Likewise.
12216 (PMXVI4GER8PP): Likewise.
12217 (PMXVI8GER4): Likewise.
12218 (PMXVI8GER4PP): Likewise.
12219 (PMXVI8GER4SPP): Likewise.
12220 (XVBF16GER2): Likewise.
12221 (XVBF16GER2NN): Likewise.
12222 (XVBF16GER2NP): Likewise.
12223 (XVBF16GER2PN): Likewise.
12224 (XVBF16GER2PP): Likewise.
12225 (XVF16GER2): Likewise.
12226 (XVF16GER2NN): Likewise.
12227 (XVF16GER2NP): Likewise.
12228 (XVF16GER2PN): Likewise.
12229 (XVF16GER2PP): Likewise.
12230 (XVF32GER): Likewise.
12231 (XVF32GERNN): Likewise.
12232 (XVF32GERNP): Likewise.
12233 (XVF32GERPN): Likewise.
12234 (XVF32GERPP): Likewise.
12235 (XVF64GER): Likewise.
12236 (XVF64GERNN): Likewise.
12237 (XVF64GERNP): Likewise.
12238 (XVF64GERPN): Likewise.
12239 (XVF64GERPP): Likewise.
12240 (XVI16GER2): Likewise.
12241 (XVI16GER2PP): Likewise.
12242 (XVI16GER2S): Likewise.
12243 (XVI16GER2SPP): Likewise.
12244 (XVI4GER8): Likewise.
12245 (XVI4GER8PP): Likewise.
12246 (XVI8GER4): Likewise.
12247 (XVI8GER4PP): Likewise.
12248 (XVI8GER4SPP): Likewise.
12249 (XXMFACC): Likewise.
12250 (XXMTACC): Likewise.
12251 (XXSETACCZ): Likewise.
12252 (ASSEMBLE_PAIR_V): Likewise.
12253 (BUILD_PAIR): Likewise.
12254 (DISASSEMBLE_PAIR_V): Likewise.
12257 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_new_mma_builtin):
12258 Handle RS6000_BIF_LXVP and RS6000_BIF_STXVP.
12259 * config/rs6000/rs6000-gen-builtins.c (attrinfo): Add ismmaint.
12260 (parse_bif_attrs): Handle ismmaint.
12261 (write_decls): Add bif_mmaint_bit and bif_is_mmaint.
12262 (write_bif_static_init): Handle ismmaint.
12264 2021-09-17 Bill Schmidt <wschmidt@linux.ibm.com>
12266 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_new_builtin): New
12268 (rs6000_gimple_fold_builtin): Call rs6000_gimple_fold_new_builtin.
12269 (rs6000_new_builtin_valid_without_lhs): New function.
12270 (rs6000_gimple_fold_new_mma_builtin): Likewise.
12271 (rs6000_gimple_fold_new_builtin): Likewise.
12273 2021-09-17 Thomas Schwinge <thomas@codesourcery.com>
12275 * hash-table.h (hash_table<Descriptor, Lazy, Allocator>::expand):
12276 Destruct stale Value objects.
12277 * hash-map-tests.c (test_map_of_type_with_ctor_and_dtor_expand):
12280 2021-09-17 Roger Sayle <roger@nextmovesoftware.com>
12283 * match.pd (shift optimizations): Disable recent sign-changing
12284 optimization for shifts by zero, these will be folded later.
12286 2021-09-17 Bill Schmidt <wschmidt@linux.ibm.com>
12288 * config/rs6000/rs6000-builtin-new.def (__builtin_mffsl): Move from
12289 [power9] to [always].
12291 2021-09-17 Richard Biener <rguenther@suse.de>
12293 * tree-vect-stmts.c (vectorizable_load): Do not frob
12296 2021-09-17 H.J. Lu <hjl.tools@gmail.com>
12298 * config/i386/i386-features.c (remove_partial_avx_dependency):
12299 Also check TARGET_SSE_PARTIAL_REG_FP_CONVERTS_DEPENDENCY and
12300 and TARGET_SSE_PARTIAL_REG_CONVERTS_DEPENDENCY before generating
12302 * config/i386/i386.h (TARGET_SSE_PARTIAL_REG_FP_CONVERTS_DEPENDENCY):
12304 (TARGET_SSE_PARTIAL_REG_CONVERTS_DEPENDENCY): Likewise.
12305 * config/i386/i386.md (SSE FP to FP splitters): Replace
12306 TARGET_SSE_PARTIAL_REG_DEPENDENCY with
12307 TARGET_SSE_PARTIAL_REG_FP_CONVERTS_DEPENDENCY.
12308 (SSE INT to FP splitter): Replace TARGET_SSE_PARTIAL_REG_DEPENDENCY
12309 with TARGET_SSE_PARTIAL_REG_CONVERTS_DEPENDENCY.
12310 * config/i386/x86-tune.def
12311 (X86_TUNE_SSE_PARTIAL_REG_FP_CONVERTS_DEPENDENCY): New.
12312 (X86_TUNE_SSE_PARTIAL_REG_CONVERTS_DEPENDENCY): Likewise.
12314 2021-09-17 H.J. Lu <hjl.tools@gmail.com>
12317 * config/i386/i386-features.c (remove_partial_avx_dependency):
12318 Check TARGET_USE_VECTOR_FP_CONVERTS and TARGET_USE_VECTOR_CONVERTS
12319 before generating vxorps.
12321 2021-09-17 H.J. Lu <hjl.tools@gmail.com>
12323 * config/i386/i386-options.c (processor_cost_table): Use
12324 tremont_cost for Tremont.
12325 * config/i386/x86-tune-costs.h (tremont_memcpy): New.
12326 (tremont_memset): Likewise.
12327 (tremont_cost): Likewise.
12328 * config/i386/x86-tune.def (X86_TUNE_PREFER_KNOWN_REP_MOVSB_STOSB):
12329 Enable for Tremont.
12331 2021-09-17 H.J. Lu <hjl.tools@gmail.com>
12333 * common/config/i386/i386-common.c: Use Haswell scheduling model
12335 * config/i386/i386.c (ix86_sched_init_global): Prepare for Tremont
12337 * config/i386/x86-tune-sched.c (ix86_issue_rate): Change Tremont
12339 (ix86_adjust_cost): Handle Tremont.
12340 * config/i386/x86-tune.def (X86_TUNE_SSE_PARTIAL_REG_DEPENDENCY):
12341 Enable for Tremont.
12342 (X86_TUNE_USE_LEAVE): Likewise.
12343 (X86_TUNE_PUSH_MEMORY): Likewise.
12344 (X86_TUNE_MISALIGNED_MOVE_STRING_PRO_EPILOGUES): Likewise.
12345 (X86_TUNE_USE_CLTD): Likewise.
12346 (X86_TUNE_AVOID_FALSE_DEP_FOR_BMI): Likewise.
12347 (X86_TUNE_AVOID_MFENCE): Likewise.
12348 (X86_TUNE_SSE_TYPELESS_STORES): Likewise.
12349 (X86_TUNE_SSE_LOAD0_BY_PXOR): Likewise.
12350 (X86_TUNE_ACCUMULATE_OUTGOING_ARGS): Disable for Tremont.
12351 (X86_TUNE_FOUR_JUMP_LIMIT): Likewise.
12352 (X86_TUNE_OPT_AGU): Likewise.
12353 (X86_TUNE_AVOID_LEA_FOR_ADDR): Likewise.
12354 (X86_TUNE_AVOID_MEM_OPND_FOR_CMOVE): Likewise.
12355 (X86_TUNE_EXPAND_ABS): Likewise.
12356 (X86_TUNE_SPLIT_MEM_OPND_FOR_FP_CONVERTS): Likewise.
12357 (X86_TUNE_SLOW_PSHUFB): Likewise.
12359 2021-09-17 Eric Botcazou <ebotcazou@adacore.com>
12361 PR rtl-optimization/102306
12362 * combine.c (try_combine): Abort the combination if we are about to
12363 duplicate volatile references.
12365 2021-09-17 liuhongt <hongtao.liu@intel.com>
12367 * config/i386/avx512fp16intrin.h (_mm_undefined_ph):
12369 (_mm256_undefined_ph): Likewise.
12370 (_mm512_undefined_ph): Likewise.
12371 (_mm_cvtsh_h): Likewise.
12372 (_mm256_cvtsh_h): Likewise.
12373 (_mm512_cvtsh_h): Likewise.
12374 (_mm512_castph_ps): Likewise.
12375 (_mm512_castph_pd): Likewise.
12376 (_mm512_castph_si512): Likewise.
12377 (_mm512_castph512_ph128): Likewise.
12378 (_mm512_castph512_ph256): Likewise.
12379 (_mm512_castph128_ph512): Likewise.
12380 (_mm512_castph256_ph512): Likewise.
12381 (_mm512_zextph128_ph512): Likewise.
12382 (_mm512_zextph256_ph512): Likewise.
12383 (_mm512_castps_ph): Likewise.
12384 (_mm512_castpd_ph): Likewise.
12385 (_mm512_castsi512_ph): Likewise.
12386 * config/i386/avx512fp16vlintrin.h (_mm_castph_ps):
12388 (_mm256_castph_ps): Likewise.
12389 (_mm_castph_pd): Likewise.
12390 (_mm256_castph_pd): Likewise.
12391 (_mm_castph_si128): Likewise.
12392 (_mm256_castph_si256): Likewise.
12393 (_mm_castps_ph): Likewise.
12394 (_mm256_castps_ph): Likewise.
12395 (_mm_castpd_ph): Likewise.
12396 (_mm256_castpd_ph): Likewise.
12397 (_mm_castsi128_ph): Likewise.
12398 (_mm256_castsi256_ph): Likewise.
12399 (_mm256_castph256_ph128): Likewise.
12400 (_mm256_castph128_ph256): Likewise.
12401 (_mm256_zextph128_ph256): Likewise.
12403 2021-09-17 liuhongt <hongtao.liu@intel.com>
12405 * config/i386/avx512fp16intrin.h (_mm_cvtsh_ss):
12407 (_mm_mask_cvtsh_ss): Likewise.
12408 (_mm_maskz_cvtsh_ss): Likewise.
12409 (_mm_cvtsh_sd): Likewise.
12410 (_mm_mask_cvtsh_sd): Likewise.
12411 (_mm_maskz_cvtsh_sd): Likewise.
12412 (_mm_cvt_roundsh_ss): Likewise.
12413 (_mm_mask_cvt_roundsh_ss): Likewise.
12414 (_mm_maskz_cvt_roundsh_ss): Likewise.
12415 (_mm_cvt_roundsh_sd): Likewise.
12416 (_mm_mask_cvt_roundsh_sd): Likewise.
12417 (_mm_maskz_cvt_roundsh_sd): Likewise.
12418 (_mm_cvtss_sh): Likewise.
12419 (_mm_mask_cvtss_sh): Likewise.
12420 (_mm_maskz_cvtss_sh): Likewise.
12421 (_mm_cvtsd_sh): Likewise.
12422 (_mm_mask_cvtsd_sh): Likewise.
12423 (_mm_maskz_cvtsd_sh): Likewise.
12424 (_mm_cvt_roundss_sh): Likewise.
12425 (_mm_mask_cvt_roundss_sh): Likewise.
12426 (_mm_maskz_cvt_roundss_sh): Likewise.
12427 (_mm_cvt_roundsd_sh): Likewise.
12428 (_mm_mask_cvt_roundsd_sh): Likewise.
12429 (_mm_maskz_cvt_roundsd_sh): Likewise.
12430 * config/i386/i386-builtin-types.def
12431 (V8HF_FTYPE_V2DF_V8HF_V8HF_UQI_INT,
12432 V8HF_FTYPE_V4SF_V8HF_V8HF_UQI_INT,
12433 V2DF_FTYPE_V8HF_V2DF_V2DF_UQI_INT,
12434 V4SF_FTYPE_V8HF_V4SF_V4SF_UQI_INT): Add new builtin types.
12435 * config/i386/i386-builtin.def: Add corrresponding new builtins.
12436 * config/i386/i386-expand.c: Handle new builtin types.
12437 * config/i386/sse.md (VF48_128): New mode iterator.
12438 (avx512fp16_vcvtsh2<ssescalarmodesuffix><mask_scalar_name><round_saeonly_scalar_name>):
12440 (avx512fp16_vcvt<ssescalarmodesuffix>2sh<mask_scalar_name><round_scalar_name>):
12443 2021-09-17 liuhongt <hongtao.liu@intel.com>
12445 * config/i386/avx512fp16intrin.h (_mm512_cvtph_pd):
12447 (_mm512_mask_cvtph_pd): Likewise.
12448 (_mm512_maskz_cvtph_pd): Likewise.
12449 (_mm512_cvt_roundph_pd): Likewise.
12450 (_mm512_mask_cvt_roundph_pd): Likewise.
12451 (_mm512_maskz_cvt_roundph_pd): Likewise.
12452 (_mm512_cvtxph_ps): Likewise.
12453 (_mm512_mask_cvtxph_ps): Likewise.
12454 (_mm512_maskz_cvtxph_ps): Likewise.
12455 (_mm512_cvtx_roundph_ps): Likewise.
12456 (_mm512_mask_cvtx_roundph_ps): Likewise.
12457 (_mm512_maskz_cvtx_roundph_ps): Likewise.
12458 (_mm512_cvtxps_ph): Likewise.
12459 (_mm512_mask_cvtxps_ph): Likewise.
12460 (_mm512_maskz_cvtxps_ph): Likewise.
12461 (_mm512_cvtx_roundps_ph): Likewise.
12462 (_mm512_mask_cvtx_roundps_ph): Likewise.
12463 (_mm512_maskz_cvtx_roundps_ph): Likewise.
12464 (_mm512_cvtpd_ph): Likewise.
12465 (_mm512_mask_cvtpd_ph): Likewise.
12466 (_mm512_maskz_cvtpd_ph): Likewise.
12467 (_mm512_cvt_roundpd_ph): Likewise.
12468 (_mm512_mask_cvt_roundpd_ph): Likewise.
12469 (_mm512_maskz_cvt_roundpd_ph): Likewise.
12470 * config/i386/avx512fp16vlintrin.h (_mm_cvtph_pd):
12472 (_mm_mask_cvtph_pd): Likewise.
12473 (_mm_maskz_cvtph_pd): Likewise.
12474 (_mm256_cvtph_pd): Likewise.
12475 (_mm256_mask_cvtph_pd): Likewise.
12476 (_mm256_maskz_cvtph_pd): Likewise.
12477 (_mm_cvtxph_ps): Likewise.
12478 (_mm_mask_cvtxph_ps): Likewise.
12479 (_mm_maskz_cvtxph_ps): Likewise.
12480 (_mm256_cvtxph_ps): Likewise.
12481 (_mm256_mask_cvtxph_ps): Likewise.
12482 (_mm256_maskz_cvtxph_ps): Likewise.
12483 (_mm_cvtxps_ph): Likewise.
12484 (_mm_mask_cvtxps_ph): Likewise.
12485 (_mm_maskz_cvtxps_ph): Likewise.
12486 (_mm256_cvtxps_ph): Likewise.
12487 (_mm256_mask_cvtxps_ph): Likewise.
12488 (_mm256_maskz_cvtxps_ph): Likewise.
12489 (_mm_cvtpd_ph): Likewise.
12490 (_mm_mask_cvtpd_ph): Likewise.
12491 (_mm_maskz_cvtpd_ph): Likewise.
12492 (_mm256_cvtpd_ph): Likewise.
12493 (_mm256_mask_cvtpd_ph): Likewise.
12494 (_mm256_maskz_cvtpd_ph): Likewise.
12495 * config/i386/i386-builtin.def: Add corresponding new builtins.
12496 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
12497 * config/i386/i386-expand.c: Handle new builtin types.
12498 * config/i386/sse.md
12499 (VF4_128_8_256): New.
12500 (VF48H_AVX512VL): Ditto.
12501 (ssePHmode): Add HF vector modes.
12502 (castmode): Add new convertable modes.
12503 (qq2phsuff): Ditto.
12504 (ph2pssuffix): New.
12505 (avx512fp16_vcvt<castmode>2ph_<mode><mask_name><round_name>): Ditto.
12506 (avx512fp16_vcvt<castmode>2ph_<mode>): Ditto.
12507 (*avx512fp16_vcvt<castmode>2ph_<mode>): Ditto.
12508 (avx512fp16_vcvt<castmode>2ph_<mode>_mask): Ditto.
12509 (*avx512fp16_vcvt<castmode>2ph_<mode>_mask): Ditto.
12510 (*avx512fp16_vcvt<castmode>2ph_<mode>_mask_1): Ditto.
12511 (avx512fp16_float_extend_ph<mode>2<mask_name><round_saeonly_name>):
12513 (avx512fp16_float_extend_ph<mode>2<mask_name>): Ditto.
12514 (*avx512fp16_float_extend_ph<mode>2_load<mask_name>): Ditto.
12515 (avx512fp16_float_extend_phv2df2<mask_name>): Ditto.
12516 (*avx512fp16_float_extend_phv2df2_load<mask_name>): Ditto.
12518 2021-09-17 liuhongt <hongtao.liu@intel.com>
12520 * config/i386/avx512fp16intrin.h (_mm_cvttsh_i32):
12522 (_mm_cvttsh_u32): Likewise.
12523 (_mm_cvtt_roundsh_i32): Likewise.
12524 (_mm_cvtt_roundsh_u32): Likewise.
12525 (_mm_cvttsh_i64): Likewise.
12526 (_mm_cvttsh_u64): Likewise.
12527 (_mm_cvtt_roundsh_i64): Likewise.
12528 (_mm_cvtt_roundsh_u64): Likewise.
12529 * config/i386/i386-builtin.def: Add corresponding new builtins.
12530 * config/i386/sse.md
12531 (avx512fp16_fix<fixunssuffix>_trunc<mode>2<round_saeonly_name>):
12534 2021-09-17 liuhongt <hongtao.liu@intel.com>
12536 * config/i386/avx512fp16intrin.h (_mm512_cvttph_epi32):
12538 (_mm512_mask_cvttph_epi32): Likewise.
12539 (_mm512_maskz_cvttph_epi32): Likewise.
12540 (_mm512_cvtt_roundph_epi32): Likewise.
12541 (_mm512_mask_cvtt_roundph_epi32): Likewise.
12542 (_mm512_maskz_cvtt_roundph_epi32): Likewise.
12543 (_mm512_cvttph_epu32): Likewise.
12544 (_mm512_mask_cvttph_epu32): Likewise.
12545 (_mm512_maskz_cvttph_epu32): Likewise.
12546 (_mm512_cvtt_roundph_epu32): Likewise.
12547 (_mm512_mask_cvtt_roundph_epu32): Likewise.
12548 (_mm512_maskz_cvtt_roundph_epu32): Likewise.
12549 (_mm512_cvttph_epi64): Likewise.
12550 (_mm512_mask_cvttph_epi64): Likewise.
12551 (_mm512_maskz_cvttph_epi64): Likewise.
12552 (_mm512_cvtt_roundph_epi64): Likewise.
12553 (_mm512_mask_cvtt_roundph_epi64): Likewise.
12554 (_mm512_maskz_cvtt_roundph_epi64): Likewise.
12555 (_mm512_cvttph_epu64): Likewise.
12556 (_mm512_mask_cvttph_epu64): Likewise.
12557 (_mm512_maskz_cvttph_epu64): Likewise.
12558 (_mm512_cvtt_roundph_epu64): Likewise.
12559 (_mm512_mask_cvtt_roundph_epu64): Likewise.
12560 (_mm512_maskz_cvtt_roundph_epu64): Likewise.
12561 (_mm512_cvttph_epi16): Likewise.
12562 (_mm512_mask_cvttph_epi16): Likewise.
12563 (_mm512_maskz_cvttph_epi16): Likewise.
12564 (_mm512_cvtt_roundph_epi16): Likewise.
12565 (_mm512_mask_cvtt_roundph_epi16): Likewise.
12566 (_mm512_maskz_cvtt_roundph_epi16): Likewise.
12567 (_mm512_cvttph_epu16): Likewise.
12568 (_mm512_mask_cvttph_epu16): Likewise.
12569 (_mm512_maskz_cvttph_epu16): Likewise.
12570 (_mm512_cvtt_roundph_epu16): Likewise.
12571 (_mm512_mask_cvtt_roundph_epu16): Likewise.
12572 (_mm512_maskz_cvtt_roundph_epu16): Likewise.
12573 * config/i386/avx512fp16vlintrin.h (_mm_cvttph_epi32):
12575 (_mm_mask_cvttph_epi32): Likewise.
12576 (_mm_maskz_cvttph_epi32): Likewise.
12577 (_mm256_cvttph_epi32): Likewise.
12578 (_mm256_mask_cvttph_epi32): Likewise.
12579 (_mm256_maskz_cvttph_epi32): Likewise.
12580 (_mm_cvttph_epu32): Likewise.
12581 (_mm_mask_cvttph_epu32): Likewise.
12582 (_mm_maskz_cvttph_epu32): Likewise.
12583 (_mm256_cvttph_epu32): Likewise.
12584 (_mm256_mask_cvttph_epu32): Likewise.
12585 (_mm256_maskz_cvttph_epu32): Likewise.
12586 (_mm_cvttph_epi64): Likewise.
12587 (_mm_mask_cvttph_epi64): Likewise.
12588 (_mm_maskz_cvttph_epi64): Likewise.
12589 (_mm256_cvttph_epi64): Likewise.
12590 (_mm256_mask_cvttph_epi64): Likewise.
12591 (_mm256_maskz_cvttph_epi64): Likewise.
12592 (_mm_cvttph_epu64): Likewise.
12593 (_mm_mask_cvttph_epu64): Likewise.
12594 (_mm_maskz_cvttph_epu64): Likewise.
12595 (_mm256_cvttph_epu64): Likewise.
12596 (_mm256_mask_cvttph_epu64): Likewise.
12597 (_mm256_maskz_cvttph_epu64): Likewise.
12598 (_mm_cvttph_epi16): Likewise.
12599 (_mm_mask_cvttph_epi16): Likewise.
12600 (_mm_maskz_cvttph_epi16): Likewise.
12601 (_mm256_cvttph_epi16): Likewise.
12602 (_mm256_mask_cvttph_epi16): Likewise.
12603 (_mm256_maskz_cvttph_epi16): Likewise.
12604 (_mm_cvttph_epu16): Likewise.
12605 (_mm_mask_cvttph_epu16): Likewise.
12606 (_mm_maskz_cvttph_epu16): Likewise.
12607 (_mm256_cvttph_epu16): Likewise.
12608 (_mm256_mask_cvttph_epu16): Likewise.
12609 (_mm256_maskz_cvttph_epu16): Likewise.
12610 * config/i386/i386-builtin.def: Add new builtins.
12611 * config/i386/sse.md
12612 (avx512fp16_fix<fixunssuffix>_trunc<mode>2<mask_name><round_saeonly_name>):
12614 (avx512fp16_fix<fixunssuffix>_trunc<mode>2<mask_name>): Ditto.
12615 (*avx512fp16_fix<fixunssuffix>_trunc<mode>2_load<mask_name>): Ditto.
12616 (avx512fp16_fix<fixunssuffix>_truncv2di2<mask_name>): Ditto.
12617 (avx512fp16_fix<fixunssuffix>_truncv2di2_load<mask_name>): Ditto.
12619 2021-09-17 liuhongt <hongtao.liu@intel.com>
12621 * config/i386/avx512fp16intrin.h (_mm_cvtsh_i32): New intrinsic.
12622 (_mm_cvtsh_u32): Likewise.
12623 (_mm_cvt_roundsh_i32): Likewise.
12624 (_mm_cvt_roundsh_u32): Likewise.
12625 (_mm_cvtsh_i64): Likewise.
12626 (_mm_cvtsh_u64): Likewise.
12627 (_mm_cvt_roundsh_i64): Likewise.
12628 (_mm_cvt_roundsh_u64): Likewise.
12629 (_mm_cvti32_sh): Likewise.
12630 (_mm_cvtu32_sh): Likewise.
12631 (_mm_cvt_roundi32_sh): Likewise.
12632 (_mm_cvt_roundu32_sh): Likewise.
12633 (_mm_cvti64_sh): Likewise.
12634 (_mm_cvtu64_sh): Likewise.
12635 (_mm_cvt_roundi64_sh): Likewise.
12636 (_mm_cvt_roundu64_sh): Likewise.
12637 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
12638 * config/i386/i386-builtin.def: Add corresponding new builtins.
12639 * config/i386/i386-expand.c (ix86_expand_round_builtin):
12640 Handle new builtin types.
12641 * config/i386/sse.md
12642 (avx512fp16_vcvtsh2<sseintconvertsignprefix>si<rex64namesuffix><round_name>):
12644 (avx512fp16_vcvtsh2<sseintconvertsignprefix>si<rex64namesuffix>_2): Likewise.
12645 (avx512fp16_vcvt<floatsuffix>si2sh<rex64namesuffix><round_name>): Likewise.
12647 2021-09-16 Bill Schmidt <wschmidt@linux.ibm.com>
12649 * config/rs6000/rs6000-c.c (rs6000-builtins.h): New include.
12650 (altivec_resolve_new_overloaded_builtin): New forward decl.
12651 (rs6000_new_builtin_type_compatible): New function.
12652 (altivec_resolve_overloaded_builtin): Call
12653 altivec_resolve_new_overloaded_builtin.
12654 (altivec_build_new_resolved_builtin): New function.
12655 (altivec_resolve_new_overloaded_builtin): Likewise.
12656 * config/rs6000/rs6000-call.c (rs6000_new_builtin_is_supported):
12658 * config/rs6000/rs6000-gen-builtins.c (write_decls): Remove _p from
12659 name of rs6000_new_builtin_is_supported.
12661 2021-09-16 Uroš Bizjak <ubizjak@gmail.com>
12663 * config/i386/i386-protos.h (ix86_decompose_address):
12664 Change return type to bool.
12665 * config/i386/i386.c (ix86_decompose_address): Ditto.
12667 2021-09-16 Tobias Burnus <tobias@codesourcery.com>
12670 * config/rs6000/t-rs6000 (build/rs6000-gen-builtins.o, build/rbtree.o):
12671 Added 'build/' to target, use build/%.o rule.
12672 (build/rs6000-gen-builtins$(build_exeext)): Add 'build/' and
12673 '$(build_exeext)' to target and 'build/' for the *.o files.
12674 (rs6000-builtins.c): Update for those changes; run rs6000-gen-builtins
12677 2021-09-16 Martin Jambor <mjambor@suse.cz>
12679 * cgraph.c (cgraph_node::dump): Do not check caller count sums if
12680 the body has been removed. Remove trailing whitespace.
12682 2021-09-16 Richard Biener <rguenther@suse.de>
12684 PR middle-end/102360
12685 * internal-fn.c (expand_DEFERRED_INIT): Make pattern-init
12686 of non-memory more robust.
12688 2021-09-16 Daniel Cederman <cederman@gaisler.com>
12690 * config/sparc/sparc-opts.h (enum sparc_processor_type): Add LEON5
12691 * config/sparc/sparc.c (struct processor_costs): Add LEON5 costs
12692 (leon5_adjust_cost): Increase cost of store with data dependency
12693 on ALU instruction and FPU anti-dependencies.
12694 (sparc_option_override): Add LEON5 costs
12695 (sparc_adjust_cost): Add LEON5 cost adjustments
12696 * config/sparc/sparc.h: Add LEON5
12697 * config/sparc/sparc.md: Include LEON5 scheduling information
12698 * config/sparc/sparc.opt: Add LEON5
12699 * doc/invoke.texi: Add LEON5
12700 * config/sparc/leon5.md: New file.
12702 2021-09-16 Daniel Cederman <cederman@gaisler.com>
12704 * config/sparc/sparc.md (stack_protect_set32): Add NOP to prevent
12705 sensitive sequence for B2BST errata workaround.
12707 2021-09-16 Daniel Cederman <cederman@gaisler.com>
12709 * config/sparc/sparc.c (sparc_do_work_around_errata): Do not begin
12710 functions with atomic instruction in the UT700 errata workaround.
12712 2021-09-16 Daniel Cederman <cederman@gaisler.com>
12714 * config/sparc/sparc.c (next_active_non_empty_insn): New function
12715 that returns next active non empty assembly instruction.
12716 (sparc_do_work_around_errata): Use new function.
12718 2021-09-16 Daniel Cederman <cederman@gaisler.com>
12720 * config/sparc/sparc.c (store_insn_p): Add predicate for store
12722 (load_insn_p): Add predicate for load attributes.
12723 (sparc_do_work_around_errata): Use new predicates.
12725 2021-09-16 Andreas Larsson <andreas@gaisler.com>
12727 * config/sparc/sparc.c (dump_target_flag_bits): Print bit names for
12730 2021-09-16 Martin Liska <mliska@suse.cz>
12732 * config/mips/netbsd.h: Fix typo in name of a macro.
12734 2021-09-16 liuhongt <hongtao.liu@intel.com>
12736 PR middle-end/102080
12737 * match.pd: Check mask type when doing cond_op related gimple
12739 * tree.c (is_truth_type_for): New function.
12740 * tree.h (is_truth_type_for): New declaration.
12742 2021-09-16 liuhongt <hongtao.liu@intel.com>
12744 * config/i386/avx512fp16intrin.h (_mm512_cvtepi32_ph): New
12746 (_mm512_mask_cvtepi32_ph): Likewise.
12747 (_mm512_maskz_cvtepi32_ph): Likewise.
12748 (_mm512_cvt_roundepi32_ph): Likewise.
12749 (_mm512_mask_cvt_roundepi32_ph): Likewise.
12750 (_mm512_maskz_cvt_roundepi32_ph): Likewise.
12751 (_mm512_cvtepu32_ph): Likewise.
12752 (_mm512_mask_cvtepu32_ph): Likewise.
12753 (_mm512_maskz_cvtepu32_ph): Likewise.
12754 (_mm512_cvt_roundepu32_ph): Likewise.
12755 (_mm512_mask_cvt_roundepu32_ph): Likewise.
12756 (_mm512_maskz_cvt_roundepu32_ph): Likewise.
12757 (_mm512_cvtepi64_ph): Likewise.
12758 (_mm512_mask_cvtepi64_ph): Likewise.
12759 (_mm512_maskz_cvtepi64_ph): Likewise.
12760 (_mm512_cvt_roundepi64_ph): Likewise.
12761 (_mm512_mask_cvt_roundepi64_ph): Likewise.
12762 (_mm512_maskz_cvt_roundepi64_ph): Likewise.
12763 (_mm512_cvtepu64_ph): Likewise.
12764 (_mm512_mask_cvtepu64_ph): Likewise.
12765 (_mm512_maskz_cvtepu64_ph): Likewise.
12766 (_mm512_cvt_roundepu64_ph): Likewise.
12767 (_mm512_mask_cvt_roundepu64_ph): Likewise.
12768 (_mm512_maskz_cvt_roundepu64_ph): Likewise.
12769 (_mm512_cvtepi16_ph): Likewise.
12770 (_mm512_mask_cvtepi16_ph): Likewise.
12771 (_mm512_maskz_cvtepi16_ph): Likewise.
12772 (_mm512_cvt_roundepi16_ph): Likewise.
12773 (_mm512_mask_cvt_roundepi16_ph): Likewise.
12774 (_mm512_maskz_cvt_roundepi16_ph): Likewise.
12775 (_mm512_cvtepu16_ph): Likewise.
12776 (_mm512_mask_cvtepu16_ph): Likewise.
12777 (_mm512_maskz_cvtepu16_ph): Likewise.
12778 (_mm512_cvt_roundepu16_ph): Likewise.
12779 (_mm512_mask_cvt_roundepu16_ph): Likewise.
12780 (_mm512_maskz_cvt_roundepu16_ph): Likewise.
12781 * config/i386/avx512fp16vlintrin.h (_mm_cvtepi32_ph): New
12783 (_mm_mask_cvtepi32_ph): Likewise.
12784 (_mm_maskz_cvtepi32_ph): Likewise.
12785 (_mm256_cvtepi32_ph): Likewise.
12786 (_mm256_mask_cvtepi32_ph): Likewise.
12787 (_mm256_maskz_cvtepi32_ph): Likewise.
12788 (_mm_cvtepu32_ph): Likewise.
12789 (_mm_mask_cvtepu32_ph): Likewise.
12790 (_mm_maskz_cvtepu32_ph): Likewise.
12791 (_mm256_cvtepu32_ph): Likewise.
12792 (_mm256_mask_cvtepu32_ph): Likewise.
12793 (_mm256_maskz_cvtepu32_ph): Likewise.
12794 (_mm_cvtepi64_ph): Likewise.
12795 (_mm_mask_cvtepi64_ph): Likewise.
12796 (_mm_maskz_cvtepi64_ph): Likewise.
12797 (_mm256_cvtepi64_ph): Likewise.
12798 (_mm256_mask_cvtepi64_ph): Likewise.
12799 (_mm256_maskz_cvtepi64_ph): Likewise.
12800 (_mm_cvtepu64_ph): Likewise.
12801 (_mm_mask_cvtepu64_ph): Likewise.
12802 (_mm_maskz_cvtepu64_ph): Likewise.
12803 (_mm256_cvtepu64_ph): Likewise.
12804 (_mm256_mask_cvtepu64_ph): Likewise.
12805 (_mm256_maskz_cvtepu64_ph): Likewise.
12806 (_mm_cvtepi16_ph): Likewise.
12807 (_mm_mask_cvtepi16_ph): Likewise.
12808 (_mm_maskz_cvtepi16_ph): Likewise.
12809 (_mm256_cvtepi16_ph): Likewise.
12810 (_mm256_mask_cvtepi16_ph): Likewise.
12811 (_mm256_maskz_cvtepi16_ph): Likewise.
12812 (_mm_cvtepu16_ph): Likewise.
12813 (_mm_mask_cvtepu16_ph): Likewise.
12814 (_mm_maskz_cvtepu16_ph): Likewise.
12815 (_mm256_cvtepu16_ph): Likewise.
12816 (_mm256_mask_cvtepu16_ph): Likewise.
12817 (_mm256_maskz_cvtepu16_ph): Likewise.
12818 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
12819 * config/i386/i386-builtin.def: Add corresponding new builtins.
12820 * config/i386/i386-expand.c
12821 (ix86_expand_args_builtin): Handle new builtin types.
12822 (ix86_expand_round_builtin): Ditto.
12823 * config/i386/i386-modes.def: Declare V2HF and V6HF.
12824 * config/i386/sse.md (VI2H_AVX512VL): New.
12825 (qq2phsuff): Ditto.
12826 (sseintvecmode): Add HF vector modes.
12827 (avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode><mask_name><round_name>):
12829 (avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>): Ditto.
12830 (*avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>): Ditto.
12831 (avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>_mask): Ditto.
12832 (*avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>_mask): Ditto.
12833 (*avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>_mask_1): Ditto.
12834 (avx512fp16_vcvt<floatsuffix>qq2ph_v2di): Ditto.
12835 (*avx512fp16_vcvt<floatsuffix>qq2ph_v2di): Ditto.
12836 (avx512fp16_vcvt<floatsuffix>qq2ph_v2di_mask): Ditto.
12837 (*avx512fp16_vcvt<floatsuffix>qq2ph_v2di_mask): Ditto.
12838 (*avx512fp16_vcvt<floatsuffix>qq2ph_v2di_mask_1): Ditto.
12839 * config/i386/subst.md (round_qq2phsuff): New subst_attr.
12841 2021-09-16 liuhongt <hongtao.liu@intel.com>
12843 * config/i386/avx512fp16intrin.h (_mm512_cvtph_epi32):
12845 (_mm512_mask_cvtph_epi32): Likewise.
12846 (_mm512_maskz_cvtph_epi32): Likewise.
12847 (_mm512_cvt_roundph_epi32): Likewise.
12848 (_mm512_mask_cvt_roundph_epi32): Likewise.
12849 (_mm512_maskz_cvt_roundph_epi32): Likewise.
12850 (_mm512_cvtph_epu32): Likewise.
12851 (_mm512_mask_cvtph_epu32): Likewise.
12852 (_mm512_maskz_cvtph_epu32): Likewise.
12853 (_mm512_cvt_roundph_epu32): Likewise.
12854 (_mm512_mask_cvt_roundph_epu32): Likewise.
12855 (_mm512_maskz_cvt_roundph_epu32): Likewise.
12856 (_mm512_cvtph_epi64): Likewise.
12857 (_mm512_mask_cvtph_epi64): Likewise.
12858 (_mm512_maskz_cvtph_epi64): Likewise.
12859 (_mm512_cvt_roundph_epi64): Likewise.
12860 (_mm512_mask_cvt_roundph_epi64): Likewise.
12861 (_mm512_maskz_cvt_roundph_epi64): Likewise.
12862 (_mm512_cvtph_epu64): Likewise.
12863 (_mm512_mask_cvtph_epu64): Likewise.
12864 (_mm512_maskz_cvtph_epu64): Likewise.
12865 (_mm512_cvt_roundph_epu64): Likewise.
12866 (_mm512_mask_cvt_roundph_epu64): Likewise.
12867 (_mm512_maskz_cvt_roundph_epu64): Likewise.
12868 (_mm512_cvtph_epi16): Likewise.
12869 (_mm512_mask_cvtph_epi16): Likewise.
12870 (_mm512_maskz_cvtph_epi16): Likewise.
12871 (_mm512_cvt_roundph_epi16): Likewise.
12872 (_mm512_mask_cvt_roundph_epi16): Likewise.
12873 (_mm512_maskz_cvt_roundph_epi16): Likewise.
12874 (_mm512_cvtph_epu16): Likewise.
12875 (_mm512_mask_cvtph_epu16): Likewise.
12876 (_mm512_maskz_cvtph_epu16): Likewise.
12877 (_mm512_cvt_roundph_epu16): Likewise.
12878 (_mm512_mask_cvt_roundph_epu16): Likewise.
12879 (_mm512_maskz_cvt_roundph_epu16): Likewise.
12880 * config/i386/avx512fp16vlintrin.h (_mm_cvtph_epi32):
12882 (_mm_mask_cvtph_epi32): Likewise.
12883 (_mm_maskz_cvtph_epi32): Likewise.
12884 (_mm256_cvtph_epi32): Likewise.
12885 (_mm256_mask_cvtph_epi32): Likewise.
12886 (_mm256_maskz_cvtph_epi32): Likewise.
12887 (_mm_cvtph_epu32): Likewise.
12888 (_mm_mask_cvtph_epu32): Likewise.
12889 (_mm_maskz_cvtph_epu32): Likewise.
12890 (_mm256_cvtph_epu32): Likewise.
12891 (_mm256_mask_cvtph_epu32): Likewise.
12892 (_mm256_maskz_cvtph_epu32): Likewise.
12893 (_mm_cvtph_epi64): Likewise.
12894 (_mm_mask_cvtph_epi64): Likewise.
12895 (_mm_maskz_cvtph_epi64): Likewise.
12896 (_mm256_cvtph_epi64): Likewise.
12897 (_mm256_mask_cvtph_epi64): Likewise.
12898 (_mm256_maskz_cvtph_epi64): Likewise.
12899 (_mm_cvtph_epu64): Likewise.
12900 (_mm_mask_cvtph_epu64): Likewise.
12901 (_mm_maskz_cvtph_epu64): Likewise.
12902 (_mm256_cvtph_epu64): Likewise.
12903 (_mm256_mask_cvtph_epu64): Likewise.
12904 (_mm256_maskz_cvtph_epu64): Likewise.
12905 (_mm_cvtph_epi16): Likewise.
12906 (_mm_mask_cvtph_epi16): Likewise.
12907 (_mm_maskz_cvtph_epi16): Likewise.
12908 (_mm256_cvtph_epi16): Likewise.
12909 (_mm256_mask_cvtph_epi16): Likewise.
12910 (_mm256_maskz_cvtph_epi16): Likewise.
12911 (_mm_cvtph_epu16): Likewise.
12912 (_mm_mask_cvtph_epu16): Likewise.
12913 (_mm_maskz_cvtph_epu16): Likewise.
12914 (_mm256_cvtph_epu16): Likewise.
12915 (_mm256_mask_cvtph_epu16): Likewise.
12916 (_mm256_maskz_cvtph_epu16): Likewise.
12917 * config/i386/i386-builtin-types.def: Add new builtin types.
12918 * config/i386/i386-builtin.def: Add new builtins.
12919 * config/i386/i386-expand.c
12920 (ix86_expand_args_builtin): Handle new builtin types.
12921 (ix86_expand_round_builtin): Ditto.
12922 * config/i386/sse.md (sseintconvert): New.
12923 (ssePHmode): Ditto.
12924 (UNSPEC_US_FIX_NOTRUNC): Ditto.
12925 (sseintconvertsignprefix): Ditto.
12926 (avx512fp16_vcvtph2<sseintconvertsignprefix><sseintconvert>_<mode><mask_name><round_name>):
12929 2021-09-16 liuhongt <hongtao.liu@intel.com>
12931 * config/i386/avx512fp16intrin.h: (_mm_cvtsi16_si128):
12933 (_mm_cvtsi128_si16): Likewise.
12934 (_mm_mask_load_sh): Likewise.
12935 (_mm_maskz_load_sh): Likewise.
12936 (_mm_mask_store_sh): Likewise.
12937 (_mm_move_sh): Likewise.
12938 (_mm_mask_move_sh): Likewise.
12939 (_mm_maskz_move_sh): Likewise.
12940 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
12941 * config/i386/i386-builtin.def: Add corresponding new builtins.
12942 * config/i386/i386-expand.c
12943 (ix86_expand_special_args_builtin): Handle new builtin types.
12944 (ix86_expand_vector_init_one_nonzero): Adjust for FP16 target.
12945 * config/i386/sse.md (VI2F): New mode iterator.
12946 (vec_set<mode>_0): Use new mode iterator.
12947 (avx512f_mov<ssescalarmodelower>_mask): Adjust for HF vector mode.
12948 (avx512f_store<mode>_mask): Ditto.
12950 2021-09-16 Kewen Lin <linkw@linux.ibm.com>
12952 * config/rs6000/rs6000.opt (-mtoc-fusion): Remove.
12954 2021-09-15 David Edelsohn <dje.gcc@gmail.com>
12956 * config/rs6000/rs6000.c (rs6000_xcoff_encode_section_info):
12957 Proceed if no symbol summary or the symbol alias flag is false.
12959 2021-09-15 Jakub Jelinek <jakub@redhat.com>
12963 * varasm.c (output_constructor_regular_field): Instead of assertion
12964 that array_size_for_constructor result is equal to size of
12965 TREE_TYPE (local->val) in bytes, assert that the type size is greater
12966 or equal to array_size_for_constructor result and use type size as
12969 2021-09-15 Martin Liska <mliska@suse.cz>
12972 * config/i386/vxworks.h: Use new macro TARGET_CPU_P.
12974 2021-09-15 Martin Liska <mliska@suse.cz>
12977 * config/rs6000/rs6000.c (rs6000_xcoff_encode_section_info):
12978 Check that we have a symbol summary for a symbol.
12980 2021-09-15 Richard Biener <rguenther@suse.de>
12983 * config/rs6000/lynx.h: Remove undef of PREFERRED_DEBUGGING_TYPE
12984 to inherit from elfos.h
12986 2021-09-15 liuhongt <hongtao.liu@intel.com>
12989 * config/i386/i386-expand.c
12990 (ix86_expand_vector_init_interleave): Use puncklwd to pack 2
12992 (ix86_expand_vector_set): Use blendw instead of pinsrw.
12993 * config/i386/i386.c (ix86_can_change_mode_class): Adjust for
12994 AVX512FP16 which supports 16bit vector load.
12995 * config/i386/sse.md (avx512bw_interleave_highv32hi<mask_name>):
12997 (avx512bw_interleave_high<mode><mask_name>): .. this, and
12998 extend to V32HFmode.
12999 (avx2_interleave_highv16hi<mask_name>): Rename to ..
13000 (avx2_interleave_high<mode><mask_name>): .. this, and extend
13002 (vec_interleave_highv8hi<mask_name>): Rename to ..
13003 (vec_interleave_high<mode><mask_name>): .. this, and extend to V8HFmode.
13004 (<mask_codefor>avx512bw_interleave_lowv32hi<mask_name>):
13006 (<mask_codefor>avx512bw_interleave_low<mode><mask_name>):
13007 this, and extend to V32HFmode.
13008 (avx2_interleave_lowv16hi<mask_name>): Rename to ..
13009 (avx2_interleave_low<mode><mask_name>): .. this, and extend to V16HFmode.
13010 (vec_interleave_lowv8hi<mask_name>): Rename to ..
13011 (vec_interleave_low<mode><mask_name>): .. this, and extend to V8HFmode.
13012 (sse4_1_pblendw): Rename to ..
13013 (sse4_1_pblend<blendsuf>): .. this, and extend to V8HFmode.
13014 (avx2_pblendph): New define_expand.
13015 (<sse2p4_1>_pinsr<ssemodesuffix>): Refactor, use
13016 sseintmodesuffix instead of ssemodesuffix.
13017 (blendsuf): New mode attr.
13019 2021-09-15 Richard Biener <rguenther@suse.de>
13021 * tree-vectorizer.h (dr_misalignment): Move out of line.
13022 (dr_target_alignment): New.
13023 (DR_TARGET_ALIGNMENT): Wrap dr_target_alignment.
13024 (set_dr_target_alignment): New.
13025 (SET_DR_TARGET_ALIGNMENT): Wrap set_dr_target_alignment.
13026 * tree-vect-data-refs.c (dr_misalignment): Compute and
13027 return the group members misalignment.
13028 (vect_compute_data_ref_alignment): Use SET_DR_TARGET_ALIGNMENT.
13029 (vect_analyze_data_refs_alignment): Compute alignment only
13030 for the first element of a DR group.
13031 (vect_slp_analyze_node_alignment): Likewise.
13033 2021-09-15 Hongyu Wang <hongyu.wang@intel.com>
13035 * config/i386/avx512fp16intrin.h: Adjust all builtin calls.
13036 * config/i386/avx512fp16vlintrin.h: Likewise.
13037 * config/i386/i386-builtin.def: Adjust builtin name and
13038 enumeration to match AVX512F style.
13040 2021-09-15 Richard Biener <rguenther@suse.de>
13042 PR tree-optimization/102318
13043 * tree-vect-loop.c (vect_transform_cycle_phi): Revert
13044 previous change and do the mode conversion separately from
13045 the sign conversion.
13047 2021-09-15 Hongtao Liu <hongtao.liu@intel.com>
13048 Peter Cordes <peter@cordes.ca>
13051 * config/i386/sse.md (extract_suf): Add V8SF/V8SI/V4DF/V4DI.
13052 (*vec_extract<mode><ssescalarmodelower>_valign): Output
13053 vextract{i,f}{32x4,64x2} instruction when byte_offset % 16 ==
13056 2021-09-15 Richard Biener <rguenther@suse.de>
13058 * config.gcc: Remove vax-*-openbsd* configuration.
13060 2021-09-15 Richard Biener <rguenther@suse.de>
13062 * config.gcc: Remove m68k-openbsd.
13064 2021-09-15 Max Filippov <jcmvbkbc@gmail.com>
13067 * config/xtensa/t-xtensa (TM_H): Add include/xtensa-config.h.
13069 2021-09-14 Peter Bergner <bergner@linux.ibm.com>
13071 * config/rs6000/mma.md (unspec): Delete UNSPEC_MMA_XXSETACCZ.
13072 (unspecv): Add UNSPECV_MMA_XXSETACCZ.
13073 (*mma_xxsetaccz): Delete.
13074 (mma_xxsetaccz): Change to define_insn. Remove operand 1.
13075 Use UNSPECV_MMA_XXSETACCZ. Update comment.
13076 * config/rs6000/rs6000.c (rs6000_rtx_costs): Use UNSPECV_MMA_XXSETACCZ.
13078 2021-09-14 Iain Sandoe <iain@sandoe.co.uk>
13080 * Makefile.in: Remove variables related to applying no-PIE
13081 to the exes on $build.
13082 * configure: Regenerate.
13083 * configure.ac: Remove configuration related to applying
13084 no-PIE to the exes on $build.
13086 2021-09-14 Claudiu Zissulescu <claziss@synopsys.com>
13088 * config/arc/arc.md (doloop_end): Add missing mode.
13089 (loop_end): Likewise.
13091 2021-09-14 Jakub Jelinek <jakub@redhat.com>
13093 * gimplify.c (goa_stabilize_expr): Add depth argument, propagate
13094 it to recursive calls, for depth above 7 just gimplify or return.
13095 Perform a test even for MODIFY_EXPR, ADDR_EXPR, COMPOUND_EXPR with
13096 __builtin_clear_padding and TARGET_EXPR.
13097 (gimplify_omp_atomic): Adjust goa_stabilize_expr callers.
13099 2021-09-14 liuhongt <hongtao.liu@intel.com>
13101 * config/i386/avx512fp16intrin.h (_mm_fpclass_sh_mask):
13103 (_mm_mask_fpclass_sh_mask): Likewise.
13104 (_mm512_mask_fpclass_ph_mask): Likewise.
13105 (_mm512_fpclass_ph_mask): Likewise.
13106 (_mm_getexp_sh): Likewise.
13107 (_mm_mask_getexp_sh): Likewise.
13108 (_mm_maskz_getexp_sh): Likewise.
13109 (_mm512_getexp_ph): Likewise.
13110 (_mm512_mask_getexp_ph): Likewise.
13111 (_mm512_maskz_getexp_ph): Likewise.
13112 (_mm_getexp_round_sh): Likewise.
13113 (_mm_mask_getexp_round_sh): Likewise.
13114 (_mm_maskz_getexp_round_sh): Likewise.
13115 (_mm512_getexp_round_ph): Likewise.
13116 (_mm512_mask_getexp_round_ph): Likewise.
13117 (_mm512_maskz_getexp_round_ph): Likewise.
13118 (_mm_getmant_sh): Likewise.
13119 (_mm_mask_getmant_sh): Likewise.
13120 (_mm_maskz_getmant_sh): Likewise.
13121 (_mm512_getmant_ph): Likewise.
13122 (_mm512_mask_getmant_ph): Likewise.
13123 (_mm512_maskz_getmant_ph): Likewise.
13124 (_mm_getmant_round_sh): Likewise.
13125 (_mm_mask_getmant_round_sh): Likewise.
13126 (_mm_maskz_getmant_round_sh): Likewise.
13127 (_mm512_getmant_round_ph): Likewise.
13128 (_mm512_mask_getmant_round_ph): Likewise.
13129 (_mm512_maskz_getmant_round_ph): Likewise.
13130 * config/i386/avx512fp16vlintrin.h (_mm_mask_fpclass_ph_mask):
13132 (_mm_fpclass_ph_mask): Likewise.
13133 (_mm256_mask_fpclass_ph_mask): Likewise.
13134 (_mm256_fpclass_ph_mask): Likewise.
13135 (_mm256_getexp_ph): Likewise.
13136 (_mm256_mask_getexp_ph): Likewise.
13137 (_mm256_maskz_getexp_ph): Likewise.
13138 (_mm_getexp_ph): Likewise.
13139 (_mm_mask_getexp_ph): Likewise.
13140 (_mm_maskz_getexp_ph): Likewise.
13141 (_mm256_getmant_ph): Likewise.
13142 (_mm256_mask_getmant_ph): Likewise.
13143 (_mm256_maskz_getmant_ph): Likewise.
13144 (_mm_getmant_ph): Likewise.
13145 (_mm_mask_getmant_ph): Likewise.
13146 (_mm_maskz_getmant_ph): Likewise.
13147 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
13148 * config/i386/i386-builtin.def: Add corresponding new builtins.
13149 * config/i386/i386-expand.c
13150 (ix86_expand_args_builtin): Handle new builtin types.
13151 (ix86_expand_round_builtin): Ditto.
13152 * config/i386/sse.md (vecmemsuffix): Add HF vector modes.
13153 (<avx512>_getexp<mode><mask_name><round_saeonly_name>): Adjust
13154 to support HF vector modes.
13155 (avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name):
13157 (avx512dq_fpclass<mode><mask_scalar_merge_name>): Ditto.
13158 (avx512dq_vmfpclass<mode><mask_scalar_merge_name>): Ditto.
13159 (<avx512>_getmant<mode><mask_name><round_saeonly_name>): Ditto.
13160 (avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
13163 2021-09-14 liuhongt <hongtao.liu@intel.com>
13165 * config/i386/avx512fp16intrin.h (_mm512_reduce_ph):
13167 (_mm512_mask_reduce_ph): Likewise.
13168 (_mm512_maskz_reduce_ph): Likewise.
13169 (_mm512_reduce_round_ph): Likewise.
13170 (_mm512_mask_reduce_round_ph): Likewise.
13171 (_mm512_maskz_reduce_round_ph): Likewise.
13172 (_mm_reduce_sh): Likewise.
13173 (_mm_mask_reduce_sh): Likewise.
13174 (_mm_maskz_reduce_sh): Likewise.
13175 (_mm_reduce_round_sh): Likewise.
13176 (_mm_mask_reduce_round_sh): Likewise.
13177 (_mm_maskz_reduce_round_sh): Likewise.
13178 (_mm512_roundscale_ph): Likewise.
13179 (_mm512_mask_roundscale_ph): Likewise.
13180 (_mm512_maskz_roundscale_ph): Likewise.
13181 (_mm512_roundscale_round_ph): Likewise.
13182 (_mm512_mask_roundscale_round_ph): Likewise.
13183 (_mm512_maskz_roundscale_round_ph): Likewise.
13184 (_mm_roundscale_sh): Likewise.
13185 (_mm_mask_roundscale_sh): Likewise.
13186 (_mm_maskz_roundscale_sh): Likewise.
13187 (_mm_roundscale_round_sh): Likewise.
13188 (_mm_mask_roundscale_round_sh): Likewise.
13189 (_mm_maskz_roundscale_round_sh): Likewise.
13190 * config/i386/avx512fp16vlintrin.h: (_mm_reduce_ph):
13192 (_mm_mask_reduce_ph): Likewise.
13193 (_mm_maskz_reduce_ph): Likewise.
13194 (_mm256_reduce_ph): Likewise.
13195 (_mm256_mask_reduce_ph): Likewise.
13196 (_mm256_maskz_reduce_ph): Likewise.
13197 (_mm_roundscale_ph): Likewise.
13198 (_mm_mask_roundscale_ph): Likewise.
13199 (_mm_maskz_roundscale_ph): Likewise.
13200 (_mm256_roundscale_ph): Likewise.
13201 (_mm256_mask_roundscale_ph): Likewise.
13202 (_mm256_maskz_roundscale_ph): Likewise.
13203 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
13204 * config/i386/i386-builtin.def: Add corresponding new builtins.
13205 * config/i386/i386-expand.c
13206 (ix86_expand_args_builtin): Handle new builtin types.
13207 (ix86_expand_round_builtin): Ditto.
13208 * config/i386/sse.md (<mask_codefor>reducep<mode><mask_name>):
13210 (<mask_codefor>reducep<mode><mask_name><round_saeonly_name>):
13211 ... this, and adjust for round operands.
13212 (reduces<mode><mask_scalar_name>): Likewise, with ...
13213 (reduces<mode><mask_scalar_name><round_saeonly_scalar_name):
13215 (<avx512>_rndscale<mode><mask_name><round_saeonly_name>):
13216 Adjust for HF vector modes.
13217 (avx512f_rndscale<mode><mask_scalar_name><round_saeonly_scalar_name>):
13219 (*avx512f_rndscale<mode><round_saeonly_name>): Ditto.
13221 2021-09-14 liuhongt <hongtao.liu@intel.com>
13223 * config/i386/avx512fp16intrin.h: (_mm512_rcp_ph):
13225 (_mm512_mask_rcp_ph): Likewise.
13226 (_mm512_maskz_rcp_ph): Likewise.
13227 (_mm_rcp_sh): Likewise.
13228 (_mm_mask_rcp_sh): Likewise.
13229 (_mm_maskz_rcp_sh): Likewise.
13230 (_mm512_scalef_ph): Likewise.
13231 (_mm512_mask_scalef_ph): Likewise.
13232 (_mm512_maskz_scalef_ph): Likewise.
13233 (_mm512_scalef_round_ph): Likewise.
13234 (_mm512_mask_scalef_round_ph): Likewise.
13235 (_mm512_maskz_scalef_round_ph): Likewise.
13236 (_mm_scalef_sh): Likewise.
13237 (_mm_mask_scalef_sh): Likewise.
13238 (_mm_maskz_scalef_sh): Likewise.
13239 (_mm_scalef_round_sh): Likewise.
13240 (_mm_mask_scalef_round_sh): Likewise.
13241 (_mm_maskz_scalef_round_sh): Likewise.
13242 * config/i386/avx512fp16vlintrin.h (_mm_rcp_ph):
13244 (_mm256_rcp_ph): Likewise.
13245 (_mm_mask_rcp_ph): Likewise.
13246 (_mm256_mask_rcp_ph): Likewise.
13247 (_mm_maskz_rcp_ph): Likewise.
13248 (_mm256_maskz_rcp_ph): Likewise.
13249 (_mm_scalef_ph): Likewise.
13250 (_mm256_scalef_ph): Likewise.
13251 (_mm_mask_scalef_ph): Likewise.
13252 (_mm256_mask_scalef_ph): Likewise.
13253 (_mm_maskz_scalef_ph): Likewise.
13254 (_mm256_maskz_scalef_ph): Likewise.
13255 * config/i386/i386-builtin.def: Add new builtins.
13256 * config/i386/sse.md (VFH_AVX512VL): New.
13257 (avx512fp16_rcp<mode>2<mask_name>): Ditto.
13258 (avx512fp16_vmrcpv8hf2<mask_scalar_name>): Ditto.
13259 (avx512f_vmscalef<mode><mask_scalar_name><round_scalar_name>):
13260 Adjust to support HF vector modes.
13261 (<avx512>_scalef<mode><mask_name><round_name>): Ditto.
13263 2021-09-14 liuhongt <hongtao.liu@intel.com>
13265 * config/i386/avx512fp16intrin.h: (_mm512_sqrt_ph):
13267 (_mm512_mask_sqrt_ph): Likewise.
13268 (_mm512_maskz_sqrt_ph): Likewise.
13269 (_mm512_sqrt_round_ph): Likewise.
13270 (_mm512_mask_sqrt_round_ph): Likewise.
13271 (_mm512_maskz_sqrt_round_ph): Likewise.
13272 (_mm512_rsqrt_ph): Likewise.
13273 (_mm512_mask_rsqrt_ph): Likewise.
13274 (_mm512_maskz_rsqrt_ph): Likewise.
13275 (_mm_rsqrt_sh): Likewise.
13276 (_mm_mask_rsqrt_sh): Likewise.
13277 (_mm_maskz_rsqrt_sh): Likewise.
13278 (_mm_sqrt_sh): Likewise.
13279 (_mm_mask_sqrt_sh): Likewise.
13280 (_mm_maskz_sqrt_sh): Likewise.
13281 (_mm_sqrt_round_sh): Likewise.
13282 (_mm_mask_sqrt_round_sh): Likewise.
13283 (_mm_maskz_sqrt_round_sh): Likewise.
13284 * config/i386/avx512fp16vlintrin.h (_mm_sqrt_ph): New intrinsic.
13285 (_mm256_sqrt_ph): Likewise.
13286 (_mm_mask_sqrt_ph): Likewise.
13287 (_mm256_mask_sqrt_ph): Likewise.
13288 (_mm_maskz_sqrt_ph): Likewise.
13289 (_mm256_maskz_sqrt_ph): Likewise.
13290 (_mm_rsqrt_ph): Likewise.
13291 (_mm256_rsqrt_ph): Likewise.
13292 (_mm_mask_rsqrt_ph): Likewise.
13293 (_mm256_mask_rsqrt_ph): Likewise.
13294 (_mm_maskz_rsqrt_ph): Likewise.
13295 (_mm256_maskz_rsqrt_ph): Likewise.
13296 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
13297 * config/i386/i386-builtin.def: Add corresponding new builtins.
13298 * config/i386/i386-expand.c
13299 (ix86_expand_args_builtin): Handle new builtins.
13300 (ix86_expand_round_builtin): Ditto.
13301 * config/i386/sse.md (VF_AVX512FP16VL): New.
13302 (sqrt<mode>2): Adjust for HF vector modes.
13303 (<sse>_sqrt<mode>2<mask_name><round_name>): Likewise.
13304 (<sse>_vmsqrt<mode>2<mask_scalar_name><round_scalar_name>):
13306 (<sse>_rsqrt<mode>2<mask_name>): New.
13307 (avx512fp16_vmrsqrtv8hf2<mask_scalar_name>): Likewise.
13309 2021-09-13 Thomas Schwinge <thomas@codesourcery.com>
13311 PR bootstrap/101574
13312 * diagnostic-spec.c (warning_suppressed_at, copy_warning): Handle
13313 'RESERVED_LOCATION_P' locations.
13314 * warning-control.cc (get_nowarn_spec, suppress_warning)
13315 (copy_warning): Likewise.
13317 2021-09-13 Thomas Schwinge <thomas@codesourcery.com>
13319 * diagnostic-spec.h (typedef xint_hash_t): Use 'location_t' instead of...
13320 (typedef key_type_t): ... this. Remove.
13321 (nowarn_map): Document.
13322 * diagnostic-spec.c (nowarn_map): Likewise.
13323 * warning-control.cc (convert_to_key): Evolve functions into...
13324 (get_location): ... these. Adjust all users.
13326 2021-09-13 Thomas Schwinge <thomas@codesourcery.com>
13328 * warning-control.cc (copy_warning): Remove 'nowarn_map' setup.
13330 2021-09-13 Jason Merrill <jason@redhat.com>
13332 * params.opt: Add destructive-interference-size and
13333 constructive-interference-size.
13334 * doc/invoke.texi: Document them.
13335 * config/aarch64/aarch64.c (aarch64_override_options_internal):
13337 * config/arm/arm.c (arm_option_override): Set them.
13338 * config/i386/i386-options.c (ix86_option_override_internal):
13341 2021-09-13 Martin Liska <mliska@suse.cz>
13342 H.J. Lu <hjl.tools@gmail.com>
13345 * common/config/i386/cpuinfo.h (cpu_indicator_init): Add support
13346 for x86-64 micro levels for __builtin_cpu_supports.
13347 * common/config/i386/i386-cpuinfo.h (enum feature_priority):
13348 Add priorities for the micro-arch levels.
13349 (enum processor_features): Add new features.
13350 * common/config/i386/i386-isas.h: Add micro-arch features.
13351 * config/i386/i386-builtins.c (get_builtin_code_for_version):
13352 Support the micro-arch levels by callsing
13353 __builtin_cpu_supports.
13354 * doc/extend.texi: Document that the levels are support by
13355 __builtin_cpu_supports.
13357 2021-09-13 Andrew Pinski <apinski@marvell.com>
13360 * config/aarch64/aarch64-builtins.c (aarch64_fold_builtin_lane_check):
13362 (aarch64_general_fold_builtin): Handle AARCH64_SIMD_BUILTIN_LANE_CHECK.
13363 (aarch64_general_gimple_fold_builtin): Likewise.
13365 2021-09-13 Andrew Pinski <apinski@marvell.com>
13367 * config.gcc: Add m32r-*-linux* and m32rle-*-linux*
13368 to the Unsupported targets list.
13369 Remove support for m32r-*-linux* and m32rle-*-linux*.
13370 * config/m32r/linux.h: Removed.
13371 * config/m32r/t-linux: Removed.
13373 2021-09-13 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13376 * config/aarch64/aarch64.c (aarch64_classify_address): Don't allow
13377 register index for SVE predicate modes.
13379 2021-09-13 Aldy Hernandez <aldyh@redhat.com>
13381 * tree-ssa-threadbackward.c
13382 (back_threader_profitability::profitable_path_p): Remove FSM
13384 (back_threader_registry::register_path): Same.
13385 * tree-ssa-threadedge.c
13386 (jump_threader::simplify_control_stmt_condition): Same.
13387 * tree-ssa-threadupdate.c (jt_path_registry::jt_path_registry):
13388 Add backedge_threads argument.
13389 (fwd_jt_path_registry::fwd_jt_path_registry): Pass
13390 backedge_threads argument.
13391 (back_jt_path_registry::back_jt_path_registry): Same.
13392 (dump_jump_thread_path): Adjust for FSM removal.
13393 (back_jt_path_registry::rewire_first_differing_edge): Same.
13394 (back_jt_path_registry::adjust_paths_after_duplication): Same.
13395 (back_jt_path_registry::update_cfg): Same.
13396 (jt_path_registry::register_jump_thread): Same.
13397 * tree-ssa-threadupdate.h (enum jump_thread_edge_type): Remove
13399 (class back_jt_path_registry): Add backedge_threads to
13402 2021-09-13 Martin Liska <mliska@suse.cz>
13405 * asan.h (sanitize_coverage_p): Handle when fn == NULL.
13407 2021-09-13 H.J. Lu <hjl.tools@gmail.com>
13410 * config/i386/i386.h (TARGET_AVX256_MOVE_BY_PIECES): New.
13411 (TARGET_AVX256_STORE_BY_PIECES): Likewise.
13412 (MOVE_MAX): Check TARGET_AVX256_MOVE_BY_PIECES and
13413 TARGET_AVX256_STORE_BY_PIECES instead of
13414 TARGET_AVX256_SPLIT_UNALIGNED_LOAD and
13415 TARGET_AVX256_SPLIT_UNALIGNED_STORE.
13416 (STORE_MAX_PIECES): Check TARGET_AVX256_STORE_BY_PIECES instead
13417 of TARGET_AVX256_SPLIT_UNALIGNED_STORE.
13418 * config/i386/x86-tune.def (X86_TUNE_AVX256_MOVE_BY_PIECES): New.
13419 (X86_TUNE_AVX256_STORE_BY_PIECES): Likewise.
13421 2021-09-13 liuhongt <hongtao.liu@intel.com>
13423 PR bootstrap/102302
13424 * expmed.c (extract_bit_field_using_extv): Use
13425 gen_lowpart_if_possible instead of gen_lowpart to avoid ICE.
13427 2021-09-13 Aldy Hernandez <aldyh@redhat.com>
13429 * Makefile.in (OBJS): Add value-pointer-equiv.o.
13430 * gimple-ssa-evrp.c (class ssa_equiv_stack): Move to
13431 value-pointer-equiv.*.
13432 (ssa_equiv_stack::ssa_equiv_stack): Same.
13433 (ssa_equiv_stack::enter): Same.
13434 (ssa_equiv_stack::leave): Same.
13435 (ssa_equiv_stack::push_replacement): Same.
13436 (ssa_equiv_stack::get_replacement): Same.
13437 (is_pointer_ssa): Same.
13438 (class pointer_equiv_analyzer): Same.
13439 (pointer_equiv_analyzer::pointer_equiv_analyzer): Same.
13440 (pointer_equiv_analyzer::~pointer_equiv_analyzer): Same.
13441 (pointer_equiv_analyzer::set_global_equiv): Same.
13442 (pointer_equiv_analyzer::set_cond_equiv): Same.
13443 (pointer_equiv_analyzer::get_equiv): Same.
13444 (pointer_equiv_analyzer::enter): Same.
13445 (pointer_equiv_analyzer::leave): Same.
13446 (pointer_equiv_analyzer::get_equiv_expr): Same.
13447 (pta_valueize): Same.
13448 (pointer_equiv_analyzer::visit_stmt): Same.
13449 (pointer_equiv_analyzer::visit_edge): Same.
13450 (hybrid_folder::value_of_expr): Same.
13451 (hybrid_folder::value_on_edge): Same.
13452 * value-pointer-equiv.cc: New file.
13453 * value-pointer-equiv.h: New file.
13455 2021-09-13 Richard Earnshaw <rearnsha@arm.com>
13458 * gimple-fold.c (gimple_fold_builtin_memory_op): Allow folding
13459 memcpy if the size is not more than MOVE_MAX * MOVE_RATIO.
13461 2021-09-13 Richard Earnshaw <rearnsha@arm.com>
13464 * config/arm/arm.md (movmisaligndi): New define_expand.
13465 * config/arm/vec-common.md (movmisalign<mode>): Iterate over VDQ mode.
13467 2021-09-13 Richard Earnshaw <rearnsha@arm.com>
13470 * emit-rtl.c (gen_highpart): Use adjust_address to handle
13471 MEM rather than calling simplify_gen_subreg.
13473 2021-09-13 Jan-Benedict Glaw <jbglaw@ług-owl.de>
13475 * config/alpha/vms.h (INIT_CUMULATIVE_ARGS): Wrap multi-statment
13476 define into a block.
13478 2021-09-13 Richard Biener <rguenther@suse.de>
13480 * config/darwin.h (DARWIN_PREFER_DWARF): Do not define.
13481 * config/i386/darwin.h (PREFERRED_DEBUGGING_TYPE): Do not
13482 change based on DARWIN_PREFER_DWARF not being defined.
13484 2021-09-13 Richard Biener <rguenther@suse.de>
13486 * config/i386/lynx.h: Remove undef of PREFERRED_DEBUGGING_TYPE
13487 to inherit from elfos.h
13489 2021-09-13 Richard Biener <rguenther@suse.de>
13491 * config.gcc: Add cr16-*-* to the list of obsoleted targets.
13493 2021-09-13 Richard Biener <rguenther@suse.de>
13495 * config/avr/elf.h (PREFERRED_DEBUGGING_TYPE): Remove
13496 override, pick up DWARF2_DEBUG define from elfos.h
13498 2021-09-13 Richard Biener <rguenther@suse.de>
13500 * config/rx/rx.h (PREFERRED_DEBUGGING_TYPE): Always define to
13503 2021-09-13 Richard Biener <rguenther@suse.de>
13505 * config/alpha/vms.h (PREFERRED_DEBUGGING_TYPE): Define to
13508 2021-09-13 Richard Biener <rguenther@suse.de>
13510 * config/i386/cygming.h: Always default to DWARF2 debugging.
13511 Do not define DBX_DEBUGGING_INFO, that's done via dbxcoff.h
13513 * doc/install.texi: Document binutils 2.16 as minimum
13514 requirement for mingw.
13516 2021-09-13 Kewen Lin <linkw@linux.ibm.com>
13518 * config/rs6000/rs6000.c (struct rs6000_cost_data): New members
13519 nstmts, nloads and extra_ctor_cost.
13520 (rs6000_density_test): Add load density related heuristics. Do
13521 extra costing on vector construction statements if need.
13522 (rs6000_init_cost): Init new members.
13523 (rs6000_update_target_cost_per_stmt): New function.
13524 (rs6000_add_stmt_cost): Factor vect_nonmem hunk out to function
13525 rs6000_update_target_cost_per_stmt and call it.
13527 2021-09-13 Kewen Lin <linkw@linux.ibm.com>
13529 * config/rs6000/rs6000.c (struct rs6000_cost_data): Remove typedef.
13530 (rs6000_init_cost): Adjust.
13532 2021-09-13 liuhongt <hongtao.liu@intel.com>
13534 * config/i386/i386.md: (UNSPEC_COPYSIGN): Remove.
13535 (UNSPEC_XORSIGN): Ditto.
13537 2021-09-12 Roger Sayle <roger@nextmovesoftware.com>
13539 * expr.c (convert_move): Preserve SUBREG_PROMOTED_VAR_P when
13540 creating a (wider) partial subreg from a SUBREG_PROMOTED_VAR_P
13543 2021-09-11 Aldy Hernandez <aldyh@redhat.com>
13545 * tree-ssa-threadbackward.c (class back_threader_registry): Use
13546 back_jt_path_registry.
13547 * tree-ssa-threadedge.c (jump_threader::jump_threader): Use
13548 fwd_jt_path_registry.
13549 * tree-ssa-threadedge.h (class jump_threader): Same..
13550 * tree-ssa-threadupdate.c
13551 (jump_thread_path_registry::jump_thread_path_registry): Rename...
13552 (jt_path_registry::jt_path_registry): ...to this.
13553 (jump_thread_path_registry::~jump_thread_path_registry): Rename...
13554 (jt_path_registry::~jt_path_registry): ...this.
13555 (fwd_jt_path_registry::fwd_jt_path_registry): New.
13556 (fwd_jt_path_registry::~fwd_jt_path_registry): New.
13557 (jump_thread_path_registry::allocate_thread_edge): Rename...
13558 (jt_path_registry::allocate_thread_edge): ...to this.
13559 (jump_thread_path_registry::allocate_thread_path): Rename...
13560 (jt_path_registry::allocate_thread_path): ...to this.
13561 (jump_thread_path_registry::lookup_redirection_data): Rename...
13562 (fwd_jt_path_registry::lookup_redirection_data): ...to this.
13563 (jump_thread_path_registry::thread_block_1): Rename...
13564 (fwd_jt_path_registry::thread_block_1): ...to this.
13565 (jump_thread_path_registry::thread_block): Rename...
13566 (fwd_jt_path_registry::thread_block): ...to this.
13567 (jt_path_registry::thread_through_loop_header): Rename...
13568 (fwd_jt_path_registry::thread_through_loop_header): ...to this.
13569 (jump_thread_path_registry::mark_threaded_blocks): Rename...
13570 (fwd_jt_path_registry::mark_threaded_blocks): ...to this.
13571 (jump_thread_path_registry::debug_path): Rename...
13572 (jt_path_registry::debug_path): ...to this.
13573 (jump_thread_path_registry::dump): Rename...
13574 (jt_path_registry::debug): ...to this.
13575 (jump_thread_path_registry::rewire_first_differing_edge): Rename...
13576 (back_jt_path_registry::rewire_first_differing_edge): ...to this.
13577 (jump_thread_path_registry::adjust_paths_after_duplication): Rename...
13578 (back_jt_path_registry::adjust_paths_after_duplication): ...to this.
13579 (jump_thread_path_registry::duplicate_thread_path): Rename...
13580 (back_jt_path_registry::duplicate_thread_path): ...to this. Also,
13581 drop ill-formed candidates.
13582 (jump_thread_path_registry::remove_jump_threads_including): Rename...
13583 (fwd_jt_path_registry::remove_jump_threads_including): ...to this.
13584 (jt_path_registry::thread_through_all_blocks): New.
13585 (back_jt_path_registry::update_cfg): New.
13586 (fwd_jt_path_registry::update_cfg): New.
13587 (jump_thread_path_registry::register_jump_thread): Rename...
13588 (jt_path_registry::register_jump_thread): ...to this.
13589 * tree-ssa-threadupdate.h (class jump_thread_path_registry):
13591 (class jt_path_registry): ...here.
13592 (class fwd_jt_path_registry): New.
13593 (class back_jt_path_registry): New.
13595 2021-09-10 liuhongt <hongtao.liu@intel.com>
13598 2021-09-01 liuhongt <hongtao.liu@intel.com>
13600 * emit-rtl.c (validate_subreg): Get rid of all float-int
13603 2021-09-10 Jakub Jelinek <jakub@redhat.com>
13605 * tree-core.h (enum omp_memory_order): Add OMP_MEMORY_ORDER_MASK,
13606 OMP_FAIL_MEMORY_ORDER_UNSPECIFIED, OMP_FAIL_MEMORY_ORDER_RELAXED,
13607 OMP_FAIL_MEMORY_ORDER_ACQUIRE, OMP_FAIL_MEMORY_ORDER_RELEASE,
13608 OMP_FAIL_MEMORY_ORDER_ACQ_REL, OMP_FAIL_MEMORY_ORDER_SEQ_CST and
13609 OMP_FAIL_MEMORY_ORDER_MASK enumerators.
13610 (OMP_FAIL_MEMORY_ORDER_SHIFT): Define.
13611 * gimple-pretty-print.c (dump_gimple_omp_atomic_load,
13612 dump_gimple_omp_atomic_store): Print [weak] for weak atomic
13614 * gimple.h (enum gf_mask): Change GF_OMP_ATOMIC_MEMORY_ORDER
13615 to 6-bit mask, adjust GF_OMP_ATOMIC_NEED_VALUE value and add
13616 GF_OMP_ATOMIC_WEAK.
13617 (gimple_omp_atomic_weak_p, gimple_omp_atomic_set_weak): New inline
13619 * tree.h (OMP_ATOMIC_WEAK): Define.
13620 * tree-pretty-print.c (dump_omp_atomic_memory_order): Adjust for
13621 fail memory order being encoded in the same enum and also print
13622 fail clause if present.
13623 (dump_generic_node): Print weak clause if OMP_ATOMIC_WEAK.
13624 * gimplify.c (goa_stabilize_expr): Add target_expr and rhs arguments,
13625 handle pre_p == NULL case as a test mode that only returns value
13626 but doesn't change gimplify nor change anything otherwise, adjust
13627 recursive calls, add MODIFY_EXPR, ADDR_EXPR, COND_EXPR, TARGET_EXPR
13628 and CALL_EXPR handling, adjust COMPOUND_EXPR handling for
13629 __builtin_clear_padding calls, for !rhs gimplify as lvalue rather
13631 (gimplify_omp_atomic): Adjust goa_stabilize_expr caller. Handle
13632 COND_EXPR rhs. Set weak flag on gimple load/store for
13634 * omp-expand.c (omp_memory_order_to_fail_memmodel): New function.
13635 (omp_memory_order_to_memmodel): Adjust for fail clause encoded
13637 (expand_omp_atomic_cas): New function.
13638 (expand_omp_atomic_pipeline): Use omp_memory_order_to_fail_memmodel
13640 (expand_omp_atomic): Attempt to optimize atomic compare and exchange
13641 using expand_omp_atomic_cas.
13643 2021-09-10 Aldy Hernandez <aldyh@redhat.com>
13644 Michael Matz <matz@suse.de>
13646 * tree-pass.h (PROP_loop_opts_done): New.
13647 * gimple-range-path.cc (path_range_query::internal_range_of_expr):
13648 Intersect with global range.
13649 * tree-ssa-loop.c (tree_ssa_loop_done): Set PROP_loop_opts_done.
13650 * tree-ssa-threadbackward.c
13651 (back_threader_profitability::profitable_path_p): Disable
13652 threading through latches until after loop optimizations have run.
13654 2021-09-10 David Faust <david.faust@oracle.com>
13656 * doc/invoke.texi: Document BPF -mcpu, -mjmpext, -mjmp32 and -malu32
13659 2021-09-10 David Faust <david.faust@oracle.com>
13661 * config/bpf/bpf-opts.h (bpf_isa_version): New enum.
13662 * config/bpf/bpf-protos.h (bpf_expand_cbranch): New.
13663 * config/bpf/bpf.c (bpf_option_override): Handle -mcpu option.
13664 (bpf_expand_cbranch): New function.
13665 * config/bpf/bpf.md (AM mode iterator): Conditionalize support for SI
13667 (zero_extendsidi2): Only use mov32 instruction if it is available.
13668 (SIM mode iterator): Conditionalize support for SI mode.
13669 (JM mode iterator): New.
13670 (cbranchdi4): Update name, use new JM iterator. Use bpf_expand_cbranch.
13671 (*branch_on_di): Update name, use new JM iterator.
13672 * config/bpf/bpf.opt: (mjmpext): New option.
13673 (malu32): Likewise.
13674 (mjmp32): Likewise.
13676 (bpf_isa): New enum.
13678 2021-09-10 David Faust <david.faust@oracle.com>
13680 * config/bpf/bpf.md (zero_extendhidi2): Add new output template
13681 for register-to-register extensions.
13682 (zero_extendqidi2): Likewise.
13684 2021-09-10 Richard Biener <rguenther@suse.de>
13686 PR middle-end/102273
13687 * internal-fn.c (expand_DEFERRED_INIT): Always expand non-SSA vars.
13689 2021-09-10 Richard Biener <rguenther@suse.de>
13691 PR middle-end/102269
13692 * gimplify.c (is_var_need_auto_init): Empty types do not need
13695 2021-09-10 Richard Biener <rguenther@suse.de>
13697 * configure.ac (--with-stabs): Remove.
13698 * configure: Regenerate.
13699 * doc/install.texi: Remove --with-stabs documentation.
13701 2021-09-10 liuhongt <hongtao.liu@intel.com>
13703 * config/i386/avx512fp16intrin.h: (_mm512_cmp_ph_mask):
13705 (_mm512_mask_cmp_ph_mask): Likewise.
13706 (_mm512_cmp_round_ph_mask): Likewise.
13707 (_mm512_mask_cmp_round_ph_mask): Likewise.
13708 (_mm_cmp_sh_mask): Likewise.
13709 (_mm_mask_cmp_sh_mask): Likewise.
13710 (_mm_cmp_round_sh_mask): Likewise.
13711 (_mm_mask_cmp_round_sh_mask): Likewise.
13712 (_mm_comieq_sh): Likewise.
13713 (_mm_comilt_sh): Likewise.
13714 (_mm_comile_sh): Likewise.
13715 (_mm_comigt_sh): Likewise.
13716 (_mm_comige_sh): Likewise.
13717 (_mm_comineq_sh): Likewise.
13718 (_mm_ucomieq_sh): Likewise.
13719 (_mm_ucomilt_sh): Likewise.
13720 (_mm_ucomile_sh): Likewise.
13721 (_mm_ucomigt_sh): Likewise.
13722 (_mm_ucomige_sh): Likewise.
13723 (_mm_ucomineq_sh): Likewise.
13724 (_mm_comi_round_sh): Likewise.
13725 (_mm_comi_sh): Likewise.
13726 * config/i386/avx512fp16vlintrin.h (_mm_cmp_ph_mask): New intrinsic.
13727 (_mm_mask_cmp_ph_mask): Likewise.
13728 (_mm256_cmp_ph_mask): Likewise.
13729 (_mm256_mask_cmp_ph_mask): Likewise.
13730 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
13731 * config/i386/i386-builtin.def: Add corresponding new builtins.
13732 * config/i386/i386-expand.c
13733 (ix86_expand_args_builtin): Handle new builtin types.
13734 (ix86_expand_round_builtin): Ditto.
13735 * config/i386/i386.md (ssevecmode): Add HF mode.
13736 (MODEFH): New mode iterator.
13737 * config/i386/sse.md
13738 (V48H_AVX512VL): New mode iterator to support HF vector modes.
13739 Ajdust corresponding description.
13740 (ssecmpintprefix): New.
13741 (VI12_AVX512VL): Adjust to support HF vector modes.
13742 (cmp_imm_predicate): Likewise.
13743 (<avx512>_cmp<mode>3<mask_scalar_merge_name><round_saeonly_name>):
13745 (avx512f_vmcmp<mode>3<round_saeonly_name>): Likewise.
13746 (avx512f_vmcmp<mode>3_mask<round_saeonly_name>): Likewise.
13747 (<sse>_<unord>comi<round_saeonly_name>): Likewise.
13749 2021-09-10 liuhongt <hongtao.liu@intel.com>
13751 * config/i386/avx512fp16intrin.h: (_mm512_max_ph): New intrinsic.
13752 (_mm512_mask_max_ph): Likewise.
13753 (_mm512_maskz_max_ph): Likewise.
13754 (_mm512_min_ph): Likewise.
13755 (_mm512_mask_min_ph): Likewise.
13756 (_mm512_maskz_min_ph): Likewise.
13757 (_mm512_max_round_ph): Likewise.
13758 (_mm512_mask_max_round_ph): Likewise.
13759 (_mm512_maskz_max_round_ph): Likewise.
13760 (_mm512_min_round_ph): Likewise.
13761 (_mm512_mask_min_round_ph): Likewise.
13762 (_mm512_maskz_min_round_ph): Likewise.
13763 (_mm_max_sh): Likewise.
13764 (_mm_mask_max_sh): Likewise.
13765 (_mm_maskz_max_sh): Likewise.
13766 (_mm_min_sh): Likewise.
13767 (_mm_mask_min_sh): Likewise.
13768 (_mm_maskz_min_sh): Likewise.
13769 (_mm_max_round_sh): Likewise.
13770 (_mm_mask_max_round_sh): Likewise.
13771 (_mm_maskz_max_round_sh): Likewise.
13772 (_mm_min_round_sh): Likewise.
13773 (_mm_mask_min_round_sh): Likewise.
13774 (_mm_maskz_min_round_sh): Likewise.
13775 * config/i386/avx512fp16vlintrin.h (_mm_max_ph): New intrinsic.
13776 (_mm256_max_ph): Likewise.
13777 (_mm_mask_max_ph): Likewise.
13778 (_mm256_mask_max_ph): Likewise.
13779 (_mm_maskz_max_ph): Likewise.
13780 (_mm256_maskz_max_ph): Likewise.
13781 (_mm_min_ph): Likewise.
13782 (_mm256_min_ph): Likewise.
13783 (_mm_mask_min_ph): Likewise.
13784 (_mm256_mask_min_ph): Likewise.
13785 (_mm_maskz_min_ph): Likewise.
13786 (_mm256_maskz_min_ph): Likewise.
13787 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
13788 * config/i386/i386-builtin.def: Add corresponding new builtins.
13789 * config/i386/i386-expand.c
13790 (ix86_expand_args_builtin): Handle new builtin types.
13791 * config/i386/sse.md
13792 (<code><mode>3<mask_name><round_saeonly_name>): Adjust to
13793 support HF vector modes.
13794 (*<code><mode>3<mask_name><round_saeonly_name>): Likewise.
13795 (ieee_<ieee_maxmin><mode>3<mask_name><round_saeonly_name>):
13797 (<sse>_vm<code><mode>3<mask_scalar_name><round_saeonly_scalar_name>):
13799 * config/i386/subst.md (round_saeonly_mode512bit_condition):
13800 Adjust for HF vector modes.
13802 2021-09-10 Liu, Hongtao <hongtao.liu@intel.com>
13804 * config/i386/avx512fp16intrin.h (_mm_add_sh): New intrinsic.
13805 (_mm_mask_add_sh): Likewise.
13806 (_mm_maskz_add_sh): Likewise.
13807 (_mm_sub_sh): Likewise.
13808 (_mm_mask_sub_sh): Likewise.
13809 (_mm_maskz_sub_sh): Likewise.
13810 (_mm_mul_sh): Likewise.
13811 (_mm_mask_mul_sh): Likewise.
13812 (_mm_maskz_mul_sh): Likewise.
13813 (_mm_div_sh): Likewise.
13814 (_mm_mask_div_sh): Likewise.
13815 (_mm_maskz_div_sh): Likewise.
13816 (_mm_add_round_sh): Likewise.
13817 (_mm_mask_add_round_sh): Likewise.
13818 (_mm_maskz_add_round_sh): Likewise.
13819 (_mm_sub_round_sh): Likewise.
13820 (_mm_mask_sub_round_sh): Likewise.
13821 (_mm_maskz_sub_round_sh): Likewise.
13822 (_mm_mul_round_sh): Likewise.
13823 (_mm_mask_mul_round_sh): Likewise.
13824 (_mm_maskz_mul_round_sh): Likewise.
13825 (_mm_div_round_sh): Likewise.
13826 (_mm_mask_div_round_sh): Likewise.
13827 (_mm_maskz_div_round_sh): Likewise.
13828 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
13829 * config/i386/i386-builtin.def: Add corresponding new builtins.
13830 * config/i386/i386-expand.c
13831 (ix86_expand_round_builtin): Handle new builtins.
13832 * config/i386/sse.md (VF_128): Change description.
13833 (<sse>_vm<plusminus_insn><mode>3<mask_scalar_name><round_scalar_name>):
13834 Adjust to support HF vector modes.
13835 (<sse>_vm<multdiv_mnemonic><mode>3<mask_scalar_name><round_scalar_name>):
13838 2021-09-10 H.J. Lu <hjl.tools@gmail.com>
13840 * config/i386/i386-expand.c
13841 (ix86_avx256_split_vector_move_misalign): Handle V16HF mode.
13842 * config/i386/i386.c
13843 (ix86_preferred_simd_mode): Handle HF mode.
13844 * config/i386/sse.md (V_256H): New mode iterator.
13845 (avx_vextractf128<mode>): Use it.
13846 (VEC_INIT_MODE): Align vector HFmode condition to vector
13847 HImodes since there're no real HF instruction used.
13848 (VEC_INIT_HALF_MODE): Ditto.
13850 (VIHF_AVX512BW): Ditto.
13851 (*vec_extracthf): Ditto.
13852 (VEC_EXTRACT_MODE): Ditto.
13854 2021-09-10 Richard Biener <rguenther@suse.de>
13857 * config/dbx.h: Remove.
13858 * config/dbxcoff.h: Do not define PREFERRED_DEBUGGING_TYPE.
13859 * config/lynx.h: Likewise.
13861 2021-09-10 liuhongt <hongtao.liu@intel.com>
13863 * config/i386/i386-expand.c (ix86_expand_copysign): Expand
13864 right into ANDNOT + AND + IOR, using paradoxical subregs.
13865 (ix86_split_copysign_const): Remove.
13866 (ix86_split_copysign_var): Ditto.
13867 * config/i386/i386-protos.h (ix86_split_copysign_const): Dotto.
13868 (ix86_split_copysign_var): Ditto.
13869 * config/i386/i386.md (@copysign<mode>3_const): Ditto.
13870 (@copysign<mode>3_var): Ditto.
13872 2021-09-09 qing zhao <qing.zhao@oracle.com>
13874 * builtins.c (expand_builtin_memset): Make external visible.
13875 * builtins.h (expand_builtin_memset): Declare extern.
13876 * common.opt (ftrivial-auto-var-init=): New option.
13877 * doc/extend.texi: Document the uninitialized attribute.
13878 * doc/invoke.texi: Document -ftrivial-auto-var-init.
13879 * flag-types.h (enum auto_init_type): New enumerated type
13881 * gimple-fold.c (clear_padding_type): Add one new parameter.
13882 (clear_padding_union): Likewise.
13883 (clear_padding_emit_loop): Likewise.
13884 (clear_type_padding_in_mask): Likewise.
13885 (gimple_fold_builtin_clear_padding): Handle this new parameter.
13886 * gimplify.c (gimple_add_init_for_auto_var): New function.
13887 (gimple_add_padding_init_for_auto_var): New function.
13888 (is_var_need_auto_init): New function.
13889 (gimplify_decl_expr): Add initialization to automatic variables per
13891 (gimplify_call_expr): Add one new parameter for call to
13892 __builtin_clear_padding.
13893 (gimplify_init_constructor): Add padding initialization in the end.
13894 * internal-fn.c (INIT_PATTERN_VALUE): New macro.
13895 (expand_DEFERRED_INIT): New function.
13896 * internal-fn.def (DEFERRED_INIT): New internal function.
13897 * tree-cfg.c (verify_gimple_call): Verify calls to .DEFERRED_INIT.
13898 * tree-sra.c (generate_subtree_deferred_init): New function.
13899 (scan_function): Avoid setting cannot_scalarize_away_bitmap for
13900 calls to .DEFERRED_INIT.
13901 (sra_modify_deferred_init): New function.
13902 (sra_modify_function_body): Handle calls to DEFERRED_INIT specially.
13903 * tree-ssa-structalias.c (find_func_aliases_for_call): Likewise.
13904 * tree-ssa-uninit.c (warn_uninit): Handle calls to DEFERRED_INIT
13906 (check_defs): Likewise.
13907 (warn_uninitialized_vars): Likewise.
13908 * tree-ssa.c (ssa_undefined_value_p): Likewise.
13909 * tree.c (build_common_builtin_nodes): Build tree node for
13910 BUILT_IN_CLEAR_PADDING when needed.
13912 2021-09-09 Richard Biener <rguenther@suse.de>
13914 * tree-ssa-loop-im.c (fill_always_executed_in_1): Walk
13917 2021-09-09 Richard Biener <rguenther@suse.de>
13919 * tree-ssa-loop-im.c (fill_always_executed_in_1): Integrate
13920 DOM walk from get_loop_body_in_dom_order using a worklist
13923 2021-09-09 liuhongt <hongtao.liu@intel.com>
13925 * config.gcc: Add avx512fp16vlintrin.h.
13926 * config/i386/avx512fp16intrin.h: (_mm512_add_ph): New intrinsic.
13927 (_mm512_mask_add_ph): Likewise.
13928 (_mm512_maskz_add_ph): Likewise.
13929 (_mm512_sub_ph): Likewise.
13930 (_mm512_mask_sub_ph): Likewise.
13931 (_mm512_maskz_sub_ph): Likewise.
13932 (_mm512_mul_ph): Likewise.
13933 (_mm512_mask_mul_ph): Likewise.
13934 (_mm512_maskz_mul_ph): Likewise.
13935 (_mm512_div_ph): Likewise.
13936 (_mm512_mask_div_ph): Likewise.
13937 (_mm512_maskz_div_ph): Likewise.
13938 (_mm512_add_round_ph): Likewise.
13939 (_mm512_mask_add_round_ph): Likewise.
13940 (_mm512_maskz_add_round_ph): Likewise.
13941 (_mm512_sub_round_ph): Likewise.
13942 (_mm512_mask_sub_round_ph): Likewise.
13943 (_mm512_maskz_sub_round_ph): Likewise.
13944 (_mm512_mul_round_ph): Likewise.
13945 (_mm512_mask_mul_round_ph): Likewise.
13946 (_mm512_maskz_mul_round_ph): Likewise.
13947 (_mm512_div_round_ph): Likewise.
13948 (_mm512_mask_div_round_ph): Likewise.
13949 (_mm512_maskz_div_round_ph): Likewise.
13950 * config/i386/avx512fp16vlintrin.h: New header.
13951 * config/i386/i386-builtin-types.def (V16HF, V8HF, V32HF):
13952 Add new builtin types.
13953 * config/i386/i386-builtin.def: Add corresponding builtins.
13954 * config/i386/i386-expand.c
13955 (ix86_expand_args_builtin): Handle new builtin types.
13956 (ix86_expand_round_builtin): Likewise.
13957 * config/i386/immintrin.h: Include avx512fp16vlintrin.h
13958 * config/i386/sse.md (VFH): New mode_iterator.
13960 (avx512fmaskmode): Add HF vector modes.
13961 (avx512fmaskhalfmode): Likewise.
13962 (<plusminus_insn><mode>3<mask_name><round_name>): Adjust to for
13964 (*<plusminus_insn><mode>3<mask_name><round_name>): Likewise.
13965 (mul<mode>3<mask_name><round_name>): Likewise.
13966 (*mul<mode>3<mask_name><round_name>): Likewise.
13967 (div<mode>3): Likewise.
13968 (<sse>_div<mode>3<mask_name><round_name>): Likewise.
13969 * config/i386/subst.md (SUBST_V): Add HF vector modes.
13970 (SUBST_A): Likewise.
13971 (round_mode512bit_condition): Adjust for V32HFmode.
13973 2021-09-09 liuhongt <hongtao.liu@intel.com>
13976 * config/i386/sse.md (reduc_plus_scal_<mode>): Split to ..
13977 (reduc_plus_scal_v4sf): .. this, New define_expand.
13978 (reduc_plus_scal_v2df): .. and this, New define_expand.
13980 2021-09-09 liuhongt <hongtao.liu@intel.com>
13983 * config/i386/sse.md (*vec_extract<mode><ssescalarmodelower>_valign):
13986 2021-09-08 Jonathan Wakely <jwakely@redhat.com>
13989 * doc/trouble.texi (Copy Assignment): Fix description of
13990 behaviour and fix code in example.
13992 2021-09-08 Segher Boessenkool <segher@kernel.crashing.org>
13995 * config/rs6000/rs6000-logue.c (rs6000_emit_epilogue): For ELFv2 use
13996 r11 instead of r12 for restoring CR.
13998 2021-09-08 Jakub Jelinek <jakub@redhat.com>
13999 liuhongt <hongtao.liu@intel.com>
14002 * config/i386/i386.md (@xorsign<mode>3_1): Remove.
14003 * config/i386/i386-expand.c (ix86_expand_xorsign): Expand right away
14004 into AND with mask and XOR, using paradoxical subregs.
14005 (ix86_split_xorsign): Remove.
14006 * config/i386/i386-protos.h (ix86_split_xorsign): Remove.
14008 2021-09-08 Di Zhao <dizhao@os.amperecomputing.com>
14010 * tree-ssa-sccvn.c (vn_nary_op_insert_into): fix result compare
14012 2021-09-08 Jakub Jelinek <jakub@redhat.com>
14015 * config/i386/i386.md (xorsign<mode>3): If operands[1] is equal to
14016 operands[2], emit abs<mode>2 instead.
14017 (@xorsign<mode>3_1): Add early-clobbers for output operand, enable
14018 first alternative even for avx, add another alternative with
14019 =&Yv <- 0, Yv, Yvm constraints.
14020 * config/i386/i386-expand.c (ix86_split_xorsign): If op0 is equal
14021 to op1, emit vpandn instead.
14023 2021-09-08 liuhongt <hongtao.liu@intel.com>
14025 * config/i386/avx512fp16intrin.h (_mm_set_ph): New intrinsic.
14026 (_mm256_set_ph): Likewise.
14027 (_mm512_set_ph): Likewise.
14028 (_mm_setr_ph): Likewise.
14029 (_mm256_setr_ph): Likewise.
14030 (_mm512_setr_ph): Likewise.
14031 (_mm_set1_ph): Likewise.
14032 (_mm256_set1_ph): Likewise.
14033 (_mm512_set1_ph): Likewise.
14034 (_mm_setzero_ph): Likewise.
14035 (_mm256_setzero_ph): Likewise.
14036 (_mm512_setzero_ph): Likewise.
14037 (_mm_set_sh): Likewise.
14038 (_mm_load_sh): Likewise.
14039 (_mm_store_sh): Likewise.
14040 * config/i386/i386-builtin-types.def (V8HF): New type.
14041 (DEF_FUNCTION_TYPE (V8HF, V8HI)): New builtin function type
14042 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
14043 Support vector HFmodes.
14044 (ix86_expand_vector_init_one_nonzero): Likewise.
14045 (ix86_expand_vector_init_one_var): Likewise.
14046 (ix86_expand_vector_init_interleave): Likewise.
14047 (ix86_expand_vector_init_general): Likewise.
14048 (ix86_expand_vector_set): Likewise.
14049 (ix86_expand_vector_extract): Likewise.
14050 (ix86_expand_vector_init_concat): Likewise.
14051 (ix86_expand_sse_movcc): Handle vector HFmodes.
14052 (ix86_expand_vector_set_var): Ditto.
14053 * config/i386/i386-modes.def: Add HF vector modes in comment.
14054 * config/i386/i386.c (classify_argument): Add HF vector modes.
14055 (ix86_hard_regno_mode_ok): Allow HF vector modes for AVX512FP16.
14056 (ix86_vector_mode_supported_p): Likewise.
14057 (ix86_set_reg_reg_cost): Handle vector HFmode.
14058 (ix86_get_ssemov): Handle vector HFmode.
14059 (function_arg_advance_64): Pass unamed V16HFmode and V32HFmode
14061 (function_arg_advance_32): Pass V8HF/V16HF/V32HF by sse reg for 32bit
14063 (function_arg_advance_32): Ditto.
14064 * config/i386/i386.h (VALID_AVX512FP16_REG_MODE): New.
14065 (VALID_AVX256_REG_OR_OI_MODE): Rename to ..
14066 (VALID_AVX256_REG_OR_OI_VHF_MODE): .. this, and add V16HF.
14067 (VALID_SSE2_REG_VHF_MODE): New.
14068 (VALID_AVX512VL_128_REG_MODE): Add V8HF and TImode.
14069 (SSE_REG_MODE_P): Add vector HFmode.
14070 * config/i386/i386.md (mode): Add HF vector modes.
14071 (MODE_SIZE): Likewise.
14072 (ssemodesuffix): Add ph suffix for HF vector modes.
14073 * config/i386/sse.md (VFH_128): New mode iterator.
14074 (VMOVE): Adjust for HF vector modes.
14076 (V_256_512): Likewise.
14077 (avx512): Likewise.
14078 (avx512fmaskmode): Likewise.
14079 (shuffletype): Likewise.
14080 (sseinsnmode): Likewise.
14081 (ssedoublevecmode): Likewise.
14082 (ssehalfvecmode): Likewise.
14083 (ssehalfvecmodelower): Likewise.
14084 (ssePScmode): Likewise.
14085 (ssescalarmode): Likewise.
14086 (ssescalarmodelower): Likewise.
14087 (sseintprefix): Likewise.
14089 (bcstscalarsuff): Likewise.
14090 (xtg_mode): Likewise.
14091 (VI12HF_AVX512VL): New mode_iterator.
14092 (VF_AVX512FP16): Likewise.
14094 (VIHF_256): Likewise.
14095 (VIHF_AVX512BW): Likewise.
14096 (V16_256): Likewise.
14097 (V32_512): Likewise.
14098 (sseintmodesuffix): New mode_attr.
14099 (sse): Add scalar and vector HFmodes.
14100 (ssescalarmode): Add vector HFmode mapping.
14101 (ssescalarmodesuffix): Add sh suffix for HFmode.
14102 (*<sse>_vm<insn><mode>3): Use VFH_128.
14103 (*<sse>_vm<multdiv_mnemonic><mode>3): Likewise.
14104 (*ieee_<ieee_maxmin><mode>3): Likewise.
14105 (<avx512>_blendm<mode>): New define_insn.
14106 (vec_setv8hf): New define_expand.
14107 (vec_set<mode>_0): New define_insn for HF vector set.
14108 (*avx512fp16_movsh): Likewise.
14109 (avx512fp16_movsh): Likewise.
14110 (vec_extract_lo_v32hi): Rename to ...
14111 (vec_extract_lo_<mode>): ... this, and adjust to allow HF
14113 (vec_extract_hi_v32hi): Likewise.
14114 (vec_extract_hi_<mode>): Likewise.
14115 (vec_extract_lo_v16hi): Likewise.
14116 (vec_extract_lo_<mode>): Likewise.
14117 (vec_extract_hi_v16hi): Likewise.
14118 (vec_extract_hi_<mode>): Likewise.
14119 (vec_set_hi_v16hi): Likewise.
14120 (vec_set_hi_<mode>): Likewise.
14121 (vec_set_lo_v16hi): Likewise.
14122 (vec_set_lo_<mode>): Likewise.
14123 (*vec_extract<mode>_0): New define_insn_and_split for HF
14125 (*vec_extracthf): New define_insn.
14126 (VEC_EXTRACT_MODE): Add HF vector modes.
14127 (PINSR_MODE): Add V8HF.
14128 (sse2p4_1): Likewise.
14129 (pinsr_evex_isa): Likewise.
14130 (<sse2p4_1>_pinsr<ssemodesuffix>): Adjust to support
14131 insert for V8HFmode.
14132 (pbroadcast_evex_isa): Add HF vector modes.
14133 (AVX2_VEC_DUP_MODE): Likewise.
14134 (VEC_INIT_MODE): Likewise.
14135 (VEC_INIT_HALF_MODE): Likewise.
14136 (avx2_pbroadcast<mode>): Adjust to support HF vector mode
14138 (avx2_pbroadcast<mode>_1): Likewise.
14139 (<avx512>_vec_dup<mode>_1): Likewise.
14140 (<avx512>_vec_dup<mode><mask_name>): Likewise.
14141 (<mask_codefor><avx512>_vec_dup_gpr<mode><mask_name>):
14144 2021-09-08 Guo, Xuepeng <xuepeng.guo@intel.com>
14145 H.J. Lu <hongjiu.lu@intel.com>
14146 Liu Hongtao <hongtao.liu@intel.com>
14147 Wang Hongyu <hongyu.wang@intel.com>
14148 Xu Dianhong <dianhong.xu@intel.com>
14150 * common/config/i386/cpuinfo.h (get_available_features):
14151 Detect FEATURE_AVX512FP16.
14152 * common/config/i386/i386-common.c
14153 (OPTION_MASK_ISA_AVX512FP16_SET,
14154 OPTION_MASK_ISA_AVX512FP16_UNSET,
14155 OPTION_MASK_ISA2_AVX512FP16_SET,
14156 OPTION_MASK_ISA2_AVX512FP16_UNSET): New.
14157 (OPTION_MASK_ISA2_AVX512BW_UNSET,
14158 OPTION_MASK_ISA2_AVX512BF16_UNSET): Add AVX512FP16.
14159 (ix86_handle_option): Handle -mavx512fp16.
14160 * common/config/i386/i386-cpuinfo.h (enum processor_features):
14161 Add FEATURE_AVX512FP16.
14162 * common/config/i386/i386-isas.h: Add entry for AVX512FP16.
14163 * config.gcc: Add avx512fp16intrin.h.
14164 * config/i386/avx512fp16intrin.h: New intrinsic header.
14165 * config/i386/cpuid.h: Add bit_AVX512FP16.
14166 * config/i386/i386-builtin-types.def: (FLOAT16): New primitive type.
14167 * config/i386/i386-builtins.c: Support _Float16 type for i386
14169 (ix86_register_float16_builtin_type): New function.
14170 (ix86_float16_type_node): New.
14171 * config/i386/i386-c.c (ix86_target_macros_internal): Define
14173 * config/i386/i386-expand.c (ix86_expand_branch): Support
14175 (ix86_prepare_fp_compare_args): Adjust TARGET_SSE_MATH &&
14176 SSE_FLOAT_MODE_P to SSE_FLOAT_MODE_SSEMATH_OR_HF_P.
14177 (ix86_expand_fp_movcc): Ditto.
14178 * config/i386/i386-isa.def: Add PTA define for AVX512FP16.
14179 * config/i386/i386-options.c (isa2_opts): Add -mavx512fp16.
14180 (ix86_valid_target_attribute_inner_p): Add avx512fp16 attribute.
14181 * config/i386/i386.c (ix86_get_ssemov): Use
14182 vmovdqu16/vmovw/vmovsh for HFmode/HImode scalar or vector.
14183 (ix86_get_excess_precision): Use
14184 FLT_EVAL_METHOD_PROMOTE_TO_FLOAT16 when TARGET_AVX512FP16
14186 (sse_store_index): Use SFmode cost for HFmode cost.
14187 (inline_memory_move_cost): Add HFmode, and perfer SSE cost over
14188 GPR cost for HFmode.
14189 (ix86_hard_regno_mode_ok): Allow HImode in sse register.
14190 (ix86_mangle_type): Add manlging for _Float16 type.
14191 (inline_secondary_memory_needed): No memory is needed for
14192 16bit movement between gpr and sse reg under
14194 (ix86_multiplication_cost): Adjust TARGET_SSE_MATH &&
14195 SSE_FLOAT_MODE_P to SSE_FLOAT_MODE_SSEMATH_OR_HF_P.
14196 (ix86_division_cost): Ditto.
14197 (ix86_rtx_costs): Ditto.
14198 (ix86_add_stmt_cost): Ditto.
14199 (ix86_optab_supported_p): Ditto.
14200 * config/i386/i386.h (VALID_AVX512F_SCALAR_MODE): Add HFmode.
14201 (SSE_FLOAT_MODE_SSEMATH_OR_HF_P): Add HFmode.
14202 (PTA_SAPPHIRERAPIDS): Add PTA_AVX512FP16.
14203 * config/i386/i386.md (mode): Add HFmode.
14204 (MODE_SIZE): Add HFmode.
14205 (isa): Add avx512fp16.
14206 (enabled): Handle avx512fp16.
14207 (ssemodesuffix): Add sh suffix for HFmode.
14208 (comm): Add mult, div.
14209 (plusminusmultdiv): New code iterator.
14210 (insn): Add mult, div.
14211 (*movhf_internal): Adjust for avx512fp16 instruction.
14212 (*movhi_internal): Ditto.
14213 (*cmpi<unord>hf): New define_insn for HFmode.
14214 (*ieee_s<ieee_maxmin>hf3): Likewise.
14215 (extendhf<mode>2): Likewise.
14216 (trunc<mode>hf2): Likewise.
14217 (float<floatunssuffix><mode>hf2): Likewise.
14218 (*<insn>hf): Likewise.
14219 (cbranchhf4): New expander.
14220 (movhfcc): Likewise.
14221 (<insn>hf3): Likewise.
14222 (mulhf3): Likewise.
14223 (divhf3): Likewise.
14224 * config/i386/i386.opt: Add mavx512fp16.
14225 * config/i386/immintrin.h: Include avx512fp16intrin.h.
14226 * doc/invoke.texi: Add mavx512fp16.
14227 * doc/extend.texi: Add avx512fp16 Usage Notes.
14229 2021-09-08 liuhongt <hongtao.liu@intel.com>
14231 * common.opt: Support -fexcess-precision=16.
14232 * config/aarch64/aarch64.c (aarch64_excess_precision): Return
14233 FLT_EVAL_METHOD_PROMOTE_TO_FLOAT16 when
14234 EXCESS_PRECISION_TYPE_FLOAT16.
14235 * config/arm/arm.c (arm_excess_precision): Ditto.
14236 * config/i386/i386.c (ix86_get_excess_precision): Ditto.
14237 * config/m68k/m68k.c (m68k_excess_precision): Issue an error
14238 when EXCESS_PRECISION_TYPE_FLOAT16.
14239 * config/s390/s390.c (s390_excess_precision): Ditto.
14240 * coretypes.h (enum excess_precision_type): Add
14241 EXCESS_PRECISION_TYPE_FLOAT16.
14242 * doc/tm.texi (TARGET_C_EXCESS_PRECISION): Update documents.
14243 * doc/tm.texi.in (TARGET_C_EXCESS_PRECISION): Ditto.
14244 * doc/extend.texi (Half-Precision): Document
14245 -fexcess-precision=16.
14246 * flag-types.h (enum excess_precision): Add
14247 EXCESS_PRECISION_FLOAT16.
14248 * target.def (excess_precision): Update document.
14249 * tree.c (excess_precision_type): Set excess_precision_type to
14250 EXCESS_PRECISION_FLOAT16 when -fexcess-precision=16.
14252 2021-09-08 liuhongt <hongtao.liu@intel.com>
14254 * doc/extend.texi: (@node Floating Types): Adjust the wording.
14255 (@node Half-Precision): Ditto.
14257 2021-09-07 Takayuki 'January June' Suwa <jjsuwa_sys3175@yahoo.co.jp>
14260 * config/xtensa/xtensa.c (xtensa_emit_move_sequence): Add
14261 'CONST_INT_P (src)' to the condition of the block that tries to
14262 eliminate literal when loading integer contant.
14264 2021-09-07 David Faust <david.faust@oracle.com>
14266 * doc/extend.texi (BPF Type Attributes) New node.
14267 Document new preserve_access_index attribute.
14268 Document new preserve_access_index builtin.
14269 * doc/invoke.texi: Document -mco-re and -mno-co-re options.
14271 2021-09-07 David Faust <david.faust@oracle.com>
14273 * config/bpf/bpf.c: Adjust includes.
14274 (bpf_handle_preserve_access_index_attribute): New function.
14275 (bpf_attribute_table): Use it here.
14276 (bpf_builtins): Add BPF_BUILTIN_PRESERVE_ACCESS_INDEX.
14277 (bpf_option_override): Handle "-mco-re" option.
14278 (bpf_asm_init_sections): New.
14279 (TARGET_ASM_INIT_SECTIONS): Redefine.
14280 (bpf_file_end): New.
14281 (TARGET_ASM_FILE_END): Redefine.
14282 (bpf_init_builtins): Add "__builtin_preserve_access_index".
14283 (bpf_core_compute, bpf_core_get_index): New.
14284 (is_attr_preserve_access): New.
14285 (bpf_expand_builtin): Handle new builtins.
14286 (bpf_core_newdecl, bpf_core_is_maybe_aggregate_access): New.
14287 (bpf_core_walk): New.
14288 (bpf_resolve_overloaded_builtin): New.
14289 (TARGET_RESOLVE_OVERLOADED_BUILTIN): Redefine.
14290 (handle_attr): New.
14291 (pass_bpf_core_attr): New RTL pass.
14292 * config/bpf/bpf-passes.def: New file.
14293 * config/bpf/bpf-protos.h (make_pass_bpf_core_attr): New.
14294 * config/bpf/coreout.c: New file.
14295 * config/bpf/coreout.h: Likewise.
14296 * config/bpf/t-bpf (TM_H): Add $(srcdir)/config/bpf/coreout.h.
14297 (coreout.o): New rule.
14298 (PASSES_EXTRA): Add $(srcdir)/config/bpf/bpf-passes.def.
14299 * config.gcc (bpf): Add coreout.h to extra_headers.
14300 Add coreout.o to extra_objs.
14301 Add $(srcdir)/config/bpf/coreout.c to target_gtfiles.
14303 2021-09-07 David Faust <david.faust@oracle.com>
14305 * btfout.c (get_btf_id): Function is no longer static.
14306 * ctfc.h: Expose it here.
14308 2021-09-07 David Faust <david.faust@oracle.com>
14310 * ctfc.c (ctf_lookup_tree_type): New function.
14311 * ctfc.h: Likewise.
14313 2021-09-07 David Faust <david.faust@oracle.com>
14315 * ctfc.c (ctf_dtd_lookup): Function is no longer static.
14316 * ctfc.h: Analogous change.
14318 2021-09-07 David Faust <david.faust@oracle.com>
14320 * dwarf2out.c (lookup_type_die): Function is no longer static.
14321 * dwarf2out.h: Expose it here.
14323 2021-09-07 Indu Bhagat <indu.bhagat@oracle.com>
14325 * dwarf2ctf.c (ctf_debug_finalize): Make it static.
14326 (ctf_debug_early_finish): New definition.
14327 (ctf_debug_finish): Likewise.
14328 * dwarf2ctf.h (ctf_debug_finalize): Remove declaration.
14329 (ctf_debug_early_finish): New declaration.
14330 (ctf_debug_finish): Likewise.
14331 * dwarf2out.c (dwarf2out_finish): Invoke ctf_debug_finish.
14332 (dwarf2out_early_finish): Invoke ctf_debug_early_finish.
14334 2021-09-07 Indu Bhagat <indu.bhagat@oracle.com>
14336 * config/bpf/bpf.c (bpf_option_override): For BPF backend, disable LTO
14337 support when compiling for CO-RE.
14338 * config/bpf/bpf.opt: Add new command line option -mco-re.
14340 2021-09-07 Indu Bhagat <indu.bhagat@oracle.com>
14342 * flag-types.h (enum debug_info_type): Add new enum
14343 DINFO_TYPE_BTF_WITH_CORE.
14344 (BTF_WITH_CORE_DEBUG): New bitmask.
14345 * flags.h (btf_with_core_debuginfo_p): New declaration.
14346 * opts.c (btf_with_core_debuginfo_p): New definition.
14348 2021-09-07 Jason Merrill <jason@redhat.com>
14350 * tree.h (error_operand_p): Change to inline function.
14352 2021-09-07 Aldy Hernandez <aldyh@redhat.com>
14354 * tree-ssa-threadedge.c (forwarder_block_p): Rename to...
14355 (empty_block_with_phis_p): ...this.
14356 (potentially_threadable_block): Same.
14357 (jump_threader::thread_through_normal_block): Same.
14359 2021-09-07 Eric Botcazou <ebotcazou@adacore.com>
14362 * dwarf2out.c (mark_base_types): New overloaded function.
14363 (dwarf2out_early_finish): Invoke it on the COMDAT type list as well
14364 as the compilation unit, and call move_marked_base_types afterward.
14366 2021-09-07 H.J. Lu <hjl.tools@gmail.com>
14369 * config/i386/i386-expand.c (ix86_expand_convert_uns_sisf_sse):
14371 (ix86_expand_vector_convert_uns_vsivsf): Likewise.
14373 2021-09-07 Richard Biener <rguenther@suse.de>
14375 PR tree-optimization/102226
14376 * tree-vect-loop.c (vect_transform_cycle_phi): Record
14377 the converted value for the epilogue PHI use.
14379 2021-09-07 Martin Liska <mliska@suse.cz>
14381 PR gcov-profile/80223
14382 * ipa-inline.c (can_inline_edge_p): Similarly to sanitizer
14383 options, do not inline when no_profile_instrument_function
14384 attributes are different in early inliner. It's fine to inline
14385 it after PGO instrumentation.
14387 2021-09-07 Richard Biener <rguenther@suse.de>
14389 PR tree-optimization/101555
14390 * tree-ssa-pre.c (translate_vuse_through_block): Do not
14391 perform an alias walk to determine the validity of the
14392 mem at the start of the block which is already guaranteed
14393 by means of prune_clobbered_mems.
14394 (phi_translate_1): Pass edge to translate_vuse_through_block.
14396 2021-09-07 Xionghu Luo <luoxhu@linux.ibm.com>
14399 * config/rs6000/rs6000.md (fmod<mode>3): New define_expand.
14400 (remainder<mode>3): Likewise.
14402 2021-09-07 YunQiang Su <yunqiang.su@cipunited.com>
14404 * config/mips/mips.c (mips_file_start): add .module for
14407 2021-09-06 Roger Sayle <roger@nextmovesoftware.com>
14409 * wide-int.cc (wi::clz): Reorder tests to ensure the result
14410 is zero for all negative values.
14412 2021-09-06 Tobias Burnus <tobias@codesourcery.com>
14414 * doc/invoke.texi (-foffload-options): Fix @opindex.
14416 2021-09-06 H.J. Lu <hjl.tools@gmail.com>
14419 * config/i386/i386-expand.c (ix86_split_xorsign): Use operands[2].
14420 * config/i386/i386.md (@xorsign<mode>3_1): Add non-destructive
14421 source alternative for AVX.
14423 2021-09-06 liuhongt <hongtao.liu@intel.com>
14425 PR middle-end/102182
14426 * optabs.c (expand_fix): Add from1 to avoid from being
14429 2021-09-06 Eric Botcazou <ebotcazou@adacore.com>
14431 * dwarf2out.c (modified_type_die): Deal with all array types earlier
14432 and use local variable consistently throughout the function.
14434 2021-09-06 Jakub Jelinek <jakub@redhat.com>
14436 PR tree-optimization/102207
14437 * match.pd: Don't demote operands of IFN_{ADD,SUB,MUL}_OVERFLOW if they
14438 were promoted from signed to wider unsigned type.
14440 2021-09-06 Andrew Pinski <apinski@marvell.com>
14442 PR tree-optimization/63184
14443 * match.pd: Add simplification of pointer_diff of two pointer_plus
14444 with addr_expr in the first operand of each pointer_plus.
14445 Add simplificatoin of ne/eq of two pointer_plus with addr_expr
14446 in the first operand of each pointer_plus.
14448 2021-09-06 Richard Biener <rguenther@suse.de>
14450 PR tree-optimization/102176
14451 * tree-vect-slp.c (vect_slp_gather_vectorized_scalar_stmts):
14453 (vect_bb_slp_scalar_cost): Use the computed set of
14454 vectorized scalar stmts instead of relying on the out-of-date
14455 and not accurate PURE_SLP_STMT.
14456 (vect_bb_vectorization_profitable_p): Compute the set
14457 of vectorized scalar stmts.
14459 2021-09-05 Aldy Hernandez <aldyh@redhat.com>
14461 * gimple-range-path.cc (path_range_query::range_of_stmt): Remove
14462 GIMPLE_COND special casing.
14463 (path_range_query::range_defined_in_block): Use range_of_stmt
14464 instead of calling fold_range directly.
14466 2021-09-05 Aldy Hernandez <aldyh@redhat.com>
14468 * gimple-range-path.cc (path_range_query::range_of_expr): Set
14469 m_undefined_path when appropriate.
14470 (path_range_query::internal_range_of_expr): Copy from range_of_expr.
14471 (path_range_query::unreachable_path_p): New.
14472 (path_range_query::precompute_ranges): Set m_undefined_path.
14473 * gimple-range-path.h (path_range_query::unreachable_path_p): New.
14474 (path_range_query::internal_range_of_expr): New.
14475 * tree-ssa-threadbackward.c (back_threader::find_taken_edge_cond):
14476 Use unreachable_path_p.
14478 2021-09-05 Aldy Hernandez <aldyh@redhat.com>
14480 * tree-ssa-threadbackward.c (back_threader::maybe_register_path):
14481 Remove argument and call find_taken_edge.
14482 (back_threader::resolve_phi): Do not calculate taken edge before
14483 calling maybe_register_path.
14484 (back_threader::find_paths_to_names): Same.
14486 2021-09-05 Jeff Law <jlaw@localhost.localdomain>
14488 * config/h8300/h8300.md (QHSI2 mode iterator): New mode iterator.
14489 * config/h8300/testcompare.md (store_c): Update name, use new
14491 (store_neg_c, store_shifted_c): New patterns.
14493 2021-09-03 Segher Boessenkool <segher@kernel.crashing.org>
14496 * config/rs6000/rs6000-logue.c (rs6000_emit_prologue): On ELFv2 use r11
14497 instead of r12 for CR save, in all cases.
14499 2021-09-03 Andrew Pinski <apinski@marvell.com>
14501 * config/aarch64/aarch64-sve-builtins.cc (register_vector_type):
14502 Handle error_mark_node as the type of the type_decl.
14504 2021-09-03 Andrew Pinski <apinski@marvell.com>
14506 * config/aarch64/aarch64-builtins.c (struct aarch64_simd_type_info):
14508 (aarch64_simd_types): Likewise.
14509 (aarch64_simd_intOI_type_node): Likewise.
14510 (aarch64_simd_intCI_type_node): Likewise.
14511 (aarch64_simd_intXI_type_node): Likewise.
14512 * config/aarch64/aarch64.h (aarch64_fp16_type_node): Likewise.
14513 (aarch64_fp16_ptr_type_node): Likewise.
14514 (aarch64_bf16_type_node): Likewise.
14515 (aarch64_bf16_ptr_type_node): Likewise.
14517 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
14519 * range-op.cc (operator_minus::op1_op2_relation_effect): Abstract
14521 (minus_op1_op2_relation_effect): ...here.
14522 (class operator_pointer_diff): New.
14523 (operator_pointer_diff::op1_op2_relation_effect): Call
14524 minus_op1_op2_relation_effect.
14525 (integral_table::integral_table): Add entry for POINTER_DIFF_EXPR.
14527 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
14529 * tree-ssa-threadbackward.c (back_threader::thread_through_all_blocks):
14530 Add may_peel_loop_headers.
14531 (back_threader_registry::thread_through_all_blocks): Same.
14532 (try_thread_blocks): Pass may_peel_loop_headers argument.
14533 (pass_early_thread_jumps::execute): Same.
14535 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
14537 * tree-ssa-threadedge.c (has_phis_p): New.
14538 (forwarder_block_p): New.
14539 (potentially_threadable_block): Call forwarder_block_p.
14540 (jump_threader::thread_around_empty_blocks): Call has_phis_p.
14541 (jump_threader::thread_through_normal_block): Call
14544 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
14546 * tree-ssa-threadbackward.c (back_threader::dump): New.
14547 (back_threader::debug): New.
14548 (back_threader_profitability::profitable_path_p): Dump blocks
14549 even if we are bailing early.
14551 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
14553 * tree-ssa-threadupdate.c (cancel_thread): New.
14554 (jump_thread_path_registry::thread_block_1): Use cancel_thread.
14555 (jump_thread_path_registry::mark_threaded_blocks): Same.
14556 (jump_thread_path_registry::register_jump_thread): Same.
14558 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
14560 * tree-ssa-threadedge.c (jt_state::push): Only call methods for
14561 which objects are available.
14562 (jt_state::pop): Same.
14563 (jt_state::register_equiv): Same.
14564 (jt_state::register_equivs_on_edge): Same.
14566 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
14568 * tree-ssa-threadedge.c (jump_threader::thread_across_edge):
14569 Move pop until after a thread is registered.
14571 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
14573 * tree-ssa-threadupdate.c (debug): New.
14575 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
14577 * gimple-range-trace.cc (push_dump_file::push_dump_file): New.
14578 (push_dump_file::~push_dump_file): New.
14579 (dump_ranger): Change dump_file temporarily while dumping
14581 * gimple-range-trace.h (class push_dump_file): New.
14583 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
14585 * gimple-range-trace.cc (debug_seed_ranger): Remove static.
14586 (dump_ranger): Dump function name.
14588 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
14590 * gimple-range-path.cc (path_range_query::range_defined_in_block):
14591 Adjust for non-null.
14592 (path_range_query::adjust_for_non_null_uses): New.
14593 (path_range_query::precompute_ranges): Call
14594 adjust_for_non_null_uses.
14595 * gimple-range-path.h: Add m_non_null and
14596 adjust_for_non_null_uses.
14598 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
14600 * gimple-range-path.cc (path_range_query::dump): Dump path
14602 (path_range_query::precompute_ranges): Dump entire path.
14604 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
14606 * value-relation.cc (relation_oracle::debug): New.
14607 * value-relation.h (relation_oracle::debug): New.
14609 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
14611 * tree-ssa-loop-ch.c: Remove unnecessary include file.
14613 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
14615 * gimple-range-fold.cc (fold_using_range::postfold_gcond_edges):
14616 Skip statements with no defining BB.
14617 * gimple-range-path.cc (path_range_query::range_defined_in_block):
14618 Do not get confused by statements with no defining BB.
14620 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
14622 * gimple-range-fold.cc (adjust_imagpart_expr): Move from
14623 gimple_range_adjustment. Add support for constants.
14624 (adjust_realpart_expr): New.
14625 (gimple_range_adjustment): Move IMAGPART_EXPR code to
14626 adjust_imagpart_expr.
14627 * range-op.cc (integral_table::integral_table): Add entry for
14630 2021-09-03 Jakub Jelinek <jakub@redhat.com>
14632 * omp-expand.c (expand_omp_atomic_pipeline): Use
14633 IFN_ATOMIC_COMPARE_EXCHANGE instead of
14634 BUILT_IN_SYNC_VAL_COMPARE_AND_SWAP_? so that memory order
14637 2021-09-03 Jakub Jelinek <jakub@redhat.com>
14640 * tree.h (DECL_FIELD_ABI_IGNORED): Changed into rvalue only macro
14641 that is false if DECL_BIT_FIELD.
14642 (SET_DECL_FIELD_ABI_IGNORED, DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD,
14643 SET_DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD): Define.
14644 * tree-streamer-out.c (pack_ts_decl_common_value_fields): For
14645 DECL_BIT_FIELD stream DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD instead
14646 of DECL_FIELD_ABI_IGNORED.
14647 * tree-streamer-in.c (unpack_ts_decl_common_value_fields): Use
14648 SET_DECL_FIELD_ABI_IGNORED instead of writing to
14649 DECL_FIELD_ABI_IGNORED and for DECL_BIT_FIELD use
14650 SET_DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD instead.
14651 * lto-streamer-out.c (hash_tree): For DECL_BIT_FIELD hash
14652 DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD instead of DECL_FIELD_ABI_IGNORED.
14654 2021-09-03 liuhongt <hongtao.liu@intel.com>
14657 * config/i386/amxbf16intrin.h : Remove macro check for __AMX_BF16__.
14658 * config/i386/amxint8intrin.h : Remove macro check for __AMX_INT8__.
14659 * config/i386/amxtileintrin.h : Remove macro check for __AMX_TILE__.
14661 2021-09-02 Martin Sebor <msebor@redhat.com>
14663 PR tree-optimization/17506
14665 * tree-ssa-uninit.c (warn_uninit): Remove conditional guarding note.
14667 2021-09-02 Richard Biener <rguenther@suse.de>
14669 * tree-ssa-loop-im.c (fill_always_executed_in_1): Refine
14670 fix for PR78185 and continue processing when leaving
14671 finite inner loops.
14673 2021-09-02 Jakub Jelinek <jakub@redhat.com>
14675 PR tree-optimization/99591
14676 * match.pd: Demote operands of IFN_{ADD,SUB,MUL}_OVERFLOW if they
14679 2021-09-02 Richard Biener <rguenther@suse.de>
14682 2021-09-02 Richard Biener <rguenther@suse.de>
14684 PR tree-optimization/102155
14685 * tree-ssa-loop-im.c (fill_always_executed_in_1): Iterate
14686 over a part of the RPO array and do not recurse here.
14687 Dump blocks marked as always executed.
14688 (fill_always_executed_in): Walk over the RPO array and
14689 process loops whose header we run into.
14690 (loop_invariant_motion_in_fun): Compute the first RPO
14691 using rev_post_order_and_mark_dfs_back_seme in iteration
14692 order and pass that to fill_always_executed_in.
14694 2021-09-02 liuhongt <hongtao.liu@intel.com>
14696 * config/i386/i386-modes.def (FLOAT_MODE): Define ieee HFmode.
14697 * config/i386/i386.c (enum x86_64_reg_class): Add
14698 X86_64_SSEHF_CLASS.
14699 (merge_classes): Handle X86_64_SSEHF_CLASS.
14700 (examine_argument): Ditto.
14701 (construct_container): Ditto.
14702 (classify_argument): Ditto, and set HFmode/HCmode to
14703 X86_64_SSEHF_CLASS.
14704 (function_value_32): Return _FLoat16/Complex Float16 by
14706 (function_value_64): Return _Float16/Complex Float16 by SSE
14708 (ix86_print_operand): Handle CONST_DOUBLE HFmode.
14709 (ix86_secondary_reload): Require gpr as intermediate register
14710 to store _Float16 from sse register when sse4 is not
14712 (ix86_libgcc_floating_mode_supported_p): Enable _FLoat16 under
14714 (ix86_scalar_mode_supported_p): Ditto.
14715 (TARGET_LIBGCC_FLOATING_MODE_SUPPORTED_P): Defined.
14716 * config/i386/i386.h (VALID_SSE2_REG_MODE): Add HFmode.
14717 (VALID_INT_MODE_P): Add HFmode and HCmode.
14718 * config/i386/i386.md (*pushhf_rex64): New define_insn.
14720 (*movhf_internal): Ditto.
14721 * doc/extend.texi (Half-Precision Floating Point): Documemt
14724 2021-09-02 Richard Biener <rguenther@suse.de>
14726 PR tree-optimization/102155
14727 * tree-ssa-loop-im.c (fill_always_executed_in_1): Iterate
14728 over a part of the RPO array and do not recurse here.
14729 Dump blocks marked as always executed.
14730 (fill_always_executed_in): Walk over the RPO array and
14731 process loops whose header we run into.
14732 (loop_invariant_motion_in_fun): Compute the first RPO
14733 using rev_post_order_and_mark_dfs_back_seme in iteration
14734 order and pass that to fill_always_executed_in.
14736 2021-09-02 YunQiang Su <syq@debian.org>
14739 2021-08-31 YunQiang Su <yunqiang.su@cipunited.com>
14741 * config/mips/mips.c (mips_module_isa_name): New.
14742 mips_file_start: add .module mipsREV to all asm output
14744 2021-09-01 Jeff Law <jlaw@localhost.localdomain>
14746 PR tree-optimization/102152
14747 * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Reduce a vector
14748 comparison to a scalar comparison before calling
14749 update_stmt_if_modified.
14751 2021-09-01 Andrew Pinski <apinski@marvell.com>
14754 * config/aarch64/aarch64.c (aarch64_expand_setmem):
14755 Check STRICT_ALIGNMENT before creating an overlapping
14758 2021-09-01 Martin Sebor <msebor@redhat.com>
14760 * gimple-ssa-warn-access.cc (get_size_range): Add argument.
14761 (check_access): Pass additional argument.
14762 (check_memop_access): Remove template and make a member function.
14763 (maybe_check_dealloc_call): Make a pass_waccess member function.
14764 (class pass_waccess): Add, rename, and remove members.
14765 (pass_waccess::pass_waccess): Adjust to name change.
14766 (pass_waccess::~pass_waccess): Same.
14767 (check_alloca): Make a member function.
14768 (check_alloc_size_call): Same.
14769 (check_strcat): Same.
14770 (check_strncat): Same.
14771 (check_stxcpy): Same.
14772 (check_stxncpy): Same.
14773 (check_strncmp): Same.
14774 (maybe_warn_rdwr_sizes): Rename...
14775 (pass_waccess::maybe_check_access_sizes): ...to this.
14776 (pass_waccess::check_call): Adjust to name changes.
14777 (pass_waccess::maybe_check_dealloc_call): Make a pass_waccess member
14779 (pass_waccess::execute): Adjust to name changes.
14780 * gimple-ssa-warn-access.h (check_memop_access): Remove.
14781 * pointer-query.cc (access_ref::phi): Handle null pointer.
14782 (access_ref::inform_access): Same.
14783 (pointer_query::put_ref): Modify a cached value, not a copy of it.
14784 (pointer_query::dump): New function.
14785 (compute_objsize_r): Avoid overwriting access_ref::bndrng. Cache
14787 * pointer-query.h (pointer_query::dump): Declare.
14788 * tree-ssa-strlen.c (get_range): Simplify. Use function query.
14789 (dump_strlen_info): Use function query.
14790 (printf_strlen_execute): Factor code out into pointer_query::put_ref.
14792 2021-09-01 Thomas Schwinge <thomas@codesourcery.com>
14794 * tree.c (walk_tree_1) <OMP_CLAUSE>: Simplify.
14796 2021-09-01 Iain Sandoe <iain@sandoe.co.uk>
14798 * doc/extend.texi: Document unavailable attribute.
14799 * print-tree.c (print_node): Handle unavailable attribute.
14800 * tree-core.h (struct tree_base): Add a bit to carry unavailability.
14801 * tree.c (error_unavailable_use): New.
14802 * tree.h (TREE_UNAVAILABLE): New.
14803 (error_unavailable_use): New.
14805 2021-09-01 Jakub Jelinek <jakub@redhat.com>
14807 PR tree-optimization/102124
14808 * tree-vect-patterns.c (vect_recog_widen_op_pattern): For ORIG_CODE
14809 MINUS_EXPR, if itype is unsigned with smaller precision than type,
14810 add an extra cast to signed variant of itype to ensure sign-extension.
14812 2021-09-01 Martin Liska <mliska@suse.cz>
14814 * graph.c (draw_cfg_node_succ_edges): Do not color fallthru
14815 edges and rather use colors for TRUE and FALSE edges.
14817 2021-09-01 Richard Biener <rguenther@suse.de>
14819 PR tree-optimization/93491
14820 * tree-ssa-pre.c (compute_avail): Set BB_MAY_NOTRETURN
14821 after processing the stmt itself. Do not consider
14822 pure functions possibly not returning. Properly avoid
14823 adding possibly trapping calls to EXP_GEN when there's
14824 a preceeding possibly not returning call.
14825 * tree-ssa-sccvn.c (vn_reference_may_trap): Conservatively
14828 2021-09-01 Richard Biener <rguenther@suse.de>
14830 PR tree-optimization/102139
14831 * tree-vectorizer.h (vec_base_alignments): Adjust hash-map
14832 type to record a std::pair of the stmt-info and the innermost
14834 (dr_vec_info::group): New member.
14835 * tree-vect-data-refs.c (vect_record_base_alignment): Adjust.
14836 (vect_compute_data_ref_alignment): Verify the recorded
14837 base alignment can be used.
14838 (data_ref_pair): Remove.
14839 (dr_group_sort_cmp): Adjust.
14840 (vect_analyze_data_ref_accesses): Store the group-ID in the
14841 dr_vec_info and operate on a vector of dr_vec_infos.
14843 2021-09-01 YunQiang Su <yunqiang.su@cipunited.com>
14845 * read-md.c (md_reader::handle_enum): support value assignation.
14846 * doc/md.texi: record define_c_enum value assignation support.
14848 2021-09-01 Jakub Jelinek <jakub@redhat.com>
14850 PR tree-optimization/102141
14851 * gimple-ssa-store-merging.c (bswap_view_convert): Add BEFORE
14852 argument. If false, emit stmts after gsi instead of before, and
14854 (bswap_replace): Adjust callers. When converting output of bswap,
14855 emit VIEW_CONVERT prepratation stmts after a copy of gsi instead
14858 2021-09-01 liuhongt <hongtao.liu@intel.com>
14860 * emit-rtl.c (validate_subreg): Get rid of all float-int
14863 2021-09-01 liuhongt <hongtao.liu@intel.com>
14866 2021-08-30 liuhongt <hongtao.liu@intel.com>
14868 * expmed.c (extract_bit_field_1): Make sure we're playing with
14869 integral modes before call extract_integral_bit_field.
14870 (extract_integral_bit_field): Add a parameter of type
14871 scalar_int_mode which corresponds to of tmode.
14872 And call extract_and_convert_fixed_bit_field instead of
14873 extract_fixed_bit_field and convert_extracted_bit_field.
14874 (extract_and_convert_fixed_bit_field): New function, it's a
14875 combination of extract_fixed_bit_field and
14876 convert_extracted_bit_field.
14878 2021-08-31 Thomas Schwinge <thomas@codesourcery.com>
14880 * tree.c (walk_tree_1) <OMP_CLAUSE_TILE>: Handle three operands.
14882 2021-08-31 Thomas Schwinge <thomas@codesourcery.com>
14884 * omp-general.h (omp_is_reference): Rename to...
14885 (omp_privatize_by_reference): ... this. Adjust all users...
14886 * omp-general.c: ... here, ...
14887 * gimplify.c: ... here, ...
14888 * omp-expand.c: ... here, ...
14889 * omp-low.c: ... here.
14891 2021-08-31 Martin Sebor <msebor@redhat.com>
14893 * gimple-ssa-warn-access.cc (maybe_warn_alloc_args_overflow): Test
14894 pointer element for equality to zero, not that of the cotaining
14897 2021-08-31 Martin Sebor <msebor@redhat.com>
14899 * gcc-rich-location.h (gcc_rich_location): Make ctor explicit.
14901 2021-08-31 Martin Sebor <msebor@redhat.com>
14903 * function.h (function): Add comments.
14904 (get_range_query): Same. Add attribute returns nonnull.
14906 2021-08-31 Roger Sayle <roger@nextmovesoftware.com>
14908 * expr.c (convert_modes): Don't use subreg_promoted_mode on a
14909 SUBREG if it can't be guaranteed to a SUBREG_PROMOTED_VAR_P set.
14910 Instead use the standard (safer) is_a <scalar_int_mode> idiom.
14912 2021-08-31 Jeff Law <jlaw@localhost.localdomain>
14914 * config.gcc (cris-*-elf, cris-*-none): Remove dbxelf.h from
14916 (m32r-*-elf, m32rle-*-elf, m32r-*-linux): Likewise.
14917 (mn10300-*-*, am33_2.0-*-linux*): Likewise.
14918 (xtensa*-*-elf, xtensa*-*-linux, xtensa*-*-uclinux): Likewise.
14919 (m32c-*-elf*, m32c-*-rtems*): Likewise.
14920 * config/cris/cris.h (DBX_NO_XREFS): Remove.
14921 (DBX_CONTIN_LENGTH, DBX_CONTIN_CHAR): Likewise.
14922 * config/m32r/m32r.h (DBXOUT_SOURCE_LINE): Likewise.
14923 (DBX_DEBUGGING_INFO, DBX_CONTIN_LENGTH): Likewise.
14924 * config/mn10300/mn10300.h (DEFAULT_GDB_EXTENSIONS): Likewise.
14925 * config/mn10300/linux.h (DBX_REGISTER_NAMES): Likewise.
14927 2021-08-31 Marcel Vollweiler <marcel@codesourcery.com>
14929 * gimplify.c (gimplify_scan_omp_clauses): Error handling. 'ancestor' only
14930 allowed on target constructs and only with particular other clauses.
14931 * omp-expand.c (expand_omp_target): Output of 'sorry, not supported' if
14932 'ancestor' is used.
14933 * omp-low.c (check_omp_nesting_restrictions): Error handling. No nested OpenMP
14934 structs when 'ancestor' is used.
14935 (scan_omp_1_stmt): No usage of OpenMP runtime routines in a target region when
14936 'ancestor' is used.
14937 * tree-pretty-print.c (dump_omp_clause): Append 'ancestor'.
14938 * tree.h (OMP_CLAUSE_DEVICE_ANCESTOR): Define macro.
14940 2021-08-31 Roger Sayle <roger@nextmovesoftware.com>
14942 * expr.c (convert_modes): Preserve SUBREG_PROMOTED_VAR_P when
14943 creating a (wider) partial subreg from a SUBREG_PROMOTED_VAR_P
14945 * simplify-rtx.c (simplify_unary_operation_1) [SIGN_EXTEND]:
14946 Likewise, preserve SUBREG_PROMOTED_VAR_P when creating a (wider)
14947 partial subreg from a SUBREG_PROMOTED_VAR_P subreg. Generate
14948 SIGN_EXTEND of the SUBREG_REG when a subreg would be paradoxical.
14949 [ZERO_EXTEND]: Likewise, preserve SUBREG_PROMOTED_VAR_P when
14950 creating a (wider) partial subreg from a SUBREG_PROMOTED_VAR_P
14951 subreg. Generate ZERO_EXTEND of the SUBREG_REG when a subreg
14952 would be paradoxical.
14954 2021-08-31 Roger Sayle <roger@nextmovesoftware.com>
14956 * combine.c (combine_simplify_rtx): Avoid converting an explicit
14957 TRUNCATE into a lowpart SUBREG on !TRULY_NOOP_TRUNCATION targets.
14958 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
14960 2021-08-31 Richard Biener <rguenther@suse.de>
14962 PR tree-optimization/102142
14963 * tree-vect-slp.c (vect_bb_vectorization_profitable_p): Fix
14964 condition under which to unset the visited flag.
14966 2021-08-31 Richard Biener <rguenther@suse.de>
14968 PR middle-end/102129
14969 * tree-ssa-ter.c (find_replaceable_in_bb): Do not move
14970 possibly trapping expressions across calls.
14972 2021-08-31 Jakub Jelinek <jakub@redhat.com>
14974 PR tree-optimization/102134
14975 * tree-ssa-ccp.c (bit_value_binop) <case RSHIFT_EXPR>: If sgn is
14976 UNSIGNED and r1val | r1mask has MSB set, ensure lzcount doesn't
14979 2021-08-31 Andrew Pinski <apinski@marvell.com>
14982 * collect-utils.c (setup_signals): New declaration.
14983 * collect-utils.h (setup_signals): New function.
14984 * collect2.c (handler): Delete.
14985 (main): Instead of manually setting up the signals,
14986 just call setup_signals.
14987 * lto-wrapper.c (main): Likewise.
14989 2021-08-31 Andrew Pinski <apinski@marvell.com>
14992 * config/i386/i386-protos.h (x86_output_aligned_bss):
14993 Change align argument to unsigned type.
14994 (x86_elf_aligned_decl_common): Likewise.
14995 * config/i386/i386.c (x86_elf_aligned_decl_common): Likewise.
14996 (x86_output_aligned_bss): Likewise.
14998 2021-08-31 YunQiang Su <yunqiang.su@cipunited.com>
15000 * config/mips/mips.c (mips_module_isa_name): New.
15001 mips_file_start: add .module mipsREV to all asm output
15003 2021-08-31 YunQiang Su <yunqiang.su@cipunited.com>
15005 * config/mips/mips.h (struct mips_cpu_info): define enum mips_isa;
15006 use enum instead of int for 'isa' member.
15007 * config.gcc, config/mips/mips.c, config/mips/mips-cpus.def,
15008 config/mips/netbsd.h: replace hardcoded numbers with enum.
15010 2021-08-31 liuhongt <hongtao.liu@intel.com>
15012 * config/i386/sse.md (*<avx512>_ucmp<mode>3_1): Change from
15013 define_split to define_insn_and_split.
15014 (*avx2_eq<mode>3): Removed.
15015 (<avx512>_eq<mode>3<mask_scalar_merge_name>): Adjust pattern
15016 (<avx512>_eq<mode>3<mask_scalar_merge_name>_1): Rename to ..
15017 (*<avx512>_eq<mode>3<mask_scalar_merge_name>_1): .. this, and
15019 (*avx2_gt<mode>3): Removed.
15020 (<avx512>_gt<mode>3<mask_scalar_merge_name>): Change from
15021 define_insn to define_expand, and adjust pattern.
15022 (UNSPEC_MASKED_EQ, UNSPEC_MASKED_GT): Removed.
15024 2021-08-30 David Malcolm <dmalcolm@redhat.com>
15027 * Makefile.in (ANALYZER_OBJS): Add analyzer/call-info.o.
15029 2021-08-30 Jason Merrill <jason@redhat.com>
15031 * doc/invoke.texi: Document -Wmissing-requires.
15033 2021-08-30 Bill Schmidt <wschmidt@linux.ibm.com>
15035 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Remove
15036 TARGET_EXTRA_BUILTINS guard.
15038 2021-08-30 Bill Schmidt <wschmidt@linux.ibm.com>
15040 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Change
15041 initialization of V2DI_type_node and unsigned_V2DI_type_node.
15043 2021-08-30 Bill Schmidt <wschmidt@linux.ibm.com>
15045 * config/rs6000/darwin.h (SUBTARGET_INIT_BUILTINS): Use the new
15046 decl when new_builtins_are_live.
15047 * config/rs6000/rs6000-builtin-new.def (__builtin_cfstring): New
15050 2021-08-30 Pat Haugen <pthaugen@linux.ibm.com>
15052 * config/rs6000/rs6000-cpus.def (ISA_3_1_MASKS_SERVER): Add
15053 OPTION_MASK_P10_FUSION_2STORE.
15054 (POWERPC_MASKS): Likewise.
15055 * config/rs6000/rs6000.c (rs6000_option_override_internal): Enable
15056 store fusion for Power10.
15057 (is_fusable_store): New.
15058 (power10_sched_reorder): Likewise.
15059 (rs6000_sched_reorder): Do Power10 specific reordering.
15060 (rs6000_sched_reorder2): Likewise.
15061 * config/rs6000/rs6000.opt: Add new option.
15063 2021-08-30 Richard Biener <rguenther@suse.de>
15065 PR tree-optimization/102128
15066 * tree-vect-slp.c (vect_bb_vectorization_profitable_p):
15067 Move scanning for if-converted scalar code to the caller
15068 and instead delay clearing the visited flag for profitable
15070 (vect_slp_region): Cost all subgraphs before scheduling.
15071 For if-converted BB vectorization scan for scalar COND_EXPRs
15072 and do not vectorize if any found and the cost model is
15075 2021-08-30 Richard Biener <rguenther@suse.de>
15077 * common.opt (fexceptions): Mark
15078 EnabledBy(fnon-call-exceptions).
15079 * doc/invoke.texi (fnon-call-exceptions): Document this
15080 enables -fexceptions.
15082 2021-08-30 Sebastian Huber <sebastian.huber@embedded-brains.de>
15084 * tsystem.h (abort): Define abort() if inhibit_libc is defined and it
15085 is not already defined.
15087 2021-08-30 liuhongt <hongtao.liu@intel.com>
15089 * expmed.c (extract_bit_field_1): Make sure we're playing with
15090 integral modes before call extract_integral_bit_field.
15091 (extract_integral_bit_field): Add a parameter of type
15092 scalar_int_mode which corresponds to of tmode.
15093 And call extract_and_convert_fixed_bit_field instead of
15094 extract_fixed_bit_field and convert_extracted_bit_field.
15095 (extract_and_convert_fixed_bit_field): New function, it's a
15096 combination of extract_fixed_bit_field and
15097 convert_extracted_bit_field.
15099 2021-08-29 Iain Sandoe <iain@sandoe.co.uk>
15101 * config/darwin.c (darwin_libc_has_function): Do not run
15102 the checks for x86 or modern Darwin. Make sure that there
15103 is a value set for darwin_macosx_version_min before testing.
15105 2021-08-29 Iain Sandoe <iain@sandoe.co.uk>
15107 * config/i386/darwin.h (CLEAR_INSN_CACHE): New.
15109 2021-08-28 Jan Hubicka <hubicka@ucw.cz>
15111 * ipa-modref-tree.h (modref_access_node::merge): Break out
15112 logic combining offsets and logic merging ranges to ...
15113 (modref_access_node::combined_offsets): ... here
15114 (modref_access_node::update2): ... here
15115 (modref_access_node::closer_pair_p): New member function.
15116 (modref_access_node::forced_merge): New member function.
15117 (modre_ref_node::insert): Do merging when table is full.
15119 2021-08-28 YunQiang Su <yunqiang.su@cipunited.com>
15122 * config.gcc: MIPS: use N64 ABI by default if the triple end
15123 with -gnuabi64, which is used by Debian since 2013.
15125 2021-08-28 Alexandre Oliva <oliva@adacore.com>
15127 * ipa-modref.c (analyze_function): Skip debug stmts.
15128 * tree-inline.c (estimate_num_insn): Consider builtins even
15129 without a cgraph_node.
15131 2021-08-27 Jeff Law <jlaw@localhost.localdomain>
15133 * config/h8300/bitfield.md (cstore<mode>4): Remove expander.
15134 * config/h8300/h8300.c (h8300_expand_branch): Remove function.
15135 * config/h8300/h8300-protos.h (h8300_expadn_branch): Remove prototype.
15136 * config/h8300/h8300.md (eqne): New code iterator.
15137 (geultu, geultu_to_c): Similarly.
15138 * config/h8300/testcompare.md (cstore<mode>4): Dummy expander.
15139 (store_c_<mode>, store_c_i_<mode>): New define_insn_and_splits
15140 (cmp<mode>_c): New pattern
15142 2021-08-27 Jeff Law <jlaw@localhost.localdomain>
15144 * tree-ssa-dom.c (reduce_vector_comparison_to_scalar_comparison): New
15146 (dom_opt_dom_walker::optimize_stmt): Use it.
15148 2021-08-27 Iain Sandoe <iain@sandoe.co.uk>
15150 * config/darwin.c (finalize_ctors): Add a section-start linker-
15152 (finalize_dtors): Likewise.
15153 * config/darwin.h (MIN_LD64_INIT_TERM_START_LABELS): New.
15155 2021-08-27 Bill Schmidt <wschmidt@linux.ibm.com>
15157 * config/rs6000/rs6000-call.c (rs6000-builtins.h): New #include.
15158 (rs6000_init_builtins): Call rs6000_init_generated_builtins. Skip the
15159 old initialization logic when new builtins are enabled.
15160 * config/rs6000/rs6000-gen-builtins.c (write_decls): Rename
15161 rs6000_autoinit_builtins to rs6000_init_generated_builtins.
15162 (write_init_file): Likewise.
15164 2021-08-27 Iain Sandoe <iain@sandoe.co.uk>
15166 * configure.ac (darwin2[[0-9]]* | darwin19*): Alter use of
15167 gcc_GAS_CHECK_FEATURE to remove an extraneous parameter.
15168 (amdgcn-* | gcn-*) Likewise.
15170 2021-08-27 Anthony Sharp <anthonysharp15@gmail.com>
15172 * symbol-summary.h: Added missing template keyword.
15174 2021-08-27 Richard Biener <rguenther@suse.de>
15176 PR tree-optimization/45178
15177 * tree-ssa-dce.c (find_obviously_necessary_stmts): For
15178 infinite loops without exit do not mark control dependent
15179 edges of the latch necessary.
15181 2021-08-27 konglin1 <lingling.kong@intel.com>
15184 * config/i386/sse.md: (<avx512>scattersi<mode>): Add mask operand to
15186 (<avx512>scattersi<mode>): Likewise.
15187 (*avx512f_scattersi<VI48F:mode>): Merge mask operand to set_dest.
15188 (*avx512f_scatterdi<VI48F:mode>): Likewise
15190 2021-08-27 Kewen Lin <linkw@linux.ibm.com>
15192 * config/rs6000/rs6000.c (rs6000_builtin_md_vectorized_function): Add
15193 support for built-in functions MISC_BUILTIN_DIVWE, MISC_BUILTIN_DIVWEU,
15194 MISC_BUILTIN_DIVDE, MISC_BUILTIN_DIVDEU, P10_BUILTIN_CFUGED,
15195 P10_BUILTIN_CNTLZDM, P10_BUILTIN_CNTTZDM, P10_BUILTIN_PDEPD and
15196 P10_BUILTIN_PEXTD on Power10.
15198 2021-08-27 Kewen Lin <linkw@linux.ibm.com>
15200 * config/rs6000/rs6000-call.c (builtin_function_type): Add unsigned
15201 signedness for some Power10 bifs.
15203 2021-08-27 David Edelsohn <dje.gcc@gmail.com>
15206 * config/rs6000/rs6000.c (rs6000_adjust_field_align): Use
15207 computed alignment if the entire struct has attribute packed.
15209 2021-08-27 liuhongt <hongtao.liu@intel.com>
15213 * config/i386/i386.c (ix86_gimple_fold_builtin): Fold
15214 IX86_BUILTIN_SHUFPD512, IX86_BUILTIN_SHUFPS512,
15215 IX86_BUILTIN_SHUFPD256, IX86_BUILTIN_SHUFPS,
15216 IX86_BUILTIN_SHUFPS256.
15217 (ix86_masked_all_ones): New function.
15219 2021-08-26 Uroš Bizjak <ubizjak@gmail.com>
15221 * config/i386/i386.md (*btr<mode>_1): Call force_reg unconditionally.
15222 (conditional moves with memory inputs splitters): Ditto.
15223 * config/i386/sse.md (one_cmpl<mode>2): Simplify.
15225 2021-08-26 Jan Hubicka <hubicka@ucw.cz>
15227 * ipa-modref-tree.h (modref_access_node::try_merge_with): Restart
15228 search after merging.
15230 2021-08-26 Bill Schmidt <wschmidt@linux.ibm.com>
15232 * config/rs6000/rs6000-overload.def: Add remaining overloads.
15234 2021-08-26 Bill Schmidt <wschmidt@linux.ibm.com>
15236 * config/rs6000/rs6000-builtin-new.def: Add cell stanza.
15238 2021-08-26 Bill Schmidt <wschmidt@linux.ibm.com>
15240 * config/rs6000/rs6000-builtin-new.def: Add ieee128-hw, dfp,
15241 crypto, and htm stanzas.
15243 2021-08-26 Bill Schmidt <wschmidt@linux.ibm.com>
15245 * config/rs6000/rs6000-builtin-new.def: Add mma stanza.
15247 2021-08-26 Martin Sebor <msebor@redhat.com>
15249 * tree-ssa-uninit.c (warn_uninit): Refactor and simplify.
15250 (warn_uninit_phi_uses): Remove argument from calls to warn_uninit.
15251 (warn_uninitialized_vars): Same. Reduce visibility of locals.
15252 (warn_uninitialized_phi): Same.
15254 2021-08-26 Roger Sayle <roger@nextmovesoftware.com>
15256 * tree-ssa-ccp.c (get_individual_bits): Helper function to
15257 extract the individual bits from a widest_int constant (mask).
15258 (gray_code_bit_flips): New read-only table for effiently
15259 enumerating permutations/combinations of bits.
15260 (bit_value_binop) [LROTATE_EXPR, RROTATE_EXPR]: Handle rotates
15261 by unknown counts that are guaranteed less than the target
15262 precision and four or fewer unknown bits by enumeration.
15263 [LSHIFT_EXPR, RSHIFT_EXPR]: Likewise, also handle shifts by
15264 enumeration under the same conditions. Handle remaining
15265 shifts as a mask based upon the minimum possible shift value.
15267 2021-08-26 Roger Sayle <roger@nextmovesoftware.com>
15268 Richard Biener <rguenther@suse.de>
15270 * match.pd (shift transformations): Remove a redundant
15271 !POINTER_TYPE_P check.
15273 2021-08-26 Uroš Bizjak <ubizjak@gmail.com>
15276 * config/i386/i386.md (cmove reg-reg move elimination peephole2s):
15277 Set all_regs to true in the call to replace_rtx.
15279 2021-08-26 Jan Hubicka <hubicka@ucw.cz>
15281 * ipa-modref-tree.c (test_insert_search_collapse): Update test.
15282 * ipa-modref-tree.h (modref_base_node::insert): Be smarter when
15283 hiting --param modref-max-refs limit.
15284 (modref_tree:insert_base): Be smarter when hitting
15285 --param modref-max-bases limit. Add new parameter REF.
15286 (modref_tree:insert): Update.
15287 (modref_tree:merge): Update.
15288 * ipa-modref.c (read_modref_records): Update.
15290 2021-08-26 Jan Hubicka <hubicka@ucw.cz>
15292 * params.opt: (modref-max-adjustments): Add full stop.
15294 2021-08-26 Jan Hubicka <hubicka@ucw.cz>
15296 * ipa-modref-tree.h (modref_ref_node::verify): New member
15298 (modref_ref_node::insert): Use it.
15299 (modref_ref_node::try_mere_with): Fix off by one error.
15301 2021-08-26 Martin Liska <mliska@suse.cz>
15302 Stefan Kneifel <stefan.kneifel@bluewin.ch>
15304 * cgraph.h (create_version_clone_with_body): Add new parameter.
15305 * cgraphclones.c: Likewise.
15306 * multiple_target.c (create_dispatcher_calls): Do not use
15308 (create_target_clone): Likewise here.
15310 2021-08-26 Jonathan Yong <10walls@gmail.com>
15312 * doc/extend.texi: Add note about reserved priorities
15313 to the constructor attribute.
15315 2021-08-25 Martin Sebor <msebor@redhat.com>
15317 * gimple-range-cache.cc (ssa_global_cache::dump): Avoid printing
15318 range table header alone.
15319 * gimple-range.cc (gimple_ranger::export_global_ranges): Same.
15321 2021-08-25 Jan Hubicka <hubicka@ucw.cz>
15323 * doc/invoke.texi: Document --param modref-max-adjustments.
15324 * ipa-modref-tree.c (test_insert_search_collapse): Update.
15325 (test_merge): Update.
15326 * ipa-modref-tree.h (struct modref_access_node): Add adjustments;
15327 (modref_access_node::operator==): Fix handling of access ranges.
15328 (modref_access_node::contains): Constify parameter; handle also
15329 mismatched parm offsets.
15330 (modref_access_node::update): New function.
15331 (modref_access_node::merge): New function.
15332 (unspecified_modref_access_node): Update constructor.
15333 (modref_ref_node::insert_access): Add record_adjustments parameter;
15335 (modref_ref_node::try_merge_with): New private function.
15336 (modref_tree::insert): New record_adjustments parameter.
15337 (modref_tree::merge): New record_adjustments parameter.
15338 (modref_tree::copy_from): Update.
15339 * ipa-modref.c (dump_access): Dump adjustments field.
15340 (get_access): Update constructor.
15341 (record_access): Update call of insert.
15342 (record_access_lto): Update call of insert.
15343 (merge_call_side_effects): Add record_adjustments parameter.
15344 (get_access_for_fnspec): Update.
15345 (process_fnspec): Update.
15346 (analyze_call): Update.
15347 (analyze_function): Update.
15348 (read_modref_records): Update.
15349 (ipa_merge_modref_summary_after_inlining): Update.
15350 (propagate_unknown_call): Update.
15351 (modref_propagate_in_scc): Update.
15352 * params.opt (param-max-modref-adjustments=): New.
15354 2021-08-25 Michael Meissner <meissner@linux.ibm.com>
15356 * config/rs6000/vsx.md (UNSPEC_XXSPLTIDP): Rename from
15358 (xxspltiw_v4si): Use vecperm type attribute.
15359 (xxspltiw_v4si_inst): Use vecperm type attribute.
15360 (xxspltiw_v4sf_inst): Likewise.
15361 (xxspltidp_v2df): Use vecperm type attribute. Use
15362 UNSPEC_XXSPLTIDP instead of UNSPEC_XXSPLTID.
15363 (xxspltidp_v2df_inst): Likewise.
15364 (xxsplti32dx_v4si): Use vecperm type attribute.
15365 (xxsplti32dx_v4si_inst): Likewise.
15366 (xxsplti32dx_v4sf_inst): Likewise.
15367 (xxblend_<mode>): Likewise.
15368 (xxpermx): Likewise.
15369 (xxpermx_inst): Likewise.
15370 (xxeval): Likewise.
15372 2021-08-25 Lewis Hyatt <lhyatt@gmail.com>
15375 * coretypes.h (typedef diagnostic_input_charset_callback): Declare.
15376 * diagnostic.c (diagnostic_initialize_input_context): New function.
15377 * diagnostic.h (diagnostic_initialize_input_context): Declare.
15378 * input.c (default_charset_callback): New function.
15379 (file_cache::initialize_input_context): New function.
15380 (file_cache_slot::create): Added ability to convert the input
15381 according to the input context.
15382 (file_cache::file_cache): Initialize the new input context.
15383 (class file_cache_slot): Added new m_alloc_offset member.
15384 (file_cache_slot::file_cache_slot): Initialize the new member.
15385 (file_cache_slot::~file_cache_slot): Handle potentially offset buffer.
15386 (file_cache_slot::maybe_grow): Likewise.
15387 (file_cache_slot::needs_read_p): Handle NULL fp, which is now possible.
15388 (file_cache_slot::get_next_line): Likewise.
15389 * input.h (class file_cache): Added input context member.
15391 2021-08-25 Richard Biener <rguenther@suse.de>
15393 PR tree-optimization/102046
15394 * tree-vect-slp.c (vect_build_slp_tree_2): Conservatively
15395 update ->any_pattern when swapping operands.
15397 2021-08-25 Hongyu Wang <hongyu.wang@intel.com>
15400 * config/i386/i386.c (ix86_live_on_entry): Adjust comment.
15401 (ix86_decompose_address): Remove retval check for ASHIFT,
15402 allow non-canonical zero extend if AND mask covers ASHIFT
15404 (ix86_legitimate_address_p): Adjust condition for decompose.
15405 (ix86_rtx_costs): Adjust cost for lea with non-canonical
15407 Co-Authored by: Uros Bizjak <ubizjak@gmail.com>
15409 2021-08-25 Jiufu Guo <guojiufu@linux.ibm.com>
15411 PR tree-optimization/101145
15412 * tree-ssa-loop-niter.c (number_of_iterations_until_wrap):
15414 (number_of_iterations_lt): Invoke above function.
15415 (adjust_cond_for_loop_until_wrap):
15416 Merge to number_of_iterations_until_wrap.
15417 (number_of_iterations_cond): Update invokes for
15418 adjust_cond_for_loop_until_wrap and number_of_iterations_lt.
15420 2021-08-25 konglin1 <lingling.kong@intel.com>
15423 * config/i386/avx512dqintrin.h (_mm512_fpclass_ps_mask): Fix
15424 macro define in O0.
15425 (_mm512_mask_fpclass_ps_mask): Ditto.
15427 2021-08-25 Kewen Lin <linkw@linux.ibm.com>
15429 * config/rs6000/altivec.md (vec_unpacku_hi_v16qi): Remove.
15430 (vec_unpacku_hi_v8hi): Likewise.
15431 (vec_unpacku_lo_v16qi): Likewise.
15432 (vec_unpacku_lo_v8hi): Likewise.
15433 (vec_unpacku_hi_<VP_small_lc>): New define_expand.
15434 (vec_unpacku_lo_<VP_small_lc>): Likewise.
15436 2021-08-24 David Edelsohn <dje.gcc@gmail.com>
15438 * config/rs6000/aix.h (SYSTEM_IMPLICIT_EXTERN_C): Delete.
15439 * config/rs6000/aix71.h (SYSTEM_IMPLICIT_EXTERN_C): Define.
15440 * config/rs6000/aix72.h (SYSTEM_IMPLICIT_EXTERN_C): Define.
15441 * config/rs6000/aix73.h (TARGET_AIX_VERSION): Increase to 73.
15443 2021-08-24 Roger Sayle <roger@nextmovesoftware.com>
15445 PR middle-end/102031
15446 * simplify-rtx.c (simplify_truncation): When comparing precisions
15447 use "subreg_prec" variable, not "subreg_mode".
15449 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
15451 * config/rs6000/rs6000-builtin-new.def: Add power10 and power10-64
15454 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
15456 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Initialize
15457 various pointer type nodes.
15458 * config/rs6000/rs6000.h (rs6000_builtin_type_index): Add enum
15459 values for various pointer types.
15460 (ptr_V16QI_type_node): New macro.
15461 (ptr_V1TI_type_node): New macro.
15462 (ptr_V2DI_type_node): New macro.
15463 (ptr_V2DF_type_node): New macro.
15464 (ptr_V4SI_type_node): New macro.
15465 (ptr_V4SF_type_node): New macro.
15466 (ptr_V8HI_type_node): New macro.
15467 (ptr_unsigned_V16QI_type_node): New macro.
15468 (ptr_unsigned_V1TI_type_node): New macro.
15469 (ptr_unsigned_V8HI_type_node): New macro.
15470 (ptr_unsigned_V4SI_type_node): New macro.
15471 (ptr_unsigned_V2DI_type_node): New macro.
15472 (ptr_bool_V16QI_type_node): New macro.
15473 (ptr_bool_V8HI_type_node): New macro.
15474 (ptr_bool_V4SI_type_node): New macro.
15475 (ptr_bool_V2DI_type_node): New macro.
15476 (ptr_bool_V1TI_type_node): New macro.
15477 (ptr_pixel_type_node): New macro.
15478 (ptr_intQI_type_node): New macro.
15479 (ptr_uintQI_type_node): New macro.
15480 (ptr_intHI_type_node): New macro.
15481 (ptr_uintHI_type_node): New macro.
15482 (ptr_intSI_type_node): New macro.
15483 (ptr_uintSI_type_node): New macro.
15484 (ptr_intDI_type_node): New macro.
15485 (ptr_uintDI_type_node): New macro.
15486 (ptr_intTI_type_node): New macro.
15487 (ptr_uintTI_type_node): New macro.
15488 (ptr_long_integer_type_node): New macro.
15489 (ptr_long_unsigned_type_node): New macro.
15490 (ptr_float_type_node): New macro.
15491 (ptr_double_type_node): New macro.
15492 (ptr_long_double_type_node): New macro.
15493 (ptr_dfloat64_type_node): New macro.
15494 (ptr_dfloat128_type_node): New macro.
15495 (ptr_ieee128_type_node): New macro.
15496 (ptr_ibm128_type_node): New macro.
15497 (ptr_vector_pair_type_node): New macro.
15498 (ptr_vector_quad_type_node): New macro.
15499 (ptr_long_long_integer_type_node): New macro.
15500 (ptr_long_long_unsigned_type_node): New macro.
15502 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
15504 * config/rs6000/rs6000-builtin-new.def: Add power9-vector, power9,
15505 and power9-64 stanzas.
15507 2021-08-24 Roger Sayle <roger@nextmovesoftware.com>
15508 Tom de Vries <tdevries@suse.de>
15510 * config.gcc (nvptx-*-*): Define {c,c++}_target_objs.
15511 * config/nvptx/nvptx-protos.h (nvptx_cpu_cpp_builtins): Prototype.
15512 * config/nvptx/nvptx.h (TARGET_CPU_CPP_BUILTINS): Implement with
15513 a call to the new nvptx_cpu_cpp_builtins function in nvptx-c.c.
15514 * config/nvptx/t-nvptx (nvptx-c.o): New rule.
15515 * config/nvptx/nvptx-c.c: New source file.
15516 (nvptx_cpu_cpp_builtins): Move implementation here.
15518 2021-08-24 Martin Sebor <msebor@redhat.com>
15520 PR middle-end/101600
15521 PR middle-end/101977
15522 * gimple-ssa-warn-access.cc (maybe_warn_for_bound): Tighten up
15523 the phrasing of a warning.
15524 (check_access): Use the remaining size after subtracting any offset
15525 rather than the whole object size.
15526 * pointer-query.cc (access_ref::get_ref): Clear BASE0 flag if it's
15527 clear for any nonnull PHI argument.
15528 (compute_objsize): Clear argument.
15530 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
15532 * config/rs6000/rs6000-builtin-new.def: Add power8-vector stanza.
15534 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
15536 * config/rs6000/rs6000-builtin-new.def: Add power7 and power7-64
15539 2021-08-24 Andrew MacLeod <amacleod@redhat.com>
15541 * value-relation.cc (rr_transitive_table): New.
15542 (relation_transitive): New.
15543 (value_relation::swap): Remove.
15544 (value_relation::apply_transitive): New.
15545 (relation_oracle::relation_oracle): Allocate a new tmp bitmap.
15546 (relation_oracle::register_relation): Call register_transitives.
15547 (relation_oracle::register_transitives): New.
15548 * value-relation.h (relation_oracle): Add new temporary bitmap and
15551 2021-08-24 H.J. Lu <hjl.tools@gmail.com>
15554 * config/i386/i386-expand.c (ix86_expand_vector_move): Broadcast
15555 from integer to a pseudo vector register.
15557 2021-08-24 Richard Biener <rguenther@suse.de>
15559 PR tree-optimization/100089
15560 * tree-vectorizer.h (vect_slp_bb): Rename to ...
15561 (vect_slp_if_converted_bb): ... this and get the original
15562 loop as new argument.
15563 * tree-vectorizer.c (try_vectorize_loop_1): Revert previous fix,
15564 pass original loop to vect_slp_if_converted_bb.
15565 * tree-vect-slp.c (vect_bb_vectorization_profitable_p):
15566 If orig_loop was passed scan the not vectorized stmts
15567 for COND_EXPRs and force not profitable if found.
15568 (vect_slp_region): Pass down all SLP instances to costing
15569 if orig_loop was specified.
15570 (vect_slp_bbs): Pass through orig_loop.
15571 (vect_slp_bb): Rename to ...
15572 (vect_slp_if_converted_bb): ... this and get the original
15573 loop as new argument.
15574 (vect_slp_function): Adjust.
15576 2021-08-24 Richard Earnshaw <rearnsha@arm.com>
15579 * config/arm/arm.md (attribute arch): Add fix_vlldm.
15580 (arch_enabled): Use it.
15581 * config/arm/vfp.md (lazy_store_multiple_insn): Add alternative to
15582 use when erratum mitigation is needed.
15584 2021-08-24 Richard Earnshaw <rearnsha@arm.com>
15587 * config/arm/arm.opt (mfix-cmse-cve-2021-35465): New option.
15588 * doc/invoke.texi (Arm Options): Document it.
15589 * config/arm/arm-cpus.in (quirk_vlldm): New feature bit.
15590 (ALL_QUIRKS): Add quirk_vlldm.
15591 (cortex-m33): Add quirk_vlldm.
15592 (cortex-m35p, cortex-m55): Likewise.
15593 * config/arm/arm.c (arm_option_override): Enable fix_vlldm if
15594 targetting an affected CPU and not explicitly controlled on
15597 2021-08-24 Richard Earnshaw <rearnsha@arm.com>
15599 * config/arm/vfp.md (lazy_store_multiple_insn): Rewrite as valid RTL.
15600 (lazy_load_multiple_insn): Likewise.
15602 2021-08-24 liuhongt <hongtao.liu@intel.com>
15605 * config/i386/sse.md (<avx512>_vternlog<mode><sd_maskz_name>):
15606 Enable avx512 embedded broadcast.
15607 (*<avx512>_vternlog<mode>_all): Ditto.
15608 (<avx512>_vternlog<mode>_mask): Ditto.
15610 2021-08-24 liuhongt <hongtao.liu@intel.com>
15613 * config/i386/i386.c (ix86_rtx_costs): Define cost for
15615 * config/i386/i386.h (STRIP_UNARY): New macro.
15616 * config/i386/predicates.md (reg_or_notreg_operand): New
15618 * config/i386/sse.md (*<avx512>_vternlog<mode>_all): New define_insn.
15619 (*<avx512>_vternlog<mode>_1): New pre_reload
15620 define_insn_and_split.
15621 (*<avx512>_vternlog<mode>_2): Ditto.
15622 (*<avx512>_vternlog<mode>_3): Ditto.
15623 (any_logic1,any_logic2): New code iterator.
15624 (logic_op): New code attribute.
15625 (ternlogsuffix): Extend to VNxDF and VNxSF.
15627 2021-08-24 Richard Biener <rguenther@suse.de>
15629 * doc/invoke.texi (vect-inner-loop-cost-factor): Adjust.
15630 * params.opt (--param vect-inner-loop-cost-factor): Adjust
15632 * tree-vect-loop.c (vect_analyze_loop_form): Initialize
15633 inner_loop_cost_factor to the minimum of the estimated number
15634 of iterations of the inner loop and vect-inner-loop-cost-factor.
15636 2021-08-24 Roger Sayle <roger@nextmovesoftware.com>
15637 Richard Biener <rguenther@suse.de>
15639 * config/i386/i386-features.c (compute_convert_gain): Provide
15640 more accurate values for CONST_INT, when optimizing for size.
15641 * config/i386/i386.c (COSTS_N_BYTES): Move definition from here...
15642 * config/i386/i386.h (COSTS_N_BYTES): to here.
15644 2021-08-24 Roger Sayle <roger@nextmovesoftware.com>
15645 Jakub Jelinek <jakub@redhat.com>
15647 PR middle-end/102029
15648 * match.pd (shift transformations): Add an additional check for
15649 !POINTER_TYPE_P in the recently added left shift transformation.
15651 2021-08-24 liuhongt <hongtao.liu@intel.com>
15653 PR tree-optimization/100089
15654 * tree-vectorizer.c (try_vectorize_loop_1): Disable slp in
15655 loop vectorizer when cost model is very-cheap.
15657 2021-08-23 Bill Schmidt <wschmidt@linux.ibm.com>
15659 * config/rs6000/rs6000-gen-builtins.c (parse_bif_entry): Don't call
15660 asprintf, which is not available on AIX.
15662 2021-08-23 Bill Schmidt <wschmidt@linux.ibm.com>
15664 * config.gcc (target_gtfiles): Add ./rs6000-builtins.h.
15665 * config/rs6000/t-rs6000 (EXTRA_GTYPE_DEPS): Set.
15667 2021-08-23 Bill Schmidt <wschmidt@linux.ibm.com>
15669 * config.gcc (powerpc*-*-*): Add rs6000-builtins.o to extra_objs.
15670 * config/rs6000/rs6000-gen-builtins.c (main): Close init_file
15672 * config/rs6000/t-rs6000 (rs6000-gen-builtins.o): New target.
15673 (rbtree.o): Likewise.
15674 (rs6000-gen-builtins): Likewise.
15675 (rs6000-builtins.c): Likewise.
15676 (rs6000-builtins.h): Likewise.
15677 (rs6000.o): Add dependency.
15678 (EXTRA_HEADERS): Add rs6000-vecdefines.h.
15679 (rs6000-vecdefines.h): New target.
15680 (rs6000-builtins.o): Likewise.
15681 (rs6000-call.o): Add rs6000-builtins.h as a dependency.
15682 (rs6000-c.o): Likewise.
15684 2021-08-23 Bill Schmidt <wschmidt@linux.ibm.com>
15687 * config/rs6000/rs6000-gen-builtins.c (consume_whitespace):
15688 Diagnose buffer overrun.
15689 (safe_inc_pos): Fix overrun detection.
15690 (match_identifier): Diagnose buffer overrun.
15691 (match_integer): Likewise.
15692 (match_to_right_bracket): Likewise.
15694 2021-08-23 Jan Hubicka <hubicka@ucw.cz>
15696 * ipa-modref-tree.h (modref_access_node::range_info_useful_p):
15697 Improve range compare.
15698 (modref_access_node::contains): New member function.
15699 (modref_access_node::search): Remove.
15700 (modref_access_node::insert): Be smarter about subaccesses.
15702 2021-08-23 Thomas Schwinge <thomas@codesourcery.com>
15704 * config/i386/i386-options.c (ix86_omp_device_kind_arch_isa)
15705 <omp_device_arch> [ACCEL_COMPILER]: Match "intel_mic".
15706 * config/i386/t-omp-device (omp-device-properties-i386) <arch>:
15709 2021-08-23 Jeff Law <jlaw@localhost.localdomain>
15711 * config/h8300/h8300-protos.h (h8300_expand_epilogue): Add new
15713 * config/h8300/jumpcall.md (call, call_value): Restrict to
15714 !SIBLING_CALL_P cases.
15715 (subcall, sibcall_value): New patterns & expanders.
15716 * config/h8300/proepi.md (epilogue): Pass new argument to
15717 h8300_expand_epilogue.
15718 (sibcall_epilogue): New expander.
15719 * config/h8300/h8300.c (h8300_expand_epilogue): Handle sibcall
15721 (h8300_ok_for_sibcall_p): New function.
15722 (TARGET_FUNCTION_OK_FOR_SIBCALL): define.
15724 2021-08-23 Roger Sayle <roger@nextmovesoftware.com>
15726 * simplify-rtx.c (simplify_unary_operation_1): [TRUNCATE]:
15727 Handle case where the operand is already the desired mode.
15729 2021-08-23 Richard Biener <rguenther@suse.de>
15732 * tree-ssa-structalias.c (ipa_pta_execute): Check in_other_partition
15733 in addition to has_gimple_body.
15735 2021-08-23 Jan Hubicka <hubicka@ucw.cz>
15737 PR middle-end/101949
15738 * ipa-modref.c (analyze_ssa_name_flags): Fix merging of
15741 2021-08-23 Martin Liska <mliska@suse.cz>
15743 * doc/invoke.texi: Put the option out of -mxl-mode-app-model
15746 2021-08-23 Richard Biener <rguenther@suse.de>
15748 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
15749 Properly scale the inner loop cost only once.
15751 2021-08-23 Roger Sayle <roger@nextmovesoftware.com>
15753 * tree-ssa-ccp.c (bit_value_binop) [TRUNC_MOD_EXPR, TRUNC_DIV_EXPR]:
15754 Provide bounds for unsigned (and signed with non-negative operands)
15755 division and modulus.
15757 2021-08-23 Roger Sayle <roger@nextmovesoftware.com>
15759 * simplify-rtx.c (simplify_truncation): Generalize simplification
15760 of (truncate:A (subreg:B X)).
15761 (simplify_unary_operation_1) [FLOAT_TRUNCATE, FLOAT_EXTEND,
15762 SIGN_EXTEND, ZERO_EXTEND]: Handle cases where the operand
15763 already has the desired machine mode.
15764 (test_scalar_int_ops): Add tests that useless extensions and
15765 truncations are optimized away.
15766 (test_scalar_int_ext_ops): New self-test function to confirm
15767 that truncations of extensions are correctly simplified.
15768 (test_scalar_int_ext_ops2): New self-test function to check
15769 truncations of truncations, extensions of extensions, and
15770 truncations of extensions.
15771 (test_scalar_ops): Call the above two functions with a
15772 representative sampling of integer machine modes.
15774 2021-08-23 Roger Sayle <roger@nextmovesoftware.com>
15776 * match.pd (shift transformations): Change the sign of an
15777 LSHIFT_EXPR if it reduces the number of explicit conversions.
15779 2021-08-23 Jakub Jelinek <jakub@redhat.com>
15781 PR tree-optimization/86723
15782 * gimple-ssa-store-merging.c (find_bswap_or_nop_finalize): Add
15783 cast64_to_32 argument, set *cast64_to_32 to false, unless n is
15784 non-memory permutation of 64-bit src which only has bytes of
15785 0 or [5..8] and n->range is 4.
15786 (find_bswap_or_nop): Add cast64_to_32 and mask arguments, adjust
15787 find_bswap_or_nop_finalize caller, support bswap with some bytes
15788 zeroed, as long as at least two bytes are not zeroed.
15789 (bswap_replace): Add mask argument and handle masking of bswap
15791 (maybe_optimize_vector_constructor): Adjust find_bswap_or_nop
15792 caller, punt if cast64_to_32 or mask is not all ones.
15793 (pass_optimize_bswap::execute): Adjust find_bswap_or_nop_finalize
15794 caller, for now punt if cast64_to_32.
15796 2021-08-23 Richard Biener <rguenther@suse.de>
15798 PR tree-optimization/79334
15799 * tree-ssa-sccvn.c (copy_reference_ops_from_ref): Record
15800 a type also for COMPONENT_REFs.
15801 (vn_reference_may_trap): Check ARRAY_REF with constant index
15802 against the array domain.
15804 2021-08-23 liuhongt <hongtao.liu@intel.com>
15807 * config/i386/sse.md (*avx512f_pshufb_truncv8hiv8qi_1): Add
15808 TARGET_AVX512BW to condition.
15810 2021-08-23 Jakub Jelinek <jakub@redhat.com>
15813 * dwarf2out.c (gen_variable_die): Add DW_AT_location for global
15814 register variables already during early_dwarf if possible.
15816 2021-08-23 Christophe Lyon <christophe.lyon@foss.st.com>
15818 * config/arm/arm_mve.h: Fix __arm_vctp16q return type.
15820 2021-08-23 Christophe Lyon <christophe.lyon@foss.st.com>
15823 * config/arm/arm.opt: Fix typo.
15824 * config/arm/t-rmprofile: Fix typo.
15826 2021-08-23 Jakub Jelinek <jakub@redhat.com>
15828 * tree.h (OMP_CLAUSE_GRAINSIZE_STRICT): Define.
15829 (OMP_CLAUSE_NUM_TASKS_STRICT): Define.
15830 * tree-pretty-print.c (dump_omp_clause) <case OMP_CLAUSE_GRAINSIZE,
15831 case OMP_CLAUSE_NUM_TASKS>: Print strict: modifier.
15832 * omp-expand.c (expand_task_call): Use GOMP_TASK_FLAG_STRICT in iflags
15833 if either grainsize or num_tasks clause has the strict modifier.
15835 2021-08-23 Martin Liska <mliska@suse.cz>
15837 * dbgcnt.def (DEBUG_COUNTER): New counter.
15838 * gimple.c (gimple_call_arg_flags): Use it in IPA PTA.
15840 2021-08-23 Jan Hubicka <hubicka@ucw.cz>
15842 * ipa-modref.c (analyze_ssa_name_flags): Improve handling of return slot.
15844 2021-08-23 Xi Ruoyao <xry111@mengyan1223.wang>
15847 * config/mips/mips-protos.h (mips_msa_output_shift_immediate):
15849 * config/mips/mips.c (mips_msa_output_shift_immediate): New
15851 * config/mips/mips-msa.md (vashl<mode>3, vashr<mode>3,
15852 vlshr<mode>3): Call it.
15854 2021-08-22 Jan Hubicka <hubicka@ucw.cz>
15855 Martin Liska <mliska@suse.cz>
15857 PR middle-end/101949
15858 * ipa-modref.c (analyze_ssa_name_flags): Indirect call implies
15861 2021-08-21 Dragan Mladjenovic <OT_Dragan.Mladjenovic@mediatek.com>
15863 * config/mips/mips.c (mips_function_rodata_section,
15864 TARGET_ASM_FUNCTION_RODATA_SECTION): Removed.
15866 2021-08-21 John David Anglin <danglin@gcc.gnu.org>
15868 * config/pa/pa.c (pa_asm_output_aligned_common): Remove warning.
15870 2021-08-20 Serge Belyshev <belyshev@depni.sinp.msu.ru>
15872 * configure.ac (thread-local storage support): Remove tls_first_major
15873 and tls_first_minor. Use "$conftest_s" to check support.
15874 * configure: Regenerate.
15876 2021-08-20 Serge Belyshev <belyshev@depni.sinp.msu.ru>
15878 * configure.ac: Fixup formatting.
15880 2021-08-20 Serge Belyshev <belyshev@depni.sinp.msu.ru>
15882 * acinclude.m4 (gcc_GAS_CHECK_FEATURE): Remove third argument and ...
15883 * configure.ac: ... update all callers.
15885 2021-08-20 Serge Belyshev <belyshev@depni.sinp.msu.ru>
15888 * acinclude.m4 (_gcc_COMPUTE_GAS_VERSION, _gcc_GAS_VERSION_GTE_IFELSE)
15889 (gcc_GAS_VERSION_GTE_IFELSE): Remove.
15890 (gcc_GAS_CHECK_FEATURE): Do not handle in-tree case specially.
15891 * configure.ac: Remove gcc_cv_gas_major_version, gcc_cv_gas_minor_version.
15892 Remove remaining checks for in-tree assembler.
15893 * configure: Regenerate.
15895 2021-08-20 Jeff Law <jlaw@localhost.localdomain>
15897 * config/h8300/h8300.c (shift_alg_hi): Improve arithmetic shift right
15898 by 15 bits for H8/300H and H8/S. Improve logical shifts by 12
15900 (shift_alg_si): Improve arithmetic right shift by 28-30 bits for
15901 H8/300H. Improve arithmetic shift right by 15 bits for H8/S.
15902 Improve logical shifts by 27 bits for H8/S.
15903 (get_shift_alg): Corresponding changes.
15904 (h8300_option_override): Revert to loops for -Os when profitable.
15906 2021-08-20 Richard Biener <rguenther@suse.de>
15908 * tree-vect-data-refs.c (dr_group_sort_cmp): Do not compare
15910 (vect_analyze_data_ref_accesses): Likewise. Assign the BB
15911 index as group_id when dataref_groups were not computed.
15912 * tree-vect-slp.c (vect_slp_bbs): Bump current_group when
15913 we advace to the next BB.
15915 2021-08-20 Jakub Jelinek <jakub@redhat.com>
15917 * omp-builtins.def (BUILT_IN_GOMP_WARNING, BUILT_IN_GOMP_ERROR): New
15920 2021-08-20 Martin Liska <mliska@suse.cz>
15922 PR gcov-profile/89961
15923 * gcov.c (make_gcov_file_name): Rewrite using std::string.
15924 (mangle_name): Simplify, do not used the second argument.
15925 (strip_extention): New function.
15926 (get_md5sum): Likewise.
15927 (get_gcov_intermediate_filename): Handle properly -p and -x
15929 (output_gcov_file): Use string type.
15930 (generate_results): Likewise.
15931 (md5sum_to_hex): Remove.
15933 2021-08-20 Michael Meissner <meissner@linux.ibm.com>
15935 * config/rs6000/altivec.md (UNSPEC_XXEVAL): Move to vsx.md.
15936 (UNSPEC_XXSPLTIW): Move to vsx.md.
15937 (UNSPEC_XXSPLTID): Move to vsx.md.
15938 (UNSPEC_XXSPLTI32DX): Move to vsx.md.
15939 (UNSPEC_XXBLEND): Move to vsx.md.
15940 (UNSPEC_XXPERMX): Move to vsx.md.
15941 (VM3): Move to vsx.md.
15942 (VM3_char): Move to vsx.md.
15943 (xxspltiw_v4si): Move to vsx.md.
15944 (xxspltiw_v4sf): Move to vsx.md.
15945 (xxspltiw_v4sf_inst): Move to vsx.md.
15946 (xxspltidp_v2df): Move to vsx.md.
15947 (xxspltidp_v2df_inst): Move to vsx.md.
15948 (xxsplti32dx_v4si_inst): Move to vsx.md.
15949 (xxsplti32dx_v4sf): Move to vsx.md.
15950 (xxsplti32dx_v4sf_inst): Move to vsx.md.
15951 (xxblend_<mode>): Move to vsx.md.
15952 (xxpermx): Move to vsx.md.
15953 (xxpermx_inst): Move to vsx.md.
15954 * config/rs6000/vsx.md (UNSPEC_XXEVAL): Move from altivec.md.
15955 (UNSPEC_XXSPLTIW): Move from altivec.md.
15956 (UNSPEC_XXSPLTID): Move from altivec.md.
15957 (UNSPEC_XXSPLTI32DX): Move from altivec.md.
15958 (UNSPEC_XXBLEND): Move from altivec.md.
15959 (UNSPEC_XXPERMX): Move from altivec.md.
15960 (VM3): Move from altivec.md.
15961 (VM3_char): Move from altivec.md.
15962 (xxspltiw_v4si): Move from altivec.md.
15963 (xxspltiw_v4sf): Move from altivec.md.
15964 (xxspltiw_v4sf_inst): Move from altivec.md.
15965 (xxspltidp_v2df): Move from altivec.md.
15966 (xxspltidp_v2df_inst): Move from altivec.md.
15967 (xxsplti32dx_v4si_inst): Move from altivec.md.
15968 (xxsplti32dx_v4sf): Move from altivec.md.
15969 (xxsplti32dx_v4sf_inst): Move from altivec.md.
15970 (xxblend_<mode>): Move from altivec.md.
15971 (xxpermx): Move from altivec.md.
15972 (xxpermx_inst): Move from altivec.md.
15974 2021-08-19 Roger Sayle <roger@nextmovesoftware.com>
15976 * tree-vect-generic.c (expand_vector_operations_1): Use either
15977 gimplify_build1 or gimplify_build2 instead of gimple_build_assign
15978 when constructing scalar splat expressions.
15980 2021-08-19 Peter Bergner <bergner@linux.ibm.com>
15983 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_mma_builtin): Cast
15984 pointer to __vector_pair *.
15986 2021-08-19 Martin Sebor <msebor@redhat.com>
15988 * gimple-range.cc: Add comments.
15989 * gimple-range.h: Same.
15991 2021-08-19 Martin Sebor <msebor@redhat.com>
15993 PR middle-end/101984
15994 * gimple-ssa-warn-access.cc (pass_waccess::execute): Also call
15997 2021-08-19 Jeff Law <jlaw@localhost.localdomain>
15999 * config.gcc (h8300-*-elf*): Do not include dbxelf.h.
16000 (h8300-*-linux*, v850-*-rtems*, v850*-elf*): Likewise.
16001 * config/v850/v850.h (DEFAULT_GDB_EXTENSIONS): Remove.
16003 2021-08-19 Jakub Jelinek <jakub@redhat.com>
16005 PR middle-end/101950
16006 * optabs.c (expand_clrsb_using_clz): New function.
16007 (expand_unop): Use it as another clrsb expansion fallback.
16009 2021-08-19 liuhongt <hongtao.liu@intel.com>
16012 2021-07-28 liuhongt <hongtao.liu@intel.com>
16015 * config/i386/i386.h (processor_costs): Add new member
16017 * config/i386/x86-tune-costs.h (ix86_size_cost, i386_cost,
16018 i486_cost, pentium_cost, lakemont_cost, pentiumpro_cost,
16019 geode_cost, k6_cost, athlon_cost, k8_cost, amdfam10_cost,
16020 bdver_cost, znver1_cost, znver2_cost, znver3_cost,
16021 btver1_cost, btver2_cost, btver3_cost, pentium4_cost,
16022 nocona_cost, atom_cost, atom_cost, slm_cost, intel_cost,
16023 generic_cost, core_cost): Initialize integer_to_sse same value
16025 (skylake_cost): Initialize integer_to_sse twice as much as sse_op.
16026 * config/i386/i386.c (ix86_builtin_vectorization_cost):
16027 Use integer_to_sse instead of sse_op to calculate the cost of
16030 2021-08-18 Iain Sandoe <iain@sandoe.co.uk>
16032 * config.gcc: Include rpath.opt for Darwin.
16033 * config/darwin.h (DRIVER_SELF_SPECS): Handle -rpath.
16035 2021-08-18 Thomas Schwinge <thomas@codesourcery.com>
16037 PR bootstrap/101959
16038 * hash-map-tests.c (test_map_of_type_with_ctor_and_dtor_expand):
16041 2021-08-18 Jonathan Wright <jonathan.wright@arm.com>
16043 * config/aarch64/arm_neon.h (vld3_lane_f64): Use float RTL
16044 pattern and type cast.
16045 (vld4_lane_f32): Use float RTL pattern.
16046 (vld4q_lane_f64): Use float type cast.
16048 2021-08-18 Jan Hubicka <hubicka@ucw.cz>
16050 * tree-ssa-uninit.c (maybe_warn_pass_by_reference): Check also
16053 2021-08-18 Thomas Schwinge <thomas@codesourcery.com>
16055 * hash-map-tests.c (test_map_of_type_with_ctor_and_dtor): Extend.
16056 (test_map_of_type_with_ctor_and_dtor_expand): Add function.
16057 (hash_map_tests_c_tests): Call it.
16059 2021-08-18 Thomas Schwinge <thomas@codesourcery.com>
16061 * ggc.h (enum ggc_collect): New.
16062 (ggc_collect): Use it.
16063 * ggc-page.c: Adjust.
16064 * ggc-common.c: Likewise.
16065 * ggc-tests.c: Likewise.
16066 * read-rtl-function.c: Likewise.
16067 * selftest-run-tests.c: Likewise.
16068 * doc/gty.texi (Invoking the garbage collector): Likewise.
16070 2021-08-18 liuhongt <hongtao.liu@intel.com>
16073 * config/i386/i386.h (TARGET_V2DF_REDUCTION_PREFER_HADDPD):
16075 * config/i386/sse.md (*sse3_haddv2df3_low): Add
16076 TARGET_V2DF_REDUCTION_PREFER_HADDPD.
16077 (*sse3_hsubv2df3_low): Ditto.
16078 * config/i386/x86-tune.def
16079 (X86_TUNE_V2DF_REDUCTION_PREFER_HADDPD): New tune.
16081 2021-08-17 Andrew MacLeod <amacleod@redhat.com>
16083 * gimple-range-gori.cc (gori_compute::gori_compute): Enable tracing.
16084 (gori_compute::compute_operand_range): Add tracing.
16085 (gori_compute::logical_combine): Ditto.
16086 (gori_compute::compute_logical_operands): Ditto.
16087 (gori_compute::compute_operand1_range): Ditto.
16088 (gori_compute::compute_operand2_range): Ditto.
16089 (gori_compute::outgoing_edge_range_p): Ditto.
16090 * gimple-range-gori.h (class gori_compute): Add range_tracer.
16092 2021-08-17 Andrew MacLeod <amacleod@redhat.com>
16094 * flag-types.h (enum evrp_mode): Adjust evrp-mode values.
16095 * gimple-range-cache.cc (DEBUG_RANGE_CACHE): Relocate from.
16096 * gimple-range-trace.h (DEBUG_RANGE_CACHE): Here.
16097 * params.opt (--param=evrp-mode): Adjust options.
16099 2021-08-17 Andrew MacLeod <amacleod@redhat.com>
16101 * Makefile.in (OBJS): Add gimple-range-trace.o.
16102 * gimple-range-cache.h (enable_new_values): Remove unused prototype.
16103 * gimple-range-fold.cc: Adjust headers.
16104 * gimple-range-trace.cc: New.
16105 * gimple-range-trace.h: New.
16106 * gimple-range.cc (gimple_ranger::gimple_ranger): Enable tracer.
16107 (gimple_ranger::range_of_expr): Add tracing.
16108 (gimple_ranger::range_on_entry): Ditto.
16109 (gimple_ranger::range_on_exit): Ditto.
16110 (gimple_ranger::range_on_edge): Ditto.
16111 (gimple_ranger::fold_range_internal): Ditto.
16112 (gimple_ranger::dump_bb): Do not calculate edge range twice.
16113 (trace_ranger::*): Remove.
16114 (enable_ranger): Never create a trace_ranger.
16115 (debug_seed_ranger): Move to gimple-range-trace.cc.
16116 (dump_ranger): Ditto.
16117 (debug_ranger): Ditto.
16118 * gimple-range.h: Include gimple-range-trace.h.
16119 (range_on_entry, range_on_exit): No longer virtual.
16120 (class trace_ranger): Remove.
16121 (DEBUG_RANGE_CACHE): Move to gimple-range-trace.h.
16123 2021-08-17 Martin Sebor <msebor@redhat.com>
16125 PR middle-end/101854
16126 * builtins.c (expand_builtin_alloca): Move warning code to check_alloca
16127 in gimple-ssa-warn-access.cc.
16128 * calls.c (alloc_max_size): Move code to check_alloca.
16129 (get_size_range): Move to pointer-query.cc.
16130 (maybe_warn_alloc_args_overflow): Move to gimple-ssa-warn-access.cc.
16131 (get_attr_nonstring_decl): Move to tree.c.
16132 (fntype_argno_type): Move to gimple-ssa-warn-access.cc.
16133 (append_attrname): Same.
16134 (maybe_warn_rdwr_sizes): Same.
16135 (initialize_argument_information): Move code to
16136 gimple-ssa-warn-access.cc.
16137 * calls.h (maybe_warn_alloc_args_overflow): Move to
16138 gimple-ssa-warn-access.h.
16139 (get_attr_nonstring_decl): Move to tree.h.
16140 (maybe_warn_nonstring_arg): Move to gimple-ssa-warn-access.h.
16141 (enum size_range_flags): Move to pointer-query.h.
16142 (get_size_range): Same.
16143 * gimple-ssa-warn-access.cc (has_location): Remove unused overload
16144 to avoid Clang -Wunused-function.
16145 (get_size_range): Declare static.
16146 (maybe_emit_free_warning): Rename...
16147 (maybe_check_dealloc_call): ...to this for consistency.
16148 (class pass_waccess): Add members.
16149 (pass_waccess::~pass_waccess): Defined.
16150 (alloc_max_size): Move here from calls.c.
16151 (maybe_warn_alloc_args_overflow): Same.
16152 (check_alloca): New function.
16153 (check_alloc_size_call): New function.
16154 (check_strncat): Handle another warning flag.
16155 (pass_waccess::check_builtin): Handle alloca.
16156 (fntype_argno_type): Move here from calls.c.
16157 (append_attrname): Same.
16158 (maybe_warn_rdwr_sizes): Same.
16159 (pass_waccess::check_call): Define.
16160 (check_nonstring_args): New function.
16161 (pass_waccess::check): Call new member functions.
16162 (pass_waccess::execute): Enable ranger.
16163 * gimple-ssa-warn-access.h (get_size_range): Move here from calls.h.
16164 (maybe_warn_nonstring_arg): Same.
16165 * gimple-ssa-warn-restrict.c: Remove #include.
16166 * pointer-query.cc (get_size_range): Move here from calls.c.
16167 * pointer-query.h (enum size_range_flags): Same.
16168 (get_size_range): Same.
16169 * tree.c (get_attr_nonstring_decl): Move here from calls.c.
16170 * tree.h (get_attr_nonstring_decl): Move here from calls.h.
16172 2021-08-17 Thomas Schwinge <thomas@codesourcery.com>
16174 * ggc.h (ggc_collect): Add 'force_collect' parameter.
16175 * ggc-page.c (ggc_collect): Use that one instead of global
16176 'ggc_force_collect'. Adjust all users.
16177 * doc/gty.texi (Invoking the garbage collector): Update.
16178 * ggc-internal.h (ggc_force_collect): Remove.
16179 * ggc-common.c (ggc_force_collect): Likewise.
16180 * selftest.h (forcibly_ggc_collect): Remove.
16181 * ggc-tests.c (selftest::forcibly_ggc_collect): Likewise.
16182 * read-rtl-function.c (test_loading_labels): Adjust.
16183 * selftest-run-tests.c (run_tests): Likewise.
16185 2021-08-17 Iain Sandoe <iain@sandoe.co.uk>
16187 * config/darwin.c (darwin_file_end): Reset and reclaim the
16188 section names table at the end of compile.
16190 2021-08-17 Iain Sandoe <iain@sandoe.co.uk>
16193 * config.in: Regenerate.
16194 * config/i386/darwin.h (EXTRA_ASM_OPTS): New
16195 (ASM_SPEC): Pass options to disable branch shortening where
16197 * configure: Regenerate.
16198 * configure.ac: Detect versions of 'as' that support the
16199 optimisation which has the bug.
16201 2021-08-17 Richard Biener <rguenther@suse.de>
16203 * optabs-query.c (supports_vec_gather_load_p): Also check
16205 (supports_vec_scatter_store_p): Likewise.
16206 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Fall
16207 back to masked variants if non-masked are not supported.
16208 * tree-vect-patterns.c (vect_recog_gather_scatter_pattern):
16209 When we need to use masked gather/scatter but do not have
16210 a mask set up a constant true one.
16211 * tree-vect-stmts.c (vect_check_scalar_mask): Also allow
16212 non-SSA_NAME masks.
16214 2021-08-17 Roger Sayle <roger@nextmovesoftware.com>
16216 * tree-ssa-ccp.c (bit_value_binop) [MINUS_EXPR]: Use same
16217 algorithm as PLUS_EXPR to improve subtraction bit bounds.
16218 [POINTER_DIFF_EXPR]: Treat as synonymous with MINUS_EXPR.
16220 2021-08-17 Roger Sayle <roger@nextmovesoftware.com>
16222 * tree-ssa-ccp.c (bit_value_mult_const): New helper function to
16223 calculate the mask-value pair result of a multiplication by an
16225 (bit_value_binop) [MULT_EXPR]: Call it from here for
16226 multiplications by (sparse) non-negative constants.
16228 2021-08-17 Christophe Lyon <christophe.lyon@foss.st.com>
16231 * config.gcc (gcc_cv_initfini_array): Leave undefined for
16232 uclinuxfdpiceabi targets.
16234 2021-08-17 Alexandre Oliva <oliva@adacore.com>
16236 * tree-inline.c (maybe_move_debug_stmts_to_successors): Don't
16237 reverse debug stmts.
16239 2021-08-17 Alexandre Oliva <oliva@adacore.com>
16241 * tree-cfg.c (dump_function_to_file): Use fun, not cfun.
16243 2021-08-17 Jonathan Wright <jonathan.wright@arm.com>
16245 * config/aarch64/arm_neon.h (__LD4_LANE_FUNC): Delete.
16246 (__LD4Q_LANE_FUNC): Likewise.
16247 (vld4_lane_u8): Define without macro.
16248 (vld4_lane_u16): Likewise.
16249 (vld4_lane_u32): Likewise.
16250 (vld4_lane_u64): Likewise.
16251 (vld4_lane_s8): Likewise.
16252 (vld4_lane_s16): Likewise.
16253 (vld4_lane_s32): Likewise.
16254 (vld4_lane_s64): Likewise.
16255 (vld4_lane_f16): Likewise.
16256 (vld4_lane_f32): Likewise.
16257 (vld4_lane_f64): Likewise.
16258 (vld4_lane_p8): Likewise.
16259 (vld4_lane_p16): Likewise.
16260 (vld4_lane_p64): Likewise.
16261 (vld4q_lane_u8): Likewise.
16262 (vld4q_lane_u16): Likewise.
16263 (vld4q_lane_u32): Likewise.
16264 (vld4q_lane_u64): Likewise.
16265 (vld4q_lane_s8): Likewise.
16266 (vld4q_lane_s16): Likewise.
16267 (vld4q_lane_s32): Likewise.
16268 (vld4q_lane_s64): Likewise.
16269 (vld4q_lane_f16): Likewise.
16270 (vld4q_lane_f32): Likewise.
16271 (vld4q_lane_f64): Likewise.
16272 (vld4q_lane_p8): Likewise.
16273 (vld4q_lane_p16): Likewise.
16274 (vld4q_lane_p64): Likewise.
16275 (vld4_lane_bf16): Likewise.
16276 (vld4q_lane_bf16): Likewise.
16278 2021-08-17 Jonathan Wright <jonathan.wright@arm.com>
16280 * config/aarch64/arm_neon.h (__LD3_LANE_FUNC): Delete.
16281 (__LD3Q_LANE_FUNC): Delete.
16282 (vld3_lane_u8): Define without macro.
16283 (vld3_lane_u16): Likewise.
16284 (vld3_lane_u32): Likewise.
16285 (vld3_lane_u64): Likewise.
16286 (vld3_lane_s8): Likewise.
16287 (vld3_lane_s16): Likewise.
16288 (vld3_lane_s32): Likewise.
16289 (vld3_lane_s64): Likewise.
16290 (vld3_lane_f16): Likewise.
16291 (vld3_lane_f32): Likewise.
16292 (vld3_lane_f64): Likewise.
16293 (vld3_lane_p8): Likewise.
16294 (vld3_lane_p16): Likewise.
16295 (vld3_lane_p64): Likewise.
16296 (vld3q_lane_u8): Likewise.
16297 (vld3q_lane_u16): Likewise.
16298 (vld3q_lane_u32): Likewise.
16299 (vld3q_lane_u64): Likewise.
16300 (vld3q_lane_s8): Likewise.
16301 (vld3q_lane_s16): Likewise.
16302 (vld3q_lane_s32): Likewise.
16303 (vld3q_lane_s64): Likewise.
16304 (vld3q_lane_f16): Likewise.
16305 (vld3q_lane_f32): Likewise.
16306 (vld3q_lane_f64): Likewise.
16307 (vld3q_lane_p8): Likewise.
16308 (vld3q_lane_p16): Likewise.
16309 (vld3q_lane_p64): Likewise.
16310 (vld3_lane_bf16): Likewise.
16311 (vld3q_lane_bf16): Likewise.
16313 2021-08-17 Jonathan Wright <jonathan.wright@arm.com>
16315 * config/aarch64/arm_neon.h (__LD2_LANE_FUNC): Delete.
16316 (__LD2Q_LANE_FUNC): Likewise.
16317 (vld2_lane_u8): Define without macro.
16318 (vld2_lane_u16): Likewise.
16319 (vld2_lane_u32): Likewise.
16320 (vld2_lane_u64): Likewise.
16321 (vld2_lane_s8): Likewise.
16322 (vld2_lane_s16): Likewise.
16323 (vld2_lane_s32): Likewise.
16324 (vld2_lane_s64): Likewise.
16325 (vld2_lane_f16): Likewise.
16326 (vld2_lane_f32): Likewise.
16327 (vld2_lane_f64): Likewise.
16328 (vld2_lane_p8): Likewise.
16329 (vld2_lane_p16): Likewise.
16330 (vld2_lane_p64): Likewise.
16331 (vld2q_lane_u8): Likewise.
16332 (vld2q_lane_u16): Likewise.
16333 (vld2q_lane_u32): Likewise.
16334 (vld2q_lane_u64): Likewise.
16335 (vld2q_lane_s8): Likewise.
16336 (vld2q_lane_s16): Likewise.
16337 (vld2q_lane_s32): Likewise.
16338 (vld2q_lane_s64): Likewise.
16339 (vld2q_lane_f16): Likewise.
16340 (vld2q_lane_f32): Likewise.
16341 (vld2q_lane_f64): Likewise.
16342 (vld2q_lane_p8): Likewise.
16343 (vld2q_lane_p16): Likewise.
16344 (vld2q_lane_p64): Likewise.
16345 (vld2_lane_bf16): Likewise.
16346 (vld2q_lane_bf16): Likewise.
16348 2021-08-17 Maxim Kuvyrkov <maxim.kuvyrkov@linaro.org>
16350 * haifa-sched.c (advance_one_cycle): Output more context-synchronization
16353 2021-08-17 Maxim Kuvyrkov <maxim.kuvyrkov@linaro.org>
16355 * haifa-sched.c (enum rfs_decision, rfs_str): Add RFS_AUTOPREF.
16356 (rank_for_schedule): Use it.
16358 2021-08-17 Maxim Kuvyrkov <maxim.kuvyrkov@linaro.org>
16360 PR rtl-optimization/91598
16361 * haifa-sched.c (autopref_rank_for_schedule): Prioritize "irrelevant"
16362 insns after memory reads and before memory writes.
16364 2021-08-17 Alistair_Lee <alistair.lee@arm.com>
16366 * rtl.h (CONST_VECTOR_P): New macro.
16367 * config/aarch64/aarch64.c (aarch64_get_sve_pred_bits): Use RTL
16368 code testing macros.
16369 (aarch64_ptrue_all_mode): Likewise.
16370 (aarch64_expand_mov_immediate): Likewise.
16371 (aarch64_const_vec_all_in_range_p): Likewise.
16372 (aarch64_rtx_costs): Likewise.
16373 (aarch64_legitimate_constant_p): Likewise.
16374 (aarch64_simd_valid_immediate): Likewise.
16375 (aarch64_simd_make_constant): Likewise.
16376 (aarch64_convert_mult_to_shift): Likewise.
16377 (aarch64_expand_sve_vec_perm): Likewise.
16378 (aarch64_vec_fpconst_pow_of_2): Likewise.
16380 2021-08-17 Andrew MacLeod <amacleod@redhat.com>
16382 PR tree-optimization/101938
16383 * range-op.cc (operator_abs::op1_range): Special case
16384 -TYPE_MIN_VALUE for flag_wrapv.
16386 2021-08-17 Kewen Lin <linkw@linux.ibm.com>
16388 * tree-vect-slp.c (vectorizable_bb_reduc_epilogue): Add the cost for
16391 2021-08-17 Jakub Jelinek <jakub@redhat.com>
16393 * tree.def (OMP_SCOPE): New tree code.
16394 * tree.h (OMP_SCOPE_BODY, OMP_SCOPE_CLAUSES): Define.
16395 * tree-nested.c (convert_nonlocal_reference_stmt,
16396 convert_local_reference_stmt, convert_gimple_call): Handle
16398 * tree-pretty-print.c (dump_generic_node): Handle OMP_SCOPE.
16399 * gimple.def (GIMPLE_OMP_SCOPE): New gimple code.
16400 * gimple.c (gimple_build_omp_scope): New function.
16401 (gimple_copy): Handle GIMPLE_OMP_SCOPE.
16402 * gimple.h (gimple_build_omp_scope): Declare.
16403 (gimple_has_substatements): Handle GIMPLE_OMP_SCOPE.
16404 (gimple_omp_scope_clauses, gimple_omp_scope_clauses_ptr,
16405 gimple_omp_scope_set_clauses): New inline functions.
16406 (CASE_GIMPLE_OMP): Add GIMPLE_OMP_SCOPE.
16407 * gimple-pretty-print.c (dump_gimple_omp_scope): New function.
16408 (pp_gimple_stmt_1): Handle GIMPLE_OMP_SCOPE.
16409 * gimple-walk.c (walk_gimple_stmt): Likewise.
16410 * gimple-low.c (lower_stmt): Likewise.
16411 * gimplify.c (is_gimple_stmt): Handle OMP_MASTER.
16412 (gimplify_scan_omp_clauses): For task reductions, handle OMP_SCOPE
16413 like ORT_WORKSHARE constructs. Adjust diagnostics for %<scope%>
16414 allowing task reductions. Reject inscan reductions on scope.
16415 (omp_find_stores_stmt): Handle GIMPLE_OMP_SCOPE.
16416 (gimplify_omp_workshare, gimplify_expr): Handle OMP_SCOPE.
16417 * tree-inline.c (remap_gimple_stmt): Handle GIMPLE_OMP_SCOPE.
16418 (estimate_num_insns): Likewise.
16419 * omp-low.c (build_outer_var_ref): Look through GIMPLE_OMP_SCOPE
16420 contexts if var isn't privatized there.
16421 (check_omp_nesting_restrictions): Handle GIMPLE_OMP_SCOPE.
16422 (scan_omp_1_stmt): Likewise.
16423 (maybe_add_implicit_barrier_cancel): Look through outer
16425 (lower_omp_scope): New function.
16426 (lower_omp_task_reductions): Handle OMP_SCOPE.
16427 (lower_omp_1): Handle GIMPLE_OMP_SCOPE.
16428 (diagnose_sb_1, diagnose_sb_2): Likewise.
16429 * omp-expand.c (expand_omp_single): Support also GIMPLE_OMP_SCOPE.
16430 (expand_omp): Handle GIMPLE_OMP_SCOPE.
16431 (omp_make_gimple_edges): Likewise.
16432 * omp-builtins.def (BUILT_IN_GOMP_SCOPE_START): New built-in.
16434 2021-08-17 Richard Biener <rguenther@suse.de>
16436 PR tree-optimization/101925
16437 * tree-ssa-sccvn.c (copy_reference_ops_from_ref): Set
16438 reverse on COMPONENT_REF and ARRAY_REF according to
16439 what reverse_storage_order_for_component_p does.
16440 (vn_reference_eq): Compare reversed on reference ops.
16441 (reverse_storage_order_for_component_p): New overload.
16442 (vn_reference_lookup_3): Check reverse_storage_order_for_component_p
16443 on the reference looked up.
16445 2021-08-17 Jeff Law <jlaw@localhost.localdomain>
16447 * config/h8300/h8300.c (shift_alg_si): Avoid loops for most SImode
16448 shifts on the H8/S.
16449 (h8300_option_override): Use loops on H8/S more often when optimizing
16451 (get_shift_alg): Handle new "special" cases on H8/S. Simplify
16452 accordingly. Handle various arithmetic right shifts with special
16453 sequences that we couldn't handle before.
16455 2021-08-16 Jeff Law <jlaw@localhost.localdomain>
16457 * config.gcc (rl78-*-elf*): Do not include dbxelf.h.
16459 2021-08-16 Sebastian Huber <sebastian.huber@embedded-brains.de>
16461 * config/sparc/rtemself.h (SPARC_GCOV_TYPE_SIZE): Define.
16462 * config/sparc/sparc.c (sparc_gcov_type_size): New.
16463 (TARGET_GCOV_TYPE_SIZE): Redefine if SPARC_GCOV_TYPE_SIZE is defined.
16464 * coverage.c (get_gcov_type): Use targetm.gcov_type_size().
16465 * doc/tm.texi (TARGET_GCOV_TYPE_SIZE): Add hook under "Misc".
16466 * doc/tm.texi.in: Regenerate.
16467 * target.def (gcov_type_size): New target hook.
16468 * targhooks.c (default_gcov_type_size): New.
16469 * targhooks.h (default_gcov_type_size): Declare.
16470 * tree-profile.c (gimple_gen_edge_profiler): Use precision of
16472 (gimple_gen_time_profiler): Likewise.
16474 2021-08-16 Eric Botcazou <ebotcazou@gcc.gnu.org>
16476 * dwarf2out.c (add_scalar_info): Deal with DW_AT_data_bit_offset.
16478 2021-08-16 Tobias Burnus <tobias@codesourcery.com>
16480 PR middle-end/101931
16481 * omp-low.c (omp_runtime_api_call): Update for routines
16482 added in the meanwhile.
16484 2021-08-16 Martin Liska <mliska@suse.cz>
16486 PR tree-optimization/100393
16487 * tree-switch-conversion.c (group_cluster::dump): Use
16488 get_comparison_count.
16489 (jump_table_cluster::find_jump_tables): Pre-compute number of
16490 comparisons and then decrement it. Cache also max_ratio.
16491 (jump_table_cluster::can_be_handled): Change signature.
16492 * tree-switch-conversion.h (get_comparison_count): New.
16494 2021-08-16 Eric Botcazou <ebotcazou@gcc.gnu.org>
16496 * dwarf2out.c (add_data_member_location_attribute): Use GNAT
16497 encodings only when -fgnat-encodings=all is specified.
16498 (add_bound_info): Likewise.
16499 (add_byte_size_attribute): Likewise.
16500 (gen_member_die): Likewise.
16502 2021-08-16 Thomas Schwinge <thomas@codesourcery.com>
16504 * omp-oacc-neuter-broadcast.cc
16505 (execute_omp_oacc_neuter_broadcast): Plug 'par' memory leak.
16507 2021-08-16 Thomas Schwinge <thomas@codesourcery.com>
16509 * omp-oacc-neuter-broadcast.cc
16510 (execute_omp_oacc_neuter_broadcast): Clarify memory management for
16513 2021-08-16 Thomas Schwinge <thomas@codesourcery.com>
16515 * omp-oacc-neuter-broadcast.cc (field_map): Move variable into...
16516 (execute_omp_oacc_neuter_broadcast): ... here.
16517 (install_var_field, build_receiver_ref, build_sender_ref): Take
16518 'field_map_t *' parameter. Adjust all users.
16519 (worker_single_copy, neuter_worker_single): Take a
16520 'record_field_map_t *' parameter. Adjust all users.
16522 2021-08-16 liuhongt <hongtao.liu@intel.com>
16525 * config/i386/i386.md (ldexp<mode>3): Force operands[1] to
16528 2021-08-16 Martin Liska <mliska@suse.cz>
16531 * multiple_target.c (create_dispatcher_calls): Make default
16532 function local only if it is a definition.
16534 2021-08-16 Martin Liska <mliska@suse.cz>
16537 * ipa-icf-gimple.c (func_checker::compare_ssa_name): Do not
16538 consider equal SSA_NAMEs when one is a param.
16540 2021-08-16 liuhongt <hongtao.liu@intel.com>
16543 * config/i386/i386-expand.c (ix86_expand_vec_perm_vpermt2):
16544 Support vpermi2b for V32QI/V16QImode.
16545 (ix86_extract_perm_from_pool_constant): New function.
16546 (ix86_expand_vec_one_operand_perm_avx512): Support
16547 vpermw/vpermb under TARGET_AVX512BW/TARGET_AVX512VBMI.
16548 (expand_vec_perm_1): Adjust comments for upper.
16549 * config/i386/i386-protos.h (ix86_extract_perm_from_pool_constant):
16551 * config/i386/predicates.md (permvar_truncate_operand): New predicate.
16552 (pshufb_truncv4siv4hi_operand): Ditto.
16553 (pshufb_truncv8hiv8qi_operand): Ditto.
16554 * config/i386/sse.md (*avx512bw_permvar_truncv16siv16hi_1):
16555 New pre_reload define_insn_and_split.
16556 (*avx512f_permvar_truncv8siv8hi_1): Ditto.
16557 (*avx512f_vpermvar_truncv8div8si_1): Ditto.
16558 (*avx512f_permvar_truncv32hiv32qi_1): Ditto.
16559 (*avx512f_permvar_truncv16hiv16qi_1): Ditto.
16560 (*avx512f_permvar_truncv4div4si_1): Ditto.
16561 (*avx512f_pshufb_truncv8hiv8qi_1): Ditto.
16562 (*avx512f_pshufb_truncv4siv4hi_1): Ditto.
16563 (*avx512f_pshufd_truncv2div2si_1): Ditto.
16565 2021-08-16 Kito Cheng <kito.cheng@sifive.com>
16567 * config/riscv/multilib-generator: Support code model option for
16569 * doc/install.texi: Add document of new option for
16570 --with-multilib-generator.
16572 2021-08-15 Clément Chigot <clement.chigot@atos.net>
16574 * config/rs6000/rs6000.c (xcoff_tls_exec_model_detected): New.
16575 (rs6000_legitimize_tls_address_aix): Use it.
16576 (rs6000_xcoff_file_end): Add ".ref __tls_get_addr" when
16577 xcoff_tls_exec_model_detected is true.
16579 2021-08-15 Jeff Law <jlaw@localhost.localdomain>
16581 * config/h8300/h8300.c (shift_alg_si): Retune H8/300H shifts
16582 to allow a bit more code growth, saving many dozens of cycles.
16583 (h8300_option_override): Adjus shift_alg_si if optimizing for
16585 (get_shift_alg): Use special + inline shifts for residuals
16588 2021-08-14 Stafford Horne <shorne@gmail.com>
16591 * config/or1k/or1k-opts.h: New file.
16592 * config/or1k/or1k.c (or1k_legitimize_address_1, print_reloc):
16593 Support generating gotha relocations if -mcmodel=large is
16595 * config/or1k/or1k.h (TARGET_CMODEL_SMALL, TARGET_CMODEL_LARGE):
16597 * config/or1k/or1k.opt (mcmodel=): New option.
16598 * doc/invoke.texi (OpenRISC Options): Document mcmodel.
16600 2021-08-14 Martin Sebor <msebor@redhat.com>
16602 PR middle-end/101791
16603 * gimple-ssa-warn-access.cc (new_delete_mismatch_p): Use new argument
16604 to valid_new_delete_pair_p.
16605 * tree.c (valid_new_delete_pair_p): Add argument.
16606 * tree.h (valid_new_delete_pair_p): Same.
16608 2021-08-14 Jakub Jelinek <jakub@redhat.com>
16611 * config/i386/i386-expand.c (expand_vec_perm_broadcast_1)
16612 <case E_V64QImode>: For this mode assert
16613 !TARGET_AVX512BW || d->perm[0] rather than !TARGET_AVX2 || d->perm[0].
16615 2021-08-13 Michael Meissner <meissner@linux.ibm.com>
16618 * config/rs6000/altivec.md (xxeval): Use register_predicate
16619 instead of altivec_register_predicate.
16621 2021-08-13 Martin Sebor <msebor@redhat.com>
16623 PR middle-end/101734
16624 * tree-ssa-uninit.c (maybe_warn_read_write_only): New function.
16625 (maybe_warn_operand): Call it.
16627 2021-08-13 Martin Liska <mliska@suse.cz>
16630 * attribs.c (decl_attributes): Make naked functions "noipa"
16633 2021-08-13 Martin Liska <mliska@suse.cz>
16636 * symtab.c (symtab_node::noninterposable_alias): Do not create
16637 local aliases for target_clone functions as the clonning pass
16640 2021-08-13 Martin Liska <mliska@suse.cz>
16642 * opts.c (LIVE_PATCHING_OPTION): Define.
16643 (control_options_for_live_patching): Use it in error messages.
16645 2021-08-13 Jan Hubicka <hubicka@ucw.cz>
16647 * ipa-modref.c (dump_eaf_flags): Dump EAF_NOREAD.
16648 (implicit_const_eaf_flags, implicit_pure_eaf_flags,
16649 ignore_stores_eaf_flags): New constants.
16650 (remove_useless_eaf_flags): New function.
16651 (eaf_flags_useful_p): Use it.
16652 (deref_flags): Add EAF_NOT_RETURNED if flag is unused;
16654 (modref_lattice::init): Add EAF_NOREAD.
16655 (modref_lattice::add_escape_point): Do not reacord escape point if
16657 (modref_lattice::merge): EAF_NOESCAPE implies EAF_NODIRECTESCAPE;
16658 use remove_useless_eaf_flags.
16659 (modref_lattice::merge_deref): Use ignore_stores_eaf_flags.
16660 (modref_lattice::merge_direct_load): Add EAF_NOREAD
16661 (analyze_ssa_name_flags): Fix handling EAF_NOT_RETURNED
16662 (analyze_parms): Use remove_useless_eaf_flags.
16663 (ipa_merge_modref_summary_after_inlining): Use ignore_stores_eaf_flags.
16664 (modref_merge_call_site_flags): Add caller and ecf_flags parameter;
16665 use remove_useless_eaf_flags.
16666 (modref_propagate_flags_in_scc): Update.
16667 * ipa-modref.h: Turn eaf_flags_t back to char.
16668 * tree-core.h (EAF_NOT_RETURNED): Fix.
16669 (EAF_NOREAD): New constant
16670 * tree-ssa-alias.c: (ref_maybe_used_by_call_p_1): Check for
16672 * tree-ssa-structalias.c (handle_rhs_call): Handle new flags.
16673 (handle_pure_call): Likewise.
16675 2021-08-12 Jakub Jelinek <jakub@redhat.com>
16677 * tree.def (OMP_MASKED): New tree code.
16678 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_FILTER.
16679 * tree.h (OMP_MASKED_BODY, OMP_MASKED_CLAUSES, OMP_MASKED_COMBINED,
16680 OMP_CLAUSE_FILTER_EXPR): Define.
16681 * tree.c (omp_clause_num_ops): Add OMP_CLAUSE_FILTER entry.
16682 (omp_clause_code_name): Likewise.
16683 (walk_tree_1): Handle OMP_CLAUSE_FILTER.
16684 * tree-nested.c (convert_nonlocal_omp_clauses,
16685 convert_local_omp_clauses): Handle OMP_CLAUSE_FILTER.
16686 (convert_nonlocal_reference_stmt, convert_local_reference_stmt,
16687 convert_gimple_call): Handle GIMPLE_OMP_MASTER.
16688 * tree-pretty-print.c (dump_omp_clause): Handle OMP_CLAUSE_FILTER.
16689 (dump_generic_node): Handle OMP_MASTER.
16690 * gimple.def (GIMPLE_OMP_MASKED): New gimple code.
16691 * gimple.c (gimple_build_omp_masked): New function.
16692 (gimple_copy): Handle GIMPLE_OMP_MASKED.
16693 * gimple.h (gimple_build_omp_masked): Declare.
16694 (gimple_has_substatements): Handle GIMPLE_OMP_MASKED.
16695 (gimple_omp_masked_clauses, gimple_omp_masked_clauses_ptr,
16696 gimple_omp_masked_set_clauses): New inline functions.
16697 (CASE_GIMPLE_OMP): Add GIMPLE_OMP_MASKED.
16698 * gimple-pretty-print.c (dump_gimple_omp_masked): New function.
16699 (pp_gimple_stmt_1): Handle GIMPLE_OMP_MASKED.
16700 * gimple-walk.c (walk_gimple_stmt): Likewise.
16701 * gimple-low.c (lower_stmt): Likewise.
16702 * gimplify.c (is_gimple_stmt): Handle OMP_MASTER.
16703 (gimplify_scan_omp_clauses): Handle OMP_CLAUSE_FILTER. For clauses
16704 that take one expression rather than decl or constant, force
16705 gimplification of that into a SSA_NAME or temporary unless min
16707 (gimplify_adjust_omp_clauses): Handle OMP_CLAUSE_FILTER.
16708 (gimplify_expr): Handle OMP_MASKED.
16709 * tree-inline.c (remap_gimple_stmt): Handle GIMPLE_OMP_MASKED.
16710 (estimate_num_insns): Likewise.
16711 * omp-low.c (scan_sharing_clauses): Handle OMP_CLAUSE_FILTER.
16712 (check_omp_nesting_restrictions): Handle GIMPLE_OMP_MASKED. Adjust
16713 diagnostics for existence of masked construct.
16714 (scan_omp_1_stmt, lower_omp_master, lower_omp_1, diagnose_sb_1,
16715 diagnose_sb_2): Handle GIMPLE_OMP_MASKED.
16716 * omp-expand.c (expand_omp_synch, expand_omp, omp_make_gimple_edges):
16719 2021-08-12 Uroš Bizjak <ubizjak@gmail.com>
16722 * config/i386/i386.md (avx512f_scalef<mode>2): New insn pattern.
16723 (ldexp<mode>3): Use avx512f_scalef<mode>2.
16724 (UNSPEC_SCALEF): Move from ...
16725 * config/i386/sse.md (UNSPEC_SCALEF): ... here.
16727 2021-08-12 Jan Hubicka <hubicka@ucw.cz>
16729 * ipa-split.c (consider_split): Fix condition testing void functions.
16731 2021-08-12 Aldy Hernandez <aldyh@redhat.com>
16733 * doc/invoke.texi: Remove docs for threader-mode param.
16734 * flag-types.h (enum threader_mode): Remove.
16735 * params.opt: Remove threader-mode param.
16736 * tree-ssa-threadbackward.c (class back_threader): Remove
16737 path_is_unreachable_p.
16738 Make find_paths private.
16739 Add maybe_thread and thread_through_all_blocks.
16740 Remove reference marker for m_registry.
16741 Remove reference marker for m_profit.
16742 (back_threader::back_threader): Adjust for registry and profit not
16744 (dump_path): Move down.
16745 (debug): Move down.
16746 (class thread_jumps): Remove.
16747 (class back_threader_registry): Remove m_all_paths.
16749 (thread_jumps::thread_through_all_blocks): Move to back_threader
16751 (fsm_find_thread_path): Remove
16752 (back_threader::maybe_thread): New.
16753 (back_threader::thread_through_all_blocks): Move from
16755 (back_threader_registry::back_threader_registry): Remove
16757 (back_threader_registry::~back_threader_registry): Remove.
16758 (thread_jumps::find_taken_edge): Remove.
16759 (thread_jumps::check_subpath_and_update_thread_path): Remove.
16760 (thread_jumps::maybe_register_path): Remove.
16761 (thread_jumps::handle_phi): Remove.
16762 (handle_assignment_p): Remove.
16763 (thread_jumps::handle_assignment): Remove.
16764 (thread_jumps::fsm_find_control_statement_thread_paths): Remove.
16765 (thread_jumps::find_jump_threads_backwards): Remove.
16766 (thread_jumps::find_jump_threads_backwards_with_ranger): Remove.
16767 (try_thread_blocks): Rename find_jump_threads_backwards to
16769 (pass_early_thread_jumps::execute): Same.
16771 2021-08-12 Tobias Burnus <tobias@codesourcery.com>
16773 * tree-core.h (omp_clause_proc_bind_kind): Add
16774 OMP_CLAUSE_PROC_BIND_PRIMARY.
16775 * tree-pretty-print.c (dump_omp_clause): Add TODO comment to
16776 change 'master' to 'primary' in proc_bind for OpenMP 5.1.
16778 2021-08-12 Claudiu Zissulescu <claziss@synopsys.com>
16780 * common/config/arc/arc-common.c (arc_option_init_struct): Remove
16781 fno-common reference.
16782 * config/arc/arc.c (arc_override_options): Remove overriding of
16785 2021-08-12 Jakub Jelinek <jakub@redhat.com>
16788 * config/i386/i386-expand.c (ix86_expand_vec_one_operand_perm_avx512):
16789 If d->testing_p, return true after performing checks instead of
16790 actually expanding the insn.
16791 (expand_vec_perm_broadcast_1): Handle V32HImode - assert
16792 !TARGET_AVX512BW and return false.
16794 2021-08-12 Eric Botcazou <ebotcazou@gcc.gnu.org>
16796 * configure.ac (PE linker --disable-dynamicbase support): New check.
16797 * configure: Regenerate.
16798 * config.in: Likewise.
16799 * config/i386/mingw32.h (LINK_SPEC_DISABLE_DYNAMICBASE): New define.
16800 (LINK_SPEC): Use it.
16801 * config/i386/mingw-w64.h (LINK_SPEC_DISABLE_DYNAMICBASE): Likewise.
16802 (LINK_SPEC): Likewise.
16804 2021-08-12 liuhongt <hongtao.liu@intel.com>
16807 * config/i386/sse.md (*avx2_zero_extendv16qiv16hi2_2): New
16808 post_reload define_insn_and_split.
16809 (*avx512bw_zero_extendv32qiv32hi2_2): Ditto.
16810 (*sse4_1_zero_extendv8qiv8hi2_4): Ditto.
16811 (*avx512f_zero_extendv16hiv16si2_2): Ditto.
16812 (*avx2_zero_extendv8hiv8si2_2): Ditto.
16813 (*sse4_1_zero_extendv4hiv4si2_4): Ditto.
16814 (*avx512f_zero_extendv8siv8di2_2): Ditto.
16815 (*avx2_zero_extendv4siv4di2_2): Ditto.
16816 (*sse4_1_zero_extendv2siv2di2_4): Ditto.
16817 (VI248_256, VI248_512, VI148_512, VI148_256, VI148_128): New
16820 2021-08-11 Bill Schmidt <wschmidt@linux.ibm.com>
16822 * config/rs6000/rs6000-builtin-new.def: Add always, power5, and
16825 2021-08-11 Bill Schmidt <wschmidt@linux.ibm.com>
16827 * config/rs6000/rs6000-builtin-new.def: Add vsx stanza.
16829 2021-08-11 Bill Schmidt <wschmidt@linux.ibm.com>
16831 * config/rs6000/rs6000-builtin-new.def: Finish altivec stanza.
16832 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Move
16833 initialization of pcvoid_type_node here...
16834 (altivec_init_builtins): ...from here.
16835 * config/rs6000/rs6000.h (rs6000_builtin_type_index): Add
16836 RS6000_BTI_const_ptr_void.
16837 (pcvoid_type_node): New macro.
16839 2021-08-11 Richard Biener <rguenther@suse.de>
16842 * tree-ssa-forwprop.c (pass_forwprop::execute): Do not decompose
16843 hard-register accesses.
16845 2021-08-11 Richard Biener <rguenther@suse.de>
16847 * tree-ssa-operands.c (operands_scanner::get_expr_operands):
16848 Do not look at COMPONENT_REF FIELD_DECLs TREE_THIS_VOLATILE
16849 to determine has_volatile_ops.
16851 2021-08-11 Eric Botcazou <ebotcazou@gcc.gnu.org>
16853 * cfgexpand.c (expand_used_vars): Reuse attribs local variable.
16855 2021-08-11 Jan Hubicka <hubicka@ucw.cz>
16856 Alexandre Oliva <oliva@adacore.com>
16858 * ipa-modref.c (modref_lattice::dump): Fix escape_point's min_flags
16860 (modref_lattice::merge_deref): Fix handling of indirect scape points.
16861 (update_escape_summary_1): Likewise.
16862 (update_escape_summary): Likewise.
16863 (ipa_merge_modref_summary_after_inlining): Likewise.
16865 2021-08-11 Richard Biener <rguenther@suse.de>
16867 PR middle-end/101858
16868 * fold-const.c (fold_binary_loc): Guard simplification
16869 of X < (cast) (1 << Y) to integer types.
16871 2021-08-11 Richard Biener <rguenther@suse.de>
16873 PR tree-optimization/101861
16874 * tree-vect-stmts.c (vectorizable_load): Fix error in
16875 previous change with regard to gather vectorization.
16877 2021-08-11 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
16880 * config/arm/arm_neon.h (vdup_n_s8): Replace call to builtin
16882 (vdup_n_s16): Likewise.
16883 (vdup_n_s32): Likewise.
16884 (vdup_n_s64): Likewise.
16885 (vdup_n_u8): Likewise.
16886 (vdup_n_u16): Likewise.
16887 (vdup_n_u32): Likewise.
16888 (vdup_n_u64): Likewise.
16889 (vdup_n_p8): Likewise.
16890 (vdup_n_p16): Likewise.
16891 (vdup_n_p64): Likewise.
16892 (vdup_n_f16): Likewise.
16893 (vdup_n_f32): Likewise.
16894 (vdupq_n_s8): Likewise.
16895 (vdupq_n_s16): Likewise.
16896 (vdupq_n_s32): Likewise.
16897 (vdupq_n_s64): Likewise.
16898 (vdupq_n_u8): Likewise.
16899 (vdupq_n_u16): Likewise.
16900 (vdupq_n_u32): Likewise.
16901 (vdupq_n_u64): Likewise.
16902 (vdupq_n_p8): Likewise.
16903 (vdupq_n_p16): Likewise.
16904 (vdupq_n_p64): Likewise.
16905 (vdupq_n_f16): Likewise.
16906 (vdupq_n_f32): Likewise.
16907 (vmov_n_s8): Replace call to builtin with call to corresponding
16909 (vmov_n_s16): Likewise.
16910 (vmov_n_s32): Likewise.
16911 (vmov_n_s64): Likewise.
16912 (vmov_n_u8): Likewise.
16913 (vmov_n_u16): Likewise.
16914 (vmov_n_u32): Likewise.
16915 (vmov_n_u64): Likewise.
16916 (vmov_n_p8): Likewise.
16917 (vmov_n_p16): Likewise.
16918 (vmov_n_f16): Likewise.
16919 (vmov_n_f32): Likewise.
16920 (vmovq_n_s8): Likewise.
16921 (vmovq_n_s16): Likewise.
16922 (vmovq_n_s32): Likewise.
16923 (vmovq_n_s64): Likewise.
16924 (vmovq_n_u8): Likewise.
16925 (vmovq_n_u16): Likewise.
16926 (vmovq_n_u32): Likewise.
16927 (vmovq_n_u64): Likewise.
16928 (vmovq_n_p8): Likewise.
16929 (vmovq_n_p16): Likewise.
16930 (vmovq_n_f16): Likewise.
16931 (vmovq_n_f32): Likewise.
16932 * config/arm/arm_neon_builtins.def: Remove entries for vdup_n.
16934 2021-08-11 liuhongt <hongtao.liu@intel.com>
16937 * config/i386/i386.md (ldexp<mode>3): Extend to vscalefs[sd]
16938 when TARGET_AVX512F and TARGET_SSE_MATH.
16940 2021-08-10 Jakub Jelinek <jakub@redhat.com>
16943 * config/i386/i386-expand.c (expand_vec_perm_even_odd): Return false
16944 for V32HImode if !TARGET_AVX512BW.
16945 (ix86_vectorize_vec_perm_const) <case E_V32HImode, case E_V64QImode>:
16946 If !TARGET_AVX512BW and TARGET_AVX512F and d.testing_p, don't fail
16947 early, but actually check the permutation.
16949 2021-08-10 Richard Biener <rguenther@suse.de>
16951 PR tree-optimization/101809
16952 * tree-vect-stmts.c (get_load_store_type): Allow emulated
16953 gathers with offset vector nunits being a constant multiple
16954 of the data vector nunits.
16955 (vect_get_gather_scatter_ops): Use the appropriate nunits
16956 for the offset vector defs.
16957 (vectorizable_store): Adjust call to
16958 vect_get_gather_scatter_ops.
16959 (vectorizable_load): Likewise. Handle the case of less
16960 offset vectors than data vectors.
16962 2021-08-10 Jakub Jelinek <jakub@redhat.com>
16965 * config/i386/sse.md (*avx512f_shuf_<shuffletype>64x2_1<mask_name>_1,
16966 *avx512f_shuf_<shuffletype>32x4_1<mask_name>_1): New define_insn
16969 2021-08-10 Richard Biener <rguenther@suse.de>
16971 PR tree-optimization/101801
16972 PR tree-optimization/101819
16973 * tree-vectorizer.h (vect_emulated_vector_p): Declare.
16974 * tree-vect-loop.c (vect_emulated_vector_p): New function.
16975 (vectorizable_reduction): Re-instantiate a check for emulated
16977 * tree-vect-stmts.c (vectorizable_shift): Likewise.
16978 (vectorizable_operation): Likewise. Cost emulated vector
16979 operations according to the scalar sequence synthesized by
16982 2021-08-10 Richard Biener <rguenther@suse.de>
16984 PR middle-end/101824
16985 * tree-nested.c (get_frame_field): Mark the COMPONENT_REF as
16986 volatile in case the variable was.
16988 2021-08-10 H.J. Lu <hjl.tools@gmail.com>
16991 * config/i386/constraints.md (BC): Document for integer SSE
16992 constant all bits set operand.
16993 (BF): New constraint for const floating-point all bits set
16995 * config/i386/i386.c (standard_sse_constant_p): Likewise.
16996 (standard_sse_constant_opcode): Likewise.
16997 * config/i386/sse.md (sseconstm1): New mode attribute.
16998 (mov<mode>_internal): Replace BC with <sseconstm1>.
17000 2021-08-10 liuhongt <hongtao.liu@intel.com>
17002 * config/i386/sse.md (cond_<insn><mode>): New expander.
17003 (VI248_AVX512VLBW): New mode iterator.
17004 * config/i386/predicates.md
17005 (nonimmediate_or_const_vec_dup_operand): New predicate.
17007 2021-08-09 Andrew MacLeod <amacleod@redhat.com>
17009 PR tree-optimization/101741
17010 * gimple-range-fold.cc (fold_using_range::range_of_builtin_call): Check
17011 type of parameter for toupper/tolower.
17013 2021-08-09 Martin Jambor <mjambor@suse.cz>
17015 PR testsuite/101654
17016 * ipa-prop.c (propagate_controlled_uses): Removed a spurious space.
17018 2021-08-09 Pat Haugen <pthaugen@linux.ibm.com>
17020 * config/rs6000/rs6000.c (is_load_insn1): Verify destination is a
17022 (is_store_insn1): Verify source is a register.
17024 2021-08-09 Uroš Bizjak <ubizjak@gmail.com>
17027 * config/i386/mmx.md (<any_logic:code>v2sf3):
17028 Rename from *mmx_<any_logic:code>v2sf3
17030 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
17032 * config/nvptx/nvptx.c: Cross-reference parts adapted in
17033 'gcc/omp-oacc-neuter-broadcast.cc'.
17034 * omp-low.c: Likewise.
17035 * omp-oacc-neuter-broadcast.cc: Cross-reference parts adapted from
17038 2021-08-09 Julian Brown <julian@codesourcery.com>
17039 Kwok Cheung Yeung <kcy@codesourcery.com>
17040 Thomas Schwinge <thomas@codesourcery.com>
17042 * config/gcn/gcn.c (gcn_init_builtins): Override decls for
17043 BUILT_IN_GOACC_SINGLE_START, BUILT_IN_GOACC_SINGLE_COPY_START,
17044 BUILT_IN_GOACC_SINGLE_COPY_END and BUILT_IN_GOACC_BARRIER.
17045 (gcn_goacc_validate_dims): Turn on worker partitioning unconditionally.
17046 (gcn_fork_join): Update comment.
17047 * config/gcn/gcn.opt (flag_worker_partitioning): Remove.
17048 (macc_experimental_workers): Remove unused option.
17050 2021-08-09 Julian Brown <julian@codesourcery.com>
17051 Nathan Sidwell <nathan@codesourcery.com> (via 'gcc/config/nvptx/nvptx.c' master)
17052 Kwok Cheung Yeung <kcy@codesourcery.com>
17053 Thomas Schwinge <thomas@codesourcery.com>
17055 * Makefile.in (OBJS): Add omp-oacc-neuter-broadcast.o.
17056 * doc/tm.texi.in (TARGET_GOACC_CREATE_WORKER_BROADCAST_RECORD):
17057 Add documentation hook.
17058 * doc/tm.texi: Regenerate.
17059 * omp-oacc-neuter-broadcast.cc: New file.
17060 * omp-builtins.def (BUILT_IN_GOACC_BARRIER)
17061 (BUILT_IN_GOACC_SINGLE_START, BUILT_IN_GOACC_SINGLE_COPY_START)
17062 (BUILT_IN_GOACC_SINGLE_COPY_END): New builtins.
17063 * passes.def (pass_omp_oacc_neuter_broadcast): Add pass.
17064 * target.def (goacc.create_worker_broadcast_record): Add target
17066 * tree-pass.h (make_pass_omp_oacc_neuter_broadcast): Add
17068 * config/gcn/gcn-protos.h (gcn_goacc_adjust_propagation_record):
17069 Rename prototype to...
17070 (gcn_goacc_create_worker_broadcast_record): ... this.
17071 * config/gcn/gcn-tree.c (gcn_goacc_adjust_propagation_record): Rename
17073 (gcn_goacc_create_worker_broadcast_record): ... this.
17074 * config/gcn/gcn.c (TARGET_GOACC_ADJUST_PROPAGATION_RECORD):
17076 (TARGET_GOACC_CREATE_WORKER_BROADCAST_RECORD): ... this.
17078 2021-08-09 Tejas Belagod <tejas.belagod@arm.com>
17081 * config/aarch64/aarch64-simd.md (vlshr<mode>3, vashr<mode>3): Use
17082 the right iterator.
17084 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
17086 * Makefile.in (GTFILES): Remove '$(srcdir)/omp-offload.c'.
17088 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
17090 * builtins.def (DEF_GOACC_BUILTIN, DEF_GOMP_BUILTIN): Don't
17091 consider '-foffload-abi'.
17092 * common.opt (-foffload-abi): Remove 'Var', 'Init'.
17093 * opts.c (common_handle_option) <-foffload-abi> [ACCEL_COMPILER]:
17096 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
17098 * optc-gen.awk: Sanity check that 'Init' doesn't appear without
17101 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
17103 * omp-builtins.def (BUILT_IN_ACC_GET_DEVICE_TYPE): Remove.
17105 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
17107 * doc/gty.texi (Files): Update.
17109 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
17111 * doc/gty.texi (Files): Fix GTY header file example.
17113 2021-08-09 Roger Sayle <roger@nextmovesoftware.com>
17115 * tree-ssa-ccp.c (value_mask_to_min_max): Helper function to
17116 determine the upper and lower bounds from a mask-value pair.
17117 (bit_value_unop) [ABS_EXPR, ABSU_EXPR]: Add support for
17118 absolute value and unsigned absolute value expressions.
17119 (bit_value_binop): Initialize *VAL's precision.
17120 [LT_EXPR, LE_EXPR]: Use value_mask_to_min_max to determine
17121 upper and lower bounds of operands. Add LE_EXPR/GE_EXPR
17122 support when the operands are unknown but potentially equal.
17123 [MIN_EXPR, MAX_EXPR]: Support minimum/maximum expressions.
17125 2021-08-09 Bin Cheng <bin.cheng@linux.alibaba.com>
17127 * config/aarch64/aarch64.md
17128 (*extend<SHORT:mode><GPI:mode>2_aarch64): Use %<GPI:w>0.
17130 2021-08-08 Sergei Trofimovich <siarheit@google.com>
17132 * lra-constraints.c: Fix s/otput/output/ typo.
17134 2021-08-06 Martin Sebor <msebor@redhat.com>
17136 * builtins.c (expand_builtin_memchr): Move to gimple-ssa-warn-access.cc.
17137 (expand_builtin_strcat): Same.
17138 (expand_builtin_stpncpy): Same.
17139 (expand_builtin_strncat): Same.
17140 (check_read_access): Same.
17141 (check_memop_access): Same.
17142 (expand_builtin_strlen): Move checks to gimple-ssa-warn-access.cc.
17143 (expand_builtin_strnlen): Same.
17144 (expand_builtin_memcpy): Same.
17145 (expand_builtin_memmove): Same.
17146 (expand_builtin_mempcpy): Same.
17147 (expand_builtin_strcpy): Same.
17148 (expand_builtin_strcpy_args): Same.
17149 (expand_builtin_stpcpy_1): Same.
17150 (expand_builtin_strncpy): Same.
17151 (expand_builtin_memset): Same.
17152 (expand_builtin_bzero): Same.
17153 (expand_builtin_strcmp): Same.
17154 (expand_builtin_strncmp): Same.
17155 (expand_builtin): Remove handlers.
17156 (fold_builtin_strlen): Add a comment.
17157 * builtins.h (check_access): Move to gimple-ssa-warn-access.cc.
17158 * calls.c (maybe_warn_nonstring_arg): Same.
17159 * diagnostic-spec.c (nowarn_spec_t::nowarn_spec_t): Add warning option.
17160 * gimple-fold.c (gimple_fold_builtin_strcpy): Pass argument to callee.
17161 (gimple_fold_builtin_stpcpy): Same.
17162 * gimple-ssa-warn-access.cc (has_location): New function.
17163 (get_location): Same.
17164 (get_callee_fndecl): Same.
17165 (call_nargs): Same.
17167 (warn_string_no_nul): Define.
17168 (unterminated_array): Same.
17169 (check_nul_terminated_array): Same.
17170 (maybe_warn_nonstring_arg): Same.
17171 (maybe_warn_for_bound): Same.
17172 (warn_for_access): Same.
17173 (check_access): Same.
17174 (check_memop_access): Same.
17175 (check_read_access): Same.
17176 (warn_dealloc_offset): Use helper functions.
17177 (maybe_emit_free_warning): Same.
17178 (class pass_waccess): Add members.
17179 (check_strcat): New function.
17180 (check_strncat): New function.
17181 (check_stxcpy): New function.
17182 (check_stxncpy): New function.
17183 (check_strncmp): New function.
17184 (pass_waccess::check_builtin): New function.
17185 (pass_waccess::check): Call it.
17186 * gimple-ssa-warn-access.h (warn_string_no_nul): Move here from
17188 (maybe_warn_for_bound): Same.
17189 (check_access): Same.
17190 (check_memop_access): Same.
17191 (check_read_access): Same.
17192 * pointer-query.h (struct access_data): Define a ctor overload.
17194 2021-08-06 Richard Biener <rguenther@suse.de>
17196 PR tree-optimization/101801
17197 * tree-vectorizer.h (vect_worthwhile_without_simd_p): Rename...
17198 (vect_can_vectorize_without_simd_p): ... to this.
17199 * tree-vect-loop.c (vect_worthwhile_without_simd_p): Rename...
17200 (vect_can_vectorize_without_simd_p): ... to this and fold
17201 in vect_min_worthwhile_factor.
17202 (vect_min_worthwhile_factor): Remove.
17203 (vectorizable_reduction): Adjust and remove the cost part.
17204 * tree-vect-stmts.c (vectorizable_shift): Likewise.
17205 (vectorizable_operation): Likewise.
17207 2021-08-06 Uroš Bizjak <ubizjak@gmail.com>
17210 * config/i386/i386.md (cmove reg-to-reg move elimination peephole2s):
17211 Add general_gr_operand predicate to operand 3.
17213 2021-08-06 Roger Sayle <roger@nextmovesoftware.com>
17215 * tree-ssa-phiopt.c (cond_removal_in_builtin_zero_pattern): Use
17216 CFN_BUILT_IN_CLRSB* instead of BUILT_IN_CLRSB* for consistency.
17218 2021-08-06 Tamar Christina <tamar.christina@arm.com>
17220 * config/aarch64/aarch64-sve-builtins.cc (register_svpattern,
17221 register_svprfop): Pass vec<> by pointer.
17222 * langhooks-def.h (lhd_simulate_enum_decl): Likewise.
17223 * langhooks.c (lhd_simulate_enum_decl): Likewise.
17224 * langhooks.h (struct lang_hooks_for_types): Likewise.
17226 2021-08-06 Jonathan Wright <jonathan.wright@arm.com>
17228 * config/aarch64/arm_neon.h (vst1_bf16_x2): Use
17229 __builtin_memcpy instead of constructing an additional
17230 __builtin_aarch64_simd_oi one vector at a time.
17231 (vst1q_bf16_x2): Likewise.
17232 (vst1_bf16_x3): Use __builtin_memcpy instead of constructing
17233 an additional __builtin_aarch64_simd_ci one vector at a time.
17234 (vst1q_bf16_x3): Likewise.
17235 (vst1_bf16_x4): Use __builtin_memcpy instead of a union.
17236 (vst1q_bf16_x4): Likewise.
17237 (vst2_bf16): Use __builtin_memcpy instead of constructing an
17238 additional __builtin_aarch64_simd_oi one vector at a time.
17239 (vst2q_bf16): Likewise.
17240 (vst3_bf16): Use __builtin_memcpy instead of constructing an
17241 additional __builtin_aarch64_simd_ci mode one vector at a
17243 (vst3q_bf16): Likewise.
17244 (vst4_bf16): Use __builtin_memcpy instead of constructing an
17245 additional __builtin_aarch64_simd_xi one vector at a time.
17246 (vst4q_bf16): Likewise.
17248 2021-08-06 Jonathan Wright <jonathan.wright@arm.com>
17250 * config/aarch64/arm_neon.h (__ST2_LANE_FUNC): Delete.
17251 (__ST2Q_LANE_FUNC): Delete.
17252 (vst2_lane_f16): Use __builtin_memcpy to copy vector
17253 structure instead of constructing __builtin_aarch64_simd_oi
17254 one vector at a time.
17255 (vst2_lane_f32): Likewise.
17256 (vst2_lane_f64): Likewise.
17257 (vst2_lane_p8): Likewise.
17258 (vst2_lane_p16): Likewise.
17259 (vst2_lane_p64): Likewise.
17260 (vst2_lane_s8): Likewise.
17261 (vst2_lane_s16): Likewise.
17262 (vst2_lane_s32): Likewise.
17263 (vst2_lane_s64): Likewise.
17264 (vst2_lane_u8): Likewise.
17265 (vst2_lane_u16): Likewise.
17266 (vst2_lane_u32): Likewise.
17267 (vst2_lane_u64): Likewise.
17268 (vst2_lane_bf16): Likewise.
17269 (vst2q_lane_f16): Use __builtin_memcpy to copy vector
17270 structure instead of using a union.
17271 (vst2q_lane_f32): Likewise.
17272 (vst2q_lane_f64): Likewise.
17273 (vst2q_lane_p8): Likewise.
17274 (vst2q_lane_p16): Likewise.
17275 (vst2q_lane_p64): Likewise.
17276 (vst2q_lane_s8): Likewise.
17277 (vst2q_lane_s16): Likewise.
17278 (vst2q_lane_s32): Likewise.
17279 (vst2q_lane_s64): Likewise.
17280 (vst2q_lane_u8): Likewise.
17281 (vst2q_lane_u16): Likewise.
17282 (vst2q_lane_u32): Likewise.
17283 (vst2q_lane_u64): Likewise.
17284 (vst2q_lane_bf16): Likewise.
17286 2021-08-06 Jonathan Wright <jonathan.wright@arm.com>
17288 * config/aarch64/arm_neon.h (__ST3_LANE_FUNC): Delete.
17289 (__ST3Q_LANE_FUNC): Delete.
17290 (vst3_lane_f16): Use __builtin_memcpy to copy vector
17291 structure instead of constructing __builtin_aarch64_simd_ci
17292 one vector at a time.
17293 (vst3_lane_f32): Likewise.
17294 (vst3_lane_f64): Likewise.
17295 (vst3_lane_p8): Likewise.
17296 (vst3_lane_p16): Likewise.
17297 (vst3_lane_p64): Likewise.
17298 (vst3_lane_s8): Likewise.
17299 (vst3_lane_s16): Likewise.
17300 (vst3_lane_s32): Likewise.
17301 (vst3_lane_s64): Likewise.
17302 (vst3_lane_u8): Likewise.
17303 (vst3_lane_u16): Likewise.
17304 (vst3_lane_u32): Likewise.
17305 (vst3_lane_u64): Likewise.
17306 (vst3_lane_bf16): Likewise.
17307 (vst3q_lane_f16): Use __builtin_memcpy to copy vector
17308 structure instead of using a union.
17309 (vst3q_lane_f32): Likewise.
17310 (vst3q_lane_f64): Likewise.
17311 (vst3q_lane_p8): Likewise.
17312 (vst3q_lane_p16): Likewise.
17313 (vst3q_lane_p64): Likewise.
17314 (vst3q_lane_s8): Likewise.
17315 (vst3q_lane_s16): Likewise.
17316 (vst3q_lane_s32): Likewise.
17317 (vst3q_lane_s64): Likewise.
17318 (vst3q_lane_u8): Likewise.
17319 (vst3q_lane_u16): Likewise.
17320 (vst3q_lane_u32): Likewise.
17321 (vst3q_lane_u64): Likewise.
17322 (vst3q_lane_bf16): Likewise.
17324 2021-08-06 Jonathan Wright <jonathan.wright@arm.com>
17326 * config/aarch64/arm_neon.h (__ST4_LANE_FUNC): Delete.
17327 (__ST4Q_LANE_FUNC): Delete.
17328 (vst4_lane_f16): Use __builtin_memcpy to copy vector
17329 structure instead of constructing __builtin_aarch64_simd_xi
17330 one vector at a time.
17331 (vst4_lane_f32): Likewise.
17332 (vst4_lane_f64): Likewise.
17333 (vst4_lane_p8): Likewise.
17334 (vst4_lane_p16): Likewise.
17335 (vst4_lane_p64): Likewise.
17336 (vst4_lane_s8): Likewise.
17337 (vst4_lane_s16): Likewise.
17338 (vst4_lane_s32): Likewise.
17339 (vst4_lane_s64): Likewise.
17340 (vst4_lane_u8): Likewise.
17341 (vst4_lane_u16): Likewise.
17342 (vst4_lane_u32): Likewise.
17343 (vst4_lane_u64): Likewise.
17344 (vst4_lane_bf16): Likewise.
17345 (vst4q_lane_f16): Use __builtin_memcpy to copy vector
17346 structure instead of using a union.
17347 (vst4q_lane_f32): Likewise.
17348 (vst4q_lane_f64): Likewise.
17349 (vst4q_lane_p8): Likewise.
17350 (vst4q_lane_p16): Likewise.
17351 (vst4q_lane_p64): Likewise.
17352 (vst4q_lane_s8): Likewise.
17353 (vst4q_lane_s16): Likewise.
17354 (vst4q_lane_s32): Likewise.
17355 (vst4q_lane_s64): Likewise.
17356 (vst4q_lane_u8): Likewise.
17357 (vst4q_lane_u16): Likewise.
17358 (vst4q_lane_u32): Likewise.
17359 (vst4q_lane_u64): Likewise.
17360 (vst4q_lane_bf16): Likewise.
17362 2021-08-06 Martin Liska <mliska@suse.cz>
17364 * config/rs6000/rs6000.c (rs6000_option_override_internal): When
17365 a target option is restored, it can have
17366 rs6000_long_double_type_size set to FLOAT_PRECISION_TFmode
17367 and error should not be emitted.
17369 2021-08-06 Sebastian Huber <sebastian.huber@embedded-brains.de>
17371 * gcov-io.h (gcov_write): Declare.
17372 * gcov-io.c (gcov_write): New.
17373 (gcov_write_counter): Remove.
17374 (gcov_write_tag_length): Likewise.
17375 (gcov_write_summary): Replace gcov_write_tag_length() with calls to
17376 gcov_write_unsigned().
17377 * doc/invoke.texi (fprofile-info-section): Mention
17378 __gcov_info_to_gdca().
17380 2021-08-06 Martin Sebor <msebor@redhat.com>
17382 * dominance.c (prune_bbs_to_update_dominators): Adjust by-value vec
17383 arguments to by-reference.
17384 (iterate_fix_dominators): Same.
17385 * dominance.h (iterate_fix_dominators): Same.
17386 * ipa-prop.h: Call auto_vec::to_vec_legacy.
17387 * tree-data-ref.c (dump_data_dependence_relation): Adjust by-value vec
17388 arguments to by-reference.
17389 (debug_data_dependence_relation): Same.
17390 (dump_data_dependence_relations): Same.
17391 * tree-data-ref.h (debug_data_dependence_relation): Same.
17392 (dump_data_dependence_relations): Same.
17393 * tree-predcom.c (dump_chains): Same.
17394 (initialize_root_vars_lm): Same.
17395 (determine_unroll_factor): Same.
17396 (replace_phis_by_defined_names): Same.
17397 (insert_init_seqs): Same.
17398 (pcom_worker::tree_predictive_commoning_loop): Call
17399 auto_vec::to_vec_legacy.
17400 * tree-ssa-pre.c (insert_into_preds_of_block): Adjust by-value vec
17401 arguments to by-reference.
17402 * tree-ssa-threadbackward.c (populate_worklist): Same.
17403 (back_threader::resolve_def): Same.
17404 * tree-vect-data-refs.c (vect_check_nonzero_value): Same.
17405 (vect_enhance_data_refs_alignment): Same.
17406 (vect_check_lower_bound): Same.
17407 (vect_prune_runtime_alias_test_list): Same.
17408 (vect_permute_store_chain): Same.
17409 * tree-vect-slp-patterns.c (vect_normalize_conj_loc): Same.
17410 * tree-vect-stmts.c (vect_create_vectorized_demotion_stmts): Same.
17411 * tree-vectorizer.h (vect_permute_store_chain): Same.
17412 * vec.c (test_init): New function.
17413 (vec_c_tests): Call new function.
17414 * vec.h (vec): Declare ctors, dtor, and assignment.
17415 (auto_vec::vec_to_legacy): New function.
17416 (vec::copy): Adjust initialization.
17418 2021-08-05 H.J. Lu <hjl.tools@gmail.com>
17421 * config/i386/i386.c (ix86_can_inline_p): Ignore MASK_80387 if
17422 callee only uses GPRs.
17423 * config/i386/ia32intrin.h: Revert commit 5463cee2770.
17424 * config/i386/serializeintrin.h: Revert commit 71958f740f1.
17425 * config/i386/x86gprintrin.h: Add
17426 #pragma GCC target("general-regs-only") and #pragma GCC pop_options
17427 to disable non-GPR ISAs.
17429 2021-08-05 Richard Sandiford <richard.sandiford@arm.com>
17431 PR middle-end/101787
17432 * doc/md.texi (cond_ashl, cond_ashr, cond_lshr): Document.
17434 2021-08-05 Richard Sandiford <richard.sandiford@arm.com>
17436 * tree-vectorizer.h (vect_is_store_elt_extraction, vect_is_reduction)
17437 (vect_reduc_type, vect_embedded_comparison_type, vect_comparison_type)
17438 (vect_is_extending_load, vect_is_integer_truncation): New functions,
17439 moved from aarch64.c but given different names.
17440 * config/aarch64/aarch64.c (aarch64_is_store_elt_extraction)
17441 (aarch64_is_reduction, aarch64_reduc_type)
17442 (aarch64_embedded_comparison_type, aarch64_comparison_type)
17443 (aarch64_extending_load_p, aarch64_integer_truncation_p): Delete
17444 in favor of the above. Update callers accordingly.
17446 2021-08-05 Richard Earnshaw <rearnsha@arm.com>
17449 * config/arm/arm-cpus.in (generic-armv7-a): Add quirk to suppress
17450 writing .cpu directive in asm output.
17451 * config/arm/arm.c (arm_identify_fpu_from_isa): New variable.
17452 (arm_last_printed_arch_string): Delete.
17453 (arm_last-printed_fpu_string): Delete.
17454 (arm_configure_build_target): If use of floating-point/SIMD is
17455 disabled, remove all fp/simd related features from the target ISA.
17456 (last_arm_targ_options): New variable.
17457 (arm_print_asm_arch_directives): Add new parameters. Change order
17458 of emitted directives and handle all cases here.
17459 (arm_file_start): Always call arm_print_asm_arch_directives, move
17460 all generation of .arch/.arch_extension here.
17461 (arm_file_end): Call arm_print_asm_arch.
17462 (arm_declare_function_name): Call arm_print_asm_arch_directives
17463 instead of printing .arch/.fpu directives directly.
17465 2021-08-05 Richard Earnshaw <rearnsha@arm.com>
17467 * config/arm/arm.c (arm_configure_build_target): Don't call
17468 arm_option_reconfigure_globals.
17469 (arm_option_restore): Call arm_option_reconfigure_globals after
17470 reconfiguring the target.
17471 * config/arm/arm-c.c (arm_pragma_target_parse): Likewise.
17473 2021-08-05 Richard Earnshaw <rearnsha@arm.com>
17475 * config/arm/arm.c (arm_configure_build_target): Ensure the target's
17476 arch_name is always set.
17478 2021-08-05 Jonathan Wright <jonathan.wright@arm.com>
17480 * config/aarch64/aarch64.c: Traverse RTL tree to prevent cost
17481 of vec_select high-half from being added into Neon subtract
17484 2021-08-05 Jonathan Wright <jonathan.wright@arm.com>
17486 * config/aarch64/aarch64.c: Traverse RTL tree to prevent cost
17487 of vec_select high-half from being added into Neon add cost.
17489 2021-08-05 Kewen Lin <linkw@linux.ibm.com>
17491 * cfgloop.h (loops_list::loops_list): Add one optional argument
17492 root and adjust accordingly, update loop tree walking and factor
17494 * cfgloop.c (loops_list::walk_loop_tree): ... this. New function.
17496 2021-08-05 Eric Botcazou <ebotcazou@gcc.gnu.org>
17498 PR tree-optimization/101626
17499 * tree-sra.c (propagate_subaccesses_from_rhs): Do not set the
17500 reverse scalar storage order on a pointer or vector component.
17502 2021-08-05 liuhongt <hongtao.liu@intel.com>
17504 * config/i386/sse.md (cond_<code><mode>): New expander.
17506 2021-08-05 liuhongt <hongtao.liu@intel.com>
17508 * config/i386/sse.md (cond_<code><mode>): New expander.
17510 2021-08-05 liuhongt <hongtao.liu@intel.com>
17512 * config/i386/sse.md (cond_<code><mode>): New expander.
17514 2021-08-04 David Malcolm <dmalcolm@redhat.com>
17517 * Makefile.in (ANALYZER_OBJS): Add analyzer/region-model-asm.o.
17519 2021-08-04 H.J. Lu <hjl.tools@gmail.com>
17522 * config/i386/i386.h (STORE_MAX_PIECES): Allow 16/32/64 bytes
17523 only if TARGET_INTER_UNIT_MOVES_TO_VEC is true.
17525 2021-08-04 H.J. Lu <hjl.tools@gmail.com>
17528 * config/i386/i386-expand.c (ix86_expand_vector_move): Call
17529 ix86_gen_scratch_sse_rtx to get a scratch SSE register to copy
17530 data with SSE register from one memory location to another.
17532 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
17534 * config/s390/s390.c (expand_perm_with_vpdi): New function.
17535 (vectorize_vec_perm_const_1): Call expand_perm_with_vpdi.
17536 * config/s390/vector.md (*vpdi1<mode>, @vpdi1<mode>): Enable a
17537 parameterized expander.
17538 (*vpdi4<mode>, @vpdi4<mode>): Likewise.
17540 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
17542 * config/s390/s390.c (MAX_VECT_LEN): Define macro.
17543 (struct expand_vec_perm_d): Define struct.
17544 (expand_perm_with_merge): New function.
17545 (vectorize_vec_perm_const_1): New function.
17546 (s390_vectorize_vec_perm_const): New function.
17547 (TARGET_VECTORIZE_VEC_PERM_CONST): Define target macro.
17549 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
17551 * config/s390/vector.md (V_HW_64): Remove mode iterator.
17552 (*vec_load_pair<mode>): Use V_HW_2 instead of V_HW_64.
17553 * config/s390/vx-builtins.md
17554 (vec_scatter_element<V_HW_2:mode>_SI): Use V_HW_2 instead of
17557 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
17559 * config/s390/s390.md (UNSPEC_VEC_PERMI): Remove constant
17561 * config/s390/vector.md (*vpdi1<mode>, *vpdi4<mode>): New pattern
17563 * config/s390/vx-builtins.md (*vec_permi<mode>): Emit generic rtx
17564 instead of an unspec.
17566 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
17568 * config/s390/s390-modes.def: Add more vector modes to support
17569 concatenation of two vectors.
17570 * config/s390/s390-protos.h (s390_expand_merge_perm_const): Add
17572 (s390_expand_merge): Likewise.
17573 * config/s390/s390.c (s390_expand_merge_perm_const): New function.
17574 (s390_expand_merge): New function.
17575 * config/s390/s390.md (UNSPEC_VEC_MERGEH, UNSPEC_VEC_MERGEL):
17576 Remove constant definitions.
17577 * config/s390/vector.md (V_HW_2): Add mode iterators.
17578 (VI_HW_4, V_HW_4): Rename VI_HW_4 to V_HW_4.
17579 (vec_2x_nelts, vec_2x_wide): New mode attributes.
17580 (*vmrhb, *vmrlb, *vmrhh, *vmrlh, *vmrhf, *vmrlf, *vmrhg, *vmrlg):
17581 New pattern definitions.
17582 (vec_widen_umult_lo_<mode>, vec_widen_umult_hi_<mode>)
17583 (vec_widen_smult_lo_<mode>, vec_widen_smult_hi_<mode>)
17584 (vec_unpacks_lo_v4sf, vec_unpacks_hi_v4sf, vec_unpacks_lo_v2df)
17585 (vec_unpacks_hi_v2df): Adjust expanders to emit non-unspec RTX for
17587 * config/s390/vx-builtins.md (V_HW_4): Remove mode iterator. Now
17589 (vec_mergeh<mode>, vec_mergel<mode>): Use s390_expand_merge to
17590 emit vec merge pattern.
17592 2021-08-04 Jonathan Wright <jonathan.wright@arm.com>
17594 * config/aarch64/aarch64.c (aarch64_strip_extend_vec_half):
17596 (aarch64_rtx_mult_cost): Traverse RTL tree to prevent cost of
17597 vec_select high-half from being added into Neon multiply
17599 * rtlanal.c (vec_series_highpart_p): Define.
17600 * rtlanal.h (vec_series_highpart_p): Declare.
17602 2021-08-04 Jonathan Wright <jonathan.wright@arm.com>
17604 * config/aarch64/aarch64.c (aarch64_strip_duplicate_vec_elt):
17606 (aarch64_rtx_mult_cost): Traverse RTL tree to prevent
17607 vec_select cost from being added into Neon multiply cost.
17609 2021-08-04 Richard Sandiford <richard.sandiford@arm.com>
17611 * tree-vect-loop.c (vect_better_loop_vinfo_p): Detect cases in
17612 which old_loop_vinfo is an epilogue loop that handles a constant
17613 number of iterations.
17615 2021-08-04 Richard Sandiford <richard.sandiford@arm.com>
17617 * tree-vect-loop.c (vect_analyze_loop): Print a dump message
17618 when a reanalyzed loop fails to be cheaper than the current
17621 2021-08-04 Richard Sandiford <richard.sandiford@arm.com>
17623 * config/aarch64/aarch64.c: Fix a typo.
17625 2021-08-04 Vincent Lefèvre <vincent-gcc@vinc17.net>
17627 PR gcov-profile/101773
17628 * gcov-io.c (gcov_close): Check return code of a fclose.
17630 2021-08-04 Bernd Edlinger <bernd.edlinger@hotmail.de>
17633 * dwarf2out.c (dwarf2out_assembly_start): Emit a dummy
17634 .file statement when needed.
17636 2021-08-04 Richard Biener <rguenther@suse.de>
17638 * tree-vect-data-refs.c (vect_check_gather_scatter):
17639 Include widening conversions only when the result is
17640 still handed by native gather or the current offset
17641 size not already matches the data size.
17642 Also succeed analysis in case there's no native support,
17643 noted by a IFN_LAST ifn and a NULL decl.
17644 (vect_analyze_data_refs): Always consider gathers.
17645 * tree-vect-patterns.c (vect_recog_gather_scatter_pattern):
17646 Test for no IFN gather rather than decl gather.
17647 * tree-vect-stmts.c (vect_model_load_cost): Pass in the
17648 gather-scatter info and cost emulated gathers accordingly.
17649 (vect_truncate_gather_scatter_offset): Properly test for
17651 (vect_use_strided_gather_scatters_p): Likewise.
17652 (get_load_store_type): Handle emulated gathers and its
17654 (vectorizable_load): Likewise. Emulate them by extracting
17655 scalar offsets, doing scalar loads and a vector construct.
17657 2021-08-04 H.J. Lu <hjl.tools@gmail.com>
17660 * expr.c (op_by_pieces_d::op_by_pieces_d): Add a max_pieces
17661 argument to set m_max_size.
17662 (move_by_pieces_d): Pass MOVE_MAX_PIECES to op_by_pieces_d.
17663 (store_by_pieces_d): Pass STORE_MAX_PIECES to op_by_pieces_d.
17664 (compare_by_pieces_d): Pass COMPARE_MAX_PIECES to op_by_pieces_d.
17666 2021-08-04 Roger Sayle <roger@nextmovesoftware.com>
17667 Marc Glisse <marc.glisse@inria.fr>
17669 * match.pd (bit_ior, bit_xor): Canonicalize (X*C1)|(X*C2) and
17670 (X*C1)^(X*C2) as X*(C1+C2), and related variants, using
17671 tree_nonzero_bits to ensure that operands are bit-wise disjoint.
17673 2021-08-04 Richard Biener <rguenther@suse.de>
17675 * tree-ssa-forwprop.c (pass_forwprop::execute): Split
17676 out code to decompose vector loads ...
17677 (optimize_vector_load): ... here. Generalize it to
17678 handle intermediate widening and TARGET_MEM_REF loads
17679 and apply it to loads with a supported vector mode as well.
17681 2021-08-04 Richard Biener <rguenther@suse.de>
17683 PR tree-optimization/101756
17684 * tree-vect-slp.c (vectorizable_bb_reduc_epilogue): Make sure
17685 the result of the reduction epilogue is compatible to the original
17688 2021-08-04 liuhongt <hongtao.liu@intel.com>
17691 * config/i386/i386.md (peephole2): Refine predicate from
17692 register_operand to general_reg_operand.
17694 2021-08-04 Aldy Hernandez <aldyh@redhat.com>
17696 * gimple-range-path.h (path_range_query::dump): Mark override.
17698 2021-08-04 Richard Biener <rguenther@suse.de>
17700 PR tree-optimization/101769
17701 * tree-tailcall.c (eliminate_tail_call): Add the created loop
17702 for the first recursion and return it via the new output parameter.
17703 (optimize_tail_call): Pass through new output param.
17704 (tree_optimize_tail_calls_1): After creating all latches,
17705 add the created loop to the loop tree. Do not mark loops for fixup.
17707 2021-08-04 Martin Liska <mliska@suse.cz>
17709 * doc/invoke.texi: Document threader-mode param.
17711 2021-08-04 liuhongt <hongtao.liu@intel.com>
17713 * config/i386/sse.md (cond_fma<mode>): New expander.
17714 (cond_fms<mode>): Ditto.
17715 (cond_fnma<mode>): Ditto.
17716 (cond_fnms<mode>): Ditto.
17718 2021-08-03 Segher Boessenkool <segher@kernel.crashing.org>
17720 * config/rs6000/vsx.md (*vsx_le_perm_store_<mode>): Use && instead of &.
17722 2021-08-03 Segher Boessenkool <segher@kernel.crashing.org>
17724 * config/rs6000/constraints.md: Remove "e" from the list of available
17725 constraint characters.
17727 2021-08-03 Eugene Rozenfeld <erozen@microsoft.com>
17729 PR gcov-profile/71672
17730 * auto-profile.c (afdo_indirect_call): Fix setup of the historgram value for indirect calls.
17732 2021-08-03 Paul A. Clarke <pc@us.ibm.com>
17734 * config/rs6000/smmintrin.h (_mm_minpos_epu16): New.
17736 2021-08-03 H.J. Lu <hjl.tools@gmail.com>
17738 * config/i386/i386.c (ix86_gen_scratch_sse_rtx): In 64-bit mode,
17739 try XMM31 to avoid vzeroupper.
17741 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
17743 * doc/invoke.texi: Document -mtune=neoverse-512tvb and
17744 -mcpu=neoverse-512tvb.
17745 * config/aarch64/aarch64-cores.def (neoverse-512tvb): New entry.
17746 * config/aarch64/aarch64-tune.md: Regenerate.
17747 * config/aarch64/aarch64.c (neoverse512tvb_sve_vector_cost)
17748 (neoverse512tvb_sve_issue_info, neoverse512tvb_vec_issue_info)
17749 (neoverse512tvb_vector_cost, neoverse512tvb_tunings): New structures.
17750 (aarch64_adjust_body_cost_sve): Handle -mtune=neoverse-512tvb.
17751 (aarch64_adjust_body_cost): Likewise.
17753 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
17755 * config/aarch64/aarch64.c (aarch64_add_stmt_cost): Only
17756 record issue information for operations that occur in the
17759 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
17761 * config/aarch64/aarch64.c (aarch64_multiply_add_p): Add a vec_flags
17762 parameter. Detect cases in which an Advanced SIMD MLA would almost
17763 certainly require a MOV.
17764 (aarch64_count_ops): Update accordingly.
17766 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
17768 * config/aarch64/aarch64.c (aarch64_is_store_elt_extraction): New
17769 function, split out from...
17770 (aarch64_detect_vector_stmt_subtype): ...here.
17771 (aarch64_add_stmt_cost): Treat extracting element 0 as free.
17773 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
17775 * config/aarch64/aarch64-protos.h (sve_vec_cost):
17776 Add gather_load_x32_cost and gather_load_x64_cost.
17777 * config/aarch64/aarch64.c (generic_sve_vector_cost)
17778 (a64fx_sve_vector_cost, neoversev1_sve_vector_cost): Update
17779 accordingly, using the values given by the scalar_load * number
17780 of elements calculation that we used previously.
17781 (aarch64_detect_vector_stmt_subtype): Use the new fields.
17783 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
17785 * config/aarch64/aarch64.c (aarch64_adjust_body_cost_sve): New
17786 function, split out from...
17787 (aarch64_adjust_body_cost): ...here.
17789 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
17791 * config/aarch64/fractional-cost.h: New file.
17792 * config/aarch64/aarch64.c: Include <algorithm> (indirectly)
17793 and cost_fraction.h.
17794 (vec_cost_fraction): New typedef.
17795 (aarch64_detect_scalar_stmt_subtype): Use it for statement costs.
17796 (aarch64_detect_vector_stmt_subtype): Likewise.
17797 (aarch64_sve_adjust_stmt_cost, aarch64_adjust_stmt_cost): Likewise.
17798 (aarch64_estimate_min_cycles_per_iter): Use vec_cost_fraction
17800 (aarch64_adjust_body_cost): Likewise.
17801 (aarch64_test_cost_fraction): New function.
17802 (aarch64_run_selftests): Call it.
17804 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
17806 * config/aarch64/aarch64-protos.h (tune_params::sve_width): Turn
17808 * config/aarch64/aarch64.c (aarch64_cmp_autovec_modes): Update
17810 (aarch64_estimated_poly_value): Likewise. Use the least significant
17811 set bit for the minimum and likely values. Use the most significant
17812 set bit for the maximum value.
17814 2021-08-03 liuhongt <hongtao.liu@intel.com>
17816 * config/i386/sse.md (cond_<insn><mode>): New expander.
17817 (cond_mul<mode>): Ditto.
17819 2021-08-03 Kewen Lin <linkw@linux.ibm.com>
17821 * tree-cfg.c (move_sese_region_to_fn): Fix typos on dloop.
17823 2021-08-03 liuhongt <hongtao.liu@intel.com>
17825 * config/i386/sse.md (cond_<insn><mode>):New expander.
17826 (cond_mul<mode>): Ditto.
17827 (cond_div<mode>): Ditto.
17829 2021-08-02 H.J. Lu <hjl.tools@gmail.com>
17831 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Also
17832 check stack_realign_needed for stack realignment.
17833 (ix86_legitimate_constant_p): Always allow CONST_WIDE_INT smaller
17834 than the largest integer supported by vector register.
17835 * config/i386/i386.h (MAX_MOVE_MAX): New. Set to 64.
17836 (MOVE_MAX): Set to bytes of the largest integer supported by
17838 (STORE_MAX_PIECES): New.
17840 2021-08-02 H.J. Lu <hjl.tools@gmail.com>
17842 * config/i386/i386-expand.c (ix86_expand_vector_move): Call
17843 ix86_gen_scratch_sse_rtx to get a scratch SSE register to copy
17844 data from one memory location to another.
17846 2021-08-02 H.J. Lu <hjl.tools@gmail.com>
17848 PR middle-end/90773
17849 * config/i386/i386.c (TARGET_GEN_MEMSET_SCRATCH_RTX): New.
17851 2021-08-02 Aldy Hernandez <aldyh@redhat.com>
17853 PR tree-optimization/101724
17854 * params.opt: Remove --param=threader-iterative.
17855 * tree-ssa-threadbackward.c (pass_thread_jumps::execute): Remove
17858 2021-08-02 Tom de Vries <tdevries@suse.de>
17860 PR middle-end/101665
17861 * doc/extend.texi (nonnull attribute): Improve documentation.
17863 2021-08-02 Andrew Pinski <apinski@marvell.com>
17865 PR rtl-optimization/101683
17866 * rtlanal.c (may_trap_p_1): Handle UNSIGNED_FIX.
17868 2021-08-02 Roger Sayle <roger@nextmovesoftware.com>
17870 * tree-ssa-phiopt.c (cond_removal_in_builtin_zero_pattern):
17871 Renamed from cond_removal_in_popcount_clz_ctz_pattern.
17872 Add support for BSWAP, FFS, PARITY and CLRSB builtins.
17873 (tree_ssa_phiop_worker): Update call to function above.
17875 2021-08-01 H.J. Lu <hjl.tools@gmail.com>
17878 * config/i386/i386.md (bsr_rex64_1_zext): New.
17879 (combine splitter for constant - clzll): Replace gen_bsr_rex64_1
17880 with gen_bsr_rex64_1_zext.
17882 2021-07-31 Jakub Jelinek <jakub@redhat.com>
17885 * config/i386/i386.md (bsr_rex64_1, bsr_1, bsr_zext_1): New
17886 define_insn patterns.
17887 (*bsr_rex64_2, *bsr_2): New define_insn_and_split patterns.
17888 Add combine splitters for constant - clz.
17889 (clz<mode>2): Use a temporary pseudo for bsr result.
17891 2021-07-30 Paul A. Clarke <pc@us.ibm.com>
17893 * config/rs6000/smmintrin.h (_mm_floor_pd, _mm_floor_ps,
17894 _mm_floor_sd, _mm_floor_ss): New.
17896 2021-07-30 Paul A. Clarke <pc@us.ibm.com>
17898 * config/rs6000/smmintrin.h (_mm_ceil_pd, _mm_ceil_ps,
17899 _mm_ceil_sd, _mm_ceil_ss): New.
17901 2021-07-30 Paul A. Clarke <pc@us.ibm.com>
17903 * config/rs6000/smmintrin.h (_mm_blend_pd, _mm_blendv_pd,
17904 _mm_blend_ps, _mm_blendv_ps): New.
17906 2021-07-30 Roger Sayle <roger@nextmovesoftware.com>
17907 Uroš Bizjak <ubizjak@gmail.com>
17909 * config/i386/i386.md (*dec_cmov<mode>): New define_insn_and_split
17910 to generate a conditional move using the carry flag after sub $1.
17911 (peephole2): Eliminate a register-to-register move by inverting
17912 the condition of a conditional move.
17914 2021-07-30 Hans-Peter Nilsson <hp@bitrange.com>
17916 * config/mmix/mmix.md ("call", "call_value", "*call_real")
17917 ("*call_value_real"): Don't generate rtx mentioning the generic
17918 operands 1 and 2 to "call", and similarly for "call_value".
17919 * config/mmix/mmix.c (mmix_print_operand_punct_valid_p)
17920 (mmix_print_operand): Use '!' instead of 'p'.
17922 2021-07-30 Hans-Peter Nilsson <hp@bitrange.com>
17924 * doc/md.texi (call): Correct information about operand 2.
17925 * config/mmix/mmix.md ("call", "call_value"): Remove fixed FIXMEs.
17927 2021-07-30 Andrew MacLeod <amacleod@redhat.com>
17929 * range-op.cc (operator_trunc_mod::wi_fold): Fold constants.
17931 2021-07-30 Andrew MacLeod <amacleod@redhat.com>
17933 * range-op.cc (operator_div::wi_fold): Return UNDEFINED for [0, 0] divisor.
17935 2021-07-30 Andrew MacLeod <amacleod@redhat.com>
17937 * gimple-range-cache.cc (*::set_bb_range): Change const basic_block to
17938 const_basic_block..
17939 (*::get_bb_range): Ditto.
17940 (*::bb_range_p): Ditto.
17941 * gimple-range-cache.h: Change prototypes.
17943 2021-07-30 H.J. Lu <hjl.tools@gmail.com>
17945 PR middle-end/90773
17946 * builtins.c (builtin_memcpy_read_str): Change the mode argument
17947 from scalar_int_mode to fixed_size_mode.
17948 (builtin_strncpy_read_str): Likewise.
17949 (gen_memset_value_from_prev): New function.
17950 (builtin_memset_read_str): Change the mode argument from
17951 scalar_int_mode to fixed_size_mode. Use gen_memset_value_from_prev
17952 and support CONST_VECTOR.
17953 (builtin_memset_gen_str): Likewise.
17954 (try_store_by_multiple_pieces): Use by_pieces_constfn to declare
17956 * builtins.h (builtin_strncpy_read_str): Replace scalar_int_mode
17957 with fixed_size_mode.
17958 (builtin_memset_read_str): Likewise.
17959 * expr.c (widest_int_mode_for_size): Renamed to ...
17960 (widest_fixed_size_mode_for_size): Add a bool argument to
17961 indicate if QI vector mode can be used.
17962 (by_pieces_ninsns): Call widest_fixed_size_mode_for_size
17963 instead of widest_int_mode_for_size.
17964 (pieces_addr::adjust): Change the mode argument from
17965 scalar_int_mode to fixed_size_mode.
17966 (op_by_pieces_d): Make m_len read-only. Add a bool member,
17967 m_qi_vector_mode, to indicate that QI vector mode can be used.
17968 (op_by_pieces_d::op_by_pieces_d): Add a bool argument to
17969 initialize m_qi_vector_mode. Call widest_fixed_size_mode_for_size
17970 instead of widest_int_mode_for_size.
17971 (op_by_pieces_d::get_usable_mode): Change the mode argument from
17972 scalar_int_mode to fixed_size_mode. Call
17973 widest_fixed_size_mode_for_size instead of
17974 widest_int_mode_for_size.
17975 (op_by_pieces_d::smallest_fixed_size_mode_for_size): New member
17976 function to return the smallest integer or QI vector mode.
17977 (op_by_pieces_d::run): Call widest_fixed_size_mode_for_size
17978 instead of widest_int_mode_for_size. Call
17979 smallest_fixed_size_mode_for_size instead of
17980 smallest_int_mode_for_size.
17981 (store_by_pieces_d::store_by_pieces_d): Add a bool argument to
17982 indicate that QI vector mode can be used and pass it to
17983 op_by_pieces_d::op_by_pieces_d.
17984 (can_store_by_pieces): Call widest_fixed_size_mode_for_size
17985 instead of widest_int_mode_for_size. Pass memsetp to
17986 widest_fixed_size_mode_for_size to support QI vector mode.
17987 Allow all CONST_VECTORs for memset if vec_duplicate is supported.
17988 (store_by_pieces): Pass memsetp to
17989 store_by_pieces_d::store_by_pieces_d.
17990 (clear_by_pieces_1): Removed.
17991 (clear_by_pieces): Replace clear_by_pieces_1 with
17992 builtin_memset_read_str and pass true to store_by_pieces_d to
17993 support vector mode broadcast.
17994 (string_cst_read_str): Change the mode argument from
17995 scalar_int_mode to fixed_size_mode.
17996 * expr.h (by_pieces_constfn): Change scalar_int_mode to
17998 (by_pieces_prev): Likewise.
17999 * rtl.h (lowpart_subreg_regno): New.
18000 * rtlanal.c (lowpart_subreg_regno): New. A wrapper around
18001 simplify_subreg_regno.
18002 * target.def (gen_memset_scratch_rtx): New hook.
18003 * doc/tm.texi.in: Add TARGET_GEN_MEMSET_SCRATCH_RTX.
18004 * doc/tm.texi: Regenerated.
18006 2021-07-30 Xi Ruoyao <xry111@mengyan1223.wang>
18009 * config/mips/mips.c (mips_atomic_assign_expand_fenv): Use
18010 TARGET_EXPR instead of MODIFY_EXPR.
18012 2021-07-30 Xi Ruoyao <xry111@mengyan1223.wang>
18015 * config/mips/mips-protos.h (mips_expand_vec_cmp_expr): Declare.
18016 * config/mips/mips.c (mips_expand_vec_cmp_expr): New function.
18017 * config/mips/mips-msa.md (vec_cmp<MSA:mode><mode_i>): New
18019 (vec_cmpu<IMSA:mode><mode_i>): New expander.
18021 2021-07-30 H.J. Lu <hjl.tools@gmail.com>
18024 * config/i386/i386-options.c (ix86_option_override_internal):
18025 Don't enable LZCNT/POPCNT if they have been disabled explicitly.
18027 2021-07-30 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
18030 * config/arm/arm_neon.h (vld1_p64): Replace call to builtin by
18031 explicitly dereferencing __a.
18032 (vld1_s64): Likewise.
18033 (vld1_u64): Likewise.
18034 * config/arm/arm_neon_builtins.def (vld1): Remove entry for di
18035 and change to VAR13.
18037 2021-07-30 Aldy Hernandez <aldyh@redhat.com>
18039 * gimple-loop-versioning.cc (lv_dom_walker::lv_dom_walker): Remove
18040 use of m_range_analyzer.
18041 (loop_versioning::lv_dom_walker::before_dom_children): Same.
18042 (loop_versioning::lv_dom_walker::after_dom_children): Remove.
18043 (loop_versioning::prune_loop_conditions): Replace vr_values use
18044 with range_query interface.
18045 (pass_loop_versioning::execute): Use ranger.
18047 2021-07-30 Xi Ruoyao <xry111@mengyan1223.wang>
18050 * ipa-devirt.c (ipa_odr_read_section): Compare the precision of
18051 enum values, and emit a warning if they mismatch.
18053 2021-07-30 Kewen Lin <linkw@linux.ibm.com>
18055 * cfgloop.h (as_const): New function.
18056 (class loop_iterator): Rename to ...
18057 (class loops_list): ... this.
18058 (loop_iterator::next): Rename to ...
18059 (loops_list::Iter::fill_curr_loop): ... this and adjust.
18060 (loop_iterator::loop_iterator): Rename to ...
18061 (loops_list::loops_list): ... this and adjust.
18062 (loops_list::Iter): New class.
18063 (loops_list::iterator): New type.
18064 (loops_list::const_iterator): New type.
18065 (loops_list::begin): New function.
18066 (loops_list::end): Likewise.
18067 (loops_list::begin const): Likewise.
18068 (loops_list::end const): Likewise.
18069 (FOR_EACH_LOOP): Remove.
18070 (FOR_EACH_LOOP_FN): Remove.
18071 * cfgloop.c (flow_loops_dump): Adjust FOR_EACH_LOOP* with range-based
18072 for loop with loops_list instance.
18073 (sort_sibling_loops): Likewise.
18074 (disambiguate_loops_with_multiple_latches): Likewise.
18075 (verify_loop_structure): Likewise.
18076 * cfgloopmanip.c (create_preheaders): Likewise.
18077 (force_single_succ_latches): Likewise.
18078 * config/aarch64/falkor-tag-collision-avoidance.c
18079 (execute_tag_collision_avoidance): Likewise.
18080 * config/mn10300/mn10300.c (mn10300_scan_for_setlb_lcc): Likewise.
18081 * config/s390/s390.c (s390_adjust_loops): Likewise.
18082 * doc/loop.texi: Likewise.
18083 * gimple-loop-interchange.cc (pass_linterchange::execute): Likewise.
18084 * gimple-loop-jam.c (tree_loop_unroll_and_jam): Likewise.
18085 * gimple-loop-versioning.cc (loop_versioning::analyze_blocks): Likewise.
18086 (loop_versioning::make_versioning_decisions): Likewise.
18087 * gimple-ssa-split-paths.c (split_paths): Likewise.
18088 * graphite-isl-ast-to-gimple.c (graphite_regenerate_ast_isl): Likewise.
18089 * graphite.c (canonicalize_loop_form): Likewise.
18090 (graphite_transform_loops): Likewise.
18091 * ipa-fnsummary.c (analyze_function_body): Likewise.
18092 * ipa-pure-const.c (analyze_function): Likewise.
18093 * loop-doloop.c (doloop_optimize_loops): Likewise.
18094 * loop-init.c (loop_optimizer_finalize): Likewise.
18095 (fix_loop_structure): Likewise.
18096 * loop-invariant.c (calculate_loop_reg_pressure): Likewise.
18097 (move_loop_invariants): Likewise.
18098 * loop-unroll.c (decide_unrolling): Likewise.
18099 (unroll_loops): Likewise.
18100 * modulo-sched.c (sms_schedule): Likewise.
18101 * predict.c (predict_loops): Likewise.
18102 (pass_profile::execute): Likewise.
18103 * profile.c (branch_prob): Likewise.
18104 * sel-sched-ir.c (sel_finish_pipelining): Likewise.
18105 (sel_find_rgns): Likewise.
18106 * tree-cfg.c (replace_loop_annotate): Likewise.
18107 (replace_uses_by): Likewise.
18108 (move_sese_region_to_fn): Likewise.
18109 * tree-if-conv.c (pass_if_conversion::execute): Likewise.
18110 * tree-loop-distribution.c (loop_distribution::execute): Likewise.
18111 * tree-parloops.c (parallelize_loops): Likewise.
18112 * tree-predcom.c (tree_predictive_commoning): Likewise.
18113 * tree-scalar-evolution.c (scev_initialize): Likewise.
18114 (scev_reset): Likewise.
18115 * tree-ssa-dce.c (find_obviously_necessary_stmts): Likewise.
18116 * tree-ssa-live.c (remove_unused_locals): Likewise.
18117 * tree-ssa-loop-ch.c (ch_base::copy_headers): Likewise.
18118 * tree-ssa-loop-im.c (analyze_memory_references): Likewise.
18119 (tree_ssa_lim_initialize): Likewise.
18120 * tree-ssa-loop-ivcanon.c (canonicalize_induction_variables): Likewise.
18121 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize): Likewise.
18122 * tree-ssa-loop-manip.c (get_loops_exits): Likewise.
18123 * tree-ssa-loop-niter.c (estimate_numbers_of_iterations): Likewise.
18124 (free_numbers_of_iterations_estimates): Likewise.
18125 * tree-ssa-loop-prefetch.c (tree_ssa_prefetch_arrays): Likewise.
18126 * tree-ssa-loop-split.c (tree_ssa_split_loops): Likewise.
18127 * tree-ssa-loop-unswitch.c (tree_ssa_unswitch_loops): Likewise.
18128 * tree-ssa-loop.c (gate_oacc_kernels): Likewise.
18129 (pass_scev_cprop::execute): Likewise.
18130 * tree-ssa-propagate.c (clean_up_loop_closed_phi): Likewise.
18131 * tree-ssa-sccvn.c (do_rpo_vn): Likewise.
18132 * tree-ssa-threadupdate.c
18133 (jump_thread_path_registry::thread_through_all_blocks): Likewise.
18134 * tree-vectorizer.c (vectorize_loops): Likewise.
18135 * tree-vrp.c (vrp_asserts::find_assert_locations): Likewise.
18137 2021-07-29 Hans-Peter Nilsson <hp@bitrange.com>
18139 * config/mmix/mmix.c (mmix_function_arg_1): Avoid
18140 generating a VOIDmode register for e.g the
18141 function_arg_info::end_marker.
18143 2021-07-29 Jeff Law <jeffreyalaw@gmail.com>
18145 * config/h8300/h8300-modes.def: Add CCZ, CCV and CCC, drop CCZNV.
18146 * config/h8300/h8300.md (H8cc mode iterator): Add CCZ.
18147 (cc mode_attr): Similarly.
18148 (ccz subst_attr): Similarly.
18149 * config/h8300/jumpcall.md: Add new patterns for branch-on-bit.
18150 * config/h8300/testcompare.md: Remove various cc0 based patterns
18151 that had been commented out. Add pattern to set CCZ from a bit
18154 2021-07-29 Thomas Schwinge <thomas@codesourcery.com>
18155 Julian Brown <julian@codesourcery.com>
18156 Kwok Cheung Yeung <kcy@codesourcery.com>
18158 * omp-offload.c (oacc_loop_xform_head_tail, oacc_loop_process):
18159 'update_stmt' after modification.
18160 (pass_oacc_loop_designation): New function, extracted out of...
18161 (pass_oacc_device_lower): ... this.
18162 (pass_data_oacc_loop_designation, pass_oacc_loop_designation)
18163 (make_pass_oacc_loop_designation): New
18164 * passes.def: Add it.
18165 * tree-parloops.c (create_parallel_loop): Adjust.
18166 * tree-pass.h (make_pass_oacc_loop_designation): New.
18168 2021-07-29 Aldy Hernandez <aldyh@redhat.com>
18170 * flag-types.h (enum threader_mode): New.
18171 * params.opt: Add entry for --param=threader-mode.
18172 * tree-ssa-threadbackward.c (THREADER_ITERATIVE_MODE): New.
18173 (class back_threader): New.
18174 (back_threader::back_threader): New.
18175 (back_threader::~back_threader): New.
18176 (back_threader::maybe_register_path): New.
18177 (back_threader::find_taken_edge): New.
18178 (back_threader::find_taken_edge_switch): New.
18179 (back_threader::find_taken_edge_cond): New.
18180 (back_threader::resolve_def): New.
18181 (back_threader::resolve_phi): New.
18182 (back_threader::find_paths_to_names): New.
18183 (back_threader::find_paths): New.
18186 (thread_jumps::find_jump_threads_backwards): Call ranger threader.
18187 (thread_jumps::find_jump_threads_backwards_with_ranger): New.
18188 (pass_thread_jumps::execute): Abstract out code...
18189 (try_thread_blocks): ...here.
18190 * tree-ssa-threadedge.c (jump_threader::thread_outgoing_edges):
18191 Abstract out threading candidate code to...
18192 (single_succ_to_potentially_threadable_block): ...here.
18193 * tree-ssa-threadedge.h (single_succ_to_potentially_threadable_block):
18195 * tree-ssa-threadupdate.c (register_jump_thread): Return boolean.
18196 * tree-ssa-threadupdate.h (class jump_thread_path_registry):
18197 Return bool from register_jump_thread.
18199 2021-07-29 Andreas Krebbel <krebbel@linux.ibm.com>
18201 * target.def: in0 and in1 do not need to be registers.
18202 * doc/tm.texi: Regenerate.
18204 2021-07-29 liuhongt <hongtao.liu@intel.com>
18207 * config/i386/i386.c (ix86_widen_mult_cost): New function.
18208 (ix86_add_stmt_cost): Use ix86_widen_mult_cost for
18211 2021-07-29 Jiufu Guo <guojiufu@linux.ibm.com>
18214 * config/rs6000/rs6000.c (TARGET_PREFERRED_DOLOOP_MODE): New hook.
18215 (rs6000_preferred_doloop_mode): New hook.
18216 * doc/tm.texi: Regenerate.
18217 * doc/tm.texi.in: Add hook preferred_doloop_mode.
18218 * target.def (preferred_doloop_mode): New hook.
18219 * targhooks.c (default_preferred_doloop_mode): New hook.
18220 * targhooks.h (default_preferred_doloop_mode): New hook.
18221 * tree-ssa-loop-ivopts.c (compute_doloop_base_on_mode): New function.
18222 (add_iv_candidate_for_doloop): Call targetm.preferred_doloop_mode
18223 and compute_doloop_base_on_mode.
18225 2021-07-28 Martin Sebor <msebor@redhat.com>
18227 PR middle-end/101494
18228 * tree-ssa-uninit.c (maybe_warn_operand): Correct object offset
18229 and size computation.
18231 2021-07-28 Martin Sebor <msebor@redhat.com>
18233 PR middle-end/101601
18234 * gimple-array-bounds.cc (array_bounds_checker::check_mem_ref): Remove
18236 Handle pointers to functions.
18238 2021-07-28 Martin Sebor <msebor@redhat.com>
18240 * Makefile.in (OBJS): Add gimple-ssa-warn-access.o and pointer-query.o.
18241 * attribs.h (fndecl_dealloc_argno): Move fndecl_dealloc_argno to tree.h.
18242 * builtins.c (compute_objsize_r): Move to pointer-query.cc.
18243 (access_ref::access_ref): Same.
18244 (access_ref::phi): Same.
18245 (access_ref::get_ref): Same.
18246 (access_ref::size_remaining): Same.
18247 (access_ref::offset_in_range): Same.
18248 (access_ref::add_offset): Same.
18249 (access_ref::inform_access): Same.
18250 (ssa_name_limit_t::visit_phi): Same.
18251 (ssa_name_limit_t::leave_phi): Same.
18252 (ssa_name_limit_t::next): Same.
18253 (ssa_name_limit_t::next_phi): Same.
18254 (ssa_name_limit_t::~ssa_name_limit_t): Same.
18255 (pointer_query::pointer_query): Same.
18256 (pointer_query::get_ref): Same.
18257 (pointer_query::put_ref): Same.
18258 (pointer_query::flush_cache): Same.
18259 (warn_string_no_nul): Move to gimple-ssa-warn-access.cc.
18260 (check_nul_terminated_array): Same.
18261 (unterminated_array): Same.
18262 (maybe_warn_for_bound): Same.
18263 (check_read_access): Same.
18264 (warn_for_access): Same.
18265 (get_size_range): Same.
18266 (check_access): Same.
18267 (gimple_call_alloc_size): Move to tree.c.
18268 (gimple_parm_array_size): Move to pointer-query.cc.
18269 (get_offset_range): Same.
18270 (gimple_call_return_array): Same.
18271 (handle_min_max_size): Same.
18272 (handle_array_ref): Same.
18273 (handle_mem_ref): Same.
18274 (compute_objsize): Same.
18275 (gimple_call_alloc_p): Move to gimple-ssa-warn-access.cc.
18276 (call_dealloc_argno): Same.
18277 (fndecl_dealloc_argno): Same.
18278 (new_delete_mismatch_p): Same.
18279 (matching_alloc_calls_p): Same.
18280 (warn_dealloc_offset): Same.
18281 (maybe_emit_free_warning): Same.
18282 * builtins.h (check_nul_terminated_array): Move to
18283 gimple-ssa-warn-access.h.
18284 (check_nul_terminated_array): Same.
18285 (warn_string_no_nul): Same.
18286 (unterminated_array): Same.
18287 (class ssa_name_limit_t): Same.
18288 (class pointer_query): Same.
18289 (struct access_ref): Same.
18290 (class range_query): Same.
18291 (struct access_data): Same.
18292 (gimple_call_alloc_size): Same.
18293 (gimple_parm_array_size): Same.
18294 (compute_objsize): Same.
18295 (class access_data): Same.
18296 (maybe_emit_free_warning): Same.
18297 * calls.c (initialize_argument_information): Remove call to
18298 maybe_emit_free_warning.
18299 * gimple-array-bounds.cc: Include new header..
18300 * gimple-fold.c: Same.
18301 * gimple-ssa-sprintf.c: Same.
18302 * gimple-ssa-warn-restrict.c: Same.
18303 * passes.def: Add pass_warn_access.
18304 * tree-pass.h (make_pass_warn_access): Declare.
18305 * tree-ssa-strlen.c: Include new headers.
18306 * tree.c (fndecl_dealloc_argno): Move here from builtins.c.
18307 * tree.h (fndecl_dealloc_argno): Move here from attribs.h.
18308 * gimple-ssa-warn-access.cc: New file.
18309 * gimple-ssa-warn-access.h: New file.
18310 * pointer-query.cc: New file.
18311 * pointer-query.h: New file.
18313 2021-07-28 Jakub Jelinek <jakub@redhat.com>
18315 PR middle-end/101624
18316 * ubsan.c (maybe_instrument_pointer_overflow,
18317 instrument_object_size): Only test DECL_REGISTER on VAR_DECLs,
18318 PARM_DECLs or RESULT_DECLs.
18319 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Likewise.
18321 2021-07-28 Jakub Jelinek <jakub@redhat.com>
18323 PR middle-end/101642
18324 * match.pd (bswap16 (x) == bswap16 (y)): Cast both operands
18325 to type of bswap16 for comparison.
18326 (bswap16 (x) == cst): Cast bswap16 operand to type of cst.
18328 2021-07-28 Richard Biener <rguenther@suse.de>
18330 PR tree-optimization/101615
18331 * tree-vect-slp.c (vect_optimize_slp): Materialize permutes
18332 at CTOR SLP graph entries.
18334 2021-07-28 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18336 * config/aarch64/aarch64.md (*extend<SHORT:mode><GPI:mode>2_aarch64):
18337 Add "r,w" alternative.
18339 2021-07-28 H.J. Lu <hjl.tools@gmail.com>
18342 * config/i386/i386.c (ix86_avx_u128_mode_needed): Don't set
18343 AVX_U128_DIRTY when all bits are zero.
18345 2021-07-28 Richard Biener <rguenther@suse.de>
18347 PR tree-optimization/101615
18348 * tree-vect-slp.c (vect_optimize_slp): Pre-existing vector
18349 external nodes cannot be permuted so make them perm_out 0.
18351 2021-07-28 Andrew Stubbs <ams@codesourcery.com>
18354 * config.in: Regenerate.
18355 * config/gcn/gcn-hsa.h (A_FIJI): New define.
18356 (A_900): New define.
18357 (A_906): New define.
18358 (A_908): New define.
18359 (ASM_SPEC): Use A_FIJI, A_900, A_906 and A_908.
18360 * config/gcn/gcn.c (output_file_start): Adjust attributes according
18361 to the assembler capabilities.
18362 * config/gcn/mkoffload.c (main): Likewise.
18363 * configure: Regenerate.
18364 * configure.ac: Add tests for LLVM assembler attribute features.
18366 2021-07-28 Andrew MacLeod <amacleod@redhat.com>
18368 * gimple-range-gori.cc (gori_compute::outgoing_edge_range_p): Check for
18369 cond_false and cond_true on branches.
18371 2021-07-28 Bin Cheng <bin.cheng@linux.alibaba.com>
18373 * config/aarch64/aarch64.c (aarch64_gen_adjusted_ldpstp): use
18376 2021-07-28 Bin Cheng <bin.cheng@linux.alibaba.com>
18378 * alias.c (init_alias_analysis): Don't skip prologue/epilogue.
18380 2021-07-28 Jakub Jelinek <jakub@redhat.com>
18383 * config/i386/sse.md (vashr<mode>3): Split into vashrv8di3 expander
18384 and vashrv4di3 expander, where the latter requires just TARGET_AVX2
18385 and has special !TARGET_AVX512VL expansion.
18386 (vashrv2di3<mask_name>): Rename to ...
18387 (vashrv2di3): ... this. Change condition to TARGET_XOP || TARGET_AVX2
18388 and add special !TARGET_XOP && !TARGET_AVX512VL expansion.
18390 2021-07-28 Martin Uecker <muecker@gwdg.de>
18392 * calls.c (maybe_warn_rdwr_sizes): Correct argument
18393 numbers in warning that were switched.
18395 2021-07-28 Kewen Lin <linkw@linux.ibm.com>
18397 PR tree-optimization/101596
18398 * tree-vect-patterns.c (vect_recog_mulhs_pattern): Fix wrong check
18399 by using new_type's precision instead.
18401 2021-07-28 liuhongt <hongtao.liu@intel.com>
18404 * config/i386/i386.h (processor_costs): Add new member
18406 * config/i386/x86-tune-costs.h (ix86_size_cost, i386_cost,
18407 i486_cost, pentium_cost, lakemont_cost, pentiumpro_cost,
18408 geode_cost, k6_cost, athlon_cost, k8_cost, amdfam10_cost,
18409 bdver_cost, znver1_cost, znver2_cost, znver3_cost,
18410 btver1_cost, btver2_cost, btver3_cost, pentium4_cost,
18411 nocona_cost, atom_cost, atom_cost, slm_cost, intel_cost,
18412 generic_cost, core_cost): Initialize integer_to_sse same value
18414 (skylake_cost): Initialize integer_to_sse twice as much as sse_op.
18415 * config/i386/i386.c (ix86_builtin_vectorization_cost):
18416 Use integer_to_sse instead of sse_op to calculate the cost of
18419 2021-07-27 Bill Schmidt <wschmidt@linux.ibm.com>
18421 * config/rs6000/rs6000-gen-builtins.c (write_ovld_static_init): New
18423 (write_init_file): Call write_ovld_static_init.
18425 2021-07-27 Bill Schmidt <wschmidt@linux.ibm.com>
18427 * config/rs6000/rs6000-gen-builtins.c (write_bif_static_init): New
18429 (write_init_file): Call write_bif_static_init.
18431 2021-07-27 Bill Schmidt <wschmidt@linux.ibm.com>
18433 * config/rs6000/rs6000-gen-builtins.c (typemap): New struct.
18434 (TYPE_MAP_SIZE): New macro.
18435 (type_map): New initialized variable.
18436 (typemap_cmp): New function.
18437 (write_type_node): Likewise.
18438 (write_fntype_init): Implement.
18440 2021-07-27 Martin Sebor <msebor@redhat.com>
18442 PR tree-optimization/101584
18443 * tree-ssa-uninit.c (builtin_call_nomodifying_p): New function.
18444 (check_defs): Call it.
18446 2021-07-27 Aldy Hernandez <aldyh@redhat.com>
18448 * tree-ssa-dom.c (dom_jump_threader_simplifier):
18449 Put avail_exprs_stack in the class, instead of passing it to
18450 jump_threader_simplifier.
18451 (dom_jump_threader_simplifier::simplify): Add state argument.
18452 (dom_opt_dom_walker): Add state.
18453 (pass_dominator::execute): Pass state to threader.
18454 (dom_opt_dom_walker::before_dom_children): Use state.
18455 * tree-ssa-threadedge.c (jump_threader::jump_threader): Replace
18456 arguments by state.
18457 (jump_threader::record_temporary_equivalences_from_phis):
18458 Register equivalences through the state variable.
18459 (jump_threader::record_temporary_equivalences_from_stmts_at_dest):
18460 Record ranges in a statement through the state variable.
18461 (jump_threader::simplify_control_stmt_condition): Pass state to
18463 (jump_threader::simplify_control_stmt_condition_1): Same.
18464 (jump_threader::thread_around_empty_blocks): Remove obsolete
18466 (jump_threader::thread_through_normal_block): Record equivalences
18467 on edge through the state variable.
18468 (jump_threader::thread_across_edge): Abstract state pushing.
18469 (jt_state::jt_state): New.
18470 (jt_state::push): New.
18471 (jt_state::pop): New.
18472 (jt_state::register_equiv): New.
18473 (jt_state::record_ranges_from_stmt): New.
18474 (jt_state::register_equivs_on_edge): New.
18475 (jump_threader_simplifier::jump_threader_simplifier): Move from
18477 (jump_threader_simplifier::simplify): Add state argument.
18478 * tree-ssa-threadedge.h (class jt_state): New.
18479 (class jump_threader): Add state to constructor.
18480 (class jump_threader_simplifier): Add state to simplify. Remove
18481 avail_exprs_stack from class.
18482 * tree-vrp.c (vrp_jump_threader_simplifier::simplify): Add state
18484 (vrp_jump_threader::vrp_jump_threader): Add state.
18485 (vrp_jump_threader::~vrp_jump_threader): Cleanup state.
18487 2021-07-27 Aldy Hernandez <aldyh@redhat.com>
18489 * Makefile.in (OBJS): Add gimple-range-path.o.
18490 * gimple-range-path.cc: New file.
18491 * gimple-range-path.h: New file.
18493 2021-07-27 Jonathan Wright <jonathan.wright@arm.com>
18495 * config/aarch64/aarch64-simd.md: Push sign/zero-extension
18496 inside vec_duplicate for all patterns.
18497 * simplify-rtx.c (simplify_context::simplify_unary_operation_1):
18498 Push sign/zero-extension inside vec_duplicate.
18500 2021-07-27 Richard Biener <rguenther@suse.de>
18502 PR tree-optimization/101573
18503 * tree-ssa-uninit.c (warn_uninit_phi_uses): New function
18504 looking at uninitialized PHI arg defs in some constrained cases.
18505 (warn_uninitialized_vars): Call it.
18506 (execute_early_warn_uninitialized): Calculate dominators.
18508 2021-07-27 Richard Biener <rguenther@suse.de>
18510 PR tree-optimization/39821
18511 * tree-vect-stmts.c (vect_model_promotion_demotion_cost): Use
18512 vector_stmt for widening arithmetic.
18513 (vectorizable_conversion): Adjust.
18515 2021-07-27 Martin Jambor <mjambor@suse.cz>
18517 * cgraph.h (ipa_replace_map): New field force_load_ref.
18518 * ipa-prop.h (ipa_param_descriptor): Reduce precision of move_cost,
18519 aded new flag load_dereferenced, adjusted comments.
18520 (ipa_get_param_dereferenced): New function.
18521 (ipa_set_param_dereferenced): Likewise.
18522 * cgraphclones.c (cgraph_node::create_virtual_clone): Follow it.
18523 * ipa-cp.c: Include gimple.h.
18524 (ipcp_discover_new_direct_edges): Take into account dereferenced flag.
18525 (get_replacement_map): New parameter force_load_ref, set the
18526 appropriate flag in ipa_replace_map if set.
18527 (struct symbol_and_index_together): New type.
18528 (adjust_refs_in_act_callers): New function.
18529 (adjust_references_in_caller): Likewise.
18530 (create_specialized_node): When appropriate, call
18531 adjust_references_in_caller and force only load references.
18532 * ipa-prop.c (load_from_dereferenced_name): New function.
18533 (ipa_analyze_controlled_uses): Also detect loads from a
18534 dereference, harden testing of call statements.
18535 (ipa_write_node_info): Stream the dereferenced flag.
18536 (ipa_read_node_info): Likewise.
18537 (ipa_set_jf_constant): Also create refdesc when jump function
18538 references a variable.
18539 (cgraph_node_for_jfunc): Rename to symtab_node_for_jfunc, work
18540 also on references of variables and return a symtab_node. Adjust
18542 (propagate_controlled_uses): Also remove references to VAR_DECLs.
18544 2021-07-27 Jakub Jelinek <jakub@redhat.com>
18546 PR middle-end/101586
18547 * gimple-fold.c (clear_padding_type): Ignore FIELD_DECLs with byte
18548 positions above or equal to sz except for diagnostics of flexible
18551 2021-07-26 Andrew MacLeod <amacleod@redhat.com>
18553 PR tree-optimization/78888
18554 * gimple-range-fold.cc (get_letter_range): New.
18555 (fold_using_range::range_of_builtin_call): Call get_letter_range.
18557 2021-07-26 Andrew MacLeod <amacleod@redhat.com>
18559 PR tree-optimization/78888
18560 * gimple-range-fold.cc (fold_using_range::range_of_builtin_call): Add cases
18561 for CFN_BUILT_IN_TOUPPER and CFN_BUILT_IN_TOLOWER.
18563 2021-07-26 Roger Sayle <roger@nextmovesoftware.com>
18564 Marc Glisse <marc.glisse@inria.fr>
18566 * match.pd (rotate): Simplify equality/inequality of rotations.
18567 (bswap): Simplify equality/inequality tests of byte swapping.
18569 2021-07-26 Aldy Hernandez <aldyh@redhat.com>
18571 * range-op.cc (operator_bitwise_xor::op1_op2_relation_effect):
18574 2021-07-26 Aldy Hernandez <aldyh@redhat.com>
18576 * range-op.cc (operator_lshift::fold_range): Pass rel to
18577 base class fold_range.
18578 (operator_rshift::fold_range): Same.
18580 2021-07-26 Ashimida <ashimida@linux.alibaba.com>
18583 * toplev.h (min_align_loops_log): Remove declaration.
18584 (min_align_jumps_log, min_align_labels_log): Likewise.
18585 (min_align_functions_log): Likewise.
18587 2021-07-26 Aldy Hernandez <aldyh@redhat.com>
18589 * tree-vrp.c (vrp_simplify_cond_using_ranges): Rename vr_values
18591 (execute_vrp): Abstract out simplification of conditionals...
18592 (simplify_casted_conds): ...here.
18594 2021-07-26 Aldy Hernandez <aldyh@redhat.com>
18596 * gimple-array-bounds.cc (array_bounds_checker::get_value_range):
18597 Add gimple argument.
18598 (array_bounds_checker::check_array_ref): Same.
18599 (array_bounds_checker::check_addr_expr): Same.
18600 (array_bounds_checker::check_array_bounds): Pass statement to
18601 check_array_bounds and check_addr_expr.
18602 * gimple-array-bounds.h (check_array_bounds): Add gimple argument.
18603 (check_addr_expr): Same.
18604 (get_value_range): Same.
18606 2021-07-26 Tamar Christina <tamar.christina@arm.com>
18608 * config/aarch64/aarch64-simd-builtins.def (sdot, udot): Rename to..
18609 (sdot_prod, udot_prod): ... This.
18610 * config/aarch64/aarch64-simd.md (aarch64_<sur>dot<vsi2qi>): Merged
18612 (<sur>dot_prod<vsi2qi>): ... this.
18613 (aarch64_<sur>dot_lane<vsi2qi>, aarch64_<sur>dot_laneq<vsi2qi>):
18614 Change operands order.
18615 (<sur>sadv16qi): Use new operands order.
18616 * config/aarch64/arm_neon.h (vdot_u32, vdotq_u32, vdot_s32,
18617 vdotq_s32): Use new RTL ordering.
18619 2021-07-26 Tamar Christina <tamar.christina@arm.com>
18621 * config/aarch64/aarch64-builtins.c (TYPES_TERNOP_SUSS,
18622 aarch64_types_ternop_suss_qualifiers): New.
18623 * config/aarch64/aarch64-simd-builtins.def (usdot_prod): Use it.
18624 * config/aarch64/aarch64-simd.md (usdot_prod<vsi2qi>): Re-organize RTL.
18625 * config/aarch64/arm_neon.h (vusdot_s32, vusdotq_s32): Use it.
18627 2021-07-23 Jakub Jelinek <jakub@redhat.com>
18629 PR rtl-optimization/101562
18630 * expmed.c (store_integral_bit_field): Only use movstrict_optab
18631 if the operand isn't paradoxical.
18633 2021-07-23 Aldy Hernandez <aldyh@redhat.com>
18635 * gimple-array-bounds.h (class array_bounds_checker): Change
18636 ranges type to range_query.
18638 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
18640 * config/aarch64/arm_neon.h (vst1_s64_x2): Use
18641 __builtin_memcpy instead of constructing
18642 __builtin_aarch64_simd_oi one vector at a time.
18643 (vst1_u64_x2): Likewise.
18644 (vst1_f64_x2): Likewise.
18645 (vst1_s8_x2): Likewise.
18646 (vst1_p8_x2): Likewise.
18647 (vst1_s16_x2): Likewise.
18648 (vst1_p16_x2): Likewise.
18649 (vst1_s32_x2): Likewise.
18650 (vst1_u8_x2): Likewise.
18651 (vst1_u16_x2): Likewise.
18652 (vst1_u32_x2): Likewise.
18653 (vst1_f16_x2): Likewise.
18654 (vst1_f32_x2): Likewise.
18655 (vst1_p64_x2): Likewise.
18656 (vst1q_s8_x2): Likewise.
18657 (vst1q_p8_x2): Likewise.
18658 (vst1q_s16_x2): Likewise.
18659 (vst1q_p16_x2): Likewise.
18660 (vst1q_s32_x2): Likewise.
18661 (vst1q_s64_x2): Likewise.
18662 (vst1q_u8_x2): Likewise.
18663 (vst1q_u16_x2): Likewise.
18664 (vst1q_u32_x2): Likewise.
18665 (vst1q_u64_x2): Likewise.
18666 (vst1q_f16_x2): Likewise.
18667 (vst1q_f32_x2): Likewise.
18668 (vst1q_f64_x2): Likewise.
18669 (vst1q_p64_x2): Likewise.
18671 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
18673 * config/aarch64/arm_neon.h (vst1_s64_x3): Use
18674 __builtin_memcpy instead of constructing
18675 __builtin_aarch64_simd_ci one vector at a time.
18676 (vst1_u64_x3): Likewise.
18677 (vst1_f64_x3): Likewise.
18678 (vst1_s8_x3): Likewise.
18679 (vst1_p8_x3): Likewise.
18680 (vst1_s16_x3): Likewise.
18681 (vst1_p16_x3): Likewise.
18682 (vst1_s32_x3): Likewise.
18683 (vst1_u8_x3): Likewise.
18684 (vst1_u16_x3): Likewise.
18685 (vst1_u32_x3): Likewise.
18686 (vst1_f16_x3): Likewise.
18687 (vst1_f32_x3): Likewise.
18688 (vst1_p64_x3): Likewise.
18689 (vst1q_s8_x3): Likewise.
18690 (vst1q_p8_x3): Likewise.
18691 (vst1q_s16_x3): Likewise.
18692 (vst1q_p16_x3): Likewise.
18693 (vst1q_s32_x3): Likewise.
18694 (vst1q_s64_x3): Likewise.
18695 (vst1q_u8_x3): Likewise.
18696 (vst1q_u16_x3): Likewise.
18697 (vst1q_u32_x3): Likewise.
18698 (vst1q_u64_x3): Likewise.
18699 (vst1q_f16_x3): Likewise.
18700 (vst1q_f32_x3): Likewise.
18701 (vst1q_f64_x3): Likewise.
18702 (vst1q_p64_x3): Likewise.
18704 2021-07-23 H.J. Lu <hjl.tools@gmail.com>
18707 * config/i386/i386.c (ix86_gen_scratch_sse_rtx): Don't return
18708 hard register when LRA is in progress.
18710 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
18712 * config/aarch64/arm_neon.h (vst1_s8_x4): Use
18713 __builtin_memcpy instead of using a union.
18714 (vst1q_s8_x4): Likewise.
18715 (vst1_s16_x4): Likewise.
18716 (vst1q_s16_x4): Likewise.
18717 (vst1_s32_x4): Likewise.
18718 (vst1q_s32_x4): Likewise.
18719 (vst1_u8_x4): Likewise.
18720 (vst1q_u8_x4): Likewise.
18721 (vst1_u16_x4): Likewise.
18722 (vst1q_u16_x4): Likewise.
18723 (vst1_u32_x4): Likewise.
18724 (vst1q_u32_x4): Likewise.
18725 (vst1_f16_x4): Likewise.
18726 (vst1q_f16_x4): Likewise.
18727 (vst1_f32_x4): Likewise.
18728 (vst1q_f32_x4): Likewise.
18729 (vst1_p8_x4): Likewise.
18730 (vst1q_p8_x4): Likewise.
18731 (vst1_p16_x4): Likewise.
18732 (vst1q_p16_x4): Likewise.
18733 (vst1_s64_x4): Likewise.
18734 (vst1_u64_x4): Likewise.
18735 (vst1_p64_x4): Likewise.
18736 (vst1q_s64_x4): Likewise.
18737 (vst1q_u64_x4): Likewise.
18738 (vst1q_p64_x4): Likewise.
18739 (vst1_f64_x4): Likewise.
18740 (vst1q_f64_x4): Likewise.
18742 2021-07-23 Jonathan Wrightt <jonathan.wright@arm.com>
18744 * config/aarch64/arm_neon.h (vst2_s64): Use __builtin_memcpy
18745 instead of constructing __builtin_aarch64_simd_oi one vector
18747 (vst2_u64): Likewise.
18748 (vst2_f64): Likewise.
18749 (vst2_s8): Likewise.
18750 (vst2_p8): Likewise.
18751 (vst2_s16): Likewise.
18752 (vst2_p16): Likewise.
18753 (vst2_s32): Likewise.
18754 (vst2_u8): Likewise.
18755 (vst2_u16): Likewise.
18756 (vst2_u32): Likewise.
18757 (vst2_f16): Likewise.
18758 (vst2_f32): Likewise.
18759 (vst2_p64): Likewise.
18760 (vst2q_s8): Likewise.
18761 (vst2q_p8): Likewise.
18762 (vst2q_s16): Likewise.
18763 (vst2q_p16): Likewise.
18764 (vst2q_s32): Likewise.
18765 (vst2q_s64): Likewise.
18766 (vst2q_u8): Likewise.
18767 (vst2q_u16): Likewise.
18768 (vst2q_u32): Likewise.
18769 (vst2q_u64): Likewise.
18770 (vst2q_f16): Likewise.
18771 (vst2q_f32): Likewise.
18772 (vst2q_f64): Likewise.
18773 (vst2q_p64): Likewise.
18775 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
18777 * config/aarch64/arm_neon.h (vst3_s64): Use __builtin_memcpy
18778 instead of constructing __builtin_aarch64_simd_ci one vector
18780 (vst3_u64): Likewise.
18781 (vst3_f64): Likewise.
18782 (vst3_s8): Likewise.
18783 (vst3_p8): Likewise.
18784 (vst3_s16): Likewise.
18785 (vst3_p16): Likewise.
18786 (vst3_s32): Likewise.
18787 (vst3_u8): Likewise.
18788 (vst3_u16): Likewise.
18789 (vst3_u32): Likewise.
18790 (vst3_f16): Likewise.
18791 (vst3_f32): Likewise.
18792 (vst3_p64): Likewise.
18793 (vst3q_s8): Likewise.
18794 (vst3q_p8): Likewise.
18795 (vst3q_s16): Likewise.
18796 (vst3q_p16): Likewise.
18797 (vst3q_s32): Likewise.
18798 (vst3q_s64): Likewise.
18799 (vst3q_u8): Likewise.
18800 (vst3q_u16): Likewise.
18801 (vst3q_u32): Likewise.
18802 (vst3q_u64): Likewise.
18803 (vst3q_f16): Likewise.
18804 (vst3q_f32): Likewise.
18805 (vst3q_f64): Likewise.
18806 (vst3q_p64): Likewise.
18808 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
18810 * config/aarch64/arm_neon.h (vst4_s64): Use __builtin_memcpy
18811 instead of constructing __builtin_aarch64_simd_xi one vector
18813 (vst4_u64): Likewise.
18814 (vst4_f64): Likewise.
18815 (vst4_s8): Likewise.
18816 (vst4_p8): Likewise.
18817 (vst4_s16): Likewise.
18818 (vst4_p16): Likewise.
18819 (vst4_s32): Likewise.
18820 (vst4_u8): Likewise.
18821 (vst4_u16): Likewise.
18822 (vst4_u32): Likewise.
18823 (vst4_f16): Likewise.
18824 (vst4_f32): Likewise.
18825 (vst4_p64): Likewise.
18826 (vst4q_s8): Likewise.
18827 (vst4q_p8): Likewise.
18828 (vst4q_s16): Likewise.
18829 (vst4q_p16): Likewise.
18830 (vst4q_s32): Likewise.
18831 (vst4q_s64): Likewise.
18832 (vst4q_u8): Likewise.
18833 (vst4q_u16): Likewise.
18834 (vst4q_u32): Likewise.
18835 (vst4q_u64): Likewise.
18836 (vst4q_f16): Likewise.
18837 (vst4q_f32): Likewise.
18838 (vst4q_f64): Likewise.
18839 (vst4q_p64): Likewise.
18841 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
18843 * config/aarch64/arm_neon.h (vtbx4_s8): Use __builtin_memcpy
18844 instead of constructing __builtin_aarch64_simd_oi one vector
18846 (vtbx4_u8): Likewise.
18847 (vtbx4_p8): Likewise.
18849 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
18851 * config/aarch64/arm_neon.h (vtbl3_s8): Use __builtin_memcpy
18852 instead of constructing __builtin_aarch64_simd_oi one vector
18854 (vtbl3_u8): Likewise.
18855 (vtbl3_p8): Likewise.
18856 (vtbl4_s8): Likewise.
18857 (vtbl4_u8): Likewise.
18858 (vtbl4_p8): Likewise.
18860 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
18862 * config/aarch64/arm_neon.h (vqtbx2_s8): Use __builtin_memcpy
18863 instead of constructing __builtin_aarch64_simd_oi one vector
18865 (vqtbx2_u8): Likewise.
18866 (vqtbx2_p8): Likewise.
18867 (vqtbx2q_s8): Likewise.
18868 (vqtbx2q_u8): Likewise.
18869 (vqtbx2q_p8): Likewise.
18870 (vqtbx3_s8): Use __builtin_memcpy instead of constructing
18871 __builtin_aarch64_simd_ci one vector at a time.
18872 (vqtbx3_u8): Likewise.
18873 (vqtbx3_p8): Likewise.
18874 (vqtbx3q_s8): Likewise.
18875 (vqtbx3q_u8): Likewise.
18876 (vqtbx3q_p8): Likewise.
18877 (vqtbx4_s8): Use __builtin_memcpy instead of constructing
18878 __builtin_aarch64_simd_xi one vector at a time.
18879 (vqtbx4_u8): Likewise.
18880 (vqtbx4_p8): Likewise.
18881 (vqtbx4q_s8): Likewise.
18882 (vqtbx4q_u8): Likewise.
18883 (vqtbx4q_p8): Likewise.
18885 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
18887 * config/aarch64/arm_neon.h (vqtbl2_s8): Use __builtin_memcpy
18888 instead of constructing __builtin_aarch64_simd_oi one vector
18890 (vqtbl2_u8): Likewise.
18891 (vqtbl2_p8): Likewise.
18892 (vqtbl2q_s8): Likewise.
18893 (vqtbl2q_u8): Likewise.
18894 (vqtbl2q_p8): Likewise.
18895 (vqtbl3_s8): Use __builtin_memcpy instead of constructing
18896 __builtin_aarch64_simd_ci one vector at a time.
18897 (vqtbl3_u8): Likewise.
18898 (vqtbl3_p8): Likewise.
18899 (vqtbl3q_s8): Likewise.
18900 (vqtbl3q_u8): Likewise.
18901 (vqtbl3q_p8): Likewise.
18902 (vqtbl4_s8): Use __builtin_memcpy instead of constructing
18903 __builtin_aarch64_simd_xi one vector at a time.
18904 (vqtbl4_u8): Likewise.
18905 (vqtbl4_p8): Likewise.
18906 (vqtbl4q_s8): Likewise.
18907 (vqtbl4q_u8): Likewise.
18908 (vqtbl4q_p8): Likewise.
18910 2021-07-23 Haochen Gui <guihaoc@gcc.gnu.org>
18913 * config/rs6000/rs6000.md (cstore<mode>4): Fix wrong fall through.
18915 2021-07-22 Andrew Pinski <apinski@marvell.com>
18917 PR tree-optimization/10153
18918 * tree-tailcall.c (create_tailcall_accumulator):
18919 Don't call fold_convert as the type should be correct already.
18920 (tree_optimize_tail_calls_1): Use build_{one,zero}_cst instead
18921 of integer_{one,zero}_node for the call of create_tailcall_accumulator.
18923 2021-07-22 Aldy Hernandez <aldyh@redhat.com>
18925 * gimple-range-cache.cc (non_null_ref::adjust_range): Replace
18926 varying_p check for null/non-null check.
18928 2021-07-22 Andrew MacLeod <amacleod@redhat.com>
18930 PR tree-optimization/101511
18931 * value-relation.cc (relation_oracle::query_relation): Check if ssa1
18932 is in ssa2's equiv set, and don't trap if so.
18934 2021-07-22 Andrew MacLeod <amacleod@redhat.com>
18936 PR tree-optimization/101497
18937 * gimple-range-fold.cc (fold_using_range::range_of_cond_expr): Check
18940 2021-07-22 Andrew MacLeod <amacleod@redhat.com>
18942 PR tree-optimization/101496
18943 * vr-values.c (simplify_using_ranges::fold_cond): Call range_of_stmt
18944 first, then vrp_visit_cond_Stmt.
18946 2021-07-22 liuhongt <hongtao.liu@intel.com>
18948 * config/i386/i386-expand.c
18949 (ix86_broadcast_from_integer_constant): Rename to ..
18950 (ix86_broadcast_from_constant): .. this, and extend it to
18952 (ix86_expand_vector_move): Extend to float mode.
18953 * config/i386/i386-features.c
18954 (replace_constant_pool_with_broadcast): Remove.
18955 (remove_partial_avx_dependency_gate): Ditto.
18956 (constant_pool_broadcast): Ditto.
18957 (class pass_constant_pool_broadcast): Ditto.
18958 (make_pass_constant_pool_broadcast): Ditto.
18959 (remove_partial_avx_dependency): Adjust gate.
18960 * config/i386/i386-passes.def: Remove pass_constant_pool_broadcast.
18961 * config/i386/i386-protos.h
18962 (make_pass_constant_pool_broadcast): Remove.
18964 2021-07-22 liuhongt <hongtao.liu@intel.com>
18966 * config/i386/constraints.md (Wb): New constraint.
18968 * config/i386/i386.md (*ashlhi3_1): Extend to avx512 mask
18970 (*ashlqi3_1): Ditto.
18971 (*<insn><mode>3_1): Split to ..
18972 (*ashr<mode>3_1): this, ...
18973 (*lshr<mode>3_1): and this, also extend this pattern to avx512
18975 (*<insn><mode>3_1): Split to ..
18976 (*ashr<mode>3_1): this, ...
18977 (*lshrqi3_1): and this, also extend this pattern to avx512
18979 (*lshrhi3_1): And this, also extend this pattern to avx512
18981 * config/i386/sse.md (k<code><mode>): New define_split after
18982 it to convert generic shift pattern to mask shift ones.
18984 2021-07-21 Thomas Schwinge <thomas@codesourcery.com>
18985 Joseph Myers <joseph@codesourcery.com>
18986 Cesar Philippidis <cesar@codesourcery.com>
18988 * tree-core.h (omp_clause_code): Add 'OMP_CLAUSE_NOHOST'.
18989 * tree.c (omp_clause_num_ops, omp_clause_code_name, walk_tree_1):
18991 * tree-pretty-print.c (dump_omp_clause): Likewise.
18992 * omp-general.c (oacc_verify_routine_clauses): Likewise.
18993 * gimplify.c (gimplify_scan_omp_clauses)
18994 (gimplify_adjust_omp_clauses): Likewise.
18995 * tree-nested.c (convert_nonlocal_omp_clauses)
18996 (convert_local_omp_clauses): Likewise.
18997 * omp-low.c (scan_sharing_clauses): Likewise.
18998 * omp-offload.c (execute_oacc_device_lower): Update.
19000 2021-07-21 Martin Sebor <msebor@redhat.com>
19002 * tree-ssa-alias.c (walk_aliased_vdefs_1): Fix typos in a comment.
19004 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
19006 * config/rs6000/rs6000-gen-builtins.c (write_init_bif_table):
19009 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
19011 * config/rs6000/rs6000-gen-builtins.c (write_fntype): New
19013 (write_fntype_init): New stub function.
19014 (write_init_bif_table): Likewise.
19015 (write_init_ovld_table): New function.
19016 (write_init_file): Implement.
19018 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
19020 * config/rs6000/rs6000-gen-builtins.c
19021 (write_autogenerated_header): New function.
19022 (write_decls): Likewise.
19023 (write_extern_fntype): New callback function.
19024 (write_header_file): Implement.
19026 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
19028 * config/rs6000/rs6000-gen-builtins.c (write_defines_file):
19031 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
19033 * config/rs6000/rs6000-gen-builtins.c (complete_vector_type): New
19035 (complete_base_type): Likewise.
19036 (construct_fntype_id): Likewise.
19037 (parse_bif_entry): Call contruct_fntype_id.
19038 (parse_ovld_entry): Likewise.
19040 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
19042 * config/rs6000/rs6000-gen-builtins.c (ovld_stanza): New struct.
19043 (MAXOVLDSTANZAS): New macro.
19044 (ovld_stanzas): New variable.
19045 (curr_ovld_stanza): Likewise.
19046 (MAXOVLDS): New macro.
19047 (ovlddata): New struct.
19048 (ovlds): New variable.
19049 (curr_ovld): Likewise.
19050 (max_ovld_args): Likewise.
19051 (parse_ovld_entry): New function.
19052 (parse_ovld_stanza): Likewise.
19053 (parse_ovld): Implement.
19055 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
19057 * config/rs6000/rs6000-gen-builtins.c (parse_bif_attrs):
19060 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
19062 * config/rs6000/rs6000-gen-builtins.c (parse_args): New function.
19063 (parse_prototype): Implement.
19065 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
19067 * config/rs6000/rs6000-gen-builtins.c (bif_stanza): New enum.
19068 (curr_bif_stanza): New variable.
19069 (stanza_entry): New struct.
19070 (stanza_map): New initialized variable.
19071 (enable_string): Likewise.
19072 (fnkinds): New enum.
19073 (typelist): New struct.
19074 (attrinfo): Likewise.
19075 (MAXRESTROPNDS): New macro.
19076 (prototype): New struct.
19077 (MAXBIFS): New macro.
19078 (bifdata): New struct.
19079 (bifs): New variable.
19080 (curr_bif): Likewise.
19081 (bif_order): Likewise.
19082 (bif_index): Likewise.
19083 (fatal): New function.
19084 (stanza_name_to_stanza): Likewise.
19085 (parse_bif_attrs): New stub function.
19086 (parse_prototype): Likewise.
19087 (parse_bif_entry): New function.
19088 (parse_bif_stanza): Likewise.
19089 (parse_bif): Implement.
19090 (set_bif_order): New function.
19091 (create_bif_order): Implement.
19093 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
19095 * config/rs6000/rs6000-gen-builtins.c (rbtree.h): New #include.
19096 (num_bifs): New variable.
19097 (num_ovld_stanzas): Likewise.
19098 (num_ovlds): Likewise.
19099 (parse_codes): New enum.
19100 (bif_rbt): New variable.
19101 (ovld_rbt): Likewise.
19102 (fntype_rbt): Likewise.
19103 (bifo_rbt): Likewise.
19104 (parse_bif): New stub function.
19105 (create_bif_order): Likewise.
19106 (parse_ovld): Likewise.
19107 (write_header_file): Likewise.
19108 (write_init_file): Likewise.
19109 (write_defines_file): Likewise.
19110 (delete_output_files): New function.
19113 2021-07-21 H.J. Lu <hjl.tools@gmail.com>
19116 * config/i386/i386-builtin.def: Remove OPTION_MASK_ISA_SSE4_2
19117 from CRC32 _builtin functions.
19119 2021-07-21 Sebastian Huber <sebastian.huber@embedded-brains.de>
19121 * coverage.c (build_gcov_info_var_registration): Mark the object placed
19122 in the linker set as referenced so that it does not get optimized away.
19124 2021-07-21 Kito Cheng <kito.cheng@sifive.com>
19127 2021-07-20 Kito Cheng <kito.cheng@sifive.com>
19129 * config.gcc (riscv*-*-*): Detect which python is available.
19131 2021-07-21 Jakub Jelinek <jakub@redhat.com>
19133 PR middle-end/101535
19134 * gimplify.c (omp_check_private): Properly skip ORT_TARGET_DATA
19135 contexts in which decl isn't privatized and for ORT_TARGET return
19136 false if decl is mapped.
19138 2021-07-21 Richard Sandiford <richard.sandiford@arm.com>
19140 * gimple-loop-jam.c: Include tree-ssa-sccvn.h.
19141 (tree_loop_unroll_and_jam): Run value-numbering on a loop that
19142 has been successfully unrolled.
19144 2021-07-21 Richard Sandiford <richard.sandiford@arm.com>
19146 * tree-ssa-loop-manip.c (determine_exit_conditions): Return a null
19147 exit condition if no tail loop is needed, and if the original exit
19148 condition should therefore be kept as-is.
19149 (tree_transform_and_unroll_loop): Handle that case here too.
19151 2021-07-21 Kewen Lin <linkw@linux.ibm.com>
19153 * tree-data-ref.c (free_dependence_relations): Adjust to pass vec
19155 (free_data_refs): Likewise.
19156 * tree-data-ref.h (free_dependence_relations): Likewise.
19157 (free_data_refs): Likewise.
19158 * tree-predcom.c (struct chain): Use auto_vec instead of vec for
19160 (struct component): Likewise.
19161 (pcom_worker::pcom_worker): Adjust for auto_vec and renaming changes.
19162 (pcom_worker::~pcom_worker): Likewise.
19163 (pcom_worker::release_chain): Adjust as auto_vec changes.
19164 (pcom_worker::loop): Rename to ...
19165 (pcom_worker::m_loop): ... this.
19166 (pcom_worker::datarefs): Rename to ...
19167 (pcom_worker::m_datarefs): ... this. Use auto_vec instead of vec.
19168 (pcom_worker::dependences): Rename to ...
19169 (pcom_worker::m_dependences): ... this. Use auto_vec instead of vec.
19170 (pcom_worker::chains): Rename to ...
19171 (pcom_worker::m_chains): ... this. Use auto_vec instead of vec.
19172 (pcom_worker::looparound_phis): Rename to ...
19173 (pcom_worker::m_looparound_phis): ... this. Use auto_vec instead of
19175 (pcom_worker::cache): Rename to ...
19176 (pcom_worker::m_cache): ... this. Use auto_vec instead of vec.
19177 (pcom_worker::release_chain): Adjust for auto_vec changes.
19178 (pcom_worker::release_chains): Adjust for auto_vec and renaming
19180 (release_component): Remove.
19181 (release_components): Adjust for release_component removal.
19182 (component_of): Adjust to use vec.
19183 (merge_comps): Likewise.
19184 (pcom_worker::aff_combination_dr_offset): Adjust for renaming changes.
19185 (pcom_worker::determine_offset): Likewise.
19186 (class comp_ptrs): Remove.
19187 (pcom_worker::split_data_refs_to_components): Adjust for renaming
19188 changes, for comp_ptrs removal with auto_vec.
19189 (pcom_worker::suitable_component_p): Adjust for renaming changes.
19190 (pcom_worker::filter_suitable_components): Adjust for release_component
19192 (pcom_worker::valid_initializer_p): Adjust for renaming changes.
19193 (pcom_worker::find_looparound_phi): Likewise.
19194 (pcom_worker::add_looparound_copies): Likewise.
19195 (pcom_worker::determine_roots_comp): Likewise.
19196 (pcom_worker::single_nonlooparound_use): Likewise.
19197 (pcom_worker::execute_pred_commoning_chain): Likewise.
19198 (pcom_worker::execute_pred_commoning): Likewise.
19199 (pcom_worker::try_combine_chains): Likewise.
19200 (pcom_worker::prepare_initializers_chain): Likewise.
19201 (pcom_worker::prepare_initializers): Likewise.
19202 (pcom_worker::prepare_finalizers_chain): Likewise.
19203 (pcom_worker::prepare_finalizers): Likewise.
19204 (pcom_worker::tree_predictive_commoning_loop): Likewise.
19206 2021-07-20 Martin Sebor <msebor@redhat.com>
19208 PR middle-end/101397
19209 * builtins.c (gimple_call_return_array): Add argument. Correct
19210 offsets for memchr, mempcpy, stpcpy, and stpncpy.
19211 (compute_objsize_r): Adjust offset computation for argument returning
19214 2021-07-20 Martin Sebor <msebor@redhat.com>
19216 PR middle-end/101300
19217 * tree-ssa-uninit.c (check_defs): Handle UBSAN built-ins.
19219 2021-07-20 Jeff Law <jlaw@localhost.localdomain>
19221 * function.c (assign_parm_setup_block): Use adjust_address instead
19222 of change_address to preserve MEM_EXPR and friends.
19224 2021-07-20 Martin Sebor <msebor@redhat.com>
19226 * cfgloop.h (single_likely_exit): Adjust by-value argument to
19227 by-const-reference.
19228 * cfgloopanal.c (single_likely_exit): Same.
19229 * cgraph.h (struct cgraph_node): Same.
19230 * cgraphclones.c (cgraph_node::create_virtual_clone): Same.
19231 * genautomata.c (merge_states): Same.
19232 * genextract.c (VEC_char_to_string): Same.
19233 * genmatch.c (dt_node::gen_kids_1): Same.
19234 (walk_captures): Adjust by-value argument to by-reference.
19235 * gimple-ssa-store-merging.c (check_no_overlap): Adjust by-value argument
19236 to by-const-reference.
19237 * gimple.c (gimple_build_call_vec): Same.
19238 (gimple_build_call_internal_vec): Same.
19239 (gimple_build_switch): Same.
19240 (sort_case_labels): Same.
19241 (preprocess_case_label_vec_for_gimple): Adjust by-value argument to
19243 * gimple.h (gimple_build_call_vec): Adjust by-value argument to
19244 by-const-reference.
19245 (gimple_build_call_internal_vec): Same.
19246 (gimple_build_switch): Same.
19247 (sort_case_labels): Same.
19248 (preprocess_case_label_vec_for_gimple): Adjust by-value argument to
19250 * haifa-sched.c (calc_priorities): Adjust by-value argument to
19251 by-const-reference.
19252 (sched_init_luids): Same.
19253 (haifa_init_h_i_d): Same.
19254 * ipa-cp.c (ipa_get_indirect_edge_target_1): Same.
19255 (adjust_callers_for_value_intersection): Adjust by-value argument to
19257 (find_more_scalar_values_for_callers_subset): Adjust by-value argument to
19258 by-const-reference.
19259 (find_more_contexts_for_caller_subset): Same.
19260 (find_aggregate_values_for_callers_subset): Same.
19261 (copy_useful_known_contexts): Same.
19262 * ipa-fnsummary.c (remap_edge_summaries): Same.
19263 (remap_freqcounting_predicate): Same.
19264 * ipa-inline.c (add_new_edges_to_heap): Adjust by-value argument to
19266 * ipa-predicate.c (predicate::remap_after_inlining): Adjust by-value argument
19267 to by-const-reference.
19268 * ipa-predicate.h (predicate::remap_after_inlining): Same.
19269 * ipa-prop.c (ipa_find_agg_cst_for_param): Same.
19270 * ipa-prop.h (ipa_find_agg_cst_for_param): Same.
19271 * ira-build.c (ira_loop_tree_body_rev_postorder): Same.
19272 * read-rtl.c (add_overload_instance): Same.
19273 * rtl.h (native_decode_rtx): Same.
19274 (native_decode_vector_rtx): Same.
19275 * sched-int.h (sched_init_luids): Same.
19276 (haifa_init_h_i_d): Same.
19277 * simplify-rtx.c (native_decode_vector_rtx): Same.
19278 (native_decode_rtx): Same.
19279 * tree-call-cdce.c (gen_shrink_wrap_conditions): Same.
19280 (shrink_wrap_one_built_in_call_with_conds): Same.
19281 (shrink_wrap_conditional_dead_built_in_calls): Same.
19282 * tree-data-ref.c (create_runtime_alias_checks): Same.
19283 (compute_all_dependences): Same.
19284 * tree-data-ref.h (compute_all_dependences): Same.
19285 (create_runtime_alias_checks): Same.
19286 (index_in_loop_nest): Same.
19287 * tree-if-conv.c (mask_exists): Same.
19288 * tree-loop-distribution.c (class loop_distribution): Same.
19289 (loop_distribution::create_rdg_vertices): Same.
19290 (dump_rdg_partitions): Same.
19291 (debug_rdg_partitions): Same.
19292 (partition_contains_all_rw): Same.
19293 (loop_distribution::distribute_loop): Same.
19294 * tree-parloops.c (oacc_entry_exit_ok_1): Same.
19295 (oacc_entry_exit_single_gang): Same.
19296 * tree-ssa-loop-im.c (hoist_memory_references): Same.
19297 (loop_suitable_for_sm): Same.
19298 * tree-ssa-loop-niter.c (bound_index): Same.
19299 * tree-ssa-reassoc.c (update_ops): Same.
19300 (swap_ops_for_binary_stmt): Same.
19301 (rewrite_expr_tree): Same.
19302 (rewrite_expr_tree_parallel): Same.
19303 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Same.
19304 * tree-ssa-sccvn.h (ao_ref_init_from_vn_reference): Same.
19305 * tree-ssa-structalias.c (process_all_all_constraints): Same.
19306 (make_constraints_to): Same.
19307 (handle_lhs_call): Same.
19308 (find_func_aliases_for_builtin_call): Same.
19309 (sort_fieldstack): Same.
19310 (check_for_overlaps): Same.
19311 * tree-vect-loop-manip.c (vect_create_cond_for_align_checks): Same.
19312 (vect_create_cond_for_unequal_addrs): Same.
19313 (vect_create_cond_for_lower_bounds): Same.
19314 (vect_create_cond_for_alias_checks): Same.
19315 * tree-vect-slp-patterns.c (vect_validate_multiplication): Same.
19316 * tree-vect-slp.c (vect_analyze_slp_instance): Same.
19317 (vect_make_slp_decision): Same.
19318 (vect_slp_bbs): Same.
19319 (duplicate_and_interleave): Same.
19320 (vect_transform_slp_perm_load): Same.
19321 (vect_schedule_slp): Same.
19322 * tree-vectorizer.h (vect_transform_slp_perm_load): Same.
19323 (vect_schedule_slp): Same.
19324 (duplicate_and_interleave): Same.
19325 * tree.c (build_vector_from_ctor): Same.
19326 (build_vector): Same.
19327 (check_vector_cst): Same.
19328 (check_vector_cst_duplicate): Same.
19329 (check_vector_cst_fill): Same.
19330 (check_vector_cst_stepped): Same.
19331 * tree.h (build_vector_from_ctor): Same.
19333 2021-07-20 Jakub Jelinek <jakub@redhat.com>
19336 * config/rs6000/rs6000-protos.h (easy_altivec_constant): Change return
19337 type from bool to int.
19338 * config/rs6000/rs6000.c (vspltis_constant): Fix up handling the
19339 EASY_VECTOR_MSB case if either step or copies is not 1.
19340 (vspltis_shifted): Fix comment typo.
19341 (easy_altivec_constant): Change return type from bool to int, instead
19342 of returning true return byte size of the element mode that should be
19343 used to synthetize the constant.
19344 * config/rs6000/predicates.md (easy_vector_constant_msb): Require
19345 that vspltis_shifted is 0, handle the case where easy_altivec_constant
19346 assumes using different vector mode from CONST_VECTOR's mode.
19347 * config/rs6000/altivec.md (easy_vector_constant_msb splitter): Use
19348 easy_altivec_constant to determine mode in which -1 >> -1 should be
19349 performed, use rs6000_expand_vector_init instead of gen_vec_initv4sisi.
19351 2021-07-20 Richard Biener <rguenther@suse.de>
19354 * dwarf2out.h (dwarf_file_data): Add key member.
19355 * dwarf2out.c (dwarf_file_hasher::equal): Compare key.
19356 (dwarf_file_hasher::hash): Hash key.
19357 (lookup_filename): Remap the filename and store it in the
19358 filename member of dwarf_file_data when creating a new
19360 (file_name_acquire): Do not remap the filename again.
19361 (maybe_emit_file): Likewise.
19363 2021-07-20 Jonathan Wright <jonathan.wright@arm.com>
19365 * config/aarch64/aarch64-simd-builtins.def: Use two variant
19366 generators for all TBL/TBX intrinsics and rename to
19367 consistent forms: qtbl[1234] or qtbx[1234].
19368 * config/aarch64/aarch64-simd.md (aarch64_tbl1<mode>):
19370 (aarch64_qtbl1<mode>): This.
19371 (aarch64_tbx1<mode>): Rename to...
19372 (aarch64_qtbx1<mode>): This.
19373 (aarch64_tbl2v16qi): Delete.
19374 (aarch64_tbl3<mode>): Rename to...
19375 (aarch64_qtbl2<mode>): This.
19376 (aarch64_tbx4<mode>): Rename to...
19377 (aarch64_qtbx2<mode>): This.
19378 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_1): Use
19379 renamed qtbl1 and qtbl2 RTL patterns.
19380 * config/aarch64/arm_neon.h (vqtbl1_p8): Use renamed qtbl1
19382 (vqtbl1_s8): Likewise.
19383 (vqtbl1_u8): Likewise.
19384 (vqtbl1q_p8): Likewise.
19385 (vqtbl1q_s8): Likewise.
19386 (vqtbl1q_u8): Likewise.
19387 (vqtbx1_s8): Use renamed qtbx1 RTL pattern.
19388 (vqtbx1_u8): Likewise.
19389 (vqtbx1_p8): Likewise.
19390 (vqtbx1q_s8): Likewise.
19391 (vqtbx1q_u8): Likewise.
19392 (vqtbx1q_p8): Likewise.
19393 (vtbl1_s8): Use renamed qtbl1 RTL pattern.
19394 (vtbl1_u8): Likewise.
19395 (vtbl1_p8): Likewise.
19396 (vtbl2_s8): Likewise
19397 (vtbl2_u8): Likewise.
19398 (vtbl2_p8): Likewise.
19399 (vtbl3_s8): Use renamed qtbl2 RTL pattern.
19400 (vtbl3_u8): Likewise.
19401 (vtbl3_p8): Likewise.
19402 (vtbl4_s8): Likewise.
19403 (vtbl4_u8): Likewise.
19404 (vtbl4_p8): Likewise.
19405 (vtbx2_s8): Use renamed qtbx2 RTL pattern.
19406 (vtbx2_u8): Likewise.
19407 (vtbx2_p8): Likewise.
19408 (vqtbl2_s8): Use renamed qtbl2 RTL pattern.
19409 (vqtbl2_u8): Likewise.
19410 (vqtbl2_p8): Likewise.
19411 (vqtbl2q_s8): Likewise.
19412 (vqtbl2q_u8): Likewise.
19413 (vqtbl2q_p8): Likewise.
19414 (vqtbx2_s8): Use renamed qtbx2 RTL pattern.
19415 (vqtbx2_u8): Likewise.
19416 (vqtbx2_p8): Likewise.
19417 (vqtbx2q_s8): Likewise.
19418 (vqtbx2q_u8): Likewise.
19419 (vqtbx2q_p8): Likewise.
19420 (vtbx4_s8): Likewise.
19421 (vtbx4_u8): Likewise.
19422 (vtbx4_p8): Likewise.
19424 2021-07-20 Uroš Bizjak <ubizjak@gmail.com>
19427 * config/i386/sync.md (define_peephole2 atomic_storedi_fpu):
19429 (define_peephole2 atomic_loaddi_fpu): Ditto.
19431 2021-07-20 Kito Cheng <kito.cheng@sifive.com>
19433 * config.gcc (riscv*-*-*): Detect which python is available.
19435 2021-07-20 Kewen Lin <linkw@linux.ibm.com>
19437 * config/rs6000/vsx.md (mulhs_<mode>): Rename to...
19438 (smul<mode>3_highpart): ... this.
19439 (mulhu_<mode>): Rename to...
19440 (umul<mode>3_highpart): ... this.
19441 * config/rs6000/rs6000-builtin.def (MULHS_V2DI, MULHS_V4SI,
19442 MULHU_V2DI, MULHU_V4SI): Adjust.
19444 2021-07-20 Kewen Lin <linkw@linux.ibm.com>
19446 PR tree-optimization/100696
19447 * internal-fn.c (first_commutative_argument): Add info for IFN_MULH.
19448 * internal-fn.def (IFN_MULH): New internal function.
19449 * tree-vect-patterns.c (vect_recog_mulhs_pattern): Add support to
19450 recog normal multiply highpart as IFN_MULH.
19451 * config/i386/i386.c (ix86_add_stmt_cost): Adjust for combined
19454 2021-07-19 Indu Bhagat <indu.bhagat@oracle.com>
19456 * config/elfos.h (CTF_DEBUGGING_INFO): New definition.
19457 (BTF_DEBUGGING_INFO): Likewise.
19458 * doc/tm.texi.in: Document the new macros.
19459 * doc/tm.texi: Regenerated.
19460 * toplev.c: Guard initialization of debug hooks.
19462 2021-07-19 Indu Bhagat <indu.bhagat@oracle.com>
19464 * flags.h (ctf_debuginfo_p): New function declaration.
19465 * opts.c (ctf_debuginfo_p): New function definition.
19467 2021-07-19 Andrew Stubbs <ams@codesourcery.com>
19470 * config/gcn/gcn-hsa.h (DRIVER_SELF_SPECS): New.
19471 (ASM_SPEC): Set -mattr for xnack and sram-ecc.
19472 * config/gcn/gcn-opts.h (enum sram_ecc_type): New.
19473 * config/gcn/gcn-valu.md: Add a warning comment.
19474 * config/gcn/gcn.c (gcn_option_override): Add "sorry" for -mxnack.
19475 (output_file_start): Add xnack and sram-ecc state to ".amdgcn_target".
19476 * config/gcn/gcn.md: Add a warning comment.
19477 * config/gcn/gcn.opt: Add -mxnack and -msram-ecc.
19478 * config/gcn/mkoffload.c (EF_AMDGPU_MACH_AMDGCN_GFX908): Remove
19480 (EF_AMDGPU_XNACK): New.
19481 (EF_AMDGPU_SRAM_ECC): New.
19483 (copy_early_debug_info): Use elf_flags.
19484 (main): Handle -mxnack and -msram-ecc options.
19485 * doc/invoke.texi: Document -mxnack and -msram-ecc.
19487 2021-07-19 Andrew Pinski <apinski@marvell.com>
19490 * config/aarch64/aarch64.md (csneg3_uxtw_insn): Rename to ...
19491 (*cs<neg_not_cs>3_uxtw_insn4): and extend to NEG_NOT.
19493 2021-07-19 Richard Biener <rguenther@suse.de>
19495 PR tree-optimization/101505
19496 * tree-vect-patterns.c (vect_determine_precisions): Walk
19497 PHIs also for loop vectorization.
19499 2021-07-19 Richard Biener <rguenther@suse.de>
19501 * gimple.h (gimple_expr_type): Remove.
19502 * doc/gimple.texi: Remove gimple_expr_type documentation.
19504 2021-07-19 Richard Biener <rguenther@suse.de>
19506 * tree-ssa-sccvn.c (vn_reference_eq): Handle NULL vr->type.
19507 (ao_ref_init_from_vn_reference): Likewise.
19508 (fully_constant_reference): Likewise.
19509 (vn_reference_lookup_call): Do not set vr->type to random
19511 * tree-ssa-pre.c (compute_avail): Do not try to PRE calls
19513 * tree-vect-generic.c (expand_vector_piecewise): Pass in
19514 whether we expanded parallel.
19515 (expand_vector_parallel): Adjust.
19516 (expand_vector_addition): Likewise.
19517 (expand_vector_comparison): Likewise.
19518 (expand_vector_operation): Likewise.
19519 (expand_vector_scalar_condition): Likewise.
19520 (expand_vector_conversion): Likewise.
19522 2021-07-19 Richard Biener <rguenther@suse.de>
19524 * tree-vrp.c (register_edge_assert_for_2): Use the
19526 (vrp_folder::fold_predicate_in): Likewise.
19527 * vr-values.c (gimple_assign_nonzero_p): Likewise.
19528 (vr_values::extract_range_from_comparison): Likewise.
19529 (vr_values::extract_range_from_ubsan_builtin): Use the
19530 type of the first operand.
19531 (vr_values::extract_range_basic): Push down type
19532 computation, use the appropriate LHS.
19533 (vr_values::extract_range_from_assignment): Use the
19536 2021-07-18 H.J. Lu <hjl.tools@gmail.com>
19539 * common/config/i386/i386-common.c (ix86_handle_option): For
19540 -mgeneral-regs-only, enable the GPR only instructions which are
19541 enabled implicitly by SSE ISAs unless they have been disabled
19544 2021-07-18 H.J. Lu <hjl.tools@gmail.com>
19547 * config/i386/i386.c (ix86_check_avx_upper_stores): Moved before
19548 ix86_avx_u128_mode_needed.
19549 (ix86_avx_u128_mode_needed): Return AVX_U128_DIRTY if callee
19550 returns AVX register.
19552 2021-07-17 Jan Hubicka <hubicka@ucw.cz>
19554 * tree-ssa-structalias.c (handle_rhs_call): Support EAF_NOT_RETURNED.
19555 (handle_const_call): Liekise
19556 (handle_pure_call): Liekise
19558 2021-07-17 Andrew MacLeod <amacleod@redhat.com>
19560 PR tree-optimization/96542
19561 * range-op.cc (range_operator::wi_fold_in_parts): New.
19562 (range_operator::fold_range): Call wi_fold_in_parts.
19563 (operator_lshift::wi_fold): Fix broken lshift by [0,0].
19564 * range-op.h (wi_fold_in_parts): Add prototype.
19566 2021-07-16 David Malcolm <dmalcolm@redhat.com>
19568 * doc/analyzer.texi: Add __analyzer_dump_state.
19570 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
19572 * config/rs6000/rbtree.c: New file.
19573 * config/rs6000/rbtree.h: New file.
19575 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
19577 * config/rs6000/rs6000-gen-builtins.c (restriction): New enum.
19578 (typeinfo): Add restr field.
19579 (match_bracketed_pair): New function.
19580 (match_const_restriction): Implement.
19582 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
19584 * config/rs6000/rs6000-gen-builtins.c (match_basetype): Implement.
19586 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
19588 * config/rs6000/rs6000-gen-builtins.c (void_status): New enum.
19589 (basetype): Likewise.
19590 (typeinfo): Likewise.
19591 (handle_pointer): New function.
19592 (match_basetype): New stub function.
19593 (match_const_restriction): Likewise.
19594 (match_type): New function.
19596 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
19598 * config/rs6000/rs6000-gen-builtins.c (consume_whitespace): New
19600 (advance_line): Likewise.
19601 (safe_inc_pos): Likewise.
19602 (match_identifier): Likewise.
19603 (match_integer): Likewise.
19604 (match_to_right_bracket): Likewise.
19606 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
19608 * config/rs6000/rs6000-gen-builtins.c (bif_file): New variable.
19609 (ovld_file): Likewise.
19610 (header_file): Likewise.
19611 (init_file): Likewise.
19612 (defines_file): Likewise.
19613 (pgm_path): Likewise.
19614 (bif_path): Likewise.
19615 (ovld_path): Likewise.
19616 (header_path): Likewise.
19617 (init_path): Likewise.
19618 (defines_path): Likewise.
19619 (LINELEN): New macro.
19620 (linebuf): New variable.
19624 (bif_diag): New function.
19625 (ovld_diag): Likewise.
19627 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
19629 * config/rs6000/rs6000-builtin-new.def: New.
19630 * config/rs6000/rs6000-overload.def: New.
19632 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
19634 * config/rs6000/rs6000-gen-builtins.c: New.
19636 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
19638 * Makefile.in (EXTRA_GTYPE_DEPS): New variable.
19639 (s-gtype): Depend on EXTRA_GTYPE_DEPS.
19640 * gengtype-state.c (state_writer::write_state_file_list): Add a
19641 parameter to the fileslist expression for the number of build
19643 (read_state_files_list): Detect build headers and strip the
19644 initial "./" or ".\" from their names.
19645 * gengtype.c (build_headers): New global variable.
19646 (num_build_headers): Likewise.
19647 (open_base_files): Emit #include for each build header.
19648 (main): Detect and count build headers.
19649 * gengtype.h (build_headers): New extern variable.
19650 (num_build_headers): Likewise.
19652 2021-07-16 Richard Biener <rguenther@suse.de>
19654 * gimple-ssa-store-merging.c (verify_symbolic_number_p): Use
19655 the type of the LHS.
19656 (find_bswap_or_nop_1): Likewise.
19657 (find_bswap_or_nop): Likewise.
19658 * tree-vectorizer.h (vect_get_smallest_scalar_type): Adjust
19660 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
19661 Remove unused parameters, pass in the scalar type. Fix
19662 internal store function handling.
19663 * tree-vect-stmts.c (vect_analyze_stmt): Remove assert.
19664 (vect_get_vector_types_for_stmt): Move down check for
19665 existing vector stmt after we've determined a scalar type.
19666 Pass down the used scalar type to vect_get_smallest_scalar_type.
19667 * tree-vect-generic.c (expand_vector_condition): Use
19668 the type of the LHS.
19669 (expand_vector_scalar_condition): Likewise.
19670 (expand_vector_operations_1): Likewise.
19671 * tree-vect-patterns.c (vect_widened_op_tree): Likewise.
19672 (vect_recog_dot_prod_pattern): Likewise.
19673 (vect_recog_sad_pattern): Likewise.
19674 (vect_recog_widen_op_pattern): Likewise.
19675 (vect_recog_widen_sum_pattern): Likewise.
19676 (vect_recog_mixed_size_cond_pattern): Likewise.
19678 2021-07-16 Jan Hubicka <hubicka@ucw.cz>
19680 * ipa-modref.c (struct escape_entry): Use eaf_fleags_t.
19681 (dump_eaf_flags): Dump EAF_NOT_RETURNED
19682 (eaf_flags_useful_p): Use eaf_fleags_t; handle const functions
19683 and EAF_NOT_RETURNED.
19684 (modref_summary::useful_p): Likewise.
19685 (modref_summary_lto::useful_p): Likewise.
19686 (struct) modref_summary_lto: Use eaf_fleags_t.
19687 (deref_flags): Handle EAF_NOT_RETURNED.
19688 (struct escape_point): Use min_flags.
19689 (modref_lattice::init): Add EAF_NOT_RETURNED.
19690 (merge_call_lhs_flags): Ignore EAF_NOT_RETURNED functions
19691 (analyze_ssa_name_flags): Clear EAF_NOT_RETURNED on return;
19693 (analyze_parms): Also analyze const functions; update conition on
19695 (modref_write): Update streaming.
19696 (read_section): Update streaming.
19697 (remap_arg_flags): Use eaf_flags_t.
19698 (modref_merge_call_site_flags): Hanlde EAF_NOT_RETURNED.
19699 * ipa-modref.h: (eaf_flags_t): New typedef.
19700 (struct modref_summary): Use eaf_flags_t.
19701 * tree-core.h (EAF_NOT_RETURNED): New constant.
19703 2021-07-16 Richard Biener <rguenther@suse.de>
19705 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Use
19706 the type of the LHS.
19707 (gimple_assign_nonnegative_warnv_p): Likewise.
19708 (gimple_call_nonnegative_warnv_p): Likewise. Return false
19709 if the call has no LHS.
19710 * gimple.c (gimple_could_trap_p_1): Use the type of the LHS.
19711 * tree-eh.c (stmt_could_throw_1_p): Likewise.
19712 * tree-inline.c (insert_init_stmt): Likewise.
19713 * tree-ssa-loop-niter.c (get_val_for): Likewise.
19714 * tree-outof-ssa.c (ssa_is_replaceable_p): Use the type of
19716 * tree-ssa-sccvn.c (init_vn_nary_op_from_stmt): Take a
19717 gassign *. Use the type of the lhs.
19718 (vn_nary_op_lookup_stmt): Adjust.
19719 (vn_nary_op_insert_stmt): Likewise.
19721 2021-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
19723 * config/s390/predicates.md (bras_sym_operand): Accept all
19724 functions in 64-bit mode, use UNSPEC_PLT31.
19725 (larl_operand): Use UNSPEC_PLT31.
19726 * config/s390/s390.c (s390_loadrelative_operand_p): Likewise.
19727 (legitimize_pic_address): Likewise.
19728 (s390_emit_tls_call_insn): Mark __tls_get_offset as function,
19730 (s390_delegitimize_address): Use UNSPEC_PLT31.
19731 (s390_output_addr_const_extra): Likewise.
19732 (print_operand): Add @PLT to TLS calls, handle %K.
19733 (s390_function_profiler): Mark __fentry__/_mcount as function,
19734 use %K, use UNSPEC_PLT31.
19735 (s390_output_mi_thunk): Use only UNSPEC_GOT, use %K.
19736 (s390_emit_call): Use UNSPEC_PLT31.
19737 (s390_emit_tpf_eh_return): Mark __tpf_eh_return as function.
19738 * config/s390/s390.md (UNSPEC_PLT31): Rename from UNSPEC_PLT.
19739 (*movdi_64): Use %K.
19740 (reload_base_64): Likewise.
19741 (*sibcall_brc): Likewise.
19742 (*sibcall_brcl): Likewise.
19743 (*sibcall_value_brc): Likewise.
19744 (*sibcall_value_brcl): Likewise.
19746 (*brasl): Likewise.
19747 (*bras_r): Likewise.
19748 (*brasl_r): Likewise.
19749 (*bras_tls): Likewise.
19750 (*brasl_tls): Likewise.
19751 (main_base_64): Likewise.
19752 (reload_base_64): Likewise.
19753 (@split_stack_call<mode>): Likewise.
19755 2021-07-16 Richard Biener <rguenther@suse.de>
19757 PR tree-optimization/101467
19758 * tree-vect-stmts.c (vect_gen_while): Properly guard
19759 make_temp_ssa_name usage.
19761 2021-07-16 Cooper Qu <cooper.qu@linux.alibaba.com>
19763 * config.gcc: Don't use forked print-sysroot-suffix.sh and
19764 t-sysroot-suffix for C-SKY.
19765 * config/csky/print-sysroot-suffix.sh: Delete.
19766 * config/csky/t-csky-linux: Delete.
19767 * config/csky/t-sysroot-suffix: Define MULTILIB_DIRNAMES
19768 instead of CSKY_MULTILIB_DIRNAMES.
19770 2021-07-16 Richard Biener <rguenther@suse.de>
19772 * tree-vect-loop.c (vect_transform_cycle_phi): Correct sign
19773 conversion issues with the partial reduction of the reused
19774 vector accumulator.
19776 2021-07-16 Richard Biener <rguenther@suse.de>
19778 * config/i386/i386-options.c (ix86_option_override_internal): Set
19779 param_vect_partial_vector_usage to zero if not set.
19781 2021-07-15 Uroš Bizjak <ubizjak@gmail.com>
19784 * config/i386/i386.h (VALID_SSE_REG_MODE): Add TDmode.
19785 (VALID_INT_MODE_P): Add SDmode and DDmode.
19786 Add TDmode for TARGET_64BIT.
19787 (VALID_DFP_MODE_P): Remove.
19788 * config/i386/i386.c (ix86_hard_regno_mode_ok):
19789 Do not use VALID_DFP_MODE_P.
19791 2021-07-15 Andrew MacLeod <amacleod@redhat.com>
19793 * gimple-range-fold.cc (adjust_pointer_diff_expr): Use
19795 (fold_using_range::fold_stmt): Ditto.
19796 (fold_using_range::range_of_range_op): Ditto.
19797 (fold_using_range::range_of_phi): Ditto.
19798 (fold_using_range::range_of_call): Ditto.
19799 (fold_using_range::range_of_builtin_ubsan_call): Ditto.
19800 (fold_using_range::range_of_builtin_call): Ditto.
19801 (fold_using_range::range_of_cond_expr): Ditto.
19802 * gimple-range-fold.h (gimple_range_type): New.
19804 2021-07-15 Martin Sebor <msebor@redhat.com>
19806 PR middle-end/97027
19807 * tree-ssa-strlen.c (handle_assign): New function.
19808 (maybe_warn_overflow): Add argument.
19809 (nonzero_bytes_for_type): New function.
19810 (count_nonzero_bytes): Handle more tree types. Call
19811 nonzero_bytes_for_tye.
19812 (count_nonzero_bytes): Handle types.
19813 (handle_store): Handle stores from function calls.
19814 (strlen_check_and_optimize_call): Move code to handle_assign. Call
19815 it for assignments from function calls.
19817 2021-07-15 David Malcolm <dmalcolm@redhat.com>
19822 * doc/invoke.texi: Add -Wanalyzer-use-of-uninitialized-value.
19824 2021-07-15 David Malcolm <dmalcolm@redhat.com>
19826 * doc/invoke.texi (-fdump-analyzer-exploded-paths): New.
19828 2021-07-15 Martin Sebor <msebor@redhat.com>
19832 * fold-const.c (operand_compare::operand_equal_p): Handle OEP_DECL_NAME.
19833 (operand_compare::verify_hash_value): Same.
19834 * tree-core.h (OEP_DECL_NAME): New.
19836 2021-07-15 Martin Jambor <mjambor@suse.cz>
19838 * profile-count.h (profile_count::value): Change the return type to
19840 * gimple-pretty-print.c (dump_gimple_bb_header): Adjust print
19842 * tree-cfg.c (dump_function_to_file): Likewise.
19844 2021-07-15 Bill Schmidt <wschmidt@linux.ibm.com>
19847 * config/rs6000/rs6000-p8swap.c (has_part_mult): New.
19848 (rs6000_analyze_swaps): Insns containing a subreg of a mult are
19851 2021-07-15 Richard Biener <rguenther@suse.de>
19853 * tree-vectorizer.h (vect_gen_while): Match up with
19854 vect_gen_while_not.
19855 * tree-vect-stmts.c (vect_gen_while): Adjust API to that
19856 of vect_gen_while_not.
19857 (vect_gen_while_not): Adjust.
19858 * tree-vect-loop-manip.c (vect_set_loop_controls_directly): Likewise.
19860 2021-07-15 Aldy Hernandez <aldyh@redhat.com>
19862 * gimple-range-cache.cc (non_null_ref::adjust_range): New.
19863 (ranger_cache::range_of_def): Call adjust_range.
19864 (ranger_cache::entry_range): Same.
19865 * gimple-range-cache.h (non_null_ref::adjust_range): New.
19866 * gimple-range.cc (gimple_ranger::range_of_expr): Call
19868 (gimple_ranger::range_on_entry): Same.
19870 2021-07-15 Tamar Christina <tamar.christina@arm.com>
19873 2021-07-14 Tamar Christina <tamar.christina@arm.com>
19875 * config/arm/neon.md (<sup>dot_prod<vsi2qi>): Drop statements.
19877 2021-07-15 Tamar Christina <tamar.christina@arm.com>
19880 2021-07-14 Tamar Christina <tamar.christina@arm.com>
19882 * config/aarch64/aarch64-simd-builtins.def (udot, sdot): Rename to...
19883 (sdot_prod, udot_prod): ...These.
19884 * config/aarch64/aarch64-simd.md (<sur>dot_prod<vsi2qi>): Remove.
19885 (aarch64_<sur>dot<vsi2qi>): Rename to...
19886 (<sur>dot_prod<vsi2qi>): ...This.
19887 * config/aarch64/arm_neon.h (vdot_u32, vdotq_u32, vdot_s32, vdotq_s32):
19890 2021-07-15 Jakub Jelinek <jakub@redhat.com>
19892 PR middle-end/101437
19893 * gimplify.c (gimplify_expr): Throw away volatile reads from empty
19894 types even if they have non-BLKmode TYPE_MODE.
19896 2021-07-15 Richard Biener <rguenther@suse.de>
19899 * gcc.c (process_command): Process -gtoggle like process_options
19900 would after parsing options.
19902 2021-07-15 Trevor Saunders <tbsaunde@tbsaunde.org>
19904 * cfgexpand.c (expand_asm_loc): Adjust.
19905 (expand_asm_stmt): Likewise.
19906 * config/arm/aarch-common-protos.h (arm_md_asm_adjust): Likewise.
19907 * config/arm/aarch-common.c (arm_md_asm_adjust): Likewise.
19908 * config/arm/arm.c (thumb1_md_asm_adjust): Likewise.
19909 * config/avr/avr.c (avr_md_asm_adjust): Likewise.
19910 * config/cris/cris.c (cris_md_asm_adjust): Likewise.
19911 * config/i386/i386.c (ix86_md_asm_adjust): Likewise.
19912 * config/mn10300/mn10300.c (mn10300_md_asm_adjust): Likewise.
19913 * config/nds32/nds32.c (nds32_md_asm_adjust): Likewise.
19914 * config/pdp11/pdp11.c (pdp11_md_asm_adjust): Likewise.
19915 * config/rs6000/rs6000.c (rs6000_md_asm_adjust): Likewise.
19916 * config/s390/s390.c (s390_md_asm_adjust): Likewise.
19917 * config/vax/vax.c (vax_md_asm_adjust): Likewise.
19918 * config/visium/visium.c (visium_md_asm_adjust): Likewise.
19919 * doc/tm.texi: Regenerate.
19920 * target.def: Add location argument to md_asm_adjust.
19922 2021-07-15 Trevor Saunders <tbsaunde@tbsaunde.org>
19924 * tree-diagnostic.c (diagnostic_report_current_function): Use the
19925 diagnostic's location, not input_location.
19927 2021-07-15 Trevor Saunders <tbsaunde@tbsaunde.org>
19929 * cfgexpand.c (tree_conflicts_with_clobbers_p): Pass location to
19931 (expand_asm_stmt): Likewise.
19933 2021-07-14 Peter Bergner <bergner@linux.ibm.com>
19935 * config/rs6000/rs6000.c (adjacent_mem_locations): Return the lower
19936 addressed memory rtx, if any.
19937 (rs6000_split_multireg_move): Fix code formatting.
19938 Handle MMA build built-ins with operands in adjacent memory locations.
19940 2021-07-14 Peter Bergner <bergner@linux.ibm.com>
19942 * config/rs6000/rs6000.c (rs6000_split_multireg_move): Move to later
19945 2021-07-14 Jason Merrill <jason@redhat.com>
19947 * sel-sched-ir.h (get_all_loop_exits): Use auto_vec.
19949 2021-07-14 Jason Merrill <jason@redhat.com>
19951 * doc/invoke.texi: -fdelete-dead-exceptions is on by default for
19954 2021-07-14 Tamar Christina <tamar.christina@arm.com>
19956 * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
19957 Remove erroneous line.
19959 2021-07-14 Andrew MacLeod <amacleod@redhat.com>
19961 * params.opt (param_evrp_mode): Change default.
19963 2021-07-14 Tamar Christina <tamar.christina@arm.com>
19965 * config/aarch64/aarch64-simd-builtins.def (udot, sdot): Rename to...
19966 (sdot_prod, udot_prod): ...These.
19967 * config/aarch64/aarch64-simd.md (<sur>dot_prod<vsi2qi>): Remove.
19968 (aarch64_<sur>dot<vsi2qi>): Rename to...
19969 (<sur>dot_prod<vsi2qi>): ...This.
19970 * config/aarch64/arm_neon.h (vdot_u32, vdotq_u32, vdot_s32, vdotq_s32):
19973 2021-07-14 Tamar Christina <tamar.christina@arm.com>
19975 * config/arm/neon.md (<sup>dot_prod<vsi2qi>): Drop statements.
19977 2021-07-14 Tamar Christina <tamar.christina@arm.com>
19979 * doc/sourcebuild.texi (arm_v8_2a_i8mm_neon_hw): Document.
19981 2021-07-14 Tamar Christina <tamar.christina@arm.com>
19983 * config/arm/neon.md (usdot_prod<vsi2qi>): New.
19985 2021-07-14 Tamar Christina <tamar.christina@arm.com>
19987 * config/aarch64/aarch64-simd.md (aarch64_usdot<vsi2qi>): Rename to...
19988 (usdot_prod<vsi2qi>): ... This.
19989 * config/aarch64/aarch64-simd-builtins.def (usdot): Rename to...
19990 (usdot_prod): ...This.
19991 * config/aarch64/arm_neon.h (vusdot_s32, vusdotq_s32): Likewise.
19992 * config/aarch64/aarch64-sve.md (@aarch64_<sur>dot_prod<vsi2qi>):
19994 (@<sur>dot_prod<vsi2qi>): ...This.
19995 * config/aarch64/aarch64-sve-builtins-base.cc
19996 (svusdot_impl::expand): Use it.
19998 2021-07-14 Tamar Christina <tamar.christina@arm.com>
20000 * optabs.def (usdot_prod_optab): New.
20001 * doc/md.texi: Document it and clarify other dot prod optabs.
20002 * optabs-tree.h (enum optab_subtype): Add optab_vector_mixed_sign.
20003 * optabs-tree.c (optab_for_tree_code): Support usdot_prod_optab.
20004 * optabs.c (expand_widen_pattern_expr): Likewise.
20005 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
20006 * tree-vect-loop.c (vectorizable_reduction): Query dot-product kind.
20007 * tree-vect-patterns.c (vect_supportable_direct_optab_p): Take optional
20009 (vect_widened_op_tree): Optionally ignore
20011 (vect_recog_dot_prod_pattern): Support usdot_prod_optab.
20013 2021-07-14 H.J. Lu <hjl.tools@gmail.com>
20016 * config/i386/driver-i386.c (host_detect_local_cpu): Check
20017 "arch [32|64]" and "tune [32|64]" for 32-bit and 64-bit codegen.
20018 Enable UINTR only for 64-bit codegen.
20019 * config/i386/i386-options.c
20020 (ix86_option_override_internal::DEF_PTA): Skip PTA_UINTR if not
20022 * config/i386/i386.h (ARCH_ARG): New.
20023 (CC1_CPU_SPEC): Pass "[arch|tune] 32" for 32-bit codegen and
20024 "[arch|tune] 64" for 64-bit codegen.
20026 2021-07-14 Richard Biener <rguenther@suse.de>
20028 PR tree-optimization/101445
20029 * tree-vect-stmts.c (vectorizable_load): Do the gap adjustment
20030 of the IV in the correct direction for negative stride
20033 2021-07-14 Jakub Jelinek <jakub@redhat.com>
20036 * godump.c (godump_str_hash): New type.
20037 (godump_container::pot_dummy_types): Use string_hash instead of
20038 ptr_hash in the hash_set.
20040 2021-07-14 Richard Biener <rguenther@suse.de>
20042 * tree-vect-loop.c (vect_find_reusable_accumulator): Handle
20043 vector types where the old vector type has a multiple of
20044 the new vector type elements.
20045 (vect_create_partial_epilog): New function, split out from...
20046 (vect_create_epilog_for_reduction): ... here.
20047 (vect_transform_cycle_phi): Reduce the re-used accumulator
20048 to the new vector type.
20050 2021-07-14 Alexandre Oliva <oliva@adacore.com>
20052 * tree-ssa-alias.c (attr_fnspec::verify): Fix index in
20053 non-'t'-sized arg check.
20055 2021-07-14 Alexandre Oliva <oliva@adacore.com>
20057 * tree-cfg.c (cleanup_dead_labels_eh): Update
20058 post_landing_pad label upon change of landing pad block's
20060 (cleanup_dead_labels): Check that a removed label is not that
20063 2021-07-13 Jonathan Wright <jonathan.wright@arm.com>
20065 * combine.c (combine_simplify_rtx): Add vec_select -> subreg
20067 * config/aarch64/aarch64.md (*zero_extend<SHORT:mode><GPI:mode>2_aarch64):
20068 Add Neon to general purpose register case for zero-extend
20070 * config/arm/vfp.md (*arm_movsi_vfp): Remove "*" from *t -> r
20071 case to prevent some cases opting to go through memory.
20072 * cse.c (fold_rtx): Add vec_select -> subreg simplification.
20073 * rtl.c (rtvec_series_p): Define predicate to determine
20074 whether a vector contains a linear series of integers.
20075 * rtl.h (rtvec_series_p): Define.
20076 * rtlanal.c (vec_series_lowpart_p): Define predicate to
20077 determine if a vector selection is equivalent to the low part
20079 * rtlanal.h (vec_series_lowpart_p): Define.
20080 * simplify-rtx.c (simplify_context::simplify_binary_operation_1):
20081 Add vec_select -> subreg simplification.
20083 2021-07-13 Paul A. Clarke <pc@us.ibm.com>
20085 * config/rs6000/smmintrin.h (_mm_testz_si128, _mm_testc_si128,
20086 _mm_testnzc_si128, _mm_test_all_ones, _mm_test_all_zeros,
20087 _mm_test_mix_ones_zeros): New.
20089 2021-07-13 Roger Sayle <roger@nextmovesoftware.com>
20090 Richard Biener <rguenther@suse.de>
20092 * gimple.c (gimple_could_trap_p_1): Make S argument a
20093 "const gimple*". Preserve constness in call to
20094 gimple_asm_volatile_p.
20095 (gimple_could_trap_p): Make S argument a "const gimple*".
20096 * gimple.h (gimple_could_trap_p_1, gimple_could_trap_p):
20097 Update function prototypes.
20099 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
20101 * tree-vectorizer.h (vect_reusable_accumulator): New structure.
20102 (_loop_vec_info::main_loop_edge): New field.
20103 (_loop_vec_info::skip_main_loop_edge): Likewise.
20104 (_loop_vec_info::skip_this_loop_edge): Likewise.
20105 (_loop_vec_info::reusable_accumulators): Likewise.
20106 (_stmt_vec_info::reduc_scalar_results): Likewise.
20107 (_stmt_vec_info::reused_accumulator): Likewise.
20108 (vect_get_main_loop_result): Declare.
20109 * tree-vectorizer.c (vec_info::new_stmt_vec_info): Initialize
20110 reduc_scalar_inputs.
20111 (vec_info::free_stmt_vec_info): Free reduc_scalar_inputs.
20112 * tree-vect-loop-manip.c (vect_get_main_loop_result): New function.
20113 (vect_do_peeling): Fill an epilogue loop's main_loop_edge,
20114 skip_main_loop_edge and skip_this_loop_edge fields.
20115 * tree-vect-loop.c (INCLUDE_ALGORITHM): Define.
20116 (vect_emit_reduction_init_stmts): New function.
20117 (get_initial_def_for_reduction): Use it.
20118 (get_initial_defs_for_reduction): Likewise. Change the vinfo
20119 parameter to a loop_vec_info.
20120 (vect_create_epilog_for_reduction): Store the scalar results
20121 in the reduc_info. If an epilogue loop is reusing an accumulator
20122 from the main loop, and if the epilogue loop can also be skipped,
20123 try to place the reduction code in the join block. Record
20124 accumulators that could potentially be reused by epilogue loops.
20125 (vect_transform_cycle_phi): When vectorizing epilogue loops,
20126 try to reuse accumulators from the main loop. Record the initial
20127 value in reduc_info for non-SLP reductions too.
20129 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
20131 * tree-vect-loop.c (get_initial_def_for_reduction): Remove
20132 adjustment handling. Take the neutral value as an argument,
20133 in place of the code argument.
20134 (vect_transform_cycle_phi): Update accordingly. Handle the
20135 initial values of cond reductions separately from code reductions.
20136 Choose the adjustment here rather than in
20137 get_initial_def_for_reduction. Sink the splat of vec_initial_def.
20139 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
20141 * tree-vect-loop.c (neutral_op_for_slp_reduction): Replace with...
20142 (neutral_op_for_reduction): ...this, providing a more general
20144 (vect_create_epilog_for_reduction): Update accordingly.
20145 (vectorizable_reduction): Likewise.
20146 (vect_transform_cycle_phi): Likewise.
20148 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
20150 * tree-vect-loop.c (get_initial_def_for_reduction): Take the
20151 reduc_info instead of the original stmt_vec_info.
20152 (vect_transform_cycle_phi): Update accordingly.
20154 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
20156 * tree-vect-loop.c (get_initial_defs_for_reduction): Take the
20157 reduc_info as an additional parameter.
20158 (vect_transform_cycle_phi): Update accordingly.
20160 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
20162 * tree-vectorizer.h: Include tree-ssa-operands.h.
20163 (vect_phi_initial_value): New function.
20164 * tree-vect-loop.c (neutral_op_for_slp_reduction): Use it.
20165 (get_initial_defs_for_reduction, info_for_reduction): Likewise.
20166 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
20167 (vect_transform_cycle_phi, vectorizable_induction): Likewise.
20169 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
20171 * tree-vect-loop.c (vect_create_epilog_for_reduction): Convert
20172 the phi results to vectype after creating them. Remove later
20173 conversion code that thus becomes redundant.
20175 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
20177 * tree-vect-loop.c (vect_create_epilog_for_reduction): Replace
20178 the new_phis vector with a reduc_inputs vector. Combine handling
20179 of reduction chains and ncopies > 1.
20181 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
20183 * tree-vect-loop.c (vect_create_epilog_for_reduction): Truncate
20184 scalar_results to group_size elements after reducing down from
20185 N*group_size elements. Construct an array_slice of the live-out
20186 stmts and assert that there is one stmt per scalar result.
20188 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
20190 * tree-vect-loop.c (vect_create_epilog_for_reduction): Remove
20191 nested_in_vect_loop and use double_reduc everywhere. Remove dead
20192 assignment to "loop".
20194 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
20196 * internal-fn.c (vectorized_internal_fn_supported_p): Handle
20197 vector types first. For scalar types, consider both the preferred
20198 vector mode and the alternative vector modes.
20199 * optabs-query.c (can_vec_mask_load_store_p): Use the same
20200 structure as above, in particular using related_vector_mode
20201 for modes provided by autovectorize_vector_modes.
20203 2021-07-13 Jakub Jelinek <jakub@redhat.com>
20204 Richard Biener <rguenther@suse.de>
20206 PR tree-optimization/101419
20207 * tree-pass.h (PROP_objsz): Define.
20208 (make_pass_early_object_sizes): Declare.
20209 * passes.def (pass_all_early_optimizations): Rename pass_object_sizes
20210 there to pass_early_object_sizes, drop parameter.
20211 (pass_all_optimizations): Move pass_object_sizes right after pass_ccp,
20212 drop parameter, move pass_post_ipa_warn right after that.
20213 * tree-object-size.c (pass_object_sizes::execute): Rename to...
20214 (object_sizes_execute): ... this. Add insert_min_max_p argument.
20215 (pass_data_object_sizes): Move after object_sizes_execute.
20216 (pass_object_sizes): Likewise. In execute method call
20217 object_sizes_execute, drop set_pass_param method and insert_min_max_p
20218 non-static data member and its initializer in the ctor.
20219 (pass_data_early_object_sizes, pass_early_object_sizes,
20220 make_pass_early_object_sizes): New.
20221 * tree-ssa-sccvn.c (copy_reference_ops_from_ref): Use
20222 (cfun->curr_properties & PROP_objsz) instead of cfun->after_inlining.
20224 2021-07-13 Kito Cheng <kito.cheng@sifive.com>
20227 * config/riscv/constraints.md ("S"): Update description and remove
20229 * doc/md.texi (Machine Constraints): Document the 'S' constraints
20232 2021-07-13 Richard Biener <rguenther@suse.de>
20235 2021-07-12 Richard Biener <rguenther@suse.de>
20237 * tree-vect-slp.c (vect_slp_region): Show the number of
20238 SLP graph entries in the optimization message.
20240 2021-07-13 Michael Meissner <meissner@linux.ibm.com>
20242 * config/rs6000/altivec.md (xxspltiw_v4sf): Change local variable
20244 * config/rs6000/rs6000-protos.h (rs6000_const_f32_to_i32): Change
20245 return type to long.
20246 * config/rs6000/rs6000.c (rs6000_const_f32_to_i32): Change return
20249 2021-07-12 Andrew MacLeod <amacleod@redhat.com>
20251 * gimple-range-fold.cc (fold_using_range::range_of_builtin_ubsan_call):
20252 Query relation between the 2 operands and use it.
20254 2021-07-12 Sergei Trofimovich <siarheit@google.com>
20256 * doc/cfg.texi: Fix s/ei_safe_safe/ei_safe_edge/ typo.
20258 2021-07-12 Uroš Bizjak <ubizjak@gmail.com>
20261 * config/i386/predicates.md (vec_setm_sse41_operand):
20262 Rename from vec_setm_operand.
20263 (vec_setm_avx2_operand): New predicate.
20264 * config/i386/sse.md (vec_set<V_128:mode>): Use V_128 mode iterator.
20265 Use vec_setm_sse41_operand as operand 2 predicate.
20266 (vec_set<V_256_512:mode): New expander.
20267 * config/i386/mmx.md (vec_setv2hi): Use vec_setm_sse41_operand
20268 as operand 2 predicate.
20270 2021-07-12 Andrew MacLeod <amacleod@redhat.com>
20272 PR tree-optimization/101335
20273 * range-op.cc (operator_cast::lhs_op1_relation): Delete.
20275 2021-07-12 Andrew Pinski <apinski@marvell.com>
20277 * tree-ssa-phiopt.c (match_simplify_replacement): Move
20278 insert of the sequence before the movement of the
20279 statement. Check if to see if the statement is used
20280 outside of the original phi to see if we should move it.
20282 2021-07-12 Richard Biener <rguenther@suse.de>
20284 * dump-context.h (debug_dump_context::debug_dump_context):
20285 Add FILE * parameter defaulted to stderr.
20286 * dumpfile.c (debug_dump_context::debug_dump_context): Adjust.
20287 * tree-vect-slp.c (dot_slp_tree): New functions.
20289 2021-07-12 Richard Biener <rguenther@suse.de>
20291 PR tree-optimization/101373
20292 * tree-ssa-pre.c (prune_clobbered_mems): Also prune trapping
20293 references when the BB may not return.
20294 (compute_avail): Pass in the function we're working on and
20295 replace cfun references with it. Externally throwing
20296 const calls also possibly terminate the function.
20297 (pass_pre::execute): Pass down the function we're working on.
20298 * gcse.c (compute_hash_table_work): Externally throwing
20299 const/pure calls also need record_last_mem_set_info.
20300 * postreload-gcse.c (record_opr_changes): Looping or externally
20301 throwing const/pure calls also need record_last_mem_set_info.
20303 2021-07-12 Uroš Bizjak <ubizjak@gmail.com>
20305 * recog.c (memory_address_addr_space_p): Change the type to bool.
20306 Return true/false instead of 1/0.
20307 (offsettable_memref_p): Ditto.
20308 (offsettable_nonstrict_memref_p): Ditto.
20309 (offsettable_address_addr_space_p): Ditto.
20310 Change the type of addressp indirect function to bool.
20311 * recog.h (memory_address_addr_space_p): Change the type to bool.
20312 (strict_memory_address_addr_space_p): Ditto.
20313 (offsettable_memref_p): Ditto.
20314 (offsettable_nonstrict_memref_p): Ditto.
20315 (offsettable_address_addr_space_p): Ditto.
20316 * reload.c (maybe_memory_address_addr_space_p): Ditto.
20317 (strict_memory_address_addr_space_p): Change the type to bool.
20318 Return true/false instead of 1/0.
20319 (maybe_memory_address_addr_space_p): Change the type to bool.
20321 2021-07-12 Richard Biener <rguenther@suse.de>
20323 * tree-vect-slp.c (vect_slp_region): Show the number of
20324 SLP graph entries in the optimization message.
20326 2021-07-12 Richard Biener <rguenther@suse.de>
20328 PR tree-optimization/101394
20329 * tree-ssa-pre.c (do_pre_regular_insertion): Avoid inserting
20330 copies from abnormals for a full redundancy.
20332 2021-07-12 Richard Biener <rguenther@suse.de>
20334 PR middle-end/101423
20335 * gimple.c (gimple_could_trap_p_1): Internal function calls
20337 * tree-eh.c (tree_could_trap_p): Likewise.
20339 2021-07-12 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
20342 * config/arm/arm_neon.h (vmul_n_u32): Replace call to builtin with
20344 (vmulq_n_u32): Likewise.
20345 (vmul_n_f32): Gate __a * __b on __FAST_MATH__.
20346 (vmulq_n_f32): Likewise.
20347 (vmul_n_f16): Likewise.
20348 (vmulq_n_f16): Likewise.
20350 2021-07-12 Martin Liska <mliska@suse.cz>
20352 PR sanitizer/101425
20353 * gcc.c (check_offload_target_name): Call
20354 candidates_list_and_hint only if we have a candidate.
20356 2021-07-12 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
20359 * config/arm/neon.md (vec_init): Move to ...
20360 * config/arm/vec-common.md (vec_init): ... here.
20361 Change the pattern's mode to VDQX and gate it on VALID_MVE_MODE.
20363 2021-07-12 Roger Sayle <roger@nextmovesoftware.com>
20365 PR tree-optimization/101403
20366 * match.pd ((T)bswap(X)>>C): Correctly handle cases where
20367 signedness of the shift is not the same as the signedness of
20368 the type extension.
20370 2021-07-09 Roger Sayle <roger@nextmovesoftware.com>
20371 Uroš Bizjak <ubizjak@gmail.com>
20373 * config/i386/i386.md (*divmodsi4_const): Optimize SImode
20374 divmod of a constant numerator with new define_insn_and_split.
20376 2021-07-09 Iain Sandoe <iain@sandoe.co.uk>
20379 * config/i386/i386-expand.c (ix86_expand_call): If a call is
20380 to a non-local-binding, or local but to a public symbol, then
20381 assume that it might be indirected via the lazy symbol binder.
20382 Mark R10 and R10 as clobbered in that case.
20384 2021-07-09 Eric Botcazou <ebotcazou@adacore.com>
20387 * gcc.c (ASM_DEBUG_DWARF_OPTION): Set again to --gdwarf2 in
20388 the case where HAVE_AS_WORKING_DWARF_N_FLAG is not defined
20389 and HAVE_LD_BROKEN_PE_DWARF5 is defined.
20391 2021-07-09 Uroš Bizjak <ubizjak@gmail.com>
20393 * config/i386/i386.md (*udivmodsi4_pow2_zext_1): Limit the
20394 log2 range of operands[3] to [1,31].
20395 (*udivmodsi4_pow2_zext_2): Ditto. Correct insn RTX pattern.
20397 2021-07-09 Sergei Trofimovich <siarheit@google.com>
20399 * doc/md.texi: Don't split @smallexample in multiple @groups.
20401 2021-07-09 Sergei Trofimovich <siarheit@google.com>
20403 * doc/md.texi: Add missing 'see' word.
20405 2021-07-09 Andrew Pinski <apinski@marvell.com>
20407 * tree-ssa-phiopt.c (phiopt_early_allow): Change arguments
20408 to take sequence and gimple_match_op. Accept the case where
20409 op is a SSA_NAME and one statement in the sequence.
20410 Also allow constants.
20411 (gimple_simplify_phiopt): Always pass a sequence to resimplify.
20412 Update call to phiopt_early_allow. Discard the sequence if not
20415 2021-07-09 Xi Ruoyao <xry111@mengyan1223.wang>
20420 * config/mips/mips.c (mips_const_insns): Use MSA_SUPPORTED_MODE_P
20421 instead of ISA_HAS_MSA.
20422 (mips_expand_vec_unpack): Likewise.
20423 (mips_expand_vector_init): Likewise.
20425 2021-07-09 Kewen Lin <linkw@linux.ibm.com>
20427 * config/rs6000/vsx.md (mods_<mode>): Rename to...
20428 (mod<mode>3): ... this.
20429 (modu_<mode>): Rename to...
20430 (umod<mode>3): ... this.
20431 * config/rs6000/rs6000-builtin.def (MODS_V2DI, MODS_V4SI, MODU_V2DI,
20432 MODU_V4SI): Adjust.
20434 2021-07-08 Jeff Law <jeffreyalaw@gmail.com>
20436 * config/h8300/shiftrotate.md (variable shifts): Expose condition
20437 code handling for the test before the loop.
20439 2021-07-08 Martin Jambor <mjambor@suse.cz>
20442 * ipa-sra.c (class isra_call_summary): New member
20443 m_before_any_store, initialize it in the constructor.
20444 (isra_call_summary::dump): Dump the new field.
20445 (ipa_sra_call_summaries::duplicate): Copy it.
20446 (process_scan_results): Set it.
20447 (isra_write_edge_summary): Stream it.
20448 (isra_read_edge_summary): Likewise.
20449 (param_splitting_across_edge): Only override
20450 safe_to_import_accesses if m_before_any_store is set.
20452 2021-07-08 Martin Sebor <msebor@redhat.com>
20454 PR bootstrap/101374
20455 * gimple-array-bounds.cc (array_bounds_checker::check_mem_ref):
20456 Use Object Size Type 0 instead of 1.
20458 2021-07-08 Richard Sandiford <richard.sandiford@arm.com>
20460 * tree-vect-loop.c (vectorizable_reduction): Remove always-true
20463 2021-07-08 Richard Sandiford <richard.sandiford@arm.com>
20465 * match.pd: Simplify an extend-operate-truncate sequence involving
20468 2021-07-08 Roger Sayle <roger@nextmovesoftware.com>
20469 Richard Biener <rguenther@suse.de>
20471 PR tree-optimization/40210
20472 * match.pd (bswap optimizations): Simplify (bswap(x)>>C1)&C2 as
20473 (x>>C3)&C2 when possible. Simplify bswap(x)>>C1 as ((T)x)>>C2
20474 when possible. Simplify bswap(x)&C1 as (x>>C2)&C1 when 0<=C1<=255.
20476 2021-07-08 Uroš Bizjak <ubizjak@gmail.com>
20479 * config/i386/i386-expand.c (ix86_expand_sse_unpack):
20481 * config/i386/mmx.md (V_32): New mode iterator.
20482 (mov<V_32:mode>): Use V_32 mode iterator.
20483 (*mov<V_32:mode>_internal): Ditto.
20484 (*push<V_32:mode>2_rex64): Ditto.
20485 (*push<V_32:mode>2): Ditto.
20486 (movmisalign<V_32:mode>): Ditto.
20487 (mmx_<any_shiftrt:insn>v1si3): New insn pattern.
20488 (sse4_1_<any_extend:code>v2qiv2hi2): Ditto.
20489 (vec_unpacks_lo_v4qi): New expander.
20490 (vec_unpacks_hi_v4qi): Ditto.
20491 (vec_unpacku_lo_v4qi): Ditto.
20492 (vec_unpacku_hi_v4qi): Ditto.
20493 * config/i386/i386.h (VALID_SSE2_REG_MODE): Add V1SImode.
20494 (VALID_INT_MODE_P): Ditto.
20496 2021-07-08 Michael Meissner <meissner@linux.ibm.com>
20499 * config/rs6000/rs6000.md (udivti3): New insn.
20500 (divti3): New insn.
20501 (umodti3): New insn.
20502 (modti3): New insn.
20504 2021-07-07 Martin Sebor <msebor@redhat.com>
20506 PR tree-optimization/100137
20507 PR tree-optimization/99121
20508 PR tree-optimization/97027
20509 * builtins.c (access_ref::access_ref): Also set offmax.
20510 (access_ref::offset_in_range): Define new function.
20511 (access_ref::add_offset): Set offmax.
20512 (access_ref::inform_access): Handle access_none.
20513 (handle_mem_ref): Clear ostype.
20514 (compute_objsize_r): Handle ASSERT_EXPR.
20515 * builtins.h (struct access_ref): Add offmax member.
20516 * gimple-array-bounds.cc (array_bounds_checker::check_mem_ref): Use
20517 compute_objsize() and simplify.
20519 2021-07-07 Peter Bergner <bergner@linux.ibm.com>
20521 * config/rs6000/rs6000-call.c (mma_init_builtins): Use VSX_BUILTIN_LXVP
20522 and VSX_BUILTIN_STXVP.
20524 2021-07-07 Martin Sebor <msebor@redhat.com>
20527 * config/aarch64/aarch64.c (aarch64_simd_lane_bounds): Remove
20528 a stray %K from error_at() missed in r12-2088.
20530 2021-07-07 Richard Biener <rguenther@suse.de>
20532 PR tree-optimization/99728
20533 * tree-ssa-loop-im.c (gather_mem_refs_stmt): Record
20535 (mem_refs_may_alias_p): Add assert we handled aggregate
20537 (sm_seq_valid_bb): Give up when running into aggregate copies.
20538 (ref_indep_loop_p): Handle aggregate copies as never
20539 being invariant themselves but allow other refs to be
20540 disambiguated against them.
20541 (can_sm_ref_p): Do not try to apply store-motion to aggregate
20544 2021-07-06 Indu Bhagat <indu.bhagat@oracle.com>
20547 * dwarf2ctf.c (ctf_get_AT_data_member_location): Multiply by 8 to get
20550 2021-07-06 Martin Sebor <msebor@redhat.com>
20552 * gimple-pretty-print.c (percent_G_format): Remove.
20553 * tree-diagnostic.c (default_tree_printer): Remove calls.
20554 * tree-pretty-print.c (percent_K_format): Remove.
20555 * tree-pretty-print.h (percent_K_format): Remove.
20557 2021-07-06 Martin Sebor <msebor@redhat.com>
20559 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_builtin):
20560 Remove %K and use error_at.
20561 (aarch64_expand_fcmla_builtin): Same.
20562 (aarch64_expand_builtin_tme): Same.
20563 (aarch64_expand_builtin_memtag): Same.
20564 * config/arm/arm-builtins.c (arm_expand_acle_builtin): Same.
20565 (arm_expand_builtin): Same.
20566 * config/arm/arm.c (bounds_check): Same.
20568 2021-07-06 Martin Sebor <msebor@redhat.com>
20570 * builtins.c (warn_string_no_nul): Remove %G.
20571 (maybe_warn_for_bound): Same.
20572 (warn_for_access): Same.
20573 (check_access): Same.
20574 (check_strncat_sizes): Same.
20575 (expand_builtin_strncat): Same.
20576 (expand_builtin_strncmp): Same.
20577 (expand_builtin): Same.
20578 (expand_builtin_object_size): Same.
20579 (warn_dealloc_offset): Same.
20580 (maybe_emit_free_warning): Same.
20581 * calls.c (maybe_warn_alloc_args_overflow): Same.
20582 (maybe_warn_nonstring_arg): Same.
20583 (maybe_warn_rdwr_sizes): Same.
20584 * expr.c (expand_expr_real_1): Remove %K.
20585 * gimple-fold.c (gimple_fold_builtin_strncpy): Remove %G.
20586 (gimple_fold_builtin_strncat): Same.
20587 * gimple-ssa-sprintf.c (format_directive): Same.
20588 (handle_printf_call): Same.
20589 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Same.
20590 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Same.
20591 (maybe_diag_access_bounds): Same. Call gimple_location.
20592 (check_bounds_or_overlap): Same.
20593 * trans-mem.c (ipa_tm_scan_irr_block): Remove %K. Simplify.
20594 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Remove %G.
20595 * tree-ssa-strlen.c (maybe_warn_overflow): Same.
20596 (maybe_diag_stxncpy_trunc): Same.
20597 (handle_builtin_stxncpy_strncat): Same.
20598 (maybe_warn_pointless_strcmp): Same.
20599 * tree-ssa-uninit.c (maybe_warn_operand): Same.
20601 2021-07-06 Uroš Bizjak <ubizjak@gmail.com>
20604 * config/i386/predicates.md (vec_setm_operand): Enable
20605 register_operand for TARGET_SSE4_1.
20606 * config/i386/mmx.md (vec_setv2hi): Use vec_setm_operand
20607 as operand 2 predicate. Call ix86_expand_vector_set_var
20608 for non-constant index operand.
20609 (vec_setv4qi): Use vec_setm_mmx_operand as operand 2 predicate.
20610 Call ix86_expand_vector_set_var for non-constant index operand.
20612 2021-07-06 Jeff Law <jeffreyalaw@gmail.com>
20614 * config/h8300/jumpcall.md (*branch): When possible, generate
20615 the comparison in CCZN mode.
20616 * config/h8300/predicates.md (simple_memory_operand): Reject all
20617 auto-increment addressing modes.
20619 2021-07-06 Iain Sandoe <iain@sandoe.co.uk>
20621 PR bootstrap/100246
20622 * config/i386/i386.h (struct stringop_algs): Define a CTOR for
20625 2021-07-06 Richard Biener <rguenther@suse.de>
20627 * doc/md.texi (vec_fmaddsub<mode>4): Document.
20628 (vec_fmsubadd<mode>4): Likewise.
20629 * optabs.def (vec_fmaddsub$a4): Add.
20630 (vec_fmsubadd$a4): Likewise.
20631 * internal-fn.def (IFN_VEC_FMADDSUB): Add.
20632 (IFN_VEC_FMSUBADD): Likewise.
20633 * tree-vect-slp-patterns.c (addsub_pattern::recognize):
20634 Refactor to handle IFN_VEC_FMADDSUB and IFN_VEC_FMSUBADD.
20635 (addsub_pattern::build): Likewise.
20636 * tree-vect-slp.c (vect_optimize_slp): CFN_VEC_FMADDSUB
20637 and CFN_VEC_FMSUBADD are not transparent for permutes.
20638 * config/i386/sse.md (vec_fmaddsub<mode>4): New expander.
20639 (vec_fmsubadd<mode>4): Likewise.
20641 2021-07-06 Richard Biener <rguenther@suse.de>
20643 * doc/invoke.texi (fmove-loop-stores): Document.
20644 * common.opt (fmove-loop-stores): New option.
20645 * opts.c (default_options_table): Enable -fmove-loop-stores
20646 at -O1 but not -Og.
20647 * tree-ssa-loop-im.c (pass_lim::execute): Pass
20648 flag_move_loop_stores instead of true to
20649 loop_invariant_motion_in_fun.
20651 2021-07-06 Iain Sandoe <iain@sandoe.co.uk>
20653 * doc/install.texi: Document --with-dsymutil.
20655 2021-07-06 Andrew Pinski <apinski@marvell.com>
20657 PR tree-optimization/101256
20658 * dbgcnt.def (phiopt_edge_range): New counter.
20659 * tree-ssa-phiopt.c (replace_phi_edge_with_variable):
20660 Check to make sure the new name is defined in the same
20661 bb as the conditional before duplicating range info.
20662 Also add debug counter.
20664 2021-07-06 Kewen Lin <linkw@linux.ibm.com>
20666 PR rtl-optimization/100328
20667 * config/i386/i386-options.c (ix86_option_override_internal):
20668 Set param_ira_consider_dup_in_all_alts to 0.
20670 2021-07-06 Kewen Lin <linkw@linux.ibm.com>
20672 PR rtl-optimization/100328
20673 * doc/invoke.texi (ira-consider-dup-in-all-alts): Document new
20675 * ira.c (ira_get_dup_out_num): Adjust as parameter
20676 param_ira_consider_dup_in_all_alts.
20677 * params.opt (ira-consider-dup-in-all-alts): New.
20678 * ira-conflicts.c (process_regs_for_copy): Add one parameter
20679 single_input_op_has_cstr_p.
20680 (get_freq_for_shuffle_copy): New function.
20681 (add_insn_allocno_copies): Adjust as single_input_op_has_cstr_p.
20682 * ira-int.h (ira_get_dup_out_num): Add one bool parameter.
20684 2021-07-05 Jeff Law <jeffreyalaw@gmail.com>
20686 * config/h8300/shiftrotate.md (shift-by-variable patterns): Update to
20687 generate condition code aware RTL directly.
20689 2021-07-05 Andrew Pinski <apinski@marvell.com>
20691 PR tree-optimization/101039
20692 * match.pd (A CMP 0 ? A : -A): New patterns.
20693 * tree-ssa-phiopt.c (abs_replacement): Delete function.
20694 (tree_ssa_phiopt_worker): Don't call abs_replacement.
20695 Update comment about abs_replacement.
20697 2021-07-05 Andrew Pinski <apinski@marvell.com>
20699 * tree-ssa-phiopt.c (gimple_simplify_phiopt):
20700 If "A ? B : C" fails to simplify, try "(!A) ? C : B".
20702 2021-07-05 Andrew Pinski <apinski@marvell.com>
20704 * tree-ssa-phiopt.c (match_simplify_replacement):
20705 Add early_p argument. Call gimple_simplify_phiopt
20706 instead of gimple_simplify.
20707 (tree_ssa_phiopt_worker): Update call to
20708 match_simplify_replacement and allow unconditionally.
20709 (phiopt_early_allow): New function.
20710 (gimple_simplify_phiopt): New function.
20712 2021-07-05 Andrew Pinski <apinski@marvell.com>
20714 PR middle-end/101237
20715 * fold-const.c (negate_expr_p): Remove call to element_mode
20716 and TREE_MODE/TREE_TYPE when calling HONOR_SIGNED_ZEROS,
20717 HONOR_SIGN_DEPENDENT_ROUNDING, and HONOR_SNANS.
20718 (fold_negate_expr_1): Likewise.
20719 (const_unop): Likewise.
20720 (fold_cond_expr_with_comparison): Likewise.
20721 (fold_binary_loc): Likewise.
20722 (fold_ternary_loc): Likewise.
20723 (tree_call_nonnegative_warnv_p): Likewise.
20724 * match.pd (-(A + B) -> (-B) - A): Likewise.
20726 2021-07-05 Iain Sandoe <iain@sandoe.co.uk>
20728 * configure.ac: Handle --with-dsymutil in the same way as we
20729 do for the assembler and linker. (DEFAULT_DSYMUTIL): New.
20730 Extract the type and version for the dsymutil configured or
20731 found by the default searches.
20732 * config.in: Regenerated.
20733 * configure: Regenerated.
20734 * collect2.c (do_dsymutil): Handle locating dsymutil in the
20735 same way as for the assembler and linker.
20736 * config/darwin.h (DSYMUTIL): Delete.
20737 * gcc.c: Report a configured dsymutil correctly.
20738 * exec-tool.in: Allow for dsymutil.
20740 2021-07-05 Uroš Bizjak <ubizjak@gmail.com>
20742 * config/i386/i386-expand.c (ix86_split_mmx_punpck):
20743 Handle V4QI and V2HI modes.
20744 (expand_vec_perm_blend): Allow 4-byte vector modes with TARGET_SSE4_1.
20745 Handle V4QI mode. Emit mmx_pblendvb32 for 4-byte modes.
20746 (expand_vec_perm_pshufb): Rewrite to use switch statemets.
20747 Handle 4-byte dual operands with TARGET_XOP and single operands
20748 with TARGET_SSSE3. Emit mmx_ppermv32 for TARGET_XOP and
20749 mmx_pshufbv4qi3 for TARGET_SSSE3.
20750 (expand_vec_perm_pblendv): Allow 4-byte vector modes with TARGET_SSE4_1.
20751 (expand_vec_perm_interleave2): Allow 4-byte vector modes.
20752 (expand_vec_perm_pshufb2): Allow 4-byte vector modes with TARGET_SSSE3.
20753 (expand_vec_perm_even_odd_1): Handle V4QI mode.
20754 (expand_vec_perm_broadcast_1): Handle V4QI mode.
20755 (ix86_vectorize_vec_perm_const): Handle V4QI mode.
20756 * config/i386/mmx.md (mmx_ppermv32): New insn pattern.
20757 (mmx_pshufbv4qi3): Ditto.
20758 (*mmx_pblendw32): Ditto.
20759 (*mmx_pblendw64): Rename from *mmx_pblendw.
20760 (mmx_punpckhbw_low): New insn_and_split pattern.
20761 (mmx_punpcklbw_low): Ditto.
20763 2021-07-05 Richard Biener <rguenther@suse.de>
20765 * tree-vect-loop-manip.c (vect_loop_versioning): Do not
20766 set LOOP_C_INFINITE on the vectorized loop.
20768 2021-07-05 Richard Biener <rguenther@suse.de>
20770 PR middle-end/101291
20771 * cfgloopmanip.c (loop_version): Set the loop copy of the
20772 versioned loop to the new loop.
20774 2021-07-04 Iain Sandoe <iain@sandoe.co.uk>
20777 * config.gcc: Ensure that Darwin biarch definitions are
20778 added before i386.h.
20779 * config/i386/darwin.h (TARGET_64BIT): Remove.
20780 (PR80556_WORKAROUND): New.
20781 (REAL_LIBGCC_SPEC): Amend to use PR80556_WORKAROUND.
20782 (DARWIN_SUBARCH_SPEC): New.
20783 * config/i386/darwin32-biarch.h (TARGET_64BIT_DEFAULT,
20784 TARGET_BI_ARCH, PR80556_WORKAROUND): New.
20785 (REAL_LIBGCC_SPEC): Remove.
20786 * config/i386/darwin64-biarch.h (TARGET_64BIT_DEFAULT,
20787 TARGET_BI_ARCH, PR80556_WORKAROUND): New.
20788 (REAL_LIBGCC_SPEC): Remove.
20790 2021-07-03 H.J. Lu <hjl.tools@gmail.com>
20792 PR middle-end/101294
20793 * expr.c (store_constructor): Don't use vec_duplicate on vector.
20795 2021-07-02 Martin Sebor <msebor@redhat.com>
20797 PR middle-end/98871
20798 PR middle-end/98512
20799 * diagnostic.c (get_any_inlining_info): New.
20800 (update_effective_level_from_pragmas): Handle inlining context.
20801 (diagnostic_enabled): Same.
20802 (diagnostic_report_diagnostic): Same.
20803 * diagnostic.h (struct diagnostic_info): Add ctor.
20804 (struct diagnostic_context): Add new member.
20805 * tree-diagnostic.c (set_inlining_locations): New.
20806 (tree_diagnostics_defaults): Set new callback pointer.
20808 2021-07-02 Peter Bergner <bergner@linux.ibm.com>
20810 * config/rs6000/rs6000-builtin.def (BU_MMA_PAIR_LD, BU_MMA_PAIR_ST):
20812 (__builtin_vsx_lxvp, __builtin_vsx_stxvp): New built-ins.
20813 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_mma_builtin): Expand
20814 lxvp and stxvp built-ins.
20815 (mma_init_builtins): Handle lxvp and stxvp built-ins.
20816 (builtin_function_type): Likewise.
20817 * doc/extend.texi (__builtin_vsx_lxvp, __builtin_mma_stxvp): Document.
20819 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
20821 * config/h8300/h8300-protos.h (compute_a_shift_cc): Accept
20822 additional argument for the code.
20823 * config/h8300/h8300.c (compute_a_shift_cc): Accept additional
20824 argument for the code. Just return if the ZN bits are useful or
20825 not rather than the old style CC_* enums.
20826 * config/h8300/shiftrotate.md (shiftqi_noscratch): Move before
20827 more generic shiftqi patterns.
20828 (shifthi_noscratch, shiftsi_noscratch): Similarly.
20829 (shiftqi_noscratch_set_flags): New pattern.
20830 (shifthi_noscratch_set_flags, shiftsi_noscratch_set_flags): Likewise.
20832 2021-07-02 Andrew MacLeod <amacleod@redhat.com>
20834 PR tree-optimization/101223
20835 * range-op.cc (build_lt): Add -1 for signed values.
20836 (built_gt): Subtract -1 for signed values.
20838 2021-07-02 David Faust <david.faust@oracle.com>
20840 * btfout.c (get_btf_kind): Support BTF_KIND_FLOAT.
20841 (btf_asm_type): Likewise.
20843 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
20845 * config/h8300/h8300-protos.h (output_a_shift): Make first argument
20846 an array of rtx rather than a pointer to rtx. Add code argument.
20847 (compute_a_shift_length): Similarly.
20848 * config/h8300/h8300.c (h8300_shift_costs): Adjust now that the
20849 shift itself isn't an operand. Create dummy operand[0] to carry
20850 a mode and pass a suitable rtx code to compute_a_shift_length.
20851 (get_shift_alg): Adjust operand number of clobber in output templates.
20852 (output_a_shift): Make first argument an array of rtx rather than
20853 a pointer to rtx. Add code argument for the type of shift.
20854 Adjust now that the shift itself is no longer an operand.
20855 (compute_a_shift_length): Similarly.
20856 * config/h8300/shiftrotate.md (shiftqi, shifthi, shiftsi): Use an
20857 iterator rather than nshift_operator.
20858 (shiftqi_noscratch, shifthi_noscratch, shiftsi_noscratch): Likewise.
20859 (shiftqi_clobber_flags): Adjust to API changes in output_a_shift
20860 and compute_a_shift_length.
20861 (shiftqi_noscratch_clobber_flags): Likewise.
20862 (shifthi_noscratch_clobber_flags): Likewise.
20863 (shiftsi_noscratch_clobber_flags): Likewise.
20865 2021-07-02 Iain Sandoe <iain@sandoe.co.uk>
20868 * config/darwin.h (DSYMUTIL_SPEC): Do not try to run
20869 dsymutil for BTF/CTF.
20871 2021-07-02 Iain Sandoe <iain@sandoe.co.uk>
20874 * config/darwin.h (CTF_INFO_SECTION_NAME): Update the
20875 segment to include BTF.
20876 (BTF_INFO_SECTION_NAME): New.
20878 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
20880 * config/m32r/m32r-protos.h (call_operand): Adjust return type.
20881 (small_data_operand, memreg_operand, small_insn_p): Likewise.
20882 * config/m32r/m32r.c (call_operand): Adjust return type.
20883 (small_data_operand, memreg_operand): Likewise.
20885 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
20887 * config/frv/frv-protos.h (integer_register_operand): Adjust return
20889 (frv_load_operand, gpr_or_fpr_operand, gpr_no_subreg_operand): Likewise.
20890 (fpr_or_int6_operand, gpr_or_int_operand); Likewise.
20891 (gpr_or_int12_operand, gpr_or_int10_operand); Likewise.
20892 (move_source_operand, move_destination_operand): Likewise.
20893 (condexec_source_operand, condexec_dest_operand): Likewise.
20894 (lr_operand, gpr_or_memory_operand, fpr_or_memory_operand): Likewise.
20895 (reg_or_0_operand, fcc_operand, icc_operand, cc_operand): Likewise.
20896 (fcr_operand, icr_operand, cr_operand, call_operand): Likewise.
20897 (fpr_operand, even_reg_operand, odd_reg_operand): Likewise.
20898 (even_gpr_operand, odd_gpr_operand, quad_fpr_operand): Likewise.
20899 (even_fpr_operand, odd_fpr_operand): Likewise.
20900 (dbl_memory_one_insn_operand, dbl_memory_two_insn_operand): Likewise.
20901 (int12_operand, int6_operand, int5_operand, uint5_operand): Likewise.
20902 (uint4_operand, uint1_operand, int_2word_operand): Likewise
20903 (upper_int16_operand, uint16_operand, symbolic_operand): Likewise.
20904 (relational_operator, float_relational_operator): Likewise.
20905 (ccr_eqne_operator, minmax_operator): Likewise.
20906 (condexec_si_binary_operator, condexec_si_media_operator): Likewise.
20907 (condexec_si_divide_operator, condexec_si_unary_operator): Likewise.
20908 (condexec_sf_conv_operator, condexec_sf_add_operator): Likewise.
20909 (intop_compare_operator, acc_operand, even_acc_operand): Likewise.
20910 (quad_acc_operand, accg_operand): Likewise.
20912 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
20914 * config/stormy16/stormy16-protos.h (xstormy16_below_100_symbol): Change
20915 return type to a bool.
20916 (nonimmediate_nonstack_operand): Likewise.
20917 (xstormy16_splittable_below100_operand): Likewise.
20918 * config/stormy16/stormy16.c (xstormy16_below_100_symbol): Fix
20920 (xstormy16_splittable_below100_operand): Likewise.
20922 2021-07-02 Richard Biener <rguenther@suse.de>
20924 PR tree-optimization/101293
20925 * tree-ssa-loop-im.c (mem_ref_hasher::equal): Compare MEM_REF bases
20926 with combined offsets.
20927 (gather_mem_refs_stmt): Hash MEM_REFs as if their offset were
20928 combined with the rest of the offset.
20930 2021-07-02 Eric Botcazou <ebotcazou@adacore.com>
20932 * config/i386/i386.c (asm_preferred_eh_data_format): Always use the
20933 PIC encodings for PE-COFF targets.
20935 2021-07-02 Jakub Jelinek <jakub@redhat.com>
20938 * config/i386/i386-expand.c (ix86_broadcast_from_integer_constant):
20939 Return nullptr for TImode inner mode.
20941 2021-07-02 Richard Biener <rguenther@suse.de>
20943 PR tree-optimization/101280
20944 PR tree-optimization/101173
20945 * gimple-loop-interchange.cc
20946 (tree_loop_interchange::valid_data_dependences): Properly
20947 guard all dependence checks with DDR_REVERSED_P or its
20950 2021-07-02 Hongyu Wang <hongyu.wang@intel.com>
20952 * config/i386/i386-expand.c (ix86_expand_builtin):
20953 Add branch to clear odata when ZF is set for asedecenc_expand
20954 and wideaesdecenc_expand.
20956 2021-07-02 Eugene Rozenfeld <erozen@microsoft.com>
20958 * config/i386/gcc-auto-profile: regenerate
20960 2021-07-02 liuhongt <hongtao.liu@intel.com>
20962 * config/i386/sse.md (trunc<mode><pmov_dst_4>2): Refined to ..
20963 (trunc<mode><pmov_dst_4_lower>2): this.
20965 2021-07-01 David Malcolm <dmalcolm@redhat.com>
20967 * diagnostic.h (diagnostic_context::m_file_cache): New field.
20968 * input.c (class fcache): Rename to...
20969 (class file_cache_slot): ...this, making most members private and
20970 prefixing fields with "m_".
20971 (file_cache_slot::get_file_path): New accessor.
20972 (file_cache_slot::get_use_count): New accessor.
20973 (file_cache_slot::missing_trailing_newline_p): New accessor.
20974 (file_cache_slot::inc_use_count): New.
20975 (fcache_buffer_size): Move to...
20976 (file_cache_slot::buffer_size): ...here.
20977 (fcache_line_record_size): Move to...
20978 (file_cache_slot::line_record_size): ...here.
20979 (fcache_tab): Delete, in favor of global_dc->m_file_cache.
20980 (fcache_tab_size): Move to file_cache::num_file_slots.
20981 (diagnostic_file_cache_init): Update for move of fcache_tab
20982 to global_dc->m_file_cache.
20983 (diagnostic_file_cache_fini): Likewise.
20984 (lookup_file_in_cache_tab): Convert to...
20985 (file_cache::lookup_file): ...this.
20986 (diagnostics_file_cache_forcibly_evict_file): Update for move of
20987 fcache_tab to global_dc->m_file_cache, moving most of
20988 implementation to...
20989 (file_cache::forcibly_evict_file): ...this new function and...
20990 (file_cache_slot::evict): ...this new function.
20991 (evicted_cache_tab_entry): Convert to...
20992 (file_cache::evicted_cache_tab_entry): ...this.
20993 (add_file_to_cache_tab): Convert to...
20994 (file_cache::add_file): ...this, moving bulk of implementation
20996 (file_cache_slot::create): ..this new function.
20997 (file_cache::file_cache): New.
20998 (file_cache::~file_cache): New.
20999 (lookup_or_add_file_to_cache_tab): Convert to...
21000 (file_cache::lookup_or_add_file): ..this new function.
21001 (fcache::fcache): Rename to...
21002 (file_cache_slot::file_cache_slot): ...this, adding "m_" prefixes
21004 (fcache::~fcache): Rename to...
21005 (file_cache_slot::~file_cache_slot): ...this, adding "m_" prefixes
21007 (needs_read): Convert to...
21008 (file_cache_slot::needs_read_p): ...this.
21009 (needs_grow): Convert to...
21010 (file_cache_slot::needs_grow_p): ...this.
21011 (maybe_grow): Convert to...
21012 (file_cache_slot::maybe_grow): ...this.
21013 (read_data): Convert to...
21014 (file_cache_slot::read_data): ...this.
21015 (maybe_read_data): Convert to...
21016 (file_cache_slot::maybe_read_data): ...this.
21017 (get_next_line): Convert to...
21018 (file_cache_slot::get_next_line): ...this.
21019 (goto_next_line): Convert to...
21020 (file_cache_slot::goto_next_line): ...this.
21021 (read_line_num): Convert to...
21022 (file_cache_slot::read_line_num): ...this.
21023 (location_get_source_line): Update for moving of globals to
21024 global_dc->m_file_cache.
21025 (location_missing_trailing_newline): Likewise.
21026 * input.h (class file_cache_slot): New forward decl.
21027 (class file_cache): New.
21029 2021-07-01 Michael Meissner <meissner@linux.ibm.com>
21031 * config/rs6000/rs6000.c (rs6000_maybe_emit_fp_cmove): Add IEEE
21032 128-bit floating point conditional move support.
21033 (have_compare_and_set_mask): Add IEEE 128-bit floating point
21035 * config/rs6000/rs6000.md (mov<mode>cc, IEEE128 iterator): New insn.
21036 (mov<mode>cc_p10, IEEE128 iterator): New insn.
21037 (mov<mode>cc_invert_p10, IEEE128 iterator): New insn.
21038 (fpmask<mode>, IEEE128 iterator): New insn.
21039 (xxsel<mode>, IEEE128 iterator): New insn.
21041 2021-07-01 Iain Sandoe <iain@sandoe.co.uk>
21044 * config/darwin.h (CTF_INFO_SECTION_NAME): New.
21046 2021-07-01 H.J. Lu <hjl.tools@gmail.com>
21048 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
21050 * config/i386/i386-protos.h (ix86_expand_vector_init_duplicate):
21052 * config/i386/sse.md (INT_BROADCAST_MODE): New mode iterator.
21053 (vec_duplicate<mode>): New expander.
21055 2021-07-01 H.J. Lu <hjl.tools@gmail.com>
21058 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
21060 (ix86_byte_broadcast): New function.
21061 (ix86_convert_const_wide_int_to_broadcast): Likewise.
21062 (ix86_expand_move): Convert CONST_WIDE_INT to broadcast if mode
21063 size is 16 bytes or bigger.
21064 (ix86_broadcast_from_integer_constant): New function.
21065 (ix86_expand_vector_move): Convert CONST_WIDE_INT and CONST_VECTOR
21066 to broadcast if mode size is 16 bytes or bigger.
21067 * config/i386/i386-protos.h (ix86_gen_scratch_sse_rtx): New
21069 * config/i386/i386.c (ix86_gen_scratch_sse_rtx): New function.
21071 2021-07-01 Uroš Bizjak <ubizjak@gmail.com>
21073 * config/i386/predicates.md (ix86_endbr_immediate_operand):
21074 Return true/false instead of 1/0.
21075 (movq_parallel): Ditto.
21077 2021-07-01 Uroš Bizjak <ubizjak@gmail.com>
21079 * recog.c (general_operand): Return true/false instead of 1/0.
21080 (register_operand): Ditto.
21081 (immediate_operand): Ditto.
21082 (const_int_operand): Ditto.
21083 (const_scalar_int_operand): Ditto.
21084 (const_double_operand): Ditto.
21085 (push_operand): Ditto.
21086 (pop_operand): Ditto.
21087 (memory_operand): Ditto.
21088 (indirect_operand): Ditto.
21090 2021-07-01 Uroš Bizjak <ubizjak@gmail.com>
21092 * genpreds.c (write_predicate_subfunction):
21093 Change the type of written subfunction to bool.
21094 (write_one_predicate_function):
21095 Change the type of written function to bool.
21096 (write_tm_preds_h): Ditto.
21097 * recog.h (*insn_operand_predicate_fn): Change the type to bool.
21098 * recog.c (general_operand): Change the type to bool.
21099 (address_operand): Ditto.
21100 (register_operand): Ditto.
21101 (pmode_register_operand): Ditto.
21102 (scratch_operand): Ditto.
21103 (immediate_operand): Ditto.
21104 (const_int_operand): Ditto.
21105 (const_scalar_int_operand): Ditto.
21106 (const_double_operand): Ditto.
21107 (nonimmediate_operand): Ditto.
21108 (nonmemory_operand): Ditto.
21109 (push_operand): Ditto.
21110 (pop_operand): Ditto.
21111 (memory_operand): Ditto.
21112 (indirect_operand): Ditto.
21113 (ordered_comparison_operator): Ditto.
21114 (comparison_operator): Ditto.
21115 * config/i386/i386-expand.c (ix86_expand_sse_cmp):
21116 Change the type of indirect predicate function to bool.
21117 * config/rs6000/rs6000.c (easy_vector_constant):
21118 Change the type to bool.
21119 * config/mips/mips-protos.h (m16_based_address_p):
21120 Change the type of operand 3 to bool.
21122 2021-07-01 Richard Biener <rguenther@suse.de>
21124 PR tree-optimization/101280
21125 PR tree-optimization/101173
21126 * gimple-loop-interchange.cc
21127 (tree_loop_interchange::valid_data_dependences): Revert
21128 previous change and instead correctly handle DDR_REVERSED_P
21131 2021-07-01 Richard Biener <rguenther@suse.de>
21133 PR tree-optimization/101278
21134 * tree-ssa-dse.c (dse_classify_store): First check for
21135 uses, then ignore stmt for chaining purposes.
21137 2021-07-01 Richard Biener <rguenther@suse.de>
21139 PR tree-optimization/100778
21140 * tree-vect-slp.c (vect_schedule_slp_node): Do not place trapping
21141 vectorized ops ahead of their scalar BB.
21143 2021-07-01 Uroš Bizjak <ubizjak@gmail.com>
21146 * config/i386/i386.md (*nabs<dwi>2_doubleword):
21147 New insn_and_split pattern.
21148 (*nabs<dwi>2_1): Ditto.
21149 * config/i386/i386-features.c
21150 (general_scalar_chain::compute_convert_gain):
21151 Handle (NEG (ABS (...))) RTX. Rewrite src code
21152 scanner as switch statement.
21153 (general_scalar_chain::convert_insn):
21154 Handle (NEG (ABS (...))) RTX.
21155 (general_scalar_to_vector_candidate_p):
21156 Detect (NEG (ABS (...))) RTX. Reorder case statements
21157 for (AND (NOT (...) ...)) fallthrough.
21159 2021-07-01 Richard Biener <rguenther@suse.de>
21161 PR tree-optimization/101178
21162 * tree-vect-slp.c (slpg_vertex::materialize): Remove.
21163 (slpg::perm_in): Add.
21164 (slpg::get_perm_in): Remove.
21165 (slpg::get_perm_materialized): Add.
21166 (vect_optimize_slp): Handle VEC_PERM nodes more optimally
21167 during permute propagation and materialization.
21169 2021-07-01 Jakub Jelinek <jakub@redhat.com>
21172 * dwarf2out.c (loc_list_from_tree_1): Handle COMPOUND_LITERAL_EXPR.
21174 2021-07-01 Jakub Jelinek <jakub@redhat.com>
21176 PR middle-end/94366
21177 * omp-low.c (lower_rec_input_clauses): Rename is_fp_and_or to
21178 is_truth_op, set it for TRUTH_*IF_EXPR regardless of new_var's type,
21179 use boolean_type_node instead of integer_type_node as NE_EXPR type.
21180 (lower_reduction_clauses): Likewise.
21182 2021-06-30 Hafiz Abid Qadeer <abidh@codesourcery.com>
21184 * config/gcn/gcn.c: Include dwarf2.h.
21185 (gcn_addr_space_debug): New function.
21186 (TARGET_ADDR_SPACE_DEBUG): New hook.
21188 2021-06-30 Hafiz Abid Qadeer <abidh@codesourcery.com>
21190 * common/config/gcn/gcn-common.c
21191 (gcn_option_optimization_table): Change OPT_fomit_frame_pointer to -O3.
21192 * config/gcn/gcn.c (gcn_expand_prologue): Prefer the frame pointer
21194 (gcn_expand_prologue): Prefer the frame pointer when emitting CFI.
21195 (gcn_frame_pointer_rqd): New function.
21196 (TARGET_FRAME_POINTER_REQUIRED): New hook.
21198 2021-06-30 Hafiz Abid Qadeer <abidh@codesourcery.com>
21200 * config/gcn/gcn.c (move_callee_saved_registers): Emit CFI notes for
21201 prologue register saves.
21202 (gcn_debug_unwind_info): Use UI_DWARF2.
21203 (gcn_dwarf_register_number): Map DWARF_LINK_REGISTER to DWARF PC.
21204 (gcn_dwarf_register_span): DWARF_LINK_REGISTER doesn't span.
21205 * config/gcn/gcn.h: (DWARF_FRAME_RETURN_COLUMN): New define.
21206 (DWARF_LINK_REGISTER): New define.
21207 (FIRST_PSEUDO_REGISTER): Increment.
21208 (FIXED_REGISTERS): Add entry for DWARF_LINK_REGISTER.
21209 (CALL_USED_REGISTERS): Likewise.
21210 (REGISTER_NAMES): Likewise.
21212 2021-06-30 Richard Biener <rguenther@suse.de>
21214 PR tree-optimization/101267
21215 * tree-vect-stmts.c (vect_check_scalar_mask): Adjust
21216 API and use SLP compatible interface of vect_is_simple_use.
21217 Reject not vectorized SLP defs for callers that do not support
21219 (vect_check_store_rhs): Handle masked stores and pass down
21220 the appropriate operator index.
21221 (vectorizable_call): Adjust.
21222 (vectorizable_store): Likewise.
21223 (vectorizable_load): Likewise. Handle SLP pecularity of
21225 (vect_is_simple_use): Remove special-casing of masked stores.
21227 2021-06-30 Tobias Burnus <tobias@codesourcery.com>
21229 * common.opt (foffload): Remove help as Driver only.
21230 * gcc.c (display_help): Add -foffload.
21232 2021-06-30 Tobias Burnus <tobias@codesourcery.com>
21234 * gcc.c (close_at_file, execute): Replace alloca by XALLOCAVEC.
21235 (check_offload_target_name): Fix splitting OFFLOAD_TARGETS into
21236 a candidate list; better inform no offload target is configured
21237 and fix hint extraction when passed target is not '\0' at [len].
21238 * common.opt (foffload): Add tailing '.'.
21239 (foffload-options): Likewise; fix flag name in the help string.
21241 2021-06-30 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
21244 * config/arm/arm_neon.h: Move vabs intrinsics before vcage_f32.
21245 (vcage_f32): Gate comparison on __FAST_MATH__.
21246 (vcageq_f32): Likewise.
21247 (vcale_f32): Likewise.
21248 (vcaleq_f32): Likewise.
21249 (vcagt_f32): Likewise.
21250 (vcagtq_f32): Likewise.
21251 (vcalt_f32): Likewise.
21252 (vcaltq_f32): Likewise.
21253 (vcage_f16): Likewise.
21254 (vcageq_f16): Likewise.
21255 (vcale_f16): Likewise.
21256 (vcaleq_f16): Likewise.
21257 (vcagt_f16): Likewise.
21258 (vcagtq_f16): Likewise.
21259 (vcalt_f16): Likewise.
21260 (vcaltq_f16): Likewise.
21262 2021-06-30 Richard Biener <rguenther@suse.de>
21264 PR tree-optimization/101264
21265 * tree-vect-slp.c (vect_optimize_slp): Propagate the
21266 computed perm_in to all "any" permute successors
21267 we cannot de-duplicate immediately.
21269 2021-06-30 liuhongt <hongtao.liu@intel.com>
21272 * config/i386/sse.md
21273 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>):
21275 (avx512f_sfixupimm<mode><maskz_scalar_name><round_saeonly_name>):
21277 (avx512f_sfixupimm<mode>_mask<round_saeonly_name>"): Refined.
21278 * config/i386/subst.md (maskz_scalar): New define_subst.
21279 (maskz_scalar_name): New subst_attr.
21280 (maskz_scalar_op5): Ditto.
21281 (round_saeonly_maskz_scalar_op5): Ditto.
21282 (round_saeonly_maskz_scalar_operand5): Ditto.
21284 2021-06-30 David Edelsohn <dje.gcc@gmail.com>
21286 * config/rs6000/rs6000.c (rs6000_xcoff_section_type_flags):
21287 Increase code CSECT alignment to at least 32 bytes.
21288 * config/rs6000/xcoff.h (TEXT_SECTION_ASM_OP): Add 32 byte
21289 alignment designation.
21291 2021-06-29 Sergei Trofimovich <siarheit@google.com>
21293 * doc/generic.texi: Fix s/net yet/not yet/ typo.
21295 2021-06-29 Andrew MacLeod <amacleod@redhat.com>
21297 PR tree-optimization/101254
21298 * range-op.cc (operator_minus::op1_op2_relation_effect): Check for
21299 wrapping/non-wrapping when setting the result range.
21301 2021-06-29 Andrew MacLeod <amacleod@redhat.com>
21303 * value-query.cc (gimple_range_global): Allow phis.
21305 2021-06-29 Andrew MacLeod <amacleod@redhat.com>
21307 * vr-values.c (vr_values::vrp_stmt_computes_nonzero): Use stmt.
21308 (simplify_using_ranges::op_with_boolean_value_range_p): Add a
21309 statement for location context.
21310 (check_for_binary_op_overflow): Ditto.
21311 (simplify_using_ranges::get_vr_for_comparison): Ditto.
21312 (simplify_using_ranges::compare_name_with_value): Ditto.
21313 (simplify_using_ranges::compare_names): Ditto.
21314 (vrp_evaluate_conditional_warnv_with_ops_using_ranges): Ditto.
21315 (simplify_using_ranges::simplify_truth_ops_using_ranges): Ditto.
21316 (simplify_using_ranges::simplify_min_or_max_using_ranges): Ditto.
21317 (simplify_using_ranges::simplify_internal_call_using_ranges): Ditto.
21318 (simplify_using_ranges::two_valued_val_range_p): Ditto.
21319 (simplify_using_ranges::simplify): Ditto.
21320 * vr-values.h: Adjust prototypes.
21322 2021-06-29 Uroš Bizjak <ubizjak@gmail.com>
21325 * config/i386/mmx.md (vec_addsubv2sf3): New insn pattern.
21327 2021-06-29 Julian Brown <julian@codesourcery.com>
21329 * config/gcn/gcn.c (gcn_init_libfuncs): New function.
21330 (TARGET_INIT_LIBFUNCS): Define target hook using above function.
21331 * config/gcn/gcn.h (UNITS_PER_WORD): Define to 8 for IN_LIBGCC2, 4
21333 (LIBGCC2_UNITS_PER_WORD, BITS_PER_WORD): Remove definitions.
21334 (MAX_FIXED_MODE_SIZE): Change to 128.
21336 2021-06-29 Julian Brown <julian@codesourcery.com>
21338 * config/gcn/gcn.md (UNSPEC_FLBIT_INT): New unspec constant.
21339 (s_mnemonic): Add clrsb.
21340 (gcn_flbit<mode>_int): Add insn pattern for SImode/DImode.
21341 (clrsb<mode>2): Add expander for SImode/DImode.
21343 2021-06-29 Julian Brown <julian@codesourcery.com>
21345 * config/gcn/gcn.md (<su>mulsidi3, <su>mulsidi3_reg, <su>mulsidi3_imm,
21346 muldi3): Add patterns.
21348 2021-06-29 Julian Brown <julian@codesourcery.com>
21350 * config/gcn/gcn.md (<su>mulsi3_highpart): Change to expander.
21351 (<su>mulsi3_highpart_reg, <su>mulsi3_highpart_imm): New patterns.
21353 2021-06-29 Julian Brown <julian@codesourcery.com>
21355 * config/gcn/gcn.md (mulsi3): Make s_mulk_i32 variant clobber SCC.
21357 2021-06-29 Joseph Myers <joseph@codesourcery.com>
21359 * btfout.c, ctfout.c: Include "memmodel.h".
21361 2021-06-29 Tobias Burnus <tobias@codesourcery.com>
21363 * gcc.c (check_offload_target_name): Cast len argument to
21364 %q.*s to 'int'; avoid -Wstringop-truncation warning.
21366 2021-06-29 Richard Biener <rguenther@suse.de>
21368 * tree-vect-slp.c (vect_optimize_slp): Forward propagate
21369 to "any" permute nodes and relax "any" permute proapgation
21370 during iterative backward propagation.
21372 2021-06-29 Tobias Burnus <tobias@codesourcery.com>
21375 * common.opt (-foffload=): Update description.
21376 (-foffload-options=): New.
21377 * doc/invoke.texi (C Language Options): Document
21378 -foffload and -foffload-options.
21379 * gcc.c (check_offload_target_name): New, split off from
21380 handle_foffload_option.
21381 (check_foffload_target_names): New.
21382 (handle_foffload_option): Handle -foffload=default.
21383 (driver_handle_option): Update for -foffload-options.
21384 * lto-opts.c (lto_write_options): Use -foffload-options
21385 instead of -foffload.
21386 * lto-wrapper.c (merge_and_complain, append_offload_options):
21388 * opts.c (common_handle_option): Likewise.
21390 2021-06-29 Tobias Burnus <tobias@codesourcery.com>
21392 * doc/invoke.texi (C Language Options): Sort options
21393 alphabetically in optlist and also the description itself.
21394 Remove leftover -fallow-single-precision from and add missing
21395 -fgnu-tm to the optlist.
21397 2021-06-29 Richard Biener <rguenther@suse.de>
21399 * tree-vect-slp.c (slpg_vertex::visited): Remove.
21400 (vect_slp_perms_eq): Handle -1 permutes.
21401 (vect_optimize_slp): Rewrite permute propagation.
21403 2021-06-29 Jakub Jelinek <jakub@redhat.com>
21406 * match.pd ((intptr_t)x eq/ne CST to x eq/ne (typeof x) CST): Don't
21407 perform the optimization in GENERIC when sanitizing and x has a
21410 2021-06-29 Richard Biener <rguenther@suse.de>
21412 PR tree-optimization/101242
21413 * tree-vect-slp.c (vect_slp_build_vertices): Force-add
21414 PHIs with not represented initial values as leafs.
21416 2021-06-29 Jan-Benedict Glaw <jbglaw@getslash.de>
21418 * config/pdp11/pdp11.h (ASM_OUTPUT_SKIP): Fix signedness warning.
21419 * config/pdp11/pdp11.c (pdp11_asm_print_operand_punct_valid_p): Remove
21420 "register" keyword.
21421 (pdp11_initial_elimination_offset) Remove unused variable.
21422 (pdp11_cmp_length) Ditto.
21423 (pdp11_insn_cost): Ditto, and fix signedness warning.
21425 2021-06-29 David Edelsohn <dje.gcc@gmail.com>
21427 * btfout.c: Include tm_p.h.
21430 2021-06-28 Indu Bhagat <indu.bhagat@oracle.com>
21432 * config/bpf/bpf.c (bpf_expand_prologue): Do not mark insns as
21434 (bpf_expand_epilogue): Likewise.
21435 * config/bpf/bpf.h (DWARF2_FRAME_INFO): Define to 0.
21436 Do not define DBX_DEBUGGING_INFO.
21438 2021-06-28 Indu Bhagat <indu.bhagat@oracle.com>
21440 * doc/invoke.texi: Document the CTF and BTF debug info options.
21442 2021-06-28 Indu Bhagat <indu.bhagat@oracle.com>
21443 David Faust <david.faust@oracle.com>
21444 Jose E. Marchesi <jose.marchesi@oracle.com>
21445 Weimin Pan <weimin.pan@oracle.com>
21447 * Makefile.in: Add ctfc.*, ctfout.c and btfout.c files to
21448 GTFILES. Add new object files.
21449 * common.opt: Add CTF and BTF debug info options.
21450 * btfout.c: New file.
21451 * ctfc.c: Likewise.
21452 * ctfc.h: Likewise.
21453 * ctfout.c: Likewise.
21454 * dwarf2ctf.c: Likewise.
21455 * dwarf2ctf.h: Likewise.
21456 * dwarf2cfi.c (dwarf2out_do_frame): Acknowledge CTF_DEBUG and
21458 * dwarf2out.c (dwarf2out_source_line): Likewise.
21459 (dwarf2out_finish): Skip emitting DWARF if CTF or BTF are to
21461 (debug_format_do_cu): New function.
21462 (dwarf2out_early_finish): Traverse DIEs and emit CTF/BTF for
21464 Include dwarf2ctf.c.
21465 * final.c (dwarf2_debug_info_emitted_p): Acknowledge DWARF-based debug
21467 * flag-types.h (enum debug_info_type): Add CTF_DEBUG and BTF_DEBUG.
21468 (CTF_DEBUG): New bitmask.
21469 (BTF_DEBUG): Likewise.
21470 (enum ctf_debug_info_levels): New enum.
21471 * gengtype.c (open_base_files): Handle ctfc.h.
21472 (main): Handle uint32_t type.
21473 * flags.h (btf_debuginfo_p): New definition.
21474 (dwarf_based_debuginfo_p): Likewise.
21475 * opts.c (debug_type_names): Add entries for CTF and BTF.
21476 (btf_debuginfo_p): New function.
21477 (dwarf_based_debuginfo_p): Likewise.
21478 (common_handle_option): Handle -gctfN and -gbtf options.
21479 (set_debug_level): Set CTF_DEBUG, BTF_DEBUG whenever appropriate.
21480 * toplev.c (process_options): Inform the user and ignore -gctfLEVEL if
21483 2021-06-28 Jose E. Marchesi <jose.marchesi@oracle.com>
21485 * dwarf2out.c (AT_class): Function is no longer static.
21486 (AT_int): Likewise.
21487 (AT_unsigned): Likewise.
21488 (AT_loc): Likewise.
21489 (get_AT): Likewise.
21490 (get_AT_string): Likewise.
21491 (get_AT_flag): Likewise.
21492 (get_AT_unsigned): Likewise.
21493 (get_AT_ref): Likewise.
21494 (new_die_raw): Likewise.
21495 (lookup_decl_die): Likewise.
21496 (base_type_die): Likewise.
21497 (add_name_attribute): Likewise.
21498 (add_AT_int): Likewise.
21499 (add_AT_unsigned): Likewise.
21500 (add_AT_loc): Likewise.
21501 (dw_get_die_tag): New function.
21502 (dw_get_die_child): Likewise.
21503 (dw_get_die_sib): Likewise.
21504 (struct dwarf_file_data): Move from here to dwarf2out.h
21505 (struct dw_attr_struct): Likewise.
21506 * dwarf2out.h: Analogous changes.
21508 2021-06-28 Martin Jambor <mjambor@suse.cz>
21511 * ipa-param-manipulation.h (class ipa_param_body_adjustments): New
21512 members m_dead_stmts and m_dead_ssas.
21513 * ipa-param-manipulation.c
21514 (ipa_param_body_adjustments::mark_dead_statements): New function.
21515 (ipa_param_body_adjustments::common_initialization): Call it on
21516 all removed but not split parameters.
21517 (ipa_param_body_adjustments::ipa_param_body_adjustments): Initialize
21519 (ipa_param_body_adjustments::modify_call_stmt): Remove arguments that
21521 * tree-inline.c (remap_gimple_stmt): Do not copy dead statements, reset
21522 dead debug statements.
21523 (copy_phis_for_bb): Do not copy dead PHI nodes.
21525 2021-06-28 Martin Jambor <mjambor@suse.cz>
21528 * symtab-clones.h (clone_info): Removed member param_adjustments.
21529 * ipa-param-manipulation.h: Adjust initial comment to reflect how we
21530 deal with pass-through splits now.
21531 (ipa_param_performed_split): Removed.
21532 (ipa_param_adjustments::modify_call): Adjusted parameters.
21533 (class ipa_param_body_adjustments): Adjusted parameters of
21534 register_replacement, modify_gimple_stmt and modify_call_stmt.
21535 (ipa_verify_edge_has_no_modifications): Declare.
21536 (ipa_edge_modifications_finalize): Declare.
21537 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Remove
21538 performed_splits processing, pas only edge to padjs->modify_call,
21539 check that call arguments were not modified if they should not have
21541 * cgraphclones.c (cgraph_node::create_clone): Do not copy performed
21543 * ipa-param-manipulation.c (struct pass_through_split_map): New type.
21544 (ipa_edge_modification_info): Likewise.
21545 (ipa_edge_modification_sum): Likewise.
21546 (ipa_edge_modifications): New edge summary.
21547 (ipa_verify_edge_has_no_modifications): New function.
21548 (transitive_split_p): Removed.
21549 (transitive_split_map): Likewise.
21550 (init_transitive_splits): Likewise.
21551 (ipa_param_adjustments::modify_call): Adjusted to use the new edge
21552 summary instead of performed_splits.
21553 (ipa_param_body_adjustments::register_replacement): Drop dummy
21554 parameter, set base_index of the created ipa_param_body_replacement.
21555 (phi_arg_will_live_p): New function.
21556 (ipa_param_body_adjustments::common_initialization): Do not create
21557 IPA_SRA dummy decls.
21558 (simple_tree_swap_info): Removed.
21559 (remap_split_decl_to_dummy): Likewise.
21560 (record_argument_state_1): New function.
21561 (record_argument_state): Likewise.
21562 (ipa_param_body_adjustments::modify_call_stmt): New parameter
21563 orig_stmt. Do not work with dummy decls, save necessary info about
21564 changes to ipa_edge_modifications.
21565 (ipa_param_body_adjustments::modify_gimple_stmt): New parameter
21566 orig_stmt, pass it to modify_call_stmt.
21567 (ipa_param_body_adjustments::modify_cfun_body): Adjust call to
21568 modify_gimple_stmt.
21569 (ipa_edge_modifications_finalize): New function.
21570 * tree-inline.c (remap_gimple_stmt): Pass original statement to
21571 modify_gimple_stmt.
21572 (copy_phis_for_bb): Do not copy dead PHI nodes.
21573 (expand_call_inline): Do not remap performed_splits.
21574 (update_clone_info): Likewise.
21575 * toplev.c: Include ipa-param-manipulation.h.
21576 (toplev::finalize): Call ipa_edge_modifications_finalize.
21578 2021-06-28 Andrew Pinski <apinski@marvell.com>
21580 * tree-ssa-phiopt.c (replace_phi_edge_with_variable): Duplicate range
21581 info if we're the only things setting the target PHI.
21582 (value_replacement): Don't duplicate range here.
21583 (minmax_replacement): Likewise.
21585 2021-06-28 Richard Biener <rguenther@suse.de>
21587 PR tree-optimization/101229
21588 * gimple-walk.c (gimple_walk_op): Handle PHIs.
21590 2021-06-28 Martin Liska <mliska@suse.cz>
21592 * config/v850/v850.c (construct_dispose_instruction): Allocate
21594 (construct_prepare_instruction): Likewise.
21596 2021-06-28 Martin Liska <mliska@suse.cz>
21598 * config/v850/v850.c (v850_option_override): Build default
21600 (v850_can_inline_p): New. Allow MASK_PROLOG_FUNCTION to be
21601 ignored for inlining.
21602 (TARGET_CAN_INLINE_P): New.
21604 2021-06-28 Richard Biener <rguenther@suse.de>
21606 PR tree-optimization/101207
21607 * tree-vect-slp.c (vect_optimize_slp): Do BB reduction
21608 permute eliding for load permutations properly.
21610 2021-06-28 Richard Biener <rguenther@suse.de>
21612 PR tree-optimization/101173
21613 * gimple-loop-interchange.cc
21614 (tree_loop_interchange::valid_data_dependences): Disallow outer
21615 loop dependence distance of zero.
21617 2021-06-28 liuhongt <hongtao.liu@intel.com>
21620 * config/i386/sse.md (*avx_cmp<mode>3_lt): New
21621 define_insn_and_split.
21622 (*avx_cmp<mode>3_ltint): Ditto.
21623 (*avx2_pcmp<mode>3_3): Ditto.
21624 (*avx2_pcmp<mode>3_4): Ditto.
21625 (*avx2_pcmp<mode>3_5): Ditto.
21627 2021-06-28 liuhongt <hongtao.liu@intel.com>
21629 * config/i386/i386-builtin.def (IX86_BUILTIN_BLENDVPD256,
21630 IX86_BUILTIN_BLENDVPS256, IX86_BUILTIN_PBLENDVB256,
21631 IX86_BUILTIN_BLENDVPD, IX86_BUILTIN_BLENDVPS,
21632 IX86_BUILTIN_PBLENDVB128): Replace icode with
21634 * config/i386/i386.c (ix86_gimple_fold_builtin): Fold blendv
21636 * config/i386/sse.md (*<sse4_1_avx2>_pblendvb_lt_subreg_not):
21637 New pre_reload splitter.
21639 2021-06-27 Andrew Pinski <apinski@marvell.com>
21641 PR middle-end/101230
21642 * fold-const.c (fold_ternary_loc): Check
21643 the return value of invert_tree_comparison.
21645 2021-06-27 David Edelsohn <dje.gcc@gmail.com>
21647 * config.gcc: Add SPDX License Identifier.
21648 (powerpc-ibm-aix789): Default to aix73.h.
21649 (powerpc-ibm-aix7.2.*.*): New stanza.
21650 * config/rs6000/aix72.h: Add SPDX License Identifier.
21651 * config/rs6000/aix73.h: New file.
21653 2021-06-26 Jason Merrill <jason@redhat.com>
21655 * except.c: #include "dwarf2.h" instead of "dwarf2out.h".
21657 2021-06-26 Andrew Pinski <apinski@marvell.com>
21659 * genmatch.c (lower_cond): Copy for_subst_vec
21660 for the simplify also.
21661 (lower): Swap the order for lower_for and lower_cond.
21663 2021-06-26 Andrew Pinski <apinski@marvell.com>
21665 * tree-ssa-phiopt.c (match_simplify_replacement): Reset
21666 flow senatitive info on the moved ssa set.
21668 2021-06-26 Andrew Pinski <apinski@marvell.com>
21670 * fold-const.c (fold_cond_expr_with_comparison):
21671 Exand arg0 into comp_code, arg00, and arg01.
21672 (fold_ternary_loc): Use invert_tree_comparison
21673 instead of fold_invert_truthvalue for the case
21674 where we have A CMP B ? C : A.
21676 2021-06-25 Martin Sebor <msebor@redhat.com>
21678 PR middle-end/101216
21679 * calls.c (maybe_warn_rdwr_sizes): Use the no_warning constant.
21681 2021-06-25 Jeff Law <jeffreyalaw@gmail.com>
21683 * config/h8300/h8300.c (select_cc_mode): Handle ASHIFTRT and LSHIFTRT.
21685 2021-06-25 Richard Biener <rguenther@suse.de>
21687 PR tree-optimization/101202
21688 * tree-vect-slp.c (vect_optimize_slp): Explicitely handle
21691 2021-06-25 Richard Biener <rguenther@suse.de>
21693 * tree-vect-slp-patterns.c (addsub_pattern::build): Copy
21694 STMT_VINFO_REDUC_DEF from the original representative.
21696 2021-06-25 Martin Sebor <msebor@redhat.com>
21698 * builtins.c (warn_string_no_nul): Replace uses of TREE_NO_WARNING,
21699 gimple_no_warning_p and gimple_set_no_warning with
21700 warning_suppressed_p, and suppress_warning.
21702 (maybe_warn_for_bound): Same.
21703 (warn_for_access): Same.
21704 (check_access): Same.
21705 (expand_builtin_strncmp): Same.
21706 (fold_builtin_varargs): Same.
21707 * calls.c (maybe_warn_nonstring_arg): Same.
21708 (maybe_warn_rdwr_sizes): Same.
21709 * cfgexpand.c (expand_call_stmt): Same.
21710 * cgraphunit.c (check_global_declaration): Same.
21711 * fold-const.c (fold_undefer_overflow_warnings): Same.
21712 (fold_truth_not_expr): Same.
21713 (fold_unary_loc): Same.
21714 (fold_checksum_tree): Same.
21715 * gimple-array-bounds.cc (array_bounds_checker::check_array_ref): Same.
21716 (array_bounds_checker::check_mem_ref): Same.
21717 (array_bounds_checker::check_addr_expr): Same.
21718 (array_bounds_checker::check_array_bounds): Same.
21719 * gimple-expr.c (copy_var_decl): Same.
21720 * gimple-fold.c (gimple_fold_builtin_strcpy): Same.
21721 (gimple_fold_builtin_strncat): Same.
21722 (gimple_fold_builtin_stxcpy_chk): Same.
21723 (gimple_fold_builtin_stpcpy): Same.
21724 (gimple_fold_builtin_sprintf): Same.
21725 (fold_stmt_1): Same.
21726 * gimple-ssa-isolate-paths.c (diag_returned_locals): Same.
21727 * gimple-ssa-nonnull-compare.c (do_warn_nonnull_compare): Same.
21728 * gimple-ssa-sprintf.c (handle_printf_call): Same.
21729 * gimple-ssa-store-merging.c (imm_store_chain_info::output_merged_store): Same.
21730 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Same.
21731 * gimple-ssa-warn-restrict.h: Adjust declarations.
21732 (maybe_diag_access_bounds): Replace uses of TREE_NO_WARNING,
21733 gimple_no_warning_p and gimple_set_no_warning with
21734 warning_suppressed_p, and suppress_warning.
21735 (check_call): Same.
21736 (check_bounds_or_overlap): Same.
21737 * gimple.c (gimple_build_call_from_tree): Same.
21738 * gimplify.c (gimplify_return_expr): Same.
21739 (gimplify_cond_expr): Same.
21740 (gimplify_modify_expr_complex_part): Same.
21741 (gimplify_modify_expr): Same.
21742 (gimple_push_cleanup): Same.
21743 (gimplify_expr): Same.
21744 * omp-expand.c (expand_omp_for_generic): Same.
21745 (expand_omp_taskloop_for_outer): Same.
21746 * omp-low.c (lower_rec_input_clauses): Same.
21747 (lower_lastprivate_clauses): Same.
21748 (lower_send_clauses): Same.
21749 (lower_omp_target): Same.
21750 * tree-cfg.c (pass_warn_function_return::execute): Same.
21751 * tree-complex.c (create_one_component_var): Same.
21752 * tree-inline.c (remap_gimple_op_r): Same.
21753 (copy_tree_body_r): Same.
21754 (declare_return_variable): Same.
21755 (expand_call_inline): Same.
21756 * tree-nested.c (lookup_field_for_decl): Same.
21757 * tree-sra.c (create_access_replacement): Same.
21758 (generate_subtree_copies): Same.
21759 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Same.
21760 * tree-ssa-forwprop.c (combine_cond_expr_cond): Same.
21761 * tree-ssa-loop-ch.c (ch_base::copy_headers): Same.
21762 * tree-ssa-loop-im.c (execute_sm): Same.
21763 * tree-ssa-phiopt.c (cond_store_replacement): Same.
21764 * tree-ssa-strlen.c (maybe_warn_overflow): Same.
21765 (handle_builtin_strcpy): Same.
21766 (maybe_diag_stxncpy_trunc): Same.
21767 (handle_builtin_stxncpy_strncat): Same.
21768 (handle_builtin_strcat): Same.
21769 * tree-ssa-uninit.c (get_no_uninit_warning): Same.
21770 (set_no_uninit_warning): Same.
21771 (uninit_undefined_value_p): Same.
21772 (warn_uninit): Same.
21773 (maybe_warn_operand): Same.
21774 * tree-vrp.c (compare_values_warnv): Same.
21775 * vr-values.c (vr_values::extract_range_for_var_from_comparison_expr): Same.
21776 (test_for_singularity): Same.
21777 * gimple.h (warning_suppressed_p): New function.
21778 (suppress_warning): Same.
21779 (copy_no_warning): Same.
21780 (gimple_set_block): Call gimple_set_location.
21781 (gimple_set_location): Call copy_warning.
21783 2021-06-25 Martin Sebor <msebor@redhat.com>
21785 * tree.h (warning_suppressed_at, copy_warning,
21786 warning_suppressed_p, suppress_warning): New functions.
21788 2021-06-25 Martin Sebor <msebor@redhat.com>
21790 * Makefile.in (OBJS-libcommon): Add diagnostic-spec.o.
21791 * gengtype.c (open_base_files): Add diagnostic-spec.h.
21792 * diagnostic-spec.c: New file.
21793 * diagnostic-spec.h: New file.
21794 * tree.h (no_warning, all_warnings, suppress_warning_at): New
21796 * warning-control.cc: New file.
21798 2021-06-25 liuhongt <hongtao.liu@intel.com>
21801 * config/i386/i386.c (x86_order_regs_for_local_alloc):
21804 2021-06-24 Andrew MacLeod <amacleod@redhat.com>
21806 PR tree-optimization/101189
21807 * gimple-range-fold.cc (fold_using_range::range_of_range_op): Pass
21808 LHS range of condition to postfold routine.
21809 (fold_using_range::postfold_gcond_edges): Only process the TRUE or
21810 FALSE edge if the LHS range supports it being taken.
21811 * gimple-range-fold.h (postfold_gcond_edges): Add range parameter.
21813 2021-06-24 Andrew MacLeod <amacleod@redhat.com>
21815 * value-relation.cc (equiv_oracle::dump): Do not dump NULL blocks.
21816 (relation_oracle::find_relation_block): Check correct bitmap.
21817 (relation_oracle::dump): Do not dump NULL blocks.
21819 2021-06-24 Andrew MacLeod <amacleod@redhat.com>
21821 * gimple-range-cache.cc (ranger_cache::propagate_cache): Call
21822 range_on_edge instead of manually calculating.
21824 2021-06-24 Andrew MacLeod <amacleod@redhat.com>
21826 * range-op.cc: Fix comment.
21828 2021-06-24 Uroš Bizjak <ubizjak@gmail.com>
21831 * config/i386/i386-expand.c (ix86_expand_sse_unpack):
21832 Handle V8QI and V4HI modes.
21833 * config/i386/mmx.md (sse4_1_<any_extend:code>v4qiv4hi2):
21835 (sse4_1_<any_extend:code>v4qiv4hi2): Ditto.
21836 (mmxpackmode): New mode attribute.
21837 (vec_pack_trunc_<mmxpackmode:mode>): New expander.
21838 (mmxunpackmode): New mode attribute.
21839 (vec_unpacks_lo_<mmxunpackmode:mode>): New expander.
21840 (vec_unpacks_hi_<mmxunpackmode:mode>): Ditto.
21841 (vec_unpacku_lo_<mmxunpackmode:mode>): Ditto.
21842 (vec_unpacku_hi_<mmxunpackmode:mode>): Ditto.
21843 * config/i386/i386.md (extsuffix): Move from ...
21844 * config/i386/sse.md: ... here.
21846 2021-06-24 Eric Botcazou <ebotcazou@adacore.com>
21848 * dwarf2out.c (dwarf2out_assembly_start): Emit .file 0 marker here..
21849 (dwarf2out_finish): ...instead of here.
21851 2021-06-24 Eric Botcazou <ebotcazou@adacore.com>
21853 * configure.ac (--gdwarf-5 option): Use objdump instead of readelf.
21854 (working --gdwarf-4/--gdwarf-5 for all sources): Likewise.
21855 (--gdwarf-4 not refusing generated .debug_line): Adjust for Windows.
21856 * configure: Regenerate.
21858 2021-06-24 Richard Biener <rguenther@suse.de>
21860 * config/i386/sse.md (vec_addsubv4df3, vec_addsubv2df3,
21861 vec_addsubv8sf3, vec_addsubv4sf3): Merge into ...
21862 (vec_addsub<mode>3): ... using a new addsub_cst mode attribute.
21864 2021-06-24 Richard Biener <rguenther@suse.de>
21866 * config/i386/sse.md (avx_addsubv4df3): Rename to
21868 (avx_addsubv8sf3): Rename to vec_addsubv8sf3.
21869 (sse3_addsubv2df3): Rename to vec_addsubv2df3.
21870 (sse3_addsubv4sf3): Rename to vec_addsubv4sf3.
21871 * config/i386/i386-builtin.def: Adjust.
21872 * internal-fn.def (VEC_ADDSUB): New internal optab fn.
21873 * optabs.def (vec_addsub_optab): New optab.
21874 * tree-vect-slp-patterns.c (class addsub_pattern): New.
21875 (slp_patterns): Add addsub_pattern.
21876 * tree-vect-slp.c (vect_optimize_slp): Disable propagation
21877 across CFN_VEC_ADDSUB.
21878 * tree-vectorizer.h (vect_pattern::vect_pattern): Make
21880 * doc/md.texi (vec_addsub<mode>3): Document.
21882 2021-06-24 Jakub Jelinek <jakub@redhat.com>
21884 PR middle-end/101170
21885 * df-scan.c (df_ref_record): For paradoxical big-endian SUBREGs
21886 where regno + subreg_regno_offset wraps around use 0 as starting
21889 2021-06-24 Jakub Jelinek <jakub@redhat.com>
21891 PR middle-end/101172
21892 * stor-layout.c (finish_bitfield_representative): If nextf has
21893 error_mark_node type, set repr type to error_mark_node too.
21895 2021-06-24 Ilya Leoshkevich <iii@linux.ibm.com>
21897 * config/s390/s390.c (s390_function_profiler): Ignore labelno
21899 * config/s390/s390.h (NO_PROFILE_COUNTERS): Define.
21901 2021-06-24 Richard Biener <rguenther@suse.de>
21903 * tree-vect-slp.c (vect_optimize_slp): Do not propagate
21904 across operations that have different semantics on different
21907 2021-06-24 Jakub Jelinek <jakub@redhat.com>
21909 * tree.h (OMP_CLAUSE_MAP_IN_REDUCTION): Document meaning for OpenMP.
21910 * gimplify.c (gimplify_scan_omp_clauses): For OpenMP map clauses
21911 with OMP_CLAUSE_MAP_IN_REDUCTION flag partially defer gimplification
21912 of non-decl OMP_CLAUSE_DECL. For OMP_CLAUSE_IN_REDUCTION on
21913 OMP_TARGET user outer_ctx instead of ctx for placeholders and
21914 initializer/combiner gimplification.
21915 * omp-low.c (scan_sharing_clauses): Handle OMP_CLAUSE_MAP_IN_REDUCTION
21916 on target constructs.
21917 (lower_rec_input_clauses): Likewise.
21918 (lower_omp_target): Likewise.
21919 * omp-expand.c (expand_omp_target): Temporarily ignore nowait clause
21920 on target if in_reduction is present.
21922 2021-06-24 Kewen Lin <linkw@linux.ibm.com>
21924 * tree-predcom.c (class pcom_worker): New class.
21925 (release_chain): Renamed to...
21926 (pcom_worker::release_chain): ...this.
21927 (release_chains): Renamed to...
21928 (pcom_worker::release_chains): ...this.
21929 (aff_combination_dr_offset): Renamed to...
21930 (pcom_worker::aff_combination_dr_offset): ...this.
21931 (determine_offset): Renamed to...
21932 (pcom_worker::determine_offset): ...this.
21933 (class comp_ptrs): New class.
21934 (split_data_refs_to_components): Renamed to...
21935 (pcom_worker::split_data_refs_to_components): ...this,
21936 and update with class comp_ptrs.
21937 (suitable_component_p): Renamed to...
21938 (pcom_worker::suitable_component_p): ...this.
21939 (filter_suitable_components): Renamed to...
21940 (pcom_worker::filter_suitable_components): ...this.
21941 (valid_initializer_p): Renamed to...
21942 (pcom_worker::valid_initializer_p): ...this.
21943 (find_looparound_phi): Renamed to...
21944 (pcom_worker::find_looparound_phi): ...this.
21945 (add_looparound_copies): Renamed to...
21946 (pcom_worker::add_looparound_copies): ...this.
21947 (determine_roots_comp): Renamed to...
21948 (pcom_worker::determine_roots_comp): ...this.
21949 (determine_roots): Renamed to...
21950 (pcom_worker::determine_roots): ...this.
21951 (single_nonlooparound_use): Renamed to...
21952 (pcom_worker::single_nonlooparound_use): ...this.
21953 (remove_stmt): Renamed to...
21954 (pcom_worker::remove_stmt): ...this.
21955 (execute_pred_commoning_chain): Renamed to...
21956 (pcom_worker::execute_pred_commoning_chain): ...this.
21957 (execute_pred_commoning): Renamed to...
21958 (pcom_worker::execute_pred_commoning): ...this.
21959 (struct epcc_data): New member worker.
21960 (execute_pred_commoning_cbck): Call execute_pred_commoning
21961 with pcom_worker pointer.
21962 (find_use_stmt): Renamed to...
21963 (pcom_worker::find_use_stmt): ...this.
21964 (find_associative_operation_root): Renamed to...
21965 (pcom_worker::find_associative_operation_root): ...this.
21966 (find_common_use_stmt): Renamed to...
21967 (pcom_worker::find_common_use_stmt): ...this.
21968 (combinable_refs_p): Renamed to...
21969 (pcom_worker::combinable_refs_p): ...this.
21970 (reassociate_to_the_same_stmt): Renamed to...
21971 (pcom_worker::reassociate_to_the_same_stmt): ...this.
21972 (stmt_combining_refs): Renamed to...
21973 (pcom_worker::stmt_combining_refs): ...this.
21974 (combine_chains): Renamed to...
21975 (pcom_worker::combine_chains): ...this.
21976 (try_combine_chains): Renamed to...
21977 (pcom_worker::try_combine_chains): ...this.
21978 (prepare_initializers_chain): Renamed to...
21979 (pcom_worker::prepare_initializers_chain): ...this.
21980 (prepare_initializers): Renamed to...
21981 (pcom_worker::prepare_initializers): ...this.
21982 (prepare_finalizers_chain): Renamed to...
21983 (pcom_worker::prepare_finalizers_chain): ...this.
21984 (prepare_finalizers): Renamed to...
21985 (pcom_worker::prepare_finalizers): ...this.
21986 (tree_predictive_commoning_loop): Renamed to...
21987 (pcom_worker::tree_predictive_commoning_loop): ...this, adjust
21988 some calls and remove some cleanup code.
21989 (tree_predictive_commoning): Adjusted to use pcom_worker instance.
21990 (static variable looparound_phis): Remove.
21991 (static variable name_expansions): Remove.
21993 2021-06-24 Richard Biener <rguenther@suse.de>
21995 * tree-vect-slp.c (slpg_vertex): New struct.
21996 (vect_slp_build_vertices): Adjust.
21997 (vect_optimize_slp): Likewise. Maintain an outgoing permute
21998 and a materialized one.
22000 2021-06-24 Richard Biener <rguenther@suse.de>
22002 PR tree-optimization/101105
22003 * tree-vect-data-refs.c (vect_prune_runtime_alias_test_list):
22004 Only ignore steps when they are equal or scalar order is preserved.
22006 2021-06-24 liuhongt <hongtao.liu@intel.com>
22009 * config/i386/i386-expand.c (ix86_expand_vec_interleave):
22010 Adjust comments for ix86_expand_vecop_qihi2.
22011 (ix86_expand_vecmul_qihi): Renamed to ..
22012 (ix86_expand_vecop_qihi2): Adjust function prototype to
22013 support shift operation, add static to definition.
22014 (ix86_expand_vec_shift_qihi_constant): Add static to definition.
22015 (ix86_expand_vecop_qihi): Call ix86_expand_vecop_qihi2 and
22016 ix86_expand_vec_shift_qihi_constant.
22017 * config/i386/i386-protos.h (ix86_expand_vecmul_qihi): Deleted.
22018 (ix86_expand_vec_shift_qihi_constant): Deleted.
22019 * config/i386/sse.md (VI12_256_512_AVX512VL): New mode
22021 (mulv8qi3): Call ix86_expand_vecop_qihi directly, add
22022 condition TARGET_64BIT.
22023 (mul<mode>3): Ditto.
22024 (<insn><mode>3): Ditto.
22025 (vlshr<mode>3): Extend to support avx512 vlshr.
22026 (v<insn><mode>3): New expander for
22028 (v<insn>v8qi3): Ditto.
22029 (vashrv8hi3<mask_name>): Renamed to ..
22030 (vashr<mode>3): And extend to support V16QImode for avx512.
22031 (vashrv16qi3): Deleted.
22032 (vashrv2di3<mask_name>): Extend expander to support avx512
22035 2021-06-23 Dimitar Dimitrov <dimitar@dinux.eu>
22037 * doc/lto.texi (Design Overview): Update that slim objects are
22040 2021-06-23 Aaron Sawdey <acsawdey@linux.ibm.com>
22042 * config/rs6000/rs6000-cpus.def: Take OPTION_MASK_PCREL_OPT out
22043 of OTHER_POWER10_MASKS so it will not be enabled by default.
22045 2021-06-23 Richard Biener <rguenther@suse.de>
22046 Martin Jambor <mjambor@suse.cz>
22048 * tree-inline.c (setup_one_parameter): Set TREE_READONLY of the
22049 param replacement unconditionally. Adjust comment.
22051 2021-06-23 Andrew MacLeod <amacleod@redhat.com>
22053 * Makefile.in (OBJS): Add gimple-range-fold.o
22054 * gimple-range-fold.cc: New.
22055 * gimple-range-fold.h: New.
22056 * gimple-range-gori.cc (gimple_range_calc_op1): Move to here.
22057 (gimple_range_calc_op2): Ditto.
22058 * gimple-range-gori.h: Move prototypes to here.
22059 * gimple-range.cc: Adjust include files.
22060 (fur_source:fur_source): Relocate to gimple-range-fold.cc.
22061 (fur_source::get_operand): Ditto.
22062 (fur_source::get_phi_operand): Ditto.
22063 (fur_source::query_relation): Ditto.
22064 (fur_source::register_relation): Ditto.
22065 (class fur_edge): Ditto.
22066 (fur_edge::fur_edge): Ditto.
22067 (fur_edge::get_operand): Ditto.
22068 (fur_edge::get_phi_operand): Ditto.
22069 (fur_stmt::fur_stmt): Ditto.
22070 (fur_stmt::get_operand): Ditto.
22071 (fur_stmt::get_phi_operand): Ditto.
22072 (fur_stmt::query_relation): Ditto.
22073 (class fur_depend): Relocate to gimple-range-fold.h.
22074 (fur_depend::fur_depend): Relocate to gimple-range-fold.cc.
22075 (fur_depend::register_relation): Ditto.
22076 (fur_depend::register_relation): Ditto.
22077 (class fur_list): Ditto.
22078 (fur_list::fur_list): Ditto.
22079 (fur_list::get_operand): Ditto.
22080 (fur_list::get_phi_operand): Ditto.
22081 (fold_range): Ditto.
22082 (adjust_pointer_diff_expr): Ditto.
22083 (gimple_range_adjustment): Ditto.
22084 (gimple_range_base_of_assignment): Ditto.
22085 (gimple_range_operand1): Ditto.
22086 (gimple_range_operand2): Ditto.
22087 (gimple_range_calc_op1): Relocate to gimple-range-gori.cc.
22088 (gimple_range_calc_op2): Ditto.
22089 (fold_using_range::fold_stmt): Relocate to gimple-range-fold.cc.
22090 (fold_using_range::range_of_range_op): Ditto.
22091 (fold_using_range::range_of_address): Ditto.
22092 (fold_using_range::range_of_phi): Ditto.
22093 (fold_using_range::range_of_call): Ditto.
22094 (fold_using_range::range_of_builtin_ubsan_call): Ditto.
22095 (fold_using_range::range_of_builtin_call): Ditto.
22096 (fold_using_range::range_of_cond_expr): Ditto.
22097 (fold_using_range::range_of_ssa_name_with_loop_info): Ditto.
22098 (fold_using_range::relation_fold_and_or): Ditto.
22099 (fold_using_range::postfold_gcond_edges): Ditto.
22100 * gimple-range.h: Add gimple-range-fold.h to include files. Change
22101 GIMPLE_RANGE_STMT_H to GIMPLE_RANGE_H.
22102 (gimple_range_handler): Relocate to gimple-range-fold.h.
22103 (gimple_range_ssa_p): Ditto.
22104 (range_compatible_p): Ditto.
22105 (class fur_source): Ditto.
22106 (class fur_stmt): Ditto.
22107 (class fold_using_range): Ditto.
22108 (gimple_range_calc_op1): Relocate to gimple-range-gori.h
22109 (gimple_range_calc_op2): Ditto.
22111 2021-06-23 Andrew MacLeod <amacleod@redhat.com>
22113 PR tree-optimization/101148
22114 PR tree-optimization/101014
22115 * gimple-range-cache.cc (ranger_cache::ranger_cache): Adjust.
22116 (ranger_cache::~ranger_cache): Adjust.
22117 (ranger_cache::block_range): Check if propagation disallowed.
22118 (ranger_cache::propagate_cache): Disallow propagation if new value
22119 can't be stored properly.
22120 * gimple-range-cache.h (ranger_cache::m_propfail): New member.
22122 2021-06-23 Andrew MacLeod <amacleod@redhat.com>
22124 * gimple-range-cache.cc (class ssa_block_ranges): Adjust prototype.
22125 (sbr_vector::set_bb_range): Return true.
22126 (class sbr_sparse_bitmap): Adjust.
22127 (sbr_sparse_bitmap::set_bb_range): Return value.
22128 (block_range_cache::set_bb_range): Return value.
22129 (ranger_cache::propagate_cache): Use return value to print msg.
22130 * gimple-range-cache.h (class block_range_cache): Adjust.
22132 2021-06-23 Andrew MacLeod <amacleod@redhat.com>
22134 * gimple-range.cc (dump_bb): Use range_on_edge from the cache.
22136 2021-06-23 Jeff Law <jeffreyalaw@gmail.com>
22138 * config/h8300/logical.md (<code><mode>3<ccnz>): Use <cczn>
22139 so this pattern can be used for test/compare removal. Pass
22140 current insn to compute_logical_op_length and output_logical_op.
22141 * config/h8300/h8300.c (compute_logical_op_cc): Remove.
22142 (h8300_and_costs): Add argument to compute_logical_op_length.
22143 (output_logical_op): Add new argument. Use it to determine if the
22144 condition codes are used and adjust the output accordingly.
22145 (compute_logical_op_length): Add new argument and update length
22146 computations when condition codes are used.
22147 * config/h8300/h8300-protos.h (compute_logical_op_length): Update
22149 (output_logical_op): Likewise.
22151 2021-06-23 Uroš Bizjak <ubizjak@gmail.com>
22154 * config/i386/i386-expand.c (expand_vec_perm_pshufb):
22155 Handle 64bit modes for TARGET_XOP. Use indirect gen_* functions.
22156 * config/i386/mmx.md (mmx_ppermv64): New insn pattern.
22157 * config/i386/i386.md (unspec): Move UNSPEC_XOP_PERMUTE from ...
22158 * config/i386/sse.md (unspec): ... here.
22160 2021-06-23 Martin Liska <mliska@suse.cz>
22163 * optc-save-gen.awk: Put back arm_fp16_format to
22166 2021-06-23 Uroš Bizjak <ubizjak@gmail.com>
22169 * config/i386/i386.md (bsr_rex64): Add zero-flag setting RTX.
22172 (clz<mode>2): Update RTX pattern for additions.
22174 2021-06-23 Jakub Jelinek <jakub@redhat.com>
22176 PR middle-end/101167
22177 * omp-low.c (lower_omp_regimplify_p): Regimplify also PARM_DECLs
22178 and RESULT_DECLs that have DECL_HAS_VALUE_EXPR_P set.
22180 2021-06-22 Sergei Trofimovich <siarheit@google.com>
22182 * doc/rtl.texi: drop unbalanced parenthesis.
22184 2021-06-22 Richard Biener <rguenther@suse.de>
22186 PR middle-end/101156
22187 * gimplify.c (gimplify_expr): Remove premature incorrect
22190 2021-06-22 Jakub Jelinek <jakub@redhat.com>
22192 PR tree-optimization/101159
22193 * tree-vect-patterns.c (vect_recog_popcount_pattern): Fix some
22196 2021-06-22 Jakub Jelinek <jakub@redhat.com>
22198 PR middle-end/101160
22199 * function.c (assign_parms): For decl_result with TYPE_EMPTY_P type
22200 clear crtl->return_rtx instead of keeping it referencing a pseudo.
22202 2021-06-22 Jakub Jelinek <jakub@redhat.com>
22203 Andrew Pinski <apinski@marvell.com>
22205 PR tree-optimization/101162
22206 * fold-const.c (range_check_type): Handle OFFSET_TYPE like pointer
22209 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
22211 * range-op.cc (range_relational_tests): New.
22212 (range_op_tests): Call range_relational_tests.
22214 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
22216 * range-op.cc (operator_cast::lhs_op1_relation): New.
22217 (operator_identity::lhs_op1_relation): Mew.
22219 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
22221 * range-op.cc (operator_minus::op1_op2_relation_effect): New.
22223 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
22225 * range-op.cc (operator_plus::lhs_op1_relation): New.
22226 (operator_plus::lhs_op2_relation): New.
22228 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
22230 * gimple-range-cache.cc (ranger_cache::ranger_cache): Create a
22231 relation_oracle if dominators exist.
22232 (ranger_cache::~ranger_cache): Dispose of oracle.
22233 (ranger_cache::dump_bb): Dump oracle.
22234 * gimple-range.cc (fur_source::fur_source): New.
22235 (fur_source::get_operand): Use mmeber query.
22236 (fur_source::get_phi_operand): Use member_query.
22237 (fur_source::query_relation): New.
22238 (fur_source::register_dependency): Delete.
22239 (fur_source::register_relation): New.
22240 (fur_edge::fur_edge): Adjust.
22241 (fur_edge::get_phi_operand): Fix comment.
22242 (fur_edge::query): Delete.
22243 (fur_stmt::fur_stmt): Adjust.
22244 (fur_stmt::query): Delete.
22245 (fur_depend::fur_depend): Adjust.
22246 (fur_depend::register_relation): New.
22247 (fur_depend::register_relation): New.
22248 (fur_list::fur_list): Adjust.
22249 (fur_list::get_operand): Use member query.
22250 (fold_using_range::range_of_range_op): Process and query relations.
22251 (fold_using_range::range_of_address): Adjust dependency call.
22252 (fold_using_range::range_of_phi): Ditto.
22253 (gimple_ranger::gimple_ranger): New. Use ranger_ache oracle.
22254 (fold_using_range::relation_fold_and_or): New.
22255 (fold_using_range::postfold_gcond_edges): New.
22256 * gimple-range.h (class gimple_ranger): Adjust.
22257 (class fur_source): Adjust members.
22258 (class fur_stmt): Ditto.
22259 (class fold_using_range): Ditto.
22261 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
22263 * range-op.cc (range_operator::wi_fold): Apply relation effect.
22264 (range_operator::fold_range): Adjust and apply relation effect.
22265 (*::fold_range): Add relation parameters.
22266 (*::op1_range): Ditto.
22267 (*::op2_range): Ditto.
22268 (range_operator::lhs_op1_relation): New.
22269 (range_operator::lhs_op2_relation): New.
22270 (range_operator::op1_op2_relation): New.
22271 (range_operator::op1_op2_relation_effect): New.
22272 (relop_early_resolve): New.
22273 (operator_equal::op1_op2_relation): New.
22274 (operator_equal::fold_range): Call relop_early_resolve.
22275 (operator_not_equal::op1_op2_relation): New.
22276 (operator_not_equal::fold_range): Call relop_early_resolve.
22277 (operator_lt::op1_op2_relation): New.
22278 (operator_lt::fold_range): Call relop_early_resolve.
22279 (operator_le::op1_op2_relation): New.
22280 (operator_le::fold_range): Call relop_early_resolve.
22281 (operator_gt::op1_op2_relation): New.
22282 (operator_gt::fold_range): Call relop_early_resolve.
22283 (operator_ge::op1_op2_relation): New.
22284 (operator_ge::fold_range): Call relop_early_resolve.
22285 * range-op.h (class range_operator): Adjust parameters and methods.
22287 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
22289 * Makefile.in (OBJS): Add value-relation.o.
22290 * gimple-range.h: Adjust include files.
22291 * tree-data-ref.c: Adjust include file order.
22292 * value-query.cc (range_query::get_value_range): Default to no oracle.
22293 (range_query::query_relation): New.
22294 (range_query::query_relation): New.
22295 * value-query.h (class range_query): Adjust.
22296 * value-relation.cc: New.
22297 * value-relation.h: New.
22299 2021-06-22 Richard Biener <rguenther@suse.de>
22301 PR tree-optimization/101151
22302 * tree-ssa-sink.c (statement_sink_location): Expand irreducible
22305 2021-06-22 Jojo R <rjiejie@linux.alibaba.com>
22307 * config/riscv/riscv.c (thead_c906_tune_info): New.
22308 (riscv_tune_info_table): Use new tune.
22310 2021-06-22 Richard Biener <rguenther@suse.de>
22312 PR tree-optimization/101158
22313 * tree-vect-slp.c (vect_build_slp_tree_1): Move same operand
22314 checking after checking for matching operation.
22316 2021-06-22 Richard Biener <rguenther@suse.de>
22318 PR tree-optimization/101159
22319 * tree-vect-patterns.c (vect_recog_popcount_pattern): Add
22320 missing NULL vectype check.
22322 2021-06-22 Richard Biener <rguenther@suse.de>
22324 PR tree-optimization/101154
22325 * tree-vect-slp.c (vect_build_slp_tree_2): Fix out-of-bound access.
22327 2021-06-22 Jakub Jelinek <jakub@redhat.com>
22330 * config/i386/i386-protos.h (ix86_last_zero_store_uid): Declare.
22331 * config/i386/i386-expand.c (ix86_last_zero_store_uid): New variable.
22332 * config/i386/i386.c (ix86_expand_prologue): Clear it.
22333 * config/i386/i386.md (peephole2s for 1/2/4 stores of const0_rtx):
22334 Remove "" from match_operand. Emit new insns using emit_move_insn and
22335 set ix86_last_zero_store_uid to INSN_UID of the last store.
22336 Add peephole2s for 1/2/4 stores of const0_rtx following previous
22339 2021-06-22 Martin Liska <mliska@suse.cz>
22341 * auto-profile.c (AUTO_PROFILE_VERSION): Bump as string format
22344 2021-06-22 Martin Liska <mliska@suse.cz>
22346 * gcov-io.h: Remove padding entries.
22348 2021-06-22 liuhongt <hongtao.liu@intel.com>
22350 PR tree-optimization/97770
22351 * tree-vect-patterns.c (vect_recog_popcount_pattern):
22353 (vect_recog_func vect_vect_recog_func_ptrs): Add new pattern.
22355 2021-06-22 liuhongt <hongtao.liu@intel.com>
22358 * config/i386/i386-builtin.def (BDESC): Adjust builtin name.
22359 * config/i386/sse.md (<avx512>_expand<mode>_mask): Rename to ..
22360 (expand<mode>_mask): this ..
22361 (*expand<mode>_mask): New pre_reload splitter to transform
22362 v{,p}expand* to vmov* when mask is zero, all ones, or has all
22363 ones in it's lower part, otherwise still generate
22366 2021-06-22 liuhongt <hongtao.liu@intel.com>
22369 * config/i386/i386-expand.c
22370 (ix86_expand_special_args_builtin): Keep constm1_operand only
22371 if it satisfies insn's operand predicate.
22373 2021-06-21 Jason Merrill <jason@redhat.com>
22376 * df-scan.c (df_ref_record): Check that regno < endregno.
22377 * function.c (assign_parms, expand_function_end): Do nothing with a
22378 TYPE_EMPTY_P result.
22380 2021-06-21 Richard Biener <rguenther@suse.de>
22382 PR tree-optimization/101120
22383 * tree-vect-data-refs.c (bump_vector_ptr): Fold the
22385 * tree-vect-slp.c (vect_transform_slp_perm_load): Add
22386 DR chain DCE capability.
22387 * tree-vectorizer.h (vect_transform_slp_perm_load): Adjust.
22388 * tree-vect-stmts.c (vectorizable_load): Remove unused
22389 loads in the DR chain for SLP.
22391 2021-06-21 Jakub Jelinek <jakub@redhat.com>
22393 PR inline-asm/100785
22394 * gimplify.c (gimplify_asm_expr): Don't diagnose errors if
22395 output or input operands were already error_mark_node.
22396 * cfgexpand.c (expand_asm_stmt): If errors are emitted,
22397 remove all inputs, outputs and clobbers from the asm and
22398 set template to "".
22400 2021-06-21 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
22402 * config/arm/arm_neon.h (vceq_s8): Replace builtin with __a == __b.
22403 (vceq_s16): Likewise.
22404 (vceq_s32): Likewise.
22405 (vceq_u8): Likewise.
22406 (vceq_u16): Likewise.
22407 (vceq_u32): Likewise.
22408 (vceq_p8): Likewise.
22409 (vceqq_s8): Likewise.
22410 (vceqq_s16): Likewise.
22411 (vceqq_s32): Likewise.
22412 (vceqq_u8): Likewise.
22413 (vceqq_u16): Likewise.
22414 (vceqq_u32): Likewise.
22415 (vceqq_p8): Likewise.
22416 (vceq_f32): Gate __a == __b on __FAST_MATH__.
22417 (vceqq_f32): Likewise.
22418 (vceq_f16): Likewise.
22419 (vceqq_f16): Likewise.
22421 2021-06-21 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
22424 * config/arm/iterators.md (NEON_VACMP): Remove.
22425 * config/arm/neon.md (neon_vca<cmp_op><mode>): Use GLTE instead of GTGE
22427 (neon_vca<cmp_op><mode>_insn): Likewise.
22428 (neon_vca<cmp_op_unsp><mode>_insn_unspec): Use NEON_VAGLTE instead of
22431 2021-06-21 Richard Biener <rguenther@suse.de>
22433 PR tree-optimization/101121
22434 * tree-vect-slp.c (vect_build_slp_tree_2): To not fail fatally
22435 when we just lack a stmt with the desired op when doing permutation.
22436 (vect_build_slp_tree): When caching a failed SLP build attempt
22437 assert that at least one lane is marked as not matching.
22439 2021-06-21 liuhongt <hongtao.liu@intel.com>
22442 * config/i386/i386.md: (*anddi_1): Disparage slightly the mask
22443 register alternative.
22444 (*and<mode>_1): Ditto.
22446 (*andn<mode>_1): Ditto.
22447 (*<code><mode>_1): Ditto.
22448 (*<code>qi_1): Ditto.
22449 (*one_cmpl<mode>2_1): Ditto.
22450 (*one_cmplsi2_1_zext): Ditto.
22451 (*one_cmplqi2_1): Ditto.
22452 * config/i386/i386.c (x86_order_regs_for_local_alloc): Change
22453 the order of mask registers to be before general registers.
22455 2021-06-21 Roger Sayle <roger@nextmovesoftware.com>
22458 * config/i386/i386.md: New define_peephole2s to shrink writing
22459 1, 2 or 4 consecutive zeros to memory when optimizing for size.
22461 2021-06-18 Jeff Law <jeffreyalaw@gmail.com>
22463 * config/h8300/h8300.c (h8300_select_cc_mode): Handle SYMBOL_REF.
22464 * config/h8300/logical.md (<code><mode>3 logcial expander): Generate
22465 more efficient code when the source can be trivially simplified.
22467 2021-06-18 Andrew MacLeod <amacleod@redhat.com>
22469 * gimple-range-cache.cc (ranger_cache::range_of_def): Calculate
22470 a range if global is not available.
22471 (ranger_cache::entry_range): Fallback to range_of_def.
22472 * gimple-range-cache.h (range_of_def): Adjust prototype.
22474 2021-06-18 Andrew MacLeod <amacleod@redhat.com>
22476 PR tree-optimization/101014
22477 * gimple-range-cache.cc (ranger_cache::ranger_cache): Remove poor
22479 (ranger_cache::~ranger_cache): Ditto.
22480 (ranger_cache::enable_new_values): Delete.
22481 (ranger_cache::push_poor_value): Delete.
22482 (ranger_cache::range_of_def): Remove poor value processing.
22483 (ranger_cache::entry_range): Ditto.
22484 (ranger_cache::fill_block_cache): Ditto.
22485 * gimple-range-cache.h (class ranger_cache): Remove poor value members.
22486 * gimple-range.cc (gimple_ranger::range_of_expr): Remove call.
22487 * gimple-range.h (class gimple_ranger): Adjust.
22489 2021-06-18 Srinath Parvathaneni <srinath.parvathaneni@arm.com>
22492 * common/config/arm/arm-common.c (arm_canon_arch_option_1): New function
22493 derived from arm_canon_arch.
22494 (arm_canon_arch_option): Call it.
22495 (arm_canon_arch_multilib_option): New function.
22496 * config/arm/arm-cpus.in (IGNORE_FOR_MULTILIB): New fgroup.
22497 * config/arm/arm.h (arm_canon_arch_multilib_option): New prototype.
22498 (CANON_ARCH_MULTILIB_SPEC_FUNCTION): New macro.
22499 (MULTILIB_ARCH_CANONICAL_SPECS): New macro.
22500 (DRIVER_SELF_SPECS): Add MULTILIB_ARCH_CANONICAL_SPECS.
22501 * config/arm/arm.opt (mlibarch): New option.
22502 * config/arm/t-rmprofile (MULTILIB_MATCHES): For armv8*-m, replace use
22503 of march on RHS with mlibarch.
22505 2021-06-18 Marcel Vollweiler <marcel@codesourcery.com>
22507 * config.in: Regenerate.
22508 * config/gcn/gcn.c (print_operand_address): Fix for global_load assembler
22510 * configure: Regenerate.
22511 * configure.ac: Fix for global_load assembler functions.
22513 2021-06-18 Richard Biener <rguenther@suse.de>
22515 PR tree-optimization/101112
22516 * tree-vect-slp.c (vect_slp_linearize_chain): Fix condition
22517 to lookup a pattern stmt def.
22519 2021-06-18 Jakub Jelinek <jakub@redhat.com>
22521 PR middle-end/101062
22522 * stor-layout.c (finish_bitfield_layout): Don't add bitfield
22523 representatives in QUAL_UNION_TYPE.
22525 2021-06-18 Andrew Pinski <apinski@marvell.com>
22527 * tree-ssa-phiopt.c (replace_phi_edge_with_variable):
22528 Add counting of how many times it is done.
22529 (factor_out_conditional_conversion): Likewise.
22530 (match_simplify_replacement): Likewise.
22531 (value_replacement): Likewise.
22532 (spaceship_replacement): Likewise.
22533 (cond_store_replacement): Likewise.
22534 (cond_if_else_store_replacement_1): Likewise.
22535 (hoist_adjacent_loads): Likewise.
22537 2021-06-18 Andrew Pinski <apinski@marvell.com>
22539 * tree-cfg.c (verify_gimple_assign_unary): Reject point and offset
22540 types on NEGATE_EXPR, ABS_EXPR, BIT_NOT_EXPR, PAREN_EXPR and CNONJ_EXPR.
22541 (verify_gimple_assign_binary): Reject point and offset types on
22542 MULT_EXPR, MULT_HIGHPART_EXPR, TRUNC_DIV_EXPR, CEIL_DIV_EXPR,
22543 FLOOR_DIV_EXPR, ROUND_DIV_EXPR, TRUNC_MOD_EXPR, CEIL_MOD_EXPR,
22544 FLOOR_MOD_EXPR, ROUND_MOD_EXPR, RDIV_EXPR, and EXACT_DIV_EXPR.
22546 2021-06-18 Michael Meissner <meissner@linux.ibm.com>
22548 * config/rs6000/rs6000.c (rs6000_emit_minmax): Add support for ISA
22549 3.1 IEEE 128-bit floating point xsmaxcqp/xsmincqp instructions.
22550 * config/rs6000/rs6000.md (s<minmax><mode>3, IEEE128 iterator):
22553 2021-06-17 Aaron Sawdey <acsawdey@linux.ibm.com>
22555 * config/rs6000/genfusion.pl (gen_logical_addsubf): Add
22556 earlyclobber to alts 0/1.
22557 (gen_addadd): Add earlyclobber to alts 0/1.
22558 * config/rs6000/fusion.md: Regenerate file.
22560 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
22562 * cfgloopanal.c (get_loop_hot_path): Make path an auto_vec.
22564 2021-06-17 Andrew MacLeod <amacleod@redhat.com>
22566 * gimple-range-cache.cc: Comment cleanups.
22567 * gimple-range-gori.cc: Comment cleanups.
22568 * gimple-range.cc: Comment/spacing cleanups
22569 * value-range.h: Comment cleanups.
22571 2021-06-17 H.J. Lu <hjl.tools@gmail.com>
22574 * calls.c (expand_call): Replace PUSH_ARGS with
22575 targetm.calls.push_argument (0).
22576 (emit_library_call_value_1): Likewise.
22577 * defaults.h (PUSH_ARGS): Removed.
22578 (PUSH_ARGS_REVERSED): Replace PUSH_ARGS with
22579 targetm.calls.push_argument (0).
22580 * expr.c (block_move_libcall_safe_for_call_parm): Likewise.
22581 (emit_push_insn): Pass the number bytes to push to
22582 targetm.calls.push_argument and pass 0 if ARGS_ADDR is 0.
22583 * hooks.c (hook_bool_uint_true): New.
22584 * hooks.h (hook_bool_uint_true): Likewise.
22585 * rtlanal.c (nonzero_bits1): Replace PUSH_ARGS with
22586 targetm.calls.push_argument (0).
22587 * target.def (push_argument): Add a targetm.calls hook.
22588 * targhooks.c (default_push_argument): New.
22589 * targhooks.h (default_push_argument): Likewise.
22590 * config/bpf/bpf.h (PUSH_ARGS): Removed.
22591 * config/cr16/cr16.c (TARGET_PUSH_ARGUMENT): New.
22592 * config/cr16/cr16.h (PUSH_ARGS): Removed.
22593 * config/i386/i386.c (ix86_push_argument): New.
22594 (TARGET_PUSH_ARGUMENT): Likewise.
22595 * config/i386/i386.h (PUSH_ARGS): Removed.
22596 * config/m32c/m32c.c (TARGET_PUSH_ARGUMENT): New.
22597 * config/m32c/m32c.h (PUSH_ARGS): Removed.
22598 * config/nios2/nios2.h (PUSH_ARGS): Likewise.
22599 * config/pru/pru.h (PUSH_ARGS): Likewise.
22600 * doc/tm.texi.in: Remove PUSH_ARGS documentation. Add
22601 TARGET_PUSH_ARGUMENT hook.
22602 * doc/tm.texi: Regenerated.
22604 2021-06-17 Uroš Bizjak <ubizjak@gmail.com>
22607 * config/i386/i386-expand.c (expand_vector_set_var):
22608 Handle V2FS mode remapping. Pass TARGET_MMX_WITH_SSE to
22609 ix86_expand_vector_init_duplicate.
22610 (ix86_expand_vector_init_duplicate): Emit insv_1 for
22611 QImode for !TARGET_PARTIAL_REG_STALL.
22612 * config/i386/predicates.md (vec_setm_mmx_operand): New predicate.
22613 * config/i386/mmx.md (vec_setv2sf): Use vec_setm_mmx_operand
22614 as operand 2 predicate. Call ix86_expand_vector_set_var
22615 for non-constant index operand.
22616 (vec_setv2si): Ditto.
22617 (vec_setv4hi): Ditto.
22618 (vec_setv8qi): ditto.
22620 2021-06-17 Aldy Hernandez <aldyh@redhat.com>
22622 PR tree-optimization/100790
22623 * gimple-range.cc (range_of_builtin_call): Cleanup clz and ctz
22626 2021-06-17 Martin Liska <mliska@suse.cz>
22628 * doc/invoke.texi: Use consistently -O1 instead of -O.
22630 2021-06-17 Martin Liska <mliska@suse.cz>
22632 * gcov-io.h: Update documentation entry about string format.
22634 2021-06-17 Marius Hillenbrand <mhillen@linux.ibm.com>
22637 * config/s390/vecintrin.h (vec_doublee): Fix to use
22638 __builtin_s390_vflls.
22639 (vec_floate): Fix to use __builtin_s390_vflrd.
22641 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
22643 * dominance.c (get_dominated_to_depth): Return auto_vec<basic_block>.
22644 * dominance.h (get_dominated_to_depth): Likewise.
22645 (get_all_dominated_blocks): Likewise.
22646 * cfgcleanup.c (delete_unreachable_blocks): Adjust.
22647 * gcse.c (hoist_code): Likewise.
22648 * tree-cfg.c (remove_edge_and_dominated_blocks): Likewise.
22649 * tree-parloops.c (oacc_entry_exit_ok): Likewise.
22650 * tree-ssa-dce.c (eliminate_unnecessary_stmts): Likewise.
22651 * tree-ssa-phiprop.c (pass_phiprop::execute): Likewise.
22653 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
22655 * dominance.c (get_dominated_by_region): Return auto_vec<basic_block>.
22656 * dominance.h (get_dominated_by_region): Likewise.
22657 * tree-cfg.c (gimple_duplicate_sese_region): Adjust.
22658 (gimple_duplicate_sese_tail): Likewise.
22659 (move_sese_region_to_fn): Likewise.
22661 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
22663 * dominance.c (get_dominated_by): Return auto_vec<basic_block>.
22664 * dominance.h (get_dominated_by): Likewise.
22665 * auto-profile.c (afdo_find_equiv_class): Adjust.
22666 * cfgloopmanip.c (duplicate_loop_to_header_edge): Likewise.
22667 * loop-unroll.c (unroll_loop_runtime_iterations): Likewise.
22668 * tree-cfg.c (test_linear_chain): Likewise.
22669 (test_diamond): Likewise.
22671 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
22673 * cfgloop.h (get_loop_hot_path): Return auto_vec<basic_block>.
22674 * cfgloopanal.c (get_loop_hot_path): Likewise.
22675 * tree-ssa-loop-ivcanon.c (tree_estimate_loop_size): Likewise.
22677 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
22679 * cgraph.c (cgraph_node::collect_callers): Return
22680 auto_vec<cgraph_edge *>.
22681 * cgraph.h (cgraph_node::collect_callers): Likewise.
22682 * ipa-cp.c (create_specialized_node): Adjust.
22683 (decide_about_value): Likewise.
22684 (decide_whether_version_node): Likewise.
22685 * ipa-sra.c (process_isra_node_results): Likewise.
22687 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
22689 * vec.h (vl_ptr>::using_auto_storage): Handle null m_vec.
22690 (auto_vec<T, 0>::auto_vec): Define move constructor, and delete copy
22692 (auto_vec<T, 0>::operator=): Define move assignment and delete copy
22695 2021-06-17 Aldy Hernandez <aldyh@redhat.com>
22697 * gimple-range.cc (debug_seed_ranger): New.
22698 (dump_ranger): New.
22699 (debug_ranger): New.
22701 2021-06-17 Richard Biener <rguenther@suse.de>
22703 PR tree-optimization/54400
22704 * tree-vectorizer.h (enum slp_instance_kind): Add
22705 slp_inst_kind_bb_reduc.
22706 (reduction_fn_for_scalar_code): Declare.
22707 * tree-vect-data-refs.c (vect_slp_analyze_instance_dependence):
22708 Check SLP_INSTANCE_KIND instead of looking at the
22710 (vect_slp_analyze_instance_alignment): Likewise.
22711 * tree-vect-loop.c (reduction_fn_for_scalar_code): Export.
22712 * tree-vect-slp.c (vect_slp_linearize_chain): Split out
22713 chain linearization from vect_build_slp_tree_2 and generalize
22714 for the use of BB reduction vectorization.
22715 (vect_build_slp_tree_2): Adjust accordingly.
22716 (vect_optimize_slp): Elide permutes at the root of BB reduction
22718 (vectorizable_bb_reduc_epilogue): New function.
22719 (vect_slp_prune_covered_roots): Likewise.
22720 (vect_slp_analyze_operations): Use them.
22721 (vect_slp_check_for_constructors): Recognize associatable
22722 chains for BB reduction vectorization.
22723 (vectorize_slp_instance_root_stmt): Generate code for the
22724 BB reduction epilogue.
22726 2021-06-17 Andrew MacLeod <amacleod@redhat.com>
22728 * gimple-range-gori.cc (gori_compute::has_edge_range_p): Check with
22730 (gori_compute::may_recompute_p): New.
22731 (gori_compute::outgoing_edge_range_p): Perform recomputations.
22732 * gimple-range-gori.h (class gori_compute): Add prototype.
22734 2021-06-17 Andrew MacLeod <amacleod@redhat.com>
22736 * gimple-range-cache.cc (ranger_cache::range_on_edge): Always return
22737 true when a range can be calculated.
22738 * gimple-range.cc (gimple_ranger::dump_bb): Check has_edge_range_p.
22740 2021-06-16 Martin Sebor <msebor@redhat.com>
22742 * doc/invoke.texi (-Wmismatched-dealloc, -Wmismatched-new-delete):
22743 Correct documented defaults.
22745 2021-06-16 Andrew MacLeod <amacleod@redhat.com>
22747 * gimple-range-cache.cc (ranger_cache::ranger_cache): Initialize
22748 m_new_value_p directly.
22750 2021-06-16 Uroš Bizjak <ubizjak@gmail.com>
22753 * config/i386/i386-expand.c (expand_vec_perm_2perm_pblendv):
22754 Handle 64bit modes for TARGET_SSE4_1.
22755 (expand_vec_perm_pshufb2): Handle 64bit modes for TARGET_SSSE3.
22756 (expand_vec_perm_even_odd_pack): Handle V4HI mode.
22757 (expand_vec_perm_even_odd_1) <case E_V4HImode>: Expand via
22758 expand_vec_perm_pshufb2 for TARGET_SSSE3 and via
22759 expand_vec_perm_even_odd_pack for TARGET_SSE4_1.
22760 * config/i386/mmx.md (mmx_packusdw): New insn pattern.
22762 2021-06-16 Jonathan Wright <jonathan.wright@arm.com>
22764 * config/aarch64/aarch64-simd.md (aarch64_<sur><addsub>hn<mode>):
22765 Change to an expander that emits the correct instruction
22766 depending on endianness.
22767 (aarch64_<sur><addsub>hn<mode>_insn_le): Define.
22768 (aarch64_<sur><addsub>hn<mode>_insn_be): Define.
22770 2021-06-16 Jonathan Wright <jonathan.wright@arm.com>
22772 * config/aarch64/aarch64-simd-builtins.def: Split generator
22773 for aarch64_<su>qmovn builtins into scalar and vector
22775 * config/aarch64/aarch64-simd.md (aarch64_<su>qmovn<mode>_insn_le):
22777 (aarch64_<su>qmovn<mode>_insn_be): Define.
22778 (aarch64_<su>qmovn<mode>): Split into scalar and vector
22779 variants. Change vector variant to an expander that emits the
22780 correct instruction depending on endianness.
22782 2021-06-16 Jonathan Wright <jonathan.wright@arm.com>
22784 * config/aarch64/aarch64-simd-builtins.def: Split generator
22785 for aarch64_sqmovun builtins into scalar and vector variants.
22786 * config/aarch64/aarch64-simd.md (aarch64_sqmovun<mode>):
22787 Split into scalar and vector variants. Change vector variant
22788 to an expander that emits the correct instruction depending
22790 (aarch64_sqmovun<mode>_insn_le): Define.
22791 (aarch64_sqmovun<mode>_insn_be): Define.
22793 2021-06-16 Jonathan Wright <jonathan.wright@arm.com>
22795 * config/aarch64/aarch64-simd.md (aarch64_xtn<mode>_insn_le):
22796 Define - modeling zero-high-half semantics.
22797 (aarch64_xtn<mode>): Change to an expander that emits the
22798 appropriate instruction depending on endianness.
22799 (aarch64_xtn<mode>_insn_be): Define - modeling zero-high-half
22801 (aarch64_xtn2<mode>_le): Rename to...
22802 (aarch64_xtn2<mode>_insn_le): This.
22803 (aarch64_xtn2<mode>_be): Rename to...
22804 (aarch64_xtn2<mode>_insn_be): This.
22805 (vec_pack_trunc_<mode>): Emit truncation instruction instead
22807 * config/aarch64/iterators.md (Vnarrowd): Add Vnarrowd mode
22808 attribute iterator.
22810 2021-06-16 Martin Jambor <mjambor@suse.cz>
22812 PR tree-optimization/100453
22813 * tree-sra.c (create_access): Disqualify any const candidates
22814 which are written to.
22815 (sra_modify_expr): Do not store sub-replacements back to a const base.
22816 (handle_unscalarized_data_in_subtree): Likewise.
22817 (sra_modify_assign): Likewise. Earlier, use TREE_READONLy test
22818 instead of constant_decl_p.
22820 2021-06-16 Jakub Jelinek <jakub@redhat.com>
22822 PR middle-end/101062
22823 * stor-layout.c (finish_bitfield_representative): For fields in unions
22824 assume nextf is always NULL.
22825 (finish_bitfield_layout): Compute bit field representatives also in
22826 unions, but handle it as if each bitfield was the only field in the
22829 2021-06-16 Richard Biener <rguenther@suse.de>
22831 PR tree-optimization/101088
22832 * tree-ssa-loop-im.c (sm_seq_valid_bb): Only look for
22833 supported refs on edges. Do not assert same ref but
22834 different kind stores are unsuported but mark them so.
22835 (hoist_memory_references): Only look for supported refs
22838 2021-06-16 Roger Sayle <roger@nextmovesoftware.com>
22840 PR rtl-optimization/46235
22841 * config/i386/i386.md: New define_split for bt followed by cmov.
22842 (*bt<mode>_setcqi): New define_insn_and_split for bt followed by setc.
22843 (*bt<mode>_setncqi): New define_insn_and_split for bt then setnc.
22844 (*bt<mode>_setnc<mode>): New define_insn_and_split for bt followed
22845 by setnc with zero extension.
22847 2021-06-16 Richard Biener <rguenther@suse.de>
22849 PR tree-optimization/101083
22850 * tree-vect-slp.c (vect_slp_build_two_operator_nodes): Get
22851 vectype as argument.
22852 (vect_build_slp_tree_2): Adjust.
22854 2021-06-15 Martin Sebor <msebor@redhat.com>
22856 PR middle-end/100876
22857 * builtins.c: (gimple_call_return_array): Account for size_t
22858 mangling as either unsigned int or unsigned long
22860 2021-06-15 Jeff Law <jeffreyalaw@gmail.com>
22862 * compare-elim.c (try_eliminate_compare): Run DCE to clean things
22863 up before eliminating comparisons.
22865 2021-06-15 Aldy Hernandez <aldyh@redhat.com>
22867 * range-op.cc (operator_bitwise_or::wi_fold): Make sure
22868 nonzero|X is nonzero.
22869 (range_op_bitwise_and_tests): Add tests for above.
22871 2021-06-15 Carl Love <cel@us.ibm.com>
22874 * config/rs6000/rs6000-builtin.def (VCMPEQUT): Fix the ICODE for the
22876 (VRLQ, VSLQ, VSRQ, VSRAQ): Remove unused BU_P10_OVERLOAD_2
22879 2021-06-15 Tobias Burnus <tobias@codesourcery.com>
22882 * gimplify.c (enum gimplify_defaultmap_kind): Add GDMK_SCALAR_TARGET.
22883 (struct gimplify_omp_ctx): Extend defaultmap array by one.
22884 (new_omp_context): Init defaultmap[GDMK_SCALAR_TARGET].
22885 (omp_notice_variable): Update type classification for Fortran.
22886 (gimplify_scan_omp_clauses): Update calls for new argument; handle
22887 GDMK_SCALAR_TARGET; for Fortran, GDMK_POINTER avoid GOVD_MAP_0LEN_ARRAY.
22888 * langhooks-def.h (lhd_omp_scalar_p): Add 'ptr_ok' argument.
22889 * langhooks.c (lhd_omp_scalar_p): Likewise.
22890 (LANG_HOOKS_OMP_ALLOCATABLE_P, LANG_HOOKS_OMP_SCALAR_TARGET_P): New.
22891 (LANG_HOOKS_DECLS): Add them.
22892 * langhooks.h (struct lang_hooks_for_decls): Add new hooks, update
22893 omp_scalar_p pointer type to include the new bool argument.
22895 2021-06-15 David Malcolm <dmalcolm@redhat.com>
22897 * doc/analyzer.texi
22898 (Special Functions for Debugging the Analyzer): Add
22899 __analyzer_dump_capacity.
22901 2021-06-15 Jakub Jelinek <jakub@redhat.com>
22904 * expr.c (expand_expr_real_2) <case VEC_PACK_FIX_TRUNC_EXPR,
22905 case VEC_PACK_TRUNC_EXPR>: Clear subtarget when changing mode.
22907 2021-06-15 Richard Biener <rguenther@suse.de>
22909 * cfgloopanal.c (mark_irreducible_loops): Use a dominance
22910 check to identify loop latches.
22911 * cfgloop.c (verify_loop_structure): Likewise.
22912 * loop-init.c (apply_loop_flags): Allow marked irreducible
22913 regions even with multiple latches.
22914 * predict.c (rebuild_frequencies): Simplify.
22916 2021-06-15 Richard Biener <rguenther@suse.de>
22918 * tree-ssa-threadupdate.c
22919 (jump_thread_path_registry::mark_threaded_blocks): Assert we
22920 have marked irreducible regions.
22922 2021-06-14 Martin Sebor <msebor@redhat.com>
22925 * builtins.c (gimple_call_return_array): Check for attribute fn spec.
22926 Handle calls to placement new.
22927 (ndecl_dealloc_argno): Avoid placement delete.
22929 2021-06-14 Peter Bergner <bergner@linux.ibm.com>
22932 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_mma_builtin): Use
22933 create_tmp_reg_or_ssa_name().
22935 2021-06-14 Andrew MacLeod <amacleod@redhat.com>
22937 * gimple-range-cache.cc (ranger_cache::ranger_cache): Adjust.
22938 (ranger_cache::enable_new_values): Set to specified value and
22939 return the old value.
22940 (ranger_cache::disable_new_values): Delete.
22941 (ranger_cache::fill_block_cache): Disable non 1st order derived
22943 * gimple-range-cache.h (ranger_cache): Adjust prototypes.
22944 * gimple-range.cc (gimple_ranger::range_of_expr): Adjust.
22946 2021-06-14 Uroš Bizjak <ubizjak@gmail.com>
22949 * config/i386/i386-expand.c (ix86_vectorize_vec_perm_const):
22950 Return true early when testing with V2HImode.
22951 * config/i386/mmx.md (*punpckwd): Split to sse2_pshuflw_1.
22953 2021-06-14 Christophe Lyon <christophe.lyon@linaro.org>
22955 * config/arm/mve.md (mve_vec_unpack<US>_lo_<mode>): New pattern.
22956 (mve_vec_unpack<US>_hi_<mode>): New pattern.
22957 (@mve_vec_pack_trunc_lo_<mode>): New pattern.
22958 (mve_vmovntq_<supf><mode>): Prefix with '@'.
22959 * config/arm/neon.md (vec_unpack<US>_hi_<mode>): Move to
22961 (vec_unpack<US>_lo_<mode>): Likewise.
22962 (vec_pack_trunc_<mode>): Rename to
22963 neon_quad_vec_pack_trunc_<mode>.
22964 * config/arm/vec-common.md (vec_unpack<US>_hi_<mode>): New
22966 (vec_unpack<US>_lo_<mode>): New.
22967 (vec_pack_trunc_<mode>): New.
22969 2021-06-14 Richard Biener <rguenther@suse.de>
22971 PR tree-optimization/100934
22972 * tree-ssa-dom.c (pass_dominator::execute): Properly
22973 mark irreducible regions.
22975 2021-06-14 Martin Liska <mliska@suse.cz>
22977 * doc/invoke.texi: Put r{...} on the same line as @item.
22979 2021-06-14 Martin Liska <mliska@suse.cz>
22981 * doc/invoke.texi: Add missing newline.
22983 2021-06-14 Martin Liska <mliska@suse.cz>
22985 * doc/invoke.texi: Remove '+' charasters.
22987 2021-06-14 Claudiu Zissulescu <claziss@synopsys.com>
22989 * config.gcc (arc): Add support for with_cpu option.
22990 * config/arc/arc.h (OPTION_DEFAULT_SPECS): Add fpu.
22992 2021-06-14 Richard Biener <rguenther@suse.de>
22994 PR tree-optimization/101031
22995 * tree-ssa-strlen.c (maybe_invalidate): Increment max_size
22996 instead of size when accounting for a possibly string
22999 2021-06-14 Martin Liska <mliska@suse.cz>
23001 * gimple-ssa-evrp.c (pointer_equiv_analyzer::~pointer_equiv_analyzer): Use delete[].
23003 2021-06-14 Aldy Hernandez <aldyh@redhat.com>
23005 * value-query.cc (gimple_range_global): Call get_range_global
23006 if called after inlining.
23008 2021-06-13 Uroš Bizjak <ubizjak@gmail.com>
23011 * config/i386/i386-expand.c (expand_vec_perm_pshufb):
23012 Emit constant permutation insn directly from here.
23014 2021-06-13 Trevor Saunders <tbsaunde@tbsaunde.org>
23016 * attribs.c (find_attribute_namespace): Iterate over vec<> with
23018 * auto-profile.c (afdo_find_equiv_class): Likewise.
23019 * gcc.c (do_specs_vec): Likewise.
23020 (do_spec_1): Likewise.
23021 (driver::set_up_specs): Likewise.
23022 * gimple-loop-jam.c (any_access_function_variant_p): Likewise.
23023 * gimple-ssa-store-merging.c (compatible_load_p): Likewise.
23024 (imm_store_chain_info::try_coalesce_bswap): Likewise.
23025 (imm_store_chain_info::coalesce_immediate_stores): Likewise.
23026 (get_location_for_stmts): Likewise.
23027 * graphite-poly.c (print_iteration_domains): Likewise.
23028 (free_poly_bb): Likewise.
23029 (remove_gbbs_in_scop): Likewise.
23030 (free_scop): Likewise.
23031 (dump_gbb_cases): Likewise.
23032 (dump_gbb_conditions): Likewise.
23033 (print_pdrs): Likewise.
23034 (print_scop): Likewise.
23035 * ifcvt.c (cond_move_process_if_block): Likewise.
23036 * lower-subreg.c (decompose_multiword_subregs): Likewise.
23037 * regcprop.c (pass_cprop_hardreg::execute): Likewise.
23038 * sanopt.c (sanitize_rewrite_addressable_params): Likewise.
23039 * sel-sched-dump.c (dump_insn_vector): Likewise.
23040 * store-motion.c (store_ops_ok): Likewise.
23041 (store_killed_in_insn): Likewise.
23042 * timevar.c (timer::named_items::print): Likewise.
23043 * tree-cfgcleanup.c (cleanup_control_flow_pre): Likewise.
23044 (cleanup_tree_cfg_noloop): Likewise.
23045 * tree-data-ref.c (dump_data_references): Likewise.
23046 (print_dir_vectors): Likewise.
23047 (print_dist_vectors): Likewise.
23048 (dump_data_dependence_relations): Likewise.
23049 (dump_dist_dir_vectors): Likewise.
23050 (dump_ddrs): Likewise.
23051 (create_runtime_alias_checks): Likewise.
23052 (free_subscripts): Likewise.
23053 (save_dist_v): Likewise.
23054 (save_dir_v): Likewise.
23055 (invariant_access_functions): Likewise.
23056 (same_access_functions): Likewise.
23057 (access_functions_are_affine_or_constant_p): Likewise.
23058 (find_data_references_in_stmt): Likewise.
23059 (graphite_find_data_references_in_stmt): Likewise.
23060 (free_dependence_relations): Likewise.
23061 (free_data_refs): Likewise.
23062 * tree-inline.c (copy_debug_stmts): Likewise.
23063 * tree-into-ssa.c (dump_currdefs): Likewise.
23064 (rewrite_update_phi_arguments): Likewise.
23065 * tree-ssa-propagate.c (clean_up_loop_closed_phi): Likewise.
23066 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
23068 (vect_slp_analyze_node_dependences): Likewise.
23069 (vect_slp_analyze_instance_dependence): Likewise.
23070 (vect_record_base_alignments): Likewise.
23071 (vect_get_peeling_costs_all_drs): Likewise.
23072 (vect_peeling_supportable): Likewise.
23073 * tree-vectorizer.c (vec_info::~vec_info): Likewise.
23074 (vec_info::free_stmt_vec_infos): Likewise.
23076 2021-06-13 Jeff Law <jeffreyalaw@gmail.com>
23078 * config/h8300/logical.md (<code>qi3_1<cczn>): New pattern.
23079 (andqi3_1<cczn>): Removed.
23080 (<ors>qi3_1): Do not split for IOR/XOR a single bit.
23081 (H8/SX bit logicals): Split out from other patterns.
23082 * config/h8300/multiply.md (mulqihi3_const<cczn>): Renamed from
23083 mulqihi3_const_clobber_flags.
23084 (mulqihi3<cczn>, mulhisi3_const<cczn>, mulhisi3<cczn>): Similarly
23086 2021-06-13 H.J. Lu <hjl.tools@gmail.com>
23089 * config/i386/i386.c (ix86_expand_prologue): Set red_zone_used
23090 to true if red zone is used.
23091 (ix86_output_indirect_jmp): Replace ix86_red_zone_size with
23092 ix86_red_zone_used.
23093 * config/i386/i386.h (machine_function): Add red_zone_used.
23094 (ix86_red_zone_size): Removed.
23095 (ix86_red_zone_used): New.
23096 * config/i386/i386.md (peephole2 patterns): Replace
23097 ix86_red_zone_size with ix86_red_zone_used.
23099 2021-06-12 Jason Merrill <jason@redhat.com>
23101 * doc/extend.texi (unused variable attribute): Applies to
23102 structure fields as well.
23104 2021-06-12 Eugene Rozenfeld <erozen@microsoft.com>
23106 * auto-profile.c (read_profile): fix a typo in an error string
23108 2021-06-11 Thomas Schwinge <thomas@codesourcery.com>
23110 * tree-pretty-print.h (dump_omp_clauses): Add 'bool = true'
23112 * tree-pretty-print.c (dump_omp_clauses): Update.
23113 (dump_generic_node) <OMP_CLAUSE>: Use it.
23115 2021-06-11 Srinath Parvathaneni <srinath.parvathaneni@arm.com>
23118 * config/arm/arm_mve.h (__arm_vld1q): Change __ARM_mve_coerce(p0,
23119 int8_t const *) to __ARM_mve_coerce1(p0, int8_t *) in the argument for
23120 the polymorphic variants matching code.
23121 (__arm_vld1q_z): Likewise.
23122 (__arm_vld2q): Likewise.
23123 (__arm_vld4q): Likewise.
23124 (__arm_vldrbq_gather_offset): Likewise.
23125 (__arm_vldrbq_gather_offset_z): Likewise.
23127 2021-06-11 Roger Sayle <roger@nextmovesoftware.com>
23129 PR tree-optimization/96392
23130 * fold-const.h (tree_expr_maybe_real_minus_zero_p): Fix prototype.
23132 2021-06-11 Roger Sayle <roger@nextmovesoftware.com>
23134 PR tree-optimization/96392
23135 * fold-const.c (fold_real_zero_addition_p): Take both arguments
23136 of the addition or subtraction, not just the zero. Use this
23137 other argument in tests for signaling NaNs and signed zeros.
23138 (tree_expr_maybe_real_minus_zero_p): New predicate.
23139 * fold-const.h (fold_real_zero_addition_p): Update prototype.
23140 (tree_expr_maybe_real_minus_zero_p): New function prototype.
23141 * match.pd: Update calls to fold_real_zero_addition_p.
23142 Replace HONOR_NANS with tree_expr_maybe_nan_p.
23143 Replace HONOR_SIGNED_ZEROS with tree_expr_maybe_real_minus_zero_p.
23144 Replace HONOR_SNANS with tree_expr_maybe_signaling_nan_p.
23145 * tree-ssa-reassoc.c (eliminate_using_constants): Update
23146 call to fold_real_zero_addition_p.
23148 2021-06-11 Richard Biener <rguenther@suse.de>
23150 PR tree-optimization/101025
23151 * tree-ssa-loop-im.c (sm_seq_valid_bb): Make sure to process
23152 all refs that require dependence checking.
23154 2021-06-11 Richard Biener <rguenther@suse.de>
23156 PR tree-optimization/101028
23157 * tree-vect-slp.c (vect_build_slp_tree_2): When SLP
23158 reassoc discovery fails fatally, mark appropriate lanes
23161 2021-06-11 Richard Biener <rguenther@suse.de>
23163 PR tree-optimization/101026
23164 * tree-vect-slp.c (vect_build_slp_tree_2): Make sure we
23165 have a representative for the associated chain nodes.
23167 2021-06-11 Jakub Jelinek <jakub@redhat.com>
23169 PR rtl-optimization/101008
23170 * simplify-rtx.c (relational_result): New function.
23171 (simplify_logical_relational_operation,
23172 simplify_relational_operation): Use it.
23174 2021-06-11 Jakub Jelinek <jakub@redhat.com>
23177 * config/i386/sse.md (*vec_concat<mode>_0_1): Require TARGET_SSE2.
23179 2021-06-11 Uroš Bizjak <ubizjak@gmail.com>
23182 * config/i386/i386-expand.c (expand_vec_perm_pshufb): Return
23183 false if the permutation can be implemented with constant
23184 permutation instruction in wider mode.
23185 (canonicalize_vector_int_perm): Move above expand_vec_perm_pshufb.
23186 Handle V8QImode and V4HImode.
23188 2021-06-11 Martin Liska <mliska@suse.cz>
23190 PR gcov-profile/100788
23191 * common.opt: Add new option.
23192 * coverage.c (coverage_begin_function): Emit warning instead on
23193 the internal compiler error.
23194 * doc/invoke.texi: Document the option.
23195 * toplev.c (process_options): Enable it by default.
23197 2021-06-11 Richard Biener <rguenther@suse.de>
23199 PR middle-end/101009
23200 * tree-data-ref.c (build_classic_dist_vector_1): Make sure
23201 to set *init_b to true when we encounter a constant equal
23203 (compute_affine_dependence): Also dump the actual DR_REF.
23205 2021-06-10 Aldy Hernandez <aldyh@redhat.com>
23207 PR tree-optimization/100984
23208 * gimple-ssa-evrp.c (ssa_equiv_stack): Use auto_vec for
23209 replacements table.
23210 (ssa_equiv_stack::~ssa_equiv_stack): Remove.
23212 2021-06-11 Kewen Lin <linkw@linux.ibm.com>
23214 * config/rs6000/rs6000.md
23215 (floatsi<SFDF:mode>2_lfiwax_<QHI:mode>_mem_zext): New
23216 define_insn_and_split.
23218 2021-06-11 Richard Biener <rguenther@suse.de>
23220 * tree-vect-slp.c (vect_build_slp_tree_2): Use stablesort
23221 to sort operands of the associative chain.
23223 2021-06-11 Richard Biener <rguenther@suse.de>
23225 * system.h (gcc_stablesort_r): Declare.
23226 * sort.cc (gcc_sort_r): Support stable sort.
23227 (gcc_stablesort_r): Define.
23228 * vec.h (vec<>::stablesort): Add.
23230 2021-06-10 Uroš Bizjak <ubizjak@gmail.com>
23233 * config/i386/i386-expand.c (ix86_split_mmx_punpck):
23234 Handle V2SF mode. Emit SHUFPS to fixup unpack-high for V2SF mode.
23235 (expand_vec_perm_blend): Handle 64bit modes for TARGET_SSE4_1.
23236 (expand_vec_perm_pshufb): Handle 64bit modes for TARGET_SSSE3.
23237 (expand_vec_perm_pblendv): Handle 64bit modes for TARGET_SSE4_1.
23238 (expand_vec_perm_interleave2): Handle 64bit modes.
23239 (expand_vec_perm_even_odd_pack): Handle V8QI mode.
23240 (expand_vec_perm_even_odd_1): Ditto.
23241 (ix86_vectorize_vec_perm_const): Ditto.
23242 * config/i386/i386.md (UNSPEC_PSHUFB): Move from ...
23243 * config/i386/sse.md: ... here.
23244 * config/i386/mmx.md (*vec_interleave_lowv2sf):
23245 New insn_and_split pattern.
23246 (*vec_interleave_highv2sf): Ditto.
23247 (mmx_pshufbv8qi3): New insn pattern.
23248 (*mmx_pblendw): Ditto.
23250 2021-06-10 Peter Bergner <bergner@linux.ibm.com>
23252 * config/rs6000/rs6000-builtin.def (build_pair): New built-in.
23253 (build_acc): Likewise.
23254 * config/rs6000/rs6000-call.c (mma_expand_builtin): Swap assemble
23255 source operands in little-endian mode.
23256 (rs6000_gimple_fold_mma_builtin): Handle VSX_BUILTIN_BUILD_PAIR.
23257 (mma_init_builtins): Likewise.
23258 * config/rs6000/rs6000.c (rs6000_split_multireg_move): Handle endianness
23259 ordering for the MMA assemble and build source operands.
23260 * doc/extend.texi (__builtin_vsx_build_acc, __builtin_mma_build_pair):
23262 (__builtin_mma_assemble_acc, __builtin_mma_assemble_pair): Remove
23265 2021-06-10 Jeff Law <jeffreyalaw@gmail.com>
23267 * config/h8300/h8300.c (select_cc_mode): Handle MEM. Use
23269 * config/h8300/extensions.md: Replace _clobber_flags patterns
23272 2021-06-10 Robin Dapp <rdapp@linux.ibm.com>
23274 * config/s390/vector.md (vcond_mask_<mode><mode>): Change to
23275 (vcond_mask_<mode><tointvec>): this.
23277 2021-06-10 Andrew Stubbs <ams@codesourcery.com>
23278 Thomas Schwinge <thomas@codesourcery.com>
23280 * omp-builtins.def (BUILT_IN_GOACC_ENTER_EXIT_DATA): Split into...
23281 (BUILT_IN_GOACC_ENTER_DATA, BUILT_IN_GOACC_EXIT_DATA): ... these.
23282 * gimple.h (enum gf_mask): Split
23283 'GF_OMP_TARGET_KIND_OACC_ENTER_EXIT_DATA' into
23284 'GF_OMP_TARGET_KIND_OACC_ENTER_DATA' and
23285 'GF_OMP_TARGET_KIND_OACC_EXIT_DATA'.
23286 (is_gimple_omp_oacc): Update.
23287 * gimple-pretty-print.c (dump_gimple_omp_target): Likewise.
23288 * gimplify.c (gimplify_omp_target_update): Likewise.
23289 * omp-expand.c (expand_omp_target, build_omp_regions_1)
23290 (omp_make_gimple_edges): Likewise.
23291 * omp-low.c (check_omp_nesting_restrictions, lower_omp_target):
23294 2021-06-10 Aldy Hernandez <aldyh@redhat.com>
23296 * value-query.cc (value_query::value_on_edge): Rename name to
23298 (range_query::range_on_edge): Same.
23299 (range_query::value_of_expr): Same.
23300 (range_query::value_on_edge): Same.
23301 * value-query.h (class value_query): Same.
23302 (class range_query): Same.
23304 2021-06-10 Richard Biener <rguenther@suse.de>
23306 PR tree-optimization/101003
23307 * tree-vect-slp.c (vect_build_slp_tree_2): Appropriately
23308 use the pattern stmt defs when linearizing a chain.
23310 2021-06-10 Jakub Jelinek <jakub@redhat.com>
23313 * ifcvt.c (noce_get_alt_condition, noce_try_abs): Use
23314 prev_nonnote_nondebug_insn instead of prev_nonnote_insn.
23316 2021-06-10 Clement Chigot <clement.chigot@atos.net>
23318 * config/rs6000/aix71.h (ASM_CPU_SPEC): Add Power10 directive.
23319 * config/rs6000/aix72.h (ASM_CPU_SPEC): Likewise.
23321 2021-06-09 Andrew Pinski <apinski@marvell.com>
23323 PR tree-optimization/100925
23324 * match.pd (a ? CST1 : CST2): Limit transformations
23325 that would produce a negative to integeral types only.
23326 Change !POINTER_TYPE_P to INTEGRAL_TYPE_P also.
23328 2021-06-09 Jeff Law <jeffreyalaw@gmail.com>
23331 2021-06-09 Jeff Law <jeffreyalaw@gmail.com>
23333 * doc/tm.texi: Correctly update.
23335 2021-06-09 Jeff Law <jeffreyalaw@gmail.com>
23337 * doc/tm.texi: Correctly update.
23339 2021-06-09 H.J. Lu <hjl.tools@gmail.com>
23342 * doc/tm.texi.in (Trampolines): Add a missing blank line.
23344 2021-06-09 Paul Eggert <eggert@cs.ucla.edu>
23347 * doc/invoke.texi (Code Gen Options); Document that -fno-trampolines
23348 and -ftrampolines work only with Ada.
23349 * doc/tm.texi.in (Trampolines): Likewise.
23350 * doc/tm.texi: Regenerated.
23352 2021-06-09 Carl Love <cel@us.ibm.com>
23354 * config/rs6000/altivec.h (vec_signextll, vec_signexti, vec_signextq):
23355 Add define for new builtins.
23356 * config/rs6000/altivec.md(altivec_vreveti2): Add define_expand.
23357 * config/rs6000/rs6000-builtin.def (VSIGNEXTI, VSIGNEXTLL): Add
23358 overloaded builtin definitions.
23359 (VSIGNEXTSB2W, VSIGNEXTSH2W, VSIGNEXTSB2D, VSIGNEXTSH2D,VSIGNEXTSW2D,
23360 VSIGNEXTSD2Q): Add builtin expansions.
23361 (SIGNEXT): Add P10 overload definition.
23362 * config/rs6000/rs6000-call.c (P9V_BUILTIN_VEC_VSIGNEXTI, P9V_BUILTIN_VEC_VSIGNEXTLL,
23363 P10_BUILTIN_VEC_SIGNEXT): Add overloaded argument definitions.
23364 * config/rs6000/vsx.md (vsx_sign_extend_v2di_v1ti): Add define_insn.
23365 (vsignextend_v2di_v1ti, vsignextend_qi_<mode>, vsignextend_hi_<mode>,
23366 vsignextend_si_v2di)[VIlong]: Add define_expand.
23367 Make define_insn vsx_sign_extend_si_v2di visible.
23368 * doc/extend.texi: Add documentation for the vec_signexti,
23369 vec_signextll builtins and vec_signextq.
23371 2021-06-09 Carl Love <cel@us.ibm.com>
23373 * config/rs6000/rs6000.c (__fixkfti, __fixunskfti, __floattikf,
23374 __floatuntikf): Names changed to __fixkfti_sw, __fixunskfti_sw,
23375 __floattikf_sw, __floatuntikf_sw respectively.
23376 * config/rs6000/rs6000.md (floatti<mode>2, floatunsti<mode>2,
23377 fix_trunc<mode>ti2, fixuns_trunc<mode>ti2): Add
23378 define_insn for mode IEEE 128.
23380 2021-06-09 Carl Love <cel@us.ibm.com>
23382 * config/rs6000/altivec.md (altivec_vslq, altivec_vsrq):
23383 Rename to altivec_vslq_<mode>, altivec_vsrq_<mode>, mode VEC_TI.
23384 * config/rs6000/vector.md (VEC_TI): Was named VSX_TI in vsx.md.
23385 (vashlv1ti3): Change to vashl<mode>3, mode VEC_TI.
23386 (vlshrv1ti3): Change to vlshr<mode>3, mode VEC_TI.
23387 * config/rs6000/vsx.md (VSX_TI): Remove define_mode_iterator. Update
23388 uses of VSX_TI to VEC_TI.
23390 2021-06-09 Carl Love <cel@us.ibm.com>
23392 * config/rs6000/dfp.md (floattitd2, fixtdti2): New define_insns.
23394 2021-06-09 Carl Love <cel@us.ibm.com>
23396 * config/rs6000/altivec.h (vec_dive, vec_mod): Add define for new
23398 * config/rs6000/altivec.md (UNSPEC_VMULEUD, UNSPEC_VMULESD,
23399 UNSPEC_VMULOUD, UNSPEC_VMULOSD): New unspecs.
23400 (altivec_eqv1ti, altivec_gtv1ti, altivec_gtuv1ti, altivec_vmuleud,
23401 altivec_vmuloud, altivec_vmulesd, altivec_vmulosd, altivec_vrlq,
23402 altivec_vrlqmi, altivec_vrlqmi_inst, altivec_vrlqnm,
23403 altivec_vrlqnm_inst, altivec_vslq, altivec_vsrq, altivec_vsraq,
23404 altivec_vcmpequt_p, altivec_vcmpgtst_p, altivec_vcmpgtut_p): New
23406 (vec_widen_umult_even_v2di, vec_widen_smult_even_v2di,
23407 vec_widen_umult_odd_v2di, vec_widen_smult_odd_v2di, altivec_vrlqmi,
23408 altivec_vrlqnm): New define_expands.
23409 * config/rs6000/rs6000-builtin.def (VCMPEQUT_P, VCMPGTST_P,
23410 VCMPGTUT_P): Add macro expansions.
23411 (BU_P10V_AV_P): Add builtin predicate definition.
23412 (VCMPGTUT, VCMPGTST, VCMPEQUT, CMPNET, CMPGE_1TI,
23413 CMPGE_U1TI, CMPLE_1TI, CMPLE_U1TI, VNOR_V1TI_UNS, VNOR_V1TI, VCMPNET_P,
23414 VCMPAET_P, VMULEUD, VMULESD, VMULOUD, VMULOSD, VRLQ,
23415 VSLQ, VSRQ, VSRAQ, VRLQNM, DIV_V1TI, UDIV_V1TI, DIVES_V1TI, DIVEU_V1TI,
23416 MODS_V1TI, MODU_V1TI, VRLQMI): New macro expansions.
23417 (VRLQ, VSLQ, VSRQ, VSRAQ, DIVE, MOD): New overload expansions.
23418 * config/rs6000/rs6000-call.c (P10_BUILTIN_VCMPEQUT,
23419 P10V_BUILTIN_CMPGE_1TI, P10V_BUILTIN_CMPGE_U1TI,
23420 P10V_BUILTIN_VCMPGTUT, P10V_BUILTIN_VCMPGTST,
23421 P10V_BUILTIN_CMPLE_1TI, P10V_BUILTIN_VCMPLE_U1TI,
23422 P10V_BUILTIN_DIV_V1TI, P10V_BUILTIN_UDIV_V1TI,
23423 P10V_BUILTIN_VMULESD, P10V_BUILTIN_VMULEUD,
23424 P10V_BUILTIN_VMULOSD, P10V_BUILTIN_VMULOUD,
23425 P10V_BUILTIN_VNOR_V1TI, P10V_BUILTIN_VNOR_V1TI_UNS,
23426 P10V_BUILTIN_VRLQ, P10V_BUILTIN_VRLQMI,
23427 P10V_BUILTIN_VRLQNM, P10V_BUILTIN_VSLQ,
23428 P10V_BUILTIN_VSRQ, P10V_BUILTIN_VSRAQ,
23429 P10V_BUILTIN_VCMPGTUT_P, P10V_BUILTIN_VCMPGTST_P,
23430 P10V_BUILTIN_VCMPEQUT_P, P10V_BUILTIN_VCMPGTUT_P,
23431 P10V_BUILTIN_VCMPGTST_P, P10V_BUILTIN_CMPNET,
23432 P10V_BUILTIN_VCMPNET_P, P10V_BUILTIN_VCMPAET_P,
23433 P10V_BUILTIN_DIVES_V1TI, P10V_BUILTIN_MODS_V1TI,
23434 P10V_BUILTIN_MODU_V1TI):
23435 New overloaded definitions.
23436 (rs6000_gimple_fold_builtin) [P10V_BUILTIN_VCMPEQUT,
23437 P10V_BUILTIN_CMPNET, P10V_BUILTIN_CMPGE_1TI,
23438 P10V_BUILTIN_CMPGE_U1TI, P10V_BUILTIN_VCMPGTUT,
23439 P10V_BUILTIN_VCMPGTST, P10V_BUILTIN_CMPLE_1TI,
23440 P10V_BUILTIN_CMPLE_U1TI]: New case statements.
23441 (rs6000_init_builtins) [bool_V1TI_type_node, int_ftype_int_v1ti_v1ti]:
23443 (altivec_init_builtins): New E_V1TImode case statement.
23444 (builtin_function_type)[P10_BUILTIN_128BIT_VMULEUD,
23445 P10_BUILTIN_128BIT_VMULOUD, P10_BUILTIN_128BIT_DIVEU_V1TI,
23446 P10_BUILTIN_128BIT_MODU_V1TI, P10_BUILTIN_CMPGE_U1TI,
23447 P10_BUILTIN_VCMPGTUT, P10_BUILTIN_VCMPEQUT]: New case statements.
23448 * config/rs6000/rs6000.c (rs6000_handle_altivec_attribute) [E_TImode,
23449 E_V1TImode]: New case statements.
23450 * config/rs6000/rs6000.h (rs6000_builtin_type_index): New enum
23451 value RS6000_BTI_bool_V1TI.
23452 * config/rs6000/vector.md (vector_gtv1ti,vector_nltv1ti,
23453 vector_gtuv1ti, vector_nltuv1ti, vector_ngtv1ti, vector_ngtuv1ti,
23454 vector_eq_v1ti_p, vector_ne_v1ti_p, vector_ae_v1ti_p,
23455 vector_gt_v1ti_p, vector_gtu_v1ti_p, vrotlv1ti3, vashlv1ti3,
23456 vlshrv1ti3, vashrv1ti3): New define_expands.
23457 * config/rs6000/vsx.md (UNSPEC_VSX_DIVSQ, UNSPEC_VSX_DIVUQ,
23458 UNSPEC_VSX_DIVESQ, UNSPEC_VSX_DIVEUQ, UNSPEC_VSX_MODSQ,
23459 UNSPEC_VSX_MODUQ): New unspecs.
23460 (mulv2di3, vsx_div_v1ti, vsx_udiv_v1ti, vsx_dives_v1ti,
23461 vsx_diveu_v1ti, vsx_mods_v1ti, vsx_modu_v1ti, xxswapd_v1ti): New
23463 (vcmpnet): New define_expand.
23464 * doc/extend.texi: Add documentation for the new builtins vec_rl,
23465 vec_rlmi, vec_rlnm, vec_sl, vec_sr, vec_sra, vec_mule, vec_mulo,
23466 vec_div, vec_dive, vec_mod, vec_cmpeq, vec_cmpne, vec_cmpgt, vec_cmplt,
23467 vec_cmpge, vec_cmple, vec_all_eq, vec_all_ne, vec_all_gt, vec_all_lt,
23468 vec_all_ge, vec_all_le, vec_any_eq, vec_any_ne, vec_any_gt, vec_any_lt,
23469 vec_any_ge, vec_any_le.
23471 2021-06-09 Carl Love <cel@us.ibm.com>
23473 * config/rs6000/altivec.md (altivec_vrl<VI_char>mi): Fix
23474 bug in argument generation.
23476 2021-06-09 Christophe Lyon <christophe.lyon@linaro.org>
23478 * config/arm/iterators.md (<supf>): Remove VCLZQ_U, VCLZQ_S.
23480 * config/arm/mve.md (mve_vclzq_<supf><mode>): Add '@' prefix,
23481 remove <supf> iterator.
23482 (mve_vclzq_u<mode>): New.
23483 * config/arm/neon.md (clz<mode>2): Rename to neon_vclz<mode>.
23484 (neon_vclz<mode): Move to ...
23485 * config/arm/unspecs.md (VCLZQ_U, VCLZQ_S): Remove.
23486 * config/arm/vec-common.md: ... here. Add support for MVE.
23488 2021-06-09 Christophe Lyon <christophe.lyon@linaro.org>
23490 * config/arm/mve.md (mve_vhaddq_<supf><mode>): Prefix with '@'.
23491 (@mve_vrhaddq_<supf><mode): Likewise.
23492 * config/arm/neon.md (neon_v<r>hadd<sup><mode>): Likewise.
23493 * config/arm/vec-common.md (avg<mode>3_floor, uavg<mode>3_floor)
23494 (avg<mode>3_ceil", uavg<mode>3_ceil): New patterns.
23496 2021-06-09 imba-tjd <109224573@qq.com>
23498 * doc/invoke.texi: Fix typo.
23500 2021-06-09 Roger Sayle <roger@nextmovesoftware.com>
23502 PR middle-end/53267
23503 * fold-const-call.c (fold_const_call_sss) [CASE_CFN_FMOD]:
23504 Support evaluation of fmod/fmodf/fmodl at compile-time.
23506 2021-06-09 Richard Biener <rguenther@suse.de>
23508 PR tree-optimization/100981
23509 * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
23510 gimple_get_lhs to also handle calls.
23511 * tree-vect-slp-patterns.c (complex_pattern::build): Transfer
23514 2021-06-09 Richard Biener <rguenther@suse.de>
23516 PR tree-optimization/97832
23517 * tree-vectorizer.h (_slp_tree::failed): New.
23518 * tree-vect-slp.c (_slp_tree::_slp_tree): Initialize
23520 (_slp_tree::~_slp_tree): Free failed.
23521 (vect_build_slp_tree): Retain failed nodes and record
23522 matches in them, copying that back out when running
23523 into a cached fail. Dump start and end of discovery.
23524 (dt_sort_cmp): New.
23525 (vect_build_slp_tree_2): Handle associatable chains
23526 together doing more aggressive operand swapping.
23528 2021-06-09 H.J. Lu <hjl.tools@gmail.com>
23531 * config.gcc (gcc_cv_initfini_array): Set to yes for Linux and
23533 * doc/install.texi: Require glibc 2.1 and binutils 2.12 for
23534 Linux and GNU targets.
23536 2021-06-09 Richard Biener <rguenther@suse.de>
23538 * tree-vect-stmts.c (vect_is_simple_use): Always get dt
23541 2021-06-09 Claudiu Zissulescu <claziss@synopsys.com>
23543 * config/arc/arc.md (loop_end): Change it to
23544 define_insn_and_split.
23546 2021-06-09 Claudiu Zissulescu <claziss@synopsys.com>
23548 * config/arc/arc.md (maddhisi4): Use VMAC2H instruction.
23549 (machi): New pattern.
23550 (umaddhisi4): Use VMAC2HU instruction.
23551 (umachi): New pattern.
23553 2021-06-09 Claudiu Zissulescu <claziss@synopsys.com>
23555 * config/arc/arc-protos.h (arc_split_move_p): New prototype.
23556 * config/arc/arc.c (arc_split_move_p): New function.
23557 (arc_split_move): Clean up.
23558 * config/arc/arc.md (movdi_insn): Clean up, use arc_split_move_p.
23559 (movdf_insn): Likewise.
23560 * config/arc/simdext.md (mov<VWH>_insn): Likewise.
23562 2021-06-09 Uroš Bizjak <ubizjak@gmail.com>
23565 * config/i386/i386.c (print_operand_address_as): Rename "no_rip"
23566 argument to "raw". Do not emit segment overrides when "raw" is true.
23568 2021-06-09 Martin Liska <mliska@suse.cz>
23570 * doc/gcov.texi: Create a proper JSON files.
23571 * doc/invoke.texi: Remove dots in order to make it a valid
23574 2021-06-09 Xionghu Luo <luoxhu@linux.ibm.com>
23576 * config/rs6000/rs6000-p8swap.c (pattern_is_rotate64): New.
23577 (insn_is_load_p): Use pattern_is_rotate64.
23578 (insn_is_swap_p): Likewise.
23579 (quad_aligned_load_p): Likewise.
23580 (const_load_sequence_p): Likewise.
23581 (replace_swapped_aligned_load): Likewise.
23582 (recombine_lvx_pattern): Likewise.
23583 (recombine_stvx_pattern): Likewise.
23585 2021-06-09 Andrew MacLeod <amacleod@redhat.com>
23587 * gimple-range-gori.cc (gori_compute::outgoing_edge_range_p): Use a
23588 fur_stmt source record.
23589 * gimple-range.cc (fur_source::get_operand): Generic range query.
23590 (fur_source::get_phi_operand): New.
23591 (fur_source::register_dependency): New.
23592 (fur_source::query): New.
23593 (class fur_edge): New. Edge source for operands.
23594 (fur_edge::fur_edge): New.
23595 (fur_edge::get_operand): New.
23596 (fur_edge::get_phi_operand): New.
23597 (fur_edge::query): New.
23598 (fur_stmt::fur_stmt): New.
23599 (fur_stmt::get_operand): New.
23600 (fur_stmt::get_phi_operand): New.
23601 (fur_stmt::query): New.
23602 (class fur_depend): New. Statement source and process dependencies.
23603 (fur_depend::fur_depend): New.
23604 (fur_depend::register_dependency): New.
23605 (class fur_list): New. List source for operands.
23606 (fur_list::fur_list): New.
23607 (fur_list::get_operand): New.
23608 (fur_list::get_phi_operand): New.
23609 (fold_range): New. Instantiate appropriate fur_source class and fold.
23610 (fold_using_range::range_of_range_op): Use new API.
23611 (fold_using_range::range_of_address): Ditto.
23612 (fold_using_range::range_of_phi): Ditto.
23613 (imple_ranger::fold_range_internal): Use fur_depend class.
23614 (fold_using_range::range_of_ssa_name_with_loop_info): Use new API.
23615 * gimple-range.h (class fur_source): Now a base class.
23616 (class fur_stmt): New.
23617 (fold_range): New prototypes.
23618 (fur_source::fur_source): Delete.
23620 2021-06-08 Andrew Pinski <apinski@marvell.com>
23622 PR tree-optimization/25290
23623 * tree-ssa-phiopt.c (xor_replacement): Delete.
23624 (tree_ssa_phiopt_worker): Delete use of xor_replacement.
23625 (match_simplify_replacement): Allow one cheap preparation
23626 statement that can be moved to before the if.
23628 2021-06-08 Pat Haugen <pthaugen@linux.ibm.com>
23630 * config/rs6000/power10.md (power10-fused-load, power10-fused-store,
23631 power10-fused_alu, power10-fused-vec, power10-fused-branch): New.
23633 2021-06-08 Jeff Law <jeffreyalaw@gmail.com>
23635 * config/h8300/logical.md (andqi3_1): Move BCLR case into define_insn_and_split.
23636 Create length attribute on define_insn_and_split. Only split for cases which we
23638 (andqi3_1<cczn>): Renamed from andqi3_1_clobber_flags. Only handle AND here and
23639 fix length computation.
23640 (b<code><mode>msx): Combine QImode and HImode H8/SX patterns using iterator.
23642 2021-06-08 Richard Biener <rguenther@suse.de>
23644 PR tree-optimization/100923
23645 * tree-ssa-sccvn.c (valueize_refs_1): Take a pointer to
23646 the operand vector to be valueized.
23647 (valueize_refs): Likewise.
23648 (valueize_shared_reference_ops_from_ref): Adjust.
23649 (valueize_shared_reference_ops_from_call): Likewise.
23650 (vn_reference_lookup_3): Likewise.
23651 (vn_reference_lookup_pieces): Likewise. Re-valueize
23652 with honoring availability when we are about to create
23653 the ao_ref and valueized before.
23654 (vn_reference_lookup): Likewise.
23655 (vn_reference_insert_pieces): Adjust.
23657 2021-06-08 Richard Biener <rguenther@suse.de>
23659 * tree-vectorizer.h (_slp_instance::root_stmt): Change to...
23660 (_slp_instance::root_stmts): ... a vector.
23661 (SLP_INSTANCE_ROOT_STMT): Rename to ...
23662 (SLP_INSTANCE_ROOT_STMTS): ... this.
23663 (slp_root::root): Change to...
23664 (slp_root::roots): ... a vector.
23665 (slp_root::slp_root): Adjust.
23666 * tree-vect-slp.c (_slp_instance::location): Adjust.
23667 (vect_free_slp_instance): Release the root stmt vector.
23668 (vect_build_slp_instance): Adjust.
23669 (vect_analyze_slp): Likewise.
23670 (_bb_vec_info::~_bb_vec_info): Likewise.
23671 (vect_slp_analyze_operations): Likewise.
23672 (vect_bb_vectorization_profitable_p): Likewise. Adjust
23673 costs for the root stmt.
23674 (vect_slp_check_for_constructors): Gather all BIT_INSERT_EXPRs
23676 (vect_slp_analyze_bb_1): Simplify by marking all root stmts
23678 (vectorize_slp_instance_root_stmt): Adjust.
23679 (vect_schedule_slp): Likewise.
23681 2021-06-08 Aldy Hernandez <aldyh@redhat.com>
23683 * gimple-ssa-evrp.c (class ssa_equiv_stack): New.
23684 (ssa_equiv_stack::ssa_equiv_stack): New.
23685 (ssa_equiv_stack::~ssa_equiv_stack): New.
23686 (ssa_equiv_stack::enter): New.
23687 (ssa_equiv_stack::leave): New.
23688 (ssa_equiv_stack::push_replacement): New.
23689 (ssa_equiv_stack::get_replacement): New.
23690 (is_pointer_ssa): New.
23691 (class pointer_equiv_analyzer): New.
23692 (pointer_equiv_analyzer::pointer_equiv_analyzer): New.
23693 (pointer_equiv_analyzer::~pointer_equiv_analyzer): New.
23694 (pointer_equiv_analyzer::set_global_equiv): New.
23695 (pointer_equiv_analyzer::set_cond_equiv): New.
23696 (pointer_equiv_analyzer::get_equiv): New.
23697 (pointer_equiv_analyzer::enter): New.
23698 (pointer_equiv_analyzer::leave): New.
23699 (pointer_equiv_analyzer::get_equiv_expr): New.
23700 (pta_valueize): New.
23701 (pointer_equiv_analyzer::visit_stmt): New.
23702 (pointer_equiv_analyzer::visit_edge): New.
23703 (hybrid_folder::value_of_expr): Call PTA.
23704 (hybrid_folder::value_on_edge): Same.
23705 (hybrid_folder::pre_fold_bb): New.
23706 (hybrid_folder::post_fold_bb): New.
23707 (hybrid_folder::pre_fold_stmt): New.
23708 (rvrp_folder::pre_fold_bb): New.
23709 (rvrp_folder::post_fold_bb): New.
23710 (rvrp_folder::pre_fold_stmt): New.
23711 (rvrp_folder::value_of_expr): Call PTA.
23712 (rvrp_folder::value_on_edge): Same.
23714 2021-06-08 Jakub Jelinek <jakub@redhat.com>
23717 * tree-inline.c (copy_tree_body_r): For OMP_CLAUSE_DEPEND don't
23718 check TREE_CODE if OMP_CLAUSE_DECL is NULL.
23720 2021-06-08 Richard Biener <rguenther@suse.de>
23722 PR middle-end/100951
23723 * tree-vect-generic.c (expand_vector_piecewise): Build a
23724 VECTOR_CST if all elements are constant.
23725 (expand_vector_condition): Likewise.
23726 (lower_vec_perm): Likewise.
23727 (expand_vector_conversion): Likewise.
23729 2021-06-08 Martin Liska <mliska@suse.cz>
23731 * doc/invoke.texi: Document new param evrp-sparse-threshold.
23733 2021-06-08 Martin Liska <mliska@suse.cz>
23735 * genautomata.c (create_automata): Fix typo.
23737 2021-06-08 Kewen Lin <linkw@linux.ibm.com>
23739 PR tree-optimization/100794
23740 * tree-predcom.c (tree_predictive_commoning_loop): Add parameter
23741 allow_unroll_p and only allow unrolling when it's true.
23742 (tree_predictive_commoning): Add parameter allow_unroll_p and
23744 (run_tree_predictive_commoning): Likewise.
23745 (pass_predcom::gate): Check flag_tree_loop_vectorize and
23746 global_options_set.x_flag_predictive_commoning.
23747 (pass_predcom::execute): Adjust for allow_unroll_p.
23749 2021-06-08 Kewen Lin <linkw@linux.ibm.com>
23751 * tree-predcom.c (execute_pred_commoning): Remove update_ssa call.
23752 (tree_predictive_commoning_loop): Factor some cleanup stuffs into
23753 lambda function cleanup, remove scev_reset call, and adjust return
23755 (tree_predictive_commoning): Adjust for different changed values,
23756 only set flag TODO_update_ssa_only_virtuals if changed.
23757 (pass_data pass_data_predcom): Remove TODO_update_ssa_only_virtuals
23758 from todo_flags_finish.
23760 2021-06-07 Andrew MacLeod <amacleod@redhat.com>
23762 * gimple-range-cache.cc (class sbr_sparse_bitmap): New.
23763 (sbr_sparse_bitmap::sbr_sparse_bitmap): New.
23764 (sbr_sparse_bitmap::bitmap_set_quad): New.
23765 (sbr_sparse_bitmap::bitmap_get_quad): New.
23766 (sbr_sparse_bitmap::set_bb_range): New.
23767 (sbr_sparse_bitmap::get_bb_range): New.
23768 (sbr_sparse_bitmap::bb_range_p): New.
23769 (block_range_cache::block_range_cache): initialize bitmap obstack.
23770 (block_range_cache::~block_range_cache): Destruct obstack.
23771 (block_range_cache::set_bb_range): Decide when to utilze the
23772 sparse on entry cache.
23773 * gimple-range-cache.h (block_range_cache): Add bitmap obstack.
23774 * params.opt (-param=evrp-sparse-threshold): New.
23776 2021-06-07 Andrew MacLeod <amacleod@redhat.com>
23778 * bitmap.c (bitmap_set_aligned_chunk): New.
23779 (bitmap_get_aligned_chunk): New.
23780 (test_aligned_chunk): New.
23781 (bitmap_c_tests): Call test_aligned_chunk.
23782 * bitmap.h (bitmap_set_aligned_chunk, bitmap_get_aligned_chunk): New.
23784 2021-06-07 Uroš Bizjak <ubizjak@gmail.com>
23787 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
23789 (ix86_expand_vector_init_one_nonzero): Ditto.
23790 (ix86_expand_vector_init_one_var): Ditto.
23791 (ix86_expand_vector_init_general): Ditto.
23792 * config/i386/mmx.md (vec_initv4qiqi): New expander.
23794 2021-06-07 Jeff Law <jeffreyalaw@gmail.com>
23796 * config/h8300/movepush.md: Change most _clobber_flags
23797 patterns to instead use <cczn> subst.
23798 (movsi_cczn): New pattern with usable CC cases split out.
23799 (movsi_h8sx_cczn): Likewise.
23801 2021-06-07 Martin Liska <mliska@suse.cz>
23803 * common/common-target.def: Split long lines and replace them
23805 * target.def: Likewise.
23806 * doc/tm.texi: Re-generated.
23808 2021-06-07 Jakub Jelinek <jakub@redhat.com>
23811 * fold-const.c (fold_read_from_vector): Return NULL if trying to
23812 read from a CONSTRUCTOR with vector type elements.
23814 2021-06-07 Jakub Jelinek <jakub@redhat.com>
23816 PR middle-end/100898
23817 * tree-inline.c (copy_bb): Only use gimple_call_arg_ptr if memcpy
23818 should copy any arguments. Don't call gimple_call_num_args
23819 on id->call_stmt or call_stmt more than once.
23821 2021-06-07 liuhongt <hongtao.liu@intel.com>
23824 * config/i386/sse.md (*sse4_1_zero_extendv8qiv8hi2_3): Refine
23826 (<insn>v4siv4di2): Delete constraints for define_expand.
23828 2021-06-07 liuhongt <hongtao.liu@intel.com>
23831 * config/i386/i386-expand.c (ix86_expand_builtin): Remove
23832 assignment of cfun->machine->has_explicit_vzeroupper.
23833 * config/i386/i386-features.c
23834 (ix86_add_reg_usage_to_vzerouppers): Delete.
23835 (ix86_add_reg_usage_to_vzeroupper): Ditto.
23836 (rest_of_handle_insert_vzeroupper): Remove
23837 ix86_add_reg_usage_to_vzerouppers, add df_analyze at the end
23839 (gate): Remove cfun->machine->has_explicit_vzeroupper.
23840 * config/i386/i386-protos.h (ix86_expand_avx_vzeroupper):
23842 * config/i386/i386.c (ix86_insn_callee_abi): New function.
23843 (ix86_initialize_callee_abi): Ditto.
23844 (ix86_expand_avx_vzeroupper): Ditto.
23845 (ix86_hard_regno_call_part_clobbered): Adjust for vzeroupper
23847 (TARGET_INSN_CALLEE_ABI): Define as ix86_insn_callee_abi.
23848 (ix86_emit_mode_set): Call ix86_expand_avx_vzeroupper
23850 * config/i386/i386.h (struct GTY(()) machine_function): Delete
23851 has_explicit_vzeroupper.
23852 * config/i386/i386.md (enum unspec): New member
23854 (ABI_DEFAULT,ABI_VZEROUPPER,ABI_UNKNOWN): New
23855 define_constants for insn callee abi index.
23856 * config/i386/predicates.md (vzeroupper_pattern): Adjust.
23857 * config/i386/sse.md (UNSPECV_VZEROUPPER): Deleted.
23858 (avx_vzeroupper): Call ix86_expand_avx_vzeroupper.
23859 (*avx_vzeroupper): Rename to ..
23860 (avx_vzeroupper_callee_abi): .. this, and adjust pattern as
23861 call_insn which has a special vzeroupper ABI.
23862 (*avx_vzeroupper_1): Deleted.
23864 2021-06-07 liuhongt <hongtao.liu@intel.com>
23867 * df-scan.c (df_get_call_refs): When call_insn is a fake call,
23868 it won't use stack pointer reg.
23869 * final.c (leaf_function_p): When call_insn is a fake call, it
23870 won't affect caller as a leaf function.
23871 * reg-stack.c (callee_clobbers_any_stack_reg): New.
23872 (subst_stack_regs): When call_insn doesn't clobber any stack
23873 reg, don't clear the arguments.
23874 * rtl.c (shallow_copy_rtx): Don't clear flag used when orig is
23876 * shrink-wrap.c (requires_stack_frame_p): No need for stack
23877 frame for a fake call.
23878 * rtl.h (FAKE_CALL_P): New macro.
23880 2021-06-06 Eric Botcazou <ebotcazou@adacore.com>
23882 * config/sparc/sparc-protos.h (order_regs_for_local_alloc): Rename
23884 (sparc_order_regs_for_local_alloc): ...this.
23885 (sparc_leaf_reg_remap): Declare.
23886 * config/sparc/sparc.h (ADJUST_REG_ALLOC_ORDER): Adjust.
23887 (LEAF_REG_REMAP): Reimplement as call to sparc_leaf_reg_remap.
23888 * config/sparc/sparc.c (leaf_reg_remap): Delete.
23889 (order_regs_for_local_alloc): Rename to...
23890 (sparc_order_regs_for_local_alloc): ...this.
23891 (sparc_leaf_reg_remap): New function.
23892 (sparc_conditional_register_usage): Do not modify leaf_reg_remap.
23894 2021-06-06 David Edelsohn <dje.gcc@gmail.com>
23896 * config/rs6000/rs6000.c (rs6000_xcoff_asm_output_aligned_decl_common):
23897 Use assemble_name to output BSS section name.
23899 2021-06-06 Uroš Bizjak <ubizjak@gmail.com>
23901 * config/i386/constraints.md (Bs):
23902 Remove boolean operators from match_test RTX.
23905 (M): Use "mode" variable instead of GET_MODE (op) in match_test RTX.
23908 2021-06-06 Martin Liska <mliska@suse.cz>
23910 * doc/extend.texi: Add missing @headitem.
23911 * doc/invoke.texi: Likewise.
23912 * doc/objc.texi: Likewise.
23914 2021-06-06 Martin Liska <mliska@suse.cz>
23916 * genhooks.c (emit_findices): Remove unused function.
23917 (emit_documentation): Do not call emit_findices
23918 and do not search for @Fcode directives.
23920 2021-06-06 Martin Liska <mliska@suse.cz>
23922 * doc/invoke.texi: Remove extra character.
23924 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
23926 * config/sh/sh.md (doloop_end_split): Fix empty split condition.
23928 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
23930 * config/sparc/sparc.md (*snedi<W:mode>_zero_vis3,
23931 *neg_snedi<W:mode>_zero_subxc, *plus_snedi<W:mode>_zero,
23932 *plus_plus_snedi<W:mode>_zero, *minus_snedi<W:mode>_zero,
23933 *minus_minus_snedi<W:mode>_zero): Fix empty split condition.
23935 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
23937 * config/or1k/or1k.md (*movdi): Fix empty split condition.
23939 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
23941 * config/mips/mips.md (<anonymous>, bswapsi2, bswapdi2): Fix empty
23944 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
23946 * config/m68k/m68k.md (*zero_extend_inc, *zero_extend_dec,
23947 *zero_extendsidi2): Fix empty split condition.
23949 2021-06-05 Jeff Law <jeffreyalaw@gmail.com>
23951 * config/h8300/addsub.md: Fix split condition in define_insn_and_split
23953 * config/h8300/bitfield.md: Likewise.
23954 * config/h8300/combiner.md: Likewise.
23955 * config/h8300/divmod.md: Likewise.
23956 * config/h8300/extensions.md: Likewise.
23957 * config/h8300/jumpcall.md: Likewise.
23958 * config/h8300/movepush.md: Likewise.
23959 * config/h8300/multiply.md: Likewise.
23960 * config/h8300/other.md: Likewise.
23961 * config/h8300/shiftrotate.md: Likewise.
23962 * config/h8300/logical.md: Likewise. Fix split pattern to use
23963 code iterator that somehow slipped through.
23965 2021-06-04 Tobias Burnus <tobias@codesourcery.com>
23967 PR middle-end/100905
23968 * tree-nested.c (convert_nonlocal_omp_clauses,
23969 convert_local_omp_clauses): Handle OMP_CLAUSE_BIND.
23971 2021-06-04 Martin Sebor <msebor@redhat.com>
23973 PR middle-end/100732
23974 * gimple-fold.c (gimple_fold_builtin_sprintf): Avoid folding calls
23975 with either source or destination argument of invalid type.
23976 * tree-ssa-uninit.c (maybe_warn_pass_by_reference): Avoid checking
23977 calls with arguments of invalid type.
23979 2021-06-04 Martin Sebor <msebor@redhat.com>
23981 * attribs.c (init_attr_rdwr_indices): Use VLA bounds in the expected
23983 (attr_access::vla_bounds): Also handle VLA bounds.
23985 2021-06-04 Uroš Bizjak <ubizjak@gmail.com>
23987 * config/i386/predicates.md (GOT_memory_operand):
23988 Implement using match_code RTXes.
23989 (GOT32_symbol_operand): Ditto.
23991 2021-06-04 Uroš Bizjak <ubizjak@gmail.com>
23994 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
23996 (ix86_expand_vector_init_general): Ditto.
23997 Use SImode instead of word_mode for logic operations
23998 when GET_MODE_SIZE (mode) < UNITS_PER_WORD.
23999 (expand_vec_perm_even_odd_1): Assert that V2HI mode should be
24000 implemented by expand_vec_perm_1.
24001 (expand_vec_perm_broadcast_1): Assert that V2HI and V4HI modes
24002 should be implemented using standard shuffle patterns.
24003 (ix86_vectorize_vec_perm_const): Handle V2HImode. Add V4HI and
24004 V2HI modes to modes, implementable with shuffle for one operand.
24005 * config/i386/mmx.md (*punpckwd): New insn_and_split pattern.
24006 (*pshufw_1): New insn pattern.
24007 (*vec_dupv2hi): Ditto.
24008 (vec_initv2hihi): New expander.
24010 2021-06-04 Kewen Lin <linkw@linux.ibm.com>
24012 * config/arm/vfp.md (no_literal_pool_df_immediate,
24013 no_literal_pool_sf_immediate): Fix empty split condition.
24015 2021-06-04 Kewen Lin <linkw@linux.ibm.com>
24017 * config/i386/i386.md (*load_tp_x32_zext, *add_tp_x32_zext,
24018 *tls_dynamic_gnu2_combine_32): Fix empty split condition.
24019 * config/i386/sse.md (*<sse2_avx2>_pmovmskb_lt,
24020 *<sse2_avx2>_pmovmskb_zext_lt, *sse2_pmovmskb_ext_lt,
24021 *<sse4_1_avx2>_pblendvb_lt): Likewise.
24023 2021-06-04 Jakub Jelinek <jakub@redhat.com>
24026 * config/i386/i386-expand.c (ix86_expand_vector_init): Handle
24027 concatenation from half-sized modes with TImode elements.
24029 2021-06-04 Claudiu Zissulescu <claziss@synopsys.com>
24031 * config/arc/arc.c (arc_override_options): Disable millicode
24032 thunks when RF16 is on.
24034 2021-06-04 Haochen Gui <guihaoc@gcc.gnu.org>
24036 * config/rs6000/rs6000.h (PROMOTE_MODE): Remove.
24038 2021-06-04 Haochen Gui <guihaoc@gcc.gnu.org>
24040 * config/rs6000/rs6000-call.c (rs6000_promote_function_mode):
24041 Replace PROMOTE_MODE marco with its content.
24043 2021-06-03 Kewen Lin <linkw@linux.ibm.com>
24045 * config/cris/cris.md (*addi_reload): Fix empty split condition.
24047 2021-06-03 Jim Wilson <jimw@sifive.com>
24049 * config.gcc (riscv*-*-*): If --with-riscv-attribute not used,
24050 turn it on for all riscv targets.
24052 2021-06-03 Uroš Bizjak <ubizjak@gmail.com>
24055 * config/i386/i386-expand.c (ix86_expand_vector_set):
24056 Handle V2HI and V4QI modes.
24057 (ix86_expand_vector_extract): Ditto.
24058 * config/i386/mmx.md (*pinsrw): New insn pattern.
24061 (*pextrw_zext): Ditto.
24063 (*pextrb_zext): Ditto.
24064 (vec_setv2hi): New expander.
24065 (vec_extractv2hihi): Ditto.
24066 (vec_setv4qi): Ditto.
24067 (vec_extractv4qiqi): Ditto.
24068 (vec_setv8qi): Enable only for TARGET_SSE4_1.
24069 (vec_extractv8qiqi): Ditto.
24071 2021-06-03 Aaron Sawdey <acsawdey@linux.ibm.com>
24073 * config/rs6000/genfusion.pl (gen_logical_addsubf): Fix input
24074 order to subf instruction.
24075 * config/rs6000/fusion.md: Regenerate.
24077 2021-06-03 Aldy Hernandez <aldyh@redhat.com>
24079 * calls.c (get_size_range): Use range_of_expr instead of
24080 determine_value_range.
24081 * tree-affine.c (expr_to_aff_combination): Same.
24082 * tree-data-ref.c (split_constant_offset): Same.
24083 * tree-vrp.c (determine_value_range_1): Remove.
24084 (determine_value_range): Remove.
24085 * tree-vrp.h (determine_value_range): Remove.
24087 2021-06-03 Aldy Hernandez <aldyh@redhat.com>
24089 * function-tests.c (test_ranges): Call gimple_range_tests.
24090 * gimple-range-cache.cc (ranger_cache::range_of_expr): Pass stmt
24092 * gimple-range.cc (fur_source::get_operand): Do not call
24093 get_tree_range or gimple_range_global.
24095 (get_tree_range): Move to value-query.cc.
24096 Call get_arith_expr_range.
24097 (gimple_ranger::range_of_expr): Add argument to get_tree_range.
24098 Include gimple-range-tests.cc.
24099 * gimple-range.h (fold_range): Add argument.
24100 (get_tree_range): Remove.
24101 * selftest.h (gimple_range_tests): New.
24102 * value-query.cc (global_range_query::range_of_expr): Add
24104 (range_query::get_tree_range): Move from gimple-range.cc.
24105 * value-query.h (class range_query): Add get_tree_range and
24106 get_arith_expr_range. Make fur_source a friend.
24107 * vr-values.c (vr_values::range_of_expr): Pass stmt to
24109 * gimple-range-tests.cc: New file.
24111 2021-06-03 Aldy Hernandez <aldyh@redhat.com>
24113 * gimple-range.cc (gimple_ranger::export_global_ranges): Call
24114 update_global_range.
24115 * value-query.cc (update_global_range): New.
24116 * value-query.h (update_global_range): New.
24118 2021-06-03 David Malcolm <dmalcolm@redhat.com>
24120 * diagnostic-show-locus.c (diagnostic_show_locus): Don't reject
24121 printing the same location twice if there are fix-it hints,
24122 multiple locations, or a label.
24124 2021-06-03 Andre Vieira <andre.simoesdiasvieira@arm.com>
24126 * tree-vect-loop.c (vect_transform_loop): Use main loop's various'
24127 thresholds to narrow the upper bound on epilogue iterations.
24129 2021-06-03 Christophe Lyon <christophe.lyon@linaro.org>
24131 * config/arm/mve.md (mve_vabsq_f<mode>): Use 'abs' instead of unspec.
24132 (mve_vabsq_s<mode>): Likewise.
24133 * config/arm/neon.md (abs<mode>2): Rename to neon_abs<mode>2.
24134 * config/arm/unspecs.md (VABSQ_F, VABSQ_S): Delete.
24135 * config/arm/vec-common.md (neg<mode>2): Rename to
24136 <absneg_str><mode>2.
24138 2021-06-03 Claudiu Zissulescu <claziss@synopsys.com>
24140 * common/config/arc/arc-common.c (arc_option_optimization_table):
24141 Remove malign-call.
24142 * config/arc/arc.c (arc_unalign_branch_p): Remove unused function.
24143 * config/arc/arc.h (TARGET_MIXED_CODE): Remove macro.
24144 (INDEX_REG_CLASS): Only refer to GENERAL_REGS.
24145 * config/arc/arc.md (abssi2_mixed): Remove pattern.
24146 * config/arc/arc.opt (munalign-prob-threshold): Mark it obsolete.
24147 (malign-call): Likewise.
24148 (mmixed-code): Likewise.
24149 * doc/invoke.texi (ARC): Update doc.
24151 2021-06-03 Martin Liska <mliska@suse.cz>
24153 * common.opt: Use proper Enum values.
24154 * opts.c (COVERAGE_SANITIZER_OPT): Remove.
24155 (parse_sanitizer_options): Handle only sanitizer_opts.
24156 (common_handle_option): Just assign value.
24158 2021-06-03 Eric Botcazou <ebotcazou@adacore.com>
24161 * tree-inline.c (inline_forbidden_p): Remove test on return type.
24163 2021-06-03 Eric Botcazou <ebotcazou@adacore.com>
24165 * dwarf2out.c (loc_list_from_tree_1) <FUNCTION_DECL>: Also generate
24166 DW_OP_GNU_variable_value referencing an existing DIE at file scope.
24167 (type_byte_size): Inline into...
24168 (add_byte_size_attribute): ...this and call add_scalar_info.
24170 2021-06-03 Eric Botcazou <ebotcazou@adacore.com>
24172 * dwarf2out.c (mem_loc_descriptor) <UDIV>: Fix typo.
24173 (typed_binop_from_tree): New function.
24174 (loc_list_from_tree_1) <EXACT_DIV_EXPR>: For an unsigned type,
24175 turn a divide by a power of 2 into a shift.
24176 <CEIL_DIV_EXPR>: For an unsigned type, use a signed divide if the
24177 size of the mode is lower than DWARF2_ADDR_SIZE; otherwise, do a
24178 typed divide by calling typed_binop_from_tree.
24180 2021-06-03 Eric Botcazou <ebotcazou@adacore.com>
24182 * dwarf2out.c (scompare_loc_descriptor): Fix head comment.
24183 (is_handled_procedure_type): Likewise.
24184 (struct loc_descr_context): Add strict_signedness field.
24185 (resolve_args_picking_1): Deal with DW_OP_[GNU_]deref_type,
24186 DW_OP_[GNU_]convert and DW_OP_[GNU_]reinterpret.
24187 (resolve_args_picking): Minor tweak.
24188 (function_to_dwarf_procedure): Initialize strict_signedness field.
24189 (type_byte_size): Likewise.
24190 (field_byte_offset): Likewise.
24191 (gen_descr_array_type_die): Likewise.
24192 (gen_variant_part): Likewise.
24193 (loc_list_from_tree_1) <CALL_EXPR>: Tidy up and set strict_signedness
24194 to true when a context is present before evaluating the arguments.
24195 <COND_EXPR>: Do not generate a useless comparison with zero.
24196 When dereferencing an address, if strict_signedness is true and the
24197 type is small and signed, use DW_OP_deref_type to do the dereference
24198 and then DW_OP_convert to convert back to the generic type.
24200 2021-06-03 Jakub Jelinek <jakub@redhat.com>
24203 * tree-inline.c (copy_tree_body_r): Handle iterators on
24204 OMP_CLAUSE_AFFINITY or OMP_CLAUSE_DEPEND.
24206 2021-06-03 Kewen Lin <linkw@linux.ibm.com>
24208 * config/arc/arc.md (*bbit_di): Remove.
24210 2021-06-02 Christoph Muellner <cmuellner@gcc.gnu.org>
24212 PR rtl-optimization/100264
24213 * ree.c (get_sub_rtx): Ignore SET expressions without register
24214 destinations and remove assertion, as it is not valid anymore
24215 with this new behaviour.
24216 (merge_def_and_ext): Eliminate destination check for register
24217 as such SET expressions can't occur anymore.
24218 (combine_reaching_defs): Likewise.
24220 2021-06-02 Jakub Jelinek <jakub@redhat.com>
24223 * config/xtensa/xtensa.h (LEAF_REG_REMAP): Cast REGNO to int to avoid
24224 -Wtype-limits warnings.
24225 (DWARF_FRAME_REGISTER): Rewrite into ternary operator with addition
24226 in operands to avoid -Wsign-compare warnings.
24228 2021-06-02 Pat Haugen <pthaugen@linux.ibm.com>
24230 * config/rs6000/rs6000-logue.c (rs6000_emit_prologue): Use
24233 2021-06-02 Vineet Gupta <vgupta@synopsys.com>
24235 * config/arc/arc.h (TARGET_CPU_DEFAULT): Change to hs38_linux.
24237 2021-06-02 Ilya Leoshkevich <iii@linux.ibm.com>
24239 * config/s390/s390.md(*ashrdi3_31<setcc><cconly>): Use a single
24241 * config/s390/subst.md(cconly_subst): Use a single constraint
24242 in (match_scratch).
24244 2021-06-02 Martin Liska <mliska@suse.cz>
24246 * ipa-icf.h: Use auto_vec for memory_access_types.
24248 2021-06-02 Jeff Law <jeffreyalaw@gmail.com>
24250 * config/h8300/h8300-protos.h (compute_a_shift_length): Drop unused
24251 argument from prototype.
24252 (output_logical_op): Add rtx_code argument.
24253 (compute_logical_op_length): Likewise.
24254 * config/h8300/h8300.c (h8300_and_costs): Pass additional argument
24255 to compute_a_shift_length.
24256 (output_logical_op); New argument with the rtx code rather than
24257 extracting it from an operand. Handle QImode too.
24258 (compute_logical_op_length): Similary.
24259 (compute_a_shift_length): Drop unused argument.
24260 * config/h8300/h8300.md (logicals): New code iterator.
24261 * config/h8300/logical.md (<code><mode>3 expander): Combine
24262 the "and" expander with the "ior"/"xor" expander.
24263 (bclr<mode>msx): Combine the QI/HI mode patterns.
24264 (<logical><mode>3 insns): Use code iterator rather than match_operator.
24265 Handle QImode as well. Update call to output_logical_op and
24266 compute_logical_op_length to pass in rtx_code
24267 Fix split condition on all define_insn_and_split patterns.
24268 (one_cmpl<mode>2<cczn>): Use <cczn> to support both clobbering
24269 the flags and setting ZN via existing define_subst.
24270 * config/h8300/shiftrotate.md: Drop unused argument from
24271 calls to compute_a_shift_length.
24272 Signed-off-by: Jeff Law <jeffreyalaw@gmail.com>
24274 2021-06-01 Andrew Pinski <apinski@marvell.com>
24276 PR tree-optimization/25290
24277 * tree-ssa-phiopt.c (match_simplify_replacement):
24279 (tree_ssa_phiopt_worker): Use match_simplify_replacement.
24280 (two_value_replacement): Change the comment about
24281 conditional_replacement.
24282 (conditional_replacement): Delete.
24284 2021-06-01 Andrew Pinski <apinski@marvell.com>
24286 PR tree-optimization/95481
24287 * tree-tailcall.c (find_tail_calls): Handle empty typed
24290 2021-06-01 Andrew Pinski <apinski@marvell.com>
24292 * gimplify.c (zero_sized_field_decl): Delete
24293 (zero_sized_type): Delete
24294 (gimplify_init_ctor_eval): Use is_empty_type instead
24295 of zero_sized_field_decl.
24296 (gimplify_modify_expr): Use is_empty_type instead of
24299 2021-06-01 Jason Merrill <jason@redhat.com>
24302 * tree.h (CALL_FROM_NEW_OR_DELETE_P): Adjust comment.
24304 2021-06-01 Jason Merrill <jason@redhat.com>
24307 * diagnostic.h (warning_enabled_at): Declare.
24308 * diagnostic.c (diagnostic_enabled): Factor out from...
24309 (diagnostic_report_diagnostic): ...here.
24310 (warning_enabled_at): New.
24312 2021-06-01 Aldy Hernandez <aldyh@redhat.com>
24314 * gimple-ssa-evrp.c: Enable exporting of global ranges.
24316 2021-06-01 Martin Liska <mliska@suse.cz>
24319 * doc/invoke.texi: Mention that -fgcse-after-reload
24320 is enabled with -O3.
24322 2021-06-01 liuhongt <hongtao.liu@intel.com>
24324 PR tree-optimization/98365
24325 * tree-if-conv.c (strip_nop_cond_scalar_reduction): New function.
24326 (is_cond_scalar_reduction): Handle nop_expr in cond scalar reduction.
24327 (convert_scalar_cond_reduction): Ditto.
24328 (predicate_scalar_phi): Ditto.
24330 2021-06-01 Andrew MacLeod <amacleod@redhat.com>
24332 PR tree-optimization/100781
24333 * gimple-range-cache.cc (ranger_cache::ranger_cache): Enable new
24334 value calculation by default.
24335 (ranger_cache::enable_new_values): New.
24336 (ranger_cache::disable_new_values): New.
24337 (ranger_cache::push_poor_value): Check if new values are allowed.
24338 * gimple-range-cache.h (class ranger_cache): New member/methods.
24339 * gimple-range.cc (gimple_ranger::range_of_expr): Check for debug
24340 statement, and disable/renable new value calculation.
24342 2021-06-01 Andrew MacLeod <amacleod@redhat.com>
24344 * gimple-range-cache.cc (ranger_cache::ssa_range_in_bb): Delete.
24345 (ranger_cache::range_of_def): New.
24346 (ranger_cache::entry_range): New.
24347 (ranger_cache::exit_range): New.
24348 (ranger_cache::range_of_expr): Adjust.
24349 (ranger_cache::range_on_edge): Adjust.
24350 (ranger_cache::propagate_cache): Call exit_range directly.
24351 * gimple-range-cache.h (class ranger_cache): Adjust.
24353 2021-06-01 Andrew MacLeod <amacleod@redhat.com>
24355 * gimple-range-cache.cc (ranger_cache::ranger_cache): Adjust for
24356 gori_compute being a member rather than base class.
24357 dervied call to member call.
24358 (ranger_cache::dump): No longer dump gori_map.
24359 (ranger_cache::dump_bb): New.
24360 (ranger_cache::get_non_stale_global_range): Adjust for gori_compute
24361 being a member rather than base class.
24362 (ranger_cache::set_global_range): Ditto.
24363 (ranger_cache::ssa_range_in_bb): Ditto.
24364 (ranger_cache::range_of_expr): New.
24365 (ranger_cache::range_on_edge): New.
24366 (ranger_cache::block_range): Adjust for gori_computes. Debug changes.
24367 (ranger_cache::propagate_cache): Adjust debugging output.
24368 (ranger_cache::fill_block_cache): Adjust for gori_computes. Debug
24370 * gimple-range-cache.h (class ranger_cache): Make gori_compute a
24371 member, and inherit from range_query instead.
24372 (ranger_cache::dump_bb): New. split from dump.
24373 * gimple-range-gori.cc (gori_compute::ssa_range_in_bb): Delete.
24374 (gori_compute::expr_range_at_stmt): Delete.
24375 (gori_compute::compute_name_range_op): Delete.
24376 (gori_compute::compute_operand_range_switch): Add fur_source.
24377 (gori_compute::compute_operand_range): Add fur_source param, inline
24378 old compute_name_range_op and optimize_logical_operands.
24379 (struct tf_range): Delete.
24380 (gori_compute::logical_combine): Adjust
24381 (gori_compute::optimize_logical_operands): Delete.
24382 (gori_compute::compute_logical_operands_in_chain): Delete.
24383 (gori_compute::compute_logical_operands): Adjust.
24384 (gori_compute::compute_operand1_range): Adjust to fur_source.
24385 (gori_compute::compute_operand2_range): Ditto.
24386 (gori_compute::compute_operand1_and_operand2_range): Ditto.
24387 (gori_compute::outgoing_edge_range_p): Add range_query parameter,
24388 and adjust to fur_source.
24389 * gimple-range-gori.h (class gori_compute): Simplify and adjust to
24390 range_query and fur_source.
24391 * gimple-range.cc (gimple_ranger::range_on_edge): Query range_on_edge
24392 from the ranger_cache..
24393 (gimple_ranger::fold_range_internal): Adjust to base class change of
24395 (gimple_ranger::dump_bb): Adjust dump.
24396 * gimple-range.h (gimple_ranger):export gori computes object.
24398 2021-06-01 Andrew MacLeod <amacleod@redhat.com>
24400 PR tree-optimization/100774
24401 * gimple-range-cache.cc (ranger_cache::get_non_stale_global_range):
24402 Constant values are also not stale.
24403 (ranger_cache::set_global_range): Range invariant values should also
24404 have the correct timestamp.
24406 2021-05-31 Martin Liska <mliska@suse.cz>
24408 * tree-streamer-in.c (unpack_ts_function_decl_value_fields):
24409 Unpack FUNCTION_DECL_DECL_TYPE.
24410 * tree-streamer-out.c (pack_ts_function_decl_value_fields):
24411 Stream FUNCTION_DECL_DECL_TYPE instead of
24412 DECL_IS_OPERATOR_NEW_P.
24413 * tree.h (set_function_decl_type): Use FUNCTION_DECL_DECL_TYPE
24415 (DECL_IS_OPERATOR_NEW_P): Likewise.
24416 (DECL_IS_OPERATOR_DELETE_P): Likewise.
24417 (DECL_LAMBDA_FUNCTION_P): Likewise.
24419 2021-05-31 Richard Biener <rguenther@suse.de>
24422 * internal-fn.c (expand_SHUFFLEVECTOR): Define.
24423 * internal-fn.def (SHUFFLEVECTOR): New.
24424 * internal-fn.h (expand_SHUFFLEVECTOR): Declare.
24425 * doc/extend.texi: Document __builtin_shufflevector.
24427 2021-05-31 Peter Bergner <bergner@linux.ibm.com>
24430 * config/rs6000/predicates.md(mma_assemble_input_operand): Allow
24431 indexed form addresses.
24433 2021-05-29 Jeff Law <jlaw@tachyum.com>
24435 * config/h8300/h8300.c (h8300_emit_stack_adjustment): Drop unused
24436 parameter. Call callers fixed.
24438 (output_plussi): Add FALLTHRU markers.
24439 (h8300_shift_needs_scratch_p): Add gcc_unreachable marker.
24441 2021-05-29 Jakub Jelinek <jakub@redhat.com>
24443 PR middle-end/99928
24444 * gimplify.c (gimplify_scan_omp_clauses): For taskloop simd
24445 combined with parallel, make sure to add shared clause to
24446 parallel for explicit linear clause.
24448 2021-05-29 Aldy Hernandez <aldyh@redhat.com>
24450 PR tree-optimization/100787
24451 * gimple-ssa-evrp.c: Disable exporting of global ranges.
24453 2021-05-28 Jason Merrill <jason@redhat.com>
24455 * tree-iterator.h (struct tree_stmt_iterator): Add operator++,
24456 operator--, operator*, operator==, and operator!=.
24457 (class tsi_range): New.
24459 2021-05-28 Richard Biener <rguenther@suse.de>
24461 PR tree-optimization/100778
24462 * tree-vect-slp.c (vect_build_slp_tree_1): Prevent possibly
24463 trapping ops in different BBs.
24465 2021-05-28 Richard Biener <rguenther@suse.de>
24468 * tree-inline.c (copy_bb): When processing __builtin_va_arg_pack
24469 copy fntype from original call.
24471 2021-05-28 Martin Liska <mliska@suse.cz>
24473 PR gcov-profile/100751
24474 * doc/gcov.texi: Revert partially a hunk that was wrong.
24476 2021-05-28 Cooper Qu <cooper.qu@linux.alibaba.com>
24478 * config/csky/csky-linux-elf.h (HAVE_sync_compare_and_swapqi):
24480 (HAVE_sync_compare_and_swaphi): Likewise.
24481 (HAVE_sync_compare_and_swapsi): Likewise.
24483 2021-05-28 Jakub Jelinek <jakub@redhat.com>
24485 PR middle-end/99928
24486 * tree.h (OMP_CLAUSE_MAP_IMPLICIT): Define.
24488 2021-05-28 Tobias Burnus <tobias@codesourcery.com>
24490 * gimplify.c (gimplify_omp_affinity): New.
24491 (gimplify_scan_omp_clauses): Call it; remove affinity clause afterwards.
24492 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_AFFINITY.
24493 * tree-pretty-print.c (dump_omp_clause): Handle OMP_CLAUSE_AFFINITY.
24494 * tree.c (omp_clause_num_ops, omp_clause_code_name): Add clause.
24495 (walk_tree_1): Handle OMP_CLAUSE_AFFINITY.
24497 2021-05-28 Joern Rennecke <joern.rennecke@riscy-ip.com>
24498 Richard Biener <rguenther@suse.de>
24500 * match.pd <popcount & / + pattern matching>:
24501 When generating popcount directly fails, try doing it in two halves.
24503 2021-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
24505 * Makefile.in (generated_files): Add gimple-match.c and
24508 2021-05-28 Joern Rennecke <joern.rennecke@embecosm.com>
24510 * gensupport.c (alter_predicate_for_insn): Handle MATCH_DUP.
24512 2021-05-28 Joern Rennecke <joern.rennecke@embecosm.com>
24514 * gensupport.c (alter_constraints): Add MATCH_SCRATCH case.
24516 2021-05-28 Kewen Lin <linkw@linux.ibm.com>
24518 PR tree-optimization/99398
24519 * tree-ssa-forwprop.c (simplify_permutation): Optimize some cases
24520 where the fed operands are CTOR/CST and propagated through
24521 VIEW_CONVERT_EXPR. Call vec_perm_indices::new_shrunk_vector.
24522 * vec-perm-indices.c (vec_perm_indices::new_shrunk_vector): New
24524 * vec-perm-indices.h (vec_perm_indices::new_shrunk_vector): New
24527 2021-05-27 Uroš Bizjak <ubizjak@gmail.com>
24529 * config/i386/mmx.md (addv2sf3): Do not call
24530 ix86_fixup_binary_operands_no_copy.
24533 (<smaxmin:code>v2sf3): Ditto.
24534 (<plusminus:insn><MMXMODEI:mode>3): Ditto.
24535 (<plusminus:insn><VI_32:mode>3): Remove expander.
24536 (<plusminus:insn><VI_32:mode>3): Rename from
24537 "*<plusminus:insn><VI_32:mode>3".
24538 (mulv4hi): Do not call ix86_fixup_binary_operands_no_copy.
24539 (mulv2hi3): Remove expander.
24540 (mulv2hi3): Rename from *mulv2hi3.
24541 (<s>mulv2hi3_highpart): Remove expander.
24542 (<s>mulv2hi3_highpart): Rename from *<s>mulv2hi3_highpart.
24543 (<smaxmin:code><MMXMODE14:mode>3): Rename from
24544 "*mmx_<smaxmin:code><MMXMODE14:mode>3".
24545 (<smaxmin:code><SMAXMIN_MMXMODEI:mode>3): Remove expander.
24546 (SMAXMIN_MMXMODEI): Remove mode iterator.
24547 (<smaxmin:code>v4hi3): New expander.
24548 (<smaxmin:code>v4qi3): Rename from *<smaxmin:code>v4qi3.
24549 (<smaxmin:code>v2hi3): Rename from *<smaxmin:code>v2hi3.
24550 (<smaxmin:code><SMAXMIN_VI_32:mode>3): Remove expander.
24551 (SMAXMIN_VI_32): Remove mode iterator.
24552 (<umaxmin:code><MMXMODE24:mode>3): Rename from
24553 "*mmx_<umaxmin:code><MMXMODE24:mode>3".
24554 (<umaxmin:code><UMAXMIN_MMXMODEI:mode>3): Remove expander.
24555 (UMAXMIN_MMXMODEI): Remove mode iterator.
24556 (<umaxmin:code>v8qi3): New expander.
24557 (<umaxmin:code>v4qi3): Rename from *<umaxmin:code>v4qi3.
24558 (<umaxmin:code>v2hi3): Rename from *<umaxmin:code>v2hi3.
24559 (<umaxmin:code><SMAXMIN_VI_32:mode>3): Remove expander.
24560 (UMAXMIN_VI_32): Remove mode iterator.
24561 (<any_shift:insn>v2hi3): Remove expander.
24562 (<any_shift:insn>v2hi3): Rename from *<any_shift:insn>v2hi3.
24563 (<any_logic:code><MMXMODEI:mode>3): Do not call
24564 ix86_fixup_binary_operands_no_copy.
24565 (<any_logic:code><VI_32:mode>3): Remove expander.
24566 (<any_logic:code><VI_32:mode>3): Rename from
24567 "*<any_logic:code><VI_32:mode>3".
24568 (uavg<mode>3_ceil): Do not call ix86_fixup_binary_operands_no_copy.
24569 * config/i386/sse.md (div<VF2:mode>3): Do not call
24570 ix86_fixup_binary_operands_no_copy.
24571 (div<VF1:mode>3): Ditto.
24572 (<maxmin:code><VI8_AVX2_AVX512F:mode>3): Ditto.
24573 (smulhrsv4hi3): Ditto.
24574 (smulhrsv2hi3): Ditto.
24576 2021-05-27 Martin Sebor <msebor@redhat.com>
24578 * ggc.h (gt_ggc_mx): Add overloads for all integers.
24580 * hash-map.h (class hash_map): Add pch_nx_helper overloads for all
24582 (hash_map::operator==): New function.
24584 2021-05-27 Uroš Bizjak <ubizjak@gmail.com>
24587 * config/i386/i386-expand.c (ix86_expand_int_sse_cmp):
24588 For TARGET_XOP bypass SSE comparisons for all supported vector modes.
24589 * config/i386/mmx.md (*xop_maskcmp<MMXMODEI:mode>3): New insn pattern.
24590 (*xop_maskcmp<VI_32:mode>3): Ditto.
24591 (*xop_maskcmp_uns<MMXMODEI:mode>3): Ditto.
24592 (*xop_maskcmp_uns<VI_32:mode>3): Ditto.
24594 2021-05-27 Richard Earnshaw <rearnsha@arm.com>
24597 * config/arm/arm.c (arm_configure_build_target): Remove parameter
24598 opts_set, directly check opts parameters for being non-null.
24599 (arm_option_restore): Update call to arm_configure_build_target.
24600 (arm_option_override): Likewise.
24601 (arm_can_inline_p): Likewise.
24602 (arm_valid_target_attribute_tree): Likewise.
24603 * config/arm/arm-c.c (arm_pragma_target_parse): Likewise.
24604 * config/arm/arm-protos.h (arm_configure_build_target): Adjust
24607 2021-05-27 Aldy Hernandez <aldyh@redhat.com>
24609 * vr-values.c (simplify_conversion_using_ranges): Use
24610 get_range_query instead of get_global_range_query.
24612 2021-05-27 Aldy Hernandez <aldyh@redhat.com>
24614 * gimple-range.cc (get_range_global): Move to value-query.cc.
24615 (gimple_range_global): Same.
24616 (get_global_range_query): Same.
24617 (global_range_query::range_of_expr): Same.
24618 * gimple-range.h (class global_range_query): Move to
24620 (gimple_range_global): Same.
24621 * tree-ssanames.c (get_range_info): Move to value-query.cc.
24622 (get_ptr_nonnull): Same.
24623 * tree-ssanames.h (get_range_info): Remove.
24624 (get_ptr_nonnull): Remove.
24625 * value-query.cc (get_ssa_name_range_info): Move from
24627 (get_ssa_name_ptr_info_nonnull): Same.
24628 (get_range_global): Move from gimple-range.cc.
24629 (gimple_range_global): Same.
24630 (get_global_range_query): Same.
24631 (global_range_query::range_of_expr): Same.
24632 * value-query.h (class global_range_query): Move from
24634 (gimple_range_global): Same.
24636 2021-05-27 Uroš Bizjak <ubizjak@gmail.com>
24639 * config/i386/mmx.md (uavgv4qi3_ceil): New insn pattern.
24640 (uavgv2hi3_ceil): Ditto.
24642 2021-05-26 Eric Botcazou <ebotcazou@adacore.com>
24645 * doc/extend.texi (scalar_storage_order): Rephrase slightly.
24647 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
24649 * tree-ssanames.c (get_range_info): Merge both copies of
24650 get_range_info into one that works with irange.
24651 * tree-ssanames.h (get_range_info): Remove version that works on
24654 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
24656 * builtins.c (check_nul_terminated_array): Convert to get_range_query.
24657 (expand_builtin_strnlen): Same.
24658 (determine_block_size): Same.
24659 * fold-const.c (expr_not_equal_to): Same.
24660 * gimple-fold.c (size_must_be_zero_p): Same.
24661 * gimple-match-head.c: Include gimple-range.h.
24662 * gimple-pretty-print.c (dump_ssaname_info): Convert to get_range_query.
24663 * gimple-ssa-warn-restrict.c
24664 (builtin_memref::extend_offset_range): Same.
24665 * graphite-sese-to-poly.c (add_param_constraints): Same.
24666 * internal-fn.c (get_min_precision): Same.
24667 * ipa-fnsummary.c (set_switch_stmt_execution_predicate): Same.
24668 * ipa-prop.c (ipa_compute_jump_functions_for_edge): Same.
24670 * tree-data-ref.c (split_constant_offset): Same.
24671 (dr_step_indicator): Same.
24672 * tree-dfa.c (get_ref_base_and_extent): Same.
24673 * tree-scalar-evolution.c (iv_can_overflow_p): Same.
24674 * tree-ssa-loop-niter.c (refine_value_range_using_guard): Same.
24675 (determine_value_range): Same.
24676 (record_nonwrapping_iv): Same.
24677 (infer_loop_bounds_from_signedness): Same.
24678 (scev_var_range_cant_overflow): Same.
24679 * tree-ssa-phiopt.c (two_value_replacement): Same.
24680 * tree-ssa-pre.c (insert_into_preds_of_block): Same.
24681 * tree-ssa-reassoc.c (optimize_range_tests_to_bit_test): Same.
24682 * tree-ssa-strlen.c (handle_builtin_stxncpy_strncat): Same.
24684 (dump_strlen_info): Same.
24685 (set_strlen_range): Same.
24686 (maybe_diag_stxncpy_trunc): Same.
24687 (get_len_or_size): Same.
24688 (handle_integral_assign): Same.
24689 * tree-ssa-structalias.c (find_what_p_points_to): Same.
24690 * tree-ssa-uninit.c (find_var_cmp_const): Same.
24691 * tree-switch-conversion.c (bit_test_cluster::emit): Same.
24692 * tree-vect-patterns.c (vect_get_range_info): Same.
24693 (vect_recog_divmod_pattern): Same.
24694 * tree-vrp.c (intersect_range_with_nonzero_bits): Same.
24695 (register_edge_assert_for_2): Same.
24696 (determine_value_range_1): Same.
24697 * tree.c (get_range_pos_neg): Same.
24698 * vr-values.c (vr_values::get_lattice_entry): Same.
24699 (vr_values::update_value_range): Same.
24700 (simplify_conversion_using_ranges): Same.
24702 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
24704 * gimple-ssa-warn-alloca.c (alloca_call_type): Use
24705 get_range_query instead of query argument.
24706 (pass_walloca::execute): Enable and disable global ranger.
24708 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
24710 * gimple-ssa-evrp.c (rvrp_folder::rvrp_folder): Call
24712 (rvrp_folder::~rvrp_folder): Call disable_ranger.
24713 (hybrid_folder::hybrid_folder): Call enable_ranger.
24714 (hybrid_folder::~hybrid_folder): Call disable_ranger.
24716 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
24718 * function.c (allocate_struct_function): Set cfun->x_range_query.
24719 * function.h (struct function): Declare x_range_query.
24720 (get_range_query): New.
24721 (get_global_range_query): New.
24722 * gimple-range-cache.cc (ssa_global_cache::ssa_global_cache):
24723 Remove call to safe_grow_cleared.
24724 * gimple-range.cc (get_range_global): New.
24725 (gimple_range_global): Move from gimple-range.h.
24726 (get_global_range_query): New.
24727 (global_range_query::range_of_expr): New.
24728 (enable_ranger): New.
24729 (disable_ranger): New.
24730 * gimple-range.h (gimple_range_global): Move to gimple-range.cc.
24731 (class global_range_query): New.
24732 (enable_ranger): New.
24733 (disable_ranger): New.
24734 * gimple-ssa-evrp.c (evrp_folder::~evrp_folder): Rename
24735 dump_all_value_ranges to dump.
24736 * tree-vrp.c (vrp_prop::finalize): Same.
24737 * value-query.cc (range_query::dump): New.
24738 * value-query.h (range_query::dump): New.
24739 * vr-values.c (vr_values::dump_all_value_ranges): Rename to...
24740 (vr_values::dump): ...this.
24741 * vr-values.h (class vr_values): Rename dump_all_value_ranges to
24742 dump and make virtual.
24744 2021-05-26 Uroš Bizjak <ubizjak@gmail.com>
24746 * config/i386/i386.c (ix86_autovectorize_vector_modes):
24747 Add V4QImode and V16QImode for TARGET_SSE2.
24748 * doc/sourcebuild.texi (Vector-specific attributes):
24749 Add vect64 and vect32 description.
24751 2021-05-26 Bernd Edlinger <bernd.edlinger@hotmail.de>
24753 * gimple-range-gori.cc (range_def_chain::register_dependency):
24754 Resize m_def_chain when needed.
24756 2021-05-26 Christophe Lyon <christophe.lyon@linaro.org>
24758 * config/arm/mve.md (mve_vaddvq_<supf><mode>): Prefix with '@'.
24759 * config/arm/neon.md (reduc_plus_scal_<mode>): Move to ..
24760 * config/arm/vec-common.md: .. here. Add support for MVE.
24762 2021-05-26 Jakub Jelinek <jakub@redhat.com>
24764 * config/epiphany/epiphany.c (epiphany_print_operand_address): Remove
24766 * config/microblaze/microblaze.c (microblaze_legitimize_address,
24768 microblaze_option_override, print_operand): Likewise.
24769 * config/microblaze/microblaze.md (call_internal_plt,
24770 call_value_intern_plt, call_value_intern): Likewise.
24771 * config/arm/aout.h (ASM_OUTPUT_ALIGN): Likewise.
24772 * config/iq2000/iq2000.md (call_internal1, call_value_internal1,
24773 call_value_multiple_internal1): Likewise.
24774 * config/bfin/bfin.c (symbolic_reference_mentioned_p): Likewise.
24776 2021-05-26 Jan-Benedict Glaw <jbglaw@lug-owl.de>
24778 * config/arc/arc.c (arc_address_cost, arc_print_operand_address,
24779 arc_ccfsm_advance, symbolic_reference_mentioned_p,
24780 arc_raw_symbolic_reference_mentioned_p): Remove register
24783 2021-05-26 Jakub Jelinek <jakub@redhat.com>
24786 * omp-low.c: Include omp-offload.h.
24787 (create_omp_child_function): If current_function_decl has
24788 "omp declare target" attribute and is_gimple_omp_offloaded,
24789 remove that attribute from the copy of attribute list and
24790 add "omp target entrypoint" attribute instead.
24791 (lower_omp_target): Mark .omp_data_sizes.* and .omp_data_kinds.*
24792 variables for offloading if in omp_maybe_offloaded_ctx.
24793 * omp-offload.c (pass_omp_target_link::execute): Nullify second
24794 argument to GOMP_target_data_ext in offloaded code.
24796 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
24798 * config/csky/csky.c (csky_can_change_mode_class): Delete.
24799 For csky, HF/SF mode use the low bits of VREGS.
24801 2021-05-26 Eric Botcazou <ebotcazou@adacore.com>
24803 * gimplify.c (gimplify_decl_expr): Do not clear TREE_READONLY on a
24804 DECL which is a reference for OMP.
24806 2021-05-26 Martin Liska <mliska@suse.cz>
24808 PR gcov-profile/100751
24809 * doc/gcov.texi: Document that __gcov_dump can be called just
24810 once and that __gcov_reset resets run-time counters.
24812 2021-05-26 Martin Liska <mliska@suse.cz>
24814 * doc/install.texi: Port relevant part from install-old.texi
24815 and re-generate list of CPUs and systems.
24817 2021-05-26 Martin Liska <mliska@suse.cz>
24819 * Makefile.in: Remove it.
24820 * doc/include/fdl.texi: Update next/previous chapters.
24821 * doc/install.texi: Likewise.
24822 * doc/install-old.texi: Removed.
24824 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
24826 * config/csky/csky.c (ck810_legitimate_index_p): Support
24827 "base + index" with DF mode.
24828 * config/csky/constraints.md ("Y"): New constraint for memory operands
24829 without index register.
24830 * config/csky/csky_insn_fpuv2.md (fpuv3_movdf): Use "Y" instead of "m"
24831 when mov between memory and general registers, and lower their priority.
24832 * config/csky/csky_insn_fpuv3.md (fpuv2_movdf): Likewise.
24834 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
24836 * config/csky/csky.c (TARGET_PROMOTE_PROTOTYPES): Delete.
24838 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
24840 * config/csky/csky.md (untyped_call): Emit clobber for return
24841 registers to mark them used.
24843 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
24845 * config/csky/csky.md (cskyv2_sextend_ldbs): New.
24847 2021-05-26 Andrew Pinski <apinski@marvell.com>
24849 * match.pd (x < 0 ? ~y : y): New patterns.
24851 2021-05-26 Andrew Pinski <apinski@marvell.com>
24853 * match.pd (A?CST1:CST2): Add simplifcations for A?0:+-1, A?+-1:0,
24854 A?POW2:0 and A?0:POW2.
24856 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
24858 * gimple-range-gori.cc (class logical_stmt_cache): Delete
24859 (logical_stmt_cache::logical_stmt_cache ): Delete.
24860 (logical_stmt_cache::~logical_stmt_cache): Delete.
24861 (logical_stmt_cache::cache_entry::dump): Delete.
24862 (logical_stmt_cache::get_range): Delete.
24863 (logical_stmt_cache::cached_name ): Delete.
24864 (logical_stmt_cache::same_cached_name): Delete.
24865 (logical_stmt_cache::cacheable_p): Delete.
24866 (logical_stmt_cache::slot_diagnostics ): Delete.
24867 (logical_stmt_cache::dump): Delete.
24868 (gori_compute_cache::gori_compute_cache): Delete.
24869 (gori_compute_cache::~gori_compute_cache): Delete.
24870 (gori_compute_cache::compute_operand_range): Delete.
24871 (gori_compute_cache::cache_stmt): Delete.
24872 * gimple-range-gori.h (gori_compute::compute_operand_range): Remove
24874 (class gori_compute_cache): Delete.
24876 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
24878 * gimple-range.cc (fold_using_range::range_of_range_op): Use m_gori
24880 (fold_using_range::range_of_address): Adjust.
24881 (fold_using_range::range_of_phi): Adjust.
24882 * gimple-range.h (class fur_source): Adjust.
24883 (fur_source::fur_source): Adjust.
24885 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
24887 * gimple-range-gori.cc (gori_compute::expr_range_at_stmt): Rename
24888 from expr_range_in_bb and adjust.
24889 (gori_compute::compute_name_range_op): Adjust.
24890 (gori_compute::optimize_logical_operands): Adjust.
24891 (gori_compute::compute_logical_operands_in_chain): Adjust.
24892 (gori_compute::compute_operand1_range): Adjust.
24893 (gori_compute::compute_operand2_range): Adjust.
24894 (ori_compute_cache::cache_stmt): Adjust.
24895 * gimple-range-gori.h (gori_compute): Rename prototype.
24897 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
24899 * gimple-range.cc (gimple_ranger::range_of_expr): Non-null should be
24900 checked only after range_of_stmt, not range_on_entry.
24901 (gimple_ranger::range_on_entry): Check for non-null in any
24902 predecessor block, if it is not already non-null.
24903 (gimple_ranger::range_on_exit): DOnt check for non-null after
24904 range on entry call.
24905 (gimple_ranger::dump_bb): New. Split from dump.
24906 (gimple_ranger::dump): Adjust.
24907 * gimple-range.h (class gimple_ranger): Adjust.
24909 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
24911 * gimple-range-cache.cc (struct range_timestamp): Delete.
24912 (class temporal_cache): Adjust.
24913 (temporal_cache::get_timestamp): Delete.
24914 (temporal_cache::set_dependency): Delete.
24915 (temporal_cache::temporal_value): Adjust.
24916 (temporal_cache::current_p): Take dependencies as params.
24917 (temporal_cache::set_timestamp): Adjust.
24918 (temporal_cache::set_always_current): Adjust.
24919 (ranger_cache::get_non_stale_global_range): Adjust.
24920 (ranger_cache::register_dependency): Delete.
24921 * gimple-range-cache.h (class range_cache): Adjust.
24923 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
24925 * gimple-range-gori.cc (range_def_chain::range_def_chain): init
24927 (range_def_chain::~range_def_chain): Dispose of obstack rather than
24928 each individual bitmap.
24929 (range_def_chain::set_import): New.
24930 (range_def_chain::get_imports): New.
24931 (range_def_chain::chain_import_p): New.
24932 (range_def_chain::register_dependency): Rename from build_def_chain
24934 (range_def_chain::def_chain_in_bitmap_p): New.
24935 (range_def_chain::add_def_chain_to_bitmap): New.
24936 (range_def_chain::has_def_chain): Just check first depenedence.
24937 (range_def_chain::get_def_chain): Process imports, use generic
24938 register_dependency routine.
24939 (range_def_chain::dump): New.
24940 (gori_map::gori_map): Allocate import list.
24941 (gori_map::~gori_map): Release imports.
24942 (gori_map::exports): Check for past allocated block size.
24943 (gori_map::imports): New.
24944 (gori_map::def_chain_in_export_p): Delete.
24945 (gori_map::is_import_p): New.
24946 (gori_map::maybe_add_gori): Handle imports.
24947 (gori_map::dump): Adjust output, add imports.
24948 (gori_compute::has_edge_range_p): Remove def_chain_in_export call.
24949 (gori_export_iterator::gori_export_iterator): New.
24950 (gori_export_iterator::next): New.
24951 (gori_export_iterator::get_name): New.
24952 * gimple-range-gori.h (range_def_chain): Add imports and direct
24953 dependecies via struct rdc.
24954 (range_def_chain::depend1): New.
24955 (range_def_chain::depend2): New.
24956 (class gori_map): Adjust.
24957 (FOR_EACH_GORI_IMPORT_NAME): New.
24958 (FOR_EACH_GORI_EXPORT_NAME): New.
24959 (class gori_export_iterator): New.
24961 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
24963 * gimple-range-cache.cc (ranger_cache::ranger_cache): Move initial
24964 export cache filling to here.
24965 * gimple-range-gori.cc (gori_compute::gori_compute) : From Here.
24967 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
24969 * gimple-range-gori.cc (range_def_chain): Move to gimple-range-gori.h.
24970 (gori_map): Move to gimple-range-gori.h.
24971 (gori_compute::gori_compute): Adjust.
24972 (gori_compute::~gori_compute): Delete.
24973 (gori_compute::compute_operand_range_switch): Adjust.
24974 (gori_compute::compute_operand_range): Adjust.
24975 (gori_compute::compute_logical_operands): Adjust.
24976 (gori_compute::has_edge_range_p ): Adjust.
24977 (gori_compute::set_range_invariant): Delete.
24978 (gori_compute::dump): Adjust.
24979 (gori_compute::outgoing_edge_range_p): Adjust.
24980 * gimple-range-gori.h (class range_def_chain): Relocate here.
24981 (class gori_map): Relocate here.
24982 (class gori_compute): Inherit from gori_map, and adjust.
24984 2021-05-25 Aldy Hernandez <aldyh@redhat.com>
24986 * value-range.cc (range_tests_legacy): Use
24987 build_nonstandard_integer_type instead of int and short.
24989 2021-05-25 Eric Botcazou <ebotcazou@adacore.com>
24991 * gimplify.c (gimplify_decl_expr): Clear TREE_READONLY on the DECL
24992 when really creating an initialization statement for it.
24994 2021-05-25 Eric Botcazou <ebotcazou@adacore.com>
24996 * tree-inline.c (setup_one_parameter): Fix thinko in new condition.
24998 2021-05-25 Kito Cheng <kito.cheng@sifive.com>
25000 * config/riscv/riscv.h (ASM_SPEC): Pass -mno-relax.
25002 2021-05-25 Martin Liska <mliska@suse.cz>
25004 PR tree-optimization/92860
25006 * optc-save-gen.awk: Remove exceptions.
25008 2021-05-25 Martin Liska <mliska@suse.cz>
25010 * asan.h (sanitize_coverage_p): New function.
25011 * doc/extend.texi: Document it.
25012 * fold-const.c (fold_range_test): Use sanitize_flags_p
25013 instead of flag_sanitize_coverage.
25014 (fold_truth_andor): Likewise.
25015 * sancov.c: Likewise.
25016 * tree-ssa-ifcombine.c (ifcombine_ifandif): Likewise.
25017 * ipa-inline.c (sanitize_attrs_match_for_inline_p): Handle
25018 -fsanitize-coverage when inlining.
25020 2021-05-25 Cooper Qu <cooper.qu@linux.alibaba.com>
25022 * config/csky/csky-modes.def : Fix copyright.
25024 2021-05-25 Cooper Qu <cooper.qu@linux.alibaba.com>
25026 * config/csky/csky-modes.def : Amend copyright.
25027 * config/csky/csky_insn_fpuv2.md : Likewise.
25028 * config/csky/csky_insn_fpuv3.md : Likewise.
25030 2021-05-25 Richard Biener <rguenther@suse.de>
25032 PR middle-end/100727
25033 * calls.c (initialize_argument_information): Explicitely test
25034 for WITH_SIZE_EXPR.
25035 * gimple-expr.c (mark_addressable): Skip outer WITH_SIZE_EXPR.
25037 2021-05-25 Geng Qi <gengqi@linux.alibaba.com>
25039 * config/csky/csky.h (FRAME_POINTER_REGNUM): Use
25040 HARD_FRAME_POINTER_REGNUM and FRAME_POINTER_REGNUM instead of
25041 the signle definition. The signle definition may not work well
25042 at simplify_subreg_regno().
25043 (HARD_FRAME_POINTER_REGNUM): New.
25044 (ELIMINABLE_REGS): Add for HARD_FRAME_POINTER_REGNUM.
25045 * config/csky/csky.c (get_csky_live_regs, csky_can_eliminate,
25046 csky_initial_elimination_offset, csky_expand_prologue,
25047 csky_expand_epilogue): Add for HARD_FRAME_POINTER_REGNUM.
25049 2021-05-25 Geng Qi <gengqi@linux.alibaba.com>
25051 * config/csky/csky.c (csky_option_override):
25052 Init csky_arch_isa_features[] in advance, so TARGET_DSP
25053 and TARGET_DIV can be set well.
25055 2021-05-25 Geng Qi <gengqi@linux.alibaba.com>
25057 * config/csky/constraints.md ("l", "h"): Delete.
25058 * config/csky/csky.h (reg_class, REG_CLASS_NAMES,
25059 REG_CLASS_CONTENTS): Delete LO_REGS and HI_REGS.
25060 * config/csky/csky.c (regno_reg_classm,
25061 csky_secondary_reload, csky_register_move_cost):
25062 Use HILO_REGS instead of LO_REGS and HI_REGS.
25064 2021-05-25 Geng Qi <gengqi@linux.alibaba.com>
25066 * config/csky/constraints.md ("W"): New constriant for mem operand
25067 with base reg, index register.
25068 ("Q"): Renamed and modified "csky_valid_fpuv2_mem_operand" to
25069 "csky_valid_mem_constraint_operand" to deal with both "Q" and "W"
25071 ("Dv"): New constraint for const double value that can be used at
25073 * config/csky/csky-modes.def (HFmode): New mode.
25074 * config/csky/csky-protos.h (csky_valid_fpuv2_mem_operand): Rename
25075 to "csky_valid_mem_constraint_operand" and support new constraint
25077 (csky_get_movedouble_length): New.
25078 (fpuv3_output_move): New.
25079 (fpuv3_const_double): New.
25080 * config/csky/csky.c (csky_option_override): New arch CK860 with fpv3.
25081 (decompose_csky_address): Refine.
25082 (csky_print_operand): New "CONST_DOUBLE" operand.
25083 (csky_output_move): Support fpv3 instructions.
25084 (csky_get_movedouble_length): New.
25085 (fpuv3_output_move): New.
25086 (fpuv3_const_double): New.
25087 (csky_emit_compare): Cover float comparsion.
25088 (csky_emit_compare_float): Refine.
25089 (csky_vaild_fpuv2_mem_operand): Rename to
25090 "csky_valid_mem_constraint_operand" and support new constraint "W".
25091 (ck860_rtx_costs): New.
25092 (csky_rtx_costs): Add the cost calculation of CK860.
25093 (regno_reg_class): New vregs for fpuv3.
25094 (csky_dbx_regno): Likewise.
25095 (csky_cpu_cpp_builtins): New builtin macro for fpuv3.
25096 (csky_conditional_register_usage): Suporrot fpuv3.
25097 (csky_dwarf_register_span): Suporrot fpuv3.
25098 (csky_init_builtins, csky_mangle_type): Support "__fp16" type.
25099 (ck810_legitimate_index_p): Support fp16.
25100 * config/csky/csky.h (TARGET_TLS): ADD CK860.
25101 (CSKY_VREG_P, CSKY_VREG_LO_P, CSKY_VREG_HI_P): Support fpuv3.
25102 (TARGET_SINGLE_FPU): Support fpuv3.
25103 (TARGET_SUPPORT_FPV3): New.
25104 (FIRST_PSEUDO_REGISTER): Change to 202 to hold the new fpuv3 registers.
25105 (FIXED_REGISTERS, CALL_REALLY_USED_REGISTERS, REGISTER_NAMES,
25106 REG_CLASS_CONTENTS): Support fpuv3.
25107 * config/csky/csky.md (movsf): Move to cksy_insn_fpu.md and refine.
25108 (csky_movsf_fpv2): Likewise.
25109 (ck801_movsf): Likewise.
25110 (csky_movsf): Likewise.
25112 (csky_movdf_fpv2): Likewise.
25113 (ck801_movdf): Likewise.
25114 (csky_movdf): Likewise.
25115 (movsicc): Refine. Use "comparison_operatior" instead of
25116 "ordered_comparison_operatior".
25117 (addsicc): Likewise.
25118 (CSKY_FIRST_VFP3_REGNUM, CSKY_LAST_VFP3_REGNUM): New constant.
25119 (call_value_internal_vh): New.
25120 * config/csky/csky_cores.def (CK860): New arch and cpu.
25125 * config/csky/csky_insn_fpu.md: Refactor. Separate all float patterns
25126 into emit-patterns and match-patterns, remain the emit-patterns here,
25127 and move the match-patterns to csky_insn_fpuv2.md or
25128 csky_insn_fpuv3.md.
25129 * config/csky/csky_insn_fpuv2.md: New file for fpuv2 instructions.
25130 * config/csky/csky_insn_fpuv3.md: New file and new patterns for fpuv3
25132 * config/csky/csky_isa.def (fcr): New.
25137 (CK860): New definition for ck860.
25138 * config/csky/csky_tables.opt (ck860): New processors ck860,
25139 ck860f. And new arch ck860.
25144 * config/csky/predicates.md (csky_float_comparsion_operator): Delete
25145 "geu", "gtu", "leu", "ltu", which will never appear at float comparison.
25146 * config/csky/t-csky-elf: Support 860.
25147 * config/csky/t-csky-linux: Likewise.
25148 * doc/md.texi: Add "Q" and "W" constraints for C-SKY.
25150 2021-05-24 Aaron Sawdey <acsawdey@linux.ibm.com>
25152 * config/rs6000/genfusion.pl (gen_logical_addsubf): Refactor to
25153 add generation of logical-add and add-logical fusion pairs.
25154 * config/rs6000/rs6000-cpus.def: Add new fusion to ISA 3.1 mask
25156 * config/rs6000/rs6000.c (rs6000_option_override_internal): Turn on
25157 logical-add and add-logical fusion by default.
25158 * config/rs6000/rs6000.opt: Add -mpower10-fusion-logical-add and
25159 -mpower10-fusion-add-logical options.
25160 * config/rs6000/fusion.md: Regenerate file.
25162 2021-05-24 Aldy Hernandez <aldyh@redhat.com>
25164 * value-range.cc (irange::legacy_equal_p): Check type when
25165 comparing VR_VARYING types.
25166 (range_tests_legacy): Test comparing VARYING ranges of different
25169 2021-05-24 Wilco Dijkstra <wdijkstr@arm.com>
25171 * config/aarch64/aarch64.c (neoversen1_tunings):
25172 Enable AARCH64_EXTRA_TUNE_CHEAP_SHIFT_EXTEND.
25174 2021-05-24 Wilco Dijkstra <wdijkstr@arm.com>
25176 * config/aarch64/aarch64.c (aarch64_classify_symbol): Use GOT for
25177 extern weak symbols. Limit symbol offsets for non-GOT symbols with
25180 2021-05-24 Christophe Lyon <christophe.lyon@linaro.org>
25182 * config/arm/neon.md (vec_load_lanesxi<mode>)
25183 (vec_store_lanexoi<mode>): Move ...
25184 * config/arm/vec-common.md: here.
25186 2021-05-24 Christophe Lyon <christophe.lyon@linaro.org>
25188 * config/arm/neon.md (vec_load_lanesoi<mode>)
25189 (vec_store_lanesoi<mode>): Move ...
25190 * config/arm/vec-common.md: here.
25192 2021-05-24 liuhongt <hongtao.liu@intel.com>
25195 * config/i386/i386.c (ix86_gimple_fold_builtin): Replacing
25196 stmt with GIMPLE_NOP when lhs doesn't exist.
25198 2021-05-23 Uroš Bizjak <ubizjak@gmail.com>
25201 * config/i386/mmx.md (*push<VI_32:mode>2_rex64):
25202 New instruction pattern.
25203 (*push<VI_32:mode>2): Ditto.
25204 (push splitter for SSE registers): New splitter.
25206 2021-05-23 Andrew Pinski <apinski@marvell.com>
25208 * match.pd ((A & C) != 0 ? D : 0): Limit to non pointer types.
25210 2021-05-22 Aaron Sawdey <acsawdey@linux.ibm.com>
25212 * config/rs6000/genfusion.pl (gen_addadd): Fix incorrect attr types.
25213 * config/rs6000/fusion.md: Regenerate file.
25215 2021-05-21 Aaron Sawdey <acsawdey@linux.ibm.com>
25217 * config/rs6000/genfusion.pl (gen_addadd): New function.
25218 * config/rs6000/fusion.md: Regenerate file.
25219 * config/rs6000/rs6000-cpus.def: Add
25220 OPTION_MASK_P10_FUSION_2ADD to masks.
25221 * config/rs6000/rs6000.c (rs6000_option_override_internal):
25222 Handle default value of OPTION_MASK_P10_FUSION_2ADD.
25223 * config/rs6000/rs6000.opt: Add -mpower10-fusion-2add.
25225 2021-05-21 Jakub Jelinek <jakub@redhat.com>
25227 PR middle-end/99928
25228 * tree.h (OMP_CLAUSE_FIRSTPRIVATE_IMPLICIT_TARGET): Define.
25229 * gimplify.c (enum gimplify_omp_var_data): Fix up
25230 GOVD_MAP_HAS_ATTACHMENTS value, add GOVD_FIRSTPRIVATE_IMPLICIT.
25231 (omp_lastprivate_for_combined_outer_constructs): If combined target
25232 has GOVD_FIRSTPRIVATE_IMPLICIT set for the decl, change it to
25233 GOVD_MAP | GOVD_SEEN.
25234 (gimplify_scan_omp_clauses): Set GOVD_FIRSTPRIVATE_IMPLICIT for
25235 firstprivate clauses with OMP_CLAUSE_FIRSTPRIVATE_IMPLICIT.
25236 (gimplify_adjust_omp_clauses): For firstprivate clauses with
25237 OMP_CLAUSE_FIRSTPRIVATE_IMPLICIT either clear that bit and
25238 OMP_CLAUSE_FIRSTPRIVATE_IMPLICIT_TARGET too, or remove it and
25239 let it be replaced by implicit map clause.
25241 2021-05-21 Jakub Jelinek <jakub@redhat.com>
25243 PR middle-end/99928
25244 * gimplify.c (omp_lastprivate_for_combined_outer_constructs): New
25246 (gimplify_scan_omp_clauses) <case OMP_CLAUSE_LASTPRIVATE>: Use it.
25247 (gimplify_omp_for): Likewise.
25249 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
25251 PR middle-end/90115
25252 * omp-low.c (oacc_privatization_candidate_p): Reject 'static',
25253 'external' in blocks.
25255 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
25257 PR middle-end/90115
25258 * flag-types.h (enum openacc_privatization): New.
25259 * params.opt (-param=openacc-privatization): New.
25260 * doc/invoke.texi (openacc-privatization): Document it.
25261 * omp-general.h (get_openacc_privatization_dump_flags): New
25263 * omp-low.c (oacc_privatization_candidate_p): Add diagnostics.
25264 * omp-offload.c (execute_oacc_device_lower)
25265 <IFN_UNIQUE_OACC_PRIVATE>: Re-work diagnostics.
25266 * target.def (goacc.adjust_private_decl): Add 'location_t'
25268 * doc/tm.texi: Regenerate.
25269 * config/gcn/gcn-protos.h (gcn_goacc_adjust_private_decl): Adjust.
25270 * config/gcn/gcn-tree.c (gcn_goacc_adjust_private_decl): Likewise.
25271 * config/nvptx/nvptx.c (nvptx_goacc_adjust_private_decl):
25272 Likewise. Preserve it for...
25273 (nvptx_goacc_expand_var_decl): ... use here.
25275 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
25277 * doc/sourcebuild.texi (Other attributes): Document '__OPTIMIZE__'
25280 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
25282 PR middle-end/90115
25283 * omp-low.c (oacc_privatization_candidate_p): New function.
25284 (oacc_privatization_scan_clause_chain)
25285 (oacc_privatization_scan_decl_chain): Use it. Also
25286 'gcc_checking_assert' that we're not seeing duplicates.
25288 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
25290 PR middle-end/90115
25291 * omp-offload.c (execute_oacc_device_lower): Skip processing if no
25294 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
25296 PR middle-end/90115
25297 * omp-offload.c (execute_oacc_device_lower): Explain.
25299 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
25301 PR middle-end/90115
25302 * omp-offload.c (execute_oacc_device_lower)
25303 <IFN_UNIQUE_OACC_PRIVATE>: Diagnose and handle for 'level == -1'
25305 * internal-fn.c (expand_UNIQUE): Don't expect
25306 'IFN_UNIQUE_OACC_PRIVATE'.
25308 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
25310 PR middle-end/90115
25311 * omp-low.c (lower_omp_for): Don't evaluate OpenMP 'for' clauses.
25313 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
25315 PR middle-end/90115
25316 * config/nvptx/nvptx.c (nvptx_goacc_adjust_private_decl)
25317 (nvptx_goacc_expand_var_decl): Tighten.
25319 2021-05-21 Julian Brown <julian@codesourcery.com>
25320 Chung-Lin Tang <cltang@codesourcery.com>
25321 Thomas Schwinge <thomas@codesourcery.com>
25323 PR middle-end/90115
25324 * doc/tm.texi.in (TARGET_GOACC_EXPAND_VAR_DECL)
25325 (TARGET_GOACC_ADJUST_PRIVATE_DECL): Add documentation hooks.
25326 * doc/tm.texi: Regenerate.
25327 * expr.c (expand_expr_real_1): Expand decls using the
25328 expand_var_decl OpenACC hook if defined.
25329 * internal-fn.c (expand_UNIQUE): Handle IFN_UNIQUE_OACC_PRIVATE.
25330 * internal-fn.h (IFN_UNIQUE_CODES): Add OACC_PRIVATE.
25331 * omp-low.c (omp_context): Add oacc_privatization_candidates
25333 (lower_oacc_reductions): Add PRIVATE_MARKER parameter. Insert
25335 (lower_oacc_head_tail): Add PRIVATE_MARKER parameter. Modify
25336 private marker's gimple call arguments, and pass it to
25337 lower_oacc_reductions.
25338 (oacc_privatization_scan_clause_chain)
25339 (oacc_privatization_scan_decl_chain, lower_oacc_private_marker):
25341 (lower_omp_for, lower_omp_target, lower_omp_1): Use these.
25342 * omp-offload.c (convert.h): Include.
25343 (oacc_loop_xform_head_tail): Treat private-variable markers like
25344 fork/join when transforming head/tail sequences.
25345 (struct var_decl_rewrite_info): Add struct.
25346 (oacc_rewrite_var_decl, is_sync_builtin_call): New functions.
25347 (execute_oacc_device_lower): Support rewriting gang-private
25348 variables using target hook, and fix up addr_expr and var_decl
25350 * target.def (adjust_private_decl, expand_var_decl): New hooks.
25351 * config/gcn/gcn-protos.h (gcn_goacc_adjust_gangprivate_decl):
25353 (gcn_goacc_adjust_private_decl): ...this.
25354 * config/gcn/gcn-tree.c (gcn_goacc_adjust_gangprivate_decl):
25356 (gcn_goacc_adjust_private_decl): ...this. Add LEVEL parameter.
25357 * config/gcn/gcn.c (TARGET_GOACC_ADJUST_GANGPRIVATE_DECL): Rename
25358 definition using gcn_goacc_adjust_gangprivate_decl...
25359 (TARGET_GOACC_ADJUST_PRIVATE_DECL): ...to this, using
25360 gcn_goacc_adjust_private_decl.
25361 * config/nvptx/nvptx.c (tree-pretty-print.h): Include.
25362 (gang_private_shared_size): New global variable.
25363 (gang_private_shared_align): Likewise.
25364 (gang_private_shared_sym): Likewise.
25365 (gang_private_shared_hmap): Likewise.
25366 (nvptx_option_override): Initialize these.
25367 (nvptx_file_end): Output gang_private_shared_sym.
25368 (nvptx_goacc_adjust_private_decl, nvptx_goacc_expand_var_decl):
25370 (nvptx_set_current_function): Clear gang_private_shared_hmap.
25371 (TARGET_GOACC_ADJUST_PRIVATE_DECL): Define hook.
25372 (TARGET_GOACC_EXPAND_VAR_DECL): Likewise.
25374 2021-05-21 H.J. Lu <hjl.tools@gmail.com>
25376 * config/i386/i386-modes.def (MAX_BITSIZE_MODE_ANY_INT): Removed.
25378 2021-05-21 Richard Biener <rguenther@suse.de>
25379 H.J. Lu <hjl.tools@gmail.com>
25381 PR middle-end/90773
25382 * expr.c (expand_constructor): Elide expand_constructor if
25383 move by pieces is preferred.
25385 2021-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
25387 * config/aarch64/aarch64-builtins.c (aarch64_call_properties):
25388 Take a flag and mode value as arguments.
25389 (aarch64_modifies_global_state_p): Likewise.
25390 (aarch64_reads_global_state_p): Likewise.
25391 (aarch64_could_trap_p): Likewise.
25392 (aarch64_get_attributes): Likewise.
25393 (aarch64_init_simd_builtins): Adjust callsite of above.
25394 (aarch64_init_fcmla_laneq_builtins): Use aarch64_get_attributes to get
25395 function attributes to apply to builtins.
25396 (aarch64_init_crc32_builtins): Likewise.
25397 (aarch64_init_builtin_rsqrt): Likewise.
25399 2021-05-21 Aaron Sawdey <acsawdey@linux.ibm.com>
25401 * config/rs6000/rs6000.md (define_attr "type"): Add types for fusion.
25402 * config/rs6000/genfusion.pl (gen_ld_cmpi_p10): Use new fusion types.
25403 (gen_2logical): Use new fusion types.
25404 * config/rs6000/fusion.md: Regenerate.
25406 2021-05-21 Uroš Bizjak <ubizjak@gmail.com>
25409 * config/i386/i386-expand.c (ix86_expand_sse_movcc):
25410 Handle V4QI and V2HI modes.
25411 (ix86_expand_sse_movcc): Ditto.
25412 * config/i386/mmx.md (*<sat_plusminus:insn><VI_32:mode>3):
25413 New instruction pattern.
25414 (*eq<VI_32:mode>3): Ditto.
25415 (*gt<VI_32:mode>3): Ditto.
25416 (*xop_pcmov_<VI_32:mode>): Ditto.
25417 (mmx_pblendvb32): Ditto.
25418 (mmx_pblendvb64): Rename from mmx_pblendvb.
25419 (vec_cmp<VI_32:mode><VI_32:mode>): New expander.
25420 (vec_cmpu<VI_32:mode><VI_32:mode>): Ditto.
25421 (vcond<VI_32:mode><VI_32:mode>): Ditto.
25422 (vcondu<VI_32:mode><VI_32:mode>): Ditto.
25423 (vcond_mask_<VI_32:mode><VI_32:mode>): Ditto.
25425 2021-05-21 Jakub Jelinek <jakub@redhat.com>
25427 PR tree-optimization/94589
25428 * tree-ssa-phiopt.c (spaceship_replacement): For integral rhs1 and
25429 rhs2, treat x <= 4 equivalently to x < 5 etc. In cmp1 and cmp2 (if
25430 not the same as cmp3) treat <= the same as < and >= the same as >.
25431 Don't require that cond2_phi_edge is true edge, instead take
25432 false/true edges into account based on cmp1/cmp2 comparison kinds.
25434 2021-05-21 Uroš Bizjak <ubizjak@gmail.com>
25437 * config/i386/mmx.md (SMAXMIN_MMXMODEI): New mode iterator.
25438 (<smaxmin:code><SMAXMIN_MMXMODEI:mode>3): Macroize expander
25439 from <smaxmin:code>v4hi3> and <smaxmin:code><MMXMODE14:mode>3
25440 using SMAXMIN_MMXMODEI mode iterator.
25441 (*<smaxmin:code>v4qi3): New insn pattern.
25442 (*<smaxmin:code>v2hi3): Ditto.
25443 (SMAXMIN_VI_32): New mode iterator.
25444 (<smaxmin:code><SMAXMIN_VI_32>mode3): New expander.
25445 (UMAXMIN_MMXMODEI): New mode iterator.
25446 (<umaxmin:code><UMAXMIN_MMXMODEI:mode>3): Macroize expander
25447 from <umaxmin:code>v8qi3> and <umaxmin:code><MMXMODE24:mode>3
25448 using UMAXMIN_MMXMODEI mode iterator.
25449 (*<umaxmin:code>v4qi3): New insn pattern.
25450 (*<umaxmin:code>v2hi3): Ditto.
25451 (UMAXMIN_VI_32): New mode iterator.
25452 (<umaxmin:code><UMAXMIN_VI_32>mode3): New expander.
25453 (abs<VI_32:mode>2): New insn pattern.
25454 (ssse3_abs<MMXMODEI:mode>2, abs<MMXMODEI:mode>2): Move from ...
25455 * config/i386/sse.md: ... here.
25457 2021-05-20 Clement Chigot <clement.chigot@atos.net>
25458 David Edelsohn <dje.gcc@gmail.com>
25460 * collect2.c (scan_prog_file): Issue non-fatal warning for
25463 2021-05-20 Jonathan Wakely <jwakely@redhat.com>
25465 * doc/invoke.texi (-Wno-c++11-extensions)
25466 (-Wno-c++14-extensions, -Wno-c++17-extensions)
25467 (-Wno-c++20-extensions, -Wno-c++23-extensions): Document
25470 2021-05-20 Indu Bhagat <indu.bhagat@oracle.com>
25472 * config/c6x/c6x.c (c6x_output_file_unwind): Use dwarf_debuginfo_p.
25473 * config/darwin.c (darwin_override_options): Likewise.
25474 * config/i386/cygming.h (DBX_REGISTER_NUMBER): Likewise.
25475 * config/i386/darwin.h (DBX_REGISTER_NUMBER): Likewise.
25476 (DWARF2_FRAME_REG_OUT): Likewise.
25477 * config/mips/mips.c (mips_output_filename): Likewise.
25478 * config/rs6000/rs6000.c (rs6000_xcoff_declare_function_name):
25480 (rs6000_dbx_register_number): Likewise.
25481 * dbxout.c: Include flags.h.
25482 * dwarf2cfi.c (cfi_label_required_p): Likewise.
25483 (dwarf2out_do_frame): Likewise.
25484 * except.c: Include flags.h.
25485 * final.c (dwarf2_debug_info_emitted_p): Likewise.
25486 (final_scan_insn_1): Likewise.
25487 * flags.h (dwarf_debuginfo_p): New function declaration.
25488 * opts.c (dwarf_debuginfo_p): New function definition.
25489 * targhooks.c (default_debug_unwind_info): Use dwarf_debuginfo_p.
25490 * toplev.c (process_options): Likewise.
25492 2021-05-20 Indu Bhagat <indu.bhagat@oracle.com>
25494 * common.opt: Change type to support bitmasks.
25495 * flag-types.h (enum debug_info_type): Rename enumerator constants.
25496 (NO_DEBUG): New bitmask.
25497 (DBX_DEBUG): Likewise.
25498 (DWARF2_DEBUG): Likewise.
25499 (XCOFF_DEBUG): Likewise.
25500 (VMS_DEBUG): Likewise.
25501 (VMS_AND_DWARF2_DEBUG): Likewise.
25502 * flags.h (debug_set_to_format): New function declaration.
25503 (debug_set_count): Likewise.
25504 (debug_set_names): Likewise.
25505 * opts.c (debug_type_masks): Array of bitmasks for debug formats.
25506 (debug_set_to_format): New function definition.
25507 (debug_set_count): Likewise.
25508 (debug_set_names): Likewise.
25509 (set_debug_level): Update access to debug_type_names.
25510 * toplev.c: Likewise.
25512 2021-05-20 Martin Sebor <msebor@redhat.com>
25514 PR middle-end/100684
25515 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Handle C++ lambda.
25517 2021-05-20 Uroš Bizjak <ubizjak@gmail.com>
25520 * config/i386/i386.md (isa): Remove x64_bmi.
25521 (enabled): Remove x64_bmi.
25522 * config/i386/mmx.md (mmx_andnot<MMXMODEI:mode>3):
25523 Remove general register alternative.
25524 (*andnot<VI_32:mode>3): Ditto.
25525 (*mmx_<any_logic:code><MMXMODEI:mode>3): Ditto.
25526 (*<any_logic:code><VI_32:mode>3): Ditto.
25528 2021-05-20 Kewen Lin <linkw@linux.ibm.com>
25530 * config/arm/arm.c: Include head files tree-vectorizer.h and
25533 2021-05-20 Uroš Bizjak <ubizjak@gmail.com>
25536 * config/i386/mmx.md (Yv_Yw): Revert adding V4QI and V2HI modes.
25537 (*<plusminus:insn><VI_32:mode>3): Use Yw instad of <Yv_Yw> constrint.
25538 (<s>mulv4hi3_highpart): New expander.
25539 (*<s>mulv2hi3_highpart): New insn pattern.
25540 (<s>mulv2hi3_higpart): New expander.
25541 (*<any_shift:insn>v2hi3): New insn pattern.
25542 (<any_shift:insn>v2hi3): New expander.
25543 * config/i386/sse.md (smulhrsv2hi3): New expander.
25544 (*smulhrsv2hi3): New insn pattern.
25546 2021-05-20 Kewen Lin <linkw@linux.ibm.com>
25548 * doc/invoke.texi (vect-inner-loop-cost-factor): Document new
25550 * params.opt (vect-inner-loop-cost-factor): New.
25551 * targhooks.c (default_add_stmt_cost): Replace hardcoded factor
25552 50 with LOOP_VINFO_INNER_LOOP_COST_FACTOR, include head file
25553 tree-vectorizer.h and its required ones.
25554 * config/aarch64/aarch64.c (aarch64_add_stmt_cost): Replace
25555 hardcoded factor 50 with LOOP_VINFO_INNER_LOOP_COST_FACTOR.
25556 * config/arm/arm.c (arm_add_stmt_cost): Likewise.
25557 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
25558 * config/rs6000/rs6000.c (rs6000_add_stmt_cost): Likewise.
25559 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
25561 (_loop_vec_info::_loop_vec_info): Init inner_loop_cost_factor.
25562 * tree-vectorizer.h (_loop_vec_info): Add inner_loop_cost_factor.
25563 (LOOP_VINFO_INNER_LOOP_COST_FACTOR): New macro.
25565 2021-05-20 Christophe Lyon <christophe.lyon@linaro.org>
25566 Torbjörn Svensson <torbjorn.svensson@st.com>
25569 * doc/cpp.texi (Common Predefined Macros): Document __FILE_NAME__.
25571 2021-05-20 Jakub Jelinek <jakub@redhat.com>
25573 PR middle-end/99928
25574 * gimplify.c (gimplify_scan_omp_clauses) <case OMP_CLAUSE_LINEAR>: For
25575 explicit linear clause when combined with target, make it map(tofrom:)
25576 instead of no clause or firstprivate.
25578 2021-05-20 Jakub Jelinek <jakub@redhat.com>
25580 PR tree-optimization/94589
25581 * match.pd ((X & Y) == X -> (X & ~Y) == 0): Simplify even in presence
25582 of integral conversions.
25584 2021-05-19 Andrew MacLeod <amacleod@redhat.com>
25586 * gimple-range.cc (fur_source::get_operand): New.
25587 (gimple_range_fold): Delete.
25588 (fold_using_range::fold_stmt): Move from gimple_ranger::calc_stmt.
25589 (fold_using_range::range_of_range_op): Move from gimple_ranger.
25590 (fold_using_range::range_of_address): Ditto.
25591 (fold_using_range::range_of_phi): Ditto.
25592 (fold_using_range::range_of_call): Ditto.
25593 (fold_using_range::range_of_builtin_ubsan_call): Move from
25594 range_of_builtin_ubsan_call.
25595 (fold_using_range::range_of_builtin_call): Move from
25596 range_of_builtin_call.
25597 (gimple_ranger::range_of_builtin_call): Delete.
25598 (fold_using_range::range_of_cond_expr): Move from gimple_ranger.
25599 (gimple_ranger::fold_range_internal): New.
25600 (gimple_ranger::range_of_stmt): Use new fold_using_range API.
25601 (fold_using_range::range_of_ssa_name_with_loop_info): Move from
25602 gimple_ranger. Improve ranges of SSA_NAMES when possible.
25603 * gimple-range.h (gimple_ranger): Remove various range_of routines.
25604 (class fur_source): New.
25605 (class fold_using_range): New.
25606 (fur_source::fur_source): New.
25608 * vr-values.c (vr_values::extract_range_basic): Use fold_using_range
25609 instead of range_of_builtin_call.
25611 2021-05-19 Jonathan Wakely <jwakely@redhat.com>
25613 * doc/cpp.texi (Common Predefined Macros): Update documentation
25614 for the __GXX_EXPERIMENTAL_CXX0X__ macro.
25616 2021-05-19 Alex Coplan <alex.coplan@arm.com>
25619 * config/arm/arm.md (nonsecure_call_internal): Always ensure
25620 callee's address is in a register.
25622 2021-05-19 Geng Qi <gengqi@linux.alibaba.com>
25624 * common/config/riscv/riscv-common.c
25625 (riscv_subset_list::parsing_subset_version): Properly parse the letter
25627 (riscv_subset_list::parse_std_ext,
25628 riscv_subset_list::parse_multiletter_ext): To handle errors generated
25629 in riscv_subset_list::parsing_subset_version.
25631 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
25633 * config/aarch64/aarch64-simd.md: Use "neon_move_narrow_q"
25634 type attribute in patterns generating XTN(2).
25636 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
25638 * config/aarch64/aarch64-simd.md (aarch64_simd_vec_pack_trunc_<mode>):
25639 Remove as duplicate of...
25640 (aarch64_xtn<mode>): This.
25641 (aarch64_xtn2<mode>_le): Move position in file.
25642 (aarch64_xtn2<mode>_be): Move position in file.
25643 (aarch64_xtn2<mode>): Move position in file.
25644 (vec_pack_trunc_<mode>): Define as an expander.
25646 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
25648 * config/aarch64/aarch64-simd-builtins.def: Split builtin
25649 generation for aarch64_<sur>q<r>shr<u>n_n<mode> pattern into
25650 separate scalar and vector generators.
25651 * config/aarch64/aarch64-simd.md
25652 (aarch64_<sur>q<r>shr<u>n_n<mode>): Define as an expander and
25654 (aarch64_<sur>q<r>shr<u>n_n<mode>_insn_le): This and...
25655 (aarch64_<sur>q<r>shr<u>n_n<mode>_insn_be): This.
25656 * config/aarch64/iterators.md: Define SD_HSDI iterator.
25658 2021-05-19 Jonathn Wright <jonathan.wright@arm.com>
25660 * config/aarch64/aarch64-simd.md: Use UNSPEC_SQXTUN instead
25662 * config/aarch64/iterators.md: Remove UNSPEC_SQXTUN2.
25664 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
25666 * config/aarch64/aarch64-simd.md (aarch64_<sur>q<r>shr<u>n2_n<mode>):
25667 Implement as an expand emitting a big/little endian
25668 instruction pattern.
25669 (aarch64_<sur>q<r>shr<u>n2_n<mode>_insn_le): Define.
25670 (aarch64_<sur>q<r>shr<u>n2_n<mode>_insn_be): Define.
25672 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
25674 * config/aarch64/aarch64-simd.md (aarch64_<sur><addsub>hn2<mode>):
25675 Implement as an expand emitting a big/little endian
25676 instruction pattern.
25677 (aarch64_<sur><addsub>hn2<mode>_insn_le): Define.
25678 (aarch64_<sur><addsub>hn2<mode>_insn_be): Define.
25679 * config/aarch64/iterators.md: Remove UNSPEC_[R]ADDHN2 and
25680 UNSPEC_[R]SUBHN2 unspecs and ADDSUBHN2 iterator.
25682 2021-05-19 Richard Biener <rguenther@suse.de>
25684 PR middle-end/100672
25685 * fold-const.c (fold_negate_expr_1): Use element_precision.
25686 (negate_expr_p): Likewise.
25688 2021-05-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
25690 * config/aarch64/iterators.md (SVE_PRED_LOAD): New iterator.
25691 (pred_load): New int attribute.
25692 * config/aarch64/aarch64-sve.md
25693 (aarch64_load_<ANY_EXTEND:optab><SVE_HSDI:mode><SVE_PARTIAL_I:mode>): Use
25694 SVE_PRED_LOAD enum iterator and corresponding pred_load attribute.
25695 * config/aarch64/aarch64-sve-builtins-base.cc (expand): Update call to
25696 code_for_aarch64_load.
25698 2021-05-19 Richard Biener <rguenther@suse.de>
25700 * cfgexpand.c (discover_nonconstant_array_refs_r): Make
25701 sure TARGET_MEM_REF bases are expanded as memory.
25702 * tree-ssa-operands.c (operands_scanner::get_tmr_operands):
25703 Do not mark TARGET_MEM_REF bases addressable.
25704 * tree-ssa.c (non_rewritable_mem_ref_base): Handle
25705 TARGET_MEM_REF bases as never rewritable.
25706 * gimple-walk.c (walk_stmt_load_store_addr_ops): Do not
25707 walk TARGET_MEM_REF bases as address-takens.
25708 * tree-ssa-dce.c (ref_may_be_aliased): Handle TARGET_MEM_REF.
25710 2021-05-19 Richard Biener <rguenther@suse.de>
25712 * builtins.c (get_object_alignment_1): Strip outer
25714 * tree-dfa.c (get_ref_base_and_extent): Handle outer
25715 WITH_SIZE_EXPR for size processing and process the
25717 * tree-ssa-alias.c (ao_ref_base_alias_set): Strip
25718 outer WITH_SIZE_EXPR.
25719 (ao_ref_base_alias_ptr_type): Likewise.
25720 (refs_may_alias_p_2): Allow WITH_SIZE_EXPR in ref->ref
25721 and handle that accordingly, stripping it for the
25722 core alias workers.
25723 * tree.c (get_base_address): Handle WITH_SIZE_EXPR by
25724 looking through it instead of returning NULL.
25726 2021-05-19 Jakub Jelinek <jakub@redhat.com>
25728 PR middle-end/100576
25729 * builtins.c (check_read_access): Convert bound to size_type_node if
25732 2021-05-19 Richard Biener <rguenther@suse.de>
25734 * tree-cfg.c (verify_types_in_gimple_min_lval): Inline...
25735 (verify_types_in_gimple_reference): ... here. Sanitize.
25736 (verify_gimple_call): Verify references in LHS and arguments.
25737 (verify_gimple_assign_single): Reject WITH_SIZE_EXPR.
25739 2021-05-19 Uroš Bizjak <ubizjak@gmail.com>
25741 * config/i386/i386.h (VALID_INT_MODE_P):
25742 Add V8QI, V4HI and V2SI modes for TARGET_64BIT.
25743 * config/i386/i386.md (isa): Add x64_bmi.
25744 (enabled): Handle x64_bmi.
25745 * config/i386/mmx.md (mmx_andnot<MMXMODEI:mode>3):
25746 Add alternative using 64bit general registers.
25747 (*mmx_<any_logic:code><MMXMODEI:mode>3): Ditto.
25749 2021-05-19 Jakub Jelinek <jakub@redhat.com>
25751 PR middle-end/99928
25752 * tree.h (OMP_MASTER_COMBINED): Define.
25753 * gimplify.c (gimplify_scan_omp_clauses): Rewrite lastprivate
25754 handling for outer combined/composite constructs to a loop.
25755 Handle lastprivate on combined target.
25756 (gimplify_expr): Formatting fix.
25758 2021-05-19 Xionghu Luo <luoxhu@linux.ibm.com>
25760 * passes.def: Add sink_code pass before store_merging.
25761 * tree-ssa-sink.c (pass_sink_code:clone): New.
25763 2021-05-18 Bill Schmidt <wschmidt@linux.ibm.com>
25765 * config/rs6000/freebsd64.h (ADJUST_FIELD_ALIGN): Remove call to
25766 rs6000_special_adjust_field_align_p.
25767 * config/rs6000/linux64.h (ADJUST_FIELD_ALIGN): Likewise.
25768 * config/rs6000/rs6000-call.c (rs6000_function_arg_boundary):
25769 Remove ABI warning.
25770 (rs6000_function_arg): Likewise.
25771 * config/rs6000/rs6000-protos.h
25772 (rs6000_special_adjust_field_align_p): Remove prototype.
25773 * config/rs6000/rs6000.c (rs6000_special_adjust_field_align_p):
25775 * config/rs6000/sysv4.h (ADJUST_FIELD_ALIGN): Remove call to
25776 rs6000_special_adjust_field_align_p.
25778 2021-05-18 Uroš Bizjak <ubizjak@gmail.com>
25781 * config/i386/i386.h (VALID_SSE2_REG_MODE):
25782 Add V4QI and V2HI modes.
25783 (VALID_INT_MODE_P): Ditto.
25784 * config/i386/mmx.md (VI_32): New mode iterator.
25785 (mmxvecsize): Handle V4QI and V2HI.
25787 (mov<VI_32:mode>): New expander.
25788 (*mov<mode>_internal): New insn pattern.
25789 (movmisalign<VI_32:mode>): New expander.
25790 (neg<VI_32:mode>): New expander.
25791 (<plusminus:insn><VI_32:mode>3): New expander.
25792 (*<plusminus:insn><VI_32:mode>3): New insn pattern.
25793 (mulv2hi3): New expander.
25794 (*mulv2hi3): New insn pattern.
25795 (one_cmpl<VI_32:mode>2): New expander.
25796 (*andnot<VI_32:mode>3): New insn pattern.
25797 (<any_logic:code><VI_32:mode>3): New expander.
25798 (*<any_logic:code><VI_32:mode>3): New insn pattern.
25800 2021-05-18 Uroš Bizjak <ubizjak@gmail.com>
25802 * config/i386/sse.md (<any_extend:insn>v4qiv4di2):
25803 Fix a mode mismatch with operand 1.
25805 2021-05-18 Uroš Bizjak <ubizjak@gmail.com>
25808 * config/i386/i386-expand.c (split_double_mode): Return
25809 temporary register when simplify_gen_subreg fails with
25810 the high half od the paradoxical subreg.
25812 2021-05-18 Richard Biener <rguenther@suse.de>
25814 * cfgexpand.c (expand_one_var): Pass in forced_stack_var
25815 and honor it when expanding.
25816 (expand_used_vars_for_block): Pass through forced_stack_var.
25817 (expand_used_vars): Likewise.
25818 (discover_nonconstant_array_refs_r): Set bits in
25819 forced_stack_vars instead of marking vars TREE_ADDRESSABLE.
25820 (avoid_type_punning_on_regs): Likewise.
25821 (discover_nonconstant_array_refs): Likewise.
25822 (pass_expand::execute): Create and pass down forced_stack_var
25823 bitmap. For parameters and returns temporarily set
25824 TREE_ADDRESSABLE when expand_function_start.
25826 2021-05-18 Thomas Schwinge <thomas@codesourcery.com>
25828 * doc/sourcebuild.texi: Document 'dg-note'.
25830 2021-05-18 Tobias Burnus <tobias@codesourcery.com>
25833 * configure: Regenerate.
25834 * configure.ac (BUILD_CFLAG, BUILD_CXXFLAGS): Add $(CFLAGS-$@).
25836 2021-05-18 Thomas Schwinge <thomas@codesourcery.com>
25838 * gimple.h (is_gimple_omp_oacc): Tighten.
25839 * omp-low.c (check_omp_nesting_restrictions): Adjust.
25841 2021-05-18 Richard Biener <rguenther@suse.de>
25843 * tree-ssa-operands.c (mark_address_taken): Simplify.
25845 2021-05-18 Martin Liska <mliska@suse.cz>
25847 * config/gcn/mkoffload.c (STR): Redefine.
25848 * config/i386/intelmic-mkoffload.c (STR): Likewise.
25849 * config/nvptx/mkoffload.c (STR): Likewise.
25851 2021-05-18 Martin Liska <mliska@suse.cz>
25853 * common/config/aarch64/aarch64-common.c (aarch64_parse_extension):
25854 Use startswith function instead of strncmp.
25855 * common/config/bfin/bfin-common.c (bfin_handle_option): Likewise.
25856 * common/config/riscv/riscv-common.c (riscv_subset_list::parse): Likewise.
25857 * config/aarch64/aarch64-sve-builtins-shapes.cc (parse_type): Likewise.
25858 * config/aarch64/aarch64.c (aarch64_process_one_target_attr): Likewise.
25859 * config/alpha/alpha.c (alpha_elf_section_type_flags): Likewise.
25860 * config/arm/aarch-common.c (arm_md_asm_adjust): Likewise.
25861 * config/arm/arm.c (arm_file_start): Likewise.
25862 (arm_valid_target_attribute_rec): Likewise.
25863 (thumb1_md_asm_adjust): Likewise.
25864 * config/arm/driver-arm.c (host_detect_local_cpu): Likewise.
25865 * config/avr/avr.c (STR_PREFIX_P): Likewise.
25866 (avr_set_current_function): Likewise.
25867 (avr_handle_addr_attribute): Likewise.
25868 (avr_asm_output_aligned_decl_common): Likewise.
25869 (avr_asm_named_section): Likewise.
25870 (avr_section_type_flags): Likewise.
25871 (avr_asm_select_section): Likewise.
25872 * config/c6x/c6x.c (c6x_in_small_data_p): Likewise.
25873 (c6x_section_type_flags): Likewise.
25874 * config/darwin-c.c (darwin_cfstring_ref_p): Likewise.
25875 (darwin_objc_declare_unresolved_class_reference): Likewise.
25876 (darwin_objc_declare_class_definition): Likewise.
25877 * config/darwin.c (indirect_data): Likewise.
25878 (darwin_encode_section_info): Likewise.
25879 (darwin_objc2_section): Likewise.
25880 (darwin_objc1_section): Likewise.
25881 (machopic_select_section): Likewise.
25882 (darwin_globalize_label): Likewise.
25883 (darwin_label_is_anonymous_local_objc_name): Likewise.
25884 (darwin_asm_named_section): Likewise.
25885 (darwin_asm_output_dwarf_offset): Likewise.
25886 * config/frv/frv.c (frv_string_begins_with): Likewise.
25887 (frv_in_small_data_p): Likewise.
25888 * config/gcn/mkoffload.c (STR): Likewise.
25890 * config/i386/i386-builtins.c (get_builtin_code_for_version): Likewise.
25891 * config/i386/i386-options.c (ix86_option_override_internal): Likewise.
25892 * config/i386/i386.c (x86_64_elf_section_type_flags): Likewise.
25893 (ix86_md_asm_adjust): Likewise.
25894 * config/i386/intelmic-mkoffload.c (STR): Likewise.
25895 * config/i386/winnt.c (i386_pe_asm_named_section): Likewise.
25896 (i386_pe_file_end): Likewise.
25897 * config/ia64/ia64.c (ia64_in_small_data_p): Likewise.
25898 (ia64_section_type_flags): Likewise.
25899 * config/mips/driver-native.c (host_detect_local_cpu): Likewise.
25900 * config/mips/mips.c (mips_handle_interrupt_attr): Likewise.
25901 (mips16_stub_function_p): Likewise.
25902 (mips_function_rodata_section): Likewise.
25903 * config/msp430/msp430.c (msp430_mcu_name): Likewise.
25904 (msp430_function_section): Likewise.
25905 (msp430_section_type_flags): Likewise.
25906 (msp430_expand_helper): Likewise.
25907 * config/nios2/nios2.c (nios2_small_section_name_p): Likewise.
25908 (nios2_valid_target_attribute_rec): Likewise.
25909 * config/nvptx/mkoffload.c (process): Likewise.
25911 * config/pa/som.h: Likewise.
25912 * config/pdp11/pdp11.c (pdp11_output_ident): Likewise.
25913 * config/riscv/riscv.c (riscv_elf_select_rtx_section): Likewise.
25914 * config/rs6000/rs6000.c (VTABLE_NAME_P): Likewise.
25915 (rs6000_inner_target_options): Likewise.
25916 * config/s390/driver-native.c (s390_host_detect_local_cpu): Likewise.
25917 * config/sparc/driver-sparc.c (host_detect_local_cpu): Likewise.
25918 * config/vax/vax.c (vax_output_int_move): Likewise.
25919 * config/vms/vms-ld.c (startswith): Likewise.
25920 (process_args): Likewise.
25922 * config/vms/vms.c: Likewise.
25924 2021-05-18 Jakub Jelinek <jakub@redhat.com>
25926 PR rtl-optimization/100590
25927 * regcprop.c (copyprop_hardreg_forward_1): Only DCE dead sets if
25928 they are NONJUMP_INSN_P.
25930 2021-05-18 Jakub Jelinek <jakub@redhat.com>
25933 * function.c (push_dummy_function): Set DECL_ARTIFICIAL and
25934 DECL_ASSEMBLER_NAME on the fn_decl.
25936 2021-05-18 Jakub Jelinek <jakub@redhat.com>
25938 PR tree-optimization/94589
25939 * tree-ssa-phiopt.c (spaceship_replacement): Pattern match
25940 phi result used in (res & ~1) == 0 comparison as res >= 0 as
25941 res == 2 would be UB with -ffinite-math-only.
25943 2021-05-18 Martin Liska <mliska@suse.cz>
25945 * Makefile.in: genversion.o should depend on DATESTAMP.
25947 2021-05-18 Claudiu Zissulescu <claziss@synopsys.com>
25949 * config/arc/simdext.md (negv2si2): Remove round bracket.
25951 2021-05-18 Andreas Krebbel <krebbel@linux.ibm.com>
25953 * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal): Define
25954 _Bool as macro expanding to _Bool.
25956 2021-05-18 Andreas Krebbel <krebbel@linux.ibm.com>
25959 * tree.c (build_reference_type_for_mode)
25960 (build_pointer_type_for_mode): Pick pointer mode if MODE argument
25962 (build_reference_type, build_pointer_type): Invoke
25963 build_*_type_for_mode with VOIDmode.
25965 2021-05-17 Andrew MacLeod <amacleod@redhat.com>
25967 PR tree-optimization/100512
25968 * gimple-range-cache.cc (ranger_cache::set_global_range): Mark const
25969 and non-zero pointer ranges as invariant.
25970 * gimple-range.cc (gimple_ranger::range_of_stmt): Remove pointer
25971 processing from here.
25973 2021-05-17 Tom de Vries <tdevries@suse.de>
25976 * config/nvptx/nvptx-protos.h (nvptx_output_atomic_insn): Declare
25977 * config/nvptx/nvptx.c (nvptx_output_barrier)
25978 (nvptx_output_atomic_insn): New function.
25979 (nvptx_print_operand): Add support for 'B'.
25980 * config/nvptx/nvptx.md: Use nvptx_output_atomic_insn for atomic
25983 2021-05-17 Aldy Hernandez <aldyh@redhat.com>
25985 PR tree-optimization/100349
25986 * vr-values.c (bounds_of_var_in_loop): Bail if scev returns
25989 2021-05-17 Tamar Christina <tamar.christina@arm.com>
25991 * config/aarch64/driver-aarch64.c (DEFAULT_ARCH): New.
25992 (host_detect_local_cpu): Use it.
25994 2021-05-17 Martin Liska <mliska@suse.cz>
25996 * doc/invoke.texi: Add 2 missing dots.
25998 2021-05-17 Marius Hillenbrand <mhillen@linux.ibm.com>
26000 PR bootstrap/100552
26001 * configure.ac: Replace pattern substitution with call to sed.
26002 * configure: Regenerate.
26004 2021-05-17 Richard Biener <rguenther@suse.de>
26006 PR middle-end/100582
26007 * tree.c (array_at_struct_end_p): Get to the base of the
26008 reference before looking for the underlying decl.
26010 2021-05-17 Joern Rennecke <joern.rennecke@embecosm.com>
26012 * genoutput.c (validate_insn_alternatives) Make "wrong number of
26013 alternatives" message more specific, and remove assumption on where
26016 2021-05-17 Christophe Lyon <christophe.lyon@linaro.org>
26018 * config/arm/iterators.md (V16): New iterator.
26019 (VH_cvtto): New iterator.
26020 (v_cmp_result): Added V4HF and V8HF support.
26021 * config/arm/vec-common.md (vec_cmp<mode><v_cmp_result>): Use VDQWH.
26022 (vcond<mode><mode>): Likewise.
26023 (vcond_mask_<mode><v_cmp_result>): Likewise.
26024 (vcond<VH_cvtto><mode>): New expander.
26026 2021-05-17 Christophe Lyon <christophe.lyon@linaro.org>
26028 * config/arm/arm-protos.h (arm_expand_vector_compare): Update
26030 * config/arm/arm.c (arm_expand_vector_compare): Add support for
26032 (arm_expand_vcond): Likewise.
26033 * config/arm/iterators.md (supf): Remove VCMPNEQ_S, VCMPEQQ_S,
26034 VCMPEQQ_N_S, VCMPNEQ_N_S.
26035 (VCMPNEQ, VCMPEQQ, VCMPEQQ_N, VCMPNEQ_N): Remove.
26036 * config/arm/mve.md (@mve_vcmp<mve_cmp_op>q_<mode>): Add '@' prefix.
26037 (@mve_vcmp<mve_cmp_op>q_f<mode>): Likewise.
26038 (@mve_vcmp<mve_cmp_op>q_n_f<mode>): Likewise.
26039 (@mve_vpselq_<supf><mode>): Likewise.
26040 (@mve_vpselq_f<mode>"): Likewise.
26041 * config/arm/neon.md (vec_cmp<mode><v_cmp_result): Enable for MVE
26042 and move to vec-common.md.
26043 (vec_cmpu<mode><mode>): Likewise.
26044 (vcond<mode><mode>): Likewise.
26045 (vcond<V_cvtto><mode>): Likewise.
26046 (vcondu<mode><v_cmp_result>): Likewise.
26047 (vcond_mask_<mode><v_cmp_result>): Likewise.
26048 * config/arm/unspecs.md (VCMPNEQ_U, VCMPNEQ_S, VCMPEQQ_S)
26049 (VCMPEQQ_N_S, VCMPNEQ_N_S, VCMPEQQ_U, CMPEQQ_N_U, VCMPNEQ_N_U)
26050 (VCMPGEQ_N_S, VCMPGEQ_S, VCMPGTQ_N_S, VCMPGTQ_S, VCMPLEQ_N_S)
26051 (VCMPLEQ_S, VCMPLTQ_N_S, VCMPLTQ_S, VCMPCSQ_N_U, VCMPCSQ_U)
26052 (VCMPHIQ_N_U, VCMPHIQ_U): Remove.
26053 * config/arm/vec-common.md (vec_cmp<mode><v_cmp_result): Moved
26055 (vec_cmpu<mode><mode>): Likewise.
26056 (vcond<mode><mode>): Likewise.
26057 (vcond<V_cvtto><mode>): Likewise.
26058 (vcondu<mode><v_cmp_result>): Likewise.
26059 (vcond_mask_<mode><v_cmp_result>): Likewise. Added unsafe math
26062 2021-05-17 liuhongt <hongtao.liu@intel.com>
26065 * config/i386/i386.c (ix86_gimple_fold_builtin): Use
26066 gsi_insert_seq_before instead.
26068 2021-05-17 Christophe Lyon <christophe.lyon@linaro.org>
26070 * doc/sourcebuild.texi (arm_qbit_ok): Rename into...
26071 (arm_sat_ok): ...this.
26073 2021-05-17 Martin Liska <mliska@suse.cz>
26075 * lto-wrapper.c (merge_flto_options): Factor out a new function.
26076 (merge_and_complain): Use it.
26077 (run_gcc): Merge also linker command line -flto=foo argument
26080 2021-05-16 Christophe Lyon <christophe.lyon@linaro.org>
26082 * config/arm/arm.h (CPP_SPEC): Remove error message about
26083 -mlittle-endian/-mbig-endian conflict.
26085 2021-05-15 Bill Schmidt <wschmidt@linux.ibm.com>
26087 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Define
26088 __ROP_PROTECT__ if -mrop-protect is selected.
26090 2021-05-15 Bill Schmidt <wschmidt@linux.ibm.com>
26092 * config/rs6000/rs6000-internal.h (rs6000_stack): Add
26093 rop_hash_save_offset and rop_hash_size.
26094 * config/rs6000/rs6000-logue.c (rs6000_stack_info): Compute
26095 rop_hash_size and rop_hash_save_offset.
26096 (debug_stack_info): Dump rop_hash_save_offset and rop_hash_size.
26097 (rs6000_emit_prologue): Emit hashst[p] in prologue.
26098 (rs6000_emit_epilogue): Emit hashchk[p] in epilogue.
26099 * config/rs6000/rs6000.md (unspec): Add UNSPEC_HASHST and
26101 (hashst): New define_insn.
26102 (hashchk): Likewise.
26104 2021-05-15 Bill Schmidt <wschmidt@linux.ibm.com>
26106 * config/rs6000/rs6000.c (rs6000_option_override_internal):
26107 Disable shrink wrap when inserting ROP-protect instructions.
26108 * config/rs6000/rs6000.opt (mrop-protect): New option.
26109 (mprivileged): Likewise.
26110 * doc/invoke.texi: Document mrop-protect and mprivileged.
26112 2021-05-15 Hans-Peter Nilsson <hp@axis.com>
26114 * reorg.c (fill_slots_from_thread): Reinstate code typoed out in
26117 2021-05-15 Martin Jambor <mjambor@suse.cz>
26120 2021-05-13 Martin Jambor <mjambor@suse.cz>
26122 PR tree-optimization/100453
26123 * tree-sra.c (sra_modify_assign): All const base accesses do not
26124 need refreshing, not just those from decl_pool.
26125 (sra_modify_assign): Do not refresh into a const base decl.
26127 2021-05-15 Jakub Jelinek <jakub@redhat.com>
26129 PR rtl-optimization/100342
26130 * regcprop.c (copy_value): When copying a source reg in a wider
26131 mode than it has recorded for the value, adjust recorded destination
26132 mode too or punt if !REG_CAN_CHANGE_MODE_P.
26134 2021-05-14 Jason Merrill <jason@redhat.com>
26136 * intl.h: Add comments.
26138 2021-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
26140 * config/aarch64/aarch64-simd.md
26141 (aarch64_sqdml<SBINQOPS:as>l2_lane<mode>_internal): Split into...
26142 (aarch64_sqdmlsl2_lane<mode>_internal): ... This...
26143 (aarch64_sqdmlal2_lane<mode>_internal): ... And this.
26144 (aarch64_sqdml<SBINQOPS:as>l2_laneq<mode>_internal): Split into ...
26145 (aarch64_sqdmlsl2_laneq<mode>_internal): ... This...
26146 (aarch64_sqdmlal2_laneq<mode>_internal): ... And this.
26147 (aarch64_sqdml<SBINQOPS:as>l2_n<mode>_internal): Split into...
26148 (aarch64_sqdmlsl2_n<mode>_internal): ... This...
26149 (aarch64_sqdmlal2_n<mode>_internal): ... And this.
26151 2021-05-14 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
26154 * config/arm/arm_neon.h (vtst_s8): Replace call to vtst builtin with it's
26155 boolean logic equivalent.
26156 (vtst_s16): Likewise.
26157 (vtst_s32): Likewise.
26158 (vtst_u8): Likewise.
26159 (vtst_u16): Likewise.
26160 (vtst_u32): Likewise.
26161 (vtst_p8): Likewise.
26162 (vtst_p16): Likewise.
26163 (vtstq_s8): Likewise.
26164 (vtstq_s16): Likewise.
26165 (vtstq_s32): Likewise.
26166 (vtstq_u8): Likewise.
26167 (vtstq_u16): Likewise.
26168 (vtstq_u32): Likewise.
26169 (vtstq_p8): Likewise.
26170 (vtstq_p16): Likewise.
26171 * config/arm/arm_neon_builtins.def: Remove entry for vtst.
26172 * config/arm/neon.md (neon_vtst<mode>): Remove pattern.
26174 2021-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
26176 * config/aarch64/aarch64-simd.md (aarch64_sqdmlal2<mode>): Merge into...
26177 (aarch64_sqdml<SBINQOPS:as>l2<mode>): ... This.
26178 (aarch64_sqdmlsl2<mode>): Delete.
26179 (aarch64_sqdmlal2_lane<mode>): Merge this...
26180 (aarch64_sqdmlsl2_lane<mode>): ... And this...
26181 (aarch64_sqdml<SBINQOPS:as>l2_lane<mode>): ... Into this.
26182 (aarch64_sqdmlal2_laneq<mode>): Merge this...
26183 (aarch64_sqdmlsl2_laneq<mode>): ... And this...
26184 (aarch64_sqdml<SBINQOPS:as>l2_laneq<mode>): ... Into this.
26185 (aarch64_sqdmlal2_n<mode>): Merge this...
26186 (aarch64_sqdmlsl2_n<mode>): ... And this...
26187 (aarch64_sqdml<SBINQOPS:as>l2_n<mode>): ... Into this.
26189 2021-05-13 Martin Sebor <msebor@redhat.com>
26191 PR middle-end/100574
26192 * builtins.c (access_ref::get_ref): Improve detection of PHIs with
26193 all null arguments.
26195 2021-05-13 Martin Sebor <msebor@redhat.com>
26197 PR tree-optimization/93100
26198 PR middle-end/98583
26199 * tree-ssa-uninit.c (check_defs): Exclude intrinsic functions that
26200 don't modify referenced objects.
26202 2021-05-13 Martin Jambor <mjambor@suse.cz>
26204 PR tree-optimization/100453
26205 * tree-sra.c (sra_modify_assign): All const base accesses do not
26206 need refreshing, not just those from decl_pool.
26207 (sra_modify_assign): Do not refresh into a const base decl.
26209 2021-05-13 Martin Liska <mliska@suse.cz>
26211 * tree-ssa-dom.c: Remove m_simplifier.
26213 2021-05-13 Richard Earnshaw <rearnsha@arm.com>
26216 * config/arm/arm.c (arm_canonicalize_comparison): Correctly
26217 canonicalize DImode inequality comparisons against the
26218 maximum integral value.
26220 2021-05-13 Jakub Jelinek <jakub@redhat.com>
26222 PR tree-optimization/98856
26223 * config/i386/i386.c (ix86_shift_rotate_cost): Add CODE argument.
26224 Expect V2DI and V4DI arithmetic right shifts to be emulated.
26225 (ix86_rtx_costs, ix86_add_stmt_cost): Adjust ix86_shift_rotate_cost
26227 * config/i386/i386-expand.c (expand_vec_perm_2perm_interleave,
26228 expand_vec_perm_2perm_pblendv): New functions.
26229 (ix86_expand_vec_perm_const_1): Use them.
26230 * config/i386/sse.md (ashr<mode>3<mask_name>): Rename to ...
26231 (<mask_codefor>ashr<mode>3<mask_name>): ... this.
26232 (ashr<mode>3): New define_expand with VI248_AVX512BW iterator.
26233 (ashrv4di3): New define_expand.
26234 (ashrv2di3): Change condition to TARGET_SSE2, handle !TARGET_XOP
26235 and !TARGET_AVX512VL expansion.
26237 2021-05-13 Uroš Bizjak <ubizjak@gmail.com>
26240 * config/i386/i386-expand.c (ix86_expand_sse_movcc): Force mode
26241 sizes < 16 to a register when constructing vpcmov pattern.
26242 * config/i386/mmx.md (*xop_pcmov_<mode>): Use MMXMODE124 mode.
26244 2021-05-13 Martin Liska <mliska@suse.cz>
26246 * gcov-io.c (gcov_write_block): Remove.
26247 (gcov_write_words): Likewise.
26248 (gcov_read_words): Re-implement using gcov_read_bytes.
26249 (gcov_allocate): Remove.
26250 (GCOV_BLOCK_SIZE): Likewise.
26251 (struct gcov_var): Remove most of the fields.
26252 (gcov_position): Implement with ftell.
26253 (gcov_rewrite): Remove setting of start and offset fields.
26254 (from_file): Re-format.
26255 (gcov_open): Remove setbuf call. It should not be needed.
26256 (gcov_close): Remove internal buffer handling.
26257 (gcov_magic): Use __builtin_bswap32.
26258 (gcov_write_counter): Use directly gcov_write_unsigned.
26259 (gcov_write_string): Use direct fwrite and do not round
26261 (gcov_seek): Use directly fseek.
26262 (gcov_write_tag): Use gcov_write_unsigned directly.
26263 (gcov_write_length): Likewise.
26264 (gcov_write_tag_length): Likewise.
26265 (gcov_read_bytes): Use directly fread.
26266 (gcov_read_unsigned): Use gcov_read_words.
26267 (gcov_read_counter): Likewise.
26268 (gcov_read_string): Use gcov_read_bytes.
26269 * gcov-io.h (GCOV_WORD_SIZE): Adjust to reflect
26270 that size is not in bytes, but words (4B).
26271 (GCOV_TAG_FUNCTION_LENGTH): Likewise.
26272 (GCOV_TAG_ARCS_LENGTH): Likewise.
26273 (GCOV_TAG_ARCS_NUM): Likewise.
26274 (GCOV_TAG_COUNTER_LENGTH): Likewise.
26275 (GCOV_TAG_COUNTER_NUM): Likewise.
26276 (GCOV_TAG_SUMMARY_LENGTH): Likewise.
26278 2021-05-13 liuhongt <hongtao.liu@intel.com>
26281 * config/i386/sse.md (ssedoublevecmode): Add attribute for
26282 V64QI/V32HI/V16SI/V4DI.
26283 (ssehalfvecmode): Add attribute for V2DI/V2DF.
26284 (*vec_concatv4si_0): Extend to VI124_128.
26285 (*vec_concat<mode>_0): New pre-reload splitter.
26286 * config/i386/predicates.md (movq_parallel): New predicate.
26288 2021-05-13 Alexandre Oliva <oliva@adacore.com>
26290 * targhooks.c (default_zero_call_used_regs): Retry using
26291 successfully-zeroed registers as sources.
26293 2021-05-12 Tobias Burnus <tobias@codesourcery.com>
26295 * omp-low.c (finish_taskreg_scan): Use the proper detach decl.
26297 2021-05-12 Aldy Hernandez <aldyh@redhat.com>
26300 * gimple-range.cc (range_of_builtin_call): Skip out on
26301 processing __builtin_clz when varying.
26303 2021-05-12 Tom de Vries <tdevries@suse.de>
26306 * config/nvptx/nvptx-opts.h (enum ptx_version): New enum.
26307 * config/nvptx/nvptx.c (nvptx_file_start): Print .version according
26308 to ptx_version_option.
26309 * config/nvptx/nvptx.h (TARGET_PTX_6_3): Define.
26310 * config/nvptx/nvptx.md (define_insn "nvptx_shuffle<mode>")
26311 (define_insn "nvptx_vote_ballot"): Use sync variant for
26313 * config/nvptx/nvptx.opt (ptx_version): Add enum.
26314 (mptx): Add option.
26315 * doc/invoke.texi (Nvidia PTX Options): Add mptx item.
26317 2021-05-12 Richard Biener <rguenther@suse.de>
26319 PR tree-optimization/100566
26320 * tree-ssa-sccvn.c (dominated_by_p_w_unex): Properly handle
26321 allow_back for all edge queries.
26323 2021-05-12 liuhongt <hongtao.liu@intel.com>
26326 * config/i386/sse.md (<sse4_1_avx2>_pblendvb): Add
26327 splitters for pblendvb of NOT mask register.
26329 2021-05-12 Richard Biener <rguenther@suse.de>
26331 PR tree-optimization/100519
26332 * tree-ssa-reassoc.c (can_associate_p): Split into...
26333 (can_associate_op_p): ... this
26334 (can_associate_type_p): ... and this.
26335 (is_reassociable_op): Call can_associate_op_p.
26336 (break_up_subtract_bb): Call the appropriate predicates.
26337 (reassociate_bb): Likewise.
26339 2021-05-12 Martin Liska <mliska@suse.cz>
26341 * lto-wrapper.c (merge_and_complain): Merge -flto=arg options.
26342 (run_gcc): Use -flto argument detection for merged
26345 2021-05-12 Martin Liska <mliska@suse.cz>
26347 * lto-wrapper.c (print_lto_docs_link): New function.
26348 (run_gcc): Print warning about missing job server detection
26349 after we know NR of partitions. Do the same for -flto{,=1}.
26350 * opts.c (get_option_html_page): Support -flto option.
26352 2021-05-12 Martin Liska <mliska@suse.cz>
26354 * lto-wrapper.c (get_options_from_collect_gcc_options): Change
26356 (append_option): Remove.
26357 (find_option): Rework to use the vector type.
26358 (remove_option): Remove.
26359 (merge_and_complain): Use vectors for cl_decoded_option data
26361 (append_compiler_options): Likewise.
26362 (append_diag_options): Likewise.
26363 (append_linker_options): Likewise.
26364 (append_offload_options): Likewise.
26365 (compile_offload_image): Likewise.
26366 (compile_images_for_offload_targets): Likewise.
26367 (find_and_merge_options): Likewise.
26368 (run_gcc): Likewise.
26370 2021-05-12 Bernd Edlinger <bernd.edlinger@hotmail.de>
26373 * dwarf2out.c (dwarf2out_finish): Set
26374 have_multiple_function_sections with multi-range text_section.
26376 2021-05-12 Martin Liska <mliska@suse.cz>
26378 PR bootstrap/100560
26379 * Makefile.in: Remove version.h from linker command line.
26381 2021-05-12 Richard Biener <rguenther@suse.de>
26383 PR middle-end/100547
26384 * rtl.h (rtvec_alloc): Make argument size_t.
26385 * rtl.c (rtvec_alloc): Verify the count is less than INT_MAX.
26387 2021-05-12 Jakub Jelinek <jakub@redhat.com>
26389 PR middle-end/100508
26390 * cfgexpand.c (expand_debug_expr): For DEBUG_EXPR_DECL with vector
26391 type, don't reuse DECL_RTL if it has different mode, instead force
26392 creation of a new DEBUG_EXPR.
26394 2021-05-12 Jakub Jelinek <jakub@redhat.com>
26395 Marc Glisse <marc.glisse@inria.fr>
26397 PR tree-optimization/94589
26398 * match.pd ((X & Y) == X -> (X & ~Y) == 0,
26399 (X | Y) == Y -> (X & ~Y) == 0): New GIMPLE simplifications.
26401 2021-05-12 Uroš Bizjak <ubizjak@gmail.com>
26404 * config/i386/i386-expand.c (ix86_expand_sse_movcc): Handle V2SF mode.
26405 * config/i386/mmx.md (MMXMODE124): New mode iterator.
26407 (mmxintvecmode): New mode attribute.
26408 (mmxintvecmodelower): Ditto.
26409 (*mmx_maskcmpv2sf3_comm): New insn pattern.
26410 (*mmx_maskcmpv2sf3): Ditto.
26411 (vec_cmpv2sfv2si): New expander.
26412 (vcond<V2FI:mode>v2si): Ditto.
26413 (mmx_vlendvps): New insn pattern.
26414 (vcond<MMXMODE124:mode><MMXMODEI:mode>): Also handle V2SFmode.
26415 (vcondu<MMXMODE124:mode><MMXMODEI:mode>): Ditto.
26416 (vcond_mask_<mode><mmxintvecmodelower>): Ditto.
26418 2021-05-11 Martin Sebor <msebor@redhat.com>
26420 PR middle-end/21433
26421 * expr.c (expand_expr_real_1): Replace unreachable code with an assert.
26423 2021-05-11 Richard Biener <rguenther@suse.de>
26425 * gimple-fold.c (gimple_fold_call): Do not call
26426 maybe_fold_reference on call arguments or the static chain.
26427 (fold_stmt_1): Do not call maybe_fold_reference on GIMPLE_ASM
26430 2021-05-11 Martin Liska <mliska@suse.cz>
26432 * builtins.def (DEF_HSAIL_BUILTIN): Remove.
26433 (DEF_HSAIL_ATOMIC_BUILTIN): Likewise.
26434 (DEF_HSAIL_SAT_BUILTIN): Likewise.
26435 (DEF_HSAIL_INTR_BUILTIN): Likewise.
26436 (DEF_HSAIL_CVT_ZEROI_SAT_BUILTIN): Likewise.
26437 * doc/frontends.texi: Remove BRIG.
26438 * doc/install.texi: Likewise.
26439 * doc/invoke.texi: Likewise.
26440 * doc/standards.texi: Likewise.
26441 * brig-builtins.def: Removed.
26442 * brig/ChangeLog: Removed.
26443 * brig/Make-lang.in: Removed.
26444 * brig/brig-builtins.h: Removed.
26445 * brig/brig-c.h: Removed.
26446 * brig/brig-lang.c: Removed.
26447 * brig/brigfrontend/brig-arg-block-handler.cc: Removed.
26448 * brig/brigfrontend/brig-atomic-inst-handler.cc: Removed.
26449 * brig/brigfrontend/brig-basic-inst-handler.cc: Removed.
26450 * brig/brigfrontend/brig-branch-inst-handler.cc: Removed.
26451 * brig/brigfrontend/brig-cmp-inst-handler.cc: Removed.
26452 * brig/brigfrontend/brig-code-entry-handler.cc: Removed.
26453 * brig/brigfrontend/brig-code-entry-handler.h: Removed.
26454 * brig/brigfrontend/brig-comment-handler.cc: Removed.
26455 * brig/brigfrontend/brig-control-handler.cc: Removed.
26456 * brig/brigfrontend/brig-copy-move-inst-handler.cc: Removed.
26457 * brig/brigfrontend/brig-cvt-inst-handler.cc: Removed.
26458 * brig/brigfrontend/brig-fbarrier-handler.cc: Removed.
26459 * brig/brigfrontend/brig-function-handler.cc: Removed.
26460 * brig/brigfrontend/brig-function.cc: Removed.
26461 * brig/brigfrontend/brig-function.h: Removed.
26462 * brig/brigfrontend/brig-inst-mod-handler.cc: Removed.
26463 * brig/brigfrontend/brig-label-handler.cc: Removed.
26464 * brig/brigfrontend/brig-lane-inst-handler.cc: Removed.
26465 * brig/brigfrontend/brig-machine.c: Removed.
26466 * brig/brigfrontend/brig-machine.h: Removed.
26467 * brig/brigfrontend/brig-mem-inst-handler.cc: Removed.
26468 * brig/brigfrontend/brig-module-handler.cc: Removed.
26469 * brig/brigfrontend/brig-queue-inst-handler.cc: Removed.
26470 * brig/brigfrontend/brig-seg-inst-handler.cc: Removed.
26471 * brig/brigfrontend/brig-signal-inst-handler.cc: Removed.
26472 * brig/brigfrontend/brig-to-generic.cc: Removed.
26473 * brig/brigfrontend/brig-to-generic.h: Removed.
26474 * brig/brigfrontend/brig-util.cc: Removed.
26475 * brig/brigfrontend/brig-util.h: Removed.
26476 * brig/brigfrontend/brig-variable-handler.cc: Removed.
26477 * brig/brigfrontend/hsa-brig-format.h: Removed.
26478 * brig/brigfrontend/phsa.h: Removed.
26479 * brig/brigspec.c: Removed.
26480 * brig/config-lang.in: Removed.
26481 * brig/gccbrig.texi: Removed.
26482 * brig/lang-specs.h: Removed.
26483 * brig/lang.opt: Removed.
26485 2021-05-11 Richard Biener <rguenther@suse.de>
26488 * ipa-param-manipulation.c
26489 (ipa_param_body_adjustments::modify_call_stmt): Avoid
26490 altering SSA_NAME_DEF_STMT by adjusting the calls LHS
26491 via gimple_call_lhs_ptr.
26493 2021-05-11 Alex Coplan <alex.coplan@arm.com>
26496 * config/arm/arm.c (cmse_nonsecure_call_inline_register_clear):
26497 Avoid emitting CFA adjusts on the sp if we have the fp.
26499 2021-05-11 Richard Sandiford <richard.sandiford@arm.com>
26501 * config/aarch64/iterators.md (VMUL_CHANGE_NLANES): Delete.
26502 (VMULD): New iterator.
26503 (VCOND): Handle V4HF and V8HF.
26504 (VCONQ): Fix entry for V2SF.
26505 * config/aarch64/aarch64-simd.md (mul_lane<mode>3): Use VMULD
26506 instead of VMUL. Use a 64-bit vector mode for the indexed operand.
26507 (*aarch64_mul3_elt_<vswap_width_name><mode>): Merge with...
26508 (mul_laneq<mode>3): ...this define_insn. Use VMUL instead of VDQSF.
26509 Use a 128-bit vector mode for the indexed operand. Use stype for
26510 the scheduling type.
26512 2021-05-11 Richard Biener <rguenther@suse.de>
26514 * gimple-fold.c (maybe_fold_reference): Only return
26515 is_gimple_min_invariant values.
26517 2021-05-11 Richard Biener <rguenther@suse.de>
26519 PR middle-end/100509
26520 * gimple-fold.c (fold_gimple_assign): Only call
26521 get_symbol_constant_value on register type symbols.
26523 2021-05-11 Srinath Parvathaneni <srinath.parvathaneni@arm.com>
26524 Joe Ramsay <joe.ramsay@arm.com>
26527 * config/arm/arm_mve.h (__arm_vstrwq_scatter_offset): Fix wrong arguments.
26528 (__arm_vcmpneq): Remove duplicate definition.
26529 (__arm_vstrwq_scatter_offset_p): Likewise.
26530 (__arm_vmaxq_x): Likewise.
26531 (__arm_vmlsdavaq): Likewise.
26532 (__arm_vmlsdavaxq): Likewise.
26533 (__arm_vmlsdavq_p): Likewise.
26534 (__arm_vmlsdavxq_p): Likewise.
26535 (__arm_vrmlaldavhaq): Likewise.
26536 (__arm_vstrbq_p): Likewise.
26537 (__arm_vstrbq_scatter_offset): Likewise.
26538 (__arm_vstrbq_scatter_offset_p): Likewise.
26539 (__arm_vstrdq_scatter_offset): Likewise.
26540 (__arm_vstrdq_scatter_offset_p): Likewise.
26541 (__arm_vstrdq_scatter_shifted_offset): Likewise.
26542 (__arm_vstrdq_scatter_shifted_offset_p): Likewise.
26544 2021-05-11 Jakub Jelinek <jakub@redhat.com>
26546 PR middle-end/100471
26547 * omp-low.c (lower_omp_task_reductions): For OMP_TASKLOOP, if data
26548 is 0, bypass the reduction loop including
26549 GOMP_taskgroup_reduction_unregister call.
26551 2021-05-11 Kewen Lin <linkw@linux.ibm.com>
26553 * config/rs6000/rs6000.c (struct rs6000_cost_data): New member
26554 costing_for_scalar.
26555 (rs6000_density_test): Early return if costing_for_scalar is true.
26556 (rs6000_init_cost): Init costing_for_scalar of rs6000_cost_data.
26558 2021-05-11 Kewen Lin <linkw@linux.ibm.com>
26560 * doc/tm.texi: Regenerated.
26561 * target.def (init_cost): Add new parameter costing_for_scalar.
26562 * targhooks.c (default_init_cost): Adjust for new parameter.
26563 * targhooks.h (default_init_cost): Likewise.
26564 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Likewise.
26565 (vect_compute_single_scalar_iteration_cost): Likewise.
26566 (vect_analyze_loop_2): Likewise.
26567 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Likewise.
26568 (vect_bb_vectorization_profitable_p): Likewise.
26569 * tree-vectorizer.h (init_cost): Likewise.
26570 * config/aarch64/aarch64.c (aarch64_init_cost): Likewise.
26571 * config/i386/i386.c (ix86_init_cost): Likewise.
26572 * config/rs6000/rs6000.c (rs6000_init_cost): Likewise.
26574 2021-05-11 Kewen Lin <linkw@linux.ibm.com>
26576 * config/rs6000/rs6000.c (rs6000_vect_nonmem): Renamed to
26577 vect_nonmem and moved into...
26578 (struct rs6000_cost_data): ...here.
26579 (rs6000_init_cost): Use vect_nonmem of cost_data instead.
26580 (rs6000_add_stmt_cost): Likewise.
26581 (rs6000_finish_cost): Likewise.
26583 2021-05-10 Eric Botcazou <ebotcazou@adacore.com>
26585 * range-op.cc (get_bool_state): Adjust head comment.
26586 (operator_not_equal::op1_range): Fix comment.
26587 (operator_bitwise_xor::op1_range): Remove call to gcc_unreachable.
26589 2021-05-10 Martin Sebor <msebor@redhat.com>
26591 PR middle-end/100425
26592 PR middle-end/100510
26593 * gimple-ssa-warn-alloca.c (pass_walloca::firast_time_p): Rename...
26594 (pass_walloca::xlimit_certain_p): ...to this.
26595 (pass_walloca::gate): Execute for any kind of handled warning.
26596 (pass_walloca::execute): Avoid issuing "maybe" and "unbounded"
26597 warnings when xlimit_certain_p is set.
26599 2021-05-10 Pat Haugen <pthaugen@linux.ibm.com>
26601 * config/rs6000/rs6000.c (rs6000_ira_change_pseudo_allocno_class):
26602 Return ALTIVEC_REGS if that is best_class.
26603 (rs6000_compute_pressure_classes): Add ALTIVEC_REGS.
26605 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
26607 * config/arm/arm.h (CPP_SPEC): Remove error message about
26610 2021-05-10 Martin Jambor <mjambor@suse.cz>
26612 * ipa-prop.h (IPA_NODE_REF): Removed.
26613 (IPA_NODE_REF_GET_CREATE): Likewise.
26614 (IPA_EDGE_REF): Likewise.
26615 (IPA_EDGE_REF_GET_CREATE): Likewise.
26616 (IS_VALID_JUMP_FUNC_INDEX): Likewise.
26617 * ipa-cp.c (print_all_lattices): Replaced IPA_NODE_REF with a direct
26618 use of ipa_node_params_sum.
26619 (ipcp_versionable_function_p): Likewise.
26620 (push_node_to_stack): Likewise.
26621 (pop_node_from_stack): Likewise.
26622 (set_single_call_flag): Replaced two IPA_NODE_REF with one single
26623 direct use of ipa_node_params_sum.
26624 (initialize_node_lattices): Replaced IPA_NODE_REF with a direct use of
26625 ipa_node_params_sum.
26626 (ipa_context_from_jfunc): Replaced IPA_EDGE_REF with a direct use of
26628 (ipcp_verify_propagated_values): Replaced IPA_NODE_REF with a direct
26629 use of ipa_node_params_sum.
26630 (self_recursively_generated_p): Likewise.
26631 (propagate_scalar_across_jump_function): Likewise.
26632 (propagate_context_across_jump_function): Replaced IPA_EDGE_REF with a
26633 direct use of ipa_edge_args_sum, moved the lookup after the early
26634 exit. Replaced IPA_NODE_REF with a direct use of ipa_node_params_sum.
26635 (propagate_bits_across_jump_function): Replaced IPA_NODE_REF with
26636 direct uses of ipa_node_params_sum.
26637 (propagate_vr_across_jump_function): Likewise.
26638 (propagate_aggregate_lattice): Likewise.
26639 (propagate_aggs_across_jump_function): Likewise.
26640 (propagate_constants_across_call): Likewise, also replaced
26641 IPA_EDGE_REF with a direct use of ipa_edge_args_sum.
26642 (good_cloning_opportunity_p): Replaced IPA_NODE_REF with a direct use
26643 of ipa_node_params_sum.
26644 (estimate_local_effects): Likewise.
26645 (add_all_node_vals_to_toposort): Likewise.
26646 (propagate_constants_topo): Likewise.
26647 (ipcp_propagate_stage): Likewise.
26648 (ipcp_discover_new_direct_edges): Likewise.
26649 (calls_same_node_or_its_all_contexts_clone_p): Likewise.
26650 (cgraph_edge_brings_value_p): Likewise (in both overloaded functions).
26651 (get_info_about_necessary_edges): Likewise.
26652 (want_remove_some_param_p): Likewise.
26653 (create_specialized_node): Likewise.
26654 (self_recursive_pass_through_p): Likewise.
26655 (self_recursive_agg_pass_through_p): Likewise.
26656 (find_more_scalar_values_for_callers_subset): Likewise and also
26657 replaced IPA_EDGE_REF with direct uses of ipa_edge_args_sum, in one
26658 case replacing two of those with a single query.
26659 (find_more_contexts_for_caller_subset): Likewise for the
26660 ipa_polymorphic_call_context overload.
26661 (intersect_aggregates_with_edge): Replaced IPA_EDGE_REF with a direct
26662 use of ipa_edge_args_sum. Replaced IPA_NODE_REF with direct uses of
26663 ipa_node_params_sum.
26664 (find_aggregate_values_for_callers_subset): Likewise, also reusing
26665 results of ipa_edge_args_sum->get.
26666 (cgraph_edge_brings_all_scalars_for_node): Replaced IPA_NODE_REF with
26667 direct uses of ipa_node_params_sum, replaced IPA_EDGE_REF with a
26668 direct use of ipa_edge_args_sum.
26669 (cgraph_edge_brings_all_agg_vals_for_node): Likewise, moved node
26670 summary query after the early exit and reused the result later.
26671 (decide_about_value): Replaced IPA_NODE_REF with a direct use of
26672 ipa_node_params_sum.
26673 (decide_whether_version_node): Likewise. Removed re-querying for
26674 summaries after cloning.
26675 (spread_undeadness): Replaced IPA_NODE_REF with a direct use of
26676 ipa_node_params_sum.
26677 (has_undead_caller_from_outside_scc_p): Likewise, reusing results of
26679 (identify_dead_nodes): Likewise.
26680 (ipcp_store_bits_results): Replaced IPA_NODE_REF with direct uses of
26681 ipa_node_params_sum.
26682 (ipcp_store_vr_results): Likewise.
26683 * ipa-fnsummary.c (evaluate_properties_for_edge): Likewise.
26684 (ipa_fn_summary_t::duplicate): Likewise.
26685 (analyze_function_body): Likewise.
26686 (estimate_calls_size_and_time): Likewise.
26687 (ipa_cached_call_context::duplicate_from): Likewise.
26688 (ipa_call_context::equal_to): Likewise.
26689 (remap_edge_params): Likewise.
26690 (ipa_merge_fn_summary_after_inlining): Likewise.
26691 (inline_read_section): Likewise.
26692 * ipa-icf.c (sem_function::param_used_p): Likewise.
26693 * ipa-modref.c (compute_parm_map): Likewise.
26694 (compute_parm_map): Replaced IPA_EDGE_REF with a direct use of
26696 (get_access_for_fnspec): Replaced IPA_NODE_REF with a direct use of
26697 ipa_node_params_sum and replaced IPA_EDGE_REF with a direct use of
26699 * ipa-profile.c (check_argument_count): Likewise.
26700 * ipa-prop.c (ipa_alloc_node_params): Replaced IPA_NODE_REF_GET_CREATE
26701 with a direct use of ipa_node_params_sum.
26702 (ipa_initialize_node_params): Likewise.
26703 (ipa_print_node_jump_functions_for_edge): Replaced IPA_EDGE_REF with a
26704 direct use of ipa_edge_args_sum and reused the query result.
26705 (ipa_compute_jump_functions_for_edge): Replaced IPA_NODE_REF with a
26706 direct use of ipa_node_params_sum and replaced IPA_EDGE_REF with a
26707 direct use of ipa_edge_args_sum.
26708 (ipa_note_param_call): Replaced IPA_NODE_REF with a direct use of
26709 ipa_node_params_sum and reused the result of the query.
26710 (ipa_analyze_node): Likewise.
26711 (ipa_analyze_controlled_uses): Replaced IPA_NODE_REF with a direct use
26712 of ipa_node_params_sum.
26713 (update_jump_functions_after_inlining): Replaced IPA_EDGE_REF with
26714 direct uses of ipa_edge_args_sum.
26715 (update_indirect_edges_after_inlining): Replaced IPA_NODE_REF with
26716 direct uses of ipa_node_params_sum and replaced IPA_EDGE_REF with a
26717 direct use of ipa_edge_args_sum. Removed superficial re-querying the
26719 (propagate_controlled_uses): Replaced IPA_NODE_REF with direct uses of
26720 ipa_node_params_sum and replaced IPA_EDGE_REF with a direct use of
26722 (ipa_propagate_indirect_call_infos): Replaced IPA_EDGE_REF with a
26723 direct use of ipa_edge_args_sum.
26724 (ipa_edge_args_sum_t::duplicate): Replaced IPA_NODE_REF with a direct
26725 use of ipa_node_params_sum.
26726 (ipa_print_node_params): Likewise.
26727 (ipa_write_node_info): Likewise and also replaced IPA_EDGE_REF with
26728 direct uses of ipa_edge_args_sum.
26729 (ipa_read_edge_info): Replaced IPA_EDGE_REF with a direct use of
26731 (ipa_read_node_info): Replaced IPA_NODE_REF with a direct use of
26732 ipa_node_params_sum.
26733 (ipa_prop_write_jump_functions): Likewise. Move variable node to the
26734 scopes where it is used.
26736 2021-05-10 Uroš Bizjak <ubizjak@gmail.com>
26738 * config/i386/i386-expand.c (ix86_expand_sse_movcc)
26739 <case E_V2SImode>: Force op_true to register.
26741 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
26743 * config/arm/iterators.md (MVE_FP_COMPARISONS): New.
26744 * config/arm/mve.md (mve_vcmp<mve_cmp_op>q_f<mode>)
26745 (mve_vcmp<mve_cmp_op>q_n_f<mode>): New, merge all vcmp_*f*
26747 (mve_vcmpeqq_f<mode>, mve_vcmpeqq_n_f<mode>, mve_vcmpgeq_f<mode>)
26748 (mve_vcmpgeq_n_f<mode>, mve_vcmpgtq_f<mode>)
26749 (mve_vcmpgtq_n_f<mode>, mve_vcmpleq_f<mode>)
26750 (mve_vcmpleq_n_f<mode>, mve_vcmpltq_f<mode>)
26751 (mve_vcmpltq_n_f<mode>, mve_vcmpneq_f<mode>)
26752 (mve_vcmpneq_n_f<mode>): Remove.
26753 * config/arm/unspecs.md (VCMPEQQ_F, VCMPEQQ_N_F, VCMPGEQ_F)
26754 (VCMPGEQ_N_F, VCMPGTQ_F, VCMPGTQ_N_F, VCMPLEQ_F, VCMPLEQ_N_F)
26755 (VCMPLTQ_F, VCMPLTQ_N_F, VCMPNEQ_F, VCMPNEQ_N_F): Remove.
26757 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
26759 * config/arm/iterators.md (MVE_COMPARISONS): New.
26761 (mve_cmp_type): New.
26762 * config/arm/mve.md (mve_vcmp<mve_cmp_op>q_<mode>): New, merge all
26764 (mve_vcmpneq_<mode>, mve_vcmpcsq_n_<mode>, mve_vcmpcsq_<mode>)
26765 (mve_vcmpeqq_n_<mode>, mve_vcmpeqq_<mode>, mve_vcmpgeq_n_<mode>)
26766 (mve_vcmpgeq_<mode>, mve_vcmpgtq_n_<mode>, mve_vcmpgtq_<mode>)
26767 (mve_vcmphiq_n_<mode>, mve_vcmphiq_<mode>, mve_vcmpleq_n_<mode>)
26768 (mve_vcmpleq_<mode>, mve_vcmpltq_n_<mode>, mve_vcmpltq_<mode>)
26769 (mve_vcmpneq_n_<mode>, mve_vcmpltq_n_<mode>, mve_vcmpltq_<mode>)
26770 (mve_vcmpneq_n_<mode>): Remove.
26772 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
26774 * config/arm/arm_mve.h (__arm_vcmp*): Remove 's' suffix.
26775 * config/arm/arm_mve_builtins.def (vcmp*): Remove 's' suffix.
26776 * config/arm/mve.md (mve_vcmp*): Remove 's' suffix in pattern
26779 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
26781 * config/arm/arm_mve_builtins.def (vcmpneq_u): Remove.
26782 (vcmpneq_n_u): Likewise.
26783 (vcmpeqq_u,): Likewise.
26784 (vcmpeqq_n_u): Likewise.
26785 * config/arm/iterators.md (supf): Remove VCMPNEQ_U, VCMPEQQ_U,
26786 VCMPEQQ_N_U and VCMPNEQ_N_U.
26787 * config/arm/mve.md (mve_vcmpneq): Remove <supf> iteration.
26788 (mve_vcmpeqq_n): Likewise.
26789 (mve_vcmpeqq): Likewise.
26790 (mve_vcmpneq_n): Likewise.
26792 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
26794 * config/arm/arm_mve.h (__arm_vcmpeq*u*, __arm_vcmpne*u*): Call
26795 the 's' version of the builtin.
26797 2021-05-10 Richard Biener <rguenther@suse.de>
26799 PR tree-optimization/100492
26800 * tree-loop-distribution.c (find_seed_stmts_for_distribution):
26801 Find nothing when the loop contains an irreducible region.
26803 2021-05-10 Richard Biener <rguenther@suse.de>
26805 PR middle-end/100464
26807 * gimple-fold.c (canonicalize_constructor_val): Do not set
26810 2021-05-10 Richard Biener <rguenther@suse.de>
26812 PR tree-optimization/100434
26813 * tree-ssa-dse.c (initialize_ao_ref_for_dse): Handle
26815 (dse_optimize_stmt): Handle call LHS by dropping the
26816 LHS or the whole call if it doesn't have other
26818 (pass_dse::execute): Adjust.
26820 2021-05-10 Martin Liska <mliska@suse.cz>
26822 * Makefile.in: Add missing genversion rule.
26824 2021-05-10 Alex Coplan <alex.coplan@arm.com>
26827 * config/arm/mve.md (*mve_mov<mode>): Simplify output code. Use
26828 vldrw.u32 and vstrw.32 for V2D[IF]mode loads and stores.
26830 2021-05-10 Martin Liska <mliska@suse.cz>
26832 * builtins.c (is_builtin_name): Use startswith
26833 function instead of strncmp.
26834 * collect2.c (main): Likewise.
26835 (has_lto_section): Likewise.
26836 (scan_libraries): Likewise.
26837 * coverage.c (coverage_checksum_string): Likewise.
26838 (coverage_init): Likewise.
26839 * dwarf2out.c (is_cxx): Likewise.
26840 (gen_compile_unit_die): Likewise.
26841 * gcc-ar.c (main): Likewise.
26842 * gcc.c (init_spec): Likewise.
26843 (read_specs): Likewise.
26844 (execute): Likewise.
26845 (check_live_switch): Likewise.
26846 * genattrtab.c (write_attr_case): Likewise.
26847 (IS_ATTR_GROUP): Likewise.
26848 * gencfn-macros.c (main): Likewise.
26849 * gengtype.c (type_for_name): Likewise.
26850 (gen_rtx_next): Likewise.
26851 (get_file_langdir): Likewise.
26852 (write_local): Likewise.
26853 * genmatch.c (get_operator): Likewise.
26854 (get_operand_type): Likewise.
26855 (expr::gen_transform): Likewise.
26856 * genoutput.c (validate_optab_operands): Likewise.
26857 * incpath.c (add_sysroot_to_chain): Likewise.
26858 * langhooks.c (lang_GNU_C): Likewise.
26859 (lang_GNU_CXX): Likewise.
26860 (lang_GNU_Fortran): Likewise.
26861 (lang_GNU_OBJC): Likewise.
26862 * lto-wrapper.c (run_gcc): Likewise.
26863 * omp-general.c (omp_max_simt_vf): Likewise.
26864 * omp-low.c (omp_runtime_api_call): Likewise.
26865 * opts-common.c (parse_options_from_collect_gcc_options): Likewise.
26866 * read-rtl-function.c (function_reader::read_rtx_operand_r): Likewise.
26867 * real.c (real_from_string): Likewise.
26868 * selftest.c (assert_str_startswith): Likewise.
26869 * timevar.c (timer::validate_phases): Likewise.
26870 * tree.c (get_file_function_name): Likewise.
26871 * ubsan.c (ubsan_use_new_style_p): Likewise.
26872 * varasm.c (default_function_rodata_section): Likewise.
26873 (incorporeal_function_p): Likewise.
26874 (default_section_type_flags): Likewise.
26875 * system.h (startswith): Define startswith.
26877 2021-05-10 Martin Liska <mliska@suse.cz>
26879 * bitmap.h (class auto_bitmap): Remove
26880 __cplusplus >= 201103.
26881 * config/aarch64/aarch64.c: Likewise.
26882 * gimple-ssa-store-merging.c (store_immediate_info::store_immediate_info):
26884 * sbitmap.h: Likewise.
26886 2021-05-10 Martin Liska <mliska@suse.cz>
26888 * Makefile.in: Rename gcov-iov to genversion and depend
26889 on version.h (instead of gcov-iov.h).
26890 * gcov-io.h: Include version.h instread of gcov-iov.h.
26891 * gengtype-state.c (read_state_version): Likewise.
26892 * gcov-iov.c: Moved to...
26893 * genversion.c: ...here.
26894 * lto-streamer.h (LTO_major_version): Define it with
26896 * version.c: Removed.
26897 * version.h: Removed.
26899 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
26901 * config/arc/arc.md (UNSPEC_ARC_DMPYWH): Define.
26902 * config/arc/simdext.md (VCT): Add predicates for iterator
26905 (voptab): Likewise.
26906 (vec_widen_<V_US>mult_hi_v4hi): Change pattern predicate.
26907 (<voptab>v2si3): New patterns.
26909 (reduc_plus_scal_v4hi): Likewise.
26910 (reduc_plus_scal_v2si): Likewise.
26911 (vec_duplicatev2si): Likewise.
26912 (vec_duplicatev4hi): Likewise.
26914 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
26916 * config/arc/simdext.md: Format and cleanup file.
26918 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
26920 * config/arc/simdext.md (movmisalignv2hi): Allow misaligned access
26921 only when munaligned-access option is on.
26922 (movmisalign<mode>): Likewise.
26924 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
26926 * common/config/arc/arc-common.c (arc_handle_option): Remove dot
26928 * config/arc/arc.c (arc_reorg): Remove underscore from string.
26930 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
26932 * config/arc/arc.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
26933 (CTZ_DEFINED_VALUE_AT_ZERO): Likewise.
26934 * config/arc/arc.md (clrsbsi2): Cleanup pattern.
26935 (norm_f): Likewise.
26938 (clzsi2): Use fls instruction when available.
26939 (arc_clzsi2): Likewise.
26941 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
26943 * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add r26 and r27.
26945 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
26947 * doc/extend.texi (__builtin_arc_sr): Swap arguments.
26949 2021-05-10 Bernd Edlinger <bernd.edlinger@hotmail.de>
26951 PR middle-end/100467
26952 * toplev.c (compile_file): Call insn_locations_init before
26953 targetm.asm_out.code_end.
26955 2021-05-07 Andrew Stubbs <ams@codesourcery.com>
26958 2021-05-07 Andrew Stubbs <ams@codesourcery.com>
26960 * config/gcn/gcn.c (gcn_scalar_mode_supported_p): Disable TImode.
26962 2021-05-07 Jakub Jelinek <jakub@redhat.com>
26963 Andrew Stubbs <amd@codesourcery.com>
26966 * builtins.c (try_store_by_multiple_pieces): Use force_operand for
26967 emit_move_insn operands.
26969 2021-05-07 Eric Botcazou <ebotcazou@adacore.com>
26971 * cfgexpand.c (expand_gimple_basic_block): Do not inherit a current
26972 location for the outgoing edges of an empty block.
26973 * dwarf2out.c (add_subscript_info): Retrieve the bounds and index
26974 type by means of the get_array_descr_info langhook, if it is set and
26975 returns true. Remove obsolete code dealing with unnamed subtypes.
26977 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
26979 * gimple-range-cache.cc (ssa_block_ranges): Virtualize.
26980 (sbr_vector): Renamed from ssa_block_cache.
26981 (sbr_vector::sbr_vector): Allocate from obstack abd initialize.
26982 (ssa_block_ranges::~ssa_block_ranges): Remove.
26983 (sbr_vector::set_bb_range): Use varying and undefined cached values.
26984 (ssa_block_ranges::set_bb_varying): Remove.
26985 (sbr_vector::get_bb_range): Adjust assert.
26986 (sbr_vector::bb_range_p): Adjust assert.
26987 (~block_range_cache): No freeing loop required.
26988 (block_range_cache::get_block_ranges): Remove.
26989 (block_range_cache::set_bb_range): Inline get_block_ranges.
26990 (block_range_cache::set_bb_varying): Remove.
26991 * gimple-range-cache.h (set_bb_varying): Remove prototype.
26992 * value-range.h (irange_allocator::get_memory): New.
26994 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
26996 * gimple-range-cache.cc (non_null_ref::non_null_deref_p): Search
26997 dominator tree is available and requested.
26998 (ranger_cache::ssa_range_in_bb): Don't search dom tree here.
26999 (ranger_cache::fill_block_cache): Don't search dom tree here either.
27000 * gimple-range-cache.h (non_null_deref_p): Add dom_search param.
27002 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
27004 * gimple-range.cc (gimple_ranger::range_on_exit): Handle block with
27005 only PHI nodes better.
27007 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
27009 * gimple-range-edge.h (gimple_outgoing_range): Rename from
27011 (gcond_edge_range): Export prototype.
27012 * gimple-range-edge.cc (gcond_edge_range): New.
27013 (gimple_outgoing_range::edge_range_p): Use gcond_edge_range.
27014 * gimple-range-gori.h (gori_compute): Use gimple_outgoing_range.
27016 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
27018 * gimple-range-edge.cc (outgoing_range::calc_switch_ranges): Compute
27019 default range into a temp and allocate only what is needed.
27021 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
27023 * range-op.cc (operator_trunc_mod::wi_fold): x % 0 is UNDEFINED.
27025 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
27027 * gimple-range.h (gimple_range_global): Pick up parameter initial
27028 values, and use-before defined locals are UNDEFINED.
27030 2021-05-07 Eric Botcazou <ebotcazou@adacore.com>
27032 * doc/extend.texi (scalar_storage_order): Mention effect on pointer
27034 * tree.h (reverse_storage_order_for_component_p): Return false if
27035 the type is a pointer.
27037 2021-05-07 Andrew Stubbs <ams@codesourcery.com>
27039 * config/gcn/gcn.c (gcn_scalar_mode_supported_p): Disable TImode.
27041 2021-05-07 Uroš Bizjak <ubizjak@gmail.com>
27044 * config/i386/i386-expand.c (ix86_expand_sse_movcc):
27045 Handle V8QI, V4HI and V2SI modes.
27046 * config/i386/mmx.md (mmx_pblendvb): New insn pattern.
27047 * config/i386/sse.md (unspec): Move UNSPEC_BLENDV ...
27048 * config/i386/i386.md (unspec): ... here.
27050 2021-05-07 Tobias Burnus <tobias@codesourcery.com>
27051 Tom de Vries <tdevries@suse.de>
27053 * omp-low.c (lower_rec_simd_input_clauses): Set max_vf = 1 if
27054 a truth_value_p reduction variable is nonintegral.
27056 2021-05-07 Uroš Bizjak <ubizjak@gmail.com>
27059 * config/i386/i386-expand.c (ix86_use_mask_cmp_p):
27060 Return false for mode sizes < 16.
27062 2021-05-07 Jakub Jelinek <jakub@redhat.com>
27065 * config/i386/mmx.md (*xop_pcmov_<mode>): New define_insn.
27067 2021-05-06 Martin Jambor <mjambor@suse.cz>
27069 * ipa-sra.c (ipa_sra_dump_all_summaries): Dump edge summaries even
27070 when there is no function summary.
27071 (ipa_sra_summarize_function): produce edge summaries even when
27074 2021-05-06 Tom Tromey <tom@tromey.com>
27076 * godump.c (string_hash_eq): Remove.
27077 (go_finish): Use htab_eq_string.
27079 2021-05-06 Tom Tromey <tom@tromey.com>
27081 * gengtype-state.c (read_state): Use htab_eq_string.
27082 (string_eq): Remove.
27084 2021-05-06 Tom Tromey <tom@tromey.com>
27086 * gensupport.c (htab_eq_string): Remove.
27088 2021-05-06 Bernd Edlinger <bernd.edlinger@hotmail.de>
27091 * debug.h (gcc_debug_hooks): Add set_ignored_loc function pointer.
27092 * dwarf2out.h (dw_fde_node::ignored_debug): New data item.
27093 * dbxout.c (dbx_debug_hooks, xcoff_debug_hooks): Add dummy
27094 set_ignored_loc callbacks.
27095 * debug.c (do_nothing_debug_hooks): Likewise.
27096 * vmsdbgout.c (vmsdbg_debug_hooks): Likewise.
27097 * dwarf2out.c (text_section_used, cold_text_section_used): Remove.
27098 (in_text_section_p, last_text_label, last_cold_label,
27099 switch_text_ranges, switch_cold_ranges): New data items.
27100 (dwarf2out_note_section_used): Remove.
27101 (dwarf2out_begin_prologue): Set fde->ignored_debug and
27103 (mark_ignored_debug_section): New helper function.
27104 (dwarf2out_end_epilogue, dwarf2out_switch_text_section): Call
27105 mark_ignored_debug_section.
27106 (dwarf2_debug_hooks): Use dwarf2out_set_ignored_loc.
27107 (dwarf2_lineno_debug_hooks): Use dummy for set_ignored_loc.
27108 (size_of_aranges): Adjust formula for multi-part text ranges size.
27109 (output_aranges): Output multi-part text ranges.
27110 (dwarf2out_set_ignored_loc): New callback function.
27111 (dwarf2out_finish): Output multi-part text ranges.
27112 (dwarf2out_c_finalize): Clear new data items.
27113 * final.c (final_start_function_1): Call set_ignored_loc callback.
27114 (final_scan_insn_1): Likewise.
27115 * ggc-page.c (gt_ggc_mx): New helper function.
27116 * stringpool.c (gt_pch_nx): Likewise.
27118 2021-05-06 Richard Biener <rguenther@suse.de>
27120 * timevar.def (TV_TREE_INSERT_PHI_NODES): Remove.
27121 (TV_TREE_SSA_REWRITE_BLOCKS): Likewise.
27122 (TV_TREE_INTO_SSA): New.
27123 * tree-into-ssa.c (insert_phi_nodes): Do not account separately.
27124 (rewrite_blocks): Likewise.
27125 (pass_data_build_ssa): Account to TV_TREE_INTO_SSA.
27127 2021-05-06 Jakub Jelinek <jakub@redhat.com>
27129 * tree-ssa-phiopt.c (value_replacement, minmax_replacement,
27130 abs_replacement, xor_replacement,
27131 cond_removal_in_popcount_clz_ctz_pattern,
27132 replace_phi_edge_with_variable): Change type of phi argument from
27133 gimple * to gphi *.
27135 2021-05-06 Richard Biener <rguenther@suse.de>
27137 * tree-ssa-loop-split.c (split_loop): Delay updating SSA form.
27138 Output an opt-info message.
27139 (do_split_loop_on_cond): Likewise.
27140 (tree_ssa_split_loops): Update SSA form here.
27142 2021-05-06 Richard Biener <rguenther@suse.de>
27144 * tree-inline.c (tree_function_versioning): Fix DECL_BY_REFERENCE
27145 return variable removal.
27147 2021-05-06 Marius Hillenbrand <mhillen@linux.ibm.com>
27149 * config/s390/s390-builtins.def (O_M5, O1_M5, ...): Remove unused macros.
27150 (s390_vec_permi_s64, s390_vec_permi_b64, s390_vec_permi_u64)
27151 (s390_vec_permi_dbl, s390_vpdi): Use the O3_U2 type for the immediate
27153 * config/s390/s390.c (s390_const_operand_ok): Remove unused
27156 2021-05-06 Jakub Jelinek <jakub@redhat.com>
27158 PR tree-optimization/94589
27159 * tree-ssa-phiopt.c (tree_ssa_phiopt_worker): Call
27160 spaceship_replacement.
27161 (cond_only_block_p, spaceship_replacement): New functions.
27163 2021-05-06 Richard Biener <rguenther@suse.de>
27166 * tree-emutls.c (gen_emutls_addr): Pass in whether we're
27167 dealing with a debug use and only query existing addresses
27169 (lower_emutls_1): Avoid splitting out addresses for debug
27170 stmts, reset the debug stmt when we fail to find existing
27172 (lower_emutls_phi_arg): Set wi.stmt.
27174 2021-05-06 Christoph Muellner <cmuellner@gcc.gnu.org>
27177 * config/riscv/riscv.c (riscv_block_move_loop): Use cbranch helper.
27178 * config/riscv/riscv.md (cbranch<mode>4): Generate helpers.
27179 (stack_protect_test): Use cbranch helper.
27181 2021-05-05 Eric Botcazou <ebotcazou@adacore.com>
27184 * config/i386/i386.c (ix86_compute_frame_layout): For a SEH target,
27185 always return the establisher frame for __builtin_frame_address (0).
27187 2021-05-05 Ivan Sorokin <vanyacpp@gmail.com>
27190 * config/i386/i386-builtins.c (ix86_cpu_model_type_node): New.
27191 (ix86_cpu_model_var): Likewise.
27192 (ix86_cpu_features2_type_node): Likewise.
27193 (ix86_cpu_features2_var): Likewise.
27194 (fold_builtin_cpu): Cache __cpu_model and __cpu_features2 with
27197 2021-05-05 Martin Sebor <msebor@redhat.com>
27199 * passes.def (pass_warn_printf): Run after SSA.
27201 2021-05-05 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
27203 * config/arm/neon.md (neon_vtst_combine<mode>): New pattern.
27204 * config/arm/predicates.md (minus_one_operand): New predicate.
27206 2021-05-05 Jeff Law <jlaw@tachyum.com>
27208 * config/avr/avr.md: Remove references to CC_STATUS_INIT.
27210 2021-05-05 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
27212 PR rtl-optimization/100263
27213 * postreload.c (move2add_valid_value_p): Ensure register can
27216 2021-05-05 Eric Botcazou <ebotcazou@adacore.com>
27218 PR rtl-optimization/100411
27219 * cfgcleanup.c (try_crossjump_to_edge): Also skip end of prologue
27220 and beginning of function markers.
27222 2021-05-05 Jeff Law <jlaw@tachyum.com>
27224 * config/cr16/cr16.h (NOTICE_UPDATE_CC): Remove.
27225 * config/cr16/cr16.c (notice_update_cc): Remove.
27226 * config/cr16/cr16-protos.h (notice_update_cc): Remove.
27228 2021-05-05 Uroš Bizjak <ubizjak@gmail.com>
27231 * config/i386/i386-expand.c (ix86_expand_int_sse_cmp):
27232 Handle V8QI, V4HI and V2SI modes.
27233 * config/i386/i386.c (ix86_build_const_vector): Handle V2SImode.
27234 (ix86_build_signbit_mask): Ditto.
27235 * config/i386/mmx.md (MMXMODE14): New mode iterator.
27236 (<smaxmin:code><MMXMODE14:mode>3): New expander.
27237 (*mmx_<smaxmin:code><MMXMODE14:mode>3): New insn pattern.
27238 (<umaxmin:code><MMXMODE24:mode>3): New expander.
27239 (*mmx_<umaxmin:code><MMXMODE24:mode>3): New insn pattern.
27240 (vec_cmp<MMXMODEI:mode><MMXMODEI:mode>): New expander.
27241 (vec_cmpu<MMXMODEI:mode><MMXMODEI:mode>): Ditto.
27242 (vcond<MMXMODEI:mode><MMXMODEI:mode>): Ditto.
27243 (vcondu<MMXMODEI:mode><MMXMODEI:mode>): Ditto.
27244 (vcond_mask_<MMXMODEI:mode><MMXMODEI:mode>): Ditto.
27246 2021-05-05 Eric Botcazou <ebotcazou@adacore.com>
27248 * dwarf2out.c (loc_list_from_tree_1) <DECL>: During early DWARF, do
27249 not expand the VALUE_EXPR of variables put in the non-local frame.
27250 * gimplify.c (gimplify_type_sizes) <RECORD_TYPE>: If the type is not
27251 to be ignored for debug info, ensure its variable offsets are not.
27253 2021-05-05 Richard Biener <rguenther@suse.de>
27255 PR tree-optimization/79333
27256 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_stmt):
27257 Fold stmt following SSA edges.
27259 2021-05-05 Richard Biener <rguenther@suse.de>
27261 PR middle-end/100394
27262 * calls.c (expand_call): Preserve possibly throwing calls.
27263 * cfgexpand.c (expand_call_stmt): When a call can throw signal
27264 RTL expansion there are side-effects.
27265 * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Simplify,
27266 mark all possibly throwing stmts necessary unless we can elide
27268 * tree-ssa-dse.c (pass_dse::execute): Preserve exceptions unless
27269 -fdelete-dead-exceptions.
27270 * tree.h (DECL_PURE_P): Add note about exceptions.
27272 2021-05-05 Alexandre Oliva <oliva@adacore.com>
27274 * config/i386/vxworks.h (DBX_REGISTER_NUMBER): Make it
27277 2021-05-04 David Edelsohn <dje.gcc@gmail.com>
27279 * config/rs6000/rs6000-call.c (rs6000_output_mi_thunk): Use
27280 get_fnname_from_decl for name of thunk.
27281 * config/rs6000/rs6000.c (rs6000_declare_alias): Use assemble_name
27282 and ASM_OUTPUT_LABEL.
27283 (rs6000_xcoff_declare_function_name): Use assemble_name and
27285 (rs6000_xcoff_declare_object_name): Use ASM_OUTPUT_LABEL.
27286 (rs6000_xcoff_encode_section_info): Don't add mapping class
27287 for aliases. Always add [DS] mapping class to primary
27289 (rs6000_asm_weaken_decl): Don't explicitly add [DS].
27291 2021-05-04 Martin Sebor <msebor@redhat.com>
27293 PR middle-end/100307
27294 * builtins.c (compute_objsize_r): Clear base0 for pointers.
27296 2021-05-04 Jeff Law <jlaw@tachyum.com>
27298 * config/bfin/bfin.h (NOTICE_UPDATE_CC): Remove.
27300 2021-05-04 Segher Boessenkool <segher@kernel.crashing.org>
27302 * caller-save.c: Remove CC0.
27303 * cfgcleanup.c: Remove CC0.
27304 * cfgrtl.c: Remove CC0.
27305 * combine.c: Remove CC0.
27306 * compare-elim.c: Remove CC0.
27307 * conditions.h: Remove CC0.
27308 * config/h8300/h8300.h: Remove CC0.
27309 * config/h8300/h8300-protos.h: Remove CC0.
27310 * config/h8300/peepholes.md: Remove CC0.
27311 * config/i386/x86-tune-sched.c: Remove CC0.
27312 * config/m68k/m68k.c: Remove CC0.
27313 * config/rl78/rl78.c: Remove CC0.
27314 * config/sparc/sparc.c: Remove CC0.
27315 * config/xtensa/xtensa.c: Remove CC0.
27316 (gen_conditional_move): Use pc_rtx instead of cc0_rtx in a piece of
27317 RTL where that is used as a placeholder only.
27318 * cprop.c: Remove CC0.
27319 * cse.c: Remove CC0.
27320 * cselib.c: Remove CC0.
27321 * df-problems.c: Remove CC0.
27322 * df-scan.c: Remove CC0.
27323 * doc/md.texi: Remove CC0. Adjust an example.
27324 * doc/rtl.texi: Remove CC0. Adjust an example.
27325 * doc/tm.texi: Regenerate.
27326 * doc/tm.texi.in: Remove CC0.
27327 * emit-rtl.c: Remove CC0.
27328 * final.c: Remove CC0.
27329 * fwprop.c: Remove CC0.
27330 * gcse-common.c: Remove CC0.
27331 * gcse.c: Remove CC0.
27332 * genattrtab.c: Remove CC0.
27333 * genconfig.c: Remove CC0.
27334 * genemit.c: Remove CC0.
27335 * genextract.c: Remove CC0.
27336 * gengenrtl.c: Remove CC0.
27337 * genrecog.c: Remove CC0.
27338 * haifa-sched.c: Remove CC0.
27339 * ifcvt.c: Remove CC0.
27340 * ira-costs.c: Remove CC0.
27341 * ira.c: Remove CC0.
27342 * jump.c: Remove CC0.
27343 * loop-invariant.c: Remove CC0.
27344 * lra-constraints.c: Remove CC0.
27345 * lra-eliminations.c: Remove CC0.
27346 * optabs.c: Remove CC0.
27347 * postreload-gcse.c: Remove CC0.
27348 * postreload.c: Remove CC0.
27349 * print-rtl.c: Remove CC0.
27350 * read-rtl-function.c: Remove CC0.
27351 * reg-notes.def: Remove CC0.
27352 * reg-stack.c: Remove CC0.
27353 * reginfo.c: Remove CC0.
27354 * regrename.c: Remove CC0.
27355 * reload.c: Remove CC0.
27356 * reload1.c: Remove CC0.
27357 * reorg.c: Remove CC0.
27358 * resource.c: Remove CC0.
27359 * rtl.c: Remove CC0.
27360 * rtl.def: Remove CC0.
27361 * rtl.h: Remove CC0.
27362 * rtlanal.c: Remove CC0.
27363 * sched-deps.c: Remove CC0.
27364 * sched-rgn.c: Remove CC0.
27365 * shrink-wrap.c: Remove CC0.
27366 * simplify-rtx.c: Remove CC0.
27367 * system.h: Remove CC0. Poison NOTICE_UPDATE_CC, CC_STATUS_MDEP_INIT,
27368 CC_STATUS_MDEP, and CC_STATUS.
27369 * target.def: Remove CC0.
27370 * valtrack.c: Remove CC0.
27371 * var-tracking.c: Remove CC0.
27373 2021-05-04 Richard Biener <rguenther@suse.de>
27375 PR tree-optimization/100414
27376 * tree-ssa-phiopt.c (get_non_trapping): Do not compute dominance
27378 (tree_ssa_phiopt_worker): But unconditionally here.
27380 2021-05-04 Tobias Burnus <tobias@codesourcery.com>
27382 * omp-low.c (lower_rec_input_clauses, lower_reduction_clauses): Handle
27383 && and || with floating-point and complex arguments.
27385 2021-05-04 Eric Botcazou <ebotcazou@adacore.com>
27387 * tree-inline.c (insert_debug_decl_map): Delete.
27388 (copy_debug_stmt): Minor tweak.
27389 (setup_one_parameter): Do not use a variable if the value is either
27390 a read-only DECL or a non-addressable local variable in the caller.
27391 In this case, insert the debug-only variable in the map manually.
27392 (expand_call_inline): Do not generate a CLOBBER for these values.
27393 * tree-inline.h (debug_map): Minor tweak.
27395 2021-05-04 Eric Botcazou <ebotcazou@adacore.com>
27397 * builtins.c (builtin_with_linkage_p): Return true for stp[n]cpy.
27398 * symtab.c (symtab_node::output_to_lto_symbol_table_p): Tidy up.
27400 2021-05-04 Richard Biener <rguenther@suse.de>
27402 PR tree-optimization/100329
27403 * tree-ssa-reassoc.c (can_reassociate_p): Do not reassociate
27405 (insert_stmt_after): Assert we're not running into asm goto.
27407 2021-05-04 Richard Biener <rguenther@suse.de>
27409 PR tree-optimization/100398
27410 * tree-ssa-dse.c (pass_dse::execute): Preserve control
27413 2021-05-04 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
27415 * builtins.c (try_store_by_multiple_pieces): Fix constfun's prototype.
27417 2021-05-04 Alexandre Oliva <oliva@adacore.com>
27419 * builtins.c (try_store_by_multiple_pieces): New.
27420 (expand_builtin_memset_args): Use it. If target_char_cast
27421 fails, proceed as for non-constant val. Pass len's ctz to...
27422 * expr.c (clear_storage_hints): ... this. Try store by
27423 multiple pieces after setmem.
27424 (clear_storage): Adjust.
27425 * expr.h (clear_storage_hints): Likewise.
27426 (try_store_by_multiple_pieces): Declare.
27427 * passes.def: Replace the last copy_prop with ccp.
27429 2021-05-03 Tom de Vries <tdevries@suse.de>
27432 * omp-low.c (lower_rec_input_clauses): Disable SIMT for user-defined
27435 2021-05-03 Richard Biener <rguenther@suse.de>
27437 * tree-ssa-dse.c (dse_classify_store): Track two PHI defs.
27439 2021-05-03 Richard Biener <rguenther@suse.de>
27441 * tree-ssa-dse.c: Do not include domwalk.h but cfganal.h.
27442 (dse_dom_walker): Remove.
27443 (dse_dom_walker::dse_optimize_stmt): Rename...
27444 (dse_optimize_stmt): ... to this, pass in live_bytes sbitmap.
27445 (dse_dom_walker::before_dom_children): Inline ...
27446 (pass_dse::execute): ... here. Perform a reverse program
27449 2021-05-03 H.J. Lu <hjl.tools@gmail.com>
27452 * configure: Regenerated.
27454 2021-05-03 Ilya Leoshkevich <iii@linux.ibm.com>
27457 * config/s390/s390.c (s390_hard_fp_reg_p): New function.
27458 (s390_md_asm_adjust): Handle hard registers.
27460 2021-05-03 Jakub Jelinek <jakub@redhat.com>
27462 PR tree-optimization/100382
27463 * tree-ssa-dse.c: Include tree-eh.h.
27464 (dse_dom_walker::before_dom_children): Don't remove stmts if
27465 stmt_unremovable_because_of_non_call_eh_p is true.
27467 2021-05-02 David Edelsohn <dje.gcc@gmail.com>
27469 * varasm.c (compute_reloc_for_var): Split out from...
27470 (get_variable_section): Use it.
27471 * output.h (compute_reloc_for_var): Declare.
27472 * config/rs6000/rs6000-protos.h
27473 (rs6000_xcoff_asm_output_aligned_decl_common): Change alignment to
27475 * config/rs6000/rs6000.c (rs6000_legitimize_tls_address_aix):
27476 Don't append storage mapping class to symbol.
27477 (rs6000_xcoff_asm_named_section): Add BS and UL mapping classes.
27478 Don't convert TLS BSS to common.
27479 (rs6000_xcoff_unique_section): Don't fall back to select_secton.
27480 (rs6000_xcoff_section_type_flags): Add SECTION_BSS if DECL is
27482 (rs6000_xcoff_asm_globalize_decl_name): Don't strip storage
27484 (rs6000_xcoff_asm_output_aligned_decl_common): Align is unsigned int.
27485 If align is 0 from TLS class, use the same rules as varasm.c
27486 If not common, switch to BSS section manually.
27487 If common, emit appropriate comm or lcomm directive.
27488 (rs6000_xcoff_encode_section_info): Add logic to append all
27489 storage mapping classes.
27490 (rs6000_asm_weaken_decl): Adjust for qualname symbols.
27491 * config/rs6000/xcoff.h (ASM_OUTPUT_ALIGNED_DECL_LOCAL): Use
27492 rs6000_xcoff_asm_output_aligned_decl_common.
27493 (ASM_OUTPUT_ALIGNED_DECL_LOCAL): Use
27494 rs6000_xcoff_asm_output_aligned_decl_common.
27495 (ASM_OUTPUT_TLS_COMMON): Use
27496 rs6000_xcoff_asm_output_aligned_decl_common.
27498 2021-05-02 Jakub Jelinek <jakub@redhat.com>
27501 * config/nvptx/nvptx.c (nvptx_sese_pseudo): Use nullptr instead of 0
27502 as first argument of pseudo_node_t constructors.
27504 2021-05-02 Jakub Jelinek <jakub@redhat.com>
27507 * config/i386/t-i386 (TM_H): Add $(srcdir)/config/i386/i386-isa.def.
27509 2021-05-01 Aldy Hernandez <aldyh@redhat.com>
27511 * value-range.cc (DEFINE_INT_RANGE_GC_STUBS): Remove.
27512 (gt_pch_nx (int_range<1> *&)): New.
27513 (gt_ggc_mx (int_range<1> *&)): New.
27514 * value-range.h (class irange): Add GTY support for
27517 2021-05-01 Geng Qi <gengqi@linux.alibaba.com>
27519 * doc/options.texi (Negative): Change either or to both and.
27521 2021-04-30 Jonathan Wright <jonathan.wright@arm.com>
27523 * config/aarch64/aarch64-simd-builtins.def: Add
27524 float_ml[as][q]_laneq builtin generator macros.
27525 * config/aarch64/aarch64-simd.md (mul_laneq<mode>3): Define.
27526 (aarch64_float_mla_laneq<mode>): Define.
27527 (aarch64_float_mls_laneq<mode>): Define.
27528 * config/aarch64/arm_neon.h (vmla_laneq_f32): Use RTL builtin
27529 instead of GCC vector extensions.
27530 (vmlaq_laneq_f32): Likewise.
27531 (vmls_laneq_f32): Likewise.
27532 (vmlsq_laneq_f32): Likewise.
27534 2021-04-30 Jonathan Wright <jonathan.wright@arm.com>
27536 * config/aarch64/aarch64-simd-builtins.def: Add
27537 float_ml[as]_lane builtin generator macros.
27538 * config/aarch64/aarch64-simd.md (*aarch64_mul3_elt<mode>):
27540 (mul_lane<mode>3): This, and re-order arguments.
27541 (aarch64_float_mla_lane<mode>): Define.
27542 (aarch64_float_mls_lane<mode>): Define.
27543 * config/aarch64/arm_neon.h (vmla_lane_f32): Use RTL builtin
27544 instead of GCC vector extensions.
27545 (vmlaq_lane_f32): Likewise.
27546 (vmls_lane_f32): Likewise.
27547 (vmlsq_lane_f32): Likewise.
27549 2021-04-30 Jonathan Wright <jonathan.wright@arm.com>
27551 * config/aarch64/aarch64-simd-builtins.def: Add float_ml[as]
27552 builtin generator macros.
27553 * config/aarch64/aarch64-simd.md (aarch64_float_mla<mode>):
27555 (aarch64_float_mls<mode>): Define.
27556 * config/aarch64/arm_neon.h (vmla_f32): Use RTL builtin
27557 instead of relying on GCC vector extensions.
27558 (vmla_f64): Likewise.
27559 (vmlaq_f32): Likewise.
27560 (vmlaq_f64): Likewise.
27561 (vmls_f32): Likewise.
27562 (vmls_f64): Likewise.
27563 (vmlsq_f32): Likewise.
27564 (vmlsq_f64): Likewise.
27565 * config/aarch64/iterators.md: Define VDQF_DF mode iterator.
27567 2021-04-30 Jonathan Wright <jonathan.wright@arm.com>
27569 * config/aarch64/aarch64-simd-builtins.def: Add
27570 float_ml[as]_n_builtin generator macros.
27571 * config/aarch64/aarch64-simd.md (*aarch64_mul3_elt_from_dup<mode>):
27573 (mul_n<mode>3): This, and re-order arguments.
27574 (aarch64_float_mla_n<mode>): Define.
27575 (aarch64_float_mls_n<mode>): Define.
27576 * config/aarch64/arm_neon.h (vmla_n_f32): Use RTL builtin
27577 instead of inline asm.
27578 (vmlaq_n_f32): Likewise.
27579 (vmls_n_f32): Likewise.
27580 (vmlsq_n_f32): Likewise.
27582 2021-04-30 Jonathan Wright <joanthan.wright@arm.com>
27584 * config/aarch64/aarch64-simd-builtins.def: Add pmull[2]
27585 builtin generator macros.
27586 * config/aarch64/aarch64-simd.md (aarch64_pmullv8qi): Define.
27587 (aarch64_pmull_hiv16qi_insn): Define.
27588 (aarch64_pmull_hiv16qi): Define.
27589 * config/aarch64/arm_neon.h (vmull_high_p8): Use RTL builtin
27590 instead of inline asm.
27591 (vmull_p8): Likewise.
27593 2021-04-30 Senthil Kumar Selvaraj <saaadhu@gcc.gnu.org>
27595 * config/avr/avr.md: Adjust peepholes to match and
27596 generate parallels with clobber of REG_CC.
27597 (mov<mode>_insn): Rename to mov<mode>_insn_split.
27598 (*mov<mode>_insn): Rename to mov<mode>_insn.
27600 2021-04-30 David Edelsohn <dje.gcc@gmail.com>
27602 * varasm.c (use_blocks_for_decl_p): Don't use section anchors
27603 for VAR_DECLs if -fdata-sections enabled.
27605 2021-04-30 Michael Meissner <meissner@linux.ibm.com>
27607 PR bootstrap/100327
27608 * config/rs6000/rs6000.c
27609 (TARGET_LIBGCC_FLOATING_MODE_SUPPORTED_P): Define.
27610 (rs6000_libgcc_floating_mode_supported_p): New target hook.
27612 2021-04-30 Aldy Hernandez <aldyh@redhat.com>
27614 * tree-ssa-threadbackward.c (class thread_jumps): Split out code
27616 (class back_threader_registry): ...to here...
27617 (class back_threader_profitability): ...and here...
27618 (thread_jumps::thread_through_all_blocks): Remove argument.
27619 (back_threader_registry::back_threader_registry): New.
27620 (back_threader_registry::~back_threader_registry): New.
27621 (back_threader_registry::thread_through_all_blocks): New.
27622 (thread_jumps::profitable_jump_thread_path): Move from here...
27623 (back_threader_profitability::profitable_path_p): ...to here.
27624 (thread_jumps::find_taken_edge): New.
27625 (thread_jumps::convert_and_register_current_path): Move...
27626 (back_threader_registry::register_path): ...to here.
27627 (thread_jumps::register_jump_thread_path_if_profitable): Move...
27628 (thread_jumps::maybe_register_path): ...to here.
27629 (thread_jumps::handle_phi): Call find_taken_edge and
27630 maybe_register_path.
27631 (thread_jumps::handle_assignment): Same.
27632 (thread_jumps::fsm_find_control_statement_thread_paths): Remove
27633 tree argument to handle_phi and handle_assignment.
27634 (thread_jumps::find_jump_threads_backwards): Set m_name. Remove
27635 set of m_speed_p and m_max_threaded_paths.
27636 (pass_thread_jumps::execute): Remove second argument from
27637 find_jump_threads_backwards.
27638 (pass_early_thread_jumps::execute): Same.
27640 2021-04-30 Aldy Hernandez <aldyh@redhat.com>
27642 * tree-ssa-dom.c (class dom_jump_threader_simplifier): New.
27643 (class dom_opt_dom_walker): Initialize some class variables.
27644 (pass_dominator::execute): Pass evrp_range_analyzer and
27645 dom_jump_threader_simplifier to dom_opt_dom_walker.
27646 Adjust for some functions moving into classes.
27647 (simplify_stmt_for_jump_threading): Adjust and move to...
27648 (jump_threader_simplifier::simplify): ...here.
27649 (dom_opt_dom_walker::before_dom_children): Adjust for
27650 m_evrp_range_analyzer.
27651 (dom_opt_dom_walker::after_dom_children): Remove x_vr_values hack.
27652 (test_for_singularity): Place in dom_opt_dom_walker class.
27653 (dom_opt_dom_walker::optimize_stmt): The argument
27654 evrp_range_analyzer is now a class field.
27655 * tree-ssa-threadbackward.c (class thread_jumps): Add m_registry.
27656 (thread_jumps::thread_through_all_blocks): New.
27657 (thread_jumps::convert_and_register_current_path): Use m_registry.
27658 (pass_thread_jumps::execute): Adjust for thread_through_all_blocks
27659 being in the threader class.
27660 (pass_early_thread_jumps::execute): Same.
27661 * tree-ssa-threadedge.c (threadedge_initialize_values): Move...
27662 (jump_threader::jump_threader): ...here.
27663 (threadedge_finalize_values): Move...
27664 (jump_threader::~jump_threader): ...here.
27665 (jump_threader::remove_jump_threads_including): New.
27666 (jump_threader::thread_through_all_blocks): New.
27667 (record_temporary_equivalences_from_phis): Move...
27668 (jump_threader::record_temporary_equivalences_from_phis): ...here.
27669 (record_temporary_equivalences_from_stmts_at_dest): Move...
27670 (jump_threader::record_temporary_equivalences_from_stmts_at_dest):
27672 (simplify_control_stmt_condition_1): Move to jump_threader class.
27673 (simplify_control_stmt_condition): Move...
27674 (jump_threader::simplify_control_stmt_condition): ...here.
27675 (thread_around_empty_blocks): Move...
27676 (jump_threader::thread_around_empty_blocks): ...here.
27677 (thread_through_normal_block): Move...
27678 (jump_threader::thread_through_normal_block): ...here.
27679 (thread_across_edge): Move...
27680 (jump_threader::thread_across_edge): ...here.
27681 (thread_outgoing_edges): Move...
27682 (jump_threader::thread_outgoing_edges): ...here.
27683 * tree-ssa-threadedge.h: Move externally facing functings...
27684 (class jump_threader): ...here...
27685 (class jump_threader_simplifier): ...and here.
27686 * tree-ssa-threadupdate.c (struct redirection_data): Remove comment.
27687 (jump_thread_path_allocator::jump_thread_path_allocator): New.
27688 (jump_thread_path_allocator::~jump_thread_path_allocator): New.
27689 (jump_thread_path_allocator::allocate_thread_edge): New.
27690 (jump_thread_path_allocator::allocate_thread_path): New.
27691 (jump_thread_path_registry::jump_thread_path_registry): New.
27692 (jump_thread_path_registry::~jump_thread_path_registry): New.
27693 (jump_thread_path_registry::allocate_thread_edge): New.
27694 (jump_thread_path_registry::allocate_thread_path): New.
27695 (dump_jump_thread_path): Make extern.
27696 (debug (const vec<jump_thread_edge *> &path)): New.
27697 (struct removed_edges): Move to tree-ssa-threadupdate.h.
27698 (struct thread_stats_d): Remove.
27699 (remove_ctrl_stmt_and_useless_edges): Make static.
27700 (lookup_redirection_data): Move...
27701 (jump_thread_path_registry::lookup_redirection_data): ...here.
27702 (ssa_redirect_edges): Make static.
27703 (thread_block_1): Move...
27704 (jump_thread_path_registry::thread_block_1): ...here.
27705 (thread_block): Move...
27706 (jump_thread_path_registry::thread_block): ...here.
27707 (thread_through_loop_header): Move...
27708 (jump_thread_path_registry::thread_through_loop_header): ...here.
27709 (mark_threaded_blocks): Move...
27710 (jump_thread_path_registry::mark_threaded_blocks): ...here.
27711 (debug_path): Move...
27712 (jump_thread_path_registry::debug_path): ...here.
27713 (debug_all_paths): Move...
27714 (jump_thread_path_registry::dump): ..here.
27715 (rewire_first_differing_edge): Move...
27716 (jump_thread_path_registry::rewire_first_differing_edge): ...here.
27717 (adjust_paths_after_duplication): Move...
27718 (jump_thread_path_registry::adjust_paths_after_duplication): ...here.
27719 (duplicate_thread_path): Move...
27720 (jump_thread_path_registry::duplicate_thread_path): ..here.
27721 (remove_jump_threads_including): Move...
27722 (jump_thread_path_registry::remove_jump_threads_including): ...here.
27723 (thread_through_all_blocks): Move to...
27724 (jump_thread_path_registry::thread_through_all_blocks): ...here.
27725 (delete_jump_thread_path): Remove.
27726 (register_jump_thread): Move...
27727 (jump_thread_path_registry::register_jump_thread): ...here.
27728 * tree-ssa-threadupdate.h: Move externally facing functions...
27729 (class jump_thread_path_allocator): ...here...
27730 (class jump_thread_path_registry): ...and here.
27731 (thread_through_all_blocks): Remove.
27732 (struct removed_edges): New.
27733 (register_jump_thread): Remove.
27734 (remove_jump_threads_including): Remove.
27735 (delete_jump_thread_path): Remove.
27736 (remove_ctrl_stmt_and_useless_edges): Remove.
27737 (free_dom_edge_info): New prototype.
27738 * tree-vrp.c: Remove x_vr_values hack.
27739 (class vrp_jump_threader_simplifier): New.
27740 (vrp_jump_threader_simplifier::simplify): New.
27741 (vrp_jump_threader::vrp_jump_threader): Adjust method signature.
27742 Remove m_dummy_cond.
27743 Instantiate m_simplifier and m_threader.
27744 (vrp_jump_threader::thread_through_all_blocks): New.
27745 (vrp_jump_threader::simplify_stmt): Remove.
27746 (vrp_jump_threader::after_dom_children): Do not set m_dummy_cond.
27747 Remove x_vr_values hack.
27748 (execute_vrp): Adjust for thread_through_all_blocks being in a
27751 2021-04-30 Christophe Lyon <christophe.lyon@linaro.org>
27753 * genflags.c (gen_insn): Print failed expansion string.
27755 2021-04-30 H.J. Lu <hjl.tools@gmail.com>
27757 * expr.c (alignment_for_piecewise_move): Call mode_for_size
27758 without limit to MAX_FIXED_MODE_SIZE.
27760 2021-04-30 H.J. Lu <hjl.tools@gmail.com>
27762 PR middle-end/90773
27763 * builtins.c (builtin_memset_gen_str): Don't use return from
27764 simplify_gen_subreg.
27766 2021-04-30 Uroš Bizjak <ubizjak@gmail.com>
27769 * config/i386/i386.md (*add<mode>3_carry_0r): New insn pattern.
27770 (*addsi3_carry_zext_0r): Ditto.
27771 (*sub<mode>3_carry_0): Ditto.
27772 (*subsi3_carry_zext_0r): Ditto.
27773 * config/i386/predicates.md (ix86_carry_flag_unset_operator):
27775 * config/i386/i386.c (ix86_rtx_costs) <case PLUS, case MINUS>:
27776 Also consider ix86_carry_flag_unset_operator to calculate
27777 the cost of adc/sbb insn.
27779 2021-04-30 Roman Zhuykov <zhroma@ispras.ru>
27781 PR rtl-optimization/100225
27782 PR rtl-optimization/84878
27783 * modulo-sched.c (sms_schedule): Use note_stores to skip loops
27784 where we have an instruction which touches (writes) any hard
27785 register from df->regular_block_artificial_uses set.
27786 Allow not-single-set instruction only right before basic block
27789 2021-04-30 Geng Qi <gengqi@linux.alibaba.com>
27791 * config/riscv/riscv.opt (march=,mabi=): Negative itself.
27793 2021-04-30 LevyHsu <admin@levyhsu.com>
27795 * config/riscv/riscv.c (riscv_min_arithmetic_precision): New.
27796 * config/riscv/riscv.h (TARGET_MIN_ARITHMETIC_PRECISION): New.
27797 * config/riscv/riscv.md (addv<mode>4, uaddv<mode>4): New.
27798 (subv<mode>4, usubv<mode>4, mulv<mode>4, umulv<mode>4): New.
27800 2021-04-29 Alexandre Oliva <oliva@adacore.com>
27802 * config.gcc: Merged x86 and x86_64 cpu_type-setting cases.
27804 2021-04-29 Alexandre Oliva <oliva@adacore.com>
27806 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Rename to...
27807 (ASM_OUTPUT_MAX_SKIP_ALIGN): ... this. Enclose in do/while(0).
27808 * config/i386/i386.c: Adjust.
27809 * config/i386/i386.md: Adjust.
27810 * config/i386/darwin.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Drop.
27811 * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
27812 * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
27813 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
27814 * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
27815 * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
27816 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
27817 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
27818 * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
27819 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
27820 (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
27822 2021-04-29 Uroš Bizjak <ubizjak@gmail.com>
27824 * config/i386/i386-expand.c (ix86_expand_int_compare):
27825 Swap operands of GTU and LEU comparison to emit carry flag comparison.
27826 * config/i386/i386.md (*add<mode>3_carry_0): Change insn
27827 predicate to allow more combine opportunities with memory operands.
27828 (*sub<mode>3_carry_0): Ditto.
27830 2021-04-29 Richard Sandiford <richard.sandiford@arm.com>
27832 PR rtl-optimization/100303
27833 * rtl-ssa/accesses.cc (function_info::make_use_available): Take a
27834 boolean that indicates whether the use will only be used in
27835 debug instructions. Treat it in the same way that existing
27836 cross-EBB debug references would be handled if so.
27837 (function_info::make_uses_available): Likewise.
27838 * rtl-ssa/functions.h (function_info::make_uses_available): Update
27839 prototype accordingly.
27840 (function_info::make_uses_available): Likewise.
27841 * fwprop.c (try_fwprop_subst): Update call accordingly.
27843 2021-04-29 Jeff Law <jlaw@tachyum.com>
27845 * config/nios2/nios2-protos.h (nios2_fpu_insn_enabled): Move outside
27848 2021-04-29 Uroš Bizjak <ubizjak@gmail.com>
27849 Richard Biener <rguenther@suse.de>
27852 * config/i386/i386-builtin.def (IX86_BUILTIN_MASKLOADPD)
27853 (IX86_BUILTIN_MASKLOADPS, IX86_BUILTIN_MASKLOADPD256)
27854 (IX86_BUILTIN_MASKLOADPS256, IX86_BUILTIN_MASKLOADD)
27855 (IX86_BUILTIN_MASKLOADQ, IX86_BUILTIN_MASKLOADD256)
27856 (IX86_BUILTIN_MASKLOADQ256): Move from SPECIAL_ARGS
27857 to PURE_ARGS category.
27858 * config/i386/i386-builtins.c (ix86_init_mmx_sse_builtins):
27859 Handle PURE_ARGS category.
27860 * config/i386/i386-expand.c (ix86_expand_builtin): Ditto.
27862 2021-04-29 Eric Botcazou <ebotcazou@adacore.com>
27864 * configure.ac: Check for the presence of sys/locking.h header and
27865 for whether _LK_LOCK is supported by _locking.
27866 * configure: Regenerate.
27867 * config.in: Likewise.
27868 * gcov-io.h: Define GCOV_LOCKED_WITH_LOCKING if HOST_HAS_LK_LOCK.
27869 * gcov-io.c (gcov_open): Add support for GCOV_LOCKED_WITH_LOCKING.
27870 * system.h: Include <sys/locking.h> if HAVE_SYS_LOCKING_H.
27872 2021-04-29 Uroš Bizjak <ubizjak@gmail.com>
27874 * config/i386/predicates.md (fcmov_comparison_operator):
27875 Do not check for trivial FP comparison operator.
27876 <case GEU, case LTU>: Allow CCGZmode.
27877 <case GTU, case LEU>: Do not allow CCCmode.
27878 (ix86_comparison_operator) <case GTU, case LEU>: Allow only CCmode.
27879 (ix86_carry_flag_operator): Match only LTU and UNLT code.
27880 Do not check for trivial FP comparison operator. Allow CCGZmode.
27882 2021-04-29 Tom de Vries <tdevries@suse.de>
27884 * omp-expand.c (expand_omp_simd): Add step_orig, and replace uses of
27885 fd->loop.step by either step or orig_step.
27887 2021-04-29 Eric Botcazou <ebotcazou@adacore.com>
27889 * config/sparc/sparc.c (gen_load_pcrel_sym): Delete.
27890 (load_got_register): Do the PIC dance here.
27891 (sparc_legitimize_tls_address): Simplify.
27892 (sparc_emit_probe_stack_range): Likewise.
27893 (sparc32_initialize_trampoline): Likewise.
27894 (sparc64_initialize_trampoline): Likewise.
27895 * config/sparc/sparc.md (load_pcrel_sym<P:mode>): Add @ marker.
27896 (probe_stack_range<P:mode>): Likewise.
27897 (flush<P:mode>): Likewise.
27898 (tgd_hi22<P:mode>): Likewise.
27899 (tgd_lo10<P:mode>): Likewise.
27900 (tgd_add<P:mode>): Likewise.
27901 (tgd_call<P:mode>): Likewise.
27902 (tldm_hi22<P:mode>): Likewise.
27903 (tldm_lo10<P:mode>): Likewise.
27904 (tldm_add<P:mode>): Likewise.
27905 (tldm_call<P:mode>): Likewise.
27906 (tldo_hix22<P:mode>): Likewise.
27907 (tldo_lox10<P:mode>): Likewise.
27908 (tldo_add<P:mode>): Likewise.
27909 (tie_hi22<P:mode>): Likewise.
27910 (tie_lo10<P:mode>): Likewise.
27911 (tie_add<P:mode>): Likewise.
27912 (tle_hix22<P:mode>): Likewise.
27913 (tle_lox10<P:mode>): Likewise.
27914 (stack_protect_setsi): Rename to...
27915 (stack_protect_set32): ...this.
27916 (stack_protect_setdi): Rename to...
27917 (stack_protect_set64): ...this.
27918 (stack_protect_set): Adjust calls to above.
27919 (stack_protect_testsi): Rename to...
27920 (stack_protect_test32): ...this.
27921 (stack_protect_testdi): Rename to...
27922 (stack_protect_test64): ...this.
27923 (stack_protect_test): Adjust calls to above.
27925 2021-04-29 H.J. Lu <hjl.tools@gmail.com>
27927 PR middle-end/90773
27928 * builtins.c (builtin_memcpy_read_str): Add a dummy argument.
27929 (builtin_strncpy_read_str): Likewise.
27930 (builtin_memset_read_str): Add an argument for the previous RTL
27931 information and generate the new RTL from the previous RTL info.
27932 (builtin_memset_gen_str): Likewise.
27933 * builtins.h (builtin_strncpy_read_str): Update the prototype.
27934 (builtin_memset_read_str): Likewise.
27935 * expr.c (by_pieces_ninsns): If targetm.overlap_op_by_pieces_p()
27936 returns true, round up size and alignment to the widest integer
27937 mode for maximum size.
27938 (pieces_addr::adjust): Add a pointer to by_pieces_prev argument
27939 and pass it to m_constfn.
27940 (op_by_pieces_d): Add m_push and m_overlap_op_by_pieces.
27941 (op_by_pieces_d::op_by_pieces_d): Add a bool argument to
27942 initialize m_push. Initialize m_overlap_op_by_pieces with
27943 targetm.overlap_op_by_pieces_p ().
27944 (op_by_pieces_d::run): Pass the previous RTL information to
27945 pieces_addr::adjust and generate overlapping operations if
27946 m_overlap_op_by_pieces is true.
27948 (move_by_pieces_d::move_by_pieces_d): Updated for op_by_pieces_d
27950 (store_by_pieces_d::store_by_pieces_d): Updated for op_by_pieces_d
27952 (can_store_by_pieces): Use by_pieces_constfn on constfun.
27953 (store_by_pieces): Use by_pieces_constfn on constfun. Updated
27954 for op_by_pieces_d change.
27955 (clear_by_pieces_1): Add a dummy argument.
27956 (clear_by_pieces): Updated for op_by_pieces_d change.
27957 (compare_by_pieces_d::compare_by_pieces_d): Likewise.
27958 (string_cst_read_str): Add a dummy argument.
27959 * expr.h (by_pieces_constfn): Add a dummy argument.
27960 (by_pieces_prev): New.
27961 * target.def (overlap_op_by_pieces_p): New target hook.
27962 * config/i386/i386.c (TARGET_OVERLAP_OP_BY_PIECES_P): New.
27963 * doc/tm.texi.in: Add TARGET_OVERLAP_OP_BY_PIECES_P.
27964 * doc/tm.texi: Regenerated.
27966 2021-04-29 Richard Biener <rguenther@suse.de>
27968 PR tree-optimization/100253
27969 * tree-vect-stmts.c (vectorizable_load): Do not assume
27970 element alignment when DR_MISALIGNMENT is -1.
27971 (vectorizable_store): Likewise.
27973 2021-04-29 Jakub Jelinek <jakub@redhat.com>
27976 * config/aarch64/aarch64.c (aarch64_add_offset_1_temporaries): Use
27977 absu_hwi instead of abs_hwi.
27979 2021-04-29 Richard Biener <rguenther@suse.de>
27981 PR middle-end/38474
27982 * tree-ssa-structalias.c (add_graph_edge): Avoid direct
27983 forwarding when indirect forwarding through ESCAPED
27986 2021-04-29 Tom de Vries <tdevries@suse.de>
27989 * internal-fn.c (expand_GOMP_SIMT_ENTER_ALLOC)
27990 (expand_GOMP_SIMT_LAST_LANE, expand_GOMP_SIMT_ORDERED_PRED)
27991 (expand_GOMP_SIMT_VOTE_ANY, expand_GOMP_SIMT_XCHG_BFLY)
27992 (expand_GOMP_SIMT_XCHG_IDX): Ensure target is assigned to.
27994 2021-04-29 Richard Biener <rguenther@suse.de>
27996 PR tree-optimization/99912
27997 * tree-ssa-dse.c (dse_dom_walker::m_need_cfg_cleanup): New.
27998 (dse_dom_walker::todo): Likewise.
27999 (dse_dom_walker::dse_optimize_stmt): Move VDEF check to the
28001 (dse_dom_walker::before_dom_children): Remove trivially
28002 dead SSA defs and schedule CFG cleanup if we removed all
28004 (pass_dse::execute): Get TODO as computed by the DOM walker
28005 and return it. Wipe dominator info earlier.
28007 2021-04-29 Richard Biener <rguenther@suse.de>
28010 * ipa-prop.c (ipcp_modif_dom_walker::before_dom_children):
28011 Track blocks to cleanup EH in new m_need_eh_cleanup.
28012 (ipcp_modif_dom_walker::cleanup_eh): New.
28013 (ipcp_transform_function): Release dominator info before
28016 2021-04-29 Martin Sebor <msebor@redhat.com>
28018 PR middle-end/100250
28019 * attribs.c (attr_access::array_as_string): Avoid dereferencing
28020 a pointer when it's null.
28022 2021-04-29 Martin Sebor <msebor@redhat.com>
28024 * Makefile.in (OBJS): Add ipa-free-lang-data.o.
28025 * ipa-free-lang-data.cc: New file.
28026 * tree.c: Move pass free_lang_data to file above.
28027 (build_array_type_1): Declare extern.
28028 * tree.h (build_array_type_1): Declare.
28030 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
28032 * config/aarch64/aarch64-simd-builtins.def: Modify comment to
28033 make consistent with updated RTL pattern.
28034 * config/aarch64/aarch64-simd.md (aarch64_<sur>qmovn<mode>):
28035 Implement using ss_truncate and us_truncate rather than
28037 * config/aarch64/iterators.md: Remove redundant unspecs and
28038 iterator: UNSPEC_[SU]QXTN and SUQMOVN respectively.
28040 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
28042 * config/aarch64/arm_acle.h (__attribute__): Make intrinsic
28043 attributes consistent with those defined in arm_neon.h.
28045 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
28047 * config/aarch64/arm_fp16.h (__attribute__): Make intrinsic
28048 attributes consistent with those defined in arm_neon.h.
28050 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
28052 * config/aarch64/aarch64-simd-builtins.def: Add
28053 float_trunc_rodd builtin generator macros.
28054 * config/aarch64/aarch64-simd.md (aarch64_float_trunc_rodd_df):
28056 (aarch64_float_trunc_rodd_lo_v2sf): Define.
28057 (aarch64_float_trunc_rodd_hi_v4sf_le): Define.
28058 (aarch64_float_trunc_rodd_hi_v4sf_be): Define.
28059 (aarch64_float_trunc_rodd_hi_v4sf): Define.
28060 * config/aarch64/arm_neon.h (vcvtx_f32_f64): Use RTL builtin
28061 instead of inline asm.
28062 (vcvtx_high_f32_f64): Likewise.
28063 (vcvtxd_f32_f64): Likewise.
28064 * config/aarch64/iterators.md: Add FCVTXN unspec.
28066 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
28068 * config/aarch64/aarch64-simd-builtins.def: Add tbx1 builtin
28070 * config/aarch64/aarch64-simd.md (aarch64_tbx1<mode>):
28072 * config/aarch64/arm_neon.h (vqtbx1_s8): USE RTL builtin
28073 instead of inline asm.
28074 (vqtbx1_u8): Likewise.
28075 (vqtbx1_p8): Likewise.
28076 (vqtbx1q_s8): Likewise.
28077 (vqtbx1q_u8): Likewise.
28078 (vqtbx1q_p8): Likewise.
28079 (vtbx2_s8): Likewise.
28080 (vtbx2_u8): Likewise.
28081 (vtbx2_p8): Likewise.
28083 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
28085 * config/aarch64/aarch64-simd-builtins.def: Add tbl1 builtin
28087 * config/aarch64/arm_neon.h (vqtbl1_p8): Use RTL builtin
28088 instead of inline asm.
28089 (vqtbl1_s8): Likewise.
28090 (vqtbl1_u8): Likewise.
28091 (vqtbl1q_p8): Likewise.
28092 (vqtbl1q_s8): Likewise.
28093 (vqtbl1q_u8): Likewise.
28094 (vtbl1_s8): Likewise.
28095 (vtbl1_u8): Likewise.
28096 (vtbl1_p8): Likewise.
28097 (vtbl2_s8): Likewise.
28098 (vtbl2_u8): Likewise.
28099 (vtbl2_p8): Likewise.
28101 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
28103 * config/aarch64/aarch64-simd-builtins.def: Add polynomial
28104 ssri_n buitin generator macro.
28105 * config/aarch64/arm_neon.h (vsri_n_p8): Use RTL builtin
28106 instead of inline asm.
28107 (vsri_n_p16): Likewise.
28108 (vsri_n_p64): Likewise.
28109 (vsriq_n_p8): Likewise.
28110 (vsriq_n_p16): Likewise.
28111 (vsriq_n_p64): Likewise.
28113 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
28115 * config/aarch64/aarch64-simd-builtins.def: Use VALLP mode
28116 iterator for polynomial ssli_n builtin generator macro.
28117 * config/aarch64/arm_neon.h (vsli_n_p8): Use RTL builtin
28118 instead of inline asm.
28119 (vsli_n_p16): Likewise.
28120 (vsliq_n_p8): Likewise.
28121 (vsliq_n_p16): Likewise.
28122 * config/aarch64/iterators.md: Define VALLP mode iterator.
28124 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
28126 * config/aarch64/aarch64-simd-builtins.def: Use VDQV_L
28127 iterator to generate [su]adalp RTL builtins.
28128 * config/aarch64/aarch64-simd.md: Use VDQV_L iterator in
28129 [su]adalp RTL pattern.
28130 * config/aarch64/arm_neon.h (vpadal_s32): Use RTL builtin
28131 instead of inline asm.
28132 (vpadal_u32): Likewise.
28134 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
28136 * config/aarch64/aarch64-simd-builtins.def: Add [su]addlp
28137 builtin generator macros.
28138 * config/aarch64/aarch64-simd.md (aarch64_<su>addlp<mode>):
28140 * config/aarch64/arm_neon.h (vpaddl_s8): Use RTL builtin
28141 instead of inline asm.
28142 (vpaddl_s16): Likewise.
28143 (vpaddl_s32): Likewise.
28144 (vpaddl_u8): Likewise.
28145 (vpaddl_u16): Likewise.
28146 (vpaddl_u32): Likewise.
28147 (vpaddlq_s8): Likewise.
28148 (vpaddlq_s16): Likewise.
28149 (vpaddlq_s32): Likewise.
28150 (vpaddlq_u8): Likewise.
28151 (vpaddlq_u16): Likewise.
28152 (vpaddlq_u32): Liwewise.
28153 * config/aarch64/iterators.md: Define [SU]ADDLP unspecs with
28154 appropriate attributes.
28156 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
28158 * config/aarch64/aarch64-simd-builtins.def: Use VDQ_I iterator
28159 for aarch64_addp<mode> builtin macro generator.
28160 * config/aarch64/aarch64-simd.md: Use VDQ_I iterator in
28161 aarch64_addp<mode> RTL pattern.
28162 * config/aarch64/arm_neon.h (vpaddq_s8): Use RTL builtin
28163 instead of inline asm.
28164 (vpaddq_s16): Likewise.
28165 (vpaddq_s32): Likewise.
28166 (vpaddq_s64): Likewise.
28167 (vpaddq_u8): Likewise.
28168 (vpaddq_u16): Likewise.
28169 (vpaddq_u32): Likewise.
28170 (vpaddq_u64): Likewise.
28172 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
28174 * config/aarch64/aarch64-simd-builtins.def: Add sq[r]dmulh_n
28175 builtin generator macros.
28176 * config/aarch64/aarch64-simd.md (aarch64_sq<r>dmulh_n<mode>):
28178 * config/aarch64/arm_neon.h (vqdmulh_n_s16): Use RTL builtin
28179 instead of inline asm.
28180 (vqdmulh_n_s32): Likewise.
28181 (vqdmulhq_n_s16): Likewise.
28182 (vqdmulhq_n_s32): Likewise.
28183 (vqrdmulh_n_s16): Likewise.
28184 (vqrdmulh_n_s32): Likewise.
28185 (vqrdmulhq_n_s16): Likewise.
28186 (vqrdmulhq_n_s32): Likewise.
28188 2021-04-28 Tobias Burnus <tobias@codesourcery.com>
28190 * doc/install.texi (--enable-offload-defaulted): Document.
28192 2021-04-28 Senthil Kumar Selvaraj <saaadhu@gcc.gnu.org>
28194 * config/avr/avr-dimode.md: Turn existing patterns into
28195 define_insn_and_split style patterns where the splitter
28196 adds a clobber of the condition code register. Drop "cc"
28197 attribute. Add new patterns to match output of
28199 * config/avr/avr-fixed.md: Likewise.
28200 * config/avr/avr.c (cc_reg_rtx): New.
28201 (avr_parallel_insn_from_insns): Adjust insn count
28202 for removal of set of cc0.
28203 (avr_is_casesi_sequence): Likewise.
28204 (avr_casei_sequence_check_operands): Likewise.
28205 (avr_optimize_casesi): Likewise. Also insert
28206 new insns after jump_insn.
28207 (avr_pass_casesi::avr_rest_of_handle_casesi): Adjust
28208 for removal of set of cc0.
28209 (avr_init_expanders): Initialize cc_reg_rtx.
28210 (avr_regno_reg_class): Handle REG_CC.
28211 (cond_string): Remove usage of CC_OVERFLOW_UNUSABLE.
28212 (avr_notice_update_cc): Remove function.
28213 (ret_cond_branch): Remove usage of CC_OVERFLOW_UNUSABLE.
28214 (compare_condition): Adjust for PARALLEL with
28216 (out_shift_with_cnt): Likewise.
28217 (ashlhi3_out): Likewise.
28218 (ashrhi3_out): Likewise.
28219 (lshrhi3_out): Likewise.
28220 (avr_class_max_nregs): Return single reg for REG_CC.
28221 (avr_compare_pattern): Check for REG_CC instead
28223 (avr_reorg_remove_redundant_compare): Likewise.
28224 (avr_reorg):Adjust for PARALLEL with REG_CC clobber.
28225 (avr_hard_regno_nregs): Return single reg for REG_CC.
28226 (avr_hard_regno_mode_ok): Allow only CCmode for REG_CC.
28227 (avr_md_asm_adjust): Clobber REG_CC.
28228 (TARGET_HARD_REGNO_NREGS): Define.
28229 (TARGET_CLASS_MAX_NREGS): Define.
28230 (TARGET_MD_ASM_ADJUST): Define.
28231 * config/avr/avr.h (FIRST_PSEUDO_REGISTER): Adjust
28233 (enum reg_class): Add CC_REG class.
28234 (NOTICE_UPDATE_CC): Remove.
28235 (CC_OVERFLOW_UNUSABLE): Remove.
28236 (CC_NO_CARRY): Remove.
28237 * config/avr/avr.md: Turn existing patterns into
28238 define_insn_and_split style patterns where the splitter
28239 adds a clobber of the condition code register. Drop "cc"
28240 attribute. Add new patterns to match output of
28242 (sez): Remove unused pattern.
28244 2021-04-28 Richard Earnshaw <rearnsha@arm.com>
28247 * config/arm/arm.c (arm_hard_regno_mode_ok): Only allow VPR to be
28250 2021-04-28 Richard Sandiford <richard.sandiford@arm.com>
28253 * config/aarch64/constraints.md (Utq): Require the address to
28254 be valid for both the element mode and for V2DImode.
28256 2021-04-28 Jakub Jelinek <jakub@redhat.com>
28257 Tobias Burnus <tobias@codesourcery.com>
28259 * configure.ac (OFFLOAD_DEFAULTED): AC_DEFINE if offload-defaulted.
28260 * gcc.c (process_command): New variable.
28261 (driver::maybe_putenv_OFFLOAD_TARGETS): If OFFLOAD_DEFAULTED,
28262 set it if -foffload is defaulted.
28263 * lto-wrapper.c (OFFLOAD_TARGET_DEFAULT_ENV): Define.
28264 (compile_offload_image): If OFFLOAD_DEFAULTED and
28265 OFFLOAD_TARGET_DEFAULT is in the environment, don't fail
28266 if corresponding mkoffload can't be found.
28267 (compile_images_for_offload_targets): Likewise. Free and clear
28268 offload_names if no valid offload is found.
28269 * config.in: Regenerate.
28270 * configure: Regenerate.
28272 2021-04-28 Richard Biener <rguenther@suse.de>
28274 PR tree-optimization/100292
28275 * tree-vect-generic.c (expand_vector_condition): Do not fold
28278 2021-04-27 David Edelsohn <dje.gcc@gmail.com>
28280 * config/rs6000/aix.h (SUBTARGET_DRIVER_SELF_SPECS): New.
28281 * config/rs6000/aix64.opt (m64): New.
28284 2021-04-27 Maciej W. Rozycki <macro@orcam.me.uk>
28286 * config/vax/vax.c (print_operand_address, vax_address_cost_1)
28287 (index_term_p): Handle ASHIFT too.
28289 2021-04-27 Maciej W. Rozycki <macro@orcam.me.uk>
28291 * config/vax/builtins.md (jbb<ccss>i<mode>): Remove operand #3.
28292 (sync_lock_test_and_set<mode>): Adjust accordingly.
28293 (sync_lock_release<mode>): Likewise.
28295 2021-04-27 Maciej W. Rozycki <macro@orcam.me.uk>
28297 * config/vax/vax-protos.h (adjacent_operands_p): Remove
28299 * config/vax/vax.c (adjacent_operands_p): Remove.
28301 2021-04-27 Maciej W. Rozycki <macro@linux-mips.org>
28303 * ifcvt.c (dead_or_predicable) [!IFCVT_MODIFY_TESTS]: Fall
28304 through to the non-conditional execution case if getting the
28305 condition for conditional execution has failed.
28307 2021-04-27 Richard Sandiford <richard.sandiford@arm.com>
28309 PR middle-end/100284
28310 * gimple.c (gimple_could_trap_p_1): Remove VEC_COND_EXPR test.
28311 * tree-eh.c (operation_could_trap_p): Handle VEC_COND_EXPR rather
28312 than asserting on it.
28314 2021-04-27 David Edelsohn <dje.gcc@gmail.com>
28316 * config/rs6000/rs6000.c (rs6000_aix_precompute_tls_p): Protect
28317 with TARGET_AIX_OS.
28319 2021-04-27 David Edelsohn <dje.gcc@gmail.com>
28322 * calls.c (precompute_register_parameters): Additionally test
28323 targetm.precompute_tls_p to pre-compute argument.
28324 * config/rs6000/aix.h (TARGET_PRECOMPUTE_TLS_P): Define.
28325 * config/rs6000/rs6000.c (rs6000_aix_precompute_tls_p): New.
28326 * target.def (precompute_tls_p): New.
28327 * doc/tm.texi.in (TARGET_PRECOMPUTE_TLS_P): Add hook documentation.
28328 * doc/tm.texi: Regenerated.
28330 2021-04-27 Jakub Jelinek <jakub@redhat.com>
28333 * config/aarch64/aarch64.c (aarch64_print_operand): Cast -UINTVAL
28334 back to HOST_WIDE_INT.
28336 2021-04-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
28339 * simplify-rtx.c (simplify_context::simplify_subreg): Check the
28340 memory alignment for the outer mode.
28342 2021-04-27 H.J. Lu <hjl.tools@gmail.com>
28344 PR middle-end/90773
28345 * expr.c (op_by_pieces_d::get_usable_mode): New member function.
28346 (op_by_pieces_d::run): Cange a while loop to a do-while loop.
28348 2021-04-27 Alex Coplan <alex.coplan@arm.com>
28351 * config/arm/arm.c (arm_split_compare_and_swap): Fix up codegen
28352 with negative immediates: ensure we expand cbranchsi4_scratch
28353 correctly and ensure we satisfy its constraints.
28354 * config/arm/sync.md
28355 (@atomic_compare_and_swap<CCSI:arch><NARROW:mode>_1): Don't
28356 attempt to tie two output operands together with constraints;
28357 collapse two alternatives.
28358 (@atomic_compare_and_swap<CCSI:arch><SIDI:mode>_1): Likewise.
28359 * config/arm/thumb1.md (cbranchsi4_neg_late): New.
28361 2021-04-27 Jakub Jelinek <jakub@redhat.com>
28364 * config/aarch64/predicates.md (aarch64_sub_immediate,
28365 aarch64_plus_immediate): Use -UINTVAL instead of -INTVAL.
28366 * config/aarch64/aarch64.md (casesi, rotl<mode>3): Likewise.
28367 * config/aarch64/aarch64.c (aarch64_print_operand,
28368 aarch64_split_atomic_op, aarch64_expand_subvti): Likewise.
28370 2021-04-27 Jakub Jelinek <jakub@redhat.com>
28372 PR tree-optimization/100239
28373 * tree-vect-generic.c (lower_vec_perm): Don't accept constant
28374 permutations with all indices from the first zero element as vec_shl.
28376 2021-04-27 Jakub Jelinek <jakub@redhat.com>
28378 PR rtl-optimization/100254
28379 * cfgcleanup.c (outgoing_edges_match): Check REG_EH_REGION on
28380 last1 and last2 insns rather than BB_END (bb1) and BB_END (bb2) insns.
28382 2021-04-27 Richard Biener <rguenther@suse.de>
28384 PR tree-optimization/99912
28385 * passes.def: Add comment about new TODO_remove_unused_locals.
28386 * tree-stdarg.c (pass_data_stdarg): Run TODO_remove_unused_locals
28389 2021-04-27 Richard Biener <rguenther@suse.de>
28391 PR tree-optimization/99912
28392 * passes.def (pass_all_optimizations): Add pass_dse before
28393 the first pass_dce, move the first pass_dse before the
28394 pass_dce following pass_pre.
28396 2021-04-27 Jakub Jelinek <jakub@redhat.com>
28398 PR tree-optimization/95527
28399 * generic-match-head.c: Include tm.h.
28400 * gimple-match-head.c: Include tm.h.
28401 * match.pd (CLZ == INTEGER_CST): Don't use
28402 #ifdef CLZ_DEFINED_VALUE_AT_ZERO, only test CLZ_DEFINED_VALUE_AT_ZERO
28403 if clz == CFN_CLZ. Add missing val declaration.
28404 (CTZ cmp CST): New simplifications.
28406 2021-04-27 Jakub Jelinek <jakub@redhat.com>
28408 PR tree-optimization/96696
28409 * expr.c (expand_expr_divmod): New function.
28410 (expand_expr_real_2) <case TRUNC_DIV_EXPR>: Use it for truncations and
28411 divisions. Formatting fixes.
28412 <case MULT_EXPR>: Optimize x / y * y as x - x % y if the latter is
28415 2021-04-27 Martin Jambor <mjambor@suse.cz>
28418 * ipa-param-manipulation.c (ipa_param_adjustments::modify_call):
28419 If removing a call statement LHS SSA name, release it.
28421 2021-04-27 Richard Earnshaw <rearnsha@arm.com>
28424 * config/arm/arm.c (THUMB2_WORK_REGS): Check PIC_OFFSET_TABLE_REGNUM
28425 is valid before including it in the mask.
28427 2021-04-27 Richard Sandiford <richard.sandiford@arm.com>
28430 * config/aarch64/aarch64.c (aarch64_comp_type_attributes): Handle
28433 2021-04-27 Richard Biener <rguenther@suse.de>
28435 PR tree-optimization/100051
28436 * tree-ssa-alias.c (indirect_ref_may_alias_decl_p): Add
28437 disambiguator based on access size vs. decl size.
28439 2021-04-27 Richard Biener <rguenther@suse.de>
28441 PR tree-optimization/100278
28442 * tree-ssa-pre.c (compute_avail): Give up when we cannot
28443 adjust TBAA beacuse of mismatching bases.
28445 2021-04-27 Jakub Jelinek <jakub@redhat.com>
28448 * config/i386/i386.md (*<insn><mode>3_mask, *<insn><mode>3_mask_1):
28449 For any_rotate define_insn_split and following splitters, use
28450 SWI iterator instead of SWI48.
28452 2021-04-27 Richard Biener <rguenther@suse.de>
28454 PR tree-optimization/99776
28455 * match.pd (bit_field_ref (ctor)): Relax element extract
28456 type compatibility checks.
28458 2021-04-27 Cui,Lili <lili.cui@intel.com>
28460 * common/config/i386/i386-common.c (processor_names):
28461 Sync processor_names with processor_type.
28462 * config/i386/i386-options.c (processor_cost_table):
28463 Sync processor_cost_table with processor_type.
28465 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
28467 * value-range.cc (irange::irange_set_1bit_anti_range): Add assert.
28468 (irange::set): Call irange_set_1bit_anti_range for handling all
28469 1-bit ranges. Fall through on ~[MIN,MAX].
28471 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
28473 * value-range.cc (irange::legacy_num_pairs): Remove.
28474 (irange::invert): Change gcc_assert to gcc_checking_assert.
28475 * value-range.h (irange::num_pairs): Adjust for a cached
28476 num_pairs(). Also, rename all gcc_assert's to
28477 gcc_checking_assert's.
28479 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
28481 * value-range.cc (irange::operator=): Set m_kind.
28482 (irange::copy_to_legacy): Handle varying and undefined sources
28483 as a legacy copy since they can be easily copied.
28484 (irange::irange_set): Set m_kind.
28485 (irange::irange_set_anti_range): Same.
28486 (irange::set): Rename normalize_min_max to normalize_kind.
28487 (irange::verify_range): Adjust for multi-ranges having the
28489 (irange::irange_union): Set m_kind.
28490 (irange::irange_intersect): Same.
28491 (irange::invert): Same.
28492 * value-range.h (irange::kind): Always return m_kind.
28493 (irange::varying_p): Rename to...
28494 (irange::varying_comptaible_p): ...this.
28495 (irange::undefined_p): Only look at m_kind.
28496 (irange::irange): Always set VR_UNDEFINED if applicable.
28497 (irange::set_undefined): Always set VR_UNDEFINED.
28498 (irange::set_varying): Always set m_kind to VR_VARYING.
28499 (irange::normalize_min_max): Rename to...
28500 (irange::normalize_kind): ...this.
28502 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
28504 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::set_ssa_range_info):
28505 Adjust for constant_p including varying_p.
28506 * tree-vrp.c (vrp_prop::finalize): Same.
28507 (determine_value_range): Same.
28508 * vr-values.c (vr_values::range_of_expr): Same.
28509 * value-range.cc (irange::symbolic_p): Do not check varying_p.
28510 (irange::constant_p): Same.
28512 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
28514 * value-range.cc (irange::legacy_lower_bound): Replace
28515 !undefined_p check with num_ranges > 0.
28516 (irange::legacy_upper_bound): Same.
28517 * value-range.h (irange::type): Same.
28518 (irange::lower_bound): Same.
28519 (irange::upper_bound): Same.
28521 2021-04-26 Richard Biener <rguenther@suse.de>
28523 PR tree-optimization/99956
28524 * gimple-loop-interchange.cc (compute_access_stride):
28525 Try instantiating the access in a shallower loop nest
28526 if instantiating failed.
28527 (compute_access_strides): Pass adjustable loop_nest
28528 to compute_access_stride.
28530 2021-04-26 Christophe Lyon <christophe.lyon@linaro.org>
28532 * doc/sourcebuild.texi (arm_cmse_hw): Document.
28534 2021-04-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
28536 * config/aarch64/iterators.md (vwcore): Handle V4BF, V8BF.
28538 2021-04-26 Thomas Schwinge <thomas@codesourcery.com>
28539 Nathan Sidwell <nathan@codesourcery.com>
28540 Tom de Vries <vries@codesourcery.com>
28541 Julian Brown <julian@codesourcery.com>
28542 Kwok Cheung Yeung <kcy@codesourcery.com>
28544 * omp-offload.c (oacc_validate_dims): Implement
28545 '-Wopenacc-parallelism'.
28546 * doc/invoke.texi (-Wopenacc-parallelism): Document.
28548 2021-04-26 Richard Biener <rguenther@suse.de>
28550 * tree-cfg.h (gimplify_build1): Remove.
28551 (gimplify_build2): Likewise.
28552 (gimplify_build3): Likewise.
28553 * tree-cfg.c (gimplify_build1): Move to tree-vect-generic.c.
28554 (gimplify_build2): Likewise.
28555 (gimplify_build3): Likewise.
28556 * tree-vect-generic.c (gimplify_build1): Move from tree-cfg.c.
28558 (gimplify_build2): Likewise.
28559 (gimplify_build3): Likewise.
28560 (tree_vec_extract): Use resimplify with following SSA edges.
28561 (expand_vector_parallel): Avoid passing NULL size/bitpos
28562 to tree_vec_extract.
28563 * expr.c (store_constructor): Deal with zero-element CTORs.
28564 * match.pd (bit_field_ref <vector CTOR>): Make sure to
28565 produce vector constants when possible.
28567 2021-04-26 Richard Biener <rguenther@suse.de>
28569 * tree-complex.c: Include gimple-fold.h.
28570 (expand_complex_addition): Use gimple_build.
28571 (expand_complex_multiplication_components): Likewise.
28572 (expand_complex_multiplication): Likewise.
28573 (expand_complex_div_straight): Likewise.
28574 (expand_complex_div_wide): Likewise.
28575 (expand_complex_division): Likewise.
28576 (expand_complex_conjugate): Likewise.
28577 (expand_complex_comparison): Likewise.
28579 2021-04-26 Richard Biener <rguenther@suse.de>
28581 * tree-ssa-phiopt.c (two_value_replacement): Remove use
28582 of legacy gimplify_buildN API.
28584 2021-04-26 Richard Biener <rguenther@suse.de>
28586 PR tree-optimization/99473
28587 * tree-ssa-phiopt.c (cond_store_replacement): Handle all
28590 2021-04-26 Richard Biener <rguenther@suse.de>
28592 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_builtin):
28593 Use replace_call_with_value.
28595 2021-04-26 Richard Biener <rguenther@suse.de>
28597 * tree-ssa-propagate.h (valid_gimple_rhs_p): Remove.
28598 (update_gimple_call): Likewise.
28599 (update_call_from_tree): Likewise.
28600 * tree-ssa-propagate.c (valid_gimple_rhs_p): Remove.
28601 (valid_gimple_call_p): Likewise.
28602 (move_ssa_defining_stmt_for_defs): Likewise.
28603 (finish_update_gimple_call): Likewise.
28604 (update_gimple_call): Likewise.
28605 (update_call_from_tree): Likewise.
28606 (propagate_tree_value_into_stmt): Use replace_call_with_value.
28607 * gimple-fold.h (update_gimple_call): Declare.
28608 * gimple-fold.c (valid_gimple_rhs_p): Move here from
28609 tree-ssa-propagate.c.
28610 (update_gimple_call): Likewise.
28611 (valid_gimple_call_p): Likewise.
28612 (finish_update_gimple_call): Likewise, and simplify.
28613 (gimplify_and_update_call_from_tree): Implement
28614 update_call_from_tree functionality, avoid excessive
28615 push/pop_gimplify_context.
28616 (gimple_fold_builtin): Use only gimplify_and_update_call_from_tree.
28617 (gimple_fold_call): Likewise.
28618 * gimple-ssa-sprintf.c (try_substitute_return_value): Likewise.
28619 * tree-ssa-ccp.c (ccp_folder::fold_stmt): Likewise.
28620 (pass_fold_builtins::execute): Likewise.
28621 (optimize_stack_restore): Use replace_call_with_value.
28622 * tree-cfg.c (fold_loop_internal_call): Likewise.
28623 * tree-ssa-dce.c (maybe_optimize_arith_overflow): Use
28624 only gimplify_and_update_call_from_tree.
28625 * tree-ssa-strlen.c (handle_builtin_strlen): Likewise.
28626 (handle_builtin_strchr): Likewise.
28627 * tsan.c: Include gimple-fold.h instead of tree-ssa-propagate.h.
28629 2021-04-26 Jakub Jelinek <jakub@redhat.com>
28632 * vmsdbgout.c (ASM_OUTPUT_DEBUG_STRING, vmsdbgout_begin_block,
28633 vmsdbgout_end_block, lookup_filename, vmsdbgout_source_line): Remove
28636 2021-04-25 liuhongt <hongtao.liu@intel.com>
28639 * config/i386/i386-builtin.def (BDESC): Change the icode of
28640 the following builtins to CODE_FOR_nothing.
28641 * config/i386/i386.c (ix86_gimple_fold_builtin): Fold
28642 IX86_BUILTIN_PCMPEQB128, IX86_BUILTIN_PCMPEQW128,
28643 IX86_BUILTIN_PCMPEQD128, IX86_BUILTIN_PCMPEQQ,
28644 IX86_BUILTIN_PCMPEQB256, IX86_BUILTIN_PCMPEQW256,
28645 IX86_BUILTIN_PCMPEQD256, IX86_BUILTIN_PCMPEQQ256,
28646 IX86_BUILTIN_PCMPGTB128, IX86_BUILTIN_PCMPGTW128,
28647 IX86_BUILTIN_PCMPGTD128, IX86_BUILTIN_PCMPGTQ,
28648 IX86_BUILTIN_PCMPGTB256, IX86_BUILTIN_PCMPGTW256,
28649 IX86_BUILTIN_PCMPGTD256, IX86_BUILTIN_PCMPGTQ256.
28650 * config/i386/sse.md (avx2_eq<mode>3): Deleted.
28651 (sse2_eq<mode>3): Ditto.
28652 (sse4_1_eqv2di3): Ditto.
28653 (sse2_gt<mode>3): Rename to ..
28654 (*sse2_gt<mode>3): .. this.
28656 2021-04-24 Iain Sandoe <iain@sandoe.co.uk>
28659 2021-04-24 Iain Sandoe <iain@sandoe.co.uk>
28662 * config/darwin.c (darwin_binds_local_p): Assume that any
28663 public symbol might be interposed for PIC code. Update function
28664 header comment to reflect current Darwin capability.
28666 2021-04-24 Iain Sandoe <iain@sandoe.co.uk>
28669 * config/darwin.c (darwin_binds_local_p): Assume that any
28670 public symbol might be interposed for PIC code. Update function
28671 header comment to reflect current Darwin capability.
28673 2021-04-24 Richard Sandiford <richard.sandiford@arm.com>
28675 * doc/sourcebuild.texi: Document no-opts and any-opts target
28678 2021-04-23 YiFei Zhu <zhuyifei1999@gmail.com>
28680 * config/bpf/bpf.h (ASM_OUTPUT_ALIGNED_BSS): Use .type and .lcomm.
28682 2021-04-23 YiFei Zhu <zhuyifei1999@gmail.com>
28684 * config/bpf/bpf.h (FUNCTION_BOUNDARY): Set to 64.
28686 2021-04-23 Uroš Bizjak <ubizjak@gmail.com>
28689 * config/i386/i386-options.c (ix86_option_override_internal):
28690 Error out when -m96bit-long-double is used with 64bit targets.
28691 * config/i386/i386.md (*pushxf_rounded): Remove pattern.
28693 2021-04-23 Martin Liska <mliska@suse.cz>
28695 * lto-wrapper.c: Remove FIXME about usage of
28696 hardware_concurrency. The function is not on par with
28699 2021-04-23 Uroš Bizjak <ubizjak@gmail.com>
28702 * config/i386/sync.md (FILD_ATOMIC/FIST_ATOMIC FP load peephole2):
28703 Copy operand 3 to operand 4. Use sse_reg_operand
28704 as operand 3 predicate.
28705 (FILD_ATOMIC/FIST_ATOMIC FP load peephole2 with mem blockage): Ditto.
28706 (LDX_ATOMIC/STX_ATOMIC FP load peephole2): Ditto.
28707 (LDX_ATOMIC/LDX_ATOMIC FP load peephole2 with mem blockage): Ditto.
28708 (FILD_ATOMIC/FIST_ATOMIC FP store peephole2):
28709 Copy operand 1 to operand 0.
28710 (FILD_ATOMIC/FIST_ATOMIC FP store peephole2 with mem blockage): Ditto.
28711 (LDX_ATOMIC/STX_ATOMIC FP store peephole2): Ditto.
28712 (LDX_ATOMIC/LDX_ATOMIC FP store peephole2 with mem blockage): Ditto.
28714 2021-04-23 Alex Coplan <alex.coplan@arm.com>
28716 PR rtl-optimization/100230
28717 * early-remat.c (early_remat::sort_candidates): Use delete[]
28718 instead of delete for array allocated with new[].
28720 2021-04-23 Richard Biener <rguenther@suse.de>
28722 * genmatch.c (lower_cond): Remove VEC_COND_EXPR special-casing.
28723 (capture_info::capture_info): Likewise.
28724 (capture_info::walk_match): Likewise.
28725 (expr::gen_transform): Likewise.
28726 (dt_simplify::gen_1): Likewise.
28727 * gimple-match-head.c (maybe_resimplify_conditional_op):
28728 Remove VEC_COND_EXPR special-casing.
28729 (gimple_simplify): Likewise.
28730 * gimple.c (gimple_could_trap_p_1): Adjust.
28731 * tree-ssa-pre.c (compute_avail): Allow VEC_COND_EXPR
28732 to participate in PRE.
28734 2021-04-23 Richard Biener <rguenther@suse.de>
28736 * cfganal.c (connect_infinite_loops_to_exit): First call
28737 add_noreturn_fake_exit_edges.
28738 * ipa-sra.c (process_scan_results): Do not call the now redundant
28739 add_noreturn_fake_exit_edges.
28740 * predict.c (tree_estimate_probability): Likewise.
28741 (rebuild_frequencies): Likewise.
28742 * store-motion.c (one_store_motion_pass): Likewise.
28744 2021-04-23 Richard Biener <rguenther@suse.de>
28746 PR tree-optimization/100222
28747 * predict.c (pass_profile::execute): Remove redundant call to
28748 mark_irreducible_loops.
28749 (report_predictor_hitrates): Likewise.
28751 2021-04-23 Richard Biener <rguenther@suse.de>
28753 * tree-ssa-loop-ivopts.c (rewrite_use_nonlinear_expr): Avoid
28754 valid_gimple_rhs_p by instead gimplifying to one.
28756 2021-04-23 Richard Biener <rguenther@suse.de>
28758 PR tree-optimization/99971
28759 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences):
28760 Always use TBAA for loads.
28762 2021-04-23 liuhongt <hongtao.liu@intel.com>
28765 * config/i386/i386-options.c (ix86_option_override_internal):
28766 Clear MASK_AVX256_SPLIT_UNALIGNED_LOAD/STORE in x_target_flags
28767 when X86_TUNE_AVX256_UNALIGNED_LOAD/STORE_OPTIMAL is enabled
28768 by target attribute.
28770 2021-04-23 David Edelsohn <dje.gcc@gmail.com>
28772 * config/rs6000/aix71.h (PREFERRED_DEBUGGING_TYPE): Change to
28774 * config/rs6000/aix72.h (PREFERRED_DEBUGGING_TYPE): Same.
28776 2021-04-22 David Edelsohn <dje.gcc@gmail.com>
28778 * config.gcc (powerpc-ibm-aix6.*): Remove.
28779 * config/rs6000/aix61.h: Delete.
28781 2021-04-22 Martin Liska <mliska@suse.cz>
28783 PR testsuite/100159
28784 PR testsuite/100192
28785 * builtins.c (expand_builtin): Fix typos and missing comments.
28786 * dwarf2out.c (gen_subprogram_die): Likewise.
28787 (gen_struct_or_union_type_die): Likewise.
28789 2021-04-22 Uroš Bizjak <ubizjak@gmail.com>
28792 * config/i386/i386-expand.c (ix86_expand_convert_uns_sidf_sse):
28793 Remove the sign with FE_DOWNWARD, where x - x = -0.0.
28795 2021-04-21 Iain Sandoe <iain@sandoe.co.uk>
28797 * config/i386/darwin.h (TARGET_64BIT): Remove definition
28798 based on TARGET_ISA_64BIT.
28799 (TARGET_64BIT_P): Remove definition based on
28800 TARGET_ISA_64BIT_P().
28802 2021-04-21 Martin Liska <mliska@suse.cz>
28805 2021-04-21 Martin Liska <mliska@suse.cz>
28807 * lto-wrapper.c (cpuset_popcount): Remove.
28808 (init_num_threads): Remove and use hardware_concurrency.
28810 2021-04-21 Martin Liska <mliska@suse.cz>
28813 * main.c (main): Call toplev::finalize in CHECKING_P mode.
28814 * ipa-modref.c (ipa_modref_c_finalize): summaries are NULL
28815 when incremental LTO linking happens.
28817 2021-04-21 Martin Liska <mliska@suse.cz>
28819 * lto-wrapper.c (run_gcc): When -flto=jobserver is used, but the
28820 makeserver cannot be detected, then use -flto=N fallback.
28822 2021-04-21 Richard Sandiford <richard.sandiford@arm.com>
28824 * acinclude.m4 (gcc_AC_INITFINI_ARRAY): When cross-compiling,
28825 default to yes for aarch64-linux-gnu.
28826 * configure: Regenerate.
28828 2021-04-21 Martin Liska <mliska@suse.cz>
28830 * lto-wrapper.c (cpuset_popcount): Remove.
28831 (init_num_threads): Remove and use hardware_concurrency.
28833 2021-04-21 Martin Liska <mliska@suse.cz>
28835 * config/i386/i386.c: Remove superfluous || TARGET_MACHO
28836 which remains to be '(... || 0)' and clang complains about it.
28837 * dwarf2out.c (AT_vms_delta): Declare conditionally.
28838 (add_AT_vms_delta): Likewise.
28839 * tree.c (fld_simplified_type): Use rather more common pattern
28840 for disabling of something (#if 0).
28841 (get_tree_code_name): Likewise.
28842 (verify_type_variant): Likewise.
28844 2021-04-21 Martin Liska <mliska@suse.cz>
28846 * config/i386/i386-expand.c (decide_alignment): Use newly named
28847 macro TARGET_CPU_P.
28848 * config/i386/i386.c (ix86_decompose_address): Likewise.
28849 (ix86_address_cost): Likewise.
28850 (ix86_lea_outperforms): Likewise.
28851 (ix86_avoid_lea_for_addr): Likewise.
28852 (ix86_add_stmt_cost): Likewise.
28853 * config/i386/i386.h (TARGET_*): Remove.
28854 (TARGET_CPU_P): New macro.
28855 * config/i386/i386.md: Use newly named macro TARGET_CPU_P.
28856 * config/i386/x86-tune-sched-atom.c (do_reorder_for_imul): Likewise.
28857 (swap_top_of_ready_list): Likewise.
28858 (ix86_atom_sched_reorder): Likewise.
28859 * config/i386/x86-tune-sched-bd.c (ix86_bd_has_dispatch): Likewise.
28860 * config/i386/x86-tune-sched.c (ix86_adjust_cost): Likewise.
28862 2021-04-21 Martin Liska <mliska@suse.cz>
28864 * config/i386/i386-options.c (TARGET_EXPLICIT_NO_SAHF_P):
28866 (SET_TARGET_NO_SAHF): Likewise.
28867 (TARGET_EXPLICIT_PREFETCH_SSE_P): Likewise.
28868 (SET_TARGET_PREFETCH_SSE): Likewise.
28869 (TARGET_EXPLICIT_NO_TUNE_P): Likewise.
28870 (SET_TARGET_NO_TUNE): Likewise.
28871 (TARGET_EXPLICIT_NO_80387_P): Likewise.
28872 (SET_TARGET_NO_80387): Likewise.
28874 * config/i386/i386.h (TARGET_*): Remove.
28875 * opth-gen.awk: Generate new used macros.
28877 2021-04-21 Martin Liska <mliska@suse.cz>
28879 * config/i386/i386.h (PTA_*): Remove.
28880 (enum pta_flag): New.
28881 (DEF_PTA): Generate PTA_* values from i386-isa.def.
28882 * config/i386/i386-isa.def: New file.
28884 2021-04-21 Alex Coplan <alex.coplan@arm.com>
28887 * config/aarch64/aarch64-bti-insert.c (aarch64_bti_j_insn_p): New.
28888 (rest_of_insert_bti): Avoid inserting duplicate bti j insns for
28889 jump table targets.
28891 2021-04-21 H.J. Lu <hjl.tools@gmail.com>
28893 * config.gcc: Install mwaitintrin.h for i[34567]86-*-* and
28894 x86_64-*-* targets.
28895 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_MWAIT_SET):
28897 (OPTION_MASK_ISA2_MWAIT_UNSET): Likewise.
28898 (ix86_handle_option): Handle -mmwait.
28899 * config/i386/i386-builtins.c (ix86_init_mmx_sse_builtins):
28900 Replace OPTION_MASK_ISA_SSE3 with OPTION_MASK_ISA2_MWAIT on
28901 __builtin_ia32_monitor and __builtin_ia32_mwait.
28902 * config/i386/i386-options.c (isa2_opts): Add -mmwait.
28903 (ix86_valid_target_attribute_inner_p): Likewise.
28904 (ix86_option_override_internal): Enable mwait/monitor
28905 instructions for -msse3.
28906 * config/i386/i386.h (TARGET_MWAIT): New.
28907 (TARGET_MWAIT_P): Likewise.
28908 * config/i386/i386.opt: Add -mmwait.
28909 * config/i386/mwaitintrin.h: New file.
28910 * config/i386/pmmintrin.h: Include <mwaitintrin.h>.
28911 * config/i386/sse.md (sse3_mwait): Replace TARGET_SSE3 with
28913 (@sse3_monitor_<mode>): Likewise.
28914 * config/i386/x86gprintrin.h: Include <mwaitintrin.h>.
28915 * doc/extend.texi: Document mwait target attribute.
28916 * doc/invoke.texi: Document -mmwait.
28918 2021-04-21 Martin Liska <mliska@suse.cz>
28920 * config/i386/i386-options.c (DEF_ENUM): Remove it.
28921 * config/i386/i386-opts.h (DEF_ENUM): Likewise.
28922 * config/i386/stringop.def (DEF_ENUM): Likewise.
28924 2021-04-21 Martin Liska <mliska@suse.cz>
28926 * tree-cfg.c (gimple_verify_flow_info): Use qD instead
28927 of print_generic_expr.
28929 2021-04-21 Jakub Jelinek <jakub@redhat.com>
28931 PR rtl-optimization/100148
28932 * cprop.c (constprop_register): Use next_nondebug_insn instead of
28935 2021-04-21 Martin Liska <mliska@suse.cz>
28938 * cgraphunit.c (cgraph_node::analyze): Remove duplicate
28939 free_dominance_info calls.
28941 2021-04-21 Richard Biener <rguenther@suse.de>
28943 * gimple-fold.c (maybe_fold_reference): Remove is_lhs
28944 parameter (and assume it to be false).
28945 (fold_gimple_assign): Adjust, remove all callers of
28946 maybe_fold_reference calling it with is_lhs true.
28947 (gimple_fold_call): Likewise.
28948 (fold_stmt_1): Likewise.
28950 2021-04-21 Richard Biener <rguenther@suse.de>
28952 * fold-const.c (pedantic_non_lvalue_loc): Remove.
28953 (fold_binary_loc): Adjust.
28954 (fold_ternary_loc): Likewise.
28956 2021-04-21 Richard Sandiford <richard.sandiford@arm.com>
28958 PR middle-end/100130
28959 * varasm.c (get_block_for_decl): Make sure that any use of the
28960 retain attribute matches the section's retain flag.
28961 (switch_to_section): Check for retain mismatches even when
28962 changing sections, but do not warn if the given decl is the
28963 section's named.decl.
28964 (output_object_block): Pass the first decl in the block (if any)
28965 to switch_to_section.
28967 2021-04-20 H.J. Lu <hjl.tools@gmail.com>
28969 * config/i386/i386-c.c (ix86_target_macros_internal): Define
28970 __CRC32__ for -mcrc32.
28971 * config/i386/i386-options.c (ix86_option_override_internal):
28972 Enable crc32 instruction for -msse4.2.
28973 * config/i386/i386.md (sse4_2_crc32<mode>): Remove TARGET_SSE4_2
28975 (sse4_2_crc32di): Likewise.
28976 * config/i386/ia32intrin.h: Use crc32 target option for CRC32
28979 2021-04-20 Segher Boessenkool <segher@kernel.crashing.org>
28982 * config/rs6000/rs6000.c (rs6000_machine_from_flags): Do not consider
28985 2021-04-20 Martin Liska <mliska@suse.cz>
28987 * doc/invoke.texi: Fix typo.
28988 * params.opt: Likewise.
28990 2021-04-20 Martin Liska <mliska@suse.cz>
28992 * doc/invoke.texi: Document new param.
28994 2021-04-19 Andrew MacLeod <amacleod@redhat.com>
28996 PR tree-optimization/100081
28997 * gimple-range-cache.h (ranger_cache): Inherit from gori_compute
28998 rather than gori_compute_cache.
28999 * gimple-range-gori.cc (is_gimple_logical_p): Move to top of file.
29000 (range_def_chain::m_logical_depth): New member.
29001 (range_def_chain::range_def_chain): Initialize m_logical_depth.
29002 (range_def_chain::get_def_chain): Don't build defchains through more
29003 than LOGICAL_LIMIT logical expressions.
29004 * params.opt (param_ranger_logical_depth): New.
29006 2021-04-19 Richard Earnshaw <rearnsha@arm.com>
29009 * config/arm/arm.c (arm_configure_build_target): Do not strip
29010 extended FPU/SIMD feature bits from the target ISA when -mfpu
29011 is specified (partial revert of r11-8168).
29013 2021-04-19 Thomas Schwinge <thomas@codesourcery.com>
29015 * params.opt (-param=openacc-kernels=): Add.
29016 * omp-oacc-kernels-decompose.cc
29017 (pass_omp_oacc_kernels_decompose::gate): Use it.
29018 * doc/invoke.texi (-fopenacc-kernels=@var{mode}): Move...
29019 (--param): ... here, 'openacc-kernels'.
29021 2021-04-19 Martin Liska <mliska@suse.cz>
29024 * gengtype.c (finish_root_table): Align function arguments
29025 in between declaration and definition.
29027 2021-04-19 Eric Botcazou <ebotcazou@adacore.com>
29029 * config/i386/winnt.c (i386_pe_seh_cold_init): Properly deal with
29030 frames larger than the SEH maximum frame size.
29032 2021-04-18 Segher Boessenkool <segher@kernel.crashing.org>
29034 PR rtl-optimization/99927
29035 * combine.c (distribute_notes) [REG_UNUSED]: If the register already
29036 is dead, just drop it.
29038 2021-04-17 Iain Buclaw <ibuclaw@gdcproject.org>
29041 * config/i386/winnt-d.c (TARGET_D_TEMPLATES_ALWAYS_COMDAT): Define.
29042 * doc/tm.texi: Regenerate.
29043 * doc/tm.texi.in (D language and ABI): Add @hook for
29044 TARGET_D_TEMPLATES_ALWAYS_COMDAT.
29046 2021-04-17 Iain Buclaw <ibuclaw@gdcproject.org>
29048 * config/darwin-d.c (darwin_d_handle_target_object_format): New
29050 (darwin_d_register_target_info): New function.
29051 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
29052 * config/dragonfly-d.c (dragonfly_d_handle_target_object_format): New
29054 (dragonfly_d_register_target_info): New function.
29055 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
29056 * config/freebsd-d.c (freebsd_d_handle_target_object_format): New
29058 (freebsd_d_register_target_info): New function.
29059 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
29060 * config/glibc-d.c (glibc_d_handle_target_object_format): New
29062 (glibc_d_register_target_info): New function.
29063 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
29064 * config/i386/i386-d.c (ix86_d_handle_target_object_format): New
29066 (ix86_d_register_target_info): Add ix86_d_handle_target_object_format
29067 as handler for objectFormat key.
29068 * config/i386/winnt-d.c (winnt_d_handle_target_object_format): New
29070 (winnt_d_register_target_info): New function.
29071 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
29072 * config/netbsd-d.c (netbsd_d_handle_target_object_format): New
29074 (netbsd_d_register_target_info): New function.
29075 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
29076 * config/openbsd-d.c (openbsd_d_handle_target_object_format): New
29078 (openbsd_d_register_target_info): New function.
29079 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
29080 * config/pa/pa-d.c (pa_d_handle_target_object_format): New function.
29081 (pa_d_register_target_info): Add pa_d_handle_target_object_format as
29082 handler for objectFormat key.
29083 * config/rs6000/rs6000-d.c (rs6000_d_handle_target_object_format): New
29085 (rs6000_d_register_target_info): Add
29086 rs6000_d_handle_target_object_format as handler for objectFormat key.
29087 * config/sol2-d.c (solaris_d_handle_target_object_format): New
29089 (solaris_d_register_target_info): New function.
29090 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
29092 2021-04-16 Jakub Jelinek <jakub@redhat.com>
29095 * config/aarch64/aarch64.c (aarch64_function_arg_alignment): Change
29096 abi_break argument from bool * to unsigned *, store there the pre-GCC 9
29098 (aarch64_layout_arg, aarch64_gimplify_va_arg_expr): Adjust callers.
29099 (aarch64_function_arg_regno_p): Likewise. Only emit -Wpsabi note if
29100 the old and new alignment after applying MIN/MAX to it is different.
29102 2021-04-16 Tamar Christina <tamar.christina@arm.com>
29105 * config/aarch64/aarch64-sve.md (@aarch64_sve_trn1_conv<mode>): New.
29106 * config/aarch64/aarch64.c (aarch64_expand_sve_const_pred_trn): Use new
29108 * config/aarch64/iterators.md (UNSPEC_TRN1_CONV): New.
29110 2021-04-16 Bill Schmidt <wschmidt@linux.ibm.com>
29112 * doc/extend.texi (PowerPC AltiVec/VSX Built-in Functions): Revise
29113 this section and its subsections.
29115 2021-04-16 Jakub Jelinek <jakub@redhat.com>
29118 * config/aarch64/aarch64.md (*neg_asr_si2_extr, *extrsi5_insn_di): New
29119 define_insn patterns.
29121 2021-04-16 Richard Sandiford <richard.sandiford@arm.com>
29123 PR rtl-optimization/98689
29124 * reg-notes.def (UNTYPED_CALL): New note.
29125 * combine.c (distribute_notes): Handle it.
29126 * emit-rtl.c (try_split): Likewise.
29127 * rtlanal.c (rtx_properties::try_to_add_insn): Likewise. Assume
29128 that calls with the note implicitly set all return value registers.
29129 * builtins.c (expand_builtin_apply): Add a REG_UNTYPED_CALL
29132 2021-04-16 Richard Sandiford <richard.sandiford@arm.com>
29134 PR rtl-optimization/99596
29135 * rtlanal.c (rtx_properties::try_to_add_insn): Don't add global
29136 register accesses for const calls. Assume that pure functions
29137 can only read from global registers. Ignore cases in which
29138 the stack pointer has been marked global.
29140 2021-04-16 Jakub Jelinek <jakub@redhat.com>
29143 * tree-vect-loop.c (vect_transform_loop): Don't remove just
29144 dead scalar .MASK_LOAD calls, but also dead .COND_* calls - replace
29145 them by their last argument.
29147 2021-04-15 Martin Liska <mliska@suse.cz>
29149 * doc/invoke.texi: Other params don't use it, remove it.
29151 2021-04-15 Richard Biener <rguenther@suse.de>
29153 * gimple-builder.h: Add deprecation note.
29155 2021-04-15 Richard Sandiford <richard.sandiford@arm.com>
29158 * attribs.h (restrict_type_identity_attributes_to): Declare.
29159 * attribs.c (restrict_type_identity_attributes_to): New function.
29161 2021-04-15 Richard Sandiford <richard.sandiford@arm.com>
29164 * attribs.h (affects_type_identity_attributes): Declare.
29165 * attribs.c (remove_attributes_matching): New function.
29166 (affects_type_identity_attributes): Likewise.
29168 2021-04-15 Jakub Jelinek <jakub@redhat.com>
29171 * config/aarch64/aarch64.md (*<LOGICAL:optab>_<SHIFT:optab><mode>3):
29172 Add combine splitters for *<LOGICAL:optab>_ashl<mode>3 with
29173 ZERO_EXTEND, SIGN_EXTEND or AND.
29175 2021-04-14 Richard Sandiford <richard.sandiford@arm.com>
29177 PR rtl-optimization/99929
29178 * rtl.h (same_vector_encodings_p): New function.
29179 * cse.c (exp_equiv_p): Check that CONST_VECTORs have the same encoding.
29180 * cselib.c (rtx_equal_for_cselib_1): Likewise.
29181 * jump.c (rtx_renumbered_equal_p): Likewise.
29182 * lra-constraints.c (operands_match_p): Likewise.
29183 * reload.c (operands_match_p): Likewise.
29184 * rtl.c (rtx_equal_p_cb, rtx_equal_p): Likewise.
29186 2021-04-14 Richard Sandiford <richard.sandiford@arm.com>
29188 * print-rtl.c (rtx_writer::print_rtx_operand_codes_E_and_V): Print
29189 more information about variable-length CONST_VECTORs.
29191 2021-04-14 Vladimir N. Makarov <vmakarov@redhat.com>
29193 PR rtl-optimization/100066
29194 * lra-constraints.c (split_reg): Check paradoxical_subreg_p for
29195 ordered modes when choosing splitting mode for hard reg.
29197 2021-04-14 Richard Sandiford <richard.sandiford@arm.com>
29200 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector_sel):
29202 (aarch64_expand_sve_const_vector): Use it for nelts_per_pattern==2.
29204 2021-04-14 Andreas Krebbel <krebbel@linux.ibm.com>
29206 * config/s390/s390-builtins.def (O_M5, O_M12, ...): Add new macros
29207 for mask operand types.
29208 (s390_vec_permi_s64, s390_vec_permi_b64, s390_vec_permi_u64)
29209 (s390_vec_permi_dbl, s390_vpdi): Use the M5 type for the immediate
29211 (s390_vec_msum_u128, s390_vmslg): Use the M12 type for the
29213 * config/s390/s390.c (s390_const_operand_ok): Check the new
29214 operand types and generate a list of valid values.
29216 2021-04-14 Iain Buclaw <ibuclaw@gdcproject.org>
29218 * doc/tm.texi: Regenerate.
29219 * doc/tm.texi.in (D language and ABI): Add @hook for
29220 TARGET_D_REGISTER_OS_TARGET_INFO.
29222 2021-04-14 Iain Buclaw <ibuclaw@gdcproject.org>
29224 * config/aarch64/aarch64-d.c (aarch64_d_handle_target_float_abi): New
29226 (aarch64_d_register_target_info): New function.
29227 * config/aarch64/aarch64-protos.h (aarch64_d_register_target_info):
29229 * config/aarch64/aarch64.h (TARGET_D_REGISTER_CPU_TARGET_INFO):
29231 * config/arm/arm-d.c (arm_d_handle_target_float_abi): New function.
29232 (arm_d_register_target_info): New function.
29233 * config/arm/arm-protos.h (arm_d_register_target_info): Declare.
29234 * config/arm/arm.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
29235 * config/i386/i386-d.c (ix86_d_handle_target_float_abi): New function.
29236 (ix86_d_register_target_info): New function.
29237 * config/i386/i386-protos.h (ix86_d_register_target_info): Declare.
29238 * config/i386/i386.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
29239 * config/mips/mips-d.c (mips_d_handle_target_float_abi): New function.
29240 (mips_d_register_target_info): New function.
29241 * config/mips/mips-protos.h (mips_d_register_target_info): Declare.
29242 * config/mips/mips.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
29243 * config/pa/pa-d.c (pa_d_handle_target_float_abi): New function.
29244 (pa_d_register_target_info): New function.
29245 * config/pa/pa-protos.h (pa_d_register_target_info): Declare.
29246 * config/pa/pa.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
29247 * config/riscv/riscv-d.c (riscv_d_handle_target_float_abi): New
29249 (riscv_d_register_target_info): New function.
29250 * config/riscv/riscv-protos.h (riscv_d_register_target_info): Declare.
29251 * config/riscv/riscv.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
29252 * config/rs6000/rs6000-d.c (rs6000_d_handle_target_float_abi): New
29254 (rs6000_d_register_target_info): New function.
29255 * config/rs6000/rs6000-protos.h (rs6000_d_register_target_info):
29257 * config/rs6000/rs6000.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
29258 * config/s390/s390-d.c (s390_d_handle_target_float_abi): New function.
29259 (s390_d_register_target_info): New function.
29260 * config/s390/s390-protos.h (s390_d_register_target_info): Declare.
29261 * config/s390/s390.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
29262 * config/sparc/sparc-d.c (sparc_d_handle_target_float_abi): New
29264 (sparc_d_register_target_info): New function.
29265 * config/sparc/sparc-protos.h (sparc_d_register_target_info): Declare.
29266 * config/sparc/sparc.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
29267 * doc/tm.texi: Regenerate.
29268 * doc/tm.texi.in (D language and ABI): Add @hook for
29269 TARGET_D_REGISTER_CPU_TARGET_INFO.
29271 2021-04-14 Iain Buclaw <ibuclaw@gdcproject.org>
29273 * config/i386/i386-d.c (ix86_d_has_stdcall_convention): New function.
29274 * config/i386/i386-protos.h (ix86_d_has_stdcall_convention): Declare.
29275 * config/i386/i386.h (TARGET_D_HAS_STDCALL_CONVENTION): Define.
29276 * doc/tm.texi: Regenerate.
29277 * doc/tm.texi.in (D language and ABI): Add @hook for
29278 TARGET_D_HAS_STDCALL_CONVENTION.
29280 2021-04-14 Richard Biener <rguenther@suse.de>
29282 * tree-cfg.c (verify_gimple_assign_ternary): Verify that
29283 VEC_COND_EXPRs have a gimple_val condition.
29284 * tree-ssa-propagate.c (valid_gimple_rhs_p): VEC_COND_EXPR
29285 can no longer have a GENERIC condition.
29287 2021-04-14 Richard Earnshaw <rearnsha@arm.com>
29290 * config/arm/arm.c (arm_configure_build_target): Strip isa_all_fpbits
29291 from the isa_delta when -mfpu has been used.
29292 (arm_options_perform_arch_sanity_checks): It's the architecture that
29293 lacks an FPU not the processor.
29295 2021-04-13 Richard Biener <rguenther@suse.de>
29297 PR tree-optimization/100053
29298 * tree-ssa-sccvn.c (vn_nary_op_get_predicated_value): Do
29299 not use optimistic dominance queries for backedges to validate
29301 (dominated_by_p_w_unex): Add parameter to ignore executable
29302 state on backedges.
29303 (rpo_elim::eliminate_avail): Adjust.
29305 2021-04-13 Jakub Jelinek <jakub@redhat.com>
29308 * config/aarch64/aarch64.md (*aarch64_bfxil<mode>_extr,
29309 *aarch64_bfxilsi_extrdi): New define_insn patterns.
29311 2021-04-13 Jakub Jelinek <jakub@redhat.com>
29314 * simplify-rtx.c (simplify_immed_subreg): For MODE_COMPOSITE_P
29315 outermode, return NULL if the result doesn't encode back to the
29316 original byte sequence.
29317 (simplify_gen_subreg): Don't create SUBREGs from constants to
29318 MODE_COMPOSITE_P outermode.
29320 2021-04-12 Jakub Jelinek <jakub@redhat.com>
29322 PR rtl-optimization/99905
29323 * combine.c (expand_compound_operation): If pos + len > modewidth,
29324 perform the right shift by pos in inner_mode and then convert to mode,
29325 instead of trying to simplify a shift of rtx with inner_mode by pos
29326 as if it was a shift in mode.
29328 2021-04-12 Jakub Jelinek <jakub@redhat.com>
29331 * combine.c (simplify_and_const_int_1): Don't optimize varop
29332 away if it has side-effects.
29334 2021-04-12 Martin Liska <mliska@suse.cz>
29336 * doc/extend.texi: Escape @smallexample content.
29338 2021-04-12 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
29340 * config/s390/s390.md ("*movdi_31", "*movdi_64"): Add
29341 alternative in order to load a DFP zero.
29343 2021-04-12 Martin Liska <mliska@suse.cz>
29345 * doc/extend.texi: Be more precise in documentation
29346 of symver attribute.
29348 2021-04-12 Martin Liska <mliska@suse.cz>
29351 * gimplify.c (gimplify_expr): Right now, we unpoison all
29352 variables before a goto <dest>. We should not do it if we are
29355 2021-04-12 Cui,Lili <lili.cui@intel.com>
29357 * common/config/i386/cpuinfo.h (get_intel_cpu): Handle
29359 * common/config/i386/i386-common.c (processor_names): Add
29361 (processor_alias_table): Add rocketlake.
29362 * common/config/i386/i386-cpuinfo.h (processor_subtypes): Add
29363 INTEL_COREI7_ROCKETLAKE.
29364 * config.gcc: Add -march=rocketlake.
29365 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
29367 * config/i386/i386-options.c (m_ROCKETLAKE) : Define.
29368 (processor_cost_table): Add rocketlake cost.
29369 * config/i386/i386.h (ix86_size_cost) : Define
29371 (processor_type) : Add PROCESSOR_ROCKETLAKE.
29372 (PTA_ROCKETLAKE): Ditto.
29373 * doc/extend.texi: Add rocketlake.
29374 * doc/invoke.texi: Add rocketlake.
29376 2021-04-12 Cui,Lili <lili.cui@intel.com>
29378 * config/i386/i386.h (PTA_ALDERLAKE): Change alderlake ISA list.
29379 * config/i386/i386-options.c (m_CORE_AVX2): Add m_ALDERLAKE.
29380 * common/config/i386/cpuinfo.h (get_intel_cpu): Add AlderLake model.
29381 * doc/invoke.texi: Change alderlake ISA list.
29383 2021-04-11 Hafiz Abid Qadeer <abidh@codesourcery.com>
29385 PR middle-end/98088
29386 * omp-expand.c (expand_oacc_collapse_init): Update condition in
29389 2021-04-10 H.J. Lu <hjl.tools@gmail.com>
29392 * config/i386/serializeintrin.h (_serialize): Defined as macro.
29394 2021-04-10 Jakub Jelinek <jakub@redhat.com>
29397 * expr.c (expand_expr_addr_expr_1): Test is_global_var rather than
29398 just TREE_STATIC on COMPOUND_LITERAL_EXPR_DECLs.
29400 2021-04-10 Jakub Jelinek <jakub@redhat.com>
29402 PR middle-end/99989
29403 * gimple-ssa-warn-alloca.c
29404 (alloca_type_and_limit::alloca_type_and_limit): Initialize limit to
29405 0 with integer precision unconditionally.
29407 2021-04-10 Jakub Jelinek <jakub@redhat.com>
29409 PR rtl-optimization/98601
29410 * rtlanal.c (rtx_addr_can_trap_p_1): Allow in assert unknown size
29411 not just for BLKmode, but also for VOIDmode. For STRICT_ALIGNMENT
29412 unaligned_mems handle VOIDmode like BLKmode.
29414 2021-04-10 Jan Hubicka <hubicka@ucw.cz>
29417 * tree.c (free_lang_data_in_decl): Do not release body of
29418 declare_variant_alt.
29420 2021-04-09 Richard Sandiford <richard.sandiford@arm.com>
29422 * config/aarch64/aarch64.c (aarch64_option_restore): If the
29423 architecture was specified explicitly and the tuning wasn't,
29424 tune for the architecture rather than the configured default CPU.
29426 2021-04-09 Richard Sandiford <richard.sandiford@arm.com>
29428 * config/aarch64/aarch64.md (tlsdesc_small_sve_<mode>): Use X30
29429 as the temporary register.
29431 2021-04-09 Martin Liska <mliska@suse.cz>
29433 * doc/extend.texi: Move non-target attributes on the top level.
29435 2021-04-09 Martin Liska <mliska@suse.cz>
29437 * doc/invoke.texi: Document minimum and maximum value of the
29438 argument for both supported compression algorithms.
29440 2021-04-08 David Edelsohn <dje.gcc@gmail.com>
29442 * config/rs6000/rs6000.c (rs6000_xcoff_select_section): Select
29443 TLS BSS before TLS data.
29444 * config/rs6000/xcoff.h (ASM_OUTPUT_TLS_COMMON): Use .comm.
29446 2021-04-08 Richard Sandiford <richard.sandiford@arm.com>
29448 * doc/sourcebuild.texi (stdint_types_mbig_endian): Document.
29450 2021-04-08 Richard Sandiford <richard.sandiford@arm.com>
29452 * match.pd: Extend vec_cond folds to handle shifts.
29454 2021-04-08 Maciej W. Rozycki <macro@orcam.me.uk>
29456 * config/vax/vax.md: Fix comment for `*bit<mode>' pattern's
29459 2021-04-08 Alex Coplan <alex.coplan@arm.com>
29462 * config/arm/iterators.md (MVE_vecs): New.
29463 (V_elem): Also handle V2DF.
29464 * config/arm/mve.md (*mve_mov<mode>): Rename to ...
29465 (*mve_vdup<mode>): ... this. Remove second alternative since
29466 vec_duplicate of const_int is not canonical RTL, and we don't
29467 want to match symbol_refs.
29468 (*mve_vec_duplicate<mode>): Delete (pattern is redundant).
29470 2021-04-08 Xionghu Luo <luoxhu@linux.ibm.com>
29472 * fold-const.c (fold_single_bit_test): Fix typo.
29473 * print-rtl.c (print_rtx_insn_vec): Call print_rtl_single
29476 2021-04-07 Richard Sandiford <richard.sandiford@arm.com>
29478 PR tree-optimization/97513
29479 * tree-vect-slp.c (vect_add_slp_permutation): New function,
29481 (vectorizable_slp_permutation): ...here. Detect cases in which
29482 all VEC_PERM_EXPRs are guaranteed to have the same stepped
29483 permute vector and only generate one permute vector for that case.
29484 Extend that case to handle variable-length vectors.
29486 2021-04-07 Richard Sandiford <richard.sandiford@arm.com>
29488 PR tree-optimization/99873
29489 * tree-vect-slp.c (vect_slp_prefer_store_lanes_p): New function.
29490 (vect_build_slp_instance): Don't split store groups that could
29491 use IFN_STORE_LANES.
29493 2021-04-07 Jakub Jelinek <jakub@redhat.com>
29496 * varasm.c (output_constant_pool_contents): Don't strip name encoding
29497 from XSTR (desc->sym, 0) or from label before passing those to
29500 2021-04-07 Richard Biener <rguenther@suse.de>
29502 PR tree-optimization/99954
29503 * tree-loop-distribution.c: Include tree-affine.h.
29504 (generate_memcpy_builtin): Try using tree-affine to prove
29506 (loop_distribution::classify_builtin_ldst): Always classify
29509 2021-04-07 Richard Biener <rguenther@suse.de>
29511 PR tree-optimization/99947
29512 * tree-vect-loop.c (vectorizable_induction): Pre-allocate
29513 steps vector to avoid pushing elements from the reallocated
29516 2021-04-07 Richard Biener <rguenther@suse.de>
29518 * tree-ssa-sccvn.h (print_vn_reference_ops): Declare.
29519 * tree-ssa-pre.c (print_pre_expr): Factor out VN reference operand
29521 * tree-ssa-sccvn.c (print_vn_reference_ops): ... into this new
29523 (debug_vn_reference_ops): New.
29525 2021-04-07 Bin Cheng <bin.cheng@linux.alibaba.com>
29527 PR tree-optimization/98736
29528 * tree-loop-distribution.c
29529 * (loop_distribution::bb_top_order_init):
29530 Compute RPO with programing order preserved by calling function
29531 rev_post_order_and_mark_dfs_back_seme.
29533 2021-04-06 Vladimir N. Makarov <vmakarov@redhat.com>
29536 * lra-constraints.c (split_reg): Don't check paradoxical_subreg_p.
29537 * lra-lives.c (clear_sparseset_regnos, regnos_in_sparseset_p): New
29539 (process_bb_lives): Don't update biggest mode of hard reg for
29540 implicit in multi-register group. Use the new functions for
29541 updating dead_set and unused_set by register notes.
29543 2021-04-06 Xianmiao Qu <xianmiao_qu@c-sky.com>
29545 * config/csky/csky_pipeline_ck802.md : Use insn reservation name
29548 2021-04-06 H.J. Lu <hjl.tools@gmail.com>
29550 * config/i386/x86-tune-costs.h (skylake_memcpy): Updated.
29551 (skylake_memset): Likewise.
29552 (skylake_cost): Change CLEAR_RATIO to 17.
29553 * config/i386/x86-tune.def (X86_TUNE_PREFER_KNOWN_REP_MOVSB_STOSB):
29554 Replace m_CANNONLAKE, m_ICELAKE_CLIENT, m_ICELAKE_SERVER,
29555 m_TIGERLAKE and m_SAPPHIRERAPIDS with m_SKYLAKE and m_CORE_AVX512.
29557 2021-04-06 Richard Biener <rguenther@suse.de>
29559 PR tree-optimization/99880
29560 * tree-vect-loop.c (maybe_set_vectorized_backedge_value): Only
29561 set vectorized defs of relevant PHIs.
29563 2021-04-06 Richard Biener <rguenther@suse.de>
29565 PR tree-optimization/99924
29566 * tree-vect-slp.c (vect_bb_partition_graph_r): Do not mark
29567 nodes w/o scalar stmts as visited.
29569 2021-04-06 Alex Coplan <alex.coplan@arm.com>
29572 * config/arm/arm.c (arm_libcall_uses_aapcs_base): Also use base
29573 PCS for [su]fix_optab.
29575 2021-04-03 Iain Sandoe <iain@sandoe.co.uk>
29577 * config/darwin.c (machopic_legitimize_pic_address): Check
29578 that the current pic register is one of the hard reg set
29579 before setting liveness.
29581 2021-04-03 Iain Sandoe <iain@sandoe.co.uk>
29583 * config/darwin.c (machopic_legitimize_pic_address): Fix
29584 whitespace, remove unused code.
29586 2021-04-03 Jakub Jelinek <jakub@redhat.com>
29588 PR tree-optimization/99882
29589 * gimple-ssa-store-merging.c (bswap_view_convert): Handle val with
29592 2021-04-03 Jakub Jelinek <jakub@redhat.com>
29594 PR rtl-optimization/99863
29595 * dse.c (replace_read): Drop regs_live argument. Instead of
29596 regs_live, use store_insn->fixed_regs_live if non-NULL,
29597 otherwise punt if insns sequence clobbers or sets any hard
29600 2021-04-03 Jakub Jelinek <jakub@redhat.com>
29603 * targhooks.h (default_print_patchable_function_entry_1): Declare.
29604 * targhooks.c (default_print_patchable_function_entry_1): New function,
29605 copied from default_print_patchable_function_entry with an added flags
29607 (default_print_patchable_function_entry): Rewritten into a small
29608 wrapper around default_print_patchable_function_entry_1.
29609 * config/rs6000/rs6000.c (TARGET_ASM_PRINT_PATCHABLE_FUNCTION_ENTRY):
29611 (rs6000_print_patchable_function_entry): New function.
29613 2021-04-02 Eric Botcazou <ebotcazou@adacore.com>
29615 * doc/invoke.texi (fdelete-dead-exceptions): Minor tweak.
29617 2021-04-01 Jason Merrill <jason@redhat.com>
29620 * common.opt: Document v15 and v16.
29622 2021-04-01 Richard Biener <rguenther@suse.de>
29624 PR tree-optimization/99863
29625 * gimplify.c (gimplify_init_constructor): Recompute vector
29628 2021-04-01 Jakub Jelinek <jakub@redhat.com>
29630 * doc/extend.texi (symver attribute): Fix up syntax errors
29633 2021-04-01 Jakub Jelinek <jakub@redhat.com>
29635 PR tree-optimization/96573
29636 * gimple-ssa-store-merging.c (init_symbolic_number): Handle
29637 also pointer types.
29639 2021-04-01 Richard Biener <rguenther@suse.de>
29641 PR tree-optimization/99856
29642 * tree-vect-patterns.c (vect_recog_over_widening_pattern): Promote
29643 precision to vector element precision.
29645 2021-04-01 Martin Jambor <mjambor@suse.cz>
29647 PR tree-optimization/97009
29648 * tree-sra.c (access_or_its_child_written): New function.
29649 (propagate_subaccesses_from_rhs): Use it instead of a simple grp_write
29652 2021-03-31 Jan Hubicka <hubicka@ucw.cz>
29655 * cif-code.def (USES_COMDAT_LOCAL): Make CIF_FINAL_NORMAL.
29657 2021-03-31 Pat Haugen <pthaugen@linux.ibm.com>
29660 * config/rs6000/altivec.md (xxspltiw_v4si, xxspltiw_v4sf_inst,
29661 xxspltidp_v2df_inst, xxsplti32dx_v4si_inst, xxsplti32dx_v4sf_inst,
29662 xxblend_<mode>, xxpermx_inst, xxeval): Mark prefixed.
29663 * config/rs6000/mma.md (mma_<vvi4i4i8>, mma_<avvi4i4i8>,
29664 mma_<vvi4i4i2>, mma_<avvi4i4i2>, mma_<vvi4i4>, mma_<avvi4i4>,
29665 mma_<pvi4i2>, mma_<apvi4i2>, mma_<vvi4i4i4>, mma_<avvi4i4i4>):
29667 * config/rs6000/rs6000.c (rs6000_final_prescan_insn): Adjust test.
29668 * config/rs6000/rs6000.md (define_attr "maybe_prefixed"): New.
29669 (define_attr "prefixed"): Update initializer.
29671 2021-03-31 Jakub Jelinek <jakub@redhat.com>
29674 * dwarf2out.c (debug_ranges_dwo_section): New variable.
29675 (DW_RANGES_IDX_SKELETON): Define.
29676 (struct dw_ranges): Add begin_entry and end_entry members.
29677 (DEBUG_DWO_RNGLISTS_SECTION): Define.
29678 (add_ranges_num): Adjust r initializer for addition of *_entry
29680 (add_ranges_by_labels): For -gsplit-dwarf and force_direct,
29681 set idx to DW_RANGES_IDX_SKELETON.
29682 (use_distinct_base_address_for_range): New function.
29683 (index_rnglists): Don't set r->idx if it is equal to
29684 DW_RANGES_IDX_SKELETON. Initialize r->begin_entry and
29685 r->end_entry for -gsplit-dwarf if those will be needed by
29687 (output_rnglists): Add DWO argument. If true, switch to
29688 debug_ranges_dwo_section rather than debug_ranges_section.
29689 Adjust l1/l2 label indexes. Only output the offset table when
29690 dwo is true and don't include in there the skeleton range
29691 entry if present. For -gsplit-dwarf, skip ranges that belong
29692 to the other rnglists section. Change return type from void
29693 to bool and return true if there are any range entries for
29694 the other section. For dwarf_split_debug_info use
29695 DW_RLE_startx_endx, DW_RLE_startx_length and DW_RLE_base_addressx
29696 entries instead of DW_RLE_start_end, DW_RLE_start_length and
29697 DW_RLE_base_address. Use use_distinct_base_address_for_range.
29698 (init_sections_and_labels): Initialize debug_ranges_dwo_section
29699 if -gsplit-dwarf and DWARF >= 5. Adjust ranges_section_label
29700 and range_base_label indexes.
29701 (dwarf2out_finish): Call index_rnglists earlier before finalizing
29702 .debug_addr. Never emit DW_AT_rnglists_base attribute. For
29703 -gsplit-dwarf and DWARF >= 5 call output_rnglists up to twice
29704 with different dwo arguments.
29705 (dwarf2out_c_finalize): Clear debug_ranges_dwo_section.
29707 2021-03-31 Richard Sandiford <richard.sandiford@arm.com>
29709 PR tree-optimization/98268
29710 * gimple-fold.c (maybe_canonicalize_mem_ref_addr): Call
29711 recompute_tree_invariant_for_addr_expr after successfully
29712 folding a TARGET_MEM_REF that occurs inside an ADDR_EXPR.
29714 2021-03-31 Richard Sandiford <richard.sandiford@arm.com>
29716 PR tree-optimization/99726
29717 * tree-data-ref.c (create_intersect_range_checks_index): Bail
29718 out if there is more than one access function SCEV for the loop
29721 2021-03-31 Richard Sandiford <richard.sandiford@arm.com>
29723 PR rtl-optimization/97141
29724 PR rtl-optimization/98726
29725 * emit-rtl.c (valid_for_const_vector_p): Return true for
29727 * rtx-vector-builder.h (rtx_vector_builder::step): Return a
29728 poly_wide_int instead of a wide_int.
29729 (rtx_vector_builder::apply_set): Take a poly_wide_int instead
29731 * rtx-vector-builder.c (rtx_vector_builder::apply_set): Likewise.
29732 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Return
29733 false for CONST_VECTORs that cannot be forced to memory.
29734 * config/aarch64/aarch64-simd.md (mov<mode>): If a CONST_VECTOR
29735 is too complex to force to memory, build it up from individual
29738 2021-03-31 Jan Hubicka <jh@suse.cz>
29741 * cgraph.c (cgraph_node::release_body): Fix overactive check.
29743 2021-03-31 Christophe Lyon <christophe.lyon@linaro.org>
29746 * config/arm/vec-common.md (mul<mode>3): Disable on iwMMXT, expect
29749 2021-03-31 H.J. Lu <hjl.tools@gmail.com>
29751 * config/i386/i386-expand.c (expand_set_or_cpymem_via_rep):
29752 For TARGET_PREFER_KNOWN_REP_MOVSB_STOSB, don't convert QImode
29754 (decide_alg): For TARGET_PREFER_KNOWN_REP_MOVSB_STOSB, use
29755 "rep movsb/stosb" only for known sizes.
29756 * config/i386/i386-options.c (processor_cost_table): Use Ice
29757 Lake cost for Cannon Lake, Ice Lake, Tiger Lake, Sapphire
29758 Rapids and Alder Lake.
29759 * config/i386/i386.h (TARGET_PREFER_KNOWN_REP_MOVSB_STOSB): New.
29760 * config/i386/x86-tune-costs.h (icelake_memcpy): New.
29761 (icelake_memset): Likewise.
29762 (icelake_cost): Likewise.
29763 * config/i386/x86-tune.def (X86_TUNE_PREFER_KNOWN_REP_MOVSB_STOSB):
29766 2021-03-31 Richard Sandiford <richard.sandiford@arm.com>
29769 * config/aarch64/aarch64.c
29770 (aarch64_vectorize_preferred_vector_alignment): Query the size
29771 of the provided SVE vector; do not assume that all SVE vectors
29772 have the same size.
29774 2021-03-31 Jan Hubicka <jh@suse.cz>
29777 * cgraph.c (cgraph_node::release_body): Remove all callers and
29779 * cgraphclones.c (cgraph_node::materialize_clone): Do not do it here.
29780 * cgraphunit.c (cgraph_node::expand): And here.
29782 2021-03-31 Martin Liska <mliska@suse.cz>
29784 * ipa-modref.c (analyze_ssa_name_flags): Fix coding style
29785 and one negated condition.
29787 2021-03-31 Jakub Jelinek <jakub@redhat.com>
29788 Richard Sandiford <richard.sandiford@arm.com>
29791 * config/aarch64/aarch64.md (*add<mode>3_poly_1): Swap Uai and Uav
29792 constraints on operands[2] and similarly 0 and rk constraints
29793 on operands[1] corresponding to that.
29795 2021-03-31 Jakub Jelinek <jakub@redhat.com>
29798 * configure.ac (HAVE_LD_BROKEN_PE_DWARF5): New AC_DEFINE if PECOFF
29799 linker doesn't support DWARF sections new in DWARF5.
29800 * config/i386/i386-options.c (ix86_option_override_internal): Default
29801 to dwarf_version 4 if HAVE_LD_BROKEN_PE_DWARF5 for TARGET_PECOFF
29803 * config.in: Regenerated.
29804 * configure: Regenerated.
29806 2021-03-30 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
29809 * config/aarch64/aarch64.c (aarch64_analyze_loop_vinfo): Check for
29810 available issue_info before using it.
29812 2021-03-30 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
29815 * config/aarch64/aarch64.md (sub<mode>3_compare1_imm): Do not allow zero
29818 2021-03-30 Xionghu Luo <luoxhu@linux.ibm.com>
29821 * config/rs6000/altivec.md (altivec_lvsl_reg): Change to ...
29822 (altivec_lvsl_reg_<mode>): ... this.
29823 (altivec_lvsr_reg): Change to ...
29824 (altivec_lvsr_reg_<mode>): ... this.
29825 * config/rs6000/predicates.md (vec_set_index_operand): New.
29826 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
29827 Enable 32bit variable vec_insert for all TARGET_VSX.
29828 * config/rs6000/rs6000.c (rs6000_expand_vector_set_var_p9):
29829 Enable 32bit variable vec_insert for p9 and above.
29830 (rs6000_expand_vector_set_var_p8): Rename to ...
29831 (rs6000_expand_vector_set_var_p7): ... this.
29832 (rs6000_expand_vector_set): Use TARGET_VSX and adjust assert
29834 * config/rs6000/vector.md (vec_set<mode>): Use vec_set_index_operand.
29835 * config/rs6000/vsx.md (xl_len_r): Use gen_altivec_lvsl_reg_di and
29836 gen_altivec_lvsr_reg_di.
29838 2021-03-30 H.J. Lu <hjl.tools@gmail.com>
29841 * config/i386/ia32intrin.h (__rdtsc): Defined as macro.
29842 (__rdtscp): Likewise.
29844 2021-03-30 Tamar Christina <tamar.christina@arm.com>
29846 PR tree-optimization/99825
29847 * tree-vect-slp-patterns.c (vect_check_evenodd_blend):
29848 Reject non-mult 2 lanes.
29850 2021-03-30 Richard Earnshaw <rearnsha@arm.com>
29853 * config/arm/arm.c (arm_file_start): Fix emission of
29854 Tag_ABI_VFP_args attribute.
29856 2021-03-30 Richard Biener <rguenther@suse.de>
29858 PR tree-optimization/99824
29859 * stor-layout.c (set_min_and_max_values_for_integral_type):
29860 Assert the precision is within the bounds of
29861 WIDE_INT_MAX_PRECISION.
29862 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
29863 the outermost component ref only to lower the access size
29864 and initialize that from the access type.
29866 2021-03-30 Richard Sandiford <richard.sandiford@arm.com>
29869 * config/aarch64/aarch64.md (mov<mode>): Pass multi-instruction
29870 CONST_INTs to aarch64_expand_mov_immediate when called after RA.
29872 2021-03-30 Mihailo Stojanovic <mihailo.stojanovic@typhoon-hil.com>
29874 * config/aarch64/aarch64.md
29875 (<optab>_trunc<fcvt_target><GPI:mode>2): Set the "arch"
29876 attribute to disambiguate between SIMD and FP variants of the
29879 2021-03-29 Jan Hubicka <hubicka@ucw.cz>
29881 * ipa-modref.c (merge_call_lhs_flags): Correct handling of deref.
29882 (analyze_ssa_name_flags): Fix typo in comment.
29884 2021-03-29 Alex Coplan <alex.coplan@arm.com>
29887 * config/aarch64/aarch64-sve-builtins.cc
29888 (function_builder::add_function): Add placeholder_p argument, use
29889 placeholder decls if this is set.
29890 (function_builder::add_unique_function): Instead of conditionally adding
29891 direct overloads, unconditionally add either a direct overload or a
29893 (function_builder::add_overloaded_function): Set placeholder_p if we're
29894 using C++ overloads. Use the obstack for string storage instead
29895 of relying on the tree nodes.
29896 (function_builder::add_overloaded_functions): Don't return early for
29897 m_direct_overloads: we need to add placeholders.
29898 * config/aarch64/aarch64-sve-builtins.h
29899 (function_builder::add_function): Add placeholder_p argument.
29901 2021-03-29 Richard Biener <rguenther@suse.de>
29903 PR tree-optimization/99807
29904 * tree-vect-slp.c (vect_slp_analyze_node_operations_1): Move
29905 assert below VEC_PERM handling.
29907 2021-03-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
29910 * config/aarch64/aarch64-simd.md (move_lo_quad_internal_<mode>): Use
29911 aarch64_simd_or_scalar_imm_zero to match zeroes. Remove pattern
29912 matching const_int 0.
29913 (move_lo_quad_internal_be_<mode>): Likewise.
29914 (move_lo_quad_<mode>): Update for the above.
29915 * config/aarch64/iterators.md (VQ_2E): Delete.
29917 2021-03-29 Jakub Jelinek <jakub@redhat.com>
29919 PR tree-optimization/99777
29920 * fold-const.c (extract_muldiv_1): For conversions, punt on casts from
29921 types other than scalar integral types.
29923 2021-03-28 David Edelsohn <dje.gcc@gmail.com>
29925 * config/rs6000/rs6000.c (rs6000_output_dwarf_dtprel): Do not add
29926 XCOFF TLS reloc decorations.
29928 2021-03-28 Gerald Pfeifer <gerald@pfeifer.com>
29930 * doc/analyzer.texi (Analyzer Internals): Update link to
29931 "A Memory Model for Static Analysis of C Programs".
29933 2021-03-26 David Edelsohn <dje.gcc@gmail.com>
29935 * config/rs6000/aix.h (ADJUST_FIELD_ALIGN): Call function.
29936 * config/rs6000/rs6000-protos.h (rs6000_special_adjust_field_align):
29938 * config/rs6000/rs6000.c (rs6000_special_adjust_field_align): New.
29939 (rs6000_special_round_type_align): Recursively check innermost first
29942 2021-03-26 Jakub Jelinek <jakub@redhat.com>
29945 * dwarf2out.h (struct dw_fde_node): Add rule18 member.
29946 * dwarf2cfi.c (dwarf2out_frame_debug_expr): When handling (set hfp sp)
29947 assignment with drap_reg active, queue reg save for hfp with offset 0
29948 and flush queued reg saves. When handling a push with rule18,
29949 defer queueing reg save for hfp and just assert the offset is 0.
29950 (scan_trace): Assert that fde->rule18 is false.
29952 2021-03-26 Vladimir Makarov <vmakarov@redhat.com>
29955 * ira-costs.c (record_reg_classes): Put case with
29956 CT_RELAXED_MEMORY adjacent to one with CT_MEMORY.
29957 * ira.c (ira_setup_alts): Ditto.
29958 * lra-constraints.c (process_alt_operands): Ditto.
29959 * recog.c (asm_operand_ok): Ditto.
29960 * reload.c (find_reloads): Ditto.
29962 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
29964 * config/aarch64/aarch64-protos.h
29965 (cpu_addrcost_table::post_modify_ld3_st3): New member variable.
29966 (cpu_addrcost_table::post_modify_ld4_st4): Likewise.
29967 * config/aarch64/aarch64.c (generic_addrcost_table): Update
29968 accordingly, using the same costs as for post_modify.
29969 (exynosm1_addrcost_table, xgene1_addrcost_table): Likewise.
29970 (thunderx2t99_addrcost_table, thunderx3t110_addrcost_table):
29971 (tsv110_addrcost_table, qdf24xx_addrcost_table): Likewise.
29972 (a64fx_addrcost_table): Likewise.
29973 (neoversev1_addrcost_table): New.
29974 (neoversev1_tunings): Use neoversev1_addrcost_table.
29975 (aarch64_address_cost): Use the new post_modify costs for CImode
29978 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
29980 * config/aarch64/aarch64.opt
29981 (-param=aarch64-loop-vect-issue-rate-niters=): New parameter.
29982 * doc/invoke.texi: Document it.
29983 * config/aarch64/aarch64-protos.h (aarch64_base_vec_issue_info)
29984 (aarch64_scalar_vec_issue_info, aarch64_simd_vec_issue_info)
29985 (aarch64_advsimd_vec_issue_info, aarch64_sve_vec_issue_info)
29986 (aarch64_vec_issue_info): New structures.
29987 (cpu_vector_cost): Write comments above the variables rather
29989 (cpu_vector_cost::issue_info): New member variable.
29990 * config/aarch64/aarch64.c: Include gimple-pretty-print.h
29991 and tree-ssa-loop-niter.h.
29992 (generic_vector_cost, a64fx_vector_cost, qdf24xx_vector_cost)
29993 (thunderx_vector_cost, tsv110_vector_cost, cortexa57_vector_cost)
29994 (exynosm1_vector_cost, xgene1_vector_cost, thunderx2t99_vector_cost)
29995 (thunderx3t110_vector_cost): Initialize issue_info to null.
29996 (neoversev1_scalar_issue_info, neoversev1_advsimd_issue_info)
29997 (neoversev1_sve_issue_info, neoversev1_vec_issue_info): New structures.
29998 (neoversev1_vector_cost): Use them.
29999 (aarch64_vec_op_count, aarch64_sve_op_count): New structures.
30000 (aarch64_vector_costs::saw_sve_only_op): New member variable.
30001 (aarch64_vector_costs::num_vector_iterations): Likewise.
30002 (aarch64_vector_costs::scalar_ops): Likewise.
30003 (aarch64_vector_costs::advsimd_ops): Likewise.
30004 (aarch64_vector_costs::sve_ops): Likewise.
30005 (aarch64_vector_costs::seen_loads): Likewise.
30006 (aarch64_simd_vec_costs_for_flags): New function.
30007 (aarch64_analyze_loop_vinfo): Initialize num_vector_iterations.
30008 Count the number of predicate operations required by SVE WHILE
30010 (aarch64_comparison_type, aarch64_multiply_add_p): New functions.
30011 (aarch64_sve_only_stmt_p, aarch64_in_loop_reduction_latency): Likewise.
30012 (aarch64_count_ops): Likewise.
30013 (aarch64_add_stmt_cost): Record whether see an SVE operation
30014 that cannot currently be implementing using Advanced SIMD.
30015 Record issue information about the scalar, Advanced SIMD
30016 and (where relevant) SVE versions of a loop.
30017 (aarch64_vec_op_count::dump): New function.
30018 (aarch64_sve_op_count::dump): Likewise.
30019 (aarch64_estimate_min_cycles_per_iter): Likewise.
30020 (aarch64_adjust_body_cost): If issue information is available,
30021 try to compare the issue rates of the various loop implementations
30022 and increase or decrease the vector body cost accordingly.
30024 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
30026 * config/aarch64/aarch64.c (aarch64_detect_vector_stmt_subtype):
30027 Assume a zero cost for induction phis.
30029 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
30031 * config/aarch64/aarch64.c (aarch64_embedded_comparison_type): New
30033 (aarch64_adjust_stmt_cost): Add the costs of embedded scalar and
30034 vector comparisons.
30036 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
30038 * config/aarch64/aarch64.c (aarch64_detect_scalar_stmt_subtype):
30040 (aarch64_add_stmt_cost): Call it.
30042 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
30044 * config/aarch64/aarch64-tuning-flags.def (matched_vector_throughput):
30045 New tuning parameter.
30046 * config/aarch64/aarch64.c (neoversev1_tunings): Use it.
30047 (aarch64_estimated_sve_vq): New function.
30048 (aarch64_vector_costs::analyzed_vinfo): New member variable.
30049 (aarch64_vector_costs::is_loop): Likewise.
30050 (aarch64_vector_costs::unrolled_advsimd_niters): Likewise.
30051 (aarch64_vector_costs::unrolled_advsimd_stmts): Likewise.
30052 (aarch64_record_potential_advsimd_unrolling): New function.
30053 (aarch64_analyze_loop_vinfo, aarch64_analyze_bb_vinfo): Likewise.
30054 (aarch64_add_stmt_cost): Call aarch64_analyze_loop_vinfo or
30055 aarch64_analyze_bb_vinfo on the first use of a costs structure.
30056 Detect whether we're vectorizing a loop for SVE that might be
30057 completely unrolled if it used Advanced SIMD instead.
30058 (aarch64_adjust_body_cost_for_latency): New function.
30059 (aarch64_finish_cost): Call it.
30061 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
30063 * config/aarch64/aarch64.c (aarch64_vector_costs): New structure.
30064 (aarch64_init_cost): New function.
30065 (aarch64_add_stmt_cost): Use aarch64_vector_costs instead of
30066 the default unsigned[3].
30067 (aarch64_finish_cost, aarch64_destroy_cost_data): New functions.
30068 (TARGET_VECTORIZE_INIT_COST): Override.
30069 (TARGET_VECTORIZE_FINISH_COST): Likewise.
30070 (TARGET_VECTORIZE_DESTROY_COST_DATA): Likewise.
30072 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
30074 * config/aarch64/aarch64.c (neoversev1_advsimd_vector_cost)
30075 (neoversev1_sve_vector_cost): New cost structures.
30076 (neoversev1_vector_cost): Likewise.
30077 (neoversev1_tunings): Use them. Enable use_new_vector_costs.
30079 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
30081 * config/aarch64/aarch64-protos.h
30082 (sve_vec_cost::scatter_store_elt_cost): New member variable.
30083 * config/aarch64/aarch64.c (generic_sve_vector_cost): Update
30084 accordingly, taking the cost from the cost of a scalar_store.
30085 (a64fx_sve_vector_cost): Likewise.
30086 (aarch64_detect_vector_stmt_subtype): Detect scatter stores.
30088 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
30090 * config/aarch64/aarch64-protos.h
30091 (simd_vec_cost::store_elt_extra_cost): New member variable.
30092 * config/aarch64/aarch64.c (generic_advsimd_vector_cost): Update
30093 accordingly, using the vec_to_scalar cost for the new field.
30094 (generic_sve_vector_cost, a64fx_advsimd_vector_cost): Likewise.
30095 (a64fx_sve_vector_cost, qdf24xx_advsimd_vector_cost): Likewise.
30096 (thunderx_advsimd_vector_cost, tsv110_advsimd_vector_cost): Likewise.
30097 (cortexa57_advsimd_vector_cost, exynosm1_advsimd_vector_cost)
30098 (xgene1_advsimd_vector_cost, thunderx2t99_advsimd_vector_cost)
30099 (thunderx3t110_advsimd_vector_cost): Likewise.
30100 (aarch64_detect_vector_stmt_subtype): Detect single-element stores.
30102 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
30104 * config/aarch64/aarch64-protos.h (simd_vec_cost::ld2_st2_permute_cost)
30105 (simd_vec_cost::ld3_st3_permute_cost): New member variables.
30106 (simd_vec_cost::ld4_st4_permute_cost): Likewise.
30107 * config/aarch64/aarch64.c (generic_advsimd_vector_cost): Update
30108 accordingly, using zero for the new costs.
30109 (generic_sve_vector_cost, a64fx_advsimd_vector_cost): Likewise.
30110 (a64fx_sve_vector_cost, qdf24xx_advsimd_vector_cost): Likewise.
30111 (thunderx_advsimd_vector_cost, tsv110_advsimd_vector_cost): Likewise.
30112 (cortexa57_advsimd_vector_cost, exynosm1_advsimd_vector_cost)
30113 (xgene1_advsimd_vector_cost, thunderx2t99_advsimd_vector_cost)
30114 (thunderx3t110_advsimd_vector_cost): Likewise.
30115 (aarch64_ld234_st234_vectors): New function.
30116 (aarch64_adjust_stmt_cost): Likewise.
30117 (aarch64_add_stmt_cost): Call aarch64_adjust_stmt_cost if using
30118 the new vector costs.
30120 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
30122 * config/aarch64/aarch64-protos.h (sve_vec_cost): Turn into a
30123 derived class of simd_vec_cost. Add information about CLAST[AB]
30124 and FADDA instructions.
30125 * config/aarch64/aarch64.c (generic_sve_vector_cost): Update
30126 accordingly, using the vec_to_scalar costs for the new fields.
30127 (a64fx_sve_vector_cost): Likewise.
30128 (aarch64_reduc_type): New function.
30129 (aarch64_sve_in_loop_reduction_latency): Likewise.
30130 (aarch64_detect_vector_stmt_subtype): Take a vinfo parameter.
30131 Use aarch64_sve_in_loop_reduction_latency to handle SVE reductions
30132 that occur in the loop body.
30133 (aarch64_add_stmt_cost): Update call accordingly.
30135 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
30137 * config/aarch64/aarch64-tuning-flags.def (use_new_vector_costs):
30139 * config/aarch64/aarch64-protos.h (simd_vec_cost): Put comments
30140 above the fields rather than to the right.
30141 (simd_vec_cost::reduc_i8_cost): New member variable.
30142 (simd_vec_cost::reduc_i16_cost): Likewise.
30143 (simd_vec_cost::reduc_i32_cost): Likewise.
30144 (simd_vec_cost::reduc_i64_cost): Likewise.
30145 (simd_vec_cost::reduc_f16_cost): Likewise.
30146 (simd_vec_cost::reduc_f32_cost): Likewise.
30147 (simd_vec_cost::reduc_f64_cost): Likewise.
30148 * config/aarch64/aarch64.c (generic_advsimd_vector_cost): Update
30149 accordingly, using the vec_to_scalar_cost for the new fields.
30150 (generic_sve_vector_cost, a64fx_advsimd_vector_cost): Likewise.
30151 (a64fx_sve_vector_cost, qdf24xx_advsimd_vector_cost): Likewise.
30152 (thunderx_advsimd_vector_cost, tsv110_advsimd_vector_cost): Likewise.
30153 (cortexa57_advsimd_vector_cost, exynosm1_advsimd_vector_cost)
30154 (xgene1_advsimd_vector_cost, thunderx2t99_advsimd_vector_cost)
30155 (thunderx3t110_advsimd_vector_cost): Likewise.
30156 (aarch64_use_new_vector_costs_p): New function.
30157 (aarch64_simd_vec_costs): New function, split out from...
30158 (aarch64_builtin_vectorization_cost): ...here.
30159 (aarch64_is_reduction): New function.
30160 (aarch64_detect_vector_stmt_subtype): Likewise.
30161 (aarch64_add_stmt_cost): Call aarch64_detect_vector_stmt_subtype if
30162 using the new vector costs.
30164 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
30167 * tree-emutls.c (get_emutls_init_templ_addr): Mark initializer of weak
30168 TLS declarations as public.
30170 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
30172 * config/aarch64/aarch64-d.c (IN_TARGET_CODE): Define.
30173 * config/arm/arm-d.c (IN_TARGET_CODE): Likewise.
30174 * config/i386/i386-d.c (IN_TARGET_CODE): Likewise.
30175 * config/mips/mips-d.c (IN_TARGET_CODE): Likewise.
30176 * config/pa/pa-d.c (IN_TARGET_CODE): Likewise.
30177 * config/riscv/riscv-d.c (IN_TARGET_CODE): Likewise.
30178 * config/rs6000/rs6000-d.c (IN_TARGET_CODE): Likewise.
30179 * config/s390/s390-d.c (IN_TARGET_CODE): Likewise.
30180 * config/sparc/sparc-d.c (IN_TARGET_CODE): Likewise.
30182 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
30185 * config.gcc (*-*-cygwin*): Add winnt-d.o
30186 (*-*-mingw*): Likewise.
30187 * config/i386/cygwin.h (EXTRA_TARGET_D_OS_VERSIONS): New macro.
30188 * config/i386/mingw32.h (EXTRA_TARGET_D_OS_VERSIONS): Likewise.
30189 * config/i386/t-cygming: Add winnt-d.o.
30190 * config/i386/winnt-d.c: New file.
30192 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
30194 * config/freebsd-d.c: Include memmodel.h.
30196 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
30199 * config.gcc (*-*-openbsd*): Add openbsd-d.o.
30200 * config/t-openbsd: Add openbsd-d.o.
30201 * config/openbsd-d.c: New file.
30203 2021-03-25 Stam Markianos-Wright <stam.markianos-wright@arm.com>
30205 PR tree-optimization/96974
30206 * tree-vect-stmts.c (vect_get_vector_types_for_stmt): Replace assert
30207 with graceful exit.
30209 2021-03-25 H.J. Lu <hjl.tools@gmail.com>
30212 2021-03-25 H.J. Lu <hjl.tools@gmail.com>
30216 * config/i386/i386.c (ix86_can_inline_p): Don't check ISA for
30217 always_inline in system headers.
30219 2021-03-25 Kewen Lin <linkw@linux.ibm.com>
30221 * tree-vect-loop.c (vect_model_reduction_cost): Init inside_cost.
30223 2021-03-25 Jakub Jelinek <jakub@redhat.com>
30226 * tree-core.h (enum operand_equal_flag): Add OEP_ADDRESS_OF_SAME_FIELD.
30227 * fold-const.c (operand_compare::operand_equal_p): Don't compare
30228 field offsets if OEP_ADDRESS_OF_SAME_FIELD.
30230 2021-03-25 H.J. Lu <hjl.tools@gmail.com>
30234 * config/i386/i386.c (ix86_can_inline_p): Don't check ISA for
30235 always_inline in system headers.
30237 2021-03-25 Richard Biener <rguenther@suse.de>
30239 PR tree-optimization/99746
30240 * tree-vect-slp-patterns.c (complex_pattern::build): Do not mark
30241 the scalar stmt as patterned. Instead set up required things
30244 2021-03-25 Xionghu Luo <luoxhu@linux.ibm.com>
30246 * config/rs6000/rs6000.c (power8_costs): Change l2 cache
30249 2021-03-24 Martin Liska <mliska@suse.cz>
30252 * common/config/i386/i386-common.c (ARRAY_SIZE): Fix off-by-one
30254 * config/i386/i386-options.c (ix86_option_override_internal):
30255 Add run-time assert.
30257 2021-03-24 Martin Jambor <mjambor@suse.cz>
30260 * ipa-cp.c (initialize_node_lattices): Mark as bottom all
30261 parameters with unknown type.
30262 (ipacp_value_safe_for_type): New function.
30263 (propagate_vals_across_arith_jfunc): Verify that the constant type
30264 can be used for a type of the formal parameter.
30265 (propagate_vals_across_ancestor): Likewise.
30266 (propagate_scalar_across_jump_function): Likewise. Pass the type
30267 also to propagate_vals_across_ancestor.
30269 2021-03-24 Christophe Lyon <christophe.lyon@linaro.org>
30272 * config/arm/mve.md (movmisalign<mode>_mve_store): Use Ux
30274 (movmisalign<mode>_mve_load): Likewise.
30276 2021-03-24 Jakub Jelinek <jakub@redhat.com>
30279 * config/arm/vec-common.md (one_cmpl<mode>2, neg<mode>2,
30280 movmisalign<mode>): Disable expanders for TARGET_REALLY_IWMMXT.
30282 2021-03-24 Alexandre Oliva <oliva@adacore.com>
30284 * doc/sourcebuild.texi (sysconf): New effective target.
30286 2021-03-24 Alexandre Oliva <oliva@adacore.com>
30288 * config/i386/predicates.md (reg_or_const_vec_operand): New.
30289 * config/i386/sse.md (ssse3_pshufbv8qi3): Add an expander for
30290 the now *-prefixed insn_and_split, turn the splitter const vec
30291 into an input for the insn, making it an ignored immediate for
30292 non-split cases, and loaded into the scratch register
30295 2021-03-23 Vladimir N. Makarov <vmakarov@redhat.com>
30298 * config/aarch64/constraints.md (Utq, UOb, UOh, UOw, UOd, UOty):
30299 Use define_relaxed_memory_constraint for them.
30301 2021-03-23 Iain Sandoe <iain@sandoe.co.uk>
30304 * config/host-darwin.c (darwin_gt_pch_use_address): Add a
30305 colon to the diagnostic message.
30307 2021-03-23 Ilya Leoshkevich <iii@linux.ibm.com>
30309 * fwprop.c (fwprop_propagation::fwprop_propagation): Look at
30311 (try_fwprop_subst_note): Use set_info instead of insn_info.
30312 (try_fwprop_subst_pattern): Likewise.
30313 (try_fwprop_subst_notes): Likewise.
30314 (try_fwprop_subst): Likewise.
30315 (forward_propagate_subreg): Likewise.
30316 (forward_propagate_and_simplify): Likewise.
30317 (forward_propagate_into): Likewise.
30318 * rtl-ssa/accesses.h (set_info::single_nondebug_use) New
30320 (set_info::single_nondebug_insn_use): Likewise.
30321 (set_info::single_phi_use): Likewise.
30322 * rtl-ssa/member-fns.inl (set_info::single_nondebug_use) New
30324 (set_info::single_nondebug_insn_use): Likewise.
30325 (set_info::single_phi_use): Likewise.
30327 2021-03-23 Christophe Lyon <christophe.lyon@linaro.org>
30329 * doc/sourcebuild.texi (arm_dsp_ok, arm_dsp): Document.
30331 2021-03-23 Jakub Jelinek <jakub@redhat.com>
30334 * config/aarch64/aarch64.c (aarch64_add_offset): Tell
30335 expand_mult to perform an unsigned rather than a signed
30338 2021-03-23 H.J. Lu <hjl.tools@gmail.com>
30341 * config/i386/cpuid.h (__cpuid): Add __volatile__.
30342 (__cpuid_count): Likewise.
30344 2021-03-23 Richard Biener <rguenther@suse.de>
30346 PR tree-optimization/99721
30347 * tree-vect-slp.c (vect_slp_analyze_node_operations):
30348 Make sure we can schedule the node.
30350 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
30352 * config/riscv/riscv.c (riscv_subword): Take endianness into
30353 account when calculating the byte offset.
30355 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
30357 * config/riscv/predicates.md (subreg_lowpart_operator): New predicate
30358 * config/riscv/riscv.md (*addsi3_extended2, *subsi3_extended2)
30359 (*negsi2_extended2, *mulsi3_extended2, *<optab>si3_mask)
30360 (*<optab>si3_mask_1, *<optab>di3_mask, *<optab>di3_mask_1)
30361 (*<optab>si3_extend_mask, *<optab>si3_extend_mask_1): Use
30362 new predicate "subreg_lowpart_operator"
30364 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
30366 * config/riscv/riscv.c (riscv_swap_instruction): New function
30367 to byteswap an SImode rtx containing an instruction.
30368 (riscv_trampoline_init): Byteswap the generated instructions
30371 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
30373 * common/config/riscv/riscv-common.c
30374 (TARGET_DEFAULT_TARGET_FLAGS): Set default endianness.
30375 * config.gcc (riscv32be-*, riscv64be-*): Set
30376 TARGET_BIG_ENDIAN_DEFAULT to 1.
30377 * config/riscv/elf.h (LINK_SPEC): Change -melf* value
30378 depending on default endianness.
30379 * config/riscv/freebsd.h (LINK_SPEC): Likewise.
30380 * config/riscv/linux.h (LINK_SPEC): Likewise.
30381 * config/riscv/riscv.c (TARGET_DEFAULT_TARGET_FLAGS): Set
30382 default endianness.
30383 * config/riscv/riscv.h (DEFAULT_ENDIAN_SPEC): New macro.
30385 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
30387 * config/riscv/elf.h (LINK_SPEC): Pass linker endianness flag.
30388 * config/riscv/freebsd.h (LINK_SPEC): Likewise.
30389 * config/riscv/linux.h (LINK_SPEC): Likewise.
30390 * config/riscv/riscv.h (ASM_SPEC): Pass -mbig-endian and
30392 (BYTES_BIG_ENDIAN): Handle big endian.
30393 (WORDS_BIG_ENDIAN): Define to BYTES_BIG_ENDIAN.
30394 * config/riscv/riscv.opt (-mbig-endian, -mlittle-endian): New
30396 * doc/invoke.texi (-mbig-endian, -mlittle-endian): Document.
30398 2021-03-23 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
30400 * regcprop.c (find_oldest_value_reg): Ask target whether
30401 different mode is fine for replacement register.
30403 2021-03-23 Aldy Hernandez <aldyh@redhat.com>
30405 PR tree-optimization/99296
30406 * value-range.cc (irange::irange_set_1bit_anti_range): New.
30407 (irange::irange_set_anti_range): Call irange_set_1bit_anti_range
30408 * value-range.h (irange::irange_set_1bit_anti_range): New.
30410 2021-03-22 Vladimir N. Makarov <vmakarov@redhat.com>
30413 * config/aarch64/constraints.md (UtQ): Use
30414 define_relaxed_memory_constraint for it.
30415 * doc/md.texi (define_relaxed_memory_constraint): Describe it.
30416 * genoutput.c (main): Process DEFINE_RELAXED_MEMORY_CONSTRAINT.
30417 * genpreds.c (constraint_data): Add bitfield is_relaxed_memory.
30418 (have_relaxed_memory_constraints): New static var.
30419 (relaxed_memory_start, relaxed_memory_end): Ditto.
30420 (add_constraint): Add arg is_relaxed_memory. Check name for
30421 relaxed memory. Set up is_relaxed_memory in constraint_data and
30422 have_relaxed_memory_constraints. Adjust calls.
30423 (choose_enum_order): Process relaxed memory.
30424 (write_tm_preds_h): Ditto.
30425 (main): Process DEFINE_RELAXED_MEMORY_CONSTRAINT.
30426 * gensupport.c (process_rtx): Process DEFINE_RELAXED_MEMORY_CONSTRAINT.
30427 * ira-costs.c (record_reg_classes): Process CT_RELAXED_MEMORY.
30428 * ira-lives.c (single_reg_class): Use
30429 insn_extra_relaxed_memory_constraint.
30430 * ira.c (ira_setup_alts): CT_RELAXED_MEMORY.
30431 * lra-constraints.c (valid_address_p): Use
30432 insn_extra_relaxed_memory_constraint instead of other memory
30434 (process_alt_operands): Process CT_RELAXED_MEMORY.
30435 (curr_insn_transform): Use insn_extra_relaxed_memory_constraint.
30436 * recog.c (asm_operand_ok, preprocess_constraints): Process
30438 * reload.c (find_reloads): Ditto.
30439 * rtl.def (DEFINE_RELAXED_MEMORY_CONSTRAINT): New.
30440 * stmt.c (parse_input_constraint): Use
30441 insn_extra_relaxed_memory_constraint.
30443 2021-03-22 Segher Boessenkool <segher@kernel.crashing.org>
30446 * ubsan.c (ubsan_instrument_float_cast): Don't test for unordered if
30449 2021-03-22 Alex Coplan <alex.coplan@arm.com>
30452 * config/arm/arm-protos.h (neon_make_constant): Add generate
30453 argument to guard emitting insns, default to true.
30454 * config/arm/arm.c (arm_legitimate_constant_p_1): Reject
30455 CONST_VECTORs which neon_make_constant can't handle.
30456 (neon_vdup_constant): Add generate argument, avoid emitting
30457 insns if it's not set.
30458 (neon_make_constant): Plumb new generate argument through.
30459 * config/arm/constraints.md (Ui): New. Use it...
30460 * config/arm/mve.md (*mve_mov<mode>): ... here.
30461 * config/arm/vec-common.md (movv8hf): Use neon_make_constant to
30462 synthesize constants.
30464 2021-03-22 Richard Biener <rguenther@suse.de>
30466 * debug.h: Add deprecation warning.
30468 2021-03-22 Richard Biener <rguenther@suse.de>
30470 PR tree-optimization/99694
30471 * tree-ssa-sccvn.c (visit_phi): Ignore edges with the
30474 2021-03-22 Kito Cheng <kito.cheng@sifive.com>
30477 * config/riscv/riscv.c (riscv_expand_block_move): Get RTL value
30478 after type checking.
30480 2021-03-22 Jakub Jelinek <jakub@redhat.com>
30484 * dwarf2out.c (get_full_len): Use get_precision rather than
30486 (add_const_value_attribute): Make sure add_AT_wide argument has
30487 precision prec rather than some very wide one.
30489 2021-03-22 Kewen Lin <linkw@linux.ibm.com>
30491 * config/rs6000/rs6000.md (*rotldi3_insert_sf,
30492 *mov<SFDF:mode><SFDF2:mode>cc_p9, floatsi<mode>2_lfiwax,
30493 floatsi<mode>2_lfiwax_mem, floatunssi<mode>2_lfiwzx,
30494 floatunssi<mode>2_lfiwzx_mem, *floatsidf2_internal,
30495 *floatunssidf2_internal, fix_trunc<mode>si2_stfiwx,
30496 fix_trunc<mode>si2_internal, fixuns_trunc<mode>si2_stfiwx,
30497 *round32<mode>2_fprs, *roundu32<mode>2_fprs,
30498 *fix_trunc<mode>si2_internal): Fix empty split condition.
30499 * config/rs6000/vsx.md (*vsx_le_undo_permute_<mode>,
30500 vsx_reduc_<VEC_reduc_name>_v2df, vsx_reduc_<VEC_reduc_name>_v4sf,
30501 *vsx_reduc_<VEC_reduc_name>_v2df_scalar,
30502 *vsx_reduc_<VEC_reduc_name>_v4sf_scalar): Likewise.
30504 2021-03-22 Xionghu Luo <luoxhu@linux.ibm.com>
30507 * config/rs6000/rs6000.c (rs6000_expand_vector_set_var_p9):
30508 Convert idx to DImode.
30509 (rs6000_expand_vector_set_var_p8): Likewise.
30511 2021-03-21 Jakub Jelinek <jakub@redhat.com>
30514 * dwarf2out.c (insert_float): Change return type from void to
30515 unsigned, handle GET_MODE_SIZE (mode) == 2 and return element size.
30516 (mem_loc_descriptor, loc_descriptor, add_const_value_attribute):
30519 2021-03-20 H.J. Lu <hjl.tools@gmail.com>
30522 * config/i386/i386.c (construct_container): Check cfun != NULL
30523 before accessing silent_p.
30525 2021-03-20 Ahamed Husni <ahamedhusni73@gmail.com>
30527 * asan.c: Fix typos in comments.
30529 2021-03-20 Vladimir N. Makarov <vmakarov@redhat.com>
30531 PR rtl-optimization/99680
30532 * lra-constraints.c (skip_contraint_modifiers): Rename to skip_constraint_modifiers.
30533 (process_address_1): Check empty constraint before using
30536 2021-03-19 Pat Haugen <pthaugen@linux.ibm.com>
30538 * config/rs6000/rs6000.c (power10_cost): New.
30539 (rs6000_option_override_internal): Set Power10 costs.
30540 (rs6000_issue_rate): Set Power10 issue rate.
30541 * config/rs6000/power10.md: Rewrite for Power10.
30543 2021-03-19 Vladimir N. Makarov <vmakarov@redhat.com>
30546 * lra-constraints.c (process_address_1): Don't use unknown
30547 constraint for address constraint.
30549 2021-03-19 Iain Sandoe <iain@sandoe.co.uk>
30552 * config.gcc (powerpc-*-darwin8): Delete the reference to
30553 the now removed darwin8.h.
30555 2021-03-19 Olivier Hainque <hainque@adacore.com>
30558 * config/vxworksae.h (VX_CPU_PREFIX): Define.
30560 2021-03-19 John David Anglin <danglin@gcc.gnu.org>
30562 * config/pa/pa.c (import_milli): Use memcpy instead of strncpy.
30564 2021-03-19 Tamar Christina <tamar.christina@arm.com>
30566 PR tree-optimization/99656
30567 * tree-vect-slp-patterns.c (linear_loads_p,
30568 complex_add_pattern::matches, is_eq_or_top,
30569 vect_validate_multiplication, complex_mul_pattern::matches,
30570 complex_fms_pattern::matches): Remove complex_perm_kinds_t.
30571 * tree-vectorizer.h: (complex_load_perm_t): Removed.
30572 (slp_tree_to_load_perm_map_t): Use complex_perm_kinds_t instead of
30573 complex_load_perm_t.
30575 2021-03-19 H.J. Lu <hjl.tools@gmail.com>
30578 * config/i386/i386-options.c (ix86_init_machine_status): Set
30580 * config/i386/i386.c (init_cumulative_args): Set silent_p to
30582 (construct_container): Return early for return and argument
30583 errors if silent_p is true.
30584 * config/i386/i386.h (machine_function): Add silent_p.
30586 2021-03-19 Jakub Jelinek <jakub@redhat.com>
30589 * config/arm/constraints.md (Ds): New constraint.
30590 * config/arm/vec-common.md (mve_vshlq_<supf><mode>): Use w,Ds
30591 constraint instead of w,Dm.
30593 2021-03-19 Andrew Stubbs <ams@codesourcery.com>
30595 * config/gcn/gcn.c (gcn_parse_amdgpu_hsa_kernel_attribute): Fix quotes
30598 2021-03-19 Eric Botcazou <ebotcazou@adacore.com>
30600 PR middle-end/99641
30601 * fold-const.c (native_encode_initializer) <CONSTRUCTOR>: For an
30602 array type, do the computation of the current position in sizetype.
30604 2021-03-18 Vladimir N. Makarov <vmakarov@redhat.com>
30607 * lra-constraints.c (process_address_1): Use lookup_constraint
30608 only for a single constraint.
30610 2021-03-18 Martin Sebor <msebor@redhat.com>
30612 PR middle-end/99502
30613 * gimple-array-bounds.cc (inbounds_vbase_memaccess_p): Rename...
30614 (inbounds_memaccess_p): ...to this. Check the ending offset of
30615 the accessed member.
30617 2021-03-18 Andrew Stubbs <ams@codesourcery.com>
30619 * config/gcn/gcn.c (gcn_parse_amdgpu_hsa_kernel_attribute): Add %< and
30620 %> quote markers to error messages.
30621 (gcn_goacc_validate_dims): Likewise.
30622 (gcn_conditional_register_usage): Remove exclaimation mark from error
30624 (gcn_vectorize_vec_perm_const): Ensure perm is fully uninitialized.
30626 2021-03-18 Jan Hubicka <hubicka@ucw.cz>
30628 * config/i386/x86-tune-costs.h (struct processor_costs): Fix costs of
30631 2021-03-18 Sinan Lin <sinan@isrc.iscas.ac.cn>
30632 Kito Cheng <kito.cheng@sifive.com>
30634 * config/riscv/riscv.c (riscv_block_move_straight): Change type
30635 to unsigned HOST_WIDE_INT for parameter and local variable with
30636 HOST_WIDE_INT type.
30637 (riscv_adjust_block_mem): Ditto.
30638 (riscv_block_move_loop): Ditto.
30639 (riscv_expand_block_move): Ditto.
30641 2021-03-18 Nick Clifton <nickc@redhat.com>
30643 * config/v850/v850.c (construct_restore_jr): Increase static
30645 (construct_save_jarl): Likewise.
30646 * config/v850/v850.h (DWARF2_DEBUGGING_INFO): Define.
30648 2021-03-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
30650 * config/aarch64/aarch64.c (aarch64_adjust_generic_arch_tuning): Define.
30651 (aarch64_override_options_internal): Use it.
30652 (generic_tunings): Add AARCH64_EXTRA_TUNE_CSE_SVE_VL_CONSTANTS to
30655 2021-03-17 Sandra Loosemore <sandra@codesourcery.com>
30657 * config/nios2/nios2.c (nios2_custom_check_insns): Clean up
30658 error message format issues.
30659 (nios2_option_override): Likewise.
30660 (nios2_expand_fpu_builtin): Likewise.
30661 (nios2_init_custom_builtins): Adjust to avoid bogus strncpy
30662 truncation warning.
30663 (nios2_expand_custom_builtin): More error message format fixes.
30664 (nios2_expand_rdwrctl_builtin): Likewise.
30665 (nios2_expand_rdprs_builtin): Likewise.
30666 (nios2_expand_eni_builtin): Likewise.
30667 (nios2_expand_builtin): Likewise.
30668 (nios2_register_custom_code): Likewise.
30669 (nios2_valid_target_attribute_rec): Likewise.
30670 (nios2_add_insn_asm): Fix uninitialized variable warning.
30672 2021-03-17 Jan Hubicka <jh@suse.cz>
30674 * config/i386/x86-tune-costs.h (struct processor_costs): Update costs
30675 of gather to match reality.
30676 * config/i386/x86-tune.def (X86_TUNE_USE_GATHER): Enable for znver3.
30678 2021-03-17 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
30680 * config/aarch64/aarch64-builtins.c (aarch64_expand_rng_builtin): Use EQ
30681 to compare against CC_REG rather than NE.
30683 2021-03-17 H.J. Lu <hjl.tools@gmail.com>
30686 * config/i386/i386.c (ix86_force_load_from_GOT_p): Support
30687 inline assembly statements.
30688 (ix86_print_operand): Update 'P' handling for -fno-plt.
30690 2021-03-17 Tamar Christina <tamar.christina@arm.com>
30693 * config/aarch64/aarch64.c
30694 (aarch64_simd_clone_compute_vecsize_and_simdlen): Remove unused var.
30696 2021-03-16 Segher Boessenkool <segher@kernel.crashing.org>
30699 * config/rs6000/predicates.md (branch_comparison_operator): Allow
30700 ordered and unordered for CCFPmode, if flag_finite_math_only.
30702 2021-03-16 Jakub Jelinek <jakub@redhat.com>
30705 * config/i386/i386-expand.c (ix86_split_lea_for_addr): Emit a MULT
30706 rather than ASHIFT.
30707 * config/i386/i386.md (mult by 1248 into ashift): New splitter.
30709 2021-03-16 Martin Liska <mliska@suse.cz>
30712 * optc-save-gen.awk: Add flag_ipa_ra to exceptions for
30713 cl_optimization_compare function.
30715 2021-03-16 Ilya Leoshkevich <iii@linux.ibm.com>
30717 * config/s390/s390.c (f_constraint_p): Treat "fv" constraints
30720 2021-03-16 Jakub Jelinek <jakub@redhat.com>
30723 * config/i386/i386.h (struct machine_function): Add
30724 has_explicit_vzeroupper bitfield.
30725 * config/i386/i386-expand.c (ix86_expand_builtin): Set
30726 cfun->machine->has_explicit_vzeroupper when expanding
30727 IX86_BUILTIN_VZEROUPPER.
30728 * config/i386/i386-features.c (rest_of_handle_insert_vzeroupper):
30729 Do the mode switching only when TARGET_VZEROUPPER, expensive
30730 optimizations turned on and not optimizing for size.
30731 (pass_insert_vzeroupper::gate): Enable even when
30732 cfun->machine->has_explicit_vzeroupper is set.
30734 2021-03-16 Jakub Jelinek <jakub@redhat.com>
30737 * config/aarch64/aarch64.c
30738 (aarch64_simd_clone_compute_vecsize_and_simdlen): If not a function
30739 definition, walk TYPE_ARG_TYPES list if non-NULL for argument types
30740 instead of DECL_ARGUMENTS. Ignore types for uniform arguments.
30742 2021-03-15 Richard Biener <rguenther@suse.de>
30744 PR tree-optimization/98834
30745 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle missing
30746 subsetting by truncating the access size.
30748 2021-03-15 Jan Hubicka <hubicka@ucw.cz>
30750 * config/i386/i386-options.c (processor_cost_table): Add znver3_cost.
30751 * config/i386/x86-tune-costs.h (znver3_cost): New gobal variable; copy
30754 2021-03-15 Martin Liska <mliska@suse.cz>
30756 * spellcheck.c: Add missing comma in initialization.
30758 2021-03-14 Uroš Bizjak <ubizjak@gmail.com>
30760 * config/i386/sse.md (*vec_extract<mode>): Merge alternative 0 with
30761 alternative 2 and alternative 1 with alternative 3 using
30762 YW register constraint.
30763 (*vec_extract<PEXTR_MODE12:mode>_zext): Merge alternatives
30764 using YW register constraint.
30765 (*vec_extractv16qi_zext): Ditto.
30766 (*vec_extractv4si): Merge alternatives 4 and 5
30767 using Yw register constraint.
30768 (*ssse3_palignr<mode>_perm): Use Yw instead of v for alternative 3.
30770 2021-03-13 Martin Sebor <msebor@redhat.com>
30772 PR tree-optimization/99489
30773 * builtins.c (gimple_call_alloc_size): Fail gracefully when argument
30774 is not a call statement.
30776 2021-03-13 Jakub Jelinek <jakub@redhat.com>
30778 PR tree-optimization/99544
30779 * match.pd (X + (X << C) -> X * (1 + (1 << C))): Don't simplify
30780 if for vector types multiplication can't be done in type's mode.
30782 2021-03-12 Eric Botcazou <ebotcazou@adacore.com>
30785 * config/sparc/constraints.md (w): Rename to...
30786 (W): ... this and ditch previous implementation.
30787 * config/sparc/sparc.md (*movdi_insn_sp64): Replace W with m.
30788 (*movdf_insn_sp64): Likewise.
30789 (*mov<VM64:mode>_insn_sp64): Likewise.
30790 * config/sparc/sync.md (*atomic_compare_and_swap<mode>_1): Replace
30792 (atomic_compare_and_swap_leon3_1): Likewise.
30793 (*atomic_compare_and_swapdi_v8plus): Likewise.
30794 * config/sparc/sparc.c (memory_ok_for_ldd): Remove useless test on
30795 architecture and add missing address validity check during LRA.
30797 2021-03-12 Tobias Burnus <tobias@codesourcery.com>
30800 * gimplify.c (omp_add_variable): Handle NULL_TREE as size
30801 occuring for assumed-size arrays in use_device_{ptr,addr}.
30803 2021-03-12 Jakub Jelinek <jakub@redhat.com>
30806 * config/i386/constraints.md (YW): New internal constraint.
30807 * config/i386/sse.md (v_Yw): Add V4TI, V2TI, V1TI and TI cases.
30808 (*<sse2_avx2>_<insn><mode>3<mask_name>,
30809 *<sse2_avx2>_uavg<mode>3<mask_name>, *abs<mode>2,
30810 *<s>mul<mode>3_highpart<mask_name>): Use <v_Yw> instead of v in
30812 (<sse2_avx2>_psadbw): Use YW instead of v in constraints.
30813 (*avx2_pmaddwd, *sse2_pmaddwd, *<code>v8hi3, *<code>v16qi3,
30814 avx2_pmaddubsw256, ssse3_pmaddubsw128): Merge last two alternatives
30815 into one, use Yw instead of former x,v.
30816 (ashr<mode>3, <insn><mode>3): Use <v_Yw> instead of x in constraints of
30817 the last alternative.
30818 (<sse2_avx2>_packsswb<mask_name>, <sse2_avx2>_packssdw<mask_name>,
30819 <sse2_avx2>_packuswb<mask_name>, <sse4_1_avx2>_packusdw<mask_name>,
30820 *<ssse3_avx2>_pmulhrsw<mode>3<mask_name>, <ssse3_avx2>_palignr<mode>,
30821 <ssse3_avx2>_pshufb<mode>3<mask_name>): Merge last two alternatives
30822 into one, use <v_Yw> instead of former x,v.
30823 (avx2_interleave_highv32qi<mask_name>,
30824 vec_interleave_highv16qi<mask_name>): Use Yw instead of v in
30825 constraints. Add && <mask_avx512bw_condition> to condition.
30826 (avx2_interleave_lowv32qi<mask_name>,
30827 vec_interleave_lowv16qi<mask_name>,
30828 avx2_interleave_highv16hi<mask_name>,
30829 vec_interleave_highv8hi<mask_name>,
30830 avx2_interleave_lowv16hi<mask_name>, vec_interleave_lowv8hi<mask_name>,
30831 avx2_pshuflw_1<mask_name>, sse2_pshuflw_1<mask_name>,
30832 avx2_pshufhw_1<mask_name>, sse2_pshufhw_1<mask_name>,
30833 avx2_<code>v16qiv16hi2<mask_name>, sse4_1_<code>v8qiv8hi2<mask_name>,
30834 *sse4_1_<code>v8qiv8hi2<mask_name>_1, <sse2_avx2>_<insn><mode>3): Use
30835 Yw instead of v in constraints.
30836 * config/i386/mmx.md (Yv_Yw): New define_mode_attr.
30837 (*mmx_<insn><mode>3, mmx_ashr<mode>3, mmx_<insn><mode>3): Use <Yv_Yw>
30838 instead of Yv in constraints.
30839 (*mmx_<insn><mode>3, *mmx_mulv4hi3, *mmx_smulv4hi3_highpart,
30840 *mmx_umulv4hi3_highpart, *mmx_pmaddwd, *mmx_<code>v4hi3,
30841 *mmx_<code>v8qi3, mmx_pack<s_trunsuffix>swb, mmx_packssdw,
30842 mmx_punpckhbw, mmx_punpcklbw, mmx_punpckhwd, mmx_punpcklwd,
30843 *mmx_uavgv8qi3, *mmx_uavgv4hi3, mmx_psadbw): Use Yw instead of Yv in
30845 (*mmx_pinsrw, *mmx_pinsrb, *mmx_pextrw, *mmx_pextrw_zext, *mmx_pextrb,
30846 *mmx_pextrb_zext): Use YW instead of Yv in constraints.
30847 (*mmx_eq<mode>3, mmx_gt<mode>3): Use x instead of Yv in constraints.
30848 (mmx_andnot<mode>3, *mmx_<code><mode>3): Split last alternative into
30849 two, one with just x, another isa avx512vl with v.
30851 2021-03-12 Martin Liska <mliska@suse.cz>
30853 * doc/invoke.texi: Add missing param documentation.
30855 2021-03-11 David Malcolm <dmalcolm@redhat.com>
30858 * Makefile.in (ANALYZER_OBJS): Add analyzer/feasible-graph.o and
30859 analyzer/trimmed-graph.o.
30860 * doc/analyzer.texi (Analyzer Paths): Rewrite description of
30861 feasibility checking to reflect new implementation.
30862 * doc/invoke.texi (-fdump-analyzer-feasibility): Document new
30864 * shortest-paths.h (shortest_paths::get_shortest_distance): New.
30866 2021-03-11 David Malcolm <dmalcolm@redhat.com>
30868 * digraph.cc (selftest::test_shortest_paths): Update
30869 shortest_paths init for new param. Add test of
30870 SPS_TO_GIVEN_TARGET.
30871 * shortest-paths.h (enum shortest_path_sense): New.
30872 (shortest_paths::shortest_paths): Add "sense" param.
30873 Update for renamings. Generalize to use "sense" param.
30874 (shortest_paths::get_shortest_path): Rename param.
30875 (shortest_paths::m_sense): New field.
30876 (shortest_paths::m_prev): Rename...
30877 (shortest_paths::m_best_edge): ...to this.
30878 (shortest_paths::get_shortest_path): Update for renamings.
30879 Conditionalize flipping of path on sense of traversal.
30881 2021-03-11 David Malcolm <dmalcolm@redhat.com>
30883 * digraph.cc (selftest::test_shortest_paths): Add test coverage
30884 for paths from B and C.
30885 * shortest-paths.h (shortest_paths::shortest_paths): Handle
30886 unreachable nodes, rather than asserting.
30888 2021-03-11 David Edelsohn <dje.gcc@gmail.com>
30891 * config/rs6000/rs6000.c (rs6000_xcoff_file_start): Don't create
30892 xcoff_tbss_section_name.
30893 * config/rs6000/xcoff.h (ASM_OUTPUT_TLS_COMMON): Use .lcomm.
30894 * xcoffout.c (xcoff_tbss_section_name): Delete.
30895 * xcoffout.h (xcoff_tbss_section_name): Delete.
30897 2021-03-11 Richard Biener <rguenther@suse.de>
30899 PR tree-optimization/99523
30900 * tree-cfg.c (dump_function_to_file): Dump SSA names
30901 w/o identifier to the decls section as well, not only those
30902 without a VAR_DECL.
30904 2021-03-11 Jakub Jelinek <jakub@redhat.com>
30907 * ipa-icf-gimple.c (func_checker::compare_gimple_call): For internal
30908 function calls with lhs fail if the lhs don't have compatible types.
30910 2021-03-11 Hans-Peter Nilsson <hp@axis.com>
30912 * config/cris/cris.h (HARD_FRAME_POINTER_REGNUM): Define.
30913 Change FRAME_POINTER_REGNUM to correspond to a new faked
30914 register faked_fp, part of GENNONACR_REGS like faked_ap.
30915 (CRIS_FAKED_REGS_CONTENTS): New helper macro.
30916 (FIRST_PSEUDO_REGISTER, FIXED_REGISTERS, CALL_USED_REGISTERS):
30917 (REG_ALLOC_ORDER, REG_CLASS_CONTENTS, REGNO_OK_FOR_BASE_P)
30918 (ELIMINABLE_REGS, REGISTER_NAMES): Adjust accordingly.
30919 * config/cris/cris.md (CRIS_FP_REGNUM): Renumber to new faked
30921 (CRIS_REAL_FP_REGNUM): New constant.
30922 * config/cris/cris.c (cris_reg_saved_in_regsave_area): Check
30923 for HARD_FRAME_POINTER_REGNUM instead of FRAME_POINTER_REGNUM.
30924 (cris_initial_elimination_offset): Handle elimination changes
30925 to HARD_FRAME_POINTER_REGNUM instead of FRAME_POINTER_REGNUM
30926 and add one from FRAME_POINTER_REGNUM to
30927 HARD_FRAME_POINTER_REGNUM.
30928 (cris_expand_prologue, cris_expand_epilogue): Emit code for
30929 hard_frame_pointer_rtx instead of frame_pointer_rtx.
30931 2021-03-10 David Edelsohn <dje.gcc@gmail.com>
30934 * config/rs6000/aix.h (ADJUST_FIELD_ALIGN): Add check for DCmode.
30935 * config/rs6000/rs6000.c (rs6000_special_round_type_align): Same.
30937 2021-03-10 Vladimir N. Makarov <vmakarov@redhat.com>
30940 * lra-constraints.c (process_address_1): Don't check unknown
30941 constraint, use X for empty constraint.
30943 2021-03-10 Alex Coplan <alex.coplan@arm.com>
30945 * config/aarch64/aarch64.c (aarch64_vfp_is_call_or_return_candidate):
30946 Fix typo in comment describing "is_ha" argument.
30948 2021-03-10 John David Anglin <danglin@gcc.gnu.org>
30950 * doc/sourcebuild.texi: Document LRA target selector.
30952 2021-03-10 David Malcolm <dmalcolm@redhat.com>
30954 * doc/ux.texi: Add subsection contrasting interactive versus
30955 batch usage of GCC.
30957 2021-03-10 Joel Hutton <joel.hutton@arm.com>
30960 * tree-vect-stmts.c (vectorizable_store): Fix scatter store mask
30962 (vectorizable_load): Fix gather load mask check condition.
30964 2021-03-10 Richard Biener <rguenther@suse.de>
30966 PR tree-optimization/99510
30967 * tree.c (check_aligned_type): Check that the candidate
30968 has TYPE_USER_ALIGN set instead of matching with the
30971 2021-03-10 Eric Botcazou <ebotcazou@adacore.com>
30973 * config/sparc/sparc.c (sparc_regmode_natural_size): Return 4 for
30974 float and vector integer modes only if the mode is not larger.
30976 2021-03-10 Hans-Peter Nilsson <hp@axis.com>
30978 * config/cris/cris.h (DWARF_FRAME_REGISTERS): Define.
30980 2021-03-09 Vladimir N. Makarov <vmakarov@redhat.com>
30982 * ira.c (ira_setup_alts, ira_get_dup_out_num): Process digital
30984 * ira-lives.c (single_reg_class): Ditto.
30986 2021-03-09 Sebastian Huber <sebastian.huber@embedded-brains.de>
30988 * config.gcc (aarch64-*-rtems*): Include general rtems.h after
30989 the architecture-specific rtems.h.
30990 (aarch64-*-rtems*): Likewise.
30991 (arm*-*-rtems*): Likewise.
30992 (epiphany-*-rtems*): Likewise.
30993 (riscv*-*-rtems*): Likewise.
30995 2021-03-09 Jakub Jelinek <jakub@redhat.com>
30997 PR tree-optimization/99305
30998 * tree-ssa-phiopt.c (conditional_replacement): Test integer_pow2p
30999 before integer_all_onesp instead of vice versa.
31001 2021-03-09 Richard Earnshaw <rearnsha@arm.com>
31003 * common/config/arm/arm-common.c (arm_config_default): Change type
31004 of 'i' to unsigned.
31006 2021-03-09 Vladimir N. Makarov <vmakarov@redhat.com>
31009 * lra-constraints.c (process_address_1): Process constraint 'g'
31010 separately and digital constraints containing more one digit.
31012 2021-03-09 Nick Clifton <nickc@redhat.com>
31014 * config/rx/rx.h (DBX_DEBUGGING_INFO): Define.
31015 (DWARF"_DEBUGGING_INFO): Define.
31017 2021-03-09 Eric Botcazou <ebotcazou@adacore.com>
31020 * calls.c (initialize_argument_information): When the argument
31021 is passed by reference, do not make a copy in a thunk only if
31022 the argument is already in memory. Remove redundant test for
31023 the case of callee copy.
31025 2021-03-09 Vladimir N. Makarov <vmakarov@redhat.com>
31028 * lra-constraints.c (process_address_1): Process 0..9 constraints
31029 in process_address_1.
31031 2021-03-09 Andreas Krebbel <krebbel@linux.ibm.com>
31033 * config/s390/s390.c (struct s390_processor processor_table):
31034 Binutils name string must not be empty.
31036 2021-03-09 Claudiu Zissulescu <claziss@synopsys.com>
31038 * config/arc/arc.c (arc_attr_type): Remove function.
31040 2021-03-09 Martin Liska <mliska@suse.cz>
31043 * config/i386/i386-options.c (ix86_option_override_internal):
31044 Set isa_flags for OPTS argument and not for the global
31047 2021-03-09 Aaron Sawdey <acsawdey@linux.ibm.com>
31049 * config/rs6000/predicates.md (ds_form_mem_operand): Check
31052 2021-03-09 Aaron Sawdey <acsawdey@linux.ibm.com>
31055 * config/rs6000/predicates.md (ds_form_mem_operand) New
31057 * config/rs6000/genfusion.pl (gen_ld_cmpi_p10) Use
31058 ds_form_mem_operand in ld/lwa patterns.
31059 * config/rs6000/fusion.md: Regenerate file.
31061 2021-03-08 Martin Sebor <msebor@redhat.com>
31063 PR middle-end/98266
31064 * gimple-array-bounds.cc (inbounds_vbase_memaccess_p): New function.
31065 (array_bounds_checker::check_array_bounds): Call it.
31067 2021-03-08 Martin Sebor <msebor@redhat.com>
31069 PR middle-end/97631
31070 * tree-ssa-strlen.c (maybe_warn_overflow): Test rawmem.
31071 (handle_builtin_stxncpy_strncat): Rename locals. Determine
31072 destination size from allocation calls. Issue a more appropriate
31074 (handle_builtin_memcpy): Pass true as rawmem to maybe_warn_overflow.
31075 (handle_builtin_memset): Same.
31077 2021-03-08 Peter Bergner <bergner@linux.ibm.com>
31080 * config/rs6000/rs6000.c (rs6000_emit_le_vsx_permute): Add an assert
31081 to ensure we do not have an Altivec style address.
31082 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): Disable if passed
31083 an Altivec style address.
31084 (*vsx_le_perm_store_<mode>): Likewise.
31085 (splitters after *vsx_le_perm_store_<mode>): Likewise.
31086 (vsx_load_<mode>): Disable special expander if passed an Altivec
31088 (vsx_store_<mode>): Likewise.
31090 2021-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
31093 * config/aarch64/predicates.md (aarch64_simd_shift_imm_vec_qi): Define.
31094 (aarch64_simd_shift_imm_vec_hi): Likewise.
31095 (aarch64_simd_shift_imm_vec_si): Likewise.
31096 (aarch64_simd_shift_imm_vec_di): Likewise.
31097 * config/aarch64/aarch64-simd.md (aarch64_shrn<mode>_insn_le): Use
31098 predicate from above.
31099 (aarch64_shrn<mode>_insn_be): Likewise.
31100 (aarch64_rshrn<mode>_insn_le): Likewise.
31101 (aarch64_rshrn<mode>_insn_be): Likewise.
31102 (aarch64_shrn2<mode>_insn_le): Likewise.
31103 (aarch64_shrn2<mode>_insn_be): Likewise.
31104 (aarch64_rshrn2<mode>_insn_le): Likewise.
31105 (aarch64_rshrn2<mode>_insn_be): Likewise.
31107 2021-03-08 Vladimir N. Makarov <vmakarov@redhat.com>
31110 * lra-constraints.c (skip_contraint_modifiers): New function.
31111 (process_address_1): Use it before lookup_constraint call.
31113 2021-03-08 Martin Liska <mliska@suse.cz>
31116 * config/i386/i386-options.c (ix86_option_override_internal):
31117 Enable UINTR and HRESET for -march that supports it.
31119 2021-03-08 Ilya Leoshkevich <iii@linux.ibm.com>
31121 * config/s390/s390.c (f_constraint_p): New function.
31122 (s390_md_asm_adjust): Implement TARGET_MD_ASM_ADJUST.
31123 (TARGET_MD_ASM_ADJUST): Likewise.
31125 2021-03-08 Tobias Burnus <tobias@codesourcery.com>
31128 * tree-nested.c (convert_local_reference_stmt): Avoid calling
31129 lookup_field_for_decl for Fortran module (= namespace context).
31131 2021-03-08 Andreas Krebbel <krebbel@linux.ibm.com>
31133 * config/s390/s390.c (s390_expand_vec_compare): Implement <0
31134 comparison with arithmetic right shift.
31135 (s390_expand_vcond): No need for a force_reg anymore.
31136 s390_vec_compare will do it.
31137 * config/s390/vector.md ("vec_cmp<mode><tointvec>"): Accept also
31138 immediate operands.
31140 2021-03-07 Jakub Jelinek <jakub@redhat.com>
31143 * config/i386/constraints.md (Yw): Use SSE_REGS if TARGET_SSE
31144 but TARGET_AVX512BW or TARGET_AVX512VL is not set. Adjust description
31146 * config/i386/sse.md (v_Yw): New define_mode_attr.
31147 (*<insn><mode>3, *mul<mode>3<mask_name>, *avx2_<code><mode>3,
31148 *sse4_1_<code><mode>3<mask_name>): Use <v_Yw> instead of v
31150 * config/i386/mmx.md (mmx_pshufw_1, *vec_dupv4hi): Use Yw instead of
31151 xYw in constraints.
31153 2021-03-06 Julian Brown <julian@codesourcery.com>
31155 * tree-pretty-print.c (dump_generic_node): Emit non-generic
31156 address space info for aggregates.
31158 2021-03-06 Hans-Peter Nilsson <hp@axis.com>
31160 * config/cris/cris.h (MAX_FIXED_MODE_SIZE): Don't define.
31162 2021-03-05 Jakub Jelinek <jakub@redhat.com>
31164 PR middle-end/99322
31165 * tree-cfg.c (bb_to_omp_idx): New variable.
31166 (execute_build_cfg): Release the bb_to_omp_idx vector after
31167 cleanup_tree_cfg returns.
31168 (handle_abnormal_edges): Remove bb_to_omp_idx argument, adjust
31169 for bb_to_omp_idx being a vec<int> instead of pointer to array
31171 (make_edges): Remove bb_to_omp_idx local variable, don't pass
31172 it to handle_abnormal_edges, adjust for bb_to_omp_idx being a
31173 vec<int> instead of pointer to array of ints and don't free/release
31175 (remove_bb): When removing a bb and placing forced label somewhere
31176 else, ensure it is put into the same OpenMP region during cfg
31177 pass if possible or to entry successor as fallback. Unregister
31178 bb from bb_to_omp_idx.
31180 2021-03-05 Vladimir N. Makarov <vmakarov@redhat.com>
31183 * lra-constraints.c (process_address_1): Skip decomposing address
31184 for asm insn operand with unknown constraint.
31186 2021-03-05 Martin Jambor <mjambor@suse.cz>
31189 * cgraph.c (cgraph_edge::set_call_stmt): Do not update all
31190 corresponding speculative edges if we are about to resolve
31191 sepculation. Make edge direct (and so resolve speculations) before
31192 removing it from call_site_hash.
31193 (cgraph_edge::make_direct): Relax the initial assert to allow calling
31194 the function on speculative direct edges.
31196 2021-03-05 Eric Botcazou <ebotcazou@adacore.com>
31198 PR rtl-optimization/99376
31199 * rtlanal.c (nonzero_bits1) <arithmetic operators>: If the number
31200 of low-order zero bits is too large, set the result to 0 directly.
31202 2021-03-04 Jakub Jelinek <jakub@redhat.com>
31204 PR middle-end/93235
31205 * expmed.c (store_bit_field_using_insv): Return false of xop0 is a
31206 SUBREG and a SUBREG to op_mode can't be created.
31208 2021-03-04 Alex Coplan <alex.coplan@arm.com>
31211 * config/aarch64/aarch64-sve-builtins.cc
31212 (function_resolver::require_vector_type): Handle error_mark_node.
31214 2021-03-04 Ilya Leoshkevich <iii@linux.ibm.com>
31216 * cfgexpand.c (expand_asm_loc): Pass new parameter.
31217 (expand_asm_stmt): Likewise.
31218 * config/arm/aarch-common-protos.h (arm_md_asm_adjust): Add new
31220 * config/arm/aarch-common.c (arm_md_asm_adjust): Likewise.
31221 * config/arm/arm.c (thumb1_md_asm_adjust): Likewise.
31222 * config/cris/cris.c (cris_md_asm_adjust): Likewise.
31223 * config/i386/i386.c (ix86_md_asm_adjust): Likewise.
31224 * config/mn10300/mn10300.c (mn10300_md_asm_adjust): Likewise.
31225 * config/nds32/nds32.c (nds32_md_asm_adjust): Likewise.
31226 * config/pdp11/pdp11.c (pdp11_md_asm_adjust): Likewise.
31227 * config/rs6000/rs6000.c (rs6000_md_asm_adjust): Likewise.
31228 * config/vax/vax.c (vax_md_asm_adjust): Likewise.
31229 * config/visium/visium.c (visium_md_asm_adjust): Likewise.
31230 * doc/tm.texi (md_asm_adjust): Likewise.
31231 * target.def (md_asm_adjust): Likewise.
31233 2021-03-04 Richard Biener <rguenther@suse.de>
31235 PR middle-end/97855
31236 * tree-pretty-print.c: Poison pp_printf.
31237 (dump_decl_name): Avoid use of pp_printf.
31238 (dump_block_node): Likewise.
31239 (dump_generic_node): Likewise.
31241 2021-03-04 Martin Sebor <msebor@redhat.com>
31243 PR middle-end/96963
31244 PR middle-end/94655
31245 * builtins.c (handle_array_ref): New helper.
31246 (handle_mem_ref): New helper.
31247 (compute_objsize_r): Factor out ARRAY_REF and MEM_REF handling
31248 into new helper functions. Correct a workaround for vectorized
31251 2021-03-03 Pat Haugen <pthaugen@linux.ibm.com>
31253 * config/rs6000/dfp.md (extendddtd2, trunctddd2, *cmp<mode>_internal1,
31254 floatditd2, ftrunc<mode>2, fix<mode>di2, dfp_ddedpd_<mode>,
31255 dfp_denbcd_<mode>, dfp_dxex_<mode>, dfp_diex_<mode>,
31256 *dfp_sgnfcnc_<mode>, dfp_dscli_<mode>, dfp_dscri_<mode>): Update size
31257 attribute for Power10.
31258 * config/rs6000/mma.md (*movoo): Likewise.
31259 * config/rs6000/rs6000.md (define_attr "size"): Add 256.
31260 (define_mode_attr bits): Add DD/TD modes.
31261 * config/rs6000/sync.md (load_quadpti, store_quadpti, load_lockedpti,
31262 store_conditionalpti): Update size attribute for Power10.
31264 2021-03-03 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
31267 * config/sparc/t-sparc (tree-ssanames.o-warn): Don't error for
31268 -Wuninitialized, -Wmaybe-uninitialized.
31269 (wide-int.o-warn): Likewise.
31271 2021-03-03 Richard Earnshaw <rearnsha@arm.com>
31273 * common/config/arm/arm-common.c: Include configargs.h.
31274 (arm_config_default): New function.
31275 (arm_target_mode): Renamed from arm_target_thumb_only. Handle
31276 processors that do not support Thumb. Take into account the
31277 --with-mode configuration setting for selecting the default.
31278 * config/arm/arm.h (OPTION_DEFAULT_SPECS): Remove entry for 'mode'.
31279 (TARGET_MODE_SPEC_FUNCTIONS): Update for function name change.
31281 2021-03-03 Martin Liska <mliska@suse.cz>
31283 PR gcov-profile/97461
31284 * gcov-io.h (GCOV_PREALLOCATED_KVP): Remove.
31286 2021-03-03 Eric Botcazou <ebotcazou@adacore.com>
31289 * config/i386/i386.c (ix86_compute_frame_layout): For a SEH target,
31290 point back the hard frame pointer to its default location when the
31291 frame is larger than SEH_MAX_FRAME_SIZE.
31293 2021-03-03 Jakub Jelinek <jakub@redhat.com>
31296 * config/i386/predicates.md (logic_operator): New define_predicate.
31297 * config/i386/i386.md (mov + mem using comm arith peephole2):
31298 Punt if operands[1] is EXT_REX_SSE_REGNO_P, AVX512BW is not enabled
31299 and the inner mode is [QH]Imode.
31301 2021-03-03 Jakub Jelinek <jakub@redhat.com>
31304 * dwarf2out.c (dw_loc_list_struct): Add end_entry member.
31305 (new_loc_list): Clear end_entry.
31306 (output_loc_list): Only use DW_LLE_startx_length for -gsplit-dwarf
31307 if HAVE_AS_LEB128, otherwise use DW_LLE_startx_endx. Fix comment
31309 (index_location_lists): For dwarf_version >= 5 without HAVE_AS_LEB128,
31310 initialize also end_entry.
31312 2021-03-03 Jakub Jelinek <jakub@redhat.com>
31315 * cfgrtl.c (fixup_partitions): When changing some bbs from hot to cold
31316 partitions, if in non-layout mode after reorder_blocks also move
31317 affected blocks to ensure a single partition transition.
31319 2021-03-03 Jason Merrill <jason@redhat.com>
31322 * cgraphunit.c (process_function_and_variable_attributes): Don't
31323 warn about flatten on an alias if the target also has it.
31324 * cgraph.h (symtab_node::get_alias_target_tree): New.
31326 2021-03-02 David Edelsohn <dje.gcc@gmail.com>
31328 * config/rs6000/rs6000.md (tls_get_tpointer_internal): Prepend
31329 period to symbol name.
31330 (tls_get_addr_internal<mode>): Same.
31332 2021-03-02 David Malcolm <dmalcolm@redhat.com>
31335 * diagnostic-show-locus.c
31336 (selftest::test_one_liner_many_fixits_2): Fix accidental usage of
31339 2021-03-02 Martin Sebor <msebor@redhat.com>
31341 PR middle-end/99276
31342 * builtins.c (warn_for_access): Remove stray warning text.
31344 2021-03-02 Martin Sebor <msebor@redhat.com>
31346 PR middle-end/99295
31347 * doc/extend.texi (attribute malloc): Reword and clarify nonaliasing
31350 2021-03-02 Jakub Jelinek <jakub@redhat.com>
31353 * dwarf2out.c (output_macinfo_op): Use DW_MACRO_*_str* even with
31354 -gdwarf-5 -gstrict-dwarf. For -gsplit-dwarf -gdwarf-5 use
31355 DW_MACRO_*_strx instead of DW_MACRO_*_strp. Handle
31356 DW_MACRO_define_strx and DW_MACRO_undef_strx.
31357 (save_macinfo_strings): Use DW_MACRO_*_str* even with
31358 -gdwarf-5 -gstrict-dwarf. Handle DW_MACRO_define_strx and
31359 DW_MACRO_undef_strx.
31361 2021-03-02 Andreas Krebbel <krebbel@linux.ibm.com>
31363 * config/s390/s390-builtin-types.def (BT_FN_V4SF_V8HI_UINT): New
31365 (BT_FN_V8HI_V8HI_UINT): Likewise.
31366 (BT_FN_V8HI_V4SF_V4SF_UINT): Likewise.
31367 * config/s390/s390-builtins.def (B_NNPA): New macro definition.
31368 (s390_vclfnhs, s390_vclfnls, s390_vcrnfs, s390_vcfn, s390_vcnf):
31369 New builtin definitions.
31370 * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal): Bump
31371 vector extension version.
31372 * config/s390/s390.c (s390_expand_builtin): Check if builtins are
31373 available with current -march level.
31374 * config/s390/s390.md (UNSPEC_NNPA_VCLFNHS_V8HI)
31375 (UNSPEC_NNPA_VCLFNLS_V8HI, UNSPEC_NNPA_VCRNFS_V8HI)
31376 (UNSPEC_NNPA_VCFN_V8HI, UNSPEC_NNPA_VCNF_V8HI): New constants.
31377 * config/s390/vecintrin.h (vec_extend_to_fp32_hi): New macro.
31378 (vec_extend_to_fp32_lo): Likewise.
31379 (vec_round_from_fp32): Likewise.
31380 (vec_convert_to_fp16): Likewise.
31381 (vec_convert_from_fp16): Likewise.
31382 * config/s390/vx-builtins.md (vclfnhs_v8hi): New insn pattern.
31383 (vclfnls_v8hi): Likewise.
31384 (vcrnfs_v8hi): Likewise.
31385 (vcfn_v8hi): Likewise.
31386 (vcnf_v8hi): Likewise.
31388 2021-03-02 Andreas Krebbel <krebbel@linux.ibm.com>
31390 * common/config/s390/s390-common.c (processor_flags_table): New entry.
31391 * config.gcc: Enable arch14 for --with-arch and --with-tune.
31392 * config/s390/driver-native.c (s390_host_detect_local_cpu): Pick
31393 arch14 for unknown CPU models.
31394 * config/s390/s390-opts.h (enum processor_type): Add PROCESSOR_ARCH14.
31395 * config/s390/s390.c (s390_issue_rate): Add case for PROCESSOR_ARCH14.
31396 (s390_get_sched_attrmask): Likewise.
31397 (s390_get_unit_mask): Likewise.
31398 * config/s390/s390.h (enum processor_flags): Add PF_NNPA and PF_ARCH14.
31399 (TARGET_CPU_ARCH14, TARGET_CPU_ARCH14_P, TARGET_CPU_NNPA)
31400 (TARGET_CPU_NNPA_P, TARGET_ARCH14, TARGET_ARCH14_P, TARGET_NNPA)
31401 (TARGET_NNPA_P): New macro definitions.
31402 * config/s390/s390.md ("cpu_facility", "enabled"): Add arch14 and nnpa.
31403 * config/s390/s390.opt: Add PROCESSOR_ARCH14.
31405 2021-03-02 Jakub Jelinek <jakub@redhat.com>
31407 PR middle-end/95757
31408 * tree-vrp.c (register_edge_assert_for): Remove superfluous ()s around
31409 condition. Call register_edge_assert_for_1 for == 0, != 0, == 1 and
31410 != 1 comparisons if name is lhs of a comparison.
31412 2021-03-01 Iain Sandoe <iain@sandoe.co.uk>
31416 * config/darwin-protos.h (darwin_should_restore_cfa_state): New.
31417 * config/darwin.c (darwin_should_restore_cfa_state): New.
31418 * config/darwin.h (TARGET_ASM_SHOULD_RESTORE_CFA_STATE): New.
31419 * doc/tm.texi: Regenerated.
31420 * doc/tm.texi.in: Document TARGET_ASM_SHOULD_RESTORE_CFA_STATE.
31421 * dwarf2cfi.c (connect_traces): If the target requests, restore
31422 the CFA expression after a DW_CFA_restore.
31423 * target.def (TARGET_ASM_SHOULD_RESTORE_CFA_STATE): New hook.
31425 2021-03-01 Martin Liska <mliska@suse.cz>
31428 * optc-save-gen.awk: Add 4 more exceptions.
31430 2021-03-01 Nathan Sidwell <nathan@acm.org>
31433 * tree.h (TYPE_ALIGN_RAW): New accessor.
31434 (TYPE_ALIGN): Use it.
31436 2021-03-01 Jan Hubicka <jh@suse.cz>
31439 * ipa-fnsummary.c (compute_fn_summary): Fix sanity check.
31441 2021-03-01 Eric Botcazou <ebotcazou@adacore.com>
31444 * config/i386/i386.c (ix86_compute_frame_layout): For a SEH target,
31445 point the hard frame pointer to the SSE register save area instead
31446 of the general register save area. Perform only minimal adjustment
31447 for small frames if it is initially not correctly aligned.
31448 (ix86_expand_prologue): Remove early saves for a SEH target.
31449 * config/i386/winnt.c (struct seh_frame_state): Document constraint.
31451 2021-02-28 Jakub Jelinek <jakub@redhat.com>
31454 * ipa.c (symbol_table::remove_unreachable_nodes): Fix a comment
31455 typo - referneced -> referenced.
31456 * tree.c (component_ref_size): Fix comment typo -
31457 refernce -> reference.
31458 * tree-ssa-alias.c (access_path_may_continue_p): Fix comment typo -
31459 traling -> trailing.
31460 (aliasing_component_refs_p): Fix comment typos -
31461 refernce -> reference and refernece -> reference and
31462 traling -> trailing.
31463 (nonoverlapping_refs_since_match_p): Fix comment typo -
31464 referneces -> references.
31465 * doc/invoke.texi (--param modref-max-bases): Fix a typo -
31466 referneces -> references.
31468 2021-02-27 Iain Sandoe <iain@sandoe.co.uk>
31470 * config/host-darwin.c (darwin_gt_pch_use_address): Modify
31471 diagnostic message to avoid use of a contraction and format
31474 2021-02-27 Jakub Jelinek <jakub@redhat.com>
31477 * gcse.c (gcse_or_cprop_is_too_expensive): Use %wu instead of
31478 HOST_WIDE_INT_PRINT_UNSIGNED in warning format string.
31479 * ipa-devirt.c (ipa_odr_read_section): Use %wd instead of
31480 HOST_WIDE_INT_PRINT_DEC in inform format string. Fix comment
31483 2021-02-26 Richard Biener <rguenther@suse.de>
31485 PR middle-end/99281
31486 * expr.c (store_field): For calls with return-slot optimization
31487 and addressable return type expand the store directly.
31489 2021-02-26 Richard Biener <rguenther@suse.de>
31492 * builtins.c (warn_string_no_nul): Fix diagnostic formatting.
31494 2021-02-26 Peter Bergner <bergner@linux.ibm.com>
31497 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Replace assert
31500 2021-02-26 Aaron Sawdey <acsawdey@linux.ibm.com>
31502 * config.gcc: Add rs6000-pcrel-opt.o.
31503 * config/rs6000/rs6000-pcrel-opt.c: New file.
31504 * config/rs6000/pcrel-opt.md: New file.
31505 * config/rs6000/predicates.md: Add d_form_memory predicate.
31506 * config/rs6000/rs6000-cpus.def: Add OPTION_MASK_PCREL_OPT.
31507 * config/rs6000/rs6000-passes.def: Add pass_pcrel_opt.
31508 * config/rs6000/rs6000-protos.h: Add reg_to_non_prefixed(),
31509 pcrel_opt_valid_mem_p(), output_pcrel_opt_reloc(),
31510 and make_pass_pcrel_opt().
31511 * config/rs6000/rs6000.c (reg_to_non_prefixed): Make global.
31512 (rs6000_option_override_internal): Add pcrel-opt.
31513 (rs6000_delegitimize_address): Support pcrel-opt.
31514 (rs6000_opt_masks): Add pcrel-opt.
31515 (pcrel_opt_valid_mem_p): New function.
31516 (reg_to_non_prefixed): Make global.
31517 (rs6000_asm_output_opcode): Reset prepend_p_to_next_insn.
31518 (output_pcrel_opt_reloc): New function.
31519 * config/rs6000/rs6000.md (loads_extern_addr): New attr.
31520 (pcrel_extern_addr): Set loads_extern_addr.
31521 Add include for pcrel-opt.md.
31522 * config/rs6000/rs6000.opt: Add -mpcrel-opt.
31523 * config/rs6000/t-rs6000: Add rules for pcrel-opt.c and
31526 2021-02-26 YunQiang Su <yunqiang.su@cipunited.com>
31529 * config/mips/mips.c (mips_expand_ext_as_unaligned_load):
31530 If TARGET_64BIT and dest is SUBREG, we check the width, if it
31531 equal to SImode, we use SImode operation, just like what we are
31534 2021-02-26 Marek Polacek <polacek@redhat.com>
31536 * builtins.c (warn_for_access): Fix typos.
31538 2021-02-25 Iain Sandoe <iain@sandoe.co.uk>
31540 * config/aarch64/aarch64.md (<optab>_rol<mode>3): Add a '#'
31541 mark in front of the immediate quantity.
31542 (<optab>_rolsi3_uxtw): Likewise.
31544 2021-02-25 Richard Earnshaw <rearnsha@arm.com>
31547 * config/arm/thumb2.md (nonsecure_call_reg_thumb2_fpcxt): New pattern.
31548 (nonsecure_call_value_reg_thumb2_fpcxt): Likewise.
31549 (nonsecure_call_reg_thumb2): Restrict to using r4 for the callee
31550 address and disable when the FPCXT is not available.
31551 (nonsecure_call_value_reg_thumb2): Likewise.
31553 2021-02-25 Nathan Sidwell <nathan@acm.org>
31556 * doc/invoke.texi (flang-info-module-cmi): Renamed option.
31558 2021-02-25 Tamar Christina <tamar.christina@arm.com>
31560 * tree-vect-slp.c (optimize_load_redistribution_1): Abort on NULL nodes.
31562 2021-02-25 Richard Biener <rguenther@suse.de>
31564 PR tree-optimization/99253
31565 * tree-vect-loop.c (check_reduction_path): First compute
31566 code, then verify out-of-loop uses.
31568 2021-02-25 Jakub Jelinek <jakub@redhat.com>
31571 * match.pd ((T)(A) + CST -> (T)(A + CST)): Add :s to convert.
31573 2021-02-25 Jakub Jelinek <jakub@redhat.com>
31575 PR tree-optimization/80635
31576 * tree-vrp.c (vrp_simplify_cond_using_ranges): Also handle
31577 VIEW_CONVERT_EXPR if modes are the same, innerop is integral and
31578 has mode precision.
31580 2021-02-25 Richard Biener <rguenther@suse.de>
31582 * tree-vect-slp.c (optimize_load_redistribution_1): Delay
31583 load_map population.
31584 (vect_match_slp_patterns_2): Revert part of last change.
31585 (vect_analyze_slp): Do not interleave optimize_load_redistribution
31586 with pattern detection but do it afterwards. Dump the
31587 whole SLP graph after pattern recognition and load
31588 redistribution optimization finished.
31590 2021-02-24 Jakub Jelinek <jakub@redhat.com>
31593 * omp-low.c (struct omp_context): Add teams_nested_p and
31594 nonteams_nested_p members.
31595 (scan_omp_target): Diagnose teams nested inside of target with other
31596 directives strictly nested inside of the same target.
31597 (check_omp_nesting_restrictions): Set ctx->teams_nested_p or
31598 ctx->nonteams_nested_p as needed.
31600 2021-02-24 Vladimir N. Makarov <vmakarov@redhat.com>
31602 PR inline-asm/99123
31603 * lra-constraints.c (uses_hard_regs_p): Don't use decompose_mem_address.
31605 2021-02-24 Hans-Peter Nilsson <hp@axis.com>
31607 * config/cris/cris.c (cris_expand_prologue): Set
31608 current_function_static_stack_size, if flag_stack_usage_info.
31610 2021-02-24 Pat Haugen <pthaugen@linux.ibm.com>
31612 * config/rs6000/rs6000.c (next_insn_prefixed_p): Rename.
31613 (rs6000_final_prescan_insn): Adjust.
31614 (rs6000_asm_output_opcode): Likewise.
31616 2021-02-24 Martin Sebor <msebor@redhat.com>
31618 PR middle-end/97172
31619 * attribs.c (attr_access::free_lang_data): Clear attribute arg spec
31620 from function arguments.
31622 2021-02-24 Tamar Christina <tamar.christina@arm.com>
31624 PR tree-optimization/99220
31625 * tree-vect-slp.c (optimize_load_redistribution_1): Remove
31626 node from cache when it's about to be deleted.
31628 2021-02-24 Jakub Jelinek <jakub@redhat.com>
31630 PR tree-optimization/99225
31631 * fold-const.c (fold_binary_loc) <case NE_EXPR>: In (x & (1 << y)) != 0
31632 to ((x >> y) & 1) != 0 simplifications use build_one_cst instead of
31633 build_int_cst (..., 1). Formatting fixes.
31635 2021-02-24 Tamar Christina <tamar.christina@arm.com>
31637 PR tree-optimization/99149
31638 * tree-vect-slp-patterns.c (vect_detect_pair_op): Don't recreate the
31640 (vect_slp_reset_pattern): Remove.
31641 (complex_fma_pattern::matches): Remove call to vect_slp_reset_pattern.
31642 (complex_mul_pattern::build, complex_fma_pattern::build,
31643 complex_fms_pattern::build): Fix ref counts.
31644 * tree-vect-slp.c (vect_free_slp_tree): Undo SLP only pattern relevancy
31645 when node is being deleted.
31646 (vect_match_slp_patterns_2): Correct result of cache hit on patterns.
31647 (vect_schedule_slp): Invalidate SLP_TREE_REPRESENTATIVE of removed
31649 * tree-vectorizer.c (vec_info::new_stmt_vec_info): Initialize value.
31651 2021-02-24 Matthias Klose <doko@ubuntu.com>
31654 2020-12-07 Matthias Klose <doko@ubuntu.com>
31656 * genextract.c (print_header): Undefine ENABLE_RTL_CHECKING
31657 and ENABLE_RTL_FLAG_CHECKING.
31659 2021-02-24 Richard Biener <rguenther@suse.de>
31662 * builtins.c (fold_builtin_next_arg): Avoid NULL arg.
31664 2021-02-23 Peter Bergner <bergner@linux.ibm.com>
31666 * config/rs6000/mma.md (mma_assemble_pair): Rename from this...
31667 (vsx_assemble_pair): ...to this.
31668 (*mma_assemble_pair): Rename from this...
31669 (*vsx_assemble_pair): ...to this.
31670 (mma_disassemble_pair): Rename from this...
31671 (vsx_disassemble_pair): ...to this.
31672 (*mma_disassemble_pair): Rename from this...
31673 (*vsx_disassemble_pair): ...to this.
31674 * config/rs6000/rs6000-builtin.def (BU_MMA_V2, BU_MMA_V3,
31675 BU_COMPAT): New macros.
31676 (mma_assemble_pair): Rename from this...
31677 (vsx_assemble_pair): ...to this.
31678 (mma_disassemble_pair): Rename from this...
31679 (vsx_disassemble_pair): ...to this.
31680 (mma_assemble_pair): New compatibility built-in.
31681 (mma_disassemble_pair): Likewise.
31682 * config/rs6000/rs6000-call.c (struct builtin_compatibility): New.
31683 (RS6000_BUILTIN_COMPAT): Define.
31684 (bdesc_compat): New.
31685 (mma_expand_builtin): Use VSX_BUILTIN_DISASSEMBLE_PAIR_INTERNAL.
31686 (rs6000_gimple_fold_mma_builtin): Use MMA_BUILTIN_DISASSEMBLE_PAIR
31687 and VSX_BUILTIN_ASSEMBLE_PAIR.
31688 (rs6000_init_builtins): Register compatibility built-ins.
31689 (mma_init_builtins): Use VSX_BUILTIN_ASSEMBLE_PAIR,
31690 VSX_BUILTIN_ASSEMBLE_PAIR_INTERNAL, VSX_BUILTIN_DISASSEMBLE_PAIR and
31691 VSX_BUILTIN_DISASSEMBLE_PAIR_INTERNAL.
31692 * doc/extend.texi (__builtin_mma_assemble_pair): Rename from this...
31693 (__builtin_vsx_assemble_pair): ...to this.
31694 (__builtin_mma_disassemble_pair): Rename from this...
31695 (__builtin_vsx_disassemble_pair): ...to this.
31697 2021-02-23 Martin Liska <mliska@suse.cz>
31700 * ipa-icf.c (sem_variable::merge): Do not merge 2 variables
31701 with different alignment. That leads to an invalid red zone
31702 size allocated in runtime.
31704 2021-02-23 Jakub Jelinek <jakub@redhat.com>
31706 PR tree-optimization/99204
31707 * fold-const.c (fold_read_from_constant_string): Check that
31708 tree_fits_uhwi_p (index) rather than just that index is INTEGER_CST.
31710 2021-02-23 Segher Boessenkool <segher@kernel.crashing.org>
31711 Kewen Lin <linkw@gcc.gnu.org>
31713 * config/rs6000/rs6000.md (*rotl<mode>3_insert_3): Renamed to...
31714 (rotl<mode>3_insert_3): ...this.
31715 (plus_ior_xor): New code_iterator.
31716 (define_split for GPR rl*imi): New splitter.
31717 * config/rs6000/vsx.md (vsx_init_v4si): Use gen_rotldi3_insert_3
31718 for integer merging.
31720 2021-02-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
31722 * config/aarch64/aarch64-tuning-flags.def (cse_sve_vl_constants):
31724 * config/aarch64/aarch64.md (add<mode>3): Force CONST_POLY_INT immediates
31725 into a register when the above is enabled.
31726 * config/aarch64/aarch64.c (neoversev1_tunings):
31727 AARCH64_EXTRA_TUNE_CSE_SVE_VL_CONSTANTS.
31728 (aarch64_rtx_costs): Use AARCH64_EXTRA_TUNE_CSE_SVE_VL_CONSTANTS.
31730 2021-02-22 Hans-Peter Nilsson <hp@axis.com>
31732 * config/cris/cris.c (cris_print_operand) <'T'>: Change
31733 valid operand from is now an addi mult-value to shift-value.
31734 * config/cris/cris.md (*addi): Change expression of scaled
31735 operand from mult to ashift.
31736 * config/cris/cris.md (*addi_reload): New insn_and_split.
31738 2021-02-22 John David Anglin <danglin@gcc.gnu.org>
31741 * config/pa/pa.c (TARGET_ASM_CAN_OUTPUT_MI_THUNK): Define as
31742 hook_bool_const_tree_hwi_hwi_const_tree_true.
31743 (pa_asm_output_mi_thunk): Add support for nonzero vcall_offset.
31745 2021-02-22 Andre Vieira <andre.simoesdiasvieira@arm.com>
31747 PR rtl-optimization/98791
31748 * ira-conflicts.c (process_regs_for_copy): Don't create allocno copies
31749 for unordered modes.
31751 2021-02-22 Martin Liska <mliska@suse.cz>
31753 * tree-inline.c (inline_forbidden_p): Set
31754 inline_forbidden_reason.
31756 2021-02-22 Richard Biener <rguenther@suse.de>
31758 * tree-vect-slp.c (vect_bb_vectorization_profitable_p): Dump
31761 2021-02-22 Richard Biener <rguenther@suse.de>
31763 PR tree-optimization/99165
31764 * gimple-ssa-store-merging.c (pass_store_merging::process_store):
31765 Accumulate changed to ret.
31767 2021-02-21 Uros Bizjak <ubizjak@gmail.com>
31770 2020-12-09 Uroš Bizjak <ubizjak@gmail.com>
31772 * config/i386/i386.h (REG_ALLOC_ORDER): Remove
31774 2021-02-20 Ilya Leoshkevich <iii@linux.ibm.com>
31777 * config/s390/vector.md (trunctf<DFP_ALL:mode>2_vr): New
31779 (trunctf<DFP_ALL:mode>2): Likewise.
31780 (trunctdtf2_vr): Likewise.
31781 (trunctdtf2): Likewise.
31782 (extend<DFP_ALL:mode>tf2_vr): Likewise.
31783 (extend<DFP_ALL:mode>tf2): Likewise.
31784 (extendtftd2_vr): Likewise.
31785 (extendtftd2): Likewise.
31787 2021-02-20 Ilya Leoshkevich <iii@linux.ibm.com>
31789 * config/s390/vector.md (*fprx2_to_tf): Rename to fprx2_to_tf,
31790 add memory alternative.
31791 (tf_to_fprx2): New pattern.
31793 2021-02-19 Martin Sebor <msebor@redhat.com>
31796 * attribs.c (init_attr_rdwr_indices): Guard vblist use.
31797 (attr_access::free_lang_data): Remove a spurious test.
31799 2021-02-19 Nathan Sidwell <nathan@acm.org>
31801 * doc/invoke.texi (flang-info-module-read): Document.
31803 2021-02-19 Martin Liska <mliska@suse.cz>
31805 PR translation/99167
31806 * params.opt: Fix typo.
31808 2021-02-19 Richard Biener <rguenther@suse.de>
31810 PR middle-end/99122
31811 * tree-inline.c (inline_forbidden_p): Do not inline functions
31812 with VLA arguments or return value.
31814 2021-02-19 Jakub Jelinek <jakub@redhat.com>
31817 * config/arm/arm.md (*stack_protect_combined_set_insn,
31818 *stack_protect_combined_test_insn): If force_const_mem result
31819 is not valid general operand, force its address into the destination
31822 2021-02-19 Jakub Jelinek <jakub@redhat.com>
31825 * tree-cfg.c (gimple_merge_blocks): If bb a starts with eh landing
31826 pad or non-local label, put FORCED_LABELs from bb b after that label
31827 rather than before it.
31829 2021-02-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
31832 * config/aarch64/aarch64-sve.md (<ASHIFT:optab><mode>3): Use
31833 expand_vector_broadcast' to emit the vec_duplicate operand.
31835 2021-02-18 Vladimir N. Makarov <vmakarov@redhat.com>
31837 PR rtl-optimization/96264
31838 * lra-remat.c (reg_overlap_for_remat_p): Check also output insn
31841 2021-02-18 H.J. Lu <hjl.tools@gmail.com>
31844 * varasm.c (get_section): Replace SUPPORTS_SHF_GNU_RETAIN with
31845 looking up the retain attribute.
31846 (resolve_unique_section): Likewise.
31847 (get_variable_section): Likewise.
31848 (switch_to_section): Likewise. Warn when a symbol without the
31849 retain attribute and a symbol with the retain attribute are
31850 placed in the section with the same name, instead of the used
31852 * doc/extend.texi: Document the "retain" attribute.
31854 2021-02-18 Nathan Sidwell <nathan@acm.org>
31857 * doc/invoke.texi (flang-info-include-translate): Document header
31860 2021-02-18 Richard Biener <rguenther@suse.de>
31862 PR middle-end/99122
31863 * ipa-fnsummary.c (analyze_function_body): Set
31864 CIF_FUNCTION_NOT_INLINABLE for VLA parameter calls.
31865 * tree-inline.c (insert_init_debug_bind): Pass NULL for
31866 error_mark_node values.
31867 (force_value_to_type): Do not build V_C_Es for WITH_SIZE_EXPR
31869 (setup_one_parameter): Delay force_value_to_type until when
31872 2021-02-18 Hans-Peter Nilsson <hp@axis.com>
31874 PR tree-optimization/99142
31875 * match.pd (clz cmp 0): Gate replacement on single_use of clz result.
31877 2021-02-18 Jakub Jelinek <jakub@redhat.com>
31879 * wide-int-bitmask.h (wide_int_bitmask::wide_int_bitmask (),
31880 wide_int_bitmask::wide_int_bitmask (uint64_t),
31881 wide_int_bitmask::wide_int_bitmask (uint64_t, uint64_t),
31882 wide_int_bitmask::operator ~ () const,
31883 wide_int_bitmask::operator | (wide_int_bitmask) const,
31884 wide_int_bitmask::operator & (wide_int_bitmask) const): Use constexpr
31886 * config/i386/i386.h (PTA_3DNOW, PTA_3DNOW_A, PTA_64BIT, PTA_ABM,
31887 PTA_AES, PTA_AVX, PTA_BMI, PTA_CX16, PTA_F16C, PTA_FMA, PTA_FMA4,
31888 PTA_FSGSBASE, PTA_LWP, PTA_LZCNT, PTA_MMX, PTA_MOVBE, PTA_NO_SAHF,
31889 PTA_PCLMUL, PTA_POPCNT, PTA_PREFETCH_SSE, PTA_RDRND, PTA_SSE, PTA_SSE2,
31890 PTA_SSE3, PTA_SSE4_1, PTA_SSE4_2, PTA_SSE4A, PTA_SSSE3, PTA_TBM,
31891 PTA_XOP, PTA_AVX2, PTA_BMI2, PTA_RTM, PTA_HLE, PTA_PRFCHW, PTA_RDSEED,
31892 PTA_ADX, PTA_FXSR, PTA_XSAVE, PTA_XSAVEOPT, PTA_AVX512F, PTA_AVX512ER,
31893 PTA_AVX512PF, PTA_AVX512CD, PTA_NO_TUNE, PTA_SHA, PTA_PREFETCHWT1,
31894 PTA_CLFLUSHOPT, PTA_XSAVEC, PTA_XSAVES, PTA_AVX512DQ, PTA_AVX512BW,
31895 PTA_AVX512VL, PTA_AVX512IFMA, PTA_AVX512VBMI, PTA_CLWB, PTA_MWAITX,
31896 PTA_CLZERO, PTA_NO_80387, PTA_PKU, PTA_AVX5124VNNIW, PTA_AVX5124FMAPS,
31897 PTA_AVX512VPOPCNTDQ, PTA_SGX, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES,
31898 PTA_AVX512VBMI2, PTA_VPCLMULQDQ, PTA_AVX512BITALG, PTA_RDPID,
31899 PTA_PCONFIG, PTA_WBNOINVD, PTA_AVX512VP2INTERSECT, PTA_PTWRITE,
31900 PTA_AVX512BF16, PTA_WAITPKG, PTA_MOVDIRI, PTA_MOVDIR64B, PTA_ENQCMD,
31901 PTA_CLDEMOTE, PTA_SERIALIZE, PTA_TSXLDTRK, PTA_AMX_TILE, PTA_AMX_INT8,
31902 PTA_AMX_BF16, PTA_UINTR, PTA_HRESET, PTA_KL, PTA_WIDEKL, PTA_AVXVNNI,
31903 PTA_X86_64_BASELINE, PTA_X86_64_V2, PTA_X86_64_V3, PTA_X86_64_V4,
31904 PTA_CORE2, PTA_NEHALEM, PTA_WESTMERE, PTA_SANDYBRIDGE, PTA_IVYBRIDGE,
31905 PTA_HASWELL, PTA_BROADWELL, PTA_SKYLAKE, PTA_SKYLAKE_AVX512,
31906 PTA_CASCADELAKE, PTA_COOPERLAKE, PTA_CANNONLAKE, PTA_ICELAKE_CLIENT,
31907 PTA_ICELAKE_SERVER, PTA_TIGERLAKE, PTA_SAPPHIRERAPIDS, PTA_ALDERLAKE,
31908 PTA_KNL, PTA_BONNELL, PTA_SILVERMONT, PTA_GOLDMONT, PTA_GOLDMONT_PLUS,
31909 PTA_TREMONT, PTA_KNM): Use constexpr instead of const.
31911 2021-02-18 Jakub Jelinek <jakub@redhat.com>
31913 PR middle-end/99109
31914 * gimple-array-bounds.cc (build_zero_elt_array_type): Rename to ...
31915 (build_printable_array_type): ... this. Add nelts argument. For
31916 overaligned eltype, use TYPE_MAIN_VARIANT (eltype) instead. If
31917 nelts, call build_array_type_nelts.
31918 (array_bounds_checker::check_mem_ref): Use build_printable_array_type
31919 instead of build_zero_elt_array_type and build_array_type_nelts.
31921 2021-02-18 Jakub Jelinek <jakub@redhat.com>
31924 * config/i386/i386.c (distance_non_agu_define): Don't call
31925 extract_insn_cached here.
31926 (ix86_lea_outperforms): Save and restore recog_data around call
31927 to distance_non_agu_define and distance_agu_use.
31928 (ix86_ok_to_clobber_flags): Remove.
31929 (ix86_avoid_lea_for_add): Don't call ix86_ok_to_clobber_flags.
31930 (ix86_avoid_lea_for_addr): Likewise. Adjust function comment.
31931 * config/i386/i386.md (*lea<mode>): Change from define_insn_and_split
31932 into define_insn. Move the splitting to define_peephole2 and
31933 check there using peep2_regno_dead_p if FLAGS_REG is dead.
31935 2021-02-17 Julian Brown <julian@codesourcery.com>
31937 * gimplify.c (gimplify_scan_omp_clauses): Handle ATTACH_DETACH
31940 2021-02-17 Xi Ruoyao <xry111@mengyan1223.wang>
31943 * config/mips/mips.c (mips_symbol_insns): Do not use
31944 MSA_SUPPORTED_MODE_P if mode is MAX_MACHINE_MODE.
31946 2021-02-16 Vladimir N. Makarov <vmakarov@redhat.com>
31948 PR inline-asm/98096
31949 * stmt.c (resolve_operand_name_1): Take inout operands into account
31950 for access to labels by names.
31951 * doc/extend.texi: Describe counting operands for accessing labels.
31953 2021-02-16 Richard Biener <rguenther@suse.de>
31955 PR tree-optimization/38474
31956 * tree-ssa-structalias.c (variable_info::address_taken): New.
31957 (new_var_info): Initialize address_taken.
31958 (process_constraint): Set address_taken.
31959 (solve_constraints): Use the new address_taken flag rather
31960 than is_reg_var for sorting variables.
31961 (dump_constraint): Dump the variable number if the name
31964 2021-02-16 Jakub Jelinek <jakub@redhat.com>
31967 * tree-vect-stmts.c (vectorizable_simd_clone_call): For num_calls != 1
31968 multiply by 4096 and for inbranch by 8192.
31969 * config/i386/i386.c (ix86_simd_clone_usable): For TARGET_AVX512F,
31970 return 3, 2 or 1 for mangle letters 'b', 'c' or 'd'.
31972 2021-02-15 Maya Rashish <coypu@sdf.org>
31974 * config/aarch64/aarch64.c (aarch64_init_builtins):
31975 Call SUBTARGET_INIT_BUILTINS.
31977 2021-02-15 Peter Bergner <bergner@linux.ibm.com>
31979 PR rtl-optimization/98872
31980 * init-regs.c (initialize_uninitialized_regs): Skip initialization
31981 if CONST0_RTX is NULL.
31983 2021-02-15 Richard Sandiford <richard.sandiford@arm.com>
31985 PR rtl-optimization/98863
31986 * rtl-ssa/functions.h (function_info::bb_live_out_info): Delete.
31987 (function_info::build_info): Turn into a declaration, moving the
31988 definition to internals.h.
31989 (function_info::bb_walker): Declare.
31990 (function_info::create_reg_use): Likewise.
31991 (function_info::calculate_potential_phi_regs): Take a build_info
31993 (function_info::place_phis, function_info::create_ebbs): Declare.
31994 (function_info::calculate_ebb_live_in_for_debug): Likewise.
31995 (function_info::populate_backedge_phis): Delete.
31996 (function_info::start_block, function_info::end_block): Declare.
31997 (function_info::populate_phi_inputs): Delete.
31998 (function_info::m_potential_phi_regs): Move information to build_info.
31999 * rtl-ssa/internals.h: New file.
32000 (function_info::bb_phi_info): New class.
32001 (function_info::build_info): Moved from functions.h.
32002 Add a constructor and destructor.
32003 (function_info::build_info::ebb_use): Delete.
32004 (function_info::build_info::ebb_def): Likewise.
32005 (function_info::build_info::bb_live_out): Likewise.
32006 (function_info::build_info::tmp_ebb_live_in_for_debug): New variable.
32007 (function_info::build_info::potential_phi_regs): Likewise.
32008 (function_info::build_info::potential_phi_regs_for_debug): Likewise.
32009 (function_info::build_info::ebb_def_regs): Likewise.
32010 (function_info::build_info::bb_phis): Likewise.
32011 (function_info::build_info::bb_mem_live_out): Likewise.
32012 (function_info::build_info::bb_to_rpo): Likewise.
32013 (function_info::build_info::def_stack): Likewise.
32014 (function_info::build_info::old_def_stack_limit): Likewise.
32015 * rtl-ssa/internals.inl (function_info::build_info::record_reg_def):
32016 Remove the regno argument. Push the previous definition onto the
32017 definition stack where necessary.
32018 * rtl-ssa/accesses.cc: Include internals.h.
32019 * rtl-ssa/changes.cc: Likewise.
32020 * rtl-ssa/blocks.cc: Likewise.
32021 (function_info::build_info::build_info): Define.
32022 (function_info::build_info::~build_info): Likewise.
32023 (function_info::bb_walker): New class.
32024 (function_info::bb_walker::bb_walker): Define.
32025 (function_info::add_live_out_use): Convert a logarithmic-complexity
32026 test into a linear one. Allow the same definition to be passed
32028 (function_info::calculate_potential_phi_regs): Moved from
32029 functions.cc. Take a build_info parameter and store the
32030 information there instead.
32031 (function_info::place_phis): New function.
32032 (function_info::add_entry_block_defs): Update call to record_reg_def.
32033 (function_info::calculate_ebb_live_in_for_debug): New function.
32034 (function_info::add_phi_nodes): Use bb_phis to decide which
32035 registers need phi nodes and initialize ebb_def_regs accordingly.
32036 Do not add degenerate phis here.
32037 (function_info::add_artificial_accesses): Use create_reg_use.
32038 Assert that all definitions are listed in the DF LR sets.
32039 Update call to record_reg_def.
32040 (function_info::record_block_live_out): Record live-out register
32041 values in the phis of successor blocks. Use the live-out set
32042 when processing the last block in an EBB, instead of always
32043 using the live-in sets of successor blocks. AND the live sets
32044 with the set of registers that have been defined in the EBB,
32045 rather than with all potential phi registers. Cope correctly
32046 with branches back to the start of the current EBB.
32047 (function_info::start_block): New function.
32048 (function_info::end_block): Likewise.
32049 (function_info::populate_phi_inputs): Likewise.
32050 (function_info::create_ebbs): Likewise.
32051 (function_info::process_all_blocks): Rewrite into a multi-phase
32053 * rtl-ssa/functions.cc: Include internals.h.
32054 (function_info::calculate_potential_phi_regs): Move to blocks.cc.
32055 (function_info::init_function_data): Remove caller.
32056 * rtl-ssa/insns.cc: Include internals.h
32057 (function_info::create_reg_use): New function. Lazily any
32058 degenerate phis needed by the linear RPO view.
32059 (function_info::record_use): Use create_reg_use. When processing
32060 debug uses, use potential_phi_regs and test it before checking
32061 whether the register is live on entry to the current EBB. Lazily
32062 calculate ebb_live_in_for_debug.
32063 (function_info::record_call_clobbers): Update call to record_reg_def.
32064 (function_info::record_def): Likewise.
32066 2021-02-15 Martin Liska <mliska@suse.cz>
32068 * toplev.c (init_asm_output): Free output of
32069 gen_command_line_string function.
32070 (process_options): Likewise.
32072 2021-02-15 Martin Liska <mliska@suse.cz>
32074 * params.opt: Add 2 missing Param keywords.
32076 2021-02-15 Eric Botcazou <ebotcazou@adacore.com>
32078 * df-core.c (df_worklist_dataflow_doublequeue): Use proper cast.
32080 2021-02-15 Jakub Jelinek <jakub@redhat.com>
32082 PR tree-optimization/99079
32083 * match.pd (A % (pow2pcst << N) -> A & ((pow2pcst << N) - 1)): Remove
32084 useless tree_nop_conversion_p (type, TREE_TYPE (@3)) check. Instead
32085 require both type and TREE_TYPE (@1) to be integral types and either
32086 type having smaller or equal precision, or TREE_TYPE (@1) being
32087 unsigned type, or type being signed type. If TREE_TYPE (@1)
32088 doesn't have wrapping overflow, perform the subtraction of one in
32091 2021-02-14 Jan Hubicka <hubicka@ucw.cz>
32092 Richard Biener <rguether@suse.de>
32095 * ipa-reference.c (ipa_init): Only conditinally initialize
32096 reference_vars_to_consider.
32097 (propagate): Conditionally deninitialize reference_vars_to_consider.
32098 (ipa_reference_write_optimization_summary): Sanity check that
32099 reference_vars_to_consider is not allocated.
32101 2021-02-13 Levy Hsu <admin@levyhsu.com>
32104 * config/riscv/riscv-shorten-memrefs.c (pass_shorten_memrefs): Add
32105 extend parameter to get_si_mem_base_reg declaration.
32106 (get_si_mem_base_reg): Add extend parameter. Set it.
32107 (analyze): Pass extend arg to get_si_mem_base_reg.
32108 (transform): Likewise. Use it when rewriting mems.
32109 * config/riscv/riscv.c (riscv_legitimize_move): Check for subword
32110 loads and emit sign/zero extending load followed by subreg move.
32112 2021-02-13 Jim Wilson <jimw@sifive.com>
32115 * config/riscv/riscv.c (riscv_compressed_lw_address_p): Drop early
32116 exit when !reload_completed. Only perform check for compressed reg
32117 if reload_completed.
32118 (riscv_rtx_costs): In MEM case, when optimizing for size and
32119 shorten memrefs, if not compressible, then increase cost.
32121 2021-02-13 Jakub Jelinek <jakub@redhat.com>
32123 PR rtl-optimization/98439
32124 * recog.c (pass_split_before_regstack::gate): Enable even when
32125 pass_split_before_sched2 is enabled if -fselective-scheduling2 is
32128 2021-02-13 Jakub Jelinek <jakub@redhat.com>
32131 * config/i386/mmx.md (*mmx_pshufd_1): Add a combine splitter for
32132 swap of V2SImode elements in memory into DImode memory rotate by 32.
32134 2021-02-12 Martin Sebor <msebor@redhat.com>
32136 * tree-pretty-print.c (print_generic_expr_to_str): Update comment.
32138 2021-02-12 Richard Sandiford <richard.sandiford@arm.com>
32140 * rtl-ssa/accesses.cc (function_info::make_use_available): Use
32141 m_temp_obstack rather than m_obstack to allocate the temporary use.
32143 2021-02-12 Richard Sandiford <richard.sandiford@arm.com>
32145 * df-problems.c (df_lr_bb_local_compute): Treat partial definitions
32146 as read-modify operations.
32148 2021-02-12 Richard Biener <rguenther@suse.de>
32150 PR middle-end/38474
32151 * ipa-fnsummary.c (unmodified_parm_1): Only walk when
32152 fbi->aa_walk_budget is bigger than zero. Update
32153 fbi->aa_walk_budget.
32154 (param_change_prob): Likewise.
32155 * ipa-prop.c (detect_type_change_from_memory_writes):
32156 Properly account walk_aliased_vdefs.
32157 (parm_preserved_before_stmt_p): Canonicalize updates.
32158 (parm_ref_data_preserved_p): Likewise.
32159 (parm_ref_data_pass_through_p): Likewise.
32160 (determine_known_aggregate_parts): Account own alias queries.
32162 2021-02-12 Martin Liska <mliska@suse.cz>
32164 * opts-common.c (decode_cmdline_option): Release werror_arg.
32165 * opts.c (gen_producer_string): Release output of
32166 gen_command_line_string.
32168 2021-02-12 Richard Biener <rguenther@suse.de>
32170 PR tree-optimization/38474
32171 * params.opt (-param=max-store-chains-to-track=): New param.
32172 (-param=max-stores-to-track=): Likewise.
32173 * doc/invoke.texi (max-store-chains-to-track): Document.
32174 (max-stores-to-track): Likewise.
32175 * gimple-ssa-store-merging.c (pass_store_merging::m_n_chains):
32177 (pass_store_merging::m_n_stores): Likewise.
32178 (pass_store_merging::terminate_and_process_chain): Update
32179 m_n_stores and m_n_chains.
32180 (pass_store_merging::process_store): Likewise. Terminate
32181 oldest chains if the number of stores or chains get too large.
32182 (imm_store_chain_info::terminate_and_process_chain): Dump
32185 2021-02-11 Eric Botcazou <ebotcazou@adacore.com>
32187 * config/i386/winnt.c (i386_pe_seh_unwind_emit): When switching to
32188 the cold section, emit a nop before the directive if the previous
32189 active instruction can throw.
32191 2021-02-11 Peter Bergner <bergner@linux.ibm.com>
32194 * config/rs6000/predicates.md (mma_assemble_input_operand): Restrict
32195 memory addresses that are legal for quad word accesses.
32197 2021-02-11 Andrea Corallo <andrea.corallo@arm.com>
32200 * config/arm/thumb2.md (*doloop_end_internal): Generate
32201 alternative sequence to handle long range branches.
32203 2021-02-11 Joel Hutton <joel.hutton@arm.com>
32205 PR tree-optimization/98772
32206 * optabs-tree.c (supportable_half_widening_operation): New function
32207 to check for supportable V8QI->V8HI widening patterns.
32208 * optabs-tree.h (supportable_half_widening_operation): New function.
32209 * tree-vect-stmts.c (vect_create_half_widening_stmts): New function
32210 to create promotion stmts for V8QI->V8HI widening patterns.
32211 (vectorizable_conversion): Add case for V8QI->V8HI.
32213 2021-02-11 Richard Biener <rguenther@suse.de>
32215 * sparseset.h (SPARSESET_ELT_BITS): Remove.
32216 (SPARSESET_ELT_TYPE): Use unsigned int.
32217 * fwprop.c: Do not include sparseset.h.
32219 2021-02-10 Jakub Jelinek <jakub@redhat.com>
32222 * varasm.c (declare_weak): For -fsyntax-only, allow even
32223 TREE_ASM_WRITTEN function decls.
32225 2021-02-10 Jakub Jelinek <jakub@redhat.com>
32228 * config/i386/sse.md (fix<fixunssuffix>_truncv2sfv2di2,
32229 <insn>v8qiv8hi2, <insn>v8qiv8si2, <insn>v4qiv4si2, <insn>v4hiv4si2,
32230 <insn>v8qiv8di2, <insn>v4qiv4di2, <insn>v2qiv2di2, <insn>v4hiv4di2,
32231 <insn>v2hiv2di2, <insn>v2siv2di2): Force operands[1] into REG before
32232 calling simplify_gen_subreg on it.
32234 2021-02-10 Martin Liska <mliska@suse.cz>
32236 * config/nvptx/nvptx.c (nvptx_option_override): Use
32237 flag_patchable_function_entry instead of the removed
32238 function_entry_patch_area_size.
32240 2021-02-10 Martin Liska <mliska@suse.cz>
32242 PR tree-optimization/99002
32243 PR tree-optimization/99026
32244 * gimple-if-to-switch.cc (if_chain::is_beneficial): Fix memory
32245 leak when adjacent cases are merged.
32246 * tree-switch-conversion.c (switch_decision_tree::analyze_switch_statement): Use
32248 (make_pass_lower_switch): Remove trailing whitespace.
32249 * tree-switch-conversion.h (release_clusters): New.
32251 2021-02-10 Richard Biener <rguenther@suse.de>
32253 PR rtl-optimization/99054
32254 * cfgrtl.c (rtl-optimization/99054): Return an auto_vec.
32255 (fixup_partitions): Adjust.
32256 (rtl_verify_edges): Likewise.
32258 2021-02-10 Jakub Jelinek <jakub@redhat.com>
32260 PR middle-end/99007
32261 * gimplify.c (gimplify_scan_omp_clauses): For MEM_REF on reductions,
32262 temporarily disable gimplify_ctxp->into_ssa around gimplify_expr
32265 2021-02-10 Richard Biener <rguenther@suse.de>
32268 * ipa-pure-const.c (propagate_malloc): Use an auto_vec<>
32271 2021-02-10 Richard Biener <rguenther@suse.de>
32273 PR tree-optimization/99024
32274 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Only
32275 clear loop->aux if it is associated with the destroyed loop_vinfo.
32277 2021-02-10 Martin Liska <mliska@suse.cz>
32279 PR tree-optimization/99002
32280 * gimple-if-to-switch.cc (find_conditions): Fix memory leak
32283 2021-02-10 Martin Liska <mliska@suse.cz>
32286 * ipa-icf.c (sem_item::add_reference): Fix memory leak when
32287 a reference exists.
32289 2021-02-10 Jakub Jelinek <jakub@redhat.com>
32292 * dwarf2out.c (prune_unused_types_walk): Mark DW_TAG_variable DIEs
32293 at class scope for DWARF5+.
32295 2021-02-09 Eric Botcazou <ebotcazou@adacore.com>
32297 PR rtl-optimization/96015
32298 * reorg.c (skip_consecutive_labels): Minor comment tweaks.
32299 (relax_delay_slots): When deleting a jump to the next active
32300 instruction over a barrier, first delete the barrier if the
32301 jump is the only way to reach the target label.
32303 2021-02-09 Andre Vieira <andre.simoesdiasvieira@arm.com>
32305 * config/aarch64/aarch64-cost-tables.h: Add entries for vect.mul.
32306 * config/aarch64/aarch64.c (aarch64_rtx_mult_cost): Use vect.mul for
32307 vector multiplies and vect.alu for SSRA.
32308 * config/arm/aarch-common-protos.h (struct vector_cost_table): Define
32309 vect.mul cost field.
32310 * config/arm/aarch-cost-tables.h: Add entries for vect.mul.
32311 * config/arm/arm.c: Likewise.
32313 2021-02-09 Richard Biener <rguenther@suse.de>
32315 PR tree-optimization/98863
32316 * tree-ssa-sccvn.h (vn_avail::next_undo): Add.
32317 * tree-ssa-sccvn.c (last_pushed_avail): New global.
32318 (rpo_elim::eliminate_push_avail): Chain pushed avails.
32319 (unwind_state::avail_top): Add.
32320 (do_unwind): Rewrite unwinding of avail entries.
32321 (do_rpo_vn): Initialize last_pushed_avail and
32322 avail_top of the undo state.
32324 2021-02-09 Jakub Jelinek <jakub@redhat.com>
32326 PR middle-end/99004
32327 * calls.c (maybe_warn_rdwr_sizes): Change s0 and s1 type from
32328 const char * to char * and free those pointers after use.
32330 2021-02-09 Richard Biener <rguenther@suse.de>
32332 PR tree-optimization/99017
32333 * tree-vect-slp.c (vect_bb_vectorization_profitable_p): Allow
32334 zero vector cost entries.
32336 2021-02-08 Andre Vieira <andre.simoesdiasvieira@arm.com>
32338 PR middle-end/98974
32339 * tree-vect-stmts.c (vectorizable_condition): Remove shadow vec_num
32340 parameter in vectorizable_condition.
32342 2021-02-08 Richard Biener <rguenther@suse.de>
32345 * tree.c (walk_tree_1): Walk VECTOR_CST elements.
32347 2021-02-08 Martin Liska <mliska@suse.cz>
32350 * cfgexpand.c (pass_expand::execute): Parse per-function option
32351 flag_patchable_function_entry and use it.
32352 * common.opt: Remove function_entry_patch_area_size and
32353 function_entry_patch_area_start global variables.
32354 * opts.c (parse_and_check_patch_area): New function.
32355 (common_handle_option): Use it.
32356 * opts.h (parse_and_check_patch_area): New function.
32357 * toplev.c (process_options): Parse and use
32358 function_entry_patch_area_size.
32360 2021-02-08 Martin Sebor <msebor@redhat.com>
32362 * doc/extend.texi (attribute malloc): Correct typos.
32364 2021-02-05 Nathan Sidwell <nathan@acm.org>
32367 * gcc.c (driver::maybe_run_linker): Check for input file
32368 accessibility if not linking.
32370 2021-02-05 Richard Biener <rguenther@suse.de>
32372 PR tree-optimization/98855
32373 * tree-vectorizer.h (add_stmt_cost): New overload.
32374 * tree-vect-slp.c (li_cost_vec_cmp): New.
32375 (vect_bb_slp_scalar_cost): Cost individual loop regions
32376 separately. Account for the scalar instance root stmt.
32378 2021-02-05 Tom de Vries <tdevries@suse.de>
32381 * tree-switch-conversion.c (jump_table_cluster::emit): Add loc
32383 (bit_test_cluster::emit): Reuse location_t for newly created
32385 (switch_decision_tree::try_switch_expansion): Preserve
32387 * tree-switch-conversion.h: Change function signatures.
32389 2021-02-05 Jakub Jelinek <jakub@redhat.com>
32392 * config/i386/i386-options.c (m_NONE, m_ALL): Define.
32393 * config/i386/x86-tune.def (X86_TUNE_BRANCH_PREDICTION_HINTS,
32394 X86_TUNE_PROMOTE_QI_REGS): Use m_NONE instead of 0U.
32395 (X86_TUNE_QIMODE_MATH): Use m_ALL instead of ~0U.
32397 2021-02-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
32399 * config/aarch64/aarch64-simd-builtins.def (get_high): Define builtin.
32400 * config/aarch64/aarch64-simd.md (aarch64_get_high<mode>): Define.
32401 * config/aarch64/arm_neon.h (__GET_HIGH): Delete.
32402 (vget_high_f16): Reimplement using new builtin.
32403 (vget_high_f32): Likewise.
32404 (vget_high_f64): Likewise.
32405 (vget_high_p8): Likewise.
32406 (vget_high_p16): Likewise.
32407 (vget_high_p64): Likewise.
32408 (vget_high_s8): Likewise.
32409 (vget_high_s16): Likewise.
32410 (vget_high_s32): Likewise.
32411 (vget_high_s64): Likewise.
32412 (vget_high_u8): Likewise.
32413 (vget_high_u16): Likewise.
32414 (vget_high_u32): Likewise.
32415 (vget_high_u64): Likewise.
32417 2021-02-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
32419 * config/aarch64/aarch64-simd-builtins.def (get_low): Define builtin.
32420 * config/aarch64/aarch64-simd.md (aarch64_get_low<mode>): Define.
32421 * config/aarch64/arm_neon.h (__GET_LOW): Delete.
32422 (vget_low_f16): Reimplement using new builtin.
32423 (vget_low_f32): Likewise.
32424 (vget_low_f64): Likewise.
32425 (vget_low_p8): Likewise.
32426 (vget_low_p16): Likewise.
32427 (vget_low_p64): Likewise.
32428 (vget_low_s8): Likewise.
32429 (vget_low_s16): Likewise.
32430 (vget_low_s32): Likewise.
32431 (vget_low_s64): Likewise.
32432 (vget_low_u8): Likewise.
32433 (vget_low_u16): Likewise.
32434 (vget_low_u32): Likewise.
32435 (vget_low_u64): Likewise.
32437 2021-02-05 Kito Cheng <kito.cheng@sifive.com>
32439 * gcc.c (print_multilib_info): Check all required argument is provided
32442 2021-02-05 liuhongt <hongtao.liu@intel.com>
32445 * config/i386/i386-expand.c (ix86_expand_sse_cmp): Don't
32446 generate integer mask comparison for 128/256-bits vector when
32447 op_true/op_false is NULL_RTX or CONSTM1_RTX/CONST0_RTX. Also
32448 delete redundant !maskcmp condition.
32449 (ix86_expand_int_vec_cmp): Ditto but no redundant deletion
32451 (ix86_expand_sse_movcc): Delete definition of maskcmp, add the
32452 condition directly to if (maskcmp), add extra check for
32453 cmpmode, it should be MODE_INT.
32454 (ix86_expand_fp_vec_cmp): Pass NULL to ix86_expand_sse_cmp's
32455 parameters op_true/op_false.
32456 (ix86_use_mask_cmp_p): New.
32458 2021-02-05 liuhongt <hongtao.liu@intel.com>
32461 * config/i386/x86-tune.def (X86_TUNE_AVX256_UNALIGNED_LOAD_OPTIMAL):
32462 Remove m_GENERIC from ~list.
32463 (X86_TUNE_AVX256_UNALIGNED_STORE_OPTIMAL): Ditto.
32465 2021-02-04 David Malcolm <dmalcolm@redhat.com>
32468 * diagnostic-show-locus.c (compatible_locations_p): Require
32469 locations in the same macro map to be either both from the
32470 macro definition, or both from the macro arguments.
32472 2021-02-04 Jonathan Wright <jonathan.wright@arm.com>
32474 * config/aarch64/aarch64-simd-builtins.def: Add
32475 [su]mull_hi_lane[q] builtin generator macros.
32476 * config/aarch64/aarch64-simd.md
32477 (aarch64_<su>mull_hi_lane<mode>_insn): Define.
32478 (aarch64_<su>mull_hi_lane<mode>): Define.
32479 (aarch64_<su>mull_hi_laneq<mode>_insn): Define.
32480 (aarch64_<su>mull_hi_laneq<mode>): Define.
32481 * config/aarch64/arm_neon.h (vmull_high_lane_s16): Use RTL
32482 builtin instead of inline asm.
32483 (vmull_high_lane_s32): Likewise.
32484 (vmull_high_lane_u16): Likewise.
32485 (vmull_high_lane_u32): Likewise.
32486 (vmull_high_laneq_s16): Likewise.
32487 (vmull_high_laneq_s32): Likewise.
32488 (vmull_high_laneq_u16): Likewise.
32489 (vmull_high_laneq_u32): Liekwise.
32491 2021-02-04 Jonathan Wright <jonathan.wright@arm.com>
32493 * config/aarch64/aarch64-simd-builtins.def: Add [su]mull_hi_n
32494 builtin generator macros.
32495 * config/aarch64/aarch64-simd.md
32496 (aarch64_<su>mull_hi_n<mode>_insn): Define.
32497 (aarch64_<su>mull_hi_n<mode>): Define.
32498 * config/aarch64/arm_neon.h (vmull_high_n_s16): Use RTL builtin
32499 instead of inline asm.
32500 (vmull_high_n_s32): Likewise.
32501 (vmull_high_n_u16): Likewise.
32502 (vmull_high_n_u32): Likewise.
32504 2021-02-04 Richard Biener <rguenther@suse.de>
32506 PR tree-optimization/98855
32507 * tree-vect-loop.c (vectorizable_phi): Do not cost
32508 single-argument PHIs.
32509 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Likewise.
32510 * tree-vect-stmts.c (vectorizable_bswap): Also perform
32511 costing for SLP operation.
32513 2021-02-04 Martin Liska <mliska@suse.cz>
32515 * doc/extend.texi: Mention -mprefer-vector-width in target
32518 2021-02-03 Martin Sebor <msebor@redhat.com>
32520 PR tree-optimization/98937
32521 * tree-ssa-strlen.c (strlen_dom_walker::~strlen_dom_walker): Define.
32522 Flush pointer_query cache.
32524 2021-02-03 Aaron Sawdey <acsawdey@linux.ibm.com>
32526 * config/rs6000/genfusion.pl (gen_2logical): Add missing
32527 fixes based on patch review.
32528 * config/rs6000/fusion.md: Regenerate file.
32530 2021-02-03 Aaron Sawdey <acsawdey@linux.ibm.com>
32532 * config/rs6000/t-rs6000: Comment out auto generation of
32535 2021-02-03 Andrew Stubbs <ams@codesourcery.com>
32537 * config/gcn/gcn-opts.h (enum processor_type): Add PROCESSOR_GFX908.
32538 * config/gcn/gcn.c (gcn_omp_device_kind_arch_isa): Add gfx908.
32539 (output_file_start): Add gfx908.
32540 * config/gcn/gcn.opt (gpu_type): Add gfx908.
32541 * config/gcn/t-gcn-hsa (MULTILIB_OPTIONS): Add march=gfx908.
32542 (MULTILIB_DIRNAMES): Add gfx908.
32543 * config/gcn/mkoffload.c (EF_AMDGPU_MACH_AMDGCN_GFX908): New define.
32544 (main): Recognize gfx908.
32545 * config/gcn/t-omp-device: Add gfx908.
32547 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
32549 * config/aarch64/aarch64-simd-builtins.def: Add
32550 [su]mlsl_hi_lane[q] builtin macro generators.
32551 * config/aarch64/aarch64-simd.md
32552 (aarch64_<su>mlsl_hi_lane<mode>_insn): Define.
32553 (aarch64_<su>mlsl_hi_lane<mode>): Define.
32554 (aarch64_<su>mlsl_hi_laneq<mode>_insn): Define.
32555 (aarch64_<su>mlsl_hi_laneq<mode>): Define.
32556 * config/aarch64/arm_neon.h (vmlsl_high_lane_s16): Use RTL
32557 builtin instead of inline asm.
32558 (vmlsl_high_lane_s32): Likewise.
32559 (vmlsl_high_lane_u16): Likewise.
32560 (vmlsl_high_lane_u32): Likewise.
32561 (vmlsl_high_laneq_s16): Likewise.
32562 (vmlsl_high_laneq_s32): Likewise.
32563 (vmlsl_high_laneq_u16): Likewise.
32564 (vmlsl_high_laneq_u32): Likewise.
32565 (vmlal_high_laneq_u32): Likewise.
32567 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
32569 * config/aarch64/aarch64-simd-builtins.def: Add
32570 [su]mlal_hi_lane[q] builtin generator macros.
32571 * config/aarch64/aarch64-simd.md
32572 (aarch64_<su>mlal_hi_lane<mode>_insn): Define.
32573 (aarch64_<su>mlal_hi_lane<mode>): Define.
32574 (aarch64_<su>mlal_hi_laneq<mode>_insn): Define.
32575 (aarch64_<su>mlal_hi_laneq<mode>): Define.
32576 * config/aarch64/arm_neon.h (vmlal_high_lane_s16): Use RTL
32577 builtin instead of inline asm.
32578 (vmlal_high_lane_s32): Likewise.
32579 (vmlal_high_lane_u16): Likewise.
32580 (vmlal_high_lane_u32): Likewise.
32581 (vmlal_high_laneq_s16): Likewise.
32582 (vmlal_high_laneq_s32): Likewise.
32583 (vmlal_high_laneq_u16): Likewise.
32584 (vmlal_high_laneq_u32): Likewise.
32586 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
32588 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlsl_hi_n
32589 builtin generator macros.
32590 * config/aarch64/aarch64-simd.md (aarch64_<su>mlsl_hi_n<mode>_insn):
32592 (aarch64_<su>mlsl_hi_n<mode>): Define.
32593 * config/aarch64/arm_neon.h (vmlsl_high_n_s16): Use RTL builtin
32594 instead of inline asm.
32595 (vmlsl_high_n_s32): Likewise.
32596 (vmlsl_high_n_u16): Likewise.
32597 (vmlsl_high_n_u32): Likewise.
32599 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
32601 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlal_hi_n
32602 builtin generator macros.
32603 * config/aarch64/aarch64-simd.md (aarch64_<su>mlal_hi_n<mode>_insn):
32605 (aarch64_<su>mlal_hi_n<mode>): Define.
32606 * config/aarch64/arm_neon.h (vmlal_high_n_s16): Use RTL builtin
32607 instead of inline asm.
32608 (vmlal_high_n_s32): Likewise.
32609 (vmlal_high_n_u16): Likewise.
32610 (vmlal_high_n_u32): Likewise.
32612 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
32614 * config/aarch64/aarch64-simd-builtins.def: Add RTL builtin
32616 * config/aarch64/aarch64-simd.md (*aarch64_<su>mlal_hi<mode>):
32618 (aarch64_<su>mlal_hi<mode>_insn): This.
32619 (aarch64_<su>mlal_hi<mode>): Define.
32620 * config/aarch64/arm_neon.h (vmlal_high_s8): Use RTL builtin
32621 instead of inline asm.
32622 (vmlal_high_s16): Likewise.
32623 (vmlal_high_s32): Likewise.
32624 (vmlal_high_u8): Likewise.
32625 (vmlal_high_u16): Likewise.
32626 (vmlal_high_u32): Likewise.
32628 2021-02-03 Ilya Leoshkevich <iii@linux.ibm.com>
32630 * lra-spills.c (remove_pseudos): Call lra_update_insn_recog_data()
32631 after calling alter_subreg() on a (mem).
32633 2021-02-03 Martin Liska <mliska@suse.cz>
32636 * lto-streamer-out.c (produce_lto_section): Fill up missing
32638 * lto-streamer.h (struct lto_section): Add _padding field.
32640 2021-02-03 Richard Biener <rguenther@suse.de>
32642 * lto-streamer.c (lto_get_section_name): Free temporary
32644 * tree-loop-distribution.c
32645 (loop_distribution::merge_dep_scc_partitions): Free edge data.
32647 2021-02-03 Jakub Jelinek <jakub@redhat.com>
32649 PR middle-end/97487
32650 * ifcvt.c (noce_can_force_operand): New function.
32651 (noce_emit_move_insn): Use it.
32652 (noce_try_sign_mask): Likewise. Formatting fix.
32654 2021-02-03 Jakub Jelinek <jakub@redhat.com>
32656 PR middle-end/97971
32657 * lra-constraints.c (process_alt_operands): For inline asm, don't call
32658 fatal_insn, but instead return false.
32660 2021-02-03 Jakub Jelinek <jakub@redhat.com>
32662 PR tree-optimization/98287
32663 * config/i386/mmx.md (<insn><mode>3): For shifts don't enable expander
32666 2021-02-03 Tamar Christina <tamar.christina@arm.com>
32668 PR tree-optimization/98928
32669 * tree-vect-loop.c (vect_analyze_loop_2): Change
32670 STMT_VINFO_SLP_VECT_ONLY to STMT_VINFO_SLP_VECT_ONLY_PATTERN.
32671 * tree-vect-slp-patterns.c (complex_pattern::build): Likewise.
32672 * tree-vectorizer.h (STMT_VINFO_SLP_VECT_ONLY_PATTERN): New.
32673 (class _stmt_vec_info): Add slp_vect_pattern_only_p.
32675 2021-02-02 Richard Biener <rguenther@suse.de>
32677 * gimple-loop-interchange.cc (prepare_data_references):
32679 * gimple-loop-jam.c (tree_loop_unroll_and_jam): Likewise.
32680 * tree-ssa-loop-im.c (hoist_memory_references): Likewise.
32681 * tree-vect-stmts.c (vectorizable_condition): Do not
32683 (vectorizable_comparison): Likewise.
32685 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
32687 * config/aarch64/aarch64-simd-builtins.def (ursqrte): Define builtin.
32688 * config/aarch64/aarch64-simd.md (aarch64_ursqrte<mode>): New pattern.
32689 * config/aarch64/arm_neon.h (vrsqrte_u32): Reimplement using builtin.
32690 (vrsqrteq_u32): Likewise.
32692 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
32694 * config/aarch64/aarch64-simd-builtins.def (sqxtun2): Define builtin.
32695 * config/aarch64/aarch64-simd.md (aarch64_sqxtun2<mode>_le): Define.
32696 (aarch64_sqxtun2<mode>_be): Likewise.
32697 (aarch64_sqxtun2<mode>): Likewise.
32698 * config/aarch64/arm_neon.h (vqmovun_high_s16): Reimplement using builtin.
32699 (vqmovun_high_s32): Likewise.
32700 (vqmovun_high_s64): Likewise.
32701 * config/aarch64/iterators.md (UNSPEC_SQXTUN2): Define.
32703 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
32705 * config/aarch64/aarch64-simd-builtins.def (bfdot_lane, bfdot_laneq): Use
32707 (bfmlalb_lane, bfmlalt_lane, bfmlalb_lane_q, bfmlalt_lane_q): Use FP flags.
32709 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
32711 * config/aarch64/aarch64-simd-builtins.def (fcmla_lane0, fcmla_lane90,
32712 fcmla_lane180, fcmla_lane270, fcmlaq_lane0, fcmlaq_lane90, fcmlaq_lane180,
32713 fcmlaq_lane270, scvtf, ucvtf, fcvtzs, fcvtzu, scvtfsi, scvtfdi, ucvtfsi,
32714 ucvtfdi, fcvtzshf, fcvtzuhf, fmlal_lane_low, fmlsl_lane_low,
32715 fmlal_laneq_low, fmlsl_laneq_low, fmlalq_lane_low, fmlslq_lane_low,
32716 fmlalq_laneq_low, fmlslq_laneq_low, fmlal_lane_high, fmlsl_lane_high,
32717 fmlal_laneq_high, fmlsl_laneq_high, fmlalq_lane_high, fmlslq_lane_high,
32718 fmlalq_laneq_high, fmlslq_laneq_high): Use FP flags.
32720 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
32722 * config/aarch64/aarch64-builtins.c (FLAG_LOAD): Define.
32723 * config/aarch64/aarch64-simd-builtins.def (ld1x2, ld2, ld3, ld4, ld2r,
32724 ld3r, ld4r, ld1, ld1x3, ld1x4): Use LOAD flags.
32726 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
32728 * config/aarch64/aarch64-simd-builtins.def (combine, zip1, zip2,
32729 uzp1, uzp2, trn1, trn2, simd_bsl): Use AUTO_FP flags.
32731 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
32733 * config/aarch64/aarch64-simd-builtins.def (clrsb, clz, ctz, popcount,
32734 vec_smult_lane_, vec_smlal_lane_, vec_smult_laneq_, vec_smlal_laneq_,
32735 vec_umult_lane_, vec_umlal_lane_, vec_umult_laneq_, vec_umlal_laneq_,
32736 ashl, sshl, ushl, srshl, urshl, sdot_lane, udot_lane, sdot_laneq,
32737 udot_laneq, usdot_lane, usdot_laneq, sudot_lane, sudot_laneq, ashr,
32738 ashr_simd, lshr, lshr_simd, srshr_n, urshr_n, ssra_n, usra_n, srsra_n,
32739 ursra_n, sshll_n, ushll_n, sshll2_n, ushll2_n, ssri_n, usri_n, ssli_n,
32740 ssli_n, usli_n, bswap, rbit, simd_bsl, eor3q, rax1q, xarq, bcaxq): Use
32741 NONE builtin flags.
32743 2021-02-02 Jakub Jelinek <jakub@redhat.com>
32745 PR tree-optimization/98848
32746 * tree-vect-patterns.c (vect_recog_over_widening_pattern): Punt if
32747 STMT_VINFO_DEF_TYPE (last_stmt_info) is vect_reduction_def.
32749 2021-02-02 Kito Cheng <kito.cheng@sifive.com>
32752 * expr.c: Check mode before calling store_expr.
32754 2021-02-02 Christophe Lyon <christophe.lyon@linaro.org>
32756 * config/arm/iterators.md (supf): Remove VORNQ_S and VORNQ_U.
32758 * config/arm/mve.md (mve_vornq_s<mode>): New entry for vorn
32759 instruction using expression ior.
32760 (mve_vornq_u<mode>): New expander.
32761 (mve_vornq_f<mode>): Use ior code instead of unspec.
32762 * config/arm/unspecs.md (VORNQ_S, VORNQ_U, VORNQ_F): Remove.
32764 2021-02-02 Alexandre Oliva <oliva@adacore.com>
32766 * tree-nested.c (convert_nonlocal_reference_op): Move
32767 current_function_decl restore after re-gimplification.
32768 (convert_local_reference_op): Likewise.
32770 2021-02-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
32772 * config/aarch64/aarch64-simd-builtins.def (rshrn, rshrn2):
32774 * config/aarch64/aarch64-simd.md (aarch64_rshrn<mode>_insn_le):
32776 (aarch64_rshrn<mode>_insn_be): Likewise.
32777 (aarch64_rshrn<mode>): Likewise.
32778 (aarch64_rshrn2<mode>_insn_le): Likewise.
32779 (aarch64_rshrn2<mode>_insn_be): Likewise.
32780 (aarch64_rshrn2<mode>): Likewise.
32781 * config/aarch64/aarch64.md (unspec): Add UNSPEC_RSHRN.
32782 * config/aarch64/arm_neon.h (vrshrn_high_n_s16): Reimplement
32784 (vrshrn_high_n_s32): Likewise.
32785 (vrshrn_high_n_s64): Likewise.
32786 (vrshrn_high_n_u16): Likewise.
32787 (vrshrn_high_n_u32): Likewise.
32788 (vrshrn_high_n_u64): Likewise.
32789 (vrshrn_n_s16): Likewise.
32790 (vrshrn_n_s32): Likewise.
32791 (vrshrn_n_s64): Likewise.
32792 (vrshrn_n_u16): Likewise.
32793 (vrshrn_n_u32): Likewise.
32794 (vrshrn_n_u64): Likewise.
32796 2021-02-01 Sergei Trofimovich <siarheit@google.com>
32798 PR tree-optimization/98499
32799 * ipa-modref.c (analyze_ssa_name_flags): treat RVO
32800 conservatively and assume all possible side-effects.
32802 2021-02-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
32804 * config/aarch64/aarch64-simd-builtins.def (vec_unpacks_hi,
32805 vec_unpacku_hi_): Define builtins.
32806 * config/aarch64/arm_neon.h (vmovl_high_s8): Reimplement using
32808 (vmovl_high_s16): Likewise.
32809 (vmovl_high_s32): Likewise.
32810 (vmovl_high_u8): Likewise.
32811 (vmovl_high_u16): Likewise.
32812 (vmovl_high_u32): Likewise.
32814 2021-02-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
32816 * config/aarch64/aarch64-simd-builtins.def (sabdl, uabdl):
32818 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl<mode>): New
32820 * config/aarch64/aarch64.md (unspec): Define UNSPEC_SABDL,
32822 * config/aarch64/arm_neon.h (vabdl_s8): Reimplemet using
32824 (vabdl_s16): Likewise.
32825 (vabdl_s32): Likewise.
32826 (vabdl_u8): Likewise.
32827 (vabdl_u16): Likewise.
32828 (vabdl_u32): Likewise.
32829 * config/aarch64/iterators.md (ABDL): New int iterator.
32830 (sur): Handle UNSPEC_SABDL, UNSPEC_UABDL.
32832 2021-02-01 Martin Sebor <msebor@redhat.com>
32834 * tree.h (BLOCK_VARS): Add comment.
32835 (BLOCK_SUBBLOCKS): Same.
32836 (BLOCK_SUPERCONTEXT): Same.
32837 (BLOCK_ABSTRACT_ORIGIN): Same.
32838 (inlined_function_outer_scope_p): Same.
32840 2021-02-01 Martin Sebor <msebor@redhat.com>
32842 PR middle-end/97172
32843 * attribs.c (attr_access::free_lang_data): Define new function.
32844 * attribs.h (attr_access::free_lang_data): Declare new function.
32846 2021-02-01 Richard Biener <rguenther@suse.de>
32848 * vec.h (auto_vec::auto_vec): Add memory stat parameters
32850 * bitmap.h (auto_bitmap::auto_bitmap): Likewise.
32852 2021-02-01 Tamar Christina <tamar.christina@arm.com>
32854 * config/aarch64/aarch64-simd.md (aarch64_<su>mlal_n<mode>,
32855 aarch64_<su>mlsl<mode>, aarch64_<su>mlsl_n<mode>): Flip mult operands.
32857 2021-02-01 Richard Biener <rguenther@suse.de>
32859 PR rtl-optimization/98863
32860 * config/i386/i386-features.c (convert_scalars_to_vector):
32861 Set DF_RD_PRUNE_DEAD_DEFS.
32863 2021-01-31 Eric Botcazou <ebotcazou@adacore.com>
32865 * system.h (SIZE_MAX): Define if not already defined.
32867 2021-01-30 Aaron Sawdey <acsawdey@linux.ibm.com>
32869 * config/rs6000/genfusion.pl (gen_2logical): New function to
32870 generate patterns for logical-logical fusion.
32871 * config/rs6000/fusion.md: Regenerated patterns.
32872 * config/rs6000/rs6000-cpus.def: Add
32873 OPTION_MASK_P10_FUSION_2LOGICAL.
32874 * config/rs6000/rs6000.c (rs6000_option_override_internal):
32875 Enable logical-logical fusion for p10.
32876 * config/rs6000/rs6000.opt: Add -mpower10-fusion-2logical.
32878 2021-01-30 David Edelsohn <dje.gcc@gmail.com>
32880 * config/rs6000/rs6000.opt: Add periods to new AIX options.
32882 2021-01-30 David Edelsohn <dje.gcc@gmail.com>
32884 * config/rs6000/rs6000.opt (mabi=vec-extabi): New.
32885 (mabi=vec-default): New.
32886 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Define
32887 __EXTABI__ for AIX Vector extended ABI.
32888 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Print AIX Vector
32890 (conditional_register_usage): If AIX vec_extabi enabled, vs20-vs31
32892 * doc/invoke.texi (PowerPC mabi): Add AIX vec-extabi and vec-default.
32894 2021-01-30 Jakub Jelinek <jakub@redhat.com>
32896 * config/i386/i386-features.c (remove_partial_avx_dependency): Clear
32897 DF_DEFER_INSN_RESCAN after calling df_process_deferred_rescans.
32899 2021-01-29 Vladimir N. Makarov <vmakarov@redhat.com>
32902 * lra-constraints.c (in_class_p): Don't narrow class only for REG
32905 2021-01-29 Will Schmidt <will_schmidt@vnet.ibm.com>
32907 * config/rs6000/rs6000-call.c (rs6000_expand_binup_builtin): Add
32908 clauses for CODE_FOR_vsx_xvcvuxddp_scale and
32909 CODE_FOR_vsx_xvcvsxddp_scale to the parameter checking code.
32911 2021-01-29 Andrew MacLeod <amacleod@redhat.com>
32913 PR tree-optimization/98866
32914 * gimple-range-gori.h (gori_compute:set_range_invariant): New.
32915 * gimple-range-gori.cc (gori_map::set_range_invariant): New.
32916 (gori_map::m_maybe_invariant): Rename from all_outgoing.
32917 (gori_map::gori_map): Rename all_outgoing to m_maybe_invariant.
32918 (gori_map::is_export_p): Ditto.
32919 (gori_map::calculate_gori): Ditto.
32920 (gori_compute::set_range_invariant): New.
32921 * gimple-range.cc (gimple_ranger::range_of_stmt): Set range
32922 invariant for pointers evaluating to [1, +INF].
32924 2021-01-29 Richard Biener <rguenther@suse.de>
32926 PR rtl-optimization/98863
32927 * config/i386/i386-features.c (remove_partial_avx_dependency):
32928 Do not perform DF analysis.
32929 (pass_data_remove_partial_avx_dependency): Remove
32932 2021-01-29 Jonathan Wright <jonathan.wright@arm.com>
32934 * config/aarch64/aarch64-simd-builtins.def: Add [su]mull_n
32935 builtin generator macros.
32936 * config/aarch64/aarch64-simd.md (aarch64_<su>mull_n<mode>):
32938 * config/aarch64/arm_neon.h (vmull_n_s16): Use RTL builtin
32939 instead of inline asm.
32940 (vmull_n_s32): Likewise.
32941 (vmull_n_u16): Likewise.
32942 (vmull_n_u32): Likewise.
32944 2021-01-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
32946 * config/aarch64/aarch64-simd-builtins.def (sabdl2, uabdl2):
32948 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
32950 (aarch64_<sur>abdl2<mode>): ... This.
32951 (<sur>sadv16qi): Adjust use of above.
32952 * config/aarch64/arm_neon.h (vabdl_high_s8): Reimplement using
32954 (vabdl_high_s16): Likewise.
32955 (vabdl_high_s32): Likewise.
32956 (vabdl_high_u8): Likewise.
32957 (vabdl_high_u16): Likewise.
32958 (vabdl_high_u32): Likewise.
32960 2021-01-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
32962 * config/aarch64/aarch64-simd-builtins.def (sabal2): Define
32964 (uabal2): Likewise.
32965 * config/aarch64/aarch64-simd.md (aarch64_<sur>abal2<mode>): New
32967 * config/aarch64/aarch64.md (unspec): Add UNSPEC_SABAL2 and
32969 * config/aarch64/arm_neon.h (vabal_high_s8): Reimplement using
32971 (vabal_high_s16): Likewise.
32972 (vabal_high_s32): Likewise.
32973 (vabal_high_u8): Likewise.
32974 (vabal_high_u16): Likewise.
32975 (vabal_high_u32): Likewise.
32976 * config/aarch64/iterators.md (ABAL2): New mode iterator.
32977 (sur): Handle UNSPEC_SABAL2, UNSPEC_UABAL2.
32979 2021-01-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
32981 * config/aarch64/aarch64-simd-builtins.def (sabal): Define
32984 * config/aarch64/aarch64-simd.md (aarch64_<sur>abal<mode>_4):
32986 (aarch64_<sur>abal<mode>): ... This
32987 (<sur>sadv16qi): Adust use of the above.
32988 * config/aarch64/arm_neon.h (vabal_s8): Reimplement using
32990 (vabal_s16): Likewise.
32991 (vabal_s32): Likewise.
32992 (vabal_u8): Likewise.
32993 (vabal_u16): Likewise.
32994 (vabal_u32): Likewise.
32996 2021-01-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
32998 * config/aarch64/aarch64-simd-builtins.def (saddlv, uaddlv):
33000 * config/aarch64/aarch64-simd.md (aarch64_<su>addlv<mode>):
33002 * config/aarch64/arm_neon.h (vaddlv_s8): Reimplement using
33004 (vaddlv_s16): Likewise.
33005 (vaddlv_u8): Likewise.
33006 (vaddlv_u16): Likewise.
33007 (vaddlvq_s8): Likewise.
33008 (vaddlvq_s16): Likewise.
33009 (vaddlvq_s32): Likewise.
33010 (vaddlvq_u8): Likewise.
33011 (vaddlvq_u16): Likewise.
33012 (vaddlvq_u32): Likewise.
33013 (vaddlv_s32): Likewise.
33014 (vaddlv_u32): Likewise.
33015 * config/aarch64/iterators.md (VDQV_L): New mode iterator.
33016 (unspec): Add UNSPEC_SADDLV, UNSPEC_UADDLV.
33017 (Vwstype): New mode attribute.
33019 (VWIDE_S): Likewise.
33020 (USADDLV): New int iterator.
33021 (su): Handle UNSPEC_SADDLV, UNSPEC_UADDLV.
33023 2021-01-29 Jonathan Wright <jonathan.wright@arm.com>
33025 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlsl_lane[q]
33026 builtin generator macros.
33027 * config/aarch64/aarch64-simd.md (aarch64_vec_<su>mlsl_lane<Qlane>):
33029 * config/aarch64/arm_neon.h (vmlsl_lane_s16): Use RTL builtin
33030 instead of inline asm.
33031 (vmlsl_lane_s32): Likewise.
33032 (vmlsl_lane_u16): Likewise.
33033 (vmlsl_lane_u32): Likewise.
33034 (vmlsl_laneq_s16): Likewise.
33035 (vmlsl_laneq_s32): Likewise.
33036 (vmlsl_laneq_u16): Likewise.
33037 (vmlsl_laneq_u32): Likewise.
33039 2021-01-29 Richard Biener <rguenther@suse.de>
33041 * doc/invoke.texi (--param max-gcse-memory): Document unit
33043 * gcse.c (gcse_or_cprop_is_too_expensive): Adjust.
33044 * params.opt (--param max-gcse-memory): Adjust default and
33045 document unit of size.
33047 2021-01-29 Richard Biener <rguenther@suse.de>
33049 PR rtl-optimization/98863
33050 * gcse.c (gcse_or_cprop_is_too_expensive): Use unsigned
33051 HOST_WIDE_INT for the memory estimate.
33053 2021-01-29 Bin Cheng <bin.cheng@linux.alibaba.com>
33054 Richard Biener <rguenther@suse.de>
33056 PR tree-optimization/97627
33057 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
33058 Do not analyze fake edges.
33060 2021-01-29 Richard Biener <rguenther@suse.de>
33062 PR rtl-optimization/98144
33063 * df.h (df_mir_bb_info): Add con_visited member.
33064 * df-problems.c (df_mir_alloc): Initialize con_visited,
33065 do not fully populate IN and OUT.
33066 (df_mir_reset): Likewise.
33067 (df_mir_confluence_0): Set con_visited.
33068 (df_mir_confluence_n): Properly handle implicitely
33069 fully populated IN and OUT as designated by con_visited
33070 and update con_visited accordingly.
33072 2021-01-29 Jakub Jelinek <jakub@redhat.com>
33075 * config/arm/vec-common.md (mve_vshlq_<supf><mode>,
33076 vashl<mode>3, vashr<mode>3, vlshr<mode>3): Add
33077 && !TARGET_REALLY_IWMMXT to conditions.
33079 2021-01-29 Jakub Jelinek <jakub@redhat.com>
33082 * cfgbuild.c (find_bb_boundaries): Reset debug_insn when seeing
33085 2021-01-28 Marek Polacek <polacek@redhat.com>
33088 * stor-layout.c (finalize_type_size): If we reset TYPE_USER_ALIGN in
33089 the main variant, maybe reset it in its variants too.
33090 * tree.c (check_base_type): Return true only if TYPE_USER_ALIGN match.
33091 (check_aligned_type): Check if TYPE_USER_ALIGN match.
33093 2021-01-28 Christophe Lyon <christophe.lyon@linaro.org>
33096 * config/arm/arm.c (arm_rtx_costs_internal): Adjust cost of vector
33097 of constant zero for comparisons.
33099 2021-01-28 Michael Meissner <meissner@linux.ibm.com>
33101 * config/rs6000/rs6000.c (rs6000_mangle_decl_assembler_name): Add
33102 support for mapping built-in function names for long double
33103 built-in functions if long double is IEEE 128-bit.
33105 2021-01-28 Jonathan Wright <jonathan.wright@arm.com>
33107 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlsl_n
33108 builtin generator macros.
33109 * config/aarch64/aarch64-simd.md (aarch64_<su>mlsl_n<mode>):
33111 * config/aarch64/arm_neon.h (vmlsl_n_s16): Use RTL builtin
33112 instead of inline asm.
33113 (vmlsl_n_s32): Likewise.
33114 (vmlsl_n_u16): Likewise.
33115 (vmlsl_n_u32): Likewise.
33117 2021-01-28 Jonathan Wright <jonathan.wright@arm.com>
33119 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlal_n
33120 builtin generator macros.
33121 * config/aarch64/aarch64-simd.md (aarch64_<su>mlal_n<mode>):
33123 * config/aarch64/arm_neon.h (vmlal_n_s16): Use RTL builtin
33124 instead of inline asm.
33125 (vmlal_n_s32): Likewise.
33126 (vmlal_n_u16): Likewise.
33127 (vmlal_n_u32): Likewise.
33129 2021-01-28 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
33131 * config/aarch64/aarch64-simd-builtins.def (shrn2): Define
33133 * config/aarch64/aarch64-simd.md (aarch64_shrn2<mode>_insn_le):
33135 (aarch64_shrn2<mode>_insn_be): Likewise.
33136 (aarch64_shrn2<mode>): Likewise.
33137 * config/aarch64/arm_neon.h (vshrn_high_n_s16): Reimlplement
33139 (vshrn_high_n_s32): Likewise.
33140 (vshrn_high_n_s64): Likewise.
33141 (vshrn_high_n_u16): Likewise.
33142 (vshrn_high_n_u32): Likewise.
33143 (vshrn_high_n_u64): Likewise.
33145 2021-01-28 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
33147 * config/aarch64/aarch64-simd-builtins.def (shrn): Define
33149 * config/aarch64/aarch64-simd.md (aarch64_shrn<mode>_insn_le):
33151 (aarch64_shrn<mode>_insn_be): Likewise.
33152 (aarch64_shrn<mode>): Likewise.
33153 * config/aarch64/arm_neon.h (vshrn_n_s16): Reimplement using
33155 (vshrn_n_s32): Likewise.
33156 (vshrn_n_s64): Likewise.
33157 (vshrn_n_u16): Likewise.
33158 (vshrn_n_u32): Likewise.
33159 (vshrn_n_u64): Likewise.
33160 * config/aarch64/iterators.md (vn_mode): New mode attribute.
33162 2021-01-28 Richard Biener <rguenther@suse.de>
33164 PR rtl-optimization/80960
33165 * dse.c (check_mem_read_rtx): Call get_addr on the
33168 2021-01-28 Xionghu Luo <luoxhu@linux.ibm.com>
33169 David Edelsohn <dje.gcc@gmail.com>
33172 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
33173 Don't generate VIEW_CONVERT_EXPR for fcode ALTIVEC_BUILTIN_VEC_INSERT
33175 * config/rs6000/rs6000-protos.h (rs6000_expand_vector_set_var):
33177 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Remove the
33178 wrapper call rs6000_expand_vector_set_var for cleanup. Call
33179 rs6000_expand_vector_set_var_p9 and rs6000_expand_vector_set_var_p8
33181 (rs6000_expand_vector_set_var): Delete.
33182 (rs6000_expand_vector_set_var_p9): Make static.
33183 (rs6000_expand_vector_set_var_p8): Make static.
33185 2021-01-28 Xing GUO <higuoxing@gmail.com>
33187 * common/config/riscv/riscv-common.c
33188 (riscv_subset_list::parsing_subset_version): Fix -march option parsing
33189 when `p` extension exists.
33191 2021-01-27 Vladimir N. Makarov <vmakarov@redhat.com>
33193 PR rtl-optimization/97684
33194 * ira.c (ira): Call ira_set_pseudo_classes before
33195 update_equiv_regs when it is necessary.
33197 2021-01-27 Jakub Jelinek <jakub@redhat.com>
33200 * config/aarch64/aarch64.md (*aarch64_bfxilsi_uxtw): Use
33201 %w0, %w1 and %2 instead of %0, %1 and %2.
33203 2021-01-27 Aaron Sawdey <acsawdey@linux.ibm.com>
33205 * config/rs6000/genfusion.pl: New script to generate
33206 define_insn_and_split patterns so combine can arrange fused
33207 instructions next to each other.
33208 * config/rs6000/fusion.md: New file, generated fused instruction
33209 patterns for combine.
33210 * config/rs6000/predicates.md (const_m1_to_1_operand): New predicate.
33211 (non_update_memory_operand): New predicate.
33212 * config/rs6000/rs6000-cpus.def: Add OPTION_MASK_P10_FUSION and
33213 OPTION_MASK_P10_FUSION_LD_CMPI to ISA_3_1_MASKS_SERVER and
33215 * config/rs6000/rs6000-protos.h (address_is_non_pfx_d_or_x): Add
33217 * config/rs6000/rs6000.c (rs6000_option_override_internal):
33218 Automatically set OPTION_MASK_P10_FUSION and
33219 OPTION_MASK_P10_FUSION_LD_CMPI if target is power10.
33220 (rs600_opt_masks): Allow -mpower10-fusion
33221 in function attributes.
33222 (address_is_non_pfx_d_or_x): New function.
33223 * config/rs6000/rs6000.h: Add MASK_P10_FUSION.
33224 * config/rs6000/rs6000.md: Include fusion.md.
33225 * config/rs6000/rs6000.opt: Add -mpower10-fusion
33226 and -mpower10-fusion-ld-cmpi.
33227 * config/rs6000/t-rs6000: Add dependencies involving fusion.md.
33229 2021-01-27 Jonathan Wright <jonathan.wright@arm.com>
33231 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlal
33232 builtin generator macros.
33233 * config/aarch64/aarch64-simd.md (*aarch64_<su>mlal<mode>):
33235 (aarch64_<su>mlal<mode>): This.
33236 * config/aarch64/arm_neon.h (vmlal_s8): Use RTL builtin
33237 instead of inline asm.
33238 (vmlal_s16): Likewise.
33239 (vmlal_s32): Likewise.
33240 (vmlal_u8): Likewise.
33241 (vmlal_u16): Likewise.
33242 (vmlal_u32): Likewise.
33244 2021-01-27 Richard Biener <rguenther@suse.de>
33246 PR tree-optimization/98854
33247 * tree-vect-slp.c (vect_build_slp_tree_2): Also build
33248 PHIs from scalars when the number of CTORs matches the
33249 number of children.
33251 2021-01-27 Jonathan Wright <jonathan.wright@arm.com>
33253 * config/aarch64/aarch64-simd-builtins.def: Add mls_n builtin
33255 * config/aarch64/aarch64-simd.md (*aarch64_mls_elt_merge<mode>):
33257 (aarch64_mls_n<mode>): This.
33258 * config/aarch64/arm_neon.h (vmls_n_s16): Use RTL builtin
33260 (vmls_n_s32): Likewise.
33261 (vmls_n_u16): Likewise.
33262 (vmls_n_u32): Likewise.
33263 (vmlsq_n_s16): Likewise.
33264 (vmlsq_n_s32): Likewise.
33265 (vmlsq_n_u16): Likewise.
33266 (vmlsq_n_u32): Likewise.
33268 2021-01-27 Jonathan Wright <jonathan.wright@arm.com>
33270 * config/aarch64/aarch64-simd-builtins.def: Add mls builtin
33272 * config/aarch64/arm_neon.h (vmls_s8): Use RTL builtin rather
33274 (vmls_s16): Likewise.
33275 (vmls_s32): Likewise.
33276 (vmls_u8): Likewise.
33277 (vmls_u16): Likewise.
33278 (vmls_u32): Likewise.
33279 (vmlsq_s8): Likewise.
33280 (vmlsq_s16): Likewise.
33281 (vmlsq_s32): Likewise.
33282 (vmlsq_u8): Likewise.
33283 (vmlsq_u16): Likewise.
33284 (vmlsq_u32): Likewise.
33286 2021-01-27 Jonathan Wright <jonathan.wright@arm.com>
33288 * config/aarch64/aarch64-simd-builtins.def: Add mla_n builtin
33290 * config/aarch64/aarch64-simd.md (*aarch64_mla_elt_merge<mode>):
33292 (aarch64_mla_n<mode>): This.
33293 * config/aarch64/arm_neon.h (vmla_n_s16): Use RTL builtin
33295 (vmla_n_s32): Likewise.
33296 (vmla_n_u16): Likewise.
33297 (vmla_n_u32): Likewise.
33298 (vmlaq_n_s16): Likewise.
33299 (vmlaq_n_s32): Likewise.
33300 (vmlaq_n_u16): Likewise.
33301 (vmlaq_n_u32): Likewise.
33303 2021-01-27 liuhongt <hongtao.liu@intel.com>
33306 * config/i386/sse.md (sse2_gt<mode>3): Drop !TARGET_XOP in condition.
33307 (*sse2_eq<mode>3): Ditto.
33309 2021-01-27 Jakub Jelinek <jakub@redhat.com>
33311 * tree-pass.h (PROP_trees): Rename to ...
33312 (PROP_gimple): ... this.
33313 * cfgexpand.c (pass_data_expand): Replace PROP_trees with PROP_gimple.
33314 * passes.c (execute_function_dump, execute_function_todo,
33315 execute_one_ipa_transform_pass, execute_one_pass): Likewise.
33316 * varpool.c (ctor_for_folding): Likewise.
33318 2021-01-27 Jakub Jelinek <jakub@redhat.com>
33320 PR tree-optimization/97260
33321 * varpool.c: Include tree-pass.h.
33322 (ctor_for_folding): In GENERIC return DECL_INITIAL for TREE_READONLY
33323 non-TREE_SIDE_EFFECTS automatic variables.
33325 2021-01-26 Paul Fee <paul.f.fee@gmail.com>
33327 * doc/cpp.texi (__cplusplus): Document value for -std=c++23
33329 * doc/invoke.texi: Document -std=c++23 and -std=gnu++23.
33330 * dwarf2out.c (highest_c_language): Recognise C++20 and C++23.
33331 (gen_compile_unit_die): Recognise C++23.
33333 2021-01-26 Jakub Jelinek <jakub@redhat.com>
33336 * dwarf2asm.c (dw2_assemble_integer): Cast DWARF2_ADDR_SIZE to int
33339 2021-01-26 Jakub Jelinek <jakub@redhat.com>
33342 * config/aarch64/aarch64.c (aarch64_mask_and_shift_for_ubfiz_p):
33343 Use UINTVAL (shft_amnt) and UINTVAL (mask) instead of INTVAL (shft_amnt)
33344 and INTVAL (mask). Add && INTVAL (mask) > 0 condition.
33346 2021-01-26 Richard Biener <rguenther@suse.de>
33348 * gimple-pretty-print.c (dump_binary_rhs): Handle
33349 VEC_WIDEN_{PLUS,MINUS}_{LO,HI}_EXPR.
33351 2021-01-26 Richard Biener <rguenther@suse.de>
33353 PR middle-end/98726
33354 * tree.h (vector_cst_int_elt): Remove.
33355 * tree.c (vector_cst_int_elt): Use poly_wide_int for computations,
33358 2021-01-26 Andrew Stubbs <ams@codesourcery.com>
33360 * config/gcn/gcn.c (gcn_expand_reduc_scalar): Use move instructions
33361 for V64DFmode min/max reductions.
33363 2021-01-26 Jakub Jelinek <jakub@redhat.com>
33365 * dwarf2asm.c (dw2_assemble_integer): Handle size twice as large
33366 as DWARF2_ADDR_SIZE if x is not a scalar int by emitting it as
33367 two halves, one with x and the other with const0_rtx, ordered
33368 depending on endianity.
33370 2021-01-26 Alexandre Oliva <oliva@adacore.com>
33372 * gimplify.c (gimplify_decl_expr): Skip asan marking calls for
33373 temporaries not seen in binding block, and not about to be
33374 added as gimple variables.
33376 2021-01-25 Martin Sebor <msebor@redhat.com>
33379 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Adjust warning text.
33381 2021-01-25 Martin Liska <mliska@suse.cz>
33383 * value-prof.c (get_nth_most_common_value): Use %s instead
33386 2021-01-25 Jakub Jelinek <jakub@redhat.com>
33389 * configure.ac (HAVE_AS_GDWARF_5_DEBUG_FLAG): Only define if
33390 readelf -wi is able to read the emitted .debug_info back.
33391 * configure: Regenerated.
33393 2021-01-25 Martin Liska <mliska@suse.cz>
33395 PR gcov-profile/98739
33396 * common.opt: Add missing sign symbol.
33397 * value-prof.c (get_nth_most_common_value): Restore handling
33398 of PROFILE_REPRODUCIBILITY_PARALLEL_RUNS and
33399 PROFILE_REPRODUCIBILITY_MULTITHREADED.
33401 2021-01-25 Richard Biener <rguenther@suse.de>
33403 PR middle-end/98807
33404 * tree.c (vector_element_bits): Always use precision of
33405 the element type for boolean vectors.
33407 2021-01-25 Sebastian Huber <sebastian.huber@embedded-brains.de>
33409 * config/rtems.h (STARTFILE_SPEC): Remove qnolinkcmds.
33410 (ENDFILE_SPEC): Evaluate qnolinkcmds.
33412 2021-01-25 Sebastian Huber <sebastian.huber@embedded-brains.de>
33414 * config/rtems.h (STARTFILE_SPEC): Remove nostdlib and
33415 nostartfiles handling since this is already done by
33416 LINK_COMMAND_SPEC. Evaluate qnolinkcmds.
33417 (ENDFILE_SPEC): Remove nostdlib and nostartfiles handling since this
33418 is already done by LINK_COMMAND_SPEC.
33419 (LIB_SPECS): Remove nostdlib and nodefaultlibs handling since
33420 this is already done by LINK_COMMAND_SPEC. Remove qnolinkcmds
33423 2021-01-25 Jakub Jelinek <jakub@redhat.com>
33426 * fold-const-call.c (host_size_t_cst_p): Renamed to ...
33427 (size_t_cst_p): ... this. Check and store unsigned HOST_WIDE_INT
33428 value rather than host size_t.
33429 (fold_const_call): Change type of s2 from size_t to
33430 unsigned HOST_WIDE_INT. Use size_t_cst_p instead of
33431 host_size_t_cst_p. For strncmp calls, pass MIN (s2, SIZE_MAX)
33432 instead of s2 as last argument.
33434 2021-01-25 Tamar Christina <tamar.christina@arm.com>
33436 * config/arm/iterators.md (rotsplit1, rotsplit2, conj_op, fcmac1,
33437 VCMLA_OP, VCMUL_OP): New.
33438 * config/arm/mve.md (mve_vcmlaq<mve_rot><mode>): Support vec_dup 0.
33439 * config/arm/neon.md (cmul<conj_op><mode>3): New.
33440 * config/arm/unspecs.md (UNSPEC_VCMLA_CONJ, UNSPEC_VCMLA180_CONJ,
33441 UNSPEC_VCMUL_CONJ): New.
33442 * config/arm/vec-common.md (cmul<conj_op><mode>3, arm_vcmla<rot><mode>,
33443 cml<fcmac1><conj_op><mode>4): New.
33445 2021-01-23 Jakub Jelinek <jakub@redhat.com>
33448 * config/rs6000/mmintrin.h (__m64): Add __may_alias__ attribute.
33450 2021-01-22 Jonathan Wright <jonathan.wright@arm.com>
33452 * config/aarch64/aarch64-simd-builtins.def: Add mla builtin
33454 * config/aarch64/arm_neon.h (vmla_s8): Use RTL builtin rather
33456 (vmla_s16): Likewise.
33457 (vmla_s32): Likewise.
33458 (vmla_u8): Likewise.
33459 (vmla_u16): Likewise.
33460 (vmla_u32): Likewise.
33461 (vmlaq_s8): Likewise.
33462 (vmlaq_s16): Likewise.
33463 (vmlaq_s32): Likewise.
33464 (vmlaq_u8): Likewise.
33465 (vmlaq_u16): Likewise.
33466 (vmlaq_u32): Likewise.
33468 2021-01-22 David Malcolm <dmalcolm@redhat.com>
33470 * doc/invoke.texi (GCC_EXTRA_DIAGNOSTIC_OUTPUT): Add @findex
33473 2021-01-22 Jakub Jelinek <jakub@redhat.com>
33476 * dwarf2out.c (output_file_names): For -gdwarf-5, if there are no
33477 filenames to emit, still emit the required 0 index directory and
33478 filename entries that match DW_AT_comp_dir and DW_AT_name of the
33481 2021-01-22 Marek Polacek <polacek@redhat.com>
33484 * doc/invoke.texi: Update C++ ABI Version 15 description.
33486 2021-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
33488 PR tree-optimization/98766
33489 * tree-ssa-math-opts.c (convert_mult_to_fma): Use maybe_le when
33490 comparing against type size with param_avoid_fma_max_bits.
33492 2021-01-22 Richard Biener <rguenther@suse.de>
33494 PR middle-end/98793
33495 * tree.c (vector_element_bits): Key single-bit bool vector on
33496 integer mode rather than not vector mode.
33498 2021-01-22 Xionghu Luo <luoxhu@linux.ibm.com>
33501 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
33502 Generate ARRAY_REF(VIEW_CONVERT_EXPR) for P8 and later
33504 * config/rs6000/rs6000.c (rs6000_expand_vector_set_var): Update
33505 to call different path for P8 and P9.
33506 (rs6000_expand_vector_set_var_p9): New function.
33507 (rs6000_expand_vector_set_var_p8): New function.
33509 2021-01-22 Xionghu Luo <luoxhu@linux.ibm.com>
33513 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
33514 Ajdust variable index vec_insert from address dereference to
33515 ARRAY_REF(VIEW_CONVERT_EXPR) tree expression.
33516 * config/rs6000/rs6000-protos.h (rs6000_expand_vector_set_var):
33518 * config/rs6000/rs6000.c (rs6000_expand_vector_set_var): New function.
33520 2021-01-22 Martin Liska <mliska@suse.cz>
33522 PR gcov-profile/98739
33523 * profile.c (compute_value_histograms): Drop time profile for
33524 -fprofile-reproducible=multithreaded.
33526 2021-01-22 Nathan Sidwell <nathan@acm.org>
33528 * gcc.c (process_command): Don't check OPT_SPECIAL_input_file
33531 2021-01-22 Richard Biener <rguenther@suse.de>
33533 PR middle-end/98773
33534 * tree-data-ref.c (initalize_matrix_A): Revert previous
33535 change, retaining failing on HOST_WIDE_INT_MIN CHREC_RIGHT.
33537 2021-01-22 Jakub Jelinek <jakub@redhat.com>
33539 PR tree-optimization/90248
33540 * match.pd (X cmp 0.0 ? 1.0 : -1.0 -> copysign(1, +-X),
33541 X cmp 0.0 ? -1.0 : +1.0 -> copysign(1, -+X)): Remove
33543 (X * (X cmp 0.0 ? 1.0 : -1.0) -> +-abs(X),
33544 X * (X cmp 0.0 ? -1.0 : 1.0) -> +-abs(X)): New simplifications.
33546 2021-01-22 Jakub Jelinek <jakub@redhat.com>
33548 PR tree-optimization/98255
33549 * tree-dfa.c (get_ref_base_and_extent): For ARRAY_REFs, sign
33550 extend index - low_bound from sizetype's precision rather than index
33552 (get_addr_base_and_unit_offset_1): Likewise.
33553 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Likewise.
33554 * gimple-fold.c (fold_const_aggregate_ref_1): Likewise.
33556 2021-01-22 Richard Biener <rguenther@suse.de>
33558 PR tree-optimization/98786
33559 * tree-ssa-phiopt.c (factor_out_conditional_conversion): Avoid
33560 adding new uses of abnormals. Verify we deal with a conditional
33563 2021-01-22 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
33566 * optc-save-gen.awk: Add arm_fp16_format to checked_options.
33568 2021-01-22 liuhongt <hongtao.liu@intel.com>
33572 * config/i386/sse.md (VI_128_256): New mode iterator.
33573 (*avx_cmp<mode>3_1, *avx_cmp<mode>3_2, *avx_cmp<mode>3_3,
33574 *avx_cmp<mode>3_4, *avx2_eq<mode>3, *avx2_pcmp<mode>3_1,
33575 *avx2_pcmp<mode>3_2, *avx2_gt<mode>3): New
33576 define_insn_and_split to lower avx512 vector comparison to avx
33577 version when dest is vector.
33578 (*<avx512>_cmp<mode>3,*<avx512>_cmp<mode>3,*<avx512>_ucmp<mode>3):
33579 define_insn_and_split for negating the comparison result.
33580 * config/i386/predicates.md (float_vector_all_ones_operand):
33582 * config/i386/i386-expand.c (ix86_expand_sse_movcc): Use
33583 general NOT operator without UNSPEC_MASKOP.
33585 2021-01-21 Vladimir N. Makarov <vmakarov@redhat.com>
33587 PR rtl-optimization/98777
33588 * lra-int.h (lra_pmode_pseudo): New extern.
33589 * lra.c (lra_pmode_pseudo): New global.
33591 * lra-eliminations.c (eliminate_regs_in_insn): Use it.
33593 2021-01-21 Ilya Leoshkevich <iii@linux.ibm.com>
33595 * fwprop.c (fwprop_propagation::classify_result): Allow
33596 (subreg (mem)) simplifications.
33598 2021-01-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
33600 * config/aarch64/aarch64-simd.md (aarch64_sqdml<SBINQOPS:as>l<mode>):
33602 (aarch64_sqdmlal<mode>): ... This...
33603 (aarch64_sqdmlsl<mode>): ... And this.
33604 (aarch64_sqdml<SBINQOPS:as>l_lane<mode>): Split into...
33605 (aarch64_sqdmlal_lane<mode>): ... This...
33606 (aarch64_sqdmlsl_lane<mode>): ... And this.
33607 (aarch64_sqdml<SBINQOPS:as>l_laneq<mode>): Split into...
33608 (aarch64_sqdmlsl_laneq<mode>): ... This...
33609 (aarch64_sqdmlal_laneq<mode>): ... And this.
33610 (aarch64_sqdml<SBINQOPS:as>l_n<mode>): Split into...
33611 (aarch64_sqdmlsl_n<mode>): ... This...
33612 (aarch64_sqdmlal_n<mode>): ... And this.
33613 (aarch64_sqdml<SBINQOPS:as>l2<mode>_internal): Split into...
33614 (aarch64_sqdmlal2<mode>_internal): ... This...
33615 (aarch64_sqdmlsl2<mode>_internal): ... And this.
33617 2021-01-21 Christophe Lyon <christophe.lyon@linaro.org>
33619 * config/arm/arm_mve.h (__arm_vcmpneq_s8): Fix return type.
33621 2021-01-21 Andrea Corallo <andrea.corallo@arm.com>
33624 * doc/sourcebuild.texi (arm_thumb2_no_arm_v8_1_lob): Document.
33626 2021-01-21 liuhongt <hongtao.liu@intel.com>
33628 PR rtl-optimization/98694
33629 * regcprop.c (copy_value): If SRC had been assigned a mode
33630 narrower than the copy, we can't link DEST into the chain even
33631 they have same hard_regno_nregs(i.e. HImode/SImode in i386
33634 2021-01-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
33636 * config/aarch64/aarch64-simd.md (aarch64_get_lane<mode>):
33637 Convert to define_insn_and_split. Split into simple move when moving
33640 2021-01-20 Segher Boessenkool <segher@kernel.crashing.org>
33642 * config/rs6000/rs6000.c (rs6000_emit_le_vsx_store): Change assert.
33643 Adjust comment. Simplify code.
33645 2021-01-20 Jakub Jelinek <jakub@redhat.com>
33648 * dwarf2out.c (reset_indirect_string): Also reset indirect strings
33649 with DW_FORM_line_strp form.
33650 (prune_unused_types_update_strings): Don't add into debug_str_hash
33651 indirect strings with DW_FORM_line_strp form.
33652 (adjust_name_comp_dir): New function.
33653 (dwarf2out_finish): Call it on CU DIEs after resetting
33654 debug_line_str_hash.
33656 2021-01-20 Vladimir N. Makarov <vmakarov@redhat.com>
33658 PR rtl-optimization/98722
33659 * lra-eliminations.c (eliminate_regs_in_insn): Check that target
33660 has no 3-op add insn to transform insns containing two pluses.
33662 2021-01-20 Richard Biener <rguenther@suse.de>
33664 * hwint.h (add_hwi): New function.
33665 (mul_hwi): Likewise.
33666 * tree-data-ref.c (initialize_matrix_A): Properly translate
33667 tree constants and avoid HOST_WIDE_INT_MIN.
33668 (lambda_matrix_row_add): Avoid undefined integer overflow
33669 and return true on such overflow.
33670 (lambda_matrix_right_hermite): Handle overflow from
33671 lambda_matrix_row_add gracefully. Simplify previous fix.
33672 (analyze_subscript_affine_affine): Likewise.
33674 2021-01-20 Eugene Rozenfeld <erozen@microsoft.com>
33676 PR tree-optimization/96674
33677 * match.pd: New patterns: x < y || y == XXX_MIN --> x <= y - 1
33678 x >= y && y != XXX_MIN --> x > y - 1
33680 2021-01-20 Richard Sandiford <richard.sandiford@arm.com>
33682 PR tree-optimization/98535
33683 * tree-vect-slp.c (duplicate_and_interleave): Use quick_grow_cleared.
33684 If the high and low permutes are the same, remove the high permutes
33685 from the working set and only continue with the low ones.
33687 2021-01-20 Jakub Jelinek <jakub@redhat.com>
33689 PR tree-optimization/98721
33690 * builtins.c (access_ref::inform_access): Don't assume
33691 SSA_NAME_IDENTIFIER must be non-NULL. Print messages about
33692 object whenever allocfn is NULL, rather than only when DECL_P
33693 is true. Use %qE instead of %qD for that. Formatting fixes.
33695 2021-01-20 Richard Biener <rguenther@suse.de>
33697 PR tree-optimization/98758
33698 * tree-data-ref.c (int_divides_p): Use lambda_int arguments.
33699 (lambda_matrix_right_hermite): Avoid undefinedness with
33700 signed integer abs and multiplication.
33701 (analyze_subscript_affine_affine): Use lambda_int.
33703 2021-01-20 David Malcolm <dmalcolm@redhat.com>
33706 * dwarf2out.c (output_line_info): Rename static variable
33707 "generation", moving it out of the function to...
33708 (output_line_info_generation): New.
33709 (init_sections_and_labels): Likewise, renaming the variable to...
33710 (init_sections_and_labels_generation): New.
33711 (dwarf2out_c_finalize): Reset the new variables.
33713 2021-01-19 Martin Sebor <msebor@redhat.com>
33715 PR middle-end/98664
33716 * tree-ssa-live.c (remove_unused_scope_block_p): Keep scopes for
33717 all functions, even if they're not declared artificial or inline.
33718 * tree.c (tree_inlined_location): Use macro expansion location
33719 only if scope traversal fails to expose one.
33721 2021-01-19 Richard Sandiford <richard.sandiford@arm.com>
33723 PR rtl-optimization/92294
33724 * alias.c (compare_base_symbol_refs): Take an extra parameter
33725 and add the distance between two symbols to it. Enshrine in
33726 comments that -1 means "either 0 or 1, but we can't tell
33727 which at compile time".
33728 (memrefs_conflict_p): Update call accordingly.
33729 (rtx_equal_for_memref_p): Likewise. Take the distance between symbols
33732 2021-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
33734 * config/aarch64/aarch64-simd-builtins.def (sqshl, uqshl,
33735 sqrshl, uqrshl, sqadd, uqadd, sqsub, uqsub, suqadd, usqadd, sqmovn,
33736 uqmovn, sqxtn2, uqxtn2, sqabs, sqneg, sqdmlal, sqdmlsl, sqdmlal_lane,
33737 sqdmlsl_lane, sqdmlal_laneq, sqdmlsl_laneq, sqdmlal_n, sqdmlsl_n,
33738 sqdmlal2, sqdmlsl2, sqdmlal2_lane, sqdmlsl2_lane, sqdmlal2_laneq,
33739 sqdmlsl2_laneq, sqdmlal2_n, sqdmlsl2_n, sqdmull, sqdmull_lane,
33740 sqdmull_laneq, sqdmull_n, sqdmull2, sqdmull2_lane, sqdmull2_laneq,
33741 sqdmull2_n, sqdmulh, sqrdmulh, sqdmulh_lane, sqdmulh_laneq,
33742 sqrdmulh_lane, sqrdmulh_laneq, sqshrun_n, sqrshrun_n, sqshrn_n,
33743 uqshrn_n, sqrshrn_n, uqrshrn_n, sqshlu_n, sqshl_n, uqshl_n, sqrdmlah,
33744 sqrdmlsh, sqrdmlah_lane, sqrdmlsh_lane, sqrdmlah_laneq, sqrdmlsh_laneq,
33745 sqmovun): Use NONE flags.
33747 2021-01-19 Richard Biener <rguenther@suse.de>
33750 * ipa-modref.c (analyze_stmt): Only record a summary for a
33753 2021-01-19 Richard Biener <rguenther@suse.de>
33755 PR middle-end/98638
33756 * tree-ssanames.c (fini_ssanames): Zero SSA_NAME_DEF_STMT.
33758 2021-01-19 Daniel Hellstrom <daniel@gaisler.com>
33760 * config/sparc/rtemself.h (TARGET_OS_CPP_BUILTINS): Add
33761 built-in define __FIX_LEON3FT_TN0018.
33763 2021-01-19 Richard Biener <rguenther@suse.de>
33766 * tree-inline.c (tree_function_versioning): Set input_location
33767 to UNKNOWN_LOCATION throughout the function.
33769 2021-01-19 Tobias Burnus <tobias@codesourcery.com>
33772 * omp-low.c (lower_omp_target): Handle nonpointer is_device_ptr.
33774 2021-01-19 Martin Jambor <mjambor@suse.cz>
33777 * ipa-sra.c (ssa_name_only_returned_p): New parameter fun. Check
33778 whether non-call exceptions allow removal of a statement.
33779 (isra_analyze_call): Pass the appropriate function to
33780 ssa_name_only_returned_p.
33782 2021-01-19 Geng Qi <gengqi@linux.alibaba.com>
33784 * config/riscv/arch-canonicalize (longext_sort): New function for
33785 sorting 'multi-letter'.
33786 * config/riscv/multilib-generator: Adjusting the loop of 'alt' in
33787 'alts'. The 'arch' may not be the first of 'alts'.
33788 (_expand_combination): Add underline for the 'ext' without '*'.
33789 This is because, a single-letter extension can always be treated well
33790 with a '_' prefix, but it cannot be separated out if it is appended
33793 2021-01-18 Vladimir N. Makarov <vmakarov@redhat.com>
33796 * ira.c (ira): Skip abnormal critical edge splitting.
33798 2021-01-18 Jakub Jelinek <jakub@redhat.com>
33800 PR tree-optimization/98727
33801 * tree-ssa-math-opts.c (match_arith_overflow): Fix up computation of
33802 second .MUL_OVERFLOW operand for signed multiplication with overflow
33803 checking if the second operand of multiplication is not constant.
33805 2021-01-18 David Edelsohn <dje.gcc@gmail.com>
33807 * doc/invoke.texi (-gdwarf): TPF defaults to version 2 and AIX
33808 defaults to version 4.
33810 2021-01-18 David Malcolm <dmalcolm@redhat.com>
33812 * attribs.h (fndecl_dealloc_argno): New decl.
33813 * builtins.c (call_dealloc_argno): Split out second half of
33815 (fndecl_dealloc_argno): New.
33816 * doc/extend.texi (Common Function Attributes): Document the
33817 interaction between the analyzer and the malloc attribute.
33818 * doc/invoke.texi (Static Analyzer Options): Likewise.
33820 2021-01-17 David Edelsohn <dje.gcc@gmail.com>
33822 * config/rs6000/aix71.h (SUBTARGET_OVERRIDE_OPTIONS): Override
33823 dwarf_version to 4.
33824 * config/rs6000/aix72.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
33826 2021-01-17 Martin Jambor <mjambor@suse.cz>
33829 * cgraph.c (clone_of_p): Check also former_clone_of as we climb
33832 2021-01-17 Mark Wielaard <mark@klomp.org>
33834 * common.opt (gdwarf-): Init(5).
33835 * doc/invoke.texi (-gdwarf): Document default to 5.
33837 2021-01-16 Kwok Cheung Yeung <kcy@codesourcery.com>
33839 * builtin-types.def
33840 (BT_FN_VOID_OMPFN_PTR_OMPCPYFN_LONG_LONG_BOOL_UINT_PTR_INT): Rename
33842 (BT_FN_VOID_OMPFN_PTR_OMPCPYFN_LONG_LONG_BOOL_UINT_PTR_INT_PTR):
33843 ...this. Add extra argument.
33844 * gimplify.c (omp_default_clause): Ensure that event handle is
33845 firstprivate in a task region.
33846 (gimplify_scan_omp_clauses): Handle OMP_CLAUSE_DETACH.
33847 (gimplify_adjust_omp_clauses): Likewise.
33848 * omp-builtins.def (BUILT_IN_GOMP_TASK): Change function type to
33849 BT_FN_VOID_OMPFN_PTR_OMPCPYFN_LONG_LONG_BOOL_UINT_PTR_INT_PTR.
33850 * omp-expand.c (expand_task_call): Add GOMP_TASK_FLAG_DETACH to flags
33851 if detach clause specified. Add detach argument when generating
33853 * omp-low.c (scan_sharing_clauses): Setup data environment for detach
33855 (finish_taskreg_scan): Move field for variable containing the event
33856 handle to the front of the struct.
33857 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_DETACH. Fix
33859 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
33860 OMP_CLAUSE_DETACH clause.
33861 (convert_local_omp_clauses): Handle OMP_CLAUSE_DETACH clause.
33862 * tree-pretty-print.c (dump_omp_clause): Handle OMP_CLAUSE_DETACH.
33863 * tree.c (omp_clause_num_ops): Add entry for OMP_CLAUSE_DETACH.
33865 (omp_clause_code_name): Add entry for OMP_CLAUSE_DETACH. Fix
33867 (walk_tree_1): Handle OMP_CLAUSE_DETACH.
33869 2021-01-16 Sebastian Huber <sebastian.huber@embedded-brains.de>
33871 * config/nios2/t-rtems: Reset all MULTILIB_* variables. Shorten
33872 multilib directory names. Use MULTILIB_REQUIRED instead of
33873 MULTILIB_EXCEPTIONS. Add -mhw-mul -mhw-mulx -mhw-div
33874 -mcustom-fpu-cfg=fph2 multilib.
33876 2021-01-16 Sebastian Huber <sebastian.huber@embedded-brains.de>
33878 * config/nios2/nios2.c (NIOS2_FPU_CONFIG_NUM): Adjust value.
33879 (nios2_init_fpu_configs): Provide register values for new
33880 -mcustom-fpu-cfg=fph2 option variant.
33881 * doc/invoke.texi (-mcustom-fpu-cfg=fph2): Document new option
33884 2021-01-16 Sebastian Huber <sebastian.huber@embedded-brains.de>
33886 * config/nios2/nios2.c (nios2_custom_check_insns): Remove
33887 custom instruction warnings.
33889 2021-01-16 Jakub Jelinek <jakub@redhat.com>
33891 PR tree-optimization/96669
33892 * match.pd ((CST << x) & 1 -> x == 0): New simplification.
33894 2021-01-16 Jakub Jelinek <jakub@redhat.com>
33896 PR tree-optimization/96271
33897 * passes.def: Pass false argument to first two pass_cd_dce
33898 instances and true to last instance. Add comment that
33899 last instance rewrites no longer addressed locals.
33900 * tree-ssa-dce.c (pass_cd_dce): Add update_address_taken_p member and
33902 (pass_cd_dce::set_pass_param): New method.
33903 (pass_cd_dce::execute): Return TODO_update_address_taken from
33904 last cd_dce instance.
33906 2021-01-15 Carl Love <cel@us.ibm.com>
33908 * config/rs6000/altivec.h (vec_mulh, vec_div, vec_dive, vec_mod):
33910 * config/rs6000/altivec.md (VIlong): Move define to file vsx.md.
33911 * config/rs6000/rs6000-builtin.def (DIVES_V4SI, DIVES_V2DI,
33912 DIVEU_V4SI, DIVEU_V2DI, DIVS_V4SI, DIVS_V2DI, DIVU_V4SI,
33913 DIVU_V2DI, MODS_V2DI, MODS_V4SI, MODU_V2DI, MODU_V4SI,
33914 MULHS_V2DI, MULHS_V4SI, MULHU_V2DI, MULHU_V4SI, MULLD_V2DI):
33915 Add builtin define.
33916 (MULH, DIVE, MOD): Add new BU_P10_OVERLOAD_2 definitions.
33917 * config/rs6000/rs6000-call.c (VSX_BUILTIN_VEC_DIV,
33918 VSX_BUILTIN_VEC_DIVE, P10_BUILTIN_VEC_MOD, P10_BUILTIN_VEC_MULH):
33919 New overloaded definitions.
33920 (builtin_function_type) [P10V_BUILTIN_DIVEU_V4SI,
33921 P10V_BUILTIN_DIVEU_V2DI, P10V_BUILTIN_DIVU_V4SI,
33922 P10V_BUILTIN_DIVU_V2DI, P10V_BUILTIN_MODU_V2DI,
33923 P10V_BUILTIN_MODU_V4SI, P10V_BUILTIN_MULHU_V2DI,
33924 P10V_BUILTIN_MULHU_V4SI]: Add case
33925 statement for builtins.
33926 * config/rs6000/rs6000.md (bits): Add new attribute sizes V4SI, V2DI.
33927 * config/rs6000/vsx.md (VIlong): Moved from config/rs6000/altivec.md.
33928 (UNSPEC_VDIVES, UNSPEC_VDIVEU): New unspec definitions.
33929 (vsx_mul_v2di): Add if TARGET_POWER10 statement.
33930 (vsx_udiv_v2di): Add if TARGET_POWER10 statement.
33931 (dives_<mode>, diveu_<mode>, div<mode>3, uvdiv<mode>3,
33932 mods_<mode>, modu_<mode>, mulhs_<mode>, mulhu_<mode>, mulv2di3):
33933 Add define_insn, mode is VIlong.
33934 * doc/extend.texi (vec_mulh, vec_mul, vec_div, vec_dive, vec_mod):
33935 Add builtin descriptions.
33937 2021-01-15 Eric Botcazou <ebotcazou@adacore.com>
33939 * final.c (final_start_function_1): Reset force_source_line.
33941 2021-01-15 Jakub Jelinek <jakub@redhat.com>
33943 PR tree-optimization/96669
33944 * match.pd (((1 << A) & 1) != 0 -> A == 0,
33945 ((1 << A) & 1) == 0 -> A != 0): Generalize for 1s replaced by
33946 possibly different power of two constants and to right shift too.
33948 2021-01-15 Jakub Jelinek <jakub@redhat.com>
33950 PR tree-optimization/96681
33951 * match.pd ((x < 0) ^ (y < 0) to (x ^ y) < 0): New simplification.
33952 ((x >= 0) ^ (y >= 0) to (x ^ y) < 0): Likewise.
33953 ((x < 0) ^ (y >= 0) to (x ^ y) >= 0): Likewise.
33954 ((x >= 0) ^ (y < 0) to (x ^ y) >= 0): Likewise.
33956 2021-01-15 Alexandre Oliva <oliva@adacore.com>
33958 * opts.c (gen_command_line_string): Exclude -dumpbase-ext.
33960 2021-01-15 Tamar Christina <tamar.christina@arm.com>
33962 * config/aarch64/aarch64-simd.md (cml<fcmac1><conj_op><mode>4,
33963 cmul<conj_op><mode>3): New.
33964 * config/aarch64/iterators.md (UNSPEC_FCMUL,
33965 UNSPEC_FCMUL180, UNSPEC_FCMLA_CONJ, UNSPEC_FCMLA180_CONJ,
33966 UNSPEC_CMLA_CONJ, UNSPEC_CMLA180_CONJ, UNSPEC_CMUL, UNSPEC_CMUL180,
33967 FCMLA_OP, FCMUL_OP, conj_op, rotsplit1, rotsplit2, fcmac1, sve_rot1,
33968 sve_rot2, SVE2_INT_CMLA_OP, SVE2_INT_CMUL_OP, SVE2_INT_CADD_OP): New.
33969 (rot): Add UNSPEC_FCMUL, UNSPEC_FCMUL180.
33970 (rot_op): Renamed to conj_op.
33971 * config/aarch64/aarch64-sve.md (cml<fcmac1><conj_op><mode>4,
33972 cmul<conj_op><mode>3): New.
33973 * config/aarch64/aarch64-sve2.md (cml<fcmac1><conj_op><mode>4,
33974 cmul<conj_op><mode>3): New.
33976 2021-01-15 David Malcolm <dmalcolm@redhat.com>
33980 (selftest::test_print_parseable_fixits_bytes_vs_display_columns):
33981 Escape the tempfile name when constructing the expected output.
33983 2021-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
33985 * config/aarch64/aarch64-simd.md (*aarch64_<su>mlsl_hi<mode>):
33987 (aarch64_<su>mlsl_hi<mode>): ... This.
33988 (aarch64_<su>mlsl_hi<mode>): Define.
33989 (*aarch64_<su>mlsl<mode): Rename to...
33990 (aarch64_<su>mlsl<mode): ... This.
33991 * config/aarch64/aarch64-simd-builtins.def (smlsl, umlsl,
33992 smlsl_hi, umlsl_hi): Define builtins.
33993 * config/aarch64/arm_neon.h (vmlsl_high_s8, vmlsl_high_s16,
33994 vmlsl_high_s32, vmlsl_high_u8, vmlsl_high_u16, vmlsl_high_u32,
33995 vmlsl_s8, vmlsl_s16, vmlsl_s32, vmlsl_u8,
33996 vmlsl_u16, vmlsl_u32): Reimplement with builtins.
33998 2021-01-15 Uroš Bizjak <ubizjak@gmail.com>
34000 * config/i386/i386-c.c (ix86_target_macros):
34001 Use cpp_define_formatted for __SIZEOF_FLOAT80__ definition.
34003 2021-01-15 Richard Sandiford <richard.sandiford@arm.com>
34006 * config.gcc (aarch64*-*-*): Add aarch64-cc-fusion.o to extra_objs.
34007 * Makefile.in (RTL_SSA_H): New variable.
34008 * config/aarch64/t-aarch64 (aarch64-cc-fusion.o): New rule.
34009 * config/aarch64/aarch64-protos.h (make_pass_cc_fusion): Declare.
34010 * config/aarch64/aarch64-passes.def: Add pass_cc_fusion after
34012 * config/aarch64/aarch64-cc-fusion.cc: New file.
34014 2021-01-15 Richard Sandiford <richard.sandiford@arm.com>
34016 * recog.h (insn_change_watermark::~insn_change_watermark): Avoid
34017 calling cancel_changes for changes that no longer exist.
34019 2021-01-15 Richard Sandiford <richard.sandiford@arm.com>
34021 * rtl-ssa/functions.h (function_info::ref_defs): Rename to...
34022 (function_info::reg_defs): ...this.
34023 * rtl-ssa/member-fns.inl (function_info::ref_defs): Rename to...
34024 (function_info::reg_defs): ...this.
34026 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
34029 * config/arm/arm_neon.h (vceqz_p64, vceqq_p64, vceqzq_p64): New.
34031 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
34034 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
34037 * config/arm/arm_neon.h (vceqz_p64, vceqq_p64, vceqzq_p64): New.
34039 2021-01-15 Richard Biener <rguenther@suse.de>
34041 PR tree-optimization/96376
34042 * tree-vect-stmts.c (get_load_store_type): Disregard alignment
34043 for VMAT_INVARIANT.
34045 2021-01-15 Martin Liska <mliska@suse.cz>
34047 * doc/install.texi: Document that some tests need pytest module.
34048 * doc/sourcebuild.texi: Likewise.
34050 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
34053 * config/arm/arm_neon.h (vceqz_p64, vceqq_p64, vceqzq_p64): New.
34055 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
34057 * config/arm/mve.md (mve_vshrq_n_s<mode>_imm): New entry.
34058 (mve_vshrq_n_u<mode>_imm): Likewise.
34059 * config/arm/neon.md (vashr<mode>3, vlshr<mode>3): Move to ...
34060 * config/arm/vec-common.md: ... here.
34062 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
34064 * config/arm/mve.md (mve_vshlq_<supf><mode>): Move to
34066 * config/arm/neon.md (vashl<mode>3): Delete.
34067 * config/arm/vec-common.md (mve_vshlq_<supf><mode>): New.
34068 (vasl<mode>3): New expander.
34070 2021-01-15 Richard Biener <rguenther@suse.de>
34072 PR tree-optimization/98685
34073 * tree-vect-slp.c (vect_schedule_slp_node): Refactor handling
34074 of vector extern defs.
34076 2021-01-14 David Malcolm <dmalcolm@redhat.com>
34079 * diagnostic.c (diagnostic_kind_text): Break out this array
34081 (diagnostic_build_prefix): ...here.
34082 (fancy_abort): Detect when diagnostic_initialize has not yet been
34083 called and fall back to a minimal implementation of printing the
34084 ICE, rather than segfaulting in internal_error.
34086 2021-01-14 David Malcolm <dmalcolm@redhat.com>
34088 * diagnostic.c (diagnostic_initialize): Eliminate
34089 parseable_fixits_p in favor of initializing extra_output_kind from
34090 GCC_EXTRA_DIAGNOSTIC_OUTPUT.
34091 (convert_column_unit): New function, split out from...
34092 (diagnostic_converted_column): ...this.
34093 (print_parseable_fixits): Add "column_unit" and "tabstop" params.
34094 Use them to call convert_column_unit on the column values.
34095 (diagnostic_report_diagnostic): Eliminate conditional on
34096 parseable_fixits_p in favor of a switch statement on
34097 extra_output_kind, passing the appropriate values to the new
34098 params of print_parseable_fixits.
34099 (selftest::test_print_parseable_fixits_none): Update for new
34100 params of print_parseable_fixits.
34101 (selftest::test_print_parseable_fixits_insert): Likewise.
34102 (selftest::test_print_parseable_fixits_remove): Likewise.
34103 (selftest::test_print_parseable_fixits_replace): Likewise.
34104 (selftest::test_print_parseable_fixits_bytes_vs_display_columns):
34106 (selftest::diagnostic_c_tests): Call it.
34107 * diagnostic.h (enum diagnostics_extra_output_kind): New.
34108 (diagnostic_context::parseable_fixits_p): Delete field in favor
34110 (diagnostic_context::extra_output_kind): ...this new field.
34111 * doc/invoke.texi (Environment Variables): Add
34112 GCC_EXTRA_DIAGNOSTIC_OUTPUT.
34113 * opts.c (common_handle_option): Update handling of
34114 OPT_fdiagnostics_parseable_fixits for change to diagnostic_context
34117 2021-01-14 Tamar Christina <tamar.christina@arm.com>
34119 * tree-vect-slp-patterns.c (class complex_operations_pattern,
34120 complex_operations_pattern::matches,
34121 complex_operations_pattern::recognize,
34122 complex_operations_pattern::build): New.
34123 (slp_patterns): Use it.
34125 2021-01-14 Tamar Christina <tamar.christina@arm.com>
34127 * internal-fn.def (COMPLEX_FMS, COMPLEX_FMS_CONJ): New.
34128 * optabs.def (cmls_optab, cmls_conj_optab): New.
34129 * doc/md.texi: Document them.
34130 * tree-vect-slp-patterns.c (class complex_fms_pattern,
34131 complex_fms_pattern::matches, complex_fms_pattern::recognize,
34132 complex_fms_pattern::build): New.
34134 2021-01-14 Tamar Christina <tamar.christina@arm.com>
34136 * internal-fn.def (COMPLEX_FMA, COMPLEX_FMA_CONJ): New.
34137 * optabs.def (cmla_optab, cmla_conj_optab): New.
34138 * doc/md.texi: Document them.
34139 * tree-vect-slp-patterns.c (vect_match_call_p,
34140 class complex_fma_pattern, vect_slp_reset_pattern,
34141 complex_fma_pattern::matches, complex_fma_pattern::recognize,
34142 complex_fma_pattern::build): New.
34144 2021-01-14 Tamar Christina <tamar.christina@arm.com>
34146 * internal-fn.def (COMPLEX_MUL, COMPLEX_MUL_CONJ): New.
34147 * optabs.def (cmul_optab, cmul_conj_optab): New.
34148 * doc/md.texi: Document them.
34149 * tree-vect-slp-patterns.c (vect_match_call_complex_mla,
34150 vect_normalize_conj_loc, is_eq_or_top, vect_validate_multiplication,
34151 vect_build_combine_node, class complex_mul_pattern,
34152 complex_mul_pattern::matches, complex_mul_pattern::recognize,
34153 complex_mul_pattern::build): New.
34155 2021-01-14 Tamar Christina <tamar.christina@arm.com>
34157 * tree-vect-slp.c (optimize_load_redistribution_1): New.
34158 (optimize_load_redistribution, vect_is_slp_load_node): New.
34159 (vect_match_slp_patterns): Use it.
34161 2021-01-14 Tamar Christina <tamar.christina@arm.com>
34163 * tree-vect-slp-patterns.c (complex_add_pattern::build):
34166 2021-01-14 Thomas Schwinge <thomas@codesourcery.com>
34168 * config/gcn/mkoffload.c (main): Create an offload image only in
34169 64-bit configurations.
34171 2021-01-14 H.J. Lu <hjl.tools@gmail.com>
34174 * config/i386/i386-options.c (ix86_option_override_internal):
34175 Issue an error for -fcf-protection with CF_BRANCH when compiling
34176 for 32-bit non-TARGET_CMOV targets.
34178 2021-01-14 Uroš Bizjak <ubizjak@gmail.com>
34181 * config/i386/i386-options.c (ix86_valid_target_attribute_inner_p):
34182 Remove declaration and initialization of shadow variable "ret".
34183 (ix86_option_override_internal): Remove delcaration of
34184 shadow variable "i". Redeclare shadowed variable to unsigned.
34185 * common/config/i386/i386-common.c (pta_size): Redeclare to unsigned.
34186 * config/i386/i386-builtins.c (get_builtin_code_for_version):
34187 Update for redeclaration.
34188 * config/i386/i386.h (pta_size): Ditto.
34190 2021-01-14 Richard Biener <rguenther@suse.de>
34192 PR tree-optimization/98674
34193 * tree-data-ref.c (base_supports_access_fn_components_p): New.
34194 (initialize_data_dependence_relation): For two bases without
34195 possible access fns resort to type size equality when determining
34196 shape compatibility.
34198 2021-01-14 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
34201 * config/arm/arm_neon.h: Replace calls to __builtin_vcge* by
34202 <=, >= operators in vcle and vcge intrinsics respectively.
34203 * config/arm/arm_neon_builtins.def: Remove entry for
34206 2021-01-14 Uroš Bizjak <ubizjak@gmail.com>
34209 * config/i386/i386-options.c (ix86_function_specific_save):
34210 Remove redundant assignment to opts->x_ix86_branch_cost.
34211 * config/i386/i386.c (ix86_prefetch_sse):
34212 Rename from x86_prefetch_sse. Update all uses.
34213 * config/i386/i386.h: Update for rename.
34214 * config/i386/i386-options.h: Ditto.
34216 2021-01-14 Jakub Jelinek <jakub@redhat.com>
34219 * config/i386/sse.md (*sse4_1_zero_extendv8qiv8hi2_3,
34220 *sse4_1_zero_extendv4hiv4si2_3, *sse4_1_zero_extendv2siv2di2_3):
34221 Use Bm instead of m for non-avx. Add isa attribute.
34223 2021-01-14 Jakub Jelinek <jakub@redhat.com>
34225 PR tree-optimization/96688
34226 * match.pd (~(X >> Y) -> ~X >> Y): New simplification if
34227 ~X can be simplified.
34229 2021-01-14 Richard Sandiford <richard.sandiford@arm.com>
34231 * tree-vect-stmts.c (vect_model_load_cost): Account for unused
34232 IFN_LOAD_LANES results.
34234 2021-01-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
34236 * config/aarch64/aarch64-simd.md (aarch64_<su>xtl<mode>):
34238 (aarch64_xtn<mode>): Likewise.
34239 * config/aarch64/aarch64-simd-builtins.def (sxtl, uxtl, xtn):
34242 * config/aarch64/arm_neon.h (vmovl_s8): Reimplement using
34244 (vmovl_s16): Likewise.
34245 (vmovl_s32): Likewise.
34246 (vmovl_u8): Likewise.
34247 (vmovl_u16): Likewise.
34248 (vmovl_u32): Likewise.
34249 (vmovn_s16): Likewise.
34250 (vmovn_s32): Likewise.
34251 (vmovn_s64): Likewise.
34252 (vmovn_u16): Likewise.
34253 (vmovn_u32): Likewise.
34254 (vmovn_u64): Likewise.
34256 2021-01-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
34258 * config/aarch64/aarch64-simd.md (aarch64_<su>qxtn2<mode>_le):
34260 (aarch64_<su>qxtn2<mode>_be): Likewise.
34261 (aarch64_<su>qxtn2<mode>): Likewise.
34262 * config/aarch64/aarch64-simd-builtins.def (sqxtn2, uqxtn2):
34264 * config/aarch64/iterators.md (SAT_TRUNC): Define code_iterator.
34265 (su): Handle ss_truncate and us_truncate.
34266 * config/aarch64/arm_neon.h (vqmovn_high_s16): Reimplement using
34268 (vqmovn_high_s32): Likewise.
34269 (vqmovn_high_s64): Likewise.
34270 (vqmovn_high_u16): Likewise.
34271 (vqmovn_high_u32): Likewise.
34272 (vqmovn_high_u64): Likewise.
34274 2021-01-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
34276 * config/aarch64/aarch64-simd.md (aarch64_xtn2<mode>_le):
34278 (aarch64_xtn2<mode>_be): Likewise.
34279 (aarch64_xtn2<mode>): Likewise.
34280 * config/aarch64/aarch64-simd-builtins.def (xtn2): Define
34282 * config/aarch64/arm_neon.h (vmovn_high_s16): Reimplement using
34284 (vmovn_high_s32): Likewise.
34285 (vmovn_high_s64): Likewise.
34286 (vmovn_high_u16): Likewise.
34287 (vmovn_high_u32): Likewise.
34288 (vmovn_high_u64): Likewise.
34290 2021-01-13 Stafford Horne <shorne@gmail.com>
34292 * config/or1k/or1k.h (ASM_PREFERRED_EH_DATA_FORMAT): New macro.
34294 2021-01-13 Stafford Horne <shorne@gmail.com>
34296 * config/or1k/linux.h (TARGET_ASM_FILE_END): Define macro.
34298 2021-01-13 Stafford Horne <shorne@gmail.com>
34300 * config/or1k/or1k.h (TARGET_CPU_CPP_BUILTINS): Add builtin
34301 define for __or1k_hard_float__.
34303 2021-01-13 Stafford Horne <shorne@gmail.com>
34305 * config/or1k/or1k.h (NO_PROFILE_COUNTERS): Define as 1.
34306 (PROFILE_HOOK): Define to call _mcount.
34307 (FUNCTION_PROFILER): Change from abort to no-op.
34309 2021-01-13 Jakub Jelinek <jakub@redhat.com>
34311 PR tree-optimization/96691
34312 * match.pd ((~X | C) ^ D -> (X | C) ^ (~D ^ C),
34313 (~X & C) ^ D -> (X & C) ^ (D ^ C)): New simplifications if
34314 (~D ^ C) or (D ^ C) can be simplified.
34316 2021-01-13 Richard Biener <rguenther@suse.de>
34318 PR tree-optimization/92645
34319 * match.pd (BIT_FIELD_REF to conversion): Delay canonicalization
34320 until after vector lowering.
34322 2021-01-13 Richard Sandiford <richard.sandiford@arm.com>
34324 * config/aarch64/aarch64-sve.md (fnma<mode>4): Extend from SVE_FULL_I
34326 (@aarch64_pred_fnma<mode>, cond_fnma<mode>, *cond_fnma<mode>_2)
34327 (*cond_fnma<mode>_4, *cond_fnma<mode>_any): Likewise.
34329 2021-01-13 Richard Sandiford <richard.sandiford@arm.com>
34331 * config/aarch64/aarch64-sve.md (fma<mode>4): Extend from SVE_FULL_I
34333 (@aarch64_pred_fma<mode>, cond_fma<mode>, *cond_fma<mode>_2)
34334 (*cond_fma<mode>_4, *cond_fma<mode>_any): Likewise.
34336 2021-01-13 Richard Biener <rguenther@suse.de>
34338 PR tree-optimization/92645
34339 * tree-vect-slp.c (vect_build_slp_tree_1): Relax supported
34340 BIT_FIELD_REF argument.
34341 (vect_build_slp_tree_2): Record the desired vector type
34342 on the external vector def.
34343 (vectorizable_slp_permutation): Handle required punning
34344 of existing vector defs.
34346 2021-01-13 Richard Sandiford <richard.sandiford@arm.com>
34348 * rtl-ssa/accesses.h (def_lookup): Fix order of comparison results.
34350 2021-01-13 Richard Sandiford <richard.sandiford@arm.com>
34352 * config/sh/sh.md (movsf_ie): Remove operands[2] test.
34354 2021-01-13 Samuel Thibault <samuel.thibault@ens-lyon.org>
34356 * config.gcc [$target == *-*-gnu*]: Enable
34357 'default_gnu_indirect_function'.
34359 2021-01-13 Jakub Jelinek <jakub@redhat.com>
34362 * optabs.c (expand_vec_perm_const): Don't force v0 and v1 into
34363 registers before calling targetm.vectorize.vec_perm_const, only after
34365 * config/i386/i386-expand.c (ix86_vectorize_vec_perm_const): Handle
34366 two argument permutation when one operand is zero vector and only
34367 after that force operands into registers.
34368 * config/i386/sse.md (*avx2_zero_extendv16qiv16hi2_1): New
34369 define_insn_and_split pattern.
34370 (*avx512bw_zero_extendv32qiv32hi2_1): Likewise.
34371 (*avx512f_zero_extendv16hiv16si2_1): Likewise.
34372 (*avx2_zero_extendv8hiv8si2_1): Likewise.
34373 (*avx512f_zero_extendv8siv8di2_1): Likewise.
34374 (*avx2_zero_extendv4siv4di2_1): Likewise.
34375 * config/mips/mips.c (mips_vectorize_vec_perm_const): Force operands
34377 * config/arm/arm.c (arm_vectorize_vec_perm_const): Likewise.
34378 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Likewise.
34379 * config/ia64/ia64.c (ia64_vectorize_vec_perm_const): Likewise.
34380 * config/aarch64/aarch64.c (aarch64_vectorize_vec_perm_const): Likewise.
34381 * config/rs6000/rs6000.c (rs6000_vectorize_vec_perm_const): Likewise.
34382 * config/gcn/gcn.c (gcn_vectorize_vec_perm_const): Likewise. Use std::swap.
34384 2021-01-13 Martin Liska <mliska@suse.cz>
34386 PR tree-optimization/98455
34387 * gimple-if-to-switch.cc (condition_info::record_phi_mapping):
34388 Record also virtual PHIs.
34389 (pass_if_to_switch::execute): Return TODO_cleanup_cfg only
34392 2021-01-13 Jonathan Wakely <jwakely@redhat.com>
34394 * doc/invoke.texi (C++ Modules): Fix typos.
34396 2021-01-13 Richard Biener <rguenther@suse.de>
34398 PR tree-optimization/98640
34399 * tree-ssa-sccvn.c (visit_nary_op): Do not try to
34400 handle plus or minus from a truncated operand to be
34403 2021-01-13 Jakub Jelinek <jakub@redhat.com>
34406 * config/i386/i386.md (*btr<mode>_1, *btr<mode>_2): New
34407 define_insn_and_split patterns.
34408 (splitter after *btr<mode>_2): New splitter.
34410 2021-01-13 Martin Liska <mliska@suse.cz>
34413 * cgraphunit.c (analyze_functions): Remove dead code.
34415 2021-01-13 Qian Jianhua <qianjh@cn.fujitsu.com>
34417 * config/aarch64/aarch64-cost-tables.h (a64fx_extra_costs): New.
34418 * config/aarch64/aarch64.c (a64fx_addrcost_table): New.
34419 (a64fx_regmove_cost, a64fx_vector_cost): New.
34420 (a64fx_tunings): Use the new added cost tables.
34422 2021-01-13 Jakub Jelinek <jakub@redhat.com>
34425 * config/i386/predicates.md (pmovzx_parallel): New predicate.
34426 * config/i386/sse.md (*sse4_1_zero_extendv8qiv8hi2_3): New
34427 define_insn_and_split pattern.
34428 (*sse4_1_zero_extendv4hiv4si2_3): Likewise.
34429 (*sse4_1_zero_extendv2siv2di2_3): Likewise.
34431 2021-01-13 Julian Brown <julian@codesourcery.com>
34433 * config/gcn/gcn.c (gcn_conditional_register_usage): Remove dead code
34434 to fix v0 register.
34436 2021-01-13 Julian Brown <julian@codesourcery.com>
34438 * config/gcn/gcn.c (gcn_md_reorg): Fix case where EXEC reg is live
34441 2021-01-13 Julian Brown <julian@codesourcery.com>
34443 * config/gcn/gcn-valu.md (recip<mode>2<exec>, recip<mode>2): Use unspec
34444 for reciprocal-approximation instructions.
34445 (div<mode>3): Use fused multiply-accumulate operations for reciprocal
34446 refinement and division result.
34447 * config/gcn/gcn.md (UNSPEC_RCP): New unspec constant.
34449 2021-01-13 Julian Brown <julian@codesourcery.com>
34451 * config/gcn/gcn-valu.md (subdf): Rename to...
34454 2021-01-12 Martin Liska <mliska@suse.cz>
34456 * gcov.c (source_info::debug): Fix printf format for 32-bit hosts.
34458 2021-01-12 Andrea Corallo <andrea.corallo@arm.com>
34460 * function-abi.h: Fix typo.
34462 2021-01-12 Christophe Lyon <christophe.lyon@linaro.org>
34466 * config/arm/arm.h (ARM_HAVE_NEON_V8QI_LDST): New macro.
34467 (ARM_HAVE_NEON_V16QI_LDST, ARM_HAVE_NEON_V4HI_LDST): Likewise.
34468 (ARM_HAVE_NEON_V8HI_LDST, ARM_HAVE_NEON_V2SI_LDST): Likewise.
34469 (ARM_HAVE_NEON_V4SI_LDST, ARM_HAVE_NEON_V4HF_LDST): Likewise.
34470 (ARM_HAVE_NEON_V8HF_LDST, ARM_HAVE_NEON_V4BF_LDST): Likewise.
34471 (ARM_HAVE_NEON_V8BF_LDST, ARM_HAVE_NEON_V2SF_LDST): Likewise.
34472 (ARM_HAVE_NEON_V4SF_LDST, ARM_HAVE_NEON_DI_LDST): Likewise.
34473 (ARM_HAVE_NEON_V2DI_LDST): Likewise.
34474 (ARM_HAVE_V8QI_LDST, ARM_HAVE_V16QI_LDST): Likewise.
34475 (ARM_HAVE_V4HI_LDST, ARM_HAVE_V8HI_LDST): Likewise.
34476 (ARM_HAVE_V2SI_LDST, ARM_HAVE_V4SI_LDST, ARM_HAVE_V4HF_LDST): Likewise.
34477 (ARM_HAVE_V8HF_LDST, ARM_HAVE_V4BF_LDST, ARM_HAVE_V8BF_LDST): Likewise.
34478 (ARM_HAVE_V2SF_LDST, ARM_HAVE_V4SF_LDST, ARM_HAVE_DI_LDST): Likewise.
34479 (ARM_HAVE_V2DI_LDST): Likewise.
34480 * config/arm/mve.md (*movmisalign<mode>_mve_store): New pattern.
34481 (*movmisalign<mode>_mve_load): New pattern.
34482 * config/arm/neon.md (movmisalign<mode>): Move to ...
34483 * config/arm/vec-common.md: ... here.
34485 2021-01-12 Vladimir N. Makarov <vmakarov@redhat.com>
34488 * lra-eliminations.c (eliminate_regs_in_insn): Add transformation
34489 of pattern 'plus (plus (hard reg, const), pseudo)'.
34491 2021-01-12 Richard Biener <rguenther@suse.de>
34493 PR tree-optimization/98550
34494 * tree-vect-slp.c (vect_record_max_nunits): Check whether
34495 the group size is a multiple of the vector element count.
34496 (vect_build_slp_tree_1): When we need to fail because
34497 the vector type choosen causes unrolling do so lazily
34498 without affecting matches only at the end to guide group splitting.
34500 2021-01-12 Martin Liska <mliska@suse.cz>
34503 * optc-save-gen.awk: Compare also n_target_save vars with
34506 2021-01-12 Martin Liska <mliska@suse.cz>
34508 * gcov.c (source_info::debug): New.
34509 (print_usage): Add --debug (-D) option.
34510 (process_args): Likewise.
34511 (generate_results): Call src->debug after
34512 accumulate_line_counts.
34513 (read_graph_file): Properly assign id for EXIT_BLOCK.
34514 * profile.c (branch_prob): Dump function body before it is
34517 2021-01-12 Jakub Jelinek <jakub@redhat.com>
34519 PR tree-optimization/98629
34520 * tree-ssa-math-opts.c (arith_overflow_check_p): Don't update use_stmt
34521 unless returning non-zero.
34523 2021-01-12 Jakub Jelinek <jakub@redhat.com>
34525 PR tree-optimization/95731
34526 * tree-ssa-reassoc.c (optimize_range_tests_cmp_bitwise): Also optimize
34527 x < 0 && y < 0 && z < 0 into (x | y | z) < 0 for signed x, y, z.
34528 (optimize_range_tests): Call optimize_range_tests_cmp_bitwise
34529 only after optimize_range_tests_var_bound.
34531 2021-01-12 Jakub Jelinek <jakub@redhat.com>
34533 * configure.ac: Ensure c/Make-lang.in comes first in @all_lang_makefrags@.
34534 * configure: Regenerated.
34536 2021-01-12 liuhongt <hongtao.liu@intel.com>
34539 * config/i386/i386-builtins.h (BUILTIN_DESC_SWAP_OPERANDS):
34541 * config/i386/i386-expand.c (ix86_expand_sse_comi): Delete
34544 2021-01-12 Alexandre Oliva <oliva@adacore.com>
34546 * ssa-iterators.h (end_imm_use_stmt_traverse): Forward
34548 (auto_end_imm_use_stmt_traverse): New struct.
34549 (FOR_EACH_IMM_USE_STMT): Use it.
34550 (BREAK_FROM_IMM_USE_STMT, RETURN_FROM_IMM_USE_STMT): Remove,
34552 * gimple-ssa-strength-reduction.c: ... here, ...
34553 * graphite-scop-detection.c: ... here, ...
34554 * ipa-modref.c, ipa-pure-const.c, ipa-sra.c: ... here, ...
34555 * tree-predcom.c, tree-ssa-ccp.c: ... here, ...
34556 * tree-ssa-dce.c, tree-ssa-dse.c: ... here, ...
34557 * tree-ssa-loop-ivopts.c, tree-ssa-math-opts.c: ... here, ...
34558 * tree-ssa-phiprop.c, tree-ssa.c: ... here, ...
34559 * tree-vect-slp.c: ... and here, ...
34560 * doc/tree-ssa.texi: ... and the example here.
34562 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
34564 * config/aarch64/aarch64-sve.md (sdiv_pow2<mode>3): Extend from
34565 SVE_FULL_I to SVE_I. Generate an UNSPEC_PRED_X.
34566 (*sdiv_pow2<mode>3): New pattern.
34567 (@cond_<sve_int_op><mode>): Extend from SVE_FULL_I to SVE_I.
34568 Wrap the ASRD in an UNSPEC_PRED_X.
34569 (*cond_<sve_int_op><mode>_2): Likewise. Replace the UNSPEC_PRED_X
34570 predicate with a constant PTRUE, if it isn't already.
34571 (*cond_<sve_int_op><mode>_z): Replace with...
34572 (*cond_<sve_int_op><mode>_any): ...this new pattern.
34574 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
34576 * config/aarch64/aarch64-sve.md (*cond_bic<mode>_2): Extend from
34577 SVE_FULL_I to SVE_I.
34578 (*cond_bic<mode>_any): Likewise.
34580 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
34582 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart)
34583 (@aarch64_pred_<MUL_HIGHPART:optab><mode>): Extend from SVE_FULL_I
34586 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
34588 * config/aarch64/aarch64-sve.md (<su>abd<mode>_3): Extend from
34589 SVE_FULL_I to SVE_I.
34590 (*aarch64_cond_<su>abd<mode>_2): Likewise.
34591 (*aarch64_cond_<su>abd<mode>_any): Likewise.
34592 (@aarch64_pred_<su>abd<mode>): Likewise. Use UNSPEC_PRED_X
34593 for the max and min but not for the minus.
34594 (*aarch64_cond_<su>abd<mode>_3): New pattern.
34596 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
34598 * config/aarch64/iterators.md (SVE_24I): New iterator.
34599 * config/aarch64/aarch64-sve.md (*aarch64_adr<mode>_shift): Extend from
34600 SVE_FULL_SDI to SVE_24I. Use containers rather than elements.
34602 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
34604 * config/aarch64/aarch64-sve.md (@cond_<SVE_INT_BINARY:optab><mode>)
34605 (*cond_<SVE_INT_BINARY:optab><mode>_2): Extend from SVE_FULL_I
34607 (*cond_<SVE_INT_BINARY:optab><mode>_3): Likewise.
34608 (*cond_<SVE_INT_BINARY:optab><mode>_any): Likewise.
34609 (*cond_<SVE_INT_BINARY:optab><mode>_2_const): Likewise.
34610 (*cond_<SVE_INT_BINARY:optab><mode>_any_const): Likewise.
34612 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
34614 * config/aarch64/aarch64-sve.md (<SVE_INT_BINARY_IMM:optab><mode>3)
34615 (@aarch64_pred_<SVE_INT_BINARY_IMM:optab><mode>)
34616 (*post_ra_<SVE_INT_BINARY_IMM:optab><mode>3): Extend from SVE_FULL_I
34619 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
34621 * config/aarch64/aarch64-sve.md (<ASHIFT:optab><mode>3)
34622 (v<ASHIFT:optab><mode>3, @aarch64_pred_<optab><mode>)
34623 (*post_ra_v<ASHIFT:optab><mode>3): Extend from SVE_FULL_I to SVE_I.
34625 2021-01-11 Martin Liska <mliska@suse.cz>
34628 * symtab-clones.h (clone_info::release): Release
34629 symtab::m_clones with ggc_delete as it's a GGC memory.
34631 2021-01-11 Matthias Klose <doko@ubuntu.com>
34633 * Makefile.in (LINK_PROGRESS): Show the link target.
34635 2021-01-11 Richard Biener <rguenther@suse.de>
34637 PR tree-optimization/91403
34638 * tree-vect-data-refs.c (vect_analyze_group_access_1): Cap
34639 single-element interleaving group size at 4096 elements.
34641 2021-01-11 Richard Biener <rguenther@suse.de>
34643 PR tree-optimization/98526
34644 * tree-vect-loop.c (vect_model_reduction_cost): Remove costing
34645 of the actual reduction op for the regular case.
34646 (vectorizable_reduction): Cost the stmts
34647 vect_transform_reduction produces here.
34649 2021-01-11 Andreas Krebbel <krebbel@linux.ibm.com>
34651 * tree-ssa-forwprop.c (simplify_vector_constructor): For
34652 big-endian, use UNPACK[_FLOAT]_HI.
34654 2021-01-11 Tamar Christina <tamar.christina@arm.com>
34656 * tree-vect-slp-patterns.c (class complex_pattern,
34657 class complex_add_pattern): Add parameters to matches.
34658 (complex_add_pattern::build): Free memory.
34659 (complex_add_pattern::matches): Move validation end of match.
34660 (complex_add_pattern::recognize): Likewise.
34662 2021-01-11 Tamar Christina <tamar.christina@arm.com>
34664 * tree-vect-slp-patterns.c (linear_loads_p): Fix externals.
34666 2021-01-11 Tamar Christina <tamar.christina@arm.com>
34668 * tree-vect-slp-patterns.c (is_linear_load_p): Fix ambiguity.
34670 2021-01-11 Jakub Jelinek <jakub@redhat.com>
34672 PR tree-optimization/95867
34673 * tree-ssa-math-opts.h: New header.
34674 * tree-ssa-math-opts.c: Include tree-ssa-math-opts.h.
34675 (powi_as_mults): No longer static. Use build_one_cst instead of
34676 build_real. Formatting fix.
34677 * tree-ssa-reassoc.c: Include tree-ssa-math-opts.h.
34678 (attempt_builtin_powi): Handle multiplication reassociation without
34679 powi_fndecl using powi_as_mults.
34680 (reassociate_bb): For integral types don't require
34681 -funsafe-math-optimizations to call attempt_builtin_powi.
34683 2021-01-11 Jakub Jelinek <jakub@redhat.com>
34685 PR tree-optimization/95852
34686 * tree-ssa-math-opts.c (maybe_optimize_guarding_check): Change
34687 mul_stmts parameter type to vec<gimple *> &. Before cond_stmt
34688 allow in the bb any of the stmts in that vector, div_stmt and
34689 up to 3 cast stmts.
34690 (arith_cast_equal_p): New function.
34691 (arith_overflow_check_p): Add cast_stmt argument, handle signed
34692 multiply overflow checks.
34693 (match_arith_overflow): Adjust caller. Handle signed multiply
34696 2021-01-11 Jakub Jelinek <jakub@redhat.com>
34698 PR tree-optimization/95852
34699 * tree-ssa-math-opts.c (maybe_optimize_guarding_check): New function.
34700 (uaddsub_overflow_check_p): Renamed to ...
34701 (arith_overflow_check_p): ... this. Handle also multiplication
34702 with overflow check.
34703 (match_uaddsub_overflow): Renamed to ...
34704 (match_arith_overflow): ... this. Add cfg_changed argument. Handle
34705 also multiplication with overflow check. Adjust function comment.
34706 (math_opts_dom_walker::after_dom_children): Adjust callers. Call
34707 match_arith_overflow also for MULT_EXPR.
34709 2021-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
34711 * config/aarch64/arm_neon.h (vmovl_s8): Reimplement using
34712 __builtin_convertvector.
34713 (vmovl_s16): Likewise.
34714 (vmovl_s32): Likewise.
34715 (vmovl_u8): Likewise.
34716 (vmovl_u16): Likewise.
34717 (vmovl_u32): Likewise.
34718 (vmovn_s16): Likewise.
34719 (vmovn_s32): Likewise.
34720 (vmovn_s64): Likewise.
34721 (vmovn_u16): Likewise.
34722 (vmovn_u32): Likewise.
34723 (vmovn_u64): Likewise.
34725 2021-01-11 Martin Liska <mliska@suse.cz>
34727 * gimple-if-to-switch.cc (struct condition_info): Use auto_var.
34728 (if_chain::is_beneficial): Delete clusters
34729 (find_conditions): Make second argument of conditions_in_bbs a
34730 pointer so that we control over it's lifetime.
34731 (pass_if_to_switch::execute): Delete them.
34733 2021-01-11 Kewen Lin <linkw@linux.ibm.com>
34735 * ira.c (move_unallocated_pseudos): Check other_reg and skip if
34738 2021-01-09 Maciej W. Rozycki <macro@linux-mips.org>
34740 * config/vax/vax.md (cc): Remove mode attribute.
34741 (subst_<cc>, subst_f<cc>): Rename to...
34742 (subst_<mode>, subst_f<VAXccnz:mode>): ... these respectively.
34743 (*cbranch<VAXint:mode>4_<VAXcc:mode>): Update for `cc' removal.
34744 (*cbranch<VAXfp:mode>4_<VAXccnz:mode>): Likewise.
34745 (*branch_<mode>, *branch_<mode>_reversed): Likewise.
34747 2021-01-09 Maciej W. Rozycki <macro@linux-mips.org>
34749 * config/vax/vax.md (subst_f<cc>): Add mode to operands and
34750 `const_double_zero'.
34752 2021-01-09 Maciej W. Rozycki <macro@linux-mips.org>
34754 * config/pdp11/pdp11.md (PDPfp): New mode iterator.
34755 (fcc_cc, fcc_ccnz): Use it. Add mode to `const_double_zero' and
34758 2021-01-09 Maciej W. Rozycki <macro@linux-mips.org>
34760 * genemit.c (gen_exp) <CONST_DOUBLE>: Handle `const_double_zero'
34762 * read-rtl.c (rtx_reader::read_rtx_code): Handle machine mode
34763 with `const_double_zero'.
34764 * doc/rtl.texi (Constant Expression Types): Document it.
34766 2021-01-09 Jakub Jelinek <jakub@redhat.com>
34769 * tree-cfg.c (verify_gimple_assign_binary): Allow lhs of
34770 POINTER_DIFF_EXPR to be any integral type.
34772 2021-01-09 Jakub Jelinek <jakub@redhat.com>
34774 PR rtl-optimization/98603
34775 * function.c (instantiate_virtual_regs_in_insn): For asm goto
34776 with impossible constraints, drop all SETs, CLOBBERs, drop PARALLEL
34777 if any, set ASM_OPERANDS mode to VOIDmode and change
34778 ASM_OPERANDS_OUTPUT_CONSTRAINT and ASM_OPERANDS_OUTPUT_IDX.
34780 2021-01-09 Alexandre Oliva <oliva@gnu.org>
34783 * final.c (notice_source_line): Narrow down the condition to
34784 skip a line-0 marker.
34786 2021-01-08 Sergei Trofimovich <siarheit@google.com>
34788 * ipa-modref.c (merge_call_side_effects): Fix
34789 linebreak split by reordering two print calls.
34791 2021-01-08 Ilya Leoshkevich <iii@linux.ibm.com>
34793 * config/s390/vector.md (*tf_to_fprx2_0): Rename from
34794 "*mov_tf_to_fprx2_0" for consistency, fix constraint.
34795 (*tf_to_fprx2_1): Rename from "*mov_tf_to_fprx2_1" for
34796 consistency, fix constraint.
34798 2021-01-08 Ilya Leoshkevich <iii@linux.ibm.com>
34800 * config/s390/s390-c.c (s390_def_or_undef_macro): Accept
34801 callables instead of mask values.
34802 (struct target_flag_set_p): New predicate.
34803 (s390_cpu_cpp_builtins_internal): Define or undefine
34804 __LONG_DOUBLE_VX__ macro.
34806 2021-01-08 H.J. Lu <hjl.tools@gmail.com>
34809 * config/i386/i386.c (x86_function_profiler): Use R10 and R11
34810 to call mcount in large model with PIC for NO_PROFILE_COUNTERS
34813 2021-01-08 Richard Biener <rguenther@suse.de>
34815 * tree-ssa-sccvn.c (pass_fre::execute): Reset the SCEV hash table.
34817 2021-01-08 Richard Biener <rguenther@suse.de>
34819 * tree-vect-slp.c (scalar_stmts_to_slp_tree_map_t): Fix.
34820 (vect_build_slp_tree): On cache hit release the matched
34821 scalar stmts vector.
34822 * tree-vect-stmts.c (vectorizable_store): Properly free
34823 vec_oprnds before possibly gathering them again.
34825 2021-01-08 Richard Biener <rguenther@suse.de>
34827 PR tree-optimization/98544
34828 * tree-vect-slp.c (vect_optimize_slp): Always materialize
34829 permutes at a permute node.
34831 2021-01-08 H.J. Lu <hjl.tools@gmail.com>
34834 * config/i386/i386.c (x86_function_profiler): Use R10 to call
34835 mcount in large model. Sorry for large model with PIC.
34837 2021-01-08 Jakub Jelinek <jakub@redhat.com>
34840 * config/i386/i386.opt (ix86_cmodel, ix86_incoming_stack_boundary_arg,
34841 ix86_pmode, ix86_preferred_stack_boundary_arg, ix86_regparm,
34842 ix86_veclibabi_type): Remove x_ prefix, use TargetVariable instead of
34843 TargetSave and initialize for variables with enum types.
34844 (mfentry, mstack-protector-guard-reg=, mstack-protector-guard-offset=,
34845 mstack-protector-guard-symbol=): Add Save.
34846 * config/i386/i386-options.c (ix86_function_specific_save,
34847 ix86_function_specific_restore): Don't save or restore x_ix86_cmodel,
34848 x_ix86_incoming_stack_boundary_arg, x_ix86_pmode,
34849 x_ix86_preferred_stack_boundary_arg, x_ix86_regparm,
34850 x_ix86_veclibabi_type.
34852 2021-01-08 Richard Sandiford <richard.sandiford@arm.com>
34854 * config/aarch64/aarch64-sve.md (*cnot<mode>): Extend from
34855 SVE_FULL_I to SVE_I.
34856 (*cond_cnot<mode>_2, *cond_cnot<mode>_any): Likewise.
34858 2021-01-08 Richard Sandiford <richard.sandiford@arm.com>
34860 * config/aarch64/aarch64-sve.md (*cond_uxt<mode>_2): Extend from
34861 SVE_FULL_I to SVE_I.
34862 (*cond_uxt<mode>_any): Likewise.
34864 2021-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
34866 * config/aarch64/iterators.md (Vwhalf): New iterator.
34867 * config/aarch64/aarch64-simd.md (aarch64_<sur>adalp<mode>_3):
34869 (aarch64_<sur>adalp<mode>): ... This. Make more
34871 (<sur>sadv16qi): Adjust callsite of the above.
34872 * config/aarch64/aarch64-simd-builtins.def (sadalp, uadalp): New
34874 * config/aarch64/arm_neon.h (vpadal_s8): Reimplement using
34876 (vpadal_s16): Likewise.
34877 (vpadal_u8): Likewise.
34878 (vpadal_u16): Likewise.
34879 (vpadalq_s8): Likewise.
34880 (vpadalq_s16): Likewise.
34881 (vpadalq_s32): Likewise.
34882 (vpadalq_u8): Likewise.
34883 (vpadalq_u16): Likewise.
34884 (vpadalq_u32): Likewise.
34886 2021-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
34888 * config/aarch64/aarch64-simd.md (aarch64_<su>abd<mode>_3):
34890 (aarch64_<su>abd<mode>): ... This.
34891 (<sur>sadv16qi): Adjust callsite of the above.
34892 * config/aarch64/aarch64-simd-builtins.def (sabd, uabd): Define
34894 * config/aarch64/arm_neon.h (vabd_s8): Reimplement using
34896 (vabd_s16): Likewise.
34897 (vabd_s32): Likewise.
34898 (vabd_u8): Likewise.
34899 (vabd_u16): Likewise.
34900 (vabd_u32): Likewise.
34901 (vabdq_s8): Likewise.
34902 (vabdq_s16): Likewise.
34903 (vabdq_s32): Likewise.
34904 (vabdq_u8): Likewise.
34905 (vabdq_u16): Likewise.
34906 (vabdq_u32): Likewise.
34908 2021-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
34910 * config/aarch64/aarch64-simd-builtins.def (saba, uaba): Define
34912 * config/aarch64/arm_neon.h (vaba_s8): Implement using builtin.
34913 (vaba_s16): Likewise.
34914 (vaba_s32): Likewise.
34915 (vaba_u8): Likewise.
34916 (vaba_u16): Likewise.
34917 (vaba_u32): Likewise.
34918 (vabaq_s8): Likewise.
34919 (vabaq_s16): Likewise.
34920 (vabaq_s32): Likewise.
34921 (vabaq_u8): Likewise.
34922 (vabaq_u16): Likewise.
34923 (vabaq_u32): Likewise.
34925 2021-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
34927 * config/aarch64/aarch64-simd.md (aba<mode>_3): Rename to...
34928 (aarch64_<su>aba<mode>): ... This. Handle uaba as well.
34929 Change RTL pattern to match.
34931 2021-01-08 Kito Cheng <kito.cheng@sifive.com>
34933 * common/config/riscv/riscv-common.c (riscv_current_subset_list): New.
34934 * config/riscv/riscv-c.c (riscv-subset.h): New.
34935 (INCLUDE_STRING): Define.
34936 (riscv_cpu_cpp_builtins): Add new style architecture extension
34938 * config/riscv/riscv-subset.h (riscv_subset_list::begin): New.
34939 (riscv_subset_list::end): New.
34940 (riscv_current_subset_list): New.
34942 2021-01-08 Kito Cheng <kito.cheng@sifive.com>
34944 * common/config/riscv/riscv-common.c (RISCV_DONT_CARE_VERSION):
34945 Move to riscv-subset.h.
34946 (struct riscv_subset_t): Ditto.
34947 (class riscv_subset_list): Ditto.
34948 * config/riscv/riscv-subset.h (RISCV_DONT_CARE_VERSION): Move
34949 from riscv-common.c.
34950 (struct riscv_subset_t): Ditto.
34951 (class riscv_subset_list): Ditto.
34952 * config/riscv/t-riscv ($(common_out_file)): Add file
34955 2021-01-07 Jakub Jelinek <jakub@redhat.com>
34958 * config/i386/i386.md (*bmi_blsi_<mode>_cmp, *bmi_blsi_<mode>_ccno):
34959 New define_insn patterns.
34961 2021-01-07 Richard Sandiford <richard.sandiford@arm.com>
34963 * config/aarch64/aarch64-sve.md (@cond_<SVE_INT_UNARY:optab><mode>)
34964 (*cond_<SVE_INT_UNARY:optab><mode>_2): Extend from SVE_FULL_I to SVE_I.
34965 (*cond_<SVE_INT_UNARY:optab><mode>_any): Likewise.
34967 2021-01-07 Richard Sandiford <richard.sandiford@arm.com>
34969 PR tree-optimization/98560
34970 * internal-fn.def (IFN_VCONDU, IFN_VCONDEQ): Use type vec_cond.
34971 * internal-fn.c (vec_cond_mask_direct): Get the data mode from
34973 (vec_cond_direct): Likewise argument 2.
34974 (vec_condu_direct, vec_condeq_direct): Delete.
34975 (expand_vect_cond_optab_fn): Rename to...
34976 (expand_vec_cond_optab_fn): ...this, replacing old macro.
34977 (expand_vec_condu_optab_fn, expand_vec_condeq_optab_fn): Delete.
34978 (expand_vect_cond_mask_optab_fn): Rename to...
34979 (expand_vec_cond_mask_optab_fn): ...this, replacing old macro.
34980 (direct_vec_cond_mask_optab_supported_p): Treat the optab as a
34982 (direct_vec_cond_optab_supported_p): Likewise.
34983 (direct_vec_condu_optab_supported_p): Delete.
34984 (direct_vec_condeq_optab_supported_p): Delete.
34985 * gimple-isel.cc: Include internal-fn.h.
34986 (gimple_expand_vec_cond_expr): Check that IFN_VCONDEQ is supported
34989 2021-01-07 Richard Sandiford <richard.sandiford@arm.com>
34991 PR tree-optimization/98560
34992 * gimple-isel.cc (gimple_expand_vec_cond_expr): If we fail to use
34993 IFN_VCOND{,U,EQ}, fall back on IFN_VCOND_MASK.
34995 2021-01-07 Uroš Bizjak <ubizjak@gmail.com>
34997 * config/i386/i386.md (insn): Merge from plusminus_insn, shift_insn,
34998 rotate_insn and optab code attributes.
34999 Update all uses to merged code attribute.
35000 * config/i386/sse.md: Update all uses to merged code attribute.
35001 * config/i386/mmx.md: Update all uses to merged code attribute.
35003 2021-01-07 Jakub Jelinek <jakub@redhat.com>
35005 PR tree-optimization/98568
35006 * gimple-ssa-store-merging.c (bswap_view_convert): New function.
35007 (bswap_replace): Use it.
35009 2021-01-06 Vladimir N. Makarov <vmakarov@redhat.com>
35011 PR rtl-optimization/97978
35012 * lra-int.h (lra_hard_reg_split_p): New external.
35013 * lra.c (lra_hard_reg_split_p): New global.
35014 (lra): Set up lra_hard_reg_split_p after splitting a hard reg.
35015 * lra-assigns.c (lra_assign): Don't check allocation correctness
35016 after hard reg splitting.
35018 2021-01-06 Martin Sebor <msebor@redhat.com>
35021 * builtins.c (new_delete_mismatch_p): New overload.
35022 (new_delete_mismatch_p (tree, tree)): Call it.
35024 2021-01-06 Alexandre Oliva <oliva@adacore.com>
35026 * Makefile.in (T_GLIMITS_H): New.
35027 (stmp-int-hdrs): Depend on it, use it.
35028 * config/t-vxworks (T_GLIMITS_H): Override it.
35029 (vxw-glimits.h): New.
35031 2021-01-06 Richard Biener <rguenther@suse.de>
35033 PR tree-optimization/98513
35034 * value-range.cc (intersect_ranges): Compare the upper bounds
35035 for the expected relation.
35037 2021-01-06 Gerald Pfeifer <gerald@pfeifer.com>
35040 2020-12-28 Gerald Pfeifer <gerald@pfeifer.com>
35042 * doc/standards.texi (HSAIL): Remove section.
35044 2021-01-05 Samuel Thibault <samuel.thibault@ens-lyon.org>
35046 * configure: Re-generate.
35048 2021-01-05 Jakub Jelinek <jakub@redhat.com>
35050 * doc/invoke.texi (-std=c++20): Adjust for the publication of
35051 ISO 14882:2020 standard.
35052 * doc/standards.texi: Likewise.
35054 2021-01-05 Jakub Jelinek <jakub@redhat.com>
35056 PR tree-optimization/94802
35057 * expr.h (maybe_optimize_sub_cmp_0): Declare.
35058 * expr.c: Include tree-pretty-print.h and flags.h.
35059 (maybe_optimize_sub_cmp_0): New function.
35060 (do_store_flag): Use it.
35061 * cfgexpand.c (expand_gimple_cond): Likewise.
35063 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
35065 * mux-utils.h (pointer_mux::m_ptr): Tweak description of contents.
35066 * rtlanal.c (simple_regno_set): Tweak description to clarify the
35069 2021-01-05 Richard Biener <rguenther@suse.de>
35071 PR tree-optimization/98516
35072 * tree-vect-slp.c (vect_optimize_slp): Permute the incoming
35073 lanes when materializing on a VEC_PERM node.
35074 (vectorizable_slp_permutation): Dump the permute properly.
35076 2021-01-05 Richard Biener <rguenther@suse.de>
35078 * tree-vect-slp.c (vect_slp_region): Move debug counter
35079 to cover individual subgraphs.
35081 2021-01-05 Richard Biener <rguenther@suse.de>
35083 PR tree-optimization/98428
35084 * tree-vect-slp.c (vect_build_slp_tree_1): Properly reject
35085 vector lane extracts for loop vectorization.
35087 2021-01-05 Jakub Jelinek <jakub@redhat.com>
35089 PR tree-optimization/98514
35090 * tree-ssa-reassoc.c (bb_rank): Change type from long * to
35092 (operand_rank): Change type from hash_map<tree, long> to
35093 hash_map<tree, int64_t>.
35094 (phi_rank): Change return type from long to int64_t.
35095 (loop_carried_phi): Change block_rank variable type from long to
35097 (propagate_rank): Change return type, rank parameter type and
35098 op_rank variable type from long to int64_t.
35099 (find_operand_rank): Change return type from long to int64_t
35100 and change slot variable type from long * to int64_t *.
35101 (insert_operand_rank): Change rank parameter type from long to
35103 (get_rank): Change return type and rank variable type from long to
35104 int64_t. Use PRId64 instead of ld to print the rank.
35105 (init_reassoc): Change rank variable type from long to int64_t
35106 and adjust correspondingly bb_rank and operand_rank initialization.
35108 2021-01-05 Jakub Jelinek <jakub@redhat.com>
35110 PR tree-optimization/96928
35111 * tree-ssa-phiopt.c (xor_replacement): New function.
35112 (tree_ssa_phiopt_worker): Call it.
35114 2021-01-05 Jakub Jelinek <jakub@redhat.com>
35116 PR tree-optimization/96930
35117 * match.pd ((A / (1 << B)) -> (A >> B)): If A is extended
35118 from narrower value which has the same type as 1 << B, perform
35119 the right shift on the narrower value followed by extension.
35121 2021-01-05 Jakub Jelinek <jakub@redhat.com>
35123 PR tree-optimization/96239
35124 * gimple-ssa-store-merging.c (maybe_optimize_vector_constructor): New
35126 (get_status_for_store_merging): Don't return BB_INVALID for blocks
35127 with potential bswap optimizable CONSTRUCTORs.
35128 (pass_store_merging::execute): Optimize vector CONSTRUCTORs with bswap
35131 2021-01-05 Richard Biener <rguenther@suse.de>
35133 PR tree-optimization/98381
35134 * tree.c (vector_element_bits): Properly compute bool vector
35136 * tree-vect-loop.c (vectorizable_live_operation): Properly
35137 compute the last lane bit offset.
35139 2021-01-05 Uroš Bizjak <ubizjak@gmail.com>
35142 * config/i386/sse.md (sse_cvtps2pi): Redefine as define_insn_and_split.
35143 Clear the top 64 bytes of the input XMM register.
35144 (sse_cvttps2pi): Ditto.
35146 2021-01-05 Uroš Bizjak <ubizjak@gmail.com>
35149 * config/i386/xopintrin.h (_mm256_cmov_si256): New.
35151 2021-01-05 H.J. Lu <hjl.tools@gmail.com>
35154 * config/i386/xmmintrin.h (_mm_extract_pi16): Cast to unsigned
35157 2021-01-05 Claudiu Zissulescu <claziss@synopsys.com>
35159 * config/arc/arc.md (maddsidi4_split): Use ACC_REG_FIRST.
35160 (umaddsidi4_split): Likewise.
35162 2021-01-05 liuhongt <hongtao.liu@intel.com>
35165 * config/i386/sse.md (*sse2_pmovskb_zexthisi): New
35166 define_insn_and_split for zero_extend of subreg HI of pmovskb
35168 (*sse2_pmovskb_zexthisi): Add new combine splitters for
35169 zero_extend of not of subreg HI of pmovskb result.
35171 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
35174 * explow.c (convert_memory_address_addr_space_1): Handle UNSPECs
35176 * config/aarch64/aarch64.c (aarch64_expand_mov_immediate): Use
35177 convert_memory_address to convert symbolic immediates to ptr_mode
35178 before forcing them to memory.
35180 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
35182 PR rtl-optimization/97144
35183 * recog.c (constrain_operands): Initialize matching_operand
35184 for each alternative, rather than only doing it once.
35186 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
35188 PR rtl-optimization/98403
35189 * rtl-ssa/changes.cc (function_info::finalize_new_accesses): Explain
35190 why we don't remove call clobbers.
35191 (function_info::apply_changes_to_insn): Don't attempt to add
35192 call clobbers here.
35194 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
35196 PR tree-optimization/98371
35197 * tree-vect-loop.c (vect_reanalyze_as_main_loop): New function.
35198 (vect_analyze_loop): If an epilogue loop appears to be cheaper
35199 than the main loop, re-analyze it as a main loop before adopting
35202 2021-01-05 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
35205 * configure.ac (NETLIBS): Determine using AX_LIB_SOCKET_NSL.
35206 * aclocal.m4, configure: Regenerate.
35207 * Makefile.in (NETLIBS): Define.
35208 (BACKEND): Remove $(CODYLIB).
35210 2021-01-05 Jakub Jelinek <jakub@redhat.com>
35212 PR rtl-optimization/98334
35213 * simplify-rtx.c (simplify_context::simplify_binary_operation_1):
35214 Optimize (X - 1) * Y + Y to X * Y or (X + 1) * Y - Y to X * Y.
35216 2021-01-05 Bernd Edlinger <bernd.edlinger@hotmail.de>
35218 * tree-inline.c (expand_call_inline): Restore input_location.
35219 Return result from recursive call.
35221 2021-01-04 Richard Sandiford <richard.sandiford@arm.com>
35223 PR tree-optimization/95401
35224 * config/aarch64/aarch64-sve-builtins.cc
35225 (gimple_folder::load_store_cookie): Use bits rather than bytes
35226 for the alignment argument to IFN_MASK_LOAD and IFN_MASK_STORE.
35227 * gimple-fold.c (gimple_fold_mask_load_store_mem_ref): Likewise.
35228 * tree-vect-stmts.c (vectorizable_store): Likewise.
35229 (vectorizable_load): Likewise.
35231 2021-01-04 Richard Biener <rguenther@suse.de>
35233 PR tree-optimization/98308
35234 * tree-vect-stmts.c (vectorizable_load): Set invariant mask
35237 2021-01-04 Jakub Jelinek <jakub@redhat.com>
35239 PR tree-optimization/95771
35240 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Handle types
35241 with precision smaller than int's precision and types with precision
35242 twice as large as long long. Formatting fixes.
35244 2021-01-04 Richard Biener <rguenther@suse.de>
35246 PR tree-optimization/98464
35247 * tree-ssa-sccvn.c (vn_valueize_for_srt): Rename from ...
35248 (vn_valueize_wrapper): ... this. Temporarily adjust vn_context_bb.
35249 (process_bb): Adjust.
35251 2021-01-04 Matthew Malcomson <matthew.malcomson@arm.com>
35254 * doc/invoke.texi (-fsanitize=address): Fix wording describing
35255 clash with -fsanitize=hwaddress.
35257 2021-01-04 Richard Biener <rguenther@suse.de>
35259 PR tree-optimization/98282
35260 * tree-ssa-sccvn.c (vn_get_stmt_kind): Classify tcc_reference on
35261 invariants as VN_NARY.
35263 2021-01-04 Richard Sandiford <richard.sandiford@arm.com>
35266 * config/aarch64/aarch64-simd.md (aarch64_combine<mode>): Accept
35267 aarch64_simd_reg_or_zero for operand 2. Use the combinez patterns
35268 to handle zero operands.
35270 2021-01-04 Richard Sandiford <richard.sandiford@arm.com>
35272 * config/aarch64/aarch64.c (offset_6bit_signed_scaled_p): New function.
35273 (offset_6bit_unsigned_scaled_p): Fix typo in comment.
35274 (aarch64_sve_prefetch_operand_p): Accept MUL VLs in the range
35277 2021-01-04 Richard Biener <rguenther@suse.de>
35279 PR tree-optimization/98393
35280 * tree-vect-slp.c (vect_build_slp_tree): Properly zero matches
35281 when hitting the limit.
35283 2021-01-04 Richard Biener <rguenther@suse.de>
35285 PR tree-optimization/98291
35286 * tree-vect-loop.c (vectorizable_reduction): Bypass
35287 associativity check for SLP reductions with VF 1.
35289 2021-01-04 Jakub Jelinek <jakub@redhat.com>
35291 PR tree-optimization/96782
35292 * match.pd (x == ~x -> false, x != ~x -> true): New simplifications.
35294 2021-01-04 Bernd Edlinger <bernd.edlinger@hotmail.de>
35296 * collect-utils.c (collect_execute): Check dumppfx.
35297 * collect2.c (maybe_run_lto_and_relink, do_link): Pass atsuffix
35298 to collect_execute.
35299 (do_link): Add new parameter atsuffix.
35300 (main): Handle -dumpdir option. Skip one argument for
35301 -o, -isystem and -B options.
35302 * gcc.c (make_at_file): New helper function.
35303 (close_at_file): Use it.
35305 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
35307 * config/darwin.h (MIN_LD64_NO_COAL_SECTS): Adjust.
35308 Amend handling for LD64_VERSION fallback defaults.
35310 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
35312 * config.gcc: Compute default version information
35313 from the configured target. Likewise defaults for
35315 * config/darwin10.h: Removed.
35316 * config/darwin12.h: Removed.
35317 * config/darwin9.h: Removed.
35318 * config/rs6000/darwin8.h: Removed.
35320 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
35322 * config/darwin9.h (ASM_OUTPUT_ALIGNED_COMMON): Delete.
35324 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
35326 * config/darwin9.h (STACK_CHECK_STATIC_BUILTIN): Move from here..
35327 * config/darwin.h (STACK_CHECK_STATIC_BUILTIN): .. to here.
35329 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
35331 * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Move from
35333 * config/darwin.h (LINK_GCC_C_SEQUENCE_SPEC): ... to here.
35335 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
35337 * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Move the spec
35338 for the Darwin10 unwinder stub from here ...
35339 * config/darwin.h (LINK_COMMAND_SPEC_A): ... to here.
35341 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
35343 * config/darwin.h (DSYMUTIL_SPEC): Default to DWARF
35344 (ASM_DEBUG_SPEC):Only define if the assembler supports
35346 (PREFERRED_DEBUGGING_TYPE): Default to DWARF.
35347 (DARWIN_PREFER_DWARF): Define.
35348 * config/darwin9.h (PREFERRED_DEBUGGING_TYPE): Remove.
35349 (DARWIN_PREFER_DWARF): Likewise
35350 (DSYMUTIL_SPEC): Likewise.
35351 (COLLECT_RUN_DSYMUTIL): Likewise.
35352 (ASM_DEBUG_SPEC): Likewise.
35353 (ASM_DEBUG_OPTION_SPEC): Likewise.
35355 2021-01-02 Jan Hubicka <jh@suse.cz>
35357 * cfg.c (free_block): ggc_free bb.
35359 2021-01-01 Jakub Jelinek <jakub@redhat.com>
35361 * gcc.c (process_command): Update copyright notice dates.
35362 * gcov-dump.c (print_version): Ditto.
35363 * gcov.c (print_version): Ditto.
35364 * gcov-tool.c (print_version): Ditto.
35365 * gengtype.c (create_file): Ditto.
35366 * doc/cpp.texi: Bump @copying's copyright year.
35367 * doc/cppinternals.texi: Ditto.
35368 * doc/gcc.texi: Ditto.
35369 * doc/gccint.texi: Ditto.
35370 * doc/gcov.texi: Ditto.
35371 * doc/install.texi: Ditto.
35372 * doc/invoke.texi: Ditto.
35374 2021-01-01 Jakub Jelinek <jakub@redhat.com>
35376 * ChangeLog-2020: Rotate ChangeLog. New file.
35379 Copyright (C) 2021 Free Software Foundation, Inc.
35381 Copying and distribution of this file, with or without modification,
35382 are permitted in any medium without royalty provided the copyright
35383 notice and this notice are preserved.