1 2021-12-22 Murray Steele <murray.steele@arm.com>
3 * config.gcc (arm*-*-*): Add arm-mve-builtins.o to extra_objs.
4 * config/arm/arm-c.c (arm_pragma_arm): Handle "#pragma GCC arm".
5 (arm_register_target_pragmas): Register it.
6 * config/arm/arm-protos.h: (arm_mve::arm_handle_mve_types_h): New
8 * config/arm/arm_mve_types.h: Replace MVE type definitions with
10 * config/arm/t-arm: (arm-mve-builtins.o): New target rule.
11 * config/arm/arm-mve-builtins.cc: New file.
12 * config/arm/arm-mve-builtins.def: New file.
13 * config/arm/arm-mve-builtins.h: New file.
15 2021-12-22 Murray Steele <murray.steele@arm.com>
17 * config/arm/arm-builtins.c (enum arm_type_qualifiers): Move to
19 (enum arm_simd_type): Move to arm-builtins.h.
20 (struct arm_simd_type_info): Move to arm-builtins.h.
21 * config/arm/arm-builtins.h (enum arm_simd_type): Move from
23 (enum arm_type_qualifiers): Move from arm-builtins.c.
24 (struct arm_simd_type_info): Move from arm-builtins.c.
26 2021-12-22 Martin Liska <mliska@suse.cz>
28 * doc/extend.texi: Unify all function declarations in examples
29 where some miss trailing ';'.
31 2021-12-22 Martin Liska <mliska@suse.cz>
33 * doc/extend.texi: Unify all function declarations in examples
34 where some miss trailing ';'.
36 2021-12-22 Martin Liska <mliska@suse.cz>
38 * doc/extend.texi: Unify all function declarations in examples
39 where some miss trailing ';'.
41 2021-12-22 Martin Liska <mliska@suse.cz>
43 * doc/extend.texi: Use uppercase letters for SSEx.
45 2021-12-21 Jiang Haochen <haochen.jiang@intel.com>
47 * config/i386/bmiintrin.h (_tzcnt_u16): New intrinsic.
51 2021-12-21 Xionghu Luo <luoxhu@linux.ibm.com>
53 * tree-ssa-loop-im.c (bb_colder_than_loop_preheader): New
55 (get_coldest_out_loop): New function.
56 (determine_max_movement): Use get_coldest_out_loop.
57 (move_computations_worker): Adjust and fix iteration udpate.
58 (class ref_in_loop_hot_body): New functor.
59 (ref_in_loop_hot_body::operator): New.
60 (can_sm_ref_p): Use for_all_locs_in_loop.
61 (fill_coldest_and_hotter_out_loop): New.
62 (tree_ssa_lim_finalize): Free coldest_outermost_loop and
63 hotter_than_inner_loop.
64 (loop_invariant_motion_in_fun): Call fill_coldest_and_hotter_out_loop.
66 2021-12-21 Xionghu Luo <luoxhu@linux.ibm.com>
68 * tree-ssa-loop-split.c (split_loop): Fix incorrect
69 profile_count and probability.
70 (do_split_loop_on_cond): Likewise.
72 2021-12-21 Xionghu Luo <luoxhu@linux.ibm.com>
75 * predict.c (predict_extra_loop_exits): Add loop parameter.
76 (predict_loops): Call with loop argument.
78 2021-12-21 Xionghu Luo <luoxhu@linux.ibm.com>
80 * config/rs6000/altivec.md (altivec_vaddu<VI_char>s): Replace
81 UNSPEC_VADDU with us_plus.
82 (altivec_vadds<VI_char>s): Replace UNSPEC_VADDS with ss_plus.
83 (altivec_vsubu<VI_char>s): Replace UNSPEC_VSUBU with us_minus.
84 (altivec_vsubs<VI_char>s): Replace UNSPEC_VSUBS with ss_minus.
85 (altivec_abss_<mode>): Likewise.
87 2021-12-20 Uroš Bizjak <ubizjak@gmail.com>
90 * config/i386/sse.md (<sse2p4_1>_pinsr<ssemodesuffix>): Add
91 earlyclobber to (x,x,x,i) alternative.
92 (<sse2p4_1>_pinsr<ssemodesuffix> peephole2): Remove.
93 (<sse2p4_1>_pinsr<ssemodesuffix> splitter): Use output
94 operand as a temporary register. Split after reload_completed.
96 2021-12-20 Roger Sayle <roger@nextmovesoftware.com>
97 Uroš Bizjak <ubizjak@gmail.com>
99 * config/i386/i386.md (any_mul_highpart): New code iterator.
100 (sgnprefix, s): Add attribute support for [su]mul_highpart.
101 (<s>mul<mode>3_highpart): Delete expander.
102 (<s>mul<mode>3_highpart, <s>mulsi32_highpart_zext):
103 New define_insn patterns.
104 (define_peephole2): Tweak the register allocation for the above
105 instructions after reload.
107 2021-12-20 Joel Sherrill <joel@rtems.org>
109 * config.gcc: Obsolete m32c-*-rtems* target.
111 2021-12-20 Martin Liska <mliska@suse.cz>
113 * opts.c (default_options_optimization): Support -Oz in -Ox option hints.
115 2021-12-20 Jan Hubicka <hubicka@ucw.cz>
118 * ipa-modref.c (modref_eaf_analysis::analyze_ssa_name): Add deferred
120 (modref_eaf_analysis::propagate): Use it.
122 2021-12-20 liuhongt <hongtao.liu@intel.com>
125 * config/i386/sse.md (*bit_and_float_vector_all_ones): New
128 2021-12-19 Jan Hubicka <hubicka@ucw.cz>
131 * ipa-modref.c (modref_merge_call_site_flags): Fix early exit condition
133 2021-12-19 Patrick Palka <ppalka@redhat.com>
135 * print-tree.c (print_node) <case tcc_declaration>: Dump
138 2021-12-19 Andrew Pinski <apinski@marvell.com>
140 * tree-ssa-phiopt.c (gimple_simplify_phiopt): Annotate the
141 new sequence with the location of the conditional statement.
143 2021-12-18 Roger Sayle <roger@nextmovesoftware.com>
144 Uroš Bizjak <ubizjak@gmail.com>
147 * config/i386/i386.md (any_or_plus): New code iterator.
148 (define_split): Split (HI<<32)|zext(LO) into piece-wise
149 move instructions on !TARGET_64BIT.
150 * config/i386/sse.md (*vec_extractv4si_0_zext_sse4):
151 Restrict to TARGET_64BIT.
153 2021-12-18 Roger Sayle <roger@nextmovesoftware.com>
156 * common.opt (Oz): New command line option.
157 * doc/invoke.texi: Document the new -Oz option.
158 * lto-wrapper.c (merge_and_complain, append_compiler_options):
159 Treat OPT_Oz as synonymous with OPT_Os.
160 * optc-save-gen.awk: Increase maximum value of optimize_size to 2.
161 * opts.c (default_options_optimization) [OPT_Oz]: Handle OPT_Oz
162 just like OPT_Os, except set opt->x_optimize_size to 2.
163 (common_handle_option): Skip OPT_Oz just like OPT_Os.
164 * config/i386/i386.md (*movdi_internal): Use a push/pop sequence
165 for suitable SImode TYPE_IMOV moves when optimize_size > 1.
166 (*movsi_internal): Likewise.
168 2021-12-18 Siddhesh Poyarekar <siddhesh@gotplt.org>
170 PR tree-optimization/103759
171 * tree-object-size.c (unknown, initval): Remove functions.
172 (size_unknown, size_initval, size_unknown_p): Operate directly
175 2021-12-18 François-Xavier Coudert <fxcoudert@gcc.gnu.org>
177 * config/darwin-driver.c: Make version code more future-proof.
178 * config.gcc: Homogeneize darwin versions.
179 * configure.ac: Homogeneize darwin versions.
180 * configure: Regenerate.
182 2021-12-17 Marek Polacek <polacek@redhat.com>
185 * attribs.c (handle_ignored_attributes_option): Create the fake
186 attribute with max_length == -2.
187 (attribute_ignored_p): New overloads.
188 * attribs.h (attribute_ignored_p): Declare them.
189 * tree-core.h (struct attribute_spec): Document that max_length
192 2021-12-17 Segher Boessenkool <segher@kernel.crashing.org>
195 * config/rs6000/rs6000-builtins.def (__builtin_darn): Expand to
196 darn_64_di. Add {32bit} attribute. Return long.
197 (__builtin_darn_32): Expand to darn_32_di. Add {32bit} attribute.
199 (__builtin_darn_raw): Expand to darn_raw_di. Add {32bit} attribute.
201 * config/rs6000/rs6000-call.c (rs6000_expand_builtin): Expand the darn
202 builtins to the _si variants for -m32.
203 * config/rs6000/rs6000.md (UNSPECV_DARN_32, UNSPECV_DARN_RAW): Delete.
204 (UNSPECV_DARN): Update comment.
205 (darn_32, darn_raw, darn): Delete.
206 (darn_32_<mode>, darn_64_<mode>, darn_raw_<mode> for GPR): New.
207 (@darn<mode> for GPR): New.
209 2021-12-17 Bill Schmidt <wschmidt@linux.ibm.com>
211 * config/rs6000/rs6000-overload.def (__builtin_vec_promote): Add second
214 2021-12-17 Richard Sandiford <richard.sandiford@arm.com>
216 PR tree-optimization/103744
217 * tree-vect-stmts.c (vectorizable_load): Handle multi-vector
220 2021-12-17 Martin Liska <mliska@suse.cz>
222 * doc/invoke.texi: Rename to -fstack-protector.
224 2021-12-17 Martin Liska <mliska@suse.cz>
226 * doc/invoke.texi: Fix spelling issues.
228 2021-12-17 Tamar Christina <tamar.christina@arm.com>
230 PR tree-optimization/103741
231 * tree-vect-stmts.c (vectorizable_operation): Check for boolean.
233 2021-12-17 Iain Sandoe <iain@sandoe.co.uk>
235 * config/darwin-driver.c (darwin_driver_init): Exit from the
236 option handling early if the command line is definitely enpty.
237 * config/darwin.h (SUBTARGET_DRIVER_SELF_SPECS): Remove
238 setting for the default content of weak_reference_mismatches.
240 2021-12-17 Iain Sandoe <iain@sandoe.co.uk>
242 * config/rs6000/darwin.h: Drop trailing _x from the
243 builtin_decls array name.
245 2021-12-17 Haochen Jiang <haochen.jiang@intel.com>
248 * config/i386/sse.md (*avx_cmp<mode>3_lt, *avx_cmp<mode>3_ltint):
249 Remove MEM_P restriction and add force_reg for operands[2].
250 (*avx_cmp<mode>3_ltint_not): Add new define_insn_and_split.
252 2021-12-17 Siddhesh Poyarekar <siddhesh@gotplt.org>
254 * builtins.def (BUILT_IN_DYNAMIC_OBJECT_SIZE): New builtin.
255 * tree-object-size.h: Move object size type bits enum from
256 tree-object-size.c and add new value OST_DYNAMIC.
257 * builtins.c (expand_builtin, fold_builtin_2): Handle it.
258 (fold_builtin_object_size): Handle new builtin and adjust for
259 change to compute_builtin_object_size.
260 * tree-object-size.c: Include builtins.h.
261 (compute_builtin_object_size): Adjust.
262 (early_object_sizes_execute_one,
263 dynamic_object_sizes_execute_one): New functions.
264 (object_sizes_execute): Rename insert_min_max_p argument to
265 early. Handle BUILT_IN_DYNAMIC_OBJECT_SIZE and call the new
267 * doc/extend.texi (__builtin_dynamic_object_size): Document new
270 2021-12-17 Siddhesh Poyarekar <siddhesh@gotplt.org>
272 * tree-object-size.h (compute_builtin_object_size): Return tree
273 instead of HOST_WIDE_INT.
274 * builtins.c (fold_builtin_object_size): Adjust.
275 * gimple-fold.c (gimple_fold_builtin_strncat): Likewise.
276 * ubsan.c (instrument_object_size): Likewise.
277 * tree-object-size.c (object_size): New structure.
278 (object_sizes): Change type to vec<object_size>.
279 (initval): New function.
281 (size_unknown_p, size_initval, size_unknown): New functions.
282 (object_sizes_unknown_p): Use it.
283 (object_sizes_get): Return tree.
284 (object_sizes_initialize): Rename from object_sizes_set_force
285 and set VAL parameter type as tree. Add new parameter WHOLEVAL.
286 (object_sizes_set): Set VAL parameter type as tree and adjust
287 implementation. Add new parameter WHOLEVAL.
288 (size_for_offset): New function.
289 (decl_init_size): Adjust comment.
290 (addr_object_size): Change PSIZE parameter to tree and adjust
291 implementation. Add new parameter PWHOLESIZE.
292 (alloc_object_size): Return tree.
293 (compute_builtin_object_size): Return tree in PSIZE.
294 (expr_object_size, call_object_size, unknown_object_size):
295 Adjust for object_sizes_set change.
296 (merge_object_sizes): Drop OFFSET parameter and adjust
297 implementation for tree change.
298 (plus_stmt_object_size): Call collect_object_sizes_for directly
299 instead of merge_object_size and call size_for_offset to get net
301 (cond_expr_object_size, collect_object_sizes_for,
302 object_sizes_execute): Adjust for change of type from
303 HOST_WIDE_INT to tree.
304 (check_for_plus_in_loops_1): Likewise and skip non-positive
307 2021-12-17 Jason Merrill <jason@redhat.com>
310 * common.opt (fabi-version): Add v17.
312 2021-12-16 Uroš Bizjak <ubizjak@gmail.com>
315 * config/i386/i386.h (VALID_AVX256_REG_MODE): Add V16HFmode.
316 (VALID_AVX256_REG_OR_OI_VHF_MODE): Replace with ...
317 (VALID_AVX256_REG_OR_OI_MODE): ... this. Remove V16HFmode.
318 (VALID_AVX512F_SCALAR_MODE): Remove HImode and HFmode.
319 (VALID_AVX512FP16_SCALAR_MODE): New.
320 (VALID_AVX512F_REG_MODE): Add V32HFmode.
321 (VALID_SSE2_REG_MODE): Add V8HFmode, V4HFmode and V2HFmode.
322 (VALID_SSE2_REG_VHF_MODE): Remove.
323 (VALID_INT_MODE_P): Add V2HFmode.
324 * config/i386/i386.c (function_arg_advance_64):
325 Remove explicit mention of V16HFmode and V32HFmode.
326 (ix86_hard_regno_mode_ok): Remove explicit mention of XImode
327 and V32HFmode, use VALID_AVX512F_REG_OR_XI_MODE instead.
328 Use VALID_AVX512FP_SCALAR_MODE for TARGET_aVX512FP16.
329 Use VALID_AVX256_REG_OR_OI_MODE instead of
330 VALID_AVX256_REG_OR_OI_VHF_MODE and VALID_SSE2_REG_MODE instead
331 of VALID_SSE2_REG_VHF_MODE.
332 (ix86_set_reg_reg_cost): Remove usge of VALID_AVX512FP16_REG_MODE.
333 (ix86_vector_mode_supported): Ditto.
335 2021-12-16 Martin Liska <mliska@suse.cz>
338 * attribs.c (decl_attributes): Check if
339 target_option_current_node is changed.
341 2021-12-16 Martin Liska <mliska@suse.cz>
343 * doc/invoke.texi: Add missing dash.
345 2021-12-16 Bill Schmidt <wschmidt@linux.ibm.com>
347 * config/rs6000/rs6000-c.c (resolution): New enum.
348 (resolve_vec_mul): New function.
349 (resolve_vec_cmpne): Likewise.
350 (resolve_vec_adde_sube): Likewise.
351 (resolve_vec_addec_subec): Likewise.
352 (resolve_vec_splats): Likewise.
353 (resolve_vec_extract): Likewise.
354 (resolve_vec_insert): Likewise.
355 (resolve_vec_step): Likewise.
356 (find_instance): Likewise.
357 (altivec_resolve_overloaded_builtin): Many cleanups. Call factored-out
358 functions. Move variable declarations closer to uses. Add commentary.
359 Remove unnecessary levels of braces. Avoid use of gotos. Change
360 misleading variable names. Use switches over if-else-if chains.
362 2021-12-16 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
365 * config/aarch64/aarch64-simd.md (aarch64_movv8di): Allow big endian
366 targets to move V8DI.
368 2021-12-16 Tobias Burnus <tobias@codesourcery.com>
371 * gcc.c (check_offload_target_name): Add 'default' and 'disable'
372 to the candidate list.
374 2021-12-15 Iain Sandoe <iain@sandoe.co.uk>
376 * configure: Regenerate.
378 2021-12-15 Roger Sayle <roger@nextmovesoftware.com>
380 * config/nvptx/nvptx-opts.h (ptx_isa): PTX_ISA_SM75 and PTX_ISA_SM80
382 * config/nvptx/nvptx.opt: Add sm_75 and sm_80 to -misa.
383 * config/nvptx/nvptx.h (TARGET_SM75, TARGET_SM80):
384 New helper macros to conditionalize functionality on target ISA.
385 * config/nvptx/nvptx-c.c (nvptx_cpu_cpp_builtins): Add __PTX_SM__
386 support for the new ISA levels.
387 * config/nvptx/nvptx.c (nvptx_file_start): Add support for TARGET_SM75
389 * config/nvptx/nvptx.md (define_c_enum "unspec"): New UNSPEC_TANH.
390 (define_mode_iterator HSFM): New iterator for HFmode and SFmode.
391 (exp2hf2): New define_insn controlled by TARGET_SM75.
392 (tanh<mode>2): New define_insn controlled by TARGET_SM75.
393 (sminhf3, smaxhf3): New define_isnns controlled by TARGET_SM80.
395 2021-12-15 Tom de Vries <tdevries@suse.de>
397 * config/nvptx/nvptx-opts.h (enum ptx_version): Add PTX_VERSION_7_0.
398 * config/nvptx/nvptx.c (nvptx_file_start): Handle TARGET_PTX_7_0.
399 * config/nvptx/nvptx.h (TARGET_PTX_7_0): New macro.
400 * config/nvptx/nvptx.opt (ptx_version): Add 7.0.
402 2021-12-15 Richard Sandiford <richard.sandiford@arm.com>
403 Tamar Christina <tamar.christina@arm.com>
406 * config/aarch64/aarch64.c (aarch64_short_vector_p): Return false
407 for structure modes, rather than ignoring the type in that case.
409 2021-12-15 Tamar Christina <tamar.christina@arm.com>
411 PR rtl-optimization/103350
412 * ree.c (add_removable_extension): Don't stop at first definition but
415 2021-12-15 Jakub Jelinek <jakub@redhat.com>
418 * dwarf2cfi.c (dwf_cfa_reg): Remove gcc_assert.
419 (operator==, operator!=): New overloaded operators.
420 (dwarf2out_frame_debug_adjust_cfa, dwarf2out_frame_debug_cfa_offset,
421 dwarf2out_frame_debug_expr): Compare vars with cfa_reg type directly
422 with REG rtxes rather than with dwf_cfa_reg results on those REGs.
423 (create_cie_data): Use stack_pointer_rtx instead of
424 gen_rtx_REG (Pmode, STACK_POINTER_REGNUM).
425 (execute_dwarf2_frame): Use hard_frame_pointer_rtx instead of
426 gen_rtx_REG (Pmode, HARD_FRAME_POINTER_REGNUM).
428 2021-12-15 Martin Liska <mliska@suse.cz>
431 * config/i386/i386-builtins.c (fold_builtin_cpu): Compare to 0
432 as API expects that non-zero values are returned (do that
434 For "avx512vbmi2" argument, we return now 1 << 31, which is a
435 negative integer value.
437 2021-12-15 Haochen Jiang <haochen.jiang@intel.com>
440 * config/i386/predicates.md (const_vector_operand):
442 * config/i386/sse.md(<insn><mode>3<mask_name>):
443 Add new define_split below.
445 2021-12-15 Michael Meissner <meissner@the-meissners.org>
447 * config/rs6000/rs6000.md (UNSPEC_XXSPLTIDP_CONST): New unspec.
448 (UNSPEC_XXSPLTIW_CONST): New unspec.
449 (movsf_hardfloat): Add support for generating XXSPLTIDP.
450 (mov<mode>_hardfloat32): Likewise.
451 (mov<mode>_hardfloat64): Likewise.
452 (xxspltidp_<mode>_internal): New insns.
453 (xxspltiw_<mode>_internal): New insns.
454 (splitters for SF/DFmode): Add new splitters for XXSPLTIDP.
456 2021-12-15 Michael Meissner <meissner@the-meissners.org>
458 * config/rs6000/predicates.md (easy_fp_constant): Add support for
459 generating XXSPLTIDP.
460 (vsx_prefixed_constant): Likewise.
461 (easy_vector_constant): Likewise.
462 * config/rs6000/rs6000-protos.h (constant_generates_xxspltidp):
464 * config/rs6000/rs6000.c (output_vec_const_move): Add support for
465 generating XXSPLTIDP.
466 (prefixed_xxsplti_p): Likewise.
467 (constant_generates_xxspltidp): New function.
468 * config/rs6000/rs6000.opt (-msplat-float-constant): New debug option.
470 2021-12-15 Michael Meissner <meissner@linux.ibm.com>
472 * config/rs6000/constraints.md (eP): Update comment.
473 * config/rs6000/predicates.md (easy_fp_constant): Add support for
475 (vsx_prefixed_constant): New predicate.
476 (easy_vector_constant): Add support for
478 * config/rs6000/rs6000-protos.h (prefixed_xxsplti_p): New
480 (constant_generates_xxspltiw): Likewise.
481 * config/rs6000/rs6000.c (xxspltib_constant_p): Generate XXSPLTIW
482 if possible instead of XXSPLTIB and sign extending the constant.
483 (output_vec_const_move): Add support for XXSPLTIW.
484 (prefixed_xxsplti_p): New function.
485 (constant_generates_xxspltiw): New function.
486 * config/rs6000/rs6000.md (prefixed attribute): Add support to
487 mark XXSPLTI* instructions as being prefixed.
488 * config/rs6000/rs6000.opt (-msplat-word-constant): New debug
490 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Add support for
491 generating XXSPLTIW or XXSPLTIDP.
492 (vsx_mov<mode>_32bit): Likewise.
493 * doc/md.texi (PowerPC and IBM RS6000 constraints): Document the
496 2021-12-15 Michael Meissner <meissner@the-meissners.org>
498 * config/rs6000/constraints.md (eQ): New constraint.
499 * config/rs6000/predicates.md (easy_fp_constant): Add support for
500 generating the LXVKQ instruction.
501 (easy_vector_constant_ieee128): New predicate.
502 (easy_vector_constant): Add support for generating the LXVKQ
504 * config/rs6000/rs6000-protos.h (constant_generates_lxvkq): New
506 * config/rs6000/rs6000.c (output_vec_const_move): Add support for
508 (constant_generates_lxvkq): New function.
509 * config/rs6000/rs6000.opt (-mieee128-constant): New debug
511 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Add support for
513 (vsx_mov<mode>_32bit): Likewise.
514 * doc/md.texi (PowerPC and IBM RS6000 constraints): Document the
517 2021-12-15 Michael Meissner <meissner@the-meissners.org>
519 * config/rs6000/rs6000-protos.h (VECTOR_128BIT_BITS): New macro.
520 (VECTOR_128BIT_BYTES): Likewise.
521 (VECTOR_128BIT_HALF_WORDS): Likewise.
522 (VECTOR_128BIT_WORDS): Likewise.
523 (VECTOR_128BIT_DOUBLE_WORDS): Likewise.
524 (vec_const_128bit_type): New structure type.
525 (vec_const_128bit_to_bytes): New declaration.
526 * config/rs6000/rs6000.c (constant_int_to_128bit_vector): New
528 (constant_fp_to_128bit_vector): New helper function.
529 (vec_const_128bit_to_bytes): New function.
531 2021-12-15 Alexandre Oliva <oliva@adacore.com>
534 * builtins.c (try_store_by_multiple_pieces): Drop address
535 conversion to ptr_mode.
537 2021-12-15 Alexandre Oliva <oliva@adacore.com>
540 * builtins.c (try_store_by_multiple_pieces): Fail if min_len
541 is greater than max_len.
543 2021-12-14 liuhongt <hongtao.liu@intel.com>
546 * tree-ssa-ccp.c (optimize_atomic_bit_test_and): Check
547 is_gimple_assign before gimple_assign_rhs_code.
549 2021-12-14 Sören Tempel <soeren@soeren-tempel.net>
551 * ginclude/stddef.h (__DEFINED_ptrdiff_t): Add support for musl
552 libc typedef macro guard.
553 (__DEFINED_size_t): Ditto.
554 (__DEFINED_wchar_t): Ditto.
556 2021-12-14 JoJo R <rjiejie@linux.alibaba.com>
558 * regrename.c (find_rename_reg): Return satisfied regno
559 if instruction is noop move.
561 2021-12-14 Bill Schmidt <wschmidt@linux.ibm.com>
563 * config/rs6000/rs6000-c.c (altivec_build_resolved_builtin): Rename
564 rs6000_builtin_decls_x to rs6000_builtin_decls.
565 (altivec_resolve_overloaded_builtin): Likewise. Also rename
566 rs6000_builtin_info_x to rs6000_builtin_info.
567 * config/rs6000/rs6000-call.c (rs6000_invalid_builtin): Rename
568 rs6000_builtin_info_x to rs6000_builtin_info.
569 (rs6000_builtin_is_supported): Likewise.
570 (rs6000_gimple_fold_mma_builtin): Likewise. Also rename
571 rs6000_builtin_decls_x to rs6000_builtin_decls.
572 (rs6000_gimple_fold_builtin): Rename rs6000_builtin_info_x to
574 (cpu_expand_builtin): Likewise.
575 (rs6000_expand_builtin): Likewise.
576 (rs6000_init_builtins): Likewise. Also rename rs6000_builtin_decls_x
577 to rs6000_builtin_decls.
578 (rs6000_builtin_decl): Rename rs6000_builtin_decls_x to
579 rs6000_builtin_decls.
580 * config/rs6000/rs6000-gen-builtins.c (write_decls): In generated code,
581 rename rs6000_builtin_decls_x to rs6000_builtin_decls, and rename
582 rs6000_builtin_info_x to rs6000_builtin_info.
583 (write_bif_static_init): In generated code, rename
584 rs6000_builtin_info_x to rs6000_builtin_info.
585 (write_init_bif_table): In generated code, rename
586 rs6000_builtin_decls_x to rs6000_builtin_decls, and rename
587 rs6000_builtin_info_x to rs6000_builtin_info.
588 (write_init_ovld_table): In generated code, rename
589 rs6000_builtin_decls_x to rs6000_builtin_decls.
590 (write_init_file): Likewise.
591 * config/rs6000/rs6000.c (rs6000_builtin_vectorized_function):
593 (rs6000_builtin_md_vectorized_function): Likewise.
594 (rs6000_builtin_reciprocal): Likewise.
595 (add_condition_to_bb): Likewise.
596 (rs6000_atomic_assign_expand_fenv): Likewise.
598 2021-12-14 Bill Schmidt <wschmidt@linux.ibm.com>
600 * config/rs6000/rs6000-c.c (altivec_resolve_new_overloaded_builtin):
601 Remove forward declaration.
602 (rs6000_new_builtin_type_compatible): Rename to
603 rs6000_builtin_type_compatible.
604 (rs6000_builtin_type_compatible): Remove.
605 (altivec_resolve_overloaded_builtin): Remove.
606 (altivec_build_new_resolved_builtin): Rename to
607 altivec_build_resolved_builtin.
608 (altivec_resolve_new_overloaded_builtin): Rename to
609 altivec_resolve_overloaded_builtin. Remove static keyword. Adjust
610 called function names.
611 * config/rs6000/rs6000-call.c (rs6000_expand_new_builtin): Remove
613 (rs6000_gimple_fold_new_builtin): Likewise.
614 (rs6000_invalid_new_builtin): Rename to rs6000_invalid_builtin.
615 (rs6000_gimple_fold_builtin): Remove.
616 (rs6000_new_builtin_valid_without_lhs): Rename to
617 rs6000_builtin_valid_without_lhs.
618 (rs6000_new_builtin_is_supported): Rename to
619 rs6000_builtin_is_supported.
620 (rs6000_gimple_fold_new_mma_builtin): Rename to
621 rs6000_gimple_fold_mma_builtin.
622 (rs6000_gimple_fold_new_builtin): Rename to
623 rs6000_gimple_fold_builtin. Remove static keyword. Adjust called
625 (rs6000_expand_builtin): Remove.
626 (new_cpu_expand_builtin): Rename to cpu_expand_builtin.
627 (new_mma_expand_builtin): Rename to mma_expand_builtin.
628 (new_htm_spr_num): Rename to htm_spr_num.
629 (new_htm_expand_builtin): Rename to htm_expand_builtin. Change name
631 (rs6000_expand_new_builtin): Rename to rs6000_expand_builtin. Remove
632 static keyword. Adjust called function names.
633 (rs6000_new_builtin_decl): Rename to rs6000_builtin_decl. Remove
635 (rs6000_builtin_decl): Remove.
636 * config/rs6000/rs6000-gen-builtins.c (write_decls): In gnerated code,
637 rename rs6000_new_builtin_is_supported to rs6000_builtin_is_supported.
638 * config/rs6000/rs6000-internal.h (rs6000_invalid_new_builtin): Rename
639 to rs6000_invalid_builtin.
640 * config/rs6000/rs6000.c (rs6000_new_builtin_vectorized_function):
641 Rename to rs6000_builtin_vectorized_function.
642 (rs6000_new_builtin_md_vectorized_function): Rename to
643 rs6000_builtin_md_vectorized_function.
644 (rs6000_builtin_vectorized_function): Remove.
645 (rs6000_builtin_md_vectorized_function): Remove.
647 2021-12-14 Bill Schmidt <wschmidt@linux.ibm.com>
649 * config/rs6000/rs6000-builtin.def: Delete.
650 * config/rs6000/rs6000-call.c (builtin_compatibility): Delete.
651 (builtin_description): Delete.
652 (builtin_hash_struct): Delete.
653 (builtin_hasher): Delete.
654 (builtin_hash_table): Delete.
655 (builtin_hasher::hash): Delete.
656 (builtin_hasher::equal): Delete.
657 (rs6000_builtin_info_type): Delete.
658 (rs6000_builtin_info): Delete.
659 (bdesc_compat): Delete.
660 (bdesc_3arg): Delete.
661 (bdesc_4arg): Delete.
663 (bdesc_2arg): Delete.
664 (bdesc_altivec_preds): Delete.
666 (bdesc_1arg): Delete.
667 (bdesc_0arg): Delete.
670 (rs6000_overloaded_builtin_p): Delete.
671 (rs6000_overloaded_builtin_name): Delete.
672 (htm_spr_num): Delete.
673 (rs6000_builtin_is_supported_p): Delete.
674 (rs6000_gimple_fold_mma_builtin): Delete.
675 (gt-rs6000-call.h): Remove include directive.
676 * config/rs6000/rs6000-protos.h (rs6000_overloaded_builtin_p): Delete.
677 (rs6000_builtin_is_supported_p): Delete.
678 (rs6000_overloaded_builtin_name): Delete.
679 * config/rs6000/rs6000.c (rs6000_builtin_decls): Delete.
680 (rs6000_debug_reg_global): Remove reference to RS6000_BUILTIN_COUNT.
681 * config/rs6000/rs6000.h (rs6000_builtins): Delete.
682 (altivec_builtin_types): Delete.
683 (rs6000_builtin_decls): Delete.
684 * config/rs6000/t-rs6000 (TM_H): Don't add rs6000-builtin.def.
686 2021-12-14 Bill Schmidt <wschmidt@linux.ibm.com>
688 * config/rs6000/rs6000-builtin-new.def: Rename to...
689 * config/rs6000/rs6000-builtins.def: ...this.
690 * config/rs6000/rs6000-gen-builtins.c: Adjust header commentary.
691 * config/rs6000/t-rs6000 (EXTRA_GTYPE_DEPS): Rename
692 rs6000-builtin-new.def to rs6000-builtins.def.
693 (rs6000-builtins.c): Likewise.
695 2021-12-14 Bill Schmidt <wschmidt@linux.ibm.com>
697 * config/rs6000/rs6000-call.c (altivec_overloaded_builtins): Remove.
698 * config/rs6000/rs6000.h (altivec_overloaded_builtins): Remove.
700 2021-12-14 Peter Bergner <bergner@linux.ibm.com>
703 * config/rs6000/mma.md (UNSPEC_MMA_ASSEMBLE): Rename unspec from this...
704 (UNSPEC_VSX_ASSEMBLE): ...to this.
705 (UNSPECV_MMA_ASSEMBLE): New unspecv.
706 (vsx_assemble_pair): Use UNSPEC_VSX_ASSEMBLE.
707 (*vsx_assemble_pair): Likewise.
708 (mma_assemble_acc): Use UNSPECV_MMA_ASSEMBLE.
709 (*mma_assemble_acc): Likewise.
710 * config/rs6000/rs6000.c (rs6000_split_multireg_move): Handle
711 UNSPEC_VOLATILE. Use UNSPEC_VSX_ASSEMBLE and UNSPECV_MMA_ASSEMBLE.
713 2021-12-14 Uroš Bizjak <ubizjak@gmail.com>
716 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate)
717 <case E_V8HFmode>: Implement for TARGET_SSE2.
718 <case E_V16HFmode>: Implement for TARGET_AVX.
719 <case E_V32HFmode>: Implement for TARGET_AVX512F.
720 (ix86_expand_vector_set_var): Handle V32HFmode
721 without TARGET_AVX512BW.
722 (ix86_expand_vector_extract)
723 <case E_V8HFmode>: Implement for TARGET_SSE2.
724 <case E_V16HFmode>: Implement for TARGET_AVX.
725 <case E_V32HFmode>: Implement for TARGET_AVX512BW.
726 (expand_vec_perm_broadcast_1) <case E_V8HFmode>: New.
727 * config/i386/sse.md (VI12HF_AVX512VL): Remove
728 TARGET_AVX512FP16 condition.
731 (avx_vbroadcastf128_<mode>): Use V_256H mode iterator.
733 2021-12-14 Bill Schmidt <wschmidt@linux.ibm.com>
735 * config/rs6000/darwin.h (SUBTARGET_INIT_BUILTINS): Remove
736 test for new_builtins_are_live and simplify.
737 * config/rs6000/rs6000-c.c (altivec_build_resolved_builtin): Remove
739 (altivec_resolve_overloaded_builtin): Remove test for
740 new_builtins_are_live and simplify.
741 * config/rs6000/rs6000-call.c (altivec_init_builtins): Remove forward
743 (builtin_function_type): Likewise.
744 (rs6000_common_init_builtins): Likewise.
745 (htm_init_builtins): Likewise.
746 (mma_init_builtins): Likewise.
747 (def_builtin): Remove dead function.
748 (rs6000_expand_zeroop_builtin): Likewise.
749 (rs6000_expand_mtfsf_builtin): Likewise.
750 (rs6000_expand_mtfsb_builtin): Likewise.
751 (rs6000_expand_set_fpscr_rn_builtin): Likewise.
752 (rs6000_expand_set_fpscr_drn_builtin): Likewise.
753 (rs6000_expand_unop_builtin): Likewise.
754 (altivec_expand_abs_builtin): Likewise.
755 (rs6000_expand_binop_builtin): Likewise.
756 (altivec_expand_lxvr_builtin): Likewise.
757 (altivec_expand_lv_builtin): Likewise.
758 (altivec_expand_stxvl_builtin): Likewise.
759 (altivec_expand_stv_builtin): Likewise.
760 (mma_expand_builtin): Likewise.
761 (htm_expand_builtin): Likewise.
762 (cpu_expand_builtin): Likewise.
763 (rs6000_expand_quaternop_builtin): Likewise.
764 (rs6000_expand_ternop_builtin): Likewise.
765 (altivec_expand_dst_builtin): Likewise.
766 (altivec_expand_vec_sel_builtin): Likewise.
767 (altivec_expand_builtin): Likewise.
768 (rs6000_invalid_builtin): Likewise.
769 (rs6000_builtin_valid_without_lhs): Likewise.
770 (rs6000_gimple_fold_builtin): Remove test for new_builtins_are_live and
772 (rs6000_expand_builtin): Likewise.
773 (rs6000_init_builtins): Remove tests for new_builtins_are_live and
775 (rs6000_builtin_decl): Likewise.
776 (altivec_init_builtins): Remove dead function.
777 (mma_init_builtins): Likewise.
778 (htm_init_builtins): Likewise.
779 (builtin_quaternary_function_type): Likewise.
780 (builtin_function_type): Likewise.
781 (rs6000_common_init_builtins): Likewise.
782 * config/rs6000/rs6000-gen-builtins.c (write_header_file): Don't
783 declare new_builtins_are_live.
784 (write_init_bif_table): In generated code, remove test for
785 new_builtins_are_live and simplify.
786 (write_init_ovld_table): Likewise.
787 (write_init_file): Don't initialize new_builtins_are_live.
788 * config/rs6000/rs6000.c (rs6000_builtin_vectorized_function): Remove
789 test for new_builtins_are_live and simplify.
790 (rs6000_builtin_md_vectorized_function): Likewise.
791 (rs6000_builtin_reciprocal): Likewise.
792 (add_condition_to_bb): Likewise.
793 (rs6000_atomic_assign_expand_fenv): Likewise.
795 2021-12-14 Bill Schmidt <wschmidt@linux.ibm.com>
798 * config/rs6000/rs6000-builtin-new.def (__builtin_altivec_vcmpequd):
799 Move to power8-vector stanza.
800 (__builtin_altivec_vcmpequd_p): Likewise.
801 (__builtin_altivec_vcmpgtsd): Likewise.
802 (__builtin_altivec_vcmpgtsd_p): Likewise.
803 (__builtin_altivec_vcmpgtud): Likewise.
804 (__builtin_altivec_vcmpgtud_p): Likewise.
806 2021-12-14 Bill Schmidt <wschmidt@linux.ibm.com>
809 * config/rs6000/rs6000-builtin-new.def (__builtin_pack_longdouble): Add
811 (__builtin_unpack_longdouble): Likewise.
812 * config/rs6000/rs6000-call.c (rs6000_expand_new_builtin): Add special
813 handling for ibmld attribute.
814 * config/rs6000/rs6000-gen-builtins.c (attrinfo): Add isibmld.
815 (parse_bif_attrs): Handle ibmld.
816 (write_decls): Likewise.
817 (write_bif_static_init): Likewise.
819 2021-12-14 Jan Hubicka <hubicka@ucw.cz>
822 * ipa-modref-tree.c (modref_access_node::range_info_useful_p): Handle
823 MODREF_GLOBAL_MEMORY_PARM.
824 (modref_access_node::dump): Likewise.
825 (modref_access_node::get_call_arg): Likewise.
826 * ipa-modref-tree.h (enum modref_special_parms): Add
827 MODREF_GLOBAL_MEMORY_PARM.
828 (modref_access_node::useful_for_kill): Handle
829 MODREF_GLOBAL_MEMORY_PARM.
830 (modref:tree::merge): Add promote_unknown_to_global.
831 * ipa-modref.c (verify_arg):New function.
832 (may_access_nonescaping_parm_p): New function.
833 (modref_access_analysis::record_global_memory_load): New member
835 (modref_access_analysis::record_global_memory_store): Likewise.
836 (modref_access_analysis::process_fnspec): Distingush global and local
838 (modref_access_analysis::analyze_call): Likewise.
839 * tree-ssa-alias.c (ref_may_access_global_memory_p): New function.
840 (modref_may_conflict): Use it.
842 2021-12-14 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
844 * config/aarch64/aarch64-builtins.c (enum aarch64_builtins):
845 Define AARCH64_LS64_BUILTIN_LD64B, AARCH64_LS64_BUILTIN_ST64B,
846 AARCH64_LS64_BUILTIN_ST64BV, AARCH64_LS64_BUILTIN_ST64BV0.
847 (aarch64_init_ls64_builtin_decl): Helper function.
848 (aarch64_init_ls64_builtins): Helper function.
849 (aarch64_init_ls64_builtins_types): Helper function.
850 (aarch64_general_init_builtins): Init LS64 intrisics for
852 (aarch64_expand_builtin_ls64): LS64 intrinsics expander.
853 (aarch64_general_expand_builtin): Handle aarch64_expand_builtin_ls64.
854 (ls64_builtins_data): New helper struct.
855 (v8di_UP): New define.
856 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Define
858 * config/aarch64/aarch64.c (aarch64_classify_address): Enforce the
859 V8DI range (7-bit signed scaled) for both ends of the range.
860 * config/aarch64/aarch64-simd.md (movv8di): New pattern.
861 (aarch64_movv8di): New pattern.
862 * config/aarch64/aarch64.h (AARCH64_ISA_LS64): New define.
863 (TARGET_LS64): New define.
864 * config/aarch64/aarch64.md: Add UNSPEC_LD64B, UNSPEC_ST64B,
865 UNSPEC_ST64BV and UNSPEC_ST64BV0.
866 (ld64b): New define_insn.
867 (st64b): New define_insn.
868 (st64bv): New define_insn.
869 (st64bv0): New define_insn.
870 * config/aarch64/arm_acle.h (data512_t): New type derived from
872 (__arm_data512_t): New internal type.
873 (__arm_ld64b): New intrinsic.
874 (__arm_st64b): New intrinsic.
875 (__arm_st64bv): New intrinsic.
876 (__arm_st64bv0): New intrinsic.
877 * config/arm/types.md: Add new type ls64.
879 2021-12-14 Olivier Hainque <hainque@adacore.com>
881 * config/i386/t-vxworks: Drop the fPIC multilibs.
883 2021-12-14 Fred Konrad <konrad@adacore.com>
885 * config/rs6000/t-vxworks: Drop the fPIC multilib.
887 2021-12-13 Vladimir N. Makarov <vmakarov@redhat.com>
890 * ira-costs.c (record_operand_costs): Do not take pseudo class
891 calculated on the 1st iteration into account when processing move
892 involving the pseudo and a hard register.
894 2021-12-13 Roger Sayle <roger@nextmovesoftware.com>
896 * config/i386/i386.md (define_split any_or:SWI248 -> orb %?h):
897 Optimize the case where the integer constant operand is zero.
899 2021-12-13 Doug Rupp <rupp@adacore.com>
900 Olivier Hainque <hainque@adacore.com>
902 * config/vxworks.h (VXWORKS_LINK_OS_SPEC): New spec.
903 (VXWORKS_BASE_LINK_SPEC): New spec, using the former.
904 (VXWORKS_EXTRA_LINK_SPEC): New spec for old and new VxWorks.
905 (VXWORKS_LINK_SPEC): Combo of BASE and EXTRA specs.
906 * config/rs6000/vxworks.h (VXWORKS_LINK_OS_SPEC): Empty.
907 (LINK_OS_EXTRA_SPEC32): Use VXWORKS_LINK_SPEC.
908 (LINK_OS_EXTRA_SPEC64): Likewise.
910 2021-12-13 Fred Konrad <konrad@adacore.com>
911 Olivier Hainque <hainque@adacore.com>
913 * config/vxworks.h (VXWORKS_BASE_LIBS_RTP): Guard -lc_internal
914 on !shared+!non-static and document.
915 (VXWORKS_LIB_SPEC): Remove the bits intended to drag the
916 init/fini functions from libc_internal in the shared lib case.
917 (VX_CRTBEGIN_SPEC/VX_CRTEND_SPEC): Use vxcrtstuff objects also in
918 configurations with shared lib and INITFINI_ARRAY support.
920 2021-12-13 Fred Konrad <konrad@adacore.com>
921 Olivier Hainque <hainque@adacore.com>
923 * config/vx-common.h: Define REAL_LIBGCC_SPEC since the
924 '-non-static' option is not standard.
925 * config/vxworks.h (VXWORKS_LIBGCC_SPEC): Implement the LIBGCC_SPEC
926 since REAL_LIBGCC_SPEC is used now.
927 (STARTFILE_PREFIX_SPEC): Use the PIC VSB when building shared libraries
928 or non-static binaries.
930 2021-12-13 Jan Hubicka <hubicka@ucw.cz>
932 * common.opt: Add -fipa-strict-aliasing.
933 * doc/invoke.texi: Document -fipa-strict-aliasing.
934 * ipa-modref.c (modref_access_analysis::record_access): Honor
935 -fipa-strict-aliasing.
936 (modref_access_analysis::record_access_lto): Likewise.
938 2021-12-13 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
940 * config/aarch64/aarch64-arches.def (armv8.8-a): Define.
941 * config/aarch64/aarch64.h (AARCH64_FL_V8_8): Define.
942 (AARCH64_FL_FOR_ARCH8_8): Define.
943 * doc/invoke.texi: Document -march=armv8.8-a.
945 2021-12-13 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
947 * config/aarch64/aarch64.c (aarch64_expand_setmem_mops): Define.
948 (aarch64_expand_setmem): Adjust for TARGET_MOPS.
949 * config/aarch64/aarch64.h (CLEAR_RATIO): Adjust for TARGET_MOPS.
950 (SET_RATIO): Likewise.
951 * config/aarch64/aarch64.md ("unspec"): Add UNSPEC_SETMEM.
952 (aarch64_setmemdi): Define.
953 (setmemdi): Adjust for TARGET_MOPS.
954 * config/aarch64/aarch64.opt (aarch64-mops-memset-size-threshold):
957 2021-12-13 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
959 * config/aarch64/aarch64.md (aarch64_movmemdi): Define.
961 (unspec): Add UNSPEC_MOVMEM.
962 * config/aarch64/aarch64.opt (aarch64-mops-memmove-size-threshold):
965 2021-12-13 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
967 * config/aarch64/aarch64-option-extensions.def (mops): Define.
968 * config/aarch64/aarch64.c (aarch64_expand_cpymem_mops): Define.
969 (aarch64_expand_cpymem): Define.
970 * config/aarch64/aarch64.h (AARCH64_FL_MOPS): Define.
971 (AARCH64_ISA_MOPS): Define.
972 (TARGET_MOPS): Define.
973 (MOVE_RATIO): Adjust for TARGET_MOPS.
974 * config/aarch64/aarch64.md ("unspec"): Add UNSPEC_CPYMEM.
975 (aarch64_cpymemdi): New pattern.
976 (cpymemdi): Adjust for TARGET_MOPS.
977 * config/aarch64/aarch64.opt (aarch64-mops-memcpy-size-threshol):
979 * doc/invoke.texi (AArch64 Options): Document +mops.
981 2021-12-13 Martin Liska <mliska@suse.cz>
984 * ipa-inline.c (can_inline_edge_p): Move logic checking
985 no_profile_instrument_function logic to ...
986 (can_early_inline_edge_p): ... here.
988 2021-12-13 Olivier Hainque <hainque@adacore.com>
990 * config/vxworks/_yvals.h: #include yvals.h also if
993 2021-12-13 Olivier Hainque <hainque@adacore.com>
995 * config/vxworks.h (VXWORKS_OS_CPP_BUILTINS): Define
998 2021-12-13 Olivier Hainque <hainque@adacore.com>
1000 * config/t-vxworks: Clear NATIVE_SYSTEM_HEADER_DIR.
1001 * config/vxworks.h (SYSROOT_HEADERS_SUFFIX_SPEC): Define, for
1002 VxWorks 7 and earlier.
1003 (VXWORKS_ADDITIONAL_CPP_SPEC): Simplify accordingly.
1004 (STARTFILE_PREFIX_SPEC): Adjust accordingly.
1005 * config/rs6000/vxworks.h (STARTFILE_PREFIX_SPEC): Adjust.
1007 2021-12-13 Martin Liska <mliska@suse.cz>
1009 * doc/extend.texi: Use @item for the first @itemx entry.
1011 2021-12-13 Jakub Jelinek <jakub@redhat.com>
1013 * machmode.h (gt_pch_nx): Use gt_pointer_operator as type of second
1014 argument instead of equivalent void (*) (void *, void *, void *).
1015 * poly-int.h (gt_pch_nx): Likewise.
1016 * wide-int.h (gt_pch_nx): Likewise.
1017 * config/aarch64/aarch64-sve-builtins.cc (gt_pch_nx): Likewise.
1019 2021-12-13 Jan Hubicka <hubicka@ucw.cz>
1022 * ipa-fnsummary.c (evaluate_conditions_for_known_args): Do not ICE
1023 on ternary expression.
1025 2021-12-13 Kewen Lin <linkw@linux.ibm.com>
1028 * attribs.c (decl_attributes): Check if target options change and
1029 create one node if so.
1031 2021-12-12 Jonathan Wakely <jwakely@redhat.com>
1033 * Makefile.in: Remove unique-ptr-tests.o.
1034 * selftest-run-tests.c (selftest::run_tests): Remove
1035 unique_ptr_tests_cc_tests.
1036 * selftest.h (unique_ptr_tests_cc_tests): Remove.
1037 * system.h: Check INCLUDE_MEMORY instead of INCLUDE_UNIQUE_PTR
1038 and include <memory> instead of "unique-ptr.h".
1039 * unique-ptr-tests.cc: Removed.
1041 2021-12-12 Roger Sayle <roger@nextmovesoftware.com>
1043 * config/nvptx/nvptx-opts.h (ptx_isa): Add PTX_ISA_SM53 ISA level
1045 * config/nvptx/nvptx.opt: Add sm_53 to -misa.
1046 * config/nvptx/nvptx-modes.def: Add support for HFmode.
1047 * config/nvptx/nvptx.h (TARGET_SM53):
1048 New helper macro to conditionalize functionality on target ISA.
1049 * config/nvptx/nvptx-c.c (nvptx_cpu_cpp_builtins): Add __PTX_SM__
1050 support for the new ISA levels.
1051 * config/nvptx/nvptx.c (nvtx_ptx_type_from_mode): Support new HFmode
1052 with the ".f16" suffix/qualifier.
1053 (nvptx_file_start): Add support for TARGET_SM53.
1054 (nvptx_omp_device_kind_arch_isa): Add support for TARGET_SM53
1055 and tweak TARGET_SM35.
1056 (nvptx_scalar_mode_supported_p): Target hook with conditional
1057 HFmode support on TARGET_SM53 and higher.
1058 (nvptx_libgcc_floating_mode_supported_p): Likewise.
1059 (TARGET_SCALAR_MODE_SUPPORTED_P): Use nvptx_scalar_mode_supported_p.
1060 (TARGET_LIBGCC_FLOATING_MODE_SUPPORTED_P): Likewise, use new hook.
1061 * config/nvptx/nvptx.md (*movhf_insn): New define_insn.
1062 (movhf): New define_expand for HFmode moves.
1063 (addhf3, subhf3, mulhf, extendhf<mode>2, trunc<mode>hf2): New
1064 instructions conditional on TARGET_SM53 (i.e. -misa=sm_53).
1066 2021-12-12 Jan Hubicka <hubicka@ucw.cz>
1069 * ipa-modref.c (modref_access_analysis::analyze): Terminate BB
1070 analysis on NULL memory access.
1071 * ipa-pure-const.c (analyze_function): Likewise.
1073 2021-12-11 Jan Hubicka <hubicka@ucw.cz>
1075 * ipa-profile.c (ipa_profile): Do not update hot bb threshold.
1077 2021-12-11 Jan Hubicka <hubicka@ucw.cz>
1079 * ipa-modref.c (get_modref_function_summary): Use ultimate_alias_target.
1080 (ignore_edge): Likewise.
1081 (compute_parm_map): Likewise.
1082 (modref_propagate_in_scc): Likewise.
1083 (modref_propagate_flags_in_scc): Likewise.
1085 2021-12-10 Jason Merrill <jason@redhat.com>
1087 * symtab.c (symtab_node::equal_address_to): Fix comment typo.
1089 2021-12-10 Doug Rupp <rupp@adacore.com>
1091 * config/vxworks.h (LINK_SPEC): Remove %(link_target).
1092 Change %{v:-v} to %{v:-V}.
1094 2021-12-10 Olivier Hainque <hainque@adacore.com>
1096 * config/t-vxworks: Remove assignment to STMP_FIXINC.
1098 2021-12-10 Martin Liska <mliska@suse.cz>
1100 * params.opt: Add missing dot.
1102 2021-12-10 Roger Sayle <roger@nextmovesoftware.com>
1105 * ipa-modref-tree.h (useful_for_kill_p): Zero width accesses aren't
1106 useful for kill tracking.
1108 2021-12-10 Andrew Stubbs <ams@codesourcery.com>
1110 * config/gcn/mkoffload.c (process_asm): Process the variable table
1111 completely differently.
1112 (process_obj): Encode the varaible data differently.
1114 2021-12-10 Joel Hutton <joel.hutton@arm.com>
1116 PR tree-optimization/103523
1117 * tree-vect-loop.c (vectorizable_induction): Check for
1118 PLUS_EXPR/MINUS_EXPR support.
1120 2021-12-10 Cui,Lili <lili.cui@intel.com>
1122 * config/i386/i386.c (ix86_vector_costs::add_stmt_cost): Remove Tremont.
1124 2021-12-09 Jan Hubicka <hubicka@ucw.cz>
1126 * doc/invoke.texi (max-inline-functions-called-once-loop-depth,
1127 max-inline-functions-called-once-insns): New parameters.
1128 * ipa-inline.c (check_callers): Handle
1129 param_inline_functions_called_once_loop_depth and
1130 param_inline_functions_called_once_insns.
1131 (edge_badness): Fix linebreaks.
1132 * params.opt (param=max-inline-functions-called-once-loop-depth,
1133 param=max-inline-functions-called-once-insn): New params.
1135 2021-12-09 Martin Sebor <msebor@redhat.com>
1137 PR tree-optimization/103215
1138 * pointer-query.cc (access_ref::merge_ref): Extend the offset and
1139 size of the merged object instead of using the larger.
1141 2021-12-09 Martin Sebor <msebor@redhat.com>
1143 PR middle-end/101751
1144 * doc/extend.texi (attribute access): Adjust.
1145 * gimple-ssa-warn-access.cc (pass_waccess::maybe_check_access_sizes):
1146 Treat access mode none on a void* argument as expecting as few as
1149 2021-12-09 Jakub Jelinek <jakub@redhat.com>
1152 * config/aarch64/aarch64-sve-builtins.cc (gt_pch_nx): Change type of
1153 second argument from function with 2 pointer arguments to function
1154 with 3 pointer arguments.
1156 2021-12-09 Olivier Hainque <hainque@adacore.com>
1158 * config/aarch64/aarch64-vxworks.h (TARGET_OS_CPP_BUILTINS):
1159 Use VX_CPU_PREFIX in CPU definitions.
1161 2021-12-09 Martin Sebor <msebor@redhat.com>
1163 * pointer-query.cc (access_ref::dump): Define new function
1164 (pointer_query::dump): Call it.
1165 * pointer-query.h (access_ref::dump): Declare new function.
1167 2021-12-09 Martin Sebor <msebor@redhat.com>
1169 * pointer-query.cc (compute_objsize_r): Add an argument.
1170 (gimple_call_return_array): Pass a new argument to compute_objsize_r.
1171 (access_ref::merge_ref): Same.
1172 (access_ref::inform_access): Add an argument and use it.
1173 (access_data::access_data): Initialize new member.
1174 (handle_min_max_size): Pass a new argument to compute_objsize_r.
1175 (handle_decl): New function.
1176 (handle_array_ref): Pass a new argument to compute_objsize_r.
1177 Avoid incrementing deref.
1178 (set_component_ref_size): New function.
1179 (handle_component_ref): New function.
1180 (handle_mem_ref): Pass a new argument to compute_objsize_r.
1181 Only increment deref after successfully computing object size.
1182 (handle_ssa_name): New function.
1183 (compute_objsize_r): Move code into helpers and call them.
1184 (compute_objsize): Pass a new argument to compute_objsize_r.
1185 * pointer-query.h (access_ref::inform_access): Add an argument.
1186 (access_data::ostype): New member.
1188 2021-12-09 Martin Sebor <msebor@redhat.com>
1190 * pointer-query.cc (access_ref::merge_ref): Define new function.
1191 (access_ref::get_ref): Move code into merge_ref and call it.
1192 * pointer-query.h (access_ref::merge_ref): Declare new function.
1194 2021-12-09 Martin Sebor <msebor@redhat.com>
1196 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Pass
1197 GIMPLE statement to compute_objsize.
1198 * pointer-query.cc (compute_objsize): Add a statement argument.
1199 * pointer-query.h (compute_objsize): Define a new overload.
1201 2021-12-09 Martin Sebor <msebor@redhat.com>
1203 * gimple-ssa-warn-access.cc (check_access): Adjust to member name
1205 (pass_waccess::check_strncmp): Same.
1206 * pointer-query.cc (access_ref::access_ref): Remove arguments.
1208 (access_data::access_data): Define new ctors.
1209 (access_data::set_bound): Define new member function.
1210 (compute_objsize_r): Remove unnecessary code.
1211 * pointer-query.h (struct access_ref): Remove ctor arguments.
1212 (struct access_data): Declare ctor overloads.
1213 (access_data::dst_bndrng): New member.
1214 (access_data::src_bndrng): New member.
1216 2021-12-09 Martin Sebor <msebor@redhat.com>
1218 PR middle-end/103143
1219 * pointer-query.cc (gimple_call_return_array): Call compute_objsize_r.
1221 2021-12-09 Olivier Hainque <hainque@adacore.com>
1222 Rasmus Villemoes <rv@rasmusvillemoes.dk>
1224 * Makefile.in (T_STDINT_GCC_H): New variable, path to
1225 stdint-gcc.h that a target configuration may override when
1226 use_gcc_stdint is "provide".
1227 (stmp-int-hdrs): Depend on it and copy that for
1228 USE_GCC_INT=provide.
1229 * config.gcc (vxworks): Revert to use_gcc_stdint=provide.
1230 * config/t-vxworks (T_STDINT_GCC_H): Define, as vxw-stdint-gcc.h.
1231 (vxw-stdint-gcc.h): New target, produced from the original
1233 (vxw-glimits.h): Use an automatic variable to designate the
1234 first and only prerequisite.
1235 * config/vxworks/stdint.h: Remove.
1237 2021-12-09 Iain Sandoe <iain@sandoe.co.uk>
1240 * config/host-darwin.c (SAFE_ALLOC_SIZE): Remove.
1241 (darwin_gt_pch_get_address): Rework for relocatable PCH.
1242 (darwin_gt_pch_use_address): Likewise.
1244 2021-12-09 Jakub Jelinek <jakub@redhat.com>
1247 * config/host-darwin.c (darwin_gt_pch_use_address): When reading
1248 manually the file into mapped area, update mapped_addr as
1249 an automatic variable rather than addr which is a reference parameter.
1250 * config/host-hpux.c (hpux_gt_pch_use_address): When reading
1251 manually the file into mapped area, update addr as
1252 an automatic variable rather than base which is a reference parameter.
1254 2021-12-09 Jakub Jelinek <jakub@redhat.com>
1257 * coretypes.h (gt_pointer_operator): Use 3 pointer arguments instead
1259 * gengtype.c (struct walk_type_data): Add in_nested_ptr argument.
1260 (walk_type): Temporarily set d->in_nested_ptr around nested_ptr
1262 (write_types_local_user_process_field): Pass a new middle pointer
1263 to gt_pointer_operator op calls, if d->in_nested_ptr pass there
1264 address of d->prev_val[2], otherwise NULL.
1265 (write_types_local_process_field): Likewise.
1266 * ggc-common.c (relocate_ptrs): Add real_ptr_p argument. If equal
1267 to ptr_p, do nothing, otherwise if NULL remember ptr_p's
1268 or if non-NULL real_ptr_p's corresponding new address in
1270 (reloc_addrs_vec): New variable.
1271 (compare_ptr, read_uleb128, write_uleb128): New functions.
1272 (gt_pch_save): When iterating over objects through relocate_ptrs,
1273 save current i into state.ptrs_i. Sort reloc_addrs_vec and emit
1274 it as uleb128 of differences between pointer addresses into the
1276 (gt_pch_restore): Allow restoring of PCH to a different address
1277 than the preferred one, in that case adjust global pointers by bias
1278 and also adjust by bias addresses read from the relocation table
1279 as uleb128 differences. Otherwise fseek over it. Perform
1280 gt_pch_restore_stringpool only after adjusting callbacks and for
1281 callback adjustments also take into account the bias.
1282 (default_gt_pch_use_address): Change type of first argument from
1284 (mmap_gt_pch_use_address): Likewise.
1285 * ggc-tests.c (gt_pch_nx): Pass NULL as new middle argument to op.
1286 * hash-map.h (hash_map::pch_nx_helper): Likewise.
1287 (gt_pch_nx): Likewise.
1288 * hash-set.h (gt_pch_nx): Likewise.
1289 * hash-table.h (gt_pch_nx): Likewise.
1290 * hash-traits.h (ggc_remove::pch_nx): Likewise.
1291 * hosthooks-def.h (default_gt_pch_use_address): Change type of first
1292 argument from void * to void *&.
1293 (mmap_gt_pch_use_address): Likewise.
1294 * hosthooks.h (struct host_hooks): Change type of first argument of
1295 gt_pch_use_address hook from void * to void *&.
1296 * machmode.h (gt_pch_nx): Expect a callback with 3 pointers instead of
1297 two in the middle argument.
1298 * poly-int.h (gt_pch_nx): Likewise.
1299 * stringpool.c (gt_pch_nx): Pass NULL as new middle argument to op.
1300 * tree-cfg.c (gt_pch_nx): Likewise, except for LOCATION_BLOCK pass
1301 the same &(block) twice.
1302 * value-range.h (gt_pch_nx): Pass NULL as new middle argument to op.
1303 * vec.h (gt_pch_nx): Likewise.
1304 * wide-int.h (gt_pch_nx): Likewise.
1305 * config/host-darwin.c (darwin_gt_pch_use_address): Change type of
1306 first argument from void * to void *&.
1307 * config/host-darwin.h (darwin_gt_pch_use_address): Likewise.
1308 * config/host-hpux.c (hpux_gt_pch_use_address): Likewise.
1309 * config/host-linux.c (linux_gt_pch_use_address): Likewise. If
1310 it couldn't succeed to mmap at the preferred location, set base
1311 to the actual one. Update addr in the manual reading loop instead of
1313 * config/host-netbsd.c (netbsd_gt_pch_use_address): Change type of
1314 first argument from void * to void *&.
1315 * config/host-openbsd.c (openbsd_gt_pch_use_address): Likewise.
1316 * config/host-solaris.c (sol_gt_pch_use_address): Likewise.
1317 * config/i386/host-mingw32.c (mingw32_gt_pch_use_address): Likewise.
1318 * config/rs6000/rs6000-gen-builtins.c (write_init_file): Pass NULL
1319 as new middle argument to op in the generated code.
1320 * doc/gty.texi: Adjust samples for the addition of middle pointer
1321 to gt_pointer_operator callback.
1323 2021-12-09 Alexandre Oliva <oliva@adacore.com>
1326 * reg-stack.c (convert_regs_1): Move any_malformed_asm
1328 (reg_to_stack): ... here.
1330 2021-12-09 Alexandre Oliva <oliva@adacore.com>
1333 * expr.c (emit_move_multi_word): Skip clobber during lra.
1335 2021-12-09 Alexandre Oliva <oliva@adacore.com>
1337 PR tree-optimization/103024
1338 PR middle-end/103530
1339 * gimple-harden-conditionals.cc (non_eh_succ_edge): New.
1340 (pass_harden_compares::execute): Accept 1-bit integral types,
1341 and cope with throwing compares.
1343 2021-12-08 Iain Sandoe <iain@sandoe.co.uk>
1345 * config/darwin.h (DARWIN_PIE_SPEC): Add -no_pie when
1346 linking mdynamic-no-pic code on macOS > 10.7.
1348 2021-12-08 Dimitar Dimitrov <dimitar@dinux.eu>
1350 * config/pru/pru.c (pru_section_type_flags): New function.
1351 (TARGET_SECTION_TYPE_FLAGS): Wire it.
1353 2021-12-08 David Faust <david.faust@oracle.com>
1355 * config/bpf/bpf.c (handle_attr_preserve): Avoid calling
1356 is_gimple_assign with a NULL pointer.
1358 2021-12-08 Martin Liska <mliska@suse.cz>
1360 * profile-count.c (profile_count::dump): Add function
1361 that can dump to a provided buffer.
1362 (profile_probability::dump): Likewise.
1363 * profile-count.h: Likewise.
1364 * tree-ssa-loop-unswitch.c (tree_unswitch_single_loop):
1365 Use dump_printf_loc infrastructure.
1366 (tree_unswitch_outer_loop): Likewise.
1367 (find_loop_guard): Likewise.
1368 (hoist_guard): Likewise.
1370 2021-12-08 Chung-Lin Tang <cltang@codesourcery.com>
1372 * gimplify.c (extract_base_bit_offset): Add 'tree *offsetp' parameter,
1373 accomodate case where 'offset' return of get_inner_reference is
1375 (is_or_contains_p): Further robustify conditions.
1376 (omp_target_reorder_clauses): In alloc/to/from sorting phase, also
1377 move following GOMP_MAP_ALWAYS_POINTER maps along. Add new sorting
1378 phase where we make sure pointers with an attach/detach map are ordered
1380 (gimplify_scan_omp_clauses): Add modifications to avoid creating
1381 GOMP_MAP_STRUCT and associated alloc map for attach/detach maps.
1383 2021-12-08 Roger Sayle <roger@nextmovesoftware.com>
1385 * config/nvptx/nvptx.md (*extend_trunc_<mode>2_qi,
1386 *extend_trunc_<mode>2_hi, *extend_trunc_di2_si): New insns.
1387 Use cvt to perform sign-extension of truncation in one step.
1389 2021-12-08 Chung-Lin Tang <cltang@codesourcery.com>
1392 * gimplify.c ("tree-hash-traits.h"): Add include.
1393 (gimplify_scan_omp_clauses): Change struct_map_to_clause to type
1394 hash_map<tree_operand, tree> *. Adjust struct map handling to handle
1395 cases of *A and A->B expressions. Under !DECL_P case of
1396 GOMP_CLAUSE_MAP handling, add STRIP_NOPS for indir_p case, add to
1397 struct_deref_set for map(*ptr_to_struct) cases. Add MEM_REF case when
1398 handling component_ref_p case. Add unshare_expr and gimplification
1399 when created GOMP_MAP_STRUCT is not a DECL. Add code to add
1400 firstprivate pointer for *pointer-to-struct case.
1401 (gimplify_adjust_omp_clauses): Move GOMP_MAP_STRUCT removal code for
1402 exit data directives code to earlier position.
1403 * omp-low.c (lower_omp_target):
1404 Handle GOMP_MAP_ATTACH_ZERO_LENGTH_ARRAY_SECTION, and
1405 GOMP_MAP_POINTER_TO_ZERO_LENGTH_ARRAY_SECTION map kinds.
1406 * tree-pretty-print.c (dump_omp_clause): Likewise.
1408 2021-12-08 Andrew Stubbs <ams@codesourcery.com>
1409 Hafiz Abid Qadeer <abidh@codesourcery.com>
1411 * dwarf2cfi.c (dw_stack_pointer_regnum): Change type to struct cfa_reg.
1412 (dw_frame_pointer_regnum): Likewise.
1413 (new_cfi_row): Use set_by_dwreg.
1414 (get_cfa_from_loc_descr): Use set_by_dwreg. Support register spans.
1415 handle DW_OP_bregx with DW_OP_breg{0-31}. Support DW_OP_lit*,
1416 DW_OP_const*, DW_OP_minus, DW_OP_shl and DW_OP_plus.
1417 (lookup_cfa_1): Use set_by_dwreg.
1418 (def_cfa_0): Update for cfa_reg and support register spans.
1419 (reg_save): Change sreg parameter to struct cfa_reg. Support register
1421 (dwf_cfa_reg): New function.
1422 (dwarf2out_flush_queued_reg_saves): Use dwf_cfa_reg instead of
1424 (dwarf2out_frame_debug_def_cfa): Likewise.
1425 (dwarf2out_frame_debug_adjust_cfa): Likewise.
1426 (dwarf2out_frame_debug_cfa_offset): Likewise. Update reg_save usage.
1427 (dwarf2out_frame_debug_cfa_register): Likewise.
1428 (dwarf2out_frame_debug_expr): Likewise.
1429 (create_pseudo_cfg): Use set_by_dwreg.
1430 (initial_return_save): Use set_by_dwreg and dwf_cfa_reg,
1431 (create_cie_data): Use dwf_cfa_reg.
1432 (execute_dwarf2_frame): Use dwf_cfa_reg.
1433 (dump_cfi_row): Use set_by_dwreg.
1434 * dwarf2out.c (build_span_loc, build_breg_loc): New function.
1435 (build_cfa_loc): Support register spans.
1436 (build_cfa_aligned_loc): Update cfa_reg usage.
1437 (convert_cfa_to_fb_loc_list): Use set_by_dwreg.
1438 * dwarf2out.h (struct cfa_reg): New type.
1439 (struct dw_cfa_location): Use struct cfa_reg.
1440 (build_span_loc): New prototype.
1442 2021-12-08 Haochen Jiang <haochen.jiang@intel.com>
1445 * config/i386/sse.md
1446 (*<sse4_1>_blendv<ssefltmodesuffix><avxsizesuffix>_not_ltint):
1447 Add new define_insn_and_split.
1449 2021-12-08 Alexandre Oliva <oliva@adacore.com>
1451 PR middle-end/103149
1452 * gimple-harden-conditionals.cc (detach_value): Use memory if
1453 general regs won't do.
1455 2021-12-07 Martin Liska <mliska@suse.cz>
1457 PR middle-end/103438
1458 * config/s390/s390.c (s390_valid_target_attribute_inner_p):
1459 Use new enum CLVC_INTEGER.
1460 * opt-functions.awk: Use new CLVC_INTEGER.
1461 * opts-common.c (set_option): Likewise.
1462 (option_enabled): Return -1,0,1 for CLVC_INTEGER.
1463 (get_option_state): Use new CLVC_INTEGER.
1464 (control_warning_option): Likewise.
1465 * opts.h (enum cl_var_type): Likewise.
1467 2021-12-07 H.J. Lu <hjl.tools@gmail.com>
1470 * config/i386/i386.c (ix86_call_use_plt_p): Check FUNCTION_DECL
1471 before calling cgraph_node::get.
1473 2021-12-07 Richard Biener <rguenther@suse.de>
1475 PR tree-optimization/103596
1476 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_stmt):
1477 Note we are not propagating into a PHI argument to may_propagate_copy.
1478 * tree-ssa-propagate.h (may_propagate_copy): Add
1479 argument specifying whether we propagate into a PHI arg.
1480 * tree-ssa-propagate.c (may_propagate_copy): Likewise.
1481 When not doing so we can replace an abnormal with
1483 (may_propagate_into_stmt): Update may_propagate_copy calls.
1484 (replace_exp_1): Move propagation checking code to
1485 propagate_value and rename to ...
1486 (replace_exp): ... this and elide previous wrapper.
1487 (propagate_value): Perform checking with adjusted
1488 may_propagate_copy call and dispatch to replace_exp.
1490 2021-12-07 Matthias Kretz <m.kretz@gsi.de>
1492 * hash-map.h (hash_map::traverse): Let both overloads behave the
1494 * predict.c (assert_is_empty): Return true, thus not changing
1497 2021-12-07 YunQiang Su <yunqiang.su@cipunited.com>
1499 * config/mips/mips.h (ISA_HAS_UNALIGNED_ACCESS, STRICT_ALIGNMENT):
1500 R6 can unaligned access.
1501 * config/mips/mips.md (movmisalign<mode>): Likewise.
1502 * config/mips/mips.opt: add -m(no-)unaligned-access
1503 * doc/invoke.texi: Likewise.
1505 2021-12-07 Eugene Rozenfeld <erozen@microsoft.com>
1507 * auto-profile.c (afdo_propagate_edge): Improve count propagation algorithm.
1509 2021-12-06 Paul A. Clarke <pc@us.ibm.com>
1512 * config/rs6000/xmmintrin.h (_mm_movemask_ps): Replace "vector" with
1515 2021-12-06 Jose E. Marchesi <jose.marchesi@oracle.com>
1517 * config/bpf/bpf.c (bpf_handle_preserve_access_index_attribute):
1518 Mark arguments `args' and flags' as unused.
1519 (bpf_core_newdecl): Remove unused local `newdecl'.
1520 (bpf_core_newdecl): Remove unused argument `loc'.
1521 (ctfc_debuginfo_early_finish_p): Remove unused function.
1522 (TARGET_CTFC_DEBUGINFO_EARLY_FINISH_P): Remove definition.
1523 (bpf_core_walk): Do not pass a location to bpf_core_newdecl.
1525 2021-12-06 Richard Sandiford <richard.sandiford@arm.com>
1527 * gimple-range-edge.cc (gimple_outgoing_range::edge_range_p): Add
1528 a shortcut for blocks with single successors.
1529 * gimple-range-gori.cc (gori_map::calculate_gori): Likewise.
1531 2021-12-06 Richard Sandiford <richard.sandiford@arm.com>
1533 * value-range.cc (irange::irange_union): Use quick_push rather
1534 than safe_push. Use widest_int rather than wide_int. Avoid
1535 assigning wi::to_* results to wide*_int temporaries.
1537 2021-12-06 Andrew MacLeod <amacleod@redhat.com>
1539 * gimple-range-cache.cc (ranger_cache::fill_block_cache): Check for
1540 a range from dominators before filling the cache.
1541 (ranger_cache::range_from_dom): New.
1542 * gimple-range-cache.h (ranger_cache::range_from_dom): Add prototype.
1544 2021-12-06 Andrew MacLeod <amacleod@redhat.com>
1546 * gimple-range-gori.h (class gori_compute):: Add prototypes.
1547 * gimple-range-gori.cc (gori_compute::has_edge_range_p): Add alternate
1548 API for basic block. Call for edge alterantive.
1549 (gori_compute::may_recompute_p): Ditto.
1551 2021-12-06 Richard Biener <rguenther@suse.de>
1553 PR tree-optimization/103581
1554 * tree-vect-stmts.c (vect_build_gather_load_calls): Properly
1555 guard all the AVX512 mask cases.
1557 2021-12-06 Richard Biener <rguenther@suse.de>
1559 PR tree-optimization/103544
1560 * tree-vect-slp.c (vect_analyze_slp): Only add a SLP reduction
1561 opportunity if the stmt in question is the reduction root.
1562 (dot_slp_tree): Add missing check for NULL child.
1564 2021-12-06 Jakub Jelinek <jakub@redhat.com>
1567 * config/avr/avr.c (avr_output_data_section_asm_op,
1568 avr_output_bss_section_asm_op): Change argument type from const void *
1571 2021-12-06 Tamar Christina <tamar.christina@arm.com>
1573 PR rtl-optimization/103404
1574 * cse.c (find_sets_in_insn): Don't select elements out of a V1 mode
1577 2021-12-06 Hongtao Liu <Hongtao.liu@intel.com>
1578 Uroš Bizjak <ubizjak@gmail.com>
1581 * config/i386/i386.c (ix86_preferred_reload_class): Allow
1582 integer regs when moves between register units are cheap.
1583 * config/i386/i386.h (INT_SSE_CLASS_P): New.
1585 2021-12-05 David Edelsohn <dje.gcc@gmail.com>
1587 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Move
1588 AIX math builtin initialization before new_builtins_are_live.
1590 2021-12-04 Jakub Jelinek <jakub@redhat.com>
1592 * config/i386/x86-tune.def (X86_TUNE_PARTIAL_REG_DEPENDENCY): Fix
1593 comment typo, Preffer -> prefer.
1594 * ipa-modref-tree.c (modref_access_node::closer_pair_p): Likewise.
1596 2021-12-04 Alexandre Oliva <oliva@adacore.com>
1598 PR rtl-optimization/103028
1599 * ifcvt.c (find_cond_trap): Validate new insns more strictly
1602 2021-12-03 Martin Liska <mliska@suse.cz>
1603 Peter Bergner <bergner@linux.ibm.com>
1606 * config/rs6000/rs6000.c (rs6000_option_override_internal): Move the
1607 disabling of shrink-wrapping when using -mrop-protect from here...
1608 (rs6000_override_options_after_change): ...to here.
1610 2021-12-03 H.J. Lu <hjl.tools@gmail.com>
1613 * config/i386/i386-expand.c (ix86_expand_builtin): Pass PVW_NONE
1614 and PVW_NONE to ix86_target_string.
1615 * config/i386/i386-options.c (ix86_target_string): Add arguments
1616 for move_max and store_max.
1617 (ix86_target_string::add_vector_width): New lambda.
1618 (ix86_debug_options): Pass ix86_move_max and ix86_store_max to
1620 (ix86_function_specific_print): Pass ptr->x_ix86_move_max and
1621 ptr->x_ix86_store_max to ix86_target_string.
1622 (ix86_valid_target_attribute_tree): Handle x_ix86_move_max and
1624 (ix86_option_override_internal): Set the default x_ix86_move_max
1625 and x_ix86_store_max.
1626 * config/i386/i386-options.h (ix86_target_string): Add
1627 prefer_vector_width and prefer_vector_width.
1628 * config/i386/i386.h (TARGET_AVX256_MOVE_BY_PIECES): Removed.
1629 (TARGET_AVX256_STORE_BY_PIECES): Likewise.
1630 (MOVE_MAX): Use 64 if ix86_move_max or ix86_store_max ==
1631 PVW_AVX512. Use 32 if ix86_move_max or ix86_store_max >=
1633 (STORE_MAX_PIECES): Use 64 if ix86_store_max == PVW_AVX512.
1634 Use 32 if ix86_store_max >= PVW_AVX256.
1635 * config/i386/i386.opt: Add -mmove-max=bits and -mstore-max=bits.
1636 * config/i386/x86-tune.def (X86_TUNE_AVX512_MOVE_BY_PIECES): New.
1637 (X86_TUNE_AVX512_STORE_BY_PIECES): Likewise.
1638 * doc/invoke.texi: Document -mmove-max=bits and -mstore-max=bits.
1640 2021-12-03 Bill Schmidt <wschmidt@linux.ibm.com>
1642 * config/rs6000/rs6000.c (rs6000_builtin_reciprocal): Fix builtin
1645 2021-12-03 SiYu Wu <siyu@isrc.iscas.ac.cn>
1647 * common/config/riscv/riscv-common.c (riscv_implied_info):
1648 Add K-ext related entry.
1649 (riscv_supported_std_ext): Add 'k'.
1650 * config/riscv/arch-canonicalize (CANONICAL_ORDER): Add 'k'.
1651 (IMPLIED_EXT): Add K-ext related entry.
1653 2021-12-03 SiYu Wu <siyu@isrc.iscas.ac.cn>
1655 * common/config/riscv/riscv-common.c (riscv_ext_version_table):
1657 * config/riscv/riscv-opts.h (MASK_ZBKB): New.
1664 (MASK_ZKSED): Ditto.
1667 (TARGET_ZBKB): Ditto.
1668 (TARGET_ZBKC): Ditto.
1669 (TARGET_ZBKX): Ditto.
1670 (TARGET_ZKNE): Ditto.
1671 (TARGET_ZKND): Ditto.
1672 (TARGET_ZKNH): Ditto.
1673 (TARGET_ZKR): Ditto.
1674 (TARGET_ZKSED): Ditto.
1675 (TARGET_ZKSH): Ditto.
1676 (TARGET_ZKT): Ditto.
1677 * config/riscv/riscv.opt (riscv_zk_subext): New.
1679 2021-12-03 Tamar Christina <tamar.christina@arm.com>
1681 * tree-vect-stmts.c (prepare_load_store_mask): Rename to...
1682 (prepare_vec_mask): ...This and record operations that have already been
1684 (vectorizable_call): Use it.
1685 (vectorizable_operation): Likewise.
1686 (vectorizable_store): Likewise.
1687 (vectorizable_load): Likewise.
1688 * tree-vectorizer.h (class _loop_vec_info): Add vec_cond_masked_set.
1689 (vec_cond_masked_set_type, tree_cond_mask_hash): New.
1691 2021-12-03 H.J. Lu <hjl.tools@gmail.com>
1695 * target.def (ifunc_ref_local_ok): Add a target hook.
1696 * varasm.c (default_binds_local_p_3): Force indirect function
1697 resolver non-local only if targetm.ifunc_ref_local_ok returns
1699 * config/i386/i386-expand.c (ix86_expand_call): Call
1700 ix86_call_use_plt_p to check if PLT should be used.
1701 * config/i386/i386-protos.h (ix86_call_use_plt_p): New.
1702 * config/i386/i386.c (output_pic_addr_const): Call
1703 ix86_call_use_plt_p to check if "@PLT" is needed.
1704 (ix86_call_use_plt_p): New.
1705 (TARGET_IFUNC_REF_LOCAL_OK): New.
1706 * doc/tm.texi.in: Add TARGET_IFUNC_REF_LOCAL_OK.
1707 * doc/tm.texi: Regenerated.
1709 2021-12-03 Jakub Jelinek <jakub@redhat.com>
1711 * attribs.h (simple_cst_list_equal): Declare.
1712 * attribs.c (simple_cst_list_equal): No longer static.
1713 * config/i386/i386-options.c (target_attribute_cache): New variable.
1714 (ix86_valid_target_attribute_p): Cache DECL_FUNCTION_SPECIFIC_TARGET
1715 and DECL_FUNCTION_SPECIFIC_OPTIMIZATION based on args.
1717 2021-12-03 Jakub Jelinek <jakub@redhat.com>
1720 * ggc.h (gt_pch_note_callback): Declare.
1721 * gengtype.h (enum typekind): Add TYPE_CALLBACK.
1722 (callback_type): Declare.
1723 * gengtype.c (dbgprint_count_type_at): Handle TYPE_CALLBACK.
1724 (callback_type): New variable.
1725 (process_gc_options): Add CALLBACK argument, handle callback
1727 (set_gc_used_type): Adjust process_gc_options caller, if callback,
1728 set type to &callback_type.
1729 (output_mangled_typename): Handle TYPE_CALLBACK.
1730 (walk_type): Likewise. Handle callback option.
1731 (write_types_process_field): Handle TYPE_CALLBACK.
1732 (write_types_local_user_process_field): Likewise.
1733 (write_types_local_process_field): Likewise.
1734 (write_root): Likewise.
1735 (dump_typekind): Likewise.
1736 (dump_type): Likewise.
1737 * gengtype-state.c (type_lineloc): Handle TYPE_CALLBACK.
1738 (state_writer::write_state_callback_type): New method.
1739 (state_writer::write_state_type): Handle TYPE_CALLBACK.
1740 (read_state_callback_type): New function.
1741 (read_state_type): Handle TYPE_CALLBACK.
1742 * ggc-common.c (callback_vec): New variable.
1743 (gt_pch_note_callback): New function.
1744 (gt_pch_save): Stream out gt_pch_save function address and relocation
1746 (gt_pch_restore): Stream in saved gt_pch_save function address and
1747 relocation table and apply relocations if needed.
1748 * doc/gty.texi (callback): Document new GTY option.
1749 * varasm.c (get_unnamed_section): Change callback argument's type and
1750 last argument's type from const void * to const char *.
1751 (output_section_asm_op): Change argument's type from const void *
1752 to const char *, remove unnecessary cast.
1753 * tree-core.h (struct tree_translation_unit_decl): Drop GTY((skip))
1754 from language member.
1755 * output.h (unnamed_section_callback): Change argument type from
1756 const void * to const char *.
1757 (struct unnamed_section): Use GTY((callback)) instead of GTY((skip))
1758 for callback member. Change data member type from const void *
1760 (struct noswitch_section): Use GTY((callback)) instead of GTY((skip))
1761 for callback member.
1762 (get_unnamed_section): Change callback argument's type and
1763 last argument's type from const void * to const char *.
1764 (output_section_asm_op): Change argument's type from const void *
1766 * config/avr/avr.c (avr_output_progmem_section_asm_op): Likewise.
1767 Remove unneeded cast.
1768 * config/darwin.c (output_objc_section_asm_op): Change argument's type
1769 from const void * to const char *.
1770 * config/pa/pa.c (som_output_text_section_asm_op): Likewise.
1771 (som_output_comdat_data_section_asm_op): Likewise.
1772 * config/rs6000/rs6000.c (rs6000_elf_output_toc_section_asm_op):
1774 (rs6000_xcoff_output_readonly_section_asm_op): Likewise. Instead
1775 of dereferencing directive hardcode variable names and decide based on
1776 whether directive is NULL or not.
1777 (rs6000_xcoff_output_readwrite_section_asm_op): Change argument's type
1778 from const void * to const char *.
1779 (rs6000_xcoff_output_tls_section_asm_op): Likewise. Instead
1780 of dereferencing directive hardcode variable names and decide based on
1781 whether directive is NULL or not.
1782 (rs6000_xcoff_output_toc_section_asm_op): Change argument's type
1783 from const void * to const char *.
1784 (rs6000_xcoff_asm_init_sections): Adjust get_unnamed_section callers.
1786 2021-12-03 Richard Purdie <richard.purdie@linuxfoundation.org>
1788 * Makefile.in: Fix "argument list too long" from install-plugins.
1790 2021-12-03 Matt Jacobson <mhjacobson@me.com>
1792 * config.gcc: For the AVR target, populate TM_MULTILIB_CONFIG.
1793 * config/avr/genmultilib.awk: Add ability to filter generated multilib
1795 * config/avr/t-avr: Pass TM_MULTILIB_CONFIG to genmultilib.awk.
1796 * configure.ac: Update help string for --with-multilib-list.
1797 * configure: Regenerate.
1799 2021-12-02 Ilya Lipnitskiy <ilya.lipnitskiy@gmail.com>
1801 * configure: Regenerate.
1802 * configure.ac: Define TARGET_LIBC_GNUSTACK on musl.
1804 2021-12-02 Iain Sandoe <iain@sandoe.co.uk>
1807 * config/host-darwin.c (TRY_EMPTY_VM_SPACE,
1808 SAFE_ALLOC_SIZE): New.
1809 (darwin_gt_pch_get_address): Rewrite to use nominated
1810 memory segments rather than part of the compiler __DATA
1812 (darwin_gt_pch_use_address): Likewise.
1814 2021-12-02 Gerald Pfeifer <gerald@pfeifer.com>
1816 * doc/install.texi (*-*-freebsd*): Remove references to
1817 FreeBSD 1 and FreeBSD 2.
1819 2021-12-02 Vladimir N. Makarov <vmakarov@redhat.com>
1821 PR rtl-optimization/103437
1822 * ira-color.c (setup_allocno_priorities): Switch off backup code
1823 for overflow if compiler has __builtin_smul_overflow. Use <
1824 for comparison with -INT_MAX.
1826 2021-12-02 Bill Schmidt <wschmidt@linux.ibm.com>
1828 * config/rs6000/altivec.h: Delete a number of #defines that are now
1829 superfluous. Alphabetize. Include rs6000-vecdefines.h. Include some
1831 * config/rs6000/rs6000-builtin-new.def (CMPB): Flag as no32bit.
1832 (BPERMD): Flag as 32bit (needing special handling for 32-bit).
1833 (UNPACK_TD): Return unsigned long long instead of unsigned long.
1834 (GET_TEXASR): Return unsigned long instead of unsigned long long.
1835 (GET_TEXASRU): Likewise.
1836 (GET_TFHAR): Likewise.
1837 (GET_TFIAR): Likewise.
1838 (SET_TEXASR): Pass unsigned long instead of unsigned long long.
1839 (SET_TEXASRU): Likewise.
1840 (SET_TFHAR): Likewise.
1841 (SET_TFIAR): Likewise.
1842 (TABORTDC): Likewise.
1843 (TABORTDCI): Likewise.
1844 * config/rs6000/rs6000-call.c (rs6000_expand_new_builtin): Fix error
1845 handling for no32bit. Add 32bit handling for RS6000_BIF_BPERMD.
1846 * config/rs6000/rs6000-gen-builtins.c (write_init_file): Initialize
1847 new_builtins_are_live to 1.
1849 2021-12-02 Vladimir N. Makarov <vmakarov@redhat.com>
1851 PR rtl-optimization/103437
1852 * ira-color.c (setup_allocno_priorities): Use long long
1853 multiplication as backup for overflow processing.
1855 2021-12-02 Tamar Christina <tamar.christina@arm.com>
1857 * config/aarch64/aarch64-protos.h (aarch64_gen_shareable_zero): New.
1858 * config/aarch64/aarch64-simd.md (aarch64_rshrn<mode>,
1859 aarch64_rshrn2<mode>): Generate rounding half-ing add when appropriate.
1860 * config/aarch64/aarch64.c (aarch64_gen_shareable_zero): New.
1862 2021-12-02 Vladimir N. Makarov <vmakarov@redhat.com>
1864 PR rtl-optimization/103437
1865 * ira-color.c (setup_allocno_priorities): Process multiplication
1868 2021-12-02 Martin Liska <mliska@suse.cz>
1870 PR gcov-profile/96092
1871 * common.opt: New option.
1872 * coverage.c (coverage_begin_function): Emit filename with
1873 remap_profile_filename.
1874 * doc/invoke.texi: Document the new option.
1875 * file-prefix-map.c (add_profile_prefix_map): New.
1876 (remap_profile_filename): Likewise.
1877 * file-prefix-map.h (add_profile_prefix_map): Likewise.
1878 (remap_profile_filename): Likewise.
1879 * lto-opts.c (lto_write_options): Handle
1880 OPT_fprofile_prefix_map_.
1881 * opts-global.c (handle_common_deferred_options): Likewise.
1882 * opts.c (common_handle_option): Likewise.
1883 (gen_command_line_string): Likewise.
1884 * profile.c (output_location): Emit filename with
1885 remap_profile_filename.
1887 2021-12-02 Richard Biener <rguenther@suse.de>
1889 PR middle-end/103271
1890 * internal-fn.c (expand_DEFERRED_INIT): When the base
1891 of the LHS is a decl with matching constant size use
1892 that as the initialization target instead of an
1893 eventual VLA typed one.
1895 2021-12-02 Richard Biener <rguenther@suse.de>
1897 PR tree-optimization/103527
1898 * tree-vect-stmts.c (vect_build_gather_load_calls): Always
1899 use a truth type for building the vector mask.
1901 2021-12-02 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
1903 * config/vxworks.h (LTO_PLUGIN_SPEC): Adapt to corrected
1904 spelling of -linker-output-auto-nolto-rel.
1906 2021-12-01 Uroš Bizjak <ubizjak@gmail.com>
1909 * config/i386/sse.md (VI2F): Remove mode iterator.
1910 (VI2F_256_512): New mode iterator.
1911 (vec_set<V8_128:mode>_0): New insn pattern.
1912 (vec_set<VI2F_256_512:mode>_0>): Rename from vec_set<VI2F:mode>mode.
1913 Use VI2F_256_512 mode iterator instead of VI2F.
1914 (*axv512fp16_movsh): Remove.
1915 (<sse2p4_1>_pinsr<ssemodesuffix>): Add (x,x,x) AVX2 alternative.
1916 Do not disable V8HF mode insn on AVX2 targets.
1917 (pinsrw -> pbroadcast + pblendw peephole2): New peephole.
1918 (pinsrw -> pbroadcast + pblendw splitter): New post-reload splitter.
1919 * config/i386/i386.md (extendhfsf): Call gen_vec_setv8hf_0.
1920 * config/i386/i386-expand.c (ix86_expand_vector_set)
1921 <case E_V8HFmode>: Use vec_merge path for TARGET_AVX2.
1923 2021-12-01 Roger Sayle <roger@nextmovesoftware.com>
1924 Richard Biener <rguenther@suse.de>
1926 * tree-ssa-loop-niter.c (number_of_iterations_until_wrap):
1927 Check if simplify_using_initial_conditions allows us to
1928 simplify the expression for may_be_zero.
1929 * match.pd (X != C ? -X : -C -> -X): New transform.
1930 (X != C ? ~X : ~C -> ~X): Likewise.
1931 ((X+1) > Y ? -X : 1 -> X >= Y ? -X : 1): Likewise.
1933 2021-12-01 Peter Bergner <bergner@linux.ibm.com>
1935 PR middle-end/103127
1936 * gimplify.c (is_var_need_auto_init): Handle opaque types.
1938 2021-12-01 Iain Buclaw <ibuclaw@gdcproject.org>
1940 * common.opt (static-libphobos): Add option.
1941 * config/darwin.h (LINK_SPEC): Substitute -lgphobos with libgphobos.a
1942 when linking statically.
1943 * gcc.c (driver_handle_option): Set -static-libphobos as always valid.
1945 2021-12-01 Jason Merrill <jason@redhat.com>
1948 * fold-const.c (maybe_nonzero_address): Use get_create or get
1949 depending on folding_initializer.
1950 (fold_binary_initializer_loc): New.
1951 * fold-const.h (fold_binary_initializer_loc): Declare.
1952 * varasm.c (mark_weak): Don't use the decl location.
1953 * doc/invoke.texi: Document -fconstexpr-fp-except.
1955 2021-12-01 Bill Schmidt <wschmidt@linux.ibm.com>
1958 * config/rs6000/rs6000-c.c (rs6000-builtins.h): Stop including.
1959 (rs6000-internal.h): Include.
1960 (altivec_resolve_new_overloaded_builtin): Move call to
1961 rs6000_invalid_new_builtin here from rs6000_new_builtin_decl.
1962 * config/rs6000/rs6000-call.c (rs6000-builtins.h): Stop including.
1963 (rs6000_invalid_new_builtin): Remove static qualifier.
1964 (rs6000_new_builtin_decl): Remove test for supported builtin.
1965 * config/rs6000/rs6000-internal.h (rs6000-builtins.h): Include.
1966 (rs6000_invalid_new_builtin): Declare.
1967 * config/rs6000/rs6000.c (rs6000-builtins.h): Don't include.
1969 2021-12-01 Aldy Hernandez <aldyh@redhat.com>
1971 PR tree-optimization/103409
1972 * gimple-range-path.cc (path_range_query::compute_ranges_in_phis):
1973 Do all the work with just one ssa_global_cache.
1974 * gimple-range-path.h: Remove m_tmp_phi_cache.
1976 2021-12-01 Richard Sandiford <richard.sandiford@arm.com>
1978 PR tree-optimization/103517
1979 * tree-vect-slp.c (vect_build_slp_tree_1): When allowing two
1980 different component references, check the codes of both them,
1981 rather than just the first.
1983 2021-12-01 Andrew MacLeod <amacleod@redhat.com>
1985 PR tree-optimization/103464
1986 * gimple-range.cc (gimple_ranger::prefill_name): Process phis also.
1987 (gimple_ranger::prefill_stmt_dependencies): Ditto.
1989 2021-12-01 Iain Buclaw <ibuclaw@gdcproject.org>
1991 * doc/install.texi (Prerequisites): Add note that D front end now
1992 requires GDC installed in order to bootstrap.
1993 (Building): Add D compiler section, referencing prerequisites.
1995 2021-12-01 Jakub Jelinek <jakub@redhat.com>
1997 * config/i386/x86-tune.def (X86_TUNE_SCHEDULE,
1998 X86_TUNE_SSE_PARTIAL_REG_FP_CONVERTS_DEPENDENCY,
1999 X86_TUNE_MISALIGNED_MOVE_STRING_PRO_EPILOGUES): Formatting fixes.
2000 (X86_TUNE_USE_GATHER): Put m_GENERIC last for consistency.
2002 2021-12-01 Jakub Jelinek <jakub@redhat.com>
2004 PR rtl-optimization/102356
2005 * rtl.h (simplify_context): Add assoc_count member and
2006 max_assoc_count static member.
2007 * simplify-rtx.c (simplify_associative_operation): Don't reassociate
2008 more than max_assoc_count times within one outermost simplify_* call.
2009 * dwarf2out.c (mem_loc_descriptor): Optimize binary operation
2010 with both operands the same using DW_OP_dup.
2012 2021-12-01 Tamar Christina <tamar.christina@arm.com>
2014 * match.pd: Move below pattern that rewrites to EQ, NE.
2015 * tree.c (bitmask_inv_cst_vector_p): Correct do .. while indentation.
2017 2021-12-01 Siddhesh Poyarekar <siddhesh@gotplt.org>
2018 Martin Liška <mliska@suse.cz>
2020 PR tree-optimization/103456
2021 * tree-object-size.c (merge_object_sizes): Update osi->changed
2022 only if object_sizes_set succeeded.
2024 2021-11-30 liuhongt <hongtao.liu@intel.com>
2028 * config/i386/i386.md (*x86_64_shld_1): Set_attr
2029 length_immediate to 1.
2030 (*x86_shld_1): Ditto.
2031 (*x86_64_shrd_1): Ditto.
2032 (*x86_shrd_1): Ditto.
2034 2021-11-30 Bill Schmidt <wschmidt@linux.ibm.com>
2036 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
2038 (altivec_resolve_new_overloaded_builtin): Likewise.
2040 2021-11-30 Martin Jambor <mjambor@suse.cz>
2043 * ipa-sra.c (scan_function): Also check ECF_LOOPING_CONST_OR_PURE flag.
2045 2021-11-30 Richard Sandiford <richard.sandiford@arm.com>
2047 PR tree-optimization/103494
2048 * tree-vect-stmts.c (vect_get_gather_scatter_ops): Remove ncopies
2049 argument and calculate ncopies from gs_info->offset_vectype
2051 (vectorizable_store, vectorizable_load): Update accordingly.
2053 2021-11-30 Iain Buclaw <ibuclaw@gdcproject.org>
2055 * Makefile.in (GDC): New variable.
2056 (GDCFLAGS): New variable.
2057 * configure: Regenerate.
2058 * configure.ac: Add call to ACX_PROG_GDC. Substitute GDCFLAGS.
2060 2021-11-30 Martin Liska <mliska@suse.cz>
2061 Martin Jambor <mjambor@suse.cz>
2064 * ipa-param-manipulation.c
2065 (ipa_param_body_adjustments::prepare_debug_expressions): Be
2066 careful about hash_map reallocating itself. Simpify a return
2067 which always returns true.
2069 2021-11-30 Andrew MacLeod <amacleod@redhat.com>
2071 PR tree-optimization/103440
2072 * gimple-range-fold.cc (fold_using_range::range_of_phi): Continue
2073 normal param processing for equiv params.
2075 2021-11-30 Richard Biener <rguenther@suse.de>
2077 * config/i386/i386.c (ix86_shift_rotate_cost): Remove
2079 * tree-chrec.c (evolution_function_is_invariant_rec_p):
2081 * tree-if-conv.c (if_convertible_stmt_p): Likewise.
2082 * tree-ssa-pre.c (fully_constant_expression): Likewise.
2083 * tree-vrp.c (operand_less_p): Likewise.
2084 * reload.c (reg_overlap_mentioned_for_reload_p): Remove
2085 unreachable gcc_unreachable ().
2086 * sel-sched-ir.h (bb_next_bb): Likewise.
2087 * varasm.c (compare_constant): Likewise.
2089 2021-11-30 Richard Biener <rguenther@suse.de>
2091 PR tree-optimization/103489
2092 * tree-vect-loop.c (vectorizable_phi): Verify argument
2093 vector type compatibility to mitigate bool pattern recog
2096 2021-11-30 Thomas Schwinge <thomas@codesourcery.com>
2098 * omp-offload.c (oacc_loop_auto_partitions): Remove erroneous
2099 "Orphan reductions cannot have gang partitioning" handling.
2101 2021-11-30 Thomas Schwinge <thomas@codesourcery.com>
2103 * omp-offload.c (oacc_loop_process): Implement "gang reduction on
2104 an orphan loop" checking.
2106 2021-11-30 Cesar Philippidis <cesar@codesourcery.com>
2107 Thomas Schwinge <thomas@codesourcery.com>
2109 * omp-general.h (enum oacc_loop_flags): Add OLF_REDUCTION enum.
2110 * omp-low.c (lower_oacc_head_mark): Use it to mark OpenACC
2112 * omp-offload.c (oacc_loop_auto_partitions): Don't assign gang
2113 level parallelism to orphan reductions.
2115 2021-11-30 Aldy Hernandez <aldyh@redhat.com>
2117 PR tree-optimization/103451
2118 * range-op.cc (operator_div::wi_fold): Remove
2119 can_throw_non_call_exceptions special case.
2120 * tree-ssa-sink.c (sink_code_in_bb): Same.
2122 2021-11-30 Richard Sandiford <richard.sandiford@arm.com>
2124 * tree-vect-slp.c (arg1_arg4_map): New variable.
2125 (vect_get_operand_map): Handle IFN_MASK_GATHER_LOAD.
2126 (vect_build_slp_tree_1): Likewise.
2127 (vect_build_slp_tree_2): Likewise.
2128 * tree-vect-stmts.c (vectorizable_load): Expect the mask to be
2129 the last SLP child node rather than the first.
2131 2021-11-30 Richard Sandiford <richard.sandiford@arm.com>
2133 * tree-if-conv.c: Include tree-eh.h.
2134 (predicate_statements): Remove pe argument. Don't hoist
2136 (combine_blocks): Remove pe argument.
2137 (ifcvt_available_on_edge_p, ifcvt_can_hoist): New functions.
2138 (ifcvt_hoist_invariants): Likewise.
2139 (tree_if_conversion): Update call to combine_blocks. Call
2140 ifcvt_hoist_invariants after VN.
2142 2021-11-30 Richard Sandiford <richard.sandiford@arm.com>
2144 PR tree-optimization/102467
2145 * doc/sourcebuild.texi (vect_gather_load_ifn): Document.
2146 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
2147 Commonize safelen handling. Punt for anything involving
2148 gathers and scatters unless safelen says otherwise.
2149 * tree-vect-slp.c (arg1_map): New variable.
2150 (vect_get_operand_map): Handle IFN_GATHER_LOAD.
2151 (vect_build_slp_tree_1): Likewise.
2152 (vect_build_slp_tree_2): Likewise.
2153 (compatible_calls_p): If vect_get_operand_map returns nonnull,
2154 check that any skipped arguments are equal.
2155 (vect_slp_analyze_node_operations_1): Tighten reduction check.
2156 * tree-vect-stmts.c (check_load_store_for_partial_vectors): Take
2157 an ncopies argument.
2158 (vect_get_gather_scatter_ops): Take slp_node and ncopies arguments.
2160 (vectorizable_store, vectorizable_load): Adjust accordingly.
2162 2021-11-30 Richard Sandiford <richard.sandiford@arm.com>
2164 * gimple.h (gimple_num_args, gimple_arg, gimple_arg_ptr): New
2166 * tree-vect-slp.c (cond_expr_maps, arg2_map): New variables.
2167 (vect_get_operand_map): New function.
2168 (vect_get_and_check_slp_defs): Fix outdated comment.
2169 Use vect_get_operand_map and new gimple argument accessors.
2170 (vect_build_slp_tree_2): Likewise.
2172 2021-11-30 Richard Sandiford <richard.sandiford@arm.com>
2174 * tree-vect-slp.c (vect_build_slp_tree_1): Use code_helper
2175 to record the operations performed by statements, only using
2176 CALL_EXPR for things that don't map to built-in or internal
2177 functions. For shifts, require all shift amounts to be equal
2178 if optab_vector is not supported but optab_scalar is.
2180 2021-11-30 Richard Sandiford <richard.sandiford@arm.com>
2182 * tree-vect-data-refs.c (vect_check_gather_scatter): Continue
2183 processing conversions if the current offset is a pointer.
2185 2021-11-30 Richard Sandiford <richard.sandiford@arm.com>
2187 * tree-vectorizer.h (vect_is_reduction): Use STMT_VINFO_REDUC_IDX.
2189 2021-11-30 Richard Sandiford <richard.sandiford@arm.com>
2191 * genopinit.c (main): Turn supports_vec_gather_load and
2192 supports_vec_scatter_store into signed char arrays and remove
2193 supports_vec_gather_load_cached and supports_vec_scatter_store_cached.
2194 * optabs-query.c (supports_vec_convert_optab_p): Add a mode parameter.
2195 If the mode is not VOIDmode, test only for that mode.
2196 (supports_vec_gather_load_p): Likewise.
2197 (supports_vec_scatter_store_p): Likewise.
2198 * optabs-query.h (supports_vec_gather_load_p): Likewise.
2199 (supports_vec_scatter_store_p): Likewise.
2200 * tree-vect-data-refs.c (vect_check_gather_scatter): Pass the
2201 vector mode to supports_vec_gather_load_p and
2202 supports_vec_scatter_store_p.
2204 2021-11-30 Richard Sandiford <richard.sandiford@arm.com>
2206 * internal-fn.c (commutative_binary_fn_p): Handle IFN_ADD_OVERFLOW
2207 and IFN_MUL_OVERFLOW.
2209 2021-11-30 Richard Sandiford <richard.sandiford@arm.com>
2211 * internal-fn.c (commutative_binary_fn_p): Handle IFN_UBSAN_CHECK_ADD
2212 and IFN_UBSAN_CHECK_MUL.
2214 2021-11-30 Richard Sandiford <richard.sandiford@arm.com>
2216 * internal-fn.c (commutative_binary_fn_p): Handle IFN_COMPLEX_MUL.
2218 2021-11-30 Richard Sandiford <richard.sandiford@arm.com>
2220 * gimple-fold.c: Include internal-fn.h.
2221 (fold_stmt_1): If a function maps to an internal one, use
2222 first_commutative_argument to canonicalize the order of
2223 commutative arguments.
2224 * gimple-match-head.c (gimple_resimplify2, gimple_resimplify3)
2225 (gimple_resimplify4, gimple_resimplify5): Extend commutativity
2226 checks to functions.
2228 2021-11-30 Richard Sandiford <richard.sandiford@arm.com>
2230 * doc/md.texi (reduc_fmin_scal_@var{m}): Document.
2231 (reduc_fmax_scal_@var{m}): Likewise.
2232 * optabs.def (reduc_fmax_scal_optab): New optab.
2233 (reduc_fmin_scal_optab): Likewise
2234 * internal-fn.def (REDUC_FMAX, REDUC_FMIN): New functions.
2235 * tree-vect-loop.c (reduction_fn_for_scalar_code): Handle
2236 CASE_CFN_FMAX and CASE_CFN_FMIN.
2237 (neutral_op_for_reduction): Likewise.
2238 (needs_fold_left_reduction_p): Likewise.
2239 * config/aarch64/iterators.md (FMAXMINV): New iterator.
2240 (fmaxmin): Handle UNSPEC_FMAXNMV and UNSPEC_FMINNMV.
2241 * config/aarch64/aarch64-simd.md (reduc_<optab>_scal_<mode>): Fix
2243 (reduc_<fmaxmin>_scal_<mode>): New pattern.
2244 * config/aarch64/aarch64-sve.md (reduc_<fmaxmin>_scal_<mode>):
2247 2021-11-30 Richard Sandiford <richard.sandiford@arm.com>
2249 * builtins.h (associated_internal_fn): Declare overload that
2250 takes a (combined_cfn, return type) pair.
2251 * builtins.c (associated_internal_fn): Split new overload out
2252 of original fndecl version. Also provide an overload that takes
2253 a (combined_cfn, return type) pair.
2254 * internal-fn.h (commutative_binary_fn_p): Declare.
2255 (commutative_ternary_fn_p): Likewise.
2256 (associative_binary_fn_p): Likewise.
2257 * internal-fn.c (commutative_binary_fn_p, commutative_ternary_fn_p):
2258 New functions, split out from...
2259 (first_commutative_argument): ...here.
2260 (associative_binary_fn_p): New function.
2261 * gimple-match.h (code_helper): Add a constructor that takes
2263 (commutative_binary_op_p): Declare.
2264 (commutative_ternary_op_p): Likewise.
2265 (first_commutative_argument): Likewise.
2266 (associative_binary_op_p): Likewise.
2267 (canonicalize_code): Likewise.
2268 (directly_supported_p): Likewise.
2269 (get_conditional_internal_fn): Likewise.
2270 (gimple_build): New overloads that takes a code_helper.
2271 * gimple-fold.c (gimple_build): Likewise.
2272 * gimple-match-head.c (commutative_binary_op_p): New function.
2273 (commutative_ternary_op_p): Likewise.
2274 (first_commutative_argument): Likewise.
2275 (associative_binary_op_p): Likewise.
2276 (canonicalize_code): Likewise.
2277 (directly_supported_p): Likewise.
2278 (get_conditional_internal_fn): Likewise.
2279 * tree-vectorizer.h: Include gimple-match.h.
2280 (neutral_op_for_reduction): Take a code_helper instead of a tree_code.
2281 (needs_fold_left_reduction_p): Likewise.
2282 (reduction_fn_for_scalar_code): Likewise.
2283 (vect_can_vectorize_without_simd_p): Declare a nNew overload that takes
2285 * tree-vect-loop.c: Include case-cfn-macros.h.
2286 (fold_left_reduction_fn): Take a code_helper instead of a tree_code.
2287 (reduction_fn_for_scalar_code): Likewise.
2288 (neutral_op_for_reduction): Likewise.
2289 (needs_fold_left_reduction_p): Likewise.
2290 (use_mask_by_cond_expr_p): Likewise.
2291 (build_vect_cond_expr): Likewise.
2292 (vect_create_partial_epilog): Likewise. Use gimple_build rather
2293 than gimple_build_assign.
2294 (check_reduction_path): Handle calls and operate on code_helpers
2295 rather than tree_codes.
2296 (vect_is_simple_reduction): Likewise.
2297 (vect_model_reduction_cost): Likewise.
2298 (vect_find_reusable_accumulator): Likewise.
2299 (vect_create_epilog_for_reduction): Likewise.
2300 (vect_transform_cycle_phi): Likewise.
2301 (vectorizable_reduction): Likewise. Make more use of
2303 (vect_transform_reduction): Use gimple_extract_op but expect
2304 a tree_code for now.
2305 (vect_can_vectorize_without_simd_p): New overload that takes
2307 * tree-vect-stmts.c (vectorizable_call): Handle reductions in
2309 * tree-vect-patterns.c (vect_mark_pattern_stmts): Use
2310 gimple_extract_op when updating STMT_VINFO_REDUC_IDX.
2312 2021-11-30 Richard Sandiford <richard.sandiford@arm.com>
2314 * gimple-match.h (code_helper): Provide == and != overloads.
2315 (code_helper::operator tree_code): Make explicit.
2316 (code_helper::operator combined_fn): Likewise.
2317 * gimple-match-head.c (convert_conditional_op): Use explicit
2318 conversions where necessary.
2319 (gimple_resimplify1, gimple_resimplify2, gimple_resimplify3): Likewise.
2320 (maybe_push_res_to_seq, gimple_simplify): Likewise.
2321 * gimple-fold.c (replace_stmt_with_simplification): Likewise.
2323 2021-11-30 Richard Sandiford <richard.sandiford@arm.com>
2325 * gimple-match.h (code_helper): Add functions for querying whether
2326 the code represents an internal_fn or a built_in_function.
2327 Provide explicit conversion operators for both cases.
2328 (gimple_extract_op): Declare.
2329 * gimple-match-head.c (gimple_extract): New function, extracted from...
2330 (gimple_simplify): ...here.
2331 (gimple_extract_op): New function.
2333 2021-11-30 Eric Botcazou <ebotcazou@adacore.com>
2336 * config/i386/i386.c (ix86_output_call_insn): Beef up comment about
2337 nops emitted with SEH.
2338 * config/i386/winnt.c (i386_pe_seh_unwind_emit): When switching to
2339 the cold section, emit a nop before the directive if the previous
2340 active instruction is a call.
2342 2021-11-30 Roger Sayle <roger@nextmovesoftware.com>
2343 Uroš Bizjak <ubizjak@gmail.com>
2346 * config/i386/sse.md (define_split): New splitters to simplify
2347 not;vec_duplicate;and as vec_duplicate;andn.
2349 2021-11-30 Richard Biener <rguenther@suse.de>
2351 PR middle-end/103485
2352 * match.pd (-((int)x >> (prec - 1)) to (unsigned)x >> (prec - 1)):
2353 Use VIEW_CONVERT_EXPR for vectors.
2355 2021-11-30 Richard Biener <rguenther@suse.de>
2357 * cfgrtl.c (skip_insns_after_block): Refactor code to
2358 be more easily readable.
2359 * expr.c (op_by_pieces_d::run): Remove unreachable
2361 * sched-deps.c (sched_analyze): Remove unreachable
2363 * sel-sched-ir.c (in_same_ebb_p): Likewise.
2364 * tree-ssa-alias.c (nonoverlapping_refs_since_match_p):
2365 Remove unreachable code.
2366 * tree-vect-slp.c (vectorize_slp_instance_root_stmt):
2367 Refactor to avoid unreachable loop iteration.
2368 * tree.c (walk_tree_1): Remove unreachable break.
2369 * vec-perm-indices.c (vec_perm_indices::series_p): Remove
2372 2021-11-30 Kewen Lin <linkw@linux.ibm.com>
2375 * config/rs6000/rs6000-call.c (rs6000_builtin_decl): Remove builtin mask
2378 2021-11-30 Kewen Lin <linkw@linux.ibm.com>
2380 * config/rs6000/rs6000.c
2381 (rs6000_cost_data::update_target_cost_per_stmt): Adjust the way to
2382 compute extra penalized cost. Remove useless parameter.
2383 (rs6000_cost_data::rs6000_add_stmt_cost): Adjust the call to function
2384 update_target_cost_per_stmt.
2386 2021-11-30 Kewen Lin <linkw@linux.ibm.com>
2389 2021-11-17 Kewen Lin <linkw@linux.ibm.com>
2391 * config/visium/visium.md (*add<mode>3_insn, *addsi3_insn, *addi3_insn,
2392 *sub<mode>3_insn, *subsi3_insn, *subdi3_insn, *neg<mode>2_insn,
2393 *negdi2_insn, *and<mode>3_insn, *ior<mode>3_insn, *xor<mode>3_insn,
2394 *one_cmpl<mode>2_insn, *ashl<mode>3_insn, *ashr<mode>3_insn,
2395 *lshr<mode>3_insn, *trunchiqi2_insn, *truncsihi2_insn,
2396 *truncdisi2_insn, *extendqihi2_insn, *extendqisi2_insn,
2397 *extendhisi2_insn, *extendsidi2_insn, *zero_extendqihi2_insn,
2398 *zero_extendqisi2_insn, *zero_extendsidi2_insn): Fix split condition.
2400 2021-11-30 Andrew MacLeod <amacleod@redhat.com>
2402 PR tree-optimization/103467
2403 * gimple-range-gori.cc (range_def_chain::register_dependency): Don't
2404 use an object reference after a potential resize.
2406 2021-11-29 Uroš Bizjak <ubizjak@gmail.com>
2409 * config/i386/i386.md (*movhi_internal): Introduce (*v,C) alternative.
2410 Do not allocate non-GPR registers. Optimize xmm->xmm moves when
2411 optimizing for size. Fix vpinsrw insn template.
2412 (*movhf_internal): Fix pinsrw and pextrw insn templates for
2413 AVX targets. Use sselog1 type instead of sselog. Optimize GPR moves.
2414 Optimize xmm->xmm moves for TARGET_SSE_PARTIAL_REG_DEPENDENCY
2415 and TARGET_SSE_SPLIT_REGS targets.
2417 2021-11-29 Eric Gallager <egallager@gcc.gnu.org>
2419 * configure: Re-regenerate.
2421 2021-11-29 Eric Gallager <egallager@gcc.gnu.org>
2424 * Makefile.in: Substitute CTAGS, ETAGS, and CSCOPE
2425 variables. Use ETAGS variable in TAGS target.
2426 * configure: Regenerate.
2427 * configure.ac: Allow CTAGS, ETAGS, and CSCOPE
2428 variables to be overridden.
2430 2021-11-29 Paul A. Clarke <pc@us.ibm.com>
2432 * config/rs6000/xmmintrin.h (_mm_movemask_ps): Use vec_extractm
2434 * config/rs6000/emmintrin.h (_mm_movemask_pd): Likewise.
2435 (_mm_movemask_epi8): Likewise.
2437 2021-11-29 Richard Biener <rguenther@suse.de>
2439 * read-rtl-function.c (function_reader::read_rtx_operand):
2440 Return only after resetting m_in_call_function_usage.
2442 2021-11-29 Roger Sayle <roger@nextmovesoftware.com>
2444 * config/i386/i386-expand.c (ix86_expand_v1ti_to_ti): Perform the
2445 conversion via V2DImode using vec_extractv2didi on TARGET_SSE2.
2446 * config/i386/sse.md (rotlv1ti3, rotrv1ti3): Change constraint
2447 on QImode shift amounts from const_int_operand to general_operand.
2449 2021-11-29 Richard Biener <rguenther@suse.de>
2451 * tree.h (reverse_storage_order_for_component_p): Remove
2452 spurious gcc_unreachable.
2453 * cfganal.c (dfs_find_deadend): Likewise.
2454 * fold-const-call.c (fold_const_logb): Likewise.
2455 (fold_const_significand): Likewise.
2456 * gimple-ssa-store-merging.c (lhs_valid_for_store_merging_p):
2459 2021-11-29 Richard Biener <rguenther@suse.de>
2461 * vec.c (qsort_chk): Do not return the void return value
2462 from the noreturn qsort_chk_error.
2463 * ccmp.c (expand_ccmp_expr_1): Remove unreachable return.
2464 * df-scan.c (df_ref_equal_p): Likewise.
2465 * dwarf2out.c (is_base_type): Likewise.
2466 (add_const_value_attribute): Likewise.
2467 * fixed-value.c (fixed_arithmetic): Likewise.
2468 * gimple-fold.c (gimple_fold_builtin_fputs): Likewise.
2469 * gimple-ssa-strength-reduction.c (stmt_cost): Likewise.
2470 * graphite-isl-ast-to-gimple.c
2471 (gcc_expression_from_isl_expr_op): Likewise.
2472 (gcc_expression_from_isl_expression): Likewise.
2473 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate):
2475 * lto-streamer-in.c (lto_input_mode_table): Likewise.
2477 2021-11-29 liuhongt <hongtao.liu@intel.com>
2480 * config/i386/i386.c (inline_secondary_memory_needed): HImode
2481 move between GPR and SSE registers is supported under
2482 TARGET_SSE2 and above.
2483 * config/i386/i386.md (extendhfsf2): Optimize expander.
2484 (truncsfhf2): Ditto.
2485 * config/i386/sse.md (sse2p4_1): Adjust attr for V8HFmode to
2486 align with V8HImode.
2488 2021-11-29 liuhongt <hongtao.liu@intel.com>
2492 * config/i386/i386.c (ix86_secondary_reload): Without
2493 TARGET_SSE4_1, General register is needed to move HImode from
2494 sse register to memory.
2495 * config/i386/sse.md (*vec_extrachf): Use %vpextrw instead of
2496 pextrw in output templates.
2497 * config/i386/i386.md (movhi_internal): Ditto, also fix typo of
2498 MEM_P (operands[1]) and adjust mode/prefix/type attribute for
2499 alternatives related to sse register.
2501 2021-11-29 Richard Biener <rguenther@suse.de>
2503 PR tree-optimization/103458
2504 * tree-ssa-dce.c (make_forwarders_with_degenerate_phis): Do not
2505 create forwarders for blocks with abnormal predecessors.
2507 2021-11-29 Richard Biener <rguenther@suse.de>
2509 * gimple-predicate-analysis.cc (can_be_invalidated_p):
2510 Restore semantics to the one before the split from
2513 2021-11-28 Jan Hubicka <hubicka@ucw.cz>
2515 * profile.c: Include sreal.h
2516 (struct bb_stats): New.
2517 (cmp_stats): New function.
2518 (compute_branch_probabilities): Output bb stats.
2520 2021-11-28 Jan Hubicka <hubicka@ucw.cz>
2522 * cfghooks.c: Include sreal.h, profile.h.
2523 (profile_record_check_consistency): Fix checking of count counsistency;
2524 record also dynamic mismatches.
2525 * cfgrtl.c (rtl_account_profile_record): Similarly.
2526 * tree-cfg.c (gimple_account_profile_record): Likewise.
2527 * cfghooks.h (struct profile_record): Remove num_mismatched_freq_in,
2528 num_mismatched_freq_out, turn time to double, add
2529 dyn_mismatched_prob_out, dyn_mismatched_count_in,
2530 num_mismatched_prob_out; remove num_mismatched_count_out.
2531 * passes.c (account_profile_1): New function.
2532 (account_profile_in_list): New function.
2533 (pass_manager::dump_profile_report): Rewrite.
2534 (execute_one_ipa_transform_pass): Check profile consistency after
2536 (execute_all_ipa_transforms): Remove cfun test; record all transform
2538 (execute_one_pass): Fix collecting of profile stats.
2540 2021-11-27 Jakub Jelinek <jakub@redhat.com>
2542 PR middle-end/103431
2543 * config/i386/i386.md (x86_64_shld, x86_shld, x86_64_shrd, x86_shrd):
2544 Change insn pattern to accurately describe the instructions.
2545 (*x86_64_shld_1, *x86_shld_1, *x86_64_shrd_1, *x86_shrd_1): New
2546 define_insn patterns.
2547 (*x86_64_shld_2, *x86_shld_2, *x86_64_shrd_2, *x86_shrd_2): New
2548 define_insn_and_split patterns.
2549 (*ashl<dwi>3_doubleword_mask, *ashl<dwi>3_doubleword_mask_1,
2550 *<insn><dwi>3_doubleword_mask, *<insn><dwi>3_doubleword_mask_1,
2551 ix86_rotl<dwi>3_doubleword, ix86_rotr<dwi>3_doubleword): Adjust
2552 splitters for x86_{,64_}sh{l,r}d pattern changes.
2554 2021-11-27 Jakub Jelinek <jakub@redhat.com>
2556 PR tree-optimization/103435
2557 * gimple-ssa-store-merging.c (find_bswap_or_nop_finalize): Avoid UB if
2558 n->range - rsize == 8, just clear both *cmpnop and *cmpxchg in that
2561 2021-11-27 Martin Jambor <mjambor@suse.cz>
2564 * ipa-prop.c (ipcp_transform_function): Call
2565 delete_unreachable_blocks_update_callgraph instead of returning
2568 2021-11-26 Siddhesh Poyarekar <siddhesh@gotplt.org>
2570 * tree-object-size.c (object_sizes_grow, object_sizes_release,
2571 object_sizes_unknown_p, object_sizes_get, object_size_set_force,
2572 object_sizes_set): New functions.
2573 (addr_object_size, compute_builtin_object_size,
2574 expr_object_size, call_object_size, unknown_object_size,
2575 merge_object_sizes, plus_stmt_object_size,
2576 cond_expr_object_size, collect_object_sizes_for,
2577 check_for_plus_in_loops_1, init_object_sizes,
2578 fini_object_sizes): Adjust.
2580 2021-11-26 Siddhesh Poyarekar <siddhesh@gotplt.org>
2582 * tree-object-size.c: New enum.
2583 (object_sizes, computed, addr_object_size,
2584 compute_builtin_object_size, expr_object_size, call_object_size,
2585 merge_object_sizes, plus_stmt_object_size,
2586 collect_object_sizes_for, init_object_sizes, fini_object_sizes,
2587 object_sizes_execute): Replace magic numbers with enums.
2589 2021-11-26 Roger Sayle <roger@nextmovesoftware.com>
2591 * tree-ssa-loop-ivopts.c (cand_value_at): Take a class
2592 tree_niter_desc* argument instead of just a tree for NITER.
2593 If we require the iv candidate value at the end of the final
2594 loop iteration, try using the original loop bound as the
2595 NITER for sufficiently simple loops.
2596 (may_eliminate_iv): Update (only) call to cand_value_at.
2598 2021-11-26 Jan Hubicka <hubicka@ucw.cz>
2600 * ipa-modref.c (analyze_function): Drop parameter F and dump
2601 cgraph node name rather than cfun name.
2602 (modref_generate): Update.
2603 (modref_summaries::insert):Update.
2604 (modref_summaries_lto::insert):Update.
2605 (pass_modref::execute):Update.
2606 (ipa_merge_modref_summary_after_inlining): Improve combining of
2609 2021-11-26 Jan Hubicka <jh@suse.cz>
2612 * ipa-modref.c (update_escape_summary_1): Fix handling of min_flags.
2614 2021-11-26 konglin1 <lingling.kong@intel.com>
2617 * config/i386/i386.c (ix86_can_change_mode_class): Allow 16 bit data in XMM register
2619 * config/i386/i386.md (extendhfsf2): Add extenndhfsf2 for TARGET_F16C.
2620 (extendhfdf2): Restrict extendhfdf for TARGET_AVX512FP16 only.
2621 (*extendhf<mode>2): Rename from extendhf<mode>2.
2622 (truncsfhf2): Likewise.
2623 (truncdfhf2): Likewise.
2624 (*trunc<mode>2): Likewise.
2626 2021-11-26 liuhongt <hongtao.liu@intel.com>
2628 PR middle-end/103419
2629 * match.pd: Fix typo, use the type of second parameter, not
2632 2021-11-25 Jan Hubicka <jh@suse.cz>
2634 * ipa-cp.c (ipa_value_range_from_jfunc): Remove forgotten early return.
2636 2021-11-25 Roger Sayle <roger@nextmovesoftware.com>
2638 PR middle-end/103406
2639 * match.pd (minus @0 @0): Check tree_expr_maybe_infinite_p.
2641 2021-11-25 Martin Jambor <mjambor@suse.cz>
2644 * ipa-prop.h (ipa_get_param): New overload. Move bits of the existing
2646 * ipa-param-manipulation.h (ipa_param_adjustments): New member
2647 function get_updated_index_or_split.
2648 * ipa-param-manipulation.c
2649 (ipa_param_adjustments::get_updated_index_or_split): New function.
2650 * ipa-prop.c (adjust_agg_replacement_values): Reimplement, add
2651 capability to identify scalarized parameters and perform substitution
2653 (ipcp_transform_function): Create descriptors earlier, handle new
2654 return values of adjust_agg_replacement_values.
2656 2021-11-25 Aldy Hernandez <aldyh@redhat.com>
2658 * gimple-range-path.cc (path_range_query::compute_ranges_defined): Remove.
2659 (path_range_query::compute_ranges_in_block): Revert to bitmap order.
2660 * gimple-range-path.h: Remove compute_ranges_defined.
2662 2021-11-25 Andrew Stubbs <ams@codesourcery.com>
2665 * config/gcn/gcn.c (move_callee_saved_registers): Ensure that the
2666 number of spilled registers is counted correctly.
2668 2021-11-25 Jan Hubicka <hubicka@ucw.cz>
2670 * ipa-prop.h (ipa_node_params::ipa_node_params): Initialize
2673 2021-11-25 Andrew MacLeod <amacleod@redhat.com>
2675 PR tree-optimization/103359
2676 * gimple-range-fold.cc (fold_using_range::range_of_phi): If arg is
2677 equivalent to def, don't initially include it's range.
2679 2021-11-25 Jan Hubicka <hubicka@ucw.cz>
2681 * tree-ssa-alias.c (ref_maybe_used_by_call_p_1): Do not check
2682 gimple_static_chain.
2684 2021-11-25 Richard Biener <rguenther@suse.de>
2686 * cgraphunit.c (symbol_table::output_weakrefs): Remove
2688 (get_alias_symbol): Remove now unused function.
2690 2021-11-25 Richard Biener <rguenther@suse.de>
2692 * cfgrtl.c (rtl_verify_fallthru): Do not stop verifying
2694 (skip_insns_after_block): Remove unreachable break and continue.
2696 2021-11-25 Richard Biener <rguenther@suse.de>
2698 * cfgexpand.c (label_rtx_for_bb): Remove dead loop construct.
2700 2021-11-25 Richard Biener <rguenther@suse.de>
2702 * regset.h (REG_SET_EMPTY_P): New macro.
2703 * cfgcleanup.c (thread_jump): Use REG_SET_EMPTY_P.
2705 2021-11-25 Martin Liska <mliska@suse.cz>
2707 * doc/invoke.texi: Use @option for -Wuninitialized.
2709 2021-11-25 Aldy Hernandez <aldyh@redhat.com>
2711 PR tree-optimization/103254
2712 * gimple-range-path.cc (path_range_query::compute_ranges): Move
2713 exported boolean code...
2714 (path_range_query::compute_imports): ...here.
2716 2021-11-25 Aldy Hernandez <aldyh@redhat.com>
2718 PR tree-optimization/103254
2719 * gimple-range-path.cc (path_range_query::compute_ranges_defined): New
2720 (path_range_query::compute_ranges_in_block): Move to
2721 compute_ranges_defined.
2722 * gimple-range-path.h (compute_ranges_defined): New.
2724 2021-11-25 Jakub Jelinek <jakub@redhat.com>
2726 PR tree-optimization/103417
2727 * match.pd ((X & Y) CMP 0): Only handle eq and ne. Commonalize
2730 2021-11-25 Jakub Jelinek <jakub@redhat.com>
2732 PR tree-optimization/103376
2733 * gimple-ssa-store-merging.c (perform_symbolic_merge): For
2734 BIT_IOR_EXPR, if masked1 && masked2 && masked1 != masked2, don't
2735 punt, but set the corresponding result byte to MARKER_BYTE_UNKNOWN.
2736 For BIT_XOR_EXPR similarly and if masked1 == masked2 and the
2737 byte isn't MARKER_BYTE_UNKNOWN, set the corresponding result byte to
2740 2021-11-25 Jakub Jelinek <jakub@redhat.com>
2743 * doc/invoke.texi (-Wcomma-subscript): Document that for
2744 -std=c++20 the option isn't enabled by default with -Wno-deprecated
2745 but for -std=c++23 it is.
2747 2021-11-24 Jan Hubicka <hubicka@ucw.cz>
2749 * ipa-modref.c (implicit_eaf_flags_for_edge_and_arg): Break out from...
2750 (modref_merge_call_site_flags): ... here.
2751 (ipa_merge_modref_summary_after_inlining): Use it.
2753 2021-11-24 Thomas Schwinge <thomas@codesourcery.com>
2755 * cfgloop.c (verify_loop_structure): Reduce scope of
2756 'class loop *loop' variable.
2757 * ipa-fnsummary.c (analyze_function_body): Likewise.
2758 * loop-init.c (fix_loop_structure): Likewise.
2759 * loop-invariant.c (calculate_loop_reg_pressure): Likewise.
2760 * predict.c (predict_loops): Likewise.
2761 * tree-loop-distribution.c (loop_distribution::execute): Likewise.
2762 * tree-vectorizer.c (pass_vectorize::execute): Likewise.
2764 2021-11-24 Andrew MacLeod <amacleod@redhat.com>
2766 PR tree-optimization/103231
2767 * gimple-range.cc (gimple_ranger::gimple_ranger): Create stmt stack.
2768 (gimple_ranger::gimple_ranger): Delete stmt stack.
2769 (gimple_ranger::range_of_stmt): Process depenedencies if they have no
2771 (gimple_ranger::prefill_name): New.
2772 (gimple_ranger::prefill_stmt_dependencies): New.
2773 * gimple-range.h (class gimple_ranger): Add prototypes.
2775 2021-11-24 Andrew MacLeod <amacleod@redhat.com>
2777 * gimple-range-cache.cc (ranger_cache::get_global_range): Always
2778 return a range, return if it came from the cache or not.
2779 (get_non_stale_global_range): Rename to get_global_range, and return
2780 the temporal state in a flag.
2781 * gimple-range-cache.h (get_non_stale_global_range): Rename and adjust.
2782 * gimple-range.cc (gimple_ranger::range_of_expr): No need to query
2784 (gimple_ranger::range_of_stmt): Adjust for global cache temporal state
2787 2021-11-24 Andrew MacLeod <amacleod@redhat.com>
2789 * gimple-range.cc (gimple_ranger::range_on_edge): Call trailer when
2790 a constant is encountered to terminate the trace.
2792 2021-11-24 Maciej W. Rozycki <macro@embecosm.com>
2794 PR middle-end/103059
2795 * reload.c (find_reloads_address_1): Also accept the ASHIFT form
2796 of indexed addressing.
2797 (find_reloads): Adjust accordingly.
2799 2021-11-24 Richard Biener <rguenther@suse.de>
2800 Jan Hubicka <jh@suse.cz>
2802 PR tree-optimization/103168
2803 * ipa-modref.h (struct modref_summary): Add load_accesses.
2804 * ipa-modref.c (modref_summary::finalize): Initialize load_accesses.
2805 * tree-ssa-sccvn.c (visit_reference_op_call): Use modref
2806 info to walk the virtual use->def chain to CSE const/pure
2807 function calls possibly reading from memory.
2809 2021-11-24 Thomas Schwinge <thomas@codesourcery.com>
2811 PR middle-end/103244
2812 * gimplify.c (gimplify_adjust_omp_clauses): Restore previous
2815 2021-11-24 Richard Biener <rguenther@suse.de>
2817 PR middle-end/103193
2818 * match.pd: Avoid canonicalizing (le/ge @0 @0) to (eq @0 @0)
2819 with NaNs and -ftrapping-math.
2821 2021-11-24 Jakub Jelinek <jakub@redhat.com>
2823 PR middle-end/103384
2824 * omp-general.c (omp_context_selector_matches): For ACCEL_COMPILER,
2825 return 0 for kind(host) and continue for kind(nohost).
2827 2021-11-24 Jakub Jelinek <jakub@redhat.com>
2829 PR middle-end/103365
2830 * attribs.h (lookup_attribute): Allow attr_name to start with
2831 underscore, as long as canonicalize_attr_name returns false.
2832 (lookup_attribute_by_prefix): Don't call get_attribute_name twice.
2833 * attribs.c (extract_attribute_substring): Reimplement using
2834 canonicalize_attr_name.
2835 (register_scoped_attribute): Change gcc_assert into
2836 gcc_checking_assert, verify !canonicalize_attr_name rather than
2837 that str.str doesn't start with '_'.
2839 2021-11-24 Jakub Jelinek <jakub@redhat.com>
2841 PR tree-optimization/103376
2842 * gimple-ssa-store-merging.c (perform_symbolic_merge): Add CODE
2843 argument. If CODE is not BIT_IOR_EXPR, ensure that one of masked1
2845 (find_bswap_or_nop_1, find_bswap_or_nop,
2846 imm_store_chain_info::try_coalesce_bswap): Adjust
2847 perform_symbolic_merge callers.
2849 2021-11-24 Richard Biener <rguenther@suse.de>
2851 * tree-ssa-loop-ivopts.c (find_givs): Take loop body as
2852 argument instead of re-computing it.
2853 (find_interesting_uses): Likewise.
2854 (find_induction_variables): Pass through loop body.
2855 (tree_ssa_iv_optimize_loop): Pass down loop body.
2857 2021-11-24 Tamar Christina <tamar.christina@arm.com>
2859 * tree-ssa-phiopt.c (spaceship_replacement): Handle new canonical
2862 2021-11-24 Tamar Christina <tamar.christina@arm.com>
2864 * tree.c (bitmask_inv_cst_vector_p): New.
2865 * tree.h (bitmask_inv_cst_vector_p): New.
2866 * match.pd: Use it in new bitmask compare pattern.
2868 2021-11-24 Jason Merrill <jason@redhat.com>
2870 * timevar.h (class auto_cond_timevar): New.
2872 2021-11-24 Hongtao Liu <hongtao.liu@intel.com>
2873 H.J. Lu <hongjiu.lu@intel.com>
2875 PR tree-optimization/103194
2876 * match.pd (gimple_nop_atomic_bit_test_and_p): Extended to
2878 * tree-ssa-ccp.c (gimple_nop_convert): Declare.
2879 (optimize_atomic_bit_test_and): Enhance
2880 optimize_atomic_bit_test_and to handle truncation.
2882 2021-11-23 Martin Sebor <msebor@redhat.com>
2885 * Makefile.in (OBJS): Add gimple-warn-recursion.o.
2886 * common.opt: Add -Winfinite-recursion.
2887 * doc/invoke.texi (-Winfinite-recursion): Document.
2888 * passes.def (pass_warn_recursion): Schedule a new pass.
2889 * tree-pass.h (make_pass_warn_recursion): Declare.
2890 * gimple-warn-recursion.c: New file.
2892 2021-11-23 Bill Schmidt <wschmidt@linux.ibm.com>
2894 * config/rs6000/rs6000-call.c (rs6000_invalid_new_builtin): Change
2895 error messages for ENB_P8V and ENB_P9V.
2897 2021-11-23 Bill Schmidt <wschmidt@linux.ibm.com>
2899 * config/rs6000/rs6000-builtin-new.def: Add power6-64 stanza. Move
2900 CMPB to power6-64 stanza.
2901 * config/rs6000/rs6000-call.c (rs6000_invalid_new_builtin): Handle
2903 (rs6000_new_builtin_is_supported): Likewise.
2904 (rs6000_expand_new_builtin): Likewise. Clean up formatting.
2905 (rs6000_init_builtins): Handle ENB_P6_64 case.
2906 * config/rs6000/rs6000-gen-builtins.c (bif_stanza): Add BSTZ_P6_64.
2907 (stanza_map): Add entry mapping power6-64 to BSTZ_P6_64.
2908 (enable_string): Add "ENB_P6_64".
2909 (write_decls): Add ENB_P6_64 to bif_enable enum.
2911 2021-11-23 Jan Hubicka <jh@suse.cz>
2913 * ipa-modref-tree.h (struct modref_tree): Remove max_bases, max_refs
2915 (modref_tree::modref_tree): Remove parametr.
2916 (modref_tree::insert_base): Add max_bases parameter.
2917 (modref_tree::insert): Add max_bases, max_refs, max_accesses
2919 (modref_tree::insert): New member function.
2920 (modref_tree::merge): Add max_bases, max_refs, max_accesses
2922 (modref_tree::insert): New member function.
2923 * ipa-modref-tree.c (test_insert_search_collapse): Update.
2924 (test_merge): Update.
2925 * ipa-modref.c (dump_records): Don't dump max_refs and max_bases.
2926 (dump_lto_records): Likewise.
2927 (modref_summary::finalize): Fix whitespace.
2928 (get_modref_function_summary): Likewise.
2929 (modref_access_analysis::record_access): Update.
2930 (modref_access_analysis::record_access_lto): Update.
2931 (modref_access_analysis::process_fnspec): Update.
2932 (analyze_function): Update.
2933 (modref_summaries::duplicate): Update.
2934 (modref_summaries_lto::duplicate): Update.
2935 (write_modref_records): Update.
2936 (read_modref_records): Update.
2937 (read_section): Update.
2938 (propagate_unknown_call): Update.
2939 (modref_propagate_in_scc): Update.
2940 (ipa_merge_modref_summary_after_inlining): Update.
2942 2021-11-23 Martin Liska <mliska@suse.cz>
2944 * doc/invoke.texi: Remove 2 more duplicite param descriptions.
2946 2021-11-23 Richard Biener <rguenther@suse.de>
2948 PR tree-optimization/103361
2949 * gimple-loop-jam.c (adjust_unroll_factor): Use lambda_int
2950 for the dependence distance.
2951 * tree-data-ref.c (print_lambda_vector): Properly print a lambda_int.
2953 2021-11-23 Jakub Jelinek <jakub@redhat.com>
2955 * tree-inline.h (struct copy_body_data): Remove
2956 transform_lang_insert_block member.
2957 * tree-inline.c (remap_block): Don't call
2958 id->transform_lang_insert_block.
2959 (optimize_inline_calls, copy_gimple_seq_and_replace_locals,
2960 tree_function_versioning, maybe_inline_call_in_expr,
2961 copy_fn): Don't initialize id.transform_lang_insert_block.
2962 * gimplify.c (gimplify_omp_loop): Likewise.
2964 2021-11-23 Jan Hubicka <hubicka@ucw.cz>
2966 PR tree-optimization/103335
2967 * tree-ssa-dse.c (valid_ao_ref_for_dse): Rename to ...
2968 (valid_ao_ref_kill_for_dse): ... this; do not check that boundaries
2969 are divisible by BITS_PER_UNIT.
2970 (get_byte_aligned_range_containing_ref): New function.
2971 (get_byte_aligned_range_contained_in_ref): New function.
2972 (normalize_ref): Rename to ...
2973 (get_byte_range): ... this one; handle accesses not aligned to byte
2974 boundary; return range in bytes rater than updating ao_ref.
2975 (clear_live_bytes_for_ref): Take write ref by reference; simplify using
2977 (setup_live_bytes_from_ref): Likewise.
2978 (clear_bytes_written_by): Update.
2979 (live_bytes_read): Update.
2980 (dse_classify_store): Simplify tech before live_bytes_read checks.
2982 2021-11-23 Andrew Pinski <apinski@marvell.com>
2984 PR tree-optimization/102216
2985 * gimple-fold.c (fold_stmt_1): Add canonicalization
2986 of "&MEM[ssa_n, CST]" to "ssa_n p+ CST", note this
2987 can only be done if !in_place.
2989 2021-11-23 Jakub Jelinek <jakub@redhat.com>
2991 PR middle-end/102431
2992 * gimplify.c (replace_reduction_placeholders): Remove.
2993 (note_no_context_vars): New function.
2994 (gimplify_omp_loop): For OMP_PARALLEL's BIND_EXPR create a new
2995 BLOCK. Use copy_tree_body_r with walk_tree instead of unshare_expr
2996 and replace_reduction_placeholders for duplication of
2997 OMP_CLAUSE_REDUCTION_{INIT,MERGE} expressions. Ensure all mentioned
2998 automatic vars have DECL_CONTEXT set to non-NULL before doing so
2999 and reset it afterwards for those vars and their corresponding
3002 2021-11-23 Haochen Gui <guihaoc@gcc.gnu.org>
3005 * config/rs6000/altivec.md (altivec_vreve<mode>2 for VEC_K): Use
3006 xxbrq for v16qi, xxbrq + xxbrh for v8hi and xxbrq + xxbrw for v4si
3007 or v4sf when p9_vector is set.
3008 (altivec_vreve<mode>2 for VEC_64): Defined. Implemented by xxswapd.
3010 2021-11-23 Navid Rahimi <navidrahimi@microsoft.com>
3012 PR tree-optimization/102232
3013 * match.pd (x * (1 + y / x) - y) -> (x - y % x): New optimization.
3015 2021-11-23 Navid Rahimi <navidrahimi@microsoft.com>
3017 PR tree-optimization/96779
3018 * match.pd (-x == x) -> (x == 0): New optimization.
3020 2021-11-22 Roger Sayle <roger@nextmovesoftware.com>
3022 PR tree-optimization/98953
3023 PR tree-optimization/103345
3024 * gimple-ssa-store-merging.c (find_bswap_or_nop_1): Handle
3025 BIT_XOR_EXPR and PLUS_EXPR the same as BIT_IOR_EXPR.
3026 (pass_optimize_bswap::execute): Likewise.
3028 2021-11-22 Martin Liska <mliska@suse.cz>
3030 * doc/invoke.texi: Remove duplicate documentation for 3 params.
3032 2021-11-22 Richard Biener <rguenther@suse.de>
3034 PR tree-optimization/103351
3035 * tree-ssa-dce.c (sort_phi_args): Sort after e->dest_idx as
3038 2021-11-22 liuhongt <hongtao.liu@intel.com>
3041 * config/i386/constraints.md (Bk): New
3042 define_memory_constraint.
3043 * config/i386/i386-protos.h (ix86_gpr_tls_address_pattern_p):
3045 * config/i386/i386.c (ix86_gpr_tls_address_pattern_p): New
3047 * config/i386/i386.md (*movsi_internal): Don't allow
3048 mask/sse/mmx move in TLS code sequences.
3049 (*movdi_internal): Ditto.
3051 2021-11-22 Kewen Lin <linkw@linux.ibm.com>
3053 * config/xtensa/xtensa.md (movdi_internal, movdf_internal): Fix split
3056 2021-11-21 Jakub Jelinek <jakub@redhat.com>
3059 * config/i386/i386-options.c (ix86_valid_target_attribute_p): If
3060 fndecl already has DECL_FUNCTION_SPECIFIC_TARGET, use that as base
3061 instead of target_option_default_node.
3063 2021-11-21 Jan Hubicka <hubicka@ucw.cz>
3066 * ipa-modref.c (parm_map_for_arg): Rename to ...
3067 (parm_map_for_ptr): .. this one; handle static chain and calls to
3069 (modref_access_analysis::get_access): Use parm_map_for_ptr.
3070 (modref_access_analysis::process_fnspec): Update.
3071 (modref_access_analysis::analyze_load): Update.
3072 (modref_access_analysis::analyze_store): Update.
3074 2021-11-21 Jan Hubicka <hubicka@ucw.cz>
3076 * ipa-modref.c (ignore_nondeterminism_p): Move earlier in source
3078 (ignore_retval_p): Likewise.
3079 (ignore_stores_p): Likewise.
3080 (parm_map_for_arg): Likewise.
3081 (class modref_access_analysis): New class.
3082 (modref_access_analysis::set_side_effects): New member function.
3083 (modref_access_analysis::set_nondeterministic): New member function.
3084 (get_access): Turn to ...
3085 (modref_access_analysis::get_access): ... this one.
3086 (record_access): Turn to ...
3087 (modref_access_analysis::record_access): ... this one.
3088 (record_access_lto): Turn to ...
3089 (modref_access_analysis::record_access_lto): ... This one.
3090 (record_access_p): Turn to ...
3091 (modref_access_analysis::record_access_p): ... This one
3092 (modref_access_analysis::record_unknown_load): New member function.
3093 (modref_access_analysis::record_unknown_store): New member function.
3094 (get_access_for_fnspec): Turn to ...
3095 (modref_access_analysis::get_access_for_fnspec): ... this one.
3096 (merge_call_side_effects): Turn to ...
3097 (moderf_access_analysis::merge_call_side_effects): Turn to ...
3098 (collapse_loads): Move later in source code.
3099 (collapse_stores): Move later in source code.
3100 (process_fnspec): Turn to ...
3101 (modref_access_analysis::process_fnspec): ... this one.
3102 (analyze_call): Turn to ...
3103 (modref_access_analysis::analyze_call): ... this one.
3104 (struct summary_ptrs): Remove.
3105 (analyze_load): Turn to ...
3106 (modref_access_analysis::analyze_load): ... this one.
3107 (analyze_store): Turn to ...
3108 (modref_access_analysis::analyze_store): ... this one.
3109 (analyze_stmt): Turn to ...
3110 (modref_access_analysis::analyze_stmt): ... This one.
3111 (remove_summary): Remove.
3112 (modref_access_analysis::propagate): Break out from ...
3113 (modref_access_analysis::analyze): Break out from ...
3114 (analyze_function): ... here.
3116 2021-11-21 Roger Sayle <roger@nextmovesoftware.com>
3117 Robin Dapp <rdapp@linux.ibm.com>
3120 * tree-ssa-math-opts.c (convert_mult_to_widen): Recognize
3121 signed WIDEN_MULT_EXPR if the target supports umul_widen_optab.
3123 2021-11-20 Jan Hubicka <hubicka@ucw.cz>
3126 * ipa-modref.c (ignore_nondeterminism_p): Allow looping pure/cont.
3127 (merge_call_side_effects): Improve debug output.
3129 2021-11-20 Jan Hubicka <hubicka@ucw.cz>
3132 * ipa-pure-const.c (propagate_pure_const): Fix merging of loping flag.
3134 2021-11-20 Jeff Law <jeffreyalaw@gmail.com>
3136 PR tree-optimization/103226
3137 * config/bfin/bfin.md (doloop pattern, splitter and expander): Clobber
3140 2021-11-20 Andrew Pinski <apinski@marvell.com>
3142 PR tree-optimization/103220
3143 * match.pd ((type) X bitop CST): Don't check if CST
3144 fits into the type if only the sign changes.
3146 2021-11-20 Alexandre Oliva <oliva@adacore.com>
3148 PR tree-optimization/102988
3149 * gimple-harden-conditionals.cc (detach_value): Copy SSA_NAME
3150 without decl sharing.
3152 2021-11-19 Iain Sandoe <iain@sandoe.co.uk>
3154 * doc/invoke.texi: Remove whitespace after an @option.
3156 2021-11-19 Paul A. Clarke <pc@us.ibm.com>
3158 * config/rs6000/emmintrin.h (_mm_sad_epu8): Use vec_absd when
3159 _ARCH_PWR9, optimize vec_sum2s when LE.
3161 2021-11-19 Iain Sandoe <iain@sandoe.co.uk>
3164 * config/darwin-driver.c (darwin_driver_init): Handle exported
3165 symbols and symbol lists (suppress automatic export of the TLS
3167 * config/darwin.c (darwin_rename_builtins): Remove workaround.
3168 * config/darwin.h (LINK_GCC_C_SEQUENCE_SPEC): Likewise.
3169 (REAL_LIBGCC_SPEC): Handle revised library uses.
3170 * config/darwin.opt (nodefaultexport): New.
3171 * config/i386/darwin.h (PR80556_WORKAROUND): Remove.
3172 * config/i386/darwin32-biarch.h (PR80556_WORKAROUND): Likewise.
3173 * config/i386/darwin64-biarch.h (PR80556_WORKAROUND): Likewise.
3175 2021-11-19 Martin Jambor <mjambor@suse.cz>
3177 * opts.c (default_options_table): Switch off
3178 flag_semantic_interposition at Ofast.
3179 * doc/invoke.texi (Optimize Options): Document that Ofast switches off
3180 -fsemantic-interposition.
3182 2021-11-19 Jan Hubicka <hubicka@ucw.cz>
3184 * tree-ssa-alias.c (ref_maybe_used_by_call_p_1): Do not guard modref
3185 by !gimple_call_chain.
3187 2021-11-19 Martin Sebor <msebor@redhat.com>
3191 * doc/invoke.texi (-Waddress): Update.
3193 2021-11-19 Andrew MacLeod <amacleod@redhat.com>
3195 PR tree-optimization/103254
3196 * gimple-range-gori.cc (range_def_chain::get_def_chain): Limit the
3197 depth for all statements with multple ssa names.
3199 2021-11-19 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
3201 * config/s390/s390.md (define_peephole2): Variable insn points
3202 to the first matched insn. Use peep2_next_insn(1) to refer to
3203 the second matched insn.
3205 2021-11-19 Tamar Christina <tamar.christina@arm.com>
3207 PR tree-optimization/103311
3209 * tree-vect-slp-patterns.c (vect_validate_multiplication): Fix CONJ
3210 test to new codegen.
3211 (complex_mul_pattern::matches): Move check downwards.
3213 2021-11-19 Martin Liska <mliska@suse.cz>
3216 2021-11-19 Martin Liska <mliska@suse.cz>
3218 * cfgexpand.c (pass_expand::execute): Use option directly.
3219 * function.c (allocate_struct_function): Likewise.
3220 * gimple-low.c (lower_function_body): Likewise.
3221 (lower_stmt): Likewise.
3222 * gimple-ssa-backprop.c (backprop::prepare_change): Likewise.
3223 * ipa-param-manipulation.c (ipa_param_adjustments::modify_call): Likewise.
3224 * ipa-split.c (split_function): Likewise.
3225 * lto-streamer-in.c (input_function): Likewise.
3226 * sese.c (sese_insert_phis_for_liveouts): Likewise.
3227 * ssa-iterators.h (num_imm_uses): Likewise.
3228 * tree-cfg.c (make_blocks): Likewise.
3229 (gimple_merge_blocks): Likewise.
3230 * tree-inline.c (tree_function_versioning): Likewise.
3231 * tree-loop-distribution.c (generate_loops_for_partition): Likewise.
3232 * tree-sra.c (analyze_access_subtree): Likewise.
3233 * tree-ssa-dce.c (remove_dead_stmt): Likewise.
3234 * tree-ssa-loop-ivopts.c (remove_unused_ivs): Likewise.
3235 * tree-ssa-phiopt.c (spaceship_replacement): Likewise.
3236 * tree-ssa-reassoc.c (reassoc_remove_stmt): Likewise.
3237 * tree-ssa-tail-merge.c (tail_merge_optimize): Likewise.
3238 * tree-ssa-threadedge.c (propagate_threaded_block_debug_into): Likewise.
3239 * tree-ssa.c (gimple_replace_ssa_lhs): Likewise.
3240 (target_for_debug_bind): Likewise.
3241 (insert_debug_temp_for_var_def): Likewise.
3242 (insert_debug_temps_for_defs): Likewise.
3243 (reset_debug_uses): Likewise.
3244 * tree-ssanames.c (release_ssa_name_fn): Likewise.
3245 * tree-vect-loop-manip.c (adjust_vec_debug_stmts): Likewise.
3246 (adjust_debug_stmts): Likewise.
3247 (adjust_phi_and_debug_stmts): Likewise.
3248 (vect_do_peeling): Likewise.
3249 * tree-vect-loop.c (vect_transform_loop_stmt): Likewise.
3250 (vect_transform_loop): Likewise.
3251 * tree.h (MAY_HAVE_DEBUG_MARKER_STMTS): Remove
3252 (MAY_HAVE_DEBUG_BIND_STMTS): Remove.
3253 (MAY_HAVE_DEBUG_STMTS): Use options directly.
3255 2021-11-19 Giuliano Belinassi <gbelinassi@suse.de>
3257 * gcc.c (process_command): Skip dumpdir override if file is a
3259 * doc/invoke.texi: Update -dumpdir documentation.
3261 2021-11-19 Andrew Pinski <apinski@marvell.com>
3263 PR tree-optimization/103314
3264 * match.pd ((type) X op CST): Restrict the equal
3265 TYPE_PRECISION case to GIMPLE only.
3267 2021-11-19 Martin Liska <mliska@suse.cz>
3270 * ipa-modref-tree.h (struct modref_parm_map): Add default
3272 * ipa-modref.c (ipa_merge_modref_summary_after_inlining): Use it.
3274 2021-11-19 Richard Biener <rguenther@suse.de>
3276 PR middle-end/103248
3277 * tree-eh.c (operation_could_trap_helper_p): Properly handle
3278 fixed-point RDIV_EXPR.
3280 2021-11-19 Richard Biener <rguenther@suse.de>
3282 PR tree-optimization/102436
3283 * tree-ssa-loop-im.c (execute_sm_if_changed): Add mode
3284 to just create the if structure and return the then block.
3285 (execute_sm): Add flag to indicate the var will re-use
3287 (hoist_memory_references): Support a single conditional
3288 block with all stores as special case.
3290 2021-11-19 Andrew Pinski <apinski@marvell.com>
3292 PR tree-optimization/103317
3293 * tree-ssa-phiopt.c (minmax_replacement): For the non empty
3294 middle bb case, check to make sure it has a single predecessor.
3296 2021-11-19 Andrew Pinski <apinski@marvell.com>
3298 PR tree-optimization/103257
3300 ((m1 >/</>=/<= m2) * d -> (m1 >/</>=/<= m2) ? d : 0):
3301 Disable until !canonicalize_math_p ().
3303 2021-11-19 Marek Polacek <polacek@redhat.com>
3307 * doc/invoke.texi: Update documentation for -Wuninitialized.
3308 * tree.c (stabilize_reference): Set location.
3310 2021-11-19 liuhongt <hongtao.liu@intel.com>
3313 * config/i386/x86-tune-costs.h (skylake_cost): Reduce cost of
3314 storing 256/512-bit SSE register to be equal to cost of
3315 unaligned store to avoid odd alignment peeling.
3316 (icelake_cost): Ditto.
3318 2021-11-18 David Edelsohn <dje.gcc@gmail.com>
3320 * config/rs6000/predicates.md (current_file_function_operand):
3321 Add flag_semantic_interposition to call of decl_replaceable_p.
3323 2021-11-18 Martin Liska <mliska@suse.cz>
3325 * ipa-modref.c (analyze_function): Do not execute the code
3326 only if dump_file != NULL.
3328 2021-11-18 Martin Liska <mliska@suse.cz>
3330 * ipa-modref.c (analyze_function): Use fnode instead of repeated
3331 cgraph_node::get (current_function_decl).
3333 2021-11-18 Jan Hubicka <hubicka@ucw.cz>
3335 * cgraph.c (cgraph_node::get_availability): Update call of
3337 (cgraph_node::verify_node): Verify that semantic_interposition flag
3339 * cgraph.h: (symtab_node): Add semantic_interposition flag.
3340 * cgraphclones.c (set_new_clone_decl_and_node_flags): Clear
3341 semantic_interposition flag.
3342 * cgraphunit.c (cgraph_node::finalize_function): Set
3343 semantic_interposition flag.
3344 (cgraph_node::add_new_function): Likewise.
3345 (varpool_node::finalize_decl): Likewise.
3346 (cgraph_node::create_wrapper): Likewise.
3347 * common.opt (fsemantic-interposition): Turn to optimization node.
3348 * lto-cgraph.c (lto_output_node): Stream semantic_interposition.
3349 (lto_output_varpool_node): Likewise.
3350 (input_overwrite_node): Likewise.
3351 (input_varpool_node): Likewise.
3352 * symtab.c (symtab_node::dump_base): Dump new flag.
3353 * varasm.c (decl_replaceable_p): Add semantic_interposition_p
3355 * varasm.h (decl_replaceable_p): Update declaration.
3356 * varpool.c (varpool_node::ctor_useable_for_folding_p):
3357 Use semantic_interposition flag.
3358 (varpool_node::get_availability): Likewise.
3359 (varpool_node::create_alias): Copy semantic_interposition flag.
3361 2021-11-18 Jan Hubicka <hubicka@ucw.cz>
3364 * ipa-modref.c (modref_eaf_analysis::merge_call_lhs_flags): Unused
3365 parameter may still be returned.
3366 (modref_eaf_analysis::analyze_ssa_name): Call merge_call_lhs_flags
3367 even for unused function args.
3369 2021-11-18 Tamar Christina <tamar.christina@arm.com>
3371 PR tree-optimization/103311
3372 * tree-vect-slp-patterns.c (complex_mul_pattern::matches,
3373 complex_fms_pattern::matches): Check for multiplications.
3375 2021-11-18 H.J. Lu <hjl.tools@gmail.com>
3378 * config/i386/i386.c (ix86_output_jmp_thunk_or_indirect): Emit
3379 CS prefix for -mindirect-branch-cs-prefix.
3380 (ix86_output_indirect_branch_via_reg): Likewise.
3381 * config/i386/i386.opt: Add -mindirect-branch-cs-prefix.
3382 * doc/invoke.texi: Document -mindirect-branch-cs-prefix.
3384 2021-11-18 David Edelsohn <dje.gcc@gmail.com>
3386 * config/rs6000/driver-rs6000.c (detect_processor_aix): Add
3389 2021-11-18 Tamar Christina <tamar.christina@arm.com>
3391 PR tree-optimization/103253
3392 * tree-ssa-math-opts.c (convert_mult_to_fma): Check for LHS.
3394 2021-11-18 Matthias Kretz <m.kretz@gsi.de>
3396 * doc/extend.texi: Document __builtin_assoc_barrier.
3398 2021-11-18 Martin Liska <mliska@suse.cz>
3400 * cfgexpand.c (pass_expand::execute): Use option directly.
3401 * function.c (allocate_struct_function): Likewise.
3402 * gimple-low.c (lower_function_body): Likewise.
3403 (lower_stmt): Likewise.
3404 * gimple-ssa-backprop.c (backprop::prepare_change): Likewise.
3405 * ipa-param-manipulation.c (ipa_param_adjustments::modify_call): Likewise.
3406 * ipa-split.c (split_function): Likewise.
3407 * lto-streamer-in.c (input_function): Likewise.
3408 * sese.c (sese_insert_phis_for_liveouts): Likewise.
3409 * ssa-iterators.h (num_imm_uses): Likewise.
3410 * tree-cfg.c (make_blocks): Likewise.
3411 (gimple_merge_blocks): Likewise.
3412 * tree-inline.c (tree_function_versioning): Likewise.
3413 * tree-loop-distribution.c (generate_loops_for_partition): Likewise.
3414 * tree-sra.c (analyze_access_subtree): Likewise.
3415 * tree-ssa-dce.c (remove_dead_stmt): Likewise.
3416 * tree-ssa-loop-ivopts.c (remove_unused_ivs): Likewise.
3417 * tree-ssa-phiopt.c (spaceship_replacement): Likewise.
3418 * tree-ssa-reassoc.c (reassoc_remove_stmt): Likewise.
3419 * tree-ssa-tail-merge.c (tail_merge_optimize): Likewise.
3420 * tree-ssa-threadedge.c (propagate_threaded_block_debug_into): Likewise.
3421 * tree-ssa.c (gimple_replace_ssa_lhs): Likewise.
3422 (target_for_debug_bind): Likewise.
3423 (insert_debug_temp_for_var_def): Likewise.
3424 (insert_debug_temps_for_defs): Likewise.
3425 (reset_debug_uses): Likewise.
3426 * tree-ssanames.c (release_ssa_name_fn): Likewise.
3427 * tree-vect-loop-manip.c (adjust_vec_debug_stmts): Likewise.
3428 (adjust_debug_stmts): Likewise.
3429 (adjust_phi_and_debug_stmts): Likewise.
3430 (vect_do_peeling): Likewise.
3431 * tree-vect-loop.c (vect_transform_loop_stmt): Likewise.
3432 (vect_transform_loop): Likewise.
3433 * tree.h (MAY_HAVE_DEBUG_MARKER_STMTS): Remove
3434 (MAY_HAVE_DEBUG_BIND_STMTS): Remove.
3435 (MAY_HAVE_DEBUG_STMTS): Use options directly.
3437 2021-11-18 Richard Biener <rguenther@suse.de>
3439 PR tree-optimization/103277
3440 * tree-ssa-dse.c (need_ab_cleanup): New.
3441 (dse_optimize_redundant_stores): Adjust.
3442 (delete_dead_or_redundant_assignment): Get extra
3443 need_ab_cleanup argument and set when abnormal cleanup is
3445 (dse_optimize_call): Adjust.
3446 (dse_optimize_stmt): Likewise.
3447 (pass_dse::execute): Allocate and deallocate need_ab_cleanup.
3448 Perform abnormal cleanup.
3449 * tree-ssa-dse.h (delete_dead_or_redundant_assignment): Adjust.
3451 2021-11-18 Hongyu Wang <hongyu.wang@intel.com>
3453 * config/i386/i386-expand.c (ix86_expand_atomic_fetch_op_loop):
3454 Adjust generated cfg to avoid infinite loop.
3456 2021-11-18 konglin1 <lingling.kong@intel.com>
3458 * config/i386/avx512fp16intrin.h (_mm512_mul_pch): Add alias for _mm512_fmul_pch.
3459 (_mm512_mask_mul_pch): Likewise.
3460 (_mm512_maskz_mul_pch): Likewise.
3461 (_mm512_mul_round_pch): Likewise.
3462 (_mm512_mask_mul_round_pch): Likewise.
3463 (_mm512_maskz_mul_round_pch): Likewise.
3464 (_mm512_cmul_pch): Likewise.
3465 (_mm512_mask_cmul_pch): Likewise.
3466 (_mm512_maskz_cmul_pch): Likewise.
3467 (_mm512_cmul_round_pch): Likewise.
3468 (_mm512_mask_cmul_round_pch): Likewise.
3469 (_mm512_maskz_cmul_round_pch): Likewise.
3470 (_mm_mul_sch): Likewise.
3471 (_mm_mask_mul_sch): Likewise.
3472 (_mm_maskz_mul_sch): Likewise.
3473 (_mm_mul_round_sch): Likewise.
3474 (_mm_mask_mul_round_sch): Likewise.
3475 (_mm_maskz_mul_round_sch): Likewise.
3476 (_mm_cmul_sch): Likewise.
3477 (_mm_mask_cmul_sch): Likewise.
3478 (_mm_maskz_cmul_sch): Likewise.
3479 (_mm_cmul_round_sch): Likewise.
3480 (_mm_mask_cmul_round_sch): Likewise.
3481 (_mm_maskz_cmul_round_sch): Likewise.
3482 * config/i386/avx512fp16vlintrin.h (_mm_mul_pch): Likewise.
3483 (_mm_mask_mul_pch): Likewise.
3484 (_mm_maskz_mul_pch): Likewise.
3485 (_mm256_mul_pch): Likewise.
3486 (_mm256_mask_mul_pch): Likewise.
3487 (_mm256_maskz_mul_pch): Likewise.
3488 (_mm_cmul_pch): Likewise.
3489 (_mm_mask_cmul_pch): Likewise.
3490 (_mm_maskz_cmul_pch): Likewise.
3491 (_mm256_cmul_pch): Likewise.
3492 (_mm256_mask_cmul_pch): Likewise.
3493 (_mm256_maskz_cmul_pch): Likewise.
3495 2021-11-17 Andrew Pinski <apinski@marvell.com>
3497 PR tree-optimization/103228
3498 PR tree-optimization/55177
3499 * match.pd ((type) X bitop CST): Also do this
3500 transformation for nop conversions.
3502 2021-11-17 Martin Sebor <msebor@redhat.com>
3504 PR tree-optimization/102759
3505 * gimple-array-bounds.cc (build_printable_array_type): Move...
3506 * gimple-ssa-warn-access.cc (build_printable_array_type): Avoid
3507 pathological function redeclarations that remove a previously
3509 Improve formatting of function arguments in informational notes.
3510 * pointer-query.cc (build_printable_array_type): ...to here.
3511 * pointer-query.h (build_printable_array_type): Declared.
3513 2021-11-17 H.J. Lu <hjl.tools@gmail.com>
3516 * config/i386/i386-opts.h (harden_sls): New enum.
3517 * config/i386/i386.c (output_indirect_thunk): Mitigate against
3518 SLS for function return.
3519 (ix86_output_function_return): Likewise.
3520 (ix86_output_jmp_thunk_or_indirect): Mitigate against indirect
3522 (ix86_output_indirect_jmp): Likewise.
3523 (ix86_output_call_insn): Likewise.
3524 * config/i386/i386.opt: Add -mharden-sls=.
3525 * doc/invoke.texi: Document -mharden-sls=.
3527 2021-11-17 H.J. Lu <hjl.tools@gmail.com>
3530 * config/i386/i386.c (ix86_code_end): Remove "%!" before ret.
3531 (ix86_output_function_return): Likewise.
3532 * config/i386/i386.md (simple_return_pop_internal): Likewise.
3534 2021-11-17 Jan Hubicka <hubicka@ucw.cz>
3537 * ipa-modref.c (read_modref_records): Fix streaminig in of every_access
3540 2021-11-17 Uroš Bizjak <ubizjak@gmail.com>
3542 * config/i386/i386.c (indirect_thunks_used): Redefine as HARD_REG_SET.
3543 (ix86_code_end): Use TEST_HARD_REG_BIT on indirect_thunks_used.
3544 (ix86_output_indirect_branch_via_reg): Use SET_HARD_REG_BIT
3545 on indirect_thunks_used.
3546 (ix86_output_indirect_function_return): Ditto.
3548 2021-11-17 Jan Hubicka <hubicka@ucw.cz>
3550 * ipa-modref-tree.c: Include cgraph.h and tree-streamer.h.
3551 (modref_access_node::stream_out): New member function.
3552 (modref_access_node::stream_in): New member function.
3553 * ipa-modref-tree.h (modref_access_node::stream_out,
3554 modref_access_node::stream_in): Declare.
3555 * ipa-modref.c (modref_summary_lto::useful_p): Free useless kills.
3556 (modref_summary_lto::dump): Dump kills.
3557 (analyze_store): Record kills for LTO
3558 (analyze_stmt): Likewise.
3559 (modref_summaries_lto::duplicate): Duplicate kills.
3560 (write_modref_records): Use new stream_out member function.
3561 (read_modref_records): Likewise.
3562 (modref_write): Stream out kills.
3563 (read_section): Stream in kills
3564 (remap_kills): New function.
3565 (update_signature): Use it.
3567 2021-11-17 Uroš Bizjak <ubizjak@gmail.com>
3569 * config/i386/i386.h (LEGACY_SSE_REGNO_P): New predicate.
3570 (SSE_REGNO_P): Use LEGACY_SSE_REGNO_P predicate.
3571 * config/i386/i386.c (zero_all_vector_registers):
3572 Use LEGACY_SSE_REGNO_P predicate.
3573 (ix86_register_priority): Use REX_INT_REGNO_P, REX_SSE_REGNO_P
3574 and EXT_REG_SSE_REGNO_P predicates.
3575 (ix86_hard_regno_call_part_clobbered): Use REX_SSE_REGNO_P
3576 and LEGACY_SSE_REGNO_P predicates.
3578 2021-11-17 Jason Merrill <jason@redhat.com>
3580 * doc/invoke.texi (C++ Dialect Options): Document
3581 -fimplicit-constexpr.
3583 2021-11-17 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
3585 * config/aarch64/aarch64-modes.def (VECTOR_MODE): New V8DI mode.
3586 * config/aarch64/aarch64.c (aarch64_hard_regno_mode_ok): Handle
3588 * config/aarch64/iterators.md (define_mode_attr nunits): Add entry
3591 2021-11-17 Martin Uecker <uecker@gcc.gnu.org>
3595 * gimplify.c (gimplify_var_or_parm_decl): Update comment.
3596 (gimplify_compound_lval): Gimplify base expression first.
3597 (gimplify_target_expr): Add comment.
3599 2021-11-17 Jakub Jelinek <jakub@redhat.com>
3601 PR tree-optimization/103192
3602 * tree-ssa-loop-im.c (move_computations_worker): Use
3603 reset_flow_sensitive_info instead of manually clearing
3604 SSA_NAME_RANGE_INFO and do it for all SSA_NAMEs, not just ones
3605 with integral types.
3607 2021-11-17 Jakub Jelinek <jakub@redhat.com>
3609 PR tree-optimization/103255
3610 * gimple-range-fold.cc (fold_using_range::range_of_address): Return
3611 range_nonzero rather than unadjusted base's range. Formatting fixes.
3613 2021-11-17 Richard Sandiford <richard.sandiford@arm.com>
3615 * doc/md.texi (cond_fmin@var{mode}, cond_fmax@var{mode}): Document.
3616 * optabs.def (cond_fmin_optab, cond_fmax_optab): New optabs.
3617 * internal-fn.def (COND_FMIN, COND_FMAX): New functions.
3618 * internal-fn.c (first_commutative_argument): Handle them.
3619 (FOR_EACH_COND_FN_PAIR): Likewise.
3620 * match.pd (UNCOND_BINARY, COND_BINARY): Likewise.
3621 * config/aarch64/aarch64-sve.md (cond_<fmaxmin><mode>): New
3624 2021-11-17 Kewen Lin <linkw@linux.ibm.com>
3626 * config/i386/i386.md (*add<dwi>3_doubleword, *addv<dwi>4_doubleword,
3627 *addv<dwi>4_doubleword_1, *sub<dwi>3_doubleword,
3628 *subv<dwi>4_doubleword, *subv<dwi>4_doubleword_1,
3629 *add<dwi>3_doubleword_cc_overflow_1, *divmodsi4_const,
3630 *neg<dwi>2_doubleword, *tls_dynamic_gnu2_combine_64_<mode>): Fix split
3633 2021-11-17 Andrew Pinski <apinski@marvell.com>
3635 PR tree-optimization/103288
3636 * tree-ssa-phiopt.c (value_replacement): Return early if middle
3637 block has more than one pred.
3639 2021-11-17 Kewen Lin <linkw@linux.ibm.com>
3641 * config/visium/visium.md (*add<mode>3_insn, *addsi3_insn, *addi3_insn,
3642 *sub<mode>3_insn, *subsi3_insn, *subdi3_insn, *neg<mode>2_insn,
3643 *negdi2_insn, *and<mode>3_insn, *ior<mode>3_insn, *xor<mode>3_insn,
3644 *one_cmpl<mode>2_insn, *ashl<mode>3_insn, *ashr<mode>3_insn,
3645 *lshr<mode>3_insn, *trunchiqi2_insn, *truncsihi2_insn,
3646 *truncdisi2_insn, *extendqihi2_insn, *extendqisi2_insn,
3647 *extendhisi2_insn, *extendsidi2_insn, *zero_extendqihi2_insn,
3648 *zero_extendqisi2_insn, *zero_extendsidi2_insn): Fix split condition.
3650 2021-11-17 Marek Polacek <polacek@redhat.com>
3652 PR preprocessor/103026
3653 * doc/invoke.texi: Document -Wbidi-chars.
3655 2021-11-17 Jan Hubicka <hubicka@ucw.cz>
3658 * ipa-modref.c (ipa_merge_modref_summary_after_inlining): Fix clearing
3661 2021-11-16 Jan Hubicka <jh@suse.cz>
3663 * ipa-modref.c (get_modref_function_summary): Declare.
3664 * ipa-modref.h (get_modref_function_summary): New function.
3665 * tree-ssa-dse.c (clear_live_bytes_for_ref): Break out from ...
3666 (clear_bytes_written_by): ... here; also clear memory killed by
3669 2021-11-16 Iain Sandoe <iain@sandoe.co.uk>
3671 * ggc-common.c (gt_pch_save): If we cannot find a suitable
3672 memory segment for save, then error-out, do not try to
3674 (gt_pch_restore): Save the existing line table, and when
3675 the replacement is being read, use that when constructing
3678 2021-11-16 Peter Bergner <bergner@linux.ibm.com>
3681 * config/rs6000/mma.md (*vsx_assemble_pair): Add early-clobber for
3683 (*mma_assemble_acc): Likewise.
3685 2021-11-16 Martin Sebor <msebor@redhat.com>
3687 PR tree-optimization/102960
3688 * gimple-fold.c (get_range_strlen): Take bitmap as an argument rather
3689 than a pointer to it.
3690 (get_range_strlen_tree): Same. Remove bitmap allocation. Use
3692 (get_maxval_strlen): Use an auto_bitmap.
3693 * tree-ssa-strlen.c (get_range_strlen_dynamic): Factor out PHI
3695 (get_range_strlen_phi): ...into this function.
3696 Avoid assuming maximum string length is constant
3697 (printf_strlen_execute): Dump pointer query cache contents when
3698 details are requisted.
3700 2021-11-16 Jason Merrill <jason@redhat.com>
3702 * langhooks.h (struct lang_hooks): Adjust comment.
3703 * print-tree.c (print_node): Also call print_xnode hook for
3706 2021-11-16 Andrew Pinski <apinski@marvell.com>
3708 PR tree-optimization/103218
3709 * match.pd: New pattern for "((type)(a<0)) << SIGNBITOFA".
3711 2021-11-16 Claudiu Zissulescu <claziss@synopsys.com>
3713 * config/arc/arc.md (maddhisi4): Use a single move to accumulator.
3714 (umaddhisi4): Likewise.
3715 (machi): Update pattern.
3718 2021-11-16 Richard Biener <rguenther@suse.de>
3720 PR tree-optimization/102880
3721 * tree-ssa-dce.c (sort_phi_args): New function.
3722 (make_forwarders_with_degenerate_phis): Likewise.
3723 (perform_tree_ssa_dce): Call
3724 make_forwarders_with_degenerate_phis.
3726 2021-11-16 Richard Biener <rguenther@suse.de>
3728 PR tree-optimization/102880
3729 * tree-ssa-phiopt.c (tree_ssa_phiopt_worker): Push
3730 single_pred (bb1) condition to places that really need it.
3731 (match_simplify_replacement): Likewise.
3732 (value_replacement): Likewise.
3733 (replace_phi_edge_with_variable): Deal with extra edges
3736 2021-11-16 Martin Jambor <mjambor@suse.cz>
3738 * cfgexpand.c (expand_gimple_basic_block): Use build_debug_expr_decl,
3739 add a fixme note about the mode assignment perhaps being unnecessary.
3740 * ipa-param-manipulation.c (ipa_param_adjustments::modify_call):
3742 (ipa_param_body_adjustments::mark_dead_statements): Likewise.
3743 (ipa_param_body_adjustments::reset_debug_stmts): Likewise.
3744 * tree-inline.c (remap_ssa_name): Likewise.
3745 (tree_function_versioning): Likewise.
3746 * tree-into-ssa.c (rewrite_debug_stmt_uses): Likewise.
3747 * tree-ssa-loop-ivopts.c (remove_unused_ivs): Likewise.
3748 * tree-ssa.c (insert_debug_temp_for_var_def): Likewise.
3750 2021-11-16 Jakub Jelinek <jakub@redhat.com>
3752 PR tree-optimization/103208
3753 * omp-expand.c (expand_omp_build_cond): New function.
3754 (expand_omp_for_init_counts, expand_omp_for_init_vars,
3755 expand_omp_for_static_nochunk, expand_omp_for_static_chunk): Use it.
3757 2021-11-16 Jakub Jelinek <jakub@redhat.com>
3759 PR tree-optimization/102009
3760 * gimple-ssa-warn-access.cc (pass_waccess::check_alloc_size_call):
3761 Punt if any of alloc_size arguments is out of bounds vs. number of
3764 2021-11-16 Roger Sayle <roger@nextmovesoftware.com>
3766 * config/i386/i386.md (*bmi2_rorx<mode3>_1): Make conditional
3767 on !optimize_function_for_size_p.
3768 (*<any_rotate><mode>3_1): Add preferred_for_size attribute.
3769 (define_splits): Conditionalize on !optimize_function_for_size_p.
3770 (*bmi2_rorxsi3_1_zext): Likewise.
3771 (*<any_rotate>si2_1_zext): Add preferred_for_size attribute.
3772 (define_splits): Conditionalize on !optimize_function_for_size_p.
3774 2021-11-16 Jan Hubicka <jh@suse.cz>
3777 * ipa-modref.c (merge_call_side_effects): Fix uninitialized
3780 2021-11-16 Andrew Pinski <apinski@marvell.com>
3782 PR tree-optimization/103245
3783 * match.pd: Combine the abs pattern matching using multiplication.
3784 Adding optional nop_convert too.
3786 2021-11-16 H.J. Lu <hjl.tools@gmail.com>
3788 PR middle-end/103268
3789 * tree-ssa-ccp.c (optimize_atomic_bit_test_and): Add a missing
3792 2021-11-15 Siddhesh Poyarekar <siddhesh@gotplt.org>
3794 * gimple-fold.c (gimple_fold_builtin_strncat): Use ranges to
3795 determine if it is safe to transform to strcat.
3796 (gimple_fold_builtin_snprintf): Likewise.
3798 2021-11-15 Siddhesh Poyarekar <siddhesh@gotplt.org>
3800 * gimple-fold.c (known_lower): New function.
3801 (gimple_fold_builtin_strncat_chk,
3802 gimple_fold_builtin_memory_chk, gimple_fold_builtin_stxcpy_chk,
3803 gimple_fold_builtin_stxncpy_chk,
3804 gimple_fold_builtin_snprintf_chk,
3805 gimple_fold_builtin_sprintf_chk): Use it.
3807 2021-11-15 Siddhesh Poyarekar <siddhesh@gotplt.org>
3809 * gimple-fold.c (dump_transformation): New function.
3810 (gimple_fold_builtin_stxcpy_chk,
3811 gimple_fold_builtin_stxncpy_chk): Use it. Simplify to
3812 BUILT_IN_STRNCPY if return value is not used.
3814 2021-11-15 H.J. Lu <hjl.tools@gmail.com>
3816 PR middle-end/103184
3817 * tree-ssa-ccp.c (optimize_atomic_bit_test_and): Check optab
3818 before transforming equivalent, but slighly different cases to
3819 their canonical forms.
3821 2021-11-15 Iain Sandoe <iain@sandoe.co.uk>
3824 * config/darwin.h (TARGET_DTORS_FROM_CXA_ATEXIT): New.
3825 * doc/tm.texi: Regenerated.
3826 * doc/tm.texi.in: Add TARGET_DTORS_FROM_CXA_ATEXIT hook.
3827 * ipa.c (cgraph_build_static_cdtor_1): Return the built
3829 (build_cxa_atexit_decl): New.
3830 (build_dso_handle_decl): New.
3831 (build_cxa_dtor_registrations): New.
3832 (compare_cdtor_tu_order): New.
3833 (build_cxa_atexit_fns): New.
3834 (ipa_cdtor_merge): If dtors_from_cxa_atexit is set,
3835 process the DTORs/CTORs accordingly.
3836 (pass_ipa_cdtor_merge::gate): Also run if
3837 dtors_from_cxa_atexit is set.
3838 * target.def (dtors_from_cxa_atexit): New hook.
3840 2021-11-15 Iain Sandoe <iain@sandoe.co.uk>
3842 * config.in: Regenerate.
3843 * configure: Regenerate.
3844 * configure.ac: Test ld64 for -platform-version support.
3846 2021-11-15 Thomas Schwinge <thomas@codesourcery.com>
3848 * diagnostic-spec.h (typedef xint_hash_t)
3849 (typedef xint_hash_map_t): Replace with...
3850 (typedef nowarn_map_t): ... this.
3851 (nowarn_map): Adjust.
3852 * diagnostic-spec.c (nowarn_map, suppress_warning_at): Likewise.
3854 2021-11-15 Thomas Schwinge <thomas@codesourcery.com>
3856 * profile.c (branch_prob): Use 'location_hash' for
3859 2021-11-15 Aldy Hernandez <aldyh@redhat.com>
3861 PR tree-optimization/103207
3862 * value-range.cc (irange::set): Drop overflow.
3864 2021-11-15 Jakub Jelinek <jakub@redhat.com>
3866 * gimplify.c (optimize_target_teams): Only add OMP_CLAUSE_THREAD_LIMIT
3867 to OMP_TARGET_CLAUSES if it isn't there already.
3869 2021-11-15 Aldy Hernandez <aldyh@redhat.com>
3871 * gimple-range-path.cc (path_range_query::internal_range_of_expr):
3872 Remove useless code.
3873 (path_range_query::ssa_defined_in_bb): New.
3874 (path_range_query::ssa_range_in_phi): Avoid fold_range call that
3875 could trigger additional lookups.
3876 Do not use the cache for ARGs defined in this block.
3877 (path_range_query::compute_ranges_in_block): Use ssa_defined_in_bb.
3878 (path_range_query::maybe_register_phi_relation): Same.
3879 (path_range_query::range_of_stmt): Adjust comment.
3880 * gimple-range-path.h (ssa_defined_in_bb): New.
3882 2021-11-15 Aldy Hernandez <aldyh@redhat.com>
3884 * gimple-range-path.cc (path_range_query::internal_range_of_expr):
3885 Default to global range if nothing found.
3887 2021-11-15 Richard Biener <rguenther@suse.de>
3889 PR tree-optimization/103237
3890 * tree-vect-loop.c (vect_is_simple_reduction): Fail for
3891 double reductions with multiple inner loop LC PHI nodes.
3893 2021-11-15 Hongyu Wang <hongyu.wang@intel.com>
3896 * config/i386/i386-expand.c (ix86_expand_atomic_fetch_op_loop):
3897 New expand function.
3898 * config/i386/i386-options.c (ix86_target_string): Add
3899 -mrelax-cmpxchg-loop flag.
3900 (ix86_valid_target_attribute_inner_p): Likewise.
3901 * config/i386/i386-protos.h (ix86_expand_atomic_fetch_op_loop):
3902 New expand function prototype.
3903 * config/i386/i386.opt: Add -mrelax-cmpxchg-loop.
3904 * config/i386/sync.md (atomic_fetch_<logic><mode>): New expander
3906 (atomic_<logic>_fetch<mode>): Likewise.
3907 (atomic_fetch_nand<mode>): Likewise.
3908 (atomic_nand_fetch<mode>): Likewise.
3909 (atomic_fetch_<logic><mode>): New expander for DI,TI modes.
3910 (atomic_<logic>_fetch<mode>): Likewise.
3911 (atomic_fetch_nand<mode>): Likewise.
3912 (atomic_nand_fetch<mode>): Likewise.
3913 * doc/invoke.texi: Document -mrelax-cmpxchg-loop.
3915 2021-11-15 Richard Biener <rguenther@suse.de>
3917 PR tree-optimization/103219
3918 * gimple-loop-jam.c (tree_loop_unroll_and_jam): Use single_exit
3919 to determine the exit for the VN region.
3921 2021-11-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
3923 * tree-ssa-loop.c (pass_vectorize): Move to tree-vectorizer.c.
3924 (pass_data_vectorize): Likewise.
3925 (make_pass_vectorize): Likewise.
3926 * tree-vectorizer.c (vectorize_loops): Merge with
3927 pass_vectorize::execute and replace cfun occurences with fun param.
3928 (adjust_simduid_builtins): Add fun param, replace cfun occurences with
3929 fun, and adjust callers approrpiately.
3930 (note_simd_array_uses): Likewise.
3931 (vect_loop_dist_alias_call): Likewise.
3932 (set_uid_loop_bbs): Likewise.
3933 (vect_transform_loops): Likewise.
3934 (try_vectorize_loop_1): Likewise.
3935 (try_vectorize_loop): Likewise.
3937 2021-11-15 Jakub Jelinek <jakub@redhat.com>
3940 * config/i386/sync.md (atomic_bit_test_and_set<mode>,
3941 atomic_bit_test_and_complement<mode>,
3942 atomic_bit_test_and_reset<mode>): Use OPTAB_WIDEN instead of
3945 2021-11-15 Maciej W. Rozycki <macro@embecosm.com>
3947 * config/vax/vax.c (vax_lra_p): New prototype and function.
3948 (TARGET_LRA_P): Wire it.
3949 * config/vax/vax.opt (mlra): New option.
3950 * doc/invoke.texi (Option Summary, VAX Options): Document the
3953 2021-11-14 Jan Hubicka <jh@suse.cz>
3955 * ipa-modref.h (struct modref_summary): Add nondeterministic
3956 and calls_interposable flags.
3957 * ipa-modref.c (modref_summary::modref_summary): Initialize new flags.
3958 (modref_summary::useful_p): Check new flags.
3959 (struct modref_summary_lto): Add nondeterministic and
3960 calls_interposable flags.
3961 (modref_summary_lto::modref_summary_lto): Initialize new flags.
3962 (modref_summary_lto::useful_p): Check new flags.
3963 (modref_summary::dump): Dump new flags.
3964 (modref_summary_lto::dump): Dump new flags.
3965 (ignore_nondeterminism_p): New function.
3966 (merge_call_side_effects): Merge new flags.
3967 (process_fnspec): Likewise.
3968 (analyze_load): Volatile access is nondeterministic.
3969 (analyze_store): Liekwise.
3970 (analyze_stmt): Volatile ASM is nondeterministic.
3971 (analyze_function): Clear new flags.
3972 (modref_summaries::duplicate): Duplicate new flags.
3973 (modref_summaries_lto::duplicate): Duplicate new flags.
3974 (modref_write): Stream new flags.
3975 (read_section): Stream new flags.
3976 (propagate_unknown_call): Update new flags.
3977 (modref_propagate_in_scc): Propagate new flags.
3978 * tree-ssa-alias.c (ref_maybe_used_by_call_p_1): Check
3980 * tree-ssa-structalias.c (determine_global_memory_access):
3983 2021-11-14 Maciej W. Rozycki <macro@embecosm.com>
3985 * config/vax/vax.h (SET_RATIO): New macro.
3986 * config/vax/vax.md (UNSPEC_SETMEM_FILL): New constant.
3987 (setmemhi): New expander.
3988 (setmemhi1): New insn and splitter.
3989 (*setmemhi1): New insn.
3991 2021-11-14 Jan Hubicka <hubicka@ucw.cz>
3993 * ipa-modref-tree.c (modref_access_node::update_for_kills): New
3995 (modref_access_node::merge_for_kills): Likewise.
3996 (modref_access_node::insert_kill): Likewise.
3997 * ipa-modref-tree.h (modref_access_node::update_for_kills,
3998 modref_access_node::merge_for_kills, modref_access_node::insert_kill):
4000 (modref_access_node::useful_for_kill): New member function.
4001 * ipa-modref.c (modref_summary::useful_p): Release useless kills.
4002 (lto_modref_summary): Add kills.
4003 (modref_summary::dump): Dump kills.
4004 (record_access): Add mdoref_access_node parameter.
4005 (record_access_lto): Likewise.
4006 (merge_call_side_effects): Merge kills.
4007 (analyze_call): Add ALWAYS_EXECUTED param and pass it around.
4008 (struct summary_ptrs): Add always_executed filed.
4009 (analyze_load): Update.
4010 (analyze_store): Update; record kills.
4011 (analyze_stmt): Add always_executed; record kills in clobbers.
4012 (analyze_function): Track always_executed.
4013 (modref_summaries::duplicate): Duplicate kills.
4014 (update_signature): Release kills.
4015 * ipa-modref.h (struct modref_summary): Add kills.
4016 * tree-ssa-alias.c (alias_stats): Add kill stats.
4017 (dump_alias_stats): Dump kill stats.
4018 (store_kills_ref_p): Break out from ...
4019 (stmt_kills_ref_p): Use it; handle modref info based kills.
4021 2021-11-14 Aldy Hernandez <aldyh@redhat.com>
4023 PR tree-optimization/103229
4024 * gimple-range-cache.cc (ssa_global_cache::clear): Do not pass
4025 null value to memset.
4027 2021-11-14 Jan Hubicka <hubicka@ucw.cz>
4029 * ipa-modref-tree.c (modref_access_node::get_call_arg): New member
4031 (modref_access_node::get_ao_ref): Likewise.
4032 * ipa-modref-tree.h (modref_access_node::get_call_arg): Declare.
4033 (modref_access_node::get_ao_ref): Declare.
4034 * tree-ssa-alias.c (modref_may_conflict): Use new accessors.
4035 * tree-ssa-dse.c (dse_optimize_call): Use new accessors.
4037 2021-11-13 Jan Hubicka <jh@suse.cz>
4040 * dbgcnt.def (ipa_attr): New counters.
4041 * ipa-pure-const.c: Include dbgcnt.c
4042 (ipa_make_function_const): Use debug counter.
4043 (ipa_make_function_pure): Likewise.
4044 (propagate_pure_const): Fix bug in my previous change.
4046 2021-11-13 Jan Hubicka <hubicka@ucw.cz>
4048 * ipa-modref-tree.c (modref_access_node::range_info_useful_p):
4049 Offline from ipa-modref-tree.h.
4050 (modref_access_node::dump): Move from ipa-modref.c; make member
4052 * ipa-modref-tree.h (modref_access_node::range_info_useful_p.
4053 modref_access_node::dump): Declare.
4054 * ipa-modref.c (dump_access): Remove.
4055 (dump_records): Update.
4056 (dump_lto_records): Update.
4057 (record_access): Update.
4058 (record_access_lto): Update.
4060 2021-11-13 Jan Hubicka <hubicka@ucw.cz>
4062 * ipa-modref.c (modref_summary::modref_summary): Clear new flags.
4063 (modref_summary::dump): Dump try_dse.
4064 (modref_summary::finalize): Add FUN attribute; compute try-dse.
4065 (analyze_function): Update.
4066 (read_section): Update.
4067 (update_signature): Update.
4068 (pass_ipa_modref::execute): Update.
4069 * ipa-modref.h (struct modref_summary):
4070 * tree-ssa-alias.c (ao_ref_init_from_ptr_and_range): Export.
4071 * tree-ssa-alias.h (ao_ref_init_from_ptr_and_range): Declare.
4072 * tree-ssa-dse.c (dse_optimize_call): New function.
4073 (dse_optimize_stmt): Use it.
4075 2021-11-13 Jan Hubicka <hubicka@ucw.cz>
4077 * ipa-modref-tree.c: Move #if CHECKING_P to proper place.
4079 2021-11-13 Jan Hubicka <jh@suse.cz>
4082 (struct modref_access_node): Move longer member functions to
4084 (modref_ref_node::try_merge_with): Turn into modreef_acces_node member
4086 * ipa-modref-tree.c (modref_access_node::contains): Move here
4087 from ipa-modref-tree.h.
4088 (modref_access_node::update): Likewise.
4089 (modref_access_node::merge): Likewise.
4090 (modref_access_node::closer_pair_p): Likewise.
4091 (modref_access_node::forced_merge): Likewise.
4092 (modref_access_node::update2): Likewise.
4093 (modref_access_node::combined_offsets): Likewise.
4094 (modref_access_node::try_merge_with): Likewise.
4095 (modref_access_node::insert): Likewise.
4097 2021-11-13 Jan Hubicka <jh@suse.cz>
4099 * ipa-modref.c (modref_summary::global_memory_read_p): Remove.
4100 (modref_summary::global_memory_written_p): Remove.
4101 (modref_summary::dump): Dump new flags.
4102 (modref_summary::finalize): New member function.
4103 (analyze_function): Call it.
4104 (read_section): Call it.
4105 (update_signature): Call it.
4106 (pass_ipa_modref::execute): Call it.
4107 * ipa-modref.h (struct modref_summary): Remove
4108 global_memory_read_p and global_memory_written_p.
4109 Add global_memory_read, global_memory_written.
4110 * tree-ssa-structalias.c (determine_global_memory_access):
4113 2021-11-13 Jan Hubicka <jh@suse.cz>
4115 * ipa-fnsummary.c (compute_fn_summary): Use type_attribut_allowed_p
4116 * ipa-param-manipulation.c
4117 (ipa_param_adjustments::type_attribute_allowed_p):
4118 New member function.
4119 (drop_type_attribute_if_params_changed_p): New function.
4120 (build_adjusted_function_type): Use it.
4121 * ipa-param-manipulation.h: Add type_attribute_allowed_p.
4123 2021-11-13 David Malcolm <dmalcolm@redhat.com>
4125 * doc/invoke.texi (Static Analyzer Options): Add
4126 -Wno-analyzer-tainted-allocation-size,
4127 -Wno-analyzer-tainted-divisor, -Wno-analyzer-tainted-offset, and
4128 -Wno-analyzer-tainted-size to list. Add
4129 -Wanalyzer-tainted-allocation-size, -Wanalyzer-tainted-divisor,
4130 -Wanalyzer-tainted-offset, and -Wanalyzer-tainted-size to list
4131 of options effectively enabled by -fanalyzer.
4132 (-Wanalyzer-tainted-allocation-size): New.
4133 (-Wanalyzer-tainted-array-index): Tweak wording; add link to CWE.
4134 (-Wanalyzer-tainted-divisor): New.
4135 (-Wanalyzer-tainted-offset): New.
4136 (-Wanalyzer-tainted-size): New.
4138 2021-11-13 Jan Hubicka <jh@suse.cz>
4140 * attr-fnspec.h (attr_fnspec::arg_eaf_flags): Break out from ...
4141 * gimple.c (gimple_call_arg_flags): ... here.
4142 * ipa-modref.c (analyze_parms): Record flags known from fnspec.
4143 (modref_merge_call_site_flags): Use arg_eaf_flags.
4145 2021-11-13 Aldy Hernandez <aldyh@redhat.com>
4147 PR tree-optimization/103222
4148 * gimple-range-path.cc (path_range_query::compute_ranges_in_phis):
4150 (path_range_query::compute_ranges_in_block): Call
4151 compute_ranges_in_phis.
4152 * gimple-range-path.h (path_range_query::compute_ranges_in_phis):
4155 2021-11-13 Jan Hubicka <jh@suse.cz>
4157 * ipa-fnsummary.c (compute_fn_summary): Do not give up on signature
4158 changes on "fn spec" attribute; give up on varadic types.
4159 * ipa-param-manipulation.c: Include attribs.h.
4160 (build_adjusted_function_type): New parameter ARG_MODIFIED; if it is
4161 true remove "fn spec" attribute.
4162 (ipa_param_adjustments::build_new_function_type): Update.
4163 (ipa_param_body_adjustments::modify_formal_parameters): update.
4164 * ipa-sra.c: Include attribs.h.
4165 (ipa_sra_preliminary_function_checks): Do not check for TYPE_ATTRIBUTES.
4167 2021-11-13 Aldy Hernandez <aldyh@redhat.com>
4169 * gimple-range-path.cc (path_range_query::path_range_query): Merge
4171 (path_range_query::import_p): Move from header file.
4172 (path_range_query::~path_range_query): Adjust for combined ctors.
4173 * gimple-range-path.h: Merge ctors.
4174 (path_range_query::import_p): Move to .cc file.
4176 2021-11-13 Jan Hubicka <jh@suse.cz>
4178 PR tree-optimization/103182
4179 * ipa-modref.c (callee_to_caller_flags): Fix merging of flags.
4180 (modref_eaf_analysis::analyze_ssa_name): Fix merging of flags.
4182 2021-11-12 Stafford Horne <shorne@gmail.com>
4184 * config/or1k/or1k-protos.h (or1k_profile_hook): New function.
4185 * config/or1k/or1k.h (PROFILE_HOOK): Change macro to reference
4186 new function or1k_profile_hook.
4187 * config/or1k/or1k.c (struct machine_function): Add new field
4188 set_mcount_arg_insn.
4189 (or1k_profile_hook): New function.
4190 (or1k_init_pic_reg): Update to inject pic rtx after _mcount arg
4192 (or1k_frame_pointer_required): Frame pointer no longer needed
4195 2021-11-12 Jan Hubicka <jh@suse.cz>
4197 PR tree-optimization/103209
4198 * tree-ssa-structalias.c (find_func_aliases_for_call): Fix
4199 use of handle_rhs_call
4201 2021-11-12 Aldy Hernandez <aldyh@redhat.com>
4203 PR tree-optimization/103202
4204 * gimple-range-path.cc
4205 (path_range_query::compute_ranges_in_block): Solve PHI imports first.
4207 2021-11-12 Jan Hubicka <jh@suse.cz>
4209 * ipa-pure-const.c (propagate_pure_const): Remove redundant check;
4210 fix call of ipa_make_function_const and ipa_make_function_pure.
4212 2021-11-12 Richard Sandiford <richard.sandiford@arm.com>
4214 * config/aarch64/aarch64.c (aarch64_vector_op::n_advsimd_ops): Delete.
4215 (aarch64_vector_op::m_seen_loads): Likewise.
4216 (aarch64_vector_costs::aarch64_vector_costs): Don't push to
4218 (aarch64_vector_op::count_ops): Remove vectype and factor parameters.
4219 Remove code that tries to predict different vec_flags from the
4221 (aarch64_vector_costs::add_stmt_cost): Update accordingly.
4222 Remove m_advsimd_ops handling.
4224 2021-11-12 Richard Sandiford <richard.sandiford@arm.com>
4226 * config/aarch64/aarch64.c (aarch64_vector_costs::m_saw_sve_only_op)
4227 (aarch64_sve_only_stmt_p): Delete.
4228 (aarch64_vector_costs::prefer_unrolled_loop): New function,
4229 extracted from adjust_body_cost.
4230 (aarch64_vector_costs::better_main_loop_than_p): New function,
4231 using heuristics extracted from adjust_body_cost and
4232 adjust_body_cost_sve.
4233 (aarch64_vector_costs::adjust_body_cost_sve): Remove
4234 advsimd_cycles_per_iter and could_use_advsimd parameters.
4235 Update after changes above.
4236 (aarch64_vector_costs::adjust_body_cost): Update after changes above.
4238 2021-11-12 Richard Sandiford <richard.sandiford@arm.com>
4240 * config/aarch64/aarch64.c (aarch64_vec_op_count::m_vf_factor):
4241 New member variable.
4242 (aarch64_vec_op_count::aarch64_vec_op_count): Add a parameter for it.
4243 (aarch64_vec_op_count::vf_factor): New function.
4244 (aarch64_vector_costs::aarch64_vector_costs): When costing for
4245 neoverse-512tvb, pass a vf_factor of 2 for the Neoverse V1 version
4247 (aarch64_vector_costs::adjust_body_cost): Read the vf factor
4248 instead of hard-coding 2.
4250 2021-11-12 Richard Sandiford <richard.sandiford@arm.com>
4252 * config/aarch64/aarch64.c
4253 (aarch64_vec_op_count::rename_cycles_per_iter): New function.
4254 (aarch64_vec_op_count::min_nonpred_cycles_per_iter): Likewise.
4255 (aarch64_vec_op_count::min_pred_cycles_per_iter): Likewise.
4256 (aarch64_vec_op_count::min_cycles_per_iter): Likewise.
4257 (aarch64_vec_op_count::dump): Move earlier in file. Dump the
4258 above properties too.
4259 (aarch64_estimate_min_cycles_per_iter): Delete.
4260 (adjust_body_cost): Use aarch64_vec_op_count::min_cycles_per_iter
4261 instead of aarch64_estimate_min_cycles_per_iter. Rely on the dump
4262 routine to print CPI estimates.
4263 (adjust_body_cost_sve): Likewise. Use the other functions above
4264 instead of doing the work inline.
4266 2021-11-12 Richard Sandiford <richard.sandiford@arm.com>
4268 * config/aarch64/aarch64.c (aarch64_vec_op_count): Allow default
4270 (aarch64_vec_op_count::base_issue_info): Remove handling of null
4272 (aarch64_vec_op_count::simd_issue_info): Likewise.
4273 (aarch64_vec_op_count::sve_issue_info): Likewise.
4274 (aarch64_vector_costs::m_ops): Turn into a vector.
4275 (aarch64_vector_costs::m_advsimd_ops): Likewise.
4276 (aarch64_vector_costs::aarch64_vector_costs): Add entries to
4277 the vectors based on aarch64_tune_params.
4278 (aarch64_vector_costs::analyze_loop_vinfo): Update the pred_ops
4279 of all entries in m_ops.
4280 (aarch64_vector_costs::add_stmt_cost): Call count_ops for all
4282 (aarch64_estimate_min_cycles_per_iter): Remove issue_info
4283 parameter and get the information from the ops instead.
4284 (aarch64_vector_costs::adjust_body_cost_sve): Take a
4285 aarch64_vec_issue_info instead of a aarch64_vec_op_count.
4286 (aarch64_vector_costs::adjust_body_cost): Update call accordingly.
4287 Exit earlier if m_ops is empty for either cost structure.
4289 2021-11-12 Richard Sandiford <richard.sandiford@arm.com>
4291 * config/aarch64/aarch64.c (aarch64_vector_costs::m_scalar_ops)
4292 (aarch64_vector_costs::m_sve_ops): Replace with...
4293 (aarch64_vector_costs::m_ops): ...this.
4294 (aarch64_vector_costs::analyze_loop_vinfo): Update accordingly.
4295 (aarch64_vector_costs::adjust_body_cost_sve): Likewise.
4296 (aarch64_vector_costs::aarch64_vector_costs): Likewise.
4297 Initialize m_vec_flags here rather than in add_stmt_cost.
4298 (aarch64_vector_costs::count_ops): Test for scalar reductions too.
4299 Allow vectype to be null.
4300 (aarch64_vector_costs::add_stmt_cost): Call count_ops for scalar
4301 code too. Don't require vectype to be nonnull.
4302 (aarch64_vector_costs::adjust_body_cost): Take the loop_vec_info
4303 and scalar costs as parameters. Use the scalar costs to determine
4304 the cycles per iteration of the scalar loop, then multiply it
4305 by the estimated VF.
4306 (aarch64_vector_costs::finish_cost): Update call accordingly.
4308 2021-11-12 Richard Sandiford <richard.sandiford@arm.com>
4310 * config/aarch64/aarch64.c (aarch64_dr_type): New function.
4311 (aarch64_vector_costs::count_ops): Use it rather than the
4312 vectype to determine floatness.
4314 2021-11-12 Richard Sandiford <richard.sandiford@arm.com>
4316 * config/aarch64/aarch64.c (aarch64_sve_in_loop_reduction_latency):
4317 Remove vectype parameter and get floatness from the type of the
4319 (arch64_in_loop_reduction_latency): Likewise.
4320 (aarch64_detect_vector_stmt_subtype): Update caller.
4321 (aarch64_vector_costs::count_ops): Likewise.
4323 2021-11-12 Richard Sandiford <richard.sandiford@arm.com>
4325 * config/aarch64/aarch64.c (aarch64_sve_op_count): Fold into...
4326 (aarch64_vec_op_count): ...this. Add a constructor.
4327 (aarch64_vec_op_count::vec_flags): New function.
4328 (aarch64_vec_op_count::base_issue_info): Likewise.
4329 (aarch64_vec_op_count::simd_issue_info): Likewise.
4330 (aarch64_vec_op_count::sve_issue_info): Likewise.
4331 (aarch64_vec_op_count::m_issue_info): New member variable.
4332 (aarch64_vec_op_count::m_vec_flags): Likewise.
4333 (aarch64_vector_costs): Add a constructor.
4334 (aarch64_vector_costs::m_sve_ops): Change type to aarch64_vec_op_count.
4335 (aarch64_vector_costs::aarch64_vector_costs): New function.
4336 Initialize m_scalar_ops, m_advsimd_ops and m_sve_ops.
4337 (aarch64_vector_costs::count_ops): Remove vec_flags and
4338 issue_info parameters, using the new aarch64_vec_op_count
4340 (aarch64_vector_costs::add_stmt_cost): Update call accordingly.
4341 (aarch64_sve_op_count::dump): Fold into...
4342 (aarch64_vec_op_count::dump): ..here.
4344 2021-11-12 Richard Sandiford <richard.sandiford@arm.com>
4346 * config/aarch64/aarch64.c: Include tree-dfa.h.
4347 (aarch64_check_consecutive_mems): New function that takes MEM_EXPR
4348 and MEM_OFFSET into account.
4349 (aarch64_swap_ldrstr_operands): Use it.
4350 (aarch64_operands_ok_for_ldpstp): Likewise. Check that the
4351 address of the second memory doesn't depend on the result of
4354 2021-11-12 Jan Hubicka <hubicka@ucw.cz>
4356 * ipa-pure-const.c (ipa_make_function_pure): Fix exit condition.
4358 2021-11-12 Jan Hubicka <jh@suse.cz>
4360 PR tree-optimization/103175
4361 * ipa-modref.c (modref_lattice::merge): Add sanity check.
4362 (callee_to_caller_flags): Make flags adjustment sane.
4363 (modref_eaf_analysis::analyze_ssa_name): Likewise.
4365 2021-11-12 Jan Hubicka <jh@suse.cz>
4368 * ipa-modref.c (analyze_function, modref_propagate_in_scc): Do
4369 not mark pure/const function if there are side-effects.
4371 2021-11-12 Chung-Lin Tang <cltang@codesourcery.com>
4373 * tree.h (OMP_CLAUSE_MAP_RUNTIME_IMPLICIT_P): New access macro for
4374 'implicit' bit, using 'base.deprecated_flag' field of tree_node.
4375 * tree-pretty-print.c (dump_omp_clause): Add support for printing
4376 implicit attribute in tree dumping.
4377 * gimplify.c (gimplify_adjust_omp_clauses_1):
4378 Set OMP_CLAUSE_MAP_RUNTIME_IMPLICIT_P to 1 if map clause is implicitly
4380 (gimplify_adjust_omp_clauses): Adjust place of adding implicitly created
4381 clauses, from simple append, to starting of list, after non-map clauses.
4382 * omp-low.c (lower_omp_target): Add GOMP_MAP_IMPLICIT bits into kind
4383 values passed to libgomp for implicit maps.
4385 2021-11-12 Jakub Jelinek <jakub@redhat.com>
4387 * omp-builtins.def (BUILT_IN_GOMP_TEAMS): Remove.
4388 (BUILT_IN_GOMP_TEAMS4): New.
4389 * builtin-types.def (BT_FN_VOID_UINT_UINT): Remove.
4390 (BT_FN_BOOL_UINT_UINT_UINT_BOOL): New.
4391 * omp-low.c (lower_omp_teams): Use GOMP_teams4 instead of
4392 GOMP_teams, pass to it also num_teams lower-bound expression
4393 or a dup of upper-bound if it is missing and a flag whether
4394 it is the first call or not.
4396 2021-11-12 Martin Liska <mliska@suse.cz>
4398 PR tree-optimization/102497
4399 * gimple-predicate-analysis.cc (add_pred): Remove unused
4402 2021-11-12 Richard Biener <rguenther@suse.de>
4404 PR tree-optimization/103204
4405 * tree-ssa-sccvn.c (valueize_refs_1): Re-valueize the
4406 top operand after folding in an address.
4408 2021-11-11 Aldy Hernandez <aldyh@redhat.com>
4410 * gimple-range-path.cc (path_range_query::path_range_query): New
4411 ctor without a ranger.
4412 (path_range_query::~path_range_query): Free ranger if necessary.
4413 (path_range_query::range_on_path_entry): Adjust m_ranger for pointer.
4414 (path_range_query::ssa_range_in_phi): Same.
4415 (path_range_query::compute_ranges_in_block): Same.
4416 (path_range_query::compute_imports): Same.
4417 (path_range_query::compute_ranges): Same.
4418 (path_range_query::range_of_stmt): Same.
4419 (path_range_query::compute_outgoing_relations): Same.
4420 * gimple-range-path.h (class path_range_query): New ctor.
4421 * tree-ssa-loop-ch.c (ch_base::copy_headers): Remove gimple_ranger
4422 as path_range_query allocates one.
4423 * tree-ssa-threadbackward.c (class back_threader): Remove m_ranger.
4424 (back_threader::~back_threader): Same.
4426 2021-11-11 Aldy Hernandez <aldyh@redhat.com>
4428 * tree-ssa-threadbackward.c
4429 (back_threader_profitability::profitable_path_p): Remove loop
4430 crossing restriction.
4432 2021-11-11 Jan Hubicka <hubicka@ucw.cz>
4434 * ipa-modref.c (analyze_function): Do pure/const discovery, return
4436 (pass_modref::execute): If pure/const is discovered fixup cfg.
4437 (ignore_edge): Do not ignore pure/const edges.
4438 (modref_propagate_in_scc): Do pure/const discovery, return true if
4439 cdtor was promoted pure/const.
4440 (pass_ipa_modref::execute): If needed remove unreachable functions.
4441 * ipa-pure-const.c (warn_function_noreturn): Fix whitespace.
4442 (warn_function_cold): Likewise.
4443 (skip_function_for_local_pure_const): Move earlier.
4444 (ipa_make_function_const): Break out from ...
4445 (ipa_make_function_pure): Break out from ...
4446 (propagate_pure_const): ... here.
4447 (pass_local_pure_const::execute): Use it.
4448 * ipa-utils.h (ipa_make_function_const): Declare.
4449 (ipa_make_function_pure): Declare.
4450 * passes.def: Move early modref after pure-const.
4452 2021-11-11 David Malcolm <dmalcolm@redhat.com>
4455 * diagnostic-show-locus.c (def_policy): Use def_tabstop.
4457 2021-11-11 Jonathan Wright <jonathan.wright@arm.com>
4459 * config/aarch64/aarch64-builtins.c (TYPES_COMBINE): Delete.
4460 (TYPES_COMBINEP): Delete.
4461 * config/aarch64/aarch64-simd-builtins.def: Declare type-
4462 qualified builtins for vcombine_* intrinsics.
4463 * config/aarch64/arm_neon.h (vcombine_s8): Remove unnecessary
4465 (vcombine_s16): Likewise.
4466 (vcombine_s32): Likewise.
4467 (vcombine_f32): Likewise.
4468 (vcombine_u8): Use type-qualified builtin and remove casts.
4469 (vcombine_u16): Likewise.
4470 (vcombine_u32): Likewise.
4471 (vcombine_u64): Likewise.
4472 (vcombine_p8): Likewise.
4473 (vcombine_p16): Likewise.
4474 (vcombine_p64): Likewise.
4475 (vcombine_bf16): Remove unnecessary cast.
4476 * config/aarch64/iterators.md (VD_I): New mode iterator.
4477 (VDC_P): New mode iterator.
4479 2021-11-11 Jonathan Wright <jonathan.wright@arm.com>
4481 * config/aarch64/aarch64-builtins.c (TYPES_LOAD1_U): Define.
4482 (TYPES_LOAD1_P): Define.
4483 (TYPES_STORE1_U): Define.
4484 (TYPES_STORE1P): Rename to...
4485 (TYPES_STORE1_P): This.
4486 (get_mem_type_for_load_store): Add unsigned and poly types.
4487 (aarch64_general_gimple_fold_builtin): Add unsigned and poly
4488 type-qualified builtin declarations.
4489 * config/aarch64/aarch64-simd-builtins.def: Declare type-
4490 qualified builtins for LD1/ST1.
4491 * config/aarch64/arm_neon.h (vld1_p8): Use type-qualified
4492 builtin and remove cast.
4493 (vld1_p16): Likewise.
4494 (vld1_u8): Likewise.
4495 (vld1_u16): Likewise.
4496 (vld1_u32): Likewise.
4497 (vld1q_p8): Likewise.
4498 (vld1q_p16): Likewise.
4499 (vld1q_p64): Likewise.
4500 (vld1q_u8): Likewise.
4501 (vld1q_u16): Likewise.
4502 (vld1q_u32): Likewise.
4503 (vld1q_u64): Likewise.
4504 (vst1_p8): Likewise.
4505 (vst1_p16): Likewise.
4506 (vst1_u8): Likewise.
4507 (vst1_u16): Likewise.
4508 (vst1_u32): Likewise.
4509 (vst1q_p8): Likewise.
4510 (vst1q_p16): Likewise.
4511 (vst1q_p64): Likewise.
4512 (vst1q_u8): Likewise.
4513 (vst1q_u16): Likewise.
4514 (vst1q_u32): Likewise.
4515 (vst1q_u64): Likewise.
4516 * config/aarch64/iterators.md (VALLP_NO_DI): New iterator.
4518 2021-11-11 Jonathan Wright <jonathan.wright@arm.com>
4520 * config/aarch64/aarch64-simd-builtins.def: Declare unsigned
4521 builtins for vector reduction.
4522 * config/aarch64/arm_neon.h (vaddv_u8): Use type-qualified
4523 builtin and remove casts.
4524 (vaddv_u16): Likewise.
4525 (vaddv_u32): Likewise.
4526 (vaddvq_u8): Likewise.
4527 (vaddvq_u16): Likewise.
4528 (vaddvq_u32): Likewise.
4529 (vaddvq_u64): Likewise.
4531 2021-11-11 Jonathan Wright <jonathan.wright@arm.com>
4533 * config/aarch64/aarch64-simd-builtins.def:
4534 * config/aarch64/arm_neon.h (vpaddq_u8): Use type-qualified
4535 builtin and remove casts.
4536 (vpaddq_u16): Likewise.
4537 (vpaddq_u32): Likewise.
4538 (vpaddq_u64): Likewise.
4539 (vpadd_u8): Likewise.
4540 (vpadd_u16): Likewise.
4541 (vpadd_u32): Likewise.
4542 (vpaddd_u64): Likewise.
4544 2021-11-11 Jonathan Wright <jonathan.wright@arm.com>
4546 * config/aarch64/aarch64-simd-builtins.def: Declare unsigned
4547 builtins for [r]subhn[2].
4548 * config/aarch64/arm_neon.h (vsubhn_s16): Remove unnecessary
4550 (vsubhn_s32): Likewise.
4551 (vsubhn_s64): Likewise.
4552 (vsubhn_u16): Use type-qualified builtin and remove casts.
4553 (vsubhn_u32): Likewise.
4554 (vsubhn_u64): Likewise.
4555 (vrsubhn_s16): Remove unnecessary cast.
4556 (vrsubhn_s32): Likewise.
4557 (vrsubhn_s64): Likewise.
4558 (vrsubhn_u16): Use type-qualified builtin and remove casts.
4559 (vrsubhn_u32): Likewise.
4560 (vrsubhn_u64): Likewise.
4561 (vrsubhn_high_s16): Remove unnecessary cast.
4562 (vrsubhn_high_s32): Likewise.
4563 (vrsubhn_high_s64): Likewise.
4564 (vrsubhn_high_u16): Use type-qualified builtin and remove
4566 (vrsubhn_high_u32): Likewise.
4567 (vrsubhn_high_u64): Likewise.
4568 (vsubhn_high_s16): Remove unnecessary cast.
4569 (vsubhn_high_s32): Likewise.
4570 (vsubhn_high_s64): Likewise.
4571 (vsubhn_high_u16): Use type-qualified builtin and remove
4573 (vsubhn_high_u32): Likewise.
4574 (vsubhn_high_u64): Likewise.
4576 2021-11-11 Jonathan Wright <jonathan.wright@arm.com>
4578 * config/aarch64/aarch64-simd-builtins.def: Declare unsigned
4579 builtins for [r]addhn[2].
4580 * config/aarch64/arm_neon.h (vaddhn_s16): Remove unnecessary
4582 (vaddhn_s32): Likewise.
4583 (vaddhn_s64): Likewise.
4584 (vaddhn_u16): Use type-qualified builtin and remove casts.
4585 (vaddhn_u32): Likewise.
4586 (vaddhn_u64): Likewise.
4587 (vraddhn_s16): Remove unnecessary cast.
4588 (vraddhn_s32): Likewise.
4589 (vraddhn_s64): Likewise.
4590 (vraddhn_u16): Use type-qualified builtin and remove casts.
4591 (vraddhn_u32): Likewise.
4592 (vraddhn_u64): Likewise.
4593 (vaddhn_high_s16): Remove unnecessary cast.
4594 (vaddhn_high_s32): Likewise.
4595 (vaddhn_high_s64): Likewise.
4596 (vaddhn_high_u16): Use type-qualified builtin and remove
4598 (vaddhn_high_u32): Likewise.
4599 (vaddhn_high_u64): Likewise.
4600 (vraddhn_high_s16): Remove unnecessary cast.
4601 (vraddhn_high_s32): Likewise.
4602 (vraddhn_high_s64): Likewise.
4603 (vraddhn_high_u16): Use type-qualified builtin and remove
4605 (vraddhn_high_u32): Likewise.
4606 (vraddhn_high_u64): Likewise.
4608 2021-11-11 Jonathan Wright <jonathan.wright@arm.com>
4610 * config/aarch64/aarch64-simd-builtins.def: Use BINOPU type
4611 qualifiers in generator macros for uhsub builtins.
4612 * config/aarch64/arm_neon.h (vhsub_s8): Remove unnecessary
4614 (vhsub_s16): Likewise.
4615 (vhsub_s32): Likewise.
4616 (vhsub_u8): Use type-qualified builtin and remove casts.
4617 (vhsub_u16): Likewise.
4618 (vhsub_u32): Likewise.
4619 (vhsubq_s8): Remove unnecessary cast.
4620 (vhsubq_s16): Likewise.
4621 (vhsubq_s32): Likewise.
4622 (vhsubq_u8): Use type-qualified builtin and remove casts.
4623 (vhsubq_u16): Likewise.
4624 (vhsubq_u32): Likewise.
4626 2021-11-11 Jonathan Wright <jonathan.wright@arm.com>
4628 * config/aarch64/aarch64-simd-builtins.def: Use BINOPU type
4629 qualifiers in generator macros for u[r]hadd builtins.
4630 * config/aarch64/arm_neon.h (vhadd_s8): Remove unnecessary
4632 (vhadd_s16): Likewise.
4633 (vhadd_s32): Likewise.
4634 (vhadd_u8): Use type-qualified builtin and remove casts.
4635 (vhadd_u16): Likewise.
4636 (vhadd_u32): Likewise.
4637 (vhaddq_s8): Remove unnecessary cast.
4638 (vhaddq_s16): Likewise.
4639 (vhaddq_s32): Likewise.
4640 (vhaddq_u8): Use type-qualified builtin and remove casts.
4641 (vhaddq_u16): Likewise.
4642 (vhaddq_u32): Likewise.
4643 (vrhadd_s8): Remove unnecessary cast.
4644 (vrhadd_s16): Likewise.
4645 (vrhadd_s32): Likewise.
4646 (vrhadd_u8): Use type-qualified builtin and remove casts.
4647 (vrhadd_u16): Likewise.
4648 (vrhadd_u32): Likewise.
4649 (vrhaddq_s8): Remove unnecessary cast.
4650 (vrhaddq_s16): Likewise.
4651 (vrhaddq_s32): Likewise.
4652 (vrhaddq_u8): Use type-wualified builtin and remove casts.
4653 (vrhaddq_u16): Likewise.
4654 (vrhaddq_u32): Likewise.
4656 2021-11-11 Jonathan Wright <jonathan.wright@arm.com>
4658 * config/aarch64/aarch64-simd-builtins.def: Use BINOPU type
4659 qualifiers in generator macros for usub[lw][2] builtins.
4660 * config/aarch64/arm_neon.h (vsubl_s8): Remove unnecessary
4662 (vsubl_s16): Likewise.
4663 (vsubl_s32): Likewise.
4664 (vsubl_u8): Use type-qualified builtin and remove casts.
4665 (vsubl_u16): Likewise.
4666 (vsubl_u32): Likewise.
4667 (vsubl_high_s8): Remove unnecessary cast.
4668 (vsubl_high_s16): Likewise.
4669 (vsubl_high_s32): Likewise.
4670 (vsubl_high_u8): Use type-qualified builtin and remove casts.
4671 (vsubl_high_u16): Likewise.
4672 (vsubl_high_u32): Likewise.
4673 (vsubw_s8): Remove unnecessary casts.
4674 (vsubw_s16): Likewise.
4675 (vsubw_s32): Likewise.
4676 (vsubw_u8): Use type-qualified builtin and remove casts.
4677 (vsubw_u16): Likewise.
4678 (vsubw_u32): Likewise.
4679 (vsubw_high_s8): Remove unnecessary cast.
4680 (vsubw_high_s16): Likewise.
4681 (vsubw_high_s32): Likewise.
4682 (vsubw_high_u8): Use type-qualified builtin and remove casts.
4683 (vsubw_high_u16): Likewise.
4684 (vsubw_high_u32): Likewise.
4686 2021-11-11 Jonathan Wright <jonathan.wright@arm.com>
4688 * config/aarch64/aarch64-simd-builtins.def: Use BINOPU type
4689 qualifiers in generator macros for uadd[lw][2] builtins.
4690 * config/aarch64/arm_neon.h (vaddl_s8): Remove unnecessary
4692 (vaddl_s16): Likewise.
4693 (vaddl_s32): Likewise.
4694 (vaddl_u8): Use type-qualified builtin and remove casts.
4695 (vaddl_u16): Likewise.
4696 (vaddl_u32): Likewise.
4697 (vaddl_high_s8): Remove unnecessary cast.
4698 (vaddl_high_s16): Likewise.
4699 (vaddl_high_s32): Likewise.
4700 (vaddl_high_u8): Use type-qualified builtin and remove casts.
4701 (vaddl_high_u16): Likewise.
4702 (vaddl_high_u32): Likewise.
4703 (vaddw_s8): Remove unnecessary cast.
4704 (vaddw_s16): Likewise.
4705 (vaddw_s32): Likewise.
4706 (vaddw_u8): Use type-qualified builtin and remove casts.
4707 (vaddw_u16): Likewise.
4708 (vaddw_u32): Likewise.
4709 (vaddw_high_s8): Remove unnecessary cast.
4710 (vaddw_high_s16): Likewise.
4711 (vaddw_high_s32): Likewise.
4712 (vaddw_high_u8): Use type-qualified builtin and remove casts.
4713 (vaddw_high_u16): Likewise.
4714 (vaddw_high_u32): Likewise.
4716 2021-11-11 Jonathan Wright <jonathan.wright@arm.com>
4718 * config/aarch64/aarch64-simd-builtins.def: Declare type-
4719 qualified builtins for [R]SHRN[2].
4720 * config/aarch64/arm_neon.h (vshrn_n_u16): Use type-qualified
4721 builtin and remove casts.
4722 (vshrn_n_u32): Likewise.
4723 (vshrn_n_u64): Likewise.
4724 (vrshrn_high_n_u16): Likewise.
4725 (vrshrn_high_n_u32): Likewise.
4726 (vrshrn_high_n_u64): Likewise.
4727 (vrshrn_n_u16): Likewise.
4728 (vrshrn_n_u32): Likewise.
4729 (vrshrn_n_u64): Likewise.
4730 (vshrn_high_n_u16): Likewise.
4731 (vshrn_high_n_u32): Likewise.
4732 (vshrn_high_n_u64): Likewise.
4734 2021-11-11 Jonathan Wright <jonathan.wright@arm.com>
4736 * config/aarch64/aarch64-simd-builtins.def: Declare unsigned
4737 type-qualified builtins for XTN[2].
4738 * config/aarch64/arm_neon.h (vmovn_high_u16): Use type-
4739 qualified builtin and remove casts.
4740 (vmovn_high_u32): Likewise.
4741 (vmovn_high_u64): Likewise.
4742 (vmovn_u16): Likewise.
4743 (vmovn_u32): Likewise.
4744 (vmovn_u64): Likewise.
4746 2021-11-11 Jonathan Wright <jonathan.wright@arm.com>
4748 * config/aarch64/aarch64-simd-builtins.def: Use poly type
4749 qualifier in builtin generator macros.
4750 * config/aarch64/arm_neon.h (vmul_p8): Use type-qualified
4751 builtin and remove casts.
4752 (vmulq_p8): Likewise.
4753 (vmull_high_p8): Likewise.
4754 (vmull_p8): Likewise.
4756 2021-11-11 Jonathan Wright <jonathan.wright@arm.com>
4758 * config/aarch64/aarch64-simd-builtins.def: Declare type-
4759 qualified builtin generators for unsigned MLA/MLS intrinsics.
4760 * config/aarch64/arm_neon.h (vmla_n_u16): Use type-qualified
4762 (vmla_n_u32): Likewise.
4763 (vmla_u8): Likewise.
4764 (vmla_u16): Likewise.
4765 (vmla_u32): Likewise.
4766 (vmlaq_n_u16): Likewise.
4767 (vmlaq_n_u32): Likewise.
4768 (vmlaq_u8): Likewise.
4769 (vmlaq_u16): Likewise.
4770 (vmlaq_u32): Likewise.
4771 (vmls_n_u16): Likewise.
4772 (vmls_n_u32): Likewise.
4773 (vmls_u8): Likewise.
4774 (vmls_u16): Likewise.
4775 (vmls_u32): Likewise.
4776 (vmlsq_n_u16): Likewise.
4777 (vmlsq_n_u32): Likewise.
4778 (vmlsq_u8): Likewise.
4779 (vmlsq_u16): Likewise.
4780 (vmlsq_u32): Likewise.
4782 2021-11-11 Jan Hubicka <hubicka@ucw.cz>
4784 * ipa-modref.c (modref_summary::useful_p): Check also for side-effects
4785 with looping const/pure.
4786 (modref_summary_lto::useful_p): Likewise.
4787 (merge_call_side_effects): Merge side effects before early exit
4789 (process_fnspec): Also handle pure functions.
4790 (analyze_call): Do not early exit on looping pure const.
4791 (propagate_unknown_call): Also handle nontrivial SCC as side-effect.
4792 (modref_propagate_in_scc): Update.
4794 2021-11-11 Richard Biener <rguenther@suse.de>
4796 PR tree-optimization/103190
4797 * tree-ssa-reassoc.c (insert_stmt_after): Only assert on asm goto.
4799 2021-11-11 Aldy Hernandez <aldyh@redhat.com>
4801 * gimple-range-path.cc (path_range_query::add_copies_to_imports):
4803 (path_range_query::compute_imports): ...this. Adapt it so it can
4804 be passed the imports bitmap instead of working on m_imports.
4805 (path_range_query::compute_ranges): Call compute_imports in all
4806 cases unless an imports bitmap is passed.
4807 * gimple-range-path.h (path_range_query::compute_imports): New.
4808 (path_range_query::add_copies_to_imports): Remove.
4809 * tree-ssa-threadbackward.c (back_threader::resolve_def): Remove.
4810 (back_threader::find_paths_to_names): Inline resolve_def.
4811 (back_threader::find_paths): Call compute_imports.
4812 (back_threader::resolve_phi): Adjust comment.
4814 2021-11-11 Richard Biener <rguenther@suse.de>
4816 PR tree-optimization/103188
4817 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p):
4818 Remove query parameter, split out check for size
4820 (ch_base::m_ranger, cb_base::m_query): Remove.
4821 (ch_base::copy_headers): Split processing loop into
4822 analysis around which we allocate and use ranger and
4823 transform where we do not.
4824 (pass_ch::execute): Do not allocate/free ranger here.
4825 (pass_ch_vect::execute): Likewise.
4827 2021-11-11 Jan Hubicka <hubicka@ucw.cz>
4829 * ipa-pure-const.c (propagate_pure_const): Self recursion is
4832 2021-11-11 Jan Hubicka <hubicka@ucw.cz>
4834 * cgraph.c (set_noreturn_flag_1): New function.
4835 (cgraph_node::set_noreturn_flag): New member function
4836 * cgraph.h (cgraph_node::set_noreturn_flags): Declare.
4837 * ipa-pure-const.c (pass_local_pure_const::execute): Use it.
4839 2021-11-11 Aldy Hernandez <aldyh@redhat.com>
4840 Richard Biener <rguenther@suse.de>
4842 * tree-ssa-loop-ch.c (entry_loop_condition_is_static): Resolve
4843 statically to the edge remaining in the loop.
4845 2021-11-11 Richard Biener <rguenther@suse.de>
4847 PR middle-end/103181
4848 * tree-eh.c (operation_could_trap_helper_p): Properly
4849 check vector constants for a zero element for integer
4850 division. Separate floating point and integer division code.
4852 2021-11-11 Jakub Jelinek <jakub@redhat.com>
4855 * dwarf2out.c (field_byte_offset): Do the PCC_BITFIELD_TYPE_MATTERS
4856 handling only for DECL_BIT_FIELD_TYPE decls.
4858 2021-11-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
4861 * config/aarch64/aarch64.c (aarch64_process_target_attr): Check if
4862 token is arch extension without leading '+' and emit appropriate
4863 diagnostic for the same.
4865 2021-11-11 Jakub Jelinek <jakub@redhat.com>
4867 * tree.h (OMP_CLAUSE_NUM_TEAMS_EXPR): Rename to ...
4868 (OMP_CLAUSE_NUM_TEAMS_UPPER_EXPR): ... this.
4869 (OMP_CLAUSE_NUM_TEAMS_LOWER_EXPR): Define.
4870 * tree.c (omp_clause_num_ops): Increase num ops for
4871 OMP_CLAUSE_NUM_TEAMS to 2.
4872 * tree-pretty-print.c (dump_omp_clause): Print optional lower bound
4873 for OMP_CLAUSE_NUM_TEAMS.
4874 * gimplify.c (gimplify_scan_omp_clauses): Gimplify
4875 OMP_CLAUSE_NUM_TEAMS_LOWER_EXPR if non-NULL.
4876 (optimize_target_teams): Use OMP_CLAUSE_NUM_TEAMS_UPPER_EXPR instead
4877 of OMP_CLAUSE_NUM_TEAMS_EXPR. Handle OMP_CLAUSE_NUM_TEAMS_LOWER_EXPR.
4878 * omp-low.c (lower_omp_teams): Use OMP_CLAUSE_NUM_TEAMS_UPPER_EXPR
4879 instead of OMP_CLAUSE_NUM_TEAMS_EXPR.
4880 * omp-expand.c (expand_teams_call, get_target_arguments): Likewise.
4882 2021-11-11 Richard Biener <rguenther@suse.de>
4884 * cfganal.c (find_pdom): Remove.
4885 (control_dependences::find_control_dependence): Remove
4886 special-casing of entry block, call get_immediate_dominator
4888 * gimple-predicate-analysis.cc (find_pdom): Remove.
4889 (find_dom): Likewise.
4890 (find_control_equiv_block): Call get_immediate_dominator
4892 (compute_control_dep_chain): Likewise.
4893 (predicate::init_from_phi_def): Likewise.
4895 2021-11-11 Richard Biener <rguenther@suse.de>
4897 * cfganal.h (control_dependences::control_dependence_map):
4899 (control_dependences::m_bitmaps): New.
4900 * cfganal.c (control_dependences::set_control_dependence_map_bit):
4902 (control_dependences::clear_control_dependence_bitmap):
4904 (control_dependences::find_control_dependence): Do not
4905 find_edge for the abnormal edge test.
4906 (control_dependences::control_dependences): Instead do not
4907 add abnormal edges to the edge list. Adjust.
4908 (control_dependences::~control_dependences): Likewise.
4909 (control_dependences::get_edges_dependent_on): Likewise.
4910 * function-tests.c: Include bitmap.h.
4912 2021-11-11 Kewen Lin <linkw@linux.ibm.com>
4914 * doc/invoke.texi: Change references to "future cpu" to "power10",
4915 "-mcpu=future" to "-mcpu=power10". Adjust words for float128.
4917 2021-11-11 Cui,Lili <lili.cui@intel.com>
4919 * config/i386/i386-options.c (m_CORE_AVX2): Remove Alderlake
4921 (processor_cost_table): Use alderlake_cost for Alderlake.
4922 * config/i386/i386.c (ix86_sched_init_global): Handle Alderlake.
4923 * config/i386/x86-tune-costs.h (struct processor_costs): Add alderlake
4925 * config/i386/x86-tune-sched.c (ix86_issue_rate): Change Alderlake
4927 (ix86_adjust_cost): Handle Alderlake.
4928 * config/i386/x86-tune.def (X86_TUNE_SCHEDULE): Enable for Alderlake.
4929 (X86_TUNE_PARTIAL_REG_DEPENDENCY): Likewise.
4930 (X86_TUNE_SSE_PARTIAL_REG_DEPENDENCY): Likewise.
4931 (X86_TUNE_SSE_PARTIAL_REG_FP_CONVERTS_DEPENDENCY): Likewise.
4932 (X86_TUNE_SSE_PARTIAL_REG_CONVERTS_DEPENDENCY): Likewise.
4933 (X86_TUNE_MEMORY_MISMATCH_STALL): Likewise.
4934 (X86_TUNE_USE_LEAVE): Likewise.
4935 (X86_TUNE_PUSH_MEMORY): Likewise.
4936 (X86_TUNE_USE_INCDEC): Likewise.
4937 (X86_TUNE_INTEGER_DFMODE_MOVES): Likewise.
4938 (X86_TUNE_MISALIGNED_MOVE_STRING_PRO_EPILOGUES): Likewise.
4939 (X86_TUNE_USE_SAHF): Likewise.
4940 (X86_TUNE_USE_BT): Likewise.
4941 (X86_TUNE_AVOID_FALSE_DEP_FOR_BMI): Likewise.
4942 (X86_TUNE_ONE_IF_CONV_INSN): Likewise.
4943 (X86_TUNE_AVOID_MFENCE): Likewise.
4944 (X86_TUNE_USE_SIMODE_FIOP): Likewise.
4945 (X86_TUNE_EXT_80387_CONSTANTS): Likewise.
4946 (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL): Likewise.
4947 (X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Likewise.
4948 (X86_TUNE_SSE_TYPELESS_STORES): Likewise.
4949 (X86_TUNE_SSE_LOAD0_BY_PXOR): Likewise.
4950 (X86_TUNE_AVOID_4BYTE_PREFIXES): Likewise.
4951 (X86_TUNE_USE_GATHER): Disable for Alderlake.
4952 (X86_TUNE_AVX256_MOVE_BY_PIECES): Likewise.
4953 (X86_TUNE_AVX256_STORE_BY_PIECES): Likewise.
4955 2021-11-11 liuhongt <hongtao.liu@intel.com>
4958 * config/i386/sse.md (V_128_256): Extend to V8HF/V16HF.
4959 (avxsizesuffix): Ditto.
4961 2021-11-11 Kito Cheng <kito.cheng@sifive.com>
4963 * common/config/riscv/riscv-common.c (riscv_subset_list::to_string): Fix
4964 wrong marco checking.
4966 2021-11-10 Aldy Hernandez <aldyh@redhat.com>
4968 PR tree-optimization/102906
4969 * tree-ssa-loop-ch.c (entry_loop_condition_is_static): New.
4970 (should_duplicate_loop_header_p): Call entry_loop_condition_is_static.
4971 (class ch_base): Add m_ranger and m_query.
4972 (ch_base::copy_headers): Pass m_query to
4973 entry_loop_condition_is_static.
4974 (pass_ch::execute): Allocate and deallocate m_ranger and
4976 (pass_ch_vect::execute): Same.
4978 2021-11-10 Andrew Pinski <apinski@marvell.com>
4981 * config/aarch64/aarch64-simd.md (aarch64_simd_dup<mode>):
4982 Use vwcore iterator for the r constraint output string.
4984 2021-11-10 qing zhao <qing.zhao@oracle.com>
4986 * internal-fn.c (expand_DEFERRED_INIT): Apply pattern initialization
4987 only when have_insn_for return true for the mode. Fix a memory leak.
4989 2021-11-10 Christophe Lyon <christophe.lyon@foss.st.com>
4991 * config/arm/arm.c (cortexa9_extra_costs, cortexa8_extra_costs,
4992 cortexa5_extra_costs, cortexa7_extra_costs,
4993 cortexa12_extra_costs, cortexa15_extra_costs, v7m_extra_costs):
4994 Initialize movi, dup and extract costing fields.
4996 2021-11-10 Aldy Hernandez <aldyh@redhat.com>
4998 * gimple-range-path.cc (path_range_query::path_range_query): Do
5000 (path_range_query::dump): Change m_path uses to non-pointer.
5001 (path_range_query::defined_outside_path): Same.
5002 (path_range_query::set_path): Same.
5003 (path_range_query::add_copies_to_imports): Same.
5004 (path_range_query::range_of_stmt): Same.
5005 (path_range_query::compute_outgoing_relations): Same.
5006 (path_range_query::compute_ranges): Imports are now optional.
5007 Implement overload that takes an edge.
5008 * gimple-range-path.h (class path_range_query): Make imports
5009 optional for compute_ranges. Add compute_ranges(edge) overload.
5010 Make m_path an auto_vec instead of a pointer and adjust
5013 2021-11-10 Tamar Christina <tamar.christina@arm.com>
5015 * tree-vectorizer.h (struct scalar_cond_masked_key): Add inverted_p.
5016 (default_hash_traits<scalar_conf_masked_key>): Likewise.
5017 * tree-vect-stmts.c (vectorizable_condition): Check if inverse of mask
5019 * tree-vectorizer.c (scalar_cond_masked_key::get_cond_ops_from_tree):
5020 Register mask inverses.
5022 2021-11-10 Tamar Christina <tamar.christina@arm.com>
5024 * tree-vectorizer.c (vectorize_loops): Do local CSE through RPVN upon
5025 successful vectorization.
5027 2021-11-10 Andrew MacLeod <amacleod@redhat.com>
5029 * gimple-range-cache.cc (sbr_vector::grow): New.
5030 (sbr_vector::set_bb_range): Call grow.
5031 (sbr_vector::get_bb_range): Same.
5032 (sbr_vector::bb_range_p): Remove assert.
5034 2021-11-10 Tamar Christina <tamar.christina@arm.com>
5036 * config/aarch64/aarch64-simd.md (*aarch64_topbits_shuffle<mode>_le
5037 ,*aarch64_topbits_shuffle<mode>_be): Remove.
5039 2021-11-10 Jan Hubicka <jh@suse.cz>
5041 * ipa-modref.c: Include tree-eh.h
5042 (modref_summary::modref_summary): Initialize side_effects.
5043 (struct modref_summary_lto): New bool field side_effects.
5044 (modref_summary_lto::modref_summary_lto): Initialize side_effects.
5045 (modref_summary::dump): Dump side_effects.
5046 (modref_summary_lto::dump): Dump side_effects.
5047 (merge_call_side_effects): Merge side effects.
5048 (process_fnspec): Calls to non-const/pure or looping
5049 function is a side effect.
5050 (analyze_call): Self-recursion is a side-effect; handle
5052 (analyze_load): Watch for volatile and throwing memory.
5053 (analyze_store): Likewise.
5054 (analyze_stmt): Watch for volatitle asm.
5055 (analyze_function): Handle side_effects.
5056 (modref_summaries::duplicate): Duplicate side_effects.
5057 (modref_summaries_lto::duplicate): Likewise.
5058 (modref_write): Stream side_effects.
5059 (read_section): Likewise.
5060 (update_signature): Update.
5061 (propagate_unknown_call): Handle side_effects.
5062 (modref_propagate_in_scc): Likewise.
5063 * ipa-modref.h (struct modref_summary): Add side_effects.
5064 * ipa-pure-const.c (special_builtin_state): Rename to ...
5065 (builtin_safe_for_const_function_p): ... this one.
5066 (check_call): Update.
5067 (finite_function_p): Break out from ...
5068 (propagate_pure_const): ... here
5069 * ipa-utils.h (finite_function): Declare.
5071 2021-11-10 Lucas A. M. Magalhães <lamm@linux.ibm.com>
5073 * config.gcc (powerpc*-*-*): Remove -rpath from
5074 --with-advance-toolchain.
5076 2021-11-10 Marek Polacek <polacek@redhat.com>
5079 * attribs.c (struct scoped_attributes): Add a bool member.
5080 (lookup_scoped_attribute_spec): Forward declare.
5081 (register_scoped_attributes): New bool parameter, defaulted to
5083 (handle_ignored_attributes_option): New function.
5084 (free_attr_data): New function.
5085 (init_attributes): Call handle_ignored_attributes_option.
5086 (attr_namespace_ignored_p): New function.
5087 (decl_attributes): Check attr_namespace_ignored_p before
5089 * attribs.h (free_attr_data): Declare.
5090 (register_scoped_attributes): Adjust declaration.
5091 (handle_ignored_attributes_option): Declare.
5092 (canonicalize_attr_name): New function template.
5093 (canonicalize_attr_name): Use it.
5094 * common.opt (Wattributes=): New option with a variable.
5095 * doc/extend.texi: Document #pragma GCC diagnostic ignored_attributes.
5096 * doc/invoke.texi: Document -Wno-attributes=.
5097 * opts.c (common_handle_option) <case OPT_Wattributes_>: Handle.
5098 * plugin.h (register_scoped_attributes): Adjust declaration.
5099 * toplev.c (compile_file): Call free_attr_data.
5101 2021-11-10 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
5103 * config/arm/arm-cpus.in (cortex-a710): New CPU.
5104 * config/arm/arm-tables.opt: Regenerate.
5105 * config/arm/arm-tune.md: Regenerate.
5106 * doc/invoke.texi: Update docs.
5108 2021-11-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
5110 * config/aarch64/aarch64-builtins.c
5111 (aarch64_general_gimple_fold_builtin): Mark argument as unused.
5113 2021-11-10 Martin Liska <mliska@suse.cz>
5115 * lto-wrapper.c (merge_and_complain): Make the first argument
5118 2021-11-10 Richard Sandiford <richard.sandiford@arm.com>
5120 * config/aarch64/iterators.md (optab): Use fmax_nan instead of
5121 smax_nan and fmin_nan instead of smin_nan.
5122 (maxmin_uns): Rename to...
5123 (fmaxmin): ...this and make the same changes. Remove entries
5124 unrelated to fmax* and fmin*.
5125 * config/aarch64/aarch64.md (<maxmin_uns><mode>3): Rename to...
5126 (<fmaxmin><mode>3): ...this.
5127 * config/aarch64/aarch64-simd.md (aarch64_<maxmin_uns>p<mode>):
5129 (aarch64_<optab>p<mode>): ...this.
5130 (<maxmin_uns><mode>3): Rename to...
5131 (<fmaxmin><mode>3): ...this.
5132 (reduc_<maxmin_uns>_scal_<mode>): Rename to...
5133 (reduc_<optab>_scal_<mode>): ...this and update gen* call.
5134 (aarch64_reduc_<maxmin_uns>_internal<mode>): Rename to...
5135 (aarch64_reduc_<optab>_internal<mode>): ...this.
5136 (aarch64_reduc_<maxmin_uns>_internalv2si): Rename to...
5137 (aarch64_reduc_<optab>_internalv2si): ...this.
5138 * config/aarch64/aarch64-sve.md (<maxmin_uns><mode>3): Rename to...
5139 (<fmaxmin><mode>3): ...this.
5140 * config/aarch64/aarch64-simd-builtins.def (smax_nan, smin_nan)
5142 (fmax_nan, fmin_nan): ...this.
5143 * config/aarch64/arm_neon.h (vmax_f32, vmax_f64, vmaxq_f32, vmaxq_f64)
5144 (vmin_f32, vmin_f64, vminq_f32, vminq_f64, vmax_f16, vmaxq_f16)
5145 (vmin_f16, vminq_f16): Update accordingly.
5147 2021-11-10 Richard Sandiford <richard.sandiford@arm.com>
5149 * tree-vectorizer.h (vector_costs::finish_cost): Take the
5150 corresponding scalar costs as a parameter.
5151 (finish_cost): Likewise.
5152 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
5153 (vect_estimate_min_profitable_iters): Update accordingly.
5154 * tree-vect-slp.c (vect_bb_vectorization_profitable_p): Likewise.
5155 * tree-vectorizer.c (vector_costs::finish_cost): Likewise.
5156 * config/aarch64/aarch64.c (aarch64_vector_costs::finish_cost):
5158 * config/rs6000/rs6000.c (rs6000_cost_data::finish_cost): Likewise.
5160 2021-11-10 Richard Sandiford <richard.sandiford@arm.com>
5162 * tree-vectorizer.h (_loop_vec_info::scalar_costs): New member
5164 (_loop_vec_info::single_scalar_iteration_cost): Delete.
5165 (LOOP_VINFO_SINGLE_SCALAR_ITERATION_COST): Delete.
5166 (vector_costs::total_cost): New function.
5167 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Update
5168 after above changes.
5169 (_loop_vec_info::~_loop_vec_info): Delete scalar_costs.
5170 (vect_compute_single_scalar_iteration_cost): Store the costs
5171 in loop_vinfo->scalar_costs.
5172 (vect_estimate_min_profitable_iters): Get the scalar cost from
5173 loop_vinfo->scalar_costs.
5175 2021-11-10 Richard Sandiford <richard.sandiford@arm.com>
5177 * tree-vectorizer.h (vector_costs::better_main_loop_than_p)
5178 (vector_costs::better_epilogue_loop_than_p)
5179 (vector_costs::compare_inside_loop_cost)
5180 (vector_costs::compare_outside_loop_cost): Likewise.
5181 * tree-vectorizer.c (vector_costs::better_main_loop_than_p)
5182 (vector_costs::better_epilogue_loop_than_p)
5183 (vector_costs::compare_inside_loop_cost)
5184 (vector_costs::compare_outside_loop_cost): New functions,
5185 containing code moved from...
5186 * tree-vect-loop.c (vect_better_loop_vinfo_p): ...here.
5188 2021-11-10 Richard Sandiford <richard.sandiford@arm.com>
5190 * tree-vectorizer.h (_loop_vec_info): Remove vec_outside_cost
5191 and vec_inside_cost.
5192 (vector_costs::outside_cost): New function.
5193 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Update
5195 (vect_estimate_min_profitable_iters): Likewise.
5196 (vect_better_loop_vinfo_p): Get the inside and outside costs
5197 from the loop_vec_infos' vector_costs.
5199 2021-11-10 Richard Sandiford <richard.sandiford@arm.com>
5201 * tree-vectorizer.h (vec_info::target_cost_data): Replace with...
5202 (_loop_vec_info::vector_costs): ...this.
5203 (LOOP_VINFO_TARGET_COST_DATA): Delete.
5204 * tree-vectorizer.c (vec_info::vec_info): Remove target_cost_data
5206 (vec_info::~vec_info): Remove corresponding delete.
5207 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
5208 vector_costs to null.
5209 (_loop_vec_info::~_loop_vec_info): Delete vector_costs.
5210 (vect_analyze_loop_operations): Update after above changes.
5211 (vect_analyze_loop_2): Likewise.
5212 (vect_estimate_min_profitable_iters): Likewise.
5213 * tree-vect-slp.c (vect_slp_analyze_operations): Likewise.
5215 2021-11-10 Jan Hubicka <hubicka@ucw.cz>
5217 * tree-core.h (EAF_DIRECT): Remove.
5218 (EAF_NOCLOBBER): Remove.
5219 (EAF_UNUSED): Remove.
5220 (EAF_NOESCAPE): Remove.
5221 (EAF_NO_DIRECT_CLOBBER): New.
5222 (EAF_NO_INDIRECT_CLOBBER): New.
5223 (EAF_NODIRECTESCAPE): Remove.
5224 (EAF_NO_DIRECT_ESCAPE): New.
5225 (EAF_NO_INDIRECT_ESCAPE): New.
5226 (EAF_NOT_RETURNED): Remove.
5227 (EAF_NOT_RETURNED_INDIRECTLY): New.
5228 (EAF_NOREAD): Remove.
5229 (EAF_NO_DIRECT_READ): New.
5230 (EAF_NO_INDIRECT_READ): New.
5231 * gimple.c (gimple_call_arg_flags): Update for new flags.
5232 (gimple_call_retslot_flags): Update for new flags.
5233 * ipa-modref.c (dump_eaf_flags): Likewise.
5234 (remove_useless_eaf_flags): Likewise.
5235 (deref_flags): Likewise.
5236 (modref_lattice::init): Likewise.
5237 (modref_lattice::merge): Likewise.
5238 (modref_lattice::merge_direct_load): Likewise.
5239 (modref_lattice::merge_direct_store): Likewise.
5240 (modref_eaf_analysis::merge_call_lhs_flags): Likewise.
5241 (callee_to_caller_flags): Likewise.
5242 (modref_eaf_analysis::analyze_ssa_name): Likewise.
5243 (modref_eaf_analysis::propagate): Likewise.
5244 (modref_merge_call_site_flags): Likewise.
5245 * ipa-modref.h (interposable_eaf_flags): Likewise.
5246 * tree-ssa-alias.c: (ref_maybe_used_by_call_p_1) Likewise.
5247 * tree-ssa-structalias.c (handle_call_arg): Likewise.
5248 (handle_rhs_call): Likewise.
5249 * tree-ssa-uninit.c (maybe_warn_pass_by_reference): Likewise.
5251 2021-11-10 Tamar Christina <tamar.christina@arm.com>
5253 * tree-vect-slp-patterns.c (complex_mul_pattern::matches): Remove l1node.
5255 2021-11-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
5257 * config/aarch64/aarch64-builtins.c
5258 (aarch64_general_gimple_fold_builtin): Change pointer alignment and
5261 2021-11-10 Jan Hubicka <jh@suse.cz>
5263 * ipa-modref-tree.h (modref_tree::remap_params): Fix off-by-one error.
5265 2021-11-10 H.J. Lu <hongjiu.lu@intel.com>
5266 Hongtao Liu <hongtao.liu@intel.com>
5268 PR middle-end/102566
5269 * match.pd (nop_atomic_bit_test_and_p): New match.
5270 * tree-ssa-ccp.c (convert_atomic_bit_not): New function.
5271 (gimple_nop_atomic_bit_test_and_p): New prototype.
5272 (optimize_atomic_bit_test_and): Transform equivalent, but slighly
5273 different cases to their canonical forms.
5275 2021-11-10 liuhongt <hongtao.liu@intel.com>
5277 PR tree-optimization/103126
5278 * tree-vect-loop.c (neutral_op_for_reduction): Remove static.
5279 * tree-vectorizer.h (neutral_op_for_reduction): Declare.
5280 * tree-if-conv.c : Include tree-vectorizer.h.
5281 (is_cond_scalar_reduction): Handle
5282 BIT_XOR_EXPR/BIT_IOR_EXPR/BIT_AND_EXPR.
5283 (convert_scalar_cond_reduction): Ditto.
5285 2021-11-10 konglin1 <lingling.kong@intel.com>
5287 * config/i386/sse.md (cmul<conj_op><mode>3): add new define_expand.
5288 (cmla<conj_op><mode>4): Likewise
5290 2021-11-10 Aldy Hernandez <aldyh@redhat.com>
5292 * tree-ssa-threadedge.c: Do not include
5293 gimple-ssa-evrp-analyze.h.
5294 * value-pointer-equiv.cc: Same.
5296 2021-11-10 Aldy Hernandez <aldyh@redhat.com>
5298 * tree-ssa-threadbackward.c
5299 (back_threader::maybe_register_path_dump): Abstract path dumping...
5300 (dump_path): ...here.
5301 (back_threader::resolve_phi): Call dump_path.
5304 2021-11-10 konglin1 <lingling.kong@intel.com>
5306 * config/i386/sse.md (fma_<complexpairopname>_<mode>_pair):
5307 Add new define_insn.
5308 (fma_<mode>_fmaddc_bcst): Add new define_insn_and_split.
5309 (fma_<mode>_fcmaddc_bcst): Likewise
5311 2021-11-10 liuhongt <hongtao.liu@intel.com>
5314 * match.pd: Simplify (trunc)fmax/fmin((extend)a, (extend)b) to
5317 2021-11-10 Andrew Pinski <apinski@marvell.com>
5320 * config/aarch64/aarch64.c (aarch64_evpc_ins): Don't use target
5321 as an input, use original one.
5323 2021-11-10 Sandra Loosemore <sandra@codesourcery.com>
5325 * config/nios2/nios2.c (nios2_can_inline_p): New.
5326 (TARGET_CAN_INLINE_P): Define.
5328 2021-11-09 Bill Schmidt <wschmidt@linux.ibm.com>
5330 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_new_builtin):
5331 Disable gimple fold for RS6000_BIF_{XVMINDP,XVMINSP,VMINFP} and
5332 RS6000_BIF_{XVMAXDP,XVMAXSP,VMAXFP} when fast-math is not set.
5333 (lxvrse_expand_builtin): Modify the expansion for sign extension.
5334 All extensions are done within VSX registers.
5336 2021-11-09 Andrew Pinski <apinski@marvell.com>
5338 PR tree-optimization/10352
5339 * match.pd: Remove check of TYPE_PRECISION for
5342 2021-11-09 Andrew MacLeod <amacleod@redhat.com>
5344 * function.c (allocate_struct_function): Don't set x_range_query.
5345 * function.h (get_range_query): Move to value-query.h.
5346 * gimple-range.cc (enable_ranger): Check that query is currently NULL.
5347 (disable_ranger): Clear function current query field.
5348 * value-query.cc (get_global_range_query): Relocate to:
5349 * value-query.h (get_global_range_query): Here and inline.
5350 (get_range_query): Relocate here from function.h.
5352 2021-11-09 Aldy Hernandez <aldyh@redhat.com>
5354 * tree-ssa-threadbackward.c
5355 (back_threader::maybe_register_path_dump): New.
5356 (back_threader::maybe_register_path): Call maybe_register_path_dump.
5358 2021-11-09 Aldy Hernandez <aldyh@redhat.com>
5360 * tree-ssa-threadbackward.c (back_threader::maybe_register_path):
5361 Return NULL when unprofitable.
5363 2021-11-09 Martin Jambor <mjambor@suse.cz>
5365 * tree.h (build_debug_expr_decl): Declare.
5366 * tree.c (build_debug_expr_decl): New function.
5367 * cfgexpand.c (avoid_deep_ter_for_debug): Use build_debug_expr_decl
5368 instead of building a DEBUG_EXPR_DECL.
5369 * ipa-param-manipulation.c
5370 (ipa_param_body_adjustments::prepare_debug_expressions): Likewise.
5371 * omp-simd-clone.c (ipa_simd_modify_stmt_ops): Likewise.
5372 * tree-ssa-ccp.c (optimize_atomic_bit_test_and): Likewise.
5373 * tree-ssa-phiopt.c (spaceship_replacement): Likewise.
5374 * tree-ssa-reassoc.c (make_new_ssa_for_def): Likewise.
5376 2021-11-09 Bill Schmidt <wschmidt@linux.ibm.com>
5378 * config/rs6000/rs6000-call.c (rs6000_type_string): New function.
5379 (def_builtin): Change debug formatting for easier parsing and
5380 include more information.
5381 (rs6000_init_builtins): Add dump of autogenerated builtins.
5382 (altivec_init_builtins): Dump __builtin_altivec_mask_for_load for
5385 2021-11-09 Marek Polacek <polacek@redhat.com>
5387 * ipa-modref.c (analyze_parms): Fix a typo.
5389 2021-11-09 Kito Cheng <kito.cheng@sifive.com>
5392 * common/config/riscv/riscv-common.c (multi_letter_subset_rank): Remove
5393 assertion for Z*-ext.
5395 2021-11-09 Jan Hubicka <hubicka@ucw.cz>
5397 * ipa-modref.c (analyze_parms): Add past_flags, past_retslot_flags
5398 and past_static_chain; merge past summary with current one.
5399 (analyze_function): Update.
5401 2021-11-09 Aldy Hernandez <aldyh@redhat.com>
5403 * doc/invoke.texi (Invoking GCC): Document --param=threader-debug.
5405 2021-11-09 Martin Liska <mliska@suse.cz>
5407 * print-rtl.c (rtx_writer::rtx_writer): Make the compilation
5408 conditional based on
5409 * print-rtl.h (class rtx_writer): Likewise.
5411 2021-11-09 Thomas Schwinge <thomas@codesourcery.com>
5413 * input.h (location_hash): Use 'BUILTINS_LOCATION' as spare value
5414 for 'Deleted'. Turn into a '#define'.
5416 2021-11-09 Aldy Hernandez <aldyh@redhat.com>
5418 * gimple-range-path.cc (path_range_query::dump): Clean up.
5419 (path_range_query::compute_ranges): Same.
5420 * value-relation.cc (path_oracle::dump): Same.
5422 2021-11-09 Aldy Hernandez <aldyh@redhat.com>
5424 * dumpfile.c (dump_options): Remove TDF_THREADING entry.
5425 * dumpfile.h (enum dump_flag): Remove TDF_THREADING and adjust
5427 * flag-types.h (enum threader_debug): New.
5428 * gimple-range-path.cc (DEBUG_SOLVER): Use param_threader_debug.
5429 * params.opt: Add entry for --param=threader-debug=.
5431 2021-11-09 Thomas Schwinge <thomas@codesourcery.com>
5433 PR middle-end/101204
5435 * diagnostic-spec.h (typedef xint_hash_t): Turn into...
5436 (struct xint_hash_t): ... this.
5437 * doc/gty.texi: Update.
5439 2021-11-09 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
5441 * config/arm/arm-cpus.in (armv9): New define.
5442 (ARMv9a): New group.
5443 (armv9-a): New arch definition.
5444 * config/arm/arm-tables.opt: Regenerate.
5445 * config/arm/arm.h (BASE_ARCH_9A): New arch enum value.
5446 * config/arm/t-aprofile: Added armv9-a and armv9+simd.
5447 * config/arm/t-arm-elf: Added arm9-a, v9_fps and all_v9_archs
5448 to MULTILIB_MATCHES.
5449 * config/arm/t-multilib: Added v9_a_nosimd_variants and
5450 v9_a_simd_variants to MULTILIB_MATCHES.
5451 * doc/invoke.texi: Update docs.
5453 2021-11-09 Martin Jambor <mjambor@suse.cz>
5456 * ipa-param-manipulation.c (replace_with_mapped_expr): Early
5457 return with error_mark_mode when part of expression is mapped to
5459 (ipa_param_body_adjustments::remap_with_debug_expressions): Set
5460 mapped value to NULL if walk_tree returns error_mark_mode.
5462 2021-11-09 Aldy Hernandez <aldyh@redhat.com>
5464 * tree-ssa-strlen.c (compare_nonzero_chars): Pass statement
5466 (get_addr_stridx): Same.
5468 (get_range_strlen_dynamic): Same.
5469 (handle_builtin_strlen): Same.
5470 (handle_builtin_strchr): Same.
5471 (handle_builtin_strcpy): Same.
5472 (maybe_diag_stxncpy_trunc): Same.
5473 (handle_builtin_stxncpy_strncat): Same.
5474 (handle_builtin_memcpy): Same.
5475 (handle_builtin_strcat): Same.
5476 (handle_alloc_call): Same.
5477 (handle_builtin_memset): Same.
5478 (handle_builtin_string_cmp): Same.
5479 (handle_pointer_plus): Same.
5480 (count_nonzero_bytes_addr): Same.
5481 (count_nonzero_bytes): Same.
5482 (handle_store): Same.
5483 (fold_strstr_to_strncmp): Same.
5484 (handle_integral_assign): Same.
5485 (check_and_optimize_stmt): Same.
5486 (class strlen_dom_walker): Replace evrp with ranger.
5487 (strlen_dom_walker::before_dom_children): Remove evrp.
5488 (strlen_dom_walker::after_dom_children): Remove evrp.
5489 * gimple-ssa-warn-access.cc (maybe_check_access_sizes):
5490 Restrict sprintf output.
5492 2021-11-09 Martin Liska <mliska@suse.cz>
5494 * genconditions.c (write_one_condition): Add const qualifier
5497 2021-11-09 Jeff Law <jeffreyalaw@gmail.com>
5499 * match.pd: New pattern to simplify (1 << n) & M ==/!= 0 for M
5502 2021-11-08 Bill Schmidt <wschmidt@linux.ibm.com>
5504 * config/rs6000/rs6000-builtin-new.def (VEC_INIT_V16QI): Use
5505 escape-newline support.
5506 (VEC_INIT_V4SI): Likewise.
5507 (VEC_INIT_V8HI): Likewise.
5508 (PACK_V1TI): Likewise.
5510 (VFIRSTMISMATCHOREOSINDEX_V16QI): Likewise.
5511 (VFIRSTMISMATCHOREOSINDEX_V8HI): Likewise.
5512 (VFIRSTMISMATCHOREOSINDEX_V4SI): Likewise.
5516 (FMAF128_ODD): Likewise.
5517 (VSCEQPUO): Likewise.
5519 (VSIEQPF): Likewise.
5520 (VSTDCQP): Likewise.
5521 (PACK_TD): Likewise.
5522 (TABORTDC): Likewise.
5523 (TABORTDCI): Likewise.
5524 (SE_LXVRBX): Likewise.
5525 (SE_LXVRHX): Likewise.
5526 (SE_LXVRWX): Likewise.
5527 (SE_LXVRDX): Likewise.
5528 (VREPLACE_UN_UV2DI): Likewise.
5529 (VREPLACE_UN_UV4SI): Likewise.
5530 (VREPLACE_UN_V2DI): Likewise.
5531 (VREPLACE_ELT_UV2DI): Likewise.
5532 (VREPLACE_ELT_V2DI): Likewise.
5533 (ZE_LXVRBX): Likewise.
5534 (ZE_LXVRHX): Likewise.
5535 (ZE_LXVRWX): Likewise.
5536 (ZE_LXVRDX): Likewise.
5538 (CNTLZDM): Likewise.
5539 (CNTTZDM): Likewise.
5542 (PMXVBF16GER2): Likewise.
5543 (PMXVBF16GER2_INTERNAL): Likewise.
5544 (PMXVBF16GER2NN): Likewise.
5545 (PMXVBF16GER2NN_INTERNAL): Likewise.
5546 (PMXVBF16GER2NP): Likewise.
5547 (PMXVBF16GER2NP_INTERNAL): Likewise.
5548 (PMXVBF16GER2PN): Likewise.
5549 (PMXVBF16GER2PN_INTERNAL): Likewise.
5550 (PMXVBF16GER2PP): Likewise.
5551 (PMXVBF16GER2PP_INTERNAL): Likewise.
5552 (PMXVF16GER2): Likewise.
5553 (PMXVF16GER2_INTERNAL): Likewise.
5554 (PMXVF16GER2NN): Likewise.
5555 (PMXVF16GER2NN_INTERNAL): Likewise.
5556 (PMXVF16GER2NP): Likewise.
5557 (PMXVF16GER2NP_INTERNAL): Likewise.
5558 (PMXVF16GER2PN): Likewise.
5559 (PMXVF16GER2PN_INTERNAL): Likewise.
5560 (PMXVF16GER2PP): Likewise.
5561 (PMXVF16GER2PP_INTERNAL): Likewise.
5562 (PMXVF32GER_INTERNAL): Likewise.
5563 (PMXVF32GERNN): Likewise.
5564 (PMXVF32GERNN_INTERNAL): Likewise.
5565 (PMXVF32GERNP): Likewise.
5566 (PMXVF32GERNP_INTERNAL): Likewise.
5567 (PMXVF32GERPN): Likewise.
5568 (PMXVF32GERPN_INTERNAL): Likewise.
5569 (PMXVF32GERPP): Likewise.
5570 (PMXVF32GERPP_INTERNAL): Likewise.
5571 (PMXVF64GER): Likewise.
5572 (PMXVF64GER_INTERNAL): Likewise.
5573 (PMXVF64GERNN): Likewise.
5574 (PMXVF64GERNN_INTERNAL): Likewise.
5575 (PMXVF64GERNP): Likewise.
5576 (PMXVF64GERNP_INTERNAL): Likewise.
5577 (PMXVF64GERPN): Likewise.
5578 (PMXVF64GERPN_INTERNAL): Likewise.
5579 (PMXVF64GERPP): Likewise.
5580 (PMXVF64GERPP_INTERNAL): Likewise.
5581 (PMXVI16GER2): Likewise.
5582 (PMXVI16GER2_INTERNAL): Likewise.
5583 (PMXVI16GER2PP): Likewise.
5584 (PMXVI16GER2PP_INTERNAL): Likewise.
5585 (PMXVI16GER2S): Likewise.
5586 (PMXVI16GER2S_INTERNAL): Likewise.
5587 (PMXVI16GER2SPP): Likewise.
5588 (PMXVI16GER2SPP_INTERNAL): Likewise.
5589 (PMXVI4GER8): Likewise.
5590 (PMXVI4GER8_INTERNAL): Likewise.
5591 (PMXVI4GER8PP): Likewise.
5592 (PMXVI4GER8PP_INTERNAL): Likewise.
5593 (PMXVI8GER4): Likewise.
5594 (PMXVI8GER4_INTERNAL): Likewise.
5595 (PMXVI8GER4PP): Likewise.
5596 (PMXVI8GER4PP_INTERNAL): Likewise.
5597 (PMXVI8GER4SPP): Likewise.
5598 (PMXVI8GER4SPP_INTERNAL): Likewise.
5599 * config/rs6000/rs6000-gen-builtins.c (MAXLINES): New macro.
5600 (linebuf): Increase size.
5601 (lines): New variable.
5602 (lastline): Likewise.
5603 (real_line_pos): New function.
5604 (diag): Change signature.
5605 (bif_diag): Change signature; support escape-newline handling.
5606 (ovld_diag): Likewise.
5607 (fatal): Move earlier.
5608 (consume_whitespace): Adjust diag call.
5609 (advance_line): Add escape-newline handling; call fatal.
5610 (safe_inc_pos): Adjust diag call.
5611 (match_identifier): Likewise.
5612 (match_integer): Likewise.
5613 (match_to_right_bracket): Call fatal instead of diag; adjust diag
5615 (match_basetype): Adjust diag calls.
5616 (match_bracketed_pair): Likewise.
5617 (match_const_restriction): Likewise.
5618 (match_type): Likewise.
5619 (parse_args): Likewise.
5620 (parse_bif_attrs): Likewise.
5621 (complete_vector_type): Likewise.
5622 (complete_base_type): Likewise.
5623 (parse_prototype): Likewise.
5624 (parse_bif_entry): Likewise.
5625 (parse_bif_stanza): Likewise.
5626 (parse_ovld_entry): Likewise.
5627 (parse_ovld_stanza): Likewise.
5628 (main): Allocate buffers for lines[].
5630 2021-11-08 Eric Botcazou <ebotcazou@adacore.com>
5632 * config/i386/i386.c (legitimize_pic_address): Adjust comment and
5633 use the REG argument on the CM_LARGE_PIC code path as well.
5634 * config/i386/predicates.md (gotoff_operand): Do not treat VxWorks
5635 specially with the large code models.
5637 2021-11-08 Jan Hubicka <hubicka@ucw.cz>
5639 * gimple.c (gimple_call_static_chain_flags): Revert the workaround
5640 allowing interposition since issues with binds_to_local_def were
5643 2021-11-08 Andrew MacLeod <amacleod@redhat.com>
5645 PR tree-optimization/103122
5646 * gimple-range.cc (gimple_ranger::range_of_expr): Request the cache
5647 entry with "calulate new values" set to false.
5649 2021-11-08 Jan Hubicka <hubicka@ucw.cz>
5651 * builtins.c (is_simple_builtin): Add builitin_dwarf_cfa
5652 and builtin_return_address.
5653 (builtin_fnspec): Annotate builtin_return,
5654 bulitin_eh_pointer, builtin_eh_filter, builtin_unwind_resume,
5655 builtin_cxa_end_cleanup, builtin_eh_copy_values,
5656 builtin_frame_address, builtin_apply_args,
5657 builtin_asan_before_dynamic_init, builtin_asan_after_dynamic_init,
5658 builtin_prefetch, builtin_dwarf_cfa, builtin_return_addrss
5660 * ipa-pure-const.c (special_builtin_state): Add builtin_dwarf_cfa
5661 and builtin_return_address.
5663 2021-11-08 Jan Hubicka <hubicka@ucw.cz>
5665 PR tree-optimization/103177
5666 * passes.def: Move uncprop after pure/const and modref.
5668 2021-11-08 Martin Jambor <mjambor@suse.cz>
5672 * tree-inline.c (remap_gimple_stmt): Unshare the expression without
5673 location before invoking remap_with_debug_expressions on it.
5674 * ipa-param-manipulation.c
5675 (ipa_param_body_adjustments::prepare_debug_expressions): Likewise.
5677 2021-11-08 David Edelsohn <dje.gcc@gmail.com>
5679 * config/rs6000/vsx.md (vsx_splat_v4si_di): Revert "wa"
5682 2021-11-08 Richard Biener <rguenther@suse.de>
5684 * cfgloop.c (verify_loop_structure): Use a temporary BB flag
5685 instead of an sbitmap to cache irreducible state.
5687 2021-11-08 Aldy Hernandez <aldyh@redhat.com>
5689 PR tree-optimization/103120
5690 * gimple-range-path.cc (path_range_query::range_defined_in_block):
5691 Bail if there's a cache entry.
5693 2021-11-08 Bill Schmidt <wschmidt@linux.ibm.com>
5695 * config/rs6000/rs6000.c (rs6000_builtin_reciprocal): Use
5696 rs6000_builtin_decls_x when appropriate.
5697 (add_condition_to_bb): Likewise.
5698 (rs6000_atomic_assign_expand_fenv): Likewise.
5700 2021-11-08 Bill Schmidt <wschmidt@linux.ibm.com>
5702 * config/rs6000/rs6000-call.c (rs6000_new_builtin_decl): New function.
5703 (rs6000_builtin_decl): Call it.
5705 2021-11-08 Martin Liska <mliska@suse.cz>
5707 * cgraph.c (cgraph_node::dump): Dump it from decl.
5709 2021-11-08 Martin Liska <mliska@suse.cz>
5711 PR gcov-profile/100520
5712 * coverage.c (coverage_compute_profile_id): Strip .gk when
5713 compare debug is used.
5714 * system.h (endswith): New function.
5716 2021-11-08 Martin Liska <mliska@suse.cz>
5718 * cgraph.c (cgraph_node::dump): Dump static_chain_decl.
5720 2021-11-08 Thomas Schwinge <thomas@codesourcery.com>
5722 * config/rs6000/rbtree.c: Fix 'Copyright (C) 2020-21' into '2020-2021'
5723 * config/rs6000/rbtree.h: Likewise.
5724 * config/rs6000/rs6000-builtin-new.def: Likewise.
5725 * config/rs6000/rs6000-gen-builtins.c: Likewise.
5726 * config/rs6000/rs6000-overload.def: Likewise.
5728 2021-11-08 Richard Sandiford <richard.sandiford@arm.com>
5730 * config/aarch64/aarch64.c (aarch64_ldn_stn_vectors): New function.
5731 (aarch64_address_cost): Use it instead of testing for CImode and
5734 2021-11-08 Richard Sandiford <richard.sandiford@arm.com>
5736 * genmodes.c (emit_insn_modes_h): Define NUM_MODE_* macros.
5737 * expmed.h (NUM_MODE_INT): Delete in favor of genmodes definitions.
5738 (NUM_MODE_PARTIAL_INT, NUM_MODE_VECTOR_INT): Likewise.
5739 * real.h (real_format_for_mode): Use NUM_MODE_FLOAT and
5740 NUM_MODE_DECIMAL_FLOAT.
5741 (REAL_MODE_FORMAT): Likewise.
5743 2021-11-08 Andre Vieira <andre.simoesdiasvieira@arm.com>
5745 * tree-vect-loop.c (vect_better_loop_vinfo_p): Change how epilogue loop
5748 2021-11-08 Richard Biener <rguenther@suse.de>
5750 * tree-vectorizer.h (vect_create_loop_vinfo): Add main_loop_info
5752 * tree-vect-loop.c (vect_create_loop_vinfo): Likewise. Set
5753 LOOP_VINFO_ORIG_LOOP_INFO and conditionalize set of
5754 LOOP_VINFO_NITERS_ASSUMPTIONS.
5755 (vect_analyze_loop_1): Adjust.
5756 (vect_analyze_loop): Move loop constraint setting and
5757 SCEV/niter reset here from vect_create_loop_vinfo to perform
5759 (vect_analyze_loop_form): Move dumping of symbolic niters
5760 here from vect_create_loop_vinfo.
5762 2021-11-08 Jan Hubicka <hubicka@ucw.cz>
5764 * ipa-modref-tree.h (enum modref_special_parms): New enum.
5765 (struct modref_access_node): update for special parms.
5766 (struct modref_ref_node): Likewise.
5767 (struct modref_parm_map): Likewise.
5768 (struct modref_tree): Likewise.
5769 * ipa-modref.c (dump_access): Likewise.
5770 (get_access): Detect static chain.
5771 (parm_map_for_arg): Take tree as arg instead of
5773 (merge_call_side_effects): Compute map for static chain.
5774 (process_fnspec): Update.
5775 (struct escape_point): Remove retslot_arg and static_chain_arg.
5776 (analyze_parms): Update.
5777 (compute_parm_map): Update.
5778 (propagate_unknown_call): Update.
5779 (modref_propagate_in_scc): Update.
5780 (modref_merge_call_site_flags): Update.
5781 (ipa_merge_modref_summary_after_inlining): Update.
5782 * tree-ssa-alias.c (modref_may_conflict): Handle static chain.
5783 * ipa-modref-tree.c (test_merge): Update.
5785 2021-11-08 Haochen Gui <guihaoc@gcc.gnu.org>
5787 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_builtin): Disable
5788 gimple fold for VSX_BUILTIN_XVMINDP, ALTIVEC_BUILTIN_VMINFP,
5789 VSX_BUILTIN_XVMAXDP, ALTIVEC_BUILTIN_VMAXFP when fast-math is not
5792 2021-11-08 liuhongt <hongtao.liu@intel.com>
5794 PR tree-optimization/103077
5795 * doc/invoke.texi (Options That Control Optimization):
5796 Update documentation for -ftree-loop-vectorize and
5797 -ftree-slp-vectorize which are enabled by default at -02.
5799 2021-11-08 liuhongt <hongtao.liu@intel.com>
5802 * match.pd (Simplifcation (trunc)copysign((extend)a, (extend)b)
5803 to .COPYSIGN (a, b)): Add !HONOR_SNANS.
5805 2021-11-08 liuhongt <hongtao.liu@intel.com>
5808 * match.pd: Simplify
5809 (trunc)fma ((extend)a, (extend)b, (extend)c) to IFN_FMA (a, b,
5810 c) under flag_unsafe_math_optimizations.
5812 2021-11-07 Jan Hubicka <hubicka@ucw.cz>
5815 * params.opt (modref-max-depth): Add range.
5816 (modref-max-adjustments): Fix range.
5818 2021-11-07 Aldy Hernandez <aldyh@redhat.com>
5820 * tree-pass.h (make_pass_vrp_threader): Remove.
5821 * tree-ssa-threadbackward.c
5822 (back_threader_profitability::profitable_path_p): Remove
5823 ASSERT_EXPR references.
5824 * tree-ssa-threadedge.c (jt_state::register_equivs_stmt): Same.
5825 * tree-vrp.c (vrp_folder::simplify_casted_conds): Same.
5826 (execute_vrp): Same.
5827 (class hybrid_threader): Remove.
5828 (hybrid_threader::hybrid_threader): Remove.
5829 (hybrid_threader::~hybrid_threader): Remove.
5830 (hybrid_threader::before_dom_children): Remove.
5831 (hybrid_threader::after_dom_children): Remove.
5832 (execute_vrp_threader): Remove.
5833 (class pass_vrp_threader): Remove.
5834 (make_pass_vrp_threader): Remove.
5836 2021-11-07 Jan Hubicka <hubicka@ucw.cz>
5838 * gimple.c (gimple_call_arg_flags): Use interposable_eaf_flags.
5839 (gimple_call_retslot_flags): Likewise.
5840 (gimple_call_static_chain_flags): Likewise.
5841 * ipa-modref.c (remove_useless_eaf_flags): Do not remove everything for
5843 (modref_summary::useful_p): Likewise.
5844 (modref_summary_lto::useful_p): Likewise.
5845 (analyze_parms): Do not give up on NOVOPS.
5846 (analyze_function): When dumping report chnages in EAF flags
5847 between IPA and local pass.
5848 (modref_merge_call_site_flags): Compute implicit eaf flags
5849 based on callee ecf_flags and fnspec; if the function does not
5850 bind to current defs use interposable_eaf_flags.
5851 (modref_propagate_flags_in_scc): Update.
5852 * ipa-modref.h (interposable_eaf_flags): New function.
5854 2021-11-07 Bill Schmidt <wschmidt@linux.ibm.com>
5856 * config/rs6000/rs6000-call.c (rs6000_expand_new_builtin): New
5858 (rs6000_invalid_new_builtin): New function.
5859 (rs6000_expand_builtin): Call rs6000_expand_new_builtin.
5860 (rs6000_expand_ldst_mask): New function.
5861 (new_cpu_expand_builtin): Likewise.
5862 (elemrev_icode): Likewise.
5863 (ldv_expand_builtin): Likewise.
5864 (lxvrse_expand_builtin): Likewise.
5865 (lxvrze_expand_builtin): Likewise.
5866 (stv_expand_builtin): Likewise.
5867 (new_mma_expand_builtin): Likewise.
5868 (new_htm_spr_num): Likewise.
5869 (new_htm_expand_builtin): Likewise.
5870 (rs6000_expand_new_builtin): Likewise.
5871 (rs6000_init_builtins): Initialize altivec_builtin_mask_for_load.
5873 2021-11-07 Jan Hubicka <hubicka@ucw.cz>
5875 * ipa-modref.c (modref_lattice): Add do_dataflow,
5876 changed and propagate_to fields.
5877 (modref_lattice::release): Free propagate_to
5878 (modref_lattice::merge): Do not give up early on unknown
5880 (modref_lattice::merge_deref): Likewise.
5881 (modref_eaf_analysis): Update toplevel comment.
5882 (modref_eaf_analysis::analyze_ssa_name): Record postponned ssa names;
5883 do optimistic dataflow initialization.
5884 (modref_eaf_analysis::merge_with_ssa_name): Build dataflow graph.
5885 (modref_eaf_analysis::propagate): New member function.
5886 (analyze_parms): Update to new API of modref_eaf_analysis.
5888 2021-11-06 Jan Hubicka <hubicka@ucw.cz>
5890 * cgraph.h (cgraph_node::can_be_discarded_p): Do not
5891 return true on functions from other partition.
5893 2021-11-06 David Edelsohn <dje.gcc@gmail.com>
5895 * config/rs6000/vsx.md (vsx_splat_v4si): Change constraints to "wa".
5896 (vsx_splat_v4si_di): Change constraint to "wa".
5898 2021-11-06 Aldy Hernandez <aldyh@redhat.com>
5900 PR tree-optimization/103061
5901 * value-relation.cc (path_oracle::path_oracle): Initialize
5903 (path_oracle::killing_def): Set m_killed_defs.
5904 (path_oracle::query_relation): Do not look at the root oracle for
5906 * value-relation.h (class path_oracle): Add m_killed_defs.
5908 2021-11-06 Aldy Hernandez <aldyh@redhat.com>
5910 * tree-ssa-threadbackward.c (back_threader::find_paths_to_names):
5911 Remove gotos and other cleanups.
5913 2021-11-05 Jan Hubicka <hubicka@ucw.cz>
5916 * ipa-modref-tree.h (modref_tree::insert): Do nothing for
5917 paradoxical and zero sized accesses.
5919 2021-11-05 Jan Hubicka <hubicka@ucw.cz>
5922 * ipa-modref-tree.h (struct modref_access_node): Avoid left shift
5925 2021-11-05 Iain Sandoe <iain@sandoe.co.uk>
5927 * config/darwin.h (ASM_GENERATE_INTERNAL_LABEL): Add LTRAMP
5928 to the list of symbol prefixes that must be made linker-
5931 2021-11-05 Iain Sandoe <iain@sandoe.co.uk>
5933 * config.host: Add support for aarch64-*-darwin.
5934 * config/aarch64/host-aarch64-darwin.c: New file.
5935 * config/aarch64/x-darwin: New file.
5937 2021-11-05 Andrew MacLeod <amacleod@redhat.com>
5939 PR tree-optimization/103093
5940 * gimple-range-gori.cc (range_def_chain::get_imports): Remove assert.
5942 2021-11-05 Andrew MacLeod <amacleod@redhat.com>
5944 PR tree-optimization/102943
5945 * gimple-range-cache.cc (class update_list): New.
5946 (update_list::add): Replace add_to_update.
5947 (update_list::pop): New.
5948 (ranger_cache::ranger_cache): Adjust.
5949 (ranger_cache::~ranger_cache): Adjust.
5950 (ranger_cache::add_to_update): Delete.
5951 (ranger_cache::propagate_cache): Adjust to new class.
5952 (ranger_cache::propagate_updated_value): Ditto.
5953 (ranger_cache::fill_block_cache): Ditto.
5954 * gimple-range-cache.h (class ranger_cache): Adjust to update class.
5956 2021-11-05 Richard Biener <rguenther@suse.de>
5958 * tree-vect-loop.c (vect_analyze_loop): Remove obsolete
5959 comment and expand on another one. Combine nested if.
5961 2021-11-05 John David Anglin <danglin@gcc.gnu.org>
5964 * config/pa/pa-modes.def: Add OImode integer type.
5965 * config/pa/pa.c (pa_scalar_mode_supported_p): Allow TImode
5967 * config/pa/pa.h (MIN_UNITS_PER_WORD) Define to MIN_UNITS_PER_WORD
5968 to UNITS_PER_WORD if IN_LIBGCC2.
5969 * config/pa/pa.md (addti3, addvti3, subti3, subvti3, negti2,
5970 negvti2, ashlti3, shrpd_internal): New patterns.
5971 Change some multi instruction types to multi.
5973 2021-11-05 Jakub Jelinek <jakub@redhat.com>
5976 * config/i386/i386.h
5977 (stringop_algs::stringop_strategy::stringop_strategy): Make the ctor
5980 2021-11-05 Wilco Dijkstra <wdijkstr@arm.com>
5983 * config/aarch64/aarch64.c (aarch64_mov_operand_p): Strip the salt
5985 * config/aarch64/constraints.md: Support const in Usw.
5987 2021-11-05 John David Anglin <danglin@gcc.gnu.org>
5989 * config/pa/pa.h (PREFERRED_DEBUGGING_TYPE): Define to DWARF2_DEBUG.
5990 * config/pa/pa64-hpux.h (PREFERRED_DEBUGGING_TYPE): Remove define.
5992 2021-11-05 Richard Biener <rguenther@suse.de>
5994 * tree-vectorizer.h (vec_info_shared::n_stmts): Add.
5995 (LOOP_VINFO_N_STMTS): Likewise.
5996 (vec_info_for_bb): Remove unused function.
5997 * tree-vectorizer.c (vec_info_shared::vec_info_shared):
5998 Initialize n_stmts member.
5999 * tree-vect-loop.c: Remove INCLUDE_FUNCTIONAL.
6000 (vect_create_loop_vinfo): Do not set loop->aux.
6001 (vect_analyze_loop_2): Do not get n_stmts as argument,
6002 instead use LOOP_VINFO_N_STMTS. Set LOOP_VINFO_VECTORIZABLE_P
6004 (vect_analyze_loop_1): Remove callback, get the mode iterator
6005 and autodetected_vector_mode as argument, advancing the
6006 iterator and initializing autodetected_vector_mode here.
6007 (vect_analyze_loop): Split analysis loop into two, first
6008 processing main loops only and then epilogues.
6010 2021-11-05 Martin Jambor <mjambor@suse.cz>
6012 * ipa-prop.c (compute_complex_assign_jump_func): Remove
6013 unnecessary check for RECORD_TYPE.
6015 2021-11-05 Gerald Pfeifer <gerald@pfeifer.com>
6017 * doc/install.texi2html: Do not generate old.html any longer.
6019 2021-11-05 Martin Liska <mliska@suse.cz>
6022 * opts.c (finish_options): Reset flag_gtoggle when it is used.
6024 2021-11-05 Jakub Jelinek <jakub@redhat.com>
6027 * dwarf2out.c (add_const_value_attribute): Add MODE argument, use it
6028 in CONST_WIDE_INT handling. Adjust recursive calls.
6029 (add_location_or_const_value_attribute): Pass DECL_MODE (decl) to
6030 new add_const_value_attribute argument.
6031 (tree_add_const_value_attribute): Pass TYPE_MODE (type) to new
6032 add_const_value_attribute argument.
6034 2021-11-05 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
6036 * config/vx-common.h: Test value of TARGET_VXWORKS7 rather
6039 2021-11-05 Richard Biener <rguenther@suse.de>
6041 * tree-vectorizer.h (struct vect_loop_form_info): New.
6042 (vect_analyze_loop_form): Adjust.
6043 (vect_create_loop_vinfo): New.
6044 * tree-parloops.c (gather_scalar_reductions): Adjust for
6045 vect_analyze_loop_form API change.
6046 * tree-vect-loop.c: Include <functional>.
6047 (vect_analyze_loop_form_1): Rename to vect_analyze_loop_form,
6048 take struct vect_loop_form_info as output parameter and adjust.
6049 (vect_analyze_loop_form): Rename to vect_create_loop_vinfo and
6050 split out call to the original vect_analyze_loop_form_1.
6051 (vect_reanalyze_as_main_loop): Rename to...
6052 (vect_analyze_loop_1): ... this, factor out the call to
6053 vect_analyze_loop_form and generalize to be able to use it twice ...
6054 (vect_analyze_loop): ... here. Perform vect_analyze_loop_form
6057 2021-11-05 Xionghu Luo <luoxhu@linux.ibm.com>
6060 * config/rs6000/fusion.md: Regenerate.
6061 * config/rs6000/genfusion.pl: Fix incorrect clobber constraint.
6063 2021-11-04 Andreas Krebbel <krebbel@linux.ibm.com>
6065 * config/s390/s390.h (STACK_CHECK_MOVING_SP): New macro
6068 2021-11-04 Tamar Christina <tamar.christina@arm.com>
6070 * config/aarch64/aarch64-builtins.c
6071 (aarch64_general_gimple_fold_builtin): Add ashl, sshl, ushl, ashr,
6072 ashr_simd, lshr, lshr_simd.
6073 * config/aarch64/aarch64-simd-builtins.def (lshr): Use USHIFTIMM.
6074 * config/aarch64/arm_neon.h (vshr_n_u8, vshr_n_u16, vshr_n_u32,
6075 vshrq_n_u8, vshrq_n_u16, vshrq_n_u32, vshrq_n_u64): Fix type hack.
6077 2021-11-04 Tamar Christina <tamar.christina@arm.com>
6079 * match.pd: New negate+shift pattern.
6081 2021-11-04 Andrew MacLeod <amacleod@redhat.com>
6083 PR tree-optimization/103079
6084 * gimple-range-gori.cc (gimple_range_calc_op1): Treat undefined as
6086 (gimple_range_calc_op2): Ditto.
6088 2021-11-04 Martin Jambor <mjambor@suse.cz>
6091 * ipa-param-manipulation.h (class ipa_param_body_adjustments): New
6092 members remap_with_debug_expressions, m_dead_ssa_debug_equiv,
6093 m_dead_stmt_debug_equiv and prepare_debug_expressions. Added
6094 parameter to mark_dead_statements.
6095 * ipa-param-manipulation.c: Include tree-phinodes.h and cfgexpand.h.
6096 (ipa_param_body_adjustments::mark_dead_statements): New parameter
6097 debugstack, push into it all SSA names used in debug statements,
6098 produce m_dead_ssa_debug_equiv mapping for the removed param.
6099 (replace_with_mapped_expr): New function.
6100 (ipa_param_body_adjustments::remap_with_debug_expressions): Likewise.
6101 (ipa_param_body_adjustments::prepare_debug_expressions): Likewise.
6102 (ipa_param_body_adjustments::common_initialization): Gather and
6103 procecc SSA which will be removed but are in debug statements. Simplify.
6104 (ipa_param_body_adjustments::ipa_param_body_adjustments): Initialize
6106 * tree-inline.c (remap_gimple_stmt): Create a debug bind when possible
6107 when avoiding a copy of an unnecessary statement. Remap removed SSA
6108 names in existing debug statements.
6109 (tree_function_versioning): Do not create DEBUG_EXPR_DECL for removed
6110 parameters if we have already done so.
6112 2021-11-04 Jan Hubicka <hubicka@ucw.cz>
6115 * gimple.c (gimple_call_static_chain_flags): Handle case when
6116 nested function does not bind locally.
6118 2021-11-04 Jonathan Wright <jonathan.wright@arm.com>
6120 * config/aarch64/aarch64.c (aarch64_function_value): Generate
6121 a register rtx for Neon vector-tuple modes.
6122 (aarch64_layout_arg): Likewise.
6124 2021-11-04 Jonathan Wright <jonathan.wright@arm.com>
6126 * lower-subreg.c (simple_move): Prevent decomposition if
6127 modes are not tieable.
6129 2021-11-04 Jonathan Wright <jonathan.wright@arm.com>
6130 Richard Sandiford <richard.sandiford@arm.com>
6132 * config/aarch64/aarch64-builtins.c (v2x8qi_UP): Define.
6133 (v2x4hi_UP): Likewise.
6134 (v2x4hf_UP): Likewise.
6135 (v2x4bf_UP): Likewise.
6136 (v2x2si_UP): Likewise.
6137 (v2x2sf_UP): Likewise.
6138 (v2x1di_UP): Likewise.
6139 (v2x1df_UP): Likewise.
6140 (v2x16qi_UP): Likewise.
6141 (v2x8hi_UP): Likewise.
6142 (v2x8hf_UP): Likewise.
6143 (v2x8bf_UP): Likewise.
6144 (v2x4si_UP): Likewise.
6145 (v2x4sf_UP): Likewise.
6146 (v2x2di_UP): Likewise.
6147 (v2x2df_UP): Likewise.
6148 (v3x8qi_UP): Likewise.
6149 (v3x4hi_UP): Likewise.
6150 (v3x4hf_UP): Likewise.
6151 (v3x4bf_UP): Likewise.
6152 (v3x2si_UP): Likewise.
6153 (v3x2sf_UP): Likewise.
6154 (v3x1di_UP): Likewise.
6155 (v3x1df_UP): Likewise.
6156 (v3x16qi_UP): Likewise.
6157 (v3x8hi_UP): Likewise.
6158 (v3x8hf_UP): Likewise.
6159 (v3x8bf_UP): Likewise.
6160 (v3x4si_UP): Likewise.
6161 (v3x4sf_UP): Likewise.
6162 (v3x2di_UP): Likewise.
6163 (v3x2df_UP): Likewise.
6164 (v4x8qi_UP): Likewise.
6165 (v4x4hi_UP): Likewise.
6166 (v4x4hf_UP): Likewise.
6167 (v4x4bf_UP): Likewise.
6168 (v4x2si_UP): Likewise.
6169 (v4x2sf_UP): Likewise.
6170 (v4x1di_UP): Likewise.
6171 (v4x1df_UP): Likewise.
6172 (v4x16qi_UP): Likewise.
6173 (v4x8hi_UP): Likewise.
6174 (v4x8hf_UP): Likewise.
6175 (v4x8bf_UP): Likewise.
6176 (v4x4si_UP): Likewise.
6177 (v4x4sf_UP): Likewise.
6178 (v4x2di_UP): Likewise.
6179 (v4x2df_UP): Likewise.
6180 (TYPES_GETREGP): Delete.
6181 (TYPES_SETREGP): Likewise.
6182 (TYPES_LOADSTRUCT_U): Define.
6183 (TYPES_LOADSTRUCT_P): Likewise.
6184 (TYPES_LOADSTRUCT_LANE_U): Likewise.
6185 (TYPES_LOADSTRUCT_LANE_P): Likewise.
6186 (TYPES_STORE1P): Move for consistency.
6187 (TYPES_STORESTRUCT_U): Define.
6188 (TYPES_STORESTRUCT_P): Likewise.
6189 (TYPES_STORESTRUCT_LANE_U): Likewise.
6190 (TYPES_STORESTRUCT_LANE_P): Likewise.
6191 (aarch64_simd_tuple_types): Define.
6192 (aarch64_lookup_simd_builtin_type): Handle tuple type lookup.
6193 (aarch64_init_simd_builtin_functions): Update frontend lookup
6194 for builtin functions after handling arm_neon.h pragma.
6195 (register_tuple_type): Manually set modes of single-integer
6196 tuple types. Record tuple types.
6197 * config/aarch64/aarch64-modes.def
6198 (ADV_SIMD_D_REG_STRUCT_MODES): Define D-register tuple modes.
6199 (ADV_SIMD_Q_REG_STRUCT_MODES): Define Q-register tuple modes.
6200 (SVE_MODES): Give single-vector modes priority over vector-
6202 (VECTOR_MODES_WITH_PREFIX): Set partial-vector mode order to
6203 be after all single-vector modes.
6204 * config/aarch64/aarch64-simd-builtins.def: Update builtin
6205 generator macros to reflect modifications to the backend
6207 * config/aarch64/aarch64-simd.md (aarch64_simd_ld2<mode>):
6208 Use vector-tuple mode iterator and rename to...
6209 (aarch64_simd_ld2<vstruct_elt>): This.
6210 (aarch64_simd_ld2r<mode>): Use vector-tuple mode iterator and
6212 (aarch64_simd_ld2r<vstruct_elt>): This.
6213 (aarch64_vec_load_lanesoi_lane<mode>): Use vector-tuple mode
6214 iterator and rename to...
6215 (aarch64_vec_load_lanes<mode>_lane<vstruct_elt>): This.
6216 (vec_load_lanesoi<mode>): Use vector-tuple mode iterator and
6218 (vec_load_lanes<mode><vstruct_elt>): This.
6219 (aarch64_simd_st2<mode>): Use vector-tuple mode iterator and
6221 (aarch64_simd_st2<vstruct_elt>): This.
6222 (aarch64_vec_store_lanesoi_lane<mode>): Use vector-tuple mode
6223 iterator and rename to...
6224 (aarch64_vec_store_lanes<mode>_lane<vstruct_elt>): This.
6225 (vec_store_lanesoi<mode>): Use vector-tuple mode iterator and
6227 (vec_store_lanes<mode><vstruct_elt>): This.
6228 (aarch64_simd_ld3<mode>): Use vector-tuple mode iterator and
6230 (aarch64_simd_ld3<vstruct_elt>): This.
6231 (aarch64_simd_ld3r<mode>): Use vector-tuple mode iterator and
6233 (aarch64_simd_ld3r<vstruct_elt>): This.
6234 (aarch64_vec_load_lanesci_lane<mode>): Use vector-tuple mode
6235 iterator and rename to...
6236 (vec_load_lanesci<mode>): This.
6237 (aarch64_simd_st3<mode>): Use vector-tuple mode iterator and
6239 (aarch64_simd_st3<vstruct_elt>): This.
6240 (aarch64_vec_store_lanesci_lane<mode>): Use vector-tuple mode
6241 iterator and rename to...
6242 (vec_store_lanesci<mode>): This.
6243 (aarch64_simd_ld4<mode>): Use vector-tuple mode iterator and
6245 (aarch64_simd_ld4<vstruct_elt>): This.
6246 (aarch64_simd_ld4r<mode>): Use vector-tuple mode iterator and
6248 (aarch64_simd_ld4r<vstruct_elt>): This.
6249 (aarch64_vec_load_lanesxi_lane<mode>): Use vector-tuple mode
6250 iterator and rename to...
6251 (vec_load_lanesxi<mode>): This.
6252 (aarch64_simd_st4<mode>): Use vector-tuple mode iterator and
6254 (aarch64_simd_st4<vstruct_elt>): This.
6255 (aarch64_vec_store_lanesxi_lane<mode>): Use vector-tuple mode
6256 iterator and rename to...
6257 (vec_store_lanesxi<mode>): This.
6258 (mov<mode>): Define for Neon vector-tuple modes.
6259 (aarch64_ld1x3<VALLDIF:mode>): Use vector-tuple mode iterator
6261 (aarch64_ld1x3<vstruct_elt>): This.
6262 (aarch64_ld1_x3_<mode>): Use vector-tuple mode iterator and
6264 (aarch64_ld1_x3_<vstruct_elt>): This.
6265 (aarch64_ld1x4<VALLDIF:mode>): Use vector-tuple mode iterator
6267 (aarch64_ld1x4<vstruct_elt>): This.
6268 (aarch64_ld1_x4_<mode>): Use vector-tuple mode iterator and
6270 (aarch64_ld1_x4_<vstruct_elt>): This.
6271 (aarch64_st1x2<VALLDIF:mode>): Use vector-tuple mode iterator
6273 (aarch64_st1x2<vstruct_elt>): This.
6274 (aarch64_st1_x2_<mode>): Use vector-tuple mode iterator and
6276 (aarch64_st1_x2_<vstruct_elt>): This.
6277 (aarch64_st1x3<VALLDIF:mode>): Use vector-tuple mode iterator
6279 (aarch64_st1x3<vstruct_elt>): This.
6280 (aarch64_st1_x3_<mode>): Use vector-tuple mode iterator and
6282 (aarch64_st1_x3_<vstruct_elt>): This.
6283 (aarch64_st1x4<VALLDIF:mode>): Use vector-tuple mode iterator
6285 (aarch64_st1x4<vstruct_elt>): This.
6286 (aarch64_st1_x4_<mode>): Use vector-tuple mode iterator and
6288 (aarch64_st1_x4_<vstruct_elt>): This.
6289 (*aarch64_mov<mode>): Define for vector-tuple modes.
6290 (*aarch64_be_mov<mode>): Likewise.
6291 (aarch64_ld<VSTRUCT:nregs>r<VALLDIF:mode>): Use vector-tuple
6292 mode iterator and rename to...
6293 (aarch64_ld<nregs>r<vstruct_elt>): This.
6294 (aarch64_ld2<mode>_dreg): Use vector-tuple mode iterator and
6296 (aarch64_ld2<vstruct_elt>_dreg): This.
6297 (aarch64_ld3<mode>_dreg): Use vector-tuple mode iterator and
6299 (aarch64_ld3<vstruct_elt>_dreg): This.
6300 (aarch64_ld4<mode>_dreg): Use vector-tuple mode iterator and
6302 (aarch64_ld4<vstruct_elt>_dreg): This.
6303 (aarch64_ld<VSTRUCT:nregs><VDC:mode>): Use vector-tuple mode
6304 iterator and rename to...
6305 (aarch64_ld<nregs><vstruct_elt>): Use vector-tuple mode
6306 iterator and rename to...
6307 (aarch64_ld<VSTRUCT:nregs><VQ:mode>): Use vector-tuple mode
6308 (aarch64_ld1x2<VQ:mode>): Delete.
6309 (aarch64_ld1x2<VDC:mode>): Use vector-tuple mode iterator and
6311 (aarch64_ld1x2<vstruct_elt>): This.
6312 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use vector-
6313 tuple mode iterator and rename to...
6314 (aarch64_ld<nregs>_lane<vstruct_elt>): This.
6315 (aarch64_get_dreg<VSTRUCT:mode><VDC:mode>): Delete.
6316 (aarch64_get_qreg<VSTRUCT:mode><VQ:mode>): Likewise.
6317 (aarch64_st2<mode>_dreg): Use vector-tuple mode iterator and
6319 (aarch64_st2<vstruct_elt>_dreg): This.
6320 (aarch64_st3<mode>_dreg): Use vector-tuple mode iterator and
6322 (aarch64_st3<vstruct_elt>_dreg): This.
6323 (aarch64_st4<mode>_dreg): Use vector-tuple mode iterator and
6325 (aarch64_st4<vstruct_elt>_dreg): This.
6326 (aarch64_st<VSTRUCT:nregs><VDC:mode>): Use vector-tuple mode
6327 iterator and rename to...
6328 (aarch64_st<nregs><vstruct_elt>): This.
6329 (aarch64_st<VSTRUCT:nregs><VQ:mode>): Use vector-tuple mode
6330 iterator and rename to aarch64_st<nregs><vstruct_elt>.
6331 (aarch64_st<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use vector-
6332 tuple mode iterator and rename to...
6333 (aarch64_st<nregs>_lane<vstruct_elt>): This.
6334 (aarch64_set_qreg<VSTRUCT:mode><VQ:mode>): Delete.
6335 (aarch64_simd_ld1<mode>_x2): Use vector-tuple mode iterator
6337 (aarch64_simd_ld1<vstruct_elt>_x2): This.
6338 * config/aarch64/aarch64.c (aarch64_advsimd_struct_mode_p):
6339 Refactor to include new vector-tuple modes.
6340 (aarch64_classify_vector_mode): Add cases for new vector-
6342 (aarch64_advsimd_partial_struct_mode_p): Define.
6343 (aarch64_advsimd_full_struct_mode_p): Likewise.
6344 (aarch64_advsimd_vector_array_mode): Likewise.
6345 (aarch64_sve_data_mode): Change location in file.
6346 (aarch64_array_mode): Handle case of Neon vector-tuple modes.
6347 (aarch64_hard_regno_nregs): Handle case of partial Neon
6349 (aarch64_classify_address): Refactor to include handling of
6350 Neon vector-tuple modes.
6351 (aarch64_print_operand): Print "d" for "%R" for a partial
6352 Neon vector structure.
6353 (aarch64_expand_vec_perm_1): Use new vector-tuple mode.
6354 (aarch64_modes_tieable_p): Prevent tieing Neon partial struct
6355 modes with scalar machines modes larger than 8 bytes.
6356 (aarch64_can_change_mode_class): Don't allow changes between
6357 partial and full Neon vector-structure modes.
6358 * config/aarch64/arm_neon.h (vst2_lane_f16): Use updated
6359 builtin and remove boiler-plate code for opaque mode.
6360 (vst2_lane_f32): Likewise.
6361 (vst2_lane_f64): Likewise.
6362 (vst2_lane_p8): Likewise.
6363 (vst2_lane_p16): Likewise.
6364 (vst2_lane_p64): Likewise.
6365 (vst2_lane_s8): Likewise.
6366 (vst2_lane_s16): Likewise.
6367 (vst2_lane_s32): Likewise.
6368 (vst2_lane_s64): Likewise.
6369 (vst2_lane_u8): Likewise.
6370 (vst2_lane_u16): Likewise.
6371 (vst2_lane_u32): Likewise.
6372 (vst2_lane_u64): Likewise.
6373 (vst2q_lane_f16): Likewise.
6374 (vst2q_lane_f32): Likewise.
6375 (vst2q_lane_f64): Likewise.
6376 (vst2q_lane_p8): Likewise.
6377 (vst2q_lane_p16): Likewise.
6378 (vst2q_lane_p64): Likewise.
6379 (vst2q_lane_s8): Likewise.
6380 (vst2q_lane_s16): Likewise.
6381 (vst2q_lane_s32): Likewise.
6382 (vst2q_lane_s64): Likewise.
6383 (vst2q_lane_u8): Likewise.
6384 (vst2q_lane_u16): Likewise.
6385 (vst2q_lane_u32): Likewise.
6386 (vst2q_lane_u64): Likewise.
6387 (vst3_lane_f16): Likewise.
6388 (vst3_lane_f32): Likewise.
6389 (vst3_lane_f64): Likewise.
6390 (vst3_lane_p8): Likewise.
6391 (vst3_lane_p16): Likewise.
6392 (vst3_lane_p64): Likewise.
6393 (vst3_lane_s8): Likewise.
6394 (vst3_lane_s16): Likewise.
6395 (vst3_lane_s32): Likewise.
6396 (vst3_lane_s64): Likewise.
6397 (vst3_lane_u8): Likewise.
6398 (vst3_lane_u16): Likewise.
6399 (vst3_lane_u32): Likewise.
6400 (vst3_lane_u64): Likewise.
6401 (vst3q_lane_f16): Likewise.
6402 (vst3q_lane_f32): Likewise.
6403 (vst3q_lane_f64): Likewise.
6404 (vst3q_lane_p8): Likewise.
6405 (vst3q_lane_p16): Likewise.
6406 (vst3q_lane_p64): Likewise.
6407 (vst3q_lane_s8): Likewise.
6408 (vst3q_lane_s16): Likewise.
6409 (vst3q_lane_s32): Likewise.
6410 (vst3q_lane_s64): Likewise.
6411 (vst3q_lane_u8): Likewise.
6412 (vst3q_lane_u16): Likewise.
6413 (vst3q_lane_u32): Likewise.
6414 (vst3q_lane_u64): Likewise.
6415 (vst4_lane_f16): Likewise.
6416 (vst4_lane_f32): Likewise.
6417 (vst4_lane_f64): Likewise.
6418 (vst4_lane_p8): Likewise.
6419 (vst4_lane_p16): Likewise.
6420 (vst4_lane_p64): Likewise.
6421 (vst4_lane_s8): Likewise.
6422 (vst4_lane_s16): Likewise.
6423 (vst4_lane_s32): Likewise.
6424 (vst4_lane_s64): Likewise.
6425 (vst4_lane_u8): Likewise.
6426 (vst4_lane_u16): Likewise.
6427 (vst4_lane_u32): Likewise.
6428 (vst4_lane_u64): Likewise.
6429 (vst4q_lane_f16): Likewise.
6430 (vst4q_lane_f32): Likewise.
6431 (vst4q_lane_f64): Likewise.
6432 (vst4q_lane_p8): Likewise.
6433 (vst4q_lane_p16): Likewise.
6434 (vst4q_lane_p64): Likewise.
6435 (vst4q_lane_s8): Likewise.
6436 (vst4q_lane_s16): Likewise.
6437 (vst4q_lane_s32): Likewise.
6438 (vst4q_lane_s64): Likewise.
6439 (vst4q_lane_u8): Likewise.
6440 (vst4q_lane_u16): Likewise.
6441 (vst4q_lane_u32): Likewise.
6442 (vst4q_lane_u64): Likewise.
6443 (vtbl3_s8): Likewise.
6444 (vtbl3_u8): Likewise.
6445 (vtbl3_p8): Likewise.
6446 (vtbl4_s8): Likewise.
6447 (vtbl4_u8): Likewise.
6448 (vtbl4_p8): Likewise.
6449 (vld1_u8_x3): Likewise.
6450 (vld1_s8_x3): Likewise.
6451 (vld1_u16_x3): Likewise.
6452 (vld1_s16_x3): Likewise.
6453 (vld1_u32_x3): Likewise.
6454 (vld1_s32_x3): Likewise.
6455 (vld1_u64_x3): Likewise.
6456 (vld1_s64_x3): Likewise.
6457 (vld1_f16_x3): Likewise.
6458 (vld1_f32_x3): Likewise.
6459 (vld1_f64_x3): Likewise.
6460 (vld1_p8_x3): Likewise.
6461 (vld1_p16_x3): Likewise.
6462 (vld1_p64_x3): Likewise.
6463 (vld1q_u8_x3): Likewise.
6464 (vld1q_s8_x3): Likewise.
6465 (vld1q_u16_x3): Likewise.
6466 (vld1q_s16_x3): Likewise.
6467 (vld1q_u32_x3): Likewise.
6468 (vld1q_s32_x3): Likewise.
6469 (vld1q_u64_x3): Likewise.
6470 (vld1q_s64_x3): Likewise.
6471 (vld1q_f16_x3): Likewise.
6472 (vld1q_f32_x3): Likewise.
6473 (vld1q_f64_x3): Likewise.
6474 (vld1q_p8_x3): Likewise.
6475 (vld1q_p16_x3): Likewise.
6476 (vld1q_p64_x3): Likewise.
6477 (vld1_u8_x2): Likewise.
6478 (vld1_s8_x2): Likewise.
6479 (vld1_u16_x2): Likewise.
6480 (vld1_s16_x2): Likewise.
6481 (vld1_u32_x2): Likewise.
6482 (vld1_s32_x2): Likewise.
6483 (vld1_u64_x2): Likewise.
6484 (vld1_s64_x2): Likewise.
6485 (vld1_f16_x2): Likewise.
6486 (vld1_f32_x2): Likewise.
6487 (vld1_f64_x2): Likewise.
6488 (vld1_p8_x2): Likewise.
6489 (vld1_p16_x2): Likewise.
6490 (vld1_p64_x2): Likewise.
6491 (vld1q_u8_x2): Likewise.
6492 (vld1q_s8_x2): Likewise.
6493 (vld1q_u16_x2): Likewise.
6494 (vld1q_s16_x2): Likewise.
6495 (vld1q_u32_x2): Likewise.
6496 (vld1q_s32_x2): Likewise.
6497 (vld1q_u64_x2): Likewise.
6498 (vld1q_s64_x2): Likewise.
6499 (vld1q_f16_x2): Likewise.
6500 (vld1q_f32_x2): Likewise.
6501 (vld1q_f64_x2): Likewise.
6502 (vld1q_p8_x2): Likewise.
6503 (vld1q_p16_x2): Likewise.
6504 (vld1q_p64_x2): Likewise.
6505 (vld1_s8_x4): Likewise.
6506 (vld1q_s8_x4): Likewise.
6507 (vld1_s16_x4): Likewise.
6508 (vld1q_s16_x4): Likewise.
6509 (vld1_s32_x4): Likewise.
6510 (vld1q_s32_x4): Likewise.
6511 (vld1_u8_x4): Likewise.
6512 (vld1q_u8_x4): Likewise.
6513 (vld1_u16_x4): Likewise.
6514 (vld1q_u16_x4): Likewise.
6515 (vld1_u32_x4): Likewise.
6516 (vld1q_u32_x4): Likewise.
6517 (vld1_f16_x4): Likewise.
6518 (vld1q_f16_x4): Likewise.
6519 (vld1_f32_x4): Likewise.
6520 (vld1q_f32_x4): Likewise.
6521 (vld1_p8_x4): Likewise.
6522 (vld1q_p8_x4): Likewise.
6523 (vld1_p16_x4): Likewise.
6524 (vld1q_p16_x4): Likewise.
6525 (vld1_s64_x4): Likewise.
6526 (vld1_u64_x4): Likewise.
6527 (vld1_p64_x4): Likewise.
6528 (vld1q_s64_x4): Likewise.
6529 (vld1q_u64_x4): Likewise.
6530 (vld1q_p64_x4): Likewise.
6531 (vld1_f64_x4): Likewise.
6532 (vld1q_f64_x4): Likewise.
6533 (vld2_s64): Likewise.
6534 (vld2_u64): Likewise.
6535 (vld2_f64): Likewise.
6536 (vld2_s8): Likewise.
6537 (vld2_p8): Likewise.
6538 (vld2_p64): Likewise.
6539 (vld2_s16): Likewise.
6540 (vld2_p16): Likewise.
6541 (vld2_s32): Likewise.
6542 (vld2_u8): Likewise.
6543 (vld2_u16): Likewise.
6544 (vld2_u32): Likewise.
6545 (vld2_f16): Likewise.
6546 (vld2_f32): Likewise.
6547 (vld2q_s8): Likewise.
6548 (vld2q_p8): Likewise.
6549 (vld2q_s16): Likewise.
6550 (vld2q_p16): Likewise.
6551 (vld2q_p64): Likewise.
6552 (vld2q_s32): Likewise.
6553 (vld2q_s64): Likewise.
6554 (vld2q_u8): Likewise.
6555 (vld2q_u16): Likewise.
6556 (vld2q_u32): Likewise.
6557 (vld2q_u64): Likewise.
6558 (vld2q_f16): Likewise.
6559 (vld2q_f32): Likewise.
6560 (vld2q_f64): Likewise.
6561 (vld3_s64): Likewise.
6562 (vld3_u64): Likewise.
6563 (vld3_f64): Likewise.
6564 (vld3_s8): Likewise.
6565 (vld3_p8): Likewise.
6566 (vld3_s16): Likewise.
6567 (vld3_p16): Likewise.
6568 (vld3_s32): Likewise.
6569 (vld3_u8): Likewise.
6570 (vld3_u16): Likewise.
6571 (vld3_u32): Likewise.
6572 (vld3_f16): Likewise.
6573 (vld3_f32): Likewise.
6574 (vld3_p64): Likewise.
6575 (vld3q_s8): Likewise.
6576 (vld3q_p8): Likewise.
6577 (vld3q_s16): Likewise.
6578 (vld3q_p16): Likewise.
6579 (vld3q_s32): Likewise.
6580 (vld3q_s64): Likewise.
6581 (vld3q_u8): Likewise.
6582 (vld3q_u16): Likewise.
6583 (vld3q_u32): Likewise.
6584 (vld3q_u64): Likewise.
6585 (vld3q_f16): Likewise.
6586 (vld3q_f32): Likewise.
6587 (vld3q_f64): Likewise.
6588 (vld3q_p64): Likewise.
6589 (vld4_s64): Likewise.
6590 (vld4_u64): Likewise.
6591 (vld4_f64): Likewise.
6592 (vld4_s8): Likewise.
6593 (vld4_p8): Likewise.
6594 (vld4_s16): Likewise.
6595 (vld4_p16): Likewise.
6596 (vld4_s32): Likewise.
6597 (vld4_u8): Likewise.
6598 (vld4_u16): Likewise.
6599 (vld4_u32): Likewise.
6600 (vld4_f16): Likewise.
6601 (vld4_f32): Likewise.
6602 (vld4_p64): Likewise.
6603 (vld4q_s8): Likewise.
6604 (vld4q_p8): Likewise.
6605 (vld4q_s16): Likewise.
6606 (vld4q_p16): Likewise.
6607 (vld4q_s32): Likewise.
6608 (vld4q_s64): Likewise.
6609 (vld4q_u8): Likewise.
6610 (vld4q_u16): Likewise.
6611 (vld4q_u32): Likewise.
6612 (vld4q_u64): Likewise.
6613 (vld4q_f16): Likewise.
6614 (vld4q_f32): Likewise.
6615 (vld4q_f64): Likewise.
6616 (vld4q_p64): Likewise.
6617 (vld2_dup_s8): Likewise.
6618 (vld2_dup_s16): Likewise.
6619 (vld2_dup_s32): Likewise.
6620 (vld2_dup_f16): Likewise.
6621 (vld2_dup_f32): Likewise.
6622 (vld2_dup_f64): Likewise.
6623 (vld2_dup_u8): Likewise.
6624 (vld2_dup_u16): Likewise.
6625 (vld2_dup_u32): Likewise.
6626 (vld2_dup_p8): Likewise.
6627 (vld2_dup_p16): Likewise.
6628 (vld2_dup_p64): Likewise.
6629 (vld2_dup_s64): Likewise.
6630 (vld2_dup_u64): Likewise.
6631 (vld2q_dup_s8): Likewise.
6632 (vld2q_dup_p8): Likewise.
6633 (vld2q_dup_s16): Likewise.
6634 (vld2q_dup_p16): Likewise.
6635 (vld2q_dup_s32): Likewise.
6636 (vld2q_dup_s64): Likewise.
6637 (vld2q_dup_u8): Likewise.
6638 (vld2q_dup_u16): Likewise.
6639 (vld2q_dup_u32): Likewise.
6640 (vld2q_dup_u64): Likewise.
6641 (vld2q_dup_f16): Likewise.
6642 (vld2q_dup_f32): Likewise.
6643 (vld2q_dup_f64): Likewise.
6644 (vld2q_dup_p64): Likewise.
6645 (vld3_dup_s64): Likewise.
6646 (vld3_dup_u64): Likewise.
6647 (vld3_dup_f64): Likewise.
6648 (vld3_dup_s8): Likewise.
6649 (vld3_dup_p8): Likewise.
6650 (vld3_dup_s16): Likewise.
6651 (vld3_dup_p16): Likewise.
6652 (vld3_dup_s32): Likewise.
6653 (vld3_dup_u8): Likewise.
6654 (vld3_dup_u16): Likewise.
6655 (vld3_dup_u32): Likewise.
6656 (vld3_dup_f16): Likewise.
6657 (vld3_dup_f32): Likewise.
6658 (vld3_dup_p64): Likewise.
6659 (vld3q_dup_s8): Likewise.
6660 (vld3q_dup_p8): Likewise.
6661 (vld3q_dup_s16): Likewise.
6662 (vld3q_dup_p16): Likewise.
6663 (vld3q_dup_s32): Likewise.
6664 (vld3q_dup_s64): Likewise.
6665 (vld3q_dup_u8): Likewise.
6666 (vld3q_dup_u16): Likewise.
6667 (vld3q_dup_u32): Likewise.
6668 (vld3q_dup_u64): Likewise.
6669 (vld3q_dup_f16): Likewise.
6670 (vld3q_dup_f32): Likewise.
6671 (vld3q_dup_f64): Likewise.
6672 (vld3q_dup_p64): Likewise.
6673 (vld4_dup_s64): Likewise.
6674 (vld4_dup_u64): Likewise.
6675 (vld4_dup_f64): Likewise.
6676 (vld4_dup_s8): Likewise.
6677 (vld4_dup_p8): Likewise.
6678 (vld4_dup_s16): Likewise.
6679 (vld4_dup_p16): Likewise.
6680 (vld4_dup_s32): Likewise.
6681 (vld4_dup_u8): Likewise.
6682 (vld4_dup_u16): Likewise.
6683 (vld4_dup_u32): Likewise.
6684 (vld4_dup_f16): Likewise.
6685 (vld4_dup_f32): Likewise.
6686 (vld4_dup_p64): Likewise.
6687 (vld4q_dup_s8): Likewise.
6688 (vld4q_dup_p8): Likewise.
6689 (vld4q_dup_s16): Likewise.
6690 (vld4q_dup_p16): Likewise.
6691 (vld4q_dup_s32): Likewise.
6692 (vld4q_dup_s64): Likewise.
6693 (vld4q_dup_u8): Likewise.
6694 (vld4q_dup_u16): Likewise.
6695 (vld4q_dup_u32): Likewise.
6696 (vld4q_dup_u64): Likewise.
6697 (vld4q_dup_f16): Likewise.
6698 (vld4q_dup_f32): Likewise.
6699 (vld4q_dup_f64): Likewise.
6700 (vld4q_dup_p64): Likewise.
6701 (vld2_lane_u8): Likewise.
6702 (vld2_lane_u16): Likewise.
6703 (vld2_lane_u32): Likewise.
6704 (vld2_lane_u64): Likewise.
6705 (vld2_lane_s8): Likewise.
6706 (vld2_lane_s16): Likewise.
6707 (vld2_lane_s32): Likewise.
6708 (vld2_lane_s64): Likewise.
6709 (vld2_lane_f16): Likewise.
6710 (vld2_lane_f32): Likewise.
6711 (vld2_lane_f64): Likewise.
6712 (vld2_lane_p8): Likewise.
6713 (vld2_lane_p16): Likewise.
6714 (vld2_lane_p64): Likewise.
6715 (vld2q_lane_u8): Likewise.
6716 (vld2q_lane_u16): Likewise.
6717 (vld2q_lane_u32): Likewise.
6718 (vld2q_lane_u64): Likewise.
6719 (vld2q_lane_s8): Likewise.
6720 (vld2q_lane_s16): Likewise.
6721 (vld2q_lane_s32): Likewise.
6722 (vld2q_lane_s64): Likewise.
6723 (vld2q_lane_f16): Likewise.
6724 (vld2q_lane_f32): Likewise.
6725 (vld2q_lane_f64): Likewise.
6726 (vld2q_lane_p8): Likewise.
6727 (vld2q_lane_p16): Likewise.
6728 (vld2q_lane_p64): Likewise.
6729 (vld3_lane_u8): Likewise.
6730 (vld3_lane_u16): Likewise.
6731 (vld3_lane_u32): Likewise.
6732 (vld3_lane_u64): Likewise.
6733 (vld3_lane_s8): Likewise.
6734 (vld3_lane_s16): Likewise.
6735 (vld3_lane_s32): Likewise.
6736 (vld3_lane_s64): Likewise.
6737 (vld3_lane_f16): Likewise.
6738 (vld3_lane_f32): Likewise.
6739 (vld3_lane_f64): Likewise.
6740 (vld3_lane_p8): Likewise.
6741 (vld3_lane_p16): Likewise.
6742 (vld3_lane_p64): Likewise.
6743 (vld3q_lane_u8): Likewise.
6744 (vld3q_lane_u16): Likewise.
6745 (vld3q_lane_u32): Likewise.
6746 (vld3q_lane_u64): Likewise.
6747 (vld3q_lane_s8): Likewise.
6748 (vld3q_lane_s16): Likewise.
6749 (vld3q_lane_s32): Likewise.
6750 (vld3q_lane_s64): Likewise.
6751 (vld3q_lane_f16): Likewise.
6752 (vld3q_lane_f32): Likewise.
6753 (vld3q_lane_f64): Likewise.
6754 (vld3q_lane_p8): Likewise.
6755 (vld3q_lane_p16): Likewise.
6756 (vld3q_lane_p64): Likewise.
6757 (vld4_lane_u8): Likewise.
6758 (vld4_lane_u16): Likewise.
6759 (vld4_lane_u32): Likewise.
6760 (vld4_lane_u64): Likewise.
6761 (vld4_lane_s8): Likewise.
6762 (vld4_lane_s16): Likewise.
6763 (vld4_lane_s32): Likewise.
6764 (vld4_lane_s64): Likewise.
6765 (vld4_lane_f16): Likewise.
6766 (vld4_lane_f32): Likewise.
6767 (vld4_lane_f64): Likewise.
6768 (vld4_lane_p8): Likewise.
6769 (vld4_lane_p16): Likewise.
6770 (vld4_lane_p64): Likewise.
6771 (vld4q_lane_u8): Likewise.
6772 (vld4q_lane_u16): Likewise.
6773 (vld4q_lane_u32): Likewise.
6774 (vld4q_lane_u64): Likewise.
6775 (vld4q_lane_s8): Likewise.
6776 (vld4q_lane_s16): Likewise.
6777 (vld4q_lane_s32): Likewise.
6778 (vld4q_lane_s64): Likewise.
6779 (vld4q_lane_f16): Likewise.
6780 (vld4q_lane_f32): Likewise.
6781 (vld4q_lane_f64): Likewise.
6782 (vld4q_lane_p8): Likewise.
6783 (vld4q_lane_p16): Likewise.
6784 (vld4q_lane_p64): Likewise.
6785 (vqtbl2_s8): Likewise.
6786 (vqtbl2_u8): Likewise.
6787 (vqtbl2_p8): Likewise.
6788 (vqtbl2q_s8): Likewise.
6789 (vqtbl2q_u8): Likewise.
6790 (vqtbl2q_p8): Likewise.
6791 (vqtbl3_s8): Likewise.
6792 (vqtbl3_u8): Likewise.
6793 (vqtbl3_p8): Likewise.
6794 (vqtbl3q_s8): Likewise.
6795 (vqtbl3q_u8): Likewise.
6796 (vqtbl3q_p8): Likewise.
6797 (vqtbl4_s8): Likewise.
6798 (vqtbl4_u8): Likewise.
6799 (vqtbl4_p8): Likewise.
6800 (vqtbl4q_s8): Likewise.
6801 (vqtbl4q_u8): Likewise.
6802 (vqtbl4q_p8): Likewise.
6803 (vqtbx2_s8): Likewise.
6804 (vqtbx2_u8): Likewise.
6805 (vqtbx2_p8): Likewise.
6806 (vqtbx2q_s8): Likewise.
6807 (vqtbx2q_u8): Likewise.
6808 (vqtbx2q_p8): Likewise.
6809 (vqtbx3_s8): Likewise.
6810 (vqtbx3_u8): Likewise.
6811 (vqtbx3_p8): Likewise.
6812 (vqtbx3q_s8): Likewise.
6813 (vqtbx3q_u8): Likewise.
6814 (vqtbx3q_p8): Likewise.
6815 (vqtbx4_s8): Likewise.
6816 (vqtbx4_u8): Likewise.
6817 (vqtbx4_p8): Likewise.
6818 (vqtbx4q_s8): Likewise.
6819 (vqtbx4q_u8): Likewise.
6820 (vqtbx4q_p8): Likewise.
6821 (vst1_s64_x2): Likewise.
6822 (vst1_u64_x2): Likewise.
6823 (vst1_f64_x2): Likewise.
6824 (vst1_s8_x2): Likewise.
6825 (vst1_p8_x2): Likewise.
6826 (vst1_s16_x2): Likewise.
6827 (vst1_p16_x2): Likewise.
6828 (vst1_s32_x2): Likewise.
6829 (vst1_u8_x2): Likewise.
6830 (vst1_u16_x2): Likewise.
6831 (vst1_u32_x2): Likewise.
6832 (vst1_f16_x2): Likewise.
6833 (vst1_f32_x2): Likewise.
6834 (vst1_p64_x2): Likewise.
6835 (vst1q_s8_x2): Likewise.
6836 (vst1q_p8_x2): Likewise.
6837 (vst1q_s16_x2): Likewise.
6838 (vst1q_p16_x2): Likewise.
6839 (vst1q_s32_x2): Likewise.
6840 (vst1q_s64_x2): Likewise.
6841 (vst1q_u8_x2): Likewise.
6842 (vst1q_u16_x2): Likewise.
6843 (vst1q_u32_x2): Likewise.
6844 (vst1q_u64_x2): Likewise.
6845 (vst1q_f16_x2): Likewise.
6846 (vst1q_f32_x2): Likewise.
6847 (vst1q_f64_x2): Likewise.
6848 (vst1q_p64_x2): Likewise.
6849 (vst1_s64_x3): Likewise.
6850 (vst1_u64_x3): Likewise.
6851 (vst1_f64_x3): Likewise.
6852 (vst1_s8_x3): Likewise.
6853 (vst1_p8_x3): Likewise.
6854 (vst1_s16_x3): Likewise.
6855 (vst1_p16_x3): Likewise.
6856 (vst1_s32_x3): Likewise.
6857 (vst1_u8_x3): Likewise.
6858 (vst1_u16_x3): Likewise.
6859 (vst1_u32_x3): Likewise.
6860 (vst1_f16_x3): Likewise.
6861 (vst1_f32_x3): Likewise.
6862 (vst1_p64_x3): Likewise.
6863 (vst1q_s8_x3): Likewise.
6864 (vst1q_p8_x3): Likewise.
6865 (vst1q_s16_x3): Likewise.
6866 (vst1q_p16_x3): Likewise.
6867 (vst1q_s32_x3): Likewise.
6868 (vst1q_s64_x3): Likewise.
6869 (vst1q_u8_x3): Likewise.
6870 (vst1q_u16_x3): Likewise.
6871 (vst1q_u32_x3): Likewise.
6872 (vst1q_u64_x3): Likewise.
6873 (vst1q_f16_x3): Likewise.
6874 (vst1q_f32_x3): Likewise.
6875 (vst1q_f64_x3): Likewise.
6876 (vst1q_p64_x3): Likewise.
6877 (vst1_s8_x4): Likewise.
6878 (vst1q_s8_x4): Likewise.
6879 (vst1_s16_x4): Likewise.
6880 (vst1q_s16_x4): Likewise.
6881 (vst1_s32_x4): Likewise.
6882 (vst1q_s32_x4): Likewise.
6883 (vst1_u8_x4): Likewise.
6884 (vst1q_u8_x4): Likewise.
6885 (vst1_u16_x4): Likewise.
6886 (vst1q_u16_x4): Likewise.
6887 (vst1_u32_x4): Likewise.
6888 (vst1q_u32_x4): Likewise.
6889 (vst1_f16_x4): Likewise.
6890 (vst1q_f16_x4): Likewise.
6891 (vst1_f32_x4): Likewise.
6892 (vst1q_f32_x4): Likewise.
6893 (vst1_p8_x4): Likewise.
6894 (vst1q_p8_x4): Likewise.
6895 (vst1_p16_x4): Likewise.
6896 (vst1q_p16_x4): Likewise.
6897 (vst1_s64_x4): Likewise.
6898 (vst1_u64_x4): Likewise.
6899 (vst1_p64_x4): Likewise.
6900 (vst1q_s64_x4): Likewise.
6901 (vst1q_u64_x4): Likewise.
6902 (vst1q_p64_x4): Likewise.
6903 (vst1_f64_x4): Likewise.
6904 (vst1q_f64_x4): Likewise.
6905 (vst2_s64): Likewise.
6906 (vst2_u64): Likewise.
6907 (vst2_f64): Likewise.
6908 (vst2_s8): Likewise.
6909 (vst2_p8): Likewise.
6910 (vst2_s16): Likewise.
6911 (vst2_p16): Likewise.
6912 (vst2_s32): Likewise.
6913 (vst2_u8): Likewise.
6914 (vst2_u16): Likewise.
6915 (vst2_u32): Likewise.
6916 (vst2_f16): Likewise.
6917 (vst2_f32): Likewise.
6918 (vst2_p64): Likewise.
6919 (vst2q_s8): Likewise.
6920 (vst2q_p8): Likewise.
6921 (vst2q_s16): Likewise.
6922 (vst2q_p16): Likewise.
6923 (vst2q_s32): Likewise.
6924 (vst2q_s64): Likewise.
6925 (vst2q_u8): Likewise.
6926 (vst2q_u16): Likewise.
6927 (vst2q_u32): Likewise.
6928 (vst2q_u64): Likewise.
6929 (vst2q_f16): Likewise.
6930 (vst2q_f32): Likewise.
6931 (vst2q_f64): Likewise.
6932 (vst2q_p64): Likewise.
6933 (vst3_s64): Likewise.
6934 (vst3_u64): Likewise.
6935 (vst3_f64): Likewise.
6936 (vst3_s8): Likewise.
6937 (vst3_p8): Likewise.
6938 (vst3_s16): Likewise.
6939 (vst3_p16): Likewise.
6940 (vst3_s32): Likewise.
6941 (vst3_u8): Likewise.
6942 (vst3_u16): Likewise.
6943 (vst3_u32): Likewise.
6944 (vst3_f16): Likewise.
6945 (vst3_f32): Likewise.
6946 (vst3_p64): Likewise.
6947 (vst3q_s8): Likewise.
6948 (vst3q_p8): Likewise.
6949 (vst3q_s16): Likewise.
6950 (vst3q_p16): Likewise.
6951 (vst3q_s32): Likewise.
6952 (vst3q_s64): Likewise.
6953 (vst3q_u8): Likewise.
6954 (vst3q_u16): Likewise.
6955 (vst3q_u32): Likewise.
6956 (vst3q_u64): Likewise.
6957 (vst3q_f16): Likewise.
6958 (vst3q_f32): Likewise.
6959 (vst3q_f64): Likewise.
6960 (vst3q_p64): Likewise.
6961 (vst4_s64): Likewise.
6962 (vst4_u64): Likewise.
6963 (vst4_f64): Likewise.
6964 (vst4_s8): Likewise.
6965 (vst4_p8): Likewise.
6966 (vst4_s16): Likewise.
6967 (vst4_p16): Likewise.
6968 (vst4_s32): Likewise.
6969 (vst4_u8): Likewise.
6970 (vst4_u16): Likewise.
6971 (vst4_u32): Likewise.
6972 (vst4_f16): Likewise.
6973 (vst4_f32): Likewise.
6974 (vst4_p64): Likewise.
6975 (vst4q_s8): Likewise.
6976 (vst4q_p8): Likewise.
6977 (vst4q_s16): Likewise.
6978 (vst4q_p16): Likewise.
6979 (vst4q_s32): Likewise.
6980 (vst4q_s64): Likewise.
6981 (vst4q_u8): Likewise.
6982 (vst4q_u16): Likewise.
6983 (vst4q_u32): Likewise.
6984 (vst4q_u64): Likewise.
6985 (vst4q_f16): Likewise.
6986 (vst4q_f32): Likewise.
6987 (vst4q_f64): Likewise.
6988 (vst4q_p64): Likewise.
6989 (vtbx4_s8): Likewise.
6990 (vtbx4_u8): Likewise.
6991 (vtbx4_p8): Likewise.
6992 (vld1_bf16_x2): Likewise.
6993 (vld1q_bf16_x2): Likewise.
6994 (vld1_bf16_x3): Likewise.
6995 (vld1q_bf16_x3): Likewise.
6996 (vld1_bf16_x4): Likewise.
6997 (vld1q_bf16_x4): Likewise.
6998 (vld2_bf16): Likewise.
6999 (vld2q_bf16): Likewise.
7000 (vld2_dup_bf16): Likewise.
7001 (vld2q_dup_bf16): Likewise.
7002 (vld3_bf16): Likewise.
7003 (vld3q_bf16): Likewise.
7004 (vld3_dup_bf16): Likewise.
7005 (vld3q_dup_bf16): Likewise.
7006 (vld4_bf16): Likewise.
7007 (vld4q_bf16): Likewise.
7008 (vld4_dup_bf16): Likewise.
7009 (vld4q_dup_bf16): Likewise.
7010 (vst1_bf16_x2): Likewise.
7011 (vst1q_bf16_x2): Likewise.
7012 (vst1_bf16_x3): Likewise.
7013 (vst1q_bf16_x3): Likewise.
7014 (vst1_bf16_x4): Likewise.
7015 (vst1q_bf16_x4): Likewise.
7016 (vst2_bf16): Likewise.
7017 (vst2q_bf16): Likewise.
7018 (vst3_bf16): Likewise.
7019 (vst3q_bf16): Likewise.
7020 (vst4_bf16): Likewise.
7021 (vst4q_bf16): Likewise.
7022 (vld2_lane_bf16): Likewise.
7023 (vld2q_lane_bf16): Likewise.
7024 (vld3_lane_bf16): Likewise.
7025 (vld3q_lane_bf16): Likewise.
7026 (vld4_lane_bf16): Likewise.
7027 (vld4q_lane_bf16): Likewise.
7028 (vst2_lane_bf16): Likewise.
7029 (vst2q_lane_bf16): Likewise.
7030 (vst3_lane_bf16): Likewise.
7031 (vst3q_lane_bf16): Likewise.
7032 (vst4_lane_bf16): Likewise.
7033 (vst4q_lane_bf16): Likewise.
7034 * config/aarch64/geniterators.sh: Modify iterator regex to
7035 match new vector-tuple modes.
7036 * config/aarch64/iterators.md (insn_count): Extend mode
7037 attribute with vector-tuple type information.
7039 (Vendreg): Likewise.
7042 (VSTRUCT_2D): New mode iterator.
7043 (VSTRUCT_2DNX): Likewise.
7044 (VSTRUCT_2DX): Likewise.
7045 (VSTRUCT_2Q): Likewise.
7046 (VSTRUCT_2QD): Likewise.
7047 (VSTRUCT_3D): Likewise.
7048 (VSTRUCT_3DNX): Likewise.
7049 (VSTRUCT_3DX): Likewise.
7050 (VSTRUCT_3Q): Likewise.
7051 (VSTRUCT_3QD): Likewise.
7052 (VSTRUCT_4D): Likewise.
7053 (VSTRUCT_4DNX): Likewise.
7054 (VSTRUCT_4DX): Likewise.
7055 (VSTRUCT_4Q): Likewise.
7056 (VSTRUCT_4QD): Likewise.
7057 (VSTRUCT_D): Likewise.
7058 (VSTRUCT_Q): Likewise.
7059 (VSTRUCT_QD): Likewise.
7060 (VSTRUCT_ELT): New mode attribute.
7061 (vstruct_elt): Likewise.
7062 * genmodes.c (VECTOR_MODE): Add default prefix and order
7064 (VECTOR_MODE_WITH_PREFIX): Define.
7065 (make_vector_mode): Add mode prefix and order parameters.
7067 2021-11-04 Jonathan Wright <jonathan.wright@arm.com>
7069 * expmed.c (extract_bit_field_1): Ensure modes are tieable.
7071 2021-11-04 Jonathan Wright <jonathan.wright@arm.com>
7073 * expr.c (emit_group_load_1): Remove historic workaround.
7075 2021-11-04 Jonathan Wright <jonathan.wright@arm.com>
7077 * config/aarch64/aarch64-builtins.c (aarch64_init_simd_builtins):
7078 Factor out main loop to...
7079 (aarch64_init_simd_builtin_functions): This new function.
7080 (register_tuple_type): Define.
7081 (aarch64_scalar_builtin_type_p): Define.
7082 (handle_arm_neon_h): Define.
7083 * config/aarch64/aarch64-c.c (aarch64_pragma_aarch64): Handle
7084 pragma for arm_neon.h.
7085 * config/aarch64/aarch64-protos.h (aarch64_advsimd_struct_mode_p):
7087 (handle_arm_neon_h): Likewise.
7088 * config/aarch64/aarch64.c (aarch64_advsimd_struct_mode_p):
7089 Remove static modifier.
7090 * config/aarch64/arm_neon.h (target): Remove Neon vector
7091 structure type definitions.
7093 2021-11-04 Aldy Hernandez <aldyh@redhat.com>
7095 PR tree-optimization/102943
7096 * gimple-range-path.cc (path_range_query::range_on_path_entry):
7097 Prefer range_of_expr unless there are no statements in the BB.
7099 2021-11-04 Aldy Hernandez <aldyh@redhat.com>
7101 PR tree-optimization/102943
7102 * tree-ssa-threadbackward.c (back_threader::find_paths_to_names):
7103 Avoid duplicate calculation of paths.
7105 2021-11-04 Aldy Hernandez <aldyh@redhat.com>
7107 PR tree-optimization/102943
7108 * gimple-range-path.cc (path_range_query::compute_phi_relations):
7109 Only compute relations for SSA names in the import list.
7110 (path_range_query::compute_outgoing_relations): Same.
7111 * gimple-range-path.h (path_range_query::import_p): New.
7113 2021-11-04 Richard Biener <rguenther@suse.de>
7115 PR rtl-optimization/103075
7116 * simplify-rtx.c (exact_int_to_float_conversion_p): Return
7117 false for a VOIDmode operand.
7119 2021-11-04 Richard Sandiford <richard.sandiford@arm.com>
7121 * config/aarch64/aarch64.c (aarch64_vector_costs): Make member
7122 variables private and add "m_" to their names. Remove is_loop.
7123 (aarch64_record_potential_advsimd_unrolling): Replace with...
7124 (aarch64_vector_costs::record_potential_advsimd_unrolling): ...this.
7125 (aarch64_analyze_loop_vinfo): Replace with...
7126 (aarch64_vector_costs::analyze_loop_vinfo): ...this.
7127 Move initialization of (m_)vec_flags to add_stmt_cost.
7128 (aarch64_analyze_bb_vinfo): Delete.
7129 (aarch64_count_ops): Replace with...
7130 (aarch64_vector_costs::count_ops): ...this.
7131 (aarch64_vector_costs::add_stmt_cost): Set m_vec_flags,
7132 using m_costing_for_scalar to test whether we're costing
7133 scalar or vector code.
7134 (aarch64_adjust_body_cost_sve): Replace with...
7135 (aarch64_vector_costs::adjust_body_cost_sve): ...this.
7136 (aarch64_adjust_body_cost): Replace with...
7137 (aarch64_vector_costs::adjust_body_cost): ...this.
7138 (aarch64_vector_costs::finish_cost): Use m_vinfo instead of is_loop.
7140 2021-11-04 Richard Sandiford <richard.sandiford@arm.com>
7142 * target.def (targetm.vectorize.init_cost): Replace with...
7143 (targetm.vectorize.create_costs): ...this.
7144 (targetm.vectorize.add_stmt_cost): Delete.
7145 (targetm.vectorize.finish_cost): Likewise.
7146 (targetm.vectorize.destroy_cost_data): Likewise.
7147 * doc/tm.texi.in (TARGET_VECTORIZE_INIT_COST): Replace with...
7148 (TARGET_VECTORIZE_CREATE_COSTS): ...this.
7149 (TARGET_VECTORIZE_ADD_STMT_COST): Delete.
7150 (TARGET_VECTORIZE_FINISH_COST): Likewise.
7151 (TARGET_VECTORIZE_DESTROY_COST_DATA): Likewise.
7152 * doc/tm.texi: Regenerate.
7153 * tree-vectorizer.h (vec_info::vec_info): Remove target_cost_data
7155 (vec_info::target_cost_data): Change from a void * to a vector_costs *.
7156 (vector_costs): New class.
7157 (init_cost): Take a vec_info and return a vector_costs.
7158 (dump_stmt_cost): Remove data parameter.
7159 (add_stmt_cost): Replace vinfo and data parameters with a vector_costs.
7160 (add_stmt_costs): Likewise.
7161 (finish_cost): Replace data parameter with a vector_costs.
7162 (destroy_cost_data): Delete.
7163 * tree-vectorizer.c (dump_stmt_cost): Remove data argument and
7165 (vec_info::vec_info): Remove the target_cost_data parameter and
7166 initialize the member variable to null instead.
7167 (vec_info::~vec_info): Delete target_cost_data instead of calling
7169 (vector_costs::add_stmt_cost): New function.
7170 (vector_costs::finish_cost): Likewise.
7171 (vector_costs::record_stmt_cost): Likewise.
7172 (vector_costs::adjust_cost_for_freq): Likewise.
7173 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Update
7174 call to vec_info::vec_info.
7175 (vect_compute_single_scalar_iteration_cost): Update after above
7176 changes to costing interface.
7177 (vect_analyze_loop_operations): Likewise.
7178 (vect_estimate_min_profitable_iters): Likewise.
7179 (vect_analyze_loop_2): Initialize LOOP_VINFO_TARGET_COST_DATA
7180 at the start_over point, where it needs to be recreated after
7181 trying without slp. Update retry code accordingly.
7182 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Update call
7183 to vec_info::vec_info.
7184 (vect_slp_analyze_operation): Update after above changes to costing
7186 (vect_bb_vectorization_profitable_p): Likewise.
7187 * targhooks.h (default_init_cost): Replace with...
7188 (default_vectorize_create_costs): ...this.
7189 (default_add_stmt_cost): Delete.
7190 (default_finish_cost, default_destroy_cost_data): Likewise.
7191 * targhooks.c (default_init_cost): Replace with...
7192 (default_vectorize_create_costs): ...this.
7193 (default_add_stmt_cost): Delete, moving logic to vector_costs instead.
7194 (default_finish_cost, default_destroy_cost_data): Delete.
7195 * config/aarch64/aarch64.c (aarch64_vector_costs): Inherit from
7196 vector_costs. Add a constructor.
7197 (aarch64_init_cost): Replace with...
7198 (aarch64_vectorize_create_costs): ...this.
7199 (aarch64_add_stmt_cost): Replace with...
7200 (aarch64_vector_costs::add_stmt_cost): ...this. Use record_stmt_cost
7201 to adjust the cost for inner loops.
7202 (aarch64_finish_cost): Replace with...
7203 (aarch64_vector_costs::finish_cost): ...this.
7204 (aarch64_destroy_cost_data): Delete.
7205 (TARGET_VECTORIZE_INIT_COST): Replace with...
7206 (TARGET_VECTORIZE_CREATE_COSTS): ...this.
7207 (TARGET_VECTORIZE_ADD_STMT_COST): Delete.
7208 (TARGET_VECTORIZE_FINISH_COST): Likewise.
7209 (TARGET_VECTORIZE_DESTROY_COST_DATA): Likewise.
7210 * config/i386/i386.c (ix86_vector_costs): New structure.
7211 (ix86_init_cost): Replace with...
7212 (ix86_vectorize_create_costs): ...this.
7213 (ix86_add_stmt_cost): Replace with...
7214 (ix86_vector_costs::add_stmt_cost): ...this. Use adjust_cost_for_freq
7215 to adjust the cost for inner loops.
7216 (ix86_finish_cost, ix86_destroy_cost_data): Delete.
7217 (TARGET_VECTORIZE_INIT_COST): Replace with...
7218 (TARGET_VECTORIZE_CREATE_COSTS): ...this.
7219 (TARGET_VECTORIZE_ADD_STMT_COST): Delete.
7220 (TARGET_VECTORIZE_FINISH_COST): Likewise.
7221 (TARGET_VECTORIZE_DESTROY_COST_DATA): Likewise.
7222 * config/rs6000/rs6000.c (TARGET_VECTORIZE_INIT_COST): Replace with...
7223 (TARGET_VECTORIZE_CREATE_COSTS): ...this.
7224 (TARGET_VECTORIZE_ADD_STMT_COST): Delete.
7225 (TARGET_VECTORIZE_FINISH_COST): Likewise.
7226 (TARGET_VECTORIZE_DESTROY_COST_DATA): Likewise.
7227 (rs6000_cost_data): Inherit from vector_costs.
7228 Add a constructor. Drop loop_info, cost and costing_for_scalar
7229 in favor of the corresponding vector_costs member variables.
7230 Add "m_" to the names of the remaining member variables and
7232 (rs6000_density_test): Replace with...
7233 (rs6000_cost_data::density_test): ...this.
7234 (rs6000_init_cost): Replace with...
7235 (rs6000_vectorize_create_costs): ...this.
7236 (rs6000_update_target_cost_per_stmt): Replace with...
7237 (rs6000_cost_data::update_target_cost_per_stmt): ...this.
7238 (rs6000_add_stmt_cost): Replace with...
7239 (rs6000_cost_data::add_stmt_cost): ...this. Use adjust_cost_for_freq
7240 to adjust the cost for inner loops.
7241 (rs6000_adjust_vect_cost_per_loop): Replace with...
7242 (rs6000_cost_data::adjust_vect_cost_per_loop): ...this.
7243 (rs6000_finish_cost): Replace with...
7244 (rs6000_cost_data::finish_cost): ...this. Group loop code
7245 into a single if statement and pass the loop_vinfo down to
7247 (rs6000_destroy_cost_data): Delete.
7249 2021-11-04 Aldy Hernandez <aldyh@redhat.com>
7251 PR tree-optimization/103062
7252 PR tree-optimization/103062
7253 * value-pointer-equiv.cc (ssa_equiv_stack::ssa_equiv_stack):
7254 Increase size of allocation by 1.
7255 (ssa_equiv_stack::push_replacement): Grow as needed.
7256 (ssa_equiv_stack::get_replacement): Same.
7257 (pointer_equiv_analyzer::pointer_equiv_analyzer): Same.
7258 (pointer_equiv_analyzer::~pointer_equiv_analyzer): Remove delete.
7259 (pointer_equiv_analyzer::set_global_equiv): Grow as needed.
7260 (pointer_equiv_analyzer::get_equiv): Same.
7261 (pointer_equiv_analyzer::get_equiv_expr): Remove const.
7262 * value-pointer-equiv.h (class pointer_equiv_analyzer): Remove
7263 const markers. Use auto_vec instead of tree *.
7265 2021-11-04 Richard Biener <rguenther@suse.de>
7267 * tree-ssa-sccvn.c (vn_nary_op_insert_into): Remove always
7268 true parameter and inline valueization.
7269 (vn_nary_op_lookup_1): Inline valueization from ...
7270 (vn_nary_op_compute_hash): ... here and remove it here.
7271 * tree-ssa-pre.c (phi_translate_1): Do not valueize
7272 before vn_nary_lookup_pieces.
7273 (get_representative_for): Mark created SSA representatives
7276 2021-11-04 Richard Sandiford <richard.sandiford@arm.com>
7278 * simplify-rtx.c (simplify_context::simplify_gen_vec_select): Assert
7279 that the operand has a vector mode. Use subreg_lowpart_offset
7280 to test whether an index corresponds to the low part.
7282 2021-11-04 Richard Sandiford <richard.sandiford@arm.com>
7284 * read-rtl.c: Remove dead !GENERATOR_FILE block.
7285 * read-rtl-function.c (function_reader::consolidate_singletons):
7286 Generate canonical CONST_VECTORs.
7288 2021-11-04 liuhongt <hongtao.liu@intel.com>
7291 * config/i386/predicates.md (reg_or_notreg_operand): Rename to ..
7292 (regmem_or_bitnot_regmem_operand): .. and extend to handle
7294 * config/i386/sse.md (*<avx512>_vpternlog<mode>_1): Force_reg
7295 the operands which are required to be register_operand.
7296 (*<avx512>_vpternlog<mode>_2): Ditto.
7297 (*<avx512>_vpternlog<mode>_3): Ditto.
7298 (*<avx512>_vternlog<mode>_all): Disallow embeded broadcast for
7299 vector HFmodes since it's not a real AVX512FP16 instruction.
7301 2021-11-04 liuhongt <hongtao.liu@intel.com>
7304 * match.pd: simplify (trunc)copysign((extend)a, (extend)b) to
7305 .COPYSIGN (a,b) when a and b are same type as the truncation
7306 type and has less precision than extend type.
7308 2021-11-04 Richard Biener <rguenther@suse.de>
7310 * doc/generic.texi: Update TARGET_MEM_REF and MEM_REF
7313 2021-11-04 Hongyu Wang <hongyu.wang@intel.com>
7315 * config/i386/sse.md (VI2_AVX512VNNIBW): New mode iterator.
7316 (VI1_AVX512VNNI): Likewise.
7317 (SDOT_VPDP_SUF): New mode_attr.
7320 (sdot_prod<mode>): Use VI2_AVX512F iterator, expand to
7321 vpdpwssd when VNNI targets available.
7322 (usdot_prod<mode>): New expander for vector QImode.
7324 2021-11-04 Hongyu Wang <hongyu.wang@intel.com>
7326 * config/i386/amxtileintrin.h (_tile_loadd_internal): Add
7327 parentheses to base and stride.
7328 (_tile_stream_loadd_internal): Likewise.
7329 (_tile_stored_internal): Likewise.
7331 2021-11-03 Maciej W. Rozycki <macro@embecosm.com>
7333 * config/riscv/riscv.c (riscv_class_max_nregs): Swap the
7334 arguments to `reg_class_subset_p'.
7336 2021-11-03 Joseph Myers <joseph@codesourcery.com>
7339 * fold-const.c (fold_init): New function.
7340 * fold-const.h (fold_init): New prototype.
7342 2021-11-03 Andrew MacLeod <amacleod@redhat.com>
7344 * params.opt (param_vrp2_mode): Make ranger the default for VRP2.
7346 2021-11-03 Andrew MacLeod <amacleod@redhat.com>
7348 * gimple-range-fold.cc (fold_using_range::range_of_phi): Don't import
7349 a range from edge if arg == phidef.
7351 2021-11-03 Andrew MacLeod <amacleod@redhat.com>
7353 * gimple-range-fold.cc (fold_using_range::range_of_builtin_call): Test
7354 for constant before any other processing.
7356 2021-11-03 Andrew MacLeod <amacleod@redhat.com>
7358 * flag-types.h (RANGER_DEBUG_ALL): Fix values.
7360 2021-11-03 Andrew MacLeod <amacleod@redhat.com>
7362 * gimple-range.cc (gimple_ranger::gimple_ranger): Initialize current_bb.
7363 (gimple_ranger::range_of_expr): Pick up range_on_entry when there is
7364 no explcit context and current_bb is set.
7365 (gimple_ranger::fold_stmt): New.
7366 * gimple-range.h (current_bb, fold_stmt): New.
7367 * tree-vrp.c (rvrp_folder::fold_stmt): Call ranger's fold_stmt.
7369 2021-11-03 Richard Biener <rguenther@suse.de>
7371 PR tree-optimization/102970
7372 * tree-ssa-pre.c (phi_translate_1): Drop clique and base
7373 when translating a MEM_REF over a backedge.
7375 2021-11-03 Philipp Tomsich <philipp.tomsich@vrull.eu>
7377 * config/aarch64/aarch64-cores.def (AARCH64_CORE): New Ampere-1 core.
7378 * config/aarch64/aarch64-tune.md: Regenerate.
7379 * config/aarch64/aarch64-cost-tables.h: Add extra costs for Ampere-1.
7380 * config/aarch64/aarch64.c: Add tuning structures for Ampere-1.
7381 * doc/invoke.texi: Add documentation for Ampere-1 core.
7383 2021-11-03 Wilco Dijkstra <wdijkstr@arm.com>
7385 * config/aarch64/aarch64.md (movsi): Add alternative for GOT accesses.
7387 (ldr_got_small_<mode>): Remove pattern.
7388 (ldr_got_small_sidi): Likewise.
7389 * config/aarch64/aarch64.c (aarch64_load_symref_appropriately): Keep
7390 GOT accesses as moves.
7391 (aarch64_print_operand): Correctly print got_lo12 in L specifier.
7392 (aarch64_mov_operand_p): Make GOT accesses valid move operands.
7393 * config/aarch64/constraints.md: Add new constraint Usw for GOT access.
7395 2021-11-03 Martin Liska <mliska@suse.cz>
7397 * gcov.c (read_line): Remove dead variable.
7399 2021-11-03 Martin Liska <mliska@suse.cz>
7402 * ipa-fnsummary.c (edge_predicate_pool): Rename predicate class to ipa_predicate.
7403 (ipa_fn_summary::account_size_time): Likewise.
7404 (edge_set_predicate): Likewise.
7405 (set_hint_predicate): Likewise.
7406 (add_freqcounting_predicate): Likewise.
7407 (evaluate_conditions_for_known_args): Likewise.
7408 (evaluate_properties_for_edge): Likewise.
7409 (remap_freqcounting_preds_after_dup): Likewise.
7410 (ipa_fn_summary_t::duplicate): Likewise.
7411 (set_cond_stmt_execution_predicate): Likewise.
7412 (set_switch_stmt_execution_predicate): Likewise.
7413 (compute_bb_predicates): Likewise.
7414 (will_be_nonconstant_expr_predicate): Likewise.
7415 (will_be_nonconstant_predicate): Likewise.
7416 (phi_result_unknown_predicate): Likewise.
7417 (predicate_for_phi_result): Likewise.
7418 (analyze_function_body): Likewise.
7419 (compute_fn_summary): Likewise.
7420 (summarize_calls_size_and_time): Likewise.
7421 (estimate_calls_size_and_time): Likewise.
7422 (ipa_call_context::estimate_size_and_time): Likewise.
7423 (remap_edge_summaries): Likewise.
7424 (remap_freqcounting_predicate): Likewise.
7425 (ipa_merge_fn_summary_after_inlining): Likewise.
7426 (ipa_update_overall_fn_summary): Likewise.
7427 (read_ipa_call_summary): Likewise.
7428 (inline_read_section): Likewise.
7429 * ipa-fnsummary.h (struct ipa_freqcounting_predicate): Likewise.
7430 * ipa-predicate.c (predicate::add_clause): Likewise.
7431 (ipa_predicate::add_clause): Likewise.
7432 (predicate::or_with): Likewise.
7433 (ipa_predicate::or_with): Likewise.
7434 (predicate::evaluate): Likewise.
7435 (ipa_predicate::evaluate): Likewise.
7436 (predicate::probability): Likewise.
7437 (ipa_predicate::probability): Likewise.
7438 (dump_condition): Likewise.
7439 (dump_clause): Likewise.
7440 (predicate::dump): Likewise.
7441 (ipa_predicate::dump): Likewise.
7442 (predicate::debug): Likewise.
7443 (ipa_predicate::debug): Likewise.
7444 (predicate::remap_after_duplication): Likewise.
7445 (ipa_predicate::remap_after_duplication): Likewise.
7446 (predicate::remap_after_inlining): Likewise.
7447 (ipa_predicate::remap_after_inlining): Likewise.
7448 (predicate::stream_in): Likewise.
7449 (ipa_predicate::stream_in): Likewise.
7450 (predicate::stream_out): Likewise.
7451 (ipa_predicate::stream_out): Likewise.
7452 (add_condition): Likewise.
7453 * ipa-predicate.h (class predicate): Likewise.
7454 (class ipa_predicate): Likewise.
7455 (add_condition): Likewise.
7457 2021-11-03 Richard Biener <rguenther@suse.de>
7459 * bitmap.h (bitmap_bit_p): Change the return type to bool.
7460 * bitmap.c (bitmap_bit_p): Likewise.
7461 * sbitmap.h (bitmap_bit_p): Likewise.
7462 (bitmap_set_bit): Return whether the bit changed.
7463 (bitmap_clear_bit): Likewise.
7464 * tree-ssa.c (verify_vssa): Make use of the changed state
7465 from bitmap_set_bit.
7467 2021-11-03 Richard Biener <rguenther@suse.de>
7469 PR middle-end/103033
7470 * internal-fn.c (expand_DEFERRED_INIT): Elide the
7471 native_interpret_expr path in favor of folding the
7472 VIEW_CONVERT_EXPR generated when punning the RHS.
7474 2021-11-03 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
7476 * config/s390/s390.c (s390_loop_unroll_adjust): In case of early
7479 2021-11-03 Jan Hubicka <hubicka@ucw.cz>
7482 * ipa-modref.c (callee_to_caller_flags): New function.
7483 (modref_eaf_analysis::analyze_ssa_name): Use it.
7484 (ipa_merge_modref_summary_after_inlining): Fix whitespace.
7486 2021-11-02 Jan Hubicka <hubicka@ucw.cz>
7488 * ipa-modref.c (modref_eaf_analysis::analyze_ssa_name): Revert
7491 2021-11-02 Roger Sayle <roger@nextmovesoftware.com>
7492 Uroš Bizjak <ubizjak@gmail.com>
7494 * config/i386/i386.md (<any_rotate>ti3): Provide expansion for
7495 rotations by non-constant amounts.
7497 2021-11-02 Jan Hubicka <hubicka@ucw.cz>
7499 * ipa-modref.c: Fix anonymous namespace placement.
7500 (class modref_eaf_analysis): New class.
7501 (analyze_ssa_name_flags): Turn to ...
7502 (modref_eaf_analysis::analyze_ssa_name): ... this one.
7503 (merge_call_lhs_flags): Turn to ...
7504 (modref_eaf_analysis::merge_call_lhs_flags): .. this one
7505 (modref_eaf_analysis::merge_with_ssa_name): New member function.
7506 (record_escape_points): Turn to ...
7507 (modref_eaf_analysis::record_escape_points): ... this one.
7508 (analyze_parms): Updat
7509 (ipa_merge_modref_summary_after_inlining): Move to the end of file.
7511 2021-11-02 Jan Hubicka <hubicka@ucw.cz>
7513 * gimple.c (gimple_call_static_chain_flags): New function.
7514 * gimple.h (gimple_call_static_chain_flags): Declare
7515 * ipa-modref.c (modref_summary::modref_summary): Initialize
7517 (modref_summary_lto::modref_summary_lto): Likewise.
7518 (modref_summary::useful_p): Test static_chain_flags.
7519 (modref_summary_lto::useful_p): Likewise.
7520 (struct modref_summary_lto): Add static_chain_flags.
7521 (modref_summary::dump): Dump static_chain_flags.
7522 (modref_summary_lto::dump): Likewise.
7523 (struct escape_point): Add static_cahin_arg.
7524 (analyze_ssa_name_flags): Use gimple_call_static_chain_flags.
7525 (analyze_parms): Handle static chains.
7526 (modref_summaries::duplicate): Duplicate static_chain_flags.
7527 (modref_summaries_lto::duplicate): Likewise.
7528 (modref_write): Stream static_chain_flags.
7529 (read_section): Likewise.
7530 (modref_merge_call_site_flags): Handle static_chain_flags.
7531 * ipa-modref.h (struct modref_summary): Add static_chain_flags.
7532 * tree-ssa-structalias.c (handle_rhs_call): Use
7533 gimple_static_chain_flags.
7535 2021-11-02 Richard Biener <rguenther@suse.de>
7537 PR tree-optimization/103029
7538 * tree-vect-loop-manip.c (vect_loop_versioning): Ensure
7539 the PHI nodes in the loop maintain their original operand
7542 2021-11-02 Jan Hubicka <hubicka@ucw.cz>
7544 * tree-core.h (EAF_NOT_RETURNED_DIRECTLY): New flag.
7545 (EAF_NOREAD): Renumber.
7546 * ipa-modref.c (dump_eaf_flags): Dump EAF_NOT_RETURNED_DIRECTLY.
7547 (remove_useless_eaf_flags): Handle EAF_NOT_RETURNED_DIRECTLY
7548 (deref_flags): Likewise.
7549 (modref_lattice::init): Likewise.
7550 (modref_lattice::merge): Likewise.
7551 (merge_call_lhs_flags): Likewise.
7552 (analyze_ssa_name_flags): Likewise.
7553 (modref_merge_call_site_flags): Likewise.
7554 * tree-ssa-structalias.c (handle_call_arg): Likewise.
7556 2021-11-02 Maciej W. Rozycki <macro@embecosm.com>
7558 * config/riscv/riscv.c (riscv_rtx_costs): Correct a CONST_INT_P
7559 check and remove an unused local variable with shNadd/shNadd.uw
7562 2021-11-02 Tamar Christina <tamar.christina@arm.com>
7564 PR tree-optimization/103007
7565 * tree-vect-slp-patterns.c (complex_fms_pattern::matches): Add elem
7568 2021-11-02 Richard Biener <rguenther@suse.de>
7570 PR middle-end/103038
7571 * fold-const.c (native_interpret_expr): Handle OFFSET_TYPE.
7572 (can_native_interpret_type_p): Likewise.
7573 * internal-fn.c (expand_DEFERRED_INIT): View-convert the
7574 RHS if the LHS is an SSA name.
7576 2021-11-02 Richard Sandiford <richard.sandiford@arm.com>
7578 * langhooks.h (lang_hooks_for_types::simulate_record_decl): New hook.
7579 * langhooks-def.h (lhd_simulate_record_decl): Declare.
7580 (LANG_HOOKS_SIMULATE_RECORD_DECL): Define.
7581 (LANG_HOOKS_FOR_TYPES_INITIALIZER): Include it.
7582 * langhooks.c (lhd_simulate_record_decl): New function.
7584 2021-11-02 Jakub Jelinek <jakub@redhat.com>
7587 * config/i386/i386.c (ix86_vector_mode_supported_p): Reject vector
7588 modes with TImode inner mode if 32-bit.
7590 2021-11-02 liuhongt <hongtao.liu@intel.com>
7592 * doc/sourcebuild.texi (vect_slp_v4qi_store_unalign,
7593 vect_slp_v2hi_store_unalign, vect_slp_v4hi_store_unalign,
7594 vect_slp_v4si_store_unalign): Document efficient target.
7595 (vect_slp_v4qi_store_unalign_1, vect_slp_v8qi_store_unalign_1,
7596 vect_slp_v16qi_store_unalign_1): Ditto.
7597 (vect_slp_v2hi_store_align,vect_slp_v2qi_store_align,
7598 vect_slp_v2si_store_align, vect_slp_v4qi_store_align): Ditto.
7599 (struct_4char_block_move, struct_8char_block_move,
7600 struct_16char_block_move): Ditto.
7602 2021-11-02 Roger Sayle <roger@nextmovesoftware.com>
7603 Jakub Jelinek <jakub@redhat.com>
7606 * config/i386/i386-expand.c (ix86_expand_v1ti_to_ti,
7607 ix86_expand_ti_to_v1ti): New helper functions.
7608 (ix86_expand_v1ti_shift): Check if the amount operand is an
7609 integer constant, and expand as a TImode shift if it isn't.
7610 (ix86_expand_v1ti_rotate): Check if the amount operand is an
7611 integer constant, and expand as a TImode rotate if it isn't.
7612 (ix86_expand_v1ti_ashiftrt): New function to expand arithmetic
7613 right shifts of V1TImode quantities.
7614 * config/i386/i386-protos.h (ix86_expand_v1ti_ashift): Prototype.
7615 * config/i386/sse.md (ashlv1ti3, lshrv1ti3): Change constraints
7616 to QImode general_operand, and let the helper functions lower
7617 shifts by non-constant operands, as TImode shifts. Make
7618 conditional on TARGET_64BIT.
7619 (ashrv1ti3): New expander calling ix86_expand_v1ti_ashiftrt.
7620 (rotlv1ti3, rotrv1ti3): Change shift operand to QImode.
7621 Make conditional on TARGET_64BIT.
7623 2021-11-02 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
7625 * config/s390/s390.md ("*cc_to_int", "tabort", "*tabort_1",
7626 "*tabort_1_plus"): Remove operands non-null check.
7628 2021-11-01 Martin Liska <mliska@suse.cz>
7630 * opt-functions.awk: Add new sanity checking.
7631 * optc-gen.awk: Add new argument to integer_range_info.
7632 * params.opt: Update 2 params which have negative IntegerRange.
7634 2021-11-01 qing zhao <qing.zhao@oracle.com>
7636 * gimplify.c (gimplify_decl_expr): Do not add call to
7637 __builtin_clear_padding when a variable is a gimple register
7638 or it might not have padding.
7639 (gimplify_init_constructor): Likewise.
7641 2021-11-01 Tamar Christina <tamar.christina@arm.com>
7643 * config/arm/aarch-common-protos.h (struct vector_cost_table): Add
7644 movi, dup and extract costing fields.
7645 * config/aarch64/aarch64-cost-tables.h (qdf24xx_extra_costs,
7646 thunderx_extra_costs, thunderx2t99_extra_costs,
7647 thunderx3t110_extra_costs, tsv110_extra_costs, a64fx_extra_costs): Use
7649 * config/arm/aarch-cost-tables.h (generic_extra_costs,
7650 cortexa53_extra_costs, cortexa57_extra_costs, cortexa76_extra_costs,
7651 exynosm1_extra_costs, xgene1_extra_costs): Likewise
7652 * config/aarch64/aarch64-simd.md (aarch64_simd_dup<mode>): Add r->w dup.
7653 * config/aarch64/aarch64.c (aarch64_rtx_costs): Add extra costs.
7655 2021-11-01 Tamar Christina <tamar.christina@arm.com>
7657 * cse.c (add_to_set): New.
7658 (find_sets_in_insn): Register constants in sets.
7659 (canonicalize_insn): Use auto_vec instead.
7660 (cse_insn): Try materializing using vec_dup.
7661 * rtl.h (simplify_context::simplify_gen_vec_select,
7662 simplify_gen_vec_select): New.
7663 * simplify-rtx.c (simplify_context::simplify_gen_vec_select): New.
7665 2021-11-01 David Malcolm <dmalcolm@redhat.com>
7667 * common.opt (fdiagnostics-escape-format=): New.
7668 (diagnostics_escape_format): New enum.
7669 (DIAGNOSTICS_ESCAPE_FORMAT_UNICODE): New enum value.
7670 (DIAGNOSTICS_ESCAPE_FORMAT_BYTES): Likewise.
7671 * diagnostic-format-json.cc (json_end_diagnostic): Add
7672 "escape-source" attribute.
7673 * diagnostic-show-locus.c
7674 (exploc_with_display_col::exploc_with_display_col): Replace
7675 "tabstop" param with a cpp_char_column_policy and add an "aspect"
7676 param. Use these to compute m_display_col accordingly.
7677 (struct char_display_policy): New struct.
7678 (layout::m_policy): New field.
7679 (layout::m_escape_on_output): New field.
7680 (def_policy): New function.
7681 (make_range): Update for changes to exploc_with_display_col ctor.
7682 (default_print_decoded_ch): New.
7683 (width_per_escaped_byte): New.
7684 (escape_as_bytes_width): New.
7685 (escape_as_bytes_print): New.
7686 (escape_as_unicode_width): New.
7687 (escape_as_unicode_print): New.
7689 (layout::layout): Initialize new fields. Update m_exploc ctor
7690 call for above change to ctor.
7691 (layout::maybe_add_location_range): Update for changes to
7692 exploc_with_display_col ctor.
7693 (layout::calculate_x_offset_display): Update for change to
7695 (layout::print_source_line): Pass policy
7696 to cpp_display_width_computation. Capture cpp_decoded_char when
7697 calling process_next_codepoint. Move printing of source code to
7698 m_policy.m_print_cb.
7699 (line_label::line_label): Pass in policy rather than context.
7700 (layout::print_any_labels): Update for change to line_label ctor.
7701 (get_affected_range): Pass in policy rather than context, updating
7702 calls to location_compute_display_column accordingly.
7703 (get_printed_columns): Likewise, also for cpp_display_width.
7704 (correction::correction): Pass in policy rather than tabstop.
7705 (correction::compute_display_cols): Pass m_policy rather than
7706 m_tabstop to cpp_display_width.
7707 (correction::m_tabstop): Replace with...
7708 (correction::m_policy): ...this.
7709 (line_corrections::line_corrections): Pass in policy rather than
7711 (line_corrections::m_context): Replace with...
7712 (line_corrections::m_policy): ...this.
7713 (line_corrections::add_hint): Update to use m_policy rather than
7715 (line_corrections::add_hint): Likewise.
7716 (layout::print_trailing_fixits): Likewise.
7717 (selftest::test_display_widths): New.
7718 (selftest::test_layout_x_offset_display_utf8): Update to use
7719 policy rather than tabstop.
7720 (selftest::test_one_liner_labels_utf8): Add test of escaping
7722 (selftest::test_diagnostic_show_locus_one_liner_utf8): Update to
7723 use policy rather than tabstop.
7724 (selftest::test_overlapped_fixit_printing): Likewise.
7725 (selftest::test_overlapped_fixit_printing_utf8): Likewise.
7726 (selftest::test_overlapped_fixit_printing_2): Likewise.
7727 (selftest::test_tab_expansion): Likewise.
7728 (selftest::test_escaping_bytes_1): New.
7729 (selftest::test_escaping_bytes_2): New.
7730 (selftest::diagnostic_show_locus_c_tests): Call the new tests.
7731 * diagnostic.c (diagnostic_initialize): Initialize
7732 context->escape_format.
7733 (convert_column_unit): Update to use default character width policy.
7734 (selftest::test_diagnostic_get_location_text): Likewise.
7735 * diagnostic.h (enum diagnostics_escape_format): New enum.
7736 (diagnostic_context::escape_format): New field.
7737 * doc/invoke.texi (-fdiagnostics-escape-format=): New option.
7738 (-fdiagnostics-format=): Add "escape-source" attribute to examples
7739 of JSON output, and document it.
7740 * input.c (location_compute_display_column): Pass in "policy"
7741 rather than "tabstop", passing to
7742 cpp_byte_column_to_display_column.
7743 (selftest::test_cpp_utf8): Update to use cpp_char_column_policy.
7744 * input.h (class cpp_char_column_policy): New forward decl.
7745 (location_compute_display_column): Pass in "policy" rather than
7747 * opts.c (common_handle_option): Handle
7748 OPT_fdiagnostics_escape_format_.
7749 * selftest.c (temp_source_file::temp_source_file): New ctor
7750 overload taking a size_t.
7751 * selftest.h (temp_source_file::temp_source_file): Likewise.
7753 2021-11-01 Aldy Hernandez <aldyh@redhat.com>
7755 * dbgcnt.def: Add debug counter for back_thread[12] and
7756 back_threadfull[12].
7757 * passes.def: Pass "first" argument to each back threading pass.
7758 * tree-ssa-threadbackward.c (back_threader::back_threader): Add
7760 (back_threader::debug_counter): New.
7761 (back_threader::maybe_register_path): Call debug_counter.
7763 2021-11-01 Aldy Hernandez <aldyh@redhat.com>
7765 * tree-ssa-threadbackward.c (BT_NONE): New.
7768 (back_threader::back_threader): Add flags.
7769 Move loop initialization here.
7770 (back_threader::~back_threader): New.
7771 (back_threader::find_taken_edge_switch): Change solver and ranger
7773 (back_threader::find_taken_edge_cond): Same.
7774 (back_threader::find_paths_to_names): Same.
7775 (back_threader::find_paths): Same.
7776 (back_threader::dump): Same.
7777 (try_thread_blocks): Merge into thread_blocks.
7778 (back_threader::thread_blocks): New.
7779 (do_early_thread_jumps): Merge into thread_blocks.
7780 (do_thread_jumps): Merge into thread_blocks.
7781 (back_threader::thread_through_all_blocks): Remove.
7783 2021-11-01 Andrew MacLeod <amacleod@redhat.com>
7785 PR tree-optimization/103003
7786 * value-relation.cc (dom_oracle::register_relation): If the 2
7787 ssa names are the same, don't register any relation.
7789 2021-11-01 Dan Li <ashimida@linux.alibaba.com>
7791 * config/aarch64/aarch64.c (aarch64_expand_epilogue): Remove
7792 redundant check for calls_eh_return.
7793 * config/aarch64/aarch64.md (*do_return): Likewise.
7795 2021-11-01 Xionghu Luo <luoxhu@linux.ibm.com>
7797 * cfghooks.c (cfg_hook_duplicate_loop_to_header_edge): Rename
7798 duplicate_loop_to_header_edge to
7799 duplicate_loop_body_to_header_edge.
7800 (cfg_hook_duplicate_loop_body_to_header_edge): Likewise.
7801 * cfghooks.h (struct cfg_hooks): Likewise.
7802 (cfg_hook_duplicate_loop_body_to_header_edge): Likewise.
7803 * cfgloopmanip.c (duplicate_loop_body_to_header_edge): Likewise.
7804 (clone_loop_to_header_edge): Likewise.
7805 * cfgloopmanip.h (duplicate_loop_body_to_header_edge): Likewise.
7806 * cfgrtl.c (struct cfg_hooks): Likewise.
7807 * doc/loop.texi: Likewise.
7808 * loop-unroll.c (unroll_loop_constant_iterations): Likewise.
7809 (unroll_loop_runtime_iterations): Likewise.
7810 (unroll_loop_stupid): Likewise.
7811 (apply_opt_in_copies): Likewise.
7812 * tree-cfg.c (struct cfg_hooks): Likewise.
7813 * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Likewise.
7814 (try_peel_loop): Likewise.
7815 * tree-ssa-loop-manip.c (copy_phi_node_args): Likewise.
7816 (gimple_duplicate_loop_body_to_header_edge): Likewise.
7817 (tree_transform_and_unroll_loop): Likewise.
7818 * tree-ssa-loop-manip.h (gimple_duplicate_loop_body_to_header_edge):
7821 2021-11-01 Xionghu Luo <luoxhu@linux.ibm.com>
7823 * cfgloopmanip.c (loop_version): Refactor loopify to
7824 loop_version. Move condition generation after loopify.
7826 * cfgloopmanip.h (loopify): Delete.
7828 2021-10-31 Jan Hubicka <hubicka@ucw.cz>
7830 * ipa-fnsummary.c: Include tree-dfa.h.
7831 (points_to_local_or_readonly_memory_p): Return true on return
7833 * ipa-modref.c (analyze_ssa_name_flags): Fix handling of copy
7836 2021-10-30 Tobias Burnus <tobias@codesourcery.com>
7838 PR middle-end/102972
7839 * omp-low.c (omp_runtime_api_call): Use DECL_ASSEMBLER_NAME to get
7840 internal Fortran name; new permit_num_teams arg to permit
7841 omp_get_num_teams and omp_get_team_num.
7842 (scan_omp_1_stmt): Update call to it, add missing call for
7843 reverse offload, and check for strictly nested API calls in teams.
7845 2021-10-30 Jakub Jelinek <jakub@redhat.com>
7847 * gimplify.c (gimplify_omp_for): Diagnose threadprivate iterators.
7849 2021-10-29 David Malcolm <dmalcolm@redhat.com>
7851 * selftest.c (assert_streq): Add newlines when emitting non-equal
7854 2021-10-29 David Malcolm <dmalcolm@redhat.com>
7856 * Makefile.in: Fix syntax for reference to LIBDEPS in
7859 2021-10-29 Bernhard Reutner-Fischer <aldot@gcc.gnu.org>
7861 * doc/install.texi: Bump required minimum DejaGnu version.
7863 2021-10-29 Aldy Hernandez <aldyh@redhat.com>
7864 Andrew MacLeod <amacleod@redhat.com>
7866 * value-relation.cc (path_oracle::killing_def): Add a
7867 self-equivalence so we don't look to the root oracle.
7869 2021-10-29 Aldy Hernandez <aldyh@redhat.com>
7871 * passes.def: Replace the pass_thread_jumps before VRP* with
7872 pass_thread_jumps_full. Remove all pass_vrp_threader instances.
7873 * tree-ssa-threadbackward.c (pass_data_thread_jumps_full):
7874 Remove hyphen from "thread-full" name.
7876 2021-10-29 Andrew MacLeod <amacleod@redhat.com>
7878 PR tree-optimization/102983
7879 * gimple-range-cache.h (propagate_updated_value): Make public.
7880 * gimple-range.cc (gimple_ranger::range_of_stmt): Propagate exports
7881 when processing gcond stmts.
7883 2021-10-29 Jan Hubicka <hubicka@ucw.cz>
7885 * gimple.c (gimple_call_retslot_flags): New function.
7886 * gimple.h (gimple_call_retslot_flags): Declare.
7887 * ipa-modref.c: Include tree-cfg.h.
7888 (struct escape_entry): Turn parm_index to signed.
7889 (modref_summary_lto::modref_summary_lto): Add retslot_flags.
7890 (modref_summary::modref_summary): Initialize retslot_flags.
7891 (struct modref_summary_lto): Likewise.
7892 (modref_summary::useful_p): Check retslot_flags.
7893 (modref_summary_lto::useful_p): Likewise.
7894 (modref_summary::dump): Dump retslot_flags.
7895 (modref_summary_lto::dump): Likewise.
7896 (struct escape_point): Add hidden_args enum.
7897 (analyze_ssa_name_flags): Ignore return slot return;
7898 use gimple_call_retslot_flags.
7899 (record_escape_points): Break out from ...
7900 (analyze_parms): ... here; handle retslot_flags.
7901 (modref_summaries::duplicate): Duplicate retslot_flags.
7902 (modref_summaries_lto::duplicate): Likewise.
7903 (modref_write_escape_summary): Stream parm_index as signed.
7904 (modref_read_escape_summary): Likewise.
7905 (modref_write): Stream retslot_flags.
7906 (read_section): Likewise.
7907 (struct escape_map): Fix typo in comment.
7908 (update_escape_summary_1): Fix whitespace.
7909 (ipa_merge_modref_summary_after_inlining): Drop retslot_flags.
7910 (modref_merge_call_site_flags): Merge retslot_flags.
7911 * ipa-modref.h (struct modref_summary): Add retslot_flags.
7912 * tree-ssa-structalias.c (handle_rhs_call): Handle retslot_flags.
7914 2021-10-29 Tamar Christina <tamar.christina@arm.com>
7916 PR tree-optimization/102977
7917 * tree-vect-slp-patterns.c (vect_match_call_p): Remove.
7918 (vect_detect_pair_op): Add crosslane check.
7919 (vect_match_call_complex_mla): Remove.
7920 (class complex_mul_pattern): Update comment.
7921 (complex_mul_pattern::matches): Update detection.
7922 (class complex_fma_pattern): Remove.
7923 (complex_fma_pattern::matches): Remove.
7924 (complex_fma_pattern::recognize): Remove.
7925 (complex_fma_pattern::build): Remove.
7926 (class complex_fms_pattern): Update comment.
7927 (complex_fms_pattern::matches): Remove.
7928 (complex_operations_pattern::recognize): Remove complex_fma_pattern
7930 2021-10-29 Jakub Jelinek <jakub@redhat.com>
7932 * gimple-fold.c (gimple_fold_builtin_memset): Copy over location from
7935 2021-10-29 Xionghu Luo <luoxhu@linux.ibm.com>
7938 * config/rs6000/rs6000.c (altivec_expand_vec_perm_const): Add
7939 patterns match and emit for VSX xxpermdi.
7941 2021-10-29 liuhongt <hongtao.liu@intel.com>
7944 * config/i386/i386-builtin-types.def (V8HF_FTYPE_V8HF): New
7946 (V16HF_FTYPE_V16HF): Ditto.
7947 (V32HF_FTYPE_V32HF): Ditto.
7948 (V8HF_FTYPE_V8HF_ROUND): Ditto.
7949 (V16HF_FTYPE_V16HF_ROUND): Ditto.
7950 (V32HF_FTYPE_V32HF_ROUND): Ditto.
7951 * config/i386/i386-builtin.def ( IX86_BUILTIN_FLOORPH,
7952 IX86_BUILTIN_CEILPH, IX86_BUILTIN_TRUNCPH,
7953 IX86_BUILTIN_FLOORPH256, IX86_BUILTIN_CEILPH256,
7954 IX86_BUILTIN_TRUNCPH256, IX86_BUILTIN_FLOORPH512,
7955 IX86_BUILTIN_CEILPH512, IX86_BUILTIN_TRUNCPH512): New builtin.
7956 * config/i386/i386-builtins.c
7957 (ix86_builtin_vectorized_function): Enable vectorization for
7958 HFmode FLOOR/CEIL/TRUNC operation.
7959 * config/i386/i386-expand.c (ix86_expand_args_builtin): Handle
7961 * config/i386/sse.md (rint<mode>2, nearbyint<mode>2): Extend
7964 2021-10-28 Aldy Hernandez <aldyh@redhat.com>
7965 Andrew MacLeod <amacleod@redhat.com>
7967 * value-relation.cc (path_oracle::killing_def): Walk the
7968 equivalency list and remove SSA from any equivalencies.
7970 2021-10-28 Stafford Horne <shorne@gmail.com>
7972 * config/or1k/or1k.h (PROFILE_HOOK): Add return address argument
7975 2021-10-28 Jakub Jelinek <jakub@redhat.com>
7977 PR tree-optimization/102951
7978 * fold-const.h (address_compare): Declare.
7979 * fold-const.c (address_compare): New function.
7980 * match.pd (cmp (convert1?@2 addr@0) (convert2? addr@1)): Use
7981 address_compare helper.
7982 (minmax cmp (convert1?@2 addr@0) (convert2?@3 addr@1)): New
7985 2021-10-28 Andrew MacLeod <amacleod@redhat.com>
7987 * vr-values.c (simplify_using_ranges::fold_cond): Change fold message.
7989 2021-10-28 Andrew MacLeod <amacleod@redhat.com>
7991 PR tree-optimization/102940
7992 * tree-vrp.c (execute_ranger_vrp): Reset scev.
7994 2021-10-28 Richard Purdie <richard.purdie@linuxfoundation.org>
7996 * config/nios2/linux.h (MUSL_DYNAMIC_LINKER): Add musl linker
7998 2021-10-28 Richard Purdie <richard.purdie@linuxfoundation.org>
8000 * configure: Regenerate.
8001 * configure.ac: Use CPPFLAGS_FOR_BUILD for GMPINC
8003 2021-10-28 Eric Botcazou <ebotcazou@adacore.com>
8005 * doc/invoke.texi (%X): Remove obsolete reference to -Wl.
8007 2021-10-28 Richard Biener <rguenther@suse.de>
8010 * fold-const.c (fold_convert_const): Avoid int to float
8011 constant folding with -frounding-math and inexact result.
8012 * simplify-rtx.c (simplify_const_unary_operation): Likewise
8013 for both float and unsigned_float.
8015 2021-10-28 Aldy Hernandez <aldyh@redhat.com>
8017 * tree-ssa-threadbackward.c
8018 (back_threader::find_taken_edge_switch): Use find_case_label_range
8019 instead of find_taken_edge.
8021 2021-10-28 Aldy Hernandez <aldyh@redhat.com>
8023 * tree-ssa-threadbackward.c (class back_threader_registry):
8024 Inherit from back_jt_path_registry.
8025 (back_threader_registry::thread_through_all_blocks): Remove.
8026 (back_threader_registry::register_path): Remove
8027 m_lowlevel_registry prefix.
8029 2021-10-28 Richard Biener <rguenther@suse.de>
8032 * fold-const.c (fold_convert_const_real_from_real): Honor
8033 -frounding-math if the conversion is not exact.
8034 * simplify-rtx.c (simplify_const_unary_operation): Do not
8035 simplify FLOAT_TRUNCATE with sign dependent rounding.
8037 2021-10-28 Richard Biener <rguenther@suse.de>
8039 PR tree-optimization/102949
8040 * tree-vect-stmts.c (ensure_base_align): Look at the
8041 dr_info of a group leader and assert we are looking at
8042 one with analyzed alignment.
8044 2021-10-28 Kewen Lin <linkw@linux.ibm.com>
8047 * config/rs6000/rs6000.c (rs6000_builtin_vectorization_cost): Consider
8048 V1T1 mode for unaligned load and store.
8050 2021-10-28 Kito Cheng <kito.cheng@sifive.com>
8052 * config/riscv/riscv.md (zero_extendsidi2_internal): Allow ZBB
8055 2021-10-28 Kito Cheng <kito.cheng@sifive.com>
8057 * config/riscv/arch-canonicalize (CANONICAL_ORDER): Add `i` to
8060 2021-10-28 Alexandre Oliva <oliva@adacore.com>
8062 * common.opt (fharden-compares): New.
8063 (fharden-conditional-branches): New.
8064 * doc/invoke.texi: Document new options.
8065 * gimple-harden-conditionals.cc: New.
8066 * Makefile.in (OBJS): Build it.
8067 * passes.def: Add new passes.
8068 * tree-pass.h (make_pass_harden_compares): Declare.
8069 (make_pass_harden_conditional_branches): Declare.
8071 2021-10-28 Xionghu Luo <luoxhu@linux.ibm.com>
8074 * config/rs6000/altivec.md: Add vsx register constraints.
8075 * config/rs6000/vsx.md (vsx_xxsel<mode>): Delete.
8076 (vsx_xxsel<mode>2): Likewise.
8077 (vsx_xxsel<mode>3): Likewise.
8078 (vsx_xxsel<mode>4): Likewise.
8080 2021-10-28 Xionghu Luo <luoxhu@linux.ibm.com>
8083 * config/rs6000/altivec.md (*altivec_vsel<mode>): Change to ...
8084 (altivec_vsel<mode>): ... this and update define.
8085 (*altivec_vsel<mode>_uns): Delete.
8086 (altivec_vsel<mode>2): New define_insn.
8087 (altivec_vsel<mode>3): Likewise.
8088 (altivec_vsel<mode>4): Likewise.
8089 * config/rs6000/rs6000-call.c (altivec_expand_vec_sel_builtin): New.
8090 (altivec_expand_builtin): Call altivec_expand_vec_sel_builtin to expand
8092 * config/rs6000/rs6000.c (rs6000_emit_vector_cond_expr): Use bit-wise
8093 selection instead of per element.
8094 * config/rs6000/vector.md:
8095 * config/rs6000/vsx.md (*vsx_xxsel<mode>): Change to ...
8096 (vsx_xxsel<mode>): ... this and update define.
8097 (*vsx_xxsel<mode>_uns): Delete.
8098 (vsx_xxsel<mode>2): New define_insn.
8099 (vsx_xxsel<mode>3): Likewise.
8100 (vsx_xxsel<mode>4): Likewise.
8102 2021-10-28 Hongyu Wang <hongyu.wang@intel.com>
8104 * config/i386/i386.c (use_rsqrt_p): Add mode parameter, enable
8105 HFmode rsqrt without TARGET_SSE_MATH.
8106 (ix86_optab_supported_p): Refactor rint, adjust floor, ceil,
8107 btrunc condition to be restricted by -ftrapping-math, adjust
8108 use_rsqrt_p function call.
8109 * config/i386/i386.md (rcphf2): New define_insn.
8110 (rsqrthf2): Likewise.
8111 * config/i386/sse.md (div<mode>3): Change VF2H to VF2.
8112 (div<mode>3): New expander for HF mode.
8113 (rsqrt<mode>2): Likewise.
8114 (*avx512fp16_vmrcpv8hf2): New define_insn for rpad pass.
8115 (*avx512fp16_vmrsqrtv8hf2): Likewise.
8117 2021-10-27 Saagar Jha <saagar@saagarjha.com>
8119 * config.gcc: Adjust for Darwin21.
8120 * config/darwin-c.c (macosx_version_as_macro): Likewise.
8121 * config/darwin-driver.c (validate_macosx_version_min):
8123 (darwin_find_version_from_kernel): Likewise.
8125 2021-10-27 Aldy Hernandez <aldyh@redhat.com>
8127 * gimple-range-path.cc
8128 (path_range_query::range_defined_in_block): Call killing_def.
8130 2021-10-27 Aldy Hernandez <aldyh@redhat.com>
8132 * gimple-range-fold.cc (fold_using_range::range_of_range_op): Dump
8133 operands as well as relation.
8134 * gimple-range-path.cc
8135 (path_range_query::compute_ranges_in_block): Compute PHI relations
8136 first. Compute outgoing relations at the end.
8137 (path_range_query::compute_ranges): Remove call to compute_relations.
8138 (path_range_query::compute_relations): Remove.
8139 (path_range_query::maybe_register_phi_relation): New.
8140 (path_range_query::compute_phi_relations): Abstract out
8141 registering one PHI relation to...
8142 (path_range_query::compute_outgoing_relations): ...here.
8143 * gimple-range-path.h (class path_range_query): Remove
8145 Add maybe_register_phi_relation.
8147 2021-10-27 Aldy Hernandez <aldyh@redhat.com>
8148 Andrew MacLeod <amacleod@redhat.com>
8150 * value-relation.cc (path_oracle::killing_def): Kill second
8153 2021-10-27 Martin Jambor <mjambor@suse.cz>
8155 * ipa-cp.c (good_cloning_opportunity_p): Decide whether to use
8156 profile feedback depending on their local availability.
8158 2021-10-27 Martin Jambor <mjambor@suse.cz>
8160 * params.opt (param_ipa_cp_profile_count_base): New parameter.
8161 * doc/invoke.texi (Optimize Options): Add entry for
8162 ipa-cp-profile-count-base.
8163 * ipa-cp.c (max_count): Replace with base_count, replace all
8164 occurrences too, unless otherwise stated.
8165 (ipcp_cloning_candidate_p): identify mostly-directly called
8166 functions based on their counts, not max_count.
8167 (compare_edge_profile_counts): New function.
8168 (ipcp_propagate_stage): Instead of setting max_count, find the
8169 appropriate edge count in a sorted vector of counts of eligible
8170 edges and make it the base_count.
8172 2021-10-27 Martin Jambor <mjambor@suse.cz>
8174 * ipa-cp.c (struct caller_statistics): New fields rec_count_sum,
8175 n_nonrec_calls and itself, document all fields.
8176 (init_caller_stats): Initialize the above new fields.
8177 (gather_caller_stats): Gather self-recursive counts and calls number.
8178 (get_info_about_necessary_edges): Gather counts of self-recursive and
8179 other edges bringing in the requested value separately.
8180 (dump_profile_updates): Rework to dump info about a single node only.
8181 (lenient_count_portion_handling): New function.
8182 (struct gather_other_count_struct): New type.
8183 (gather_count_of_non_rec_edges): New function.
8184 (struct desc_incoming_count_struct): New type.
8185 (analyze_clone_icoming_counts): New function.
8186 (adjust_clone_incoming_counts): Likewise.
8187 (update_counts_for_self_gen_clones): Likewise.
8188 (update_profiling_info): Rewritten.
8189 (update_specialized_profile): Adjust call to dump_profile_updates.
8190 (create_specialized_node): Do not update profiling info.
8191 (decide_about_value): New parameter self_gen_clones, either push new
8192 clones into it or updat their profile counts. For self-recursively
8193 generated values, use a portion of the node count instead of count
8194 from self-recursive edges to estimate goodness.
8195 (decide_whether_version_node): Gather clones for self-generated values
8196 in a new vector, update their profiles at once at the end.
8198 2021-10-27 Richard Biener <rguenther@suse.de>
8200 * tree-vectorizer.c (vect_transform_loops): New function,
8202 (try_vectorize_loop_1): ... here. Simplify as epilogues
8203 are now fully handled in the split part.
8205 2021-10-27 Jakub Jelinek <jakub@redhat.com>
8207 * omp-expand.c (expand_omp_for_init_counts): Handle non-rectangular
8208 iterators with pointer types.
8209 (expand_omp_for_init_vars, extract_omp_for_update_vars): Likewise.
8211 2021-10-26 Martin Sebor <msebor@redhat.com>
8213 PR tree-optimization/102238
8214 PR tree-optimization/102919
8215 * gimple-ssa-sprintf.c (get_string_length): Add an argument.
8216 (array_elt_at_offset): Move to pointer-query.
8217 (set_aggregate_size_and_offset): New function.
8218 (field_at_offset): Move to pointer-query.
8219 (get_origin_and_offset): Rename...
8220 (get_origin_and_offset_r): this. Add an argument. Make aggregate
8221 handling more robust.
8222 (get_origin_and_offset): New.
8223 (alias_offset): Add an argument.
8224 (format_string): Use subobject size determined by get_origin_and_offset.
8225 * pointer-query.cc (field_at_offset): Move from gimple-ssa-sprintf.c.
8226 Improve/correct handling of aggregates.
8227 (array_elt_at_offset): Same.
8228 * pointer-query.h (field_at_offset): Declare.
8229 (array_elt_at_offset): Declare.
8231 2021-10-26 Martin Sebor <msebor@redhat.com>
8233 * builtins.c (check_strncat_sizes): Pass access_data ctor additional
8235 (expand_builtin_memcmp): Move code to gimple-ssa-warn-access.cc.
8236 (expand_builtin_fork_or_exec): Same.
8237 * gimple-array-bounds.cc (array_bounds_checker::check_mem_ref): Pass
8238 compute_objsize additional arguments.
8239 (inbounds_memaccess_p): Same.
8240 (array_bounds_checker::check_array_bounds): Add an assert. Stash
8241 statement in a member.
8242 (check_array_bounds_dom_walker::before_dom_children): Same.
8243 * gimple-array-bounds.h (array_bounds_checker::m_stmt): New member.
8244 * gimple-ssa-sprintf.c (get_destination_size): Add an argument.
8245 (handle_printf_call): Pass a new argument.
8246 * gimple-ssa-warn-access.cc (get_size_range): Add an argument.
8247 (check_access): Add an argument and pass it along to callees.
8248 (check_read_access): Make a member function.
8249 (pass_waccess::check_strcat): Pass access_data ctor additional
8251 (pass_waccess::check_strncat): Same.
8252 (pass_waccess::check_stxcpy): Same.
8253 (pass_waccess::check_stxncpy): Same.
8254 (pass_waccess::check_strncmp): Same.
8255 (pass_waccess::check_read_access): Same.
8256 (pass_waccess::check_builtin): Same.
8257 (pass_waccess::maybe_check_access_sizes): Same.
8258 (pass_waccess::maybe_check_dealloc_call): Same.
8259 * gimple-ssa-warn-access.h (check_read_access): Declare a new
8261 * pointer-query.cc (compute_objsize_r): Add an argument.
8262 (gimple_call_return_array): Same.
8263 (gimple_call_alloc_size): Same.
8264 (access_ref::access_ref): Same.
8265 (access_ref::get_ref): Same.
8266 (pointer_query::get_ref): Same.
8267 (handle_min_max_size): Pass an arguments to callees.
8268 (handle_array_ref): Add an argument.
8269 (handle_mem_ref): Same.
8270 (compute_objsize): Same.
8271 * pointer-query.h (struct access_ref): Adjust signatures.
8272 (struct access_data): Same.
8273 (gimple_call_alloc_size): Add an argument.
8274 (gimple_parm_array_size): Same.
8275 (compute_objsize): Same.
8276 * tree-ssa-strlen.c (strlen_pass::adjust_last_stmt): Pass an additional
8277 argument to compute_objsize.
8278 (strlen_pass::maybe_warn_overflow): Same.
8279 (maybe_diag_stxncpy_trunc): Same.
8281 2021-10-26 Martin Sebor <msebor@redhat.com>
8283 PR middle-end/102453
8284 * gimple-ssa-warn-access.cc (pass_waccess::check_atomic_builtin): New.
8285 (pass_waccess::check_atomic_builtin): Call it.
8287 2021-10-26 Vladimir N. Makarov <vmakarov@redhat.com>
8289 PR rtl-optimization/102842
8290 * lra-constraints.c (match_reload): Ignore out in checking values
8292 (curr_insn_transform): Collect outputs before doing reloads of operands.
8294 2021-10-26 Paul A. Clarke <pc@us.ibm.com>
8297 * config/rs6000/x86intrin.h: Move some included headers to new
8298 headers. Include new immintrin.h instead of those headers.
8299 * config/rs6000/immintrin.h: New.
8300 * config/rs6000/x86gprintrin.h: New.
8301 * config.gcc (powerpc*-*-*): Add new headers to extra_headers.
8303 2021-10-26 Richard Biener <rguenther@suse.de>
8305 * tree-vectorizer.h (vect_create_addr_base_for_vector_ref):
8306 Remove byte_offset parameter.
8307 (vect_create_data_ref_ptr): Likewise.
8308 * tree-vect-data-refs.c (vect_create_addr_base_for_vector_ref):
8310 (vect_create_data_ref_ptr): Likewise.
8311 * tree-vect-stmts.c (vectorizable_store): Adjust.
8312 (vectorizable_load): Likewise.
8314 2021-10-26 Richard Biener <rguenther@suse.de>
8316 PR tree-optimization/96109
8317 * tree-vectorizer.h (dr_misalignment): Add optional offset
8319 * tree-vect-data-refs.c (dr_misalignment): Likewise. Remove
8320 offset applied for negative stride accesses.
8321 (vect_enhance_data_refs_alignment): Compute negative stride
8322 access offset and pass it to dr_misalignment.
8323 * tree-vect-stmts.c (get_negative_load_store_type): Pass
8324 negative offset to dr_misalignment.
8325 (get_group_load_store_type): Likewise.
8326 (get_load_store_type): Likewise.
8327 (vectorizable_store): Remove asserts about alignment.
8328 (vectorizable_load): Likewise.
8330 2021-10-26 Kewen Lin <linkw@linux.ibm.com>
8332 PR tree-optimization/102897
8333 * tree-ssa-forwprop.c (simplify_permutation): Remove a wrong assertion.
8335 2021-10-26 Richard Biener <rguenther@suse.de>
8337 * tree-vect-data-refs.c (vect_create_addr_base_for_vector_ref):
8338 Take offset in bytes.
8339 (vect_create_data_ref_ptr): Likewise.
8340 * tree-vect-loop-manip.c (get_misalign_in_elems): Multiply
8341 offset by element size.
8342 (vect_create_cond_for_align_checks): Likewise.
8343 * tree-vect-stmts.c (get_negative_load_store_type): Likewise.
8344 (vectorizable_load): Remove duplicate leftover from merge
8347 2021-10-26 Roger Sayle <roger@nextmovesoftware.com>
8349 * config/i386/i386-expand.c (ix86_expand_v1ti_shift): New helper
8350 function to expand V1TI mode logical shifts by integer constants.
8351 (ix86_expand_v1ti_rotate): New helper function to expand V1TI
8352 mode rotations by integer constants.
8353 * config/i386/i386-protos.h (ix86_expand_v1ti_shift,
8354 ix86_expand_v1ti_rotate): Prototype new functions here.
8355 * config/i386/sse.md (ashlv1ti3, lshrv1ti3, rotlv1ti3, rotrv1ti3):
8356 New TARGET_SSE2 expanders to implement V1TI shifts and rotations.
8358 2021-10-26 Aldy Hernandez <aldyh@redhat.com>
8360 * tree-ssa-threadbackward.c (back_threader::maybe_register_path):
8361 Avoid threading circular paths.
8363 2021-10-26 Aldy Hernandez <aldyh@redhat.com>
8365 * tree-ssa-threadbackward.c (back_threader::resolve_phi):
8366 Attempt to resolve all incoming paths to a PHI.
8367 (back_threader::resolve_def): Always return true for PHIs.
8369 2021-10-26 Aldy Hernandez <aldyh@redhat.com>
8371 * tree-ssa-threadbackward.c (back_threader::find_paths_to_names):
8372 Always try to resolve path without looking back.
8373 * tree-ssa-threadupdate.c (dump_jump_thread): Indidicate whether
8374 edge is a back edge.
8376 2021-10-26 Kewen Lin <linkw@linux.ibm.com>
8378 PR tree-optimization/102789
8379 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
8380 update inits of simd_lane_access.
8382 2021-10-25 Andrew MacLeod <amacleod@redhat.com>
8384 * tree-vrp.c (vrp_simplify_cond_using_ranges): Add return type and
8385 move to vr-values.c.
8386 (simplify_casted_conds): Move to vrp_folder class.
8387 (execute_vrp): Call via vrp_folder now.
8388 * vr-values.c (simplify_cond_using_ranges_1): Call simplify_casted_cond.
8389 (simplify_using_ranges::simplify_casted_cond): Relocate from tree-vrp.c.
8390 * vr-values.h (simplify_casted_cond): Add prototype.
8392 2021-10-25 Andrew MacLeod <amacleod@redhat.com>
8394 * tree-vrp.c (rvrp_folder::fold_stmt): If simplification fails, try
8397 2021-10-25 Paul A. Clarke <pc@us.ibm.com>
8399 * config/rs6000/smmintrin.h (_mm_testz_si128): Add "extern" to
8401 (_mm_testc_si128): Likewise.
8402 (_mm_testnzc_si128): Likewise.
8403 (_mm_blend_ps): Likewise.
8404 (_mm_blendv_ps): Likewise.
8405 (_mm_blend_pd): Likewise.
8406 (_mm_blendv_pd): Likewise.
8407 (_mm_ceil_pd): Likewise.
8408 (_mm_ceil_sd): Likewise.
8409 (_mm_ceil_ps): Likewise.
8410 (_mm_ceil_ss): Likewise.
8411 (_mm_floor_pd): Likewise.
8412 (_mm_floor_sd): Likewise.
8413 (_mm_floor_ps): Likewise.
8414 (_mm_floor_ss): Likewise.
8415 (_mm_minpos_epu16): Likewise.
8416 (_mm_mul_epi32): Likewise.
8417 (_mm_cvtepi8_epi16): Likewise.
8418 (_mm_packus_epi32): Likewise.
8419 (_mm_cmpgt_epi64): Likewise.
8421 2021-10-25 Roger Sayle <roger@nextmovesoftware.com>
8423 * simplify-rtx.c (simplify_binary_operation_1) [SS_ASHIFT]: Simplify
8424 shifts of the mode's smin_value and smax_value when the bit count
8425 operand doesn't have side-effects.
8426 [US_ASHIFT]: Likewise, simplify shifts of the mode's umax_value
8427 when the bit count operand doesn't have side-effects.
8428 (simplify_const_binary_operation) [SS_ASHIFT, US_ASHIFT]: Perform
8429 compile-time evaluation of saturating left shifts with constant
8432 2021-10-25 Andrew MacLeod <amacleod@redhat.com>
8434 * gimple-range.cc (gimple_ranger::export_global_ranges): Remove check
8437 2021-10-25 Andrew MacLeod <amacleod@redhat.com>
8439 * flag-types.h (enum ranger_debug): Adjust values.
8440 * params.opt (ranger_debug): Ditto.
8442 2021-10-25 Martin Jambor <mjambor@suse.cz>
8444 PR tree-optimization/102886
8445 * tree-sra.c (totally_scalarize_subtree): Fix the out of
8448 2021-10-25 Andrew Pinski <apinski@marvell.com>
8450 * tree-ssa-dce.c (simple_dce_from_worklist):
8451 Check stmt_unremovable_because_of_non_call_eh_p also
8452 before removing the statement.
8454 2021-10-25 Richard Biener <rguenther@suse.de>
8456 PR tree-optimization/102905
8457 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment):
8458 Use vect_supportable_dr_alignment again to determine whether
8459 an access is supported when not aligned.
8461 2021-10-25 Kito Cheng <kito.cheng@sifive.com>
8463 * config/riscv/riscv.c (riscv_rtx_costs): Handle cost model
8466 2021-10-25 Jim Wilson <jimw@sifive.com>
8467 Kito Cheng <kito.cheng@sifive.com>
8468 Jia-Wei Chen <jiawei@iscas.ac.cn>
8469 Shi-Hua Liao <shihua@iscas.ac.cn>
8471 * config/riscv/bitmanip.md (shiftm1): New.
8472 (*bset<mode>): Ditto.
8473 (*bset<mode>_mask): Ditto.
8474 (*bset<mode>_1): Ditto.
8475 (*bset<mode>_1_mask): Ditto.
8476 (*bseti<mode>): Ditto.
8477 (*bclr<mode>): Ditto.
8478 (*bclri<mode>): Ditto.
8479 (*binv<mode>): Ditto.
8480 (*binvi<mode>): Ditto.
8481 (*bext<mode>): Ditto.
8483 * config/riscv/predicates.md (splittable_const_int_operand):
8485 (single_bit_mask_operand): New.
8486 (not_single_bit_mask_operand): Ditto.
8487 (const31_operand): Ditto.
8488 (const63_operand): Ditto.
8489 * config/riscv/riscv.c (riscv_build_integer_1): Handle bseti.
8490 (riscv_output_move): Ditto.
8491 (riscv_print_operand): Handle new operand type: T and S.
8492 * config/riscv/riscv.h (SINGLE_BIT_MASK_OPERAND): New.
8494 2021-10-25 Jim Wilson <jimw@sifive.com>
8496 * config/riscv/riscv.c (riscv_build_integer_1): Build integer
8499 2021-10-25 Kito Cheng <kito.cheng@sifive.com>
8501 * config/riscv/riscv.c (riscv_extend_cost): Handle cost model
8503 (riscv_rtx_costs): Ditto.
8505 2021-10-25 Jim Wilson <jimw@sifive.com>
8506 Kito Cheng <kito.cheng@sifive.com>
8507 Jia-Wei Chen <jiawei@iscas.ac.cn>
8509 * config/riscv/bitmanip.md (bitmanip_bitwise): New.
8510 (bitmanip_minmax): New.
8511 (clz_ctz_pcnt): New.
8512 (bitmanip_optab): New.
8513 (bitmanip_insn): New.
8514 (*<optab>_not<mode>): New.
8515 (*xor_not<mode>): New.
8516 (<bitmanip_optab>si2): New.
8517 (*<bitmanip_optab>disi2): New.
8518 (<bitmanip_optab>di2): New.
8519 (*zero_extendhi<GPR:mode>2_bitmanip): New.
8520 (*extend<SHORT:mode><SUPERQI:mode>2_zbb): New.
8521 (*zero_extendhi<GPR:mode>2_zbb): New.
8524 (rotrsi3_sext): New.
8527 (rotlsi3_sext): New.
8528 (bswap<mode>2): New.
8529 (<bitmanip_optab><mode>3): New.
8530 * config/riscv/riscv.md (type): Add rotate.
8531 (zero_extendhi<GPR:mode>2): Change to define_expand pattern.
8532 (*zero_extendhi<GPR:mode>2): New.
8533 (extend<SHORT:mode><SUPERQI:mode>2): Change to define_expand pattern.
8534 (*extend<SHORT:mode><SUPERQI:mode>2): New.
8536 2021-10-25 Kito Cheng <kito.cheng@sifive.com>
8538 * config/riscv/riscv.c (riscv_extend_cost): Handle cost model
8540 (riscv_rtx_costs): Ditto.
8542 2021-10-25 Jim Wilson <jimw@sifive.com>
8543 Kito Cheng <kito.cheng@sifive.com>
8544 Jia-Wei Chen <jiawei@iscas.ac.cn>
8546 * config/riscv/bitmanip.md (*zero_extendsidi2_bitmanip): New.
8551 (riscv_rtx_costs): Ditto.
8552 * config/riscv/riscv.md: Include bitmanip.md
8553 (type): Add bitmanip bype.
8554 (zero_extendsidi2): Change to define_expand pattern.
8555 (*zero_extendsidi2_internal): New.
8556 (zero_extendsidi2_shifted): Disable for ZBA.
8558 2021-10-25 Kito Cheng <kito.cheng@sifive.com>
8560 * common/config/riscv/riscv-common.c (riscv_ext_version_table):
8561 Add zba, zbb, zbc and zbs.
8562 (riscv_ext_flag_table): Ditto.
8563 * config/riscv/riscv-opts.h (MASK_ZBA): New.
8567 (TARGET_ZBA): Ditto.
8568 (TARGET_ZBB): Ditto.
8569 (TARGET_ZBC): Ditto.
8570 (TARGET_ZBS): Ditto.
8571 * config/riscv/riscv.opt (riscv_zb_subext): New.
8573 2021-10-25 liuhongt <hongtao.liu@intel.com>
8576 * match.pd: Simplify (_Float16) sqrtf((float) a) to .SQRT(a)
8577 when direct_internal_fn_supported_p, similar for sqrt/sqrtl.
8579 2021-10-25 Richard Biener <rguenther@suse.de>
8581 PR tree-optimization/102920
8582 * tree-ssa-sccvn.h (expressions_equal_p): Add argument
8583 controlling VN_TOP matching behavior.
8584 * tree-ssa-sccvn.c (expressions_equal_p): Likewise.
8585 (vn_phi_eq): Do not optimistically match VN_TOP.
8587 2021-10-25 konglin1 <lingling.kong@intel.com>
8589 * config/i386/sse.md (fma_<mode>_fadd_fmul): Add new
8590 define_insn_and_split.
8591 (fma_<mode>_fadd_fcmul):Likewise
8592 (fma_<complexopname>_<mode>_fma_zero):Likewise
8594 2021-10-24 John David Anglin <danglin@gcc.gnu.org>
8596 * config/pa/pa-d.c (pa_d_handle_target_float_abi): Don't check
8597 TARGET_DISABLE_FPREGS.
8598 * config/pa/pa.c (fix_range): Use MASK_SOFT_FLOAT instead of
8599 MASK_DISABLE_FPREGS.
8600 (hppa_rtx_costs): Don't check TARGET_DISABLE_FPREGS. Adjust
8601 cost of hardware integer multiplication.
8602 (pa_conditional_register_usage): Don't check TARGET_DISABLE_FPREGS.
8603 * config/pa/pa.h (INT14_OK_STRICT): Likewise.
8604 * config/pa/pa.md: Don't check TARGET_DISABLE_FPREGS. Check
8605 TARGET_SOFT_FLOAT in patterns that use xmpyu instruction.
8606 * config/pa/pa.opt (mdisable-fpregs): Change target mask to
8607 SOFT_FLOAT. Revise comment.
8608 (msoft-float): New option.
8610 2021-10-24 John David Anglin <danglin@gcc.gnu.org>
8612 * config/pa/pa.md: Don't use 'G' constraint in integer move patterns.
8614 2021-10-24 Gerald Pfeifer <gerald@pfeifer.com>
8616 * doc/install.texi (Specific): Remove obsolete details
8617 around GNU/Linux on Itanium.
8618 (Specific): Remove reference to Windows for Itanium.
8620 2021-10-23 Bernhard Reutner-Fischer <aldot@gcc.gnu.org>
8622 * config/i386/x86-tune-sched-bd.c (dispatch_group): Commentary
8625 2021-10-23 Jan Hubicka <hubicka@ucw.cz>
8627 * tree-ssa-structalias.c (compute_points_to_sets): Cleanup.
8629 2021-10-23 Roger Sayle <roger@nextmovesoftware.com>
8631 * config/i386/sse.md (<any_logic>v1ti3): New define_insn to
8632 implement V1TImode AND, IOR and XOR on TARGET_SSE2 (and above).
8633 (one_cmplv1ti2): New define expand.
8635 2021-10-22 Eric Gallager <egallager@gcc.gnu.org>
8638 * Makefile.in: Handle dvidir and install-dvi target.
8639 * configure: Regenerate.
8640 * configure.ac: Add install-dvi to target_list.
8642 2021-10-22 Gerald Pfeifer <gerald@pfeifer.com>
8644 * doc/install.texi (Binaries): Convert mingw-w64.org to https.
8647 2021-10-22 Richard Biener <rguenther@suse.de>
8649 PR tree-optimization/102893
8650 * tree-ssa-dce.c (find_obviously_necessary_stmts): Fix the
8651 test for an exit edge.
8653 2021-10-22 Aldy Hernandez <aldyh@redhat.com>
8654 Andrew MacLeod <amacleod@redhat.com>
8656 * gimple-range-path.cc (path_range_query::compute_phi_relations):
8657 Kill any global relations we may know before registering a new
8659 * value-relation.cc (path_oracle::killing_def): New.
8660 * value-relation.h (path_oracle::killing_def): New.
8662 2021-10-22 Richard Biener <rguenther@suse.de>
8665 * tree-ssa-sccvn.c (vn_phi_insert): For undefined SSA args
8667 (vn_phi_lookup): Likewise.
8669 2021-10-21 H.J. Lu <hjl.tools@gmail.com>
8672 * doc/invoke.texi: Document -fcf-protection requires i686 or
8675 2021-10-21 Eric Botcazou <ebotcazou@adacore.com>
8677 PR middle-end/102764
8678 * cfgexpand.c (expand_gimple_basic_block): Robustify latest change.
8680 2021-10-21 Jonathan Wright <jonathan.wright@arm.com>
8682 * config/aarch64/arm_neon.h (__STRUCTN): Delete function
8683 macro and all invocations.
8685 2021-10-21 Andrew MacLeod <amacleod@redhat.com>
8687 * doc/invoke.texi (ranger-debug): Document.
8688 * flag-types.h (enum ranger_debug): New.
8689 (enum evrp_mode): Remove debug values.
8690 * gimple-range-cache.cc (DEBUG_RANGE_CACHE): Use new debug flag.
8691 * gimple-range-gori.cc (gori_compute::gori_compute): Ditto.
8692 * gimple-range.cc (gimple_ranger::gimple_ranger): Ditto.
8693 * gimple-ssa-evrp.c (hybrid_folder::choose_value): Ditto.
8694 (execute_early_vrp): Use evrp-mode directly.
8695 * params.opt (enum evrp_mode): Remove debug values.
8696 (ranger-debug): New.
8697 (ranger-logical-depth): Relocate to be in alphabetical order.
8699 2021-10-21 Andrew MacLeod <amacleod@redhat.com>
8701 * doc/invoke.texi: (vrp1-mode, vrp2-mode): Document.
8702 * flag-types.h: (enum vrp_mode): New.
8703 * params.opt: (vrp1-mode, vrp2-mode): New.
8704 * tree-vrp.c (vrp_pass_num): New.
8705 (pass_vrp::pass_vrp): Set pass number.
8706 (pass_vrp::execute): Choose which VRP mode to execute.
8708 2021-10-21 Andrew MacLeod <amacleod@redhat.com>
8710 * gimple-ssa-evrp.c (class rvrp_folder): Move to tree-vrp.c.
8711 (execute_early_vrp): For ranger only mode, invoke ranger_vrp.
8712 * tree-vrp.c (class rvrp_folder): Relocate here.
8713 (execute_ranger_vrp): New.
8714 * tree-vrp.h (execute_ranger_vrp): Export.
8716 2021-10-21 Martin Liska <mliska@suse.cz>
8720 * opts.c (finish_options): Process flag_var_tracking* options
8721 here as they can be adjusted by optimize attribute.
8722 Process also flag_syntax_only and flag_gtoggle.
8723 * toplev.c (process_options): Remove it here.
8724 * common.opt: Make debug_nonbind_markers_p as PerFunction
8725 attribute as it depends on optimization level.
8727 2021-10-21 Martin Jambor <mjambor@suse.cz>
8729 PR tree-optimization/102505
8730 * tree-sra.c (totally_scalarize_subtree): Check that the
8731 encountered field fits within the acces we would like to put it
8734 2021-10-21 Aldy Hernandez <aldyh@redhat.com>
8736 * tree-ssa-threadbackward.c
8737 (back_threader::maybe_register_path): Remove circular paths check.
8739 2021-10-21 Richard Biener <rguenther@suse.de>
8741 * toplev.c (process_options): Move the initial debug_hooks
8743 (toplev::main): ... before the call of the post_options
8746 2021-10-21 Richard Biener <rguenther@suse.de>
8748 PR tree-optimization/102847
8749 * tree-vect-stmts.c (vect_model_load_cost): Add the scalar
8750 load cost in the prologue for VMAT_INVARIANT.
8752 2021-10-21 Richard Biener <rguenther@suse.de>
8754 PR tree-optimization/102847
8755 * tree-vect-stmts.c (vect_model_load_cost): Explicitely
8756 handle VMAT_INVARIANT as a splat in the prologue.
8758 2021-10-21 Hongyu Wang <hongyu.wang@intel.com>
8761 * config/i386/i386.c (ix86_get_ssemov): Adjust HFmode vector
8762 move to use the same logic as HImode.
8764 2021-10-21 Richard Biener <rguenther@suse.de>
8766 * tree-vect-slp.c (vect_build_slp_tree_1): Remove
8767 superfluous gimple_call_nothrow_p check.
8769 2021-10-21 Andrew Pinski <apinski@marvell.com>
8771 * tree-cfg.c (maybe_remove_writeonly_store): Add dce_ssa_names argument.
8772 Mark the ssa-name of the rhs as one to be removed.
8773 (execute_fixup_cfg): Update call to maybe_remove_writeonly_store.
8774 Call simple_dce_from_worklist at the end to a simple dce.
8776 2021-10-21 Andrew Pinski <apinski@marvell.com>
8778 * tree-cfg.c (maybe_remove_writeonly_store): New function
8779 factored out from ...
8780 (execute_fixup_cfg): Here. Call maybe_remove_writeonly_store.
8782 2021-10-21 Andrew Pinski <apinski@marvell.com>
8784 * tree-cfg.c (execute_fixup_cfg): Remove comment
8785 about standalone pass.
8787 2021-10-21 Andrew Pinski <apinski@marvell.com>
8789 * tree-cfg.c (execute_fixup_cfg): Output when the statement
8790 is removed when it is a write only var.
8792 2021-10-21 Aldy Hernandez <aldyh@redhat.com>
8794 * tree-ssa-threadbackward.c (back_threader::maybe_register_path):
8795 Avoid threading circular paths.
8797 2021-10-20 Alex Coplan <alex.coplan@arm.com>
8799 * calls.c (initialize_argument_information): Remove some dead
8800 code, remove handling for function_arg returning const_int.
8801 * doc/tm.texi: Delete documentation for unused target hooks.
8802 * doc/tm.texi.in: Likewise.
8803 * target.def (load_bounds_for_arg): Delete.
8804 (store_bounds_for_arg): Delete.
8805 (load_returned_bounds): Delete.
8806 (store_returned_bounds): Delete.
8807 * targhooks.c (default_load_bounds_for_arg): Delete.
8808 (default_store_bounds_for_arg): Delete.
8809 (default_load_returned_bounds): Delete.
8810 (default_store_returned_bounds): Delete.
8811 * targhooks.h (default_load_bounds_for_arg): Delete.
8812 (default_store_bounds_for_arg): Delete.
8813 (default_load_returned_bounds): Delete.
8814 (default_store_returned_bounds): Delete.
8816 2021-10-20 Jonathan Wakely <jwakely@redhat.com>
8818 * doc/extend.texi (Basic Asm): Clarify that asm is not an
8820 * doc/invoke.texi (-fno-asm): Fix description for C++.
8822 2021-10-20 Jonathan Wakely <jwakely@redhat.com>
8824 * doc/install.texi: Remove link to old.html
8826 2021-10-20 Tamar Christina <tamar.christina@arm.com>
8828 * config/aarch64/aarch64-simd.md (*aarch64_cmtst_same_<mode>): New.
8830 2021-10-20 Tamar Christina <tamar.christina@arm.com>
8832 * config/aarch64/aarch64-simd.md (*aarch64_narrow_trunc<mode>): New.
8834 2021-10-20 Tamar Christina <tamar.christina@arm.com>
8836 * config/aarch64/aarch64-simd.md (aarch64_simd_ashr<mode>): Add case cmp
8838 * config/aarch64/constraints.md (D1): New.
8840 2021-10-20 Tamar Christina <tamar.christina@arm.com>
8842 * config/aarch64/aarch64-simd.md
8843 (*aarch64_<srn_op>topbits_shuffle<mode>_le): New.
8844 (*aarch64_topbits_shuffle<mode>_le): New.
8845 (*aarch64_<srn_op>topbits_shuffle<mode>_be): New.
8846 (*aarch64_topbits_shuffle<mode>_be): New.
8847 * config/aarch64/predicates.md
8848 (aarch64_simd_shift_imm_vec_exact_top): New.
8850 2021-10-20 Tamar Christina <tamar.christina@arm.com>
8852 * config/aarch64/aarch64-simd.md (*aarch64_<srn_op>shrn<mode>_vect,
8853 *aarch64_<srn_op>shrn<mode>2_vect_le,
8854 *aarch64_<srn_op>shrn<mode>2_vect_be): New.
8855 * config/aarch64/iterators.md (srn_op): New.
8857 2021-10-20 Chung-Lin Tang <cltang@codesourcery.com>
8859 * omp-low.c (omp_copy_decl_2): For !ctx, use record_vars to add new copy
8861 (scan_sharing_clauses): Place copy of OMP_CLAUSE_IN_REDUCTION decl in
8862 ctx->outer instead of ctx.
8864 2021-10-20 Martin Liska <mliska@suse.cz>
8867 2021-10-19 Martin Liska <mliska@suse.cz>
8870 * config/i386/i386-options.c (ix86_valid_target_attribute_inner_p): Strip whitespaces.
8871 * system.h (strip_whilespaces): New function.
8873 2021-10-20 Martin Liska <mliska@suse.cz>
8876 2021-10-19 Martin Liska <mliska@suse.cz>
8879 * config/aarch64/aarch64.c (aarch64_process_one_target_attr):
8882 2021-10-20 Wilco Dijkstra <wdijkstr@arm.com>
8884 * config/aarch64/aarch64.c (aarch64_case_values_threshold):
8885 Change to 8 with -Os, 11 otherwise.
8887 2021-10-20 Wilco Dijkstra <wdijkstr@arm.com>
8889 * config/aarch64/aarch64.c (neoversev1_tunings):
8890 Enable AARCH64_EXTRA_TUNE_CHEAP_SHIFT_EXTEND.
8891 (neoversen2_tunings): Likewise.
8893 2021-10-20 Wilco Dijkstra <wdijkstr@arm.com>
8896 * config/aarch64/aarch64.md (frint_pattern): Update comment.
8897 * config/aarch64/aarch64-simd-builtins.def: Change frintn to roundeven.
8898 * config/aarch64/arm_fp16.h: Change frintn to roundeven.
8899 * config/aarch64/arm_neon.h: Likewise.
8900 * config/aarch64/iterators.md (frint_pattern): Use roundeven for FRINTN.
8902 2021-10-20 Martin Liska <mliska@suse.cz>
8904 * config/arm/arm.c (arm_unwind_emit_sequence): Do not declare
8905 already declared global variable.
8906 (arm_unwind_emit_set): Use out_file as function argument.
8907 (arm_unwind_emit): Likewise.
8908 * config/darwin.c (machopic_output_data_section_indirection): Likewise.
8909 (machopic_output_stub_indirection): Likewise.
8910 (machopic_output_indirection): Likewise.
8911 (machopic_finish): Likewise.
8912 * config/i386/i386.c (ix86_asm_output_function_label): Likewise.
8913 * config/i386/winnt.c (i386_pe_seh_unwind_emit): Likewise.
8914 * config/ia64/ia64.c (process_epilogue): Likewise.
8915 (process_cfa_adjust_cfa): Likewise.
8916 (process_cfa_register): Likewise.
8917 (process_cfa_offset): Likewise.
8918 (ia64_asm_unwind_emit): Likewise.
8919 * config/s390/s390.c (s390_asm_output_function_label): Likewise.
8921 2021-10-20 Andre Simoes Dias Vieira <andre.simoesdiasvieira@arm.com>
8923 * config/aarch64/aarch64-builtins.c
8924 (aarch64_general_gimple_fold_builtin):
8925 lower vld1 and vst1 variants of the neon builtins
8926 * config/aarch64/aarch64-protos.h:
8927 (aarch64_general_gimple_fold_builtin): Add gsi parameter.
8928 * config/aarch64/aarch64.c (aarch64_general_gimple_fold_builtin):
8931 2021-10-20 Andre Simoes Dias Vieira <andre.simoesdiasvieira@arm.com>
8933 * match.pd: Generate IFN_TRUNC.
8935 2021-10-20 Richard Biener <rguenther@suse.de>
8937 PR tree-optimization/102853
8938 * tree-data-ref.c (split_constant_offset_1): Bail out
8939 immediately if the expression traps on overflow.
8941 2021-10-20 Aldy Hernandez <aldyh@redhat.com>
8943 * tree-ssa-threadbackward.c (back_threader::~back_threader): Remove.
8945 2021-10-20 Aldy Hernandez <aldyh@redhat.com>
8947 * tree-ssa-threadupdate.c (back_jt_path_registry::adjust_paths_after_duplication):
8948 Remove superflous debugging message.
8949 (back_jt_path_registry::duplicate_thread_path): Same.
8951 2021-10-20 Aldy Hernandez <aldyh@redhat.com>
8953 * tree-ssa-threadbackward.c (back_threader_registry::back_threader_registry):
8955 (back_threader_registry::register_path): Remove m_threaded_paths.
8957 2021-10-20 Aldy Hernandez <aldyh@redhat.com>
8959 PR tree-optimization/102814
8960 * doc/invoke.texi: Document --param=max-fsm-thread-length.
8961 * params.opt: Add --param=max-fsm-thread-length.
8962 * tree-ssa-threadbackward.c
8963 (back_threader_profitability::profitable_path_p): Fail on paths
8964 longer than max-fsm-thread-length.
8966 2021-10-20 Eric Botcazou <ebotcazou@adacore.com>
8968 PR middle-end/102764
8969 * cfgexpand.c (expand_gimple_basic_block): Disregard a final debug
8970 statement to reset the current location for the outgoing edges.
8972 2021-10-20 Aldy Hernandez <aldyh@redhat.com>
8973 Richard Biener <rguenther@suse.de>
8975 * tree-ssa-threadupdate.c (cancel_thread): Dump threading reason
8976 on the same line as the threading cancellation.
8977 (jt_path_registry::cancel_invalid_paths): Avoid rotating loops.
8978 Avoid threading through loop headers where the path remains in the
8981 2021-10-20 Siddhesh Poyarekar <siddhesh@gotplt.org>
8983 * tree-object-size.c (unknown): Make into a function. Adjust
8985 (unknown_object_size): Simplify implementation.
8987 2021-10-20 Hongtao Liu <hongtao.liu@intel.com>
8988 Kewen Lin <linkw@linux.ibm.com>
8990 * doc/sourcebuild.texi (Effective-Target Keywords): Document
8991 vect_slp_v2qi_store, vect_slp_v4qi_store, vect_slp_v8qi_store,
8992 vect_slp_v16qi_store, vect_slp_v2hi_store,
8993 vect_slp_v4hi_store, vect_slp_v2si_store, vect_slp_v4si_store.
8995 2021-10-19 Jonathan Wakely <jwakely@redhat.com>
8997 * doc/extend.texi (Basic PowerPC Built-in Functions): Fix typo.
8999 2021-10-19 Paul A. Clarke <pc@us.ibm.com>
9003 * config/rs6000/emmintrin.h: Guard POWER8 intrinsics.
9004 * config/rs6000/pmmintrin.h: Same.
9005 * config/rs6000/smmintrin.h: Same.
9006 * config/rs6000/tmmintrin.h: Same.
9008 2021-10-19 Paul A. Clarke <pc@us.ibm.com>
9010 * config.gcc (extra_headers): Add nmmintrin.h.
9012 2021-10-19 Richard Biener <rguenther@suse.de>
9014 * tree-vectorizer.h (vect_supportable_dr_alignment): Add
9015 misalignment parameter.
9016 * tree-vect-data-refs.c (vect_get_peeling_costs_all_drs):
9017 Do not change DR_MISALIGNMENT in place, instead pass the
9018 adjusted misalignment to vect_supportable_dr_alignment.
9019 (vect_peeling_supportable): Likewise.
9020 (vect_peeling_hash_get_lowest_cost): Adjust.
9021 (vect_enhance_data_refs_alignment): Likewise.
9022 (vect_vfa_access_size): Likewise.
9023 (vect_supportable_dr_alignment): Add misalignment
9024 parameter and simplify.
9025 * tree-vect-stmts.c (get_negative_load_store_type): Adjust.
9026 (get_group_load_store_type): Likewise.
9027 (get_load_store_type): Likewise.
9029 2021-10-19 Clément Chigot <clement.chigot@atos.net>
9031 * config/rs6000/rs6000.c (rs6000_xcoff_file_end): Move
9032 __tls_get_addr reference to .text csect.
9034 2021-10-19 Martin Liska <mliska@suse.cz>
9037 * config/aarch64/aarch64.c (aarch64_process_one_target_attr):
9040 2021-10-19 Richard Biener <rguenther@suse.de>
9042 * tree-vectorizer.h (vect_get_store_cost): Adjust signature.
9043 (vect_get_load_cost): Likewise.
9044 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
9045 alignment support scheme and misalignment as arguments
9047 (vect_get_peeling_costs_all_drs): Compute that info here
9048 and note that we shouldn't need to.
9049 * tree-vect-stmts.c (vect_model_store_cost): Get
9050 alignment support scheme and misalignment as arguments.
9051 (vect_get_store_cost): Likewise.
9052 (vect_model_load_cost): Likewise.
9053 (vect_get_load_cost): Likewise.
9054 (vectorizable_store): Pass down alignment support scheme
9055 and misalignment to costing.
9056 (vectorizable_load): Likewise.
9058 2021-10-19 Richard Biener <rguenther@suse.de>
9060 * tree-vect-stmts.c (get_negative_load_store_type): Add
9061 offset output parameter and initialize it.
9062 (get_group_load_store_type): Likewise.
9063 (get_load_store_type): Likewise.
9064 (vectorizable_store): Use offset as computed by
9065 get_load_store_type.
9066 (vectorizable_load): Likewise.
9068 2021-10-19 Richard Biener <rguenther@suse.de>
9070 PR tree-optimization/102827
9071 * tree-if-conv.c (predicate_statements): Add pe parameter
9072 and use that edge to insert invariant stmts on.
9073 (combine_blocks): Pass through pe.
9074 (tree_if_conversion): Compute the edge to insert invariant
9075 stmts on and pass it along.
9077 2021-10-19 Roger Sayle <roger@nextmovesoftware.com>
9080 * config/bfin/bfin.md (addsubv2hi3, subaddv2hi3, ssaddsubv2hi3,
9081 sssubaddv2hi3): Swap the order of operators in vec_concat.
9083 2021-10-19 Xionghu Luo <luoxhu@linux.ibm.com>
9085 * config/rs6000/altivec.md (*altivec_vmrghb_internal): Delete.
9086 (altivec_vmrghb_direct): New.
9087 (*altivec_vmrghh_internal): Delete.
9088 (altivec_vmrghh_direct): New.
9089 (*altivec_vmrghw_internal): Delete.
9090 (altivec_vmrghw_direct_<mode>): New.
9091 (altivec_vmrghw_direct): Delete.
9092 (*altivec_vmrglb_internal): Delete.
9093 (altivec_vmrglb_direct): New.
9094 (*altivec_vmrglh_internal): Delete.
9095 (altivec_vmrglh_direct): New.
9096 (*altivec_vmrglw_internal): Delete.
9097 (altivec_vmrglw_direct_<mode>): New.
9098 (altivec_vmrglw_direct): Delete.
9099 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Adjust.
9100 * config/rs6000/rs6000.c (altivec_expand_vec_perm_const):
9102 * config/rs6000/vsx.md (vsx_xxmrghw_<mode>): Adjust.
9103 (vsx_xxmrglw_<mode>): Adjust.
9105 2021-10-19 Aldy Hernandez <aldyh@redhat.com>
9107 * passes.def: Change threading comment before pass_ccp pass.
9109 2021-10-19 Haochen Gui <guihaoc@gcc.gnu.org>
9111 * config/rs6000/rs6000-call.c (altivec_expand_lxvr_builtin):
9112 Modify the expansion for sign extension. All extensions are done
9113 within VSX registers.
9115 2021-10-19 Richard Biener <rguenther@suse.de>
9117 * tree-vect-stmts.c (get_group_load_store_type): Add
9118 misalignment output parameter and initialize it.
9119 (get_group_load_store_type): Likewise.
9120 (vectorizable_store): Remove now redundant queries.
9121 (vectorizable_load): Likewise.
9123 2021-10-19 Richard Biener <rguenther@suse.de>
9125 * tree-vectorizer.h (vect_supportable_dr_alignment): Remove
9126 check_aligned argument.
9127 * tree-vect-data-refs.c (vect_supportable_dr_alignment):
9129 (vect_peeling_hash_insert): Add supportable_if_not_aligned
9130 argument and do not call vect_supportable_dr_alignment here.
9131 (vect_peeling_supportable): Adjust.
9132 (vect_enhance_data_refs_alignment): Compute whether the
9133 access is supported with different alignment here and
9134 pass that down to vect_peeling_hash_insert.
9135 (vect_vfa_access_size): Adjust.
9136 * tree-vect-stmts.c (vect_get_store_cost): Likewise.
9137 (vect_get_load_cost): Likewise.
9138 (get_negative_load_store_type): Likewise.
9139 (get_group_load_store_type): Likewise.
9140 (get_load_store_type): Likewise.
9142 2021-10-19 Martin Liska <mliska@suse.cz>
9145 * config/i386/i386-options.c (ix86_valid_target_attribute_inner_p): Strip whitespaces.
9146 * system.h (strip_whilespaces): New function.
9148 2021-10-19 dianhong xu <dianhong.xu@intel.com>
9150 * config/i386/avx512fp16intrin.h:
9151 (_mm512_set1_pch): New intrinsic.
9152 * config/i386/avx512fp16vlintrin.h:
9153 (_mm256_set1_pch): New intrinsic.
9154 (_mm_set1_pch): Ditto.
9156 2021-10-18 Andrew MacLeod <amacleod@redhat.com>
9158 PR tree-optimization/102796
9159 * gimple-range.cc (gimple_ranger::range_on_edge): Process EH edges
9160 normally. Return get_tree_range for non gimple_range_ssa_p names.
9161 (gimple_ranger::range_of_stmt): Use get_tree_range for non
9162 gimple_range_ssa_p names.
9164 2021-10-18 Uroš Bizjak <ubizjak@gmail.com>
9167 * config/i386/i386.c (ix86_print_operand_address):
9168 Error out for non-address_operand asm operands.
9170 2021-10-18 Richard Biener <rguenther@suse.de>
9172 * tree-vect-data-refs.c (vect_peeling_hash_insert): Do
9173 not auto-convert dr_alignment_support to bool.
9174 (vect_peeling_supportable): Likewise.
9175 (vect_enhance_data_refs_alignment): Likewise.
9176 (vect_supportable_dr_alignment): Commonize read/write case.
9177 * tree-vect-stmts.c (vect_get_store_cost): Use
9178 dr_alignment_support, not int, for the vect_supportable_dr_alignment
9180 (vect_get_load_cost): Likewise.
9182 2021-10-18 Siddhesh Poyarekar <siddhesh@gotplt.org>
9184 * tree-object-size.c (object_sizes_execute): Consolidate LHS
9185 null check and do it early.
9187 2021-10-18 Richard Biener <rguenther@suse.de>
9189 * tree-vect-stmts.c (vectorizable_store): Use the
9190 computed alignment scheme instead of querying
9193 2021-10-18 Richard Biener <rguenther@suse.de>
9195 * tree-vect-stmts.c (vectorizable_store): Do not recompute
9196 alignment scheme already determined by get_load_store_type.
9198 2021-10-18 Aldy Hernandez <aldyh@redhat.com>
9200 * tree-ssa-threadbackward.c (class pass_thread_jumps_full):
9201 Clone corresponding pass.
9203 2021-10-18 Roger Sayle <roger@nextmovesoftware.com>
9205 * combine.c (recog_for_combine): For an unrecognized move/set of
9206 a constant, try force_const_mem to place it in the constant pool.
9207 * cse.c (constant_pool_entries_cost, constant_pool_entries_regcost):
9208 Delete global variables (that are no longer assigned a cost value).
9209 (cse_insn): Simplify logic for deciding whether to place a folded
9210 constant in the constant pool using force_const_mem.
9211 (cse_main): Remove zero initialization of constant_pool_entries_cost
9212 and constant_pool_entries_regcost.
9213 * config/i386/i386.c (ix86_rtx_costs): Make memory accesses
9214 fractionally more expensive, when optimizing for speed.
9216 2021-10-18 Martin Liska <mliska@suse.cz>
9218 PR gcov-profile/102746
9219 PR gcov-profile/102747
9220 * gcov.c (main): Return return_code.
9221 (output_gcov_file): Mark return_code when error happens.
9222 (generate_results): Likewise.
9223 (read_graph_file): Likewise.
9224 (read_count_file): Likewise.
9226 2021-10-18 Roger Sayle <roger@nextmovesoftware.com>
9228 * config/bfin/bfin.md (define_constants): Remove UNSPEC_ONES.
9229 (define_insn "ones"): Replace UNSPEC_ONES with a truncate of
9230 a popcount, allowing compile-time evaluation/simplification.
9231 (popcountsi2, popcounthi2): New expanders using a "ones" insn.
9233 2021-10-18 Richard Biener <rguenther@suse.de>
9235 PR tree-optimization/102788
9236 * tree-vect-patterns.c (vect_init_pattern_stmt): Allow
9238 (vect_pattern_recog_1): Likewise.
9239 (vect_recog_bool_pattern): Continue matching the pattern
9240 even if we do not have a vector type for a conversion
9243 2021-10-18 Roger Sayle <roger@nextmovesoftware.com>
9245 * simplify-rtx.c (simplify_const_unary_operation) [SS_NEG, SS_ABS]:
9246 Evalute SS_NEG and SS_ABS of a constant argument.
9248 2021-10-18 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
9251 * gimple-match-head.c (try_conditional_simplification): Add case for single operand.
9252 * internal-fn.def: Add entry for COND_NEG internal function.
9253 * internal-fn.c (FOR_EACH_CODE_MAPPING): Add entry for
9254 NEGATE_EXPR, COND_NEG mapping.
9255 * optabs.def: Add entry for cond_neg_optab.
9256 * match.pd (UNCOND_UNARY, COND_UNARY): New operator lists.
9257 (vec_cond COND (foo A) B) -> (IFN_COND_FOO COND A B): New pattern.
9258 (vec_cond COND B (foo A)) -> (IFN_COND_FOO ~COND A B): Likewise.
9260 2021-10-18 Aldy Hernandez <aldyh@redhat.com>
9262 * tree-ssa-strlen.c (class strlen_pass): Rename from
9264 (handle_builtin_stxncpy_strncat): Move to strlen_pass.
9265 (handle_assign): Same.
9266 (adjust_last_stmt): Same.
9267 (maybe_warn_overflow): Same.
9268 (handle_builtin_strlen): Same.
9269 (handle_builtin_strchr): Same.
9270 (handle_builtin_strcpy): Same.
9271 (handle_builtin_strncat): Same.
9272 (handle_builtin_stxncpy_strncat): Same.
9273 (handle_builtin_memcpy): Same.
9274 (handle_builtin_strcat): Same.
9275 (handle_alloc_call): Same.
9276 (handle_builtin_memset): Same.
9277 (handle_builtin_memcmp): Same.
9278 (get_len_or_size): Same.
9279 (strxcmp_eqz_result): Same.
9280 (handle_builtin_string_cmp): Same.
9281 (handle_pointer_plus): Same.
9282 (count_nonzero_bytes_addr): Same.
9283 (count_nonzero_bytes): Same.
9284 (handle_store): Same.
9285 (strlen_check_and_optimize_call): Same.
9286 (handle_integral_assign): Same.
9287 (check_and_optimize_stmt): Same.
9288 (printf_strlen_execute): Rename strlen_dom_walker to strlen_pass.
9290 2021-10-18 Richard Biener <rguenther@suse.de>
9292 PR tree-optimization/102798
9293 * tree-vect-data-refs.c (vect_create_addr_base_for_vector_ref):
9294 Only copy points-to info to newly generated SSA names.
9296 2021-10-18 Martin Liska <mliska@suse.cz>
9298 * dbgcnt.c (dbg_cnt_process_opt): Remove unused but set variable.
9299 * gcov.c (get_cycles_count): Likewise.
9300 * lto-compress.c (lto_compression_zlib): Likewise.
9301 (lto_uncompression_zlib): Likewise.
9302 * targhooks.c (default_pch_valid_p): Likewise.
9304 2021-10-17 Aldy Hernandez <aldyh@redhat.com>
9306 * tree-pass.h (make_pass_thread_jumps_full): New.
9307 * tree-ssa-threadbackward.c (pass_thread_jumps::gate): Inline.
9308 (try_thread_blocks): Add resolve and speed arguments.
9309 (pass_thread_jumps::execute): Inline.
9310 (do_early_thread_jumps): New.
9311 (do_thread_jumps): New.
9312 (make_pass_thread_jumps): Move.
9313 (pass_early_thread_jumps::gate): Inline.
9314 (pass_early_thread_jumps::execute): Inline.
9315 (class pass_thread_jumps_full): New.
9317 2021-10-16 Piotr Kubaj <pkubaj@FreeBSD.org>
9319 * configure.ac: Treat powerpc64*-*-freebsd* the same as
9320 powerpc64-*-freebsd*.
9321 * configure: Regenerate.
9323 2021-10-16 H.J. Lu <hjl.tools@gmail.com>
9325 * value-query.cc (get_ssa_name_ptr_info_nonnull): Change
9326 set_ptr_nonull to set_ptr_nonnull in comments.
9328 2021-10-16 Jan Hubicka <hubicka@ucw.cz>
9330 PR tree-optimization/102720
9331 * tree-ssa-structalias.c (compute_points_to_sets): Fix producing
9332 of call used and clobbered sets.
9334 2021-10-15 Iain Sandoe <iain@sandoe.co.uk>
9336 * config/darwin.h (LINK_COMMAND_SPEC_A): Update 'r' handling to
9337 skip gomp and itm when r or nodefaultlibs is given.
9338 (DSYMUTIL_SPEC): Do not call dsymutil for '-r' link lines.
9339 Update ordering of exclusions, remove duplicate 'v' addition
9340 (collect2 will add this from the main command line).
9342 2021-10-15 Iain Sandoe <iain@sandoe.co.uk>
9344 * config/darwin-driver.c (darwin_driver_init): Revise comments, handle
9345 filelist and framework options in specs instead of code.
9346 * config/darwin.h (SUBTARGET_DRIVER_SELF_SPECS): Update to handle link
9347 specs that are really driver ones.
9348 (DARWIN_CC1_SPEC): Likewise.
9349 (CPP_SPEC): Likewise.
9350 (SYSROOT_SPEC): Append space.
9351 (LINK_SYSROOT_SPEC): Remove most driver link specs.
9352 (STANDARD_STARTFILE_PREFIX_2): Update link-related specs.
9353 (STARTFILE_SPEC): Likewise.
9354 (ASM_MMACOSX_VERSION_MIN_SPEC): Fix line wrap.
9355 (ASM_SPEC): Update driver-related specs.
9356 (ASM_FINAL_SPEC): Likewise.
9357 * config/darwin.opt: Remove now unused option aliases.
9358 * config/i386/darwin.h (EXTRA_ASM_OPTS): Ensure space after opt.
9359 (ASM_SPEC): Update driver-related specs.
9361 2021-10-15 Roger Sayle <roger@nextmovesoftware.com>
9363 * config/i386/i386.c (ix86_hardreg_mov_ok): For vector modes,
9364 allow standard_sse_constant_p immediate constants.
9366 2021-10-15 Iain Sandoe <iain@sandoe.co.uk>
9368 * config.gcc: Add tm-dwarf2.h to tm_d-file.
9370 2021-10-15 Andrew MacLeod <amacleod@redhat.com>
9372 * gimple-range-fold.h (gimple_range_ssa_p): Don't process names
9373 that occur in abnormal phis.
9374 * gimple-range.cc (gimple_ranger::range_on_edge): Return false for
9375 abnormal and EH edges.
9376 * gimple-ssa-evrp.c (rvrp_folder::value_of_expr): Ditto.
9377 (rvrp_folder::value_on_edge): Ditto.
9378 (rvrp_folder::value_of_stmt): Ditto.
9379 (hybrid_folder::value_of_expr): Ditto for ranger queries.
9380 (hybrid_folder::value_on_edge): Ditto.
9381 (hybrid_folder::value_of_stmt): Ditto.
9382 * value-query.cc (gimple_range_global): Always return a range if
9383 the type is supported.
9385 2021-10-15 John David Anglin <danglin@gcc.gnu.org>
9387 * config/pa/pa.md: Consistently use "rG" constraint for copy
9388 instruction in move patterns.
9390 2021-10-15 Andrew Stubbs <ams@codesourcery.com>
9392 * config/gcn/gcn-hsa.h (S_FIJI): Set unconditionally.
9395 * config/gcn/gcn.c: Hard code SRAM ECC settings for old architectures.
9396 * config/gcn/mkoffload.c (ELFABIVERSION_AMDGPU_HSA): Rename to ...
9397 (ELFABIVERSION_AMDGPU_HSA_V3): ... this.
9398 (ELFABIVERSION_AMDGPU_HSA_V4): New.
9399 (SET_SRAM_ECC_UNSUPPORTED): New.
9400 (copy_early_debug_info): Create elf flags to match the other objects.
9401 (main): Just let the attribute flags pass through.
9403 2021-10-15 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
9405 * tree-loop-distribution.c (reduction_var_overflows_first):
9406 Pass the type of reduction_var as first argument as it is also
9407 done for the load type.
9408 (loop_distribution::transform_reduction_loop): Add missing
9409 TREE_TYPE while determining precission of reduction_var.
9411 2021-10-15 Richard Biener <rguenther@suse.de>
9413 * defaults.h (PREFERRED_DEBUGGING_TYPE): Choose DWARF2_DEBUG
9415 * toplev.c (process_options): Warn when STABS debugging is
9416 enabled but not the preferred format.
9417 * config/pa/som.h (PREFERRED_DEBUGGING_TYPE): Define to
9419 * config/pdp11/pdp11.h (PREFERRED_DEBUGGING_TYPE): Likewise.
9421 2021-10-15 Richard Biener <rguenther@suse.de>
9424 * tree-inline.c (copy_bb): Avoid underflowing nargs.
9426 2021-10-15 Hongyu Wang <hongyu.wang@intel.com>
9428 * config/i386/i386-expand.c (ix86_expand_vec_perm): Convert
9429 HFmode input operand to HImode.
9430 (ix86_vectorize_vec_perm_const): Likewise.
9431 * config/i386/sse.md (*avx512bw_permvar_truncv16siv16hi_1_hf):
9433 (*avx512f_permvar_truncv8siv8hi_1_hf):
9436 2021-10-15 Richard Biener <rguenther@suse.de>
9438 PR middle-end/102682
9439 * expmed.c (store_bit_field_1): Ensure a LHS subreg would
9440 not create a paradoxical subreg.
9442 2021-10-15 Hongyu Wang <hongyu.wang@intel.com>
9444 * config/i386/i386-expand.c (ix86_expand_vector_init):
9445 For half_vector concat for HFmode, handle them like HImode.
9447 2021-10-14 Aldy Hernandez <aldyh@redhat.com>
9449 * tree-ssa-threadbackward.c (class back_threader): Add m_resolve.
9450 (back_threader::back_threader): Same.
9451 (back_threader::resolve_phi): Try to solve without looking back if
9453 (back_threader::find_paths_to_names): Same.
9454 (try_thread_blocks): Pass resolve argument to back threader.
9455 (pass_early_thread_jumps::execute): Same.
9457 2021-10-14 Aldy Hernandez <aldyh@redhat.com>
9459 * doc/invoke.texi: Remove max-fsm-thread-length,
9460 max-fsm-thread-paths, and fsm-maximum-phi-arguments.
9462 * tree-ssa-threadbackward.c (back_threader::back_threader): Remove
9464 (back_threader_registry::back_threader_registry): Same.
9465 (back_threader_profitability::profitable_path_p): Remove
9466 param_max_fsm_thread-length.
9467 (back_threader_registry::register_path): Remove
9468 m_max_allowable_paths.
9470 2021-10-14 Aldy Hernandez <aldyh@redhat.com>
9472 * tree-ssa-threadbackward.c (class back_threader): Make m_imports
9474 (back_threader::~back_threader): Do not release m_path.
9476 2021-10-14 Andrew MacLeod <amacleod@redhat.com>
9478 PR tree-optimization/102738
9479 * vr-values.c (simplify_using_ranges::simplify): Handle RSHIFT_EXPR.
9481 2021-10-14 Kwok Cheung Yeung <kcy@codesourcery.com>
9483 * omp-general.c (omp_check_context_selector): Move from c-omp.c.
9484 (omp_mark_declare_variant): Move from c-omp.c.
9485 (omp_context_name_list_prop): Update for Fortran strings.
9486 * omp-general.h (omp_check_context_selector): New prototype.
9487 (omp_mark_declare_variant): New prototype.
9489 2021-10-14 Richard Sandiford <richard.sandiford@arm.com>
9491 * config/rs6000/rs6000.c (rs6000_density_test): Move early
9492 exit test further up the function.
9494 2021-10-14 Richard Sandiford <richard.sandiford@arm.com>
9496 * config/arm/arm.c (arm_add_stmt_cost): Delete.
9497 (TARGET_VECTORIZE_ADD_STMT_COST): Delete.
9499 2021-10-14 Martin Jambor <mjambor@suse.cz>
9501 * doc/invoke.texi (Optimize Options): Add entry for
9502 ipa-cp-recursive-freq-factor.
9504 2021-10-14 Tamar Christina <tamar.christina@arm.com>
9506 * match.pd: New rule.
9508 2021-10-14 Jan Hubicka <hubicka@ucw.cz>
9511 * ipa-modref-tree.h (modref_access_node::update2):
9512 Also check that parm_offset is unchanged.
9513 (modref_ref_node::insert_access): Fix updating of
9516 2021-10-14 Aldy Hernandez <aldyh@redhat.com>
9518 * tree-ssa-threadbackward.c (back_threader::resolve_phi): Add
9521 2021-10-14 Richard Biener <rguenther@suse.de>
9523 PR tree-optimization/102659
9524 * tree-if-conv.c (if_convertible_gimple_assign_stmt_p): Also
9525 rewrite pointer typed undefined overflow operations.
9526 (predicate_statements): Likewise. Make sure to emit invariant
9527 conversions in the preheader.
9528 * tree-vectorizer.c (vect_loop_vectorized_call): Look through
9529 non-empty preheaders.
9530 * tree-data-ref.c (dr_analyze_indices): Strip useless
9531 conversions to the MEM_REF base type.
9533 2021-10-14 Martin Liska <mliska@suse.cz>
9535 * common.opt: Stop using AUTODETECT_VALUE
9536 and use EnabledBy where possible.
9537 * opts.c: Enable OPT_fvar_tracking with optimize >= 1.
9538 * toplev.c (AUTODETECT_VALUE): Remove macro.
9539 (process_options): Simplify by using EnabledBy and
9540 OPT_fvar_tracking. Use OPTION_SET_P macro instead of
9543 2021-10-14 Jonathan Wright <jonathan.wright@arm.com>
9545 * config/aarch64/arm_neon.h (vld1_s8_x3): Use signed type for
9547 (vld1_s32_x3): Likewise.
9549 2021-10-14 Aldy Hernandez <aldyh@redhat.com>
9551 PR tree-optimization/102736
9552 PR tree-optimization/102736
9553 * gimple-range-path.cc (path_range_query::range_on_path_entry):
9554 Assert that the requested range is defined outside the path.
9555 (path_range_query::ssa_range_in_phi): Do not call
9556 range_on_path_entry for SSA names that are defined within the
9559 2021-10-14 Iain Sandoe <iain@sandoe.co.uk>
9561 * config/darwin-driver.c (darwin_find_version_from_kernel):
9562 Quote internal identifiers and avoid contractions in
9564 (darwin_default_min_version): Likewise.
9565 (darwin_driver_init): Likewise.
9567 2021-10-14 Martin Jambor <mjambor@suse.cz>
9569 * params.opt (ipa-cp-recursive-freq-factor): New.
9570 * ipa-cp.c (ipcp_value): Switch to inline initialization. New members
9571 scc_no, self_recursion_generated_level, same_scc and
9572 self_recursion_generated_p.
9573 (ipcp_lattice::add_value): Replaced parameter unlimited with
9574 same_lat_gen_level, usit it determine limit of values and store it to
9576 (ipcp_lattice<valtype>::print): Dump the new fileds.
9577 (allocate_and_init_ipcp_value): Take same_lat_gen_level as a new
9578 parameter and store it to the new value.
9579 (self_recursively_generated_p): Removed.
9580 (propagate_vals_across_arith_jfunc): Use self_recursion_generated_p
9581 instead of self_recursively_generated_p, store self generation level
9583 (value_topo_info<valtype>::add_val): Set scc_no.
9584 (value_topo_info<valtype>::propagate_effects): Multiply frequencies of
9585 recursively feeding values and self generated values by appropriate
9588 2021-10-14 Richard Sandiford <richard.sandiford@arm.com>
9590 * config/aarch64/aarch64.c (aarch64_add_stmt_cost): Remove
9591 redundant test for flag_vect_cost_model.
9593 2021-10-14 Aldy Hernandez <aldyh@redhat.com>
9595 * bitmap.c (debug): New overloaded function for auto_bitmaps.
9596 * bitmap.h (debug): Same.
9598 2021-10-14 Hongyu Wang <hongyu.wang@intel.com>
9600 * config/i386/avx512fp16intrin.h (_mm512_mask_fcmadd_pch):
9601 Adjust builtin call.
9602 (_mm512_mask3_fcmadd_pch): Likewise.
9603 (_mm512_mask_fmadd_pch): Likewise
9604 (_mm512_mask3_fmadd_pch): Likewise
9605 (_mm512_mask_fcmadd_round_pch): Likewise
9606 (_mm512_mask3_fcmadd_round_pch): Likewise
9607 (_mm512_mask_fmadd_round_pch): Likewise
9608 (_mm512_mask3_fmadd_round_pch): Likewise
9609 (_mm_mask_fcmadd_sch): Likewise
9610 (_mm_mask3_fcmadd_sch): Likewise
9611 (_mm_mask_fmadd_sch): Likewise
9612 (_mm_mask3_fmadd_sch): Likewise
9613 (_mm_mask_fcmadd_round_sch): Likewise
9614 (_mm_mask3_fcmadd_round_sch): Likewise
9615 (_mm_mask_fmadd_round_sch): Likewise
9616 (_mm_mask3_fmadd_round_sch): Likewise
9617 (_mm_fcmadd_round_sch): Likewise
9618 * config/i386/avx512fp16vlintrin.h (_mm_mask_fmadd_pch):
9619 Adjust builtin call.
9620 (_mm_mask3_fmadd_pch): Likewise
9621 (_mm256_mask_fmadd_pch): Likewise
9622 (_mm256_mask3_fmadd_pch): Likewise
9623 (_mm_mask_fcmadd_pch): Likewise
9624 (_mm_mask3_fcmadd_pch): Likewise
9625 (_mm256_mask_fcmadd_pch): Likewise
9626 (_mm256_mask3_fcmadd_pch): Likewise
9627 * config/i386/i386-builtin.def: Add mask3 builtin for complex
9628 fma, and adjust mask_builtin to corresponding expander.
9629 * config/i386/i386-expand.c (ix86_expand_round_builtin):
9630 Skip eraseing embedded rounding for expanders that emits
9632 * config/i386/sse.md (complexmove): New mode_attr.
9633 (<avx512>_fmaddc_<mode>_mask1<round_expand_name>): New expander.
9634 (<avx512>_fcmaddc_<mode>_mask1<round_expand_name>): Likewise.
9635 (avx512fp16_fmaddcsh_v8hf_mask1<round_expand_name>): Likewise.
9636 (avx512fp16_fcmaddcsh_v8hf_mask1<round_expand_name>): Likewise.
9637 (avx512fp16_fcmaddcsh_v8hf_mask3<round_expand_name>): Likewise.
9638 (avx512fp16_fmaddcsh_v8hf_mask3<round_expand_name>): Likewise.
9639 * config/i386/subst.md (round_embedded_complex): New subst.
9641 2021-10-14 John David Anglin <danglin@gcc.gnu.org>
9643 * config/pa/pa.md (cbranchsf4): Disable if TARGET_SOFT_FLOAT.
9644 (cbranchdf4): Likewise.
9645 Add missing move patterns for TARGET_SOFT_FLOAT.
9647 2021-10-13 Roger Sayle <roger@nextmovesoftware.com>
9649 * config/i386/i386-expand.c (ix86_expand_vector_move): Use a
9650 pseudo intermediate when moving a SUBREG into a hard register,
9651 by checking ix86_hardreg_mov_ok.
9652 (ix86_expand_vector_extract): Store zero-extended SImode
9653 intermediate in a pseudo, then set target using a SUBREG_PROMOTED
9655 * config/i386/sse.md (mov<VMOVE>_internal): Prevent CSE creating
9656 complex (SUBREG) sets of (vector) hard registers before reload, by
9657 checking ix86_hardreg_mov_ok.
9659 2021-10-13 Indu Bhagat <indu.bhagat@oracle.com>
9661 * ctfc.h (enum ctf_dtu_d_union_enum): Remove redundant comma.
9663 2021-10-13 Indu Bhagat <indu.bhagat@oracle.com>
9665 * dwarf2ctf.c (gen_ctf_array_type): Fix typo in comment.
9667 2021-10-13 Martin Sebor <msebor@redhat.com>
9669 PR middle-end/102630
9670 * pointer-query.cc (compute_objsize_r): Handle named address spaces.
9672 2021-10-13 Iain Sandoe <iain@sandoe.co.uk>
9674 * collect2.c (is_lto_object_file): Release simple-object
9675 resources, close files.
9677 2021-10-13 Andre Vieira <andre.simoesdiasvieira@arm.com>
9679 * config/arm/arm.c (thumb2_legitimate_address_p): Use VALID_MVE_MODE
9680 when checking mve addressing modes.
9681 (mve_vector_mem_operand): Fix the way we handle pre, post and offset
9683 (arm_print_operand): Fix printing of POST_ and PRE_MODIFY.
9684 * config/arm/mve.md: Use mve_memory_operand predicate everywhere where
9685 there is a single Ux constraint.
9687 2021-10-13 John David Anglin <danglin@gcc.gnu.org>
9689 * config/pa/pa.md (muldi3): Add support for inlining 64-bit
9690 multiplication on 32-bit PA 1.1 and 2.0 targets.
9692 2021-10-13 Martin Liska <mliska@suse.cz>
9694 PR gcov-profile/90364
9695 * coverage.c (build_info): Emit checksum to the global variable.
9696 (build_info_type): Add new field for checksum.
9697 (coverage_obj_finish): Pass object_checksum.
9698 (coverage_init): Use 0 as checksum for .gcno files.
9699 * gcov-dump.c (dump_gcov_file): Dump also new checksum field.
9700 * gcov.c (read_graph_file): Read also checksum.
9701 * doc/invoke.texi: Document the behaviour change.
9703 2021-10-13 Richard Biener <rguenther@suse.de>
9705 * gimple-iterator.h (gsi_iterator_update): Add GSI_LAST_NEW_STMT,
9706 start at integer value 2.
9707 * gimple-iterator.c (gsi_insert_seq_nodes_before): Update
9708 the iterator for GSI_LAST_NEW_STMT.
9709 (gsi_insert_seq_nodes_after): Likewise.
9710 * tree-if-conv.c (predicate_statements): Use GSI_LAST_NEW_STMT.
9711 * tree-ssa.c (execute_update_addresses_taken): Correct bogus
9712 arguments to gsi_replace.
9714 2021-10-13 Martin Liska <mliska@suse.cz>
9717 * common.opt: Use EnabledBy instead of detection in
9718 finish_options and process_options.
9719 * opts.c (finish_options): Remove handling of
9720 x_flag_unroll_all_loops.
9721 * toplev.c (process_options): Likewise for flag_web and
9722 flag_rename_registers.
9724 2021-10-13 Richard Biener <rguenther@suse.de>
9726 PR tree-optimization/102659
9727 * tree-if-conv.c (need_to_rewrite_undefined): New flag.
9728 (if_convertible_gimple_assign_stmt_p): Mark the loop for
9729 rewrite when stmts with undefined behavior on integer
9731 (combine_blocks): Predicate also when we need to rewrite stmts.
9732 (predicate_statements): Rewrite affected stmts to something
9733 with well-defined behavior on overflow.
9734 (tree_if_conversion): Initialize need_to_rewrite_undefined.
9736 2021-10-13 Richard Biener <rguenther@suse.de>
9739 * ipa-sra.c (ptr_parm_has_nonarg_uses): Fix volatileness
9742 2021-10-13 Jose E. Marchesi <jose.marchesi@oracle.com>
9744 * dwarf2ctf.c: Fix typo in comment.
9746 2021-10-12 Bill Schmidt <wschmidt@linux.ibm.com>
9749 * config/rs6000/altivec.h (vec_cpsgn): Swap operand order.
9750 * config/rs6000/rs6000-overload.def (VEC_COPYSIGN): Use SKIP to
9751 avoid generating an automatic #define of vec_cpsgn. Use the
9752 correct built-in for V4SFmode that doesn't depend on VSX.
9754 2021-10-12 Uroš Bizjak <ubizjak@gmail.com>
9758 * config/i386/i386.md (*add<mode>_1_slp): Rewrite as
9759 define_insn_and_split pattern. Add alternative 1 and split it
9760 post reload to insert operand 1 into the low part of operand 0.
9761 (*sub<mode>_1_slp): Ditto.
9762 (*and<mode>_1_slp): Ditto.
9763 (*<any_or:code><mode>_1_slp): Ditto.
9764 (*ashl<mode>3_1_slp): Ditto.
9765 (*<any_shiftrt:insn><mode>3_1_slp): Ditto.
9766 (*<any_rotate:insn><mode>3_1_slp): Ditto.
9767 (*neg<mode>_1_slp): New insn_and_split pattern.
9768 (*one_cmpl<mode>_1_slp): Ditto.
9770 2021-10-12 David Edelsohn <dje.gcc@gmail.com>
9772 * doc/install.texi: Update MinGW and mingw-64 Binaries
9775 2021-10-12 Daniel Le Duc Khoi Nguyen <greenrecyclebin@gmail.com>
9777 * doc/extend.texi (Common Variable Attributes): Fix typos in
9778 alloc_size documentation.
9780 2021-10-12 Richard Biener <rguenther@suse.de>
9782 PR tree-optimization/102696
9783 * tree-vect-slp.c (vect_build_slp_tree_2): Properly mark
9784 the tree fatally failed when we reject a BIT_FIELD_REF.
9786 2021-10-12 Richard Biener <rguenther@suse.de>
9788 PR tree-optimization/102572
9789 * tree-vect-stmts.c (vect_build_gather_load_calls): When
9790 gathering the vectorized defs for the mask pass in the
9791 desired mask vector type so invariants will be handled
9794 2021-10-12 Tamar Christina <tamar.christina@arm.com>
9796 * config/aarch64/aarch64-sve.md (*fcm<cmp_op><mode>_bic_combine,
9797 *fcm<cmp_op><mode>_nor_combine, *fcmuo<mode>_bic_combine,
9798 *fcmuo<mode>_nor_combine): New.
9800 2021-10-12 Eric Botcazou <ebotcazou@adacore.com>
9803 * config/sparc/sparc-modes.def (OI): New integer mode.
9805 2021-10-12 Jakub Jelinek <jakub@redhat.com>
9807 * gimple-fold.h (clear_padding_type_may_have_padding_p): Declare.
9808 * gimple-fold.c (clear_padding_type_may_have_padding_p): No longer
9811 2021-10-12 Jakub Jelinek <jakub@redhat.com>
9813 * tree-vectorizer.h (loop_cost_model): New function.
9814 (unlimited_cost_model): Use it.
9815 * tree-vect-loop.c (vect_analyze_loop_costing): Use loop_cost_model
9816 call instead of flag_vect_cost_model.
9817 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
9818 (vect_prune_runtime_alias_test_list): Likewise. Also use it instead
9819 of flag_simd_cost_model.
9821 2021-10-12 liuhongt <hongtao.liu@intel.com>
9824 * config/i386/i386-expand.c (emit_reduc_half): Handle
9826 * config/i386/mmx.md (reduc_<code>_scal_v4qi): New expander.
9827 (reduc_plus_scal_v4qi): Ditto.
9829 2021-10-12 Paul A. Clarke <pc@us.ibm.com>
9831 * config/rs6000/smmintrin.h (_mm_cmpeq_epi64, _mm_cmpgt_epi64,
9832 _mm_mullo_epi32, _mm_mul_epi32, _mm_packus_epi32): New.
9833 * config/rs6000/nmmintrin.h: Copy from i386, tweak to suit.
9835 2021-10-12 Paul A. Clarke <pc@us.ibm.com>
9837 * config/rs6000/smmintrin.h (_mm_cvtepi8_epi16, _mm_cvtepi8_epi32,
9838 _mm_cvtepi8_epi64, _mm_cvtepi16_epi32, _mm_cvtepi16_epi64,
9839 _mm_cvtepi32_epi64, _mm_cvtepu8_epi16, _mm_cvtepu8_epi32,
9840 _mm_cvtepu8_epi64, _mm_cvtepu16_epi32, _mm_cvtepu16_epi64,
9841 _mm_cvtepu32_epi64): New.
9843 2021-10-12 Paul A. Clarke <pc@us.ibm.com>
9845 * config/rs6000/smmintrin.h (_mm_test_all_zeros,
9846 _mm_test_all_ones, _mm_test_mix_ones_zeros): Rewrite as macro.
9848 2021-10-12 Paul A. Clarke <pc@us.ibm.com>
9850 * config/rs6000/smmintrin.h (_mm_min_epi8, _mm_min_epu16,
9851 _mm_min_epi32, _mm_min_epu32, _mm_max_epi8, _mm_max_epu16,
9852 _mm_max_epi32, _mm_max_epu32): New.
9854 2021-10-11 Jan Hubicka <hubicka@ucw.cz>
9856 * ipa-modref-tree.h (struct modref_access_node): Revert
9858 (struct modref_ref_node): Likewise.
9860 2021-10-11 Jan Hubicka <hubicka@ucw.cz>
9862 * ipa-modref-tree.h (modref_tree::global_access_p): New member
9865 (implicint_const_eaf_flags,implicit_pure_eaf_flags,
9866 ignore_stores_eaf_flags): Move to ipa-modref.h
9867 (remove_useless_eaf_flags): Remove early exit on NOCLOBBER.
9868 (modref_summary::global_memory_read_p): New member function.
9869 (modref_summary::global_memory_written_p): New member function.
9870 * ipa-modref.h (modref_summary::global_memory_read_p,
9871 modref_summary::global_memory_written_p): Declare.
9872 (implicint_const_eaf_flags,implicit_pure_eaf_flags,
9873 ignore_stores_eaf_flags): move here.
9874 * tree-ssa-structalias.c: Include ipa-modref-tree.h, ipa-modref.h
9876 (handle_rhs_call): Rewrite.
9877 (handle_call_arg): New function.
9878 (determine_global_memory_access): New function.
9879 (handle_const_call): Remove
9880 (handle_pure_call): Remove
9881 (find_func_aliases_for_call): Update use of handle_rhs_call.
9882 (compute_points_to_sets): Handle global memory acccesses
9885 2021-10-11 Diane Meirowitz <diane.meirowitz@oracle.com>
9887 * doc/invoke.texi: Add link to UndefinedBehaviorSanitizer
9888 documentation, mention UBSAN_OPTIONS, similar to what is done
9889 for AddressSanitizer.
9891 2021-10-11 Richard Biener <rguenther@suse.de>
9893 PR middle-end/102683
9894 * internal-fn.c (expand_DEFERRED_INIT): Check for mode
9895 availability before building an integer type for storage
9898 2021-10-11 Richard Biener <rguenther@suse.de>
9900 PR middle-end/101480
9901 * gimple.c (gimple_call_fnspec): Do not mark operator new/delete
9904 2021-10-11 Martin Liska <mliska@suse.cz>
9906 * common.opt: Remove Init(2) for some options.
9907 * toplev.c (process_options): Do not use AUTODETECT_VALUE, but
9908 use rather OPTION_SET_P.
9910 2021-10-11 Martin Liska <mliska@suse.cz>
9912 * common.opt: Remove usage of IRA_REGION_AUTODETECT.
9913 * flag-types.h (enum ira_region): Likewise.
9914 * toplev.c (process_options): Use OPTION_SET_P instead of
9915 IRA_REGION_AUTODETECT.
9917 2021-10-11 Jakub Jelinek <jakub@redhat.com>
9919 * omp-low.c (omp_runtime_api_call): Handle omp_get_max_teams,
9920 omp_[sg]et_teams_thread_limit and omp_set_num_teams.
9922 2021-10-11 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
9924 * config/s390/s390-protos.h (s390_rawmemchr): Add prototype.
9925 * config/s390/s390.c (s390_rawmemchr): New function.
9926 * config/s390/s390.md (rawmemchr<SINT:mode>): New expander.
9927 * config/s390/vector.md (@vec_vfees<mode>): Basically a copy of
9928 the pattern vfees<mode> from vx-builtins.md.
9929 * config/s390/vx-builtins.md (*vfees<mode>): Remove.
9931 2021-10-11 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
9933 * builtins.c (get_memory_rtx): Change to external linkage.
9934 * builtins.h (get_memory_rtx): Add function prototype.
9935 * doc/md.texi (rawmemchr<mode>): Document.
9936 * internal-fn.c (expand_RAWMEMCHR): Define.
9937 * internal-fn.def (RAWMEMCHR): Add.
9938 * optabs.def (rawmemchr_optab): Add.
9939 * tree-loop-distribution.c (find_single_drs): Change return code
9940 behaviour by also returning true if no single store was found
9942 (loop_distribution::classify_partition): Respect the new return
9943 code behaviour of function find_single_drs.
9944 (loop_distribution::execute): Call new function
9945 transform_reduction_loop in order to replace rawmemchr or strlen
9946 like loops by calls into builtins.
9947 (generate_reduction_builtin_1): New function.
9948 (generate_rawmemchr_builtin): New function.
9949 (generate_strlen_builtin_1): New function.
9950 (generate_strlen_builtin): New function.
9951 (generate_strlen_builtin_using_rawmemchr): New function.
9952 (reduction_var_overflows_first): New function.
9953 (determine_reduction_stmt_1): New function.
9954 (determine_reduction_stmt): New function.
9955 (loop_distribution::transform_reduction_loop): New function.
9957 2021-10-11 Martin Liska <mliska@suse.cz>
9959 * tree.c (cl_option_hasher::hash): Use cl_optimization_hash
9960 and remove legacy hashing code.
9962 2021-10-11 Kito Cheng <kito.cheng@sifive.com>
9965 * builtins.c (maybe_emit_call_builtin___clear_cache): Allow
9966 CONST_INT for BEGIN and END, and use gcc_assert rather than
9969 2021-10-10 Jakub Jelinek <jakub@redhat.com>
9972 * var-tracking.c (add_stores): For cselib_sp_derived_value_p values
9973 use MO_VAL_SET if loc is not sp.
9975 2021-10-10 Andrew Pinski <apinski@marvell.com>
9977 PR tree-optimization/102622
9978 * match.pd: Swap the order of a?pow2cst:0 and a?-1:0 transformations.
9979 Swap the order of a?0:pow2cst and a?0:-1 transformations.
9981 2021-10-09 liuhongt <hongtao.liu@intel.com>
9984 * config/i386/i386-expand.c (ix86_valid_mask_cmp_mode): Handle
9986 (ix86_use_mask_cmp_p): Ditto.
9987 (ix86_expand_sse_movcc): Ditto.
9988 * config/i386/i386.md (setcc_hf_mask): New define_insn.
9989 (movhf_mask): Ditto.
9990 (UNSPEC_MOVCC_MASK): New unspec.
9991 * config/i386/sse.md (UNSPEC_PCMP): Move to i386.md.
9993 2021-10-08 Vladimir N. Makarov <vmakarov@redhat.com>
9995 PR rtl-optimization/102627
9996 * lra-constraints.c (split_reg): Use at least natural mode of hard reg.
9998 2021-10-08 Aldy Hernandez <aldyh@redhat.com>
10000 * gimple-range-cache.cc (non_null_ref::non_null_deref_p): Grow
10003 2021-10-08 Aldy Hernandez <aldyh@redhat.com>
10005 * value-range.cc (irange::debug): New.
10006 * value-range.h (irange::debug): New.
10008 2021-10-08 Richard Sandiford <richard.sandiford@arm.com>
10010 PR tree-optimization/102385
10011 * predict.h (change_edge_frequency): Declare.
10012 * predict.c (change_edge_frequency): New function.
10013 * tree-ssa-loop-manip.h (tree_transform_and_unroll_loop): Remove
10015 (tree_unroll_loop): Likewise.
10016 * gimple-loop-jam.c (tree_loop_unroll_and_jam): Update accordingly.
10017 * tree-predcom.c (pcom_worker::tree_predictive_commoning_loop):
10019 * tree-ssa-loop-prefetch.c (loop_prefetch_arrays): Likewise.
10020 * tree-ssa-loop-manip.c (tree_unroll_loop): Likewise.
10021 (tree_transform_and_unroll_loop): Likewise. Use single_dom_exit
10022 to retrieve the exit edges. Make all the old profile update code
10023 conditional on !single_loop_p -- the case it was written for --
10024 and use a different approach for the single-loop case.
10026 2021-10-08 Martin Liska <mliska@suse.cz>
10028 * config/alpha/alpha.c (alpha_option_override): Use new macro
10030 * config/arc/arc.c (arc_override_options): Likewise.
10031 * config/arm/arm.c (arm_option_override): Likewise.
10032 * config/bfin/bfin.c (bfin_load_pic_reg): Likewise.
10033 * config/c6x/c6x.c (c6x_option_override): Likewise.
10034 * config/csky/csky.c: Likewise.
10035 * config/darwin.c (darwin_override_options): Likewise.
10036 * config/frv/frv.c (frv_option_override): Likewise.
10037 * config/i386/djgpp.h: Likewise.
10038 * config/i386/i386.c (ix86_stack_protect_guard): Likewise.
10039 (ix86_max_noce_ifcvt_seq_cost): Likewise.
10040 * config/ia64/ia64.c (ia64_option_override): Likewise.
10041 (ia64_override_options_after_change): Likewise.
10042 * config/m32c/m32c.c (m32c_option_override): Likewise.
10043 * config/m32r/m32r.c (m32r_init): Likewise.
10044 * config/m68k/m68k.c (m68k_option_override): Likewise.
10045 * config/microblaze/microblaze.c (microblaze_option_override): Likewise.
10046 * config/mips/mips.c (mips_option_override): Likewise.
10047 * config/nios2/nios2.c (nios2_option_override): Likewise.
10048 * config/nvptx/nvptx.c (nvptx_option_override): Likewise.
10049 * config/pa/pa.c (pa_option_override): Likewise.
10050 * config/riscv/riscv.c (riscv_option_override): Likewise.
10051 * config/rs6000/aix71.h: Likewise.
10052 * config/rs6000/aix72.h: Likewise.
10053 * config/rs6000/aix73.h: Likewise.
10054 * config/rs6000/rs6000.c (darwin_rs6000_override_options): Likewise.
10055 (rs6000_override_options_after_change): Likewise.
10056 (rs6000_linux64_override_options): Likewise.
10057 (glibc_supports_ieee_128bit): Likewise.
10058 (rs6000_option_override_internal): Likewise.
10059 (rs6000_file_start): Likewise.
10060 (rs6000_darwin_file_start): Likewise.
10061 * config/rs6000/rtems.h: Likewise.
10062 * config/rs6000/sysv4.h: Likewise.
10063 * config/rs6000/vxworks.h (SUB3TARGET_OVERRIDE_OPTIONS): Likewise.
10064 * config/s390/s390.c (s390_option_override): Likewise.
10065 * config/sh/linux.h: Likewise.
10066 * config/sh/netbsd-elf.h (while): Likewise.
10067 * config/sh/sh.c (sh_option_override): Likewise.
10068 * config/sol2.c (solaris_override_options): Likewise.
10069 * config/sparc/sparc.c (sparc_option_override): Likewise.
10070 * config/tilegx/tilegx.c (tilegx_option_override): Likewise.
10071 * config/visium/visium.c (visium_option_override): Likewise.
10072 * config/vxworks.c (vxworks_override_options): Likewise.
10073 * lto-opts.c (lto_write_options): Likewise.
10074 * omp-expand.c (expand_omp_simd): Likewise.
10075 * omp-general.c (omp_max_vf): Likewise.
10076 * omp-offload.c (oacc_xform_loop): Likewise.
10077 * opts.h (OPTION_SET_P): Likewise.
10078 * targhooks.c (default_max_noce_ifcvt_seq_cost): Likewise.
10079 * toplev.c (process_options): Likewise.
10080 * tree-predcom.c: Likewise.
10081 * tree-sra.c (analyze_all_variable_accesses): Likewise.
10083 2021-10-08 liuhongt <hongtao.liu@intel.com>
10086 * config/i386/i386.c (ix86_optab_supported_p):
10087 Return true for HFmode.
10088 * match.pd: Simplify (_Float16) ceil ((double) x) to
10089 __builtin_ceilf16 (a) when a is _Float16 type and
10090 direct_internal_fn_supported_p.
10092 2021-10-08 liuhongt <hongtao.liu@intel.com>
10095 * config/i386/i386-expand.c (emit_reduc_half): Hanlde V4HImode.
10096 * config/i386/mmx.md (reduc_plus_scal_v4hi): New.
10097 (reduc_<code>_scal_v4hi): New.
10099 2021-10-08 liuhongt <hongtao.liu@intel.com>
10101 * common.opt (ftree-vectorize): Add Var(flag_tree_vectorize).
10102 * doc/invoke.texi (Options That Control Optimization): Update
10104 * opts.c (default_options_table): Enable auto-vectorization at
10105 O2 with very-cheap cost model.
10106 (finish_options): Use cheap cost model for
10107 explicit -ftree{,-loop}-vectorize.
10109 2021-10-07 Indu Bhagat <indu.bhagat@oracle.com>
10111 * ctfc.c (ctfc_delete_container): Free hash table contents.
10113 2021-10-07 Indu Bhagat <indu.bhagat@oracle.com>
10115 * toplev.c (process_options): Do not warn for GNU GIMPLE.
10117 2021-10-07 Siddhesh Poyarekar <siddhesh@gotplt.org>
10119 * tree-object-size.c (addr_object_size,
10120 compute_builtin_object_size): Drop PDECL and POFF arguments.
10121 (addr_object_size): Adjust calls.
10122 * tree-object-size.h (compute_builtin_object_size): Drop PDECL
10123 and POFF arguments.
10125 2021-10-07 Roger Sayle <roger@nextmovesoftware.com>
10127 * rtl.def (SMUL_HIGHPART, UMUL_HIGHPART): New RTX codes for
10128 representing signed and unsigned high-part multiplication resp.
10129 * simplify-rtx.c (simplify_binary_operation_1) [SMUL_HIGHPART,
10130 UMUL_HIGHPART]: Simplify high-part multiplications by zero.
10131 [SS_PLUS, US_PLUS, SS_MINUS, US_MINUS, SS_MULT, US_MULT,
10132 SS_DIV, US_DIV]: Similar simplifications for saturating
10134 (simplify_const_binary_operation) [SS_PLUS, US_PLUS, SS_MINUS,
10135 US_MINUS, SS_MULT, US_MULT, SMUL_HIGHPART, UMUL_HIGHPART]:
10136 Implement compile-time evaluation for constant operands.
10137 * dwarf2out.c (mem_loc_descriptor): Skip SMUL_HIGHPART and
10139 * doc/rtl.texi (smul_highpart, umul_highpart): Document RTX codes.
10140 * doc/md.texi (smul@var{m}3_highpart, umul@var{m3}_highpart):
10141 Mention the new smul_highpart and umul_highpart RTX codes.
10142 * doc/invoke.texi: Silence @xref "compilation" warnings.
10144 2021-10-07 Martin Jambor <mjambor@suse.cz>
10147 * ipa-prop.c (ipa_edge_args_sum_t::duplicate): Also handle the
10148 case when the source reference description corresponds to a
10149 referance taken in a function src->caller is inlined to.
10151 2021-10-07 Jan Hubicka <hubicka@ucw.cz>
10154 * ipa-modref-tree.h (modref_access_node::contains_p): Handle offsets
10156 (modref_access_node::try_merge_with): Add sanity check that there
10157 are no redundant entries in the list.
10159 2021-10-07 Richard Biener <rguenther@suse.de>
10161 PR tree-optimization/102608
10162 * tree-ssa-sccvn.c (visit_stmt): Drop .DEFERRED_INIT to
10165 2021-10-07 Martin Liska <mliska@suse.cz>
10167 * toplev.c (toplev::main): Make
10168 save_opt_decoded_options a pointer type
10169 * toplev.h: Likewise.
10171 2021-10-07 Andrew Stubbs <ams@codesourcery.com>
10173 * config/gcn/gcn-valu.md (gather<mode>_insn_2offsets<exec>): Apply
10174 HAVE_GCN_ASM_GLOBAL_LOAD_FIXED.
10175 (scatter<mode>_insn_2offsets<exec_scatter>): Likewise.
10177 2021-10-07 Andrew Stubbs <ams@codesourcery.com>
10179 * config/gcn/gcn-hsa.h (SRAMOPT): Include the whole option string.
10180 Adjust for new -msram-ecc=any behaviour.
10181 (ASM_SPEC): Adjust -mxnack and -msram-ecc usage.
10182 * config/gcn/gcn.c (output_file_start): Implement -msram-ecc=any.
10183 * config/gcn/mkoffload.c (EF_AMDGPU_XNACK): Rename to ...
10184 (EF_AMDGPU_XNACK_V3): ... this.
10185 (EF_AMDGPU_SRAM_ECC): Rename to ...
10186 (EF_AMDGPU_SRAM_ECC_V3): ... this.
10187 (EF_AMDGPU_FEATURE_XNACK_V4): New.
10188 (EF_AMDGPU_FEATURE_XNACK_UNSUPPORTED_V4): New.
10189 (EF_AMDGPU_FEATURE_XNACK_ANY_V4): New.
10190 (EF_AMDGPU_FEATURE_XNACK_OFF_V4): New.
10191 (EF_AMDGPU_FEATURE_XNACK_ON_V4): New.
10192 (EF_AMDGPU_FEATURE_SRAMECC_V4): New.
10193 (EF_AMDGPU_FEATURE_SRAMECC_UNSUPPORTED_V4): New.
10194 (EF_AMDGPU_FEATURE_SRAMECC_ANY_V4): New.
10195 (EF_AMDGPU_FEATURE_SRAMECC_OFF_V4): New.
10196 (EF_AMDGPU_FEATURE_SRAMECC_ON_V4): New.
10197 (SET_XNACK_ON): New.
10198 (SET_XNACK_OFF): New.
10200 (SET_SRAM_ECC_ON): New.
10201 (SET_SRAM_ECC_ANY): New.
10202 (SET_SRAM_ECC_OFF): New.
10203 (TEST_SRAM_ECC_ANY): New.
10204 (TEST_SRAM_ECC_ON): New.
10205 (main): Implement HSACOv4 and -msram-ecc=any.
10207 2021-10-07 Andrew Stubbs <ams@codesourcery.com>
10209 * config.in: Regenerate.
10210 * config/gcn/gcn-hsa.h (X_FIJI): New macro.
10211 (X_900): New macro.
10212 (X_906): New macro.
10213 (X_908): New macro.
10214 (A_FIJI): Rename to ...
10215 (S_FIJI): ... this.
10216 (A_900): Rename to ...
10218 (A_906): Rename to ...
10220 (A_908): Rename to ...
10222 (SRAMOPT): New macro.
10223 (ASM_SPEC): Adjust xnack option usage.
10224 * config/gcn/gcn.c (output_file_start): Adjust amdgcn_target usage.
10225 * configure: Regenerate.
10226 * configure.ac: Detect LLVM assembler dialect.
10228 2021-10-07 Richard Biener <rguenther@suse.de>
10230 * tree-pretty-print.c (dump_generic_node): Do not elide
10231 printing '&' when dumping with -gimple.
10233 2021-10-06 Andrew MacLeod <amacleod@redhat.com>
10235 * gimple-range-cache.cc (non_null_ref::adjust_range): Call new
10237 * gimple-range-fold.cc (adjust_pointer_diff_expr): Ditto.
10238 (adjust_imagpart_expr): Ditto.
10239 * value-range.cc (irange::irange_intersect): Call new routine if
10240 RHS is a single pair.
10241 (irange::intersect): New wide_int version.
10242 * value-range.h (class irange): New prototype.
10244 2021-10-06 Andrew MacLeod <amacleod@redhat.com>
10246 * gimple-range-edge.cc (gimple_outgoing_range::gimple_outgoing_range):
10247 Add parameter to limit size when recognizing switches.
10248 (gimple_outgoing_range::edge_range_p): Check size limit.
10249 * gimple-range-edge.h (gimple_outgoing_range): Add size field.
10250 * gimple-range-gori.cc (gori_map::calculate_gori): Ignore switches
10251 that exceed the size limit.
10252 (gori_compute::gori_compute): Add initializer.
10253 * params.opt (evrp-switch-limit): New.
10254 * doc/invoke.texi: Update docs.
10256 2021-10-06 Andrew MacLeod <amacleod@redhat.com>
10258 * value-range.h (irange::set_varying): Use TYPE_MIN_VALUE and
10259 TYPE_MAX_VALUE instead of creating new trees when possible.
10261 2021-10-06 Andrew MacLeod <amacleod@redhat.com>
10263 * gimple-range-cache.cc (non_null_ref::adjust_range): Check for
10264 zero and non-zero more efficently.
10266 2021-10-06 Richard Biener <rguenther@suse.de>
10269 * dumpfile.h (TDF_GIMPLE_VAL): New.
10270 (dump_flag): Re-order and adjust TDF_* flags. Make
10271 the enum uint32_t. Use std::underlying_type in the
10272 operator overloads.
10273 (optgroup_flag): Likewise for the operator overloads.
10274 * tree-pretty-print.c (dump_generic_node): Wrap ADDR_EXPR
10275 in _Literal if TDF_GIMPLE_VAL.
10276 * gimple-pretty-print.c (dump_gimple_assign): Add
10277 TDF_GIMPLE_VAL to flags when dumping operands where only
10278 is_gimple_val are allowed.
10279 (dump_gimple_cond): Likewise.
10281 2021-10-06 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
10283 * gimple-isel.cc (gimple_expand_vec_cond_expr): Remove redundant if
10286 2021-10-05 qing zhao <qing.zhao@oracle.com>
10288 PR middle-end/102359
10289 * gimplify.c (gimplify_decl_expr): Not add initialization for an
10290 auto variable when it has been initialized by frontend.
10292 2021-10-05 Aldy Hernandez <aldyh@redhat.com>
10294 * tree-ssa-threadupdate.c (jt_path_registry::cancel_invalid_paths):
10295 Loosen restrictions
10297 2021-10-05 Jan-Benedict Glaw <jbglaw@lug-owl.de>
10299 * common/config/avr/avr-common.c (avr_handle_option): Mark
10300 argument as ATTRIBUTE_UNUSED.
10302 2021-10-05 Jan-Benedict Glaw <jbglaw@lug-owl.de>
10304 * config/lm32/uclinux-elf.h (LINK_GCC_C_SEQUENCE_SPEC):
10305 Undefine before redefinition.
10307 2021-10-05 Richard Biener <rguenther@suse.de>
10309 * toplev.c (no_backend): Remove global var.
10310 (process_options): Pass in no_backend, move post_options
10311 langhook call to toplev::main.
10312 (do_compile): Pass in no_backend, move process_options call
10314 (toplev::run_self_tests): Check no_backend at the caller.
10315 (toplev::main): Call post_options and process_options
10316 split out from do_compile, do self-tests only if
10317 no_backend is initialized.
10319 2021-10-05 Richard Biener <rguenther@suse.de>
10321 * tree-cfg.c (dump_function_to_file): Dump the UID of the
10322 function as part of the name when requested.
10323 * tree-pretty-print.c (dump_function_name): Dump the UID when
10324 requested and the langhook produced the actual name.
10326 2021-10-05 Richard Biener <rguenther@suse.de>
10328 PR middle-end/102587
10329 PR middle-end/102285
10330 * internal-fn.c (expand_DEFERRED_INIT): Fall back to
10331 zero-initialization as last resort, use the constant
10332 size as given by the DEFERRED_INIT argument to build
10335 2021-10-04 Marek Polacek <polacek@redhat.com>
10338 * doc/invoke.texi: Document -Warray-compare.
10340 2021-10-04 Richard Biener <rguenther@suse.de>
10342 * gimplify.c (is_var_need_auto_init): DECL_HARD_REGISTER
10343 variables are not to be initialized.
10345 2021-10-04 Richard Biener <rguenther@suse.de>
10347 * expr.h (non_mem_decl_p): Declare.
10348 (mem_ref_refers_to_non_mem_p): Likewise.
10349 * expr.c (non_mem_decl_p): Export.
10350 (mem_ref_refers_to_non_mem_p): Likewise.
10351 * internal-fn.c (expand_DEFERRED_INIT): Do not expand the LHS
10352 but check the base with mem_ref_refers_to_non_mem_p
10353 and non_mem_decl_p.
10355 2021-10-04 Richard Biener <rguenther@suse.de>
10357 PR tree-optimization/102570
10358 * tree-ssa-sccvn.h (vn_reference_op_struct): Document
10359 we are using clique for the internal function code.
10360 * tree-ssa-sccvn.c (vn_reference_op_eq): Compare the
10361 internal function code.
10362 (print_vn_reference_ops): Print the internal function code.
10363 (vn_reference_op_compute_hash): Hash it.
10364 (copy_reference_ops_from_call): Record it.
10365 (visit_stmt): Remove the restriction around internal function
10367 (fully_constant_vn_reference_p): Use fold_const_call and handle
10368 internal functions.
10369 (vn_reference_eq): Compare call return types.
10370 * tree-ssa-pre.c (create_expression_by_pieces): Handle
10371 generating calls to internal functions.
10372 (compute_avail): Remove the restriction around internal function
10375 2021-10-04 Aldy Hernandez <aldyh@redhat.com>
10377 PR tree-optimization/102560
10378 * gimple-ssa-warn-alloca.c (alloca_call_type): Remove static
10379 marker for invalid_range.
10381 2021-10-04 Richard Biener <rguenther@suse.de>
10383 PR middle-end/102587
10384 * internal-fn.c (expand_DEFERRED_INIT): Guard register
10385 initialization path an avoid initializing VLA registers
10388 2021-10-04 Eric Botcazou <ebotcazou@adacore.com>
10390 * config/rs6000/vxworks.h (TARGET_INIT_LIBFUNCS): Delete.
10392 2021-10-03 Martin Liska <mliska@suse.cz>
10394 * toplev.c (toplev::main): Check opt_index if it is a part
10397 2021-10-02 Aldy Hernandez <aldyh@redhat.com>
10399 PR tree-optimization/102563
10400 * range-op.cc (operator_lshift::op1_range): Do not clobber
10403 2021-10-02 Martin Liska <mliska@suse.cz>
10405 * toplev.c (toplev::main): save_decoded_options[0] is program
10406 name and so it should be skipped.
10408 2021-10-01 Aldy Hernandez <aldyh@redhat.com>
10410 PR tree-optimization/102546
10411 * range-op.cc (operator_lshift::op1_range): Teach range-ops that
10412 X << Y is non-zero implies X is also non-zero.
10414 2021-10-01 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
10416 * config/aarch64/aarch64-cores.def (AARCH64_CORE): New
10418 * config/aarch64/aarch64-tune.md: Regenerate.
10419 * doc/invoke.texi: Update docs.
10421 2021-10-01 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
10423 * config/aarch64/aarch64-cores.def (AARCH64_CORE): New
10425 * config/aarch64/aarch64-tune.md: Regenerate.
10426 * doc/invoke.texi: Update docs.
10428 2021-10-01 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
10430 * config/aarch64/aarch64-cores.def (AARCH64_CORE): New
10432 * config/aarch64/aarch64-tune.md: Regenerate.
10433 * doc/invoke.texi: Update docs.
10435 2021-10-01 Martin Sebor <msebor@redhat.com>
10438 * doc/invoke.texi (-Waddress): Update.
10439 * gengtype.c (write_types): Avoid -Waddress.
10440 * poly-int.h (POLY_SET_COEFF): Avoid using null.
10442 2021-10-01 John David Anglin <danglin@gcc.gnu.org>
10445 * config/pa/pa.c (pa_option_override): Default to dwarf version 4
10448 2021-10-01 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
10450 * config/aarch64/aarch64.h (AARCH64_FL_V9): Update value.
10452 2021-10-01 Aldy Hernandez <aldyh@redhat.com>
10454 * gimple-range-path.cc (path_range_query::compute_ranges): Use
10456 * gimple-range-path.h (class path_range_query): Remove shadowed
10458 (path_range_query::get_path_oracle): New.
10460 2021-10-01 Jakub Jelinek <jakub@redhat.com>
10461 Richard Biener <rguenther@suse.de>
10463 PR sanitizer/102515
10464 * doc/invoke.texi (-fsanitize=integer-divide-by-zero): Remove
10465 INT_MIN / -1 division detection from here ...
10466 (-fsanitize=signed-integer-overflow): ... and add it here.
10468 2021-10-01 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
10470 * config/aarch64/aarch64-arches.def (AARCH64_ARCH): Added
10472 * config/aarch64/aarch64.h (AARCH64_FL_V9): New.
10473 (AARCH64_FL_FOR_ARCH9): New flags for Armv9-A.
10474 (AARCH64_ISA_V9): New ISA flag.
10475 * doc/invoke.texi: Update docs.
10477 2021-10-01 Martin Liska <mliska@suse.cz>
10479 * toplev.c (toplev::main): Save decoded optimization options.
10480 * toplev.h (save_opt_decoded_options): New.
10481 * doc/extend.texi: Be more clear about optimize and target
10484 2021-10-01 Eric Botcazou <ebotcazou@adacore.com>
10486 * explow.c: Include langhooks.h.
10487 (set_stack_check_libfunc): Build a proper function type.
10489 2021-10-01 Eric Botcazou <ebotcazou@adacore.com>
10492 * config/i386/i386.c (legitimate_pic_address_disp_p): For PE-COFF do
10493 not return true for external weak function symbols in medium model.
10495 2021-10-01 Jakub Jelinek <jakub@redhat.com>
10497 * tree.h (OMP_CLAUSE_ORDER_REPRODUCIBLE): Define.
10498 * tree-pretty-print.c (dump_omp_clause) <case OMP_CLAUSE_ORDER>: Print
10499 reproducible: for OMP_CLAUSE_ORDER_REPRODUCIBLE.
10500 * omp-general.c (omp_extract_for_data): If OMP_CLAUSE_ORDER is seen
10501 without OMP_CLAUSE_ORDER_UNCONSTRAINED, overwrite sched_kind to
10502 OMP_CLAUSE_SCHEDULE_STATIC.
10504 2021-10-01 Richard Biener <rguenther@suse.de>
10506 PR middle-end/102518
10507 * tree-inline.c (setup_one_parameter): Avoid substituting
10508 an invariant into contexts where a GIMPLE register is not valid.
10510 2021-09-30 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
10512 * config/arm/arm-cpus.in: Add Cortex-R52+ CPU.
10513 * config/arm/arm-tables.opt: Regenerate.
10514 * config/arm/arm-tune.md: Regenerate.
10515 * doc/invoke.texi: Update docs.
10517 2021-09-30 Uroš Bizjak <ubizjak@gmail.com>
10520 * config/i386/i386.md
10521 (sign_extend:WIDE (any_logic:NARROW (memory, immediate)) splitters):
10524 2021-09-30 Tobias Burnus <tobias@codesourcery.com>
10526 * omp-low.c (omp_runtime_api_call): Add omp_aligned_{,c}alloc and
10527 omp_{c,re}alloc, fix omp_alloc/omp_free.
10529 2021-09-30 Martin Liska <mliska@suse.cz>
10531 * defaults.h (ASM_OUTPUT_ASCII): Do not hide global variable
10532 asm_out_file and stream directly to MYFILE.
10534 2021-09-30 Richard Biener <rguenther@suse.de>
10536 * tree-vect-data-refs.c (vect_update_misalignment_for_peel):
10537 Restore and fix condition under which we apply npeel to
10538 the DRs misalignment value.
10540 2021-09-30 Richard Biener <rguenther@suse.de>
10542 * tree-vect-data-refs.c (vect_update_misalignment_for_peel):
10543 Fix npeel check for variable amount of peeling.
10545 2021-09-30 Aldy Hernandez <aldyh@redhat.com>
10547 * lto-wrapper.c (run_gcc): Plug snprintf overflow.
10549 2021-09-30 Aldy Hernandez <aldyh@redhat.com>
10551 * gimple-range.cc (gimple_ranger::debug): New.
10552 * gimple-range.h (class gimple_ranger): Add debug.
10554 2021-09-30 Aldy Hernandez <aldyh@redhat.com>
10556 PR middle-end/102519
10557 * tree-vrp.c (hybrid_threader::~hybrid_threader): Free m_query.
10559 2021-09-29 Indu Bhagat <indu.bhagat@oracle.com>
10562 * btfout.c (GTY): Add GTY (()) albeit for cosmetic only purpose.
10563 (btf_finalize): Empty the hash_map btf_var_ids.
10565 2021-09-29 Aldy Hernandez <aldyh@redhat.com>
10567 * tree-vrp.c (thread_through_all_blocks): Return bool.
10568 (execute_vrp_threader): Return TODO_* flags.
10569 (pass_data_vrp_threader): Set todo_flags_finish to 0.
10571 2021-09-29 Aldy Hernandez <aldyh@redhat.com>
10573 * timevar.def (TV_TREE_VRP_THREADER): New.
10574 * tree-vrp.c: Use TV_TREE_VRP_THREADER for VRP threader pass.
10576 2021-09-29 David Faust <david.faust@oracle.com>
10578 * config.gcc (bpf-*-*): Do not overwrite extra_headers.
10580 2021-09-29 Jonathan Wright <jonathan.wright@arm.com>
10582 * config/aarch64/aarch64-builtins.c (TYPES_BINOP_PPU): Define
10583 new type qualifier enum.
10584 (TYPES_TERNOP_SSSU): Likewise.
10585 (TYPES_TERNOP_PPPU): Likewise.
10586 * config/aarch64/aarch64-simd-builtins.def: Define PPU, SSU,
10587 PPPU and SSSU builtin generator macros for qtbl1 and qtbx1
10589 * config/aarch64/arm_neon.h (vqtbl1_p8): Use type-qualified
10590 builtin and remove casts.
10591 (vqtbl1_s8): Likewise.
10592 (vqtbl1q_p8): Likewise.
10593 (vqtbl1q_s8): Likewise.
10594 (vqtbx1_s8): Likewise.
10595 (vqtbx1_p8): Likewise.
10596 (vqtbx1q_s8): Likewise.
10597 (vqtbx1q_p8): Likewise.
10598 (vtbl1_p8): Likewise.
10599 (vtbl2_p8): Likewise.
10600 (vtbx2_p8): Likewise.
10602 2021-09-29 Richard Biener <rguenther@suse.de>
10604 * tree-vect-data-refs.c (vect_dr_misalign_for_aligned_access):
10606 (vect_update_misalignment_for_peel): Use it to update
10607 misaligned to the value necessary for an aligned access.
10608 (vect_get_peeling_costs_all_drs): Likewise.
10609 (vect_enhance_data_refs_alignment): Likewise.
10611 2021-09-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10613 * config/aarch64/aarch64.c (aarch64_expand_cpymem): Count number of
10614 emitted operations and adjust heuristic for code size.
10616 2021-09-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10618 * config/aarch64/aarch64.c (aarch64_expand_setmem): Count number of
10619 emitted operations and adjust heuristic for code size.
10621 2021-09-29 Jakub Jelinek <jakub@redhat.com>
10623 PR middle-end/102504
10624 * gimplify.c (gimplify_scan_omp_clauses): Use omp_check_private even
10625 in OMP_SCOPE clauses, not just on worksharing construct clauses.
10627 2021-09-28 Geng Qi <gengqi@linux.alibaba.com>
10629 * config/riscv/riscv.md (mulv<mode>4): Call gen_smul<mode>3_highpart.
10630 (<u>mulditi3): Call <su>muldi3_highpart.
10631 (<u>muldi3_highpart): Rename to <su>muldi3_highpart.
10632 (<u>mulsidi3): Call <su>mulsi3_highpart.
10633 (<u>mulsi3_highpart): Rename to <su>mulsi3_highpart.
10635 2021-09-28 Iain Sandoe <iain@sandoe.co.uk>
10637 * config/darwin.h (DSYMUTIL_SPEC): Recognize D sources.
10639 2021-09-28 Iain Sandoe <iain@sandoe.co.uk>
10641 * config/rs6000/darwin.h (FIXED_R13): Add for PPC64.
10642 (FIRST_SAVED_GP_REGNO): Save from R13 even when it is one
10645 2021-09-28 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10647 * config/aarch64/aarch64.h (AARCH64_FL_LS64): Define
10648 (AARCH64_FL_V8_7): Likewise.
10649 (AARCH64_FL_FOR_ARCH8_7): Likewise.
10650 * config/aarch64/aarch64-arches.def (armv8.7-a): Define.
10651 * config/aarch64/aarch64-option-extensions.def (ls64): Define.
10652 * doc/invoke.texi: Document the above.
10654 2021-09-28 Aldy Hernandez <aldyh@redhat.com>
10656 * dbgcnt.c (dbg_cnt_counter): New.
10657 * dbgcnt.h (dbg_cnt_counter): New.
10658 * dumpfile.c (dump_options): Add entry for TDF_THREADING.
10659 * dumpfile.h (enum dump_flag): Add TDF_THREADING.
10660 * gimple-range-path.cc (DEBUG_SOLVER): Use TDF_THREADING.
10661 * tree-ssa-threadupdate.c (dump_jump_thread_path): Dump out
10664 2021-09-28 Aldy Hernandez <aldyh@redhat.com>
10666 * cfgcleanup.c (pass_jump::execute): Check
10667 flag_expensive_optimizations.
10668 (pass_jump_after_combine::gate): Same.
10669 * doc/invoke.texi (-fthread-jumps): Enable for -O1.
10670 * opts.c (default_options_table): Enable -fthread-jumps at -O1.
10671 * tree-ssa-threadupdate.c
10672 (fwd_jt_path_registry::remove_jump_threads_including): Bail unless
10675 2021-09-28 Ilya Leoshkevich <iii@linux.ibm.com>
10677 * tree-ssa-reassoc.c (biased_names): New global.
10678 (propagate_bias_p): New function.
10679 (loop_carried_phi): Remove.
10680 (propagate_rank): Propagate bias along single uses.
10681 (get_rank): Update biased_names when needed.
10683 2021-09-28 Ilya Leoshkevich <iii@linux.ibm.com>
10685 * passes.def (pass_reassoc): Rename parameter to early_p.
10686 * tree-ssa-reassoc.c (reassoc_bias_loop_carried_phi_ranks_p):
10688 (phi_rank): Don't bias loop-carried phi ranks
10689 before vectorization pass.
10690 (execute_reassoc): Add bias_loop_carried_phi_ranks_p parameter.
10691 (pass_reassoc::pass_reassoc): Add bias_loop_carried_phi_ranks_p
10693 (pass_reassoc::set_param): Set bias_loop_carried_phi_ranks_p
10695 (pass_reassoc::execute): Pass bias_loop_carried_phi_ranks_p to
10697 (pass_reassoc::bias_loop_carried_phi_ranks_p): New member.
10699 2021-09-28 Jakub Jelinek <jakub@redhat.com>
10702 * config/i386/i386.c (standard_80387_constant_p): Don't recognize
10703 special 80387 instruction XFmode constants if flag_rounding_math.
10705 2021-09-28 Richard Biener <rguenther@suse.de>
10707 PR tree-optimization/100112
10708 * tree-ssa-sccvn.c (visit_reference_op_load): Record the
10709 referece into the hashtable twice in case last_vuse is
10710 different from the original vuse on the stmt.
10712 2021-09-28 Jakub Jelinek <jakub@redhat.com>
10714 PR middle-end/102492
10715 * gimplify.c (gimplify_adjust_omp_clauses_1): Don't call the
10716 omp_finish_clause langhook on implicitly added OMP_CLAUSE_PRIVATE
10717 clauses on SIMD constructs.
10719 2021-09-28 Aldy Hernandez <aldyh@redhat.com>
10721 PR tree-optimization/102511
10722 * gimple-range-path.cc (path_range_query::range_on_path_entry):
10723 Return VARYING when nothing found.
10725 2021-09-28 Hongyu Wang <hongyu.wang@intel.com>
10728 * config/i386/i386.h (VALID_AVX512FP16_REG_MODE): Add
10730 (VALID_SSE2_REG_VHF_MODE): Add V4HFmode and V2HFmode.
10731 (VALID_MMX_REG_MODE): Add V4HFmode.
10732 (SSE_REG_MODE_P): Replace VALID_AVX512FP16_REG_MODE with
10733 vector mode condition.
10734 * config/i386/i386.c (classify_argument): Parse V4HF/V2HF
10736 (function_arg_32): Add V4HFmode.
10737 (function_arg_advance_32): Likewise.
10738 * config/i386/i386.md (mode): Add V4HF/V2HF.
10739 (MODE_SIZE): Likewise.
10740 * config/i386/mmx.md (MMXMODE): Add V4HF mode.
10741 (V_32): Add V2HF mode.
10742 (VHF_32_64): New mode iterator.
10743 (*mov<mode>_internal): Adjust sse alternatives to support
10745 (*mov<mode>_internal): Adjust sse alternatives to support
10747 (<insn><mode>3): New define_insn for add/sub/mul/div.
10749 2021-09-28 Aldy Hernandez <aldyh@redhat.com>
10751 * tree-ssa-threadbackward.c (pass_thread_jumps::gate): Check
10753 (pass_early_thread_jumps::gate): Same.
10754 * tree-ssa-threadedge.c (jump_threader::thread_outgoing_edges):
10755 Return if !flag_thread_jumps.
10756 * tree-ssa-threadupdate.c
10757 (jt_path_registry::register_jump_thread): Assert that
10758 flag_thread_jumps is true.
10760 2021-09-28 liuhongt <hongtao.liu@intel.com>
10763 (simplify_context::simplify_binary_operation_1): Relax
10764 condition of simplifying (vec_concat:M (vec_select op0
10765 index0)(vec_select op1 index1)) to allow different modes
10766 between op0 and M, but have same inner mode.
10768 2021-09-28 liuhongt <hongtao.liu@intel.com>
10770 * config/i386/i386-expand.c (emit_reduc_half): Handle
10771 V8HF/V16HF/V32HFmode.
10772 * config/i386/sse.md (REDUC_SSE_PLUS_MODE): Add V8HF.
10773 (REDUC_SSE_SMINMAX_MODE): Ditto.
10774 (REDUC_PLUS_MODE): Add V16HF and V32HF.
10775 (REDUC_SMINMAX_MODE): Ditto.
10777 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
10779 * gimple-range-path.cc
10780 (path_range_query::precompute_ranges_in_block): Rename to...
10781 (path_range_query::compute_ranges_in_block): ...this.
10782 (path_range_query::precompute_ranges): Rename to...
10783 (path_range_query::compute_ranges): ...this.
10784 (path_range_query::precompute_relations): Rename to...
10785 (path_range_query::compute_relations): ...this.
10786 (path_range_query::precompute_phi_relations): Rename to...
10787 (path_range_query::compute_phi_relations): ...this.
10788 * gimple-range-path.h: Rename precompute* to compute*.
10789 * tree-ssa-threadbackward.c
10790 (back_threader::find_taken_edge_switch): Same.
10791 (back_threader::find_taken_edge_cond): Same.
10792 * tree-ssa-threadedge.c
10793 (hybrid_jt_simplifier::compute_ranges_from_state): Same.
10794 (hybrid_jt_state::register_equivs_stmt): Inline...
10795 * tree-ssa-threadedge.h: ...here.
10797 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
10799 * tree-vrp.c (lhs_of_dominating_assert): Remove.
10800 (class vrp_jt_state): Remove.
10801 (class vrp_jt_simplifier): Remove.
10802 (vrp_jt_simplifier::simplify): Remove.
10803 (class vrp_jump_threader): Remove.
10804 (vrp_jump_threader::vrp_jump_threader): Remove.
10805 (vrp_jump_threader::~vrp_jump_threader): Remove.
10806 (vrp_jump_threader::before_dom_children): Remove.
10807 (vrp_jump_threader::after_dom_children): Remove.
10809 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
10811 * passes.def (pass_vrp_threader): New.
10812 * tree-pass.h (make_pass_vrp_threader): Add make_pass_vrp_threader.
10813 * tree-ssa-threadedge.c (hybrid_jt_state::register_equivs_stmt): New.
10814 (hybrid_jt_simplifier::hybrid_jt_simplifier): New.
10815 (hybrid_jt_simplifier::simplify): New.
10816 (hybrid_jt_simplifier::compute_ranges_from_state): New.
10817 * tree-ssa-threadedge.h (class hybrid_jt_state): New.
10818 (class hybrid_jt_simplifier): New.
10819 * tree-vrp.c (execute_vrp): Remove ASSERT_EXPR based jump
10821 (class hybrid_threader): New.
10822 (hybrid_threader::hybrid_threader): New.
10823 (hybrid_threader::~hybrid_threader): New.
10824 (hybrid_threader::before_dom_children): New.
10825 (hybrid_threader::after_dom_children): New.
10826 (execute_vrp_threader): New.
10827 (class pass_vrp_threader): New.
10828 (make_pass_vrp_threader): New.
10830 2021-09-27 Martin Liska <mliska@suse.cz>
10832 * output.h (enum section_flag): New.
10833 (SECTION_FORGET): Remove.
10834 (SECTION_ENTSIZE): Make it (1UL << 8) - 1.
10835 (SECTION_STYLE_MASK): Define it based on other enum
10837 * varasm.c (switch_to_section): Remove unused handling of
10840 2021-09-27 Martin Liska <mliska@suse.cz>
10842 * common.opt: Add new variable flag_default_complex_method.
10843 * opts.c (finish_options): Handle flags related to
10844 x_flag_complex_method.
10845 * toplev.c (process_options): Remove option handling related
10846 to flag_complex_method.
10848 2021-09-27 Richard Biener <rguenther@suse.de>
10850 PR middle-end/102450
10851 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid using
10852 type_for_size, instead use int_mode_for_size.
10854 2021-09-27 Andrew Pinski <apinski@marvell.com>
10857 * gimplify.c (gimplify_save_expr): Return early
10858 if the type of val is error_mark_node.
10860 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
10862 * tree-ssanames.c (ssa_name_has_boolean_range): Use
10865 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
10867 * gimple-ssa-evrp-analyze.h (class evrp_range_analyzer): Remove
10868 vrp_visit_cond_stmt.
10869 * tree-ssa-dom.c (cprop_operand): Convert to range_query API.
10870 (cprop_into_stmt): Same.
10871 (dom_opt_dom_walker::optimize_stmt): Same.
10873 2021-09-27 Richard Biener <rguenther@suse.de>
10875 PR tree-optimization/97351
10876 PR tree-optimization/97352
10877 PR tree-optimization/82426
10878 * tree-vectorizer.h (dr_misalignment): Add vector type
10880 (aligned_access_p): Likewise.
10881 (known_alignment_for_access_p): Likewise.
10882 (vect_supportable_dr_alignment): Likewise.
10883 (vect_known_alignment_in_bytes): Likewise. Refactor.
10884 (DR_MISALIGNMENT): Remove.
10885 (vect_update_shared_vectype): Likewise.
10886 * tree-vect-data-refs.c (dr_misalignment): Refactor, handle
10887 a vector type with larger alignment requirement and apply
10888 the negative step adjustment here.
10889 (vect_calculate_target_alignment): Remove.
10890 (vect_compute_data_ref_alignment): Get explicit vector type
10891 argument, do not apply a negative step alignment adjustment
10893 (vect_slp_analyze_node_alignment): Re-analyze alignment
10894 when we re-visit the DR with a bigger desired alignment but
10895 keep more precise results from smaller alignments.
10896 * tree-vect-slp.c (vect_update_shared_vectype): Remove.
10897 (vect_slp_analyze_node_operations_1): Do not update the
10898 shared vector type on stmts.
10899 * tree-vect-stmts.c (vect_analyze_stmt): Push/pop the
10900 vector type of an SLP node to the representative stmt-info.
10901 (vect_transform_stmt): Likewise.
10903 2021-09-27 liuhongt <hongtao.liu@intel.com>
10906 2021-09-09 liuhongt <hongtao.liu@intel.com>
10909 * config/i386/sse.md (reduc_plus_scal_<mode>): Split to ..
10910 (reduc_plus_scal_v4sf): .. this, New define_expand.
10911 (reduc_plus_scal_v2df): .. and this, New define_expand.
10913 2021-09-26 liuhongt <hongtao.liu@intel.com>
10915 * doc/extend.texi (Half-Precision): Remove storage only
10916 description for _Float16 w/o avx512fp16.
10918 2021-09-25 Dimitar Dimitrov <dimitar@dinux.eu>
10920 * config/pru/constraints.md (Rrio): New constraint.
10921 * config/pru/predicates.md (regio_operand): New predicate.
10922 * config/pru/pru-pragma.c (pru_register_pragmas): Register
10923 the __regio_symbol address space.
10924 * config/pru/pru-protos.h (pru_symref2ioregno): Declaration.
10925 * config/pru/pru.c (pru_symref2ioregno): New helper function.
10926 (pru_legitimate_address_p): Remove.
10927 (pru_addr_space_legitimate_address_p): Use the address space
10928 aware hook variant.
10929 (pru_nongeneric_pointer_addrspace): New helper function.
10930 (pru_insert_attributes): New function to validate __regio_symbol
10932 (TARGET_INSERT_ATTRIBUTES): New macro.
10933 (TARGET_LEGITIMATE_ADDRESS_P): Remove.
10934 (TARGET_ADDR_SPACE_LEGITIMATE_ADDRESS_P): New macro.
10935 * config/pru/pru.h (enum reg_class): Add REGIO_REGS class.
10936 * config/pru/pru.md (*regio_readsi): New pattern to read I/O
10938 (*regio_nozext_writesi): New pattern to write to I/O registers.
10939 (*regio_zext_write_r30<EQS0:mode>): Ditto.
10940 * doc/extend.texi: Document the new PRU Named Address Space.
10942 2021-09-24 Patrick Palka <ppalka@redhat.com>
10946 * real.c (encode_ieee_double): Avoid unwanted sign extension.
10947 (encode_ieee_quad): Likewise.
10949 2021-09-24 Vladimir Makarov <vmakarov@redhat.com>
10951 PR rtl-optimization/102147
10952 * ira-build.c (ira_conflict_vector_profitable_p): Make
10953 profitability calculation independent of host compiler pointer and
10954 IRA_INT_BITS sizes.
10956 2021-09-24 Aldy Hernandez <aldyh@redhat.com>
10958 * gimple-range-path.cc (path_range_query::path_range_query):
10959 Move debugging header...
10960 (path_range_query::precompute_ranges): ...here.
10961 (path_range_query::internal_range_of_expr): Do not call
10962 range_on_path_entry if NAME is defined in the current block.
10964 2021-09-24 Richard Biener <rguenther@suse.de>
10966 * cfghooks.c (verify_flow_info): Verify unallocated BB and
10967 edge flags are not set.
10969 2021-09-24 Aldy Hernandez <aldyh@redhat.com>
10971 * tree-ssa-threadupdate.c (jt_path_registry::cancel_invalid_paths):
10973 (jt_path_registry::register_jump_thread): Call
10974 cancel_invalid_paths.
10975 * tree-ssa-threadupdate.h (class jt_path_registry): Add
10976 cancel_invalid_paths.
10978 2021-09-24 Feng Xue <fxue@os.amperecomputing.com>
10980 PR tree-optimization/102400
10981 * tree-ssa-sccvn.c (vn_reference_insert_pieces): Initialize
10982 result_vdef to zero value.
10984 2021-09-24 Feng Xue <fxue@os.amperecomputing.com>
10986 PR tree-optimization/102451
10987 * tree-ssa-dse.c (delete_dead_or_redundant_call): Record bb of stmt
10990 2021-09-24 Hongyu Wang <hongyu.wang@intel.com>
10992 * config/i386/sse.md (cond_<insn><mode>): Extend to support
10994 (cond_mul<mode>): Likewise.
10995 (cond_div<mode>): Likewise.
10996 (cond_<code><mode>): Likewise.
10997 (cond_fma<mode>): Likewise.
10998 (cond_fms<mode>): Likewise.
10999 (cond_fnma<mode>): Likewise.
11000 (cond_fnms<mode>): Likewise.
11002 2021-09-23 Andrew MacLeod <amacleod@redhat.com>
11004 PR tree-optimization/102463
11005 * gimple-range-fold.cc (fold_using_range::relation_fold_and_or): If
11006 there is no range-ops handler, don't look for a relation.
11008 2021-09-23 Andrew MacLeod <amacleod@redhat.com>
11010 * gimple-range-cache.cc (ranger_cache::ranger_cache): Take
11011 non-executable_edge flag as parameter.
11012 * gimple-range-cache.h (ranger_cache): Adjust prototype.
11013 * gimple-range-gori.cc (gori_compute::gori_compute): Take
11014 non-executable_edge flag as parameter.
11015 (gori_compute::outgoing_edge_range_p): Check new flag.
11016 * gimple-range-gori.h (gori_compute): Adjust prototype.
11017 * gimple-range.cc (gimple_ranger::gimple_ranger): Create new flag.
11018 (gimple_ranger::range_on_edge): Check new flag.
11019 * gimple-range.h (gimple_ranger::non_executable_edge_flag): New.
11020 * gimple-ssa-evrp.c (rvrp_folder): Pass ranger flag to simplifer.
11021 (hybrid_folder::hybrid_folder): Set ranger non-executable flag value.
11022 (hybrid_folder::fold_stmt): Set flag value in the simplifer.
11023 * vr-values.c (simplify_using_ranges::set_and_propagate_unexecutable):
11024 Use not_executable flag if provided inmstead of EDGE_EXECUTABLE.
11025 (simplify_using_ranges::simplify_switch_using_ranges): Clear
11026 EDGE_EXECUTABLE like it originally did.
11027 (simplify_using_ranges::cleanup_edges_and_switches): Clear any
11028 NON_EXECUTABLE flags.
11029 (simplify_using_ranges::simplify_using_ranges): Adjust.
11030 * vr-values.h (class simplify_using_ranges): Adjust.
11031 (simplify_using_ranges::set_range_query): Add non-executable flag param.
11033 2021-09-23 Bill Schmidt <wschmidt@linux.ibm.com>
11036 * config/rs6000/rs6000-call.c (rs6000_aggregate_candidate): Detect
11037 zero-width bit fields and return indicator.
11038 (rs6000_discover_homogeneous_aggregate): Diagnose when the
11039 presence of a zero-width bit field changes parameter passing in
11042 2021-09-23 Aldy Hernandez <aldyh@redhat.com>
11044 * gimple-range-fold.cc (fold_using_range::range_of_phi):
11045 Remove dominator check.
11047 2021-09-23 Aldy Hernandez <aldyh@redhat.com>
11049 * gimple-range-path.cc (path_range_query::precompute_relations):
11050 Hoist edge calculations before using EDGE_SUCC.
11052 2021-09-23 Jonathan Wakely <jwakely@redhat.com>
11054 * configure.ac: Fix --with-multilib-list description.
11055 * configure: Regenerate.
11057 2021-09-23 Richard Biener <rguenther@suse.de>
11059 PR tree-optimization/102448
11060 * tree-vect-data-refs.c (vect_duplicate_ssa_name_ptr_info):
11061 Clear alignment info copied from DR_PTR_INFO.
11063 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
11065 * config/i386/i386-expand.c (ix86_use_mask_cmp_p): Enable
11067 * config/i386/sse.md (sseintvecmodelower): Add HF vector modes.
11068 (<avx512>_store<mode>_mask): Extend to support HF vector modes.
11069 (vec_cmp<mode><avx512fmaskmodelower>): Likewise.
11070 (vcond_mask_<mode><avx512fmaskmodelower>): Likewise.
11071 (vcond<mode><mode>): New expander.
11072 (vcond<mode><sseintvecmodelower>): Likewise.
11073 (vcond<sseintvecmodelower><mode>): Likewise.
11074 (vcondu<mode><sseintvecmodelower>): Likewise.
11076 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
11078 * config/i386/sse.md (extend<ssePHmodelower><mode>2):
11080 (extendv4hf<mode>2): Likewise.
11081 (extendv2hfv2df2): Likewise.
11082 (trunc<mode><ssePHmodelower>2): Likewise.
11083 (avx512fp16_vcvt<castmode>2ph_<mode>): Rename to ...
11084 (trunc<mode>v4hf2): ... this, and drop constraints.
11085 (avx512fp16_vcvtpd2ph_v2df): Rename to ...
11086 (truncv2dfv2hf2): ... this, and likewise.
11088 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
11090 * config/i386/sse.md (float<floatunssuffix><mode><ssePHmodelower>2):
11092 (avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>):
11094 (float<floatunssuffix><mode>v4hf2): ... this, and drop constraints.
11095 (avx512fp16_vcvt<floatsuffix>qq2ph_v2di): Rename to ...
11096 (float<floatunssuffix>v2div2hf2): ... this, and likewise.
11098 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
11100 * config/i386/i386.md (fix<fixunssuffix>_trunchf<mode>2): New expander.
11101 (fixuns_trunchfhi2): Likewise.
11102 (*fixuns_trunchfsi2zext): New define_insn.
11103 * config/i386/sse.md (ssePHmodelower): New mode_attr.
11104 (fix<fixunssuffix>_trunc<ssePHmodelower><mode>2):
11105 New expander for same element vector fix_truncate.
11106 (fix<fixunssuffix>_trunc<ssePHmodelower><mode>2):
11107 Likewise for V4HF to V4SI/V4DI fix_truncate.
11108 (fix<fixunssuffix>_truncv2hfv2di2):
11109 Likeise for V2HF to V2DI fix_truncate.
11111 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
11113 * config/i386/i386.md (<code>hf3): New expander.
11115 2021-09-23 liuhongt <hongtao.liu@intel.com>
11117 * config/i386/sse.md (FMAMODEM): extend to handle FP16.
11118 (VFH_SF_AVX512VL): Extend to handle HFmode.
11119 (VF_SF_AVX512VL): Deleted.
11121 2021-09-23 liuhongt <hongtao.liu@intel.com>
11123 * config/i386/i386.md (rinthf2): New expander.
11124 (nearbyinthf2): New expander.
11126 2021-09-23 Aldy Hernandez <aldyh@redhat.com>
11128 * tree-ssa-dom.c (class dom_jump_threader_simplifier): Rename...
11129 (class dom_jt_state): ...this and provide virtual overrides.
11130 (dom_jt_state::register_equiv): New.
11131 (class dom_jt_simplifier): Rename from
11132 dom_jump_threader_simplifier.
11133 (dom_jump_threader_simplifier::simplify): Rename...
11134 (dom_jt_simplifier::simplify): ...to this.
11135 (pass_dominator::execute): Use dom_jt_simplifier and
11137 * tree-ssa-threadedge.c (jump_threader::jump_threader):
11139 (jt_state::register_equivs_stmt): Abstract out...
11140 (jump_threader::record_temporary_equivalences_from_stmts_at_dest):
11142 (jump_threader::thread_around_empty_blocks): Update state.
11143 (jump_threader::thread_through_normal_block): Same.
11144 (jt_state::jt_state): Remove.
11145 (jt_state::push): Remove pass specific bits. Keep block vector
11147 (jt_state::append_path): New.
11148 (jt_state::pop): Remove pass specific bits.
11149 (jt_state::register_equiv): Same.
11150 (jt_state::record_ranges_from_stmt): Same.
11151 (jt_state::register_equivs_on_edge): Same. Rename...
11152 (jt_state::register_equivs_edge): ...to this.
11153 (jt_state::dump): New.
11154 (jt_state::debug): New.
11155 (jump_threader_simplifier::simplify): Remove.
11156 (jt_state::get_path): New.
11157 * tree-ssa-threadedge.h (class jt_simplifier): Make into a base
11158 class. Expose common functionality as virtual methods.
11159 (class jump_threader_simplifier): Same. Rename...
11160 (class jt_simplifier): ...to this.
11161 * tree-vrp.c (class vrp_jump_threader_simplifier): Rename...
11162 (class vrp_jt_simplifier): ...to this. Provide pass specific
11164 (class vrp_jt_state): New.
11165 (vrp_jump_threader_simplifier::simplify): Rename...
11166 (vrp_jt_simplifier::simplify): ...to this. Inline code from
11167 what used to be the base class.
11168 (vrp_jump_threader::vrp_jump_threader): Use vrp_jt_state and
11171 2021-09-22 Tobias Burnus <tobias@codesourcery.com>
11174 * doc/invoke.texi (-Wno-missing-include-dirs.): Document Fortran
11177 2021-09-22 Roger Sayle <roger@nextmovesoftware.com>
11178 Richard Biener <rguenther@suse.de>
11180 * match.pd (negation simplifications): Implement some negation
11181 folding transformations from fold-const.c's fold_negate_expr.
11182 * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Add a SIMPLIFY
11183 argument, to control whether the op should be simplified prior
11184 to looking up/assigning a value number.
11185 (vn_nary_build_or_lookup): Update call to vn_nary_build_or_lookup_1.
11186 (vn_nary_simplify): Likewise.
11187 (visit_nary_op): Likewise, but when constructing a NEGATE_EXPR
11188 now call vn_nary_build_or_lookup_1 disabling simplification.
11190 2021-09-22 Jiufu Guo <guojiufu@linux.ibm.com>
11192 PR tree-optimization/102087
11193 * tree-ssa-loop-niter.c (number_of_iterations_until_wrap):
11194 Update bound/cmp/control for niter.
11196 2021-09-22 Aldy Hernandez <aldyh@redhat.com>
11198 * gimple-range-fold.cc (fold_using_range::range_of_range_op):
11199 Move check for non-empty BB here.
11200 (fur_source::register_outgoing_edges): ...from here.
11202 2021-09-22 Aldy Hernandez <aldyh@redhat.com>
11204 * gimple-range-path.cc (path_range_query::internal_range_of_expr):
11205 Remove call to improve_range_with_equivs.
11206 (path_range_query::improve_range_with_equivs): Remove
11207 * gimple-range-path.h: Remove improve_range_with_equivs.
11209 2021-09-22 dianhong xu <dianhong.xu@intel.com>
11211 * config/i386/avx512fp16intrin.h:
11212 (_mm512_mask_blend_ph): New intrinsic.
11213 (_mm512_permutex2var_ph): Ditto.
11214 (_mm512_permutexvar_ph): Ditto.
11215 * config/i386/avx512fp16vlintrin.h:
11216 (_mm256_mask_blend_ph): New intrinsic.
11217 (_mm256_permutex2var_ph): Ditto.
11218 (_mm256_permutexvar_ph): Ditto.
11219 (_mm_mask_blend_ph): Ditto.
11220 (_mm_permutex2var_ph): Ditto.
11221 (_mm_permutexvar_ph): Ditto.
11223 2021-09-22 dianhong xu <dianhong.xu@intel.com>
11225 * config/i386/avx512fp16intrin.h: Add new intrinsics.
11226 (_mm512_conj_pch): New intrinsic.
11227 (_mm512_mask_conj_pch): Ditto.
11228 (_mm512_maskz_conj_pch): Ditto.
11229 * config/i386/avx512fp16vlintrin.h: Add new intrinsics.
11230 (_mm256_conj_pch): New intrinsic.
11231 (_mm256_mask_conj_pch): Ditto.
11232 (_mm256_maskz_conj_pch): Ditto.
11233 (_mm_conj_pch): Ditto.
11234 (_mm_mask_conj_pch): Ditto.
11235 (_mm_maskz_conj_pch): Ditto.
11237 2021-09-22 dianhong xu <dianhong.xu@intel.com>
11239 * config/i386/avx512fp16intrin.h (_MM512_REDUCE_OP): New macro
11240 (_mm512_reduce_add_ph): New intrinsic.
11241 (_mm512_reduce_mul_ph): Ditto.
11242 (_mm512_reduce_min_ph): Ditto.
11243 (_mm512_reduce_max_ph): Ditto.
11244 * config/i386/avx512fp16vlintrin.h
11245 (_MM256_REDUCE_OP/_MM_REDUCE_OP): New macro.
11246 (_mm256_reduce_add_ph): New intrinsic.
11247 (_mm256_reduce_mul_ph): Ditto.
11248 (_mm256_reduce_min_ph): Ditto.
11249 (_mm256_reduce_max_ph): Ditto.
11250 (_mm_reduce_add_ph): Ditto.
11251 (_mm_reduce_mul_ph): Ditto.
11252 (_mm_reduce_min_ph): Ditto.
11253 (_mm_reduce_max_ph): Ditto.
11255 2021-09-22 dianhong xu <dianhong.xu@intel.com>
11257 * config/i386/avx512fp16intrin.h (__m512h_u, __m256h_u,
11258 __m128h_u): New typedef.
11259 (_mm512_load_ph): New intrinsic.
11260 (_mm256_load_ph): Ditto.
11261 (_mm_load_ph): Ditto.
11262 (_mm512_loadu_ph): Ditto.
11263 (_mm256_loadu_ph): Ditto.
11264 (_mm_loadu_ph): Ditto.
11265 (_mm512_store_ph): Ditto.
11266 (_mm256_store_ph): Ditto.
11267 (_mm_store_ph): Ditto.
11268 (_mm512_storeu_ph): Ditto.
11269 (_mm256_storeu_ph): Ditto.
11270 (_mm_storeu_ph): Ditto.
11271 (_mm512_abs_ph): Ditto.
11272 * config/i386/avx512fp16vlintrin.h
11273 (_mm_abs_ph): Ditto.
11274 (_mm256_abs_ph): Ditto.
11276 2021-09-22 Andreas Krebbel <krebbel@linux.ibm.com>
11278 * config/s390/tpf.md (prologue_tpf, epilogue_tpf): Add cc clobber.
11280 2021-09-22 Andreas Krebbel <krebbel@linux.ibm.com>
11283 * config/s390/s390.c (s390_expand_insv): Emit a normal move if it
11284 is actually a full copy of the source operand into the target.
11285 Don't emit a strict low part move if source and target mode match.
11287 2021-09-22 Jakub Jelinek <jakub@redhat.com>
11289 PR middle-end/102415
11290 * omp-expand.c (expand_omp_single): If region->exit is NULL,
11291 assert region->entry is GIMPLE_OMP_SCOPE region and return.
11293 2021-09-22 Jakub Jelinek <jakub@redhat.com>
11295 * tree.h (OMP_CLAUSE_ALLOCATE_ALIGN): Define.
11296 * tree.c (omp_clause_num_ops): Change number of OMP_CLAUSE_ALLOCATE
11297 arguments from 2 to 3.
11298 * tree-pretty-print.c (dump_omp_clause): Print allocator() around
11299 allocate clause allocator and print align if present.
11300 * omp-low.c (scan_sharing_clauses): Force allocate_map entry even
11301 for omp_default_mem_alloc if align modifier is present. If align
11302 modifier is present, use TREE_LIST to encode both allocator and
11304 (lower_private_allocate, lower_rec_input_clauses, create_task_copyfn):
11305 Handle align modifier on allocator clause if present.
11307 2021-09-22 liuhongt <hongtao.liu@intel.com>
11309 * config/i386/i386.md (define_attr "isa"): Add
11311 (define_attr "enabled"): Correspond fma_or_avx512vl to
11312 TARGET_FMA || TARGET_AVX512VL.
11313 * config/i386/mmx.md (fmav2sf4): Extend to AVX512 fma.
11315 (fnmav2sf4): Ditto.
11316 (fnmsv2sf4): Ditto.
11318 2021-09-22 liuhongt <hongtao.liu@intel.com>
11320 * config/i386/i386.md (cstorehf3): New define_expand.
11322 2021-09-22 liuhongt <hongtao.liu@intel.com>
11324 * config/i386/i386.md (<rounding_insn>hf2): New expander.
11325 (sse4_1_round<mode>2): Extend from MODEF to MODEFH.
11326 * config/i386/sse.md (*sse4_1_round<ssescalarmodesuffix>):
11327 Extend from VF_128 to VFH_128.
11329 2021-09-22 liuhongt <hongtao.liu@intel.com>
11331 * config/i386/i386-features.c (i386-features.c): Handle
11333 * config/i386/i386.md (sqrthf2): New expander.
11334 (*sqrthf2): New define_insn.
11335 * config/i386/sse.md
11336 (*<sse>_vmsqrt<mode>2<mask_scalar_name><round_scalar_name>):
11339 2021-09-22 liuhongt <hongtao.liu@intel.com>
11341 * config/i386/avx512fp16intrin.h (_mm_mask_fcmadd_sch):
11343 (_mm_mask3_fcmadd_sch): Likewise.
11344 (_mm_maskz_fcmadd_sch): Likewise.
11345 (_mm_fcmadd_sch): Likewise.
11346 (_mm_mask_fmadd_sch): Likewise.
11347 (_mm_mask3_fmadd_sch): Likewise.
11348 (_mm_maskz_fmadd_sch): Likewise.
11349 (_mm_fmadd_sch): Likewise.
11350 (_mm_mask_fcmadd_round_sch): Likewise.
11351 (_mm_mask3_fcmadd_round_sch): Likewise.
11352 (_mm_maskz_fcmadd_round_sch): Likewise.
11353 (_mm_fcmadd_round_sch): Likewise.
11354 (_mm_mask_fmadd_round_sch): Likewise.
11355 (_mm_mask3_fmadd_round_sch): Likewise.
11356 (_mm_maskz_fmadd_round_sch): Likewise.
11357 (_mm_fmadd_round_sch): Likewise.
11358 (_mm_fcmul_sch): Likewise.
11359 (_mm_mask_fcmul_sch): Likewise.
11360 (_mm_maskz_fcmul_sch): Likewise.
11361 (_mm_fmul_sch): Likewise.
11362 (_mm_mask_fmul_sch): Likewise.
11363 (_mm_maskz_fmul_sch): Likewise.
11364 (_mm_fcmul_round_sch): Likewise.
11365 (_mm_mask_fcmul_round_sch): Likewise.
11366 (_mm_maskz_fcmul_round_sch): Likewise.
11367 (_mm_fmul_round_sch): Likewise.
11368 (_mm_mask_fmul_round_sch): Likewise.
11369 (_mm_maskz_fmul_round_sch): Likewise.
11370 * config/i386/i386-builtin.def: Add corresponding new builtins.
11371 * config/i386/sse.md
11372 (avx512fp16_fmaddcsh_v8hf_maskz<round_expand_name>): New expander.
11373 (avx512fp16_fcmaddcsh_v8hf_maskz<round_expand_name>): Ditto.
11374 (avx512fp16_fma_<complexopname>sh_v8hf<mask_scalarcz_name><round_scalarcz_name>):
11376 (avx512fp16_<complexopname>sh_v8hf_mask<round_name>): Ditto.
11377 (avx512fp16_<complexopname>sh_v8hf<mask_scalarc_name><round_scalarcz_name>):
11379 * config/i386/subst.md (mask_scalarcz_name): New.
11380 (mask_scalarc_name): Ditto.
11381 (mask_scalarc_operand3): Ditto.
11382 (mask_scalarcz_operand4): Ditto.
11383 (round_scalarcz_name): Ditto.
11384 (round_scalarc_mask_operand3): Ditto.
11385 (round_scalarcz_mask_operand4): Ditto.
11386 (round_scalarc_mask_op3): Ditto.
11387 (round_scalarcz_mask_op4): Ditto.
11388 (round_scalarcz_constraint): Ditto.
11389 (round_scalarcz_nimm_predicate): Ditto.
11390 (mask_scalarcz): Ditto.
11391 (mask_scalarc): Ditto.
11392 (round_scalarcz): Ditto.
11394 2021-09-22 liuhongt <hongtao.liu@intel.com>
11396 * config/i386/avx512fp16intrin.h (_mm512_fcmadd_pch):
11398 (_mm512_mask_fcmadd_pch): Likewise.
11399 (_mm512_mask3_fcmadd_pch): Likewise.
11400 (_mm512_maskz_fcmadd_pch): Likewise.
11401 (_mm512_fmadd_pch): Likewise.
11402 (_mm512_mask_fmadd_pch): Likewise.
11403 (_mm512_mask3_fmadd_pch): Likewise.
11404 (_mm512_maskz_fmadd_pch): Likewise.
11405 (_mm512_fcmadd_round_pch): Likewise.
11406 (_mm512_mask_fcmadd_round_pch): Likewise.
11407 (_mm512_mask3_fcmadd_round_pch): Likewise.
11408 (_mm512_maskz_fcmadd_round_pch): Likewise.
11409 (_mm512_fmadd_round_pch): Likewise.
11410 (_mm512_mask_fmadd_round_pch): Likewise.
11411 (_mm512_mask3_fmadd_round_pch): Likewise.
11412 (_mm512_maskz_fmadd_round_pch): Likewise.
11413 (_mm512_fcmul_pch): Likewise.
11414 (_mm512_mask_fcmul_pch): Likewise.
11415 (_mm512_maskz_fcmul_pch): Likewise.
11416 (_mm512_fmul_pch): Likewise.
11417 (_mm512_mask_fmul_pch): Likewise.
11418 (_mm512_maskz_fmul_pch): Likewise.
11419 (_mm512_fcmul_round_pch): Likewise.
11420 (_mm512_mask_fcmul_round_pch): Likewise.
11421 (_mm512_maskz_fcmul_round_pch): Likewise.
11422 (_mm512_fmul_round_pch): Likewise.
11423 (_mm512_mask_fmul_round_pch): Likewise.
11424 (_mm512_maskz_fmul_round_pch): Likewise.
11425 * config/i386/avx512fp16vlintrin.h (_mm_fmadd_pch):
11427 (_mm_mask_fmadd_pch): Likewise.
11428 (_mm_mask3_fmadd_pch): Likewise.
11429 (_mm_maskz_fmadd_pch): Likewise.
11430 (_mm256_fmadd_pch): Likewise.
11431 (_mm256_mask_fmadd_pch): Likewise.
11432 (_mm256_mask3_fmadd_pch): Likewise.
11433 (_mm256_maskz_fmadd_pch): Likewise.
11434 (_mm_fcmadd_pch): Likewise.
11435 (_mm_mask_fcmadd_pch): Likewise.
11436 (_mm_mask3_fcmadd_pch): Likewise.
11437 (_mm_maskz_fcmadd_pch): Likewise.
11438 (_mm256_fcmadd_pch): Likewise.
11439 (_mm256_mask_fcmadd_pch): Likewise.
11440 (_mm256_mask3_fcmadd_pch): Likewise.
11441 (_mm256_maskz_fcmadd_pch): Likewise.
11442 (_mm_fmul_pch): Likewise.
11443 (_mm_mask_fmul_pch): Likewise.
11444 (_mm_maskz_fmul_pch): Likewise.
11445 (_mm256_fmul_pch): Likewise.
11446 (_mm256_mask_fmul_pch): Likewise.
11447 (_mm256_maskz_fmul_pch): Likewise.
11448 (_mm_fcmul_pch): Likewise.
11449 (_mm_mask_fcmul_pch): Likewise.
11450 (_mm_maskz_fcmul_pch): Likewise.
11451 (_mm256_fcmul_pch): Likewise.
11452 (_mm256_mask_fcmul_pch): Likewise.
11453 (_mm256_maskz_fcmul_pch): Likewise.
11454 * config/i386/i386-builtin-types.def (V8HF_FTYPE_V8HF_V8HF_V8HF,
11455 V8HF_FTYPE_V16HF_V16HF_V16HF, V16HF_FTYPE_V16HF_V16HF_V16HF_UQI,
11456 V32HF_FTYPE_V32HF_V32HF_V32HF_INT,
11457 V32HF_FTYPE_V32HF_V32HF_V32HF_UHI_INT): Add new builtin types.
11458 * config/i386/i386-builtin.def: Add new builtins.
11459 * config/i386/i386-expand.c: Handle new builtin types.
11460 * config/i386/subst.md (SUBST_CV): New.
11461 (maskc_name): Ditto.
11462 (maskc_operand3): Ditto.
11464 (sdc_maskz_name): Ditto.
11465 (sdc_mask_op4): Ditto.
11466 (sdc_mask_op5): Ditto.
11467 (sdc_mask_mode512bit_condition): Ditto.
11469 (round_maskc_operand3): Ditto.
11470 (round_sdc_mask_operand4): Ditto.
11471 (round_maskc_op3): Ditto.
11472 (round_sdc_mask_op4): Ditto.
11473 (round_saeonly_sdc_mask_operand5): Ditto.
11474 * config/i386/sse.md (unspec): Add complex fma unspecs.
11475 (avx512fmaskcmode): New.
11476 (UNSPEC_COMPLEX_F_C_MA): Ditto.
11477 (UNSPEC_COMPLEX_F_C_MUL): Ditto.
11478 (complexopname): Ditto.
11479 (<avx512>_fmaddc_<mode>_maskz<round_expand_name>): New expander.
11480 (<avx512>_fcmaddc_<mode>_maskz<round_expand_name>): Ditto.
11481 (fma_<complexopname>_<mode><sdc_maskz_name><round_name>): New
11483 (<avx512>_<complexopname>_<mode>_mask<round_name>): Ditto.
11484 (<avx512>_<complexopname>_<mode><maskc_name><round_name>): Ditto.
11486 2021-09-22 Kewen Lin <linkw@linux.ibm.com>
11488 * config/rs6000/rs6000.opt (rs6000-density-pct-threshold,
11489 rs6000-density-size-threshold, rs6000-density-penalty,
11490 rs6000-density-load-pct-threshold,
11491 rs6000-density-load-num-threshold): New parameter.
11492 * config/rs6000/rs6000.c (rs6000_density_test): Adjust with
11493 corresponding parameters.
11495 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
11497 * gimple-range-path.cc (path_range_query::defined_outside_path):
11499 (path_range_query::range_on_path_entry): New.
11500 (path_range_query::internal_range_of_expr): Resolve unknowns
11502 (path_range_query::improve_range_with_equivs): New.
11503 (path_range_query::ssa_range_in_phi): Resolve unknowns with
11505 * gimple-range-path.h (class path_range_query): Add
11506 defined_outside_path, range_on_path_entry, and
11507 improve_range_with_equivs.
11509 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
11511 * gimple-range-path.cc (path_range_query::add_to_imports): New.
11512 (path_range_query::add_copies_to_imports): New.
11513 (path_range_query::precompute_ranges): Call
11514 add_copies_to_imports.
11515 * gimple-range-path.h (class path_range_query): Add prototypes
11516 for add_copies_to_imports and add_to_imports.
11518 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
11520 * gimple-range-path.cc (path_range_query::range_defined_in_block):
11521 Remove useless code.
11523 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
11525 * gimple-range-fold.h (class fur_source): Make oracle protected.
11526 * gimple-range-path.cc (path_range_query::path_range_query): Add
11527 resolve argument. Initialize oracle.
11528 (path_range_query::~path_range_query): Delete oracle.
11529 (path_range_query::range_of_stmt): Adapt to use relations.
11530 (path_range_query::precompute_ranges): Pre-compute relations.
11531 (class jt_fur_source): New
11532 (jt_fur_source::jt_fur_source): New.
11533 (jt_fur_source::register_relation): New.
11534 (jt_fur_source::query_relation): New.
11535 (path_range_query::precompute_relations): New.
11536 (path_range_query::precompute_phi_relations): New.
11537 * gimple-range-path.h (path_range_query): Add resolve argument.
11538 Add oracle, precompute_relations, precompute_phi_relations.
11539 * tree-ssa-threadbackward.c (back_threader::back_threader): Pass
11540 resolve argument to solver.
11542 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
11544 * gimple-range-fold.cc (fold_using_range::range_of_range_op):
11545 Rename postfold_gcond_edges to register_outgoing_edges and
11547 (fold_using_range::postfold_gcond_edges): Rename...
11548 (fur_source::register_outgoing_edges): ...to this.
11549 * gimple-range-fold.h (postfold_gcond_edges): Rename to
11550 register_outgoing_edges and move to fur_source.
11552 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
11554 * gimple-range-fold.cc (fold_using_range::range_of_phi): Check
11555 dom_info_available_p.
11557 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
11559 * gimple-range-cache.cc (non_null_ref::non_null_ref): Use create
11560 and quick_grow_cleared instead of safe_grow_cleared.
11562 2021-09-21 Thomas Schwinge <thomas@codesourcery.com>
11565 * omp-oacc-neuter-broadcast.cc (oacc_do_neutering): Evaluate
11566 'random ()' to '0'.
11568 2021-09-21 Richard Earnshaw <rearnsha@arm.com>
11570 * configure.ac: Detect when the assembler supports new-style
11571 architecture extensions.
11572 * common/config/arm/arm-common.c (arm_rewrite_mcpu): Return
11573 the full CPU string if the assembler can grok it.
11574 (arm_rewrite_march): Likewise but for the architecture.
11575 * config.in: Regenerate.
11576 * configure: Regenerate.
11578 2021-09-21 Richard Biener <rguenther@suse.de>
11580 PR tree-optimization/102421
11581 * tree-vect-loop.c (vect_dissolve_slp_only_groups): Copy and
11582 adjust alignment info.
11584 2021-09-21 Kewen Lin <linkw@linux.ibm.com>
11586 * ipa-fnsummary.c (ipa_fn_summary_write): Remove inconsistent
11587 bitfield stream out.
11589 2021-09-20 Andrew MacLeod <amacleod@redhat.com>
11591 * gimple-range-fold.cc (fold_using_range::range_of_phi): Ignore
11592 undefined edges, apply an equivalence if appropriate.
11593 * gimple-range-gori.cc (gori_compute::outgoing_edge_range_p): Return
11594 UNDEFINED if EDGE_EXECUTABLE is not set.
11595 * gimple-range.cc (gimple_ranger::gimple_ranger): Set all edges
11596 as EXECUTABLE upon startup.
11597 (gimple_ranger::range_on_edge): Return UNDEFINED for edges without
11598 EDGE_EXECUTABLE set.
11599 * vr-values.c (set_and_propagate_unexecutable): New.
11600 (simplify_using_ranges::fold_cond): Call set_and_propagate.
11601 (simplify_using_ranges::simplify_switch_using_ranges): Ditto.
11602 * vr-values.h: Add prototype.
11604 2021-09-20 Andrew MacLeod <amacleod@redhat.com>
11606 * value-relation.cc (equiv_oracle::register_initial_def): New.
11607 (equiv_oracle::register_relation): Call register_initial_def.
11608 (equiv_oracle::add_equiv_to_block): New. Split register_relation.
11609 (relation_oracle::register_stmt): Check def block of PHI arguments.
11610 * value-relation.h (equiv_oracle): Add new prototypes.
11612 2021-09-20 Matthias Kretz <m.kretz@gsi.de>
11614 * cppbuiltin.c (define_builtin_macros_for_compilation_flags):
11615 Define __RECIPROCAL_MATH__, __NO_SIGNED_ZEROS__,
11616 __NO_TRAPPING_MATH__, __ASSOCIATIVE_MATH__, and
11617 __ROUNDING_MATH__ according to their corresponding flags.
11618 * doc/cpp.texi: Document __RECIPROCAL_MATH__,
11619 __NO_SIGNED_ZEROS__, __NO_TRAPPING_MATH__, __ASSOCIATIVE_MATH__,
11620 and __ROUNDING_MATH__.
11622 2021-09-20 Richard Biener <rguenther@suse.de>
11624 * tree-vect-stmts.c (vectorizable_load): Use the vectype
11627 2021-09-20 Richard Biener <rguenther@suse.de>
11629 * tree-vect-data-refs.c (vect_duplicate_ssa_name_ptr_info):
11630 Do not compute alignment of the vectorized access here.
11632 2021-09-20 Richard Biener <rguenther@suse.de>
11634 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment):
11635 Store -1 for runtime alias peeling iterations.
11637 2021-09-20 Richard Biener <rguenther@suse.de>
11639 * config.gcc: Obsolete hppa[12]*-*-hpux10* and hppa[12]*-*-hpux11*.
11641 2021-09-20 Thomas Schwinge <thomas@codesourcery.com>
11643 * input.c (string_concat_db::record_string_concatenation)
11644 (string_concat_db::get_string_concatenation): Skip for
11645 'RESERVED_LOCATION_P'.
11647 2021-09-20 Richard Biener <rguenther@suse.de>
11649 PR tree-optimization/65206
11650 * tree-data-ref.h (struct data_reference): Add alt_indices,
11652 * tree-data-ref.c (free_data_ref): Release alt_indices.
11653 (dr_analyze_indices): Work on struct indices and get DR_REF as tree.
11654 (create_data_ref): Adjust.
11655 (initialize_data_dependence_relation): Split into head
11656 and tail. When the base objects fail to match up try
11657 again with pointer-based analysis of indices.
11658 * tree-vectorizer.c (vec_info_shared::check_datarefs): Do
11659 not compare the lazily computed alternate set of indices.
11661 2021-09-20 Iain Sandoe <iain@sandoe.co.uk>
11663 * gcc.c: Test for execute OK when we find the
11664 programs for assembler linker and dsymutil and those
11665 were specified at configure-time.
11667 2021-09-19 Martin Sebor <msebor@redhat.com>
11669 PR middle-end/102403
11670 * gimple-predicate-analysis.cc (predicate::init_from_control_deps):
11671 Correct a function pre/postcondition.
11673 2021-09-19 Martin Sebor <msebor@redhat.com>
11675 PR middle-end/102243
11676 * tree-ssa-strlen.c (get_range): Handle null cfun.
11678 2021-09-19 Iain Sandoe <iain@sandoe.co.uk>
11680 * config/darwin.h (LINK_COMMAND_SPEC_A): Use Darwin10
11681 unwinder shim as a convenience library.
11683 2021-09-19 Andrew Pinski <apinski@marvell.com>
11685 * doc/install.texi: Add note about
11686 binutils 2.35 is required for LTO usage.
11688 2021-09-19 Aldy Hernandez <aldyh@redhat.com>
11690 * tree-ssa-threadbackward.c
11691 (back_threader_registry::register_path): Use push_edge.
11692 * tree-ssa-threadedge.c
11693 (jump_threader::thread_around_empty_blocks): Same.
11694 (jump_threader::thread_through_normal_block): Same.
11695 (jump_threader::thread_across_edge): Same. Also, use auto_bitmap.
11697 * tree-ssa-threadupdate.c
11698 (jt_path_registry::allocate_thread_edge): Remove.
11699 (jt_path_registry::push_edge): New.
11700 (dump_jump_thread_path): Make static.
11701 * tree-ssa-threadupdate.h (allocate_thread_edge): Remove.
11704 2021-09-19 Aldy Hernandez <aldyh@redhat.com>
11706 * gimple-range-path.cc (path_range_query::path_range_query): Add
11708 (path_range_query::dump): Remove extern declaration of dump_ranger.
11709 * gimple-range-trace.cc (dump_ranger): Add DEBUG_FUNCTION marker.
11710 * gimple-range-trace.h (dump_ranger): Add prototype.
11712 2021-09-19 John Ericson <git@JohnEricson.me>
11714 * gcc.c (find_a_program): New function, factored out of...
11715 (find_a_file): Here.
11716 (execute): Use find_a_program when looking for programs rather
11719 2021-09-19 Matwey V. Kornilov <matwey.kornilov@gmail.com>
11721 * config/avr/avr-mcus.def: Add atmega324pb.
11722 * doc/avr-mmcu.texi: Corresponding changes.
11724 2021-09-19 Roger Sayle <roger@nextmovesoftware.com>
11726 PR middle-end/88173
11727 * match.pd (cmp @0 REAL_CST@1): When @0 is also REAL_CST, apply
11728 the same transformations as to @1. For comparisons against NaN,
11729 don't check HONOR_SNANS but confirm that neither operand is a
11732 2021-09-19 Benjamin Peterson <benjamin@locrian.net>
11734 * attribs.c (make_unique_name): Delete.
11735 * attribs.h (make_unique_name): Delete.
11737 2021-09-19 Andrew Pinski <apinski@marvell.com>
11739 * lra-constraints.c (check_and_process_move): Assert
11740 that dclass and sclass are greater than or equal to NO_REGS.
11742 2021-09-18 Jakub Jelinek <jakub@redhat.com>
11744 * tree.h (OMP_CLAUSE_ORDER_UNCONSTRAINED): Define.
11745 * tree-pretty-print.c (dump_omp_clause): Print unconstrained:
11746 for OMP_CLAUSE_ORDER_UNCONSTRAINED.
11748 2021-09-18 liuhongt <hongtao.liu@intel.com>
11750 * config/i386/i386-features.c (remove_partial_avx_dependency):
11751 Restrict TARGET_USE_VECTOR_FP_CONVERTS and
11752 TARGET_USE_VECTOR_CONVERTS to conversion instructions only.
11754 2021-09-18 Jakub Jelinek <jakub@redhat.com>
11756 * gimplify.c (omp_default_clause): For C/C++ default({,first}private),
11757 if file/namespace scope variable doesn't have predetermined sharing,
11758 treat it as if there was default(none).
11760 2021-09-18 liuhongt <hongtao.liu@intel.com>
11762 * config/i386/avx512fp16intrin.h (_mm_fmadd_sh):
11764 (_mm_mask_fmadd_sh): Likewise.
11765 (_mm_mask3_fmadd_sh): Likewise.
11766 (_mm_maskz_fmadd_sh): Likewise.
11767 (_mm_fmadd_round_sh): Likewise.
11768 (_mm_mask_fmadd_round_sh): Likewise.
11769 (_mm_mask3_fmadd_round_sh): Likewise.
11770 (_mm_maskz_fmadd_round_sh): Likewise.
11771 (_mm_fnmadd_sh): Likewise.
11772 (_mm_mask_fnmadd_sh): Likewise.
11773 (_mm_mask3_fnmadd_sh): Likewise.
11774 (_mm_maskz_fnmadd_sh): Likewise.
11775 (_mm_fnmadd_round_sh): Likewise.
11776 (_mm_mask_fnmadd_round_sh): Likewise.
11777 (_mm_mask3_fnmadd_round_sh): Likewise.
11778 (_mm_maskz_fnmadd_round_sh): Likewise.
11779 (_mm_fmsub_sh): Likewise.
11780 (_mm_mask_fmsub_sh): Likewise.
11781 (_mm_mask3_fmsub_sh): Likewise.
11782 (_mm_maskz_fmsub_sh): Likewise.
11783 (_mm_fmsub_round_sh): Likewise.
11784 (_mm_mask_fmsub_round_sh): Likewise.
11785 (_mm_mask3_fmsub_round_sh): Likewise.
11786 (_mm_maskz_fmsub_round_sh): Likewise.
11787 (_mm_fnmsub_sh): Likewise.
11788 (_mm_mask_fnmsub_sh): Likewise.
11789 (_mm_mask3_fnmsub_sh): Likewise.
11790 (_mm_maskz_fnmsub_sh): Likewise.
11791 (_mm_fnmsub_round_sh): Likewise.
11792 (_mm_mask_fnmsub_round_sh): Likewise.
11793 (_mm_mask3_fnmsub_round_sh): Likewise.
11794 (_mm_maskz_fnmsub_round_sh): Likewise.
11795 * config/i386/i386-builtin-types.def
11796 (V8HF_FTYPE_V8HF_V8HF_V8HF_UQI_INT): New builtin type.
11797 * config/i386/i386-builtin.def: Add new builtins.
11798 * config/i386/i386-expand.c: Handle new builtin type.
11799 * config/i386/sse.md (fmai_vmfmadd_<mode><round_name>):
11800 Ajdust to support FP16.
11801 (fmai_vmfmsub_<mode><round_name>): Ditto.
11802 (fmai_vmfnmadd_<mode><round_name>): Ditto.
11803 (fmai_vmfnmsub_<mode><round_name>): Ditto.
11804 (*fmai_fmadd_<mode>): Ditto.
11805 (*fmai_fmsub_<mode>): Ditto.
11806 (*fmai_fnmadd_<mode><round_name>): Ditto.
11807 (*fmai_fnmsub_<mode><round_name>): Ditto.
11808 (avx512f_vmfmadd_<mode>_mask<round_name>): Ditto.
11809 (avx512f_vmfmadd_<mode>_mask3<round_name>): Ditto.
11810 (avx512f_vmfmadd_<mode>_maskz<round_expand_name>): Ditto.
11811 (avx512f_vmfmadd_<mode>_maskz_1<round_name>): Ditto.
11812 (*avx512f_vmfmsub_<mode>_mask<round_name>): Ditto.
11813 (avx512f_vmfmsub_<mode>_mask3<round_name>): Ditto.
11814 (*avx512f_vmfmsub_<mode>_maskz_1<round_name>): Ditto.
11815 (*avx512f_vmfnmsub_<mode>_mask<round_name>): Ditto.
11816 (*avx512f_vmfnmsub_<mode>_mask3<round_name>): Ditto.
11817 (*avx512f_vmfnmsub_<mode>_mask<round_name>): Ditto.
11818 (*avx512f_vmfnmadd_<mode>_mask<round_name>): Renamed to ...
11819 (avx512f_vmfnmadd_<mode>_mask<round_name>) ... this, and
11820 adjust to support FP16.
11821 (avx512f_vmfnmadd_<mode>_mask3<round_name>): Ditto.
11822 (avx512f_vmfnmadd_<mode>_maskz_1<round_name>): Ditto.
11823 (avx512f_vmfnmadd_<mode>_maskz<round_expand_name>): New
11826 2021-09-18 H.J. Lu <hjl.tools@gmail.com>
11828 * config/i386/sse.md (avx512fmaskmodelower): Extend to support
11830 (maskload<mode><avx512fmaskmodelower>): Ditto.
11831 (maskstore<mode><avx512fmaskmodelower>): Ditto.
11833 2021-09-18 H.J. Lu <hjl.tools@gmail.com>
11835 * config/i386/i386-expand.c (ix86_expand_fp_absneg_operator):
11837 (ix86_expand_copysign): Ditto.
11838 (ix86_expand_xorsign): Ditto.
11839 * config/i386/i386.c (ix86_build_const_vector): Handle HF vector
11841 (ix86_build_signbit_mask): Ditto.
11842 (ix86_can_change_mode_class): Ditto.
11843 * config/i386/i386.md
11844 (SSEMODEF): Add HFmode.
11845 (ssevecmodef): Ditto.
11846 (<code>hf2): New define_expand.
11847 (*<code>hf2_1): New define_insn_and_split.
11848 (copysign<mode>): Extend to support HFmode under AVX512FP16.
11849 (xorsign<mode>): Ditto.
11850 * config/i386/sse.md (VFB): New mode iterator.
11851 (VFB_128_256): Ditto.
11853 (sseintvecmode2): Support HF vector mode.
11854 (<code><mode>2): Use new mode iterator.
11855 (*<code><mode>2): Ditto.
11856 (copysign<mode>3): Ditto.
11857 (xorsign<mode>3): Ditto.
11858 (<code><mode>3<mask_name>): Ditto.
11859 (<code><mode>3<mask_name>): Ditto.
11860 (<sse>_andnot<mode>3<mask_name>): Adjust for HF vector mode.
11861 (<sse>_andnot<mode>3<mask_name>): Ditto.
11862 (*<code><mode>3<mask_name>): Ditto.
11863 (*<code><mode>3<mask_name>): Ditto.
11865 2021-09-18 liuhongt <hongtao.liu@intel.com>
11867 * config/i386/avx512fp16intrin.h (_mm512_mask_fmadd_ph):
11869 (_mm512_mask3_fmadd_ph): Likewise.
11870 (_mm512_maskz_fmadd_ph): Likewise.
11871 (_mm512_fmadd_round_ph): Likewise.
11872 (_mm512_mask_fmadd_round_ph): Likewise.
11873 (_mm512_mask3_fmadd_round_ph): Likewise.
11874 (_mm512_maskz_fmadd_round_ph): Likewise.
11875 (_mm512_fnmadd_ph): Likewise.
11876 (_mm512_mask_fnmadd_ph): Likewise.
11877 (_mm512_mask3_fnmadd_ph): Likewise.
11878 (_mm512_maskz_fnmadd_ph): Likewise.
11879 (_mm512_fnmadd_round_ph): Likewise.
11880 (_mm512_mask_fnmadd_round_ph): Likewise.
11881 (_mm512_mask3_fnmadd_round_ph): Likewise.
11882 (_mm512_maskz_fnmadd_round_ph): Likewise.
11883 (_mm512_fmsub_ph): Likewise.
11884 (_mm512_mask_fmsub_ph): Likewise.
11885 (_mm512_mask3_fmsub_ph): Likewise.
11886 (_mm512_maskz_fmsub_ph): Likewise.
11887 (_mm512_fmsub_round_ph): Likewise.
11888 (_mm512_mask_fmsub_round_ph): Likewise.
11889 (_mm512_mask3_fmsub_round_ph): Likewise.
11890 (_mm512_maskz_fmsub_round_ph): Likewise.
11891 (_mm512_fnmsub_ph): Likewise.
11892 (_mm512_mask_fnmsub_ph): Likewise.
11893 (_mm512_mask3_fnmsub_ph): Likewise.
11894 (_mm512_maskz_fnmsub_ph): Likewise.
11895 (_mm512_fnmsub_round_ph): Likewise.
11896 (_mm512_mask_fnmsub_round_ph): Likewise.
11897 (_mm512_mask3_fnmsub_round_ph): Likewise.
11898 (_mm512_maskz_fnmsub_round_ph): Likewise.
11899 * config/i386/avx512fp16vlintrin.h (_mm256_fmadd_ph):
11901 (_mm256_mask_fmadd_ph): Likewise.
11902 (_mm256_mask3_fmadd_ph): Likewise.
11903 (_mm256_maskz_fmadd_ph): Likewise.
11904 (_mm_fmadd_ph): Likewise.
11905 (_mm_mask_fmadd_ph): Likewise.
11906 (_mm_mask3_fmadd_ph): Likewise.
11907 (_mm_maskz_fmadd_ph): Likewise.
11908 (_mm256_fnmadd_ph): Likewise.
11909 (_mm256_mask_fnmadd_ph): Likewise.
11910 (_mm256_mask3_fnmadd_ph): Likewise.
11911 (_mm256_maskz_fnmadd_ph): Likewise.
11912 (_mm_fnmadd_ph): Likewise.
11913 (_mm_mask_fnmadd_ph): Likewise.
11914 (_mm_mask3_fnmadd_ph): Likewise.
11915 (_mm_maskz_fnmadd_ph): Likewise.
11916 (_mm256_fmsub_ph): Likewise.
11917 (_mm256_mask_fmsub_ph): Likewise.
11918 (_mm256_mask3_fmsub_ph): Likewise.
11919 (_mm256_maskz_fmsub_ph): Likewise.
11920 (_mm_fmsub_ph): Likewise.
11921 (_mm_mask_fmsub_ph): Likewise.
11922 (_mm_mask3_fmsub_ph): Likewise.
11923 (_mm_maskz_fmsub_ph): Likewise.
11924 (_mm256_fnmsub_ph): Likewise.
11925 (_mm256_mask_fnmsub_ph): Likewise.
11926 (_mm256_mask3_fnmsub_ph): Likewise.
11927 (_mm256_maskz_fnmsub_ph): Likewise.
11928 (_mm_fnmsub_ph): Likewise.
11929 (_mm_mask_fnmsub_ph): Likewise.
11930 (_mm_mask3_fnmsub_ph): Likewise.
11931 (_mm_maskz_fnmsub_ph): Likewise.
11932 * config/i386/i386-builtin.def: Add corresponding new builtins.
11933 * config/i386/sse.md
11934 (<avx512>_fmadd_<mode>_maskz<round_expand_name>): Adjust to
11935 support HF vector modes.
11936 (<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name><round_name>):
11938 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1): Ditto.
11939 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_2): Ditto.
11940 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_3): Ditto.
11941 (<avx512>_fmadd_<mode>_mask<round_name>): Ditto.
11942 (<avx512>_fmadd_<mode>_mask3<round_name>): Ditto.
11943 (<avx512>_fmsub_<mode>_maskz<round_expand_name>): Ditto.
11944 (<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name><round_name>):
11946 (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1): Ditto.
11947 (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_2): Ditto.
11948 (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_3): Ditto.
11949 (<avx512>_fmsub_<mode>_mask<round_name>): Ditto.
11950 (<avx512>_fmsub_<mode>_mask3<round_name>): Ditto.
11951 (<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name><round_name>):
11953 (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1): Ditto.
11954 (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_2): Ditto.
11955 (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_3): Ditto.
11956 (<avx512>_fnmadd_<mode>_mask<round_name>): Ditto.
11957 (<avx512>_fnmadd_<mode>_mask3<round_name>): Ditto.
11958 (<avx512>_fnmsub_<mode>_maskz<round_expand_name>): Ditto.
11959 (<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name><round_name>):
11961 (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1): Ditto.
11962 (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_2): Ditto.
11963 (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_3): Ditto.
11964 (<avx512>_fnmsub_<mode>_mask<round_name>): Ditto.
11965 (<avx512>_fnmsub_<mode>_mask3<round_name>): Ditto.
11967 2021-09-18 liuhongt <hongtao.liu@intel.com>
11969 * config/i386/avx512fp16intrin.h (_mm512_fmaddsub_ph):
11971 (_mm512_mask_fmaddsub_ph): Likewise.
11972 (_mm512_mask3_fmaddsub_ph): Likewise.
11973 (_mm512_maskz_fmaddsub_ph): Likewise.
11974 (_mm512_fmaddsub_round_ph): Likewise.
11975 (_mm512_mask_fmaddsub_round_ph): Likewise.
11976 (_mm512_mask3_fmaddsub_round_ph): Likewise.
11977 (_mm512_maskz_fmaddsub_round_ph): Likewise.
11978 (_mm512_mask_fmsubadd_ph): Likewise.
11979 (_mm512_mask3_fmsubadd_ph): Likewise.
11980 (_mm512_maskz_fmsubadd_ph): Likewise.
11981 (_mm512_fmsubadd_round_ph): Likewise.
11982 (_mm512_mask_fmsubadd_round_ph): Likewise.
11983 (_mm512_mask3_fmsubadd_round_ph): Likewise.
11984 (_mm512_maskz_fmsubadd_round_ph): Likewise.
11985 * config/i386/avx512fp16vlintrin.h (_mm256_fmaddsub_ph):
11987 (_mm256_mask_fmaddsub_ph): Likewise.
11988 (_mm256_mask3_fmaddsub_ph): Likewise.
11989 (_mm256_maskz_fmaddsub_ph): Likewise.
11990 (_mm_fmaddsub_ph): Likewise.
11991 (_mm_mask_fmaddsub_ph): Likewise.
11992 (_mm_mask3_fmaddsub_ph): Likewise.
11993 (_mm_maskz_fmaddsub_ph): Likewise.
11994 (_mm256_fmsubadd_ph): Likewise.
11995 (_mm256_mask_fmsubadd_ph): Likewise.
11996 (_mm256_mask3_fmsubadd_ph): Likewise.
11997 (_mm256_maskz_fmsubadd_ph): Likewise.
11998 (_mm_fmsubadd_ph): Likewise.
11999 (_mm_mask_fmsubadd_ph): Likewise.
12000 (_mm_mask3_fmsubadd_ph): Likewise.
12001 (_mm_maskz_fmsubadd_ph): Likewise.
12002 * config/i386/i386-builtin.def: Add corresponding new builtins.
12003 * config/i386/sse.md (VFH_SF_AVX512VL): New mode iterator.
12004 * (<avx512>_fmsubadd_<mode>_maskz<round_expand_name>): New expander.
12005 * (<avx512>_fmaddsub_<mode>_maskz<round_expand_name>): Use
12007 * (<sd_mask_codefor>fma_fmaddsub_<mode><sd_maskz_name><round_name>):
12009 * (<avx512>_fmaddsub_<mode>_mask<round_name>): Ditto.
12010 * (<avx512>_fmaddsub_<mode>_mask3<round_name>): Ditto.
12011 * (<sd_mask_codefor>fma_fmsubadd_<mode><sd_maskz_name><round_name>):
12013 * (<avx512>_fmsubadd_<mode>_mask<round_name>): Ditto.
12014 * (<avx512>_fmsubadd_<mode>_mask3<round_name>): Ditto.
12016 2021-09-18 liuhongt <hongtao.liu@intel.com>
12019 * config/i386/i386.c (ix86_print_operand): Handle
12020 V8HF/V16HF/V32HFmode.
12021 * config/i386/i386.h (VALID_BCST_MODE_P): Add HFmode.
12022 * config/i386/sse.md (avx512bcst): Remove.
12024 2021-09-17 Martin Sebor <msebor@redhat.com>
12026 * Makefile.in (OBJS): Add gimple-predicate-analysis.o.
12027 * tree-ssa-uninit.c (max_phi_args): Move to gimple-predicate-analysis.
12028 (MASK_SET_BIT, MASK_TEST_BIT, MASK_EMPTY): Same.
12029 (check_defs): Add comment.
12030 (can_skip_redundant_opnd): Update comment.
12031 (compute_uninit_opnds_pos): Adjust to namespace change.
12032 (find_pdom): Move to gimple-predicate-analysis.cc.
12034 (struct uninit_undef_val_t): New.
12035 (is_non_loop_exit_postdominating): Move to gimple-predicate-analysis.cc.
12036 (find_control_equiv_block): Same.
12037 (MAX_NUM_CHAINS, MAX_CHAIN_LEN, MAX_POSTDOM_CHECK): Same.
12038 (MAX_SWITCH_CASES): Same.
12039 (compute_control_dep_chain): Same.
12040 (find_uninit_use): Use predicate analyzer.
12041 (struct pred_info): Move to gimple-predicate-analysis.
12042 (convert_control_dep_chain_into_preds): Same.
12043 (find_predicates): Same.
12044 (collect_phi_def_edges): Same.
12045 (warn_uninitialized_phi): Use predicate analyzer.
12046 (find_def_preds): Move to gimple-predicate-analysis.
12047 (dump_pred_info): Same.
12048 (dump_pred_chain): Same.
12049 (dump_predicates): Same.
12050 (destroy_predicate_vecs): Remove.
12051 (execute_late_warn_uninitialized): New.
12052 (get_cmp_code): Move to gimple-predicate-analysis.
12053 (is_value_included_in): Same.
12054 (value_sat_pred_p): Same.
12055 (find_matching_predicate_in_rest_chains): Same.
12056 (is_use_properly_guarded): Same.
12057 (prune_uninit_phi_opnds): Same.
12058 (find_var_cmp_const): Same.
12059 (use_pred_not_overlap_with_undef_path_pred): Same.
12060 (pred_equal_p): Same.
12061 (is_neq_relop_p): Same.
12062 (is_neq_zero_form_p): Same.
12063 (pred_expr_equal_p): Same.
12064 (is_pred_expr_subset_of): Same.
12065 (is_pred_chain_subset_of): Same.
12066 (is_included_in): Same.
12067 (is_superset_of): Same.
12068 (pred_neg_p): Same.
12069 (simplify_pred): Same.
12070 (simplify_preds_2): Same.
12071 (simplify_preds_3): Same.
12072 (simplify_preds_4): Same.
12073 (simplify_preds): Same.
12075 (push_to_worklist): Same.
12076 (get_pred_info_from_cmp): Same.
12077 (is_degenerated_phi): Same.
12078 (normalize_one_pred_1): Same.
12079 (normalize_one_pred): Same.
12080 (normalize_one_pred_chain): Same.
12081 (normalize_preds): Same.
12082 (can_one_predicate_be_invalidated_p): Same.
12083 (can_chain_union_be_invalidated_p): Same.
12084 (uninit_uses_cannot_happen): Same.
12085 (pass_late_warn_uninitialized::execute): Define.
12086 * gimple-predicate-analysis.cc: New file.
12087 * gimple-predicate-analysis.h: New file.
12089 2021-09-17 Julian Brown <julian@codesourcery.com>
12091 * config/gcn/gcn.c (gimple.h): Include.
12092 (gcn_fork_join): Emit barrier for worker-level joins.
12093 * omp-oacc-neuter-broadcast.cc (find_local_vars_to_propagate): Add
12094 writes_gang_private bitmap parameter. Set bit for blocks
12095 containing gang-private variable writes.
12096 (worker_single_simple): Don't emit barrier after predicated block.
12097 (worker_single_copy): Don't emit barrier if we're not broadcasting
12098 anything and the block contains no gang-private writes.
12099 (neuter_worker_single): Don't predicate blocks that only contain
12100 NOPs or internal marker functions. Pass has_gang_private_write
12101 argument to worker_single_copy.
12102 (oacc_do_neutering): Add writes_gang_private bitmap handling.
12104 2021-09-17 Julian Brown <julian@codesourcery.com>
12106 * config/gcn/gcn-protos.h
12107 (gcn_goacc_create_worker_broadcast_record): Update prototype.
12108 * config/gcn/gcn-tree.c (gcn_goacc_get_worker_red_decl): Use
12109 preallocated block of LDS memory. Do not cache/share decls for
12110 reduction temporaries between invocations.
12111 (gcn_goacc_reduction_teardown): Unshare VAR on second use.
12112 (gcn_goacc_create_worker_broadcast_record): Add OFFSET parameter
12113 and return temporary LDS space at that offset. Return pointer in
12115 * config/gcn/gcn.c (acc_lds_size, gang_private_hwm, lds_allocs):
12117 (ACC_LDS_SIZE): Define as acc_lds_size.
12118 (gcn_init_machine_status): Don't initialise lds_allocated,
12119 lds_allocs, reduc_decls fields of machine function struct.
12120 (gcn_option_override): Handle default size for gang-private
12121 variables and -mgang-private-size option.
12122 (gcn_expand_prologue): Use LDS_SIZE instead of LDS_SIZE-1 when
12123 initialising M0_REG.
12124 (gcn_shared_mem_layout): New function.
12125 (gcn_print_lds_decl): Update comment. Use global lds_allocs map and
12126 gang_private_hwm variable.
12127 (TARGET_GOACC_SHARED_MEM_LAYOUT): Define target hook.
12128 * config/gcn/gcn.h (machine_function): Remove lds_allocated,
12129 lds_allocs, reduc_decls. Add reduction_base, reduction_limit.
12130 * config/gcn/gcn.opt (gang_private_size_opt): New global.
12131 (mgang-private-size=): New option.
12132 * doc/tm.texi.in (TARGET_GOACC_SHARED_MEM_LAYOUT): Place
12133 documentation hook.
12134 * doc/tm.texi: Regenerate.
12135 * omp-oacc-neuter-broadcast.cc (targhooks.h, diagnostic-core.h):
12137 (build_sender_ref): Handle sender_decl being pointer.
12138 (worker_single_copy): Add PLACEMENT and ISOLATE_BROADCASTS
12139 parameters. Pass placement argument to
12140 create_worker_broadcast_record hook invocations. Handle
12141 sender_decl being pointer and isolate_broadcasts inserting extra
12143 (blk_offset_map_t): Add typedef.
12144 (neuter_worker_single): Add BLK_OFFSET_MAP parameter. Pass
12145 preallocated range to worker_single_copy call.
12146 (dfs_broadcast_reachable_1): New function.
12147 (idx_decl_pair_t, used_range_vec_t): New typedefs.
12148 (sort_size_descending): New function.
12149 (addr_range): New class.
12150 (splay_tree_compare_addr_range, splay_tree_free_key)
12151 (first_fit_range, merge_ranges_1, merge_ranges): New functions.
12152 (execute_omp_oacc_neuter_broadcast): Rename to...
12153 (oacc_do_neutering): ... this. Add BOUNDS_LO, BOUNDS_HI
12154 parameters. Arrange layout of shared memory for broadcast
12156 (execute_omp_oacc_neuter_broadcast): New function.
12157 (pass_omp_oacc_neuter_broadcast::gate): Remove num_workers==1
12158 handling from here. Enable pass for all OpenACC routines in order
12159 to call shared memory-layout hook.
12160 * target.def (create_worker_broadcast_record): Add OFFSET
12162 (shared_mem_layout): New hook.
12164 2021-09-17 Julian Brown <julian@codesourcery.com>
12165 Thomas Schwinge <thomas@codesourcery.com>
12167 * omp-oacc-neuter-broadcast.cc
12168 (pass_omp_oacc_neuter_broadcast::gate): Disable if num_workers is
12170 (execute_omp_oacc_neuter_broadcast): Adjust.
12172 2021-09-17 Andrew MacLeod <amacleod@redhat.com>
12174 * value-relation.cc (class equiv_chain): Move to header file.
12175 (path_oracle::path_oracle): New.
12176 (path_oracle::~path_oracle): New.
12177 (path_oracle::register_relation): New.
12178 (path_oracle::query_relation): New.
12179 (path_oracle::reset_path): New.
12180 (path_oracle::dump): New.
12181 * value-relation.h (class equiv_chain): Move to here.
12182 (class path_oracle): New.
12184 2021-09-17 Andrew MacLeod <amacleod@redhat.com>
12186 * gimple-range-cache.cc (ranger_cache::ranger_cache): Create a DOM
12188 * gimple-range-fold.cc (fur_depend::register_relation): Use
12189 register_stmt/edge routines.
12190 * value-relation.cc (equiv_chain::find): Relocate from equiv_oracle.
12191 (equiv_oracle::equiv_oracle): Create self equivalence cache.
12192 (equiv_oracle::~equiv_oracle): Release same.
12193 (equiv_oracle::equiv_set): Return entry from self equiv cache if there
12194 are no equivalences.
12195 (equiv_oracle::find_equiv_block): Move list find to equiv_chain.
12196 (equiv_oracle::register_relation): Rename from register_equiv.
12197 (relation_chain_head::find_relation): Relocate from dom_oracle.
12198 (relation_oracle::register_stmt): New.
12199 (relation_oracle::register_edge): New.
12200 (dom_oracle::*): Rename from relation_oracle.
12201 (dom_oracle::register_relation): Adjust to call equiv_oracle.
12202 (dom_oracle::set_one_relation): Split from register_relation.
12203 (dom_oracle::register_transitives): Consolidate 2 methods.
12204 (dom_oracle::find_relation_block): Move core to relation_chain.
12205 (dom_oracle::query_relation): Rename from find_relation_dom and adjust.
12206 * value-relation.h (class relation_oracle): New pure virtual base.
12207 (class equiv_oracle): Inherit from relation_oracle and adjust.
12208 (class dom_oracle): Rename from old relation_oracle and adjust.
12210 2021-09-17 Martin Sebor <msebor@redhat.com>
12212 PR middle-end/102200
12213 * pointer-query.cc (access_ref::inform_access): Handle MIN/MAX_EXPR.
12214 (handle_min_max_size): Change argument. Store original SSA_NAME for
12215 operands to potentially distinct (sub)objects.
12216 (compute_objsize_r): Adjust call to the above.
12218 2021-09-17 Bill Schmidt <wschmidt@linux.ibm.com>
12220 * config/rs6000/rs6000.c (rs6000-builtins.h): New include.
12221 (rs6000_new_builtin_vectorized_function): New function.
12222 (rs6000_new_builtin_md_vectorized_function): Likewise.
12223 (rs6000_builtin_vectorized_function): Call
12224 rs6000_new_builtin_vectorized_function.
12225 (rs6000_builtin_md_vectorized_function): Call
12226 rs6000_new_builtin_md_vectorized_function.
12228 2021-09-17 Bill Schmidt <wschmidt@linux.ibm.com>
12230 * config/rs6000/rs6000-builtin-new.def (ASSEMBLE_ACC): Add mmaint flag.
12231 (ASSEMBLE_PAIR): Likewise.
12232 (BUILD_ACC): Likewise.
12233 (DISASSEMBLE_ACC): Likewise.
12234 (DISASSEMBLE_PAIR): Likewise.
12235 (PMXVBF16GER2): Likewise.
12236 (PMXVBF16GER2NN): Likewise.
12237 (PMXVBF16GER2NP): Likewise.
12238 (PMXVBF16GER2PN): Likewise.
12239 (PMXVBF16GER2PP): Likewise.
12240 (PMXVF16GER2): Likewise.
12241 (PMXVF16GER2NN): Likewise.
12242 (PMXVF16GER2NP): Likewise.
12243 (PMXVF16GER2PN): Likewise.
12244 (PMXVF16GER2PP): Likewise.
12245 (PMXVF32GER): Likewise.
12246 (PMXVF32GERNN): Likewise.
12247 (PMXVF32GERNP): Likewise.
12248 (PMXVF32GERPN): Likewise.
12249 (PMXVF32GERPP): Likewise.
12250 (PMXVF64GER): Likewise.
12251 (PMXVF64GERNN): Likewise.
12252 (PMXVF64GERNP): Likewise.
12253 (PMXVF64GERPN): Likewise.
12254 (PMXVF64GERPP): Likewise.
12255 (PMXVI16GER2): Likewise.
12256 (PMXVI16GER2PP): Likewise.
12257 (PMXVI16GER2S): Likewise.
12258 (PMXVI16GER2SPP): Likewise.
12259 (PMXVI4GER8): Likewise.
12260 (PMXVI4GER8PP): Likewise.
12261 (PMXVI8GER4): Likewise.
12262 (PMXVI8GER4PP): Likewise.
12263 (PMXVI8GER4SPP): Likewise.
12264 (XVBF16GER2): Likewise.
12265 (XVBF16GER2NN): Likewise.
12266 (XVBF16GER2NP): Likewise.
12267 (XVBF16GER2PN): Likewise.
12268 (XVBF16GER2PP): Likewise.
12269 (XVF16GER2): Likewise.
12270 (XVF16GER2NN): Likewise.
12271 (XVF16GER2NP): Likewise.
12272 (XVF16GER2PN): Likewise.
12273 (XVF16GER2PP): Likewise.
12274 (XVF32GER): Likewise.
12275 (XVF32GERNN): Likewise.
12276 (XVF32GERNP): Likewise.
12277 (XVF32GERPN): Likewise.
12278 (XVF32GERPP): Likewise.
12279 (XVF64GER): Likewise.
12280 (XVF64GERNN): Likewise.
12281 (XVF64GERNP): Likewise.
12282 (XVF64GERPN): Likewise.
12283 (XVF64GERPP): Likewise.
12284 (XVI16GER2): Likewise.
12285 (XVI16GER2PP): Likewise.
12286 (XVI16GER2S): Likewise.
12287 (XVI16GER2SPP): Likewise.
12288 (XVI4GER8): Likewise.
12289 (XVI4GER8PP): Likewise.
12290 (XVI8GER4): Likewise.
12291 (XVI8GER4PP): Likewise.
12292 (XVI8GER4SPP): Likewise.
12293 (XXMFACC): Likewise.
12294 (XXMTACC): Likewise.
12295 (XXSETACCZ): Likewise.
12296 (ASSEMBLE_PAIR_V): Likewise.
12297 (BUILD_PAIR): Likewise.
12298 (DISASSEMBLE_PAIR_V): Likewise.
12301 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_new_mma_builtin):
12302 Handle RS6000_BIF_LXVP and RS6000_BIF_STXVP.
12303 * config/rs6000/rs6000-gen-builtins.c (attrinfo): Add ismmaint.
12304 (parse_bif_attrs): Handle ismmaint.
12305 (write_decls): Add bif_mmaint_bit and bif_is_mmaint.
12306 (write_bif_static_init): Handle ismmaint.
12308 2021-09-17 Bill Schmidt <wschmidt@linux.ibm.com>
12310 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_new_builtin): New
12312 (rs6000_gimple_fold_builtin): Call rs6000_gimple_fold_new_builtin.
12313 (rs6000_new_builtin_valid_without_lhs): New function.
12314 (rs6000_gimple_fold_new_mma_builtin): Likewise.
12315 (rs6000_gimple_fold_new_builtin): Likewise.
12317 2021-09-17 Thomas Schwinge <thomas@codesourcery.com>
12319 * hash-table.h (hash_table<Descriptor, Lazy, Allocator>::expand):
12320 Destruct stale Value objects.
12321 * hash-map-tests.c (test_map_of_type_with_ctor_and_dtor_expand):
12324 2021-09-17 Roger Sayle <roger@nextmovesoftware.com>
12327 * match.pd (shift optimizations): Disable recent sign-changing
12328 optimization for shifts by zero, these will be folded later.
12330 2021-09-17 Bill Schmidt <wschmidt@linux.ibm.com>
12332 * config/rs6000/rs6000-builtin-new.def (__builtin_mffsl): Move from
12333 [power9] to [always].
12335 2021-09-17 Richard Biener <rguenther@suse.de>
12337 * tree-vect-stmts.c (vectorizable_load): Do not frob
12340 2021-09-17 H.J. Lu <hjl.tools@gmail.com>
12342 * config/i386/i386-features.c (remove_partial_avx_dependency):
12343 Also check TARGET_SSE_PARTIAL_REG_FP_CONVERTS_DEPENDENCY and
12344 and TARGET_SSE_PARTIAL_REG_CONVERTS_DEPENDENCY before generating
12346 * config/i386/i386.h (TARGET_SSE_PARTIAL_REG_FP_CONVERTS_DEPENDENCY):
12348 (TARGET_SSE_PARTIAL_REG_CONVERTS_DEPENDENCY): Likewise.
12349 * config/i386/i386.md (SSE FP to FP splitters): Replace
12350 TARGET_SSE_PARTIAL_REG_DEPENDENCY with
12351 TARGET_SSE_PARTIAL_REG_FP_CONVERTS_DEPENDENCY.
12352 (SSE INT to FP splitter): Replace TARGET_SSE_PARTIAL_REG_DEPENDENCY
12353 with TARGET_SSE_PARTIAL_REG_CONVERTS_DEPENDENCY.
12354 * config/i386/x86-tune.def
12355 (X86_TUNE_SSE_PARTIAL_REG_FP_CONVERTS_DEPENDENCY): New.
12356 (X86_TUNE_SSE_PARTIAL_REG_CONVERTS_DEPENDENCY): Likewise.
12358 2021-09-17 H.J. Lu <hjl.tools@gmail.com>
12361 * config/i386/i386-features.c (remove_partial_avx_dependency):
12362 Check TARGET_USE_VECTOR_FP_CONVERTS and TARGET_USE_VECTOR_CONVERTS
12363 before generating vxorps.
12365 2021-09-17 H.J. Lu <hjl.tools@gmail.com>
12367 * config/i386/i386-options.c (processor_cost_table): Use
12368 tremont_cost for Tremont.
12369 * config/i386/x86-tune-costs.h (tremont_memcpy): New.
12370 (tremont_memset): Likewise.
12371 (tremont_cost): Likewise.
12372 * config/i386/x86-tune.def (X86_TUNE_PREFER_KNOWN_REP_MOVSB_STOSB):
12373 Enable for Tremont.
12375 2021-09-17 H.J. Lu <hjl.tools@gmail.com>
12377 * common/config/i386/i386-common.c: Use Haswell scheduling model
12379 * config/i386/i386.c (ix86_sched_init_global): Prepare for Tremont
12381 * config/i386/x86-tune-sched.c (ix86_issue_rate): Change Tremont
12383 (ix86_adjust_cost): Handle Tremont.
12384 * config/i386/x86-tune.def (X86_TUNE_SSE_PARTIAL_REG_DEPENDENCY):
12385 Enable for Tremont.
12386 (X86_TUNE_USE_LEAVE): Likewise.
12387 (X86_TUNE_PUSH_MEMORY): Likewise.
12388 (X86_TUNE_MISALIGNED_MOVE_STRING_PRO_EPILOGUES): Likewise.
12389 (X86_TUNE_USE_CLTD): Likewise.
12390 (X86_TUNE_AVOID_FALSE_DEP_FOR_BMI): Likewise.
12391 (X86_TUNE_AVOID_MFENCE): Likewise.
12392 (X86_TUNE_SSE_TYPELESS_STORES): Likewise.
12393 (X86_TUNE_SSE_LOAD0_BY_PXOR): Likewise.
12394 (X86_TUNE_ACCUMULATE_OUTGOING_ARGS): Disable for Tremont.
12395 (X86_TUNE_FOUR_JUMP_LIMIT): Likewise.
12396 (X86_TUNE_OPT_AGU): Likewise.
12397 (X86_TUNE_AVOID_LEA_FOR_ADDR): Likewise.
12398 (X86_TUNE_AVOID_MEM_OPND_FOR_CMOVE): Likewise.
12399 (X86_TUNE_EXPAND_ABS): Likewise.
12400 (X86_TUNE_SPLIT_MEM_OPND_FOR_FP_CONVERTS): Likewise.
12401 (X86_TUNE_SLOW_PSHUFB): Likewise.
12403 2021-09-17 Eric Botcazou <ebotcazou@adacore.com>
12405 PR rtl-optimization/102306
12406 * combine.c (try_combine): Abort the combination if we are about to
12407 duplicate volatile references.
12409 2021-09-17 liuhongt <hongtao.liu@intel.com>
12411 * config/i386/avx512fp16intrin.h (_mm_undefined_ph):
12413 (_mm256_undefined_ph): Likewise.
12414 (_mm512_undefined_ph): Likewise.
12415 (_mm_cvtsh_h): Likewise.
12416 (_mm256_cvtsh_h): Likewise.
12417 (_mm512_cvtsh_h): Likewise.
12418 (_mm512_castph_ps): Likewise.
12419 (_mm512_castph_pd): Likewise.
12420 (_mm512_castph_si512): Likewise.
12421 (_mm512_castph512_ph128): Likewise.
12422 (_mm512_castph512_ph256): Likewise.
12423 (_mm512_castph128_ph512): Likewise.
12424 (_mm512_castph256_ph512): Likewise.
12425 (_mm512_zextph128_ph512): Likewise.
12426 (_mm512_zextph256_ph512): Likewise.
12427 (_mm512_castps_ph): Likewise.
12428 (_mm512_castpd_ph): Likewise.
12429 (_mm512_castsi512_ph): Likewise.
12430 * config/i386/avx512fp16vlintrin.h (_mm_castph_ps):
12432 (_mm256_castph_ps): Likewise.
12433 (_mm_castph_pd): Likewise.
12434 (_mm256_castph_pd): Likewise.
12435 (_mm_castph_si128): Likewise.
12436 (_mm256_castph_si256): Likewise.
12437 (_mm_castps_ph): Likewise.
12438 (_mm256_castps_ph): Likewise.
12439 (_mm_castpd_ph): Likewise.
12440 (_mm256_castpd_ph): Likewise.
12441 (_mm_castsi128_ph): Likewise.
12442 (_mm256_castsi256_ph): Likewise.
12443 (_mm256_castph256_ph128): Likewise.
12444 (_mm256_castph128_ph256): Likewise.
12445 (_mm256_zextph128_ph256): Likewise.
12447 2021-09-17 liuhongt <hongtao.liu@intel.com>
12449 * config/i386/avx512fp16intrin.h (_mm_cvtsh_ss):
12451 (_mm_mask_cvtsh_ss): Likewise.
12452 (_mm_maskz_cvtsh_ss): Likewise.
12453 (_mm_cvtsh_sd): Likewise.
12454 (_mm_mask_cvtsh_sd): Likewise.
12455 (_mm_maskz_cvtsh_sd): Likewise.
12456 (_mm_cvt_roundsh_ss): Likewise.
12457 (_mm_mask_cvt_roundsh_ss): Likewise.
12458 (_mm_maskz_cvt_roundsh_ss): Likewise.
12459 (_mm_cvt_roundsh_sd): Likewise.
12460 (_mm_mask_cvt_roundsh_sd): Likewise.
12461 (_mm_maskz_cvt_roundsh_sd): Likewise.
12462 (_mm_cvtss_sh): Likewise.
12463 (_mm_mask_cvtss_sh): Likewise.
12464 (_mm_maskz_cvtss_sh): Likewise.
12465 (_mm_cvtsd_sh): Likewise.
12466 (_mm_mask_cvtsd_sh): Likewise.
12467 (_mm_maskz_cvtsd_sh): Likewise.
12468 (_mm_cvt_roundss_sh): Likewise.
12469 (_mm_mask_cvt_roundss_sh): Likewise.
12470 (_mm_maskz_cvt_roundss_sh): Likewise.
12471 (_mm_cvt_roundsd_sh): Likewise.
12472 (_mm_mask_cvt_roundsd_sh): Likewise.
12473 (_mm_maskz_cvt_roundsd_sh): Likewise.
12474 * config/i386/i386-builtin-types.def
12475 (V8HF_FTYPE_V2DF_V8HF_V8HF_UQI_INT,
12476 V8HF_FTYPE_V4SF_V8HF_V8HF_UQI_INT,
12477 V2DF_FTYPE_V8HF_V2DF_V2DF_UQI_INT,
12478 V4SF_FTYPE_V8HF_V4SF_V4SF_UQI_INT): Add new builtin types.
12479 * config/i386/i386-builtin.def: Add corrresponding new builtins.
12480 * config/i386/i386-expand.c: Handle new builtin types.
12481 * config/i386/sse.md (VF48_128): New mode iterator.
12482 (avx512fp16_vcvtsh2<ssescalarmodesuffix><mask_scalar_name><round_saeonly_scalar_name>):
12484 (avx512fp16_vcvt<ssescalarmodesuffix>2sh<mask_scalar_name><round_scalar_name>):
12487 2021-09-17 liuhongt <hongtao.liu@intel.com>
12489 * config/i386/avx512fp16intrin.h (_mm512_cvtph_pd):
12491 (_mm512_mask_cvtph_pd): Likewise.
12492 (_mm512_maskz_cvtph_pd): Likewise.
12493 (_mm512_cvt_roundph_pd): Likewise.
12494 (_mm512_mask_cvt_roundph_pd): Likewise.
12495 (_mm512_maskz_cvt_roundph_pd): Likewise.
12496 (_mm512_cvtxph_ps): Likewise.
12497 (_mm512_mask_cvtxph_ps): Likewise.
12498 (_mm512_maskz_cvtxph_ps): Likewise.
12499 (_mm512_cvtx_roundph_ps): Likewise.
12500 (_mm512_mask_cvtx_roundph_ps): Likewise.
12501 (_mm512_maskz_cvtx_roundph_ps): Likewise.
12502 (_mm512_cvtxps_ph): Likewise.
12503 (_mm512_mask_cvtxps_ph): Likewise.
12504 (_mm512_maskz_cvtxps_ph): Likewise.
12505 (_mm512_cvtx_roundps_ph): Likewise.
12506 (_mm512_mask_cvtx_roundps_ph): Likewise.
12507 (_mm512_maskz_cvtx_roundps_ph): Likewise.
12508 (_mm512_cvtpd_ph): Likewise.
12509 (_mm512_mask_cvtpd_ph): Likewise.
12510 (_mm512_maskz_cvtpd_ph): Likewise.
12511 (_mm512_cvt_roundpd_ph): Likewise.
12512 (_mm512_mask_cvt_roundpd_ph): Likewise.
12513 (_mm512_maskz_cvt_roundpd_ph): Likewise.
12514 * config/i386/avx512fp16vlintrin.h (_mm_cvtph_pd):
12516 (_mm_mask_cvtph_pd): Likewise.
12517 (_mm_maskz_cvtph_pd): Likewise.
12518 (_mm256_cvtph_pd): Likewise.
12519 (_mm256_mask_cvtph_pd): Likewise.
12520 (_mm256_maskz_cvtph_pd): Likewise.
12521 (_mm_cvtxph_ps): Likewise.
12522 (_mm_mask_cvtxph_ps): Likewise.
12523 (_mm_maskz_cvtxph_ps): Likewise.
12524 (_mm256_cvtxph_ps): Likewise.
12525 (_mm256_mask_cvtxph_ps): Likewise.
12526 (_mm256_maskz_cvtxph_ps): Likewise.
12527 (_mm_cvtxps_ph): Likewise.
12528 (_mm_mask_cvtxps_ph): Likewise.
12529 (_mm_maskz_cvtxps_ph): Likewise.
12530 (_mm256_cvtxps_ph): Likewise.
12531 (_mm256_mask_cvtxps_ph): Likewise.
12532 (_mm256_maskz_cvtxps_ph): Likewise.
12533 (_mm_cvtpd_ph): Likewise.
12534 (_mm_mask_cvtpd_ph): Likewise.
12535 (_mm_maskz_cvtpd_ph): Likewise.
12536 (_mm256_cvtpd_ph): Likewise.
12537 (_mm256_mask_cvtpd_ph): Likewise.
12538 (_mm256_maskz_cvtpd_ph): Likewise.
12539 * config/i386/i386-builtin.def: Add corresponding new builtins.
12540 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
12541 * config/i386/i386-expand.c: Handle new builtin types.
12542 * config/i386/sse.md
12543 (VF4_128_8_256): New.
12544 (VF48H_AVX512VL): Ditto.
12545 (ssePHmode): Add HF vector modes.
12546 (castmode): Add new convertable modes.
12547 (qq2phsuff): Ditto.
12548 (ph2pssuffix): New.
12549 (avx512fp16_vcvt<castmode>2ph_<mode><mask_name><round_name>): Ditto.
12550 (avx512fp16_vcvt<castmode>2ph_<mode>): Ditto.
12551 (*avx512fp16_vcvt<castmode>2ph_<mode>): Ditto.
12552 (avx512fp16_vcvt<castmode>2ph_<mode>_mask): Ditto.
12553 (*avx512fp16_vcvt<castmode>2ph_<mode>_mask): Ditto.
12554 (*avx512fp16_vcvt<castmode>2ph_<mode>_mask_1): Ditto.
12555 (avx512fp16_float_extend_ph<mode>2<mask_name><round_saeonly_name>):
12557 (avx512fp16_float_extend_ph<mode>2<mask_name>): Ditto.
12558 (*avx512fp16_float_extend_ph<mode>2_load<mask_name>): Ditto.
12559 (avx512fp16_float_extend_phv2df2<mask_name>): Ditto.
12560 (*avx512fp16_float_extend_phv2df2_load<mask_name>): Ditto.
12562 2021-09-17 liuhongt <hongtao.liu@intel.com>
12564 * config/i386/avx512fp16intrin.h (_mm_cvttsh_i32):
12566 (_mm_cvttsh_u32): Likewise.
12567 (_mm_cvtt_roundsh_i32): Likewise.
12568 (_mm_cvtt_roundsh_u32): Likewise.
12569 (_mm_cvttsh_i64): Likewise.
12570 (_mm_cvttsh_u64): Likewise.
12571 (_mm_cvtt_roundsh_i64): Likewise.
12572 (_mm_cvtt_roundsh_u64): Likewise.
12573 * config/i386/i386-builtin.def: Add corresponding new builtins.
12574 * config/i386/sse.md
12575 (avx512fp16_fix<fixunssuffix>_trunc<mode>2<round_saeonly_name>):
12578 2021-09-17 liuhongt <hongtao.liu@intel.com>
12580 * config/i386/avx512fp16intrin.h (_mm512_cvttph_epi32):
12582 (_mm512_mask_cvttph_epi32): Likewise.
12583 (_mm512_maskz_cvttph_epi32): Likewise.
12584 (_mm512_cvtt_roundph_epi32): Likewise.
12585 (_mm512_mask_cvtt_roundph_epi32): Likewise.
12586 (_mm512_maskz_cvtt_roundph_epi32): Likewise.
12587 (_mm512_cvttph_epu32): Likewise.
12588 (_mm512_mask_cvttph_epu32): Likewise.
12589 (_mm512_maskz_cvttph_epu32): Likewise.
12590 (_mm512_cvtt_roundph_epu32): Likewise.
12591 (_mm512_mask_cvtt_roundph_epu32): Likewise.
12592 (_mm512_maskz_cvtt_roundph_epu32): Likewise.
12593 (_mm512_cvttph_epi64): Likewise.
12594 (_mm512_mask_cvttph_epi64): Likewise.
12595 (_mm512_maskz_cvttph_epi64): Likewise.
12596 (_mm512_cvtt_roundph_epi64): Likewise.
12597 (_mm512_mask_cvtt_roundph_epi64): Likewise.
12598 (_mm512_maskz_cvtt_roundph_epi64): Likewise.
12599 (_mm512_cvttph_epu64): Likewise.
12600 (_mm512_mask_cvttph_epu64): Likewise.
12601 (_mm512_maskz_cvttph_epu64): Likewise.
12602 (_mm512_cvtt_roundph_epu64): Likewise.
12603 (_mm512_mask_cvtt_roundph_epu64): Likewise.
12604 (_mm512_maskz_cvtt_roundph_epu64): Likewise.
12605 (_mm512_cvttph_epi16): Likewise.
12606 (_mm512_mask_cvttph_epi16): Likewise.
12607 (_mm512_maskz_cvttph_epi16): Likewise.
12608 (_mm512_cvtt_roundph_epi16): Likewise.
12609 (_mm512_mask_cvtt_roundph_epi16): Likewise.
12610 (_mm512_maskz_cvtt_roundph_epi16): Likewise.
12611 (_mm512_cvttph_epu16): Likewise.
12612 (_mm512_mask_cvttph_epu16): Likewise.
12613 (_mm512_maskz_cvttph_epu16): Likewise.
12614 (_mm512_cvtt_roundph_epu16): Likewise.
12615 (_mm512_mask_cvtt_roundph_epu16): Likewise.
12616 (_mm512_maskz_cvtt_roundph_epu16): Likewise.
12617 * config/i386/avx512fp16vlintrin.h (_mm_cvttph_epi32):
12619 (_mm_mask_cvttph_epi32): Likewise.
12620 (_mm_maskz_cvttph_epi32): Likewise.
12621 (_mm256_cvttph_epi32): Likewise.
12622 (_mm256_mask_cvttph_epi32): Likewise.
12623 (_mm256_maskz_cvttph_epi32): Likewise.
12624 (_mm_cvttph_epu32): Likewise.
12625 (_mm_mask_cvttph_epu32): Likewise.
12626 (_mm_maskz_cvttph_epu32): Likewise.
12627 (_mm256_cvttph_epu32): Likewise.
12628 (_mm256_mask_cvttph_epu32): Likewise.
12629 (_mm256_maskz_cvttph_epu32): Likewise.
12630 (_mm_cvttph_epi64): Likewise.
12631 (_mm_mask_cvttph_epi64): Likewise.
12632 (_mm_maskz_cvttph_epi64): Likewise.
12633 (_mm256_cvttph_epi64): Likewise.
12634 (_mm256_mask_cvttph_epi64): Likewise.
12635 (_mm256_maskz_cvttph_epi64): Likewise.
12636 (_mm_cvttph_epu64): Likewise.
12637 (_mm_mask_cvttph_epu64): Likewise.
12638 (_mm_maskz_cvttph_epu64): Likewise.
12639 (_mm256_cvttph_epu64): Likewise.
12640 (_mm256_mask_cvttph_epu64): Likewise.
12641 (_mm256_maskz_cvttph_epu64): Likewise.
12642 (_mm_cvttph_epi16): Likewise.
12643 (_mm_mask_cvttph_epi16): Likewise.
12644 (_mm_maskz_cvttph_epi16): Likewise.
12645 (_mm256_cvttph_epi16): Likewise.
12646 (_mm256_mask_cvttph_epi16): Likewise.
12647 (_mm256_maskz_cvttph_epi16): Likewise.
12648 (_mm_cvttph_epu16): Likewise.
12649 (_mm_mask_cvttph_epu16): Likewise.
12650 (_mm_maskz_cvttph_epu16): Likewise.
12651 (_mm256_cvttph_epu16): Likewise.
12652 (_mm256_mask_cvttph_epu16): Likewise.
12653 (_mm256_maskz_cvttph_epu16): Likewise.
12654 * config/i386/i386-builtin.def: Add new builtins.
12655 * config/i386/sse.md
12656 (avx512fp16_fix<fixunssuffix>_trunc<mode>2<mask_name><round_saeonly_name>):
12658 (avx512fp16_fix<fixunssuffix>_trunc<mode>2<mask_name>): Ditto.
12659 (*avx512fp16_fix<fixunssuffix>_trunc<mode>2_load<mask_name>): Ditto.
12660 (avx512fp16_fix<fixunssuffix>_truncv2di2<mask_name>): Ditto.
12661 (avx512fp16_fix<fixunssuffix>_truncv2di2_load<mask_name>): Ditto.
12663 2021-09-17 liuhongt <hongtao.liu@intel.com>
12665 * config/i386/avx512fp16intrin.h (_mm_cvtsh_i32): New intrinsic.
12666 (_mm_cvtsh_u32): Likewise.
12667 (_mm_cvt_roundsh_i32): Likewise.
12668 (_mm_cvt_roundsh_u32): Likewise.
12669 (_mm_cvtsh_i64): Likewise.
12670 (_mm_cvtsh_u64): Likewise.
12671 (_mm_cvt_roundsh_i64): Likewise.
12672 (_mm_cvt_roundsh_u64): Likewise.
12673 (_mm_cvti32_sh): Likewise.
12674 (_mm_cvtu32_sh): Likewise.
12675 (_mm_cvt_roundi32_sh): Likewise.
12676 (_mm_cvt_roundu32_sh): Likewise.
12677 (_mm_cvti64_sh): Likewise.
12678 (_mm_cvtu64_sh): Likewise.
12679 (_mm_cvt_roundi64_sh): Likewise.
12680 (_mm_cvt_roundu64_sh): Likewise.
12681 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
12682 * config/i386/i386-builtin.def: Add corresponding new builtins.
12683 * config/i386/i386-expand.c (ix86_expand_round_builtin):
12684 Handle new builtin types.
12685 * config/i386/sse.md
12686 (avx512fp16_vcvtsh2<sseintconvertsignprefix>si<rex64namesuffix><round_name>):
12688 (avx512fp16_vcvtsh2<sseintconvertsignprefix>si<rex64namesuffix>_2): Likewise.
12689 (avx512fp16_vcvt<floatsuffix>si2sh<rex64namesuffix><round_name>): Likewise.
12691 2021-09-16 Bill Schmidt <wschmidt@linux.ibm.com>
12693 * config/rs6000/rs6000-c.c (rs6000-builtins.h): New include.
12694 (altivec_resolve_new_overloaded_builtin): New forward decl.
12695 (rs6000_new_builtin_type_compatible): New function.
12696 (altivec_resolve_overloaded_builtin): Call
12697 altivec_resolve_new_overloaded_builtin.
12698 (altivec_build_new_resolved_builtin): New function.
12699 (altivec_resolve_new_overloaded_builtin): Likewise.
12700 * config/rs6000/rs6000-call.c (rs6000_new_builtin_is_supported):
12702 * config/rs6000/rs6000-gen-builtins.c (write_decls): Remove _p from
12703 name of rs6000_new_builtin_is_supported.
12705 2021-09-16 Uroš Bizjak <ubizjak@gmail.com>
12707 * config/i386/i386-protos.h (ix86_decompose_address):
12708 Change return type to bool.
12709 * config/i386/i386.c (ix86_decompose_address): Ditto.
12711 2021-09-16 Tobias Burnus <tobias@codesourcery.com>
12714 * config/rs6000/t-rs6000 (build/rs6000-gen-builtins.o, build/rbtree.o):
12715 Added 'build/' to target, use build/%.o rule.
12716 (build/rs6000-gen-builtins$(build_exeext)): Add 'build/' and
12717 '$(build_exeext)' to target and 'build/' for the *.o files.
12718 (rs6000-builtins.c): Update for those changes; run rs6000-gen-builtins
12721 2021-09-16 Martin Jambor <mjambor@suse.cz>
12723 * cgraph.c (cgraph_node::dump): Do not check caller count sums if
12724 the body has been removed. Remove trailing whitespace.
12726 2021-09-16 Richard Biener <rguenther@suse.de>
12728 PR middle-end/102360
12729 * internal-fn.c (expand_DEFERRED_INIT): Make pattern-init
12730 of non-memory more robust.
12732 2021-09-16 Daniel Cederman <cederman@gaisler.com>
12734 * config/sparc/sparc-opts.h (enum sparc_processor_type): Add LEON5
12735 * config/sparc/sparc.c (struct processor_costs): Add LEON5 costs
12736 (leon5_adjust_cost): Increase cost of store with data dependency
12737 on ALU instruction and FPU anti-dependencies.
12738 (sparc_option_override): Add LEON5 costs
12739 (sparc_adjust_cost): Add LEON5 cost adjustments
12740 * config/sparc/sparc.h: Add LEON5
12741 * config/sparc/sparc.md: Include LEON5 scheduling information
12742 * config/sparc/sparc.opt: Add LEON5
12743 * doc/invoke.texi: Add LEON5
12744 * config/sparc/leon5.md: New file.
12746 2021-09-16 Daniel Cederman <cederman@gaisler.com>
12748 * config/sparc/sparc.md (stack_protect_set32): Add NOP to prevent
12749 sensitive sequence for B2BST errata workaround.
12751 2021-09-16 Daniel Cederman <cederman@gaisler.com>
12753 * config/sparc/sparc.c (sparc_do_work_around_errata): Do not begin
12754 functions with atomic instruction in the UT700 errata workaround.
12756 2021-09-16 Daniel Cederman <cederman@gaisler.com>
12758 * config/sparc/sparc.c (next_active_non_empty_insn): New function
12759 that returns next active non empty assembly instruction.
12760 (sparc_do_work_around_errata): Use new function.
12762 2021-09-16 Daniel Cederman <cederman@gaisler.com>
12764 * config/sparc/sparc.c (store_insn_p): Add predicate for store
12766 (load_insn_p): Add predicate for load attributes.
12767 (sparc_do_work_around_errata): Use new predicates.
12769 2021-09-16 Andreas Larsson <andreas@gaisler.com>
12771 * config/sparc/sparc.c (dump_target_flag_bits): Print bit names for
12774 2021-09-16 Martin Liska <mliska@suse.cz>
12776 * config/mips/netbsd.h: Fix typo in name of a macro.
12778 2021-09-16 liuhongt <hongtao.liu@intel.com>
12780 PR middle-end/102080
12781 * match.pd: Check mask type when doing cond_op related gimple
12783 * tree.c (is_truth_type_for): New function.
12784 * tree.h (is_truth_type_for): New declaration.
12786 2021-09-16 liuhongt <hongtao.liu@intel.com>
12788 * config/i386/avx512fp16intrin.h (_mm512_cvtepi32_ph): New
12790 (_mm512_mask_cvtepi32_ph): Likewise.
12791 (_mm512_maskz_cvtepi32_ph): Likewise.
12792 (_mm512_cvt_roundepi32_ph): Likewise.
12793 (_mm512_mask_cvt_roundepi32_ph): Likewise.
12794 (_mm512_maskz_cvt_roundepi32_ph): Likewise.
12795 (_mm512_cvtepu32_ph): Likewise.
12796 (_mm512_mask_cvtepu32_ph): Likewise.
12797 (_mm512_maskz_cvtepu32_ph): Likewise.
12798 (_mm512_cvt_roundepu32_ph): Likewise.
12799 (_mm512_mask_cvt_roundepu32_ph): Likewise.
12800 (_mm512_maskz_cvt_roundepu32_ph): Likewise.
12801 (_mm512_cvtepi64_ph): Likewise.
12802 (_mm512_mask_cvtepi64_ph): Likewise.
12803 (_mm512_maskz_cvtepi64_ph): Likewise.
12804 (_mm512_cvt_roundepi64_ph): Likewise.
12805 (_mm512_mask_cvt_roundepi64_ph): Likewise.
12806 (_mm512_maskz_cvt_roundepi64_ph): Likewise.
12807 (_mm512_cvtepu64_ph): Likewise.
12808 (_mm512_mask_cvtepu64_ph): Likewise.
12809 (_mm512_maskz_cvtepu64_ph): Likewise.
12810 (_mm512_cvt_roundepu64_ph): Likewise.
12811 (_mm512_mask_cvt_roundepu64_ph): Likewise.
12812 (_mm512_maskz_cvt_roundepu64_ph): Likewise.
12813 (_mm512_cvtepi16_ph): Likewise.
12814 (_mm512_mask_cvtepi16_ph): Likewise.
12815 (_mm512_maskz_cvtepi16_ph): Likewise.
12816 (_mm512_cvt_roundepi16_ph): Likewise.
12817 (_mm512_mask_cvt_roundepi16_ph): Likewise.
12818 (_mm512_maskz_cvt_roundepi16_ph): Likewise.
12819 (_mm512_cvtepu16_ph): Likewise.
12820 (_mm512_mask_cvtepu16_ph): Likewise.
12821 (_mm512_maskz_cvtepu16_ph): Likewise.
12822 (_mm512_cvt_roundepu16_ph): Likewise.
12823 (_mm512_mask_cvt_roundepu16_ph): Likewise.
12824 (_mm512_maskz_cvt_roundepu16_ph): Likewise.
12825 * config/i386/avx512fp16vlintrin.h (_mm_cvtepi32_ph): New
12827 (_mm_mask_cvtepi32_ph): Likewise.
12828 (_mm_maskz_cvtepi32_ph): Likewise.
12829 (_mm256_cvtepi32_ph): Likewise.
12830 (_mm256_mask_cvtepi32_ph): Likewise.
12831 (_mm256_maskz_cvtepi32_ph): Likewise.
12832 (_mm_cvtepu32_ph): Likewise.
12833 (_mm_mask_cvtepu32_ph): Likewise.
12834 (_mm_maskz_cvtepu32_ph): Likewise.
12835 (_mm256_cvtepu32_ph): Likewise.
12836 (_mm256_mask_cvtepu32_ph): Likewise.
12837 (_mm256_maskz_cvtepu32_ph): Likewise.
12838 (_mm_cvtepi64_ph): Likewise.
12839 (_mm_mask_cvtepi64_ph): Likewise.
12840 (_mm_maskz_cvtepi64_ph): Likewise.
12841 (_mm256_cvtepi64_ph): Likewise.
12842 (_mm256_mask_cvtepi64_ph): Likewise.
12843 (_mm256_maskz_cvtepi64_ph): Likewise.
12844 (_mm_cvtepu64_ph): Likewise.
12845 (_mm_mask_cvtepu64_ph): Likewise.
12846 (_mm_maskz_cvtepu64_ph): Likewise.
12847 (_mm256_cvtepu64_ph): Likewise.
12848 (_mm256_mask_cvtepu64_ph): Likewise.
12849 (_mm256_maskz_cvtepu64_ph): Likewise.
12850 (_mm_cvtepi16_ph): Likewise.
12851 (_mm_mask_cvtepi16_ph): Likewise.
12852 (_mm_maskz_cvtepi16_ph): Likewise.
12853 (_mm256_cvtepi16_ph): Likewise.
12854 (_mm256_mask_cvtepi16_ph): Likewise.
12855 (_mm256_maskz_cvtepi16_ph): Likewise.
12856 (_mm_cvtepu16_ph): Likewise.
12857 (_mm_mask_cvtepu16_ph): Likewise.
12858 (_mm_maskz_cvtepu16_ph): Likewise.
12859 (_mm256_cvtepu16_ph): Likewise.
12860 (_mm256_mask_cvtepu16_ph): Likewise.
12861 (_mm256_maskz_cvtepu16_ph): Likewise.
12862 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
12863 * config/i386/i386-builtin.def: Add corresponding new builtins.
12864 * config/i386/i386-expand.c
12865 (ix86_expand_args_builtin): Handle new builtin types.
12866 (ix86_expand_round_builtin): Ditto.
12867 * config/i386/i386-modes.def: Declare V2HF and V6HF.
12868 * config/i386/sse.md (VI2H_AVX512VL): New.
12869 (qq2phsuff): Ditto.
12870 (sseintvecmode): Add HF vector modes.
12871 (avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode><mask_name><round_name>):
12873 (avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>): Ditto.
12874 (*avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>): Ditto.
12875 (avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>_mask): Ditto.
12876 (*avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>_mask): Ditto.
12877 (*avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>_mask_1): Ditto.
12878 (avx512fp16_vcvt<floatsuffix>qq2ph_v2di): Ditto.
12879 (*avx512fp16_vcvt<floatsuffix>qq2ph_v2di): Ditto.
12880 (avx512fp16_vcvt<floatsuffix>qq2ph_v2di_mask): Ditto.
12881 (*avx512fp16_vcvt<floatsuffix>qq2ph_v2di_mask): Ditto.
12882 (*avx512fp16_vcvt<floatsuffix>qq2ph_v2di_mask_1): Ditto.
12883 * config/i386/subst.md (round_qq2phsuff): New subst_attr.
12885 2021-09-16 liuhongt <hongtao.liu@intel.com>
12887 * config/i386/avx512fp16intrin.h (_mm512_cvtph_epi32):
12889 (_mm512_mask_cvtph_epi32): Likewise.
12890 (_mm512_maskz_cvtph_epi32): Likewise.
12891 (_mm512_cvt_roundph_epi32): Likewise.
12892 (_mm512_mask_cvt_roundph_epi32): Likewise.
12893 (_mm512_maskz_cvt_roundph_epi32): Likewise.
12894 (_mm512_cvtph_epu32): Likewise.
12895 (_mm512_mask_cvtph_epu32): Likewise.
12896 (_mm512_maskz_cvtph_epu32): Likewise.
12897 (_mm512_cvt_roundph_epu32): Likewise.
12898 (_mm512_mask_cvt_roundph_epu32): Likewise.
12899 (_mm512_maskz_cvt_roundph_epu32): Likewise.
12900 (_mm512_cvtph_epi64): Likewise.
12901 (_mm512_mask_cvtph_epi64): Likewise.
12902 (_mm512_maskz_cvtph_epi64): Likewise.
12903 (_mm512_cvt_roundph_epi64): Likewise.
12904 (_mm512_mask_cvt_roundph_epi64): Likewise.
12905 (_mm512_maskz_cvt_roundph_epi64): Likewise.
12906 (_mm512_cvtph_epu64): Likewise.
12907 (_mm512_mask_cvtph_epu64): Likewise.
12908 (_mm512_maskz_cvtph_epu64): Likewise.
12909 (_mm512_cvt_roundph_epu64): Likewise.
12910 (_mm512_mask_cvt_roundph_epu64): Likewise.
12911 (_mm512_maskz_cvt_roundph_epu64): Likewise.
12912 (_mm512_cvtph_epi16): Likewise.
12913 (_mm512_mask_cvtph_epi16): Likewise.
12914 (_mm512_maskz_cvtph_epi16): Likewise.
12915 (_mm512_cvt_roundph_epi16): Likewise.
12916 (_mm512_mask_cvt_roundph_epi16): Likewise.
12917 (_mm512_maskz_cvt_roundph_epi16): Likewise.
12918 (_mm512_cvtph_epu16): Likewise.
12919 (_mm512_mask_cvtph_epu16): Likewise.
12920 (_mm512_maskz_cvtph_epu16): Likewise.
12921 (_mm512_cvt_roundph_epu16): Likewise.
12922 (_mm512_mask_cvt_roundph_epu16): Likewise.
12923 (_mm512_maskz_cvt_roundph_epu16): Likewise.
12924 * config/i386/avx512fp16vlintrin.h (_mm_cvtph_epi32):
12926 (_mm_mask_cvtph_epi32): Likewise.
12927 (_mm_maskz_cvtph_epi32): Likewise.
12928 (_mm256_cvtph_epi32): Likewise.
12929 (_mm256_mask_cvtph_epi32): Likewise.
12930 (_mm256_maskz_cvtph_epi32): Likewise.
12931 (_mm_cvtph_epu32): Likewise.
12932 (_mm_mask_cvtph_epu32): Likewise.
12933 (_mm_maskz_cvtph_epu32): Likewise.
12934 (_mm256_cvtph_epu32): Likewise.
12935 (_mm256_mask_cvtph_epu32): Likewise.
12936 (_mm256_maskz_cvtph_epu32): Likewise.
12937 (_mm_cvtph_epi64): Likewise.
12938 (_mm_mask_cvtph_epi64): Likewise.
12939 (_mm_maskz_cvtph_epi64): Likewise.
12940 (_mm256_cvtph_epi64): Likewise.
12941 (_mm256_mask_cvtph_epi64): Likewise.
12942 (_mm256_maskz_cvtph_epi64): Likewise.
12943 (_mm_cvtph_epu64): Likewise.
12944 (_mm_mask_cvtph_epu64): Likewise.
12945 (_mm_maskz_cvtph_epu64): Likewise.
12946 (_mm256_cvtph_epu64): Likewise.
12947 (_mm256_mask_cvtph_epu64): Likewise.
12948 (_mm256_maskz_cvtph_epu64): Likewise.
12949 (_mm_cvtph_epi16): Likewise.
12950 (_mm_mask_cvtph_epi16): Likewise.
12951 (_mm_maskz_cvtph_epi16): Likewise.
12952 (_mm256_cvtph_epi16): Likewise.
12953 (_mm256_mask_cvtph_epi16): Likewise.
12954 (_mm256_maskz_cvtph_epi16): Likewise.
12955 (_mm_cvtph_epu16): Likewise.
12956 (_mm_mask_cvtph_epu16): Likewise.
12957 (_mm_maskz_cvtph_epu16): Likewise.
12958 (_mm256_cvtph_epu16): Likewise.
12959 (_mm256_mask_cvtph_epu16): Likewise.
12960 (_mm256_maskz_cvtph_epu16): Likewise.
12961 * config/i386/i386-builtin-types.def: Add new builtin types.
12962 * config/i386/i386-builtin.def: Add new builtins.
12963 * config/i386/i386-expand.c
12964 (ix86_expand_args_builtin): Handle new builtin types.
12965 (ix86_expand_round_builtin): Ditto.
12966 * config/i386/sse.md (sseintconvert): New.
12967 (ssePHmode): Ditto.
12968 (UNSPEC_US_FIX_NOTRUNC): Ditto.
12969 (sseintconvertsignprefix): Ditto.
12970 (avx512fp16_vcvtph2<sseintconvertsignprefix><sseintconvert>_<mode><mask_name><round_name>):
12973 2021-09-16 liuhongt <hongtao.liu@intel.com>
12975 * config/i386/avx512fp16intrin.h: (_mm_cvtsi16_si128):
12977 (_mm_cvtsi128_si16): Likewise.
12978 (_mm_mask_load_sh): Likewise.
12979 (_mm_maskz_load_sh): Likewise.
12980 (_mm_mask_store_sh): Likewise.
12981 (_mm_move_sh): Likewise.
12982 (_mm_mask_move_sh): Likewise.
12983 (_mm_maskz_move_sh): Likewise.
12984 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
12985 * config/i386/i386-builtin.def: Add corresponding new builtins.
12986 * config/i386/i386-expand.c
12987 (ix86_expand_special_args_builtin): Handle new builtin types.
12988 (ix86_expand_vector_init_one_nonzero): Adjust for FP16 target.
12989 * config/i386/sse.md (VI2F): New mode iterator.
12990 (vec_set<mode>_0): Use new mode iterator.
12991 (avx512f_mov<ssescalarmodelower>_mask): Adjust for HF vector mode.
12992 (avx512f_store<mode>_mask): Ditto.
12994 2021-09-16 Kewen Lin <linkw@linux.ibm.com>
12996 * config/rs6000/rs6000.opt (-mtoc-fusion): Remove.
12998 2021-09-15 David Edelsohn <dje.gcc@gmail.com>
13000 * config/rs6000/rs6000.c (rs6000_xcoff_encode_section_info):
13001 Proceed if no symbol summary or the symbol alias flag is false.
13003 2021-09-15 Jakub Jelinek <jakub@redhat.com>
13007 * varasm.c (output_constructor_regular_field): Instead of assertion
13008 that array_size_for_constructor result is equal to size of
13009 TREE_TYPE (local->val) in bytes, assert that the type size is greater
13010 or equal to array_size_for_constructor result and use type size as
13013 2021-09-15 Martin Liska <mliska@suse.cz>
13016 * config/i386/vxworks.h: Use new macro TARGET_CPU_P.
13018 2021-09-15 Martin Liska <mliska@suse.cz>
13021 * config/rs6000/rs6000.c (rs6000_xcoff_encode_section_info):
13022 Check that we have a symbol summary for a symbol.
13024 2021-09-15 Richard Biener <rguenther@suse.de>
13027 * config/rs6000/lynx.h: Remove undef of PREFERRED_DEBUGGING_TYPE
13028 to inherit from elfos.h
13030 2021-09-15 liuhongt <hongtao.liu@intel.com>
13033 * config/i386/i386-expand.c
13034 (ix86_expand_vector_init_interleave): Use puncklwd to pack 2
13036 (ix86_expand_vector_set): Use blendw instead of pinsrw.
13037 * config/i386/i386.c (ix86_can_change_mode_class): Adjust for
13038 AVX512FP16 which supports 16bit vector load.
13039 * config/i386/sse.md (avx512bw_interleave_highv32hi<mask_name>):
13041 (avx512bw_interleave_high<mode><mask_name>): .. this, and
13042 extend to V32HFmode.
13043 (avx2_interleave_highv16hi<mask_name>): Rename to ..
13044 (avx2_interleave_high<mode><mask_name>): .. this, and extend
13046 (vec_interleave_highv8hi<mask_name>): Rename to ..
13047 (vec_interleave_high<mode><mask_name>): .. this, and extend to V8HFmode.
13048 (<mask_codefor>avx512bw_interleave_lowv32hi<mask_name>):
13050 (<mask_codefor>avx512bw_interleave_low<mode><mask_name>):
13051 this, and extend to V32HFmode.
13052 (avx2_interleave_lowv16hi<mask_name>): Rename to ..
13053 (avx2_interleave_low<mode><mask_name>): .. this, and extend to V16HFmode.
13054 (vec_interleave_lowv8hi<mask_name>): Rename to ..
13055 (vec_interleave_low<mode><mask_name>): .. this, and extend to V8HFmode.
13056 (sse4_1_pblendw): Rename to ..
13057 (sse4_1_pblend<blendsuf>): .. this, and extend to V8HFmode.
13058 (avx2_pblendph): New define_expand.
13059 (<sse2p4_1>_pinsr<ssemodesuffix>): Refactor, use
13060 sseintmodesuffix instead of ssemodesuffix.
13061 (blendsuf): New mode attr.
13063 2021-09-15 Richard Biener <rguenther@suse.de>
13065 * tree-vectorizer.h (dr_misalignment): Move out of line.
13066 (dr_target_alignment): New.
13067 (DR_TARGET_ALIGNMENT): Wrap dr_target_alignment.
13068 (set_dr_target_alignment): New.
13069 (SET_DR_TARGET_ALIGNMENT): Wrap set_dr_target_alignment.
13070 * tree-vect-data-refs.c (dr_misalignment): Compute and
13071 return the group members misalignment.
13072 (vect_compute_data_ref_alignment): Use SET_DR_TARGET_ALIGNMENT.
13073 (vect_analyze_data_refs_alignment): Compute alignment only
13074 for the first element of a DR group.
13075 (vect_slp_analyze_node_alignment): Likewise.
13077 2021-09-15 Hongyu Wang <hongyu.wang@intel.com>
13079 * config/i386/avx512fp16intrin.h: Adjust all builtin calls.
13080 * config/i386/avx512fp16vlintrin.h: Likewise.
13081 * config/i386/i386-builtin.def: Adjust builtin name and
13082 enumeration to match AVX512F style.
13084 2021-09-15 Richard Biener <rguenther@suse.de>
13086 PR tree-optimization/102318
13087 * tree-vect-loop.c (vect_transform_cycle_phi): Revert
13088 previous change and do the mode conversion separately from
13089 the sign conversion.
13091 2021-09-15 Hongtao Liu <hongtao.liu@intel.com>
13092 Peter Cordes <peter@cordes.ca>
13095 * config/i386/sse.md (extract_suf): Add V8SF/V8SI/V4DF/V4DI.
13096 (*vec_extract<mode><ssescalarmodelower>_valign): Output
13097 vextract{i,f}{32x4,64x2} instruction when byte_offset % 16 ==
13100 2021-09-15 Richard Biener <rguenther@suse.de>
13102 * config.gcc: Remove vax-*-openbsd* configuration.
13104 2021-09-15 Richard Biener <rguenther@suse.de>
13106 * config.gcc: Remove m68k-openbsd.
13108 2021-09-15 Max Filippov <jcmvbkbc@gmail.com>
13111 * config/xtensa/t-xtensa (TM_H): Add include/xtensa-config.h.
13113 2021-09-14 Peter Bergner <bergner@linux.ibm.com>
13115 * config/rs6000/mma.md (unspec): Delete UNSPEC_MMA_XXSETACCZ.
13116 (unspecv): Add UNSPECV_MMA_XXSETACCZ.
13117 (*mma_xxsetaccz): Delete.
13118 (mma_xxsetaccz): Change to define_insn. Remove operand 1.
13119 Use UNSPECV_MMA_XXSETACCZ. Update comment.
13120 * config/rs6000/rs6000.c (rs6000_rtx_costs): Use UNSPECV_MMA_XXSETACCZ.
13122 2021-09-14 Iain Sandoe <iain@sandoe.co.uk>
13124 * Makefile.in: Remove variables related to applying no-PIE
13125 to the exes on $build.
13126 * configure: Regenerate.
13127 * configure.ac: Remove configuration related to applying
13128 no-PIE to the exes on $build.
13130 2021-09-14 Claudiu Zissulescu <claziss@synopsys.com>
13132 * config/arc/arc.md (doloop_end): Add missing mode.
13133 (loop_end): Likewise.
13135 2021-09-14 Jakub Jelinek <jakub@redhat.com>
13137 * gimplify.c (goa_stabilize_expr): Add depth argument, propagate
13138 it to recursive calls, for depth above 7 just gimplify or return.
13139 Perform a test even for MODIFY_EXPR, ADDR_EXPR, COMPOUND_EXPR with
13140 __builtin_clear_padding and TARGET_EXPR.
13141 (gimplify_omp_atomic): Adjust goa_stabilize_expr callers.
13143 2021-09-14 liuhongt <hongtao.liu@intel.com>
13145 * config/i386/avx512fp16intrin.h (_mm_fpclass_sh_mask):
13147 (_mm_mask_fpclass_sh_mask): Likewise.
13148 (_mm512_mask_fpclass_ph_mask): Likewise.
13149 (_mm512_fpclass_ph_mask): Likewise.
13150 (_mm_getexp_sh): Likewise.
13151 (_mm_mask_getexp_sh): Likewise.
13152 (_mm_maskz_getexp_sh): Likewise.
13153 (_mm512_getexp_ph): Likewise.
13154 (_mm512_mask_getexp_ph): Likewise.
13155 (_mm512_maskz_getexp_ph): Likewise.
13156 (_mm_getexp_round_sh): Likewise.
13157 (_mm_mask_getexp_round_sh): Likewise.
13158 (_mm_maskz_getexp_round_sh): Likewise.
13159 (_mm512_getexp_round_ph): Likewise.
13160 (_mm512_mask_getexp_round_ph): Likewise.
13161 (_mm512_maskz_getexp_round_ph): Likewise.
13162 (_mm_getmant_sh): Likewise.
13163 (_mm_mask_getmant_sh): Likewise.
13164 (_mm_maskz_getmant_sh): Likewise.
13165 (_mm512_getmant_ph): Likewise.
13166 (_mm512_mask_getmant_ph): Likewise.
13167 (_mm512_maskz_getmant_ph): Likewise.
13168 (_mm_getmant_round_sh): Likewise.
13169 (_mm_mask_getmant_round_sh): Likewise.
13170 (_mm_maskz_getmant_round_sh): Likewise.
13171 (_mm512_getmant_round_ph): Likewise.
13172 (_mm512_mask_getmant_round_ph): Likewise.
13173 (_mm512_maskz_getmant_round_ph): Likewise.
13174 * config/i386/avx512fp16vlintrin.h (_mm_mask_fpclass_ph_mask):
13176 (_mm_fpclass_ph_mask): Likewise.
13177 (_mm256_mask_fpclass_ph_mask): Likewise.
13178 (_mm256_fpclass_ph_mask): Likewise.
13179 (_mm256_getexp_ph): Likewise.
13180 (_mm256_mask_getexp_ph): Likewise.
13181 (_mm256_maskz_getexp_ph): Likewise.
13182 (_mm_getexp_ph): Likewise.
13183 (_mm_mask_getexp_ph): Likewise.
13184 (_mm_maskz_getexp_ph): Likewise.
13185 (_mm256_getmant_ph): Likewise.
13186 (_mm256_mask_getmant_ph): Likewise.
13187 (_mm256_maskz_getmant_ph): Likewise.
13188 (_mm_getmant_ph): Likewise.
13189 (_mm_mask_getmant_ph): Likewise.
13190 (_mm_maskz_getmant_ph): Likewise.
13191 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
13192 * config/i386/i386-builtin.def: Add corresponding new builtins.
13193 * config/i386/i386-expand.c
13194 (ix86_expand_args_builtin): Handle new builtin types.
13195 (ix86_expand_round_builtin): Ditto.
13196 * config/i386/sse.md (vecmemsuffix): Add HF vector modes.
13197 (<avx512>_getexp<mode><mask_name><round_saeonly_name>): Adjust
13198 to support HF vector modes.
13199 (avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name):
13201 (avx512dq_fpclass<mode><mask_scalar_merge_name>): Ditto.
13202 (avx512dq_vmfpclass<mode><mask_scalar_merge_name>): Ditto.
13203 (<avx512>_getmant<mode><mask_name><round_saeonly_name>): Ditto.
13204 (avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
13207 2021-09-14 liuhongt <hongtao.liu@intel.com>
13209 * config/i386/avx512fp16intrin.h (_mm512_reduce_ph):
13211 (_mm512_mask_reduce_ph): Likewise.
13212 (_mm512_maskz_reduce_ph): Likewise.
13213 (_mm512_reduce_round_ph): Likewise.
13214 (_mm512_mask_reduce_round_ph): Likewise.
13215 (_mm512_maskz_reduce_round_ph): Likewise.
13216 (_mm_reduce_sh): Likewise.
13217 (_mm_mask_reduce_sh): Likewise.
13218 (_mm_maskz_reduce_sh): Likewise.
13219 (_mm_reduce_round_sh): Likewise.
13220 (_mm_mask_reduce_round_sh): Likewise.
13221 (_mm_maskz_reduce_round_sh): Likewise.
13222 (_mm512_roundscale_ph): Likewise.
13223 (_mm512_mask_roundscale_ph): Likewise.
13224 (_mm512_maskz_roundscale_ph): Likewise.
13225 (_mm512_roundscale_round_ph): Likewise.
13226 (_mm512_mask_roundscale_round_ph): Likewise.
13227 (_mm512_maskz_roundscale_round_ph): Likewise.
13228 (_mm_roundscale_sh): Likewise.
13229 (_mm_mask_roundscale_sh): Likewise.
13230 (_mm_maskz_roundscale_sh): Likewise.
13231 (_mm_roundscale_round_sh): Likewise.
13232 (_mm_mask_roundscale_round_sh): Likewise.
13233 (_mm_maskz_roundscale_round_sh): Likewise.
13234 * config/i386/avx512fp16vlintrin.h: (_mm_reduce_ph):
13236 (_mm_mask_reduce_ph): Likewise.
13237 (_mm_maskz_reduce_ph): Likewise.
13238 (_mm256_reduce_ph): Likewise.
13239 (_mm256_mask_reduce_ph): Likewise.
13240 (_mm256_maskz_reduce_ph): Likewise.
13241 (_mm_roundscale_ph): Likewise.
13242 (_mm_mask_roundscale_ph): Likewise.
13243 (_mm_maskz_roundscale_ph): Likewise.
13244 (_mm256_roundscale_ph): Likewise.
13245 (_mm256_mask_roundscale_ph): Likewise.
13246 (_mm256_maskz_roundscale_ph): Likewise.
13247 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
13248 * config/i386/i386-builtin.def: Add corresponding new builtins.
13249 * config/i386/i386-expand.c
13250 (ix86_expand_args_builtin): Handle new builtin types.
13251 (ix86_expand_round_builtin): Ditto.
13252 * config/i386/sse.md (<mask_codefor>reducep<mode><mask_name>):
13254 (<mask_codefor>reducep<mode><mask_name><round_saeonly_name>):
13255 ... this, and adjust for round operands.
13256 (reduces<mode><mask_scalar_name>): Likewise, with ...
13257 (reduces<mode><mask_scalar_name><round_saeonly_scalar_name):
13259 (<avx512>_rndscale<mode><mask_name><round_saeonly_name>):
13260 Adjust for HF vector modes.
13261 (avx512f_rndscale<mode><mask_scalar_name><round_saeonly_scalar_name>):
13263 (*avx512f_rndscale<mode><round_saeonly_name>): Ditto.
13265 2021-09-14 liuhongt <hongtao.liu@intel.com>
13267 * config/i386/avx512fp16intrin.h: (_mm512_rcp_ph):
13269 (_mm512_mask_rcp_ph): Likewise.
13270 (_mm512_maskz_rcp_ph): Likewise.
13271 (_mm_rcp_sh): Likewise.
13272 (_mm_mask_rcp_sh): Likewise.
13273 (_mm_maskz_rcp_sh): Likewise.
13274 (_mm512_scalef_ph): Likewise.
13275 (_mm512_mask_scalef_ph): Likewise.
13276 (_mm512_maskz_scalef_ph): Likewise.
13277 (_mm512_scalef_round_ph): Likewise.
13278 (_mm512_mask_scalef_round_ph): Likewise.
13279 (_mm512_maskz_scalef_round_ph): Likewise.
13280 (_mm_scalef_sh): Likewise.
13281 (_mm_mask_scalef_sh): Likewise.
13282 (_mm_maskz_scalef_sh): Likewise.
13283 (_mm_scalef_round_sh): Likewise.
13284 (_mm_mask_scalef_round_sh): Likewise.
13285 (_mm_maskz_scalef_round_sh): Likewise.
13286 * config/i386/avx512fp16vlintrin.h (_mm_rcp_ph):
13288 (_mm256_rcp_ph): Likewise.
13289 (_mm_mask_rcp_ph): Likewise.
13290 (_mm256_mask_rcp_ph): Likewise.
13291 (_mm_maskz_rcp_ph): Likewise.
13292 (_mm256_maskz_rcp_ph): Likewise.
13293 (_mm_scalef_ph): Likewise.
13294 (_mm256_scalef_ph): Likewise.
13295 (_mm_mask_scalef_ph): Likewise.
13296 (_mm256_mask_scalef_ph): Likewise.
13297 (_mm_maskz_scalef_ph): Likewise.
13298 (_mm256_maskz_scalef_ph): Likewise.
13299 * config/i386/i386-builtin.def: Add new builtins.
13300 * config/i386/sse.md (VFH_AVX512VL): New.
13301 (avx512fp16_rcp<mode>2<mask_name>): Ditto.
13302 (avx512fp16_vmrcpv8hf2<mask_scalar_name>): Ditto.
13303 (avx512f_vmscalef<mode><mask_scalar_name><round_scalar_name>):
13304 Adjust to support HF vector modes.
13305 (<avx512>_scalef<mode><mask_name><round_name>): Ditto.
13307 2021-09-14 liuhongt <hongtao.liu@intel.com>
13309 * config/i386/avx512fp16intrin.h: (_mm512_sqrt_ph):
13311 (_mm512_mask_sqrt_ph): Likewise.
13312 (_mm512_maskz_sqrt_ph): Likewise.
13313 (_mm512_sqrt_round_ph): Likewise.
13314 (_mm512_mask_sqrt_round_ph): Likewise.
13315 (_mm512_maskz_sqrt_round_ph): Likewise.
13316 (_mm512_rsqrt_ph): Likewise.
13317 (_mm512_mask_rsqrt_ph): Likewise.
13318 (_mm512_maskz_rsqrt_ph): Likewise.
13319 (_mm_rsqrt_sh): Likewise.
13320 (_mm_mask_rsqrt_sh): Likewise.
13321 (_mm_maskz_rsqrt_sh): Likewise.
13322 (_mm_sqrt_sh): Likewise.
13323 (_mm_mask_sqrt_sh): Likewise.
13324 (_mm_maskz_sqrt_sh): Likewise.
13325 (_mm_sqrt_round_sh): Likewise.
13326 (_mm_mask_sqrt_round_sh): Likewise.
13327 (_mm_maskz_sqrt_round_sh): Likewise.
13328 * config/i386/avx512fp16vlintrin.h (_mm_sqrt_ph): New intrinsic.
13329 (_mm256_sqrt_ph): Likewise.
13330 (_mm_mask_sqrt_ph): Likewise.
13331 (_mm256_mask_sqrt_ph): Likewise.
13332 (_mm_maskz_sqrt_ph): Likewise.
13333 (_mm256_maskz_sqrt_ph): Likewise.
13334 (_mm_rsqrt_ph): Likewise.
13335 (_mm256_rsqrt_ph): Likewise.
13336 (_mm_mask_rsqrt_ph): Likewise.
13337 (_mm256_mask_rsqrt_ph): Likewise.
13338 (_mm_maskz_rsqrt_ph): Likewise.
13339 (_mm256_maskz_rsqrt_ph): Likewise.
13340 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
13341 * config/i386/i386-builtin.def: Add corresponding new builtins.
13342 * config/i386/i386-expand.c
13343 (ix86_expand_args_builtin): Handle new builtins.
13344 (ix86_expand_round_builtin): Ditto.
13345 * config/i386/sse.md (VF_AVX512FP16VL): New.
13346 (sqrt<mode>2): Adjust for HF vector modes.
13347 (<sse>_sqrt<mode>2<mask_name><round_name>): Likewise.
13348 (<sse>_vmsqrt<mode>2<mask_scalar_name><round_scalar_name>):
13350 (<sse>_rsqrt<mode>2<mask_name>): New.
13351 (avx512fp16_vmrsqrtv8hf2<mask_scalar_name>): Likewise.
13353 2021-09-13 Thomas Schwinge <thomas@codesourcery.com>
13355 PR bootstrap/101574
13356 * diagnostic-spec.c (warning_suppressed_at, copy_warning): Handle
13357 'RESERVED_LOCATION_P' locations.
13358 * warning-control.cc (get_nowarn_spec, suppress_warning)
13359 (copy_warning): Likewise.
13361 2021-09-13 Thomas Schwinge <thomas@codesourcery.com>
13363 * diagnostic-spec.h (typedef xint_hash_t): Use 'location_t' instead of...
13364 (typedef key_type_t): ... this. Remove.
13365 (nowarn_map): Document.
13366 * diagnostic-spec.c (nowarn_map): Likewise.
13367 * warning-control.cc (convert_to_key): Evolve functions into...
13368 (get_location): ... these. Adjust all users.
13370 2021-09-13 Thomas Schwinge <thomas@codesourcery.com>
13372 * warning-control.cc (copy_warning): Remove 'nowarn_map' setup.
13374 2021-09-13 Jason Merrill <jason@redhat.com>
13376 * params.opt: Add destructive-interference-size and
13377 constructive-interference-size.
13378 * doc/invoke.texi: Document them.
13379 * config/aarch64/aarch64.c (aarch64_override_options_internal):
13381 * config/arm/arm.c (arm_option_override): Set them.
13382 * config/i386/i386-options.c (ix86_option_override_internal):
13385 2021-09-13 Martin Liska <mliska@suse.cz>
13386 H.J. Lu <hjl.tools@gmail.com>
13389 * common/config/i386/cpuinfo.h (cpu_indicator_init): Add support
13390 for x86-64 micro levels for __builtin_cpu_supports.
13391 * common/config/i386/i386-cpuinfo.h (enum feature_priority):
13392 Add priorities for the micro-arch levels.
13393 (enum processor_features): Add new features.
13394 * common/config/i386/i386-isas.h: Add micro-arch features.
13395 * config/i386/i386-builtins.c (get_builtin_code_for_version):
13396 Support the micro-arch levels by callsing
13397 __builtin_cpu_supports.
13398 * doc/extend.texi: Document that the levels are support by
13399 __builtin_cpu_supports.
13401 2021-09-13 Andrew Pinski <apinski@marvell.com>
13404 * config/aarch64/aarch64-builtins.c (aarch64_fold_builtin_lane_check):
13406 (aarch64_general_fold_builtin): Handle AARCH64_SIMD_BUILTIN_LANE_CHECK.
13407 (aarch64_general_gimple_fold_builtin): Likewise.
13409 2021-09-13 Andrew Pinski <apinski@marvell.com>
13411 * config.gcc: Add m32r-*-linux* and m32rle-*-linux*
13412 to the Unsupported targets list.
13413 Remove support for m32r-*-linux* and m32rle-*-linux*.
13414 * config/m32r/linux.h: Removed.
13415 * config/m32r/t-linux: Removed.
13417 2021-09-13 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13420 * config/aarch64/aarch64.c (aarch64_classify_address): Don't allow
13421 register index for SVE predicate modes.
13423 2021-09-13 Aldy Hernandez <aldyh@redhat.com>
13425 * tree-ssa-threadbackward.c
13426 (back_threader_profitability::profitable_path_p): Remove FSM
13428 (back_threader_registry::register_path): Same.
13429 * tree-ssa-threadedge.c
13430 (jump_threader::simplify_control_stmt_condition): Same.
13431 * tree-ssa-threadupdate.c (jt_path_registry::jt_path_registry):
13432 Add backedge_threads argument.
13433 (fwd_jt_path_registry::fwd_jt_path_registry): Pass
13434 backedge_threads argument.
13435 (back_jt_path_registry::back_jt_path_registry): Same.
13436 (dump_jump_thread_path): Adjust for FSM removal.
13437 (back_jt_path_registry::rewire_first_differing_edge): Same.
13438 (back_jt_path_registry::adjust_paths_after_duplication): Same.
13439 (back_jt_path_registry::update_cfg): Same.
13440 (jt_path_registry::register_jump_thread): Same.
13441 * tree-ssa-threadupdate.h (enum jump_thread_edge_type): Remove
13443 (class back_jt_path_registry): Add backedge_threads to
13446 2021-09-13 Martin Liska <mliska@suse.cz>
13449 * asan.h (sanitize_coverage_p): Handle when fn == NULL.
13451 2021-09-13 H.J. Lu <hjl.tools@gmail.com>
13454 * config/i386/i386.h (TARGET_AVX256_MOVE_BY_PIECES): New.
13455 (TARGET_AVX256_STORE_BY_PIECES): Likewise.
13456 (MOVE_MAX): Check TARGET_AVX256_MOVE_BY_PIECES and
13457 TARGET_AVX256_STORE_BY_PIECES instead of
13458 TARGET_AVX256_SPLIT_UNALIGNED_LOAD and
13459 TARGET_AVX256_SPLIT_UNALIGNED_STORE.
13460 (STORE_MAX_PIECES): Check TARGET_AVX256_STORE_BY_PIECES instead
13461 of TARGET_AVX256_SPLIT_UNALIGNED_STORE.
13462 * config/i386/x86-tune.def (X86_TUNE_AVX256_MOVE_BY_PIECES): New.
13463 (X86_TUNE_AVX256_STORE_BY_PIECES): Likewise.
13465 2021-09-13 liuhongt <hongtao.liu@intel.com>
13467 PR bootstrap/102302
13468 * expmed.c (extract_bit_field_using_extv): Use
13469 gen_lowpart_if_possible instead of gen_lowpart to avoid ICE.
13471 2021-09-13 Aldy Hernandez <aldyh@redhat.com>
13473 * Makefile.in (OBJS): Add value-pointer-equiv.o.
13474 * gimple-ssa-evrp.c (class ssa_equiv_stack): Move to
13475 value-pointer-equiv.*.
13476 (ssa_equiv_stack::ssa_equiv_stack): Same.
13477 (ssa_equiv_stack::enter): Same.
13478 (ssa_equiv_stack::leave): Same.
13479 (ssa_equiv_stack::push_replacement): Same.
13480 (ssa_equiv_stack::get_replacement): Same.
13481 (is_pointer_ssa): Same.
13482 (class pointer_equiv_analyzer): Same.
13483 (pointer_equiv_analyzer::pointer_equiv_analyzer): Same.
13484 (pointer_equiv_analyzer::~pointer_equiv_analyzer): Same.
13485 (pointer_equiv_analyzer::set_global_equiv): Same.
13486 (pointer_equiv_analyzer::set_cond_equiv): Same.
13487 (pointer_equiv_analyzer::get_equiv): Same.
13488 (pointer_equiv_analyzer::enter): Same.
13489 (pointer_equiv_analyzer::leave): Same.
13490 (pointer_equiv_analyzer::get_equiv_expr): Same.
13491 (pta_valueize): Same.
13492 (pointer_equiv_analyzer::visit_stmt): Same.
13493 (pointer_equiv_analyzer::visit_edge): Same.
13494 (hybrid_folder::value_of_expr): Same.
13495 (hybrid_folder::value_on_edge): Same.
13496 * value-pointer-equiv.cc: New file.
13497 * value-pointer-equiv.h: New file.
13499 2021-09-13 Richard Earnshaw <rearnsha@arm.com>
13502 * gimple-fold.c (gimple_fold_builtin_memory_op): Allow folding
13503 memcpy if the size is not more than MOVE_MAX * MOVE_RATIO.
13505 2021-09-13 Richard Earnshaw <rearnsha@arm.com>
13508 * config/arm/arm.md (movmisaligndi): New define_expand.
13509 * config/arm/vec-common.md (movmisalign<mode>): Iterate over VDQ mode.
13511 2021-09-13 Richard Earnshaw <rearnsha@arm.com>
13514 * emit-rtl.c (gen_highpart): Use adjust_address to handle
13515 MEM rather than calling simplify_gen_subreg.
13517 2021-09-13 Jan-Benedict Glaw <jbglaw@ług-owl.de>
13519 * config/alpha/vms.h (INIT_CUMULATIVE_ARGS): Wrap multi-statment
13520 define into a block.
13522 2021-09-13 Richard Biener <rguenther@suse.de>
13524 * config/darwin.h (DARWIN_PREFER_DWARF): Do not define.
13525 * config/i386/darwin.h (PREFERRED_DEBUGGING_TYPE): Do not
13526 change based on DARWIN_PREFER_DWARF not being defined.
13528 2021-09-13 Richard Biener <rguenther@suse.de>
13530 * config/i386/lynx.h: Remove undef of PREFERRED_DEBUGGING_TYPE
13531 to inherit from elfos.h
13533 2021-09-13 Richard Biener <rguenther@suse.de>
13535 * config.gcc: Add cr16-*-* to the list of obsoleted targets.
13537 2021-09-13 Richard Biener <rguenther@suse.de>
13539 * config/avr/elf.h (PREFERRED_DEBUGGING_TYPE): Remove
13540 override, pick up DWARF2_DEBUG define from elfos.h
13542 2021-09-13 Richard Biener <rguenther@suse.de>
13544 * config/rx/rx.h (PREFERRED_DEBUGGING_TYPE): Always define to
13547 2021-09-13 Richard Biener <rguenther@suse.de>
13549 * config/alpha/vms.h (PREFERRED_DEBUGGING_TYPE): Define to
13552 2021-09-13 Richard Biener <rguenther@suse.de>
13554 * config/i386/cygming.h: Always default to DWARF2 debugging.
13555 Do not define DBX_DEBUGGING_INFO, that's done via dbxcoff.h
13557 * doc/install.texi: Document binutils 2.16 as minimum
13558 requirement for mingw.
13560 2021-09-13 Kewen Lin <linkw@linux.ibm.com>
13562 * config/rs6000/rs6000.c (struct rs6000_cost_data): New members
13563 nstmts, nloads and extra_ctor_cost.
13564 (rs6000_density_test): Add load density related heuristics. Do
13565 extra costing on vector construction statements if need.
13566 (rs6000_init_cost): Init new members.
13567 (rs6000_update_target_cost_per_stmt): New function.
13568 (rs6000_add_stmt_cost): Factor vect_nonmem hunk out to function
13569 rs6000_update_target_cost_per_stmt and call it.
13571 2021-09-13 Kewen Lin <linkw@linux.ibm.com>
13573 * config/rs6000/rs6000.c (struct rs6000_cost_data): Remove typedef.
13574 (rs6000_init_cost): Adjust.
13576 2021-09-13 liuhongt <hongtao.liu@intel.com>
13578 * config/i386/i386.md: (UNSPEC_COPYSIGN): Remove.
13579 (UNSPEC_XORSIGN): Ditto.
13581 2021-09-12 Roger Sayle <roger@nextmovesoftware.com>
13583 * expr.c (convert_move): Preserve SUBREG_PROMOTED_VAR_P when
13584 creating a (wider) partial subreg from a SUBREG_PROMOTED_VAR_P
13587 2021-09-11 Aldy Hernandez <aldyh@redhat.com>
13589 * tree-ssa-threadbackward.c (class back_threader_registry): Use
13590 back_jt_path_registry.
13591 * tree-ssa-threadedge.c (jump_threader::jump_threader): Use
13592 fwd_jt_path_registry.
13593 * tree-ssa-threadedge.h (class jump_threader): Same..
13594 * tree-ssa-threadupdate.c
13595 (jump_thread_path_registry::jump_thread_path_registry): Rename...
13596 (jt_path_registry::jt_path_registry): ...to this.
13597 (jump_thread_path_registry::~jump_thread_path_registry): Rename...
13598 (jt_path_registry::~jt_path_registry): ...this.
13599 (fwd_jt_path_registry::fwd_jt_path_registry): New.
13600 (fwd_jt_path_registry::~fwd_jt_path_registry): New.
13601 (jump_thread_path_registry::allocate_thread_edge): Rename...
13602 (jt_path_registry::allocate_thread_edge): ...to this.
13603 (jump_thread_path_registry::allocate_thread_path): Rename...
13604 (jt_path_registry::allocate_thread_path): ...to this.
13605 (jump_thread_path_registry::lookup_redirection_data): Rename...
13606 (fwd_jt_path_registry::lookup_redirection_data): ...to this.
13607 (jump_thread_path_registry::thread_block_1): Rename...
13608 (fwd_jt_path_registry::thread_block_1): ...to this.
13609 (jump_thread_path_registry::thread_block): Rename...
13610 (fwd_jt_path_registry::thread_block): ...to this.
13611 (jt_path_registry::thread_through_loop_header): Rename...
13612 (fwd_jt_path_registry::thread_through_loop_header): ...to this.
13613 (jump_thread_path_registry::mark_threaded_blocks): Rename...
13614 (fwd_jt_path_registry::mark_threaded_blocks): ...to this.
13615 (jump_thread_path_registry::debug_path): Rename...
13616 (jt_path_registry::debug_path): ...to this.
13617 (jump_thread_path_registry::dump): Rename...
13618 (jt_path_registry::debug): ...to this.
13619 (jump_thread_path_registry::rewire_first_differing_edge): Rename...
13620 (back_jt_path_registry::rewire_first_differing_edge): ...to this.
13621 (jump_thread_path_registry::adjust_paths_after_duplication): Rename...
13622 (back_jt_path_registry::adjust_paths_after_duplication): ...to this.
13623 (jump_thread_path_registry::duplicate_thread_path): Rename...
13624 (back_jt_path_registry::duplicate_thread_path): ...to this. Also,
13625 drop ill-formed candidates.
13626 (jump_thread_path_registry::remove_jump_threads_including): Rename...
13627 (fwd_jt_path_registry::remove_jump_threads_including): ...to this.
13628 (jt_path_registry::thread_through_all_blocks): New.
13629 (back_jt_path_registry::update_cfg): New.
13630 (fwd_jt_path_registry::update_cfg): New.
13631 (jump_thread_path_registry::register_jump_thread): Rename...
13632 (jt_path_registry::register_jump_thread): ...to this.
13633 * tree-ssa-threadupdate.h (class jump_thread_path_registry):
13635 (class jt_path_registry): ...here.
13636 (class fwd_jt_path_registry): New.
13637 (class back_jt_path_registry): New.
13639 2021-09-10 liuhongt <hongtao.liu@intel.com>
13642 2021-09-01 liuhongt <hongtao.liu@intel.com>
13644 * emit-rtl.c (validate_subreg): Get rid of all float-int
13647 2021-09-10 Jakub Jelinek <jakub@redhat.com>
13649 * tree-core.h (enum omp_memory_order): Add OMP_MEMORY_ORDER_MASK,
13650 OMP_FAIL_MEMORY_ORDER_UNSPECIFIED, OMP_FAIL_MEMORY_ORDER_RELAXED,
13651 OMP_FAIL_MEMORY_ORDER_ACQUIRE, OMP_FAIL_MEMORY_ORDER_RELEASE,
13652 OMP_FAIL_MEMORY_ORDER_ACQ_REL, OMP_FAIL_MEMORY_ORDER_SEQ_CST and
13653 OMP_FAIL_MEMORY_ORDER_MASK enumerators.
13654 (OMP_FAIL_MEMORY_ORDER_SHIFT): Define.
13655 * gimple-pretty-print.c (dump_gimple_omp_atomic_load,
13656 dump_gimple_omp_atomic_store): Print [weak] for weak atomic
13658 * gimple.h (enum gf_mask): Change GF_OMP_ATOMIC_MEMORY_ORDER
13659 to 6-bit mask, adjust GF_OMP_ATOMIC_NEED_VALUE value and add
13660 GF_OMP_ATOMIC_WEAK.
13661 (gimple_omp_atomic_weak_p, gimple_omp_atomic_set_weak): New inline
13663 * tree.h (OMP_ATOMIC_WEAK): Define.
13664 * tree-pretty-print.c (dump_omp_atomic_memory_order): Adjust for
13665 fail memory order being encoded in the same enum and also print
13666 fail clause if present.
13667 (dump_generic_node): Print weak clause if OMP_ATOMIC_WEAK.
13668 * gimplify.c (goa_stabilize_expr): Add target_expr and rhs arguments,
13669 handle pre_p == NULL case as a test mode that only returns value
13670 but doesn't change gimplify nor change anything otherwise, adjust
13671 recursive calls, add MODIFY_EXPR, ADDR_EXPR, COND_EXPR, TARGET_EXPR
13672 and CALL_EXPR handling, adjust COMPOUND_EXPR handling for
13673 __builtin_clear_padding calls, for !rhs gimplify as lvalue rather
13675 (gimplify_omp_atomic): Adjust goa_stabilize_expr caller. Handle
13676 COND_EXPR rhs. Set weak flag on gimple load/store for
13678 * omp-expand.c (omp_memory_order_to_fail_memmodel): New function.
13679 (omp_memory_order_to_memmodel): Adjust for fail clause encoded
13681 (expand_omp_atomic_cas): New function.
13682 (expand_omp_atomic_pipeline): Use omp_memory_order_to_fail_memmodel
13684 (expand_omp_atomic): Attempt to optimize atomic compare and exchange
13685 using expand_omp_atomic_cas.
13687 2021-09-10 Aldy Hernandez <aldyh@redhat.com>
13688 Michael Matz <matz@suse.de>
13690 * tree-pass.h (PROP_loop_opts_done): New.
13691 * gimple-range-path.cc (path_range_query::internal_range_of_expr):
13692 Intersect with global range.
13693 * tree-ssa-loop.c (tree_ssa_loop_done): Set PROP_loop_opts_done.
13694 * tree-ssa-threadbackward.c
13695 (back_threader_profitability::profitable_path_p): Disable
13696 threading through latches until after loop optimizations have run.
13698 2021-09-10 David Faust <david.faust@oracle.com>
13700 * doc/invoke.texi: Document BPF -mcpu, -mjmpext, -mjmp32 and -malu32
13703 2021-09-10 David Faust <david.faust@oracle.com>
13705 * config/bpf/bpf-opts.h (bpf_isa_version): New enum.
13706 * config/bpf/bpf-protos.h (bpf_expand_cbranch): New.
13707 * config/bpf/bpf.c (bpf_option_override): Handle -mcpu option.
13708 (bpf_expand_cbranch): New function.
13709 * config/bpf/bpf.md (AM mode iterator): Conditionalize support for SI
13711 (zero_extendsidi2): Only use mov32 instruction if it is available.
13712 (SIM mode iterator): Conditionalize support for SI mode.
13713 (JM mode iterator): New.
13714 (cbranchdi4): Update name, use new JM iterator. Use bpf_expand_cbranch.
13715 (*branch_on_di): Update name, use new JM iterator.
13716 * config/bpf/bpf.opt: (mjmpext): New option.
13717 (malu32): Likewise.
13718 (mjmp32): Likewise.
13720 (bpf_isa): New enum.
13722 2021-09-10 David Faust <david.faust@oracle.com>
13724 * config/bpf/bpf.md (zero_extendhidi2): Add new output template
13725 for register-to-register extensions.
13726 (zero_extendqidi2): Likewise.
13728 2021-09-10 Richard Biener <rguenther@suse.de>
13730 PR middle-end/102273
13731 * internal-fn.c (expand_DEFERRED_INIT): Always expand non-SSA vars.
13733 2021-09-10 Richard Biener <rguenther@suse.de>
13735 PR middle-end/102269
13736 * gimplify.c (is_var_need_auto_init): Empty types do not need
13739 2021-09-10 Richard Biener <rguenther@suse.de>
13741 * configure.ac (--with-stabs): Remove.
13742 * configure: Regenerate.
13743 * doc/install.texi: Remove --with-stabs documentation.
13745 2021-09-10 liuhongt <hongtao.liu@intel.com>
13747 * config/i386/avx512fp16intrin.h: (_mm512_cmp_ph_mask):
13749 (_mm512_mask_cmp_ph_mask): Likewise.
13750 (_mm512_cmp_round_ph_mask): Likewise.
13751 (_mm512_mask_cmp_round_ph_mask): Likewise.
13752 (_mm_cmp_sh_mask): Likewise.
13753 (_mm_mask_cmp_sh_mask): Likewise.
13754 (_mm_cmp_round_sh_mask): Likewise.
13755 (_mm_mask_cmp_round_sh_mask): Likewise.
13756 (_mm_comieq_sh): Likewise.
13757 (_mm_comilt_sh): Likewise.
13758 (_mm_comile_sh): Likewise.
13759 (_mm_comigt_sh): Likewise.
13760 (_mm_comige_sh): Likewise.
13761 (_mm_comineq_sh): Likewise.
13762 (_mm_ucomieq_sh): Likewise.
13763 (_mm_ucomilt_sh): Likewise.
13764 (_mm_ucomile_sh): Likewise.
13765 (_mm_ucomigt_sh): Likewise.
13766 (_mm_ucomige_sh): Likewise.
13767 (_mm_ucomineq_sh): Likewise.
13768 (_mm_comi_round_sh): Likewise.
13769 (_mm_comi_sh): Likewise.
13770 * config/i386/avx512fp16vlintrin.h (_mm_cmp_ph_mask): New intrinsic.
13771 (_mm_mask_cmp_ph_mask): Likewise.
13772 (_mm256_cmp_ph_mask): Likewise.
13773 (_mm256_mask_cmp_ph_mask): Likewise.
13774 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
13775 * config/i386/i386-builtin.def: Add corresponding new builtins.
13776 * config/i386/i386-expand.c
13777 (ix86_expand_args_builtin): Handle new builtin types.
13778 (ix86_expand_round_builtin): Ditto.
13779 * config/i386/i386.md (ssevecmode): Add HF mode.
13780 (MODEFH): New mode iterator.
13781 * config/i386/sse.md
13782 (V48H_AVX512VL): New mode iterator to support HF vector modes.
13783 Ajdust corresponding description.
13784 (ssecmpintprefix): New.
13785 (VI12_AVX512VL): Adjust to support HF vector modes.
13786 (cmp_imm_predicate): Likewise.
13787 (<avx512>_cmp<mode>3<mask_scalar_merge_name><round_saeonly_name>):
13789 (avx512f_vmcmp<mode>3<round_saeonly_name>): Likewise.
13790 (avx512f_vmcmp<mode>3_mask<round_saeonly_name>): Likewise.
13791 (<sse>_<unord>comi<round_saeonly_name>): Likewise.
13793 2021-09-10 liuhongt <hongtao.liu@intel.com>
13795 * config/i386/avx512fp16intrin.h: (_mm512_max_ph): New intrinsic.
13796 (_mm512_mask_max_ph): Likewise.
13797 (_mm512_maskz_max_ph): Likewise.
13798 (_mm512_min_ph): Likewise.
13799 (_mm512_mask_min_ph): Likewise.
13800 (_mm512_maskz_min_ph): Likewise.
13801 (_mm512_max_round_ph): Likewise.
13802 (_mm512_mask_max_round_ph): Likewise.
13803 (_mm512_maskz_max_round_ph): Likewise.
13804 (_mm512_min_round_ph): Likewise.
13805 (_mm512_mask_min_round_ph): Likewise.
13806 (_mm512_maskz_min_round_ph): Likewise.
13807 (_mm_max_sh): Likewise.
13808 (_mm_mask_max_sh): Likewise.
13809 (_mm_maskz_max_sh): Likewise.
13810 (_mm_min_sh): Likewise.
13811 (_mm_mask_min_sh): Likewise.
13812 (_mm_maskz_min_sh): Likewise.
13813 (_mm_max_round_sh): Likewise.
13814 (_mm_mask_max_round_sh): Likewise.
13815 (_mm_maskz_max_round_sh): Likewise.
13816 (_mm_min_round_sh): Likewise.
13817 (_mm_mask_min_round_sh): Likewise.
13818 (_mm_maskz_min_round_sh): Likewise.
13819 * config/i386/avx512fp16vlintrin.h (_mm_max_ph): New intrinsic.
13820 (_mm256_max_ph): Likewise.
13821 (_mm_mask_max_ph): Likewise.
13822 (_mm256_mask_max_ph): Likewise.
13823 (_mm_maskz_max_ph): Likewise.
13824 (_mm256_maskz_max_ph): Likewise.
13825 (_mm_min_ph): Likewise.
13826 (_mm256_min_ph): Likewise.
13827 (_mm_mask_min_ph): Likewise.
13828 (_mm256_mask_min_ph): Likewise.
13829 (_mm_maskz_min_ph): Likewise.
13830 (_mm256_maskz_min_ph): Likewise.
13831 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
13832 * config/i386/i386-builtin.def: Add corresponding new builtins.
13833 * config/i386/i386-expand.c
13834 (ix86_expand_args_builtin): Handle new builtin types.
13835 * config/i386/sse.md
13836 (<code><mode>3<mask_name><round_saeonly_name>): Adjust to
13837 support HF vector modes.
13838 (*<code><mode>3<mask_name><round_saeonly_name>): Likewise.
13839 (ieee_<ieee_maxmin><mode>3<mask_name><round_saeonly_name>):
13841 (<sse>_vm<code><mode>3<mask_scalar_name><round_saeonly_scalar_name>):
13843 * config/i386/subst.md (round_saeonly_mode512bit_condition):
13844 Adjust for HF vector modes.
13846 2021-09-10 Liu, Hongtao <hongtao.liu@intel.com>
13848 * config/i386/avx512fp16intrin.h (_mm_add_sh): New intrinsic.
13849 (_mm_mask_add_sh): Likewise.
13850 (_mm_maskz_add_sh): Likewise.
13851 (_mm_sub_sh): Likewise.
13852 (_mm_mask_sub_sh): Likewise.
13853 (_mm_maskz_sub_sh): Likewise.
13854 (_mm_mul_sh): Likewise.
13855 (_mm_mask_mul_sh): Likewise.
13856 (_mm_maskz_mul_sh): Likewise.
13857 (_mm_div_sh): Likewise.
13858 (_mm_mask_div_sh): Likewise.
13859 (_mm_maskz_div_sh): Likewise.
13860 (_mm_add_round_sh): Likewise.
13861 (_mm_mask_add_round_sh): Likewise.
13862 (_mm_maskz_add_round_sh): Likewise.
13863 (_mm_sub_round_sh): Likewise.
13864 (_mm_mask_sub_round_sh): Likewise.
13865 (_mm_maskz_sub_round_sh): Likewise.
13866 (_mm_mul_round_sh): Likewise.
13867 (_mm_mask_mul_round_sh): Likewise.
13868 (_mm_maskz_mul_round_sh): Likewise.
13869 (_mm_div_round_sh): Likewise.
13870 (_mm_mask_div_round_sh): Likewise.
13871 (_mm_maskz_div_round_sh): Likewise.
13872 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
13873 * config/i386/i386-builtin.def: Add corresponding new builtins.
13874 * config/i386/i386-expand.c
13875 (ix86_expand_round_builtin): Handle new builtins.
13876 * config/i386/sse.md (VF_128): Change description.
13877 (<sse>_vm<plusminus_insn><mode>3<mask_scalar_name><round_scalar_name>):
13878 Adjust to support HF vector modes.
13879 (<sse>_vm<multdiv_mnemonic><mode>3<mask_scalar_name><round_scalar_name>):
13882 2021-09-10 H.J. Lu <hjl.tools@gmail.com>
13884 * config/i386/i386-expand.c
13885 (ix86_avx256_split_vector_move_misalign): Handle V16HF mode.
13886 * config/i386/i386.c
13887 (ix86_preferred_simd_mode): Handle HF mode.
13888 * config/i386/sse.md (V_256H): New mode iterator.
13889 (avx_vextractf128<mode>): Use it.
13890 (VEC_INIT_MODE): Align vector HFmode condition to vector
13891 HImodes since there're no real HF instruction used.
13892 (VEC_INIT_HALF_MODE): Ditto.
13894 (VIHF_AVX512BW): Ditto.
13895 (*vec_extracthf): Ditto.
13896 (VEC_EXTRACT_MODE): Ditto.
13898 2021-09-10 Richard Biener <rguenther@suse.de>
13901 * config/dbx.h: Remove.
13902 * config/dbxcoff.h: Do not define PREFERRED_DEBUGGING_TYPE.
13903 * config/lynx.h: Likewise.
13905 2021-09-10 liuhongt <hongtao.liu@intel.com>
13907 * config/i386/i386-expand.c (ix86_expand_copysign): Expand
13908 right into ANDNOT + AND + IOR, using paradoxical subregs.
13909 (ix86_split_copysign_const): Remove.
13910 (ix86_split_copysign_var): Ditto.
13911 * config/i386/i386-protos.h (ix86_split_copysign_const): Dotto.
13912 (ix86_split_copysign_var): Ditto.
13913 * config/i386/i386.md (@copysign<mode>3_const): Ditto.
13914 (@copysign<mode>3_var): Ditto.
13916 2021-09-09 qing zhao <qing.zhao@oracle.com>
13918 * builtins.c (expand_builtin_memset): Make external visible.
13919 * builtins.h (expand_builtin_memset): Declare extern.
13920 * common.opt (ftrivial-auto-var-init=): New option.
13921 * doc/extend.texi: Document the uninitialized attribute.
13922 * doc/invoke.texi: Document -ftrivial-auto-var-init.
13923 * flag-types.h (enum auto_init_type): New enumerated type
13925 * gimple-fold.c (clear_padding_type): Add one new parameter.
13926 (clear_padding_union): Likewise.
13927 (clear_padding_emit_loop): Likewise.
13928 (clear_type_padding_in_mask): Likewise.
13929 (gimple_fold_builtin_clear_padding): Handle this new parameter.
13930 * gimplify.c (gimple_add_init_for_auto_var): New function.
13931 (gimple_add_padding_init_for_auto_var): New function.
13932 (is_var_need_auto_init): New function.
13933 (gimplify_decl_expr): Add initialization to automatic variables per
13935 (gimplify_call_expr): Add one new parameter for call to
13936 __builtin_clear_padding.
13937 (gimplify_init_constructor): Add padding initialization in the end.
13938 * internal-fn.c (INIT_PATTERN_VALUE): New macro.
13939 (expand_DEFERRED_INIT): New function.
13940 * internal-fn.def (DEFERRED_INIT): New internal function.
13941 * tree-cfg.c (verify_gimple_call): Verify calls to .DEFERRED_INIT.
13942 * tree-sra.c (generate_subtree_deferred_init): New function.
13943 (scan_function): Avoid setting cannot_scalarize_away_bitmap for
13944 calls to .DEFERRED_INIT.
13945 (sra_modify_deferred_init): New function.
13946 (sra_modify_function_body): Handle calls to DEFERRED_INIT specially.
13947 * tree-ssa-structalias.c (find_func_aliases_for_call): Likewise.
13948 * tree-ssa-uninit.c (warn_uninit): Handle calls to DEFERRED_INIT
13950 (check_defs): Likewise.
13951 (warn_uninitialized_vars): Likewise.
13952 * tree-ssa.c (ssa_undefined_value_p): Likewise.
13953 * tree.c (build_common_builtin_nodes): Build tree node for
13954 BUILT_IN_CLEAR_PADDING when needed.
13956 2021-09-09 Richard Biener <rguenther@suse.de>
13958 * tree-ssa-loop-im.c (fill_always_executed_in_1): Walk
13961 2021-09-09 Richard Biener <rguenther@suse.de>
13963 * tree-ssa-loop-im.c (fill_always_executed_in_1): Integrate
13964 DOM walk from get_loop_body_in_dom_order using a worklist
13967 2021-09-09 liuhongt <hongtao.liu@intel.com>
13969 * config.gcc: Add avx512fp16vlintrin.h.
13970 * config/i386/avx512fp16intrin.h: (_mm512_add_ph): New intrinsic.
13971 (_mm512_mask_add_ph): Likewise.
13972 (_mm512_maskz_add_ph): Likewise.
13973 (_mm512_sub_ph): Likewise.
13974 (_mm512_mask_sub_ph): Likewise.
13975 (_mm512_maskz_sub_ph): Likewise.
13976 (_mm512_mul_ph): Likewise.
13977 (_mm512_mask_mul_ph): Likewise.
13978 (_mm512_maskz_mul_ph): Likewise.
13979 (_mm512_div_ph): Likewise.
13980 (_mm512_mask_div_ph): Likewise.
13981 (_mm512_maskz_div_ph): Likewise.
13982 (_mm512_add_round_ph): Likewise.
13983 (_mm512_mask_add_round_ph): Likewise.
13984 (_mm512_maskz_add_round_ph): Likewise.
13985 (_mm512_sub_round_ph): Likewise.
13986 (_mm512_mask_sub_round_ph): Likewise.
13987 (_mm512_maskz_sub_round_ph): Likewise.
13988 (_mm512_mul_round_ph): Likewise.
13989 (_mm512_mask_mul_round_ph): Likewise.
13990 (_mm512_maskz_mul_round_ph): Likewise.
13991 (_mm512_div_round_ph): Likewise.
13992 (_mm512_mask_div_round_ph): Likewise.
13993 (_mm512_maskz_div_round_ph): Likewise.
13994 * config/i386/avx512fp16vlintrin.h: New header.
13995 * config/i386/i386-builtin-types.def (V16HF, V8HF, V32HF):
13996 Add new builtin types.
13997 * config/i386/i386-builtin.def: Add corresponding builtins.
13998 * config/i386/i386-expand.c
13999 (ix86_expand_args_builtin): Handle new builtin types.
14000 (ix86_expand_round_builtin): Likewise.
14001 * config/i386/immintrin.h: Include avx512fp16vlintrin.h
14002 * config/i386/sse.md (VFH): New mode_iterator.
14004 (avx512fmaskmode): Add HF vector modes.
14005 (avx512fmaskhalfmode): Likewise.
14006 (<plusminus_insn><mode>3<mask_name><round_name>): Adjust to for
14008 (*<plusminus_insn><mode>3<mask_name><round_name>): Likewise.
14009 (mul<mode>3<mask_name><round_name>): Likewise.
14010 (*mul<mode>3<mask_name><round_name>): Likewise.
14011 (div<mode>3): Likewise.
14012 (<sse>_div<mode>3<mask_name><round_name>): Likewise.
14013 * config/i386/subst.md (SUBST_V): Add HF vector modes.
14014 (SUBST_A): Likewise.
14015 (round_mode512bit_condition): Adjust for V32HFmode.
14017 2021-09-09 liuhongt <hongtao.liu@intel.com>
14020 * config/i386/sse.md (reduc_plus_scal_<mode>): Split to ..
14021 (reduc_plus_scal_v4sf): .. this, New define_expand.
14022 (reduc_plus_scal_v2df): .. and this, New define_expand.
14024 2021-09-09 liuhongt <hongtao.liu@intel.com>
14027 * config/i386/sse.md (*vec_extract<mode><ssescalarmodelower>_valign):
14030 2021-09-08 Jonathan Wakely <jwakely@redhat.com>
14033 * doc/trouble.texi (Copy Assignment): Fix description of
14034 behaviour and fix code in example.
14036 2021-09-08 Segher Boessenkool <segher@kernel.crashing.org>
14039 * config/rs6000/rs6000-logue.c (rs6000_emit_epilogue): For ELFv2 use
14040 r11 instead of r12 for restoring CR.
14042 2021-09-08 Jakub Jelinek <jakub@redhat.com>
14043 liuhongt <hongtao.liu@intel.com>
14046 * config/i386/i386.md (@xorsign<mode>3_1): Remove.
14047 * config/i386/i386-expand.c (ix86_expand_xorsign): Expand right away
14048 into AND with mask and XOR, using paradoxical subregs.
14049 (ix86_split_xorsign): Remove.
14050 * config/i386/i386-protos.h (ix86_split_xorsign): Remove.
14052 2021-09-08 Di Zhao <dizhao@os.amperecomputing.com>
14054 * tree-ssa-sccvn.c (vn_nary_op_insert_into): fix result compare
14056 2021-09-08 Jakub Jelinek <jakub@redhat.com>
14059 * config/i386/i386.md (xorsign<mode>3): If operands[1] is equal to
14060 operands[2], emit abs<mode>2 instead.
14061 (@xorsign<mode>3_1): Add early-clobbers for output operand, enable
14062 first alternative even for avx, add another alternative with
14063 =&Yv <- 0, Yv, Yvm constraints.
14064 * config/i386/i386-expand.c (ix86_split_xorsign): If op0 is equal
14065 to op1, emit vpandn instead.
14067 2021-09-08 liuhongt <hongtao.liu@intel.com>
14069 * config/i386/avx512fp16intrin.h (_mm_set_ph): New intrinsic.
14070 (_mm256_set_ph): Likewise.
14071 (_mm512_set_ph): Likewise.
14072 (_mm_setr_ph): Likewise.
14073 (_mm256_setr_ph): Likewise.
14074 (_mm512_setr_ph): Likewise.
14075 (_mm_set1_ph): Likewise.
14076 (_mm256_set1_ph): Likewise.
14077 (_mm512_set1_ph): Likewise.
14078 (_mm_setzero_ph): Likewise.
14079 (_mm256_setzero_ph): Likewise.
14080 (_mm512_setzero_ph): Likewise.
14081 (_mm_set_sh): Likewise.
14082 (_mm_load_sh): Likewise.
14083 (_mm_store_sh): Likewise.
14084 * config/i386/i386-builtin-types.def (V8HF): New type.
14085 (DEF_FUNCTION_TYPE (V8HF, V8HI)): New builtin function type
14086 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
14087 Support vector HFmodes.
14088 (ix86_expand_vector_init_one_nonzero): Likewise.
14089 (ix86_expand_vector_init_one_var): Likewise.
14090 (ix86_expand_vector_init_interleave): Likewise.
14091 (ix86_expand_vector_init_general): Likewise.
14092 (ix86_expand_vector_set): Likewise.
14093 (ix86_expand_vector_extract): Likewise.
14094 (ix86_expand_vector_init_concat): Likewise.
14095 (ix86_expand_sse_movcc): Handle vector HFmodes.
14096 (ix86_expand_vector_set_var): Ditto.
14097 * config/i386/i386-modes.def: Add HF vector modes in comment.
14098 * config/i386/i386.c (classify_argument): Add HF vector modes.
14099 (ix86_hard_regno_mode_ok): Allow HF vector modes for AVX512FP16.
14100 (ix86_vector_mode_supported_p): Likewise.
14101 (ix86_set_reg_reg_cost): Handle vector HFmode.
14102 (ix86_get_ssemov): Handle vector HFmode.
14103 (function_arg_advance_64): Pass unamed V16HFmode and V32HFmode
14105 (function_arg_advance_32): Pass V8HF/V16HF/V32HF by sse reg for 32bit
14107 (function_arg_advance_32): Ditto.
14108 * config/i386/i386.h (VALID_AVX512FP16_REG_MODE): New.
14109 (VALID_AVX256_REG_OR_OI_MODE): Rename to ..
14110 (VALID_AVX256_REG_OR_OI_VHF_MODE): .. this, and add V16HF.
14111 (VALID_SSE2_REG_VHF_MODE): New.
14112 (VALID_AVX512VL_128_REG_MODE): Add V8HF and TImode.
14113 (SSE_REG_MODE_P): Add vector HFmode.
14114 * config/i386/i386.md (mode): Add HF vector modes.
14115 (MODE_SIZE): Likewise.
14116 (ssemodesuffix): Add ph suffix for HF vector modes.
14117 * config/i386/sse.md (VFH_128): New mode iterator.
14118 (VMOVE): Adjust for HF vector modes.
14120 (V_256_512): Likewise.
14121 (avx512): Likewise.
14122 (avx512fmaskmode): Likewise.
14123 (shuffletype): Likewise.
14124 (sseinsnmode): Likewise.
14125 (ssedoublevecmode): Likewise.
14126 (ssehalfvecmode): Likewise.
14127 (ssehalfvecmodelower): Likewise.
14128 (ssePScmode): Likewise.
14129 (ssescalarmode): Likewise.
14130 (ssescalarmodelower): Likewise.
14131 (sseintprefix): Likewise.
14133 (bcstscalarsuff): Likewise.
14134 (xtg_mode): Likewise.
14135 (VI12HF_AVX512VL): New mode_iterator.
14136 (VF_AVX512FP16): Likewise.
14138 (VIHF_256): Likewise.
14139 (VIHF_AVX512BW): Likewise.
14140 (V16_256): Likewise.
14141 (V32_512): Likewise.
14142 (sseintmodesuffix): New mode_attr.
14143 (sse): Add scalar and vector HFmodes.
14144 (ssescalarmode): Add vector HFmode mapping.
14145 (ssescalarmodesuffix): Add sh suffix for HFmode.
14146 (*<sse>_vm<insn><mode>3): Use VFH_128.
14147 (*<sse>_vm<multdiv_mnemonic><mode>3): Likewise.
14148 (*ieee_<ieee_maxmin><mode>3): Likewise.
14149 (<avx512>_blendm<mode>): New define_insn.
14150 (vec_setv8hf): New define_expand.
14151 (vec_set<mode>_0): New define_insn for HF vector set.
14152 (*avx512fp16_movsh): Likewise.
14153 (avx512fp16_movsh): Likewise.
14154 (vec_extract_lo_v32hi): Rename to ...
14155 (vec_extract_lo_<mode>): ... this, and adjust to allow HF
14157 (vec_extract_hi_v32hi): Likewise.
14158 (vec_extract_hi_<mode>): Likewise.
14159 (vec_extract_lo_v16hi): Likewise.
14160 (vec_extract_lo_<mode>): Likewise.
14161 (vec_extract_hi_v16hi): Likewise.
14162 (vec_extract_hi_<mode>): Likewise.
14163 (vec_set_hi_v16hi): Likewise.
14164 (vec_set_hi_<mode>): Likewise.
14165 (vec_set_lo_v16hi): Likewise.
14166 (vec_set_lo_<mode>): Likewise.
14167 (*vec_extract<mode>_0): New define_insn_and_split for HF
14169 (*vec_extracthf): New define_insn.
14170 (VEC_EXTRACT_MODE): Add HF vector modes.
14171 (PINSR_MODE): Add V8HF.
14172 (sse2p4_1): Likewise.
14173 (pinsr_evex_isa): Likewise.
14174 (<sse2p4_1>_pinsr<ssemodesuffix>): Adjust to support
14175 insert for V8HFmode.
14176 (pbroadcast_evex_isa): Add HF vector modes.
14177 (AVX2_VEC_DUP_MODE): Likewise.
14178 (VEC_INIT_MODE): Likewise.
14179 (VEC_INIT_HALF_MODE): Likewise.
14180 (avx2_pbroadcast<mode>): Adjust to support HF vector mode
14182 (avx2_pbroadcast<mode>_1): Likewise.
14183 (<avx512>_vec_dup<mode>_1): Likewise.
14184 (<avx512>_vec_dup<mode><mask_name>): Likewise.
14185 (<mask_codefor><avx512>_vec_dup_gpr<mode><mask_name>):
14188 2021-09-08 Guo, Xuepeng <xuepeng.guo@intel.com>
14189 H.J. Lu <hongjiu.lu@intel.com>
14190 Liu Hongtao <hongtao.liu@intel.com>
14191 Wang Hongyu <hongyu.wang@intel.com>
14192 Xu Dianhong <dianhong.xu@intel.com>
14194 * common/config/i386/cpuinfo.h (get_available_features):
14195 Detect FEATURE_AVX512FP16.
14196 * common/config/i386/i386-common.c
14197 (OPTION_MASK_ISA_AVX512FP16_SET,
14198 OPTION_MASK_ISA_AVX512FP16_UNSET,
14199 OPTION_MASK_ISA2_AVX512FP16_SET,
14200 OPTION_MASK_ISA2_AVX512FP16_UNSET): New.
14201 (OPTION_MASK_ISA2_AVX512BW_UNSET,
14202 OPTION_MASK_ISA2_AVX512BF16_UNSET): Add AVX512FP16.
14203 (ix86_handle_option): Handle -mavx512fp16.
14204 * common/config/i386/i386-cpuinfo.h (enum processor_features):
14205 Add FEATURE_AVX512FP16.
14206 * common/config/i386/i386-isas.h: Add entry for AVX512FP16.
14207 * config.gcc: Add avx512fp16intrin.h.
14208 * config/i386/avx512fp16intrin.h: New intrinsic header.
14209 * config/i386/cpuid.h: Add bit_AVX512FP16.
14210 * config/i386/i386-builtin-types.def: (FLOAT16): New primitive type.
14211 * config/i386/i386-builtins.c: Support _Float16 type for i386
14213 (ix86_register_float16_builtin_type): New function.
14214 (ix86_float16_type_node): New.
14215 * config/i386/i386-c.c (ix86_target_macros_internal): Define
14217 * config/i386/i386-expand.c (ix86_expand_branch): Support
14219 (ix86_prepare_fp_compare_args): Adjust TARGET_SSE_MATH &&
14220 SSE_FLOAT_MODE_P to SSE_FLOAT_MODE_SSEMATH_OR_HF_P.
14221 (ix86_expand_fp_movcc): Ditto.
14222 * config/i386/i386-isa.def: Add PTA define for AVX512FP16.
14223 * config/i386/i386-options.c (isa2_opts): Add -mavx512fp16.
14224 (ix86_valid_target_attribute_inner_p): Add avx512fp16 attribute.
14225 * config/i386/i386.c (ix86_get_ssemov): Use
14226 vmovdqu16/vmovw/vmovsh for HFmode/HImode scalar or vector.
14227 (ix86_get_excess_precision): Use
14228 FLT_EVAL_METHOD_PROMOTE_TO_FLOAT16 when TARGET_AVX512FP16
14230 (sse_store_index): Use SFmode cost for HFmode cost.
14231 (inline_memory_move_cost): Add HFmode, and perfer SSE cost over
14232 GPR cost for HFmode.
14233 (ix86_hard_regno_mode_ok): Allow HImode in sse register.
14234 (ix86_mangle_type): Add manlging for _Float16 type.
14235 (inline_secondary_memory_needed): No memory is needed for
14236 16bit movement between gpr and sse reg under
14238 (ix86_multiplication_cost): Adjust TARGET_SSE_MATH &&
14239 SSE_FLOAT_MODE_P to SSE_FLOAT_MODE_SSEMATH_OR_HF_P.
14240 (ix86_division_cost): Ditto.
14241 (ix86_rtx_costs): Ditto.
14242 (ix86_add_stmt_cost): Ditto.
14243 (ix86_optab_supported_p): Ditto.
14244 * config/i386/i386.h (VALID_AVX512F_SCALAR_MODE): Add HFmode.
14245 (SSE_FLOAT_MODE_SSEMATH_OR_HF_P): Add HFmode.
14246 (PTA_SAPPHIRERAPIDS): Add PTA_AVX512FP16.
14247 * config/i386/i386.md (mode): Add HFmode.
14248 (MODE_SIZE): Add HFmode.
14249 (isa): Add avx512fp16.
14250 (enabled): Handle avx512fp16.
14251 (ssemodesuffix): Add sh suffix for HFmode.
14252 (comm): Add mult, div.
14253 (plusminusmultdiv): New code iterator.
14254 (insn): Add mult, div.
14255 (*movhf_internal): Adjust for avx512fp16 instruction.
14256 (*movhi_internal): Ditto.
14257 (*cmpi<unord>hf): New define_insn for HFmode.
14258 (*ieee_s<ieee_maxmin>hf3): Likewise.
14259 (extendhf<mode>2): Likewise.
14260 (trunc<mode>hf2): Likewise.
14261 (float<floatunssuffix><mode>hf2): Likewise.
14262 (*<insn>hf): Likewise.
14263 (cbranchhf4): New expander.
14264 (movhfcc): Likewise.
14265 (<insn>hf3): Likewise.
14266 (mulhf3): Likewise.
14267 (divhf3): Likewise.
14268 * config/i386/i386.opt: Add mavx512fp16.
14269 * config/i386/immintrin.h: Include avx512fp16intrin.h.
14270 * doc/invoke.texi: Add mavx512fp16.
14271 * doc/extend.texi: Add avx512fp16 Usage Notes.
14273 2021-09-08 liuhongt <hongtao.liu@intel.com>
14275 * common.opt: Support -fexcess-precision=16.
14276 * config/aarch64/aarch64.c (aarch64_excess_precision): Return
14277 FLT_EVAL_METHOD_PROMOTE_TO_FLOAT16 when
14278 EXCESS_PRECISION_TYPE_FLOAT16.
14279 * config/arm/arm.c (arm_excess_precision): Ditto.
14280 * config/i386/i386.c (ix86_get_excess_precision): Ditto.
14281 * config/m68k/m68k.c (m68k_excess_precision): Issue an error
14282 when EXCESS_PRECISION_TYPE_FLOAT16.
14283 * config/s390/s390.c (s390_excess_precision): Ditto.
14284 * coretypes.h (enum excess_precision_type): Add
14285 EXCESS_PRECISION_TYPE_FLOAT16.
14286 * doc/tm.texi (TARGET_C_EXCESS_PRECISION): Update documents.
14287 * doc/tm.texi.in (TARGET_C_EXCESS_PRECISION): Ditto.
14288 * doc/extend.texi (Half-Precision): Document
14289 -fexcess-precision=16.
14290 * flag-types.h (enum excess_precision): Add
14291 EXCESS_PRECISION_FLOAT16.
14292 * target.def (excess_precision): Update document.
14293 * tree.c (excess_precision_type): Set excess_precision_type to
14294 EXCESS_PRECISION_FLOAT16 when -fexcess-precision=16.
14296 2021-09-08 liuhongt <hongtao.liu@intel.com>
14298 * doc/extend.texi: (@node Floating Types): Adjust the wording.
14299 (@node Half-Precision): Ditto.
14301 2021-09-07 Takayuki 'January June' Suwa <jjsuwa_sys3175@yahoo.co.jp>
14304 * config/xtensa/xtensa.c (xtensa_emit_move_sequence): Add
14305 'CONST_INT_P (src)' to the condition of the block that tries to
14306 eliminate literal when loading integer contant.
14308 2021-09-07 David Faust <david.faust@oracle.com>
14310 * doc/extend.texi (BPF Type Attributes) New node.
14311 Document new preserve_access_index attribute.
14312 Document new preserve_access_index builtin.
14313 * doc/invoke.texi: Document -mco-re and -mno-co-re options.
14315 2021-09-07 David Faust <david.faust@oracle.com>
14317 * config/bpf/bpf.c: Adjust includes.
14318 (bpf_handle_preserve_access_index_attribute): New function.
14319 (bpf_attribute_table): Use it here.
14320 (bpf_builtins): Add BPF_BUILTIN_PRESERVE_ACCESS_INDEX.
14321 (bpf_option_override): Handle "-mco-re" option.
14322 (bpf_asm_init_sections): New.
14323 (TARGET_ASM_INIT_SECTIONS): Redefine.
14324 (bpf_file_end): New.
14325 (TARGET_ASM_FILE_END): Redefine.
14326 (bpf_init_builtins): Add "__builtin_preserve_access_index".
14327 (bpf_core_compute, bpf_core_get_index): New.
14328 (is_attr_preserve_access): New.
14329 (bpf_expand_builtin): Handle new builtins.
14330 (bpf_core_newdecl, bpf_core_is_maybe_aggregate_access): New.
14331 (bpf_core_walk): New.
14332 (bpf_resolve_overloaded_builtin): New.
14333 (TARGET_RESOLVE_OVERLOADED_BUILTIN): Redefine.
14334 (handle_attr): New.
14335 (pass_bpf_core_attr): New RTL pass.
14336 * config/bpf/bpf-passes.def: New file.
14337 * config/bpf/bpf-protos.h (make_pass_bpf_core_attr): New.
14338 * config/bpf/coreout.c: New file.
14339 * config/bpf/coreout.h: Likewise.
14340 * config/bpf/t-bpf (TM_H): Add $(srcdir)/config/bpf/coreout.h.
14341 (coreout.o): New rule.
14342 (PASSES_EXTRA): Add $(srcdir)/config/bpf/bpf-passes.def.
14343 * config.gcc (bpf): Add coreout.h to extra_headers.
14344 Add coreout.o to extra_objs.
14345 Add $(srcdir)/config/bpf/coreout.c to target_gtfiles.
14347 2021-09-07 David Faust <david.faust@oracle.com>
14349 * btfout.c (get_btf_id): Function is no longer static.
14350 * ctfc.h: Expose it here.
14352 2021-09-07 David Faust <david.faust@oracle.com>
14354 * ctfc.c (ctf_lookup_tree_type): New function.
14355 * ctfc.h: Likewise.
14357 2021-09-07 David Faust <david.faust@oracle.com>
14359 * ctfc.c (ctf_dtd_lookup): Function is no longer static.
14360 * ctfc.h: Analogous change.
14362 2021-09-07 David Faust <david.faust@oracle.com>
14364 * dwarf2out.c (lookup_type_die): Function is no longer static.
14365 * dwarf2out.h: Expose it here.
14367 2021-09-07 Indu Bhagat <indu.bhagat@oracle.com>
14369 * dwarf2ctf.c (ctf_debug_finalize): Make it static.
14370 (ctf_debug_early_finish): New definition.
14371 (ctf_debug_finish): Likewise.
14372 * dwarf2ctf.h (ctf_debug_finalize): Remove declaration.
14373 (ctf_debug_early_finish): New declaration.
14374 (ctf_debug_finish): Likewise.
14375 * dwarf2out.c (dwarf2out_finish): Invoke ctf_debug_finish.
14376 (dwarf2out_early_finish): Invoke ctf_debug_early_finish.
14378 2021-09-07 Indu Bhagat <indu.bhagat@oracle.com>
14380 * config/bpf/bpf.c (bpf_option_override): For BPF backend, disable LTO
14381 support when compiling for CO-RE.
14382 * config/bpf/bpf.opt: Add new command line option -mco-re.
14384 2021-09-07 Indu Bhagat <indu.bhagat@oracle.com>
14386 * flag-types.h (enum debug_info_type): Add new enum
14387 DINFO_TYPE_BTF_WITH_CORE.
14388 (BTF_WITH_CORE_DEBUG): New bitmask.
14389 * flags.h (btf_with_core_debuginfo_p): New declaration.
14390 * opts.c (btf_with_core_debuginfo_p): New definition.
14392 2021-09-07 Jason Merrill <jason@redhat.com>
14394 * tree.h (error_operand_p): Change to inline function.
14396 2021-09-07 Aldy Hernandez <aldyh@redhat.com>
14398 * tree-ssa-threadedge.c (forwarder_block_p): Rename to...
14399 (empty_block_with_phis_p): ...this.
14400 (potentially_threadable_block): Same.
14401 (jump_threader::thread_through_normal_block): Same.
14403 2021-09-07 Eric Botcazou <ebotcazou@adacore.com>
14406 * dwarf2out.c (mark_base_types): New overloaded function.
14407 (dwarf2out_early_finish): Invoke it on the COMDAT type list as well
14408 as the compilation unit, and call move_marked_base_types afterward.
14410 2021-09-07 H.J. Lu <hjl.tools@gmail.com>
14413 * config/i386/i386-expand.c (ix86_expand_convert_uns_sisf_sse):
14415 (ix86_expand_vector_convert_uns_vsivsf): Likewise.
14417 2021-09-07 Richard Biener <rguenther@suse.de>
14419 PR tree-optimization/102226
14420 * tree-vect-loop.c (vect_transform_cycle_phi): Record
14421 the converted value for the epilogue PHI use.
14423 2021-09-07 Martin Liska <mliska@suse.cz>
14425 PR gcov-profile/80223
14426 * ipa-inline.c (can_inline_edge_p): Similarly to sanitizer
14427 options, do not inline when no_profile_instrument_function
14428 attributes are different in early inliner. It's fine to inline
14429 it after PGO instrumentation.
14431 2021-09-07 Richard Biener <rguenther@suse.de>
14433 PR tree-optimization/101555
14434 * tree-ssa-pre.c (translate_vuse_through_block): Do not
14435 perform an alias walk to determine the validity of the
14436 mem at the start of the block which is already guaranteed
14437 by means of prune_clobbered_mems.
14438 (phi_translate_1): Pass edge to translate_vuse_through_block.
14440 2021-09-07 Xionghu Luo <luoxhu@linux.ibm.com>
14443 * config/rs6000/rs6000.md (fmod<mode>3): New define_expand.
14444 (remainder<mode>3): Likewise.
14446 2021-09-07 YunQiang Su <yunqiang.su@cipunited.com>
14448 * config/mips/mips.c (mips_file_start): add .module for
14451 2021-09-06 Roger Sayle <roger@nextmovesoftware.com>
14453 * wide-int.cc (wi::clz): Reorder tests to ensure the result
14454 is zero for all negative values.
14456 2021-09-06 Tobias Burnus <tobias@codesourcery.com>
14458 * doc/invoke.texi (-foffload-options): Fix @opindex.
14460 2021-09-06 H.J. Lu <hjl.tools@gmail.com>
14463 * config/i386/i386-expand.c (ix86_split_xorsign): Use operands[2].
14464 * config/i386/i386.md (@xorsign<mode>3_1): Add non-destructive
14465 source alternative for AVX.
14467 2021-09-06 liuhongt <hongtao.liu@intel.com>
14469 PR middle-end/102182
14470 * optabs.c (expand_fix): Add from1 to avoid from being
14473 2021-09-06 Eric Botcazou <ebotcazou@adacore.com>
14475 * dwarf2out.c (modified_type_die): Deal with all array types earlier
14476 and use local variable consistently throughout the function.
14478 2021-09-06 Jakub Jelinek <jakub@redhat.com>
14480 PR tree-optimization/102207
14481 * match.pd: Don't demote operands of IFN_{ADD,SUB,MUL}_OVERFLOW if they
14482 were promoted from signed to wider unsigned type.
14484 2021-09-06 Andrew Pinski <apinski@marvell.com>
14486 PR tree-optimization/63184
14487 * match.pd: Add simplification of pointer_diff of two pointer_plus
14488 with addr_expr in the first operand of each pointer_plus.
14489 Add simplificatoin of ne/eq of two pointer_plus with addr_expr
14490 in the first operand of each pointer_plus.
14492 2021-09-06 Richard Biener <rguenther@suse.de>
14494 PR tree-optimization/102176
14495 * tree-vect-slp.c (vect_slp_gather_vectorized_scalar_stmts):
14497 (vect_bb_slp_scalar_cost): Use the computed set of
14498 vectorized scalar stmts instead of relying on the out-of-date
14499 and not accurate PURE_SLP_STMT.
14500 (vect_bb_vectorization_profitable_p): Compute the set
14501 of vectorized scalar stmts.
14503 2021-09-05 Aldy Hernandez <aldyh@redhat.com>
14505 * gimple-range-path.cc (path_range_query::range_of_stmt): Remove
14506 GIMPLE_COND special casing.
14507 (path_range_query::range_defined_in_block): Use range_of_stmt
14508 instead of calling fold_range directly.
14510 2021-09-05 Aldy Hernandez <aldyh@redhat.com>
14512 * gimple-range-path.cc (path_range_query::range_of_expr): Set
14513 m_undefined_path when appropriate.
14514 (path_range_query::internal_range_of_expr): Copy from range_of_expr.
14515 (path_range_query::unreachable_path_p): New.
14516 (path_range_query::precompute_ranges): Set m_undefined_path.
14517 * gimple-range-path.h (path_range_query::unreachable_path_p): New.
14518 (path_range_query::internal_range_of_expr): New.
14519 * tree-ssa-threadbackward.c (back_threader::find_taken_edge_cond):
14520 Use unreachable_path_p.
14522 2021-09-05 Aldy Hernandez <aldyh@redhat.com>
14524 * tree-ssa-threadbackward.c (back_threader::maybe_register_path):
14525 Remove argument and call find_taken_edge.
14526 (back_threader::resolve_phi): Do not calculate taken edge before
14527 calling maybe_register_path.
14528 (back_threader::find_paths_to_names): Same.
14530 2021-09-05 Jeff Law <jlaw@localhost.localdomain>
14532 * config/h8300/h8300.md (QHSI2 mode iterator): New mode iterator.
14533 * config/h8300/testcompare.md (store_c): Update name, use new
14535 (store_neg_c, store_shifted_c): New patterns.
14537 2021-09-03 Segher Boessenkool <segher@kernel.crashing.org>
14540 * config/rs6000/rs6000-logue.c (rs6000_emit_prologue): On ELFv2 use r11
14541 instead of r12 for CR save, in all cases.
14543 2021-09-03 Andrew Pinski <apinski@marvell.com>
14545 * config/aarch64/aarch64-sve-builtins.cc (register_vector_type):
14546 Handle error_mark_node as the type of the type_decl.
14548 2021-09-03 Andrew Pinski <apinski@marvell.com>
14550 * config/aarch64/aarch64-builtins.c (struct aarch64_simd_type_info):
14552 (aarch64_simd_types): Likewise.
14553 (aarch64_simd_intOI_type_node): Likewise.
14554 (aarch64_simd_intCI_type_node): Likewise.
14555 (aarch64_simd_intXI_type_node): Likewise.
14556 * config/aarch64/aarch64.h (aarch64_fp16_type_node): Likewise.
14557 (aarch64_fp16_ptr_type_node): Likewise.
14558 (aarch64_bf16_type_node): Likewise.
14559 (aarch64_bf16_ptr_type_node): Likewise.
14561 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
14563 * range-op.cc (operator_minus::op1_op2_relation_effect): Abstract
14565 (minus_op1_op2_relation_effect): ...here.
14566 (class operator_pointer_diff): New.
14567 (operator_pointer_diff::op1_op2_relation_effect): Call
14568 minus_op1_op2_relation_effect.
14569 (integral_table::integral_table): Add entry for POINTER_DIFF_EXPR.
14571 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
14573 * tree-ssa-threadbackward.c (back_threader::thread_through_all_blocks):
14574 Add may_peel_loop_headers.
14575 (back_threader_registry::thread_through_all_blocks): Same.
14576 (try_thread_blocks): Pass may_peel_loop_headers argument.
14577 (pass_early_thread_jumps::execute): Same.
14579 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
14581 * tree-ssa-threadedge.c (has_phis_p): New.
14582 (forwarder_block_p): New.
14583 (potentially_threadable_block): Call forwarder_block_p.
14584 (jump_threader::thread_around_empty_blocks): Call has_phis_p.
14585 (jump_threader::thread_through_normal_block): Call
14588 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
14590 * tree-ssa-threadbackward.c (back_threader::dump): New.
14591 (back_threader::debug): New.
14592 (back_threader_profitability::profitable_path_p): Dump blocks
14593 even if we are bailing early.
14595 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
14597 * tree-ssa-threadupdate.c (cancel_thread): New.
14598 (jump_thread_path_registry::thread_block_1): Use cancel_thread.
14599 (jump_thread_path_registry::mark_threaded_blocks): Same.
14600 (jump_thread_path_registry::register_jump_thread): Same.
14602 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
14604 * tree-ssa-threadedge.c (jt_state::push): Only call methods for
14605 which objects are available.
14606 (jt_state::pop): Same.
14607 (jt_state::register_equiv): Same.
14608 (jt_state::register_equivs_on_edge): Same.
14610 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
14612 * tree-ssa-threadedge.c (jump_threader::thread_across_edge):
14613 Move pop until after a thread is registered.
14615 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
14617 * tree-ssa-threadupdate.c (debug): New.
14619 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
14621 * gimple-range-trace.cc (push_dump_file::push_dump_file): New.
14622 (push_dump_file::~push_dump_file): New.
14623 (dump_ranger): Change dump_file temporarily while dumping
14625 * gimple-range-trace.h (class push_dump_file): New.
14627 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
14629 * gimple-range-trace.cc (debug_seed_ranger): Remove static.
14630 (dump_ranger): Dump function name.
14632 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
14634 * gimple-range-path.cc (path_range_query::range_defined_in_block):
14635 Adjust for non-null.
14636 (path_range_query::adjust_for_non_null_uses): New.
14637 (path_range_query::precompute_ranges): Call
14638 adjust_for_non_null_uses.
14639 * gimple-range-path.h: Add m_non_null and
14640 adjust_for_non_null_uses.
14642 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
14644 * gimple-range-path.cc (path_range_query::dump): Dump path
14646 (path_range_query::precompute_ranges): Dump entire path.
14648 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
14650 * value-relation.cc (relation_oracle::debug): New.
14651 * value-relation.h (relation_oracle::debug): New.
14653 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
14655 * tree-ssa-loop-ch.c: Remove unnecessary include file.
14657 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
14659 * gimple-range-fold.cc (fold_using_range::postfold_gcond_edges):
14660 Skip statements with no defining BB.
14661 * gimple-range-path.cc (path_range_query::range_defined_in_block):
14662 Do not get confused by statements with no defining BB.
14664 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
14666 * gimple-range-fold.cc (adjust_imagpart_expr): Move from
14667 gimple_range_adjustment. Add support for constants.
14668 (adjust_realpart_expr): New.
14669 (gimple_range_adjustment): Move IMAGPART_EXPR code to
14670 adjust_imagpart_expr.
14671 * range-op.cc (integral_table::integral_table): Add entry for
14674 2021-09-03 Jakub Jelinek <jakub@redhat.com>
14676 * omp-expand.c (expand_omp_atomic_pipeline): Use
14677 IFN_ATOMIC_COMPARE_EXCHANGE instead of
14678 BUILT_IN_SYNC_VAL_COMPARE_AND_SWAP_? so that memory order
14681 2021-09-03 Jakub Jelinek <jakub@redhat.com>
14684 * tree.h (DECL_FIELD_ABI_IGNORED): Changed into rvalue only macro
14685 that is false if DECL_BIT_FIELD.
14686 (SET_DECL_FIELD_ABI_IGNORED, DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD,
14687 SET_DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD): Define.
14688 * tree-streamer-out.c (pack_ts_decl_common_value_fields): For
14689 DECL_BIT_FIELD stream DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD instead
14690 of DECL_FIELD_ABI_IGNORED.
14691 * tree-streamer-in.c (unpack_ts_decl_common_value_fields): Use
14692 SET_DECL_FIELD_ABI_IGNORED instead of writing to
14693 DECL_FIELD_ABI_IGNORED and for DECL_BIT_FIELD use
14694 SET_DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD instead.
14695 * lto-streamer-out.c (hash_tree): For DECL_BIT_FIELD hash
14696 DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD instead of DECL_FIELD_ABI_IGNORED.
14698 2021-09-03 liuhongt <hongtao.liu@intel.com>
14701 * config/i386/amxbf16intrin.h : Remove macro check for __AMX_BF16__.
14702 * config/i386/amxint8intrin.h : Remove macro check for __AMX_INT8__.
14703 * config/i386/amxtileintrin.h : Remove macro check for __AMX_TILE__.
14705 2021-09-02 Martin Sebor <msebor@redhat.com>
14707 PR tree-optimization/17506
14709 * tree-ssa-uninit.c (warn_uninit): Remove conditional guarding note.
14711 2021-09-02 Richard Biener <rguenther@suse.de>
14713 * tree-ssa-loop-im.c (fill_always_executed_in_1): Refine
14714 fix for PR78185 and continue processing when leaving
14715 finite inner loops.
14717 2021-09-02 Jakub Jelinek <jakub@redhat.com>
14719 PR tree-optimization/99591
14720 * match.pd: Demote operands of IFN_{ADD,SUB,MUL}_OVERFLOW if they
14723 2021-09-02 Richard Biener <rguenther@suse.de>
14726 2021-09-02 Richard Biener <rguenther@suse.de>
14728 PR tree-optimization/102155
14729 * tree-ssa-loop-im.c (fill_always_executed_in_1): Iterate
14730 over a part of the RPO array and do not recurse here.
14731 Dump blocks marked as always executed.
14732 (fill_always_executed_in): Walk over the RPO array and
14733 process loops whose header we run into.
14734 (loop_invariant_motion_in_fun): Compute the first RPO
14735 using rev_post_order_and_mark_dfs_back_seme in iteration
14736 order and pass that to fill_always_executed_in.
14738 2021-09-02 liuhongt <hongtao.liu@intel.com>
14740 * config/i386/i386-modes.def (FLOAT_MODE): Define ieee HFmode.
14741 * config/i386/i386.c (enum x86_64_reg_class): Add
14742 X86_64_SSEHF_CLASS.
14743 (merge_classes): Handle X86_64_SSEHF_CLASS.
14744 (examine_argument): Ditto.
14745 (construct_container): Ditto.
14746 (classify_argument): Ditto, and set HFmode/HCmode to
14747 X86_64_SSEHF_CLASS.
14748 (function_value_32): Return _FLoat16/Complex Float16 by
14750 (function_value_64): Return _Float16/Complex Float16 by SSE
14752 (ix86_print_operand): Handle CONST_DOUBLE HFmode.
14753 (ix86_secondary_reload): Require gpr as intermediate register
14754 to store _Float16 from sse register when sse4 is not
14756 (ix86_libgcc_floating_mode_supported_p): Enable _FLoat16 under
14758 (ix86_scalar_mode_supported_p): Ditto.
14759 (TARGET_LIBGCC_FLOATING_MODE_SUPPORTED_P): Defined.
14760 * config/i386/i386.h (VALID_SSE2_REG_MODE): Add HFmode.
14761 (VALID_INT_MODE_P): Add HFmode and HCmode.
14762 * config/i386/i386.md (*pushhf_rex64): New define_insn.
14764 (*movhf_internal): Ditto.
14765 * doc/extend.texi (Half-Precision Floating Point): Documemt
14768 2021-09-02 Richard Biener <rguenther@suse.de>
14770 PR tree-optimization/102155
14771 * tree-ssa-loop-im.c (fill_always_executed_in_1): Iterate
14772 over a part of the RPO array and do not recurse here.
14773 Dump blocks marked as always executed.
14774 (fill_always_executed_in): Walk over the RPO array and
14775 process loops whose header we run into.
14776 (loop_invariant_motion_in_fun): Compute the first RPO
14777 using rev_post_order_and_mark_dfs_back_seme in iteration
14778 order and pass that to fill_always_executed_in.
14780 2021-09-02 YunQiang Su <syq@debian.org>
14783 2021-08-31 YunQiang Su <yunqiang.su@cipunited.com>
14785 * config/mips/mips.c (mips_module_isa_name): New.
14786 mips_file_start: add .module mipsREV to all asm output
14788 2021-09-01 Jeff Law <jlaw@localhost.localdomain>
14790 PR tree-optimization/102152
14791 * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Reduce a vector
14792 comparison to a scalar comparison before calling
14793 update_stmt_if_modified.
14795 2021-09-01 Andrew Pinski <apinski@marvell.com>
14798 * config/aarch64/aarch64.c (aarch64_expand_setmem):
14799 Check STRICT_ALIGNMENT before creating an overlapping
14802 2021-09-01 Martin Sebor <msebor@redhat.com>
14804 * gimple-ssa-warn-access.cc (get_size_range): Add argument.
14805 (check_access): Pass additional argument.
14806 (check_memop_access): Remove template and make a member function.
14807 (maybe_check_dealloc_call): Make a pass_waccess member function.
14808 (class pass_waccess): Add, rename, and remove members.
14809 (pass_waccess::pass_waccess): Adjust to name change.
14810 (pass_waccess::~pass_waccess): Same.
14811 (check_alloca): Make a member function.
14812 (check_alloc_size_call): Same.
14813 (check_strcat): Same.
14814 (check_strncat): Same.
14815 (check_stxcpy): Same.
14816 (check_stxncpy): Same.
14817 (check_strncmp): Same.
14818 (maybe_warn_rdwr_sizes): Rename...
14819 (pass_waccess::maybe_check_access_sizes): ...to this.
14820 (pass_waccess::check_call): Adjust to name changes.
14821 (pass_waccess::maybe_check_dealloc_call): Make a pass_waccess member
14823 (pass_waccess::execute): Adjust to name changes.
14824 * gimple-ssa-warn-access.h (check_memop_access): Remove.
14825 * pointer-query.cc (access_ref::phi): Handle null pointer.
14826 (access_ref::inform_access): Same.
14827 (pointer_query::put_ref): Modify a cached value, not a copy of it.
14828 (pointer_query::dump): New function.
14829 (compute_objsize_r): Avoid overwriting access_ref::bndrng. Cache
14831 * pointer-query.h (pointer_query::dump): Declare.
14832 * tree-ssa-strlen.c (get_range): Simplify. Use function query.
14833 (dump_strlen_info): Use function query.
14834 (printf_strlen_execute): Factor code out into pointer_query::put_ref.
14836 2021-09-01 Thomas Schwinge <thomas@codesourcery.com>
14838 * tree.c (walk_tree_1) <OMP_CLAUSE>: Simplify.
14840 2021-09-01 Iain Sandoe <iain@sandoe.co.uk>
14842 * doc/extend.texi: Document unavailable attribute.
14843 * print-tree.c (print_node): Handle unavailable attribute.
14844 * tree-core.h (struct tree_base): Add a bit to carry unavailability.
14845 * tree.c (error_unavailable_use): New.
14846 * tree.h (TREE_UNAVAILABLE): New.
14847 (error_unavailable_use): New.
14849 2021-09-01 Jakub Jelinek <jakub@redhat.com>
14851 PR tree-optimization/102124
14852 * tree-vect-patterns.c (vect_recog_widen_op_pattern): For ORIG_CODE
14853 MINUS_EXPR, if itype is unsigned with smaller precision than type,
14854 add an extra cast to signed variant of itype to ensure sign-extension.
14856 2021-09-01 Martin Liska <mliska@suse.cz>
14858 * graph.c (draw_cfg_node_succ_edges): Do not color fallthru
14859 edges and rather use colors for TRUE and FALSE edges.
14861 2021-09-01 Richard Biener <rguenther@suse.de>
14863 PR tree-optimization/93491
14864 * tree-ssa-pre.c (compute_avail): Set BB_MAY_NOTRETURN
14865 after processing the stmt itself. Do not consider
14866 pure functions possibly not returning. Properly avoid
14867 adding possibly trapping calls to EXP_GEN when there's
14868 a preceeding possibly not returning call.
14869 * tree-ssa-sccvn.c (vn_reference_may_trap): Conservatively
14872 2021-09-01 Richard Biener <rguenther@suse.de>
14874 PR tree-optimization/102139
14875 * tree-vectorizer.h (vec_base_alignments): Adjust hash-map
14876 type to record a std::pair of the stmt-info and the innermost
14878 (dr_vec_info::group): New member.
14879 * tree-vect-data-refs.c (vect_record_base_alignment): Adjust.
14880 (vect_compute_data_ref_alignment): Verify the recorded
14881 base alignment can be used.
14882 (data_ref_pair): Remove.
14883 (dr_group_sort_cmp): Adjust.
14884 (vect_analyze_data_ref_accesses): Store the group-ID in the
14885 dr_vec_info and operate on a vector of dr_vec_infos.
14887 2021-09-01 YunQiang Su <yunqiang.su@cipunited.com>
14889 * read-md.c (md_reader::handle_enum): support value assignation.
14890 * doc/md.texi: record define_c_enum value assignation support.
14892 2021-09-01 Jakub Jelinek <jakub@redhat.com>
14894 PR tree-optimization/102141
14895 * gimple-ssa-store-merging.c (bswap_view_convert): Add BEFORE
14896 argument. If false, emit stmts after gsi instead of before, and
14898 (bswap_replace): Adjust callers. When converting output of bswap,
14899 emit VIEW_CONVERT prepratation stmts after a copy of gsi instead
14902 2021-09-01 liuhongt <hongtao.liu@intel.com>
14904 * emit-rtl.c (validate_subreg): Get rid of all float-int
14907 2021-09-01 liuhongt <hongtao.liu@intel.com>
14910 2021-08-30 liuhongt <hongtao.liu@intel.com>
14912 * expmed.c (extract_bit_field_1): Make sure we're playing with
14913 integral modes before call extract_integral_bit_field.
14914 (extract_integral_bit_field): Add a parameter of type
14915 scalar_int_mode which corresponds to of tmode.
14916 And call extract_and_convert_fixed_bit_field instead of
14917 extract_fixed_bit_field and convert_extracted_bit_field.
14918 (extract_and_convert_fixed_bit_field): New function, it's a
14919 combination of extract_fixed_bit_field and
14920 convert_extracted_bit_field.
14922 2021-08-31 Thomas Schwinge <thomas@codesourcery.com>
14924 * tree.c (walk_tree_1) <OMP_CLAUSE_TILE>: Handle three operands.
14926 2021-08-31 Thomas Schwinge <thomas@codesourcery.com>
14928 * omp-general.h (omp_is_reference): Rename to...
14929 (omp_privatize_by_reference): ... this. Adjust all users...
14930 * omp-general.c: ... here, ...
14931 * gimplify.c: ... here, ...
14932 * omp-expand.c: ... here, ...
14933 * omp-low.c: ... here.
14935 2021-08-31 Martin Sebor <msebor@redhat.com>
14937 * gimple-ssa-warn-access.cc (maybe_warn_alloc_args_overflow): Test
14938 pointer element for equality to zero, not that of the cotaining
14941 2021-08-31 Martin Sebor <msebor@redhat.com>
14943 * gcc-rich-location.h (gcc_rich_location): Make ctor explicit.
14945 2021-08-31 Martin Sebor <msebor@redhat.com>
14947 * function.h (function): Add comments.
14948 (get_range_query): Same. Add attribute returns nonnull.
14950 2021-08-31 Roger Sayle <roger@nextmovesoftware.com>
14952 * expr.c (convert_modes): Don't use subreg_promoted_mode on a
14953 SUBREG if it can't be guaranteed to a SUBREG_PROMOTED_VAR_P set.
14954 Instead use the standard (safer) is_a <scalar_int_mode> idiom.
14956 2021-08-31 Jeff Law <jlaw@localhost.localdomain>
14958 * config.gcc (cris-*-elf, cris-*-none): Remove dbxelf.h from
14960 (m32r-*-elf, m32rle-*-elf, m32r-*-linux): Likewise.
14961 (mn10300-*-*, am33_2.0-*-linux*): Likewise.
14962 (xtensa*-*-elf, xtensa*-*-linux, xtensa*-*-uclinux): Likewise.
14963 (m32c-*-elf*, m32c-*-rtems*): Likewise.
14964 * config/cris/cris.h (DBX_NO_XREFS): Remove.
14965 (DBX_CONTIN_LENGTH, DBX_CONTIN_CHAR): Likewise.
14966 * config/m32r/m32r.h (DBXOUT_SOURCE_LINE): Likewise.
14967 (DBX_DEBUGGING_INFO, DBX_CONTIN_LENGTH): Likewise.
14968 * config/mn10300/mn10300.h (DEFAULT_GDB_EXTENSIONS): Likewise.
14969 * config/mn10300/linux.h (DBX_REGISTER_NAMES): Likewise.
14971 2021-08-31 Marcel Vollweiler <marcel@codesourcery.com>
14973 * gimplify.c (gimplify_scan_omp_clauses): Error handling. 'ancestor' only
14974 allowed on target constructs and only with particular other clauses.
14975 * omp-expand.c (expand_omp_target): Output of 'sorry, not supported' if
14976 'ancestor' is used.
14977 * omp-low.c (check_omp_nesting_restrictions): Error handling. No nested OpenMP
14978 structs when 'ancestor' is used.
14979 (scan_omp_1_stmt): No usage of OpenMP runtime routines in a target region when
14980 'ancestor' is used.
14981 * tree-pretty-print.c (dump_omp_clause): Append 'ancestor'.
14982 * tree.h (OMP_CLAUSE_DEVICE_ANCESTOR): Define macro.
14984 2021-08-31 Roger Sayle <roger@nextmovesoftware.com>
14986 * expr.c (convert_modes): Preserve SUBREG_PROMOTED_VAR_P when
14987 creating a (wider) partial subreg from a SUBREG_PROMOTED_VAR_P
14989 * simplify-rtx.c (simplify_unary_operation_1) [SIGN_EXTEND]:
14990 Likewise, preserve SUBREG_PROMOTED_VAR_P when creating a (wider)
14991 partial subreg from a SUBREG_PROMOTED_VAR_P subreg. Generate
14992 SIGN_EXTEND of the SUBREG_REG when a subreg would be paradoxical.
14993 [ZERO_EXTEND]: Likewise, preserve SUBREG_PROMOTED_VAR_P when
14994 creating a (wider) partial subreg from a SUBREG_PROMOTED_VAR_P
14995 subreg. Generate ZERO_EXTEND of the SUBREG_REG when a subreg
14996 would be paradoxical.
14998 2021-08-31 Roger Sayle <roger@nextmovesoftware.com>
15000 * combine.c (combine_simplify_rtx): Avoid converting an explicit
15001 TRUNCATE into a lowpart SUBREG on !TRULY_NOOP_TRUNCATION targets.
15002 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
15004 2021-08-31 Richard Biener <rguenther@suse.de>
15006 PR tree-optimization/102142
15007 * tree-vect-slp.c (vect_bb_vectorization_profitable_p): Fix
15008 condition under which to unset the visited flag.
15010 2021-08-31 Richard Biener <rguenther@suse.de>
15012 PR middle-end/102129
15013 * tree-ssa-ter.c (find_replaceable_in_bb): Do not move
15014 possibly trapping expressions across calls.
15016 2021-08-31 Jakub Jelinek <jakub@redhat.com>
15018 PR tree-optimization/102134
15019 * tree-ssa-ccp.c (bit_value_binop) <case RSHIFT_EXPR>: If sgn is
15020 UNSIGNED and r1val | r1mask has MSB set, ensure lzcount doesn't
15023 2021-08-31 Andrew Pinski <apinski@marvell.com>
15026 * collect-utils.c (setup_signals): New declaration.
15027 * collect-utils.h (setup_signals): New function.
15028 * collect2.c (handler): Delete.
15029 (main): Instead of manually setting up the signals,
15030 just call setup_signals.
15031 * lto-wrapper.c (main): Likewise.
15033 2021-08-31 Andrew Pinski <apinski@marvell.com>
15036 * config/i386/i386-protos.h (x86_output_aligned_bss):
15037 Change align argument to unsigned type.
15038 (x86_elf_aligned_decl_common): Likewise.
15039 * config/i386/i386.c (x86_elf_aligned_decl_common): Likewise.
15040 (x86_output_aligned_bss): Likewise.
15042 2021-08-31 YunQiang Su <yunqiang.su@cipunited.com>
15044 * config/mips/mips.c (mips_module_isa_name): New.
15045 mips_file_start: add .module mipsREV to all asm output
15047 2021-08-31 YunQiang Su <yunqiang.su@cipunited.com>
15049 * config/mips/mips.h (struct mips_cpu_info): define enum mips_isa;
15050 use enum instead of int for 'isa' member.
15051 * config.gcc, config/mips/mips.c, config/mips/mips-cpus.def,
15052 config/mips/netbsd.h: replace hardcoded numbers with enum.
15054 2021-08-31 liuhongt <hongtao.liu@intel.com>
15056 * config/i386/sse.md (*<avx512>_ucmp<mode>3_1): Change from
15057 define_split to define_insn_and_split.
15058 (*avx2_eq<mode>3): Removed.
15059 (<avx512>_eq<mode>3<mask_scalar_merge_name>): Adjust pattern
15060 (<avx512>_eq<mode>3<mask_scalar_merge_name>_1): Rename to ..
15061 (*<avx512>_eq<mode>3<mask_scalar_merge_name>_1): .. this, and
15063 (*avx2_gt<mode>3): Removed.
15064 (<avx512>_gt<mode>3<mask_scalar_merge_name>): Change from
15065 define_insn to define_expand, and adjust pattern.
15066 (UNSPEC_MASKED_EQ, UNSPEC_MASKED_GT): Removed.
15068 2021-08-30 David Malcolm <dmalcolm@redhat.com>
15071 * Makefile.in (ANALYZER_OBJS): Add analyzer/call-info.o.
15073 2021-08-30 Jason Merrill <jason@redhat.com>
15075 * doc/invoke.texi: Document -Wmissing-requires.
15077 2021-08-30 Bill Schmidt <wschmidt@linux.ibm.com>
15079 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Remove
15080 TARGET_EXTRA_BUILTINS guard.
15082 2021-08-30 Bill Schmidt <wschmidt@linux.ibm.com>
15084 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Change
15085 initialization of V2DI_type_node and unsigned_V2DI_type_node.
15087 2021-08-30 Bill Schmidt <wschmidt@linux.ibm.com>
15089 * config/rs6000/darwin.h (SUBTARGET_INIT_BUILTINS): Use the new
15090 decl when new_builtins_are_live.
15091 * config/rs6000/rs6000-builtin-new.def (__builtin_cfstring): New
15094 2021-08-30 Pat Haugen <pthaugen@linux.ibm.com>
15096 * config/rs6000/rs6000-cpus.def (ISA_3_1_MASKS_SERVER): Add
15097 OPTION_MASK_P10_FUSION_2STORE.
15098 (POWERPC_MASKS): Likewise.
15099 * config/rs6000/rs6000.c (rs6000_option_override_internal): Enable
15100 store fusion for Power10.
15101 (is_fusable_store): New.
15102 (power10_sched_reorder): Likewise.
15103 (rs6000_sched_reorder): Do Power10 specific reordering.
15104 (rs6000_sched_reorder2): Likewise.
15105 * config/rs6000/rs6000.opt: Add new option.
15107 2021-08-30 Richard Biener <rguenther@suse.de>
15109 PR tree-optimization/102128
15110 * tree-vect-slp.c (vect_bb_vectorization_profitable_p):
15111 Move scanning for if-converted scalar code to the caller
15112 and instead delay clearing the visited flag for profitable
15114 (vect_slp_region): Cost all subgraphs before scheduling.
15115 For if-converted BB vectorization scan for scalar COND_EXPRs
15116 and do not vectorize if any found and the cost model is
15119 2021-08-30 Richard Biener <rguenther@suse.de>
15121 * common.opt (fexceptions): Mark
15122 EnabledBy(fnon-call-exceptions).
15123 * doc/invoke.texi (fnon-call-exceptions): Document this
15124 enables -fexceptions.
15126 2021-08-30 Sebastian Huber <sebastian.huber@embedded-brains.de>
15128 * tsystem.h (abort): Define abort() if inhibit_libc is defined and it
15129 is not already defined.
15131 2021-08-30 liuhongt <hongtao.liu@intel.com>
15133 * expmed.c (extract_bit_field_1): Make sure we're playing with
15134 integral modes before call extract_integral_bit_field.
15135 (extract_integral_bit_field): Add a parameter of type
15136 scalar_int_mode which corresponds to of tmode.
15137 And call extract_and_convert_fixed_bit_field instead of
15138 extract_fixed_bit_field and convert_extracted_bit_field.
15139 (extract_and_convert_fixed_bit_field): New function, it's a
15140 combination of extract_fixed_bit_field and
15141 convert_extracted_bit_field.
15143 2021-08-29 Iain Sandoe <iain@sandoe.co.uk>
15145 * config/darwin.c (darwin_libc_has_function): Do not run
15146 the checks for x86 or modern Darwin. Make sure that there
15147 is a value set for darwin_macosx_version_min before testing.
15149 2021-08-29 Iain Sandoe <iain@sandoe.co.uk>
15151 * config/i386/darwin.h (CLEAR_INSN_CACHE): New.
15153 2021-08-28 Jan Hubicka <hubicka@ucw.cz>
15155 * ipa-modref-tree.h (modref_access_node::merge): Break out
15156 logic combining offsets and logic merging ranges to ...
15157 (modref_access_node::combined_offsets): ... here
15158 (modref_access_node::update2): ... here
15159 (modref_access_node::closer_pair_p): New member function.
15160 (modref_access_node::forced_merge): New member function.
15161 (modre_ref_node::insert): Do merging when table is full.
15163 2021-08-28 YunQiang Su <yunqiang.su@cipunited.com>
15166 * config.gcc: MIPS: use N64 ABI by default if the triple end
15167 with -gnuabi64, which is used by Debian since 2013.
15169 2021-08-28 Alexandre Oliva <oliva@adacore.com>
15171 * ipa-modref.c (analyze_function): Skip debug stmts.
15172 * tree-inline.c (estimate_num_insn): Consider builtins even
15173 without a cgraph_node.
15175 2021-08-27 Jeff Law <jlaw@localhost.localdomain>
15177 * config/h8300/bitfield.md (cstore<mode>4): Remove expander.
15178 * config/h8300/h8300.c (h8300_expand_branch): Remove function.
15179 * config/h8300/h8300-protos.h (h8300_expadn_branch): Remove prototype.
15180 * config/h8300/h8300.md (eqne): New code iterator.
15181 (geultu, geultu_to_c): Similarly.
15182 * config/h8300/testcompare.md (cstore<mode>4): Dummy expander.
15183 (store_c_<mode>, store_c_i_<mode>): New define_insn_and_splits
15184 (cmp<mode>_c): New pattern
15186 2021-08-27 Jeff Law <jlaw@localhost.localdomain>
15188 * tree-ssa-dom.c (reduce_vector_comparison_to_scalar_comparison): New
15190 (dom_opt_dom_walker::optimize_stmt): Use it.
15192 2021-08-27 Iain Sandoe <iain@sandoe.co.uk>
15194 * config/darwin.c (finalize_ctors): Add a section-start linker-
15196 (finalize_dtors): Likewise.
15197 * config/darwin.h (MIN_LD64_INIT_TERM_START_LABELS): New.
15199 2021-08-27 Bill Schmidt <wschmidt@linux.ibm.com>
15201 * config/rs6000/rs6000-call.c (rs6000-builtins.h): New #include.
15202 (rs6000_init_builtins): Call rs6000_init_generated_builtins. Skip the
15203 old initialization logic when new builtins are enabled.
15204 * config/rs6000/rs6000-gen-builtins.c (write_decls): Rename
15205 rs6000_autoinit_builtins to rs6000_init_generated_builtins.
15206 (write_init_file): Likewise.
15208 2021-08-27 Iain Sandoe <iain@sandoe.co.uk>
15210 * configure.ac (darwin2[[0-9]]* | darwin19*): Alter use of
15211 gcc_GAS_CHECK_FEATURE to remove an extraneous parameter.
15212 (amdgcn-* | gcn-*) Likewise.
15214 2021-08-27 Anthony Sharp <anthonysharp15@gmail.com>
15216 * symbol-summary.h: Added missing template keyword.
15218 2021-08-27 Richard Biener <rguenther@suse.de>
15220 PR tree-optimization/45178
15221 * tree-ssa-dce.c (find_obviously_necessary_stmts): For
15222 infinite loops without exit do not mark control dependent
15223 edges of the latch necessary.
15225 2021-08-27 konglin1 <lingling.kong@intel.com>
15228 * config/i386/sse.md: (<avx512>scattersi<mode>): Add mask operand to
15230 (<avx512>scattersi<mode>): Likewise.
15231 (*avx512f_scattersi<VI48F:mode>): Merge mask operand to set_dest.
15232 (*avx512f_scatterdi<VI48F:mode>): Likewise
15234 2021-08-27 Kewen Lin <linkw@linux.ibm.com>
15236 * config/rs6000/rs6000.c (rs6000_builtin_md_vectorized_function): Add
15237 support for built-in functions MISC_BUILTIN_DIVWE, MISC_BUILTIN_DIVWEU,
15238 MISC_BUILTIN_DIVDE, MISC_BUILTIN_DIVDEU, P10_BUILTIN_CFUGED,
15239 P10_BUILTIN_CNTLZDM, P10_BUILTIN_CNTTZDM, P10_BUILTIN_PDEPD and
15240 P10_BUILTIN_PEXTD on Power10.
15242 2021-08-27 Kewen Lin <linkw@linux.ibm.com>
15244 * config/rs6000/rs6000-call.c (builtin_function_type): Add unsigned
15245 signedness for some Power10 bifs.
15247 2021-08-27 David Edelsohn <dje.gcc@gmail.com>
15250 * config/rs6000/rs6000.c (rs6000_adjust_field_align): Use
15251 computed alignment if the entire struct has attribute packed.
15253 2021-08-27 liuhongt <hongtao.liu@intel.com>
15257 * config/i386/i386.c (ix86_gimple_fold_builtin): Fold
15258 IX86_BUILTIN_SHUFPD512, IX86_BUILTIN_SHUFPS512,
15259 IX86_BUILTIN_SHUFPD256, IX86_BUILTIN_SHUFPS,
15260 IX86_BUILTIN_SHUFPS256.
15261 (ix86_masked_all_ones): New function.
15263 2021-08-26 Uroš Bizjak <ubizjak@gmail.com>
15265 * config/i386/i386.md (*btr<mode>_1): Call force_reg unconditionally.
15266 (conditional moves with memory inputs splitters): Ditto.
15267 * config/i386/sse.md (one_cmpl<mode>2): Simplify.
15269 2021-08-26 Jan Hubicka <hubicka@ucw.cz>
15271 * ipa-modref-tree.h (modref_access_node::try_merge_with): Restart
15272 search after merging.
15274 2021-08-26 Bill Schmidt <wschmidt@linux.ibm.com>
15276 * config/rs6000/rs6000-overload.def: Add remaining overloads.
15278 2021-08-26 Bill Schmidt <wschmidt@linux.ibm.com>
15280 * config/rs6000/rs6000-builtin-new.def: Add cell stanza.
15282 2021-08-26 Bill Schmidt <wschmidt@linux.ibm.com>
15284 * config/rs6000/rs6000-builtin-new.def: Add ieee128-hw, dfp,
15285 crypto, and htm stanzas.
15287 2021-08-26 Bill Schmidt <wschmidt@linux.ibm.com>
15289 * config/rs6000/rs6000-builtin-new.def: Add mma stanza.
15291 2021-08-26 Martin Sebor <msebor@redhat.com>
15293 * tree-ssa-uninit.c (warn_uninit): Refactor and simplify.
15294 (warn_uninit_phi_uses): Remove argument from calls to warn_uninit.
15295 (warn_uninitialized_vars): Same. Reduce visibility of locals.
15296 (warn_uninitialized_phi): Same.
15298 2021-08-26 Roger Sayle <roger@nextmovesoftware.com>
15300 * tree-ssa-ccp.c (get_individual_bits): Helper function to
15301 extract the individual bits from a widest_int constant (mask).
15302 (gray_code_bit_flips): New read-only table for effiently
15303 enumerating permutations/combinations of bits.
15304 (bit_value_binop) [LROTATE_EXPR, RROTATE_EXPR]: Handle rotates
15305 by unknown counts that are guaranteed less than the target
15306 precision and four or fewer unknown bits by enumeration.
15307 [LSHIFT_EXPR, RSHIFT_EXPR]: Likewise, also handle shifts by
15308 enumeration under the same conditions. Handle remaining
15309 shifts as a mask based upon the minimum possible shift value.
15311 2021-08-26 Roger Sayle <roger@nextmovesoftware.com>
15312 Richard Biener <rguenther@suse.de>
15314 * match.pd (shift transformations): Remove a redundant
15315 !POINTER_TYPE_P check.
15317 2021-08-26 Uroš Bizjak <ubizjak@gmail.com>
15320 * config/i386/i386.md (cmove reg-reg move elimination peephole2s):
15321 Set all_regs to true in the call to replace_rtx.
15323 2021-08-26 Jan Hubicka <hubicka@ucw.cz>
15325 * ipa-modref-tree.c (test_insert_search_collapse): Update test.
15326 * ipa-modref-tree.h (modref_base_node::insert): Be smarter when
15327 hiting --param modref-max-refs limit.
15328 (modref_tree:insert_base): Be smarter when hitting
15329 --param modref-max-bases limit. Add new parameter REF.
15330 (modref_tree:insert): Update.
15331 (modref_tree:merge): Update.
15332 * ipa-modref.c (read_modref_records): Update.
15334 2021-08-26 Jan Hubicka <hubicka@ucw.cz>
15336 * params.opt: (modref-max-adjustments): Add full stop.
15338 2021-08-26 Jan Hubicka <hubicka@ucw.cz>
15340 * ipa-modref-tree.h (modref_ref_node::verify): New member
15342 (modref_ref_node::insert): Use it.
15343 (modref_ref_node::try_mere_with): Fix off by one error.
15345 2021-08-26 Martin Liska <mliska@suse.cz>
15346 Stefan Kneifel <stefan.kneifel@bluewin.ch>
15348 * cgraph.h (create_version_clone_with_body): Add new parameter.
15349 * cgraphclones.c: Likewise.
15350 * multiple_target.c (create_dispatcher_calls): Do not use
15352 (create_target_clone): Likewise here.
15354 2021-08-26 Jonathan Yong <10walls@gmail.com>
15356 * doc/extend.texi: Add note about reserved priorities
15357 to the constructor attribute.
15359 2021-08-25 Martin Sebor <msebor@redhat.com>
15361 * gimple-range-cache.cc (ssa_global_cache::dump): Avoid printing
15362 range table header alone.
15363 * gimple-range.cc (gimple_ranger::export_global_ranges): Same.
15365 2021-08-25 Jan Hubicka <hubicka@ucw.cz>
15367 * doc/invoke.texi: Document --param modref-max-adjustments.
15368 * ipa-modref-tree.c (test_insert_search_collapse): Update.
15369 (test_merge): Update.
15370 * ipa-modref-tree.h (struct modref_access_node): Add adjustments;
15371 (modref_access_node::operator==): Fix handling of access ranges.
15372 (modref_access_node::contains): Constify parameter; handle also
15373 mismatched parm offsets.
15374 (modref_access_node::update): New function.
15375 (modref_access_node::merge): New function.
15376 (unspecified_modref_access_node): Update constructor.
15377 (modref_ref_node::insert_access): Add record_adjustments parameter;
15379 (modref_ref_node::try_merge_with): New private function.
15380 (modref_tree::insert): New record_adjustments parameter.
15381 (modref_tree::merge): New record_adjustments parameter.
15382 (modref_tree::copy_from): Update.
15383 * ipa-modref.c (dump_access): Dump adjustments field.
15384 (get_access): Update constructor.
15385 (record_access): Update call of insert.
15386 (record_access_lto): Update call of insert.
15387 (merge_call_side_effects): Add record_adjustments parameter.
15388 (get_access_for_fnspec): Update.
15389 (process_fnspec): Update.
15390 (analyze_call): Update.
15391 (analyze_function): Update.
15392 (read_modref_records): Update.
15393 (ipa_merge_modref_summary_after_inlining): Update.
15394 (propagate_unknown_call): Update.
15395 (modref_propagate_in_scc): Update.
15396 * params.opt (param-max-modref-adjustments=): New.
15398 2021-08-25 Michael Meissner <meissner@linux.ibm.com>
15400 * config/rs6000/vsx.md (UNSPEC_XXSPLTIDP): Rename from
15402 (xxspltiw_v4si): Use vecperm type attribute.
15403 (xxspltiw_v4si_inst): Use vecperm type attribute.
15404 (xxspltiw_v4sf_inst): Likewise.
15405 (xxspltidp_v2df): Use vecperm type attribute. Use
15406 UNSPEC_XXSPLTIDP instead of UNSPEC_XXSPLTID.
15407 (xxspltidp_v2df_inst): Likewise.
15408 (xxsplti32dx_v4si): Use vecperm type attribute.
15409 (xxsplti32dx_v4si_inst): Likewise.
15410 (xxsplti32dx_v4sf_inst): Likewise.
15411 (xxblend_<mode>): Likewise.
15412 (xxpermx): Likewise.
15413 (xxpermx_inst): Likewise.
15414 (xxeval): Likewise.
15416 2021-08-25 Lewis Hyatt <lhyatt@gmail.com>
15419 * coretypes.h (typedef diagnostic_input_charset_callback): Declare.
15420 * diagnostic.c (diagnostic_initialize_input_context): New function.
15421 * diagnostic.h (diagnostic_initialize_input_context): Declare.
15422 * input.c (default_charset_callback): New function.
15423 (file_cache::initialize_input_context): New function.
15424 (file_cache_slot::create): Added ability to convert the input
15425 according to the input context.
15426 (file_cache::file_cache): Initialize the new input context.
15427 (class file_cache_slot): Added new m_alloc_offset member.
15428 (file_cache_slot::file_cache_slot): Initialize the new member.
15429 (file_cache_slot::~file_cache_slot): Handle potentially offset buffer.
15430 (file_cache_slot::maybe_grow): Likewise.
15431 (file_cache_slot::needs_read_p): Handle NULL fp, which is now possible.
15432 (file_cache_slot::get_next_line): Likewise.
15433 * input.h (class file_cache): Added input context member.
15435 2021-08-25 Richard Biener <rguenther@suse.de>
15437 PR tree-optimization/102046
15438 * tree-vect-slp.c (vect_build_slp_tree_2): Conservatively
15439 update ->any_pattern when swapping operands.
15441 2021-08-25 Hongyu Wang <hongyu.wang@intel.com>
15444 * config/i386/i386.c (ix86_live_on_entry): Adjust comment.
15445 (ix86_decompose_address): Remove retval check for ASHIFT,
15446 allow non-canonical zero extend if AND mask covers ASHIFT
15448 (ix86_legitimate_address_p): Adjust condition for decompose.
15449 (ix86_rtx_costs): Adjust cost for lea with non-canonical
15451 Co-Authored by: Uros Bizjak <ubizjak@gmail.com>
15453 2021-08-25 Jiufu Guo <guojiufu@linux.ibm.com>
15455 PR tree-optimization/101145
15456 * tree-ssa-loop-niter.c (number_of_iterations_until_wrap):
15458 (number_of_iterations_lt): Invoke above function.
15459 (adjust_cond_for_loop_until_wrap):
15460 Merge to number_of_iterations_until_wrap.
15461 (number_of_iterations_cond): Update invokes for
15462 adjust_cond_for_loop_until_wrap and number_of_iterations_lt.
15464 2021-08-25 konglin1 <lingling.kong@intel.com>
15467 * config/i386/avx512dqintrin.h (_mm512_fpclass_ps_mask): Fix
15468 macro define in O0.
15469 (_mm512_mask_fpclass_ps_mask): Ditto.
15471 2021-08-25 Kewen Lin <linkw@linux.ibm.com>
15473 * config/rs6000/altivec.md (vec_unpacku_hi_v16qi): Remove.
15474 (vec_unpacku_hi_v8hi): Likewise.
15475 (vec_unpacku_lo_v16qi): Likewise.
15476 (vec_unpacku_lo_v8hi): Likewise.
15477 (vec_unpacku_hi_<VP_small_lc>): New define_expand.
15478 (vec_unpacku_lo_<VP_small_lc>): Likewise.
15480 2021-08-24 David Edelsohn <dje.gcc@gmail.com>
15482 * config/rs6000/aix.h (SYSTEM_IMPLICIT_EXTERN_C): Delete.
15483 * config/rs6000/aix71.h (SYSTEM_IMPLICIT_EXTERN_C): Define.
15484 * config/rs6000/aix72.h (SYSTEM_IMPLICIT_EXTERN_C): Define.
15485 * config/rs6000/aix73.h (TARGET_AIX_VERSION): Increase to 73.
15487 2021-08-24 Roger Sayle <roger@nextmovesoftware.com>
15489 PR middle-end/102031
15490 * simplify-rtx.c (simplify_truncation): When comparing precisions
15491 use "subreg_prec" variable, not "subreg_mode".
15493 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
15495 * config/rs6000/rs6000-builtin-new.def: Add power10 and power10-64
15498 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
15500 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Initialize
15501 various pointer type nodes.
15502 * config/rs6000/rs6000.h (rs6000_builtin_type_index): Add enum
15503 values for various pointer types.
15504 (ptr_V16QI_type_node): New macro.
15505 (ptr_V1TI_type_node): New macro.
15506 (ptr_V2DI_type_node): New macro.
15507 (ptr_V2DF_type_node): New macro.
15508 (ptr_V4SI_type_node): New macro.
15509 (ptr_V4SF_type_node): New macro.
15510 (ptr_V8HI_type_node): New macro.
15511 (ptr_unsigned_V16QI_type_node): New macro.
15512 (ptr_unsigned_V1TI_type_node): New macro.
15513 (ptr_unsigned_V8HI_type_node): New macro.
15514 (ptr_unsigned_V4SI_type_node): New macro.
15515 (ptr_unsigned_V2DI_type_node): New macro.
15516 (ptr_bool_V16QI_type_node): New macro.
15517 (ptr_bool_V8HI_type_node): New macro.
15518 (ptr_bool_V4SI_type_node): New macro.
15519 (ptr_bool_V2DI_type_node): New macro.
15520 (ptr_bool_V1TI_type_node): New macro.
15521 (ptr_pixel_type_node): New macro.
15522 (ptr_intQI_type_node): New macro.
15523 (ptr_uintQI_type_node): New macro.
15524 (ptr_intHI_type_node): New macro.
15525 (ptr_uintHI_type_node): New macro.
15526 (ptr_intSI_type_node): New macro.
15527 (ptr_uintSI_type_node): New macro.
15528 (ptr_intDI_type_node): New macro.
15529 (ptr_uintDI_type_node): New macro.
15530 (ptr_intTI_type_node): New macro.
15531 (ptr_uintTI_type_node): New macro.
15532 (ptr_long_integer_type_node): New macro.
15533 (ptr_long_unsigned_type_node): New macro.
15534 (ptr_float_type_node): New macro.
15535 (ptr_double_type_node): New macro.
15536 (ptr_long_double_type_node): New macro.
15537 (ptr_dfloat64_type_node): New macro.
15538 (ptr_dfloat128_type_node): New macro.
15539 (ptr_ieee128_type_node): New macro.
15540 (ptr_ibm128_type_node): New macro.
15541 (ptr_vector_pair_type_node): New macro.
15542 (ptr_vector_quad_type_node): New macro.
15543 (ptr_long_long_integer_type_node): New macro.
15544 (ptr_long_long_unsigned_type_node): New macro.
15546 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
15548 * config/rs6000/rs6000-builtin-new.def: Add power9-vector, power9,
15549 and power9-64 stanzas.
15551 2021-08-24 Roger Sayle <roger@nextmovesoftware.com>
15552 Tom de Vries <tdevries@suse.de>
15554 * config.gcc (nvptx-*-*): Define {c,c++}_target_objs.
15555 * config/nvptx/nvptx-protos.h (nvptx_cpu_cpp_builtins): Prototype.
15556 * config/nvptx/nvptx.h (TARGET_CPU_CPP_BUILTINS): Implement with
15557 a call to the new nvptx_cpu_cpp_builtins function in nvptx-c.c.
15558 * config/nvptx/t-nvptx (nvptx-c.o): New rule.
15559 * config/nvptx/nvptx-c.c: New source file.
15560 (nvptx_cpu_cpp_builtins): Move implementation here.
15562 2021-08-24 Martin Sebor <msebor@redhat.com>
15564 PR middle-end/101600
15565 PR middle-end/101977
15566 * gimple-ssa-warn-access.cc (maybe_warn_for_bound): Tighten up
15567 the phrasing of a warning.
15568 (check_access): Use the remaining size after subtracting any offset
15569 rather than the whole object size.
15570 * pointer-query.cc (access_ref::get_ref): Clear BASE0 flag if it's
15571 clear for any nonnull PHI argument.
15572 (compute_objsize): Clear argument.
15574 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
15576 * config/rs6000/rs6000-builtin-new.def: Add power8-vector stanza.
15578 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
15580 * config/rs6000/rs6000-builtin-new.def: Add power7 and power7-64
15583 2021-08-24 Andrew MacLeod <amacleod@redhat.com>
15585 * value-relation.cc (rr_transitive_table): New.
15586 (relation_transitive): New.
15587 (value_relation::swap): Remove.
15588 (value_relation::apply_transitive): New.
15589 (relation_oracle::relation_oracle): Allocate a new tmp bitmap.
15590 (relation_oracle::register_relation): Call register_transitives.
15591 (relation_oracle::register_transitives): New.
15592 * value-relation.h (relation_oracle): Add new temporary bitmap and
15595 2021-08-24 H.J. Lu <hjl.tools@gmail.com>
15598 * config/i386/i386-expand.c (ix86_expand_vector_move): Broadcast
15599 from integer to a pseudo vector register.
15601 2021-08-24 Richard Biener <rguenther@suse.de>
15603 PR tree-optimization/100089
15604 * tree-vectorizer.h (vect_slp_bb): Rename to ...
15605 (vect_slp_if_converted_bb): ... this and get the original
15606 loop as new argument.
15607 * tree-vectorizer.c (try_vectorize_loop_1): Revert previous fix,
15608 pass original loop to vect_slp_if_converted_bb.
15609 * tree-vect-slp.c (vect_bb_vectorization_profitable_p):
15610 If orig_loop was passed scan the not vectorized stmts
15611 for COND_EXPRs and force not profitable if found.
15612 (vect_slp_region): Pass down all SLP instances to costing
15613 if orig_loop was specified.
15614 (vect_slp_bbs): Pass through orig_loop.
15615 (vect_slp_bb): Rename to ...
15616 (vect_slp_if_converted_bb): ... this and get the original
15617 loop as new argument.
15618 (vect_slp_function): Adjust.
15620 2021-08-24 Richard Earnshaw <rearnsha@arm.com>
15623 * config/arm/arm.md (attribute arch): Add fix_vlldm.
15624 (arch_enabled): Use it.
15625 * config/arm/vfp.md (lazy_store_multiple_insn): Add alternative to
15626 use when erratum mitigation is needed.
15628 2021-08-24 Richard Earnshaw <rearnsha@arm.com>
15631 * config/arm/arm.opt (mfix-cmse-cve-2021-35465): New option.
15632 * doc/invoke.texi (Arm Options): Document it.
15633 * config/arm/arm-cpus.in (quirk_vlldm): New feature bit.
15634 (ALL_QUIRKS): Add quirk_vlldm.
15635 (cortex-m33): Add quirk_vlldm.
15636 (cortex-m35p, cortex-m55): Likewise.
15637 * config/arm/arm.c (arm_option_override): Enable fix_vlldm if
15638 targetting an affected CPU and not explicitly controlled on
15641 2021-08-24 Richard Earnshaw <rearnsha@arm.com>
15643 * config/arm/vfp.md (lazy_store_multiple_insn): Rewrite as valid RTL.
15644 (lazy_load_multiple_insn): Likewise.
15646 2021-08-24 liuhongt <hongtao.liu@intel.com>
15649 * config/i386/sse.md (<avx512>_vternlog<mode><sd_maskz_name>):
15650 Enable avx512 embedded broadcast.
15651 (*<avx512>_vternlog<mode>_all): Ditto.
15652 (<avx512>_vternlog<mode>_mask): Ditto.
15654 2021-08-24 liuhongt <hongtao.liu@intel.com>
15657 * config/i386/i386.c (ix86_rtx_costs): Define cost for
15659 * config/i386/i386.h (STRIP_UNARY): New macro.
15660 * config/i386/predicates.md (reg_or_notreg_operand): New
15662 * config/i386/sse.md (*<avx512>_vternlog<mode>_all): New define_insn.
15663 (*<avx512>_vternlog<mode>_1): New pre_reload
15664 define_insn_and_split.
15665 (*<avx512>_vternlog<mode>_2): Ditto.
15666 (*<avx512>_vternlog<mode>_3): Ditto.
15667 (any_logic1,any_logic2): New code iterator.
15668 (logic_op): New code attribute.
15669 (ternlogsuffix): Extend to VNxDF and VNxSF.
15671 2021-08-24 Richard Biener <rguenther@suse.de>
15673 * doc/invoke.texi (vect-inner-loop-cost-factor): Adjust.
15674 * params.opt (--param vect-inner-loop-cost-factor): Adjust
15676 * tree-vect-loop.c (vect_analyze_loop_form): Initialize
15677 inner_loop_cost_factor to the minimum of the estimated number
15678 of iterations of the inner loop and vect-inner-loop-cost-factor.
15680 2021-08-24 Roger Sayle <roger@nextmovesoftware.com>
15681 Richard Biener <rguenther@suse.de>
15683 * config/i386/i386-features.c (compute_convert_gain): Provide
15684 more accurate values for CONST_INT, when optimizing for size.
15685 * config/i386/i386.c (COSTS_N_BYTES): Move definition from here...
15686 * config/i386/i386.h (COSTS_N_BYTES): to here.
15688 2021-08-24 Roger Sayle <roger@nextmovesoftware.com>
15689 Jakub Jelinek <jakub@redhat.com>
15691 PR middle-end/102029
15692 * match.pd (shift transformations): Add an additional check for
15693 !POINTER_TYPE_P in the recently added left shift transformation.
15695 2021-08-24 liuhongt <hongtao.liu@intel.com>
15697 PR tree-optimization/100089
15698 * tree-vectorizer.c (try_vectorize_loop_1): Disable slp in
15699 loop vectorizer when cost model is very-cheap.
15701 2021-08-23 Bill Schmidt <wschmidt@linux.ibm.com>
15703 * config/rs6000/rs6000-gen-builtins.c (parse_bif_entry): Don't call
15704 asprintf, which is not available on AIX.
15706 2021-08-23 Bill Schmidt <wschmidt@linux.ibm.com>
15708 * config.gcc (target_gtfiles): Add ./rs6000-builtins.h.
15709 * config/rs6000/t-rs6000 (EXTRA_GTYPE_DEPS): Set.
15711 2021-08-23 Bill Schmidt <wschmidt@linux.ibm.com>
15713 * config.gcc (powerpc*-*-*): Add rs6000-builtins.o to extra_objs.
15714 * config/rs6000/rs6000-gen-builtins.c (main): Close init_file
15716 * config/rs6000/t-rs6000 (rs6000-gen-builtins.o): New target.
15717 (rbtree.o): Likewise.
15718 (rs6000-gen-builtins): Likewise.
15719 (rs6000-builtins.c): Likewise.
15720 (rs6000-builtins.h): Likewise.
15721 (rs6000.o): Add dependency.
15722 (EXTRA_HEADERS): Add rs6000-vecdefines.h.
15723 (rs6000-vecdefines.h): New target.
15724 (rs6000-builtins.o): Likewise.
15725 (rs6000-call.o): Add rs6000-builtins.h as a dependency.
15726 (rs6000-c.o): Likewise.
15728 2021-08-23 Bill Schmidt <wschmidt@linux.ibm.com>
15731 * config/rs6000/rs6000-gen-builtins.c (consume_whitespace):
15732 Diagnose buffer overrun.
15733 (safe_inc_pos): Fix overrun detection.
15734 (match_identifier): Diagnose buffer overrun.
15735 (match_integer): Likewise.
15736 (match_to_right_bracket): Likewise.
15738 2021-08-23 Jan Hubicka <hubicka@ucw.cz>
15740 * ipa-modref-tree.h (modref_access_node::range_info_useful_p):
15741 Improve range compare.
15742 (modref_access_node::contains): New member function.
15743 (modref_access_node::search): Remove.
15744 (modref_access_node::insert): Be smarter about subaccesses.
15746 2021-08-23 Thomas Schwinge <thomas@codesourcery.com>
15748 * config/i386/i386-options.c (ix86_omp_device_kind_arch_isa)
15749 <omp_device_arch> [ACCEL_COMPILER]: Match "intel_mic".
15750 * config/i386/t-omp-device (omp-device-properties-i386) <arch>:
15753 2021-08-23 Jeff Law <jlaw@localhost.localdomain>
15755 * config/h8300/h8300-protos.h (h8300_expand_epilogue): Add new
15757 * config/h8300/jumpcall.md (call, call_value): Restrict to
15758 !SIBLING_CALL_P cases.
15759 (subcall, sibcall_value): New patterns & expanders.
15760 * config/h8300/proepi.md (epilogue): Pass new argument to
15761 h8300_expand_epilogue.
15762 (sibcall_epilogue): New expander.
15763 * config/h8300/h8300.c (h8300_expand_epilogue): Handle sibcall
15765 (h8300_ok_for_sibcall_p): New function.
15766 (TARGET_FUNCTION_OK_FOR_SIBCALL): define.
15768 2021-08-23 Roger Sayle <roger@nextmovesoftware.com>
15770 * simplify-rtx.c (simplify_unary_operation_1): [TRUNCATE]:
15771 Handle case where the operand is already the desired mode.
15773 2021-08-23 Richard Biener <rguenther@suse.de>
15776 * tree-ssa-structalias.c (ipa_pta_execute): Check in_other_partition
15777 in addition to has_gimple_body.
15779 2021-08-23 Jan Hubicka <hubicka@ucw.cz>
15781 PR middle-end/101949
15782 * ipa-modref.c (analyze_ssa_name_flags): Fix merging of
15785 2021-08-23 Martin Liska <mliska@suse.cz>
15787 * doc/invoke.texi: Put the option out of -mxl-mode-app-model
15790 2021-08-23 Richard Biener <rguenther@suse.de>
15792 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
15793 Properly scale the inner loop cost only once.
15795 2021-08-23 Roger Sayle <roger@nextmovesoftware.com>
15797 * tree-ssa-ccp.c (bit_value_binop) [TRUNC_MOD_EXPR, TRUNC_DIV_EXPR]:
15798 Provide bounds for unsigned (and signed with non-negative operands)
15799 division and modulus.
15801 2021-08-23 Roger Sayle <roger@nextmovesoftware.com>
15803 * simplify-rtx.c (simplify_truncation): Generalize simplification
15804 of (truncate:A (subreg:B X)).
15805 (simplify_unary_operation_1) [FLOAT_TRUNCATE, FLOAT_EXTEND,
15806 SIGN_EXTEND, ZERO_EXTEND]: Handle cases where the operand
15807 already has the desired machine mode.
15808 (test_scalar_int_ops): Add tests that useless extensions and
15809 truncations are optimized away.
15810 (test_scalar_int_ext_ops): New self-test function to confirm
15811 that truncations of extensions are correctly simplified.
15812 (test_scalar_int_ext_ops2): New self-test function to check
15813 truncations of truncations, extensions of extensions, and
15814 truncations of extensions.
15815 (test_scalar_ops): Call the above two functions with a
15816 representative sampling of integer machine modes.
15818 2021-08-23 Roger Sayle <roger@nextmovesoftware.com>
15820 * match.pd (shift transformations): Change the sign of an
15821 LSHIFT_EXPR if it reduces the number of explicit conversions.
15823 2021-08-23 Jakub Jelinek <jakub@redhat.com>
15825 PR tree-optimization/86723
15826 * gimple-ssa-store-merging.c (find_bswap_or_nop_finalize): Add
15827 cast64_to_32 argument, set *cast64_to_32 to false, unless n is
15828 non-memory permutation of 64-bit src which only has bytes of
15829 0 or [5..8] and n->range is 4.
15830 (find_bswap_or_nop): Add cast64_to_32 and mask arguments, adjust
15831 find_bswap_or_nop_finalize caller, support bswap with some bytes
15832 zeroed, as long as at least two bytes are not zeroed.
15833 (bswap_replace): Add mask argument and handle masking of bswap
15835 (maybe_optimize_vector_constructor): Adjust find_bswap_or_nop
15836 caller, punt if cast64_to_32 or mask is not all ones.
15837 (pass_optimize_bswap::execute): Adjust find_bswap_or_nop_finalize
15838 caller, for now punt if cast64_to_32.
15840 2021-08-23 Richard Biener <rguenther@suse.de>
15842 PR tree-optimization/79334
15843 * tree-ssa-sccvn.c (copy_reference_ops_from_ref): Record
15844 a type also for COMPONENT_REFs.
15845 (vn_reference_may_trap): Check ARRAY_REF with constant index
15846 against the array domain.
15848 2021-08-23 liuhongt <hongtao.liu@intel.com>
15851 * config/i386/sse.md (*avx512f_pshufb_truncv8hiv8qi_1): Add
15852 TARGET_AVX512BW to condition.
15854 2021-08-23 Jakub Jelinek <jakub@redhat.com>
15857 * dwarf2out.c (gen_variable_die): Add DW_AT_location for global
15858 register variables already during early_dwarf if possible.
15860 2021-08-23 Christophe Lyon <christophe.lyon@foss.st.com>
15862 * config/arm/arm_mve.h: Fix __arm_vctp16q return type.
15864 2021-08-23 Christophe Lyon <christophe.lyon@foss.st.com>
15867 * config/arm/arm.opt: Fix typo.
15868 * config/arm/t-rmprofile: Fix typo.
15870 2021-08-23 Jakub Jelinek <jakub@redhat.com>
15872 * tree.h (OMP_CLAUSE_GRAINSIZE_STRICT): Define.
15873 (OMP_CLAUSE_NUM_TASKS_STRICT): Define.
15874 * tree-pretty-print.c (dump_omp_clause) <case OMP_CLAUSE_GRAINSIZE,
15875 case OMP_CLAUSE_NUM_TASKS>: Print strict: modifier.
15876 * omp-expand.c (expand_task_call): Use GOMP_TASK_FLAG_STRICT in iflags
15877 if either grainsize or num_tasks clause has the strict modifier.
15879 2021-08-23 Martin Liska <mliska@suse.cz>
15881 * dbgcnt.def (DEBUG_COUNTER): New counter.
15882 * gimple.c (gimple_call_arg_flags): Use it in IPA PTA.
15884 2021-08-23 Jan Hubicka <hubicka@ucw.cz>
15886 * ipa-modref.c (analyze_ssa_name_flags): Improve handling of return slot.
15888 2021-08-23 Xi Ruoyao <xry111@mengyan1223.wang>
15891 * config/mips/mips-protos.h (mips_msa_output_shift_immediate):
15893 * config/mips/mips.c (mips_msa_output_shift_immediate): New
15895 * config/mips/mips-msa.md (vashl<mode>3, vashr<mode>3,
15896 vlshr<mode>3): Call it.
15898 2021-08-22 Jan Hubicka <hubicka@ucw.cz>
15899 Martin Liska <mliska@suse.cz>
15901 PR middle-end/101949
15902 * ipa-modref.c (analyze_ssa_name_flags): Indirect call implies
15905 2021-08-21 Dragan Mladjenovic <OT_Dragan.Mladjenovic@mediatek.com>
15907 * config/mips/mips.c (mips_function_rodata_section,
15908 TARGET_ASM_FUNCTION_RODATA_SECTION): Removed.
15910 2021-08-21 John David Anglin <danglin@gcc.gnu.org>
15912 * config/pa/pa.c (pa_asm_output_aligned_common): Remove warning.
15914 2021-08-20 Serge Belyshev <belyshev@depni.sinp.msu.ru>
15916 * configure.ac (thread-local storage support): Remove tls_first_major
15917 and tls_first_minor. Use "$conftest_s" to check support.
15918 * configure: Regenerate.
15920 2021-08-20 Serge Belyshev <belyshev@depni.sinp.msu.ru>
15922 * configure.ac: Fixup formatting.
15924 2021-08-20 Serge Belyshev <belyshev@depni.sinp.msu.ru>
15926 * acinclude.m4 (gcc_GAS_CHECK_FEATURE): Remove third argument and ...
15927 * configure.ac: ... update all callers.
15929 2021-08-20 Serge Belyshev <belyshev@depni.sinp.msu.ru>
15932 * acinclude.m4 (_gcc_COMPUTE_GAS_VERSION, _gcc_GAS_VERSION_GTE_IFELSE)
15933 (gcc_GAS_VERSION_GTE_IFELSE): Remove.
15934 (gcc_GAS_CHECK_FEATURE): Do not handle in-tree case specially.
15935 * configure.ac: Remove gcc_cv_gas_major_version, gcc_cv_gas_minor_version.
15936 Remove remaining checks for in-tree assembler.
15937 * configure: Regenerate.
15939 2021-08-20 Jeff Law <jlaw@localhost.localdomain>
15941 * config/h8300/h8300.c (shift_alg_hi): Improve arithmetic shift right
15942 by 15 bits for H8/300H and H8/S. Improve logical shifts by 12
15944 (shift_alg_si): Improve arithmetic right shift by 28-30 bits for
15945 H8/300H. Improve arithmetic shift right by 15 bits for H8/S.
15946 Improve logical shifts by 27 bits for H8/S.
15947 (get_shift_alg): Corresponding changes.
15948 (h8300_option_override): Revert to loops for -Os when profitable.
15950 2021-08-20 Richard Biener <rguenther@suse.de>
15952 * tree-vect-data-refs.c (dr_group_sort_cmp): Do not compare
15954 (vect_analyze_data_ref_accesses): Likewise. Assign the BB
15955 index as group_id when dataref_groups were not computed.
15956 * tree-vect-slp.c (vect_slp_bbs): Bump current_group when
15957 we advace to the next BB.
15959 2021-08-20 Jakub Jelinek <jakub@redhat.com>
15961 * omp-builtins.def (BUILT_IN_GOMP_WARNING, BUILT_IN_GOMP_ERROR): New
15964 2021-08-20 Martin Liska <mliska@suse.cz>
15966 PR gcov-profile/89961
15967 * gcov.c (make_gcov_file_name): Rewrite using std::string.
15968 (mangle_name): Simplify, do not used the second argument.
15969 (strip_extention): New function.
15970 (get_md5sum): Likewise.
15971 (get_gcov_intermediate_filename): Handle properly -p and -x
15973 (output_gcov_file): Use string type.
15974 (generate_results): Likewise.
15975 (md5sum_to_hex): Remove.
15977 2021-08-20 Michael Meissner <meissner@linux.ibm.com>
15979 * config/rs6000/altivec.md (UNSPEC_XXEVAL): Move to vsx.md.
15980 (UNSPEC_XXSPLTIW): Move to vsx.md.
15981 (UNSPEC_XXSPLTID): Move to vsx.md.
15982 (UNSPEC_XXSPLTI32DX): Move to vsx.md.
15983 (UNSPEC_XXBLEND): Move to vsx.md.
15984 (UNSPEC_XXPERMX): Move to vsx.md.
15985 (VM3): Move to vsx.md.
15986 (VM3_char): Move to vsx.md.
15987 (xxspltiw_v4si): Move to vsx.md.
15988 (xxspltiw_v4sf): Move to vsx.md.
15989 (xxspltiw_v4sf_inst): Move to vsx.md.
15990 (xxspltidp_v2df): Move to vsx.md.
15991 (xxspltidp_v2df_inst): Move to vsx.md.
15992 (xxsplti32dx_v4si_inst): Move to vsx.md.
15993 (xxsplti32dx_v4sf): Move to vsx.md.
15994 (xxsplti32dx_v4sf_inst): Move to vsx.md.
15995 (xxblend_<mode>): Move to vsx.md.
15996 (xxpermx): Move to vsx.md.
15997 (xxpermx_inst): Move to vsx.md.
15998 * config/rs6000/vsx.md (UNSPEC_XXEVAL): Move from altivec.md.
15999 (UNSPEC_XXSPLTIW): Move from altivec.md.
16000 (UNSPEC_XXSPLTID): Move from altivec.md.
16001 (UNSPEC_XXSPLTI32DX): Move from altivec.md.
16002 (UNSPEC_XXBLEND): Move from altivec.md.
16003 (UNSPEC_XXPERMX): Move from altivec.md.
16004 (VM3): Move from altivec.md.
16005 (VM3_char): Move from altivec.md.
16006 (xxspltiw_v4si): Move from altivec.md.
16007 (xxspltiw_v4sf): Move from altivec.md.
16008 (xxspltiw_v4sf_inst): Move from altivec.md.
16009 (xxspltidp_v2df): Move from altivec.md.
16010 (xxspltidp_v2df_inst): Move from altivec.md.
16011 (xxsplti32dx_v4si_inst): Move from altivec.md.
16012 (xxsplti32dx_v4sf): Move from altivec.md.
16013 (xxsplti32dx_v4sf_inst): Move from altivec.md.
16014 (xxblend_<mode>): Move from altivec.md.
16015 (xxpermx): Move from altivec.md.
16016 (xxpermx_inst): Move from altivec.md.
16018 2021-08-19 Roger Sayle <roger@nextmovesoftware.com>
16020 * tree-vect-generic.c (expand_vector_operations_1): Use either
16021 gimplify_build1 or gimplify_build2 instead of gimple_build_assign
16022 when constructing scalar splat expressions.
16024 2021-08-19 Peter Bergner <bergner@linux.ibm.com>
16027 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_mma_builtin): Cast
16028 pointer to __vector_pair *.
16030 2021-08-19 Martin Sebor <msebor@redhat.com>
16032 * gimple-range.cc: Add comments.
16033 * gimple-range.h: Same.
16035 2021-08-19 Martin Sebor <msebor@redhat.com>
16037 PR middle-end/101984
16038 * gimple-ssa-warn-access.cc (pass_waccess::execute): Also call
16041 2021-08-19 Jeff Law <jlaw@localhost.localdomain>
16043 * config.gcc (h8300-*-elf*): Do not include dbxelf.h.
16044 (h8300-*-linux*, v850-*-rtems*, v850*-elf*): Likewise.
16045 * config/v850/v850.h (DEFAULT_GDB_EXTENSIONS): Remove.
16047 2021-08-19 Jakub Jelinek <jakub@redhat.com>
16049 PR middle-end/101950
16050 * optabs.c (expand_clrsb_using_clz): New function.
16051 (expand_unop): Use it as another clrsb expansion fallback.
16053 2021-08-19 liuhongt <hongtao.liu@intel.com>
16056 2021-07-28 liuhongt <hongtao.liu@intel.com>
16059 * config/i386/i386.h (processor_costs): Add new member
16061 * config/i386/x86-tune-costs.h (ix86_size_cost, i386_cost,
16062 i486_cost, pentium_cost, lakemont_cost, pentiumpro_cost,
16063 geode_cost, k6_cost, athlon_cost, k8_cost, amdfam10_cost,
16064 bdver_cost, znver1_cost, znver2_cost, znver3_cost,
16065 btver1_cost, btver2_cost, btver3_cost, pentium4_cost,
16066 nocona_cost, atom_cost, atom_cost, slm_cost, intel_cost,
16067 generic_cost, core_cost): Initialize integer_to_sse same value
16069 (skylake_cost): Initialize integer_to_sse twice as much as sse_op.
16070 * config/i386/i386.c (ix86_builtin_vectorization_cost):
16071 Use integer_to_sse instead of sse_op to calculate the cost of
16074 2021-08-18 Iain Sandoe <iain@sandoe.co.uk>
16076 * config.gcc: Include rpath.opt for Darwin.
16077 * config/darwin.h (DRIVER_SELF_SPECS): Handle -rpath.
16079 2021-08-18 Thomas Schwinge <thomas@codesourcery.com>
16081 PR bootstrap/101959
16082 * hash-map-tests.c (test_map_of_type_with_ctor_and_dtor_expand):
16085 2021-08-18 Jonathan Wright <jonathan.wright@arm.com>
16087 * config/aarch64/arm_neon.h (vld3_lane_f64): Use float RTL
16088 pattern and type cast.
16089 (vld4_lane_f32): Use float RTL pattern.
16090 (vld4q_lane_f64): Use float type cast.
16092 2021-08-18 Jan Hubicka <hubicka@ucw.cz>
16094 * tree-ssa-uninit.c (maybe_warn_pass_by_reference): Check also
16097 2021-08-18 Thomas Schwinge <thomas@codesourcery.com>
16099 * hash-map-tests.c (test_map_of_type_with_ctor_and_dtor): Extend.
16100 (test_map_of_type_with_ctor_and_dtor_expand): Add function.
16101 (hash_map_tests_c_tests): Call it.
16103 2021-08-18 Thomas Schwinge <thomas@codesourcery.com>
16105 * ggc.h (enum ggc_collect): New.
16106 (ggc_collect): Use it.
16107 * ggc-page.c: Adjust.
16108 * ggc-common.c: Likewise.
16109 * ggc-tests.c: Likewise.
16110 * read-rtl-function.c: Likewise.
16111 * selftest-run-tests.c: Likewise.
16112 * doc/gty.texi (Invoking the garbage collector): Likewise.
16114 2021-08-18 liuhongt <hongtao.liu@intel.com>
16117 * config/i386/i386.h (TARGET_V2DF_REDUCTION_PREFER_HADDPD):
16119 * config/i386/sse.md (*sse3_haddv2df3_low): Add
16120 TARGET_V2DF_REDUCTION_PREFER_HADDPD.
16121 (*sse3_hsubv2df3_low): Ditto.
16122 * config/i386/x86-tune.def
16123 (X86_TUNE_V2DF_REDUCTION_PREFER_HADDPD): New tune.
16125 2021-08-17 Andrew MacLeod <amacleod@redhat.com>
16127 * gimple-range-gori.cc (gori_compute::gori_compute): Enable tracing.
16128 (gori_compute::compute_operand_range): Add tracing.
16129 (gori_compute::logical_combine): Ditto.
16130 (gori_compute::compute_logical_operands): Ditto.
16131 (gori_compute::compute_operand1_range): Ditto.
16132 (gori_compute::compute_operand2_range): Ditto.
16133 (gori_compute::outgoing_edge_range_p): Ditto.
16134 * gimple-range-gori.h (class gori_compute): Add range_tracer.
16136 2021-08-17 Andrew MacLeod <amacleod@redhat.com>
16138 * flag-types.h (enum evrp_mode): Adjust evrp-mode values.
16139 * gimple-range-cache.cc (DEBUG_RANGE_CACHE): Relocate from.
16140 * gimple-range-trace.h (DEBUG_RANGE_CACHE): Here.
16141 * params.opt (--param=evrp-mode): Adjust options.
16143 2021-08-17 Andrew MacLeod <amacleod@redhat.com>
16145 * Makefile.in (OBJS): Add gimple-range-trace.o.
16146 * gimple-range-cache.h (enable_new_values): Remove unused prototype.
16147 * gimple-range-fold.cc: Adjust headers.
16148 * gimple-range-trace.cc: New.
16149 * gimple-range-trace.h: New.
16150 * gimple-range.cc (gimple_ranger::gimple_ranger): Enable tracer.
16151 (gimple_ranger::range_of_expr): Add tracing.
16152 (gimple_ranger::range_on_entry): Ditto.
16153 (gimple_ranger::range_on_exit): Ditto.
16154 (gimple_ranger::range_on_edge): Ditto.
16155 (gimple_ranger::fold_range_internal): Ditto.
16156 (gimple_ranger::dump_bb): Do not calculate edge range twice.
16157 (trace_ranger::*): Remove.
16158 (enable_ranger): Never create a trace_ranger.
16159 (debug_seed_ranger): Move to gimple-range-trace.cc.
16160 (dump_ranger): Ditto.
16161 (debug_ranger): Ditto.
16162 * gimple-range.h: Include gimple-range-trace.h.
16163 (range_on_entry, range_on_exit): No longer virtual.
16164 (class trace_ranger): Remove.
16165 (DEBUG_RANGE_CACHE): Move to gimple-range-trace.h.
16167 2021-08-17 Martin Sebor <msebor@redhat.com>
16169 PR middle-end/101854
16170 * builtins.c (expand_builtin_alloca): Move warning code to check_alloca
16171 in gimple-ssa-warn-access.cc.
16172 * calls.c (alloc_max_size): Move code to check_alloca.
16173 (get_size_range): Move to pointer-query.cc.
16174 (maybe_warn_alloc_args_overflow): Move to gimple-ssa-warn-access.cc.
16175 (get_attr_nonstring_decl): Move to tree.c.
16176 (fntype_argno_type): Move to gimple-ssa-warn-access.cc.
16177 (append_attrname): Same.
16178 (maybe_warn_rdwr_sizes): Same.
16179 (initialize_argument_information): Move code to
16180 gimple-ssa-warn-access.cc.
16181 * calls.h (maybe_warn_alloc_args_overflow): Move to
16182 gimple-ssa-warn-access.h.
16183 (get_attr_nonstring_decl): Move to tree.h.
16184 (maybe_warn_nonstring_arg): Move to gimple-ssa-warn-access.h.
16185 (enum size_range_flags): Move to pointer-query.h.
16186 (get_size_range): Same.
16187 * gimple-ssa-warn-access.cc (has_location): Remove unused overload
16188 to avoid Clang -Wunused-function.
16189 (get_size_range): Declare static.
16190 (maybe_emit_free_warning): Rename...
16191 (maybe_check_dealloc_call): ...to this for consistency.
16192 (class pass_waccess): Add members.
16193 (pass_waccess::~pass_waccess): Defined.
16194 (alloc_max_size): Move here from calls.c.
16195 (maybe_warn_alloc_args_overflow): Same.
16196 (check_alloca): New function.
16197 (check_alloc_size_call): New function.
16198 (check_strncat): Handle another warning flag.
16199 (pass_waccess::check_builtin): Handle alloca.
16200 (fntype_argno_type): Move here from calls.c.
16201 (append_attrname): Same.
16202 (maybe_warn_rdwr_sizes): Same.
16203 (pass_waccess::check_call): Define.
16204 (check_nonstring_args): New function.
16205 (pass_waccess::check): Call new member functions.
16206 (pass_waccess::execute): Enable ranger.
16207 * gimple-ssa-warn-access.h (get_size_range): Move here from calls.h.
16208 (maybe_warn_nonstring_arg): Same.
16209 * gimple-ssa-warn-restrict.c: Remove #include.
16210 * pointer-query.cc (get_size_range): Move here from calls.c.
16211 * pointer-query.h (enum size_range_flags): Same.
16212 (get_size_range): Same.
16213 * tree.c (get_attr_nonstring_decl): Move here from calls.c.
16214 * tree.h (get_attr_nonstring_decl): Move here from calls.h.
16216 2021-08-17 Thomas Schwinge <thomas@codesourcery.com>
16218 * ggc.h (ggc_collect): Add 'force_collect' parameter.
16219 * ggc-page.c (ggc_collect): Use that one instead of global
16220 'ggc_force_collect'. Adjust all users.
16221 * doc/gty.texi (Invoking the garbage collector): Update.
16222 * ggc-internal.h (ggc_force_collect): Remove.
16223 * ggc-common.c (ggc_force_collect): Likewise.
16224 * selftest.h (forcibly_ggc_collect): Remove.
16225 * ggc-tests.c (selftest::forcibly_ggc_collect): Likewise.
16226 * read-rtl-function.c (test_loading_labels): Adjust.
16227 * selftest-run-tests.c (run_tests): Likewise.
16229 2021-08-17 Iain Sandoe <iain@sandoe.co.uk>
16231 * config/darwin.c (darwin_file_end): Reset and reclaim the
16232 section names table at the end of compile.
16234 2021-08-17 Iain Sandoe <iain@sandoe.co.uk>
16237 * config.in: Regenerate.
16238 * config/i386/darwin.h (EXTRA_ASM_OPTS): New
16239 (ASM_SPEC): Pass options to disable branch shortening where
16241 * configure: Regenerate.
16242 * configure.ac: Detect versions of 'as' that support the
16243 optimisation which has the bug.
16245 2021-08-17 Richard Biener <rguenther@suse.de>
16247 * optabs-query.c (supports_vec_gather_load_p): Also check
16249 (supports_vec_scatter_store_p): Likewise.
16250 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Fall
16251 back to masked variants if non-masked are not supported.
16252 * tree-vect-patterns.c (vect_recog_gather_scatter_pattern):
16253 When we need to use masked gather/scatter but do not have
16254 a mask set up a constant true one.
16255 * tree-vect-stmts.c (vect_check_scalar_mask): Also allow
16256 non-SSA_NAME masks.
16258 2021-08-17 Roger Sayle <roger@nextmovesoftware.com>
16260 * tree-ssa-ccp.c (bit_value_binop) [MINUS_EXPR]: Use same
16261 algorithm as PLUS_EXPR to improve subtraction bit bounds.
16262 [POINTER_DIFF_EXPR]: Treat as synonymous with MINUS_EXPR.
16264 2021-08-17 Roger Sayle <roger@nextmovesoftware.com>
16266 * tree-ssa-ccp.c (bit_value_mult_const): New helper function to
16267 calculate the mask-value pair result of a multiplication by an
16269 (bit_value_binop) [MULT_EXPR]: Call it from here for
16270 multiplications by (sparse) non-negative constants.
16272 2021-08-17 Christophe Lyon <christophe.lyon@foss.st.com>
16275 * config.gcc (gcc_cv_initfini_array): Leave undefined for
16276 uclinuxfdpiceabi targets.
16278 2021-08-17 Alexandre Oliva <oliva@adacore.com>
16280 * tree-inline.c (maybe_move_debug_stmts_to_successors): Don't
16281 reverse debug stmts.
16283 2021-08-17 Alexandre Oliva <oliva@adacore.com>
16285 * tree-cfg.c (dump_function_to_file): Use fun, not cfun.
16287 2021-08-17 Jonathan Wright <jonathan.wright@arm.com>
16289 * config/aarch64/arm_neon.h (__LD4_LANE_FUNC): Delete.
16290 (__LD4Q_LANE_FUNC): Likewise.
16291 (vld4_lane_u8): Define without macro.
16292 (vld4_lane_u16): Likewise.
16293 (vld4_lane_u32): Likewise.
16294 (vld4_lane_u64): Likewise.
16295 (vld4_lane_s8): Likewise.
16296 (vld4_lane_s16): Likewise.
16297 (vld4_lane_s32): Likewise.
16298 (vld4_lane_s64): Likewise.
16299 (vld4_lane_f16): Likewise.
16300 (vld4_lane_f32): Likewise.
16301 (vld4_lane_f64): Likewise.
16302 (vld4_lane_p8): Likewise.
16303 (vld4_lane_p16): Likewise.
16304 (vld4_lane_p64): Likewise.
16305 (vld4q_lane_u8): Likewise.
16306 (vld4q_lane_u16): Likewise.
16307 (vld4q_lane_u32): Likewise.
16308 (vld4q_lane_u64): Likewise.
16309 (vld4q_lane_s8): Likewise.
16310 (vld4q_lane_s16): Likewise.
16311 (vld4q_lane_s32): Likewise.
16312 (vld4q_lane_s64): Likewise.
16313 (vld4q_lane_f16): Likewise.
16314 (vld4q_lane_f32): Likewise.
16315 (vld4q_lane_f64): Likewise.
16316 (vld4q_lane_p8): Likewise.
16317 (vld4q_lane_p16): Likewise.
16318 (vld4q_lane_p64): Likewise.
16319 (vld4_lane_bf16): Likewise.
16320 (vld4q_lane_bf16): Likewise.
16322 2021-08-17 Jonathan Wright <jonathan.wright@arm.com>
16324 * config/aarch64/arm_neon.h (__LD3_LANE_FUNC): Delete.
16325 (__LD3Q_LANE_FUNC): Delete.
16326 (vld3_lane_u8): Define without macro.
16327 (vld3_lane_u16): Likewise.
16328 (vld3_lane_u32): Likewise.
16329 (vld3_lane_u64): Likewise.
16330 (vld3_lane_s8): Likewise.
16331 (vld3_lane_s16): Likewise.
16332 (vld3_lane_s32): Likewise.
16333 (vld3_lane_s64): Likewise.
16334 (vld3_lane_f16): Likewise.
16335 (vld3_lane_f32): Likewise.
16336 (vld3_lane_f64): Likewise.
16337 (vld3_lane_p8): Likewise.
16338 (vld3_lane_p16): Likewise.
16339 (vld3_lane_p64): Likewise.
16340 (vld3q_lane_u8): Likewise.
16341 (vld3q_lane_u16): Likewise.
16342 (vld3q_lane_u32): Likewise.
16343 (vld3q_lane_u64): Likewise.
16344 (vld3q_lane_s8): Likewise.
16345 (vld3q_lane_s16): Likewise.
16346 (vld3q_lane_s32): Likewise.
16347 (vld3q_lane_s64): Likewise.
16348 (vld3q_lane_f16): Likewise.
16349 (vld3q_lane_f32): Likewise.
16350 (vld3q_lane_f64): Likewise.
16351 (vld3q_lane_p8): Likewise.
16352 (vld3q_lane_p16): Likewise.
16353 (vld3q_lane_p64): Likewise.
16354 (vld3_lane_bf16): Likewise.
16355 (vld3q_lane_bf16): Likewise.
16357 2021-08-17 Jonathan Wright <jonathan.wright@arm.com>
16359 * config/aarch64/arm_neon.h (__LD2_LANE_FUNC): Delete.
16360 (__LD2Q_LANE_FUNC): Likewise.
16361 (vld2_lane_u8): Define without macro.
16362 (vld2_lane_u16): Likewise.
16363 (vld2_lane_u32): Likewise.
16364 (vld2_lane_u64): Likewise.
16365 (vld2_lane_s8): Likewise.
16366 (vld2_lane_s16): Likewise.
16367 (vld2_lane_s32): Likewise.
16368 (vld2_lane_s64): Likewise.
16369 (vld2_lane_f16): Likewise.
16370 (vld2_lane_f32): Likewise.
16371 (vld2_lane_f64): Likewise.
16372 (vld2_lane_p8): Likewise.
16373 (vld2_lane_p16): Likewise.
16374 (vld2_lane_p64): Likewise.
16375 (vld2q_lane_u8): Likewise.
16376 (vld2q_lane_u16): Likewise.
16377 (vld2q_lane_u32): Likewise.
16378 (vld2q_lane_u64): Likewise.
16379 (vld2q_lane_s8): Likewise.
16380 (vld2q_lane_s16): Likewise.
16381 (vld2q_lane_s32): Likewise.
16382 (vld2q_lane_s64): Likewise.
16383 (vld2q_lane_f16): Likewise.
16384 (vld2q_lane_f32): Likewise.
16385 (vld2q_lane_f64): Likewise.
16386 (vld2q_lane_p8): Likewise.
16387 (vld2q_lane_p16): Likewise.
16388 (vld2q_lane_p64): Likewise.
16389 (vld2_lane_bf16): Likewise.
16390 (vld2q_lane_bf16): Likewise.
16392 2021-08-17 Maxim Kuvyrkov <maxim.kuvyrkov@linaro.org>
16394 * haifa-sched.c (advance_one_cycle): Output more context-synchronization
16397 2021-08-17 Maxim Kuvyrkov <maxim.kuvyrkov@linaro.org>
16399 * haifa-sched.c (enum rfs_decision, rfs_str): Add RFS_AUTOPREF.
16400 (rank_for_schedule): Use it.
16402 2021-08-17 Maxim Kuvyrkov <maxim.kuvyrkov@linaro.org>
16404 PR rtl-optimization/91598
16405 * haifa-sched.c (autopref_rank_for_schedule): Prioritize "irrelevant"
16406 insns after memory reads and before memory writes.
16408 2021-08-17 Alistair_Lee <alistair.lee@arm.com>
16410 * rtl.h (CONST_VECTOR_P): New macro.
16411 * config/aarch64/aarch64.c (aarch64_get_sve_pred_bits): Use RTL
16412 code testing macros.
16413 (aarch64_ptrue_all_mode): Likewise.
16414 (aarch64_expand_mov_immediate): Likewise.
16415 (aarch64_const_vec_all_in_range_p): Likewise.
16416 (aarch64_rtx_costs): Likewise.
16417 (aarch64_legitimate_constant_p): Likewise.
16418 (aarch64_simd_valid_immediate): Likewise.
16419 (aarch64_simd_make_constant): Likewise.
16420 (aarch64_convert_mult_to_shift): Likewise.
16421 (aarch64_expand_sve_vec_perm): Likewise.
16422 (aarch64_vec_fpconst_pow_of_2): Likewise.
16424 2021-08-17 Andrew MacLeod <amacleod@redhat.com>
16426 PR tree-optimization/101938
16427 * range-op.cc (operator_abs::op1_range): Special case
16428 -TYPE_MIN_VALUE for flag_wrapv.
16430 2021-08-17 Kewen Lin <linkw@linux.ibm.com>
16432 * tree-vect-slp.c (vectorizable_bb_reduc_epilogue): Add the cost for
16435 2021-08-17 Jakub Jelinek <jakub@redhat.com>
16437 * tree.def (OMP_SCOPE): New tree code.
16438 * tree.h (OMP_SCOPE_BODY, OMP_SCOPE_CLAUSES): Define.
16439 * tree-nested.c (convert_nonlocal_reference_stmt,
16440 convert_local_reference_stmt, convert_gimple_call): Handle
16442 * tree-pretty-print.c (dump_generic_node): Handle OMP_SCOPE.
16443 * gimple.def (GIMPLE_OMP_SCOPE): New gimple code.
16444 * gimple.c (gimple_build_omp_scope): New function.
16445 (gimple_copy): Handle GIMPLE_OMP_SCOPE.
16446 * gimple.h (gimple_build_omp_scope): Declare.
16447 (gimple_has_substatements): Handle GIMPLE_OMP_SCOPE.
16448 (gimple_omp_scope_clauses, gimple_omp_scope_clauses_ptr,
16449 gimple_omp_scope_set_clauses): New inline functions.
16450 (CASE_GIMPLE_OMP): Add GIMPLE_OMP_SCOPE.
16451 * gimple-pretty-print.c (dump_gimple_omp_scope): New function.
16452 (pp_gimple_stmt_1): Handle GIMPLE_OMP_SCOPE.
16453 * gimple-walk.c (walk_gimple_stmt): Likewise.
16454 * gimple-low.c (lower_stmt): Likewise.
16455 * gimplify.c (is_gimple_stmt): Handle OMP_MASTER.
16456 (gimplify_scan_omp_clauses): For task reductions, handle OMP_SCOPE
16457 like ORT_WORKSHARE constructs. Adjust diagnostics for %<scope%>
16458 allowing task reductions. Reject inscan reductions on scope.
16459 (omp_find_stores_stmt): Handle GIMPLE_OMP_SCOPE.
16460 (gimplify_omp_workshare, gimplify_expr): Handle OMP_SCOPE.
16461 * tree-inline.c (remap_gimple_stmt): Handle GIMPLE_OMP_SCOPE.
16462 (estimate_num_insns): Likewise.
16463 * omp-low.c (build_outer_var_ref): Look through GIMPLE_OMP_SCOPE
16464 contexts if var isn't privatized there.
16465 (check_omp_nesting_restrictions): Handle GIMPLE_OMP_SCOPE.
16466 (scan_omp_1_stmt): Likewise.
16467 (maybe_add_implicit_barrier_cancel): Look through outer
16469 (lower_omp_scope): New function.
16470 (lower_omp_task_reductions): Handle OMP_SCOPE.
16471 (lower_omp_1): Handle GIMPLE_OMP_SCOPE.
16472 (diagnose_sb_1, diagnose_sb_2): Likewise.
16473 * omp-expand.c (expand_omp_single): Support also GIMPLE_OMP_SCOPE.
16474 (expand_omp): Handle GIMPLE_OMP_SCOPE.
16475 (omp_make_gimple_edges): Likewise.
16476 * omp-builtins.def (BUILT_IN_GOMP_SCOPE_START): New built-in.
16478 2021-08-17 Richard Biener <rguenther@suse.de>
16480 PR tree-optimization/101925
16481 * tree-ssa-sccvn.c (copy_reference_ops_from_ref): Set
16482 reverse on COMPONENT_REF and ARRAY_REF according to
16483 what reverse_storage_order_for_component_p does.
16484 (vn_reference_eq): Compare reversed on reference ops.
16485 (reverse_storage_order_for_component_p): New overload.
16486 (vn_reference_lookup_3): Check reverse_storage_order_for_component_p
16487 on the reference looked up.
16489 2021-08-17 Jeff Law <jlaw@localhost.localdomain>
16491 * config/h8300/h8300.c (shift_alg_si): Avoid loops for most SImode
16492 shifts on the H8/S.
16493 (h8300_option_override): Use loops on H8/S more often when optimizing
16495 (get_shift_alg): Handle new "special" cases on H8/S. Simplify
16496 accordingly. Handle various arithmetic right shifts with special
16497 sequences that we couldn't handle before.
16499 2021-08-16 Jeff Law <jlaw@localhost.localdomain>
16501 * config.gcc (rl78-*-elf*): Do not include dbxelf.h.
16503 2021-08-16 Sebastian Huber <sebastian.huber@embedded-brains.de>
16505 * config/sparc/rtemself.h (SPARC_GCOV_TYPE_SIZE): Define.
16506 * config/sparc/sparc.c (sparc_gcov_type_size): New.
16507 (TARGET_GCOV_TYPE_SIZE): Redefine if SPARC_GCOV_TYPE_SIZE is defined.
16508 * coverage.c (get_gcov_type): Use targetm.gcov_type_size().
16509 * doc/tm.texi (TARGET_GCOV_TYPE_SIZE): Add hook under "Misc".
16510 * doc/tm.texi.in: Regenerate.
16511 * target.def (gcov_type_size): New target hook.
16512 * targhooks.c (default_gcov_type_size): New.
16513 * targhooks.h (default_gcov_type_size): Declare.
16514 * tree-profile.c (gimple_gen_edge_profiler): Use precision of
16516 (gimple_gen_time_profiler): Likewise.
16518 2021-08-16 Eric Botcazou <ebotcazou@gcc.gnu.org>
16520 * dwarf2out.c (add_scalar_info): Deal with DW_AT_data_bit_offset.
16522 2021-08-16 Tobias Burnus <tobias@codesourcery.com>
16524 PR middle-end/101931
16525 * omp-low.c (omp_runtime_api_call): Update for routines
16526 added in the meanwhile.
16528 2021-08-16 Martin Liska <mliska@suse.cz>
16530 PR tree-optimization/100393
16531 * tree-switch-conversion.c (group_cluster::dump): Use
16532 get_comparison_count.
16533 (jump_table_cluster::find_jump_tables): Pre-compute number of
16534 comparisons and then decrement it. Cache also max_ratio.
16535 (jump_table_cluster::can_be_handled): Change signature.
16536 * tree-switch-conversion.h (get_comparison_count): New.
16538 2021-08-16 Eric Botcazou <ebotcazou@gcc.gnu.org>
16540 * dwarf2out.c (add_data_member_location_attribute): Use GNAT
16541 encodings only when -fgnat-encodings=all is specified.
16542 (add_bound_info): Likewise.
16543 (add_byte_size_attribute): Likewise.
16544 (gen_member_die): Likewise.
16546 2021-08-16 Thomas Schwinge <thomas@codesourcery.com>
16548 * omp-oacc-neuter-broadcast.cc
16549 (execute_omp_oacc_neuter_broadcast): Plug 'par' memory leak.
16551 2021-08-16 Thomas Schwinge <thomas@codesourcery.com>
16553 * omp-oacc-neuter-broadcast.cc
16554 (execute_omp_oacc_neuter_broadcast): Clarify memory management for
16557 2021-08-16 Thomas Schwinge <thomas@codesourcery.com>
16559 * omp-oacc-neuter-broadcast.cc (field_map): Move variable into...
16560 (execute_omp_oacc_neuter_broadcast): ... here.
16561 (install_var_field, build_receiver_ref, build_sender_ref): Take
16562 'field_map_t *' parameter. Adjust all users.
16563 (worker_single_copy, neuter_worker_single): Take a
16564 'record_field_map_t *' parameter. Adjust all users.
16566 2021-08-16 liuhongt <hongtao.liu@intel.com>
16569 * config/i386/i386.md (ldexp<mode>3): Force operands[1] to
16572 2021-08-16 Martin Liska <mliska@suse.cz>
16575 * multiple_target.c (create_dispatcher_calls): Make default
16576 function local only if it is a definition.
16578 2021-08-16 Martin Liska <mliska@suse.cz>
16581 * ipa-icf-gimple.c (func_checker::compare_ssa_name): Do not
16582 consider equal SSA_NAMEs when one is a param.
16584 2021-08-16 liuhongt <hongtao.liu@intel.com>
16587 * config/i386/i386-expand.c (ix86_expand_vec_perm_vpermt2):
16588 Support vpermi2b for V32QI/V16QImode.
16589 (ix86_extract_perm_from_pool_constant): New function.
16590 (ix86_expand_vec_one_operand_perm_avx512): Support
16591 vpermw/vpermb under TARGET_AVX512BW/TARGET_AVX512VBMI.
16592 (expand_vec_perm_1): Adjust comments for upper.
16593 * config/i386/i386-protos.h (ix86_extract_perm_from_pool_constant):
16595 * config/i386/predicates.md (permvar_truncate_operand): New predicate.
16596 (pshufb_truncv4siv4hi_operand): Ditto.
16597 (pshufb_truncv8hiv8qi_operand): Ditto.
16598 * config/i386/sse.md (*avx512bw_permvar_truncv16siv16hi_1):
16599 New pre_reload define_insn_and_split.
16600 (*avx512f_permvar_truncv8siv8hi_1): Ditto.
16601 (*avx512f_vpermvar_truncv8div8si_1): Ditto.
16602 (*avx512f_permvar_truncv32hiv32qi_1): Ditto.
16603 (*avx512f_permvar_truncv16hiv16qi_1): Ditto.
16604 (*avx512f_permvar_truncv4div4si_1): Ditto.
16605 (*avx512f_pshufb_truncv8hiv8qi_1): Ditto.
16606 (*avx512f_pshufb_truncv4siv4hi_1): Ditto.
16607 (*avx512f_pshufd_truncv2div2si_1): Ditto.
16609 2021-08-16 Kito Cheng <kito.cheng@sifive.com>
16611 * config/riscv/multilib-generator: Support code model option for
16613 * doc/install.texi: Add document of new option for
16614 --with-multilib-generator.
16616 2021-08-15 Clément Chigot <clement.chigot@atos.net>
16618 * config/rs6000/rs6000.c (xcoff_tls_exec_model_detected): New.
16619 (rs6000_legitimize_tls_address_aix): Use it.
16620 (rs6000_xcoff_file_end): Add ".ref __tls_get_addr" when
16621 xcoff_tls_exec_model_detected is true.
16623 2021-08-15 Jeff Law <jlaw@localhost.localdomain>
16625 * config/h8300/h8300.c (shift_alg_si): Retune H8/300H shifts
16626 to allow a bit more code growth, saving many dozens of cycles.
16627 (h8300_option_override): Adjus shift_alg_si if optimizing for
16629 (get_shift_alg): Use special + inline shifts for residuals
16632 2021-08-14 Stafford Horne <shorne@gmail.com>
16635 * config/or1k/or1k-opts.h: New file.
16636 * config/or1k/or1k.c (or1k_legitimize_address_1, print_reloc):
16637 Support generating gotha relocations if -mcmodel=large is
16639 * config/or1k/or1k.h (TARGET_CMODEL_SMALL, TARGET_CMODEL_LARGE):
16641 * config/or1k/or1k.opt (mcmodel=): New option.
16642 * doc/invoke.texi (OpenRISC Options): Document mcmodel.
16644 2021-08-14 Martin Sebor <msebor@redhat.com>
16646 PR middle-end/101791
16647 * gimple-ssa-warn-access.cc (new_delete_mismatch_p): Use new argument
16648 to valid_new_delete_pair_p.
16649 * tree.c (valid_new_delete_pair_p): Add argument.
16650 * tree.h (valid_new_delete_pair_p): Same.
16652 2021-08-14 Jakub Jelinek <jakub@redhat.com>
16655 * config/i386/i386-expand.c (expand_vec_perm_broadcast_1)
16656 <case E_V64QImode>: For this mode assert
16657 !TARGET_AVX512BW || d->perm[0] rather than !TARGET_AVX2 || d->perm[0].
16659 2021-08-13 Michael Meissner <meissner@linux.ibm.com>
16662 * config/rs6000/altivec.md (xxeval): Use register_predicate
16663 instead of altivec_register_predicate.
16665 2021-08-13 Martin Sebor <msebor@redhat.com>
16667 PR middle-end/101734
16668 * tree-ssa-uninit.c (maybe_warn_read_write_only): New function.
16669 (maybe_warn_operand): Call it.
16671 2021-08-13 Martin Liska <mliska@suse.cz>
16674 * attribs.c (decl_attributes): Make naked functions "noipa"
16677 2021-08-13 Martin Liska <mliska@suse.cz>
16680 * symtab.c (symtab_node::noninterposable_alias): Do not create
16681 local aliases for target_clone functions as the clonning pass
16684 2021-08-13 Martin Liska <mliska@suse.cz>
16686 * opts.c (LIVE_PATCHING_OPTION): Define.
16687 (control_options_for_live_patching): Use it in error messages.
16689 2021-08-13 Jan Hubicka <hubicka@ucw.cz>
16691 * ipa-modref.c (dump_eaf_flags): Dump EAF_NOREAD.
16692 (implicit_const_eaf_flags, implicit_pure_eaf_flags,
16693 ignore_stores_eaf_flags): New constants.
16694 (remove_useless_eaf_flags): New function.
16695 (eaf_flags_useful_p): Use it.
16696 (deref_flags): Add EAF_NOT_RETURNED if flag is unused;
16698 (modref_lattice::init): Add EAF_NOREAD.
16699 (modref_lattice::add_escape_point): Do not reacord escape point if
16701 (modref_lattice::merge): EAF_NOESCAPE implies EAF_NODIRECTESCAPE;
16702 use remove_useless_eaf_flags.
16703 (modref_lattice::merge_deref): Use ignore_stores_eaf_flags.
16704 (modref_lattice::merge_direct_load): Add EAF_NOREAD
16705 (analyze_ssa_name_flags): Fix handling EAF_NOT_RETURNED
16706 (analyze_parms): Use remove_useless_eaf_flags.
16707 (ipa_merge_modref_summary_after_inlining): Use ignore_stores_eaf_flags.
16708 (modref_merge_call_site_flags): Add caller and ecf_flags parameter;
16709 use remove_useless_eaf_flags.
16710 (modref_propagate_flags_in_scc): Update.
16711 * ipa-modref.h: Turn eaf_flags_t back to char.
16712 * tree-core.h (EAF_NOT_RETURNED): Fix.
16713 (EAF_NOREAD): New constant
16714 * tree-ssa-alias.c: (ref_maybe_used_by_call_p_1): Check for
16716 * tree-ssa-structalias.c (handle_rhs_call): Handle new flags.
16717 (handle_pure_call): Likewise.
16719 2021-08-12 Jakub Jelinek <jakub@redhat.com>
16721 * tree.def (OMP_MASKED): New tree code.
16722 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_FILTER.
16723 * tree.h (OMP_MASKED_BODY, OMP_MASKED_CLAUSES, OMP_MASKED_COMBINED,
16724 OMP_CLAUSE_FILTER_EXPR): Define.
16725 * tree.c (omp_clause_num_ops): Add OMP_CLAUSE_FILTER entry.
16726 (omp_clause_code_name): Likewise.
16727 (walk_tree_1): Handle OMP_CLAUSE_FILTER.
16728 * tree-nested.c (convert_nonlocal_omp_clauses,
16729 convert_local_omp_clauses): Handle OMP_CLAUSE_FILTER.
16730 (convert_nonlocal_reference_stmt, convert_local_reference_stmt,
16731 convert_gimple_call): Handle GIMPLE_OMP_MASTER.
16732 * tree-pretty-print.c (dump_omp_clause): Handle OMP_CLAUSE_FILTER.
16733 (dump_generic_node): Handle OMP_MASTER.
16734 * gimple.def (GIMPLE_OMP_MASKED): New gimple code.
16735 * gimple.c (gimple_build_omp_masked): New function.
16736 (gimple_copy): Handle GIMPLE_OMP_MASKED.
16737 * gimple.h (gimple_build_omp_masked): Declare.
16738 (gimple_has_substatements): Handle GIMPLE_OMP_MASKED.
16739 (gimple_omp_masked_clauses, gimple_omp_masked_clauses_ptr,
16740 gimple_omp_masked_set_clauses): New inline functions.
16741 (CASE_GIMPLE_OMP): Add GIMPLE_OMP_MASKED.
16742 * gimple-pretty-print.c (dump_gimple_omp_masked): New function.
16743 (pp_gimple_stmt_1): Handle GIMPLE_OMP_MASKED.
16744 * gimple-walk.c (walk_gimple_stmt): Likewise.
16745 * gimple-low.c (lower_stmt): Likewise.
16746 * gimplify.c (is_gimple_stmt): Handle OMP_MASTER.
16747 (gimplify_scan_omp_clauses): Handle OMP_CLAUSE_FILTER. For clauses
16748 that take one expression rather than decl or constant, force
16749 gimplification of that into a SSA_NAME or temporary unless min
16751 (gimplify_adjust_omp_clauses): Handle OMP_CLAUSE_FILTER.
16752 (gimplify_expr): Handle OMP_MASKED.
16753 * tree-inline.c (remap_gimple_stmt): Handle GIMPLE_OMP_MASKED.
16754 (estimate_num_insns): Likewise.
16755 * omp-low.c (scan_sharing_clauses): Handle OMP_CLAUSE_FILTER.
16756 (check_omp_nesting_restrictions): Handle GIMPLE_OMP_MASKED. Adjust
16757 diagnostics for existence of masked construct.
16758 (scan_omp_1_stmt, lower_omp_master, lower_omp_1, diagnose_sb_1,
16759 diagnose_sb_2): Handle GIMPLE_OMP_MASKED.
16760 * omp-expand.c (expand_omp_synch, expand_omp, omp_make_gimple_edges):
16763 2021-08-12 Uroš Bizjak <ubizjak@gmail.com>
16766 * config/i386/i386.md (avx512f_scalef<mode>2): New insn pattern.
16767 (ldexp<mode>3): Use avx512f_scalef<mode>2.
16768 (UNSPEC_SCALEF): Move from ...
16769 * config/i386/sse.md (UNSPEC_SCALEF): ... here.
16771 2021-08-12 Jan Hubicka <hubicka@ucw.cz>
16773 * ipa-split.c (consider_split): Fix condition testing void functions.
16775 2021-08-12 Aldy Hernandez <aldyh@redhat.com>
16777 * doc/invoke.texi: Remove docs for threader-mode param.
16778 * flag-types.h (enum threader_mode): Remove.
16779 * params.opt: Remove threader-mode param.
16780 * tree-ssa-threadbackward.c (class back_threader): Remove
16781 path_is_unreachable_p.
16782 Make find_paths private.
16783 Add maybe_thread and thread_through_all_blocks.
16784 Remove reference marker for m_registry.
16785 Remove reference marker for m_profit.
16786 (back_threader::back_threader): Adjust for registry and profit not
16788 (dump_path): Move down.
16789 (debug): Move down.
16790 (class thread_jumps): Remove.
16791 (class back_threader_registry): Remove m_all_paths.
16793 (thread_jumps::thread_through_all_blocks): Move to back_threader
16795 (fsm_find_thread_path): Remove
16796 (back_threader::maybe_thread): New.
16797 (back_threader::thread_through_all_blocks): Move from
16799 (back_threader_registry::back_threader_registry): Remove
16801 (back_threader_registry::~back_threader_registry): Remove.
16802 (thread_jumps::find_taken_edge): Remove.
16803 (thread_jumps::check_subpath_and_update_thread_path): Remove.
16804 (thread_jumps::maybe_register_path): Remove.
16805 (thread_jumps::handle_phi): Remove.
16806 (handle_assignment_p): Remove.
16807 (thread_jumps::handle_assignment): Remove.
16808 (thread_jumps::fsm_find_control_statement_thread_paths): Remove.
16809 (thread_jumps::find_jump_threads_backwards): Remove.
16810 (thread_jumps::find_jump_threads_backwards_with_ranger): Remove.
16811 (try_thread_blocks): Rename find_jump_threads_backwards to
16813 (pass_early_thread_jumps::execute): Same.
16815 2021-08-12 Tobias Burnus <tobias@codesourcery.com>
16817 * tree-core.h (omp_clause_proc_bind_kind): Add
16818 OMP_CLAUSE_PROC_BIND_PRIMARY.
16819 * tree-pretty-print.c (dump_omp_clause): Add TODO comment to
16820 change 'master' to 'primary' in proc_bind for OpenMP 5.1.
16822 2021-08-12 Claudiu Zissulescu <claziss@synopsys.com>
16824 * common/config/arc/arc-common.c (arc_option_init_struct): Remove
16825 fno-common reference.
16826 * config/arc/arc.c (arc_override_options): Remove overriding of
16829 2021-08-12 Jakub Jelinek <jakub@redhat.com>
16832 * config/i386/i386-expand.c (ix86_expand_vec_one_operand_perm_avx512):
16833 If d->testing_p, return true after performing checks instead of
16834 actually expanding the insn.
16835 (expand_vec_perm_broadcast_1): Handle V32HImode - assert
16836 !TARGET_AVX512BW and return false.
16838 2021-08-12 Eric Botcazou <ebotcazou@gcc.gnu.org>
16840 * configure.ac (PE linker --disable-dynamicbase support): New check.
16841 * configure: Regenerate.
16842 * config.in: Likewise.
16843 * config/i386/mingw32.h (LINK_SPEC_DISABLE_DYNAMICBASE): New define.
16844 (LINK_SPEC): Use it.
16845 * config/i386/mingw-w64.h (LINK_SPEC_DISABLE_DYNAMICBASE): Likewise.
16846 (LINK_SPEC): Likewise.
16848 2021-08-12 liuhongt <hongtao.liu@intel.com>
16851 * config/i386/sse.md (*avx2_zero_extendv16qiv16hi2_2): New
16852 post_reload define_insn_and_split.
16853 (*avx512bw_zero_extendv32qiv32hi2_2): Ditto.
16854 (*sse4_1_zero_extendv8qiv8hi2_4): Ditto.
16855 (*avx512f_zero_extendv16hiv16si2_2): Ditto.
16856 (*avx2_zero_extendv8hiv8si2_2): Ditto.
16857 (*sse4_1_zero_extendv4hiv4si2_4): Ditto.
16858 (*avx512f_zero_extendv8siv8di2_2): Ditto.
16859 (*avx2_zero_extendv4siv4di2_2): Ditto.
16860 (*sse4_1_zero_extendv2siv2di2_4): Ditto.
16861 (VI248_256, VI248_512, VI148_512, VI148_256, VI148_128): New
16864 2021-08-11 Bill Schmidt <wschmidt@linux.ibm.com>
16866 * config/rs6000/rs6000-builtin-new.def: Add always, power5, and
16869 2021-08-11 Bill Schmidt <wschmidt@linux.ibm.com>
16871 * config/rs6000/rs6000-builtin-new.def: Add vsx stanza.
16873 2021-08-11 Bill Schmidt <wschmidt@linux.ibm.com>
16875 * config/rs6000/rs6000-builtin-new.def: Finish altivec stanza.
16876 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Move
16877 initialization of pcvoid_type_node here...
16878 (altivec_init_builtins): ...from here.
16879 * config/rs6000/rs6000.h (rs6000_builtin_type_index): Add
16880 RS6000_BTI_const_ptr_void.
16881 (pcvoid_type_node): New macro.
16883 2021-08-11 Richard Biener <rguenther@suse.de>
16886 * tree-ssa-forwprop.c (pass_forwprop::execute): Do not decompose
16887 hard-register accesses.
16889 2021-08-11 Richard Biener <rguenther@suse.de>
16891 * tree-ssa-operands.c (operands_scanner::get_expr_operands):
16892 Do not look at COMPONENT_REF FIELD_DECLs TREE_THIS_VOLATILE
16893 to determine has_volatile_ops.
16895 2021-08-11 Eric Botcazou <ebotcazou@gcc.gnu.org>
16897 * cfgexpand.c (expand_used_vars): Reuse attribs local variable.
16899 2021-08-11 Jan Hubicka <hubicka@ucw.cz>
16900 Alexandre Oliva <oliva@adacore.com>
16902 * ipa-modref.c (modref_lattice::dump): Fix escape_point's min_flags
16904 (modref_lattice::merge_deref): Fix handling of indirect scape points.
16905 (update_escape_summary_1): Likewise.
16906 (update_escape_summary): Likewise.
16907 (ipa_merge_modref_summary_after_inlining): Likewise.
16909 2021-08-11 Richard Biener <rguenther@suse.de>
16911 PR middle-end/101858
16912 * fold-const.c (fold_binary_loc): Guard simplification
16913 of X < (cast) (1 << Y) to integer types.
16915 2021-08-11 Richard Biener <rguenther@suse.de>
16917 PR tree-optimization/101861
16918 * tree-vect-stmts.c (vectorizable_load): Fix error in
16919 previous change with regard to gather vectorization.
16921 2021-08-11 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
16924 * config/arm/arm_neon.h (vdup_n_s8): Replace call to builtin
16926 (vdup_n_s16): Likewise.
16927 (vdup_n_s32): Likewise.
16928 (vdup_n_s64): Likewise.
16929 (vdup_n_u8): Likewise.
16930 (vdup_n_u16): Likewise.
16931 (vdup_n_u32): Likewise.
16932 (vdup_n_u64): Likewise.
16933 (vdup_n_p8): Likewise.
16934 (vdup_n_p16): Likewise.
16935 (vdup_n_p64): Likewise.
16936 (vdup_n_f16): Likewise.
16937 (vdup_n_f32): Likewise.
16938 (vdupq_n_s8): Likewise.
16939 (vdupq_n_s16): Likewise.
16940 (vdupq_n_s32): Likewise.
16941 (vdupq_n_s64): Likewise.
16942 (vdupq_n_u8): Likewise.
16943 (vdupq_n_u16): Likewise.
16944 (vdupq_n_u32): Likewise.
16945 (vdupq_n_u64): Likewise.
16946 (vdupq_n_p8): Likewise.
16947 (vdupq_n_p16): Likewise.
16948 (vdupq_n_p64): Likewise.
16949 (vdupq_n_f16): Likewise.
16950 (vdupq_n_f32): Likewise.
16951 (vmov_n_s8): Replace call to builtin with call to corresponding
16953 (vmov_n_s16): Likewise.
16954 (vmov_n_s32): Likewise.
16955 (vmov_n_s64): Likewise.
16956 (vmov_n_u8): Likewise.
16957 (vmov_n_u16): Likewise.
16958 (vmov_n_u32): Likewise.
16959 (vmov_n_u64): Likewise.
16960 (vmov_n_p8): Likewise.
16961 (vmov_n_p16): Likewise.
16962 (vmov_n_f16): Likewise.
16963 (vmov_n_f32): Likewise.
16964 (vmovq_n_s8): Likewise.
16965 (vmovq_n_s16): Likewise.
16966 (vmovq_n_s32): Likewise.
16967 (vmovq_n_s64): Likewise.
16968 (vmovq_n_u8): Likewise.
16969 (vmovq_n_u16): Likewise.
16970 (vmovq_n_u32): Likewise.
16971 (vmovq_n_u64): Likewise.
16972 (vmovq_n_p8): Likewise.
16973 (vmovq_n_p16): Likewise.
16974 (vmovq_n_f16): Likewise.
16975 (vmovq_n_f32): Likewise.
16976 * config/arm/arm_neon_builtins.def: Remove entries for vdup_n.
16978 2021-08-11 liuhongt <hongtao.liu@intel.com>
16981 * config/i386/i386.md (ldexp<mode>3): Extend to vscalefs[sd]
16982 when TARGET_AVX512F and TARGET_SSE_MATH.
16984 2021-08-10 Jakub Jelinek <jakub@redhat.com>
16987 * config/i386/i386-expand.c (expand_vec_perm_even_odd): Return false
16988 for V32HImode if !TARGET_AVX512BW.
16989 (ix86_vectorize_vec_perm_const) <case E_V32HImode, case E_V64QImode>:
16990 If !TARGET_AVX512BW and TARGET_AVX512F and d.testing_p, don't fail
16991 early, but actually check the permutation.
16993 2021-08-10 Richard Biener <rguenther@suse.de>
16995 PR tree-optimization/101809
16996 * tree-vect-stmts.c (get_load_store_type): Allow emulated
16997 gathers with offset vector nunits being a constant multiple
16998 of the data vector nunits.
16999 (vect_get_gather_scatter_ops): Use the appropriate nunits
17000 for the offset vector defs.
17001 (vectorizable_store): Adjust call to
17002 vect_get_gather_scatter_ops.
17003 (vectorizable_load): Likewise. Handle the case of less
17004 offset vectors than data vectors.
17006 2021-08-10 Jakub Jelinek <jakub@redhat.com>
17009 * config/i386/sse.md (*avx512f_shuf_<shuffletype>64x2_1<mask_name>_1,
17010 *avx512f_shuf_<shuffletype>32x4_1<mask_name>_1): New define_insn
17013 2021-08-10 Richard Biener <rguenther@suse.de>
17015 PR tree-optimization/101801
17016 PR tree-optimization/101819
17017 * tree-vectorizer.h (vect_emulated_vector_p): Declare.
17018 * tree-vect-loop.c (vect_emulated_vector_p): New function.
17019 (vectorizable_reduction): Re-instantiate a check for emulated
17021 * tree-vect-stmts.c (vectorizable_shift): Likewise.
17022 (vectorizable_operation): Likewise. Cost emulated vector
17023 operations according to the scalar sequence synthesized by
17026 2021-08-10 Richard Biener <rguenther@suse.de>
17028 PR middle-end/101824
17029 * tree-nested.c (get_frame_field): Mark the COMPONENT_REF as
17030 volatile in case the variable was.
17032 2021-08-10 H.J. Lu <hjl.tools@gmail.com>
17035 * config/i386/constraints.md (BC): Document for integer SSE
17036 constant all bits set operand.
17037 (BF): New constraint for const floating-point all bits set
17039 * config/i386/i386.c (standard_sse_constant_p): Likewise.
17040 (standard_sse_constant_opcode): Likewise.
17041 * config/i386/sse.md (sseconstm1): New mode attribute.
17042 (mov<mode>_internal): Replace BC with <sseconstm1>.
17044 2021-08-10 liuhongt <hongtao.liu@intel.com>
17046 * config/i386/sse.md (cond_<insn><mode>): New expander.
17047 (VI248_AVX512VLBW): New mode iterator.
17048 * config/i386/predicates.md
17049 (nonimmediate_or_const_vec_dup_operand): New predicate.
17051 2021-08-09 Andrew MacLeod <amacleod@redhat.com>
17053 PR tree-optimization/101741
17054 * gimple-range-fold.cc (fold_using_range::range_of_builtin_call): Check
17055 type of parameter for toupper/tolower.
17057 2021-08-09 Martin Jambor <mjambor@suse.cz>
17059 PR testsuite/101654
17060 * ipa-prop.c (propagate_controlled_uses): Removed a spurious space.
17062 2021-08-09 Pat Haugen <pthaugen@linux.ibm.com>
17064 * config/rs6000/rs6000.c (is_load_insn1): Verify destination is a
17066 (is_store_insn1): Verify source is a register.
17068 2021-08-09 Uroš Bizjak <ubizjak@gmail.com>
17071 * config/i386/mmx.md (<any_logic:code>v2sf3):
17072 Rename from *mmx_<any_logic:code>v2sf3
17074 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
17076 * config/nvptx/nvptx.c: Cross-reference parts adapted in
17077 'gcc/omp-oacc-neuter-broadcast.cc'.
17078 * omp-low.c: Likewise.
17079 * omp-oacc-neuter-broadcast.cc: Cross-reference parts adapted from
17082 2021-08-09 Julian Brown <julian@codesourcery.com>
17083 Kwok Cheung Yeung <kcy@codesourcery.com>
17084 Thomas Schwinge <thomas@codesourcery.com>
17086 * config/gcn/gcn.c (gcn_init_builtins): Override decls for
17087 BUILT_IN_GOACC_SINGLE_START, BUILT_IN_GOACC_SINGLE_COPY_START,
17088 BUILT_IN_GOACC_SINGLE_COPY_END and BUILT_IN_GOACC_BARRIER.
17089 (gcn_goacc_validate_dims): Turn on worker partitioning unconditionally.
17090 (gcn_fork_join): Update comment.
17091 * config/gcn/gcn.opt (flag_worker_partitioning): Remove.
17092 (macc_experimental_workers): Remove unused option.
17094 2021-08-09 Julian Brown <julian@codesourcery.com>
17095 Nathan Sidwell <nathan@codesourcery.com> (via 'gcc/config/nvptx/nvptx.c' master)
17096 Kwok Cheung Yeung <kcy@codesourcery.com>
17097 Thomas Schwinge <thomas@codesourcery.com>
17099 * Makefile.in (OBJS): Add omp-oacc-neuter-broadcast.o.
17100 * doc/tm.texi.in (TARGET_GOACC_CREATE_WORKER_BROADCAST_RECORD):
17101 Add documentation hook.
17102 * doc/tm.texi: Regenerate.
17103 * omp-oacc-neuter-broadcast.cc: New file.
17104 * omp-builtins.def (BUILT_IN_GOACC_BARRIER)
17105 (BUILT_IN_GOACC_SINGLE_START, BUILT_IN_GOACC_SINGLE_COPY_START)
17106 (BUILT_IN_GOACC_SINGLE_COPY_END): New builtins.
17107 * passes.def (pass_omp_oacc_neuter_broadcast): Add pass.
17108 * target.def (goacc.create_worker_broadcast_record): Add target
17110 * tree-pass.h (make_pass_omp_oacc_neuter_broadcast): Add
17112 * config/gcn/gcn-protos.h (gcn_goacc_adjust_propagation_record):
17113 Rename prototype to...
17114 (gcn_goacc_create_worker_broadcast_record): ... this.
17115 * config/gcn/gcn-tree.c (gcn_goacc_adjust_propagation_record): Rename
17117 (gcn_goacc_create_worker_broadcast_record): ... this.
17118 * config/gcn/gcn.c (TARGET_GOACC_ADJUST_PROPAGATION_RECORD):
17120 (TARGET_GOACC_CREATE_WORKER_BROADCAST_RECORD): ... this.
17122 2021-08-09 Tejas Belagod <tejas.belagod@arm.com>
17125 * config/aarch64/aarch64-simd.md (vlshr<mode>3, vashr<mode>3): Use
17126 the right iterator.
17128 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
17130 * Makefile.in (GTFILES): Remove '$(srcdir)/omp-offload.c'.
17132 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
17134 * builtins.def (DEF_GOACC_BUILTIN, DEF_GOMP_BUILTIN): Don't
17135 consider '-foffload-abi'.
17136 * common.opt (-foffload-abi): Remove 'Var', 'Init'.
17137 * opts.c (common_handle_option) <-foffload-abi> [ACCEL_COMPILER]:
17140 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
17142 * optc-gen.awk: Sanity check that 'Init' doesn't appear without
17145 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
17147 * omp-builtins.def (BUILT_IN_ACC_GET_DEVICE_TYPE): Remove.
17149 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
17151 * doc/gty.texi (Files): Update.
17153 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
17155 * doc/gty.texi (Files): Fix GTY header file example.
17157 2021-08-09 Roger Sayle <roger@nextmovesoftware.com>
17159 * tree-ssa-ccp.c (value_mask_to_min_max): Helper function to
17160 determine the upper and lower bounds from a mask-value pair.
17161 (bit_value_unop) [ABS_EXPR, ABSU_EXPR]: Add support for
17162 absolute value and unsigned absolute value expressions.
17163 (bit_value_binop): Initialize *VAL's precision.
17164 [LT_EXPR, LE_EXPR]: Use value_mask_to_min_max to determine
17165 upper and lower bounds of operands. Add LE_EXPR/GE_EXPR
17166 support when the operands are unknown but potentially equal.
17167 [MIN_EXPR, MAX_EXPR]: Support minimum/maximum expressions.
17169 2021-08-09 Bin Cheng <bin.cheng@linux.alibaba.com>
17171 * config/aarch64/aarch64.md
17172 (*extend<SHORT:mode><GPI:mode>2_aarch64): Use %<GPI:w>0.
17174 2021-08-08 Sergei Trofimovich <siarheit@google.com>
17176 * lra-constraints.c: Fix s/otput/output/ typo.
17178 2021-08-06 Martin Sebor <msebor@redhat.com>
17180 * builtins.c (expand_builtin_memchr): Move to gimple-ssa-warn-access.cc.
17181 (expand_builtin_strcat): Same.
17182 (expand_builtin_stpncpy): Same.
17183 (expand_builtin_strncat): Same.
17184 (check_read_access): Same.
17185 (check_memop_access): Same.
17186 (expand_builtin_strlen): Move checks to gimple-ssa-warn-access.cc.
17187 (expand_builtin_strnlen): Same.
17188 (expand_builtin_memcpy): Same.
17189 (expand_builtin_memmove): Same.
17190 (expand_builtin_mempcpy): Same.
17191 (expand_builtin_strcpy): Same.
17192 (expand_builtin_strcpy_args): Same.
17193 (expand_builtin_stpcpy_1): Same.
17194 (expand_builtin_strncpy): Same.
17195 (expand_builtin_memset): Same.
17196 (expand_builtin_bzero): Same.
17197 (expand_builtin_strcmp): Same.
17198 (expand_builtin_strncmp): Same.
17199 (expand_builtin): Remove handlers.
17200 (fold_builtin_strlen): Add a comment.
17201 * builtins.h (check_access): Move to gimple-ssa-warn-access.cc.
17202 * calls.c (maybe_warn_nonstring_arg): Same.
17203 * diagnostic-spec.c (nowarn_spec_t::nowarn_spec_t): Add warning option.
17204 * gimple-fold.c (gimple_fold_builtin_strcpy): Pass argument to callee.
17205 (gimple_fold_builtin_stpcpy): Same.
17206 * gimple-ssa-warn-access.cc (has_location): New function.
17207 (get_location): Same.
17208 (get_callee_fndecl): Same.
17209 (call_nargs): Same.
17211 (warn_string_no_nul): Define.
17212 (unterminated_array): Same.
17213 (check_nul_terminated_array): Same.
17214 (maybe_warn_nonstring_arg): Same.
17215 (maybe_warn_for_bound): Same.
17216 (warn_for_access): Same.
17217 (check_access): Same.
17218 (check_memop_access): Same.
17219 (check_read_access): Same.
17220 (warn_dealloc_offset): Use helper functions.
17221 (maybe_emit_free_warning): Same.
17222 (class pass_waccess): Add members.
17223 (check_strcat): New function.
17224 (check_strncat): New function.
17225 (check_stxcpy): New function.
17226 (check_stxncpy): New function.
17227 (check_strncmp): New function.
17228 (pass_waccess::check_builtin): New function.
17229 (pass_waccess::check): Call it.
17230 * gimple-ssa-warn-access.h (warn_string_no_nul): Move here from
17232 (maybe_warn_for_bound): Same.
17233 (check_access): Same.
17234 (check_memop_access): Same.
17235 (check_read_access): Same.
17236 * pointer-query.h (struct access_data): Define a ctor overload.
17238 2021-08-06 Richard Biener <rguenther@suse.de>
17240 PR tree-optimization/101801
17241 * tree-vectorizer.h (vect_worthwhile_without_simd_p): Rename...
17242 (vect_can_vectorize_without_simd_p): ... to this.
17243 * tree-vect-loop.c (vect_worthwhile_without_simd_p): Rename...
17244 (vect_can_vectorize_without_simd_p): ... to this and fold
17245 in vect_min_worthwhile_factor.
17246 (vect_min_worthwhile_factor): Remove.
17247 (vectorizable_reduction): Adjust and remove the cost part.
17248 * tree-vect-stmts.c (vectorizable_shift): Likewise.
17249 (vectorizable_operation): Likewise.
17251 2021-08-06 Uroš Bizjak <ubizjak@gmail.com>
17254 * config/i386/i386.md (cmove reg-to-reg move elimination peephole2s):
17255 Add general_gr_operand predicate to operand 3.
17257 2021-08-06 Roger Sayle <roger@nextmovesoftware.com>
17259 * tree-ssa-phiopt.c (cond_removal_in_builtin_zero_pattern): Use
17260 CFN_BUILT_IN_CLRSB* instead of BUILT_IN_CLRSB* for consistency.
17262 2021-08-06 Tamar Christina <tamar.christina@arm.com>
17264 * config/aarch64/aarch64-sve-builtins.cc (register_svpattern,
17265 register_svprfop): Pass vec<> by pointer.
17266 * langhooks-def.h (lhd_simulate_enum_decl): Likewise.
17267 * langhooks.c (lhd_simulate_enum_decl): Likewise.
17268 * langhooks.h (struct lang_hooks_for_types): Likewise.
17270 2021-08-06 Jonathan Wright <jonathan.wright@arm.com>
17272 * config/aarch64/arm_neon.h (vst1_bf16_x2): Use
17273 __builtin_memcpy instead of constructing an additional
17274 __builtin_aarch64_simd_oi one vector at a time.
17275 (vst1q_bf16_x2): Likewise.
17276 (vst1_bf16_x3): Use __builtin_memcpy instead of constructing
17277 an additional __builtin_aarch64_simd_ci one vector at a time.
17278 (vst1q_bf16_x3): Likewise.
17279 (vst1_bf16_x4): Use __builtin_memcpy instead of a union.
17280 (vst1q_bf16_x4): Likewise.
17281 (vst2_bf16): Use __builtin_memcpy instead of constructing an
17282 additional __builtin_aarch64_simd_oi one vector at a time.
17283 (vst2q_bf16): Likewise.
17284 (vst3_bf16): Use __builtin_memcpy instead of constructing an
17285 additional __builtin_aarch64_simd_ci mode one vector at a
17287 (vst3q_bf16): Likewise.
17288 (vst4_bf16): Use __builtin_memcpy instead of constructing an
17289 additional __builtin_aarch64_simd_xi one vector at a time.
17290 (vst4q_bf16): Likewise.
17292 2021-08-06 Jonathan Wright <jonathan.wright@arm.com>
17294 * config/aarch64/arm_neon.h (__ST2_LANE_FUNC): Delete.
17295 (__ST2Q_LANE_FUNC): Delete.
17296 (vst2_lane_f16): Use __builtin_memcpy to copy vector
17297 structure instead of constructing __builtin_aarch64_simd_oi
17298 one vector at a time.
17299 (vst2_lane_f32): Likewise.
17300 (vst2_lane_f64): Likewise.
17301 (vst2_lane_p8): Likewise.
17302 (vst2_lane_p16): Likewise.
17303 (vst2_lane_p64): Likewise.
17304 (vst2_lane_s8): Likewise.
17305 (vst2_lane_s16): Likewise.
17306 (vst2_lane_s32): Likewise.
17307 (vst2_lane_s64): Likewise.
17308 (vst2_lane_u8): Likewise.
17309 (vst2_lane_u16): Likewise.
17310 (vst2_lane_u32): Likewise.
17311 (vst2_lane_u64): Likewise.
17312 (vst2_lane_bf16): Likewise.
17313 (vst2q_lane_f16): Use __builtin_memcpy to copy vector
17314 structure instead of using a union.
17315 (vst2q_lane_f32): Likewise.
17316 (vst2q_lane_f64): Likewise.
17317 (vst2q_lane_p8): Likewise.
17318 (vst2q_lane_p16): Likewise.
17319 (vst2q_lane_p64): Likewise.
17320 (vst2q_lane_s8): Likewise.
17321 (vst2q_lane_s16): Likewise.
17322 (vst2q_lane_s32): Likewise.
17323 (vst2q_lane_s64): Likewise.
17324 (vst2q_lane_u8): Likewise.
17325 (vst2q_lane_u16): Likewise.
17326 (vst2q_lane_u32): Likewise.
17327 (vst2q_lane_u64): Likewise.
17328 (vst2q_lane_bf16): Likewise.
17330 2021-08-06 Jonathan Wright <jonathan.wright@arm.com>
17332 * config/aarch64/arm_neon.h (__ST3_LANE_FUNC): Delete.
17333 (__ST3Q_LANE_FUNC): Delete.
17334 (vst3_lane_f16): Use __builtin_memcpy to copy vector
17335 structure instead of constructing __builtin_aarch64_simd_ci
17336 one vector at a time.
17337 (vst3_lane_f32): Likewise.
17338 (vst3_lane_f64): Likewise.
17339 (vst3_lane_p8): Likewise.
17340 (vst3_lane_p16): Likewise.
17341 (vst3_lane_p64): Likewise.
17342 (vst3_lane_s8): Likewise.
17343 (vst3_lane_s16): Likewise.
17344 (vst3_lane_s32): Likewise.
17345 (vst3_lane_s64): Likewise.
17346 (vst3_lane_u8): Likewise.
17347 (vst3_lane_u16): Likewise.
17348 (vst3_lane_u32): Likewise.
17349 (vst3_lane_u64): Likewise.
17350 (vst3_lane_bf16): Likewise.
17351 (vst3q_lane_f16): Use __builtin_memcpy to copy vector
17352 structure instead of using a union.
17353 (vst3q_lane_f32): Likewise.
17354 (vst3q_lane_f64): Likewise.
17355 (vst3q_lane_p8): Likewise.
17356 (vst3q_lane_p16): Likewise.
17357 (vst3q_lane_p64): Likewise.
17358 (vst3q_lane_s8): Likewise.
17359 (vst3q_lane_s16): Likewise.
17360 (vst3q_lane_s32): Likewise.
17361 (vst3q_lane_s64): Likewise.
17362 (vst3q_lane_u8): Likewise.
17363 (vst3q_lane_u16): Likewise.
17364 (vst3q_lane_u32): Likewise.
17365 (vst3q_lane_u64): Likewise.
17366 (vst3q_lane_bf16): Likewise.
17368 2021-08-06 Jonathan Wright <jonathan.wright@arm.com>
17370 * config/aarch64/arm_neon.h (__ST4_LANE_FUNC): Delete.
17371 (__ST4Q_LANE_FUNC): Delete.
17372 (vst4_lane_f16): Use __builtin_memcpy to copy vector
17373 structure instead of constructing __builtin_aarch64_simd_xi
17374 one vector at a time.
17375 (vst4_lane_f32): Likewise.
17376 (vst4_lane_f64): Likewise.
17377 (vst4_lane_p8): Likewise.
17378 (vst4_lane_p16): Likewise.
17379 (vst4_lane_p64): Likewise.
17380 (vst4_lane_s8): Likewise.
17381 (vst4_lane_s16): Likewise.
17382 (vst4_lane_s32): Likewise.
17383 (vst4_lane_s64): Likewise.
17384 (vst4_lane_u8): Likewise.
17385 (vst4_lane_u16): Likewise.
17386 (vst4_lane_u32): Likewise.
17387 (vst4_lane_u64): Likewise.
17388 (vst4_lane_bf16): Likewise.
17389 (vst4q_lane_f16): Use __builtin_memcpy to copy vector
17390 structure instead of using a union.
17391 (vst4q_lane_f32): Likewise.
17392 (vst4q_lane_f64): Likewise.
17393 (vst4q_lane_p8): Likewise.
17394 (vst4q_lane_p16): Likewise.
17395 (vst4q_lane_p64): Likewise.
17396 (vst4q_lane_s8): Likewise.
17397 (vst4q_lane_s16): Likewise.
17398 (vst4q_lane_s32): Likewise.
17399 (vst4q_lane_s64): Likewise.
17400 (vst4q_lane_u8): Likewise.
17401 (vst4q_lane_u16): Likewise.
17402 (vst4q_lane_u32): Likewise.
17403 (vst4q_lane_u64): Likewise.
17404 (vst4q_lane_bf16): Likewise.
17406 2021-08-06 Martin Liska <mliska@suse.cz>
17408 * config/rs6000/rs6000.c (rs6000_option_override_internal): When
17409 a target option is restored, it can have
17410 rs6000_long_double_type_size set to FLOAT_PRECISION_TFmode
17411 and error should not be emitted.
17413 2021-08-06 Sebastian Huber <sebastian.huber@embedded-brains.de>
17415 * gcov-io.h (gcov_write): Declare.
17416 * gcov-io.c (gcov_write): New.
17417 (gcov_write_counter): Remove.
17418 (gcov_write_tag_length): Likewise.
17419 (gcov_write_summary): Replace gcov_write_tag_length() with calls to
17420 gcov_write_unsigned().
17421 * doc/invoke.texi (fprofile-info-section): Mention
17422 __gcov_info_to_gdca().
17424 2021-08-06 Martin Sebor <msebor@redhat.com>
17426 * dominance.c (prune_bbs_to_update_dominators): Adjust by-value vec
17427 arguments to by-reference.
17428 (iterate_fix_dominators): Same.
17429 * dominance.h (iterate_fix_dominators): Same.
17430 * ipa-prop.h: Call auto_vec::to_vec_legacy.
17431 * tree-data-ref.c (dump_data_dependence_relation): Adjust by-value vec
17432 arguments to by-reference.
17433 (debug_data_dependence_relation): Same.
17434 (dump_data_dependence_relations): Same.
17435 * tree-data-ref.h (debug_data_dependence_relation): Same.
17436 (dump_data_dependence_relations): Same.
17437 * tree-predcom.c (dump_chains): Same.
17438 (initialize_root_vars_lm): Same.
17439 (determine_unroll_factor): Same.
17440 (replace_phis_by_defined_names): Same.
17441 (insert_init_seqs): Same.
17442 (pcom_worker::tree_predictive_commoning_loop): Call
17443 auto_vec::to_vec_legacy.
17444 * tree-ssa-pre.c (insert_into_preds_of_block): Adjust by-value vec
17445 arguments to by-reference.
17446 * tree-ssa-threadbackward.c (populate_worklist): Same.
17447 (back_threader::resolve_def): Same.
17448 * tree-vect-data-refs.c (vect_check_nonzero_value): Same.
17449 (vect_enhance_data_refs_alignment): Same.
17450 (vect_check_lower_bound): Same.
17451 (vect_prune_runtime_alias_test_list): Same.
17452 (vect_permute_store_chain): Same.
17453 * tree-vect-slp-patterns.c (vect_normalize_conj_loc): Same.
17454 * tree-vect-stmts.c (vect_create_vectorized_demotion_stmts): Same.
17455 * tree-vectorizer.h (vect_permute_store_chain): Same.
17456 * vec.c (test_init): New function.
17457 (vec_c_tests): Call new function.
17458 * vec.h (vec): Declare ctors, dtor, and assignment.
17459 (auto_vec::vec_to_legacy): New function.
17460 (vec::copy): Adjust initialization.
17462 2021-08-05 H.J. Lu <hjl.tools@gmail.com>
17465 * config/i386/i386.c (ix86_can_inline_p): Ignore MASK_80387 if
17466 callee only uses GPRs.
17467 * config/i386/ia32intrin.h: Revert commit 5463cee2770.
17468 * config/i386/serializeintrin.h: Revert commit 71958f740f1.
17469 * config/i386/x86gprintrin.h: Add
17470 #pragma GCC target("general-regs-only") and #pragma GCC pop_options
17471 to disable non-GPR ISAs.
17473 2021-08-05 Richard Sandiford <richard.sandiford@arm.com>
17475 PR middle-end/101787
17476 * doc/md.texi (cond_ashl, cond_ashr, cond_lshr): Document.
17478 2021-08-05 Richard Sandiford <richard.sandiford@arm.com>
17480 * tree-vectorizer.h (vect_is_store_elt_extraction, vect_is_reduction)
17481 (vect_reduc_type, vect_embedded_comparison_type, vect_comparison_type)
17482 (vect_is_extending_load, vect_is_integer_truncation): New functions,
17483 moved from aarch64.c but given different names.
17484 * config/aarch64/aarch64.c (aarch64_is_store_elt_extraction)
17485 (aarch64_is_reduction, aarch64_reduc_type)
17486 (aarch64_embedded_comparison_type, aarch64_comparison_type)
17487 (aarch64_extending_load_p, aarch64_integer_truncation_p): Delete
17488 in favor of the above. Update callers accordingly.
17490 2021-08-05 Richard Earnshaw <rearnsha@arm.com>
17493 * config/arm/arm-cpus.in (generic-armv7-a): Add quirk to suppress
17494 writing .cpu directive in asm output.
17495 * config/arm/arm.c (arm_identify_fpu_from_isa): New variable.
17496 (arm_last_printed_arch_string): Delete.
17497 (arm_last-printed_fpu_string): Delete.
17498 (arm_configure_build_target): If use of floating-point/SIMD is
17499 disabled, remove all fp/simd related features from the target ISA.
17500 (last_arm_targ_options): New variable.
17501 (arm_print_asm_arch_directives): Add new parameters. Change order
17502 of emitted directives and handle all cases here.
17503 (arm_file_start): Always call arm_print_asm_arch_directives, move
17504 all generation of .arch/.arch_extension here.
17505 (arm_file_end): Call arm_print_asm_arch.
17506 (arm_declare_function_name): Call arm_print_asm_arch_directives
17507 instead of printing .arch/.fpu directives directly.
17509 2021-08-05 Richard Earnshaw <rearnsha@arm.com>
17511 * config/arm/arm.c (arm_configure_build_target): Don't call
17512 arm_option_reconfigure_globals.
17513 (arm_option_restore): Call arm_option_reconfigure_globals after
17514 reconfiguring the target.
17515 * config/arm/arm-c.c (arm_pragma_target_parse): Likewise.
17517 2021-08-05 Richard Earnshaw <rearnsha@arm.com>
17519 * config/arm/arm.c (arm_configure_build_target): Ensure the target's
17520 arch_name is always set.
17522 2021-08-05 Jonathan Wright <jonathan.wright@arm.com>
17524 * config/aarch64/aarch64.c: Traverse RTL tree to prevent cost
17525 of vec_select high-half from being added into Neon subtract
17528 2021-08-05 Jonathan Wright <jonathan.wright@arm.com>
17530 * config/aarch64/aarch64.c: Traverse RTL tree to prevent cost
17531 of vec_select high-half from being added into Neon add cost.
17533 2021-08-05 Kewen Lin <linkw@linux.ibm.com>
17535 * cfgloop.h (loops_list::loops_list): Add one optional argument
17536 root and adjust accordingly, update loop tree walking and factor
17538 * cfgloop.c (loops_list::walk_loop_tree): ... this. New function.
17540 2021-08-05 Eric Botcazou <ebotcazou@gcc.gnu.org>
17542 PR tree-optimization/101626
17543 * tree-sra.c (propagate_subaccesses_from_rhs): Do not set the
17544 reverse scalar storage order on a pointer or vector component.
17546 2021-08-05 liuhongt <hongtao.liu@intel.com>
17548 * config/i386/sse.md (cond_<code><mode>): New expander.
17550 2021-08-05 liuhongt <hongtao.liu@intel.com>
17552 * config/i386/sse.md (cond_<code><mode>): New expander.
17554 2021-08-05 liuhongt <hongtao.liu@intel.com>
17556 * config/i386/sse.md (cond_<code><mode>): New expander.
17558 2021-08-04 David Malcolm <dmalcolm@redhat.com>
17561 * Makefile.in (ANALYZER_OBJS): Add analyzer/region-model-asm.o.
17563 2021-08-04 H.J. Lu <hjl.tools@gmail.com>
17566 * config/i386/i386.h (STORE_MAX_PIECES): Allow 16/32/64 bytes
17567 only if TARGET_INTER_UNIT_MOVES_TO_VEC is true.
17569 2021-08-04 H.J. Lu <hjl.tools@gmail.com>
17572 * config/i386/i386-expand.c (ix86_expand_vector_move): Call
17573 ix86_gen_scratch_sse_rtx to get a scratch SSE register to copy
17574 data with SSE register from one memory location to another.
17576 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
17578 * config/s390/s390.c (expand_perm_with_vpdi): New function.
17579 (vectorize_vec_perm_const_1): Call expand_perm_with_vpdi.
17580 * config/s390/vector.md (*vpdi1<mode>, @vpdi1<mode>): Enable a
17581 parameterized expander.
17582 (*vpdi4<mode>, @vpdi4<mode>): Likewise.
17584 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
17586 * config/s390/s390.c (MAX_VECT_LEN): Define macro.
17587 (struct expand_vec_perm_d): Define struct.
17588 (expand_perm_with_merge): New function.
17589 (vectorize_vec_perm_const_1): New function.
17590 (s390_vectorize_vec_perm_const): New function.
17591 (TARGET_VECTORIZE_VEC_PERM_CONST): Define target macro.
17593 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
17595 * config/s390/vector.md (V_HW_64): Remove mode iterator.
17596 (*vec_load_pair<mode>): Use V_HW_2 instead of V_HW_64.
17597 * config/s390/vx-builtins.md
17598 (vec_scatter_element<V_HW_2:mode>_SI): Use V_HW_2 instead of
17601 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
17603 * config/s390/s390.md (UNSPEC_VEC_PERMI): Remove constant
17605 * config/s390/vector.md (*vpdi1<mode>, *vpdi4<mode>): New pattern
17607 * config/s390/vx-builtins.md (*vec_permi<mode>): Emit generic rtx
17608 instead of an unspec.
17610 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
17612 * config/s390/s390-modes.def: Add more vector modes to support
17613 concatenation of two vectors.
17614 * config/s390/s390-protos.h (s390_expand_merge_perm_const): Add
17616 (s390_expand_merge): Likewise.
17617 * config/s390/s390.c (s390_expand_merge_perm_const): New function.
17618 (s390_expand_merge): New function.
17619 * config/s390/s390.md (UNSPEC_VEC_MERGEH, UNSPEC_VEC_MERGEL):
17620 Remove constant definitions.
17621 * config/s390/vector.md (V_HW_2): Add mode iterators.
17622 (VI_HW_4, V_HW_4): Rename VI_HW_4 to V_HW_4.
17623 (vec_2x_nelts, vec_2x_wide): New mode attributes.
17624 (*vmrhb, *vmrlb, *vmrhh, *vmrlh, *vmrhf, *vmrlf, *vmrhg, *vmrlg):
17625 New pattern definitions.
17626 (vec_widen_umult_lo_<mode>, vec_widen_umult_hi_<mode>)
17627 (vec_widen_smult_lo_<mode>, vec_widen_smult_hi_<mode>)
17628 (vec_unpacks_lo_v4sf, vec_unpacks_hi_v4sf, vec_unpacks_lo_v2df)
17629 (vec_unpacks_hi_v2df): Adjust expanders to emit non-unspec RTX for
17631 * config/s390/vx-builtins.md (V_HW_4): Remove mode iterator. Now
17633 (vec_mergeh<mode>, vec_mergel<mode>): Use s390_expand_merge to
17634 emit vec merge pattern.
17636 2021-08-04 Jonathan Wright <jonathan.wright@arm.com>
17638 * config/aarch64/aarch64.c (aarch64_strip_extend_vec_half):
17640 (aarch64_rtx_mult_cost): Traverse RTL tree to prevent cost of
17641 vec_select high-half from being added into Neon multiply
17643 * rtlanal.c (vec_series_highpart_p): Define.
17644 * rtlanal.h (vec_series_highpart_p): Declare.
17646 2021-08-04 Jonathan Wright <jonathan.wright@arm.com>
17648 * config/aarch64/aarch64.c (aarch64_strip_duplicate_vec_elt):
17650 (aarch64_rtx_mult_cost): Traverse RTL tree to prevent
17651 vec_select cost from being added into Neon multiply cost.
17653 2021-08-04 Richard Sandiford <richard.sandiford@arm.com>
17655 * tree-vect-loop.c (vect_better_loop_vinfo_p): Detect cases in
17656 which old_loop_vinfo is an epilogue loop that handles a constant
17657 number of iterations.
17659 2021-08-04 Richard Sandiford <richard.sandiford@arm.com>
17661 * tree-vect-loop.c (vect_analyze_loop): Print a dump message
17662 when a reanalyzed loop fails to be cheaper than the current
17665 2021-08-04 Richard Sandiford <richard.sandiford@arm.com>
17667 * config/aarch64/aarch64.c: Fix a typo.
17669 2021-08-04 Vincent Lefèvre <vincent-gcc@vinc17.net>
17671 PR gcov-profile/101773
17672 * gcov-io.c (gcov_close): Check return code of a fclose.
17674 2021-08-04 Bernd Edlinger <bernd.edlinger@hotmail.de>
17677 * dwarf2out.c (dwarf2out_assembly_start): Emit a dummy
17678 .file statement when needed.
17680 2021-08-04 Richard Biener <rguenther@suse.de>
17682 * tree-vect-data-refs.c (vect_check_gather_scatter):
17683 Include widening conversions only when the result is
17684 still handed by native gather or the current offset
17685 size not already matches the data size.
17686 Also succeed analysis in case there's no native support,
17687 noted by a IFN_LAST ifn and a NULL decl.
17688 (vect_analyze_data_refs): Always consider gathers.
17689 * tree-vect-patterns.c (vect_recog_gather_scatter_pattern):
17690 Test for no IFN gather rather than decl gather.
17691 * tree-vect-stmts.c (vect_model_load_cost): Pass in the
17692 gather-scatter info and cost emulated gathers accordingly.
17693 (vect_truncate_gather_scatter_offset): Properly test for
17695 (vect_use_strided_gather_scatters_p): Likewise.
17696 (get_load_store_type): Handle emulated gathers and its
17698 (vectorizable_load): Likewise. Emulate them by extracting
17699 scalar offsets, doing scalar loads and a vector construct.
17701 2021-08-04 H.J. Lu <hjl.tools@gmail.com>
17704 * expr.c (op_by_pieces_d::op_by_pieces_d): Add a max_pieces
17705 argument to set m_max_size.
17706 (move_by_pieces_d): Pass MOVE_MAX_PIECES to op_by_pieces_d.
17707 (store_by_pieces_d): Pass STORE_MAX_PIECES to op_by_pieces_d.
17708 (compare_by_pieces_d): Pass COMPARE_MAX_PIECES to op_by_pieces_d.
17710 2021-08-04 Roger Sayle <roger@nextmovesoftware.com>
17711 Marc Glisse <marc.glisse@inria.fr>
17713 * match.pd (bit_ior, bit_xor): Canonicalize (X*C1)|(X*C2) and
17714 (X*C1)^(X*C2) as X*(C1+C2), and related variants, using
17715 tree_nonzero_bits to ensure that operands are bit-wise disjoint.
17717 2021-08-04 Richard Biener <rguenther@suse.de>
17719 * tree-ssa-forwprop.c (pass_forwprop::execute): Split
17720 out code to decompose vector loads ...
17721 (optimize_vector_load): ... here. Generalize it to
17722 handle intermediate widening and TARGET_MEM_REF loads
17723 and apply it to loads with a supported vector mode as well.
17725 2021-08-04 Richard Biener <rguenther@suse.de>
17727 PR tree-optimization/101756
17728 * tree-vect-slp.c (vectorizable_bb_reduc_epilogue): Make sure
17729 the result of the reduction epilogue is compatible to the original
17732 2021-08-04 liuhongt <hongtao.liu@intel.com>
17735 * config/i386/i386.md (peephole2): Refine predicate from
17736 register_operand to general_reg_operand.
17738 2021-08-04 Aldy Hernandez <aldyh@redhat.com>
17740 * gimple-range-path.h (path_range_query::dump): Mark override.
17742 2021-08-04 Richard Biener <rguenther@suse.de>
17744 PR tree-optimization/101769
17745 * tree-tailcall.c (eliminate_tail_call): Add the created loop
17746 for the first recursion and return it via the new output parameter.
17747 (optimize_tail_call): Pass through new output param.
17748 (tree_optimize_tail_calls_1): After creating all latches,
17749 add the created loop to the loop tree. Do not mark loops for fixup.
17751 2021-08-04 Martin Liska <mliska@suse.cz>
17753 * doc/invoke.texi: Document threader-mode param.
17755 2021-08-04 liuhongt <hongtao.liu@intel.com>
17757 * config/i386/sse.md (cond_fma<mode>): New expander.
17758 (cond_fms<mode>): Ditto.
17759 (cond_fnma<mode>): Ditto.
17760 (cond_fnms<mode>): Ditto.
17762 2021-08-03 Segher Boessenkool <segher@kernel.crashing.org>
17764 * config/rs6000/vsx.md (*vsx_le_perm_store_<mode>): Use && instead of &.
17766 2021-08-03 Segher Boessenkool <segher@kernel.crashing.org>
17768 * config/rs6000/constraints.md: Remove "e" from the list of available
17769 constraint characters.
17771 2021-08-03 Eugene Rozenfeld <erozen@microsoft.com>
17773 PR gcov-profile/71672
17774 * auto-profile.c (afdo_indirect_call): Fix setup of the historgram value for indirect calls.
17776 2021-08-03 Paul A. Clarke <pc@us.ibm.com>
17778 * config/rs6000/smmintrin.h (_mm_minpos_epu16): New.
17780 2021-08-03 H.J. Lu <hjl.tools@gmail.com>
17782 * config/i386/i386.c (ix86_gen_scratch_sse_rtx): In 64-bit mode,
17783 try XMM31 to avoid vzeroupper.
17785 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
17787 * doc/invoke.texi: Document -mtune=neoverse-512tvb and
17788 -mcpu=neoverse-512tvb.
17789 * config/aarch64/aarch64-cores.def (neoverse-512tvb): New entry.
17790 * config/aarch64/aarch64-tune.md: Regenerate.
17791 * config/aarch64/aarch64.c (neoverse512tvb_sve_vector_cost)
17792 (neoverse512tvb_sve_issue_info, neoverse512tvb_vec_issue_info)
17793 (neoverse512tvb_vector_cost, neoverse512tvb_tunings): New structures.
17794 (aarch64_adjust_body_cost_sve): Handle -mtune=neoverse-512tvb.
17795 (aarch64_adjust_body_cost): Likewise.
17797 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
17799 * config/aarch64/aarch64.c (aarch64_add_stmt_cost): Only
17800 record issue information for operations that occur in the
17803 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
17805 * config/aarch64/aarch64.c (aarch64_multiply_add_p): Add a vec_flags
17806 parameter. Detect cases in which an Advanced SIMD MLA would almost
17807 certainly require a MOV.
17808 (aarch64_count_ops): Update accordingly.
17810 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
17812 * config/aarch64/aarch64.c (aarch64_is_store_elt_extraction): New
17813 function, split out from...
17814 (aarch64_detect_vector_stmt_subtype): ...here.
17815 (aarch64_add_stmt_cost): Treat extracting element 0 as free.
17817 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
17819 * config/aarch64/aarch64-protos.h (sve_vec_cost):
17820 Add gather_load_x32_cost and gather_load_x64_cost.
17821 * config/aarch64/aarch64.c (generic_sve_vector_cost)
17822 (a64fx_sve_vector_cost, neoversev1_sve_vector_cost): Update
17823 accordingly, using the values given by the scalar_load * number
17824 of elements calculation that we used previously.
17825 (aarch64_detect_vector_stmt_subtype): Use the new fields.
17827 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
17829 * config/aarch64/aarch64.c (aarch64_adjust_body_cost_sve): New
17830 function, split out from...
17831 (aarch64_adjust_body_cost): ...here.
17833 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
17835 * config/aarch64/fractional-cost.h: New file.
17836 * config/aarch64/aarch64.c: Include <algorithm> (indirectly)
17837 and cost_fraction.h.
17838 (vec_cost_fraction): New typedef.
17839 (aarch64_detect_scalar_stmt_subtype): Use it for statement costs.
17840 (aarch64_detect_vector_stmt_subtype): Likewise.
17841 (aarch64_sve_adjust_stmt_cost, aarch64_adjust_stmt_cost): Likewise.
17842 (aarch64_estimate_min_cycles_per_iter): Use vec_cost_fraction
17844 (aarch64_adjust_body_cost): Likewise.
17845 (aarch64_test_cost_fraction): New function.
17846 (aarch64_run_selftests): Call it.
17848 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
17850 * config/aarch64/aarch64-protos.h (tune_params::sve_width): Turn
17852 * config/aarch64/aarch64.c (aarch64_cmp_autovec_modes): Update
17854 (aarch64_estimated_poly_value): Likewise. Use the least significant
17855 set bit for the minimum and likely values. Use the most significant
17856 set bit for the maximum value.
17858 2021-08-03 liuhongt <hongtao.liu@intel.com>
17860 * config/i386/sse.md (cond_<insn><mode>): New expander.
17861 (cond_mul<mode>): Ditto.
17863 2021-08-03 Kewen Lin <linkw@linux.ibm.com>
17865 * tree-cfg.c (move_sese_region_to_fn): Fix typos on dloop.
17867 2021-08-03 liuhongt <hongtao.liu@intel.com>
17869 * config/i386/sse.md (cond_<insn><mode>):New expander.
17870 (cond_mul<mode>): Ditto.
17871 (cond_div<mode>): Ditto.
17873 2021-08-02 H.J. Lu <hjl.tools@gmail.com>
17875 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Also
17876 check stack_realign_needed for stack realignment.
17877 (ix86_legitimate_constant_p): Always allow CONST_WIDE_INT smaller
17878 than the largest integer supported by vector register.
17879 * config/i386/i386.h (MAX_MOVE_MAX): New. Set to 64.
17880 (MOVE_MAX): Set to bytes of the largest integer supported by
17882 (STORE_MAX_PIECES): New.
17884 2021-08-02 H.J. Lu <hjl.tools@gmail.com>
17886 * config/i386/i386-expand.c (ix86_expand_vector_move): Call
17887 ix86_gen_scratch_sse_rtx to get a scratch SSE register to copy
17888 data from one memory location to another.
17890 2021-08-02 H.J. Lu <hjl.tools@gmail.com>
17892 PR middle-end/90773
17893 * config/i386/i386.c (TARGET_GEN_MEMSET_SCRATCH_RTX): New.
17895 2021-08-02 Aldy Hernandez <aldyh@redhat.com>
17897 PR tree-optimization/101724
17898 * params.opt: Remove --param=threader-iterative.
17899 * tree-ssa-threadbackward.c (pass_thread_jumps::execute): Remove
17902 2021-08-02 Tom de Vries <tdevries@suse.de>
17904 PR middle-end/101665
17905 * doc/extend.texi (nonnull attribute): Improve documentation.
17907 2021-08-02 Andrew Pinski <apinski@marvell.com>
17909 PR rtl-optimization/101683
17910 * rtlanal.c (may_trap_p_1): Handle UNSIGNED_FIX.
17912 2021-08-02 Roger Sayle <roger@nextmovesoftware.com>
17914 * tree-ssa-phiopt.c (cond_removal_in_builtin_zero_pattern):
17915 Renamed from cond_removal_in_popcount_clz_ctz_pattern.
17916 Add support for BSWAP, FFS, PARITY and CLRSB builtins.
17917 (tree_ssa_phiop_worker): Update call to function above.
17919 2021-08-01 H.J. Lu <hjl.tools@gmail.com>
17922 * config/i386/i386.md (bsr_rex64_1_zext): New.
17923 (combine splitter for constant - clzll): Replace gen_bsr_rex64_1
17924 with gen_bsr_rex64_1_zext.
17926 2021-07-31 Jakub Jelinek <jakub@redhat.com>
17929 * config/i386/i386.md (bsr_rex64_1, bsr_1, bsr_zext_1): New
17930 define_insn patterns.
17931 (*bsr_rex64_2, *bsr_2): New define_insn_and_split patterns.
17932 Add combine splitters for constant - clz.
17933 (clz<mode>2): Use a temporary pseudo for bsr result.
17935 2021-07-30 Paul A. Clarke <pc@us.ibm.com>
17937 * config/rs6000/smmintrin.h (_mm_floor_pd, _mm_floor_ps,
17938 _mm_floor_sd, _mm_floor_ss): New.
17940 2021-07-30 Paul A. Clarke <pc@us.ibm.com>
17942 * config/rs6000/smmintrin.h (_mm_ceil_pd, _mm_ceil_ps,
17943 _mm_ceil_sd, _mm_ceil_ss): New.
17945 2021-07-30 Paul A. Clarke <pc@us.ibm.com>
17947 * config/rs6000/smmintrin.h (_mm_blend_pd, _mm_blendv_pd,
17948 _mm_blend_ps, _mm_blendv_ps): New.
17950 2021-07-30 Roger Sayle <roger@nextmovesoftware.com>
17951 Uroš Bizjak <ubizjak@gmail.com>
17953 * config/i386/i386.md (*dec_cmov<mode>): New define_insn_and_split
17954 to generate a conditional move using the carry flag after sub $1.
17955 (peephole2): Eliminate a register-to-register move by inverting
17956 the condition of a conditional move.
17958 2021-07-30 Hans-Peter Nilsson <hp@bitrange.com>
17960 * config/mmix/mmix.md ("call", "call_value", "*call_real")
17961 ("*call_value_real"): Don't generate rtx mentioning the generic
17962 operands 1 and 2 to "call", and similarly for "call_value".
17963 * config/mmix/mmix.c (mmix_print_operand_punct_valid_p)
17964 (mmix_print_operand): Use '!' instead of 'p'.
17966 2021-07-30 Hans-Peter Nilsson <hp@bitrange.com>
17968 * doc/md.texi (call): Correct information about operand 2.
17969 * config/mmix/mmix.md ("call", "call_value"): Remove fixed FIXMEs.
17971 2021-07-30 Andrew MacLeod <amacleod@redhat.com>
17973 * range-op.cc (operator_trunc_mod::wi_fold): Fold constants.
17975 2021-07-30 Andrew MacLeod <amacleod@redhat.com>
17977 * range-op.cc (operator_div::wi_fold): Return UNDEFINED for [0, 0] divisor.
17979 2021-07-30 Andrew MacLeod <amacleod@redhat.com>
17981 * gimple-range-cache.cc (*::set_bb_range): Change const basic_block to
17982 const_basic_block..
17983 (*::get_bb_range): Ditto.
17984 (*::bb_range_p): Ditto.
17985 * gimple-range-cache.h: Change prototypes.
17987 2021-07-30 H.J. Lu <hjl.tools@gmail.com>
17989 PR middle-end/90773
17990 * builtins.c (builtin_memcpy_read_str): Change the mode argument
17991 from scalar_int_mode to fixed_size_mode.
17992 (builtin_strncpy_read_str): Likewise.
17993 (gen_memset_value_from_prev): New function.
17994 (builtin_memset_read_str): Change the mode argument from
17995 scalar_int_mode to fixed_size_mode. Use gen_memset_value_from_prev
17996 and support CONST_VECTOR.
17997 (builtin_memset_gen_str): Likewise.
17998 (try_store_by_multiple_pieces): Use by_pieces_constfn to declare
18000 * builtins.h (builtin_strncpy_read_str): Replace scalar_int_mode
18001 with fixed_size_mode.
18002 (builtin_memset_read_str): Likewise.
18003 * expr.c (widest_int_mode_for_size): Renamed to ...
18004 (widest_fixed_size_mode_for_size): Add a bool argument to
18005 indicate if QI vector mode can be used.
18006 (by_pieces_ninsns): Call widest_fixed_size_mode_for_size
18007 instead of widest_int_mode_for_size.
18008 (pieces_addr::adjust): Change the mode argument from
18009 scalar_int_mode to fixed_size_mode.
18010 (op_by_pieces_d): Make m_len read-only. Add a bool member,
18011 m_qi_vector_mode, to indicate that QI vector mode can be used.
18012 (op_by_pieces_d::op_by_pieces_d): Add a bool argument to
18013 initialize m_qi_vector_mode. Call widest_fixed_size_mode_for_size
18014 instead of widest_int_mode_for_size.
18015 (op_by_pieces_d::get_usable_mode): Change the mode argument from
18016 scalar_int_mode to fixed_size_mode. Call
18017 widest_fixed_size_mode_for_size instead of
18018 widest_int_mode_for_size.
18019 (op_by_pieces_d::smallest_fixed_size_mode_for_size): New member
18020 function to return the smallest integer or QI vector mode.
18021 (op_by_pieces_d::run): Call widest_fixed_size_mode_for_size
18022 instead of widest_int_mode_for_size. Call
18023 smallest_fixed_size_mode_for_size instead of
18024 smallest_int_mode_for_size.
18025 (store_by_pieces_d::store_by_pieces_d): Add a bool argument to
18026 indicate that QI vector mode can be used and pass it to
18027 op_by_pieces_d::op_by_pieces_d.
18028 (can_store_by_pieces): Call widest_fixed_size_mode_for_size
18029 instead of widest_int_mode_for_size. Pass memsetp to
18030 widest_fixed_size_mode_for_size to support QI vector mode.
18031 Allow all CONST_VECTORs for memset if vec_duplicate is supported.
18032 (store_by_pieces): Pass memsetp to
18033 store_by_pieces_d::store_by_pieces_d.
18034 (clear_by_pieces_1): Removed.
18035 (clear_by_pieces): Replace clear_by_pieces_1 with
18036 builtin_memset_read_str and pass true to store_by_pieces_d to
18037 support vector mode broadcast.
18038 (string_cst_read_str): Change the mode argument from
18039 scalar_int_mode to fixed_size_mode.
18040 * expr.h (by_pieces_constfn): Change scalar_int_mode to
18042 (by_pieces_prev): Likewise.
18043 * rtl.h (lowpart_subreg_regno): New.
18044 * rtlanal.c (lowpart_subreg_regno): New. A wrapper around
18045 simplify_subreg_regno.
18046 * target.def (gen_memset_scratch_rtx): New hook.
18047 * doc/tm.texi.in: Add TARGET_GEN_MEMSET_SCRATCH_RTX.
18048 * doc/tm.texi: Regenerated.
18050 2021-07-30 Xi Ruoyao <xry111@mengyan1223.wang>
18053 * config/mips/mips.c (mips_atomic_assign_expand_fenv): Use
18054 TARGET_EXPR instead of MODIFY_EXPR.
18056 2021-07-30 Xi Ruoyao <xry111@mengyan1223.wang>
18059 * config/mips/mips-protos.h (mips_expand_vec_cmp_expr): Declare.
18060 * config/mips/mips.c (mips_expand_vec_cmp_expr): New function.
18061 * config/mips/mips-msa.md (vec_cmp<MSA:mode><mode_i>): New
18063 (vec_cmpu<IMSA:mode><mode_i>): New expander.
18065 2021-07-30 H.J. Lu <hjl.tools@gmail.com>
18068 * config/i386/i386-options.c (ix86_option_override_internal):
18069 Don't enable LZCNT/POPCNT if they have been disabled explicitly.
18071 2021-07-30 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
18074 * config/arm/arm_neon.h (vld1_p64): Replace call to builtin by
18075 explicitly dereferencing __a.
18076 (vld1_s64): Likewise.
18077 (vld1_u64): Likewise.
18078 * config/arm/arm_neon_builtins.def (vld1): Remove entry for di
18079 and change to VAR13.
18081 2021-07-30 Aldy Hernandez <aldyh@redhat.com>
18083 * gimple-loop-versioning.cc (lv_dom_walker::lv_dom_walker): Remove
18084 use of m_range_analyzer.
18085 (loop_versioning::lv_dom_walker::before_dom_children): Same.
18086 (loop_versioning::lv_dom_walker::after_dom_children): Remove.
18087 (loop_versioning::prune_loop_conditions): Replace vr_values use
18088 with range_query interface.
18089 (pass_loop_versioning::execute): Use ranger.
18091 2021-07-30 Xi Ruoyao <xry111@mengyan1223.wang>
18094 * ipa-devirt.c (ipa_odr_read_section): Compare the precision of
18095 enum values, and emit a warning if they mismatch.
18097 2021-07-30 Kewen Lin <linkw@linux.ibm.com>
18099 * cfgloop.h (as_const): New function.
18100 (class loop_iterator): Rename to ...
18101 (class loops_list): ... this.
18102 (loop_iterator::next): Rename to ...
18103 (loops_list::Iter::fill_curr_loop): ... this and adjust.
18104 (loop_iterator::loop_iterator): Rename to ...
18105 (loops_list::loops_list): ... this and adjust.
18106 (loops_list::Iter): New class.
18107 (loops_list::iterator): New type.
18108 (loops_list::const_iterator): New type.
18109 (loops_list::begin): New function.
18110 (loops_list::end): Likewise.
18111 (loops_list::begin const): Likewise.
18112 (loops_list::end const): Likewise.
18113 (FOR_EACH_LOOP): Remove.
18114 (FOR_EACH_LOOP_FN): Remove.
18115 * cfgloop.c (flow_loops_dump): Adjust FOR_EACH_LOOP* with range-based
18116 for loop with loops_list instance.
18117 (sort_sibling_loops): Likewise.
18118 (disambiguate_loops_with_multiple_latches): Likewise.
18119 (verify_loop_structure): Likewise.
18120 * cfgloopmanip.c (create_preheaders): Likewise.
18121 (force_single_succ_latches): Likewise.
18122 * config/aarch64/falkor-tag-collision-avoidance.c
18123 (execute_tag_collision_avoidance): Likewise.
18124 * config/mn10300/mn10300.c (mn10300_scan_for_setlb_lcc): Likewise.
18125 * config/s390/s390.c (s390_adjust_loops): Likewise.
18126 * doc/loop.texi: Likewise.
18127 * gimple-loop-interchange.cc (pass_linterchange::execute): Likewise.
18128 * gimple-loop-jam.c (tree_loop_unroll_and_jam): Likewise.
18129 * gimple-loop-versioning.cc (loop_versioning::analyze_blocks): Likewise.
18130 (loop_versioning::make_versioning_decisions): Likewise.
18131 * gimple-ssa-split-paths.c (split_paths): Likewise.
18132 * graphite-isl-ast-to-gimple.c (graphite_regenerate_ast_isl): Likewise.
18133 * graphite.c (canonicalize_loop_form): Likewise.
18134 (graphite_transform_loops): Likewise.
18135 * ipa-fnsummary.c (analyze_function_body): Likewise.
18136 * ipa-pure-const.c (analyze_function): Likewise.
18137 * loop-doloop.c (doloop_optimize_loops): Likewise.
18138 * loop-init.c (loop_optimizer_finalize): Likewise.
18139 (fix_loop_structure): Likewise.
18140 * loop-invariant.c (calculate_loop_reg_pressure): Likewise.
18141 (move_loop_invariants): Likewise.
18142 * loop-unroll.c (decide_unrolling): Likewise.
18143 (unroll_loops): Likewise.
18144 * modulo-sched.c (sms_schedule): Likewise.
18145 * predict.c (predict_loops): Likewise.
18146 (pass_profile::execute): Likewise.
18147 * profile.c (branch_prob): Likewise.
18148 * sel-sched-ir.c (sel_finish_pipelining): Likewise.
18149 (sel_find_rgns): Likewise.
18150 * tree-cfg.c (replace_loop_annotate): Likewise.
18151 (replace_uses_by): Likewise.
18152 (move_sese_region_to_fn): Likewise.
18153 * tree-if-conv.c (pass_if_conversion::execute): Likewise.
18154 * tree-loop-distribution.c (loop_distribution::execute): Likewise.
18155 * tree-parloops.c (parallelize_loops): Likewise.
18156 * tree-predcom.c (tree_predictive_commoning): Likewise.
18157 * tree-scalar-evolution.c (scev_initialize): Likewise.
18158 (scev_reset): Likewise.
18159 * tree-ssa-dce.c (find_obviously_necessary_stmts): Likewise.
18160 * tree-ssa-live.c (remove_unused_locals): Likewise.
18161 * tree-ssa-loop-ch.c (ch_base::copy_headers): Likewise.
18162 * tree-ssa-loop-im.c (analyze_memory_references): Likewise.
18163 (tree_ssa_lim_initialize): Likewise.
18164 * tree-ssa-loop-ivcanon.c (canonicalize_induction_variables): Likewise.
18165 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize): Likewise.
18166 * tree-ssa-loop-manip.c (get_loops_exits): Likewise.
18167 * tree-ssa-loop-niter.c (estimate_numbers_of_iterations): Likewise.
18168 (free_numbers_of_iterations_estimates): Likewise.
18169 * tree-ssa-loop-prefetch.c (tree_ssa_prefetch_arrays): Likewise.
18170 * tree-ssa-loop-split.c (tree_ssa_split_loops): Likewise.
18171 * tree-ssa-loop-unswitch.c (tree_ssa_unswitch_loops): Likewise.
18172 * tree-ssa-loop.c (gate_oacc_kernels): Likewise.
18173 (pass_scev_cprop::execute): Likewise.
18174 * tree-ssa-propagate.c (clean_up_loop_closed_phi): Likewise.
18175 * tree-ssa-sccvn.c (do_rpo_vn): Likewise.
18176 * tree-ssa-threadupdate.c
18177 (jump_thread_path_registry::thread_through_all_blocks): Likewise.
18178 * tree-vectorizer.c (vectorize_loops): Likewise.
18179 * tree-vrp.c (vrp_asserts::find_assert_locations): Likewise.
18181 2021-07-29 Hans-Peter Nilsson <hp@bitrange.com>
18183 * config/mmix/mmix.c (mmix_function_arg_1): Avoid
18184 generating a VOIDmode register for e.g the
18185 function_arg_info::end_marker.
18187 2021-07-29 Jeff Law <jeffreyalaw@gmail.com>
18189 * config/h8300/h8300-modes.def: Add CCZ, CCV and CCC, drop CCZNV.
18190 * config/h8300/h8300.md (H8cc mode iterator): Add CCZ.
18191 (cc mode_attr): Similarly.
18192 (ccz subst_attr): Similarly.
18193 * config/h8300/jumpcall.md: Add new patterns for branch-on-bit.
18194 * config/h8300/testcompare.md: Remove various cc0 based patterns
18195 that had been commented out. Add pattern to set CCZ from a bit
18198 2021-07-29 Thomas Schwinge <thomas@codesourcery.com>
18199 Julian Brown <julian@codesourcery.com>
18200 Kwok Cheung Yeung <kcy@codesourcery.com>
18202 * omp-offload.c (oacc_loop_xform_head_tail, oacc_loop_process):
18203 'update_stmt' after modification.
18204 (pass_oacc_loop_designation): New function, extracted out of...
18205 (pass_oacc_device_lower): ... this.
18206 (pass_data_oacc_loop_designation, pass_oacc_loop_designation)
18207 (make_pass_oacc_loop_designation): New
18208 * passes.def: Add it.
18209 * tree-parloops.c (create_parallel_loop): Adjust.
18210 * tree-pass.h (make_pass_oacc_loop_designation): New.
18212 2021-07-29 Aldy Hernandez <aldyh@redhat.com>
18214 * flag-types.h (enum threader_mode): New.
18215 * params.opt: Add entry for --param=threader-mode.
18216 * tree-ssa-threadbackward.c (THREADER_ITERATIVE_MODE): New.
18217 (class back_threader): New.
18218 (back_threader::back_threader): New.
18219 (back_threader::~back_threader): New.
18220 (back_threader::maybe_register_path): New.
18221 (back_threader::find_taken_edge): New.
18222 (back_threader::find_taken_edge_switch): New.
18223 (back_threader::find_taken_edge_cond): New.
18224 (back_threader::resolve_def): New.
18225 (back_threader::resolve_phi): New.
18226 (back_threader::find_paths_to_names): New.
18227 (back_threader::find_paths): New.
18230 (thread_jumps::find_jump_threads_backwards): Call ranger threader.
18231 (thread_jumps::find_jump_threads_backwards_with_ranger): New.
18232 (pass_thread_jumps::execute): Abstract out code...
18233 (try_thread_blocks): ...here.
18234 * tree-ssa-threadedge.c (jump_threader::thread_outgoing_edges):
18235 Abstract out threading candidate code to...
18236 (single_succ_to_potentially_threadable_block): ...here.
18237 * tree-ssa-threadedge.h (single_succ_to_potentially_threadable_block):
18239 * tree-ssa-threadupdate.c (register_jump_thread): Return boolean.
18240 * tree-ssa-threadupdate.h (class jump_thread_path_registry):
18241 Return bool from register_jump_thread.
18243 2021-07-29 Andreas Krebbel <krebbel@linux.ibm.com>
18245 * target.def: in0 and in1 do not need to be registers.
18246 * doc/tm.texi: Regenerate.
18248 2021-07-29 liuhongt <hongtao.liu@intel.com>
18251 * config/i386/i386.c (ix86_widen_mult_cost): New function.
18252 (ix86_add_stmt_cost): Use ix86_widen_mult_cost for
18255 2021-07-29 Jiufu Guo <guojiufu@linux.ibm.com>
18258 * config/rs6000/rs6000.c (TARGET_PREFERRED_DOLOOP_MODE): New hook.
18259 (rs6000_preferred_doloop_mode): New hook.
18260 * doc/tm.texi: Regenerate.
18261 * doc/tm.texi.in: Add hook preferred_doloop_mode.
18262 * target.def (preferred_doloop_mode): New hook.
18263 * targhooks.c (default_preferred_doloop_mode): New hook.
18264 * targhooks.h (default_preferred_doloop_mode): New hook.
18265 * tree-ssa-loop-ivopts.c (compute_doloop_base_on_mode): New function.
18266 (add_iv_candidate_for_doloop): Call targetm.preferred_doloop_mode
18267 and compute_doloop_base_on_mode.
18269 2021-07-28 Martin Sebor <msebor@redhat.com>
18271 PR middle-end/101494
18272 * tree-ssa-uninit.c (maybe_warn_operand): Correct object offset
18273 and size computation.
18275 2021-07-28 Martin Sebor <msebor@redhat.com>
18277 PR middle-end/101601
18278 * gimple-array-bounds.cc (array_bounds_checker::check_mem_ref): Remove
18280 Handle pointers to functions.
18282 2021-07-28 Martin Sebor <msebor@redhat.com>
18284 * Makefile.in (OBJS): Add gimple-ssa-warn-access.o and pointer-query.o.
18285 * attribs.h (fndecl_dealloc_argno): Move fndecl_dealloc_argno to tree.h.
18286 * builtins.c (compute_objsize_r): Move to pointer-query.cc.
18287 (access_ref::access_ref): Same.
18288 (access_ref::phi): Same.
18289 (access_ref::get_ref): Same.
18290 (access_ref::size_remaining): Same.
18291 (access_ref::offset_in_range): Same.
18292 (access_ref::add_offset): Same.
18293 (access_ref::inform_access): Same.
18294 (ssa_name_limit_t::visit_phi): Same.
18295 (ssa_name_limit_t::leave_phi): Same.
18296 (ssa_name_limit_t::next): Same.
18297 (ssa_name_limit_t::next_phi): Same.
18298 (ssa_name_limit_t::~ssa_name_limit_t): Same.
18299 (pointer_query::pointer_query): Same.
18300 (pointer_query::get_ref): Same.
18301 (pointer_query::put_ref): Same.
18302 (pointer_query::flush_cache): Same.
18303 (warn_string_no_nul): Move to gimple-ssa-warn-access.cc.
18304 (check_nul_terminated_array): Same.
18305 (unterminated_array): Same.
18306 (maybe_warn_for_bound): Same.
18307 (check_read_access): Same.
18308 (warn_for_access): Same.
18309 (get_size_range): Same.
18310 (check_access): Same.
18311 (gimple_call_alloc_size): Move to tree.c.
18312 (gimple_parm_array_size): Move to pointer-query.cc.
18313 (get_offset_range): Same.
18314 (gimple_call_return_array): Same.
18315 (handle_min_max_size): Same.
18316 (handle_array_ref): Same.
18317 (handle_mem_ref): Same.
18318 (compute_objsize): Same.
18319 (gimple_call_alloc_p): Move to gimple-ssa-warn-access.cc.
18320 (call_dealloc_argno): Same.
18321 (fndecl_dealloc_argno): Same.
18322 (new_delete_mismatch_p): Same.
18323 (matching_alloc_calls_p): Same.
18324 (warn_dealloc_offset): Same.
18325 (maybe_emit_free_warning): Same.
18326 * builtins.h (check_nul_terminated_array): Move to
18327 gimple-ssa-warn-access.h.
18328 (check_nul_terminated_array): Same.
18329 (warn_string_no_nul): Same.
18330 (unterminated_array): Same.
18331 (class ssa_name_limit_t): Same.
18332 (class pointer_query): Same.
18333 (struct access_ref): Same.
18334 (class range_query): Same.
18335 (struct access_data): Same.
18336 (gimple_call_alloc_size): Same.
18337 (gimple_parm_array_size): Same.
18338 (compute_objsize): Same.
18339 (class access_data): Same.
18340 (maybe_emit_free_warning): Same.
18341 * calls.c (initialize_argument_information): Remove call to
18342 maybe_emit_free_warning.
18343 * gimple-array-bounds.cc: Include new header..
18344 * gimple-fold.c: Same.
18345 * gimple-ssa-sprintf.c: Same.
18346 * gimple-ssa-warn-restrict.c: Same.
18347 * passes.def: Add pass_warn_access.
18348 * tree-pass.h (make_pass_warn_access): Declare.
18349 * tree-ssa-strlen.c: Include new headers.
18350 * tree.c (fndecl_dealloc_argno): Move here from builtins.c.
18351 * tree.h (fndecl_dealloc_argno): Move here from attribs.h.
18352 * gimple-ssa-warn-access.cc: New file.
18353 * gimple-ssa-warn-access.h: New file.
18354 * pointer-query.cc: New file.
18355 * pointer-query.h: New file.
18357 2021-07-28 Jakub Jelinek <jakub@redhat.com>
18359 PR middle-end/101624
18360 * ubsan.c (maybe_instrument_pointer_overflow,
18361 instrument_object_size): Only test DECL_REGISTER on VAR_DECLs,
18362 PARM_DECLs or RESULT_DECLs.
18363 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Likewise.
18365 2021-07-28 Jakub Jelinek <jakub@redhat.com>
18367 PR middle-end/101642
18368 * match.pd (bswap16 (x) == bswap16 (y)): Cast both operands
18369 to type of bswap16 for comparison.
18370 (bswap16 (x) == cst): Cast bswap16 operand to type of cst.
18372 2021-07-28 Richard Biener <rguenther@suse.de>
18374 PR tree-optimization/101615
18375 * tree-vect-slp.c (vect_optimize_slp): Materialize permutes
18376 at CTOR SLP graph entries.
18378 2021-07-28 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18380 * config/aarch64/aarch64.md (*extend<SHORT:mode><GPI:mode>2_aarch64):
18381 Add "r,w" alternative.
18383 2021-07-28 H.J. Lu <hjl.tools@gmail.com>
18386 * config/i386/i386.c (ix86_avx_u128_mode_needed): Don't set
18387 AVX_U128_DIRTY when all bits are zero.
18389 2021-07-28 Richard Biener <rguenther@suse.de>
18391 PR tree-optimization/101615
18392 * tree-vect-slp.c (vect_optimize_slp): Pre-existing vector
18393 external nodes cannot be permuted so make them perm_out 0.
18395 2021-07-28 Andrew Stubbs <ams@codesourcery.com>
18398 * config.in: Regenerate.
18399 * config/gcn/gcn-hsa.h (A_FIJI): New define.
18400 (A_900): New define.
18401 (A_906): New define.
18402 (A_908): New define.
18403 (ASM_SPEC): Use A_FIJI, A_900, A_906 and A_908.
18404 * config/gcn/gcn.c (output_file_start): Adjust attributes according
18405 to the assembler capabilities.
18406 * config/gcn/mkoffload.c (main): Likewise.
18407 * configure: Regenerate.
18408 * configure.ac: Add tests for LLVM assembler attribute features.
18410 2021-07-28 Andrew MacLeod <amacleod@redhat.com>
18412 * gimple-range-gori.cc (gori_compute::outgoing_edge_range_p): Check for
18413 cond_false and cond_true on branches.
18415 2021-07-28 Bin Cheng <bin.cheng@linux.alibaba.com>
18417 * config/aarch64/aarch64.c (aarch64_gen_adjusted_ldpstp): use
18420 2021-07-28 Bin Cheng <bin.cheng@linux.alibaba.com>
18422 * alias.c (init_alias_analysis): Don't skip prologue/epilogue.
18424 2021-07-28 Jakub Jelinek <jakub@redhat.com>
18427 * config/i386/sse.md (vashr<mode>3): Split into vashrv8di3 expander
18428 and vashrv4di3 expander, where the latter requires just TARGET_AVX2
18429 and has special !TARGET_AVX512VL expansion.
18430 (vashrv2di3<mask_name>): Rename to ...
18431 (vashrv2di3): ... this. Change condition to TARGET_XOP || TARGET_AVX2
18432 and add special !TARGET_XOP && !TARGET_AVX512VL expansion.
18434 2021-07-28 Martin Uecker <muecker@gwdg.de>
18436 * calls.c (maybe_warn_rdwr_sizes): Correct argument
18437 numbers in warning that were switched.
18439 2021-07-28 Kewen Lin <linkw@linux.ibm.com>
18441 PR tree-optimization/101596
18442 * tree-vect-patterns.c (vect_recog_mulhs_pattern): Fix wrong check
18443 by using new_type's precision instead.
18445 2021-07-28 liuhongt <hongtao.liu@intel.com>
18448 * config/i386/i386.h (processor_costs): Add new member
18450 * config/i386/x86-tune-costs.h (ix86_size_cost, i386_cost,
18451 i486_cost, pentium_cost, lakemont_cost, pentiumpro_cost,
18452 geode_cost, k6_cost, athlon_cost, k8_cost, amdfam10_cost,
18453 bdver_cost, znver1_cost, znver2_cost, znver3_cost,
18454 btver1_cost, btver2_cost, btver3_cost, pentium4_cost,
18455 nocona_cost, atom_cost, atom_cost, slm_cost, intel_cost,
18456 generic_cost, core_cost): Initialize integer_to_sse same value
18458 (skylake_cost): Initialize integer_to_sse twice as much as sse_op.
18459 * config/i386/i386.c (ix86_builtin_vectorization_cost):
18460 Use integer_to_sse instead of sse_op to calculate the cost of
18463 2021-07-27 Bill Schmidt <wschmidt@linux.ibm.com>
18465 * config/rs6000/rs6000-gen-builtins.c (write_ovld_static_init): New
18467 (write_init_file): Call write_ovld_static_init.
18469 2021-07-27 Bill Schmidt <wschmidt@linux.ibm.com>
18471 * config/rs6000/rs6000-gen-builtins.c (write_bif_static_init): New
18473 (write_init_file): Call write_bif_static_init.
18475 2021-07-27 Bill Schmidt <wschmidt@linux.ibm.com>
18477 * config/rs6000/rs6000-gen-builtins.c (typemap): New struct.
18478 (TYPE_MAP_SIZE): New macro.
18479 (type_map): New initialized variable.
18480 (typemap_cmp): New function.
18481 (write_type_node): Likewise.
18482 (write_fntype_init): Implement.
18484 2021-07-27 Martin Sebor <msebor@redhat.com>
18486 PR tree-optimization/101584
18487 * tree-ssa-uninit.c (builtin_call_nomodifying_p): New function.
18488 (check_defs): Call it.
18490 2021-07-27 Aldy Hernandez <aldyh@redhat.com>
18492 * tree-ssa-dom.c (dom_jump_threader_simplifier):
18493 Put avail_exprs_stack in the class, instead of passing it to
18494 jump_threader_simplifier.
18495 (dom_jump_threader_simplifier::simplify): Add state argument.
18496 (dom_opt_dom_walker): Add state.
18497 (pass_dominator::execute): Pass state to threader.
18498 (dom_opt_dom_walker::before_dom_children): Use state.
18499 * tree-ssa-threadedge.c (jump_threader::jump_threader): Replace
18500 arguments by state.
18501 (jump_threader::record_temporary_equivalences_from_phis):
18502 Register equivalences through the state variable.
18503 (jump_threader::record_temporary_equivalences_from_stmts_at_dest):
18504 Record ranges in a statement through the state variable.
18505 (jump_threader::simplify_control_stmt_condition): Pass state to
18507 (jump_threader::simplify_control_stmt_condition_1): Same.
18508 (jump_threader::thread_around_empty_blocks): Remove obsolete
18510 (jump_threader::thread_through_normal_block): Record equivalences
18511 on edge through the state variable.
18512 (jump_threader::thread_across_edge): Abstract state pushing.
18513 (jt_state::jt_state): New.
18514 (jt_state::push): New.
18515 (jt_state::pop): New.
18516 (jt_state::register_equiv): New.
18517 (jt_state::record_ranges_from_stmt): New.
18518 (jt_state::register_equivs_on_edge): New.
18519 (jump_threader_simplifier::jump_threader_simplifier): Move from
18521 (jump_threader_simplifier::simplify): Add state argument.
18522 * tree-ssa-threadedge.h (class jt_state): New.
18523 (class jump_threader): Add state to constructor.
18524 (class jump_threader_simplifier): Add state to simplify. Remove
18525 avail_exprs_stack from class.
18526 * tree-vrp.c (vrp_jump_threader_simplifier::simplify): Add state
18528 (vrp_jump_threader::vrp_jump_threader): Add state.
18529 (vrp_jump_threader::~vrp_jump_threader): Cleanup state.
18531 2021-07-27 Aldy Hernandez <aldyh@redhat.com>
18533 * Makefile.in (OBJS): Add gimple-range-path.o.
18534 * gimple-range-path.cc: New file.
18535 * gimple-range-path.h: New file.
18537 2021-07-27 Jonathan Wright <jonathan.wright@arm.com>
18539 * config/aarch64/aarch64-simd.md: Push sign/zero-extension
18540 inside vec_duplicate for all patterns.
18541 * simplify-rtx.c (simplify_context::simplify_unary_operation_1):
18542 Push sign/zero-extension inside vec_duplicate.
18544 2021-07-27 Richard Biener <rguenther@suse.de>
18546 PR tree-optimization/101573
18547 * tree-ssa-uninit.c (warn_uninit_phi_uses): New function
18548 looking at uninitialized PHI arg defs in some constrained cases.
18549 (warn_uninitialized_vars): Call it.
18550 (execute_early_warn_uninitialized): Calculate dominators.
18552 2021-07-27 Richard Biener <rguenther@suse.de>
18554 PR tree-optimization/39821
18555 * tree-vect-stmts.c (vect_model_promotion_demotion_cost): Use
18556 vector_stmt for widening arithmetic.
18557 (vectorizable_conversion): Adjust.
18559 2021-07-27 Martin Jambor <mjambor@suse.cz>
18561 * cgraph.h (ipa_replace_map): New field force_load_ref.
18562 * ipa-prop.h (ipa_param_descriptor): Reduce precision of move_cost,
18563 aded new flag load_dereferenced, adjusted comments.
18564 (ipa_get_param_dereferenced): New function.
18565 (ipa_set_param_dereferenced): Likewise.
18566 * cgraphclones.c (cgraph_node::create_virtual_clone): Follow it.
18567 * ipa-cp.c: Include gimple.h.
18568 (ipcp_discover_new_direct_edges): Take into account dereferenced flag.
18569 (get_replacement_map): New parameter force_load_ref, set the
18570 appropriate flag in ipa_replace_map if set.
18571 (struct symbol_and_index_together): New type.
18572 (adjust_refs_in_act_callers): New function.
18573 (adjust_references_in_caller): Likewise.
18574 (create_specialized_node): When appropriate, call
18575 adjust_references_in_caller and force only load references.
18576 * ipa-prop.c (load_from_dereferenced_name): New function.
18577 (ipa_analyze_controlled_uses): Also detect loads from a
18578 dereference, harden testing of call statements.
18579 (ipa_write_node_info): Stream the dereferenced flag.
18580 (ipa_read_node_info): Likewise.
18581 (ipa_set_jf_constant): Also create refdesc when jump function
18582 references a variable.
18583 (cgraph_node_for_jfunc): Rename to symtab_node_for_jfunc, work
18584 also on references of variables and return a symtab_node. Adjust
18586 (propagate_controlled_uses): Also remove references to VAR_DECLs.
18588 2021-07-27 Jakub Jelinek <jakub@redhat.com>
18590 PR middle-end/101586
18591 * gimple-fold.c (clear_padding_type): Ignore FIELD_DECLs with byte
18592 positions above or equal to sz except for diagnostics of flexible
18595 2021-07-26 Andrew MacLeod <amacleod@redhat.com>
18597 PR tree-optimization/78888
18598 * gimple-range-fold.cc (get_letter_range): New.
18599 (fold_using_range::range_of_builtin_call): Call get_letter_range.
18601 2021-07-26 Andrew MacLeod <amacleod@redhat.com>
18603 PR tree-optimization/78888
18604 * gimple-range-fold.cc (fold_using_range::range_of_builtin_call): Add cases
18605 for CFN_BUILT_IN_TOUPPER and CFN_BUILT_IN_TOLOWER.
18607 2021-07-26 Roger Sayle <roger@nextmovesoftware.com>
18608 Marc Glisse <marc.glisse@inria.fr>
18610 * match.pd (rotate): Simplify equality/inequality of rotations.
18611 (bswap): Simplify equality/inequality tests of byte swapping.
18613 2021-07-26 Aldy Hernandez <aldyh@redhat.com>
18615 * range-op.cc (operator_bitwise_xor::op1_op2_relation_effect):
18618 2021-07-26 Aldy Hernandez <aldyh@redhat.com>
18620 * range-op.cc (operator_lshift::fold_range): Pass rel to
18621 base class fold_range.
18622 (operator_rshift::fold_range): Same.
18624 2021-07-26 Ashimida <ashimida@linux.alibaba.com>
18627 * toplev.h (min_align_loops_log): Remove declaration.
18628 (min_align_jumps_log, min_align_labels_log): Likewise.
18629 (min_align_functions_log): Likewise.
18631 2021-07-26 Aldy Hernandez <aldyh@redhat.com>
18633 * tree-vrp.c (vrp_simplify_cond_using_ranges): Rename vr_values
18635 (execute_vrp): Abstract out simplification of conditionals...
18636 (simplify_casted_conds): ...here.
18638 2021-07-26 Aldy Hernandez <aldyh@redhat.com>
18640 * gimple-array-bounds.cc (array_bounds_checker::get_value_range):
18641 Add gimple argument.
18642 (array_bounds_checker::check_array_ref): Same.
18643 (array_bounds_checker::check_addr_expr): Same.
18644 (array_bounds_checker::check_array_bounds): Pass statement to
18645 check_array_bounds and check_addr_expr.
18646 * gimple-array-bounds.h (check_array_bounds): Add gimple argument.
18647 (check_addr_expr): Same.
18648 (get_value_range): Same.
18650 2021-07-26 Tamar Christina <tamar.christina@arm.com>
18652 * config/aarch64/aarch64-simd-builtins.def (sdot, udot): Rename to..
18653 (sdot_prod, udot_prod): ... This.
18654 * config/aarch64/aarch64-simd.md (aarch64_<sur>dot<vsi2qi>): Merged
18656 (<sur>dot_prod<vsi2qi>): ... this.
18657 (aarch64_<sur>dot_lane<vsi2qi>, aarch64_<sur>dot_laneq<vsi2qi>):
18658 Change operands order.
18659 (<sur>sadv16qi): Use new operands order.
18660 * config/aarch64/arm_neon.h (vdot_u32, vdotq_u32, vdot_s32,
18661 vdotq_s32): Use new RTL ordering.
18663 2021-07-26 Tamar Christina <tamar.christina@arm.com>
18665 * config/aarch64/aarch64-builtins.c (TYPES_TERNOP_SUSS,
18666 aarch64_types_ternop_suss_qualifiers): New.
18667 * config/aarch64/aarch64-simd-builtins.def (usdot_prod): Use it.
18668 * config/aarch64/aarch64-simd.md (usdot_prod<vsi2qi>): Re-organize RTL.
18669 * config/aarch64/arm_neon.h (vusdot_s32, vusdotq_s32): Use it.
18671 2021-07-23 Jakub Jelinek <jakub@redhat.com>
18673 PR rtl-optimization/101562
18674 * expmed.c (store_integral_bit_field): Only use movstrict_optab
18675 if the operand isn't paradoxical.
18677 2021-07-23 Aldy Hernandez <aldyh@redhat.com>
18679 * gimple-array-bounds.h (class array_bounds_checker): Change
18680 ranges type to range_query.
18682 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
18684 * config/aarch64/arm_neon.h (vst1_s64_x2): Use
18685 __builtin_memcpy instead of constructing
18686 __builtin_aarch64_simd_oi one vector at a time.
18687 (vst1_u64_x2): Likewise.
18688 (vst1_f64_x2): Likewise.
18689 (vst1_s8_x2): Likewise.
18690 (vst1_p8_x2): Likewise.
18691 (vst1_s16_x2): Likewise.
18692 (vst1_p16_x2): Likewise.
18693 (vst1_s32_x2): Likewise.
18694 (vst1_u8_x2): Likewise.
18695 (vst1_u16_x2): Likewise.
18696 (vst1_u32_x2): Likewise.
18697 (vst1_f16_x2): Likewise.
18698 (vst1_f32_x2): Likewise.
18699 (vst1_p64_x2): Likewise.
18700 (vst1q_s8_x2): Likewise.
18701 (vst1q_p8_x2): Likewise.
18702 (vst1q_s16_x2): Likewise.
18703 (vst1q_p16_x2): Likewise.
18704 (vst1q_s32_x2): Likewise.
18705 (vst1q_s64_x2): Likewise.
18706 (vst1q_u8_x2): Likewise.
18707 (vst1q_u16_x2): Likewise.
18708 (vst1q_u32_x2): Likewise.
18709 (vst1q_u64_x2): Likewise.
18710 (vst1q_f16_x2): Likewise.
18711 (vst1q_f32_x2): Likewise.
18712 (vst1q_f64_x2): Likewise.
18713 (vst1q_p64_x2): Likewise.
18715 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
18717 * config/aarch64/arm_neon.h (vst1_s64_x3): Use
18718 __builtin_memcpy instead of constructing
18719 __builtin_aarch64_simd_ci one vector at a time.
18720 (vst1_u64_x3): Likewise.
18721 (vst1_f64_x3): Likewise.
18722 (vst1_s8_x3): Likewise.
18723 (vst1_p8_x3): Likewise.
18724 (vst1_s16_x3): Likewise.
18725 (vst1_p16_x3): Likewise.
18726 (vst1_s32_x3): Likewise.
18727 (vst1_u8_x3): Likewise.
18728 (vst1_u16_x3): Likewise.
18729 (vst1_u32_x3): Likewise.
18730 (vst1_f16_x3): Likewise.
18731 (vst1_f32_x3): Likewise.
18732 (vst1_p64_x3): Likewise.
18733 (vst1q_s8_x3): Likewise.
18734 (vst1q_p8_x3): Likewise.
18735 (vst1q_s16_x3): Likewise.
18736 (vst1q_p16_x3): Likewise.
18737 (vst1q_s32_x3): Likewise.
18738 (vst1q_s64_x3): Likewise.
18739 (vst1q_u8_x3): Likewise.
18740 (vst1q_u16_x3): Likewise.
18741 (vst1q_u32_x3): Likewise.
18742 (vst1q_u64_x3): Likewise.
18743 (vst1q_f16_x3): Likewise.
18744 (vst1q_f32_x3): Likewise.
18745 (vst1q_f64_x3): Likewise.
18746 (vst1q_p64_x3): Likewise.
18748 2021-07-23 H.J. Lu <hjl.tools@gmail.com>
18751 * config/i386/i386.c (ix86_gen_scratch_sse_rtx): Don't return
18752 hard register when LRA is in progress.
18754 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
18756 * config/aarch64/arm_neon.h (vst1_s8_x4): Use
18757 __builtin_memcpy instead of using a union.
18758 (vst1q_s8_x4): Likewise.
18759 (vst1_s16_x4): Likewise.
18760 (vst1q_s16_x4): Likewise.
18761 (vst1_s32_x4): Likewise.
18762 (vst1q_s32_x4): Likewise.
18763 (vst1_u8_x4): Likewise.
18764 (vst1q_u8_x4): Likewise.
18765 (vst1_u16_x4): Likewise.
18766 (vst1q_u16_x4): Likewise.
18767 (vst1_u32_x4): Likewise.
18768 (vst1q_u32_x4): Likewise.
18769 (vst1_f16_x4): Likewise.
18770 (vst1q_f16_x4): Likewise.
18771 (vst1_f32_x4): Likewise.
18772 (vst1q_f32_x4): Likewise.
18773 (vst1_p8_x4): Likewise.
18774 (vst1q_p8_x4): Likewise.
18775 (vst1_p16_x4): Likewise.
18776 (vst1q_p16_x4): Likewise.
18777 (vst1_s64_x4): Likewise.
18778 (vst1_u64_x4): Likewise.
18779 (vst1_p64_x4): Likewise.
18780 (vst1q_s64_x4): Likewise.
18781 (vst1q_u64_x4): Likewise.
18782 (vst1q_p64_x4): Likewise.
18783 (vst1_f64_x4): Likewise.
18784 (vst1q_f64_x4): Likewise.
18786 2021-07-23 Jonathan Wrightt <jonathan.wright@arm.com>
18788 * config/aarch64/arm_neon.h (vst2_s64): Use __builtin_memcpy
18789 instead of constructing __builtin_aarch64_simd_oi one vector
18791 (vst2_u64): Likewise.
18792 (vst2_f64): Likewise.
18793 (vst2_s8): Likewise.
18794 (vst2_p8): Likewise.
18795 (vst2_s16): Likewise.
18796 (vst2_p16): Likewise.
18797 (vst2_s32): Likewise.
18798 (vst2_u8): Likewise.
18799 (vst2_u16): Likewise.
18800 (vst2_u32): Likewise.
18801 (vst2_f16): Likewise.
18802 (vst2_f32): Likewise.
18803 (vst2_p64): Likewise.
18804 (vst2q_s8): Likewise.
18805 (vst2q_p8): Likewise.
18806 (vst2q_s16): Likewise.
18807 (vst2q_p16): Likewise.
18808 (vst2q_s32): Likewise.
18809 (vst2q_s64): Likewise.
18810 (vst2q_u8): Likewise.
18811 (vst2q_u16): Likewise.
18812 (vst2q_u32): Likewise.
18813 (vst2q_u64): Likewise.
18814 (vst2q_f16): Likewise.
18815 (vst2q_f32): Likewise.
18816 (vst2q_f64): Likewise.
18817 (vst2q_p64): Likewise.
18819 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
18821 * config/aarch64/arm_neon.h (vst3_s64): Use __builtin_memcpy
18822 instead of constructing __builtin_aarch64_simd_ci one vector
18824 (vst3_u64): Likewise.
18825 (vst3_f64): Likewise.
18826 (vst3_s8): Likewise.
18827 (vst3_p8): Likewise.
18828 (vst3_s16): Likewise.
18829 (vst3_p16): Likewise.
18830 (vst3_s32): Likewise.
18831 (vst3_u8): Likewise.
18832 (vst3_u16): Likewise.
18833 (vst3_u32): Likewise.
18834 (vst3_f16): Likewise.
18835 (vst3_f32): Likewise.
18836 (vst3_p64): Likewise.
18837 (vst3q_s8): Likewise.
18838 (vst3q_p8): Likewise.
18839 (vst3q_s16): Likewise.
18840 (vst3q_p16): Likewise.
18841 (vst3q_s32): Likewise.
18842 (vst3q_s64): Likewise.
18843 (vst3q_u8): Likewise.
18844 (vst3q_u16): Likewise.
18845 (vst3q_u32): Likewise.
18846 (vst3q_u64): Likewise.
18847 (vst3q_f16): Likewise.
18848 (vst3q_f32): Likewise.
18849 (vst3q_f64): Likewise.
18850 (vst3q_p64): Likewise.
18852 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
18854 * config/aarch64/arm_neon.h (vst4_s64): Use __builtin_memcpy
18855 instead of constructing __builtin_aarch64_simd_xi one vector
18857 (vst4_u64): Likewise.
18858 (vst4_f64): Likewise.
18859 (vst4_s8): Likewise.
18860 (vst4_p8): Likewise.
18861 (vst4_s16): Likewise.
18862 (vst4_p16): Likewise.
18863 (vst4_s32): Likewise.
18864 (vst4_u8): Likewise.
18865 (vst4_u16): Likewise.
18866 (vst4_u32): Likewise.
18867 (vst4_f16): Likewise.
18868 (vst4_f32): Likewise.
18869 (vst4_p64): Likewise.
18870 (vst4q_s8): Likewise.
18871 (vst4q_p8): Likewise.
18872 (vst4q_s16): Likewise.
18873 (vst4q_p16): Likewise.
18874 (vst4q_s32): Likewise.
18875 (vst4q_s64): Likewise.
18876 (vst4q_u8): Likewise.
18877 (vst4q_u16): Likewise.
18878 (vst4q_u32): Likewise.
18879 (vst4q_u64): Likewise.
18880 (vst4q_f16): Likewise.
18881 (vst4q_f32): Likewise.
18882 (vst4q_f64): Likewise.
18883 (vst4q_p64): Likewise.
18885 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
18887 * config/aarch64/arm_neon.h (vtbx4_s8): Use __builtin_memcpy
18888 instead of constructing __builtin_aarch64_simd_oi one vector
18890 (vtbx4_u8): Likewise.
18891 (vtbx4_p8): Likewise.
18893 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
18895 * config/aarch64/arm_neon.h (vtbl3_s8): Use __builtin_memcpy
18896 instead of constructing __builtin_aarch64_simd_oi one vector
18898 (vtbl3_u8): Likewise.
18899 (vtbl3_p8): Likewise.
18900 (vtbl4_s8): Likewise.
18901 (vtbl4_u8): Likewise.
18902 (vtbl4_p8): Likewise.
18904 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
18906 * config/aarch64/arm_neon.h (vqtbx2_s8): Use __builtin_memcpy
18907 instead of constructing __builtin_aarch64_simd_oi one vector
18909 (vqtbx2_u8): Likewise.
18910 (vqtbx2_p8): Likewise.
18911 (vqtbx2q_s8): Likewise.
18912 (vqtbx2q_u8): Likewise.
18913 (vqtbx2q_p8): Likewise.
18914 (vqtbx3_s8): Use __builtin_memcpy instead of constructing
18915 __builtin_aarch64_simd_ci one vector at a time.
18916 (vqtbx3_u8): Likewise.
18917 (vqtbx3_p8): Likewise.
18918 (vqtbx3q_s8): Likewise.
18919 (vqtbx3q_u8): Likewise.
18920 (vqtbx3q_p8): Likewise.
18921 (vqtbx4_s8): Use __builtin_memcpy instead of constructing
18922 __builtin_aarch64_simd_xi one vector at a time.
18923 (vqtbx4_u8): Likewise.
18924 (vqtbx4_p8): Likewise.
18925 (vqtbx4q_s8): Likewise.
18926 (vqtbx4q_u8): Likewise.
18927 (vqtbx4q_p8): Likewise.
18929 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
18931 * config/aarch64/arm_neon.h (vqtbl2_s8): Use __builtin_memcpy
18932 instead of constructing __builtin_aarch64_simd_oi one vector
18934 (vqtbl2_u8): Likewise.
18935 (vqtbl2_p8): Likewise.
18936 (vqtbl2q_s8): Likewise.
18937 (vqtbl2q_u8): Likewise.
18938 (vqtbl2q_p8): Likewise.
18939 (vqtbl3_s8): Use __builtin_memcpy instead of constructing
18940 __builtin_aarch64_simd_ci one vector at a time.
18941 (vqtbl3_u8): Likewise.
18942 (vqtbl3_p8): Likewise.
18943 (vqtbl3q_s8): Likewise.
18944 (vqtbl3q_u8): Likewise.
18945 (vqtbl3q_p8): Likewise.
18946 (vqtbl4_s8): Use __builtin_memcpy instead of constructing
18947 __builtin_aarch64_simd_xi one vector at a time.
18948 (vqtbl4_u8): Likewise.
18949 (vqtbl4_p8): Likewise.
18950 (vqtbl4q_s8): Likewise.
18951 (vqtbl4q_u8): Likewise.
18952 (vqtbl4q_p8): Likewise.
18954 2021-07-23 Haochen Gui <guihaoc@gcc.gnu.org>
18957 * config/rs6000/rs6000.md (cstore<mode>4): Fix wrong fall through.
18959 2021-07-22 Andrew Pinski <apinski@marvell.com>
18961 PR tree-optimization/10153
18962 * tree-tailcall.c (create_tailcall_accumulator):
18963 Don't call fold_convert as the type should be correct already.
18964 (tree_optimize_tail_calls_1): Use build_{one,zero}_cst instead
18965 of integer_{one,zero}_node for the call of create_tailcall_accumulator.
18967 2021-07-22 Aldy Hernandez <aldyh@redhat.com>
18969 * gimple-range-cache.cc (non_null_ref::adjust_range): Replace
18970 varying_p check for null/non-null check.
18972 2021-07-22 Andrew MacLeod <amacleod@redhat.com>
18974 PR tree-optimization/101511
18975 * value-relation.cc (relation_oracle::query_relation): Check if ssa1
18976 is in ssa2's equiv set, and don't trap if so.
18978 2021-07-22 Andrew MacLeod <amacleod@redhat.com>
18980 PR tree-optimization/101497
18981 * gimple-range-fold.cc (fold_using_range::range_of_cond_expr): Check
18984 2021-07-22 Andrew MacLeod <amacleod@redhat.com>
18986 PR tree-optimization/101496
18987 * vr-values.c (simplify_using_ranges::fold_cond): Call range_of_stmt
18988 first, then vrp_visit_cond_Stmt.
18990 2021-07-22 liuhongt <hongtao.liu@intel.com>
18992 * config/i386/i386-expand.c
18993 (ix86_broadcast_from_integer_constant): Rename to ..
18994 (ix86_broadcast_from_constant): .. this, and extend it to
18996 (ix86_expand_vector_move): Extend to float mode.
18997 * config/i386/i386-features.c
18998 (replace_constant_pool_with_broadcast): Remove.
18999 (remove_partial_avx_dependency_gate): Ditto.
19000 (constant_pool_broadcast): Ditto.
19001 (class pass_constant_pool_broadcast): Ditto.
19002 (make_pass_constant_pool_broadcast): Ditto.
19003 (remove_partial_avx_dependency): Adjust gate.
19004 * config/i386/i386-passes.def: Remove pass_constant_pool_broadcast.
19005 * config/i386/i386-protos.h
19006 (make_pass_constant_pool_broadcast): Remove.
19008 2021-07-22 liuhongt <hongtao.liu@intel.com>
19010 * config/i386/constraints.md (Wb): New constraint.
19012 * config/i386/i386.md (*ashlhi3_1): Extend to avx512 mask
19014 (*ashlqi3_1): Ditto.
19015 (*<insn><mode>3_1): Split to ..
19016 (*ashr<mode>3_1): this, ...
19017 (*lshr<mode>3_1): and this, also extend this pattern to avx512
19019 (*<insn><mode>3_1): Split to ..
19020 (*ashr<mode>3_1): this, ...
19021 (*lshrqi3_1): and this, also extend this pattern to avx512
19023 (*lshrhi3_1): And this, also extend this pattern to avx512
19025 * config/i386/sse.md (k<code><mode>): New define_split after
19026 it to convert generic shift pattern to mask shift ones.
19028 2021-07-21 Thomas Schwinge <thomas@codesourcery.com>
19029 Joseph Myers <joseph@codesourcery.com>
19030 Cesar Philippidis <cesar@codesourcery.com>
19032 * tree-core.h (omp_clause_code): Add 'OMP_CLAUSE_NOHOST'.
19033 * tree.c (omp_clause_num_ops, omp_clause_code_name, walk_tree_1):
19035 * tree-pretty-print.c (dump_omp_clause): Likewise.
19036 * omp-general.c (oacc_verify_routine_clauses): Likewise.
19037 * gimplify.c (gimplify_scan_omp_clauses)
19038 (gimplify_adjust_omp_clauses): Likewise.
19039 * tree-nested.c (convert_nonlocal_omp_clauses)
19040 (convert_local_omp_clauses): Likewise.
19041 * omp-low.c (scan_sharing_clauses): Likewise.
19042 * omp-offload.c (execute_oacc_device_lower): Update.
19044 2021-07-21 Martin Sebor <msebor@redhat.com>
19046 * tree-ssa-alias.c (walk_aliased_vdefs_1): Fix typos in a comment.
19048 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
19050 * config/rs6000/rs6000-gen-builtins.c (write_init_bif_table):
19053 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
19055 * config/rs6000/rs6000-gen-builtins.c (write_fntype): New
19057 (write_fntype_init): New stub function.
19058 (write_init_bif_table): Likewise.
19059 (write_init_ovld_table): New function.
19060 (write_init_file): Implement.
19062 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
19064 * config/rs6000/rs6000-gen-builtins.c
19065 (write_autogenerated_header): New function.
19066 (write_decls): Likewise.
19067 (write_extern_fntype): New callback function.
19068 (write_header_file): Implement.
19070 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
19072 * config/rs6000/rs6000-gen-builtins.c (write_defines_file):
19075 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
19077 * config/rs6000/rs6000-gen-builtins.c (complete_vector_type): New
19079 (complete_base_type): Likewise.
19080 (construct_fntype_id): Likewise.
19081 (parse_bif_entry): Call contruct_fntype_id.
19082 (parse_ovld_entry): Likewise.
19084 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
19086 * config/rs6000/rs6000-gen-builtins.c (ovld_stanza): New struct.
19087 (MAXOVLDSTANZAS): New macro.
19088 (ovld_stanzas): New variable.
19089 (curr_ovld_stanza): Likewise.
19090 (MAXOVLDS): New macro.
19091 (ovlddata): New struct.
19092 (ovlds): New variable.
19093 (curr_ovld): Likewise.
19094 (max_ovld_args): Likewise.
19095 (parse_ovld_entry): New function.
19096 (parse_ovld_stanza): Likewise.
19097 (parse_ovld): Implement.
19099 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
19101 * config/rs6000/rs6000-gen-builtins.c (parse_bif_attrs):
19104 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
19106 * config/rs6000/rs6000-gen-builtins.c (parse_args): New function.
19107 (parse_prototype): Implement.
19109 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
19111 * config/rs6000/rs6000-gen-builtins.c (bif_stanza): New enum.
19112 (curr_bif_stanza): New variable.
19113 (stanza_entry): New struct.
19114 (stanza_map): New initialized variable.
19115 (enable_string): Likewise.
19116 (fnkinds): New enum.
19117 (typelist): New struct.
19118 (attrinfo): Likewise.
19119 (MAXRESTROPNDS): New macro.
19120 (prototype): New struct.
19121 (MAXBIFS): New macro.
19122 (bifdata): New struct.
19123 (bifs): New variable.
19124 (curr_bif): Likewise.
19125 (bif_order): Likewise.
19126 (bif_index): Likewise.
19127 (fatal): New function.
19128 (stanza_name_to_stanza): Likewise.
19129 (parse_bif_attrs): New stub function.
19130 (parse_prototype): Likewise.
19131 (parse_bif_entry): New function.
19132 (parse_bif_stanza): Likewise.
19133 (parse_bif): Implement.
19134 (set_bif_order): New function.
19135 (create_bif_order): Implement.
19137 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
19139 * config/rs6000/rs6000-gen-builtins.c (rbtree.h): New #include.
19140 (num_bifs): New variable.
19141 (num_ovld_stanzas): Likewise.
19142 (num_ovlds): Likewise.
19143 (parse_codes): New enum.
19144 (bif_rbt): New variable.
19145 (ovld_rbt): Likewise.
19146 (fntype_rbt): Likewise.
19147 (bifo_rbt): Likewise.
19148 (parse_bif): New stub function.
19149 (create_bif_order): Likewise.
19150 (parse_ovld): Likewise.
19151 (write_header_file): Likewise.
19152 (write_init_file): Likewise.
19153 (write_defines_file): Likewise.
19154 (delete_output_files): New function.
19157 2021-07-21 H.J. Lu <hjl.tools@gmail.com>
19160 * config/i386/i386-builtin.def: Remove OPTION_MASK_ISA_SSE4_2
19161 from CRC32 _builtin functions.
19163 2021-07-21 Sebastian Huber <sebastian.huber@embedded-brains.de>
19165 * coverage.c (build_gcov_info_var_registration): Mark the object placed
19166 in the linker set as referenced so that it does not get optimized away.
19168 2021-07-21 Kito Cheng <kito.cheng@sifive.com>
19171 2021-07-20 Kito Cheng <kito.cheng@sifive.com>
19173 * config.gcc (riscv*-*-*): Detect which python is available.
19175 2021-07-21 Jakub Jelinek <jakub@redhat.com>
19177 PR middle-end/101535
19178 * gimplify.c (omp_check_private): Properly skip ORT_TARGET_DATA
19179 contexts in which decl isn't privatized and for ORT_TARGET return
19180 false if decl is mapped.
19182 2021-07-21 Richard Sandiford <richard.sandiford@arm.com>
19184 * gimple-loop-jam.c: Include tree-ssa-sccvn.h.
19185 (tree_loop_unroll_and_jam): Run value-numbering on a loop that
19186 has been successfully unrolled.
19188 2021-07-21 Richard Sandiford <richard.sandiford@arm.com>
19190 * tree-ssa-loop-manip.c (determine_exit_conditions): Return a null
19191 exit condition if no tail loop is needed, and if the original exit
19192 condition should therefore be kept as-is.
19193 (tree_transform_and_unroll_loop): Handle that case here too.
19195 2021-07-21 Kewen Lin <linkw@linux.ibm.com>
19197 * tree-data-ref.c (free_dependence_relations): Adjust to pass vec
19199 (free_data_refs): Likewise.
19200 * tree-data-ref.h (free_dependence_relations): Likewise.
19201 (free_data_refs): Likewise.
19202 * tree-predcom.c (struct chain): Use auto_vec instead of vec for
19204 (struct component): Likewise.
19205 (pcom_worker::pcom_worker): Adjust for auto_vec and renaming changes.
19206 (pcom_worker::~pcom_worker): Likewise.
19207 (pcom_worker::release_chain): Adjust as auto_vec changes.
19208 (pcom_worker::loop): Rename to ...
19209 (pcom_worker::m_loop): ... this.
19210 (pcom_worker::datarefs): Rename to ...
19211 (pcom_worker::m_datarefs): ... this. Use auto_vec instead of vec.
19212 (pcom_worker::dependences): Rename to ...
19213 (pcom_worker::m_dependences): ... this. Use auto_vec instead of vec.
19214 (pcom_worker::chains): Rename to ...
19215 (pcom_worker::m_chains): ... this. Use auto_vec instead of vec.
19216 (pcom_worker::looparound_phis): Rename to ...
19217 (pcom_worker::m_looparound_phis): ... this. Use auto_vec instead of
19219 (pcom_worker::cache): Rename to ...
19220 (pcom_worker::m_cache): ... this. Use auto_vec instead of vec.
19221 (pcom_worker::release_chain): Adjust for auto_vec changes.
19222 (pcom_worker::release_chains): Adjust for auto_vec and renaming
19224 (release_component): Remove.
19225 (release_components): Adjust for release_component removal.
19226 (component_of): Adjust to use vec.
19227 (merge_comps): Likewise.
19228 (pcom_worker::aff_combination_dr_offset): Adjust for renaming changes.
19229 (pcom_worker::determine_offset): Likewise.
19230 (class comp_ptrs): Remove.
19231 (pcom_worker::split_data_refs_to_components): Adjust for renaming
19232 changes, for comp_ptrs removal with auto_vec.
19233 (pcom_worker::suitable_component_p): Adjust for renaming changes.
19234 (pcom_worker::filter_suitable_components): Adjust for release_component
19236 (pcom_worker::valid_initializer_p): Adjust for renaming changes.
19237 (pcom_worker::find_looparound_phi): Likewise.
19238 (pcom_worker::add_looparound_copies): Likewise.
19239 (pcom_worker::determine_roots_comp): Likewise.
19240 (pcom_worker::single_nonlooparound_use): Likewise.
19241 (pcom_worker::execute_pred_commoning_chain): Likewise.
19242 (pcom_worker::execute_pred_commoning): Likewise.
19243 (pcom_worker::try_combine_chains): Likewise.
19244 (pcom_worker::prepare_initializers_chain): Likewise.
19245 (pcom_worker::prepare_initializers): Likewise.
19246 (pcom_worker::prepare_finalizers_chain): Likewise.
19247 (pcom_worker::prepare_finalizers): Likewise.
19248 (pcom_worker::tree_predictive_commoning_loop): Likewise.
19250 2021-07-20 Martin Sebor <msebor@redhat.com>
19252 PR middle-end/101397
19253 * builtins.c (gimple_call_return_array): Add argument. Correct
19254 offsets for memchr, mempcpy, stpcpy, and stpncpy.
19255 (compute_objsize_r): Adjust offset computation for argument returning
19258 2021-07-20 Martin Sebor <msebor@redhat.com>
19260 PR middle-end/101300
19261 * tree-ssa-uninit.c (check_defs): Handle UBSAN built-ins.
19263 2021-07-20 Jeff Law <jlaw@localhost.localdomain>
19265 * function.c (assign_parm_setup_block): Use adjust_address instead
19266 of change_address to preserve MEM_EXPR and friends.
19268 2021-07-20 Martin Sebor <msebor@redhat.com>
19270 * cfgloop.h (single_likely_exit): Adjust by-value argument to
19271 by-const-reference.
19272 * cfgloopanal.c (single_likely_exit): Same.
19273 * cgraph.h (struct cgraph_node): Same.
19274 * cgraphclones.c (cgraph_node::create_virtual_clone): Same.
19275 * genautomata.c (merge_states): Same.
19276 * genextract.c (VEC_char_to_string): Same.
19277 * genmatch.c (dt_node::gen_kids_1): Same.
19278 (walk_captures): Adjust by-value argument to by-reference.
19279 * gimple-ssa-store-merging.c (check_no_overlap): Adjust by-value argument
19280 to by-const-reference.
19281 * gimple.c (gimple_build_call_vec): Same.
19282 (gimple_build_call_internal_vec): Same.
19283 (gimple_build_switch): Same.
19284 (sort_case_labels): Same.
19285 (preprocess_case_label_vec_for_gimple): Adjust by-value argument to
19287 * gimple.h (gimple_build_call_vec): Adjust by-value argument to
19288 by-const-reference.
19289 (gimple_build_call_internal_vec): Same.
19290 (gimple_build_switch): Same.
19291 (sort_case_labels): Same.
19292 (preprocess_case_label_vec_for_gimple): Adjust by-value argument to
19294 * haifa-sched.c (calc_priorities): Adjust by-value argument to
19295 by-const-reference.
19296 (sched_init_luids): Same.
19297 (haifa_init_h_i_d): Same.
19298 * ipa-cp.c (ipa_get_indirect_edge_target_1): Same.
19299 (adjust_callers_for_value_intersection): Adjust by-value argument to
19301 (find_more_scalar_values_for_callers_subset): Adjust by-value argument to
19302 by-const-reference.
19303 (find_more_contexts_for_caller_subset): Same.
19304 (find_aggregate_values_for_callers_subset): Same.
19305 (copy_useful_known_contexts): Same.
19306 * ipa-fnsummary.c (remap_edge_summaries): Same.
19307 (remap_freqcounting_predicate): Same.
19308 * ipa-inline.c (add_new_edges_to_heap): Adjust by-value argument to
19310 * ipa-predicate.c (predicate::remap_after_inlining): Adjust by-value argument
19311 to by-const-reference.
19312 * ipa-predicate.h (predicate::remap_after_inlining): Same.
19313 * ipa-prop.c (ipa_find_agg_cst_for_param): Same.
19314 * ipa-prop.h (ipa_find_agg_cst_for_param): Same.
19315 * ira-build.c (ira_loop_tree_body_rev_postorder): Same.
19316 * read-rtl.c (add_overload_instance): Same.
19317 * rtl.h (native_decode_rtx): Same.
19318 (native_decode_vector_rtx): Same.
19319 * sched-int.h (sched_init_luids): Same.
19320 (haifa_init_h_i_d): Same.
19321 * simplify-rtx.c (native_decode_vector_rtx): Same.
19322 (native_decode_rtx): Same.
19323 * tree-call-cdce.c (gen_shrink_wrap_conditions): Same.
19324 (shrink_wrap_one_built_in_call_with_conds): Same.
19325 (shrink_wrap_conditional_dead_built_in_calls): Same.
19326 * tree-data-ref.c (create_runtime_alias_checks): Same.
19327 (compute_all_dependences): Same.
19328 * tree-data-ref.h (compute_all_dependences): Same.
19329 (create_runtime_alias_checks): Same.
19330 (index_in_loop_nest): Same.
19331 * tree-if-conv.c (mask_exists): Same.
19332 * tree-loop-distribution.c (class loop_distribution): Same.
19333 (loop_distribution::create_rdg_vertices): Same.
19334 (dump_rdg_partitions): Same.
19335 (debug_rdg_partitions): Same.
19336 (partition_contains_all_rw): Same.
19337 (loop_distribution::distribute_loop): Same.
19338 * tree-parloops.c (oacc_entry_exit_ok_1): Same.
19339 (oacc_entry_exit_single_gang): Same.
19340 * tree-ssa-loop-im.c (hoist_memory_references): Same.
19341 (loop_suitable_for_sm): Same.
19342 * tree-ssa-loop-niter.c (bound_index): Same.
19343 * tree-ssa-reassoc.c (update_ops): Same.
19344 (swap_ops_for_binary_stmt): Same.
19345 (rewrite_expr_tree): Same.
19346 (rewrite_expr_tree_parallel): Same.
19347 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Same.
19348 * tree-ssa-sccvn.h (ao_ref_init_from_vn_reference): Same.
19349 * tree-ssa-structalias.c (process_all_all_constraints): Same.
19350 (make_constraints_to): Same.
19351 (handle_lhs_call): Same.
19352 (find_func_aliases_for_builtin_call): Same.
19353 (sort_fieldstack): Same.
19354 (check_for_overlaps): Same.
19355 * tree-vect-loop-manip.c (vect_create_cond_for_align_checks): Same.
19356 (vect_create_cond_for_unequal_addrs): Same.
19357 (vect_create_cond_for_lower_bounds): Same.
19358 (vect_create_cond_for_alias_checks): Same.
19359 * tree-vect-slp-patterns.c (vect_validate_multiplication): Same.
19360 * tree-vect-slp.c (vect_analyze_slp_instance): Same.
19361 (vect_make_slp_decision): Same.
19362 (vect_slp_bbs): Same.
19363 (duplicate_and_interleave): Same.
19364 (vect_transform_slp_perm_load): Same.
19365 (vect_schedule_slp): Same.
19366 * tree-vectorizer.h (vect_transform_slp_perm_load): Same.
19367 (vect_schedule_slp): Same.
19368 (duplicate_and_interleave): Same.
19369 * tree.c (build_vector_from_ctor): Same.
19370 (build_vector): Same.
19371 (check_vector_cst): Same.
19372 (check_vector_cst_duplicate): Same.
19373 (check_vector_cst_fill): Same.
19374 (check_vector_cst_stepped): Same.
19375 * tree.h (build_vector_from_ctor): Same.
19377 2021-07-20 Jakub Jelinek <jakub@redhat.com>
19380 * config/rs6000/rs6000-protos.h (easy_altivec_constant): Change return
19381 type from bool to int.
19382 * config/rs6000/rs6000.c (vspltis_constant): Fix up handling the
19383 EASY_VECTOR_MSB case if either step or copies is not 1.
19384 (vspltis_shifted): Fix comment typo.
19385 (easy_altivec_constant): Change return type from bool to int, instead
19386 of returning true return byte size of the element mode that should be
19387 used to synthetize the constant.
19388 * config/rs6000/predicates.md (easy_vector_constant_msb): Require
19389 that vspltis_shifted is 0, handle the case where easy_altivec_constant
19390 assumes using different vector mode from CONST_VECTOR's mode.
19391 * config/rs6000/altivec.md (easy_vector_constant_msb splitter): Use
19392 easy_altivec_constant to determine mode in which -1 >> -1 should be
19393 performed, use rs6000_expand_vector_init instead of gen_vec_initv4sisi.
19395 2021-07-20 Richard Biener <rguenther@suse.de>
19398 * dwarf2out.h (dwarf_file_data): Add key member.
19399 * dwarf2out.c (dwarf_file_hasher::equal): Compare key.
19400 (dwarf_file_hasher::hash): Hash key.
19401 (lookup_filename): Remap the filename and store it in the
19402 filename member of dwarf_file_data when creating a new
19404 (file_name_acquire): Do not remap the filename again.
19405 (maybe_emit_file): Likewise.
19407 2021-07-20 Jonathan Wright <jonathan.wright@arm.com>
19409 * config/aarch64/aarch64-simd-builtins.def: Use two variant
19410 generators for all TBL/TBX intrinsics and rename to
19411 consistent forms: qtbl[1234] or qtbx[1234].
19412 * config/aarch64/aarch64-simd.md (aarch64_tbl1<mode>):
19414 (aarch64_qtbl1<mode>): This.
19415 (aarch64_tbx1<mode>): Rename to...
19416 (aarch64_qtbx1<mode>): This.
19417 (aarch64_tbl2v16qi): Delete.
19418 (aarch64_tbl3<mode>): Rename to...
19419 (aarch64_qtbl2<mode>): This.
19420 (aarch64_tbx4<mode>): Rename to...
19421 (aarch64_qtbx2<mode>): This.
19422 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_1): Use
19423 renamed qtbl1 and qtbl2 RTL patterns.
19424 * config/aarch64/arm_neon.h (vqtbl1_p8): Use renamed qtbl1
19426 (vqtbl1_s8): Likewise.
19427 (vqtbl1_u8): Likewise.
19428 (vqtbl1q_p8): Likewise.
19429 (vqtbl1q_s8): Likewise.
19430 (vqtbl1q_u8): Likewise.
19431 (vqtbx1_s8): Use renamed qtbx1 RTL pattern.
19432 (vqtbx1_u8): Likewise.
19433 (vqtbx1_p8): Likewise.
19434 (vqtbx1q_s8): Likewise.
19435 (vqtbx1q_u8): Likewise.
19436 (vqtbx1q_p8): Likewise.
19437 (vtbl1_s8): Use renamed qtbl1 RTL pattern.
19438 (vtbl1_u8): Likewise.
19439 (vtbl1_p8): Likewise.
19440 (vtbl2_s8): Likewise
19441 (vtbl2_u8): Likewise.
19442 (vtbl2_p8): Likewise.
19443 (vtbl3_s8): Use renamed qtbl2 RTL pattern.
19444 (vtbl3_u8): Likewise.
19445 (vtbl3_p8): Likewise.
19446 (vtbl4_s8): Likewise.
19447 (vtbl4_u8): Likewise.
19448 (vtbl4_p8): Likewise.
19449 (vtbx2_s8): Use renamed qtbx2 RTL pattern.
19450 (vtbx2_u8): Likewise.
19451 (vtbx2_p8): Likewise.
19452 (vqtbl2_s8): Use renamed qtbl2 RTL pattern.
19453 (vqtbl2_u8): Likewise.
19454 (vqtbl2_p8): Likewise.
19455 (vqtbl2q_s8): Likewise.
19456 (vqtbl2q_u8): Likewise.
19457 (vqtbl2q_p8): Likewise.
19458 (vqtbx2_s8): Use renamed qtbx2 RTL pattern.
19459 (vqtbx2_u8): Likewise.
19460 (vqtbx2_p8): Likewise.
19461 (vqtbx2q_s8): Likewise.
19462 (vqtbx2q_u8): Likewise.
19463 (vqtbx2q_p8): Likewise.
19464 (vtbx4_s8): Likewise.
19465 (vtbx4_u8): Likewise.
19466 (vtbx4_p8): Likewise.
19468 2021-07-20 Uroš Bizjak <ubizjak@gmail.com>
19471 * config/i386/sync.md (define_peephole2 atomic_storedi_fpu):
19473 (define_peephole2 atomic_loaddi_fpu): Ditto.
19475 2021-07-20 Kito Cheng <kito.cheng@sifive.com>
19477 * config.gcc (riscv*-*-*): Detect which python is available.
19479 2021-07-20 Kewen Lin <linkw@linux.ibm.com>
19481 * config/rs6000/vsx.md (mulhs_<mode>): Rename to...
19482 (smul<mode>3_highpart): ... this.
19483 (mulhu_<mode>): Rename to...
19484 (umul<mode>3_highpart): ... this.
19485 * config/rs6000/rs6000-builtin.def (MULHS_V2DI, MULHS_V4SI,
19486 MULHU_V2DI, MULHU_V4SI): Adjust.
19488 2021-07-20 Kewen Lin <linkw@linux.ibm.com>
19490 PR tree-optimization/100696
19491 * internal-fn.c (first_commutative_argument): Add info for IFN_MULH.
19492 * internal-fn.def (IFN_MULH): New internal function.
19493 * tree-vect-patterns.c (vect_recog_mulhs_pattern): Add support to
19494 recog normal multiply highpart as IFN_MULH.
19495 * config/i386/i386.c (ix86_add_stmt_cost): Adjust for combined
19498 2021-07-19 Indu Bhagat <indu.bhagat@oracle.com>
19500 * config/elfos.h (CTF_DEBUGGING_INFO): New definition.
19501 (BTF_DEBUGGING_INFO): Likewise.
19502 * doc/tm.texi.in: Document the new macros.
19503 * doc/tm.texi: Regenerated.
19504 * toplev.c: Guard initialization of debug hooks.
19506 2021-07-19 Indu Bhagat <indu.bhagat@oracle.com>
19508 * flags.h (ctf_debuginfo_p): New function declaration.
19509 * opts.c (ctf_debuginfo_p): New function definition.
19511 2021-07-19 Andrew Stubbs <ams@codesourcery.com>
19514 * config/gcn/gcn-hsa.h (DRIVER_SELF_SPECS): New.
19515 (ASM_SPEC): Set -mattr for xnack and sram-ecc.
19516 * config/gcn/gcn-opts.h (enum sram_ecc_type): New.
19517 * config/gcn/gcn-valu.md: Add a warning comment.
19518 * config/gcn/gcn.c (gcn_option_override): Add "sorry" for -mxnack.
19519 (output_file_start): Add xnack and sram-ecc state to ".amdgcn_target".
19520 * config/gcn/gcn.md: Add a warning comment.
19521 * config/gcn/gcn.opt: Add -mxnack and -msram-ecc.
19522 * config/gcn/mkoffload.c (EF_AMDGPU_MACH_AMDGCN_GFX908): Remove
19524 (EF_AMDGPU_XNACK): New.
19525 (EF_AMDGPU_SRAM_ECC): New.
19527 (copy_early_debug_info): Use elf_flags.
19528 (main): Handle -mxnack and -msram-ecc options.
19529 * doc/invoke.texi: Document -mxnack and -msram-ecc.
19531 2021-07-19 Andrew Pinski <apinski@marvell.com>
19534 * config/aarch64/aarch64.md (csneg3_uxtw_insn): Rename to ...
19535 (*cs<neg_not_cs>3_uxtw_insn4): and extend to NEG_NOT.
19537 2021-07-19 Richard Biener <rguenther@suse.de>
19539 PR tree-optimization/101505
19540 * tree-vect-patterns.c (vect_determine_precisions): Walk
19541 PHIs also for loop vectorization.
19543 2021-07-19 Richard Biener <rguenther@suse.de>
19545 * gimple.h (gimple_expr_type): Remove.
19546 * doc/gimple.texi: Remove gimple_expr_type documentation.
19548 2021-07-19 Richard Biener <rguenther@suse.de>
19550 * tree-ssa-sccvn.c (vn_reference_eq): Handle NULL vr->type.
19551 (ao_ref_init_from_vn_reference): Likewise.
19552 (fully_constant_reference): Likewise.
19553 (vn_reference_lookup_call): Do not set vr->type to random
19555 * tree-ssa-pre.c (compute_avail): Do not try to PRE calls
19557 * tree-vect-generic.c (expand_vector_piecewise): Pass in
19558 whether we expanded parallel.
19559 (expand_vector_parallel): Adjust.
19560 (expand_vector_addition): Likewise.
19561 (expand_vector_comparison): Likewise.
19562 (expand_vector_operation): Likewise.
19563 (expand_vector_scalar_condition): Likewise.
19564 (expand_vector_conversion): Likewise.
19566 2021-07-19 Richard Biener <rguenther@suse.de>
19568 * tree-vrp.c (register_edge_assert_for_2): Use the
19570 (vrp_folder::fold_predicate_in): Likewise.
19571 * vr-values.c (gimple_assign_nonzero_p): Likewise.
19572 (vr_values::extract_range_from_comparison): Likewise.
19573 (vr_values::extract_range_from_ubsan_builtin): Use the
19574 type of the first operand.
19575 (vr_values::extract_range_basic): Push down type
19576 computation, use the appropriate LHS.
19577 (vr_values::extract_range_from_assignment): Use the
19580 2021-07-18 H.J. Lu <hjl.tools@gmail.com>
19583 * common/config/i386/i386-common.c (ix86_handle_option): For
19584 -mgeneral-regs-only, enable the GPR only instructions which are
19585 enabled implicitly by SSE ISAs unless they have been disabled
19588 2021-07-18 H.J. Lu <hjl.tools@gmail.com>
19591 * config/i386/i386.c (ix86_check_avx_upper_stores): Moved before
19592 ix86_avx_u128_mode_needed.
19593 (ix86_avx_u128_mode_needed): Return AVX_U128_DIRTY if callee
19594 returns AVX register.
19596 2021-07-17 Jan Hubicka <hubicka@ucw.cz>
19598 * tree-ssa-structalias.c (handle_rhs_call): Support EAF_NOT_RETURNED.
19599 (handle_const_call): Liekise
19600 (handle_pure_call): Liekise
19602 2021-07-17 Andrew MacLeod <amacleod@redhat.com>
19604 PR tree-optimization/96542
19605 * range-op.cc (range_operator::wi_fold_in_parts): New.
19606 (range_operator::fold_range): Call wi_fold_in_parts.
19607 (operator_lshift::wi_fold): Fix broken lshift by [0,0].
19608 * range-op.h (wi_fold_in_parts): Add prototype.
19610 2021-07-16 David Malcolm <dmalcolm@redhat.com>
19612 * doc/analyzer.texi: Add __analyzer_dump_state.
19614 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
19616 * config/rs6000/rbtree.c: New file.
19617 * config/rs6000/rbtree.h: New file.
19619 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
19621 * config/rs6000/rs6000-gen-builtins.c (restriction): New enum.
19622 (typeinfo): Add restr field.
19623 (match_bracketed_pair): New function.
19624 (match_const_restriction): Implement.
19626 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
19628 * config/rs6000/rs6000-gen-builtins.c (match_basetype): Implement.
19630 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
19632 * config/rs6000/rs6000-gen-builtins.c (void_status): New enum.
19633 (basetype): Likewise.
19634 (typeinfo): Likewise.
19635 (handle_pointer): New function.
19636 (match_basetype): New stub function.
19637 (match_const_restriction): Likewise.
19638 (match_type): New function.
19640 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
19642 * config/rs6000/rs6000-gen-builtins.c (consume_whitespace): New
19644 (advance_line): Likewise.
19645 (safe_inc_pos): Likewise.
19646 (match_identifier): Likewise.
19647 (match_integer): Likewise.
19648 (match_to_right_bracket): Likewise.
19650 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
19652 * config/rs6000/rs6000-gen-builtins.c (bif_file): New variable.
19653 (ovld_file): Likewise.
19654 (header_file): Likewise.
19655 (init_file): Likewise.
19656 (defines_file): Likewise.
19657 (pgm_path): Likewise.
19658 (bif_path): Likewise.
19659 (ovld_path): Likewise.
19660 (header_path): Likewise.
19661 (init_path): Likewise.
19662 (defines_path): Likewise.
19663 (LINELEN): New macro.
19664 (linebuf): New variable.
19668 (bif_diag): New function.
19669 (ovld_diag): Likewise.
19671 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
19673 * config/rs6000/rs6000-builtin-new.def: New.
19674 * config/rs6000/rs6000-overload.def: New.
19676 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
19678 * config/rs6000/rs6000-gen-builtins.c: New.
19680 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
19682 * Makefile.in (EXTRA_GTYPE_DEPS): New variable.
19683 (s-gtype): Depend on EXTRA_GTYPE_DEPS.
19684 * gengtype-state.c (state_writer::write_state_file_list): Add a
19685 parameter to the fileslist expression for the number of build
19687 (read_state_files_list): Detect build headers and strip the
19688 initial "./" or ".\" from their names.
19689 * gengtype.c (build_headers): New global variable.
19690 (num_build_headers): Likewise.
19691 (open_base_files): Emit #include for each build header.
19692 (main): Detect and count build headers.
19693 * gengtype.h (build_headers): New extern variable.
19694 (num_build_headers): Likewise.
19696 2021-07-16 Richard Biener <rguenther@suse.de>
19698 * gimple-ssa-store-merging.c (verify_symbolic_number_p): Use
19699 the type of the LHS.
19700 (find_bswap_or_nop_1): Likewise.
19701 (find_bswap_or_nop): Likewise.
19702 * tree-vectorizer.h (vect_get_smallest_scalar_type): Adjust
19704 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
19705 Remove unused parameters, pass in the scalar type. Fix
19706 internal store function handling.
19707 * tree-vect-stmts.c (vect_analyze_stmt): Remove assert.
19708 (vect_get_vector_types_for_stmt): Move down check for
19709 existing vector stmt after we've determined a scalar type.
19710 Pass down the used scalar type to vect_get_smallest_scalar_type.
19711 * tree-vect-generic.c (expand_vector_condition): Use
19712 the type of the LHS.
19713 (expand_vector_scalar_condition): Likewise.
19714 (expand_vector_operations_1): Likewise.
19715 * tree-vect-patterns.c (vect_widened_op_tree): Likewise.
19716 (vect_recog_dot_prod_pattern): Likewise.
19717 (vect_recog_sad_pattern): Likewise.
19718 (vect_recog_widen_op_pattern): Likewise.
19719 (vect_recog_widen_sum_pattern): Likewise.
19720 (vect_recog_mixed_size_cond_pattern): Likewise.
19722 2021-07-16 Jan Hubicka <hubicka@ucw.cz>
19724 * ipa-modref.c (struct escape_entry): Use eaf_fleags_t.
19725 (dump_eaf_flags): Dump EAF_NOT_RETURNED
19726 (eaf_flags_useful_p): Use eaf_fleags_t; handle const functions
19727 and EAF_NOT_RETURNED.
19728 (modref_summary::useful_p): Likewise.
19729 (modref_summary_lto::useful_p): Likewise.
19730 (struct) modref_summary_lto: Use eaf_fleags_t.
19731 (deref_flags): Handle EAF_NOT_RETURNED.
19732 (struct escape_point): Use min_flags.
19733 (modref_lattice::init): Add EAF_NOT_RETURNED.
19734 (merge_call_lhs_flags): Ignore EAF_NOT_RETURNED functions
19735 (analyze_ssa_name_flags): Clear EAF_NOT_RETURNED on return;
19737 (analyze_parms): Also analyze const functions; update conition on
19739 (modref_write): Update streaming.
19740 (read_section): Update streaming.
19741 (remap_arg_flags): Use eaf_flags_t.
19742 (modref_merge_call_site_flags): Hanlde EAF_NOT_RETURNED.
19743 * ipa-modref.h: (eaf_flags_t): New typedef.
19744 (struct modref_summary): Use eaf_flags_t.
19745 * tree-core.h (EAF_NOT_RETURNED): New constant.
19747 2021-07-16 Richard Biener <rguenther@suse.de>
19749 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Use
19750 the type of the LHS.
19751 (gimple_assign_nonnegative_warnv_p): Likewise.
19752 (gimple_call_nonnegative_warnv_p): Likewise. Return false
19753 if the call has no LHS.
19754 * gimple.c (gimple_could_trap_p_1): Use the type of the LHS.
19755 * tree-eh.c (stmt_could_throw_1_p): Likewise.
19756 * tree-inline.c (insert_init_stmt): Likewise.
19757 * tree-ssa-loop-niter.c (get_val_for): Likewise.
19758 * tree-outof-ssa.c (ssa_is_replaceable_p): Use the type of
19760 * tree-ssa-sccvn.c (init_vn_nary_op_from_stmt): Take a
19761 gassign *. Use the type of the lhs.
19762 (vn_nary_op_lookup_stmt): Adjust.
19763 (vn_nary_op_insert_stmt): Likewise.
19765 2021-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
19767 * config/s390/predicates.md (bras_sym_operand): Accept all
19768 functions in 64-bit mode, use UNSPEC_PLT31.
19769 (larl_operand): Use UNSPEC_PLT31.
19770 * config/s390/s390.c (s390_loadrelative_operand_p): Likewise.
19771 (legitimize_pic_address): Likewise.
19772 (s390_emit_tls_call_insn): Mark __tls_get_offset as function,
19774 (s390_delegitimize_address): Use UNSPEC_PLT31.
19775 (s390_output_addr_const_extra): Likewise.
19776 (print_operand): Add @PLT to TLS calls, handle %K.
19777 (s390_function_profiler): Mark __fentry__/_mcount as function,
19778 use %K, use UNSPEC_PLT31.
19779 (s390_output_mi_thunk): Use only UNSPEC_GOT, use %K.
19780 (s390_emit_call): Use UNSPEC_PLT31.
19781 (s390_emit_tpf_eh_return): Mark __tpf_eh_return as function.
19782 * config/s390/s390.md (UNSPEC_PLT31): Rename from UNSPEC_PLT.
19783 (*movdi_64): Use %K.
19784 (reload_base_64): Likewise.
19785 (*sibcall_brc): Likewise.
19786 (*sibcall_brcl): Likewise.
19787 (*sibcall_value_brc): Likewise.
19788 (*sibcall_value_brcl): Likewise.
19790 (*brasl): Likewise.
19791 (*bras_r): Likewise.
19792 (*brasl_r): Likewise.
19793 (*bras_tls): Likewise.
19794 (*brasl_tls): Likewise.
19795 (main_base_64): Likewise.
19796 (reload_base_64): Likewise.
19797 (@split_stack_call<mode>): Likewise.
19799 2021-07-16 Richard Biener <rguenther@suse.de>
19801 PR tree-optimization/101467
19802 * tree-vect-stmts.c (vect_gen_while): Properly guard
19803 make_temp_ssa_name usage.
19805 2021-07-16 Cooper Qu <cooper.qu@linux.alibaba.com>
19807 * config.gcc: Don't use forked print-sysroot-suffix.sh and
19808 t-sysroot-suffix for C-SKY.
19809 * config/csky/print-sysroot-suffix.sh: Delete.
19810 * config/csky/t-csky-linux: Delete.
19811 * config/csky/t-sysroot-suffix: Define MULTILIB_DIRNAMES
19812 instead of CSKY_MULTILIB_DIRNAMES.
19814 2021-07-16 Richard Biener <rguenther@suse.de>
19816 * tree-vect-loop.c (vect_transform_cycle_phi): Correct sign
19817 conversion issues with the partial reduction of the reused
19818 vector accumulator.
19820 2021-07-16 Richard Biener <rguenther@suse.de>
19822 * config/i386/i386-options.c (ix86_option_override_internal): Set
19823 param_vect_partial_vector_usage to zero if not set.
19825 2021-07-15 Uroš Bizjak <ubizjak@gmail.com>
19828 * config/i386/i386.h (VALID_SSE_REG_MODE): Add TDmode.
19829 (VALID_INT_MODE_P): Add SDmode and DDmode.
19830 Add TDmode for TARGET_64BIT.
19831 (VALID_DFP_MODE_P): Remove.
19832 * config/i386/i386.c (ix86_hard_regno_mode_ok):
19833 Do not use VALID_DFP_MODE_P.
19835 2021-07-15 Andrew MacLeod <amacleod@redhat.com>
19837 * gimple-range-fold.cc (adjust_pointer_diff_expr): Use
19839 (fold_using_range::fold_stmt): Ditto.
19840 (fold_using_range::range_of_range_op): Ditto.
19841 (fold_using_range::range_of_phi): Ditto.
19842 (fold_using_range::range_of_call): Ditto.
19843 (fold_using_range::range_of_builtin_ubsan_call): Ditto.
19844 (fold_using_range::range_of_builtin_call): Ditto.
19845 (fold_using_range::range_of_cond_expr): Ditto.
19846 * gimple-range-fold.h (gimple_range_type): New.
19848 2021-07-15 Martin Sebor <msebor@redhat.com>
19850 PR middle-end/97027
19851 * tree-ssa-strlen.c (handle_assign): New function.
19852 (maybe_warn_overflow): Add argument.
19853 (nonzero_bytes_for_type): New function.
19854 (count_nonzero_bytes): Handle more tree types. Call
19855 nonzero_bytes_for_tye.
19856 (count_nonzero_bytes): Handle types.
19857 (handle_store): Handle stores from function calls.
19858 (strlen_check_and_optimize_call): Move code to handle_assign. Call
19859 it for assignments from function calls.
19861 2021-07-15 David Malcolm <dmalcolm@redhat.com>
19866 * doc/invoke.texi: Add -Wanalyzer-use-of-uninitialized-value.
19868 2021-07-15 David Malcolm <dmalcolm@redhat.com>
19870 * doc/invoke.texi (-fdump-analyzer-exploded-paths): New.
19872 2021-07-15 Martin Sebor <msebor@redhat.com>
19876 * fold-const.c (operand_compare::operand_equal_p): Handle OEP_DECL_NAME.
19877 (operand_compare::verify_hash_value): Same.
19878 * tree-core.h (OEP_DECL_NAME): New.
19880 2021-07-15 Martin Jambor <mjambor@suse.cz>
19882 * profile-count.h (profile_count::value): Change the return type to
19884 * gimple-pretty-print.c (dump_gimple_bb_header): Adjust print
19886 * tree-cfg.c (dump_function_to_file): Likewise.
19888 2021-07-15 Bill Schmidt <wschmidt@linux.ibm.com>
19891 * config/rs6000/rs6000-p8swap.c (has_part_mult): New.
19892 (rs6000_analyze_swaps): Insns containing a subreg of a mult are
19895 2021-07-15 Richard Biener <rguenther@suse.de>
19897 * tree-vectorizer.h (vect_gen_while): Match up with
19898 vect_gen_while_not.
19899 * tree-vect-stmts.c (vect_gen_while): Adjust API to that
19900 of vect_gen_while_not.
19901 (vect_gen_while_not): Adjust.
19902 * tree-vect-loop-manip.c (vect_set_loop_controls_directly): Likewise.
19904 2021-07-15 Aldy Hernandez <aldyh@redhat.com>
19906 * gimple-range-cache.cc (non_null_ref::adjust_range): New.
19907 (ranger_cache::range_of_def): Call adjust_range.
19908 (ranger_cache::entry_range): Same.
19909 * gimple-range-cache.h (non_null_ref::adjust_range): New.
19910 * gimple-range.cc (gimple_ranger::range_of_expr): Call
19912 (gimple_ranger::range_on_entry): Same.
19914 2021-07-15 Tamar Christina <tamar.christina@arm.com>
19917 2021-07-14 Tamar Christina <tamar.christina@arm.com>
19919 * config/arm/neon.md (<sup>dot_prod<vsi2qi>): Drop statements.
19921 2021-07-15 Tamar Christina <tamar.christina@arm.com>
19924 2021-07-14 Tamar Christina <tamar.christina@arm.com>
19926 * config/aarch64/aarch64-simd-builtins.def (udot, sdot): Rename to...
19927 (sdot_prod, udot_prod): ...These.
19928 * config/aarch64/aarch64-simd.md (<sur>dot_prod<vsi2qi>): Remove.
19929 (aarch64_<sur>dot<vsi2qi>): Rename to...
19930 (<sur>dot_prod<vsi2qi>): ...This.
19931 * config/aarch64/arm_neon.h (vdot_u32, vdotq_u32, vdot_s32, vdotq_s32):
19934 2021-07-15 Jakub Jelinek <jakub@redhat.com>
19936 PR middle-end/101437
19937 * gimplify.c (gimplify_expr): Throw away volatile reads from empty
19938 types even if they have non-BLKmode TYPE_MODE.
19940 2021-07-15 Richard Biener <rguenther@suse.de>
19943 * gcc.c (process_command): Process -gtoggle like process_options
19944 would after parsing options.
19946 2021-07-15 Trevor Saunders <tbsaunde@tbsaunde.org>
19948 * cfgexpand.c (expand_asm_loc): Adjust.
19949 (expand_asm_stmt): Likewise.
19950 * config/arm/aarch-common-protos.h (arm_md_asm_adjust): Likewise.
19951 * config/arm/aarch-common.c (arm_md_asm_adjust): Likewise.
19952 * config/arm/arm.c (thumb1_md_asm_adjust): Likewise.
19953 * config/avr/avr.c (avr_md_asm_adjust): Likewise.
19954 * config/cris/cris.c (cris_md_asm_adjust): Likewise.
19955 * config/i386/i386.c (ix86_md_asm_adjust): Likewise.
19956 * config/mn10300/mn10300.c (mn10300_md_asm_adjust): Likewise.
19957 * config/nds32/nds32.c (nds32_md_asm_adjust): Likewise.
19958 * config/pdp11/pdp11.c (pdp11_md_asm_adjust): Likewise.
19959 * config/rs6000/rs6000.c (rs6000_md_asm_adjust): Likewise.
19960 * config/s390/s390.c (s390_md_asm_adjust): Likewise.
19961 * config/vax/vax.c (vax_md_asm_adjust): Likewise.
19962 * config/visium/visium.c (visium_md_asm_adjust): Likewise.
19963 * doc/tm.texi: Regenerate.
19964 * target.def: Add location argument to md_asm_adjust.
19966 2021-07-15 Trevor Saunders <tbsaunde@tbsaunde.org>
19968 * tree-diagnostic.c (diagnostic_report_current_function): Use the
19969 diagnostic's location, not input_location.
19971 2021-07-15 Trevor Saunders <tbsaunde@tbsaunde.org>
19973 * cfgexpand.c (tree_conflicts_with_clobbers_p): Pass location to
19975 (expand_asm_stmt): Likewise.
19977 2021-07-14 Peter Bergner <bergner@linux.ibm.com>
19979 * config/rs6000/rs6000.c (adjacent_mem_locations): Return the lower
19980 addressed memory rtx, if any.
19981 (rs6000_split_multireg_move): Fix code formatting.
19982 Handle MMA build built-ins with operands in adjacent memory locations.
19984 2021-07-14 Peter Bergner <bergner@linux.ibm.com>
19986 * config/rs6000/rs6000.c (rs6000_split_multireg_move): Move to later
19989 2021-07-14 Jason Merrill <jason@redhat.com>
19991 * sel-sched-ir.h (get_all_loop_exits): Use auto_vec.
19993 2021-07-14 Jason Merrill <jason@redhat.com>
19995 * doc/invoke.texi: -fdelete-dead-exceptions is on by default for
19998 2021-07-14 Tamar Christina <tamar.christina@arm.com>
20000 * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
20001 Remove erroneous line.
20003 2021-07-14 Andrew MacLeod <amacleod@redhat.com>
20005 * params.opt (param_evrp_mode): Change default.
20007 2021-07-14 Tamar Christina <tamar.christina@arm.com>
20009 * config/aarch64/aarch64-simd-builtins.def (udot, sdot): Rename to...
20010 (sdot_prod, udot_prod): ...These.
20011 * config/aarch64/aarch64-simd.md (<sur>dot_prod<vsi2qi>): Remove.
20012 (aarch64_<sur>dot<vsi2qi>): Rename to...
20013 (<sur>dot_prod<vsi2qi>): ...This.
20014 * config/aarch64/arm_neon.h (vdot_u32, vdotq_u32, vdot_s32, vdotq_s32):
20017 2021-07-14 Tamar Christina <tamar.christina@arm.com>
20019 * config/arm/neon.md (<sup>dot_prod<vsi2qi>): Drop statements.
20021 2021-07-14 Tamar Christina <tamar.christina@arm.com>
20023 * doc/sourcebuild.texi (arm_v8_2a_i8mm_neon_hw): Document.
20025 2021-07-14 Tamar Christina <tamar.christina@arm.com>
20027 * config/arm/neon.md (usdot_prod<vsi2qi>): New.
20029 2021-07-14 Tamar Christina <tamar.christina@arm.com>
20031 * config/aarch64/aarch64-simd.md (aarch64_usdot<vsi2qi>): Rename to...
20032 (usdot_prod<vsi2qi>): ... This.
20033 * config/aarch64/aarch64-simd-builtins.def (usdot): Rename to...
20034 (usdot_prod): ...This.
20035 * config/aarch64/arm_neon.h (vusdot_s32, vusdotq_s32): Likewise.
20036 * config/aarch64/aarch64-sve.md (@aarch64_<sur>dot_prod<vsi2qi>):
20038 (@<sur>dot_prod<vsi2qi>): ...This.
20039 * config/aarch64/aarch64-sve-builtins-base.cc
20040 (svusdot_impl::expand): Use it.
20042 2021-07-14 Tamar Christina <tamar.christina@arm.com>
20044 * optabs.def (usdot_prod_optab): New.
20045 * doc/md.texi: Document it and clarify other dot prod optabs.
20046 * optabs-tree.h (enum optab_subtype): Add optab_vector_mixed_sign.
20047 * optabs-tree.c (optab_for_tree_code): Support usdot_prod_optab.
20048 * optabs.c (expand_widen_pattern_expr): Likewise.
20049 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
20050 * tree-vect-loop.c (vectorizable_reduction): Query dot-product kind.
20051 * tree-vect-patterns.c (vect_supportable_direct_optab_p): Take optional
20053 (vect_widened_op_tree): Optionally ignore
20055 (vect_recog_dot_prod_pattern): Support usdot_prod_optab.
20057 2021-07-14 H.J. Lu <hjl.tools@gmail.com>
20060 * config/i386/driver-i386.c (host_detect_local_cpu): Check
20061 "arch [32|64]" and "tune [32|64]" for 32-bit and 64-bit codegen.
20062 Enable UINTR only for 64-bit codegen.
20063 * config/i386/i386-options.c
20064 (ix86_option_override_internal::DEF_PTA): Skip PTA_UINTR if not
20066 * config/i386/i386.h (ARCH_ARG): New.
20067 (CC1_CPU_SPEC): Pass "[arch|tune] 32" for 32-bit codegen and
20068 "[arch|tune] 64" for 64-bit codegen.
20070 2021-07-14 Richard Biener <rguenther@suse.de>
20072 PR tree-optimization/101445
20073 * tree-vect-stmts.c (vectorizable_load): Do the gap adjustment
20074 of the IV in the correct direction for negative stride
20077 2021-07-14 Jakub Jelinek <jakub@redhat.com>
20080 * godump.c (godump_str_hash): New type.
20081 (godump_container::pot_dummy_types): Use string_hash instead of
20082 ptr_hash in the hash_set.
20084 2021-07-14 Richard Biener <rguenther@suse.de>
20086 * tree-vect-loop.c (vect_find_reusable_accumulator): Handle
20087 vector types where the old vector type has a multiple of
20088 the new vector type elements.
20089 (vect_create_partial_epilog): New function, split out from...
20090 (vect_create_epilog_for_reduction): ... here.
20091 (vect_transform_cycle_phi): Reduce the re-used accumulator
20092 to the new vector type.
20094 2021-07-14 Alexandre Oliva <oliva@adacore.com>
20096 * tree-ssa-alias.c (attr_fnspec::verify): Fix index in
20097 non-'t'-sized arg check.
20099 2021-07-14 Alexandre Oliva <oliva@adacore.com>
20101 * tree-cfg.c (cleanup_dead_labels_eh): Update
20102 post_landing_pad label upon change of landing pad block's
20104 (cleanup_dead_labels): Check that a removed label is not that
20107 2021-07-13 Jonathan Wright <jonathan.wright@arm.com>
20109 * combine.c (combine_simplify_rtx): Add vec_select -> subreg
20111 * config/aarch64/aarch64.md (*zero_extend<SHORT:mode><GPI:mode>2_aarch64):
20112 Add Neon to general purpose register case for zero-extend
20114 * config/arm/vfp.md (*arm_movsi_vfp): Remove "*" from *t -> r
20115 case to prevent some cases opting to go through memory.
20116 * cse.c (fold_rtx): Add vec_select -> subreg simplification.
20117 * rtl.c (rtvec_series_p): Define predicate to determine
20118 whether a vector contains a linear series of integers.
20119 * rtl.h (rtvec_series_p): Define.
20120 * rtlanal.c (vec_series_lowpart_p): Define predicate to
20121 determine if a vector selection is equivalent to the low part
20123 * rtlanal.h (vec_series_lowpart_p): Define.
20124 * simplify-rtx.c (simplify_context::simplify_binary_operation_1):
20125 Add vec_select -> subreg simplification.
20127 2021-07-13 Paul A. Clarke <pc@us.ibm.com>
20129 * config/rs6000/smmintrin.h (_mm_testz_si128, _mm_testc_si128,
20130 _mm_testnzc_si128, _mm_test_all_ones, _mm_test_all_zeros,
20131 _mm_test_mix_ones_zeros): New.
20133 2021-07-13 Roger Sayle <roger@nextmovesoftware.com>
20134 Richard Biener <rguenther@suse.de>
20136 * gimple.c (gimple_could_trap_p_1): Make S argument a
20137 "const gimple*". Preserve constness in call to
20138 gimple_asm_volatile_p.
20139 (gimple_could_trap_p): Make S argument a "const gimple*".
20140 * gimple.h (gimple_could_trap_p_1, gimple_could_trap_p):
20141 Update function prototypes.
20143 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
20145 * tree-vectorizer.h (vect_reusable_accumulator): New structure.
20146 (_loop_vec_info::main_loop_edge): New field.
20147 (_loop_vec_info::skip_main_loop_edge): Likewise.
20148 (_loop_vec_info::skip_this_loop_edge): Likewise.
20149 (_loop_vec_info::reusable_accumulators): Likewise.
20150 (_stmt_vec_info::reduc_scalar_results): Likewise.
20151 (_stmt_vec_info::reused_accumulator): Likewise.
20152 (vect_get_main_loop_result): Declare.
20153 * tree-vectorizer.c (vec_info::new_stmt_vec_info): Initialize
20154 reduc_scalar_inputs.
20155 (vec_info::free_stmt_vec_info): Free reduc_scalar_inputs.
20156 * tree-vect-loop-manip.c (vect_get_main_loop_result): New function.
20157 (vect_do_peeling): Fill an epilogue loop's main_loop_edge,
20158 skip_main_loop_edge and skip_this_loop_edge fields.
20159 * tree-vect-loop.c (INCLUDE_ALGORITHM): Define.
20160 (vect_emit_reduction_init_stmts): New function.
20161 (get_initial_def_for_reduction): Use it.
20162 (get_initial_defs_for_reduction): Likewise. Change the vinfo
20163 parameter to a loop_vec_info.
20164 (vect_create_epilog_for_reduction): Store the scalar results
20165 in the reduc_info. If an epilogue loop is reusing an accumulator
20166 from the main loop, and if the epilogue loop can also be skipped,
20167 try to place the reduction code in the join block. Record
20168 accumulators that could potentially be reused by epilogue loops.
20169 (vect_transform_cycle_phi): When vectorizing epilogue loops,
20170 try to reuse accumulators from the main loop. Record the initial
20171 value in reduc_info for non-SLP reductions too.
20173 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
20175 * tree-vect-loop.c (get_initial_def_for_reduction): Remove
20176 adjustment handling. Take the neutral value as an argument,
20177 in place of the code argument.
20178 (vect_transform_cycle_phi): Update accordingly. Handle the
20179 initial values of cond reductions separately from code reductions.
20180 Choose the adjustment here rather than in
20181 get_initial_def_for_reduction. Sink the splat of vec_initial_def.
20183 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
20185 * tree-vect-loop.c (neutral_op_for_slp_reduction): Replace with...
20186 (neutral_op_for_reduction): ...this, providing a more general
20188 (vect_create_epilog_for_reduction): Update accordingly.
20189 (vectorizable_reduction): Likewise.
20190 (vect_transform_cycle_phi): Likewise.
20192 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
20194 * tree-vect-loop.c (get_initial_def_for_reduction): Take the
20195 reduc_info instead of the original stmt_vec_info.
20196 (vect_transform_cycle_phi): Update accordingly.
20198 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
20200 * tree-vect-loop.c (get_initial_defs_for_reduction): Take the
20201 reduc_info as an additional parameter.
20202 (vect_transform_cycle_phi): Update accordingly.
20204 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
20206 * tree-vectorizer.h: Include tree-ssa-operands.h.
20207 (vect_phi_initial_value): New function.
20208 * tree-vect-loop.c (neutral_op_for_slp_reduction): Use it.
20209 (get_initial_defs_for_reduction, info_for_reduction): Likewise.
20210 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
20211 (vect_transform_cycle_phi, vectorizable_induction): Likewise.
20213 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
20215 * tree-vect-loop.c (vect_create_epilog_for_reduction): Convert
20216 the phi results to vectype after creating them. Remove later
20217 conversion code that thus becomes redundant.
20219 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
20221 * tree-vect-loop.c (vect_create_epilog_for_reduction): Replace
20222 the new_phis vector with a reduc_inputs vector. Combine handling
20223 of reduction chains and ncopies > 1.
20225 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
20227 * tree-vect-loop.c (vect_create_epilog_for_reduction): Truncate
20228 scalar_results to group_size elements after reducing down from
20229 N*group_size elements. Construct an array_slice of the live-out
20230 stmts and assert that there is one stmt per scalar result.
20232 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
20234 * tree-vect-loop.c (vect_create_epilog_for_reduction): Remove
20235 nested_in_vect_loop and use double_reduc everywhere. Remove dead
20236 assignment to "loop".
20238 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
20240 * internal-fn.c (vectorized_internal_fn_supported_p): Handle
20241 vector types first. For scalar types, consider both the preferred
20242 vector mode and the alternative vector modes.
20243 * optabs-query.c (can_vec_mask_load_store_p): Use the same
20244 structure as above, in particular using related_vector_mode
20245 for modes provided by autovectorize_vector_modes.
20247 2021-07-13 Jakub Jelinek <jakub@redhat.com>
20248 Richard Biener <rguenther@suse.de>
20250 PR tree-optimization/101419
20251 * tree-pass.h (PROP_objsz): Define.
20252 (make_pass_early_object_sizes): Declare.
20253 * passes.def (pass_all_early_optimizations): Rename pass_object_sizes
20254 there to pass_early_object_sizes, drop parameter.
20255 (pass_all_optimizations): Move pass_object_sizes right after pass_ccp,
20256 drop parameter, move pass_post_ipa_warn right after that.
20257 * tree-object-size.c (pass_object_sizes::execute): Rename to...
20258 (object_sizes_execute): ... this. Add insert_min_max_p argument.
20259 (pass_data_object_sizes): Move after object_sizes_execute.
20260 (pass_object_sizes): Likewise. In execute method call
20261 object_sizes_execute, drop set_pass_param method and insert_min_max_p
20262 non-static data member and its initializer in the ctor.
20263 (pass_data_early_object_sizes, pass_early_object_sizes,
20264 make_pass_early_object_sizes): New.
20265 * tree-ssa-sccvn.c (copy_reference_ops_from_ref): Use
20266 (cfun->curr_properties & PROP_objsz) instead of cfun->after_inlining.
20268 2021-07-13 Kito Cheng <kito.cheng@sifive.com>
20271 * config/riscv/constraints.md ("S"): Update description and remove
20273 * doc/md.texi (Machine Constraints): Document the 'S' constraints
20276 2021-07-13 Richard Biener <rguenther@suse.de>
20279 2021-07-12 Richard Biener <rguenther@suse.de>
20281 * tree-vect-slp.c (vect_slp_region): Show the number of
20282 SLP graph entries in the optimization message.
20284 2021-07-13 Michael Meissner <meissner@linux.ibm.com>
20286 * config/rs6000/altivec.md (xxspltiw_v4sf): Change local variable
20288 * config/rs6000/rs6000-protos.h (rs6000_const_f32_to_i32): Change
20289 return type to long.
20290 * config/rs6000/rs6000.c (rs6000_const_f32_to_i32): Change return
20293 2021-07-12 Andrew MacLeod <amacleod@redhat.com>
20295 * gimple-range-fold.cc (fold_using_range::range_of_builtin_ubsan_call):
20296 Query relation between the 2 operands and use it.
20298 2021-07-12 Sergei Trofimovich <siarheit@google.com>
20300 * doc/cfg.texi: Fix s/ei_safe_safe/ei_safe_edge/ typo.
20302 2021-07-12 Uroš Bizjak <ubizjak@gmail.com>
20305 * config/i386/predicates.md (vec_setm_sse41_operand):
20306 Rename from vec_setm_operand.
20307 (vec_setm_avx2_operand): New predicate.
20308 * config/i386/sse.md (vec_set<V_128:mode>): Use V_128 mode iterator.
20309 Use vec_setm_sse41_operand as operand 2 predicate.
20310 (vec_set<V_256_512:mode): New expander.
20311 * config/i386/mmx.md (vec_setv2hi): Use vec_setm_sse41_operand
20312 as operand 2 predicate.
20314 2021-07-12 Andrew MacLeod <amacleod@redhat.com>
20316 PR tree-optimization/101335
20317 * range-op.cc (operator_cast::lhs_op1_relation): Delete.
20319 2021-07-12 Andrew Pinski <apinski@marvell.com>
20321 * tree-ssa-phiopt.c (match_simplify_replacement): Move
20322 insert of the sequence before the movement of the
20323 statement. Check if to see if the statement is used
20324 outside of the original phi to see if we should move it.
20326 2021-07-12 Richard Biener <rguenther@suse.de>
20328 * dump-context.h (debug_dump_context::debug_dump_context):
20329 Add FILE * parameter defaulted to stderr.
20330 * dumpfile.c (debug_dump_context::debug_dump_context): Adjust.
20331 * tree-vect-slp.c (dot_slp_tree): New functions.
20333 2021-07-12 Richard Biener <rguenther@suse.de>
20335 PR tree-optimization/101373
20336 * tree-ssa-pre.c (prune_clobbered_mems): Also prune trapping
20337 references when the BB may not return.
20338 (compute_avail): Pass in the function we're working on and
20339 replace cfun references with it. Externally throwing
20340 const calls also possibly terminate the function.
20341 (pass_pre::execute): Pass down the function we're working on.
20342 * gcse.c (compute_hash_table_work): Externally throwing
20343 const/pure calls also need record_last_mem_set_info.
20344 * postreload-gcse.c (record_opr_changes): Looping or externally
20345 throwing const/pure calls also need record_last_mem_set_info.
20347 2021-07-12 Uroš Bizjak <ubizjak@gmail.com>
20349 * recog.c (memory_address_addr_space_p): Change the type to bool.
20350 Return true/false instead of 1/0.
20351 (offsettable_memref_p): Ditto.
20352 (offsettable_nonstrict_memref_p): Ditto.
20353 (offsettable_address_addr_space_p): Ditto.
20354 Change the type of addressp indirect function to bool.
20355 * recog.h (memory_address_addr_space_p): Change the type to bool.
20356 (strict_memory_address_addr_space_p): Ditto.
20357 (offsettable_memref_p): Ditto.
20358 (offsettable_nonstrict_memref_p): Ditto.
20359 (offsettable_address_addr_space_p): Ditto.
20360 * reload.c (maybe_memory_address_addr_space_p): Ditto.
20361 (strict_memory_address_addr_space_p): Change the type to bool.
20362 Return true/false instead of 1/0.
20363 (maybe_memory_address_addr_space_p): Change the type to bool.
20365 2021-07-12 Richard Biener <rguenther@suse.de>
20367 * tree-vect-slp.c (vect_slp_region): Show the number of
20368 SLP graph entries in the optimization message.
20370 2021-07-12 Richard Biener <rguenther@suse.de>
20372 PR tree-optimization/101394
20373 * tree-ssa-pre.c (do_pre_regular_insertion): Avoid inserting
20374 copies from abnormals for a full redundancy.
20376 2021-07-12 Richard Biener <rguenther@suse.de>
20378 PR middle-end/101423
20379 * gimple.c (gimple_could_trap_p_1): Internal function calls
20381 * tree-eh.c (tree_could_trap_p): Likewise.
20383 2021-07-12 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
20386 * config/arm/arm_neon.h (vmul_n_u32): Replace call to builtin with
20388 (vmulq_n_u32): Likewise.
20389 (vmul_n_f32): Gate __a * __b on __FAST_MATH__.
20390 (vmulq_n_f32): Likewise.
20391 (vmul_n_f16): Likewise.
20392 (vmulq_n_f16): Likewise.
20394 2021-07-12 Martin Liska <mliska@suse.cz>
20396 PR sanitizer/101425
20397 * gcc.c (check_offload_target_name): Call
20398 candidates_list_and_hint only if we have a candidate.
20400 2021-07-12 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
20403 * config/arm/neon.md (vec_init): Move to ...
20404 * config/arm/vec-common.md (vec_init): ... here.
20405 Change the pattern's mode to VDQX and gate it on VALID_MVE_MODE.
20407 2021-07-12 Roger Sayle <roger@nextmovesoftware.com>
20409 PR tree-optimization/101403
20410 * match.pd ((T)bswap(X)>>C): Correctly handle cases where
20411 signedness of the shift is not the same as the signedness of
20412 the type extension.
20414 2021-07-09 Roger Sayle <roger@nextmovesoftware.com>
20415 Uroš Bizjak <ubizjak@gmail.com>
20417 * config/i386/i386.md (*divmodsi4_const): Optimize SImode
20418 divmod of a constant numerator with new define_insn_and_split.
20420 2021-07-09 Iain Sandoe <iain@sandoe.co.uk>
20423 * config/i386/i386-expand.c (ix86_expand_call): If a call is
20424 to a non-local-binding, or local but to a public symbol, then
20425 assume that it might be indirected via the lazy symbol binder.
20426 Mark R10 and R10 as clobbered in that case.
20428 2021-07-09 Eric Botcazou <ebotcazou@adacore.com>
20431 * gcc.c (ASM_DEBUG_DWARF_OPTION): Set again to --gdwarf2 in
20432 the case where HAVE_AS_WORKING_DWARF_N_FLAG is not defined
20433 and HAVE_LD_BROKEN_PE_DWARF5 is defined.
20435 2021-07-09 Uroš Bizjak <ubizjak@gmail.com>
20437 * config/i386/i386.md (*udivmodsi4_pow2_zext_1): Limit the
20438 log2 range of operands[3] to [1,31].
20439 (*udivmodsi4_pow2_zext_2): Ditto. Correct insn RTX pattern.
20441 2021-07-09 Sergei Trofimovich <siarheit@google.com>
20443 * doc/md.texi: Don't split @smallexample in multiple @groups.
20445 2021-07-09 Sergei Trofimovich <siarheit@google.com>
20447 * doc/md.texi: Add missing 'see' word.
20449 2021-07-09 Andrew Pinski <apinski@marvell.com>
20451 * tree-ssa-phiopt.c (phiopt_early_allow): Change arguments
20452 to take sequence and gimple_match_op. Accept the case where
20453 op is a SSA_NAME and one statement in the sequence.
20454 Also allow constants.
20455 (gimple_simplify_phiopt): Always pass a sequence to resimplify.
20456 Update call to phiopt_early_allow. Discard the sequence if not
20459 2021-07-09 Xi Ruoyao <xry111@mengyan1223.wang>
20464 * config/mips/mips.c (mips_const_insns): Use MSA_SUPPORTED_MODE_P
20465 instead of ISA_HAS_MSA.
20466 (mips_expand_vec_unpack): Likewise.
20467 (mips_expand_vector_init): Likewise.
20469 2021-07-09 Kewen Lin <linkw@linux.ibm.com>
20471 * config/rs6000/vsx.md (mods_<mode>): Rename to...
20472 (mod<mode>3): ... this.
20473 (modu_<mode>): Rename to...
20474 (umod<mode>3): ... this.
20475 * config/rs6000/rs6000-builtin.def (MODS_V2DI, MODS_V4SI, MODU_V2DI,
20476 MODU_V4SI): Adjust.
20478 2021-07-08 Jeff Law <jeffreyalaw@gmail.com>
20480 * config/h8300/shiftrotate.md (variable shifts): Expose condition
20481 code handling for the test before the loop.
20483 2021-07-08 Martin Jambor <mjambor@suse.cz>
20486 * ipa-sra.c (class isra_call_summary): New member
20487 m_before_any_store, initialize it in the constructor.
20488 (isra_call_summary::dump): Dump the new field.
20489 (ipa_sra_call_summaries::duplicate): Copy it.
20490 (process_scan_results): Set it.
20491 (isra_write_edge_summary): Stream it.
20492 (isra_read_edge_summary): Likewise.
20493 (param_splitting_across_edge): Only override
20494 safe_to_import_accesses if m_before_any_store is set.
20496 2021-07-08 Martin Sebor <msebor@redhat.com>
20498 PR bootstrap/101374
20499 * gimple-array-bounds.cc (array_bounds_checker::check_mem_ref):
20500 Use Object Size Type 0 instead of 1.
20502 2021-07-08 Richard Sandiford <richard.sandiford@arm.com>
20504 * tree-vect-loop.c (vectorizable_reduction): Remove always-true
20507 2021-07-08 Richard Sandiford <richard.sandiford@arm.com>
20509 * match.pd: Simplify an extend-operate-truncate sequence involving
20512 2021-07-08 Roger Sayle <roger@nextmovesoftware.com>
20513 Richard Biener <rguenther@suse.de>
20515 PR tree-optimization/40210
20516 * match.pd (bswap optimizations): Simplify (bswap(x)>>C1)&C2 as
20517 (x>>C3)&C2 when possible. Simplify bswap(x)>>C1 as ((T)x)>>C2
20518 when possible. Simplify bswap(x)&C1 as (x>>C2)&C1 when 0<=C1<=255.
20520 2021-07-08 Uroš Bizjak <ubizjak@gmail.com>
20523 * config/i386/i386-expand.c (ix86_expand_sse_unpack):
20525 * config/i386/mmx.md (V_32): New mode iterator.
20526 (mov<V_32:mode>): Use V_32 mode iterator.
20527 (*mov<V_32:mode>_internal): Ditto.
20528 (*push<V_32:mode>2_rex64): Ditto.
20529 (*push<V_32:mode>2): Ditto.
20530 (movmisalign<V_32:mode>): Ditto.
20531 (mmx_<any_shiftrt:insn>v1si3): New insn pattern.
20532 (sse4_1_<any_extend:code>v2qiv2hi2): Ditto.
20533 (vec_unpacks_lo_v4qi): New expander.
20534 (vec_unpacks_hi_v4qi): Ditto.
20535 (vec_unpacku_lo_v4qi): Ditto.
20536 (vec_unpacku_hi_v4qi): Ditto.
20537 * config/i386/i386.h (VALID_SSE2_REG_MODE): Add V1SImode.
20538 (VALID_INT_MODE_P): Ditto.
20540 2021-07-08 Michael Meissner <meissner@linux.ibm.com>
20543 * config/rs6000/rs6000.md (udivti3): New insn.
20544 (divti3): New insn.
20545 (umodti3): New insn.
20546 (modti3): New insn.
20548 2021-07-07 Martin Sebor <msebor@redhat.com>
20550 PR tree-optimization/100137
20551 PR tree-optimization/99121
20552 PR tree-optimization/97027
20553 * builtins.c (access_ref::access_ref): Also set offmax.
20554 (access_ref::offset_in_range): Define new function.
20555 (access_ref::add_offset): Set offmax.
20556 (access_ref::inform_access): Handle access_none.
20557 (handle_mem_ref): Clear ostype.
20558 (compute_objsize_r): Handle ASSERT_EXPR.
20559 * builtins.h (struct access_ref): Add offmax member.
20560 * gimple-array-bounds.cc (array_bounds_checker::check_mem_ref): Use
20561 compute_objsize() and simplify.
20563 2021-07-07 Peter Bergner <bergner@linux.ibm.com>
20565 * config/rs6000/rs6000-call.c (mma_init_builtins): Use VSX_BUILTIN_LXVP
20566 and VSX_BUILTIN_STXVP.
20568 2021-07-07 Martin Sebor <msebor@redhat.com>
20571 * config/aarch64/aarch64.c (aarch64_simd_lane_bounds): Remove
20572 a stray %K from error_at() missed in r12-2088.
20574 2021-07-07 Richard Biener <rguenther@suse.de>
20576 PR tree-optimization/99728
20577 * tree-ssa-loop-im.c (gather_mem_refs_stmt): Record
20579 (mem_refs_may_alias_p): Add assert we handled aggregate
20581 (sm_seq_valid_bb): Give up when running into aggregate copies.
20582 (ref_indep_loop_p): Handle aggregate copies as never
20583 being invariant themselves but allow other refs to be
20584 disambiguated against them.
20585 (can_sm_ref_p): Do not try to apply store-motion to aggregate
20588 2021-07-06 Indu Bhagat <indu.bhagat@oracle.com>
20591 * dwarf2ctf.c (ctf_get_AT_data_member_location): Multiply by 8 to get
20594 2021-07-06 Martin Sebor <msebor@redhat.com>
20596 * gimple-pretty-print.c (percent_G_format): Remove.
20597 * tree-diagnostic.c (default_tree_printer): Remove calls.
20598 * tree-pretty-print.c (percent_K_format): Remove.
20599 * tree-pretty-print.h (percent_K_format): Remove.
20601 2021-07-06 Martin Sebor <msebor@redhat.com>
20603 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_builtin):
20604 Remove %K and use error_at.
20605 (aarch64_expand_fcmla_builtin): Same.
20606 (aarch64_expand_builtin_tme): Same.
20607 (aarch64_expand_builtin_memtag): Same.
20608 * config/arm/arm-builtins.c (arm_expand_acle_builtin): Same.
20609 (arm_expand_builtin): Same.
20610 * config/arm/arm.c (bounds_check): Same.
20612 2021-07-06 Martin Sebor <msebor@redhat.com>
20614 * builtins.c (warn_string_no_nul): Remove %G.
20615 (maybe_warn_for_bound): Same.
20616 (warn_for_access): Same.
20617 (check_access): Same.
20618 (check_strncat_sizes): Same.
20619 (expand_builtin_strncat): Same.
20620 (expand_builtin_strncmp): Same.
20621 (expand_builtin): Same.
20622 (expand_builtin_object_size): Same.
20623 (warn_dealloc_offset): Same.
20624 (maybe_emit_free_warning): Same.
20625 * calls.c (maybe_warn_alloc_args_overflow): Same.
20626 (maybe_warn_nonstring_arg): Same.
20627 (maybe_warn_rdwr_sizes): Same.
20628 * expr.c (expand_expr_real_1): Remove %K.
20629 * gimple-fold.c (gimple_fold_builtin_strncpy): Remove %G.
20630 (gimple_fold_builtin_strncat): Same.
20631 * gimple-ssa-sprintf.c (format_directive): Same.
20632 (handle_printf_call): Same.
20633 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Same.
20634 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Same.
20635 (maybe_diag_access_bounds): Same. Call gimple_location.
20636 (check_bounds_or_overlap): Same.
20637 * trans-mem.c (ipa_tm_scan_irr_block): Remove %K. Simplify.
20638 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Remove %G.
20639 * tree-ssa-strlen.c (maybe_warn_overflow): Same.
20640 (maybe_diag_stxncpy_trunc): Same.
20641 (handle_builtin_stxncpy_strncat): Same.
20642 (maybe_warn_pointless_strcmp): Same.
20643 * tree-ssa-uninit.c (maybe_warn_operand): Same.
20645 2021-07-06 Uroš Bizjak <ubizjak@gmail.com>
20648 * config/i386/predicates.md (vec_setm_operand): Enable
20649 register_operand for TARGET_SSE4_1.
20650 * config/i386/mmx.md (vec_setv2hi): Use vec_setm_operand
20651 as operand 2 predicate. Call ix86_expand_vector_set_var
20652 for non-constant index operand.
20653 (vec_setv4qi): Use vec_setm_mmx_operand as operand 2 predicate.
20654 Call ix86_expand_vector_set_var for non-constant index operand.
20656 2021-07-06 Jeff Law <jeffreyalaw@gmail.com>
20658 * config/h8300/jumpcall.md (*branch): When possible, generate
20659 the comparison in CCZN mode.
20660 * config/h8300/predicates.md (simple_memory_operand): Reject all
20661 auto-increment addressing modes.
20663 2021-07-06 Iain Sandoe <iain@sandoe.co.uk>
20665 PR bootstrap/100246
20666 * config/i386/i386.h (struct stringop_algs): Define a CTOR for
20669 2021-07-06 Richard Biener <rguenther@suse.de>
20671 * doc/md.texi (vec_fmaddsub<mode>4): Document.
20672 (vec_fmsubadd<mode>4): Likewise.
20673 * optabs.def (vec_fmaddsub$a4): Add.
20674 (vec_fmsubadd$a4): Likewise.
20675 * internal-fn.def (IFN_VEC_FMADDSUB): Add.
20676 (IFN_VEC_FMSUBADD): Likewise.
20677 * tree-vect-slp-patterns.c (addsub_pattern::recognize):
20678 Refactor to handle IFN_VEC_FMADDSUB and IFN_VEC_FMSUBADD.
20679 (addsub_pattern::build): Likewise.
20680 * tree-vect-slp.c (vect_optimize_slp): CFN_VEC_FMADDSUB
20681 and CFN_VEC_FMSUBADD are not transparent for permutes.
20682 * config/i386/sse.md (vec_fmaddsub<mode>4): New expander.
20683 (vec_fmsubadd<mode>4): Likewise.
20685 2021-07-06 Richard Biener <rguenther@suse.de>
20687 * doc/invoke.texi (fmove-loop-stores): Document.
20688 * common.opt (fmove-loop-stores): New option.
20689 * opts.c (default_options_table): Enable -fmove-loop-stores
20690 at -O1 but not -Og.
20691 * tree-ssa-loop-im.c (pass_lim::execute): Pass
20692 flag_move_loop_stores instead of true to
20693 loop_invariant_motion_in_fun.
20695 2021-07-06 Iain Sandoe <iain@sandoe.co.uk>
20697 * doc/install.texi: Document --with-dsymutil.
20699 2021-07-06 Andrew Pinski <apinski@marvell.com>
20701 PR tree-optimization/101256
20702 * dbgcnt.def (phiopt_edge_range): New counter.
20703 * tree-ssa-phiopt.c (replace_phi_edge_with_variable):
20704 Check to make sure the new name is defined in the same
20705 bb as the conditional before duplicating range info.
20706 Also add debug counter.
20708 2021-07-06 Kewen Lin <linkw@linux.ibm.com>
20710 PR rtl-optimization/100328
20711 * config/i386/i386-options.c (ix86_option_override_internal):
20712 Set param_ira_consider_dup_in_all_alts to 0.
20714 2021-07-06 Kewen Lin <linkw@linux.ibm.com>
20716 PR rtl-optimization/100328
20717 * doc/invoke.texi (ira-consider-dup-in-all-alts): Document new
20719 * ira.c (ira_get_dup_out_num): Adjust as parameter
20720 param_ira_consider_dup_in_all_alts.
20721 * params.opt (ira-consider-dup-in-all-alts): New.
20722 * ira-conflicts.c (process_regs_for_copy): Add one parameter
20723 single_input_op_has_cstr_p.
20724 (get_freq_for_shuffle_copy): New function.
20725 (add_insn_allocno_copies): Adjust as single_input_op_has_cstr_p.
20726 * ira-int.h (ira_get_dup_out_num): Add one bool parameter.
20728 2021-07-05 Jeff Law <jeffreyalaw@gmail.com>
20730 * config/h8300/shiftrotate.md (shift-by-variable patterns): Update to
20731 generate condition code aware RTL directly.
20733 2021-07-05 Andrew Pinski <apinski@marvell.com>
20735 PR tree-optimization/101039
20736 * match.pd (A CMP 0 ? A : -A): New patterns.
20737 * tree-ssa-phiopt.c (abs_replacement): Delete function.
20738 (tree_ssa_phiopt_worker): Don't call abs_replacement.
20739 Update comment about abs_replacement.
20741 2021-07-05 Andrew Pinski <apinski@marvell.com>
20743 * tree-ssa-phiopt.c (gimple_simplify_phiopt):
20744 If "A ? B : C" fails to simplify, try "(!A) ? C : B".
20746 2021-07-05 Andrew Pinski <apinski@marvell.com>
20748 * tree-ssa-phiopt.c (match_simplify_replacement):
20749 Add early_p argument. Call gimple_simplify_phiopt
20750 instead of gimple_simplify.
20751 (tree_ssa_phiopt_worker): Update call to
20752 match_simplify_replacement and allow unconditionally.
20753 (phiopt_early_allow): New function.
20754 (gimple_simplify_phiopt): New function.
20756 2021-07-05 Andrew Pinski <apinski@marvell.com>
20758 PR middle-end/101237
20759 * fold-const.c (negate_expr_p): Remove call to element_mode
20760 and TREE_MODE/TREE_TYPE when calling HONOR_SIGNED_ZEROS,
20761 HONOR_SIGN_DEPENDENT_ROUNDING, and HONOR_SNANS.
20762 (fold_negate_expr_1): Likewise.
20763 (const_unop): Likewise.
20764 (fold_cond_expr_with_comparison): Likewise.
20765 (fold_binary_loc): Likewise.
20766 (fold_ternary_loc): Likewise.
20767 (tree_call_nonnegative_warnv_p): Likewise.
20768 * match.pd (-(A + B) -> (-B) - A): Likewise.
20770 2021-07-05 Iain Sandoe <iain@sandoe.co.uk>
20772 * configure.ac: Handle --with-dsymutil in the same way as we
20773 do for the assembler and linker. (DEFAULT_DSYMUTIL): New.
20774 Extract the type and version for the dsymutil configured or
20775 found by the default searches.
20776 * config.in: Regenerated.
20777 * configure: Regenerated.
20778 * collect2.c (do_dsymutil): Handle locating dsymutil in the
20779 same way as for the assembler and linker.
20780 * config/darwin.h (DSYMUTIL): Delete.
20781 * gcc.c: Report a configured dsymutil correctly.
20782 * exec-tool.in: Allow for dsymutil.
20784 2021-07-05 Uroš Bizjak <ubizjak@gmail.com>
20786 * config/i386/i386-expand.c (ix86_split_mmx_punpck):
20787 Handle V4QI and V2HI modes.
20788 (expand_vec_perm_blend): Allow 4-byte vector modes with TARGET_SSE4_1.
20789 Handle V4QI mode. Emit mmx_pblendvb32 for 4-byte modes.
20790 (expand_vec_perm_pshufb): Rewrite to use switch statemets.
20791 Handle 4-byte dual operands with TARGET_XOP and single operands
20792 with TARGET_SSSE3. Emit mmx_ppermv32 for TARGET_XOP and
20793 mmx_pshufbv4qi3 for TARGET_SSSE3.
20794 (expand_vec_perm_pblendv): Allow 4-byte vector modes with TARGET_SSE4_1.
20795 (expand_vec_perm_interleave2): Allow 4-byte vector modes.
20796 (expand_vec_perm_pshufb2): Allow 4-byte vector modes with TARGET_SSSE3.
20797 (expand_vec_perm_even_odd_1): Handle V4QI mode.
20798 (expand_vec_perm_broadcast_1): Handle V4QI mode.
20799 (ix86_vectorize_vec_perm_const): Handle V4QI mode.
20800 * config/i386/mmx.md (mmx_ppermv32): New insn pattern.
20801 (mmx_pshufbv4qi3): Ditto.
20802 (*mmx_pblendw32): Ditto.
20803 (*mmx_pblendw64): Rename from *mmx_pblendw.
20804 (mmx_punpckhbw_low): New insn_and_split pattern.
20805 (mmx_punpcklbw_low): Ditto.
20807 2021-07-05 Richard Biener <rguenther@suse.de>
20809 * tree-vect-loop-manip.c (vect_loop_versioning): Do not
20810 set LOOP_C_INFINITE on the vectorized loop.
20812 2021-07-05 Richard Biener <rguenther@suse.de>
20814 PR middle-end/101291
20815 * cfgloopmanip.c (loop_version): Set the loop copy of the
20816 versioned loop to the new loop.
20818 2021-07-04 Iain Sandoe <iain@sandoe.co.uk>
20821 * config.gcc: Ensure that Darwin biarch definitions are
20822 added before i386.h.
20823 * config/i386/darwin.h (TARGET_64BIT): Remove.
20824 (PR80556_WORKAROUND): New.
20825 (REAL_LIBGCC_SPEC): Amend to use PR80556_WORKAROUND.
20826 (DARWIN_SUBARCH_SPEC): New.
20827 * config/i386/darwin32-biarch.h (TARGET_64BIT_DEFAULT,
20828 TARGET_BI_ARCH, PR80556_WORKAROUND): New.
20829 (REAL_LIBGCC_SPEC): Remove.
20830 * config/i386/darwin64-biarch.h (TARGET_64BIT_DEFAULT,
20831 TARGET_BI_ARCH, PR80556_WORKAROUND): New.
20832 (REAL_LIBGCC_SPEC): Remove.
20834 2021-07-03 H.J. Lu <hjl.tools@gmail.com>
20836 PR middle-end/101294
20837 * expr.c (store_constructor): Don't use vec_duplicate on vector.
20839 2021-07-02 Martin Sebor <msebor@redhat.com>
20841 PR middle-end/98871
20842 PR middle-end/98512
20843 * diagnostic.c (get_any_inlining_info): New.
20844 (update_effective_level_from_pragmas): Handle inlining context.
20845 (diagnostic_enabled): Same.
20846 (diagnostic_report_diagnostic): Same.
20847 * diagnostic.h (struct diagnostic_info): Add ctor.
20848 (struct diagnostic_context): Add new member.
20849 * tree-diagnostic.c (set_inlining_locations): New.
20850 (tree_diagnostics_defaults): Set new callback pointer.
20852 2021-07-02 Peter Bergner <bergner@linux.ibm.com>
20854 * config/rs6000/rs6000-builtin.def (BU_MMA_PAIR_LD, BU_MMA_PAIR_ST):
20856 (__builtin_vsx_lxvp, __builtin_vsx_stxvp): New built-ins.
20857 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_mma_builtin): Expand
20858 lxvp and stxvp built-ins.
20859 (mma_init_builtins): Handle lxvp and stxvp built-ins.
20860 (builtin_function_type): Likewise.
20861 * doc/extend.texi (__builtin_vsx_lxvp, __builtin_mma_stxvp): Document.
20863 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
20865 * config/h8300/h8300-protos.h (compute_a_shift_cc): Accept
20866 additional argument for the code.
20867 * config/h8300/h8300.c (compute_a_shift_cc): Accept additional
20868 argument for the code. Just return if the ZN bits are useful or
20869 not rather than the old style CC_* enums.
20870 * config/h8300/shiftrotate.md (shiftqi_noscratch): Move before
20871 more generic shiftqi patterns.
20872 (shifthi_noscratch, shiftsi_noscratch): Similarly.
20873 (shiftqi_noscratch_set_flags): New pattern.
20874 (shifthi_noscratch_set_flags, shiftsi_noscratch_set_flags): Likewise.
20876 2021-07-02 Andrew MacLeod <amacleod@redhat.com>
20878 PR tree-optimization/101223
20879 * range-op.cc (build_lt): Add -1 for signed values.
20880 (built_gt): Subtract -1 for signed values.
20882 2021-07-02 David Faust <david.faust@oracle.com>
20884 * btfout.c (get_btf_kind): Support BTF_KIND_FLOAT.
20885 (btf_asm_type): Likewise.
20887 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
20889 * config/h8300/h8300-protos.h (output_a_shift): Make first argument
20890 an array of rtx rather than a pointer to rtx. Add code argument.
20891 (compute_a_shift_length): Similarly.
20892 * config/h8300/h8300.c (h8300_shift_costs): Adjust now that the
20893 shift itself isn't an operand. Create dummy operand[0] to carry
20894 a mode and pass a suitable rtx code to compute_a_shift_length.
20895 (get_shift_alg): Adjust operand number of clobber in output templates.
20896 (output_a_shift): Make first argument an array of rtx rather than
20897 a pointer to rtx. Add code argument for the type of shift.
20898 Adjust now that the shift itself is no longer an operand.
20899 (compute_a_shift_length): Similarly.
20900 * config/h8300/shiftrotate.md (shiftqi, shifthi, shiftsi): Use an
20901 iterator rather than nshift_operator.
20902 (shiftqi_noscratch, shifthi_noscratch, shiftsi_noscratch): Likewise.
20903 (shiftqi_clobber_flags): Adjust to API changes in output_a_shift
20904 and compute_a_shift_length.
20905 (shiftqi_noscratch_clobber_flags): Likewise.
20906 (shifthi_noscratch_clobber_flags): Likewise.
20907 (shiftsi_noscratch_clobber_flags): Likewise.
20909 2021-07-02 Iain Sandoe <iain@sandoe.co.uk>
20912 * config/darwin.h (DSYMUTIL_SPEC): Do not try to run
20913 dsymutil for BTF/CTF.
20915 2021-07-02 Iain Sandoe <iain@sandoe.co.uk>
20918 * config/darwin.h (CTF_INFO_SECTION_NAME): Update the
20919 segment to include BTF.
20920 (BTF_INFO_SECTION_NAME): New.
20922 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
20924 * config/m32r/m32r-protos.h (call_operand): Adjust return type.
20925 (small_data_operand, memreg_operand, small_insn_p): Likewise.
20926 * config/m32r/m32r.c (call_operand): Adjust return type.
20927 (small_data_operand, memreg_operand): Likewise.
20929 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
20931 * config/frv/frv-protos.h (integer_register_operand): Adjust return
20933 (frv_load_operand, gpr_or_fpr_operand, gpr_no_subreg_operand): Likewise.
20934 (fpr_or_int6_operand, gpr_or_int_operand); Likewise.
20935 (gpr_or_int12_operand, gpr_or_int10_operand); Likewise.
20936 (move_source_operand, move_destination_operand): Likewise.
20937 (condexec_source_operand, condexec_dest_operand): Likewise.
20938 (lr_operand, gpr_or_memory_operand, fpr_or_memory_operand): Likewise.
20939 (reg_or_0_operand, fcc_operand, icc_operand, cc_operand): Likewise.
20940 (fcr_operand, icr_operand, cr_operand, call_operand): Likewise.
20941 (fpr_operand, even_reg_operand, odd_reg_operand): Likewise.
20942 (even_gpr_operand, odd_gpr_operand, quad_fpr_operand): Likewise.
20943 (even_fpr_operand, odd_fpr_operand): Likewise.
20944 (dbl_memory_one_insn_operand, dbl_memory_two_insn_operand): Likewise.
20945 (int12_operand, int6_operand, int5_operand, uint5_operand): Likewise.
20946 (uint4_operand, uint1_operand, int_2word_operand): Likewise
20947 (upper_int16_operand, uint16_operand, symbolic_operand): Likewise.
20948 (relational_operator, float_relational_operator): Likewise.
20949 (ccr_eqne_operator, minmax_operator): Likewise.
20950 (condexec_si_binary_operator, condexec_si_media_operator): Likewise.
20951 (condexec_si_divide_operator, condexec_si_unary_operator): Likewise.
20952 (condexec_sf_conv_operator, condexec_sf_add_operator): Likewise.
20953 (intop_compare_operator, acc_operand, even_acc_operand): Likewise.
20954 (quad_acc_operand, accg_operand): Likewise.
20956 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
20958 * config/stormy16/stormy16-protos.h (xstormy16_below_100_symbol): Change
20959 return type to a bool.
20960 (nonimmediate_nonstack_operand): Likewise.
20961 (xstormy16_splittable_below100_operand): Likewise.
20962 * config/stormy16/stormy16.c (xstormy16_below_100_symbol): Fix
20964 (xstormy16_splittable_below100_operand): Likewise.
20966 2021-07-02 Richard Biener <rguenther@suse.de>
20968 PR tree-optimization/101293
20969 * tree-ssa-loop-im.c (mem_ref_hasher::equal): Compare MEM_REF bases
20970 with combined offsets.
20971 (gather_mem_refs_stmt): Hash MEM_REFs as if their offset were
20972 combined with the rest of the offset.
20974 2021-07-02 Eric Botcazou <ebotcazou@adacore.com>
20976 * config/i386/i386.c (asm_preferred_eh_data_format): Always use the
20977 PIC encodings for PE-COFF targets.
20979 2021-07-02 Jakub Jelinek <jakub@redhat.com>
20982 * config/i386/i386-expand.c (ix86_broadcast_from_integer_constant):
20983 Return nullptr for TImode inner mode.
20985 2021-07-02 Richard Biener <rguenther@suse.de>
20987 PR tree-optimization/101280
20988 PR tree-optimization/101173
20989 * gimple-loop-interchange.cc
20990 (tree_loop_interchange::valid_data_dependences): Properly
20991 guard all dependence checks with DDR_REVERSED_P or its
20994 2021-07-02 Hongyu Wang <hongyu.wang@intel.com>
20996 * config/i386/i386-expand.c (ix86_expand_builtin):
20997 Add branch to clear odata when ZF is set for asedecenc_expand
20998 and wideaesdecenc_expand.
21000 2021-07-02 Eugene Rozenfeld <erozen@microsoft.com>
21002 * config/i386/gcc-auto-profile: regenerate
21004 2021-07-02 liuhongt <hongtao.liu@intel.com>
21006 * config/i386/sse.md (trunc<mode><pmov_dst_4>2): Refined to ..
21007 (trunc<mode><pmov_dst_4_lower>2): this.
21009 2021-07-01 David Malcolm <dmalcolm@redhat.com>
21011 * diagnostic.h (diagnostic_context::m_file_cache): New field.
21012 * input.c (class fcache): Rename to...
21013 (class file_cache_slot): ...this, making most members private and
21014 prefixing fields with "m_".
21015 (file_cache_slot::get_file_path): New accessor.
21016 (file_cache_slot::get_use_count): New accessor.
21017 (file_cache_slot::missing_trailing_newline_p): New accessor.
21018 (file_cache_slot::inc_use_count): New.
21019 (fcache_buffer_size): Move to...
21020 (file_cache_slot::buffer_size): ...here.
21021 (fcache_line_record_size): Move to...
21022 (file_cache_slot::line_record_size): ...here.
21023 (fcache_tab): Delete, in favor of global_dc->m_file_cache.
21024 (fcache_tab_size): Move to file_cache::num_file_slots.
21025 (diagnostic_file_cache_init): Update for move of fcache_tab
21026 to global_dc->m_file_cache.
21027 (diagnostic_file_cache_fini): Likewise.
21028 (lookup_file_in_cache_tab): Convert to...
21029 (file_cache::lookup_file): ...this.
21030 (diagnostics_file_cache_forcibly_evict_file): Update for move of
21031 fcache_tab to global_dc->m_file_cache, moving most of
21032 implementation to...
21033 (file_cache::forcibly_evict_file): ...this new function and...
21034 (file_cache_slot::evict): ...this new function.
21035 (evicted_cache_tab_entry): Convert to...
21036 (file_cache::evicted_cache_tab_entry): ...this.
21037 (add_file_to_cache_tab): Convert to...
21038 (file_cache::add_file): ...this, moving bulk of implementation
21040 (file_cache_slot::create): ..this new function.
21041 (file_cache::file_cache): New.
21042 (file_cache::~file_cache): New.
21043 (lookup_or_add_file_to_cache_tab): Convert to...
21044 (file_cache::lookup_or_add_file): ..this new function.
21045 (fcache::fcache): Rename to...
21046 (file_cache_slot::file_cache_slot): ...this, adding "m_" prefixes
21048 (fcache::~fcache): Rename to...
21049 (file_cache_slot::~file_cache_slot): ...this, adding "m_" prefixes
21051 (needs_read): Convert to...
21052 (file_cache_slot::needs_read_p): ...this.
21053 (needs_grow): Convert to...
21054 (file_cache_slot::needs_grow_p): ...this.
21055 (maybe_grow): Convert to...
21056 (file_cache_slot::maybe_grow): ...this.
21057 (read_data): Convert to...
21058 (file_cache_slot::read_data): ...this.
21059 (maybe_read_data): Convert to...
21060 (file_cache_slot::maybe_read_data): ...this.
21061 (get_next_line): Convert to...
21062 (file_cache_slot::get_next_line): ...this.
21063 (goto_next_line): Convert to...
21064 (file_cache_slot::goto_next_line): ...this.
21065 (read_line_num): Convert to...
21066 (file_cache_slot::read_line_num): ...this.
21067 (location_get_source_line): Update for moving of globals to
21068 global_dc->m_file_cache.
21069 (location_missing_trailing_newline): Likewise.
21070 * input.h (class file_cache_slot): New forward decl.
21071 (class file_cache): New.
21073 2021-07-01 Michael Meissner <meissner@linux.ibm.com>
21075 * config/rs6000/rs6000.c (rs6000_maybe_emit_fp_cmove): Add IEEE
21076 128-bit floating point conditional move support.
21077 (have_compare_and_set_mask): Add IEEE 128-bit floating point
21079 * config/rs6000/rs6000.md (mov<mode>cc, IEEE128 iterator): New insn.
21080 (mov<mode>cc_p10, IEEE128 iterator): New insn.
21081 (mov<mode>cc_invert_p10, IEEE128 iterator): New insn.
21082 (fpmask<mode>, IEEE128 iterator): New insn.
21083 (xxsel<mode>, IEEE128 iterator): New insn.
21085 2021-07-01 Iain Sandoe <iain@sandoe.co.uk>
21088 * config/darwin.h (CTF_INFO_SECTION_NAME): New.
21090 2021-07-01 H.J. Lu <hjl.tools@gmail.com>
21092 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
21094 * config/i386/i386-protos.h (ix86_expand_vector_init_duplicate):
21096 * config/i386/sse.md (INT_BROADCAST_MODE): New mode iterator.
21097 (vec_duplicate<mode>): New expander.
21099 2021-07-01 H.J. Lu <hjl.tools@gmail.com>
21102 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
21104 (ix86_byte_broadcast): New function.
21105 (ix86_convert_const_wide_int_to_broadcast): Likewise.
21106 (ix86_expand_move): Convert CONST_WIDE_INT to broadcast if mode
21107 size is 16 bytes or bigger.
21108 (ix86_broadcast_from_integer_constant): New function.
21109 (ix86_expand_vector_move): Convert CONST_WIDE_INT and CONST_VECTOR
21110 to broadcast if mode size is 16 bytes or bigger.
21111 * config/i386/i386-protos.h (ix86_gen_scratch_sse_rtx): New
21113 * config/i386/i386.c (ix86_gen_scratch_sse_rtx): New function.
21115 2021-07-01 Uroš Bizjak <ubizjak@gmail.com>
21117 * config/i386/predicates.md (ix86_endbr_immediate_operand):
21118 Return true/false instead of 1/0.
21119 (movq_parallel): Ditto.
21121 2021-07-01 Uroš Bizjak <ubizjak@gmail.com>
21123 * recog.c (general_operand): Return true/false instead of 1/0.
21124 (register_operand): Ditto.
21125 (immediate_operand): Ditto.
21126 (const_int_operand): Ditto.
21127 (const_scalar_int_operand): Ditto.
21128 (const_double_operand): Ditto.
21129 (push_operand): Ditto.
21130 (pop_operand): Ditto.
21131 (memory_operand): Ditto.
21132 (indirect_operand): Ditto.
21134 2021-07-01 Uroš Bizjak <ubizjak@gmail.com>
21136 * genpreds.c (write_predicate_subfunction):
21137 Change the type of written subfunction to bool.
21138 (write_one_predicate_function):
21139 Change the type of written function to bool.
21140 (write_tm_preds_h): Ditto.
21141 * recog.h (*insn_operand_predicate_fn): Change the type to bool.
21142 * recog.c (general_operand): Change the type to bool.
21143 (address_operand): Ditto.
21144 (register_operand): Ditto.
21145 (pmode_register_operand): Ditto.
21146 (scratch_operand): Ditto.
21147 (immediate_operand): Ditto.
21148 (const_int_operand): Ditto.
21149 (const_scalar_int_operand): Ditto.
21150 (const_double_operand): Ditto.
21151 (nonimmediate_operand): Ditto.
21152 (nonmemory_operand): Ditto.
21153 (push_operand): Ditto.
21154 (pop_operand): Ditto.
21155 (memory_operand): Ditto.
21156 (indirect_operand): Ditto.
21157 (ordered_comparison_operator): Ditto.
21158 (comparison_operator): Ditto.
21159 * config/i386/i386-expand.c (ix86_expand_sse_cmp):
21160 Change the type of indirect predicate function to bool.
21161 * config/rs6000/rs6000.c (easy_vector_constant):
21162 Change the type to bool.
21163 * config/mips/mips-protos.h (m16_based_address_p):
21164 Change the type of operand 3 to bool.
21166 2021-07-01 Richard Biener <rguenther@suse.de>
21168 PR tree-optimization/101280
21169 PR tree-optimization/101173
21170 * gimple-loop-interchange.cc
21171 (tree_loop_interchange::valid_data_dependences): Revert
21172 previous change and instead correctly handle DDR_REVERSED_P
21175 2021-07-01 Richard Biener <rguenther@suse.de>
21177 PR tree-optimization/101278
21178 * tree-ssa-dse.c (dse_classify_store): First check for
21179 uses, then ignore stmt for chaining purposes.
21181 2021-07-01 Richard Biener <rguenther@suse.de>
21183 PR tree-optimization/100778
21184 * tree-vect-slp.c (vect_schedule_slp_node): Do not place trapping
21185 vectorized ops ahead of their scalar BB.
21187 2021-07-01 Uroš Bizjak <ubizjak@gmail.com>
21190 * config/i386/i386.md (*nabs<dwi>2_doubleword):
21191 New insn_and_split pattern.
21192 (*nabs<dwi>2_1): Ditto.
21193 * config/i386/i386-features.c
21194 (general_scalar_chain::compute_convert_gain):
21195 Handle (NEG (ABS (...))) RTX. Rewrite src code
21196 scanner as switch statement.
21197 (general_scalar_chain::convert_insn):
21198 Handle (NEG (ABS (...))) RTX.
21199 (general_scalar_to_vector_candidate_p):
21200 Detect (NEG (ABS (...))) RTX. Reorder case statements
21201 for (AND (NOT (...) ...)) fallthrough.
21203 2021-07-01 Richard Biener <rguenther@suse.de>
21205 PR tree-optimization/101178
21206 * tree-vect-slp.c (slpg_vertex::materialize): Remove.
21207 (slpg::perm_in): Add.
21208 (slpg::get_perm_in): Remove.
21209 (slpg::get_perm_materialized): Add.
21210 (vect_optimize_slp): Handle VEC_PERM nodes more optimally
21211 during permute propagation and materialization.
21213 2021-07-01 Jakub Jelinek <jakub@redhat.com>
21216 * dwarf2out.c (loc_list_from_tree_1): Handle COMPOUND_LITERAL_EXPR.
21218 2021-07-01 Jakub Jelinek <jakub@redhat.com>
21220 PR middle-end/94366
21221 * omp-low.c (lower_rec_input_clauses): Rename is_fp_and_or to
21222 is_truth_op, set it for TRUTH_*IF_EXPR regardless of new_var's type,
21223 use boolean_type_node instead of integer_type_node as NE_EXPR type.
21224 (lower_reduction_clauses): Likewise.
21226 2021-06-30 Hafiz Abid Qadeer <abidh@codesourcery.com>
21228 * config/gcn/gcn.c: Include dwarf2.h.
21229 (gcn_addr_space_debug): New function.
21230 (TARGET_ADDR_SPACE_DEBUG): New hook.
21232 2021-06-30 Hafiz Abid Qadeer <abidh@codesourcery.com>
21234 * common/config/gcn/gcn-common.c
21235 (gcn_option_optimization_table): Change OPT_fomit_frame_pointer to -O3.
21236 * config/gcn/gcn.c (gcn_expand_prologue): Prefer the frame pointer
21238 (gcn_expand_prologue): Prefer the frame pointer when emitting CFI.
21239 (gcn_frame_pointer_rqd): New function.
21240 (TARGET_FRAME_POINTER_REQUIRED): New hook.
21242 2021-06-30 Hafiz Abid Qadeer <abidh@codesourcery.com>
21244 * config/gcn/gcn.c (move_callee_saved_registers): Emit CFI notes for
21245 prologue register saves.
21246 (gcn_debug_unwind_info): Use UI_DWARF2.
21247 (gcn_dwarf_register_number): Map DWARF_LINK_REGISTER to DWARF PC.
21248 (gcn_dwarf_register_span): DWARF_LINK_REGISTER doesn't span.
21249 * config/gcn/gcn.h: (DWARF_FRAME_RETURN_COLUMN): New define.
21250 (DWARF_LINK_REGISTER): New define.
21251 (FIRST_PSEUDO_REGISTER): Increment.
21252 (FIXED_REGISTERS): Add entry for DWARF_LINK_REGISTER.
21253 (CALL_USED_REGISTERS): Likewise.
21254 (REGISTER_NAMES): Likewise.
21256 2021-06-30 Richard Biener <rguenther@suse.de>
21258 PR tree-optimization/101267
21259 * tree-vect-stmts.c (vect_check_scalar_mask): Adjust
21260 API and use SLP compatible interface of vect_is_simple_use.
21261 Reject not vectorized SLP defs for callers that do not support
21263 (vect_check_store_rhs): Handle masked stores and pass down
21264 the appropriate operator index.
21265 (vectorizable_call): Adjust.
21266 (vectorizable_store): Likewise.
21267 (vectorizable_load): Likewise. Handle SLP pecularity of
21269 (vect_is_simple_use): Remove special-casing of masked stores.
21271 2021-06-30 Tobias Burnus <tobias@codesourcery.com>
21273 * common.opt (foffload): Remove help as Driver only.
21274 * gcc.c (display_help): Add -foffload.
21276 2021-06-30 Tobias Burnus <tobias@codesourcery.com>
21278 * gcc.c (close_at_file, execute): Replace alloca by XALLOCAVEC.
21279 (check_offload_target_name): Fix splitting OFFLOAD_TARGETS into
21280 a candidate list; better inform no offload target is configured
21281 and fix hint extraction when passed target is not '\0' at [len].
21282 * common.opt (foffload): Add tailing '.'.
21283 (foffload-options): Likewise; fix flag name in the help string.
21285 2021-06-30 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
21288 * config/arm/arm_neon.h: Move vabs intrinsics before vcage_f32.
21289 (vcage_f32): Gate comparison on __FAST_MATH__.
21290 (vcageq_f32): Likewise.
21291 (vcale_f32): Likewise.
21292 (vcaleq_f32): Likewise.
21293 (vcagt_f32): Likewise.
21294 (vcagtq_f32): Likewise.
21295 (vcalt_f32): Likewise.
21296 (vcaltq_f32): Likewise.
21297 (vcage_f16): Likewise.
21298 (vcageq_f16): Likewise.
21299 (vcale_f16): Likewise.
21300 (vcaleq_f16): Likewise.
21301 (vcagt_f16): Likewise.
21302 (vcagtq_f16): Likewise.
21303 (vcalt_f16): Likewise.
21304 (vcaltq_f16): Likewise.
21306 2021-06-30 Richard Biener <rguenther@suse.de>
21308 PR tree-optimization/101264
21309 * tree-vect-slp.c (vect_optimize_slp): Propagate the
21310 computed perm_in to all "any" permute successors
21311 we cannot de-duplicate immediately.
21313 2021-06-30 liuhongt <hongtao.liu@intel.com>
21316 * config/i386/sse.md
21317 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>):
21319 (avx512f_sfixupimm<mode><maskz_scalar_name><round_saeonly_name>):
21321 (avx512f_sfixupimm<mode>_mask<round_saeonly_name>"): Refined.
21322 * config/i386/subst.md (maskz_scalar): New define_subst.
21323 (maskz_scalar_name): New subst_attr.
21324 (maskz_scalar_op5): Ditto.
21325 (round_saeonly_maskz_scalar_op5): Ditto.
21326 (round_saeonly_maskz_scalar_operand5): Ditto.
21328 2021-06-30 David Edelsohn <dje.gcc@gmail.com>
21330 * config/rs6000/rs6000.c (rs6000_xcoff_section_type_flags):
21331 Increase code CSECT alignment to at least 32 bytes.
21332 * config/rs6000/xcoff.h (TEXT_SECTION_ASM_OP): Add 32 byte
21333 alignment designation.
21335 2021-06-29 Sergei Trofimovich <siarheit@google.com>
21337 * doc/generic.texi: Fix s/net yet/not yet/ typo.
21339 2021-06-29 Andrew MacLeod <amacleod@redhat.com>
21341 PR tree-optimization/101254
21342 * range-op.cc (operator_minus::op1_op2_relation_effect): Check for
21343 wrapping/non-wrapping when setting the result range.
21345 2021-06-29 Andrew MacLeod <amacleod@redhat.com>
21347 * value-query.cc (gimple_range_global): Allow phis.
21349 2021-06-29 Andrew MacLeod <amacleod@redhat.com>
21351 * vr-values.c (vr_values::vrp_stmt_computes_nonzero): Use stmt.
21352 (simplify_using_ranges::op_with_boolean_value_range_p): Add a
21353 statement for location context.
21354 (check_for_binary_op_overflow): Ditto.
21355 (simplify_using_ranges::get_vr_for_comparison): Ditto.
21356 (simplify_using_ranges::compare_name_with_value): Ditto.
21357 (simplify_using_ranges::compare_names): Ditto.
21358 (vrp_evaluate_conditional_warnv_with_ops_using_ranges): Ditto.
21359 (simplify_using_ranges::simplify_truth_ops_using_ranges): Ditto.
21360 (simplify_using_ranges::simplify_min_or_max_using_ranges): Ditto.
21361 (simplify_using_ranges::simplify_internal_call_using_ranges): Ditto.
21362 (simplify_using_ranges::two_valued_val_range_p): Ditto.
21363 (simplify_using_ranges::simplify): Ditto.
21364 * vr-values.h: Adjust prototypes.
21366 2021-06-29 Uroš Bizjak <ubizjak@gmail.com>
21369 * config/i386/mmx.md (vec_addsubv2sf3): New insn pattern.
21371 2021-06-29 Julian Brown <julian@codesourcery.com>
21373 * config/gcn/gcn.c (gcn_init_libfuncs): New function.
21374 (TARGET_INIT_LIBFUNCS): Define target hook using above function.
21375 * config/gcn/gcn.h (UNITS_PER_WORD): Define to 8 for IN_LIBGCC2, 4
21377 (LIBGCC2_UNITS_PER_WORD, BITS_PER_WORD): Remove definitions.
21378 (MAX_FIXED_MODE_SIZE): Change to 128.
21380 2021-06-29 Julian Brown <julian@codesourcery.com>
21382 * config/gcn/gcn.md (UNSPEC_FLBIT_INT): New unspec constant.
21383 (s_mnemonic): Add clrsb.
21384 (gcn_flbit<mode>_int): Add insn pattern for SImode/DImode.
21385 (clrsb<mode>2): Add expander for SImode/DImode.
21387 2021-06-29 Julian Brown <julian@codesourcery.com>
21389 * config/gcn/gcn.md (<su>mulsidi3, <su>mulsidi3_reg, <su>mulsidi3_imm,
21390 muldi3): Add patterns.
21392 2021-06-29 Julian Brown <julian@codesourcery.com>
21394 * config/gcn/gcn.md (<su>mulsi3_highpart): Change to expander.
21395 (<su>mulsi3_highpart_reg, <su>mulsi3_highpart_imm): New patterns.
21397 2021-06-29 Julian Brown <julian@codesourcery.com>
21399 * config/gcn/gcn.md (mulsi3): Make s_mulk_i32 variant clobber SCC.
21401 2021-06-29 Joseph Myers <joseph@codesourcery.com>
21403 * btfout.c, ctfout.c: Include "memmodel.h".
21405 2021-06-29 Tobias Burnus <tobias@codesourcery.com>
21407 * gcc.c (check_offload_target_name): Cast len argument to
21408 %q.*s to 'int'; avoid -Wstringop-truncation warning.
21410 2021-06-29 Richard Biener <rguenther@suse.de>
21412 * tree-vect-slp.c (vect_optimize_slp): Forward propagate
21413 to "any" permute nodes and relax "any" permute proapgation
21414 during iterative backward propagation.
21416 2021-06-29 Tobias Burnus <tobias@codesourcery.com>
21419 * common.opt (-foffload=): Update description.
21420 (-foffload-options=): New.
21421 * doc/invoke.texi (C Language Options): Document
21422 -foffload and -foffload-options.
21423 * gcc.c (check_offload_target_name): New, split off from
21424 handle_foffload_option.
21425 (check_foffload_target_names): New.
21426 (handle_foffload_option): Handle -foffload=default.
21427 (driver_handle_option): Update for -foffload-options.
21428 * lto-opts.c (lto_write_options): Use -foffload-options
21429 instead of -foffload.
21430 * lto-wrapper.c (merge_and_complain, append_offload_options):
21432 * opts.c (common_handle_option): Likewise.
21434 2021-06-29 Tobias Burnus <tobias@codesourcery.com>
21436 * doc/invoke.texi (C Language Options): Sort options
21437 alphabetically in optlist and also the description itself.
21438 Remove leftover -fallow-single-precision from and add missing
21439 -fgnu-tm to the optlist.
21441 2021-06-29 Richard Biener <rguenther@suse.de>
21443 * tree-vect-slp.c (slpg_vertex::visited): Remove.
21444 (vect_slp_perms_eq): Handle -1 permutes.
21445 (vect_optimize_slp): Rewrite permute propagation.
21447 2021-06-29 Jakub Jelinek <jakub@redhat.com>
21450 * match.pd ((intptr_t)x eq/ne CST to x eq/ne (typeof x) CST): Don't
21451 perform the optimization in GENERIC when sanitizing and x has a
21454 2021-06-29 Richard Biener <rguenther@suse.de>
21456 PR tree-optimization/101242
21457 * tree-vect-slp.c (vect_slp_build_vertices): Force-add
21458 PHIs with not represented initial values as leafs.
21460 2021-06-29 Jan-Benedict Glaw <jbglaw@getslash.de>
21462 * config/pdp11/pdp11.h (ASM_OUTPUT_SKIP): Fix signedness warning.
21463 * config/pdp11/pdp11.c (pdp11_asm_print_operand_punct_valid_p): Remove
21464 "register" keyword.
21465 (pdp11_initial_elimination_offset) Remove unused variable.
21466 (pdp11_cmp_length) Ditto.
21467 (pdp11_insn_cost): Ditto, and fix signedness warning.
21469 2021-06-29 David Edelsohn <dje.gcc@gmail.com>
21471 * btfout.c: Include tm_p.h.
21474 2021-06-28 Indu Bhagat <indu.bhagat@oracle.com>
21476 * config/bpf/bpf.c (bpf_expand_prologue): Do not mark insns as
21478 (bpf_expand_epilogue): Likewise.
21479 * config/bpf/bpf.h (DWARF2_FRAME_INFO): Define to 0.
21480 Do not define DBX_DEBUGGING_INFO.
21482 2021-06-28 Indu Bhagat <indu.bhagat@oracle.com>
21484 * doc/invoke.texi: Document the CTF and BTF debug info options.
21486 2021-06-28 Indu Bhagat <indu.bhagat@oracle.com>
21487 David Faust <david.faust@oracle.com>
21488 Jose E. Marchesi <jose.marchesi@oracle.com>
21489 Weimin Pan <weimin.pan@oracle.com>
21491 * Makefile.in: Add ctfc.*, ctfout.c and btfout.c files to
21492 GTFILES. Add new object files.
21493 * common.opt: Add CTF and BTF debug info options.
21494 * btfout.c: New file.
21495 * ctfc.c: Likewise.
21496 * ctfc.h: Likewise.
21497 * ctfout.c: Likewise.
21498 * dwarf2ctf.c: Likewise.
21499 * dwarf2ctf.h: Likewise.
21500 * dwarf2cfi.c (dwarf2out_do_frame): Acknowledge CTF_DEBUG and
21502 * dwarf2out.c (dwarf2out_source_line): Likewise.
21503 (dwarf2out_finish): Skip emitting DWARF if CTF or BTF are to
21505 (debug_format_do_cu): New function.
21506 (dwarf2out_early_finish): Traverse DIEs and emit CTF/BTF for
21508 Include dwarf2ctf.c.
21509 * final.c (dwarf2_debug_info_emitted_p): Acknowledge DWARF-based debug
21511 * flag-types.h (enum debug_info_type): Add CTF_DEBUG and BTF_DEBUG.
21512 (CTF_DEBUG): New bitmask.
21513 (BTF_DEBUG): Likewise.
21514 (enum ctf_debug_info_levels): New enum.
21515 * gengtype.c (open_base_files): Handle ctfc.h.
21516 (main): Handle uint32_t type.
21517 * flags.h (btf_debuginfo_p): New definition.
21518 (dwarf_based_debuginfo_p): Likewise.
21519 * opts.c (debug_type_names): Add entries for CTF and BTF.
21520 (btf_debuginfo_p): New function.
21521 (dwarf_based_debuginfo_p): Likewise.
21522 (common_handle_option): Handle -gctfN and -gbtf options.
21523 (set_debug_level): Set CTF_DEBUG, BTF_DEBUG whenever appropriate.
21524 * toplev.c (process_options): Inform the user and ignore -gctfLEVEL if
21527 2021-06-28 Jose E. Marchesi <jose.marchesi@oracle.com>
21529 * dwarf2out.c (AT_class): Function is no longer static.
21530 (AT_int): Likewise.
21531 (AT_unsigned): Likewise.
21532 (AT_loc): Likewise.
21533 (get_AT): Likewise.
21534 (get_AT_string): Likewise.
21535 (get_AT_flag): Likewise.
21536 (get_AT_unsigned): Likewise.
21537 (get_AT_ref): Likewise.
21538 (new_die_raw): Likewise.
21539 (lookup_decl_die): Likewise.
21540 (base_type_die): Likewise.
21541 (add_name_attribute): Likewise.
21542 (add_AT_int): Likewise.
21543 (add_AT_unsigned): Likewise.
21544 (add_AT_loc): Likewise.
21545 (dw_get_die_tag): New function.
21546 (dw_get_die_child): Likewise.
21547 (dw_get_die_sib): Likewise.
21548 (struct dwarf_file_data): Move from here to dwarf2out.h
21549 (struct dw_attr_struct): Likewise.
21550 * dwarf2out.h: Analogous changes.
21552 2021-06-28 Martin Jambor <mjambor@suse.cz>
21555 * ipa-param-manipulation.h (class ipa_param_body_adjustments): New
21556 members m_dead_stmts and m_dead_ssas.
21557 * ipa-param-manipulation.c
21558 (ipa_param_body_adjustments::mark_dead_statements): New function.
21559 (ipa_param_body_adjustments::common_initialization): Call it on
21560 all removed but not split parameters.
21561 (ipa_param_body_adjustments::ipa_param_body_adjustments): Initialize
21563 (ipa_param_body_adjustments::modify_call_stmt): Remove arguments that
21565 * tree-inline.c (remap_gimple_stmt): Do not copy dead statements, reset
21566 dead debug statements.
21567 (copy_phis_for_bb): Do not copy dead PHI nodes.
21569 2021-06-28 Martin Jambor <mjambor@suse.cz>
21572 * symtab-clones.h (clone_info): Removed member param_adjustments.
21573 * ipa-param-manipulation.h: Adjust initial comment to reflect how we
21574 deal with pass-through splits now.
21575 (ipa_param_performed_split): Removed.
21576 (ipa_param_adjustments::modify_call): Adjusted parameters.
21577 (class ipa_param_body_adjustments): Adjusted parameters of
21578 register_replacement, modify_gimple_stmt and modify_call_stmt.
21579 (ipa_verify_edge_has_no_modifications): Declare.
21580 (ipa_edge_modifications_finalize): Declare.
21581 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Remove
21582 performed_splits processing, pas only edge to padjs->modify_call,
21583 check that call arguments were not modified if they should not have
21585 * cgraphclones.c (cgraph_node::create_clone): Do not copy performed
21587 * ipa-param-manipulation.c (struct pass_through_split_map): New type.
21588 (ipa_edge_modification_info): Likewise.
21589 (ipa_edge_modification_sum): Likewise.
21590 (ipa_edge_modifications): New edge summary.
21591 (ipa_verify_edge_has_no_modifications): New function.
21592 (transitive_split_p): Removed.
21593 (transitive_split_map): Likewise.
21594 (init_transitive_splits): Likewise.
21595 (ipa_param_adjustments::modify_call): Adjusted to use the new edge
21596 summary instead of performed_splits.
21597 (ipa_param_body_adjustments::register_replacement): Drop dummy
21598 parameter, set base_index of the created ipa_param_body_replacement.
21599 (phi_arg_will_live_p): New function.
21600 (ipa_param_body_adjustments::common_initialization): Do not create
21601 IPA_SRA dummy decls.
21602 (simple_tree_swap_info): Removed.
21603 (remap_split_decl_to_dummy): Likewise.
21604 (record_argument_state_1): New function.
21605 (record_argument_state): Likewise.
21606 (ipa_param_body_adjustments::modify_call_stmt): New parameter
21607 orig_stmt. Do not work with dummy decls, save necessary info about
21608 changes to ipa_edge_modifications.
21609 (ipa_param_body_adjustments::modify_gimple_stmt): New parameter
21610 orig_stmt, pass it to modify_call_stmt.
21611 (ipa_param_body_adjustments::modify_cfun_body): Adjust call to
21612 modify_gimple_stmt.
21613 (ipa_edge_modifications_finalize): New function.
21614 * tree-inline.c (remap_gimple_stmt): Pass original statement to
21615 modify_gimple_stmt.
21616 (copy_phis_for_bb): Do not copy dead PHI nodes.
21617 (expand_call_inline): Do not remap performed_splits.
21618 (update_clone_info): Likewise.
21619 * toplev.c: Include ipa-param-manipulation.h.
21620 (toplev::finalize): Call ipa_edge_modifications_finalize.
21622 2021-06-28 Andrew Pinski <apinski@marvell.com>
21624 * tree-ssa-phiopt.c (replace_phi_edge_with_variable): Duplicate range
21625 info if we're the only things setting the target PHI.
21626 (value_replacement): Don't duplicate range here.
21627 (minmax_replacement): Likewise.
21629 2021-06-28 Richard Biener <rguenther@suse.de>
21631 PR tree-optimization/101229
21632 * gimple-walk.c (gimple_walk_op): Handle PHIs.
21634 2021-06-28 Martin Liska <mliska@suse.cz>
21636 * config/v850/v850.c (construct_dispose_instruction): Allocate
21638 (construct_prepare_instruction): Likewise.
21640 2021-06-28 Martin Liska <mliska@suse.cz>
21642 * config/v850/v850.c (v850_option_override): Build default
21644 (v850_can_inline_p): New. Allow MASK_PROLOG_FUNCTION to be
21645 ignored for inlining.
21646 (TARGET_CAN_INLINE_P): New.
21648 2021-06-28 Richard Biener <rguenther@suse.de>
21650 PR tree-optimization/101207
21651 * tree-vect-slp.c (vect_optimize_slp): Do BB reduction
21652 permute eliding for load permutations properly.
21654 2021-06-28 Richard Biener <rguenther@suse.de>
21656 PR tree-optimization/101173
21657 * gimple-loop-interchange.cc
21658 (tree_loop_interchange::valid_data_dependences): Disallow outer
21659 loop dependence distance of zero.
21661 2021-06-28 liuhongt <hongtao.liu@intel.com>
21664 * config/i386/sse.md (*avx_cmp<mode>3_lt): New
21665 define_insn_and_split.
21666 (*avx_cmp<mode>3_ltint): Ditto.
21667 (*avx2_pcmp<mode>3_3): Ditto.
21668 (*avx2_pcmp<mode>3_4): Ditto.
21669 (*avx2_pcmp<mode>3_5): Ditto.
21671 2021-06-28 liuhongt <hongtao.liu@intel.com>
21673 * config/i386/i386-builtin.def (IX86_BUILTIN_BLENDVPD256,
21674 IX86_BUILTIN_BLENDVPS256, IX86_BUILTIN_PBLENDVB256,
21675 IX86_BUILTIN_BLENDVPD, IX86_BUILTIN_BLENDVPS,
21676 IX86_BUILTIN_PBLENDVB128): Replace icode with
21678 * config/i386/i386.c (ix86_gimple_fold_builtin): Fold blendv
21680 * config/i386/sse.md (*<sse4_1_avx2>_pblendvb_lt_subreg_not):
21681 New pre_reload splitter.
21683 2021-06-27 Andrew Pinski <apinski@marvell.com>
21685 PR middle-end/101230
21686 * fold-const.c (fold_ternary_loc): Check
21687 the return value of invert_tree_comparison.
21689 2021-06-27 David Edelsohn <dje.gcc@gmail.com>
21691 * config.gcc: Add SPDX License Identifier.
21692 (powerpc-ibm-aix789): Default to aix73.h.
21693 (powerpc-ibm-aix7.2.*.*): New stanza.
21694 * config/rs6000/aix72.h: Add SPDX License Identifier.
21695 * config/rs6000/aix73.h: New file.
21697 2021-06-26 Jason Merrill <jason@redhat.com>
21699 * except.c: #include "dwarf2.h" instead of "dwarf2out.h".
21701 2021-06-26 Andrew Pinski <apinski@marvell.com>
21703 * genmatch.c (lower_cond): Copy for_subst_vec
21704 for the simplify also.
21705 (lower): Swap the order for lower_for and lower_cond.
21707 2021-06-26 Andrew Pinski <apinski@marvell.com>
21709 * tree-ssa-phiopt.c (match_simplify_replacement): Reset
21710 flow senatitive info on the moved ssa set.
21712 2021-06-26 Andrew Pinski <apinski@marvell.com>
21714 * fold-const.c (fold_cond_expr_with_comparison):
21715 Exand arg0 into comp_code, arg00, and arg01.
21716 (fold_ternary_loc): Use invert_tree_comparison
21717 instead of fold_invert_truthvalue for the case
21718 where we have A CMP B ? C : A.
21720 2021-06-25 Martin Sebor <msebor@redhat.com>
21722 PR middle-end/101216
21723 * calls.c (maybe_warn_rdwr_sizes): Use the no_warning constant.
21725 2021-06-25 Jeff Law <jeffreyalaw@gmail.com>
21727 * config/h8300/h8300.c (select_cc_mode): Handle ASHIFTRT and LSHIFTRT.
21729 2021-06-25 Richard Biener <rguenther@suse.de>
21731 PR tree-optimization/101202
21732 * tree-vect-slp.c (vect_optimize_slp): Explicitely handle
21735 2021-06-25 Richard Biener <rguenther@suse.de>
21737 * tree-vect-slp-patterns.c (addsub_pattern::build): Copy
21738 STMT_VINFO_REDUC_DEF from the original representative.
21740 2021-06-25 Martin Sebor <msebor@redhat.com>
21742 * builtins.c (warn_string_no_nul): Replace uses of TREE_NO_WARNING,
21743 gimple_no_warning_p and gimple_set_no_warning with
21744 warning_suppressed_p, and suppress_warning.
21746 (maybe_warn_for_bound): Same.
21747 (warn_for_access): Same.
21748 (check_access): Same.
21749 (expand_builtin_strncmp): Same.
21750 (fold_builtin_varargs): Same.
21751 * calls.c (maybe_warn_nonstring_arg): Same.
21752 (maybe_warn_rdwr_sizes): Same.
21753 * cfgexpand.c (expand_call_stmt): Same.
21754 * cgraphunit.c (check_global_declaration): Same.
21755 * fold-const.c (fold_undefer_overflow_warnings): Same.
21756 (fold_truth_not_expr): Same.
21757 (fold_unary_loc): Same.
21758 (fold_checksum_tree): Same.
21759 * gimple-array-bounds.cc (array_bounds_checker::check_array_ref): Same.
21760 (array_bounds_checker::check_mem_ref): Same.
21761 (array_bounds_checker::check_addr_expr): Same.
21762 (array_bounds_checker::check_array_bounds): Same.
21763 * gimple-expr.c (copy_var_decl): Same.
21764 * gimple-fold.c (gimple_fold_builtin_strcpy): Same.
21765 (gimple_fold_builtin_strncat): Same.
21766 (gimple_fold_builtin_stxcpy_chk): Same.
21767 (gimple_fold_builtin_stpcpy): Same.
21768 (gimple_fold_builtin_sprintf): Same.
21769 (fold_stmt_1): Same.
21770 * gimple-ssa-isolate-paths.c (diag_returned_locals): Same.
21771 * gimple-ssa-nonnull-compare.c (do_warn_nonnull_compare): Same.
21772 * gimple-ssa-sprintf.c (handle_printf_call): Same.
21773 * gimple-ssa-store-merging.c (imm_store_chain_info::output_merged_store): Same.
21774 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Same.
21775 * gimple-ssa-warn-restrict.h: Adjust declarations.
21776 (maybe_diag_access_bounds): Replace uses of TREE_NO_WARNING,
21777 gimple_no_warning_p and gimple_set_no_warning with
21778 warning_suppressed_p, and suppress_warning.
21779 (check_call): Same.
21780 (check_bounds_or_overlap): Same.
21781 * gimple.c (gimple_build_call_from_tree): Same.
21782 * gimplify.c (gimplify_return_expr): Same.
21783 (gimplify_cond_expr): Same.
21784 (gimplify_modify_expr_complex_part): Same.
21785 (gimplify_modify_expr): Same.
21786 (gimple_push_cleanup): Same.
21787 (gimplify_expr): Same.
21788 * omp-expand.c (expand_omp_for_generic): Same.
21789 (expand_omp_taskloop_for_outer): Same.
21790 * omp-low.c (lower_rec_input_clauses): Same.
21791 (lower_lastprivate_clauses): Same.
21792 (lower_send_clauses): Same.
21793 (lower_omp_target): Same.
21794 * tree-cfg.c (pass_warn_function_return::execute): Same.
21795 * tree-complex.c (create_one_component_var): Same.
21796 * tree-inline.c (remap_gimple_op_r): Same.
21797 (copy_tree_body_r): Same.
21798 (declare_return_variable): Same.
21799 (expand_call_inline): Same.
21800 * tree-nested.c (lookup_field_for_decl): Same.
21801 * tree-sra.c (create_access_replacement): Same.
21802 (generate_subtree_copies): Same.
21803 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Same.
21804 * tree-ssa-forwprop.c (combine_cond_expr_cond): Same.
21805 * tree-ssa-loop-ch.c (ch_base::copy_headers): Same.
21806 * tree-ssa-loop-im.c (execute_sm): Same.
21807 * tree-ssa-phiopt.c (cond_store_replacement): Same.
21808 * tree-ssa-strlen.c (maybe_warn_overflow): Same.
21809 (handle_builtin_strcpy): Same.
21810 (maybe_diag_stxncpy_trunc): Same.
21811 (handle_builtin_stxncpy_strncat): Same.
21812 (handle_builtin_strcat): Same.
21813 * tree-ssa-uninit.c (get_no_uninit_warning): Same.
21814 (set_no_uninit_warning): Same.
21815 (uninit_undefined_value_p): Same.
21816 (warn_uninit): Same.
21817 (maybe_warn_operand): Same.
21818 * tree-vrp.c (compare_values_warnv): Same.
21819 * vr-values.c (vr_values::extract_range_for_var_from_comparison_expr): Same.
21820 (test_for_singularity): Same.
21821 * gimple.h (warning_suppressed_p): New function.
21822 (suppress_warning): Same.
21823 (copy_no_warning): Same.
21824 (gimple_set_block): Call gimple_set_location.
21825 (gimple_set_location): Call copy_warning.
21827 2021-06-25 Martin Sebor <msebor@redhat.com>
21829 * tree.h (warning_suppressed_at, copy_warning,
21830 warning_suppressed_p, suppress_warning): New functions.
21832 2021-06-25 Martin Sebor <msebor@redhat.com>
21834 * Makefile.in (OBJS-libcommon): Add diagnostic-spec.o.
21835 * gengtype.c (open_base_files): Add diagnostic-spec.h.
21836 * diagnostic-spec.c: New file.
21837 * diagnostic-spec.h: New file.
21838 * tree.h (no_warning, all_warnings, suppress_warning_at): New
21840 * warning-control.cc: New file.
21842 2021-06-25 liuhongt <hongtao.liu@intel.com>
21845 * config/i386/i386.c (x86_order_regs_for_local_alloc):
21848 2021-06-24 Andrew MacLeod <amacleod@redhat.com>
21850 PR tree-optimization/101189
21851 * gimple-range-fold.cc (fold_using_range::range_of_range_op): Pass
21852 LHS range of condition to postfold routine.
21853 (fold_using_range::postfold_gcond_edges): Only process the TRUE or
21854 FALSE edge if the LHS range supports it being taken.
21855 * gimple-range-fold.h (postfold_gcond_edges): Add range parameter.
21857 2021-06-24 Andrew MacLeod <amacleod@redhat.com>
21859 * value-relation.cc (equiv_oracle::dump): Do not dump NULL blocks.
21860 (relation_oracle::find_relation_block): Check correct bitmap.
21861 (relation_oracle::dump): Do not dump NULL blocks.
21863 2021-06-24 Andrew MacLeod <amacleod@redhat.com>
21865 * gimple-range-cache.cc (ranger_cache::propagate_cache): Call
21866 range_on_edge instead of manually calculating.
21868 2021-06-24 Andrew MacLeod <amacleod@redhat.com>
21870 * range-op.cc: Fix comment.
21872 2021-06-24 Uroš Bizjak <ubizjak@gmail.com>
21875 * config/i386/i386-expand.c (ix86_expand_sse_unpack):
21876 Handle V8QI and V4HI modes.
21877 * config/i386/mmx.md (sse4_1_<any_extend:code>v4qiv4hi2):
21879 (sse4_1_<any_extend:code>v4qiv4hi2): Ditto.
21880 (mmxpackmode): New mode attribute.
21881 (vec_pack_trunc_<mmxpackmode:mode>): New expander.
21882 (mmxunpackmode): New mode attribute.
21883 (vec_unpacks_lo_<mmxunpackmode:mode>): New expander.
21884 (vec_unpacks_hi_<mmxunpackmode:mode>): Ditto.
21885 (vec_unpacku_lo_<mmxunpackmode:mode>): Ditto.
21886 (vec_unpacku_hi_<mmxunpackmode:mode>): Ditto.
21887 * config/i386/i386.md (extsuffix): Move from ...
21888 * config/i386/sse.md: ... here.
21890 2021-06-24 Eric Botcazou <ebotcazou@adacore.com>
21892 * dwarf2out.c (dwarf2out_assembly_start): Emit .file 0 marker here..
21893 (dwarf2out_finish): ...instead of here.
21895 2021-06-24 Eric Botcazou <ebotcazou@adacore.com>
21897 * configure.ac (--gdwarf-5 option): Use objdump instead of readelf.
21898 (working --gdwarf-4/--gdwarf-5 for all sources): Likewise.
21899 (--gdwarf-4 not refusing generated .debug_line): Adjust for Windows.
21900 * configure: Regenerate.
21902 2021-06-24 Richard Biener <rguenther@suse.de>
21904 * config/i386/sse.md (vec_addsubv4df3, vec_addsubv2df3,
21905 vec_addsubv8sf3, vec_addsubv4sf3): Merge into ...
21906 (vec_addsub<mode>3): ... using a new addsub_cst mode attribute.
21908 2021-06-24 Richard Biener <rguenther@suse.de>
21910 * config/i386/sse.md (avx_addsubv4df3): Rename to
21912 (avx_addsubv8sf3): Rename to vec_addsubv8sf3.
21913 (sse3_addsubv2df3): Rename to vec_addsubv2df3.
21914 (sse3_addsubv4sf3): Rename to vec_addsubv4sf3.
21915 * config/i386/i386-builtin.def: Adjust.
21916 * internal-fn.def (VEC_ADDSUB): New internal optab fn.
21917 * optabs.def (vec_addsub_optab): New optab.
21918 * tree-vect-slp-patterns.c (class addsub_pattern): New.
21919 (slp_patterns): Add addsub_pattern.
21920 * tree-vect-slp.c (vect_optimize_slp): Disable propagation
21921 across CFN_VEC_ADDSUB.
21922 * tree-vectorizer.h (vect_pattern::vect_pattern): Make
21924 * doc/md.texi (vec_addsub<mode>3): Document.
21926 2021-06-24 Jakub Jelinek <jakub@redhat.com>
21928 PR middle-end/101170
21929 * df-scan.c (df_ref_record): For paradoxical big-endian SUBREGs
21930 where regno + subreg_regno_offset wraps around use 0 as starting
21933 2021-06-24 Jakub Jelinek <jakub@redhat.com>
21935 PR middle-end/101172
21936 * stor-layout.c (finish_bitfield_representative): If nextf has
21937 error_mark_node type, set repr type to error_mark_node too.
21939 2021-06-24 Ilya Leoshkevich <iii@linux.ibm.com>
21941 * config/s390/s390.c (s390_function_profiler): Ignore labelno
21943 * config/s390/s390.h (NO_PROFILE_COUNTERS): Define.
21945 2021-06-24 Richard Biener <rguenther@suse.de>
21947 * tree-vect-slp.c (vect_optimize_slp): Do not propagate
21948 across operations that have different semantics on different
21951 2021-06-24 Jakub Jelinek <jakub@redhat.com>
21953 * tree.h (OMP_CLAUSE_MAP_IN_REDUCTION): Document meaning for OpenMP.
21954 * gimplify.c (gimplify_scan_omp_clauses): For OpenMP map clauses
21955 with OMP_CLAUSE_MAP_IN_REDUCTION flag partially defer gimplification
21956 of non-decl OMP_CLAUSE_DECL. For OMP_CLAUSE_IN_REDUCTION on
21957 OMP_TARGET user outer_ctx instead of ctx for placeholders and
21958 initializer/combiner gimplification.
21959 * omp-low.c (scan_sharing_clauses): Handle OMP_CLAUSE_MAP_IN_REDUCTION
21960 on target constructs.
21961 (lower_rec_input_clauses): Likewise.
21962 (lower_omp_target): Likewise.
21963 * omp-expand.c (expand_omp_target): Temporarily ignore nowait clause
21964 on target if in_reduction is present.
21966 2021-06-24 Kewen Lin <linkw@linux.ibm.com>
21968 * tree-predcom.c (class pcom_worker): New class.
21969 (release_chain): Renamed to...
21970 (pcom_worker::release_chain): ...this.
21971 (release_chains): Renamed to...
21972 (pcom_worker::release_chains): ...this.
21973 (aff_combination_dr_offset): Renamed to...
21974 (pcom_worker::aff_combination_dr_offset): ...this.
21975 (determine_offset): Renamed to...
21976 (pcom_worker::determine_offset): ...this.
21977 (class comp_ptrs): New class.
21978 (split_data_refs_to_components): Renamed to...
21979 (pcom_worker::split_data_refs_to_components): ...this,
21980 and update with class comp_ptrs.
21981 (suitable_component_p): Renamed to...
21982 (pcom_worker::suitable_component_p): ...this.
21983 (filter_suitable_components): Renamed to...
21984 (pcom_worker::filter_suitable_components): ...this.
21985 (valid_initializer_p): Renamed to...
21986 (pcom_worker::valid_initializer_p): ...this.
21987 (find_looparound_phi): Renamed to...
21988 (pcom_worker::find_looparound_phi): ...this.
21989 (add_looparound_copies): Renamed to...
21990 (pcom_worker::add_looparound_copies): ...this.
21991 (determine_roots_comp): Renamed to...
21992 (pcom_worker::determine_roots_comp): ...this.
21993 (determine_roots): Renamed to...
21994 (pcom_worker::determine_roots): ...this.
21995 (single_nonlooparound_use): Renamed to...
21996 (pcom_worker::single_nonlooparound_use): ...this.
21997 (remove_stmt): Renamed to...
21998 (pcom_worker::remove_stmt): ...this.
21999 (execute_pred_commoning_chain): Renamed to...
22000 (pcom_worker::execute_pred_commoning_chain): ...this.
22001 (execute_pred_commoning): Renamed to...
22002 (pcom_worker::execute_pred_commoning): ...this.
22003 (struct epcc_data): New member worker.
22004 (execute_pred_commoning_cbck): Call execute_pred_commoning
22005 with pcom_worker pointer.
22006 (find_use_stmt): Renamed to...
22007 (pcom_worker::find_use_stmt): ...this.
22008 (find_associative_operation_root): Renamed to...
22009 (pcom_worker::find_associative_operation_root): ...this.
22010 (find_common_use_stmt): Renamed to...
22011 (pcom_worker::find_common_use_stmt): ...this.
22012 (combinable_refs_p): Renamed to...
22013 (pcom_worker::combinable_refs_p): ...this.
22014 (reassociate_to_the_same_stmt): Renamed to...
22015 (pcom_worker::reassociate_to_the_same_stmt): ...this.
22016 (stmt_combining_refs): Renamed to...
22017 (pcom_worker::stmt_combining_refs): ...this.
22018 (combine_chains): Renamed to...
22019 (pcom_worker::combine_chains): ...this.
22020 (try_combine_chains): Renamed to...
22021 (pcom_worker::try_combine_chains): ...this.
22022 (prepare_initializers_chain): Renamed to...
22023 (pcom_worker::prepare_initializers_chain): ...this.
22024 (prepare_initializers): Renamed to...
22025 (pcom_worker::prepare_initializers): ...this.
22026 (prepare_finalizers_chain): Renamed to...
22027 (pcom_worker::prepare_finalizers_chain): ...this.
22028 (prepare_finalizers): Renamed to...
22029 (pcom_worker::prepare_finalizers): ...this.
22030 (tree_predictive_commoning_loop): Renamed to...
22031 (pcom_worker::tree_predictive_commoning_loop): ...this, adjust
22032 some calls and remove some cleanup code.
22033 (tree_predictive_commoning): Adjusted to use pcom_worker instance.
22034 (static variable looparound_phis): Remove.
22035 (static variable name_expansions): Remove.
22037 2021-06-24 Richard Biener <rguenther@suse.de>
22039 * tree-vect-slp.c (slpg_vertex): New struct.
22040 (vect_slp_build_vertices): Adjust.
22041 (vect_optimize_slp): Likewise. Maintain an outgoing permute
22042 and a materialized one.
22044 2021-06-24 Richard Biener <rguenther@suse.de>
22046 PR tree-optimization/101105
22047 * tree-vect-data-refs.c (vect_prune_runtime_alias_test_list):
22048 Only ignore steps when they are equal or scalar order is preserved.
22050 2021-06-24 liuhongt <hongtao.liu@intel.com>
22053 * config/i386/i386-expand.c (ix86_expand_vec_interleave):
22054 Adjust comments for ix86_expand_vecop_qihi2.
22055 (ix86_expand_vecmul_qihi): Renamed to ..
22056 (ix86_expand_vecop_qihi2): Adjust function prototype to
22057 support shift operation, add static to definition.
22058 (ix86_expand_vec_shift_qihi_constant): Add static to definition.
22059 (ix86_expand_vecop_qihi): Call ix86_expand_vecop_qihi2 and
22060 ix86_expand_vec_shift_qihi_constant.
22061 * config/i386/i386-protos.h (ix86_expand_vecmul_qihi): Deleted.
22062 (ix86_expand_vec_shift_qihi_constant): Deleted.
22063 * config/i386/sse.md (VI12_256_512_AVX512VL): New mode
22065 (mulv8qi3): Call ix86_expand_vecop_qihi directly, add
22066 condition TARGET_64BIT.
22067 (mul<mode>3): Ditto.
22068 (<insn><mode>3): Ditto.
22069 (vlshr<mode>3): Extend to support avx512 vlshr.
22070 (v<insn><mode>3): New expander for
22072 (v<insn>v8qi3): Ditto.
22073 (vashrv8hi3<mask_name>): Renamed to ..
22074 (vashr<mode>3): And extend to support V16QImode for avx512.
22075 (vashrv16qi3): Deleted.
22076 (vashrv2di3<mask_name>): Extend expander to support avx512
22079 2021-06-23 Dimitar Dimitrov <dimitar@dinux.eu>
22081 * doc/lto.texi (Design Overview): Update that slim objects are
22084 2021-06-23 Aaron Sawdey <acsawdey@linux.ibm.com>
22086 * config/rs6000/rs6000-cpus.def: Take OPTION_MASK_PCREL_OPT out
22087 of OTHER_POWER10_MASKS so it will not be enabled by default.
22089 2021-06-23 Richard Biener <rguenther@suse.de>
22090 Martin Jambor <mjambor@suse.cz>
22092 * tree-inline.c (setup_one_parameter): Set TREE_READONLY of the
22093 param replacement unconditionally. Adjust comment.
22095 2021-06-23 Andrew MacLeod <amacleod@redhat.com>
22097 * Makefile.in (OBJS): Add gimple-range-fold.o
22098 * gimple-range-fold.cc: New.
22099 * gimple-range-fold.h: New.
22100 * gimple-range-gori.cc (gimple_range_calc_op1): Move to here.
22101 (gimple_range_calc_op2): Ditto.
22102 * gimple-range-gori.h: Move prototypes to here.
22103 * gimple-range.cc: Adjust include files.
22104 (fur_source:fur_source): Relocate to gimple-range-fold.cc.
22105 (fur_source::get_operand): Ditto.
22106 (fur_source::get_phi_operand): Ditto.
22107 (fur_source::query_relation): Ditto.
22108 (fur_source::register_relation): Ditto.
22109 (class fur_edge): Ditto.
22110 (fur_edge::fur_edge): Ditto.
22111 (fur_edge::get_operand): Ditto.
22112 (fur_edge::get_phi_operand): Ditto.
22113 (fur_stmt::fur_stmt): Ditto.
22114 (fur_stmt::get_operand): Ditto.
22115 (fur_stmt::get_phi_operand): Ditto.
22116 (fur_stmt::query_relation): Ditto.
22117 (class fur_depend): Relocate to gimple-range-fold.h.
22118 (fur_depend::fur_depend): Relocate to gimple-range-fold.cc.
22119 (fur_depend::register_relation): Ditto.
22120 (fur_depend::register_relation): Ditto.
22121 (class fur_list): Ditto.
22122 (fur_list::fur_list): Ditto.
22123 (fur_list::get_operand): Ditto.
22124 (fur_list::get_phi_operand): Ditto.
22125 (fold_range): Ditto.
22126 (adjust_pointer_diff_expr): Ditto.
22127 (gimple_range_adjustment): Ditto.
22128 (gimple_range_base_of_assignment): Ditto.
22129 (gimple_range_operand1): Ditto.
22130 (gimple_range_operand2): Ditto.
22131 (gimple_range_calc_op1): Relocate to gimple-range-gori.cc.
22132 (gimple_range_calc_op2): Ditto.
22133 (fold_using_range::fold_stmt): Relocate to gimple-range-fold.cc.
22134 (fold_using_range::range_of_range_op): Ditto.
22135 (fold_using_range::range_of_address): Ditto.
22136 (fold_using_range::range_of_phi): Ditto.
22137 (fold_using_range::range_of_call): Ditto.
22138 (fold_using_range::range_of_builtin_ubsan_call): Ditto.
22139 (fold_using_range::range_of_builtin_call): Ditto.
22140 (fold_using_range::range_of_cond_expr): Ditto.
22141 (fold_using_range::range_of_ssa_name_with_loop_info): Ditto.
22142 (fold_using_range::relation_fold_and_or): Ditto.
22143 (fold_using_range::postfold_gcond_edges): Ditto.
22144 * gimple-range.h: Add gimple-range-fold.h to include files. Change
22145 GIMPLE_RANGE_STMT_H to GIMPLE_RANGE_H.
22146 (gimple_range_handler): Relocate to gimple-range-fold.h.
22147 (gimple_range_ssa_p): Ditto.
22148 (range_compatible_p): Ditto.
22149 (class fur_source): Ditto.
22150 (class fur_stmt): Ditto.
22151 (class fold_using_range): Ditto.
22152 (gimple_range_calc_op1): Relocate to gimple-range-gori.h
22153 (gimple_range_calc_op2): Ditto.
22155 2021-06-23 Andrew MacLeod <amacleod@redhat.com>
22157 PR tree-optimization/101148
22158 PR tree-optimization/101014
22159 * gimple-range-cache.cc (ranger_cache::ranger_cache): Adjust.
22160 (ranger_cache::~ranger_cache): Adjust.
22161 (ranger_cache::block_range): Check if propagation disallowed.
22162 (ranger_cache::propagate_cache): Disallow propagation if new value
22163 can't be stored properly.
22164 * gimple-range-cache.h (ranger_cache::m_propfail): New member.
22166 2021-06-23 Andrew MacLeod <amacleod@redhat.com>
22168 * gimple-range-cache.cc (class ssa_block_ranges): Adjust prototype.
22169 (sbr_vector::set_bb_range): Return true.
22170 (class sbr_sparse_bitmap): Adjust.
22171 (sbr_sparse_bitmap::set_bb_range): Return value.
22172 (block_range_cache::set_bb_range): Return value.
22173 (ranger_cache::propagate_cache): Use return value to print msg.
22174 * gimple-range-cache.h (class block_range_cache): Adjust.
22176 2021-06-23 Andrew MacLeod <amacleod@redhat.com>
22178 * gimple-range.cc (dump_bb): Use range_on_edge from the cache.
22180 2021-06-23 Jeff Law <jeffreyalaw@gmail.com>
22182 * config/h8300/logical.md (<code><mode>3<ccnz>): Use <cczn>
22183 so this pattern can be used for test/compare removal. Pass
22184 current insn to compute_logical_op_length and output_logical_op.
22185 * config/h8300/h8300.c (compute_logical_op_cc): Remove.
22186 (h8300_and_costs): Add argument to compute_logical_op_length.
22187 (output_logical_op): Add new argument. Use it to determine if the
22188 condition codes are used and adjust the output accordingly.
22189 (compute_logical_op_length): Add new argument and update length
22190 computations when condition codes are used.
22191 * config/h8300/h8300-protos.h (compute_logical_op_length): Update
22193 (output_logical_op): Likewise.
22195 2021-06-23 Uroš Bizjak <ubizjak@gmail.com>
22198 * config/i386/i386-expand.c (expand_vec_perm_pshufb):
22199 Handle 64bit modes for TARGET_XOP. Use indirect gen_* functions.
22200 * config/i386/mmx.md (mmx_ppermv64): New insn pattern.
22201 * config/i386/i386.md (unspec): Move UNSPEC_XOP_PERMUTE from ...
22202 * config/i386/sse.md (unspec): ... here.
22204 2021-06-23 Martin Liska <mliska@suse.cz>
22207 * optc-save-gen.awk: Put back arm_fp16_format to
22210 2021-06-23 Uroš Bizjak <ubizjak@gmail.com>
22213 * config/i386/i386.md (bsr_rex64): Add zero-flag setting RTX.
22216 (clz<mode>2): Update RTX pattern for additions.
22218 2021-06-23 Jakub Jelinek <jakub@redhat.com>
22220 PR middle-end/101167
22221 * omp-low.c (lower_omp_regimplify_p): Regimplify also PARM_DECLs
22222 and RESULT_DECLs that have DECL_HAS_VALUE_EXPR_P set.
22224 2021-06-22 Sergei Trofimovich <siarheit@google.com>
22226 * doc/rtl.texi: drop unbalanced parenthesis.
22228 2021-06-22 Richard Biener <rguenther@suse.de>
22230 PR middle-end/101156
22231 * gimplify.c (gimplify_expr): Remove premature incorrect
22234 2021-06-22 Jakub Jelinek <jakub@redhat.com>
22236 PR tree-optimization/101159
22237 * tree-vect-patterns.c (vect_recog_popcount_pattern): Fix some
22240 2021-06-22 Jakub Jelinek <jakub@redhat.com>
22242 PR middle-end/101160
22243 * function.c (assign_parms): For decl_result with TYPE_EMPTY_P type
22244 clear crtl->return_rtx instead of keeping it referencing a pseudo.
22246 2021-06-22 Jakub Jelinek <jakub@redhat.com>
22247 Andrew Pinski <apinski@marvell.com>
22249 PR tree-optimization/101162
22250 * fold-const.c (range_check_type): Handle OFFSET_TYPE like pointer
22253 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
22255 * range-op.cc (range_relational_tests): New.
22256 (range_op_tests): Call range_relational_tests.
22258 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
22260 * range-op.cc (operator_cast::lhs_op1_relation): New.
22261 (operator_identity::lhs_op1_relation): Mew.
22263 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
22265 * range-op.cc (operator_minus::op1_op2_relation_effect): New.
22267 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
22269 * range-op.cc (operator_plus::lhs_op1_relation): New.
22270 (operator_plus::lhs_op2_relation): New.
22272 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
22274 * gimple-range-cache.cc (ranger_cache::ranger_cache): Create a
22275 relation_oracle if dominators exist.
22276 (ranger_cache::~ranger_cache): Dispose of oracle.
22277 (ranger_cache::dump_bb): Dump oracle.
22278 * gimple-range.cc (fur_source::fur_source): New.
22279 (fur_source::get_operand): Use mmeber query.
22280 (fur_source::get_phi_operand): Use member_query.
22281 (fur_source::query_relation): New.
22282 (fur_source::register_dependency): Delete.
22283 (fur_source::register_relation): New.
22284 (fur_edge::fur_edge): Adjust.
22285 (fur_edge::get_phi_operand): Fix comment.
22286 (fur_edge::query): Delete.
22287 (fur_stmt::fur_stmt): Adjust.
22288 (fur_stmt::query): Delete.
22289 (fur_depend::fur_depend): Adjust.
22290 (fur_depend::register_relation): New.
22291 (fur_depend::register_relation): New.
22292 (fur_list::fur_list): Adjust.
22293 (fur_list::get_operand): Use member query.
22294 (fold_using_range::range_of_range_op): Process and query relations.
22295 (fold_using_range::range_of_address): Adjust dependency call.
22296 (fold_using_range::range_of_phi): Ditto.
22297 (gimple_ranger::gimple_ranger): New. Use ranger_ache oracle.
22298 (fold_using_range::relation_fold_and_or): New.
22299 (fold_using_range::postfold_gcond_edges): New.
22300 * gimple-range.h (class gimple_ranger): Adjust.
22301 (class fur_source): Adjust members.
22302 (class fur_stmt): Ditto.
22303 (class fold_using_range): Ditto.
22305 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
22307 * range-op.cc (range_operator::wi_fold): Apply relation effect.
22308 (range_operator::fold_range): Adjust and apply relation effect.
22309 (*::fold_range): Add relation parameters.
22310 (*::op1_range): Ditto.
22311 (*::op2_range): Ditto.
22312 (range_operator::lhs_op1_relation): New.
22313 (range_operator::lhs_op2_relation): New.
22314 (range_operator::op1_op2_relation): New.
22315 (range_operator::op1_op2_relation_effect): New.
22316 (relop_early_resolve): New.
22317 (operator_equal::op1_op2_relation): New.
22318 (operator_equal::fold_range): Call relop_early_resolve.
22319 (operator_not_equal::op1_op2_relation): New.
22320 (operator_not_equal::fold_range): Call relop_early_resolve.
22321 (operator_lt::op1_op2_relation): New.
22322 (operator_lt::fold_range): Call relop_early_resolve.
22323 (operator_le::op1_op2_relation): New.
22324 (operator_le::fold_range): Call relop_early_resolve.
22325 (operator_gt::op1_op2_relation): New.
22326 (operator_gt::fold_range): Call relop_early_resolve.
22327 (operator_ge::op1_op2_relation): New.
22328 (operator_ge::fold_range): Call relop_early_resolve.
22329 * range-op.h (class range_operator): Adjust parameters and methods.
22331 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
22333 * Makefile.in (OBJS): Add value-relation.o.
22334 * gimple-range.h: Adjust include files.
22335 * tree-data-ref.c: Adjust include file order.
22336 * value-query.cc (range_query::get_value_range): Default to no oracle.
22337 (range_query::query_relation): New.
22338 (range_query::query_relation): New.
22339 * value-query.h (class range_query): Adjust.
22340 * value-relation.cc: New.
22341 * value-relation.h: New.
22343 2021-06-22 Richard Biener <rguenther@suse.de>
22345 PR tree-optimization/101151
22346 * tree-ssa-sink.c (statement_sink_location): Expand irreducible
22349 2021-06-22 Jojo R <rjiejie@linux.alibaba.com>
22351 * config/riscv/riscv.c (thead_c906_tune_info): New.
22352 (riscv_tune_info_table): Use new tune.
22354 2021-06-22 Richard Biener <rguenther@suse.de>
22356 PR tree-optimization/101158
22357 * tree-vect-slp.c (vect_build_slp_tree_1): Move same operand
22358 checking after checking for matching operation.
22360 2021-06-22 Richard Biener <rguenther@suse.de>
22362 PR tree-optimization/101159
22363 * tree-vect-patterns.c (vect_recog_popcount_pattern): Add
22364 missing NULL vectype check.
22366 2021-06-22 Richard Biener <rguenther@suse.de>
22368 PR tree-optimization/101154
22369 * tree-vect-slp.c (vect_build_slp_tree_2): Fix out-of-bound access.
22371 2021-06-22 Jakub Jelinek <jakub@redhat.com>
22374 * config/i386/i386-protos.h (ix86_last_zero_store_uid): Declare.
22375 * config/i386/i386-expand.c (ix86_last_zero_store_uid): New variable.
22376 * config/i386/i386.c (ix86_expand_prologue): Clear it.
22377 * config/i386/i386.md (peephole2s for 1/2/4 stores of const0_rtx):
22378 Remove "" from match_operand. Emit new insns using emit_move_insn and
22379 set ix86_last_zero_store_uid to INSN_UID of the last store.
22380 Add peephole2s for 1/2/4 stores of const0_rtx following previous
22383 2021-06-22 Martin Liska <mliska@suse.cz>
22385 * auto-profile.c (AUTO_PROFILE_VERSION): Bump as string format
22388 2021-06-22 Martin Liska <mliska@suse.cz>
22390 * gcov-io.h: Remove padding entries.
22392 2021-06-22 liuhongt <hongtao.liu@intel.com>
22394 PR tree-optimization/97770
22395 * tree-vect-patterns.c (vect_recog_popcount_pattern):
22397 (vect_recog_func vect_vect_recog_func_ptrs): Add new pattern.
22399 2021-06-22 liuhongt <hongtao.liu@intel.com>
22402 * config/i386/i386-builtin.def (BDESC): Adjust builtin name.
22403 * config/i386/sse.md (<avx512>_expand<mode>_mask): Rename to ..
22404 (expand<mode>_mask): this ..
22405 (*expand<mode>_mask): New pre_reload splitter to transform
22406 v{,p}expand* to vmov* when mask is zero, all ones, or has all
22407 ones in it's lower part, otherwise still generate
22410 2021-06-22 liuhongt <hongtao.liu@intel.com>
22413 * config/i386/i386-expand.c
22414 (ix86_expand_special_args_builtin): Keep constm1_operand only
22415 if it satisfies insn's operand predicate.
22417 2021-06-21 Jason Merrill <jason@redhat.com>
22420 * df-scan.c (df_ref_record): Check that regno < endregno.
22421 * function.c (assign_parms, expand_function_end): Do nothing with a
22422 TYPE_EMPTY_P result.
22424 2021-06-21 Richard Biener <rguenther@suse.de>
22426 PR tree-optimization/101120
22427 * tree-vect-data-refs.c (bump_vector_ptr): Fold the
22429 * tree-vect-slp.c (vect_transform_slp_perm_load): Add
22430 DR chain DCE capability.
22431 * tree-vectorizer.h (vect_transform_slp_perm_load): Adjust.
22432 * tree-vect-stmts.c (vectorizable_load): Remove unused
22433 loads in the DR chain for SLP.
22435 2021-06-21 Jakub Jelinek <jakub@redhat.com>
22437 PR inline-asm/100785
22438 * gimplify.c (gimplify_asm_expr): Don't diagnose errors if
22439 output or input operands were already error_mark_node.
22440 * cfgexpand.c (expand_asm_stmt): If errors are emitted,
22441 remove all inputs, outputs and clobbers from the asm and
22442 set template to "".
22444 2021-06-21 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
22446 * config/arm/arm_neon.h (vceq_s8): Replace builtin with __a == __b.
22447 (vceq_s16): Likewise.
22448 (vceq_s32): Likewise.
22449 (vceq_u8): Likewise.
22450 (vceq_u16): Likewise.
22451 (vceq_u32): Likewise.
22452 (vceq_p8): Likewise.
22453 (vceqq_s8): Likewise.
22454 (vceqq_s16): Likewise.
22455 (vceqq_s32): Likewise.
22456 (vceqq_u8): Likewise.
22457 (vceqq_u16): Likewise.
22458 (vceqq_u32): Likewise.
22459 (vceqq_p8): Likewise.
22460 (vceq_f32): Gate __a == __b on __FAST_MATH__.
22461 (vceqq_f32): Likewise.
22462 (vceq_f16): Likewise.
22463 (vceqq_f16): Likewise.
22465 2021-06-21 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
22468 * config/arm/iterators.md (NEON_VACMP): Remove.
22469 * config/arm/neon.md (neon_vca<cmp_op><mode>): Use GLTE instead of GTGE
22471 (neon_vca<cmp_op><mode>_insn): Likewise.
22472 (neon_vca<cmp_op_unsp><mode>_insn_unspec): Use NEON_VAGLTE instead of
22475 2021-06-21 Richard Biener <rguenther@suse.de>
22477 PR tree-optimization/101121
22478 * tree-vect-slp.c (vect_build_slp_tree_2): To not fail fatally
22479 when we just lack a stmt with the desired op when doing permutation.
22480 (vect_build_slp_tree): When caching a failed SLP build attempt
22481 assert that at least one lane is marked as not matching.
22483 2021-06-21 liuhongt <hongtao.liu@intel.com>
22486 * config/i386/i386.md: (*anddi_1): Disparage slightly the mask
22487 register alternative.
22488 (*and<mode>_1): Ditto.
22490 (*andn<mode>_1): Ditto.
22491 (*<code><mode>_1): Ditto.
22492 (*<code>qi_1): Ditto.
22493 (*one_cmpl<mode>2_1): Ditto.
22494 (*one_cmplsi2_1_zext): Ditto.
22495 (*one_cmplqi2_1): Ditto.
22496 * config/i386/i386.c (x86_order_regs_for_local_alloc): Change
22497 the order of mask registers to be before general registers.
22499 2021-06-21 Roger Sayle <roger@nextmovesoftware.com>
22502 * config/i386/i386.md: New define_peephole2s to shrink writing
22503 1, 2 or 4 consecutive zeros to memory when optimizing for size.
22505 2021-06-18 Jeff Law <jeffreyalaw@gmail.com>
22507 * config/h8300/h8300.c (h8300_select_cc_mode): Handle SYMBOL_REF.
22508 * config/h8300/logical.md (<code><mode>3 logcial expander): Generate
22509 more efficient code when the source can be trivially simplified.
22511 2021-06-18 Andrew MacLeod <amacleod@redhat.com>
22513 * gimple-range-cache.cc (ranger_cache::range_of_def): Calculate
22514 a range if global is not available.
22515 (ranger_cache::entry_range): Fallback to range_of_def.
22516 * gimple-range-cache.h (range_of_def): Adjust prototype.
22518 2021-06-18 Andrew MacLeod <amacleod@redhat.com>
22520 PR tree-optimization/101014
22521 * gimple-range-cache.cc (ranger_cache::ranger_cache): Remove poor
22523 (ranger_cache::~ranger_cache): Ditto.
22524 (ranger_cache::enable_new_values): Delete.
22525 (ranger_cache::push_poor_value): Delete.
22526 (ranger_cache::range_of_def): Remove poor value processing.
22527 (ranger_cache::entry_range): Ditto.
22528 (ranger_cache::fill_block_cache): Ditto.
22529 * gimple-range-cache.h (class ranger_cache): Remove poor value members.
22530 * gimple-range.cc (gimple_ranger::range_of_expr): Remove call.
22531 * gimple-range.h (class gimple_ranger): Adjust.
22533 2021-06-18 Srinath Parvathaneni <srinath.parvathaneni@arm.com>
22536 * common/config/arm/arm-common.c (arm_canon_arch_option_1): New function
22537 derived from arm_canon_arch.
22538 (arm_canon_arch_option): Call it.
22539 (arm_canon_arch_multilib_option): New function.
22540 * config/arm/arm-cpus.in (IGNORE_FOR_MULTILIB): New fgroup.
22541 * config/arm/arm.h (arm_canon_arch_multilib_option): New prototype.
22542 (CANON_ARCH_MULTILIB_SPEC_FUNCTION): New macro.
22543 (MULTILIB_ARCH_CANONICAL_SPECS): New macro.
22544 (DRIVER_SELF_SPECS): Add MULTILIB_ARCH_CANONICAL_SPECS.
22545 * config/arm/arm.opt (mlibarch): New option.
22546 * config/arm/t-rmprofile (MULTILIB_MATCHES): For armv8*-m, replace use
22547 of march on RHS with mlibarch.
22549 2021-06-18 Marcel Vollweiler <marcel@codesourcery.com>
22551 * config.in: Regenerate.
22552 * config/gcn/gcn.c (print_operand_address): Fix for global_load assembler
22554 * configure: Regenerate.
22555 * configure.ac: Fix for global_load assembler functions.
22557 2021-06-18 Richard Biener <rguenther@suse.de>
22559 PR tree-optimization/101112
22560 * tree-vect-slp.c (vect_slp_linearize_chain): Fix condition
22561 to lookup a pattern stmt def.
22563 2021-06-18 Jakub Jelinek <jakub@redhat.com>
22565 PR middle-end/101062
22566 * stor-layout.c (finish_bitfield_layout): Don't add bitfield
22567 representatives in QUAL_UNION_TYPE.
22569 2021-06-18 Andrew Pinski <apinski@marvell.com>
22571 * tree-ssa-phiopt.c (replace_phi_edge_with_variable):
22572 Add counting of how many times it is done.
22573 (factor_out_conditional_conversion): Likewise.
22574 (match_simplify_replacement): Likewise.
22575 (value_replacement): Likewise.
22576 (spaceship_replacement): Likewise.
22577 (cond_store_replacement): Likewise.
22578 (cond_if_else_store_replacement_1): Likewise.
22579 (hoist_adjacent_loads): Likewise.
22581 2021-06-18 Andrew Pinski <apinski@marvell.com>
22583 * tree-cfg.c (verify_gimple_assign_unary): Reject point and offset
22584 types on NEGATE_EXPR, ABS_EXPR, BIT_NOT_EXPR, PAREN_EXPR and CNONJ_EXPR.
22585 (verify_gimple_assign_binary): Reject point and offset types on
22586 MULT_EXPR, MULT_HIGHPART_EXPR, TRUNC_DIV_EXPR, CEIL_DIV_EXPR,
22587 FLOOR_DIV_EXPR, ROUND_DIV_EXPR, TRUNC_MOD_EXPR, CEIL_MOD_EXPR,
22588 FLOOR_MOD_EXPR, ROUND_MOD_EXPR, RDIV_EXPR, and EXACT_DIV_EXPR.
22590 2021-06-18 Michael Meissner <meissner@linux.ibm.com>
22592 * config/rs6000/rs6000.c (rs6000_emit_minmax): Add support for ISA
22593 3.1 IEEE 128-bit floating point xsmaxcqp/xsmincqp instructions.
22594 * config/rs6000/rs6000.md (s<minmax><mode>3, IEEE128 iterator):
22597 2021-06-17 Aaron Sawdey <acsawdey@linux.ibm.com>
22599 * config/rs6000/genfusion.pl (gen_logical_addsubf): Add
22600 earlyclobber to alts 0/1.
22601 (gen_addadd): Add earlyclobber to alts 0/1.
22602 * config/rs6000/fusion.md: Regenerate file.
22604 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
22606 * cfgloopanal.c (get_loop_hot_path): Make path an auto_vec.
22608 2021-06-17 Andrew MacLeod <amacleod@redhat.com>
22610 * gimple-range-cache.cc: Comment cleanups.
22611 * gimple-range-gori.cc: Comment cleanups.
22612 * gimple-range.cc: Comment/spacing cleanups
22613 * value-range.h: Comment cleanups.
22615 2021-06-17 H.J. Lu <hjl.tools@gmail.com>
22618 * calls.c (expand_call): Replace PUSH_ARGS with
22619 targetm.calls.push_argument (0).
22620 (emit_library_call_value_1): Likewise.
22621 * defaults.h (PUSH_ARGS): Removed.
22622 (PUSH_ARGS_REVERSED): Replace PUSH_ARGS with
22623 targetm.calls.push_argument (0).
22624 * expr.c (block_move_libcall_safe_for_call_parm): Likewise.
22625 (emit_push_insn): Pass the number bytes to push to
22626 targetm.calls.push_argument and pass 0 if ARGS_ADDR is 0.
22627 * hooks.c (hook_bool_uint_true): New.
22628 * hooks.h (hook_bool_uint_true): Likewise.
22629 * rtlanal.c (nonzero_bits1): Replace PUSH_ARGS with
22630 targetm.calls.push_argument (0).
22631 * target.def (push_argument): Add a targetm.calls hook.
22632 * targhooks.c (default_push_argument): New.
22633 * targhooks.h (default_push_argument): Likewise.
22634 * config/bpf/bpf.h (PUSH_ARGS): Removed.
22635 * config/cr16/cr16.c (TARGET_PUSH_ARGUMENT): New.
22636 * config/cr16/cr16.h (PUSH_ARGS): Removed.
22637 * config/i386/i386.c (ix86_push_argument): New.
22638 (TARGET_PUSH_ARGUMENT): Likewise.
22639 * config/i386/i386.h (PUSH_ARGS): Removed.
22640 * config/m32c/m32c.c (TARGET_PUSH_ARGUMENT): New.
22641 * config/m32c/m32c.h (PUSH_ARGS): Removed.
22642 * config/nios2/nios2.h (PUSH_ARGS): Likewise.
22643 * config/pru/pru.h (PUSH_ARGS): Likewise.
22644 * doc/tm.texi.in: Remove PUSH_ARGS documentation. Add
22645 TARGET_PUSH_ARGUMENT hook.
22646 * doc/tm.texi: Regenerated.
22648 2021-06-17 Uroš Bizjak <ubizjak@gmail.com>
22651 * config/i386/i386-expand.c (expand_vector_set_var):
22652 Handle V2FS mode remapping. Pass TARGET_MMX_WITH_SSE to
22653 ix86_expand_vector_init_duplicate.
22654 (ix86_expand_vector_init_duplicate): Emit insv_1 for
22655 QImode for !TARGET_PARTIAL_REG_STALL.
22656 * config/i386/predicates.md (vec_setm_mmx_operand): New predicate.
22657 * config/i386/mmx.md (vec_setv2sf): Use vec_setm_mmx_operand
22658 as operand 2 predicate. Call ix86_expand_vector_set_var
22659 for non-constant index operand.
22660 (vec_setv2si): Ditto.
22661 (vec_setv4hi): Ditto.
22662 (vec_setv8qi): ditto.
22664 2021-06-17 Aldy Hernandez <aldyh@redhat.com>
22666 PR tree-optimization/100790
22667 * gimple-range.cc (range_of_builtin_call): Cleanup clz and ctz
22670 2021-06-17 Martin Liska <mliska@suse.cz>
22672 * doc/invoke.texi: Use consistently -O1 instead of -O.
22674 2021-06-17 Martin Liska <mliska@suse.cz>
22676 * gcov-io.h: Update documentation entry about string format.
22678 2021-06-17 Marius Hillenbrand <mhillen@linux.ibm.com>
22681 * config/s390/vecintrin.h (vec_doublee): Fix to use
22682 __builtin_s390_vflls.
22683 (vec_floate): Fix to use __builtin_s390_vflrd.
22685 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
22687 * dominance.c (get_dominated_to_depth): Return auto_vec<basic_block>.
22688 * dominance.h (get_dominated_to_depth): Likewise.
22689 (get_all_dominated_blocks): Likewise.
22690 * cfgcleanup.c (delete_unreachable_blocks): Adjust.
22691 * gcse.c (hoist_code): Likewise.
22692 * tree-cfg.c (remove_edge_and_dominated_blocks): Likewise.
22693 * tree-parloops.c (oacc_entry_exit_ok): Likewise.
22694 * tree-ssa-dce.c (eliminate_unnecessary_stmts): Likewise.
22695 * tree-ssa-phiprop.c (pass_phiprop::execute): Likewise.
22697 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
22699 * dominance.c (get_dominated_by_region): Return auto_vec<basic_block>.
22700 * dominance.h (get_dominated_by_region): Likewise.
22701 * tree-cfg.c (gimple_duplicate_sese_region): Adjust.
22702 (gimple_duplicate_sese_tail): Likewise.
22703 (move_sese_region_to_fn): Likewise.
22705 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
22707 * dominance.c (get_dominated_by): Return auto_vec<basic_block>.
22708 * dominance.h (get_dominated_by): Likewise.
22709 * auto-profile.c (afdo_find_equiv_class): Adjust.
22710 * cfgloopmanip.c (duplicate_loop_to_header_edge): Likewise.
22711 * loop-unroll.c (unroll_loop_runtime_iterations): Likewise.
22712 * tree-cfg.c (test_linear_chain): Likewise.
22713 (test_diamond): Likewise.
22715 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
22717 * cfgloop.h (get_loop_hot_path): Return auto_vec<basic_block>.
22718 * cfgloopanal.c (get_loop_hot_path): Likewise.
22719 * tree-ssa-loop-ivcanon.c (tree_estimate_loop_size): Likewise.
22721 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
22723 * cgraph.c (cgraph_node::collect_callers): Return
22724 auto_vec<cgraph_edge *>.
22725 * cgraph.h (cgraph_node::collect_callers): Likewise.
22726 * ipa-cp.c (create_specialized_node): Adjust.
22727 (decide_about_value): Likewise.
22728 (decide_whether_version_node): Likewise.
22729 * ipa-sra.c (process_isra_node_results): Likewise.
22731 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
22733 * vec.h (vl_ptr>::using_auto_storage): Handle null m_vec.
22734 (auto_vec<T, 0>::auto_vec): Define move constructor, and delete copy
22736 (auto_vec<T, 0>::operator=): Define move assignment and delete copy
22739 2021-06-17 Aldy Hernandez <aldyh@redhat.com>
22741 * gimple-range.cc (debug_seed_ranger): New.
22742 (dump_ranger): New.
22743 (debug_ranger): New.
22745 2021-06-17 Richard Biener <rguenther@suse.de>
22747 PR tree-optimization/54400
22748 * tree-vectorizer.h (enum slp_instance_kind): Add
22749 slp_inst_kind_bb_reduc.
22750 (reduction_fn_for_scalar_code): Declare.
22751 * tree-vect-data-refs.c (vect_slp_analyze_instance_dependence):
22752 Check SLP_INSTANCE_KIND instead of looking at the
22754 (vect_slp_analyze_instance_alignment): Likewise.
22755 * tree-vect-loop.c (reduction_fn_for_scalar_code): Export.
22756 * tree-vect-slp.c (vect_slp_linearize_chain): Split out
22757 chain linearization from vect_build_slp_tree_2 and generalize
22758 for the use of BB reduction vectorization.
22759 (vect_build_slp_tree_2): Adjust accordingly.
22760 (vect_optimize_slp): Elide permutes at the root of BB reduction
22762 (vectorizable_bb_reduc_epilogue): New function.
22763 (vect_slp_prune_covered_roots): Likewise.
22764 (vect_slp_analyze_operations): Use them.
22765 (vect_slp_check_for_constructors): Recognize associatable
22766 chains for BB reduction vectorization.
22767 (vectorize_slp_instance_root_stmt): Generate code for the
22768 BB reduction epilogue.
22770 2021-06-17 Andrew MacLeod <amacleod@redhat.com>
22772 * gimple-range-gori.cc (gori_compute::has_edge_range_p): Check with
22774 (gori_compute::may_recompute_p): New.
22775 (gori_compute::outgoing_edge_range_p): Perform recomputations.
22776 * gimple-range-gori.h (class gori_compute): Add prototype.
22778 2021-06-17 Andrew MacLeod <amacleod@redhat.com>
22780 * gimple-range-cache.cc (ranger_cache::range_on_edge): Always return
22781 true when a range can be calculated.
22782 * gimple-range.cc (gimple_ranger::dump_bb): Check has_edge_range_p.
22784 2021-06-16 Martin Sebor <msebor@redhat.com>
22786 * doc/invoke.texi (-Wmismatched-dealloc, -Wmismatched-new-delete):
22787 Correct documented defaults.
22789 2021-06-16 Andrew MacLeod <amacleod@redhat.com>
22791 * gimple-range-cache.cc (ranger_cache::ranger_cache): Initialize
22792 m_new_value_p directly.
22794 2021-06-16 Uroš Bizjak <ubizjak@gmail.com>
22797 * config/i386/i386-expand.c (expand_vec_perm_2perm_pblendv):
22798 Handle 64bit modes for TARGET_SSE4_1.
22799 (expand_vec_perm_pshufb2): Handle 64bit modes for TARGET_SSSE3.
22800 (expand_vec_perm_even_odd_pack): Handle V4HI mode.
22801 (expand_vec_perm_even_odd_1) <case E_V4HImode>: Expand via
22802 expand_vec_perm_pshufb2 for TARGET_SSSE3 and via
22803 expand_vec_perm_even_odd_pack for TARGET_SSE4_1.
22804 * config/i386/mmx.md (mmx_packusdw): New insn pattern.
22806 2021-06-16 Jonathan Wright <jonathan.wright@arm.com>
22808 * config/aarch64/aarch64-simd.md (aarch64_<sur><addsub>hn<mode>):
22809 Change to an expander that emits the correct instruction
22810 depending on endianness.
22811 (aarch64_<sur><addsub>hn<mode>_insn_le): Define.
22812 (aarch64_<sur><addsub>hn<mode>_insn_be): Define.
22814 2021-06-16 Jonathan Wright <jonathan.wright@arm.com>
22816 * config/aarch64/aarch64-simd-builtins.def: Split generator
22817 for aarch64_<su>qmovn builtins into scalar and vector
22819 * config/aarch64/aarch64-simd.md (aarch64_<su>qmovn<mode>_insn_le):
22821 (aarch64_<su>qmovn<mode>_insn_be): Define.
22822 (aarch64_<su>qmovn<mode>): Split into scalar and vector
22823 variants. Change vector variant to an expander that emits the
22824 correct instruction depending on endianness.
22826 2021-06-16 Jonathan Wright <jonathan.wright@arm.com>
22828 * config/aarch64/aarch64-simd-builtins.def: Split generator
22829 for aarch64_sqmovun builtins into scalar and vector variants.
22830 * config/aarch64/aarch64-simd.md (aarch64_sqmovun<mode>):
22831 Split into scalar and vector variants. Change vector variant
22832 to an expander that emits the correct instruction depending
22834 (aarch64_sqmovun<mode>_insn_le): Define.
22835 (aarch64_sqmovun<mode>_insn_be): Define.
22837 2021-06-16 Jonathan Wright <jonathan.wright@arm.com>
22839 * config/aarch64/aarch64-simd.md (aarch64_xtn<mode>_insn_le):
22840 Define - modeling zero-high-half semantics.
22841 (aarch64_xtn<mode>): Change to an expander that emits the
22842 appropriate instruction depending on endianness.
22843 (aarch64_xtn<mode>_insn_be): Define - modeling zero-high-half
22845 (aarch64_xtn2<mode>_le): Rename to...
22846 (aarch64_xtn2<mode>_insn_le): This.
22847 (aarch64_xtn2<mode>_be): Rename to...
22848 (aarch64_xtn2<mode>_insn_be): This.
22849 (vec_pack_trunc_<mode>): Emit truncation instruction instead
22851 * config/aarch64/iterators.md (Vnarrowd): Add Vnarrowd mode
22852 attribute iterator.
22854 2021-06-16 Martin Jambor <mjambor@suse.cz>
22856 PR tree-optimization/100453
22857 * tree-sra.c (create_access): Disqualify any const candidates
22858 which are written to.
22859 (sra_modify_expr): Do not store sub-replacements back to a const base.
22860 (handle_unscalarized_data_in_subtree): Likewise.
22861 (sra_modify_assign): Likewise. Earlier, use TREE_READONLy test
22862 instead of constant_decl_p.
22864 2021-06-16 Jakub Jelinek <jakub@redhat.com>
22866 PR middle-end/101062
22867 * stor-layout.c (finish_bitfield_representative): For fields in unions
22868 assume nextf is always NULL.
22869 (finish_bitfield_layout): Compute bit field representatives also in
22870 unions, but handle it as if each bitfield was the only field in the
22873 2021-06-16 Richard Biener <rguenther@suse.de>
22875 PR tree-optimization/101088
22876 * tree-ssa-loop-im.c (sm_seq_valid_bb): Only look for
22877 supported refs on edges. Do not assert same ref but
22878 different kind stores are unsuported but mark them so.
22879 (hoist_memory_references): Only look for supported refs
22882 2021-06-16 Roger Sayle <roger@nextmovesoftware.com>
22884 PR rtl-optimization/46235
22885 * config/i386/i386.md: New define_split for bt followed by cmov.
22886 (*bt<mode>_setcqi): New define_insn_and_split for bt followed by setc.
22887 (*bt<mode>_setncqi): New define_insn_and_split for bt then setnc.
22888 (*bt<mode>_setnc<mode>): New define_insn_and_split for bt followed
22889 by setnc with zero extension.
22891 2021-06-16 Richard Biener <rguenther@suse.de>
22893 PR tree-optimization/101083
22894 * tree-vect-slp.c (vect_slp_build_two_operator_nodes): Get
22895 vectype as argument.
22896 (vect_build_slp_tree_2): Adjust.
22898 2021-06-15 Martin Sebor <msebor@redhat.com>
22900 PR middle-end/100876
22901 * builtins.c: (gimple_call_return_array): Account for size_t
22902 mangling as either unsigned int or unsigned long
22904 2021-06-15 Jeff Law <jeffreyalaw@gmail.com>
22906 * compare-elim.c (try_eliminate_compare): Run DCE to clean things
22907 up before eliminating comparisons.
22909 2021-06-15 Aldy Hernandez <aldyh@redhat.com>
22911 * range-op.cc (operator_bitwise_or::wi_fold): Make sure
22912 nonzero|X is nonzero.
22913 (range_op_bitwise_and_tests): Add tests for above.
22915 2021-06-15 Carl Love <cel@us.ibm.com>
22918 * config/rs6000/rs6000-builtin.def (VCMPEQUT): Fix the ICODE for the
22920 (VRLQ, VSLQ, VSRQ, VSRAQ): Remove unused BU_P10_OVERLOAD_2
22923 2021-06-15 Tobias Burnus <tobias@codesourcery.com>
22926 * gimplify.c (enum gimplify_defaultmap_kind): Add GDMK_SCALAR_TARGET.
22927 (struct gimplify_omp_ctx): Extend defaultmap array by one.
22928 (new_omp_context): Init defaultmap[GDMK_SCALAR_TARGET].
22929 (omp_notice_variable): Update type classification for Fortran.
22930 (gimplify_scan_omp_clauses): Update calls for new argument; handle
22931 GDMK_SCALAR_TARGET; for Fortran, GDMK_POINTER avoid GOVD_MAP_0LEN_ARRAY.
22932 * langhooks-def.h (lhd_omp_scalar_p): Add 'ptr_ok' argument.
22933 * langhooks.c (lhd_omp_scalar_p): Likewise.
22934 (LANG_HOOKS_OMP_ALLOCATABLE_P, LANG_HOOKS_OMP_SCALAR_TARGET_P): New.
22935 (LANG_HOOKS_DECLS): Add them.
22936 * langhooks.h (struct lang_hooks_for_decls): Add new hooks, update
22937 omp_scalar_p pointer type to include the new bool argument.
22939 2021-06-15 David Malcolm <dmalcolm@redhat.com>
22941 * doc/analyzer.texi
22942 (Special Functions for Debugging the Analyzer): Add
22943 __analyzer_dump_capacity.
22945 2021-06-15 Jakub Jelinek <jakub@redhat.com>
22948 * expr.c (expand_expr_real_2) <case VEC_PACK_FIX_TRUNC_EXPR,
22949 case VEC_PACK_TRUNC_EXPR>: Clear subtarget when changing mode.
22951 2021-06-15 Richard Biener <rguenther@suse.de>
22953 * cfgloopanal.c (mark_irreducible_loops): Use a dominance
22954 check to identify loop latches.
22955 * cfgloop.c (verify_loop_structure): Likewise.
22956 * loop-init.c (apply_loop_flags): Allow marked irreducible
22957 regions even with multiple latches.
22958 * predict.c (rebuild_frequencies): Simplify.
22960 2021-06-15 Richard Biener <rguenther@suse.de>
22962 * tree-ssa-threadupdate.c
22963 (jump_thread_path_registry::mark_threaded_blocks): Assert we
22964 have marked irreducible regions.
22966 2021-06-14 Martin Sebor <msebor@redhat.com>
22969 * builtins.c (gimple_call_return_array): Check for attribute fn spec.
22970 Handle calls to placement new.
22971 (ndecl_dealloc_argno): Avoid placement delete.
22973 2021-06-14 Peter Bergner <bergner@linux.ibm.com>
22976 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_mma_builtin): Use
22977 create_tmp_reg_or_ssa_name().
22979 2021-06-14 Andrew MacLeod <amacleod@redhat.com>
22981 * gimple-range-cache.cc (ranger_cache::ranger_cache): Adjust.
22982 (ranger_cache::enable_new_values): Set to specified value and
22983 return the old value.
22984 (ranger_cache::disable_new_values): Delete.
22985 (ranger_cache::fill_block_cache): Disable non 1st order derived
22987 * gimple-range-cache.h (ranger_cache): Adjust prototypes.
22988 * gimple-range.cc (gimple_ranger::range_of_expr): Adjust.
22990 2021-06-14 Uroš Bizjak <ubizjak@gmail.com>
22993 * config/i386/i386-expand.c (ix86_vectorize_vec_perm_const):
22994 Return true early when testing with V2HImode.
22995 * config/i386/mmx.md (*punpckwd): Split to sse2_pshuflw_1.
22997 2021-06-14 Christophe Lyon <christophe.lyon@linaro.org>
22999 * config/arm/mve.md (mve_vec_unpack<US>_lo_<mode>): New pattern.
23000 (mve_vec_unpack<US>_hi_<mode>): New pattern.
23001 (@mve_vec_pack_trunc_lo_<mode>): New pattern.
23002 (mve_vmovntq_<supf><mode>): Prefix with '@'.
23003 * config/arm/neon.md (vec_unpack<US>_hi_<mode>): Move to
23005 (vec_unpack<US>_lo_<mode>): Likewise.
23006 (vec_pack_trunc_<mode>): Rename to
23007 neon_quad_vec_pack_trunc_<mode>.
23008 * config/arm/vec-common.md (vec_unpack<US>_hi_<mode>): New
23010 (vec_unpack<US>_lo_<mode>): New.
23011 (vec_pack_trunc_<mode>): New.
23013 2021-06-14 Richard Biener <rguenther@suse.de>
23015 PR tree-optimization/100934
23016 * tree-ssa-dom.c (pass_dominator::execute): Properly
23017 mark irreducible regions.
23019 2021-06-14 Martin Liska <mliska@suse.cz>
23021 * doc/invoke.texi: Put r{...} on the same line as @item.
23023 2021-06-14 Martin Liska <mliska@suse.cz>
23025 * doc/invoke.texi: Add missing newline.
23027 2021-06-14 Martin Liska <mliska@suse.cz>
23029 * doc/invoke.texi: Remove '+' charasters.
23031 2021-06-14 Claudiu Zissulescu <claziss@synopsys.com>
23033 * config.gcc (arc): Add support for with_cpu option.
23034 * config/arc/arc.h (OPTION_DEFAULT_SPECS): Add fpu.
23036 2021-06-14 Richard Biener <rguenther@suse.de>
23038 PR tree-optimization/101031
23039 * tree-ssa-strlen.c (maybe_invalidate): Increment max_size
23040 instead of size when accounting for a possibly string
23043 2021-06-14 Martin Liska <mliska@suse.cz>
23045 * gimple-ssa-evrp.c (pointer_equiv_analyzer::~pointer_equiv_analyzer): Use delete[].
23047 2021-06-14 Aldy Hernandez <aldyh@redhat.com>
23049 * value-query.cc (gimple_range_global): Call get_range_global
23050 if called after inlining.
23052 2021-06-13 Uroš Bizjak <ubizjak@gmail.com>
23055 * config/i386/i386-expand.c (expand_vec_perm_pshufb):
23056 Emit constant permutation insn directly from here.
23058 2021-06-13 Trevor Saunders <tbsaunde@tbsaunde.org>
23060 * attribs.c (find_attribute_namespace): Iterate over vec<> with
23062 * auto-profile.c (afdo_find_equiv_class): Likewise.
23063 * gcc.c (do_specs_vec): Likewise.
23064 (do_spec_1): Likewise.
23065 (driver::set_up_specs): Likewise.
23066 * gimple-loop-jam.c (any_access_function_variant_p): Likewise.
23067 * gimple-ssa-store-merging.c (compatible_load_p): Likewise.
23068 (imm_store_chain_info::try_coalesce_bswap): Likewise.
23069 (imm_store_chain_info::coalesce_immediate_stores): Likewise.
23070 (get_location_for_stmts): Likewise.
23071 * graphite-poly.c (print_iteration_domains): Likewise.
23072 (free_poly_bb): Likewise.
23073 (remove_gbbs_in_scop): Likewise.
23074 (free_scop): Likewise.
23075 (dump_gbb_cases): Likewise.
23076 (dump_gbb_conditions): Likewise.
23077 (print_pdrs): Likewise.
23078 (print_scop): Likewise.
23079 * ifcvt.c (cond_move_process_if_block): Likewise.
23080 * lower-subreg.c (decompose_multiword_subregs): Likewise.
23081 * regcprop.c (pass_cprop_hardreg::execute): Likewise.
23082 * sanopt.c (sanitize_rewrite_addressable_params): Likewise.
23083 * sel-sched-dump.c (dump_insn_vector): Likewise.
23084 * store-motion.c (store_ops_ok): Likewise.
23085 (store_killed_in_insn): Likewise.
23086 * timevar.c (timer::named_items::print): Likewise.
23087 * tree-cfgcleanup.c (cleanup_control_flow_pre): Likewise.
23088 (cleanup_tree_cfg_noloop): Likewise.
23089 * tree-data-ref.c (dump_data_references): Likewise.
23090 (print_dir_vectors): Likewise.
23091 (print_dist_vectors): Likewise.
23092 (dump_data_dependence_relations): Likewise.
23093 (dump_dist_dir_vectors): Likewise.
23094 (dump_ddrs): Likewise.
23095 (create_runtime_alias_checks): Likewise.
23096 (free_subscripts): Likewise.
23097 (save_dist_v): Likewise.
23098 (save_dir_v): Likewise.
23099 (invariant_access_functions): Likewise.
23100 (same_access_functions): Likewise.
23101 (access_functions_are_affine_or_constant_p): Likewise.
23102 (find_data_references_in_stmt): Likewise.
23103 (graphite_find_data_references_in_stmt): Likewise.
23104 (free_dependence_relations): Likewise.
23105 (free_data_refs): Likewise.
23106 * tree-inline.c (copy_debug_stmts): Likewise.
23107 * tree-into-ssa.c (dump_currdefs): Likewise.
23108 (rewrite_update_phi_arguments): Likewise.
23109 * tree-ssa-propagate.c (clean_up_loop_closed_phi): Likewise.
23110 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
23112 (vect_slp_analyze_node_dependences): Likewise.
23113 (vect_slp_analyze_instance_dependence): Likewise.
23114 (vect_record_base_alignments): Likewise.
23115 (vect_get_peeling_costs_all_drs): Likewise.
23116 (vect_peeling_supportable): Likewise.
23117 * tree-vectorizer.c (vec_info::~vec_info): Likewise.
23118 (vec_info::free_stmt_vec_infos): Likewise.
23120 2021-06-13 Jeff Law <jeffreyalaw@gmail.com>
23122 * config/h8300/logical.md (<code>qi3_1<cczn>): New pattern.
23123 (andqi3_1<cczn>): Removed.
23124 (<ors>qi3_1): Do not split for IOR/XOR a single bit.
23125 (H8/SX bit logicals): Split out from other patterns.
23126 * config/h8300/multiply.md (mulqihi3_const<cczn>): Renamed from
23127 mulqihi3_const_clobber_flags.
23128 (mulqihi3<cczn>, mulhisi3_const<cczn>, mulhisi3<cczn>): Similarly
23130 2021-06-13 H.J. Lu <hjl.tools@gmail.com>
23133 * config/i386/i386.c (ix86_expand_prologue): Set red_zone_used
23134 to true if red zone is used.
23135 (ix86_output_indirect_jmp): Replace ix86_red_zone_size with
23136 ix86_red_zone_used.
23137 * config/i386/i386.h (machine_function): Add red_zone_used.
23138 (ix86_red_zone_size): Removed.
23139 (ix86_red_zone_used): New.
23140 * config/i386/i386.md (peephole2 patterns): Replace
23141 ix86_red_zone_size with ix86_red_zone_used.
23143 2021-06-12 Jason Merrill <jason@redhat.com>
23145 * doc/extend.texi (unused variable attribute): Applies to
23146 structure fields as well.
23148 2021-06-12 Eugene Rozenfeld <erozen@microsoft.com>
23150 * auto-profile.c (read_profile): fix a typo in an error string
23152 2021-06-11 Thomas Schwinge <thomas@codesourcery.com>
23154 * tree-pretty-print.h (dump_omp_clauses): Add 'bool = true'
23156 * tree-pretty-print.c (dump_omp_clauses): Update.
23157 (dump_generic_node) <OMP_CLAUSE>: Use it.
23159 2021-06-11 Srinath Parvathaneni <srinath.parvathaneni@arm.com>
23162 * config/arm/arm_mve.h (__arm_vld1q): Change __ARM_mve_coerce(p0,
23163 int8_t const *) to __ARM_mve_coerce1(p0, int8_t *) in the argument for
23164 the polymorphic variants matching code.
23165 (__arm_vld1q_z): Likewise.
23166 (__arm_vld2q): Likewise.
23167 (__arm_vld4q): Likewise.
23168 (__arm_vldrbq_gather_offset): Likewise.
23169 (__arm_vldrbq_gather_offset_z): Likewise.
23171 2021-06-11 Roger Sayle <roger@nextmovesoftware.com>
23173 PR tree-optimization/96392
23174 * fold-const.h (tree_expr_maybe_real_minus_zero_p): Fix prototype.
23176 2021-06-11 Roger Sayle <roger@nextmovesoftware.com>
23178 PR tree-optimization/96392
23179 * fold-const.c (fold_real_zero_addition_p): Take both arguments
23180 of the addition or subtraction, not just the zero. Use this
23181 other argument in tests for signaling NaNs and signed zeros.
23182 (tree_expr_maybe_real_minus_zero_p): New predicate.
23183 * fold-const.h (fold_real_zero_addition_p): Update prototype.
23184 (tree_expr_maybe_real_minus_zero_p): New function prototype.
23185 * match.pd: Update calls to fold_real_zero_addition_p.
23186 Replace HONOR_NANS with tree_expr_maybe_nan_p.
23187 Replace HONOR_SIGNED_ZEROS with tree_expr_maybe_real_minus_zero_p.
23188 Replace HONOR_SNANS with tree_expr_maybe_signaling_nan_p.
23189 * tree-ssa-reassoc.c (eliminate_using_constants): Update
23190 call to fold_real_zero_addition_p.
23192 2021-06-11 Richard Biener <rguenther@suse.de>
23194 PR tree-optimization/101025
23195 * tree-ssa-loop-im.c (sm_seq_valid_bb): Make sure to process
23196 all refs that require dependence checking.
23198 2021-06-11 Richard Biener <rguenther@suse.de>
23200 PR tree-optimization/101028
23201 * tree-vect-slp.c (vect_build_slp_tree_2): When SLP
23202 reassoc discovery fails fatally, mark appropriate lanes
23205 2021-06-11 Richard Biener <rguenther@suse.de>
23207 PR tree-optimization/101026
23208 * tree-vect-slp.c (vect_build_slp_tree_2): Make sure we
23209 have a representative for the associated chain nodes.
23211 2021-06-11 Jakub Jelinek <jakub@redhat.com>
23213 PR rtl-optimization/101008
23214 * simplify-rtx.c (relational_result): New function.
23215 (simplify_logical_relational_operation,
23216 simplify_relational_operation): Use it.
23218 2021-06-11 Jakub Jelinek <jakub@redhat.com>
23221 * config/i386/sse.md (*vec_concat<mode>_0_1): Require TARGET_SSE2.
23223 2021-06-11 Uroš Bizjak <ubizjak@gmail.com>
23226 * config/i386/i386-expand.c (expand_vec_perm_pshufb): Return
23227 false if the permutation can be implemented with constant
23228 permutation instruction in wider mode.
23229 (canonicalize_vector_int_perm): Move above expand_vec_perm_pshufb.
23230 Handle V8QImode and V4HImode.
23232 2021-06-11 Martin Liska <mliska@suse.cz>
23234 PR gcov-profile/100788
23235 * common.opt: Add new option.
23236 * coverage.c (coverage_begin_function): Emit warning instead on
23237 the internal compiler error.
23238 * doc/invoke.texi: Document the option.
23239 * toplev.c (process_options): Enable it by default.
23241 2021-06-11 Richard Biener <rguenther@suse.de>
23243 PR middle-end/101009
23244 * tree-data-ref.c (build_classic_dist_vector_1): Make sure
23245 to set *init_b to true when we encounter a constant equal
23247 (compute_affine_dependence): Also dump the actual DR_REF.
23249 2021-06-10 Aldy Hernandez <aldyh@redhat.com>
23251 PR tree-optimization/100984
23252 * gimple-ssa-evrp.c (ssa_equiv_stack): Use auto_vec for
23253 replacements table.
23254 (ssa_equiv_stack::~ssa_equiv_stack): Remove.
23256 2021-06-11 Kewen Lin <linkw@linux.ibm.com>
23258 * config/rs6000/rs6000.md
23259 (floatsi<SFDF:mode>2_lfiwax_<QHI:mode>_mem_zext): New
23260 define_insn_and_split.
23262 2021-06-11 Richard Biener <rguenther@suse.de>
23264 * tree-vect-slp.c (vect_build_slp_tree_2): Use stablesort
23265 to sort operands of the associative chain.
23267 2021-06-11 Richard Biener <rguenther@suse.de>
23269 * system.h (gcc_stablesort_r): Declare.
23270 * sort.cc (gcc_sort_r): Support stable sort.
23271 (gcc_stablesort_r): Define.
23272 * vec.h (vec<>::stablesort): Add.
23274 2021-06-10 Uroš Bizjak <ubizjak@gmail.com>
23277 * config/i386/i386-expand.c (ix86_split_mmx_punpck):
23278 Handle V2SF mode. Emit SHUFPS to fixup unpack-high for V2SF mode.
23279 (expand_vec_perm_blend): Handle 64bit modes for TARGET_SSE4_1.
23280 (expand_vec_perm_pshufb): Handle 64bit modes for TARGET_SSSE3.
23281 (expand_vec_perm_pblendv): Handle 64bit modes for TARGET_SSE4_1.
23282 (expand_vec_perm_interleave2): Handle 64bit modes.
23283 (expand_vec_perm_even_odd_pack): Handle V8QI mode.
23284 (expand_vec_perm_even_odd_1): Ditto.
23285 (ix86_vectorize_vec_perm_const): Ditto.
23286 * config/i386/i386.md (UNSPEC_PSHUFB): Move from ...
23287 * config/i386/sse.md: ... here.
23288 * config/i386/mmx.md (*vec_interleave_lowv2sf):
23289 New insn_and_split pattern.
23290 (*vec_interleave_highv2sf): Ditto.
23291 (mmx_pshufbv8qi3): New insn pattern.
23292 (*mmx_pblendw): Ditto.
23294 2021-06-10 Peter Bergner <bergner@linux.ibm.com>
23296 * config/rs6000/rs6000-builtin.def (build_pair): New built-in.
23297 (build_acc): Likewise.
23298 * config/rs6000/rs6000-call.c (mma_expand_builtin): Swap assemble
23299 source operands in little-endian mode.
23300 (rs6000_gimple_fold_mma_builtin): Handle VSX_BUILTIN_BUILD_PAIR.
23301 (mma_init_builtins): Likewise.
23302 * config/rs6000/rs6000.c (rs6000_split_multireg_move): Handle endianness
23303 ordering for the MMA assemble and build source operands.
23304 * doc/extend.texi (__builtin_vsx_build_acc, __builtin_mma_build_pair):
23306 (__builtin_mma_assemble_acc, __builtin_mma_assemble_pair): Remove
23309 2021-06-10 Jeff Law <jeffreyalaw@gmail.com>
23311 * config/h8300/h8300.c (select_cc_mode): Handle MEM. Use
23313 * config/h8300/extensions.md: Replace _clobber_flags patterns
23316 2021-06-10 Robin Dapp <rdapp@linux.ibm.com>
23318 * config/s390/vector.md (vcond_mask_<mode><mode>): Change to
23319 (vcond_mask_<mode><tointvec>): this.
23321 2021-06-10 Andrew Stubbs <ams@codesourcery.com>
23322 Thomas Schwinge <thomas@codesourcery.com>
23324 * omp-builtins.def (BUILT_IN_GOACC_ENTER_EXIT_DATA): Split into...
23325 (BUILT_IN_GOACC_ENTER_DATA, BUILT_IN_GOACC_EXIT_DATA): ... these.
23326 * gimple.h (enum gf_mask): Split
23327 'GF_OMP_TARGET_KIND_OACC_ENTER_EXIT_DATA' into
23328 'GF_OMP_TARGET_KIND_OACC_ENTER_DATA' and
23329 'GF_OMP_TARGET_KIND_OACC_EXIT_DATA'.
23330 (is_gimple_omp_oacc): Update.
23331 * gimple-pretty-print.c (dump_gimple_omp_target): Likewise.
23332 * gimplify.c (gimplify_omp_target_update): Likewise.
23333 * omp-expand.c (expand_omp_target, build_omp_regions_1)
23334 (omp_make_gimple_edges): Likewise.
23335 * omp-low.c (check_omp_nesting_restrictions, lower_omp_target):
23338 2021-06-10 Aldy Hernandez <aldyh@redhat.com>
23340 * value-query.cc (value_query::value_on_edge): Rename name to
23342 (range_query::range_on_edge): Same.
23343 (range_query::value_of_expr): Same.
23344 (range_query::value_on_edge): Same.
23345 * value-query.h (class value_query): Same.
23346 (class range_query): Same.
23348 2021-06-10 Richard Biener <rguenther@suse.de>
23350 PR tree-optimization/101003
23351 * tree-vect-slp.c (vect_build_slp_tree_2): Appropriately
23352 use the pattern stmt defs when linearizing a chain.
23354 2021-06-10 Jakub Jelinek <jakub@redhat.com>
23357 * ifcvt.c (noce_get_alt_condition, noce_try_abs): Use
23358 prev_nonnote_nondebug_insn instead of prev_nonnote_insn.
23360 2021-06-10 Clement Chigot <clement.chigot@atos.net>
23362 * config/rs6000/aix71.h (ASM_CPU_SPEC): Add Power10 directive.
23363 * config/rs6000/aix72.h (ASM_CPU_SPEC): Likewise.
23365 2021-06-09 Andrew Pinski <apinski@marvell.com>
23367 PR tree-optimization/100925
23368 * match.pd (a ? CST1 : CST2): Limit transformations
23369 that would produce a negative to integeral types only.
23370 Change !POINTER_TYPE_P to INTEGRAL_TYPE_P also.
23372 2021-06-09 Jeff Law <jeffreyalaw@gmail.com>
23375 2021-06-09 Jeff Law <jeffreyalaw@gmail.com>
23377 * doc/tm.texi: Correctly update.
23379 2021-06-09 Jeff Law <jeffreyalaw@gmail.com>
23381 * doc/tm.texi: Correctly update.
23383 2021-06-09 H.J. Lu <hjl.tools@gmail.com>
23386 * doc/tm.texi.in (Trampolines): Add a missing blank line.
23388 2021-06-09 Paul Eggert <eggert@cs.ucla.edu>
23391 * doc/invoke.texi (Code Gen Options); Document that -fno-trampolines
23392 and -ftrampolines work only with Ada.
23393 * doc/tm.texi.in (Trampolines): Likewise.
23394 * doc/tm.texi: Regenerated.
23396 2021-06-09 Carl Love <cel@us.ibm.com>
23398 * config/rs6000/altivec.h (vec_signextll, vec_signexti, vec_signextq):
23399 Add define for new builtins.
23400 * config/rs6000/altivec.md(altivec_vreveti2): Add define_expand.
23401 * config/rs6000/rs6000-builtin.def (VSIGNEXTI, VSIGNEXTLL): Add
23402 overloaded builtin definitions.
23403 (VSIGNEXTSB2W, VSIGNEXTSH2W, VSIGNEXTSB2D, VSIGNEXTSH2D,VSIGNEXTSW2D,
23404 VSIGNEXTSD2Q): Add builtin expansions.
23405 (SIGNEXT): Add P10 overload definition.
23406 * config/rs6000/rs6000-call.c (P9V_BUILTIN_VEC_VSIGNEXTI, P9V_BUILTIN_VEC_VSIGNEXTLL,
23407 P10_BUILTIN_VEC_SIGNEXT): Add overloaded argument definitions.
23408 * config/rs6000/vsx.md (vsx_sign_extend_v2di_v1ti): Add define_insn.
23409 (vsignextend_v2di_v1ti, vsignextend_qi_<mode>, vsignextend_hi_<mode>,
23410 vsignextend_si_v2di)[VIlong]: Add define_expand.
23411 Make define_insn vsx_sign_extend_si_v2di visible.
23412 * doc/extend.texi: Add documentation for the vec_signexti,
23413 vec_signextll builtins and vec_signextq.
23415 2021-06-09 Carl Love <cel@us.ibm.com>
23417 * config/rs6000/rs6000.c (__fixkfti, __fixunskfti, __floattikf,
23418 __floatuntikf): Names changed to __fixkfti_sw, __fixunskfti_sw,
23419 __floattikf_sw, __floatuntikf_sw respectively.
23420 * config/rs6000/rs6000.md (floatti<mode>2, floatunsti<mode>2,
23421 fix_trunc<mode>ti2, fixuns_trunc<mode>ti2): Add
23422 define_insn for mode IEEE 128.
23424 2021-06-09 Carl Love <cel@us.ibm.com>
23426 * config/rs6000/altivec.md (altivec_vslq, altivec_vsrq):
23427 Rename to altivec_vslq_<mode>, altivec_vsrq_<mode>, mode VEC_TI.
23428 * config/rs6000/vector.md (VEC_TI): Was named VSX_TI in vsx.md.
23429 (vashlv1ti3): Change to vashl<mode>3, mode VEC_TI.
23430 (vlshrv1ti3): Change to vlshr<mode>3, mode VEC_TI.
23431 * config/rs6000/vsx.md (VSX_TI): Remove define_mode_iterator. Update
23432 uses of VSX_TI to VEC_TI.
23434 2021-06-09 Carl Love <cel@us.ibm.com>
23436 * config/rs6000/dfp.md (floattitd2, fixtdti2): New define_insns.
23438 2021-06-09 Carl Love <cel@us.ibm.com>
23440 * config/rs6000/altivec.h (vec_dive, vec_mod): Add define for new
23442 * config/rs6000/altivec.md (UNSPEC_VMULEUD, UNSPEC_VMULESD,
23443 UNSPEC_VMULOUD, UNSPEC_VMULOSD): New unspecs.
23444 (altivec_eqv1ti, altivec_gtv1ti, altivec_gtuv1ti, altivec_vmuleud,
23445 altivec_vmuloud, altivec_vmulesd, altivec_vmulosd, altivec_vrlq,
23446 altivec_vrlqmi, altivec_vrlqmi_inst, altivec_vrlqnm,
23447 altivec_vrlqnm_inst, altivec_vslq, altivec_vsrq, altivec_vsraq,
23448 altivec_vcmpequt_p, altivec_vcmpgtst_p, altivec_vcmpgtut_p): New
23450 (vec_widen_umult_even_v2di, vec_widen_smult_even_v2di,
23451 vec_widen_umult_odd_v2di, vec_widen_smult_odd_v2di, altivec_vrlqmi,
23452 altivec_vrlqnm): New define_expands.
23453 * config/rs6000/rs6000-builtin.def (VCMPEQUT_P, VCMPGTST_P,
23454 VCMPGTUT_P): Add macro expansions.
23455 (BU_P10V_AV_P): Add builtin predicate definition.
23456 (VCMPGTUT, VCMPGTST, VCMPEQUT, CMPNET, CMPGE_1TI,
23457 CMPGE_U1TI, CMPLE_1TI, CMPLE_U1TI, VNOR_V1TI_UNS, VNOR_V1TI, VCMPNET_P,
23458 VCMPAET_P, VMULEUD, VMULESD, VMULOUD, VMULOSD, VRLQ,
23459 VSLQ, VSRQ, VSRAQ, VRLQNM, DIV_V1TI, UDIV_V1TI, DIVES_V1TI, DIVEU_V1TI,
23460 MODS_V1TI, MODU_V1TI, VRLQMI): New macro expansions.
23461 (VRLQ, VSLQ, VSRQ, VSRAQ, DIVE, MOD): New overload expansions.
23462 * config/rs6000/rs6000-call.c (P10_BUILTIN_VCMPEQUT,
23463 P10V_BUILTIN_CMPGE_1TI, P10V_BUILTIN_CMPGE_U1TI,
23464 P10V_BUILTIN_VCMPGTUT, P10V_BUILTIN_VCMPGTST,
23465 P10V_BUILTIN_CMPLE_1TI, P10V_BUILTIN_VCMPLE_U1TI,
23466 P10V_BUILTIN_DIV_V1TI, P10V_BUILTIN_UDIV_V1TI,
23467 P10V_BUILTIN_VMULESD, P10V_BUILTIN_VMULEUD,
23468 P10V_BUILTIN_VMULOSD, P10V_BUILTIN_VMULOUD,
23469 P10V_BUILTIN_VNOR_V1TI, P10V_BUILTIN_VNOR_V1TI_UNS,
23470 P10V_BUILTIN_VRLQ, P10V_BUILTIN_VRLQMI,
23471 P10V_BUILTIN_VRLQNM, P10V_BUILTIN_VSLQ,
23472 P10V_BUILTIN_VSRQ, P10V_BUILTIN_VSRAQ,
23473 P10V_BUILTIN_VCMPGTUT_P, P10V_BUILTIN_VCMPGTST_P,
23474 P10V_BUILTIN_VCMPEQUT_P, P10V_BUILTIN_VCMPGTUT_P,
23475 P10V_BUILTIN_VCMPGTST_P, P10V_BUILTIN_CMPNET,
23476 P10V_BUILTIN_VCMPNET_P, P10V_BUILTIN_VCMPAET_P,
23477 P10V_BUILTIN_DIVES_V1TI, P10V_BUILTIN_MODS_V1TI,
23478 P10V_BUILTIN_MODU_V1TI):
23479 New overloaded definitions.
23480 (rs6000_gimple_fold_builtin) [P10V_BUILTIN_VCMPEQUT,
23481 P10V_BUILTIN_CMPNET, P10V_BUILTIN_CMPGE_1TI,
23482 P10V_BUILTIN_CMPGE_U1TI, P10V_BUILTIN_VCMPGTUT,
23483 P10V_BUILTIN_VCMPGTST, P10V_BUILTIN_CMPLE_1TI,
23484 P10V_BUILTIN_CMPLE_U1TI]: New case statements.
23485 (rs6000_init_builtins) [bool_V1TI_type_node, int_ftype_int_v1ti_v1ti]:
23487 (altivec_init_builtins): New E_V1TImode case statement.
23488 (builtin_function_type)[P10_BUILTIN_128BIT_VMULEUD,
23489 P10_BUILTIN_128BIT_VMULOUD, P10_BUILTIN_128BIT_DIVEU_V1TI,
23490 P10_BUILTIN_128BIT_MODU_V1TI, P10_BUILTIN_CMPGE_U1TI,
23491 P10_BUILTIN_VCMPGTUT, P10_BUILTIN_VCMPEQUT]: New case statements.
23492 * config/rs6000/rs6000.c (rs6000_handle_altivec_attribute) [E_TImode,
23493 E_V1TImode]: New case statements.
23494 * config/rs6000/rs6000.h (rs6000_builtin_type_index): New enum
23495 value RS6000_BTI_bool_V1TI.
23496 * config/rs6000/vector.md (vector_gtv1ti,vector_nltv1ti,
23497 vector_gtuv1ti, vector_nltuv1ti, vector_ngtv1ti, vector_ngtuv1ti,
23498 vector_eq_v1ti_p, vector_ne_v1ti_p, vector_ae_v1ti_p,
23499 vector_gt_v1ti_p, vector_gtu_v1ti_p, vrotlv1ti3, vashlv1ti3,
23500 vlshrv1ti3, vashrv1ti3): New define_expands.
23501 * config/rs6000/vsx.md (UNSPEC_VSX_DIVSQ, UNSPEC_VSX_DIVUQ,
23502 UNSPEC_VSX_DIVESQ, UNSPEC_VSX_DIVEUQ, UNSPEC_VSX_MODSQ,
23503 UNSPEC_VSX_MODUQ): New unspecs.
23504 (mulv2di3, vsx_div_v1ti, vsx_udiv_v1ti, vsx_dives_v1ti,
23505 vsx_diveu_v1ti, vsx_mods_v1ti, vsx_modu_v1ti, xxswapd_v1ti): New
23507 (vcmpnet): New define_expand.
23508 * doc/extend.texi: Add documentation for the new builtins vec_rl,
23509 vec_rlmi, vec_rlnm, vec_sl, vec_sr, vec_sra, vec_mule, vec_mulo,
23510 vec_div, vec_dive, vec_mod, vec_cmpeq, vec_cmpne, vec_cmpgt, vec_cmplt,
23511 vec_cmpge, vec_cmple, vec_all_eq, vec_all_ne, vec_all_gt, vec_all_lt,
23512 vec_all_ge, vec_all_le, vec_any_eq, vec_any_ne, vec_any_gt, vec_any_lt,
23513 vec_any_ge, vec_any_le.
23515 2021-06-09 Carl Love <cel@us.ibm.com>
23517 * config/rs6000/altivec.md (altivec_vrl<VI_char>mi): Fix
23518 bug in argument generation.
23520 2021-06-09 Christophe Lyon <christophe.lyon@linaro.org>
23522 * config/arm/iterators.md (<supf>): Remove VCLZQ_U, VCLZQ_S.
23524 * config/arm/mve.md (mve_vclzq_<supf><mode>): Add '@' prefix,
23525 remove <supf> iterator.
23526 (mve_vclzq_u<mode>): New.
23527 * config/arm/neon.md (clz<mode>2): Rename to neon_vclz<mode>.
23528 (neon_vclz<mode): Move to ...
23529 * config/arm/unspecs.md (VCLZQ_U, VCLZQ_S): Remove.
23530 * config/arm/vec-common.md: ... here. Add support for MVE.
23532 2021-06-09 Christophe Lyon <christophe.lyon@linaro.org>
23534 * config/arm/mve.md (mve_vhaddq_<supf><mode>): Prefix with '@'.
23535 (@mve_vrhaddq_<supf><mode): Likewise.
23536 * config/arm/neon.md (neon_v<r>hadd<sup><mode>): Likewise.
23537 * config/arm/vec-common.md (avg<mode>3_floor, uavg<mode>3_floor)
23538 (avg<mode>3_ceil", uavg<mode>3_ceil): New patterns.
23540 2021-06-09 imba-tjd <109224573@qq.com>
23542 * doc/invoke.texi: Fix typo.
23544 2021-06-09 Roger Sayle <roger@nextmovesoftware.com>
23546 PR middle-end/53267
23547 * fold-const-call.c (fold_const_call_sss) [CASE_CFN_FMOD]:
23548 Support evaluation of fmod/fmodf/fmodl at compile-time.
23550 2021-06-09 Richard Biener <rguenther@suse.de>
23552 PR tree-optimization/100981
23553 * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
23554 gimple_get_lhs to also handle calls.
23555 * tree-vect-slp-patterns.c (complex_pattern::build): Transfer
23558 2021-06-09 Richard Biener <rguenther@suse.de>
23560 PR tree-optimization/97832
23561 * tree-vectorizer.h (_slp_tree::failed): New.
23562 * tree-vect-slp.c (_slp_tree::_slp_tree): Initialize
23564 (_slp_tree::~_slp_tree): Free failed.
23565 (vect_build_slp_tree): Retain failed nodes and record
23566 matches in them, copying that back out when running
23567 into a cached fail. Dump start and end of discovery.
23568 (dt_sort_cmp): New.
23569 (vect_build_slp_tree_2): Handle associatable chains
23570 together doing more aggressive operand swapping.
23572 2021-06-09 H.J. Lu <hjl.tools@gmail.com>
23575 * config.gcc (gcc_cv_initfini_array): Set to yes for Linux and
23577 * doc/install.texi: Require glibc 2.1 and binutils 2.12 for
23578 Linux and GNU targets.
23580 2021-06-09 Richard Biener <rguenther@suse.de>
23582 * tree-vect-stmts.c (vect_is_simple_use): Always get dt
23585 2021-06-09 Claudiu Zissulescu <claziss@synopsys.com>
23587 * config/arc/arc.md (loop_end): Change it to
23588 define_insn_and_split.
23590 2021-06-09 Claudiu Zissulescu <claziss@synopsys.com>
23592 * config/arc/arc.md (maddhisi4): Use VMAC2H instruction.
23593 (machi): New pattern.
23594 (umaddhisi4): Use VMAC2HU instruction.
23595 (umachi): New pattern.
23597 2021-06-09 Claudiu Zissulescu <claziss@synopsys.com>
23599 * config/arc/arc-protos.h (arc_split_move_p): New prototype.
23600 * config/arc/arc.c (arc_split_move_p): New function.
23601 (arc_split_move): Clean up.
23602 * config/arc/arc.md (movdi_insn): Clean up, use arc_split_move_p.
23603 (movdf_insn): Likewise.
23604 * config/arc/simdext.md (mov<VWH>_insn): Likewise.
23606 2021-06-09 Uroš Bizjak <ubizjak@gmail.com>
23609 * config/i386/i386.c (print_operand_address_as): Rename "no_rip"
23610 argument to "raw". Do not emit segment overrides when "raw" is true.
23612 2021-06-09 Martin Liska <mliska@suse.cz>
23614 * doc/gcov.texi: Create a proper JSON files.
23615 * doc/invoke.texi: Remove dots in order to make it a valid
23618 2021-06-09 Xionghu Luo <luoxhu@linux.ibm.com>
23620 * config/rs6000/rs6000-p8swap.c (pattern_is_rotate64): New.
23621 (insn_is_load_p): Use pattern_is_rotate64.
23622 (insn_is_swap_p): Likewise.
23623 (quad_aligned_load_p): Likewise.
23624 (const_load_sequence_p): Likewise.
23625 (replace_swapped_aligned_load): Likewise.
23626 (recombine_lvx_pattern): Likewise.
23627 (recombine_stvx_pattern): Likewise.
23629 2021-06-09 Andrew MacLeod <amacleod@redhat.com>
23631 * gimple-range-gori.cc (gori_compute::outgoing_edge_range_p): Use a
23632 fur_stmt source record.
23633 * gimple-range.cc (fur_source::get_operand): Generic range query.
23634 (fur_source::get_phi_operand): New.
23635 (fur_source::register_dependency): New.
23636 (fur_source::query): New.
23637 (class fur_edge): New. Edge source for operands.
23638 (fur_edge::fur_edge): New.
23639 (fur_edge::get_operand): New.
23640 (fur_edge::get_phi_operand): New.
23641 (fur_edge::query): New.
23642 (fur_stmt::fur_stmt): New.
23643 (fur_stmt::get_operand): New.
23644 (fur_stmt::get_phi_operand): New.
23645 (fur_stmt::query): New.
23646 (class fur_depend): New. Statement source and process dependencies.
23647 (fur_depend::fur_depend): New.
23648 (fur_depend::register_dependency): New.
23649 (class fur_list): New. List source for operands.
23650 (fur_list::fur_list): New.
23651 (fur_list::get_operand): New.
23652 (fur_list::get_phi_operand): New.
23653 (fold_range): New. Instantiate appropriate fur_source class and fold.
23654 (fold_using_range::range_of_range_op): Use new API.
23655 (fold_using_range::range_of_address): Ditto.
23656 (fold_using_range::range_of_phi): Ditto.
23657 (imple_ranger::fold_range_internal): Use fur_depend class.
23658 (fold_using_range::range_of_ssa_name_with_loop_info): Use new API.
23659 * gimple-range.h (class fur_source): Now a base class.
23660 (class fur_stmt): New.
23661 (fold_range): New prototypes.
23662 (fur_source::fur_source): Delete.
23664 2021-06-08 Andrew Pinski <apinski@marvell.com>
23666 PR tree-optimization/25290
23667 * tree-ssa-phiopt.c (xor_replacement): Delete.
23668 (tree_ssa_phiopt_worker): Delete use of xor_replacement.
23669 (match_simplify_replacement): Allow one cheap preparation
23670 statement that can be moved to before the if.
23672 2021-06-08 Pat Haugen <pthaugen@linux.ibm.com>
23674 * config/rs6000/power10.md (power10-fused-load, power10-fused-store,
23675 power10-fused_alu, power10-fused-vec, power10-fused-branch): New.
23677 2021-06-08 Jeff Law <jeffreyalaw@gmail.com>
23679 * config/h8300/logical.md (andqi3_1): Move BCLR case into define_insn_and_split.
23680 Create length attribute on define_insn_and_split. Only split for cases which we
23682 (andqi3_1<cczn>): Renamed from andqi3_1_clobber_flags. Only handle AND here and
23683 fix length computation.
23684 (b<code><mode>msx): Combine QImode and HImode H8/SX patterns using iterator.
23686 2021-06-08 Richard Biener <rguenther@suse.de>
23688 PR tree-optimization/100923
23689 * tree-ssa-sccvn.c (valueize_refs_1): Take a pointer to
23690 the operand vector to be valueized.
23691 (valueize_refs): Likewise.
23692 (valueize_shared_reference_ops_from_ref): Adjust.
23693 (valueize_shared_reference_ops_from_call): Likewise.
23694 (vn_reference_lookup_3): Likewise.
23695 (vn_reference_lookup_pieces): Likewise. Re-valueize
23696 with honoring availability when we are about to create
23697 the ao_ref and valueized before.
23698 (vn_reference_lookup): Likewise.
23699 (vn_reference_insert_pieces): Adjust.
23701 2021-06-08 Richard Biener <rguenther@suse.de>
23703 * tree-vectorizer.h (_slp_instance::root_stmt): Change to...
23704 (_slp_instance::root_stmts): ... a vector.
23705 (SLP_INSTANCE_ROOT_STMT): Rename to ...
23706 (SLP_INSTANCE_ROOT_STMTS): ... this.
23707 (slp_root::root): Change to...
23708 (slp_root::roots): ... a vector.
23709 (slp_root::slp_root): Adjust.
23710 * tree-vect-slp.c (_slp_instance::location): Adjust.
23711 (vect_free_slp_instance): Release the root stmt vector.
23712 (vect_build_slp_instance): Adjust.
23713 (vect_analyze_slp): Likewise.
23714 (_bb_vec_info::~_bb_vec_info): Likewise.
23715 (vect_slp_analyze_operations): Likewise.
23716 (vect_bb_vectorization_profitable_p): Likewise. Adjust
23717 costs for the root stmt.
23718 (vect_slp_check_for_constructors): Gather all BIT_INSERT_EXPRs
23720 (vect_slp_analyze_bb_1): Simplify by marking all root stmts
23722 (vectorize_slp_instance_root_stmt): Adjust.
23723 (vect_schedule_slp): Likewise.
23725 2021-06-08 Aldy Hernandez <aldyh@redhat.com>
23727 * gimple-ssa-evrp.c (class ssa_equiv_stack): New.
23728 (ssa_equiv_stack::ssa_equiv_stack): New.
23729 (ssa_equiv_stack::~ssa_equiv_stack): New.
23730 (ssa_equiv_stack::enter): New.
23731 (ssa_equiv_stack::leave): New.
23732 (ssa_equiv_stack::push_replacement): New.
23733 (ssa_equiv_stack::get_replacement): New.
23734 (is_pointer_ssa): New.
23735 (class pointer_equiv_analyzer): New.
23736 (pointer_equiv_analyzer::pointer_equiv_analyzer): New.
23737 (pointer_equiv_analyzer::~pointer_equiv_analyzer): New.
23738 (pointer_equiv_analyzer::set_global_equiv): New.
23739 (pointer_equiv_analyzer::set_cond_equiv): New.
23740 (pointer_equiv_analyzer::get_equiv): New.
23741 (pointer_equiv_analyzer::enter): New.
23742 (pointer_equiv_analyzer::leave): New.
23743 (pointer_equiv_analyzer::get_equiv_expr): New.
23744 (pta_valueize): New.
23745 (pointer_equiv_analyzer::visit_stmt): New.
23746 (pointer_equiv_analyzer::visit_edge): New.
23747 (hybrid_folder::value_of_expr): Call PTA.
23748 (hybrid_folder::value_on_edge): Same.
23749 (hybrid_folder::pre_fold_bb): New.
23750 (hybrid_folder::post_fold_bb): New.
23751 (hybrid_folder::pre_fold_stmt): New.
23752 (rvrp_folder::pre_fold_bb): New.
23753 (rvrp_folder::post_fold_bb): New.
23754 (rvrp_folder::pre_fold_stmt): New.
23755 (rvrp_folder::value_of_expr): Call PTA.
23756 (rvrp_folder::value_on_edge): Same.
23758 2021-06-08 Jakub Jelinek <jakub@redhat.com>
23761 * tree-inline.c (copy_tree_body_r): For OMP_CLAUSE_DEPEND don't
23762 check TREE_CODE if OMP_CLAUSE_DECL is NULL.
23764 2021-06-08 Richard Biener <rguenther@suse.de>
23766 PR middle-end/100951
23767 * tree-vect-generic.c (expand_vector_piecewise): Build a
23768 VECTOR_CST if all elements are constant.
23769 (expand_vector_condition): Likewise.
23770 (lower_vec_perm): Likewise.
23771 (expand_vector_conversion): Likewise.
23773 2021-06-08 Martin Liska <mliska@suse.cz>
23775 * doc/invoke.texi: Document new param evrp-sparse-threshold.
23777 2021-06-08 Martin Liska <mliska@suse.cz>
23779 * genautomata.c (create_automata): Fix typo.
23781 2021-06-08 Kewen Lin <linkw@linux.ibm.com>
23783 PR tree-optimization/100794
23784 * tree-predcom.c (tree_predictive_commoning_loop): Add parameter
23785 allow_unroll_p and only allow unrolling when it's true.
23786 (tree_predictive_commoning): Add parameter allow_unroll_p and
23788 (run_tree_predictive_commoning): Likewise.
23789 (pass_predcom::gate): Check flag_tree_loop_vectorize and
23790 global_options_set.x_flag_predictive_commoning.
23791 (pass_predcom::execute): Adjust for allow_unroll_p.
23793 2021-06-08 Kewen Lin <linkw@linux.ibm.com>
23795 * tree-predcom.c (execute_pred_commoning): Remove update_ssa call.
23796 (tree_predictive_commoning_loop): Factor some cleanup stuffs into
23797 lambda function cleanup, remove scev_reset call, and adjust return
23799 (tree_predictive_commoning): Adjust for different changed values,
23800 only set flag TODO_update_ssa_only_virtuals if changed.
23801 (pass_data pass_data_predcom): Remove TODO_update_ssa_only_virtuals
23802 from todo_flags_finish.
23804 2021-06-07 Andrew MacLeod <amacleod@redhat.com>
23806 * gimple-range-cache.cc (class sbr_sparse_bitmap): New.
23807 (sbr_sparse_bitmap::sbr_sparse_bitmap): New.
23808 (sbr_sparse_bitmap::bitmap_set_quad): New.
23809 (sbr_sparse_bitmap::bitmap_get_quad): New.
23810 (sbr_sparse_bitmap::set_bb_range): New.
23811 (sbr_sparse_bitmap::get_bb_range): New.
23812 (sbr_sparse_bitmap::bb_range_p): New.
23813 (block_range_cache::block_range_cache): initialize bitmap obstack.
23814 (block_range_cache::~block_range_cache): Destruct obstack.
23815 (block_range_cache::set_bb_range): Decide when to utilze the
23816 sparse on entry cache.
23817 * gimple-range-cache.h (block_range_cache): Add bitmap obstack.
23818 * params.opt (-param=evrp-sparse-threshold): New.
23820 2021-06-07 Andrew MacLeod <amacleod@redhat.com>
23822 * bitmap.c (bitmap_set_aligned_chunk): New.
23823 (bitmap_get_aligned_chunk): New.
23824 (test_aligned_chunk): New.
23825 (bitmap_c_tests): Call test_aligned_chunk.
23826 * bitmap.h (bitmap_set_aligned_chunk, bitmap_get_aligned_chunk): New.
23828 2021-06-07 Uroš Bizjak <ubizjak@gmail.com>
23831 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
23833 (ix86_expand_vector_init_one_nonzero): Ditto.
23834 (ix86_expand_vector_init_one_var): Ditto.
23835 (ix86_expand_vector_init_general): Ditto.
23836 * config/i386/mmx.md (vec_initv4qiqi): New expander.
23838 2021-06-07 Jeff Law <jeffreyalaw@gmail.com>
23840 * config/h8300/movepush.md: Change most _clobber_flags
23841 patterns to instead use <cczn> subst.
23842 (movsi_cczn): New pattern with usable CC cases split out.
23843 (movsi_h8sx_cczn): Likewise.
23845 2021-06-07 Martin Liska <mliska@suse.cz>
23847 * common/common-target.def: Split long lines and replace them
23849 * target.def: Likewise.
23850 * doc/tm.texi: Re-generated.
23852 2021-06-07 Jakub Jelinek <jakub@redhat.com>
23855 * fold-const.c (fold_read_from_vector): Return NULL if trying to
23856 read from a CONSTRUCTOR with vector type elements.
23858 2021-06-07 Jakub Jelinek <jakub@redhat.com>
23860 PR middle-end/100898
23861 * tree-inline.c (copy_bb): Only use gimple_call_arg_ptr if memcpy
23862 should copy any arguments. Don't call gimple_call_num_args
23863 on id->call_stmt or call_stmt more than once.
23865 2021-06-07 liuhongt <hongtao.liu@intel.com>
23868 * config/i386/sse.md (*sse4_1_zero_extendv8qiv8hi2_3): Refine
23870 (<insn>v4siv4di2): Delete constraints for define_expand.
23872 2021-06-07 liuhongt <hongtao.liu@intel.com>
23875 * config/i386/i386-expand.c (ix86_expand_builtin): Remove
23876 assignment of cfun->machine->has_explicit_vzeroupper.
23877 * config/i386/i386-features.c
23878 (ix86_add_reg_usage_to_vzerouppers): Delete.
23879 (ix86_add_reg_usage_to_vzeroupper): Ditto.
23880 (rest_of_handle_insert_vzeroupper): Remove
23881 ix86_add_reg_usage_to_vzerouppers, add df_analyze at the end
23883 (gate): Remove cfun->machine->has_explicit_vzeroupper.
23884 * config/i386/i386-protos.h (ix86_expand_avx_vzeroupper):
23886 * config/i386/i386.c (ix86_insn_callee_abi): New function.
23887 (ix86_initialize_callee_abi): Ditto.
23888 (ix86_expand_avx_vzeroupper): Ditto.
23889 (ix86_hard_regno_call_part_clobbered): Adjust for vzeroupper
23891 (TARGET_INSN_CALLEE_ABI): Define as ix86_insn_callee_abi.
23892 (ix86_emit_mode_set): Call ix86_expand_avx_vzeroupper
23894 * config/i386/i386.h (struct GTY(()) machine_function): Delete
23895 has_explicit_vzeroupper.
23896 * config/i386/i386.md (enum unspec): New member
23898 (ABI_DEFAULT,ABI_VZEROUPPER,ABI_UNKNOWN): New
23899 define_constants for insn callee abi index.
23900 * config/i386/predicates.md (vzeroupper_pattern): Adjust.
23901 * config/i386/sse.md (UNSPECV_VZEROUPPER): Deleted.
23902 (avx_vzeroupper): Call ix86_expand_avx_vzeroupper.
23903 (*avx_vzeroupper): Rename to ..
23904 (avx_vzeroupper_callee_abi): .. this, and adjust pattern as
23905 call_insn which has a special vzeroupper ABI.
23906 (*avx_vzeroupper_1): Deleted.
23908 2021-06-07 liuhongt <hongtao.liu@intel.com>
23911 * df-scan.c (df_get_call_refs): When call_insn is a fake call,
23912 it won't use stack pointer reg.
23913 * final.c (leaf_function_p): When call_insn is a fake call, it
23914 won't affect caller as a leaf function.
23915 * reg-stack.c (callee_clobbers_any_stack_reg): New.
23916 (subst_stack_regs): When call_insn doesn't clobber any stack
23917 reg, don't clear the arguments.
23918 * rtl.c (shallow_copy_rtx): Don't clear flag used when orig is
23920 * shrink-wrap.c (requires_stack_frame_p): No need for stack
23921 frame for a fake call.
23922 * rtl.h (FAKE_CALL_P): New macro.
23924 2021-06-06 Eric Botcazou <ebotcazou@adacore.com>
23926 * config/sparc/sparc-protos.h (order_regs_for_local_alloc): Rename
23928 (sparc_order_regs_for_local_alloc): ...this.
23929 (sparc_leaf_reg_remap): Declare.
23930 * config/sparc/sparc.h (ADJUST_REG_ALLOC_ORDER): Adjust.
23931 (LEAF_REG_REMAP): Reimplement as call to sparc_leaf_reg_remap.
23932 * config/sparc/sparc.c (leaf_reg_remap): Delete.
23933 (order_regs_for_local_alloc): Rename to...
23934 (sparc_order_regs_for_local_alloc): ...this.
23935 (sparc_leaf_reg_remap): New function.
23936 (sparc_conditional_register_usage): Do not modify leaf_reg_remap.
23938 2021-06-06 David Edelsohn <dje.gcc@gmail.com>
23940 * config/rs6000/rs6000.c (rs6000_xcoff_asm_output_aligned_decl_common):
23941 Use assemble_name to output BSS section name.
23943 2021-06-06 Uroš Bizjak <ubizjak@gmail.com>
23945 * config/i386/constraints.md (Bs):
23946 Remove boolean operators from match_test RTX.
23949 (M): Use "mode" variable instead of GET_MODE (op) in match_test RTX.
23952 2021-06-06 Martin Liska <mliska@suse.cz>
23954 * doc/extend.texi: Add missing @headitem.
23955 * doc/invoke.texi: Likewise.
23956 * doc/objc.texi: Likewise.
23958 2021-06-06 Martin Liska <mliska@suse.cz>
23960 * genhooks.c (emit_findices): Remove unused function.
23961 (emit_documentation): Do not call emit_findices
23962 and do not search for @Fcode directives.
23964 2021-06-06 Martin Liska <mliska@suse.cz>
23966 * doc/invoke.texi: Remove extra character.
23968 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
23970 * config/sh/sh.md (doloop_end_split): Fix empty split condition.
23972 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
23974 * config/sparc/sparc.md (*snedi<W:mode>_zero_vis3,
23975 *neg_snedi<W:mode>_zero_subxc, *plus_snedi<W:mode>_zero,
23976 *plus_plus_snedi<W:mode>_zero, *minus_snedi<W:mode>_zero,
23977 *minus_minus_snedi<W:mode>_zero): Fix empty split condition.
23979 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
23981 * config/or1k/or1k.md (*movdi): Fix empty split condition.
23983 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
23985 * config/mips/mips.md (<anonymous>, bswapsi2, bswapdi2): Fix empty
23988 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
23990 * config/m68k/m68k.md (*zero_extend_inc, *zero_extend_dec,
23991 *zero_extendsidi2): Fix empty split condition.
23993 2021-06-05 Jeff Law <jeffreyalaw@gmail.com>
23995 * config/h8300/addsub.md: Fix split condition in define_insn_and_split
23997 * config/h8300/bitfield.md: Likewise.
23998 * config/h8300/combiner.md: Likewise.
23999 * config/h8300/divmod.md: Likewise.
24000 * config/h8300/extensions.md: Likewise.
24001 * config/h8300/jumpcall.md: Likewise.
24002 * config/h8300/movepush.md: Likewise.
24003 * config/h8300/multiply.md: Likewise.
24004 * config/h8300/other.md: Likewise.
24005 * config/h8300/shiftrotate.md: Likewise.
24006 * config/h8300/logical.md: Likewise. Fix split pattern to use
24007 code iterator that somehow slipped through.
24009 2021-06-04 Tobias Burnus <tobias@codesourcery.com>
24011 PR middle-end/100905
24012 * tree-nested.c (convert_nonlocal_omp_clauses,
24013 convert_local_omp_clauses): Handle OMP_CLAUSE_BIND.
24015 2021-06-04 Martin Sebor <msebor@redhat.com>
24017 PR middle-end/100732
24018 * gimple-fold.c (gimple_fold_builtin_sprintf): Avoid folding calls
24019 with either source or destination argument of invalid type.
24020 * tree-ssa-uninit.c (maybe_warn_pass_by_reference): Avoid checking
24021 calls with arguments of invalid type.
24023 2021-06-04 Martin Sebor <msebor@redhat.com>
24025 * attribs.c (init_attr_rdwr_indices): Use VLA bounds in the expected
24027 (attr_access::vla_bounds): Also handle VLA bounds.
24029 2021-06-04 Uroš Bizjak <ubizjak@gmail.com>
24031 * config/i386/predicates.md (GOT_memory_operand):
24032 Implement using match_code RTXes.
24033 (GOT32_symbol_operand): Ditto.
24035 2021-06-04 Uroš Bizjak <ubizjak@gmail.com>
24038 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
24040 (ix86_expand_vector_init_general): Ditto.
24041 Use SImode instead of word_mode for logic operations
24042 when GET_MODE_SIZE (mode) < UNITS_PER_WORD.
24043 (expand_vec_perm_even_odd_1): Assert that V2HI mode should be
24044 implemented by expand_vec_perm_1.
24045 (expand_vec_perm_broadcast_1): Assert that V2HI and V4HI modes
24046 should be implemented using standard shuffle patterns.
24047 (ix86_vectorize_vec_perm_const): Handle V2HImode. Add V4HI and
24048 V2HI modes to modes, implementable with shuffle for one operand.
24049 * config/i386/mmx.md (*punpckwd): New insn_and_split pattern.
24050 (*pshufw_1): New insn pattern.
24051 (*vec_dupv2hi): Ditto.
24052 (vec_initv2hihi): New expander.
24054 2021-06-04 Kewen Lin <linkw@linux.ibm.com>
24056 * config/arm/vfp.md (no_literal_pool_df_immediate,
24057 no_literal_pool_sf_immediate): Fix empty split condition.
24059 2021-06-04 Kewen Lin <linkw@linux.ibm.com>
24061 * config/i386/i386.md (*load_tp_x32_zext, *add_tp_x32_zext,
24062 *tls_dynamic_gnu2_combine_32): Fix empty split condition.
24063 * config/i386/sse.md (*<sse2_avx2>_pmovmskb_lt,
24064 *<sse2_avx2>_pmovmskb_zext_lt, *sse2_pmovmskb_ext_lt,
24065 *<sse4_1_avx2>_pblendvb_lt): Likewise.
24067 2021-06-04 Jakub Jelinek <jakub@redhat.com>
24070 * config/i386/i386-expand.c (ix86_expand_vector_init): Handle
24071 concatenation from half-sized modes with TImode elements.
24073 2021-06-04 Claudiu Zissulescu <claziss@synopsys.com>
24075 * config/arc/arc.c (arc_override_options): Disable millicode
24076 thunks when RF16 is on.
24078 2021-06-04 Haochen Gui <guihaoc@gcc.gnu.org>
24080 * config/rs6000/rs6000.h (PROMOTE_MODE): Remove.
24082 2021-06-04 Haochen Gui <guihaoc@gcc.gnu.org>
24084 * config/rs6000/rs6000-call.c (rs6000_promote_function_mode):
24085 Replace PROMOTE_MODE marco with its content.
24087 2021-06-03 Kewen Lin <linkw@linux.ibm.com>
24089 * config/cris/cris.md (*addi_reload): Fix empty split condition.
24091 2021-06-03 Jim Wilson <jimw@sifive.com>
24093 * config.gcc (riscv*-*-*): If --with-riscv-attribute not used,
24094 turn it on for all riscv targets.
24096 2021-06-03 Uroš Bizjak <ubizjak@gmail.com>
24099 * config/i386/i386-expand.c (ix86_expand_vector_set):
24100 Handle V2HI and V4QI modes.
24101 (ix86_expand_vector_extract): Ditto.
24102 * config/i386/mmx.md (*pinsrw): New insn pattern.
24105 (*pextrw_zext): Ditto.
24107 (*pextrb_zext): Ditto.
24108 (vec_setv2hi): New expander.
24109 (vec_extractv2hihi): Ditto.
24110 (vec_setv4qi): Ditto.
24111 (vec_extractv4qiqi): Ditto.
24112 (vec_setv8qi): Enable only for TARGET_SSE4_1.
24113 (vec_extractv8qiqi): Ditto.
24115 2021-06-03 Aaron Sawdey <acsawdey@linux.ibm.com>
24117 * config/rs6000/genfusion.pl (gen_logical_addsubf): Fix input
24118 order to subf instruction.
24119 * config/rs6000/fusion.md: Regenerate.
24121 2021-06-03 Aldy Hernandez <aldyh@redhat.com>
24123 * calls.c (get_size_range): Use range_of_expr instead of
24124 determine_value_range.
24125 * tree-affine.c (expr_to_aff_combination): Same.
24126 * tree-data-ref.c (split_constant_offset): Same.
24127 * tree-vrp.c (determine_value_range_1): Remove.
24128 (determine_value_range): Remove.
24129 * tree-vrp.h (determine_value_range): Remove.
24131 2021-06-03 Aldy Hernandez <aldyh@redhat.com>
24133 * function-tests.c (test_ranges): Call gimple_range_tests.
24134 * gimple-range-cache.cc (ranger_cache::range_of_expr): Pass stmt
24136 * gimple-range.cc (fur_source::get_operand): Do not call
24137 get_tree_range or gimple_range_global.
24139 (get_tree_range): Move to value-query.cc.
24140 Call get_arith_expr_range.
24141 (gimple_ranger::range_of_expr): Add argument to get_tree_range.
24142 Include gimple-range-tests.cc.
24143 * gimple-range.h (fold_range): Add argument.
24144 (get_tree_range): Remove.
24145 * selftest.h (gimple_range_tests): New.
24146 * value-query.cc (global_range_query::range_of_expr): Add
24148 (range_query::get_tree_range): Move from gimple-range.cc.
24149 * value-query.h (class range_query): Add get_tree_range and
24150 get_arith_expr_range. Make fur_source a friend.
24151 * vr-values.c (vr_values::range_of_expr): Pass stmt to
24153 * gimple-range-tests.cc: New file.
24155 2021-06-03 Aldy Hernandez <aldyh@redhat.com>
24157 * gimple-range.cc (gimple_ranger::export_global_ranges): Call
24158 update_global_range.
24159 * value-query.cc (update_global_range): New.
24160 * value-query.h (update_global_range): New.
24162 2021-06-03 David Malcolm <dmalcolm@redhat.com>
24164 * diagnostic-show-locus.c (diagnostic_show_locus): Don't reject
24165 printing the same location twice if there are fix-it hints,
24166 multiple locations, or a label.
24168 2021-06-03 Andre Vieira <andre.simoesdiasvieira@arm.com>
24170 * tree-vect-loop.c (vect_transform_loop): Use main loop's various'
24171 thresholds to narrow the upper bound on epilogue iterations.
24173 2021-06-03 Christophe Lyon <christophe.lyon@linaro.org>
24175 * config/arm/mve.md (mve_vabsq_f<mode>): Use 'abs' instead of unspec.
24176 (mve_vabsq_s<mode>): Likewise.
24177 * config/arm/neon.md (abs<mode>2): Rename to neon_abs<mode>2.
24178 * config/arm/unspecs.md (VABSQ_F, VABSQ_S): Delete.
24179 * config/arm/vec-common.md (neg<mode>2): Rename to
24180 <absneg_str><mode>2.
24182 2021-06-03 Claudiu Zissulescu <claziss@synopsys.com>
24184 * common/config/arc/arc-common.c (arc_option_optimization_table):
24185 Remove malign-call.
24186 * config/arc/arc.c (arc_unalign_branch_p): Remove unused function.
24187 * config/arc/arc.h (TARGET_MIXED_CODE): Remove macro.
24188 (INDEX_REG_CLASS): Only refer to GENERAL_REGS.
24189 * config/arc/arc.md (abssi2_mixed): Remove pattern.
24190 * config/arc/arc.opt (munalign-prob-threshold): Mark it obsolete.
24191 (malign-call): Likewise.
24192 (mmixed-code): Likewise.
24193 * doc/invoke.texi (ARC): Update doc.
24195 2021-06-03 Martin Liska <mliska@suse.cz>
24197 * common.opt: Use proper Enum values.
24198 * opts.c (COVERAGE_SANITIZER_OPT): Remove.
24199 (parse_sanitizer_options): Handle only sanitizer_opts.
24200 (common_handle_option): Just assign value.
24202 2021-06-03 Eric Botcazou <ebotcazou@adacore.com>
24205 * tree-inline.c (inline_forbidden_p): Remove test on return type.
24207 2021-06-03 Eric Botcazou <ebotcazou@adacore.com>
24209 * dwarf2out.c (loc_list_from_tree_1) <FUNCTION_DECL>: Also generate
24210 DW_OP_GNU_variable_value referencing an existing DIE at file scope.
24211 (type_byte_size): Inline into...
24212 (add_byte_size_attribute): ...this and call add_scalar_info.
24214 2021-06-03 Eric Botcazou <ebotcazou@adacore.com>
24216 * dwarf2out.c (mem_loc_descriptor) <UDIV>: Fix typo.
24217 (typed_binop_from_tree): New function.
24218 (loc_list_from_tree_1) <EXACT_DIV_EXPR>: For an unsigned type,
24219 turn a divide by a power of 2 into a shift.
24220 <CEIL_DIV_EXPR>: For an unsigned type, use a signed divide if the
24221 size of the mode is lower than DWARF2_ADDR_SIZE; otherwise, do a
24222 typed divide by calling typed_binop_from_tree.
24224 2021-06-03 Eric Botcazou <ebotcazou@adacore.com>
24226 * dwarf2out.c (scompare_loc_descriptor): Fix head comment.
24227 (is_handled_procedure_type): Likewise.
24228 (struct loc_descr_context): Add strict_signedness field.
24229 (resolve_args_picking_1): Deal with DW_OP_[GNU_]deref_type,
24230 DW_OP_[GNU_]convert and DW_OP_[GNU_]reinterpret.
24231 (resolve_args_picking): Minor tweak.
24232 (function_to_dwarf_procedure): Initialize strict_signedness field.
24233 (type_byte_size): Likewise.
24234 (field_byte_offset): Likewise.
24235 (gen_descr_array_type_die): Likewise.
24236 (gen_variant_part): Likewise.
24237 (loc_list_from_tree_1) <CALL_EXPR>: Tidy up and set strict_signedness
24238 to true when a context is present before evaluating the arguments.
24239 <COND_EXPR>: Do not generate a useless comparison with zero.
24240 When dereferencing an address, if strict_signedness is true and the
24241 type is small and signed, use DW_OP_deref_type to do the dereference
24242 and then DW_OP_convert to convert back to the generic type.
24244 2021-06-03 Jakub Jelinek <jakub@redhat.com>
24247 * tree-inline.c (copy_tree_body_r): Handle iterators on
24248 OMP_CLAUSE_AFFINITY or OMP_CLAUSE_DEPEND.
24250 2021-06-03 Kewen Lin <linkw@linux.ibm.com>
24252 * config/arc/arc.md (*bbit_di): Remove.
24254 2021-06-02 Christoph Muellner <cmuellner@gcc.gnu.org>
24256 PR rtl-optimization/100264
24257 * ree.c (get_sub_rtx): Ignore SET expressions without register
24258 destinations and remove assertion, as it is not valid anymore
24259 with this new behaviour.
24260 (merge_def_and_ext): Eliminate destination check for register
24261 as such SET expressions can't occur anymore.
24262 (combine_reaching_defs): Likewise.
24264 2021-06-02 Jakub Jelinek <jakub@redhat.com>
24267 * config/xtensa/xtensa.h (LEAF_REG_REMAP): Cast REGNO to int to avoid
24268 -Wtype-limits warnings.
24269 (DWARF_FRAME_REGISTER): Rewrite into ternary operator with addition
24270 in operands to avoid -Wsign-compare warnings.
24272 2021-06-02 Pat Haugen <pthaugen@linux.ibm.com>
24274 * config/rs6000/rs6000-logue.c (rs6000_emit_prologue): Use
24277 2021-06-02 Vineet Gupta <vgupta@synopsys.com>
24279 * config/arc/arc.h (TARGET_CPU_DEFAULT): Change to hs38_linux.
24281 2021-06-02 Ilya Leoshkevich <iii@linux.ibm.com>
24283 * config/s390/s390.md(*ashrdi3_31<setcc><cconly>): Use a single
24285 * config/s390/subst.md(cconly_subst): Use a single constraint
24286 in (match_scratch).
24288 2021-06-02 Martin Liska <mliska@suse.cz>
24290 * ipa-icf.h: Use auto_vec for memory_access_types.
24292 2021-06-02 Jeff Law <jeffreyalaw@gmail.com>
24294 * config/h8300/h8300-protos.h (compute_a_shift_length): Drop unused
24295 argument from prototype.
24296 (output_logical_op): Add rtx_code argument.
24297 (compute_logical_op_length): Likewise.
24298 * config/h8300/h8300.c (h8300_and_costs): Pass additional argument
24299 to compute_a_shift_length.
24300 (output_logical_op); New argument with the rtx code rather than
24301 extracting it from an operand. Handle QImode too.
24302 (compute_logical_op_length): Similary.
24303 (compute_a_shift_length): Drop unused argument.
24304 * config/h8300/h8300.md (logicals): New code iterator.
24305 * config/h8300/logical.md (<code><mode>3 expander): Combine
24306 the "and" expander with the "ior"/"xor" expander.
24307 (bclr<mode>msx): Combine the QI/HI mode patterns.
24308 (<logical><mode>3 insns): Use code iterator rather than match_operator.
24309 Handle QImode as well. Update call to output_logical_op and
24310 compute_logical_op_length to pass in rtx_code
24311 Fix split condition on all define_insn_and_split patterns.
24312 (one_cmpl<mode>2<cczn>): Use <cczn> to support both clobbering
24313 the flags and setting ZN via existing define_subst.
24314 * config/h8300/shiftrotate.md: Drop unused argument from
24315 calls to compute_a_shift_length.
24316 Signed-off-by: Jeff Law <jeffreyalaw@gmail.com>
24318 2021-06-01 Andrew Pinski <apinski@marvell.com>
24320 PR tree-optimization/25290
24321 * tree-ssa-phiopt.c (match_simplify_replacement):
24323 (tree_ssa_phiopt_worker): Use match_simplify_replacement.
24324 (two_value_replacement): Change the comment about
24325 conditional_replacement.
24326 (conditional_replacement): Delete.
24328 2021-06-01 Andrew Pinski <apinski@marvell.com>
24330 PR tree-optimization/95481
24331 * tree-tailcall.c (find_tail_calls): Handle empty typed
24334 2021-06-01 Andrew Pinski <apinski@marvell.com>
24336 * gimplify.c (zero_sized_field_decl): Delete
24337 (zero_sized_type): Delete
24338 (gimplify_init_ctor_eval): Use is_empty_type instead
24339 of zero_sized_field_decl.
24340 (gimplify_modify_expr): Use is_empty_type instead of
24343 2021-06-01 Jason Merrill <jason@redhat.com>
24346 * tree.h (CALL_FROM_NEW_OR_DELETE_P): Adjust comment.
24348 2021-06-01 Jason Merrill <jason@redhat.com>
24351 * diagnostic.h (warning_enabled_at): Declare.
24352 * diagnostic.c (diagnostic_enabled): Factor out from...
24353 (diagnostic_report_diagnostic): ...here.
24354 (warning_enabled_at): New.
24356 2021-06-01 Aldy Hernandez <aldyh@redhat.com>
24358 * gimple-ssa-evrp.c: Enable exporting of global ranges.
24360 2021-06-01 Martin Liska <mliska@suse.cz>
24363 * doc/invoke.texi: Mention that -fgcse-after-reload
24364 is enabled with -O3.
24366 2021-06-01 liuhongt <hongtao.liu@intel.com>
24368 PR tree-optimization/98365
24369 * tree-if-conv.c (strip_nop_cond_scalar_reduction): New function.
24370 (is_cond_scalar_reduction): Handle nop_expr in cond scalar reduction.
24371 (convert_scalar_cond_reduction): Ditto.
24372 (predicate_scalar_phi): Ditto.
24374 2021-06-01 Andrew MacLeod <amacleod@redhat.com>
24376 PR tree-optimization/100781
24377 * gimple-range-cache.cc (ranger_cache::ranger_cache): Enable new
24378 value calculation by default.
24379 (ranger_cache::enable_new_values): New.
24380 (ranger_cache::disable_new_values): New.
24381 (ranger_cache::push_poor_value): Check if new values are allowed.
24382 * gimple-range-cache.h (class ranger_cache): New member/methods.
24383 * gimple-range.cc (gimple_ranger::range_of_expr): Check for debug
24384 statement, and disable/renable new value calculation.
24386 2021-06-01 Andrew MacLeod <amacleod@redhat.com>
24388 * gimple-range-cache.cc (ranger_cache::ssa_range_in_bb): Delete.
24389 (ranger_cache::range_of_def): New.
24390 (ranger_cache::entry_range): New.
24391 (ranger_cache::exit_range): New.
24392 (ranger_cache::range_of_expr): Adjust.
24393 (ranger_cache::range_on_edge): Adjust.
24394 (ranger_cache::propagate_cache): Call exit_range directly.
24395 * gimple-range-cache.h (class ranger_cache): Adjust.
24397 2021-06-01 Andrew MacLeod <amacleod@redhat.com>
24399 * gimple-range-cache.cc (ranger_cache::ranger_cache): Adjust for
24400 gori_compute being a member rather than base class.
24401 dervied call to member call.
24402 (ranger_cache::dump): No longer dump gori_map.
24403 (ranger_cache::dump_bb): New.
24404 (ranger_cache::get_non_stale_global_range): Adjust for gori_compute
24405 being a member rather than base class.
24406 (ranger_cache::set_global_range): Ditto.
24407 (ranger_cache::ssa_range_in_bb): Ditto.
24408 (ranger_cache::range_of_expr): New.
24409 (ranger_cache::range_on_edge): New.
24410 (ranger_cache::block_range): Adjust for gori_computes. Debug changes.
24411 (ranger_cache::propagate_cache): Adjust debugging output.
24412 (ranger_cache::fill_block_cache): Adjust for gori_computes. Debug
24414 * gimple-range-cache.h (class ranger_cache): Make gori_compute a
24415 member, and inherit from range_query instead.
24416 (ranger_cache::dump_bb): New. split from dump.
24417 * gimple-range-gori.cc (gori_compute::ssa_range_in_bb): Delete.
24418 (gori_compute::expr_range_at_stmt): Delete.
24419 (gori_compute::compute_name_range_op): Delete.
24420 (gori_compute::compute_operand_range_switch): Add fur_source.
24421 (gori_compute::compute_operand_range): Add fur_source param, inline
24422 old compute_name_range_op and optimize_logical_operands.
24423 (struct tf_range): Delete.
24424 (gori_compute::logical_combine): Adjust
24425 (gori_compute::optimize_logical_operands): Delete.
24426 (gori_compute::compute_logical_operands_in_chain): Delete.
24427 (gori_compute::compute_logical_operands): Adjust.
24428 (gori_compute::compute_operand1_range): Adjust to fur_source.
24429 (gori_compute::compute_operand2_range): Ditto.
24430 (gori_compute::compute_operand1_and_operand2_range): Ditto.
24431 (gori_compute::outgoing_edge_range_p): Add range_query parameter,
24432 and adjust to fur_source.
24433 * gimple-range-gori.h (class gori_compute): Simplify and adjust to
24434 range_query and fur_source.
24435 * gimple-range.cc (gimple_ranger::range_on_edge): Query range_on_edge
24436 from the ranger_cache..
24437 (gimple_ranger::fold_range_internal): Adjust to base class change of
24439 (gimple_ranger::dump_bb): Adjust dump.
24440 * gimple-range.h (gimple_ranger):export gori computes object.
24442 2021-06-01 Andrew MacLeod <amacleod@redhat.com>
24444 PR tree-optimization/100774
24445 * gimple-range-cache.cc (ranger_cache::get_non_stale_global_range):
24446 Constant values are also not stale.
24447 (ranger_cache::set_global_range): Range invariant values should also
24448 have the correct timestamp.
24450 2021-05-31 Martin Liska <mliska@suse.cz>
24452 * tree-streamer-in.c (unpack_ts_function_decl_value_fields):
24453 Unpack FUNCTION_DECL_DECL_TYPE.
24454 * tree-streamer-out.c (pack_ts_function_decl_value_fields):
24455 Stream FUNCTION_DECL_DECL_TYPE instead of
24456 DECL_IS_OPERATOR_NEW_P.
24457 * tree.h (set_function_decl_type): Use FUNCTION_DECL_DECL_TYPE
24459 (DECL_IS_OPERATOR_NEW_P): Likewise.
24460 (DECL_IS_OPERATOR_DELETE_P): Likewise.
24461 (DECL_LAMBDA_FUNCTION_P): Likewise.
24463 2021-05-31 Richard Biener <rguenther@suse.de>
24466 * internal-fn.c (expand_SHUFFLEVECTOR): Define.
24467 * internal-fn.def (SHUFFLEVECTOR): New.
24468 * internal-fn.h (expand_SHUFFLEVECTOR): Declare.
24469 * doc/extend.texi: Document __builtin_shufflevector.
24471 2021-05-31 Peter Bergner <bergner@linux.ibm.com>
24474 * config/rs6000/predicates.md(mma_assemble_input_operand): Allow
24475 indexed form addresses.
24477 2021-05-29 Jeff Law <jlaw@tachyum.com>
24479 * config/h8300/h8300.c (h8300_emit_stack_adjustment): Drop unused
24480 parameter. Call callers fixed.
24482 (output_plussi): Add FALLTHRU markers.
24483 (h8300_shift_needs_scratch_p): Add gcc_unreachable marker.
24485 2021-05-29 Jakub Jelinek <jakub@redhat.com>
24487 PR middle-end/99928
24488 * gimplify.c (gimplify_scan_omp_clauses): For taskloop simd
24489 combined with parallel, make sure to add shared clause to
24490 parallel for explicit linear clause.
24492 2021-05-29 Aldy Hernandez <aldyh@redhat.com>
24494 PR tree-optimization/100787
24495 * gimple-ssa-evrp.c: Disable exporting of global ranges.
24497 2021-05-28 Jason Merrill <jason@redhat.com>
24499 * tree-iterator.h (struct tree_stmt_iterator): Add operator++,
24500 operator--, operator*, operator==, and operator!=.
24501 (class tsi_range): New.
24503 2021-05-28 Richard Biener <rguenther@suse.de>
24505 PR tree-optimization/100778
24506 * tree-vect-slp.c (vect_build_slp_tree_1): Prevent possibly
24507 trapping ops in different BBs.
24509 2021-05-28 Richard Biener <rguenther@suse.de>
24512 * tree-inline.c (copy_bb): When processing __builtin_va_arg_pack
24513 copy fntype from original call.
24515 2021-05-28 Martin Liska <mliska@suse.cz>
24517 PR gcov-profile/100751
24518 * doc/gcov.texi: Revert partially a hunk that was wrong.
24520 2021-05-28 Cooper Qu <cooper.qu@linux.alibaba.com>
24522 * config/csky/csky-linux-elf.h (HAVE_sync_compare_and_swapqi):
24524 (HAVE_sync_compare_and_swaphi): Likewise.
24525 (HAVE_sync_compare_and_swapsi): Likewise.
24527 2021-05-28 Jakub Jelinek <jakub@redhat.com>
24529 PR middle-end/99928
24530 * tree.h (OMP_CLAUSE_MAP_IMPLICIT): Define.
24532 2021-05-28 Tobias Burnus <tobias@codesourcery.com>
24534 * gimplify.c (gimplify_omp_affinity): New.
24535 (gimplify_scan_omp_clauses): Call it; remove affinity clause afterwards.
24536 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_AFFINITY.
24537 * tree-pretty-print.c (dump_omp_clause): Handle OMP_CLAUSE_AFFINITY.
24538 * tree.c (omp_clause_num_ops, omp_clause_code_name): Add clause.
24539 (walk_tree_1): Handle OMP_CLAUSE_AFFINITY.
24541 2021-05-28 Joern Rennecke <joern.rennecke@riscy-ip.com>
24542 Richard Biener <rguenther@suse.de>
24544 * match.pd <popcount & / + pattern matching>:
24545 When generating popcount directly fails, try doing it in two halves.
24547 2021-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
24549 * Makefile.in (generated_files): Add gimple-match.c and
24552 2021-05-28 Joern Rennecke <joern.rennecke@embecosm.com>
24554 * gensupport.c (alter_predicate_for_insn): Handle MATCH_DUP.
24556 2021-05-28 Joern Rennecke <joern.rennecke@embecosm.com>
24558 * gensupport.c (alter_constraints): Add MATCH_SCRATCH case.
24560 2021-05-28 Kewen Lin <linkw@linux.ibm.com>
24562 PR tree-optimization/99398
24563 * tree-ssa-forwprop.c (simplify_permutation): Optimize some cases
24564 where the fed operands are CTOR/CST and propagated through
24565 VIEW_CONVERT_EXPR. Call vec_perm_indices::new_shrunk_vector.
24566 * vec-perm-indices.c (vec_perm_indices::new_shrunk_vector): New
24568 * vec-perm-indices.h (vec_perm_indices::new_shrunk_vector): New
24571 2021-05-27 Uroš Bizjak <ubizjak@gmail.com>
24573 * config/i386/mmx.md (addv2sf3): Do not call
24574 ix86_fixup_binary_operands_no_copy.
24577 (<smaxmin:code>v2sf3): Ditto.
24578 (<plusminus:insn><MMXMODEI:mode>3): Ditto.
24579 (<plusminus:insn><VI_32:mode>3): Remove expander.
24580 (<plusminus:insn><VI_32:mode>3): Rename from
24581 "*<plusminus:insn><VI_32:mode>3".
24582 (mulv4hi): Do not call ix86_fixup_binary_operands_no_copy.
24583 (mulv2hi3): Remove expander.
24584 (mulv2hi3): Rename from *mulv2hi3.
24585 (<s>mulv2hi3_highpart): Remove expander.
24586 (<s>mulv2hi3_highpart): Rename from *<s>mulv2hi3_highpart.
24587 (<smaxmin:code><MMXMODE14:mode>3): Rename from
24588 "*mmx_<smaxmin:code><MMXMODE14:mode>3".
24589 (<smaxmin:code><SMAXMIN_MMXMODEI:mode>3): Remove expander.
24590 (SMAXMIN_MMXMODEI): Remove mode iterator.
24591 (<smaxmin:code>v4hi3): New expander.
24592 (<smaxmin:code>v4qi3): Rename from *<smaxmin:code>v4qi3.
24593 (<smaxmin:code>v2hi3): Rename from *<smaxmin:code>v2hi3.
24594 (<smaxmin:code><SMAXMIN_VI_32:mode>3): Remove expander.
24595 (SMAXMIN_VI_32): Remove mode iterator.
24596 (<umaxmin:code><MMXMODE24:mode>3): Rename from
24597 "*mmx_<umaxmin:code><MMXMODE24:mode>3".
24598 (<umaxmin:code><UMAXMIN_MMXMODEI:mode>3): Remove expander.
24599 (UMAXMIN_MMXMODEI): Remove mode iterator.
24600 (<umaxmin:code>v8qi3): New expander.
24601 (<umaxmin:code>v4qi3): Rename from *<umaxmin:code>v4qi3.
24602 (<umaxmin:code>v2hi3): Rename from *<umaxmin:code>v2hi3.
24603 (<umaxmin:code><SMAXMIN_VI_32:mode>3): Remove expander.
24604 (UMAXMIN_VI_32): Remove mode iterator.
24605 (<any_shift:insn>v2hi3): Remove expander.
24606 (<any_shift:insn>v2hi3): Rename from *<any_shift:insn>v2hi3.
24607 (<any_logic:code><MMXMODEI:mode>3): Do not call
24608 ix86_fixup_binary_operands_no_copy.
24609 (<any_logic:code><VI_32:mode>3): Remove expander.
24610 (<any_logic:code><VI_32:mode>3): Rename from
24611 "*<any_logic:code><VI_32:mode>3".
24612 (uavg<mode>3_ceil): Do not call ix86_fixup_binary_operands_no_copy.
24613 * config/i386/sse.md (div<VF2:mode>3): Do not call
24614 ix86_fixup_binary_operands_no_copy.
24615 (div<VF1:mode>3): Ditto.
24616 (<maxmin:code><VI8_AVX2_AVX512F:mode>3): Ditto.
24617 (smulhrsv4hi3): Ditto.
24618 (smulhrsv2hi3): Ditto.
24620 2021-05-27 Martin Sebor <msebor@redhat.com>
24622 * ggc.h (gt_ggc_mx): Add overloads for all integers.
24624 * hash-map.h (class hash_map): Add pch_nx_helper overloads for all
24626 (hash_map::operator==): New function.
24628 2021-05-27 Uroš Bizjak <ubizjak@gmail.com>
24631 * config/i386/i386-expand.c (ix86_expand_int_sse_cmp):
24632 For TARGET_XOP bypass SSE comparisons for all supported vector modes.
24633 * config/i386/mmx.md (*xop_maskcmp<MMXMODEI:mode>3): New insn pattern.
24634 (*xop_maskcmp<VI_32:mode>3): Ditto.
24635 (*xop_maskcmp_uns<MMXMODEI:mode>3): Ditto.
24636 (*xop_maskcmp_uns<VI_32:mode>3): Ditto.
24638 2021-05-27 Richard Earnshaw <rearnsha@arm.com>
24641 * config/arm/arm.c (arm_configure_build_target): Remove parameter
24642 opts_set, directly check opts parameters for being non-null.
24643 (arm_option_restore): Update call to arm_configure_build_target.
24644 (arm_option_override): Likewise.
24645 (arm_can_inline_p): Likewise.
24646 (arm_valid_target_attribute_tree): Likewise.
24647 * config/arm/arm-c.c (arm_pragma_target_parse): Likewise.
24648 * config/arm/arm-protos.h (arm_configure_build_target): Adjust
24651 2021-05-27 Aldy Hernandez <aldyh@redhat.com>
24653 * vr-values.c (simplify_conversion_using_ranges): Use
24654 get_range_query instead of get_global_range_query.
24656 2021-05-27 Aldy Hernandez <aldyh@redhat.com>
24658 * gimple-range.cc (get_range_global): Move to value-query.cc.
24659 (gimple_range_global): Same.
24660 (get_global_range_query): Same.
24661 (global_range_query::range_of_expr): Same.
24662 * gimple-range.h (class global_range_query): Move to
24664 (gimple_range_global): Same.
24665 * tree-ssanames.c (get_range_info): Move to value-query.cc.
24666 (get_ptr_nonnull): Same.
24667 * tree-ssanames.h (get_range_info): Remove.
24668 (get_ptr_nonnull): Remove.
24669 * value-query.cc (get_ssa_name_range_info): Move from
24671 (get_ssa_name_ptr_info_nonnull): Same.
24672 (get_range_global): Move from gimple-range.cc.
24673 (gimple_range_global): Same.
24674 (get_global_range_query): Same.
24675 (global_range_query::range_of_expr): Same.
24676 * value-query.h (class global_range_query): Move from
24678 (gimple_range_global): Same.
24680 2021-05-27 Uroš Bizjak <ubizjak@gmail.com>
24683 * config/i386/mmx.md (uavgv4qi3_ceil): New insn pattern.
24684 (uavgv2hi3_ceil): Ditto.
24686 2021-05-26 Eric Botcazou <ebotcazou@adacore.com>
24689 * doc/extend.texi (scalar_storage_order): Rephrase slightly.
24691 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
24693 * tree-ssanames.c (get_range_info): Merge both copies of
24694 get_range_info into one that works with irange.
24695 * tree-ssanames.h (get_range_info): Remove version that works on
24698 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
24700 * builtins.c (check_nul_terminated_array): Convert to get_range_query.
24701 (expand_builtin_strnlen): Same.
24702 (determine_block_size): Same.
24703 * fold-const.c (expr_not_equal_to): Same.
24704 * gimple-fold.c (size_must_be_zero_p): Same.
24705 * gimple-match-head.c: Include gimple-range.h.
24706 * gimple-pretty-print.c (dump_ssaname_info): Convert to get_range_query.
24707 * gimple-ssa-warn-restrict.c
24708 (builtin_memref::extend_offset_range): Same.
24709 * graphite-sese-to-poly.c (add_param_constraints): Same.
24710 * internal-fn.c (get_min_precision): Same.
24711 * ipa-fnsummary.c (set_switch_stmt_execution_predicate): Same.
24712 * ipa-prop.c (ipa_compute_jump_functions_for_edge): Same.
24714 * tree-data-ref.c (split_constant_offset): Same.
24715 (dr_step_indicator): Same.
24716 * tree-dfa.c (get_ref_base_and_extent): Same.
24717 * tree-scalar-evolution.c (iv_can_overflow_p): Same.
24718 * tree-ssa-loop-niter.c (refine_value_range_using_guard): Same.
24719 (determine_value_range): Same.
24720 (record_nonwrapping_iv): Same.
24721 (infer_loop_bounds_from_signedness): Same.
24722 (scev_var_range_cant_overflow): Same.
24723 * tree-ssa-phiopt.c (two_value_replacement): Same.
24724 * tree-ssa-pre.c (insert_into_preds_of_block): Same.
24725 * tree-ssa-reassoc.c (optimize_range_tests_to_bit_test): Same.
24726 * tree-ssa-strlen.c (handle_builtin_stxncpy_strncat): Same.
24728 (dump_strlen_info): Same.
24729 (set_strlen_range): Same.
24730 (maybe_diag_stxncpy_trunc): Same.
24731 (get_len_or_size): Same.
24732 (handle_integral_assign): Same.
24733 * tree-ssa-structalias.c (find_what_p_points_to): Same.
24734 * tree-ssa-uninit.c (find_var_cmp_const): Same.
24735 * tree-switch-conversion.c (bit_test_cluster::emit): Same.
24736 * tree-vect-patterns.c (vect_get_range_info): Same.
24737 (vect_recog_divmod_pattern): Same.
24738 * tree-vrp.c (intersect_range_with_nonzero_bits): Same.
24739 (register_edge_assert_for_2): Same.
24740 (determine_value_range_1): Same.
24741 * tree.c (get_range_pos_neg): Same.
24742 * vr-values.c (vr_values::get_lattice_entry): Same.
24743 (vr_values::update_value_range): Same.
24744 (simplify_conversion_using_ranges): Same.
24746 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
24748 * gimple-ssa-warn-alloca.c (alloca_call_type): Use
24749 get_range_query instead of query argument.
24750 (pass_walloca::execute): Enable and disable global ranger.
24752 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
24754 * gimple-ssa-evrp.c (rvrp_folder::rvrp_folder): Call
24756 (rvrp_folder::~rvrp_folder): Call disable_ranger.
24757 (hybrid_folder::hybrid_folder): Call enable_ranger.
24758 (hybrid_folder::~hybrid_folder): Call disable_ranger.
24760 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
24762 * function.c (allocate_struct_function): Set cfun->x_range_query.
24763 * function.h (struct function): Declare x_range_query.
24764 (get_range_query): New.
24765 (get_global_range_query): New.
24766 * gimple-range-cache.cc (ssa_global_cache::ssa_global_cache):
24767 Remove call to safe_grow_cleared.
24768 * gimple-range.cc (get_range_global): New.
24769 (gimple_range_global): Move from gimple-range.h.
24770 (get_global_range_query): New.
24771 (global_range_query::range_of_expr): New.
24772 (enable_ranger): New.
24773 (disable_ranger): New.
24774 * gimple-range.h (gimple_range_global): Move to gimple-range.cc.
24775 (class global_range_query): New.
24776 (enable_ranger): New.
24777 (disable_ranger): New.
24778 * gimple-ssa-evrp.c (evrp_folder::~evrp_folder): Rename
24779 dump_all_value_ranges to dump.
24780 * tree-vrp.c (vrp_prop::finalize): Same.
24781 * value-query.cc (range_query::dump): New.
24782 * value-query.h (range_query::dump): New.
24783 * vr-values.c (vr_values::dump_all_value_ranges): Rename to...
24784 (vr_values::dump): ...this.
24785 * vr-values.h (class vr_values): Rename dump_all_value_ranges to
24786 dump and make virtual.
24788 2021-05-26 Uroš Bizjak <ubizjak@gmail.com>
24790 * config/i386/i386.c (ix86_autovectorize_vector_modes):
24791 Add V4QImode and V16QImode for TARGET_SSE2.
24792 * doc/sourcebuild.texi (Vector-specific attributes):
24793 Add vect64 and vect32 description.
24795 2021-05-26 Bernd Edlinger <bernd.edlinger@hotmail.de>
24797 * gimple-range-gori.cc (range_def_chain::register_dependency):
24798 Resize m_def_chain when needed.
24800 2021-05-26 Christophe Lyon <christophe.lyon@linaro.org>
24802 * config/arm/mve.md (mve_vaddvq_<supf><mode>): Prefix with '@'.
24803 * config/arm/neon.md (reduc_plus_scal_<mode>): Move to ..
24804 * config/arm/vec-common.md: .. here. Add support for MVE.
24806 2021-05-26 Jakub Jelinek <jakub@redhat.com>
24808 * config/epiphany/epiphany.c (epiphany_print_operand_address): Remove
24810 * config/microblaze/microblaze.c (microblaze_legitimize_address,
24812 microblaze_option_override, print_operand): Likewise.
24813 * config/microblaze/microblaze.md (call_internal_plt,
24814 call_value_intern_plt, call_value_intern): Likewise.
24815 * config/arm/aout.h (ASM_OUTPUT_ALIGN): Likewise.
24816 * config/iq2000/iq2000.md (call_internal1, call_value_internal1,
24817 call_value_multiple_internal1): Likewise.
24818 * config/bfin/bfin.c (symbolic_reference_mentioned_p): Likewise.
24820 2021-05-26 Jan-Benedict Glaw <jbglaw@lug-owl.de>
24822 * config/arc/arc.c (arc_address_cost, arc_print_operand_address,
24823 arc_ccfsm_advance, symbolic_reference_mentioned_p,
24824 arc_raw_symbolic_reference_mentioned_p): Remove register
24827 2021-05-26 Jakub Jelinek <jakub@redhat.com>
24830 * omp-low.c: Include omp-offload.h.
24831 (create_omp_child_function): If current_function_decl has
24832 "omp declare target" attribute and is_gimple_omp_offloaded,
24833 remove that attribute from the copy of attribute list and
24834 add "omp target entrypoint" attribute instead.
24835 (lower_omp_target): Mark .omp_data_sizes.* and .omp_data_kinds.*
24836 variables for offloading if in omp_maybe_offloaded_ctx.
24837 * omp-offload.c (pass_omp_target_link::execute): Nullify second
24838 argument to GOMP_target_data_ext in offloaded code.
24840 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
24842 * config/csky/csky.c (csky_can_change_mode_class): Delete.
24843 For csky, HF/SF mode use the low bits of VREGS.
24845 2021-05-26 Eric Botcazou <ebotcazou@adacore.com>
24847 * gimplify.c (gimplify_decl_expr): Do not clear TREE_READONLY on a
24848 DECL which is a reference for OMP.
24850 2021-05-26 Martin Liska <mliska@suse.cz>
24852 PR gcov-profile/100751
24853 * doc/gcov.texi: Document that __gcov_dump can be called just
24854 once and that __gcov_reset resets run-time counters.
24856 2021-05-26 Martin Liska <mliska@suse.cz>
24858 * doc/install.texi: Port relevant part from install-old.texi
24859 and re-generate list of CPUs and systems.
24861 2021-05-26 Martin Liska <mliska@suse.cz>
24863 * Makefile.in: Remove it.
24864 * doc/include/fdl.texi: Update next/previous chapters.
24865 * doc/install.texi: Likewise.
24866 * doc/install-old.texi: Removed.
24868 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
24870 * config/csky/csky.c (ck810_legitimate_index_p): Support
24871 "base + index" with DF mode.
24872 * config/csky/constraints.md ("Y"): New constraint for memory operands
24873 without index register.
24874 * config/csky/csky_insn_fpuv2.md (fpuv3_movdf): Use "Y" instead of "m"
24875 when mov between memory and general registers, and lower their priority.
24876 * config/csky/csky_insn_fpuv3.md (fpuv2_movdf): Likewise.
24878 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
24880 * config/csky/csky.c (TARGET_PROMOTE_PROTOTYPES): Delete.
24882 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
24884 * config/csky/csky.md (untyped_call): Emit clobber for return
24885 registers to mark them used.
24887 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
24889 * config/csky/csky.md (cskyv2_sextend_ldbs): New.
24891 2021-05-26 Andrew Pinski <apinski@marvell.com>
24893 * match.pd (x < 0 ? ~y : y): New patterns.
24895 2021-05-26 Andrew Pinski <apinski@marvell.com>
24897 * match.pd (A?CST1:CST2): Add simplifcations for A?0:+-1, A?+-1:0,
24898 A?POW2:0 and A?0:POW2.
24900 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
24902 * gimple-range-gori.cc (class logical_stmt_cache): Delete
24903 (logical_stmt_cache::logical_stmt_cache ): Delete.
24904 (logical_stmt_cache::~logical_stmt_cache): Delete.
24905 (logical_stmt_cache::cache_entry::dump): Delete.
24906 (logical_stmt_cache::get_range): Delete.
24907 (logical_stmt_cache::cached_name ): Delete.
24908 (logical_stmt_cache::same_cached_name): Delete.
24909 (logical_stmt_cache::cacheable_p): Delete.
24910 (logical_stmt_cache::slot_diagnostics ): Delete.
24911 (logical_stmt_cache::dump): Delete.
24912 (gori_compute_cache::gori_compute_cache): Delete.
24913 (gori_compute_cache::~gori_compute_cache): Delete.
24914 (gori_compute_cache::compute_operand_range): Delete.
24915 (gori_compute_cache::cache_stmt): Delete.
24916 * gimple-range-gori.h (gori_compute::compute_operand_range): Remove
24918 (class gori_compute_cache): Delete.
24920 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
24922 * gimple-range.cc (fold_using_range::range_of_range_op): Use m_gori
24924 (fold_using_range::range_of_address): Adjust.
24925 (fold_using_range::range_of_phi): Adjust.
24926 * gimple-range.h (class fur_source): Adjust.
24927 (fur_source::fur_source): Adjust.
24929 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
24931 * gimple-range-gori.cc (gori_compute::expr_range_at_stmt): Rename
24932 from expr_range_in_bb and adjust.
24933 (gori_compute::compute_name_range_op): Adjust.
24934 (gori_compute::optimize_logical_operands): Adjust.
24935 (gori_compute::compute_logical_operands_in_chain): Adjust.
24936 (gori_compute::compute_operand1_range): Adjust.
24937 (gori_compute::compute_operand2_range): Adjust.
24938 (ori_compute_cache::cache_stmt): Adjust.
24939 * gimple-range-gori.h (gori_compute): Rename prototype.
24941 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
24943 * gimple-range.cc (gimple_ranger::range_of_expr): Non-null should be
24944 checked only after range_of_stmt, not range_on_entry.
24945 (gimple_ranger::range_on_entry): Check for non-null in any
24946 predecessor block, if it is not already non-null.
24947 (gimple_ranger::range_on_exit): DOnt check for non-null after
24948 range on entry call.
24949 (gimple_ranger::dump_bb): New. Split from dump.
24950 (gimple_ranger::dump): Adjust.
24951 * gimple-range.h (class gimple_ranger): Adjust.
24953 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
24955 * gimple-range-cache.cc (struct range_timestamp): Delete.
24956 (class temporal_cache): Adjust.
24957 (temporal_cache::get_timestamp): Delete.
24958 (temporal_cache::set_dependency): Delete.
24959 (temporal_cache::temporal_value): Adjust.
24960 (temporal_cache::current_p): Take dependencies as params.
24961 (temporal_cache::set_timestamp): Adjust.
24962 (temporal_cache::set_always_current): Adjust.
24963 (ranger_cache::get_non_stale_global_range): Adjust.
24964 (ranger_cache::register_dependency): Delete.
24965 * gimple-range-cache.h (class range_cache): Adjust.
24967 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
24969 * gimple-range-gori.cc (range_def_chain::range_def_chain): init
24971 (range_def_chain::~range_def_chain): Dispose of obstack rather than
24972 each individual bitmap.
24973 (range_def_chain::set_import): New.
24974 (range_def_chain::get_imports): New.
24975 (range_def_chain::chain_import_p): New.
24976 (range_def_chain::register_dependency): Rename from build_def_chain
24978 (range_def_chain::def_chain_in_bitmap_p): New.
24979 (range_def_chain::add_def_chain_to_bitmap): New.
24980 (range_def_chain::has_def_chain): Just check first depenedence.
24981 (range_def_chain::get_def_chain): Process imports, use generic
24982 register_dependency routine.
24983 (range_def_chain::dump): New.
24984 (gori_map::gori_map): Allocate import list.
24985 (gori_map::~gori_map): Release imports.
24986 (gori_map::exports): Check for past allocated block size.
24987 (gori_map::imports): New.
24988 (gori_map::def_chain_in_export_p): Delete.
24989 (gori_map::is_import_p): New.
24990 (gori_map::maybe_add_gori): Handle imports.
24991 (gori_map::dump): Adjust output, add imports.
24992 (gori_compute::has_edge_range_p): Remove def_chain_in_export call.
24993 (gori_export_iterator::gori_export_iterator): New.
24994 (gori_export_iterator::next): New.
24995 (gori_export_iterator::get_name): New.
24996 * gimple-range-gori.h (range_def_chain): Add imports and direct
24997 dependecies via struct rdc.
24998 (range_def_chain::depend1): New.
24999 (range_def_chain::depend2): New.
25000 (class gori_map): Adjust.
25001 (FOR_EACH_GORI_IMPORT_NAME): New.
25002 (FOR_EACH_GORI_EXPORT_NAME): New.
25003 (class gori_export_iterator): New.
25005 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
25007 * gimple-range-cache.cc (ranger_cache::ranger_cache): Move initial
25008 export cache filling to here.
25009 * gimple-range-gori.cc (gori_compute::gori_compute) : From Here.
25011 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
25013 * gimple-range-gori.cc (range_def_chain): Move to gimple-range-gori.h.
25014 (gori_map): Move to gimple-range-gori.h.
25015 (gori_compute::gori_compute): Adjust.
25016 (gori_compute::~gori_compute): Delete.
25017 (gori_compute::compute_operand_range_switch): Adjust.
25018 (gori_compute::compute_operand_range): Adjust.
25019 (gori_compute::compute_logical_operands): Adjust.
25020 (gori_compute::has_edge_range_p ): Adjust.
25021 (gori_compute::set_range_invariant): Delete.
25022 (gori_compute::dump): Adjust.
25023 (gori_compute::outgoing_edge_range_p): Adjust.
25024 * gimple-range-gori.h (class range_def_chain): Relocate here.
25025 (class gori_map): Relocate here.
25026 (class gori_compute): Inherit from gori_map, and adjust.
25028 2021-05-25 Aldy Hernandez <aldyh@redhat.com>
25030 * value-range.cc (range_tests_legacy): Use
25031 build_nonstandard_integer_type instead of int and short.
25033 2021-05-25 Eric Botcazou <ebotcazou@adacore.com>
25035 * gimplify.c (gimplify_decl_expr): Clear TREE_READONLY on the DECL
25036 when really creating an initialization statement for it.
25038 2021-05-25 Eric Botcazou <ebotcazou@adacore.com>
25040 * tree-inline.c (setup_one_parameter): Fix thinko in new condition.
25042 2021-05-25 Kito Cheng <kito.cheng@sifive.com>
25044 * config/riscv/riscv.h (ASM_SPEC): Pass -mno-relax.
25046 2021-05-25 Martin Liska <mliska@suse.cz>
25048 PR tree-optimization/92860
25050 * optc-save-gen.awk: Remove exceptions.
25052 2021-05-25 Martin Liska <mliska@suse.cz>
25054 * asan.h (sanitize_coverage_p): New function.
25055 * doc/extend.texi: Document it.
25056 * fold-const.c (fold_range_test): Use sanitize_flags_p
25057 instead of flag_sanitize_coverage.
25058 (fold_truth_andor): Likewise.
25059 * sancov.c: Likewise.
25060 * tree-ssa-ifcombine.c (ifcombine_ifandif): Likewise.
25061 * ipa-inline.c (sanitize_attrs_match_for_inline_p): Handle
25062 -fsanitize-coverage when inlining.
25064 2021-05-25 Cooper Qu <cooper.qu@linux.alibaba.com>
25066 * config/csky/csky-modes.def : Fix copyright.
25068 2021-05-25 Cooper Qu <cooper.qu@linux.alibaba.com>
25070 * config/csky/csky-modes.def : Amend copyright.
25071 * config/csky/csky_insn_fpuv2.md : Likewise.
25072 * config/csky/csky_insn_fpuv3.md : Likewise.
25074 2021-05-25 Richard Biener <rguenther@suse.de>
25076 PR middle-end/100727
25077 * calls.c (initialize_argument_information): Explicitely test
25078 for WITH_SIZE_EXPR.
25079 * gimple-expr.c (mark_addressable): Skip outer WITH_SIZE_EXPR.
25081 2021-05-25 Geng Qi <gengqi@linux.alibaba.com>
25083 * config/csky/csky.h (FRAME_POINTER_REGNUM): Use
25084 HARD_FRAME_POINTER_REGNUM and FRAME_POINTER_REGNUM instead of
25085 the signle definition. The signle definition may not work well
25086 at simplify_subreg_regno().
25087 (HARD_FRAME_POINTER_REGNUM): New.
25088 (ELIMINABLE_REGS): Add for HARD_FRAME_POINTER_REGNUM.
25089 * config/csky/csky.c (get_csky_live_regs, csky_can_eliminate,
25090 csky_initial_elimination_offset, csky_expand_prologue,
25091 csky_expand_epilogue): Add for HARD_FRAME_POINTER_REGNUM.
25093 2021-05-25 Geng Qi <gengqi@linux.alibaba.com>
25095 * config/csky/csky.c (csky_option_override):
25096 Init csky_arch_isa_features[] in advance, so TARGET_DSP
25097 and TARGET_DIV can be set well.
25099 2021-05-25 Geng Qi <gengqi@linux.alibaba.com>
25101 * config/csky/constraints.md ("l", "h"): Delete.
25102 * config/csky/csky.h (reg_class, REG_CLASS_NAMES,
25103 REG_CLASS_CONTENTS): Delete LO_REGS and HI_REGS.
25104 * config/csky/csky.c (regno_reg_classm,
25105 csky_secondary_reload, csky_register_move_cost):
25106 Use HILO_REGS instead of LO_REGS and HI_REGS.
25108 2021-05-25 Geng Qi <gengqi@linux.alibaba.com>
25110 * config/csky/constraints.md ("W"): New constriant for mem operand
25111 with base reg, index register.
25112 ("Q"): Renamed and modified "csky_valid_fpuv2_mem_operand" to
25113 "csky_valid_mem_constraint_operand" to deal with both "Q" and "W"
25115 ("Dv"): New constraint for const double value that can be used at
25117 * config/csky/csky-modes.def (HFmode): New mode.
25118 * config/csky/csky-protos.h (csky_valid_fpuv2_mem_operand): Rename
25119 to "csky_valid_mem_constraint_operand" and support new constraint
25121 (csky_get_movedouble_length): New.
25122 (fpuv3_output_move): New.
25123 (fpuv3_const_double): New.
25124 * config/csky/csky.c (csky_option_override): New arch CK860 with fpv3.
25125 (decompose_csky_address): Refine.
25126 (csky_print_operand): New "CONST_DOUBLE" operand.
25127 (csky_output_move): Support fpv3 instructions.
25128 (csky_get_movedouble_length): New.
25129 (fpuv3_output_move): New.
25130 (fpuv3_const_double): New.
25131 (csky_emit_compare): Cover float comparsion.
25132 (csky_emit_compare_float): Refine.
25133 (csky_vaild_fpuv2_mem_operand): Rename to
25134 "csky_valid_mem_constraint_operand" and support new constraint "W".
25135 (ck860_rtx_costs): New.
25136 (csky_rtx_costs): Add the cost calculation of CK860.
25137 (regno_reg_class): New vregs for fpuv3.
25138 (csky_dbx_regno): Likewise.
25139 (csky_cpu_cpp_builtins): New builtin macro for fpuv3.
25140 (csky_conditional_register_usage): Suporrot fpuv3.
25141 (csky_dwarf_register_span): Suporrot fpuv3.
25142 (csky_init_builtins, csky_mangle_type): Support "__fp16" type.
25143 (ck810_legitimate_index_p): Support fp16.
25144 * config/csky/csky.h (TARGET_TLS): ADD CK860.
25145 (CSKY_VREG_P, CSKY_VREG_LO_P, CSKY_VREG_HI_P): Support fpuv3.
25146 (TARGET_SINGLE_FPU): Support fpuv3.
25147 (TARGET_SUPPORT_FPV3): New.
25148 (FIRST_PSEUDO_REGISTER): Change to 202 to hold the new fpuv3 registers.
25149 (FIXED_REGISTERS, CALL_REALLY_USED_REGISTERS, REGISTER_NAMES,
25150 REG_CLASS_CONTENTS): Support fpuv3.
25151 * config/csky/csky.md (movsf): Move to cksy_insn_fpu.md and refine.
25152 (csky_movsf_fpv2): Likewise.
25153 (ck801_movsf): Likewise.
25154 (csky_movsf): Likewise.
25156 (csky_movdf_fpv2): Likewise.
25157 (ck801_movdf): Likewise.
25158 (csky_movdf): Likewise.
25159 (movsicc): Refine. Use "comparison_operatior" instead of
25160 "ordered_comparison_operatior".
25161 (addsicc): Likewise.
25162 (CSKY_FIRST_VFP3_REGNUM, CSKY_LAST_VFP3_REGNUM): New constant.
25163 (call_value_internal_vh): New.
25164 * config/csky/csky_cores.def (CK860): New arch and cpu.
25169 * config/csky/csky_insn_fpu.md: Refactor. Separate all float patterns
25170 into emit-patterns and match-patterns, remain the emit-patterns here,
25171 and move the match-patterns to csky_insn_fpuv2.md or
25172 csky_insn_fpuv3.md.
25173 * config/csky/csky_insn_fpuv2.md: New file for fpuv2 instructions.
25174 * config/csky/csky_insn_fpuv3.md: New file and new patterns for fpuv3
25176 * config/csky/csky_isa.def (fcr): New.
25181 (CK860): New definition for ck860.
25182 * config/csky/csky_tables.opt (ck860): New processors ck860,
25183 ck860f. And new arch ck860.
25188 * config/csky/predicates.md (csky_float_comparsion_operator): Delete
25189 "geu", "gtu", "leu", "ltu", which will never appear at float comparison.
25190 * config/csky/t-csky-elf: Support 860.
25191 * config/csky/t-csky-linux: Likewise.
25192 * doc/md.texi: Add "Q" and "W" constraints for C-SKY.
25194 2021-05-24 Aaron Sawdey <acsawdey@linux.ibm.com>
25196 * config/rs6000/genfusion.pl (gen_logical_addsubf): Refactor to
25197 add generation of logical-add and add-logical fusion pairs.
25198 * config/rs6000/rs6000-cpus.def: Add new fusion to ISA 3.1 mask
25200 * config/rs6000/rs6000.c (rs6000_option_override_internal): Turn on
25201 logical-add and add-logical fusion by default.
25202 * config/rs6000/rs6000.opt: Add -mpower10-fusion-logical-add and
25203 -mpower10-fusion-add-logical options.
25204 * config/rs6000/fusion.md: Regenerate file.
25206 2021-05-24 Aldy Hernandez <aldyh@redhat.com>
25208 * value-range.cc (irange::legacy_equal_p): Check type when
25209 comparing VR_VARYING types.
25210 (range_tests_legacy): Test comparing VARYING ranges of different
25213 2021-05-24 Wilco Dijkstra <wdijkstr@arm.com>
25215 * config/aarch64/aarch64.c (neoversen1_tunings):
25216 Enable AARCH64_EXTRA_TUNE_CHEAP_SHIFT_EXTEND.
25218 2021-05-24 Wilco Dijkstra <wdijkstr@arm.com>
25220 * config/aarch64/aarch64.c (aarch64_classify_symbol): Use GOT for
25221 extern weak symbols. Limit symbol offsets for non-GOT symbols with
25224 2021-05-24 Christophe Lyon <christophe.lyon@linaro.org>
25226 * config/arm/neon.md (vec_load_lanesxi<mode>)
25227 (vec_store_lanexoi<mode>): Move ...
25228 * config/arm/vec-common.md: here.
25230 2021-05-24 Christophe Lyon <christophe.lyon@linaro.org>
25232 * config/arm/neon.md (vec_load_lanesoi<mode>)
25233 (vec_store_lanesoi<mode>): Move ...
25234 * config/arm/vec-common.md: here.
25236 2021-05-24 liuhongt <hongtao.liu@intel.com>
25239 * config/i386/i386.c (ix86_gimple_fold_builtin): Replacing
25240 stmt with GIMPLE_NOP when lhs doesn't exist.
25242 2021-05-23 Uroš Bizjak <ubizjak@gmail.com>
25245 * config/i386/mmx.md (*push<VI_32:mode>2_rex64):
25246 New instruction pattern.
25247 (*push<VI_32:mode>2): Ditto.
25248 (push splitter for SSE registers): New splitter.
25250 2021-05-23 Andrew Pinski <apinski@marvell.com>
25252 * match.pd ((A & C) != 0 ? D : 0): Limit to non pointer types.
25254 2021-05-22 Aaron Sawdey <acsawdey@linux.ibm.com>
25256 * config/rs6000/genfusion.pl (gen_addadd): Fix incorrect attr types.
25257 * config/rs6000/fusion.md: Regenerate file.
25259 2021-05-21 Aaron Sawdey <acsawdey@linux.ibm.com>
25261 * config/rs6000/genfusion.pl (gen_addadd): New function.
25262 * config/rs6000/fusion.md: Regenerate file.
25263 * config/rs6000/rs6000-cpus.def: Add
25264 OPTION_MASK_P10_FUSION_2ADD to masks.
25265 * config/rs6000/rs6000.c (rs6000_option_override_internal):
25266 Handle default value of OPTION_MASK_P10_FUSION_2ADD.
25267 * config/rs6000/rs6000.opt: Add -mpower10-fusion-2add.
25269 2021-05-21 Jakub Jelinek <jakub@redhat.com>
25271 PR middle-end/99928
25272 * tree.h (OMP_CLAUSE_FIRSTPRIVATE_IMPLICIT_TARGET): Define.
25273 * gimplify.c (enum gimplify_omp_var_data): Fix up
25274 GOVD_MAP_HAS_ATTACHMENTS value, add GOVD_FIRSTPRIVATE_IMPLICIT.
25275 (omp_lastprivate_for_combined_outer_constructs): If combined target
25276 has GOVD_FIRSTPRIVATE_IMPLICIT set for the decl, change it to
25277 GOVD_MAP | GOVD_SEEN.
25278 (gimplify_scan_omp_clauses): Set GOVD_FIRSTPRIVATE_IMPLICIT for
25279 firstprivate clauses with OMP_CLAUSE_FIRSTPRIVATE_IMPLICIT.
25280 (gimplify_adjust_omp_clauses): For firstprivate clauses with
25281 OMP_CLAUSE_FIRSTPRIVATE_IMPLICIT either clear that bit and
25282 OMP_CLAUSE_FIRSTPRIVATE_IMPLICIT_TARGET too, or remove it and
25283 let it be replaced by implicit map clause.
25285 2021-05-21 Jakub Jelinek <jakub@redhat.com>
25287 PR middle-end/99928
25288 * gimplify.c (omp_lastprivate_for_combined_outer_constructs): New
25290 (gimplify_scan_omp_clauses) <case OMP_CLAUSE_LASTPRIVATE>: Use it.
25291 (gimplify_omp_for): Likewise.
25293 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
25295 PR middle-end/90115
25296 * omp-low.c (oacc_privatization_candidate_p): Reject 'static',
25297 'external' in blocks.
25299 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
25301 PR middle-end/90115
25302 * flag-types.h (enum openacc_privatization): New.
25303 * params.opt (-param=openacc-privatization): New.
25304 * doc/invoke.texi (openacc-privatization): Document it.
25305 * omp-general.h (get_openacc_privatization_dump_flags): New
25307 * omp-low.c (oacc_privatization_candidate_p): Add diagnostics.
25308 * omp-offload.c (execute_oacc_device_lower)
25309 <IFN_UNIQUE_OACC_PRIVATE>: Re-work diagnostics.
25310 * target.def (goacc.adjust_private_decl): Add 'location_t'
25312 * doc/tm.texi: Regenerate.
25313 * config/gcn/gcn-protos.h (gcn_goacc_adjust_private_decl): Adjust.
25314 * config/gcn/gcn-tree.c (gcn_goacc_adjust_private_decl): Likewise.
25315 * config/nvptx/nvptx.c (nvptx_goacc_adjust_private_decl):
25316 Likewise. Preserve it for...
25317 (nvptx_goacc_expand_var_decl): ... use here.
25319 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
25321 * doc/sourcebuild.texi (Other attributes): Document '__OPTIMIZE__'
25324 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
25326 PR middle-end/90115
25327 * omp-low.c (oacc_privatization_candidate_p): New function.
25328 (oacc_privatization_scan_clause_chain)
25329 (oacc_privatization_scan_decl_chain): Use it. Also
25330 'gcc_checking_assert' that we're not seeing duplicates.
25332 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
25334 PR middle-end/90115
25335 * omp-offload.c (execute_oacc_device_lower): Skip processing if no
25338 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
25340 PR middle-end/90115
25341 * omp-offload.c (execute_oacc_device_lower): Explain.
25343 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
25345 PR middle-end/90115
25346 * omp-offload.c (execute_oacc_device_lower)
25347 <IFN_UNIQUE_OACC_PRIVATE>: Diagnose and handle for 'level == -1'
25349 * internal-fn.c (expand_UNIQUE): Don't expect
25350 'IFN_UNIQUE_OACC_PRIVATE'.
25352 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
25354 PR middle-end/90115
25355 * omp-low.c (lower_omp_for): Don't evaluate OpenMP 'for' clauses.
25357 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
25359 PR middle-end/90115
25360 * config/nvptx/nvptx.c (nvptx_goacc_adjust_private_decl)
25361 (nvptx_goacc_expand_var_decl): Tighten.
25363 2021-05-21 Julian Brown <julian@codesourcery.com>
25364 Chung-Lin Tang <cltang@codesourcery.com>
25365 Thomas Schwinge <thomas@codesourcery.com>
25367 PR middle-end/90115
25368 * doc/tm.texi.in (TARGET_GOACC_EXPAND_VAR_DECL)
25369 (TARGET_GOACC_ADJUST_PRIVATE_DECL): Add documentation hooks.
25370 * doc/tm.texi: Regenerate.
25371 * expr.c (expand_expr_real_1): Expand decls using the
25372 expand_var_decl OpenACC hook if defined.
25373 * internal-fn.c (expand_UNIQUE): Handle IFN_UNIQUE_OACC_PRIVATE.
25374 * internal-fn.h (IFN_UNIQUE_CODES): Add OACC_PRIVATE.
25375 * omp-low.c (omp_context): Add oacc_privatization_candidates
25377 (lower_oacc_reductions): Add PRIVATE_MARKER parameter. Insert
25379 (lower_oacc_head_tail): Add PRIVATE_MARKER parameter. Modify
25380 private marker's gimple call arguments, and pass it to
25381 lower_oacc_reductions.
25382 (oacc_privatization_scan_clause_chain)
25383 (oacc_privatization_scan_decl_chain, lower_oacc_private_marker):
25385 (lower_omp_for, lower_omp_target, lower_omp_1): Use these.
25386 * omp-offload.c (convert.h): Include.
25387 (oacc_loop_xform_head_tail): Treat private-variable markers like
25388 fork/join when transforming head/tail sequences.
25389 (struct var_decl_rewrite_info): Add struct.
25390 (oacc_rewrite_var_decl, is_sync_builtin_call): New functions.
25391 (execute_oacc_device_lower): Support rewriting gang-private
25392 variables using target hook, and fix up addr_expr and var_decl
25394 * target.def (adjust_private_decl, expand_var_decl): New hooks.
25395 * config/gcn/gcn-protos.h (gcn_goacc_adjust_gangprivate_decl):
25397 (gcn_goacc_adjust_private_decl): ...this.
25398 * config/gcn/gcn-tree.c (gcn_goacc_adjust_gangprivate_decl):
25400 (gcn_goacc_adjust_private_decl): ...this. Add LEVEL parameter.
25401 * config/gcn/gcn.c (TARGET_GOACC_ADJUST_GANGPRIVATE_DECL): Rename
25402 definition using gcn_goacc_adjust_gangprivate_decl...
25403 (TARGET_GOACC_ADJUST_PRIVATE_DECL): ...to this, using
25404 gcn_goacc_adjust_private_decl.
25405 * config/nvptx/nvptx.c (tree-pretty-print.h): Include.
25406 (gang_private_shared_size): New global variable.
25407 (gang_private_shared_align): Likewise.
25408 (gang_private_shared_sym): Likewise.
25409 (gang_private_shared_hmap): Likewise.
25410 (nvptx_option_override): Initialize these.
25411 (nvptx_file_end): Output gang_private_shared_sym.
25412 (nvptx_goacc_adjust_private_decl, nvptx_goacc_expand_var_decl):
25414 (nvptx_set_current_function): Clear gang_private_shared_hmap.
25415 (TARGET_GOACC_ADJUST_PRIVATE_DECL): Define hook.
25416 (TARGET_GOACC_EXPAND_VAR_DECL): Likewise.
25418 2021-05-21 H.J. Lu <hjl.tools@gmail.com>
25420 * config/i386/i386-modes.def (MAX_BITSIZE_MODE_ANY_INT): Removed.
25422 2021-05-21 Richard Biener <rguenther@suse.de>
25423 H.J. Lu <hjl.tools@gmail.com>
25425 PR middle-end/90773
25426 * expr.c (expand_constructor): Elide expand_constructor if
25427 move by pieces is preferred.
25429 2021-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
25431 * config/aarch64/aarch64-builtins.c (aarch64_call_properties):
25432 Take a flag and mode value as arguments.
25433 (aarch64_modifies_global_state_p): Likewise.
25434 (aarch64_reads_global_state_p): Likewise.
25435 (aarch64_could_trap_p): Likewise.
25436 (aarch64_get_attributes): Likewise.
25437 (aarch64_init_simd_builtins): Adjust callsite of above.
25438 (aarch64_init_fcmla_laneq_builtins): Use aarch64_get_attributes to get
25439 function attributes to apply to builtins.
25440 (aarch64_init_crc32_builtins): Likewise.
25441 (aarch64_init_builtin_rsqrt): Likewise.
25443 2021-05-21 Aaron Sawdey <acsawdey@linux.ibm.com>
25445 * config/rs6000/rs6000.md (define_attr "type"): Add types for fusion.
25446 * config/rs6000/genfusion.pl (gen_ld_cmpi_p10): Use new fusion types.
25447 (gen_2logical): Use new fusion types.
25448 * config/rs6000/fusion.md: Regenerate.
25450 2021-05-21 Uroš Bizjak <ubizjak@gmail.com>
25453 * config/i386/i386-expand.c (ix86_expand_sse_movcc):
25454 Handle V4QI and V2HI modes.
25455 (ix86_expand_sse_movcc): Ditto.
25456 * config/i386/mmx.md (*<sat_plusminus:insn><VI_32:mode>3):
25457 New instruction pattern.
25458 (*eq<VI_32:mode>3): Ditto.
25459 (*gt<VI_32:mode>3): Ditto.
25460 (*xop_pcmov_<VI_32:mode>): Ditto.
25461 (mmx_pblendvb32): Ditto.
25462 (mmx_pblendvb64): Rename from mmx_pblendvb.
25463 (vec_cmp<VI_32:mode><VI_32:mode>): New expander.
25464 (vec_cmpu<VI_32:mode><VI_32:mode>): Ditto.
25465 (vcond<VI_32:mode><VI_32:mode>): Ditto.
25466 (vcondu<VI_32:mode><VI_32:mode>): Ditto.
25467 (vcond_mask_<VI_32:mode><VI_32:mode>): Ditto.
25469 2021-05-21 Jakub Jelinek <jakub@redhat.com>
25471 PR tree-optimization/94589
25472 * tree-ssa-phiopt.c (spaceship_replacement): For integral rhs1 and
25473 rhs2, treat x <= 4 equivalently to x < 5 etc. In cmp1 and cmp2 (if
25474 not the same as cmp3) treat <= the same as < and >= the same as >.
25475 Don't require that cond2_phi_edge is true edge, instead take
25476 false/true edges into account based on cmp1/cmp2 comparison kinds.
25478 2021-05-21 Uroš Bizjak <ubizjak@gmail.com>
25481 * config/i386/mmx.md (SMAXMIN_MMXMODEI): New mode iterator.
25482 (<smaxmin:code><SMAXMIN_MMXMODEI:mode>3): Macroize expander
25483 from <smaxmin:code>v4hi3> and <smaxmin:code><MMXMODE14:mode>3
25484 using SMAXMIN_MMXMODEI mode iterator.
25485 (*<smaxmin:code>v4qi3): New insn pattern.
25486 (*<smaxmin:code>v2hi3): Ditto.
25487 (SMAXMIN_VI_32): New mode iterator.
25488 (<smaxmin:code><SMAXMIN_VI_32>mode3): New expander.
25489 (UMAXMIN_MMXMODEI): New mode iterator.
25490 (<umaxmin:code><UMAXMIN_MMXMODEI:mode>3): Macroize expander
25491 from <umaxmin:code>v8qi3> and <umaxmin:code><MMXMODE24:mode>3
25492 using UMAXMIN_MMXMODEI mode iterator.
25493 (*<umaxmin:code>v4qi3): New insn pattern.
25494 (*<umaxmin:code>v2hi3): Ditto.
25495 (UMAXMIN_VI_32): New mode iterator.
25496 (<umaxmin:code><UMAXMIN_VI_32>mode3): New expander.
25497 (abs<VI_32:mode>2): New insn pattern.
25498 (ssse3_abs<MMXMODEI:mode>2, abs<MMXMODEI:mode>2): Move from ...
25499 * config/i386/sse.md: ... here.
25501 2021-05-20 Clement Chigot <clement.chigot@atos.net>
25502 David Edelsohn <dje.gcc@gmail.com>
25504 * collect2.c (scan_prog_file): Issue non-fatal warning for
25507 2021-05-20 Jonathan Wakely <jwakely@redhat.com>
25509 * doc/invoke.texi (-Wno-c++11-extensions)
25510 (-Wno-c++14-extensions, -Wno-c++17-extensions)
25511 (-Wno-c++20-extensions, -Wno-c++23-extensions): Document
25514 2021-05-20 Indu Bhagat <indu.bhagat@oracle.com>
25516 * config/c6x/c6x.c (c6x_output_file_unwind): Use dwarf_debuginfo_p.
25517 * config/darwin.c (darwin_override_options): Likewise.
25518 * config/i386/cygming.h (DBX_REGISTER_NUMBER): Likewise.
25519 * config/i386/darwin.h (DBX_REGISTER_NUMBER): Likewise.
25520 (DWARF2_FRAME_REG_OUT): Likewise.
25521 * config/mips/mips.c (mips_output_filename): Likewise.
25522 * config/rs6000/rs6000.c (rs6000_xcoff_declare_function_name):
25524 (rs6000_dbx_register_number): Likewise.
25525 * dbxout.c: Include flags.h.
25526 * dwarf2cfi.c (cfi_label_required_p): Likewise.
25527 (dwarf2out_do_frame): Likewise.
25528 * except.c: Include flags.h.
25529 * final.c (dwarf2_debug_info_emitted_p): Likewise.
25530 (final_scan_insn_1): Likewise.
25531 * flags.h (dwarf_debuginfo_p): New function declaration.
25532 * opts.c (dwarf_debuginfo_p): New function definition.
25533 * targhooks.c (default_debug_unwind_info): Use dwarf_debuginfo_p.
25534 * toplev.c (process_options): Likewise.
25536 2021-05-20 Indu Bhagat <indu.bhagat@oracle.com>
25538 * common.opt: Change type to support bitmasks.
25539 * flag-types.h (enum debug_info_type): Rename enumerator constants.
25540 (NO_DEBUG): New bitmask.
25541 (DBX_DEBUG): Likewise.
25542 (DWARF2_DEBUG): Likewise.
25543 (XCOFF_DEBUG): Likewise.
25544 (VMS_DEBUG): Likewise.
25545 (VMS_AND_DWARF2_DEBUG): Likewise.
25546 * flags.h (debug_set_to_format): New function declaration.
25547 (debug_set_count): Likewise.
25548 (debug_set_names): Likewise.
25549 * opts.c (debug_type_masks): Array of bitmasks for debug formats.
25550 (debug_set_to_format): New function definition.
25551 (debug_set_count): Likewise.
25552 (debug_set_names): Likewise.
25553 (set_debug_level): Update access to debug_type_names.
25554 * toplev.c: Likewise.
25556 2021-05-20 Martin Sebor <msebor@redhat.com>
25558 PR middle-end/100684
25559 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Handle C++ lambda.
25561 2021-05-20 Uroš Bizjak <ubizjak@gmail.com>
25564 * config/i386/i386.md (isa): Remove x64_bmi.
25565 (enabled): Remove x64_bmi.
25566 * config/i386/mmx.md (mmx_andnot<MMXMODEI:mode>3):
25567 Remove general register alternative.
25568 (*andnot<VI_32:mode>3): Ditto.
25569 (*mmx_<any_logic:code><MMXMODEI:mode>3): Ditto.
25570 (*<any_logic:code><VI_32:mode>3): Ditto.
25572 2021-05-20 Kewen Lin <linkw@linux.ibm.com>
25574 * config/arm/arm.c: Include head files tree-vectorizer.h and
25577 2021-05-20 Uroš Bizjak <ubizjak@gmail.com>
25580 * config/i386/mmx.md (Yv_Yw): Revert adding V4QI and V2HI modes.
25581 (*<plusminus:insn><VI_32:mode>3): Use Yw instad of <Yv_Yw> constrint.
25582 (<s>mulv4hi3_highpart): New expander.
25583 (*<s>mulv2hi3_highpart): New insn pattern.
25584 (<s>mulv2hi3_higpart): New expander.
25585 (*<any_shift:insn>v2hi3): New insn pattern.
25586 (<any_shift:insn>v2hi3): New expander.
25587 * config/i386/sse.md (smulhrsv2hi3): New expander.
25588 (*smulhrsv2hi3): New insn pattern.
25590 2021-05-20 Kewen Lin <linkw@linux.ibm.com>
25592 * doc/invoke.texi (vect-inner-loop-cost-factor): Document new
25594 * params.opt (vect-inner-loop-cost-factor): New.
25595 * targhooks.c (default_add_stmt_cost): Replace hardcoded factor
25596 50 with LOOP_VINFO_INNER_LOOP_COST_FACTOR, include head file
25597 tree-vectorizer.h and its required ones.
25598 * config/aarch64/aarch64.c (aarch64_add_stmt_cost): Replace
25599 hardcoded factor 50 with LOOP_VINFO_INNER_LOOP_COST_FACTOR.
25600 * config/arm/arm.c (arm_add_stmt_cost): Likewise.
25601 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
25602 * config/rs6000/rs6000.c (rs6000_add_stmt_cost): Likewise.
25603 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
25605 (_loop_vec_info::_loop_vec_info): Init inner_loop_cost_factor.
25606 * tree-vectorizer.h (_loop_vec_info): Add inner_loop_cost_factor.
25607 (LOOP_VINFO_INNER_LOOP_COST_FACTOR): New macro.
25609 2021-05-20 Christophe Lyon <christophe.lyon@linaro.org>
25610 Torbjörn Svensson <torbjorn.svensson@st.com>
25613 * doc/cpp.texi (Common Predefined Macros): Document __FILE_NAME__.
25615 2021-05-20 Jakub Jelinek <jakub@redhat.com>
25617 PR middle-end/99928
25618 * gimplify.c (gimplify_scan_omp_clauses) <case OMP_CLAUSE_LINEAR>: For
25619 explicit linear clause when combined with target, make it map(tofrom:)
25620 instead of no clause or firstprivate.
25622 2021-05-20 Jakub Jelinek <jakub@redhat.com>
25624 PR tree-optimization/94589
25625 * match.pd ((X & Y) == X -> (X & ~Y) == 0): Simplify even in presence
25626 of integral conversions.
25628 2021-05-19 Andrew MacLeod <amacleod@redhat.com>
25630 * gimple-range.cc (fur_source::get_operand): New.
25631 (gimple_range_fold): Delete.
25632 (fold_using_range::fold_stmt): Move from gimple_ranger::calc_stmt.
25633 (fold_using_range::range_of_range_op): Move from gimple_ranger.
25634 (fold_using_range::range_of_address): Ditto.
25635 (fold_using_range::range_of_phi): Ditto.
25636 (fold_using_range::range_of_call): Ditto.
25637 (fold_using_range::range_of_builtin_ubsan_call): Move from
25638 range_of_builtin_ubsan_call.
25639 (fold_using_range::range_of_builtin_call): Move from
25640 range_of_builtin_call.
25641 (gimple_ranger::range_of_builtin_call): Delete.
25642 (fold_using_range::range_of_cond_expr): Move from gimple_ranger.
25643 (gimple_ranger::fold_range_internal): New.
25644 (gimple_ranger::range_of_stmt): Use new fold_using_range API.
25645 (fold_using_range::range_of_ssa_name_with_loop_info): Move from
25646 gimple_ranger. Improve ranges of SSA_NAMES when possible.
25647 * gimple-range.h (gimple_ranger): Remove various range_of routines.
25648 (class fur_source): New.
25649 (class fold_using_range): New.
25650 (fur_source::fur_source): New.
25652 * vr-values.c (vr_values::extract_range_basic): Use fold_using_range
25653 instead of range_of_builtin_call.
25655 2021-05-19 Jonathan Wakely <jwakely@redhat.com>
25657 * doc/cpp.texi (Common Predefined Macros): Update documentation
25658 for the __GXX_EXPERIMENTAL_CXX0X__ macro.
25660 2021-05-19 Alex Coplan <alex.coplan@arm.com>
25663 * config/arm/arm.md (nonsecure_call_internal): Always ensure
25664 callee's address is in a register.
25666 2021-05-19 Geng Qi <gengqi@linux.alibaba.com>
25668 * common/config/riscv/riscv-common.c
25669 (riscv_subset_list::parsing_subset_version): Properly parse the letter
25671 (riscv_subset_list::parse_std_ext,
25672 riscv_subset_list::parse_multiletter_ext): To handle errors generated
25673 in riscv_subset_list::parsing_subset_version.
25675 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
25677 * config/aarch64/aarch64-simd.md: Use "neon_move_narrow_q"
25678 type attribute in patterns generating XTN(2).
25680 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
25682 * config/aarch64/aarch64-simd.md (aarch64_simd_vec_pack_trunc_<mode>):
25683 Remove as duplicate of...
25684 (aarch64_xtn<mode>): This.
25685 (aarch64_xtn2<mode>_le): Move position in file.
25686 (aarch64_xtn2<mode>_be): Move position in file.
25687 (aarch64_xtn2<mode>): Move position in file.
25688 (vec_pack_trunc_<mode>): Define as an expander.
25690 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
25692 * config/aarch64/aarch64-simd-builtins.def: Split builtin
25693 generation for aarch64_<sur>q<r>shr<u>n_n<mode> pattern into
25694 separate scalar and vector generators.
25695 * config/aarch64/aarch64-simd.md
25696 (aarch64_<sur>q<r>shr<u>n_n<mode>): Define as an expander and
25698 (aarch64_<sur>q<r>shr<u>n_n<mode>_insn_le): This and...
25699 (aarch64_<sur>q<r>shr<u>n_n<mode>_insn_be): This.
25700 * config/aarch64/iterators.md: Define SD_HSDI iterator.
25702 2021-05-19 Jonathn Wright <jonathan.wright@arm.com>
25704 * config/aarch64/aarch64-simd.md: Use UNSPEC_SQXTUN instead
25706 * config/aarch64/iterators.md: Remove UNSPEC_SQXTUN2.
25708 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
25710 * config/aarch64/aarch64-simd.md (aarch64_<sur>q<r>shr<u>n2_n<mode>):
25711 Implement as an expand emitting a big/little endian
25712 instruction pattern.
25713 (aarch64_<sur>q<r>shr<u>n2_n<mode>_insn_le): Define.
25714 (aarch64_<sur>q<r>shr<u>n2_n<mode>_insn_be): Define.
25716 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
25718 * config/aarch64/aarch64-simd.md (aarch64_<sur><addsub>hn2<mode>):
25719 Implement as an expand emitting a big/little endian
25720 instruction pattern.
25721 (aarch64_<sur><addsub>hn2<mode>_insn_le): Define.
25722 (aarch64_<sur><addsub>hn2<mode>_insn_be): Define.
25723 * config/aarch64/iterators.md: Remove UNSPEC_[R]ADDHN2 and
25724 UNSPEC_[R]SUBHN2 unspecs and ADDSUBHN2 iterator.
25726 2021-05-19 Richard Biener <rguenther@suse.de>
25728 PR middle-end/100672
25729 * fold-const.c (fold_negate_expr_1): Use element_precision.
25730 (negate_expr_p): Likewise.
25732 2021-05-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
25734 * config/aarch64/iterators.md (SVE_PRED_LOAD): New iterator.
25735 (pred_load): New int attribute.
25736 * config/aarch64/aarch64-sve.md
25737 (aarch64_load_<ANY_EXTEND:optab><SVE_HSDI:mode><SVE_PARTIAL_I:mode>): Use
25738 SVE_PRED_LOAD enum iterator and corresponding pred_load attribute.
25739 * config/aarch64/aarch64-sve-builtins-base.cc (expand): Update call to
25740 code_for_aarch64_load.
25742 2021-05-19 Richard Biener <rguenther@suse.de>
25744 * cfgexpand.c (discover_nonconstant_array_refs_r): Make
25745 sure TARGET_MEM_REF bases are expanded as memory.
25746 * tree-ssa-operands.c (operands_scanner::get_tmr_operands):
25747 Do not mark TARGET_MEM_REF bases addressable.
25748 * tree-ssa.c (non_rewritable_mem_ref_base): Handle
25749 TARGET_MEM_REF bases as never rewritable.
25750 * gimple-walk.c (walk_stmt_load_store_addr_ops): Do not
25751 walk TARGET_MEM_REF bases as address-takens.
25752 * tree-ssa-dce.c (ref_may_be_aliased): Handle TARGET_MEM_REF.
25754 2021-05-19 Richard Biener <rguenther@suse.de>
25756 * builtins.c (get_object_alignment_1): Strip outer
25758 * tree-dfa.c (get_ref_base_and_extent): Handle outer
25759 WITH_SIZE_EXPR for size processing and process the
25761 * tree-ssa-alias.c (ao_ref_base_alias_set): Strip
25762 outer WITH_SIZE_EXPR.
25763 (ao_ref_base_alias_ptr_type): Likewise.
25764 (refs_may_alias_p_2): Allow WITH_SIZE_EXPR in ref->ref
25765 and handle that accordingly, stripping it for the
25766 core alias workers.
25767 * tree.c (get_base_address): Handle WITH_SIZE_EXPR by
25768 looking through it instead of returning NULL.
25770 2021-05-19 Jakub Jelinek <jakub@redhat.com>
25772 PR middle-end/100576
25773 * builtins.c (check_read_access): Convert bound to size_type_node if
25776 2021-05-19 Richard Biener <rguenther@suse.de>
25778 * tree-cfg.c (verify_types_in_gimple_min_lval): Inline...
25779 (verify_types_in_gimple_reference): ... here. Sanitize.
25780 (verify_gimple_call): Verify references in LHS and arguments.
25781 (verify_gimple_assign_single): Reject WITH_SIZE_EXPR.
25783 2021-05-19 Uroš Bizjak <ubizjak@gmail.com>
25785 * config/i386/i386.h (VALID_INT_MODE_P):
25786 Add V8QI, V4HI and V2SI modes for TARGET_64BIT.
25787 * config/i386/i386.md (isa): Add x64_bmi.
25788 (enabled): Handle x64_bmi.
25789 * config/i386/mmx.md (mmx_andnot<MMXMODEI:mode>3):
25790 Add alternative using 64bit general registers.
25791 (*mmx_<any_logic:code><MMXMODEI:mode>3): Ditto.
25793 2021-05-19 Jakub Jelinek <jakub@redhat.com>
25795 PR middle-end/99928
25796 * tree.h (OMP_MASTER_COMBINED): Define.
25797 * gimplify.c (gimplify_scan_omp_clauses): Rewrite lastprivate
25798 handling for outer combined/composite constructs to a loop.
25799 Handle lastprivate on combined target.
25800 (gimplify_expr): Formatting fix.
25802 2021-05-19 Xionghu Luo <luoxhu@linux.ibm.com>
25804 * passes.def: Add sink_code pass before store_merging.
25805 * tree-ssa-sink.c (pass_sink_code:clone): New.
25807 2021-05-18 Bill Schmidt <wschmidt@linux.ibm.com>
25809 * config/rs6000/freebsd64.h (ADJUST_FIELD_ALIGN): Remove call to
25810 rs6000_special_adjust_field_align_p.
25811 * config/rs6000/linux64.h (ADJUST_FIELD_ALIGN): Likewise.
25812 * config/rs6000/rs6000-call.c (rs6000_function_arg_boundary):
25813 Remove ABI warning.
25814 (rs6000_function_arg): Likewise.
25815 * config/rs6000/rs6000-protos.h
25816 (rs6000_special_adjust_field_align_p): Remove prototype.
25817 * config/rs6000/rs6000.c (rs6000_special_adjust_field_align_p):
25819 * config/rs6000/sysv4.h (ADJUST_FIELD_ALIGN): Remove call to
25820 rs6000_special_adjust_field_align_p.
25822 2021-05-18 Uroš Bizjak <ubizjak@gmail.com>
25825 * config/i386/i386.h (VALID_SSE2_REG_MODE):
25826 Add V4QI and V2HI modes.
25827 (VALID_INT_MODE_P): Ditto.
25828 * config/i386/mmx.md (VI_32): New mode iterator.
25829 (mmxvecsize): Handle V4QI and V2HI.
25831 (mov<VI_32:mode>): New expander.
25832 (*mov<mode>_internal): New insn pattern.
25833 (movmisalign<VI_32:mode>): New expander.
25834 (neg<VI_32:mode>): New expander.
25835 (<plusminus:insn><VI_32:mode>3): New expander.
25836 (*<plusminus:insn><VI_32:mode>3): New insn pattern.
25837 (mulv2hi3): New expander.
25838 (*mulv2hi3): New insn pattern.
25839 (one_cmpl<VI_32:mode>2): New expander.
25840 (*andnot<VI_32:mode>3): New insn pattern.
25841 (<any_logic:code><VI_32:mode>3): New expander.
25842 (*<any_logic:code><VI_32:mode>3): New insn pattern.
25844 2021-05-18 Uroš Bizjak <ubizjak@gmail.com>
25846 * config/i386/sse.md (<any_extend:insn>v4qiv4di2):
25847 Fix a mode mismatch with operand 1.
25849 2021-05-18 Uroš Bizjak <ubizjak@gmail.com>
25852 * config/i386/i386-expand.c (split_double_mode): Return
25853 temporary register when simplify_gen_subreg fails with
25854 the high half od the paradoxical subreg.
25856 2021-05-18 Richard Biener <rguenther@suse.de>
25858 * cfgexpand.c (expand_one_var): Pass in forced_stack_var
25859 and honor it when expanding.
25860 (expand_used_vars_for_block): Pass through forced_stack_var.
25861 (expand_used_vars): Likewise.
25862 (discover_nonconstant_array_refs_r): Set bits in
25863 forced_stack_vars instead of marking vars TREE_ADDRESSABLE.
25864 (avoid_type_punning_on_regs): Likewise.
25865 (discover_nonconstant_array_refs): Likewise.
25866 (pass_expand::execute): Create and pass down forced_stack_var
25867 bitmap. For parameters and returns temporarily set
25868 TREE_ADDRESSABLE when expand_function_start.
25870 2021-05-18 Thomas Schwinge <thomas@codesourcery.com>
25872 * doc/sourcebuild.texi: Document 'dg-note'.
25874 2021-05-18 Tobias Burnus <tobias@codesourcery.com>
25877 * configure: Regenerate.
25878 * configure.ac (BUILD_CFLAG, BUILD_CXXFLAGS): Add $(CFLAGS-$@).
25880 2021-05-18 Thomas Schwinge <thomas@codesourcery.com>
25882 * gimple.h (is_gimple_omp_oacc): Tighten.
25883 * omp-low.c (check_omp_nesting_restrictions): Adjust.
25885 2021-05-18 Richard Biener <rguenther@suse.de>
25887 * tree-ssa-operands.c (mark_address_taken): Simplify.
25889 2021-05-18 Martin Liska <mliska@suse.cz>
25891 * config/gcn/mkoffload.c (STR): Redefine.
25892 * config/i386/intelmic-mkoffload.c (STR): Likewise.
25893 * config/nvptx/mkoffload.c (STR): Likewise.
25895 2021-05-18 Martin Liska <mliska@suse.cz>
25897 * common/config/aarch64/aarch64-common.c (aarch64_parse_extension):
25898 Use startswith function instead of strncmp.
25899 * common/config/bfin/bfin-common.c (bfin_handle_option): Likewise.
25900 * common/config/riscv/riscv-common.c (riscv_subset_list::parse): Likewise.
25901 * config/aarch64/aarch64-sve-builtins-shapes.cc (parse_type): Likewise.
25902 * config/aarch64/aarch64.c (aarch64_process_one_target_attr): Likewise.
25903 * config/alpha/alpha.c (alpha_elf_section_type_flags): Likewise.
25904 * config/arm/aarch-common.c (arm_md_asm_adjust): Likewise.
25905 * config/arm/arm.c (arm_file_start): Likewise.
25906 (arm_valid_target_attribute_rec): Likewise.
25907 (thumb1_md_asm_adjust): Likewise.
25908 * config/arm/driver-arm.c (host_detect_local_cpu): Likewise.
25909 * config/avr/avr.c (STR_PREFIX_P): Likewise.
25910 (avr_set_current_function): Likewise.
25911 (avr_handle_addr_attribute): Likewise.
25912 (avr_asm_output_aligned_decl_common): Likewise.
25913 (avr_asm_named_section): Likewise.
25914 (avr_section_type_flags): Likewise.
25915 (avr_asm_select_section): Likewise.
25916 * config/c6x/c6x.c (c6x_in_small_data_p): Likewise.
25917 (c6x_section_type_flags): Likewise.
25918 * config/darwin-c.c (darwin_cfstring_ref_p): Likewise.
25919 (darwin_objc_declare_unresolved_class_reference): Likewise.
25920 (darwin_objc_declare_class_definition): Likewise.
25921 * config/darwin.c (indirect_data): Likewise.
25922 (darwin_encode_section_info): Likewise.
25923 (darwin_objc2_section): Likewise.
25924 (darwin_objc1_section): Likewise.
25925 (machopic_select_section): Likewise.
25926 (darwin_globalize_label): Likewise.
25927 (darwin_label_is_anonymous_local_objc_name): Likewise.
25928 (darwin_asm_named_section): Likewise.
25929 (darwin_asm_output_dwarf_offset): Likewise.
25930 * config/frv/frv.c (frv_string_begins_with): Likewise.
25931 (frv_in_small_data_p): Likewise.
25932 * config/gcn/mkoffload.c (STR): Likewise.
25934 * config/i386/i386-builtins.c (get_builtin_code_for_version): Likewise.
25935 * config/i386/i386-options.c (ix86_option_override_internal): Likewise.
25936 * config/i386/i386.c (x86_64_elf_section_type_flags): Likewise.
25937 (ix86_md_asm_adjust): Likewise.
25938 * config/i386/intelmic-mkoffload.c (STR): Likewise.
25939 * config/i386/winnt.c (i386_pe_asm_named_section): Likewise.
25940 (i386_pe_file_end): Likewise.
25941 * config/ia64/ia64.c (ia64_in_small_data_p): Likewise.
25942 (ia64_section_type_flags): Likewise.
25943 * config/mips/driver-native.c (host_detect_local_cpu): Likewise.
25944 * config/mips/mips.c (mips_handle_interrupt_attr): Likewise.
25945 (mips16_stub_function_p): Likewise.
25946 (mips_function_rodata_section): Likewise.
25947 * config/msp430/msp430.c (msp430_mcu_name): Likewise.
25948 (msp430_function_section): Likewise.
25949 (msp430_section_type_flags): Likewise.
25950 (msp430_expand_helper): Likewise.
25951 * config/nios2/nios2.c (nios2_small_section_name_p): Likewise.
25952 (nios2_valid_target_attribute_rec): Likewise.
25953 * config/nvptx/mkoffload.c (process): Likewise.
25955 * config/pa/som.h: Likewise.
25956 * config/pdp11/pdp11.c (pdp11_output_ident): Likewise.
25957 * config/riscv/riscv.c (riscv_elf_select_rtx_section): Likewise.
25958 * config/rs6000/rs6000.c (VTABLE_NAME_P): Likewise.
25959 (rs6000_inner_target_options): Likewise.
25960 * config/s390/driver-native.c (s390_host_detect_local_cpu): Likewise.
25961 * config/sparc/driver-sparc.c (host_detect_local_cpu): Likewise.
25962 * config/vax/vax.c (vax_output_int_move): Likewise.
25963 * config/vms/vms-ld.c (startswith): Likewise.
25964 (process_args): Likewise.
25966 * config/vms/vms.c: Likewise.
25968 2021-05-18 Jakub Jelinek <jakub@redhat.com>
25970 PR rtl-optimization/100590
25971 * regcprop.c (copyprop_hardreg_forward_1): Only DCE dead sets if
25972 they are NONJUMP_INSN_P.
25974 2021-05-18 Jakub Jelinek <jakub@redhat.com>
25977 * function.c (push_dummy_function): Set DECL_ARTIFICIAL and
25978 DECL_ASSEMBLER_NAME on the fn_decl.
25980 2021-05-18 Jakub Jelinek <jakub@redhat.com>
25982 PR tree-optimization/94589
25983 * tree-ssa-phiopt.c (spaceship_replacement): Pattern match
25984 phi result used in (res & ~1) == 0 comparison as res >= 0 as
25985 res == 2 would be UB with -ffinite-math-only.
25987 2021-05-18 Martin Liska <mliska@suse.cz>
25989 * Makefile.in: genversion.o should depend on DATESTAMP.
25991 2021-05-18 Claudiu Zissulescu <claziss@synopsys.com>
25993 * config/arc/simdext.md (negv2si2): Remove round bracket.
25995 2021-05-18 Andreas Krebbel <krebbel@linux.ibm.com>
25997 * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal): Define
25998 _Bool as macro expanding to _Bool.
26000 2021-05-18 Andreas Krebbel <krebbel@linux.ibm.com>
26003 * tree.c (build_reference_type_for_mode)
26004 (build_pointer_type_for_mode): Pick pointer mode if MODE argument
26006 (build_reference_type, build_pointer_type): Invoke
26007 build_*_type_for_mode with VOIDmode.
26009 2021-05-17 Andrew MacLeod <amacleod@redhat.com>
26011 PR tree-optimization/100512
26012 * gimple-range-cache.cc (ranger_cache::set_global_range): Mark const
26013 and non-zero pointer ranges as invariant.
26014 * gimple-range.cc (gimple_ranger::range_of_stmt): Remove pointer
26015 processing from here.
26017 2021-05-17 Tom de Vries <tdevries@suse.de>
26020 * config/nvptx/nvptx-protos.h (nvptx_output_atomic_insn): Declare
26021 * config/nvptx/nvptx.c (nvptx_output_barrier)
26022 (nvptx_output_atomic_insn): New function.
26023 (nvptx_print_operand): Add support for 'B'.
26024 * config/nvptx/nvptx.md: Use nvptx_output_atomic_insn for atomic
26027 2021-05-17 Aldy Hernandez <aldyh@redhat.com>
26029 PR tree-optimization/100349
26030 * vr-values.c (bounds_of_var_in_loop): Bail if scev returns
26033 2021-05-17 Tamar Christina <tamar.christina@arm.com>
26035 * config/aarch64/driver-aarch64.c (DEFAULT_ARCH): New.
26036 (host_detect_local_cpu): Use it.
26038 2021-05-17 Martin Liska <mliska@suse.cz>
26040 * doc/invoke.texi: Add 2 missing dots.
26042 2021-05-17 Marius Hillenbrand <mhillen@linux.ibm.com>
26044 PR bootstrap/100552
26045 * configure.ac: Replace pattern substitution with call to sed.
26046 * configure: Regenerate.
26048 2021-05-17 Richard Biener <rguenther@suse.de>
26050 PR middle-end/100582
26051 * tree.c (array_at_struct_end_p): Get to the base of the
26052 reference before looking for the underlying decl.
26054 2021-05-17 Joern Rennecke <joern.rennecke@embecosm.com>
26056 * genoutput.c (validate_insn_alternatives) Make "wrong number of
26057 alternatives" message more specific, and remove assumption on where
26060 2021-05-17 Christophe Lyon <christophe.lyon@linaro.org>
26062 * config/arm/iterators.md (V16): New iterator.
26063 (VH_cvtto): New iterator.
26064 (v_cmp_result): Added V4HF and V8HF support.
26065 * config/arm/vec-common.md (vec_cmp<mode><v_cmp_result>): Use VDQWH.
26066 (vcond<mode><mode>): Likewise.
26067 (vcond_mask_<mode><v_cmp_result>): Likewise.
26068 (vcond<VH_cvtto><mode>): New expander.
26070 2021-05-17 Christophe Lyon <christophe.lyon@linaro.org>
26072 * config/arm/arm-protos.h (arm_expand_vector_compare): Update
26074 * config/arm/arm.c (arm_expand_vector_compare): Add support for
26076 (arm_expand_vcond): Likewise.
26077 * config/arm/iterators.md (supf): Remove VCMPNEQ_S, VCMPEQQ_S,
26078 VCMPEQQ_N_S, VCMPNEQ_N_S.
26079 (VCMPNEQ, VCMPEQQ, VCMPEQQ_N, VCMPNEQ_N): Remove.
26080 * config/arm/mve.md (@mve_vcmp<mve_cmp_op>q_<mode>): Add '@' prefix.
26081 (@mve_vcmp<mve_cmp_op>q_f<mode>): Likewise.
26082 (@mve_vcmp<mve_cmp_op>q_n_f<mode>): Likewise.
26083 (@mve_vpselq_<supf><mode>): Likewise.
26084 (@mve_vpselq_f<mode>"): Likewise.
26085 * config/arm/neon.md (vec_cmp<mode><v_cmp_result): Enable for MVE
26086 and move to vec-common.md.
26087 (vec_cmpu<mode><mode>): Likewise.
26088 (vcond<mode><mode>): Likewise.
26089 (vcond<V_cvtto><mode>): Likewise.
26090 (vcondu<mode><v_cmp_result>): Likewise.
26091 (vcond_mask_<mode><v_cmp_result>): Likewise.
26092 * config/arm/unspecs.md (VCMPNEQ_U, VCMPNEQ_S, VCMPEQQ_S)
26093 (VCMPEQQ_N_S, VCMPNEQ_N_S, VCMPEQQ_U, CMPEQQ_N_U, VCMPNEQ_N_U)
26094 (VCMPGEQ_N_S, VCMPGEQ_S, VCMPGTQ_N_S, VCMPGTQ_S, VCMPLEQ_N_S)
26095 (VCMPLEQ_S, VCMPLTQ_N_S, VCMPLTQ_S, VCMPCSQ_N_U, VCMPCSQ_U)
26096 (VCMPHIQ_N_U, VCMPHIQ_U): Remove.
26097 * config/arm/vec-common.md (vec_cmp<mode><v_cmp_result): Moved
26099 (vec_cmpu<mode><mode>): Likewise.
26100 (vcond<mode><mode>): Likewise.
26101 (vcond<V_cvtto><mode>): Likewise.
26102 (vcondu<mode><v_cmp_result>): Likewise.
26103 (vcond_mask_<mode><v_cmp_result>): Likewise. Added unsafe math
26106 2021-05-17 liuhongt <hongtao.liu@intel.com>
26109 * config/i386/i386.c (ix86_gimple_fold_builtin): Use
26110 gsi_insert_seq_before instead.
26112 2021-05-17 Christophe Lyon <christophe.lyon@linaro.org>
26114 * doc/sourcebuild.texi (arm_qbit_ok): Rename into...
26115 (arm_sat_ok): ...this.
26117 2021-05-17 Martin Liska <mliska@suse.cz>
26119 * lto-wrapper.c (merge_flto_options): Factor out a new function.
26120 (merge_and_complain): Use it.
26121 (run_gcc): Merge also linker command line -flto=foo argument
26124 2021-05-16 Christophe Lyon <christophe.lyon@linaro.org>
26126 * config/arm/arm.h (CPP_SPEC): Remove error message about
26127 -mlittle-endian/-mbig-endian conflict.
26129 2021-05-15 Bill Schmidt <wschmidt@linux.ibm.com>
26131 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Define
26132 __ROP_PROTECT__ if -mrop-protect is selected.
26134 2021-05-15 Bill Schmidt <wschmidt@linux.ibm.com>
26136 * config/rs6000/rs6000-internal.h (rs6000_stack): Add
26137 rop_hash_save_offset and rop_hash_size.
26138 * config/rs6000/rs6000-logue.c (rs6000_stack_info): Compute
26139 rop_hash_size and rop_hash_save_offset.
26140 (debug_stack_info): Dump rop_hash_save_offset and rop_hash_size.
26141 (rs6000_emit_prologue): Emit hashst[p] in prologue.
26142 (rs6000_emit_epilogue): Emit hashchk[p] in epilogue.
26143 * config/rs6000/rs6000.md (unspec): Add UNSPEC_HASHST and
26145 (hashst): New define_insn.
26146 (hashchk): Likewise.
26148 2021-05-15 Bill Schmidt <wschmidt@linux.ibm.com>
26150 * config/rs6000/rs6000.c (rs6000_option_override_internal):
26151 Disable shrink wrap when inserting ROP-protect instructions.
26152 * config/rs6000/rs6000.opt (mrop-protect): New option.
26153 (mprivileged): Likewise.
26154 * doc/invoke.texi: Document mrop-protect and mprivileged.
26156 2021-05-15 Hans-Peter Nilsson <hp@axis.com>
26158 * reorg.c (fill_slots_from_thread): Reinstate code typoed out in
26161 2021-05-15 Martin Jambor <mjambor@suse.cz>
26164 2021-05-13 Martin Jambor <mjambor@suse.cz>
26166 PR tree-optimization/100453
26167 * tree-sra.c (sra_modify_assign): All const base accesses do not
26168 need refreshing, not just those from decl_pool.
26169 (sra_modify_assign): Do not refresh into a const base decl.
26171 2021-05-15 Jakub Jelinek <jakub@redhat.com>
26173 PR rtl-optimization/100342
26174 * regcprop.c (copy_value): When copying a source reg in a wider
26175 mode than it has recorded for the value, adjust recorded destination
26176 mode too or punt if !REG_CAN_CHANGE_MODE_P.
26178 2021-05-14 Jason Merrill <jason@redhat.com>
26180 * intl.h: Add comments.
26182 2021-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
26184 * config/aarch64/aarch64-simd.md
26185 (aarch64_sqdml<SBINQOPS:as>l2_lane<mode>_internal): Split into...
26186 (aarch64_sqdmlsl2_lane<mode>_internal): ... This...
26187 (aarch64_sqdmlal2_lane<mode>_internal): ... And this.
26188 (aarch64_sqdml<SBINQOPS:as>l2_laneq<mode>_internal): Split into ...
26189 (aarch64_sqdmlsl2_laneq<mode>_internal): ... This...
26190 (aarch64_sqdmlal2_laneq<mode>_internal): ... And this.
26191 (aarch64_sqdml<SBINQOPS:as>l2_n<mode>_internal): Split into...
26192 (aarch64_sqdmlsl2_n<mode>_internal): ... This...
26193 (aarch64_sqdmlal2_n<mode>_internal): ... And this.
26195 2021-05-14 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
26198 * config/arm/arm_neon.h (vtst_s8): Replace call to vtst builtin with it's
26199 boolean logic equivalent.
26200 (vtst_s16): Likewise.
26201 (vtst_s32): Likewise.
26202 (vtst_u8): Likewise.
26203 (vtst_u16): Likewise.
26204 (vtst_u32): Likewise.
26205 (vtst_p8): Likewise.
26206 (vtst_p16): Likewise.
26207 (vtstq_s8): Likewise.
26208 (vtstq_s16): Likewise.
26209 (vtstq_s32): Likewise.
26210 (vtstq_u8): Likewise.
26211 (vtstq_u16): Likewise.
26212 (vtstq_u32): Likewise.
26213 (vtstq_p8): Likewise.
26214 (vtstq_p16): Likewise.
26215 * config/arm/arm_neon_builtins.def: Remove entry for vtst.
26216 * config/arm/neon.md (neon_vtst<mode>): Remove pattern.
26218 2021-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
26220 * config/aarch64/aarch64-simd.md (aarch64_sqdmlal2<mode>): Merge into...
26221 (aarch64_sqdml<SBINQOPS:as>l2<mode>): ... This.
26222 (aarch64_sqdmlsl2<mode>): Delete.
26223 (aarch64_sqdmlal2_lane<mode>): Merge this...
26224 (aarch64_sqdmlsl2_lane<mode>): ... And this...
26225 (aarch64_sqdml<SBINQOPS:as>l2_lane<mode>): ... Into this.
26226 (aarch64_sqdmlal2_laneq<mode>): Merge this...
26227 (aarch64_sqdmlsl2_laneq<mode>): ... And this...
26228 (aarch64_sqdml<SBINQOPS:as>l2_laneq<mode>): ... Into this.
26229 (aarch64_sqdmlal2_n<mode>): Merge this...
26230 (aarch64_sqdmlsl2_n<mode>): ... And this...
26231 (aarch64_sqdml<SBINQOPS:as>l2_n<mode>): ... Into this.
26233 2021-05-13 Martin Sebor <msebor@redhat.com>
26235 PR middle-end/100574
26236 * builtins.c (access_ref::get_ref): Improve detection of PHIs with
26237 all null arguments.
26239 2021-05-13 Martin Sebor <msebor@redhat.com>
26241 PR tree-optimization/93100
26242 PR middle-end/98583
26243 * tree-ssa-uninit.c (check_defs): Exclude intrinsic functions that
26244 don't modify referenced objects.
26246 2021-05-13 Martin Jambor <mjambor@suse.cz>
26248 PR tree-optimization/100453
26249 * tree-sra.c (sra_modify_assign): All const base accesses do not
26250 need refreshing, not just those from decl_pool.
26251 (sra_modify_assign): Do not refresh into a const base decl.
26253 2021-05-13 Martin Liska <mliska@suse.cz>
26255 * tree-ssa-dom.c: Remove m_simplifier.
26257 2021-05-13 Richard Earnshaw <rearnsha@arm.com>
26260 * config/arm/arm.c (arm_canonicalize_comparison): Correctly
26261 canonicalize DImode inequality comparisons against the
26262 maximum integral value.
26264 2021-05-13 Jakub Jelinek <jakub@redhat.com>
26266 PR tree-optimization/98856
26267 * config/i386/i386.c (ix86_shift_rotate_cost): Add CODE argument.
26268 Expect V2DI and V4DI arithmetic right shifts to be emulated.
26269 (ix86_rtx_costs, ix86_add_stmt_cost): Adjust ix86_shift_rotate_cost
26271 * config/i386/i386-expand.c (expand_vec_perm_2perm_interleave,
26272 expand_vec_perm_2perm_pblendv): New functions.
26273 (ix86_expand_vec_perm_const_1): Use them.
26274 * config/i386/sse.md (ashr<mode>3<mask_name>): Rename to ...
26275 (<mask_codefor>ashr<mode>3<mask_name>): ... this.
26276 (ashr<mode>3): New define_expand with VI248_AVX512BW iterator.
26277 (ashrv4di3): New define_expand.
26278 (ashrv2di3): Change condition to TARGET_SSE2, handle !TARGET_XOP
26279 and !TARGET_AVX512VL expansion.
26281 2021-05-13 Uroš Bizjak <ubizjak@gmail.com>
26284 * config/i386/i386-expand.c (ix86_expand_sse_movcc): Force mode
26285 sizes < 16 to a register when constructing vpcmov pattern.
26286 * config/i386/mmx.md (*xop_pcmov_<mode>): Use MMXMODE124 mode.
26288 2021-05-13 Martin Liska <mliska@suse.cz>
26290 * gcov-io.c (gcov_write_block): Remove.
26291 (gcov_write_words): Likewise.
26292 (gcov_read_words): Re-implement using gcov_read_bytes.
26293 (gcov_allocate): Remove.
26294 (GCOV_BLOCK_SIZE): Likewise.
26295 (struct gcov_var): Remove most of the fields.
26296 (gcov_position): Implement with ftell.
26297 (gcov_rewrite): Remove setting of start and offset fields.
26298 (from_file): Re-format.
26299 (gcov_open): Remove setbuf call. It should not be needed.
26300 (gcov_close): Remove internal buffer handling.
26301 (gcov_magic): Use __builtin_bswap32.
26302 (gcov_write_counter): Use directly gcov_write_unsigned.
26303 (gcov_write_string): Use direct fwrite and do not round
26305 (gcov_seek): Use directly fseek.
26306 (gcov_write_tag): Use gcov_write_unsigned directly.
26307 (gcov_write_length): Likewise.
26308 (gcov_write_tag_length): Likewise.
26309 (gcov_read_bytes): Use directly fread.
26310 (gcov_read_unsigned): Use gcov_read_words.
26311 (gcov_read_counter): Likewise.
26312 (gcov_read_string): Use gcov_read_bytes.
26313 * gcov-io.h (GCOV_WORD_SIZE): Adjust to reflect
26314 that size is not in bytes, but words (4B).
26315 (GCOV_TAG_FUNCTION_LENGTH): Likewise.
26316 (GCOV_TAG_ARCS_LENGTH): Likewise.
26317 (GCOV_TAG_ARCS_NUM): Likewise.
26318 (GCOV_TAG_COUNTER_LENGTH): Likewise.
26319 (GCOV_TAG_COUNTER_NUM): Likewise.
26320 (GCOV_TAG_SUMMARY_LENGTH): Likewise.
26322 2021-05-13 liuhongt <hongtao.liu@intel.com>
26325 * config/i386/sse.md (ssedoublevecmode): Add attribute for
26326 V64QI/V32HI/V16SI/V4DI.
26327 (ssehalfvecmode): Add attribute for V2DI/V2DF.
26328 (*vec_concatv4si_0): Extend to VI124_128.
26329 (*vec_concat<mode>_0): New pre-reload splitter.
26330 * config/i386/predicates.md (movq_parallel): New predicate.
26332 2021-05-13 Alexandre Oliva <oliva@adacore.com>
26334 * targhooks.c (default_zero_call_used_regs): Retry using
26335 successfully-zeroed registers as sources.
26337 2021-05-12 Tobias Burnus <tobias@codesourcery.com>
26339 * omp-low.c (finish_taskreg_scan): Use the proper detach decl.
26341 2021-05-12 Aldy Hernandez <aldyh@redhat.com>
26344 * gimple-range.cc (range_of_builtin_call): Skip out on
26345 processing __builtin_clz when varying.
26347 2021-05-12 Tom de Vries <tdevries@suse.de>
26350 * config/nvptx/nvptx-opts.h (enum ptx_version): New enum.
26351 * config/nvptx/nvptx.c (nvptx_file_start): Print .version according
26352 to ptx_version_option.
26353 * config/nvptx/nvptx.h (TARGET_PTX_6_3): Define.
26354 * config/nvptx/nvptx.md (define_insn "nvptx_shuffle<mode>")
26355 (define_insn "nvptx_vote_ballot"): Use sync variant for
26357 * config/nvptx/nvptx.opt (ptx_version): Add enum.
26358 (mptx): Add option.
26359 * doc/invoke.texi (Nvidia PTX Options): Add mptx item.
26361 2021-05-12 Richard Biener <rguenther@suse.de>
26363 PR tree-optimization/100566
26364 * tree-ssa-sccvn.c (dominated_by_p_w_unex): Properly handle
26365 allow_back for all edge queries.
26367 2021-05-12 liuhongt <hongtao.liu@intel.com>
26370 * config/i386/sse.md (<sse4_1_avx2>_pblendvb): Add
26371 splitters for pblendvb of NOT mask register.
26373 2021-05-12 Richard Biener <rguenther@suse.de>
26375 PR tree-optimization/100519
26376 * tree-ssa-reassoc.c (can_associate_p): Split into...
26377 (can_associate_op_p): ... this
26378 (can_associate_type_p): ... and this.
26379 (is_reassociable_op): Call can_associate_op_p.
26380 (break_up_subtract_bb): Call the appropriate predicates.
26381 (reassociate_bb): Likewise.
26383 2021-05-12 Martin Liska <mliska@suse.cz>
26385 * lto-wrapper.c (merge_and_complain): Merge -flto=arg options.
26386 (run_gcc): Use -flto argument detection for merged
26389 2021-05-12 Martin Liska <mliska@suse.cz>
26391 * lto-wrapper.c (print_lto_docs_link): New function.
26392 (run_gcc): Print warning about missing job server detection
26393 after we know NR of partitions. Do the same for -flto{,=1}.
26394 * opts.c (get_option_html_page): Support -flto option.
26396 2021-05-12 Martin Liska <mliska@suse.cz>
26398 * lto-wrapper.c (get_options_from_collect_gcc_options): Change
26400 (append_option): Remove.
26401 (find_option): Rework to use the vector type.
26402 (remove_option): Remove.
26403 (merge_and_complain): Use vectors for cl_decoded_option data
26405 (append_compiler_options): Likewise.
26406 (append_diag_options): Likewise.
26407 (append_linker_options): Likewise.
26408 (append_offload_options): Likewise.
26409 (compile_offload_image): Likewise.
26410 (compile_images_for_offload_targets): Likewise.
26411 (find_and_merge_options): Likewise.
26412 (run_gcc): Likewise.
26414 2021-05-12 Bernd Edlinger <bernd.edlinger@hotmail.de>
26417 * dwarf2out.c (dwarf2out_finish): Set
26418 have_multiple_function_sections with multi-range text_section.
26420 2021-05-12 Martin Liska <mliska@suse.cz>
26422 PR bootstrap/100560
26423 * Makefile.in: Remove version.h from linker command line.
26425 2021-05-12 Richard Biener <rguenther@suse.de>
26427 PR middle-end/100547
26428 * rtl.h (rtvec_alloc): Make argument size_t.
26429 * rtl.c (rtvec_alloc): Verify the count is less than INT_MAX.
26431 2021-05-12 Jakub Jelinek <jakub@redhat.com>
26433 PR middle-end/100508
26434 * cfgexpand.c (expand_debug_expr): For DEBUG_EXPR_DECL with vector
26435 type, don't reuse DECL_RTL if it has different mode, instead force
26436 creation of a new DEBUG_EXPR.
26438 2021-05-12 Jakub Jelinek <jakub@redhat.com>
26439 Marc Glisse <marc.glisse@inria.fr>
26441 PR tree-optimization/94589
26442 * match.pd ((X & Y) == X -> (X & ~Y) == 0,
26443 (X | Y) == Y -> (X & ~Y) == 0): New GIMPLE simplifications.
26445 2021-05-12 Uroš Bizjak <ubizjak@gmail.com>
26448 * config/i386/i386-expand.c (ix86_expand_sse_movcc): Handle V2SF mode.
26449 * config/i386/mmx.md (MMXMODE124): New mode iterator.
26451 (mmxintvecmode): New mode attribute.
26452 (mmxintvecmodelower): Ditto.
26453 (*mmx_maskcmpv2sf3_comm): New insn pattern.
26454 (*mmx_maskcmpv2sf3): Ditto.
26455 (vec_cmpv2sfv2si): New expander.
26456 (vcond<V2FI:mode>v2si): Ditto.
26457 (mmx_vlendvps): New insn pattern.
26458 (vcond<MMXMODE124:mode><MMXMODEI:mode>): Also handle V2SFmode.
26459 (vcondu<MMXMODE124:mode><MMXMODEI:mode>): Ditto.
26460 (vcond_mask_<mode><mmxintvecmodelower>): Ditto.
26462 2021-05-11 Martin Sebor <msebor@redhat.com>
26464 PR middle-end/21433
26465 * expr.c (expand_expr_real_1): Replace unreachable code with an assert.
26467 2021-05-11 Richard Biener <rguenther@suse.de>
26469 * gimple-fold.c (gimple_fold_call): Do not call
26470 maybe_fold_reference on call arguments or the static chain.
26471 (fold_stmt_1): Do not call maybe_fold_reference on GIMPLE_ASM
26474 2021-05-11 Martin Liska <mliska@suse.cz>
26476 * builtins.def (DEF_HSAIL_BUILTIN): Remove.
26477 (DEF_HSAIL_ATOMIC_BUILTIN): Likewise.
26478 (DEF_HSAIL_SAT_BUILTIN): Likewise.
26479 (DEF_HSAIL_INTR_BUILTIN): Likewise.
26480 (DEF_HSAIL_CVT_ZEROI_SAT_BUILTIN): Likewise.
26481 * doc/frontends.texi: Remove BRIG.
26482 * doc/install.texi: Likewise.
26483 * doc/invoke.texi: Likewise.
26484 * doc/standards.texi: Likewise.
26485 * brig-builtins.def: Removed.
26486 * brig/ChangeLog: Removed.
26487 * brig/Make-lang.in: Removed.
26488 * brig/brig-builtins.h: Removed.
26489 * brig/brig-c.h: Removed.
26490 * brig/brig-lang.c: Removed.
26491 * brig/brigfrontend/brig-arg-block-handler.cc: Removed.
26492 * brig/brigfrontend/brig-atomic-inst-handler.cc: Removed.
26493 * brig/brigfrontend/brig-basic-inst-handler.cc: Removed.
26494 * brig/brigfrontend/brig-branch-inst-handler.cc: Removed.
26495 * brig/brigfrontend/brig-cmp-inst-handler.cc: Removed.
26496 * brig/brigfrontend/brig-code-entry-handler.cc: Removed.
26497 * brig/brigfrontend/brig-code-entry-handler.h: Removed.
26498 * brig/brigfrontend/brig-comment-handler.cc: Removed.
26499 * brig/brigfrontend/brig-control-handler.cc: Removed.
26500 * brig/brigfrontend/brig-copy-move-inst-handler.cc: Removed.
26501 * brig/brigfrontend/brig-cvt-inst-handler.cc: Removed.
26502 * brig/brigfrontend/brig-fbarrier-handler.cc: Removed.
26503 * brig/brigfrontend/brig-function-handler.cc: Removed.
26504 * brig/brigfrontend/brig-function.cc: Removed.
26505 * brig/brigfrontend/brig-function.h: Removed.
26506 * brig/brigfrontend/brig-inst-mod-handler.cc: Removed.
26507 * brig/brigfrontend/brig-label-handler.cc: Removed.
26508 * brig/brigfrontend/brig-lane-inst-handler.cc: Removed.
26509 * brig/brigfrontend/brig-machine.c: Removed.
26510 * brig/brigfrontend/brig-machine.h: Removed.
26511 * brig/brigfrontend/brig-mem-inst-handler.cc: Removed.
26512 * brig/brigfrontend/brig-module-handler.cc: Removed.
26513 * brig/brigfrontend/brig-queue-inst-handler.cc: Removed.
26514 * brig/brigfrontend/brig-seg-inst-handler.cc: Removed.
26515 * brig/brigfrontend/brig-signal-inst-handler.cc: Removed.
26516 * brig/brigfrontend/brig-to-generic.cc: Removed.
26517 * brig/brigfrontend/brig-to-generic.h: Removed.
26518 * brig/brigfrontend/brig-util.cc: Removed.
26519 * brig/brigfrontend/brig-util.h: Removed.
26520 * brig/brigfrontend/brig-variable-handler.cc: Removed.
26521 * brig/brigfrontend/hsa-brig-format.h: Removed.
26522 * brig/brigfrontend/phsa.h: Removed.
26523 * brig/brigspec.c: Removed.
26524 * brig/config-lang.in: Removed.
26525 * brig/gccbrig.texi: Removed.
26526 * brig/lang-specs.h: Removed.
26527 * brig/lang.opt: Removed.
26529 2021-05-11 Richard Biener <rguenther@suse.de>
26532 * ipa-param-manipulation.c
26533 (ipa_param_body_adjustments::modify_call_stmt): Avoid
26534 altering SSA_NAME_DEF_STMT by adjusting the calls LHS
26535 via gimple_call_lhs_ptr.
26537 2021-05-11 Alex Coplan <alex.coplan@arm.com>
26540 * config/arm/arm.c (cmse_nonsecure_call_inline_register_clear):
26541 Avoid emitting CFA adjusts on the sp if we have the fp.
26543 2021-05-11 Richard Sandiford <richard.sandiford@arm.com>
26545 * config/aarch64/iterators.md (VMUL_CHANGE_NLANES): Delete.
26546 (VMULD): New iterator.
26547 (VCOND): Handle V4HF and V8HF.
26548 (VCONQ): Fix entry for V2SF.
26549 * config/aarch64/aarch64-simd.md (mul_lane<mode>3): Use VMULD
26550 instead of VMUL. Use a 64-bit vector mode for the indexed operand.
26551 (*aarch64_mul3_elt_<vswap_width_name><mode>): Merge with...
26552 (mul_laneq<mode>3): ...this define_insn. Use VMUL instead of VDQSF.
26553 Use a 128-bit vector mode for the indexed operand. Use stype for
26554 the scheduling type.
26556 2021-05-11 Richard Biener <rguenther@suse.de>
26558 * gimple-fold.c (maybe_fold_reference): Only return
26559 is_gimple_min_invariant values.
26561 2021-05-11 Richard Biener <rguenther@suse.de>
26563 PR middle-end/100509
26564 * gimple-fold.c (fold_gimple_assign): Only call
26565 get_symbol_constant_value on register type symbols.
26567 2021-05-11 Srinath Parvathaneni <srinath.parvathaneni@arm.com>
26568 Joe Ramsay <joe.ramsay@arm.com>
26571 * config/arm/arm_mve.h (__arm_vstrwq_scatter_offset): Fix wrong arguments.
26572 (__arm_vcmpneq): Remove duplicate definition.
26573 (__arm_vstrwq_scatter_offset_p): Likewise.
26574 (__arm_vmaxq_x): Likewise.
26575 (__arm_vmlsdavaq): Likewise.
26576 (__arm_vmlsdavaxq): Likewise.
26577 (__arm_vmlsdavq_p): Likewise.
26578 (__arm_vmlsdavxq_p): Likewise.
26579 (__arm_vrmlaldavhaq): Likewise.
26580 (__arm_vstrbq_p): Likewise.
26581 (__arm_vstrbq_scatter_offset): Likewise.
26582 (__arm_vstrbq_scatter_offset_p): Likewise.
26583 (__arm_vstrdq_scatter_offset): Likewise.
26584 (__arm_vstrdq_scatter_offset_p): Likewise.
26585 (__arm_vstrdq_scatter_shifted_offset): Likewise.
26586 (__arm_vstrdq_scatter_shifted_offset_p): Likewise.
26588 2021-05-11 Jakub Jelinek <jakub@redhat.com>
26590 PR middle-end/100471
26591 * omp-low.c (lower_omp_task_reductions): For OMP_TASKLOOP, if data
26592 is 0, bypass the reduction loop including
26593 GOMP_taskgroup_reduction_unregister call.
26595 2021-05-11 Kewen Lin <linkw@linux.ibm.com>
26597 * config/rs6000/rs6000.c (struct rs6000_cost_data): New member
26598 costing_for_scalar.
26599 (rs6000_density_test): Early return if costing_for_scalar is true.
26600 (rs6000_init_cost): Init costing_for_scalar of rs6000_cost_data.
26602 2021-05-11 Kewen Lin <linkw@linux.ibm.com>
26604 * doc/tm.texi: Regenerated.
26605 * target.def (init_cost): Add new parameter costing_for_scalar.
26606 * targhooks.c (default_init_cost): Adjust for new parameter.
26607 * targhooks.h (default_init_cost): Likewise.
26608 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Likewise.
26609 (vect_compute_single_scalar_iteration_cost): Likewise.
26610 (vect_analyze_loop_2): Likewise.
26611 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Likewise.
26612 (vect_bb_vectorization_profitable_p): Likewise.
26613 * tree-vectorizer.h (init_cost): Likewise.
26614 * config/aarch64/aarch64.c (aarch64_init_cost): Likewise.
26615 * config/i386/i386.c (ix86_init_cost): Likewise.
26616 * config/rs6000/rs6000.c (rs6000_init_cost): Likewise.
26618 2021-05-11 Kewen Lin <linkw@linux.ibm.com>
26620 * config/rs6000/rs6000.c (rs6000_vect_nonmem): Renamed to
26621 vect_nonmem and moved into...
26622 (struct rs6000_cost_data): ...here.
26623 (rs6000_init_cost): Use vect_nonmem of cost_data instead.
26624 (rs6000_add_stmt_cost): Likewise.
26625 (rs6000_finish_cost): Likewise.
26627 2021-05-10 Eric Botcazou <ebotcazou@adacore.com>
26629 * range-op.cc (get_bool_state): Adjust head comment.
26630 (operator_not_equal::op1_range): Fix comment.
26631 (operator_bitwise_xor::op1_range): Remove call to gcc_unreachable.
26633 2021-05-10 Martin Sebor <msebor@redhat.com>
26635 PR middle-end/100425
26636 PR middle-end/100510
26637 * gimple-ssa-warn-alloca.c (pass_walloca::firast_time_p): Rename...
26638 (pass_walloca::xlimit_certain_p): ...to this.
26639 (pass_walloca::gate): Execute for any kind of handled warning.
26640 (pass_walloca::execute): Avoid issuing "maybe" and "unbounded"
26641 warnings when xlimit_certain_p is set.
26643 2021-05-10 Pat Haugen <pthaugen@linux.ibm.com>
26645 * config/rs6000/rs6000.c (rs6000_ira_change_pseudo_allocno_class):
26646 Return ALTIVEC_REGS if that is best_class.
26647 (rs6000_compute_pressure_classes): Add ALTIVEC_REGS.
26649 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
26651 * config/arm/arm.h (CPP_SPEC): Remove error message about
26654 2021-05-10 Martin Jambor <mjambor@suse.cz>
26656 * ipa-prop.h (IPA_NODE_REF): Removed.
26657 (IPA_NODE_REF_GET_CREATE): Likewise.
26658 (IPA_EDGE_REF): Likewise.
26659 (IPA_EDGE_REF_GET_CREATE): Likewise.
26660 (IS_VALID_JUMP_FUNC_INDEX): Likewise.
26661 * ipa-cp.c (print_all_lattices): Replaced IPA_NODE_REF with a direct
26662 use of ipa_node_params_sum.
26663 (ipcp_versionable_function_p): Likewise.
26664 (push_node_to_stack): Likewise.
26665 (pop_node_from_stack): Likewise.
26666 (set_single_call_flag): Replaced two IPA_NODE_REF with one single
26667 direct use of ipa_node_params_sum.
26668 (initialize_node_lattices): Replaced IPA_NODE_REF with a direct use of
26669 ipa_node_params_sum.
26670 (ipa_context_from_jfunc): Replaced IPA_EDGE_REF with a direct use of
26672 (ipcp_verify_propagated_values): Replaced IPA_NODE_REF with a direct
26673 use of ipa_node_params_sum.
26674 (self_recursively_generated_p): Likewise.
26675 (propagate_scalar_across_jump_function): Likewise.
26676 (propagate_context_across_jump_function): Replaced IPA_EDGE_REF with a
26677 direct use of ipa_edge_args_sum, moved the lookup after the early
26678 exit. Replaced IPA_NODE_REF with a direct use of ipa_node_params_sum.
26679 (propagate_bits_across_jump_function): Replaced IPA_NODE_REF with
26680 direct uses of ipa_node_params_sum.
26681 (propagate_vr_across_jump_function): Likewise.
26682 (propagate_aggregate_lattice): Likewise.
26683 (propagate_aggs_across_jump_function): Likewise.
26684 (propagate_constants_across_call): Likewise, also replaced
26685 IPA_EDGE_REF with a direct use of ipa_edge_args_sum.
26686 (good_cloning_opportunity_p): Replaced IPA_NODE_REF with a direct use
26687 of ipa_node_params_sum.
26688 (estimate_local_effects): Likewise.
26689 (add_all_node_vals_to_toposort): Likewise.
26690 (propagate_constants_topo): Likewise.
26691 (ipcp_propagate_stage): Likewise.
26692 (ipcp_discover_new_direct_edges): Likewise.
26693 (calls_same_node_or_its_all_contexts_clone_p): Likewise.
26694 (cgraph_edge_brings_value_p): Likewise (in both overloaded functions).
26695 (get_info_about_necessary_edges): Likewise.
26696 (want_remove_some_param_p): Likewise.
26697 (create_specialized_node): Likewise.
26698 (self_recursive_pass_through_p): Likewise.
26699 (self_recursive_agg_pass_through_p): Likewise.
26700 (find_more_scalar_values_for_callers_subset): Likewise and also
26701 replaced IPA_EDGE_REF with direct uses of ipa_edge_args_sum, in one
26702 case replacing two of those with a single query.
26703 (find_more_contexts_for_caller_subset): Likewise for the
26704 ipa_polymorphic_call_context overload.
26705 (intersect_aggregates_with_edge): Replaced IPA_EDGE_REF with a direct
26706 use of ipa_edge_args_sum. Replaced IPA_NODE_REF with direct uses of
26707 ipa_node_params_sum.
26708 (find_aggregate_values_for_callers_subset): Likewise, also reusing
26709 results of ipa_edge_args_sum->get.
26710 (cgraph_edge_brings_all_scalars_for_node): Replaced IPA_NODE_REF with
26711 direct uses of ipa_node_params_sum, replaced IPA_EDGE_REF with a
26712 direct use of ipa_edge_args_sum.
26713 (cgraph_edge_brings_all_agg_vals_for_node): Likewise, moved node
26714 summary query after the early exit and reused the result later.
26715 (decide_about_value): Replaced IPA_NODE_REF with a direct use of
26716 ipa_node_params_sum.
26717 (decide_whether_version_node): Likewise. Removed re-querying for
26718 summaries after cloning.
26719 (spread_undeadness): Replaced IPA_NODE_REF with a direct use of
26720 ipa_node_params_sum.
26721 (has_undead_caller_from_outside_scc_p): Likewise, reusing results of
26723 (identify_dead_nodes): Likewise.
26724 (ipcp_store_bits_results): Replaced IPA_NODE_REF with direct uses of
26725 ipa_node_params_sum.
26726 (ipcp_store_vr_results): Likewise.
26727 * ipa-fnsummary.c (evaluate_properties_for_edge): Likewise.
26728 (ipa_fn_summary_t::duplicate): Likewise.
26729 (analyze_function_body): Likewise.
26730 (estimate_calls_size_and_time): Likewise.
26731 (ipa_cached_call_context::duplicate_from): Likewise.
26732 (ipa_call_context::equal_to): Likewise.
26733 (remap_edge_params): Likewise.
26734 (ipa_merge_fn_summary_after_inlining): Likewise.
26735 (inline_read_section): Likewise.
26736 * ipa-icf.c (sem_function::param_used_p): Likewise.
26737 * ipa-modref.c (compute_parm_map): Likewise.
26738 (compute_parm_map): Replaced IPA_EDGE_REF with a direct use of
26740 (get_access_for_fnspec): Replaced IPA_NODE_REF with a direct use of
26741 ipa_node_params_sum and replaced IPA_EDGE_REF with a direct use of
26743 * ipa-profile.c (check_argument_count): Likewise.
26744 * ipa-prop.c (ipa_alloc_node_params): Replaced IPA_NODE_REF_GET_CREATE
26745 with a direct use of ipa_node_params_sum.
26746 (ipa_initialize_node_params): Likewise.
26747 (ipa_print_node_jump_functions_for_edge): Replaced IPA_EDGE_REF with a
26748 direct use of ipa_edge_args_sum and reused the query result.
26749 (ipa_compute_jump_functions_for_edge): Replaced IPA_NODE_REF with a
26750 direct use of ipa_node_params_sum and replaced IPA_EDGE_REF with a
26751 direct use of ipa_edge_args_sum.
26752 (ipa_note_param_call): Replaced IPA_NODE_REF with a direct use of
26753 ipa_node_params_sum and reused the result of the query.
26754 (ipa_analyze_node): Likewise.
26755 (ipa_analyze_controlled_uses): Replaced IPA_NODE_REF with a direct use
26756 of ipa_node_params_sum.
26757 (update_jump_functions_after_inlining): Replaced IPA_EDGE_REF with
26758 direct uses of ipa_edge_args_sum.
26759 (update_indirect_edges_after_inlining): Replaced IPA_NODE_REF with
26760 direct uses of ipa_node_params_sum and replaced IPA_EDGE_REF with a
26761 direct use of ipa_edge_args_sum. Removed superficial re-querying the
26763 (propagate_controlled_uses): Replaced IPA_NODE_REF with direct uses of
26764 ipa_node_params_sum and replaced IPA_EDGE_REF with a direct use of
26766 (ipa_propagate_indirect_call_infos): Replaced IPA_EDGE_REF with a
26767 direct use of ipa_edge_args_sum.
26768 (ipa_edge_args_sum_t::duplicate): Replaced IPA_NODE_REF with a direct
26769 use of ipa_node_params_sum.
26770 (ipa_print_node_params): Likewise.
26771 (ipa_write_node_info): Likewise and also replaced IPA_EDGE_REF with
26772 direct uses of ipa_edge_args_sum.
26773 (ipa_read_edge_info): Replaced IPA_EDGE_REF with a direct use of
26775 (ipa_read_node_info): Replaced IPA_NODE_REF with a direct use of
26776 ipa_node_params_sum.
26777 (ipa_prop_write_jump_functions): Likewise. Move variable node to the
26778 scopes where it is used.
26780 2021-05-10 Uroš Bizjak <ubizjak@gmail.com>
26782 * config/i386/i386-expand.c (ix86_expand_sse_movcc)
26783 <case E_V2SImode>: Force op_true to register.
26785 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
26787 * config/arm/iterators.md (MVE_FP_COMPARISONS): New.
26788 * config/arm/mve.md (mve_vcmp<mve_cmp_op>q_f<mode>)
26789 (mve_vcmp<mve_cmp_op>q_n_f<mode>): New, merge all vcmp_*f*
26791 (mve_vcmpeqq_f<mode>, mve_vcmpeqq_n_f<mode>, mve_vcmpgeq_f<mode>)
26792 (mve_vcmpgeq_n_f<mode>, mve_vcmpgtq_f<mode>)
26793 (mve_vcmpgtq_n_f<mode>, mve_vcmpleq_f<mode>)
26794 (mve_vcmpleq_n_f<mode>, mve_vcmpltq_f<mode>)
26795 (mve_vcmpltq_n_f<mode>, mve_vcmpneq_f<mode>)
26796 (mve_vcmpneq_n_f<mode>): Remove.
26797 * config/arm/unspecs.md (VCMPEQQ_F, VCMPEQQ_N_F, VCMPGEQ_F)
26798 (VCMPGEQ_N_F, VCMPGTQ_F, VCMPGTQ_N_F, VCMPLEQ_F, VCMPLEQ_N_F)
26799 (VCMPLTQ_F, VCMPLTQ_N_F, VCMPNEQ_F, VCMPNEQ_N_F): Remove.
26801 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
26803 * config/arm/iterators.md (MVE_COMPARISONS): New.
26805 (mve_cmp_type): New.
26806 * config/arm/mve.md (mve_vcmp<mve_cmp_op>q_<mode>): New, merge all
26808 (mve_vcmpneq_<mode>, mve_vcmpcsq_n_<mode>, mve_vcmpcsq_<mode>)
26809 (mve_vcmpeqq_n_<mode>, mve_vcmpeqq_<mode>, mve_vcmpgeq_n_<mode>)
26810 (mve_vcmpgeq_<mode>, mve_vcmpgtq_n_<mode>, mve_vcmpgtq_<mode>)
26811 (mve_vcmphiq_n_<mode>, mve_vcmphiq_<mode>, mve_vcmpleq_n_<mode>)
26812 (mve_vcmpleq_<mode>, mve_vcmpltq_n_<mode>, mve_vcmpltq_<mode>)
26813 (mve_vcmpneq_n_<mode>, mve_vcmpltq_n_<mode>, mve_vcmpltq_<mode>)
26814 (mve_vcmpneq_n_<mode>): Remove.
26816 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
26818 * config/arm/arm_mve.h (__arm_vcmp*): Remove 's' suffix.
26819 * config/arm/arm_mve_builtins.def (vcmp*): Remove 's' suffix.
26820 * config/arm/mve.md (mve_vcmp*): Remove 's' suffix in pattern
26823 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
26825 * config/arm/arm_mve_builtins.def (vcmpneq_u): Remove.
26826 (vcmpneq_n_u): Likewise.
26827 (vcmpeqq_u,): Likewise.
26828 (vcmpeqq_n_u): Likewise.
26829 * config/arm/iterators.md (supf): Remove VCMPNEQ_U, VCMPEQQ_U,
26830 VCMPEQQ_N_U and VCMPNEQ_N_U.
26831 * config/arm/mve.md (mve_vcmpneq): Remove <supf> iteration.
26832 (mve_vcmpeqq_n): Likewise.
26833 (mve_vcmpeqq): Likewise.
26834 (mve_vcmpneq_n): Likewise.
26836 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
26838 * config/arm/arm_mve.h (__arm_vcmpeq*u*, __arm_vcmpne*u*): Call
26839 the 's' version of the builtin.
26841 2021-05-10 Richard Biener <rguenther@suse.de>
26843 PR tree-optimization/100492
26844 * tree-loop-distribution.c (find_seed_stmts_for_distribution):
26845 Find nothing when the loop contains an irreducible region.
26847 2021-05-10 Richard Biener <rguenther@suse.de>
26849 PR middle-end/100464
26851 * gimple-fold.c (canonicalize_constructor_val): Do not set
26854 2021-05-10 Richard Biener <rguenther@suse.de>
26856 PR tree-optimization/100434
26857 * tree-ssa-dse.c (initialize_ao_ref_for_dse): Handle
26859 (dse_optimize_stmt): Handle call LHS by dropping the
26860 LHS or the whole call if it doesn't have other
26862 (pass_dse::execute): Adjust.
26864 2021-05-10 Martin Liska <mliska@suse.cz>
26866 * Makefile.in: Add missing genversion rule.
26868 2021-05-10 Alex Coplan <alex.coplan@arm.com>
26871 * config/arm/mve.md (*mve_mov<mode>): Simplify output code. Use
26872 vldrw.u32 and vstrw.32 for V2D[IF]mode loads and stores.
26874 2021-05-10 Martin Liska <mliska@suse.cz>
26876 * builtins.c (is_builtin_name): Use startswith
26877 function instead of strncmp.
26878 * collect2.c (main): Likewise.
26879 (has_lto_section): Likewise.
26880 (scan_libraries): Likewise.
26881 * coverage.c (coverage_checksum_string): Likewise.
26882 (coverage_init): Likewise.
26883 * dwarf2out.c (is_cxx): Likewise.
26884 (gen_compile_unit_die): Likewise.
26885 * gcc-ar.c (main): Likewise.
26886 * gcc.c (init_spec): Likewise.
26887 (read_specs): Likewise.
26888 (execute): Likewise.
26889 (check_live_switch): Likewise.
26890 * genattrtab.c (write_attr_case): Likewise.
26891 (IS_ATTR_GROUP): Likewise.
26892 * gencfn-macros.c (main): Likewise.
26893 * gengtype.c (type_for_name): Likewise.
26894 (gen_rtx_next): Likewise.
26895 (get_file_langdir): Likewise.
26896 (write_local): Likewise.
26897 * genmatch.c (get_operator): Likewise.
26898 (get_operand_type): Likewise.
26899 (expr::gen_transform): Likewise.
26900 * genoutput.c (validate_optab_operands): Likewise.
26901 * incpath.c (add_sysroot_to_chain): Likewise.
26902 * langhooks.c (lang_GNU_C): Likewise.
26903 (lang_GNU_CXX): Likewise.
26904 (lang_GNU_Fortran): Likewise.
26905 (lang_GNU_OBJC): Likewise.
26906 * lto-wrapper.c (run_gcc): Likewise.
26907 * omp-general.c (omp_max_simt_vf): Likewise.
26908 * omp-low.c (omp_runtime_api_call): Likewise.
26909 * opts-common.c (parse_options_from_collect_gcc_options): Likewise.
26910 * read-rtl-function.c (function_reader::read_rtx_operand_r): Likewise.
26911 * real.c (real_from_string): Likewise.
26912 * selftest.c (assert_str_startswith): Likewise.
26913 * timevar.c (timer::validate_phases): Likewise.
26914 * tree.c (get_file_function_name): Likewise.
26915 * ubsan.c (ubsan_use_new_style_p): Likewise.
26916 * varasm.c (default_function_rodata_section): Likewise.
26917 (incorporeal_function_p): Likewise.
26918 (default_section_type_flags): Likewise.
26919 * system.h (startswith): Define startswith.
26921 2021-05-10 Martin Liska <mliska@suse.cz>
26923 * bitmap.h (class auto_bitmap): Remove
26924 __cplusplus >= 201103.
26925 * config/aarch64/aarch64.c: Likewise.
26926 * gimple-ssa-store-merging.c (store_immediate_info::store_immediate_info):
26928 * sbitmap.h: Likewise.
26930 2021-05-10 Martin Liska <mliska@suse.cz>
26932 * Makefile.in: Rename gcov-iov to genversion and depend
26933 on version.h (instead of gcov-iov.h).
26934 * gcov-io.h: Include version.h instread of gcov-iov.h.
26935 * gengtype-state.c (read_state_version): Likewise.
26936 * gcov-iov.c: Moved to...
26937 * genversion.c: ...here.
26938 * lto-streamer.h (LTO_major_version): Define it with
26940 * version.c: Removed.
26941 * version.h: Removed.
26943 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
26945 * config/arc/arc.md (UNSPEC_ARC_DMPYWH): Define.
26946 * config/arc/simdext.md (VCT): Add predicates for iterator
26949 (voptab): Likewise.
26950 (vec_widen_<V_US>mult_hi_v4hi): Change pattern predicate.
26951 (<voptab>v2si3): New patterns.
26953 (reduc_plus_scal_v4hi): Likewise.
26954 (reduc_plus_scal_v2si): Likewise.
26955 (vec_duplicatev2si): Likewise.
26956 (vec_duplicatev4hi): Likewise.
26958 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
26960 * config/arc/simdext.md: Format and cleanup file.
26962 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
26964 * config/arc/simdext.md (movmisalignv2hi): Allow misaligned access
26965 only when munaligned-access option is on.
26966 (movmisalign<mode>): Likewise.
26968 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
26970 * common/config/arc/arc-common.c (arc_handle_option): Remove dot
26972 * config/arc/arc.c (arc_reorg): Remove underscore from string.
26974 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
26976 * config/arc/arc.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
26977 (CTZ_DEFINED_VALUE_AT_ZERO): Likewise.
26978 * config/arc/arc.md (clrsbsi2): Cleanup pattern.
26979 (norm_f): Likewise.
26982 (clzsi2): Use fls instruction when available.
26983 (arc_clzsi2): Likewise.
26985 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
26987 * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add r26 and r27.
26989 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
26991 * doc/extend.texi (__builtin_arc_sr): Swap arguments.
26993 2021-05-10 Bernd Edlinger <bernd.edlinger@hotmail.de>
26995 PR middle-end/100467
26996 * toplev.c (compile_file): Call insn_locations_init before
26997 targetm.asm_out.code_end.
26999 2021-05-07 Andrew Stubbs <ams@codesourcery.com>
27002 2021-05-07 Andrew Stubbs <ams@codesourcery.com>
27004 * config/gcn/gcn.c (gcn_scalar_mode_supported_p): Disable TImode.
27006 2021-05-07 Jakub Jelinek <jakub@redhat.com>
27007 Andrew Stubbs <amd@codesourcery.com>
27010 * builtins.c (try_store_by_multiple_pieces): Use force_operand for
27011 emit_move_insn operands.
27013 2021-05-07 Eric Botcazou <ebotcazou@adacore.com>
27015 * cfgexpand.c (expand_gimple_basic_block): Do not inherit a current
27016 location for the outgoing edges of an empty block.
27017 * dwarf2out.c (add_subscript_info): Retrieve the bounds and index
27018 type by means of the get_array_descr_info langhook, if it is set and
27019 returns true. Remove obsolete code dealing with unnamed subtypes.
27021 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
27023 * gimple-range-cache.cc (ssa_block_ranges): Virtualize.
27024 (sbr_vector): Renamed from ssa_block_cache.
27025 (sbr_vector::sbr_vector): Allocate from obstack abd initialize.
27026 (ssa_block_ranges::~ssa_block_ranges): Remove.
27027 (sbr_vector::set_bb_range): Use varying and undefined cached values.
27028 (ssa_block_ranges::set_bb_varying): Remove.
27029 (sbr_vector::get_bb_range): Adjust assert.
27030 (sbr_vector::bb_range_p): Adjust assert.
27031 (~block_range_cache): No freeing loop required.
27032 (block_range_cache::get_block_ranges): Remove.
27033 (block_range_cache::set_bb_range): Inline get_block_ranges.
27034 (block_range_cache::set_bb_varying): Remove.
27035 * gimple-range-cache.h (set_bb_varying): Remove prototype.
27036 * value-range.h (irange_allocator::get_memory): New.
27038 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
27040 * gimple-range-cache.cc (non_null_ref::non_null_deref_p): Search
27041 dominator tree is available and requested.
27042 (ranger_cache::ssa_range_in_bb): Don't search dom tree here.
27043 (ranger_cache::fill_block_cache): Don't search dom tree here either.
27044 * gimple-range-cache.h (non_null_deref_p): Add dom_search param.
27046 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
27048 * gimple-range.cc (gimple_ranger::range_on_exit): Handle block with
27049 only PHI nodes better.
27051 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
27053 * gimple-range-edge.h (gimple_outgoing_range): Rename from
27055 (gcond_edge_range): Export prototype.
27056 * gimple-range-edge.cc (gcond_edge_range): New.
27057 (gimple_outgoing_range::edge_range_p): Use gcond_edge_range.
27058 * gimple-range-gori.h (gori_compute): Use gimple_outgoing_range.
27060 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
27062 * gimple-range-edge.cc (outgoing_range::calc_switch_ranges): Compute
27063 default range into a temp and allocate only what is needed.
27065 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
27067 * range-op.cc (operator_trunc_mod::wi_fold): x % 0 is UNDEFINED.
27069 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
27071 * gimple-range.h (gimple_range_global): Pick up parameter initial
27072 values, and use-before defined locals are UNDEFINED.
27074 2021-05-07 Eric Botcazou <ebotcazou@adacore.com>
27076 * doc/extend.texi (scalar_storage_order): Mention effect on pointer
27078 * tree.h (reverse_storage_order_for_component_p): Return false if
27079 the type is a pointer.
27081 2021-05-07 Andrew Stubbs <ams@codesourcery.com>
27083 * config/gcn/gcn.c (gcn_scalar_mode_supported_p): Disable TImode.
27085 2021-05-07 Uroš Bizjak <ubizjak@gmail.com>
27088 * config/i386/i386-expand.c (ix86_expand_sse_movcc):
27089 Handle V8QI, V4HI and V2SI modes.
27090 * config/i386/mmx.md (mmx_pblendvb): New insn pattern.
27091 * config/i386/sse.md (unspec): Move UNSPEC_BLENDV ...
27092 * config/i386/i386.md (unspec): ... here.
27094 2021-05-07 Tobias Burnus <tobias@codesourcery.com>
27095 Tom de Vries <tdevries@suse.de>
27097 * omp-low.c (lower_rec_simd_input_clauses): Set max_vf = 1 if
27098 a truth_value_p reduction variable is nonintegral.
27100 2021-05-07 Uroš Bizjak <ubizjak@gmail.com>
27103 * config/i386/i386-expand.c (ix86_use_mask_cmp_p):
27104 Return false for mode sizes < 16.
27106 2021-05-07 Jakub Jelinek <jakub@redhat.com>
27109 * config/i386/mmx.md (*xop_pcmov_<mode>): New define_insn.
27111 2021-05-06 Martin Jambor <mjambor@suse.cz>
27113 * ipa-sra.c (ipa_sra_dump_all_summaries): Dump edge summaries even
27114 when there is no function summary.
27115 (ipa_sra_summarize_function): produce edge summaries even when
27118 2021-05-06 Tom Tromey <tom@tromey.com>
27120 * godump.c (string_hash_eq): Remove.
27121 (go_finish): Use htab_eq_string.
27123 2021-05-06 Tom Tromey <tom@tromey.com>
27125 * gengtype-state.c (read_state): Use htab_eq_string.
27126 (string_eq): Remove.
27128 2021-05-06 Tom Tromey <tom@tromey.com>
27130 * gensupport.c (htab_eq_string): Remove.
27132 2021-05-06 Bernd Edlinger <bernd.edlinger@hotmail.de>
27135 * debug.h (gcc_debug_hooks): Add set_ignored_loc function pointer.
27136 * dwarf2out.h (dw_fde_node::ignored_debug): New data item.
27137 * dbxout.c (dbx_debug_hooks, xcoff_debug_hooks): Add dummy
27138 set_ignored_loc callbacks.
27139 * debug.c (do_nothing_debug_hooks): Likewise.
27140 * vmsdbgout.c (vmsdbg_debug_hooks): Likewise.
27141 * dwarf2out.c (text_section_used, cold_text_section_used): Remove.
27142 (in_text_section_p, last_text_label, last_cold_label,
27143 switch_text_ranges, switch_cold_ranges): New data items.
27144 (dwarf2out_note_section_used): Remove.
27145 (dwarf2out_begin_prologue): Set fde->ignored_debug and
27147 (mark_ignored_debug_section): New helper function.
27148 (dwarf2out_end_epilogue, dwarf2out_switch_text_section): Call
27149 mark_ignored_debug_section.
27150 (dwarf2_debug_hooks): Use dwarf2out_set_ignored_loc.
27151 (dwarf2_lineno_debug_hooks): Use dummy for set_ignored_loc.
27152 (size_of_aranges): Adjust formula for multi-part text ranges size.
27153 (output_aranges): Output multi-part text ranges.
27154 (dwarf2out_set_ignored_loc): New callback function.
27155 (dwarf2out_finish): Output multi-part text ranges.
27156 (dwarf2out_c_finalize): Clear new data items.
27157 * final.c (final_start_function_1): Call set_ignored_loc callback.
27158 (final_scan_insn_1): Likewise.
27159 * ggc-page.c (gt_ggc_mx): New helper function.
27160 * stringpool.c (gt_pch_nx): Likewise.
27162 2021-05-06 Richard Biener <rguenther@suse.de>
27164 * timevar.def (TV_TREE_INSERT_PHI_NODES): Remove.
27165 (TV_TREE_SSA_REWRITE_BLOCKS): Likewise.
27166 (TV_TREE_INTO_SSA): New.
27167 * tree-into-ssa.c (insert_phi_nodes): Do not account separately.
27168 (rewrite_blocks): Likewise.
27169 (pass_data_build_ssa): Account to TV_TREE_INTO_SSA.
27171 2021-05-06 Jakub Jelinek <jakub@redhat.com>
27173 * tree-ssa-phiopt.c (value_replacement, minmax_replacement,
27174 abs_replacement, xor_replacement,
27175 cond_removal_in_popcount_clz_ctz_pattern,
27176 replace_phi_edge_with_variable): Change type of phi argument from
27177 gimple * to gphi *.
27179 2021-05-06 Richard Biener <rguenther@suse.de>
27181 * tree-ssa-loop-split.c (split_loop): Delay updating SSA form.
27182 Output an opt-info message.
27183 (do_split_loop_on_cond): Likewise.
27184 (tree_ssa_split_loops): Update SSA form here.
27186 2021-05-06 Richard Biener <rguenther@suse.de>
27188 * tree-inline.c (tree_function_versioning): Fix DECL_BY_REFERENCE
27189 return variable removal.
27191 2021-05-06 Marius Hillenbrand <mhillen@linux.ibm.com>
27193 * config/s390/s390-builtins.def (O_M5, O1_M5, ...): Remove unused macros.
27194 (s390_vec_permi_s64, s390_vec_permi_b64, s390_vec_permi_u64)
27195 (s390_vec_permi_dbl, s390_vpdi): Use the O3_U2 type for the immediate
27197 * config/s390/s390.c (s390_const_operand_ok): Remove unused
27200 2021-05-06 Jakub Jelinek <jakub@redhat.com>
27202 PR tree-optimization/94589
27203 * tree-ssa-phiopt.c (tree_ssa_phiopt_worker): Call
27204 spaceship_replacement.
27205 (cond_only_block_p, spaceship_replacement): New functions.
27207 2021-05-06 Richard Biener <rguenther@suse.de>
27210 * tree-emutls.c (gen_emutls_addr): Pass in whether we're
27211 dealing with a debug use and only query existing addresses
27213 (lower_emutls_1): Avoid splitting out addresses for debug
27214 stmts, reset the debug stmt when we fail to find existing
27216 (lower_emutls_phi_arg): Set wi.stmt.
27218 2021-05-06 Christoph Muellner <cmuellner@gcc.gnu.org>
27221 * config/riscv/riscv.c (riscv_block_move_loop): Use cbranch helper.
27222 * config/riscv/riscv.md (cbranch<mode>4): Generate helpers.
27223 (stack_protect_test): Use cbranch helper.
27225 2021-05-05 Eric Botcazou <ebotcazou@adacore.com>
27228 * config/i386/i386.c (ix86_compute_frame_layout): For a SEH target,
27229 always return the establisher frame for __builtin_frame_address (0).
27231 2021-05-05 Ivan Sorokin <vanyacpp@gmail.com>
27234 * config/i386/i386-builtins.c (ix86_cpu_model_type_node): New.
27235 (ix86_cpu_model_var): Likewise.
27236 (ix86_cpu_features2_type_node): Likewise.
27237 (ix86_cpu_features2_var): Likewise.
27238 (fold_builtin_cpu): Cache __cpu_model and __cpu_features2 with
27241 2021-05-05 Martin Sebor <msebor@redhat.com>
27243 * passes.def (pass_warn_printf): Run after SSA.
27245 2021-05-05 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
27247 * config/arm/neon.md (neon_vtst_combine<mode>): New pattern.
27248 * config/arm/predicates.md (minus_one_operand): New predicate.
27250 2021-05-05 Jeff Law <jlaw@tachyum.com>
27252 * config/avr/avr.md: Remove references to CC_STATUS_INIT.
27254 2021-05-05 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
27256 PR rtl-optimization/100263
27257 * postreload.c (move2add_valid_value_p): Ensure register can
27260 2021-05-05 Eric Botcazou <ebotcazou@adacore.com>
27262 PR rtl-optimization/100411
27263 * cfgcleanup.c (try_crossjump_to_edge): Also skip end of prologue
27264 and beginning of function markers.
27266 2021-05-05 Jeff Law <jlaw@tachyum.com>
27268 * config/cr16/cr16.h (NOTICE_UPDATE_CC): Remove.
27269 * config/cr16/cr16.c (notice_update_cc): Remove.
27270 * config/cr16/cr16-protos.h (notice_update_cc): Remove.
27272 2021-05-05 Uroš Bizjak <ubizjak@gmail.com>
27275 * config/i386/i386-expand.c (ix86_expand_int_sse_cmp):
27276 Handle V8QI, V4HI and V2SI modes.
27277 * config/i386/i386.c (ix86_build_const_vector): Handle V2SImode.
27278 (ix86_build_signbit_mask): Ditto.
27279 * config/i386/mmx.md (MMXMODE14): New mode iterator.
27280 (<smaxmin:code><MMXMODE14:mode>3): New expander.
27281 (*mmx_<smaxmin:code><MMXMODE14:mode>3): New insn pattern.
27282 (<umaxmin:code><MMXMODE24:mode>3): New expander.
27283 (*mmx_<umaxmin:code><MMXMODE24:mode>3): New insn pattern.
27284 (vec_cmp<MMXMODEI:mode><MMXMODEI:mode>): New expander.
27285 (vec_cmpu<MMXMODEI:mode><MMXMODEI:mode>): Ditto.
27286 (vcond<MMXMODEI:mode><MMXMODEI:mode>): Ditto.
27287 (vcondu<MMXMODEI:mode><MMXMODEI:mode>): Ditto.
27288 (vcond_mask_<MMXMODEI:mode><MMXMODEI:mode>): Ditto.
27290 2021-05-05 Eric Botcazou <ebotcazou@adacore.com>
27292 * dwarf2out.c (loc_list_from_tree_1) <DECL>: During early DWARF, do
27293 not expand the VALUE_EXPR of variables put in the non-local frame.
27294 * gimplify.c (gimplify_type_sizes) <RECORD_TYPE>: If the type is not
27295 to be ignored for debug info, ensure its variable offsets are not.
27297 2021-05-05 Richard Biener <rguenther@suse.de>
27299 PR tree-optimization/79333
27300 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_stmt):
27301 Fold stmt following SSA edges.
27303 2021-05-05 Richard Biener <rguenther@suse.de>
27305 PR middle-end/100394
27306 * calls.c (expand_call): Preserve possibly throwing calls.
27307 * cfgexpand.c (expand_call_stmt): When a call can throw signal
27308 RTL expansion there are side-effects.
27309 * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Simplify,
27310 mark all possibly throwing stmts necessary unless we can elide
27312 * tree-ssa-dse.c (pass_dse::execute): Preserve exceptions unless
27313 -fdelete-dead-exceptions.
27314 * tree.h (DECL_PURE_P): Add note about exceptions.
27316 2021-05-05 Alexandre Oliva <oliva@adacore.com>
27318 * config/i386/vxworks.h (DBX_REGISTER_NUMBER): Make it
27321 2021-05-04 David Edelsohn <dje.gcc@gmail.com>
27323 * config/rs6000/rs6000-call.c (rs6000_output_mi_thunk): Use
27324 get_fnname_from_decl for name of thunk.
27325 * config/rs6000/rs6000.c (rs6000_declare_alias): Use assemble_name
27326 and ASM_OUTPUT_LABEL.
27327 (rs6000_xcoff_declare_function_name): Use assemble_name and
27329 (rs6000_xcoff_declare_object_name): Use ASM_OUTPUT_LABEL.
27330 (rs6000_xcoff_encode_section_info): Don't add mapping class
27331 for aliases. Always add [DS] mapping class to primary
27333 (rs6000_asm_weaken_decl): Don't explicitly add [DS].
27335 2021-05-04 Martin Sebor <msebor@redhat.com>
27337 PR middle-end/100307
27338 * builtins.c (compute_objsize_r): Clear base0 for pointers.
27340 2021-05-04 Jeff Law <jlaw@tachyum.com>
27342 * config/bfin/bfin.h (NOTICE_UPDATE_CC): Remove.
27344 2021-05-04 Segher Boessenkool <segher@kernel.crashing.org>
27346 * caller-save.c: Remove CC0.
27347 * cfgcleanup.c: Remove CC0.
27348 * cfgrtl.c: Remove CC0.
27349 * combine.c: Remove CC0.
27350 * compare-elim.c: Remove CC0.
27351 * conditions.h: Remove CC0.
27352 * config/h8300/h8300.h: Remove CC0.
27353 * config/h8300/h8300-protos.h: Remove CC0.
27354 * config/h8300/peepholes.md: Remove CC0.
27355 * config/i386/x86-tune-sched.c: Remove CC0.
27356 * config/m68k/m68k.c: Remove CC0.
27357 * config/rl78/rl78.c: Remove CC0.
27358 * config/sparc/sparc.c: Remove CC0.
27359 * config/xtensa/xtensa.c: Remove CC0.
27360 (gen_conditional_move): Use pc_rtx instead of cc0_rtx in a piece of
27361 RTL where that is used as a placeholder only.
27362 * cprop.c: Remove CC0.
27363 * cse.c: Remove CC0.
27364 * cselib.c: Remove CC0.
27365 * df-problems.c: Remove CC0.
27366 * df-scan.c: Remove CC0.
27367 * doc/md.texi: Remove CC0. Adjust an example.
27368 * doc/rtl.texi: Remove CC0. Adjust an example.
27369 * doc/tm.texi: Regenerate.
27370 * doc/tm.texi.in: Remove CC0.
27371 * emit-rtl.c: Remove CC0.
27372 * final.c: Remove CC0.
27373 * fwprop.c: Remove CC0.
27374 * gcse-common.c: Remove CC0.
27375 * gcse.c: Remove CC0.
27376 * genattrtab.c: Remove CC0.
27377 * genconfig.c: Remove CC0.
27378 * genemit.c: Remove CC0.
27379 * genextract.c: Remove CC0.
27380 * gengenrtl.c: Remove CC0.
27381 * genrecog.c: Remove CC0.
27382 * haifa-sched.c: Remove CC0.
27383 * ifcvt.c: Remove CC0.
27384 * ira-costs.c: Remove CC0.
27385 * ira.c: Remove CC0.
27386 * jump.c: Remove CC0.
27387 * loop-invariant.c: Remove CC0.
27388 * lra-constraints.c: Remove CC0.
27389 * lra-eliminations.c: Remove CC0.
27390 * optabs.c: Remove CC0.
27391 * postreload-gcse.c: Remove CC0.
27392 * postreload.c: Remove CC0.
27393 * print-rtl.c: Remove CC0.
27394 * read-rtl-function.c: Remove CC0.
27395 * reg-notes.def: Remove CC0.
27396 * reg-stack.c: Remove CC0.
27397 * reginfo.c: Remove CC0.
27398 * regrename.c: Remove CC0.
27399 * reload.c: Remove CC0.
27400 * reload1.c: Remove CC0.
27401 * reorg.c: Remove CC0.
27402 * resource.c: Remove CC0.
27403 * rtl.c: Remove CC0.
27404 * rtl.def: Remove CC0.
27405 * rtl.h: Remove CC0.
27406 * rtlanal.c: Remove CC0.
27407 * sched-deps.c: Remove CC0.
27408 * sched-rgn.c: Remove CC0.
27409 * shrink-wrap.c: Remove CC0.
27410 * simplify-rtx.c: Remove CC0.
27411 * system.h: Remove CC0. Poison NOTICE_UPDATE_CC, CC_STATUS_MDEP_INIT,
27412 CC_STATUS_MDEP, and CC_STATUS.
27413 * target.def: Remove CC0.
27414 * valtrack.c: Remove CC0.
27415 * var-tracking.c: Remove CC0.
27417 2021-05-04 Richard Biener <rguenther@suse.de>
27419 PR tree-optimization/100414
27420 * tree-ssa-phiopt.c (get_non_trapping): Do not compute dominance
27422 (tree_ssa_phiopt_worker): But unconditionally here.
27424 2021-05-04 Tobias Burnus <tobias@codesourcery.com>
27426 * omp-low.c (lower_rec_input_clauses, lower_reduction_clauses): Handle
27427 && and || with floating-point and complex arguments.
27429 2021-05-04 Eric Botcazou <ebotcazou@adacore.com>
27431 * tree-inline.c (insert_debug_decl_map): Delete.
27432 (copy_debug_stmt): Minor tweak.
27433 (setup_one_parameter): Do not use a variable if the value is either
27434 a read-only DECL or a non-addressable local variable in the caller.
27435 In this case, insert the debug-only variable in the map manually.
27436 (expand_call_inline): Do not generate a CLOBBER for these values.
27437 * tree-inline.h (debug_map): Minor tweak.
27439 2021-05-04 Eric Botcazou <ebotcazou@adacore.com>
27441 * builtins.c (builtin_with_linkage_p): Return true for stp[n]cpy.
27442 * symtab.c (symtab_node::output_to_lto_symbol_table_p): Tidy up.
27444 2021-05-04 Richard Biener <rguenther@suse.de>
27446 PR tree-optimization/100329
27447 * tree-ssa-reassoc.c (can_reassociate_p): Do not reassociate
27449 (insert_stmt_after): Assert we're not running into asm goto.
27451 2021-05-04 Richard Biener <rguenther@suse.de>
27453 PR tree-optimization/100398
27454 * tree-ssa-dse.c (pass_dse::execute): Preserve control
27457 2021-05-04 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
27459 * builtins.c (try_store_by_multiple_pieces): Fix constfun's prototype.
27461 2021-05-04 Alexandre Oliva <oliva@adacore.com>
27463 * builtins.c (try_store_by_multiple_pieces): New.
27464 (expand_builtin_memset_args): Use it. If target_char_cast
27465 fails, proceed as for non-constant val. Pass len's ctz to...
27466 * expr.c (clear_storage_hints): ... this. Try store by
27467 multiple pieces after setmem.
27468 (clear_storage): Adjust.
27469 * expr.h (clear_storage_hints): Likewise.
27470 (try_store_by_multiple_pieces): Declare.
27471 * passes.def: Replace the last copy_prop with ccp.
27473 2021-05-03 Tom de Vries <tdevries@suse.de>
27476 * omp-low.c (lower_rec_input_clauses): Disable SIMT for user-defined
27479 2021-05-03 Richard Biener <rguenther@suse.de>
27481 * tree-ssa-dse.c (dse_classify_store): Track two PHI defs.
27483 2021-05-03 Richard Biener <rguenther@suse.de>
27485 * tree-ssa-dse.c: Do not include domwalk.h but cfganal.h.
27486 (dse_dom_walker): Remove.
27487 (dse_dom_walker::dse_optimize_stmt): Rename...
27488 (dse_optimize_stmt): ... to this, pass in live_bytes sbitmap.
27489 (dse_dom_walker::before_dom_children): Inline ...
27490 (pass_dse::execute): ... here. Perform a reverse program
27493 2021-05-03 H.J. Lu <hjl.tools@gmail.com>
27496 * configure: Regenerated.
27498 2021-05-03 Ilya Leoshkevich <iii@linux.ibm.com>
27501 * config/s390/s390.c (s390_hard_fp_reg_p): New function.
27502 (s390_md_asm_adjust): Handle hard registers.
27504 2021-05-03 Jakub Jelinek <jakub@redhat.com>
27506 PR tree-optimization/100382
27507 * tree-ssa-dse.c: Include tree-eh.h.
27508 (dse_dom_walker::before_dom_children): Don't remove stmts if
27509 stmt_unremovable_because_of_non_call_eh_p is true.
27511 2021-05-02 David Edelsohn <dje.gcc@gmail.com>
27513 * varasm.c (compute_reloc_for_var): Split out from...
27514 (get_variable_section): Use it.
27515 * output.h (compute_reloc_for_var): Declare.
27516 * config/rs6000/rs6000-protos.h
27517 (rs6000_xcoff_asm_output_aligned_decl_common): Change alignment to
27519 * config/rs6000/rs6000.c (rs6000_legitimize_tls_address_aix):
27520 Don't append storage mapping class to symbol.
27521 (rs6000_xcoff_asm_named_section): Add BS and UL mapping classes.
27522 Don't convert TLS BSS to common.
27523 (rs6000_xcoff_unique_section): Don't fall back to select_secton.
27524 (rs6000_xcoff_section_type_flags): Add SECTION_BSS if DECL is
27526 (rs6000_xcoff_asm_globalize_decl_name): Don't strip storage
27528 (rs6000_xcoff_asm_output_aligned_decl_common): Align is unsigned int.
27529 If align is 0 from TLS class, use the same rules as varasm.c
27530 If not common, switch to BSS section manually.
27531 If common, emit appropriate comm or lcomm directive.
27532 (rs6000_xcoff_encode_section_info): Add logic to append all
27533 storage mapping classes.
27534 (rs6000_asm_weaken_decl): Adjust for qualname symbols.
27535 * config/rs6000/xcoff.h (ASM_OUTPUT_ALIGNED_DECL_LOCAL): Use
27536 rs6000_xcoff_asm_output_aligned_decl_common.
27537 (ASM_OUTPUT_ALIGNED_DECL_LOCAL): Use
27538 rs6000_xcoff_asm_output_aligned_decl_common.
27539 (ASM_OUTPUT_TLS_COMMON): Use
27540 rs6000_xcoff_asm_output_aligned_decl_common.
27542 2021-05-02 Jakub Jelinek <jakub@redhat.com>
27545 * config/nvptx/nvptx.c (nvptx_sese_pseudo): Use nullptr instead of 0
27546 as first argument of pseudo_node_t constructors.
27548 2021-05-02 Jakub Jelinek <jakub@redhat.com>
27551 * config/i386/t-i386 (TM_H): Add $(srcdir)/config/i386/i386-isa.def.
27553 2021-05-01 Aldy Hernandez <aldyh@redhat.com>
27555 * value-range.cc (DEFINE_INT_RANGE_GC_STUBS): Remove.
27556 (gt_pch_nx (int_range<1> *&)): New.
27557 (gt_ggc_mx (int_range<1> *&)): New.
27558 * value-range.h (class irange): Add GTY support for
27561 2021-05-01 Geng Qi <gengqi@linux.alibaba.com>
27563 * doc/options.texi (Negative): Change either or to both and.
27565 2021-04-30 Jonathan Wright <jonathan.wright@arm.com>
27567 * config/aarch64/aarch64-simd-builtins.def: Add
27568 float_ml[as][q]_laneq builtin generator macros.
27569 * config/aarch64/aarch64-simd.md (mul_laneq<mode>3): Define.
27570 (aarch64_float_mla_laneq<mode>): Define.
27571 (aarch64_float_mls_laneq<mode>): Define.
27572 * config/aarch64/arm_neon.h (vmla_laneq_f32): Use RTL builtin
27573 instead of GCC vector extensions.
27574 (vmlaq_laneq_f32): Likewise.
27575 (vmls_laneq_f32): Likewise.
27576 (vmlsq_laneq_f32): Likewise.
27578 2021-04-30 Jonathan Wright <jonathan.wright@arm.com>
27580 * config/aarch64/aarch64-simd-builtins.def: Add
27581 float_ml[as]_lane builtin generator macros.
27582 * config/aarch64/aarch64-simd.md (*aarch64_mul3_elt<mode>):
27584 (mul_lane<mode>3): This, and re-order arguments.
27585 (aarch64_float_mla_lane<mode>): Define.
27586 (aarch64_float_mls_lane<mode>): Define.
27587 * config/aarch64/arm_neon.h (vmla_lane_f32): Use RTL builtin
27588 instead of GCC vector extensions.
27589 (vmlaq_lane_f32): Likewise.
27590 (vmls_lane_f32): Likewise.
27591 (vmlsq_lane_f32): Likewise.
27593 2021-04-30 Jonathan Wright <jonathan.wright@arm.com>
27595 * config/aarch64/aarch64-simd-builtins.def: Add float_ml[as]
27596 builtin generator macros.
27597 * config/aarch64/aarch64-simd.md (aarch64_float_mla<mode>):
27599 (aarch64_float_mls<mode>): Define.
27600 * config/aarch64/arm_neon.h (vmla_f32): Use RTL builtin
27601 instead of relying on GCC vector extensions.
27602 (vmla_f64): Likewise.
27603 (vmlaq_f32): Likewise.
27604 (vmlaq_f64): Likewise.
27605 (vmls_f32): Likewise.
27606 (vmls_f64): Likewise.
27607 (vmlsq_f32): Likewise.
27608 (vmlsq_f64): Likewise.
27609 * config/aarch64/iterators.md: Define VDQF_DF mode iterator.
27611 2021-04-30 Jonathan Wright <jonathan.wright@arm.com>
27613 * config/aarch64/aarch64-simd-builtins.def: Add
27614 float_ml[as]_n_builtin generator macros.
27615 * config/aarch64/aarch64-simd.md (*aarch64_mul3_elt_from_dup<mode>):
27617 (mul_n<mode>3): This, and re-order arguments.
27618 (aarch64_float_mla_n<mode>): Define.
27619 (aarch64_float_mls_n<mode>): Define.
27620 * config/aarch64/arm_neon.h (vmla_n_f32): Use RTL builtin
27621 instead of inline asm.
27622 (vmlaq_n_f32): Likewise.
27623 (vmls_n_f32): Likewise.
27624 (vmlsq_n_f32): Likewise.
27626 2021-04-30 Jonathan Wright <joanthan.wright@arm.com>
27628 * config/aarch64/aarch64-simd-builtins.def: Add pmull[2]
27629 builtin generator macros.
27630 * config/aarch64/aarch64-simd.md (aarch64_pmullv8qi): Define.
27631 (aarch64_pmull_hiv16qi_insn): Define.
27632 (aarch64_pmull_hiv16qi): Define.
27633 * config/aarch64/arm_neon.h (vmull_high_p8): Use RTL builtin
27634 instead of inline asm.
27635 (vmull_p8): Likewise.
27637 2021-04-30 Senthil Kumar Selvaraj <saaadhu@gcc.gnu.org>
27639 * config/avr/avr.md: Adjust peepholes to match and
27640 generate parallels with clobber of REG_CC.
27641 (mov<mode>_insn): Rename to mov<mode>_insn_split.
27642 (*mov<mode>_insn): Rename to mov<mode>_insn.
27644 2021-04-30 David Edelsohn <dje.gcc@gmail.com>
27646 * varasm.c (use_blocks_for_decl_p): Don't use section anchors
27647 for VAR_DECLs if -fdata-sections enabled.
27649 2021-04-30 Michael Meissner <meissner@linux.ibm.com>
27651 PR bootstrap/100327
27652 * config/rs6000/rs6000.c
27653 (TARGET_LIBGCC_FLOATING_MODE_SUPPORTED_P): Define.
27654 (rs6000_libgcc_floating_mode_supported_p): New target hook.
27656 2021-04-30 Aldy Hernandez <aldyh@redhat.com>
27658 * tree-ssa-threadbackward.c (class thread_jumps): Split out code
27660 (class back_threader_registry): ...to here...
27661 (class back_threader_profitability): ...and here...
27662 (thread_jumps::thread_through_all_blocks): Remove argument.
27663 (back_threader_registry::back_threader_registry): New.
27664 (back_threader_registry::~back_threader_registry): New.
27665 (back_threader_registry::thread_through_all_blocks): New.
27666 (thread_jumps::profitable_jump_thread_path): Move from here...
27667 (back_threader_profitability::profitable_path_p): ...to here.
27668 (thread_jumps::find_taken_edge): New.
27669 (thread_jumps::convert_and_register_current_path): Move...
27670 (back_threader_registry::register_path): ...to here.
27671 (thread_jumps::register_jump_thread_path_if_profitable): Move...
27672 (thread_jumps::maybe_register_path): ...to here.
27673 (thread_jumps::handle_phi): Call find_taken_edge and
27674 maybe_register_path.
27675 (thread_jumps::handle_assignment): Same.
27676 (thread_jumps::fsm_find_control_statement_thread_paths): Remove
27677 tree argument to handle_phi and handle_assignment.
27678 (thread_jumps::find_jump_threads_backwards): Set m_name. Remove
27679 set of m_speed_p and m_max_threaded_paths.
27680 (pass_thread_jumps::execute): Remove second argument from
27681 find_jump_threads_backwards.
27682 (pass_early_thread_jumps::execute): Same.
27684 2021-04-30 Aldy Hernandez <aldyh@redhat.com>
27686 * tree-ssa-dom.c (class dom_jump_threader_simplifier): New.
27687 (class dom_opt_dom_walker): Initialize some class variables.
27688 (pass_dominator::execute): Pass evrp_range_analyzer and
27689 dom_jump_threader_simplifier to dom_opt_dom_walker.
27690 Adjust for some functions moving into classes.
27691 (simplify_stmt_for_jump_threading): Adjust and move to...
27692 (jump_threader_simplifier::simplify): ...here.
27693 (dom_opt_dom_walker::before_dom_children): Adjust for
27694 m_evrp_range_analyzer.
27695 (dom_opt_dom_walker::after_dom_children): Remove x_vr_values hack.
27696 (test_for_singularity): Place in dom_opt_dom_walker class.
27697 (dom_opt_dom_walker::optimize_stmt): The argument
27698 evrp_range_analyzer is now a class field.
27699 * tree-ssa-threadbackward.c (class thread_jumps): Add m_registry.
27700 (thread_jumps::thread_through_all_blocks): New.
27701 (thread_jumps::convert_and_register_current_path): Use m_registry.
27702 (pass_thread_jumps::execute): Adjust for thread_through_all_blocks
27703 being in the threader class.
27704 (pass_early_thread_jumps::execute): Same.
27705 * tree-ssa-threadedge.c (threadedge_initialize_values): Move...
27706 (jump_threader::jump_threader): ...here.
27707 (threadedge_finalize_values): Move...
27708 (jump_threader::~jump_threader): ...here.
27709 (jump_threader::remove_jump_threads_including): New.
27710 (jump_threader::thread_through_all_blocks): New.
27711 (record_temporary_equivalences_from_phis): Move...
27712 (jump_threader::record_temporary_equivalences_from_phis): ...here.
27713 (record_temporary_equivalences_from_stmts_at_dest): Move...
27714 (jump_threader::record_temporary_equivalences_from_stmts_at_dest):
27716 (simplify_control_stmt_condition_1): Move to jump_threader class.
27717 (simplify_control_stmt_condition): Move...
27718 (jump_threader::simplify_control_stmt_condition): ...here.
27719 (thread_around_empty_blocks): Move...
27720 (jump_threader::thread_around_empty_blocks): ...here.
27721 (thread_through_normal_block): Move...
27722 (jump_threader::thread_through_normal_block): ...here.
27723 (thread_across_edge): Move...
27724 (jump_threader::thread_across_edge): ...here.
27725 (thread_outgoing_edges): Move...
27726 (jump_threader::thread_outgoing_edges): ...here.
27727 * tree-ssa-threadedge.h: Move externally facing functings...
27728 (class jump_threader): ...here...
27729 (class jump_threader_simplifier): ...and here.
27730 * tree-ssa-threadupdate.c (struct redirection_data): Remove comment.
27731 (jump_thread_path_allocator::jump_thread_path_allocator): New.
27732 (jump_thread_path_allocator::~jump_thread_path_allocator): New.
27733 (jump_thread_path_allocator::allocate_thread_edge): New.
27734 (jump_thread_path_allocator::allocate_thread_path): New.
27735 (jump_thread_path_registry::jump_thread_path_registry): New.
27736 (jump_thread_path_registry::~jump_thread_path_registry): New.
27737 (jump_thread_path_registry::allocate_thread_edge): New.
27738 (jump_thread_path_registry::allocate_thread_path): New.
27739 (dump_jump_thread_path): Make extern.
27740 (debug (const vec<jump_thread_edge *> &path)): New.
27741 (struct removed_edges): Move to tree-ssa-threadupdate.h.
27742 (struct thread_stats_d): Remove.
27743 (remove_ctrl_stmt_and_useless_edges): Make static.
27744 (lookup_redirection_data): Move...
27745 (jump_thread_path_registry::lookup_redirection_data): ...here.
27746 (ssa_redirect_edges): Make static.
27747 (thread_block_1): Move...
27748 (jump_thread_path_registry::thread_block_1): ...here.
27749 (thread_block): Move...
27750 (jump_thread_path_registry::thread_block): ...here.
27751 (thread_through_loop_header): Move...
27752 (jump_thread_path_registry::thread_through_loop_header): ...here.
27753 (mark_threaded_blocks): Move...
27754 (jump_thread_path_registry::mark_threaded_blocks): ...here.
27755 (debug_path): Move...
27756 (jump_thread_path_registry::debug_path): ...here.
27757 (debug_all_paths): Move...
27758 (jump_thread_path_registry::dump): ..here.
27759 (rewire_first_differing_edge): Move...
27760 (jump_thread_path_registry::rewire_first_differing_edge): ...here.
27761 (adjust_paths_after_duplication): Move...
27762 (jump_thread_path_registry::adjust_paths_after_duplication): ...here.
27763 (duplicate_thread_path): Move...
27764 (jump_thread_path_registry::duplicate_thread_path): ..here.
27765 (remove_jump_threads_including): Move...
27766 (jump_thread_path_registry::remove_jump_threads_including): ...here.
27767 (thread_through_all_blocks): Move to...
27768 (jump_thread_path_registry::thread_through_all_blocks): ...here.
27769 (delete_jump_thread_path): Remove.
27770 (register_jump_thread): Move...
27771 (jump_thread_path_registry::register_jump_thread): ...here.
27772 * tree-ssa-threadupdate.h: Move externally facing functions...
27773 (class jump_thread_path_allocator): ...here...
27774 (class jump_thread_path_registry): ...and here.
27775 (thread_through_all_blocks): Remove.
27776 (struct removed_edges): New.
27777 (register_jump_thread): Remove.
27778 (remove_jump_threads_including): Remove.
27779 (delete_jump_thread_path): Remove.
27780 (remove_ctrl_stmt_and_useless_edges): Remove.
27781 (free_dom_edge_info): New prototype.
27782 * tree-vrp.c: Remove x_vr_values hack.
27783 (class vrp_jump_threader_simplifier): New.
27784 (vrp_jump_threader_simplifier::simplify): New.
27785 (vrp_jump_threader::vrp_jump_threader): Adjust method signature.
27786 Remove m_dummy_cond.
27787 Instantiate m_simplifier and m_threader.
27788 (vrp_jump_threader::thread_through_all_blocks): New.
27789 (vrp_jump_threader::simplify_stmt): Remove.
27790 (vrp_jump_threader::after_dom_children): Do not set m_dummy_cond.
27791 Remove x_vr_values hack.
27792 (execute_vrp): Adjust for thread_through_all_blocks being in a
27795 2021-04-30 Christophe Lyon <christophe.lyon@linaro.org>
27797 * genflags.c (gen_insn): Print failed expansion string.
27799 2021-04-30 H.J. Lu <hjl.tools@gmail.com>
27801 * expr.c (alignment_for_piecewise_move): Call mode_for_size
27802 without limit to MAX_FIXED_MODE_SIZE.
27804 2021-04-30 H.J. Lu <hjl.tools@gmail.com>
27806 PR middle-end/90773
27807 * builtins.c (builtin_memset_gen_str): Don't use return from
27808 simplify_gen_subreg.
27810 2021-04-30 Uroš Bizjak <ubizjak@gmail.com>
27813 * config/i386/i386.md (*add<mode>3_carry_0r): New insn pattern.
27814 (*addsi3_carry_zext_0r): Ditto.
27815 (*sub<mode>3_carry_0): Ditto.
27816 (*subsi3_carry_zext_0r): Ditto.
27817 * config/i386/predicates.md (ix86_carry_flag_unset_operator):
27819 * config/i386/i386.c (ix86_rtx_costs) <case PLUS, case MINUS>:
27820 Also consider ix86_carry_flag_unset_operator to calculate
27821 the cost of adc/sbb insn.
27823 2021-04-30 Roman Zhuykov <zhroma@ispras.ru>
27825 PR rtl-optimization/100225
27826 PR rtl-optimization/84878
27827 * modulo-sched.c (sms_schedule): Use note_stores to skip loops
27828 where we have an instruction which touches (writes) any hard
27829 register from df->regular_block_artificial_uses set.
27830 Allow not-single-set instruction only right before basic block
27833 2021-04-30 Geng Qi <gengqi@linux.alibaba.com>
27835 * config/riscv/riscv.opt (march=,mabi=): Negative itself.
27837 2021-04-30 LevyHsu <admin@levyhsu.com>
27839 * config/riscv/riscv.c (riscv_min_arithmetic_precision): New.
27840 * config/riscv/riscv.h (TARGET_MIN_ARITHMETIC_PRECISION): New.
27841 * config/riscv/riscv.md (addv<mode>4, uaddv<mode>4): New.
27842 (subv<mode>4, usubv<mode>4, mulv<mode>4, umulv<mode>4): New.
27844 2021-04-29 Alexandre Oliva <oliva@adacore.com>
27846 * config.gcc: Merged x86 and x86_64 cpu_type-setting cases.
27848 2021-04-29 Alexandre Oliva <oliva@adacore.com>
27850 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Rename to...
27851 (ASM_OUTPUT_MAX_SKIP_ALIGN): ... this. Enclose in do/while(0).
27852 * config/i386/i386.c: Adjust.
27853 * config/i386/i386.md: Adjust.
27854 * config/i386/darwin.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Drop.
27855 * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
27856 * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
27857 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
27858 * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
27859 * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
27860 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
27861 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
27862 * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
27863 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
27864 (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
27866 2021-04-29 Uroš Bizjak <ubizjak@gmail.com>
27868 * config/i386/i386-expand.c (ix86_expand_int_compare):
27869 Swap operands of GTU and LEU comparison to emit carry flag comparison.
27870 * config/i386/i386.md (*add<mode>3_carry_0): Change insn
27871 predicate to allow more combine opportunities with memory operands.
27872 (*sub<mode>3_carry_0): Ditto.
27874 2021-04-29 Richard Sandiford <richard.sandiford@arm.com>
27876 PR rtl-optimization/100303
27877 * rtl-ssa/accesses.cc (function_info::make_use_available): Take a
27878 boolean that indicates whether the use will only be used in
27879 debug instructions. Treat it in the same way that existing
27880 cross-EBB debug references would be handled if so.
27881 (function_info::make_uses_available): Likewise.
27882 * rtl-ssa/functions.h (function_info::make_uses_available): Update
27883 prototype accordingly.
27884 (function_info::make_uses_available): Likewise.
27885 * fwprop.c (try_fwprop_subst): Update call accordingly.
27887 2021-04-29 Jeff Law <jlaw@tachyum.com>
27889 * config/nios2/nios2-protos.h (nios2_fpu_insn_enabled): Move outside
27892 2021-04-29 Uroš Bizjak <ubizjak@gmail.com>
27893 Richard Biener <rguenther@suse.de>
27896 * config/i386/i386-builtin.def (IX86_BUILTIN_MASKLOADPD)
27897 (IX86_BUILTIN_MASKLOADPS, IX86_BUILTIN_MASKLOADPD256)
27898 (IX86_BUILTIN_MASKLOADPS256, IX86_BUILTIN_MASKLOADD)
27899 (IX86_BUILTIN_MASKLOADQ, IX86_BUILTIN_MASKLOADD256)
27900 (IX86_BUILTIN_MASKLOADQ256): Move from SPECIAL_ARGS
27901 to PURE_ARGS category.
27902 * config/i386/i386-builtins.c (ix86_init_mmx_sse_builtins):
27903 Handle PURE_ARGS category.
27904 * config/i386/i386-expand.c (ix86_expand_builtin): Ditto.
27906 2021-04-29 Eric Botcazou <ebotcazou@adacore.com>
27908 * configure.ac: Check for the presence of sys/locking.h header and
27909 for whether _LK_LOCK is supported by _locking.
27910 * configure: Regenerate.
27911 * config.in: Likewise.
27912 * gcov-io.h: Define GCOV_LOCKED_WITH_LOCKING if HOST_HAS_LK_LOCK.
27913 * gcov-io.c (gcov_open): Add support for GCOV_LOCKED_WITH_LOCKING.
27914 * system.h: Include <sys/locking.h> if HAVE_SYS_LOCKING_H.
27916 2021-04-29 Uroš Bizjak <ubizjak@gmail.com>
27918 * config/i386/predicates.md (fcmov_comparison_operator):
27919 Do not check for trivial FP comparison operator.
27920 <case GEU, case LTU>: Allow CCGZmode.
27921 <case GTU, case LEU>: Do not allow CCCmode.
27922 (ix86_comparison_operator) <case GTU, case LEU>: Allow only CCmode.
27923 (ix86_carry_flag_operator): Match only LTU and UNLT code.
27924 Do not check for trivial FP comparison operator. Allow CCGZmode.
27926 2021-04-29 Tom de Vries <tdevries@suse.de>
27928 * omp-expand.c (expand_omp_simd): Add step_orig, and replace uses of
27929 fd->loop.step by either step or orig_step.
27931 2021-04-29 Eric Botcazou <ebotcazou@adacore.com>
27933 * config/sparc/sparc.c (gen_load_pcrel_sym): Delete.
27934 (load_got_register): Do the PIC dance here.
27935 (sparc_legitimize_tls_address): Simplify.
27936 (sparc_emit_probe_stack_range): Likewise.
27937 (sparc32_initialize_trampoline): Likewise.
27938 (sparc64_initialize_trampoline): Likewise.
27939 * config/sparc/sparc.md (load_pcrel_sym<P:mode>): Add @ marker.
27940 (probe_stack_range<P:mode>): Likewise.
27941 (flush<P:mode>): Likewise.
27942 (tgd_hi22<P:mode>): Likewise.
27943 (tgd_lo10<P:mode>): Likewise.
27944 (tgd_add<P:mode>): Likewise.
27945 (tgd_call<P:mode>): Likewise.
27946 (tldm_hi22<P:mode>): Likewise.
27947 (tldm_lo10<P:mode>): Likewise.
27948 (tldm_add<P:mode>): Likewise.
27949 (tldm_call<P:mode>): Likewise.
27950 (tldo_hix22<P:mode>): Likewise.
27951 (tldo_lox10<P:mode>): Likewise.
27952 (tldo_add<P:mode>): Likewise.
27953 (tie_hi22<P:mode>): Likewise.
27954 (tie_lo10<P:mode>): Likewise.
27955 (tie_add<P:mode>): Likewise.
27956 (tle_hix22<P:mode>): Likewise.
27957 (tle_lox10<P:mode>): Likewise.
27958 (stack_protect_setsi): Rename to...
27959 (stack_protect_set32): ...this.
27960 (stack_protect_setdi): Rename to...
27961 (stack_protect_set64): ...this.
27962 (stack_protect_set): Adjust calls to above.
27963 (stack_protect_testsi): Rename to...
27964 (stack_protect_test32): ...this.
27965 (stack_protect_testdi): Rename to...
27966 (stack_protect_test64): ...this.
27967 (stack_protect_test): Adjust calls to above.
27969 2021-04-29 H.J. Lu <hjl.tools@gmail.com>
27971 PR middle-end/90773
27972 * builtins.c (builtin_memcpy_read_str): Add a dummy argument.
27973 (builtin_strncpy_read_str): Likewise.
27974 (builtin_memset_read_str): Add an argument for the previous RTL
27975 information and generate the new RTL from the previous RTL info.
27976 (builtin_memset_gen_str): Likewise.
27977 * builtins.h (builtin_strncpy_read_str): Update the prototype.
27978 (builtin_memset_read_str): Likewise.
27979 * expr.c (by_pieces_ninsns): If targetm.overlap_op_by_pieces_p()
27980 returns true, round up size and alignment to the widest integer
27981 mode for maximum size.
27982 (pieces_addr::adjust): Add a pointer to by_pieces_prev argument
27983 and pass it to m_constfn.
27984 (op_by_pieces_d): Add m_push and m_overlap_op_by_pieces.
27985 (op_by_pieces_d::op_by_pieces_d): Add a bool argument to
27986 initialize m_push. Initialize m_overlap_op_by_pieces with
27987 targetm.overlap_op_by_pieces_p ().
27988 (op_by_pieces_d::run): Pass the previous RTL information to
27989 pieces_addr::adjust and generate overlapping operations if
27990 m_overlap_op_by_pieces is true.
27992 (move_by_pieces_d::move_by_pieces_d): Updated for op_by_pieces_d
27994 (store_by_pieces_d::store_by_pieces_d): Updated for op_by_pieces_d
27996 (can_store_by_pieces): Use by_pieces_constfn on constfun.
27997 (store_by_pieces): Use by_pieces_constfn on constfun. Updated
27998 for op_by_pieces_d change.
27999 (clear_by_pieces_1): Add a dummy argument.
28000 (clear_by_pieces): Updated for op_by_pieces_d change.
28001 (compare_by_pieces_d::compare_by_pieces_d): Likewise.
28002 (string_cst_read_str): Add a dummy argument.
28003 * expr.h (by_pieces_constfn): Add a dummy argument.
28004 (by_pieces_prev): New.
28005 * target.def (overlap_op_by_pieces_p): New target hook.
28006 * config/i386/i386.c (TARGET_OVERLAP_OP_BY_PIECES_P): New.
28007 * doc/tm.texi.in: Add TARGET_OVERLAP_OP_BY_PIECES_P.
28008 * doc/tm.texi: Regenerated.
28010 2021-04-29 Richard Biener <rguenther@suse.de>
28012 PR tree-optimization/100253
28013 * tree-vect-stmts.c (vectorizable_load): Do not assume
28014 element alignment when DR_MISALIGNMENT is -1.
28015 (vectorizable_store): Likewise.
28017 2021-04-29 Jakub Jelinek <jakub@redhat.com>
28020 * config/aarch64/aarch64.c (aarch64_add_offset_1_temporaries): Use
28021 absu_hwi instead of abs_hwi.
28023 2021-04-29 Richard Biener <rguenther@suse.de>
28025 PR middle-end/38474
28026 * tree-ssa-structalias.c (add_graph_edge): Avoid direct
28027 forwarding when indirect forwarding through ESCAPED
28030 2021-04-29 Tom de Vries <tdevries@suse.de>
28033 * internal-fn.c (expand_GOMP_SIMT_ENTER_ALLOC)
28034 (expand_GOMP_SIMT_LAST_LANE, expand_GOMP_SIMT_ORDERED_PRED)
28035 (expand_GOMP_SIMT_VOTE_ANY, expand_GOMP_SIMT_XCHG_BFLY)
28036 (expand_GOMP_SIMT_XCHG_IDX): Ensure target is assigned to.
28038 2021-04-29 Richard Biener <rguenther@suse.de>
28040 PR tree-optimization/99912
28041 * tree-ssa-dse.c (dse_dom_walker::m_need_cfg_cleanup): New.
28042 (dse_dom_walker::todo): Likewise.
28043 (dse_dom_walker::dse_optimize_stmt): Move VDEF check to the
28045 (dse_dom_walker::before_dom_children): Remove trivially
28046 dead SSA defs and schedule CFG cleanup if we removed all
28048 (pass_dse::execute): Get TODO as computed by the DOM walker
28049 and return it. Wipe dominator info earlier.
28051 2021-04-29 Richard Biener <rguenther@suse.de>
28054 * ipa-prop.c (ipcp_modif_dom_walker::before_dom_children):
28055 Track blocks to cleanup EH in new m_need_eh_cleanup.
28056 (ipcp_modif_dom_walker::cleanup_eh): New.
28057 (ipcp_transform_function): Release dominator info before
28060 2021-04-29 Martin Sebor <msebor@redhat.com>
28062 PR middle-end/100250
28063 * attribs.c (attr_access::array_as_string): Avoid dereferencing
28064 a pointer when it's null.
28066 2021-04-29 Martin Sebor <msebor@redhat.com>
28068 * Makefile.in (OBJS): Add ipa-free-lang-data.o.
28069 * ipa-free-lang-data.cc: New file.
28070 * tree.c: Move pass free_lang_data to file above.
28071 (build_array_type_1): Declare extern.
28072 * tree.h (build_array_type_1): Declare.
28074 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
28076 * config/aarch64/aarch64-simd-builtins.def: Modify comment to
28077 make consistent with updated RTL pattern.
28078 * config/aarch64/aarch64-simd.md (aarch64_<sur>qmovn<mode>):
28079 Implement using ss_truncate and us_truncate rather than
28081 * config/aarch64/iterators.md: Remove redundant unspecs and
28082 iterator: UNSPEC_[SU]QXTN and SUQMOVN respectively.
28084 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
28086 * config/aarch64/arm_acle.h (__attribute__): Make intrinsic
28087 attributes consistent with those defined in arm_neon.h.
28089 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
28091 * config/aarch64/arm_fp16.h (__attribute__): Make intrinsic
28092 attributes consistent with those defined in arm_neon.h.
28094 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
28096 * config/aarch64/aarch64-simd-builtins.def: Add
28097 float_trunc_rodd builtin generator macros.
28098 * config/aarch64/aarch64-simd.md (aarch64_float_trunc_rodd_df):
28100 (aarch64_float_trunc_rodd_lo_v2sf): Define.
28101 (aarch64_float_trunc_rodd_hi_v4sf_le): Define.
28102 (aarch64_float_trunc_rodd_hi_v4sf_be): Define.
28103 (aarch64_float_trunc_rodd_hi_v4sf): Define.
28104 * config/aarch64/arm_neon.h (vcvtx_f32_f64): Use RTL builtin
28105 instead of inline asm.
28106 (vcvtx_high_f32_f64): Likewise.
28107 (vcvtxd_f32_f64): Likewise.
28108 * config/aarch64/iterators.md: Add FCVTXN unspec.
28110 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
28112 * config/aarch64/aarch64-simd-builtins.def: Add tbx1 builtin
28114 * config/aarch64/aarch64-simd.md (aarch64_tbx1<mode>):
28116 * config/aarch64/arm_neon.h (vqtbx1_s8): USE RTL builtin
28117 instead of inline asm.
28118 (vqtbx1_u8): Likewise.
28119 (vqtbx1_p8): Likewise.
28120 (vqtbx1q_s8): Likewise.
28121 (vqtbx1q_u8): Likewise.
28122 (vqtbx1q_p8): Likewise.
28123 (vtbx2_s8): Likewise.
28124 (vtbx2_u8): Likewise.
28125 (vtbx2_p8): Likewise.
28127 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
28129 * config/aarch64/aarch64-simd-builtins.def: Add tbl1 builtin
28131 * config/aarch64/arm_neon.h (vqtbl1_p8): Use RTL builtin
28132 instead of inline asm.
28133 (vqtbl1_s8): Likewise.
28134 (vqtbl1_u8): Likewise.
28135 (vqtbl1q_p8): Likewise.
28136 (vqtbl1q_s8): Likewise.
28137 (vqtbl1q_u8): Likewise.
28138 (vtbl1_s8): Likewise.
28139 (vtbl1_u8): Likewise.
28140 (vtbl1_p8): Likewise.
28141 (vtbl2_s8): Likewise.
28142 (vtbl2_u8): Likewise.
28143 (vtbl2_p8): Likewise.
28145 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
28147 * config/aarch64/aarch64-simd-builtins.def: Add polynomial
28148 ssri_n buitin generator macro.
28149 * config/aarch64/arm_neon.h (vsri_n_p8): Use RTL builtin
28150 instead of inline asm.
28151 (vsri_n_p16): Likewise.
28152 (vsri_n_p64): Likewise.
28153 (vsriq_n_p8): Likewise.
28154 (vsriq_n_p16): Likewise.
28155 (vsriq_n_p64): Likewise.
28157 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
28159 * config/aarch64/aarch64-simd-builtins.def: Use VALLP mode
28160 iterator for polynomial ssli_n builtin generator macro.
28161 * config/aarch64/arm_neon.h (vsli_n_p8): Use RTL builtin
28162 instead of inline asm.
28163 (vsli_n_p16): Likewise.
28164 (vsliq_n_p8): Likewise.
28165 (vsliq_n_p16): Likewise.
28166 * config/aarch64/iterators.md: Define VALLP mode iterator.
28168 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
28170 * config/aarch64/aarch64-simd-builtins.def: Use VDQV_L
28171 iterator to generate [su]adalp RTL builtins.
28172 * config/aarch64/aarch64-simd.md: Use VDQV_L iterator in
28173 [su]adalp RTL pattern.
28174 * config/aarch64/arm_neon.h (vpadal_s32): Use RTL builtin
28175 instead of inline asm.
28176 (vpadal_u32): Likewise.
28178 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
28180 * config/aarch64/aarch64-simd-builtins.def: Add [su]addlp
28181 builtin generator macros.
28182 * config/aarch64/aarch64-simd.md (aarch64_<su>addlp<mode>):
28184 * config/aarch64/arm_neon.h (vpaddl_s8): Use RTL builtin
28185 instead of inline asm.
28186 (vpaddl_s16): Likewise.
28187 (vpaddl_s32): Likewise.
28188 (vpaddl_u8): Likewise.
28189 (vpaddl_u16): Likewise.
28190 (vpaddl_u32): Likewise.
28191 (vpaddlq_s8): Likewise.
28192 (vpaddlq_s16): Likewise.
28193 (vpaddlq_s32): Likewise.
28194 (vpaddlq_u8): Likewise.
28195 (vpaddlq_u16): Likewise.
28196 (vpaddlq_u32): Liwewise.
28197 * config/aarch64/iterators.md: Define [SU]ADDLP unspecs with
28198 appropriate attributes.
28200 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
28202 * config/aarch64/aarch64-simd-builtins.def: Use VDQ_I iterator
28203 for aarch64_addp<mode> builtin macro generator.
28204 * config/aarch64/aarch64-simd.md: Use VDQ_I iterator in
28205 aarch64_addp<mode> RTL pattern.
28206 * config/aarch64/arm_neon.h (vpaddq_s8): Use RTL builtin
28207 instead of inline asm.
28208 (vpaddq_s16): Likewise.
28209 (vpaddq_s32): Likewise.
28210 (vpaddq_s64): Likewise.
28211 (vpaddq_u8): Likewise.
28212 (vpaddq_u16): Likewise.
28213 (vpaddq_u32): Likewise.
28214 (vpaddq_u64): Likewise.
28216 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
28218 * config/aarch64/aarch64-simd-builtins.def: Add sq[r]dmulh_n
28219 builtin generator macros.
28220 * config/aarch64/aarch64-simd.md (aarch64_sq<r>dmulh_n<mode>):
28222 * config/aarch64/arm_neon.h (vqdmulh_n_s16): Use RTL builtin
28223 instead of inline asm.
28224 (vqdmulh_n_s32): Likewise.
28225 (vqdmulhq_n_s16): Likewise.
28226 (vqdmulhq_n_s32): Likewise.
28227 (vqrdmulh_n_s16): Likewise.
28228 (vqrdmulh_n_s32): Likewise.
28229 (vqrdmulhq_n_s16): Likewise.
28230 (vqrdmulhq_n_s32): Likewise.
28232 2021-04-28 Tobias Burnus <tobias@codesourcery.com>
28234 * doc/install.texi (--enable-offload-defaulted): Document.
28236 2021-04-28 Senthil Kumar Selvaraj <saaadhu@gcc.gnu.org>
28238 * config/avr/avr-dimode.md: Turn existing patterns into
28239 define_insn_and_split style patterns where the splitter
28240 adds a clobber of the condition code register. Drop "cc"
28241 attribute. Add new patterns to match output of
28243 * config/avr/avr-fixed.md: Likewise.
28244 * config/avr/avr.c (cc_reg_rtx): New.
28245 (avr_parallel_insn_from_insns): Adjust insn count
28246 for removal of set of cc0.
28247 (avr_is_casesi_sequence): Likewise.
28248 (avr_casei_sequence_check_operands): Likewise.
28249 (avr_optimize_casesi): Likewise. Also insert
28250 new insns after jump_insn.
28251 (avr_pass_casesi::avr_rest_of_handle_casesi): Adjust
28252 for removal of set of cc0.
28253 (avr_init_expanders): Initialize cc_reg_rtx.
28254 (avr_regno_reg_class): Handle REG_CC.
28255 (cond_string): Remove usage of CC_OVERFLOW_UNUSABLE.
28256 (avr_notice_update_cc): Remove function.
28257 (ret_cond_branch): Remove usage of CC_OVERFLOW_UNUSABLE.
28258 (compare_condition): Adjust for PARALLEL with
28260 (out_shift_with_cnt): Likewise.
28261 (ashlhi3_out): Likewise.
28262 (ashrhi3_out): Likewise.
28263 (lshrhi3_out): Likewise.
28264 (avr_class_max_nregs): Return single reg for REG_CC.
28265 (avr_compare_pattern): Check for REG_CC instead
28267 (avr_reorg_remove_redundant_compare): Likewise.
28268 (avr_reorg):Adjust for PARALLEL with REG_CC clobber.
28269 (avr_hard_regno_nregs): Return single reg for REG_CC.
28270 (avr_hard_regno_mode_ok): Allow only CCmode for REG_CC.
28271 (avr_md_asm_adjust): Clobber REG_CC.
28272 (TARGET_HARD_REGNO_NREGS): Define.
28273 (TARGET_CLASS_MAX_NREGS): Define.
28274 (TARGET_MD_ASM_ADJUST): Define.
28275 * config/avr/avr.h (FIRST_PSEUDO_REGISTER): Adjust
28277 (enum reg_class): Add CC_REG class.
28278 (NOTICE_UPDATE_CC): Remove.
28279 (CC_OVERFLOW_UNUSABLE): Remove.
28280 (CC_NO_CARRY): Remove.
28281 * config/avr/avr.md: Turn existing patterns into
28282 define_insn_and_split style patterns where the splitter
28283 adds a clobber of the condition code register. Drop "cc"
28284 attribute. Add new patterns to match output of
28286 (sez): Remove unused pattern.
28288 2021-04-28 Richard Earnshaw <rearnsha@arm.com>
28291 * config/arm/arm.c (arm_hard_regno_mode_ok): Only allow VPR to be
28294 2021-04-28 Richard Sandiford <richard.sandiford@arm.com>
28297 * config/aarch64/constraints.md (Utq): Require the address to
28298 be valid for both the element mode and for V2DImode.
28300 2021-04-28 Jakub Jelinek <jakub@redhat.com>
28301 Tobias Burnus <tobias@codesourcery.com>
28303 * configure.ac (OFFLOAD_DEFAULTED): AC_DEFINE if offload-defaulted.
28304 * gcc.c (process_command): New variable.
28305 (driver::maybe_putenv_OFFLOAD_TARGETS): If OFFLOAD_DEFAULTED,
28306 set it if -foffload is defaulted.
28307 * lto-wrapper.c (OFFLOAD_TARGET_DEFAULT_ENV): Define.
28308 (compile_offload_image): If OFFLOAD_DEFAULTED and
28309 OFFLOAD_TARGET_DEFAULT is in the environment, don't fail
28310 if corresponding mkoffload can't be found.
28311 (compile_images_for_offload_targets): Likewise. Free and clear
28312 offload_names if no valid offload is found.
28313 * config.in: Regenerate.
28314 * configure: Regenerate.
28316 2021-04-28 Richard Biener <rguenther@suse.de>
28318 PR tree-optimization/100292
28319 * tree-vect-generic.c (expand_vector_condition): Do not fold
28322 2021-04-27 David Edelsohn <dje.gcc@gmail.com>
28324 * config/rs6000/aix.h (SUBTARGET_DRIVER_SELF_SPECS): New.
28325 * config/rs6000/aix64.opt (m64): New.
28328 2021-04-27 Maciej W. Rozycki <macro@orcam.me.uk>
28330 * config/vax/vax.c (print_operand_address, vax_address_cost_1)
28331 (index_term_p): Handle ASHIFT too.
28333 2021-04-27 Maciej W. Rozycki <macro@orcam.me.uk>
28335 * config/vax/builtins.md (jbb<ccss>i<mode>): Remove operand #3.
28336 (sync_lock_test_and_set<mode>): Adjust accordingly.
28337 (sync_lock_release<mode>): Likewise.
28339 2021-04-27 Maciej W. Rozycki <macro@orcam.me.uk>
28341 * config/vax/vax-protos.h (adjacent_operands_p): Remove
28343 * config/vax/vax.c (adjacent_operands_p): Remove.
28345 2021-04-27 Maciej W. Rozycki <macro@linux-mips.org>
28347 * ifcvt.c (dead_or_predicable) [!IFCVT_MODIFY_TESTS]: Fall
28348 through to the non-conditional execution case if getting the
28349 condition for conditional execution has failed.
28351 2021-04-27 Richard Sandiford <richard.sandiford@arm.com>
28353 PR middle-end/100284
28354 * gimple.c (gimple_could_trap_p_1): Remove VEC_COND_EXPR test.
28355 * tree-eh.c (operation_could_trap_p): Handle VEC_COND_EXPR rather
28356 than asserting on it.
28358 2021-04-27 David Edelsohn <dje.gcc@gmail.com>
28360 * config/rs6000/rs6000.c (rs6000_aix_precompute_tls_p): Protect
28361 with TARGET_AIX_OS.
28363 2021-04-27 David Edelsohn <dje.gcc@gmail.com>
28366 * calls.c (precompute_register_parameters): Additionally test
28367 targetm.precompute_tls_p to pre-compute argument.
28368 * config/rs6000/aix.h (TARGET_PRECOMPUTE_TLS_P): Define.
28369 * config/rs6000/rs6000.c (rs6000_aix_precompute_tls_p): New.
28370 * target.def (precompute_tls_p): New.
28371 * doc/tm.texi.in (TARGET_PRECOMPUTE_TLS_P): Add hook documentation.
28372 * doc/tm.texi: Regenerated.
28374 2021-04-27 Jakub Jelinek <jakub@redhat.com>
28377 * config/aarch64/aarch64.c (aarch64_print_operand): Cast -UINTVAL
28378 back to HOST_WIDE_INT.
28380 2021-04-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
28383 * simplify-rtx.c (simplify_context::simplify_subreg): Check the
28384 memory alignment for the outer mode.
28386 2021-04-27 H.J. Lu <hjl.tools@gmail.com>
28388 PR middle-end/90773
28389 * expr.c (op_by_pieces_d::get_usable_mode): New member function.
28390 (op_by_pieces_d::run): Cange a while loop to a do-while loop.
28392 2021-04-27 Alex Coplan <alex.coplan@arm.com>
28395 * config/arm/arm.c (arm_split_compare_and_swap): Fix up codegen
28396 with negative immediates: ensure we expand cbranchsi4_scratch
28397 correctly and ensure we satisfy its constraints.
28398 * config/arm/sync.md
28399 (@atomic_compare_and_swap<CCSI:arch><NARROW:mode>_1): Don't
28400 attempt to tie two output operands together with constraints;
28401 collapse two alternatives.
28402 (@atomic_compare_and_swap<CCSI:arch><SIDI:mode>_1): Likewise.
28403 * config/arm/thumb1.md (cbranchsi4_neg_late): New.
28405 2021-04-27 Jakub Jelinek <jakub@redhat.com>
28408 * config/aarch64/predicates.md (aarch64_sub_immediate,
28409 aarch64_plus_immediate): Use -UINTVAL instead of -INTVAL.
28410 * config/aarch64/aarch64.md (casesi, rotl<mode>3): Likewise.
28411 * config/aarch64/aarch64.c (aarch64_print_operand,
28412 aarch64_split_atomic_op, aarch64_expand_subvti): Likewise.
28414 2021-04-27 Jakub Jelinek <jakub@redhat.com>
28416 PR tree-optimization/100239
28417 * tree-vect-generic.c (lower_vec_perm): Don't accept constant
28418 permutations with all indices from the first zero element as vec_shl.
28420 2021-04-27 Jakub Jelinek <jakub@redhat.com>
28422 PR rtl-optimization/100254
28423 * cfgcleanup.c (outgoing_edges_match): Check REG_EH_REGION on
28424 last1 and last2 insns rather than BB_END (bb1) and BB_END (bb2) insns.
28426 2021-04-27 Richard Biener <rguenther@suse.de>
28428 PR tree-optimization/99912
28429 * passes.def: Add comment about new TODO_remove_unused_locals.
28430 * tree-stdarg.c (pass_data_stdarg): Run TODO_remove_unused_locals
28433 2021-04-27 Richard Biener <rguenther@suse.de>
28435 PR tree-optimization/99912
28436 * passes.def (pass_all_optimizations): Add pass_dse before
28437 the first pass_dce, move the first pass_dse before the
28438 pass_dce following pass_pre.
28440 2021-04-27 Jakub Jelinek <jakub@redhat.com>
28442 PR tree-optimization/95527
28443 * generic-match-head.c: Include tm.h.
28444 * gimple-match-head.c: Include tm.h.
28445 * match.pd (CLZ == INTEGER_CST): Don't use
28446 #ifdef CLZ_DEFINED_VALUE_AT_ZERO, only test CLZ_DEFINED_VALUE_AT_ZERO
28447 if clz == CFN_CLZ. Add missing val declaration.
28448 (CTZ cmp CST): New simplifications.
28450 2021-04-27 Jakub Jelinek <jakub@redhat.com>
28452 PR tree-optimization/96696
28453 * expr.c (expand_expr_divmod): New function.
28454 (expand_expr_real_2) <case TRUNC_DIV_EXPR>: Use it for truncations and
28455 divisions. Formatting fixes.
28456 <case MULT_EXPR>: Optimize x / y * y as x - x % y if the latter is
28459 2021-04-27 Martin Jambor <mjambor@suse.cz>
28462 * ipa-param-manipulation.c (ipa_param_adjustments::modify_call):
28463 If removing a call statement LHS SSA name, release it.
28465 2021-04-27 Richard Earnshaw <rearnsha@arm.com>
28468 * config/arm/arm.c (THUMB2_WORK_REGS): Check PIC_OFFSET_TABLE_REGNUM
28469 is valid before including it in the mask.
28471 2021-04-27 Richard Sandiford <richard.sandiford@arm.com>
28474 * config/aarch64/aarch64.c (aarch64_comp_type_attributes): Handle
28477 2021-04-27 Richard Biener <rguenther@suse.de>
28479 PR tree-optimization/100051
28480 * tree-ssa-alias.c (indirect_ref_may_alias_decl_p): Add
28481 disambiguator based on access size vs. decl size.
28483 2021-04-27 Richard Biener <rguenther@suse.de>
28485 PR tree-optimization/100278
28486 * tree-ssa-pre.c (compute_avail): Give up when we cannot
28487 adjust TBAA beacuse of mismatching bases.
28489 2021-04-27 Jakub Jelinek <jakub@redhat.com>
28492 * config/i386/i386.md (*<insn><mode>3_mask, *<insn><mode>3_mask_1):
28493 For any_rotate define_insn_split and following splitters, use
28494 SWI iterator instead of SWI48.
28496 2021-04-27 Richard Biener <rguenther@suse.de>
28498 PR tree-optimization/99776
28499 * match.pd (bit_field_ref (ctor)): Relax element extract
28500 type compatibility checks.
28502 2021-04-27 Cui,Lili <lili.cui@intel.com>
28504 * common/config/i386/i386-common.c (processor_names):
28505 Sync processor_names with processor_type.
28506 * config/i386/i386-options.c (processor_cost_table):
28507 Sync processor_cost_table with processor_type.
28509 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
28511 * value-range.cc (irange::irange_set_1bit_anti_range): Add assert.
28512 (irange::set): Call irange_set_1bit_anti_range for handling all
28513 1-bit ranges. Fall through on ~[MIN,MAX].
28515 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
28517 * value-range.cc (irange::legacy_num_pairs): Remove.
28518 (irange::invert): Change gcc_assert to gcc_checking_assert.
28519 * value-range.h (irange::num_pairs): Adjust for a cached
28520 num_pairs(). Also, rename all gcc_assert's to
28521 gcc_checking_assert's.
28523 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
28525 * value-range.cc (irange::operator=): Set m_kind.
28526 (irange::copy_to_legacy): Handle varying and undefined sources
28527 as a legacy copy since they can be easily copied.
28528 (irange::irange_set): Set m_kind.
28529 (irange::irange_set_anti_range): Same.
28530 (irange::set): Rename normalize_min_max to normalize_kind.
28531 (irange::verify_range): Adjust for multi-ranges having the
28533 (irange::irange_union): Set m_kind.
28534 (irange::irange_intersect): Same.
28535 (irange::invert): Same.
28536 * value-range.h (irange::kind): Always return m_kind.
28537 (irange::varying_p): Rename to...
28538 (irange::varying_comptaible_p): ...this.
28539 (irange::undefined_p): Only look at m_kind.
28540 (irange::irange): Always set VR_UNDEFINED if applicable.
28541 (irange::set_undefined): Always set VR_UNDEFINED.
28542 (irange::set_varying): Always set m_kind to VR_VARYING.
28543 (irange::normalize_min_max): Rename to...
28544 (irange::normalize_kind): ...this.
28546 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
28548 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::set_ssa_range_info):
28549 Adjust for constant_p including varying_p.
28550 * tree-vrp.c (vrp_prop::finalize): Same.
28551 (determine_value_range): Same.
28552 * vr-values.c (vr_values::range_of_expr): Same.
28553 * value-range.cc (irange::symbolic_p): Do not check varying_p.
28554 (irange::constant_p): Same.
28556 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
28558 * value-range.cc (irange::legacy_lower_bound): Replace
28559 !undefined_p check with num_ranges > 0.
28560 (irange::legacy_upper_bound): Same.
28561 * value-range.h (irange::type): Same.
28562 (irange::lower_bound): Same.
28563 (irange::upper_bound): Same.
28565 2021-04-26 Richard Biener <rguenther@suse.de>
28567 PR tree-optimization/99956
28568 * gimple-loop-interchange.cc (compute_access_stride):
28569 Try instantiating the access in a shallower loop nest
28570 if instantiating failed.
28571 (compute_access_strides): Pass adjustable loop_nest
28572 to compute_access_stride.
28574 2021-04-26 Christophe Lyon <christophe.lyon@linaro.org>
28576 * doc/sourcebuild.texi (arm_cmse_hw): Document.
28578 2021-04-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
28580 * config/aarch64/iterators.md (vwcore): Handle V4BF, V8BF.
28582 2021-04-26 Thomas Schwinge <thomas@codesourcery.com>
28583 Nathan Sidwell <nathan@codesourcery.com>
28584 Tom de Vries <vries@codesourcery.com>
28585 Julian Brown <julian@codesourcery.com>
28586 Kwok Cheung Yeung <kcy@codesourcery.com>
28588 * omp-offload.c (oacc_validate_dims): Implement
28589 '-Wopenacc-parallelism'.
28590 * doc/invoke.texi (-Wopenacc-parallelism): Document.
28592 2021-04-26 Richard Biener <rguenther@suse.de>
28594 * tree-cfg.h (gimplify_build1): Remove.
28595 (gimplify_build2): Likewise.
28596 (gimplify_build3): Likewise.
28597 * tree-cfg.c (gimplify_build1): Move to tree-vect-generic.c.
28598 (gimplify_build2): Likewise.
28599 (gimplify_build3): Likewise.
28600 * tree-vect-generic.c (gimplify_build1): Move from tree-cfg.c.
28602 (gimplify_build2): Likewise.
28603 (gimplify_build3): Likewise.
28604 (tree_vec_extract): Use resimplify with following SSA edges.
28605 (expand_vector_parallel): Avoid passing NULL size/bitpos
28606 to tree_vec_extract.
28607 * expr.c (store_constructor): Deal with zero-element CTORs.
28608 * match.pd (bit_field_ref <vector CTOR>): Make sure to
28609 produce vector constants when possible.
28611 2021-04-26 Richard Biener <rguenther@suse.de>
28613 * tree-complex.c: Include gimple-fold.h.
28614 (expand_complex_addition): Use gimple_build.
28615 (expand_complex_multiplication_components): Likewise.
28616 (expand_complex_multiplication): Likewise.
28617 (expand_complex_div_straight): Likewise.
28618 (expand_complex_div_wide): Likewise.
28619 (expand_complex_division): Likewise.
28620 (expand_complex_conjugate): Likewise.
28621 (expand_complex_comparison): Likewise.
28623 2021-04-26 Richard Biener <rguenther@suse.de>
28625 * tree-ssa-phiopt.c (two_value_replacement): Remove use
28626 of legacy gimplify_buildN API.
28628 2021-04-26 Richard Biener <rguenther@suse.de>
28630 PR tree-optimization/99473
28631 * tree-ssa-phiopt.c (cond_store_replacement): Handle all
28634 2021-04-26 Richard Biener <rguenther@suse.de>
28636 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_builtin):
28637 Use replace_call_with_value.
28639 2021-04-26 Richard Biener <rguenther@suse.de>
28641 * tree-ssa-propagate.h (valid_gimple_rhs_p): Remove.
28642 (update_gimple_call): Likewise.
28643 (update_call_from_tree): Likewise.
28644 * tree-ssa-propagate.c (valid_gimple_rhs_p): Remove.
28645 (valid_gimple_call_p): Likewise.
28646 (move_ssa_defining_stmt_for_defs): Likewise.
28647 (finish_update_gimple_call): Likewise.
28648 (update_gimple_call): Likewise.
28649 (update_call_from_tree): Likewise.
28650 (propagate_tree_value_into_stmt): Use replace_call_with_value.
28651 * gimple-fold.h (update_gimple_call): Declare.
28652 * gimple-fold.c (valid_gimple_rhs_p): Move here from
28653 tree-ssa-propagate.c.
28654 (update_gimple_call): Likewise.
28655 (valid_gimple_call_p): Likewise.
28656 (finish_update_gimple_call): Likewise, and simplify.
28657 (gimplify_and_update_call_from_tree): Implement
28658 update_call_from_tree functionality, avoid excessive
28659 push/pop_gimplify_context.
28660 (gimple_fold_builtin): Use only gimplify_and_update_call_from_tree.
28661 (gimple_fold_call): Likewise.
28662 * gimple-ssa-sprintf.c (try_substitute_return_value): Likewise.
28663 * tree-ssa-ccp.c (ccp_folder::fold_stmt): Likewise.
28664 (pass_fold_builtins::execute): Likewise.
28665 (optimize_stack_restore): Use replace_call_with_value.
28666 * tree-cfg.c (fold_loop_internal_call): Likewise.
28667 * tree-ssa-dce.c (maybe_optimize_arith_overflow): Use
28668 only gimplify_and_update_call_from_tree.
28669 * tree-ssa-strlen.c (handle_builtin_strlen): Likewise.
28670 (handle_builtin_strchr): Likewise.
28671 * tsan.c: Include gimple-fold.h instead of tree-ssa-propagate.h.
28673 2021-04-26 Jakub Jelinek <jakub@redhat.com>
28676 * vmsdbgout.c (ASM_OUTPUT_DEBUG_STRING, vmsdbgout_begin_block,
28677 vmsdbgout_end_block, lookup_filename, vmsdbgout_source_line): Remove
28680 2021-04-25 liuhongt <hongtao.liu@intel.com>
28683 * config/i386/i386-builtin.def (BDESC): Change the icode of
28684 the following builtins to CODE_FOR_nothing.
28685 * config/i386/i386.c (ix86_gimple_fold_builtin): Fold
28686 IX86_BUILTIN_PCMPEQB128, IX86_BUILTIN_PCMPEQW128,
28687 IX86_BUILTIN_PCMPEQD128, IX86_BUILTIN_PCMPEQQ,
28688 IX86_BUILTIN_PCMPEQB256, IX86_BUILTIN_PCMPEQW256,
28689 IX86_BUILTIN_PCMPEQD256, IX86_BUILTIN_PCMPEQQ256,
28690 IX86_BUILTIN_PCMPGTB128, IX86_BUILTIN_PCMPGTW128,
28691 IX86_BUILTIN_PCMPGTD128, IX86_BUILTIN_PCMPGTQ,
28692 IX86_BUILTIN_PCMPGTB256, IX86_BUILTIN_PCMPGTW256,
28693 IX86_BUILTIN_PCMPGTD256, IX86_BUILTIN_PCMPGTQ256.
28694 * config/i386/sse.md (avx2_eq<mode>3): Deleted.
28695 (sse2_eq<mode>3): Ditto.
28696 (sse4_1_eqv2di3): Ditto.
28697 (sse2_gt<mode>3): Rename to ..
28698 (*sse2_gt<mode>3): .. this.
28700 2021-04-24 Iain Sandoe <iain@sandoe.co.uk>
28703 2021-04-24 Iain Sandoe <iain@sandoe.co.uk>
28706 * config/darwin.c (darwin_binds_local_p): Assume that any
28707 public symbol might be interposed for PIC code. Update function
28708 header comment to reflect current Darwin capability.
28710 2021-04-24 Iain Sandoe <iain@sandoe.co.uk>
28713 * config/darwin.c (darwin_binds_local_p): Assume that any
28714 public symbol might be interposed for PIC code. Update function
28715 header comment to reflect current Darwin capability.
28717 2021-04-24 Richard Sandiford <richard.sandiford@arm.com>
28719 * doc/sourcebuild.texi: Document no-opts and any-opts target
28722 2021-04-23 YiFei Zhu <zhuyifei1999@gmail.com>
28724 * config/bpf/bpf.h (ASM_OUTPUT_ALIGNED_BSS): Use .type and .lcomm.
28726 2021-04-23 YiFei Zhu <zhuyifei1999@gmail.com>
28728 * config/bpf/bpf.h (FUNCTION_BOUNDARY): Set to 64.
28730 2021-04-23 Uroš Bizjak <ubizjak@gmail.com>
28733 * config/i386/i386-options.c (ix86_option_override_internal):
28734 Error out when -m96bit-long-double is used with 64bit targets.
28735 * config/i386/i386.md (*pushxf_rounded): Remove pattern.
28737 2021-04-23 Martin Liska <mliska@suse.cz>
28739 * lto-wrapper.c: Remove FIXME about usage of
28740 hardware_concurrency. The function is not on par with
28743 2021-04-23 Uroš Bizjak <ubizjak@gmail.com>
28746 * config/i386/sync.md (FILD_ATOMIC/FIST_ATOMIC FP load peephole2):
28747 Copy operand 3 to operand 4. Use sse_reg_operand
28748 as operand 3 predicate.
28749 (FILD_ATOMIC/FIST_ATOMIC FP load peephole2 with mem blockage): Ditto.
28750 (LDX_ATOMIC/STX_ATOMIC FP load peephole2): Ditto.
28751 (LDX_ATOMIC/LDX_ATOMIC FP load peephole2 with mem blockage): Ditto.
28752 (FILD_ATOMIC/FIST_ATOMIC FP store peephole2):
28753 Copy operand 1 to operand 0.
28754 (FILD_ATOMIC/FIST_ATOMIC FP store peephole2 with mem blockage): Ditto.
28755 (LDX_ATOMIC/STX_ATOMIC FP store peephole2): Ditto.
28756 (LDX_ATOMIC/LDX_ATOMIC FP store peephole2 with mem blockage): Ditto.
28758 2021-04-23 Alex Coplan <alex.coplan@arm.com>
28760 PR rtl-optimization/100230
28761 * early-remat.c (early_remat::sort_candidates): Use delete[]
28762 instead of delete for array allocated with new[].
28764 2021-04-23 Richard Biener <rguenther@suse.de>
28766 * genmatch.c (lower_cond): Remove VEC_COND_EXPR special-casing.
28767 (capture_info::capture_info): Likewise.
28768 (capture_info::walk_match): Likewise.
28769 (expr::gen_transform): Likewise.
28770 (dt_simplify::gen_1): Likewise.
28771 * gimple-match-head.c (maybe_resimplify_conditional_op):
28772 Remove VEC_COND_EXPR special-casing.
28773 (gimple_simplify): Likewise.
28774 * gimple.c (gimple_could_trap_p_1): Adjust.
28775 * tree-ssa-pre.c (compute_avail): Allow VEC_COND_EXPR
28776 to participate in PRE.
28778 2021-04-23 Richard Biener <rguenther@suse.de>
28780 * cfganal.c (connect_infinite_loops_to_exit): First call
28781 add_noreturn_fake_exit_edges.
28782 * ipa-sra.c (process_scan_results): Do not call the now redundant
28783 add_noreturn_fake_exit_edges.
28784 * predict.c (tree_estimate_probability): Likewise.
28785 (rebuild_frequencies): Likewise.
28786 * store-motion.c (one_store_motion_pass): Likewise.
28788 2021-04-23 Richard Biener <rguenther@suse.de>
28790 PR tree-optimization/100222
28791 * predict.c (pass_profile::execute): Remove redundant call to
28792 mark_irreducible_loops.
28793 (report_predictor_hitrates): Likewise.
28795 2021-04-23 Richard Biener <rguenther@suse.de>
28797 * tree-ssa-loop-ivopts.c (rewrite_use_nonlinear_expr): Avoid
28798 valid_gimple_rhs_p by instead gimplifying to one.
28800 2021-04-23 Richard Biener <rguenther@suse.de>
28802 PR tree-optimization/99971
28803 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences):
28804 Always use TBAA for loads.
28806 2021-04-23 liuhongt <hongtao.liu@intel.com>
28809 * config/i386/i386-options.c (ix86_option_override_internal):
28810 Clear MASK_AVX256_SPLIT_UNALIGNED_LOAD/STORE in x_target_flags
28811 when X86_TUNE_AVX256_UNALIGNED_LOAD/STORE_OPTIMAL is enabled
28812 by target attribute.
28814 2021-04-23 David Edelsohn <dje.gcc@gmail.com>
28816 * config/rs6000/aix71.h (PREFERRED_DEBUGGING_TYPE): Change to
28818 * config/rs6000/aix72.h (PREFERRED_DEBUGGING_TYPE): Same.
28820 2021-04-22 David Edelsohn <dje.gcc@gmail.com>
28822 * config.gcc (powerpc-ibm-aix6.*): Remove.
28823 * config/rs6000/aix61.h: Delete.
28825 2021-04-22 Martin Liska <mliska@suse.cz>
28827 PR testsuite/100159
28828 PR testsuite/100192
28829 * builtins.c (expand_builtin): Fix typos and missing comments.
28830 * dwarf2out.c (gen_subprogram_die): Likewise.
28831 (gen_struct_or_union_type_die): Likewise.
28833 2021-04-22 Uroš Bizjak <ubizjak@gmail.com>
28836 * config/i386/i386-expand.c (ix86_expand_convert_uns_sidf_sse):
28837 Remove the sign with FE_DOWNWARD, where x - x = -0.0.
28839 2021-04-21 Iain Sandoe <iain@sandoe.co.uk>
28841 * config/i386/darwin.h (TARGET_64BIT): Remove definition
28842 based on TARGET_ISA_64BIT.
28843 (TARGET_64BIT_P): Remove definition based on
28844 TARGET_ISA_64BIT_P().
28846 2021-04-21 Martin Liska <mliska@suse.cz>
28849 2021-04-21 Martin Liska <mliska@suse.cz>
28851 * lto-wrapper.c (cpuset_popcount): Remove.
28852 (init_num_threads): Remove and use hardware_concurrency.
28854 2021-04-21 Martin Liska <mliska@suse.cz>
28857 * main.c (main): Call toplev::finalize in CHECKING_P mode.
28858 * ipa-modref.c (ipa_modref_c_finalize): summaries are NULL
28859 when incremental LTO linking happens.
28861 2021-04-21 Martin Liska <mliska@suse.cz>
28863 * lto-wrapper.c (run_gcc): When -flto=jobserver is used, but the
28864 makeserver cannot be detected, then use -flto=N fallback.
28866 2021-04-21 Richard Sandiford <richard.sandiford@arm.com>
28868 * acinclude.m4 (gcc_AC_INITFINI_ARRAY): When cross-compiling,
28869 default to yes for aarch64-linux-gnu.
28870 * configure: Regenerate.
28872 2021-04-21 Martin Liska <mliska@suse.cz>
28874 * lto-wrapper.c (cpuset_popcount): Remove.
28875 (init_num_threads): Remove and use hardware_concurrency.
28877 2021-04-21 Martin Liska <mliska@suse.cz>
28879 * config/i386/i386.c: Remove superfluous || TARGET_MACHO
28880 which remains to be '(... || 0)' and clang complains about it.
28881 * dwarf2out.c (AT_vms_delta): Declare conditionally.
28882 (add_AT_vms_delta): Likewise.
28883 * tree.c (fld_simplified_type): Use rather more common pattern
28884 for disabling of something (#if 0).
28885 (get_tree_code_name): Likewise.
28886 (verify_type_variant): Likewise.
28888 2021-04-21 Martin Liska <mliska@suse.cz>
28890 * config/i386/i386-expand.c (decide_alignment): Use newly named
28891 macro TARGET_CPU_P.
28892 * config/i386/i386.c (ix86_decompose_address): Likewise.
28893 (ix86_address_cost): Likewise.
28894 (ix86_lea_outperforms): Likewise.
28895 (ix86_avoid_lea_for_addr): Likewise.
28896 (ix86_add_stmt_cost): Likewise.
28897 * config/i386/i386.h (TARGET_*): Remove.
28898 (TARGET_CPU_P): New macro.
28899 * config/i386/i386.md: Use newly named macro TARGET_CPU_P.
28900 * config/i386/x86-tune-sched-atom.c (do_reorder_for_imul): Likewise.
28901 (swap_top_of_ready_list): Likewise.
28902 (ix86_atom_sched_reorder): Likewise.
28903 * config/i386/x86-tune-sched-bd.c (ix86_bd_has_dispatch): Likewise.
28904 * config/i386/x86-tune-sched.c (ix86_adjust_cost): Likewise.
28906 2021-04-21 Martin Liska <mliska@suse.cz>
28908 * config/i386/i386-options.c (TARGET_EXPLICIT_NO_SAHF_P):
28910 (SET_TARGET_NO_SAHF): Likewise.
28911 (TARGET_EXPLICIT_PREFETCH_SSE_P): Likewise.
28912 (SET_TARGET_PREFETCH_SSE): Likewise.
28913 (TARGET_EXPLICIT_NO_TUNE_P): Likewise.
28914 (SET_TARGET_NO_TUNE): Likewise.
28915 (TARGET_EXPLICIT_NO_80387_P): Likewise.
28916 (SET_TARGET_NO_80387): Likewise.
28918 * config/i386/i386.h (TARGET_*): Remove.
28919 * opth-gen.awk: Generate new used macros.
28921 2021-04-21 Martin Liska <mliska@suse.cz>
28923 * config/i386/i386.h (PTA_*): Remove.
28924 (enum pta_flag): New.
28925 (DEF_PTA): Generate PTA_* values from i386-isa.def.
28926 * config/i386/i386-isa.def: New file.
28928 2021-04-21 Alex Coplan <alex.coplan@arm.com>
28931 * config/aarch64/aarch64-bti-insert.c (aarch64_bti_j_insn_p): New.
28932 (rest_of_insert_bti): Avoid inserting duplicate bti j insns for
28933 jump table targets.
28935 2021-04-21 H.J. Lu <hjl.tools@gmail.com>
28937 * config.gcc: Install mwaitintrin.h for i[34567]86-*-* and
28938 x86_64-*-* targets.
28939 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_MWAIT_SET):
28941 (OPTION_MASK_ISA2_MWAIT_UNSET): Likewise.
28942 (ix86_handle_option): Handle -mmwait.
28943 * config/i386/i386-builtins.c (ix86_init_mmx_sse_builtins):
28944 Replace OPTION_MASK_ISA_SSE3 with OPTION_MASK_ISA2_MWAIT on
28945 __builtin_ia32_monitor and __builtin_ia32_mwait.
28946 * config/i386/i386-options.c (isa2_opts): Add -mmwait.
28947 (ix86_valid_target_attribute_inner_p): Likewise.
28948 (ix86_option_override_internal): Enable mwait/monitor
28949 instructions for -msse3.
28950 * config/i386/i386.h (TARGET_MWAIT): New.
28951 (TARGET_MWAIT_P): Likewise.
28952 * config/i386/i386.opt: Add -mmwait.
28953 * config/i386/mwaitintrin.h: New file.
28954 * config/i386/pmmintrin.h: Include <mwaitintrin.h>.
28955 * config/i386/sse.md (sse3_mwait): Replace TARGET_SSE3 with
28957 (@sse3_monitor_<mode>): Likewise.
28958 * config/i386/x86gprintrin.h: Include <mwaitintrin.h>.
28959 * doc/extend.texi: Document mwait target attribute.
28960 * doc/invoke.texi: Document -mmwait.
28962 2021-04-21 Martin Liska <mliska@suse.cz>
28964 * config/i386/i386-options.c (DEF_ENUM): Remove it.
28965 * config/i386/i386-opts.h (DEF_ENUM): Likewise.
28966 * config/i386/stringop.def (DEF_ENUM): Likewise.
28968 2021-04-21 Martin Liska <mliska@suse.cz>
28970 * tree-cfg.c (gimple_verify_flow_info): Use qD instead
28971 of print_generic_expr.
28973 2021-04-21 Jakub Jelinek <jakub@redhat.com>
28975 PR rtl-optimization/100148
28976 * cprop.c (constprop_register): Use next_nondebug_insn instead of
28979 2021-04-21 Martin Liska <mliska@suse.cz>
28982 * cgraphunit.c (cgraph_node::analyze): Remove duplicate
28983 free_dominance_info calls.
28985 2021-04-21 Richard Biener <rguenther@suse.de>
28987 * gimple-fold.c (maybe_fold_reference): Remove is_lhs
28988 parameter (and assume it to be false).
28989 (fold_gimple_assign): Adjust, remove all callers of
28990 maybe_fold_reference calling it with is_lhs true.
28991 (gimple_fold_call): Likewise.
28992 (fold_stmt_1): Likewise.
28994 2021-04-21 Richard Biener <rguenther@suse.de>
28996 * fold-const.c (pedantic_non_lvalue_loc): Remove.
28997 (fold_binary_loc): Adjust.
28998 (fold_ternary_loc): Likewise.
29000 2021-04-21 Richard Sandiford <richard.sandiford@arm.com>
29002 PR middle-end/100130
29003 * varasm.c (get_block_for_decl): Make sure that any use of the
29004 retain attribute matches the section's retain flag.
29005 (switch_to_section): Check for retain mismatches even when
29006 changing sections, but do not warn if the given decl is the
29007 section's named.decl.
29008 (output_object_block): Pass the first decl in the block (if any)
29009 to switch_to_section.
29011 2021-04-20 H.J. Lu <hjl.tools@gmail.com>
29013 * config/i386/i386-c.c (ix86_target_macros_internal): Define
29014 __CRC32__ for -mcrc32.
29015 * config/i386/i386-options.c (ix86_option_override_internal):
29016 Enable crc32 instruction for -msse4.2.
29017 * config/i386/i386.md (sse4_2_crc32<mode>): Remove TARGET_SSE4_2
29019 (sse4_2_crc32di): Likewise.
29020 * config/i386/ia32intrin.h: Use crc32 target option for CRC32
29023 2021-04-20 Segher Boessenkool <segher@kernel.crashing.org>
29026 * config/rs6000/rs6000.c (rs6000_machine_from_flags): Do not consider
29029 2021-04-20 Martin Liska <mliska@suse.cz>
29031 * doc/invoke.texi: Fix typo.
29032 * params.opt: Likewise.
29034 2021-04-20 Martin Liska <mliska@suse.cz>
29036 * doc/invoke.texi: Document new param.
29038 2021-04-19 Andrew MacLeod <amacleod@redhat.com>
29040 PR tree-optimization/100081
29041 * gimple-range-cache.h (ranger_cache): Inherit from gori_compute
29042 rather than gori_compute_cache.
29043 * gimple-range-gori.cc (is_gimple_logical_p): Move to top of file.
29044 (range_def_chain::m_logical_depth): New member.
29045 (range_def_chain::range_def_chain): Initialize m_logical_depth.
29046 (range_def_chain::get_def_chain): Don't build defchains through more
29047 than LOGICAL_LIMIT logical expressions.
29048 * params.opt (param_ranger_logical_depth): New.
29050 2021-04-19 Richard Earnshaw <rearnsha@arm.com>
29053 * config/arm/arm.c (arm_configure_build_target): Do not strip
29054 extended FPU/SIMD feature bits from the target ISA when -mfpu
29055 is specified (partial revert of r11-8168).
29057 2021-04-19 Thomas Schwinge <thomas@codesourcery.com>
29059 * params.opt (-param=openacc-kernels=): Add.
29060 * omp-oacc-kernels-decompose.cc
29061 (pass_omp_oacc_kernels_decompose::gate): Use it.
29062 * doc/invoke.texi (-fopenacc-kernels=@var{mode}): Move...
29063 (--param): ... here, 'openacc-kernels'.
29065 2021-04-19 Martin Liska <mliska@suse.cz>
29068 * gengtype.c (finish_root_table): Align function arguments
29069 in between declaration and definition.
29071 2021-04-19 Eric Botcazou <ebotcazou@adacore.com>
29073 * config/i386/winnt.c (i386_pe_seh_cold_init): Properly deal with
29074 frames larger than the SEH maximum frame size.
29076 2021-04-18 Segher Boessenkool <segher@kernel.crashing.org>
29078 PR rtl-optimization/99927
29079 * combine.c (distribute_notes) [REG_UNUSED]: If the register already
29080 is dead, just drop it.
29082 2021-04-17 Iain Buclaw <ibuclaw@gdcproject.org>
29085 * config/i386/winnt-d.c (TARGET_D_TEMPLATES_ALWAYS_COMDAT): Define.
29086 * doc/tm.texi: Regenerate.
29087 * doc/tm.texi.in (D language and ABI): Add @hook for
29088 TARGET_D_TEMPLATES_ALWAYS_COMDAT.
29090 2021-04-17 Iain Buclaw <ibuclaw@gdcproject.org>
29092 * config/darwin-d.c (darwin_d_handle_target_object_format): New
29094 (darwin_d_register_target_info): New function.
29095 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
29096 * config/dragonfly-d.c (dragonfly_d_handle_target_object_format): New
29098 (dragonfly_d_register_target_info): New function.
29099 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
29100 * config/freebsd-d.c (freebsd_d_handle_target_object_format): New
29102 (freebsd_d_register_target_info): New function.
29103 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
29104 * config/glibc-d.c (glibc_d_handle_target_object_format): New
29106 (glibc_d_register_target_info): New function.
29107 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
29108 * config/i386/i386-d.c (ix86_d_handle_target_object_format): New
29110 (ix86_d_register_target_info): Add ix86_d_handle_target_object_format
29111 as handler for objectFormat key.
29112 * config/i386/winnt-d.c (winnt_d_handle_target_object_format): New
29114 (winnt_d_register_target_info): New function.
29115 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
29116 * config/netbsd-d.c (netbsd_d_handle_target_object_format): New
29118 (netbsd_d_register_target_info): New function.
29119 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
29120 * config/openbsd-d.c (openbsd_d_handle_target_object_format): New
29122 (openbsd_d_register_target_info): New function.
29123 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
29124 * config/pa/pa-d.c (pa_d_handle_target_object_format): New function.
29125 (pa_d_register_target_info): Add pa_d_handle_target_object_format as
29126 handler for objectFormat key.
29127 * config/rs6000/rs6000-d.c (rs6000_d_handle_target_object_format): New
29129 (rs6000_d_register_target_info): Add
29130 rs6000_d_handle_target_object_format as handler for objectFormat key.
29131 * config/sol2-d.c (solaris_d_handle_target_object_format): New
29133 (solaris_d_register_target_info): New function.
29134 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
29136 2021-04-16 Jakub Jelinek <jakub@redhat.com>
29139 * config/aarch64/aarch64.c (aarch64_function_arg_alignment): Change
29140 abi_break argument from bool * to unsigned *, store there the pre-GCC 9
29142 (aarch64_layout_arg, aarch64_gimplify_va_arg_expr): Adjust callers.
29143 (aarch64_function_arg_regno_p): Likewise. Only emit -Wpsabi note if
29144 the old and new alignment after applying MIN/MAX to it is different.
29146 2021-04-16 Tamar Christina <tamar.christina@arm.com>
29149 * config/aarch64/aarch64-sve.md (@aarch64_sve_trn1_conv<mode>): New.
29150 * config/aarch64/aarch64.c (aarch64_expand_sve_const_pred_trn): Use new
29152 * config/aarch64/iterators.md (UNSPEC_TRN1_CONV): New.
29154 2021-04-16 Bill Schmidt <wschmidt@linux.ibm.com>
29156 * doc/extend.texi (PowerPC AltiVec/VSX Built-in Functions): Revise
29157 this section and its subsections.
29159 2021-04-16 Jakub Jelinek <jakub@redhat.com>
29162 * config/aarch64/aarch64.md (*neg_asr_si2_extr, *extrsi5_insn_di): New
29163 define_insn patterns.
29165 2021-04-16 Richard Sandiford <richard.sandiford@arm.com>
29167 PR rtl-optimization/98689
29168 * reg-notes.def (UNTYPED_CALL): New note.
29169 * combine.c (distribute_notes): Handle it.
29170 * emit-rtl.c (try_split): Likewise.
29171 * rtlanal.c (rtx_properties::try_to_add_insn): Likewise. Assume
29172 that calls with the note implicitly set all return value registers.
29173 * builtins.c (expand_builtin_apply): Add a REG_UNTYPED_CALL
29176 2021-04-16 Richard Sandiford <richard.sandiford@arm.com>
29178 PR rtl-optimization/99596
29179 * rtlanal.c (rtx_properties::try_to_add_insn): Don't add global
29180 register accesses for const calls. Assume that pure functions
29181 can only read from global registers. Ignore cases in which
29182 the stack pointer has been marked global.
29184 2021-04-16 Jakub Jelinek <jakub@redhat.com>
29187 * tree-vect-loop.c (vect_transform_loop): Don't remove just
29188 dead scalar .MASK_LOAD calls, but also dead .COND_* calls - replace
29189 them by their last argument.
29191 2021-04-15 Martin Liska <mliska@suse.cz>
29193 * doc/invoke.texi: Other params don't use it, remove it.
29195 2021-04-15 Richard Biener <rguenther@suse.de>
29197 * gimple-builder.h: Add deprecation note.
29199 2021-04-15 Richard Sandiford <richard.sandiford@arm.com>
29202 * attribs.h (restrict_type_identity_attributes_to): Declare.
29203 * attribs.c (restrict_type_identity_attributes_to): New function.
29205 2021-04-15 Richard Sandiford <richard.sandiford@arm.com>
29208 * attribs.h (affects_type_identity_attributes): Declare.
29209 * attribs.c (remove_attributes_matching): New function.
29210 (affects_type_identity_attributes): Likewise.
29212 2021-04-15 Jakub Jelinek <jakub@redhat.com>
29215 * config/aarch64/aarch64.md (*<LOGICAL:optab>_<SHIFT:optab><mode>3):
29216 Add combine splitters for *<LOGICAL:optab>_ashl<mode>3 with
29217 ZERO_EXTEND, SIGN_EXTEND or AND.
29219 2021-04-14 Richard Sandiford <richard.sandiford@arm.com>
29221 PR rtl-optimization/99929
29222 * rtl.h (same_vector_encodings_p): New function.
29223 * cse.c (exp_equiv_p): Check that CONST_VECTORs have the same encoding.
29224 * cselib.c (rtx_equal_for_cselib_1): Likewise.
29225 * jump.c (rtx_renumbered_equal_p): Likewise.
29226 * lra-constraints.c (operands_match_p): Likewise.
29227 * reload.c (operands_match_p): Likewise.
29228 * rtl.c (rtx_equal_p_cb, rtx_equal_p): Likewise.
29230 2021-04-14 Richard Sandiford <richard.sandiford@arm.com>
29232 * print-rtl.c (rtx_writer::print_rtx_operand_codes_E_and_V): Print
29233 more information about variable-length CONST_VECTORs.
29235 2021-04-14 Vladimir N. Makarov <vmakarov@redhat.com>
29237 PR rtl-optimization/100066
29238 * lra-constraints.c (split_reg): Check paradoxical_subreg_p for
29239 ordered modes when choosing splitting mode for hard reg.
29241 2021-04-14 Richard Sandiford <richard.sandiford@arm.com>
29244 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector_sel):
29246 (aarch64_expand_sve_const_vector): Use it for nelts_per_pattern==2.
29248 2021-04-14 Andreas Krebbel <krebbel@linux.ibm.com>
29250 * config/s390/s390-builtins.def (O_M5, O_M12, ...): Add new macros
29251 for mask operand types.
29252 (s390_vec_permi_s64, s390_vec_permi_b64, s390_vec_permi_u64)
29253 (s390_vec_permi_dbl, s390_vpdi): Use the M5 type for the immediate
29255 (s390_vec_msum_u128, s390_vmslg): Use the M12 type for the
29257 * config/s390/s390.c (s390_const_operand_ok): Check the new
29258 operand types and generate a list of valid values.
29260 2021-04-14 Iain Buclaw <ibuclaw@gdcproject.org>
29262 * doc/tm.texi: Regenerate.
29263 * doc/tm.texi.in (D language and ABI): Add @hook for
29264 TARGET_D_REGISTER_OS_TARGET_INFO.
29266 2021-04-14 Iain Buclaw <ibuclaw@gdcproject.org>
29268 * config/aarch64/aarch64-d.c (aarch64_d_handle_target_float_abi): New
29270 (aarch64_d_register_target_info): New function.
29271 * config/aarch64/aarch64-protos.h (aarch64_d_register_target_info):
29273 * config/aarch64/aarch64.h (TARGET_D_REGISTER_CPU_TARGET_INFO):
29275 * config/arm/arm-d.c (arm_d_handle_target_float_abi): New function.
29276 (arm_d_register_target_info): New function.
29277 * config/arm/arm-protos.h (arm_d_register_target_info): Declare.
29278 * config/arm/arm.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
29279 * config/i386/i386-d.c (ix86_d_handle_target_float_abi): New function.
29280 (ix86_d_register_target_info): New function.
29281 * config/i386/i386-protos.h (ix86_d_register_target_info): Declare.
29282 * config/i386/i386.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
29283 * config/mips/mips-d.c (mips_d_handle_target_float_abi): New function.
29284 (mips_d_register_target_info): New function.
29285 * config/mips/mips-protos.h (mips_d_register_target_info): Declare.
29286 * config/mips/mips.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
29287 * config/pa/pa-d.c (pa_d_handle_target_float_abi): New function.
29288 (pa_d_register_target_info): New function.
29289 * config/pa/pa-protos.h (pa_d_register_target_info): Declare.
29290 * config/pa/pa.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
29291 * config/riscv/riscv-d.c (riscv_d_handle_target_float_abi): New
29293 (riscv_d_register_target_info): New function.
29294 * config/riscv/riscv-protos.h (riscv_d_register_target_info): Declare.
29295 * config/riscv/riscv.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
29296 * config/rs6000/rs6000-d.c (rs6000_d_handle_target_float_abi): New
29298 (rs6000_d_register_target_info): New function.
29299 * config/rs6000/rs6000-protos.h (rs6000_d_register_target_info):
29301 * config/rs6000/rs6000.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
29302 * config/s390/s390-d.c (s390_d_handle_target_float_abi): New function.
29303 (s390_d_register_target_info): New function.
29304 * config/s390/s390-protos.h (s390_d_register_target_info): Declare.
29305 * config/s390/s390.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
29306 * config/sparc/sparc-d.c (sparc_d_handle_target_float_abi): New
29308 (sparc_d_register_target_info): New function.
29309 * config/sparc/sparc-protos.h (sparc_d_register_target_info): Declare.
29310 * config/sparc/sparc.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
29311 * doc/tm.texi: Regenerate.
29312 * doc/tm.texi.in (D language and ABI): Add @hook for
29313 TARGET_D_REGISTER_CPU_TARGET_INFO.
29315 2021-04-14 Iain Buclaw <ibuclaw@gdcproject.org>
29317 * config/i386/i386-d.c (ix86_d_has_stdcall_convention): New function.
29318 * config/i386/i386-protos.h (ix86_d_has_stdcall_convention): Declare.
29319 * config/i386/i386.h (TARGET_D_HAS_STDCALL_CONVENTION): Define.
29320 * doc/tm.texi: Regenerate.
29321 * doc/tm.texi.in (D language and ABI): Add @hook for
29322 TARGET_D_HAS_STDCALL_CONVENTION.
29324 2021-04-14 Richard Biener <rguenther@suse.de>
29326 * tree-cfg.c (verify_gimple_assign_ternary): Verify that
29327 VEC_COND_EXPRs have a gimple_val condition.
29328 * tree-ssa-propagate.c (valid_gimple_rhs_p): VEC_COND_EXPR
29329 can no longer have a GENERIC condition.
29331 2021-04-14 Richard Earnshaw <rearnsha@arm.com>
29334 * config/arm/arm.c (arm_configure_build_target): Strip isa_all_fpbits
29335 from the isa_delta when -mfpu has been used.
29336 (arm_options_perform_arch_sanity_checks): It's the architecture that
29337 lacks an FPU not the processor.
29339 2021-04-13 Richard Biener <rguenther@suse.de>
29341 PR tree-optimization/100053
29342 * tree-ssa-sccvn.c (vn_nary_op_get_predicated_value): Do
29343 not use optimistic dominance queries for backedges to validate
29345 (dominated_by_p_w_unex): Add parameter to ignore executable
29346 state on backedges.
29347 (rpo_elim::eliminate_avail): Adjust.
29349 2021-04-13 Jakub Jelinek <jakub@redhat.com>
29352 * config/aarch64/aarch64.md (*aarch64_bfxil<mode>_extr,
29353 *aarch64_bfxilsi_extrdi): New define_insn patterns.
29355 2021-04-13 Jakub Jelinek <jakub@redhat.com>
29358 * simplify-rtx.c (simplify_immed_subreg): For MODE_COMPOSITE_P
29359 outermode, return NULL if the result doesn't encode back to the
29360 original byte sequence.
29361 (simplify_gen_subreg): Don't create SUBREGs from constants to
29362 MODE_COMPOSITE_P outermode.
29364 2021-04-12 Jakub Jelinek <jakub@redhat.com>
29366 PR rtl-optimization/99905
29367 * combine.c (expand_compound_operation): If pos + len > modewidth,
29368 perform the right shift by pos in inner_mode and then convert to mode,
29369 instead of trying to simplify a shift of rtx with inner_mode by pos
29370 as if it was a shift in mode.
29372 2021-04-12 Jakub Jelinek <jakub@redhat.com>
29375 * combine.c (simplify_and_const_int_1): Don't optimize varop
29376 away if it has side-effects.
29378 2021-04-12 Martin Liska <mliska@suse.cz>
29380 * doc/extend.texi: Escape @smallexample content.
29382 2021-04-12 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
29384 * config/s390/s390.md ("*movdi_31", "*movdi_64"): Add
29385 alternative in order to load a DFP zero.
29387 2021-04-12 Martin Liska <mliska@suse.cz>
29389 * doc/extend.texi: Be more precise in documentation
29390 of symver attribute.
29392 2021-04-12 Martin Liska <mliska@suse.cz>
29395 * gimplify.c (gimplify_expr): Right now, we unpoison all
29396 variables before a goto <dest>. We should not do it if we are
29399 2021-04-12 Cui,Lili <lili.cui@intel.com>
29401 * common/config/i386/cpuinfo.h (get_intel_cpu): Handle
29403 * common/config/i386/i386-common.c (processor_names): Add
29405 (processor_alias_table): Add rocketlake.
29406 * common/config/i386/i386-cpuinfo.h (processor_subtypes): Add
29407 INTEL_COREI7_ROCKETLAKE.
29408 * config.gcc: Add -march=rocketlake.
29409 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
29411 * config/i386/i386-options.c (m_ROCKETLAKE) : Define.
29412 (processor_cost_table): Add rocketlake cost.
29413 * config/i386/i386.h (ix86_size_cost) : Define
29415 (processor_type) : Add PROCESSOR_ROCKETLAKE.
29416 (PTA_ROCKETLAKE): Ditto.
29417 * doc/extend.texi: Add rocketlake.
29418 * doc/invoke.texi: Add rocketlake.
29420 2021-04-12 Cui,Lili <lili.cui@intel.com>
29422 * config/i386/i386.h (PTA_ALDERLAKE): Change alderlake ISA list.
29423 * config/i386/i386-options.c (m_CORE_AVX2): Add m_ALDERLAKE.
29424 * common/config/i386/cpuinfo.h (get_intel_cpu): Add AlderLake model.
29425 * doc/invoke.texi: Change alderlake ISA list.
29427 2021-04-11 Hafiz Abid Qadeer <abidh@codesourcery.com>
29429 PR middle-end/98088
29430 * omp-expand.c (expand_oacc_collapse_init): Update condition in
29433 2021-04-10 H.J. Lu <hjl.tools@gmail.com>
29436 * config/i386/serializeintrin.h (_serialize): Defined as macro.
29438 2021-04-10 Jakub Jelinek <jakub@redhat.com>
29441 * expr.c (expand_expr_addr_expr_1): Test is_global_var rather than
29442 just TREE_STATIC on COMPOUND_LITERAL_EXPR_DECLs.
29444 2021-04-10 Jakub Jelinek <jakub@redhat.com>
29446 PR middle-end/99989
29447 * gimple-ssa-warn-alloca.c
29448 (alloca_type_and_limit::alloca_type_and_limit): Initialize limit to
29449 0 with integer precision unconditionally.
29451 2021-04-10 Jakub Jelinek <jakub@redhat.com>
29453 PR rtl-optimization/98601
29454 * rtlanal.c (rtx_addr_can_trap_p_1): Allow in assert unknown size
29455 not just for BLKmode, but also for VOIDmode. For STRICT_ALIGNMENT
29456 unaligned_mems handle VOIDmode like BLKmode.
29458 2021-04-10 Jan Hubicka <hubicka@ucw.cz>
29461 * tree.c (free_lang_data_in_decl): Do not release body of
29462 declare_variant_alt.
29464 2021-04-09 Richard Sandiford <richard.sandiford@arm.com>
29466 * config/aarch64/aarch64.c (aarch64_option_restore): If the
29467 architecture was specified explicitly and the tuning wasn't,
29468 tune for the architecture rather than the configured default CPU.
29470 2021-04-09 Richard Sandiford <richard.sandiford@arm.com>
29472 * config/aarch64/aarch64.md (tlsdesc_small_sve_<mode>): Use X30
29473 as the temporary register.
29475 2021-04-09 Martin Liska <mliska@suse.cz>
29477 * doc/extend.texi: Move non-target attributes on the top level.
29479 2021-04-09 Martin Liska <mliska@suse.cz>
29481 * doc/invoke.texi: Document minimum and maximum value of the
29482 argument for both supported compression algorithms.
29484 2021-04-08 David Edelsohn <dje.gcc@gmail.com>
29486 * config/rs6000/rs6000.c (rs6000_xcoff_select_section): Select
29487 TLS BSS before TLS data.
29488 * config/rs6000/xcoff.h (ASM_OUTPUT_TLS_COMMON): Use .comm.
29490 2021-04-08 Richard Sandiford <richard.sandiford@arm.com>
29492 * doc/sourcebuild.texi (stdint_types_mbig_endian): Document.
29494 2021-04-08 Richard Sandiford <richard.sandiford@arm.com>
29496 * match.pd: Extend vec_cond folds to handle shifts.
29498 2021-04-08 Maciej W. Rozycki <macro@orcam.me.uk>
29500 * config/vax/vax.md: Fix comment for `*bit<mode>' pattern's
29503 2021-04-08 Alex Coplan <alex.coplan@arm.com>
29506 * config/arm/iterators.md (MVE_vecs): New.
29507 (V_elem): Also handle V2DF.
29508 * config/arm/mve.md (*mve_mov<mode>): Rename to ...
29509 (*mve_vdup<mode>): ... this. Remove second alternative since
29510 vec_duplicate of const_int is not canonical RTL, and we don't
29511 want to match symbol_refs.
29512 (*mve_vec_duplicate<mode>): Delete (pattern is redundant).
29514 2021-04-08 Xionghu Luo <luoxhu@linux.ibm.com>
29516 * fold-const.c (fold_single_bit_test): Fix typo.
29517 * print-rtl.c (print_rtx_insn_vec): Call print_rtl_single
29520 2021-04-07 Richard Sandiford <richard.sandiford@arm.com>
29522 PR tree-optimization/97513
29523 * tree-vect-slp.c (vect_add_slp_permutation): New function,
29525 (vectorizable_slp_permutation): ...here. Detect cases in which
29526 all VEC_PERM_EXPRs are guaranteed to have the same stepped
29527 permute vector and only generate one permute vector for that case.
29528 Extend that case to handle variable-length vectors.
29530 2021-04-07 Richard Sandiford <richard.sandiford@arm.com>
29532 PR tree-optimization/99873
29533 * tree-vect-slp.c (vect_slp_prefer_store_lanes_p): New function.
29534 (vect_build_slp_instance): Don't split store groups that could
29535 use IFN_STORE_LANES.
29537 2021-04-07 Jakub Jelinek <jakub@redhat.com>
29540 * varasm.c (output_constant_pool_contents): Don't strip name encoding
29541 from XSTR (desc->sym, 0) or from label before passing those to
29544 2021-04-07 Richard Biener <rguenther@suse.de>
29546 PR tree-optimization/99954
29547 * tree-loop-distribution.c: Include tree-affine.h.
29548 (generate_memcpy_builtin): Try using tree-affine to prove
29550 (loop_distribution::classify_builtin_ldst): Always classify
29553 2021-04-07 Richard Biener <rguenther@suse.de>
29555 PR tree-optimization/99947
29556 * tree-vect-loop.c (vectorizable_induction): Pre-allocate
29557 steps vector to avoid pushing elements from the reallocated
29560 2021-04-07 Richard Biener <rguenther@suse.de>
29562 * tree-ssa-sccvn.h (print_vn_reference_ops): Declare.
29563 * tree-ssa-pre.c (print_pre_expr): Factor out VN reference operand
29565 * tree-ssa-sccvn.c (print_vn_reference_ops): ... into this new
29567 (debug_vn_reference_ops): New.
29569 2021-04-07 Bin Cheng <bin.cheng@linux.alibaba.com>
29571 PR tree-optimization/98736
29572 * tree-loop-distribution.c
29573 * (loop_distribution::bb_top_order_init):
29574 Compute RPO with programing order preserved by calling function
29575 rev_post_order_and_mark_dfs_back_seme.
29577 2021-04-06 Vladimir N. Makarov <vmakarov@redhat.com>
29580 * lra-constraints.c (split_reg): Don't check paradoxical_subreg_p.
29581 * lra-lives.c (clear_sparseset_regnos, regnos_in_sparseset_p): New
29583 (process_bb_lives): Don't update biggest mode of hard reg for
29584 implicit in multi-register group. Use the new functions for
29585 updating dead_set and unused_set by register notes.
29587 2021-04-06 Xianmiao Qu <xianmiao_qu@c-sky.com>
29589 * config/csky/csky_pipeline_ck802.md : Use insn reservation name
29592 2021-04-06 H.J. Lu <hjl.tools@gmail.com>
29594 * config/i386/x86-tune-costs.h (skylake_memcpy): Updated.
29595 (skylake_memset): Likewise.
29596 (skylake_cost): Change CLEAR_RATIO to 17.
29597 * config/i386/x86-tune.def (X86_TUNE_PREFER_KNOWN_REP_MOVSB_STOSB):
29598 Replace m_CANNONLAKE, m_ICELAKE_CLIENT, m_ICELAKE_SERVER,
29599 m_TIGERLAKE and m_SAPPHIRERAPIDS with m_SKYLAKE and m_CORE_AVX512.
29601 2021-04-06 Richard Biener <rguenther@suse.de>
29603 PR tree-optimization/99880
29604 * tree-vect-loop.c (maybe_set_vectorized_backedge_value): Only
29605 set vectorized defs of relevant PHIs.
29607 2021-04-06 Richard Biener <rguenther@suse.de>
29609 PR tree-optimization/99924
29610 * tree-vect-slp.c (vect_bb_partition_graph_r): Do not mark
29611 nodes w/o scalar stmts as visited.
29613 2021-04-06 Alex Coplan <alex.coplan@arm.com>
29616 * config/arm/arm.c (arm_libcall_uses_aapcs_base): Also use base
29617 PCS for [su]fix_optab.
29619 2021-04-03 Iain Sandoe <iain@sandoe.co.uk>
29621 * config/darwin.c (machopic_legitimize_pic_address): Check
29622 that the current pic register is one of the hard reg set
29623 before setting liveness.
29625 2021-04-03 Iain Sandoe <iain@sandoe.co.uk>
29627 * config/darwin.c (machopic_legitimize_pic_address): Fix
29628 whitespace, remove unused code.
29630 2021-04-03 Jakub Jelinek <jakub@redhat.com>
29632 PR tree-optimization/99882
29633 * gimple-ssa-store-merging.c (bswap_view_convert): Handle val with
29636 2021-04-03 Jakub Jelinek <jakub@redhat.com>
29638 PR rtl-optimization/99863
29639 * dse.c (replace_read): Drop regs_live argument. Instead of
29640 regs_live, use store_insn->fixed_regs_live if non-NULL,
29641 otherwise punt if insns sequence clobbers or sets any hard
29644 2021-04-03 Jakub Jelinek <jakub@redhat.com>
29647 * targhooks.h (default_print_patchable_function_entry_1): Declare.
29648 * targhooks.c (default_print_patchable_function_entry_1): New function,
29649 copied from default_print_patchable_function_entry with an added flags
29651 (default_print_patchable_function_entry): Rewritten into a small
29652 wrapper around default_print_patchable_function_entry_1.
29653 * config/rs6000/rs6000.c (TARGET_ASM_PRINT_PATCHABLE_FUNCTION_ENTRY):
29655 (rs6000_print_patchable_function_entry): New function.
29657 2021-04-02 Eric Botcazou <ebotcazou@adacore.com>
29659 * doc/invoke.texi (fdelete-dead-exceptions): Minor tweak.
29661 2021-04-01 Jason Merrill <jason@redhat.com>
29664 * common.opt: Document v15 and v16.
29666 2021-04-01 Richard Biener <rguenther@suse.de>
29668 PR tree-optimization/99863
29669 * gimplify.c (gimplify_init_constructor): Recompute vector
29672 2021-04-01 Jakub Jelinek <jakub@redhat.com>
29674 * doc/extend.texi (symver attribute): Fix up syntax errors
29677 2021-04-01 Jakub Jelinek <jakub@redhat.com>
29679 PR tree-optimization/96573
29680 * gimple-ssa-store-merging.c (init_symbolic_number): Handle
29681 also pointer types.
29683 2021-04-01 Richard Biener <rguenther@suse.de>
29685 PR tree-optimization/99856
29686 * tree-vect-patterns.c (vect_recog_over_widening_pattern): Promote
29687 precision to vector element precision.
29689 2021-04-01 Martin Jambor <mjambor@suse.cz>
29691 PR tree-optimization/97009
29692 * tree-sra.c (access_or_its_child_written): New function.
29693 (propagate_subaccesses_from_rhs): Use it instead of a simple grp_write
29696 2021-03-31 Jan Hubicka <hubicka@ucw.cz>
29699 * cif-code.def (USES_COMDAT_LOCAL): Make CIF_FINAL_NORMAL.
29701 2021-03-31 Pat Haugen <pthaugen@linux.ibm.com>
29704 * config/rs6000/altivec.md (xxspltiw_v4si, xxspltiw_v4sf_inst,
29705 xxspltidp_v2df_inst, xxsplti32dx_v4si_inst, xxsplti32dx_v4sf_inst,
29706 xxblend_<mode>, xxpermx_inst, xxeval): Mark prefixed.
29707 * config/rs6000/mma.md (mma_<vvi4i4i8>, mma_<avvi4i4i8>,
29708 mma_<vvi4i4i2>, mma_<avvi4i4i2>, mma_<vvi4i4>, mma_<avvi4i4>,
29709 mma_<pvi4i2>, mma_<apvi4i2>, mma_<vvi4i4i4>, mma_<avvi4i4i4>):
29711 * config/rs6000/rs6000.c (rs6000_final_prescan_insn): Adjust test.
29712 * config/rs6000/rs6000.md (define_attr "maybe_prefixed"): New.
29713 (define_attr "prefixed"): Update initializer.
29715 2021-03-31 Jakub Jelinek <jakub@redhat.com>
29718 * dwarf2out.c (debug_ranges_dwo_section): New variable.
29719 (DW_RANGES_IDX_SKELETON): Define.
29720 (struct dw_ranges): Add begin_entry and end_entry members.
29721 (DEBUG_DWO_RNGLISTS_SECTION): Define.
29722 (add_ranges_num): Adjust r initializer for addition of *_entry
29724 (add_ranges_by_labels): For -gsplit-dwarf and force_direct,
29725 set idx to DW_RANGES_IDX_SKELETON.
29726 (use_distinct_base_address_for_range): New function.
29727 (index_rnglists): Don't set r->idx if it is equal to
29728 DW_RANGES_IDX_SKELETON. Initialize r->begin_entry and
29729 r->end_entry for -gsplit-dwarf if those will be needed by
29731 (output_rnglists): Add DWO argument. If true, switch to
29732 debug_ranges_dwo_section rather than debug_ranges_section.
29733 Adjust l1/l2 label indexes. Only output the offset table when
29734 dwo is true and don't include in there the skeleton range
29735 entry if present. For -gsplit-dwarf, skip ranges that belong
29736 to the other rnglists section. Change return type from void
29737 to bool and return true if there are any range entries for
29738 the other section. For dwarf_split_debug_info use
29739 DW_RLE_startx_endx, DW_RLE_startx_length and DW_RLE_base_addressx
29740 entries instead of DW_RLE_start_end, DW_RLE_start_length and
29741 DW_RLE_base_address. Use use_distinct_base_address_for_range.
29742 (init_sections_and_labels): Initialize debug_ranges_dwo_section
29743 if -gsplit-dwarf and DWARF >= 5. Adjust ranges_section_label
29744 and range_base_label indexes.
29745 (dwarf2out_finish): Call index_rnglists earlier before finalizing
29746 .debug_addr. Never emit DW_AT_rnglists_base attribute. For
29747 -gsplit-dwarf and DWARF >= 5 call output_rnglists up to twice
29748 with different dwo arguments.
29749 (dwarf2out_c_finalize): Clear debug_ranges_dwo_section.
29751 2021-03-31 Richard Sandiford <richard.sandiford@arm.com>
29753 PR tree-optimization/98268
29754 * gimple-fold.c (maybe_canonicalize_mem_ref_addr): Call
29755 recompute_tree_invariant_for_addr_expr after successfully
29756 folding a TARGET_MEM_REF that occurs inside an ADDR_EXPR.
29758 2021-03-31 Richard Sandiford <richard.sandiford@arm.com>
29760 PR tree-optimization/99726
29761 * tree-data-ref.c (create_intersect_range_checks_index): Bail
29762 out if there is more than one access function SCEV for the loop
29765 2021-03-31 Richard Sandiford <richard.sandiford@arm.com>
29767 PR rtl-optimization/97141
29768 PR rtl-optimization/98726
29769 * emit-rtl.c (valid_for_const_vector_p): Return true for
29771 * rtx-vector-builder.h (rtx_vector_builder::step): Return a
29772 poly_wide_int instead of a wide_int.
29773 (rtx_vector_builder::apply_set): Take a poly_wide_int instead
29775 * rtx-vector-builder.c (rtx_vector_builder::apply_set): Likewise.
29776 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Return
29777 false for CONST_VECTORs that cannot be forced to memory.
29778 * config/aarch64/aarch64-simd.md (mov<mode>): If a CONST_VECTOR
29779 is too complex to force to memory, build it up from individual
29782 2021-03-31 Jan Hubicka <jh@suse.cz>
29785 * cgraph.c (cgraph_node::release_body): Fix overactive check.
29787 2021-03-31 Christophe Lyon <christophe.lyon@linaro.org>
29790 * config/arm/vec-common.md (mul<mode>3): Disable on iwMMXT, expect
29793 2021-03-31 H.J. Lu <hjl.tools@gmail.com>
29795 * config/i386/i386-expand.c (expand_set_or_cpymem_via_rep):
29796 For TARGET_PREFER_KNOWN_REP_MOVSB_STOSB, don't convert QImode
29798 (decide_alg): For TARGET_PREFER_KNOWN_REP_MOVSB_STOSB, use
29799 "rep movsb/stosb" only for known sizes.
29800 * config/i386/i386-options.c (processor_cost_table): Use Ice
29801 Lake cost for Cannon Lake, Ice Lake, Tiger Lake, Sapphire
29802 Rapids and Alder Lake.
29803 * config/i386/i386.h (TARGET_PREFER_KNOWN_REP_MOVSB_STOSB): New.
29804 * config/i386/x86-tune-costs.h (icelake_memcpy): New.
29805 (icelake_memset): Likewise.
29806 (icelake_cost): Likewise.
29807 * config/i386/x86-tune.def (X86_TUNE_PREFER_KNOWN_REP_MOVSB_STOSB):
29810 2021-03-31 Richard Sandiford <richard.sandiford@arm.com>
29813 * config/aarch64/aarch64.c
29814 (aarch64_vectorize_preferred_vector_alignment): Query the size
29815 of the provided SVE vector; do not assume that all SVE vectors
29816 have the same size.
29818 2021-03-31 Jan Hubicka <jh@suse.cz>
29821 * cgraph.c (cgraph_node::release_body): Remove all callers and
29823 * cgraphclones.c (cgraph_node::materialize_clone): Do not do it here.
29824 * cgraphunit.c (cgraph_node::expand): And here.
29826 2021-03-31 Martin Liska <mliska@suse.cz>
29828 * ipa-modref.c (analyze_ssa_name_flags): Fix coding style
29829 and one negated condition.
29831 2021-03-31 Jakub Jelinek <jakub@redhat.com>
29832 Richard Sandiford <richard.sandiford@arm.com>
29835 * config/aarch64/aarch64.md (*add<mode>3_poly_1): Swap Uai and Uav
29836 constraints on operands[2] and similarly 0 and rk constraints
29837 on operands[1] corresponding to that.
29839 2021-03-31 Jakub Jelinek <jakub@redhat.com>
29842 * configure.ac (HAVE_LD_BROKEN_PE_DWARF5): New AC_DEFINE if PECOFF
29843 linker doesn't support DWARF sections new in DWARF5.
29844 * config/i386/i386-options.c (ix86_option_override_internal): Default
29845 to dwarf_version 4 if HAVE_LD_BROKEN_PE_DWARF5 for TARGET_PECOFF
29847 * config.in: Regenerated.
29848 * configure: Regenerated.
29850 2021-03-30 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
29853 * config/aarch64/aarch64.c (aarch64_analyze_loop_vinfo): Check for
29854 available issue_info before using it.
29856 2021-03-30 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
29859 * config/aarch64/aarch64.md (sub<mode>3_compare1_imm): Do not allow zero
29862 2021-03-30 Xionghu Luo <luoxhu@linux.ibm.com>
29865 * config/rs6000/altivec.md (altivec_lvsl_reg): Change to ...
29866 (altivec_lvsl_reg_<mode>): ... this.
29867 (altivec_lvsr_reg): Change to ...
29868 (altivec_lvsr_reg_<mode>): ... this.
29869 * config/rs6000/predicates.md (vec_set_index_operand): New.
29870 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
29871 Enable 32bit variable vec_insert for all TARGET_VSX.
29872 * config/rs6000/rs6000.c (rs6000_expand_vector_set_var_p9):
29873 Enable 32bit variable vec_insert for p9 and above.
29874 (rs6000_expand_vector_set_var_p8): Rename to ...
29875 (rs6000_expand_vector_set_var_p7): ... this.
29876 (rs6000_expand_vector_set): Use TARGET_VSX and adjust assert
29878 * config/rs6000/vector.md (vec_set<mode>): Use vec_set_index_operand.
29879 * config/rs6000/vsx.md (xl_len_r): Use gen_altivec_lvsl_reg_di and
29880 gen_altivec_lvsr_reg_di.
29882 2021-03-30 H.J. Lu <hjl.tools@gmail.com>
29885 * config/i386/ia32intrin.h (__rdtsc): Defined as macro.
29886 (__rdtscp): Likewise.
29888 2021-03-30 Tamar Christina <tamar.christina@arm.com>
29890 PR tree-optimization/99825
29891 * tree-vect-slp-patterns.c (vect_check_evenodd_blend):
29892 Reject non-mult 2 lanes.
29894 2021-03-30 Richard Earnshaw <rearnsha@arm.com>
29897 * config/arm/arm.c (arm_file_start): Fix emission of
29898 Tag_ABI_VFP_args attribute.
29900 2021-03-30 Richard Biener <rguenther@suse.de>
29902 PR tree-optimization/99824
29903 * stor-layout.c (set_min_and_max_values_for_integral_type):
29904 Assert the precision is within the bounds of
29905 WIDE_INT_MAX_PRECISION.
29906 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
29907 the outermost component ref only to lower the access size
29908 and initialize that from the access type.
29910 2021-03-30 Richard Sandiford <richard.sandiford@arm.com>
29913 * config/aarch64/aarch64.md (mov<mode>): Pass multi-instruction
29914 CONST_INTs to aarch64_expand_mov_immediate when called after RA.
29916 2021-03-30 Mihailo Stojanovic <mihailo.stojanovic@typhoon-hil.com>
29918 * config/aarch64/aarch64.md
29919 (<optab>_trunc<fcvt_target><GPI:mode>2): Set the "arch"
29920 attribute to disambiguate between SIMD and FP variants of the
29923 2021-03-29 Jan Hubicka <hubicka@ucw.cz>
29925 * ipa-modref.c (merge_call_lhs_flags): Correct handling of deref.
29926 (analyze_ssa_name_flags): Fix typo in comment.
29928 2021-03-29 Alex Coplan <alex.coplan@arm.com>
29931 * config/aarch64/aarch64-sve-builtins.cc
29932 (function_builder::add_function): Add placeholder_p argument, use
29933 placeholder decls if this is set.
29934 (function_builder::add_unique_function): Instead of conditionally adding
29935 direct overloads, unconditionally add either a direct overload or a
29937 (function_builder::add_overloaded_function): Set placeholder_p if we're
29938 using C++ overloads. Use the obstack for string storage instead
29939 of relying on the tree nodes.
29940 (function_builder::add_overloaded_functions): Don't return early for
29941 m_direct_overloads: we need to add placeholders.
29942 * config/aarch64/aarch64-sve-builtins.h
29943 (function_builder::add_function): Add placeholder_p argument.
29945 2021-03-29 Richard Biener <rguenther@suse.de>
29947 PR tree-optimization/99807
29948 * tree-vect-slp.c (vect_slp_analyze_node_operations_1): Move
29949 assert below VEC_PERM handling.
29951 2021-03-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
29954 * config/aarch64/aarch64-simd.md (move_lo_quad_internal_<mode>): Use
29955 aarch64_simd_or_scalar_imm_zero to match zeroes. Remove pattern
29956 matching const_int 0.
29957 (move_lo_quad_internal_be_<mode>): Likewise.
29958 (move_lo_quad_<mode>): Update for the above.
29959 * config/aarch64/iterators.md (VQ_2E): Delete.
29961 2021-03-29 Jakub Jelinek <jakub@redhat.com>
29963 PR tree-optimization/99777
29964 * fold-const.c (extract_muldiv_1): For conversions, punt on casts from
29965 types other than scalar integral types.
29967 2021-03-28 David Edelsohn <dje.gcc@gmail.com>
29969 * config/rs6000/rs6000.c (rs6000_output_dwarf_dtprel): Do not add
29970 XCOFF TLS reloc decorations.
29972 2021-03-28 Gerald Pfeifer <gerald@pfeifer.com>
29974 * doc/analyzer.texi (Analyzer Internals): Update link to
29975 "A Memory Model for Static Analysis of C Programs".
29977 2021-03-26 David Edelsohn <dje.gcc@gmail.com>
29979 * config/rs6000/aix.h (ADJUST_FIELD_ALIGN): Call function.
29980 * config/rs6000/rs6000-protos.h (rs6000_special_adjust_field_align):
29982 * config/rs6000/rs6000.c (rs6000_special_adjust_field_align): New.
29983 (rs6000_special_round_type_align): Recursively check innermost first
29986 2021-03-26 Jakub Jelinek <jakub@redhat.com>
29989 * dwarf2out.h (struct dw_fde_node): Add rule18 member.
29990 * dwarf2cfi.c (dwarf2out_frame_debug_expr): When handling (set hfp sp)
29991 assignment with drap_reg active, queue reg save for hfp with offset 0
29992 and flush queued reg saves. When handling a push with rule18,
29993 defer queueing reg save for hfp and just assert the offset is 0.
29994 (scan_trace): Assert that fde->rule18 is false.
29996 2021-03-26 Vladimir Makarov <vmakarov@redhat.com>
29999 * ira-costs.c (record_reg_classes): Put case with
30000 CT_RELAXED_MEMORY adjacent to one with CT_MEMORY.
30001 * ira.c (ira_setup_alts): Ditto.
30002 * lra-constraints.c (process_alt_operands): Ditto.
30003 * recog.c (asm_operand_ok): Ditto.
30004 * reload.c (find_reloads): Ditto.
30006 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
30008 * config/aarch64/aarch64-protos.h
30009 (cpu_addrcost_table::post_modify_ld3_st3): New member variable.
30010 (cpu_addrcost_table::post_modify_ld4_st4): Likewise.
30011 * config/aarch64/aarch64.c (generic_addrcost_table): Update
30012 accordingly, using the same costs as for post_modify.
30013 (exynosm1_addrcost_table, xgene1_addrcost_table): Likewise.
30014 (thunderx2t99_addrcost_table, thunderx3t110_addrcost_table):
30015 (tsv110_addrcost_table, qdf24xx_addrcost_table): Likewise.
30016 (a64fx_addrcost_table): Likewise.
30017 (neoversev1_addrcost_table): New.
30018 (neoversev1_tunings): Use neoversev1_addrcost_table.
30019 (aarch64_address_cost): Use the new post_modify costs for CImode
30022 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
30024 * config/aarch64/aarch64.opt
30025 (-param=aarch64-loop-vect-issue-rate-niters=): New parameter.
30026 * doc/invoke.texi: Document it.
30027 * config/aarch64/aarch64-protos.h (aarch64_base_vec_issue_info)
30028 (aarch64_scalar_vec_issue_info, aarch64_simd_vec_issue_info)
30029 (aarch64_advsimd_vec_issue_info, aarch64_sve_vec_issue_info)
30030 (aarch64_vec_issue_info): New structures.
30031 (cpu_vector_cost): Write comments above the variables rather
30033 (cpu_vector_cost::issue_info): New member variable.
30034 * config/aarch64/aarch64.c: Include gimple-pretty-print.h
30035 and tree-ssa-loop-niter.h.
30036 (generic_vector_cost, a64fx_vector_cost, qdf24xx_vector_cost)
30037 (thunderx_vector_cost, tsv110_vector_cost, cortexa57_vector_cost)
30038 (exynosm1_vector_cost, xgene1_vector_cost, thunderx2t99_vector_cost)
30039 (thunderx3t110_vector_cost): Initialize issue_info to null.
30040 (neoversev1_scalar_issue_info, neoversev1_advsimd_issue_info)
30041 (neoversev1_sve_issue_info, neoversev1_vec_issue_info): New structures.
30042 (neoversev1_vector_cost): Use them.
30043 (aarch64_vec_op_count, aarch64_sve_op_count): New structures.
30044 (aarch64_vector_costs::saw_sve_only_op): New member variable.
30045 (aarch64_vector_costs::num_vector_iterations): Likewise.
30046 (aarch64_vector_costs::scalar_ops): Likewise.
30047 (aarch64_vector_costs::advsimd_ops): Likewise.
30048 (aarch64_vector_costs::sve_ops): Likewise.
30049 (aarch64_vector_costs::seen_loads): Likewise.
30050 (aarch64_simd_vec_costs_for_flags): New function.
30051 (aarch64_analyze_loop_vinfo): Initialize num_vector_iterations.
30052 Count the number of predicate operations required by SVE WHILE
30054 (aarch64_comparison_type, aarch64_multiply_add_p): New functions.
30055 (aarch64_sve_only_stmt_p, aarch64_in_loop_reduction_latency): Likewise.
30056 (aarch64_count_ops): Likewise.
30057 (aarch64_add_stmt_cost): Record whether see an SVE operation
30058 that cannot currently be implementing using Advanced SIMD.
30059 Record issue information about the scalar, Advanced SIMD
30060 and (where relevant) SVE versions of a loop.
30061 (aarch64_vec_op_count::dump): New function.
30062 (aarch64_sve_op_count::dump): Likewise.
30063 (aarch64_estimate_min_cycles_per_iter): Likewise.
30064 (aarch64_adjust_body_cost): If issue information is available,
30065 try to compare the issue rates of the various loop implementations
30066 and increase or decrease the vector body cost accordingly.
30068 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
30070 * config/aarch64/aarch64.c (aarch64_detect_vector_stmt_subtype):
30071 Assume a zero cost for induction phis.
30073 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
30075 * config/aarch64/aarch64.c (aarch64_embedded_comparison_type): New
30077 (aarch64_adjust_stmt_cost): Add the costs of embedded scalar and
30078 vector comparisons.
30080 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
30082 * config/aarch64/aarch64.c (aarch64_detect_scalar_stmt_subtype):
30084 (aarch64_add_stmt_cost): Call it.
30086 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
30088 * config/aarch64/aarch64-tuning-flags.def (matched_vector_throughput):
30089 New tuning parameter.
30090 * config/aarch64/aarch64.c (neoversev1_tunings): Use it.
30091 (aarch64_estimated_sve_vq): New function.
30092 (aarch64_vector_costs::analyzed_vinfo): New member variable.
30093 (aarch64_vector_costs::is_loop): Likewise.
30094 (aarch64_vector_costs::unrolled_advsimd_niters): Likewise.
30095 (aarch64_vector_costs::unrolled_advsimd_stmts): Likewise.
30096 (aarch64_record_potential_advsimd_unrolling): New function.
30097 (aarch64_analyze_loop_vinfo, aarch64_analyze_bb_vinfo): Likewise.
30098 (aarch64_add_stmt_cost): Call aarch64_analyze_loop_vinfo or
30099 aarch64_analyze_bb_vinfo on the first use of a costs structure.
30100 Detect whether we're vectorizing a loop for SVE that might be
30101 completely unrolled if it used Advanced SIMD instead.
30102 (aarch64_adjust_body_cost_for_latency): New function.
30103 (aarch64_finish_cost): Call it.
30105 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
30107 * config/aarch64/aarch64.c (aarch64_vector_costs): New structure.
30108 (aarch64_init_cost): New function.
30109 (aarch64_add_stmt_cost): Use aarch64_vector_costs instead of
30110 the default unsigned[3].
30111 (aarch64_finish_cost, aarch64_destroy_cost_data): New functions.
30112 (TARGET_VECTORIZE_INIT_COST): Override.
30113 (TARGET_VECTORIZE_FINISH_COST): Likewise.
30114 (TARGET_VECTORIZE_DESTROY_COST_DATA): Likewise.
30116 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
30118 * config/aarch64/aarch64.c (neoversev1_advsimd_vector_cost)
30119 (neoversev1_sve_vector_cost): New cost structures.
30120 (neoversev1_vector_cost): Likewise.
30121 (neoversev1_tunings): Use them. Enable use_new_vector_costs.
30123 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
30125 * config/aarch64/aarch64-protos.h
30126 (sve_vec_cost::scatter_store_elt_cost): New member variable.
30127 * config/aarch64/aarch64.c (generic_sve_vector_cost): Update
30128 accordingly, taking the cost from the cost of a scalar_store.
30129 (a64fx_sve_vector_cost): Likewise.
30130 (aarch64_detect_vector_stmt_subtype): Detect scatter stores.
30132 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
30134 * config/aarch64/aarch64-protos.h
30135 (simd_vec_cost::store_elt_extra_cost): New member variable.
30136 * config/aarch64/aarch64.c (generic_advsimd_vector_cost): Update
30137 accordingly, using the vec_to_scalar cost for the new field.
30138 (generic_sve_vector_cost, a64fx_advsimd_vector_cost): Likewise.
30139 (a64fx_sve_vector_cost, qdf24xx_advsimd_vector_cost): Likewise.
30140 (thunderx_advsimd_vector_cost, tsv110_advsimd_vector_cost): Likewise.
30141 (cortexa57_advsimd_vector_cost, exynosm1_advsimd_vector_cost)
30142 (xgene1_advsimd_vector_cost, thunderx2t99_advsimd_vector_cost)
30143 (thunderx3t110_advsimd_vector_cost): Likewise.
30144 (aarch64_detect_vector_stmt_subtype): Detect single-element stores.
30146 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
30148 * config/aarch64/aarch64-protos.h (simd_vec_cost::ld2_st2_permute_cost)
30149 (simd_vec_cost::ld3_st3_permute_cost): New member variables.
30150 (simd_vec_cost::ld4_st4_permute_cost): Likewise.
30151 * config/aarch64/aarch64.c (generic_advsimd_vector_cost): Update
30152 accordingly, using zero for the new costs.
30153 (generic_sve_vector_cost, a64fx_advsimd_vector_cost): Likewise.
30154 (a64fx_sve_vector_cost, qdf24xx_advsimd_vector_cost): Likewise.
30155 (thunderx_advsimd_vector_cost, tsv110_advsimd_vector_cost): Likewise.
30156 (cortexa57_advsimd_vector_cost, exynosm1_advsimd_vector_cost)
30157 (xgene1_advsimd_vector_cost, thunderx2t99_advsimd_vector_cost)
30158 (thunderx3t110_advsimd_vector_cost): Likewise.
30159 (aarch64_ld234_st234_vectors): New function.
30160 (aarch64_adjust_stmt_cost): Likewise.
30161 (aarch64_add_stmt_cost): Call aarch64_adjust_stmt_cost if using
30162 the new vector costs.
30164 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
30166 * config/aarch64/aarch64-protos.h (sve_vec_cost): Turn into a
30167 derived class of simd_vec_cost. Add information about CLAST[AB]
30168 and FADDA instructions.
30169 * config/aarch64/aarch64.c (generic_sve_vector_cost): Update
30170 accordingly, using the vec_to_scalar costs for the new fields.
30171 (a64fx_sve_vector_cost): Likewise.
30172 (aarch64_reduc_type): New function.
30173 (aarch64_sve_in_loop_reduction_latency): Likewise.
30174 (aarch64_detect_vector_stmt_subtype): Take a vinfo parameter.
30175 Use aarch64_sve_in_loop_reduction_latency to handle SVE reductions
30176 that occur in the loop body.
30177 (aarch64_add_stmt_cost): Update call accordingly.
30179 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
30181 * config/aarch64/aarch64-tuning-flags.def (use_new_vector_costs):
30183 * config/aarch64/aarch64-protos.h (simd_vec_cost): Put comments
30184 above the fields rather than to the right.
30185 (simd_vec_cost::reduc_i8_cost): New member variable.
30186 (simd_vec_cost::reduc_i16_cost): Likewise.
30187 (simd_vec_cost::reduc_i32_cost): Likewise.
30188 (simd_vec_cost::reduc_i64_cost): Likewise.
30189 (simd_vec_cost::reduc_f16_cost): Likewise.
30190 (simd_vec_cost::reduc_f32_cost): Likewise.
30191 (simd_vec_cost::reduc_f64_cost): Likewise.
30192 * config/aarch64/aarch64.c (generic_advsimd_vector_cost): Update
30193 accordingly, using the vec_to_scalar_cost for the new fields.
30194 (generic_sve_vector_cost, a64fx_advsimd_vector_cost): Likewise.
30195 (a64fx_sve_vector_cost, qdf24xx_advsimd_vector_cost): Likewise.
30196 (thunderx_advsimd_vector_cost, tsv110_advsimd_vector_cost): Likewise.
30197 (cortexa57_advsimd_vector_cost, exynosm1_advsimd_vector_cost)
30198 (xgene1_advsimd_vector_cost, thunderx2t99_advsimd_vector_cost)
30199 (thunderx3t110_advsimd_vector_cost): Likewise.
30200 (aarch64_use_new_vector_costs_p): New function.
30201 (aarch64_simd_vec_costs): New function, split out from...
30202 (aarch64_builtin_vectorization_cost): ...here.
30203 (aarch64_is_reduction): New function.
30204 (aarch64_detect_vector_stmt_subtype): Likewise.
30205 (aarch64_add_stmt_cost): Call aarch64_detect_vector_stmt_subtype if
30206 using the new vector costs.
30208 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
30211 * tree-emutls.c (get_emutls_init_templ_addr): Mark initializer of weak
30212 TLS declarations as public.
30214 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
30216 * config/aarch64/aarch64-d.c (IN_TARGET_CODE): Define.
30217 * config/arm/arm-d.c (IN_TARGET_CODE): Likewise.
30218 * config/i386/i386-d.c (IN_TARGET_CODE): Likewise.
30219 * config/mips/mips-d.c (IN_TARGET_CODE): Likewise.
30220 * config/pa/pa-d.c (IN_TARGET_CODE): Likewise.
30221 * config/riscv/riscv-d.c (IN_TARGET_CODE): Likewise.
30222 * config/rs6000/rs6000-d.c (IN_TARGET_CODE): Likewise.
30223 * config/s390/s390-d.c (IN_TARGET_CODE): Likewise.
30224 * config/sparc/sparc-d.c (IN_TARGET_CODE): Likewise.
30226 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
30229 * config.gcc (*-*-cygwin*): Add winnt-d.o
30230 (*-*-mingw*): Likewise.
30231 * config/i386/cygwin.h (EXTRA_TARGET_D_OS_VERSIONS): New macro.
30232 * config/i386/mingw32.h (EXTRA_TARGET_D_OS_VERSIONS): Likewise.
30233 * config/i386/t-cygming: Add winnt-d.o.
30234 * config/i386/winnt-d.c: New file.
30236 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
30238 * config/freebsd-d.c: Include memmodel.h.
30240 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
30243 * config.gcc (*-*-openbsd*): Add openbsd-d.o.
30244 * config/t-openbsd: Add openbsd-d.o.
30245 * config/openbsd-d.c: New file.
30247 2021-03-25 Stam Markianos-Wright <stam.markianos-wright@arm.com>
30249 PR tree-optimization/96974
30250 * tree-vect-stmts.c (vect_get_vector_types_for_stmt): Replace assert
30251 with graceful exit.
30253 2021-03-25 H.J. Lu <hjl.tools@gmail.com>
30256 2021-03-25 H.J. Lu <hjl.tools@gmail.com>
30260 * config/i386/i386.c (ix86_can_inline_p): Don't check ISA for
30261 always_inline in system headers.
30263 2021-03-25 Kewen Lin <linkw@linux.ibm.com>
30265 * tree-vect-loop.c (vect_model_reduction_cost): Init inside_cost.
30267 2021-03-25 Jakub Jelinek <jakub@redhat.com>
30270 * tree-core.h (enum operand_equal_flag): Add OEP_ADDRESS_OF_SAME_FIELD.
30271 * fold-const.c (operand_compare::operand_equal_p): Don't compare
30272 field offsets if OEP_ADDRESS_OF_SAME_FIELD.
30274 2021-03-25 H.J. Lu <hjl.tools@gmail.com>
30278 * config/i386/i386.c (ix86_can_inline_p): Don't check ISA for
30279 always_inline in system headers.
30281 2021-03-25 Richard Biener <rguenther@suse.de>
30283 PR tree-optimization/99746
30284 * tree-vect-slp-patterns.c (complex_pattern::build): Do not mark
30285 the scalar stmt as patterned. Instead set up required things
30288 2021-03-25 Xionghu Luo <luoxhu@linux.ibm.com>
30290 * config/rs6000/rs6000.c (power8_costs): Change l2 cache
30293 2021-03-24 Martin Liska <mliska@suse.cz>
30296 * common/config/i386/i386-common.c (ARRAY_SIZE): Fix off-by-one
30298 * config/i386/i386-options.c (ix86_option_override_internal):
30299 Add run-time assert.
30301 2021-03-24 Martin Jambor <mjambor@suse.cz>
30304 * ipa-cp.c (initialize_node_lattices): Mark as bottom all
30305 parameters with unknown type.
30306 (ipacp_value_safe_for_type): New function.
30307 (propagate_vals_across_arith_jfunc): Verify that the constant type
30308 can be used for a type of the formal parameter.
30309 (propagate_vals_across_ancestor): Likewise.
30310 (propagate_scalar_across_jump_function): Likewise. Pass the type
30311 also to propagate_vals_across_ancestor.
30313 2021-03-24 Christophe Lyon <christophe.lyon@linaro.org>
30316 * config/arm/mve.md (movmisalign<mode>_mve_store): Use Ux
30318 (movmisalign<mode>_mve_load): Likewise.
30320 2021-03-24 Jakub Jelinek <jakub@redhat.com>
30323 * config/arm/vec-common.md (one_cmpl<mode>2, neg<mode>2,
30324 movmisalign<mode>): Disable expanders for TARGET_REALLY_IWMMXT.
30326 2021-03-24 Alexandre Oliva <oliva@adacore.com>
30328 * doc/sourcebuild.texi (sysconf): New effective target.
30330 2021-03-24 Alexandre Oliva <oliva@adacore.com>
30332 * config/i386/predicates.md (reg_or_const_vec_operand): New.
30333 * config/i386/sse.md (ssse3_pshufbv8qi3): Add an expander for
30334 the now *-prefixed insn_and_split, turn the splitter const vec
30335 into an input for the insn, making it an ignored immediate for
30336 non-split cases, and loaded into the scratch register
30339 2021-03-23 Vladimir N. Makarov <vmakarov@redhat.com>
30342 * config/aarch64/constraints.md (Utq, UOb, UOh, UOw, UOd, UOty):
30343 Use define_relaxed_memory_constraint for them.
30345 2021-03-23 Iain Sandoe <iain@sandoe.co.uk>
30348 * config/host-darwin.c (darwin_gt_pch_use_address): Add a
30349 colon to the diagnostic message.
30351 2021-03-23 Ilya Leoshkevich <iii@linux.ibm.com>
30353 * fwprop.c (fwprop_propagation::fwprop_propagation): Look at
30355 (try_fwprop_subst_note): Use set_info instead of insn_info.
30356 (try_fwprop_subst_pattern): Likewise.
30357 (try_fwprop_subst_notes): Likewise.
30358 (try_fwprop_subst): Likewise.
30359 (forward_propagate_subreg): Likewise.
30360 (forward_propagate_and_simplify): Likewise.
30361 (forward_propagate_into): Likewise.
30362 * rtl-ssa/accesses.h (set_info::single_nondebug_use) New
30364 (set_info::single_nondebug_insn_use): Likewise.
30365 (set_info::single_phi_use): Likewise.
30366 * rtl-ssa/member-fns.inl (set_info::single_nondebug_use) New
30368 (set_info::single_nondebug_insn_use): Likewise.
30369 (set_info::single_phi_use): Likewise.
30371 2021-03-23 Christophe Lyon <christophe.lyon@linaro.org>
30373 * doc/sourcebuild.texi (arm_dsp_ok, arm_dsp): Document.
30375 2021-03-23 Jakub Jelinek <jakub@redhat.com>
30378 * config/aarch64/aarch64.c (aarch64_add_offset): Tell
30379 expand_mult to perform an unsigned rather than a signed
30382 2021-03-23 H.J. Lu <hjl.tools@gmail.com>
30385 * config/i386/cpuid.h (__cpuid): Add __volatile__.
30386 (__cpuid_count): Likewise.
30388 2021-03-23 Richard Biener <rguenther@suse.de>
30390 PR tree-optimization/99721
30391 * tree-vect-slp.c (vect_slp_analyze_node_operations):
30392 Make sure we can schedule the node.
30394 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
30396 * config/riscv/riscv.c (riscv_subword): Take endianness into
30397 account when calculating the byte offset.
30399 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
30401 * config/riscv/predicates.md (subreg_lowpart_operator): New predicate
30402 * config/riscv/riscv.md (*addsi3_extended2, *subsi3_extended2)
30403 (*negsi2_extended2, *mulsi3_extended2, *<optab>si3_mask)
30404 (*<optab>si3_mask_1, *<optab>di3_mask, *<optab>di3_mask_1)
30405 (*<optab>si3_extend_mask, *<optab>si3_extend_mask_1): Use
30406 new predicate "subreg_lowpart_operator"
30408 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
30410 * config/riscv/riscv.c (riscv_swap_instruction): New function
30411 to byteswap an SImode rtx containing an instruction.
30412 (riscv_trampoline_init): Byteswap the generated instructions
30415 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
30417 * common/config/riscv/riscv-common.c
30418 (TARGET_DEFAULT_TARGET_FLAGS): Set default endianness.
30419 * config.gcc (riscv32be-*, riscv64be-*): Set
30420 TARGET_BIG_ENDIAN_DEFAULT to 1.
30421 * config/riscv/elf.h (LINK_SPEC): Change -melf* value
30422 depending on default endianness.
30423 * config/riscv/freebsd.h (LINK_SPEC): Likewise.
30424 * config/riscv/linux.h (LINK_SPEC): Likewise.
30425 * config/riscv/riscv.c (TARGET_DEFAULT_TARGET_FLAGS): Set
30426 default endianness.
30427 * config/riscv/riscv.h (DEFAULT_ENDIAN_SPEC): New macro.
30429 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
30431 * config/riscv/elf.h (LINK_SPEC): Pass linker endianness flag.
30432 * config/riscv/freebsd.h (LINK_SPEC): Likewise.
30433 * config/riscv/linux.h (LINK_SPEC): Likewise.
30434 * config/riscv/riscv.h (ASM_SPEC): Pass -mbig-endian and
30436 (BYTES_BIG_ENDIAN): Handle big endian.
30437 (WORDS_BIG_ENDIAN): Define to BYTES_BIG_ENDIAN.
30438 * config/riscv/riscv.opt (-mbig-endian, -mlittle-endian): New
30440 * doc/invoke.texi (-mbig-endian, -mlittle-endian): Document.
30442 2021-03-23 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
30444 * regcprop.c (find_oldest_value_reg): Ask target whether
30445 different mode is fine for replacement register.
30447 2021-03-23 Aldy Hernandez <aldyh@redhat.com>
30449 PR tree-optimization/99296
30450 * value-range.cc (irange::irange_set_1bit_anti_range): New.
30451 (irange::irange_set_anti_range): Call irange_set_1bit_anti_range
30452 * value-range.h (irange::irange_set_1bit_anti_range): New.
30454 2021-03-22 Vladimir N. Makarov <vmakarov@redhat.com>
30457 * config/aarch64/constraints.md (UtQ): Use
30458 define_relaxed_memory_constraint for it.
30459 * doc/md.texi (define_relaxed_memory_constraint): Describe it.
30460 * genoutput.c (main): Process DEFINE_RELAXED_MEMORY_CONSTRAINT.
30461 * genpreds.c (constraint_data): Add bitfield is_relaxed_memory.
30462 (have_relaxed_memory_constraints): New static var.
30463 (relaxed_memory_start, relaxed_memory_end): Ditto.
30464 (add_constraint): Add arg is_relaxed_memory. Check name for
30465 relaxed memory. Set up is_relaxed_memory in constraint_data and
30466 have_relaxed_memory_constraints. Adjust calls.
30467 (choose_enum_order): Process relaxed memory.
30468 (write_tm_preds_h): Ditto.
30469 (main): Process DEFINE_RELAXED_MEMORY_CONSTRAINT.
30470 * gensupport.c (process_rtx): Process DEFINE_RELAXED_MEMORY_CONSTRAINT.
30471 * ira-costs.c (record_reg_classes): Process CT_RELAXED_MEMORY.
30472 * ira-lives.c (single_reg_class): Use
30473 insn_extra_relaxed_memory_constraint.
30474 * ira.c (ira_setup_alts): CT_RELAXED_MEMORY.
30475 * lra-constraints.c (valid_address_p): Use
30476 insn_extra_relaxed_memory_constraint instead of other memory
30478 (process_alt_operands): Process CT_RELAXED_MEMORY.
30479 (curr_insn_transform): Use insn_extra_relaxed_memory_constraint.
30480 * recog.c (asm_operand_ok, preprocess_constraints): Process
30482 * reload.c (find_reloads): Ditto.
30483 * rtl.def (DEFINE_RELAXED_MEMORY_CONSTRAINT): New.
30484 * stmt.c (parse_input_constraint): Use
30485 insn_extra_relaxed_memory_constraint.
30487 2021-03-22 Segher Boessenkool <segher@kernel.crashing.org>
30490 * ubsan.c (ubsan_instrument_float_cast): Don't test for unordered if
30493 2021-03-22 Alex Coplan <alex.coplan@arm.com>
30496 * config/arm/arm-protos.h (neon_make_constant): Add generate
30497 argument to guard emitting insns, default to true.
30498 * config/arm/arm.c (arm_legitimate_constant_p_1): Reject
30499 CONST_VECTORs which neon_make_constant can't handle.
30500 (neon_vdup_constant): Add generate argument, avoid emitting
30501 insns if it's not set.
30502 (neon_make_constant): Plumb new generate argument through.
30503 * config/arm/constraints.md (Ui): New. Use it...
30504 * config/arm/mve.md (*mve_mov<mode>): ... here.
30505 * config/arm/vec-common.md (movv8hf): Use neon_make_constant to
30506 synthesize constants.
30508 2021-03-22 Richard Biener <rguenther@suse.de>
30510 * debug.h: Add deprecation warning.
30512 2021-03-22 Richard Biener <rguenther@suse.de>
30514 PR tree-optimization/99694
30515 * tree-ssa-sccvn.c (visit_phi): Ignore edges with the
30518 2021-03-22 Kito Cheng <kito.cheng@sifive.com>
30521 * config/riscv/riscv.c (riscv_expand_block_move): Get RTL value
30522 after type checking.
30524 2021-03-22 Jakub Jelinek <jakub@redhat.com>
30528 * dwarf2out.c (get_full_len): Use get_precision rather than
30530 (add_const_value_attribute): Make sure add_AT_wide argument has
30531 precision prec rather than some very wide one.
30533 2021-03-22 Kewen Lin <linkw@linux.ibm.com>
30535 * config/rs6000/rs6000.md (*rotldi3_insert_sf,
30536 *mov<SFDF:mode><SFDF2:mode>cc_p9, floatsi<mode>2_lfiwax,
30537 floatsi<mode>2_lfiwax_mem, floatunssi<mode>2_lfiwzx,
30538 floatunssi<mode>2_lfiwzx_mem, *floatsidf2_internal,
30539 *floatunssidf2_internal, fix_trunc<mode>si2_stfiwx,
30540 fix_trunc<mode>si2_internal, fixuns_trunc<mode>si2_stfiwx,
30541 *round32<mode>2_fprs, *roundu32<mode>2_fprs,
30542 *fix_trunc<mode>si2_internal): Fix empty split condition.
30543 * config/rs6000/vsx.md (*vsx_le_undo_permute_<mode>,
30544 vsx_reduc_<VEC_reduc_name>_v2df, vsx_reduc_<VEC_reduc_name>_v4sf,
30545 *vsx_reduc_<VEC_reduc_name>_v2df_scalar,
30546 *vsx_reduc_<VEC_reduc_name>_v4sf_scalar): Likewise.
30548 2021-03-22 Xionghu Luo <luoxhu@linux.ibm.com>
30551 * config/rs6000/rs6000.c (rs6000_expand_vector_set_var_p9):
30552 Convert idx to DImode.
30553 (rs6000_expand_vector_set_var_p8): Likewise.
30555 2021-03-21 Jakub Jelinek <jakub@redhat.com>
30558 * dwarf2out.c (insert_float): Change return type from void to
30559 unsigned, handle GET_MODE_SIZE (mode) == 2 and return element size.
30560 (mem_loc_descriptor, loc_descriptor, add_const_value_attribute):
30563 2021-03-20 H.J. Lu <hjl.tools@gmail.com>
30566 * config/i386/i386.c (construct_container): Check cfun != NULL
30567 before accessing silent_p.
30569 2021-03-20 Ahamed Husni <ahamedhusni73@gmail.com>
30571 * asan.c: Fix typos in comments.
30573 2021-03-20 Vladimir N. Makarov <vmakarov@redhat.com>
30575 PR rtl-optimization/99680
30576 * lra-constraints.c (skip_contraint_modifiers): Rename to skip_constraint_modifiers.
30577 (process_address_1): Check empty constraint before using
30580 2021-03-19 Pat Haugen <pthaugen@linux.ibm.com>
30582 * config/rs6000/rs6000.c (power10_cost): New.
30583 (rs6000_option_override_internal): Set Power10 costs.
30584 (rs6000_issue_rate): Set Power10 issue rate.
30585 * config/rs6000/power10.md: Rewrite for Power10.
30587 2021-03-19 Vladimir N. Makarov <vmakarov@redhat.com>
30590 * lra-constraints.c (process_address_1): Don't use unknown
30591 constraint for address constraint.
30593 2021-03-19 Iain Sandoe <iain@sandoe.co.uk>
30596 * config.gcc (powerpc-*-darwin8): Delete the reference to
30597 the now removed darwin8.h.
30599 2021-03-19 Olivier Hainque <hainque@adacore.com>
30602 * config/vxworksae.h (VX_CPU_PREFIX): Define.
30604 2021-03-19 John David Anglin <danglin@gcc.gnu.org>
30606 * config/pa/pa.c (import_milli): Use memcpy instead of strncpy.
30608 2021-03-19 Tamar Christina <tamar.christina@arm.com>
30610 PR tree-optimization/99656
30611 * tree-vect-slp-patterns.c (linear_loads_p,
30612 complex_add_pattern::matches, is_eq_or_top,
30613 vect_validate_multiplication, complex_mul_pattern::matches,
30614 complex_fms_pattern::matches): Remove complex_perm_kinds_t.
30615 * tree-vectorizer.h: (complex_load_perm_t): Removed.
30616 (slp_tree_to_load_perm_map_t): Use complex_perm_kinds_t instead of
30617 complex_load_perm_t.
30619 2021-03-19 H.J. Lu <hjl.tools@gmail.com>
30622 * config/i386/i386-options.c (ix86_init_machine_status): Set
30624 * config/i386/i386.c (init_cumulative_args): Set silent_p to
30626 (construct_container): Return early for return and argument
30627 errors if silent_p is true.
30628 * config/i386/i386.h (machine_function): Add silent_p.
30630 2021-03-19 Jakub Jelinek <jakub@redhat.com>
30633 * config/arm/constraints.md (Ds): New constraint.
30634 * config/arm/vec-common.md (mve_vshlq_<supf><mode>): Use w,Ds
30635 constraint instead of w,Dm.
30637 2021-03-19 Andrew Stubbs <ams@codesourcery.com>
30639 * config/gcn/gcn.c (gcn_parse_amdgpu_hsa_kernel_attribute): Fix quotes
30642 2021-03-19 Eric Botcazou <ebotcazou@adacore.com>
30644 PR middle-end/99641
30645 * fold-const.c (native_encode_initializer) <CONSTRUCTOR>: For an
30646 array type, do the computation of the current position in sizetype.
30648 2021-03-18 Vladimir N. Makarov <vmakarov@redhat.com>
30651 * lra-constraints.c (process_address_1): Use lookup_constraint
30652 only for a single constraint.
30654 2021-03-18 Martin Sebor <msebor@redhat.com>
30656 PR middle-end/99502
30657 * gimple-array-bounds.cc (inbounds_vbase_memaccess_p): Rename...
30658 (inbounds_memaccess_p): ...to this. Check the ending offset of
30659 the accessed member.
30661 2021-03-18 Andrew Stubbs <ams@codesourcery.com>
30663 * config/gcn/gcn.c (gcn_parse_amdgpu_hsa_kernel_attribute): Add %< and
30664 %> quote markers to error messages.
30665 (gcn_goacc_validate_dims): Likewise.
30666 (gcn_conditional_register_usage): Remove exclaimation mark from error
30668 (gcn_vectorize_vec_perm_const): Ensure perm is fully uninitialized.
30670 2021-03-18 Jan Hubicka <hubicka@ucw.cz>
30672 * config/i386/x86-tune-costs.h (struct processor_costs): Fix costs of
30675 2021-03-18 Sinan Lin <sinan@isrc.iscas.ac.cn>
30676 Kito Cheng <kito.cheng@sifive.com>
30678 * config/riscv/riscv.c (riscv_block_move_straight): Change type
30679 to unsigned HOST_WIDE_INT for parameter and local variable with
30680 HOST_WIDE_INT type.
30681 (riscv_adjust_block_mem): Ditto.
30682 (riscv_block_move_loop): Ditto.
30683 (riscv_expand_block_move): Ditto.
30685 2021-03-18 Nick Clifton <nickc@redhat.com>
30687 * config/v850/v850.c (construct_restore_jr): Increase static
30689 (construct_save_jarl): Likewise.
30690 * config/v850/v850.h (DWARF2_DEBUGGING_INFO): Define.
30692 2021-03-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
30694 * config/aarch64/aarch64.c (aarch64_adjust_generic_arch_tuning): Define.
30695 (aarch64_override_options_internal): Use it.
30696 (generic_tunings): Add AARCH64_EXTRA_TUNE_CSE_SVE_VL_CONSTANTS to
30699 2021-03-17 Sandra Loosemore <sandra@codesourcery.com>
30701 * config/nios2/nios2.c (nios2_custom_check_insns): Clean up
30702 error message format issues.
30703 (nios2_option_override): Likewise.
30704 (nios2_expand_fpu_builtin): Likewise.
30705 (nios2_init_custom_builtins): Adjust to avoid bogus strncpy
30706 truncation warning.
30707 (nios2_expand_custom_builtin): More error message format fixes.
30708 (nios2_expand_rdwrctl_builtin): Likewise.
30709 (nios2_expand_rdprs_builtin): Likewise.
30710 (nios2_expand_eni_builtin): Likewise.
30711 (nios2_expand_builtin): Likewise.
30712 (nios2_register_custom_code): Likewise.
30713 (nios2_valid_target_attribute_rec): Likewise.
30714 (nios2_add_insn_asm): Fix uninitialized variable warning.
30716 2021-03-17 Jan Hubicka <jh@suse.cz>
30718 * config/i386/x86-tune-costs.h (struct processor_costs): Update costs
30719 of gather to match reality.
30720 * config/i386/x86-tune.def (X86_TUNE_USE_GATHER): Enable for znver3.
30722 2021-03-17 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
30724 * config/aarch64/aarch64-builtins.c (aarch64_expand_rng_builtin): Use EQ
30725 to compare against CC_REG rather than NE.
30727 2021-03-17 H.J. Lu <hjl.tools@gmail.com>
30730 * config/i386/i386.c (ix86_force_load_from_GOT_p): Support
30731 inline assembly statements.
30732 (ix86_print_operand): Update 'P' handling for -fno-plt.
30734 2021-03-17 Tamar Christina <tamar.christina@arm.com>
30737 * config/aarch64/aarch64.c
30738 (aarch64_simd_clone_compute_vecsize_and_simdlen): Remove unused var.
30740 2021-03-16 Segher Boessenkool <segher@kernel.crashing.org>
30743 * config/rs6000/predicates.md (branch_comparison_operator): Allow
30744 ordered and unordered for CCFPmode, if flag_finite_math_only.
30746 2021-03-16 Jakub Jelinek <jakub@redhat.com>
30749 * config/i386/i386-expand.c (ix86_split_lea_for_addr): Emit a MULT
30750 rather than ASHIFT.
30751 * config/i386/i386.md (mult by 1248 into ashift): New splitter.
30753 2021-03-16 Martin Liska <mliska@suse.cz>
30756 * optc-save-gen.awk: Add flag_ipa_ra to exceptions for
30757 cl_optimization_compare function.
30759 2021-03-16 Ilya Leoshkevich <iii@linux.ibm.com>
30761 * config/s390/s390.c (f_constraint_p): Treat "fv" constraints
30764 2021-03-16 Jakub Jelinek <jakub@redhat.com>
30767 * config/i386/i386.h (struct machine_function): Add
30768 has_explicit_vzeroupper bitfield.
30769 * config/i386/i386-expand.c (ix86_expand_builtin): Set
30770 cfun->machine->has_explicit_vzeroupper when expanding
30771 IX86_BUILTIN_VZEROUPPER.
30772 * config/i386/i386-features.c (rest_of_handle_insert_vzeroupper):
30773 Do the mode switching only when TARGET_VZEROUPPER, expensive
30774 optimizations turned on and not optimizing for size.
30775 (pass_insert_vzeroupper::gate): Enable even when
30776 cfun->machine->has_explicit_vzeroupper is set.
30778 2021-03-16 Jakub Jelinek <jakub@redhat.com>
30781 * config/aarch64/aarch64.c
30782 (aarch64_simd_clone_compute_vecsize_and_simdlen): If not a function
30783 definition, walk TYPE_ARG_TYPES list if non-NULL for argument types
30784 instead of DECL_ARGUMENTS. Ignore types for uniform arguments.
30786 2021-03-15 Richard Biener <rguenther@suse.de>
30788 PR tree-optimization/98834
30789 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle missing
30790 subsetting by truncating the access size.
30792 2021-03-15 Jan Hubicka <hubicka@ucw.cz>
30794 * config/i386/i386-options.c (processor_cost_table): Add znver3_cost.
30795 * config/i386/x86-tune-costs.h (znver3_cost): New gobal variable; copy
30798 2021-03-15 Martin Liska <mliska@suse.cz>
30800 * spellcheck.c: Add missing comma in initialization.
30802 2021-03-14 Uroš Bizjak <ubizjak@gmail.com>
30804 * config/i386/sse.md (*vec_extract<mode>): Merge alternative 0 with
30805 alternative 2 and alternative 1 with alternative 3 using
30806 YW register constraint.
30807 (*vec_extract<PEXTR_MODE12:mode>_zext): Merge alternatives
30808 using YW register constraint.
30809 (*vec_extractv16qi_zext): Ditto.
30810 (*vec_extractv4si): Merge alternatives 4 and 5
30811 using Yw register constraint.
30812 (*ssse3_palignr<mode>_perm): Use Yw instead of v for alternative 3.
30814 2021-03-13 Martin Sebor <msebor@redhat.com>
30816 PR tree-optimization/99489
30817 * builtins.c (gimple_call_alloc_size): Fail gracefully when argument
30818 is not a call statement.
30820 2021-03-13 Jakub Jelinek <jakub@redhat.com>
30822 PR tree-optimization/99544
30823 * match.pd (X + (X << C) -> X * (1 + (1 << C))): Don't simplify
30824 if for vector types multiplication can't be done in type's mode.
30826 2021-03-12 Eric Botcazou <ebotcazou@adacore.com>
30829 * config/sparc/constraints.md (w): Rename to...
30830 (W): ... this and ditch previous implementation.
30831 * config/sparc/sparc.md (*movdi_insn_sp64): Replace W with m.
30832 (*movdf_insn_sp64): Likewise.
30833 (*mov<VM64:mode>_insn_sp64): Likewise.
30834 * config/sparc/sync.md (*atomic_compare_and_swap<mode>_1): Replace
30836 (atomic_compare_and_swap_leon3_1): Likewise.
30837 (*atomic_compare_and_swapdi_v8plus): Likewise.
30838 * config/sparc/sparc.c (memory_ok_for_ldd): Remove useless test on
30839 architecture and add missing address validity check during LRA.
30841 2021-03-12 Tobias Burnus <tobias@codesourcery.com>
30844 * gimplify.c (omp_add_variable): Handle NULL_TREE as size
30845 occuring for assumed-size arrays in use_device_{ptr,addr}.
30847 2021-03-12 Jakub Jelinek <jakub@redhat.com>
30850 * config/i386/constraints.md (YW): New internal constraint.
30851 * config/i386/sse.md (v_Yw): Add V4TI, V2TI, V1TI and TI cases.
30852 (*<sse2_avx2>_<insn><mode>3<mask_name>,
30853 *<sse2_avx2>_uavg<mode>3<mask_name>, *abs<mode>2,
30854 *<s>mul<mode>3_highpart<mask_name>): Use <v_Yw> instead of v in
30856 (<sse2_avx2>_psadbw): Use YW instead of v in constraints.
30857 (*avx2_pmaddwd, *sse2_pmaddwd, *<code>v8hi3, *<code>v16qi3,
30858 avx2_pmaddubsw256, ssse3_pmaddubsw128): Merge last two alternatives
30859 into one, use Yw instead of former x,v.
30860 (ashr<mode>3, <insn><mode>3): Use <v_Yw> instead of x in constraints of
30861 the last alternative.
30862 (<sse2_avx2>_packsswb<mask_name>, <sse2_avx2>_packssdw<mask_name>,
30863 <sse2_avx2>_packuswb<mask_name>, <sse4_1_avx2>_packusdw<mask_name>,
30864 *<ssse3_avx2>_pmulhrsw<mode>3<mask_name>, <ssse3_avx2>_palignr<mode>,
30865 <ssse3_avx2>_pshufb<mode>3<mask_name>): Merge last two alternatives
30866 into one, use <v_Yw> instead of former x,v.
30867 (avx2_interleave_highv32qi<mask_name>,
30868 vec_interleave_highv16qi<mask_name>): Use Yw instead of v in
30869 constraints. Add && <mask_avx512bw_condition> to condition.
30870 (avx2_interleave_lowv32qi<mask_name>,
30871 vec_interleave_lowv16qi<mask_name>,
30872 avx2_interleave_highv16hi<mask_name>,
30873 vec_interleave_highv8hi<mask_name>,
30874 avx2_interleave_lowv16hi<mask_name>, vec_interleave_lowv8hi<mask_name>,
30875 avx2_pshuflw_1<mask_name>, sse2_pshuflw_1<mask_name>,
30876 avx2_pshufhw_1<mask_name>, sse2_pshufhw_1<mask_name>,
30877 avx2_<code>v16qiv16hi2<mask_name>, sse4_1_<code>v8qiv8hi2<mask_name>,
30878 *sse4_1_<code>v8qiv8hi2<mask_name>_1, <sse2_avx2>_<insn><mode>3): Use
30879 Yw instead of v in constraints.
30880 * config/i386/mmx.md (Yv_Yw): New define_mode_attr.
30881 (*mmx_<insn><mode>3, mmx_ashr<mode>3, mmx_<insn><mode>3): Use <Yv_Yw>
30882 instead of Yv in constraints.
30883 (*mmx_<insn><mode>3, *mmx_mulv4hi3, *mmx_smulv4hi3_highpart,
30884 *mmx_umulv4hi3_highpart, *mmx_pmaddwd, *mmx_<code>v4hi3,
30885 *mmx_<code>v8qi3, mmx_pack<s_trunsuffix>swb, mmx_packssdw,
30886 mmx_punpckhbw, mmx_punpcklbw, mmx_punpckhwd, mmx_punpcklwd,
30887 *mmx_uavgv8qi3, *mmx_uavgv4hi3, mmx_psadbw): Use Yw instead of Yv in
30889 (*mmx_pinsrw, *mmx_pinsrb, *mmx_pextrw, *mmx_pextrw_zext, *mmx_pextrb,
30890 *mmx_pextrb_zext): Use YW instead of Yv in constraints.
30891 (*mmx_eq<mode>3, mmx_gt<mode>3): Use x instead of Yv in constraints.
30892 (mmx_andnot<mode>3, *mmx_<code><mode>3): Split last alternative into
30893 two, one with just x, another isa avx512vl with v.
30895 2021-03-12 Martin Liska <mliska@suse.cz>
30897 * doc/invoke.texi: Add missing param documentation.
30899 2021-03-11 David Malcolm <dmalcolm@redhat.com>
30902 * Makefile.in (ANALYZER_OBJS): Add analyzer/feasible-graph.o and
30903 analyzer/trimmed-graph.o.
30904 * doc/analyzer.texi (Analyzer Paths): Rewrite description of
30905 feasibility checking to reflect new implementation.
30906 * doc/invoke.texi (-fdump-analyzer-feasibility): Document new
30908 * shortest-paths.h (shortest_paths::get_shortest_distance): New.
30910 2021-03-11 David Malcolm <dmalcolm@redhat.com>
30912 * digraph.cc (selftest::test_shortest_paths): Update
30913 shortest_paths init for new param. Add test of
30914 SPS_TO_GIVEN_TARGET.
30915 * shortest-paths.h (enum shortest_path_sense): New.
30916 (shortest_paths::shortest_paths): Add "sense" param.
30917 Update for renamings. Generalize to use "sense" param.
30918 (shortest_paths::get_shortest_path): Rename param.
30919 (shortest_paths::m_sense): New field.
30920 (shortest_paths::m_prev): Rename...
30921 (shortest_paths::m_best_edge): ...to this.
30922 (shortest_paths::get_shortest_path): Update for renamings.
30923 Conditionalize flipping of path on sense of traversal.
30925 2021-03-11 David Malcolm <dmalcolm@redhat.com>
30927 * digraph.cc (selftest::test_shortest_paths): Add test coverage
30928 for paths from B and C.
30929 * shortest-paths.h (shortest_paths::shortest_paths): Handle
30930 unreachable nodes, rather than asserting.
30932 2021-03-11 David Edelsohn <dje.gcc@gmail.com>
30935 * config/rs6000/rs6000.c (rs6000_xcoff_file_start): Don't create
30936 xcoff_tbss_section_name.
30937 * config/rs6000/xcoff.h (ASM_OUTPUT_TLS_COMMON): Use .lcomm.
30938 * xcoffout.c (xcoff_tbss_section_name): Delete.
30939 * xcoffout.h (xcoff_tbss_section_name): Delete.
30941 2021-03-11 Richard Biener <rguenther@suse.de>
30943 PR tree-optimization/99523
30944 * tree-cfg.c (dump_function_to_file): Dump SSA names
30945 w/o identifier to the decls section as well, not only those
30946 without a VAR_DECL.
30948 2021-03-11 Jakub Jelinek <jakub@redhat.com>
30951 * ipa-icf-gimple.c (func_checker::compare_gimple_call): For internal
30952 function calls with lhs fail if the lhs don't have compatible types.
30954 2021-03-11 Hans-Peter Nilsson <hp@axis.com>
30956 * config/cris/cris.h (HARD_FRAME_POINTER_REGNUM): Define.
30957 Change FRAME_POINTER_REGNUM to correspond to a new faked
30958 register faked_fp, part of GENNONACR_REGS like faked_ap.
30959 (CRIS_FAKED_REGS_CONTENTS): New helper macro.
30960 (FIRST_PSEUDO_REGISTER, FIXED_REGISTERS, CALL_USED_REGISTERS):
30961 (REG_ALLOC_ORDER, REG_CLASS_CONTENTS, REGNO_OK_FOR_BASE_P)
30962 (ELIMINABLE_REGS, REGISTER_NAMES): Adjust accordingly.
30963 * config/cris/cris.md (CRIS_FP_REGNUM): Renumber to new faked
30965 (CRIS_REAL_FP_REGNUM): New constant.
30966 * config/cris/cris.c (cris_reg_saved_in_regsave_area): Check
30967 for HARD_FRAME_POINTER_REGNUM instead of FRAME_POINTER_REGNUM.
30968 (cris_initial_elimination_offset): Handle elimination changes
30969 to HARD_FRAME_POINTER_REGNUM instead of FRAME_POINTER_REGNUM
30970 and add one from FRAME_POINTER_REGNUM to
30971 HARD_FRAME_POINTER_REGNUM.
30972 (cris_expand_prologue, cris_expand_epilogue): Emit code for
30973 hard_frame_pointer_rtx instead of frame_pointer_rtx.
30975 2021-03-10 David Edelsohn <dje.gcc@gmail.com>
30978 * config/rs6000/aix.h (ADJUST_FIELD_ALIGN): Add check for DCmode.
30979 * config/rs6000/rs6000.c (rs6000_special_round_type_align): Same.
30981 2021-03-10 Vladimir N. Makarov <vmakarov@redhat.com>
30984 * lra-constraints.c (process_address_1): Don't check unknown
30985 constraint, use X for empty constraint.
30987 2021-03-10 Alex Coplan <alex.coplan@arm.com>
30989 * config/aarch64/aarch64.c (aarch64_vfp_is_call_or_return_candidate):
30990 Fix typo in comment describing "is_ha" argument.
30992 2021-03-10 John David Anglin <danglin@gcc.gnu.org>
30994 * doc/sourcebuild.texi: Document LRA target selector.
30996 2021-03-10 David Malcolm <dmalcolm@redhat.com>
30998 * doc/ux.texi: Add subsection contrasting interactive versus
30999 batch usage of GCC.
31001 2021-03-10 Joel Hutton <joel.hutton@arm.com>
31004 * tree-vect-stmts.c (vectorizable_store): Fix scatter store mask
31006 (vectorizable_load): Fix gather load mask check condition.
31008 2021-03-10 Richard Biener <rguenther@suse.de>
31010 PR tree-optimization/99510
31011 * tree.c (check_aligned_type): Check that the candidate
31012 has TYPE_USER_ALIGN set instead of matching with the
31015 2021-03-10 Eric Botcazou <ebotcazou@adacore.com>
31017 * config/sparc/sparc.c (sparc_regmode_natural_size): Return 4 for
31018 float and vector integer modes only if the mode is not larger.
31020 2021-03-10 Hans-Peter Nilsson <hp@axis.com>
31022 * config/cris/cris.h (DWARF_FRAME_REGISTERS): Define.
31024 2021-03-09 Vladimir N. Makarov <vmakarov@redhat.com>
31026 * ira.c (ira_setup_alts, ira_get_dup_out_num): Process digital
31028 * ira-lives.c (single_reg_class): Ditto.
31030 2021-03-09 Sebastian Huber <sebastian.huber@embedded-brains.de>
31032 * config.gcc (aarch64-*-rtems*): Include general rtems.h after
31033 the architecture-specific rtems.h.
31034 (aarch64-*-rtems*): Likewise.
31035 (arm*-*-rtems*): Likewise.
31036 (epiphany-*-rtems*): Likewise.
31037 (riscv*-*-rtems*): Likewise.
31039 2021-03-09 Jakub Jelinek <jakub@redhat.com>
31041 PR tree-optimization/99305
31042 * tree-ssa-phiopt.c (conditional_replacement): Test integer_pow2p
31043 before integer_all_onesp instead of vice versa.
31045 2021-03-09 Richard Earnshaw <rearnsha@arm.com>
31047 * common/config/arm/arm-common.c (arm_config_default): Change type
31048 of 'i' to unsigned.
31050 2021-03-09 Vladimir N. Makarov <vmakarov@redhat.com>
31053 * lra-constraints.c (process_address_1): Process constraint 'g'
31054 separately and digital constraints containing more one digit.
31056 2021-03-09 Nick Clifton <nickc@redhat.com>
31058 * config/rx/rx.h (DBX_DEBUGGING_INFO): Define.
31059 (DWARF"_DEBUGGING_INFO): Define.
31061 2021-03-09 Eric Botcazou <ebotcazou@adacore.com>
31064 * calls.c (initialize_argument_information): When the argument
31065 is passed by reference, do not make a copy in a thunk only if
31066 the argument is already in memory. Remove redundant test for
31067 the case of callee copy.
31069 2021-03-09 Vladimir N. Makarov <vmakarov@redhat.com>
31072 * lra-constraints.c (process_address_1): Process 0..9 constraints
31073 in process_address_1.
31075 2021-03-09 Andreas Krebbel <krebbel@linux.ibm.com>
31077 * config/s390/s390.c (struct s390_processor processor_table):
31078 Binutils name string must not be empty.
31080 2021-03-09 Claudiu Zissulescu <claziss@synopsys.com>
31082 * config/arc/arc.c (arc_attr_type): Remove function.
31084 2021-03-09 Martin Liska <mliska@suse.cz>
31087 * config/i386/i386-options.c (ix86_option_override_internal):
31088 Set isa_flags for OPTS argument and not for the global
31091 2021-03-09 Aaron Sawdey <acsawdey@linux.ibm.com>
31093 * config/rs6000/predicates.md (ds_form_mem_operand): Check
31096 2021-03-09 Aaron Sawdey <acsawdey@linux.ibm.com>
31099 * config/rs6000/predicates.md (ds_form_mem_operand) New
31101 * config/rs6000/genfusion.pl (gen_ld_cmpi_p10) Use
31102 ds_form_mem_operand in ld/lwa patterns.
31103 * config/rs6000/fusion.md: Regenerate file.
31105 2021-03-08 Martin Sebor <msebor@redhat.com>
31107 PR middle-end/98266
31108 * gimple-array-bounds.cc (inbounds_vbase_memaccess_p): New function.
31109 (array_bounds_checker::check_array_bounds): Call it.
31111 2021-03-08 Martin Sebor <msebor@redhat.com>
31113 PR middle-end/97631
31114 * tree-ssa-strlen.c (maybe_warn_overflow): Test rawmem.
31115 (handle_builtin_stxncpy_strncat): Rename locals. Determine
31116 destination size from allocation calls. Issue a more appropriate
31118 (handle_builtin_memcpy): Pass true as rawmem to maybe_warn_overflow.
31119 (handle_builtin_memset): Same.
31121 2021-03-08 Peter Bergner <bergner@linux.ibm.com>
31124 * config/rs6000/rs6000.c (rs6000_emit_le_vsx_permute): Add an assert
31125 to ensure we do not have an Altivec style address.
31126 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): Disable if passed
31127 an Altivec style address.
31128 (*vsx_le_perm_store_<mode>): Likewise.
31129 (splitters after *vsx_le_perm_store_<mode>): Likewise.
31130 (vsx_load_<mode>): Disable special expander if passed an Altivec
31132 (vsx_store_<mode>): Likewise.
31134 2021-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
31137 * config/aarch64/predicates.md (aarch64_simd_shift_imm_vec_qi): Define.
31138 (aarch64_simd_shift_imm_vec_hi): Likewise.
31139 (aarch64_simd_shift_imm_vec_si): Likewise.
31140 (aarch64_simd_shift_imm_vec_di): Likewise.
31141 * config/aarch64/aarch64-simd.md (aarch64_shrn<mode>_insn_le): Use
31142 predicate from above.
31143 (aarch64_shrn<mode>_insn_be): Likewise.
31144 (aarch64_rshrn<mode>_insn_le): Likewise.
31145 (aarch64_rshrn<mode>_insn_be): Likewise.
31146 (aarch64_shrn2<mode>_insn_le): Likewise.
31147 (aarch64_shrn2<mode>_insn_be): Likewise.
31148 (aarch64_rshrn2<mode>_insn_le): Likewise.
31149 (aarch64_rshrn2<mode>_insn_be): Likewise.
31151 2021-03-08 Vladimir N. Makarov <vmakarov@redhat.com>
31154 * lra-constraints.c (skip_contraint_modifiers): New function.
31155 (process_address_1): Use it before lookup_constraint call.
31157 2021-03-08 Martin Liska <mliska@suse.cz>
31160 * config/i386/i386-options.c (ix86_option_override_internal):
31161 Enable UINTR and HRESET for -march that supports it.
31163 2021-03-08 Ilya Leoshkevich <iii@linux.ibm.com>
31165 * config/s390/s390.c (f_constraint_p): New function.
31166 (s390_md_asm_adjust): Implement TARGET_MD_ASM_ADJUST.
31167 (TARGET_MD_ASM_ADJUST): Likewise.
31169 2021-03-08 Tobias Burnus <tobias@codesourcery.com>
31172 * tree-nested.c (convert_local_reference_stmt): Avoid calling
31173 lookup_field_for_decl for Fortran module (= namespace context).
31175 2021-03-08 Andreas Krebbel <krebbel@linux.ibm.com>
31177 * config/s390/s390.c (s390_expand_vec_compare): Implement <0
31178 comparison with arithmetic right shift.
31179 (s390_expand_vcond): No need for a force_reg anymore.
31180 s390_vec_compare will do it.
31181 * config/s390/vector.md ("vec_cmp<mode><tointvec>"): Accept also
31182 immediate operands.
31184 2021-03-07 Jakub Jelinek <jakub@redhat.com>
31187 * config/i386/constraints.md (Yw): Use SSE_REGS if TARGET_SSE
31188 but TARGET_AVX512BW or TARGET_AVX512VL is not set. Adjust description
31190 * config/i386/sse.md (v_Yw): New define_mode_attr.
31191 (*<insn><mode>3, *mul<mode>3<mask_name>, *avx2_<code><mode>3,
31192 *sse4_1_<code><mode>3<mask_name>): Use <v_Yw> instead of v
31194 * config/i386/mmx.md (mmx_pshufw_1, *vec_dupv4hi): Use Yw instead of
31195 xYw in constraints.
31197 2021-03-06 Julian Brown <julian@codesourcery.com>
31199 * tree-pretty-print.c (dump_generic_node): Emit non-generic
31200 address space info for aggregates.
31202 2021-03-06 Hans-Peter Nilsson <hp@axis.com>
31204 * config/cris/cris.h (MAX_FIXED_MODE_SIZE): Don't define.
31206 2021-03-05 Jakub Jelinek <jakub@redhat.com>
31208 PR middle-end/99322
31209 * tree-cfg.c (bb_to_omp_idx): New variable.
31210 (execute_build_cfg): Release the bb_to_omp_idx vector after
31211 cleanup_tree_cfg returns.
31212 (handle_abnormal_edges): Remove bb_to_omp_idx argument, adjust
31213 for bb_to_omp_idx being a vec<int> instead of pointer to array
31215 (make_edges): Remove bb_to_omp_idx local variable, don't pass
31216 it to handle_abnormal_edges, adjust for bb_to_omp_idx being a
31217 vec<int> instead of pointer to array of ints and don't free/release
31219 (remove_bb): When removing a bb and placing forced label somewhere
31220 else, ensure it is put into the same OpenMP region during cfg
31221 pass if possible or to entry successor as fallback. Unregister
31222 bb from bb_to_omp_idx.
31224 2021-03-05 Vladimir N. Makarov <vmakarov@redhat.com>
31227 * lra-constraints.c (process_address_1): Skip decomposing address
31228 for asm insn operand with unknown constraint.
31230 2021-03-05 Martin Jambor <mjambor@suse.cz>
31233 * cgraph.c (cgraph_edge::set_call_stmt): Do not update all
31234 corresponding speculative edges if we are about to resolve
31235 sepculation. Make edge direct (and so resolve speculations) before
31236 removing it from call_site_hash.
31237 (cgraph_edge::make_direct): Relax the initial assert to allow calling
31238 the function on speculative direct edges.
31240 2021-03-05 Eric Botcazou <ebotcazou@adacore.com>
31242 PR rtl-optimization/99376
31243 * rtlanal.c (nonzero_bits1) <arithmetic operators>: If the number
31244 of low-order zero bits is too large, set the result to 0 directly.
31246 2021-03-04 Jakub Jelinek <jakub@redhat.com>
31248 PR middle-end/93235
31249 * expmed.c (store_bit_field_using_insv): Return false of xop0 is a
31250 SUBREG and a SUBREG to op_mode can't be created.
31252 2021-03-04 Alex Coplan <alex.coplan@arm.com>
31255 * config/aarch64/aarch64-sve-builtins.cc
31256 (function_resolver::require_vector_type): Handle error_mark_node.
31258 2021-03-04 Ilya Leoshkevich <iii@linux.ibm.com>
31260 * cfgexpand.c (expand_asm_loc): Pass new parameter.
31261 (expand_asm_stmt): Likewise.
31262 * config/arm/aarch-common-protos.h (arm_md_asm_adjust): Add new
31264 * config/arm/aarch-common.c (arm_md_asm_adjust): Likewise.
31265 * config/arm/arm.c (thumb1_md_asm_adjust): Likewise.
31266 * config/cris/cris.c (cris_md_asm_adjust): Likewise.
31267 * config/i386/i386.c (ix86_md_asm_adjust): Likewise.
31268 * config/mn10300/mn10300.c (mn10300_md_asm_adjust): Likewise.
31269 * config/nds32/nds32.c (nds32_md_asm_adjust): Likewise.
31270 * config/pdp11/pdp11.c (pdp11_md_asm_adjust): Likewise.
31271 * config/rs6000/rs6000.c (rs6000_md_asm_adjust): Likewise.
31272 * config/vax/vax.c (vax_md_asm_adjust): Likewise.
31273 * config/visium/visium.c (visium_md_asm_adjust): Likewise.
31274 * doc/tm.texi (md_asm_adjust): Likewise.
31275 * target.def (md_asm_adjust): Likewise.
31277 2021-03-04 Richard Biener <rguenther@suse.de>
31279 PR middle-end/97855
31280 * tree-pretty-print.c: Poison pp_printf.
31281 (dump_decl_name): Avoid use of pp_printf.
31282 (dump_block_node): Likewise.
31283 (dump_generic_node): Likewise.
31285 2021-03-04 Martin Sebor <msebor@redhat.com>
31287 PR middle-end/96963
31288 PR middle-end/94655
31289 * builtins.c (handle_array_ref): New helper.
31290 (handle_mem_ref): New helper.
31291 (compute_objsize_r): Factor out ARRAY_REF and MEM_REF handling
31292 into new helper functions. Correct a workaround for vectorized
31295 2021-03-03 Pat Haugen <pthaugen@linux.ibm.com>
31297 * config/rs6000/dfp.md (extendddtd2, trunctddd2, *cmp<mode>_internal1,
31298 floatditd2, ftrunc<mode>2, fix<mode>di2, dfp_ddedpd_<mode>,
31299 dfp_denbcd_<mode>, dfp_dxex_<mode>, dfp_diex_<mode>,
31300 *dfp_sgnfcnc_<mode>, dfp_dscli_<mode>, dfp_dscri_<mode>): Update size
31301 attribute for Power10.
31302 * config/rs6000/mma.md (*movoo): Likewise.
31303 * config/rs6000/rs6000.md (define_attr "size"): Add 256.
31304 (define_mode_attr bits): Add DD/TD modes.
31305 * config/rs6000/sync.md (load_quadpti, store_quadpti, load_lockedpti,
31306 store_conditionalpti): Update size attribute for Power10.
31308 2021-03-03 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
31311 * config/sparc/t-sparc (tree-ssanames.o-warn): Don't error for
31312 -Wuninitialized, -Wmaybe-uninitialized.
31313 (wide-int.o-warn): Likewise.
31315 2021-03-03 Richard Earnshaw <rearnsha@arm.com>
31317 * common/config/arm/arm-common.c: Include configargs.h.
31318 (arm_config_default): New function.
31319 (arm_target_mode): Renamed from arm_target_thumb_only. Handle
31320 processors that do not support Thumb. Take into account the
31321 --with-mode configuration setting for selecting the default.
31322 * config/arm/arm.h (OPTION_DEFAULT_SPECS): Remove entry for 'mode'.
31323 (TARGET_MODE_SPEC_FUNCTIONS): Update for function name change.
31325 2021-03-03 Martin Liska <mliska@suse.cz>
31327 PR gcov-profile/97461
31328 * gcov-io.h (GCOV_PREALLOCATED_KVP): Remove.
31330 2021-03-03 Eric Botcazou <ebotcazou@adacore.com>
31333 * config/i386/i386.c (ix86_compute_frame_layout): For a SEH target,
31334 point back the hard frame pointer to its default location when the
31335 frame is larger than SEH_MAX_FRAME_SIZE.
31337 2021-03-03 Jakub Jelinek <jakub@redhat.com>
31340 * config/i386/predicates.md (logic_operator): New define_predicate.
31341 * config/i386/i386.md (mov + mem using comm arith peephole2):
31342 Punt if operands[1] is EXT_REX_SSE_REGNO_P, AVX512BW is not enabled
31343 and the inner mode is [QH]Imode.
31345 2021-03-03 Jakub Jelinek <jakub@redhat.com>
31348 * dwarf2out.c (dw_loc_list_struct): Add end_entry member.
31349 (new_loc_list): Clear end_entry.
31350 (output_loc_list): Only use DW_LLE_startx_length for -gsplit-dwarf
31351 if HAVE_AS_LEB128, otherwise use DW_LLE_startx_endx. Fix comment
31353 (index_location_lists): For dwarf_version >= 5 without HAVE_AS_LEB128,
31354 initialize also end_entry.
31356 2021-03-03 Jakub Jelinek <jakub@redhat.com>
31359 * cfgrtl.c (fixup_partitions): When changing some bbs from hot to cold
31360 partitions, if in non-layout mode after reorder_blocks also move
31361 affected blocks to ensure a single partition transition.
31363 2021-03-03 Jason Merrill <jason@redhat.com>
31366 * cgraphunit.c (process_function_and_variable_attributes): Don't
31367 warn about flatten on an alias if the target also has it.
31368 * cgraph.h (symtab_node::get_alias_target_tree): New.
31370 2021-03-02 David Edelsohn <dje.gcc@gmail.com>
31372 * config/rs6000/rs6000.md (tls_get_tpointer_internal): Prepend
31373 period to symbol name.
31374 (tls_get_addr_internal<mode>): Same.
31376 2021-03-02 David Malcolm <dmalcolm@redhat.com>
31379 * diagnostic-show-locus.c
31380 (selftest::test_one_liner_many_fixits_2): Fix accidental usage of
31383 2021-03-02 Martin Sebor <msebor@redhat.com>
31385 PR middle-end/99276
31386 * builtins.c (warn_for_access): Remove stray warning text.
31388 2021-03-02 Martin Sebor <msebor@redhat.com>
31390 PR middle-end/99295
31391 * doc/extend.texi (attribute malloc): Reword and clarify nonaliasing
31394 2021-03-02 Jakub Jelinek <jakub@redhat.com>
31397 * dwarf2out.c (output_macinfo_op): Use DW_MACRO_*_str* even with
31398 -gdwarf-5 -gstrict-dwarf. For -gsplit-dwarf -gdwarf-5 use
31399 DW_MACRO_*_strx instead of DW_MACRO_*_strp. Handle
31400 DW_MACRO_define_strx and DW_MACRO_undef_strx.
31401 (save_macinfo_strings): Use DW_MACRO_*_str* even with
31402 -gdwarf-5 -gstrict-dwarf. Handle DW_MACRO_define_strx and
31403 DW_MACRO_undef_strx.
31405 2021-03-02 Andreas Krebbel <krebbel@linux.ibm.com>
31407 * config/s390/s390-builtin-types.def (BT_FN_V4SF_V8HI_UINT): New
31409 (BT_FN_V8HI_V8HI_UINT): Likewise.
31410 (BT_FN_V8HI_V4SF_V4SF_UINT): Likewise.
31411 * config/s390/s390-builtins.def (B_NNPA): New macro definition.
31412 (s390_vclfnhs, s390_vclfnls, s390_vcrnfs, s390_vcfn, s390_vcnf):
31413 New builtin definitions.
31414 * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal): Bump
31415 vector extension version.
31416 * config/s390/s390.c (s390_expand_builtin): Check if builtins are
31417 available with current -march level.
31418 * config/s390/s390.md (UNSPEC_NNPA_VCLFNHS_V8HI)
31419 (UNSPEC_NNPA_VCLFNLS_V8HI, UNSPEC_NNPA_VCRNFS_V8HI)
31420 (UNSPEC_NNPA_VCFN_V8HI, UNSPEC_NNPA_VCNF_V8HI): New constants.
31421 * config/s390/vecintrin.h (vec_extend_to_fp32_hi): New macro.
31422 (vec_extend_to_fp32_lo): Likewise.
31423 (vec_round_from_fp32): Likewise.
31424 (vec_convert_to_fp16): Likewise.
31425 (vec_convert_from_fp16): Likewise.
31426 * config/s390/vx-builtins.md (vclfnhs_v8hi): New insn pattern.
31427 (vclfnls_v8hi): Likewise.
31428 (vcrnfs_v8hi): Likewise.
31429 (vcfn_v8hi): Likewise.
31430 (vcnf_v8hi): Likewise.
31432 2021-03-02 Andreas Krebbel <krebbel@linux.ibm.com>
31434 * common/config/s390/s390-common.c (processor_flags_table): New entry.
31435 * config.gcc: Enable arch14 for --with-arch and --with-tune.
31436 * config/s390/driver-native.c (s390_host_detect_local_cpu): Pick
31437 arch14 for unknown CPU models.
31438 * config/s390/s390-opts.h (enum processor_type): Add PROCESSOR_ARCH14.
31439 * config/s390/s390.c (s390_issue_rate): Add case for PROCESSOR_ARCH14.
31440 (s390_get_sched_attrmask): Likewise.
31441 (s390_get_unit_mask): Likewise.
31442 * config/s390/s390.h (enum processor_flags): Add PF_NNPA and PF_ARCH14.
31443 (TARGET_CPU_ARCH14, TARGET_CPU_ARCH14_P, TARGET_CPU_NNPA)
31444 (TARGET_CPU_NNPA_P, TARGET_ARCH14, TARGET_ARCH14_P, TARGET_NNPA)
31445 (TARGET_NNPA_P): New macro definitions.
31446 * config/s390/s390.md ("cpu_facility", "enabled"): Add arch14 and nnpa.
31447 * config/s390/s390.opt: Add PROCESSOR_ARCH14.
31449 2021-03-02 Jakub Jelinek <jakub@redhat.com>
31451 PR middle-end/95757
31452 * tree-vrp.c (register_edge_assert_for): Remove superfluous ()s around
31453 condition. Call register_edge_assert_for_1 for == 0, != 0, == 1 and
31454 != 1 comparisons if name is lhs of a comparison.
31456 2021-03-01 Iain Sandoe <iain@sandoe.co.uk>
31460 * config/darwin-protos.h (darwin_should_restore_cfa_state): New.
31461 * config/darwin.c (darwin_should_restore_cfa_state): New.
31462 * config/darwin.h (TARGET_ASM_SHOULD_RESTORE_CFA_STATE): New.
31463 * doc/tm.texi: Regenerated.
31464 * doc/tm.texi.in: Document TARGET_ASM_SHOULD_RESTORE_CFA_STATE.
31465 * dwarf2cfi.c (connect_traces): If the target requests, restore
31466 the CFA expression after a DW_CFA_restore.
31467 * target.def (TARGET_ASM_SHOULD_RESTORE_CFA_STATE): New hook.
31469 2021-03-01 Martin Liska <mliska@suse.cz>
31472 * optc-save-gen.awk: Add 4 more exceptions.
31474 2021-03-01 Nathan Sidwell <nathan@acm.org>
31477 * tree.h (TYPE_ALIGN_RAW): New accessor.
31478 (TYPE_ALIGN): Use it.
31480 2021-03-01 Jan Hubicka <jh@suse.cz>
31483 * ipa-fnsummary.c (compute_fn_summary): Fix sanity check.
31485 2021-03-01 Eric Botcazou <ebotcazou@adacore.com>
31488 * config/i386/i386.c (ix86_compute_frame_layout): For a SEH target,
31489 point the hard frame pointer to the SSE register save area instead
31490 of the general register save area. Perform only minimal adjustment
31491 for small frames if it is initially not correctly aligned.
31492 (ix86_expand_prologue): Remove early saves for a SEH target.
31493 * config/i386/winnt.c (struct seh_frame_state): Document constraint.
31495 2021-02-28 Jakub Jelinek <jakub@redhat.com>
31498 * ipa.c (symbol_table::remove_unreachable_nodes): Fix a comment
31499 typo - referneced -> referenced.
31500 * tree.c (component_ref_size): Fix comment typo -
31501 refernce -> reference.
31502 * tree-ssa-alias.c (access_path_may_continue_p): Fix comment typo -
31503 traling -> trailing.
31504 (aliasing_component_refs_p): Fix comment typos -
31505 refernce -> reference and refernece -> reference and
31506 traling -> trailing.
31507 (nonoverlapping_refs_since_match_p): Fix comment typo -
31508 referneces -> references.
31509 * doc/invoke.texi (--param modref-max-bases): Fix a typo -
31510 referneces -> references.
31512 2021-02-27 Iain Sandoe <iain@sandoe.co.uk>
31514 * config/host-darwin.c (darwin_gt_pch_use_address): Modify
31515 diagnostic message to avoid use of a contraction and format
31518 2021-02-27 Jakub Jelinek <jakub@redhat.com>
31521 * gcse.c (gcse_or_cprop_is_too_expensive): Use %wu instead of
31522 HOST_WIDE_INT_PRINT_UNSIGNED in warning format string.
31523 * ipa-devirt.c (ipa_odr_read_section): Use %wd instead of
31524 HOST_WIDE_INT_PRINT_DEC in inform format string. Fix comment
31527 2021-02-26 Richard Biener <rguenther@suse.de>
31529 PR middle-end/99281
31530 * expr.c (store_field): For calls with return-slot optimization
31531 and addressable return type expand the store directly.
31533 2021-02-26 Richard Biener <rguenther@suse.de>
31536 * builtins.c (warn_string_no_nul): Fix diagnostic formatting.
31538 2021-02-26 Peter Bergner <bergner@linux.ibm.com>
31541 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Replace assert
31544 2021-02-26 Aaron Sawdey <acsawdey@linux.ibm.com>
31546 * config.gcc: Add rs6000-pcrel-opt.o.
31547 * config/rs6000/rs6000-pcrel-opt.c: New file.
31548 * config/rs6000/pcrel-opt.md: New file.
31549 * config/rs6000/predicates.md: Add d_form_memory predicate.
31550 * config/rs6000/rs6000-cpus.def: Add OPTION_MASK_PCREL_OPT.
31551 * config/rs6000/rs6000-passes.def: Add pass_pcrel_opt.
31552 * config/rs6000/rs6000-protos.h: Add reg_to_non_prefixed(),
31553 pcrel_opt_valid_mem_p(), output_pcrel_opt_reloc(),
31554 and make_pass_pcrel_opt().
31555 * config/rs6000/rs6000.c (reg_to_non_prefixed): Make global.
31556 (rs6000_option_override_internal): Add pcrel-opt.
31557 (rs6000_delegitimize_address): Support pcrel-opt.
31558 (rs6000_opt_masks): Add pcrel-opt.
31559 (pcrel_opt_valid_mem_p): New function.
31560 (reg_to_non_prefixed): Make global.
31561 (rs6000_asm_output_opcode): Reset prepend_p_to_next_insn.
31562 (output_pcrel_opt_reloc): New function.
31563 * config/rs6000/rs6000.md (loads_extern_addr): New attr.
31564 (pcrel_extern_addr): Set loads_extern_addr.
31565 Add include for pcrel-opt.md.
31566 * config/rs6000/rs6000.opt: Add -mpcrel-opt.
31567 * config/rs6000/t-rs6000: Add rules for pcrel-opt.c and
31570 2021-02-26 YunQiang Su <yunqiang.su@cipunited.com>
31573 * config/mips/mips.c (mips_expand_ext_as_unaligned_load):
31574 If TARGET_64BIT and dest is SUBREG, we check the width, if it
31575 equal to SImode, we use SImode operation, just like what we are
31578 2021-02-26 Marek Polacek <polacek@redhat.com>
31580 * builtins.c (warn_for_access): Fix typos.
31582 2021-02-25 Iain Sandoe <iain@sandoe.co.uk>
31584 * config/aarch64/aarch64.md (<optab>_rol<mode>3): Add a '#'
31585 mark in front of the immediate quantity.
31586 (<optab>_rolsi3_uxtw): Likewise.
31588 2021-02-25 Richard Earnshaw <rearnsha@arm.com>
31591 * config/arm/thumb2.md (nonsecure_call_reg_thumb2_fpcxt): New pattern.
31592 (nonsecure_call_value_reg_thumb2_fpcxt): Likewise.
31593 (nonsecure_call_reg_thumb2): Restrict to using r4 for the callee
31594 address and disable when the FPCXT is not available.
31595 (nonsecure_call_value_reg_thumb2): Likewise.
31597 2021-02-25 Nathan Sidwell <nathan@acm.org>
31600 * doc/invoke.texi (flang-info-module-cmi): Renamed option.
31602 2021-02-25 Tamar Christina <tamar.christina@arm.com>
31604 * tree-vect-slp.c (optimize_load_redistribution_1): Abort on NULL nodes.
31606 2021-02-25 Richard Biener <rguenther@suse.de>
31608 PR tree-optimization/99253
31609 * tree-vect-loop.c (check_reduction_path): First compute
31610 code, then verify out-of-loop uses.
31612 2021-02-25 Jakub Jelinek <jakub@redhat.com>
31615 * match.pd ((T)(A) + CST -> (T)(A + CST)): Add :s to convert.
31617 2021-02-25 Jakub Jelinek <jakub@redhat.com>
31619 PR tree-optimization/80635
31620 * tree-vrp.c (vrp_simplify_cond_using_ranges): Also handle
31621 VIEW_CONVERT_EXPR if modes are the same, innerop is integral and
31622 has mode precision.
31624 2021-02-25 Richard Biener <rguenther@suse.de>
31626 * tree-vect-slp.c (optimize_load_redistribution_1): Delay
31627 load_map population.
31628 (vect_match_slp_patterns_2): Revert part of last change.
31629 (vect_analyze_slp): Do not interleave optimize_load_redistribution
31630 with pattern detection but do it afterwards. Dump the
31631 whole SLP graph after pattern recognition and load
31632 redistribution optimization finished.
31634 2021-02-24 Jakub Jelinek <jakub@redhat.com>
31637 * omp-low.c (struct omp_context): Add teams_nested_p and
31638 nonteams_nested_p members.
31639 (scan_omp_target): Diagnose teams nested inside of target with other
31640 directives strictly nested inside of the same target.
31641 (check_omp_nesting_restrictions): Set ctx->teams_nested_p or
31642 ctx->nonteams_nested_p as needed.
31644 2021-02-24 Vladimir N. Makarov <vmakarov@redhat.com>
31646 PR inline-asm/99123
31647 * lra-constraints.c (uses_hard_regs_p): Don't use decompose_mem_address.
31649 2021-02-24 Hans-Peter Nilsson <hp@axis.com>
31651 * config/cris/cris.c (cris_expand_prologue): Set
31652 current_function_static_stack_size, if flag_stack_usage_info.
31654 2021-02-24 Pat Haugen <pthaugen@linux.ibm.com>
31656 * config/rs6000/rs6000.c (next_insn_prefixed_p): Rename.
31657 (rs6000_final_prescan_insn): Adjust.
31658 (rs6000_asm_output_opcode): Likewise.
31660 2021-02-24 Martin Sebor <msebor@redhat.com>
31662 PR middle-end/97172
31663 * attribs.c (attr_access::free_lang_data): Clear attribute arg spec
31664 from function arguments.
31666 2021-02-24 Tamar Christina <tamar.christina@arm.com>
31668 PR tree-optimization/99220
31669 * tree-vect-slp.c (optimize_load_redistribution_1): Remove
31670 node from cache when it's about to be deleted.
31672 2021-02-24 Jakub Jelinek <jakub@redhat.com>
31674 PR tree-optimization/99225
31675 * fold-const.c (fold_binary_loc) <case NE_EXPR>: In (x & (1 << y)) != 0
31676 to ((x >> y) & 1) != 0 simplifications use build_one_cst instead of
31677 build_int_cst (..., 1). Formatting fixes.
31679 2021-02-24 Tamar Christina <tamar.christina@arm.com>
31681 PR tree-optimization/99149
31682 * tree-vect-slp-patterns.c (vect_detect_pair_op): Don't recreate the
31684 (vect_slp_reset_pattern): Remove.
31685 (complex_fma_pattern::matches): Remove call to vect_slp_reset_pattern.
31686 (complex_mul_pattern::build, complex_fma_pattern::build,
31687 complex_fms_pattern::build): Fix ref counts.
31688 * tree-vect-slp.c (vect_free_slp_tree): Undo SLP only pattern relevancy
31689 when node is being deleted.
31690 (vect_match_slp_patterns_2): Correct result of cache hit on patterns.
31691 (vect_schedule_slp): Invalidate SLP_TREE_REPRESENTATIVE of removed
31693 * tree-vectorizer.c (vec_info::new_stmt_vec_info): Initialize value.
31695 2021-02-24 Matthias Klose <doko@ubuntu.com>
31698 2020-12-07 Matthias Klose <doko@ubuntu.com>
31700 * genextract.c (print_header): Undefine ENABLE_RTL_CHECKING
31701 and ENABLE_RTL_FLAG_CHECKING.
31703 2021-02-24 Richard Biener <rguenther@suse.de>
31706 * builtins.c (fold_builtin_next_arg): Avoid NULL arg.
31708 2021-02-23 Peter Bergner <bergner@linux.ibm.com>
31710 * config/rs6000/mma.md (mma_assemble_pair): Rename from this...
31711 (vsx_assemble_pair): ...to this.
31712 (*mma_assemble_pair): Rename from this...
31713 (*vsx_assemble_pair): ...to this.
31714 (mma_disassemble_pair): Rename from this...
31715 (vsx_disassemble_pair): ...to this.
31716 (*mma_disassemble_pair): Rename from this...
31717 (*vsx_disassemble_pair): ...to this.
31718 * config/rs6000/rs6000-builtin.def (BU_MMA_V2, BU_MMA_V3,
31719 BU_COMPAT): New macros.
31720 (mma_assemble_pair): Rename from this...
31721 (vsx_assemble_pair): ...to this.
31722 (mma_disassemble_pair): Rename from this...
31723 (vsx_disassemble_pair): ...to this.
31724 (mma_assemble_pair): New compatibility built-in.
31725 (mma_disassemble_pair): Likewise.
31726 * config/rs6000/rs6000-call.c (struct builtin_compatibility): New.
31727 (RS6000_BUILTIN_COMPAT): Define.
31728 (bdesc_compat): New.
31729 (mma_expand_builtin): Use VSX_BUILTIN_DISASSEMBLE_PAIR_INTERNAL.
31730 (rs6000_gimple_fold_mma_builtin): Use MMA_BUILTIN_DISASSEMBLE_PAIR
31731 and VSX_BUILTIN_ASSEMBLE_PAIR.
31732 (rs6000_init_builtins): Register compatibility built-ins.
31733 (mma_init_builtins): Use VSX_BUILTIN_ASSEMBLE_PAIR,
31734 VSX_BUILTIN_ASSEMBLE_PAIR_INTERNAL, VSX_BUILTIN_DISASSEMBLE_PAIR and
31735 VSX_BUILTIN_DISASSEMBLE_PAIR_INTERNAL.
31736 * doc/extend.texi (__builtin_mma_assemble_pair): Rename from this...
31737 (__builtin_vsx_assemble_pair): ...to this.
31738 (__builtin_mma_disassemble_pair): Rename from this...
31739 (__builtin_vsx_disassemble_pair): ...to this.
31741 2021-02-23 Martin Liska <mliska@suse.cz>
31744 * ipa-icf.c (sem_variable::merge): Do not merge 2 variables
31745 with different alignment. That leads to an invalid red zone
31746 size allocated in runtime.
31748 2021-02-23 Jakub Jelinek <jakub@redhat.com>
31750 PR tree-optimization/99204
31751 * fold-const.c (fold_read_from_constant_string): Check that
31752 tree_fits_uhwi_p (index) rather than just that index is INTEGER_CST.
31754 2021-02-23 Segher Boessenkool <segher@kernel.crashing.org>
31755 Kewen Lin <linkw@gcc.gnu.org>
31757 * config/rs6000/rs6000.md (*rotl<mode>3_insert_3): Renamed to...
31758 (rotl<mode>3_insert_3): ...this.
31759 (plus_ior_xor): New code_iterator.
31760 (define_split for GPR rl*imi): New splitter.
31761 * config/rs6000/vsx.md (vsx_init_v4si): Use gen_rotldi3_insert_3
31762 for integer merging.
31764 2021-02-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
31766 * config/aarch64/aarch64-tuning-flags.def (cse_sve_vl_constants):
31768 * config/aarch64/aarch64.md (add<mode>3): Force CONST_POLY_INT immediates
31769 into a register when the above is enabled.
31770 * config/aarch64/aarch64.c (neoversev1_tunings):
31771 AARCH64_EXTRA_TUNE_CSE_SVE_VL_CONSTANTS.
31772 (aarch64_rtx_costs): Use AARCH64_EXTRA_TUNE_CSE_SVE_VL_CONSTANTS.
31774 2021-02-22 Hans-Peter Nilsson <hp@axis.com>
31776 * config/cris/cris.c (cris_print_operand) <'T'>: Change
31777 valid operand from is now an addi mult-value to shift-value.
31778 * config/cris/cris.md (*addi): Change expression of scaled
31779 operand from mult to ashift.
31780 * config/cris/cris.md (*addi_reload): New insn_and_split.
31782 2021-02-22 John David Anglin <danglin@gcc.gnu.org>
31785 * config/pa/pa.c (TARGET_ASM_CAN_OUTPUT_MI_THUNK): Define as
31786 hook_bool_const_tree_hwi_hwi_const_tree_true.
31787 (pa_asm_output_mi_thunk): Add support for nonzero vcall_offset.
31789 2021-02-22 Andre Vieira <andre.simoesdiasvieira@arm.com>
31791 PR rtl-optimization/98791
31792 * ira-conflicts.c (process_regs_for_copy): Don't create allocno copies
31793 for unordered modes.
31795 2021-02-22 Martin Liska <mliska@suse.cz>
31797 * tree-inline.c (inline_forbidden_p): Set
31798 inline_forbidden_reason.
31800 2021-02-22 Richard Biener <rguenther@suse.de>
31802 * tree-vect-slp.c (vect_bb_vectorization_profitable_p): Dump
31805 2021-02-22 Richard Biener <rguenther@suse.de>
31807 PR tree-optimization/99165
31808 * gimple-ssa-store-merging.c (pass_store_merging::process_store):
31809 Accumulate changed to ret.
31811 2021-02-21 Uros Bizjak <ubizjak@gmail.com>
31814 2020-12-09 Uroš Bizjak <ubizjak@gmail.com>
31816 * config/i386/i386.h (REG_ALLOC_ORDER): Remove
31818 2021-02-20 Ilya Leoshkevich <iii@linux.ibm.com>
31821 * config/s390/vector.md (trunctf<DFP_ALL:mode>2_vr): New
31823 (trunctf<DFP_ALL:mode>2): Likewise.
31824 (trunctdtf2_vr): Likewise.
31825 (trunctdtf2): Likewise.
31826 (extend<DFP_ALL:mode>tf2_vr): Likewise.
31827 (extend<DFP_ALL:mode>tf2): Likewise.
31828 (extendtftd2_vr): Likewise.
31829 (extendtftd2): Likewise.
31831 2021-02-20 Ilya Leoshkevich <iii@linux.ibm.com>
31833 * config/s390/vector.md (*fprx2_to_tf): Rename to fprx2_to_tf,
31834 add memory alternative.
31835 (tf_to_fprx2): New pattern.
31837 2021-02-19 Martin Sebor <msebor@redhat.com>
31840 * attribs.c (init_attr_rdwr_indices): Guard vblist use.
31841 (attr_access::free_lang_data): Remove a spurious test.
31843 2021-02-19 Nathan Sidwell <nathan@acm.org>
31845 * doc/invoke.texi (flang-info-module-read): Document.
31847 2021-02-19 Martin Liska <mliska@suse.cz>
31849 PR translation/99167
31850 * params.opt: Fix typo.
31852 2021-02-19 Richard Biener <rguenther@suse.de>
31854 PR middle-end/99122
31855 * tree-inline.c (inline_forbidden_p): Do not inline functions
31856 with VLA arguments or return value.
31858 2021-02-19 Jakub Jelinek <jakub@redhat.com>
31861 * config/arm/arm.md (*stack_protect_combined_set_insn,
31862 *stack_protect_combined_test_insn): If force_const_mem result
31863 is not valid general operand, force its address into the destination
31866 2021-02-19 Jakub Jelinek <jakub@redhat.com>
31869 * tree-cfg.c (gimple_merge_blocks): If bb a starts with eh landing
31870 pad or non-local label, put FORCED_LABELs from bb b after that label
31871 rather than before it.
31873 2021-02-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
31876 * config/aarch64/aarch64-sve.md (<ASHIFT:optab><mode>3): Use
31877 expand_vector_broadcast' to emit the vec_duplicate operand.
31879 2021-02-18 Vladimir N. Makarov <vmakarov@redhat.com>
31881 PR rtl-optimization/96264
31882 * lra-remat.c (reg_overlap_for_remat_p): Check also output insn
31885 2021-02-18 H.J. Lu <hjl.tools@gmail.com>
31888 * varasm.c (get_section): Replace SUPPORTS_SHF_GNU_RETAIN with
31889 looking up the retain attribute.
31890 (resolve_unique_section): Likewise.
31891 (get_variable_section): Likewise.
31892 (switch_to_section): Likewise. Warn when a symbol without the
31893 retain attribute and a symbol with the retain attribute are
31894 placed in the section with the same name, instead of the used
31896 * doc/extend.texi: Document the "retain" attribute.
31898 2021-02-18 Nathan Sidwell <nathan@acm.org>
31901 * doc/invoke.texi (flang-info-include-translate): Document header
31904 2021-02-18 Richard Biener <rguenther@suse.de>
31906 PR middle-end/99122
31907 * ipa-fnsummary.c (analyze_function_body): Set
31908 CIF_FUNCTION_NOT_INLINABLE for VLA parameter calls.
31909 * tree-inline.c (insert_init_debug_bind): Pass NULL for
31910 error_mark_node values.
31911 (force_value_to_type): Do not build V_C_Es for WITH_SIZE_EXPR
31913 (setup_one_parameter): Delay force_value_to_type until when
31916 2021-02-18 Hans-Peter Nilsson <hp@axis.com>
31918 PR tree-optimization/99142
31919 * match.pd (clz cmp 0): Gate replacement on single_use of clz result.
31921 2021-02-18 Jakub Jelinek <jakub@redhat.com>
31923 * wide-int-bitmask.h (wide_int_bitmask::wide_int_bitmask (),
31924 wide_int_bitmask::wide_int_bitmask (uint64_t),
31925 wide_int_bitmask::wide_int_bitmask (uint64_t, uint64_t),
31926 wide_int_bitmask::operator ~ () const,
31927 wide_int_bitmask::operator | (wide_int_bitmask) const,
31928 wide_int_bitmask::operator & (wide_int_bitmask) const): Use constexpr
31930 * config/i386/i386.h (PTA_3DNOW, PTA_3DNOW_A, PTA_64BIT, PTA_ABM,
31931 PTA_AES, PTA_AVX, PTA_BMI, PTA_CX16, PTA_F16C, PTA_FMA, PTA_FMA4,
31932 PTA_FSGSBASE, PTA_LWP, PTA_LZCNT, PTA_MMX, PTA_MOVBE, PTA_NO_SAHF,
31933 PTA_PCLMUL, PTA_POPCNT, PTA_PREFETCH_SSE, PTA_RDRND, PTA_SSE, PTA_SSE2,
31934 PTA_SSE3, PTA_SSE4_1, PTA_SSE4_2, PTA_SSE4A, PTA_SSSE3, PTA_TBM,
31935 PTA_XOP, PTA_AVX2, PTA_BMI2, PTA_RTM, PTA_HLE, PTA_PRFCHW, PTA_RDSEED,
31936 PTA_ADX, PTA_FXSR, PTA_XSAVE, PTA_XSAVEOPT, PTA_AVX512F, PTA_AVX512ER,
31937 PTA_AVX512PF, PTA_AVX512CD, PTA_NO_TUNE, PTA_SHA, PTA_PREFETCHWT1,
31938 PTA_CLFLUSHOPT, PTA_XSAVEC, PTA_XSAVES, PTA_AVX512DQ, PTA_AVX512BW,
31939 PTA_AVX512VL, PTA_AVX512IFMA, PTA_AVX512VBMI, PTA_CLWB, PTA_MWAITX,
31940 PTA_CLZERO, PTA_NO_80387, PTA_PKU, PTA_AVX5124VNNIW, PTA_AVX5124FMAPS,
31941 PTA_AVX512VPOPCNTDQ, PTA_SGX, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES,
31942 PTA_AVX512VBMI2, PTA_VPCLMULQDQ, PTA_AVX512BITALG, PTA_RDPID,
31943 PTA_PCONFIG, PTA_WBNOINVD, PTA_AVX512VP2INTERSECT, PTA_PTWRITE,
31944 PTA_AVX512BF16, PTA_WAITPKG, PTA_MOVDIRI, PTA_MOVDIR64B, PTA_ENQCMD,
31945 PTA_CLDEMOTE, PTA_SERIALIZE, PTA_TSXLDTRK, PTA_AMX_TILE, PTA_AMX_INT8,
31946 PTA_AMX_BF16, PTA_UINTR, PTA_HRESET, PTA_KL, PTA_WIDEKL, PTA_AVXVNNI,
31947 PTA_X86_64_BASELINE, PTA_X86_64_V2, PTA_X86_64_V3, PTA_X86_64_V4,
31948 PTA_CORE2, PTA_NEHALEM, PTA_WESTMERE, PTA_SANDYBRIDGE, PTA_IVYBRIDGE,
31949 PTA_HASWELL, PTA_BROADWELL, PTA_SKYLAKE, PTA_SKYLAKE_AVX512,
31950 PTA_CASCADELAKE, PTA_COOPERLAKE, PTA_CANNONLAKE, PTA_ICELAKE_CLIENT,
31951 PTA_ICELAKE_SERVER, PTA_TIGERLAKE, PTA_SAPPHIRERAPIDS, PTA_ALDERLAKE,
31952 PTA_KNL, PTA_BONNELL, PTA_SILVERMONT, PTA_GOLDMONT, PTA_GOLDMONT_PLUS,
31953 PTA_TREMONT, PTA_KNM): Use constexpr instead of const.
31955 2021-02-18 Jakub Jelinek <jakub@redhat.com>
31957 PR middle-end/99109
31958 * gimple-array-bounds.cc (build_zero_elt_array_type): Rename to ...
31959 (build_printable_array_type): ... this. Add nelts argument. For
31960 overaligned eltype, use TYPE_MAIN_VARIANT (eltype) instead. If
31961 nelts, call build_array_type_nelts.
31962 (array_bounds_checker::check_mem_ref): Use build_printable_array_type
31963 instead of build_zero_elt_array_type and build_array_type_nelts.
31965 2021-02-18 Jakub Jelinek <jakub@redhat.com>
31968 * config/i386/i386.c (distance_non_agu_define): Don't call
31969 extract_insn_cached here.
31970 (ix86_lea_outperforms): Save and restore recog_data around call
31971 to distance_non_agu_define and distance_agu_use.
31972 (ix86_ok_to_clobber_flags): Remove.
31973 (ix86_avoid_lea_for_add): Don't call ix86_ok_to_clobber_flags.
31974 (ix86_avoid_lea_for_addr): Likewise. Adjust function comment.
31975 * config/i386/i386.md (*lea<mode>): Change from define_insn_and_split
31976 into define_insn. Move the splitting to define_peephole2 and
31977 check there using peep2_regno_dead_p if FLAGS_REG is dead.
31979 2021-02-17 Julian Brown <julian@codesourcery.com>
31981 * gimplify.c (gimplify_scan_omp_clauses): Handle ATTACH_DETACH
31984 2021-02-17 Xi Ruoyao <xry111@mengyan1223.wang>
31987 * config/mips/mips.c (mips_symbol_insns): Do not use
31988 MSA_SUPPORTED_MODE_P if mode is MAX_MACHINE_MODE.
31990 2021-02-16 Vladimir N. Makarov <vmakarov@redhat.com>
31992 PR inline-asm/98096
31993 * stmt.c (resolve_operand_name_1): Take inout operands into account
31994 for access to labels by names.
31995 * doc/extend.texi: Describe counting operands for accessing labels.
31997 2021-02-16 Richard Biener <rguenther@suse.de>
31999 PR tree-optimization/38474
32000 * tree-ssa-structalias.c (variable_info::address_taken): New.
32001 (new_var_info): Initialize address_taken.
32002 (process_constraint): Set address_taken.
32003 (solve_constraints): Use the new address_taken flag rather
32004 than is_reg_var for sorting variables.
32005 (dump_constraint): Dump the variable number if the name
32008 2021-02-16 Jakub Jelinek <jakub@redhat.com>
32011 * tree-vect-stmts.c (vectorizable_simd_clone_call): For num_calls != 1
32012 multiply by 4096 and for inbranch by 8192.
32013 * config/i386/i386.c (ix86_simd_clone_usable): For TARGET_AVX512F,
32014 return 3, 2 or 1 for mangle letters 'b', 'c' or 'd'.
32016 2021-02-15 Maya Rashish <coypu@sdf.org>
32018 * config/aarch64/aarch64.c (aarch64_init_builtins):
32019 Call SUBTARGET_INIT_BUILTINS.
32021 2021-02-15 Peter Bergner <bergner@linux.ibm.com>
32023 PR rtl-optimization/98872
32024 * init-regs.c (initialize_uninitialized_regs): Skip initialization
32025 if CONST0_RTX is NULL.
32027 2021-02-15 Richard Sandiford <richard.sandiford@arm.com>
32029 PR rtl-optimization/98863
32030 * rtl-ssa/functions.h (function_info::bb_live_out_info): Delete.
32031 (function_info::build_info): Turn into a declaration, moving the
32032 definition to internals.h.
32033 (function_info::bb_walker): Declare.
32034 (function_info::create_reg_use): Likewise.
32035 (function_info::calculate_potential_phi_regs): Take a build_info
32037 (function_info::place_phis, function_info::create_ebbs): Declare.
32038 (function_info::calculate_ebb_live_in_for_debug): Likewise.
32039 (function_info::populate_backedge_phis): Delete.
32040 (function_info::start_block, function_info::end_block): Declare.
32041 (function_info::populate_phi_inputs): Delete.
32042 (function_info::m_potential_phi_regs): Move information to build_info.
32043 * rtl-ssa/internals.h: New file.
32044 (function_info::bb_phi_info): New class.
32045 (function_info::build_info): Moved from functions.h.
32046 Add a constructor and destructor.
32047 (function_info::build_info::ebb_use): Delete.
32048 (function_info::build_info::ebb_def): Likewise.
32049 (function_info::build_info::bb_live_out): Likewise.
32050 (function_info::build_info::tmp_ebb_live_in_for_debug): New variable.
32051 (function_info::build_info::potential_phi_regs): Likewise.
32052 (function_info::build_info::potential_phi_regs_for_debug): Likewise.
32053 (function_info::build_info::ebb_def_regs): Likewise.
32054 (function_info::build_info::bb_phis): Likewise.
32055 (function_info::build_info::bb_mem_live_out): Likewise.
32056 (function_info::build_info::bb_to_rpo): Likewise.
32057 (function_info::build_info::def_stack): Likewise.
32058 (function_info::build_info::old_def_stack_limit): Likewise.
32059 * rtl-ssa/internals.inl (function_info::build_info::record_reg_def):
32060 Remove the regno argument. Push the previous definition onto the
32061 definition stack where necessary.
32062 * rtl-ssa/accesses.cc: Include internals.h.
32063 * rtl-ssa/changes.cc: Likewise.
32064 * rtl-ssa/blocks.cc: Likewise.
32065 (function_info::build_info::build_info): Define.
32066 (function_info::build_info::~build_info): Likewise.
32067 (function_info::bb_walker): New class.
32068 (function_info::bb_walker::bb_walker): Define.
32069 (function_info::add_live_out_use): Convert a logarithmic-complexity
32070 test into a linear one. Allow the same definition to be passed
32072 (function_info::calculate_potential_phi_regs): Moved from
32073 functions.cc. Take a build_info parameter and store the
32074 information there instead.
32075 (function_info::place_phis): New function.
32076 (function_info::add_entry_block_defs): Update call to record_reg_def.
32077 (function_info::calculate_ebb_live_in_for_debug): New function.
32078 (function_info::add_phi_nodes): Use bb_phis to decide which
32079 registers need phi nodes and initialize ebb_def_regs accordingly.
32080 Do not add degenerate phis here.
32081 (function_info::add_artificial_accesses): Use create_reg_use.
32082 Assert that all definitions are listed in the DF LR sets.
32083 Update call to record_reg_def.
32084 (function_info::record_block_live_out): Record live-out register
32085 values in the phis of successor blocks. Use the live-out set
32086 when processing the last block in an EBB, instead of always
32087 using the live-in sets of successor blocks. AND the live sets
32088 with the set of registers that have been defined in the EBB,
32089 rather than with all potential phi registers. Cope correctly
32090 with branches back to the start of the current EBB.
32091 (function_info::start_block): New function.
32092 (function_info::end_block): Likewise.
32093 (function_info::populate_phi_inputs): Likewise.
32094 (function_info::create_ebbs): Likewise.
32095 (function_info::process_all_blocks): Rewrite into a multi-phase
32097 * rtl-ssa/functions.cc: Include internals.h.
32098 (function_info::calculate_potential_phi_regs): Move to blocks.cc.
32099 (function_info::init_function_data): Remove caller.
32100 * rtl-ssa/insns.cc: Include internals.h
32101 (function_info::create_reg_use): New function. Lazily any
32102 degenerate phis needed by the linear RPO view.
32103 (function_info::record_use): Use create_reg_use. When processing
32104 debug uses, use potential_phi_regs and test it before checking
32105 whether the register is live on entry to the current EBB. Lazily
32106 calculate ebb_live_in_for_debug.
32107 (function_info::record_call_clobbers): Update call to record_reg_def.
32108 (function_info::record_def): Likewise.
32110 2021-02-15 Martin Liska <mliska@suse.cz>
32112 * toplev.c (init_asm_output): Free output of
32113 gen_command_line_string function.
32114 (process_options): Likewise.
32116 2021-02-15 Martin Liska <mliska@suse.cz>
32118 * params.opt: Add 2 missing Param keywords.
32120 2021-02-15 Eric Botcazou <ebotcazou@adacore.com>
32122 * df-core.c (df_worklist_dataflow_doublequeue): Use proper cast.
32124 2021-02-15 Jakub Jelinek <jakub@redhat.com>
32126 PR tree-optimization/99079
32127 * match.pd (A % (pow2pcst << N) -> A & ((pow2pcst << N) - 1)): Remove
32128 useless tree_nop_conversion_p (type, TREE_TYPE (@3)) check. Instead
32129 require both type and TREE_TYPE (@1) to be integral types and either
32130 type having smaller or equal precision, or TREE_TYPE (@1) being
32131 unsigned type, or type being signed type. If TREE_TYPE (@1)
32132 doesn't have wrapping overflow, perform the subtraction of one in
32135 2021-02-14 Jan Hubicka <hubicka@ucw.cz>
32136 Richard Biener <rguether@suse.de>
32139 * ipa-reference.c (ipa_init): Only conditinally initialize
32140 reference_vars_to_consider.
32141 (propagate): Conditionally deninitialize reference_vars_to_consider.
32142 (ipa_reference_write_optimization_summary): Sanity check that
32143 reference_vars_to_consider is not allocated.
32145 2021-02-13 Levy Hsu <admin@levyhsu.com>
32148 * config/riscv/riscv-shorten-memrefs.c (pass_shorten_memrefs): Add
32149 extend parameter to get_si_mem_base_reg declaration.
32150 (get_si_mem_base_reg): Add extend parameter. Set it.
32151 (analyze): Pass extend arg to get_si_mem_base_reg.
32152 (transform): Likewise. Use it when rewriting mems.
32153 * config/riscv/riscv.c (riscv_legitimize_move): Check for subword
32154 loads and emit sign/zero extending load followed by subreg move.
32156 2021-02-13 Jim Wilson <jimw@sifive.com>
32159 * config/riscv/riscv.c (riscv_compressed_lw_address_p): Drop early
32160 exit when !reload_completed. Only perform check for compressed reg
32161 if reload_completed.
32162 (riscv_rtx_costs): In MEM case, when optimizing for size and
32163 shorten memrefs, if not compressible, then increase cost.
32165 2021-02-13 Jakub Jelinek <jakub@redhat.com>
32167 PR rtl-optimization/98439
32168 * recog.c (pass_split_before_regstack::gate): Enable even when
32169 pass_split_before_sched2 is enabled if -fselective-scheduling2 is
32172 2021-02-13 Jakub Jelinek <jakub@redhat.com>
32175 * config/i386/mmx.md (*mmx_pshufd_1): Add a combine splitter for
32176 swap of V2SImode elements in memory into DImode memory rotate by 32.
32178 2021-02-12 Martin Sebor <msebor@redhat.com>
32180 * tree-pretty-print.c (print_generic_expr_to_str): Update comment.
32182 2021-02-12 Richard Sandiford <richard.sandiford@arm.com>
32184 * rtl-ssa/accesses.cc (function_info::make_use_available): Use
32185 m_temp_obstack rather than m_obstack to allocate the temporary use.
32187 2021-02-12 Richard Sandiford <richard.sandiford@arm.com>
32189 * df-problems.c (df_lr_bb_local_compute): Treat partial definitions
32190 as read-modify operations.
32192 2021-02-12 Richard Biener <rguenther@suse.de>
32194 PR middle-end/38474
32195 * ipa-fnsummary.c (unmodified_parm_1): Only walk when
32196 fbi->aa_walk_budget is bigger than zero. Update
32197 fbi->aa_walk_budget.
32198 (param_change_prob): Likewise.
32199 * ipa-prop.c (detect_type_change_from_memory_writes):
32200 Properly account walk_aliased_vdefs.
32201 (parm_preserved_before_stmt_p): Canonicalize updates.
32202 (parm_ref_data_preserved_p): Likewise.
32203 (parm_ref_data_pass_through_p): Likewise.
32204 (determine_known_aggregate_parts): Account own alias queries.
32206 2021-02-12 Martin Liska <mliska@suse.cz>
32208 * opts-common.c (decode_cmdline_option): Release werror_arg.
32209 * opts.c (gen_producer_string): Release output of
32210 gen_command_line_string.
32212 2021-02-12 Richard Biener <rguenther@suse.de>
32214 PR tree-optimization/38474
32215 * params.opt (-param=max-store-chains-to-track=): New param.
32216 (-param=max-stores-to-track=): Likewise.
32217 * doc/invoke.texi (max-store-chains-to-track): Document.
32218 (max-stores-to-track): Likewise.
32219 * gimple-ssa-store-merging.c (pass_store_merging::m_n_chains):
32221 (pass_store_merging::m_n_stores): Likewise.
32222 (pass_store_merging::terminate_and_process_chain): Update
32223 m_n_stores and m_n_chains.
32224 (pass_store_merging::process_store): Likewise. Terminate
32225 oldest chains if the number of stores or chains get too large.
32226 (imm_store_chain_info::terminate_and_process_chain): Dump
32229 2021-02-11 Eric Botcazou <ebotcazou@adacore.com>
32231 * config/i386/winnt.c (i386_pe_seh_unwind_emit): When switching to
32232 the cold section, emit a nop before the directive if the previous
32233 active instruction can throw.
32235 2021-02-11 Peter Bergner <bergner@linux.ibm.com>
32238 * config/rs6000/predicates.md (mma_assemble_input_operand): Restrict
32239 memory addresses that are legal for quad word accesses.
32241 2021-02-11 Andrea Corallo <andrea.corallo@arm.com>
32244 * config/arm/thumb2.md (*doloop_end_internal): Generate
32245 alternative sequence to handle long range branches.
32247 2021-02-11 Joel Hutton <joel.hutton@arm.com>
32249 PR tree-optimization/98772
32250 * optabs-tree.c (supportable_half_widening_operation): New function
32251 to check for supportable V8QI->V8HI widening patterns.
32252 * optabs-tree.h (supportable_half_widening_operation): New function.
32253 * tree-vect-stmts.c (vect_create_half_widening_stmts): New function
32254 to create promotion stmts for V8QI->V8HI widening patterns.
32255 (vectorizable_conversion): Add case for V8QI->V8HI.
32257 2021-02-11 Richard Biener <rguenther@suse.de>
32259 * sparseset.h (SPARSESET_ELT_BITS): Remove.
32260 (SPARSESET_ELT_TYPE): Use unsigned int.
32261 * fwprop.c: Do not include sparseset.h.
32263 2021-02-10 Jakub Jelinek <jakub@redhat.com>
32266 * varasm.c (declare_weak): For -fsyntax-only, allow even
32267 TREE_ASM_WRITTEN function decls.
32269 2021-02-10 Jakub Jelinek <jakub@redhat.com>
32272 * config/i386/sse.md (fix<fixunssuffix>_truncv2sfv2di2,
32273 <insn>v8qiv8hi2, <insn>v8qiv8si2, <insn>v4qiv4si2, <insn>v4hiv4si2,
32274 <insn>v8qiv8di2, <insn>v4qiv4di2, <insn>v2qiv2di2, <insn>v4hiv4di2,
32275 <insn>v2hiv2di2, <insn>v2siv2di2): Force operands[1] into REG before
32276 calling simplify_gen_subreg on it.
32278 2021-02-10 Martin Liska <mliska@suse.cz>
32280 * config/nvptx/nvptx.c (nvptx_option_override): Use
32281 flag_patchable_function_entry instead of the removed
32282 function_entry_patch_area_size.
32284 2021-02-10 Martin Liska <mliska@suse.cz>
32286 PR tree-optimization/99002
32287 PR tree-optimization/99026
32288 * gimple-if-to-switch.cc (if_chain::is_beneficial): Fix memory
32289 leak when adjacent cases are merged.
32290 * tree-switch-conversion.c (switch_decision_tree::analyze_switch_statement): Use
32292 (make_pass_lower_switch): Remove trailing whitespace.
32293 * tree-switch-conversion.h (release_clusters): New.
32295 2021-02-10 Richard Biener <rguenther@suse.de>
32297 PR rtl-optimization/99054
32298 * cfgrtl.c (rtl-optimization/99054): Return an auto_vec.
32299 (fixup_partitions): Adjust.
32300 (rtl_verify_edges): Likewise.
32302 2021-02-10 Jakub Jelinek <jakub@redhat.com>
32304 PR middle-end/99007
32305 * gimplify.c (gimplify_scan_omp_clauses): For MEM_REF on reductions,
32306 temporarily disable gimplify_ctxp->into_ssa around gimplify_expr
32309 2021-02-10 Richard Biener <rguenther@suse.de>
32312 * ipa-pure-const.c (propagate_malloc): Use an auto_vec<>
32315 2021-02-10 Richard Biener <rguenther@suse.de>
32317 PR tree-optimization/99024
32318 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Only
32319 clear loop->aux if it is associated with the destroyed loop_vinfo.
32321 2021-02-10 Martin Liska <mliska@suse.cz>
32323 PR tree-optimization/99002
32324 * gimple-if-to-switch.cc (find_conditions): Fix memory leak
32327 2021-02-10 Martin Liska <mliska@suse.cz>
32330 * ipa-icf.c (sem_item::add_reference): Fix memory leak when
32331 a reference exists.
32333 2021-02-10 Jakub Jelinek <jakub@redhat.com>
32336 * dwarf2out.c (prune_unused_types_walk): Mark DW_TAG_variable DIEs
32337 at class scope for DWARF5+.
32339 2021-02-09 Eric Botcazou <ebotcazou@adacore.com>
32341 PR rtl-optimization/96015
32342 * reorg.c (skip_consecutive_labels): Minor comment tweaks.
32343 (relax_delay_slots): When deleting a jump to the next active
32344 instruction over a barrier, first delete the barrier if the
32345 jump is the only way to reach the target label.
32347 2021-02-09 Andre Vieira <andre.simoesdiasvieira@arm.com>
32349 * config/aarch64/aarch64-cost-tables.h: Add entries for vect.mul.
32350 * config/aarch64/aarch64.c (aarch64_rtx_mult_cost): Use vect.mul for
32351 vector multiplies and vect.alu for SSRA.
32352 * config/arm/aarch-common-protos.h (struct vector_cost_table): Define
32353 vect.mul cost field.
32354 * config/arm/aarch-cost-tables.h: Add entries for vect.mul.
32355 * config/arm/arm.c: Likewise.
32357 2021-02-09 Richard Biener <rguenther@suse.de>
32359 PR tree-optimization/98863
32360 * tree-ssa-sccvn.h (vn_avail::next_undo): Add.
32361 * tree-ssa-sccvn.c (last_pushed_avail): New global.
32362 (rpo_elim::eliminate_push_avail): Chain pushed avails.
32363 (unwind_state::avail_top): Add.
32364 (do_unwind): Rewrite unwinding of avail entries.
32365 (do_rpo_vn): Initialize last_pushed_avail and
32366 avail_top of the undo state.
32368 2021-02-09 Jakub Jelinek <jakub@redhat.com>
32370 PR middle-end/99004
32371 * calls.c (maybe_warn_rdwr_sizes): Change s0 and s1 type from
32372 const char * to char * and free those pointers after use.
32374 2021-02-09 Richard Biener <rguenther@suse.de>
32376 PR tree-optimization/99017
32377 * tree-vect-slp.c (vect_bb_vectorization_profitable_p): Allow
32378 zero vector cost entries.
32380 2021-02-08 Andre Vieira <andre.simoesdiasvieira@arm.com>
32382 PR middle-end/98974
32383 * tree-vect-stmts.c (vectorizable_condition): Remove shadow vec_num
32384 parameter in vectorizable_condition.
32386 2021-02-08 Richard Biener <rguenther@suse.de>
32389 * tree.c (walk_tree_1): Walk VECTOR_CST elements.
32391 2021-02-08 Martin Liska <mliska@suse.cz>
32394 * cfgexpand.c (pass_expand::execute): Parse per-function option
32395 flag_patchable_function_entry and use it.
32396 * common.opt: Remove function_entry_patch_area_size and
32397 function_entry_patch_area_start global variables.
32398 * opts.c (parse_and_check_patch_area): New function.
32399 (common_handle_option): Use it.
32400 * opts.h (parse_and_check_patch_area): New function.
32401 * toplev.c (process_options): Parse and use
32402 function_entry_patch_area_size.
32404 2021-02-08 Martin Sebor <msebor@redhat.com>
32406 * doc/extend.texi (attribute malloc): Correct typos.
32408 2021-02-05 Nathan Sidwell <nathan@acm.org>
32411 * gcc.c (driver::maybe_run_linker): Check for input file
32412 accessibility if not linking.
32414 2021-02-05 Richard Biener <rguenther@suse.de>
32416 PR tree-optimization/98855
32417 * tree-vectorizer.h (add_stmt_cost): New overload.
32418 * tree-vect-slp.c (li_cost_vec_cmp): New.
32419 (vect_bb_slp_scalar_cost): Cost individual loop regions
32420 separately. Account for the scalar instance root stmt.
32422 2021-02-05 Tom de Vries <tdevries@suse.de>
32425 * tree-switch-conversion.c (jump_table_cluster::emit): Add loc
32427 (bit_test_cluster::emit): Reuse location_t for newly created
32429 (switch_decision_tree::try_switch_expansion): Preserve
32431 * tree-switch-conversion.h: Change function signatures.
32433 2021-02-05 Jakub Jelinek <jakub@redhat.com>
32436 * config/i386/i386-options.c (m_NONE, m_ALL): Define.
32437 * config/i386/x86-tune.def (X86_TUNE_BRANCH_PREDICTION_HINTS,
32438 X86_TUNE_PROMOTE_QI_REGS): Use m_NONE instead of 0U.
32439 (X86_TUNE_QIMODE_MATH): Use m_ALL instead of ~0U.
32441 2021-02-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
32443 * config/aarch64/aarch64-simd-builtins.def (get_high): Define builtin.
32444 * config/aarch64/aarch64-simd.md (aarch64_get_high<mode>): Define.
32445 * config/aarch64/arm_neon.h (__GET_HIGH): Delete.
32446 (vget_high_f16): Reimplement using new builtin.
32447 (vget_high_f32): Likewise.
32448 (vget_high_f64): Likewise.
32449 (vget_high_p8): Likewise.
32450 (vget_high_p16): Likewise.
32451 (vget_high_p64): Likewise.
32452 (vget_high_s8): Likewise.
32453 (vget_high_s16): Likewise.
32454 (vget_high_s32): Likewise.
32455 (vget_high_s64): Likewise.
32456 (vget_high_u8): Likewise.
32457 (vget_high_u16): Likewise.
32458 (vget_high_u32): Likewise.
32459 (vget_high_u64): Likewise.
32461 2021-02-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
32463 * config/aarch64/aarch64-simd-builtins.def (get_low): Define builtin.
32464 * config/aarch64/aarch64-simd.md (aarch64_get_low<mode>): Define.
32465 * config/aarch64/arm_neon.h (__GET_LOW): Delete.
32466 (vget_low_f16): Reimplement using new builtin.
32467 (vget_low_f32): Likewise.
32468 (vget_low_f64): Likewise.
32469 (vget_low_p8): Likewise.
32470 (vget_low_p16): Likewise.
32471 (vget_low_p64): Likewise.
32472 (vget_low_s8): Likewise.
32473 (vget_low_s16): Likewise.
32474 (vget_low_s32): Likewise.
32475 (vget_low_s64): Likewise.
32476 (vget_low_u8): Likewise.
32477 (vget_low_u16): Likewise.
32478 (vget_low_u32): Likewise.
32479 (vget_low_u64): Likewise.
32481 2021-02-05 Kito Cheng <kito.cheng@sifive.com>
32483 * gcc.c (print_multilib_info): Check all required argument is provided
32486 2021-02-05 liuhongt <hongtao.liu@intel.com>
32489 * config/i386/i386-expand.c (ix86_expand_sse_cmp): Don't
32490 generate integer mask comparison for 128/256-bits vector when
32491 op_true/op_false is NULL_RTX or CONSTM1_RTX/CONST0_RTX. Also
32492 delete redundant !maskcmp condition.
32493 (ix86_expand_int_vec_cmp): Ditto but no redundant deletion
32495 (ix86_expand_sse_movcc): Delete definition of maskcmp, add the
32496 condition directly to if (maskcmp), add extra check for
32497 cmpmode, it should be MODE_INT.
32498 (ix86_expand_fp_vec_cmp): Pass NULL to ix86_expand_sse_cmp's
32499 parameters op_true/op_false.
32500 (ix86_use_mask_cmp_p): New.
32502 2021-02-05 liuhongt <hongtao.liu@intel.com>
32505 * config/i386/x86-tune.def (X86_TUNE_AVX256_UNALIGNED_LOAD_OPTIMAL):
32506 Remove m_GENERIC from ~list.
32507 (X86_TUNE_AVX256_UNALIGNED_STORE_OPTIMAL): Ditto.
32509 2021-02-04 David Malcolm <dmalcolm@redhat.com>
32512 * diagnostic-show-locus.c (compatible_locations_p): Require
32513 locations in the same macro map to be either both from the
32514 macro definition, or both from the macro arguments.
32516 2021-02-04 Jonathan Wright <jonathan.wright@arm.com>
32518 * config/aarch64/aarch64-simd-builtins.def: Add
32519 [su]mull_hi_lane[q] builtin generator macros.
32520 * config/aarch64/aarch64-simd.md
32521 (aarch64_<su>mull_hi_lane<mode>_insn): Define.
32522 (aarch64_<su>mull_hi_lane<mode>): Define.
32523 (aarch64_<su>mull_hi_laneq<mode>_insn): Define.
32524 (aarch64_<su>mull_hi_laneq<mode>): Define.
32525 * config/aarch64/arm_neon.h (vmull_high_lane_s16): Use RTL
32526 builtin instead of inline asm.
32527 (vmull_high_lane_s32): Likewise.
32528 (vmull_high_lane_u16): Likewise.
32529 (vmull_high_lane_u32): Likewise.
32530 (vmull_high_laneq_s16): Likewise.
32531 (vmull_high_laneq_s32): Likewise.
32532 (vmull_high_laneq_u16): Likewise.
32533 (vmull_high_laneq_u32): Liekwise.
32535 2021-02-04 Jonathan Wright <jonathan.wright@arm.com>
32537 * config/aarch64/aarch64-simd-builtins.def: Add [su]mull_hi_n
32538 builtin generator macros.
32539 * config/aarch64/aarch64-simd.md
32540 (aarch64_<su>mull_hi_n<mode>_insn): Define.
32541 (aarch64_<su>mull_hi_n<mode>): Define.
32542 * config/aarch64/arm_neon.h (vmull_high_n_s16): Use RTL builtin
32543 instead of inline asm.
32544 (vmull_high_n_s32): Likewise.
32545 (vmull_high_n_u16): Likewise.
32546 (vmull_high_n_u32): Likewise.
32548 2021-02-04 Richard Biener <rguenther@suse.de>
32550 PR tree-optimization/98855
32551 * tree-vect-loop.c (vectorizable_phi): Do not cost
32552 single-argument PHIs.
32553 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Likewise.
32554 * tree-vect-stmts.c (vectorizable_bswap): Also perform
32555 costing for SLP operation.
32557 2021-02-04 Martin Liska <mliska@suse.cz>
32559 * doc/extend.texi: Mention -mprefer-vector-width in target
32562 2021-02-03 Martin Sebor <msebor@redhat.com>
32564 PR tree-optimization/98937
32565 * tree-ssa-strlen.c (strlen_dom_walker::~strlen_dom_walker): Define.
32566 Flush pointer_query cache.
32568 2021-02-03 Aaron Sawdey <acsawdey@linux.ibm.com>
32570 * config/rs6000/genfusion.pl (gen_2logical): Add missing
32571 fixes based on patch review.
32572 * config/rs6000/fusion.md: Regenerate file.
32574 2021-02-03 Aaron Sawdey <acsawdey@linux.ibm.com>
32576 * config/rs6000/t-rs6000: Comment out auto generation of
32579 2021-02-03 Andrew Stubbs <ams@codesourcery.com>
32581 * config/gcn/gcn-opts.h (enum processor_type): Add PROCESSOR_GFX908.
32582 * config/gcn/gcn.c (gcn_omp_device_kind_arch_isa): Add gfx908.
32583 (output_file_start): Add gfx908.
32584 * config/gcn/gcn.opt (gpu_type): Add gfx908.
32585 * config/gcn/t-gcn-hsa (MULTILIB_OPTIONS): Add march=gfx908.
32586 (MULTILIB_DIRNAMES): Add gfx908.
32587 * config/gcn/mkoffload.c (EF_AMDGPU_MACH_AMDGCN_GFX908): New define.
32588 (main): Recognize gfx908.
32589 * config/gcn/t-omp-device: Add gfx908.
32591 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
32593 * config/aarch64/aarch64-simd-builtins.def: Add
32594 [su]mlsl_hi_lane[q] builtin macro generators.
32595 * config/aarch64/aarch64-simd.md
32596 (aarch64_<su>mlsl_hi_lane<mode>_insn): Define.
32597 (aarch64_<su>mlsl_hi_lane<mode>): Define.
32598 (aarch64_<su>mlsl_hi_laneq<mode>_insn): Define.
32599 (aarch64_<su>mlsl_hi_laneq<mode>): Define.
32600 * config/aarch64/arm_neon.h (vmlsl_high_lane_s16): Use RTL
32601 builtin instead of inline asm.
32602 (vmlsl_high_lane_s32): Likewise.
32603 (vmlsl_high_lane_u16): Likewise.
32604 (vmlsl_high_lane_u32): Likewise.
32605 (vmlsl_high_laneq_s16): Likewise.
32606 (vmlsl_high_laneq_s32): Likewise.
32607 (vmlsl_high_laneq_u16): Likewise.
32608 (vmlsl_high_laneq_u32): Likewise.
32609 (vmlal_high_laneq_u32): Likewise.
32611 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
32613 * config/aarch64/aarch64-simd-builtins.def: Add
32614 [su]mlal_hi_lane[q] builtin generator macros.
32615 * config/aarch64/aarch64-simd.md
32616 (aarch64_<su>mlal_hi_lane<mode>_insn): Define.
32617 (aarch64_<su>mlal_hi_lane<mode>): Define.
32618 (aarch64_<su>mlal_hi_laneq<mode>_insn): Define.
32619 (aarch64_<su>mlal_hi_laneq<mode>): Define.
32620 * config/aarch64/arm_neon.h (vmlal_high_lane_s16): Use RTL
32621 builtin instead of inline asm.
32622 (vmlal_high_lane_s32): Likewise.
32623 (vmlal_high_lane_u16): Likewise.
32624 (vmlal_high_lane_u32): Likewise.
32625 (vmlal_high_laneq_s16): Likewise.
32626 (vmlal_high_laneq_s32): Likewise.
32627 (vmlal_high_laneq_u16): Likewise.
32628 (vmlal_high_laneq_u32): Likewise.
32630 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
32632 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlsl_hi_n
32633 builtin generator macros.
32634 * config/aarch64/aarch64-simd.md (aarch64_<su>mlsl_hi_n<mode>_insn):
32636 (aarch64_<su>mlsl_hi_n<mode>): Define.
32637 * config/aarch64/arm_neon.h (vmlsl_high_n_s16): Use RTL builtin
32638 instead of inline asm.
32639 (vmlsl_high_n_s32): Likewise.
32640 (vmlsl_high_n_u16): Likewise.
32641 (vmlsl_high_n_u32): Likewise.
32643 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
32645 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlal_hi_n
32646 builtin generator macros.
32647 * config/aarch64/aarch64-simd.md (aarch64_<su>mlal_hi_n<mode>_insn):
32649 (aarch64_<su>mlal_hi_n<mode>): Define.
32650 * config/aarch64/arm_neon.h (vmlal_high_n_s16): Use RTL builtin
32651 instead of inline asm.
32652 (vmlal_high_n_s32): Likewise.
32653 (vmlal_high_n_u16): Likewise.
32654 (vmlal_high_n_u32): Likewise.
32656 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
32658 * config/aarch64/aarch64-simd-builtins.def: Add RTL builtin
32660 * config/aarch64/aarch64-simd.md (*aarch64_<su>mlal_hi<mode>):
32662 (aarch64_<su>mlal_hi<mode>_insn): This.
32663 (aarch64_<su>mlal_hi<mode>): Define.
32664 * config/aarch64/arm_neon.h (vmlal_high_s8): Use RTL builtin
32665 instead of inline asm.
32666 (vmlal_high_s16): Likewise.
32667 (vmlal_high_s32): Likewise.
32668 (vmlal_high_u8): Likewise.
32669 (vmlal_high_u16): Likewise.
32670 (vmlal_high_u32): Likewise.
32672 2021-02-03 Ilya Leoshkevich <iii@linux.ibm.com>
32674 * lra-spills.c (remove_pseudos): Call lra_update_insn_recog_data()
32675 after calling alter_subreg() on a (mem).
32677 2021-02-03 Martin Liska <mliska@suse.cz>
32680 * lto-streamer-out.c (produce_lto_section): Fill up missing
32682 * lto-streamer.h (struct lto_section): Add _padding field.
32684 2021-02-03 Richard Biener <rguenther@suse.de>
32686 * lto-streamer.c (lto_get_section_name): Free temporary
32688 * tree-loop-distribution.c
32689 (loop_distribution::merge_dep_scc_partitions): Free edge data.
32691 2021-02-03 Jakub Jelinek <jakub@redhat.com>
32693 PR middle-end/97487
32694 * ifcvt.c (noce_can_force_operand): New function.
32695 (noce_emit_move_insn): Use it.
32696 (noce_try_sign_mask): Likewise. Formatting fix.
32698 2021-02-03 Jakub Jelinek <jakub@redhat.com>
32700 PR middle-end/97971
32701 * lra-constraints.c (process_alt_operands): For inline asm, don't call
32702 fatal_insn, but instead return false.
32704 2021-02-03 Jakub Jelinek <jakub@redhat.com>
32706 PR tree-optimization/98287
32707 * config/i386/mmx.md (<insn><mode>3): For shifts don't enable expander
32710 2021-02-03 Tamar Christina <tamar.christina@arm.com>
32712 PR tree-optimization/98928
32713 * tree-vect-loop.c (vect_analyze_loop_2): Change
32714 STMT_VINFO_SLP_VECT_ONLY to STMT_VINFO_SLP_VECT_ONLY_PATTERN.
32715 * tree-vect-slp-patterns.c (complex_pattern::build): Likewise.
32716 * tree-vectorizer.h (STMT_VINFO_SLP_VECT_ONLY_PATTERN): New.
32717 (class _stmt_vec_info): Add slp_vect_pattern_only_p.
32719 2021-02-02 Richard Biener <rguenther@suse.de>
32721 * gimple-loop-interchange.cc (prepare_data_references):
32723 * gimple-loop-jam.c (tree_loop_unroll_and_jam): Likewise.
32724 * tree-ssa-loop-im.c (hoist_memory_references): Likewise.
32725 * tree-vect-stmts.c (vectorizable_condition): Do not
32727 (vectorizable_comparison): Likewise.
32729 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
32731 * config/aarch64/aarch64-simd-builtins.def (ursqrte): Define builtin.
32732 * config/aarch64/aarch64-simd.md (aarch64_ursqrte<mode>): New pattern.
32733 * config/aarch64/arm_neon.h (vrsqrte_u32): Reimplement using builtin.
32734 (vrsqrteq_u32): Likewise.
32736 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
32738 * config/aarch64/aarch64-simd-builtins.def (sqxtun2): Define builtin.
32739 * config/aarch64/aarch64-simd.md (aarch64_sqxtun2<mode>_le): Define.
32740 (aarch64_sqxtun2<mode>_be): Likewise.
32741 (aarch64_sqxtun2<mode>): Likewise.
32742 * config/aarch64/arm_neon.h (vqmovun_high_s16): Reimplement using builtin.
32743 (vqmovun_high_s32): Likewise.
32744 (vqmovun_high_s64): Likewise.
32745 * config/aarch64/iterators.md (UNSPEC_SQXTUN2): Define.
32747 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
32749 * config/aarch64/aarch64-simd-builtins.def (bfdot_lane, bfdot_laneq): Use
32751 (bfmlalb_lane, bfmlalt_lane, bfmlalb_lane_q, bfmlalt_lane_q): Use FP flags.
32753 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
32755 * config/aarch64/aarch64-simd-builtins.def (fcmla_lane0, fcmla_lane90,
32756 fcmla_lane180, fcmla_lane270, fcmlaq_lane0, fcmlaq_lane90, fcmlaq_lane180,
32757 fcmlaq_lane270, scvtf, ucvtf, fcvtzs, fcvtzu, scvtfsi, scvtfdi, ucvtfsi,
32758 ucvtfdi, fcvtzshf, fcvtzuhf, fmlal_lane_low, fmlsl_lane_low,
32759 fmlal_laneq_low, fmlsl_laneq_low, fmlalq_lane_low, fmlslq_lane_low,
32760 fmlalq_laneq_low, fmlslq_laneq_low, fmlal_lane_high, fmlsl_lane_high,
32761 fmlal_laneq_high, fmlsl_laneq_high, fmlalq_lane_high, fmlslq_lane_high,
32762 fmlalq_laneq_high, fmlslq_laneq_high): Use FP flags.
32764 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
32766 * config/aarch64/aarch64-builtins.c (FLAG_LOAD): Define.
32767 * config/aarch64/aarch64-simd-builtins.def (ld1x2, ld2, ld3, ld4, ld2r,
32768 ld3r, ld4r, ld1, ld1x3, ld1x4): Use LOAD flags.
32770 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
32772 * config/aarch64/aarch64-simd-builtins.def (combine, zip1, zip2,
32773 uzp1, uzp2, trn1, trn2, simd_bsl): Use AUTO_FP flags.
32775 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
32777 * config/aarch64/aarch64-simd-builtins.def (clrsb, clz, ctz, popcount,
32778 vec_smult_lane_, vec_smlal_lane_, vec_smult_laneq_, vec_smlal_laneq_,
32779 vec_umult_lane_, vec_umlal_lane_, vec_umult_laneq_, vec_umlal_laneq_,
32780 ashl, sshl, ushl, srshl, urshl, sdot_lane, udot_lane, sdot_laneq,
32781 udot_laneq, usdot_lane, usdot_laneq, sudot_lane, sudot_laneq, ashr,
32782 ashr_simd, lshr, lshr_simd, srshr_n, urshr_n, ssra_n, usra_n, srsra_n,
32783 ursra_n, sshll_n, ushll_n, sshll2_n, ushll2_n, ssri_n, usri_n, ssli_n,
32784 ssli_n, usli_n, bswap, rbit, simd_bsl, eor3q, rax1q, xarq, bcaxq): Use
32785 NONE builtin flags.
32787 2021-02-02 Jakub Jelinek <jakub@redhat.com>
32789 PR tree-optimization/98848
32790 * tree-vect-patterns.c (vect_recog_over_widening_pattern): Punt if
32791 STMT_VINFO_DEF_TYPE (last_stmt_info) is vect_reduction_def.
32793 2021-02-02 Kito Cheng <kito.cheng@sifive.com>
32796 * expr.c: Check mode before calling store_expr.
32798 2021-02-02 Christophe Lyon <christophe.lyon@linaro.org>
32800 * config/arm/iterators.md (supf): Remove VORNQ_S and VORNQ_U.
32802 * config/arm/mve.md (mve_vornq_s<mode>): New entry for vorn
32803 instruction using expression ior.
32804 (mve_vornq_u<mode>): New expander.
32805 (mve_vornq_f<mode>): Use ior code instead of unspec.
32806 * config/arm/unspecs.md (VORNQ_S, VORNQ_U, VORNQ_F): Remove.
32808 2021-02-02 Alexandre Oliva <oliva@adacore.com>
32810 * tree-nested.c (convert_nonlocal_reference_op): Move
32811 current_function_decl restore after re-gimplification.
32812 (convert_local_reference_op): Likewise.
32814 2021-02-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
32816 * config/aarch64/aarch64-simd-builtins.def (rshrn, rshrn2):
32818 * config/aarch64/aarch64-simd.md (aarch64_rshrn<mode>_insn_le):
32820 (aarch64_rshrn<mode>_insn_be): Likewise.
32821 (aarch64_rshrn<mode>): Likewise.
32822 (aarch64_rshrn2<mode>_insn_le): Likewise.
32823 (aarch64_rshrn2<mode>_insn_be): Likewise.
32824 (aarch64_rshrn2<mode>): Likewise.
32825 * config/aarch64/aarch64.md (unspec): Add UNSPEC_RSHRN.
32826 * config/aarch64/arm_neon.h (vrshrn_high_n_s16): Reimplement
32828 (vrshrn_high_n_s32): Likewise.
32829 (vrshrn_high_n_s64): Likewise.
32830 (vrshrn_high_n_u16): Likewise.
32831 (vrshrn_high_n_u32): Likewise.
32832 (vrshrn_high_n_u64): Likewise.
32833 (vrshrn_n_s16): Likewise.
32834 (vrshrn_n_s32): Likewise.
32835 (vrshrn_n_s64): Likewise.
32836 (vrshrn_n_u16): Likewise.
32837 (vrshrn_n_u32): Likewise.
32838 (vrshrn_n_u64): Likewise.
32840 2021-02-01 Sergei Trofimovich <siarheit@google.com>
32842 PR tree-optimization/98499
32843 * ipa-modref.c (analyze_ssa_name_flags): treat RVO
32844 conservatively and assume all possible side-effects.
32846 2021-02-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
32848 * config/aarch64/aarch64-simd-builtins.def (vec_unpacks_hi,
32849 vec_unpacku_hi_): Define builtins.
32850 * config/aarch64/arm_neon.h (vmovl_high_s8): Reimplement using
32852 (vmovl_high_s16): Likewise.
32853 (vmovl_high_s32): Likewise.
32854 (vmovl_high_u8): Likewise.
32855 (vmovl_high_u16): Likewise.
32856 (vmovl_high_u32): Likewise.
32858 2021-02-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
32860 * config/aarch64/aarch64-simd-builtins.def (sabdl, uabdl):
32862 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl<mode>): New
32864 * config/aarch64/aarch64.md (unspec): Define UNSPEC_SABDL,
32866 * config/aarch64/arm_neon.h (vabdl_s8): Reimplemet using
32868 (vabdl_s16): Likewise.
32869 (vabdl_s32): Likewise.
32870 (vabdl_u8): Likewise.
32871 (vabdl_u16): Likewise.
32872 (vabdl_u32): Likewise.
32873 * config/aarch64/iterators.md (ABDL): New int iterator.
32874 (sur): Handle UNSPEC_SABDL, UNSPEC_UABDL.
32876 2021-02-01 Martin Sebor <msebor@redhat.com>
32878 * tree.h (BLOCK_VARS): Add comment.
32879 (BLOCK_SUBBLOCKS): Same.
32880 (BLOCK_SUPERCONTEXT): Same.
32881 (BLOCK_ABSTRACT_ORIGIN): Same.
32882 (inlined_function_outer_scope_p): Same.
32884 2021-02-01 Martin Sebor <msebor@redhat.com>
32886 PR middle-end/97172
32887 * attribs.c (attr_access::free_lang_data): Define new function.
32888 * attribs.h (attr_access::free_lang_data): Declare new function.
32890 2021-02-01 Richard Biener <rguenther@suse.de>
32892 * vec.h (auto_vec::auto_vec): Add memory stat parameters
32894 * bitmap.h (auto_bitmap::auto_bitmap): Likewise.
32896 2021-02-01 Tamar Christina <tamar.christina@arm.com>
32898 * config/aarch64/aarch64-simd.md (aarch64_<su>mlal_n<mode>,
32899 aarch64_<su>mlsl<mode>, aarch64_<su>mlsl_n<mode>): Flip mult operands.
32901 2021-02-01 Richard Biener <rguenther@suse.de>
32903 PR rtl-optimization/98863
32904 * config/i386/i386-features.c (convert_scalars_to_vector):
32905 Set DF_RD_PRUNE_DEAD_DEFS.
32907 2021-01-31 Eric Botcazou <ebotcazou@adacore.com>
32909 * system.h (SIZE_MAX): Define if not already defined.
32911 2021-01-30 Aaron Sawdey <acsawdey@linux.ibm.com>
32913 * config/rs6000/genfusion.pl (gen_2logical): New function to
32914 generate patterns for logical-logical fusion.
32915 * config/rs6000/fusion.md: Regenerated patterns.
32916 * config/rs6000/rs6000-cpus.def: Add
32917 OPTION_MASK_P10_FUSION_2LOGICAL.
32918 * config/rs6000/rs6000.c (rs6000_option_override_internal):
32919 Enable logical-logical fusion for p10.
32920 * config/rs6000/rs6000.opt: Add -mpower10-fusion-2logical.
32922 2021-01-30 David Edelsohn <dje.gcc@gmail.com>
32924 * config/rs6000/rs6000.opt: Add periods to new AIX options.
32926 2021-01-30 David Edelsohn <dje.gcc@gmail.com>
32928 * config/rs6000/rs6000.opt (mabi=vec-extabi): New.
32929 (mabi=vec-default): New.
32930 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Define
32931 __EXTABI__ for AIX Vector extended ABI.
32932 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Print AIX Vector
32934 (conditional_register_usage): If AIX vec_extabi enabled, vs20-vs31
32936 * doc/invoke.texi (PowerPC mabi): Add AIX vec-extabi and vec-default.
32938 2021-01-30 Jakub Jelinek <jakub@redhat.com>
32940 * config/i386/i386-features.c (remove_partial_avx_dependency): Clear
32941 DF_DEFER_INSN_RESCAN after calling df_process_deferred_rescans.
32943 2021-01-29 Vladimir N. Makarov <vmakarov@redhat.com>
32946 * lra-constraints.c (in_class_p): Don't narrow class only for REG
32949 2021-01-29 Will Schmidt <will_schmidt@vnet.ibm.com>
32951 * config/rs6000/rs6000-call.c (rs6000_expand_binup_builtin): Add
32952 clauses for CODE_FOR_vsx_xvcvuxddp_scale and
32953 CODE_FOR_vsx_xvcvsxddp_scale to the parameter checking code.
32955 2021-01-29 Andrew MacLeod <amacleod@redhat.com>
32957 PR tree-optimization/98866
32958 * gimple-range-gori.h (gori_compute:set_range_invariant): New.
32959 * gimple-range-gori.cc (gori_map::set_range_invariant): New.
32960 (gori_map::m_maybe_invariant): Rename from all_outgoing.
32961 (gori_map::gori_map): Rename all_outgoing to m_maybe_invariant.
32962 (gori_map::is_export_p): Ditto.
32963 (gori_map::calculate_gori): Ditto.
32964 (gori_compute::set_range_invariant): New.
32965 * gimple-range.cc (gimple_ranger::range_of_stmt): Set range
32966 invariant for pointers evaluating to [1, +INF].
32968 2021-01-29 Richard Biener <rguenther@suse.de>
32970 PR rtl-optimization/98863
32971 * config/i386/i386-features.c (remove_partial_avx_dependency):
32972 Do not perform DF analysis.
32973 (pass_data_remove_partial_avx_dependency): Remove
32976 2021-01-29 Jonathan Wright <jonathan.wright@arm.com>
32978 * config/aarch64/aarch64-simd-builtins.def: Add [su]mull_n
32979 builtin generator macros.
32980 * config/aarch64/aarch64-simd.md (aarch64_<su>mull_n<mode>):
32982 * config/aarch64/arm_neon.h (vmull_n_s16): Use RTL builtin
32983 instead of inline asm.
32984 (vmull_n_s32): Likewise.
32985 (vmull_n_u16): Likewise.
32986 (vmull_n_u32): Likewise.
32988 2021-01-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
32990 * config/aarch64/aarch64-simd-builtins.def (sabdl2, uabdl2):
32992 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
32994 (aarch64_<sur>abdl2<mode>): ... This.
32995 (<sur>sadv16qi): Adjust use of above.
32996 * config/aarch64/arm_neon.h (vabdl_high_s8): Reimplement using
32998 (vabdl_high_s16): Likewise.
32999 (vabdl_high_s32): Likewise.
33000 (vabdl_high_u8): Likewise.
33001 (vabdl_high_u16): Likewise.
33002 (vabdl_high_u32): Likewise.
33004 2021-01-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
33006 * config/aarch64/aarch64-simd-builtins.def (sabal2): Define
33008 (uabal2): Likewise.
33009 * config/aarch64/aarch64-simd.md (aarch64_<sur>abal2<mode>): New
33011 * config/aarch64/aarch64.md (unspec): Add UNSPEC_SABAL2 and
33013 * config/aarch64/arm_neon.h (vabal_high_s8): Reimplement using
33015 (vabal_high_s16): Likewise.
33016 (vabal_high_s32): Likewise.
33017 (vabal_high_u8): Likewise.
33018 (vabal_high_u16): Likewise.
33019 (vabal_high_u32): Likewise.
33020 * config/aarch64/iterators.md (ABAL2): New mode iterator.
33021 (sur): Handle UNSPEC_SABAL2, UNSPEC_UABAL2.
33023 2021-01-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
33025 * config/aarch64/aarch64-simd-builtins.def (sabal): Define
33028 * config/aarch64/aarch64-simd.md (aarch64_<sur>abal<mode>_4):
33030 (aarch64_<sur>abal<mode>): ... This
33031 (<sur>sadv16qi): Adust use of the above.
33032 * config/aarch64/arm_neon.h (vabal_s8): Reimplement using
33034 (vabal_s16): Likewise.
33035 (vabal_s32): Likewise.
33036 (vabal_u8): Likewise.
33037 (vabal_u16): Likewise.
33038 (vabal_u32): Likewise.
33040 2021-01-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
33042 * config/aarch64/aarch64-simd-builtins.def (saddlv, uaddlv):
33044 * config/aarch64/aarch64-simd.md (aarch64_<su>addlv<mode>):
33046 * config/aarch64/arm_neon.h (vaddlv_s8): Reimplement using
33048 (vaddlv_s16): Likewise.
33049 (vaddlv_u8): Likewise.
33050 (vaddlv_u16): Likewise.
33051 (vaddlvq_s8): Likewise.
33052 (vaddlvq_s16): Likewise.
33053 (vaddlvq_s32): Likewise.
33054 (vaddlvq_u8): Likewise.
33055 (vaddlvq_u16): Likewise.
33056 (vaddlvq_u32): Likewise.
33057 (vaddlv_s32): Likewise.
33058 (vaddlv_u32): Likewise.
33059 * config/aarch64/iterators.md (VDQV_L): New mode iterator.
33060 (unspec): Add UNSPEC_SADDLV, UNSPEC_UADDLV.
33061 (Vwstype): New mode attribute.
33063 (VWIDE_S): Likewise.
33064 (USADDLV): New int iterator.
33065 (su): Handle UNSPEC_SADDLV, UNSPEC_UADDLV.
33067 2021-01-29 Jonathan Wright <jonathan.wright@arm.com>
33069 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlsl_lane[q]
33070 builtin generator macros.
33071 * config/aarch64/aarch64-simd.md (aarch64_vec_<su>mlsl_lane<Qlane>):
33073 * config/aarch64/arm_neon.h (vmlsl_lane_s16): Use RTL builtin
33074 instead of inline asm.
33075 (vmlsl_lane_s32): Likewise.
33076 (vmlsl_lane_u16): Likewise.
33077 (vmlsl_lane_u32): Likewise.
33078 (vmlsl_laneq_s16): Likewise.
33079 (vmlsl_laneq_s32): Likewise.
33080 (vmlsl_laneq_u16): Likewise.
33081 (vmlsl_laneq_u32): Likewise.
33083 2021-01-29 Richard Biener <rguenther@suse.de>
33085 * doc/invoke.texi (--param max-gcse-memory): Document unit
33087 * gcse.c (gcse_or_cprop_is_too_expensive): Adjust.
33088 * params.opt (--param max-gcse-memory): Adjust default and
33089 document unit of size.
33091 2021-01-29 Richard Biener <rguenther@suse.de>
33093 PR rtl-optimization/98863
33094 * gcse.c (gcse_or_cprop_is_too_expensive): Use unsigned
33095 HOST_WIDE_INT for the memory estimate.
33097 2021-01-29 Bin Cheng <bin.cheng@linux.alibaba.com>
33098 Richard Biener <rguenther@suse.de>
33100 PR tree-optimization/97627
33101 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
33102 Do not analyze fake edges.
33104 2021-01-29 Richard Biener <rguenther@suse.de>
33106 PR rtl-optimization/98144
33107 * df.h (df_mir_bb_info): Add con_visited member.
33108 * df-problems.c (df_mir_alloc): Initialize con_visited,
33109 do not fully populate IN and OUT.
33110 (df_mir_reset): Likewise.
33111 (df_mir_confluence_0): Set con_visited.
33112 (df_mir_confluence_n): Properly handle implicitely
33113 fully populated IN and OUT as designated by con_visited
33114 and update con_visited accordingly.
33116 2021-01-29 Jakub Jelinek <jakub@redhat.com>
33119 * config/arm/vec-common.md (mve_vshlq_<supf><mode>,
33120 vashl<mode>3, vashr<mode>3, vlshr<mode>3): Add
33121 && !TARGET_REALLY_IWMMXT to conditions.
33123 2021-01-29 Jakub Jelinek <jakub@redhat.com>
33126 * cfgbuild.c (find_bb_boundaries): Reset debug_insn when seeing
33129 2021-01-28 Marek Polacek <polacek@redhat.com>
33132 * stor-layout.c (finalize_type_size): If we reset TYPE_USER_ALIGN in
33133 the main variant, maybe reset it in its variants too.
33134 * tree.c (check_base_type): Return true only if TYPE_USER_ALIGN match.
33135 (check_aligned_type): Check if TYPE_USER_ALIGN match.
33137 2021-01-28 Christophe Lyon <christophe.lyon@linaro.org>
33140 * config/arm/arm.c (arm_rtx_costs_internal): Adjust cost of vector
33141 of constant zero for comparisons.
33143 2021-01-28 Michael Meissner <meissner@linux.ibm.com>
33145 * config/rs6000/rs6000.c (rs6000_mangle_decl_assembler_name): Add
33146 support for mapping built-in function names for long double
33147 built-in functions if long double is IEEE 128-bit.
33149 2021-01-28 Jonathan Wright <jonathan.wright@arm.com>
33151 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlsl_n
33152 builtin generator macros.
33153 * config/aarch64/aarch64-simd.md (aarch64_<su>mlsl_n<mode>):
33155 * config/aarch64/arm_neon.h (vmlsl_n_s16): Use RTL builtin
33156 instead of inline asm.
33157 (vmlsl_n_s32): Likewise.
33158 (vmlsl_n_u16): Likewise.
33159 (vmlsl_n_u32): Likewise.
33161 2021-01-28 Jonathan Wright <jonathan.wright@arm.com>
33163 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlal_n
33164 builtin generator macros.
33165 * config/aarch64/aarch64-simd.md (aarch64_<su>mlal_n<mode>):
33167 * config/aarch64/arm_neon.h (vmlal_n_s16): Use RTL builtin
33168 instead of inline asm.
33169 (vmlal_n_s32): Likewise.
33170 (vmlal_n_u16): Likewise.
33171 (vmlal_n_u32): Likewise.
33173 2021-01-28 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
33175 * config/aarch64/aarch64-simd-builtins.def (shrn2): Define
33177 * config/aarch64/aarch64-simd.md (aarch64_shrn2<mode>_insn_le):
33179 (aarch64_shrn2<mode>_insn_be): Likewise.
33180 (aarch64_shrn2<mode>): Likewise.
33181 * config/aarch64/arm_neon.h (vshrn_high_n_s16): Reimlplement
33183 (vshrn_high_n_s32): Likewise.
33184 (vshrn_high_n_s64): Likewise.
33185 (vshrn_high_n_u16): Likewise.
33186 (vshrn_high_n_u32): Likewise.
33187 (vshrn_high_n_u64): Likewise.
33189 2021-01-28 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
33191 * config/aarch64/aarch64-simd-builtins.def (shrn): Define
33193 * config/aarch64/aarch64-simd.md (aarch64_shrn<mode>_insn_le):
33195 (aarch64_shrn<mode>_insn_be): Likewise.
33196 (aarch64_shrn<mode>): Likewise.
33197 * config/aarch64/arm_neon.h (vshrn_n_s16): Reimplement using
33199 (vshrn_n_s32): Likewise.
33200 (vshrn_n_s64): Likewise.
33201 (vshrn_n_u16): Likewise.
33202 (vshrn_n_u32): Likewise.
33203 (vshrn_n_u64): Likewise.
33204 * config/aarch64/iterators.md (vn_mode): New mode attribute.
33206 2021-01-28 Richard Biener <rguenther@suse.de>
33208 PR rtl-optimization/80960
33209 * dse.c (check_mem_read_rtx): Call get_addr on the
33212 2021-01-28 Xionghu Luo <luoxhu@linux.ibm.com>
33213 David Edelsohn <dje.gcc@gmail.com>
33216 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
33217 Don't generate VIEW_CONVERT_EXPR for fcode ALTIVEC_BUILTIN_VEC_INSERT
33219 * config/rs6000/rs6000-protos.h (rs6000_expand_vector_set_var):
33221 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Remove the
33222 wrapper call rs6000_expand_vector_set_var for cleanup. Call
33223 rs6000_expand_vector_set_var_p9 and rs6000_expand_vector_set_var_p8
33225 (rs6000_expand_vector_set_var): Delete.
33226 (rs6000_expand_vector_set_var_p9): Make static.
33227 (rs6000_expand_vector_set_var_p8): Make static.
33229 2021-01-28 Xing GUO <higuoxing@gmail.com>
33231 * common/config/riscv/riscv-common.c
33232 (riscv_subset_list::parsing_subset_version): Fix -march option parsing
33233 when `p` extension exists.
33235 2021-01-27 Vladimir N. Makarov <vmakarov@redhat.com>
33237 PR rtl-optimization/97684
33238 * ira.c (ira): Call ira_set_pseudo_classes before
33239 update_equiv_regs when it is necessary.
33241 2021-01-27 Jakub Jelinek <jakub@redhat.com>
33244 * config/aarch64/aarch64.md (*aarch64_bfxilsi_uxtw): Use
33245 %w0, %w1 and %2 instead of %0, %1 and %2.
33247 2021-01-27 Aaron Sawdey <acsawdey@linux.ibm.com>
33249 * config/rs6000/genfusion.pl: New script to generate
33250 define_insn_and_split patterns so combine can arrange fused
33251 instructions next to each other.
33252 * config/rs6000/fusion.md: New file, generated fused instruction
33253 patterns for combine.
33254 * config/rs6000/predicates.md (const_m1_to_1_operand): New predicate.
33255 (non_update_memory_operand): New predicate.
33256 * config/rs6000/rs6000-cpus.def: Add OPTION_MASK_P10_FUSION and
33257 OPTION_MASK_P10_FUSION_LD_CMPI to ISA_3_1_MASKS_SERVER and
33259 * config/rs6000/rs6000-protos.h (address_is_non_pfx_d_or_x): Add
33261 * config/rs6000/rs6000.c (rs6000_option_override_internal):
33262 Automatically set OPTION_MASK_P10_FUSION and
33263 OPTION_MASK_P10_FUSION_LD_CMPI if target is power10.
33264 (rs600_opt_masks): Allow -mpower10-fusion
33265 in function attributes.
33266 (address_is_non_pfx_d_or_x): New function.
33267 * config/rs6000/rs6000.h: Add MASK_P10_FUSION.
33268 * config/rs6000/rs6000.md: Include fusion.md.
33269 * config/rs6000/rs6000.opt: Add -mpower10-fusion
33270 and -mpower10-fusion-ld-cmpi.
33271 * config/rs6000/t-rs6000: Add dependencies involving fusion.md.
33273 2021-01-27 Jonathan Wright <jonathan.wright@arm.com>
33275 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlal
33276 builtin generator macros.
33277 * config/aarch64/aarch64-simd.md (*aarch64_<su>mlal<mode>):
33279 (aarch64_<su>mlal<mode>): This.
33280 * config/aarch64/arm_neon.h (vmlal_s8): Use RTL builtin
33281 instead of inline asm.
33282 (vmlal_s16): Likewise.
33283 (vmlal_s32): Likewise.
33284 (vmlal_u8): Likewise.
33285 (vmlal_u16): Likewise.
33286 (vmlal_u32): Likewise.
33288 2021-01-27 Richard Biener <rguenther@suse.de>
33290 PR tree-optimization/98854
33291 * tree-vect-slp.c (vect_build_slp_tree_2): Also build
33292 PHIs from scalars when the number of CTORs matches the
33293 number of children.
33295 2021-01-27 Jonathan Wright <jonathan.wright@arm.com>
33297 * config/aarch64/aarch64-simd-builtins.def: Add mls_n builtin
33299 * config/aarch64/aarch64-simd.md (*aarch64_mls_elt_merge<mode>):
33301 (aarch64_mls_n<mode>): This.
33302 * config/aarch64/arm_neon.h (vmls_n_s16): Use RTL builtin
33304 (vmls_n_s32): Likewise.
33305 (vmls_n_u16): Likewise.
33306 (vmls_n_u32): Likewise.
33307 (vmlsq_n_s16): Likewise.
33308 (vmlsq_n_s32): Likewise.
33309 (vmlsq_n_u16): Likewise.
33310 (vmlsq_n_u32): Likewise.
33312 2021-01-27 Jonathan Wright <jonathan.wright@arm.com>
33314 * config/aarch64/aarch64-simd-builtins.def: Add mls builtin
33316 * config/aarch64/arm_neon.h (vmls_s8): Use RTL builtin rather
33318 (vmls_s16): Likewise.
33319 (vmls_s32): Likewise.
33320 (vmls_u8): Likewise.
33321 (vmls_u16): Likewise.
33322 (vmls_u32): Likewise.
33323 (vmlsq_s8): Likewise.
33324 (vmlsq_s16): Likewise.
33325 (vmlsq_s32): Likewise.
33326 (vmlsq_u8): Likewise.
33327 (vmlsq_u16): Likewise.
33328 (vmlsq_u32): Likewise.
33330 2021-01-27 Jonathan Wright <jonathan.wright@arm.com>
33332 * config/aarch64/aarch64-simd-builtins.def: Add mla_n builtin
33334 * config/aarch64/aarch64-simd.md (*aarch64_mla_elt_merge<mode>):
33336 (aarch64_mla_n<mode>): This.
33337 * config/aarch64/arm_neon.h (vmla_n_s16): Use RTL builtin
33339 (vmla_n_s32): Likewise.
33340 (vmla_n_u16): Likewise.
33341 (vmla_n_u32): Likewise.
33342 (vmlaq_n_s16): Likewise.
33343 (vmlaq_n_s32): Likewise.
33344 (vmlaq_n_u16): Likewise.
33345 (vmlaq_n_u32): Likewise.
33347 2021-01-27 liuhongt <hongtao.liu@intel.com>
33350 * config/i386/sse.md (sse2_gt<mode>3): Drop !TARGET_XOP in condition.
33351 (*sse2_eq<mode>3): Ditto.
33353 2021-01-27 Jakub Jelinek <jakub@redhat.com>
33355 * tree-pass.h (PROP_trees): Rename to ...
33356 (PROP_gimple): ... this.
33357 * cfgexpand.c (pass_data_expand): Replace PROP_trees with PROP_gimple.
33358 * passes.c (execute_function_dump, execute_function_todo,
33359 execute_one_ipa_transform_pass, execute_one_pass): Likewise.
33360 * varpool.c (ctor_for_folding): Likewise.
33362 2021-01-27 Jakub Jelinek <jakub@redhat.com>
33364 PR tree-optimization/97260
33365 * varpool.c: Include tree-pass.h.
33366 (ctor_for_folding): In GENERIC return DECL_INITIAL for TREE_READONLY
33367 non-TREE_SIDE_EFFECTS automatic variables.
33369 2021-01-26 Paul Fee <paul.f.fee@gmail.com>
33371 * doc/cpp.texi (__cplusplus): Document value for -std=c++23
33373 * doc/invoke.texi: Document -std=c++23 and -std=gnu++23.
33374 * dwarf2out.c (highest_c_language): Recognise C++20 and C++23.
33375 (gen_compile_unit_die): Recognise C++23.
33377 2021-01-26 Jakub Jelinek <jakub@redhat.com>
33380 * dwarf2asm.c (dw2_assemble_integer): Cast DWARF2_ADDR_SIZE to int
33383 2021-01-26 Jakub Jelinek <jakub@redhat.com>
33386 * config/aarch64/aarch64.c (aarch64_mask_and_shift_for_ubfiz_p):
33387 Use UINTVAL (shft_amnt) and UINTVAL (mask) instead of INTVAL (shft_amnt)
33388 and INTVAL (mask). Add && INTVAL (mask) > 0 condition.
33390 2021-01-26 Richard Biener <rguenther@suse.de>
33392 * gimple-pretty-print.c (dump_binary_rhs): Handle
33393 VEC_WIDEN_{PLUS,MINUS}_{LO,HI}_EXPR.
33395 2021-01-26 Richard Biener <rguenther@suse.de>
33397 PR middle-end/98726
33398 * tree.h (vector_cst_int_elt): Remove.
33399 * tree.c (vector_cst_int_elt): Use poly_wide_int for computations,
33402 2021-01-26 Andrew Stubbs <ams@codesourcery.com>
33404 * config/gcn/gcn.c (gcn_expand_reduc_scalar): Use move instructions
33405 for V64DFmode min/max reductions.
33407 2021-01-26 Jakub Jelinek <jakub@redhat.com>
33409 * dwarf2asm.c (dw2_assemble_integer): Handle size twice as large
33410 as DWARF2_ADDR_SIZE if x is not a scalar int by emitting it as
33411 two halves, one with x and the other with const0_rtx, ordered
33412 depending on endianity.
33414 2021-01-26 Alexandre Oliva <oliva@adacore.com>
33416 * gimplify.c (gimplify_decl_expr): Skip asan marking calls for
33417 temporaries not seen in binding block, and not about to be
33418 added as gimple variables.
33420 2021-01-25 Martin Sebor <msebor@redhat.com>
33423 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Adjust warning text.
33425 2021-01-25 Martin Liska <mliska@suse.cz>
33427 * value-prof.c (get_nth_most_common_value): Use %s instead
33430 2021-01-25 Jakub Jelinek <jakub@redhat.com>
33433 * configure.ac (HAVE_AS_GDWARF_5_DEBUG_FLAG): Only define if
33434 readelf -wi is able to read the emitted .debug_info back.
33435 * configure: Regenerated.
33437 2021-01-25 Martin Liska <mliska@suse.cz>
33439 PR gcov-profile/98739
33440 * common.opt: Add missing sign symbol.
33441 * value-prof.c (get_nth_most_common_value): Restore handling
33442 of PROFILE_REPRODUCIBILITY_PARALLEL_RUNS and
33443 PROFILE_REPRODUCIBILITY_MULTITHREADED.
33445 2021-01-25 Richard Biener <rguenther@suse.de>
33447 PR middle-end/98807
33448 * tree.c (vector_element_bits): Always use precision of
33449 the element type for boolean vectors.
33451 2021-01-25 Sebastian Huber <sebastian.huber@embedded-brains.de>
33453 * config/rtems.h (STARTFILE_SPEC): Remove qnolinkcmds.
33454 (ENDFILE_SPEC): Evaluate qnolinkcmds.
33456 2021-01-25 Sebastian Huber <sebastian.huber@embedded-brains.de>
33458 * config/rtems.h (STARTFILE_SPEC): Remove nostdlib and
33459 nostartfiles handling since this is already done by
33460 LINK_COMMAND_SPEC. Evaluate qnolinkcmds.
33461 (ENDFILE_SPEC): Remove nostdlib and nostartfiles handling since this
33462 is already done by LINK_COMMAND_SPEC.
33463 (LIB_SPECS): Remove nostdlib and nodefaultlibs handling since
33464 this is already done by LINK_COMMAND_SPEC. Remove qnolinkcmds
33467 2021-01-25 Jakub Jelinek <jakub@redhat.com>
33470 * fold-const-call.c (host_size_t_cst_p): Renamed to ...
33471 (size_t_cst_p): ... this. Check and store unsigned HOST_WIDE_INT
33472 value rather than host size_t.
33473 (fold_const_call): Change type of s2 from size_t to
33474 unsigned HOST_WIDE_INT. Use size_t_cst_p instead of
33475 host_size_t_cst_p. For strncmp calls, pass MIN (s2, SIZE_MAX)
33476 instead of s2 as last argument.
33478 2021-01-25 Tamar Christina <tamar.christina@arm.com>
33480 * config/arm/iterators.md (rotsplit1, rotsplit2, conj_op, fcmac1,
33481 VCMLA_OP, VCMUL_OP): New.
33482 * config/arm/mve.md (mve_vcmlaq<mve_rot><mode>): Support vec_dup 0.
33483 * config/arm/neon.md (cmul<conj_op><mode>3): New.
33484 * config/arm/unspecs.md (UNSPEC_VCMLA_CONJ, UNSPEC_VCMLA180_CONJ,
33485 UNSPEC_VCMUL_CONJ): New.
33486 * config/arm/vec-common.md (cmul<conj_op><mode>3, arm_vcmla<rot><mode>,
33487 cml<fcmac1><conj_op><mode>4): New.
33489 2021-01-23 Jakub Jelinek <jakub@redhat.com>
33492 * config/rs6000/mmintrin.h (__m64): Add __may_alias__ attribute.
33494 2021-01-22 Jonathan Wright <jonathan.wright@arm.com>
33496 * config/aarch64/aarch64-simd-builtins.def: Add mla builtin
33498 * config/aarch64/arm_neon.h (vmla_s8): Use RTL builtin rather
33500 (vmla_s16): Likewise.
33501 (vmla_s32): Likewise.
33502 (vmla_u8): Likewise.
33503 (vmla_u16): Likewise.
33504 (vmla_u32): Likewise.
33505 (vmlaq_s8): Likewise.
33506 (vmlaq_s16): Likewise.
33507 (vmlaq_s32): Likewise.
33508 (vmlaq_u8): Likewise.
33509 (vmlaq_u16): Likewise.
33510 (vmlaq_u32): Likewise.
33512 2021-01-22 David Malcolm <dmalcolm@redhat.com>
33514 * doc/invoke.texi (GCC_EXTRA_DIAGNOSTIC_OUTPUT): Add @findex
33517 2021-01-22 Jakub Jelinek <jakub@redhat.com>
33520 * dwarf2out.c (output_file_names): For -gdwarf-5, if there are no
33521 filenames to emit, still emit the required 0 index directory and
33522 filename entries that match DW_AT_comp_dir and DW_AT_name of the
33525 2021-01-22 Marek Polacek <polacek@redhat.com>
33528 * doc/invoke.texi: Update C++ ABI Version 15 description.
33530 2021-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
33532 PR tree-optimization/98766
33533 * tree-ssa-math-opts.c (convert_mult_to_fma): Use maybe_le when
33534 comparing against type size with param_avoid_fma_max_bits.
33536 2021-01-22 Richard Biener <rguenther@suse.de>
33538 PR middle-end/98793
33539 * tree.c (vector_element_bits): Key single-bit bool vector on
33540 integer mode rather than not vector mode.
33542 2021-01-22 Xionghu Luo <luoxhu@linux.ibm.com>
33545 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
33546 Generate ARRAY_REF(VIEW_CONVERT_EXPR) for P8 and later
33548 * config/rs6000/rs6000.c (rs6000_expand_vector_set_var): Update
33549 to call different path for P8 and P9.
33550 (rs6000_expand_vector_set_var_p9): New function.
33551 (rs6000_expand_vector_set_var_p8): New function.
33553 2021-01-22 Xionghu Luo <luoxhu@linux.ibm.com>
33557 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
33558 Ajdust variable index vec_insert from address dereference to
33559 ARRAY_REF(VIEW_CONVERT_EXPR) tree expression.
33560 * config/rs6000/rs6000-protos.h (rs6000_expand_vector_set_var):
33562 * config/rs6000/rs6000.c (rs6000_expand_vector_set_var): New function.
33564 2021-01-22 Martin Liska <mliska@suse.cz>
33566 PR gcov-profile/98739
33567 * profile.c (compute_value_histograms): Drop time profile for
33568 -fprofile-reproducible=multithreaded.
33570 2021-01-22 Nathan Sidwell <nathan@acm.org>
33572 * gcc.c (process_command): Don't check OPT_SPECIAL_input_file
33575 2021-01-22 Richard Biener <rguenther@suse.de>
33577 PR middle-end/98773
33578 * tree-data-ref.c (initalize_matrix_A): Revert previous
33579 change, retaining failing on HOST_WIDE_INT_MIN CHREC_RIGHT.
33581 2021-01-22 Jakub Jelinek <jakub@redhat.com>
33583 PR tree-optimization/90248
33584 * match.pd (X cmp 0.0 ? 1.0 : -1.0 -> copysign(1, +-X),
33585 X cmp 0.0 ? -1.0 : +1.0 -> copysign(1, -+X)): Remove
33587 (X * (X cmp 0.0 ? 1.0 : -1.0) -> +-abs(X),
33588 X * (X cmp 0.0 ? -1.0 : 1.0) -> +-abs(X)): New simplifications.
33590 2021-01-22 Jakub Jelinek <jakub@redhat.com>
33592 PR tree-optimization/98255
33593 * tree-dfa.c (get_ref_base_and_extent): For ARRAY_REFs, sign
33594 extend index - low_bound from sizetype's precision rather than index
33596 (get_addr_base_and_unit_offset_1): Likewise.
33597 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Likewise.
33598 * gimple-fold.c (fold_const_aggregate_ref_1): Likewise.
33600 2021-01-22 Richard Biener <rguenther@suse.de>
33602 PR tree-optimization/98786
33603 * tree-ssa-phiopt.c (factor_out_conditional_conversion): Avoid
33604 adding new uses of abnormals. Verify we deal with a conditional
33607 2021-01-22 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
33610 * optc-save-gen.awk: Add arm_fp16_format to checked_options.
33612 2021-01-22 liuhongt <hongtao.liu@intel.com>
33616 * config/i386/sse.md (VI_128_256): New mode iterator.
33617 (*avx_cmp<mode>3_1, *avx_cmp<mode>3_2, *avx_cmp<mode>3_3,
33618 *avx_cmp<mode>3_4, *avx2_eq<mode>3, *avx2_pcmp<mode>3_1,
33619 *avx2_pcmp<mode>3_2, *avx2_gt<mode>3): New
33620 define_insn_and_split to lower avx512 vector comparison to avx
33621 version when dest is vector.
33622 (*<avx512>_cmp<mode>3,*<avx512>_cmp<mode>3,*<avx512>_ucmp<mode>3):
33623 define_insn_and_split for negating the comparison result.
33624 * config/i386/predicates.md (float_vector_all_ones_operand):
33626 * config/i386/i386-expand.c (ix86_expand_sse_movcc): Use
33627 general NOT operator without UNSPEC_MASKOP.
33629 2021-01-21 Vladimir N. Makarov <vmakarov@redhat.com>
33631 PR rtl-optimization/98777
33632 * lra-int.h (lra_pmode_pseudo): New extern.
33633 * lra.c (lra_pmode_pseudo): New global.
33635 * lra-eliminations.c (eliminate_regs_in_insn): Use it.
33637 2021-01-21 Ilya Leoshkevich <iii@linux.ibm.com>
33639 * fwprop.c (fwprop_propagation::classify_result): Allow
33640 (subreg (mem)) simplifications.
33642 2021-01-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
33644 * config/aarch64/aarch64-simd.md (aarch64_sqdml<SBINQOPS:as>l<mode>):
33646 (aarch64_sqdmlal<mode>): ... This...
33647 (aarch64_sqdmlsl<mode>): ... And this.
33648 (aarch64_sqdml<SBINQOPS:as>l_lane<mode>): Split into...
33649 (aarch64_sqdmlal_lane<mode>): ... This...
33650 (aarch64_sqdmlsl_lane<mode>): ... And this.
33651 (aarch64_sqdml<SBINQOPS:as>l_laneq<mode>): Split into...
33652 (aarch64_sqdmlsl_laneq<mode>): ... This...
33653 (aarch64_sqdmlal_laneq<mode>): ... And this.
33654 (aarch64_sqdml<SBINQOPS:as>l_n<mode>): Split into...
33655 (aarch64_sqdmlsl_n<mode>): ... This...
33656 (aarch64_sqdmlal_n<mode>): ... And this.
33657 (aarch64_sqdml<SBINQOPS:as>l2<mode>_internal): Split into...
33658 (aarch64_sqdmlal2<mode>_internal): ... This...
33659 (aarch64_sqdmlsl2<mode>_internal): ... And this.
33661 2021-01-21 Christophe Lyon <christophe.lyon@linaro.org>
33663 * config/arm/arm_mve.h (__arm_vcmpneq_s8): Fix return type.
33665 2021-01-21 Andrea Corallo <andrea.corallo@arm.com>
33668 * doc/sourcebuild.texi (arm_thumb2_no_arm_v8_1_lob): Document.
33670 2021-01-21 liuhongt <hongtao.liu@intel.com>
33672 PR rtl-optimization/98694
33673 * regcprop.c (copy_value): If SRC had been assigned a mode
33674 narrower than the copy, we can't link DEST into the chain even
33675 they have same hard_regno_nregs(i.e. HImode/SImode in i386
33678 2021-01-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
33680 * config/aarch64/aarch64-simd.md (aarch64_get_lane<mode>):
33681 Convert to define_insn_and_split. Split into simple move when moving
33684 2021-01-20 Segher Boessenkool <segher@kernel.crashing.org>
33686 * config/rs6000/rs6000.c (rs6000_emit_le_vsx_store): Change assert.
33687 Adjust comment. Simplify code.
33689 2021-01-20 Jakub Jelinek <jakub@redhat.com>
33692 * dwarf2out.c (reset_indirect_string): Also reset indirect strings
33693 with DW_FORM_line_strp form.
33694 (prune_unused_types_update_strings): Don't add into debug_str_hash
33695 indirect strings with DW_FORM_line_strp form.
33696 (adjust_name_comp_dir): New function.
33697 (dwarf2out_finish): Call it on CU DIEs after resetting
33698 debug_line_str_hash.
33700 2021-01-20 Vladimir N. Makarov <vmakarov@redhat.com>
33702 PR rtl-optimization/98722
33703 * lra-eliminations.c (eliminate_regs_in_insn): Check that target
33704 has no 3-op add insn to transform insns containing two pluses.
33706 2021-01-20 Richard Biener <rguenther@suse.de>
33708 * hwint.h (add_hwi): New function.
33709 (mul_hwi): Likewise.
33710 * tree-data-ref.c (initialize_matrix_A): Properly translate
33711 tree constants and avoid HOST_WIDE_INT_MIN.
33712 (lambda_matrix_row_add): Avoid undefined integer overflow
33713 and return true on such overflow.
33714 (lambda_matrix_right_hermite): Handle overflow from
33715 lambda_matrix_row_add gracefully. Simplify previous fix.
33716 (analyze_subscript_affine_affine): Likewise.
33718 2021-01-20 Eugene Rozenfeld <erozen@microsoft.com>
33720 PR tree-optimization/96674
33721 * match.pd: New patterns: x < y || y == XXX_MIN --> x <= y - 1
33722 x >= y && y != XXX_MIN --> x > y - 1
33724 2021-01-20 Richard Sandiford <richard.sandiford@arm.com>
33726 PR tree-optimization/98535
33727 * tree-vect-slp.c (duplicate_and_interleave): Use quick_grow_cleared.
33728 If the high and low permutes are the same, remove the high permutes
33729 from the working set and only continue with the low ones.
33731 2021-01-20 Jakub Jelinek <jakub@redhat.com>
33733 PR tree-optimization/98721
33734 * builtins.c (access_ref::inform_access): Don't assume
33735 SSA_NAME_IDENTIFIER must be non-NULL. Print messages about
33736 object whenever allocfn is NULL, rather than only when DECL_P
33737 is true. Use %qE instead of %qD for that. Formatting fixes.
33739 2021-01-20 Richard Biener <rguenther@suse.de>
33741 PR tree-optimization/98758
33742 * tree-data-ref.c (int_divides_p): Use lambda_int arguments.
33743 (lambda_matrix_right_hermite): Avoid undefinedness with
33744 signed integer abs and multiplication.
33745 (analyze_subscript_affine_affine): Use lambda_int.
33747 2021-01-20 David Malcolm <dmalcolm@redhat.com>
33750 * dwarf2out.c (output_line_info): Rename static variable
33751 "generation", moving it out of the function to...
33752 (output_line_info_generation): New.
33753 (init_sections_and_labels): Likewise, renaming the variable to...
33754 (init_sections_and_labels_generation): New.
33755 (dwarf2out_c_finalize): Reset the new variables.
33757 2021-01-19 Martin Sebor <msebor@redhat.com>
33759 PR middle-end/98664
33760 * tree-ssa-live.c (remove_unused_scope_block_p): Keep scopes for
33761 all functions, even if they're not declared artificial or inline.
33762 * tree.c (tree_inlined_location): Use macro expansion location
33763 only if scope traversal fails to expose one.
33765 2021-01-19 Richard Sandiford <richard.sandiford@arm.com>
33767 PR rtl-optimization/92294
33768 * alias.c (compare_base_symbol_refs): Take an extra parameter
33769 and add the distance between two symbols to it. Enshrine in
33770 comments that -1 means "either 0 or 1, but we can't tell
33771 which at compile time".
33772 (memrefs_conflict_p): Update call accordingly.
33773 (rtx_equal_for_memref_p): Likewise. Take the distance between symbols
33776 2021-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
33778 * config/aarch64/aarch64-simd-builtins.def (sqshl, uqshl,
33779 sqrshl, uqrshl, sqadd, uqadd, sqsub, uqsub, suqadd, usqadd, sqmovn,
33780 uqmovn, sqxtn2, uqxtn2, sqabs, sqneg, sqdmlal, sqdmlsl, sqdmlal_lane,
33781 sqdmlsl_lane, sqdmlal_laneq, sqdmlsl_laneq, sqdmlal_n, sqdmlsl_n,
33782 sqdmlal2, sqdmlsl2, sqdmlal2_lane, sqdmlsl2_lane, sqdmlal2_laneq,
33783 sqdmlsl2_laneq, sqdmlal2_n, sqdmlsl2_n, sqdmull, sqdmull_lane,
33784 sqdmull_laneq, sqdmull_n, sqdmull2, sqdmull2_lane, sqdmull2_laneq,
33785 sqdmull2_n, sqdmulh, sqrdmulh, sqdmulh_lane, sqdmulh_laneq,
33786 sqrdmulh_lane, sqrdmulh_laneq, sqshrun_n, sqrshrun_n, sqshrn_n,
33787 uqshrn_n, sqrshrn_n, uqrshrn_n, sqshlu_n, sqshl_n, uqshl_n, sqrdmlah,
33788 sqrdmlsh, sqrdmlah_lane, sqrdmlsh_lane, sqrdmlah_laneq, sqrdmlsh_laneq,
33789 sqmovun): Use NONE flags.
33791 2021-01-19 Richard Biener <rguenther@suse.de>
33794 * ipa-modref.c (analyze_stmt): Only record a summary for a
33797 2021-01-19 Richard Biener <rguenther@suse.de>
33799 PR middle-end/98638
33800 * tree-ssanames.c (fini_ssanames): Zero SSA_NAME_DEF_STMT.
33802 2021-01-19 Daniel Hellstrom <daniel@gaisler.com>
33804 * config/sparc/rtemself.h (TARGET_OS_CPP_BUILTINS): Add
33805 built-in define __FIX_LEON3FT_TN0018.
33807 2021-01-19 Richard Biener <rguenther@suse.de>
33810 * tree-inline.c (tree_function_versioning): Set input_location
33811 to UNKNOWN_LOCATION throughout the function.
33813 2021-01-19 Tobias Burnus <tobias@codesourcery.com>
33816 * omp-low.c (lower_omp_target): Handle nonpointer is_device_ptr.
33818 2021-01-19 Martin Jambor <mjambor@suse.cz>
33821 * ipa-sra.c (ssa_name_only_returned_p): New parameter fun. Check
33822 whether non-call exceptions allow removal of a statement.
33823 (isra_analyze_call): Pass the appropriate function to
33824 ssa_name_only_returned_p.
33826 2021-01-19 Geng Qi <gengqi@linux.alibaba.com>
33828 * config/riscv/arch-canonicalize (longext_sort): New function for
33829 sorting 'multi-letter'.
33830 * config/riscv/multilib-generator: Adjusting the loop of 'alt' in
33831 'alts'. The 'arch' may not be the first of 'alts'.
33832 (_expand_combination): Add underline for the 'ext' without '*'.
33833 This is because, a single-letter extension can always be treated well
33834 with a '_' prefix, but it cannot be separated out if it is appended
33837 2021-01-18 Vladimir N. Makarov <vmakarov@redhat.com>
33840 * ira.c (ira): Skip abnormal critical edge splitting.
33842 2021-01-18 Jakub Jelinek <jakub@redhat.com>
33844 PR tree-optimization/98727
33845 * tree-ssa-math-opts.c (match_arith_overflow): Fix up computation of
33846 second .MUL_OVERFLOW operand for signed multiplication with overflow
33847 checking if the second operand of multiplication is not constant.
33849 2021-01-18 David Edelsohn <dje.gcc@gmail.com>
33851 * doc/invoke.texi (-gdwarf): TPF defaults to version 2 and AIX
33852 defaults to version 4.
33854 2021-01-18 David Malcolm <dmalcolm@redhat.com>
33856 * attribs.h (fndecl_dealloc_argno): New decl.
33857 * builtins.c (call_dealloc_argno): Split out second half of
33859 (fndecl_dealloc_argno): New.
33860 * doc/extend.texi (Common Function Attributes): Document the
33861 interaction between the analyzer and the malloc attribute.
33862 * doc/invoke.texi (Static Analyzer Options): Likewise.
33864 2021-01-17 David Edelsohn <dje.gcc@gmail.com>
33866 * config/rs6000/aix71.h (SUBTARGET_OVERRIDE_OPTIONS): Override
33867 dwarf_version to 4.
33868 * config/rs6000/aix72.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
33870 2021-01-17 Martin Jambor <mjambor@suse.cz>
33873 * cgraph.c (clone_of_p): Check also former_clone_of as we climb
33876 2021-01-17 Mark Wielaard <mark@klomp.org>
33878 * common.opt (gdwarf-): Init(5).
33879 * doc/invoke.texi (-gdwarf): Document default to 5.
33881 2021-01-16 Kwok Cheung Yeung <kcy@codesourcery.com>
33883 * builtin-types.def
33884 (BT_FN_VOID_OMPFN_PTR_OMPCPYFN_LONG_LONG_BOOL_UINT_PTR_INT): Rename
33886 (BT_FN_VOID_OMPFN_PTR_OMPCPYFN_LONG_LONG_BOOL_UINT_PTR_INT_PTR):
33887 ...this. Add extra argument.
33888 * gimplify.c (omp_default_clause): Ensure that event handle is
33889 firstprivate in a task region.
33890 (gimplify_scan_omp_clauses): Handle OMP_CLAUSE_DETACH.
33891 (gimplify_adjust_omp_clauses): Likewise.
33892 * omp-builtins.def (BUILT_IN_GOMP_TASK): Change function type to
33893 BT_FN_VOID_OMPFN_PTR_OMPCPYFN_LONG_LONG_BOOL_UINT_PTR_INT_PTR.
33894 * omp-expand.c (expand_task_call): Add GOMP_TASK_FLAG_DETACH to flags
33895 if detach clause specified. Add detach argument when generating
33897 * omp-low.c (scan_sharing_clauses): Setup data environment for detach
33899 (finish_taskreg_scan): Move field for variable containing the event
33900 handle to the front of the struct.
33901 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_DETACH. Fix
33903 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
33904 OMP_CLAUSE_DETACH clause.
33905 (convert_local_omp_clauses): Handle OMP_CLAUSE_DETACH clause.
33906 * tree-pretty-print.c (dump_omp_clause): Handle OMP_CLAUSE_DETACH.
33907 * tree.c (omp_clause_num_ops): Add entry for OMP_CLAUSE_DETACH.
33909 (omp_clause_code_name): Add entry for OMP_CLAUSE_DETACH. Fix
33911 (walk_tree_1): Handle OMP_CLAUSE_DETACH.
33913 2021-01-16 Sebastian Huber <sebastian.huber@embedded-brains.de>
33915 * config/nios2/t-rtems: Reset all MULTILIB_* variables. Shorten
33916 multilib directory names. Use MULTILIB_REQUIRED instead of
33917 MULTILIB_EXCEPTIONS. Add -mhw-mul -mhw-mulx -mhw-div
33918 -mcustom-fpu-cfg=fph2 multilib.
33920 2021-01-16 Sebastian Huber <sebastian.huber@embedded-brains.de>
33922 * config/nios2/nios2.c (NIOS2_FPU_CONFIG_NUM): Adjust value.
33923 (nios2_init_fpu_configs): Provide register values for new
33924 -mcustom-fpu-cfg=fph2 option variant.
33925 * doc/invoke.texi (-mcustom-fpu-cfg=fph2): Document new option
33928 2021-01-16 Sebastian Huber <sebastian.huber@embedded-brains.de>
33930 * config/nios2/nios2.c (nios2_custom_check_insns): Remove
33931 custom instruction warnings.
33933 2021-01-16 Jakub Jelinek <jakub@redhat.com>
33935 PR tree-optimization/96669
33936 * match.pd ((CST << x) & 1 -> x == 0): New simplification.
33938 2021-01-16 Jakub Jelinek <jakub@redhat.com>
33940 PR tree-optimization/96271
33941 * passes.def: Pass false argument to first two pass_cd_dce
33942 instances and true to last instance. Add comment that
33943 last instance rewrites no longer addressed locals.
33944 * tree-ssa-dce.c (pass_cd_dce): Add update_address_taken_p member and
33946 (pass_cd_dce::set_pass_param): New method.
33947 (pass_cd_dce::execute): Return TODO_update_address_taken from
33948 last cd_dce instance.
33950 2021-01-15 Carl Love <cel@us.ibm.com>
33952 * config/rs6000/altivec.h (vec_mulh, vec_div, vec_dive, vec_mod):
33954 * config/rs6000/altivec.md (VIlong): Move define to file vsx.md.
33955 * config/rs6000/rs6000-builtin.def (DIVES_V4SI, DIVES_V2DI,
33956 DIVEU_V4SI, DIVEU_V2DI, DIVS_V4SI, DIVS_V2DI, DIVU_V4SI,
33957 DIVU_V2DI, MODS_V2DI, MODS_V4SI, MODU_V2DI, MODU_V4SI,
33958 MULHS_V2DI, MULHS_V4SI, MULHU_V2DI, MULHU_V4SI, MULLD_V2DI):
33959 Add builtin define.
33960 (MULH, DIVE, MOD): Add new BU_P10_OVERLOAD_2 definitions.
33961 * config/rs6000/rs6000-call.c (VSX_BUILTIN_VEC_DIV,
33962 VSX_BUILTIN_VEC_DIVE, P10_BUILTIN_VEC_MOD, P10_BUILTIN_VEC_MULH):
33963 New overloaded definitions.
33964 (builtin_function_type) [P10V_BUILTIN_DIVEU_V4SI,
33965 P10V_BUILTIN_DIVEU_V2DI, P10V_BUILTIN_DIVU_V4SI,
33966 P10V_BUILTIN_DIVU_V2DI, P10V_BUILTIN_MODU_V2DI,
33967 P10V_BUILTIN_MODU_V4SI, P10V_BUILTIN_MULHU_V2DI,
33968 P10V_BUILTIN_MULHU_V4SI]: Add case
33969 statement for builtins.
33970 * config/rs6000/rs6000.md (bits): Add new attribute sizes V4SI, V2DI.
33971 * config/rs6000/vsx.md (VIlong): Moved from config/rs6000/altivec.md.
33972 (UNSPEC_VDIVES, UNSPEC_VDIVEU): New unspec definitions.
33973 (vsx_mul_v2di): Add if TARGET_POWER10 statement.
33974 (vsx_udiv_v2di): Add if TARGET_POWER10 statement.
33975 (dives_<mode>, diveu_<mode>, div<mode>3, uvdiv<mode>3,
33976 mods_<mode>, modu_<mode>, mulhs_<mode>, mulhu_<mode>, mulv2di3):
33977 Add define_insn, mode is VIlong.
33978 * doc/extend.texi (vec_mulh, vec_mul, vec_div, vec_dive, vec_mod):
33979 Add builtin descriptions.
33981 2021-01-15 Eric Botcazou <ebotcazou@adacore.com>
33983 * final.c (final_start_function_1): Reset force_source_line.
33985 2021-01-15 Jakub Jelinek <jakub@redhat.com>
33987 PR tree-optimization/96669
33988 * match.pd (((1 << A) & 1) != 0 -> A == 0,
33989 ((1 << A) & 1) == 0 -> A != 0): Generalize for 1s replaced by
33990 possibly different power of two constants and to right shift too.
33992 2021-01-15 Jakub Jelinek <jakub@redhat.com>
33994 PR tree-optimization/96681
33995 * match.pd ((x < 0) ^ (y < 0) to (x ^ y) < 0): New simplification.
33996 ((x >= 0) ^ (y >= 0) to (x ^ y) < 0): Likewise.
33997 ((x < 0) ^ (y >= 0) to (x ^ y) >= 0): Likewise.
33998 ((x >= 0) ^ (y < 0) to (x ^ y) >= 0): Likewise.
34000 2021-01-15 Alexandre Oliva <oliva@adacore.com>
34002 * opts.c (gen_command_line_string): Exclude -dumpbase-ext.
34004 2021-01-15 Tamar Christina <tamar.christina@arm.com>
34006 * config/aarch64/aarch64-simd.md (cml<fcmac1><conj_op><mode>4,
34007 cmul<conj_op><mode>3): New.
34008 * config/aarch64/iterators.md (UNSPEC_FCMUL,
34009 UNSPEC_FCMUL180, UNSPEC_FCMLA_CONJ, UNSPEC_FCMLA180_CONJ,
34010 UNSPEC_CMLA_CONJ, UNSPEC_CMLA180_CONJ, UNSPEC_CMUL, UNSPEC_CMUL180,
34011 FCMLA_OP, FCMUL_OP, conj_op, rotsplit1, rotsplit2, fcmac1, sve_rot1,
34012 sve_rot2, SVE2_INT_CMLA_OP, SVE2_INT_CMUL_OP, SVE2_INT_CADD_OP): New.
34013 (rot): Add UNSPEC_FCMUL, UNSPEC_FCMUL180.
34014 (rot_op): Renamed to conj_op.
34015 * config/aarch64/aarch64-sve.md (cml<fcmac1><conj_op><mode>4,
34016 cmul<conj_op><mode>3): New.
34017 * config/aarch64/aarch64-sve2.md (cml<fcmac1><conj_op><mode>4,
34018 cmul<conj_op><mode>3): New.
34020 2021-01-15 David Malcolm <dmalcolm@redhat.com>
34024 (selftest::test_print_parseable_fixits_bytes_vs_display_columns):
34025 Escape the tempfile name when constructing the expected output.
34027 2021-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
34029 * config/aarch64/aarch64-simd.md (*aarch64_<su>mlsl_hi<mode>):
34031 (aarch64_<su>mlsl_hi<mode>): ... This.
34032 (aarch64_<su>mlsl_hi<mode>): Define.
34033 (*aarch64_<su>mlsl<mode): Rename to...
34034 (aarch64_<su>mlsl<mode): ... This.
34035 * config/aarch64/aarch64-simd-builtins.def (smlsl, umlsl,
34036 smlsl_hi, umlsl_hi): Define builtins.
34037 * config/aarch64/arm_neon.h (vmlsl_high_s8, vmlsl_high_s16,
34038 vmlsl_high_s32, vmlsl_high_u8, vmlsl_high_u16, vmlsl_high_u32,
34039 vmlsl_s8, vmlsl_s16, vmlsl_s32, vmlsl_u8,
34040 vmlsl_u16, vmlsl_u32): Reimplement with builtins.
34042 2021-01-15 Uroš Bizjak <ubizjak@gmail.com>
34044 * config/i386/i386-c.c (ix86_target_macros):
34045 Use cpp_define_formatted for __SIZEOF_FLOAT80__ definition.
34047 2021-01-15 Richard Sandiford <richard.sandiford@arm.com>
34050 * config.gcc (aarch64*-*-*): Add aarch64-cc-fusion.o to extra_objs.
34051 * Makefile.in (RTL_SSA_H): New variable.
34052 * config/aarch64/t-aarch64 (aarch64-cc-fusion.o): New rule.
34053 * config/aarch64/aarch64-protos.h (make_pass_cc_fusion): Declare.
34054 * config/aarch64/aarch64-passes.def: Add pass_cc_fusion after
34056 * config/aarch64/aarch64-cc-fusion.cc: New file.
34058 2021-01-15 Richard Sandiford <richard.sandiford@arm.com>
34060 * recog.h (insn_change_watermark::~insn_change_watermark): Avoid
34061 calling cancel_changes for changes that no longer exist.
34063 2021-01-15 Richard Sandiford <richard.sandiford@arm.com>
34065 * rtl-ssa/functions.h (function_info::ref_defs): Rename to...
34066 (function_info::reg_defs): ...this.
34067 * rtl-ssa/member-fns.inl (function_info::ref_defs): Rename to...
34068 (function_info::reg_defs): ...this.
34070 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
34073 * config/arm/arm_neon.h (vceqz_p64, vceqq_p64, vceqzq_p64): New.
34075 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
34078 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
34081 * config/arm/arm_neon.h (vceqz_p64, vceqq_p64, vceqzq_p64): New.
34083 2021-01-15 Richard Biener <rguenther@suse.de>
34085 PR tree-optimization/96376
34086 * tree-vect-stmts.c (get_load_store_type): Disregard alignment
34087 for VMAT_INVARIANT.
34089 2021-01-15 Martin Liska <mliska@suse.cz>
34091 * doc/install.texi: Document that some tests need pytest module.
34092 * doc/sourcebuild.texi: Likewise.
34094 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
34097 * config/arm/arm_neon.h (vceqz_p64, vceqq_p64, vceqzq_p64): New.
34099 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
34101 * config/arm/mve.md (mve_vshrq_n_s<mode>_imm): New entry.
34102 (mve_vshrq_n_u<mode>_imm): Likewise.
34103 * config/arm/neon.md (vashr<mode>3, vlshr<mode>3): Move to ...
34104 * config/arm/vec-common.md: ... here.
34106 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
34108 * config/arm/mve.md (mve_vshlq_<supf><mode>): Move to
34110 * config/arm/neon.md (vashl<mode>3): Delete.
34111 * config/arm/vec-common.md (mve_vshlq_<supf><mode>): New.
34112 (vasl<mode>3): New expander.
34114 2021-01-15 Richard Biener <rguenther@suse.de>
34116 PR tree-optimization/98685
34117 * tree-vect-slp.c (vect_schedule_slp_node): Refactor handling
34118 of vector extern defs.
34120 2021-01-14 David Malcolm <dmalcolm@redhat.com>
34123 * diagnostic.c (diagnostic_kind_text): Break out this array
34125 (diagnostic_build_prefix): ...here.
34126 (fancy_abort): Detect when diagnostic_initialize has not yet been
34127 called and fall back to a minimal implementation of printing the
34128 ICE, rather than segfaulting in internal_error.
34130 2021-01-14 David Malcolm <dmalcolm@redhat.com>
34132 * diagnostic.c (diagnostic_initialize): Eliminate
34133 parseable_fixits_p in favor of initializing extra_output_kind from
34134 GCC_EXTRA_DIAGNOSTIC_OUTPUT.
34135 (convert_column_unit): New function, split out from...
34136 (diagnostic_converted_column): ...this.
34137 (print_parseable_fixits): Add "column_unit" and "tabstop" params.
34138 Use them to call convert_column_unit on the column values.
34139 (diagnostic_report_diagnostic): Eliminate conditional on
34140 parseable_fixits_p in favor of a switch statement on
34141 extra_output_kind, passing the appropriate values to the new
34142 params of print_parseable_fixits.
34143 (selftest::test_print_parseable_fixits_none): Update for new
34144 params of print_parseable_fixits.
34145 (selftest::test_print_parseable_fixits_insert): Likewise.
34146 (selftest::test_print_parseable_fixits_remove): Likewise.
34147 (selftest::test_print_parseable_fixits_replace): Likewise.
34148 (selftest::test_print_parseable_fixits_bytes_vs_display_columns):
34150 (selftest::diagnostic_c_tests): Call it.
34151 * diagnostic.h (enum diagnostics_extra_output_kind): New.
34152 (diagnostic_context::parseable_fixits_p): Delete field in favor
34154 (diagnostic_context::extra_output_kind): ...this new field.
34155 * doc/invoke.texi (Environment Variables): Add
34156 GCC_EXTRA_DIAGNOSTIC_OUTPUT.
34157 * opts.c (common_handle_option): Update handling of
34158 OPT_fdiagnostics_parseable_fixits for change to diagnostic_context
34161 2021-01-14 Tamar Christina <tamar.christina@arm.com>
34163 * tree-vect-slp-patterns.c (class complex_operations_pattern,
34164 complex_operations_pattern::matches,
34165 complex_operations_pattern::recognize,
34166 complex_operations_pattern::build): New.
34167 (slp_patterns): Use it.
34169 2021-01-14 Tamar Christina <tamar.christina@arm.com>
34171 * internal-fn.def (COMPLEX_FMS, COMPLEX_FMS_CONJ): New.
34172 * optabs.def (cmls_optab, cmls_conj_optab): New.
34173 * doc/md.texi: Document them.
34174 * tree-vect-slp-patterns.c (class complex_fms_pattern,
34175 complex_fms_pattern::matches, complex_fms_pattern::recognize,
34176 complex_fms_pattern::build): New.
34178 2021-01-14 Tamar Christina <tamar.christina@arm.com>
34180 * internal-fn.def (COMPLEX_FMA, COMPLEX_FMA_CONJ): New.
34181 * optabs.def (cmla_optab, cmla_conj_optab): New.
34182 * doc/md.texi: Document them.
34183 * tree-vect-slp-patterns.c (vect_match_call_p,
34184 class complex_fma_pattern, vect_slp_reset_pattern,
34185 complex_fma_pattern::matches, complex_fma_pattern::recognize,
34186 complex_fma_pattern::build): New.
34188 2021-01-14 Tamar Christina <tamar.christina@arm.com>
34190 * internal-fn.def (COMPLEX_MUL, COMPLEX_MUL_CONJ): New.
34191 * optabs.def (cmul_optab, cmul_conj_optab): New.
34192 * doc/md.texi: Document them.
34193 * tree-vect-slp-patterns.c (vect_match_call_complex_mla,
34194 vect_normalize_conj_loc, is_eq_or_top, vect_validate_multiplication,
34195 vect_build_combine_node, class complex_mul_pattern,
34196 complex_mul_pattern::matches, complex_mul_pattern::recognize,
34197 complex_mul_pattern::build): New.
34199 2021-01-14 Tamar Christina <tamar.christina@arm.com>
34201 * tree-vect-slp.c (optimize_load_redistribution_1): New.
34202 (optimize_load_redistribution, vect_is_slp_load_node): New.
34203 (vect_match_slp_patterns): Use it.
34205 2021-01-14 Tamar Christina <tamar.christina@arm.com>
34207 * tree-vect-slp-patterns.c (complex_add_pattern::build):
34210 2021-01-14 Thomas Schwinge <thomas@codesourcery.com>
34212 * config/gcn/mkoffload.c (main): Create an offload image only in
34213 64-bit configurations.
34215 2021-01-14 H.J. Lu <hjl.tools@gmail.com>
34218 * config/i386/i386-options.c (ix86_option_override_internal):
34219 Issue an error for -fcf-protection with CF_BRANCH when compiling
34220 for 32-bit non-TARGET_CMOV targets.
34222 2021-01-14 Uroš Bizjak <ubizjak@gmail.com>
34225 * config/i386/i386-options.c (ix86_valid_target_attribute_inner_p):
34226 Remove declaration and initialization of shadow variable "ret".
34227 (ix86_option_override_internal): Remove delcaration of
34228 shadow variable "i". Redeclare shadowed variable to unsigned.
34229 * common/config/i386/i386-common.c (pta_size): Redeclare to unsigned.
34230 * config/i386/i386-builtins.c (get_builtin_code_for_version):
34231 Update for redeclaration.
34232 * config/i386/i386.h (pta_size): Ditto.
34234 2021-01-14 Richard Biener <rguenther@suse.de>
34236 PR tree-optimization/98674
34237 * tree-data-ref.c (base_supports_access_fn_components_p): New.
34238 (initialize_data_dependence_relation): For two bases without
34239 possible access fns resort to type size equality when determining
34240 shape compatibility.
34242 2021-01-14 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
34245 * config/arm/arm_neon.h: Replace calls to __builtin_vcge* by
34246 <=, >= operators in vcle and vcge intrinsics respectively.
34247 * config/arm/arm_neon_builtins.def: Remove entry for
34250 2021-01-14 Uroš Bizjak <ubizjak@gmail.com>
34253 * config/i386/i386-options.c (ix86_function_specific_save):
34254 Remove redundant assignment to opts->x_ix86_branch_cost.
34255 * config/i386/i386.c (ix86_prefetch_sse):
34256 Rename from x86_prefetch_sse. Update all uses.
34257 * config/i386/i386.h: Update for rename.
34258 * config/i386/i386-options.h: Ditto.
34260 2021-01-14 Jakub Jelinek <jakub@redhat.com>
34263 * config/i386/sse.md (*sse4_1_zero_extendv8qiv8hi2_3,
34264 *sse4_1_zero_extendv4hiv4si2_3, *sse4_1_zero_extendv2siv2di2_3):
34265 Use Bm instead of m for non-avx. Add isa attribute.
34267 2021-01-14 Jakub Jelinek <jakub@redhat.com>
34269 PR tree-optimization/96688
34270 * match.pd (~(X >> Y) -> ~X >> Y): New simplification if
34271 ~X can be simplified.
34273 2021-01-14 Richard Sandiford <richard.sandiford@arm.com>
34275 * tree-vect-stmts.c (vect_model_load_cost): Account for unused
34276 IFN_LOAD_LANES results.
34278 2021-01-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
34280 * config/aarch64/aarch64-simd.md (aarch64_<su>xtl<mode>):
34282 (aarch64_xtn<mode>): Likewise.
34283 * config/aarch64/aarch64-simd-builtins.def (sxtl, uxtl, xtn):
34286 * config/aarch64/arm_neon.h (vmovl_s8): Reimplement using
34288 (vmovl_s16): Likewise.
34289 (vmovl_s32): Likewise.
34290 (vmovl_u8): Likewise.
34291 (vmovl_u16): Likewise.
34292 (vmovl_u32): Likewise.
34293 (vmovn_s16): Likewise.
34294 (vmovn_s32): Likewise.
34295 (vmovn_s64): Likewise.
34296 (vmovn_u16): Likewise.
34297 (vmovn_u32): Likewise.
34298 (vmovn_u64): Likewise.
34300 2021-01-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
34302 * config/aarch64/aarch64-simd.md (aarch64_<su>qxtn2<mode>_le):
34304 (aarch64_<su>qxtn2<mode>_be): Likewise.
34305 (aarch64_<su>qxtn2<mode>): Likewise.
34306 * config/aarch64/aarch64-simd-builtins.def (sqxtn2, uqxtn2):
34308 * config/aarch64/iterators.md (SAT_TRUNC): Define code_iterator.
34309 (su): Handle ss_truncate and us_truncate.
34310 * config/aarch64/arm_neon.h (vqmovn_high_s16): Reimplement using
34312 (vqmovn_high_s32): Likewise.
34313 (vqmovn_high_s64): Likewise.
34314 (vqmovn_high_u16): Likewise.
34315 (vqmovn_high_u32): Likewise.
34316 (vqmovn_high_u64): Likewise.
34318 2021-01-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
34320 * config/aarch64/aarch64-simd.md (aarch64_xtn2<mode>_le):
34322 (aarch64_xtn2<mode>_be): Likewise.
34323 (aarch64_xtn2<mode>): Likewise.
34324 * config/aarch64/aarch64-simd-builtins.def (xtn2): Define
34326 * config/aarch64/arm_neon.h (vmovn_high_s16): Reimplement using
34328 (vmovn_high_s32): Likewise.
34329 (vmovn_high_s64): Likewise.
34330 (vmovn_high_u16): Likewise.
34331 (vmovn_high_u32): Likewise.
34332 (vmovn_high_u64): Likewise.
34334 2021-01-13 Stafford Horne <shorne@gmail.com>
34336 * config/or1k/or1k.h (ASM_PREFERRED_EH_DATA_FORMAT): New macro.
34338 2021-01-13 Stafford Horne <shorne@gmail.com>
34340 * config/or1k/linux.h (TARGET_ASM_FILE_END): Define macro.
34342 2021-01-13 Stafford Horne <shorne@gmail.com>
34344 * config/or1k/or1k.h (TARGET_CPU_CPP_BUILTINS): Add builtin
34345 define for __or1k_hard_float__.
34347 2021-01-13 Stafford Horne <shorne@gmail.com>
34349 * config/or1k/or1k.h (NO_PROFILE_COUNTERS): Define as 1.
34350 (PROFILE_HOOK): Define to call _mcount.
34351 (FUNCTION_PROFILER): Change from abort to no-op.
34353 2021-01-13 Jakub Jelinek <jakub@redhat.com>
34355 PR tree-optimization/96691
34356 * match.pd ((~X | C) ^ D -> (X | C) ^ (~D ^ C),
34357 (~X & C) ^ D -> (X & C) ^ (D ^ C)): New simplifications if
34358 (~D ^ C) or (D ^ C) can be simplified.
34360 2021-01-13 Richard Biener <rguenther@suse.de>
34362 PR tree-optimization/92645
34363 * match.pd (BIT_FIELD_REF to conversion): Delay canonicalization
34364 until after vector lowering.
34366 2021-01-13 Richard Sandiford <richard.sandiford@arm.com>
34368 * config/aarch64/aarch64-sve.md (fnma<mode>4): Extend from SVE_FULL_I
34370 (@aarch64_pred_fnma<mode>, cond_fnma<mode>, *cond_fnma<mode>_2)
34371 (*cond_fnma<mode>_4, *cond_fnma<mode>_any): Likewise.
34373 2021-01-13 Richard Sandiford <richard.sandiford@arm.com>
34375 * config/aarch64/aarch64-sve.md (fma<mode>4): Extend from SVE_FULL_I
34377 (@aarch64_pred_fma<mode>, cond_fma<mode>, *cond_fma<mode>_2)
34378 (*cond_fma<mode>_4, *cond_fma<mode>_any): Likewise.
34380 2021-01-13 Richard Biener <rguenther@suse.de>
34382 PR tree-optimization/92645
34383 * tree-vect-slp.c (vect_build_slp_tree_1): Relax supported
34384 BIT_FIELD_REF argument.
34385 (vect_build_slp_tree_2): Record the desired vector type
34386 on the external vector def.
34387 (vectorizable_slp_permutation): Handle required punning
34388 of existing vector defs.
34390 2021-01-13 Richard Sandiford <richard.sandiford@arm.com>
34392 * rtl-ssa/accesses.h (def_lookup): Fix order of comparison results.
34394 2021-01-13 Richard Sandiford <richard.sandiford@arm.com>
34396 * config/sh/sh.md (movsf_ie): Remove operands[2] test.
34398 2021-01-13 Samuel Thibault <samuel.thibault@ens-lyon.org>
34400 * config.gcc [$target == *-*-gnu*]: Enable
34401 'default_gnu_indirect_function'.
34403 2021-01-13 Jakub Jelinek <jakub@redhat.com>
34406 * optabs.c (expand_vec_perm_const): Don't force v0 and v1 into
34407 registers before calling targetm.vectorize.vec_perm_const, only after
34409 * config/i386/i386-expand.c (ix86_vectorize_vec_perm_const): Handle
34410 two argument permutation when one operand is zero vector and only
34411 after that force operands into registers.
34412 * config/i386/sse.md (*avx2_zero_extendv16qiv16hi2_1): New
34413 define_insn_and_split pattern.
34414 (*avx512bw_zero_extendv32qiv32hi2_1): Likewise.
34415 (*avx512f_zero_extendv16hiv16si2_1): Likewise.
34416 (*avx2_zero_extendv8hiv8si2_1): Likewise.
34417 (*avx512f_zero_extendv8siv8di2_1): Likewise.
34418 (*avx2_zero_extendv4siv4di2_1): Likewise.
34419 * config/mips/mips.c (mips_vectorize_vec_perm_const): Force operands
34421 * config/arm/arm.c (arm_vectorize_vec_perm_const): Likewise.
34422 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Likewise.
34423 * config/ia64/ia64.c (ia64_vectorize_vec_perm_const): Likewise.
34424 * config/aarch64/aarch64.c (aarch64_vectorize_vec_perm_const): Likewise.
34425 * config/rs6000/rs6000.c (rs6000_vectorize_vec_perm_const): Likewise.
34426 * config/gcn/gcn.c (gcn_vectorize_vec_perm_const): Likewise. Use std::swap.
34428 2021-01-13 Martin Liska <mliska@suse.cz>
34430 PR tree-optimization/98455
34431 * gimple-if-to-switch.cc (condition_info::record_phi_mapping):
34432 Record also virtual PHIs.
34433 (pass_if_to_switch::execute): Return TODO_cleanup_cfg only
34436 2021-01-13 Jonathan Wakely <jwakely@redhat.com>
34438 * doc/invoke.texi (C++ Modules): Fix typos.
34440 2021-01-13 Richard Biener <rguenther@suse.de>
34442 PR tree-optimization/98640
34443 * tree-ssa-sccvn.c (visit_nary_op): Do not try to
34444 handle plus or minus from a truncated operand to be
34447 2021-01-13 Jakub Jelinek <jakub@redhat.com>
34450 * config/i386/i386.md (*btr<mode>_1, *btr<mode>_2): New
34451 define_insn_and_split patterns.
34452 (splitter after *btr<mode>_2): New splitter.
34454 2021-01-13 Martin Liska <mliska@suse.cz>
34457 * cgraphunit.c (analyze_functions): Remove dead code.
34459 2021-01-13 Qian Jianhua <qianjh@cn.fujitsu.com>
34461 * config/aarch64/aarch64-cost-tables.h (a64fx_extra_costs): New.
34462 * config/aarch64/aarch64.c (a64fx_addrcost_table): New.
34463 (a64fx_regmove_cost, a64fx_vector_cost): New.
34464 (a64fx_tunings): Use the new added cost tables.
34466 2021-01-13 Jakub Jelinek <jakub@redhat.com>
34469 * config/i386/predicates.md (pmovzx_parallel): New predicate.
34470 * config/i386/sse.md (*sse4_1_zero_extendv8qiv8hi2_3): New
34471 define_insn_and_split pattern.
34472 (*sse4_1_zero_extendv4hiv4si2_3): Likewise.
34473 (*sse4_1_zero_extendv2siv2di2_3): Likewise.
34475 2021-01-13 Julian Brown <julian@codesourcery.com>
34477 * config/gcn/gcn.c (gcn_conditional_register_usage): Remove dead code
34478 to fix v0 register.
34480 2021-01-13 Julian Brown <julian@codesourcery.com>
34482 * config/gcn/gcn.c (gcn_md_reorg): Fix case where EXEC reg is live
34485 2021-01-13 Julian Brown <julian@codesourcery.com>
34487 * config/gcn/gcn-valu.md (recip<mode>2<exec>, recip<mode>2): Use unspec
34488 for reciprocal-approximation instructions.
34489 (div<mode>3): Use fused multiply-accumulate operations for reciprocal
34490 refinement and division result.
34491 * config/gcn/gcn.md (UNSPEC_RCP): New unspec constant.
34493 2021-01-13 Julian Brown <julian@codesourcery.com>
34495 * config/gcn/gcn-valu.md (subdf): Rename to...
34498 2021-01-12 Martin Liska <mliska@suse.cz>
34500 * gcov.c (source_info::debug): Fix printf format for 32-bit hosts.
34502 2021-01-12 Andrea Corallo <andrea.corallo@arm.com>
34504 * function-abi.h: Fix typo.
34506 2021-01-12 Christophe Lyon <christophe.lyon@linaro.org>
34510 * config/arm/arm.h (ARM_HAVE_NEON_V8QI_LDST): New macro.
34511 (ARM_HAVE_NEON_V16QI_LDST, ARM_HAVE_NEON_V4HI_LDST): Likewise.
34512 (ARM_HAVE_NEON_V8HI_LDST, ARM_HAVE_NEON_V2SI_LDST): Likewise.
34513 (ARM_HAVE_NEON_V4SI_LDST, ARM_HAVE_NEON_V4HF_LDST): Likewise.
34514 (ARM_HAVE_NEON_V8HF_LDST, ARM_HAVE_NEON_V4BF_LDST): Likewise.
34515 (ARM_HAVE_NEON_V8BF_LDST, ARM_HAVE_NEON_V2SF_LDST): Likewise.
34516 (ARM_HAVE_NEON_V4SF_LDST, ARM_HAVE_NEON_DI_LDST): Likewise.
34517 (ARM_HAVE_NEON_V2DI_LDST): Likewise.
34518 (ARM_HAVE_V8QI_LDST, ARM_HAVE_V16QI_LDST): Likewise.
34519 (ARM_HAVE_V4HI_LDST, ARM_HAVE_V8HI_LDST): Likewise.
34520 (ARM_HAVE_V2SI_LDST, ARM_HAVE_V4SI_LDST, ARM_HAVE_V4HF_LDST): Likewise.
34521 (ARM_HAVE_V8HF_LDST, ARM_HAVE_V4BF_LDST, ARM_HAVE_V8BF_LDST): Likewise.
34522 (ARM_HAVE_V2SF_LDST, ARM_HAVE_V4SF_LDST, ARM_HAVE_DI_LDST): Likewise.
34523 (ARM_HAVE_V2DI_LDST): Likewise.
34524 * config/arm/mve.md (*movmisalign<mode>_mve_store): New pattern.
34525 (*movmisalign<mode>_mve_load): New pattern.
34526 * config/arm/neon.md (movmisalign<mode>): Move to ...
34527 * config/arm/vec-common.md: ... here.
34529 2021-01-12 Vladimir N. Makarov <vmakarov@redhat.com>
34532 * lra-eliminations.c (eliminate_regs_in_insn): Add transformation
34533 of pattern 'plus (plus (hard reg, const), pseudo)'.
34535 2021-01-12 Richard Biener <rguenther@suse.de>
34537 PR tree-optimization/98550
34538 * tree-vect-slp.c (vect_record_max_nunits): Check whether
34539 the group size is a multiple of the vector element count.
34540 (vect_build_slp_tree_1): When we need to fail because
34541 the vector type choosen causes unrolling do so lazily
34542 without affecting matches only at the end to guide group splitting.
34544 2021-01-12 Martin Liska <mliska@suse.cz>
34547 * optc-save-gen.awk: Compare also n_target_save vars with
34550 2021-01-12 Martin Liska <mliska@suse.cz>
34552 * gcov.c (source_info::debug): New.
34553 (print_usage): Add --debug (-D) option.
34554 (process_args): Likewise.
34555 (generate_results): Call src->debug after
34556 accumulate_line_counts.
34557 (read_graph_file): Properly assign id for EXIT_BLOCK.
34558 * profile.c (branch_prob): Dump function body before it is
34561 2021-01-12 Jakub Jelinek <jakub@redhat.com>
34563 PR tree-optimization/98629
34564 * tree-ssa-math-opts.c (arith_overflow_check_p): Don't update use_stmt
34565 unless returning non-zero.
34567 2021-01-12 Jakub Jelinek <jakub@redhat.com>
34569 PR tree-optimization/95731
34570 * tree-ssa-reassoc.c (optimize_range_tests_cmp_bitwise): Also optimize
34571 x < 0 && y < 0 && z < 0 into (x | y | z) < 0 for signed x, y, z.
34572 (optimize_range_tests): Call optimize_range_tests_cmp_bitwise
34573 only after optimize_range_tests_var_bound.
34575 2021-01-12 Jakub Jelinek <jakub@redhat.com>
34577 * configure.ac: Ensure c/Make-lang.in comes first in @all_lang_makefrags@.
34578 * configure: Regenerated.
34580 2021-01-12 liuhongt <hongtao.liu@intel.com>
34583 * config/i386/i386-builtins.h (BUILTIN_DESC_SWAP_OPERANDS):
34585 * config/i386/i386-expand.c (ix86_expand_sse_comi): Delete
34588 2021-01-12 Alexandre Oliva <oliva@adacore.com>
34590 * ssa-iterators.h (end_imm_use_stmt_traverse): Forward
34592 (auto_end_imm_use_stmt_traverse): New struct.
34593 (FOR_EACH_IMM_USE_STMT): Use it.
34594 (BREAK_FROM_IMM_USE_STMT, RETURN_FROM_IMM_USE_STMT): Remove,
34596 * gimple-ssa-strength-reduction.c: ... here, ...
34597 * graphite-scop-detection.c: ... here, ...
34598 * ipa-modref.c, ipa-pure-const.c, ipa-sra.c: ... here, ...
34599 * tree-predcom.c, tree-ssa-ccp.c: ... here, ...
34600 * tree-ssa-dce.c, tree-ssa-dse.c: ... here, ...
34601 * tree-ssa-loop-ivopts.c, tree-ssa-math-opts.c: ... here, ...
34602 * tree-ssa-phiprop.c, tree-ssa.c: ... here, ...
34603 * tree-vect-slp.c: ... and here, ...
34604 * doc/tree-ssa.texi: ... and the example here.
34606 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
34608 * config/aarch64/aarch64-sve.md (sdiv_pow2<mode>3): Extend from
34609 SVE_FULL_I to SVE_I. Generate an UNSPEC_PRED_X.
34610 (*sdiv_pow2<mode>3): New pattern.
34611 (@cond_<sve_int_op><mode>): Extend from SVE_FULL_I to SVE_I.
34612 Wrap the ASRD in an UNSPEC_PRED_X.
34613 (*cond_<sve_int_op><mode>_2): Likewise. Replace the UNSPEC_PRED_X
34614 predicate with a constant PTRUE, if it isn't already.
34615 (*cond_<sve_int_op><mode>_z): Replace with...
34616 (*cond_<sve_int_op><mode>_any): ...this new pattern.
34618 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
34620 * config/aarch64/aarch64-sve.md (*cond_bic<mode>_2): Extend from
34621 SVE_FULL_I to SVE_I.
34622 (*cond_bic<mode>_any): Likewise.
34624 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
34626 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart)
34627 (@aarch64_pred_<MUL_HIGHPART:optab><mode>): Extend from SVE_FULL_I
34630 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
34632 * config/aarch64/aarch64-sve.md (<su>abd<mode>_3): Extend from
34633 SVE_FULL_I to SVE_I.
34634 (*aarch64_cond_<su>abd<mode>_2): Likewise.
34635 (*aarch64_cond_<su>abd<mode>_any): Likewise.
34636 (@aarch64_pred_<su>abd<mode>): Likewise. Use UNSPEC_PRED_X
34637 for the max and min but not for the minus.
34638 (*aarch64_cond_<su>abd<mode>_3): New pattern.
34640 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
34642 * config/aarch64/iterators.md (SVE_24I): New iterator.
34643 * config/aarch64/aarch64-sve.md (*aarch64_adr<mode>_shift): Extend from
34644 SVE_FULL_SDI to SVE_24I. Use containers rather than elements.
34646 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
34648 * config/aarch64/aarch64-sve.md (@cond_<SVE_INT_BINARY:optab><mode>)
34649 (*cond_<SVE_INT_BINARY:optab><mode>_2): Extend from SVE_FULL_I
34651 (*cond_<SVE_INT_BINARY:optab><mode>_3): Likewise.
34652 (*cond_<SVE_INT_BINARY:optab><mode>_any): Likewise.
34653 (*cond_<SVE_INT_BINARY:optab><mode>_2_const): Likewise.
34654 (*cond_<SVE_INT_BINARY:optab><mode>_any_const): Likewise.
34656 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
34658 * config/aarch64/aarch64-sve.md (<SVE_INT_BINARY_IMM:optab><mode>3)
34659 (@aarch64_pred_<SVE_INT_BINARY_IMM:optab><mode>)
34660 (*post_ra_<SVE_INT_BINARY_IMM:optab><mode>3): Extend from SVE_FULL_I
34663 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
34665 * config/aarch64/aarch64-sve.md (<ASHIFT:optab><mode>3)
34666 (v<ASHIFT:optab><mode>3, @aarch64_pred_<optab><mode>)
34667 (*post_ra_v<ASHIFT:optab><mode>3): Extend from SVE_FULL_I to SVE_I.
34669 2021-01-11 Martin Liska <mliska@suse.cz>
34672 * symtab-clones.h (clone_info::release): Release
34673 symtab::m_clones with ggc_delete as it's a GGC memory.
34675 2021-01-11 Matthias Klose <doko@ubuntu.com>
34677 * Makefile.in (LINK_PROGRESS): Show the link target.
34679 2021-01-11 Richard Biener <rguenther@suse.de>
34681 PR tree-optimization/91403
34682 * tree-vect-data-refs.c (vect_analyze_group_access_1): Cap
34683 single-element interleaving group size at 4096 elements.
34685 2021-01-11 Richard Biener <rguenther@suse.de>
34687 PR tree-optimization/98526
34688 * tree-vect-loop.c (vect_model_reduction_cost): Remove costing
34689 of the actual reduction op for the regular case.
34690 (vectorizable_reduction): Cost the stmts
34691 vect_transform_reduction produces here.
34693 2021-01-11 Andreas Krebbel <krebbel@linux.ibm.com>
34695 * tree-ssa-forwprop.c (simplify_vector_constructor): For
34696 big-endian, use UNPACK[_FLOAT]_HI.
34698 2021-01-11 Tamar Christina <tamar.christina@arm.com>
34700 * tree-vect-slp-patterns.c (class complex_pattern,
34701 class complex_add_pattern): Add parameters to matches.
34702 (complex_add_pattern::build): Free memory.
34703 (complex_add_pattern::matches): Move validation end of match.
34704 (complex_add_pattern::recognize): Likewise.
34706 2021-01-11 Tamar Christina <tamar.christina@arm.com>
34708 * tree-vect-slp-patterns.c (linear_loads_p): Fix externals.
34710 2021-01-11 Tamar Christina <tamar.christina@arm.com>
34712 * tree-vect-slp-patterns.c (is_linear_load_p): Fix ambiguity.
34714 2021-01-11 Jakub Jelinek <jakub@redhat.com>
34716 PR tree-optimization/95867
34717 * tree-ssa-math-opts.h: New header.
34718 * tree-ssa-math-opts.c: Include tree-ssa-math-opts.h.
34719 (powi_as_mults): No longer static. Use build_one_cst instead of
34720 build_real. Formatting fix.
34721 * tree-ssa-reassoc.c: Include tree-ssa-math-opts.h.
34722 (attempt_builtin_powi): Handle multiplication reassociation without
34723 powi_fndecl using powi_as_mults.
34724 (reassociate_bb): For integral types don't require
34725 -funsafe-math-optimizations to call attempt_builtin_powi.
34727 2021-01-11 Jakub Jelinek <jakub@redhat.com>
34729 PR tree-optimization/95852
34730 * tree-ssa-math-opts.c (maybe_optimize_guarding_check): Change
34731 mul_stmts parameter type to vec<gimple *> &. Before cond_stmt
34732 allow in the bb any of the stmts in that vector, div_stmt and
34733 up to 3 cast stmts.
34734 (arith_cast_equal_p): New function.
34735 (arith_overflow_check_p): Add cast_stmt argument, handle signed
34736 multiply overflow checks.
34737 (match_arith_overflow): Adjust caller. Handle signed multiply
34740 2021-01-11 Jakub Jelinek <jakub@redhat.com>
34742 PR tree-optimization/95852
34743 * tree-ssa-math-opts.c (maybe_optimize_guarding_check): New function.
34744 (uaddsub_overflow_check_p): Renamed to ...
34745 (arith_overflow_check_p): ... this. Handle also multiplication
34746 with overflow check.
34747 (match_uaddsub_overflow): Renamed to ...
34748 (match_arith_overflow): ... this. Add cfg_changed argument. Handle
34749 also multiplication with overflow check. Adjust function comment.
34750 (math_opts_dom_walker::after_dom_children): Adjust callers. Call
34751 match_arith_overflow also for MULT_EXPR.
34753 2021-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
34755 * config/aarch64/arm_neon.h (vmovl_s8): Reimplement using
34756 __builtin_convertvector.
34757 (vmovl_s16): Likewise.
34758 (vmovl_s32): Likewise.
34759 (vmovl_u8): Likewise.
34760 (vmovl_u16): Likewise.
34761 (vmovl_u32): Likewise.
34762 (vmovn_s16): Likewise.
34763 (vmovn_s32): Likewise.
34764 (vmovn_s64): Likewise.
34765 (vmovn_u16): Likewise.
34766 (vmovn_u32): Likewise.
34767 (vmovn_u64): Likewise.
34769 2021-01-11 Martin Liska <mliska@suse.cz>
34771 * gimple-if-to-switch.cc (struct condition_info): Use auto_var.
34772 (if_chain::is_beneficial): Delete clusters
34773 (find_conditions): Make second argument of conditions_in_bbs a
34774 pointer so that we control over it's lifetime.
34775 (pass_if_to_switch::execute): Delete them.
34777 2021-01-11 Kewen Lin <linkw@linux.ibm.com>
34779 * ira.c (move_unallocated_pseudos): Check other_reg and skip if
34782 2021-01-09 Maciej W. Rozycki <macro@linux-mips.org>
34784 * config/vax/vax.md (cc): Remove mode attribute.
34785 (subst_<cc>, subst_f<cc>): Rename to...
34786 (subst_<mode>, subst_f<VAXccnz:mode>): ... these respectively.
34787 (*cbranch<VAXint:mode>4_<VAXcc:mode>): Update for `cc' removal.
34788 (*cbranch<VAXfp:mode>4_<VAXccnz:mode>): Likewise.
34789 (*branch_<mode>, *branch_<mode>_reversed): Likewise.
34791 2021-01-09 Maciej W. Rozycki <macro@linux-mips.org>
34793 * config/vax/vax.md (subst_f<cc>): Add mode to operands and
34794 `const_double_zero'.
34796 2021-01-09 Maciej W. Rozycki <macro@linux-mips.org>
34798 * config/pdp11/pdp11.md (PDPfp): New mode iterator.
34799 (fcc_cc, fcc_ccnz): Use it. Add mode to `const_double_zero' and
34802 2021-01-09 Maciej W. Rozycki <macro@linux-mips.org>
34804 * genemit.c (gen_exp) <CONST_DOUBLE>: Handle `const_double_zero'
34806 * read-rtl.c (rtx_reader::read_rtx_code): Handle machine mode
34807 with `const_double_zero'.
34808 * doc/rtl.texi (Constant Expression Types): Document it.
34810 2021-01-09 Jakub Jelinek <jakub@redhat.com>
34813 * tree-cfg.c (verify_gimple_assign_binary): Allow lhs of
34814 POINTER_DIFF_EXPR to be any integral type.
34816 2021-01-09 Jakub Jelinek <jakub@redhat.com>
34818 PR rtl-optimization/98603
34819 * function.c (instantiate_virtual_regs_in_insn): For asm goto
34820 with impossible constraints, drop all SETs, CLOBBERs, drop PARALLEL
34821 if any, set ASM_OPERANDS mode to VOIDmode and change
34822 ASM_OPERANDS_OUTPUT_CONSTRAINT and ASM_OPERANDS_OUTPUT_IDX.
34824 2021-01-09 Alexandre Oliva <oliva@gnu.org>
34827 * final.c (notice_source_line): Narrow down the condition to
34828 skip a line-0 marker.
34830 2021-01-08 Sergei Trofimovich <siarheit@google.com>
34832 * ipa-modref.c (merge_call_side_effects): Fix
34833 linebreak split by reordering two print calls.
34835 2021-01-08 Ilya Leoshkevich <iii@linux.ibm.com>
34837 * config/s390/vector.md (*tf_to_fprx2_0): Rename from
34838 "*mov_tf_to_fprx2_0" for consistency, fix constraint.
34839 (*tf_to_fprx2_1): Rename from "*mov_tf_to_fprx2_1" for
34840 consistency, fix constraint.
34842 2021-01-08 Ilya Leoshkevich <iii@linux.ibm.com>
34844 * config/s390/s390-c.c (s390_def_or_undef_macro): Accept
34845 callables instead of mask values.
34846 (struct target_flag_set_p): New predicate.
34847 (s390_cpu_cpp_builtins_internal): Define or undefine
34848 __LONG_DOUBLE_VX__ macro.
34850 2021-01-08 H.J. Lu <hjl.tools@gmail.com>
34853 * config/i386/i386.c (x86_function_profiler): Use R10 and R11
34854 to call mcount in large model with PIC for NO_PROFILE_COUNTERS
34857 2021-01-08 Richard Biener <rguenther@suse.de>
34859 * tree-ssa-sccvn.c (pass_fre::execute): Reset the SCEV hash table.
34861 2021-01-08 Richard Biener <rguenther@suse.de>
34863 * tree-vect-slp.c (scalar_stmts_to_slp_tree_map_t): Fix.
34864 (vect_build_slp_tree): On cache hit release the matched
34865 scalar stmts vector.
34866 * tree-vect-stmts.c (vectorizable_store): Properly free
34867 vec_oprnds before possibly gathering them again.
34869 2021-01-08 Richard Biener <rguenther@suse.de>
34871 PR tree-optimization/98544
34872 * tree-vect-slp.c (vect_optimize_slp): Always materialize
34873 permutes at a permute node.
34875 2021-01-08 H.J. Lu <hjl.tools@gmail.com>
34878 * config/i386/i386.c (x86_function_profiler): Use R10 to call
34879 mcount in large model. Sorry for large model with PIC.
34881 2021-01-08 Jakub Jelinek <jakub@redhat.com>
34884 * config/i386/i386.opt (ix86_cmodel, ix86_incoming_stack_boundary_arg,
34885 ix86_pmode, ix86_preferred_stack_boundary_arg, ix86_regparm,
34886 ix86_veclibabi_type): Remove x_ prefix, use TargetVariable instead of
34887 TargetSave and initialize for variables with enum types.
34888 (mfentry, mstack-protector-guard-reg=, mstack-protector-guard-offset=,
34889 mstack-protector-guard-symbol=): Add Save.
34890 * config/i386/i386-options.c (ix86_function_specific_save,
34891 ix86_function_specific_restore): Don't save or restore x_ix86_cmodel,
34892 x_ix86_incoming_stack_boundary_arg, x_ix86_pmode,
34893 x_ix86_preferred_stack_boundary_arg, x_ix86_regparm,
34894 x_ix86_veclibabi_type.
34896 2021-01-08 Richard Sandiford <richard.sandiford@arm.com>
34898 * config/aarch64/aarch64-sve.md (*cnot<mode>): Extend from
34899 SVE_FULL_I to SVE_I.
34900 (*cond_cnot<mode>_2, *cond_cnot<mode>_any): Likewise.
34902 2021-01-08 Richard Sandiford <richard.sandiford@arm.com>
34904 * config/aarch64/aarch64-sve.md (*cond_uxt<mode>_2): Extend from
34905 SVE_FULL_I to SVE_I.
34906 (*cond_uxt<mode>_any): Likewise.
34908 2021-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
34910 * config/aarch64/iterators.md (Vwhalf): New iterator.
34911 * config/aarch64/aarch64-simd.md (aarch64_<sur>adalp<mode>_3):
34913 (aarch64_<sur>adalp<mode>): ... This. Make more
34915 (<sur>sadv16qi): Adjust callsite of the above.
34916 * config/aarch64/aarch64-simd-builtins.def (sadalp, uadalp): New
34918 * config/aarch64/arm_neon.h (vpadal_s8): Reimplement using
34920 (vpadal_s16): Likewise.
34921 (vpadal_u8): Likewise.
34922 (vpadal_u16): Likewise.
34923 (vpadalq_s8): Likewise.
34924 (vpadalq_s16): Likewise.
34925 (vpadalq_s32): Likewise.
34926 (vpadalq_u8): Likewise.
34927 (vpadalq_u16): Likewise.
34928 (vpadalq_u32): Likewise.
34930 2021-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
34932 * config/aarch64/aarch64-simd.md (aarch64_<su>abd<mode>_3):
34934 (aarch64_<su>abd<mode>): ... This.
34935 (<sur>sadv16qi): Adjust callsite of the above.
34936 * config/aarch64/aarch64-simd-builtins.def (sabd, uabd): Define
34938 * config/aarch64/arm_neon.h (vabd_s8): Reimplement using
34940 (vabd_s16): Likewise.
34941 (vabd_s32): Likewise.
34942 (vabd_u8): Likewise.
34943 (vabd_u16): Likewise.
34944 (vabd_u32): Likewise.
34945 (vabdq_s8): Likewise.
34946 (vabdq_s16): Likewise.
34947 (vabdq_s32): Likewise.
34948 (vabdq_u8): Likewise.
34949 (vabdq_u16): Likewise.
34950 (vabdq_u32): Likewise.
34952 2021-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
34954 * config/aarch64/aarch64-simd-builtins.def (saba, uaba): Define
34956 * config/aarch64/arm_neon.h (vaba_s8): Implement using builtin.
34957 (vaba_s16): Likewise.
34958 (vaba_s32): Likewise.
34959 (vaba_u8): Likewise.
34960 (vaba_u16): Likewise.
34961 (vaba_u32): Likewise.
34962 (vabaq_s8): Likewise.
34963 (vabaq_s16): Likewise.
34964 (vabaq_s32): Likewise.
34965 (vabaq_u8): Likewise.
34966 (vabaq_u16): Likewise.
34967 (vabaq_u32): Likewise.
34969 2021-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
34971 * config/aarch64/aarch64-simd.md (aba<mode>_3): Rename to...
34972 (aarch64_<su>aba<mode>): ... This. Handle uaba as well.
34973 Change RTL pattern to match.
34975 2021-01-08 Kito Cheng <kito.cheng@sifive.com>
34977 * common/config/riscv/riscv-common.c (riscv_current_subset_list): New.
34978 * config/riscv/riscv-c.c (riscv-subset.h): New.
34979 (INCLUDE_STRING): Define.
34980 (riscv_cpu_cpp_builtins): Add new style architecture extension
34982 * config/riscv/riscv-subset.h (riscv_subset_list::begin): New.
34983 (riscv_subset_list::end): New.
34984 (riscv_current_subset_list): New.
34986 2021-01-08 Kito Cheng <kito.cheng@sifive.com>
34988 * common/config/riscv/riscv-common.c (RISCV_DONT_CARE_VERSION):
34989 Move to riscv-subset.h.
34990 (struct riscv_subset_t): Ditto.
34991 (class riscv_subset_list): Ditto.
34992 * config/riscv/riscv-subset.h (RISCV_DONT_CARE_VERSION): Move
34993 from riscv-common.c.
34994 (struct riscv_subset_t): Ditto.
34995 (class riscv_subset_list): Ditto.
34996 * config/riscv/t-riscv ($(common_out_file)): Add file
34999 2021-01-07 Jakub Jelinek <jakub@redhat.com>
35002 * config/i386/i386.md (*bmi_blsi_<mode>_cmp, *bmi_blsi_<mode>_ccno):
35003 New define_insn patterns.
35005 2021-01-07 Richard Sandiford <richard.sandiford@arm.com>
35007 * config/aarch64/aarch64-sve.md (@cond_<SVE_INT_UNARY:optab><mode>)
35008 (*cond_<SVE_INT_UNARY:optab><mode>_2): Extend from SVE_FULL_I to SVE_I.
35009 (*cond_<SVE_INT_UNARY:optab><mode>_any): Likewise.
35011 2021-01-07 Richard Sandiford <richard.sandiford@arm.com>
35013 PR tree-optimization/98560
35014 * internal-fn.def (IFN_VCONDU, IFN_VCONDEQ): Use type vec_cond.
35015 * internal-fn.c (vec_cond_mask_direct): Get the data mode from
35017 (vec_cond_direct): Likewise argument 2.
35018 (vec_condu_direct, vec_condeq_direct): Delete.
35019 (expand_vect_cond_optab_fn): Rename to...
35020 (expand_vec_cond_optab_fn): ...this, replacing old macro.
35021 (expand_vec_condu_optab_fn, expand_vec_condeq_optab_fn): Delete.
35022 (expand_vect_cond_mask_optab_fn): Rename to...
35023 (expand_vec_cond_mask_optab_fn): ...this, replacing old macro.
35024 (direct_vec_cond_mask_optab_supported_p): Treat the optab as a
35026 (direct_vec_cond_optab_supported_p): Likewise.
35027 (direct_vec_condu_optab_supported_p): Delete.
35028 (direct_vec_condeq_optab_supported_p): Delete.
35029 * gimple-isel.cc: Include internal-fn.h.
35030 (gimple_expand_vec_cond_expr): Check that IFN_VCONDEQ is supported
35033 2021-01-07 Richard Sandiford <richard.sandiford@arm.com>
35035 PR tree-optimization/98560
35036 * gimple-isel.cc (gimple_expand_vec_cond_expr): If we fail to use
35037 IFN_VCOND{,U,EQ}, fall back on IFN_VCOND_MASK.
35039 2021-01-07 Uroš Bizjak <ubizjak@gmail.com>
35041 * config/i386/i386.md (insn): Merge from plusminus_insn, shift_insn,
35042 rotate_insn and optab code attributes.
35043 Update all uses to merged code attribute.
35044 * config/i386/sse.md: Update all uses to merged code attribute.
35045 * config/i386/mmx.md: Update all uses to merged code attribute.
35047 2021-01-07 Jakub Jelinek <jakub@redhat.com>
35049 PR tree-optimization/98568
35050 * gimple-ssa-store-merging.c (bswap_view_convert): New function.
35051 (bswap_replace): Use it.
35053 2021-01-06 Vladimir N. Makarov <vmakarov@redhat.com>
35055 PR rtl-optimization/97978
35056 * lra-int.h (lra_hard_reg_split_p): New external.
35057 * lra.c (lra_hard_reg_split_p): New global.
35058 (lra): Set up lra_hard_reg_split_p after splitting a hard reg.
35059 * lra-assigns.c (lra_assign): Don't check allocation correctness
35060 after hard reg splitting.
35062 2021-01-06 Martin Sebor <msebor@redhat.com>
35065 * builtins.c (new_delete_mismatch_p): New overload.
35066 (new_delete_mismatch_p (tree, tree)): Call it.
35068 2021-01-06 Alexandre Oliva <oliva@adacore.com>
35070 * Makefile.in (T_GLIMITS_H): New.
35071 (stmp-int-hdrs): Depend on it, use it.
35072 * config/t-vxworks (T_GLIMITS_H): Override it.
35073 (vxw-glimits.h): New.
35075 2021-01-06 Richard Biener <rguenther@suse.de>
35077 PR tree-optimization/98513
35078 * value-range.cc (intersect_ranges): Compare the upper bounds
35079 for the expected relation.
35081 2021-01-06 Gerald Pfeifer <gerald@pfeifer.com>
35084 2020-12-28 Gerald Pfeifer <gerald@pfeifer.com>
35086 * doc/standards.texi (HSAIL): Remove section.
35088 2021-01-05 Samuel Thibault <samuel.thibault@ens-lyon.org>
35090 * configure: Re-generate.
35092 2021-01-05 Jakub Jelinek <jakub@redhat.com>
35094 * doc/invoke.texi (-std=c++20): Adjust for the publication of
35095 ISO 14882:2020 standard.
35096 * doc/standards.texi: Likewise.
35098 2021-01-05 Jakub Jelinek <jakub@redhat.com>
35100 PR tree-optimization/94802
35101 * expr.h (maybe_optimize_sub_cmp_0): Declare.
35102 * expr.c: Include tree-pretty-print.h and flags.h.
35103 (maybe_optimize_sub_cmp_0): New function.
35104 (do_store_flag): Use it.
35105 * cfgexpand.c (expand_gimple_cond): Likewise.
35107 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
35109 * mux-utils.h (pointer_mux::m_ptr): Tweak description of contents.
35110 * rtlanal.c (simple_regno_set): Tweak description to clarify the
35113 2021-01-05 Richard Biener <rguenther@suse.de>
35115 PR tree-optimization/98516
35116 * tree-vect-slp.c (vect_optimize_slp): Permute the incoming
35117 lanes when materializing on a VEC_PERM node.
35118 (vectorizable_slp_permutation): Dump the permute properly.
35120 2021-01-05 Richard Biener <rguenther@suse.de>
35122 * tree-vect-slp.c (vect_slp_region): Move debug counter
35123 to cover individual subgraphs.
35125 2021-01-05 Richard Biener <rguenther@suse.de>
35127 PR tree-optimization/98428
35128 * tree-vect-slp.c (vect_build_slp_tree_1): Properly reject
35129 vector lane extracts for loop vectorization.
35131 2021-01-05 Jakub Jelinek <jakub@redhat.com>
35133 PR tree-optimization/98514
35134 * tree-ssa-reassoc.c (bb_rank): Change type from long * to
35136 (operand_rank): Change type from hash_map<tree, long> to
35137 hash_map<tree, int64_t>.
35138 (phi_rank): Change return type from long to int64_t.
35139 (loop_carried_phi): Change block_rank variable type from long to
35141 (propagate_rank): Change return type, rank parameter type and
35142 op_rank variable type from long to int64_t.
35143 (find_operand_rank): Change return type from long to int64_t
35144 and change slot variable type from long * to int64_t *.
35145 (insert_operand_rank): Change rank parameter type from long to
35147 (get_rank): Change return type and rank variable type from long to
35148 int64_t. Use PRId64 instead of ld to print the rank.
35149 (init_reassoc): Change rank variable type from long to int64_t
35150 and adjust correspondingly bb_rank and operand_rank initialization.
35152 2021-01-05 Jakub Jelinek <jakub@redhat.com>
35154 PR tree-optimization/96928
35155 * tree-ssa-phiopt.c (xor_replacement): New function.
35156 (tree_ssa_phiopt_worker): Call it.
35158 2021-01-05 Jakub Jelinek <jakub@redhat.com>
35160 PR tree-optimization/96930
35161 * match.pd ((A / (1 << B)) -> (A >> B)): If A is extended
35162 from narrower value which has the same type as 1 << B, perform
35163 the right shift on the narrower value followed by extension.
35165 2021-01-05 Jakub Jelinek <jakub@redhat.com>
35167 PR tree-optimization/96239
35168 * gimple-ssa-store-merging.c (maybe_optimize_vector_constructor): New
35170 (get_status_for_store_merging): Don't return BB_INVALID for blocks
35171 with potential bswap optimizable CONSTRUCTORs.
35172 (pass_store_merging::execute): Optimize vector CONSTRUCTORs with bswap
35175 2021-01-05 Richard Biener <rguenther@suse.de>
35177 PR tree-optimization/98381
35178 * tree.c (vector_element_bits): Properly compute bool vector
35180 * tree-vect-loop.c (vectorizable_live_operation): Properly
35181 compute the last lane bit offset.
35183 2021-01-05 Uroš Bizjak <ubizjak@gmail.com>
35186 * config/i386/sse.md (sse_cvtps2pi): Redefine as define_insn_and_split.
35187 Clear the top 64 bytes of the input XMM register.
35188 (sse_cvttps2pi): Ditto.
35190 2021-01-05 Uroš Bizjak <ubizjak@gmail.com>
35193 * config/i386/xopintrin.h (_mm256_cmov_si256): New.
35195 2021-01-05 H.J. Lu <hjl.tools@gmail.com>
35198 * config/i386/xmmintrin.h (_mm_extract_pi16): Cast to unsigned
35201 2021-01-05 Claudiu Zissulescu <claziss@synopsys.com>
35203 * config/arc/arc.md (maddsidi4_split): Use ACC_REG_FIRST.
35204 (umaddsidi4_split): Likewise.
35206 2021-01-05 liuhongt <hongtao.liu@intel.com>
35209 * config/i386/sse.md (*sse2_pmovskb_zexthisi): New
35210 define_insn_and_split for zero_extend of subreg HI of pmovskb
35212 (*sse2_pmovskb_zexthisi): Add new combine splitters for
35213 zero_extend of not of subreg HI of pmovskb result.
35215 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
35218 * explow.c (convert_memory_address_addr_space_1): Handle UNSPECs
35220 * config/aarch64/aarch64.c (aarch64_expand_mov_immediate): Use
35221 convert_memory_address to convert symbolic immediates to ptr_mode
35222 before forcing them to memory.
35224 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
35226 PR rtl-optimization/97144
35227 * recog.c (constrain_operands): Initialize matching_operand
35228 for each alternative, rather than only doing it once.
35230 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
35232 PR rtl-optimization/98403
35233 * rtl-ssa/changes.cc (function_info::finalize_new_accesses): Explain
35234 why we don't remove call clobbers.
35235 (function_info::apply_changes_to_insn): Don't attempt to add
35236 call clobbers here.
35238 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
35240 PR tree-optimization/98371
35241 * tree-vect-loop.c (vect_reanalyze_as_main_loop): New function.
35242 (vect_analyze_loop): If an epilogue loop appears to be cheaper
35243 than the main loop, re-analyze it as a main loop before adopting
35246 2021-01-05 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
35249 * configure.ac (NETLIBS): Determine using AX_LIB_SOCKET_NSL.
35250 * aclocal.m4, configure: Regenerate.
35251 * Makefile.in (NETLIBS): Define.
35252 (BACKEND): Remove $(CODYLIB).
35254 2021-01-05 Jakub Jelinek <jakub@redhat.com>
35256 PR rtl-optimization/98334
35257 * simplify-rtx.c (simplify_context::simplify_binary_operation_1):
35258 Optimize (X - 1) * Y + Y to X * Y or (X + 1) * Y - Y to X * Y.
35260 2021-01-05 Bernd Edlinger <bernd.edlinger@hotmail.de>
35262 * tree-inline.c (expand_call_inline): Restore input_location.
35263 Return result from recursive call.
35265 2021-01-04 Richard Sandiford <richard.sandiford@arm.com>
35267 PR tree-optimization/95401
35268 * config/aarch64/aarch64-sve-builtins.cc
35269 (gimple_folder::load_store_cookie): Use bits rather than bytes
35270 for the alignment argument to IFN_MASK_LOAD and IFN_MASK_STORE.
35271 * gimple-fold.c (gimple_fold_mask_load_store_mem_ref): Likewise.
35272 * tree-vect-stmts.c (vectorizable_store): Likewise.
35273 (vectorizable_load): Likewise.
35275 2021-01-04 Richard Biener <rguenther@suse.de>
35277 PR tree-optimization/98308
35278 * tree-vect-stmts.c (vectorizable_load): Set invariant mask
35281 2021-01-04 Jakub Jelinek <jakub@redhat.com>
35283 PR tree-optimization/95771
35284 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Handle types
35285 with precision smaller than int's precision and types with precision
35286 twice as large as long long. Formatting fixes.
35288 2021-01-04 Richard Biener <rguenther@suse.de>
35290 PR tree-optimization/98464
35291 * tree-ssa-sccvn.c (vn_valueize_for_srt): Rename from ...
35292 (vn_valueize_wrapper): ... this. Temporarily adjust vn_context_bb.
35293 (process_bb): Adjust.
35295 2021-01-04 Matthew Malcomson <matthew.malcomson@arm.com>
35298 * doc/invoke.texi (-fsanitize=address): Fix wording describing
35299 clash with -fsanitize=hwaddress.
35301 2021-01-04 Richard Biener <rguenther@suse.de>
35303 PR tree-optimization/98282
35304 * tree-ssa-sccvn.c (vn_get_stmt_kind): Classify tcc_reference on
35305 invariants as VN_NARY.
35307 2021-01-04 Richard Sandiford <richard.sandiford@arm.com>
35310 * config/aarch64/aarch64-simd.md (aarch64_combine<mode>): Accept
35311 aarch64_simd_reg_or_zero for operand 2. Use the combinez patterns
35312 to handle zero operands.
35314 2021-01-04 Richard Sandiford <richard.sandiford@arm.com>
35316 * config/aarch64/aarch64.c (offset_6bit_signed_scaled_p): New function.
35317 (offset_6bit_unsigned_scaled_p): Fix typo in comment.
35318 (aarch64_sve_prefetch_operand_p): Accept MUL VLs in the range
35321 2021-01-04 Richard Biener <rguenther@suse.de>
35323 PR tree-optimization/98393
35324 * tree-vect-slp.c (vect_build_slp_tree): Properly zero matches
35325 when hitting the limit.
35327 2021-01-04 Richard Biener <rguenther@suse.de>
35329 PR tree-optimization/98291
35330 * tree-vect-loop.c (vectorizable_reduction): Bypass
35331 associativity check for SLP reductions with VF 1.
35333 2021-01-04 Jakub Jelinek <jakub@redhat.com>
35335 PR tree-optimization/96782
35336 * match.pd (x == ~x -> false, x != ~x -> true): New simplifications.
35338 2021-01-04 Bernd Edlinger <bernd.edlinger@hotmail.de>
35340 * collect-utils.c (collect_execute): Check dumppfx.
35341 * collect2.c (maybe_run_lto_and_relink, do_link): Pass atsuffix
35342 to collect_execute.
35343 (do_link): Add new parameter atsuffix.
35344 (main): Handle -dumpdir option. Skip one argument for
35345 -o, -isystem and -B options.
35346 * gcc.c (make_at_file): New helper function.
35347 (close_at_file): Use it.
35349 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
35351 * config/darwin.h (MIN_LD64_NO_COAL_SECTS): Adjust.
35352 Amend handling for LD64_VERSION fallback defaults.
35354 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
35356 * config.gcc: Compute default version information
35357 from the configured target. Likewise defaults for
35359 * config/darwin10.h: Removed.
35360 * config/darwin12.h: Removed.
35361 * config/darwin9.h: Removed.
35362 * config/rs6000/darwin8.h: Removed.
35364 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
35366 * config/darwin9.h (ASM_OUTPUT_ALIGNED_COMMON): Delete.
35368 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
35370 * config/darwin9.h (STACK_CHECK_STATIC_BUILTIN): Move from here..
35371 * config/darwin.h (STACK_CHECK_STATIC_BUILTIN): .. to here.
35373 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
35375 * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Move from
35377 * config/darwin.h (LINK_GCC_C_SEQUENCE_SPEC): ... to here.
35379 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
35381 * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Move the spec
35382 for the Darwin10 unwinder stub from here ...
35383 * config/darwin.h (LINK_COMMAND_SPEC_A): ... to here.
35385 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
35387 * config/darwin.h (DSYMUTIL_SPEC): Default to DWARF
35388 (ASM_DEBUG_SPEC):Only define if the assembler supports
35390 (PREFERRED_DEBUGGING_TYPE): Default to DWARF.
35391 (DARWIN_PREFER_DWARF): Define.
35392 * config/darwin9.h (PREFERRED_DEBUGGING_TYPE): Remove.
35393 (DARWIN_PREFER_DWARF): Likewise
35394 (DSYMUTIL_SPEC): Likewise.
35395 (COLLECT_RUN_DSYMUTIL): Likewise.
35396 (ASM_DEBUG_SPEC): Likewise.
35397 (ASM_DEBUG_OPTION_SPEC): Likewise.
35399 2021-01-02 Jan Hubicka <jh@suse.cz>
35401 * cfg.c (free_block): ggc_free bb.
35403 2021-01-01 Jakub Jelinek <jakub@redhat.com>
35405 * gcc.c (process_command): Update copyright notice dates.
35406 * gcov-dump.c (print_version): Ditto.
35407 * gcov.c (print_version): Ditto.
35408 * gcov-tool.c (print_version): Ditto.
35409 * gengtype.c (create_file): Ditto.
35410 * doc/cpp.texi: Bump @copying's copyright year.
35411 * doc/cppinternals.texi: Ditto.
35412 * doc/gcc.texi: Ditto.
35413 * doc/gccint.texi: Ditto.
35414 * doc/gcov.texi: Ditto.
35415 * doc/install.texi: Ditto.
35416 * doc/invoke.texi: Ditto.
35418 2021-01-01 Jakub Jelinek <jakub@redhat.com>
35420 * ChangeLog-2020: Rotate ChangeLog. New file.
35423 Copyright (C) 2021 Free Software Foundation, Inc.
35425 Copying and distribution of this file, with or without modification,
35426 are permitted in any medium without royalty provided the copyright
35427 notice and this notice are preserved.