1 2021-12-16 Uroš Bizjak <ubizjak@gmail.com>
4 * config/i386/i386.h (VALID_AVX256_REG_MODE): Add V16HFmode.
5 (VALID_AVX256_REG_OR_OI_VHF_MODE): Replace with ...
6 (VALID_AVX256_REG_OR_OI_MODE): ... this. Remove V16HFmode.
7 (VALID_AVX512F_SCALAR_MODE): Remove HImode and HFmode.
8 (VALID_AVX512FP16_SCALAR_MODE): New.
9 (VALID_AVX512F_REG_MODE): Add V32HFmode.
10 (VALID_SSE2_REG_MODE): Add V8HFmode, V4HFmode and V2HFmode.
11 (VALID_SSE2_REG_VHF_MODE): Remove.
12 (VALID_INT_MODE_P): Add V2HFmode.
13 * config/i386/i386.c (function_arg_advance_64):
14 Remove explicit mention of V16HFmode and V32HFmode.
15 (ix86_hard_regno_mode_ok): Remove explicit mention of XImode
16 and V32HFmode, use VALID_AVX512F_REG_OR_XI_MODE instead.
17 Use VALID_AVX512FP_SCALAR_MODE for TARGET_aVX512FP16.
18 Use VALID_AVX256_REG_OR_OI_MODE instead of
19 VALID_AVX256_REG_OR_OI_VHF_MODE and VALID_SSE2_REG_MODE instead
20 of VALID_SSE2_REG_VHF_MODE.
21 (ix86_set_reg_reg_cost): Remove usge of VALID_AVX512FP16_REG_MODE.
22 (ix86_vector_mode_supported): Ditto.
24 2021-12-16 Martin Liska <mliska@suse.cz>
27 * attribs.c (decl_attributes): Check if
28 target_option_current_node is changed.
30 2021-12-16 Martin Liska <mliska@suse.cz>
32 * doc/invoke.texi: Add missing dash.
34 2021-12-16 Bill Schmidt <wschmidt@linux.ibm.com>
36 * config/rs6000/rs6000-c.c (resolution): New enum.
37 (resolve_vec_mul): New function.
38 (resolve_vec_cmpne): Likewise.
39 (resolve_vec_adde_sube): Likewise.
40 (resolve_vec_addec_subec): Likewise.
41 (resolve_vec_splats): Likewise.
42 (resolve_vec_extract): Likewise.
43 (resolve_vec_insert): Likewise.
44 (resolve_vec_step): Likewise.
45 (find_instance): Likewise.
46 (altivec_resolve_overloaded_builtin): Many cleanups. Call factored-out
47 functions. Move variable declarations closer to uses. Add commentary.
48 Remove unnecessary levels of braces. Avoid use of gotos. Change
49 misleading variable names. Use switches over if-else-if chains.
51 2021-12-16 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
54 * config/aarch64/aarch64-simd.md (aarch64_movv8di): Allow big endian
57 2021-12-16 Tobias Burnus <tobias@codesourcery.com>
60 * gcc.c (check_offload_target_name): Add 'default' and 'disable'
61 to the candidate list.
63 2021-12-15 Iain Sandoe <iain@sandoe.co.uk>
65 * configure: Regenerate.
67 2021-12-15 Roger Sayle <roger@nextmovesoftware.com>
69 * config/nvptx/nvptx-opts.h (ptx_isa): PTX_ISA_SM75 and PTX_ISA_SM80
71 * config/nvptx/nvptx.opt: Add sm_75 and sm_80 to -misa.
72 * config/nvptx/nvptx.h (TARGET_SM75, TARGET_SM80):
73 New helper macros to conditionalize functionality on target ISA.
74 * config/nvptx/nvptx-c.c (nvptx_cpu_cpp_builtins): Add __PTX_SM__
75 support for the new ISA levels.
76 * config/nvptx/nvptx.c (nvptx_file_start): Add support for TARGET_SM75
78 * config/nvptx/nvptx.md (define_c_enum "unspec"): New UNSPEC_TANH.
79 (define_mode_iterator HSFM): New iterator for HFmode and SFmode.
80 (exp2hf2): New define_insn controlled by TARGET_SM75.
81 (tanh<mode>2): New define_insn controlled by TARGET_SM75.
82 (sminhf3, smaxhf3): New define_isnns controlled by TARGET_SM80.
84 2021-12-15 Tom de Vries <tdevries@suse.de>
86 * config/nvptx/nvptx-opts.h (enum ptx_version): Add PTX_VERSION_7_0.
87 * config/nvptx/nvptx.c (nvptx_file_start): Handle TARGET_PTX_7_0.
88 * config/nvptx/nvptx.h (TARGET_PTX_7_0): New macro.
89 * config/nvptx/nvptx.opt (ptx_version): Add 7.0.
91 2021-12-15 Richard Sandiford <richard.sandiford@arm.com>
92 Tamar Christina <tamar.christina@arm.com>
95 * config/aarch64/aarch64.c (aarch64_short_vector_p): Return false
96 for structure modes, rather than ignoring the type in that case.
98 2021-12-15 Tamar Christina <tamar.christina@arm.com>
100 PR rtl-optimization/103350
101 * ree.c (add_removable_extension): Don't stop at first definition but
104 2021-12-15 Jakub Jelinek <jakub@redhat.com>
107 * dwarf2cfi.c (dwf_cfa_reg): Remove gcc_assert.
108 (operator==, operator!=): New overloaded operators.
109 (dwarf2out_frame_debug_adjust_cfa, dwarf2out_frame_debug_cfa_offset,
110 dwarf2out_frame_debug_expr): Compare vars with cfa_reg type directly
111 with REG rtxes rather than with dwf_cfa_reg results on those REGs.
112 (create_cie_data): Use stack_pointer_rtx instead of
113 gen_rtx_REG (Pmode, STACK_POINTER_REGNUM).
114 (execute_dwarf2_frame): Use hard_frame_pointer_rtx instead of
115 gen_rtx_REG (Pmode, HARD_FRAME_POINTER_REGNUM).
117 2021-12-15 Martin Liska <mliska@suse.cz>
120 * config/i386/i386-builtins.c (fold_builtin_cpu): Compare to 0
121 as API expects that non-zero values are returned (do that
123 For "avx512vbmi2" argument, we return now 1 << 31, which is a
124 negative integer value.
126 2021-12-15 Haochen Jiang <haochen.jiang@intel.com>
129 * config/i386/predicates.md (const_vector_operand):
131 * config/i386/sse.md(<insn><mode>3<mask_name>):
132 Add new define_split below.
134 2021-12-15 Michael Meissner <meissner@the-meissners.org>
136 * config/rs6000/rs6000.md (UNSPEC_XXSPLTIDP_CONST): New unspec.
137 (UNSPEC_XXSPLTIW_CONST): New unspec.
138 (movsf_hardfloat): Add support for generating XXSPLTIDP.
139 (mov<mode>_hardfloat32): Likewise.
140 (mov<mode>_hardfloat64): Likewise.
141 (xxspltidp_<mode>_internal): New insns.
142 (xxspltiw_<mode>_internal): New insns.
143 (splitters for SF/DFmode): Add new splitters for XXSPLTIDP.
145 2021-12-15 Michael Meissner <meissner@the-meissners.org>
147 * config/rs6000/predicates.md (easy_fp_constant): Add support for
148 generating XXSPLTIDP.
149 (vsx_prefixed_constant): Likewise.
150 (easy_vector_constant): Likewise.
151 * config/rs6000/rs6000-protos.h (constant_generates_xxspltidp):
153 * config/rs6000/rs6000.c (output_vec_const_move): Add support for
154 generating XXSPLTIDP.
155 (prefixed_xxsplti_p): Likewise.
156 (constant_generates_xxspltidp): New function.
157 * config/rs6000/rs6000.opt (-msplat-float-constant): New debug option.
159 2021-12-15 Michael Meissner <meissner@linux.ibm.com>
161 * config/rs6000/constraints.md (eP): Update comment.
162 * config/rs6000/predicates.md (easy_fp_constant): Add support for
164 (vsx_prefixed_constant): New predicate.
165 (easy_vector_constant): Add support for
167 * config/rs6000/rs6000-protos.h (prefixed_xxsplti_p): New
169 (constant_generates_xxspltiw): Likewise.
170 * config/rs6000/rs6000.c (xxspltib_constant_p): Generate XXSPLTIW
171 if possible instead of XXSPLTIB and sign extending the constant.
172 (output_vec_const_move): Add support for XXSPLTIW.
173 (prefixed_xxsplti_p): New function.
174 (constant_generates_xxspltiw): New function.
175 * config/rs6000/rs6000.md (prefixed attribute): Add support to
176 mark XXSPLTI* instructions as being prefixed.
177 * config/rs6000/rs6000.opt (-msplat-word-constant): New debug
179 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Add support for
180 generating XXSPLTIW or XXSPLTIDP.
181 (vsx_mov<mode>_32bit): Likewise.
182 * doc/md.texi (PowerPC and IBM RS6000 constraints): Document the
185 2021-12-15 Michael Meissner <meissner@the-meissners.org>
187 * config/rs6000/constraints.md (eQ): New constraint.
188 * config/rs6000/predicates.md (easy_fp_constant): Add support for
189 generating the LXVKQ instruction.
190 (easy_vector_constant_ieee128): New predicate.
191 (easy_vector_constant): Add support for generating the LXVKQ
193 * config/rs6000/rs6000-protos.h (constant_generates_lxvkq): New
195 * config/rs6000/rs6000.c (output_vec_const_move): Add support for
197 (constant_generates_lxvkq): New function.
198 * config/rs6000/rs6000.opt (-mieee128-constant): New debug
200 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Add support for
202 (vsx_mov<mode>_32bit): Likewise.
203 * doc/md.texi (PowerPC and IBM RS6000 constraints): Document the
206 2021-12-15 Michael Meissner <meissner@the-meissners.org>
208 * config/rs6000/rs6000-protos.h (VECTOR_128BIT_BITS): New macro.
209 (VECTOR_128BIT_BYTES): Likewise.
210 (VECTOR_128BIT_HALF_WORDS): Likewise.
211 (VECTOR_128BIT_WORDS): Likewise.
212 (VECTOR_128BIT_DOUBLE_WORDS): Likewise.
213 (vec_const_128bit_type): New structure type.
214 (vec_const_128bit_to_bytes): New declaration.
215 * config/rs6000/rs6000.c (constant_int_to_128bit_vector): New
217 (constant_fp_to_128bit_vector): New helper function.
218 (vec_const_128bit_to_bytes): New function.
220 2021-12-15 Alexandre Oliva <oliva@adacore.com>
223 * builtins.c (try_store_by_multiple_pieces): Drop address
224 conversion to ptr_mode.
226 2021-12-15 Alexandre Oliva <oliva@adacore.com>
229 * builtins.c (try_store_by_multiple_pieces): Fail if min_len
230 is greater than max_len.
232 2021-12-14 liuhongt <hongtao.liu@intel.com>
235 * tree-ssa-ccp.c (optimize_atomic_bit_test_and): Check
236 is_gimple_assign before gimple_assign_rhs_code.
238 2021-12-14 Sören Tempel <soeren@soeren-tempel.net>
240 * ginclude/stddef.h (__DEFINED_ptrdiff_t): Add support for musl
241 libc typedef macro guard.
242 (__DEFINED_size_t): Ditto.
243 (__DEFINED_wchar_t): Ditto.
245 2021-12-14 JoJo R <rjiejie@linux.alibaba.com>
247 * regrename.c (find_rename_reg): Return satisfied regno
248 if instruction is noop move.
250 2021-12-14 Bill Schmidt <wschmidt@linux.ibm.com>
252 * config/rs6000/rs6000-c.c (altivec_build_resolved_builtin): Rename
253 rs6000_builtin_decls_x to rs6000_builtin_decls.
254 (altivec_resolve_overloaded_builtin): Likewise. Also rename
255 rs6000_builtin_info_x to rs6000_builtin_info.
256 * config/rs6000/rs6000-call.c (rs6000_invalid_builtin): Rename
257 rs6000_builtin_info_x to rs6000_builtin_info.
258 (rs6000_builtin_is_supported): Likewise.
259 (rs6000_gimple_fold_mma_builtin): Likewise. Also rename
260 rs6000_builtin_decls_x to rs6000_builtin_decls.
261 (rs6000_gimple_fold_builtin): Rename rs6000_builtin_info_x to
263 (cpu_expand_builtin): Likewise.
264 (rs6000_expand_builtin): Likewise.
265 (rs6000_init_builtins): Likewise. Also rename rs6000_builtin_decls_x
266 to rs6000_builtin_decls.
267 (rs6000_builtin_decl): Rename rs6000_builtin_decls_x to
268 rs6000_builtin_decls.
269 * config/rs6000/rs6000-gen-builtins.c (write_decls): In generated code,
270 rename rs6000_builtin_decls_x to rs6000_builtin_decls, and rename
271 rs6000_builtin_info_x to rs6000_builtin_info.
272 (write_bif_static_init): In generated code, rename
273 rs6000_builtin_info_x to rs6000_builtin_info.
274 (write_init_bif_table): In generated code, rename
275 rs6000_builtin_decls_x to rs6000_builtin_decls, and rename
276 rs6000_builtin_info_x to rs6000_builtin_info.
277 (write_init_ovld_table): In generated code, rename
278 rs6000_builtin_decls_x to rs6000_builtin_decls.
279 (write_init_file): Likewise.
280 * config/rs6000/rs6000.c (rs6000_builtin_vectorized_function):
282 (rs6000_builtin_md_vectorized_function): Likewise.
283 (rs6000_builtin_reciprocal): Likewise.
284 (add_condition_to_bb): Likewise.
285 (rs6000_atomic_assign_expand_fenv): Likewise.
287 2021-12-14 Bill Schmidt <wschmidt@linux.ibm.com>
289 * config/rs6000/rs6000-c.c (altivec_resolve_new_overloaded_builtin):
290 Remove forward declaration.
291 (rs6000_new_builtin_type_compatible): Rename to
292 rs6000_builtin_type_compatible.
293 (rs6000_builtin_type_compatible): Remove.
294 (altivec_resolve_overloaded_builtin): Remove.
295 (altivec_build_new_resolved_builtin): Rename to
296 altivec_build_resolved_builtin.
297 (altivec_resolve_new_overloaded_builtin): Rename to
298 altivec_resolve_overloaded_builtin. Remove static keyword. Adjust
299 called function names.
300 * config/rs6000/rs6000-call.c (rs6000_expand_new_builtin): Remove
302 (rs6000_gimple_fold_new_builtin): Likewise.
303 (rs6000_invalid_new_builtin): Rename to rs6000_invalid_builtin.
304 (rs6000_gimple_fold_builtin): Remove.
305 (rs6000_new_builtin_valid_without_lhs): Rename to
306 rs6000_builtin_valid_without_lhs.
307 (rs6000_new_builtin_is_supported): Rename to
308 rs6000_builtin_is_supported.
309 (rs6000_gimple_fold_new_mma_builtin): Rename to
310 rs6000_gimple_fold_mma_builtin.
311 (rs6000_gimple_fold_new_builtin): Rename to
312 rs6000_gimple_fold_builtin. Remove static keyword. Adjust called
314 (rs6000_expand_builtin): Remove.
315 (new_cpu_expand_builtin): Rename to cpu_expand_builtin.
316 (new_mma_expand_builtin): Rename to mma_expand_builtin.
317 (new_htm_spr_num): Rename to htm_spr_num.
318 (new_htm_expand_builtin): Rename to htm_expand_builtin. Change name
320 (rs6000_expand_new_builtin): Rename to rs6000_expand_builtin. Remove
321 static keyword. Adjust called function names.
322 (rs6000_new_builtin_decl): Rename to rs6000_builtin_decl. Remove
324 (rs6000_builtin_decl): Remove.
325 * config/rs6000/rs6000-gen-builtins.c (write_decls): In gnerated code,
326 rename rs6000_new_builtin_is_supported to rs6000_builtin_is_supported.
327 * config/rs6000/rs6000-internal.h (rs6000_invalid_new_builtin): Rename
328 to rs6000_invalid_builtin.
329 * config/rs6000/rs6000.c (rs6000_new_builtin_vectorized_function):
330 Rename to rs6000_builtin_vectorized_function.
331 (rs6000_new_builtin_md_vectorized_function): Rename to
332 rs6000_builtin_md_vectorized_function.
333 (rs6000_builtin_vectorized_function): Remove.
334 (rs6000_builtin_md_vectorized_function): Remove.
336 2021-12-14 Bill Schmidt <wschmidt@linux.ibm.com>
338 * config/rs6000/rs6000-builtin.def: Delete.
339 * config/rs6000/rs6000-call.c (builtin_compatibility): Delete.
340 (builtin_description): Delete.
341 (builtin_hash_struct): Delete.
342 (builtin_hasher): Delete.
343 (builtin_hash_table): Delete.
344 (builtin_hasher::hash): Delete.
345 (builtin_hasher::equal): Delete.
346 (rs6000_builtin_info_type): Delete.
347 (rs6000_builtin_info): Delete.
348 (bdesc_compat): Delete.
349 (bdesc_3arg): Delete.
350 (bdesc_4arg): Delete.
352 (bdesc_2arg): Delete.
353 (bdesc_altivec_preds): Delete.
355 (bdesc_1arg): Delete.
356 (bdesc_0arg): Delete.
359 (rs6000_overloaded_builtin_p): Delete.
360 (rs6000_overloaded_builtin_name): Delete.
361 (htm_spr_num): Delete.
362 (rs6000_builtin_is_supported_p): Delete.
363 (rs6000_gimple_fold_mma_builtin): Delete.
364 (gt-rs6000-call.h): Remove include directive.
365 * config/rs6000/rs6000-protos.h (rs6000_overloaded_builtin_p): Delete.
366 (rs6000_builtin_is_supported_p): Delete.
367 (rs6000_overloaded_builtin_name): Delete.
368 * config/rs6000/rs6000.c (rs6000_builtin_decls): Delete.
369 (rs6000_debug_reg_global): Remove reference to RS6000_BUILTIN_COUNT.
370 * config/rs6000/rs6000.h (rs6000_builtins): Delete.
371 (altivec_builtin_types): Delete.
372 (rs6000_builtin_decls): Delete.
373 * config/rs6000/t-rs6000 (TM_H): Don't add rs6000-builtin.def.
375 2021-12-14 Bill Schmidt <wschmidt@linux.ibm.com>
377 * config/rs6000/rs6000-builtin-new.def: Rename to...
378 * config/rs6000/rs6000-builtins.def: ...this.
379 * config/rs6000/rs6000-gen-builtins.c: Adjust header commentary.
380 * config/rs6000/t-rs6000 (EXTRA_GTYPE_DEPS): Rename
381 rs6000-builtin-new.def to rs6000-builtins.def.
382 (rs6000-builtins.c): Likewise.
384 2021-12-14 Bill Schmidt <wschmidt@linux.ibm.com>
386 * config/rs6000/rs6000-call.c (altivec_overloaded_builtins): Remove.
387 * config/rs6000/rs6000.h (altivec_overloaded_builtins): Remove.
389 2021-12-14 Peter Bergner <bergner@linux.ibm.com>
392 * config/rs6000/mma.md (UNSPEC_MMA_ASSEMBLE): Rename unspec from this...
393 (UNSPEC_VSX_ASSEMBLE): ...to this.
394 (UNSPECV_MMA_ASSEMBLE): New unspecv.
395 (vsx_assemble_pair): Use UNSPEC_VSX_ASSEMBLE.
396 (*vsx_assemble_pair): Likewise.
397 (mma_assemble_acc): Use UNSPECV_MMA_ASSEMBLE.
398 (*mma_assemble_acc): Likewise.
399 * config/rs6000/rs6000.c (rs6000_split_multireg_move): Handle
400 UNSPEC_VOLATILE. Use UNSPEC_VSX_ASSEMBLE and UNSPECV_MMA_ASSEMBLE.
402 2021-12-14 Uroš Bizjak <ubizjak@gmail.com>
405 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate)
406 <case E_V8HFmode>: Implement for TARGET_SSE2.
407 <case E_V16HFmode>: Implement for TARGET_AVX.
408 <case E_V32HFmode>: Implement for TARGET_AVX512F.
409 (ix86_expand_vector_set_var): Handle V32HFmode
410 without TARGET_AVX512BW.
411 (ix86_expand_vector_extract)
412 <case E_V8HFmode>: Implement for TARGET_SSE2.
413 <case E_V16HFmode>: Implement for TARGET_AVX.
414 <case E_V32HFmode>: Implement for TARGET_AVX512BW.
415 (expand_vec_perm_broadcast_1) <case E_V8HFmode>: New.
416 * config/i386/sse.md (VI12HF_AVX512VL): Remove
417 TARGET_AVX512FP16 condition.
420 (avx_vbroadcastf128_<mode>): Use V_256H mode iterator.
422 2021-12-14 Bill Schmidt <wschmidt@linux.ibm.com>
424 * config/rs6000/darwin.h (SUBTARGET_INIT_BUILTINS): Remove
425 test for new_builtins_are_live and simplify.
426 * config/rs6000/rs6000-c.c (altivec_build_resolved_builtin): Remove
428 (altivec_resolve_overloaded_builtin): Remove test for
429 new_builtins_are_live and simplify.
430 * config/rs6000/rs6000-call.c (altivec_init_builtins): Remove forward
432 (builtin_function_type): Likewise.
433 (rs6000_common_init_builtins): Likewise.
434 (htm_init_builtins): Likewise.
435 (mma_init_builtins): Likewise.
436 (def_builtin): Remove dead function.
437 (rs6000_expand_zeroop_builtin): Likewise.
438 (rs6000_expand_mtfsf_builtin): Likewise.
439 (rs6000_expand_mtfsb_builtin): Likewise.
440 (rs6000_expand_set_fpscr_rn_builtin): Likewise.
441 (rs6000_expand_set_fpscr_drn_builtin): Likewise.
442 (rs6000_expand_unop_builtin): Likewise.
443 (altivec_expand_abs_builtin): Likewise.
444 (rs6000_expand_binop_builtin): Likewise.
445 (altivec_expand_lxvr_builtin): Likewise.
446 (altivec_expand_lv_builtin): Likewise.
447 (altivec_expand_stxvl_builtin): Likewise.
448 (altivec_expand_stv_builtin): Likewise.
449 (mma_expand_builtin): Likewise.
450 (htm_expand_builtin): Likewise.
451 (cpu_expand_builtin): Likewise.
452 (rs6000_expand_quaternop_builtin): Likewise.
453 (rs6000_expand_ternop_builtin): Likewise.
454 (altivec_expand_dst_builtin): Likewise.
455 (altivec_expand_vec_sel_builtin): Likewise.
456 (altivec_expand_builtin): Likewise.
457 (rs6000_invalid_builtin): Likewise.
458 (rs6000_builtin_valid_without_lhs): Likewise.
459 (rs6000_gimple_fold_builtin): Remove test for new_builtins_are_live and
461 (rs6000_expand_builtin): Likewise.
462 (rs6000_init_builtins): Remove tests for new_builtins_are_live and
464 (rs6000_builtin_decl): Likewise.
465 (altivec_init_builtins): Remove dead function.
466 (mma_init_builtins): Likewise.
467 (htm_init_builtins): Likewise.
468 (builtin_quaternary_function_type): Likewise.
469 (builtin_function_type): Likewise.
470 (rs6000_common_init_builtins): Likewise.
471 * config/rs6000/rs6000-gen-builtins.c (write_header_file): Don't
472 declare new_builtins_are_live.
473 (write_init_bif_table): In generated code, remove test for
474 new_builtins_are_live and simplify.
475 (write_init_ovld_table): Likewise.
476 (write_init_file): Don't initialize new_builtins_are_live.
477 * config/rs6000/rs6000.c (rs6000_builtin_vectorized_function): Remove
478 test for new_builtins_are_live and simplify.
479 (rs6000_builtin_md_vectorized_function): Likewise.
480 (rs6000_builtin_reciprocal): Likewise.
481 (add_condition_to_bb): Likewise.
482 (rs6000_atomic_assign_expand_fenv): Likewise.
484 2021-12-14 Bill Schmidt <wschmidt@linux.ibm.com>
487 * config/rs6000/rs6000-builtin-new.def (__builtin_altivec_vcmpequd):
488 Move to power8-vector stanza.
489 (__builtin_altivec_vcmpequd_p): Likewise.
490 (__builtin_altivec_vcmpgtsd): Likewise.
491 (__builtin_altivec_vcmpgtsd_p): Likewise.
492 (__builtin_altivec_vcmpgtud): Likewise.
493 (__builtin_altivec_vcmpgtud_p): Likewise.
495 2021-12-14 Bill Schmidt <wschmidt@linux.ibm.com>
498 * config/rs6000/rs6000-builtin-new.def (__builtin_pack_longdouble): Add
500 (__builtin_unpack_longdouble): Likewise.
501 * config/rs6000/rs6000-call.c (rs6000_expand_new_builtin): Add special
502 handling for ibmld attribute.
503 * config/rs6000/rs6000-gen-builtins.c (attrinfo): Add isibmld.
504 (parse_bif_attrs): Handle ibmld.
505 (write_decls): Likewise.
506 (write_bif_static_init): Likewise.
508 2021-12-14 Jan Hubicka <hubicka@ucw.cz>
511 * ipa-modref-tree.c (modref_access_node::range_info_useful_p): Handle
512 MODREF_GLOBAL_MEMORY_PARM.
513 (modref_access_node::dump): Likewise.
514 (modref_access_node::get_call_arg): Likewise.
515 * ipa-modref-tree.h (enum modref_special_parms): Add
516 MODREF_GLOBAL_MEMORY_PARM.
517 (modref_access_node::useful_for_kill): Handle
518 MODREF_GLOBAL_MEMORY_PARM.
519 (modref:tree::merge): Add promote_unknown_to_global.
520 * ipa-modref.c (verify_arg):New function.
521 (may_access_nonescaping_parm_p): New function.
522 (modref_access_analysis::record_global_memory_load): New member
524 (modref_access_analysis::record_global_memory_store): Likewise.
525 (modref_access_analysis::process_fnspec): Distingush global and local
527 (modref_access_analysis::analyze_call): Likewise.
528 * tree-ssa-alias.c (ref_may_access_global_memory_p): New function.
529 (modref_may_conflict): Use it.
531 2021-12-14 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
533 * config/aarch64/aarch64-builtins.c (enum aarch64_builtins):
534 Define AARCH64_LS64_BUILTIN_LD64B, AARCH64_LS64_BUILTIN_ST64B,
535 AARCH64_LS64_BUILTIN_ST64BV, AARCH64_LS64_BUILTIN_ST64BV0.
536 (aarch64_init_ls64_builtin_decl): Helper function.
537 (aarch64_init_ls64_builtins): Helper function.
538 (aarch64_init_ls64_builtins_types): Helper function.
539 (aarch64_general_init_builtins): Init LS64 intrisics for
541 (aarch64_expand_builtin_ls64): LS64 intrinsics expander.
542 (aarch64_general_expand_builtin): Handle aarch64_expand_builtin_ls64.
543 (ls64_builtins_data): New helper struct.
544 (v8di_UP): New define.
545 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Define
547 * config/aarch64/aarch64.c (aarch64_classify_address): Enforce the
548 V8DI range (7-bit signed scaled) for both ends of the range.
549 * config/aarch64/aarch64-simd.md (movv8di): New pattern.
550 (aarch64_movv8di): New pattern.
551 * config/aarch64/aarch64.h (AARCH64_ISA_LS64): New define.
552 (TARGET_LS64): New define.
553 * config/aarch64/aarch64.md: Add UNSPEC_LD64B, UNSPEC_ST64B,
554 UNSPEC_ST64BV and UNSPEC_ST64BV0.
555 (ld64b): New define_insn.
556 (st64b): New define_insn.
557 (st64bv): New define_insn.
558 (st64bv0): New define_insn.
559 * config/aarch64/arm_acle.h (data512_t): New type derived from
561 (__arm_data512_t): New internal type.
562 (__arm_ld64b): New intrinsic.
563 (__arm_st64b): New intrinsic.
564 (__arm_st64bv): New intrinsic.
565 (__arm_st64bv0): New intrinsic.
566 * config/arm/types.md: Add new type ls64.
568 2021-12-14 Olivier Hainque <hainque@adacore.com>
570 * config/i386/t-vxworks: Drop the fPIC multilibs.
572 2021-12-14 Fred Konrad <konrad@adacore.com>
574 * config/rs6000/t-vxworks: Drop the fPIC multilib.
576 2021-12-13 Vladimir N. Makarov <vmakarov@redhat.com>
579 * ira-costs.c (record_operand_costs): Do not take pseudo class
580 calculated on the 1st iteration into account when processing move
581 involving the pseudo and a hard register.
583 2021-12-13 Roger Sayle <roger@nextmovesoftware.com>
585 * config/i386/i386.md (define_split any_or:SWI248 -> orb %?h):
586 Optimize the case where the integer constant operand is zero.
588 2021-12-13 Doug Rupp <rupp@adacore.com>
589 Olivier Hainque <hainque@adacore.com>
591 * config/vxworks.h (VXWORKS_LINK_OS_SPEC): New spec.
592 (VXWORKS_BASE_LINK_SPEC): New spec, using the former.
593 (VXWORKS_EXTRA_LINK_SPEC): New spec for old and new VxWorks.
594 (VXWORKS_LINK_SPEC): Combo of BASE and EXTRA specs.
595 * config/rs6000/vxworks.h (VXWORKS_LINK_OS_SPEC): Empty.
596 (LINK_OS_EXTRA_SPEC32): Use VXWORKS_LINK_SPEC.
597 (LINK_OS_EXTRA_SPEC64): Likewise.
599 2021-12-13 Fred Konrad <konrad@adacore.com>
600 Olivier Hainque <hainque@adacore.com>
602 * config/vxworks.h (VXWORKS_BASE_LIBS_RTP): Guard -lc_internal
603 on !shared+!non-static and document.
604 (VXWORKS_LIB_SPEC): Remove the bits intended to drag the
605 init/fini functions from libc_internal in the shared lib case.
606 (VX_CRTBEGIN_SPEC/VX_CRTEND_SPEC): Use vxcrtstuff objects also in
607 configurations with shared lib and INITFINI_ARRAY support.
609 2021-12-13 Fred Konrad <konrad@adacore.com>
610 Olivier Hainque <hainque@adacore.com>
612 * config/vx-common.h: Define REAL_LIBGCC_SPEC since the
613 '-non-static' option is not standard.
614 * config/vxworks.h (VXWORKS_LIBGCC_SPEC): Implement the LIBGCC_SPEC
615 since REAL_LIBGCC_SPEC is used now.
616 (STARTFILE_PREFIX_SPEC): Use the PIC VSB when building shared libraries
617 or non-static binaries.
619 2021-12-13 Jan Hubicka <hubicka@ucw.cz>
621 * common.opt: Add -fipa-strict-aliasing.
622 * doc/invoke.texi: Document -fipa-strict-aliasing.
623 * ipa-modref.c (modref_access_analysis::record_access): Honor
624 -fipa-strict-aliasing.
625 (modref_access_analysis::record_access_lto): Likewise.
627 2021-12-13 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
629 * config/aarch64/aarch64-arches.def (armv8.8-a): Define.
630 * config/aarch64/aarch64.h (AARCH64_FL_V8_8): Define.
631 (AARCH64_FL_FOR_ARCH8_8): Define.
632 * doc/invoke.texi: Document -march=armv8.8-a.
634 2021-12-13 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
636 * config/aarch64/aarch64.c (aarch64_expand_setmem_mops): Define.
637 (aarch64_expand_setmem): Adjust for TARGET_MOPS.
638 * config/aarch64/aarch64.h (CLEAR_RATIO): Adjust for TARGET_MOPS.
639 (SET_RATIO): Likewise.
640 * config/aarch64/aarch64.md ("unspec"): Add UNSPEC_SETMEM.
641 (aarch64_setmemdi): Define.
642 (setmemdi): Adjust for TARGET_MOPS.
643 * config/aarch64/aarch64.opt (aarch64-mops-memset-size-threshold):
646 2021-12-13 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
648 * config/aarch64/aarch64.md (aarch64_movmemdi): Define.
650 (unspec): Add UNSPEC_MOVMEM.
651 * config/aarch64/aarch64.opt (aarch64-mops-memmove-size-threshold):
654 2021-12-13 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
656 * config/aarch64/aarch64-option-extensions.def (mops): Define.
657 * config/aarch64/aarch64.c (aarch64_expand_cpymem_mops): Define.
658 (aarch64_expand_cpymem): Define.
659 * config/aarch64/aarch64.h (AARCH64_FL_MOPS): Define.
660 (AARCH64_ISA_MOPS): Define.
661 (TARGET_MOPS): Define.
662 (MOVE_RATIO): Adjust for TARGET_MOPS.
663 * config/aarch64/aarch64.md ("unspec"): Add UNSPEC_CPYMEM.
664 (aarch64_cpymemdi): New pattern.
665 (cpymemdi): Adjust for TARGET_MOPS.
666 * config/aarch64/aarch64.opt (aarch64-mops-memcpy-size-threshol):
668 * doc/invoke.texi (AArch64 Options): Document +mops.
670 2021-12-13 Martin Liska <mliska@suse.cz>
673 * ipa-inline.c (can_inline_edge_p): Move logic checking
674 no_profile_instrument_function logic to ...
675 (can_early_inline_edge_p): ... here.
677 2021-12-13 Olivier Hainque <hainque@adacore.com>
679 * config/vxworks/_yvals.h: #include yvals.h also if
682 2021-12-13 Olivier Hainque <hainque@adacore.com>
684 * config/vxworks.h (VXWORKS_OS_CPP_BUILTINS): Define
687 2021-12-13 Olivier Hainque <hainque@adacore.com>
689 * config/t-vxworks: Clear NATIVE_SYSTEM_HEADER_DIR.
690 * config/vxworks.h (SYSROOT_HEADERS_SUFFIX_SPEC): Define, for
691 VxWorks 7 and earlier.
692 (VXWORKS_ADDITIONAL_CPP_SPEC): Simplify accordingly.
693 (STARTFILE_PREFIX_SPEC): Adjust accordingly.
694 * config/rs6000/vxworks.h (STARTFILE_PREFIX_SPEC): Adjust.
696 2021-12-13 Martin Liska <mliska@suse.cz>
698 * doc/extend.texi: Use @item for the first @itemx entry.
700 2021-12-13 Jakub Jelinek <jakub@redhat.com>
702 * machmode.h (gt_pch_nx): Use gt_pointer_operator as type of second
703 argument instead of equivalent void (*) (void *, void *, void *).
704 * poly-int.h (gt_pch_nx): Likewise.
705 * wide-int.h (gt_pch_nx): Likewise.
706 * config/aarch64/aarch64-sve-builtins.cc (gt_pch_nx): Likewise.
708 2021-12-13 Jan Hubicka <hubicka@ucw.cz>
711 * ipa-fnsummary.c (evaluate_conditions_for_known_args): Do not ICE
712 on ternary expression.
714 2021-12-13 Kewen Lin <linkw@linux.ibm.com>
717 * attribs.c (decl_attributes): Check if target options change and
718 create one node if so.
720 2021-12-12 Jonathan Wakely <jwakely@redhat.com>
722 * Makefile.in: Remove unique-ptr-tests.o.
723 * selftest-run-tests.c (selftest::run_tests): Remove
724 unique_ptr_tests_cc_tests.
725 * selftest.h (unique_ptr_tests_cc_tests): Remove.
726 * system.h: Check INCLUDE_MEMORY instead of INCLUDE_UNIQUE_PTR
727 and include <memory> instead of "unique-ptr.h".
728 * unique-ptr-tests.cc: Removed.
730 2021-12-12 Roger Sayle <roger@nextmovesoftware.com>
732 * config/nvptx/nvptx-opts.h (ptx_isa): Add PTX_ISA_SM53 ISA level
734 * config/nvptx/nvptx.opt: Add sm_53 to -misa.
735 * config/nvptx/nvptx-modes.def: Add support for HFmode.
736 * config/nvptx/nvptx.h (TARGET_SM53):
737 New helper macro to conditionalize functionality on target ISA.
738 * config/nvptx/nvptx-c.c (nvptx_cpu_cpp_builtins): Add __PTX_SM__
739 support for the new ISA levels.
740 * config/nvptx/nvptx.c (nvtx_ptx_type_from_mode): Support new HFmode
741 with the ".f16" suffix/qualifier.
742 (nvptx_file_start): Add support for TARGET_SM53.
743 (nvptx_omp_device_kind_arch_isa): Add support for TARGET_SM53
744 and tweak TARGET_SM35.
745 (nvptx_scalar_mode_supported_p): Target hook with conditional
746 HFmode support on TARGET_SM53 and higher.
747 (nvptx_libgcc_floating_mode_supported_p): Likewise.
748 (TARGET_SCALAR_MODE_SUPPORTED_P): Use nvptx_scalar_mode_supported_p.
749 (TARGET_LIBGCC_FLOATING_MODE_SUPPORTED_P): Likewise, use new hook.
750 * config/nvptx/nvptx.md (*movhf_insn): New define_insn.
751 (movhf): New define_expand for HFmode moves.
752 (addhf3, subhf3, mulhf, extendhf<mode>2, trunc<mode>hf2): New
753 instructions conditional on TARGET_SM53 (i.e. -misa=sm_53).
755 2021-12-12 Jan Hubicka <hubicka@ucw.cz>
758 * ipa-modref.c (modref_access_analysis::analyze): Terminate BB
759 analysis on NULL memory access.
760 * ipa-pure-const.c (analyze_function): Likewise.
762 2021-12-11 Jan Hubicka <hubicka@ucw.cz>
764 * ipa-profile.c (ipa_profile): Do not update hot bb threshold.
766 2021-12-11 Jan Hubicka <hubicka@ucw.cz>
768 * ipa-modref.c (get_modref_function_summary): Use ultimate_alias_target.
769 (ignore_edge): Likewise.
770 (compute_parm_map): Likewise.
771 (modref_propagate_in_scc): Likewise.
772 (modref_propagate_flags_in_scc): Likewise.
774 2021-12-10 Jason Merrill <jason@redhat.com>
776 * symtab.c (symtab_node::equal_address_to): Fix comment typo.
778 2021-12-10 Doug Rupp <rupp@adacore.com>
780 * config/vxworks.h (LINK_SPEC): Remove %(link_target).
781 Change %{v:-v} to %{v:-V}.
783 2021-12-10 Olivier Hainque <hainque@adacore.com>
785 * config/t-vxworks: Remove assignment to STMP_FIXINC.
787 2021-12-10 Martin Liska <mliska@suse.cz>
789 * params.opt: Add missing dot.
791 2021-12-10 Roger Sayle <roger@nextmovesoftware.com>
794 * ipa-modref-tree.h (useful_for_kill_p): Zero width accesses aren't
795 useful for kill tracking.
797 2021-12-10 Andrew Stubbs <ams@codesourcery.com>
799 * config/gcn/mkoffload.c (process_asm): Process the variable table
800 completely differently.
801 (process_obj): Encode the varaible data differently.
803 2021-12-10 Joel Hutton <joel.hutton@arm.com>
805 PR tree-optimization/103523
806 * tree-vect-loop.c (vectorizable_induction): Check for
807 PLUS_EXPR/MINUS_EXPR support.
809 2021-12-10 Cui,Lili <lili.cui@intel.com>
811 * config/i386/i386.c (ix86_vector_costs::add_stmt_cost): Remove Tremont.
813 2021-12-09 Jan Hubicka <hubicka@ucw.cz>
815 * doc/invoke.texi (max-inline-functions-called-once-loop-depth,
816 max-inline-functions-called-once-insns): New parameters.
817 * ipa-inline.c (check_callers): Handle
818 param_inline_functions_called_once_loop_depth and
819 param_inline_functions_called_once_insns.
820 (edge_badness): Fix linebreaks.
821 * params.opt (param=max-inline-functions-called-once-loop-depth,
822 param=max-inline-functions-called-once-insn): New params.
824 2021-12-09 Martin Sebor <msebor@redhat.com>
826 PR tree-optimization/103215
827 * pointer-query.cc (access_ref::merge_ref): Extend the offset and
828 size of the merged object instead of using the larger.
830 2021-12-09 Martin Sebor <msebor@redhat.com>
833 * doc/extend.texi (attribute access): Adjust.
834 * gimple-ssa-warn-access.cc (pass_waccess::maybe_check_access_sizes):
835 Treat access mode none on a void* argument as expecting as few as
838 2021-12-09 Jakub Jelinek <jakub@redhat.com>
841 * config/aarch64/aarch64-sve-builtins.cc (gt_pch_nx): Change type of
842 second argument from function with 2 pointer arguments to function
843 with 3 pointer arguments.
845 2021-12-09 Olivier Hainque <hainque@adacore.com>
847 * config/aarch64/aarch64-vxworks.h (TARGET_OS_CPP_BUILTINS):
848 Use VX_CPU_PREFIX in CPU definitions.
850 2021-12-09 Martin Sebor <msebor@redhat.com>
852 * pointer-query.cc (access_ref::dump): Define new function
853 (pointer_query::dump): Call it.
854 * pointer-query.h (access_ref::dump): Declare new function.
856 2021-12-09 Martin Sebor <msebor@redhat.com>
858 * pointer-query.cc (compute_objsize_r): Add an argument.
859 (gimple_call_return_array): Pass a new argument to compute_objsize_r.
860 (access_ref::merge_ref): Same.
861 (access_ref::inform_access): Add an argument and use it.
862 (access_data::access_data): Initialize new member.
863 (handle_min_max_size): Pass a new argument to compute_objsize_r.
864 (handle_decl): New function.
865 (handle_array_ref): Pass a new argument to compute_objsize_r.
866 Avoid incrementing deref.
867 (set_component_ref_size): New function.
868 (handle_component_ref): New function.
869 (handle_mem_ref): Pass a new argument to compute_objsize_r.
870 Only increment deref after successfully computing object size.
871 (handle_ssa_name): New function.
872 (compute_objsize_r): Move code into helpers and call them.
873 (compute_objsize): Pass a new argument to compute_objsize_r.
874 * pointer-query.h (access_ref::inform_access): Add an argument.
875 (access_data::ostype): New member.
877 2021-12-09 Martin Sebor <msebor@redhat.com>
879 * pointer-query.cc (access_ref::merge_ref): Define new function.
880 (access_ref::get_ref): Move code into merge_ref and call it.
881 * pointer-query.h (access_ref::merge_ref): Declare new function.
883 2021-12-09 Martin Sebor <msebor@redhat.com>
885 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Pass
886 GIMPLE statement to compute_objsize.
887 * pointer-query.cc (compute_objsize): Add a statement argument.
888 * pointer-query.h (compute_objsize): Define a new overload.
890 2021-12-09 Martin Sebor <msebor@redhat.com>
892 * gimple-ssa-warn-access.cc (check_access): Adjust to member name
894 (pass_waccess::check_strncmp): Same.
895 * pointer-query.cc (access_ref::access_ref): Remove arguments.
897 (access_data::access_data): Define new ctors.
898 (access_data::set_bound): Define new member function.
899 (compute_objsize_r): Remove unnecessary code.
900 * pointer-query.h (struct access_ref): Remove ctor arguments.
901 (struct access_data): Declare ctor overloads.
902 (access_data::dst_bndrng): New member.
903 (access_data::src_bndrng): New member.
905 2021-12-09 Martin Sebor <msebor@redhat.com>
908 * pointer-query.cc (gimple_call_return_array): Call compute_objsize_r.
910 2021-12-09 Olivier Hainque <hainque@adacore.com>
911 Rasmus Villemoes <rv@rasmusvillemoes.dk>
913 * Makefile.in (T_STDINT_GCC_H): New variable, path to
914 stdint-gcc.h that a target configuration may override when
915 use_gcc_stdint is "provide".
916 (stmp-int-hdrs): Depend on it and copy that for
918 * config.gcc (vxworks): Revert to use_gcc_stdint=provide.
919 * config/t-vxworks (T_STDINT_GCC_H): Define, as vxw-stdint-gcc.h.
920 (vxw-stdint-gcc.h): New target, produced from the original
922 (vxw-glimits.h): Use an automatic variable to designate the
923 first and only prerequisite.
924 * config/vxworks/stdint.h: Remove.
926 2021-12-09 Iain Sandoe <iain@sandoe.co.uk>
929 * config/host-darwin.c (SAFE_ALLOC_SIZE): Remove.
930 (darwin_gt_pch_get_address): Rework for relocatable PCH.
931 (darwin_gt_pch_use_address): Likewise.
933 2021-12-09 Jakub Jelinek <jakub@redhat.com>
936 * config/host-darwin.c (darwin_gt_pch_use_address): When reading
937 manually the file into mapped area, update mapped_addr as
938 an automatic variable rather than addr which is a reference parameter.
939 * config/host-hpux.c (hpux_gt_pch_use_address): When reading
940 manually the file into mapped area, update addr as
941 an automatic variable rather than base which is a reference parameter.
943 2021-12-09 Jakub Jelinek <jakub@redhat.com>
946 * coretypes.h (gt_pointer_operator): Use 3 pointer arguments instead
948 * gengtype.c (struct walk_type_data): Add in_nested_ptr argument.
949 (walk_type): Temporarily set d->in_nested_ptr around nested_ptr
951 (write_types_local_user_process_field): Pass a new middle pointer
952 to gt_pointer_operator op calls, if d->in_nested_ptr pass there
953 address of d->prev_val[2], otherwise NULL.
954 (write_types_local_process_field): Likewise.
955 * ggc-common.c (relocate_ptrs): Add real_ptr_p argument. If equal
956 to ptr_p, do nothing, otherwise if NULL remember ptr_p's
957 or if non-NULL real_ptr_p's corresponding new address in
959 (reloc_addrs_vec): New variable.
960 (compare_ptr, read_uleb128, write_uleb128): New functions.
961 (gt_pch_save): When iterating over objects through relocate_ptrs,
962 save current i into state.ptrs_i. Sort reloc_addrs_vec and emit
963 it as uleb128 of differences between pointer addresses into the
965 (gt_pch_restore): Allow restoring of PCH to a different address
966 than the preferred one, in that case adjust global pointers by bias
967 and also adjust by bias addresses read from the relocation table
968 as uleb128 differences. Otherwise fseek over it. Perform
969 gt_pch_restore_stringpool only after adjusting callbacks and for
970 callback adjustments also take into account the bias.
971 (default_gt_pch_use_address): Change type of first argument from
973 (mmap_gt_pch_use_address): Likewise.
974 * ggc-tests.c (gt_pch_nx): Pass NULL as new middle argument to op.
975 * hash-map.h (hash_map::pch_nx_helper): Likewise.
976 (gt_pch_nx): Likewise.
977 * hash-set.h (gt_pch_nx): Likewise.
978 * hash-table.h (gt_pch_nx): Likewise.
979 * hash-traits.h (ggc_remove::pch_nx): Likewise.
980 * hosthooks-def.h (default_gt_pch_use_address): Change type of first
981 argument from void * to void *&.
982 (mmap_gt_pch_use_address): Likewise.
983 * hosthooks.h (struct host_hooks): Change type of first argument of
984 gt_pch_use_address hook from void * to void *&.
985 * machmode.h (gt_pch_nx): Expect a callback with 3 pointers instead of
986 two in the middle argument.
987 * poly-int.h (gt_pch_nx): Likewise.
988 * stringpool.c (gt_pch_nx): Pass NULL as new middle argument to op.
989 * tree-cfg.c (gt_pch_nx): Likewise, except for LOCATION_BLOCK pass
990 the same &(block) twice.
991 * value-range.h (gt_pch_nx): Pass NULL as new middle argument to op.
992 * vec.h (gt_pch_nx): Likewise.
993 * wide-int.h (gt_pch_nx): Likewise.
994 * config/host-darwin.c (darwin_gt_pch_use_address): Change type of
995 first argument from void * to void *&.
996 * config/host-darwin.h (darwin_gt_pch_use_address): Likewise.
997 * config/host-hpux.c (hpux_gt_pch_use_address): Likewise.
998 * config/host-linux.c (linux_gt_pch_use_address): Likewise. If
999 it couldn't succeed to mmap at the preferred location, set base
1000 to the actual one. Update addr in the manual reading loop instead of
1002 * config/host-netbsd.c (netbsd_gt_pch_use_address): Change type of
1003 first argument from void * to void *&.
1004 * config/host-openbsd.c (openbsd_gt_pch_use_address): Likewise.
1005 * config/host-solaris.c (sol_gt_pch_use_address): Likewise.
1006 * config/i386/host-mingw32.c (mingw32_gt_pch_use_address): Likewise.
1007 * config/rs6000/rs6000-gen-builtins.c (write_init_file): Pass NULL
1008 as new middle argument to op in the generated code.
1009 * doc/gty.texi: Adjust samples for the addition of middle pointer
1010 to gt_pointer_operator callback.
1012 2021-12-09 Alexandre Oliva <oliva@adacore.com>
1015 * reg-stack.c (convert_regs_1): Move any_malformed_asm
1017 (reg_to_stack): ... here.
1019 2021-12-09 Alexandre Oliva <oliva@adacore.com>
1022 * expr.c (emit_move_multi_word): Skip clobber during lra.
1024 2021-12-09 Alexandre Oliva <oliva@adacore.com>
1026 PR tree-optimization/103024
1027 PR middle-end/103530
1028 * gimple-harden-conditionals.cc (non_eh_succ_edge): New.
1029 (pass_harden_compares::execute): Accept 1-bit integral types,
1030 and cope with throwing compares.
1032 2021-12-08 Iain Sandoe <iain@sandoe.co.uk>
1034 * config/darwin.h (DARWIN_PIE_SPEC): Add -no_pie when
1035 linking mdynamic-no-pic code on macOS > 10.7.
1037 2021-12-08 Dimitar Dimitrov <dimitar@dinux.eu>
1039 * config/pru/pru.c (pru_section_type_flags): New function.
1040 (TARGET_SECTION_TYPE_FLAGS): Wire it.
1042 2021-12-08 David Faust <david.faust@oracle.com>
1044 * config/bpf/bpf.c (handle_attr_preserve): Avoid calling
1045 is_gimple_assign with a NULL pointer.
1047 2021-12-08 Martin Liska <mliska@suse.cz>
1049 * profile-count.c (profile_count::dump): Add function
1050 that can dump to a provided buffer.
1051 (profile_probability::dump): Likewise.
1052 * profile-count.h: Likewise.
1053 * tree-ssa-loop-unswitch.c (tree_unswitch_single_loop):
1054 Use dump_printf_loc infrastructure.
1055 (tree_unswitch_outer_loop): Likewise.
1056 (find_loop_guard): Likewise.
1057 (hoist_guard): Likewise.
1059 2021-12-08 Chung-Lin Tang <cltang@codesourcery.com>
1061 * gimplify.c (extract_base_bit_offset): Add 'tree *offsetp' parameter,
1062 accomodate case where 'offset' return of get_inner_reference is
1064 (is_or_contains_p): Further robustify conditions.
1065 (omp_target_reorder_clauses): In alloc/to/from sorting phase, also
1066 move following GOMP_MAP_ALWAYS_POINTER maps along. Add new sorting
1067 phase where we make sure pointers with an attach/detach map are ordered
1069 (gimplify_scan_omp_clauses): Add modifications to avoid creating
1070 GOMP_MAP_STRUCT and associated alloc map for attach/detach maps.
1072 2021-12-08 Roger Sayle <roger@nextmovesoftware.com>
1074 * config/nvptx/nvptx.md (*extend_trunc_<mode>2_qi,
1075 *extend_trunc_<mode>2_hi, *extend_trunc_di2_si): New insns.
1076 Use cvt to perform sign-extension of truncation in one step.
1078 2021-12-08 Chung-Lin Tang <cltang@codesourcery.com>
1081 * gimplify.c ("tree-hash-traits.h"): Add include.
1082 (gimplify_scan_omp_clauses): Change struct_map_to_clause to type
1083 hash_map<tree_operand, tree> *. Adjust struct map handling to handle
1084 cases of *A and A->B expressions. Under !DECL_P case of
1085 GOMP_CLAUSE_MAP handling, add STRIP_NOPS for indir_p case, add to
1086 struct_deref_set for map(*ptr_to_struct) cases. Add MEM_REF case when
1087 handling component_ref_p case. Add unshare_expr and gimplification
1088 when created GOMP_MAP_STRUCT is not a DECL. Add code to add
1089 firstprivate pointer for *pointer-to-struct case.
1090 (gimplify_adjust_omp_clauses): Move GOMP_MAP_STRUCT removal code for
1091 exit data directives code to earlier position.
1092 * omp-low.c (lower_omp_target):
1093 Handle GOMP_MAP_ATTACH_ZERO_LENGTH_ARRAY_SECTION, and
1094 GOMP_MAP_POINTER_TO_ZERO_LENGTH_ARRAY_SECTION map kinds.
1095 * tree-pretty-print.c (dump_omp_clause): Likewise.
1097 2021-12-08 Andrew Stubbs <ams@codesourcery.com>
1098 Hafiz Abid Qadeer <abidh@codesourcery.com>
1100 * dwarf2cfi.c (dw_stack_pointer_regnum): Change type to struct cfa_reg.
1101 (dw_frame_pointer_regnum): Likewise.
1102 (new_cfi_row): Use set_by_dwreg.
1103 (get_cfa_from_loc_descr): Use set_by_dwreg. Support register spans.
1104 handle DW_OP_bregx with DW_OP_breg{0-31}. Support DW_OP_lit*,
1105 DW_OP_const*, DW_OP_minus, DW_OP_shl and DW_OP_plus.
1106 (lookup_cfa_1): Use set_by_dwreg.
1107 (def_cfa_0): Update for cfa_reg and support register spans.
1108 (reg_save): Change sreg parameter to struct cfa_reg. Support register
1110 (dwf_cfa_reg): New function.
1111 (dwarf2out_flush_queued_reg_saves): Use dwf_cfa_reg instead of
1113 (dwarf2out_frame_debug_def_cfa): Likewise.
1114 (dwarf2out_frame_debug_adjust_cfa): Likewise.
1115 (dwarf2out_frame_debug_cfa_offset): Likewise. Update reg_save usage.
1116 (dwarf2out_frame_debug_cfa_register): Likewise.
1117 (dwarf2out_frame_debug_expr): Likewise.
1118 (create_pseudo_cfg): Use set_by_dwreg.
1119 (initial_return_save): Use set_by_dwreg and dwf_cfa_reg,
1120 (create_cie_data): Use dwf_cfa_reg.
1121 (execute_dwarf2_frame): Use dwf_cfa_reg.
1122 (dump_cfi_row): Use set_by_dwreg.
1123 * dwarf2out.c (build_span_loc, build_breg_loc): New function.
1124 (build_cfa_loc): Support register spans.
1125 (build_cfa_aligned_loc): Update cfa_reg usage.
1126 (convert_cfa_to_fb_loc_list): Use set_by_dwreg.
1127 * dwarf2out.h (struct cfa_reg): New type.
1128 (struct dw_cfa_location): Use struct cfa_reg.
1129 (build_span_loc): New prototype.
1131 2021-12-08 Haochen Jiang <haochen.jiang@intel.com>
1134 * config/i386/sse.md
1135 (*<sse4_1>_blendv<ssefltmodesuffix><avxsizesuffix>_not_ltint):
1136 Add new define_insn_and_split.
1138 2021-12-08 Alexandre Oliva <oliva@adacore.com>
1140 PR middle-end/103149
1141 * gimple-harden-conditionals.cc (detach_value): Use memory if
1142 general regs won't do.
1144 2021-12-07 Martin Liska <mliska@suse.cz>
1146 PR middle-end/103438
1147 * config/s390/s390.c (s390_valid_target_attribute_inner_p):
1148 Use new enum CLVC_INTEGER.
1149 * opt-functions.awk: Use new CLVC_INTEGER.
1150 * opts-common.c (set_option): Likewise.
1151 (option_enabled): Return -1,0,1 for CLVC_INTEGER.
1152 (get_option_state): Use new CLVC_INTEGER.
1153 (control_warning_option): Likewise.
1154 * opts.h (enum cl_var_type): Likewise.
1156 2021-12-07 H.J. Lu <hjl.tools@gmail.com>
1159 * config/i386/i386.c (ix86_call_use_plt_p): Check FUNCTION_DECL
1160 before calling cgraph_node::get.
1162 2021-12-07 Richard Biener <rguenther@suse.de>
1164 PR tree-optimization/103596
1165 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_stmt):
1166 Note we are not propagating into a PHI argument to may_propagate_copy.
1167 * tree-ssa-propagate.h (may_propagate_copy): Add
1168 argument specifying whether we propagate into a PHI arg.
1169 * tree-ssa-propagate.c (may_propagate_copy): Likewise.
1170 When not doing so we can replace an abnormal with
1172 (may_propagate_into_stmt): Update may_propagate_copy calls.
1173 (replace_exp_1): Move propagation checking code to
1174 propagate_value and rename to ...
1175 (replace_exp): ... this and elide previous wrapper.
1176 (propagate_value): Perform checking with adjusted
1177 may_propagate_copy call and dispatch to replace_exp.
1179 2021-12-07 Matthias Kretz <m.kretz@gsi.de>
1181 * hash-map.h (hash_map::traverse): Let both overloads behave the
1183 * predict.c (assert_is_empty): Return true, thus not changing
1186 2021-12-07 YunQiang Su <yunqiang.su@cipunited.com>
1188 * config/mips/mips.h (ISA_HAS_UNALIGNED_ACCESS, STRICT_ALIGNMENT):
1189 R6 can unaligned access.
1190 * config/mips/mips.md (movmisalign<mode>): Likewise.
1191 * config/mips/mips.opt: add -m(no-)unaligned-access
1192 * doc/invoke.texi: Likewise.
1194 2021-12-07 Eugene Rozenfeld <erozen@microsoft.com>
1196 * auto-profile.c (afdo_propagate_edge): Improve count propagation algorithm.
1198 2021-12-06 Paul A. Clarke <pc@us.ibm.com>
1201 * config/rs6000/xmmintrin.h (_mm_movemask_ps): Replace "vector" with
1204 2021-12-06 Jose E. Marchesi <jose.marchesi@oracle.com>
1206 * config/bpf/bpf.c (bpf_handle_preserve_access_index_attribute):
1207 Mark arguments `args' and flags' as unused.
1208 (bpf_core_newdecl): Remove unused local `newdecl'.
1209 (bpf_core_newdecl): Remove unused argument `loc'.
1210 (ctfc_debuginfo_early_finish_p): Remove unused function.
1211 (TARGET_CTFC_DEBUGINFO_EARLY_FINISH_P): Remove definition.
1212 (bpf_core_walk): Do not pass a location to bpf_core_newdecl.
1214 2021-12-06 Richard Sandiford <richard.sandiford@arm.com>
1216 * gimple-range-edge.cc (gimple_outgoing_range::edge_range_p): Add
1217 a shortcut for blocks with single successors.
1218 * gimple-range-gori.cc (gori_map::calculate_gori): Likewise.
1220 2021-12-06 Richard Sandiford <richard.sandiford@arm.com>
1222 * value-range.cc (irange::irange_union): Use quick_push rather
1223 than safe_push. Use widest_int rather than wide_int. Avoid
1224 assigning wi::to_* results to wide*_int temporaries.
1226 2021-12-06 Andrew MacLeod <amacleod@redhat.com>
1228 * gimple-range-cache.cc (ranger_cache::fill_block_cache): Check for
1229 a range from dominators before filling the cache.
1230 (ranger_cache::range_from_dom): New.
1231 * gimple-range-cache.h (ranger_cache::range_from_dom): Add prototype.
1233 2021-12-06 Andrew MacLeod <amacleod@redhat.com>
1235 * gimple-range-gori.h (class gori_compute):: Add prototypes.
1236 * gimple-range-gori.cc (gori_compute::has_edge_range_p): Add alternate
1237 API for basic block. Call for edge alterantive.
1238 (gori_compute::may_recompute_p): Ditto.
1240 2021-12-06 Richard Biener <rguenther@suse.de>
1242 PR tree-optimization/103581
1243 * tree-vect-stmts.c (vect_build_gather_load_calls): Properly
1244 guard all the AVX512 mask cases.
1246 2021-12-06 Richard Biener <rguenther@suse.de>
1248 PR tree-optimization/103544
1249 * tree-vect-slp.c (vect_analyze_slp): Only add a SLP reduction
1250 opportunity if the stmt in question is the reduction root.
1251 (dot_slp_tree): Add missing check for NULL child.
1253 2021-12-06 Jakub Jelinek <jakub@redhat.com>
1256 * config/avr/avr.c (avr_output_data_section_asm_op,
1257 avr_output_bss_section_asm_op): Change argument type from const void *
1260 2021-12-06 Tamar Christina <tamar.christina@arm.com>
1262 PR rtl-optimization/103404
1263 * cse.c (find_sets_in_insn): Don't select elements out of a V1 mode
1266 2021-12-06 Hongtao Liu <Hongtao.liu@intel.com>
1267 Uroš Bizjak <ubizjak@gmail.com>
1270 * config/i386/i386.c (ix86_preferred_reload_class): Allow
1271 integer regs when moves between register units are cheap.
1272 * config/i386/i386.h (INT_SSE_CLASS_P): New.
1274 2021-12-05 David Edelsohn <dje.gcc@gmail.com>
1276 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Move
1277 AIX math builtin initialization before new_builtins_are_live.
1279 2021-12-04 Jakub Jelinek <jakub@redhat.com>
1281 * config/i386/x86-tune.def (X86_TUNE_PARTIAL_REG_DEPENDENCY): Fix
1282 comment typo, Preffer -> prefer.
1283 * ipa-modref-tree.c (modref_access_node::closer_pair_p): Likewise.
1285 2021-12-04 Alexandre Oliva <oliva@adacore.com>
1287 PR rtl-optimization/103028
1288 * ifcvt.c (find_cond_trap): Validate new insns more strictly
1291 2021-12-03 Martin Liska <mliska@suse.cz>
1292 Peter Bergner <bergner@linux.ibm.com>
1295 * config/rs6000/rs6000.c (rs6000_option_override_internal): Move the
1296 disabling of shrink-wrapping when using -mrop-protect from here...
1297 (rs6000_override_options_after_change): ...to here.
1299 2021-12-03 H.J. Lu <hjl.tools@gmail.com>
1302 * config/i386/i386-expand.c (ix86_expand_builtin): Pass PVW_NONE
1303 and PVW_NONE to ix86_target_string.
1304 * config/i386/i386-options.c (ix86_target_string): Add arguments
1305 for move_max and store_max.
1306 (ix86_target_string::add_vector_width): New lambda.
1307 (ix86_debug_options): Pass ix86_move_max and ix86_store_max to
1309 (ix86_function_specific_print): Pass ptr->x_ix86_move_max and
1310 ptr->x_ix86_store_max to ix86_target_string.
1311 (ix86_valid_target_attribute_tree): Handle x_ix86_move_max and
1313 (ix86_option_override_internal): Set the default x_ix86_move_max
1314 and x_ix86_store_max.
1315 * config/i386/i386-options.h (ix86_target_string): Add
1316 prefer_vector_width and prefer_vector_width.
1317 * config/i386/i386.h (TARGET_AVX256_MOVE_BY_PIECES): Removed.
1318 (TARGET_AVX256_STORE_BY_PIECES): Likewise.
1319 (MOVE_MAX): Use 64 if ix86_move_max or ix86_store_max ==
1320 PVW_AVX512. Use 32 if ix86_move_max or ix86_store_max >=
1322 (STORE_MAX_PIECES): Use 64 if ix86_store_max == PVW_AVX512.
1323 Use 32 if ix86_store_max >= PVW_AVX256.
1324 * config/i386/i386.opt: Add -mmove-max=bits and -mstore-max=bits.
1325 * config/i386/x86-tune.def (X86_TUNE_AVX512_MOVE_BY_PIECES): New.
1326 (X86_TUNE_AVX512_STORE_BY_PIECES): Likewise.
1327 * doc/invoke.texi: Document -mmove-max=bits and -mstore-max=bits.
1329 2021-12-03 Bill Schmidt <wschmidt@linux.ibm.com>
1331 * config/rs6000/rs6000.c (rs6000_builtin_reciprocal): Fix builtin
1334 2021-12-03 SiYu Wu <siyu@isrc.iscas.ac.cn>
1336 * common/config/riscv/riscv-common.c (riscv_implied_info):
1337 Add K-ext related entry.
1338 (riscv_supported_std_ext): Add 'k'.
1339 * config/riscv/arch-canonicalize (CANONICAL_ORDER): Add 'k'.
1340 (IMPLIED_EXT): Add K-ext related entry.
1342 2021-12-03 SiYu Wu <siyu@isrc.iscas.ac.cn>
1344 * common/config/riscv/riscv-common.c (riscv_ext_version_table):
1346 * config/riscv/riscv-opts.h (MASK_ZBKB): New.
1353 (MASK_ZKSED): Ditto.
1356 (TARGET_ZBKB): Ditto.
1357 (TARGET_ZBKC): Ditto.
1358 (TARGET_ZBKX): Ditto.
1359 (TARGET_ZKNE): Ditto.
1360 (TARGET_ZKND): Ditto.
1361 (TARGET_ZKNH): Ditto.
1362 (TARGET_ZKR): Ditto.
1363 (TARGET_ZKSED): Ditto.
1364 (TARGET_ZKSH): Ditto.
1365 (TARGET_ZKT): Ditto.
1366 * config/riscv/riscv.opt (riscv_zk_subext): New.
1368 2021-12-03 Tamar Christina <tamar.christina@arm.com>
1370 * tree-vect-stmts.c (prepare_load_store_mask): Rename to...
1371 (prepare_vec_mask): ...This and record operations that have already been
1373 (vectorizable_call): Use it.
1374 (vectorizable_operation): Likewise.
1375 (vectorizable_store): Likewise.
1376 (vectorizable_load): Likewise.
1377 * tree-vectorizer.h (class _loop_vec_info): Add vec_cond_masked_set.
1378 (vec_cond_masked_set_type, tree_cond_mask_hash): New.
1380 2021-12-03 H.J. Lu <hjl.tools@gmail.com>
1384 * target.def (ifunc_ref_local_ok): Add a target hook.
1385 * varasm.c (default_binds_local_p_3): Force indirect function
1386 resolver non-local only if targetm.ifunc_ref_local_ok returns
1388 * config/i386/i386-expand.c (ix86_expand_call): Call
1389 ix86_call_use_plt_p to check if PLT should be used.
1390 * config/i386/i386-protos.h (ix86_call_use_plt_p): New.
1391 * config/i386/i386.c (output_pic_addr_const): Call
1392 ix86_call_use_plt_p to check if "@PLT" is needed.
1393 (ix86_call_use_plt_p): New.
1394 (TARGET_IFUNC_REF_LOCAL_OK): New.
1395 * doc/tm.texi.in: Add TARGET_IFUNC_REF_LOCAL_OK.
1396 * doc/tm.texi: Regenerated.
1398 2021-12-03 Jakub Jelinek <jakub@redhat.com>
1400 * attribs.h (simple_cst_list_equal): Declare.
1401 * attribs.c (simple_cst_list_equal): No longer static.
1402 * config/i386/i386-options.c (target_attribute_cache): New variable.
1403 (ix86_valid_target_attribute_p): Cache DECL_FUNCTION_SPECIFIC_TARGET
1404 and DECL_FUNCTION_SPECIFIC_OPTIMIZATION based on args.
1406 2021-12-03 Jakub Jelinek <jakub@redhat.com>
1409 * ggc.h (gt_pch_note_callback): Declare.
1410 * gengtype.h (enum typekind): Add TYPE_CALLBACK.
1411 (callback_type): Declare.
1412 * gengtype.c (dbgprint_count_type_at): Handle TYPE_CALLBACK.
1413 (callback_type): New variable.
1414 (process_gc_options): Add CALLBACK argument, handle callback
1416 (set_gc_used_type): Adjust process_gc_options caller, if callback,
1417 set type to &callback_type.
1418 (output_mangled_typename): Handle TYPE_CALLBACK.
1419 (walk_type): Likewise. Handle callback option.
1420 (write_types_process_field): Handle TYPE_CALLBACK.
1421 (write_types_local_user_process_field): Likewise.
1422 (write_types_local_process_field): Likewise.
1423 (write_root): Likewise.
1424 (dump_typekind): Likewise.
1425 (dump_type): Likewise.
1426 * gengtype-state.c (type_lineloc): Handle TYPE_CALLBACK.
1427 (state_writer::write_state_callback_type): New method.
1428 (state_writer::write_state_type): Handle TYPE_CALLBACK.
1429 (read_state_callback_type): New function.
1430 (read_state_type): Handle TYPE_CALLBACK.
1431 * ggc-common.c (callback_vec): New variable.
1432 (gt_pch_note_callback): New function.
1433 (gt_pch_save): Stream out gt_pch_save function address and relocation
1435 (gt_pch_restore): Stream in saved gt_pch_save function address and
1436 relocation table and apply relocations if needed.
1437 * doc/gty.texi (callback): Document new GTY option.
1438 * varasm.c (get_unnamed_section): Change callback argument's type and
1439 last argument's type from const void * to const char *.
1440 (output_section_asm_op): Change argument's type from const void *
1441 to const char *, remove unnecessary cast.
1442 * tree-core.h (struct tree_translation_unit_decl): Drop GTY((skip))
1443 from language member.
1444 * output.h (unnamed_section_callback): Change argument type from
1445 const void * to const char *.
1446 (struct unnamed_section): Use GTY((callback)) instead of GTY((skip))
1447 for callback member. Change data member type from const void *
1449 (struct noswitch_section): Use GTY((callback)) instead of GTY((skip))
1450 for callback member.
1451 (get_unnamed_section): Change callback argument's type and
1452 last argument's type from const void * to const char *.
1453 (output_section_asm_op): Change argument's type from const void *
1455 * config/avr/avr.c (avr_output_progmem_section_asm_op): Likewise.
1456 Remove unneeded cast.
1457 * config/darwin.c (output_objc_section_asm_op): Change argument's type
1458 from const void * to const char *.
1459 * config/pa/pa.c (som_output_text_section_asm_op): Likewise.
1460 (som_output_comdat_data_section_asm_op): Likewise.
1461 * config/rs6000/rs6000.c (rs6000_elf_output_toc_section_asm_op):
1463 (rs6000_xcoff_output_readonly_section_asm_op): Likewise. Instead
1464 of dereferencing directive hardcode variable names and decide based on
1465 whether directive is NULL or not.
1466 (rs6000_xcoff_output_readwrite_section_asm_op): Change argument's type
1467 from const void * to const char *.
1468 (rs6000_xcoff_output_tls_section_asm_op): Likewise. Instead
1469 of dereferencing directive hardcode variable names and decide based on
1470 whether directive is NULL or not.
1471 (rs6000_xcoff_output_toc_section_asm_op): Change argument's type
1472 from const void * to const char *.
1473 (rs6000_xcoff_asm_init_sections): Adjust get_unnamed_section callers.
1475 2021-12-03 Richard Purdie <richard.purdie@linuxfoundation.org>
1477 * Makefile.in: Fix "argument list too long" from install-plugins.
1479 2021-12-03 Matt Jacobson <mhjacobson@me.com>
1481 * config.gcc: For the AVR target, populate TM_MULTILIB_CONFIG.
1482 * config/avr/genmultilib.awk: Add ability to filter generated multilib
1484 * config/avr/t-avr: Pass TM_MULTILIB_CONFIG to genmultilib.awk.
1485 * configure.ac: Update help string for --with-multilib-list.
1486 * configure: Regenerate.
1488 2021-12-02 Ilya Lipnitskiy <ilya.lipnitskiy@gmail.com>
1490 * configure: Regenerate.
1491 * configure.ac: Define TARGET_LIBC_GNUSTACK on musl.
1493 2021-12-02 Iain Sandoe <iain@sandoe.co.uk>
1496 * config/host-darwin.c (TRY_EMPTY_VM_SPACE,
1497 SAFE_ALLOC_SIZE): New.
1498 (darwin_gt_pch_get_address): Rewrite to use nominated
1499 memory segments rather than part of the compiler __DATA
1501 (darwin_gt_pch_use_address): Likewise.
1503 2021-12-02 Gerald Pfeifer <gerald@pfeifer.com>
1505 * doc/install.texi (*-*-freebsd*): Remove references to
1506 FreeBSD 1 and FreeBSD 2.
1508 2021-12-02 Vladimir N. Makarov <vmakarov@redhat.com>
1510 PR rtl-optimization/103437
1511 * ira-color.c (setup_allocno_priorities): Switch off backup code
1512 for overflow if compiler has __builtin_smul_overflow. Use <
1513 for comparison with -INT_MAX.
1515 2021-12-02 Bill Schmidt <wschmidt@linux.ibm.com>
1517 * config/rs6000/altivec.h: Delete a number of #defines that are now
1518 superfluous. Alphabetize. Include rs6000-vecdefines.h. Include some
1520 * config/rs6000/rs6000-builtin-new.def (CMPB): Flag as no32bit.
1521 (BPERMD): Flag as 32bit (needing special handling for 32-bit).
1522 (UNPACK_TD): Return unsigned long long instead of unsigned long.
1523 (GET_TEXASR): Return unsigned long instead of unsigned long long.
1524 (GET_TEXASRU): Likewise.
1525 (GET_TFHAR): Likewise.
1526 (GET_TFIAR): Likewise.
1527 (SET_TEXASR): Pass unsigned long instead of unsigned long long.
1528 (SET_TEXASRU): Likewise.
1529 (SET_TFHAR): Likewise.
1530 (SET_TFIAR): Likewise.
1531 (TABORTDC): Likewise.
1532 (TABORTDCI): Likewise.
1533 * config/rs6000/rs6000-call.c (rs6000_expand_new_builtin): Fix error
1534 handling for no32bit. Add 32bit handling for RS6000_BIF_BPERMD.
1535 * config/rs6000/rs6000-gen-builtins.c (write_init_file): Initialize
1536 new_builtins_are_live to 1.
1538 2021-12-02 Vladimir N. Makarov <vmakarov@redhat.com>
1540 PR rtl-optimization/103437
1541 * ira-color.c (setup_allocno_priorities): Use long long
1542 multiplication as backup for overflow processing.
1544 2021-12-02 Tamar Christina <tamar.christina@arm.com>
1546 * config/aarch64/aarch64-protos.h (aarch64_gen_shareable_zero): New.
1547 * config/aarch64/aarch64-simd.md (aarch64_rshrn<mode>,
1548 aarch64_rshrn2<mode>): Generate rounding half-ing add when appropriate.
1549 * config/aarch64/aarch64.c (aarch64_gen_shareable_zero): New.
1551 2021-12-02 Vladimir N. Makarov <vmakarov@redhat.com>
1553 PR rtl-optimization/103437
1554 * ira-color.c (setup_allocno_priorities): Process multiplication
1557 2021-12-02 Martin Liska <mliska@suse.cz>
1559 PR gcov-profile/96092
1560 * common.opt: New option.
1561 * coverage.c (coverage_begin_function): Emit filename with
1562 remap_profile_filename.
1563 * doc/invoke.texi: Document the new option.
1564 * file-prefix-map.c (add_profile_prefix_map): New.
1565 (remap_profile_filename): Likewise.
1566 * file-prefix-map.h (add_profile_prefix_map): Likewise.
1567 (remap_profile_filename): Likewise.
1568 * lto-opts.c (lto_write_options): Handle
1569 OPT_fprofile_prefix_map_.
1570 * opts-global.c (handle_common_deferred_options): Likewise.
1571 * opts.c (common_handle_option): Likewise.
1572 (gen_command_line_string): Likewise.
1573 * profile.c (output_location): Emit filename with
1574 remap_profile_filename.
1576 2021-12-02 Richard Biener <rguenther@suse.de>
1578 PR middle-end/103271
1579 * internal-fn.c (expand_DEFERRED_INIT): When the base
1580 of the LHS is a decl with matching constant size use
1581 that as the initialization target instead of an
1582 eventual VLA typed one.
1584 2021-12-02 Richard Biener <rguenther@suse.de>
1586 PR tree-optimization/103527
1587 * tree-vect-stmts.c (vect_build_gather_load_calls): Always
1588 use a truth type for building the vector mask.
1590 2021-12-02 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
1592 * config/vxworks.h (LTO_PLUGIN_SPEC): Adapt to corrected
1593 spelling of -linker-output-auto-nolto-rel.
1595 2021-12-01 Uroš Bizjak <ubizjak@gmail.com>
1598 * config/i386/sse.md (VI2F): Remove mode iterator.
1599 (VI2F_256_512): New mode iterator.
1600 (vec_set<V8_128:mode>_0): New insn pattern.
1601 (vec_set<VI2F_256_512:mode>_0>): Rename from vec_set<VI2F:mode>mode.
1602 Use VI2F_256_512 mode iterator instead of VI2F.
1603 (*axv512fp16_movsh): Remove.
1604 (<sse2p4_1>_pinsr<ssemodesuffix>): Add (x,x,x) AVX2 alternative.
1605 Do not disable V8HF mode insn on AVX2 targets.
1606 (pinsrw -> pbroadcast + pblendw peephole2): New peephole.
1607 (pinsrw -> pbroadcast + pblendw splitter): New post-reload splitter.
1608 * config/i386/i386.md (extendhfsf): Call gen_vec_setv8hf_0.
1609 * config/i386/i386-expand.c (ix86_expand_vector_set)
1610 <case E_V8HFmode>: Use vec_merge path for TARGET_AVX2.
1612 2021-12-01 Roger Sayle <roger@nextmovesoftware.com>
1613 Richard Biener <rguenther@suse.de>
1615 * tree-ssa-loop-niter.c (number_of_iterations_until_wrap):
1616 Check if simplify_using_initial_conditions allows us to
1617 simplify the expression for may_be_zero.
1618 * match.pd (X != C ? -X : -C -> -X): New transform.
1619 (X != C ? ~X : ~C -> ~X): Likewise.
1620 ((X+1) > Y ? -X : 1 -> X >= Y ? -X : 1): Likewise.
1622 2021-12-01 Peter Bergner <bergner@linux.ibm.com>
1624 PR middle-end/103127
1625 * gimplify.c (is_var_need_auto_init): Handle opaque types.
1627 2021-12-01 Iain Buclaw <ibuclaw@gdcproject.org>
1629 * common.opt (static-libphobos): Add option.
1630 * config/darwin.h (LINK_SPEC): Substitute -lgphobos with libgphobos.a
1631 when linking statically.
1632 * gcc.c (driver_handle_option): Set -static-libphobos as always valid.
1634 2021-12-01 Jason Merrill <jason@redhat.com>
1637 * fold-const.c (maybe_nonzero_address): Use get_create or get
1638 depending on folding_initializer.
1639 (fold_binary_initializer_loc): New.
1640 * fold-const.h (fold_binary_initializer_loc): Declare.
1641 * varasm.c (mark_weak): Don't use the decl location.
1642 * doc/invoke.texi: Document -fconstexpr-fp-except.
1644 2021-12-01 Bill Schmidt <wschmidt@linux.ibm.com>
1647 * config/rs6000/rs6000-c.c (rs6000-builtins.h): Stop including.
1648 (rs6000-internal.h): Include.
1649 (altivec_resolve_new_overloaded_builtin): Move call to
1650 rs6000_invalid_new_builtin here from rs6000_new_builtin_decl.
1651 * config/rs6000/rs6000-call.c (rs6000-builtins.h): Stop including.
1652 (rs6000_invalid_new_builtin): Remove static qualifier.
1653 (rs6000_new_builtin_decl): Remove test for supported builtin.
1654 * config/rs6000/rs6000-internal.h (rs6000-builtins.h): Include.
1655 (rs6000_invalid_new_builtin): Declare.
1656 * config/rs6000/rs6000.c (rs6000-builtins.h): Don't include.
1658 2021-12-01 Aldy Hernandez <aldyh@redhat.com>
1660 PR tree-optimization/103409
1661 * gimple-range-path.cc (path_range_query::compute_ranges_in_phis):
1662 Do all the work with just one ssa_global_cache.
1663 * gimple-range-path.h: Remove m_tmp_phi_cache.
1665 2021-12-01 Richard Sandiford <richard.sandiford@arm.com>
1667 PR tree-optimization/103517
1668 * tree-vect-slp.c (vect_build_slp_tree_1): When allowing two
1669 different component references, check the codes of both them,
1670 rather than just the first.
1672 2021-12-01 Andrew MacLeod <amacleod@redhat.com>
1674 PR tree-optimization/103464
1675 * gimple-range.cc (gimple_ranger::prefill_name): Process phis also.
1676 (gimple_ranger::prefill_stmt_dependencies): Ditto.
1678 2021-12-01 Iain Buclaw <ibuclaw@gdcproject.org>
1680 * doc/install.texi (Prerequisites): Add note that D front end now
1681 requires GDC installed in order to bootstrap.
1682 (Building): Add D compiler section, referencing prerequisites.
1684 2021-12-01 Jakub Jelinek <jakub@redhat.com>
1686 * config/i386/x86-tune.def (X86_TUNE_SCHEDULE,
1687 X86_TUNE_SSE_PARTIAL_REG_FP_CONVERTS_DEPENDENCY,
1688 X86_TUNE_MISALIGNED_MOVE_STRING_PRO_EPILOGUES): Formatting fixes.
1689 (X86_TUNE_USE_GATHER): Put m_GENERIC last for consistency.
1691 2021-12-01 Jakub Jelinek <jakub@redhat.com>
1693 PR rtl-optimization/102356
1694 * rtl.h (simplify_context): Add assoc_count member and
1695 max_assoc_count static member.
1696 * simplify-rtx.c (simplify_associative_operation): Don't reassociate
1697 more than max_assoc_count times within one outermost simplify_* call.
1698 * dwarf2out.c (mem_loc_descriptor): Optimize binary operation
1699 with both operands the same using DW_OP_dup.
1701 2021-12-01 Tamar Christina <tamar.christina@arm.com>
1703 * match.pd: Move below pattern that rewrites to EQ, NE.
1704 * tree.c (bitmask_inv_cst_vector_p): Correct do .. while indentation.
1706 2021-12-01 Siddhesh Poyarekar <siddhesh@gotplt.org>
1707 Martin Liška <mliska@suse.cz>
1709 PR tree-optimization/103456
1710 * tree-object-size.c (merge_object_sizes): Update osi->changed
1711 only if object_sizes_set succeeded.
1713 2021-11-30 liuhongt <hongtao.liu@intel.com>
1717 * config/i386/i386.md (*x86_64_shld_1): Set_attr
1718 length_immediate to 1.
1719 (*x86_shld_1): Ditto.
1720 (*x86_64_shrd_1): Ditto.
1721 (*x86_shrd_1): Ditto.
1723 2021-11-30 Bill Schmidt <wschmidt@linux.ibm.com>
1725 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
1727 (altivec_resolve_new_overloaded_builtin): Likewise.
1729 2021-11-30 Martin Jambor <mjambor@suse.cz>
1732 * ipa-sra.c (scan_function): Also check ECF_LOOPING_CONST_OR_PURE flag.
1734 2021-11-30 Richard Sandiford <richard.sandiford@arm.com>
1736 PR tree-optimization/103494
1737 * tree-vect-stmts.c (vect_get_gather_scatter_ops): Remove ncopies
1738 argument and calculate ncopies from gs_info->offset_vectype
1740 (vectorizable_store, vectorizable_load): Update accordingly.
1742 2021-11-30 Iain Buclaw <ibuclaw@gdcproject.org>
1744 * Makefile.in (GDC): New variable.
1745 (GDCFLAGS): New variable.
1746 * configure: Regenerate.
1747 * configure.ac: Add call to ACX_PROG_GDC. Substitute GDCFLAGS.
1749 2021-11-30 Martin Liska <mliska@suse.cz>
1750 Martin Jambor <mjambor@suse.cz>
1753 * ipa-param-manipulation.c
1754 (ipa_param_body_adjustments::prepare_debug_expressions): Be
1755 careful about hash_map reallocating itself. Simpify a return
1756 which always returns true.
1758 2021-11-30 Andrew MacLeod <amacleod@redhat.com>
1760 PR tree-optimization/103440
1761 * gimple-range-fold.cc (fold_using_range::range_of_phi): Continue
1762 normal param processing for equiv params.
1764 2021-11-30 Richard Biener <rguenther@suse.de>
1766 * config/i386/i386.c (ix86_shift_rotate_cost): Remove
1768 * tree-chrec.c (evolution_function_is_invariant_rec_p):
1770 * tree-if-conv.c (if_convertible_stmt_p): Likewise.
1771 * tree-ssa-pre.c (fully_constant_expression): Likewise.
1772 * tree-vrp.c (operand_less_p): Likewise.
1773 * reload.c (reg_overlap_mentioned_for_reload_p): Remove
1774 unreachable gcc_unreachable ().
1775 * sel-sched-ir.h (bb_next_bb): Likewise.
1776 * varasm.c (compare_constant): Likewise.
1778 2021-11-30 Richard Biener <rguenther@suse.de>
1780 PR tree-optimization/103489
1781 * tree-vect-loop.c (vectorizable_phi): Verify argument
1782 vector type compatibility to mitigate bool pattern recog
1785 2021-11-30 Thomas Schwinge <thomas@codesourcery.com>
1787 * omp-offload.c (oacc_loop_auto_partitions): Remove erroneous
1788 "Orphan reductions cannot have gang partitioning" handling.
1790 2021-11-30 Thomas Schwinge <thomas@codesourcery.com>
1792 * omp-offload.c (oacc_loop_process): Implement "gang reduction on
1793 an orphan loop" checking.
1795 2021-11-30 Cesar Philippidis <cesar@codesourcery.com>
1796 Thomas Schwinge <thomas@codesourcery.com>
1798 * omp-general.h (enum oacc_loop_flags): Add OLF_REDUCTION enum.
1799 * omp-low.c (lower_oacc_head_mark): Use it to mark OpenACC
1801 * omp-offload.c (oacc_loop_auto_partitions): Don't assign gang
1802 level parallelism to orphan reductions.
1804 2021-11-30 Aldy Hernandez <aldyh@redhat.com>
1806 PR tree-optimization/103451
1807 * range-op.cc (operator_div::wi_fold): Remove
1808 can_throw_non_call_exceptions special case.
1809 * tree-ssa-sink.c (sink_code_in_bb): Same.
1811 2021-11-30 Richard Sandiford <richard.sandiford@arm.com>
1813 * tree-vect-slp.c (arg1_arg4_map): New variable.
1814 (vect_get_operand_map): Handle IFN_MASK_GATHER_LOAD.
1815 (vect_build_slp_tree_1): Likewise.
1816 (vect_build_slp_tree_2): Likewise.
1817 * tree-vect-stmts.c (vectorizable_load): Expect the mask to be
1818 the last SLP child node rather than the first.
1820 2021-11-30 Richard Sandiford <richard.sandiford@arm.com>
1822 * tree-if-conv.c: Include tree-eh.h.
1823 (predicate_statements): Remove pe argument. Don't hoist
1825 (combine_blocks): Remove pe argument.
1826 (ifcvt_available_on_edge_p, ifcvt_can_hoist): New functions.
1827 (ifcvt_hoist_invariants): Likewise.
1828 (tree_if_conversion): Update call to combine_blocks. Call
1829 ifcvt_hoist_invariants after VN.
1831 2021-11-30 Richard Sandiford <richard.sandiford@arm.com>
1833 PR tree-optimization/102467
1834 * doc/sourcebuild.texi (vect_gather_load_ifn): Document.
1835 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
1836 Commonize safelen handling. Punt for anything involving
1837 gathers and scatters unless safelen says otherwise.
1838 * tree-vect-slp.c (arg1_map): New variable.
1839 (vect_get_operand_map): Handle IFN_GATHER_LOAD.
1840 (vect_build_slp_tree_1): Likewise.
1841 (vect_build_slp_tree_2): Likewise.
1842 (compatible_calls_p): If vect_get_operand_map returns nonnull,
1843 check that any skipped arguments are equal.
1844 (vect_slp_analyze_node_operations_1): Tighten reduction check.
1845 * tree-vect-stmts.c (check_load_store_for_partial_vectors): Take
1846 an ncopies argument.
1847 (vect_get_gather_scatter_ops): Take slp_node and ncopies arguments.
1849 (vectorizable_store, vectorizable_load): Adjust accordingly.
1851 2021-11-30 Richard Sandiford <richard.sandiford@arm.com>
1853 * gimple.h (gimple_num_args, gimple_arg, gimple_arg_ptr): New
1855 * tree-vect-slp.c (cond_expr_maps, arg2_map): New variables.
1856 (vect_get_operand_map): New function.
1857 (vect_get_and_check_slp_defs): Fix outdated comment.
1858 Use vect_get_operand_map and new gimple argument accessors.
1859 (vect_build_slp_tree_2): Likewise.
1861 2021-11-30 Richard Sandiford <richard.sandiford@arm.com>
1863 * tree-vect-slp.c (vect_build_slp_tree_1): Use code_helper
1864 to record the operations performed by statements, only using
1865 CALL_EXPR for things that don't map to built-in or internal
1866 functions. For shifts, require all shift amounts to be equal
1867 if optab_vector is not supported but optab_scalar is.
1869 2021-11-30 Richard Sandiford <richard.sandiford@arm.com>
1871 * tree-vect-data-refs.c (vect_check_gather_scatter): Continue
1872 processing conversions if the current offset is a pointer.
1874 2021-11-30 Richard Sandiford <richard.sandiford@arm.com>
1876 * tree-vectorizer.h (vect_is_reduction): Use STMT_VINFO_REDUC_IDX.
1878 2021-11-30 Richard Sandiford <richard.sandiford@arm.com>
1880 * genopinit.c (main): Turn supports_vec_gather_load and
1881 supports_vec_scatter_store into signed char arrays and remove
1882 supports_vec_gather_load_cached and supports_vec_scatter_store_cached.
1883 * optabs-query.c (supports_vec_convert_optab_p): Add a mode parameter.
1884 If the mode is not VOIDmode, test only for that mode.
1885 (supports_vec_gather_load_p): Likewise.
1886 (supports_vec_scatter_store_p): Likewise.
1887 * optabs-query.h (supports_vec_gather_load_p): Likewise.
1888 (supports_vec_scatter_store_p): Likewise.
1889 * tree-vect-data-refs.c (vect_check_gather_scatter): Pass the
1890 vector mode to supports_vec_gather_load_p and
1891 supports_vec_scatter_store_p.
1893 2021-11-30 Richard Sandiford <richard.sandiford@arm.com>
1895 * internal-fn.c (commutative_binary_fn_p): Handle IFN_ADD_OVERFLOW
1896 and IFN_MUL_OVERFLOW.
1898 2021-11-30 Richard Sandiford <richard.sandiford@arm.com>
1900 * internal-fn.c (commutative_binary_fn_p): Handle IFN_UBSAN_CHECK_ADD
1901 and IFN_UBSAN_CHECK_MUL.
1903 2021-11-30 Richard Sandiford <richard.sandiford@arm.com>
1905 * internal-fn.c (commutative_binary_fn_p): Handle IFN_COMPLEX_MUL.
1907 2021-11-30 Richard Sandiford <richard.sandiford@arm.com>
1909 * gimple-fold.c: Include internal-fn.h.
1910 (fold_stmt_1): If a function maps to an internal one, use
1911 first_commutative_argument to canonicalize the order of
1912 commutative arguments.
1913 * gimple-match-head.c (gimple_resimplify2, gimple_resimplify3)
1914 (gimple_resimplify4, gimple_resimplify5): Extend commutativity
1915 checks to functions.
1917 2021-11-30 Richard Sandiford <richard.sandiford@arm.com>
1919 * doc/md.texi (reduc_fmin_scal_@var{m}): Document.
1920 (reduc_fmax_scal_@var{m}): Likewise.
1921 * optabs.def (reduc_fmax_scal_optab): New optab.
1922 (reduc_fmin_scal_optab): Likewise
1923 * internal-fn.def (REDUC_FMAX, REDUC_FMIN): New functions.
1924 * tree-vect-loop.c (reduction_fn_for_scalar_code): Handle
1925 CASE_CFN_FMAX and CASE_CFN_FMIN.
1926 (neutral_op_for_reduction): Likewise.
1927 (needs_fold_left_reduction_p): Likewise.
1928 * config/aarch64/iterators.md (FMAXMINV): New iterator.
1929 (fmaxmin): Handle UNSPEC_FMAXNMV and UNSPEC_FMINNMV.
1930 * config/aarch64/aarch64-simd.md (reduc_<optab>_scal_<mode>): Fix
1932 (reduc_<fmaxmin>_scal_<mode>): New pattern.
1933 * config/aarch64/aarch64-sve.md (reduc_<fmaxmin>_scal_<mode>):
1936 2021-11-30 Richard Sandiford <richard.sandiford@arm.com>
1938 * builtins.h (associated_internal_fn): Declare overload that
1939 takes a (combined_cfn, return type) pair.
1940 * builtins.c (associated_internal_fn): Split new overload out
1941 of original fndecl version. Also provide an overload that takes
1942 a (combined_cfn, return type) pair.
1943 * internal-fn.h (commutative_binary_fn_p): Declare.
1944 (commutative_ternary_fn_p): Likewise.
1945 (associative_binary_fn_p): Likewise.
1946 * internal-fn.c (commutative_binary_fn_p, commutative_ternary_fn_p):
1947 New functions, split out from...
1948 (first_commutative_argument): ...here.
1949 (associative_binary_fn_p): New function.
1950 * gimple-match.h (code_helper): Add a constructor that takes
1952 (commutative_binary_op_p): Declare.
1953 (commutative_ternary_op_p): Likewise.
1954 (first_commutative_argument): Likewise.
1955 (associative_binary_op_p): Likewise.
1956 (canonicalize_code): Likewise.
1957 (directly_supported_p): Likewise.
1958 (get_conditional_internal_fn): Likewise.
1959 (gimple_build): New overloads that takes a code_helper.
1960 * gimple-fold.c (gimple_build): Likewise.
1961 * gimple-match-head.c (commutative_binary_op_p): New function.
1962 (commutative_ternary_op_p): Likewise.
1963 (first_commutative_argument): Likewise.
1964 (associative_binary_op_p): Likewise.
1965 (canonicalize_code): Likewise.
1966 (directly_supported_p): Likewise.
1967 (get_conditional_internal_fn): Likewise.
1968 * tree-vectorizer.h: Include gimple-match.h.
1969 (neutral_op_for_reduction): Take a code_helper instead of a tree_code.
1970 (needs_fold_left_reduction_p): Likewise.
1971 (reduction_fn_for_scalar_code): Likewise.
1972 (vect_can_vectorize_without_simd_p): Declare a nNew overload that takes
1974 * tree-vect-loop.c: Include case-cfn-macros.h.
1975 (fold_left_reduction_fn): Take a code_helper instead of a tree_code.
1976 (reduction_fn_for_scalar_code): Likewise.
1977 (neutral_op_for_reduction): Likewise.
1978 (needs_fold_left_reduction_p): Likewise.
1979 (use_mask_by_cond_expr_p): Likewise.
1980 (build_vect_cond_expr): Likewise.
1981 (vect_create_partial_epilog): Likewise. Use gimple_build rather
1982 than gimple_build_assign.
1983 (check_reduction_path): Handle calls and operate on code_helpers
1984 rather than tree_codes.
1985 (vect_is_simple_reduction): Likewise.
1986 (vect_model_reduction_cost): Likewise.
1987 (vect_find_reusable_accumulator): Likewise.
1988 (vect_create_epilog_for_reduction): Likewise.
1989 (vect_transform_cycle_phi): Likewise.
1990 (vectorizable_reduction): Likewise. Make more use of
1992 (vect_transform_reduction): Use gimple_extract_op but expect
1993 a tree_code for now.
1994 (vect_can_vectorize_without_simd_p): New overload that takes
1996 * tree-vect-stmts.c (vectorizable_call): Handle reductions in
1998 * tree-vect-patterns.c (vect_mark_pattern_stmts): Use
1999 gimple_extract_op when updating STMT_VINFO_REDUC_IDX.
2001 2021-11-30 Richard Sandiford <richard.sandiford@arm.com>
2003 * gimple-match.h (code_helper): Provide == and != overloads.
2004 (code_helper::operator tree_code): Make explicit.
2005 (code_helper::operator combined_fn): Likewise.
2006 * gimple-match-head.c (convert_conditional_op): Use explicit
2007 conversions where necessary.
2008 (gimple_resimplify1, gimple_resimplify2, gimple_resimplify3): Likewise.
2009 (maybe_push_res_to_seq, gimple_simplify): Likewise.
2010 * gimple-fold.c (replace_stmt_with_simplification): Likewise.
2012 2021-11-30 Richard Sandiford <richard.sandiford@arm.com>
2014 * gimple-match.h (code_helper): Add functions for querying whether
2015 the code represents an internal_fn or a built_in_function.
2016 Provide explicit conversion operators for both cases.
2017 (gimple_extract_op): Declare.
2018 * gimple-match-head.c (gimple_extract): New function, extracted from...
2019 (gimple_simplify): ...here.
2020 (gimple_extract_op): New function.
2022 2021-11-30 Eric Botcazou <ebotcazou@adacore.com>
2025 * config/i386/i386.c (ix86_output_call_insn): Beef up comment about
2026 nops emitted with SEH.
2027 * config/i386/winnt.c (i386_pe_seh_unwind_emit): When switching to
2028 the cold section, emit a nop before the directive if the previous
2029 active instruction is a call.
2031 2021-11-30 Roger Sayle <roger@nextmovesoftware.com>
2032 Uroš Bizjak <ubizjak@gmail.com>
2035 * config/i386/sse.md (define_split): New splitters to simplify
2036 not;vec_duplicate;and as vec_duplicate;andn.
2038 2021-11-30 Richard Biener <rguenther@suse.de>
2040 PR middle-end/103485
2041 * match.pd (-((int)x >> (prec - 1)) to (unsigned)x >> (prec - 1)):
2042 Use VIEW_CONVERT_EXPR for vectors.
2044 2021-11-30 Richard Biener <rguenther@suse.de>
2046 * cfgrtl.c (skip_insns_after_block): Refactor code to
2047 be more easily readable.
2048 * expr.c (op_by_pieces_d::run): Remove unreachable
2050 * sched-deps.c (sched_analyze): Remove unreachable
2052 * sel-sched-ir.c (in_same_ebb_p): Likewise.
2053 * tree-ssa-alias.c (nonoverlapping_refs_since_match_p):
2054 Remove unreachable code.
2055 * tree-vect-slp.c (vectorize_slp_instance_root_stmt):
2056 Refactor to avoid unreachable loop iteration.
2057 * tree.c (walk_tree_1): Remove unreachable break.
2058 * vec-perm-indices.c (vec_perm_indices::series_p): Remove
2061 2021-11-30 Kewen Lin <linkw@linux.ibm.com>
2064 * config/rs6000/rs6000-call.c (rs6000_builtin_decl): Remove builtin mask
2067 2021-11-30 Kewen Lin <linkw@linux.ibm.com>
2069 * config/rs6000/rs6000.c
2070 (rs6000_cost_data::update_target_cost_per_stmt): Adjust the way to
2071 compute extra penalized cost. Remove useless parameter.
2072 (rs6000_cost_data::rs6000_add_stmt_cost): Adjust the call to function
2073 update_target_cost_per_stmt.
2075 2021-11-30 Kewen Lin <linkw@linux.ibm.com>
2078 2021-11-17 Kewen Lin <linkw@linux.ibm.com>
2080 * config/visium/visium.md (*add<mode>3_insn, *addsi3_insn, *addi3_insn,
2081 *sub<mode>3_insn, *subsi3_insn, *subdi3_insn, *neg<mode>2_insn,
2082 *negdi2_insn, *and<mode>3_insn, *ior<mode>3_insn, *xor<mode>3_insn,
2083 *one_cmpl<mode>2_insn, *ashl<mode>3_insn, *ashr<mode>3_insn,
2084 *lshr<mode>3_insn, *trunchiqi2_insn, *truncsihi2_insn,
2085 *truncdisi2_insn, *extendqihi2_insn, *extendqisi2_insn,
2086 *extendhisi2_insn, *extendsidi2_insn, *zero_extendqihi2_insn,
2087 *zero_extendqisi2_insn, *zero_extendsidi2_insn): Fix split condition.
2089 2021-11-30 Andrew MacLeod <amacleod@redhat.com>
2091 PR tree-optimization/103467
2092 * gimple-range-gori.cc (range_def_chain::register_dependency): Don't
2093 use an object reference after a potential resize.
2095 2021-11-29 Uroš Bizjak <ubizjak@gmail.com>
2098 * config/i386/i386.md (*movhi_internal): Introduce (*v,C) alternative.
2099 Do not allocate non-GPR registers. Optimize xmm->xmm moves when
2100 optimizing for size. Fix vpinsrw insn template.
2101 (*movhf_internal): Fix pinsrw and pextrw insn templates for
2102 AVX targets. Use sselog1 type instead of sselog. Optimize GPR moves.
2103 Optimize xmm->xmm moves for TARGET_SSE_PARTIAL_REG_DEPENDENCY
2104 and TARGET_SSE_SPLIT_REGS targets.
2106 2021-11-29 Eric Gallager <egallager@gcc.gnu.org>
2108 * configure: Re-regenerate.
2110 2021-11-29 Eric Gallager <egallager@gcc.gnu.org>
2113 * Makefile.in: Substitute CTAGS, ETAGS, and CSCOPE
2114 variables. Use ETAGS variable in TAGS target.
2115 * configure: Regenerate.
2116 * configure.ac: Allow CTAGS, ETAGS, and CSCOPE
2117 variables to be overridden.
2119 2021-11-29 Paul A. Clarke <pc@us.ibm.com>
2121 * config/rs6000/xmmintrin.h (_mm_movemask_ps): Use vec_extractm
2123 * config/rs6000/emmintrin.h (_mm_movemask_pd): Likewise.
2124 (_mm_movemask_epi8): Likewise.
2126 2021-11-29 Richard Biener <rguenther@suse.de>
2128 * read-rtl-function.c (function_reader::read_rtx_operand):
2129 Return only after resetting m_in_call_function_usage.
2131 2021-11-29 Roger Sayle <roger@nextmovesoftware.com>
2133 * config/i386/i386-expand.c (ix86_expand_v1ti_to_ti): Perform the
2134 conversion via V2DImode using vec_extractv2didi on TARGET_SSE2.
2135 * config/i386/sse.md (rotlv1ti3, rotrv1ti3): Change constraint
2136 on QImode shift amounts from const_int_operand to general_operand.
2138 2021-11-29 Richard Biener <rguenther@suse.de>
2140 * tree.h (reverse_storage_order_for_component_p): Remove
2141 spurious gcc_unreachable.
2142 * cfganal.c (dfs_find_deadend): Likewise.
2143 * fold-const-call.c (fold_const_logb): Likewise.
2144 (fold_const_significand): Likewise.
2145 * gimple-ssa-store-merging.c (lhs_valid_for_store_merging_p):
2148 2021-11-29 Richard Biener <rguenther@suse.de>
2150 * vec.c (qsort_chk): Do not return the void return value
2151 from the noreturn qsort_chk_error.
2152 * ccmp.c (expand_ccmp_expr_1): Remove unreachable return.
2153 * df-scan.c (df_ref_equal_p): Likewise.
2154 * dwarf2out.c (is_base_type): Likewise.
2155 (add_const_value_attribute): Likewise.
2156 * fixed-value.c (fixed_arithmetic): Likewise.
2157 * gimple-fold.c (gimple_fold_builtin_fputs): Likewise.
2158 * gimple-ssa-strength-reduction.c (stmt_cost): Likewise.
2159 * graphite-isl-ast-to-gimple.c
2160 (gcc_expression_from_isl_expr_op): Likewise.
2161 (gcc_expression_from_isl_expression): Likewise.
2162 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate):
2164 * lto-streamer-in.c (lto_input_mode_table): Likewise.
2166 2021-11-29 liuhongt <hongtao.liu@intel.com>
2169 * config/i386/i386.c (inline_secondary_memory_needed): HImode
2170 move between GPR and SSE registers is supported under
2171 TARGET_SSE2 and above.
2172 * config/i386/i386.md (extendhfsf2): Optimize expander.
2173 (truncsfhf2): Ditto.
2174 * config/i386/sse.md (sse2p4_1): Adjust attr for V8HFmode to
2175 align with V8HImode.
2177 2021-11-29 liuhongt <hongtao.liu@intel.com>
2181 * config/i386/i386.c (ix86_secondary_reload): Without
2182 TARGET_SSE4_1, General register is needed to move HImode from
2183 sse register to memory.
2184 * config/i386/sse.md (*vec_extrachf): Use %vpextrw instead of
2185 pextrw in output templates.
2186 * config/i386/i386.md (movhi_internal): Ditto, also fix typo of
2187 MEM_P (operands[1]) and adjust mode/prefix/type attribute for
2188 alternatives related to sse register.
2190 2021-11-29 Richard Biener <rguenther@suse.de>
2192 PR tree-optimization/103458
2193 * tree-ssa-dce.c (make_forwarders_with_degenerate_phis): Do not
2194 create forwarders for blocks with abnormal predecessors.
2196 2021-11-29 Richard Biener <rguenther@suse.de>
2198 * gimple-predicate-analysis.cc (can_be_invalidated_p):
2199 Restore semantics to the one before the split from
2202 2021-11-28 Jan Hubicka <hubicka@ucw.cz>
2204 * profile.c: Include sreal.h
2205 (struct bb_stats): New.
2206 (cmp_stats): New function.
2207 (compute_branch_probabilities): Output bb stats.
2209 2021-11-28 Jan Hubicka <hubicka@ucw.cz>
2211 * cfghooks.c: Include sreal.h, profile.h.
2212 (profile_record_check_consistency): Fix checking of count counsistency;
2213 record also dynamic mismatches.
2214 * cfgrtl.c (rtl_account_profile_record): Similarly.
2215 * tree-cfg.c (gimple_account_profile_record): Likewise.
2216 * cfghooks.h (struct profile_record): Remove num_mismatched_freq_in,
2217 num_mismatched_freq_out, turn time to double, add
2218 dyn_mismatched_prob_out, dyn_mismatched_count_in,
2219 num_mismatched_prob_out; remove num_mismatched_count_out.
2220 * passes.c (account_profile_1): New function.
2221 (account_profile_in_list): New function.
2222 (pass_manager::dump_profile_report): Rewrite.
2223 (execute_one_ipa_transform_pass): Check profile consistency after
2225 (execute_all_ipa_transforms): Remove cfun test; record all transform
2227 (execute_one_pass): Fix collecting of profile stats.
2229 2021-11-27 Jakub Jelinek <jakub@redhat.com>
2231 PR middle-end/103431
2232 * config/i386/i386.md (x86_64_shld, x86_shld, x86_64_shrd, x86_shrd):
2233 Change insn pattern to accurately describe the instructions.
2234 (*x86_64_shld_1, *x86_shld_1, *x86_64_shrd_1, *x86_shrd_1): New
2235 define_insn patterns.
2236 (*x86_64_shld_2, *x86_shld_2, *x86_64_shrd_2, *x86_shrd_2): New
2237 define_insn_and_split patterns.
2238 (*ashl<dwi>3_doubleword_mask, *ashl<dwi>3_doubleword_mask_1,
2239 *<insn><dwi>3_doubleword_mask, *<insn><dwi>3_doubleword_mask_1,
2240 ix86_rotl<dwi>3_doubleword, ix86_rotr<dwi>3_doubleword): Adjust
2241 splitters for x86_{,64_}sh{l,r}d pattern changes.
2243 2021-11-27 Jakub Jelinek <jakub@redhat.com>
2245 PR tree-optimization/103435
2246 * gimple-ssa-store-merging.c (find_bswap_or_nop_finalize): Avoid UB if
2247 n->range - rsize == 8, just clear both *cmpnop and *cmpxchg in that
2250 2021-11-27 Martin Jambor <mjambor@suse.cz>
2253 * ipa-prop.c (ipcp_transform_function): Call
2254 delete_unreachable_blocks_update_callgraph instead of returning
2257 2021-11-26 Siddhesh Poyarekar <siddhesh@gotplt.org>
2259 * tree-object-size.c (object_sizes_grow, object_sizes_release,
2260 object_sizes_unknown_p, object_sizes_get, object_size_set_force,
2261 object_sizes_set): New functions.
2262 (addr_object_size, compute_builtin_object_size,
2263 expr_object_size, call_object_size, unknown_object_size,
2264 merge_object_sizes, plus_stmt_object_size,
2265 cond_expr_object_size, collect_object_sizes_for,
2266 check_for_plus_in_loops_1, init_object_sizes,
2267 fini_object_sizes): Adjust.
2269 2021-11-26 Siddhesh Poyarekar <siddhesh@gotplt.org>
2271 * tree-object-size.c: New enum.
2272 (object_sizes, computed, addr_object_size,
2273 compute_builtin_object_size, expr_object_size, call_object_size,
2274 merge_object_sizes, plus_stmt_object_size,
2275 collect_object_sizes_for, init_object_sizes, fini_object_sizes,
2276 object_sizes_execute): Replace magic numbers with enums.
2278 2021-11-26 Roger Sayle <roger@nextmovesoftware.com>
2280 * tree-ssa-loop-ivopts.c (cand_value_at): Take a class
2281 tree_niter_desc* argument instead of just a tree for NITER.
2282 If we require the iv candidate value at the end of the final
2283 loop iteration, try using the original loop bound as the
2284 NITER for sufficiently simple loops.
2285 (may_eliminate_iv): Update (only) call to cand_value_at.
2287 2021-11-26 Jan Hubicka <hubicka@ucw.cz>
2289 * ipa-modref.c (analyze_function): Drop parameter F and dump
2290 cgraph node name rather than cfun name.
2291 (modref_generate): Update.
2292 (modref_summaries::insert):Update.
2293 (modref_summaries_lto::insert):Update.
2294 (pass_modref::execute):Update.
2295 (ipa_merge_modref_summary_after_inlining): Improve combining of
2298 2021-11-26 Jan Hubicka <jh@suse.cz>
2301 * ipa-modref.c (update_escape_summary_1): Fix handling of min_flags.
2303 2021-11-26 konglin1 <lingling.kong@intel.com>
2306 * config/i386/i386.c (ix86_can_change_mode_class): Allow 16 bit data in XMM register
2308 * config/i386/i386.md (extendhfsf2): Add extenndhfsf2 for TARGET_F16C.
2309 (extendhfdf2): Restrict extendhfdf for TARGET_AVX512FP16 only.
2310 (*extendhf<mode>2): Rename from extendhf<mode>2.
2311 (truncsfhf2): Likewise.
2312 (truncdfhf2): Likewise.
2313 (*trunc<mode>2): Likewise.
2315 2021-11-26 liuhongt <hongtao.liu@intel.com>
2317 PR middle-end/103419
2318 * match.pd: Fix typo, use the type of second parameter, not
2321 2021-11-25 Jan Hubicka <jh@suse.cz>
2323 * ipa-cp.c (ipa_value_range_from_jfunc): Remove forgotten early return.
2325 2021-11-25 Roger Sayle <roger@nextmovesoftware.com>
2327 PR middle-end/103406
2328 * match.pd (minus @0 @0): Check tree_expr_maybe_infinite_p.
2330 2021-11-25 Martin Jambor <mjambor@suse.cz>
2333 * ipa-prop.h (ipa_get_param): New overload. Move bits of the existing
2335 * ipa-param-manipulation.h (ipa_param_adjustments): New member
2336 function get_updated_index_or_split.
2337 * ipa-param-manipulation.c
2338 (ipa_param_adjustments::get_updated_index_or_split): New function.
2339 * ipa-prop.c (adjust_agg_replacement_values): Reimplement, add
2340 capability to identify scalarized parameters and perform substitution
2342 (ipcp_transform_function): Create descriptors earlier, handle new
2343 return values of adjust_agg_replacement_values.
2345 2021-11-25 Aldy Hernandez <aldyh@redhat.com>
2347 * gimple-range-path.cc (path_range_query::compute_ranges_defined): Remove.
2348 (path_range_query::compute_ranges_in_block): Revert to bitmap order.
2349 * gimple-range-path.h: Remove compute_ranges_defined.
2351 2021-11-25 Andrew Stubbs <ams@codesourcery.com>
2354 * config/gcn/gcn.c (move_callee_saved_registers): Ensure that the
2355 number of spilled registers is counted correctly.
2357 2021-11-25 Jan Hubicka <hubicka@ucw.cz>
2359 * ipa-prop.h (ipa_node_params::ipa_node_params): Initialize
2362 2021-11-25 Andrew MacLeod <amacleod@redhat.com>
2364 PR tree-optimization/103359
2365 * gimple-range-fold.cc (fold_using_range::range_of_phi): If arg is
2366 equivalent to def, don't initially include it's range.
2368 2021-11-25 Jan Hubicka <hubicka@ucw.cz>
2370 * tree-ssa-alias.c (ref_maybe_used_by_call_p_1): Do not check
2371 gimple_static_chain.
2373 2021-11-25 Richard Biener <rguenther@suse.de>
2375 * cgraphunit.c (symbol_table::output_weakrefs): Remove
2377 (get_alias_symbol): Remove now unused function.
2379 2021-11-25 Richard Biener <rguenther@suse.de>
2381 * cfgrtl.c (rtl_verify_fallthru): Do not stop verifying
2383 (skip_insns_after_block): Remove unreachable break and continue.
2385 2021-11-25 Richard Biener <rguenther@suse.de>
2387 * cfgexpand.c (label_rtx_for_bb): Remove dead loop construct.
2389 2021-11-25 Richard Biener <rguenther@suse.de>
2391 * regset.h (REG_SET_EMPTY_P): New macro.
2392 * cfgcleanup.c (thread_jump): Use REG_SET_EMPTY_P.
2394 2021-11-25 Martin Liska <mliska@suse.cz>
2396 * doc/invoke.texi: Use @option for -Wuninitialized.
2398 2021-11-25 Aldy Hernandez <aldyh@redhat.com>
2400 PR tree-optimization/103254
2401 * gimple-range-path.cc (path_range_query::compute_ranges): Move
2402 exported boolean code...
2403 (path_range_query::compute_imports): ...here.
2405 2021-11-25 Aldy Hernandez <aldyh@redhat.com>
2407 PR tree-optimization/103254
2408 * gimple-range-path.cc (path_range_query::compute_ranges_defined): New
2409 (path_range_query::compute_ranges_in_block): Move to
2410 compute_ranges_defined.
2411 * gimple-range-path.h (compute_ranges_defined): New.
2413 2021-11-25 Jakub Jelinek <jakub@redhat.com>
2415 PR tree-optimization/103417
2416 * match.pd ((X & Y) CMP 0): Only handle eq and ne. Commonalize
2419 2021-11-25 Jakub Jelinek <jakub@redhat.com>
2421 PR tree-optimization/103376
2422 * gimple-ssa-store-merging.c (perform_symbolic_merge): For
2423 BIT_IOR_EXPR, if masked1 && masked2 && masked1 != masked2, don't
2424 punt, but set the corresponding result byte to MARKER_BYTE_UNKNOWN.
2425 For BIT_XOR_EXPR similarly and if masked1 == masked2 and the
2426 byte isn't MARKER_BYTE_UNKNOWN, set the corresponding result byte to
2429 2021-11-25 Jakub Jelinek <jakub@redhat.com>
2432 * doc/invoke.texi (-Wcomma-subscript): Document that for
2433 -std=c++20 the option isn't enabled by default with -Wno-deprecated
2434 but for -std=c++23 it is.
2436 2021-11-24 Jan Hubicka <hubicka@ucw.cz>
2438 * ipa-modref.c (implicit_eaf_flags_for_edge_and_arg): Break out from...
2439 (modref_merge_call_site_flags): ... here.
2440 (ipa_merge_modref_summary_after_inlining): Use it.
2442 2021-11-24 Thomas Schwinge <thomas@codesourcery.com>
2444 * cfgloop.c (verify_loop_structure): Reduce scope of
2445 'class loop *loop' variable.
2446 * ipa-fnsummary.c (analyze_function_body): Likewise.
2447 * loop-init.c (fix_loop_structure): Likewise.
2448 * loop-invariant.c (calculate_loop_reg_pressure): Likewise.
2449 * predict.c (predict_loops): Likewise.
2450 * tree-loop-distribution.c (loop_distribution::execute): Likewise.
2451 * tree-vectorizer.c (pass_vectorize::execute): Likewise.
2453 2021-11-24 Andrew MacLeod <amacleod@redhat.com>
2455 PR tree-optimization/103231
2456 * gimple-range.cc (gimple_ranger::gimple_ranger): Create stmt stack.
2457 (gimple_ranger::gimple_ranger): Delete stmt stack.
2458 (gimple_ranger::range_of_stmt): Process depenedencies if they have no
2460 (gimple_ranger::prefill_name): New.
2461 (gimple_ranger::prefill_stmt_dependencies): New.
2462 * gimple-range.h (class gimple_ranger): Add prototypes.
2464 2021-11-24 Andrew MacLeod <amacleod@redhat.com>
2466 * gimple-range-cache.cc (ranger_cache::get_global_range): Always
2467 return a range, return if it came from the cache or not.
2468 (get_non_stale_global_range): Rename to get_global_range, and return
2469 the temporal state in a flag.
2470 * gimple-range-cache.h (get_non_stale_global_range): Rename and adjust.
2471 * gimple-range.cc (gimple_ranger::range_of_expr): No need to query
2473 (gimple_ranger::range_of_stmt): Adjust for global cache temporal state
2476 2021-11-24 Andrew MacLeod <amacleod@redhat.com>
2478 * gimple-range.cc (gimple_ranger::range_on_edge): Call trailer when
2479 a constant is encountered to terminate the trace.
2481 2021-11-24 Maciej W. Rozycki <macro@embecosm.com>
2483 PR middle-end/103059
2484 * reload.c (find_reloads_address_1): Also accept the ASHIFT form
2485 of indexed addressing.
2486 (find_reloads): Adjust accordingly.
2488 2021-11-24 Richard Biener <rguenther@suse.de>
2489 Jan Hubicka <jh@suse.cz>
2491 PR tree-optimization/103168
2492 * ipa-modref.h (struct modref_summary): Add load_accesses.
2493 * ipa-modref.c (modref_summary::finalize): Initialize load_accesses.
2494 * tree-ssa-sccvn.c (visit_reference_op_call): Use modref
2495 info to walk the virtual use->def chain to CSE const/pure
2496 function calls possibly reading from memory.
2498 2021-11-24 Thomas Schwinge <thomas@codesourcery.com>
2500 PR middle-end/103244
2501 * gimplify.c (gimplify_adjust_omp_clauses): Restore previous
2504 2021-11-24 Richard Biener <rguenther@suse.de>
2506 PR middle-end/103193
2507 * match.pd: Avoid canonicalizing (le/ge @0 @0) to (eq @0 @0)
2508 with NaNs and -ftrapping-math.
2510 2021-11-24 Jakub Jelinek <jakub@redhat.com>
2512 PR middle-end/103384
2513 * omp-general.c (omp_context_selector_matches): For ACCEL_COMPILER,
2514 return 0 for kind(host) and continue for kind(nohost).
2516 2021-11-24 Jakub Jelinek <jakub@redhat.com>
2518 PR middle-end/103365
2519 * attribs.h (lookup_attribute): Allow attr_name to start with
2520 underscore, as long as canonicalize_attr_name returns false.
2521 (lookup_attribute_by_prefix): Don't call get_attribute_name twice.
2522 * attribs.c (extract_attribute_substring): Reimplement using
2523 canonicalize_attr_name.
2524 (register_scoped_attribute): Change gcc_assert into
2525 gcc_checking_assert, verify !canonicalize_attr_name rather than
2526 that str.str doesn't start with '_'.
2528 2021-11-24 Jakub Jelinek <jakub@redhat.com>
2530 PR tree-optimization/103376
2531 * gimple-ssa-store-merging.c (perform_symbolic_merge): Add CODE
2532 argument. If CODE is not BIT_IOR_EXPR, ensure that one of masked1
2534 (find_bswap_or_nop_1, find_bswap_or_nop,
2535 imm_store_chain_info::try_coalesce_bswap): Adjust
2536 perform_symbolic_merge callers.
2538 2021-11-24 Richard Biener <rguenther@suse.de>
2540 * tree-ssa-loop-ivopts.c (find_givs): Take loop body as
2541 argument instead of re-computing it.
2542 (find_interesting_uses): Likewise.
2543 (find_induction_variables): Pass through loop body.
2544 (tree_ssa_iv_optimize_loop): Pass down loop body.
2546 2021-11-24 Tamar Christina <tamar.christina@arm.com>
2548 * tree-ssa-phiopt.c (spaceship_replacement): Handle new canonical
2551 2021-11-24 Tamar Christina <tamar.christina@arm.com>
2553 * tree.c (bitmask_inv_cst_vector_p): New.
2554 * tree.h (bitmask_inv_cst_vector_p): New.
2555 * match.pd: Use it in new bitmask compare pattern.
2557 2021-11-24 Jason Merrill <jason@redhat.com>
2559 * timevar.h (class auto_cond_timevar): New.
2561 2021-11-24 Hongtao Liu <hongtao.liu@intel.com>
2562 H.J. Lu <hongjiu.lu@intel.com>
2564 PR tree-optimization/103194
2565 * match.pd (gimple_nop_atomic_bit_test_and_p): Extended to
2567 * tree-ssa-ccp.c (gimple_nop_convert): Declare.
2568 (optimize_atomic_bit_test_and): Enhance
2569 optimize_atomic_bit_test_and to handle truncation.
2571 2021-11-23 Martin Sebor <msebor@redhat.com>
2574 * Makefile.in (OBJS): Add gimple-warn-recursion.o.
2575 * common.opt: Add -Winfinite-recursion.
2576 * doc/invoke.texi (-Winfinite-recursion): Document.
2577 * passes.def (pass_warn_recursion): Schedule a new pass.
2578 * tree-pass.h (make_pass_warn_recursion): Declare.
2579 * gimple-warn-recursion.c: New file.
2581 2021-11-23 Bill Schmidt <wschmidt@linux.ibm.com>
2583 * config/rs6000/rs6000-call.c (rs6000_invalid_new_builtin): Change
2584 error messages for ENB_P8V and ENB_P9V.
2586 2021-11-23 Bill Schmidt <wschmidt@linux.ibm.com>
2588 * config/rs6000/rs6000-builtin-new.def: Add power6-64 stanza. Move
2589 CMPB to power6-64 stanza.
2590 * config/rs6000/rs6000-call.c (rs6000_invalid_new_builtin): Handle
2592 (rs6000_new_builtin_is_supported): Likewise.
2593 (rs6000_expand_new_builtin): Likewise. Clean up formatting.
2594 (rs6000_init_builtins): Handle ENB_P6_64 case.
2595 * config/rs6000/rs6000-gen-builtins.c (bif_stanza): Add BSTZ_P6_64.
2596 (stanza_map): Add entry mapping power6-64 to BSTZ_P6_64.
2597 (enable_string): Add "ENB_P6_64".
2598 (write_decls): Add ENB_P6_64 to bif_enable enum.
2600 2021-11-23 Jan Hubicka <jh@suse.cz>
2602 * ipa-modref-tree.h (struct modref_tree): Remove max_bases, max_refs
2604 (modref_tree::modref_tree): Remove parametr.
2605 (modref_tree::insert_base): Add max_bases parameter.
2606 (modref_tree::insert): Add max_bases, max_refs, max_accesses
2608 (modref_tree::insert): New member function.
2609 (modref_tree::merge): Add max_bases, max_refs, max_accesses
2611 (modref_tree::insert): New member function.
2612 * ipa-modref-tree.c (test_insert_search_collapse): Update.
2613 (test_merge): Update.
2614 * ipa-modref.c (dump_records): Don't dump max_refs and max_bases.
2615 (dump_lto_records): Likewise.
2616 (modref_summary::finalize): Fix whitespace.
2617 (get_modref_function_summary): Likewise.
2618 (modref_access_analysis::record_access): Update.
2619 (modref_access_analysis::record_access_lto): Update.
2620 (modref_access_analysis::process_fnspec): Update.
2621 (analyze_function): Update.
2622 (modref_summaries::duplicate): Update.
2623 (modref_summaries_lto::duplicate): Update.
2624 (write_modref_records): Update.
2625 (read_modref_records): Update.
2626 (read_section): Update.
2627 (propagate_unknown_call): Update.
2628 (modref_propagate_in_scc): Update.
2629 (ipa_merge_modref_summary_after_inlining): Update.
2631 2021-11-23 Martin Liska <mliska@suse.cz>
2633 * doc/invoke.texi: Remove 2 more duplicite param descriptions.
2635 2021-11-23 Richard Biener <rguenther@suse.de>
2637 PR tree-optimization/103361
2638 * gimple-loop-jam.c (adjust_unroll_factor): Use lambda_int
2639 for the dependence distance.
2640 * tree-data-ref.c (print_lambda_vector): Properly print a lambda_int.
2642 2021-11-23 Jakub Jelinek <jakub@redhat.com>
2644 * tree-inline.h (struct copy_body_data): Remove
2645 transform_lang_insert_block member.
2646 * tree-inline.c (remap_block): Don't call
2647 id->transform_lang_insert_block.
2648 (optimize_inline_calls, copy_gimple_seq_and_replace_locals,
2649 tree_function_versioning, maybe_inline_call_in_expr,
2650 copy_fn): Don't initialize id.transform_lang_insert_block.
2651 * gimplify.c (gimplify_omp_loop): Likewise.
2653 2021-11-23 Jan Hubicka <hubicka@ucw.cz>
2655 PR tree-optimization/103335
2656 * tree-ssa-dse.c (valid_ao_ref_for_dse): Rename to ...
2657 (valid_ao_ref_kill_for_dse): ... this; do not check that boundaries
2658 are divisible by BITS_PER_UNIT.
2659 (get_byte_aligned_range_containing_ref): New function.
2660 (get_byte_aligned_range_contained_in_ref): New function.
2661 (normalize_ref): Rename to ...
2662 (get_byte_range): ... this one; handle accesses not aligned to byte
2663 boundary; return range in bytes rater than updating ao_ref.
2664 (clear_live_bytes_for_ref): Take write ref by reference; simplify using
2666 (setup_live_bytes_from_ref): Likewise.
2667 (clear_bytes_written_by): Update.
2668 (live_bytes_read): Update.
2669 (dse_classify_store): Simplify tech before live_bytes_read checks.
2671 2021-11-23 Andrew Pinski <apinski@marvell.com>
2673 PR tree-optimization/102216
2674 * gimple-fold.c (fold_stmt_1): Add canonicalization
2675 of "&MEM[ssa_n, CST]" to "ssa_n p+ CST", note this
2676 can only be done if !in_place.
2678 2021-11-23 Jakub Jelinek <jakub@redhat.com>
2680 PR middle-end/102431
2681 * gimplify.c (replace_reduction_placeholders): Remove.
2682 (note_no_context_vars): New function.
2683 (gimplify_omp_loop): For OMP_PARALLEL's BIND_EXPR create a new
2684 BLOCK. Use copy_tree_body_r with walk_tree instead of unshare_expr
2685 and replace_reduction_placeholders for duplication of
2686 OMP_CLAUSE_REDUCTION_{INIT,MERGE} expressions. Ensure all mentioned
2687 automatic vars have DECL_CONTEXT set to non-NULL before doing so
2688 and reset it afterwards for those vars and their corresponding
2691 2021-11-23 Haochen Gui <guihaoc@gcc.gnu.org>
2694 * config/rs6000/altivec.md (altivec_vreve<mode>2 for VEC_K): Use
2695 xxbrq for v16qi, xxbrq + xxbrh for v8hi and xxbrq + xxbrw for v4si
2696 or v4sf when p9_vector is set.
2697 (altivec_vreve<mode>2 for VEC_64): Defined. Implemented by xxswapd.
2699 2021-11-23 Navid Rahimi <navidrahimi@microsoft.com>
2701 PR tree-optimization/102232
2702 * match.pd (x * (1 + y / x) - y) -> (x - y % x): New optimization.
2704 2021-11-23 Navid Rahimi <navidrahimi@microsoft.com>
2706 PR tree-optimization/96779
2707 * match.pd (-x == x) -> (x == 0): New optimization.
2709 2021-11-22 Roger Sayle <roger@nextmovesoftware.com>
2711 PR tree-optimization/98953
2712 PR tree-optimization/103345
2713 * gimple-ssa-store-merging.c (find_bswap_or_nop_1): Handle
2714 BIT_XOR_EXPR and PLUS_EXPR the same as BIT_IOR_EXPR.
2715 (pass_optimize_bswap::execute): Likewise.
2717 2021-11-22 Martin Liska <mliska@suse.cz>
2719 * doc/invoke.texi: Remove duplicate documentation for 3 params.
2721 2021-11-22 Richard Biener <rguenther@suse.de>
2723 PR tree-optimization/103351
2724 * tree-ssa-dce.c (sort_phi_args): Sort after e->dest_idx as
2727 2021-11-22 liuhongt <hongtao.liu@intel.com>
2730 * config/i386/constraints.md (Bk): New
2731 define_memory_constraint.
2732 * config/i386/i386-protos.h (ix86_gpr_tls_address_pattern_p):
2734 * config/i386/i386.c (ix86_gpr_tls_address_pattern_p): New
2736 * config/i386/i386.md (*movsi_internal): Don't allow
2737 mask/sse/mmx move in TLS code sequences.
2738 (*movdi_internal): Ditto.
2740 2021-11-22 Kewen Lin <linkw@linux.ibm.com>
2742 * config/xtensa/xtensa.md (movdi_internal, movdf_internal): Fix split
2745 2021-11-21 Jakub Jelinek <jakub@redhat.com>
2748 * config/i386/i386-options.c (ix86_valid_target_attribute_p): If
2749 fndecl already has DECL_FUNCTION_SPECIFIC_TARGET, use that as base
2750 instead of target_option_default_node.
2752 2021-11-21 Jan Hubicka <hubicka@ucw.cz>
2755 * ipa-modref.c (parm_map_for_arg): Rename to ...
2756 (parm_map_for_ptr): .. this one; handle static chain and calls to
2758 (modref_access_analysis::get_access): Use parm_map_for_ptr.
2759 (modref_access_analysis::process_fnspec): Update.
2760 (modref_access_analysis::analyze_load): Update.
2761 (modref_access_analysis::analyze_store): Update.
2763 2021-11-21 Jan Hubicka <hubicka@ucw.cz>
2765 * ipa-modref.c (ignore_nondeterminism_p): Move earlier in source
2767 (ignore_retval_p): Likewise.
2768 (ignore_stores_p): Likewise.
2769 (parm_map_for_arg): Likewise.
2770 (class modref_access_analysis): New class.
2771 (modref_access_analysis::set_side_effects): New member function.
2772 (modref_access_analysis::set_nondeterministic): New member function.
2773 (get_access): Turn to ...
2774 (modref_access_analysis::get_access): ... this one.
2775 (record_access): Turn to ...
2776 (modref_access_analysis::record_access): ... this one.
2777 (record_access_lto): Turn to ...
2778 (modref_access_analysis::record_access_lto): ... This one.
2779 (record_access_p): Turn to ...
2780 (modref_access_analysis::record_access_p): ... This one
2781 (modref_access_analysis::record_unknown_load): New member function.
2782 (modref_access_analysis::record_unknown_store): New member function.
2783 (get_access_for_fnspec): Turn to ...
2784 (modref_access_analysis::get_access_for_fnspec): ... this one.
2785 (merge_call_side_effects): Turn to ...
2786 (moderf_access_analysis::merge_call_side_effects): Turn to ...
2787 (collapse_loads): Move later in source code.
2788 (collapse_stores): Move later in source code.
2789 (process_fnspec): Turn to ...
2790 (modref_access_analysis::process_fnspec): ... this one.
2791 (analyze_call): Turn to ...
2792 (modref_access_analysis::analyze_call): ... this one.
2793 (struct summary_ptrs): Remove.
2794 (analyze_load): Turn to ...
2795 (modref_access_analysis::analyze_load): ... this one.
2796 (analyze_store): Turn to ...
2797 (modref_access_analysis::analyze_store): ... this one.
2798 (analyze_stmt): Turn to ...
2799 (modref_access_analysis::analyze_stmt): ... This one.
2800 (remove_summary): Remove.
2801 (modref_access_analysis::propagate): Break out from ...
2802 (modref_access_analysis::analyze): Break out from ...
2803 (analyze_function): ... here.
2805 2021-11-21 Roger Sayle <roger@nextmovesoftware.com>
2806 Robin Dapp <rdapp@linux.ibm.com>
2809 * tree-ssa-math-opts.c (convert_mult_to_widen): Recognize
2810 signed WIDEN_MULT_EXPR if the target supports umul_widen_optab.
2812 2021-11-20 Jan Hubicka <hubicka@ucw.cz>
2815 * ipa-modref.c (ignore_nondeterminism_p): Allow looping pure/cont.
2816 (merge_call_side_effects): Improve debug output.
2818 2021-11-20 Jan Hubicka <hubicka@ucw.cz>
2821 * ipa-pure-const.c (propagate_pure_const): Fix merging of loping flag.
2823 2021-11-20 Jeff Law <jeffreyalaw@gmail.com>
2825 PR tree-optimization/103226
2826 * config/bfin/bfin.md (doloop pattern, splitter and expander): Clobber
2829 2021-11-20 Andrew Pinski <apinski@marvell.com>
2831 PR tree-optimization/103220
2832 * match.pd ((type) X bitop CST): Don't check if CST
2833 fits into the type if only the sign changes.
2835 2021-11-20 Alexandre Oliva <oliva@adacore.com>
2837 PR tree-optimization/102988
2838 * gimple-harden-conditionals.cc (detach_value): Copy SSA_NAME
2839 without decl sharing.
2841 2021-11-19 Iain Sandoe <iain@sandoe.co.uk>
2843 * doc/invoke.texi: Remove whitespace after an @option.
2845 2021-11-19 Paul A. Clarke <pc@us.ibm.com>
2847 * config/rs6000/emmintrin.h (_mm_sad_epu8): Use vec_absd when
2848 _ARCH_PWR9, optimize vec_sum2s when LE.
2850 2021-11-19 Iain Sandoe <iain@sandoe.co.uk>
2853 * config/darwin-driver.c (darwin_driver_init): Handle exported
2854 symbols and symbol lists (suppress automatic export of the TLS
2856 * config/darwin.c (darwin_rename_builtins): Remove workaround.
2857 * config/darwin.h (LINK_GCC_C_SEQUENCE_SPEC): Likewise.
2858 (REAL_LIBGCC_SPEC): Handle revised library uses.
2859 * config/darwin.opt (nodefaultexport): New.
2860 * config/i386/darwin.h (PR80556_WORKAROUND): Remove.
2861 * config/i386/darwin32-biarch.h (PR80556_WORKAROUND): Likewise.
2862 * config/i386/darwin64-biarch.h (PR80556_WORKAROUND): Likewise.
2864 2021-11-19 Martin Jambor <mjambor@suse.cz>
2866 * opts.c (default_options_table): Switch off
2867 flag_semantic_interposition at Ofast.
2868 * doc/invoke.texi (Optimize Options): Document that Ofast switches off
2869 -fsemantic-interposition.
2871 2021-11-19 Jan Hubicka <hubicka@ucw.cz>
2873 * tree-ssa-alias.c (ref_maybe_used_by_call_p_1): Do not guard modref
2874 by !gimple_call_chain.
2876 2021-11-19 Martin Sebor <msebor@redhat.com>
2880 * doc/invoke.texi (-Waddress): Update.
2882 2021-11-19 Andrew MacLeod <amacleod@redhat.com>
2884 PR tree-optimization/103254
2885 * gimple-range-gori.cc (range_def_chain::get_def_chain): Limit the
2886 depth for all statements with multple ssa names.
2888 2021-11-19 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
2890 * config/s390/s390.md (define_peephole2): Variable insn points
2891 to the first matched insn. Use peep2_next_insn(1) to refer to
2892 the second matched insn.
2894 2021-11-19 Tamar Christina <tamar.christina@arm.com>
2896 PR tree-optimization/103311
2898 * tree-vect-slp-patterns.c (vect_validate_multiplication): Fix CONJ
2899 test to new codegen.
2900 (complex_mul_pattern::matches): Move check downwards.
2902 2021-11-19 Martin Liska <mliska@suse.cz>
2905 2021-11-19 Martin Liska <mliska@suse.cz>
2907 * cfgexpand.c (pass_expand::execute): Use option directly.
2908 * function.c (allocate_struct_function): Likewise.
2909 * gimple-low.c (lower_function_body): Likewise.
2910 (lower_stmt): Likewise.
2911 * gimple-ssa-backprop.c (backprop::prepare_change): Likewise.
2912 * ipa-param-manipulation.c (ipa_param_adjustments::modify_call): Likewise.
2913 * ipa-split.c (split_function): Likewise.
2914 * lto-streamer-in.c (input_function): Likewise.
2915 * sese.c (sese_insert_phis_for_liveouts): Likewise.
2916 * ssa-iterators.h (num_imm_uses): Likewise.
2917 * tree-cfg.c (make_blocks): Likewise.
2918 (gimple_merge_blocks): Likewise.
2919 * tree-inline.c (tree_function_versioning): Likewise.
2920 * tree-loop-distribution.c (generate_loops_for_partition): Likewise.
2921 * tree-sra.c (analyze_access_subtree): Likewise.
2922 * tree-ssa-dce.c (remove_dead_stmt): Likewise.
2923 * tree-ssa-loop-ivopts.c (remove_unused_ivs): Likewise.
2924 * tree-ssa-phiopt.c (spaceship_replacement): Likewise.
2925 * tree-ssa-reassoc.c (reassoc_remove_stmt): Likewise.
2926 * tree-ssa-tail-merge.c (tail_merge_optimize): Likewise.
2927 * tree-ssa-threadedge.c (propagate_threaded_block_debug_into): Likewise.
2928 * tree-ssa.c (gimple_replace_ssa_lhs): Likewise.
2929 (target_for_debug_bind): Likewise.
2930 (insert_debug_temp_for_var_def): Likewise.
2931 (insert_debug_temps_for_defs): Likewise.
2932 (reset_debug_uses): Likewise.
2933 * tree-ssanames.c (release_ssa_name_fn): Likewise.
2934 * tree-vect-loop-manip.c (adjust_vec_debug_stmts): Likewise.
2935 (adjust_debug_stmts): Likewise.
2936 (adjust_phi_and_debug_stmts): Likewise.
2937 (vect_do_peeling): Likewise.
2938 * tree-vect-loop.c (vect_transform_loop_stmt): Likewise.
2939 (vect_transform_loop): Likewise.
2940 * tree.h (MAY_HAVE_DEBUG_MARKER_STMTS): Remove
2941 (MAY_HAVE_DEBUG_BIND_STMTS): Remove.
2942 (MAY_HAVE_DEBUG_STMTS): Use options directly.
2944 2021-11-19 Giuliano Belinassi <gbelinassi@suse.de>
2946 * gcc.c (process_command): Skip dumpdir override if file is a
2948 * doc/invoke.texi: Update -dumpdir documentation.
2950 2021-11-19 Andrew Pinski <apinski@marvell.com>
2952 PR tree-optimization/103314
2953 * match.pd ((type) X op CST): Restrict the equal
2954 TYPE_PRECISION case to GIMPLE only.
2956 2021-11-19 Martin Liska <mliska@suse.cz>
2959 * ipa-modref-tree.h (struct modref_parm_map): Add default
2961 * ipa-modref.c (ipa_merge_modref_summary_after_inlining): Use it.
2963 2021-11-19 Richard Biener <rguenther@suse.de>
2965 PR middle-end/103248
2966 * tree-eh.c (operation_could_trap_helper_p): Properly handle
2967 fixed-point RDIV_EXPR.
2969 2021-11-19 Richard Biener <rguenther@suse.de>
2971 PR tree-optimization/102436
2972 * tree-ssa-loop-im.c (execute_sm_if_changed): Add mode
2973 to just create the if structure and return the then block.
2974 (execute_sm): Add flag to indicate the var will re-use
2976 (hoist_memory_references): Support a single conditional
2977 block with all stores as special case.
2979 2021-11-19 Andrew Pinski <apinski@marvell.com>
2981 PR tree-optimization/103317
2982 * tree-ssa-phiopt.c (minmax_replacement): For the non empty
2983 middle bb case, check to make sure it has a single predecessor.
2985 2021-11-19 Andrew Pinski <apinski@marvell.com>
2987 PR tree-optimization/103257
2989 ((m1 >/</>=/<= m2) * d -> (m1 >/</>=/<= m2) ? d : 0):
2990 Disable until !canonicalize_math_p ().
2992 2021-11-19 Marek Polacek <polacek@redhat.com>
2996 * doc/invoke.texi: Update documentation for -Wuninitialized.
2997 * tree.c (stabilize_reference): Set location.
2999 2021-11-19 liuhongt <hongtao.liu@intel.com>
3002 * config/i386/x86-tune-costs.h (skylake_cost): Reduce cost of
3003 storing 256/512-bit SSE register to be equal to cost of
3004 unaligned store to avoid odd alignment peeling.
3005 (icelake_cost): Ditto.
3007 2021-11-18 David Edelsohn <dje.gcc@gmail.com>
3009 * config/rs6000/predicates.md (current_file_function_operand):
3010 Add flag_semantic_interposition to call of decl_replaceable_p.
3012 2021-11-18 Martin Liska <mliska@suse.cz>
3014 * ipa-modref.c (analyze_function): Do not execute the code
3015 only if dump_file != NULL.
3017 2021-11-18 Martin Liska <mliska@suse.cz>
3019 * ipa-modref.c (analyze_function): Use fnode instead of repeated
3020 cgraph_node::get (current_function_decl).
3022 2021-11-18 Jan Hubicka <hubicka@ucw.cz>
3024 * cgraph.c (cgraph_node::get_availability): Update call of
3026 (cgraph_node::verify_node): Verify that semantic_interposition flag
3028 * cgraph.h: (symtab_node): Add semantic_interposition flag.
3029 * cgraphclones.c (set_new_clone_decl_and_node_flags): Clear
3030 semantic_interposition flag.
3031 * cgraphunit.c (cgraph_node::finalize_function): Set
3032 semantic_interposition flag.
3033 (cgraph_node::add_new_function): Likewise.
3034 (varpool_node::finalize_decl): Likewise.
3035 (cgraph_node::create_wrapper): Likewise.
3036 * common.opt (fsemantic-interposition): Turn to optimization node.
3037 * lto-cgraph.c (lto_output_node): Stream semantic_interposition.
3038 (lto_output_varpool_node): Likewise.
3039 (input_overwrite_node): Likewise.
3040 (input_varpool_node): Likewise.
3041 * symtab.c (symtab_node::dump_base): Dump new flag.
3042 * varasm.c (decl_replaceable_p): Add semantic_interposition_p
3044 * varasm.h (decl_replaceable_p): Update declaration.
3045 * varpool.c (varpool_node::ctor_useable_for_folding_p):
3046 Use semantic_interposition flag.
3047 (varpool_node::get_availability): Likewise.
3048 (varpool_node::create_alias): Copy semantic_interposition flag.
3050 2021-11-18 Jan Hubicka <hubicka@ucw.cz>
3053 * ipa-modref.c (modref_eaf_analysis::merge_call_lhs_flags): Unused
3054 parameter may still be returned.
3055 (modref_eaf_analysis::analyze_ssa_name): Call merge_call_lhs_flags
3056 even for unused function args.
3058 2021-11-18 Tamar Christina <tamar.christina@arm.com>
3060 PR tree-optimization/103311
3061 * tree-vect-slp-patterns.c (complex_mul_pattern::matches,
3062 complex_fms_pattern::matches): Check for multiplications.
3064 2021-11-18 H.J. Lu <hjl.tools@gmail.com>
3067 * config/i386/i386.c (ix86_output_jmp_thunk_or_indirect): Emit
3068 CS prefix for -mindirect-branch-cs-prefix.
3069 (ix86_output_indirect_branch_via_reg): Likewise.
3070 * config/i386/i386.opt: Add -mindirect-branch-cs-prefix.
3071 * doc/invoke.texi: Document -mindirect-branch-cs-prefix.
3073 2021-11-18 David Edelsohn <dje.gcc@gmail.com>
3075 * config/rs6000/driver-rs6000.c (detect_processor_aix): Add
3078 2021-11-18 Tamar Christina <tamar.christina@arm.com>
3080 PR tree-optimization/103253
3081 * tree-ssa-math-opts.c (convert_mult_to_fma): Check for LHS.
3083 2021-11-18 Matthias Kretz <m.kretz@gsi.de>
3085 * doc/extend.texi: Document __builtin_assoc_barrier.
3087 2021-11-18 Martin Liska <mliska@suse.cz>
3089 * cfgexpand.c (pass_expand::execute): Use option directly.
3090 * function.c (allocate_struct_function): Likewise.
3091 * gimple-low.c (lower_function_body): Likewise.
3092 (lower_stmt): Likewise.
3093 * gimple-ssa-backprop.c (backprop::prepare_change): Likewise.
3094 * ipa-param-manipulation.c (ipa_param_adjustments::modify_call): Likewise.
3095 * ipa-split.c (split_function): Likewise.
3096 * lto-streamer-in.c (input_function): Likewise.
3097 * sese.c (sese_insert_phis_for_liveouts): Likewise.
3098 * ssa-iterators.h (num_imm_uses): Likewise.
3099 * tree-cfg.c (make_blocks): Likewise.
3100 (gimple_merge_blocks): Likewise.
3101 * tree-inline.c (tree_function_versioning): Likewise.
3102 * tree-loop-distribution.c (generate_loops_for_partition): Likewise.
3103 * tree-sra.c (analyze_access_subtree): Likewise.
3104 * tree-ssa-dce.c (remove_dead_stmt): Likewise.
3105 * tree-ssa-loop-ivopts.c (remove_unused_ivs): Likewise.
3106 * tree-ssa-phiopt.c (spaceship_replacement): Likewise.
3107 * tree-ssa-reassoc.c (reassoc_remove_stmt): Likewise.
3108 * tree-ssa-tail-merge.c (tail_merge_optimize): Likewise.
3109 * tree-ssa-threadedge.c (propagate_threaded_block_debug_into): Likewise.
3110 * tree-ssa.c (gimple_replace_ssa_lhs): Likewise.
3111 (target_for_debug_bind): Likewise.
3112 (insert_debug_temp_for_var_def): Likewise.
3113 (insert_debug_temps_for_defs): Likewise.
3114 (reset_debug_uses): Likewise.
3115 * tree-ssanames.c (release_ssa_name_fn): Likewise.
3116 * tree-vect-loop-manip.c (adjust_vec_debug_stmts): Likewise.
3117 (adjust_debug_stmts): Likewise.
3118 (adjust_phi_and_debug_stmts): Likewise.
3119 (vect_do_peeling): Likewise.
3120 * tree-vect-loop.c (vect_transform_loop_stmt): Likewise.
3121 (vect_transform_loop): Likewise.
3122 * tree.h (MAY_HAVE_DEBUG_MARKER_STMTS): Remove
3123 (MAY_HAVE_DEBUG_BIND_STMTS): Remove.
3124 (MAY_HAVE_DEBUG_STMTS): Use options directly.
3126 2021-11-18 Richard Biener <rguenther@suse.de>
3128 PR tree-optimization/103277
3129 * tree-ssa-dse.c (need_ab_cleanup): New.
3130 (dse_optimize_redundant_stores): Adjust.
3131 (delete_dead_or_redundant_assignment): Get extra
3132 need_ab_cleanup argument and set when abnormal cleanup is
3134 (dse_optimize_call): Adjust.
3135 (dse_optimize_stmt): Likewise.
3136 (pass_dse::execute): Allocate and deallocate need_ab_cleanup.
3137 Perform abnormal cleanup.
3138 * tree-ssa-dse.h (delete_dead_or_redundant_assignment): Adjust.
3140 2021-11-18 Hongyu Wang <hongyu.wang@intel.com>
3142 * config/i386/i386-expand.c (ix86_expand_atomic_fetch_op_loop):
3143 Adjust generated cfg to avoid infinite loop.
3145 2021-11-18 konglin1 <lingling.kong@intel.com>
3147 * config/i386/avx512fp16intrin.h (_mm512_mul_pch): Add alias for _mm512_fmul_pch.
3148 (_mm512_mask_mul_pch): Likewise.
3149 (_mm512_maskz_mul_pch): Likewise.
3150 (_mm512_mul_round_pch): Likewise.
3151 (_mm512_mask_mul_round_pch): Likewise.
3152 (_mm512_maskz_mul_round_pch): Likewise.
3153 (_mm512_cmul_pch): Likewise.
3154 (_mm512_mask_cmul_pch): Likewise.
3155 (_mm512_maskz_cmul_pch): Likewise.
3156 (_mm512_cmul_round_pch): Likewise.
3157 (_mm512_mask_cmul_round_pch): Likewise.
3158 (_mm512_maskz_cmul_round_pch): Likewise.
3159 (_mm_mul_sch): Likewise.
3160 (_mm_mask_mul_sch): Likewise.
3161 (_mm_maskz_mul_sch): Likewise.
3162 (_mm_mul_round_sch): Likewise.
3163 (_mm_mask_mul_round_sch): Likewise.
3164 (_mm_maskz_mul_round_sch): Likewise.
3165 (_mm_cmul_sch): Likewise.
3166 (_mm_mask_cmul_sch): Likewise.
3167 (_mm_maskz_cmul_sch): Likewise.
3168 (_mm_cmul_round_sch): Likewise.
3169 (_mm_mask_cmul_round_sch): Likewise.
3170 (_mm_maskz_cmul_round_sch): Likewise.
3171 * config/i386/avx512fp16vlintrin.h (_mm_mul_pch): Likewise.
3172 (_mm_mask_mul_pch): Likewise.
3173 (_mm_maskz_mul_pch): Likewise.
3174 (_mm256_mul_pch): Likewise.
3175 (_mm256_mask_mul_pch): Likewise.
3176 (_mm256_maskz_mul_pch): Likewise.
3177 (_mm_cmul_pch): Likewise.
3178 (_mm_mask_cmul_pch): Likewise.
3179 (_mm_maskz_cmul_pch): Likewise.
3180 (_mm256_cmul_pch): Likewise.
3181 (_mm256_mask_cmul_pch): Likewise.
3182 (_mm256_maskz_cmul_pch): Likewise.
3184 2021-11-17 Andrew Pinski <apinski@marvell.com>
3186 PR tree-optimization/103228
3187 PR tree-optimization/55177
3188 * match.pd ((type) X bitop CST): Also do this
3189 transformation for nop conversions.
3191 2021-11-17 Martin Sebor <msebor@redhat.com>
3193 PR tree-optimization/102759
3194 * gimple-array-bounds.cc (build_printable_array_type): Move...
3195 * gimple-ssa-warn-access.cc (build_printable_array_type): Avoid
3196 pathological function redeclarations that remove a previously
3198 Improve formatting of function arguments in informational notes.
3199 * pointer-query.cc (build_printable_array_type): ...to here.
3200 * pointer-query.h (build_printable_array_type): Declared.
3202 2021-11-17 H.J. Lu <hjl.tools@gmail.com>
3205 * config/i386/i386-opts.h (harden_sls): New enum.
3206 * config/i386/i386.c (output_indirect_thunk): Mitigate against
3207 SLS for function return.
3208 (ix86_output_function_return): Likewise.
3209 (ix86_output_jmp_thunk_or_indirect): Mitigate against indirect
3211 (ix86_output_indirect_jmp): Likewise.
3212 (ix86_output_call_insn): Likewise.
3213 * config/i386/i386.opt: Add -mharden-sls=.
3214 * doc/invoke.texi: Document -mharden-sls=.
3216 2021-11-17 H.J. Lu <hjl.tools@gmail.com>
3219 * config/i386/i386.c (ix86_code_end): Remove "%!" before ret.
3220 (ix86_output_function_return): Likewise.
3221 * config/i386/i386.md (simple_return_pop_internal): Likewise.
3223 2021-11-17 Jan Hubicka <hubicka@ucw.cz>
3226 * ipa-modref.c (read_modref_records): Fix streaminig in of every_access
3229 2021-11-17 Uroš Bizjak <ubizjak@gmail.com>
3231 * config/i386/i386.c (indirect_thunks_used): Redefine as HARD_REG_SET.
3232 (ix86_code_end): Use TEST_HARD_REG_BIT on indirect_thunks_used.
3233 (ix86_output_indirect_branch_via_reg): Use SET_HARD_REG_BIT
3234 on indirect_thunks_used.
3235 (ix86_output_indirect_function_return): Ditto.
3237 2021-11-17 Jan Hubicka <hubicka@ucw.cz>
3239 * ipa-modref-tree.c: Include cgraph.h and tree-streamer.h.
3240 (modref_access_node::stream_out): New member function.
3241 (modref_access_node::stream_in): New member function.
3242 * ipa-modref-tree.h (modref_access_node::stream_out,
3243 modref_access_node::stream_in): Declare.
3244 * ipa-modref.c (modref_summary_lto::useful_p): Free useless kills.
3245 (modref_summary_lto::dump): Dump kills.
3246 (analyze_store): Record kills for LTO
3247 (analyze_stmt): Likewise.
3248 (modref_summaries_lto::duplicate): Duplicate kills.
3249 (write_modref_records): Use new stream_out member function.
3250 (read_modref_records): Likewise.
3251 (modref_write): Stream out kills.
3252 (read_section): Stream in kills
3253 (remap_kills): New function.
3254 (update_signature): Use it.
3256 2021-11-17 Uroš Bizjak <ubizjak@gmail.com>
3258 * config/i386/i386.h (LEGACY_SSE_REGNO_P): New predicate.
3259 (SSE_REGNO_P): Use LEGACY_SSE_REGNO_P predicate.
3260 * config/i386/i386.c (zero_all_vector_registers):
3261 Use LEGACY_SSE_REGNO_P predicate.
3262 (ix86_register_priority): Use REX_INT_REGNO_P, REX_SSE_REGNO_P
3263 and EXT_REG_SSE_REGNO_P predicates.
3264 (ix86_hard_regno_call_part_clobbered): Use REX_SSE_REGNO_P
3265 and LEGACY_SSE_REGNO_P predicates.
3267 2021-11-17 Jason Merrill <jason@redhat.com>
3269 * doc/invoke.texi (C++ Dialect Options): Document
3270 -fimplicit-constexpr.
3272 2021-11-17 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
3274 * config/aarch64/aarch64-modes.def (VECTOR_MODE): New V8DI mode.
3275 * config/aarch64/aarch64.c (aarch64_hard_regno_mode_ok): Handle
3277 * config/aarch64/iterators.md (define_mode_attr nunits): Add entry
3280 2021-11-17 Martin Uecker <uecker@gcc.gnu.org>
3284 * gimplify.c (gimplify_var_or_parm_decl): Update comment.
3285 (gimplify_compound_lval): Gimplify base expression first.
3286 (gimplify_target_expr): Add comment.
3288 2021-11-17 Jakub Jelinek <jakub@redhat.com>
3290 PR tree-optimization/103192
3291 * tree-ssa-loop-im.c (move_computations_worker): Use
3292 reset_flow_sensitive_info instead of manually clearing
3293 SSA_NAME_RANGE_INFO and do it for all SSA_NAMEs, not just ones
3294 with integral types.
3296 2021-11-17 Jakub Jelinek <jakub@redhat.com>
3298 PR tree-optimization/103255
3299 * gimple-range-fold.cc (fold_using_range::range_of_address): Return
3300 range_nonzero rather than unadjusted base's range. Formatting fixes.
3302 2021-11-17 Richard Sandiford <richard.sandiford@arm.com>
3304 * doc/md.texi (cond_fmin@var{mode}, cond_fmax@var{mode}): Document.
3305 * optabs.def (cond_fmin_optab, cond_fmax_optab): New optabs.
3306 * internal-fn.def (COND_FMIN, COND_FMAX): New functions.
3307 * internal-fn.c (first_commutative_argument): Handle them.
3308 (FOR_EACH_COND_FN_PAIR): Likewise.
3309 * match.pd (UNCOND_BINARY, COND_BINARY): Likewise.
3310 * config/aarch64/aarch64-sve.md (cond_<fmaxmin><mode>): New
3313 2021-11-17 Kewen Lin <linkw@linux.ibm.com>
3315 * config/i386/i386.md (*add<dwi>3_doubleword, *addv<dwi>4_doubleword,
3316 *addv<dwi>4_doubleword_1, *sub<dwi>3_doubleword,
3317 *subv<dwi>4_doubleword, *subv<dwi>4_doubleword_1,
3318 *add<dwi>3_doubleword_cc_overflow_1, *divmodsi4_const,
3319 *neg<dwi>2_doubleword, *tls_dynamic_gnu2_combine_64_<mode>): Fix split
3322 2021-11-17 Andrew Pinski <apinski@marvell.com>
3324 PR tree-optimization/103288
3325 * tree-ssa-phiopt.c (value_replacement): Return early if middle
3326 block has more than one pred.
3328 2021-11-17 Kewen Lin <linkw@linux.ibm.com>
3330 * config/visium/visium.md (*add<mode>3_insn, *addsi3_insn, *addi3_insn,
3331 *sub<mode>3_insn, *subsi3_insn, *subdi3_insn, *neg<mode>2_insn,
3332 *negdi2_insn, *and<mode>3_insn, *ior<mode>3_insn, *xor<mode>3_insn,
3333 *one_cmpl<mode>2_insn, *ashl<mode>3_insn, *ashr<mode>3_insn,
3334 *lshr<mode>3_insn, *trunchiqi2_insn, *truncsihi2_insn,
3335 *truncdisi2_insn, *extendqihi2_insn, *extendqisi2_insn,
3336 *extendhisi2_insn, *extendsidi2_insn, *zero_extendqihi2_insn,
3337 *zero_extendqisi2_insn, *zero_extendsidi2_insn): Fix split condition.
3339 2021-11-17 Marek Polacek <polacek@redhat.com>
3341 PR preprocessor/103026
3342 * doc/invoke.texi: Document -Wbidi-chars.
3344 2021-11-17 Jan Hubicka <hubicka@ucw.cz>
3347 * ipa-modref.c (ipa_merge_modref_summary_after_inlining): Fix clearing
3350 2021-11-16 Jan Hubicka <jh@suse.cz>
3352 * ipa-modref.c (get_modref_function_summary): Declare.
3353 * ipa-modref.h (get_modref_function_summary): New function.
3354 * tree-ssa-dse.c (clear_live_bytes_for_ref): Break out from ...
3355 (clear_bytes_written_by): ... here; also clear memory killed by
3358 2021-11-16 Iain Sandoe <iain@sandoe.co.uk>
3360 * ggc-common.c (gt_pch_save): If we cannot find a suitable
3361 memory segment for save, then error-out, do not try to
3363 (gt_pch_restore): Save the existing line table, and when
3364 the replacement is being read, use that when constructing
3367 2021-11-16 Peter Bergner <bergner@linux.ibm.com>
3370 * config/rs6000/mma.md (*vsx_assemble_pair): Add early-clobber for
3372 (*mma_assemble_acc): Likewise.
3374 2021-11-16 Martin Sebor <msebor@redhat.com>
3376 PR tree-optimization/102960
3377 * gimple-fold.c (get_range_strlen): Take bitmap as an argument rather
3378 than a pointer to it.
3379 (get_range_strlen_tree): Same. Remove bitmap allocation. Use
3381 (get_maxval_strlen): Use an auto_bitmap.
3382 * tree-ssa-strlen.c (get_range_strlen_dynamic): Factor out PHI
3384 (get_range_strlen_phi): ...into this function.
3385 Avoid assuming maximum string length is constant
3386 (printf_strlen_execute): Dump pointer query cache contents when
3387 details are requisted.
3389 2021-11-16 Jason Merrill <jason@redhat.com>
3391 * langhooks.h (struct lang_hooks): Adjust comment.
3392 * print-tree.c (print_node): Also call print_xnode hook for
3395 2021-11-16 Andrew Pinski <apinski@marvell.com>
3397 PR tree-optimization/103218
3398 * match.pd: New pattern for "((type)(a<0)) << SIGNBITOFA".
3400 2021-11-16 Claudiu Zissulescu <claziss@synopsys.com>
3402 * config/arc/arc.md (maddhisi4): Use a single move to accumulator.
3403 (umaddhisi4): Likewise.
3404 (machi): Update pattern.
3407 2021-11-16 Richard Biener <rguenther@suse.de>
3409 PR tree-optimization/102880
3410 * tree-ssa-dce.c (sort_phi_args): New function.
3411 (make_forwarders_with_degenerate_phis): Likewise.
3412 (perform_tree_ssa_dce): Call
3413 make_forwarders_with_degenerate_phis.
3415 2021-11-16 Richard Biener <rguenther@suse.de>
3417 PR tree-optimization/102880
3418 * tree-ssa-phiopt.c (tree_ssa_phiopt_worker): Push
3419 single_pred (bb1) condition to places that really need it.
3420 (match_simplify_replacement): Likewise.
3421 (value_replacement): Likewise.
3422 (replace_phi_edge_with_variable): Deal with extra edges
3425 2021-11-16 Martin Jambor <mjambor@suse.cz>
3427 * cfgexpand.c (expand_gimple_basic_block): Use build_debug_expr_decl,
3428 add a fixme note about the mode assignment perhaps being unnecessary.
3429 * ipa-param-manipulation.c (ipa_param_adjustments::modify_call):
3431 (ipa_param_body_adjustments::mark_dead_statements): Likewise.
3432 (ipa_param_body_adjustments::reset_debug_stmts): Likewise.
3433 * tree-inline.c (remap_ssa_name): Likewise.
3434 (tree_function_versioning): Likewise.
3435 * tree-into-ssa.c (rewrite_debug_stmt_uses): Likewise.
3436 * tree-ssa-loop-ivopts.c (remove_unused_ivs): Likewise.
3437 * tree-ssa.c (insert_debug_temp_for_var_def): Likewise.
3439 2021-11-16 Jakub Jelinek <jakub@redhat.com>
3441 PR tree-optimization/103208
3442 * omp-expand.c (expand_omp_build_cond): New function.
3443 (expand_omp_for_init_counts, expand_omp_for_init_vars,
3444 expand_omp_for_static_nochunk, expand_omp_for_static_chunk): Use it.
3446 2021-11-16 Jakub Jelinek <jakub@redhat.com>
3448 PR tree-optimization/102009
3449 * gimple-ssa-warn-access.cc (pass_waccess::check_alloc_size_call):
3450 Punt if any of alloc_size arguments is out of bounds vs. number of
3453 2021-11-16 Roger Sayle <roger@nextmovesoftware.com>
3455 * config/i386/i386.md (*bmi2_rorx<mode3>_1): Make conditional
3456 on !optimize_function_for_size_p.
3457 (*<any_rotate><mode>3_1): Add preferred_for_size attribute.
3458 (define_splits): Conditionalize on !optimize_function_for_size_p.
3459 (*bmi2_rorxsi3_1_zext): Likewise.
3460 (*<any_rotate>si2_1_zext): Add preferred_for_size attribute.
3461 (define_splits): Conditionalize on !optimize_function_for_size_p.
3463 2021-11-16 Jan Hubicka <jh@suse.cz>
3466 * ipa-modref.c (merge_call_side_effects): Fix uninitialized
3469 2021-11-16 Andrew Pinski <apinski@marvell.com>
3471 PR tree-optimization/103245
3472 * match.pd: Combine the abs pattern matching using multiplication.
3473 Adding optional nop_convert too.
3475 2021-11-16 H.J. Lu <hjl.tools@gmail.com>
3477 PR middle-end/103268
3478 * tree-ssa-ccp.c (optimize_atomic_bit_test_and): Add a missing
3481 2021-11-15 Siddhesh Poyarekar <siddhesh@gotplt.org>
3483 * gimple-fold.c (gimple_fold_builtin_strncat): Use ranges to
3484 determine if it is safe to transform to strcat.
3485 (gimple_fold_builtin_snprintf): Likewise.
3487 2021-11-15 Siddhesh Poyarekar <siddhesh@gotplt.org>
3489 * gimple-fold.c (known_lower): New function.
3490 (gimple_fold_builtin_strncat_chk,
3491 gimple_fold_builtin_memory_chk, gimple_fold_builtin_stxcpy_chk,
3492 gimple_fold_builtin_stxncpy_chk,
3493 gimple_fold_builtin_snprintf_chk,
3494 gimple_fold_builtin_sprintf_chk): Use it.
3496 2021-11-15 Siddhesh Poyarekar <siddhesh@gotplt.org>
3498 * gimple-fold.c (dump_transformation): New function.
3499 (gimple_fold_builtin_stxcpy_chk,
3500 gimple_fold_builtin_stxncpy_chk): Use it. Simplify to
3501 BUILT_IN_STRNCPY if return value is not used.
3503 2021-11-15 H.J. Lu <hjl.tools@gmail.com>
3505 PR middle-end/103184
3506 * tree-ssa-ccp.c (optimize_atomic_bit_test_and): Check optab
3507 before transforming equivalent, but slighly different cases to
3508 their canonical forms.
3510 2021-11-15 Iain Sandoe <iain@sandoe.co.uk>
3513 * config/darwin.h (TARGET_DTORS_FROM_CXA_ATEXIT): New.
3514 * doc/tm.texi: Regenerated.
3515 * doc/tm.texi.in: Add TARGET_DTORS_FROM_CXA_ATEXIT hook.
3516 * ipa.c (cgraph_build_static_cdtor_1): Return the built
3518 (build_cxa_atexit_decl): New.
3519 (build_dso_handle_decl): New.
3520 (build_cxa_dtor_registrations): New.
3521 (compare_cdtor_tu_order): New.
3522 (build_cxa_atexit_fns): New.
3523 (ipa_cdtor_merge): If dtors_from_cxa_atexit is set,
3524 process the DTORs/CTORs accordingly.
3525 (pass_ipa_cdtor_merge::gate): Also run if
3526 dtors_from_cxa_atexit is set.
3527 * target.def (dtors_from_cxa_atexit): New hook.
3529 2021-11-15 Iain Sandoe <iain@sandoe.co.uk>
3531 * config.in: Regenerate.
3532 * configure: Regenerate.
3533 * configure.ac: Test ld64 for -platform-version support.
3535 2021-11-15 Thomas Schwinge <thomas@codesourcery.com>
3537 * diagnostic-spec.h (typedef xint_hash_t)
3538 (typedef xint_hash_map_t): Replace with...
3539 (typedef nowarn_map_t): ... this.
3540 (nowarn_map): Adjust.
3541 * diagnostic-spec.c (nowarn_map, suppress_warning_at): Likewise.
3543 2021-11-15 Thomas Schwinge <thomas@codesourcery.com>
3545 * profile.c (branch_prob): Use 'location_hash' for
3548 2021-11-15 Aldy Hernandez <aldyh@redhat.com>
3550 PR tree-optimization/103207
3551 * value-range.cc (irange::set): Drop overflow.
3553 2021-11-15 Jakub Jelinek <jakub@redhat.com>
3555 * gimplify.c (optimize_target_teams): Only add OMP_CLAUSE_THREAD_LIMIT
3556 to OMP_TARGET_CLAUSES if it isn't there already.
3558 2021-11-15 Aldy Hernandez <aldyh@redhat.com>
3560 * gimple-range-path.cc (path_range_query::internal_range_of_expr):
3561 Remove useless code.
3562 (path_range_query::ssa_defined_in_bb): New.
3563 (path_range_query::ssa_range_in_phi): Avoid fold_range call that
3564 could trigger additional lookups.
3565 Do not use the cache for ARGs defined in this block.
3566 (path_range_query::compute_ranges_in_block): Use ssa_defined_in_bb.
3567 (path_range_query::maybe_register_phi_relation): Same.
3568 (path_range_query::range_of_stmt): Adjust comment.
3569 * gimple-range-path.h (ssa_defined_in_bb): New.
3571 2021-11-15 Aldy Hernandez <aldyh@redhat.com>
3573 * gimple-range-path.cc (path_range_query::internal_range_of_expr):
3574 Default to global range if nothing found.
3576 2021-11-15 Richard Biener <rguenther@suse.de>
3578 PR tree-optimization/103237
3579 * tree-vect-loop.c (vect_is_simple_reduction): Fail for
3580 double reductions with multiple inner loop LC PHI nodes.
3582 2021-11-15 Hongyu Wang <hongyu.wang@intel.com>
3585 * config/i386/i386-expand.c (ix86_expand_atomic_fetch_op_loop):
3586 New expand function.
3587 * config/i386/i386-options.c (ix86_target_string): Add
3588 -mrelax-cmpxchg-loop flag.
3589 (ix86_valid_target_attribute_inner_p): Likewise.
3590 * config/i386/i386-protos.h (ix86_expand_atomic_fetch_op_loop):
3591 New expand function prototype.
3592 * config/i386/i386.opt: Add -mrelax-cmpxchg-loop.
3593 * config/i386/sync.md (atomic_fetch_<logic><mode>): New expander
3595 (atomic_<logic>_fetch<mode>): Likewise.
3596 (atomic_fetch_nand<mode>): Likewise.
3597 (atomic_nand_fetch<mode>): Likewise.
3598 (atomic_fetch_<logic><mode>): New expander for DI,TI modes.
3599 (atomic_<logic>_fetch<mode>): Likewise.
3600 (atomic_fetch_nand<mode>): Likewise.
3601 (atomic_nand_fetch<mode>): Likewise.
3602 * doc/invoke.texi: Document -mrelax-cmpxchg-loop.
3604 2021-11-15 Richard Biener <rguenther@suse.de>
3606 PR tree-optimization/103219
3607 * gimple-loop-jam.c (tree_loop_unroll_and_jam): Use single_exit
3608 to determine the exit for the VN region.
3610 2021-11-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
3612 * tree-ssa-loop.c (pass_vectorize): Move to tree-vectorizer.c.
3613 (pass_data_vectorize): Likewise.
3614 (make_pass_vectorize): Likewise.
3615 * tree-vectorizer.c (vectorize_loops): Merge with
3616 pass_vectorize::execute and replace cfun occurences with fun param.
3617 (adjust_simduid_builtins): Add fun param, replace cfun occurences with
3618 fun, and adjust callers approrpiately.
3619 (note_simd_array_uses): Likewise.
3620 (vect_loop_dist_alias_call): Likewise.
3621 (set_uid_loop_bbs): Likewise.
3622 (vect_transform_loops): Likewise.
3623 (try_vectorize_loop_1): Likewise.
3624 (try_vectorize_loop): Likewise.
3626 2021-11-15 Jakub Jelinek <jakub@redhat.com>
3629 * config/i386/sync.md (atomic_bit_test_and_set<mode>,
3630 atomic_bit_test_and_complement<mode>,
3631 atomic_bit_test_and_reset<mode>): Use OPTAB_WIDEN instead of
3634 2021-11-15 Maciej W. Rozycki <macro@embecosm.com>
3636 * config/vax/vax.c (vax_lra_p): New prototype and function.
3637 (TARGET_LRA_P): Wire it.
3638 * config/vax/vax.opt (mlra): New option.
3639 * doc/invoke.texi (Option Summary, VAX Options): Document the
3642 2021-11-14 Jan Hubicka <jh@suse.cz>
3644 * ipa-modref.h (struct modref_summary): Add nondeterministic
3645 and calls_interposable flags.
3646 * ipa-modref.c (modref_summary::modref_summary): Initialize new flags.
3647 (modref_summary::useful_p): Check new flags.
3648 (struct modref_summary_lto): Add nondeterministic and
3649 calls_interposable flags.
3650 (modref_summary_lto::modref_summary_lto): Initialize new flags.
3651 (modref_summary_lto::useful_p): Check new flags.
3652 (modref_summary::dump): Dump new flags.
3653 (modref_summary_lto::dump): Dump new flags.
3654 (ignore_nondeterminism_p): New function.
3655 (merge_call_side_effects): Merge new flags.
3656 (process_fnspec): Likewise.
3657 (analyze_load): Volatile access is nondeterministic.
3658 (analyze_store): Liekwise.
3659 (analyze_stmt): Volatile ASM is nondeterministic.
3660 (analyze_function): Clear new flags.
3661 (modref_summaries::duplicate): Duplicate new flags.
3662 (modref_summaries_lto::duplicate): Duplicate new flags.
3663 (modref_write): Stream new flags.
3664 (read_section): Stream new flags.
3665 (propagate_unknown_call): Update new flags.
3666 (modref_propagate_in_scc): Propagate new flags.
3667 * tree-ssa-alias.c (ref_maybe_used_by_call_p_1): Check
3669 * tree-ssa-structalias.c (determine_global_memory_access):
3672 2021-11-14 Maciej W. Rozycki <macro@embecosm.com>
3674 * config/vax/vax.h (SET_RATIO): New macro.
3675 * config/vax/vax.md (UNSPEC_SETMEM_FILL): New constant.
3676 (setmemhi): New expander.
3677 (setmemhi1): New insn and splitter.
3678 (*setmemhi1): New insn.
3680 2021-11-14 Jan Hubicka <hubicka@ucw.cz>
3682 * ipa-modref-tree.c (modref_access_node::update_for_kills): New
3684 (modref_access_node::merge_for_kills): Likewise.
3685 (modref_access_node::insert_kill): Likewise.
3686 * ipa-modref-tree.h (modref_access_node::update_for_kills,
3687 modref_access_node::merge_for_kills, modref_access_node::insert_kill):
3689 (modref_access_node::useful_for_kill): New member function.
3690 * ipa-modref.c (modref_summary::useful_p): Release useless kills.
3691 (lto_modref_summary): Add kills.
3692 (modref_summary::dump): Dump kills.
3693 (record_access): Add mdoref_access_node parameter.
3694 (record_access_lto): Likewise.
3695 (merge_call_side_effects): Merge kills.
3696 (analyze_call): Add ALWAYS_EXECUTED param and pass it around.
3697 (struct summary_ptrs): Add always_executed filed.
3698 (analyze_load): Update.
3699 (analyze_store): Update; record kills.
3700 (analyze_stmt): Add always_executed; record kills in clobbers.
3701 (analyze_function): Track always_executed.
3702 (modref_summaries::duplicate): Duplicate kills.
3703 (update_signature): Release kills.
3704 * ipa-modref.h (struct modref_summary): Add kills.
3705 * tree-ssa-alias.c (alias_stats): Add kill stats.
3706 (dump_alias_stats): Dump kill stats.
3707 (store_kills_ref_p): Break out from ...
3708 (stmt_kills_ref_p): Use it; handle modref info based kills.
3710 2021-11-14 Aldy Hernandez <aldyh@redhat.com>
3712 PR tree-optimization/103229
3713 * gimple-range-cache.cc (ssa_global_cache::clear): Do not pass
3714 null value to memset.
3716 2021-11-14 Jan Hubicka <hubicka@ucw.cz>
3718 * ipa-modref-tree.c (modref_access_node::get_call_arg): New member
3720 (modref_access_node::get_ao_ref): Likewise.
3721 * ipa-modref-tree.h (modref_access_node::get_call_arg): Declare.
3722 (modref_access_node::get_ao_ref): Declare.
3723 * tree-ssa-alias.c (modref_may_conflict): Use new accessors.
3724 * tree-ssa-dse.c (dse_optimize_call): Use new accessors.
3726 2021-11-13 Jan Hubicka <jh@suse.cz>
3729 * dbgcnt.def (ipa_attr): New counters.
3730 * ipa-pure-const.c: Include dbgcnt.c
3731 (ipa_make_function_const): Use debug counter.
3732 (ipa_make_function_pure): Likewise.
3733 (propagate_pure_const): Fix bug in my previous change.
3735 2021-11-13 Jan Hubicka <hubicka@ucw.cz>
3737 * ipa-modref-tree.c (modref_access_node::range_info_useful_p):
3738 Offline from ipa-modref-tree.h.
3739 (modref_access_node::dump): Move from ipa-modref.c; make member
3741 * ipa-modref-tree.h (modref_access_node::range_info_useful_p.
3742 modref_access_node::dump): Declare.
3743 * ipa-modref.c (dump_access): Remove.
3744 (dump_records): Update.
3745 (dump_lto_records): Update.
3746 (record_access): Update.
3747 (record_access_lto): Update.
3749 2021-11-13 Jan Hubicka <hubicka@ucw.cz>
3751 * ipa-modref.c (modref_summary::modref_summary): Clear new flags.
3752 (modref_summary::dump): Dump try_dse.
3753 (modref_summary::finalize): Add FUN attribute; compute try-dse.
3754 (analyze_function): Update.
3755 (read_section): Update.
3756 (update_signature): Update.
3757 (pass_ipa_modref::execute): Update.
3758 * ipa-modref.h (struct modref_summary):
3759 * tree-ssa-alias.c (ao_ref_init_from_ptr_and_range): Export.
3760 * tree-ssa-alias.h (ao_ref_init_from_ptr_and_range): Declare.
3761 * tree-ssa-dse.c (dse_optimize_call): New function.
3762 (dse_optimize_stmt): Use it.
3764 2021-11-13 Jan Hubicka <hubicka@ucw.cz>
3766 * ipa-modref-tree.c: Move #if CHECKING_P to proper place.
3768 2021-11-13 Jan Hubicka <jh@suse.cz>
3771 (struct modref_access_node): Move longer member functions to
3773 (modref_ref_node::try_merge_with): Turn into modreef_acces_node member
3775 * ipa-modref-tree.c (modref_access_node::contains): Move here
3776 from ipa-modref-tree.h.
3777 (modref_access_node::update): Likewise.
3778 (modref_access_node::merge): Likewise.
3779 (modref_access_node::closer_pair_p): Likewise.
3780 (modref_access_node::forced_merge): Likewise.
3781 (modref_access_node::update2): Likewise.
3782 (modref_access_node::combined_offsets): Likewise.
3783 (modref_access_node::try_merge_with): Likewise.
3784 (modref_access_node::insert): Likewise.
3786 2021-11-13 Jan Hubicka <jh@suse.cz>
3788 * ipa-modref.c (modref_summary::global_memory_read_p): Remove.
3789 (modref_summary::global_memory_written_p): Remove.
3790 (modref_summary::dump): Dump new flags.
3791 (modref_summary::finalize): New member function.
3792 (analyze_function): Call it.
3793 (read_section): Call it.
3794 (update_signature): Call it.
3795 (pass_ipa_modref::execute): Call it.
3796 * ipa-modref.h (struct modref_summary): Remove
3797 global_memory_read_p and global_memory_written_p.
3798 Add global_memory_read, global_memory_written.
3799 * tree-ssa-structalias.c (determine_global_memory_access):
3802 2021-11-13 Jan Hubicka <jh@suse.cz>
3804 * ipa-fnsummary.c (compute_fn_summary): Use type_attribut_allowed_p
3805 * ipa-param-manipulation.c
3806 (ipa_param_adjustments::type_attribute_allowed_p):
3807 New member function.
3808 (drop_type_attribute_if_params_changed_p): New function.
3809 (build_adjusted_function_type): Use it.
3810 * ipa-param-manipulation.h: Add type_attribute_allowed_p.
3812 2021-11-13 David Malcolm <dmalcolm@redhat.com>
3814 * doc/invoke.texi (Static Analyzer Options): Add
3815 -Wno-analyzer-tainted-allocation-size,
3816 -Wno-analyzer-tainted-divisor, -Wno-analyzer-tainted-offset, and
3817 -Wno-analyzer-tainted-size to list. Add
3818 -Wanalyzer-tainted-allocation-size, -Wanalyzer-tainted-divisor,
3819 -Wanalyzer-tainted-offset, and -Wanalyzer-tainted-size to list
3820 of options effectively enabled by -fanalyzer.
3821 (-Wanalyzer-tainted-allocation-size): New.
3822 (-Wanalyzer-tainted-array-index): Tweak wording; add link to CWE.
3823 (-Wanalyzer-tainted-divisor): New.
3824 (-Wanalyzer-tainted-offset): New.
3825 (-Wanalyzer-tainted-size): New.
3827 2021-11-13 Jan Hubicka <jh@suse.cz>
3829 * attr-fnspec.h (attr_fnspec::arg_eaf_flags): Break out from ...
3830 * gimple.c (gimple_call_arg_flags): ... here.
3831 * ipa-modref.c (analyze_parms): Record flags known from fnspec.
3832 (modref_merge_call_site_flags): Use arg_eaf_flags.
3834 2021-11-13 Aldy Hernandez <aldyh@redhat.com>
3836 PR tree-optimization/103222
3837 * gimple-range-path.cc (path_range_query::compute_ranges_in_phis):
3839 (path_range_query::compute_ranges_in_block): Call
3840 compute_ranges_in_phis.
3841 * gimple-range-path.h (path_range_query::compute_ranges_in_phis):
3844 2021-11-13 Jan Hubicka <jh@suse.cz>
3846 * ipa-fnsummary.c (compute_fn_summary): Do not give up on signature
3847 changes on "fn spec" attribute; give up on varadic types.
3848 * ipa-param-manipulation.c: Include attribs.h.
3849 (build_adjusted_function_type): New parameter ARG_MODIFIED; if it is
3850 true remove "fn spec" attribute.
3851 (ipa_param_adjustments::build_new_function_type): Update.
3852 (ipa_param_body_adjustments::modify_formal_parameters): update.
3853 * ipa-sra.c: Include attribs.h.
3854 (ipa_sra_preliminary_function_checks): Do not check for TYPE_ATTRIBUTES.
3856 2021-11-13 Aldy Hernandez <aldyh@redhat.com>
3858 * gimple-range-path.cc (path_range_query::path_range_query): Merge
3860 (path_range_query::import_p): Move from header file.
3861 (path_range_query::~path_range_query): Adjust for combined ctors.
3862 * gimple-range-path.h: Merge ctors.
3863 (path_range_query::import_p): Move to .cc file.
3865 2021-11-13 Jan Hubicka <jh@suse.cz>
3867 PR tree-optimization/103182
3868 * ipa-modref.c (callee_to_caller_flags): Fix merging of flags.
3869 (modref_eaf_analysis::analyze_ssa_name): Fix merging of flags.
3871 2021-11-12 Stafford Horne <shorne@gmail.com>
3873 * config/or1k/or1k-protos.h (or1k_profile_hook): New function.
3874 * config/or1k/or1k.h (PROFILE_HOOK): Change macro to reference
3875 new function or1k_profile_hook.
3876 * config/or1k/or1k.c (struct machine_function): Add new field
3877 set_mcount_arg_insn.
3878 (or1k_profile_hook): New function.
3879 (or1k_init_pic_reg): Update to inject pic rtx after _mcount arg
3881 (or1k_frame_pointer_required): Frame pointer no longer needed
3884 2021-11-12 Jan Hubicka <jh@suse.cz>
3886 PR tree-optimization/103209
3887 * tree-ssa-structalias.c (find_func_aliases_for_call): Fix
3888 use of handle_rhs_call
3890 2021-11-12 Aldy Hernandez <aldyh@redhat.com>
3892 PR tree-optimization/103202
3893 * gimple-range-path.cc
3894 (path_range_query::compute_ranges_in_block): Solve PHI imports first.
3896 2021-11-12 Jan Hubicka <jh@suse.cz>
3898 * ipa-pure-const.c (propagate_pure_const): Remove redundant check;
3899 fix call of ipa_make_function_const and ipa_make_function_pure.
3901 2021-11-12 Richard Sandiford <richard.sandiford@arm.com>
3903 * config/aarch64/aarch64.c (aarch64_vector_op::n_advsimd_ops): Delete.
3904 (aarch64_vector_op::m_seen_loads): Likewise.
3905 (aarch64_vector_costs::aarch64_vector_costs): Don't push to
3907 (aarch64_vector_op::count_ops): Remove vectype and factor parameters.
3908 Remove code that tries to predict different vec_flags from the
3910 (aarch64_vector_costs::add_stmt_cost): Update accordingly.
3911 Remove m_advsimd_ops handling.
3913 2021-11-12 Richard Sandiford <richard.sandiford@arm.com>
3915 * config/aarch64/aarch64.c (aarch64_vector_costs::m_saw_sve_only_op)
3916 (aarch64_sve_only_stmt_p): Delete.
3917 (aarch64_vector_costs::prefer_unrolled_loop): New function,
3918 extracted from adjust_body_cost.
3919 (aarch64_vector_costs::better_main_loop_than_p): New function,
3920 using heuristics extracted from adjust_body_cost and
3921 adjust_body_cost_sve.
3922 (aarch64_vector_costs::adjust_body_cost_sve): Remove
3923 advsimd_cycles_per_iter and could_use_advsimd parameters.
3924 Update after changes above.
3925 (aarch64_vector_costs::adjust_body_cost): Update after changes above.
3927 2021-11-12 Richard Sandiford <richard.sandiford@arm.com>
3929 * config/aarch64/aarch64.c (aarch64_vec_op_count::m_vf_factor):
3930 New member variable.
3931 (aarch64_vec_op_count::aarch64_vec_op_count): Add a parameter for it.
3932 (aarch64_vec_op_count::vf_factor): New function.
3933 (aarch64_vector_costs::aarch64_vector_costs): When costing for
3934 neoverse-512tvb, pass a vf_factor of 2 for the Neoverse V1 version
3936 (aarch64_vector_costs::adjust_body_cost): Read the vf factor
3937 instead of hard-coding 2.
3939 2021-11-12 Richard Sandiford <richard.sandiford@arm.com>
3941 * config/aarch64/aarch64.c
3942 (aarch64_vec_op_count::rename_cycles_per_iter): New function.
3943 (aarch64_vec_op_count::min_nonpred_cycles_per_iter): Likewise.
3944 (aarch64_vec_op_count::min_pred_cycles_per_iter): Likewise.
3945 (aarch64_vec_op_count::min_cycles_per_iter): Likewise.
3946 (aarch64_vec_op_count::dump): Move earlier in file. Dump the
3947 above properties too.
3948 (aarch64_estimate_min_cycles_per_iter): Delete.
3949 (adjust_body_cost): Use aarch64_vec_op_count::min_cycles_per_iter
3950 instead of aarch64_estimate_min_cycles_per_iter. Rely on the dump
3951 routine to print CPI estimates.
3952 (adjust_body_cost_sve): Likewise. Use the other functions above
3953 instead of doing the work inline.
3955 2021-11-12 Richard Sandiford <richard.sandiford@arm.com>
3957 * config/aarch64/aarch64.c (aarch64_vec_op_count): Allow default
3959 (aarch64_vec_op_count::base_issue_info): Remove handling of null
3961 (aarch64_vec_op_count::simd_issue_info): Likewise.
3962 (aarch64_vec_op_count::sve_issue_info): Likewise.
3963 (aarch64_vector_costs::m_ops): Turn into a vector.
3964 (aarch64_vector_costs::m_advsimd_ops): Likewise.
3965 (aarch64_vector_costs::aarch64_vector_costs): Add entries to
3966 the vectors based on aarch64_tune_params.
3967 (aarch64_vector_costs::analyze_loop_vinfo): Update the pred_ops
3968 of all entries in m_ops.
3969 (aarch64_vector_costs::add_stmt_cost): Call count_ops for all
3971 (aarch64_estimate_min_cycles_per_iter): Remove issue_info
3972 parameter and get the information from the ops instead.
3973 (aarch64_vector_costs::adjust_body_cost_sve): Take a
3974 aarch64_vec_issue_info instead of a aarch64_vec_op_count.
3975 (aarch64_vector_costs::adjust_body_cost): Update call accordingly.
3976 Exit earlier if m_ops is empty for either cost structure.
3978 2021-11-12 Richard Sandiford <richard.sandiford@arm.com>
3980 * config/aarch64/aarch64.c (aarch64_vector_costs::m_scalar_ops)
3981 (aarch64_vector_costs::m_sve_ops): Replace with...
3982 (aarch64_vector_costs::m_ops): ...this.
3983 (aarch64_vector_costs::analyze_loop_vinfo): Update accordingly.
3984 (aarch64_vector_costs::adjust_body_cost_sve): Likewise.
3985 (aarch64_vector_costs::aarch64_vector_costs): Likewise.
3986 Initialize m_vec_flags here rather than in add_stmt_cost.
3987 (aarch64_vector_costs::count_ops): Test for scalar reductions too.
3988 Allow vectype to be null.
3989 (aarch64_vector_costs::add_stmt_cost): Call count_ops for scalar
3990 code too. Don't require vectype to be nonnull.
3991 (aarch64_vector_costs::adjust_body_cost): Take the loop_vec_info
3992 and scalar costs as parameters. Use the scalar costs to determine
3993 the cycles per iteration of the scalar loop, then multiply it
3994 by the estimated VF.
3995 (aarch64_vector_costs::finish_cost): Update call accordingly.
3997 2021-11-12 Richard Sandiford <richard.sandiford@arm.com>
3999 * config/aarch64/aarch64.c (aarch64_dr_type): New function.
4000 (aarch64_vector_costs::count_ops): Use it rather than the
4001 vectype to determine floatness.
4003 2021-11-12 Richard Sandiford <richard.sandiford@arm.com>
4005 * config/aarch64/aarch64.c (aarch64_sve_in_loop_reduction_latency):
4006 Remove vectype parameter and get floatness from the type of the
4008 (arch64_in_loop_reduction_latency): Likewise.
4009 (aarch64_detect_vector_stmt_subtype): Update caller.
4010 (aarch64_vector_costs::count_ops): Likewise.
4012 2021-11-12 Richard Sandiford <richard.sandiford@arm.com>
4014 * config/aarch64/aarch64.c (aarch64_sve_op_count): Fold into...
4015 (aarch64_vec_op_count): ...this. Add a constructor.
4016 (aarch64_vec_op_count::vec_flags): New function.
4017 (aarch64_vec_op_count::base_issue_info): Likewise.
4018 (aarch64_vec_op_count::simd_issue_info): Likewise.
4019 (aarch64_vec_op_count::sve_issue_info): Likewise.
4020 (aarch64_vec_op_count::m_issue_info): New member variable.
4021 (aarch64_vec_op_count::m_vec_flags): Likewise.
4022 (aarch64_vector_costs): Add a constructor.
4023 (aarch64_vector_costs::m_sve_ops): Change type to aarch64_vec_op_count.
4024 (aarch64_vector_costs::aarch64_vector_costs): New function.
4025 Initialize m_scalar_ops, m_advsimd_ops and m_sve_ops.
4026 (aarch64_vector_costs::count_ops): Remove vec_flags and
4027 issue_info parameters, using the new aarch64_vec_op_count
4029 (aarch64_vector_costs::add_stmt_cost): Update call accordingly.
4030 (aarch64_sve_op_count::dump): Fold into...
4031 (aarch64_vec_op_count::dump): ..here.
4033 2021-11-12 Richard Sandiford <richard.sandiford@arm.com>
4035 * config/aarch64/aarch64.c: Include tree-dfa.h.
4036 (aarch64_check_consecutive_mems): New function that takes MEM_EXPR
4037 and MEM_OFFSET into account.
4038 (aarch64_swap_ldrstr_operands): Use it.
4039 (aarch64_operands_ok_for_ldpstp): Likewise. Check that the
4040 address of the second memory doesn't depend on the result of
4043 2021-11-12 Jan Hubicka <hubicka@ucw.cz>
4045 * ipa-pure-const.c (ipa_make_function_pure): Fix exit condition.
4047 2021-11-12 Jan Hubicka <jh@suse.cz>
4049 PR tree-optimization/103175
4050 * ipa-modref.c (modref_lattice::merge): Add sanity check.
4051 (callee_to_caller_flags): Make flags adjustment sane.
4052 (modref_eaf_analysis::analyze_ssa_name): Likewise.
4054 2021-11-12 Jan Hubicka <jh@suse.cz>
4057 * ipa-modref.c (analyze_function, modref_propagate_in_scc): Do
4058 not mark pure/const function if there are side-effects.
4060 2021-11-12 Chung-Lin Tang <cltang@codesourcery.com>
4062 * tree.h (OMP_CLAUSE_MAP_RUNTIME_IMPLICIT_P): New access macro for
4063 'implicit' bit, using 'base.deprecated_flag' field of tree_node.
4064 * tree-pretty-print.c (dump_omp_clause): Add support for printing
4065 implicit attribute in tree dumping.
4066 * gimplify.c (gimplify_adjust_omp_clauses_1):
4067 Set OMP_CLAUSE_MAP_RUNTIME_IMPLICIT_P to 1 if map clause is implicitly
4069 (gimplify_adjust_omp_clauses): Adjust place of adding implicitly created
4070 clauses, from simple append, to starting of list, after non-map clauses.
4071 * omp-low.c (lower_omp_target): Add GOMP_MAP_IMPLICIT bits into kind
4072 values passed to libgomp for implicit maps.
4074 2021-11-12 Jakub Jelinek <jakub@redhat.com>
4076 * omp-builtins.def (BUILT_IN_GOMP_TEAMS): Remove.
4077 (BUILT_IN_GOMP_TEAMS4): New.
4078 * builtin-types.def (BT_FN_VOID_UINT_UINT): Remove.
4079 (BT_FN_BOOL_UINT_UINT_UINT_BOOL): New.
4080 * omp-low.c (lower_omp_teams): Use GOMP_teams4 instead of
4081 GOMP_teams, pass to it also num_teams lower-bound expression
4082 or a dup of upper-bound if it is missing and a flag whether
4083 it is the first call or not.
4085 2021-11-12 Martin Liska <mliska@suse.cz>
4087 PR tree-optimization/102497
4088 * gimple-predicate-analysis.cc (add_pred): Remove unused
4091 2021-11-12 Richard Biener <rguenther@suse.de>
4093 PR tree-optimization/103204
4094 * tree-ssa-sccvn.c (valueize_refs_1): Re-valueize the
4095 top operand after folding in an address.
4097 2021-11-11 Aldy Hernandez <aldyh@redhat.com>
4099 * gimple-range-path.cc (path_range_query::path_range_query): New
4100 ctor without a ranger.
4101 (path_range_query::~path_range_query): Free ranger if necessary.
4102 (path_range_query::range_on_path_entry): Adjust m_ranger for pointer.
4103 (path_range_query::ssa_range_in_phi): Same.
4104 (path_range_query::compute_ranges_in_block): Same.
4105 (path_range_query::compute_imports): Same.
4106 (path_range_query::compute_ranges): Same.
4107 (path_range_query::range_of_stmt): Same.
4108 (path_range_query::compute_outgoing_relations): Same.
4109 * gimple-range-path.h (class path_range_query): New ctor.
4110 * tree-ssa-loop-ch.c (ch_base::copy_headers): Remove gimple_ranger
4111 as path_range_query allocates one.
4112 * tree-ssa-threadbackward.c (class back_threader): Remove m_ranger.
4113 (back_threader::~back_threader): Same.
4115 2021-11-11 Aldy Hernandez <aldyh@redhat.com>
4117 * tree-ssa-threadbackward.c
4118 (back_threader_profitability::profitable_path_p): Remove loop
4119 crossing restriction.
4121 2021-11-11 Jan Hubicka <hubicka@ucw.cz>
4123 * ipa-modref.c (analyze_function): Do pure/const discovery, return
4125 (pass_modref::execute): If pure/const is discovered fixup cfg.
4126 (ignore_edge): Do not ignore pure/const edges.
4127 (modref_propagate_in_scc): Do pure/const discovery, return true if
4128 cdtor was promoted pure/const.
4129 (pass_ipa_modref::execute): If needed remove unreachable functions.
4130 * ipa-pure-const.c (warn_function_noreturn): Fix whitespace.
4131 (warn_function_cold): Likewise.
4132 (skip_function_for_local_pure_const): Move earlier.
4133 (ipa_make_function_const): Break out from ...
4134 (ipa_make_function_pure): Break out from ...
4135 (propagate_pure_const): ... here.
4136 (pass_local_pure_const::execute): Use it.
4137 * ipa-utils.h (ipa_make_function_const): Declare.
4138 (ipa_make_function_pure): Declare.
4139 * passes.def: Move early modref after pure-const.
4141 2021-11-11 David Malcolm <dmalcolm@redhat.com>
4144 * diagnostic-show-locus.c (def_policy): Use def_tabstop.
4146 2021-11-11 Jonathan Wright <jonathan.wright@arm.com>
4148 * config/aarch64/aarch64-builtins.c (TYPES_COMBINE): Delete.
4149 (TYPES_COMBINEP): Delete.
4150 * config/aarch64/aarch64-simd-builtins.def: Declare type-
4151 qualified builtins for vcombine_* intrinsics.
4152 * config/aarch64/arm_neon.h (vcombine_s8): Remove unnecessary
4154 (vcombine_s16): Likewise.
4155 (vcombine_s32): Likewise.
4156 (vcombine_f32): Likewise.
4157 (vcombine_u8): Use type-qualified builtin and remove casts.
4158 (vcombine_u16): Likewise.
4159 (vcombine_u32): Likewise.
4160 (vcombine_u64): Likewise.
4161 (vcombine_p8): Likewise.
4162 (vcombine_p16): Likewise.
4163 (vcombine_p64): Likewise.
4164 (vcombine_bf16): Remove unnecessary cast.
4165 * config/aarch64/iterators.md (VD_I): New mode iterator.
4166 (VDC_P): New mode iterator.
4168 2021-11-11 Jonathan Wright <jonathan.wright@arm.com>
4170 * config/aarch64/aarch64-builtins.c (TYPES_LOAD1_U): Define.
4171 (TYPES_LOAD1_P): Define.
4172 (TYPES_STORE1_U): Define.
4173 (TYPES_STORE1P): Rename to...
4174 (TYPES_STORE1_P): This.
4175 (get_mem_type_for_load_store): Add unsigned and poly types.
4176 (aarch64_general_gimple_fold_builtin): Add unsigned and poly
4177 type-qualified builtin declarations.
4178 * config/aarch64/aarch64-simd-builtins.def: Declare type-
4179 qualified builtins for LD1/ST1.
4180 * config/aarch64/arm_neon.h (vld1_p8): Use type-qualified
4181 builtin and remove cast.
4182 (vld1_p16): Likewise.
4183 (vld1_u8): Likewise.
4184 (vld1_u16): Likewise.
4185 (vld1_u32): Likewise.
4186 (vld1q_p8): Likewise.
4187 (vld1q_p16): Likewise.
4188 (vld1q_p64): Likewise.
4189 (vld1q_u8): Likewise.
4190 (vld1q_u16): Likewise.
4191 (vld1q_u32): Likewise.
4192 (vld1q_u64): Likewise.
4193 (vst1_p8): Likewise.
4194 (vst1_p16): Likewise.
4195 (vst1_u8): Likewise.
4196 (vst1_u16): Likewise.
4197 (vst1_u32): Likewise.
4198 (vst1q_p8): Likewise.
4199 (vst1q_p16): Likewise.
4200 (vst1q_p64): Likewise.
4201 (vst1q_u8): Likewise.
4202 (vst1q_u16): Likewise.
4203 (vst1q_u32): Likewise.
4204 (vst1q_u64): Likewise.
4205 * config/aarch64/iterators.md (VALLP_NO_DI): New iterator.
4207 2021-11-11 Jonathan Wright <jonathan.wright@arm.com>
4209 * config/aarch64/aarch64-simd-builtins.def: Declare unsigned
4210 builtins for vector reduction.
4211 * config/aarch64/arm_neon.h (vaddv_u8): Use type-qualified
4212 builtin and remove casts.
4213 (vaddv_u16): Likewise.
4214 (vaddv_u32): Likewise.
4215 (vaddvq_u8): Likewise.
4216 (vaddvq_u16): Likewise.
4217 (vaddvq_u32): Likewise.
4218 (vaddvq_u64): Likewise.
4220 2021-11-11 Jonathan Wright <jonathan.wright@arm.com>
4222 * config/aarch64/aarch64-simd-builtins.def:
4223 * config/aarch64/arm_neon.h (vpaddq_u8): Use type-qualified
4224 builtin and remove casts.
4225 (vpaddq_u16): Likewise.
4226 (vpaddq_u32): Likewise.
4227 (vpaddq_u64): Likewise.
4228 (vpadd_u8): Likewise.
4229 (vpadd_u16): Likewise.
4230 (vpadd_u32): Likewise.
4231 (vpaddd_u64): Likewise.
4233 2021-11-11 Jonathan Wright <jonathan.wright@arm.com>
4235 * config/aarch64/aarch64-simd-builtins.def: Declare unsigned
4236 builtins for [r]subhn[2].
4237 * config/aarch64/arm_neon.h (vsubhn_s16): Remove unnecessary
4239 (vsubhn_s32): Likewise.
4240 (vsubhn_s64): Likewise.
4241 (vsubhn_u16): Use type-qualified builtin and remove casts.
4242 (vsubhn_u32): Likewise.
4243 (vsubhn_u64): Likewise.
4244 (vrsubhn_s16): Remove unnecessary cast.
4245 (vrsubhn_s32): Likewise.
4246 (vrsubhn_s64): Likewise.
4247 (vrsubhn_u16): Use type-qualified builtin and remove casts.
4248 (vrsubhn_u32): Likewise.
4249 (vrsubhn_u64): Likewise.
4250 (vrsubhn_high_s16): Remove unnecessary cast.
4251 (vrsubhn_high_s32): Likewise.
4252 (vrsubhn_high_s64): Likewise.
4253 (vrsubhn_high_u16): Use type-qualified builtin and remove
4255 (vrsubhn_high_u32): Likewise.
4256 (vrsubhn_high_u64): Likewise.
4257 (vsubhn_high_s16): Remove unnecessary cast.
4258 (vsubhn_high_s32): Likewise.
4259 (vsubhn_high_s64): Likewise.
4260 (vsubhn_high_u16): Use type-qualified builtin and remove
4262 (vsubhn_high_u32): Likewise.
4263 (vsubhn_high_u64): Likewise.
4265 2021-11-11 Jonathan Wright <jonathan.wright@arm.com>
4267 * config/aarch64/aarch64-simd-builtins.def: Declare unsigned
4268 builtins for [r]addhn[2].
4269 * config/aarch64/arm_neon.h (vaddhn_s16): Remove unnecessary
4271 (vaddhn_s32): Likewise.
4272 (vaddhn_s64): Likewise.
4273 (vaddhn_u16): Use type-qualified builtin and remove casts.
4274 (vaddhn_u32): Likewise.
4275 (vaddhn_u64): Likewise.
4276 (vraddhn_s16): Remove unnecessary cast.
4277 (vraddhn_s32): Likewise.
4278 (vraddhn_s64): Likewise.
4279 (vraddhn_u16): Use type-qualified builtin and remove casts.
4280 (vraddhn_u32): Likewise.
4281 (vraddhn_u64): Likewise.
4282 (vaddhn_high_s16): Remove unnecessary cast.
4283 (vaddhn_high_s32): Likewise.
4284 (vaddhn_high_s64): Likewise.
4285 (vaddhn_high_u16): Use type-qualified builtin and remove
4287 (vaddhn_high_u32): Likewise.
4288 (vaddhn_high_u64): Likewise.
4289 (vraddhn_high_s16): Remove unnecessary cast.
4290 (vraddhn_high_s32): Likewise.
4291 (vraddhn_high_s64): Likewise.
4292 (vraddhn_high_u16): Use type-qualified builtin and remove
4294 (vraddhn_high_u32): Likewise.
4295 (vraddhn_high_u64): Likewise.
4297 2021-11-11 Jonathan Wright <jonathan.wright@arm.com>
4299 * config/aarch64/aarch64-simd-builtins.def: Use BINOPU type
4300 qualifiers in generator macros for uhsub builtins.
4301 * config/aarch64/arm_neon.h (vhsub_s8): Remove unnecessary
4303 (vhsub_s16): Likewise.
4304 (vhsub_s32): Likewise.
4305 (vhsub_u8): Use type-qualified builtin and remove casts.
4306 (vhsub_u16): Likewise.
4307 (vhsub_u32): Likewise.
4308 (vhsubq_s8): Remove unnecessary cast.
4309 (vhsubq_s16): Likewise.
4310 (vhsubq_s32): Likewise.
4311 (vhsubq_u8): Use type-qualified builtin and remove casts.
4312 (vhsubq_u16): Likewise.
4313 (vhsubq_u32): Likewise.
4315 2021-11-11 Jonathan Wright <jonathan.wright@arm.com>
4317 * config/aarch64/aarch64-simd-builtins.def: Use BINOPU type
4318 qualifiers in generator macros for u[r]hadd builtins.
4319 * config/aarch64/arm_neon.h (vhadd_s8): Remove unnecessary
4321 (vhadd_s16): Likewise.
4322 (vhadd_s32): Likewise.
4323 (vhadd_u8): Use type-qualified builtin and remove casts.
4324 (vhadd_u16): Likewise.
4325 (vhadd_u32): Likewise.
4326 (vhaddq_s8): Remove unnecessary cast.
4327 (vhaddq_s16): Likewise.
4328 (vhaddq_s32): Likewise.
4329 (vhaddq_u8): Use type-qualified builtin and remove casts.
4330 (vhaddq_u16): Likewise.
4331 (vhaddq_u32): Likewise.
4332 (vrhadd_s8): Remove unnecessary cast.
4333 (vrhadd_s16): Likewise.
4334 (vrhadd_s32): Likewise.
4335 (vrhadd_u8): Use type-qualified builtin and remove casts.
4336 (vrhadd_u16): Likewise.
4337 (vrhadd_u32): Likewise.
4338 (vrhaddq_s8): Remove unnecessary cast.
4339 (vrhaddq_s16): Likewise.
4340 (vrhaddq_s32): Likewise.
4341 (vrhaddq_u8): Use type-wualified builtin and remove casts.
4342 (vrhaddq_u16): Likewise.
4343 (vrhaddq_u32): Likewise.
4345 2021-11-11 Jonathan Wright <jonathan.wright@arm.com>
4347 * config/aarch64/aarch64-simd-builtins.def: Use BINOPU type
4348 qualifiers in generator macros for usub[lw][2] builtins.
4349 * config/aarch64/arm_neon.h (vsubl_s8): Remove unnecessary
4351 (vsubl_s16): Likewise.
4352 (vsubl_s32): Likewise.
4353 (vsubl_u8): Use type-qualified builtin and remove casts.
4354 (vsubl_u16): Likewise.
4355 (vsubl_u32): Likewise.
4356 (vsubl_high_s8): Remove unnecessary cast.
4357 (vsubl_high_s16): Likewise.
4358 (vsubl_high_s32): Likewise.
4359 (vsubl_high_u8): Use type-qualified builtin and remove casts.
4360 (vsubl_high_u16): Likewise.
4361 (vsubl_high_u32): Likewise.
4362 (vsubw_s8): Remove unnecessary casts.
4363 (vsubw_s16): Likewise.
4364 (vsubw_s32): Likewise.
4365 (vsubw_u8): Use type-qualified builtin and remove casts.
4366 (vsubw_u16): Likewise.
4367 (vsubw_u32): Likewise.
4368 (vsubw_high_s8): Remove unnecessary cast.
4369 (vsubw_high_s16): Likewise.
4370 (vsubw_high_s32): Likewise.
4371 (vsubw_high_u8): Use type-qualified builtin and remove casts.
4372 (vsubw_high_u16): Likewise.
4373 (vsubw_high_u32): Likewise.
4375 2021-11-11 Jonathan Wright <jonathan.wright@arm.com>
4377 * config/aarch64/aarch64-simd-builtins.def: Use BINOPU type
4378 qualifiers in generator macros for uadd[lw][2] builtins.
4379 * config/aarch64/arm_neon.h (vaddl_s8): Remove unnecessary
4381 (vaddl_s16): Likewise.
4382 (vaddl_s32): Likewise.
4383 (vaddl_u8): Use type-qualified builtin and remove casts.
4384 (vaddl_u16): Likewise.
4385 (vaddl_u32): Likewise.
4386 (vaddl_high_s8): Remove unnecessary cast.
4387 (vaddl_high_s16): Likewise.
4388 (vaddl_high_s32): Likewise.
4389 (vaddl_high_u8): Use type-qualified builtin and remove casts.
4390 (vaddl_high_u16): Likewise.
4391 (vaddl_high_u32): Likewise.
4392 (vaddw_s8): Remove unnecessary cast.
4393 (vaddw_s16): Likewise.
4394 (vaddw_s32): Likewise.
4395 (vaddw_u8): Use type-qualified builtin and remove casts.
4396 (vaddw_u16): Likewise.
4397 (vaddw_u32): Likewise.
4398 (vaddw_high_s8): Remove unnecessary cast.
4399 (vaddw_high_s16): Likewise.
4400 (vaddw_high_s32): Likewise.
4401 (vaddw_high_u8): Use type-qualified builtin and remove casts.
4402 (vaddw_high_u16): Likewise.
4403 (vaddw_high_u32): Likewise.
4405 2021-11-11 Jonathan Wright <jonathan.wright@arm.com>
4407 * config/aarch64/aarch64-simd-builtins.def: Declare type-
4408 qualified builtins for [R]SHRN[2].
4409 * config/aarch64/arm_neon.h (vshrn_n_u16): Use type-qualified
4410 builtin and remove casts.
4411 (vshrn_n_u32): Likewise.
4412 (vshrn_n_u64): Likewise.
4413 (vrshrn_high_n_u16): Likewise.
4414 (vrshrn_high_n_u32): Likewise.
4415 (vrshrn_high_n_u64): Likewise.
4416 (vrshrn_n_u16): Likewise.
4417 (vrshrn_n_u32): Likewise.
4418 (vrshrn_n_u64): Likewise.
4419 (vshrn_high_n_u16): Likewise.
4420 (vshrn_high_n_u32): Likewise.
4421 (vshrn_high_n_u64): Likewise.
4423 2021-11-11 Jonathan Wright <jonathan.wright@arm.com>
4425 * config/aarch64/aarch64-simd-builtins.def: Declare unsigned
4426 type-qualified builtins for XTN[2].
4427 * config/aarch64/arm_neon.h (vmovn_high_u16): Use type-
4428 qualified builtin and remove casts.
4429 (vmovn_high_u32): Likewise.
4430 (vmovn_high_u64): Likewise.
4431 (vmovn_u16): Likewise.
4432 (vmovn_u32): Likewise.
4433 (vmovn_u64): Likewise.
4435 2021-11-11 Jonathan Wright <jonathan.wright@arm.com>
4437 * config/aarch64/aarch64-simd-builtins.def: Use poly type
4438 qualifier in builtin generator macros.
4439 * config/aarch64/arm_neon.h (vmul_p8): Use type-qualified
4440 builtin and remove casts.
4441 (vmulq_p8): Likewise.
4442 (vmull_high_p8): Likewise.
4443 (vmull_p8): Likewise.
4445 2021-11-11 Jonathan Wright <jonathan.wright@arm.com>
4447 * config/aarch64/aarch64-simd-builtins.def: Declare type-
4448 qualified builtin generators for unsigned MLA/MLS intrinsics.
4449 * config/aarch64/arm_neon.h (vmla_n_u16): Use type-qualified
4451 (vmla_n_u32): Likewise.
4452 (vmla_u8): Likewise.
4453 (vmla_u16): Likewise.
4454 (vmla_u32): Likewise.
4455 (vmlaq_n_u16): Likewise.
4456 (vmlaq_n_u32): Likewise.
4457 (vmlaq_u8): Likewise.
4458 (vmlaq_u16): Likewise.
4459 (vmlaq_u32): Likewise.
4460 (vmls_n_u16): Likewise.
4461 (vmls_n_u32): Likewise.
4462 (vmls_u8): Likewise.
4463 (vmls_u16): Likewise.
4464 (vmls_u32): Likewise.
4465 (vmlsq_n_u16): Likewise.
4466 (vmlsq_n_u32): Likewise.
4467 (vmlsq_u8): Likewise.
4468 (vmlsq_u16): Likewise.
4469 (vmlsq_u32): Likewise.
4471 2021-11-11 Jan Hubicka <hubicka@ucw.cz>
4473 * ipa-modref.c (modref_summary::useful_p): Check also for side-effects
4474 with looping const/pure.
4475 (modref_summary_lto::useful_p): Likewise.
4476 (merge_call_side_effects): Merge side effects before early exit
4478 (process_fnspec): Also handle pure functions.
4479 (analyze_call): Do not early exit on looping pure const.
4480 (propagate_unknown_call): Also handle nontrivial SCC as side-effect.
4481 (modref_propagate_in_scc): Update.
4483 2021-11-11 Richard Biener <rguenther@suse.de>
4485 PR tree-optimization/103190
4486 * tree-ssa-reassoc.c (insert_stmt_after): Only assert on asm goto.
4488 2021-11-11 Aldy Hernandez <aldyh@redhat.com>
4490 * gimple-range-path.cc (path_range_query::add_copies_to_imports):
4492 (path_range_query::compute_imports): ...this. Adapt it so it can
4493 be passed the imports bitmap instead of working on m_imports.
4494 (path_range_query::compute_ranges): Call compute_imports in all
4495 cases unless an imports bitmap is passed.
4496 * gimple-range-path.h (path_range_query::compute_imports): New.
4497 (path_range_query::add_copies_to_imports): Remove.
4498 * tree-ssa-threadbackward.c (back_threader::resolve_def): Remove.
4499 (back_threader::find_paths_to_names): Inline resolve_def.
4500 (back_threader::find_paths): Call compute_imports.
4501 (back_threader::resolve_phi): Adjust comment.
4503 2021-11-11 Richard Biener <rguenther@suse.de>
4505 PR tree-optimization/103188
4506 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p):
4507 Remove query parameter, split out check for size
4509 (ch_base::m_ranger, cb_base::m_query): Remove.
4510 (ch_base::copy_headers): Split processing loop into
4511 analysis around which we allocate and use ranger and
4512 transform where we do not.
4513 (pass_ch::execute): Do not allocate/free ranger here.
4514 (pass_ch_vect::execute): Likewise.
4516 2021-11-11 Jan Hubicka <hubicka@ucw.cz>
4518 * ipa-pure-const.c (propagate_pure_const): Self recursion is
4521 2021-11-11 Jan Hubicka <hubicka@ucw.cz>
4523 * cgraph.c (set_noreturn_flag_1): New function.
4524 (cgraph_node::set_noreturn_flag): New member function
4525 * cgraph.h (cgraph_node::set_noreturn_flags): Declare.
4526 * ipa-pure-const.c (pass_local_pure_const::execute): Use it.
4528 2021-11-11 Aldy Hernandez <aldyh@redhat.com>
4529 Richard Biener <rguenther@suse.de>
4531 * tree-ssa-loop-ch.c (entry_loop_condition_is_static): Resolve
4532 statically to the edge remaining in the loop.
4534 2021-11-11 Richard Biener <rguenther@suse.de>
4536 PR middle-end/103181
4537 * tree-eh.c (operation_could_trap_helper_p): Properly
4538 check vector constants for a zero element for integer
4539 division. Separate floating point and integer division code.
4541 2021-11-11 Jakub Jelinek <jakub@redhat.com>
4544 * dwarf2out.c (field_byte_offset): Do the PCC_BITFIELD_TYPE_MATTERS
4545 handling only for DECL_BIT_FIELD_TYPE decls.
4547 2021-11-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
4550 * config/aarch64/aarch64.c (aarch64_process_target_attr): Check if
4551 token is arch extension without leading '+' and emit appropriate
4552 diagnostic for the same.
4554 2021-11-11 Jakub Jelinek <jakub@redhat.com>
4556 * tree.h (OMP_CLAUSE_NUM_TEAMS_EXPR): Rename to ...
4557 (OMP_CLAUSE_NUM_TEAMS_UPPER_EXPR): ... this.
4558 (OMP_CLAUSE_NUM_TEAMS_LOWER_EXPR): Define.
4559 * tree.c (omp_clause_num_ops): Increase num ops for
4560 OMP_CLAUSE_NUM_TEAMS to 2.
4561 * tree-pretty-print.c (dump_omp_clause): Print optional lower bound
4562 for OMP_CLAUSE_NUM_TEAMS.
4563 * gimplify.c (gimplify_scan_omp_clauses): Gimplify
4564 OMP_CLAUSE_NUM_TEAMS_LOWER_EXPR if non-NULL.
4565 (optimize_target_teams): Use OMP_CLAUSE_NUM_TEAMS_UPPER_EXPR instead
4566 of OMP_CLAUSE_NUM_TEAMS_EXPR. Handle OMP_CLAUSE_NUM_TEAMS_LOWER_EXPR.
4567 * omp-low.c (lower_omp_teams): Use OMP_CLAUSE_NUM_TEAMS_UPPER_EXPR
4568 instead of OMP_CLAUSE_NUM_TEAMS_EXPR.
4569 * omp-expand.c (expand_teams_call, get_target_arguments): Likewise.
4571 2021-11-11 Richard Biener <rguenther@suse.de>
4573 * cfganal.c (find_pdom): Remove.
4574 (control_dependences::find_control_dependence): Remove
4575 special-casing of entry block, call get_immediate_dominator
4577 * gimple-predicate-analysis.cc (find_pdom): Remove.
4578 (find_dom): Likewise.
4579 (find_control_equiv_block): Call get_immediate_dominator
4581 (compute_control_dep_chain): Likewise.
4582 (predicate::init_from_phi_def): Likewise.
4584 2021-11-11 Richard Biener <rguenther@suse.de>
4586 * cfganal.h (control_dependences::control_dependence_map):
4588 (control_dependences::m_bitmaps): New.
4589 * cfganal.c (control_dependences::set_control_dependence_map_bit):
4591 (control_dependences::clear_control_dependence_bitmap):
4593 (control_dependences::find_control_dependence): Do not
4594 find_edge for the abnormal edge test.
4595 (control_dependences::control_dependences): Instead do not
4596 add abnormal edges to the edge list. Adjust.
4597 (control_dependences::~control_dependences): Likewise.
4598 (control_dependences::get_edges_dependent_on): Likewise.
4599 * function-tests.c: Include bitmap.h.
4601 2021-11-11 Kewen Lin <linkw@linux.ibm.com>
4603 * doc/invoke.texi: Change references to "future cpu" to "power10",
4604 "-mcpu=future" to "-mcpu=power10". Adjust words for float128.
4606 2021-11-11 Cui,Lili <lili.cui@intel.com>
4608 * config/i386/i386-options.c (m_CORE_AVX2): Remove Alderlake
4610 (processor_cost_table): Use alderlake_cost for Alderlake.
4611 * config/i386/i386.c (ix86_sched_init_global): Handle Alderlake.
4612 * config/i386/x86-tune-costs.h (struct processor_costs): Add alderlake
4614 * config/i386/x86-tune-sched.c (ix86_issue_rate): Change Alderlake
4616 (ix86_adjust_cost): Handle Alderlake.
4617 * config/i386/x86-tune.def (X86_TUNE_SCHEDULE): Enable for Alderlake.
4618 (X86_TUNE_PARTIAL_REG_DEPENDENCY): Likewise.
4619 (X86_TUNE_SSE_PARTIAL_REG_DEPENDENCY): Likewise.
4620 (X86_TUNE_SSE_PARTIAL_REG_FP_CONVERTS_DEPENDENCY): Likewise.
4621 (X86_TUNE_SSE_PARTIAL_REG_CONVERTS_DEPENDENCY): Likewise.
4622 (X86_TUNE_MEMORY_MISMATCH_STALL): Likewise.
4623 (X86_TUNE_USE_LEAVE): Likewise.
4624 (X86_TUNE_PUSH_MEMORY): Likewise.
4625 (X86_TUNE_USE_INCDEC): Likewise.
4626 (X86_TUNE_INTEGER_DFMODE_MOVES): Likewise.
4627 (X86_TUNE_MISALIGNED_MOVE_STRING_PRO_EPILOGUES): Likewise.
4628 (X86_TUNE_USE_SAHF): Likewise.
4629 (X86_TUNE_USE_BT): Likewise.
4630 (X86_TUNE_AVOID_FALSE_DEP_FOR_BMI): Likewise.
4631 (X86_TUNE_ONE_IF_CONV_INSN): Likewise.
4632 (X86_TUNE_AVOID_MFENCE): Likewise.
4633 (X86_TUNE_USE_SIMODE_FIOP): Likewise.
4634 (X86_TUNE_EXT_80387_CONSTANTS): Likewise.
4635 (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL): Likewise.
4636 (X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Likewise.
4637 (X86_TUNE_SSE_TYPELESS_STORES): Likewise.
4638 (X86_TUNE_SSE_LOAD0_BY_PXOR): Likewise.
4639 (X86_TUNE_AVOID_4BYTE_PREFIXES): Likewise.
4640 (X86_TUNE_USE_GATHER): Disable for Alderlake.
4641 (X86_TUNE_AVX256_MOVE_BY_PIECES): Likewise.
4642 (X86_TUNE_AVX256_STORE_BY_PIECES): Likewise.
4644 2021-11-11 liuhongt <hongtao.liu@intel.com>
4647 * config/i386/sse.md (V_128_256): Extend to V8HF/V16HF.
4648 (avxsizesuffix): Ditto.
4650 2021-11-11 Kito Cheng <kito.cheng@sifive.com>
4652 * common/config/riscv/riscv-common.c (riscv_subset_list::to_string): Fix
4653 wrong marco checking.
4655 2021-11-10 Aldy Hernandez <aldyh@redhat.com>
4657 PR tree-optimization/102906
4658 * tree-ssa-loop-ch.c (entry_loop_condition_is_static): New.
4659 (should_duplicate_loop_header_p): Call entry_loop_condition_is_static.
4660 (class ch_base): Add m_ranger and m_query.
4661 (ch_base::copy_headers): Pass m_query to
4662 entry_loop_condition_is_static.
4663 (pass_ch::execute): Allocate and deallocate m_ranger and
4665 (pass_ch_vect::execute): Same.
4667 2021-11-10 Andrew Pinski <apinski@marvell.com>
4670 * config/aarch64/aarch64-simd.md (aarch64_simd_dup<mode>):
4671 Use vwcore iterator for the r constraint output string.
4673 2021-11-10 qing zhao <qing.zhao@oracle.com>
4675 * internal-fn.c (expand_DEFERRED_INIT): Apply pattern initialization
4676 only when have_insn_for return true for the mode. Fix a memory leak.
4678 2021-11-10 Christophe Lyon <christophe.lyon@foss.st.com>
4680 * config/arm/arm.c (cortexa9_extra_costs, cortexa8_extra_costs,
4681 cortexa5_extra_costs, cortexa7_extra_costs,
4682 cortexa12_extra_costs, cortexa15_extra_costs, v7m_extra_costs):
4683 Initialize movi, dup and extract costing fields.
4685 2021-11-10 Aldy Hernandez <aldyh@redhat.com>
4687 * gimple-range-path.cc (path_range_query::path_range_query): Do
4689 (path_range_query::dump): Change m_path uses to non-pointer.
4690 (path_range_query::defined_outside_path): Same.
4691 (path_range_query::set_path): Same.
4692 (path_range_query::add_copies_to_imports): Same.
4693 (path_range_query::range_of_stmt): Same.
4694 (path_range_query::compute_outgoing_relations): Same.
4695 (path_range_query::compute_ranges): Imports are now optional.
4696 Implement overload that takes an edge.
4697 * gimple-range-path.h (class path_range_query): Make imports
4698 optional for compute_ranges. Add compute_ranges(edge) overload.
4699 Make m_path an auto_vec instead of a pointer and adjust
4702 2021-11-10 Tamar Christina <tamar.christina@arm.com>
4704 * tree-vectorizer.h (struct scalar_cond_masked_key): Add inverted_p.
4705 (default_hash_traits<scalar_conf_masked_key>): Likewise.
4706 * tree-vect-stmts.c (vectorizable_condition): Check if inverse of mask
4708 * tree-vectorizer.c (scalar_cond_masked_key::get_cond_ops_from_tree):
4709 Register mask inverses.
4711 2021-11-10 Tamar Christina <tamar.christina@arm.com>
4713 * tree-vectorizer.c (vectorize_loops): Do local CSE through RPVN upon
4714 successful vectorization.
4716 2021-11-10 Andrew MacLeod <amacleod@redhat.com>
4718 * gimple-range-cache.cc (sbr_vector::grow): New.
4719 (sbr_vector::set_bb_range): Call grow.
4720 (sbr_vector::get_bb_range): Same.
4721 (sbr_vector::bb_range_p): Remove assert.
4723 2021-11-10 Tamar Christina <tamar.christina@arm.com>
4725 * config/aarch64/aarch64-simd.md (*aarch64_topbits_shuffle<mode>_le
4726 ,*aarch64_topbits_shuffle<mode>_be): Remove.
4728 2021-11-10 Jan Hubicka <jh@suse.cz>
4730 * ipa-modref.c: Include tree-eh.h
4731 (modref_summary::modref_summary): Initialize side_effects.
4732 (struct modref_summary_lto): New bool field side_effects.
4733 (modref_summary_lto::modref_summary_lto): Initialize side_effects.
4734 (modref_summary::dump): Dump side_effects.
4735 (modref_summary_lto::dump): Dump side_effects.
4736 (merge_call_side_effects): Merge side effects.
4737 (process_fnspec): Calls to non-const/pure or looping
4738 function is a side effect.
4739 (analyze_call): Self-recursion is a side-effect; handle
4741 (analyze_load): Watch for volatile and throwing memory.
4742 (analyze_store): Likewise.
4743 (analyze_stmt): Watch for volatitle asm.
4744 (analyze_function): Handle side_effects.
4745 (modref_summaries::duplicate): Duplicate side_effects.
4746 (modref_summaries_lto::duplicate): Likewise.
4747 (modref_write): Stream side_effects.
4748 (read_section): Likewise.
4749 (update_signature): Update.
4750 (propagate_unknown_call): Handle side_effects.
4751 (modref_propagate_in_scc): Likewise.
4752 * ipa-modref.h (struct modref_summary): Add side_effects.
4753 * ipa-pure-const.c (special_builtin_state): Rename to ...
4754 (builtin_safe_for_const_function_p): ... this one.
4755 (check_call): Update.
4756 (finite_function_p): Break out from ...
4757 (propagate_pure_const): ... here
4758 * ipa-utils.h (finite_function): Declare.
4760 2021-11-10 Lucas A. M. Magalhães <lamm@linux.ibm.com>
4762 * config.gcc (powerpc*-*-*): Remove -rpath from
4763 --with-advance-toolchain.
4765 2021-11-10 Marek Polacek <polacek@redhat.com>
4768 * attribs.c (struct scoped_attributes): Add a bool member.
4769 (lookup_scoped_attribute_spec): Forward declare.
4770 (register_scoped_attributes): New bool parameter, defaulted to
4772 (handle_ignored_attributes_option): New function.
4773 (free_attr_data): New function.
4774 (init_attributes): Call handle_ignored_attributes_option.
4775 (attr_namespace_ignored_p): New function.
4776 (decl_attributes): Check attr_namespace_ignored_p before
4778 * attribs.h (free_attr_data): Declare.
4779 (register_scoped_attributes): Adjust declaration.
4780 (handle_ignored_attributes_option): Declare.
4781 (canonicalize_attr_name): New function template.
4782 (canonicalize_attr_name): Use it.
4783 * common.opt (Wattributes=): New option with a variable.
4784 * doc/extend.texi: Document #pragma GCC diagnostic ignored_attributes.
4785 * doc/invoke.texi: Document -Wno-attributes=.
4786 * opts.c (common_handle_option) <case OPT_Wattributes_>: Handle.
4787 * plugin.h (register_scoped_attributes): Adjust declaration.
4788 * toplev.c (compile_file): Call free_attr_data.
4790 2021-11-10 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
4792 * config/arm/arm-cpus.in (cortex-a710): New CPU.
4793 * config/arm/arm-tables.opt: Regenerate.
4794 * config/arm/arm-tune.md: Regenerate.
4795 * doc/invoke.texi: Update docs.
4797 2021-11-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
4799 * config/aarch64/aarch64-builtins.c
4800 (aarch64_general_gimple_fold_builtin): Mark argument as unused.
4802 2021-11-10 Martin Liska <mliska@suse.cz>
4804 * lto-wrapper.c (merge_and_complain): Make the first argument
4807 2021-11-10 Richard Sandiford <richard.sandiford@arm.com>
4809 * config/aarch64/iterators.md (optab): Use fmax_nan instead of
4810 smax_nan and fmin_nan instead of smin_nan.
4811 (maxmin_uns): Rename to...
4812 (fmaxmin): ...this and make the same changes. Remove entries
4813 unrelated to fmax* and fmin*.
4814 * config/aarch64/aarch64.md (<maxmin_uns><mode>3): Rename to...
4815 (<fmaxmin><mode>3): ...this.
4816 * config/aarch64/aarch64-simd.md (aarch64_<maxmin_uns>p<mode>):
4818 (aarch64_<optab>p<mode>): ...this.
4819 (<maxmin_uns><mode>3): Rename to...
4820 (<fmaxmin><mode>3): ...this.
4821 (reduc_<maxmin_uns>_scal_<mode>): Rename to...
4822 (reduc_<optab>_scal_<mode>): ...this and update gen* call.
4823 (aarch64_reduc_<maxmin_uns>_internal<mode>): Rename to...
4824 (aarch64_reduc_<optab>_internal<mode>): ...this.
4825 (aarch64_reduc_<maxmin_uns>_internalv2si): Rename to...
4826 (aarch64_reduc_<optab>_internalv2si): ...this.
4827 * config/aarch64/aarch64-sve.md (<maxmin_uns><mode>3): Rename to...
4828 (<fmaxmin><mode>3): ...this.
4829 * config/aarch64/aarch64-simd-builtins.def (smax_nan, smin_nan)
4831 (fmax_nan, fmin_nan): ...this.
4832 * config/aarch64/arm_neon.h (vmax_f32, vmax_f64, vmaxq_f32, vmaxq_f64)
4833 (vmin_f32, vmin_f64, vminq_f32, vminq_f64, vmax_f16, vmaxq_f16)
4834 (vmin_f16, vminq_f16): Update accordingly.
4836 2021-11-10 Richard Sandiford <richard.sandiford@arm.com>
4838 * tree-vectorizer.h (vector_costs::finish_cost): Take the
4839 corresponding scalar costs as a parameter.
4840 (finish_cost): Likewise.
4841 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
4842 (vect_estimate_min_profitable_iters): Update accordingly.
4843 * tree-vect-slp.c (vect_bb_vectorization_profitable_p): Likewise.
4844 * tree-vectorizer.c (vector_costs::finish_cost): Likewise.
4845 * config/aarch64/aarch64.c (aarch64_vector_costs::finish_cost):
4847 * config/rs6000/rs6000.c (rs6000_cost_data::finish_cost): Likewise.
4849 2021-11-10 Richard Sandiford <richard.sandiford@arm.com>
4851 * tree-vectorizer.h (_loop_vec_info::scalar_costs): New member
4853 (_loop_vec_info::single_scalar_iteration_cost): Delete.
4854 (LOOP_VINFO_SINGLE_SCALAR_ITERATION_COST): Delete.
4855 (vector_costs::total_cost): New function.
4856 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Update
4857 after above changes.
4858 (_loop_vec_info::~_loop_vec_info): Delete scalar_costs.
4859 (vect_compute_single_scalar_iteration_cost): Store the costs
4860 in loop_vinfo->scalar_costs.
4861 (vect_estimate_min_profitable_iters): Get the scalar cost from
4862 loop_vinfo->scalar_costs.
4864 2021-11-10 Richard Sandiford <richard.sandiford@arm.com>
4866 * tree-vectorizer.h (vector_costs::better_main_loop_than_p)
4867 (vector_costs::better_epilogue_loop_than_p)
4868 (vector_costs::compare_inside_loop_cost)
4869 (vector_costs::compare_outside_loop_cost): Likewise.
4870 * tree-vectorizer.c (vector_costs::better_main_loop_than_p)
4871 (vector_costs::better_epilogue_loop_than_p)
4872 (vector_costs::compare_inside_loop_cost)
4873 (vector_costs::compare_outside_loop_cost): New functions,
4874 containing code moved from...
4875 * tree-vect-loop.c (vect_better_loop_vinfo_p): ...here.
4877 2021-11-10 Richard Sandiford <richard.sandiford@arm.com>
4879 * tree-vectorizer.h (_loop_vec_info): Remove vec_outside_cost
4880 and vec_inside_cost.
4881 (vector_costs::outside_cost): New function.
4882 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Update
4884 (vect_estimate_min_profitable_iters): Likewise.
4885 (vect_better_loop_vinfo_p): Get the inside and outside costs
4886 from the loop_vec_infos' vector_costs.
4888 2021-11-10 Richard Sandiford <richard.sandiford@arm.com>
4890 * tree-vectorizer.h (vec_info::target_cost_data): Replace with...
4891 (_loop_vec_info::vector_costs): ...this.
4892 (LOOP_VINFO_TARGET_COST_DATA): Delete.
4893 * tree-vectorizer.c (vec_info::vec_info): Remove target_cost_data
4895 (vec_info::~vec_info): Remove corresponding delete.
4896 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
4897 vector_costs to null.
4898 (_loop_vec_info::~_loop_vec_info): Delete vector_costs.
4899 (vect_analyze_loop_operations): Update after above changes.
4900 (vect_analyze_loop_2): Likewise.
4901 (vect_estimate_min_profitable_iters): Likewise.
4902 * tree-vect-slp.c (vect_slp_analyze_operations): Likewise.
4904 2021-11-10 Jan Hubicka <hubicka@ucw.cz>
4906 * tree-core.h (EAF_DIRECT): Remove.
4907 (EAF_NOCLOBBER): Remove.
4908 (EAF_UNUSED): Remove.
4909 (EAF_NOESCAPE): Remove.
4910 (EAF_NO_DIRECT_CLOBBER): New.
4911 (EAF_NO_INDIRECT_CLOBBER): New.
4912 (EAF_NODIRECTESCAPE): Remove.
4913 (EAF_NO_DIRECT_ESCAPE): New.
4914 (EAF_NO_INDIRECT_ESCAPE): New.
4915 (EAF_NOT_RETURNED): Remove.
4916 (EAF_NOT_RETURNED_INDIRECTLY): New.
4917 (EAF_NOREAD): Remove.
4918 (EAF_NO_DIRECT_READ): New.
4919 (EAF_NO_INDIRECT_READ): New.
4920 * gimple.c (gimple_call_arg_flags): Update for new flags.
4921 (gimple_call_retslot_flags): Update for new flags.
4922 * ipa-modref.c (dump_eaf_flags): Likewise.
4923 (remove_useless_eaf_flags): Likewise.
4924 (deref_flags): Likewise.
4925 (modref_lattice::init): Likewise.
4926 (modref_lattice::merge): Likewise.
4927 (modref_lattice::merge_direct_load): Likewise.
4928 (modref_lattice::merge_direct_store): Likewise.
4929 (modref_eaf_analysis::merge_call_lhs_flags): Likewise.
4930 (callee_to_caller_flags): Likewise.
4931 (modref_eaf_analysis::analyze_ssa_name): Likewise.
4932 (modref_eaf_analysis::propagate): Likewise.
4933 (modref_merge_call_site_flags): Likewise.
4934 * ipa-modref.h (interposable_eaf_flags): Likewise.
4935 * tree-ssa-alias.c: (ref_maybe_used_by_call_p_1) Likewise.
4936 * tree-ssa-structalias.c (handle_call_arg): Likewise.
4937 (handle_rhs_call): Likewise.
4938 * tree-ssa-uninit.c (maybe_warn_pass_by_reference): Likewise.
4940 2021-11-10 Tamar Christina <tamar.christina@arm.com>
4942 * tree-vect-slp-patterns.c (complex_mul_pattern::matches): Remove l1node.
4944 2021-11-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
4946 * config/aarch64/aarch64-builtins.c
4947 (aarch64_general_gimple_fold_builtin): Change pointer alignment and
4950 2021-11-10 Jan Hubicka <jh@suse.cz>
4952 * ipa-modref-tree.h (modref_tree::remap_params): Fix off-by-one error.
4954 2021-11-10 H.J. Lu <hongjiu.lu@intel.com>
4955 Hongtao Liu <hongtao.liu@intel.com>
4957 PR middle-end/102566
4958 * match.pd (nop_atomic_bit_test_and_p): New match.
4959 * tree-ssa-ccp.c (convert_atomic_bit_not): New function.
4960 (gimple_nop_atomic_bit_test_and_p): New prototype.
4961 (optimize_atomic_bit_test_and): Transform equivalent, but slighly
4962 different cases to their canonical forms.
4964 2021-11-10 liuhongt <hongtao.liu@intel.com>
4966 PR tree-optimization/103126
4967 * tree-vect-loop.c (neutral_op_for_reduction): Remove static.
4968 * tree-vectorizer.h (neutral_op_for_reduction): Declare.
4969 * tree-if-conv.c : Include tree-vectorizer.h.
4970 (is_cond_scalar_reduction): Handle
4971 BIT_XOR_EXPR/BIT_IOR_EXPR/BIT_AND_EXPR.
4972 (convert_scalar_cond_reduction): Ditto.
4974 2021-11-10 konglin1 <lingling.kong@intel.com>
4976 * config/i386/sse.md (cmul<conj_op><mode>3): add new define_expand.
4977 (cmla<conj_op><mode>4): Likewise
4979 2021-11-10 Aldy Hernandez <aldyh@redhat.com>
4981 * tree-ssa-threadedge.c: Do not include
4982 gimple-ssa-evrp-analyze.h.
4983 * value-pointer-equiv.cc: Same.
4985 2021-11-10 Aldy Hernandez <aldyh@redhat.com>
4987 * tree-ssa-threadbackward.c
4988 (back_threader::maybe_register_path_dump): Abstract path dumping...
4989 (dump_path): ...here.
4990 (back_threader::resolve_phi): Call dump_path.
4993 2021-11-10 konglin1 <lingling.kong@intel.com>
4995 * config/i386/sse.md (fma_<complexpairopname>_<mode>_pair):
4996 Add new define_insn.
4997 (fma_<mode>_fmaddc_bcst): Add new define_insn_and_split.
4998 (fma_<mode>_fcmaddc_bcst): Likewise
5000 2021-11-10 liuhongt <hongtao.liu@intel.com>
5003 * match.pd: Simplify (trunc)fmax/fmin((extend)a, (extend)b) to
5006 2021-11-10 Andrew Pinski <apinski@marvell.com>
5009 * config/aarch64/aarch64.c (aarch64_evpc_ins): Don't use target
5010 as an input, use original one.
5012 2021-11-10 Sandra Loosemore <sandra@codesourcery.com>
5014 * config/nios2/nios2.c (nios2_can_inline_p): New.
5015 (TARGET_CAN_INLINE_P): Define.
5017 2021-11-09 Bill Schmidt <wschmidt@linux.ibm.com>
5019 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_new_builtin):
5020 Disable gimple fold for RS6000_BIF_{XVMINDP,XVMINSP,VMINFP} and
5021 RS6000_BIF_{XVMAXDP,XVMAXSP,VMAXFP} when fast-math is not set.
5022 (lxvrse_expand_builtin): Modify the expansion for sign extension.
5023 All extensions are done within VSX registers.
5025 2021-11-09 Andrew Pinski <apinski@marvell.com>
5027 PR tree-optimization/10352
5028 * match.pd: Remove check of TYPE_PRECISION for
5031 2021-11-09 Andrew MacLeod <amacleod@redhat.com>
5033 * function.c (allocate_struct_function): Don't set x_range_query.
5034 * function.h (get_range_query): Move to value-query.h.
5035 * gimple-range.cc (enable_ranger): Check that query is currently NULL.
5036 (disable_ranger): Clear function current query field.
5037 * value-query.cc (get_global_range_query): Relocate to:
5038 * value-query.h (get_global_range_query): Here and inline.
5039 (get_range_query): Relocate here from function.h.
5041 2021-11-09 Aldy Hernandez <aldyh@redhat.com>
5043 * tree-ssa-threadbackward.c
5044 (back_threader::maybe_register_path_dump): New.
5045 (back_threader::maybe_register_path): Call maybe_register_path_dump.
5047 2021-11-09 Aldy Hernandez <aldyh@redhat.com>
5049 * tree-ssa-threadbackward.c (back_threader::maybe_register_path):
5050 Return NULL when unprofitable.
5052 2021-11-09 Martin Jambor <mjambor@suse.cz>
5054 * tree.h (build_debug_expr_decl): Declare.
5055 * tree.c (build_debug_expr_decl): New function.
5056 * cfgexpand.c (avoid_deep_ter_for_debug): Use build_debug_expr_decl
5057 instead of building a DEBUG_EXPR_DECL.
5058 * ipa-param-manipulation.c
5059 (ipa_param_body_adjustments::prepare_debug_expressions): Likewise.
5060 * omp-simd-clone.c (ipa_simd_modify_stmt_ops): Likewise.
5061 * tree-ssa-ccp.c (optimize_atomic_bit_test_and): Likewise.
5062 * tree-ssa-phiopt.c (spaceship_replacement): Likewise.
5063 * tree-ssa-reassoc.c (make_new_ssa_for_def): Likewise.
5065 2021-11-09 Bill Schmidt <wschmidt@linux.ibm.com>
5067 * config/rs6000/rs6000-call.c (rs6000_type_string): New function.
5068 (def_builtin): Change debug formatting for easier parsing and
5069 include more information.
5070 (rs6000_init_builtins): Add dump of autogenerated builtins.
5071 (altivec_init_builtins): Dump __builtin_altivec_mask_for_load for
5074 2021-11-09 Marek Polacek <polacek@redhat.com>
5076 * ipa-modref.c (analyze_parms): Fix a typo.
5078 2021-11-09 Kito Cheng <kito.cheng@sifive.com>
5081 * common/config/riscv/riscv-common.c (multi_letter_subset_rank): Remove
5082 assertion for Z*-ext.
5084 2021-11-09 Jan Hubicka <hubicka@ucw.cz>
5086 * ipa-modref.c (analyze_parms): Add past_flags, past_retslot_flags
5087 and past_static_chain; merge past summary with current one.
5088 (analyze_function): Update.
5090 2021-11-09 Aldy Hernandez <aldyh@redhat.com>
5092 * doc/invoke.texi (Invoking GCC): Document --param=threader-debug.
5094 2021-11-09 Martin Liska <mliska@suse.cz>
5096 * print-rtl.c (rtx_writer::rtx_writer): Make the compilation
5097 conditional based on
5098 * print-rtl.h (class rtx_writer): Likewise.
5100 2021-11-09 Thomas Schwinge <thomas@codesourcery.com>
5102 * input.h (location_hash): Use 'BUILTINS_LOCATION' as spare value
5103 for 'Deleted'. Turn into a '#define'.
5105 2021-11-09 Aldy Hernandez <aldyh@redhat.com>
5107 * gimple-range-path.cc (path_range_query::dump): Clean up.
5108 (path_range_query::compute_ranges): Same.
5109 * value-relation.cc (path_oracle::dump): Same.
5111 2021-11-09 Aldy Hernandez <aldyh@redhat.com>
5113 * dumpfile.c (dump_options): Remove TDF_THREADING entry.
5114 * dumpfile.h (enum dump_flag): Remove TDF_THREADING and adjust
5116 * flag-types.h (enum threader_debug): New.
5117 * gimple-range-path.cc (DEBUG_SOLVER): Use param_threader_debug.
5118 * params.opt: Add entry for --param=threader-debug=.
5120 2021-11-09 Thomas Schwinge <thomas@codesourcery.com>
5122 PR middle-end/101204
5124 * diagnostic-spec.h (typedef xint_hash_t): Turn into...
5125 (struct xint_hash_t): ... this.
5126 * doc/gty.texi: Update.
5128 2021-11-09 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
5130 * config/arm/arm-cpus.in (armv9): New define.
5131 (ARMv9a): New group.
5132 (armv9-a): New arch definition.
5133 * config/arm/arm-tables.opt: Regenerate.
5134 * config/arm/arm.h (BASE_ARCH_9A): New arch enum value.
5135 * config/arm/t-aprofile: Added armv9-a and armv9+simd.
5136 * config/arm/t-arm-elf: Added arm9-a, v9_fps and all_v9_archs
5137 to MULTILIB_MATCHES.
5138 * config/arm/t-multilib: Added v9_a_nosimd_variants and
5139 v9_a_simd_variants to MULTILIB_MATCHES.
5140 * doc/invoke.texi: Update docs.
5142 2021-11-09 Martin Jambor <mjambor@suse.cz>
5145 * ipa-param-manipulation.c (replace_with_mapped_expr): Early
5146 return with error_mark_mode when part of expression is mapped to
5148 (ipa_param_body_adjustments::remap_with_debug_expressions): Set
5149 mapped value to NULL if walk_tree returns error_mark_mode.
5151 2021-11-09 Aldy Hernandez <aldyh@redhat.com>
5153 * tree-ssa-strlen.c (compare_nonzero_chars): Pass statement
5155 (get_addr_stridx): Same.
5157 (get_range_strlen_dynamic): Same.
5158 (handle_builtin_strlen): Same.
5159 (handle_builtin_strchr): Same.
5160 (handle_builtin_strcpy): Same.
5161 (maybe_diag_stxncpy_trunc): Same.
5162 (handle_builtin_stxncpy_strncat): Same.
5163 (handle_builtin_memcpy): Same.
5164 (handle_builtin_strcat): Same.
5165 (handle_alloc_call): Same.
5166 (handle_builtin_memset): Same.
5167 (handle_builtin_string_cmp): Same.
5168 (handle_pointer_plus): Same.
5169 (count_nonzero_bytes_addr): Same.
5170 (count_nonzero_bytes): Same.
5171 (handle_store): Same.
5172 (fold_strstr_to_strncmp): Same.
5173 (handle_integral_assign): Same.
5174 (check_and_optimize_stmt): Same.
5175 (class strlen_dom_walker): Replace evrp with ranger.
5176 (strlen_dom_walker::before_dom_children): Remove evrp.
5177 (strlen_dom_walker::after_dom_children): Remove evrp.
5178 * gimple-ssa-warn-access.cc (maybe_check_access_sizes):
5179 Restrict sprintf output.
5181 2021-11-09 Martin Liska <mliska@suse.cz>
5183 * genconditions.c (write_one_condition): Add const qualifier
5186 2021-11-09 Jeff Law <jeffreyalaw@gmail.com>
5188 * match.pd: New pattern to simplify (1 << n) & M ==/!= 0 for M
5191 2021-11-08 Bill Schmidt <wschmidt@linux.ibm.com>
5193 * config/rs6000/rs6000-builtin-new.def (VEC_INIT_V16QI): Use
5194 escape-newline support.
5195 (VEC_INIT_V4SI): Likewise.
5196 (VEC_INIT_V8HI): Likewise.
5197 (PACK_V1TI): Likewise.
5199 (VFIRSTMISMATCHOREOSINDEX_V16QI): Likewise.
5200 (VFIRSTMISMATCHOREOSINDEX_V8HI): Likewise.
5201 (VFIRSTMISMATCHOREOSINDEX_V4SI): Likewise.
5205 (FMAF128_ODD): Likewise.
5206 (VSCEQPUO): Likewise.
5208 (VSIEQPF): Likewise.
5209 (VSTDCQP): Likewise.
5210 (PACK_TD): Likewise.
5211 (TABORTDC): Likewise.
5212 (TABORTDCI): Likewise.
5213 (SE_LXVRBX): Likewise.
5214 (SE_LXVRHX): Likewise.
5215 (SE_LXVRWX): Likewise.
5216 (SE_LXVRDX): Likewise.
5217 (VREPLACE_UN_UV2DI): Likewise.
5218 (VREPLACE_UN_UV4SI): Likewise.
5219 (VREPLACE_UN_V2DI): Likewise.
5220 (VREPLACE_ELT_UV2DI): Likewise.
5221 (VREPLACE_ELT_V2DI): Likewise.
5222 (ZE_LXVRBX): Likewise.
5223 (ZE_LXVRHX): Likewise.
5224 (ZE_LXVRWX): Likewise.
5225 (ZE_LXVRDX): Likewise.
5227 (CNTLZDM): Likewise.
5228 (CNTTZDM): Likewise.
5231 (PMXVBF16GER2): Likewise.
5232 (PMXVBF16GER2_INTERNAL): Likewise.
5233 (PMXVBF16GER2NN): Likewise.
5234 (PMXVBF16GER2NN_INTERNAL): Likewise.
5235 (PMXVBF16GER2NP): Likewise.
5236 (PMXVBF16GER2NP_INTERNAL): Likewise.
5237 (PMXVBF16GER2PN): Likewise.
5238 (PMXVBF16GER2PN_INTERNAL): Likewise.
5239 (PMXVBF16GER2PP): Likewise.
5240 (PMXVBF16GER2PP_INTERNAL): Likewise.
5241 (PMXVF16GER2): Likewise.
5242 (PMXVF16GER2_INTERNAL): Likewise.
5243 (PMXVF16GER2NN): Likewise.
5244 (PMXVF16GER2NN_INTERNAL): Likewise.
5245 (PMXVF16GER2NP): Likewise.
5246 (PMXVF16GER2NP_INTERNAL): Likewise.
5247 (PMXVF16GER2PN): Likewise.
5248 (PMXVF16GER2PN_INTERNAL): Likewise.
5249 (PMXVF16GER2PP): Likewise.
5250 (PMXVF16GER2PP_INTERNAL): Likewise.
5251 (PMXVF32GER_INTERNAL): Likewise.
5252 (PMXVF32GERNN): Likewise.
5253 (PMXVF32GERNN_INTERNAL): Likewise.
5254 (PMXVF32GERNP): Likewise.
5255 (PMXVF32GERNP_INTERNAL): Likewise.
5256 (PMXVF32GERPN): Likewise.
5257 (PMXVF32GERPN_INTERNAL): Likewise.
5258 (PMXVF32GERPP): Likewise.
5259 (PMXVF32GERPP_INTERNAL): Likewise.
5260 (PMXVF64GER): Likewise.
5261 (PMXVF64GER_INTERNAL): Likewise.
5262 (PMXVF64GERNN): Likewise.
5263 (PMXVF64GERNN_INTERNAL): Likewise.
5264 (PMXVF64GERNP): Likewise.
5265 (PMXVF64GERNP_INTERNAL): Likewise.
5266 (PMXVF64GERPN): Likewise.
5267 (PMXVF64GERPN_INTERNAL): Likewise.
5268 (PMXVF64GERPP): Likewise.
5269 (PMXVF64GERPP_INTERNAL): Likewise.
5270 (PMXVI16GER2): Likewise.
5271 (PMXVI16GER2_INTERNAL): Likewise.
5272 (PMXVI16GER2PP): Likewise.
5273 (PMXVI16GER2PP_INTERNAL): Likewise.
5274 (PMXVI16GER2S): Likewise.
5275 (PMXVI16GER2S_INTERNAL): Likewise.
5276 (PMXVI16GER2SPP): Likewise.
5277 (PMXVI16GER2SPP_INTERNAL): Likewise.
5278 (PMXVI4GER8): Likewise.
5279 (PMXVI4GER8_INTERNAL): Likewise.
5280 (PMXVI4GER8PP): Likewise.
5281 (PMXVI4GER8PP_INTERNAL): Likewise.
5282 (PMXVI8GER4): Likewise.
5283 (PMXVI8GER4_INTERNAL): Likewise.
5284 (PMXVI8GER4PP): Likewise.
5285 (PMXVI8GER4PP_INTERNAL): Likewise.
5286 (PMXVI8GER4SPP): Likewise.
5287 (PMXVI8GER4SPP_INTERNAL): Likewise.
5288 * config/rs6000/rs6000-gen-builtins.c (MAXLINES): New macro.
5289 (linebuf): Increase size.
5290 (lines): New variable.
5291 (lastline): Likewise.
5292 (real_line_pos): New function.
5293 (diag): Change signature.
5294 (bif_diag): Change signature; support escape-newline handling.
5295 (ovld_diag): Likewise.
5296 (fatal): Move earlier.
5297 (consume_whitespace): Adjust diag call.
5298 (advance_line): Add escape-newline handling; call fatal.
5299 (safe_inc_pos): Adjust diag call.
5300 (match_identifier): Likewise.
5301 (match_integer): Likewise.
5302 (match_to_right_bracket): Call fatal instead of diag; adjust diag
5304 (match_basetype): Adjust diag calls.
5305 (match_bracketed_pair): Likewise.
5306 (match_const_restriction): Likewise.
5307 (match_type): Likewise.
5308 (parse_args): Likewise.
5309 (parse_bif_attrs): Likewise.
5310 (complete_vector_type): Likewise.
5311 (complete_base_type): Likewise.
5312 (parse_prototype): Likewise.
5313 (parse_bif_entry): Likewise.
5314 (parse_bif_stanza): Likewise.
5315 (parse_ovld_entry): Likewise.
5316 (parse_ovld_stanza): Likewise.
5317 (main): Allocate buffers for lines[].
5319 2021-11-08 Eric Botcazou <ebotcazou@adacore.com>
5321 * config/i386/i386.c (legitimize_pic_address): Adjust comment and
5322 use the REG argument on the CM_LARGE_PIC code path as well.
5323 * config/i386/predicates.md (gotoff_operand): Do not treat VxWorks
5324 specially with the large code models.
5326 2021-11-08 Jan Hubicka <hubicka@ucw.cz>
5328 * gimple.c (gimple_call_static_chain_flags): Revert the workaround
5329 allowing interposition since issues with binds_to_local_def were
5332 2021-11-08 Andrew MacLeod <amacleod@redhat.com>
5334 PR tree-optimization/103122
5335 * gimple-range.cc (gimple_ranger::range_of_expr): Request the cache
5336 entry with "calulate new values" set to false.
5338 2021-11-08 Jan Hubicka <hubicka@ucw.cz>
5340 * builtins.c (is_simple_builtin): Add builitin_dwarf_cfa
5341 and builtin_return_address.
5342 (builtin_fnspec): Annotate builtin_return,
5343 bulitin_eh_pointer, builtin_eh_filter, builtin_unwind_resume,
5344 builtin_cxa_end_cleanup, builtin_eh_copy_values,
5345 builtin_frame_address, builtin_apply_args,
5346 builtin_asan_before_dynamic_init, builtin_asan_after_dynamic_init,
5347 builtin_prefetch, builtin_dwarf_cfa, builtin_return_addrss
5349 * ipa-pure-const.c (special_builtin_state): Add builtin_dwarf_cfa
5350 and builtin_return_address.
5352 2021-11-08 Jan Hubicka <hubicka@ucw.cz>
5354 PR tree-optimization/103177
5355 * passes.def: Move uncprop after pure/const and modref.
5357 2021-11-08 Martin Jambor <mjambor@suse.cz>
5361 * tree-inline.c (remap_gimple_stmt): Unshare the expression without
5362 location before invoking remap_with_debug_expressions on it.
5363 * ipa-param-manipulation.c
5364 (ipa_param_body_adjustments::prepare_debug_expressions): Likewise.
5366 2021-11-08 David Edelsohn <dje.gcc@gmail.com>
5368 * config/rs6000/vsx.md (vsx_splat_v4si_di): Revert "wa"
5371 2021-11-08 Richard Biener <rguenther@suse.de>
5373 * cfgloop.c (verify_loop_structure): Use a temporary BB flag
5374 instead of an sbitmap to cache irreducible state.
5376 2021-11-08 Aldy Hernandez <aldyh@redhat.com>
5378 PR tree-optimization/103120
5379 * gimple-range-path.cc (path_range_query::range_defined_in_block):
5380 Bail if there's a cache entry.
5382 2021-11-08 Bill Schmidt <wschmidt@linux.ibm.com>
5384 * config/rs6000/rs6000.c (rs6000_builtin_reciprocal): Use
5385 rs6000_builtin_decls_x when appropriate.
5386 (add_condition_to_bb): Likewise.
5387 (rs6000_atomic_assign_expand_fenv): Likewise.
5389 2021-11-08 Bill Schmidt <wschmidt@linux.ibm.com>
5391 * config/rs6000/rs6000-call.c (rs6000_new_builtin_decl): New function.
5392 (rs6000_builtin_decl): Call it.
5394 2021-11-08 Martin Liska <mliska@suse.cz>
5396 * cgraph.c (cgraph_node::dump): Dump it from decl.
5398 2021-11-08 Martin Liska <mliska@suse.cz>
5400 PR gcov-profile/100520
5401 * coverage.c (coverage_compute_profile_id): Strip .gk when
5402 compare debug is used.
5403 * system.h (endswith): New function.
5405 2021-11-08 Martin Liska <mliska@suse.cz>
5407 * cgraph.c (cgraph_node::dump): Dump static_chain_decl.
5409 2021-11-08 Thomas Schwinge <thomas@codesourcery.com>
5411 * config/rs6000/rbtree.c: Fix 'Copyright (C) 2020-21' into '2020-2021'
5412 * config/rs6000/rbtree.h: Likewise.
5413 * config/rs6000/rs6000-builtin-new.def: Likewise.
5414 * config/rs6000/rs6000-gen-builtins.c: Likewise.
5415 * config/rs6000/rs6000-overload.def: Likewise.
5417 2021-11-08 Richard Sandiford <richard.sandiford@arm.com>
5419 * config/aarch64/aarch64.c (aarch64_ldn_stn_vectors): New function.
5420 (aarch64_address_cost): Use it instead of testing for CImode and
5423 2021-11-08 Richard Sandiford <richard.sandiford@arm.com>
5425 * genmodes.c (emit_insn_modes_h): Define NUM_MODE_* macros.
5426 * expmed.h (NUM_MODE_INT): Delete in favor of genmodes definitions.
5427 (NUM_MODE_PARTIAL_INT, NUM_MODE_VECTOR_INT): Likewise.
5428 * real.h (real_format_for_mode): Use NUM_MODE_FLOAT and
5429 NUM_MODE_DECIMAL_FLOAT.
5430 (REAL_MODE_FORMAT): Likewise.
5432 2021-11-08 Andre Vieira <andre.simoesdiasvieira@arm.com>
5434 * tree-vect-loop.c (vect_better_loop_vinfo_p): Change how epilogue loop
5437 2021-11-08 Richard Biener <rguenther@suse.de>
5439 * tree-vectorizer.h (vect_create_loop_vinfo): Add main_loop_info
5441 * tree-vect-loop.c (vect_create_loop_vinfo): Likewise. Set
5442 LOOP_VINFO_ORIG_LOOP_INFO and conditionalize set of
5443 LOOP_VINFO_NITERS_ASSUMPTIONS.
5444 (vect_analyze_loop_1): Adjust.
5445 (vect_analyze_loop): Move loop constraint setting and
5446 SCEV/niter reset here from vect_create_loop_vinfo to perform
5448 (vect_analyze_loop_form): Move dumping of symbolic niters
5449 here from vect_create_loop_vinfo.
5451 2021-11-08 Jan Hubicka <hubicka@ucw.cz>
5453 * ipa-modref-tree.h (enum modref_special_parms): New enum.
5454 (struct modref_access_node): update for special parms.
5455 (struct modref_ref_node): Likewise.
5456 (struct modref_parm_map): Likewise.
5457 (struct modref_tree): Likewise.
5458 * ipa-modref.c (dump_access): Likewise.
5459 (get_access): Detect static chain.
5460 (parm_map_for_arg): Take tree as arg instead of
5462 (merge_call_side_effects): Compute map for static chain.
5463 (process_fnspec): Update.
5464 (struct escape_point): Remove retslot_arg and static_chain_arg.
5465 (analyze_parms): Update.
5466 (compute_parm_map): Update.
5467 (propagate_unknown_call): Update.
5468 (modref_propagate_in_scc): Update.
5469 (modref_merge_call_site_flags): Update.
5470 (ipa_merge_modref_summary_after_inlining): Update.
5471 * tree-ssa-alias.c (modref_may_conflict): Handle static chain.
5472 * ipa-modref-tree.c (test_merge): Update.
5474 2021-11-08 Haochen Gui <guihaoc@gcc.gnu.org>
5476 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_builtin): Disable
5477 gimple fold for VSX_BUILTIN_XVMINDP, ALTIVEC_BUILTIN_VMINFP,
5478 VSX_BUILTIN_XVMAXDP, ALTIVEC_BUILTIN_VMAXFP when fast-math is not
5481 2021-11-08 liuhongt <hongtao.liu@intel.com>
5483 PR tree-optimization/103077
5484 * doc/invoke.texi (Options That Control Optimization):
5485 Update documentation for -ftree-loop-vectorize and
5486 -ftree-slp-vectorize which are enabled by default at -02.
5488 2021-11-08 liuhongt <hongtao.liu@intel.com>
5491 * match.pd (Simplifcation (trunc)copysign((extend)a, (extend)b)
5492 to .COPYSIGN (a, b)): Add !HONOR_SNANS.
5494 2021-11-08 liuhongt <hongtao.liu@intel.com>
5497 * match.pd: Simplify
5498 (trunc)fma ((extend)a, (extend)b, (extend)c) to IFN_FMA (a, b,
5499 c) under flag_unsafe_math_optimizations.
5501 2021-11-07 Jan Hubicka <hubicka@ucw.cz>
5504 * params.opt (modref-max-depth): Add range.
5505 (modref-max-adjustments): Fix range.
5507 2021-11-07 Aldy Hernandez <aldyh@redhat.com>
5509 * tree-pass.h (make_pass_vrp_threader): Remove.
5510 * tree-ssa-threadbackward.c
5511 (back_threader_profitability::profitable_path_p): Remove
5512 ASSERT_EXPR references.
5513 * tree-ssa-threadedge.c (jt_state::register_equivs_stmt): Same.
5514 * tree-vrp.c (vrp_folder::simplify_casted_conds): Same.
5515 (execute_vrp): Same.
5516 (class hybrid_threader): Remove.
5517 (hybrid_threader::hybrid_threader): Remove.
5518 (hybrid_threader::~hybrid_threader): Remove.
5519 (hybrid_threader::before_dom_children): Remove.
5520 (hybrid_threader::after_dom_children): Remove.
5521 (execute_vrp_threader): Remove.
5522 (class pass_vrp_threader): Remove.
5523 (make_pass_vrp_threader): Remove.
5525 2021-11-07 Jan Hubicka <hubicka@ucw.cz>
5527 * gimple.c (gimple_call_arg_flags): Use interposable_eaf_flags.
5528 (gimple_call_retslot_flags): Likewise.
5529 (gimple_call_static_chain_flags): Likewise.
5530 * ipa-modref.c (remove_useless_eaf_flags): Do not remove everything for
5532 (modref_summary::useful_p): Likewise.
5533 (modref_summary_lto::useful_p): Likewise.
5534 (analyze_parms): Do not give up on NOVOPS.
5535 (analyze_function): When dumping report chnages in EAF flags
5536 between IPA and local pass.
5537 (modref_merge_call_site_flags): Compute implicit eaf flags
5538 based on callee ecf_flags and fnspec; if the function does not
5539 bind to current defs use interposable_eaf_flags.
5540 (modref_propagate_flags_in_scc): Update.
5541 * ipa-modref.h (interposable_eaf_flags): New function.
5543 2021-11-07 Bill Schmidt <wschmidt@linux.ibm.com>
5545 * config/rs6000/rs6000-call.c (rs6000_expand_new_builtin): New
5547 (rs6000_invalid_new_builtin): New function.
5548 (rs6000_expand_builtin): Call rs6000_expand_new_builtin.
5549 (rs6000_expand_ldst_mask): New function.
5550 (new_cpu_expand_builtin): Likewise.
5551 (elemrev_icode): Likewise.
5552 (ldv_expand_builtin): Likewise.
5553 (lxvrse_expand_builtin): Likewise.
5554 (lxvrze_expand_builtin): Likewise.
5555 (stv_expand_builtin): Likewise.
5556 (new_mma_expand_builtin): Likewise.
5557 (new_htm_spr_num): Likewise.
5558 (new_htm_expand_builtin): Likewise.
5559 (rs6000_expand_new_builtin): Likewise.
5560 (rs6000_init_builtins): Initialize altivec_builtin_mask_for_load.
5562 2021-11-07 Jan Hubicka <hubicka@ucw.cz>
5564 * ipa-modref.c (modref_lattice): Add do_dataflow,
5565 changed and propagate_to fields.
5566 (modref_lattice::release): Free propagate_to
5567 (modref_lattice::merge): Do not give up early on unknown
5569 (modref_lattice::merge_deref): Likewise.
5570 (modref_eaf_analysis): Update toplevel comment.
5571 (modref_eaf_analysis::analyze_ssa_name): Record postponned ssa names;
5572 do optimistic dataflow initialization.
5573 (modref_eaf_analysis::merge_with_ssa_name): Build dataflow graph.
5574 (modref_eaf_analysis::propagate): New member function.
5575 (analyze_parms): Update to new API of modref_eaf_analysis.
5577 2021-11-06 Jan Hubicka <hubicka@ucw.cz>
5579 * cgraph.h (cgraph_node::can_be_discarded_p): Do not
5580 return true on functions from other partition.
5582 2021-11-06 David Edelsohn <dje.gcc@gmail.com>
5584 * config/rs6000/vsx.md (vsx_splat_v4si): Change constraints to "wa".
5585 (vsx_splat_v4si_di): Change constraint to "wa".
5587 2021-11-06 Aldy Hernandez <aldyh@redhat.com>
5589 PR tree-optimization/103061
5590 * value-relation.cc (path_oracle::path_oracle): Initialize
5592 (path_oracle::killing_def): Set m_killed_defs.
5593 (path_oracle::query_relation): Do not look at the root oracle for
5595 * value-relation.h (class path_oracle): Add m_killed_defs.
5597 2021-11-06 Aldy Hernandez <aldyh@redhat.com>
5599 * tree-ssa-threadbackward.c (back_threader::find_paths_to_names):
5600 Remove gotos and other cleanups.
5602 2021-11-05 Jan Hubicka <hubicka@ucw.cz>
5605 * ipa-modref-tree.h (modref_tree::insert): Do nothing for
5606 paradoxical and zero sized accesses.
5608 2021-11-05 Jan Hubicka <hubicka@ucw.cz>
5611 * ipa-modref-tree.h (struct modref_access_node): Avoid left shift
5614 2021-11-05 Iain Sandoe <iain@sandoe.co.uk>
5616 * config/darwin.h (ASM_GENERATE_INTERNAL_LABEL): Add LTRAMP
5617 to the list of symbol prefixes that must be made linker-
5620 2021-11-05 Iain Sandoe <iain@sandoe.co.uk>
5622 * config.host: Add support for aarch64-*-darwin.
5623 * config/aarch64/host-aarch64-darwin.c: New file.
5624 * config/aarch64/x-darwin: New file.
5626 2021-11-05 Andrew MacLeod <amacleod@redhat.com>
5628 PR tree-optimization/103093
5629 * gimple-range-gori.cc (range_def_chain::get_imports): Remove assert.
5631 2021-11-05 Andrew MacLeod <amacleod@redhat.com>
5633 PR tree-optimization/102943
5634 * gimple-range-cache.cc (class update_list): New.
5635 (update_list::add): Replace add_to_update.
5636 (update_list::pop): New.
5637 (ranger_cache::ranger_cache): Adjust.
5638 (ranger_cache::~ranger_cache): Adjust.
5639 (ranger_cache::add_to_update): Delete.
5640 (ranger_cache::propagate_cache): Adjust to new class.
5641 (ranger_cache::propagate_updated_value): Ditto.
5642 (ranger_cache::fill_block_cache): Ditto.
5643 * gimple-range-cache.h (class ranger_cache): Adjust to update class.
5645 2021-11-05 Richard Biener <rguenther@suse.de>
5647 * tree-vect-loop.c (vect_analyze_loop): Remove obsolete
5648 comment and expand on another one. Combine nested if.
5650 2021-11-05 John David Anglin <danglin@gcc.gnu.org>
5653 * config/pa/pa-modes.def: Add OImode integer type.
5654 * config/pa/pa.c (pa_scalar_mode_supported_p): Allow TImode
5656 * config/pa/pa.h (MIN_UNITS_PER_WORD) Define to MIN_UNITS_PER_WORD
5657 to UNITS_PER_WORD if IN_LIBGCC2.
5658 * config/pa/pa.md (addti3, addvti3, subti3, subvti3, negti2,
5659 negvti2, ashlti3, shrpd_internal): New patterns.
5660 Change some multi instruction types to multi.
5662 2021-11-05 Jakub Jelinek <jakub@redhat.com>
5665 * config/i386/i386.h
5666 (stringop_algs::stringop_strategy::stringop_strategy): Make the ctor
5669 2021-11-05 Wilco Dijkstra <wdijkstr@arm.com>
5672 * config/aarch64/aarch64.c (aarch64_mov_operand_p): Strip the salt
5674 * config/aarch64/constraints.md: Support const in Usw.
5676 2021-11-05 John David Anglin <danglin@gcc.gnu.org>
5678 * config/pa/pa.h (PREFERRED_DEBUGGING_TYPE): Define to DWARF2_DEBUG.
5679 * config/pa/pa64-hpux.h (PREFERRED_DEBUGGING_TYPE): Remove define.
5681 2021-11-05 Richard Biener <rguenther@suse.de>
5683 * tree-vectorizer.h (vec_info_shared::n_stmts): Add.
5684 (LOOP_VINFO_N_STMTS): Likewise.
5685 (vec_info_for_bb): Remove unused function.
5686 * tree-vectorizer.c (vec_info_shared::vec_info_shared):
5687 Initialize n_stmts member.
5688 * tree-vect-loop.c: Remove INCLUDE_FUNCTIONAL.
5689 (vect_create_loop_vinfo): Do not set loop->aux.
5690 (vect_analyze_loop_2): Do not get n_stmts as argument,
5691 instead use LOOP_VINFO_N_STMTS. Set LOOP_VINFO_VECTORIZABLE_P
5693 (vect_analyze_loop_1): Remove callback, get the mode iterator
5694 and autodetected_vector_mode as argument, advancing the
5695 iterator and initializing autodetected_vector_mode here.
5696 (vect_analyze_loop): Split analysis loop into two, first
5697 processing main loops only and then epilogues.
5699 2021-11-05 Martin Jambor <mjambor@suse.cz>
5701 * ipa-prop.c (compute_complex_assign_jump_func): Remove
5702 unnecessary check for RECORD_TYPE.
5704 2021-11-05 Gerald Pfeifer <gerald@pfeifer.com>
5706 * doc/install.texi2html: Do not generate old.html any longer.
5708 2021-11-05 Martin Liska <mliska@suse.cz>
5711 * opts.c (finish_options): Reset flag_gtoggle when it is used.
5713 2021-11-05 Jakub Jelinek <jakub@redhat.com>
5716 * dwarf2out.c (add_const_value_attribute): Add MODE argument, use it
5717 in CONST_WIDE_INT handling. Adjust recursive calls.
5718 (add_location_or_const_value_attribute): Pass DECL_MODE (decl) to
5719 new add_const_value_attribute argument.
5720 (tree_add_const_value_attribute): Pass TYPE_MODE (type) to new
5721 add_const_value_attribute argument.
5723 2021-11-05 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
5725 * config/vx-common.h: Test value of TARGET_VXWORKS7 rather
5728 2021-11-05 Richard Biener <rguenther@suse.de>
5730 * tree-vectorizer.h (struct vect_loop_form_info): New.
5731 (vect_analyze_loop_form): Adjust.
5732 (vect_create_loop_vinfo): New.
5733 * tree-parloops.c (gather_scalar_reductions): Adjust for
5734 vect_analyze_loop_form API change.
5735 * tree-vect-loop.c: Include <functional>.
5736 (vect_analyze_loop_form_1): Rename to vect_analyze_loop_form,
5737 take struct vect_loop_form_info as output parameter and adjust.
5738 (vect_analyze_loop_form): Rename to vect_create_loop_vinfo and
5739 split out call to the original vect_analyze_loop_form_1.
5740 (vect_reanalyze_as_main_loop): Rename to...
5741 (vect_analyze_loop_1): ... this, factor out the call to
5742 vect_analyze_loop_form and generalize to be able to use it twice ...
5743 (vect_analyze_loop): ... here. Perform vect_analyze_loop_form
5746 2021-11-05 Xionghu Luo <luoxhu@linux.ibm.com>
5749 * config/rs6000/fusion.md: Regenerate.
5750 * config/rs6000/genfusion.pl: Fix incorrect clobber constraint.
5752 2021-11-04 Andreas Krebbel <krebbel@linux.ibm.com>
5754 * config/s390/s390.h (STACK_CHECK_MOVING_SP): New macro
5757 2021-11-04 Tamar Christina <tamar.christina@arm.com>
5759 * config/aarch64/aarch64-builtins.c
5760 (aarch64_general_gimple_fold_builtin): Add ashl, sshl, ushl, ashr,
5761 ashr_simd, lshr, lshr_simd.
5762 * config/aarch64/aarch64-simd-builtins.def (lshr): Use USHIFTIMM.
5763 * config/aarch64/arm_neon.h (vshr_n_u8, vshr_n_u16, vshr_n_u32,
5764 vshrq_n_u8, vshrq_n_u16, vshrq_n_u32, vshrq_n_u64): Fix type hack.
5766 2021-11-04 Tamar Christina <tamar.christina@arm.com>
5768 * match.pd: New negate+shift pattern.
5770 2021-11-04 Andrew MacLeod <amacleod@redhat.com>
5772 PR tree-optimization/103079
5773 * gimple-range-gori.cc (gimple_range_calc_op1): Treat undefined as
5775 (gimple_range_calc_op2): Ditto.
5777 2021-11-04 Martin Jambor <mjambor@suse.cz>
5780 * ipa-param-manipulation.h (class ipa_param_body_adjustments): New
5781 members remap_with_debug_expressions, m_dead_ssa_debug_equiv,
5782 m_dead_stmt_debug_equiv and prepare_debug_expressions. Added
5783 parameter to mark_dead_statements.
5784 * ipa-param-manipulation.c: Include tree-phinodes.h and cfgexpand.h.
5785 (ipa_param_body_adjustments::mark_dead_statements): New parameter
5786 debugstack, push into it all SSA names used in debug statements,
5787 produce m_dead_ssa_debug_equiv mapping for the removed param.
5788 (replace_with_mapped_expr): New function.
5789 (ipa_param_body_adjustments::remap_with_debug_expressions): Likewise.
5790 (ipa_param_body_adjustments::prepare_debug_expressions): Likewise.
5791 (ipa_param_body_adjustments::common_initialization): Gather and
5792 procecc SSA which will be removed but are in debug statements. Simplify.
5793 (ipa_param_body_adjustments::ipa_param_body_adjustments): Initialize
5795 * tree-inline.c (remap_gimple_stmt): Create a debug bind when possible
5796 when avoiding a copy of an unnecessary statement. Remap removed SSA
5797 names in existing debug statements.
5798 (tree_function_versioning): Do not create DEBUG_EXPR_DECL for removed
5799 parameters if we have already done so.
5801 2021-11-04 Jan Hubicka <hubicka@ucw.cz>
5804 * gimple.c (gimple_call_static_chain_flags): Handle case when
5805 nested function does not bind locally.
5807 2021-11-04 Jonathan Wright <jonathan.wright@arm.com>
5809 * config/aarch64/aarch64.c (aarch64_function_value): Generate
5810 a register rtx for Neon vector-tuple modes.
5811 (aarch64_layout_arg): Likewise.
5813 2021-11-04 Jonathan Wright <jonathan.wright@arm.com>
5815 * lower-subreg.c (simple_move): Prevent decomposition if
5816 modes are not tieable.
5818 2021-11-04 Jonathan Wright <jonathan.wright@arm.com>
5819 Richard Sandiford <richard.sandiford@arm.com>
5821 * config/aarch64/aarch64-builtins.c (v2x8qi_UP): Define.
5822 (v2x4hi_UP): Likewise.
5823 (v2x4hf_UP): Likewise.
5824 (v2x4bf_UP): Likewise.
5825 (v2x2si_UP): Likewise.
5826 (v2x2sf_UP): Likewise.
5827 (v2x1di_UP): Likewise.
5828 (v2x1df_UP): Likewise.
5829 (v2x16qi_UP): Likewise.
5830 (v2x8hi_UP): Likewise.
5831 (v2x8hf_UP): Likewise.
5832 (v2x8bf_UP): Likewise.
5833 (v2x4si_UP): Likewise.
5834 (v2x4sf_UP): Likewise.
5835 (v2x2di_UP): Likewise.
5836 (v2x2df_UP): Likewise.
5837 (v3x8qi_UP): Likewise.
5838 (v3x4hi_UP): Likewise.
5839 (v3x4hf_UP): Likewise.
5840 (v3x4bf_UP): Likewise.
5841 (v3x2si_UP): Likewise.
5842 (v3x2sf_UP): Likewise.
5843 (v3x1di_UP): Likewise.
5844 (v3x1df_UP): Likewise.
5845 (v3x16qi_UP): Likewise.
5846 (v3x8hi_UP): Likewise.
5847 (v3x8hf_UP): Likewise.
5848 (v3x8bf_UP): Likewise.
5849 (v3x4si_UP): Likewise.
5850 (v3x4sf_UP): Likewise.
5851 (v3x2di_UP): Likewise.
5852 (v3x2df_UP): Likewise.
5853 (v4x8qi_UP): Likewise.
5854 (v4x4hi_UP): Likewise.
5855 (v4x4hf_UP): Likewise.
5856 (v4x4bf_UP): Likewise.
5857 (v4x2si_UP): Likewise.
5858 (v4x2sf_UP): Likewise.
5859 (v4x1di_UP): Likewise.
5860 (v4x1df_UP): Likewise.
5861 (v4x16qi_UP): Likewise.
5862 (v4x8hi_UP): Likewise.
5863 (v4x8hf_UP): Likewise.
5864 (v4x8bf_UP): Likewise.
5865 (v4x4si_UP): Likewise.
5866 (v4x4sf_UP): Likewise.
5867 (v4x2di_UP): Likewise.
5868 (v4x2df_UP): Likewise.
5869 (TYPES_GETREGP): Delete.
5870 (TYPES_SETREGP): Likewise.
5871 (TYPES_LOADSTRUCT_U): Define.
5872 (TYPES_LOADSTRUCT_P): Likewise.
5873 (TYPES_LOADSTRUCT_LANE_U): Likewise.
5874 (TYPES_LOADSTRUCT_LANE_P): Likewise.
5875 (TYPES_STORE1P): Move for consistency.
5876 (TYPES_STORESTRUCT_U): Define.
5877 (TYPES_STORESTRUCT_P): Likewise.
5878 (TYPES_STORESTRUCT_LANE_U): Likewise.
5879 (TYPES_STORESTRUCT_LANE_P): Likewise.
5880 (aarch64_simd_tuple_types): Define.
5881 (aarch64_lookup_simd_builtin_type): Handle tuple type lookup.
5882 (aarch64_init_simd_builtin_functions): Update frontend lookup
5883 for builtin functions after handling arm_neon.h pragma.
5884 (register_tuple_type): Manually set modes of single-integer
5885 tuple types. Record tuple types.
5886 * config/aarch64/aarch64-modes.def
5887 (ADV_SIMD_D_REG_STRUCT_MODES): Define D-register tuple modes.
5888 (ADV_SIMD_Q_REG_STRUCT_MODES): Define Q-register tuple modes.
5889 (SVE_MODES): Give single-vector modes priority over vector-
5891 (VECTOR_MODES_WITH_PREFIX): Set partial-vector mode order to
5892 be after all single-vector modes.
5893 * config/aarch64/aarch64-simd-builtins.def: Update builtin
5894 generator macros to reflect modifications to the backend
5896 * config/aarch64/aarch64-simd.md (aarch64_simd_ld2<mode>):
5897 Use vector-tuple mode iterator and rename to...
5898 (aarch64_simd_ld2<vstruct_elt>): This.
5899 (aarch64_simd_ld2r<mode>): Use vector-tuple mode iterator and
5901 (aarch64_simd_ld2r<vstruct_elt>): This.
5902 (aarch64_vec_load_lanesoi_lane<mode>): Use vector-tuple mode
5903 iterator and rename to...
5904 (aarch64_vec_load_lanes<mode>_lane<vstruct_elt>): This.
5905 (vec_load_lanesoi<mode>): Use vector-tuple mode iterator and
5907 (vec_load_lanes<mode><vstruct_elt>): This.
5908 (aarch64_simd_st2<mode>): Use vector-tuple mode iterator and
5910 (aarch64_simd_st2<vstruct_elt>): This.
5911 (aarch64_vec_store_lanesoi_lane<mode>): Use vector-tuple mode
5912 iterator and rename to...
5913 (aarch64_vec_store_lanes<mode>_lane<vstruct_elt>): This.
5914 (vec_store_lanesoi<mode>): Use vector-tuple mode iterator and
5916 (vec_store_lanes<mode><vstruct_elt>): This.
5917 (aarch64_simd_ld3<mode>): Use vector-tuple mode iterator and
5919 (aarch64_simd_ld3<vstruct_elt>): This.
5920 (aarch64_simd_ld3r<mode>): Use vector-tuple mode iterator and
5922 (aarch64_simd_ld3r<vstruct_elt>): This.
5923 (aarch64_vec_load_lanesci_lane<mode>): Use vector-tuple mode
5924 iterator and rename to...
5925 (vec_load_lanesci<mode>): This.
5926 (aarch64_simd_st3<mode>): Use vector-tuple mode iterator and
5928 (aarch64_simd_st3<vstruct_elt>): This.
5929 (aarch64_vec_store_lanesci_lane<mode>): Use vector-tuple mode
5930 iterator and rename to...
5931 (vec_store_lanesci<mode>): This.
5932 (aarch64_simd_ld4<mode>): Use vector-tuple mode iterator and
5934 (aarch64_simd_ld4<vstruct_elt>): This.
5935 (aarch64_simd_ld4r<mode>): Use vector-tuple mode iterator and
5937 (aarch64_simd_ld4r<vstruct_elt>): This.
5938 (aarch64_vec_load_lanesxi_lane<mode>): Use vector-tuple mode
5939 iterator and rename to...
5940 (vec_load_lanesxi<mode>): This.
5941 (aarch64_simd_st4<mode>): Use vector-tuple mode iterator and
5943 (aarch64_simd_st4<vstruct_elt>): This.
5944 (aarch64_vec_store_lanesxi_lane<mode>): Use vector-tuple mode
5945 iterator and rename to...
5946 (vec_store_lanesxi<mode>): This.
5947 (mov<mode>): Define for Neon vector-tuple modes.
5948 (aarch64_ld1x3<VALLDIF:mode>): Use vector-tuple mode iterator
5950 (aarch64_ld1x3<vstruct_elt>): This.
5951 (aarch64_ld1_x3_<mode>): Use vector-tuple mode iterator and
5953 (aarch64_ld1_x3_<vstruct_elt>): This.
5954 (aarch64_ld1x4<VALLDIF:mode>): Use vector-tuple mode iterator
5956 (aarch64_ld1x4<vstruct_elt>): This.
5957 (aarch64_ld1_x4_<mode>): Use vector-tuple mode iterator and
5959 (aarch64_ld1_x4_<vstruct_elt>): This.
5960 (aarch64_st1x2<VALLDIF:mode>): Use vector-tuple mode iterator
5962 (aarch64_st1x2<vstruct_elt>): This.
5963 (aarch64_st1_x2_<mode>): Use vector-tuple mode iterator and
5965 (aarch64_st1_x2_<vstruct_elt>): This.
5966 (aarch64_st1x3<VALLDIF:mode>): Use vector-tuple mode iterator
5968 (aarch64_st1x3<vstruct_elt>): This.
5969 (aarch64_st1_x3_<mode>): Use vector-tuple mode iterator and
5971 (aarch64_st1_x3_<vstruct_elt>): This.
5972 (aarch64_st1x4<VALLDIF:mode>): Use vector-tuple mode iterator
5974 (aarch64_st1x4<vstruct_elt>): This.
5975 (aarch64_st1_x4_<mode>): Use vector-tuple mode iterator and
5977 (aarch64_st1_x4_<vstruct_elt>): This.
5978 (*aarch64_mov<mode>): Define for vector-tuple modes.
5979 (*aarch64_be_mov<mode>): Likewise.
5980 (aarch64_ld<VSTRUCT:nregs>r<VALLDIF:mode>): Use vector-tuple
5981 mode iterator and rename to...
5982 (aarch64_ld<nregs>r<vstruct_elt>): This.
5983 (aarch64_ld2<mode>_dreg): Use vector-tuple mode iterator and
5985 (aarch64_ld2<vstruct_elt>_dreg): This.
5986 (aarch64_ld3<mode>_dreg): Use vector-tuple mode iterator and
5988 (aarch64_ld3<vstruct_elt>_dreg): This.
5989 (aarch64_ld4<mode>_dreg): Use vector-tuple mode iterator and
5991 (aarch64_ld4<vstruct_elt>_dreg): This.
5992 (aarch64_ld<VSTRUCT:nregs><VDC:mode>): Use vector-tuple mode
5993 iterator and rename to...
5994 (aarch64_ld<nregs><vstruct_elt>): Use vector-tuple mode
5995 iterator and rename to...
5996 (aarch64_ld<VSTRUCT:nregs><VQ:mode>): Use vector-tuple mode
5997 (aarch64_ld1x2<VQ:mode>): Delete.
5998 (aarch64_ld1x2<VDC:mode>): Use vector-tuple mode iterator and
6000 (aarch64_ld1x2<vstruct_elt>): This.
6001 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use vector-
6002 tuple mode iterator and rename to...
6003 (aarch64_ld<nregs>_lane<vstruct_elt>): This.
6004 (aarch64_get_dreg<VSTRUCT:mode><VDC:mode>): Delete.
6005 (aarch64_get_qreg<VSTRUCT:mode><VQ:mode>): Likewise.
6006 (aarch64_st2<mode>_dreg): Use vector-tuple mode iterator and
6008 (aarch64_st2<vstruct_elt>_dreg): This.
6009 (aarch64_st3<mode>_dreg): Use vector-tuple mode iterator and
6011 (aarch64_st3<vstruct_elt>_dreg): This.
6012 (aarch64_st4<mode>_dreg): Use vector-tuple mode iterator and
6014 (aarch64_st4<vstruct_elt>_dreg): This.
6015 (aarch64_st<VSTRUCT:nregs><VDC:mode>): Use vector-tuple mode
6016 iterator and rename to...
6017 (aarch64_st<nregs><vstruct_elt>): This.
6018 (aarch64_st<VSTRUCT:nregs><VQ:mode>): Use vector-tuple mode
6019 iterator and rename to aarch64_st<nregs><vstruct_elt>.
6020 (aarch64_st<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use vector-
6021 tuple mode iterator and rename to...
6022 (aarch64_st<nregs>_lane<vstruct_elt>): This.
6023 (aarch64_set_qreg<VSTRUCT:mode><VQ:mode>): Delete.
6024 (aarch64_simd_ld1<mode>_x2): Use vector-tuple mode iterator
6026 (aarch64_simd_ld1<vstruct_elt>_x2): This.
6027 * config/aarch64/aarch64.c (aarch64_advsimd_struct_mode_p):
6028 Refactor to include new vector-tuple modes.
6029 (aarch64_classify_vector_mode): Add cases for new vector-
6031 (aarch64_advsimd_partial_struct_mode_p): Define.
6032 (aarch64_advsimd_full_struct_mode_p): Likewise.
6033 (aarch64_advsimd_vector_array_mode): Likewise.
6034 (aarch64_sve_data_mode): Change location in file.
6035 (aarch64_array_mode): Handle case of Neon vector-tuple modes.
6036 (aarch64_hard_regno_nregs): Handle case of partial Neon
6038 (aarch64_classify_address): Refactor to include handling of
6039 Neon vector-tuple modes.
6040 (aarch64_print_operand): Print "d" for "%R" for a partial
6041 Neon vector structure.
6042 (aarch64_expand_vec_perm_1): Use new vector-tuple mode.
6043 (aarch64_modes_tieable_p): Prevent tieing Neon partial struct
6044 modes with scalar machines modes larger than 8 bytes.
6045 (aarch64_can_change_mode_class): Don't allow changes between
6046 partial and full Neon vector-structure modes.
6047 * config/aarch64/arm_neon.h (vst2_lane_f16): Use updated
6048 builtin and remove boiler-plate code for opaque mode.
6049 (vst2_lane_f32): Likewise.
6050 (vst2_lane_f64): Likewise.
6051 (vst2_lane_p8): Likewise.
6052 (vst2_lane_p16): Likewise.
6053 (vst2_lane_p64): Likewise.
6054 (vst2_lane_s8): Likewise.
6055 (vst2_lane_s16): Likewise.
6056 (vst2_lane_s32): Likewise.
6057 (vst2_lane_s64): Likewise.
6058 (vst2_lane_u8): Likewise.
6059 (vst2_lane_u16): Likewise.
6060 (vst2_lane_u32): Likewise.
6061 (vst2_lane_u64): Likewise.
6062 (vst2q_lane_f16): Likewise.
6063 (vst2q_lane_f32): Likewise.
6064 (vst2q_lane_f64): Likewise.
6065 (vst2q_lane_p8): Likewise.
6066 (vst2q_lane_p16): Likewise.
6067 (vst2q_lane_p64): Likewise.
6068 (vst2q_lane_s8): Likewise.
6069 (vst2q_lane_s16): Likewise.
6070 (vst2q_lane_s32): Likewise.
6071 (vst2q_lane_s64): Likewise.
6072 (vst2q_lane_u8): Likewise.
6073 (vst2q_lane_u16): Likewise.
6074 (vst2q_lane_u32): Likewise.
6075 (vst2q_lane_u64): Likewise.
6076 (vst3_lane_f16): Likewise.
6077 (vst3_lane_f32): Likewise.
6078 (vst3_lane_f64): Likewise.
6079 (vst3_lane_p8): Likewise.
6080 (vst3_lane_p16): Likewise.
6081 (vst3_lane_p64): Likewise.
6082 (vst3_lane_s8): Likewise.
6083 (vst3_lane_s16): Likewise.
6084 (vst3_lane_s32): Likewise.
6085 (vst3_lane_s64): Likewise.
6086 (vst3_lane_u8): Likewise.
6087 (vst3_lane_u16): Likewise.
6088 (vst3_lane_u32): Likewise.
6089 (vst3_lane_u64): Likewise.
6090 (vst3q_lane_f16): Likewise.
6091 (vst3q_lane_f32): Likewise.
6092 (vst3q_lane_f64): Likewise.
6093 (vst3q_lane_p8): Likewise.
6094 (vst3q_lane_p16): Likewise.
6095 (vst3q_lane_p64): Likewise.
6096 (vst3q_lane_s8): Likewise.
6097 (vst3q_lane_s16): Likewise.
6098 (vst3q_lane_s32): Likewise.
6099 (vst3q_lane_s64): Likewise.
6100 (vst3q_lane_u8): Likewise.
6101 (vst3q_lane_u16): Likewise.
6102 (vst3q_lane_u32): Likewise.
6103 (vst3q_lane_u64): Likewise.
6104 (vst4_lane_f16): Likewise.
6105 (vst4_lane_f32): Likewise.
6106 (vst4_lane_f64): Likewise.
6107 (vst4_lane_p8): Likewise.
6108 (vst4_lane_p16): Likewise.
6109 (vst4_lane_p64): Likewise.
6110 (vst4_lane_s8): Likewise.
6111 (vst4_lane_s16): Likewise.
6112 (vst4_lane_s32): Likewise.
6113 (vst4_lane_s64): Likewise.
6114 (vst4_lane_u8): Likewise.
6115 (vst4_lane_u16): Likewise.
6116 (vst4_lane_u32): Likewise.
6117 (vst4_lane_u64): Likewise.
6118 (vst4q_lane_f16): Likewise.
6119 (vst4q_lane_f32): Likewise.
6120 (vst4q_lane_f64): Likewise.
6121 (vst4q_lane_p8): Likewise.
6122 (vst4q_lane_p16): Likewise.
6123 (vst4q_lane_p64): Likewise.
6124 (vst4q_lane_s8): Likewise.
6125 (vst4q_lane_s16): Likewise.
6126 (vst4q_lane_s32): Likewise.
6127 (vst4q_lane_s64): Likewise.
6128 (vst4q_lane_u8): Likewise.
6129 (vst4q_lane_u16): Likewise.
6130 (vst4q_lane_u32): Likewise.
6131 (vst4q_lane_u64): Likewise.
6132 (vtbl3_s8): Likewise.
6133 (vtbl3_u8): Likewise.
6134 (vtbl3_p8): Likewise.
6135 (vtbl4_s8): Likewise.
6136 (vtbl4_u8): Likewise.
6137 (vtbl4_p8): Likewise.
6138 (vld1_u8_x3): Likewise.
6139 (vld1_s8_x3): Likewise.
6140 (vld1_u16_x3): Likewise.
6141 (vld1_s16_x3): Likewise.
6142 (vld1_u32_x3): Likewise.
6143 (vld1_s32_x3): Likewise.
6144 (vld1_u64_x3): Likewise.
6145 (vld1_s64_x3): Likewise.
6146 (vld1_f16_x3): Likewise.
6147 (vld1_f32_x3): Likewise.
6148 (vld1_f64_x3): Likewise.
6149 (vld1_p8_x3): Likewise.
6150 (vld1_p16_x3): Likewise.
6151 (vld1_p64_x3): Likewise.
6152 (vld1q_u8_x3): Likewise.
6153 (vld1q_s8_x3): Likewise.
6154 (vld1q_u16_x3): Likewise.
6155 (vld1q_s16_x3): Likewise.
6156 (vld1q_u32_x3): Likewise.
6157 (vld1q_s32_x3): Likewise.
6158 (vld1q_u64_x3): Likewise.
6159 (vld1q_s64_x3): Likewise.
6160 (vld1q_f16_x3): Likewise.
6161 (vld1q_f32_x3): Likewise.
6162 (vld1q_f64_x3): Likewise.
6163 (vld1q_p8_x3): Likewise.
6164 (vld1q_p16_x3): Likewise.
6165 (vld1q_p64_x3): Likewise.
6166 (vld1_u8_x2): Likewise.
6167 (vld1_s8_x2): Likewise.
6168 (vld1_u16_x2): Likewise.
6169 (vld1_s16_x2): Likewise.
6170 (vld1_u32_x2): Likewise.
6171 (vld1_s32_x2): Likewise.
6172 (vld1_u64_x2): Likewise.
6173 (vld1_s64_x2): Likewise.
6174 (vld1_f16_x2): Likewise.
6175 (vld1_f32_x2): Likewise.
6176 (vld1_f64_x2): Likewise.
6177 (vld1_p8_x2): Likewise.
6178 (vld1_p16_x2): Likewise.
6179 (vld1_p64_x2): Likewise.
6180 (vld1q_u8_x2): Likewise.
6181 (vld1q_s8_x2): Likewise.
6182 (vld1q_u16_x2): Likewise.
6183 (vld1q_s16_x2): Likewise.
6184 (vld1q_u32_x2): Likewise.
6185 (vld1q_s32_x2): Likewise.
6186 (vld1q_u64_x2): Likewise.
6187 (vld1q_s64_x2): Likewise.
6188 (vld1q_f16_x2): Likewise.
6189 (vld1q_f32_x2): Likewise.
6190 (vld1q_f64_x2): Likewise.
6191 (vld1q_p8_x2): Likewise.
6192 (vld1q_p16_x2): Likewise.
6193 (vld1q_p64_x2): Likewise.
6194 (vld1_s8_x4): Likewise.
6195 (vld1q_s8_x4): Likewise.
6196 (vld1_s16_x4): Likewise.
6197 (vld1q_s16_x4): Likewise.
6198 (vld1_s32_x4): Likewise.
6199 (vld1q_s32_x4): Likewise.
6200 (vld1_u8_x4): Likewise.
6201 (vld1q_u8_x4): Likewise.
6202 (vld1_u16_x4): Likewise.
6203 (vld1q_u16_x4): Likewise.
6204 (vld1_u32_x4): Likewise.
6205 (vld1q_u32_x4): Likewise.
6206 (vld1_f16_x4): Likewise.
6207 (vld1q_f16_x4): Likewise.
6208 (vld1_f32_x4): Likewise.
6209 (vld1q_f32_x4): Likewise.
6210 (vld1_p8_x4): Likewise.
6211 (vld1q_p8_x4): Likewise.
6212 (vld1_p16_x4): Likewise.
6213 (vld1q_p16_x4): Likewise.
6214 (vld1_s64_x4): Likewise.
6215 (vld1_u64_x4): Likewise.
6216 (vld1_p64_x4): Likewise.
6217 (vld1q_s64_x4): Likewise.
6218 (vld1q_u64_x4): Likewise.
6219 (vld1q_p64_x4): Likewise.
6220 (vld1_f64_x4): Likewise.
6221 (vld1q_f64_x4): Likewise.
6222 (vld2_s64): Likewise.
6223 (vld2_u64): Likewise.
6224 (vld2_f64): Likewise.
6225 (vld2_s8): Likewise.
6226 (vld2_p8): Likewise.
6227 (vld2_p64): Likewise.
6228 (vld2_s16): Likewise.
6229 (vld2_p16): Likewise.
6230 (vld2_s32): Likewise.
6231 (vld2_u8): Likewise.
6232 (vld2_u16): Likewise.
6233 (vld2_u32): Likewise.
6234 (vld2_f16): Likewise.
6235 (vld2_f32): Likewise.
6236 (vld2q_s8): Likewise.
6237 (vld2q_p8): Likewise.
6238 (vld2q_s16): Likewise.
6239 (vld2q_p16): Likewise.
6240 (vld2q_p64): Likewise.
6241 (vld2q_s32): Likewise.
6242 (vld2q_s64): Likewise.
6243 (vld2q_u8): Likewise.
6244 (vld2q_u16): Likewise.
6245 (vld2q_u32): Likewise.
6246 (vld2q_u64): Likewise.
6247 (vld2q_f16): Likewise.
6248 (vld2q_f32): Likewise.
6249 (vld2q_f64): Likewise.
6250 (vld3_s64): Likewise.
6251 (vld3_u64): Likewise.
6252 (vld3_f64): Likewise.
6253 (vld3_s8): Likewise.
6254 (vld3_p8): Likewise.
6255 (vld3_s16): Likewise.
6256 (vld3_p16): Likewise.
6257 (vld3_s32): Likewise.
6258 (vld3_u8): Likewise.
6259 (vld3_u16): Likewise.
6260 (vld3_u32): Likewise.
6261 (vld3_f16): Likewise.
6262 (vld3_f32): Likewise.
6263 (vld3_p64): Likewise.
6264 (vld3q_s8): Likewise.
6265 (vld3q_p8): Likewise.
6266 (vld3q_s16): Likewise.
6267 (vld3q_p16): Likewise.
6268 (vld3q_s32): Likewise.
6269 (vld3q_s64): Likewise.
6270 (vld3q_u8): Likewise.
6271 (vld3q_u16): Likewise.
6272 (vld3q_u32): Likewise.
6273 (vld3q_u64): Likewise.
6274 (vld3q_f16): Likewise.
6275 (vld3q_f32): Likewise.
6276 (vld3q_f64): Likewise.
6277 (vld3q_p64): Likewise.
6278 (vld4_s64): Likewise.
6279 (vld4_u64): Likewise.
6280 (vld4_f64): Likewise.
6281 (vld4_s8): Likewise.
6282 (vld4_p8): Likewise.
6283 (vld4_s16): Likewise.
6284 (vld4_p16): Likewise.
6285 (vld4_s32): Likewise.
6286 (vld4_u8): Likewise.
6287 (vld4_u16): Likewise.
6288 (vld4_u32): Likewise.
6289 (vld4_f16): Likewise.
6290 (vld4_f32): Likewise.
6291 (vld4_p64): Likewise.
6292 (vld4q_s8): Likewise.
6293 (vld4q_p8): Likewise.
6294 (vld4q_s16): Likewise.
6295 (vld4q_p16): Likewise.
6296 (vld4q_s32): Likewise.
6297 (vld4q_s64): Likewise.
6298 (vld4q_u8): Likewise.
6299 (vld4q_u16): Likewise.
6300 (vld4q_u32): Likewise.
6301 (vld4q_u64): Likewise.
6302 (vld4q_f16): Likewise.
6303 (vld4q_f32): Likewise.
6304 (vld4q_f64): Likewise.
6305 (vld4q_p64): Likewise.
6306 (vld2_dup_s8): Likewise.
6307 (vld2_dup_s16): Likewise.
6308 (vld2_dup_s32): Likewise.
6309 (vld2_dup_f16): Likewise.
6310 (vld2_dup_f32): Likewise.
6311 (vld2_dup_f64): Likewise.
6312 (vld2_dup_u8): Likewise.
6313 (vld2_dup_u16): Likewise.
6314 (vld2_dup_u32): Likewise.
6315 (vld2_dup_p8): Likewise.
6316 (vld2_dup_p16): Likewise.
6317 (vld2_dup_p64): Likewise.
6318 (vld2_dup_s64): Likewise.
6319 (vld2_dup_u64): Likewise.
6320 (vld2q_dup_s8): Likewise.
6321 (vld2q_dup_p8): Likewise.
6322 (vld2q_dup_s16): Likewise.
6323 (vld2q_dup_p16): Likewise.
6324 (vld2q_dup_s32): Likewise.
6325 (vld2q_dup_s64): Likewise.
6326 (vld2q_dup_u8): Likewise.
6327 (vld2q_dup_u16): Likewise.
6328 (vld2q_dup_u32): Likewise.
6329 (vld2q_dup_u64): Likewise.
6330 (vld2q_dup_f16): Likewise.
6331 (vld2q_dup_f32): Likewise.
6332 (vld2q_dup_f64): Likewise.
6333 (vld2q_dup_p64): Likewise.
6334 (vld3_dup_s64): Likewise.
6335 (vld3_dup_u64): Likewise.
6336 (vld3_dup_f64): Likewise.
6337 (vld3_dup_s8): Likewise.
6338 (vld3_dup_p8): Likewise.
6339 (vld3_dup_s16): Likewise.
6340 (vld3_dup_p16): Likewise.
6341 (vld3_dup_s32): Likewise.
6342 (vld3_dup_u8): Likewise.
6343 (vld3_dup_u16): Likewise.
6344 (vld3_dup_u32): Likewise.
6345 (vld3_dup_f16): Likewise.
6346 (vld3_dup_f32): Likewise.
6347 (vld3_dup_p64): Likewise.
6348 (vld3q_dup_s8): Likewise.
6349 (vld3q_dup_p8): Likewise.
6350 (vld3q_dup_s16): Likewise.
6351 (vld3q_dup_p16): Likewise.
6352 (vld3q_dup_s32): Likewise.
6353 (vld3q_dup_s64): Likewise.
6354 (vld3q_dup_u8): Likewise.
6355 (vld3q_dup_u16): Likewise.
6356 (vld3q_dup_u32): Likewise.
6357 (vld3q_dup_u64): Likewise.
6358 (vld3q_dup_f16): Likewise.
6359 (vld3q_dup_f32): Likewise.
6360 (vld3q_dup_f64): Likewise.
6361 (vld3q_dup_p64): Likewise.
6362 (vld4_dup_s64): Likewise.
6363 (vld4_dup_u64): Likewise.
6364 (vld4_dup_f64): Likewise.
6365 (vld4_dup_s8): Likewise.
6366 (vld4_dup_p8): Likewise.
6367 (vld4_dup_s16): Likewise.
6368 (vld4_dup_p16): Likewise.
6369 (vld4_dup_s32): Likewise.
6370 (vld4_dup_u8): Likewise.
6371 (vld4_dup_u16): Likewise.
6372 (vld4_dup_u32): Likewise.
6373 (vld4_dup_f16): Likewise.
6374 (vld4_dup_f32): Likewise.
6375 (vld4_dup_p64): Likewise.
6376 (vld4q_dup_s8): Likewise.
6377 (vld4q_dup_p8): Likewise.
6378 (vld4q_dup_s16): Likewise.
6379 (vld4q_dup_p16): Likewise.
6380 (vld4q_dup_s32): Likewise.
6381 (vld4q_dup_s64): Likewise.
6382 (vld4q_dup_u8): Likewise.
6383 (vld4q_dup_u16): Likewise.
6384 (vld4q_dup_u32): Likewise.
6385 (vld4q_dup_u64): Likewise.
6386 (vld4q_dup_f16): Likewise.
6387 (vld4q_dup_f32): Likewise.
6388 (vld4q_dup_f64): Likewise.
6389 (vld4q_dup_p64): Likewise.
6390 (vld2_lane_u8): Likewise.
6391 (vld2_lane_u16): Likewise.
6392 (vld2_lane_u32): Likewise.
6393 (vld2_lane_u64): Likewise.
6394 (vld2_lane_s8): Likewise.
6395 (vld2_lane_s16): Likewise.
6396 (vld2_lane_s32): Likewise.
6397 (vld2_lane_s64): Likewise.
6398 (vld2_lane_f16): Likewise.
6399 (vld2_lane_f32): Likewise.
6400 (vld2_lane_f64): Likewise.
6401 (vld2_lane_p8): Likewise.
6402 (vld2_lane_p16): Likewise.
6403 (vld2_lane_p64): Likewise.
6404 (vld2q_lane_u8): Likewise.
6405 (vld2q_lane_u16): Likewise.
6406 (vld2q_lane_u32): Likewise.
6407 (vld2q_lane_u64): Likewise.
6408 (vld2q_lane_s8): Likewise.
6409 (vld2q_lane_s16): Likewise.
6410 (vld2q_lane_s32): Likewise.
6411 (vld2q_lane_s64): Likewise.
6412 (vld2q_lane_f16): Likewise.
6413 (vld2q_lane_f32): Likewise.
6414 (vld2q_lane_f64): Likewise.
6415 (vld2q_lane_p8): Likewise.
6416 (vld2q_lane_p16): Likewise.
6417 (vld2q_lane_p64): Likewise.
6418 (vld3_lane_u8): Likewise.
6419 (vld3_lane_u16): Likewise.
6420 (vld3_lane_u32): Likewise.
6421 (vld3_lane_u64): Likewise.
6422 (vld3_lane_s8): Likewise.
6423 (vld3_lane_s16): Likewise.
6424 (vld3_lane_s32): Likewise.
6425 (vld3_lane_s64): Likewise.
6426 (vld3_lane_f16): Likewise.
6427 (vld3_lane_f32): Likewise.
6428 (vld3_lane_f64): Likewise.
6429 (vld3_lane_p8): Likewise.
6430 (vld3_lane_p16): Likewise.
6431 (vld3_lane_p64): Likewise.
6432 (vld3q_lane_u8): Likewise.
6433 (vld3q_lane_u16): Likewise.
6434 (vld3q_lane_u32): Likewise.
6435 (vld3q_lane_u64): Likewise.
6436 (vld3q_lane_s8): Likewise.
6437 (vld3q_lane_s16): Likewise.
6438 (vld3q_lane_s32): Likewise.
6439 (vld3q_lane_s64): Likewise.
6440 (vld3q_lane_f16): Likewise.
6441 (vld3q_lane_f32): Likewise.
6442 (vld3q_lane_f64): Likewise.
6443 (vld3q_lane_p8): Likewise.
6444 (vld3q_lane_p16): Likewise.
6445 (vld3q_lane_p64): Likewise.
6446 (vld4_lane_u8): Likewise.
6447 (vld4_lane_u16): Likewise.
6448 (vld4_lane_u32): Likewise.
6449 (vld4_lane_u64): Likewise.
6450 (vld4_lane_s8): Likewise.
6451 (vld4_lane_s16): Likewise.
6452 (vld4_lane_s32): Likewise.
6453 (vld4_lane_s64): Likewise.
6454 (vld4_lane_f16): Likewise.
6455 (vld4_lane_f32): Likewise.
6456 (vld4_lane_f64): Likewise.
6457 (vld4_lane_p8): Likewise.
6458 (vld4_lane_p16): Likewise.
6459 (vld4_lane_p64): Likewise.
6460 (vld4q_lane_u8): Likewise.
6461 (vld4q_lane_u16): Likewise.
6462 (vld4q_lane_u32): Likewise.
6463 (vld4q_lane_u64): Likewise.
6464 (vld4q_lane_s8): Likewise.
6465 (vld4q_lane_s16): Likewise.
6466 (vld4q_lane_s32): Likewise.
6467 (vld4q_lane_s64): Likewise.
6468 (vld4q_lane_f16): Likewise.
6469 (vld4q_lane_f32): Likewise.
6470 (vld4q_lane_f64): Likewise.
6471 (vld4q_lane_p8): Likewise.
6472 (vld4q_lane_p16): Likewise.
6473 (vld4q_lane_p64): Likewise.
6474 (vqtbl2_s8): Likewise.
6475 (vqtbl2_u8): Likewise.
6476 (vqtbl2_p8): Likewise.
6477 (vqtbl2q_s8): Likewise.
6478 (vqtbl2q_u8): Likewise.
6479 (vqtbl2q_p8): Likewise.
6480 (vqtbl3_s8): Likewise.
6481 (vqtbl3_u8): Likewise.
6482 (vqtbl3_p8): Likewise.
6483 (vqtbl3q_s8): Likewise.
6484 (vqtbl3q_u8): Likewise.
6485 (vqtbl3q_p8): Likewise.
6486 (vqtbl4_s8): Likewise.
6487 (vqtbl4_u8): Likewise.
6488 (vqtbl4_p8): Likewise.
6489 (vqtbl4q_s8): Likewise.
6490 (vqtbl4q_u8): Likewise.
6491 (vqtbl4q_p8): Likewise.
6492 (vqtbx2_s8): Likewise.
6493 (vqtbx2_u8): Likewise.
6494 (vqtbx2_p8): Likewise.
6495 (vqtbx2q_s8): Likewise.
6496 (vqtbx2q_u8): Likewise.
6497 (vqtbx2q_p8): Likewise.
6498 (vqtbx3_s8): Likewise.
6499 (vqtbx3_u8): Likewise.
6500 (vqtbx3_p8): Likewise.
6501 (vqtbx3q_s8): Likewise.
6502 (vqtbx3q_u8): Likewise.
6503 (vqtbx3q_p8): Likewise.
6504 (vqtbx4_s8): Likewise.
6505 (vqtbx4_u8): Likewise.
6506 (vqtbx4_p8): Likewise.
6507 (vqtbx4q_s8): Likewise.
6508 (vqtbx4q_u8): Likewise.
6509 (vqtbx4q_p8): Likewise.
6510 (vst1_s64_x2): Likewise.
6511 (vst1_u64_x2): Likewise.
6512 (vst1_f64_x2): Likewise.
6513 (vst1_s8_x2): Likewise.
6514 (vst1_p8_x2): Likewise.
6515 (vst1_s16_x2): Likewise.
6516 (vst1_p16_x2): Likewise.
6517 (vst1_s32_x2): Likewise.
6518 (vst1_u8_x2): Likewise.
6519 (vst1_u16_x2): Likewise.
6520 (vst1_u32_x2): Likewise.
6521 (vst1_f16_x2): Likewise.
6522 (vst1_f32_x2): Likewise.
6523 (vst1_p64_x2): Likewise.
6524 (vst1q_s8_x2): Likewise.
6525 (vst1q_p8_x2): Likewise.
6526 (vst1q_s16_x2): Likewise.
6527 (vst1q_p16_x2): Likewise.
6528 (vst1q_s32_x2): Likewise.
6529 (vst1q_s64_x2): Likewise.
6530 (vst1q_u8_x2): Likewise.
6531 (vst1q_u16_x2): Likewise.
6532 (vst1q_u32_x2): Likewise.
6533 (vst1q_u64_x2): Likewise.
6534 (vst1q_f16_x2): Likewise.
6535 (vst1q_f32_x2): Likewise.
6536 (vst1q_f64_x2): Likewise.
6537 (vst1q_p64_x2): Likewise.
6538 (vst1_s64_x3): Likewise.
6539 (vst1_u64_x3): Likewise.
6540 (vst1_f64_x3): Likewise.
6541 (vst1_s8_x3): Likewise.
6542 (vst1_p8_x3): Likewise.
6543 (vst1_s16_x3): Likewise.
6544 (vst1_p16_x3): Likewise.
6545 (vst1_s32_x3): Likewise.
6546 (vst1_u8_x3): Likewise.
6547 (vst1_u16_x3): Likewise.
6548 (vst1_u32_x3): Likewise.
6549 (vst1_f16_x3): Likewise.
6550 (vst1_f32_x3): Likewise.
6551 (vst1_p64_x3): Likewise.
6552 (vst1q_s8_x3): Likewise.
6553 (vst1q_p8_x3): Likewise.
6554 (vst1q_s16_x3): Likewise.
6555 (vst1q_p16_x3): Likewise.
6556 (vst1q_s32_x3): Likewise.
6557 (vst1q_s64_x3): Likewise.
6558 (vst1q_u8_x3): Likewise.
6559 (vst1q_u16_x3): Likewise.
6560 (vst1q_u32_x3): Likewise.
6561 (vst1q_u64_x3): Likewise.
6562 (vst1q_f16_x3): Likewise.
6563 (vst1q_f32_x3): Likewise.
6564 (vst1q_f64_x3): Likewise.
6565 (vst1q_p64_x3): Likewise.
6566 (vst1_s8_x4): Likewise.
6567 (vst1q_s8_x4): Likewise.
6568 (vst1_s16_x4): Likewise.
6569 (vst1q_s16_x4): Likewise.
6570 (vst1_s32_x4): Likewise.
6571 (vst1q_s32_x4): Likewise.
6572 (vst1_u8_x4): Likewise.
6573 (vst1q_u8_x4): Likewise.
6574 (vst1_u16_x4): Likewise.
6575 (vst1q_u16_x4): Likewise.
6576 (vst1_u32_x4): Likewise.
6577 (vst1q_u32_x4): Likewise.
6578 (vst1_f16_x4): Likewise.
6579 (vst1q_f16_x4): Likewise.
6580 (vst1_f32_x4): Likewise.
6581 (vst1q_f32_x4): Likewise.
6582 (vst1_p8_x4): Likewise.
6583 (vst1q_p8_x4): Likewise.
6584 (vst1_p16_x4): Likewise.
6585 (vst1q_p16_x4): Likewise.
6586 (vst1_s64_x4): Likewise.
6587 (vst1_u64_x4): Likewise.
6588 (vst1_p64_x4): Likewise.
6589 (vst1q_s64_x4): Likewise.
6590 (vst1q_u64_x4): Likewise.
6591 (vst1q_p64_x4): Likewise.
6592 (vst1_f64_x4): Likewise.
6593 (vst1q_f64_x4): Likewise.
6594 (vst2_s64): Likewise.
6595 (vst2_u64): Likewise.
6596 (vst2_f64): Likewise.
6597 (vst2_s8): Likewise.
6598 (vst2_p8): Likewise.
6599 (vst2_s16): Likewise.
6600 (vst2_p16): Likewise.
6601 (vst2_s32): Likewise.
6602 (vst2_u8): Likewise.
6603 (vst2_u16): Likewise.
6604 (vst2_u32): Likewise.
6605 (vst2_f16): Likewise.
6606 (vst2_f32): Likewise.
6607 (vst2_p64): Likewise.
6608 (vst2q_s8): Likewise.
6609 (vst2q_p8): Likewise.
6610 (vst2q_s16): Likewise.
6611 (vst2q_p16): Likewise.
6612 (vst2q_s32): Likewise.
6613 (vst2q_s64): Likewise.
6614 (vst2q_u8): Likewise.
6615 (vst2q_u16): Likewise.
6616 (vst2q_u32): Likewise.
6617 (vst2q_u64): Likewise.
6618 (vst2q_f16): Likewise.
6619 (vst2q_f32): Likewise.
6620 (vst2q_f64): Likewise.
6621 (vst2q_p64): Likewise.
6622 (vst3_s64): Likewise.
6623 (vst3_u64): Likewise.
6624 (vst3_f64): Likewise.
6625 (vst3_s8): Likewise.
6626 (vst3_p8): Likewise.
6627 (vst3_s16): Likewise.
6628 (vst3_p16): Likewise.
6629 (vst3_s32): Likewise.
6630 (vst3_u8): Likewise.
6631 (vst3_u16): Likewise.
6632 (vst3_u32): Likewise.
6633 (vst3_f16): Likewise.
6634 (vst3_f32): Likewise.
6635 (vst3_p64): Likewise.
6636 (vst3q_s8): Likewise.
6637 (vst3q_p8): Likewise.
6638 (vst3q_s16): Likewise.
6639 (vst3q_p16): Likewise.
6640 (vst3q_s32): Likewise.
6641 (vst3q_s64): Likewise.
6642 (vst3q_u8): Likewise.
6643 (vst3q_u16): Likewise.
6644 (vst3q_u32): Likewise.
6645 (vst3q_u64): Likewise.
6646 (vst3q_f16): Likewise.
6647 (vst3q_f32): Likewise.
6648 (vst3q_f64): Likewise.
6649 (vst3q_p64): Likewise.
6650 (vst4_s64): Likewise.
6651 (vst4_u64): Likewise.
6652 (vst4_f64): Likewise.
6653 (vst4_s8): Likewise.
6654 (vst4_p8): Likewise.
6655 (vst4_s16): Likewise.
6656 (vst4_p16): Likewise.
6657 (vst4_s32): Likewise.
6658 (vst4_u8): Likewise.
6659 (vst4_u16): Likewise.
6660 (vst4_u32): Likewise.
6661 (vst4_f16): Likewise.
6662 (vst4_f32): Likewise.
6663 (vst4_p64): Likewise.
6664 (vst4q_s8): Likewise.
6665 (vst4q_p8): Likewise.
6666 (vst4q_s16): Likewise.
6667 (vst4q_p16): Likewise.
6668 (vst4q_s32): Likewise.
6669 (vst4q_s64): Likewise.
6670 (vst4q_u8): Likewise.
6671 (vst4q_u16): Likewise.
6672 (vst4q_u32): Likewise.
6673 (vst4q_u64): Likewise.
6674 (vst4q_f16): Likewise.
6675 (vst4q_f32): Likewise.
6676 (vst4q_f64): Likewise.
6677 (vst4q_p64): Likewise.
6678 (vtbx4_s8): Likewise.
6679 (vtbx4_u8): Likewise.
6680 (vtbx4_p8): Likewise.
6681 (vld1_bf16_x2): Likewise.
6682 (vld1q_bf16_x2): Likewise.
6683 (vld1_bf16_x3): Likewise.
6684 (vld1q_bf16_x3): Likewise.
6685 (vld1_bf16_x4): Likewise.
6686 (vld1q_bf16_x4): Likewise.
6687 (vld2_bf16): Likewise.
6688 (vld2q_bf16): Likewise.
6689 (vld2_dup_bf16): Likewise.
6690 (vld2q_dup_bf16): Likewise.
6691 (vld3_bf16): Likewise.
6692 (vld3q_bf16): Likewise.
6693 (vld3_dup_bf16): Likewise.
6694 (vld3q_dup_bf16): Likewise.
6695 (vld4_bf16): Likewise.
6696 (vld4q_bf16): Likewise.
6697 (vld4_dup_bf16): Likewise.
6698 (vld4q_dup_bf16): Likewise.
6699 (vst1_bf16_x2): Likewise.
6700 (vst1q_bf16_x2): Likewise.
6701 (vst1_bf16_x3): Likewise.
6702 (vst1q_bf16_x3): Likewise.
6703 (vst1_bf16_x4): Likewise.
6704 (vst1q_bf16_x4): Likewise.
6705 (vst2_bf16): Likewise.
6706 (vst2q_bf16): Likewise.
6707 (vst3_bf16): Likewise.
6708 (vst3q_bf16): Likewise.
6709 (vst4_bf16): Likewise.
6710 (vst4q_bf16): Likewise.
6711 (vld2_lane_bf16): Likewise.
6712 (vld2q_lane_bf16): Likewise.
6713 (vld3_lane_bf16): Likewise.
6714 (vld3q_lane_bf16): Likewise.
6715 (vld4_lane_bf16): Likewise.
6716 (vld4q_lane_bf16): Likewise.
6717 (vst2_lane_bf16): Likewise.
6718 (vst2q_lane_bf16): Likewise.
6719 (vst3_lane_bf16): Likewise.
6720 (vst3q_lane_bf16): Likewise.
6721 (vst4_lane_bf16): Likewise.
6722 (vst4q_lane_bf16): Likewise.
6723 * config/aarch64/geniterators.sh: Modify iterator regex to
6724 match new vector-tuple modes.
6725 * config/aarch64/iterators.md (insn_count): Extend mode
6726 attribute with vector-tuple type information.
6728 (Vendreg): Likewise.
6731 (VSTRUCT_2D): New mode iterator.
6732 (VSTRUCT_2DNX): Likewise.
6733 (VSTRUCT_2DX): Likewise.
6734 (VSTRUCT_2Q): Likewise.
6735 (VSTRUCT_2QD): Likewise.
6736 (VSTRUCT_3D): Likewise.
6737 (VSTRUCT_3DNX): Likewise.
6738 (VSTRUCT_3DX): Likewise.
6739 (VSTRUCT_3Q): Likewise.
6740 (VSTRUCT_3QD): Likewise.
6741 (VSTRUCT_4D): Likewise.
6742 (VSTRUCT_4DNX): Likewise.
6743 (VSTRUCT_4DX): Likewise.
6744 (VSTRUCT_4Q): Likewise.
6745 (VSTRUCT_4QD): Likewise.
6746 (VSTRUCT_D): Likewise.
6747 (VSTRUCT_Q): Likewise.
6748 (VSTRUCT_QD): Likewise.
6749 (VSTRUCT_ELT): New mode attribute.
6750 (vstruct_elt): Likewise.
6751 * genmodes.c (VECTOR_MODE): Add default prefix and order
6753 (VECTOR_MODE_WITH_PREFIX): Define.
6754 (make_vector_mode): Add mode prefix and order parameters.
6756 2021-11-04 Jonathan Wright <jonathan.wright@arm.com>
6758 * expmed.c (extract_bit_field_1): Ensure modes are tieable.
6760 2021-11-04 Jonathan Wright <jonathan.wright@arm.com>
6762 * expr.c (emit_group_load_1): Remove historic workaround.
6764 2021-11-04 Jonathan Wright <jonathan.wright@arm.com>
6766 * config/aarch64/aarch64-builtins.c (aarch64_init_simd_builtins):
6767 Factor out main loop to...
6768 (aarch64_init_simd_builtin_functions): This new function.
6769 (register_tuple_type): Define.
6770 (aarch64_scalar_builtin_type_p): Define.
6771 (handle_arm_neon_h): Define.
6772 * config/aarch64/aarch64-c.c (aarch64_pragma_aarch64): Handle
6773 pragma for arm_neon.h.
6774 * config/aarch64/aarch64-protos.h (aarch64_advsimd_struct_mode_p):
6776 (handle_arm_neon_h): Likewise.
6777 * config/aarch64/aarch64.c (aarch64_advsimd_struct_mode_p):
6778 Remove static modifier.
6779 * config/aarch64/arm_neon.h (target): Remove Neon vector
6780 structure type definitions.
6782 2021-11-04 Aldy Hernandez <aldyh@redhat.com>
6784 PR tree-optimization/102943
6785 * gimple-range-path.cc (path_range_query::range_on_path_entry):
6786 Prefer range_of_expr unless there are no statements in the BB.
6788 2021-11-04 Aldy Hernandez <aldyh@redhat.com>
6790 PR tree-optimization/102943
6791 * tree-ssa-threadbackward.c (back_threader::find_paths_to_names):
6792 Avoid duplicate calculation of paths.
6794 2021-11-04 Aldy Hernandez <aldyh@redhat.com>
6796 PR tree-optimization/102943
6797 * gimple-range-path.cc (path_range_query::compute_phi_relations):
6798 Only compute relations for SSA names in the import list.
6799 (path_range_query::compute_outgoing_relations): Same.
6800 * gimple-range-path.h (path_range_query::import_p): New.
6802 2021-11-04 Richard Biener <rguenther@suse.de>
6804 PR rtl-optimization/103075
6805 * simplify-rtx.c (exact_int_to_float_conversion_p): Return
6806 false for a VOIDmode operand.
6808 2021-11-04 Richard Sandiford <richard.sandiford@arm.com>
6810 * config/aarch64/aarch64.c (aarch64_vector_costs): Make member
6811 variables private and add "m_" to their names. Remove is_loop.
6812 (aarch64_record_potential_advsimd_unrolling): Replace with...
6813 (aarch64_vector_costs::record_potential_advsimd_unrolling): ...this.
6814 (aarch64_analyze_loop_vinfo): Replace with...
6815 (aarch64_vector_costs::analyze_loop_vinfo): ...this.
6816 Move initialization of (m_)vec_flags to add_stmt_cost.
6817 (aarch64_analyze_bb_vinfo): Delete.
6818 (aarch64_count_ops): Replace with...
6819 (aarch64_vector_costs::count_ops): ...this.
6820 (aarch64_vector_costs::add_stmt_cost): Set m_vec_flags,
6821 using m_costing_for_scalar to test whether we're costing
6822 scalar or vector code.
6823 (aarch64_adjust_body_cost_sve): Replace with...
6824 (aarch64_vector_costs::adjust_body_cost_sve): ...this.
6825 (aarch64_adjust_body_cost): Replace with...
6826 (aarch64_vector_costs::adjust_body_cost): ...this.
6827 (aarch64_vector_costs::finish_cost): Use m_vinfo instead of is_loop.
6829 2021-11-04 Richard Sandiford <richard.sandiford@arm.com>
6831 * target.def (targetm.vectorize.init_cost): Replace with...
6832 (targetm.vectorize.create_costs): ...this.
6833 (targetm.vectorize.add_stmt_cost): Delete.
6834 (targetm.vectorize.finish_cost): Likewise.
6835 (targetm.vectorize.destroy_cost_data): Likewise.
6836 * doc/tm.texi.in (TARGET_VECTORIZE_INIT_COST): Replace with...
6837 (TARGET_VECTORIZE_CREATE_COSTS): ...this.
6838 (TARGET_VECTORIZE_ADD_STMT_COST): Delete.
6839 (TARGET_VECTORIZE_FINISH_COST): Likewise.
6840 (TARGET_VECTORIZE_DESTROY_COST_DATA): Likewise.
6841 * doc/tm.texi: Regenerate.
6842 * tree-vectorizer.h (vec_info::vec_info): Remove target_cost_data
6844 (vec_info::target_cost_data): Change from a void * to a vector_costs *.
6845 (vector_costs): New class.
6846 (init_cost): Take a vec_info and return a vector_costs.
6847 (dump_stmt_cost): Remove data parameter.
6848 (add_stmt_cost): Replace vinfo and data parameters with a vector_costs.
6849 (add_stmt_costs): Likewise.
6850 (finish_cost): Replace data parameter with a vector_costs.
6851 (destroy_cost_data): Delete.
6852 * tree-vectorizer.c (dump_stmt_cost): Remove data argument and
6854 (vec_info::vec_info): Remove the target_cost_data parameter and
6855 initialize the member variable to null instead.
6856 (vec_info::~vec_info): Delete target_cost_data instead of calling
6858 (vector_costs::add_stmt_cost): New function.
6859 (vector_costs::finish_cost): Likewise.
6860 (vector_costs::record_stmt_cost): Likewise.
6861 (vector_costs::adjust_cost_for_freq): Likewise.
6862 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Update
6863 call to vec_info::vec_info.
6864 (vect_compute_single_scalar_iteration_cost): Update after above
6865 changes to costing interface.
6866 (vect_analyze_loop_operations): Likewise.
6867 (vect_estimate_min_profitable_iters): Likewise.
6868 (vect_analyze_loop_2): Initialize LOOP_VINFO_TARGET_COST_DATA
6869 at the start_over point, where it needs to be recreated after
6870 trying without slp. Update retry code accordingly.
6871 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Update call
6872 to vec_info::vec_info.
6873 (vect_slp_analyze_operation): Update after above changes to costing
6875 (vect_bb_vectorization_profitable_p): Likewise.
6876 * targhooks.h (default_init_cost): Replace with...
6877 (default_vectorize_create_costs): ...this.
6878 (default_add_stmt_cost): Delete.
6879 (default_finish_cost, default_destroy_cost_data): Likewise.
6880 * targhooks.c (default_init_cost): Replace with...
6881 (default_vectorize_create_costs): ...this.
6882 (default_add_stmt_cost): Delete, moving logic to vector_costs instead.
6883 (default_finish_cost, default_destroy_cost_data): Delete.
6884 * config/aarch64/aarch64.c (aarch64_vector_costs): Inherit from
6885 vector_costs. Add a constructor.
6886 (aarch64_init_cost): Replace with...
6887 (aarch64_vectorize_create_costs): ...this.
6888 (aarch64_add_stmt_cost): Replace with...
6889 (aarch64_vector_costs::add_stmt_cost): ...this. Use record_stmt_cost
6890 to adjust the cost for inner loops.
6891 (aarch64_finish_cost): Replace with...
6892 (aarch64_vector_costs::finish_cost): ...this.
6893 (aarch64_destroy_cost_data): Delete.
6894 (TARGET_VECTORIZE_INIT_COST): Replace with...
6895 (TARGET_VECTORIZE_CREATE_COSTS): ...this.
6896 (TARGET_VECTORIZE_ADD_STMT_COST): Delete.
6897 (TARGET_VECTORIZE_FINISH_COST): Likewise.
6898 (TARGET_VECTORIZE_DESTROY_COST_DATA): Likewise.
6899 * config/i386/i386.c (ix86_vector_costs): New structure.
6900 (ix86_init_cost): Replace with...
6901 (ix86_vectorize_create_costs): ...this.
6902 (ix86_add_stmt_cost): Replace with...
6903 (ix86_vector_costs::add_stmt_cost): ...this. Use adjust_cost_for_freq
6904 to adjust the cost for inner loops.
6905 (ix86_finish_cost, ix86_destroy_cost_data): Delete.
6906 (TARGET_VECTORIZE_INIT_COST): Replace with...
6907 (TARGET_VECTORIZE_CREATE_COSTS): ...this.
6908 (TARGET_VECTORIZE_ADD_STMT_COST): Delete.
6909 (TARGET_VECTORIZE_FINISH_COST): Likewise.
6910 (TARGET_VECTORIZE_DESTROY_COST_DATA): Likewise.
6911 * config/rs6000/rs6000.c (TARGET_VECTORIZE_INIT_COST): Replace with...
6912 (TARGET_VECTORIZE_CREATE_COSTS): ...this.
6913 (TARGET_VECTORIZE_ADD_STMT_COST): Delete.
6914 (TARGET_VECTORIZE_FINISH_COST): Likewise.
6915 (TARGET_VECTORIZE_DESTROY_COST_DATA): Likewise.
6916 (rs6000_cost_data): Inherit from vector_costs.
6917 Add a constructor. Drop loop_info, cost and costing_for_scalar
6918 in favor of the corresponding vector_costs member variables.
6919 Add "m_" to the names of the remaining member variables and
6921 (rs6000_density_test): Replace with...
6922 (rs6000_cost_data::density_test): ...this.
6923 (rs6000_init_cost): Replace with...
6924 (rs6000_vectorize_create_costs): ...this.
6925 (rs6000_update_target_cost_per_stmt): Replace with...
6926 (rs6000_cost_data::update_target_cost_per_stmt): ...this.
6927 (rs6000_add_stmt_cost): Replace with...
6928 (rs6000_cost_data::add_stmt_cost): ...this. Use adjust_cost_for_freq
6929 to adjust the cost for inner loops.
6930 (rs6000_adjust_vect_cost_per_loop): Replace with...
6931 (rs6000_cost_data::adjust_vect_cost_per_loop): ...this.
6932 (rs6000_finish_cost): Replace with...
6933 (rs6000_cost_data::finish_cost): ...this. Group loop code
6934 into a single if statement and pass the loop_vinfo down to
6936 (rs6000_destroy_cost_data): Delete.
6938 2021-11-04 Aldy Hernandez <aldyh@redhat.com>
6940 PR tree-optimization/103062
6941 PR tree-optimization/103062
6942 * value-pointer-equiv.cc (ssa_equiv_stack::ssa_equiv_stack):
6943 Increase size of allocation by 1.
6944 (ssa_equiv_stack::push_replacement): Grow as needed.
6945 (ssa_equiv_stack::get_replacement): Same.
6946 (pointer_equiv_analyzer::pointer_equiv_analyzer): Same.
6947 (pointer_equiv_analyzer::~pointer_equiv_analyzer): Remove delete.
6948 (pointer_equiv_analyzer::set_global_equiv): Grow as needed.
6949 (pointer_equiv_analyzer::get_equiv): Same.
6950 (pointer_equiv_analyzer::get_equiv_expr): Remove const.
6951 * value-pointer-equiv.h (class pointer_equiv_analyzer): Remove
6952 const markers. Use auto_vec instead of tree *.
6954 2021-11-04 Richard Biener <rguenther@suse.de>
6956 * tree-ssa-sccvn.c (vn_nary_op_insert_into): Remove always
6957 true parameter and inline valueization.
6958 (vn_nary_op_lookup_1): Inline valueization from ...
6959 (vn_nary_op_compute_hash): ... here and remove it here.
6960 * tree-ssa-pre.c (phi_translate_1): Do not valueize
6961 before vn_nary_lookup_pieces.
6962 (get_representative_for): Mark created SSA representatives
6965 2021-11-04 Richard Sandiford <richard.sandiford@arm.com>
6967 * simplify-rtx.c (simplify_context::simplify_gen_vec_select): Assert
6968 that the operand has a vector mode. Use subreg_lowpart_offset
6969 to test whether an index corresponds to the low part.
6971 2021-11-04 Richard Sandiford <richard.sandiford@arm.com>
6973 * read-rtl.c: Remove dead !GENERATOR_FILE block.
6974 * read-rtl-function.c (function_reader::consolidate_singletons):
6975 Generate canonical CONST_VECTORs.
6977 2021-11-04 liuhongt <hongtao.liu@intel.com>
6980 * config/i386/predicates.md (reg_or_notreg_operand): Rename to ..
6981 (regmem_or_bitnot_regmem_operand): .. and extend to handle
6983 * config/i386/sse.md (*<avx512>_vpternlog<mode>_1): Force_reg
6984 the operands which are required to be register_operand.
6985 (*<avx512>_vpternlog<mode>_2): Ditto.
6986 (*<avx512>_vpternlog<mode>_3): Ditto.
6987 (*<avx512>_vternlog<mode>_all): Disallow embeded broadcast for
6988 vector HFmodes since it's not a real AVX512FP16 instruction.
6990 2021-11-04 liuhongt <hongtao.liu@intel.com>
6993 * match.pd: simplify (trunc)copysign((extend)a, (extend)b) to
6994 .COPYSIGN (a,b) when a and b are same type as the truncation
6995 type and has less precision than extend type.
6997 2021-11-04 Richard Biener <rguenther@suse.de>
6999 * doc/generic.texi: Update TARGET_MEM_REF and MEM_REF
7002 2021-11-04 Hongyu Wang <hongyu.wang@intel.com>
7004 * config/i386/sse.md (VI2_AVX512VNNIBW): New mode iterator.
7005 (VI1_AVX512VNNI): Likewise.
7006 (SDOT_VPDP_SUF): New mode_attr.
7009 (sdot_prod<mode>): Use VI2_AVX512F iterator, expand to
7010 vpdpwssd when VNNI targets available.
7011 (usdot_prod<mode>): New expander for vector QImode.
7013 2021-11-04 Hongyu Wang <hongyu.wang@intel.com>
7015 * config/i386/amxtileintrin.h (_tile_loadd_internal): Add
7016 parentheses to base and stride.
7017 (_tile_stream_loadd_internal): Likewise.
7018 (_tile_stored_internal): Likewise.
7020 2021-11-03 Maciej W. Rozycki <macro@embecosm.com>
7022 * config/riscv/riscv.c (riscv_class_max_nregs): Swap the
7023 arguments to `reg_class_subset_p'.
7025 2021-11-03 Joseph Myers <joseph@codesourcery.com>
7028 * fold-const.c (fold_init): New function.
7029 * fold-const.h (fold_init): New prototype.
7031 2021-11-03 Andrew MacLeod <amacleod@redhat.com>
7033 * params.opt (param_vrp2_mode): Make ranger the default for VRP2.
7035 2021-11-03 Andrew MacLeod <amacleod@redhat.com>
7037 * gimple-range-fold.cc (fold_using_range::range_of_phi): Don't import
7038 a range from edge if arg == phidef.
7040 2021-11-03 Andrew MacLeod <amacleod@redhat.com>
7042 * gimple-range-fold.cc (fold_using_range::range_of_builtin_call): Test
7043 for constant before any other processing.
7045 2021-11-03 Andrew MacLeod <amacleod@redhat.com>
7047 * flag-types.h (RANGER_DEBUG_ALL): Fix values.
7049 2021-11-03 Andrew MacLeod <amacleod@redhat.com>
7051 * gimple-range.cc (gimple_ranger::gimple_ranger): Initialize current_bb.
7052 (gimple_ranger::range_of_expr): Pick up range_on_entry when there is
7053 no explcit context and current_bb is set.
7054 (gimple_ranger::fold_stmt): New.
7055 * gimple-range.h (current_bb, fold_stmt): New.
7056 * tree-vrp.c (rvrp_folder::fold_stmt): Call ranger's fold_stmt.
7058 2021-11-03 Richard Biener <rguenther@suse.de>
7060 PR tree-optimization/102970
7061 * tree-ssa-pre.c (phi_translate_1): Drop clique and base
7062 when translating a MEM_REF over a backedge.
7064 2021-11-03 Philipp Tomsich <philipp.tomsich@vrull.eu>
7066 * config/aarch64/aarch64-cores.def (AARCH64_CORE): New Ampere-1 core.
7067 * config/aarch64/aarch64-tune.md: Regenerate.
7068 * config/aarch64/aarch64-cost-tables.h: Add extra costs for Ampere-1.
7069 * config/aarch64/aarch64.c: Add tuning structures for Ampere-1.
7070 * doc/invoke.texi: Add documentation for Ampere-1 core.
7072 2021-11-03 Wilco Dijkstra <wdijkstr@arm.com>
7074 * config/aarch64/aarch64.md (movsi): Add alternative for GOT accesses.
7076 (ldr_got_small_<mode>): Remove pattern.
7077 (ldr_got_small_sidi): Likewise.
7078 * config/aarch64/aarch64.c (aarch64_load_symref_appropriately): Keep
7079 GOT accesses as moves.
7080 (aarch64_print_operand): Correctly print got_lo12 in L specifier.
7081 (aarch64_mov_operand_p): Make GOT accesses valid move operands.
7082 * config/aarch64/constraints.md: Add new constraint Usw for GOT access.
7084 2021-11-03 Martin Liska <mliska@suse.cz>
7086 * gcov.c (read_line): Remove dead variable.
7088 2021-11-03 Martin Liska <mliska@suse.cz>
7091 * ipa-fnsummary.c (edge_predicate_pool): Rename predicate class to ipa_predicate.
7092 (ipa_fn_summary::account_size_time): Likewise.
7093 (edge_set_predicate): Likewise.
7094 (set_hint_predicate): Likewise.
7095 (add_freqcounting_predicate): Likewise.
7096 (evaluate_conditions_for_known_args): Likewise.
7097 (evaluate_properties_for_edge): Likewise.
7098 (remap_freqcounting_preds_after_dup): Likewise.
7099 (ipa_fn_summary_t::duplicate): Likewise.
7100 (set_cond_stmt_execution_predicate): Likewise.
7101 (set_switch_stmt_execution_predicate): Likewise.
7102 (compute_bb_predicates): Likewise.
7103 (will_be_nonconstant_expr_predicate): Likewise.
7104 (will_be_nonconstant_predicate): Likewise.
7105 (phi_result_unknown_predicate): Likewise.
7106 (predicate_for_phi_result): Likewise.
7107 (analyze_function_body): Likewise.
7108 (compute_fn_summary): Likewise.
7109 (summarize_calls_size_and_time): Likewise.
7110 (estimate_calls_size_and_time): Likewise.
7111 (ipa_call_context::estimate_size_and_time): Likewise.
7112 (remap_edge_summaries): Likewise.
7113 (remap_freqcounting_predicate): Likewise.
7114 (ipa_merge_fn_summary_after_inlining): Likewise.
7115 (ipa_update_overall_fn_summary): Likewise.
7116 (read_ipa_call_summary): Likewise.
7117 (inline_read_section): Likewise.
7118 * ipa-fnsummary.h (struct ipa_freqcounting_predicate): Likewise.
7119 * ipa-predicate.c (predicate::add_clause): Likewise.
7120 (ipa_predicate::add_clause): Likewise.
7121 (predicate::or_with): Likewise.
7122 (ipa_predicate::or_with): Likewise.
7123 (predicate::evaluate): Likewise.
7124 (ipa_predicate::evaluate): Likewise.
7125 (predicate::probability): Likewise.
7126 (ipa_predicate::probability): Likewise.
7127 (dump_condition): Likewise.
7128 (dump_clause): Likewise.
7129 (predicate::dump): Likewise.
7130 (ipa_predicate::dump): Likewise.
7131 (predicate::debug): Likewise.
7132 (ipa_predicate::debug): Likewise.
7133 (predicate::remap_after_duplication): Likewise.
7134 (ipa_predicate::remap_after_duplication): Likewise.
7135 (predicate::remap_after_inlining): Likewise.
7136 (ipa_predicate::remap_after_inlining): Likewise.
7137 (predicate::stream_in): Likewise.
7138 (ipa_predicate::stream_in): Likewise.
7139 (predicate::stream_out): Likewise.
7140 (ipa_predicate::stream_out): Likewise.
7141 (add_condition): Likewise.
7142 * ipa-predicate.h (class predicate): Likewise.
7143 (class ipa_predicate): Likewise.
7144 (add_condition): Likewise.
7146 2021-11-03 Richard Biener <rguenther@suse.de>
7148 * bitmap.h (bitmap_bit_p): Change the return type to bool.
7149 * bitmap.c (bitmap_bit_p): Likewise.
7150 * sbitmap.h (bitmap_bit_p): Likewise.
7151 (bitmap_set_bit): Return whether the bit changed.
7152 (bitmap_clear_bit): Likewise.
7153 * tree-ssa.c (verify_vssa): Make use of the changed state
7154 from bitmap_set_bit.
7156 2021-11-03 Richard Biener <rguenther@suse.de>
7158 PR middle-end/103033
7159 * internal-fn.c (expand_DEFERRED_INIT): Elide the
7160 native_interpret_expr path in favor of folding the
7161 VIEW_CONVERT_EXPR generated when punning the RHS.
7163 2021-11-03 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
7165 * config/s390/s390.c (s390_loop_unroll_adjust): In case of early
7168 2021-11-03 Jan Hubicka <hubicka@ucw.cz>
7171 * ipa-modref.c (callee_to_caller_flags): New function.
7172 (modref_eaf_analysis::analyze_ssa_name): Use it.
7173 (ipa_merge_modref_summary_after_inlining): Fix whitespace.
7175 2021-11-02 Jan Hubicka <hubicka@ucw.cz>
7177 * ipa-modref.c (modref_eaf_analysis::analyze_ssa_name): Revert
7180 2021-11-02 Roger Sayle <roger@nextmovesoftware.com>
7181 Uroš Bizjak <ubizjak@gmail.com>
7183 * config/i386/i386.md (<any_rotate>ti3): Provide expansion for
7184 rotations by non-constant amounts.
7186 2021-11-02 Jan Hubicka <hubicka@ucw.cz>
7188 * ipa-modref.c: Fix anonymous namespace placement.
7189 (class modref_eaf_analysis): New class.
7190 (analyze_ssa_name_flags): Turn to ...
7191 (modref_eaf_analysis::analyze_ssa_name): ... this one.
7192 (merge_call_lhs_flags): Turn to ...
7193 (modref_eaf_analysis::merge_call_lhs_flags): .. this one
7194 (modref_eaf_analysis::merge_with_ssa_name): New member function.
7195 (record_escape_points): Turn to ...
7196 (modref_eaf_analysis::record_escape_points): ... this one.
7197 (analyze_parms): Updat
7198 (ipa_merge_modref_summary_after_inlining): Move to the end of file.
7200 2021-11-02 Jan Hubicka <hubicka@ucw.cz>
7202 * gimple.c (gimple_call_static_chain_flags): New function.
7203 * gimple.h (gimple_call_static_chain_flags): Declare
7204 * ipa-modref.c (modref_summary::modref_summary): Initialize
7206 (modref_summary_lto::modref_summary_lto): Likewise.
7207 (modref_summary::useful_p): Test static_chain_flags.
7208 (modref_summary_lto::useful_p): Likewise.
7209 (struct modref_summary_lto): Add static_chain_flags.
7210 (modref_summary::dump): Dump static_chain_flags.
7211 (modref_summary_lto::dump): Likewise.
7212 (struct escape_point): Add static_cahin_arg.
7213 (analyze_ssa_name_flags): Use gimple_call_static_chain_flags.
7214 (analyze_parms): Handle static chains.
7215 (modref_summaries::duplicate): Duplicate static_chain_flags.
7216 (modref_summaries_lto::duplicate): Likewise.
7217 (modref_write): Stream static_chain_flags.
7218 (read_section): Likewise.
7219 (modref_merge_call_site_flags): Handle static_chain_flags.
7220 * ipa-modref.h (struct modref_summary): Add static_chain_flags.
7221 * tree-ssa-structalias.c (handle_rhs_call): Use
7222 gimple_static_chain_flags.
7224 2021-11-02 Richard Biener <rguenther@suse.de>
7226 PR tree-optimization/103029
7227 * tree-vect-loop-manip.c (vect_loop_versioning): Ensure
7228 the PHI nodes in the loop maintain their original operand
7231 2021-11-02 Jan Hubicka <hubicka@ucw.cz>
7233 * tree-core.h (EAF_NOT_RETURNED_DIRECTLY): New flag.
7234 (EAF_NOREAD): Renumber.
7235 * ipa-modref.c (dump_eaf_flags): Dump EAF_NOT_RETURNED_DIRECTLY.
7236 (remove_useless_eaf_flags): Handle EAF_NOT_RETURNED_DIRECTLY
7237 (deref_flags): Likewise.
7238 (modref_lattice::init): Likewise.
7239 (modref_lattice::merge): Likewise.
7240 (merge_call_lhs_flags): Likewise.
7241 (analyze_ssa_name_flags): Likewise.
7242 (modref_merge_call_site_flags): Likewise.
7243 * tree-ssa-structalias.c (handle_call_arg): Likewise.
7245 2021-11-02 Maciej W. Rozycki <macro@embecosm.com>
7247 * config/riscv/riscv.c (riscv_rtx_costs): Correct a CONST_INT_P
7248 check and remove an unused local variable with shNadd/shNadd.uw
7251 2021-11-02 Tamar Christina <tamar.christina@arm.com>
7253 PR tree-optimization/103007
7254 * tree-vect-slp-patterns.c (complex_fms_pattern::matches): Add elem
7257 2021-11-02 Richard Biener <rguenther@suse.de>
7259 PR middle-end/103038
7260 * fold-const.c (native_interpret_expr): Handle OFFSET_TYPE.
7261 (can_native_interpret_type_p): Likewise.
7262 * internal-fn.c (expand_DEFERRED_INIT): View-convert the
7263 RHS if the LHS is an SSA name.
7265 2021-11-02 Richard Sandiford <richard.sandiford@arm.com>
7267 * langhooks.h (lang_hooks_for_types::simulate_record_decl): New hook.
7268 * langhooks-def.h (lhd_simulate_record_decl): Declare.
7269 (LANG_HOOKS_SIMULATE_RECORD_DECL): Define.
7270 (LANG_HOOKS_FOR_TYPES_INITIALIZER): Include it.
7271 * langhooks.c (lhd_simulate_record_decl): New function.
7273 2021-11-02 Jakub Jelinek <jakub@redhat.com>
7276 * config/i386/i386.c (ix86_vector_mode_supported_p): Reject vector
7277 modes with TImode inner mode if 32-bit.
7279 2021-11-02 liuhongt <hongtao.liu@intel.com>
7281 * doc/sourcebuild.texi (vect_slp_v4qi_store_unalign,
7282 vect_slp_v2hi_store_unalign, vect_slp_v4hi_store_unalign,
7283 vect_slp_v4si_store_unalign): Document efficient target.
7284 (vect_slp_v4qi_store_unalign_1, vect_slp_v8qi_store_unalign_1,
7285 vect_slp_v16qi_store_unalign_1): Ditto.
7286 (vect_slp_v2hi_store_align,vect_slp_v2qi_store_align,
7287 vect_slp_v2si_store_align, vect_slp_v4qi_store_align): Ditto.
7288 (struct_4char_block_move, struct_8char_block_move,
7289 struct_16char_block_move): Ditto.
7291 2021-11-02 Roger Sayle <roger@nextmovesoftware.com>
7292 Jakub Jelinek <jakub@redhat.com>
7295 * config/i386/i386-expand.c (ix86_expand_v1ti_to_ti,
7296 ix86_expand_ti_to_v1ti): New helper functions.
7297 (ix86_expand_v1ti_shift): Check if the amount operand is an
7298 integer constant, and expand as a TImode shift if it isn't.
7299 (ix86_expand_v1ti_rotate): Check if the amount operand is an
7300 integer constant, and expand as a TImode rotate if it isn't.
7301 (ix86_expand_v1ti_ashiftrt): New function to expand arithmetic
7302 right shifts of V1TImode quantities.
7303 * config/i386/i386-protos.h (ix86_expand_v1ti_ashift): Prototype.
7304 * config/i386/sse.md (ashlv1ti3, lshrv1ti3): Change constraints
7305 to QImode general_operand, and let the helper functions lower
7306 shifts by non-constant operands, as TImode shifts. Make
7307 conditional on TARGET_64BIT.
7308 (ashrv1ti3): New expander calling ix86_expand_v1ti_ashiftrt.
7309 (rotlv1ti3, rotrv1ti3): Change shift operand to QImode.
7310 Make conditional on TARGET_64BIT.
7312 2021-11-02 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
7314 * config/s390/s390.md ("*cc_to_int", "tabort", "*tabort_1",
7315 "*tabort_1_plus"): Remove operands non-null check.
7317 2021-11-01 Martin Liska <mliska@suse.cz>
7319 * opt-functions.awk: Add new sanity checking.
7320 * optc-gen.awk: Add new argument to integer_range_info.
7321 * params.opt: Update 2 params which have negative IntegerRange.
7323 2021-11-01 qing zhao <qing.zhao@oracle.com>
7325 * gimplify.c (gimplify_decl_expr): Do not add call to
7326 __builtin_clear_padding when a variable is a gimple register
7327 or it might not have padding.
7328 (gimplify_init_constructor): Likewise.
7330 2021-11-01 Tamar Christina <tamar.christina@arm.com>
7332 * config/arm/aarch-common-protos.h (struct vector_cost_table): Add
7333 movi, dup and extract costing fields.
7334 * config/aarch64/aarch64-cost-tables.h (qdf24xx_extra_costs,
7335 thunderx_extra_costs, thunderx2t99_extra_costs,
7336 thunderx3t110_extra_costs, tsv110_extra_costs, a64fx_extra_costs): Use
7338 * config/arm/aarch-cost-tables.h (generic_extra_costs,
7339 cortexa53_extra_costs, cortexa57_extra_costs, cortexa76_extra_costs,
7340 exynosm1_extra_costs, xgene1_extra_costs): Likewise
7341 * config/aarch64/aarch64-simd.md (aarch64_simd_dup<mode>): Add r->w dup.
7342 * config/aarch64/aarch64.c (aarch64_rtx_costs): Add extra costs.
7344 2021-11-01 Tamar Christina <tamar.christina@arm.com>
7346 * cse.c (add_to_set): New.
7347 (find_sets_in_insn): Register constants in sets.
7348 (canonicalize_insn): Use auto_vec instead.
7349 (cse_insn): Try materializing using vec_dup.
7350 * rtl.h (simplify_context::simplify_gen_vec_select,
7351 simplify_gen_vec_select): New.
7352 * simplify-rtx.c (simplify_context::simplify_gen_vec_select): New.
7354 2021-11-01 David Malcolm <dmalcolm@redhat.com>
7356 * common.opt (fdiagnostics-escape-format=): New.
7357 (diagnostics_escape_format): New enum.
7358 (DIAGNOSTICS_ESCAPE_FORMAT_UNICODE): New enum value.
7359 (DIAGNOSTICS_ESCAPE_FORMAT_BYTES): Likewise.
7360 * diagnostic-format-json.cc (json_end_diagnostic): Add
7361 "escape-source" attribute.
7362 * diagnostic-show-locus.c
7363 (exploc_with_display_col::exploc_with_display_col): Replace
7364 "tabstop" param with a cpp_char_column_policy and add an "aspect"
7365 param. Use these to compute m_display_col accordingly.
7366 (struct char_display_policy): New struct.
7367 (layout::m_policy): New field.
7368 (layout::m_escape_on_output): New field.
7369 (def_policy): New function.
7370 (make_range): Update for changes to exploc_with_display_col ctor.
7371 (default_print_decoded_ch): New.
7372 (width_per_escaped_byte): New.
7373 (escape_as_bytes_width): New.
7374 (escape_as_bytes_print): New.
7375 (escape_as_unicode_width): New.
7376 (escape_as_unicode_print): New.
7378 (layout::layout): Initialize new fields. Update m_exploc ctor
7379 call for above change to ctor.
7380 (layout::maybe_add_location_range): Update for changes to
7381 exploc_with_display_col ctor.
7382 (layout::calculate_x_offset_display): Update for change to
7384 (layout::print_source_line): Pass policy
7385 to cpp_display_width_computation. Capture cpp_decoded_char when
7386 calling process_next_codepoint. Move printing of source code to
7387 m_policy.m_print_cb.
7388 (line_label::line_label): Pass in policy rather than context.
7389 (layout::print_any_labels): Update for change to line_label ctor.
7390 (get_affected_range): Pass in policy rather than context, updating
7391 calls to location_compute_display_column accordingly.
7392 (get_printed_columns): Likewise, also for cpp_display_width.
7393 (correction::correction): Pass in policy rather than tabstop.
7394 (correction::compute_display_cols): Pass m_policy rather than
7395 m_tabstop to cpp_display_width.
7396 (correction::m_tabstop): Replace with...
7397 (correction::m_policy): ...this.
7398 (line_corrections::line_corrections): Pass in policy rather than
7400 (line_corrections::m_context): Replace with...
7401 (line_corrections::m_policy): ...this.
7402 (line_corrections::add_hint): Update to use m_policy rather than
7404 (line_corrections::add_hint): Likewise.
7405 (layout::print_trailing_fixits): Likewise.
7406 (selftest::test_display_widths): New.
7407 (selftest::test_layout_x_offset_display_utf8): Update to use
7408 policy rather than tabstop.
7409 (selftest::test_one_liner_labels_utf8): Add test of escaping
7411 (selftest::test_diagnostic_show_locus_one_liner_utf8): Update to
7412 use policy rather than tabstop.
7413 (selftest::test_overlapped_fixit_printing): Likewise.
7414 (selftest::test_overlapped_fixit_printing_utf8): Likewise.
7415 (selftest::test_overlapped_fixit_printing_2): Likewise.
7416 (selftest::test_tab_expansion): Likewise.
7417 (selftest::test_escaping_bytes_1): New.
7418 (selftest::test_escaping_bytes_2): New.
7419 (selftest::diagnostic_show_locus_c_tests): Call the new tests.
7420 * diagnostic.c (diagnostic_initialize): Initialize
7421 context->escape_format.
7422 (convert_column_unit): Update to use default character width policy.
7423 (selftest::test_diagnostic_get_location_text): Likewise.
7424 * diagnostic.h (enum diagnostics_escape_format): New enum.
7425 (diagnostic_context::escape_format): New field.
7426 * doc/invoke.texi (-fdiagnostics-escape-format=): New option.
7427 (-fdiagnostics-format=): Add "escape-source" attribute to examples
7428 of JSON output, and document it.
7429 * input.c (location_compute_display_column): Pass in "policy"
7430 rather than "tabstop", passing to
7431 cpp_byte_column_to_display_column.
7432 (selftest::test_cpp_utf8): Update to use cpp_char_column_policy.
7433 * input.h (class cpp_char_column_policy): New forward decl.
7434 (location_compute_display_column): Pass in "policy" rather than
7436 * opts.c (common_handle_option): Handle
7437 OPT_fdiagnostics_escape_format_.
7438 * selftest.c (temp_source_file::temp_source_file): New ctor
7439 overload taking a size_t.
7440 * selftest.h (temp_source_file::temp_source_file): Likewise.
7442 2021-11-01 Aldy Hernandez <aldyh@redhat.com>
7444 * dbgcnt.def: Add debug counter for back_thread[12] and
7445 back_threadfull[12].
7446 * passes.def: Pass "first" argument to each back threading pass.
7447 * tree-ssa-threadbackward.c (back_threader::back_threader): Add
7449 (back_threader::debug_counter): New.
7450 (back_threader::maybe_register_path): Call debug_counter.
7452 2021-11-01 Aldy Hernandez <aldyh@redhat.com>
7454 * tree-ssa-threadbackward.c (BT_NONE): New.
7457 (back_threader::back_threader): Add flags.
7458 Move loop initialization here.
7459 (back_threader::~back_threader): New.
7460 (back_threader::find_taken_edge_switch): Change solver and ranger
7462 (back_threader::find_taken_edge_cond): Same.
7463 (back_threader::find_paths_to_names): Same.
7464 (back_threader::find_paths): Same.
7465 (back_threader::dump): Same.
7466 (try_thread_blocks): Merge into thread_blocks.
7467 (back_threader::thread_blocks): New.
7468 (do_early_thread_jumps): Merge into thread_blocks.
7469 (do_thread_jumps): Merge into thread_blocks.
7470 (back_threader::thread_through_all_blocks): Remove.
7472 2021-11-01 Andrew MacLeod <amacleod@redhat.com>
7474 PR tree-optimization/103003
7475 * value-relation.cc (dom_oracle::register_relation): If the 2
7476 ssa names are the same, don't register any relation.
7478 2021-11-01 Dan Li <ashimida@linux.alibaba.com>
7480 * config/aarch64/aarch64.c (aarch64_expand_epilogue): Remove
7481 redundant check for calls_eh_return.
7482 * config/aarch64/aarch64.md (*do_return): Likewise.
7484 2021-11-01 Xionghu Luo <luoxhu@linux.ibm.com>
7486 * cfghooks.c (cfg_hook_duplicate_loop_to_header_edge): Rename
7487 duplicate_loop_to_header_edge to
7488 duplicate_loop_body_to_header_edge.
7489 (cfg_hook_duplicate_loop_body_to_header_edge): Likewise.
7490 * cfghooks.h (struct cfg_hooks): Likewise.
7491 (cfg_hook_duplicate_loop_body_to_header_edge): Likewise.
7492 * cfgloopmanip.c (duplicate_loop_body_to_header_edge): Likewise.
7493 (clone_loop_to_header_edge): Likewise.
7494 * cfgloopmanip.h (duplicate_loop_body_to_header_edge): Likewise.
7495 * cfgrtl.c (struct cfg_hooks): Likewise.
7496 * doc/loop.texi: Likewise.
7497 * loop-unroll.c (unroll_loop_constant_iterations): Likewise.
7498 (unroll_loop_runtime_iterations): Likewise.
7499 (unroll_loop_stupid): Likewise.
7500 (apply_opt_in_copies): Likewise.
7501 * tree-cfg.c (struct cfg_hooks): Likewise.
7502 * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Likewise.
7503 (try_peel_loop): Likewise.
7504 * tree-ssa-loop-manip.c (copy_phi_node_args): Likewise.
7505 (gimple_duplicate_loop_body_to_header_edge): Likewise.
7506 (tree_transform_and_unroll_loop): Likewise.
7507 * tree-ssa-loop-manip.h (gimple_duplicate_loop_body_to_header_edge):
7510 2021-11-01 Xionghu Luo <luoxhu@linux.ibm.com>
7512 * cfgloopmanip.c (loop_version): Refactor loopify to
7513 loop_version. Move condition generation after loopify.
7515 * cfgloopmanip.h (loopify): Delete.
7517 2021-10-31 Jan Hubicka <hubicka@ucw.cz>
7519 * ipa-fnsummary.c: Include tree-dfa.h.
7520 (points_to_local_or_readonly_memory_p): Return true on return
7522 * ipa-modref.c (analyze_ssa_name_flags): Fix handling of copy
7525 2021-10-30 Tobias Burnus <tobias@codesourcery.com>
7527 PR middle-end/102972
7528 * omp-low.c (omp_runtime_api_call): Use DECL_ASSEMBLER_NAME to get
7529 internal Fortran name; new permit_num_teams arg to permit
7530 omp_get_num_teams and omp_get_team_num.
7531 (scan_omp_1_stmt): Update call to it, add missing call for
7532 reverse offload, and check for strictly nested API calls in teams.
7534 2021-10-30 Jakub Jelinek <jakub@redhat.com>
7536 * gimplify.c (gimplify_omp_for): Diagnose threadprivate iterators.
7538 2021-10-29 David Malcolm <dmalcolm@redhat.com>
7540 * selftest.c (assert_streq): Add newlines when emitting non-equal
7543 2021-10-29 David Malcolm <dmalcolm@redhat.com>
7545 * Makefile.in: Fix syntax for reference to LIBDEPS in
7548 2021-10-29 Bernhard Reutner-Fischer <aldot@gcc.gnu.org>
7550 * doc/install.texi: Bump required minimum DejaGnu version.
7552 2021-10-29 Aldy Hernandez <aldyh@redhat.com>
7553 Andrew MacLeod <amacleod@redhat.com>
7555 * value-relation.cc (path_oracle::killing_def): Add a
7556 self-equivalence so we don't look to the root oracle.
7558 2021-10-29 Aldy Hernandez <aldyh@redhat.com>
7560 * passes.def: Replace the pass_thread_jumps before VRP* with
7561 pass_thread_jumps_full. Remove all pass_vrp_threader instances.
7562 * tree-ssa-threadbackward.c (pass_data_thread_jumps_full):
7563 Remove hyphen from "thread-full" name.
7565 2021-10-29 Andrew MacLeod <amacleod@redhat.com>
7567 PR tree-optimization/102983
7568 * gimple-range-cache.h (propagate_updated_value): Make public.
7569 * gimple-range.cc (gimple_ranger::range_of_stmt): Propagate exports
7570 when processing gcond stmts.
7572 2021-10-29 Jan Hubicka <hubicka@ucw.cz>
7574 * gimple.c (gimple_call_retslot_flags): New function.
7575 * gimple.h (gimple_call_retslot_flags): Declare.
7576 * ipa-modref.c: Include tree-cfg.h.
7577 (struct escape_entry): Turn parm_index to signed.
7578 (modref_summary_lto::modref_summary_lto): Add retslot_flags.
7579 (modref_summary::modref_summary): Initialize retslot_flags.
7580 (struct modref_summary_lto): Likewise.
7581 (modref_summary::useful_p): Check retslot_flags.
7582 (modref_summary_lto::useful_p): Likewise.
7583 (modref_summary::dump): Dump retslot_flags.
7584 (modref_summary_lto::dump): Likewise.
7585 (struct escape_point): Add hidden_args enum.
7586 (analyze_ssa_name_flags): Ignore return slot return;
7587 use gimple_call_retslot_flags.
7588 (record_escape_points): Break out from ...
7589 (analyze_parms): ... here; handle retslot_flags.
7590 (modref_summaries::duplicate): Duplicate retslot_flags.
7591 (modref_summaries_lto::duplicate): Likewise.
7592 (modref_write_escape_summary): Stream parm_index as signed.
7593 (modref_read_escape_summary): Likewise.
7594 (modref_write): Stream retslot_flags.
7595 (read_section): Likewise.
7596 (struct escape_map): Fix typo in comment.
7597 (update_escape_summary_1): Fix whitespace.
7598 (ipa_merge_modref_summary_after_inlining): Drop retslot_flags.
7599 (modref_merge_call_site_flags): Merge retslot_flags.
7600 * ipa-modref.h (struct modref_summary): Add retslot_flags.
7601 * tree-ssa-structalias.c (handle_rhs_call): Handle retslot_flags.
7603 2021-10-29 Tamar Christina <tamar.christina@arm.com>
7605 PR tree-optimization/102977
7606 * tree-vect-slp-patterns.c (vect_match_call_p): Remove.
7607 (vect_detect_pair_op): Add crosslane check.
7608 (vect_match_call_complex_mla): Remove.
7609 (class complex_mul_pattern): Update comment.
7610 (complex_mul_pattern::matches): Update detection.
7611 (class complex_fma_pattern): Remove.
7612 (complex_fma_pattern::matches): Remove.
7613 (complex_fma_pattern::recognize): Remove.
7614 (complex_fma_pattern::build): Remove.
7615 (class complex_fms_pattern): Update comment.
7616 (complex_fms_pattern::matches): Remove.
7617 (complex_operations_pattern::recognize): Remove complex_fma_pattern
7619 2021-10-29 Jakub Jelinek <jakub@redhat.com>
7621 * gimple-fold.c (gimple_fold_builtin_memset): Copy over location from
7624 2021-10-29 Xionghu Luo <luoxhu@linux.ibm.com>
7627 * config/rs6000/rs6000.c (altivec_expand_vec_perm_const): Add
7628 patterns match and emit for VSX xxpermdi.
7630 2021-10-29 liuhongt <hongtao.liu@intel.com>
7633 * config/i386/i386-builtin-types.def (V8HF_FTYPE_V8HF): New
7635 (V16HF_FTYPE_V16HF): Ditto.
7636 (V32HF_FTYPE_V32HF): Ditto.
7637 (V8HF_FTYPE_V8HF_ROUND): Ditto.
7638 (V16HF_FTYPE_V16HF_ROUND): Ditto.
7639 (V32HF_FTYPE_V32HF_ROUND): Ditto.
7640 * config/i386/i386-builtin.def ( IX86_BUILTIN_FLOORPH,
7641 IX86_BUILTIN_CEILPH, IX86_BUILTIN_TRUNCPH,
7642 IX86_BUILTIN_FLOORPH256, IX86_BUILTIN_CEILPH256,
7643 IX86_BUILTIN_TRUNCPH256, IX86_BUILTIN_FLOORPH512,
7644 IX86_BUILTIN_CEILPH512, IX86_BUILTIN_TRUNCPH512): New builtin.
7645 * config/i386/i386-builtins.c
7646 (ix86_builtin_vectorized_function): Enable vectorization for
7647 HFmode FLOOR/CEIL/TRUNC operation.
7648 * config/i386/i386-expand.c (ix86_expand_args_builtin): Handle
7650 * config/i386/sse.md (rint<mode>2, nearbyint<mode>2): Extend
7653 2021-10-28 Aldy Hernandez <aldyh@redhat.com>
7654 Andrew MacLeod <amacleod@redhat.com>
7656 * value-relation.cc (path_oracle::killing_def): Walk the
7657 equivalency list and remove SSA from any equivalencies.
7659 2021-10-28 Stafford Horne <shorne@gmail.com>
7661 * config/or1k/or1k.h (PROFILE_HOOK): Add return address argument
7664 2021-10-28 Jakub Jelinek <jakub@redhat.com>
7666 PR tree-optimization/102951
7667 * fold-const.h (address_compare): Declare.
7668 * fold-const.c (address_compare): New function.
7669 * match.pd (cmp (convert1?@2 addr@0) (convert2? addr@1)): Use
7670 address_compare helper.
7671 (minmax cmp (convert1?@2 addr@0) (convert2?@3 addr@1)): New
7674 2021-10-28 Andrew MacLeod <amacleod@redhat.com>
7676 * vr-values.c (simplify_using_ranges::fold_cond): Change fold message.
7678 2021-10-28 Andrew MacLeod <amacleod@redhat.com>
7680 PR tree-optimization/102940
7681 * tree-vrp.c (execute_ranger_vrp): Reset scev.
7683 2021-10-28 Richard Purdie <richard.purdie@linuxfoundation.org>
7685 * config/nios2/linux.h (MUSL_DYNAMIC_LINKER): Add musl linker
7687 2021-10-28 Richard Purdie <richard.purdie@linuxfoundation.org>
7689 * configure: Regenerate.
7690 * configure.ac: Use CPPFLAGS_FOR_BUILD for GMPINC
7692 2021-10-28 Eric Botcazou <ebotcazou@adacore.com>
7694 * doc/invoke.texi (%X): Remove obsolete reference to -Wl.
7696 2021-10-28 Richard Biener <rguenther@suse.de>
7699 * fold-const.c (fold_convert_const): Avoid int to float
7700 constant folding with -frounding-math and inexact result.
7701 * simplify-rtx.c (simplify_const_unary_operation): Likewise
7702 for both float and unsigned_float.
7704 2021-10-28 Aldy Hernandez <aldyh@redhat.com>
7706 * tree-ssa-threadbackward.c
7707 (back_threader::find_taken_edge_switch): Use find_case_label_range
7708 instead of find_taken_edge.
7710 2021-10-28 Aldy Hernandez <aldyh@redhat.com>
7712 * tree-ssa-threadbackward.c (class back_threader_registry):
7713 Inherit from back_jt_path_registry.
7714 (back_threader_registry::thread_through_all_blocks): Remove.
7715 (back_threader_registry::register_path): Remove
7716 m_lowlevel_registry prefix.
7718 2021-10-28 Richard Biener <rguenther@suse.de>
7721 * fold-const.c (fold_convert_const_real_from_real): Honor
7722 -frounding-math if the conversion is not exact.
7723 * simplify-rtx.c (simplify_const_unary_operation): Do not
7724 simplify FLOAT_TRUNCATE with sign dependent rounding.
7726 2021-10-28 Richard Biener <rguenther@suse.de>
7728 PR tree-optimization/102949
7729 * tree-vect-stmts.c (ensure_base_align): Look at the
7730 dr_info of a group leader and assert we are looking at
7731 one with analyzed alignment.
7733 2021-10-28 Kewen Lin <linkw@linux.ibm.com>
7736 * config/rs6000/rs6000.c (rs6000_builtin_vectorization_cost): Consider
7737 V1T1 mode for unaligned load and store.
7739 2021-10-28 Kito Cheng <kito.cheng@sifive.com>
7741 * config/riscv/riscv.md (zero_extendsidi2_internal): Allow ZBB
7744 2021-10-28 Kito Cheng <kito.cheng@sifive.com>
7746 * config/riscv/arch-canonicalize (CANONICAL_ORDER): Add `i` to
7749 2021-10-28 Alexandre Oliva <oliva@adacore.com>
7751 * common.opt (fharden-compares): New.
7752 (fharden-conditional-branches): New.
7753 * doc/invoke.texi: Document new options.
7754 * gimple-harden-conditionals.cc: New.
7755 * Makefile.in (OBJS): Build it.
7756 * passes.def: Add new passes.
7757 * tree-pass.h (make_pass_harden_compares): Declare.
7758 (make_pass_harden_conditional_branches): Declare.
7760 2021-10-28 Xionghu Luo <luoxhu@linux.ibm.com>
7763 * config/rs6000/altivec.md: Add vsx register constraints.
7764 * config/rs6000/vsx.md (vsx_xxsel<mode>): Delete.
7765 (vsx_xxsel<mode>2): Likewise.
7766 (vsx_xxsel<mode>3): Likewise.
7767 (vsx_xxsel<mode>4): Likewise.
7769 2021-10-28 Xionghu Luo <luoxhu@linux.ibm.com>
7772 * config/rs6000/altivec.md (*altivec_vsel<mode>): Change to ...
7773 (altivec_vsel<mode>): ... this and update define.
7774 (*altivec_vsel<mode>_uns): Delete.
7775 (altivec_vsel<mode>2): New define_insn.
7776 (altivec_vsel<mode>3): Likewise.
7777 (altivec_vsel<mode>4): Likewise.
7778 * config/rs6000/rs6000-call.c (altivec_expand_vec_sel_builtin): New.
7779 (altivec_expand_builtin): Call altivec_expand_vec_sel_builtin to expand
7781 * config/rs6000/rs6000.c (rs6000_emit_vector_cond_expr): Use bit-wise
7782 selection instead of per element.
7783 * config/rs6000/vector.md:
7784 * config/rs6000/vsx.md (*vsx_xxsel<mode>): Change to ...
7785 (vsx_xxsel<mode>): ... this and update define.
7786 (*vsx_xxsel<mode>_uns): Delete.
7787 (vsx_xxsel<mode>2): New define_insn.
7788 (vsx_xxsel<mode>3): Likewise.
7789 (vsx_xxsel<mode>4): Likewise.
7791 2021-10-28 Hongyu Wang <hongyu.wang@intel.com>
7793 * config/i386/i386.c (use_rsqrt_p): Add mode parameter, enable
7794 HFmode rsqrt without TARGET_SSE_MATH.
7795 (ix86_optab_supported_p): Refactor rint, adjust floor, ceil,
7796 btrunc condition to be restricted by -ftrapping-math, adjust
7797 use_rsqrt_p function call.
7798 * config/i386/i386.md (rcphf2): New define_insn.
7799 (rsqrthf2): Likewise.
7800 * config/i386/sse.md (div<mode>3): Change VF2H to VF2.
7801 (div<mode>3): New expander for HF mode.
7802 (rsqrt<mode>2): Likewise.
7803 (*avx512fp16_vmrcpv8hf2): New define_insn for rpad pass.
7804 (*avx512fp16_vmrsqrtv8hf2): Likewise.
7806 2021-10-27 Saagar Jha <saagar@saagarjha.com>
7808 * config.gcc: Adjust for Darwin21.
7809 * config/darwin-c.c (macosx_version_as_macro): Likewise.
7810 * config/darwin-driver.c (validate_macosx_version_min):
7812 (darwin_find_version_from_kernel): Likewise.
7814 2021-10-27 Aldy Hernandez <aldyh@redhat.com>
7816 * gimple-range-path.cc
7817 (path_range_query::range_defined_in_block): Call killing_def.
7819 2021-10-27 Aldy Hernandez <aldyh@redhat.com>
7821 * gimple-range-fold.cc (fold_using_range::range_of_range_op): Dump
7822 operands as well as relation.
7823 * gimple-range-path.cc
7824 (path_range_query::compute_ranges_in_block): Compute PHI relations
7825 first. Compute outgoing relations at the end.
7826 (path_range_query::compute_ranges): Remove call to compute_relations.
7827 (path_range_query::compute_relations): Remove.
7828 (path_range_query::maybe_register_phi_relation): New.
7829 (path_range_query::compute_phi_relations): Abstract out
7830 registering one PHI relation to...
7831 (path_range_query::compute_outgoing_relations): ...here.
7832 * gimple-range-path.h (class path_range_query): Remove
7834 Add maybe_register_phi_relation.
7836 2021-10-27 Aldy Hernandez <aldyh@redhat.com>
7837 Andrew MacLeod <amacleod@redhat.com>
7839 * value-relation.cc (path_oracle::killing_def): Kill second
7842 2021-10-27 Martin Jambor <mjambor@suse.cz>
7844 * ipa-cp.c (good_cloning_opportunity_p): Decide whether to use
7845 profile feedback depending on their local availability.
7847 2021-10-27 Martin Jambor <mjambor@suse.cz>
7849 * params.opt (param_ipa_cp_profile_count_base): New parameter.
7850 * doc/invoke.texi (Optimize Options): Add entry for
7851 ipa-cp-profile-count-base.
7852 * ipa-cp.c (max_count): Replace with base_count, replace all
7853 occurrences too, unless otherwise stated.
7854 (ipcp_cloning_candidate_p): identify mostly-directly called
7855 functions based on their counts, not max_count.
7856 (compare_edge_profile_counts): New function.
7857 (ipcp_propagate_stage): Instead of setting max_count, find the
7858 appropriate edge count in a sorted vector of counts of eligible
7859 edges and make it the base_count.
7861 2021-10-27 Martin Jambor <mjambor@suse.cz>
7863 * ipa-cp.c (struct caller_statistics): New fields rec_count_sum,
7864 n_nonrec_calls and itself, document all fields.
7865 (init_caller_stats): Initialize the above new fields.
7866 (gather_caller_stats): Gather self-recursive counts and calls number.
7867 (get_info_about_necessary_edges): Gather counts of self-recursive and
7868 other edges bringing in the requested value separately.
7869 (dump_profile_updates): Rework to dump info about a single node only.
7870 (lenient_count_portion_handling): New function.
7871 (struct gather_other_count_struct): New type.
7872 (gather_count_of_non_rec_edges): New function.
7873 (struct desc_incoming_count_struct): New type.
7874 (analyze_clone_icoming_counts): New function.
7875 (adjust_clone_incoming_counts): Likewise.
7876 (update_counts_for_self_gen_clones): Likewise.
7877 (update_profiling_info): Rewritten.
7878 (update_specialized_profile): Adjust call to dump_profile_updates.
7879 (create_specialized_node): Do not update profiling info.
7880 (decide_about_value): New parameter self_gen_clones, either push new
7881 clones into it or updat their profile counts. For self-recursively
7882 generated values, use a portion of the node count instead of count
7883 from self-recursive edges to estimate goodness.
7884 (decide_whether_version_node): Gather clones for self-generated values
7885 in a new vector, update their profiles at once at the end.
7887 2021-10-27 Richard Biener <rguenther@suse.de>
7889 * tree-vectorizer.c (vect_transform_loops): New function,
7891 (try_vectorize_loop_1): ... here. Simplify as epilogues
7892 are now fully handled in the split part.
7894 2021-10-27 Jakub Jelinek <jakub@redhat.com>
7896 * omp-expand.c (expand_omp_for_init_counts): Handle non-rectangular
7897 iterators with pointer types.
7898 (expand_omp_for_init_vars, extract_omp_for_update_vars): Likewise.
7900 2021-10-26 Martin Sebor <msebor@redhat.com>
7902 PR tree-optimization/102238
7903 PR tree-optimization/102919
7904 * gimple-ssa-sprintf.c (get_string_length): Add an argument.
7905 (array_elt_at_offset): Move to pointer-query.
7906 (set_aggregate_size_and_offset): New function.
7907 (field_at_offset): Move to pointer-query.
7908 (get_origin_and_offset): Rename...
7909 (get_origin_and_offset_r): this. Add an argument. Make aggregate
7910 handling more robust.
7911 (get_origin_and_offset): New.
7912 (alias_offset): Add an argument.
7913 (format_string): Use subobject size determined by get_origin_and_offset.
7914 * pointer-query.cc (field_at_offset): Move from gimple-ssa-sprintf.c.
7915 Improve/correct handling of aggregates.
7916 (array_elt_at_offset): Same.
7917 * pointer-query.h (field_at_offset): Declare.
7918 (array_elt_at_offset): Declare.
7920 2021-10-26 Martin Sebor <msebor@redhat.com>
7922 * builtins.c (check_strncat_sizes): Pass access_data ctor additional
7924 (expand_builtin_memcmp): Move code to gimple-ssa-warn-access.cc.
7925 (expand_builtin_fork_or_exec): Same.
7926 * gimple-array-bounds.cc (array_bounds_checker::check_mem_ref): Pass
7927 compute_objsize additional arguments.
7928 (inbounds_memaccess_p): Same.
7929 (array_bounds_checker::check_array_bounds): Add an assert. Stash
7930 statement in a member.
7931 (check_array_bounds_dom_walker::before_dom_children): Same.
7932 * gimple-array-bounds.h (array_bounds_checker::m_stmt): New member.
7933 * gimple-ssa-sprintf.c (get_destination_size): Add an argument.
7934 (handle_printf_call): Pass a new argument.
7935 * gimple-ssa-warn-access.cc (get_size_range): Add an argument.
7936 (check_access): Add an argument and pass it along to callees.
7937 (check_read_access): Make a member function.
7938 (pass_waccess::check_strcat): Pass access_data ctor additional
7940 (pass_waccess::check_strncat): Same.
7941 (pass_waccess::check_stxcpy): Same.
7942 (pass_waccess::check_stxncpy): Same.
7943 (pass_waccess::check_strncmp): Same.
7944 (pass_waccess::check_read_access): Same.
7945 (pass_waccess::check_builtin): Same.
7946 (pass_waccess::maybe_check_access_sizes): Same.
7947 (pass_waccess::maybe_check_dealloc_call): Same.
7948 * gimple-ssa-warn-access.h (check_read_access): Declare a new
7950 * pointer-query.cc (compute_objsize_r): Add an argument.
7951 (gimple_call_return_array): Same.
7952 (gimple_call_alloc_size): Same.
7953 (access_ref::access_ref): Same.
7954 (access_ref::get_ref): Same.
7955 (pointer_query::get_ref): Same.
7956 (handle_min_max_size): Pass an arguments to callees.
7957 (handle_array_ref): Add an argument.
7958 (handle_mem_ref): Same.
7959 (compute_objsize): Same.
7960 * pointer-query.h (struct access_ref): Adjust signatures.
7961 (struct access_data): Same.
7962 (gimple_call_alloc_size): Add an argument.
7963 (gimple_parm_array_size): Same.
7964 (compute_objsize): Same.
7965 * tree-ssa-strlen.c (strlen_pass::adjust_last_stmt): Pass an additional
7966 argument to compute_objsize.
7967 (strlen_pass::maybe_warn_overflow): Same.
7968 (maybe_diag_stxncpy_trunc): Same.
7970 2021-10-26 Martin Sebor <msebor@redhat.com>
7972 PR middle-end/102453
7973 * gimple-ssa-warn-access.cc (pass_waccess::check_atomic_builtin): New.
7974 (pass_waccess::check_atomic_builtin): Call it.
7976 2021-10-26 Vladimir N. Makarov <vmakarov@redhat.com>
7978 PR rtl-optimization/102842
7979 * lra-constraints.c (match_reload): Ignore out in checking values
7981 (curr_insn_transform): Collect outputs before doing reloads of operands.
7983 2021-10-26 Paul A. Clarke <pc@us.ibm.com>
7986 * config/rs6000/x86intrin.h: Move some included headers to new
7987 headers. Include new immintrin.h instead of those headers.
7988 * config/rs6000/immintrin.h: New.
7989 * config/rs6000/x86gprintrin.h: New.
7990 * config.gcc (powerpc*-*-*): Add new headers to extra_headers.
7992 2021-10-26 Richard Biener <rguenther@suse.de>
7994 * tree-vectorizer.h (vect_create_addr_base_for_vector_ref):
7995 Remove byte_offset parameter.
7996 (vect_create_data_ref_ptr): Likewise.
7997 * tree-vect-data-refs.c (vect_create_addr_base_for_vector_ref):
7999 (vect_create_data_ref_ptr): Likewise.
8000 * tree-vect-stmts.c (vectorizable_store): Adjust.
8001 (vectorizable_load): Likewise.
8003 2021-10-26 Richard Biener <rguenther@suse.de>
8005 PR tree-optimization/96109
8006 * tree-vectorizer.h (dr_misalignment): Add optional offset
8008 * tree-vect-data-refs.c (dr_misalignment): Likewise. Remove
8009 offset applied for negative stride accesses.
8010 (vect_enhance_data_refs_alignment): Compute negative stride
8011 access offset and pass it to dr_misalignment.
8012 * tree-vect-stmts.c (get_negative_load_store_type): Pass
8013 negative offset to dr_misalignment.
8014 (get_group_load_store_type): Likewise.
8015 (get_load_store_type): Likewise.
8016 (vectorizable_store): Remove asserts about alignment.
8017 (vectorizable_load): Likewise.
8019 2021-10-26 Kewen Lin <linkw@linux.ibm.com>
8021 PR tree-optimization/102897
8022 * tree-ssa-forwprop.c (simplify_permutation): Remove a wrong assertion.
8024 2021-10-26 Richard Biener <rguenther@suse.de>
8026 * tree-vect-data-refs.c (vect_create_addr_base_for_vector_ref):
8027 Take offset in bytes.
8028 (vect_create_data_ref_ptr): Likewise.
8029 * tree-vect-loop-manip.c (get_misalign_in_elems): Multiply
8030 offset by element size.
8031 (vect_create_cond_for_align_checks): Likewise.
8032 * tree-vect-stmts.c (get_negative_load_store_type): Likewise.
8033 (vectorizable_load): Remove duplicate leftover from merge
8036 2021-10-26 Roger Sayle <roger@nextmovesoftware.com>
8038 * config/i386/i386-expand.c (ix86_expand_v1ti_shift): New helper
8039 function to expand V1TI mode logical shifts by integer constants.
8040 (ix86_expand_v1ti_rotate): New helper function to expand V1TI
8041 mode rotations by integer constants.
8042 * config/i386/i386-protos.h (ix86_expand_v1ti_shift,
8043 ix86_expand_v1ti_rotate): Prototype new functions here.
8044 * config/i386/sse.md (ashlv1ti3, lshrv1ti3, rotlv1ti3, rotrv1ti3):
8045 New TARGET_SSE2 expanders to implement V1TI shifts and rotations.
8047 2021-10-26 Aldy Hernandez <aldyh@redhat.com>
8049 * tree-ssa-threadbackward.c (back_threader::maybe_register_path):
8050 Avoid threading circular paths.
8052 2021-10-26 Aldy Hernandez <aldyh@redhat.com>
8054 * tree-ssa-threadbackward.c (back_threader::resolve_phi):
8055 Attempt to resolve all incoming paths to a PHI.
8056 (back_threader::resolve_def): Always return true for PHIs.
8058 2021-10-26 Aldy Hernandez <aldyh@redhat.com>
8060 * tree-ssa-threadbackward.c (back_threader::find_paths_to_names):
8061 Always try to resolve path without looking back.
8062 * tree-ssa-threadupdate.c (dump_jump_thread): Indidicate whether
8063 edge is a back edge.
8065 2021-10-26 Kewen Lin <linkw@linux.ibm.com>
8067 PR tree-optimization/102789
8068 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
8069 update inits of simd_lane_access.
8071 2021-10-25 Andrew MacLeod <amacleod@redhat.com>
8073 * tree-vrp.c (vrp_simplify_cond_using_ranges): Add return type and
8074 move to vr-values.c.
8075 (simplify_casted_conds): Move to vrp_folder class.
8076 (execute_vrp): Call via vrp_folder now.
8077 * vr-values.c (simplify_cond_using_ranges_1): Call simplify_casted_cond.
8078 (simplify_using_ranges::simplify_casted_cond): Relocate from tree-vrp.c.
8079 * vr-values.h (simplify_casted_cond): Add prototype.
8081 2021-10-25 Andrew MacLeod <amacleod@redhat.com>
8083 * tree-vrp.c (rvrp_folder::fold_stmt): If simplification fails, try
8086 2021-10-25 Paul A. Clarke <pc@us.ibm.com>
8088 * config/rs6000/smmintrin.h (_mm_testz_si128): Add "extern" to
8090 (_mm_testc_si128): Likewise.
8091 (_mm_testnzc_si128): Likewise.
8092 (_mm_blend_ps): Likewise.
8093 (_mm_blendv_ps): Likewise.
8094 (_mm_blend_pd): Likewise.
8095 (_mm_blendv_pd): Likewise.
8096 (_mm_ceil_pd): Likewise.
8097 (_mm_ceil_sd): Likewise.
8098 (_mm_ceil_ps): Likewise.
8099 (_mm_ceil_ss): Likewise.
8100 (_mm_floor_pd): Likewise.
8101 (_mm_floor_sd): Likewise.
8102 (_mm_floor_ps): Likewise.
8103 (_mm_floor_ss): Likewise.
8104 (_mm_minpos_epu16): Likewise.
8105 (_mm_mul_epi32): Likewise.
8106 (_mm_cvtepi8_epi16): Likewise.
8107 (_mm_packus_epi32): Likewise.
8108 (_mm_cmpgt_epi64): Likewise.
8110 2021-10-25 Roger Sayle <roger@nextmovesoftware.com>
8112 * simplify-rtx.c (simplify_binary_operation_1) [SS_ASHIFT]: Simplify
8113 shifts of the mode's smin_value and smax_value when the bit count
8114 operand doesn't have side-effects.
8115 [US_ASHIFT]: Likewise, simplify shifts of the mode's umax_value
8116 when the bit count operand doesn't have side-effects.
8117 (simplify_const_binary_operation) [SS_ASHIFT, US_ASHIFT]: Perform
8118 compile-time evaluation of saturating left shifts with constant
8121 2021-10-25 Andrew MacLeod <amacleod@redhat.com>
8123 * gimple-range.cc (gimple_ranger::export_global_ranges): Remove check
8126 2021-10-25 Andrew MacLeod <amacleod@redhat.com>
8128 * flag-types.h (enum ranger_debug): Adjust values.
8129 * params.opt (ranger_debug): Ditto.
8131 2021-10-25 Martin Jambor <mjambor@suse.cz>
8133 PR tree-optimization/102886
8134 * tree-sra.c (totally_scalarize_subtree): Fix the out of
8137 2021-10-25 Andrew Pinski <apinski@marvell.com>
8139 * tree-ssa-dce.c (simple_dce_from_worklist):
8140 Check stmt_unremovable_because_of_non_call_eh_p also
8141 before removing the statement.
8143 2021-10-25 Richard Biener <rguenther@suse.de>
8145 PR tree-optimization/102905
8146 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment):
8147 Use vect_supportable_dr_alignment again to determine whether
8148 an access is supported when not aligned.
8150 2021-10-25 Kito Cheng <kito.cheng@sifive.com>
8152 * config/riscv/riscv.c (riscv_rtx_costs): Handle cost model
8155 2021-10-25 Jim Wilson <jimw@sifive.com>
8156 Kito Cheng <kito.cheng@sifive.com>
8157 Jia-Wei Chen <jiawei@iscas.ac.cn>
8158 Shi-Hua Liao <shihua@iscas.ac.cn>
8160 * config/riscv/bitmanip.md (shiftm1): New.
8161 (*bset<mode>): Ditto.
8162 (*bset<mode>_mask): Ditto.
8163 (*bset<mode>_1): Ditto.
8164 (*bset<mode>_1_mask): Ditto.
8165 (*bseti<mode>): Ditto.
8166 (*bclr<mode>): Ditto.
8167 (*bclri<mode>): Ditto.
8168 (*binv<mode>): Ditto.
8169 (*binvi<mode>): Ditto.
8170 (*bext<mode>): Ditto.
8172 * config/riscv/predicates.md (splittable_const_int_operand):
8174 (single_bit_mask_operand): New.
8175 (not_single_bit_mask_operand): Ditto.
8176 (const31_operand): Ditto.
8177 (const63_operand): Ditto.
8178 * config/riscv/riscv.c (riscv_build_integer_1): Handle bseti.
8179 (riscv_output_move): Ditto.
8180 (riscv_print_operand): Handle new operand type: T and S.
8181 * config/riscv/riscv.h (SINGLE_BIT_MASK_OPERAND): New.
8183 2021-10-25 Jim Wilson <jimw@sifive.com>
8185 * config/riscv/riscv.c (riscv_build_integer_1): Build integer
8188 2021-10-25 Kito Cheng <kito.cheng@sifive.com>
8190 * config/riscv/riscv.c (riscv_extend_cost): Handle cost model
8192 (riscv_rtx_costs): Ditto.
8194 2021-10-25 Jim Wilson <jimw@sifive.com>
8195 Kito Cheng <kito.cheng@sifive.com>
8196 Jia-Wei Chen <jiawei@iscas.ac.cn>
8198 * config/riscv/bitmanip.md (bitmanip_bitwise): New.
8199 (bitmanip_minmax): New.
8200 (clz_ctz_pcnt): New.
8201 (bitmanip_optab): New.
8202 (bitmanip_insn): New.
8203 (*<optab>_not<mode>): New.
8204 (*xor_not<mode>): New.
8205 (<bitmanip_optab>si2): New.
8206 (*<bitmanip_optab>disi2): New.
8207 (<bitmanip_optab>di2): New.
8208 (*zero_extendhi<GPR:mode>2_bitmanip): New.
8209 (*extend<SHORT:mode><SUPERQI:mode>2_zbb): New.
8210 (*zero_extendhi<GPR:mode>2_zbb): New.
8213 (rotrsi3_sext): New.
8216 (rotlsi3_sext): New.
8217 (bswap<mode>2): New.
8218 (<bitmanip_optab><mode>3): New.
8219 * config/riscv/riscv.md (type): Add rotate.
8220 (zero_extendhi<GPR:mode>2): Change to define_expand pattern.
8221 (*zero_extendhi<GPR:mode>2): New.
8222 (extend<SHORT:mode><SUPERQI:mode>2): Change to define_expand pattern.
8223 (*extend<SHORT:mode><SUPERQI:mode>2): New.
8225 2021-10-25 Kito Cheng <kito.cheng@sifive.com>
8227 * config/riscv/riscv.c (riscv_extend_cost): Handle cost model
8229 (riscv_rtx_costs): Ditto.
8231 2021-10-25 Jim Wilson <jimw@sifive.com>
8232 Kito Cheng <kito.cheng@sifive.com>
8233 Jia-Wei Chen <jiawei@iscas.ac.cn>
8235 * config/riscv/bitmanip.md (*zero_extendsidi2_bitmanip): New.
8240 (riscv_rtx_costs): Ditto.
8241 * config/riscv/riscv.md: Include bitmanip.md
8242 (type): Add bitmanip bype.
8243 (zero_extendsidi2): Change to define_expand pattern.
8244 (*zero_extendsidi2_internal): New.
8245 (zero_extendsidi2_shifted): Disable for ZBA.
8247 2021-10-25 Kito Cheng <kito.cheng@sifive.com>
8249 * common/config/riscv/riscv-common.c (riscv_ext_version_table):
8250 Add zba, zbb, zbc and zbs.
8251 (riscv_ext_flag_table): Ditto.
8252 * config/riscv/riscv-opts.h (MASK_ZBA): New.
8256 (TARGET_ZBA): Ditto.
8257 (TARGET_ZBB): Ditto.
8258 (TARGET_ZBC): Ditto.
8259 (TARGET_ZBS): Ditto.
8260 * config/riscv/riscv.opt (riscv_zb_subext): New.
8262 2021-10-25 liuhongt <hongtao.liu@intel.com>
8265 * match.pd: Simplify (_Float16) sqrtf((float) a) to .SQRT(a)
8266 when direct_internal_fn_supported_p, similar for sqrt/sqrtl.
8268 2021-10-25 Richard Biener <rguenther@suse.de>
8270 PR tree-optimization/102920
8271 * tree-ssa-sccvn.h (expressions_equal_p): Add argument
8272 controlling VN_TOP matching behavior.
8273 * tree-ssa-sccvn.c (expressions_equal_p): Likewise.
8274 (vn_phi_eq): Do not optimistically match VN_TOP.
8276 2021-10-25 konglin1 <lingling.kong@intel.com>
8278 * config/i386/sse.md (fma_<mode>_fadd_fmul): Add new
8279 define_insn_and_split.
8280 (fma_<mode>_fadd_fcmul):Likewise
8281 (fma_<complexopname>_<mode>_fma_zero):Likewise
8283 2021-10-24 John David Anglin <danglin@gcc.gnu.org>
8285 * config/pa/pa-d.c (pa_d_handle_target_float_abi): Don't check
8286 TARGET_DISABLE_FPREGS.
8287 * config/pa/pa.c (fix_range): Use MASK_SOFT_FLOAT instead of
8288 MASK_DISABLE_FPREGS.
8289 (hppa_rtx_costs): Don't check TARGET_DISABLE_FPREGS. Adjust
8290 cost of hardware integer multiplication.
8291 (pa_conditional_register_usage): Don't check TARGET_DISABLE_FPREGS.
8292 * config/pa/pa.h (INT14_OK_STRICT): Likewise.
8293 * config/pa/pa.md: Don't check TARGET_DISABLE_FPREGS. Check
8294 TARGET_SOFT_FLOAT in patterns that use xmpyu instruction.
8295 * config/pa/pa.opt (mdisable-fpregs): Change target mask to
8296 SOFT_FLOAT. Revise comment.
8297 (msoft-float): New option.
8299 2021-10-24 John David Anglin <danglin@gcc.gnu.org>
8301 * config/pa/pa.md: Don't use 'G' constraint in integer move patterns.
8303 2021-10-24 Gerald Pfeifer <gerald@pfeifer.com>
8305 * doc/install.texi (Specific): Remove obsolete details
8306 around GNU/Linux on Itanium.
8307 (Specific): Remove reference to Windows for Itanium.
8309 2021-10-23 Bernhard Reutner-Fischer <aldot@gcc.gnu.org>
8311 * config/i386/x86-tune-sched-bd.c (dispatch_group): Commentary
8314 2021-10-23 Jan Hubicka <hubicka@ucw.cz>
8316 * tree-ssa-structalias.c (compute_points_to_sets): Cleanup.
8318 2021-10-23 Roger Sayle <roger@nextmovesoftware.com>
8320 * config/i386/sse.md (<any_logic>v1ti3): New define_insn to
8321 implement V1TImode AND, IOR and XOR on TARGET_SSE2 (and above).
8322 (one_cmplv1ti2): New define expand.
8324 2021-10-22 Eric Gallager <egallager@gcc.gnu.org>
8327 * Makefile.in: Handle dvidir and install-dvi target.
8328 * configure: Regenerate.
8329 * configure.ac: Add install-dvi to target_list.
8331 2021-10-22 Gerald Pfeifer <gerald@pfeifer.com>
8333 * doc/install.texi (Binaries): Convert mingw-w64.org to https.
8336 2021-10-22 Richard Biener <rguenther@suse.de>
8338 PR tree-optimization/102893
8339 * tree-ssa-dce.c (find_obviously_necessary_stmts): Fix the
8340 test for an exit edge.
8342 2021-10-22 Aldy Hernandez <aldyh@redhat.com>
8343 Andrew MacLeod <amacleod@redhat.com>
8345 * gimple-range-path.cc (path_range_query::compute_phi_relations):
8346 Kill any global relations we may know before registering a new
8348 * value-relation.cc (path_oracle::killing_def): New.
8349 * value-relation.h (path_oracle::killing_def): New.
8351 2021-10-22 Richard Biener <rguenther@suse.de>
8354 * tree-ssa-sccvn.c (vn_phi_insert): For undefined SSA args
8356 (vn_phi_lookup): Likewise.
8358 2021-10-21 H.J. Lu <hjl.tools@gmail.com>
8361 * doc/invoke.texi: Document -fcf-protection requires i686 or
8364 2021-10-21 Eric Botcazou <ebotcazou@adacore.com>
8366 PR middle-end/102764
8367 * cfgexpand.c (expand_gimple_basic_block): Robustify latest change.
8369 2021-10-21 Jonathan Wright <jonathan.wright@arm.com>
8371 * config/aarch64/arm_neon.h (__STRUCTN): Delete function
8372 macro and all invocations.
8374 2021-10-21 Andrew MacLeod <amacleod@redhat.com>
8376 * doc/invoke.texi (ranger-debug): Document.
8377 * flag-types.h (enum ranger_debug): New.
8378 (enum evrp_mode): Remove debug values.
8379 * gimple-range-cache.cc (DEBUG_RANGE_CACHE): Use new debug flag.
8380 * gimple-range-gori.cc (gori_compute::gori_compute): Ditto.
8381 * gimple-range.cc (gimple_ranger::gimple_ranger): Ditto.
8382 * gimple-ssa-evrp.c (hybrid_folder::choose_value): Ditto.
8383 (execute_early_vrp): Use evrp-mode directly.
8384 * params.opt (enum evrp_mode): Remove debug values.
8385 (ranger-debug): New.
8386 (ranger-logical-depth): Relocate to be in alphabetical order.
8388 2021-10-21 Andrew MacLeod <amacleod@redhat.com>
8390 * doc/invoke.texi: (vrp1-mode, vrp2-mode): Document.
8391 * flag-types.h: (enum vrp_mode): New.
8392 * params.opt: (vrp1-mode, vrp2-mode): New.
8393 * tree-vrp.c (vrp_pass_num): New.
8394 (pass_vrp::pass_vrp): Set pass number.
8395 (pass_vrp::execute): Choose which VRP mode to execute.
8397 2021-10-21 Andrew MacLeod <amacleod@redhat.com>
8399 * gimple-ssa-evrp.c (class rvrp_folder): Move to tree-vrp.c.
8400 (execute_early_vrp): For ranger only mode, invoke ranger_vrp.
8401 * tree-vrp.c (class rvrp_folder): Relocate here.
8402 (execute_ranger_vrp): New.
8403 * tree-vrp.h (execute_ranger_vrp): Export.
8405 2021-10-21 Martin Liska <mliska@suse.cz>
8409 * opts.c (finish_options): Process flag_var_tracking* options
8410 here as they can be adjusted by optimize attribute.
8411 Process also flag_syntax_only and flag_gtoggle.
8412 * toplev.c (process_options): Remove it here.
8413 * common.opt: Make debug_nonbind_markers_p as PerFunction
8414 attribute as it depends on optimization level.
8416 2021-10-21 Martin Jambor <mjambor@suse.cz>
8418 PR tree-optimization/102505
8419 * tree-sra.c (totally_scalarize_subtree): Check that the
8420 encountered field fits within the acces we would like to put it
8423 2021-10-21 Aldy Hernandez <aldyh@redhat.com>
8425 * tree-ssa-threadbackward.c
8426 (back_threader::maybe_register_path): Remove circular paths check.
8428 2021-10-21 Richard Biener <rguenther@suse.de>
8430 * toplev.c (process_options): Move the initial debug_hooks
8432 (toplev::main): ... before the call of the post_options
8435 2021-10-21 Richard Biener <rguenther@suse.de>
8437 PR tree-optimization/102847
8438 * tree-vect-stmts.c (vect_model_load_cost): Add the scalar
8439 load cost in the prologue for VMAT_INVARIANT.
8441 2021-10-21 Richard Biener <rguenther@suse.de>
8443 PR tree-optimization/102847
8444 * tree-vect-stmts.c (vect_model_load_cost): Explicitely
8445 handle VMAT_INVARIANT as a splat in the prologue.
8447 2021-10-21 Hongyu Wang <hongyu.wang@intel.com>
8450 * config/i386/i386.c (ix86_get_ssemov): Adjust HFmode vector
8451 move to use the same logic as HImode.
8453 2021-10-21 Richard Biener <rguenther@suse.de>
8455 * tree-vect-slp.c (vect_build_slp_tree_1): Remove
8456 superfluous gimple_call_nothrow_p check.
8458 2021-10-21 Andrew Pinski <apinski@marvell.com>
8460 * tree-cfg.c (maybe_remove_writeonly_store): Add dce_ssa_names argument.
8461 Mark the ssa-name of the rhs as one to be removed.
8462 (execute_fixup_cfg): Update call to maybe_remove_writeonly_store.
8463 Call simple_dce_from_worklist at the end to a simple dce.
8465 2021-10-21 Andrew Pinski <apinski@marvell.com>
8467 * tree-cfg.c (maybe_remove_writeonly_store): New function
8468 factored out from ...
8469 (execute_fixup_cfg): Here. Call maybe_remove_writeonly_store.
8471 2021-10-21 Andrew Pinski <apinski@marvell.com>
8473 * tree-cfg.c (execute_fixup_cfg): Remove comment
8474 about standalone pass.
8476 2021-10-21 Andrew Pinski <apinski@marvell.com>
8478 * tree-cfg.c (execute_fixup_cfg): Output when the statement
8479 is removed when it is a write only var.
8481 2021-10-21 Aldy Hernandez <aldyh@redhat.com>
8483 * tree-ssa-threadbackward.c (back_threader::maybe_register_path):
8484 Avoid threading circular paths.
8486 2021-10-20 Alex Coplan <alex.coplan@arm.com>
8488 * calls.c (initialize_argument_information): Remove some dead
8489 code, remove handling for function_arg returning const_int.
8490 * doc/tm.texi: Delete documentation for unused target hooks.
8491 * doc/tm.texi.in: Likewise.
8492 * target.def (load_bounds_for_arg): Delete.
8493 (store_bounds_for_arg): Delete.
8494 (load_returned_bounds): Delete.
8495 (store_returned_bounds): Delete.
8496 * targhooks.c (default_load_bounds_for_arg): Delete.
8497 (default_store_bounds_for_arg): Delete.
8498 (default_load_returned_bounds): Delete.
8499 (default_store_returned_bounds): Delete.
8500 * targhooks.h (default_load_bounds_for_arg): Delete.
8501 (default_store_bounds_for_arg): Delete.
8502 (default_load_returned_bounds): Delete.
8503 (default_store_returned_bounds): Delete.
8505 2021-10-20 Jonathan Wakely <jwakely@redhat.com>
8507 * doc/extend.texi (Basic Asm): Clarify that asm is not an
8509 * doc/invoke.texi (-fno-asm): Fix description for C++.
8511 2021-10-20 Jonathan Wakely <jwakely@redhat.com>
8513 * doc/install.texi: Remove link to old.html
8515 2021-10-20 Tamar Christina <tamar.christina@arm.com>
8517 * config/aarch64/aarch64-simd.md (*aarch64_cmtst_same_<mode>): New.
8519 2021-10-20 Tamar Christina <tamar.christina@arm.com>
8521 * config/aarch64/aarch64-simd.md (*aarch64_narrow_trunc<mode>): New.
8523 2021-10-20 Tamar Christina <tamar.christina@arm.com>
8525 * config/aarch64/aarch64-simd.md (aarch64_simd_ashr<mode>): Add case cmp
8527 * config/aarch64/constraints.md (D1): New.
8529 2021-10-20 Tamar Christina <tamar.christina@arm.com>
8531 * config/aarch64/aarch64-simd.md
8532 (*aarch64_<srn_op>topbits_shuffle<mode>_le): New.
8533 (*aarch64_topbits_shuffle<mode>_le): New.
8534 (*aarch64_<srn_op>topbits_shuffle<mode>_be): New.
8535 (*aarch64_topbits_shuffle<mode>_be): New.
8536 * config/aarch64/predicates.md
8537 (aarch64_simd_shift_imm_vec_exact_top): New.
8539 2021-10-20 Tamar Christina <tamar.christina@arm.com>
8541 * config/aarch64/aarch64-simd.md (*aarch64_<srn_op>shrn<mode>_vect,
8542 *aarch64_<srn_op>shrn<mode>2_vect_le,
8543 *aarch64_<srn_op>shrn<mode>2_vect_be): New.
8544 * config/aarch64/iterators.md (srn_op): New.
8546 2021-10-20 Chung-Lin Tang <cltang@codesourcery.com>
8548 * omp-low.c (omp_copy_decl_2): For !ctx, use record_vars to add new copy
8550 (scan_sharing_clauses): Place copy of OMP_CLAUSE_IN_REDUCTION decl in
8551 ctx->outer instead of ctx.
8553 2021-10-20 Martin Liska <mliska@suse.cz>
8556 2021-10-19 Martin Liska <mliska@suse.cz>
8559 * config/i386/i386-options.c (ix86_valid_target_attribute_inner_p): Strip whitespaces.
8560 * system.h (strip_whilespaces): New function.
8562 2021-10-20 Martin Liska <mliska@suse.cz>
8565 2021-10-19 Martin Liska <mliska@suse.cz>
8568 * config/aarch64/aarch64.c (aarch64_process_one_target_attr):
8571 2021-10-20 Wilco Dijkstra <wdijkstr@arm.com>
8573 * config/aarch64/aarch64.c (aarch64_case_values_threshold):
8574 Change to 8 with -Os, 11 otherwise.
8576 2021-10-20 Wilco Dijkstra <wdijkstr@arm.com>
8578 * config/aarch64/aarch64.c (neoversev1_tunings):
8579 Enable AARCH64_EXTRA_TUNE_CHEAP_SHIFT_EXTEND.
8580 (neoversen2_tunings): Likewise.
8582 2021-10-20 Wilco Dijkstra <wdijkstr@arm.com>
8585 * config/aarch64/aarch64.md (frint_pattern): Update comment.
8586 * config/aarch64/aarch64-simd-builtins.def: Change frintn to roundeven.
8587 * config/aarch64/arm_fp16.h: Change frintn to roundeven.
8588 * config/aarch64/arm_neon.h: Likewise.
8589 * config/aarch64/iterators.md (frint_pattern): Use roundeven for FRINTN.
8591 2021-10-20 Martin Liska <mliska@suse.cz>
8593 * config/arm/arm.c (arm_unwind_emit_sequence): Do not declare
8594 already declared global variable.
8595 (arm_unwind_emit_set): Use out_file as function argument.
8596 (arm_unwind_emit): Likewise.
8597 * config/darwin.c (machopic_output_data_section_indirection): Likewise.
8598 (machopic_output_stub_indirection): Likewise.
8599 (machopic_output_indirection): Likewise.
8600 (machopic_finish): Likewise.
8601 * config/i386/i386.c (ix86_asm_output_function_label): Likewise.
8602 * config/i386/winnt.c (i386_pe_seh_unwind_emit): Likewise.
8603 * config/ia64/ia64.c (process_epilogue): Likewise.
8604 (process_cfa_adjust_cfa): Likewise.
8605 (process_cfa_register): Likewise.
8606 (process_cfa_offset): Likewise.
8607 (ia64_asm_unwind_emit): Likewise.
8608 * config/s390/s390.c (s390_asm_output_function_label): Likewise.
8610 2021-10-20 Andre Simoes Dias Vieira <andre.simoesdiasvieira@arm.com>
8612 * config/aarch64/aarch64-builtins.c
8613 (aarch64_general_gimple_fold_builtin):
8614 lower vld1 and vst1 variants of the neon builtins
8615 * config/aarch64/aarch64-protos.h:
8616 (aarch64_general_gimple_fold_builtin): Add gsi parameter.
8617 * config/aarch64/aarch64.c (aarch64_general_gimple_fold_builtin):
8620 2021-10-20 Andre Simoes Dias Vieira <andre.simoesdiasvieira@arm.com>
8622 * match.pd: Generate IFN_TRUNC.
8624 2021-10-20 Richard Biener <rguenther@suse.de>
8626 PR tree-optimization/102853
8627 * tree-data-ref.c (split_constant_offset_1): Bail out
8628 immediately if the expression traps on overflow.
8630 2021-10-20 Aldy Hernandez <aldyh@redhat.com>
8632 * tree-ssa-threadbackward.c (back_threader::~back_threader): Remove.
8634 2021-10-20 Aldy Hernandez <aldyh@redhat.com>
8636 * tree-ssa-threadupdate.c (back_jt_path_registry::adjust_paths_after_duplication):
8637 Remove superflous debugging message.
8638 (back_jt_path_registry::duplicate_thread_path): Same.
8640 2021-10-20 Aldy Hernandez <aldyh@redhat.com>
8642 * tree-ssa-threadbackward.c (back_threader_registry::back_threader_registry):
8644 (back_threader_registry::register_path): Remove m_threaded_paths.
8646 2021-10-20 Aldy Hernandez <aldyh@redhat.com>
8648 PR tree-optimization/102814
8649 * doc/invoke.texi: Document --param=max-fsm-thread-length.
8650 * params.opt: Add --param=max-fsm-thread-length.
8651 * tree-ssa-threadbackward.c
8652 (back_threader_profitability::profitable_path_p): Fail on paths
8653 longer than max-fsm-thread-length.
8655 2021-10-20 Eric Botcazou <ebotcazou@adacore.com>
8657 PR middle-end/102764
8658 * cfgexpand.c (expand_gimple_basic_block): Disregard a final debug
8659 statement to reset the current location for the outgoing edges.
8661 2021-10-20 Aldy Hernandez <aldyh@redhat.com>
8662 Richard Biener <rguenther@suse.de>
8664 * tree-ssa-threadupdate.c (cancel_thread): Dump threading reason
8665 on the same line as the threading cancellation.
8666 (jt_path_registry::cancel_invalid_paths): Avoid rotating loops.
8667 Avoid threading through loop headers where the path remains in the
8670 2021-10-20 Siddhesh Poyarekar <siddhesh@gotplt.org>
8672 * tree-object-size.c (unknown): Make into a function. Adjust
8674 (unknown_object_size): Simplify implementation.
8676 2021-10-20 Hongtao Liu <hongtao.liu@intel.com>
8677 Kewen Lin <linkw@linux.ibm.com>
8679 * doc/sourcebuild.texi (Effective-Target Keywords): Document
8680 vect_slp_v2qi_store, vect_slp_v4qi_store, vect_slp_v8qi_store,
8681 vect_slp_v16qi_store, vect_slp_v2hi_store,
8682 vect_slp_v4hi_store, vect_slp_v2si_store, vect_slp_v4si_store.
8684 2021-10-19 Jonathan Wakely <jwakely@redhat.com>
8686 * doc/extend.texi (Basic PowerPC Built-in Functions): Fix typo.
8688 2021-10-19 Paul A. Clarke <pc@us.ibm.com>
8692 * config/rs6000/emmintrin.h: Guard POWER8 intrinsics.
8693 * config/rs6000/pmmintrin.h: Same.
8694 * config/rs6000/smmintrin.h: Same.
8695 * config/rs6000/tmmintrin.h: Same.
8697 2021-10-19 Paul A. Clarke <pc@us.ibm.com>
8699 * config.gcc (extra_headers): Add nmmintrin.h.
8701 2021-10-19 Richard Biener <rguenther@suse.de>
8703 * tree-vectorizer.h (vect_supportable_dr_alignment): Add
8704 misalignment parameter.
8705 * tree-vect-data-refs.c (vect_get_peeling_costs_all_drs):
8706 Do not change DR_MISALIGNMENT in place, instead pass the
8707 adjusted misalignment to vect_supportable_dr_alignment.
8708 (vect_peeling_supportable): Likewise.
8709 (vect_peeling_hash_get_lowest_cost): Adjust.
8710 (vect_enhance_data_refs_alignment): Likewise.
8711 (vect_vfa_access_size): Likewise.
8712 (vect_supportable_dr_alignment): Add misalignment
8713 parameter and simplify.
8714 * tree-vect-stmts.c (get_negative_load_store_type): Adjust.
8715 (get_group_load_store_type): Likewise.
8716 (get_load_store_type): Likewise.
8718 2021-10-19 Clément Chigot <clement.chigot@atos.net>
8720 * config/rs6000/rs6000.c (rs6000_xcoff_file_end): Move
8721 __tls_get_addr reference to .text csect.
8723 2021-10-19 Martin Liska <mliska@suse.cz>
8726 * config/aarch64/aarch64.c (aarch64_process_one_target_attr):
8729 2021-10-19 Richard Biener <rguenther@suse.de>
8731 * tree-vectorizer.h (vect_get_store_cost): Adjust signature.
8732 (vect_get_load_cost): Likewise.
8733 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
8734 alignment support scheme and misalignment as arguments
8736 (vect_get_peeling_costs_all_drs): Compute that info here
8737 and note that we shouldn't need to.
8738 * tree-vect-stmts.c (vect_model_store_cost): Get
8739 alignment support scheme and misalignment as arguments.
8740 (vect_get_store_cost): Likewise.
8741 (vect_model_load_cost): Likewise.
8742 (vect_get_load_cost): Likewise.
8743 (vectorizable_store): Pass down alignment support scheme
8744 and misalignment to costing.
8745 (vectorizable_load): Likewise.
8747 2021-10-19 Richard Biener <rguenther@suse.de>
8749 * tree-vect-stmts.c (get_negative_load_store_type): Add
8750 offset output parameter and initialize it.
8751 (get_group_load_store_type): Likewise.
8752 (get_load_store_type): Likewise.
8753 (vectorizable_store): Use offset as computed by
8754 get_load_store_type.
8755 (vectorizable_load): Likewise.
8757 2021-10-19 Richard Biener <rguenther@suse.de>
8759 PR tree-optimization/102827
8760 * tree-if-conv.c (predicate_statements): Add pe parameter
8761 and use that edge to insert invariant stmts on.
8762 (combine_blocks): Pass through pe.
8763 (tree_if_conversion): Compute the edge to insert invariant
8764 stmts on and pass it along.
8766 2021-10-19 Roger Sayle <roger@nextmovesoftware.com>
8769 * config/bfin/bfin.md (addsubv2hi3, subaddv2hi3, ssaddsubv2hi3,
8770 sssubaddv2hi3): Swap the order of operators in vec_concat.
8772 2021-10-19 Xionghu Luo <luoxhu@linux.ibm.com>
8774 * config/rs6000/altivec.md (*altivec_vmrghb_internal): Delete.
8775 (altivec_vmrghb_direct): New.
8776 (*altivec_vmrghh_internal): Delete.
8777 (altivec_vmrghh_direct): New.
8778 (*altivec_vmrghw_internal): Delete.
8779 (altivec_vmrghw_direct_<mode>): New.
8780 (altivec_vmrghw_direct): Delete.
8781 (*altivec_vmrglb_internal): Delete.
8782 (altivec_vmrglb_direct): New.
8783 (*altivec_vmrglh_internal): Delete.
8784 (altivec_vmrglh_direct): New.
8785 (*altivec_vmrglw_internal): Delete.
8786 (altivec_vmrglw_direct_<mode>): New.
8787 (altivec_vmrglw_direct): Delete.
8788 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Adjust.
8789 * config/rs6000/rs6000.c (altivec_expand_vec_perm_const):
8791 * config/rs6000/vsx.md (vsx_xxmrghw_<mode>): Adjust.
8792 (vsx_xxmrglw_<mode>): Adjust.
8794 2021-10-19 Aldy Hernandez <aldyh@redhat.com>
8796 * passes.def: Change threading comment before pass_ccp pass.
8798 2021-10-19 Haochen Gui <guihaoc@gcc.gnu.org>
8800 * config/rs6000/rs6000-call.c (altivec_expand_lxvr_builtin):
8801 Modify the expansion for sign extension. All extensions are done
8802 within VSX registers.
8804 2021-10-19 Richard Biener <rguenther@suse.de>
8806 * tree-vect-stmts.c (get_group_load_store_type): Add
8807 misalignment output parameter and initialize it.
8808 (get_group_load_store_type): Likewise.
8809 (vectorizable_store): Remove now redundant queries.
8810 (vectorizable_load): Likewise.
8812 2021-10-19 Richard Biener <rguenther@suse.de>
8814 * tree-vectorizer.h (vect_supportable_dr_alignment): Remove
8815 check_aligned argument.
8816 * tree-vect-data-refs.c (vect_supportable_dr_alignment):
8818 (vect_peeling_hash_insert): Add supportable_if_not_aligned
8819 argument and do not call vect_supportable_dr_alignment here.
8820 (vect_peeling_supportable): Adjust.
8821 (vect_enhance_data_refs_alignment): Compute whether the
8822 access is supported with different alignment here and
8823 pass that down to vect_peeling_hash_insert.
8824 (vect_vfa_access_size): Adjust.
8825 * tree-vect-stmts.c (vect_get_store_cost): Likewise.
8826 (vect_get_load_cost): Likewise.
8827 (get_negative_load_store_type): Likewise.
8828 (get_group_load_store_type): Likewise.
8829 (get_load_store_type): Likewise.
8831 2021-10-19 Martin Liska <mliska@suse.cz>
8834 * config/i386/i386-options.c (ix86_valid_target_attribute_inner_p): Strip whitespaces.
8835 * system.h (strip_whilespaces): New function.
8837 2021-10-19 dianhong xu <dianhong.xu@intel.com>
8839 * config/i386/avx512fp16intrin.h:
8840 (_mm512_set1_pch): New intrinsic.
8841 * config/i386/avx512fp16vlintrin.h:
8842 (_mm256_set1_pch): New intrinsic.
8843 (_mm_set1_pch): Ditto.
8845 2021-10-18 Andrew MacLeod <amacleod@redhat.com>
8847 PR tree-optimization/102796
8848 * gimple-range.cc (gimple_ranger::range_on_edge): Process EH edges
8849 normally. Return get_tree_range for non gimple_range_ssa_p names.
8850 (gimple_ranger::range_of_stmt): Use get_tree_range for non
8851 gimple_range_ssa_p names.
8853 2021-10-18 Uroš Bizjak <ubizjak@gmail.com>
8856 * config/i386/i386.c (ix86_print_operand_address):
8857 Error out for non-address_operand asm operands.
8859 2021-10-18 Richard Biener <rguenther@suse.de>
8861 * tree-vect-data-refs.c (vect_peeling_hash_insert): Do
8862 not auto-convert dr_alignment_support to bool.
8863 (vect_peeling_supportable): Likewise.
8864 (vect_enhance_data_refs_alignment): Likewise.
8865 (vect_supportable_dr_alignment): Commonize read/write case.
8866 * tree-vect-stmts.c (vect_get_store_cost): Use
8867 dr_alignment_support, not int, for the vect_supportable_dr_alignment
8869 (vect_get_load_cost): Likewise.
8871 2021-10-18 Siddhesh Poyarekar <siddhesh@gotplt.org>
8873 * tree-object-size.c (object_sizes_execute): Consolidate LHS
8874 null check and do it early.
8876 2021-10-18 Richard Biener <rguenther@suse.de>
8878 * tree-vect-stmts.c (vectorizable_store): Use the
8879 computed alignment scheme instead of querying
8882 2021-10-18 Richard Biener <rguenther@suse.de>
8884 * tree-vect-stmts.c (vectorizable_store): Do not recompute
8885 alignment scheme already determined by get_load_store_type.
8887 2021-10-18 Aldy Hernandez <aldyh@redhat.com>
8889 * tree-ssa-threadbackward.c (class pass_thread_jumps_full):
8890 Clone corresponding pass.
8892 2021-10-18 Roger Sayle <roger@nextmovesoftware.com>
8894 * combine.c (recog_for_combine): For an unrecognized move/set of
8895 a constant, try force_const_mem to place it in the constant pool.
8896 * cse.c (constant_pool_entries_cost, constant_pool_entries_regcost):
8897 Delete global variables (that are no longer assigned a cost value).
8898 (cse_insn): Simplify logic for deciding whether to place a folded
8899 constant in the constant pool using force_const_mem.
8900 (cse_main): Remove zero initialization of constant_pool_entries_cost
8901 and constant_pool_entries_regcost.
8902 * config/i386/i386.c (ix86_rtx_costs): Make memory accesses
8903 fractionally more expensive, when optimizing for speed.
8905 2021-10-18 Martin Liska <mliska@suse.cz>
8907 PR gcov-profile/102746
8908 PR gcov-profile/102747
8909 * gcov.c (main): Return return_code.
8910 (output_gcov_file): Mark return_code when error happens.
8911 (generate_results): Likewise.
8912 (read_graph_file): Likewise.
8913 (read_count_file): Likewise.
8915 2021-10-18 Roger Sayle <roger@nextmovesoftware.com>
8917 * config/bfin/bfin.md (define_constants): Remove UNSPEC_ONES.
8918 (define_insn "ones"): Replace UNSPEC_ONES with a truncate of
8919 a popcount, allowing compile-time evaluation/simplification.
8920 (popcountsi2, popcounthi2): New expanders using a "ones" insn.
8922 2021-10-18 Richard Biener <rguenther@suse.de>
8924 PR tree-optimization/102788
8925 * tree-vect-patterns.c (vect_init_pattern_stmt): Allow
8927 (vect_pattern_recog_1): Likewise.
8928 (vect_recog_bool_pattern): Continue matching the pattern
8929 even if we do not have a vector type for a conversion
8932 2021-10-18 Roger Sayle <roger@nextmovesoftware.com>
8934 * simplify-rtx.c (simplify_const_unary_operation) [SS_NEG, SS_ABS]:
8935 Evalute SS_NEG and SS_ABS of a constant argument.
8937 2021-10-18 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
8940 * gimple-match-head.c (try_conditional_simplification): Add case for single operand.
8941 * internal-fn.def: Add entry for COND_NEG internal function.
8942 * internal-fn.c (FOR_EACH_CODE_MAPPING): Add entry for
8943 NEGATE_EXPR, COND_NEG mapping.
8944 * optabs.def: Add entry for cond_neg_optab.
8945 * match.pd (UNCOND_UNARY, COND_UNARY): New operator lists.
8946 (vec_cond COND (foo A) B) -> (IFN_COND_FOO COND A B): New pattern.
8947 (vec_cond COND B (foo A)) -> (IFN_COND_FOO ~COND A B): Likewise.
8949 2021-10-18 Aldy Hernandez <aldyh@redhat.com>
8951 * tree-ssa-strlen.c (class strlen_pass): Rename from
8953 (handle_builtin_stxncpy_strncat): Move to strlen_pass.
8954 (handle_assign): Same.
8955 (adjust_last_stmt): Same.
8956 (maybe_warn_overflow): Same.
8957 (handle_builtin_strlen): Same.
8958 (handle_builtin_strchr): Same.
8959 (handle_builtin_strcpy): Same.
8960 (handle_builtin_strncat): Same.
8961 (handle_builtin_stxncpy_strncat): Same.
8962 (handle_builtin_memcpy): Same.
8963 (handle_builtin_strcat): Same.
8964 (handle_alloc_call): Same.
8965 (handle_builtin_memset): Same.
8966 (handle_builtin_memcmp): Same.
8967 (get_len_or_size): Same.
8968 (strxcmp_eqz_result): Same.
8969 (handle_builtin_string_cmp): Same.
8970 (handle_pointer_plus): Same.
8971 (count_nonzero_bytes_addr): Same.
8972 (count_nonzero_bytes): Same.
8973 (handle_store): Same.
8974 (strlen_check_and_optimize_call): Same.
8975 (handle_integral_assign): Same.
8976 (check_and_optimize_stmt): Same.
8977 (printf_strlen_execute): Rename strlen_dom_walker to strlen_pass.
8979 2021-10-18 Richard Biener <rguenther@suse.de>
8981 PR tree-optimization/102798
8982 * tree-vect-data-refs.c (vect_create_addr_base_for_vector_ref):
8983 Only copy points-to info to newly generated SSA names.
8985 2021-10-18 Martin Liska <mliska@suse.cz>
8987 * dbgcnt.c (dbg_cnt_process_opt): Remove unused but set variable.
8988 * gcov.c (get_cycles_count): Likewise.
8989 * lto-compress.c (lto_compression_zlib): Likewise.
8990 (lto_uncompression_zlib): Likewise.
8991 * targhooks.c (default_pch_valid_p): Likewise.
8993 2021-10-17 Aldy Hernandez <aldyh@redhat.com>
8995 * tree-pass.h (make_pass_thread_jumps_full): New.
8996 * tree-ssa-threadbackward.c (pass_thread_jumps::gate): Inline.
8997 (try_thread_blocks): Add resolve and speed arguments.
8998 (pass_thread_jumps::execute): Inline.
8999 (do_early_thread_jumps): New.
9000 (do_thread_jumps): New.
9001 (make_pass_thread_jumps): Move.
9002 (pass_early_thread_jumps::gate): Inline.
9003 (pass_early_thread_jumps::execute): Inline.
9004 (class pass_thread_jumps_full): New.
9006 2021-10-16 Piotr Kubaj <pkubaj@FreeBSD.org>
9008 * configure.ac: Treat powerpc64*-*-freebsd* the same as
9009 powerpc64-*-freebsd*.
9010 * configure: Regenerate.
9012 2021-10-16 H.J. Lu <hjl.tools@gmail.com>
9014 * value-query.cc (get_ssa_name_ptr_info_nonnull): Change
9015 set_ptr_nonull to set_ptr_nonnull in comments.
9017 2021-10-16 Jan Hubicka <hubicka@ucw.cz>
9019 PR tree-optimization/102720
9020 * tree-ssa-structalias.c (compute_points_to_sets): Fix producing
9021 of call used and clobbered sets.
9023 2021-10-15 Iain Sandoe <iain@sandoe.co.uk>
9025 * config/darwin.h (LINK_COMMAND_SPEC_A): Update 'r' handling to
9026 skip gomp and itm when r or nodefaultlibs is given.
9027 (DSYMUTIL_SPEC): Do not call dsymutil for '-r' link lines.
9028 Update ordering of exclusions, remove duplicate 'v' addition
9029 (collect2 will add this from the main command line).
9031 2021-10-15 Iain Sandoe <iain@sandoe.co.uk>
9033 * config/darwin-driver.c (darwin_driver_init): Revise comments, handle
9034 filelist and framework options in specs instead of code.
9035 * config/darwin.h (SUBTARGET_DRIVER_SELF_SPECS): Update to handle link
9036 specs that are really driver ones.
9037 (DARWIN_CC1_SPEC): Likewise.
9038 (CPP_SPEC): Likewise.
9039 (SYSROOT_SPEC): Append space.
9040 (LINK_SYSROOT_SPEC): Remove most driver link specs.
9041 (STANDARD_STARTFILE_PREFIX_2): Update link-related specs.
9042 (STARTFILE_SPEC): Likewise.
9043 (ASM_MMACOSX_VERSION_MIN_SPEC): Fix line wrap.
9044 (ASM_SPEC): Update driver-related specs.
9045 (ASM_FINAL_SPEC): Likewise.
9046 * config/darwin.opt: Remove now unused option aliases.
9047 * config/i386/darwin.h (EXTRA_ASM_OPTS): Ensure space after opt.
9048 (ASM_SPEC): Update driver-related specs.
9050 2021-10-15 Roger Sayle <roger@nextmovesoftware.com>
9052 * config/i386/i386.c (ix86_hardreg_mov_ok): For vector modes,
9053 allow standard_sse_constant_p immediate constants.
9055 2021-10-15 Iain Sandoe <iain@sandoe.co.uk>
9057 * config.gcc: Add tm-dwarf2.h to tm_d-file.
9059 2021-10-15 Andrew MacLeod <amacleod@redhat.com>
9061 * gimple-range-fold.h (gimple_range_ssa_p): Don't process names
9062 that occur in abnormal phis.
9063 * gimple-range.cc (gimple_ranger::range_on_edge): Return false for
9064 abnormal and EH edges.
9065 * gimple-ssa-evrp.c (rvrp_folder::value_of_expr): Ditto.
9066 (rvrp_folder::value_on_edge): Ditto.
9067 (rvrp_folder::value_of_stmt): Ditto.
9068 (hybrid_folder::value_of_expr): Ditto for ranger queries.
9069 (hybrid_folder::value_on_edge): Ditto.
9070 (hybrid_folder::value_of_stmt): Ditto.
9071 * value-query.cc (gimple_range_global): Always return a range if
9072 the type is supported.
9074 2021-10-15 John David Anglin <danglin@gcc.gnu.org>
9076 * config/pa/pa.md: Consistently use "rG" constraint for copy
9077 instruction in move patterns.
9079 2021-10-15 Andrew Stubbs <ams@codesourcery.com>
9081 * config/gcn/gcn-hsa.h (S_FIJI): Set unconditionally.
9084 * config/gcn/gcn.c: Hard code SRAM ECC settings for old architectures.
9085 * config/gcn/mkoffload.c (ELFABIVERSION_AMDGPU_HSA): Rename to ...
9086 (ELFABIVERSION_AMDGPU_HSA_V3): ... this.
9087 (ELFABIVERSION_AMDGPU_HSA_V4): New.
9088 (SET_SRAM_ECC_UNSUPPORTED): New.
9089 (copy_early_debug_info): Create elf flags to match the other objects.
9090 (main): Just let the attribute flags pass through.
9092 2021-10-15 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
9094 * tree-loop-distribution.c (reduction_var_overflows_first):
9095 Pass the type of reduction_var as first argument as it is also
9096 done for the load type.
9097 (loop_distribution::transform_reduction_loop): Add missing
9098 TREE_TYPE while determining precission of reduction_var.
9100 2021-10-15 Richard Biener <rguenther@suse.de>
9102 * defaults.h (PREFERRED_DEBUGGING_TYPE): Choose DWARF2_DEBUG
9104 * toplev.c (process_options): Warn when STABS debugging is
9105 enabled but not the preferred format.
9106 * config/pa/som.h (PREFERRED_DEBUGGING_TYPE): Define to
9108 * config/pdp11/pdp11.h (PREFERRED_DEBUGGING_TYPE): Likewise.
9110 2021-10-15 Richard Biener <rguenther@suse.de>
9113 * tree-inline.c (copy_bb): Avoid underflowing nargs.
9115 2021-10-15 Hongyu Wang <hongyu.wang@intel.com>
9117 * config/i386/i386-expand.c (ix86_expand_vec_perm): Convert
9118 HFmode input operand to HImode.
9119 (ix86_vectorize_vec_perm_const): Likewise.
9120 * config/i386/sse.md (*avx512bw_permvar_truncv16siv16hi_1_hf):
9122 (*avx512f_permvar_truncv8siv8hi_1_hf):
9125 2021-10-15 Richard Biener <rguenther@suse.de>
9127 PR middle-end/102682
9128 * expmed.c (store_bit_field_1): Ensure a LHS subreg would
9129 not create a paradoxical subreg.
9131 2021-10-15 Hongyu Wang <hongyu.wang@intel.com>
9133 * config/i386/i386-expand.c (ix86_expand_vector_init):
9134 For half_vector concat for HFmode, handle them like HImode.
9136 2021-10-14 Aldy Hernandez <aldyh@redhat.com>
9138 * tree-ssa-threadbackward.c (class back_threader): Add m_resolve.
9139 (back_threader::back_threader): Same.
9140 (back_threader::resolve_phi): Try to solve without looking back if
9142 (back_threader::find_paths_to_names): Same.
9143 (try_thread_blocks): Pass resolve argument to back threader.
9144 (pass_early_thread_jumps::execute): Same.
9146 2021-10-14 Aldy Hernandez <aldyh@redhat.com>
9148 * doc/invoke.texi: Remove max-fsm-thread-length,
9149 max-fsm-thread-paths, and fsm-maximum-phi-arguments.
9151 * tree-ssa-threadbackward.c (back_threader::back_threader): Remove
9153 (back_threader_registry::back_threader_registry): Same.
9154 (back_threader_profitability::profitable_path_p): Remove
9155 param_max_fsm_thread-length.
9156 (back_threader_registry::register_path): Remove
9157 m_max_allowable_paths.
9159 2021-10-14 Aldy Hernandez <aldyh@redhat.com>
9161 * tree-ssa-threadbackward.c (class back_threader): Make m_imports
9163 (back_threader::~back_threader): Do not release m_path.
9165 2021-10-14 Andrew MacLeod <amacleod@redhat.com>
9167 PR tree-optimization/102738
9168 * vr-values.c (simplify_using_ranges::simplify): Handle RSHIFT_EXPR.
9170 2021-10-14 Kwok Cheung Yeung <kcy@codesourcery.com>
9172 * omp-general.c (omp_check_context_selector): Move from c-omp.c.
9173 (omp_mark_declare_variant): Move from c-omp.c.
9174 (omp_context_name_list_prop): Update for Fortran strings.
9175 * omp-general.h (omp_check_context_selector): New prototype.
9176 (omp_mark_declare_variant): New prototype.
9178 2021-10-14 Richard Sandiford <richard.sandiford@arm.com>
9180 * config/rs6000/rs6000.c (rs6000_density_test): Move early
9181 exit test further up the function.
9183 2021-10-14 Richard Sandiford <richard.sandiford@arm.com>
9185 * config/arm/arm.c (arm_add_stmt_cost): Delete.
9186 (TARGET_VECTORIZE_ADD_STMT_COST): Delete.
9188 2021-10-14 Martin Jambor <mjambor@suse.cz>
9190 * doc/invoke.texi (Optimize Options): Add entry for
9191 ipa-cp-recursive-freq-factor.
9193 2021-10-14 Tamar Christina <tamar.christina@arm.com>
9195 * match.pd: New rule.
9197 2021-10-14 Jan Hubicka <hubicka@ucw.cz>
9200 * ipa-modref-tree.h (modref_access_node::update2):
9201 Also check that parm_offset is unchanged.
9202 (modref_ref_node::insert_access): Fix updating of
9205 2021-10-14 Aldy Hernandez <aldyh@redhat.com>
9207 * tree-ssa-threadbackward.c (back_threader::resolve_phi): Add
9210 2021-10-14 Richard Biener <rguenther@suse.de>
9212 PR tree-optimization/102659
9213 * tree-if-conv.c (if_convertible_gimple_assign_stmt_p): Also
9214 rewrite pointer typed undefined overflow operations.
9215 (predicate_statements): Likewise. Make sure to emit invariant
9216 conversions in the preheader.
9217 * tree-vectorizer.c (vect_loop_vectorized_call): Look through
9218 non-empty preheaders.
9219 * tree-data-ref.c (dr_analyze_indices): Strip useless
9220 conversions to the MEM_REF base type.
9222 2021-10-14 Martin Liska <mliska@suse.cz>
9224 * common.opt: Stop using AUTODETECT_VALUE
9225 and use EnabledBy where possible.
9226 * opts.c: Enable OPT_fvar_tracking with optimize >= 1.
9227 * toplev.c (AUTODETECT_VALUE): Remove macro.
9228 (process_options): Simplify by using EnabledBy and
9229 OPT_fvar_tracking. Use OPTION_SET_P macro instead of
9232 2021-10-14 Jonathan Wright <jonathan.wright@arm.com>
9234 * config/aarch64/arm_neon.h (vld1_s8_x3): Use signed type for
9236 (vld1_s32_x3): Likewise.
9238 2021-10-14 Aldy Hernandez <aldyh@redhat.com>
9240 PR tree-optimization/102736
9241 PR tree-optimization/102736
9242 * gimple-range-path.cc (path_range_query::range_on_path_entry):
9243 Assert that the requested range is defined outside the path.
9244 (path_range_query::ssa_range_in_phi): Do not call
9245 range_on_path_entry for SSA names that are defined within the
9248 2021-10-14 Iain Sandoe <iain@sandoe.co.uk>
9250 * config/darwin-driver.c (darwin_find_version_from_kernel):
9251 Quote internal identifiers and avoid contractions in
9253 (darwin_default_min_version): Likewise.
9254 (darwin_driver_init): Likewise.
9256 2021-10-14 Martin Jambor <mjambor@suse.cz>
9258 * params.opt (ipa-cp-recursive-freq-factor): New.
9259 * ipa-cp.c (ipcp_value): Switch to inline initialization. New members
9260 scc_no, self_recursion_generated_level, same_scc and
9261 self_recursion_generated_p.
9262 (ipcp_lattice::add_value): Replaced parameter unlimited with
9263 same_lat_gen_level, usit it determine limit of values and store it to
9265 (ipcp_lattice<valtype>::print): Dump the new fileds.
9266 (allocate_and_init_ipcp_value): Take same_lat_gen_level as a new
9267 parameter and store it to the new value.
9268 (self_recursively_generated_p): Removed.
9269 (propagate_vals_across_arith_jfunc): Use self_recursion_generated_p
9270 instead of self_recursively_generated_p, store self generation level
9272 (value_topo_info<valtype>::add_val): Set scc_no.
9273 (value_topo_info<valtype>::propagate_effects): Multiply frequencies of
9274 recursively feeding values and self generated values by appropriate
9277 2021-10-14 Richard Sandiford <richard.sandiford@arm.com>
9279 * config/aarch64/aarch64.c (aarch64_add_stmt_cost): Remove
9280 redundant test for flag_vect_cost_model.
9282 2021-10-14 Aldy Hernandez <aldyh@redhat.com>
9284 * bitmap.c (debug): New overloaded function for auto_bitmaps.
9285 * bitmap.h (debug): Same.
9287 2021-10-14 Hongyu Wang <hongyu.wang@intel.com>
9289 * config/i386/avx512fp16intrin.h (_mm512_mask_fcmadd_pch):
9290 Adjust builtin call.
9291 (_mm512_mask3_fcmadd_pch): Likewise.
9292 (_mm512_mask_fmadd_pch): Likewise
9293 (_mm512_mask3_fmadd_pch): Likewise
9294 (_mm512_mask_fcmadd_round_pch): Likewise
9295 (_mm512_mask3_fcmadd_round_pch): Likewise
9296 (_mm512_mask_fmadd_round_pch): Likewise
9297 (_mm512_mask3_fmadd_round_pch): Likewise
9298 (_mm_mask_fcmadd_sch): Likewise
9299 (_mm_mask3_fcmadd_sch): Likewise
9300 (_mm_mask_fmadd_sch): Likewise
9301 (_mm_mask3_fmadd_sch): Likewise
9302 (_mm_mask_fcmadd_round_sch): Likewise
9303 (_mm_mask3_fcmadd_round_sch): Likewise
9304 (_mm_mask_fmadd_round_sch): Likewise
9305 (_mm_mask3_fmadd_round_sch): Likewise
9306 (_mm_fcmadd_round_sch): Likewise
9307 * config/i386/avx512fp16vlintrin.h (_mm_mask_fmadd_pch):
9308 Adjust builtin call.
9309 (_mm_mask3_fmadd_pch): Likewise
9310 (_mm256_mask_fmadd_pch): Likewise
9311 (_mm256_mask3_fmadd_pch): Likewise
9312 (_mm_mask_fcmadd_pch): Likewise
9313 (_mm_mask3_fcmadd_pch): Likewise
9314 (_mm256_mask_fcmadd_pch): Likewise
9315 (_mm256_mask3_fcmadd_pch): Likewise
9316 * config/i386/i386-builtin.def: Add mask3 builtin for complex
9317 fma, and adjust mask_builtin to corresponding expander.
9318 * config/i386/i386-expand.c (ix86_expand_round_builtin):
9319 Skip eraseing embedded rounding for expanders that emits
9321 * config/i386/sse.md (complexmove): New mode_attr.
9322 (<avx512>_fmaddc_<mode>_mask1<round_expand_name>): New expander.
9323 (<avx512>_fcmaddc_<mode>_mask1<round_expand_name>): Likewise.
9324 (avx512fp16_fmaddcsh_v8hf_mask1<round_expand_name>): Likewise.
9325 (avx512fp16_fcmaddcsh_v8hf_mask1<round_expand_name>): Likewise.
9326 (avx512fp16_fcmaddcsh_v8hf_mask3<round_expand_name>): Likewise.
9327 (avx512fp16_fmaddcsh_v8hf_mask3<round_expand_name>): Likewise.
9328 * config/i386/subst.md (round_embedded_complex): New subst.
9330 2021-10-14 John David Anglin <danglin@gcc.gnu.org>
9332 * config/pa/pa.md (cbranchsf4): Disable if TARGET_SOFT_FLOAT.
9333 (cbranchdf4): Likewise.
9334 Add missing move patterns for TARGET_SOFT_FLOAT.
9336 2021-10-13 Roger Sayle <roger@nextmovesoftware.com>
9338 * config/i386/i386-expand.c (ix86_expand_vector_move): Use a
9339 pseudo intermediate when moving a SUBREG into a hard register,
9340 by checking ix86_hardreg_mov_ok.
9341 (ix86_expand_vector_extract): Store zero-extended SImode
9342 intermediate in a pseudo, then set target using a SUBREG_PROMOTED
9344 * config/i386/sse.md (mov<VMOVE>_internal): Prevent CSE creating
9345 complex (SUBREG) sets of (vector) hard registers before reload, by
9346 checking ix86_hardreg_mov_ok.
9348 2021-10-13 Indu Bhagat <indu.bhagat@oracle.com>
9350 * ctfc.h (enum ctf_dtu_d_union_enum): Remove redundant comma.
9352 2021-10-13 Indu Bhagat <indu.bhagat@oracle.com>
9354 * dwarf2ctf.c (gen_ctf_array_type): Fix typo in comment.
9356 2021-10-13 Martin Sebor <msebor@redhat.com>
9358 PR middle-end/102630
9359 * pointer-query.cc (compute_objsize_r): Handle named address spaces.
9361 2021-10-13 Iain Sandoe <iain@sandoe.co.uk>
9363 * collect2.c (is_lto_object_file): Release simple-object
9364 resources, close files.
9366 2021-10-13 Andre Vieira <andre.simoesdiasvieira@arm.com>
9368 * config/arm/arm.c (thumb2_legitimate_address_p): Use VALID_MVE_MODE
9369 when checking mve addressing modes.
9370 (mve_vector_mem_operand): Fix the way we handle pre, post and offset
9372 (arm_print_operand): Fix printing of POST_ and PRE_MODIFY.
9373 * config/arm/mve.md: Use mve_memory_operand predicate everywhere where
9374 there is a single Ux constraint.
9376 2021-10-13 John David Anglin <danglin@gcc.gnu.org>
9378 * config/pa/pa.md (muldi3): Add support for inlining 64-bit
9379 multiplication on 32-bit PA 1.1 and 2.0 targets.
9381 2021-10-13 Martin Liska <mliska@suse.cz>
9383 PR gcov-profile/90364
9384 * coverage.c (build_info): Emit checksum to the global variable.
9385 (build_info_type): Add new field for checksum.
9386 (coverage_obj_finish): Pass object_checksum.
9387 (coverage_init): Use 0 as checksum for .gcno files.
9388 * gcov-dump.c (dump_gcov_file): Dump also new checksum field.
9389 * gcov.c (read_graph_file): Read also checksum.
9390 * doc/invoke.texi: Document the behaviour change.
9392 2021-10-13 Richard Biener <rguenther@suse.de>
9394 * gimple-iterator.h (gsi_iterator_update): Add GSI_LAST_NEW_STMT,
9395 start at integer value 2.
9396 * gimple-iterator.c (gsi_insert_seq_nodes_before): Update
9397 the iterator for GSI_LAST_NEW_STMT.
9398 (gsi_insert_seq_nodes_after): Likewise.
9399 * tree-if-conv.c (predicate_statements): Use GSI_LAST_NEW_STMT.
9400 * tree-ssa.c (execute_update_addresses_taken): Correct bogus
9401 arguments to gsi_replace.
9403 2021-10-13 Martin Liska <mliska@suse.cz>
9406 * common.opt: Use EnabledBy instead of detection in
9407 finish_options and process_options.
9408 * opts.c (finish_options): Remove handling of
9409 x_flag_unroll_all_loops.
9410 * toplev.c (process_options): Likewise for flag_web and
9411 flag_rename_registers.
9413 2021-10-13 Richard Biener <rguenther@suse.de>
9415 PR tree-optimization/102659
9416 * tree-if-conv.c (need_to_rewrite_undefined): New flag.
9417 (if_convertible_gimple_assign_stmt_p): Mark the loop for
9418 rewrite when stmts with undefined behavior on integer
9420 (combine_blocks): Predicate also when we need to rewrite stmts.
9421 (predicate_statements): Rewrite affected stmts to something
9422 with well-defined behavior on overflow.
9423 (tree_if_conversion): Initialize need_to_rewrite_undefined.
9425 2021-10-13 Richard Biener <rguenther@suse.de>
9428 * ipa-sra.c (ptr_parm_has_nonarg_uses): Fix volatileness
9431 2021-10-13 Jose E. Marchesi <jose.marchesi@oracle.com>
9433 * dwarf2ctf.c: Fix typo in comment.
9435 2021-10-12 Bill Schmidt <wschmidt@linux.ibm.com>
9438 * config/rs6000/altivec.h (vec_cpsgn): Swap operand order.
9439 * config/rs6000/rs6000-overload.def (VEC_COPYSIGN): Use SKIP to
9440 avoid generating an automatic #define of vec_cpsgn. Use the
9441 correct built-in for V4SFmode that doesn't depend on VSX.
9443 2021-10-12 Uroš Bizjak <ubizjak@gmail.com>
9447 * config/i386/i386.md (*add<mode>_1_slp): Rewrite as
9448 define_insn_and_split pattern. Add alternative 1 and split it
9449 post reload to insert operand 1 into the low part of operand 0.
9450 (*sub<mode>_1_slp): Ditto.
9451 (*and<mode>_1_slp): Ditto.
9452 (*<any_or:code><mode>_1_slp): Ditto.
9453 (*ashl<mode>3_1_slp): Ditto.
9454 (*<any_shiftrt:insn><mode>3_1_slp): Ditto.
9455 (*<any_rotate:insn><mode>3_1_slp): Ditto.
9456 (*neg<mode>_1_slp): New insn_and_split pattern.
9457 (*one_cmpl<mode>_1_slp): Ditto.
9459 2021-10-12 David Edelsohn <dje.gcc@gmail.com>
9461 * doc/install.texi: Update MinGW and mingw-64 Binaries
9464 2021-10-12 Daniel Le Duc Khoi Nguyen <greenrecyclebin@gmail.com>
9466 * doc/extend.texi (Common Variable Attributes): Fix typos in
9467 alloc_size documentation.
9469 2021-10-12 Richard Biener <rguenther@suse.de>
9471 PR tree-optimization/102696
9472 * tree-vect-slp.c (vect_build_slp_tree_2): Properly mark
9473 the tree fatally failed when we reject a BIT_FIELD_REF.
9475 2021-10-12 Richard Biener <rguenther@suse.de>
9477 PR tree-optimization/102572
9478 * tree-vect-stmts.c (vect_build_gather_load_calls): When
9479 gathering the vectorized defs for the mask pass in the
9480 desired mask vector type so invariants will be handled
9483 2021-10-12 Tamar Christina <tamar.christina@arm.com>
9485 * config/aarch64/aarch64-sve.md (*fcm<cmp_op><mode>_bic_combine,
9486 *fcm<cmp_op><mode>_nor_combine, *fcmuo<mode>_bic_combine,
9487 *fcmuo<mode>_nor_combine): New.
9489 2021-10-12 Eric Botcazou <ebotcazou@adacore.com>
9492 * config/sparc/sparc-modes.def (OI): New integer mode.
9494 2021-10-12 Jakub Jelinek <jakub@redhat.com>
9496 * gimple-fold.h (clear_padding_type_may_have_padding_p): Declare.
9497 * gimple-fold.c (clear_padding_type_may_have_padding_p): No longer
9500 2021-10-12 Jakub Jelinek <jakub@redhat.com>
9502 * tree-vectorizer.h (loop_cost_model): New function.
9503 (unlimited_cost_model): Use it.
9504 * tree-vect-loop.c (vect_analyze_loop_costing): Use loop_cost_model
9505 call instead of flag_vect_cost_model.
9506 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
9507 (vect_prune_runtime_alias_test_list): Likewise. Also use it instead
9508 of flag_simd_cost_model.
9510 2021-10-12 liuhongt <hongtao.liu@intel.com>
9513 * config/i386/i386-expand.c (emit_reduc_half): Handle
9515 * config/i386/mmx.md (reduc_<code>_scal_v4qi): New expander.
9516 (reduc_plus_scal_v4qi): Ditto.
9518 2021-10-12 Paul A. Clarke <pc@us.ibm.com>
9520 * config/rs6000/smmintrin.h (_mm_cmpeq_epi64, _mm_cmpgt_epi64,
9521 _mm_mullo_epi32, _mm_mul_epi32, _mm_packus_epi32): New.
9522 * config/rs6000/nmmintrin.h: Copy from i386, tweak to suit.
9524 2021-10-12 Paul A. Clarke <pc@us.ibm.com>
9526 * config/rs6000/smmintrin.h (_mm_cvtepi8_epi16, _mm_cvtepi8_epi32,
9527 _mm_cvtepi8_epi64, _mm_cvtepi16_epi32, _mm_cvtepi16_epi64,
9528 _mm_cvtepi32_epi64, _mm_cvtepu8_epi16, _mm_cvtepu8_epi32,
9529 _mm_cvtepu8_epi64, _mm_cvtepu16_epi32, _mm_cvtepu16_epi64,
9530 _mm_cvtepu32_epi64): New.
9532 2021-10-12 Paul A. Clarke <pc@us.ibm.com>
9534 * config/rs6000/smmintrin.h (_mm_test_all_zeros,
9535 _mm_test_all_ones, _mm_test_mix_ones_zeros): Rewrite as macro.
9537 2021-10-12 Paul A. Clarke <pc@us.ibm.com>
9539 * config/rs6000/smmintrin.h (_mm_min_epi8, _mm_min_epu16,
9540 _mm_min_epi32, _mm_min_epu32, _mm_max_epi8, _mm_max_epu16,
9541 _mm_max_epi32, _mm_max_epu32): New.
9543 2021-10-11 Jan Hubicka <hubicka@ucw.cz>
9545 * ipa-modref-tree.h (struct modref_access_node): Revert
9547 (struct modref_ref_node): Likewise.
9549 2021-10-11 Jan Hubicka <hubicka@ucw.cz>
9551 * ipa-modref-tree.h (modref_tree::global_access_p): New member
9554 (implicint_const_eaf_flags,implicit_pure_eaf_flags,
9555 ignore_stores_eaf_flags): Move to ipa-modref.h
9556 (remove_useless_eaf_flags): Remove early exit on NOCLOBBER.
9557 (modref_summary::global_memory_read_p): New member function.
9558 (modref_summary::global_memory_written_p): New member function.
9559 * ipa-modref.h (modref_summary::global_memory_read_p,
9560 modref_summary::global_memory_written_p): Declare.
9561 (implicint_const_eaf_flags,implicit_pure_eaf_flags,
9562 ignore_stores_eaf_flags): move here.
9563 * tree-ssa-structalias.c: Include ipa-modref-tree.h, ipa-modref.h
9565 (handle_rhs_call): Rewrite.
9566 (handle_call_arg): New function.
9567 (determine_global_memory_access): New function.
9568 (handle_const_call): Remove
9569 (handle_pure_call): Remove
9570 (find_func_aliases_for_call): Update use of handle_rhs_call.
9571 (compute_points_to_sets): Handle global memory acccesses
9574 2021-10-11 Diane Meirowitz <diane.meirowitz@oracle.com>
9576 * doc/invoke.texi: Add link to UndefinedBehaviorSanitizer
9577 documentation, mention UBSAN_OPTIONS, similar to what is done
9578 for AddressSanitizer.
9580 2021-10-11 Richard Biener <rguenther@suse.de>
9582 PR middle-end/102683
9583 * internal-fn.c (expand_DEFERRED_INIT): Check for mode
9584 availability before building an integer type for storage
9587 2021-10-11 Richard Biener <rguenther@suse.de>
9589 PR middle-end/101480
9590 * gimple.c (gimple_call_fnspec): Do not mark operator new/delete
9593 2021-10-11 Martin Liska <mliska@suse.cz>
9595 * common.opt: Remove Init(2) for some options.
9596 * toplev.c (process_options): Do not use AUTODETECT_VALUE, but
9597 use rather OPTION_SET_P.
9599 2021-10-11 Martin Liska <mliska@suse.cz>
9601 * common.opt: Remove usage of IRA_REGION_AUTODETECT.
9602 * flag-types.h (enum ira_region): Likewise.
9603 * toplev.c (process_options): Use OPTION_SET_P instead of
9604 IRA_REGION_AUTODETECT.
9606 2021-10-11 Jakub Jelinek <jakub@redhat.com>
9608 * omp-low.c (omp_runtime_api_call): Handle omp_get_max_teams,
9609 omp_[sg]et_teams_thread_limit and omp_set_num_teams.
9611 2021-10-11 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
9613 * config/s390/s390-protos.h (s390_rawmemchr): Add prototype.
9614 * config/s390/s390.c (s390_rawmemchr): New function.
9615 * config/s390/s390.md (rawmemchr<SINT:mode>): New expander.
9616 * config/s390/vector.md (@vec_vfees<mode>): Basically a copy of
9617 the pattern vfees<mode> from vx-builtins.md.
9618 * config/s390/vx-builtins.md (*vfees<mode>): Remove.
9620 2021-10-11 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
9622 * builtins.c (get_memory_rtx): Change to external linkage.
9623 * builtins.h (get_memory_rtx): Add function prototype.
9624 * doc/md.texi (rawmemchr<mode>): Document.
9625 * internal-fn.c (expand_RAWMEMCHR): Define.
9626 * internal-fn.def (RAWMEMCHR): Add.
9627 * optabs.def (rawmemchr_optab): Add.
9628 * tree-loop-distribution.c (find_single_drs): Change return code
9629 behaviour by also returning true if no single store was found
9631 (loop_distribution::classify_partition): Respect the new return
9632 code behaviour of function find_single_drs.
9633 (loop_distribution::execute): Call new function
9634 transform_reduction_loop in order to replace rawmemchr or strlen
9635 like loops by calls into builtins.
9636 (generate_reduction_builtin_1): New function.
9637 (generate_rawmemchr_builtin): New function.
9638 (generate_strlen_builtin_1): New function.
9639 (generate_strlen_builtin): New function.
9640 (generate_strlen_builtin_using_rawmemchr): New function.
9641 (reduction_var_overflows_first): New function.
9642 (determine_reduction_stmt_1): New function.
9643 (determine_reduction_stmt): New function.
9644 (loop_distribution::transform_reduction_loop): New function.
9646 2021-10-11 Martin Liska <mliska@suse.cz>
9648 * tree.c (cl_option_hasher::hash): Use cl_optimization_hash
9649 and remove legacy hashing code.
9651 2021-10-11 Kito Cheng <kito.cheng@sifive.com>
9654 * builtins.c (maybe_emit_call_builtin___clear_cache): Allow
9655 CONST_INT for BEGIN and END, and use gcc_assert rather than
9658 2021-10-10 Jakub Jelinek <jakub@redhat.com>
9661 * var-tracking.c (add_stores): For cselib_sp_derived_value_p values
9662 use MO_VAL_SET if loc is not sp.
9664 2021-10-10 Andrew Pinski <apinski@marvell.com>
9666 PR tree-optimization/102622
9667 * match.pd: Swap the order of a?pow2cst:0 and a?-1:0 transformations.
9668 Swap the order of a?0:pow2cst and a?0:-1 transformations.
9670 2021-10-09 liuhongt <hongtao.liu@intel.com>
9673 * config/i386/i386-expand.c (ix86_valid_mask_cmp_mode): Handle
9675 (ix86_use_mask_cmp_p): Ditto.
9676 (ix86_expand_sse_movcc): Ditto.
9677 * config/i386/i386.md (setcc_hf_mask): New define_insn.
9678 (movhf_mask): Ditto.
9679 (UNSPEC_MOVCC_MASK): New unspec.
9680 * config/i386/sse.md (UNSPEC_PCMP): Move to i386.md.
9682 2021-10-08 Vladimir N. Makarov <vmakarov@redhat.com>
9684 PR rtl-optimization/102627
9685 * lra-constraints.c (split_reg): Use at least natural mode of hard reg.
9687 2021-10-08 Aldy Hernandez <aldyh@redhat.com>
9689 * gimple-range-cache.cc (non_null_ref::non_null_deref_p): Grow
9692 2021-10-08 Aldy Hernandez <aldyh@redhat.com>
9694 * value-range.cc (irange::debug): New.
9695 * value-range.h (irange::debug): New.
9697 2021-10-08 Richard Sandiford <richard.sandiford@arm.com>
9699 PR tree-optimization/102385
9700 * predict.h (change_edge_frequency): Declare.
9701 * predict.c (change_edge_frequency): New function.
9702 * tree-ssa-loop-manip.h (tree_transform_and_unroll_loop): Remove
9704 (tree_unroll_loop): Likewise.
9705 * gimple-loop-jam.c (tree_loop_unroll_and_jam): Update accordingly.
9706 * tree-predcom.c (pcom_worker::tree_predictive_commoning_loop):
9708 * tree-ssa-loop-prefetch.c (loop_prefetch_arrays): Likewise.
9709 * tree-ssa-loop-manip.c (tree_unroll_loop): Likewise.
9710 (tree_transform_and_unroll_loop): Likewise. Use single_dom_exit
9711 to retrieve the exit edges. Make all the old profile update code
9712 conditional on !single_loop_p -- the case it was written for --
9713 and use a different approach for the single-loop case.
9715 2021-10-08 Martin Liska <mliska@suse.cz>
9717 * config/alpha/alpha.c (alpha_option_override): Use new macro
9719 * config/arc/arc.c (arc_override_options): Likewise.
9720 * config/arm/arm.c (arm_option_override): Likewise.
9721 * config/bfin/bfin.c (bfin_load_pic_reg): Likewise.
9722 * config/c6x/c6x.c (c6x_option_override): Likewise.
9723 * config/csky/csky.c: Likewise.
9724 * config/darwin.c (darwin_override_options): Likewise.
9725 * config/frv/frv.c (frv_option_override): Likewise.
9726 * config/i386/djgpp.h: Likewise.
9727 * config/i386/i386.c (ix86_stack_protect_guard): Likewise.
9728 (ix86_max_noce_ifcvt_seq_cost): Likewise.
9729 * config/ia64/ia64.c (ia64_option_override): Likewise.
9730 (ia64_override_options_after_change): Likewise.
9731 * config/m32c/m32c.c (m32c_option_override): Likewise.
9732 * config/m32r/m32r.c (m32r_init): Likewise.
9733 * config/m68k/m68k.c (m68k_option_override): Likewise.
9734 * config/microblaze/microblaze.c (microblaze_option_override): Likewise.
9735 * config/mips/mips.c (mips_option_override): Likewise.
9736 * config/nios2/nios2.c (nios2_option_override): Likewise.
9737 * config/nvptx/nvptx.c (nvptx_option_override): Likewise.
9738 * config/pa/pa.c (pa_option_override): Likewise.
9739 * config/riscv/riscv.c (riscv_option_override): Likewise.
9740 * config/rs6000/aix71.h: Likewise.
9741 * config/rs6000/aix72.h: Likewise.
9742 * config/rs6000/aix73.h: Likewise.
9743 * config/rs6000/rs6000.c (darwin_rs6000_override_options): Likewise.
9744 (rs6000_override_options_after_change): Likewise.
9745 (rs6000_linux64_override_options): Likewise.
9746 (glibc_supports_ieee_128bit): Likewise.
9747 (rs6000_option_override_internal): Likewise.
9748 (rs6000_file_start): Likewise.
9749 (rs6000_darwin_file_start): Likewise.
9750 * config/rs6000/rtems.h: Likewise.
9751 * config/rs6000/sysv4.h: Likewise.
9752 * config/rs6000/vxworks.h (SUB3TARGET_OVERRIDE_OPTIONS): Likewise.
9753 * config/s390/s390.c (s390_option_override): Likewise.
9754 * config/sh/linux.h: Likewise.
9755 * config/sh/netbsd-elf.h (while): Likewise.
9756 * config/sh/sh.c (sh_option_override): Likewise.
9757 * config/sol2.c (solaris_override_options): Likewise.
9758 * config/sparc/sparc.c (sparc_option_override): Likewise.
9759 * config/tilegx/tilegx.c (tilegx_option_override): Likewise.
9760 * config/visium/visium.c (visium_option_override): Likewise.
9761 * config/vxworks.c (vxworks_override_options): Likewise.
9762 * lto-opts.c (lto_write_options): Likewise.
9763 * omp-expand.c (expand_omp_simd): Likewise.
9764 * omp-general.c (omp_max_vf): Likewise.
9765 * omp-offload.c (oacc_xform_loop): Likewise.
9766 * opts.h (OPTION_SET_P): Likewise.
9767 * targhooks.c (default_max_noce_ifcvt_seq_cost): Likewise.
9768 * toplev.c (process_options): Likewise.
9769 * tree-predcom.c: Likewise.
9770 * tree-sra.c (analyze_all_variable_accesses): Likewise.
9772 2021-10-08 liuhongt <hongtao.liu@intel.com>
9775 * config/i386/i386.c (ix86_optab_supported_p):
9776 Return true for HFmode.
9777 * match.pd: Simplify (_Float16) ceil ((double) x) to
9778 __builtin_ceilf16 (a) when a is _Float16 type and
9779 direct_internal_fn_supported_p.
9781 2021-10-08 liuhongt <hongtao.liu@intel.com>
9784 * config/i386/i386-expand.c (emit_reduc_half): Hanlde V4HImode.
9785 * config/i386/mmx.md (reduc_plus_scal_v4hi): New.
9786 (reduc_<code>_scal_v4hi): New.
9788 2021-10-08 liuhongt <hongtao.liu@intel.com>
9790 * common.opt (ftree-vectorize): Add Var(flag_tree_vectorize).
9791 * doc/invoke.texi (Options That Control Optimization): Update
9793 * opts.c (default_options_table): Enable auto-vectorization at
9794 O2 with very-cheap cost model.
9795 (finish_options): Use cheap cost model for
9796 explicit -ftree{,-loop}-vectorize.
9798 2021-10-07 Indu Bhagat <indu.bhagat@oracle.com>
9800 * ctfc.c (ctfc_delete_container): Free hash table contents.
9802 2021-10-07 Indu Bhagat <indu.bhagat@oracle.com>
9804 * toplev.c (process_options): Do not warn for GNU GIMPLE.
9806 2021-10-07 Siddhesh Poyarekar <siddhesh@gotplt.org>
9808 * tree-object-size.c (addr_object_size,
9809 compute_builtin_object_size): Drop PDECL and POFF arguments.
9810 (addr_object_size): Adjust calls.
9811 * tree-object-size.h (compute_builtin_object_size): Drop PDECL
9814 2021-10-07 Roger Sayle <roger@nextmovesoftware.com>
9816 * rtl.def (SMUL_HIGHPART, UMUL_HIGHPART): New RTX codes for
9817 representing signed and unsigned high-part multiplication resp.
9818 * simplify-rtx.c (simplify_binary_operation_1) [SMUL_HIGHPART,
9819 UMUL_HIGHPART]: Simplify high-part multiplications by zero.
9820 [SS_PLUS, US_PLUS, SS_MINUS, US_MINUS, SS_MULT, US_MULT,
9821 SS_DIV, US_DIV]: Similar simplifications for saturating
9823 (simplify_const_binary_operation) [SS_PLUS, US_PLUS, SS_MINUS,
9824 US_MINUS, SS_MULT, US_MULT, SMUL_HIGHPART, UMUL_HIGHPART]:
9825 Implement compile-time evaluation for constant operands.
9826 * dwarf2out.c (mem_loc_descriptor): Skip SMUL_HIGHPART and
9828 * doc/rtl.texi (smul_highpart, umul_highpart): Document RTX codes.
9829 * doc/md.texi (smul@var{m}3_highpart, umul@var{m3}_highpart):
9830 Mention the new smul_highpart and umul_highpart RTX codes.
9831 * doc/invoke.texi: Silence @xref "compilation" warnings.
9833 2021-10-07 Martin Jambor <mjambor@suse.cz>
9836 * ipa-prop.c (ipa_edge_args_sum_t::duplicate): Also handle the
9837 case when the source reference description corresponds to a
9838 referance taken in a function src->caller is inlined to.
9840 2021-10-07 Jan Hubicka <hubicka@ucw.cz>
9843 * ipa-modref-tree.h (modref_access_node::contains_p): Handle offsets
9845 (modref_access_node::try_merge_with): Add sanity check that there
9846 are no redundant entries in the list.
9848 2021-10-07 Richard Biener <rguenther@suse.de>
9850 PR tree-optimization/102608
9851 * tree-ssa-sccvn.c (visit_stmt): Drop .DEFERRED_INIT to
9854 2021-10-07 Martin Liska <mliska@suse.cz>
9856 * toplev.c (toplev::main): Make
9857 save_opt_decoded_options a pointer type
9858 * toplev.h: Likewise.
9860 2021-10-07 Andrew Stubbs <ams@codesourcery.com>
9862 * config/gcn/gcn-valu.md (gather<mode>_insn_2offsets<exec>): Apply
9863 HAVE_GCN_ASM_GLOBAL_LOAD_FIXED.
9864 (scatter<mode>_insn_2offsets<exec_scatter>): Likewise.
9866 2021-10-07 Andrew Stubbs <ams@codesourcery.com>
9868 * config/gcn/gcn-hsa.h (SRAMOPT): Include the whole option string.
9869 Adjust for new -msram-ecc=any behaviour.
9870 (ASM_SPEC): Adjust -mxnack and -msram-ecc usage.
9871 * config/gcn/gcn.c (output_file_start): Implement -msram-ecc=any.
9872 * config/gcn/mkoffload.c (EF_AMDGPU_XNACK): Rename to ...
9873 (EF_AMDGPU_XNACK_V3): ... this.
9874 (EF_AMDGPU_SRAM_ECC): Rename to ...
9875 (EF_AMDGPU_SRAM_ECC_V3): ... this.
9876 (EF_AMDGPU_FEATURE_XNACK_V4): New.
9877 (EF_AMDGPU_FEATURE_XNACK_UNSUPPORTED_V4): New.
9878 (EF_AMDGPU_FEATURE_XNACK_ANY_V4): New.
9879 (EF_AMDGPU_FEATURE_XNACK_OFF_V4): New.
9880 (EF_AMDGPU_FEATURE_XNACK_ON_V4): New.
9881 (EF_AMDGPU_FEATURE_SRAMECC_V4): New.
9882 (EF_AMDGPU_FEATURE_SRAMECC_UNSUPPORTED_V4): New.
9883 (EF_AMDGPU_FEATURE_SRAMECC_ANY_V4): New.
9884 (EF_AMDGPU_FEATURE_SRAMECC_OFF_V4): New.
9885 (EF_AMDGPU_FEATURE_SRAMECC_ON_V4): New.
9886 (SET_XNACK_ON): New.
9887 (SET_XNACK_OFF): New.
9889 (SET_SRAM_ECC_ON): New.
9890 (SET_SRAM_ECC_ANY): New.
9891 (SET_SRAM_ECC_OFF): New.
9892 (TEST_SRAM_ECC_ANY): New.
9893 (TEST_SRAM_ECC_ON): New.
9894 (main): Implement HSACOv4 and -msram-ecc=any.
9896 2021-10-07 Andrew Stubbs <ams@codesourcery.com>
9898 * config.in: Regenerate.
9899 * config/gcn/gcn-hsa.h (X_FIJI): New macro.
9903 (A_FIJI): Rename to ...
9905 (A_900): Rename to ...
9907 (A_906): Rename to ...
9909 (A_908): Rename to ...
9911 (SRAMOPT): New macro.
9912 (ASM_SPEC): Adjust xnack option usage.
9913 * config/gcn/gcn.c (output_file_start): Adjust amdgcn_target usage.
9914 * configure: Regenerate.
9915 * configure.ac: Detect LLVM assembler dialect.
9917 2021-10-07 Richard Biener <rguenther@suse.de>
9919 * tree-pretty-print.c (dump_generic_node): Do not elide
9920 printing '&' when dumping with -gimple.
9922 2021-10-06 Andrew MacLeod <amacleod@redhat.com>
9924 * gimple-range-cache.cc (non_null_ref::adjust_range): Call new
9926 * gimple-range-fold.cc (adjust_pointer_diff_expr): Ditto.
9927 (adjust_imagpart_expr): Ditto.
9928 * value-range.cc (irange::irange_intersect): Call new routine if
9929 RHS is a single pair.
9930 (irange::intersect): New wide_int version.
9931 * value-range.h (class irange): New prototype.
9933 2021-10-06 Andrew MacLeod <amacleod@redhat.com>
9935 * gimple-range-edge.cc (gimple_outgoing_range::gimple_outgoing_range):
9936 Add parameter to limit size when recognizing switches.
9937 (gimple_outgoing_range::edge_range_p): Check size limit.
9938 * gimple-range-edge.h (gimple_outgoing_range): Add size field.
9939 * gimple-range-gori.cc (gori_map::calculate_gori): Ignore switches
9940 that exceed the size limit.
9941 (gori_compute::gori_compute): Add initializer.
9942 * params.opt (evrp-switch-limit): New.
9943 * doc/invoke.texi: Update docs.
9945 2021-10-06 Andrew MacLeod <amacleod@redhat.com>
9947 * value-range.h (irange::set_varying): Use TYPE_MIN_VALUE and
9948 TYPE_MAX_VALUE instead of creating new trees when possible.
9950 2021-10-06 Andrew MacLeod <amacleod@redhat.com>
9952 * gimple-range-cache.cc (non_null_ref::adjust_range): Check for
9953 zero and non-zero more efficently.
9955 2021-10-06 Richard Biener <rguenther@suse.de>
9958 * dumpfile.h (TDF_GIMPLE_VAL): New.
9959 (dump_flag): Re-order and adjust TDF_* flags. Make
9960 the enum uint32_t. Use std::underlying_type in the
9962 (optgroup_flag): Likewise for the operator overloads.
9963 * tree-pretty-print.c (dump_generic_node): Wrap ADDR_EXPR
9964 in _Literal if TDF_GIMPLE_VAL.
9965 * gimple-pretty-print.c (dump_gimple_assign): Add
9966 TDF_GIMPLE_VAL to flags when dumping operands where only
9967 is_gimple_val are allowed.
9968 (dump_gimple_cond): Likewise.
9970 2021-10-06 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
9972 * gimple-isel.cc (gimple_expand_vec_cond_expr): Remove redundant if
9975 2021-10-05 qing zhao <qing.zhao@oracle.com>
9977 PR middle-end/102359
9978 * gimplify.c (gimplify_decl_expr): Not add initialization for an
9979 auto variable when it has been initialized by frontend.
9981 2021-10-05 Aldy Hernandez <aldyh@redhat.com>
9983 * tree-ssa-threadupdate.c (jt_path_registry::cancel_invalid_paths):
9986 2021-10-05 Jan-Benedict Glaw <jbglaw@lug-owl.de>
9988 * common/config/avr/avr-common.c (avr_handle_option): Mark
9989 argument as ATTRIBUTE_UNUSED.
9991 2021-10-05 Jan-Benedict Glaw <jbglaw@lug-owl.de>
9993 * config/lm32/uclinux-elf.h (LINK_GCC_C_SEQUENCE_SPEC):
9994 Undefine before redefinition.
9996 2021-10-05 Richard Biener <rguenther@suse.de>
9998 * toplev.c (no_backend): Remove global var.
9999 (process_options): Pass in no_backend, move post_options
10000 langhook call to toplev::main.
10001 (do_compile): Pass in no_backend, move process_options call
10003 (toplev::run_self_tests): Check no_backend at the caller.
10004 (toplev::main): Call post_options and process_options
10005 split out from do_compile, do self-tests only if
10006 no_backend is initialized.
10008 2021-10-05 Richard Biener <rguenther@suse.de>
10010 * tree-cfg.c (dump_function_to_file): Dump the UID of the
10011 function as part of the name when requested.
10012 * tree-pretty-print.c (dump_function_name): Dump the UID when
10013 requested and the langhook produced the actual name.
10015 2021-10-05 Richard Biener <rguenther@suse.de>
10017 PR middle-end/102587
10018 PR middle-end/102285
10019 * internal-fn.c (expand_DEFERRED_INIT): Fall back to
10020 zero-initialization as last resort, use the constant
10021 size as given by the DEFERRED_INIT argument to build
10024 2021-10-04 Marek Polacek <polacek@redhat.com>
10027 * doc/invoke.texi: Document -Warray-compare.
10029 2021-10-04 Richard Biener <rguenther@suse.de>
10031 * gimplify.c (is_var_need_auto_init): DECL_HARD_REGISTER
10032 variables are not to be initialized.
10034 2021-10-04 Richard Biener <rguenther@suse.de>
10036 * expr.h (non_mem_decl_p): Declare.
10037 (mem_ref_refers_to_non_mem_p): Likewise.
10038 * expr.c (non_mem_decl_p): Export.
10039 (mem_ref_refers_to_non_mem_p): Likewise.
10040 * internal-fn.c (expand_DEFERRED_INIT): Do not expand the LHS
10041 but check the base with mem_ref_refers_to_non_mem_p
10042 and non_mem_decl_p.
10044 2021-10-04 Richard Biener <rguenther@suse.de>
10046 PR tree-optimization/102570
10047 * tree-ssa-sccvn.h (vn_reference_op_struct): Document
10048 we are using clique for the internal function code.
10049 * tree-ssa-sccvn.c (vn_reference_op_eq): Compare the
10050 internal function code.
10051 (print_vn_reference_ops): Print the internal function code.
10052 (vn_reference_op_compute_hash): Hash it.
10053 (copy_reference_ops_from_call): Record it.
10054 (visit_stmt): Remove the restriction around internal function
10056 (fully_constant_vn_reference_p): Use fold_const_call and handle
10057 internal functions.
10058 (vn_reference_eq): Compare call return types.
10059 * tree-ssa-pre.c (create_expression_by_pieces): Handle
10060 generating calls to internal functions.
10061 (compute_avail): Remove the restriction around internal function
10064 2021-10-04 Aldy Hernandez <aldyh@redhat.com>
10066 PR tree-optimization/102560
10067 * gimple-ssa-warn-alloca.c (alloca_call_type): Remove static
10068 marker for invalid_range.
10070 2021-10-04 Richard Biener <rguenther@suse.de>
10072 PR middle-end/102587
10073 * internal-fn.c (expand_DEFERRED_INIT): Guard register
10074 initialization path an avoid initializing VLA registers
10077 2021-10-04 Eric Botcazou <ebotcazou@adacore.com>
10079 * config/rs6000/vxworks.h (TARGET_INIT_LIBFUNCS): Delete.
10081 2021-10-03 Martin Liska <mliska@suse.cz>
10083 * toplev.c (toplev::main): Check opt_index if it is a part
10086 2021-10-02 Aldy Hernandez <aldyh@redhat.com>
10088 PR tree-optimization/102563
10089 * range-op.cc (operator_lshift::op1_range): Do not clobber
10092 2021-10-02 Martin Liska <mliska@suse.cz>
10094 * toplev.c (toplev::main): save_decoded_options[0] is program
10095 name and so it should be skipped.
10097 2021-10-01 Aldy Hernandez <aldyh@redhat.com>
10099 PR tree-optimization/102546
10100 * range-op.cc (operator_lshift::op1_range): Teach range-ops that
10101 X << Y is non-zero implies X is also non-zero.
10103 2021-10-01 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
10105 * config/aarch64/aarch64-cores.def (AARCH64_CORE): New
10107 * config/aarch64/aarch64-tune.md: Regenerate.
10108 * doc/invoke.texi: Update docs.
10110 2021-10-01 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
10112 * config/aarch64/aarch64-cores.def (AARCH64_CORE): New
10114 * config/aarch64/aarch64-tune.md: Regenerate.
10115 * doc/invoke.texi: Update docs.
10117 2021-10-01 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
10119 * config/aarch64/aarch64-cores.def (AARCH64_CORE): New
10121 * config/aarch64/aarch64-tune.md: Regenerate.
10122 * doc/invoke.texi: Update docs.
10124 2021-10-01 Martin Sebor <msebor@redhat.com>
10127 * doc/invoke.texi (-Waddress): Update.
10128 * gengtype.c (write_types): Avoid -Waddress.
10129 * poly-int.h (POLY_SET_COEFF): Avoid using null.
10131 2021-10-01 John David Anglin <danglin@gcc.gnu.org>
10134 * config/pa/pa.c (pa_option_override): Default to dwarf version 4
10137 2021-10-01 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
10139 * config/aarch64/aarch64.h (AARCH64_FL_V9): Update value.
10141 2021-10-01 Aldy Hernandez <aldyh@redhat.com>
10143 * gimple-range-path.cc (path_range_query::compute_ranges): Use
10145 * gimple-range-path.h (class path_range_query): Remove shadowed
10147 (path_range_query::get_path_oracle): New.
10149 2021-10-01 Jakub Jelinek <jakub@redhat.com>
10150 Richard Biener <rguenther@suse.de>
10152 PR sanitizer/102515
10153 * doc/invoke.texi (-fsanitize=integer-divide-by-zero): Remove
10154 INT_MIN / -1 division detection from here ...
10155 (-fsanitize=signed-integer-overflow): ... and add it here.
10157 2021-10-01 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
10159 * config/aarch64/aarch64-arches.def (AARCH64_ARCH): Added
10161 * config/aarch64/aarch64.h (AARCH64_FL_V9): New.
10162 (AARCH64_FL_FOR_ARCH9): New flags for Armv9-A.
10163 (AARCH64_ISA_V9): New ISA flag.
10164 * doc/invoke.texi: Update docs.
10166 2021-10-01 Martin Liska <mliska@suse.cz>
10168 * toplev.c (toplev::main): Save decoded optimization options.
10169 * toplev.h (save_opt_decoded_options): New.
10170 * doc/extend.texi: Be more clear about optimize and target
10173 2021-10-01 Eric Botcazou <ebotcazou@adacore.com>
10175 * explow.c: Include langhooks.h.
10176 (set_stack_check_libfunc): Build a proper function type.
10178 2021-10-01 Eric Botcazou <ebotcazou@adacore.com>
10181 * config/i386/i386.c (legitimate_pic_address_disp_p): For PE-COFF do
10182 not return true for external weak function symbols in medium model.
10184 2021-10-01 Jakub Jelinek <jakub@redhat.com>
10186 * tree.h (OMP_CLAUSE_ORDER_REPRODUCIBLE): Define.
10187 * tree-pretty-print.c (dump_omp_clause) <case OMP_CLAUSE_ORDER>: Print
10188 reproducible: for OMP_CLAUSE_ORDER_REPRODUCIBLE.
10189 * omp-general.c (omp_extract_for_data): If OMP_CLAUSE_ORDER is seen
10190 without OMP_CLAUSE_ORDER_UNCONSTRAINED, overwrite sched_kind to
10191 OMP_CLAUSE_SCHEDULE_STATIC.
10193 2021-10-01 Richard Biener <rguenther@suse.de>
10195 PR middle-end/102518
10196 * tree-inline.c (setup_one_parameter): Avoid substituting
10197 an invariant into contexts where a GIMPLE register is not valid.
10199 2021-09-30 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
10201 * config/arm/arm-cpus.in: Add Cortex-R52+ CPU.
10202 * config/arm/arm-tables.opt: Regenerate.
10203 * config/arm/arm-tune.md: Regenerate.
10204 * doc/invoke.texi: Update docs.
10206 2021-09-30 Uroš Bizjak <ubizjak@gmail.com>
10209 * config/i386/i386.md
10210 (sign_extend:WIDE (any_logic:NARROW (memory, immediate)) splitters):
10213 2021-09-30 Tobias Burnus <tobias@codesourcery.com>
10215 * omp-low.c (omp_runtime_api_call): Add omp_aligned_{,c}alloc and
10216 omp_{c,re}alloc, fix omp_alloc/omp_free.
10218 2021-09-30 Martin Liska <mliska@suse.cz>
10220 * defaults.h (ASM_OUTPUT_ASCII): Do not hide global variable
10221 asm_out_file and stream directly to MYFILE.
10223 2021-09-30 Richard Biener <rguenther@suse.de>
10225 * tree-vect-data-refs.c (vect_update_misalignment_for_peel):
10226 Restore and fix condition under which we apply npeel to
10227 the DRs misalignment value.
10229 2021-09-30 Richard Biener <rguenther@suse.de>
10231 * tree-vect-data-refs.c (vect_update_misalignment_for_peel):
10232 Fix npeel check for variable amount of peeling.
10234 2021-09-30 Aldy Hernandez <aldyh@redhat.com>
10236 * lto-wrapper.c (run_gcc): Plug snprintf overflow.
10238 2021-09-30 Aldy Hernandez <aldyh@redhat.com>
10240 * gimple-range.cc (gimple_ranger::debug): New.
10241 * gimple-range.h (class gimple_ranger): Add debug.
10243 2021-09-30 Aldy Hernandez <aldyh@redhat.com>
10245 PR middle-end/102519
10246 * tree-vrp.c (hybrid_threader::~hybrid_threader): Free m_query.
10248 2021-09-29 Indu Bhagat <indu.bhagat@oracle.com>
10251 * btfout.c (GTY): Add GTY (()) albeit for cosmetic only purpose.
10252 (btf_finalize): Empty the hash_map btf_var_ids.
10254 2021-09-29 Aldy Hernandez <aldyh@redhat.com>
10256 * tree-vrp.c (thread_through_all_blocks): Return bool.
10257 (execute_vrp_threader): Return TODO_* flags.
10258 (pass_data_vrp_threader): Set todo_flags_finish to 0.
10260 2021-09-29 Aldy Hernandez <aldyh@redhat.com>
10262 * timevar.def (TV_TREE_VRP_THREADER): New.
10263 * tree-vrp.c: Use TV_TREE_VRP_THREADER for VRP threader pass.
10265 2021-09-29 David Faust <david.faust@oracle.com>
10267 * config.gcc (bpf-*-*): Do not overwrite extra_headers.
10269 2021-09-29 Jonathan Wright <jonathan.wright@arm.com>
10271 * config/aarch64/aarch64-builtins.c (TYPES_BINOP_PPU): Define
10272 new type qualifier enum.
10273 (TYPES_TERNOP_SSSU): Likewise.
10274 (TYPES_TERNOP_PPPU): Likewise.
10275 * config/aarch64/aarch64-simd-builtins.def: Define PPU, SSU,
10276 PPPU and SSSU builtin generator macros for qtbl1 and qtbx1
10278 * config/aarch64/arm_neon.h (vqtbl1_p8): Use type-qualified
10279 builtin and remove casts.
10280 (vqtbl1_s8): Likewise.
10281 (vqtbl1q_p8): Likewise.
10282 (vqtbl1q_s8): Likewise.
10283 (vqtbx1_s8): Likewise.
10284 (vqtbx1_p8): Likewise.
10285 (vqtbx1q_s8): Likewise.
10286 (vqtbx1q_p8): Likewise.
10287 (vtbl1_p8): Likewise.
10288 (vtbl2_p8): Likewise.
10289 (vtbx2_p8): Likewise.
10291 2021-09-29 Richard Biener <rguenther@suse.de>
10293 * tree-vect-data-refs.c (vect_dr_misalign_for_aligned_access):
10295 (vect_update_misalignment_for_peel): Use it to update
10296 misaligned to the value necessary for an aligned access.
10297 (vect_get_peeling_costs_all_drs): Likewise.
10298 (vect_enhance_data_refs_alignment): Likewise.
10300 2021-09-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10302 * config/aarch64/aarch64.c (aarch64_expand_cpymem): Count number of
10303 emitted operations and adjust heuristic for code size.
10305 2021-09-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10307 * config/aarch64/aarch64.c (aarch64_expand_setmem): Count number of
10308 emitted operations and adjust heuristic for code size.
10310 2021-09-29 Jakub Jelinek <jakub@redhat.com>
10312 PR middle-end/102504
10313 * gimplify.c (gimplify_scan_omp_clauses): Use omp_check_private even
10314 in OMP_SCOPE clauses, not just on worksharing construct clauses.
10316 2021-09-28 Geng Qi <gengqi@linux.alibaba.com>
10318 * config/riscv/riscv.md (mulv<mode>4): Call gen_smul<mode>3_highpart.
10319 (<u>mulditi3): Call <su>muldi3_highpart.
10320 (<u>muldi3_highpart): Rename to <su>muldi3_highpart.
10321 (<u>mulsidi3): Call <su>mulsi3_highpart.
10322 (<u>mulsi3_highpart): Rename to <su>mulsi3_highpart.
10324 2021-09-28 Iain Sandoe <iain@sandoe.co.uk>
10326 * config/darwin.h (DSYMUTIL_SPEC): Recognize D sources.
10328 2021-09-28 Iain Sandoe <iain@sandoe.co.uk>
10330 * config/rs6000/darwin.h (FIXED_R13): Add for PPC64.
10331 (FIRST_SAVED_GP_REGNO): Save from R13 even when it is one
10334 2021-09-28 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10336 * config/aarch64/aarch64.h (AARCH64_FL_LS64): Define
10337 (AARCH64_FL_V8_7): Likewise.
10338 (AARCH64_FL_FOR_ARCH8_7): Likewise.
10339 * config/aarch64/aarch64-arches.def (armv8.7-a): Define.
10340 * config/aarch64/aarch64-option-extensions.def (ls64): Define.
10341 * doc/invoke.texi: Document the above.
10343 2021-09-28 Aldy Hernandez <aldyh@redhat.com>
10345 * dbgcnt.c (dbg_cnt_counter): New.
10346 * dbgcnt.h (dbg_cnt_counter): New.
10347 * dumpfile.c (dump_options): Add entry for TDF_THREADING.
10348 * dumpfile.h (enum dump_flag): Add TDF_THREADING.
10349 * gimple-range-path.cc (DEBUG_SOLVER): Use TDF_THREADING.
10350 * tree-ssa-threadupdate.c (dump_jump_thread_path): Dump out
10353 2021-09-28 Aldy Hernandez <aldyh@redhat.com>
10355 * cfgcleanup.c (pass_jump::execute): Check
10356 flag_expensive_optimizations.
10357 (pass_jump_after_combine::gate): Same.
10358 * doc/invoke.texi (-fthread-jumps): Enable for -O1.
10359 * opts.c (default_options_table): Enable -fthread-jumps at -O1.
10360 * tree-ssa-threadupdate.c
10361 (fwd_jt_path_registry::remove_jump_threads_including): Bail unless
10364 2021-09-28 Ilya Leoshkevich <iii@linux.ibm.com>
10366 * tree-ssa-reassoc.c (biased_names): New global.
10367 (propagate_bias_p): New function.
10368 (loop_carried_phi): Remove.
10369 (propagate_rank): Propagate bias along single uses.
10370 (get_rank): Update biased_names when needed.
10372 2021-09-28 Ilya Leoshkevich <iii@linux.ibm.com>
10374 * passes.def (pass_reassoc): Rename parameter to early_p.
10375 * tree-ssa-reassoc.c (reassoc_bias_loop_carried_phi_ranks_p):
10377 (phi_rank): Don't bias loop-carried phi ranks
10378 before vectorization pass.
10379 (execute_reassoc): Add bias_loop_carried_phi_ranks_p parameter.
10380 (pass_reassoc::pass_reassoc): Add bias_loop_carried_phi_ranks_p
10382 (pass_reassoc::set_param): Set bias_loop_carried_phi_ranks_p
10384 (pass_reassoc::execute): Pass bias_loop_carried_phi_ranks_p to
10386 (pass_reassoc::bias_loop_carried_phi_ranks_p): New member.
10388 2021-09-28 Jakub Jelinek <jakub@redhat.com>
10391 * config/i386/i386.c (standard_80387_constant_p): Don't recognize
10392 special 80387 instruction XFmode constants if flag_rounding_math.
10394 2021-09-28 Richard Biener <rguenther@suse.de>
10396 PR tree-optimization/100112
10397 * tree-ssa-sccvn.c (visit_reference_op_load): Record the
10398 referece into the hashtable twice in case last_vuse is
10399 different from the original vuse on the stmt.
10401 2021-09-28 Jakub Jelinek <jakub@redhat.com>
10403 PR middle-end/102492
10404 * gimplify.c (gimplify_adjust_omp_clauses_1): Don't call the
10405 omp_finish_clause langhook on implicitly added OMP_CLAUSE_PRIVATE
10406 clauses on SIMD constructs.
10408 2021-09-28 Aldy Hernandez <aldyh@redhat.com>
10410 PR tree-optimization/102511
10411 * gimple-range-path.cc (path_range_query::range_on_path_entry):
10412 Return VARYING when nothing found.
10414 2021-09-28 Hongyu Wang <hongyu.wang@intel.com>
10417 * config/i386/i386.h (VALID_AVX512FP16_REG_MODE): Add
10419 (VALID_SSE2_REG_VHF_MODE): Add V4HFmode and V2HFmode.
10420 (VALID_MMX_REG_MODE): Add V4HFmode.
10421 (SSE_REG_MODE_P): Replace VALID_AVX512FP16_REG_MODE with
10422 vector mode condition.
10423 * config/i386/i386.c (classify_argument): Parse V4HF/V2HF
10425 (function_arg_32): Add V4HFmode.
10426 (function_arg_advance_32): Likewise.
10427 * config/i386/i386.md (mode): Add V4HF/V2HF.
10428 (MODE_SIZE): Likewise.
10429 * config/i386/mmx.md (MMXMODE): Add V4HF mode.
10430 (V_32): Add V2HF mode.
10431 (VHF_32_64): New mode iterator.
10432 (*mov<mode>_internal): Adjust sse alternatives to support
10434 (*mov<mode>_internal): Adjust sse alternatives to support
10436 (<insn><mode>3): New define_insn for add/sub/mul/div.
10438 2021-09-28 Aldy Hernandez <aldyh@redhat.com>
10440 * tree-ssa-threadbackward.c (pass_thread_jumps::gate): Check
10442 (pass_early_thread_jumps::gate): Same.
10443 * tree-ssa-threadedge.c (jump_threader::thread_outgoing_edges):
10444 Return if !flag_thread_jumps.
10445 * tree-ssa-threadupdate.c
10446 (jt_path_registry::register_jump_thread): Assert that
10447 flag_thread_jumps is true.
10449 2021-09-28 liuhongt <hongtao.liu@intel.com>
10452 (simplify_context::simplify_binary_operation_1): Relax
10453 condition of simplifying (vec_concat:M (vec_select op0
10454 index0)(vec_select op1 index1)) to allow different modes
10455 between op0 and M, but have same inner mode.
10457 2021-09-28 liuhongt <hongtao.liu@intel.com>
10459 * config/i386/i386-expand.c (emit_reduc_half): Handle
10460 V8HF/V16HF/V32HFmode.
10461 * config/i386/sse.md (REDUC_SSE_PLUS_MODE): Add V8HF.
10462 (REDUC_SSE_SMINMAX_MODE): Ditto.
10463 (REDUC_PLUS_MODE): Add V16HF and V32HF.
10464 (REDUC_SMINMAX_MODE): Ditto.
10466 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
10468 * gimple-range-path.cc
10469 (path_range_query::precompute_ranges_in_block): Rename to...
10470 (path_range_query::compute_ranges_in_block): ...this.
10471 (path_range_query::precompute_ranges): Rename to...
10472 (path_range_query::compute_ranges): ...this.
10473 (path_range_query::precompute_relations): Rename to...
10474 (path_range_query::compute_relations): ...this.
10475 (path_range_query::precompute_phi_relations): Rename to...
10476 (path_range_query::compute_phi_relations): ...this.
10477 * gimple-range-path.h: Rename precompute* to compute*.
10478 * tree-ssa-threadbackward.c
10479 (back_threader::find_taken_edge_switch): Same.
10480 (back_threader::find_taken_edge_cond): Same.
10481 * tree-ssa-threadedge.c
10482 (hybrid_jt_simplifier::compute_ranges_from_state): Same.
10483 (hybrid_jt_state::register_equivs_stmt): Inline...
10484 * tree-ssa-threadedge.h: ...here.
10486 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
10488 * tree-vrp.c (lhs_of_dominating_assert): Remove.
10489 (class vrp_jt_state): Remove.
10490 (class vrp_jt_simplifier): Remove.
10491 (vrp_jt_simplifier::simplify): Remove.
10492 (class vrp_jump_threader): Remove.
10493 (vrp_jump_threader::vrp_jump_threader): Remove.
10494 (vrp_jump_threader::~vrp_jump_threader): Remove.
10495 (vrp_jump_threader::before_dom_children): Remove.
10496 (vrp_jump_threader::after_dom_children): Remove.
10498 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
10500 * passes.def (pass_vrp_threader): New.
10501 * tree-pass.h (make_pass_vrp_threader): Add make_pass_vrp_threader.
10502 * tree-ssa-threadedge.c (hybrid_jt_state::register_equivs_stmt): New.
10503 (hybrid_jt_simplifier::hybrid_jt_simplifier): New.
10504 (hybrid_jt_simplifier::simplify): New.
10505 (hybrid_jt_simplifier::compute_ranges_from_state): New.
10506 * tree-ssa-threadedge.h (class hybrid_jt_state): New.
10507 (class hybrid_jt_simplifier): New.
10508 * tree-vrp.c (execute_vrp): Remove ASSERT_EXPR based jump
10510 (class hybrid_threader): New.
10511 (hybrid_threader::hybrid_threader): New.
10512 (hybrid_threader::~hybrid_threader): New.
10513 (hybrid_threader::before_dom_children): New.
10514 (hybrid_threader::after_dom_children): New.
10515 (execute_vrp_threader): New.
10516 (class pass_vrp_threader): New.
10517 (make_pass_vrp_threader): New.
10519 2021-09-27 Martin Liska <mliska@suse.cz>
10521 * output.h (enum section_flag): New.
10522 (SECTION_FORGET): Remove.
10523 (SECTION_ENTSIZE): Make it (1UL << 8) - 1.
10524 (SECTION_STYLE_MASK): Define it based on other enum
10526 * varasm.c (switch_to_section): Remove unused handling of
10529 2021-09-27 Martin Liska <mliska@suse.cz>
10531 * common.opt: Add new variable flag_default_complex_method.
10532 * opts.c (finish_options): Handle flags related to
10533 x_flag_complex_method.
10534 * toplev.c (process_options): Remove option handling related
10535 to flag_complex_method.
10537 2021-09-27 Richard Biener <rguenther@suse.de>
10539 PR middle-end/102450
10540 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid using
10541 type_for_size, instead use int_mode_for_size.
10543 2021-09-27 Andrew Pinski <apinski@marvell.com>
10546 * gimplify.c (gimplify_save_expr): Return early
10547 if the type of val is error_mark_node.
10549 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
10551 * tree-ssanames.c (ssa_name_has_boolean_range): Use
10554 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
10556 * gimple-ssa-evrp-analyze.h (class evrp_range_analyzer): Remove
10557 vrp_visit_cond_stmt.
10558 * tree-ssa-dom.c (cprop_operand): Convert to range_query API.
10559 (cprop_into_stmt): Same.
10560 (dom_opt_dom_walker::optimize_stmt): Same.
10562 2021-09-27 Richard Biener <rguenther@suse.de>
10564 PR tree-optimization/97351
10565 PR tree-optimization/97352
10566 PR tree-optimization/82426
10567 * tree-vectorizer.h (dr_misalignment): Add vector type
10569 (aligned_access_p): Likewise.
10570 (known_alignment_for_access_p): Likewise.
10571 (vect_supportable_dr_alignment): Likewise.
10572 (vect_known_alignment_in_bytes): Likewise. Refactor.
10573 (DR_MISALIGNMENT): Remove.
10574 (vect_update_shared_vectype): Likewise.
10575 * tree-vect-data-refs.c (dr_misalignment): Refactor, handle
10576 a vector type with larger alignment requirement and apply
10577 the negative step adjustment here.
10578 (vect_calculate_target_alignment): Remove.
10579 (vect_compute_data_ref_alignment): Get explicit vector type
10580 argument, do not apply a negative step alignment adjustment
10582 (vect_slp_analyze_node_alignment): Re-analyze alignment
10583 when we re-visit the DR with a bigger desired alignment but
10584 keep more precise results from smaller alignments.
10585 * tree-vect-slp.c (vect_update_shared_vectype): Remove.
10586 (vect_slp_analyze_node_operations_1): Do not update the
10587 shared vector type on stmts.
10588 * tree-vect-stmts.c (vect_analyze_stmt): Push/pop the
10589 vector type of an SLP node to the representative stmt-info.
10590 (vect_transform_stmt): Likewise.
10592 2021-09-27 liuhongt <hongtao.liu@intel.com>
10595 2021-09-09 liuhongt <hongtao.liu@intel.com>
10598 * config/i386/sse.md (reduc_plus_scal_<mode>): Split to ..
10599 (reduc_plus_scal_v4sf): .. this, New define_expand.
10600 (reduc_plus_scal_v2df): .. and this, New define_expand.
10602 2021-09-26 liuhongt <hongtao.liu@intel.com>
10604 * doc/extend.texi (Half-Precision): Remove storage only
10605 description for _Float16 w/o avx512fp16.
10607 2021-09-25 Dimitar Dimitrov <dimitar@dinux.eu>
10609 * config/pru/constraints.md (Rrio): New constraint.
10610 * config/pru/predicates.md (regio_operand): New predicate.
10611 * config/pru/pru-pragma.c (pru_register_pragmas): Register
10612 the __regio_symbol address space.
10613 * config/pru/pru-protos.h (pru_symref2ioregno): Declaration.
10614 * config/pru/pru.c (pru_symref2ioregno): New helper function.
10615 (pru_legitimate_address_p): Remove.
10616 (pru_addr_space_legitimate_address_p): Use the address space
10617 aware hook variant.
10618 (pru_nongeneric_pointer_addrspace): New helper function.
10619 (pru_insert_attributes): New function to validate __regio_symbol
10621 (TARGET_INSERT_ATTRIBUTES): New macro.
10622 (TARGET_LEGITIMATE_ADDRESS_P): Remove.
10623 (TARGET_ADDR_SPACE_LEGITIMATE_ADDRESS_P): New macro.
10624 * config/pru/pru.h (enum reg_class): Add REGIO_REGS class.
10625 * config/pru/pru.md (*regio_readsi): New pattern to read I/O
10627 (*regio_nozext_writesi): New pattern to write to I/O registers.
10628 (*regio_zext_write_r30<EQS0:mode>): Ditto.
10629 * doc/extend.texi: Document the new PRU Named Address Space.
10631 2021-09-24 Patrick Palka <ppalka@redhat.com>
10635 * real.c (encode_ieee_double): Avoid unwanted sign extension.
10636 (encode_ieee_quad): Likewise.
10638 2021-09-24 Vladimir Makarov <vmakarov@redhat.com>
10640 PR rtl-optimization/102147
10641 * ira-build.c (ira_conflict_vector_profitable_p): Make
10642 profitability calculation independent of host compiler pointer and
10643 IRA_INT_BITS sizes.
10645 2021-09-24 Aldy Hernandez <aldyh@redhat.com>
10647 * gimple-range-path.cc (path_range_query::path_range_query):
10648 Move debugging header...
10649 (path_range_query::precompute_ranges): ...here.
10650 (path_range_query::internal_range_of_expr): Do not call
10651 range_on_path_entry if NAME is defined in the current block.
10653 2021-09-24 Richard Biener <rguenther@suse.de>
10655 * cfghooks.c (verify_flow_info): Verify unallocated BB and
10656 edge flags are not set.
10658 2021-09-24 Aldy Hernandez <aldyh@redhat.com>
10660 * tree-ssa-threadupdate.c (jt_path_registry::cancel_invalid_paths):
10662 (jt_path_registry::register_jump_thread): Call
10663 cancel_invalid_paths.
10664 * tree-ssa-threadupdate.h (class jt_path_registry): Add
10665 cancel_invalid_paths.
10667 2021-09-24 Feng Xue <fxue@os.amperecomputing.com>
10669 PR tree-optimization/102400
10670 * tree-ssa-sccvn.c (vn_reference_insert_pieces): Initialize
10671 result_vdef to zero value.
10673 2021-09-24 Feng Xue <fxue@os.amperecomputing.com>
10675 PR tree-optimization/102451
10676 * tree-ssa-dse.c (delete_dead_or_redundant_call): Record bb of stmt
10679 2021-09-24 Hongyu Wang <hongyu.wang@intel.com>
10681 * config/i386/sse.md (cond_<insn><mode>): Extend to support
10683 (cond_mul<mode>): Likewise.
10684 (cond_div<mode>): Likewise.
10685 (cond_<code><mode>): Likewise.
10686 (cond_fma<mode>): Likewise.
10687 (cond_fms<mode>): Likewise.
10688 (cond_fnma<mode>): Likewise.
10689 (cond_fnms<mode>): Likewise.
10691 2021-09-23 Andrew MacLeod <amacleod@redhat.com>
10693 PR tree-optimization/102463
10694 * gimple-range-fold.cc (fold_using_range::relation_fold_and_or): If
10695 there is no range-ops handler, don't look for a relation.
10697 2021-09-23 Andrew MacLeod <amacleod@redhat.com>
10699 * gimple-range-cache.cc (ranger_cache::ranger_cache): Take
10700 non-executable_edge flag as parameter.
10701 * gimple-range-cache.h (ranger_cache): Adjust prototype.
10702 * gimple-range-gori.cc (gori_compute::gori_compute): Take
10703 non-executable_edge flag as parameter.
10704 (gori_compute::outgoing_edge_range_p): Check new flag.
10705 * gimple-range-gori.h (gori_compute): Adjust prototype.
10706 * gimple-range.cc (gimple_ranger::gimple_ranger): Create new flag.
10707 (gimple_ranger::range_on_edge): Check new flag.
10708 * gimple-range.h (gimple_ranger::non_executable_edge_flag): New.
10709 * gimple-ssa-evrp.c (rvrp_folder): Pass ranger flag to simplifer.
10710 (hybrid_folder::hybrid_folder): Set ranger non-executable flag value.
10711 (hybrid_folder::fold_stmt): Set flag value in the simplifer.
10712 * vr-values.c (simplify_using_ranges::set_and_propagate_unexecutable):
10713 Use not_executable flag if provided inmstead of EDGE_EXECUTABLE.
10714 (simplify_using_ranges::simplify_switch_using_ranges): Clear
10715 EDGE_EXECUTABLE like it originally did.
10716 (simplify_using_ranges::cleanup_edges_and_switches): Clear any
10717 NON_EXECUTABLE flags.
10718 (simplify_using_ranges::simplify_using_ranges): Adjust.
10719 * vr-values.h (class simplify_using_ranges): Adjust.
10720 (simplify_using_ranges::set_range_query): Add non-executable flag param.
10722 2021-09-23 Bill Schmidt <wschmidt@linux.ibm.com>
10725 * config/rs6000/rs6000-call.c (rs6000_aggregate_candidate): Detect
10726 zero-width bit fields and return indicator.
10727 (rs6000_discover_homogeneous_aggregate): Diagnose when the
10728 presence of a zero-width bit field changes parameter passing in
10731 2021-09-23 Aldy Hernandez <aldyh@redhat.com>
10733 * gimple-range-fold.cc (fold_using_range::range_of_phi):
10734 Remove dominator check.
10736 2021-09-23 Aldy Hernandez <aldyh@redhat.com>
10738 * gimple-range-path.cc (path_range_query::precompute_relations):
10739 Hoist edge calculations before using EDGE_SUCC.
10741 2021-09-23 Jonathan Wakely <jwakely@redhat.com>
10743 * configure.ac: Fix --with-multilib-list description.
10744 * configure: Regenerate.
10746 2021-09-23 Richard Biener <rguenther@suse.de>
10748 PR tree-optimization/102448
10749 * tree-vect-data-refs.c (vect_duplicate_ssa_name_ptr_info):
10750 Clear alignment info copied from DR_PTR_INFO.
10752 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
10754 * config/i386/i386-expand.c (ix86_use_mask_cmp_p): Enable
10756 * config/i386/sse.md (sseintvecmodelower): Add HF vector modes.
10757 (<avx512>_store<mode>_mask): Extend to support HF vector modes.
10758 (vec_cmp<mode><avx512fmaskmodelower>): Likewise.
10759 (vcond_mask_<mode><avx512fmaskmodelower>): Likewise.
10760 (vcond<mode><mode>): New expander.
10761 (vcond<mode><sseintvecmodelower>): Likewise.
10762 (vcond<sseintvecmodelower><mode>): Likewise.
10763 (vcondu<mode><sseintvecmodelower>): Likewise.
10765 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
10767 * config/i386/sse.md (extend<ssePHmodelower><mode>2):
10769 (extendv4hf<mode>2): Likewise.
10770 (extendv2hfv2df2): Likewise.
10771 (trunc<mode><ssePHmodelower>2): Likewise.
10772 (avx512fp16_vcvt<castmode>2ph_<mode>): Rename to ...
10773 (trunc<mode>v4hf2): ... this, and drop constraints.
10774 (avx512fp16_vcvtpd2ph_v2df): Rename to ...
10775 (truncv2dfv2hf2): ... this, and likewise.
10777 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
10779 * config/i386/sse.md (float<floatunssuffix><mode><ssePHmodelower>2):
10781 (avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>):
10783 (float<floatunssuffix><mode>v4hf2): ... this, and drop constraints.
10784 (avx512fp16_vcvt<floatsuffix>qq2ph_v2di): Rename to ...
10785 (float<floatunssuffix>v2div2hf2): ... this, and likewise.
10787 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
10789 * config/i386/i386.md (fix<fixunssuffix>_trunchf<mode>2): New expander.
10790 (fixuns_trunchfhi2): Likewise.
10791 (*fixuns_trunchfsi2zext): New define_insn.
10792 * config/i386/sse.md (ssePHmodelower): New mode_attr.
10793 (fix<fixunssuffix>_trunc<ssePHmodelower><mode>2):
10794 New expander for same element vector fix_truncate.
10795 (fix<fixunssuffix>_trunc<ssePHmodelower><mode>2):
10796 Likewise for V4HF to V4SI/V4DI fix_truncate.
10797 (fix<fixunssuffix>_truncv2hfv2di2):
10798 Likeise for V2HF to V2DI fix_truncate.
10800 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
10802 * config/i386/i386.md (<code>hf3): New expander.
10804 2021-09-23 liuhongt <hongtao.liu@intel.com>
10806 * config/i386/sse.md (FMAMODEM): extend to handle FP16.
10807 (VFH_SF_AVX512VL): Extend to handle HFmode.
10808 (VF_SF_AVX512VL): Deleted.
10810 2021-09-23 liuhongt <hongtao.liu@intel.com>
10812 * config/i386/i386.md (rinthf2): New expander.
10813 (nearbyinthf2): New expander.
10815 2021-09-23 Aldy Hernandez <aldyh@redhat.com>
10817 * tree-ssa-dom.c (class dom_jump_threader_simplifier): Rename...
10818 (class dom_jt_state): ...this and provide virtual overrides.
10819 (dom_jt_state::register_equiv): New.
10820 (class dom_jt_simplifier): Rename from
10821 dom_jump_threader_simplifier.
10822 (dom_jump_threader_simplifier::simplify): Rename...
10823 (dom_jt_simplifier::simplify): ...to this.
10824 (pass_dominator::execute): Use dom_jt_simplifier and
10826 * tree-ssa-threadedge.c (jump_threader::jump_threader):
10828 (jt_state::register_equivs_stmt): Abstract out...
10829 (jump_threader::record_temporary_equivalences_from_stmts_at_dest):
10831 (jump_threader::thread_around_empty_blocks): Update state.
10832 (jump_threader::thread_through_normal_block): Same.
10833 (jt_state::jt_state): Remove.
10834 (jt_state::push): Remove pass specific bits. Keep block vector
10836 (jt_state::append_path): New.
10837 (jt_state::pop): Remove pass specific bits.
10838 (jt_state::register_equiv): Same.
10839 (jt_state::record_ranges_from_stmt): Same.
10840 (jt_state::register_equivs_on_edge): Same. Rename...
10841 (jt_state::register_equivs_edge): ...to this.
10842 (jt_state::dump): New.
10843 (jt_state::debug): New.
10844 (jump_threader_simplifier::simplify): Remove.
10845 (jt_state::get_path): New.
10846 * tree-ssa-threadedge.h (class jt_simplifier): Make into a base
10847 class. Expose common functionality as virtual methods.
10848 (class jump_threader_simplifier): Same. Rename...
10849 (class jt_simplifier): ...to this.
10850 * tree-vrp.c (class vrp_jump_threader_simplifier): Rename...
10851 (class vrp_jt_simplifier): ...to this. Provide pass specific
10853 (class vrp_jt_state): New.
10854 (vrp_jump_threader_simplifier::simplify): Rename...
10855 (vrp_jt_simplifier::simplify): ...to this. Inline code from
10856 what used to be the base class.
10857 (vrp_jump_threader::vrp_jump_threader): Use vrp_jt_state and
10860 2021-09-22 Tobias Burnus <tobias@codesourcery.com>
10863 * doc/invoke.texi (-Wno-missing-include-dirs.): Document Fortran
10866 2021-09-22 Roger Sayle <roger@nextmovesoftware.com>
10867 Richard Biener <rguenther@suse.de>
10869 * match.pd (negation simplifications): Implement some negation
10870 folding transformations from fold-const.c's fold_negate_expr.
10871 * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Add a SIMPLIFY
10872 argument, to control whether the op should be simplified prior
10873 to looking up/assigning a value number.
10874 (vn_nary_build_or_lookup): Update call to vn_nary_build_or_lookup_1.
10875 (vn_nary_simplify): Likewise.
10876 (visit_nary_op): Likewise, but when constructing a NEGATE_EXPR
10877 now call vn_nary_build_or_lookup_1 disabling simplification.
10879 2021-09-22 Jiufu Guo <guojiufu@linux.ibm.com>
10881 PR tree-optimization/102087
10882 * tree-ssa-loop-niter.c (number_of_iterations_until_wrap):
10883 Update bound/cmp/control for niter.
10885 2021-09-22 Aldy Hernandez <aldyh@redhat.com>
10887 * gimple-range-fold.cc (fold_using_range::range_of_range_op):
10888 Move check for non-empty BB here.
10889 (fur_source::register_outgoing_edges): ...from here.
10891 2021-09-22 Aldy Hernandez <aldyh@redhat.com>
10893 * gimple-range-path.cc (path_range_query::internal_range_of_expr):
10894 Remove call to improve_range_with_equivs.
10895 (path_range_query::improve_range_with_equivs): Remove
10896 * gimple-range-path.h: Remove improve_range_with_equivs.
10898 2021-09-22 dianhong xu <dianhong.xu@intel.com>
10900 * config/i386/avx512fp16intrin.h:
10901 (_mm512_mask_blend_ph): New intrinsic.
10902 (_mm512_permutex2var_ph): Ditto.
10903 (_mm512_permutexvar_ph): Ditto.
10904 * config/i386/avx512fp16vlintrin.h:
10905 (_mm256_mask_blend_ph): New intrinsic.
10906 (_mm256_permutex2var_ph): Ditto.
10907 (_mm256_permutexvar_ph): Ditto.
10908 (_mm_mask_blend_ph): Ditto.
10909 (_mm_permutex2var_ph): Ditto.
10910 (_mm_permutexvar_ph): Ditto.
10912 2021-09-22 dianhong xu <dianhong.xu@intel.com>
10914 * config/i386/avx512fp16intrin.h: Add new intrinsics.
10915 (_mm512_conj_pch): New intrinsic.
10916 (_mm512_mask_conj_pch): Ditto.
10917 (_mm512_maskz_conj_pch): Ditto.
10918 * config/i386/avx512fp16vlintrin.h: Add new intrinsics.
10919 (_mm256_conj_pch): New intrinsic.
10920 (_mm256_mask_conj_pch): Ditto.
10921 (_mm256_maskz_conj_pch): Ditto.
10922 (_mm_conj_pch): Ditto.
10923 (_mm_mask_conj_pch): Ditto.
10924 (_mm_maskz_conj_pch): Ditto.
10926 2021-09-22 dianhong xu <dianhong.xu@intel.com>
10928 * config/i386/avx512fp16intrin.h (_MM512_REDUCE_OP): New macro
10929 (_mm512_reduce_add_ph): New intrinsic.
10930 (_mm512_reduce_mul_ph): Ditto.
10931 (_mm512_reduce_min_ph): Ditto.
10932 (_mm512_reduce_max_ph): Ditto.
10933 * config/i386/avx512fp16vlintrin.h
10934 (_MM256_REDUCE_OP/_MM_REDUCE_OP): New macro.
10935 (_mm256_reduce_add_ph): New intrinsic.
10936 (_mm256_reduce_mul_ph): Ditto.
10937 (_mm256_reduce_min_ph): Ditto.
10938 (_mm256_reduce_max_ph): Ditto.
10939 (_mm_reduce_add_ph): Ditto.
10940 (_mm_reduce_mul_ph): Ditto.
10941 (_mm_reduce_min_ph): Ditto.
10942 (_mm_reduce_max_ph): Ditto.
10944 2021-09-22 dianhong xu <dianhong.xu@intel.com>
10946 * config/i386/avx512fp16intrin.h (__m512h_u, __m256h_u,
10947 __m128h_u): New typedef.
10948 (_mm512_load_ph): New intrinsic.
10949 (_mm256_load_ph): Ditto.
10950 (_mm_load_ph): Ditto.
10951 (_mm512_loadu_ph): Ditto.
10952 (_mm256_loadu_ph): Ditto.
10953 (_mm_loadu_ph): Ditto.
10954 (_mm512_store_ph): Ditto.
10955 (_mm256_store_ph): Ditto.
10956 (_mm_store_ph): Ditto.
10957 (_mm512_storeu_ph): Ditto.
10958 (_mm256_storeu_ph): Ditto.
10959 (_mm_storeu_ph): Ditto.
10960 (_mm512_abs_ph): Ditto.
10961 * config/i386/avx512fp16vlintrin.h
10962 (_mm_abs_ph): Ditto.
10963 (_mm256_abs_ph): Ditto.
10965 2021-09-22 Andreas Krebbel <krebbel@linux.ibm.com>
10967 * config/s390/tpf.md (prologue_tpf, epilogue_tpf): Add cc clobber.
10969 2021-09-22 Andreas Krebbel <krebbel@linux.ibm.com>
10972 * config/s390/s390.c (s390_expand_insv): Emit a normal move if it
10973 is actually a full copy of the source operand into the target.
10974 Don't emit a strict low part move if source and target mode match.
10976 2021-09-22 Jakub Jelinek <jakub@redhat.com>
10978 PR middle-end/102415
10979 * omp-expand.c (expand_omp_single): If region->exit is NULL,
10980 assert region->entry is GIMPLE_OMP_SCOPE region and return.
10982 2021-09-22 Jakub Jelinek <jakub@redhat.com>
10984 * tree.h (OMP_CLAUSE_ALLOCATE_ALIGN): Define.
10985 * tree.c (omp_clause_num_ops): Change number of OMP_CLAUSE_ALLOCATE
10986 arguments from 2 to 3.
10987 * tree-pretty-print.c (dump_omp_clause): Print allocator() around
10988 allocate clause allocator and print align if present.
10989 * omp-low.c (scan_sharing_clauses): Force allocate_map entry even
10990 for omp_default_mem_alloc if align modifier is present. If align
10991 modifier is present, use TREE_LIST to encode both allocator and
10993 (lower_private_allocate, lower_rec_input_clauses, create_task_copyfn):
10994 Handle align modifier on allocator clause if present.
10996 2021-09-22 liuhongt <hongtao.liu@intel.com>
10998 * config/i386/i386.md (define_attr "isa"): Add
11000 (define_attr "enabled"): Correspond fma_or_avx512vl to
11001 TARGET_FMA || TARGET_AVX512VL.
11002 * config/i386/mmx.md (fmav2sf4): Extend to AVX512 fma.
11004 (fnmav2sf4): Ditto.
11005 (fnmsv2sf4): Ditto.
11007 2021-09-22 liuhongt <hongtao.liu@intel.com>
11009 * config/i386/i386.md (cstorehf3): New define_expand.
11011 2021-09-22 liuhongt <hongtao.liu@intel.com>
11013 * config/i386/i386.md (<rounding_insn>hf2): New expander.
11014 (sse4_1_round<mode>2): Extend from MODEF to MODEFH.
11015 * config/i386/sse.md (*sse4_1_round<ssescalarmodesuffix>):
11016 Extend from VF_128 to VFH_128.
11018 2021-09-22 liuhongt <hongtao.liu@intel.com>
11020 * config/i386/i386-features.c (i386-features.c): Handle
11022 * config/i386/i386.md (sqrthf2): New expander.
11023 (*sqrthf2): New define_insn.
11024 * config/i386/sse.md
11025 (*<sse>_vmsqrt<mode>2<mask_scalar_name><round_scalar_name>):
11028 2021-09-22 liuhongt <hongtao.liu@intel.com>
11030 * config/i386/avx512fp16intrin.h (_mm_mask_fcmadd_sch):
11032 (_mm_mask3_fcmadd_sch): Likewise.
11033 (_mm_maskz_fcmadd_sch): Likewise.
11034 (_mm_fcmadd_sch): Likewise.
11035 (_mm_mask_fmadd_sch): Likewise.
11036 (_mm_mask3_fmadd_sch): Likewise.
11037 (_mm_maskz_fmadd_sch): Likewise.
11038 (_mm_fmadd_sch): Likewise.
11039 (_mm_mask_fcmadd_round_sch): Likewise.
11040 (_mm_mask3_fcmadd_round_sch): Likewise.
11041 (_mm_maskz_fcmadd_round_sch): Likewise.
11042 (_mm_fcmadd_round_sch): Likewise.
11043 (_mm_mask_fmadd_round_sch): Likewise.
11044 (_mm_mask3_fmadd_round_sch): Likewise.
11045 (_mm_maskz_fmadd_round_sch): Likewise.
11046 (_mm_fmadd_round_sch): Likewise.
11047 (_mm_fcmul_sch): Likewise.
11048 (_mm_mask_fcmul_sch): Likewise.
11049 (_mm_maskz_fcmul_sch): Likewise.
11050 (_mm_fmul_sch): Likewise.
11051 (_mm_mask_fmul_sch): Likewise.
11052 (_mm_maskz_fmul_sch): Likewise.
11053 (_mm_fcmul_round_sch): Likewise.
11054 (_mm_mask_fcmul_round_sch): Likewise.
11055 (_mm_maskz_fcmul_round_sch): Likewise.
11056 (_mm_fmul_round_sch): Likewise.
11057 (_mm_mask_fmul_round_sch): Likewise.
11058 (_mm_maskz_fmul_round_sch): Likewise.
11059 * config/i386/i386-builtin.def: Add corresponding new builtins.
11060 * config/i386/sse.md
11061 (avx512fp16_fmaddcsh_v8hf_maskz<round_expand_name>): New expander.
11062 (avx512fp16_fcmaddcsh_v8hf_maskz<round_expand_name>): Ditto.
11063 (avx512fp16_fma_<complexopname>sh_v8hf<mask_scalarcz_name><round_scalarcz_name>):
11065 (avx512fp16_<complexopname>sh_v8hf_mask<round_name>): Ditto.
11066 (avx512fp16_<complexopname>sh_v8hf<mask_scalarc_name><round_scalarcz_name>):
11068 * config/i386/subst.md (mask_scalarcz_name): New.
11069 (mask_scalarc_name): Ditto.
11070 (mask_scalarc_operand3): Ditto.
11071 (mask_scalarcz_operand4): Ditto.
11072 (round_scalarcz_name): Ditto.
11073 (round_scalarc_mask_operand3): Ditto.
11074 (round_scalarcz_mask_operand4): Ditto.
11075 (round_scalarc_mask_op3): Ditto.
11076 (round_scalarcz_mask_op4): Ditto.
11077 (round_scalarcz_constraint): Ditto.
11078 (round_scalarcz_nimm_predicate): Ditto.
11079 (mask_scalarcz): Ditto.
11080 (mask_scalarc): Ditto.
11081 (round_scalarcz): Ditto.
11083 2021-09-22 liuhongt <hongtao.liu@intel.com>
11085 * config/i386/avx512fp16intrin.h (_mm512_fcmadd_pch):
11087 (_mm512_mask_fcmadd_pch): Likewise.
11088 (_mm512_mask3_fcmadd_pch): Likewise.
11089 (_mm512_maskz_fcmadd_pch): Likewise.
11090 (_mm512_fmadd_pch): Likewise.
11091 (_mm512_mask_fmadd_pch): Likewise.
11092 (_mm512_mask3_fmadd_pch): Likewise.
11093 (_mm512_maskz_fmadd_pch): Likewise.
11094 (_mm512_fcmadd_round_pch): Likewise.
11095 (_mm512_mask_fcmadd_round_pch): Likewise.
11096 (_mm512_mask3_fcmadd_round_pch): Likewise.
11097 (_mm512_maskz_fcmadd_round_pch): Likewise.
11098 (_mm512_fmadd_round_pch): Likewise.
11099 (_mm512_mask_fmadd_round_pch): Likewise.
11100 (_mm512_mask3_fmadd_round_pch): Likewise.
11101 (_mm512_maskz_fmadd_round_pch): Likewise.
11102 (_mm512_fcmul_pch): Likewise.
11103 (_mm512_mask_fcmul_pch): Likewise.
11104 (_mm512_maskz_fcmul_pch): Likewise.
11105 (_mm512_fmul_pch): Likewise.
11106 (_mm512_mask_fmul_pch): Likewise.
11107 (_mm512_maskz_fmul_pch): Likewise.
11108 (_mm512_fcmul_round_pch): Likewise.
11109 (_mm512_mask_fcmul_round_pch): Likewise.
11110 (_mm512_maskz_fcmul_round_pch): Likewise.
11111 (_mm512_fmul_round_pch): Likewise.
11112 (_mm512_mask_fmul_round_pch): Likewise.
11113 (_mm512_maskz_fmul_round_pch): Likewise.
11114 * config/i386/avx512fp16vlintrin.h (_mm_fmadd_pch):
11116 (_mm_mask_fmadd_pch): Likewise.
11117 (_mm_mask3_fmadd_pch): Likewise.
11118 (_mm_maskz_fmadd_pch): Likewise.
11119 (_mm256_fmadd_pch): Likewise.
11120 (_mm256_mask_fmadd_pch): Likewise.
11121 (_mm256_mask3_fmadd_pch): Likewise.
11122 (_mm256_maskz_fmadd_pch): Likewise.
11123 (_mm_fcmadd_pch): Likewise.
11124 (_mm_mask_fcmadd_pch): Likewise.
11125 (_mm_mask3_fcmadd_pch): Likewise.
11126 (_mm_maskz_fcmadd_pch): Likewise.
11127 (_mm256_fcmadd_pch): Likewise.
11128 (_mm256_mask_fcmadd_pch): Likewise.
11129 (_mm256_mask3_fcmadd_pch): Likewise.
11130 (_mm256_maskz_fcmadd_pch): Likewise.
11131 (_mm_fmul_pch): Likewise.
11132 (_mm_mask_fmul_pch): Likewise.
11133 (_mm_maskz_fmul_pch): Likewise.
11134 (_mm256_fmul_pch): Likewise.
11135 (_mm256_mask_fmul_pch): Likewise.
11136 (_mm256_maskz_fmul_pch): Likewise.
11137 (_mm_fcmul_pch): Likewise.
11138 (_mm_mask_fcmul_pch): Likewise.
11139 (_mm_maskz_fcmul_pch): Likewise.
11140 (_mm256_fcmul_pch): Likewise.
11141 (_mm256_mask_fcmul_pch): Likewise.
11142 (_mm256_maskz_fcmul_pch): Likewise.
11143 * config/i386/i386-builtin-types.def (V8HF_FTYPE_V8HF_V8HF_V8HF,
11144 V8HF_FTYPE_V16HF_V16HF_V16HF, V16HF_FTYPE_V16HF_V16HF_V16HF_UQI,
11145 V32HF_FTYPE_V32HF_V32HF_V32HF_INT,
11146 V32HF_FTYPE_V32HF_V32HF_V32HF_UHI_INT): Add new builtin types.
11147 * config/i386/i386-builtin.def: Add new builtins.
11148 * config/i386/i386-expand.c: Handle new builtin types.
11149 * config/i386/subst.md (SUBST_CV): New.
11150 (maskc_name): Ditto.
11151 (maskc_operand3): Ditto.
11153 (sdc_maskz_name): Ditto.
11154 (sdc_mask_op4): Ditto.
11155 (sdc_mask_op5): Ditto.
11156 (sdc_mask_mode512bit_condition): Ditto.
11158 (round_maskc_operand3): Ditto.
11159 (round_sdc_mask_operand4): Ditto.
11160 (round_maskc_op3): Ditto.
11161 (round_sdc_mask_op4): Ditto.
11162 (round_saeonly_sdc_mask_operand5): Ditto.
11163 * config/i386/sse.md (unspec): Add complex fma unspecs.
11164 (avx512fmaskcmode): New.
11165 (UNSPEC_COMPLEX_F_C_MA): Ditto.
11166 (UNSPEC_COMPLEX_F_C_MUL): Ditto.
11167 (complexopname): Ditto.
11168 (<avx512>_fmaddc_<mode>_maskz<round_expand_name>): New expander.
11169 (<avx512>_fcmaddc_<mode>_maskz<round_expand_name>): Ditto.
11170 (fma_<complexopname>_<mode><sdc_maskz_name><round_name>): New
11172 (<avx512>_<complexopname>_<mode>_mask<round_name>): Ditto.
11173 (<avx512>_<complexopname>_<mode><maskc_name><round_name>): Ditto.
11175 2021-09-22 Kewen Lin <linkw@linux.ibm.com>
11177 * config/rs6000/rs6000.opt (rs6000-density-pct-threshold,
11178 rs6000-density-size-threshold, rs6000-density-penalty,
11179 rs6000-density-load-pct-threshold,
11180 rs6000-density-load-num-threshold): New parameter.
11181 * config/rs6000/rs6000.c (rs6000_density_test): Adjust with
11182 corresponding parameters.
11184 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
11186 * gimple-range-path.cc (path_range_query::defined_outside_path):
11188 (path_range_query::range_on_path_entry): New.
11189 (path_range_query::internal_range_of_expr): Resolve unknowns
11191 (path_range_query::improve_range_with_equivs): New.
11192 (path_range_query::ssa_range_in_phi): Resolve unknowns with
11194 * gimple-range-path.h (class path_range_query): Add
11195 defined_outside_path, range_on_path_entry, and
11196 improve_range_with_equivs.
11198 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
11200 * gimple-range-path.cc (path_range_query::add_to_imports): New.
11201 (path_range_query::add_copies_to_imports): New.
11202 (path_range_query::precompute_ranges): Call
11203 add_copies_to_imports.
11204 * gimple-range-path.h (class path_range_query): Add prototypes
11205 for add_copies_to_imports and add_to_imports.
11207 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
11209 * gimple-range-path.cc (path_range_query::range_defined_in_block):
11210 Remove useless code.
11212 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
11214 * gimple-range-fold.h (class fur_source): Make oracle protected.
11215 * gimple-range-path.cc (path_range_query::path_range_query): Add
11216 resolve argument. Initialize oracle.
11217 (path_range_query::~path_range_query): Delete oracle.
11218 (path_range_query::range_of_stmt): Adapt to use relations.
11219 (path_range_query::precompute_ranges): Pre-compute relations.
11220 (class jt_fur_source): New
11221 (jt_fur_source::jt_fur_source): New.
11222 (jt_fur_source::register_relation): New.
11223 (jt_fur_source::query_relation): New.
11224 (path_range_query::precompute_relations): New.
11225 (path_range_query::precompute_phi_relations): New.
11226 * gimple-range-path.h (path_range_query): Add resolve argument.
11227 Add oracle, precompute_relations, precompute_phi_relations.
11228 * tree-ssa-threadbackward.c (back_threader::back_threader): Pass
11229 resolve argument to solver.
11231 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
11233 * gimple-range-fold.cc (fold_using_range::range_of_range_op):
11234 Rename postfold_gcond_edges to register_outgoing_edges and
11236 (fold_using_range::postfold_gcond_edges): Rename...
11237 (fur_source::register_outgoing_edges): ...to this.
11238 * gimple-range-fold.h (postfold_gcond_edges): Rename to
11239 register_outgoing_edges and move to fur_source.
11241 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
11243 * gimple-range-fold.cc (fold_using_range::range_of_phi): Check
11244 dom_info_available_p.
11246 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
11248 * gimple-range-cache.cc (non_null_ref::non_null_ref): Use create
11249 and quick_grow_cleared instead of safe_grow_cleared.
11251 2021-09-21 Thomas Schwinge <thomas@codesourcery.com>
11254 * omp-oacc-neuter-broadcast.cc (oacc_do_neutering): Evaluate
11255 'random ()' to '0'.
11257 2021-09-21 Richard Earnshaw <rearnsha@arm.com>
11259 * configure.ac: Detect when the assembler supports new-style
11260 architecture extensions.
11261 * common/config/arm/arm-common.c (arm_rewrite_mcpu): Return
11262 the full CPU string if the assembler can grok it.
11263 (arm_rewrite_march): Likewise but for the architecture.
11264 * config.in: Regenerate.
11265 * configure: Regenerate.
11267 2021-09-21 Richard Biener <rguenther@suse.de>
11269 PR tree-optimization/102421
11270 * tree-vect-loop.c (vect_dissolve_slp_only_groups): Copy and
11271 adjust alignment info.
11273 2021-09-21 Kewen Lin <linkw@linux.ibm.com>
11275 * ipa-fnsummary.c (ipa_fn_summary_write): Remove inconsistent
11276 bitfield stream out.
11278 2021-09-20 Andrew MacLeod <amacleod@redhat.com>
11280 * gimple-range-fold.cc (fold_using_range::range_of_phi): Ignore
11281 undefined edges, apply an equivalence if appropriate.
11282 * gimple-range-gori.cc (gori_compute::outgoing_edge_range_p): Return
11283 UNDEFINED if EDGE_EXECUTABLE is not set.
11284 * gimple-range.cc (gimple_ranger::gimple_ranger): Set all edges
11285 as EXECUTABLE upon startup.
11286 (gimple_ranger::range_on_edge): Return UNDEFINED for edges without
11287 EDGE_EXECUTABLE set.
11288 * vr-values.c (set_and_propagate_unexecutable): New.
11289 (simplify_using_ranges::fold_cond): Call set_and_propagate.
11290 (simplify_using_ranges::simplify_switch_using_ranges): Ditto.
11291 * vr-values.h: Add prototype.
11293 2021-09-20 Andrew MacLeod <amacleod@redhat.com>
11295 * value-relation.cc (equiv_oracle::register_initial_def): New.
11296 (equiv_oracle::register_relation): Call register_initial_def.
11297 (equiv_oracle::add_equiv_to_block): New. Split register_relation.
11298 (relation_oracle::register_stmt): Check def block of PHI arguments.
11299 * value-relation.h (equiv_oracle): Add new prototypes.
11301 2021-09-20 Matthias Kretz <m.kretz@gsi.de>
11303 * cppbuiltin.c (define_builtin_macros_for_compilation_flags):
11304 Define __RECIPROCAL_MATH__, __NO_SIGNED_ZEROS__,
11305 __NO_TRAPPING_MATH__, __ASSOCIATIVE_MATH__, and
11306 __ROUNDING_MATH__ according to their corresponding flags.
11307 * doc/cpp.texi: Document __RECIPROCAL_MATH__,
11308 __NO_SIGNED_ZEROS__, __NO_TRAPPING_MATH__, __ASSOCIATIVE_MATH__,
11309 and __ROUNDING_MATH__.
11311 2021-09-20 Richard Biener <rguenther@suse.de>
11313 * tree-vect-stmts.c (vectorizable_load): Use the vectype
11316 2021-09-20 Richard Biener <rguenther@suse.de>
11318 * tree-vect-data-refs.c (vect_duplicate_ssa_name_ptr_info):
11319 Do not compute alignment of the vectorized access here.
11321 2021-09-20 Richard Biener <rguenther@suse.de>
11323 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment):
11324 Store -1 for runtime alias peeling iterations.
11326 2021-09-20 Richard Biener <rguenther@suse.de>
11328 * config.gcc: Obsolete hppa[12]*-*-hpux10* and hppa[12]*-*-hpux11*.
11330 2021-09-20 Thomas Schwinge <thomas@codesourcery.com>
11332 * input.c (string_concat_db::record_string_concatenation)
11333 (string_concat_db::get_string_concatenation): Skip for
11334 'RESERVED_LOCATION_P'.
11336 2021-09-20 Richard Biener <rguenther@suse.de>
11338 PR tree-optimization/65206
11339 * tree-data-ref.h (struct data_reference): Add alt_indices,
11341 * tree-data-ref.c (free_data_ref): Release alt_indices.
11342 (dr_analyze_indices): Work on struct indices and get DR_REF as tree.
11343 (create_data_ref): Adjust.
11344 (initialize_data_dependence_relation): Split into head
11345 and tail. When the base objects fail to match up try
11346 again with pointer-based analysis of indices.
11347 * tree-vectorizer.c (vec_info_shared::check_datarefs): Do
11348 not compare the lazily computed alternate set of indices.
11350 2021-09-20 Iain Sandoe <iain@sandoe.co.uk>
11352 * gcc.c: Test for execute OK when we find the
11353 programs for assembler linker and dsymutil and those
11354 were specified at configure-time.
11356 2021-09-19 Martin Sebor <msebor@redhat.com>
11358 PR middle-end/102403
11359 * gimple-predicate-analysis.cc (predicate::init_from_control_deps):
11360 Correct a function pre/postcondition.
11362 2021-09-19 Martin Sebor <msebor@redhat.com>
11364 PR middle-end/102243
11365 * tree-ssa-strlen.c (get_range): Handle null cfun.
11367 2021-09-19 Iain Sandoe <iain@sandoe.co.uk>
11369 * config/darwin.h (LINK_COMMAND_SPEC_A): Use Darwin10
11370 unwinder shim as a convenience library.
11372 2021-09-19 Andrew Pinski <apinski@marvell.com>
11374 * doc/install.texi: Add note about
11375 binutils 2.35 is required for LTO usage.
11377 2021-09-19 Aldy Hernandez <aldyh@redhat.com>
11379 * tree-ssa-threadbackward.c
11380 (back_threader_registry::register_path): Use push_edge.
11381 * tree-ssa-threadedge.c
11382 (jump_threader::thread_around_empty_blocks): Same.
11383 (jump_threader::thread_through_normal_block): Same.
11384 (jump_threader::thread_across_edge): Same. Also, use auto_bitmap.
11386 * tree-ssa-threadupdate.c
11387 (jt_path_registry::allocate_thread_edge): Remove.
11388 (jt_path_registry::push_edge): New.
11389 (dump_jump_thread_path): Make static.
11390 * tree-ssa-threadupdate.h (allocate_thread_edge): Remove.
11393 2021-09-19 Aldy Hernandez <aldyh@redhat.com>
11395 * gimple-range-path.cc (path_range_query::path_range_query): Add
11397 (path_range_query::dump): Remove extern declaration of dump_ranger.
11398 * gimple-range-trace.cc (dump_ranger): Add DEBUG_FUNCTION marker.
11399 * gimple-range-trace.h (dump_ranger): Add prototype.
11401 2021-09-19 John Ericson <git@JohnEricson.me>
11403 * gcc.c (find_a_program): New function, factored out of...
11404 (find_a_file): Here.
11405 (execute): Use find_a_program when looking for programs rather
11408 2021-09-19 Matwey V. Kornilov <matwey.kornilov@gmail.com>
11410 * config/avr/avr-mcus.def: Add atmega324pb.
11411 * doc/avr-mmcu.texi: Corresponding changes.
11413 2021-09-19 Roger Sayle <roger@nextmovesoftware.com>
11415 PR middle-end/88173
11416 * match.pd (cmp @0 REAL_CST@1): When @0 is also REAL_CST, apply
11417 the same transformations as to @1. For comparisons against NaN,
11418 don't check HONOR_SNANS but confirm that neither operand is a
11421 2021-09-19 Benjamin Peterson <benjamin@locrian.net>
11423 * attribs.c (make_unique_name): Delete.
11424 * attribs.h (make_unique_name): Delete.
11426 2021-09-19 Andrew Pinski <apinski@marvell.com>
11428 * lra-constraints.c (check_and_process_move): Assert
11429 that dclass and sclass are greater than or equal to NO_REGS.
11431 2021-09-18 Jakub Jelinek <jakub@redhat.com>
11433 * tree.h (OMP_CLAUSE_ORDER_UNCONSTRAINED): Define.
11434 * tree-pretty-print.c (dump_omp_clause): Print unconstrained:
11435 for OMP_CLAUSE_ORDER_UNCONSTRAINED.
11437 2021-09-18 liuhongt <hongtao.liu@intel.com>
11439 * config/i386/i386-features.c (remove_partial_avx_dependency):
11440 Restrict TARGET_USE_VECTOR_FP_CONVERTS and
11441 TARGET_USE_VECTOR_CONVERTS to conversion instructions only.
11443 2021-09-18 Jakub Jelinek <jakub@redhat.com>
11445 * gimplify.c (omp_default_clause): For C/C++ default({,first}private),
11446 if file/namespace scope variable doesn't have predetermined sharing,
11447 treat it as if there was default(none).
11449 2021-09-18 liuhongt <hongtao.liu@intel.com>
11451 * config/i386/avx512fp16intrin.h (_mm_fmadd_sh):
11453 (_mm_mask_fmadd_sh): Likewise.
11454 (_mm_mask3_fmadd_sh): Likewise.
11455 (_mm_maskz_fmadd_sh): Likewise.
11456 (_mm_fmadd_round_sh): Likewise.
11457 (_mm_mask_fmadd_round_sh): Likewise.
11458 (_mm_mask3_fmadd_round_sh): Likewise.
11459 (_mm_maskz_fmadd_round_sh): Likewise.
11460 (_mm_fnmadd_sh): Likewise.
11461 (_mm_mask_fnmadd_sh): Likewise.
11462 (_mm_mask3_fnmadd_sh): Likewise.
11463 (_mm_maskz_fnmadd_sh): Likewise.
11464 (_mm_fnmadd_round_sh): Likewise.
11465 (_mm_mask_fnmadd_round_sh): Likewise.
11466 (_mm_mask3_fnmadd_round_sh): Likewise.
11467 (_mm_maskz_fnmadd_round_sh): Likewise.
11468 (_mm_fmsub_sh): Likewise.
11469 (_mm_mask_fmsub_sh): Likewise.
11470 (_mm_mask3_fmsub_sh): Likewise.
11471 (_mm_maskz_fmsub_sh): Likewise.
11472 (_mm_fmsub_round_sh): Likewise.
11473 (_mm_mask_fmsub_round_sh): Likewise.
11474 (_mm_mask3_fmsub_round_sh): Likewise.
11475 (_mm_maskz_fmsub_round_sh): Likewise.
11476 (_mm_fnmsub_sh): Likewise.
11477 (_mm_mask_fnmsub_sh): Likewise.
11478 (_mm_mask3_fnmsub_sh): Likewise.
11479 (_mm_maskz_fnmsub_sh): Likewise.
11480 (_mm_fnmsub_round_sh): Likewise.
11481 (_mm_mask_fnmsub_round_sh): Likewise.
11482 (_mm_mask3_fnmsub_round_sh): Likewise.
11483 (_mm_maskz_fnmsub_round_sh): Likewise.
11484 * config/i386/i386-builtin-types.def
11485 (V8HF_FTYPE_V8HF_V8HF_V8HF_UQI_INT): New builtin type.
11486 * config/i386/i386-builtin.def: Add new builtins.
11487 * config/i386/i386-expand.c: Handle new builtin type.
11488 * config/i386/sse.md (fmai_vmfmadd_<mode><round_name>):
11489 Ajdust to support FP16.
11490 (fmai_vmfmsub_<mode><round_name>): Ditto.
11491 (fmai_vmfnmadd_<mode><round_name>): Ditto.
11492 (fmai_vmfnmsub_<mode><round_name>): Ditto.
11493 (*fmai_fmadd_<mode>): Ditto.
11494 (*fmai_fmsub_<mode>): Ditto.
11495 (*fmai_fnmadd_<mode><round_name>): Ditto.
11496 (*fmai_fnmsub_<mode><round_name>): Ditto.
11497 (avx512f_vmfmadd_<mode>_mask<round_name>): Ditto.
11498 (avx512f_vmfmadd_<mode>_mask3<round_name>): Ditto.
11499 (avx512f_vmfmadd_<mode>_maskz<round_expand_name>): Ditto.
11500 (avx512f_vmfmadd_<mode>_maskz_1<round_name>): Ditto.
11501 (*avx512f_vmfmsub_<mode>_mask<round_name>): Ditto.
11502 (avx512f_vmfmsub_<mode>_mask3<round_name>): Ditto.
11503 (*avx512f_vmfmsub_<mode>_maskz_1<round_name>): Ditto.
11504 (*avx512f_vmfnmsub_<mode>_mask<round_name>): Ditto.
11505 (*avx512f_vmfnmsub_<mode>_mask3<round_name>): Ditto.
11506 (*avx512f_vmfnmsub_<mode>_mask<round_name>): Ditto.
11507 (*avx512f_vmfnmadd_<mode>_mask<round_name>): Renamed to ...
11508 (avx512f_vmfnmadd_<mode>_mask<round_name>) ... this, and
11509 adjust to support FP16.
11510 (avx512f_vmfnmadd_<mode>_mask3<round_name>): Ditto.
11511 (avx512f_vmfnmadd_<mode>_maskz_1<round_name>): Ditto.
11512 (avx512f_vmfnmadd_<mode>_maskz<round_expand_name>): New
11515 2021-09-18 H.J. Lu <hjl.tools@gmail.com>
11517 * config/i386/sse.md (avx512fmaskmodelower): Extend to support
11519 (maskload<mode><avx512fmaskmodelower>): Ditto.
11520 (maskstore<mode><avx512fmaskmodelower>): Ditto.
11522 2021-09-18 H.J. Lu <hjl.tools@gmail.com>
11524 * config/i386/i386-expand.c (ix86_expand_fp_absneg_operator):
11526 (ix86_expand_copysign): Ditto.
11527 (ix86_expand_xorsign): Ditto.
11528 * config/i386/i386.c (ix86_build_const_vector): Handle HF vector
11530 (ix86_build_signbit_mask): Ditto.
11531 (ix86_can_change_mode_class): Ditto.
11532 * config/i386/i386.md
11533 (SSEMODEF): Add HFmode.
11534 (ssevecmodef): Ditto.
11535 (<code>hf2): New define_expand.
11536 (*<code>hf2_1): New define_insn_and_split.
11537 (copysign<mode>): Extend to support HFmode under AVX512FP16.
11538 (xorsign<mode>): Ditto.
11539 * config/i386/sse.md (VFB): New mode iterator.
11540 (VFB_128_256): Ditto.
11542 (sseintvecmode2): Support HF vector mode.
11543 (<code><mode>2): Use new mode iterator.
11544 (*<code><mode>2): Ditto.
11545 (copysign<mode>3): Ditto.
11546 (xorsign<mode>3): Ditto.
11547 (<code><mode>3<mask_name>): Ditto.
11548 (<code><mode>3<mask_name>): Ditto.
11549 (<sse>_andnot<mode>3<mask_name>): Adjust for HF vector mode.
11550 (<sse>_andnot<mode>3<mask_name>): Ditto.
11551 (*<code><mode>3<mask_name>): Ditto.
11552 (*<code><mode>3<mask_name>): Ditto.
11554 2021-09-18 liuhongt <hongtao.liu@intel.com>
11556 * config/i386/avx512fp16intrin.h (_mm512_mask_fmadd_ph):
11558 (_mm512_mask3_fmadd_ph): Likewise.
11559 (_mm512_maskz_fmadd_ph): Likewise.
11560 (_mm512_fmadd_round_ph): Likewise.
11561 (_mm512_mask_fmadd_round_ph): Likewise.
11562 (_mm512_mask3_fmadd_round_ph): Likewise.
11563 (_mm512_maskz_fmadd_round_ph): Likewise.
11564 (_mm512_fnmadd_ph): Likewise.
11565 (_mm512_mask_fnmadd_ph): Likewise.
11566 (_mm512_mask3_fnmadd_ph): Likewise.
11567 (_mm512_maskz_fnmadd_ph): Likewise.
11568 (_mm512_fnmadd_round_ph): Likewise.
11569 (_mm512_mask_fnmadd_round_ph): Likewise.
11570 (_mm512_mask3_fnmadd_round_ph): Likewise.
11571 (_mm512_maskz_fnmadd_round_ph): Likewise.
11572 (_mm512_fmsub_ph): Likewise.
11573 (_mm512_mask_fmsub_ph): Likewise.
11574 (_mm512_mask3_fmsub_ph): Likewise.
11575 (_mm512_maskz_fmsub_ph): Likewise.
11576 (_mm512_fmsub_round_ph): Likewise.
11577 (_mm512_mask_fmsub_round_ph): Likewise.
11578 (_mm512_mask3_fmsub_round_ph): Likewise.
11579 (_mm512_maskz_fmsub_round_ph): Likewise.
11580 (_mm512_fnmsub_ph): Likewise.
11581 (_mm512_mask_fnmsub_ph): Likewise.
11582 (_mm512_mask3_fnmsub_ph): Likewise.
11583 (_mm512_maskz_fnmsub_ph): Likewise.
11584 (_mm512_fnmsub_round_ph): Likewise.
11585 (_mm512_mask_fnmsub_round_ph): Likewise.
11586 (_mm512_mask3_fnmsub_round_ph): Likewise.
11587 (_mm512_maskz_fnmsub_round_ph): Likewise.
11588 * config/i386/avx512fp16vlintrin.h (_mm256_fmadd_ph):
11590 (_mm256_mask_fmadd_ph): Likewise.
11591 (_mm256_mask3_fmadd_ph): Likewise.
11592 (_mm256_maskz_fmadd_ph): Likewise.
11593 (_mm_fmadd_ph): Likewise.
11594 (_mm_mask_fmadd_ph): Likewise.
11595 (_mm_mask3_fmadd_ph): Likewise.
11596 (_mm_maskz_fmadd_ph): Likewise.
11597 (_mm256_fnmadd_ph): Likewise.
11598 (_mm256_mask_fnmadd_ph): Likewise.
11599 (_mm256_mask3_fnmadd_ph): Likewise.
11600 (_mm256_maskz_fnmadd_ph): Likewise.
11601 (_mm_fnmadd_ph): Likewise.
11602 (_mm_mask_fnmadd_ph): Likewise.
11603 (_mm_mask3_fnmadd_ph): Likewise.
11604 (_mm_maskz_fnmadd_ph): Likewise.
11605 (_mm256_fmsub_ph): Likewise.
11606 (_mm256_mask_fmsub_ph): Likewise.
11607 (_mm256_mask3_fmsub_ph): Likewise.
11608 (_mm256_maskz_fmsub_ph): Likewise.
11609 (_mm_fmsub_ph): Likewise.
11610 (_mm_mask_fmsub_ph): Likewise.
11611 (_mm_mask3_fmsub_ph): Likewise.
11612 (_mm_maskz_fmsub_ph): Likewise.
11613 (_mm256_fnmsub_ph): Likewise.
11614 (_mm256_mask_fnmsub_ph): Likewise.
11615 (_mm256_mask3_fnmsub_ph): Likewise.
11616 (_mm256_maskz_fnmsub_ph): Likewise.
11617 (_mm_fnmsub_ph): Likewise.
11618 (_mm_mask_fnmsub_ph): Likewise.
11619 (_mm_mask3_fnmsub_ph): Likewise.
11620 (_mm_maskz_fnmsub_ph): Likewise.
11621 * config/i386/i386-builtin.def: Add corresponding new builtins.
11622 * config/i386/sse.md
11623 (<avx512>_fmadd_<mode>_maskz<round_expand_name>): Adjust to
11624 support HF vector modes.
11625 (<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name><round_name>):
11627 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1): Ditto.
11628 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_2): Ditto.
11629 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_3): Ditto.
11630 (<avx512>_fmadd_<mode>_mask<round_name>): Ditto.
11631 (<avx512>_fmadd_<mode>_mask3<round_name>): Ditto.
11632 (<avx512>_fmsub_<mode>_maskz<round_expand_name>): Ditto.
11633 (<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name><round_name>):
11635 (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1): Ditto.
11636 (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_2): Ditto.
11637 (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_3): Ditto.
11638 (<avx512>_fmsub_<mode>_mask<round_name>): Ditto.
11639 (<avx512>_fmsub_<mode>_mask3<round_name>): Ditto.
11640 (<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name><round_name>):
11642 (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1): Ditto.
11643 (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_2): Ditto.
11644 (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_3): Ditto.
11645 (<avx512>_fnmadd_<mode>_mask<round_name>): Ditto.
11646 (<avx512>_fnmadd_<mode>_mask3<round_name>): Ditto.
11647 (<avx512>_fnmsub_<mode>_maskz<round_expand_name>): Ditto.
11648 (<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name><round_name>):
11650 (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1): Ditto.
11651 (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_2): Ditto.
11652 (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_3): Ditto.
11653 (<avx512>_fnmsub_<mode>_mask<round_name>): Ditto.
11654 (<avx512>_fnmsub_<mode>_mask3<round_name>): Ditto.
11656 2021-09-18 liuhongt <hongtao.liu@intel.com>
11658 * config/i386/avx512fp16intrin.h (_mm512_fmaddsub_ph):
11660 (_mm512_mask_fmaddsub_ph): Likewise.
11661 (_mm512_mask3_fmaddsub_ph): Likewise.
11662 (_mm512_maskz_fmaddsub_ph): Likewise.
11663 (_mm512_fmaddsub_round_ph): Likewise.
11664 (_mm512_mask_fmaddsub_round_ph): Likewise.
11665 (_mm512_mask3_fmaddsub_round_ph): Likewise.
11666 (_mm512_maskz_fmaddsub_round_ph): Likewise.
11667 (_mm512_mask_fmsubadd_ph): Likewise.
11668 (_mm512_mask3_fmsubadd_ph): Likewise.
11669 (_mm512_maskz_fmsubadd_ph): Likewise.
11670 (_mm512_fmsubadd_round_ph): Likewise.
11671 (_mm512_mask_fmsubadd_round_ph): Likewise.
11672 (_mm512_mask3_fmsubadd_round_ph): Likewise.
11673 (_mm512_maskz_fmsubadd_round_ph): Likewise.
11674 * config/i386/avx512fp16vlintrin.h (_mm256_fmaddsub_ph):
11676 (_mm256_mask_fmaddsub_ph): Likewise.
11677 (_mm256_mask3_fmaddsub_ph): Likewise.
11678 (_mm256_maskz_fmaddsub_ph): Likewise.
11679 (_mm_fmaddsub_ph): Likewise.
11680 (_mm_mask_fmaddsub_ph): Likewise.
11681 (_mm_mask3_fmaddsub_ph): Likewise.
11682 (_mm_maskz_fmaddsub_ph): Likewise.
11683 (_mm256_fmsubadd_ph): Likewise.
11684 (_mm256_mask_fmsubadd_ph): Likewise.
11685 (_mm256_mask3_fmsubadd_ph): Likewise.
11686 (_mm256_maskz_fmsubadd_ph): Likewise.
11687 (_mm_fmsubadd_ph): Likewise.
11688 (_mm_mask_fmsubadd_ph): Likewise.
11689 (_mm_mask3_fmsubadd_ph): Likewise.
11690 (_mm_maskz_fmsubadd_ph): Likewise.
11691 * config/i386/i386-builtin.def: Add corresponding new builtins.
11692 * config/i386/sse.md (VFH_SF_AVX512VL): New mode iterator.
11693 * (<avx512>_fmsubadd_<mode>_maskz<round_expand_name>): New expander.
11694 * (<avx512>_fmaddsub_<mode>_maskz<round_expand_name>): Use
11696 * (<sd_mask_codefor>fma_fmaddsub_<mode><sd_maskz_name><round_name>):
11698 * (<avx512>_fmaddsub_<mode>_mask<round_name>): Ditto.
11699 * (<avx512>_fmaddsub_<mode>_mask3<round_name>): Ditto.
11700 * (<sd_mask_codefor>fma_fmsubadd_<mode><sd_maskz_name><round_name>):
11702 * (<avx512>_fmsubadd_<mode>_mask<round_name>): Ditto.
11703 * (<avx512>_fmsubadd_<mode>_mask3<round_name>): Ditto.
11705 2021-09-18 liuhongt <hongtao.liu@intel.com>
11708 * config/i386/i386.c (ix86_print_operand): Handle
11709 V8HF/V16HF/V32HFmode.
11710 * config/i386/i386.h (VALID_BCST_MODE_P): Add HFmode.
11711 * config/i386/sse.md (avx512bcst): Remove.
11713 2021-09-17 Martin Sebor <msebor@redhat.com>
11715 * Makefile.in (OBJS): Add gimple-predicate-analysis.o.
11716 * tree-ssa-uninit.c (max_phi_args): Move to gimple-predicate-analysis.
11717 (MASK_SET_BIT, MASK_TEST_BIT, MASK_EMPTY): Same.
11718 (check_defs): Add comment.
11719 (can_skip_redundant_opnd): Update comment.
11720 (compute_uninit_opnds_pos): Adjust to namespace change.
11721 (find_pdom): Move to gimple-predicate-analysis.cc.
11723 (struct uninit_undef_val_t): New.
11724 (is_non_loop_exit_postdominating): Move to gimple-predicate-analysis.cc.
11725 (find_control_equiv_block): Same.
11726 (MAX_NUM_CHAINS, MAX_CHAIN_LEN, MAX_POSTDOM_CHECK): Same.
11727 (MAX_SWITCH_CASES): Same.
11728 (compute_control_dep_chain): Same.
11729 (find_uninit_use): Use predicate analyzer.
11730 (struct pred_info): Move to gimple-predicate-analysis.
11731 (convert_control_dep_chain_into_preds): Same.
11732 (find_predicates): Same.
11733 (collect_phi_def_edges): Same.
11734 (warn_uninitialized_phi): Use predicate analyzer.
11735 (find_def_preds): Move to gimple-predicate-analysis.
11736 (dump_pred_info): Same.
11737 (dump_pred_chain): Same.
11738 (dump_predicates): Same.
11739 (destroy_predicate_vecs): Remove.
11740 (execute_late_warn_uninitialized): New.
11741 (get_cmp_code): Move to gimple-predicate-analysis.
11742 (is_value_included_in): Same.
11743 (value_sat_pred_p): Same.
11744 (find_matching_predicate_in_rest_chains): Same.
11745 (is_use_properly_guarded): Same.
11746 (prune_uninit_phi_opnds): Same.
11747 (find_var_cmp_const): Same.
11748 (use_pred_not_overlap_with_undef_path_pred): Same.
11749 (pred_equal_p): Same.
11750 (is_neq_relop_p): Same.
11751 (is_neq_zero_form_p): Same.
11752 (pred_expr_equal_p): Same.
11753 (is_pred_expr_subset_of): Same.
11754 (is_pred_chain_subset_of): Same.
11755 (is_included_in): Same.
11756 (is_superset_of): Same.
11757 (pred_neg_p): Same.
11758 (simplify_pred): Same.
11759 (simplify_preds_2): Same.
11760 (simplify_preds_3): Same.
11761 (simplify_preds_4): Same.
11762 (simplify_preds): Same.
11764 (push_to_worklist): Same.
11765 (get_pred_info_from_cmp): Same.
11766 (is_degenerated_phi): Same.
11767 (normalize_one_pred_1): Same.
11768 (normalize_one_pred): Same.
11769 (normalize_one_pred_chain): Same.
11770 (normalize_preds): Same.
11771 (can_one_predicate_be_invalidated_p): Same.
11772 (can_chain_union_be_invalidated_p): Same.
11773 (uninit_uses_cannot_happen): Same.
11774 (pass_late_warn_uninitialized::execute): Define.
11775 * gimple-predicate-analysis.cc: New file.
11776 * gimple-predicate-analysis.h: New file.
11778 2021-09-17 Julian Brown <julian@codesourcery.com>
11780 * config/gcn/gcn.c (gimple.h): Include.
11781 (gcn_fork_join): Emit barrier for worker-level joins.
11782 * omp-oacc-neuter-broadcast.cc (find_local_vars_to_propagate): Add
11783 writes_gang_private bitmap parameter. Set bit for blocks
11784 containing gang-private variable writes.
11785 (worker_single_simple): Don't emit barrier after predicated block.
11786 (worker_single_copy): Don't emit barrier if we're not broadcasting
11787 anything and the block contains no gang-private writes.
11788 (neuter_worker_single): Don't predicate blocks that only contain
11789 NOPs or internal marker functions. Pass has_gang_private_write
11790 argument to worker_single_copy.
11791 (oacc_do_neutering): Add writes_gang_private bitmap handling.
11793 2021-09-17 Julian Brown <julian@codesourcery.com>
11795 * config/gcn/gcn-protos.h
11796 (gcn_goacc_create_worker_broadcast_record): Update prototype.
11797 * config/gcn/gcn-tree.c (gcn_goacc_get_worker_red_decl): Use
11798 preallocated block of LDS memory. Do not cache/share decls for
11799 reduction temporaries between invocations.
11800 (gcn_goacc_reduction_teardown): Unshare VAR on second use.
11801 (gcn_goacc_create_worker_broadcast_record): Add OFFSET parameter
11802 and return temporary LDS space at that offset. Return pointer in
11804 * config/gcn/gcn.c (acc_lds_size, gang_private_hwm, lds_allocs):
11806 (ACC_LDS_SIZE): Define as acc_lds_size.
11807 (gcn_init_machine_status): Don't initialise lds_allocated,
11808 lds_allocs, reduc_decls fields of machine function struct.
11809 (gcn_option_override): Handle default size for gang-private
11810 variables and -mgang-private-size option.
11811 (gcn_expand_prologue): Use LDS_SIZE instead of LDS_SIZE-1 when
11812 initialising M0_REG.
11813 (gcn_shared_mem_layout): New function.
11814 (gcn_print_lds_decl): Update comment. Use global lds_allocs map and
11815 gang_private_hwm variable.
11816 (TARGET_GOACC_SHARED_MEM_LAYOUT): Define target hook.
11817 * config/gcn/gcn.h (machine_function): Remove lds_allocated,
11818 lds_allocs, reduc_decls. Add reduction_base, reduction_limit.
11819 * config/gcn/gcn.opt (gang_private_size_opt): New global.
11820 (mgang-private-size=): New option.
11821 * doc/tm.texi.in (TARGET_GOACC_SHARED_MEM_LAYOUT): Place
11822 documentation hook.
11823 * doc/tm.texi: Regenerate.
11824 * omp-oacc-neuter-broadcast.cc (targhooks.h, diagnostic-core.h):
11826 (build_sender_ref): Handle sender_decl being pointer.
11827 (worker_single_copy): Add PLACEMENT and ISOLATE_BROADCASTS
11828 parameters. Pass placement argument to
11829 create_worker_broadcast_record hook invocations. Handle
11830 sender_decl being pointer and isolate_broadcasts inserting extra
11832 (blk_offset_map_t): Add typedef.
11833 (neuter_worker_single): Add BLK_OFFSET_MAP parameter. Pass
11834 preallocated range to worker_single_copy call.
11835 (dfs_broadcast_reachable_1): New function.
11836 (idx_decl_pair_t, used_range_vec_t): New typedefs.
11837 (sort_size_descending): New function.
11838 (addr_range): New class.
11839 (splay_tree_compare_addr_range, splay_tree_free_key)
11840 (first_fit_range, merge_ranges_1, merge_ranges): New functions.
11841 (execute_omp_oacc_neuter_broadcast): Rename to...
11842 (oacc_do_neutering): ... this. Add BOUNDS_LO, BOUNDS_HI
11843 parameters. Arrange layout of shared memory for broadcast
11845 (execute_omp_oacc_neuter_broadcast): New function.
11846 (pass_omp_oacc_neuter_broadcast::gate): Remove num_workers==1
11847 handling from here. Enable pass for all OpenACC routines in order
11848 to call shared memory-layout hook.
11849 * target.def (create_worker_broadcast_record): Add OFFSET
11851 (shared_mem_layout): New hook.
11853 2021-09-17 Julian Brown <julian@codesourcery.com>
11854 Thomas Schwinge <thomas@codesourcery.com>
11856 * omp-oacc-neuter-broadcast.cc
11857 (pass_omp_oacc_neuter_broadcast::gate): Disable if num_workers is
11859 (execute_omp_oacc_neuter_broadcast): Adjust.
11861 2021-09-17 Andrew MacLeod <amacleod@redhat.com>
11863 * value-relation.cc (class equiv_chain): Move to header file.
11864 (path_oracle::path_oracle): New.
11865 (path_oracle::~path_oracle): New.
11866 (path_oracle::register_relation): New.
11867 (path_oracle::query_relation): New.
11868 (path_oracle::reset_path): New.
11869 (path_oracle::dump): New.
11870 * value-relation.h (class equiv_chain): Move to here.
11871 (class path_oracle): New.
11873 2021-09-17 Andrew MacLeod <amacleod@redhat.com>
11875 * gimple-range-cache.cc (ranger_cache::ranger_cache): Create a DOM
11877 * gimple-range-fold.cc (fur_depend::register_relation): Use
11878 register_stmt/edge routines.
11879 * value-relation.cc (equiv_chain::find): Relocate from equiv_oracle.
11880 (equiv_oracle::equiv_oracle): Create self equivalence cache.
11881 (equiv_oracle::~equiv_oracle): Release same.
11882 (equiv_oracle::equiv_set): Return entry from self equiv cache if there
11883 are no equivalences.
11884 (equiv_oracle::find_equiv_block): Move list find to equiv_chain.
11885 (equiv_oracle::register_relation): Rename from register_equiv.
11886 (relation_chain_head::find_relation): Relocate from dom_oracle.
11887 (relation_oracle::register_stmt): New.
11888 (relation_oracle::register_edge): New.
11889 (dom_oracle::*): Rename from relation_oracle.
11890 (dom_oracle::register_relation): Adjust to call equiv_oracle.
11891 (dom_oracle::set_one_relation): Split from register_relation.
11892 (dom_oracle::register_transitives): Consolidate 2 methods.
11893 (dom_oracle::find_relation_block): Move core to relation_chain.
11894 (dom_oracle::query_relation): Rename from find_relation_dom and adjust.
11895 * value-relation.h (class relation_oracle): New pure virtual base.
11896 (class equiv_oracle): Inherit from relation_oracle and adjust.
11897 (class dom_oracle): Rename from old relation_oracle and adjust.
11899 2021-09-17 Martin Sebor <msebor@redhat.com>
11901 PR middle-end/102200
11902 * pointer-query.cc (access_ref::inform_access): Handle MIN/MAX_EXPR.
11903 (handle_min_max_size): Change argument. Store original SSA_NAME for
11904 operands to potentially distinct (sub)objects.
11905 (compute_objsize_r): Adjust call to the above.
11907 2021-09-17 Bill Schmidt <wschmidt@linux.ibm.com>
11909 * config/rs6000/rs6000.c (rs6000-builtins.h): New include.
11910 (rs6000_new_builtin_vectorized_function): New function.
11911 (rs6000_new_builtin_md_vectorized_function): Likewise.
11912 (rs6000_builtin_vectorized_function): Call
11913 rs6000_new_builtin_vectorized_function.
11914 (rs6000_builtin_md_vectorized_function): Call
11915 rs6000_new_builtin_md_vectorized_function.
11917 2021-09-17 Bill Schmidt <wschmidt@linux.ibm.com>
11919 * config/rs6000/rs6000-builtin-new.def (ASSEMBLE_ACC): Add mmaint flag.
11920 (ASSEMBLE_PAIR): Likewise.
11921 (BUILD_ACC): Likewise.
11922 (DISASSEMBLE_ACC): Likewise.
11923 (DISASSEMBLE_PAIR): Likewise.
11924 (PMXVBF16GER2): Likewise.
11925 (PMXVBF16GER2NN): Likewise.
11926 (PMXVBF16GER2NP): Likewise.
11927 (PMXVBF16GER2PN): Likewise.
11928 (PMXVBF16GER2PP): Likewise.
11929 (PMXVF16GER2): Likewise.
11930 (PMXVF16GER2NN): Likewise.
11931 (PMXVF16GER2NP): Likewise.
11932 (PMXVF16GER2PN): Likewise.
11933 (PMXVF16GER2PP): Likewise.
11934 (PMXVF32GER): Likewise.
11935 (PMXVF32GERNN): Likewise.
11936 (PMXVF32GERNP): Likewise.
11937 (PMXVF32GERPN): Likewise.
11938 (PMXVF32GERPP): Likewise.
11939 (PMXVF64GER): Likewise.
11940 (PMXVF64GERNN): Likewise.
11941 (PMXVF64GERNP): Likewise.
11942 (PMXVF64GERPN): Likewise.
11943 (PMXVF64GERPP): Likewise.
11944 (PMXVI16GER2): Likewise.
11945 (PMXVI16GER2PP): Likewise.
11946 (PMXVI16GER2S): Likewise.
11947 (PMXVI16GER2SPP): Likewise.
11948 (PMXVI4GER8): Likewise.
11949 (PMXVI4GER8PP): Likewise.
11950 (PMXVI8GER4): Likewise.
11951 (PMXVI8GER4PP): Likewise.
11952 (PMXVI8GER4SPP): Likewise.
11953 (XVBF16GER2): Likewise.
11954 (XVBF16GER2NN): Likewise.
11955 (XVBF16GER2NP): Likewise.
11956 (XVBF16GER2PN): Likewise.
11957 (XVBF16GER2PP): Likewise.
11958 (XVF16GER2): Likewise.
11959 (XVF16GER2NN): Likewise.
11960 (XVF16GER2NP): Likewise.
11961 (XVF16GER2PN): Likewise.
11962 (XVF16GER2PP): Likewise.
11963 (XVF32GER): Likewise.
11964 (XVF32GERNN): Likewise.
11965 (XVF32GERNP): Likewise.
11966 (XVF32GERPN): Likewise.
11967 (XVF32GERPP): Likewise.
11968 (XVF64GER): Likewise.
11969 (XVF64GERNN): Likewise.
11970 (XVF64GERNP): Likewise.
11971 (XVF64GERPN): Likewise.
11972 (XVF64GERPP): Likewise.
11973 (XVI16GER2): Likewise.
11974 (XVI16GER2PP): Likewise.
11975 (XVI16GER2S): Likewise.
11976 (XVI16GER2SPP): Likewise.
11977 (XVI4GER8): Likewise.
11978 (XVI4GER8PP): Likewise.
11979 (XVI8GER4): Likewise.
11980 (XVI8GER4PP): Likewise.
11981 (XVI8GER4SPP): Likewise.
11982 (XXMFACC): Likewise.
11983 (XXMTACC): Likewise.
11984 (XXSETACCZ): Likewise.
11985 (ASSEMBLE_PAIR_V): Likewise.
11986 (BUILD_PAIR): Likewise.
11987 (DISASSEMBLE_PAIR_V): Likewise.
11990 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_new_mma_builtin):
11991 Handle RS6000_BIF_LXVP and RS6000_BIF_STXVP.
11992 * config/rs6000/rs6000-gen-builtins.c (attrinfo): Add ismmaint.
11993 (parse_bif_attrs): Handle ismmaint.
11994 (write_decls): Add bif_mmaint_bit and bif_is_mmaint.
11995 (write_bif_static_init): Handle ismmaint.
11997 2021-09-17 Bill Schmidt <wschmidt@linux.ibm.com>
11999 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_new_builtin): New
12001 (rs6000_gimple_fold_builtin): Call rs6000_gimple_fold_new_builtin.
12002 (rs6000_new_builtin_valid_without_lhs): New function.
12003 (rs6000_gimple_fold_new_mma_builtin): Likewise.
12004 (rs6000_gimple_fold_new_builtin): Likewise.
12006 2021-09-17 Thomas Schwinge <thomas@codesourcery.com>
12008 * hash-table.h (hash_table<Descriptor, Lazy, Allocator>::expand):
12009 Destruct stale Value objects.
12010 * hash-map-tests.c (test_map_of_type_with_ctor_and_dtor_expand):
12013 2021-09-17 Roger Sayle <roger@nextmovesoftware.com>
12016 * match.pd (shift optimizations): Disable recent sign-changing
12017 optimization for shifts by zero, these will be folded later.
12019 2021-09-17 Bill Schmidt <wschmidt@linux.ibm.com>
12021 * config/rs6000/rs6000-builtin-new.def (__builtin_mffsl): Move from
12022 [power9] to [always].
12024 2021-09-17 Richard Biener <rguenther@suse.de>
12026 * tree-vect-stmts.c (vectorizable_load): Do not frob
12029 2021-09-17 H.J. Lu <hjl.tools@gmail.com>
12031 * config/i386/i386-features.c (remove_partial_avx_dependency):
12032 Also check TARGET_SSE_PARTIAL_REG_FP_CONVERTS_DEPENDENCY and
12033 and TARGET_SSE_PARTIAL_REG_CONVERTS_DEPENDENCY before generating
12035 * config/i386/i386.h (TARGET_SSE_PARTIAL_REG_FP_CONVERTS_DEPENDENCY):
12037 (TARGET_SSE_PARTIAL_REG_CONVERTS_DEPENDENCY): Likewise.
12038 * config/i386/i386.md (SSE FP to FP splitters): Replace
12039 TARGET_SSE_PARTIAL_REG_DEPENDENCY with
12040 TARGET_SSE_PARTIAL_REG_FP_CONVERTS_DEPENDENCY.
12041 (SSE INT to FP splitter): Replace TARGET_SSE_PARTIAL_REG_DEPENDENCY
12042 with TARGET_SSE_PARTIAL_REG_CONVERTS_DEPENDENCY.
12043 * config/i386/x86-tune.def
12044 (X86_TUNE_SSE_PARTIAL_REG_FP_CONVERTS_DEPENDENCY): New.
12045 (X86_TUNE_SSE_PARTIAL_REG_CONVERTS_DEPENDENCY): Likewise.
12047 2021-09-17 H.J. Lu <hjl.tools@gmail.com>
12050 * config/i386/i386-features.c (remove_partial_avx_dependency):
12051 Check TARGET_USE_VECTOR_FP_CONVERTS and TARGET_USE_VECTOR_CONVERTS
12052 before generating vxorps.
12054 2021-09-17 H.J. Lu <hjl.tools@gmail.com>
12056 * config/i386/i386-options.c (processor_cost_table): Use
12057 tremont_cost for Tremont.
12058 * config/i386/x86-tune-costs.h (tremont_memcpy): New.
12059 (tremont_memset): Likewise.
12060 (tremont_cost): Likewise.
12061 * config/i386/x86-tune.def (X86_TUNE_PREFER_KNOWN_REP_MOVSB_STOSB):
12062 Enable for Tremont.
12064 2021-09-17 H.J. Lu <hjl.tools@gmail.com>
12066 * common/config/i386/i386-common.c: Use Haswell scheduling model
12068 * config/i386/i386.c (ix86_sched_init_global): Prepare for Tremont
12070 * config/i386/x86-tune-sched.c (ix86_issue_rate): Change Tremont
12072 (ix86_adjust_cost): Handle Tremont.
12073 * config/i386/x86-tune.def (X86_TUNE_SSE_PARTIAL_REG_DEPENDENCY):
12074 Enable for Tremont.
12075 (X86_TUNE_USE_LEAVE): Likewise.
12076 (X86_TUNE_PUSH_MEMORY): Likewise.
12077 (X86_TUNE_MISALIGNED_MOVE_STRING_PRO_EPILOGUES): Likewise.
12078 (X86_TUNE_USE_CLTD): Likewise.
12079 (X86_TUNE_AVOID_FALSE_DEP_FOR_BMI): Likewise.
12080 (X86_TUNE_AVOID_MFENCE): Likewise.
12081 (X86_TUNE_SSE_TYPELESS_STORES): Likewise.
12082 (X86_TUNE_SSE_LOAD0_BY_PXOR): Likewise.
12083 (X86_TUNE_ACCUMULATE_OUTGOING_ARGS): Disable for Tremont.
12084 (X86_TUNE_FOUR_JUMP_LIMIT): Likewise.
12085 (X86_TUNE_OPT_AGU): Likewise.
12086 (X86_TUNE_AVOID_LEA_FOR_ADDR): Likewise.
12087 (X86_TUNE_AVOID_MEM_OPND_FOR_CMOVE): Likewise.
12088 (X86_TUNE_EXPAND_ABS): Likewise.
12089 (X86_TUNE_SPLIT_MEM_OPND_FOR_FP_CONVERTS): Likewise.
12090 (X86_TUNE_SLOW_PSHUFB): Likewise.
12092 2021-09-17 Eric Botcazou <ebotcazou@adacore.com>
12094 PR rtl-optimization/102306
12095 * combine.c (try_combine): Abort the combination if we are about to
12096 duplicate volatile references.
12098 2021-09-17 liuhongt <hongtao.liu@intel.com>
12100 * config/i386/avx512fp16intrin.h (_mm_undefined_ph):
12102 (_mm256_undefined_ph): Likewise.
12103 (_mm512_undefined_ph): Likewise.
12104 (_mm_cvtsh_h): Likewise.
12105 (_mm256_cvtsh_h): Likewise.
12106 (_mm512_cvtsh_h): Likewise.
12107 (_mm512_castph_ps): Likewise.
12108 (_mm512_castph_pd): Likewise.
12109 (_mm512_castph_si512): Likewise.
12110 (_mm512_castph512_ph128): Likewise.
12111 (_mm512_castph512_ph256): Likewise.
12112 (_mm512_castph128_ph512): Likewise.
12113 (_mm512_castph256_ph512): Likewise.
12114 (_mm512_zextph128_ph512): Likewise.
12115 (_mm512_zextph256_ph512): Likewise.
12116 (_mm512_castps_ph): Likewise.
12117 (_mm512_castpd_ph): Likewise.
12118 (_mm512_castsi512_ph): Likewise.
12119 * config/i386/avx512fp16vlintrin.h (_mm_castph_ps):
12121 (_mm256_castph_ps): Likewise.
12122 (_mm_castph_pd): Likewise.
12123 (_mm256_castph_pd): Likewise.
12124 (_mm_castph_si128): Likewise.
12125 (_mm256_castph_si256): Likewise.
12126 (_mm_castps_ph): Likewise.
12127 (_mm256_castps_ph): Likewise.
12128 (_mm_castpd_ph): Likewise.
12129 (_mm256_castpd_ph): Likewise.
12130 (_mm_castsi128_ph): Likewise.
12131 (_mm256_castsi256_ph): Likewise.
12132 (_mm256_castph256_ph128): Likewise.
12133 (_mm256_castph128_ph256): Likewise.
12134 (_mm256_zextph128_ph256): Likewise.
12136 2021-09-17 liuhongt <hongtao.liu@intel.com>
12138 * config/i386/avx512fp16intrin.h (_mm_cvtsh_ss):
12140 (_mm_mask_cvtsh_ss): Likewise.
12141 (_mm_maskz_cvtsh_ss): Likewise.
12142 (_mm_cvtsh_sd): Likewise.
12143 (_mm_mask_cvtsh_sd): Likewise.
12144 (_mm_maskz_cvtsh_sd): Likewise.
12145 (_mm_cvt_roundsh_ss): Likewise.
12146 (_mm_mask_cvt_roundsh_ss): Likewise.
12147 (_mm_maskz_cvt_roundsh_ss): Likewise.
12148 (_mm_cvt_roundsh_sd): Likewise.
12149 (_mm_mask_cvt_roundsh_sd): Likewise.
12150 (_mm_maskz_cvt_roundsh_sd): Likewise.
12151 (_mm_cvtss_sh): Likewise.
12152 (_mm_mask_cvtss_sh): Likewise.
12153 (_mm_maskz_cvtss_sh): Likewise.
12154 (_mm_cvtsd_sh): Likewise.
12155 (_mm_mask_cvtsd_sh): Likewise.
12156 (_mm_maskz_cvtsd_sh): Likewise.
12157 (_mm_cvt_roundss_sh): Likewise.
12158 (_mm_mask_cvt_roundss_sh): Likewise.
12159 (_mm_maskz_cvt_roundss_sh): Likewise.
12160 (_mm_cvt_roundsd_sh): Likewise.
12161 (_mm_mask_cvt_roundsd_sh): Likewise.
12162 (_mm_maskz_cvt_roundsd_sh): Likewise.
12163 * config/i386/i386-builtin-types.def
12164 (V8HF_FTYPE_V2DF_V8HF_V8HF_UQI_INT,
12165 V8HF_FTYPE_V4SF_V8HF_V8HF_UQI_INT,
12166 V2DF_FTYPE_V8HF_V2DF_V2DF_UQI_INT,
12167 V4SF_FTYPE_V8HF_V4SF_V4SF_UQI_INT): Add new builtin types.
12168 * config/i386/i386-builtin.def: Add corrresponding new builtins.
12169 * config/i386/i386-expand.c: Handle new builtin types.
12170 * config/i386/sse.md (VF48_128): New mode iterator.
12171 (avx512fp16_vcvtsh2<ssescalarmodesuffix><mask_scalar_name><round_saeonly_scalar_name>):
12173 (avx512fp16_vcvt<ssescalarmodesuffix>2sh<mask_scalar_name><round_scalar_name>):
12176 2021-09-17 liuhongt <hongtao.liu@intel.com>
12178 * config/i386/avx512fp16intrin.h (_mm512_cvtph_pd):
12180 (_mm512_mask_cvtph_pd): Likewise.
12181 (_mm512_maskz_cvtph_pd): Likewise.
12182 (_mm512_cvt_roundph_pd): Likewise.
12183 (_mm512_mask_cvt_roundph_pd): Likewise.
12184 (_mm512_maskz_cvt_roundph_pd): Likewise.
12185 (_mm512_cvtxph_ps): Likewise.
12186 (_mm512_mask_cvtxph_ps): Likewise.
12187 (_mm512_maskz_cvtxph_ps): Likewise.
12188 (_mm512_cvtx_roundph_ps): Likewise.
12189 (_mm512_mask_cvtx_roundph_ps): Likewise.
12190 (_mm512_maskz_cvtx_roundph_ps): Likewise.
12191 (_mm512_cvtxps_ph): Likewise.
12192 (_mm512_mask_cvtxps_ph): Likewise.
12193 (_mm512_maskz_cvtxps_ph): Likewise.
12194 (_mm512_cvtx_roundps_ph): Likewise.
12195 (_mm512_mask_cvtx_roundps_ph): Likewise.
12196 (_mm512_maskz_cvtx_roundps_ph): Likewise.
12197 (_mm512_cvtpd_ph): Likewise.
12198 (_mm512_mask_cvtpd_ph): Likewise.
12199 (_mm512_maskz_cvtpd_ph): Likewise.
12200 (_mm512_cvt_roundpd_ph): Likewise.
12201 (_mm512_mask_cvt_roundpd_ph): Likewise.
12202 (_mm512_maskz_cvt_roundpd_ph): Likewise.
12203 * config/i386/avx512fp16vlintrin.h (_mm_cvtph_pd):
12205 (_mm_mask_cvtph_pd): Likewise.
12206 (_mm_maskz_cvtph_pd): Likewise.
12207 (_mm256_cvtph_pd): Likewise.
12208 (_mm256_mask_cvtph_pd): Likewise.
12209 (_mm256_maskz_cvtph_pd): Likewise.
12210 (_mm_cvtxph_ps): Likewise.
12211 (_mm_mask_cvtxph_ps): Likewise.
12212 (_mm_maskz_cvtxph_ps): Likewise.
12213 (_mm256_cvtxph_ps): Likewise.
12214 (_mm256_mask_cvtxph_ps): Likewise.
12215 (_mm256_maskz_cvtxph_ps): Likewise.
12216 (_mm_cvtxps_ph): Likewise.
12217 (_mm_mask_cvtxps_ph): Likewise.
12218 (_mm_maskz_cvtxps_ph): Likewise.
12219 (_mm256_cvtxps_ph): Likewise.
12220 (_mm256_mask_cvtxps_ph): Likewise.
12221 (_mm256_maskz_cvtxps_ph): Likewise.
12222 (_mm_cvtpd_ph): Likewise.
12223 (_mm_mask_cvtpd_ph): Likewise.
12224 (_mm_maskz_cvtpd_ph): Likewise.
12225 (_mm256_cvtpd_ph): Likewise.
12226 (_mm256_mask_cvtpd_ph): Likewise.
12227 (_mm256_maskz_cvtpd_ph): Likewise.
12228 * config/i386/i386-builtin.def: Add corresponding new builtins.
12229 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
12230 * config/i386/i386-expand.c: Handle new builtin types.
12231 * config/i386/sse.md
12232 (VF4_128_8_256): New.
12233 (VF48H_AVX512VL): Ditto.
12234 (ssePHmode): Add HF vector modes.
12235 (castmode): Add new convertable modes.
12236 (qq2phsuff): Ditto.
12237 (ph2pssuffix): New.
12238 (avx512fp16_vcvt<castmode>2ph_<mode><mask_name><round_name>): Ditto.
12239 (avx512fp16_vcvt<castmode>2ph_<mode>): Ditto.
12240 (*avx512fp16_vcvt<castmode>2ph_<mode>): Ditto.
12241 (avx512fp16_vcvt<castmode>2ph_<mode>_mask): Ditto.
12242 (*avx512fp16_vcvt<castmode>2ph_<mode>_mask): Ditto.
12243 (*avx512fp16_vcvt<castmode>2ph_<mode>_mask_1): Ditto.
12244 (avx512fp16_float_extend_ph<mode>2<mask_name><round_saeonly_name>):
12246 (avx512fp16_float_extend_ph<mode>2<mask_name>): Ditto.
12247 (*avx512fp16_float_extend_ph<mode>2_load<mask_name>): Ditto.
12248 (avx512fp16_float_extend_phv2df2<mask_name>): Ditto.
12249 (*avx512fp16_float_extend_phv2df2_load<mask_name>): Ditto.
12251 2021-09-17 liuhongt <hongtao.liu@intel.com>
12253 * config/i386/avx512fp16intrin.h (_mm_cvttsh_i32):
12255 (_mm_cvttsh_u32): Likewise.
12256 (_mm_cvtt_roundsh_i32): Likewise.
12257 (_mm_cvtt_roundsh_u32): Likewise.
12258 (_mm_cvttsh_i64): Likewise.
12259 (_mm_cvttsh_u64): Likewise.
12260 (_mm_cvtt_roundsh_i64): Likewise.
12261 (_mm_cvtt_roundsh_u64): Likewise.
12262 * config/i386/i386-builtin.def: Add corresponding new builtins.
12263 * config/i386/sse.md
12264 (avx512fp16_fix<fixunssuffix>_trunc<mode>2<round_saeonly_name>):
12267 2021-09-17 liuhongt <hongtao.liu@intel.com>
12269 * config/i386/avx512fp16intrin.h (_mm512_cvttph_epi32):
12271 (_mm512_mask_cvttph_epi32): Likewise.
12272 (_mm512_maskz_cvttph_epi32): Likewise.
12273 (_mm512_cvtt_roundph_epi32): Likewise.
12274 (_mm512_mask_cvtt_roundph_epi32): Likewise.
12275 (_mm512_maskz_cvtt_roundph_epi32): Likewise.
12276 (_mm512_cvttph_epu32): Likewise.
12277 (_mm512_mask_cvttph_epu32): Likewise.
12278 (_mm512_maskz_cvttph_epu32): Likewise.
12279 (_mm512_cvtt_roundph_epu32): Likewise.
12280 (_mm512_mask_cvtt_roundph_epu32): Likewise.
12281 (_mm512_maskz_cvtt_roundph_epu32): Likewise.
12282 (_mm512_cvttph_epi64): Likewise.
12283 (_mm512_mask_cvttph_epi64): Likewise.
12284 (_mm512_maskz_cvttph_epi64): Likewise.
12285 (_mm512_cvtt_roundph_epi64): Likewise.
12286 (_mm512_mask_cvtt_roundph_epi64): Likewise.
12287 (_mm512_maskz_cvtt_roundph_epi64): Likewise.
12288 (_mm512_cvttph_epu64): Likewise.
12289 (_mm512_mask_cvttph_epu64): Likewise.
12290 (_mm512_maskz_cvttph_epu64): Likewise.
12291 (_mm512_cvtt_roundph_epu64): Likewise.
12292 (_mm512_mask_cvtt_roundph_epu64): Likewise.
12293 (_mm512_maskz_cvtt_roundph_epu64): Likewise.
12294 (_mm512_cvttph_epi16): Likewise.
12295 (_mm512_mask_cvttph_epi16): Likewise.
12296 (_mm512_maskz_cvttph_epi16): Likewise.
12297 (_mm512_cvtt_roundph_epi16): Likewise.
12298 (_mm512_mask_cvtt_roundph_epi16): Likewise.
12299 (_mm512_maskz_cvtt_roundph_epi16): Likewise.
12300 (_mm512_cvttph_epu16): Likewise.
12301 (_mm512_mask_cvttph_epu16): Likewise.
12302 (_mm512_maskz_cvttph_epu16): Likewise.
12303 (_mm512_cvtt_roundph_epu16): Likewise.
12304 (_mm512_mask_cvtt_roundph_epu16): Likewise.
12305 (_mm512_maskz_cvtt_roundph_epu16): Likewise.
12306 * config/i386/avx512fp16vlintrin.h (_mm_cvttph_epi32):
12308 (_mm_mask_cvttph_epi32): Likewise.
12309 (_mm_maskz_cvttph_epi32): Likewise.
12310 (_mm256_cvttph_epi32): Likewise.
12311 (_mm256_mask_cvttph_epi32): Likewise.
12312 (_mm256_maskz_cvttph_epi32): Likewise.
12313 (_mm_cvttph_epu32): Likewise.
12314 (_mm_mask_cvttph_epu32): Likewise.
12315 (_mm_maskz_cvttph_epu32): Likewise.
12316 (_mm256_cvttph_epu32): Likewise.
12317 (_mm256_mask_cvttph_epu32): Likewise.
12318 (_mm256_maskz_cvttph_epu32): Likewise.
12319 (_mm_cvttph_epi64): Likewise.
12320 (_mm_mask_cvttph_epi64): Likewise.
12321 (_mm_maskz_cvttph_epi64): Likewise.
12322 (_mm256_cvttph_epi64): Likewise.
12323 (_mm256_mask_cvttph_epi64): Likewise.
12324 (_mm256_maskz_cvttph_epi64): Likewise.
12325 (_mm_cvttph_epu64): Likewise.
12326 (_mm_mask_cvttph_epu64): Likewise.
12327 (_mm_maskz_cvttph_epu64): Likewise.
12328 (_mm256_cvttph_epu64): Likewise.
12329 (_mm256_mask_cvttph_epu64): Likewise.
12330 (_mm256_maskz_cvttph_epu64): Likewise.
12331 (_mm_cvttph_epi16): Likewise.
12332 (_mm_mask_cvttph_epi16): Likewise.
12333 (_mm_maskz_cvttph_epi16): Likewise.
12334 (_mm256_cvttph_epi16): Likewise.
12335 (_mm256_mask_cvttph_epi16): Likewise.
12336 (_mm256_maskz_cvttph_epi16): Likewise.
12337 (_mm_cvttph_epu16): Likewise.
12338 (_mm_mask_cvttph_epu16): Likewise.
12339 (_mm_maskz_cvttph_epu16): Likewise.
12340 (_mm256_cvttph_epu16): Likewise.
12341 (_mm256_mask_cvttph_epu16): Likewise.
12342 (_mm256_maskz_cvttph_epu16): Likewise.
12343 * config/i386/i386-builtin.def: Add new builtins.
12344 * config/i386/sse.md
12345 (avx512fp16_fix<fixunssuffix>_trunc<mode>2<mask_name><round_saeonly_name>):
12347 (avx512fp16_fix<fixunssuffix>_trunc<mode>2<mask_name>): Ditto.
12348 (*avx512fp16_fix<fixunssuffix>_trunc<mode>2_load<mask_name>): Ditto.
12349 (avx512fp16_fix<fixunssuffix>_truncv2di2<mask_name>): Ditto.
12350 (avx512fp16_fix<fixunssuffix>_truncv2di2_load<mask_name>): Ditto.
12352 2021-09-17 liuhongt <hongtao.liu@intel.com>
12354 * config/i386/avx512fp16intrin.h (_mm_cvtsh_i32): New intrinsic.
12355 (_mm_cvtsh_u32): Likewise.
12356 (_mm_cvt_roundsh_i32): Likewise.
12357 (_mm_cvt_roundsh_u32): Likewise.
12358 (_mm_cvtsh_i64): Likewise.
12359 (_mm_cvtsh_u64): Likewise.
12360 (_mm_cvt_roundsh_i64): Likewise.
12361 (_mm_cvt_roundsh_u64): Likewise.
12362 (_mm_cvti32_sh): Likewise.
12363 (_mm_cvtu32_sh): Likewise.
12364 (_mm_cvt_roundi32_sh): Likewise.
12365 (_mm_cvt_roundu32_sh): Likewise.
12366 (_mm_cvti64_sh): Likewise.
12367 (_mm_cvtu64_sh): Likewise.
12368 (_mm_cvt_roundi64_sh): Likewise.
12369 (_mm_cvt_roundu64_sh): Likewise.
12370 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
12371 * config/i386/i386-builtin.def: Add corresponding new builtins.
12372 * config/i386/i386-expand.c (ix86_expand_round_builtin):
12373 Handle new builtin types.
12374 * config/i386/sse.md
12375 (avx512fp16_vcvtsh2<sseintconvertsignprefix>si<rex64namesuffix><round_name>):
12377 (avx512fp16_vcvtsh2<sseintconvertsignprefix>si<rex64namesuffix>_2): Likewise.
12378 (avx512fp16_vcvt<floatsuffix>si2sh<rex64namesuffix><round_name>): Likewise.
12380 2021-09-16 Bill Schmidt <wschmidt@linux.ibm.com>
12382 * config/rs6000/rs6000-c.c (rs6000-builtins.h): New include.
12383 (altivec_resolve_new_overloaded_builtin): New forward decl.
12384 (rs6000_new_builtin_type_compatible): New function.
12385 (altivec_resolve_overloaded_builtin): Call
12386 altivec_resolve_new_overloaded_builtin.
12387 (altivec_build_new_resolved_builtin): New function.
12388 (altivec_resolve_new_overloaded_builtin): Likewise.
12389 * config/rs6000/rs6000-call.c (rs6000_new_builtin_is_supported):
12391 * config/rs6000/rs6000-gen-builtins.c (write_decls): Remove _p from
12392 name of rs6000_new_builtin_is_supported.
12394 2021-09-16 Uroš Bizjak <ubizjak@gmail.com>
12396 * config/i386/i386-protos.h (ix86_decompose_address):
12397 Change return type to bool.
12398 * config/i386/i386.c (ix86_decompose_address): Ditto.
12400 2021-09-16 Tobias Burnus <tobias@codesourcery.com>
12403 * config/rs6000/t-rs6000 (build/rs6000-gen-builtins.o, build/rbtree.o):
12404 Added 'build/' to target, use build/%.o rule.
12405 (build/rs6000-gen-builtins$(build_exeext)): Add 'build/' and
12406 '$(build_exeext)' to target and 'build/' for the *.o files.
12407 (rs6000-builtins.c): Update for those changes; run rs6000-gen-builtins
12410 2021-09-16 Martin Jambor <mjambor@suse.cz>
12412 * cgraph.c (cgraph_node::dump): Do not check caller count sums if
12413 the body has been removed. Remove trailing whitespace.
12415 2021-09-16 Richard Biener <rguenther@suse.de>
12417 PR middle-end/102360
12418 * internal-fn.c (expand_DEFERRED_INIT): Make pattern-init
12419 of non-memory more robust.
12421 2021-09-16 Daniel Cederman <cederman@gaisler.com>
12423 * config/sparc/sparc-opts.h (enum sparc_processor_type): Add LEON5
12424 * config/sparc/sparc.c (struct processor_costs): Add LEON5 costs
12425 (leon5_adjust_cost): Increase cost of store with data dependency
12426 on ALU instruction and FPU anti-dependencies.
12427 (sparc_option_override): Add LEON5 costs
12428 (sparc_adjust_cost): Add LEON5 cost adjustments
12429 * config/sparc/sparc.h: Add LEON5
12430 * config/sparc/sparc.md: Include LEON5 scheduling information
12431 * config/sparc/sparc.opt: Add LEON5
12432 * doc/invoke.texi: Add LEON5
12433 * config/sparc/leon5.md: New file.
12435 2021-09-16 Daniel Cederman <cederman@gaisler.com>
12437 * config/sparc/sparc.md (stack_protect_set32): Add NOP to prevent
12438 sensitive sequence for B2BST errata workaround.
12440 2021-09-16 Daniel Cederman <cederman@gaisler.com>
12442 * config/sparc/sparc.c (sparc_do_work_around_errata): Do not begin
12443 functions with atomic instruction in the UT700 errata workaround.
12445 2021-09-16 Daniel Cederman <cederman@gaisler.com>
12447 * config/sparc/sparc.c (next_active_non_empty_insn): New function
12448 that returns next active non empty assembly instruction.
12449 (sparc_do_work_around_errata): Use new function.
12451 2021-09-16 Daniel Cederman <cederman@gaisler.com>
12453 * config/sparc/sparc.c (store_insn_p): Add predicate for store
12455 (load_insn_p): Add predicate for load attributes.
12456 (sparc_do_work_around_errata): Use new predicates.
12458 2021-09-16 Andreas Larsson <andreas@gaisler.com>
12460 * config/sparc/sparc.c (dump_target_flag_bits): Print bit names for
12463 2021-09-16 Martin Liska <mliska@suse.cz>
12465 * config/mips/netbsd.h: Fix typo in name of a macro.
12467 2021-09-16 liuhongt <hongtao.liu@intel.com>
12469 PR middle-end/102080
12470 * match.pd: Check mask type when doing cond_op related gimple
12472 * tree.c (is_truth_type_for): New function.
12473 * tree.h (is_truth_type_for): New declaration.
12475 2021-09-16 liuhongt <hongtao.liu@intel.com>
12477 * config/i386/avx512fp16intrin.h (_mm512_cvtepi32_ph): New
12479 (_mm512_mask_cvtepi32_ph): Likewise.
12480 (_mm512_maskz_cvtepi32_ph): Likewise.
12481 (_mm512_cvt_roundepi32_ph): Likewise.
12482 (_mm512_mask_cvt_roundepi32_ph): Likewise.
12483 (_mm512_maskz_cvt_roundepi32_ph): Likewise.
12484 (_mm512_cvtepu32_ph): Likewise.
12485 (_mm512_mask_cvtepu32_ph): Likewise.
12486 (_mm512_maskz_cvtepu32_ph): Likewise.
12487 (_mm512_cvt_roundepu32_ph): Likewise.
12488 (_mm512_mask_cvt_roundepu32_ph): Likewise.
12489 (_mm512_maskz_cvt_roundepu32_ph): Likewise.
12490 (_mm512_cvtepi64_ph): Likewise.
12491 (_mm512_mask_cvtepi64_ph): Likewise.
12492 (_mm512_maskz_cvtepi64_ph): Likewise.
12493 (_mm512_cvt_roundepi64_ph): Likewise.
12494 (_mm512_mask_cvt_roundepi64_ph): Likewise.
12495 (_mm512_maskz_cvt_roundepi64_ph): Likewise.
12496 (_mm512_cvtepu64_ph): Likewise.
12497 (_mm512_mask_cvtepu64_ph): Likewise.
12498 (_mm512_maskz_cvtepu64_ph): Likewise.
12499 (_mm512_cvt_roundepu64_ph): Likewise.
12500 (_mm512_mask_cvt_roundepu64_ph): Likewise.
12501 (_mm512_maskz_cvt_roundepu64_ph): Likewise.
12502 (_mm512_cvtepi16_ph): Likewise.
12503 (_mm512_mask_cvtepi16_ph): Likewise.
12504 (_mm512_maskz_cvtepi16_ph): Likewise.
12505 (_mm512_cvt_roundepi16_ph): Likewise.
12506 (_mm512_mask_cvt_roundepi16_ph): Likewise.
12507 (_mm512_maskz_cvt_roundepi16_ph): Likewise.
12508 (_mm512_cvtepu16_ph): Likewise.
12509 (_mm512_mask_cvtepu16_ph): Likewise.
12510 (_mm512_maskz_cvtepu16_ph): Likewise.
12511 (_mm512_cvt_roundepu16_ph): Likewise.
12512 (_mm512_mask_cvt_roundepu16_ph): Likewise.
12513 (_mm512_maskz_cvt_roundepu16_ph): Likewise.
12514 * config/i386/avx512fp16vlintrin.h (_mm_cvtepi32_ph): New
12516 (_mm_mask_cvtepi32_ph): Likewise.
12517 (_mm_maskz_cvtepi32_ph): Likewise.
12518 (_mm256_cvtepi32_ph): Likewise.
12519 (_mm256_mask_cvtepi32_ph): Likewise.
12520 (_mm256_maskz_cvtepi32_ph): Likewise.
12521 (_mm_cvtepu32_ph): Likewise.
12522 (_mm_mask_cvtepu32_ph): Likewise.
12523 (_mm_maskz_cvtepu32_ph): Likewise.
12524 (_mm256_cvtepu32_ph): Likewise.
12525 (_mm256_mask_cvtepu32_ph): Likewise.
12526 (_mm256_maskz_cvtepu32_ph): Likewise.
12527 (_mm_cvtepi64_ph): Likewise.
12528 (_mm_mask_cvtepi64_ph): Likewise.
12529 (_mm_maskz_cvtepi64_ph): Likewise.
12530 (_mm256_cvtepi64_ph): Likewise.
12531 (_mm256_mask_cvtepi64_ph): Likewise.
12532 (_mm256_maskz_cvtepi64_ph): Likewise.
12533 (_mm_cvtepu64_ph): Likewise.
12534 (_mm_mask_cvtepu64_ph): Likewise.
12535 (_mm_maskz_cvtepu64_ph): Likewise.
12536 (_mm256_cvtepu64_ph): Likewise.
12537 (_mm256_mask_cvtepu64_ph): Likewise.
12538 (_mm256_maskz_cvtepu64_ph): Likewise.
12539 (_mm_cvtepi16_ph): Likewise.
12540 (_mm_mask_cvtepi16_ph): Likewise.
12541 (_mm_maskz_cvtepi16_ph): Likewise.
12542 (_mm256_cvtepi16_ph): Likewise.
12543 (_mm256_mask_cvtepi16_ph): Likewise.
12544 (_mm256_maskz_cvtepi16_ph): Likewise.
12545 (_mm_cvtepu16_ph): Likewise.
12546 (_mm_mask_cvtepu16_ph): Likewise.
12547 (_mm_maskz_cvtepu16_ph): Likewise.
12548 (_mm256_cvtepu16_ph): Likewise.
12549 (_mm256_mask_cvtepu16_ph): Likewise.
12550 (_mm256_maskz_cvtepu16_ph): Likewise.
12551 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
12552 * config/i386/i386-builtin.def: Add corresponding new builtins.
12553 * config/i386/i386-expand.c
12554 (ix86_expand_args_builtin): Handle new builtin types.
12555 (ix86_expand_round_builtin): Ditto.
12556 * config/i386/i386-modes.def: Declare V2HF and V6HF.
12557 * config/i386/sse.md (VI2H_AVX512VL): New.
12558 (qq2phsuff): Ditto.
12559 (sseintvecmode): Add HF vector modes.
12560 (avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode><mask_name><round_name>):
12562 (avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>): Ditto.
12563 (*avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>): Ditto.
12564 (avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>_mask): Ditto.
12565 (*avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>_mask): Ditto.
12566 (*avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>_mask_1): Ditto.
12567 (avx512fp16_vcvt<floatsuffix>qq2ph_v2di): Ditto.
12568 (*avx512fp16_vcvt<floatsuffix>qq2ph_v2di): Ditto.
12569 (avx512fp16_vcvt<floatsuffix>qq2ph_v2di_mask): Ditto.
12570 (*avx512fp16_vcvt<floatsuffix>qq2ph_v2di_mask): Ditto.
12571 (*avx512fp16_vcvt<floatsuffix>qq2ph_v2di_mask_1): Ditto.
12572 * config/i386/subst.md (round_qq2phsuff): New subst_attr.
12574 2021-09-16 liuhongt <hongtao.liu@intel.com>
12576 * config/i386/avx512fp16intrin.h (_mm512_cvtph_epi32):
12578 (_mm512_mask_cvtph_epi32): Likewise.
12579 (_mm512_maskz_cvtph_epi32): Likewise.
12580 (_mm512_cvt_roundph_epi32): Likewise.
12581 (_mm512_mask_cvt_roundph_epi32): Likewise.
12582 (_mm512_maskz_cvt_roundph_epi32): Likewise.
12583 (_mm512_cvtph_epu32): Likewise.
12584 (_mm512_mask_cvtph_epu32): Likewise.
12585 (_mm512_maskz_cvtph_epu32): Likewise.
12586 (_mm512_cvt_roundph_epu32): Likewise.
12587 (_mm512_mask_cvt_roundph_epu32): Likewise.
12588 (_mm512_maskz_cvt_roundph_epu32): Likewise.
12589 (_mm512_cvtph_epi64): Likewise.
12590 (_mm512_mask_cvtph_epi64): Likewise.
12591 (_mm512_maskz_cvtph_epi64): Likewise.
12592 (_mm512_cvt_roundph_epi64): Likewise.
12593 (_mm512_mask_cvt_roundph_epi64): Likewise.
12594 (_mm512_maskz_cvt_roundph_epi64): Likewise.
12595 (_mm512_cvtph_epu64): Likewise.
12596 (_mm512_mask_cvtph_epu64): Likewise.
12597 (_mm512_maskz_cvtph_epu64): Likewise.
12598 (_mm512_cvt_roundph_epu64): Likewise.
12599 (_mm512_mask_cvt_roundph_epu64): Likewise.
12600 (_mm512_maskz_cvt_roundph_epu64): Likewise.
12601 (_mm512_cvtph_epi16): Likewise.
12602 (_mm512_mask_cvtph_epi16): Likewise.
12603 (_mm512_maskz_cvtph_epi16): Likewise.
12604 (_mm512_cvt_roundph_epi16): Likewise.
12605 (_mm512_mask_cvt_roundph_epi16): Likewise.
12606 (_mm512_maskz_cvt_roundph_epi16): Likewise.
12607 (_mm512_cvtph_epu16): Likewise.
12608 (_mm512_mask_cvtph_epu16): Likewise.
12609 (_mm512_maskz_cvtph_epu16): Likewise.
12610 (_mm512_cvt_roundph_epu16): Likewise.
12611 (_mm512_mask_cvt_roundph_epu16): Likewise.
12612 (_mm512_maskz_cvt_roundph_epu16): Likewise.
12613 * config/i386/avx512fp16vlintrin.h (_mm_cvtph_epi32):
12615 (_mm_mask_cvtph_epi32): Likewise.
12616 (_mm_maskz_cvtph_epi32): Likewise.
12617 (_mm256_cvtph_epi32): Likewise.
12618 (_mm256_mask_cvtph_epi32): Likewise.
12619 (_mm256_maskz_cvtph_epi32): Likewise.
12620 (_mm_cvtph_epu32): Likewise.
12621 (_mm_mask_cvtph_epu32): Likewise.
12622 (_mm_maskz_cvtph_epu32): Likewise.
12623 (_mm256_cvtph_epu32): Likewise.
12624 (_mm256_mask_cvtph_epu32): Likewise.
12625 (_mm256_maskz_cvtph_epu32): Likewise.
12626 (_mm_cvtph_epi64): Likewise.
12627 (_mm_mask_cvtph_epi64): Likewise.
12628 (_mm_maskz_cvtph_epi64): Likewise.
12629 (_mm256_cvtph_epi64): Likewise.
12630 (_mm256_mask_cvtph_epi64): Likewise.
12631 (_mm256_maskz_cvtph_epi64): Likewise.
12632 (_mm_cvtph_epu64): Likewise.
12633 (_mm_mask_cvtph_epu64): Likewise.
12634 (_mm_maskz_cvtph_epu64): Likewise.
12635 (_mm256_cvtph_epu64): Likewise.
12636 (_mm256_mask_cvtph_epu64): Likewise.
12637 (_mm256_maskz_cvtph_epu64): Likewise.
12638 (_mm_cvtph_epi16): Likewise.
12639 (_mm_mask_cvtph_epi16): Likewise.
12640 (_mm_maskz_cvtph_epi16): Likewise.
12641 (_mm256_cvtph_epi16): Likewise.
12642 (_mm256_mask_cvtph_epi16): Likewise.
12643 (_mm256_maskz_cvtph_epi16): Likewise.
12644 (_mm_cvtph_epu16): Likewise.
12645 (_mm_mask_cvtph_epu16): Likewise.
12646 (_mm_maskz_cvtph_epu16): Likewise.
12647 (_mm256_cvtph_epu16): Likewise.
12648 (_mm256_mask_cvtph_epu16): Likewise.
12649 (_mm256_maskz_cvtph_epu16): Likewise.
12650 * config/i386/i386-builtin-types.def: Add new builtin types.
12651 * config/i386/i386-builtin.def: Add new builtins.
12652 * config/i386/i386-expand.c
12653 (ix86_expand_args_builtin): Handle new builtin types.
12654 (ix86_expand_round_builtin): Ditto.
12655 * config/i386/sse.md (sseintconvert): New.
12656 (ssePHmode): Ditto.
12657 (UNSPEC_US_FIX_NOTRUNC): Ditto.
12658 (sseintconvertsignprefix): Ditto.
12659 (avx512fp16_vcvtph2<sseintconvertsignprefix><sseintconvert>_<mode><mask_name><round_name>):
12662 2021-09-16 liuhongt <hongtao.liu@intel.com>
12664 * config/i386/avx512fp16intrin.h: (_mm_cvtsi16_si128):
12666 (_mm_cvtsi128_si16): Likewise.
12667 (_mm_mask_load_sh): Likewise.
12668 (_mm_maskz_load_sh): Likewise.
12669 (_mm_mask_store_sh): Likewise.
12670 (_mm_move_sh): Likewise.
12671 (_mm_mask_move_sh): Likewise.
12672 (_mm_maskz_move_sh): Likewise.
12673 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
12674 * config/i386/i386-builtin.def: Add corresponding new builtins.
12675 * config/i386/i386-expand.c
12676 (ix86_expand_special_args_builtin): Handle new builtin types.
12677 (ix86_expand_vector_init_one_nonzero): Adjust for FP16 target.
12678 * config/i386/sse.md (VI2F): New mode iterator.
12679 (vec_set<mode>_0): Use new mode iterator.
12680 (avx512f_mov<ssescalarmodelower>_mask): Adjust for HF vector mode.
12681 (avx512f_store<mode>_mask): Ditto.
12683 2021-09-16 Kewen Lin <linkw@linux.ibm.com>
12685 * config/rs6000/rs6000.opt (-mtoc-fusion): Remove.
12687 2021-09-15 David Edelsohn <dje.gcc@gmail.com>
12689 * config/rs6000/rs6000.c (rs6000_xcoff_encode_section_info):
12690 Proceed if no symbol summary or the symbol alias flag is false.
12692 2021-09-15 Jakub Jelinek <jakub@redhat.com>
12696 * varasm.c (output_constructor_regular_field): Instead of assertion
12697 that array_size_for_constructor result is equal to size of
12698 TREE_TYPE (local->val) in bytes, assert that the type size is greater
12699 or equal to array_size_for_constructor result and use type size as
12702 2021-09-15 Martin Liska <mliska@suse.cz>
12705 * config/i386/vxworks.h: Use new macro TARGET_CPU_P.
12707 2021-09-15 Martin Liska <mliska@suse.cz>
12710 * config/rs6000/rs6000.c (rs6000_xcoff_encode_section_info):
12711 Check that we have a symbol summary for a symbol.
12713 2021-09-15 Richard Biener <rguenther@suse.de>
12716 * config/rs6000/lynx.h: Remove undef of PREFERRED_DEBUGGING_TYPE
12717 to inherit from elfos.h
12719 2021-09-15 liuhongt <hongtao.liu@intel.com>
12722 * config/i386/i386-expand.c
12723 (ix86_expand_vector_init_interleave): Use puncklwd to pack 2
12725 (ix86_expand_vector_set): Use blendw instead of pinsrw.
12726 * config/i386/i386.c (ix86_can_change_mode_class): Adjust for
12727 AVX512FP16 which supports 16bit vector load.
12728 * config/i386/sse.md (avx512bw_interleave_highv32hi<mask_name>):
12730 (avx512bw_interleave_high<mode><mask_name>): .. this, and
12731 extend to V32HFmode.
12732 (avx2_interleave_highv16hi<mask_name>): Rename to ..
12733 (avx2_interleave_high<mode><mask_name>): .. this, and extend
12735 (vec_interleave_highv8hi<mask_name>): Rename to ..
12736 (vec_interleave_high<mode><mask_name>): .. this, and extend to V8HFmode.
12737 (<mask_codefor>avx512bw_interleave_lowv32hi<mask_name>):
12739 (<mask_codefor>avx512bw_interleave_low<mode><mask_name>):
12740 this, and extend to V32HFmode.
12741 (avx2_interleave_lowv16hi<mask_name>): Rename to ..
12742 (avx2_interleave_low<mode><mask_name>): .. this, and extend to V16HFmode.
12743 (vec_interleave_lowv8hi<mask_name>): Rename to ..
12744 (vec_interleave_low<mode><mask_name>): .. this, and extend to V8HFmode.
12745 (sse4_1_pblendw): Rename to ..
12746 (sse4_1_pblend<blendsuf>): .. this, and extend to V8HFmode.
12747 (avx2_pblendph): New define_expand.
12748 (<sse2p4_1>_pinsr<ssemodesuffix>): Refactor, use
12749 sseintmodesuffix instead of ssemodesuffix.
12750 (blendsuf): New mode attr.
12752 2021-09-15 Richard Biener <rguenther@suse.de>
12754 * tree-vectorizer.h (dr_misalignment): Move out of line.
12755 (dr_target_alignment): New.
12756 (DR_TARGET_ALIGNMENT): Wrap dr_target_alignment.
12757 (set_dr_target_alignment): New.
12758 (SET_DR_TARGET_ALIGNMENT): Wrap set_dr_target_alignment.
12759 * tree-vect-data-refs.c (dr_misalignment): Compute and
12760 return the group members misalignment.
12761 (vect_compute_data_ref_alignment): Use SET_DR_TARGET_ALIGNMENT.
12762 (vect_analyze_data_refs_alignment): Compute alignment only
12763 for the first element of a DR group.
12764 (vect_slp_analyze_node_alignment): Likewise.
12766 2021-09-15 Hongyu Wang <hongyu.wang@intel.com>
12768 * config/i386/avx512fp16intrin.h: Adjust all builtin calls.
12769 * config/i386/avx512fp16vlintrin.h: Likewise.
12770 * config/i386/i386-builtin.def: Adjust builtin name and
12771 enumeration to match AVX512F style.
12773 2021-09-15 Richard Biener <rguenther@suse.de>
12775 PR tree-optimization/102318
12776 * tree-vect-loop.c (vect_transform_cycle_phi): Revert
12777 previous change and do the mode conversion separately from
12778 the sign conversion.
12780 2021-09-15 Hongtao Liu <hongtao.liu@intel.com>
12781 Peter Cordes <peter@cordes.ca>
12784 * config/i386/sse.md (extract_suf): Add V8SF/V8SI/V4DF/V4DI.
12785 (*vec_extract<mode><ssescalarmodelower>_valign): Output
12786 vextract{i,f}{32x4,64x2} instruction when byte_offset % 16 ==
12789 2021-09-15 Richard Biener <rguenther@suse.de>
12791 * config.gcc: Remove vax-*-openbsd* configuration.
12793 2021-09-15 Richard Biener <rguenther@suse.de>
12795 * config.gcc: Remove m68k-openbsd.
12797 2021-09-15 Max Filippov <jcmvbkbc@gmail.com>
12800 * config/xtensa/t-xtensa (TM_H): Add include/xtensa-config.h.
12802 2021-09-14 Peter Bergner <bergner@linux.ibm.com>
12804 * config/rs6000/mma.md (unspec): Delete UNSPEC_MMA_XXSETACCZ.
12805 (unspecv): Add UNSPECV_MMA_XXSETACCZ.
12806 (*mma_xxsetaccz): Delete.
12807 (mma_xxsetaccz): Change to define_insn. Remove operand 1.
12808 Use UNSPECV_MMA_XXSETACCZ. Update comment.
12809 * config/rs6000/rs6000.c (rs6000_rtx_costs): Use UNSPECV_MMA_XXSETACCZ.
12811 2021-09-14 Iain Sandoe <iain@sandoe.co.uk>
12813 * Makefile.in: Remove variables related to applying no-PIE
12814 to the exes on $build.
12815 * configure: Regenerate.
12816 * configure.ac: Remove configuration related to applying
12817 no-PIE to the exes on $build.
12819 2021-09-14 Claudiu Zissulescu <claziss@synopsys.com>
12821 * config/arc/arc.md (doloop_end): Add missing mode.
12822 (loop_end): Likewise.
12824 2021-09-14 Jakub Jelinek <jakub@redhat.com>
12826 * gimplify.c (goa_stabilize_expr): Add depth argument, propagate
12827 it to recursive calls, for depth above 7 just gimplify or return.
12828 Perform a test even for MODIFY_EXPR, ADDR_EXPR, COMPOUND_EXPR with
12829 __builtin_clear_padding and TARGET_EXPR.
12830 (gimplify_omp_atomic): Adjust goa_stabilize_expr callers.
12832 2021-09-14 liuhongt <hongtao.liu@intel.com>
12834 * config/i386/avx512fp16intrin.h (_mm_fpclass_sh_mask):
12836 (_mm_mask_fpclass_sh_mask): Likewise.
12837 (_mm512_mask_fpclass_ph_mask): Likewise.
12838 (_mm512_fpclass_ph_mask): Likewise.
12839 (_mm_getexp_sh): Likewise.
12840 (_mm_mask_getexp_sh): Likewise.
12841 (_mm_maskz_getexp_sh): Likewise.
12842 (_mm512_getexp_ph): Likewise.
12843 (_mm512_mask_getexp_ph): Likewise.
12844 (_mm512_maskz_getexp_ph): Likewise.
12845 (_mm_getexp_round_sh): Likewise.
12846 (_mm_mask_getexp_round_sh): Likewise.
12847 (_mm_maskz_getexp_round_sh): Likewise.
12848 (_mm512_getexp_round_ph): Likewise.
12849 (_mm512_mask_getexp_round_ph): Likewise.
12850 (_mm512_maskz_getexp_round_ph): Likewise.
12851 (_mm_getmant_sh): Likewise.
12852 (_mm_mask_getmant_sh): Likewise.
12853 (_mm_maskz_getmant_sh): Likewise.
12854 (_mm512_getmant_ph): Likewise.
12855 (_mm512_mask_getmant_ph): Likewise.
12856 (_mm512_maskz_getmant_ph): Likewise.
12857 (_mm_getmant_round_sh): Likewise.
12858 (_mm_mask_getmant_round_sh): Likewise.
12859 (_mm_maskz_getmant_round_sh): Likewise.
12860 (_mm512_getmant_round_ph): Likewise.
12861 (_mm512_mask_getmant_round_ph): Likewise.
12862 (_mm512_maskz_getmant_round_ph): Likewise.
12863 * config/i386/avx512fp16vlintrin.h (_mm_mask_fpclass_ph_mask):
12865 (_mm_fpclass_ph_mask): Likewise.
12866 (_mm256_mask_fpclass_ph_mask): Likewise.
12867 (_mm256_fpclass_ph_mask): Likewise.
12868 (_mm256_getexp_ph): Likewise.
12869 (_mm256_mask_getexp_ph): Likewise.
12870 (_mm256_maskz_getexp_ph): Likewise.
12871 (_mm_getexp_ph): Likewise.
12872 (_mm_mask_getexp_ph): Likewise.
12873 (_mm_maskz_getexp_ph): Likewise.
12874 (_mm256_getmant_ph): Likewise.
12875 (_mm256_mask_getmant_ph): Likewise.
12876 (_mm256_maskz_getmant_ph): Likewise.
12877 (_mm_getmant_ph): Likewise.
12878 (_mm_mask_getmant_ph): Likewise.
12879 (_mm_maskz_getmant_ph): Likewise.
12880 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
12881 * config/i386/i386-builtin.def: Add corresponding new builtins.
12882 * config/i386/i386-expand.c
12883 (ix86_expand_args_builtin): Handle new builtin types.
12884 (ix86_expand_round_builtin): Ditto.
12885 * config/i386/sse.md (vecmemsuffix): Add HF vector modes.
12886 (<avx512>_getexp<mode><mask_name><round_saeonly_name>): Adjust
12887 to support HF vector modes.
12888 (avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name):
12890 (avx512dq_fpclass<mode><mask_scalar_merge_name>): Ditto.
12891 (avx512dq_vmfpclass<mode><mask_scalar_merge_name>): Ditto.
12892 (<avx512>_getmant<mode><mask_name><round_saeonly_name>): Ditto.
12893 (avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
12896 2021-09-14 liuhongt <hongtao.liu@intel.com>
12898 * config/i386/avx512fp16intrin.h (_mm512_reduce_ph):
12900 (_mm512_mask_reduce_ph): Likewise.
12901 (_mm512_maskz_reduce_ph): Likewise.
12902 (_mm512_reduce_round_ph): Likewise.
12903 (_mm512_mask_reduce_round_ph): Likewise.
12904 (_mm512_maskz_reduce_round_ph): Likewise.
12905 (_mm_reduce_sh): Likewise.
12906 (_mm_mask_reduce_sh): Likewise.
12907 (_mm_maskz_reduce_sh): Likewise.
12908 (_mm_reduce_round_sh): Likewise.
12909 (_mm_mask_reduce_round_sh): Likewise.
12910 (_mm_maskz_reduce_round_sh): Likewise.
12911 (_mm512_roundscale_ph): Likewise.
12912 (_mm512_mask_roundscale_ph): Likewise.
12913 (_mm512_maskz_roundscale_ph): Likewise.
12914 (_mm512_roundscale_round_ph): Likewise.
12915 (_mm512_mask_roundscale_round_ph): Likewise.
12916 (_mm512_maskz_roundscale_round_ph): Likewise.
12917 (_mm_roundscale_sh): Likewise.
12918 (_mm_mask_roundscale_sh): Likewise.
12919 (_mm_maskz_roundscale_sh): Likewise.
12920 (_mm_roundscale_round_sh): Likewise.
12921 (_mm_mask_roundscale_round_sh): Likewise.
12922 (_mm_maskz_roundscale_round_sh): Likewise.
12923 * config/i386/avx512fp16vlintrin.h: (_mm_reduce_ph):
12925 (_mm_mask_reduce_ph): Likewise.
12926 (_mm_maskz_reduce_ph): Likewise.
12927 (_mm256_reduce_ph): Likewise.
12928 (_mm256_mask_reduce_ph): Likewise.
12929 (_mm256_maskz_reduce_ph): Likewise.
12930 (_mm_roundscale_ph): Likewise.
12931 (_mm_mask_roundscale_ph): Likewise.
12932 (_mm_maskz_roundscale_ph): Likewise.
12933 (_mm256_roundscale_ph): Likewise.
12934 (_mm256_mask_roundscale_ph): Likewise.
12935 (_mm256_maskz_roundscale_ph): Likewise.
12936 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
12937 * config/i386/i386-builtin.def: Add corresponding new builtins.
12938 * config/i386/i386-expand.c
12939 (ix86_expand_args_builtin): Handle new builtin types.
12940 (ix86_expand_round_builtin): Ditto.
12941 * config/i386/sse.md (<mask_codefor>reducep<mode><mask_name>):
12943 (<mask_codefor>reducep<mode><mask_name><round_saeonly_name>):
12944 ... this, and adjust for round operands.
12945 (reduces<mode><mask_scalar_name>): Likewise, with ...
12946 (reduces<mode><mask_scalar_name><round_saeonly_scalar_name):
12948 (<avx512>_rndscale<mode><mask_name><round_saeonly_name>):
12949 Adjust for HF vector modes.
12950 (avx512f_rndscale<mode><mask_scalar_name><round_saeonly_scalar_name>):
12952 (*avx512f_rndscale<mode><round_saeonly_name>): Ditto.
12954 2021-09-14 liuhongt <hongtao.liu@intel.com>
12956 * config/i386/avx512fp16intrin.h: (_mm512_rcp_ph):
12958 (_mm512_mask_rcp_ph): Likewise.
12959 (_mm512_maskz_rcp_ph): Likewise.
12960 (_mm_rcp_sh): Likewise.
12961 (_mm_mask_rcp_sh): Likewise.
12962 (_mm_maskz_rcp_sh): Likewise.
12963 (_mm512_scalef_ph): Likewise.
12964 (_mm512_mask_scalef_ph): Likewise.
12965 (_mm512_maskz_scalef_ph): Likewise.
12966 (_mm512_scalef_round_ph): Likewise.
12967 (_mm512_mask_scalef_round_ph): Likewise.
12968 (_mm512_maskz_scalef_round_ph): Likewise.
12969 (_mm_scalef_sh): Likewise.
12970 (_mm_mask_scalef_sh): Likewise.
12971 (_mm_maskz_scalef_sh): Likewise.
12972 (_mm_scalef_round_sh): Likewise.
12973 (_mm_mask_scalef_round_sh): Likewise.
12974 (_mm_maskz_scalef_round_sh): Likewise.
12975 * config/i386/avx512fp16vlintrin.h (_mm_rcp_ph):
12977 (_mm256_rcp_ph): Likewise.
12978 (_mm_mask_rcp_ph): Likewise.
12979 (_mm256_mask_rcp_ph): Likewise.
12980 (_mm_maskz_rcp_ph): Likewise.
12981 (_mm256_maskz_rcp_ph): Likewise.
12982 (_mm_scalef_ph): Likewise.
12983 (_mm256_scalef_ph): Likewise.
12984 (_mm_mask_scalef_ph): Likewise.
12985 (_mm256_mask_scalef_ph): Likewise.
12986 (_mm_maskz_scalef_ph): Likewise.
12987 (_mm256_maskz_scalef_ph): Likewise.
12988 * config/i386/i386-builtin.def: Add new builtins.
12989 * config/i386/sse.md (VFH_AVX512VL): New.
12990 (avx512fp16_rcp<mode>2<mask_name>): Ditto.
12991 (avx512fp16_vmrcpv8hf2<mask_scalar_name>): Ditto.
12992 (avx512f_vmscalef<mode><mask_scalar_name><round_scalar_name>):
12993 Adjust to support HF vector modes.
12994 (<avx512>_scalef<mode><mask_name><round_name>): Ditto.
12996 2021-09-14 liuhongt <hongtao.liu@intel.com>
12998 * config/i386/avx512fp16intrin.h: (_mm512_sqrt_ph):
13000 (_mm512_mask_sqrt_ph): Likewise.
13001 (_mm512_maskz_sqrt_ph): Likewise.
13002 (_mm512_sqrt_round_ph): Likewise.
13003 (_mm512_mask_sqrt_round_ph): Likewise.
13004 (_mm512_maskz_sqrt_round_ph): Likewise.
13005 (_mm512_rsqrt_ph): Likewise.
13006 (_mm512_mask_rsqrt_ph): Likewise.
13007 (_mm512_maskz_rsqrt_ph): Likewise.
13008 (_mm_rsqrt_sh): Likewise.
13009 (_mm_mask_rsqrt_sh): Likewise.
13010 (_mm_maskz_rsqrt_sh): Likewise.
13011 (_mm_sqrt_sh): Likewise.
13012 (_mm_mask_sqrt_sh): Likewise.
13013 (_mm_maskz_sqrt_sh): Likewise.
13014 (_mm_sqrt_round_sh): Likewise.
13015 (_mm_mask_sqrt_round_sh): Likewise.
13016 (_mm_maskz_sqrt_round_sh): Likewise.
13017 * config/i386/avx512fp16vlintrin.h (_mm_sqrt_ph): New intrinsic.
13018 (_mm256_sqrt_ph): Likewise.
13019 (_mm_mask_sqrt_ph): Likewise.
13020 (_mm256_mask_sqrt_ph): Likewise.
13021 (_mm_maskz_sqrt_ph): Likewise.
13022 (_mm256_maskz_sqrt_ph): Likewise.
13023 (_mm_rsqrt_ph): Likewise.
13024 (_mm256_rsqrt_ph): Likewise.
13025 (_mm_mask_rsqrt_ph): Likewise.
13026 (_mm256_mask_rsqrt_ph): Likewise.
13027 (_mm_maskz_rsqrt_ph): Likewise.
13028 (_mm256_maskz_rsqrt_ph): Likewise.
13029 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
13030 * config/i386/i386-builtin.def: Add corresponding new builtins.
13031 * config/i386/i386-expand.c
13032 (ix86_expand_args_builtin): Handle new builtins.
13033 (ix86_expand_round_builtin): Ditto.
13034 * config/i386/sse.md (VF_AVX512FP16VL): New.
13035 (sqrt<mode>2): Adjust for HF vector modes.
13036 (<sse>_sqrt<mode>2<mask_name><round_name>): Likewise.
13037 (<sse>_vmsqrt<mode>2<mask_scalar_name><round_scalar_name>):
13039 (<sse>_rsqrt<mode>2<mask_name>): New.
13040 (avx512fp16_vmrsqrtv8hf2<mask_scalar_name>): Likewise.
13042 2021-09-13 Thomas Schwinge <thomas@codesourcery.com>
13044 PR bootstrap/101574
13045 * diagnostic-spec.c (warning_suppressed_at, copy_warning): Handle
13046 'RESERVED_LOCATION_P' locations.
13047 * warning-control.cc (get_nowarn_spec, suppress_warning)
13048 (copy_warning): Likewise.
13050 2021-09-13 Thomas Schwinge <thomas@codesourcery.com>
13052 * diagnostic-spec.h (typedef xint_hash_t): Use 'location_t' instead of...
13053 (typedef key_type_t): ... this. Remove.
13054 (nowarn_map): Document.
13055 * diagnostic-spec.c (nowarn_map): Likewise.
13056 * warning-control.cc (convert_to_key): Evolve functions into...
13057 (get_location): ... these. Adjust all users.
13059 2021-09-13 Thomas Schwinge <thomas@codesourcery.com>
13061 * warning-control.cc (copy_warning): Remove 'nowarn_map' setup.
13063 2021-09-13 Jason Merrill <jason@redhat.com>
13065 * params.opt: Add destructive-interference-size and
13066 constructive-interference-size.
13067 * doc/invoke.texi: Document them.
13068 * config/aarch64/aarch64.c (aarch64_override_options_internal):
13070 * config/arm/arm.c (arm_option_override): Set them.
13071 * config/i386/i386-options.c (ix86_option_override_internal):
13074 2021-09-13 Martin Liska <mliska@suse.cz>
13075 H.J. Lu <hjl.tools@gmail.com>
13078 * common/config/i386/cpuinfo.h (cpu_indicator_init): Add support
13079 for x86-64 micro levels for __builtin_cpu_supports.
13080 * common/config/i386/i386-cpuinfo.h (enum feature_priority):
13081 Add priorities for the micro-arch levels.
13082 (enum processor_features): Add new features.
13083 * common/config/i386/i386-isas.h: Add micro-arch features.
13084 * config/i386/i386-builtins.c (get_builtin_code_for_version):
13085 Support the micro-arch levels by callsing
13086 __builtin_cpu_supports.
13087 * doc/extend.texi: Document that the levels are support by
13088 __builtin_cpu_supports.
13090 2021-09-13 Andrew Pinski <apinski@marvell.com>
13093 * config/aarch64/aarch64-builtins.c (aarch64_fold_builtin_lane_check):
13095 (aarch64_general_fold_builtin): Handle AARCH64_SIMD_BUILTIN_LANE_CHECK.
13096 (aarch64_general_gimple_fold_builtin): Likewise.
13098 2021-09-13 Andrew Pinski <apinski@marvell.com>
13100 * config.gcc: Add m32r-*-linux* and m32rle-*-linux*
13101 to the Unsupported targets list.
13102 Remove support for m32r-*-linux* and m32rle-*-linux*.
13103 * config/m32r/linux.h: Removed.
13104 * config/m32r/t-linux: Removed.
13106 2021-09-13 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13109 * config/aarch64/aarch64.c (aarch64_classify_address): Don't allow
13110 register index for SVE predicate modes.
13112 2021-09-13 Aldy Hernandez <aldyh@redhat.com>
13114 * tree-ssa-threadbackward.c
13115 (back_threader_profitability::profitable_path_p): Remove FSM
13117 (back_threader_registry::register_path): Same.
13118 * tree-ssa-threadedge.c
13119 (jump_threader::simplify_control_stmt_condition): Same.
13120 * tree-ssa-threadupdate.c (jt_path_registry::jt_path_registry):
13121 Add backedge_threads argument.
13122 (fwd_jt_path_registry::fwd_jt_path_registry): Pass
13123 backedge_threads argument.
13124 (back_jt_path_registry::back_jt_path_registry): Same.
13125 (dump_jump_thread_path): Adjust for FSM removal.
13126 (back_jt_path_registry::rewire_first_differing_edge): Same.
13127 (back_jt_path_registry::adjust_paths_after_duplication): Same.
13128 (back_jt_path_registry::update_cfg): Same.
13129 (jt_path_registry::register_jump_thread): Same.
13130 * tree-ssa-threadupdate.h (enum jump_thread_edge_type): Remove
13132 (class back_jt_path_registry): Add backedge_threads to
13135 2021-09-13 Martin Liska <mliska@suse.cz>
13138 * asan.h (sanitize_coverage_p): Handle when fn == NULL.
13140 2021-09-13 H.J. Lu <hjl.tools@gmail.com>
13143 * config/i386/i386.h (TARGET_AVX256_MOVE_BY_PIECES): New.
13144 (TARGET_AVX256_STORE_BY_PIECES): Likewise.
13145 (MOVE_MAX): Check TARGET_AVX256_MOVE_BY_PIECES and
13146 TARGET_AVX256_STORE_BY_PIECES instead of
13147 TARGET_AVX256_SPLIT_UNALIGNED_LOAD and
13148 TARGET_AVX256_SPLIT_UNALIGNED_STORE.
13149 (STORE_MAX_PIECES): Check TARGET_AVX256_STORE_BY_PIECES instead
13150 of TARGET_AVX256_SPLIT_UNALIGNED_STORE.
13151 * config/i386/x86-tune.def (X86_TUNE_AVX256_MOVE_BY_PIECES): New.
13152 (X86_TUNE_AVX256_STORE_BY_PIECES): Likewise.
13154 2021-09-13 liuhongt <hongtao.liu@intel.com>
13156 PR bootstrap/102302
13157 * expmed.c (extract_bit_field_using_extv): Use
13158 gen_lowpart_if_possible instead of gen_lowpart to avoid ICE.
13160 2021-09-13 Aldy Hernandez <aldyh@redhat.com>
13162 * Makefile.in (OBJS): Add value-pointer-equiv.o.
13163 * gimple-ssa-evrp.c (class ssa_equiv_stack): Move to
13164 value-pointer-equiv.*.
13165 (ssa_equiv_stack::ssa_equiv_stack): Same.
13166 (ssa_equiv_stack::enter): Same.
13167 (ssa_equiv_stack::leave): Same.
13168 (ssa_equiv_stack::push_replacement): Same.
13169 (ssa_equiv_stack::get_replacement): Same.
13170 (is_pointer_ssa): Same.
13171 (class pointer_equiv_analyzer): Same.
13172 (pointer_equiv_analyzer::pointer_equiv_analyzer): Same.
13173 (pointer_equiv_analyzer::~pointer_equiv_analyzer): Same.
13174 (pointer_equiv_analyzer::set_global_equiv): Same.
13175 (pointer_equiv_analyzer::set_cond_equiv): Same.
13176 (pointer_equiv_analyzer::get_equiv): Same.
13177 (pointer_equiv_analyzer::enter): Same.
13178 (pointer_equiv_analyzer::leave): Same.
13179 (pointer_equiv_analyzer::get_equiv_expr): Same.
13180 (pta_valueize): Same.
13181 (pointer_equiv_analyzer::visit_stmt): Same.
13182 (pointer_equiv_analyzer::visit_edge): Same.
13183 (hybrid_folder::value_of_expr): Same.
13184 (hybrid_folder::value_on_edge): Same.
13185 * value-pointer-equiv.cc: New file.
13186 * value-pointer-equiv.h: New file.
13188 2021-09-13 Richard Earnshaw <rearnsha@arm.com>
13191 * gimple-fold.c (gimple_fold_builtin_memory_op): Allow folding
13192 memcpy if the size is not more than MOVE_MAX * MOVE_RATIO.
13194 2021-09-13 Richard Earnshaw <rearnsha@arm.com>
13197 * config/arm/arm.md (movmisaligndi): New define_expand.
13198 * config/arm/vec-common.md (movmisalign<mode>): Iterate over VDQ mode.
13200 2021-09-13 Richard Earnshaw <rearnsha@arm.com>
13203 * emit-rtl.c (gen_highpart): Use adjust_address to handle
13204 MEM rather than calling simplify_gen_subreg.
13206 2021-09-13 Jan-Benedict Glaw <jbglaw@ług-owl.de>
13208 * config/alpha/vms.h (INIT_CUMULATIVE_ARGS): Wrap multi-statment
13209 define into a block.
13211 2021-09-13 Richard Biener <rguenther@suse.de>
13213 * config/darwin.h (DARWIN_PREFER_DWARF): Do not define.
13214 * config/i386/darwin.h (PREFERRED_DEBUGGING_TYPE): Do not
13215 change based on DARWIN_PREFER_DWARF not being defined.
13217 2021-09-13 Richard Biener <rguenther@suse.de>
13219 * config/i386/lynx.h: Remove undef of PREFERRED_DEBUGGING_TYPE
13220 to inherit from elfos.h
13222 2021-09-13 Richard Biener <rguenther@suse.de>
13224 * config.gcc: Add cr16-*-* to the list of obsoleted targets.
13226 2021-09-13 Richard Biener <rguenther@suse.de>
13228 * config/avr/elf.h (PREFERRED_DEBUGGING_TYPE): Remove
13229 override, pick up DWARF2_DEBUG define from elfos.h
13231 2021-09-13 Richard Biener <rguenther@suse.de>
13233 * config/rx/rx.h (PREFERRED_DEBUGGING_TYPE): Always define to
13236 2021-09-13 Richard Biener <rguenther@suse.de>
13238 * config/alpha/vms.h (PREFERRED_DEBUGGING_TYPE): Define to
13241 2021-09-13 Richard Biener <rguenther@suse.de>
13243 * config/i386/cygming.h: Always default to DWARF2 debugging.
13244 Do not define DBX_DEBUGGING_INFO, that's done via dbxcoff.h
13246 * doc/install.texi: Document binutils 2.16 as minimum
13247 requirement for mingw.
13249 2021-09-13 Kewen Lin <linkw@linux.ibm.com>
13251 * config/rs6000/rs6000.c (struct rs6000_cost_data): New members
13252 nstmts, nloads and extra_ctor_cost.
13253 (rs6000_density_test): Add load density related heuristics. Do
13254 extra costing on vector construction statements if need.
13255 (rs6000_init_cost): Init new members.
13256 (rs6000_update_target_cost_per_stmt): New function.
13257 (rs6000_add_stmt_cost): Factor vect_nonmem hunk out to function
13258 rs6000_update_target_cost_per_stmt and call it.
13260 2021-09-13 Kewen Lin <linkw@linux.ibm.com>
13262 * config/rs6000/rs6000.c (struct rs6000_cost_data): Remove typedef.
13263 (rs6000_init_cost): Adjust.
13265 2021-09-13 liuhongt <hongtao.liu@intel.com>
13267 * config/i386/i386.md: (UNSPEC_COPYSIGN): Remove.
13268 (UNSPEC_XORSIGN): Ditto.
13270 2021-09-12 Roger Sayle <roger@nextmovesoftware.com>
13272 * expr.c (convert_move): Preserve SUBREG_PROMOTED_VAR_P when
13273 creating a (wider) partial subreg from a SUBREG_PROMOTED_VAR_P
13276 2021-09-11 Aldy Hernandez <aldyh@redhat.com>
13278 * tree-ssa-threadbackward.c (class back_threader_registry): Use
13279 back_jt_path_registry.
13280 * tree-ssa-threadedge.c (jump_threader::jump_threader): Use
13281 fwd_jt_path_registry.
13282 * tree-ssa-threadedge.h (class jump_threader): Same..
13283 * tree-ssa-threadupdate.c
13284 (jump_thread_path_registry::jump_thread_path_registry): Rename...
13285 (jt_path_registry::jt_path_registry): ...to this.
13286 (jump_thread_path_registry::~jump_thread_path_registry): Rename...
13287 (jt_path_registry::~jt_path_registry): ...this.
13288 (fwd_jt_path_registry::fwd_jt_path_registry): New.
13289 (fwd_jt_path_registry::~fwd_jt_path_registry): New.
13290 (jump_thread_path_registry::allocate_thread_edge): Rename...
13291 (jt_path_registry::allocate_thread_edge): ...to this.
13292 (jump_thread_path_registry::allocate_thread_path): Rename...
13293 (jt_path_registry::allocate_thread_path): ...to this.
13294 (jump_thread_path_registry::lookup_redirection_data): Rename...
13295 (fwd_jt_path_registry::lookup_redirection_data): ...to this.
13296 (jump_thread_path_registry::thread_block_1): Rename...
13297 (fwd_jt_path_registry::thread_block_1): ...to this.
13298 (jump_thread_path_registry::thread_block): Rename...
13299 (fwd_jt_path_registry::thread_block): ...to this.
13300 (jt_path_registry::thread_through_loop_header): Rename...
13301 (fwd_jt_path_registry::thread_through_loop_header): ...to this.
13302 (jump_thread_path_registry::mark_threaded_blocks): Rename...
13303 (fwd_jt_path_registry::mark_threaded_blocks): ...to this.
13304 (jump_thread_path_registry::debug_path): Rename...
13305 (jt_path_registry::debug_path): ...to this.
13306 (jump_thread_path_registry::dump): Rename...
13307 (jt_path_registry::debug): ...to this.
13308 (jump_thread_path_registry::rewire_first_differing_edge): Rename...
13309 (back_jt_path_registry::rewire_first_differing_edge): ...to this.
13310 (jump_thread_path_registry::adjust_paths_after_duplication): Rename...
13311 (back_jt_path_registry::adjust_paths_after_duplication): ...to this.
13312 (jump_thread_path_registry::duplicate_thread_path): Rename...
13313 (back_jt_path_registry::duplicate_thread_path): ...to this. Also,
13314 drop ill-formed candidates.
13315 (jump_thread_path_registry::remove_jump_threads_including): Rename...
13316 (fwd_jt_path_registry::remove_jump_threads_including): ...to this.
13317 (jt_path_registry::thread_through_all_blocks): New.
13318 (back_jt_path_registry::update_cfg): New.
13319 (fwd_jt_path_registry::update_cfg): New.
13320 (jump_thread_path_registry::register_jump_thread): Rename...
13321 (jt_path_registry::register_jump_thread): ...to this.
13322 * tree-ssa-threadupdate.h (class jump_thread_path_registry):
13324 (class jt_path_registry): ...here.
13325 (class fwd_jt_path_registry): New.
13326 (class back_jt_path_registry): New.
13328 2021-09-10 liuhongt <hongtao.liu@intel.com>
13331 2021-09-01 liuhongt <hongtao.liu@intel.com>
13333 * emit-rtl.c (validate_subreg): Get rid of all float-int
13336 2021-09-10 Jakub Jelinek <jakub@redhat.com>
13338 * tree-core.h (enum omp_memory_order): Add OMP_MEMORY_ORDER_MASK,
13339 OMP_FAIL_MEMORY_ORDER_UNSPECIFIED, OMP_FAIL_MEMORY_ORDER_RELAXED,
13340 OMP_FAIL_MEMORY_ORDER_ACQUIRE, OMP_FAIL_MEMORY_ORDER_RELEASE,
13341 OMP_FAIL_MEMORY_ORDER_ACQ_REL, OMP_FAIL_MEMORY_ORDER_SEQ_CST and
13342 OMP_FAIL_MEMORY_ORDER_MASK enumerators.
13343 (OMP_FAIL_MEMORY_ORDER_SHIFT): Define.
13344 * gimple-pretty-print.c (dump_gimple_omp_atomic_load,
13345 dump_gimple_omp_atomic_store): Print [weak] for weak atomic
13347 * gimple.h (enum gf_mask): Change GF_OMP_ATOMIC_MEMORY_ORDER
13348 to 6-bit mask, adjust GF_OMP_ATOMIC_NEED_VALUE value and add
13349 GF_OMP_ATOMIC_WEAK.
13350 (gimple_omp_atomic_weak_p, gimple_omp_atomic_set_weak): New inline
13352 * tree.h (OMP_ATOMIC_WEAK): Define.
13353 * tree-pretty-print.c (dump_omp_atomic_memory_order): Adjust for
13354 fail memory order being encoded in the same enum and also print
13355 fail clause if present.
13356 (dump_generic_node): Print weak clause if OMP_ATOMIC_WEAK.
13357 * gimplify.c (goa_stabilize_expr): Add target_expr and rhs arguments,
13358 handle pre_p == NULL case as a test mode that only returns value
13359 but doesn't change gimplify nor change anything otherwise, adjust
13360 recursive calls, add MODIFY_EXPR, ADDR_EXPR, COND_EXPR, TARGET_EXPR
13361 and CALL_EXPR handling, adjust COMPOUND_EXPR handling for
13362 __builtin_clear_padding calls, for !rhs gimplify as lvalue rather
13364 (gimplify_omp_atomic): Adjust goa_stabilize_expr caller. Handle
13365 COND_EXPR rhs. Set weak flag on gimple load/store for
13367 * omp-expand.c (omp_memory_order_to_fail_memmodel): New function.
13368 (omp_memory_order_to_memmodel): Adjust for fail clause encoded
13370 (expand_omp_atomic_cas): New function.
13371 (expand_omp_atomic_pipeline): Use omp_memory_order_to_fail_memmodel
13373 (expand_omp_atomic): Attempt to optimize atomic compare and exchange
13374 using expand_omp_atomic_cas.
13376 2021-09-10 Aldy Hernandez <aldyh@redhat.com>
13377 Michael Matz <matz@suse.de>
13379 * tree-pass.h (PROP_loop_opts_done): New.
13380 * gimple-range-path.cc (path_range_query::internal_range_of_expr):
13381 Intersect with global range.
13382 * tree-ssa-loop.c (tree_ssa_loop_done): Set PROP_loop_opts_done.
13383 * tree-ssa-threadbackward.c
13384 (back_threader_profitability::profitable_path_p): Disable
13385 threading through latches until after loop optimizations have run.
13387 2021-09-10 David Faust <david.faust@oracle.com>
13389 * doc/invoke.texi: Document BPF -mcpu, -mjmpext, -mjmp32 and -malu32
13392 2021-09-10 David Faust <david.faust@oracle.com>
13394 * config/bpf/bpf-opts.h (bpf_isa_version): New enum.
13395 * config/bpf/bpf-protos.h (bpf_expand_cbranch): New.
13396 * config/bpf/bpf.c (bpf_option_override): Handle -mcpu option.
13397 (bpf_expand_cbranch): New function.
13398 * config/bpf/bpf.md (AM mode iterator): Conditionalize support for SI
13400 (zero_extendsidi2): Only use mov32 instruction if it is available.
13401 (SIM mode iterator): Conditionalize support for SI mode.
13402 (JM mode iterator): New.
13403 (cbranchdi4): Update name, use new JM iterator. Use bpf_expand_cbranch.
13404 (*branch_on_di): Update name, use new JM iterator.
13405 * config/bpf/bpf.opt: (mjmpext): New option.
13406 (malu32): Likewise.
13407 (mjmp32): Likewise.
13409 (bpf_isa): New enum.
13411 2021-09-10 David Faust <david.faust@oracle.com>
13413 * config/bpf/bpf.md (zero_extendhidi2): Add new output template
13414 for register-to-register extensions.
13415 (zero_extendqidi2): Likewise.
13417 2021-09-10 Richard Biener <rguenther@suse.de>
13419 PR middle-end/102273
13420 * internal-fn.c (expand_DEFERRED_INIT): Always expand non-SSA vars.
13422 2021-09-10 Richard Biener <rguenther@suse.de>
13424 PR middle-end/102269
13425 * gimplify.c (is_var_need_auto_init): Empty types do not need
13428 2021-09-10 Richard Biener <rguenther@suse.de>
13430 * configure.ac (--with-stabs): Remove.
13431 * configure: Regenerate.
13432 * doc/install.texi: Remove --with-stabs documentation.
13434 2021-09-10 liuhongt <hongtao.liu@intel.com>
13436 * config/i386/avx512fp16intrin.h: (_mm512_cmp_ph_mask):
13438 (_mm512_mask_cmp_ph_mask): Likewise.
13439 (_mm512_cmp_round_ph_mask): Likewise.
13440 (_mm512_mask_cmp_round_ph_mask): Likewise.
13441 (_mm_cmp_sh_mask): Likewise.
13442 (_mm_mask_cmp_sh_mask): Likewise.
13443 (_mm_cmp_round_sh_mask): Likewise.
13444 (_mm_mask_cmp_round_sh_mask): Likewise.
13445 (_mm_comieq_sh): Likewise.
13446 (_mm_comilt_sh): Likewise.
13447 (_mm_comile_sh): Likewise.
13448 (_mm_comigt_sh): Likewise.
13449 (_mm_comige_sh): Likewise.
13450 (_mm_comineq_sh): Likewise.
13451 (_mm_ucomieq_sh): Likewise.
13452 (_mm_ucomilt_sh): Likewise.
13453 (_mm_ucomile_sh): Likewise.
13454 (_mm_ucomigt_sh): Likewise.
13455 (_mm_ucomige_sh): Likewise.
13456 (_mm_ucomineq_sh): Likewise.
13457 (_mm_comi_round_sh): Likewise.
13458 (_mm_comi_sh): Likewise.
13459 * config/i386/avx512fp16vlintrin.h (_mm_cmp_ph_mask): New intrinsic.
13460 (_mm_mask_cmp_ph_mask): Likewise.
13461 (_mm256_cmp_ph_mask): Likewise.
13462 (_mm256_mask_cmp_ph_mask): Likewise.
13463 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
13464 * config/i386/i386-builtin.def: Add corresponding new builtins.
13465 * config/i386/i386-expand.c
13466 (ix86_expand_args_builtin): Handle new builtin types.
13467 (ix86_expand_round_builtin): Ditto.
13468 * config/i386/i386.md (ssevecmode): Add HF mode.
13469 (MODEFH): New mode iterator.
13470 * config/i386/sse.md
13471 (V48H_AVX512VL): New mode iterator to support HF vector modes.
13472 Ajdust corresponding description.
13473 (ssecmpintprefix): New.
13474 (VI12_AVX512VL): Adjust to support HF vector modes.
13475 (cmp_imm_predicate): Likewise.
13476 (<avx512>_cmp<mode>3<mask_scalar_merge_name><round_saeonly_name>):
13478 (avx512f_vmcmp<mode>3<round_saeonly_name>): Likewise.
13479 (avx512f_vmcmp<mode>3_mask<round_saeonly_name>): Likewise.
13480 (<sse>_<unord>comi<round_saeonly_name>): Likewise.
13482 2021-09-10 liuhongt <hongtao.liu@intel.com>
13484 * config/i386/avx512fp16intrin.h: (_mm512_max_ph): New intrinsic.
13485 (_mm512_mask_max_ph): Likewise.
13486 (_mm512_maskz_max_ph): Likewise.
13487 (_mm512_min_ph): Likewise.
13488 (_mm512_mask_min_ph): Likewise.
13489 (_mm512_maskz_min_ph): Likewise.
13490 (_mm512_max_round_ph): Likewise.
13491 (_mm512_mask_max_round_ph): Likewise.
13492 (_mm512_maskz_max_round_ph): Likewise.
13493 (_mm512_min_round_ph): Likewise.
13494 (_mm512_mask_min_round_ph): Likewise.
13495 (_mm512_maskz_min_round_ph): Likewise.
13496 (_mm_max_sh): Likewise.
13497 (_mm_mask_max_sh): Likewise.
13498 (_mm_maskz_max_sh): Likewise.
13499 (_mm_min_sh): Likewise.
13500 (_mm_mask_min_sh): Likewise.
13501 (_mm_maskz_min_sh): Likewise.
13502 (_mm_max_round_sh): Likewise.
13503 (_mm_mask_max_round_sh): Likewise.
13504 (_mm_maskz_max_round_sh): Likewise.
13505 (_mm_min_round_sh): Likewise.
13506 (_mm_mask_min_round_sh): Likewise.
13507 (_mm_maskz_min_round_sh): Likewise.
13508 * config/i386/avx512fp16vlintrin.h (_mm_max_ph): New intrinsic.
13509 (_mm256_max_ph): Likewise.
13510 (_mm_mask_max_ph): Likewise.
13511 (_mm256_mask_max_ph): Likewise.
13512 (_mm_maskz_max_ph): Likewise.
13513 (_mm256_maskz_max_ph): Likewise.
13514 (_mm_min_ph): Likewise.
13515 (_mm256_min_ph): Likewise.
13516 (_mm_mask_min_ph): Likewise.
13517 (_mm256_mask_min_ph): Likewise.
13518 (_mm_maskz_min_ph): Likewise.
13519 (_mm256_maskz_min_ph): Likewise.
13520 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
13521 * config/i386/i386-builtin.def: Add corresponding new builtins.
13522 * config/i386/i386-expand.c
13523 (ix86_expand_args_builtin): Handle new builtin types.
13524 * config/i386/sse.md
13525 (<code><mode>3<mask_name><round_saeonly_name>): Adjust to
13526 support HF vector modes.
13527 (*<code><mode>3<mask_name><round_saeonly_name>): Likewise.
13528 (ieee_<ieee_maxmin><mode>3<mask_name><round_saeonly_name>):
13530 (<sse>_vm<code><mode>3<mask_scalar_name><round_saeonly_scalar_name>):
13532 * config/i386/subst.md (round_saeonly_mode512bit_condition):
13533 Adjust for HF vector modes.
13535 2021-09-10 Liu, Hongtao <hongtao.liu@intel.com>
13537 * config/i386/avx512fp16intrin.h (_mm_add_sh): New intrinsic.
13538 (_mm_mask_add_sh): Likewise.
13539 (_mm_maskz_add_sh): Likewise.
13540 (_mm_sub_sh): Likewise.
13541 (_mm_mask_sub_sh): Likewise.
13542 (_mm_maskz_sub_sh): Likewise.
13543 (_mm_mul_sh): Likewise.
13544 (_mm_mask_mul_sh): Likewise.
13545 (_mm_maskz_mul_sh): Likewise.
13546 (_mm_div_sh): Likewise.
13547 (_mm_mask_div_sh): Likewise.
13548 (_mm_maskz_div_sh): Likewise.
13549 (_mm_add_round_sh): Likewise.
13550 (_mm_mask_add_round_sh): Likewise.
13551 (_mm_maskz_add_round_sh): Likewise.
13552 (_mm_sub_round_sh): Likewise.
13553 (_mm_mask_sub_round_sh): Likewise.
13554 (_mm_maskz_sub_round_sh): Likewise.
13555 (_mm_mul_round_sh): Likewise.
13556 (_mm_mask_mul_round_sh): Likewise.
13557 (_mm_maskz_mul_round_sh): Likewise.
13558 (_mm_div_round_sh): Likewise.
13559 (_mm_mask_div_round_sh): Likewise.
13560 (_mm_maskz_div_round_sh): Likewise.
13561 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
13562 * config/i386/i386-builtin.def: Add corresponding new builtins.
13563 * config/i386/i386-expand.c
13564 (ix86_expand_round_builtin): Handle new builtins.
13565 * config/i386/sse.md (VF_128): Change description.
13566 (<sse>_vm<plusminus_insn><mode>3<mask_scalar_name><round_scalar_name>):
13567 Adjust to support HF vector modes.
13568 (<sse>_vm<multdiv_mnemonic><mode>3<mask_scalar_name><round_scalar_name>):
13571 2021-09-10 H.J. Lu <hjl.tools@gmail.com>
13573 * config/i386/i386-expand.c
13574 (ix86_avx256_split_vector_move_misalign): Handle V16HF mode.
13575 * config/i386/i386.c
13576 (ix86_preferred_simd_mode): Handle HF mode.
13577 * config/i386/sse.md (V_256H): New mode iterator.
13578 (avx_vextractf128<mode>): Use it.
13579 (VEC_INIT_MODE): Align vector HFmode condition to vector
13580 HImodes since there're no real HF instruction used.
13581 (VEC_INIT_HALF_MODE): Ditto.
13583 (VIHF_AVX512BW): Ditto.
13584 (*vec_extracthf): Ditto.
13585 (VEC_EXTRACT_MODE): Ditto.
13587 2021-09-10 Richard Biener <rguenther@suse.de>
13590 * config/dbx.h: Remove.
13591 * config/dbxcoff.h: Do not define PREFERRED_DEBUGGING_TYPE.
13592 * config/lynx.h: Likewise.
13594 2021-09-10 liuhongt <hongtao.liu@intel.com>
13596 * config/i386/i386-expand.c (ix86_expand_copysign): Expand
13597 right into ANDNOT + AND + IOR, using paradoxical subregs.
13598 (ix86_split_copysign_const): Remove.
13599 (ix86_split_copysign_var): Ditto.
13600 * config/i386/i386-protos.h (ix86_split_copysign_const): Dotto.
13601 (ix86_split_copysign_var): Ditto.
13602 * config/i386/i386.md (@copysign<mode>3_const): Ditto.
13603 (@copysign<mode>3_var): Ditto.
13605 2021-09-09 qing zhao <qing.zhao@oracle.com>
13607 * builtins.c (expand_builtin_memset): Make external visible.
13608 * builtins.h (expand_builtin_memset): Declare extern.
13609 * common.opt (ftrivial-auto-var-init=): New option.
13610 * doc/extend.texi: Document the uninitialized attribute.
13611 * doc/invoke.texi: Document -ftrivial-auto-var-init.
13612 * flag-types.h (enum auto_init_type): New enumerated type
13614 * gimple-fold.c (clear_padding_type): Add one new parameter.
13615 (clear_padding_union): Likewise.
13616 (clear_padding_emit_loop): Likewise.
13617 (clear_type_padding_in_mask): Likewise.
13618 (gimple_fold_builtin_clear_padding): Handle this new parameter.
13619 * gimplify.c (gimple_add_init_for_auto_var): New function.
13620 (gimple_add_padding_init_for_auto_var): New function.
13621 (is_var_need_auto_init): New function.
13622 (gimplify_decl_expr): Add initialization to automatic variables per
13624 (gimplify_call_expr): Add one new parameter for call to
13625 __builtin_clear_padding.
13626 (gimplify_init_constructor): Add padding initialization in the end.
13627 * internal-fn.c (INIT_PATTERN_VALUE): New macro.
13628 (expand_DEFERRED_INIT): New function.
13629 * internal-fn.def (DEFERRED_INIT): New internal function.
13630 * tree-cfg.c (verify_gimple_call): Verify calls to .DEFERRED_INIT.
13631 * tree-sra.c (generate_subtree_deferred_init): New function.
13632 (scan_function): Avoid setting cannot_scalarize_away_bitmap for
13633 calls to .DEFERRED_INIT.
13634 (sra_modify_deferred_init): New function.
13635 (sra_modify_function_body): Handle calls to DEFERRED_INIT specially.
13636 * tree-ssa-structalias.c (find_func_aliases_for_call): Likewise.
13637 * tree-ssa-uninit.c (warn_uninit): Handle calls to DEFERRED_INIT
13639 (check_defs): Likewise.
13640 (warn_uninitialized_vars): Likewise.
13641 * tree-ssa.c (ssa_undefined_value_p): Likewise.
13642 * tree.c (build_common_builtin_nodes): Build tree node for
13643 BUILT_IN_CLEAR_PADDING when needed.
13645 2021-09-09 Richard Biener <rguenther@suse.de>
13647 * tree-ssa-loop-im.c (fill_always_executed_in_1): Walk
13650 2021-09-09 Richard Biener <rguenther@suse.de>
13652 * tree-ssa-loop-im.c (fill_always_executed_in_1): Integrate
13653 DOM walk from get_loop_body_in_dom_order using a worklist
13656 2021-09-09 liuhongt <hongtao.liu@intel.com>
13658 * config.gcc: Add avx512fp16vlintrin.h.
13659 * config/i386/avx512fp16intrin.h: (_mm512_add_ph): New intrinsic.
13660 (_mm512_mask_add_ph): Likewise.
13661 (_mm512_maskz_add_ph): Likewise.
13662 (_mm512_sub_ph): Likewise.
13663 (_mm512_mask_sub_ph): Likewise.
13664 (_mm512_maskz_sub_ph): Likewise.
13665 (_mm512_mul_ph): Likewise.
13666 (_mm512_mask_mul_ph): Likewise.
13667 (_mm512_maskz_mul_ph): Likewise.
13668 (_mm512_div_ph): Likewise.
13669 (_mm512_mask_div_ph): Likewise.
13670 (_mm512_maskz_div_ph): Likewise.
13671 (_mm512_add_round_ph): Likewise.
13672 (_mm512_mask_add_round_ph): Likewise.
13673 (_mm512_maskz_add_round_ph): Likewise.
13674 (_mm512_sub_round_ph): Likewise.
13675 (_mm512_mask_sub_round_ph): Likewise.
13676 (_mm512_maskz_sub_round_ph): Likewise.
13677 (_mm512_mul_round_ph): Likewise.
13678 (_mm512_mask_mul_round_ph): Likewise.
13679 (_mm512_maskz_mul_round_ph): Likewise.
13680 (_mm512_div_round_ph): Likewise.
13681 (_mm512_mask_div_round_ph): Likewise.
13682 (_mm512_maskz_div_round_ph): Likewise.
13683 * config/i386/avx512fp16vlintrin.h: New header.
13684 * config/i386/i386-builtin-types.def (V16HF, V8HF, V32HF):
13685 Add new builtin types.
13686 * config/i386/i386-builtin.def: Add corresponding builtins.
13687 * config/i386/i386-expand.c
13688 (ix86_expand_args_builtin): Handle new builtin types.
13689 (ix86_expand_round_builtin): Likewise.
13690 * config/i386/immintrin.h: Include avx512fp16vlintrin.h
13691 * config/i386/sse.md (VFH): New mode_iterator.
13693 (avx512fmaskmode): Add HF vector modes.
13694 (avx512fmaskhalfmode): Likewise.
13695 (<plusminus_insn><mode>3<mask_name><round_name>): Adjust to for
13697 (*<plusminus_insn><mode>3<mask_name><round_name>): Likewise.
13698 (mul<mode>3<mask_name><round_name>): Likewise.
13699 (*mul<mode>3<mask_name><round_name>): Likewise.
13700 (div<mode>3): Likewise.
13701 (<sse>_div<mode>3<mask_name><round_name>): Likewise.
13702 * config/i386/subst.md (SUBST_V): Add HF vector modes.
13703 (SUBST_A): Likewise.
13704 (round_mode512bit_condition): Adjust for V32HFmode.
13706 2021-09-09 liuhongt <hongtao.liu@intel.com>
13709 * config/i386/sse.md (reduc_plus_scal_<mode>): Split to ..
13710 (reduc_plus_scal_v4sf): .. this, New define_expand.
13711 (reduc_plus_scal_v2df): .. and this, New define_expand.
13713 2021-09-09 liuhongt <hongtao.liu@intel.com>
13716 * config/i386/sse.md (*vec_extract<mode><ssescalarmodelower>_valign):
13719 2021-09-08 Jonathan Wakely <jwakely@redhat.com>
13722 * doc/trouble.texi (Copy Assignment): Fix description of
13723 behaviour and fix code in example.
13725 2021-09-08 Segher Boessenkool <segher@kernel.crashing.org>
13728 * config/rs6000/rs6000-logue.c (rs6000_emit_epilogue): For ELFv2 use
13729 r11 instead of r12 for restoring CR.
13731 2021-09-08 Jakub Jelinek <jakub@redhat.com>
13732 liuhongt <hongtao.liu@intel.com>
13735 * config/i386/i386.md (@xorsign<mode>3_1): Remove.
13736 * config/i386/i386-expand.c (ix86_expand_xorsign): Expand right away
13737 into AND with mask and XOR, using paradoxical subregs.
13738 (ix86_split_xorsign): Remove.
13739 * config/i386/i386-protos.h (ix86_split_xorsign): Remove.
13741 2021-09-08 Di Zhao <dizhao@os.amperecomputing.com>
13743 * tree-ssa-sccvn.c (vn_nary_op_insert_into): fix result compare
13745 2021-09-08 Jakub Jelinek <jakub@redhat.com>
13748 * config/i386/i386.md (xorsign<mode>3): If operands[1] is equal to
13749 operands[2], emit abs<mode>2 instead.
13750 (@xorsign<mode>3_1): Add early-clobbers for output operand, enable
13751 first alternative even for avx, add another alternative with
13752 =&Yv <- 0, Yv, Yvm constraints.
13753 * config/i386/i386-expand.c (ix86_split_xorsign): If op0 is equal
13754 to op1, emit vpandn instead.
13756 2021-09-08 liuhongt <hongtao.liu@intel.com>
13758 * config/i386/avx512fp16intrin.h (_mm_set_ph): New intrinsic.
13759 (_mm256_set_ph): Likewise.
13760 (_mm512_set_ph): Likewise.
13761 (_mm_setr_ph): Likewise.
13762 (_mm256_setr_ph): Likewise.
13763 (_mm512_setr_ph): Likewise.
13764 (_mm_set1_ph): Likewise.
13765 (_mm256_set1_ph): Likewise.
13766 (_mm512_set1_ph): Likewise.
13767 (_mm_setzero_ph): Likewise.
13768 (_mm256_setzero_ph): Likewise.
13769 (_mm512_setzero_ph): Likewise.
13770 (_mm_set_sh): Likewise.
13771 (_mm_load_sh): Likewise.
13772 (_mm_store_sh): Likewise.
13773 * config/i386/i386-builtin-types.def (V8HF): New type.
13774 (DEF_FUNCTION_TYPE (V8HF, V8HI)): New builtin function type
13775 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
13776 Support vector HFmodes.
13777 (ix86_expand_vector_init_one_nonzero): Likewise.
13778 (ix86_expand_vector_init_one_var): Likewise.
13779 (ix86_expand_vector_init_interleave): Likewise.
13780 (ix86_expand_vector_init_general): Likewise.
13781 (ix86_expand_vector_set): Likewise.
13782 (ix86_expand_vector_extract): Likewise.
13783 (ix86_expand_vector_init_concat): Likewise.
13784 (ix86_expand_sse_movcc): Handle vector HFmodes.
13785 (ix86_expand_vector_set_var): Ditto.
13786 * config/i386/i386-modes.def: Add HF vector modes in comment.
13787 * config/i386/i386.c (classify_argument): Add HF vector modes.
13788 (ix86_hard_regno_mode_ok): Allow HF vector modes for AVX512FP16.
13789 (ix86_vector_mode_supported_p): Likewise.
13790 (ix86_set_reg_reg_cost): Handle vector HFmode.
13791 (ix86_get_ssemov): Handle vector HFmode.
13792 (function_arg_advance_64): Pass unamed V16HFmode and V32HFmode
13794 (function_arg_advance_32): Pass V8HF/V16HF/V32HF by sse reg for 32bit
13796 (function_arg_advance_32): Ditto.
13797 * config/i386/i386.h (VALID_AVX512FP16_REG_MODE): New.
13798 (VALID_AVX256_REG_OR_OI_MODE): Rename to ..
13799 (VALID_AVX256_REG_OR_OI_VHF_MODE): .. this, and add V16HF.
13800 (VALID_SSE2_REG_VHF_MODE): New.
13801 (VALID_AVX512VL_128_REG_MODE): Add V8HF and TImode.
13802 (SSE_REG_MODE_P): Add vector HFmode.
13803 * config/i386/i386.md (mode): Add HF vector modes.
13804 (MODE_SIZE): Likewise.
13805 (ssemodesuffix): Add ph suffix for HF vector modes.
13806 * config/i386/sse.md (VFH_128): New mode iterator.
13807 (VMOVE): Adjust for HF vector modes.
13809 (V_256_512): Likewise.
13810 (avx512): Likewise.
13811 (avx512fmaskmode): Likewise.
13812 (shuffletype): Likewise.
13813 (sseinsnmode): Likewise.
13814 (ssedoublevecmode): Likewise.
13815 (ssehalfvecmode): Likewise.
13816 (ssehalfvecmodelower): Likewise.
13817 (ssePScmode): Likewise.
13818 (ssescalarmode): Likewise.
13819 (ssescalarmodelower): Likewise.
13820 (sseintprefix): Likewise.
13822 (bcstscalarsuff): Likewise.
13823 (xtg_mode): Likewise.
13824 (VI12HF_AVX512VL): New mode_iterator.
13825 (VF_AVX512FP16): Likewise.
13827 (VIHF_256): Likewise.
13828 (VIHF_AVX512BW): Likewise.
13829 (V16_256): Likewise.
13830 (V32_512): Likewise.
13831 (sseintmodesuffix): New mode_attr.
13832 (sse): Add scalar and vector HFmodes.
13833 (ssescalarmode): Add vector HFmode mapping.
13834 (ssescalarmodesuffix): Add sh suffix for HFmode.
13835 (*<sse>_vm<insn><mode>3): Use VFH_128.
13836 (*<sse>_vm<multdiv_mnemonic><mode>3): Likewise.
13837 (*ieee_<ieee_maxmin><mode>3): Likewise.
13838 (<avx512>_blendm<mode>): New define_insn.
13839 (vec_setv8hf): New define_expand.
13840 (vec_set<mode>_0): New define_insn for HF vector set.
13841 (*avx512fp16_movsh): Likewise.
13842 (avx512fp16_movsh): Likewise.
13843 (vec_extract_lo_v32hi): Rename to ...
13844 (vec_extract_lo_<mode>): ... this, and adjust to allow HF
13846 (vec_extract_hi_v32hi): Likewise.
13847 (vec_extract_hi_<mode>): Likewise.
13848 (vec_extract_lo_v16hi): Likewise.
13849 (vec_extract_lo_<mode>): Likewise.
13850 (vec_extract_hi_v16hi): Likewise.
13851 (vec_extract_hi_<mode>): Likewise.
13852 (vec_set_hi_v16hi): Likewise.
13853 (vec_set_hi_<mode>): Likewise.
13854 (vec_set_lo_v16hi): Likewise.
13855 (vec_set_lo_<mode>): Likewise.
13856 (*vec_extract<mode>_0): New define_insn_and_split for HF
13858 (*vec_extracthf): New define_insn.
13859 (VEC_EXTRACT_MODE): Add HF vector modes.
13860 (PINSR_MODE): Add V8HF.
13861 (sse2p4_1): Likewise.
13862 (pinsr_evex_isa): Likewise.
13863 (<sse2p4_1>_pinsr<ssemodesuffix>): Adjust to support
13864 insert for V8HFmode.
13865 (pbroadcast_evex_isa): Add HF vector modes.
13866 (AVX2_VEC_DUP_MODE): Likewise.
13867 (VEC_INIT_MODE): Likewise.
13868 (VEC_INIT_HALF_MODE): Likewise.
13869 (avx2_pbroadcast<mode>): Adjust to support HF vector mode
13871 (avx2_pbroadcast<mode>_1): Likewise.
13872 (<avx512>_vec_dup<mode>_1): Likewise.
13873 (<avx512>_vec_dup<mode><mask_name>): Likewise.
13874 (<mask_codefor><avx512>_vec_dup_gpr<mode><mask_name>):
13877 2021-09-08 Guo, Xuepeng <xuepeng.guo@intel.com>
13878 H.J. Lu <hongjiu.lu@intel.com>
13879 Liu Hongtao <hongtao.liu@intel.com>
13880 Wang Hongyu <hongyu.wang@intel.com>
13881 Xu Dianhong <dianhong.xu@intel.com>
13883 * common/config/i386/cpuinfo.h (get_available_features):
13884 Detect FEATURE_AVX512FP16.
13885 * common/config/i386/i386-common.c
13886 (OPTION_MASK_ISA_AVX512FP16_SET,
13887 OPTION_MASK_ISA_AVX512FP16_UNSET,
13888 OPTION_MASK_ISA2_AVX512FP16_SET,
13889 OPTION_MASK_ISA2_AVX512FP16_UNSET): New.
13890 (OPTION_MASK_ISA2_AVX512BW_UNSET,
13891 OPTION_MASK_ISA2_AVX512BF16_UNSET): Add AVX512FP16.
13892 (ix86_handle_option): Handle -mavx512fp16.
13893 * common/config/i386/i386-cpuinfo.h (enum processor_features):
13894 Add FEATURE_AVX512FP16.
13895 * common/config/i386/i386-isas.h: Add entry for AVX512FP16.
13896 * config.gcc: Add avx512fp16intrin.h.
13897 * config/i386/avx512fp16intrin.h: New intrinsic header.
13898 * config/i386/cpuid.h: Add bit_AVX512FP16.
13899 * config/i386/i386-builtin-types.def: (FLOAT16): New primitive type.
13900 * config/i386/i386-builtins.c: Support _Float16 type for i386
13902 (ix86_register_float16_builtin_type): New function.
13903 (ix86_float16_type_node): New.
13904 * config/i386/i386-c.c (ix86_target_macros_internal): Define
13906 * config/i386/i386-expand.c (ix86_expand_branch): Support
13908 (ix86_prepare_fp_compare_args): Adjust TARGET_SSE_MATH &&
13909 SSE_FLOAT_MODE_P to SSE_FLOAT_MODE_SSEMATH_OR_HF_P.
13910 (ix86_expand_fp_movcc): Ditto.
13911 * config/i386/i386-isa.def: Add PTA define for AVX512FP16.
13912 * config/i386/i386-options.c (isa2_opts): Add -mavx512fp16.
13913 (ix86_valid_target_attribute_inner_p): Add avx512fp16 attribute.
13914 * config/i386/i386.c (ix86_get_ssemov): Use
13915 vmovdqu16/vmovw/vmovsh for HFmode/HImode scalar or vector.
13916 (ix86_get_excess_precision): Use
13917 FLT_EVAL_METHOD_PROMOTE_TO_FLOAT16 when TARGET_AVX512FP16
13919 (sse_store_index): Use SFmode cost for HFmode cost.
13920 (inline_memory_move_cost): Add HFmode, and perfer SSE cost over
13921 GPR cost for HFmode.
13922 (ix86_hard_regno_mode_ok): Allow HImode in sse register.
13923 (ix86_mangle_type): Add manlging for _Float16 type.
13924 (inline_secondary_memory_needed): No memory is needed for
13925 16bit movement between gpr and sse reg under
13927 (ix86_multiplication_cost): Adjust TARGET_SSE_MATH &&
13928 SSE_FLOAT_MODE_P to SSE_FLOAT_MODE_SSEMATH_OR_HF_P.
13929 (ix86_division_cost): Ditto.
13930 (ix86_rtx_costs): Ditto.
13931 (ix86_add_stmt_cost): Ditto.
13932 (ix86_optab_supported_p): Ditto.
13933 * config/i386/i386.h (VALID_AVX512F_SCALAR_MODE): Add HFmode.
13934 (SSE_FLOAT_MODE_SSEMATH_OR_HF_P): Add HFmode.
13935 (PTA_SAPPHIRERAPIDS): Add PTA_AVX512FP16.
13936 * config/i386/i386.md (mode): Add HFmode.
13937 (MODE_SIZE): Add HFmode.
13938 (isa): Add avx512fp16.
13939 (enabled): Handle avx512fp16.
13940 (ssemodesuffix): Add sh suffix for HFmode.
13941 (comm): Add mult, div.
13942 (plusminusmultdiv): New code iterator.
13943 (insn): Add mult, div.
13944 (*movhf_internal): Adjust for avx512fp16 instruction.
13945 (*movhi_internal): Ditto.
13946 (*cmpi<unord>hf): New define_insn for HFmode.
13947 (*ieee_s<ieee_maxmin>hf3): Likewise.
13948 (extendhf<mode>2): Likewise.
13949 (trunc<mode>hf2): Likewise.
13950 (float<floatunssuffix><mode>hf2): Likewise.
13951 (*<insn>hf): Likewise.
13952 (cbranchhf4): New expander.
13953 (movhfcc): Likewise.
13954 (<insn>hf3): Likewise.
13955 (mulhf3): Likewise.
13956 (divhf3): Likewise.
13957 * config/i386/i386.opt: Add mavx512fp16.
13958 * config/i386/immintrin.h: Include avx512fp16intrin.h.
13959 * doc/invoke.texi: Add mavx512fp16.
13960 * doc/extend.texi: Add avx512fp16 Usage Notes.
13962 2021-09-08 liuhongt <hongtao.liu@intel.com>
13964 * common.opt: Support -fexcess-precision=16.
13965 * config/aarch64/aarch64.c (aarch64_excess_precision): Return
13966 FLT_EVAL_METHOD_PROMOTE_TO_FLOAT16 when
13967 EXCESS_PRECISION_TYPE_FLOAT16.
13968 * config/arm/arm.c (arm_excess_precision): Ditto.
13969 * config/i386/i386.c (ix86_get_excess_precision): Ditto.
13970 * config/m68k/m68k.c (m68k_excess_precision): Issue an error
13971 when EXCESS_PRECISION_TYPE_FLOAT16.
13972 * config/s390/s390.c (s390_excess_precision): Ditto.
13973 * coretypes.h (enum excess_precision_type): Add
13974 EXCESS_PRECISION_TYPE_FLOAT16.
13975 * doc/tm.texi (TARGET_C_EXCESS_PRECISION): Update documents.
13976 * doc/tm.texi.in (TARGET_C_EXCESS_PRECISION): Ditto.
13977 * doc/extend.texi (Half-Precision): Document
13978 -fexcess-precision=16.
13979 * flag-types.h (enum excess_precision): Add
13980 EXCESS_PRECISION_FLOAT16.
13981 * target.def (excess_precision): Update document.
13982 * tree.c (excess_precision_type): Set excess_precision_type to
13983 EXCESS_PRECISION_FLOAT16 when -fexcess-precision=16.
13985 2021-09-08 liuhongt <hongtao.liu@intel.com>
13987 * doc/extend.texi: (@node Floating Types): Adjust the wording.
13988 (@node Half-Precision): Ditto.
13990 2021-09-07 Takayuki 'January June' Suwa <jjsuwa_sys3175@yahoo.co.jp>
13993 * config/xtensa/xtensa.c (xtensa_emit_move_sequence): Add
13994 'CONST_INT_P (src)' to the condition of the block that tries to
13995 eliminate literal when loading integer contant.
13997 2021-09-07 David Faust <david.faust@oracle.com>
13999 * doc/extend.texi (BPF Type Attributes) New node.
14000 Document new preserve_access_index attribute.
14001 Document new preserve_access_index builtin.
14002 * doc/invoke.texi: Document -mco-re and -mno-co-re options.
14004 2021-09-07 David Faust <david.faust@oracle.com>
14006 * config/bpf/bpf.c: Adjust includes.
14007 (bpf_handle_preserve_access_index_attribute): New function.
14008 (bpf_attribute_table): Use it here.
14009 (bpf_builtins): Add BPF_BUILTIN_PRESERVE_ACCESS_INDEX.
14010 (bpf_option_override): Handle "-mco-re" option.
14011 (bpf_asm_init_sections): New.
14012 (TARGET_ASM_INIT_SECTIONS): Redefine.
14013 (bpf_file_end): New.
14014 (TARGET_ASM_FILE_END): Redefine.
14015 (bpf_init_builtins): Add "__builtin_preserve_access_index".
14016 (bpf_core_compute, bpf_core_get_index): New.
14017 (is_attr_preserve_access): New.
14018 (bpf_expand_builtin): Handle new builtins.
14019 (bpf_core_newdecl, bpf_core_is_maybe_aggregate_access): New.
14020 (bpf_core_walk): New.
14021 (bpf_resolve_overloaded_builtin): New.
14022 (TARGET_RESOLVE_OVERLOADED_BUILTIN): Redefine.
14023 (handle_attr): New.
14024 (pass_bpf_core_attr): New RTL pass.
14025 * config/bpf/bpf-passes.def: New file.
14026 * config/bpf/bpf-protos.h (make_pass_bpf_core_attr): New.
14027 * config/bpf/coreout.c: New file.
14028 * config/bpf/coreout.h: Likewise.
14029 * config/bpf/t-bpf (TM_H): Add $(srcdir)/config/bpf/coreout.h.
14030 (coreout.o): New rule.
14031 (PASSES_EXTRA): Add $(srcdir)/config/bpf/bpf-passes.def.
14032 * config.gcc (bpf): Add coreout.h to extra_headers.
14033 Add coreout.o to extra_objs.
14034 Add $(srcdir)/config/bpf/coreout.c to target_gtfiles.
14036 2021-09-07 David Faust <david.faust@oracle.com>
14038 * btfout.c (get_btf_id): Function is no longer static.
14039 * ctfc.h: Expose it here.
14041 2021-09-07 David Faust <david.faust@oracle.com>
14043 * ctfc.c (ctf_lookup_tree_type): New function.
14044 * ctfc.h: Likewise.
14046 2021-09-07 David Faust <david.faust@oracle.com>
14048 * ctfc.c (ctf_dtd_lookup): Function is no longer static.
14049 * ctfc.h: Analogous change.
14051 2021-09-07 David Faust <david.faust@oracle.com>
14053 * dwarf2out.c (lookup_type_die): Function is no longer static.
14054 * dwarf2out.h: Expose it here.
14056 2021-09-07 Indu Bhagat <indu.bhagat@oracle.com>
14058 * dwarf2ctf.c (ctf_debug_finalize): Make it static.
14059 (ctf_debug_early_finish): New definition.
14060 (ctf_debug_finish): Likewise.
14061 * dwarf2ctf.h (ctf_debug_finalize): Remove declaration.
14062 (ctf_debug_early_finish): New declaration.
14063 (ctf_debug_finish): Likewise.
14064 * dwarf2out.c (dwarf2out_finish): Invoke ctf_debug_finish.
14065 (dwarf2out_early_finish): Invoke ctf_debug_early_finish.
14067 2021-09-07 Indu Bhagat <indu.bhagat@oracle.com>
14069 * config/bpf/bpf.c (bpf_option_override): For BPF backend, disable LTO
14070 support when compiling for CO-RE.
14071 * config/bpf/bpf.opt: Add new command line option -mco-re.
14073 2021-09-07 Indu Bhagat <indu.bhagat@oracle.com>
14075 * flag-types.h (enum debug_info_type): Add new enum
14076 DINFO_TYPE_BTF_WITH_CORE.
14077 (BTF_WITH_CORE_DEBUG): New bitmask.
14078 * flags.h (btf_with_core_debuginfo_p): New declaration.
14079 * opts.c (btf_with_core_debuginfo_p): New definition.
14081 2021-09-07 Jason Merrill <jason@redhat.com>
14083 * tree.h (error_operand_p): Change to inline function.
14085 2021-09-07 Aldy Hernandez <aldyh@redhat.com>
14087 * tree-ssa-threadedge.c (forwarder_block_p): Rename to...
14088 (empty_block_with_phis_p): ...this.
14089 (potentially_threadable_block): Same.
14090 (jump_threader::thread_through_normal_block): Same.
14092 2021-09-07 Eric Botcazou <ebotcazou@adacore.com>
14095 * dwarf2out.c (mark_base_types): New overloaded function.
14096 (dwarf2out_early_finish): Invoke it on the COMDAT type list as well
14097 as the compilation unit, and call move_marked_base_types afterward.
14099 2021-09-07 H.J. Lu <hjl.tools@gmail.com>
14102 * config/i386/i386-expand.c (ix86_expand_convert_uns_sisf_sse):
14104 (ix86_expand_vector_convert_uns_vsivsf): Likewise.
14106 2021-09-07 Richard Biener <rguenther@suse.de>
14108 PR tree-optimization/102226
14109 * tree-vect-loop.c (vect_transform_cycle_phi): Record
14110 the converted value for the epilogue PHI use.
14112 2021-09-07 Martin Liska <mliska@suse.cz>
14114 PR gcov-profile/80223
14115 * ipa-inline.c (can_inline_edge_p): Similarly to sanitizer
14116 options, do not inline when no_profile_instrument_function
14117 attributes are different in early inliner. It's fine to inline
14118 it after PGO instrumentation.
14120 2021-09-07 Richard Biener <rguenther@suse.de>
14122 PR tree-optimization/101555
14123 * tree-ssa-pre.c (translate_vuse_through_block): Do not
14124 perform an alias walk to determine the validity of the
14125 mem at the start of the block which is already guaranteed
14126 by means of prune_clobbered_mems.
14127 (phi_translate_1): Pass edge to translate_vuse_through_block.
14129 2021-09-07 Xionghu Luo <luoxhu@linux.ibm.com>
14132 * config/rs6000/rs6000.md (fmod<mode>3): New define_expand.
14133 (remainder<mode>3): Likewise.
14135 2021-09-07 YunQiang Su <yunqiang.su@cipunited.com>
14137 * config/mips/mips.c (mips_file_start): add .module for
14140 2021-09-06 Roger Sayle <roger@nextmovesoftware.com>
14142 * wide-int.cc (wi::clz): Reorder tests to ensure the result
14143 is zero for all negative values.
14145 2021-09-06 Tobias Burnus <tobias@codesourcery.com>
14147 * doc/invoke.texi (-foffload-options): Fix @opindex.
14149 2021-09-06 H.J. Lu <hjl.tools@gmail.com>
14152 * config/i386/i386-expand.c (ix86_split_xorsign): Use operands[2].
14153 * config/i386/i386.md (@xorsign<mode>3_1): Add non-destructive
14154 source alternative for AVX.
14156 2021-09-06 liuhongt <hongtao.liu@intel.com>
14158 PR middle-end/102182
14159 * optabs.c (expand_fix): Add from1 to avoid from being
14162 2021-09-06 Eric Botcazou <ebotcazou@adacore.com>
14164 * dwarf2out.c (modified_type_die): Deal with all array types earlier
14165 and use local variable consistently throughout the function.
14167 2021-09-06 Jakub Jelinek <jakub@redhat.com>
14169 PR tree-optimization/102207
14170 * match.pd: Don't demote operands of IFN_{ADD,SUB,MUL}_OVERFLOW if they
14171 were promoted from signed to wider unsigned type.
14173 2021-09-06 Andrew Pinski <apinski@marvell.com>
14175 PR tree-optimization/63184
14176 * match.pd: Add simplification of pointer_diff of two pointer_plus
14177 with addr_expr in the first operand of each pointer_plus.
14178 Add simplificatoin of ne/eq of two pointer_plus with addr_expr
14179 in the first operand of each pointer_plus.
14181 2021-09-06 Richard Biener <rguenther@suse.de>
14183 PR tree-optimization/102176
14184 * tree-vect-slp.c (vect_slp_gather_vectorized_scalar_stmts):
14186 (vect_bb_slp_scalar_cost): Use the computed set of
14187 vectorized scalar stmts instead of relying on the out-of-date
14188 and not accurate PURE_SLP_STMT.
14189 (vect_bb_vectorization_profitable_p): Compute the set
14190 of vectorized scalar stmts.
14192 2021-09-05 Aldy Hernandez <aldyh@redhat.com>
14194 * gimple-range-path.cc (path_range_query::range_of_stmt): Remove
14195 GIMPLE_COND special casing.
14196 (path_range_query::range_defined_in_block): Use range_of_stmt
14197 instead of calling fold_range directly.
14199 2021-09-05 Aldy Hernandez <aldyh@redhat.com>
14201 * gimple-range-path.cc (path_range_query::range_of_expr): Set
14202 m_undefined_path when appropriate.
14203 (path_range_query::internal_range_of_expr): Copy from range_of_expr.
14204 (path_range_query::unreachable_path_p): New.
14205 (path_range_query::precompute_ranges): Set m_undefined_path.
14206 * gimple-range-path.h (path_range_query::unreachable_path_p): New.
14207 (path_range_query::internal_range_of_expr): New.
14208 * tree-ssa-threadbackward.c (back_threader::find_taken_edge_cond):
14209 Use unreachable_path_p.
14211 2021-09-05 Aldy Hernandez <aldyh@redhat.com>
14213 * tree-ssa-threadbackward.c (back_threader::maybe_register_path):
14214 Remove argument and call find_taken_edge.
14215 (back_threader::resolve_phi): Do not calculate taken edge before
14216 calling maybe_register_path.
14217 (back_threader::find_paths_to_names): Same.
14219 2021-09-05 Jeff Law <jlaw@localhost.localdomain>
14221 * config/h8300/h8300.md (QHSI2 mode iterator): New mode iterator.
14222 * config/h8300/testcompare.md (store_c): Update name, use new
14224 (store_neg_c, store_shifted_c): New patterns.
14226 2021-09-03 Segher Boessenkool <segher@kernel.crashing.org>
14229 * config/rs6000/rs6000-logue.c (rs6000_emit_prologue): On ELFv2 use r11
14230 instead of r12 for CR save, in all cases.
14232 2021-09-03 Andrew Pinski <apinski@marvell.com>
14234 * config/aarch64/aarch64-sve-builtins.cc (register_vector_type):
14235 Handle error_mark_node as the type of the type_decl.
14237 2021-09-03 Andrew Pinski <apinski@marvell.com>
14239 * config/aarch64/aarch64-builtins.c (struct aarch64_simd_type_info):
14241 (aarch64_simd_types): Likewise.
14242 (aarch64_simd_intOI_type_node): Likewise.
14243 (aarch64_simd_intCI_type_node): Likewise.
14244 (aarch64_simd_intXI_type_node): Likewise.
14245 * config/aarch64/aarch64.h (aarch64_fp16_type_node): Likewise.
14246 (aarch64_fp16_ptr_type_node): Likewise.
14247 (aarch64_bf16_type_node): Likewise.
14248 (aarch64_bf16_ptr_type_node): Likewise.
14250 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
14252 * range-op.cc (operator_minus::op1_op2_relation_effect): Abstract
14254 (minus_op1_op2_relation_effect): ...here.
14255 (class operator_pointer_diff): New.
14256 (operator_pointer_diff::op1_op2_relation_effect): Call
14257 minus_op1_op2_relation_effect.
14258 (integral_table::integral_table): Add entry for POINTER_DIFF_EXPR.
14260 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
14262 * tree-ssa-threadbackward.c (back_threader::thread_through_all_blocks):
14263 Add may_peel_loop_headers.
14264 (back_threader_registry::thread_through_all_blocks): Same.
14265 (try_thread_blocks): Pass may_peel_loop_headers argument.
14266 (pass_early_thread_jumps::execute): Same.
14268 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
14270 * tree-ssa-threadedge.c (has_phis_p): New.
14271 (forwarder_block_p): New.
14272 (potentially_threadable_block): Call forwarder_block_p.
14273 (jump_threader::thread_around_empty_blocks): Call has_phis_p.
14274 (jump_threader::thread_through_normal_block): Call
14277 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
14279 * tree-ssa-threadbackward.c (back_threader::dump): New.
14280 (back_threader::debug): New.
14281 (back_threader_profitability::profitable_path_p): Dump blocks
14282 even if we are bailing early.
14284 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
14286 * tree-ssa-threadupdate.c (cancel_thread): New.
14287 (jump_thread_path_registry::thread_block_1): Use cancel_thread.
14288 (jump_thread_path_registry::mark_threaded_blocks): Same.
14289 (jump_thread_path_registry::register_jump_thread): Same.
14291 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
14293 * tree-ssa-threadedge.c (jt_state::push): Only call methods for
14294 which objects are available.
14295 (jt_state::pop): Same.
14296 (jt_state::register_equiv): Same.
14297 (jt_state::register_equivs_on_edge): Same.
14299 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
14301 * tree-ssa-threadedge.c (jump_threader::thread_across_edge):
14302 Move pop until after a thread is registered.
14304 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
14306 * tree-ssa-threadupdate.c (debug): New.
14308 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
14310 * gimple-range-trace.cc (push_dump_file::push_dump_file): New.
14311 (push_dump_file::~push_dump_file): New.
14312 (dump_ranger): Change dump_file temporarily while dumping
14314 * gimple-range-trace.h (class push_dump_file): New.
14316 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
14318 * gimple-range-trace.cc (debug_seed_ranger): Remove static.
14319 (dump_ranger): Dump function name.
14321 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
14323 * gimple-range-path.cc (path_range_query::range_defined_in_block):
14324 Adjust for non-null.
14325 (path_range_query::adjust_for_non_null_uses): New.
14326 (path_range_query::precompute_ranges): Call
14327 adjust_for_non_null_uses.
14328 * gimple-range-path.h: Add m_non_null and
14329 adjust_for_non_null_uses.
14331 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
14333 * gimple-range-path.cc (path_range_query::dump): Dump path
14335 (path_range_query::precompute_ranges): Dump entire path.
14337 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
14339 * value-relation.cc (relation_oracle::debug): New.
14340 * value-relation.h (relation_oracle::debug): New.
14342 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
14344 * tree-ssa-loop-ch.c: Remove unnecessary include file.
14346 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
14348 * gimple-range-fold.cc (fold_using_range::postfold_gcond_edges):
14349 Skip statements with no defining BB.
14350 * gimple-range-path.cc (path_range_query::range_defined_in_block):
14351 Do not get confused by statements with no defining BB.
14353 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
14355 * gimple-range-fold.cc (adjust_imagpart_expr): Move from
14356 gimple_range_adjustment. Add support for constants.
14357 (adjust_realpart_expr): New.
14358 (gimple_range_adjustment): Move IMAGPART_EXPR code to
14359 adjust_imagpart_expr.
14360 * range-op.cc (integral_table::integral_table): Add entry for
14363 2021-09-03 Jakub Jelinek <jakub@redhat.com>
14365 * omp-expand.c (expand_omp_atomic_pipeline): Use
14366 IFN_ATOMIC_COMPARE_EXCHANGE instead of
14367 BUILT_IN_SYNC_VAL_COMPARE_AND_SWAP_? so that memory order
14370 2021-09-03 Jakub Jelinek <jakub@redhat.com>
14373 * tree.h (DECL_FIELD_ABI_IGNORED): Changed into rvalue only macro
14374 that is false if DECL_BIT_FIELD.
14375 (SET_DECL_FIELD_ABI_IGNORED, DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD,
14376 SET_DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD): Define.
14377 * tree-streamer-out.c (pack_ts_decl_common_value_fields): For
14378 DECL_BIT_FIELD stream DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD instead
14379 of DECL_FIELD_ABI_IGNORED.
14380 * tree-streamer-in.c (unpack_ts_decl_common_value_fields): Use
14381 SET_DECL_FIELD_ABI_IGNORED instead of writing to
14382 DECL_FIELD_ABI_IGNORED and for DECL_BIT_FIELD use
14383 SET_DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD instead.
14384 * lto-streamer-out.c (hash_tree): For DECL_BIT_FIELD hash
14385 DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD instead of DECL_FIELD_ABI_IGNORED.
14387 2021-09-03 liuhongt <hongtao.liu@intel.com>
14390 * config/i386/amxbf16intrin.h : Remove macro check for __AMX_BF16__.
14391 * config/i386/amxint8intrin.h : Remove macro check for __AMX_INT8__.
14392 * config/i386/amxtileintrin.h : Remove macro check for __AMX_TILE__.
14394 2021-09-02 Martin Sebor <msebor@redhat.com>
14396 PR tree-optimization/17506
14398 * tree-ssa-uninit.c (warn_uninit): Remove conditional guarding note.
14400 2021-09-02 Richard Biener <rguenther@suse.de>
14402 * tree-ssa-loop-im.c (fill_always_executed_in_1): Refine
14403 fix for PR78185 and continue processing when leaving
14404 finite inner loops.
14406 2021-09-02 Jakub Jelinek <jakub@redhat.com>
14408 PR tree-optimization/99591
14409 * match.pd: Demote operands of IFN_{ADD,SUB,MUL}_OVERFLOW if they
14412 2021-09-02 Richard Biener <rguenther@suse.de>
14415 2021-09-02 Richard Biener <rguenther@suse.de>
14417 PR tree-optimization/102155
14418 * tree-ssa-loop-im.c (fill_always_executed_in_1): Iterate
14419 over a part of the RPO array and do not recurse here.
14420 Dump blocks marked as always executed.
14421 (fill_always_executed_in): Walk over the RPO array and
14422 process loops whose header we run into.
14423 (loop_invariant_motion_in_fun): Compute the first RPO
14424 using rev_post_order_and_mark_dfs_back_seme in iteration
14425 order and pass that to fill_always_executed_in.
14427 2021-09-02 liuhongt <hongtao.liu@intel.com>
14429 * config/i386/i386-modes.def (FLOAT_MODE): Define ieee HFmode.
14430 * config/i386/i386.c (enum x86_64_reg_class): Add
14431 X86_64_SSEHF_CLASS.
14432 (merge_classes): Handle X86_64_SSEHF_CLASS.
14433 (examine_argument): Ditto.
14434 (construct_container): Ditto.
14435 (classify_argument): Ditto, and set HFmode/HCmode to
14436 X86_64_SSEHF_CLASS.
14437 (function_value_32): Return _FLoat16/Complex Float16 by
14439 (function_value_64): Return _Float16/Complex Float16 by SSE
14441 (ix86_print_operand): Handle CONST_DOUBLE HFmode.
14442 (ix86_secondary_reload): Require gpr as intermediate register
14443 to store _Float16 from sse register when sse4 is not
14445 (ix86_libgcc_floating_mode_supported_p): Enable _FLoat16 under
14447 (ix86_scalar_mode_supported_p): Ditto.
14448 (TARGET_LIBGCC_FLOATING_MODE_SUPPORTED_P): Defined.
14449 * config/i386/i386.h (VALID_SSE2_REG_MODE): Add HFmode.
14450 (VALID_INT_MODE_P): Add HFmode and HCmode.
14451 * config/i386/i386.md (*pushhf_rex64): New define_insn.
14453 (*movhf_internal): Ditto.
14454 * doc/extend.texi (Half-Precision Floating Point): Documemt
14457 2021-09-02 Richard Biener <rguenther@suse.de>
14459 PR tree-optimization/102155
14460 * tree-ssa-loop-im.c (fill_always_executed_in_1): Iterate
14461 over a part of the RPO array and do not recurse here.
14462 Dump blocks marked as always executed.
14463 (fill_always_executed_in): Walk over the RPO array and
14464 process loops whose header we run into.
14465 (loop_invariant_motion_in_fun): Compute the first RPO
14466 using rev_post_order_and_mark_dfs_back_seme in iteration
14467 order and pass that to fill_always_executed_in.
14469 2021-09-02 YunQiang Su <syq@debian.org>
14472 2021-08-31 YunQiang Su <yunqiang.su@cipunited.com>
14474 * config/mips/mips.c (mips_module_isa_name): New.
14475 mips_file_start: add .module mipsREV to all asm output
14477 2021-09-01 Jeff Law <jlaw@localhost.localdomain>
14479 PR tree-optimization/102152
14480 * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Reduce a vector
14481 comparison to a scalar comparison before calling
14482 update_stmt_if_modified.
14484 2021-09-01 Andrew Pinski <apinski@marvell.com>
14487 * config/aarch64/aarch64.c (aarch64_expand_setmem):
14488 Check STRICT_ALIGNMENT before creating an overlapping
14491 2021-09-01 Martin Sebor <msebor@redhat.com>
14493 * gimple-ssa-warn-access.cc (get_size_range): Add argument.
14494 (check_access): Pass additional argument.
14495 (check_memop_access): Remove template and make a member function.
14496 (maybe_check_dealloc_call): Make a pass_waccess member function.
14497 (class pass_waccess): Add, rename, and remove members.
14498 (pass_waccess::pass_waccess): Adjust to name change.
14499 (pass_waccess::~pass_waccess): Same.
14500 (check_alloca): Make a member function.
14501 (check_alloc_size_call): Same.
14502 (check_strcat): Same.
14503 (check_strncat): Same.
14504 (check_stxcpy): Same.
14505 (check_stxncpy): Same.
14506 (check_strncmp): Same.
14507 (maybe_warn_rdwr_sizes): Rename...
14508 (pass_waccess::maybe_check_access_sizes): ...to this.
14509 (pass_waccess::check_call): Adjust to name changes.
14510 (pass_waccess::maybe_check_dealloc_call): Make a pass_waccess member
14512 (pass_waccess::execute): Adjust to name changes.
14513 * gimple-ssa-warn-access.h (check_memop_access): Remove.
14514 * pointer-query.cc (access_ref::phi): Handle null pointer.
14515 (access_ref::inform_access): Same.
14516 (pointer_query::put_ref): Modify a cached value, not a copy of it.
14517 (pointer_query::dump): New function.
14518 (compute_objsize_r): Avoid overwriting access_ref::bndrng. Cache
14520 * pointer-query.h (pointer_query::dump): Declare.
14521 * tree-ssa-strlen.c (get_range): Simplify. Use function query.
14522 (dump_strlen_info): Use function query.
14523 (printf_strlen_execute): Factor code out into pointer_query::put_ref.
14525 2021-09-01 Thomas Schwinge <thomas@codesourcery.com>
14527 * tree.c (walk_tree_1) <OMP_CLAUSE>: Simplify.
14529 2021-09-01 Iain Sandoe <iain@sandoe.co.uk>
14531 * doc/extend.texi: Document unavailable attribute.
14532 * print-tree.c (print_node): Handle unavailable attribute.
14533 * tree-core.h (struct tree_base): Add a bit to carry unavailability.
14534 * tree.c (error_unavailable_use): New.
14535 * tree.h (TREE_UNAVAILABLE): New.
14536 (error_unavailable_use): New.
14538 2021-09-01 Jakub Jelinek <jakub@redhat.com>
14540 PR tree-optimization/102124
14541 * tree-vect-patterns.c (vect_recog_widen_op_pattern): For ORIG_CODE
14542 MINUS_EXPR, if itype is unsigned with smaller precision than type,
14543 add an extra cast to signed variant of itype to ensure sign-extension.
14545 2021-09-01 Martin Liska <mliska@suse.cz>
14547 * graph.c (draw_cfg_node_succ_edges): Do not color fallthru
14548 edges and rather use colors for TRUE and FALSE edges.
14550 2021-09-01 Richard Biener <rguenther@suse.de>
14552 PR tree-optimization/93491
14553 * tree-ssa-pre.c (compute_avail): Set BB_MAY_NOTRETURN
14554 after processing the stmt itself. Do not consider
14555 pure functions possibly not returning. Properly avoid
14556 adding possibly trapping calls to EXP_GEN when there's
14557 a preceeding possibly not returning call.
14558 * tree-ssa-sccvn.c (vn_reference_may_trap): Conservatively
14561 2021-09-01 Richard Biener <rguenther@suse.de>
14563 PR tree-optimization/102139
14564 * tree-vectorizer.h (vec_base_alignments): Adjust hash-map
14565 type to record a std::pair of the stmt-info and the innermost
14567 (dr_vec_info::group): New member.
14568 * tree-vect-data-refs.c (vect_record_base_alignment): Adjust.
14569 (vect_compute_data_ref_alignment): Verify the recorded
14570 base alignment can be used.
14571 (data_ref_pair): Remove.
14572 (dr_group_sort_cmp): Adjust.
14573 (vect_analyze_data_ref_accesses): Store the group-ID in the
14574 dr_vec_info and operate on a vector of dr_vec_infos.
14576 2021-09-01 YunQiang Su <yunqiang.su@cipunited.com>
14578 * read-md.c (md_reader::handle_enum): support value assignation.
14579 * doc/md.texi: record define_c_enum value assignation support.
14581 2021-09-01 Jakub Jelinek <jakub@redhat.com>
14583 PR tree-optimization/102141
14584 * gimple-ssa-store-merging.c (bswap_view_convert): Add BEFORE
14585 argument. If false, emit stmts after gsi instead of before, and
14587 (bswap_replace): Adjust callers. When converting output of bswap,
14588 emit VIEW_CONVERT prepratation stmts after a copy of gsi instead
14591 2021-09-01 liuhongt <hongtao.liu@intel.com>
14593 * emit-rtl.c (validate_subreg): Get rid of all float-int
14596 2021-09-01 liuhongt <hongtao.liu@intel.com>
14599 2021-08-30 liuhongt <hongtao.liu@intel.com>
14601 * expmed.c (extract_bit_field_1): Make sure we're playing with
14602 integral modes before call extract_integral_bit_field.
14603 (extract_integral_bit_field): Add a parameter of type
14604 scalar_int_mode which corresponds to of tmode.
14605 And call extract_and_convert_fixed_bit_field instead of
14606 extract_fixed_bit_field and convert_extracted_bit_field.
14607 (extract_and_convert_fixed_bit_field): New function, it's a
14608 combination of extract_fixed_bit_field and
14609 convert_extracted_bit_field.
14611 2021-08-31 Thomas Schwinge <thomas@codesourcery.com>
14613 * tree.c (walk_tree_1) <OMP_CLAUSE_TILE>: Handle three operands.
14615 2021-08-31 Thomas Schwinge <thomas@codesourcery.com>
14617 * omp-general.h (omp_is_reference): Rename to...
14618 (omp_privatize_by_reference): ... this. Adjust all users...
14619 * omp-general.c: ... here, ...
14620 * gimplify.c: ... here, ...
14621 * omp-expand.c: ... here, ...
14622 * omp-low.c: ... here.
14624 2021-08-31 Martin Sebor <msebor@redhat.com>
14626 * gimple-ssa-warn-access.cc (maybe_warn_alloc_args_overflow): Test
14627 pointer element for equality to zero, not that of the cotaining
14630 2021-08-31 Martin Sebor <msebor@redhat.com>
14632 * gcc-rich-location.h (gcc_rich_location): Make ctor explicit.
14634 2021-08-31 Martin Sebor <msebor@redhat.com>
14636 * function.h (function): Add comments.
14637 (get_range_query): Same. Add attribute returns nonnull.
14639 2021-08-31 Roger Sayle <roger@nextmovesoftware.com>
14641 * expr.c (convert_modes): Don't use subreg_promoted_mode on a
14642 SUBREG if it can't be guaranteed to a SUBREG_PROMOTED_VAR_P set.
14643 Instead use the standard (safer) is_a <scalar_int_mode> idiom.
14645 2021-08-31 Jeff Law <jlaw@localhost.localdomain>
14647 * config.gcc (cris-*-elf, cris-*-none): Remove dbxelf.h from
14649 (m32r-*-elf, m32rle-*-elf, m32r-*-linux): Likewise.
14650 (mn10300-*-*, am33_2.0-*-linux*): Likewise.
14651 (xtensa*-*-elf, xtensa*-*-linux, xtensa*-*-uclinux): Likewise.
14652 (m32c-*-elf*, m32c-*-rtems*): Likewise.
14653 * config/cris/cris.h (DBX_NO_XREFS): Remove.
14654 (DBX_CONTIN_LENGTH, DBX_CONTIN_CHAR): Likewise.
14655 * config/m32r/m32r.h (DBXOUT_SOURCE_LINE): Likewise.
14656 (DBX_DEBUGGING_INFO, DBX_CONTIN_LENGTH): Likewise.
14657 * config/mn10300/mn10300.h (DEFAULT_GDB_EXTENSIONS): Likewise.
14658 * config/mn10300/linux.h (DBX_REGISTER_NAMES): Likewise.
14660 2021-08-31 Marcel Vollweiler <marcel@codesourcery.com>
14662 * gimplify.c (gimplify_scan_omp_clauses): Error handling. 'ancestor' only
14663 allowed on target constructs and only with particular other clauses.
14664 * omp-expand.c (expand_omp_target): Output of 'sorry, not supported' if
14665 'ancestor' is used.
14666 * omp-low.c (check_omp_nesting_restrictions): Error handling. No nested OpenMP
14667 structs when 'ancestor' is used.
14668 (scan_omp_1_stmt): No usage of OpenMP runtime routines in a target region when
14669 'ancestor' is used.
14670 * tree-pretty-print.c (dump_omp_clause): Append 'ancestor'.
14671 * tree.h (OMP_CLAUSE_DEVICE_ANCESTOR): Define macro.
14673 2021-08-31 Roger Sayle <roger@nextmovesoftware.com>
14675 * expr.c (convert_modes): Preserve SUBREG_PROMOTED_VAR_P when
14676 creating a (wider) partial subreg from a SUBREG_PROMOTED_VAR_P
14678 * simplify-rtx.c (simplify_unary_operation_1) [SIGN_EXTEND]:
14679 Likewise, preserve SUBREG_PROMOTED_VAR_P when creating a (wider)
14680 partial subreg from a SUBREG_PROMOTED_VAR_P subreg. Generate
14681 SIGN_EXTEND of the SUBREG_REG when a subreg would be paradoxical.
14682 [ZERO_EXTEND]: Likewise, preserve SUBREG_PROMOTED_VAR_P when
14683 creating a (wider) partial subreg from a SUBREG_PROMOTED_VAR_P
14684 subreg. Generate ZERO_EXTEND of the SUBREG_REG when a subreg
14685 would be paradoxical.
14687 2021-08-31 Roger Sayle <roger@nextmovesoftware.com>
14689 * combine.c (combine_simplify_rtx): Avoid converting an explicit
14690 TRUNCATE into a lowpart SUBREG on !TRULY_NOOP_TRUNCATION targets.
14691 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
14693 2021-08-31 Richard Biener <rguenther@suse.de>
14695 PR tree-optimization/102142
14696 * tree-vect-slp.c (vect_bb_vectorization_profitable_p): Fix
14697 condition under which to unset the visited flag.
14699 2021-08-31 Richard Biener <rguenther@suse.de>
14701 PR middle-end/102129
14702 * tree-ssa-ter.c (find_replaceable_in_bb): Do not move
14703 possibly trapping expressions across calls.
14705 2021-08-31 Jakub Jelinek <jakub@redhat.com>
14707 PR tree-optimization/102134
14708 * tree-ssa-ccp.c (bit_value_binop) <case RSHIFT_EXPR>: If sgn is
14709 UNSIGNED and r1val | r1mask has MSB set, ensure lzcount doesn't
14712 2021-08-31 Andrew Pinski <apinski@marvell.com>
14715 * collect-utils.c (setup_signals): New declaration.
14716 * collect-utils.h (setup_signals): New function.
14717 * collect2.c (handler): Delete.
14718 (main): Instead of manually setting up the signals,
14719 just call setup_signals.
14720 * lto-wrapper.c (main): Likewise.
14722 2021-08-31 Andrew Pinski <apinski@marvell.com>
14725 * config/i386/i386-protos.h (x86_output_aligned_bss):
14726 Change align argument to unsigned type.
14727 (x86_elf_aligned_decl_common): Likewise.
14728 * config/i386/i386.c (x86_elf_aligned_decl_common): Likewise.
14729 (x86_output_aligned_bss): Likewise.
14731 2021-08-31 YunQiang Su <yunqiang.su@cipunited.com>
14733 * config/mips/mips.c (mips_module_isa_name): New.
14734 mips_file_start: add .module mipsREV to all asm output
14736 2021-08-31 YunQiang Su <yunqiang.su@cipunited.com>
14738 * config/mips/mips.h (struct mips_cpu_info): define enum mips_isa;
14739 use enum instead of int for 'isa' member.
14740 * config.gcc, config/mips/mips.c, config/mips/mips-cpus.def,
14741 config/mips/netbsd.h: replace hardcoded numbers with enum.
14743 2021-08-31 liuhongt <hongtao.liu@intel.com>
14745 * config/i386/sse.md (*<avx512>_ucmp<mode>3_1): Change from
14746 define_split to define_insn_and_split.
14747 (*avx2_eq<mode>3): Removed.
14748 (<avx512>_eq<mode>3<mask_scalar_merge_name>): Adjust pattern
14749 (<avx512>_eq<mode>3<mask_scalar_merge_name>_1): Rename to ..
14750 (*<avx512>_eq<mode>3<mask_scalar_merge_name>_1): .. this, and
14752 (*avx2_gt<mode>3): Removed.
14753 (<avx512>_gt<mode>3<mask_scalar_merge_name>): Change from
14754 define_insn to define_expand, and adjust pattern.
14755 (UNSPEC_MASKED_EQ, UNSPEC_MASKED_GT): Removed.
14757 2021-08-30 David Malcolm <dmalcolm@redhat.com>
14760 * Makefile.in (ANALYZER_OBJS): Add analyzer/call-info.o.
14762 2021-08-30 Jason Merrill <jason@redhat.com>
14764 * doc/invoke.texi: Document -Wmissing-requires.
14766 2021-08-30 Bill Schmidt <wschmidt@linux.ibm.com>
14768 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Remove
14769 TARGET_EXTRA_BUILTINS guard.
14771 2021-08-30 Bill Schmidt <wschmidt@linux.ibm.com>
14773 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Change
14774 initialization of V2DI_type_node and unsigned_V2DI_type_node.
14776 2021-08-30 Bill Schmidt <wschmidt@linux.ibm.com>
14778 * config/rs6000/darwin.h (SUBTARGET_INIT_BUILTINS): Use the new
14779 decl when new_builtins_are_live.
14780 * config/rs6000/rs6000-builtin-new.def (__builtin_cfstring): New
14783 2021-08-30 Pat Haugen <pthaugen@linux.ibm.com>
14785 * config/rs6000/rs6000-cpus.def (ISA_3_1_MASKS_SERVER): Add
14786 OPTION_MASK_P10_FUSION_2STORE.
14787 (POWERPC_MASKS): Likewise.
14788 * config/rs6000/rs6000.c (rs6000_option_override_internal): Enable
14789 store fusion for Power10.
14790 (is_fusable_store): New.
14791 (power10_sched_reorder): Likewise.
14792 (rs6000_sched_reorder): Do Power10 specific reordering.
14793 (rs6000_sched_reorder2): Likewise.
14794 * config/rs6000/rs6000.opt: Add new option.
14796 2021-08-30 Richard Biener <rguenther@suse.de>
14798 PR tree-optimization/102128
14799 * tree-vect-slp.c (vect_bb_vectorization_profitable_p):
14800 Move scanning for if-converted scalar code to the caller
14801 and instead delay clearing the visited flag for profitable
14803 (vect_slp_region): Cost all subgraphs before scheduling.
14804 For if-converted BB vectorization scan for scalar COND_EXPRs
14805 and do not vectorize if any found and the cost model is
14808 2021-08-30 Richard Biener <rguenther@suse.de>
14810 * common.opt (fexceptions): Mark
14811 EnabledBy(fnon-call-exceptions).
14812 * doc/invoke.texi (fnon-call-exceptions): Document this
14813 enables -fexceptions.
14815 2021-08-30 Sebastian Huber <sebastian.huber@embedded-brains.de>
14817 * tsystem.h (abort): Define abort() if inhibit_libc is defined and it
14818 is not already defined.
14820 2021-08-30 liuhongt <hongtao.liu@intel.com>
14822 * expmed.c (extract_bit_field_1): Make sure we're playing with
14823 integral modes before call extract_integral_bit_field.
14824 (extract_integral_bit_field): Add a parameter of type
14825 scalar_int_mode which corresponds to of tmode.
14826 And call extract_and_convert_fixed_bit_field instead of
14827 extract_fixed_bit_field and convert_extracted_bit_field.
14828 (extract_and_convert_fixed_bit_field): New function, it's a
14829 combination of extract_fixed_bit_field and
14830 convert_extracted_bit_field.
14832 2021-08-29 Iain Sandoe <iain@sandoe.co.uk>
14834 * config/darwin.c (darwin_libc_has_function): Do not run
14835 the checks for x86 or modern Darwin. Make sure that there
14836 is a value set for darwin_macosx_version_min before testing.
14838 2021-08-29 Iain Sandoe <iain@sandoe.co.uk>
14840 * config/i386/darwin.h (CLEAR_INSN_CACHE): New.
14842 2021-08-28 Jan Hubicka <hubicka@ucw.cz>
14844 * ipa-modref-tree.h (modref_access_node::merge): Break out
14845 logic combining offsets and logic merging ranges to ...
14846 (modref_access_node::combined_offsets): ... here
14847 (modref_access_node::update2): ... here
14848 (modref_access_node::closer_pair_p): New member function.
14849 (modref_access_node::forced_merge): New member function.
14850 (modre_ref_node::insert): Do merging when table is full.
14852 2021-08-28 YunQiang Su <yunqiang.su@cipunited.com>
14855 * config.gcc: MIPS: use N64 ABI by default if the triple end
14856 with -gnuabi64, which is used by Debian since 2013.
14858 2021-08-28 Alexandre Oliva <oliva@adacore.com>
14860 * ipa-modref.c (analyze_function): Skip debug stmts.
14861 * tree-inline.c (estimate_num_insn): Consider builtins even
14862 without a cgraph_node.
14864 2021-08-27 Jeff Law <jlaw@localhost.localdomain>
14866 * config/h8300/bitfield.md (cstore<mode>4): Remove expander.
14867 * config/h8300/h8300.c (h8300_expand_branch): Remove function.
14868 * config/h8300/h8300-protos.h (h8300_expadn_branch): Remove prototype.
14869 * config/h8300/h8300.md (eqne): New code iterator.
14870 (geultu, geultu_to_c): Similarly.
14871 * config/h8300/testcompare.md (cstore<mode>4): Dummy expander.
14872 (store_c_<mode>, store_c_i_<mode>): New define_insn_and_splits
14873 (cmp<mode>_c): New pattern
14875 2021-08-27 Jeff Law <jlaw@localhost.localdomain>
14877 * tree-ssa-dom.c (reduce_vector_comparison_to_scalar_comparison): New
14879 (dom_opt_dom_walker::optimize_stmt): Use it.
14881 2021-08-27 Iain Sandoe <iain@sandoe.co.uk>
14883 * config/darwin.c (finalize_ctors): Add a section-start linker-
14885 (finalize_dtors): Likewise.
14886 * config/darwin.h (MIN_LD64_INIT_TERM_START_LABELS): New.
14888 2021-08-27 Bill Schmidt <wschmidt@linux.ibm.com>
14890 * config/rs6000/rs6000-call.c (rs6000-builtins.h): New #include.
14891 (rs6000_init_builtins): Call rs6000_init_generated_builtins. Skip the
14892 old initialization logic when new builtins are enabled.
14893 * config/rs6000/rs6000-gen-builtins.c (write_decls): Rename
14894 rs6000_autoinit_builtins to rs6000_init_generated_builtins.
14895 (write_init_file): Likewise.
14897 2021-08-27 Iain Sandoe <iain@sandoe.co.uk>
14899 * configure.ac (darwin2[[0-9]]* | darwin19*): Alter use of
14900 gcc_GAS_CHECK_FEATURE to remove an extraneous parameter.
14901 (amdgcn-* | gcn-*) Likewise.
14903 2021-08-27 Anthony Sharp <anthonysharp15@gmail.com>
14905 * symbol-summary.h: Added missing template keyword.
14907 2021-08-27 Richard Biener <rguenther@suse.de>
14909 PR tree-optimization/45178
14910 * tree-ssa-dce.c (find_obviously_necessary_stmts): For
14911 infinite loops without exit do not mark control dependent
14912 edges of the latch necessary.
14914 2021-08-27 konglin1 <lingling.kong@intel.com>
14917 * config/i386/sse.md: (<avx512>scattersi<mode>): Add mask operand to
14919 (<avx512>scattersi<mode>): Likewise.
14920 (*avx512f_scattersi<VI48F:mode>): Merge mask operand to set_dest.
14921 (*avx512f_scatterdi<VI48F:mode>): Likewise
14923 2021-08-27 Kewen Lin <linkw@linux.ibm.com>
14925 * config/rs6000/rs6000.c (rs6000_builtin_md_vectorized_function): Add
14926 support for built-in functions MISC_BUILTIN_DIVWE, MISC_BUILTIN_DIVWEU,
14927 MISC_BUILTIN_DIVDE, MISC_BUILTIN_DIVDEU, P10_BUILTIN_CFUGED,
14928 P10_BUILTIN_CNTLZDM, P10_BUILTIN_CNTTZDM, P10_BUILTIN_PDEPD and
14929 P10_BUILTIN_PEXTD on Power10.
14931 2021-08-27 Kewen Lin <linkw@linux.ibm.com>
14933 * config/rs6000/rs6000-call.c (builtin_function_type): Add unsigned
14934 signedness for some Power10 bifs.
14936 2021-08-27 David Edelsohn <dje.gcc@gmail.com>
14939 * config/rs6000/rs6000.c (rs6000_adjust_field_align): Use
14940 computed alignment if the entire struct has attribute packed.
14942 2021-08-27 liuhongt <hongtao.liu@intel.com>
14946 * config/i386/i386.c (ix86_gimple_fold_builtin): Fold
14947 IX86_BUILTIN_SHUFPD512, IX86_BUILTIN_SHUFPS512,
14948 IX86_BUILTIN_SHUFPD256, IX86_BUILTIN_SHUFPS,
14949 IX86_BUILTIN_SHUFPS256.
14950 (ix86_masked_all_ones): New function.
14952 2021-08-26 Uroš Bizjak <ubizjak@gmail.com>
14954 * config/i386/i386.md (*btr<mode>_1): Call force_reg unconditionally.
14955 (conditional moves with memory inputs splitters): Ditto.
14956 * config/i386/sse.md (one_cmpl<mode>2): Simplify.
14958 2021-08-26 Jan Hubicka <hubicka@ucw.cz>
14960 * ipa-modref-tree.h (modref_access_node::try_merge_with): Restart
14961 search after merging.
14963 2021-08-26 Bill Schmidt <wschmidt@linux.ibm.com>
14965 * config/rs6000/rs6000-overload.def: Add remaining overloads.
14967 2021-08-26 Bill Schmidt <wschmidt@linux.ibm.com>
14969 * config/rs6000/rs6000-builtin-new.def: Add cell stanza.
14971 2021-08-26 Bill Schmidt <wschmidt@linux.ibm.com>
14973 * config/rs6000/rs6000-builtin-new.def: Add ieee128-hw, dfp,
14974 crypto, and htm stanzas.
14976 2021-08-26 Bill Schmidt <wschmidt@linux.ibm.com>
14978 * config/rs6000/rs6000-builtin-new.def: Add mma stanza.
14980 2021-08-26 Martin Sebor <msebor@redhat.com>
14982 * tree-ssa-uninit.c (warn_uninit): Refactor and simplify.
14983 (warn_uninit_phi_uses): Remove argument from calls to warn_uninit.
14984 (warn_uninitialized_vars): Same. Reduce visibility of locals.
14985 (warn_uninitialized_phi): Same.
14987 2021-08-26 Roger Sayle <roger@nextmovesoftware.com>
14989 * tree-ssa-ccp.c (get_individual_bits): Helper function to
14990 extract the individual bits from a widest_int constant (mask).
14991 (gray_code_bit_flips): New read-only table for effiently
14992 enumerating permutations/combinations of bits.
14993 (bit_value_binop) [LROTATE_EXPR, RROTATE_EXPR]: Handle rotates
14994 by unknown counts that are guaranteed less than the target
14995 precision and four or fewer unknown bits by enumeration.
14996 [LSHIFT_EXPR, RSHIFT_EXPR]: Likewise, also handle shifts by
14997 enumeration under the same conditions. Handle remaining
14998 shifts as a mask based upon the minimum possible shift value.
15000 2021-08-26 Roger Sayle <roger@nextmovesoftware.com>
15001 Richard Biener <rguenther@suse.de>
15003 * match.pd (shift transformations): Remove a redundant
15004 !POINTER_TYPE_P check.
15006 2021-08-26 Uroš Bizjak <ubizjak@gmail.com>
15009 * config/i386/i386.md (cmove reg-reg move elimination peephole2s):
15010 Set all_regs to true in the call to replace_rtx.
15012 2021-08-26 Jan Hubicka <hubicka@ucw.cz>
15014 * ipa-modref-tree.c (test_insert_search_collapse): Update test.
15015 * ipa-modref-tree.h (modref_base_node::insert): Be smarter when
15016 hiting --param modref-max-refs limit.
15017 (modref_tree:insert_base): Be smarter when hitting
15018 --param modref-max-bases limit. Add new parameter REF.
15019 (modref_tree:insert): Update.
15020 (modref_tree:merge): Update.
15021 * ipa-modref.c (read_modref_records): Update.
15023 2021-08-26 Jan Hubicka <hubicka@ucw.cz>
15025 * params.opt: (modref-max-adjustments): Add full stop.
15027 2021-08-26 Jan Hubicka <hubicka@ucw.cz>
15029 * ipa-modref-tree.h (modref_ref_node::verify): New member
15031 (modref_ref_node::insert): Use it.
15032 (modref_ref_node::try_mere_with): Fix off by one error.
15034 2021-08-26 Martin Liska <mliska@suse.cz>
15035 Stefan Kneifel <stefan.kneifel@bluewin.ch>
15037 * cgraph.h (create_version_clone_with_body): Add new parameter.
15038 * cgraphclones.c: Likewise.
15039 * multiple_target.c (create_dispatcher_calls): Do not use
15041 (create_target_clone): Likewise here.
15043 2021-08-26 Jonathan Yong <10walls@gmail.com>
15045 * doc/extend.texi: Add note about reserved priorities
15046 to the constructor attribute.
15048 2021-08-25 Martin Sebor <msebor@redhat.com>
15050 * gimple-range-cache.cc (ssa_global_cache::dump): Avoid printing
15051 range table header alone.
15052 * gimple-range.cc (gimple_ranger::export_global_ranges): Same.
15054 2021-08-25 Jan Hubicka <hubicka@ucw.cz>
15056 * doc/invoke.texi: Document --param modref-max-adjustments.
15057 * ipa-modref-tree.c (test_insert_search_collapse): Update.
15058 (test_merge): Update.
15059 * ipa-modref-tree.h (struct modref_access_node): Add adjustments;
15060 (modref_access_node::operator==): Fix handling of access ranges.
15061 (modref_access_node::contains): Constify parameter; handle also
15062 mismatched parm offsets.
15063 (modref_access_node::update): New function.
15064 (modref_access_node::merge): New function.
15065 (unspecified_modref_access_node): Update constructor.
15066 (modref_ref_node::insert_access): Add record_adjustments parameter;
15068 (modref_ref_node::try_merge_with): New private function.
15069 (modref_tree::insert): New record_adjustments parameter.
15070 (modref_tree::merge): New record_adjustments parameter.
15071 (modref_tree::copy_from): Update.
15072 * ipa-modref.c (dump_access): Dump adjustments field.
15073 (get_access): Update constructor.
15074 (record_access): Update call of insert.
15075 (record_access_lto): Update call of insert.
15076 (merge_call_side_effects): Add record_adjustments parameter.
15077 (get_access_for_fnspec): Update.
15078 (process_fnspec): Update.
15079 (analyze_call): Update.
15080 (analyze_function): Update.
15081 (read_modref_records): Update.
15082 (ipa_merge_modref_summary_after_inlining): Update.
15083 (propagate_unknown_call): Update.
15084 (modref_propagate_in_scc): Update.
15085 * params.opt (param-max-modref-adjustments=): New.
15087 2021-08-25 Michael Meissner <meissner@linux.ibm.com>
15089 * config/rs6000/vsx.md (UNSPEC_XXSPLTIDP): Rename from
15091 (xxspltiw_v4si): Use vecperm type attribute.
15092 (xxspltiw_v4si_inst): Use vecperm type attribute.
15093 (xxspltiw_v4sf_inst): Likewise.
15094 (xxspltidp_v2df): Use vecperm type attribute. Use
15095 UNSPEC_XXSPLTIDP instead of UNSPEC_XXSPLTID.
15096 (xxspltidp_v2df_inst): Likewise.
15097 (xxsplti32dx_v4si): Use vecperm type attribute.
15098 (xxsplti32dx_v4si_inst): Likewise.
15099 (xxsplti32dx_v4sf_inst): Likewise.
15100 (xxblend_<mode>): Likewise.
15101 (xxpermx): Likewise.
15102 (xxpermx_inst): Likewise.
15103 (xxeval): Likewise.
15105 2021-08-25 Lewis Hyatt <lhyatt@gmail.com>
15108 * coretypes.h (typedef diagnostic_input_charset_callback): Declare.
15109 * diagnostic.c (diagnostic_initialize_input_context): New function.
15110 * diagnostic.h (diagnostic_initialize_input_context): Declare.
15111 * input.c (default_charset_callback): New function.
15112 (file_cache::initialize_input_context): New function.
15113 (file_cache_slot::create): Added ability to convert the input
15114 according to the input context.
15115 (file_cache::file_cache): Initialize the new input context.
15116 (class file_cache_slot): Added new m_alloc_offset member.
15117 (file_cache_slot::file_cache_slot): Initialize the new member.
15118 (file_cache_slot::~file_cache_slot): Handle potentially offset buffer.
15119 (file_cache_slot::maybe_grow): Likewise.
15120 (file_cache_slot::needs_read_p): Handle NULL fp, which is now possible.
15121 (file_cache_slot::get_next_line): Likewise.
15122 * input.h (class file_cache): Added input context member.
15124 2021-08-25 Richard Biener <rguenther@suse.de>
15126 PR tree-optimization/102046
15127 * tree-vect-slp.c (vect_build_slp_tree_2): Conservatively
15128 update ->any_pattern when swapping operands.
15130 2021-08-25 Hongyu Wang <hongyu.wang@intel.com>
15133 * config/i386/i386.c (ix86_live_on_entry): Adjust comment.
15134 (ix86_decompose_address): Remove retval check for ASHIFT,
15135 allow non-canonical zero extend if AND mask covers ASHIFT
15137 (ix86_legitimate_address_p): Adjust condition for decompose.
15138 (ix86_rtx_costs): Adjust cost for lea with non-canonical
15140 Co-Authored by: Uros Bizjak <ubizjak@gmail.com>
15142 2021-08-25 Jiufu Guo <guojiufu@linux.ibm.com>
15144 PR tree-optimization/101145
15145 * tree-ssa-loop-niter.c (number_of_iterations_until_wrap):
15147 (number_of_iterations_lt): Invoke above function.
15148 (adjust_cond_for_loop_until_wrap):
15149 Merge to number_of_iterations_until_wrap.
15150 (number_of_iterations_cond): Update invokes for
15151 adjust_cond_for_loop_until_wrap and number_of_iterations_lt.
15153 2021-08-25 konglin1 <lingling.kong@intel.com>
15156 * config/i386/avx512dqintrin.h (_mm512_fpclass_ps_mask): Fix
15157 macro define in O0.
15158 (_mm512_mask_fpclass_ps_mask): Ditto.
15160 2021-08-25 Kewen Lin <linkw@linux.ibm.com>
15162 * config/rs6000/altivec.md (vec_unpacku_hi_v16qi): Remove.
15163 (vec_unpacku_hi_v8hi): Likewise.
15164 (vec_unpacku_lo_v16qi): Likewise.
15165 (vec_unpacku_lo_v8hi): Likewise.
15166 (vec_unpacku_hi_<VP_small_lc>): New define_expand.
15167 (vec_unpacku_lo_<VP_small_lc>): Likewise.
15169 2021-08-24 David Edelsohn <dje.gcc@gmail.com>
15171 * config/rs6000/aix.h (SYSTEM_IMPLICIT_EXTERN_C): Delete.
15172 * config/rs6000/aix71.h (SYSTEM_IMPLICIT_EXTERN_C): Define.
15173 * config/rs6000/aix72.h (SYSTEM_IMPLICIT_EXTERN_C): Define.
15174 * config/rs6000/aix73.h (TARGET_AIX_VERSION): Increase to 73.
15176 2021-08-24 Roger Sayle <roger@nextmovesoftware.com>
15178 PR middle-end/102031
15179 * simplify-rtx.c (simplify_truncation): When comparing precisions
15180 use "subreg_prec" variable, not "subreg_mode".
15182 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
15184 * config/rs6000/rs6000-builtin-new.def: Add power10 and power10-64
15187 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
15189 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Initialize
15190 various pointer type nodes.
15191 * config/rs6000/rs6000.h (rs6000_builtin_type_index): Add enum
15192 values for various pointer types.
15193 (ptr_V16QI_type_node): New macro.
15194 (ptr_V1TI_type_node): New macro.
15195 (ptr_V2DI_type_node): New macro.
15196 (ptr_V2DF_type_node): New macro.
15197 (ptr_V4SI_type_node): New macro.
15198 (ptr_V4SF_type_node): New macro.
15199 (ptr_V8HI_type_node): New macro.
15200 (ptr_unsigned_V16QI_type_node): New macro.
15201 (ptr_unsigned_V1TI_type_node): New macro.
15202 (ptr_unsigned_V8HI_type_node): New macro.
15203 (ptr_unsigned_V4SI_type_node): New macro.
15204 (ptr_unsigned_V2DI_type_node): New macro.
15205 (ptr_bool_V16QI_type_node): New macro.
15206 (ptr_bool_V8HI_type_node): New macro.
15207 (ptr_bool_V4SI_type_node): New macro.
15208 (ptr_bool_V2DI_type_node): New macro.
15209 (ptr_bool_V1TI_type_node): New macro.
15210 (ptr_pixel_type_node): New macro.
15211 (ptr_intQI_type_node): New macro.
15212 (ptr_uintQI_type_node): New macro.
15213 (ptr_intHI_type_node): New macro.
15214 (ptr_uintHI_type_node): New macro.
15215 (ptr_intSI_type_node): New macro.
15216 (ptr_uintSI_type_node): New macro.
15217 (ptr_intDI_type_node): New macro.
15218 (ptr_uintDI_type_node): New macro.
15219 (ptr_intTI_type_node): New macro.
15220 (ptr_uintTI_type_node): New macro.
15221 (ptr_long_integer_type_node): New macro.
15222 (ptr_long_unsigned_type_node): New macro.
15223 (ptr_float_type_node): New macro.
15224 (ptr_double_type_node): New macro.
15225 (ptr_long_double_type_node): New macro.
15226 (ptr_dfloat64_type_node): New macro.
15227 (ptr_dfloat128_type_node): New macro.
15228 (ptr_ieee128_type_node): New macro.
15229 (ptr_ibm128_type_node): New macro.
15230 (ptr_vector_pair_type_node): New macro.
15231 (ptr_vector_quad_type_node): New macro.
15232 (ptr_long_long_integer_type_node): New macro.
15233 (ptr_long_long_unsigned_type_node): New macro.
15235 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
15237 * config/rs6000/rs6000-builtin-new.def: Add power9-vector, power9,
15238 and power9-64 stanzas.
15240 2021-08-24 Roger Sayle <roger@nextmovesoftware.com>
15241 Tom de Vries <tdevries@suse.de>
15243 * config.gcc (nvptx-*-*): Define {c,c++}_target_objs.
15244 * config/nvptx/nvptx-protos.h (nvptx_cpu_cpp_builtins): Prototype.
15245 * config/nvptx/nvptx.h (TARGET_CPU_CPP_BUILTINS): Implement with
15246 a call to the new nvptx_cpu_cpp_builtins function in nvptx-c.c.
15247 * config/nvptx/t-nvptx (nvptx-c.o): New rule.
15248 * config/nvptx/nvptx-c.c: New source file.
15249 (nvptx_cpu_cpp_builtins): Move implementation here.
15251 2021-08-24 Martin Sebor <msebor@redhat.com>
15253 PR middle-end/101600
15254 PR middle-end/101977
15255 * gimple-ssa-warn-access.cc (maybe_warn_for_bound): Tighten up
15256 the phrasing of a warning.
15257 (check_access): Use the remaining size after subtracting any offset
15258 rather than the whole object size.
15259 * pointer-query.cc (access_ref::get_ref): Clear BASE0 flag if it's
15260 clear for any nonnull PHI argument.
15261 (compute_objsize): Clear argument.
15263 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
15265 * config/rs6000/rs6000-builtin-new.def: Add power8-vector stanza.
15267 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
15269 * config/rs6000/rs6000-builtin-new.def: Add power7 and power7-64
15272 2021-08-24 Andrew MacLeod <amacleod@redhat.com>
15274 * value-relation.cc (rr_transitive_table): New.
15275 (relation_transitive): New.
15276 (value_relation::swap): Remove.
15277 (value_relation::apply_transitive): New.
15278 (relation_oracle::relation_oracle): Allocate a new tmp bitmap.
15279 (relation_oracle::register_relation): Call register_transitives.
15280 (relation_oracle::register_transitives): New.
15281 * value-relation.h (relation_oracle): Add new temporary bitmap and
15284 2021-08-24 H.J. Lu <hjl.tools@gmail.com>
15287 * config/i386/i386-expand.c (ix86_expand_vector_move): Broadcast
15288 from integer to a pseudo vector register.
15290 2021-08-24 Richard Biener <rguenther@suse.de>
15292 PR tree-optimization/100089
15293 * tree-vectorizer.h (vect_slp_bb): Rename to ...
15294 (vect_slp_if_converted_bb): ... this and get the original
15295 loop as new argument.
15296 * tree-vectorizer.c (try_vectorize_loop_1): Revert previous fix,
15297 pass original loop to vect_slp_if_converted_bb.
15298 * tree-vect-slp.c (vect_bb_vectorization_profitable_p):
15299 If orig_loop was passed scan the not vectorized stmts
15300 for COND_EXPRs and force not profitable if found.
15301 (vect_slp_region): Pass down all SLP instances to costing
15302 if orig_loop was specified.
15303 (vect_slp_bbs): Pass through orig_loop.
15304 (vect_slp_bb): Rename to ...
15305 (vect_slp_if_converted_bb): ... this and get the original
15306 loop as new argument.
15307 (vect_slp_function): Adjust.
15309 2021-08-24 Richard Earnshaw <rearnsha@arm.com>
15312 * config/arm/arm.md (attribute arch): Add fix_vlldm.
15313 (arch_enabled): Use it.
15314 * config/arm/vfp.md (lazy_store_multiple_insn): Add alternative to
15315 use when erratum mitigation is needed.
15317 2021-08-24 Richard Earnshaw <rearnsha@arm.com>
15320 * config/arm/arm.opt (mfix-cmse-cve-2021-35465): New option.
15321 * doc/invoke.texi (Arm Options): Document it.
15322 * config/arm/arm-cpus.in (quirk_vlldm): New feature bit.
15323 (ALL_QUIRKS): Add quirk_vlldm.
15324 (cortex-m33): Add quirk_vlldm.
15325 (cortex-m35p, cortex-m55): Likewise.
15326 * config/arm/arm.c (arm_option_override): Enable fix_vlldm if
15327 targetting an affected CPU and not explicitly controlled on
15330 2021-08-24 Richard Earnshaw <rearnsha@arm.com>
15332 * config/arm/vfp.md (lazy_store_multiple_insn): Rewrite as valid RTL.
15333 (lazy_load_multiple_insn): Likewise.
15335 2021-08-24 liuhongt <hongtao.liu@intel.com>
15338 * config/i386/sse.md (<avx512>_vternlog<mode><sd_maskz_name>):
15339 Enable avx512 embedded broadcast.
15340 (*<avx512>_vternlog<mode>_all): Ditto.
15341 (<avx512>_vternlog<mode>_mask): Ditto.
15343 2021-08-24 liuhongt <hongtao.liu@intel.com>
15346 * config/i386/i386.c (ix86_rtx_costs): Define cost for
15348 * config/i386/i386.h (STRIP_UNARY): New macro.
15349 * config/i386/predicates.md (reg_or_notreg_operand): New
15351 * config/i386/sse.md (*<avx512>_vternlog<mode>_all): New define_insn.
15352 (*<avx512>_vternlog<mode>_1): New pre_reload
15353 define_insn_and_split.
15354 (*<avx512>_vternlog<mode>_2): Ditto.
15355 (*<avx512>_vternlog<mode>_3): Ditto.
15356 (any_logic1,any_logic2): New code iterator.
15357 (logic_op): New code attribute.
15358 (ternlogsuffix): Extend to VNxDF and VNxSF.
15360 2021-08-24 Richard Biener <rguenther@suse.de>
15362 * doc/invoke.texi (vect-inner-loop-cost-factor): Adjust.
15363 * params.opt (--param vect-inner-loop-cost-factor): Adjust
15365 * tree-vect-loop.c (vect_analyze_loop_form): Initialize
15366 inner_loop_cost_factor to the minimum of the estimated number
15367 of iterations of the inner loop and vect-inner-loop-cost-factor.
15369 2021-08-24 Roger Sayle <roger@nextmovesoftware.com>
15370 Richard Biener <rguenther@suse.de>
15372 * config/i386/i386-features.c (compute_convert_gain): Provide
15373 more accurate values for CONST_INT, when optimizing for size.
15374 * config/i386/i386.c (COSTS_N_BYTES): Move definition from here...
15375 * config/i386/i386.h (COSTS_N_BYTES): to here.
15377 2021-08-24 Roger Sayle <roger@nextmovesoftware.com>
15378 Jakub Jelinek <jakub@redhat.com>
15380 PR middle-end/102029
15381 * match.pd (shift transformations): Add an additional check for
15382 !POINTER_TYPE_P in the recently added left shift transformation.
15384 2021-08-24 liuhongt <hongtao.liu@intel.com>
15386 PR tree-optimization/100089
15387 * tree-vectorizer.c (try_vectorize_loop_1): Disable slp in
15388 loop vectorizer when cost model is very-cheap.
15390 2021-08-23 Bill Schmidt <wschmidt@linux.ibm.com>
15392 * config/rs6000/rs6000-gen-builtins.c (parse_bif_entry): Don't call
15393 asprintf, which is not available on AIX.
15395 2021-08-23 Bill Schmidt <wschmidt@linux.ibm.com>
15397 * config.gcc (target_gtfiles): Add ./rs6000-builtins.h.
15398 * config/rs6000/t-rs6000 (EXTRA_GTYPE_DEPS): Set.
15400 2021-08-23 Bill Schmidt <wschmidt@linux.ibm.com>
15402 * config.gcc (powerpc*-*-*): Add rs6000-builtins.o to extra_objs.
15403 * config/rs6000/rs6000-gen-builtins.c (main): Close init_file
15405 * config/rs6000/t-rs6000 (rs6000-gen-builtins.o): New target.
15406 (rbtree.o): Likewise.
15407 (rs6000-gen-builtins): Likewise.
15408 (rs6000-builtins.c): Likewise.
15409 (rs6000-builtins.h): Likewise.
15410 (rs6000.o): Add dependency.
15411 (EXTRA_HEADERS): Add rs6000-vecdefines.h.
15412 (rs6000-vecdefines.h): New target.
15413 (rs6000-builtins.o): Likewise.
15414 (rs6000-call.o): Add rs6000-builtins.h as a dependency.
15415 (rs6000-c.o): Likewise.
15417 2021-08-23 Bill Schmidt <wschmidt@linux.ibm.com>
15420 * config/rs6000/rs6000-gen-builtins.c (consume_whitespace):
15421 Diagnose buffer overrun.
15422 (safe_inc_pos): Fix overrun detection.
15423 (match_identifier): Diagnose buffer overrun.
15424 (match_integer): Likewise.
15425 (match_to_right_bracket): Likewise.
15427 2021-08-23 Jan Hubicka <hubicka@ucw.cz>
15429 * ipa-modref-tree.h (modref_access_node::range_info_useful_p):
15430 Improve range compare.
15431 (modref_access_node::contains): New member function.
15432 (modref_access_node::search): Remove.
15433 (modref_access_node::insert): Be smarter about subaccesses.
15435 2021-08-23 Thomas Schwinge <thomas@codesourcery.com>
15437 * config/i386/i386-options.c (ix86_omp_device_kind_arch_isa)
15438 <omp_device_arch> [ACCEL_COMPILER]: Match "intel_mic".
15439 * config/i386/t-omp-device (omp-device-properties-i386) <arch>:
15442 2021-08-23 Jeff Law <jlaw@localhost.localdomain>
15444 * config/h8300/h8300-protos.h (h8300_expand_epilogue): Add new
15446 * config/h8300/jumpcall.md (call, call_value): Restrict to
15447 !SIBLING_CALL_P cases.
15448 (subcall, sibcall_value): New patterns & expanders.
15449 * config/h8300/proepi.md (epilogue): Pass new argument to
15450 h8300_expand_epilogue.
15451 (sibcall_epilogue): New expander.
15452 * config/h8300/h8300.c (h8300_expand_epilogue): Handle sibcall
15454 (h8300_ok_for_sibcall_p): New function.
15455 (TARGET_FUNCTION_OK_FOR_SIBCALL): define.
15457 2021-08-23 Roger Sayle <roger@nextmovesoftware.com>
15459 * simplify-rtx.c (simplify_unary_operation_1): [TRUNCATE]:
15460 Handle case where the operand is already the desired mode.
15462 2021-08-23 Richard Biener <rguenther@suse.de>
15465 * tree-ssa-structalias.c (ipa_pta_execute): Check in_other_partition
15466 in addition to has_gimple_body.
15468 2021-08-23 Jan Hubicka <hubicka@ucw.cz>
15470 PR middle-end/101949
15471 * ipa-modref.c (analyze_ssa_name_flags): Fix merging of
15474 2021-08-23 Martin Liska <mliska@suse.cz>
15476 * doc/invoke.texi: Put the option out of -mxl-mode-app-model
15479 2021-08-23 Richard Biener <rguenther@suse.de>
15481 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
15482 Properly scale the inner loop cost only once.
15484 2021-08-23 Roger Sayle <roger@nextmovesoftware.com>
15486 * tree-ssa-ccp.c (bit_value_binop) [TRUNC_MOD_EXPR, TRUNC_DIV_EXPR]:
15487 Provide bounds for unsigned (and signed with non-negative operands)
15488 division and modulus.
15490 2021-08-23 Roger Sayle <roger@nextmovesoftware.com>
15492 * simplify-rtx.c (simplify_truncation): Generalize simplification
15493 of (truncate:A (subreg:B X)).
15494 (simplify_unary_operation_1) [FLOAT_TRUNCATE, FLOAT_EXTEND,
15495 SIGN_EXTEND, ZERO_EXTEND]: Handle cases where the operand
15496 already has the desired machine mode.
15497 (test_scalar_int_ops): Add tests that useless extensions and
15498 truncations are optimized away.
15499 (test_scalar_int_ext_ops): New self-test function to confirm
15500 that truncations of extensions are correctly simplified.
15501 (test_scalar_int_ext_ops2): New self-test function to check
15502 truncations of truncations, extensions of extensions, and
15503 truncations of extensions.
15504 (test_scalar_ops): Call the above two functions with a
15505 representative sampling of integer machine modes.
15507 2021-08-23 Roger Sayle <roger@nextmovesoftware.com>
15509 * match.pd (shift transformations): Change the sign of an
15510 LSHIFT_EXPR if it reduces the number of explicit conversions.
15512 2021-08-23 Jakub Jelinek <jakub@redhat.com>
15514 PR tree-optimization/86723
15515 * gimple-ssa-store-merging.c (find_bswap_or_nop_finalize): Add
15516 cast64_to_32 argument, set *cast64_to_32 to false, unless n is
15517 non-memory permutation of 64-bit src which only has bytes of
15518 0 or [5..8] and n->range is 4.
15519 (find_bswap_or_nop): Add cast64_to_32 and mask arguments, adjust
15520 find_bswap_or_nop_finalize caller, support bswap with some bytes
15521 zeroed, as long as at least two bytes are not zeroed.
15522 (bswap_replace): Add mask argument and handle masking of bswap
15524 (maybe_optimize_vector_constructor): Adjust find_bswap_or_nop
15525 caller, punt if cast64_to_32 or mask is not all ones.
15526 (pass_optimize_bswap::execute): Adjust find_bswap_or_nop_finalize
15527 caller, for now punt if cast64_to_32.
15529 2021-08-23 Richard Biener <rguenther@suse.de>
15531 PR tree-optimization/79334
15532 * tree-ssa-sccvn.c (copy_reference_ops_from_ref): Record
15533 a type also for COMPONENT_REFs.
15534 (vn_reference_may_trap): Check ARRAY_REF with constant index
15535 against the array domain.
15537 2021-08-23 liuhongt <hongtao.liu@intel.com>
15540 * config/i386/sse.md (*avx512f_pshufb_truncv8hiv8qi_1): Add
15541 TARGET_AVX512BW to condition.
15543 2021-08-23 Jakub Jelinek <jakub@redhat.com>
15546 * dwarf2out.c (gen_variable_die): Add DW_AT_location for global
15547 register variables already during early_dwarf if possible.
15549 2021-08-23 Christophe Lyon <christophe.lyon@foss.st.com>
15551 * config/arm/arm_mve.h: Fix __arm_vctp16q return type.
15553 2021-08-23 Christophe Lyon <christophe.lyon@foss.st.com>
15556 * config/arm/arm.opt: Fix typo.
15557 * config/arm/t-rmprofile: Fix typo.
15559 2021-08-23 Jakub Jelinek <jakub@redhat.com>
15561 * tree.h (OMP_CLAUSE_GRAINSIZE_STRICT): Define.
15562 (OMP_CLAUSE_NUM_TASKS_STRICT): Define.
15563 * tree-pretty-print.c (dump_omp_clause) <case OMP_CLAUSE_GRAINSIZE,
15564 case OMP_CLAUSE_NUM_TASKS>: Print strict: modifier.
15565 * omp-expand.c (expand_task_call): Use GOMP_TASK_FLAG_STRICT in iflags
15566 if either grainsize or num_tasks clause has the strict modifier.
15568 2021-08-23 Martin Liska <mliska@suse.cz>
15570 * dbgcnt.def (DEBUG_COUNTER): New counter.
15571 * gimple.c (gimple_call_arg_flags): Use it in IPA PTA.
15573 2021-08-23 Jan Hubicka <hubicka@ucw.cz>
15575 * ipa-modref.c (analyze_ssa_name_flags): Improve handling of return slot.
15577 2021-08-23 Xi Ruoyao <xry111@mengyan1223.wang>
15580 * config/mips/mips-protos.h (mips_msa_output_shift_immediate):
15582 * config/mips/mips.c (mips_msa_output_shift_immediate): New
15584 * config/mips/mips-msa.md (vashl<mode>3, vashr<mode>3,
15585 vlshr<mode>3): Call it.
15587 2021-08-22 Jan Hubicka <hubicka@ucw.cz>
15588 Martin Liska <mliska@suse.cz>
15590 PR middle-end/101949
15591 * ipa-modref.c (analyze_ssa_name_flags): Indirect call implies
15594 2021-08-21 Dragan Mladjenovic <OT_Dragan.Mladjenovic@mediatek.com>
15596 * config/mips/mips.c (mips_function_rodata_section,
15597 TARGET_ASM_FUNCTION_RODATA_SECTION): Removed.
15599 2021-08-21 John David Anglin <danglin@gcc.gnu.org>
15601 * config/pa/pa.c (pa_asm_output_aligned_common): Remove warning.
15603 2021-08-20 Serge Belyshev <belyshev@depni.sinp.msu.ru>
15605 * configure.ac (thread-local storage support): Remove tls_first_major
15606 and tls_first_minor. Use "$conftest_s" to check support.
15607 * configure: Regenerate.
15609 2021-08-20 Serge Belyshev <belyshev@depni.sinp.msu.ru>
15611 * configure.ac: Fixup formatting.
15613 2021-08-20 Serge Belyshev <belyshev@depni.sinp.msu.ru>
15615 * acinclude.m4 (gcc_GAS_CHECK_FEATURE): Remove third argument and ...
15616 * configure.ac: ... update all callers.
15618 2021-08-20 Serge Belyshev <belyshev@depni.sinp.msu.ru>
15621 * acinclude.m4 (_gcc_COMPUTE_GAS_VERSION, _gcc_GAS_VERSION_GTE_IFELSE)
15622 (gcc_GAS_VERSION_GTE_IFELSE): Remove.
15623 (gcc_GAS_CHECK_FEATURE): Do not handle in-tree case specially.
15624 * configure.ac: Remove gcc_cv_gas_major_version, gcc_cv_gas_minor_version.
15625 Remove remaining checks for in-tree assembler.
15626 * configure: Regenerate.
15628 2021-08-20 Jeff Law <jlaw@localhost.localdomain>
15630 * config/h8300/h8300.c (shift_alg_hi): Improve arithmetic shift right
15631 by 15 bits for H8/300H and H8/S. Improve logical shifts by 12
15633 (shift_alg_si): Improve arithmetic right shift by 28-30 bits for
15634 H8/300H. Improve arithmetic shift right by 15 bits for H8/S.
15635 Improve logical shifts by 27 bits for H8/S.
15636 (get_shift_alg): Corresponding changes.
15637 (h8300_option_override): Revert to loops for -Os when profitable.
15639 2021-08-20 Richard Biener <rguenther@suse.de>
15641 * tree-vect-data-refs.c (dr_group_sort_cmp): Do not compare
15643 (vect_analyze_data_ref_accesses): Likewise. Assign the BB
15644 index as group_id when dataref_groups were not computed.
15645 * tree-vect-slp.c (vect_slp_bbs): Bump current_group when
15646 we advace to the next BB.
15648 2021-08-20 Jakub Jelinek <jakub@redhat.com>
15650 * omp-builtins.def (BUILT_IN_GOMP_WARNING, BUILT_IN_GOMP_ERROR): New
15653 2021-08-20 Martin Liska <mliska@suse.cz>
15655 PR gcov-profile/89961
15656 * gcov.c (make_gcov_file_name): Rewrite using std::string.
15657 (mangle_name): Simplify, do not used the second argument.
15658 (strip_extention): New function.
15659 (get_md5sum): Likewise.
15660 (get_gcov_intermediate_filename): Handle properly -p and -x
15662 (output_gcov_file): Use string type.
15663 (generate_results): Likewise.
15664 (md5sum_to_hex): Remove.
15666 2021-08-20 Michael Meissner <meissner@linux.ibm.com>
15668 * config/rs6000/altivec.md (UNSPEC_XXEVAL): Move to vsx.md.
15669 (UNSPEC_XXSPLTIW): Move to vsx.md.
15670 (UNSPEC_XXSPLTID): Move to vsx.md.
15671 (UNSPEC_XXSPLTI32DX): Move to vsx.md.
15672 (UNSPEC_XXBLEND): Move to vsx.md.
15673 (UNSPEC_XXPERMX): Move to vsx.md.
15674 (VM3): Move to vsx.md.
15675 (VM3_char): Move to vsx.md.
15676 (xxspltiw_v4si): Move to vsx.md.
15677 (xxspltiw_v4sf): Move to vsx.md.
15678 (xxspltiw_v4sf_inst): Move to vsx.md.
15679 (xxspltidp_v2df): Move to vsx.md.
15680 (xxspltidp_v2df_inst): Move to vsx.md.
15681 (xxsplti32dx_v4si_inst): Move to vsx.md.
15682 (xxsplti32dx_v4sf): Move to vsx.md.
15683 (xxsplti32dx_v4sf_inst): Move to vsx.md.
15684 (xxblend_<mode>): Move to vsx.md.
15685 (xxpermx): Move to vsx.md.
15686 (xxpermx_inst): Move to vsx.md.
15687 * config/rs6000/vsx.md (UNSPEC_XXEVAL): Move from altivec.md.
15688 (UNSPEC_XXSPLTIW): Move from altivec.md.
15689 (UNSPEC_XXSPLTID): Move from altivec.md.
15690 (UNSPEC_XXSPLTI32DX): Move from altivec.md.
15691 (UNSPEC_XXBLEND): Move from altivec.md.
15692 (UNSPEC_XXPERMX): Move from altivec.md.
15693 (VM3): Move from altivec.md.
15694 (VM3_char): Move from altivec.md.
15695 (xxspltiw_v4si): Move from altivec.md.
15696 (xxspltiw_v4sf): Move from altivec.md.
15697 (xxspltiw_v4sf_inst): Move from altivec.md.
15698 (xxspltidp_v2df): Move from altivec.md.
15699 (xxspltidp_v2df_inst): Move from altivec.md.
15700 (xxsplti32dx_v4si_inst): Move from altivec.md.
15701 (xxsplti32dx_v4sf): Move from altivec.md.
15702 (xxsplti32dx_v4sf_inst): Move from altivec.md.
15703 (xxblend_<mode>): Move from altivec.md.
15704 (xxpermx): Move from altivec.md.
15705 (xxpermx_inst): Move from altivec.md.
15707 2021-08-19 Roger Sayle <roger@nextmovesoftware.com>
15709 * tree-vect-generic.c (expand_vector_operations_1): Use either
15710 gimplify_build1 or gimplify_build2 instead of gimple_build_assign
15711 when constructing scalar splat expressions.
15713 2021-08-19 Peter Bergner <bergner@linux.ibm.com>
15716 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_mma_builtin): Cast
15717 pointer to __vector_pair *.
15719 2021-08-19 Martin Sebor <msebor@redhat.com>
15721 * gimple-range.cc: Add comments.
15722 * gimple-range.h: Same.
15724 2021-08-19 Martin Sebor <msebor@redhat.com>
15726 PR middle-end/101984
15727 * gimple-ssa-warn-access.cc (pass_waccess::execute): Also call
15730 2021-08-19 Jeff Law <jlaw@localhost.localdomain>
15732 * config.gcc (h8300-*-elf*): Do not include dbxelf.h.
15733 (h8300-*-linux*, v850-*-rtems*, v850*-elf*): Likewise.
15734 * config/v850/v850.h (DEFAULT_GDB_EXTENSIONS): Remove.
15736 2021-08-19 Jakub Jelinek <jakub@redhat.com>
15738 PR middle-end/101950
15739 * optabs.c (expand_clrsb_using_clz): New function.
15740 (expand_unop): Use it as another clrsb expansion fallback.
15742 2021-08-19 liuhongt <hongtao.liu@intel.com>
15745 2021-07-28 liuhongt <hongtao.liu@intel.com>
15748 * config/i386/i386.h (processor_costs): Add new member
15750 * config/i386/x86-tune-costs.h (ix86_size_cost, i386_cost,
15751 i486_cost, pentium_cost, lakemont_cost, pentiumpro_cost,
15752 geode_cost, k6_cost, athlon_cost, k8_cost, amdfam10_cost,
15753 bdver_cost, znver1_cost, znver2_cost, znver3_cost,
15754 btver1_cost, btver2_cost, btver3_cost, pentium4_cost,
15755 nocona_cost, atom_cost, atom_cost, slm_cost, intel_cost,
15756 generic_cost, core_cost): Initialize integer_to_sse same value
15758 (skylake_cost): Initialize integer_to_sse twice as much as sse_op.
15759 * config/i386/i386.c (ix86_builtin_vectorization_cost):
15760 Use integer_to_sse instead of sse_op to calculate the cost of
15763 2021-08-18 Iain Sandoe <iain@sandoe.co.uk>
15765 * config.gcc: Include rpath.opt for Darwin.
15766 * config/darwin.h (DRIVER_SELF_SPECS): Handle -rpath.
15768 2021-08-18 Thomas Schwinge <thomas@codesourcery.com>
15770 PR bootstrap/101959
15771 * hash-map-tests.c (test_map_of_type_with_ctor_and_dtor_expand):
15774 2021-08-18 Jonathan Wright <jonathan.wright@arm.com>
15776 * config/aarch64/arm_neon.h (vld3_lane_f64): Use float RTL
15777 pattern and type cast.
15778 (vld4_lane_f32): Use float RTL pattern.
15779 (vld4q_lane_f64): Use float type cast.
15781 2021-08-18 Jan Hubicka <hubicka@ucw.cz>
15783 * tree-ssa-uninit.c (maybe_warn_pass_by_reference): Check also
15786 2021-08-18 Thomas Schwinge <thomas@codesourcery.com>
15788 * hash-map-tests.c (test_map_of_type_with_ctor_and_dtor): Extend.
15789 (test_map_of_type_with_ctor_and_dtor_expand): Add function.
15790 (hash_map_tests_c_tests): Call it.
15792 2021-08-18 Thomas Schwinge <thomas@codesourcery.com>
15794 * ggc.h (enum ggc_collect): New.
15795 (ggc_collect): Use it.
15796 * ggc-page.c: Adjust.
15797 * ggc-common.c: Likewise.
15798 * ggc-tests.c: Likewise.
15799 * read-rtl-function.c: Likewise.
15800 * selftest-run-tests.c: Likewise.
15801 * doc/gty.texi (Invoking the garbage collector): Likewise.
15803 2021-08-18 liuhongt <hongtao.liu@intel.com>
15806 * config/i386/i386.h (TARGET_V2DF_REDUCTION_PREFER_HADDPD):
15808 * config/i386/sse.md (*sse3_haddv2df3_low): Add
15809 TARGET_V2DF_REDUCTION_PREFER_HADDPD.
15810 (*sse3_hsubv2df3_low): Ditto.
15811 * config/i386/x86-tune.def
15812 (X86_TUNE_V2DF_REDUCTION_PREFER_HADDPD): New tune.
15814 2021-08-17 Andrew MacLeod <amacleod@redhat.com>
15816 * gimple-range-gori.cc (gori_compute::gori_compute): Enable tracing.
15817 (gori_compute::compute_operand_range): Add tracing.
15818 (gori_compute::logical_combine): Ditto.
15819 (gori_compute::compute_logical_operands): Ditto.
15820 (gori_compute::compute_operand1_range): Ditto.
15821 (gori_compute::compute_operand2_range): Ditto.
15822 (gori_compute::outgoing_edge_range_p): Ditto.
15823 * gimple-range-gori.h (class gori_compute): Add range_tracer.
15825 2021-08-17 Andrew MacLeod <amacleod@redhat.com>
15827 * flag-types.h (enum evrp_mode): Adjust evrp-mode values.
15828 * gimple-range-cache.cc (DEBUG_RANGE_CACHE): Relocate from.
15829 * gimple-range-trace.h (DEBUG_RANGE_CACHE): Here.
15830 * params.opt (--param=evrp-mode): Adjust options.
15832 2021-08-17 Andrew MacLeod <amacleod@redhat.com>
15834 * Makefile.in (OBJS): Add gimple-range-trace.o.
15835 * gimple-range-cache.h (enable_new_values): Remove unused prototype.
15836 * gimple-range-fold.cc: Adjust headers.
15837 * gimple-range-trace.cc: New.
15838 * gimple-range-trace.h: New.
15839 * gimple-range.cc (gimple_ranger::gimple_ranger): Enable tracer.
15840 (gimple_ranger::range_of_expr): Add tracing.
15841 (gimple_ranger::range_on_entry): Ditto.
15842 (gimple_ranger::range_on_exit): Ditto.
15843 (gimple_ranger::range_on_edge): Ditto.
15844 (gimple_ranger::fold_range_internal): Ditto.
15845 (gimple_ranger::dump_bb): Do not calculate edge range twice.
15846 (trace_ranger::*): Remove.
15847 (enable_ranger): Never create a trace_ranger.
15848 (debug_seed_ranger): Move to gimple-range-trace.cc.
15849 (dump_ranger): Ditto.
15850 (debug_ranger): Ditto.
15851 * gimple-range.h: Include gimple-range-trace.h.
15852 (range_on_entry, range_on_exit): No longer virtual.
15853 (class trace_ranger): Remove.
15854 (DEBUG_RANGE_CACHE): Move to gimple-range-trace.h.
15856 2021-08-17 Martin Sebor <msebor@redhat.com>
15858 PR middle-end/101854
15859 * builtins.c (expand_builtin_alloca): Move warning code to check_alloca
15860 in gimple-ssa-warn-access.cc.
15861 * calls.c (alloc_max_size): Move code to check_alloca.
15862 (get_size_range): Move to pointer-query.cc.
15863 (maybe_warn_alloc_args_overflow): Move to gimple-ssa-warn-access.cc.
15864 (get_attr_nonstring_decl): Move to tree.c.
15865 (fntype_argno_type): Move to gimple-ssa-warn-access.cc.
15866 (append_attrname): Same.
15867 (maybe_warn_rdwr_sizes): Same.
15868 (initialize_argument_information): Move code to
15869 gimple-ssa-warn-access.cc.
15870 * calls.h (maybe_warn_alloc_args_overflow): Move to
15871 gimple-ssa-warn-access.h.
15872 (get_attr_nonstring_decl): Move to tree.h.
15873 (maybe_warn_nonstring_arg): Move to gimple-ssa-warn-access.h.
15874 (enum size_range_flags): Move to pointer-query.h.
15875 (get_size_range): Same.
15876 * gimple-ssa-warn-access.cc (has_location): Remove unused overload
15877 to avoid Clang -Wunused-function.
15878 (get_size_range): Declare static.
15879 (maybe_emit_free_warning): Rename...
15880 (maybe_check_dealloc_call): ...to this for consistency.
15881 (class pass_waccess): Add members.
15882 (pass_waccess::~pass_waccess): Defined.
15883 (alloc_max_size): Move here from calls.c.
15884 (maybe_warn_alloc_args_overflow): Same.
15885 (check_alloca): New function.
15886 (check_alloc_size_call): New function.
15887 (check_strncat): Handle another warning flag.
15888 (pass_waccess::check_builtin): Handle alloca.
15889 (fntype_argno_type): Move here from calls.c.
15890 (append_attrname): Same.
15891 (maybe_warn_rdwr_sizes): Same.
15892 (pass_waccess::check_call): Define.
15893 (check_nonstring_args): New function.
15894 (pass_waccess::check): Call new member functions.
15895 (pass_waccess::execute): Enable ranger.
15896 * gimple-ssa-warn-access.h (get_size_range): Move here from calls.h.
15897 (maybe_warn_nonstring_arg): Same.
15898 * gimple-ssa-warn-restrict.c: Remove #include.
15899 * pointer-query.cc (get_size_range): Move here from calls.c.
15900 * pointer-query.h (enum size_range_flags): Same.
15901 (get_size_range): Same.
15902 * tree.c (get_attr_nonstring_decl): Move here from calls.c.
15903 * tree.h (get_attr_nonstring_decl): Move here from calls.h.
15905 2021-08-17 Thomas Schwinge <thomas@codesourcery.com>
15907 * ggc.h (ggc_collect): Add 'force_collect' parameter.
15908 * ggc-page.c (ggc_collect): Use that one instead of global
15909 'ggc_force_collect'. Adjust all users.
15910 * doc/gty.texi (Invoking the garbage collector): Update.
15911 * ggc-internal.h (ggc_force_collect): Remove.
15912 * ggc-common.c (ggc_force_collect): Likewise.
15913 * selftest.h (forcibly_ggc_collect): Remove.
15914 * ggc-tests.c (selftest::forcibly_ggc_collect): Likewise.
15915 * read-rtl-function.c (test_loading_labels): Adjust.
15916 * selftest-run-tests.c (run_tests): Likewise.
15918 2021-08-17 Iain Sandoe <iain@sandoe.co.uk>
15920 * config/darwin.c (darwin_file_end): Reset and reclaim the
15921 section names table at the end of compile.
15923 2021-08-17 Iain Sandoe <iain@sandoe.co.uk>
15926 * config.in: Regenerate.
15927 * config/i386/darwin.h (EXTRA_ASM_OPTS): New
15928 (ASM_SPEC): Pass options to disable branch shortening where
15930 * configure: Regenerate.
15931 * configure.ac: Detect versions of 'as' that support the
15932 optimisation which has the bug.
15934 2021-08-17 Richard Biener <rguenther@suse.de>
15936 * optabs-query.c (supports_vec_gather_load_p): Also check
15938 (supports_vec_scatter_store_p): Likewise.
15939 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Fall
15940 back to masked variants if non-masked are not supported.
15941 * tree-vect-patterns.c (vect_recog_gather_scatter_pattern):
15942 When we need to use masked gather/scatter but do not have
15943 a mask set up a constant true one.
15944 * tree-vect-stmts.c (vect_check_scalar_mask): Also allow
15945 non-SSA_NAME masks.
15947 2021-08-17 Roger Sayle <roger@nextmovesoftware.com>
15949 * tree-ssa-ccp.c (bit_value_binop) [MINUS_EXPR]: Use same
15950 algorithm as PLUS_EXPR to improve subtraction bit bounds.
15951 [POINTER_DIFF_EXPR]: Treat as synonymous with MINUS_EXPR.
15953 2021-08-17 Roger Sayle <roger@nextmovesoftware.com>
15955 * tree-ssa-ccp.c (bit_value_mult_const): New helper function to
15956 calculate the mask-value pair result of a multiplication by an
15958 (bit_value_binop) [MULT_EXPR]: Call it from here for
15959 multiplications by (sparse) non-negative constants.
15961 2021-08-17 Christophe Lyon <christophe.lyon@foss.st.com>
15964 * config.gcc (gcc_cv_initfini_array): Leave undefined for
15965 uclinuxfdpiceabi targets.
15967 2021-08-17 Alexandre Oliva <oliva@adacore.com>
15969 * tree-inline.c (maybe_move_debug_stmts_to_successors): Don't
15970 reverse debug stmts.
15972 2021-08-17 Alexandre Oliva <oliva@adacore.com>
15974 * tree-cfg.c (dump_function_to_file): Use fun, not cfun.
15976 2021-08-17 Jonathan Wright <jonathan.wright@arm.com>
15978 * config/aarch64/arm_neon.h (__LD4_LANE_FUNC): Delete.
15979 (__LD4Q_LANE_FUNC): Likewise.
15980 (vld4_lane_u8): Define without macro.
15981 (vld4_lane_u16): Likewise.
15982 (vld4_lane_u32): Likewise.
15983 (vld4_lane_u64): Likewise.
15984 (vld4_lane_s8): Likewise.
15985 (vld4_lane_s16): Likewise.
15986 (vld4_lane_s32): Likewise.
15987 (vld4_lane_s64): Likewise.
15988 (vld4_lane_f16): Likewise.
15989 (vld4_lane_f32): Likewise.
15990 (vld4_lane_f64): Likewise.
15991 (vld4_lane_p8): Likewise.
15992 (vld4_lane_p16): Likewise.
15993 (vld4_lane_p64): Likewise.
15994 (vld4q_lane_u8): Likewise.
15995 (vld4q_lane_u16): Likewise.
15996 (vld4q_lane_u32): Likewise.
15997 (vld4q_lane_u64): Likewise.
15998 (vld4q_lane_s8): Likewise.
15999 (vld4q_lane_s16): Likewise.
16000 (vld4q_lane_s32): Likewise.
16001 (vld4q_lane_s64): Likewise.
16002 (vld4q_lane_f16): Likewise.
16003 (vld4q_lane_f32): Likewise.
16004 (vld4q_lane_f64): Likewise.
16005 (vld4q_lane_p8): Likewise.
16006 (vld4q_lane_p16): Likewise.
16007 (vld4q_lane_p64): Likewise.
16008 (vld4_lane_bf16): Likewise.
16009 (vld4q_lane_bf16): Likewise.
16011 2021-08-17 Jonathan Wright <jonathan.wright@arm.com>
16013 * config/aarch64/arm_neon.h (__LD3_LANE_FUNC): Delete.
16014 (__LD3Q_LANE_FUNC): Delete.
16015 (vld3_lane_u8): Define without macro.
16016 (vld3_lane_u16): Likewise.
16017 (vld3_lane_u32): Likewise.
16018 (vld3_lane_u64): Likewise.
16019 (vld3_lane_s8): Likewise.
16020 (vld3_lane_s16): Likewise.
16021 (vld3_lane_s32): Likewise.
16022 (vld3_lane_s64): Likewise.
16023 (vld3_lane_f16): Likewise.
16024 (vld3_lane_f32): Likewise.
16025 (vld3_lane_f64): Likewise.
16026 (vld3_lane_p8): Likewise.
16027 (vld3_lane_p16): Likewise.
16028 (vld3_lane_p64): Likewise.
16029 (vld3q_lane_u8): Likewise.
16030 (vld3q_lane_u16): Likewise.
16031 (vld3q_lane_u32): Likewise.
16032 (vld3q_lane_u64): Likewise.
16033 (vld3q_lane_s8): Likewise.
16034 (vld3q_lane_s16): Likewise.
16035 (vld3q_lane_s32): Likewise.
16036 (vld3q_lane_s64): Likewise.
16037 (vld3q_lane_f16): Likewise.
16038 (vld3q_lane_f32): Likewise.
16039 (vld3q_lane_f64): Likewise.
16040 (vld3q_lane_p8): Likewise.
16041 (vld3q_lane_p16): Likewise.
16042 (vld3q_lane_p64): Likewise.
16043 (vld3_lane_bf16): Likewise.
16044 (vld3q_lane_bf16): Likewise.
16046 2021-08-17 Jonathan Wright <jonathan.wright@arm.com>
16048 * config/aarch64/arm_neon.h (__LD2_LANE_FUNC): Delete.
16049 (__LD2Q_LANE_FUNC): Likewise.
16050 (vld2_lane_u8): Define without macro.
16051 (vld2_lane_u16): Likewise.
16052 (vld2_lane_u32): Likewise.
16053 (vld2_lane_u64): Likewise.
16054 (vld2_lane_s8): Likewise.
16055 (vld2_lane_s16): Likewise.
16056 (vld2_lane_s32): Likewise.
16057 (vld2_lane_s64): Likewise.
16058 (vld2_lane_f16): Likewise.
16059 (vld2_lane_f32): Likewise.
16060 (vld2_lane_f64): Likewise.
16061 (vld2_lane_p8): Likewise.
16062 (vld2_lane_p16): Likewise.
16063 (vld2_lane_p64): Likewise.
16064 (vld2q_lane_u8): Likewise.
16065 (vld2q_lane_u16): Likewise.
16066 (vld2q_lane_u32): Likewise.
16067 (vld2q_lane_u64): Likewise.
16068 (vld2q_lane_s8): Likewise.
16069 (vld2q_lane_s16): Likewise.
16070 (vld2q_lane_s32): Likewise.
16071 (vld2q_lane_s64): Likewise.
16072 (vld2q_lane_f16): Likewise.
16073 (vld2q_lane_f32): Likewise.
16074 (vld2q_lane_f64): Likewise.
16075 (vld2q_lane_p8): Likewise.
16076 (vld2q_lane_p16): Likewise.
16077 (vld2q_lane_p64): Likewise.
16078 (vld2_lane_bf16): Likewise.
16079 (vld2q_lane_bf16): Likewise.
16081 2021-08-17 Maxim Kuvyrkov <maxim.kuvyrkov@linaro.org>
16083 * haifa-sched.c (advance_one_cycle): Output more context-synchronization
16086 2021-08-17 Maxim Kuvyrkov <maxim.kuvyrkov@linaro.org>
16088 * haifa-sched.c (enum rfs_decision, rfs_str): Add RFS_AUTOPREF.
16089 (rank_for_schedule): Use it.
16091 2021-08-17 Maxim Kuvyrkov <maxim.kuvyrkov@linaro.org>
16093 PR rtl-optimization/91598
16094 * haifa-sched.c (autopref_rank_for_schedule): Prioritize "irrelevant"
16095 insns after memory reads and before memory writes.
16097 2021-08-17 Alistair_Lee <alistair.lee@arm.com>
16099 * rtl.h (CONST_VECTOR_P): New macro.
16100 * config/aarch64/aarch64.c (aarch64_get_sve_pred_bits): Use RTL
16101 code testing macros.
16102 (aarch64_ptrue_all_mode): Likewise.
16103 (aarch64_expand_mov_immediate): Likewise.
16104 (aarch64_const_vec_all_in_range_p): Likewise.
16105 (aarch64_rtx_costs): Likewise.
16106 (aarch64_legitimate_constant_p): Likewise.
16107 (aarch64_simd_valid_immediate): Likewise.
16108 (aarch64_simd_make_constant): Likewise.
16109 (aarch64_convert_mult_to_shift): Likewise.
16110 (aarch64_expand_sve_vec_perm): Likewise.
16111 (aarch64_vec_fpconst_pow_of_2): Likewise.
16113 2021-08-17 Andrew MacLeod <amacleod@redhat.com>
16115 PR tree-optimization/101938
16116 * range-op.cc (operator_abs::op1_range): Special case
16117 -TYPE_MIN_VALUE for flag_wrapv.
16119 2021-08-17 Kewen Lin <linkw@linux.ibm.com>
16121 * tree-vect-slp.c (vectorizable_bb_reduc_epilogue): Add the cost for
16124 2021-08-17 Jakub Jelinek <jakub@redhat.com>
16126 * tree.def (OMP_SCOPE): New tree code.
16127 * tree.h (OMP_SCOPE_BODY, OMP_SCOPE_CLAUSES): Define.
16128 * tree-nested.c (convert_nonlocal_reference_stmt,
16129 convert_local_reference_stmt, convert_gimple_call): Handle
16131 * tree-pretty-print.c (dump_generic_node): Handle OMP_SCOPE.
16132 * gimple.def (GIMPLE_OMP_SCOPE): New gimple code.
16133 * gimple.c (gimple_build_omp_scope): New function.
16134 (gimple_copy): Handle GIMPLE_OMP_SCOPE.
16135 * gimple.h (gimple_build_omp_scope): Declare.
16136 (gimple_has_substatements): Handle GIMPLE_OMP_SCOPE.
16137 (gimple_omp_scope_clauses, gimple_omp_scope_clauses_ptr,
16138 gimple_omp_scope_set_clauses): New inline functions.
16139 (CASE_GIMPLE_OMP): Add GIMPLE_OMP_SCOPE.
16140 * gimple-pretty-print.c (dump_gimple_omp_scope): New function.
16141 (pp_gimple_stmt_1): Handle GIMPLE_OMP_SCOPE.
16142 * gimple-walk.c (walk_gimple_stmt): Likewise.
16143 * gimple-low.c (lower_stmt): Likewise.
16144 * gimplify.c (is_gimple_stmt): Handle OMP_MASTER.
16145 (gimplify_scan_omp_clauses): For task reductions, handle OMP_SCOPE
16146 like ORT_WORKSHARE constructs. Adjust diagnostics for %<scope%>
16147 allowing task reductions. Reject inscan reductions on scope.
16148 (omp_find_stores_stmt): Handle GIMPLE_OMP_SCOPE.
16149 (gimplify_omp_workshare, gimplify_expr): Handle OMP_SCOPE.
16150 * tree-inline.c (remap_gimple_stmt): Handle GIMPLE_OMP_SCOPE.
16151 (estimate_num_insns): Likewise.
16152 * omp-low.c (build_outer_var_ref): Look through GIMPLE_OMP_SCOPE
16153 contexts if var isn't privatized there.
16154 (check_omp_nesting_restrictions): Handle GIMPLE_OMP_SCOPE.
16155 (scan_omp_1_stmt): Likewise.
16156 (maybe_add_implicit_barrier_cancel): Look through outer
16158 (lower_omp_scope): New function.
16159 (lower_omp_task_reductions): Handle OMP_SCOPE.
16160 (lower_omp_1): Handle GIMPLE_OMP_SCOPE.
16161 (diagnose_sb_1, diagnose_sb_2): Likewise.
16162 * omp-expand.c (expand_omp_single): Support also GIMPLE_OMP_SCOPE.
16163 (expand_omp): Handle GIMPLE_OMP_SCOPE.
16164 (omp_make_gimple_edges): Likewise.
16165 * omp-builtins.def (BUILT_IN_GOMP_SCOPE_START): New built-in.
16167 2021-08-17 Richard Biener <rguenther@suse.de>
16169 PR tree-optimization/101925
16170 * tree-ssa-sccvn.c (copy_reference_ops_from_ref): Set
16171 reverse on COMPONENT_REF and ARRAY_REF according to
16172 what reverse_storage_order_for_component_p does.
16173 (vn_reference_eq): Compare reversed on reference ops.
16174 (reverse_storage_order_for_component_p): New overload.
16175 (vn_reference_lookup_3): Check reverse_storage_order_for_component_p
16176 on the reference looked up.
16178 2021-08-17 Jeff Law <jlaw@localhost.localdomain>
16180 * config/h8300/h8300.c (shift_alg_si): Avoid loops for most SImode
16181 shifts on the H8/S.
16182 (h8300_option_override): Use loops on H8/S more often when optimizing
16184 (get_shift_alg): Handle new "special" cases on H8/S. Simplify
16185 accordingly. Handle various arithmetic right shifts with special
16186 sequences that we couldn't handle before.
16188 2021-08-16 Jeff Law <jlaw@localhost.localdomain>
16190 * config.gcc (rl78-*-elf*): Do not include dbxelf.h.
16192 2021-08-16 Sebastian Huber <sebastian.huber@embedded-brains.de>
16194 * config/sparc/rtemself.h (SPARC_GCOV_TYPE_SIZE): Define.
16195 * config/sparc/sparc.c (sparc_gcov_type_size): New.
16196 (TARGET_GCOV_TYPE_SIZE): Redefine if SPARC_GCOV_TYPE_SIZE is defined.
16197 * coverage.c (get_gcov_type): Use targetm.gcov_type_size().
16198 * doc/tm.texi (TARGET_GCOV_TYPE_SIZE): Add hook under "Misc".
16199 * doc/tm.texi.in: Regenerate.
16200 * target.def (gcov_type_size): New target hook.
16201 * targhooks.c (default_gcov_type_size): New.
16202 * targhooks.h (default_gcov_type_size): Declare.
16203 * tree-profile.c (gimple_gen_edge_profiler): Use precision of
16205 (gimple_gen_time_profiler): Likewise.
16207 2021-08-16 Eric Botcazou <ebotcazou@gcc.gnu.org>
16209 * dwarf2out.c (add_scalar_info): Deal with DW_AT_data_bit_offset.
16211 2021-08-16 Tobias Burnus <tobias@codesourcery.com>
16213 PR middle-end/101931
16214 * omp-low.c (omp_runtime_api_call): Update for routines
16215 added in the meanwhile.
16217 2021-08-16 Martin Liska <mliska@suse.cz>
16219 PR tree-optimization/100393
16220 * tree-switch-conversion.c (group_cluster::dump): Use
16221 get_comparison_count.
16222 (jump_table_cluster::find_jump_tables): Pre-compute number of
16223 comparisons and then decrement it. Cache also max_ratio.
16224 (jump_table_cluster::can_be_handled): Change signature.
16225 * tree-switch-conversion.h (get_comparison_count): New.
16227 2021-08-16 Eric Botcazou <ebotcazou@gcc.gnu.org>
16229 * dwarf2out.c (add_data_member_location_attribute): Use GNAT
16230 encodings only when -fgnat-encodings=all is specified.
16231 (add_bound_info): Likewise.
16232 (add_byte_size_attribute): Likewise.
16233 (gen_member_die): Likewise.
16235 2021-08-16 Thomas Schwinge <thomas@codesourcery.com>
16237 * omp-oacc-neuter-broadcast.cc
16238 (execute_omp_oacc_neuter_broadcast): Plug 'par' memory leak.
16240 2021-08-16 Thomas Schwinge <thomas@codesourcery.com>
16242 * omp-oacc-neuter-broadcast.cc
16243 (execute_omp_oacc_neuter_broadcast): Clarify memory management for
16246 2021-08-16 Thomas Schwinge <thomas@codesourcery.com>
16248 * omp-oacc-neuter-broadcast.cc (field_map): Move variable into...
16249 (execute_omp_oacc_neuter_broadcast): ... here.
16250 (install_var_field, build_receiver_ref, build_sender_ref): Take
16251 'field_map_t *' parameter. Adjust all users.
16252 (worker_single_copy, neuter_worker_single): Take a
16253 'record_field_map_t *' parameter. Adjust all users.
16255 2021-08-16 liuhongt <hongtao.liu@intel.com>
16258 * config/i386/i386.md (ldexp<mode>3): Force operands[1] to
16261 2021-08-16 Martin Liska <mliska@suse.cz>
16264 * multiple_target.c (create_dispatcher_calls): Make default
16265 function local only if it is a definition.
16267 2021-08-16 Martin Liska <mliska@suse.cz>
16270 * ipa-icf-gimple.c (func_checker::compare_ssa_name): Do not
16271 consider equal SSA_NAMEs when one is a param.
16273 2021-08-16 liuhongt <hongtao.liu@intel.com>
16276 * config/i386/i386-expand.c (ix86_expand_vec_perm_vpermt2):
16277 Support vpermi2b for V32QI/V16QImode.
16278 (ix86_extract_perm_from_pool_constant): New function.
16279 (ix86_expand_vec_one_operand_perm_avx512): Support
16280 vpermw/vpermb under TARGET_AVX512BW/TARGET_AVX512VBMI.
16281 (expand_vec_perm_1): Adjust comments for upper.
16282 * config/i386/i386-protos.h (ix86_extract_perm_from_pool_constant):
16284 * config/i386/predicates.md (permvar_truncate_operand): New predicate.
16285 (pshufb_truncv4siv4hi_operand): Ditto.
16286 (pshufb_truncv8hiv8qi_operand): Ditto.
16287 * config/i386/sse.md (*avx512bw_permvar_truncv16siv16hi_1):
16288 New pre_reload define_insn_and_split.
16289 (*avx512f_permvar_truncv8siv8hi_1): Ditto.
16290 (*avx512f_vpermvar_truncv8div8si_1): Ditto.
16291 (*avx512f_permvar_truncv32hiv32qi_1): Ditto.
16292 (*avx512f_permvar_truncv16hiv16qi_1): Ditto.
16293 (*avx512f_permvar_truncv4div4si_1): Ditto.
16294 (*avx512f_pshufb_truncv8hiv8qi_1): Ditto.
16295 (*avx512f_pshufb_truncv4siv4hi_1): Ditto.
16296 (*avx512f_pshufd_truncv2div2si_1): Ditto.
16298 2021-08-16 Kito Cheng <kito.cheng@sifive.com>
16300 * config/riscv/multilib-generator: Support code model option for
16302 * doc/install.texi: Add document of new option for
16303 --with-multilib-generator.
16305 2021-08-15 Clément Chigot <clement.chigot@atos.net>
16307 * config/rs6000/rs6000.c (xcoff_tls_exec_model_detected): New.
16308 (rs6000_legitimize_tls_address_aix): Use it.
16309 (rs6000_xcoff_file_end): Add ".ref __tls_get_addr" when
16310 xcoff_tls_exec_model_detected is true.
16312 2021-08-15 Jeff Law <jlaw@localhost.localdomain>
16314 * config/h8300/h8300.c (shift_alg_si): Retune H8/300H shifts
16315 to allow a bit more code growth, saving many dozens of cycles.
16316 (h8300_option_override): Adjus shift_alg_si if optimizing for
16318 (get_shift_alg): Use special + inline shifts for residuals
16321 2021-08-14 Stafford Horne <shorne@gmail.com>
16324 * config/or1k/or1k-opts.h: New file.
16325 * config/or1k/or1k.c (or1k_legitimize_address_1, print_reloc):
16326 Support generating gotha relocations if -mcmodel=large is
16328 * config/or1k/or1k.h (TARGET_CMODEL_SMALL, TARGET_CMODEL_LARGE):
16330 * config/or1k/or1k.opt (mcmodel=): New option.
16331 * doc/invoke.texi (OpenRISC Options): Document mcmodel.
16333 2021-08-14 Martin Sebor <msebor@redhat.com>
16335 PR middle-end/101791
16336 * gimple-ssa-warn-access.cc (new_delete_mismatch_p): Use new argument
16337 to valid_new_delete_pair_p.
16338 * tree.c (valid_new_delete_pair_p): Add argument.
16339 * tree.h (valid_new_delete_pair_p): Same.
16341 2021-08-14 Jakub Jelinek <jakub@redhat.com>
16344 * config/i386/i386-expand.c (expand_vec_perm_broadcast_1)
16345 <case E_V64QImode>: For this mode assert
16346 !TARGET_AVX512BW || d->perm[0] rather than !TARGET_AVX2 || d->perm[0].
16348 2021-08-13 Michael Meissner <meissner@linux.ibm.com>
16351 * config/rs6000/altivec.md (xxeval): Use register_predicate
16352 instead of altivec_register_predicate.
16354 2021-08-13 Martin Sebor <msebor@redhat.com>
16356 PR middle-end/101734
16357 * tree-ssa-uninit.c (maybe_warn_read_write_only): New function.
16358 (maybe_warn_operand): Call it.
16360 2021-08-13 Martin Liska <mliska@suse.cz>
16363 * attribs.c (decl_attributes): Make naked functions "noipa"
16366 2021-08-13 Martin Liska <mliska@suse.cz>
16369 * symtab.c (symtab_node::noninterposable_alias): Do not create
16370 local aliases for target_clone functions as the clonning pass
16373 2021-08-13 Martin Liska <mliska@suse.cz>
16375 * opts.c (LIVE_PATCHING_OPTION): Define.
16376 (control_options_for_live_patching): Use it in error messages.
16378 2021-08-13 Jan Hubicka <hubicka@ucw.cz>
16380 * ipa-modref.c (dump_eaf_flags): Dump EAF_NOREAD.
16381 (implicit_const_eaf_flags, implicit_pure_eaf_flags,
16382 ignore_stores_eaf_flags): New constants.
16383 (remove_useless_eaf_flags): New function.
16384 (eaf_flags_useful_p): Use it.
16385 (deref_flags): Add EAF_NOT_RETURNED if flag is unused;
16387 (modref_lattice::init): Add EAF_NOREAD.
16388 (modref_lattice::add_escape_point): Do not reacord escape point if
16390 (modref_lattice::merge): EAF_NOESCAPE implies EAF_NODIRECTESCAPE;
16391 use remove_useless_eaf_flags.
16392 (modref_lattice::merge_deref): Use ignore_stores_eaf_flags.
16393 (modref_lattice::merge_direct_load): Add EAF_NOREAD
16394 (analyze_ssa_name_flags): Fix handling EAF_NOT_RETURNED
16395 (analyze_parms): Use remove_useless_eaf_flags.
16396 (ipa_merge_modref_summary_after_inlining): Use ignore_stores_eaf_flags.
16397 (modref_merge_call_site_flags): Add caller and ecf_flags parameter;
16398 use remove_useless_eaf_flags.
16399 (modref_propagate_flags_in_scc): Update.
16400 * ipa-modref.h: Turn eaf_flags_t back to char.
16401 * tree-core.h (EAF_NOT_RETURNED): Fix.
16402 (EAF_NOREAD): New constant
16403 * tree-ssa-alias.c: (ref_maybe_used_by_call_p_1): Check for
16405 * tree-ssa-structalias.c (handle_rhs_call): Handle new flags.
16406 (handle_pure_call): Likewise.
16408 2021-08-12 Jakub Jelinek <jakub@redhat.com>
16410 * tree.def (OMP_MASKED): New tree code.
16411 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_FILTER.
16412 * tree.h (OMP_MASKED_BODY, OMP_MASKED_CLAUSES, OMP_MASKED_COMBINED,
16413 OMP_CLAUSE_FILTER_EXPR): Define.
16414 * tree.c (omp_clause_num_ops): Add OMP_CLAUSE_FILTER entry.
16415 (omp_clause_code_name): Likewise.
16416 (walk_tree_1): Handle OMP_CLAUSE_FILTER.
16417 * tree-nested.c (convert_nonlocal_omp_clauses,
16418 convert_local_omp_clauses): Handle OMP_CLAUSE_FILTER.
16419 (convert_nonlocal_reference_stmt, convert_local_reference_stmt,
16420 convert_gimple_call): Handle GIMPLE_OMP_MASTER.
16421 * tree-pretty-print.c (dump_omp_clause): Handle OMP_CLAUSE_FILTER.
16422 (dump_generic_node): Handle OMP_MASTER.
16423 * gimple.def (GIMPLE_OMP_MASKED): New gimple code.
16424 * gimple.c (gimple_build_omp_masked): New function.
16425 (gimple_copy): Handle GIMPLE_OMP_MASKED.
16426 * gimple.h (gimple_build_omp_masked): Declare.
16427 (gimple_has_substatements): Handle GIMPLE_OMP_MASKED.
16428 (gimple_omp_masked_clauses, gimple_omp_masked_clauses_ptr,
16429 gimple_omp_masked_set_clauses): New inline functions.
16430 (CASE_GIMPLE_OMP): Add GIMPLE_OMP_MASKED.
16431 * gimple-pretty-print.c (dump_gimple_omp_masked): New function.
16432 (pp_gimple_stmt_1): Handle GIMPLE_OMP_MASKED.
16433 * gimple-walk.c (walk_gimple_stmt): Likewise.
16434 * gimple-low.c (lower_stmt): Likewise.
16435 * gimplify.c (is_gimple_stmt): Handle OMP_MASTER.
16436 (gimplify_scan_omp_clauses): Handle OMP_CLAUSE_FILTER. For clauses
16437 that take one expression rather than decl or constant, force
16438 gimplification of that into a SSA_NAME or temporary unless min
16440 (gimplify_adjust_omp_clauses): Handle OMP_CLAUSE_FILTER.
16441 (gimplify_expr): Handle OMP_MASKED.
16442 * tree-inline.c (remap_gimple_stmt): Handle GIMPLE_OMP_MASKED.
16443 (estimate_num_insns): Likewise.
16444 * omp-low.c (scan_sharing_clauses): Handle OMP_CLAUSE_FILTER.
16445 (check_omp_nesting_restrictions): Handle GIMPLE_OMP_MASKED. Adjust
16446 diagnostics for existence of masked construct.
16447 (scan_omp_1_stmt, lower_omp_master, lower_omp_1, diagnose_sb_1,
16448 diagnose_sb_2): Handle GIMPLE_OMP_MASKED.
16449 * omp-expand.c (expand_omp_synch, expand_omp, omp_make_gimple_edges):
16452 2021-08-12 Uroš Bizjak <ubizjak@gmail.com>
16455 * config/i386/i386.md (avx512f_scalef<mode>2): New insn pattern.
16456 (ldexp<mode>3): Use avx512f_scalef<mode>2.
16457 (UNSPEC_SCALEF): Move from ...
16458 * config/i386/sse.md (UNSPEC_SCALEF): ... here.
16460 2021-08-12 Jan Hubicka <hubicka@ucw.cz>
16462 * ipa-split.c (consider_split): Fix condition testing void functions.
16464 2021-08-12 Aldy Hernandez <aldyh@redhat.com>
16466 * doc/invoke.texi: Remove docs for threader-mode param.
16467 * flag-types.h (enum threader_mode): Remove.
16468 * params.opt: Remove threader-mode param.
16469 * tree-ssa-threadbackward.c (class back_threader): Remove
16470 path_is_unreachable_p.
16471 Make find_paths private.
16472 Add maybe_thread and thread_through_all_blocks.
16473 Remove reference marker for m_registry.
16474 Remove reference marker for m_profit.
16475 (back_threader::back_threader): Adjust for registry and profit not
16477 (dump_path): Move down.
16478 (debug): Move down.
16479 (class thread_jumps): Remove.
16480 (class back_threader_registry): Remove m_all_paths.
16482 (thread_jumps::thread_through_all_blocks): Move to back_threader
16484 (fsm_find_thread_path): Remove
16485 (back_threader::maybe_thread): New.
16486 (back_threader::thread_through_all_blocks): Move from
16488 (back_threader_registry::back_threader_registry): Remove
16490 (back_threader_registry::~back_threader_registry): Remove.
16491 (thread_jumps::find_taken_edge): Remove.
16492 (thread_jumps::check_subpath_and_update_thread_path): Remove.
16493 (thread_jumps::maybe_register_path): Remove.
16494 (thread_jumps::handle_phi): Remove.
16495 (handle_assignment_p): Remove.
16496 (thread_jumps::handle_assignment): Remove.
16497 (thread_jumps::fsm_find_control_statement_thread_paths): Remove.
16498 (thread_jumps::find_jump_threads_backwards): Remove.
16499 (thread_jumps::find_jump_threads_backwards_with_ranger): Remove.
16500 (try_thread_blocks): Rename find_jump_threads_backwards to
16502 (pass_early_thread_jumps::execute): Same.
16504 2021-08-12 Tobias Burnus <tobias@codesourcery.com>
16506 * tree-core.h (omp_clause_proc_bind_kind): Add
16507 OMP_CLAUSE_PROC_BIND_PRIMARY.
16508 * tree-pretty-print.c (dump_omp_clause): Add TODO comment to
16509 change 'master' to 'primary' in proc_bind for OpenMP 5.1.
16511 2021-08-12 Claudiu Zissulescu <claziss@synopsys.com>
16513 * common/config/arc/arc-common.c (arc_option_init_struct): Remove
16514 fno-common reference.
16515 * config/arc/arc.c (arc_override_options): Remove overriding of
16518 2021-08-12 Jakub Jelinek <jakub@redhat.com>
16521 * config/i386/i386-expand.c (ix86_expand_vec_one_operand_perm_avx512):
16522 If d->testing_p, return true after performing checks instead of
16523 actually expanding the insn.
16524 (expand_vec_perm_broadcast_1): Handle V32HImode - assert
16525 !TARGET_AVX512BW and return false.
16527 2021-08-12 Eric Botcazou <ebotcazou@gcc.gnu.org>
16529 * configure.ac (PE linker --disable-dynamicbase support): New check.
16530 * configure: Regenerate.
16531 * config.in: Likewise.
16532 * config/i386/mingw32.h (LINK_SPEC_DISABLE_DYNAMICBASE): New define.
16533 (LINK_SPEC): Use it.
16534 * config/i386/mingw-w64.h (LINK_SPEC_DISABLE_DYNAMICBASE): Likewise.
16535 (LINK_SPEC): Likewise.
16537 2021-08-12 liuhongt <hongtao.liu@intel.com>
16540 * config/i386/sse.md (*avx2_zero_extendv16qiv16hi2_2): New
16541 post_reload define_insn_and_split.
16542 (*avx512bw_zero_extendv32qiv32hi2_2): Ditto.
16543 (*sse4_1_zero_extendv8qiv8hi2_4): Ditto.
16544 (*avx512f_zero_extendv16hiv16si2_2): Ditto.
16545 (*avx2_zero_extendv8hiv8si2_2): Ditto.
16546 (*sse4_1_zero_extendv4hiv4si2_4): Ditto.
16547 (*avx512f_zero_extendv8siv8di2_2): Ditto.
16548 (*avx2_zero_extendv4siv4di2_2): Ditto.
16549 (*sse4_1_zero_extendv2siv2di2_4): Ditto.
16550 (VI248_256, VI248_512, VI148_512, VI148_256, VI148_128): New
16553 2021-08-11 Bill Schmidt <wschmidt@linux.ibm.com>
16555 * config/rs6000/rs6000-builtin-new.def: Add always, power5, and
16558 2021-08-11 Bill Schmidt <wschmidt@linux.ibm.com>
16560 * config/rs6000/rs6000-builtin-new.def: Add vsx stanza.
16562 2021-08-11 Bill Schmidt <wschmidt@linux.ibm.com>
16564 * config/rs6000/rs6000-builtin-new.def: Finish altivec stanza.
16565 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Move
16566 initialization of pcvoid_type_node here...
16567 (altivec_init_builtins): ...from here.
16568 * config/rs6000/rs6000.h (rs6000_builtin_type_index): Add
16569 RS6000_BTI_const_ptr_void.
16570 (pcvoid_type_node): New macro.
16572 2021-08-11 Richard Biener <rguenther@suse.de>
16575 * tree-ssa-forwprop.c (pass_forwprop::execute): Do not decompose
16576 hard-register accesses.
16578 2021-08-11 Richard Biener <rguenther@suse.de>
16580 * tree-ssa-operands.c (operands_scanner::get_expr_operands):
16581 Do not look at COMPONENT_REF FIELD_DECLs TREE_THIS_VOLATILE
16582 to determine has_volatile_ops.
16584 2021-08-11 Eric Botcazou <ebotcazou@gcc.gnu.org>
16586 * cfgexpand.c (expand_used_vars): Reuse attribs local variable.
16588 2021-08-11 Jan Hubicka <hubicka@ucw.cz>
16589 Alexandre Oliva <oliva@adacore.com>
16591 * ipa-modref.c (modref_lattice::dump): Fix escape_point's min_flags
16593 (modref_lattice::merge_deref): Fix handling of indirect scape points.
16594 (update_escape_summary_1): Likewise.
16595 (update_escape_summary): Likewise.
16596 (ipa_merge_modref_summary_after_inlining): Likewise.
16598 2021-08-11 Richard Biener <rguenther@suse.de>
16600 PR middle-end/101858
16601 * fold-const.c (fold_binary_loc): Guard simplification
16602 of X < (cast) (1 << Y) to integer types.
16604 2021-08-11 Richard Biener <rguenther@suse.de>
16606 PR tree-optimization/101861
16607 * tree-vect-stmts.c (vectorizable_load): Fix error in
16608 previous change with regard to gather vectorization.
16610 2021-08-11 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
16613 * config/arm/arm_neon.h (vdup_n_s8): Replace call to builtin
16615 (vdup_n_s16): Likewise.
16616 (vdup_n_s32): Likewise.
16617 (vdup_n_s64): Likewise.
16618 (vdup_n_u8): Likewise.
16619 (vdup_n_u16): Likewise.
16620 (vdup_n_u32): Likewise.
16621 (vdup_n_u64): Likewise.
16622 (vdup_n_p8): Likewise.
16623 (vdup_n_p16): Likewise.
16624 (vdup_n_p64): Likewise.
16625 (vdup_n_f16): Likewise.
16626 (vdup_n_f32): Likewise.
16627 (vdupq_n_s8): Likewise.
16628 (vdupq_n_s16): Likewise.
16629 (vdupq_n_s32): Likewise.
16630 (vdupq_n_s64): Likewise.
16631 (vdupq_n_u8): Likewise.
16632 (vdupq_n_u16): Likewise.
16633 (vdupq_n_u32): Likewise.
16634 (vdupq_n_u64): Likewise.
16635 (vdupq_n_p8): Likewise.
16636 (vdupq_n_p16): Likewise.
16637 (vdupq_n_p64): Likewise.
16638 (vdupq_n_f16): Likewise.
16639 (vdupq_n_f32): Likewise.
16640 (vmov_n_s8): Replace call to builtin with call to corresponding
16642 (vmov_n_s16): Likewise.
16643 (vmov_n_s32): Likewise.
16644 (vmov_n_s64): Likewise.
16645 (vmov_n_u8): Likewise.
16646 (vmov_n_u16): Likewise.
16647 (vmov_n_u32): Likewise.
16648 (vmov_n_u64): Likewise.
16649 (vmov_n_p8): Likewise.
16650 (vmov_n_p16): Likewise.
16651 (vmov_n_f16): Likewise.
16652 (vmov_n_f32): Likewise.
16653 (vmovq_n_s8): Likewise.
16654 (vmovq_n_s16): Likewise.
16655 (vmovq_n_s32): Likewise.
16656 (vmovq_n_s64): Likewise.
16657 (vmovq_n_u8): Likewise.
16658 (vmovq_n_u16): Likewise.
16659 (vmovq_n_u32): Likewise.
16660 (vmovq_n_u64): Likewise.
16661 (vmovq_n_p8): Likewise.
16662 (vmovq_n_p16): Likewise.
16663 (vmovq_n_f16): Likewise.
16664 (vmovq_n_f32): Likewise.
16665 * config/arm/arm_neon_builtins.def: Remove entries for vdup_n.
16667 2021-08-11 liuhongt <hongtao.liu@intel.com>
16670 * config/i386/i386.md (ldexp<mode>3): Extend to vscalefs[sd]
16671 when TARGET_AVX512F and TARGET_SSE_MATH.
16673 2021-08-10 Jakub Jelinek <jakub@redhat.com>
16676 * config/i386/i386-expand.c (expand_vec_perm_even_odd): Return false
16677 for V32HImode if !TARGET_AVX512BW.
16678 (ix86_vectorize_vec_perm_const) <case E_V32HImode, case E_V64QImode>:
16679 If !TARGET_AVX512BW and TARGET_AVX512F and d.testing_p, don't fail
16680 early, but actually check the permutation.
16682 2021-08-10 Richard Biener <rguenther@suse.de>
16684 PR tree-optimization/101809
16685 * tree-vect-stmts.c (get_load_store_type): Allow emulated
16686 gathers with offset vector nunits being a constant multiple
16687 of the data vector nunits.
16688 (vect_get_gather_scatter_ops): Use the appropriate nunits
16689 for the offset vector defs.
16690 (vectorizable_store): Adjust call to
16691 vect_get_gather_scatter_ops.
16692 (vectorizable_load): Likewise. Handle the case of less
16693 offset vectors than data vectors.
16695 2021-08-10 Jakub Jelinek <jakub@redhat.com>
16698 * config/i386/sse.md (*avx512f_shuf_<shuffletype>64x2_1<mask_name>_1,
16699 *avx512f_shuf_<shuffletype>32x4_1<mask_name>_1): New define_insn
16702 2021-08-10 Richard Biener <rguenther@suse.de>
16704 PR tree-optimization/101801
16705 PR tree-optimization/101819
16706 * tree-vectorizer.h (vect_emulated_vector_p): Declare.
16707 * tree-vect-loop.c (vect_emulated_vector_p): New function.
16708 (vectorizable_reduction): Re-instantiate a check for emulated
16710 * tree-vect-stmts.c (vectorizable_shift): Likewise.
16711 (vectorizable_operation): Likewise. Cost emulated vector
16712 operations according to the scalar sequence synthesized by
16715 2021-08-10 Richard Biener <rguenther@suse.de>
16717 PR middle-end/101824
16718 * tree-nested.c (get_frame_field): Mark the COMPONENT_REF as
16719 volatile in case the variable was.
16721 2021-08-10 H.J. Lu <hjl.tools@gmail.com>
16724 * config/i386/constraints.md (BC): Document for integer SSE
16725 constant all bits set operand.
16726 (BF): New constraint for const floating-point all bits set
16728 * config/i386/i386.c (standard_sse_constant_p): Likewise.
16729 (standard_sse_constant_opcode): Likewise.
16730 * config/i386/sse.md (sseconstm1): New mode attribute.
16731 (mov<mode>_internal): Replace BC with <sseconstm1>.
16733 2021-08-10 liuhongt <hongtao.liu@intel.com>
16735 * config/i386/sse.md (cond_<insn><mode>): New expander.
16736 (VI248_AVX512VLBW): New mode iterator.
16737 * config/i386/predicates.md
16738 (nonimmediate_or_const_vec_dup_operand): New predicate.
16740 2021-08-09 Andrew MacLeod <amacleod@redhat.com>
16742 PR tree-optimization/101741
16743 * gimple-range-fold.cc (fold_using_range::range_of_builtin_call): Check
16744 type of parameter for toupper/tolower.
16746 2021-08-09 Martin Jambor <mjambor@suse.cz>
16748 PR testsuite/101654
16749 * ipa-prop.c (propagate_controlled_uses): Removed a spurious space.
16751 2021-08-09 Pat Haugen <pthaugen@linux.ibm.com>
16753 * config/rs6000/rs6000.c (is_load_insn1): Verify destination is a
16755 (is_store_insn1): Verify source is a register.
16757 2021-08-09 Uroš Bizjak <ubizjak@gmail.com>
16760 * config/i386/mmx.md (<any_logic:code>v2sf3):
16761 Rename from *mmx_<any_logic:code>v2sf3
16763 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
16765 * config/nvptx/nvptx.c: Cross-reference parts adapted in
16766 'gcc/omp-oacc-neuter-broadcast.cc'.
16767 * omp-low.c: Likewise.
16768 * omp-oacc-neuter-broadcast.cc: Cross-reference parts adapted from
16771 2021-08-09 Julian Brown <julian@codesourcery.com>
16772 Kwok Cheung Yeung <kcy@codesourcery.com>
16773 Thomas Schwinge <thomas@codesourcery.com>
16775 * config/gcn/gcn.c (gcn_init_builtins): Override decls for
16776 BUILT_IN_GOACC_SINGLE_START, BUILT_IN_GOACC_SINGLE_COPY_START,
16777 BUILT_IN_GOACC_SINGLE_COPY_END and BUILT_IN_GOACC_BARRIER.
16778 (gcn_goacc_validate_dims): Turn on worker partitioning unconditionally.
16779 (gcn_fork_join): Update comment.
16780 * config/gcn/gcn.opt (flag_worker_partitioning): Remove.
16781 (macc_experimental_workers): Remove unused option.
16783 2021-08-09 Julian Brown <julian@codesourcery.com>
16784 Nathan Sidwell <nathan@codesourcery.com> (via 'gcc/config/nvptx/nvptx.c' master)
16785 Kwok Cheung Yeung <kcy@codesourcery.com>
16786 Thomas Schwinge <thomas@codesourcery.com>
16788 * Makefile.in (OBJS): Add omp-oacc-neuter-broadcast.o.
16789 * doc/tm.texi.in (TARGET_GOACC_CREATE_WORKER_BROADCAST_RECORD):
16790 Add documentation hook.
16791 * doc/tm.texi: Regenerate.
16792 * omp-oacc-neuter-broadcast.cc: New file.
16793 * omp-builtins.def (BUILT_IN_GOACC_BARRIER)
16794 (BUILT_IN_GOACC_SINGLE_START, BUILT_IN_GOACC_SINGLE_COPY_START)
16795 (BUILT_IN_GOACC_SINGLE_COPY_END): New builtins.
16796 * passes.def (pass_omp_oacc_neuter_broadcast): Add pass.
16797 * target.def (goacc.create_worker_broadcast_record): Add target
16799 * tree-pass.h (make_pass_omp_oacc_neuter_broadcast): Add
16801 * config/gcn/gcn-protos.h (gcn_goacc_adjust_propagation_record):
16802 Rename prototype to...
16803 (gcn_goacc_create_worker_broadcast_record): ... this.
16804 * config/gcn/gcn-tree.c (gcn_goacc_adjust_propagation_record): Rename
16806 (gcn_goacc_create_worker_broadcast_record): ... this.
16807 * config/gcn/gcn.c (TARGET_GOACC_ADJUST_PROPAGATION_RECORD):
16809 (TARGET_GOACC_CREATE_WORKER_BROADCAST_RECORD): ... this.
16811 2021-08-09 Tejas Belagod <tejas.belagod@arm.com>
16814 * config/aarch64/aarch64-simd.md (vlshr<mode>3, vashr<mode>3): Use
16815 the right iterator.
16817 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
16819 * Makefile.in (GTFILES): Remove '$(srcdir)/omp-offload.c'.
16821 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
16823 * builtins.def (DEF_GOACC_BUILTIN, DEF_GOMP_BUILTIN): Don't
16824 consider '-foffload-abi'.
16825 * common.opt (-foffload-abi): Remove 'Var', 'Init'.
16826 * opts.c (common_handle_option) <-foffload-abi> [ACCEL_COMPILER]:
16829 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
16831 * optc-gen.awk: Sanity check that 'Init' doesn't appear without
16834 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
16836 * omp-builtins.def (BUILT_IN_ACC_GET_DEVICE_TYPE): Remove.
16838 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
16840 * doc/gty.texi (Files): Update.
16842 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
16844 * doc/gty.texi (Files): Fix GTY header file example.
16846 2021-08-09 Roger Sayle <roger@nextmovesoftware.com>
16848 * tree-ssa-ccp.c (value_mask_to_min_max): Helper function to
16849 determine the upper and lower bounds from a mask-value pair.
16850 (bit_value_unop) [ABS_EXPR, ABSU_EXPR]: Add support for
16851 absolute value and unsigned absolute value expressions.
16852 (bit_value_binop): Initialize *VAL's precision.
16853 [LT_EXPR, LE_EXPR]: Use value_mask_to_min_max to determine
16854 upper and lower bounds of operands. Add LE_EXPR/GE_EXPR
16855 support when the operands are unknown but potentially equal.
16856 [MIN_EXPR, MAX_EXPR]: Support minimum/maximum expressions.
16858 2021-08-09 Bin Cheng <bin.cheng@linux.alibaba.com>
16860 * config/aarch64/aarch64.md
16861 (*extend<SHORT:mode><GPI:mode>2_aarch64): Use %<GPI:w>0.
16863 2021-08-08 Sergei Trofimovich <siarheit@google.com>
16865 * lra-constraints.c: Fix s/otput/output/ typo.
16867 2021-08-06 Martin Sebor <msebor@redhat.com>
16869 * builtins.c (expand_builtin_memchr): Move to gimple-ssa-warn-access.cc.
16870 (expand_builtin_strcat): Same.
16871 (expand_builtin_stpncpy): Same.
16872 (expand_builtin_strncat): Same.
16873 (check_read_access): Same.
16874 (check_memop_access): Same.
16875 (expand_builtin_strlen): Move checks to gimple-ssa-warn-access.cc.
16876 (expand_builtin_strnlen): Same.
16877 (expand_builtin_memcpy): Same.
16878 (expand_builtin_memmove): Same.
16879 (expand_builtin_mempcpy): Same.
16880 (expand_builtin_strcpy): Same.
16881 (expand_builtin_strcpy_args): Same.
16882 (expand_builtin_stpcpy_1): Same.
16883 (expand_builtin_strncpy): Same.
16884 (expand_builtin_memset): Same.
16885 (expand_builtin_bzero): Same.
16886 (expand_builtin_strcmp): Same.
16887 (expand_builtin_strncmp): Same.
16888 (expand_builtin): Remove handlers.
16889 (fold_builtin_strlen): Add a comment.
16890 * builtins.h (check_access): Move to gimple-ssa-warn-access.cc.
16891 * calls.c (maybe_warn_nonstring_arg): Same.
16892 * diagnostic-spec.c (nowarn_spec_t::nowarn_spec_t): Add warning option.
16893 * gimple-fold.c (gimple_fold_builtin_strcpy): Pass argument to callee.
16894 (gimple_fold_builtin_stpcpy): Same.
16895 * gimple-ssa-warn-access.cc (has_location): New function.
16896 (get_location): Same.
16897 (get_callee_fndecl): Same.
16898 (call_nargs): Same.
16900 (warn_string_no_nul): Define.
16901 (unterminated_array): Same.
16902 (check_nul_terminated_array): Same.
16903 (maybe_warn_nonstring_arg): Same.
16904 (maybe_warn_for_bound): Same.
16905 (warn_for_access): Same.
16906 (check_access): Same.
16907 (check_memop_access): Same.
16908 (check_read_access): Same.
16909 (warn_dealloc_offset): Use helper functions.
16910 (maybe_emit_free_warning): Same.
16911 (class pass_waccess): Add members.
16912 (check_strcat): New function.
16913 (check_strncat): New function.
16914 (check_stxcpy): New function.
16915 (check_stxncpy): New function.
16916 (check_strncmp): New function.
16917 (pass_waccess::check_builtin): New function.
16918 (pass_waccess::check): Call it.
16919 * gimple-ssa-warn-access.h (warn_string_no_nul): Move here from
16921 (maybe_warn_for_bound): Same.
16922 (check_access): Same.
16923 (check_memop_access): Same.
16924 (check_read_access): Same.
16925 * pointer-query.h (struct access_data): Define a ctor overload.
16927 2021-08-06 Richard Biener <rguenther@suse.de>
16929 PR tree-optimization/101801
16930 * tree-vectorizer.h (vect_worthwhile_without_simd_p): Rename...
16931 (vect_can_vectorize_without_simd_p): ... to this.
16932 * tree-vect-loop.c (vect_worthwhile_without_simd_p): Rename...
16933 (vect_can_vectorize_without_simd_p): ... to this and fold
16934 in vect_min_worthwhile_factor.
16935 (vect_min_worthwhile_factor): Remove.
16936 (vectorizable_reduction): Adjust and remove the cost part.
16937 * tree-vect-stmts.c (vectorizable_shift): Likewise.
16938 (vectorizable_operation): Likewise.
16940 2021-08-06 Uroš Bizjak <ubizjak@gmail.com>
16943 * config/i386/i386.md (cmove reg-to-reg move elimination peephole2s):
16944 Add general_gr_operand predicate to operand 3.
16946 2021-08-06 Roger Sayle <roger@nextmovesoftware.com>
16948 * tree-ssa-phiopt.c (cond_removal_in_builtin_zero_pattern): Use
16949 CFN_BUILT_IN_CLRSB* instead of BUILT_IN_CLRSB* for consistency.
16951 2021-08-06 Tamar Christina <tamar.christina@arm.com>
16953 * config/aarch64/aarch64-sve-builtins.cc (register_svpattern,
16954 register_svprfop): Pass vec<> by pointer.
16955 * langhooks-def.h (lhd_simulate_enum_decl): Likewise.
16956 * langhooks.c (lhd_simulate_enum_decl): Likewise.
16957 * langhooks.h (struct lang_hooks_for_types): Likewise.
16959 2021-08-06 Jonathan Wright <jonathan.wright@arm.com>
16961 * config/aarch64/arm_neon.h (vst1_bf16_x2): Use
16962 __builtin_memcpy instead of constructing an additional
16963 __builtin_aarch64_simd_oi one vector at a time.
16964 (vst1q_bf16_x2): Likewise.
16965 (vst1_bf16_x3): Use __builtin_memcpy instead of constructing
16966 an additional __builtin_aarch64_simd_ci one vector at a time.
16967 (vst1q_bf16_x3): Likewise.
16968 (vst1_bf16_x4): Use __builtin_memcpy instead of a union.
16969 (vst1q_bf16_x4): Likewise.
16970 (vst2_bf16): Use __builtin_memcpy instead of constructing an
16971 additional __builtin_aarch64_simd_oi one vector at a time.
16972 (vst2q_bf16): Likewise.
16973 (vst3_bf16): Use __builtin_memcpy instead of constructing an
16974 additional __builtin_aarch64_simd_ci mode one vector at a
16976 (vst3q_bf16): Likewise.
16977 (vst4_bf16): Use __builtin_memcpy instead of constructing an
16978 additional __builtin_aarch64_simd_xi one vector at a time.
16979 (vst4q_bf16): Likewise.
16981 2021-08-06 Jonathan Wright <jonathan.wright@arm.com>
16983 * config/aarch64/arm_neon.h (__ST2_LANE_FUNC): Delete.
16984 (__ST2Q_LANE_FUNC): Delete.
16985 (vst2_lane_f16): Use __builtin_memcpy to copy vector
16986 structure instead of constructing __builtin_aarch64_simd_oi
16987 one vector at a time.
16988 (vst2_lane_f32): Likewise.
16989 (vst2_lane_f64): Likewise.
16990 (vst2_lane_p8): Likewise.
16991 (vst2_lane_p16): Likewise.
16992 (vst2_lane_p64): Likewise.
16993 (vst2_lane_s8): Likewise.
16994 (vst2_lane_s16): Likewise.
16995 (vst2_lane_s32): Likewise.
16996 (vst2_lane_s64): Likewise.
16997 (vst2_lane_u8): Likewise.
16998 (vst2_lane_u16): Likewise.
16999 (vst2_lane_u32): Likewise.
17000 (vst2_lane_u64): Likewise.
17001 (vst2_lane_bf16): Likewise.
17002 (vst2q_lane_f16): Use __builtin_memcpy to copy vector
17003 structure instead of using a union.
17004 (vst2q_lane_f32): Likewise.
17005 (vst2q_lane_f64): Likewise.
17006 (vst2q_lane_p8): Likewise.
17007 (vst2q_lane_p16): Likewise.
17008 (vst2q_lane_p64): Likewise.
17009 (vst2q_lane_s8): Likewise.
17010 (vst2q_lane_s16): Likewise.
17011 (vst2q_lane_s32): Likewise.
17012 (vst2q_lane_s64): Likewise.
17013 (vst2q_lane_u8): Likewise.
17014 (vst2q_lane_u16): Likewise.
17015 (vst2q_lane_u32): Likewise.
17016 (vst2q_lane_u64): Likewise.
17017 (vst2q_lane_bf16): Likewise.
17019 2021-08-06 Jonathan Wright <jonathan.wright@arm.com>
17021 * config/aarch64/arm_neon.h (__ST3_LANE_FUNC): Delete.
17022 (__ST3Q_LANE_FUNC): Delete.
17023 (vst3_lane_f16): Use __builtin_memcpy to copy vector
17024 structure instead of constructing __builtin_aarch64_simd_ci
17025 one vector at a time.
17026 (vst3_lane_f32): Likewise.
17027 (vst3_lane_f64): Likewise.
17028 (vst3_lane_p8): Likewise.
17029 (vst3_lane_p16): Likewise.
17030 (vst3_lane_p64): Likewise.
17031 (vst3_lane_s8): Likewise.
17032 (vst3_lane_s16): Likewise.
17033 (vst3_lane_s32): Likewise.
17034 (vst3_lane_s64): Likewise.
17035 (vst3_lane_u8): Likewise.
17036 (vst3_lane_u16): Likewise.
17037 (vst3_lane_u32): Likewise.
17038 (vst3_lane_u64): Likewise.
17039 (vst3_lane_bf16): Likewise.
17040 (vst3q_lane_f16): Use __builtin_memcpy to copy vector
17041 structure instead of using a union.
17042 (vst3q_lane_f32): Likewise.
17043 (vst3q_lane_f64): Likewise.
17044 (vst3q_lane_p8): Likewise.
17045 (vst3q_lane_p16): Likewise.
17046 (vst3q_lane_p64): Likewise.
17047 (vst3q_lane_s8): Likewise.
17048 (vst3q_lane_s16): Likewise.
17049 (vst3q_lane_s32): Likewise.
17050 (vst3q_lane_s64): Likewise.
17051 (vst3q_lane_u8): Likewise.
17052 (vst3q_lane_u16): Likewise.
17053 (vst3q_lane_u32): Likewise.
17054 (vst3q_lane_u64): Likewise.
17055 (vst3q_lane_bf16): Likewise.
17057 2021-08-06 Jonathan Wright <jonathan.wright@arm.com>
17059 * config/aarch64/arm_neon.h (__ST4_LANE_FUNC): Delete.
17060 (__ST4Q_LANE_FUNC): Delete.
17061 (vst4_lane_f16): Use __builtin_memcpy to copy vector
17062 structure instead of constructing __builtin_aarch64_simd_xi
17063 one vector at a time.
17064 (vst4_lane_f32): Likewise.
17065 (vst4_lane_f64): Likewise.
17066 (vst4_lane_p8): Likewise.
17067 (vst4_lane_p16): Likewise.
17068 (vst4_lane_p64): Likewise.
17069 (vst4_lane_s8): Likewise.
17070 (vst4_lane_s16): Likewise.
17071 (vst4_lane_s32): Likewise.
17072 (vst4_lane_s64): Likewise.
17073 (vst4_lane_u8): Likewise.
17074 (vst4_lane_u16): Likewise.
17075 (vst4_lane_u32): Likewise.
17076 (vst4_lane_u64): Likewise.
17077 (vst4_lane_bf16): Likewise.
17078 (vst4q_lane_f16): Use __builtin_memcpy to copy vector
17079 structure instead of using a union.
17080 (vst4q_lane_f32): Likewise.
17081 (vst4q_lane_f64): Likewise.
17082 (vst4q_lane_p8): Likewise.
17083 (vst4q_lane_p16): Likewise.
17084 (vst4q_lane_p64): Likewise.
17085 (vst4q_lane_s8): Likewise.
17086 (vst4q_lane_s16): Likewise.
17087 (vst4q_lane_s32): Likewise.
17088 (vst4q_lane_s64): Likewise.
17089 (vst4q_lane_u8): Likewise.
17090 (vst4q_lane_u16): Likewise.
17091 (vst4q_lane_u32): Likewise.
17092 (vst4q_lane_u64): Likewise.
17093 (vst4q_lane_bf16): Likewise.
17095 2021-08-06 Martin Liska <mliska@suse.cz>
17097 * config/rs6000/rs6000.c (rs6000_option_override_internal): When
17098 a target option is restored, it can have
17099 rs6000_long_double_type_size set to FLOAT_PRECISION_TFmode
17100 and error should not be emitted.
17102 2021-08-06 Sebastian Huber <sebastian.huber@embedded-brains.de>
17104 * gcov-io.h (gcov_write): Declare.
17105 * gcov-io.c (gcov_write): New.
17106 (gcov_write_counter): Remove.
17107 (gcov_write_tag_length): Likewise.
17108 (gcov_write_summary): Replace gcov_write_tag_length() with calls to
17109 gcov_write_unsigned().
17110 * doc/invoke.texi (fprofile-info-section): Mention
17111 __gcov_info_to_gdca().
17113 2021-08-06 Martin Sebor <msebor@redhat.com>
17115 * dominance.c (prune_bbs_to_update_dominators): Adjust by-value vec
17116 arguments to by-reference.
17117 (iterate_fix_dominators): Same.
17118 * dominance.h (iterate_fix_dominators): Same.
17119 * ipa-prop.h: Call auto_vec::to_vec_legacy.
17120 * tree-data-ref.c (dump_data_dependence_relation): Adjust by-value vec
17121 arguments to by-reference.
17122 (debug_data_dependence_relation): Same.
17123 (dump_data_dependence_relations): Same.
17124 * tree-data-ref.h (debug_data_dependence_relation): Same.
17125 (dump_data_dependence_relations): Same.
17126 * tree-predcom.c (dump_chains): Same.
17127 (initialize_root_vars_lm): Same.
17128 (determine_unroll_factor): Same.
17129 (replace_phis_by_defined_names): Same.
17130 (insert_init_seqs): Same.
17131 (pcom_worker::tree_predictive_commoning_loop): Call
17132 auto_vec::to_vec_legacy.
17133 * tree-ssa-pre.c (insert_into_preds_of_block): Adjust by-value vec
17134 arguments to by-reference.
17135 * tree-ssa-threadbackward.c (populate_worklist): Same.
17136 (back_threader::resolve_def): Same.
17137 * tree-vect-data-refs.c (vect_check_nonzero_value): Same.
17138 (vect_enhance_data_refs_alignment): Same.
17139 (vect_check_lower_bound): Same.
17140 (vect_prune_runtime_alias_test_list): Same.
17141 (vect_permute_store_chain): Same.
17142 * tree-vect-slp-patterns.c (vect_normalize_conj_loc): Same.
17143 * tree-vect-stmts.c (vect_create_vectorized_demotion_stmts): Same.
17144 * tree-vectorizer.h (vect_permute_store_chain): Same.
17145 * vec.c (test_init): New function.
17146 (vec_c_tests): Call new function.
17147 * vec.h (vec): Declare ctors, dtor, and assignment.
17148 (auto_vec::vec_to_legacy): New function.
17149 (vec::copy): Adjust initialization.
17151 2021-08-05 H.J. Lu <hjl.tools@gmail.com>
17154 * config/i386/i386.c (ix86_can_inline_p): Ignore MASK_80387 if
17155 callee only uses GPRs.
17156 * config/i386/ia32intrin.h: Revert commit 5463cee2770.
17157 * config/i386/serializeintrin.h: Revert commit 71958f740f1.
17158 * config/i386/x86gprintrin.h: Add
17159 #pragma GCC target("general-regs-only") and #pragma GCC pop_options
17160 to disable non-GPR ISAs.
17162 2021-08-05 Richard Sandiford <richard.sandiford@arm.com>
17164 PR middle-end/101787
17165 * doc/md.texi (cond_ashl, cond_ashr, cond_lshr): Document.
17167 2021-08-05 Richard Sandiford <richard.sandiford@arm.com>
17169 * tree-vectorizer.h (vect_is_store_elt_extraction, vect_is_reduction)
17170 (vect_reduc_type, vect_embedded_comparison_type, vect_comparison_type)
17171 (vect_is_extending_load, vect_is_integer_truncation): New functions,
17172 moved from aarch64.c but given different names.
17173 * config/aarch64/aarch64.c (aarch64_is_store_elt_extraction)
17174 (aarch64_is_reduction, aarch64_reduc_type)
17175 (aarch64_embedded_comparison_type, aarch64_comparison_type)
17176 (aarch64_extending_load_p, aarch64_integer_truncation_p): Delete
17177 in favor of the above. Update callers accordingly.
17179 2021-08-05 Richard Earnshaw <rearnsha@arm.com>
17182 * config/arm/arm-cpus.in (generic-armv7-a): Add quirk to suppress
17183 writing .cpu directive in asm output.
17184 * config/arm/arm.c (arm_identify_fpu_from_isa): New variable.
17185 (arm_last_printed_arch_string): Delete.
17186 (arm_last-printed_fpu_string): Delete.
17187 (arm_configure_build_target): If use of floating-point/SIMD is
17188 disabled, remove all fp/simd related features from the target ISA.
17189 (last_arm_targ_options): New variable.
17190 (arm_print_asm_arch_directives): Add new parameters. Change order
17191 of emitted directives and handle all cases here.
17192 (arm_file_start): Always call arm_print_asm_arch_directives, move
17193 all generation of .arch/.arch_extension here.
17194 (arm_file_end): Call arm_print_asm_arch.
17195 (arm_declare_function_name): Call arm_print_asm_arch_directives
17196 instead of printing .arch/.fpu directives directly.
17198 2021-08-05 Richard Earnshaw <rearnsha@arm.com>
17200 * config/arm/arm.c (arm_configure_build_target): Don't call
17201 arm_option_reconfigure_globals.
17202 (arm_option_restore): Call arm_option_reconfigure_globals after
17203 reconfiguring the target.
17204 * config/arm/arm-c.c (arm_pragma_target_parse): Likewise.
17206 2021-08-05 Richard Earnshaw <rearnsha@arm.com>
17208 * config/arm/arm.c (arm_configure_build_target): Ensure the target's
17209 arch_name is always set.
17211 2021-08-05 Jonathan Wright <jonathan.wright@arm.com>
17213 * config/aarch64/aarch64.c: Traverse RTL tree to prevent cost
17214 of vec_select high-half from being added into Neon subtract
17217 2021-08-05 Jonathan Wright <jonathan.wright@arm.com>
17219 * config/aarch64/aarch64.c: Traverse RTL tree to prevent cost
17220 of vec_select high-half from being added into Neon add cost.
17222 2021-08-05 Kewen Lin <linkw@linux.ibm.com>
17224 * cfgloop.h (loops_list::loops_list): Add one optional argument
17225 root and adjust accordingly, update loop tree walking and factor
17227 * cfgloop.c (loops_list::walk_loop_tree): ... this. New function.
17229 2021-08-05 Eric Botcazou <ebotcazou@gcc.gnu.org>
17231 PR tree-optimization/101626
17232 * tree-sra.c (propagate_subaccesses_from_rhs): Do not set the
17233 reverse scalar storage order on a pointer or vector component.
17235 2021-08-05 liuhongt <hongtao.liu@intel.com>
17237 * config/i386/sse.md (cond_<code><mode>): New expander.
17239 2021-08-05 liuhongt <hongtao.liu@intel.com>
17241 * config/i386/sse.md (cond_<code><mode>): New expander.
17243 2021-08-05 liuhongt <hongtao.liu@intel.com>
17245 * config/i386/sse.md (cond_<code><mode>): New expander.
17247 2021-08-04 David Malcolm <dmalcolm@redhat.com>
17250 * Makefile.in (ANALYZER_OBJS): Add analyzer/region-model-asm.o.
17252 2021-08-04 H.J. Lu <hjl.tools@gmail.com>
17255 * config/i386/i386.h (STORE_MAX_PIECES): Allow 16/32/64 bytes
17256 only if TARGET_INTER_UNIT_MOVES_TO_VEC is true.
17258 2021-08-04 H.J. Lu <hjl.tools@gmail.com>
17261 * config/i386/i386-expand.c (ix86_expand_vector_move): Call
17262 ix86_gen_scratch_sse_rtx to get a scratch SSE register to copy
17263 data with SSE register from one memory location to another.
17265 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
17267 * config/s390/s390.c (expand_perm_with_vpdi): New function.
17268 (vectorize_vec_perm_const_1): Call expand_perm_with_vpdi.
17269 * config/s390/vector.md (*vpdi1<mode>, @vpdi1<mode>): Enable a
17270 parameterized expander.
17271 (*vpdi4<mode>, @vpdi4<mode>): Likewise.
17273 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
17275 * config/s390/s390.c (MAX_VECT_LEN): Define macro.
17276 (struct expand_vec_perm_d): Define struct.
17277 (expand_perm_with_merge): New function.
17278 (vectorize_vec_perm_const_1): New function.
17279 (s390_vectorize_vec_perm_const): New function.
17280 (TARGET_VECTORIZE_VEC_PERM_CONST): Define target macro.
17282 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
17284 * config/s390/vector.md (V_HW_64): Remove mode iterator.
17285 (*vec_load_pair<mode>): Use V_HW_2 instead of V_HW_64.
17286 * config/s390/vx-builtins.md
17287 (vec_scatter_element<V_HW_2:mode>_SI): Use V_HW_2 instead of
17290 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
17292 * config/s390/s390.md (UNSPEC_VEC_PERMI): Remove constant
17294 * config/s390/vector.md (*vpdi1<mode>, *vpdi4<mode>): New pattern
17296 * config/s390/vx-builtins.md (*vec_permi<mode>): Emit generic rtx
17297 instead of an unspec.
17299 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
17301 * config/s390/s390-modes.def: Add more vector modes to support
17302 concatenation of two vectors.
17303 * config/s390/s390-protos.h (s390_expand_merge_perm_const): Add
17305 (s390_expand_merge): Likewise.
17306 * config/s390/s390.c (s390_expand_merge_perm_const): New function.
17307 (s390_expand_merge): New function.
17308 * config/s390/s390.md (UNSPEC_VEC_MERGEH, UNSPEC_VEC_MERGEL):
17309 Remove constant definitions.
17310 * config/s390/vector.md (V_HW_2): Add mode iterators.
17311 (VI_HW_4, V_HW_4): Rename VI_HW_4 to V_HW_4.
17312 (vec_2x_nelts, vec_2x_wide): New mode attributes.
17313 (*vmrhb, *vmrlb, *vmrhh, *vmrlh, *vmrhf, *vmrlf, *vmrhg, *vmrlg):
17314 New pattern definitions.
17315 (vec_widen_umult_lo_<mode>, vec_widen_umult_hi_<mode>)
17316 (vec_widen_smult_lo_<mode>, vec_widen_smult_hi_<mode>)
17317 (vec_unpacks_lo_v4sf, vec_unpacks_hi_v4sf, vec_unpacks_lo_v2df)
17318 (vec_unpacks_hi_v2df): Adjust expanders to emit non-unspec RTX for
17320 * config/s390/vx-builtins.md (V_HW_4): Remove mode iterator. Now
17322 (vec_mergeh<mode>, vec_mergel<mode>): Use s390_expand_merge to
17323 emit vec merge pattern.
17325 2021-08-04 Jonathan Wright <jonathan.wright@arm.com>
17327 * config/aarch64/aarch64.c (aarch64_strip_extend_vec_half):
17329 (aarch64_rtx_mult_cost): Traverse RTL tree to prevent cost of
17330 vec_select high-half from being added into Neon multiply
17332 * rtlanal.c (vec_series_highpart_p): Define.
17333 * rtlanal.h (vec_series_highpart_p): Declare.
17335 2021-08-04 Jonathan Wright <jonathan.wright@arm.com>
17337 * config/aarch64/aarch64.c (aarch64_strip_duplicate_vec_elt):
17339 (aarch64_rtx_mult_cost): Traverse RTL tree to prevent
17340 vec_select cost from being added into Neon multiply cost.
17342 2021-08-04 Richard Sandiford <richard.sandiford@arm.com>
17344 * tree-vect-loop.c (vect_better_loop_vinfo_p): Detect cases in
17345 which old_loop_vinfo is an epilogue loop that handles a constant
17346 number of iterations.
17348 2021-08-04 Richard Sandiford <richard.sandiford@arm.com>
17350 * tree-vect-loop.c (vect_analyze_loop): Print a dump message
17351 when a reanalyzed loop fails to be cheaper than the current
17354 2021-08-04 Richard Sandiford <richard.sandiford@arm.com>
17356 * config/aarch64/aarch64.c: Fix a typo.
17358 2021-08-04 Vincent Lefèvre <vincent-gcc@vinc17.net>
17360 PR gcov-profile/101773
17361 * gcov-io.c (gcov_close): Check return code of a fclose.
17363 2021-08-04 Bernd Edlinger <bernd.edlinger@hotmail.de>
17366 * dwarf2out.c (dwarf2out_assembly_start): Emit a dummy
17367 .file statement when needed.
17369 2021-08-04 Richard Biener <rguenther@suse.de>
17371 * tree-vect-data-refs.c (vect_check_gather_scatter):
17372 Include widening conversions only when the result is
17373 still handed by native gather or the current offset
17374 size not already matches the data size.
17375 Also succeed analysis in case there's no native support,
17376 noted by a IFN_LAST ifn and a NULL decl.
17377 (vect_analyze_data_refs): Always consider gathers.
17378 * tree-vect-patterns.c (vect_recog_gather_scatter_pattern):
17379 Test for no IFN gather rather than decl gather.
17380 * tree-vect-stmts.c (vect_model_load_cost): Pass in the
17381 gather-scatter info and cost emulated gathers accordingly.
17382 (vect_truncate_gather_scatter_offset): Properly test for
17384 (vect_use_strided_gather_scatters_p): Likewise.
17385 (get_load_store_type): Handle emulated gathers and its
17387 (vectorizable_load): Likewise. Emulate them by extracting
17388 scalar offsets, doing scalar loads and a vector construct.
17390 2021-08-04 H.J. Lu <hjl.tools@gmail.com>
17393 * expr.c (op_by_pieces_d::op_by_pieces_d): Add a max_pieces
17394 argument to set m_max_size.
17395 (move_by_pieces_d): Pass MOVE_MAX_PIECES to op_by_pieces_d.
17396 (store_by_pieces_d): Pass STORE_MAX_PIECES to op_by_pieces_d.
17397 (compare_by_pieces_d): Pass COMPARE_MAX_PIECES to op_by_pieces_d.
17399 2021-08-04 Roger Sayle <roger@nextmovesoftware.com>
17400 Marc Glisse <marc.glisse@inria.fr>
17402 * match.pd (bit_ior, bit_xor): Canonicalize (X*C1)|(X*C2) and
17403 (X*C1)^(X*C2) as X*(C1+C2), and related variants, using
17404 tree_nonzero_bits to ensure that operands are bit-wise disjoint.
17406 2021-08-04 Richard Biener <rguenther@suse.de>
17408 * tree-ssa-forwprop.c (pass_forwprop::execute): Split
17409 out code to decompose vector loads ...
17410 (optimize_vector_load): ... here. Generalize it to
17411 handle intermediate widening and TARGET_MEM_REF loads
17412 and apply it to loads with a supported vector mode as well.
17414 2021-08-04 Richard Biener <rguenther@suse.de>
17416 PR tree-optimization/101756
17417 * tree-vect-slp.c (vectorizable_bb_reduc_epilogue): Make sure
17418 the result of the reduction epilogue is compatible to the original
17421 2021-08-04 liuhongt <hongtao.liu@intel.com>
17424 * config/i386/i386.md (peephole2): Refine predicate from
17425 register_operand to general_reg_operand.
17427 2021-08-04 Aldy Hernandez <aldyh@redhat.com>
17429 * gimple-range-path.h (path_range_query::dump): Mark override.
17431 2021-08-04 Richard Biener <rguenther@suse.de>
17433 PR tree-optimization/101769
17434 * tree-tailcall.c (eliminate_tail_call): Add the created loop
17435 for the first recursion and return it via the new output parameter.
17436 (optimize_tail_call): Pass through new output param.
17437 (tree_optimize_tail_calls_1): After creating all latches,
17438 add the created loop to the loop tree. Do not mark loops for fixup.
17440 2021-08-04 Martin Liska <mliska@suse.cz>
17442 * doc/invoke.texi: Document threader-mode param.
17444 2021-08-04 liuhongt <hongtao.liu@intel.com>
17446 * config/i386/sse.md (cond_fma<mode>): New expander.
17447 (cond_fms<mode>): Ditto.
17448 (cond_fnma<mode>): Ditto.
17449 (cond_fnms<mode>): Ditto.
17451 2021-08-03 Segher Boessenkool <segher@kernel.crashing.org>
17453 * config/rs6000/vsx.md (*vsx_le_perm_store_<mode>): Use && instead of &.
17455 2021-08-03 Segher Boessenkool <segher@kernel.crashing.org>
17457 * config/rs6000/constraints.md: Remove "e" from the list of available
17458 constraint characters.
17460 2021-08-03 Eugene Rozenfeld <erozen@microsoft.com>
17462 PR gcov-profile/71672
17463 * auto-profile.c (afdo_indirect_call): Fix setup of the historgram value for indirect calls.
17465 2021-08-03 Paul A. Clarke <pc@us.ibm.com>
17467 * config/rs6000/smmintrin.h (_mm_minpos_epu16): New.
17469 2021-08-03 H.J. Lu <hjl.tools@gmail.com>
17471 * config/i386/i386.c (ix86_gen_scratch_sse_rtx): In 64-bit mode,
17472 try XMM31 to avoid vzeroupper.
17474 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
17476 * doc/invoke.texi: Document -mtune=neoverse-512tvb and
17477 -mcpu=neoverse-512tvb.
17478 * config/aarch64/aarch64-cores.def (neoverse-512tvb): New entry.
17479 * config/aarch64/aarch64-tune.md: Regenerate.
17480 * config/aarch64/aarch64.c (neoverse512tvb_sve_vector_cost)
17481 (neoverse512tvb_sve_issue_info, neoverse512tvb_vec_issue_info)
17482 (neoverse512tvb_vector_cost, neoverse512tvb_tunings): New structures.
17483 (aarch64_adjust_body_cost_sve): Handle -mtune=neoverse-512tvb.
17484 (aarch64_adjust_body_cost): Likewise.
17486 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
17488 * config/aarch64/aarch64.c (aarch64_add_stmt_cost): Only
17489 record issue information for operations that occur in the
17492 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
17494 * config/aarch64/aarch64.c (aarch64_multiply_add_p): Add a vec_flags
17495 parameter. Detect cases in which an Advanced SIMD MLA would almost
17496 certainly require a MOV.
17497 (aarch64_count_ops): Update accordingly.
17499 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
17501 * config/aarch64/aarch64.c (aarch64_is_store_elt_extraction): New
17502 function, split out from...
17503 (aarch64_detect_vector_stmt_subtype): ...here.
17504 (aarch64_add_stmt_cost): Treat extracting element 0 as free.
17506 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
17508 * config/aarch64/aarch64-protos.h (sve_vec_cost):
17509 Add gather_load_x32_cost and gather_load_x64_cost.
17510 * config/aarch64/aarch64.c (generic_sve_vector_cost)
17511 (a64fx_sve_vector_cost, neoversev1_sve_vector_cost): Update
17512 accordingly, using the values given by the scalar_load * number
17513 of elements calculation that we used previously.
17514 (aarch64_detect_vector_stmt_subtype): Use the new fields.
17516 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
17518 * config/aarch64/aarch64.c (aarch64_adjust_body_cost_sve): New
17519 function, split out from...
17520 (aarch64_adjust_body_cost): ...here.
17522 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
17524 * config/aarch64/fractional-cost.h: New file.
17525 * config/aarch64/aarch64.c: Include <algorithm> (indirectly)
17526 and cost_fraction.h.
17527 (vec_cost_fraction): New typedef.
17528 (aarch64_detect_scalar_stmt_subtype): Use it for statement costs.
17529 (aarch64_detect_vector_stmt_subtype): Likewise.
17530 (aarch64_sve_adjust_stmt_cost, aarch64_adjust_stmt_cost): Likewise.
17531 (aarch64_estimate_min_cycles_per_iter): Use vec_cost_fraction
17533 (aarch64_adjust_body_cost): Likewise.
17534 (aarch64_test_cost_fraction): New function.
17535 (aarch64_run_selftests): Call it.
17537 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
17539 * config/aarch64/aarch64-protos.h (tune_params::sve_width): Turn
17541 * config/aarch64/aarch64.c (aarch64_cmp_autovec_modes): Update
17543 (aarch64_estimated_poly_value): Likewise. Use the least significant
17544 set bit for the minimum and likely values. Use the most significant
17545 set bit for the maximum value.
17547 2021-08-03 liuhongt <hongtao.liu@intel.com>
17549 * config/i386/sse.md (cond_<insn><mode>): New expander.
17550 (cond_mul<mode>): Ditto.
17552 2021-08-03 Kewen Lin <linkw@linux.ibm.com>
17554 * tree-cfg.c (move_sese_region_to_fn): Fix typos on dloop.
17556 2021-08-03 liuhongt <hongtao.liu@intel.com>
17558 * config/i386/sse.md (cond_<insn><mode>):New expander.
17559 (cond_mul<mode>): Ditto.
17560 (cond_div<mode>): Ditto.
17562 2021-08-02 H.J. Lu <hjl.tools@gmail.com>
17564 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Also
17565 check stack_realign_needed for stack realignment.
17566 (ix86_legitimate_constant_p): Always allow CONST_WIDE_INT smaller
17567 than the largest integer supported by vector register.
17568 * config/i386/i386.h (MAX_MOVE_MAX): New. Set to 64.
17569 (MOVE_MAX): Set to bytes of the largest integer supported by
17571 (STORE_MAX_PIECES): New.
17573 2021-08-02 H.J. Lu <hjl.tools@gmail.com>
17575 * config/i386/i386-expand.c (ix86_expand_vector_move): Call
17576 ix86_gen_scratch_sse_rtx to get a scratch SSE register to copy
17577 data from one memory location to another.
17579 2021-08-02 H.J. Lu <hjl.tools@gmail.com>
17581 PR middle-end/90773
17582 * config/i386/i386.c (TARGET_GEN_MEMSET_SCRATCH_RTX): New.
17584 2021-08-02 Aldy Hernandez <aldyh@redhat.com>
17586 PR tree-optimization/101724
17587 * params.opt: Remove --param=threader-iterative.
17588 * tree-ssa-threadbackward.c (pass_thread_jumps::execute): Remove
17591 2021-08-02 Tom de Vries <tdevries@suse.de>
17593 PR middle-end/101665
17594 * doc/extend.texi (nonnull attribute): Improve documentation.
17596 2021-08-02 Andrew Pinski <apinski@marvell.com>
17598 PR rtl-optimization/101683
17599 * rtlanal.c (may_trap_p_1): Handle UNSIGNED_FIX.
17601 2021-08-02 Roger Sayle <roger@nextmovesoftware.com>
17603 * tree-ssa-phiopt.c (cond_removal_in_builtin_zero_pattern):
17604 Renamed from cond_removal_in_popcount_clz_ctz_pattern.
17605 Add support for BSWAP, FFS, PARITY and CLRSB builtins.
17606 (tree_ssa_phiop_worker): Update call to function above.
17608 2021-08-01 H.J. Lu <hjl.tools@gmail.com>
17611 * config/i386/i386.md (bsr_rex64_1_zext): New.
17612 (combine splitter for constant - clzll): Replace gen_bsr_rex64_1
17613 with gen_bsr_rex64_1_zext.
17615 2021-07-31 Jakub Jelinek <jakub@redhat.com>
17618 * config/i386/i386.md (bsr_rex64_1, bsr_1, bsr_zext_1): New
17619 define_insn patterns.
17620 (*bsr_rex64_2, *bsr_2): New define_insn_and_split patterns.
17621 Add combine splitters for constant - clz.
17622 (clz<mode>2): Use a temporary pseudo for bsr result.
17624 2021-07-30 Paul A. Clarke <pc@us.ibm.com>
17626 * config/rs6000/smmintrin.h (_mm_floor_pd, _mm_floor_ps,
17627 _mm_floor_sd, _mm_floor_ss): New.
17629 2021-07-30 Paul A. Clarke <pc@us.ibm.com>
17631 * config/rs6000/smmintrin.h (_mm_ceil_pd, _mm_ceil_ps,
17632 _mm_ceil_sd, _mm_ceil_ss): New.
17634 2021-07-30 Paul A. Clarke <pc@us.ibm.com>
17636 * config/rs6000/smmintrin.h (_mm_blend_pd, _mm_blendv_pd,
17637 _mm_blend_ps, _mm_blendv_ps): New.
17639 2021-07-30 Roger Sayle <roger@nextmovesoftware.com>
17640 Uroš Bizjak <ubizjak@gmail.com>
17642 * config/i386/i386.md (*dec_cmov<mode>): New define_insn_and_split
17643 to generate a conditional move using the carry flag after sub $1.
17644 (peephole2): Eliminate a register-to-register move by inverting
17645 the condition of a conditional move.
17647 2021-07-30 Hans-Peter Nilsson <hp@bitrange.com>
17649 * config/mmix/mmix.md ("call", "call_value", "*call_real")
17650 ("*call_value_real"): Don't generate rtx mentioning the generic
17651 operands 1 and 2 to "call", and similarly for "call_value".
17652 * config/mmix/mmix.c (mmix_print_operand_punct_valid_p)
17653 (mmix_print_operand): Use '!' instead of 'p'.
17655 2021-07-30 Hans-Peter Nilsson <hp@bitrange.com>
17657 * doc/md.texi (call): Correct information about operand 2.
17658 * config/mmix/mmix.md ("call", "call_value"): Remove fixed FIXMEs.
17660 2021-07-30 Andrew MacLeod <amacleod@redhat.com>
17662 * range-op.cc (operator_trunc_mod::wi_fold): Fold constants.
17664 2021-07-30 Andrew MacLeod <amacleod@redhat.com>
17666 * range-op.cc (operator_div::wi_fold): Return UNDEFINED for [0, 0] divisor.
17668 2021-07-30 Andrew MacLeod <amacleod@redhat.com>
17670 * gimple-range-cache.cc (*::set_bb_range): Change const basic_block to
17671 const_basic_block..
17672 (*::get_bb_range): Ditto.
17673 (*::bb_range_p): Ditto.
17674 * gimple-range-cache.h: Change prototypes.
17676 2021-07-30 H.J. Lu <hjl.tools@gmail.com>
17678 PR middle-end/90773
17679 * builtins.c (builtin_memcpy_read_str): Change the mode argument
17680 from scalar_int_mode to fixed_size_mode.
17681 (builtin_strncpy_read_str): Likewise.
17682 (gen_memset_value_from_prev): New function.
17683 (builtin_memset_read_str): Change the mode argument from
17684 scalar_int_mode to fixed_size_mode. Use gen_memset_value_from_prev
17685 and support CONST_VECTOR.
17686 (builtin_memset_gen_str): Likewise.
17687 (try_store_by_multiple_pieces): Use by_pieces_constfn to declare
17689 * builtins.h (builtin_strncpy_read_str): Replace scalar_int_mode
17690 with fixed_size_mode.
17691 (builtin_memset_read_str): Likewise.
17692 * expr.c (widest_int_mode_for_size): Renamed to ...
17693 (widest_fixed_size_mode_for_size): Add a bool argument to
17694 indicate if QI vector mode can be used.
17695 (by_pieces_ninsns): Call widest_fixed_size_mode_for_size
17696 instead of widest_int_mode_for_size.
17697 (pieces_addr::adjust): Change the mode argument from
17698 scalar_int_mode to fixed_size_mode.
17699 (op_by_pieces_d): Make m_len read-only. Add a bool member,
17700 m_qi_vector_mode, to indicate that QI vector mode can be used.
17701 (op_by_pieces_d::op_by_pieces_d): Add a bool argument to
17702 initialize m_qi_vector_mode. Call widest_fixed_size_mode_for_size
17703 instead of widest_int_mode_for_size.
17704 (op_by_pieces_d::get_usable_mode): Change the mode argument from
17705 scalar_int_mode to fixed_size_mode. Call
17706 widest_fixed_size_mode_for_size instead of
17707 widest_int_mode_for_size.
17708 (op_by_pieces_d::smallest_fixed_size_mode_for_size): New member
17709 function to return the smallest integer or QI vector mode.
17710 (op_by_pieces_d::run): Call widest_fixed_size_mode_for_size
17711 instead of widest_int_mode_for_size. Call
17712 smallest_fixed_size_mode_for_size instead of
17713 smallest_int_mode_for_size.
17714 (store_by_pieces_d::store_by_pieces_d): Add a bool argument to
17715 indicate that QI vector mode can be used and pass it to
17716 op_by_pieces_d::op_by_pieces_d.
17717 (can_store_by_pieces): Call widest_fixed_size_mode_for_size
17718 instead of widest_int_mode_for_size. Pass memsetp to
17719 widest_fixed_size_mode_for_size to support QI vector mode.
17720 Allow all CONST_VECTORs for memset if vec_duplicate is supported.
17721 (store_by_pieces): Pass memsetp to
17722 store_by_pieces_d::store_by_pieces_d.
17723 (clear_by_pieces_1): Removed.
17724 (clear_by_pieces): Replace clear_by_pieces_1 with
17725 builtin_memset_read_str and pass true to store_by_pieces_d to
17726 support vector mode broadcast.
17727 (string_cst_read_str): Change the mode argument from
17728 scalar_int_mode to fixed_size_mode.
17729 * expr.h (by_pieces_constfn): Change scalar_int_mode to
17731 (by_pieces_prev): Likewise.
17732 * rtl.h (lowpart_subreg_regno): New.
17733 * rtlanal.c (lowpart_subreg_regno): New. A wrapper around
17734 simplify_subreg_regno.
17735 * target.def (gen_memset_scratch_rtx): New hook.
17736 * doc/tm.texi.in: Add TARGET_GEN_MEMSET_SCRATCH_RTX.
17737 * doc/tm.texi: Regenerated.
17739 2021-07-30 Xi Ruoyao <xry111@mengyan1223.wang>
17742 * config/mips/mips.c (mips_atomic_assign_expand_fenv): Use
17743 TARGET_EXPR instead of MODIFY_EXPR.
17745 2021-07-30 Xi Ruoyao <xry111@mengyan1223.wang>
17748 * config/mips/mips-protos.h (mips_expand_vec_cmp_expr): Declare.
17749 * config/mips/mips.c (mips_expand_vec_cmp_expr): New function.
17750 * config/mips/mips-msa.md (vec_cmp<MSA:mode><mode_i>): New
17752 (vec_cmpu<IMSA:mode><mode_i>): New expander.
17754 2021-07-30 H.J. Lu <hjl.tools@gmail.com>
17757 * config/i386/i386-options.c (ix86_option_override_internal):
17758 Don't enable LZCNT/POPCNT if they have been disabled explicitly.
17760 2021-07-30 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
17763 * config/arm/arm_neon.h (vld1_p64): Replace call to builtin by
17764 explicitly dereferencing __a.
17765 (vld1_s64): Likewise.
17766 (vld1_u64): Likewise.
17767 * config/arm/arm_neon_builtins.def (vld1): Remove entry for di
17768 and change to VAR13.
17770 2021-07-30 Aldy Hernandez <aldyh@redhat.com>
17772 * gimple-loop-versioning.cc (lv_dom_walker::lv_dom_walker): Remove
17773 use of m_range_analyzer.
17774 (loop_versioning::lv_dom_walker::before_dom_children): Same.
17775 (loop_versioning::lv_dom_walker::after_dom_children): Remove.
17776 (loop_versioning::prune_loop_conditions): Replace vr_values use
17777 with range_query interface.
17778 (pass_loop_versioning::execute): Use ranger.
17780 2021-07-30 Xi Ruoyao <xry111@mengyan1223.wang>
17783 * ipa-devirt.c (ipa_odr_read_section): Compare the precision of
17784 enum values, and emit a warning if they mismatch.
17786 2021-07-30 Kewen Lin <linkw@linux.ibm.com>
17788 * cfgloop.h (as_const): New function.
17789 (class loop_iterator): Rename to ...
17790 (class loops_list): ... this.
17791 (loop_iterator::next): Rename to ...
17792 (loops_list::Iter::fill_curr_loop): ... this and adjust.
17793 (loop_iterator::loop_iterator): Rename to ...
17794 (loops_list::loops_list): ... this and adjust.
17795 (loops_list::Iter): New class.
17796 (loops_list::iterator): New type.
17797 (loops_list::const_iterator): New type.
17798 (loops_list::begin): New function.
17799 (loops_list::end): Likewise.
17800 (loops_list::begin const): Likewise.
17801 (loops_list::end const): Likewise.
17802 (FOR_EACH_LOOP): Remove.
17803 (FOR_EACH_LOOP_FN): Remove.
17804 * cfgloop.c (flow_loops_dump): Adjust FOR_EACH_LOOP* with range-based
17805 for loop with loops_list instance.
17806 (sort_sibling_loops): Likewise.
17807 (disambiguate_loops_with_multiple_latches): Likewise.
17808 (verify_loop_structure): Likewise.
17809 * cfgloopmanip.c (create_preheaders): Likewise.
17810 (force_single_succ_latches): Likewise.
17811 * config/aarch64/falkor-tag-collision-avoidance.c
17812 (execute_tag_collision_avoidance): Likewise.
17813 * config/mn10300/mn10300.c (mn10300_scan_for_setlb_lcc): Likewise.
17814 * config/s390/s390.c (s390_adjust_loops): Likewise.
17815 * doc/loop.texi: Likewise.
17816 * gimple-loop-interchange.cc (pass_linterchange::execute): Likewise.
17817 * gimple-loop-jam.c (tree_loop_unroll_and_jam): Likewise.
17818 * gimple-loop-versioning.cc (loop_versioning::analyze_blocks): Likewise.
17819 (loop_versioning::make_versioning_decisions): Likewise.
17820 * gimple-ssa-split-paths.c (split_paths): Likewise.
17821 * graphite-isl-ast-to-gimple.c (graphite_regenerate_ast_isl): Likewise.
17822 * graphite.c (canonicalize_loop_form): Likewise.
17823 (graphite_transform_loops): Likewise.
17824 * ipa-fnsummary.c (analyze_function_body): Likewise.
17825 * ipa-pure-const.c (analyze_function): Likewise.
17826 * loop-doloop.c (doloop_optimize_loops): Likewise.
17827 * loop-init.c (loop_optimizer_finalize): Likewise.
17828 (fix_loop_structure): Likewise.
17829 * loop-invariant.c (calculate_loop_reg_pressure): Likewise.
17830 (move_loop_invariants): Likewise.
17831 * loop-unroll.c (decide_unrolling): Likewise.
17832 (unroll_loops): Likewise.
17833 * modulo-sched.c (sms_schedule): Likewise.
17834 * predict.c (predict_loops): Likewise.
17835 (pass_profile::execute): Likewise.
17836 * profile.c (branch_prob): Likewise.
17837 * sel-sched-ir.c (sel_finish_pipelining): Likewise.
17838 (sel_find_rgns): Likewise.
17839 * tree-cfg.c (replace_loop_annotate): Likewise.
17840 (replace_uses_by): Likewise.
17841 (move_sese_region_to_fn): Likewise.
17842 * tree-if-conv.c (pass_if_conversion::execute): Likewise.
17843 * tree-loop-distribution.c (loop_distribution::execute): Likewise.
17844 * tree-parloops.c (parallelize_loops): Likewise.
17845 * tree-predcom.c (tree_predictive_commoning): Likewise.
17846 * tree-scalar-evolution.c (scev_initialize): Likewise.
17847 (scev_reset): Likewise.
17848 * tree-ssa-dce.c (find_obviously_necessary_stmts): Likewise.
17849 * tree-ssa-live.c (remove_unused_locals): Likewise.
17850 * tree-ssa-loop-ch.c (ch_base::copy_headers): Likewise.
17851 * tree-ssa-loop-im.c (analyze_memory_references): Likewise.
17852 (tree_ssa_lim_initialize): Likewise.
17853 * tree-ssa-loop-ivcanon.c (canonicalize_induction_variables): Likewise.
17854 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize): Likewise.
17855 * tree-ssa-loop-manip.c (get_loops_exits): Likewise.
17856 * tree-ssa-loop-niter.c (estimate_numbers_of_iterations): Likewise.
17857 (free_numbers_of_iterations_estimates): Likewise.
17858 * tree-ssa-loop-prefetch.c (tree_ssa_prefetch_arrays): Likewise.
17859 * tree-ssa-loop-split.c (tree_ssa_split_loops): Likewise.
17860 * tree-ssa-loop-unswitch.c (tree_ssa_unswitch_loops): Likewise.
17861 * tree-ssa-loop.c (gate_oacc_kernels): Likewise.
17862 (pass_scev_cprop::execute): Likewise.
17863 * tree-ssa-propagate.c (clean_up_loop_closed_phi): Likewise.
17864 * tree-ssa-sccvn.c (do_rpo_vn): Likewise.
17865 * tree-ssa-threadupdate.c
17866 (jump_thread_path_registry::thread_through_all_blocks): Likewise.
17867 * tree-vectorizer.c (vectorize_loops): Likewise.
17868 * tree-vrp.c (vrp_asserts::find_assert_locations): Likewise.
17870 2021-07-29 Hans-Peter Nilsson <hp@bitrange.com>
17872 * config/mmix/mmix.c (mmix_function_arg_1): Avoid
17873 generating a VOIDmode register for e.g the
17874 function_arg_info::end_marker.
17876 2021-07-29 Jeff Law <jeffreyalaw@gmail.com>
17878 * config/h8300/h8300-modes.def: Add CCZ, CCV and CCC, drop CCZNV.
17879 * config/h8300/h8300.md (H8cc mode iterator): Add CCZ.
17880 (cc mode_attr): Similarly.
17881 (ccz subst_attr): Similarly.
17882 * config/h8300/jumpcall.md: Add new patterns for branch-on-bit.
17883 * config/h8300/testcompare.md: Remove various cc0 based patterns
17884 that had been commented out. Add pattern to set CCZ from a bit
17887 2021-07-29 Thomas Schwinge <thomas@codesourcery.com>
17888 Julian Brown <julian@codesourcery.com>
17889 Kwok Cheung Yeung <kcy@codesourcery.com>
17891 * omp-offload.c (oacc_loop_xform_head_tail, oacc_loop_process):
17892 'update_stmt' after modification.
17893 (pass_oacc_loop_designation): New function, extracted out of...
17894 (pass_oacc_device_lower): ... this.
17895 (pass_data_oacc_loop_designation, pass_oacc_loop_designation)
17896 (make_pass_oacc_loop_designation): New
17897 * passes.def: Add it.
17898 * tree-parloops.c (create_parallel_loop): Adjust.
17899 * tree-pass.h (make_pass_oacc_loop_designation): New.
17901 2021-07-29 Aldy Hernandez <aldyh@redhat.com>
17903 * flag-types.h (enum threader_mode): New.
17904 * params.opt: Add entry for --param=threader-mode.
17905 * tree-ssa-threadbackward.c (THREADER_ITERATIVE_MODE): New.
17906 (class back_threader): New.
17907 (back_threader::back_threader): New.
17908 (back_threader::~back_threader): New.
17909 (back_threader::maybe_register_path): New.
17910 (back_threader::find_taken_edge): New.
17911 (back_threader::find_taken_edge_switch): New.
17912 (back_threader::find_taken_edge_cond): New.
17913 (back_threader::resolve_def): New.
17914 (back_threader::resolve_phi): New.
17915 (back_threader::find_paths_to_names): New.
17916 (back_threader::find_paths): New.
17919 (thread_jumps::find_jump_threads_backwards): Call ranger threader.
17920 (thread_jumps::find_jump_threads_backwards_with_ranger): New.
17921 (pass_thread_jumps::execute): Abstract out code...
17922 (try_thread_blocks): ...here.
17923 * tree-ssa-threadedge.c (jump_threader::thread_outgoing_edges):
17924 Abstract out threading candidate code to...
17925 (single_succ_to_potentially_threadable_block): ...here.
17926 * tree-ssa-threadedge.h (single_succ_to_potentially_threadable_block):
17928 * tree-ssa-threadupdate.c (register_jump_thread): Return boolean.
17929 * tree-ssa-threadupdate.h (class jump_thread_path_registry):
17930 Return bool from register_jump_thread.
17932 2021-07-29 Andreas Krebbel <krebbel@linux.ibm.com>
17934 * target.def: in0 and in1 do not need to be registers.
17935 * doc/tm.texi: Regenerate.
17937 2021-07-29 liuhongt <hongtao.liu@intel.com>
17940 * config/i386/i386.c (ix86_widen_mult_cost): New function.
17941 (ix86_add_stmt_cost): Use ix86_widen_mult_cost for
17944 2021-07-29 Jiufu Guo <guojiufu@linux.ibm.com>
17947 * config/rs6000/rs6000.c (TARGET_PREFERRED_DOLOOP_MODE): New hook.
17948 (rs6000_preferred_doloop_mode): New hook.
17949 * doc/tm.texi: Regenerate.
17950 * doc/tm.texi.in: Add hook preferred_doloop_mode.
17951 * target.def (preferred_doloop_mode): New hook.
17952 * targhooks.c (default_preferred_doloop_mode): New hook.
17953 * targhooks.h (default_preferred_doloop_mode): New hook.
17954 * tree-ssa-loop-ivopts.c (compute_doloop_base_on_mode): New function.
17955 (add_iv_candidate_for_doloop): Call targetm.preferred_doloop_mode
17956 and compute_doloop_base_on_mode.
17958 2021-07-28 Martin Sebor <msebor@redhat.com>
17960 PR middle-end/101494
17961 * tree-ssa-uninit.c (maybe_warn_operand): Correct object offset
17962 and size computation.
17964 2021-07-28 Martin Sebor <msebor@redhat.com>
17966 PR middle-end/101601
17967 * gimple-array-bounds.cc (array_bounds_checker::check_mem_ref): Remove
17969 Handle pointers to functions.
17971 2021-07-28 Martin Sebor <msebor@redhat.com>
17973 * Makefile.in (OBJS): Add gimple-ssa-warn-access.o and pointer-query.o.
17974 * attribs.h (fndecl_dealloc_argno): Move fndecl_dealloc_argno to tree.h.
17975 * builtins.c (compute_objsize_r): Move to pointer-query.cc.
17976 (access_ref::access_ref): Same.
17977 (access_ref::phi): Same.
17978 (access_ref::get_ref): Same.
17979 (access_ref::size_remaining): Same.
17980 (access_ref::offset_in_range): Same.
17981 (access_ref::add_offset): Same.
17982 (access_ref::inform_access): Same.
17983 (ssa_name_limit_t::visit_phi): Same.
17984 (ssa_name_limit_t::leave_phi): Same.
17985 (ssa_name_limit_t::next): Same.
17986 (ssa_name_limit_t::next_phi): Same.
17987 (ssa_name_limit_t::~ssa_name_limit_t): Same.
17988 (pointer_query::pointer_query): Same.
17989 (pointer_query::get_ref): Same.
17990 (pointer_query::put_ref): Same.
17991 (pointer_query::flush_cache): Same.
17992 (warn_string_no_nul): Move to gimple-ssa-warn-access.cc.
17993 (check_nul_terminated_array): Same.
17994 (unterminated_array): Same.
17995 (maybe_warn_for_bound): Same.
17996 (check_read_access): Same.
17997 (warn_for_access): Same.
17998 (get_size_range): Same.
17999 (check_access): Same.
18000 (gimple_call_alloc_size): Move to tree.c.
18001 (gimple_parm_array_size): Move to pointer-query.cc.
18002 (get_offset_range): Same.
18003 (gimple_call_return_array): Same.
18004 (handle_min_max_size): Same.
18005 (handle_array_ref): Same.
18006 (handle_mem_ref): Same.
18007 (compute_objsize): Same.
18008 (gimple_call_alloc_p): Move to gimple-ssa-warn-access.cc.
18009 (call_dealloc_argno): Same.
18010 (fndecl_dealloc_argno): Same.
18011 (new_delete_mismatch_p): Same.
18012 (matching_alloc_calls_p): Same.
18013 (warn_dealloc_offset): Same.
18014 (maybe_emit_free_warning): Same.
18015 * builtins.h (check_nul_terminated_array): Move to
18016 gimple-ssa-warn-access.h.
18017 (check_nul_terminated_array): Same.
18018 (warn_string_no_nul): Same.
18019 (unterminated_array): Same.
18020 (class ssa_name_limit_t): Same.
18021 (class pointer_query): Same.
18022 (struct access_ref): Same.
18023 (class range_query): Same.
18024 (struct access_data): Same.
18025 (gimple_call_alloc_size): Same.
18026 (gimple_parm_array_size): Same.
18027 (compute_objsize): Same.
18028 (class access_data): Same.
18029 (maybe_emit_free_warning): Same.
18030 * calls.c (initialize_argument_information): Remove call to
18031 maybe_emit_free_warning.
18032 * gimple-array-bounds.cc: Include new header..
18033 * gimple-fold.c: Same.
18034 * gimple-ssa-sprintf.c: Same.
18035 * gimple-ssa-warn-restrict.c: Same.
18036 * passes.def: Add pass_warn_access.
18037 * tree-pass.h (make_pass_warn_access): Declare.
18038 * tree-ssa-strlen.c: Include new headers.
18039 * tree.c (fndecl_dealloc_argno): Move here from builtins.c.
18040 * tree.h (fndecl_dealloc_argno): Move here from attribs.h.
18041 * gimple-ssa-warn-access.cc: New file.
18042 * gimple-ssa-warn-access.h: New file.
18043 * pointer-query.cc: New file.
18044 * pointer-query.h: New file.
18046 2021-07-28 Jakub Jelinek <jakub@redhat.com>
18048 PR middle-end/101624
18049 * ubsan.c (maybe_instrument_pointer_overflow,
18050 instrument_object_size): Only test DECL_REGISTER on VAR_DECLs,
18051 PARM_DECLs or RESULT_DECLs.
18052 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Likewise.
18054 2021-07-28 Jakub Jelinek <jakub@redhat.com>
18056 PR middle-end/101642
18057 * match.pd (bswap16 (x) == bswap16 (y)): Cast both operands
18058 to type of bswap16 for comparison.
18059 (bswap16 (x) == cst): Cast bswap16 operand to type of cst.
18061 2021-07-28 Richard Biener <rguenther@suse.de>
18063 PR tree-optimization/101615
18064 * tree-vect-slp.c (vect_optimize_slp): Materialize permutes
18065 at CTOR SLP graph entries.
18067 2021-07-28 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18069 * config/aarch64/aarch64.md (*extend<SHORT:mode><GPI:mode>2_aarch64):
18070 Add "r,w" alternative.
18072 2021-07-28 H.J. Lu <hjl.tools@gmail.com>
18075 * config/i386/i386.c (ix86_avx_u128_mode_needed): Don't set
18076 AVX_U128_DIRTY when all bits are zero.
18078 2021-07-28 Richard Biener <rguenther@suse.de>
18080 PR tree-optimization/101615
18081 * tree-vect-slp.c (vect_optimize_slp): Pre-existing vector
18082 external nodes cannot be permuted so make them perm_out 0.
18084 2021-07-28 Andrew Stubbs <ams@codesourcery.com>
18087 * config.in: Regenerate.
18088 * config/gcn/gcn-hsa.h (A_FIJI): New define.
18089 (A_900): New define.
18090 (A_906): New define.
18091 (A_908): New define.
18092 (ASM_SPEC): Use A_FIJI, A_900, A_906 and A_908.
18093 * config/gcn/gcn.c (output_file_start): Adjust attributes according
18094 to the assembler capabilities.
18095 * config/gcn/mkoffload.c (main): Likewise.
18096 * configure: Regenerate.
18097 * configure.ac: Add tests for LLVM assembler attribute features.
18099 2021-07-28 Andrew MacLeod <amacleod@redhat.com>
18101 * gimple-range-gori.cc (gori_compute::outgoing_edge_range_p): Check for
18102 cond_false and cond_true on branches.
18104 2021-07-28 Bin Cheng <bin.cheng@linux.alibaba.com>
18106 * config/aarch64/aarch64.c (aarch64_gen_adjusted_ldpstp): use
18109 2021-07-28 Bin Cheng <bin.cheng@linux.alibaba.com>
18111 * alias.c (init_alias_analysis): Don't skip prologue/epilogue.
18113 2021-07-28 Jakub Jelinek <jakub@redhat.com>
18116 * config/i386/sse.md (vashr<mode>3): Split into vashrv8di3 expander
18117 and vashrv4di3 expander, where the latter requires just TARGET_AVX2
18118 and has special !TARGET_AVX512VL expansion.
18119 (vashrv2di3<mask_name>): Rename to ...
18120 (vashrv2di3): ... this. Change condition to TARGET_XOP || TARGET_AVX2
18121 and add special !TARGET_XOP && !TARGET_AVX512VL expansion.
18123 2021-07-28 Martin Uecker <muecker@gwdg.de>
18125 * calls.c (maybe_warn_rdwr_sizes): Correct argument
18126 numbers in warning that were switched.
18128 2021-07-28 Kewen Lin <linkw@linux.ibm.com>
18130 PR tree-optimization/101596
18131 * tree-vect-patterns.c (vect_recog_mulhs_pattern): Fix wrong check
18132 by using new_type's precision instead.
18134 2021-07-28 liuhongt <hongtao.liu@intel.com>
18137 * config/i386/i386.h (processor_costs): Add new member
18139 * config/i386/x86-tune-costs.h (ix86_size_cost, i386_cost,
18140 i486_cost, pentium_cost, lakemont_cost, pentiumpro_cost,
18141 geode_cost, k6_cost, athlon_cost, k8_cost, amdfam10_cost,
18142 bdver_cost, znver1_cost, znver2_cost, znver3_cost,
18143 btver1_cost, btver2_cost, btver3_cost, pentium4_cost,
18144 nocona_cost, atom_cost, atom_cost, slm_cost, intel_cost,
18145 generic_cost, core_cost): Initialize integer_to_sse same value
18147 (skylake_cost): Initialize integer_to_sse twice as much as sse_op.
18148 * config/i386/i386.c (ix86_builtin_vectorization_cost):
18149 Use integer_to_sse instead of sse_op to calculate the cost of
18152 2021-07-27 Bill Schmidt <wschmidt@linux.ibm.com>
18154 * config/rs6000/rs6000-gen-builtins.c (write_ovld_static_init): New
18156 (write_init_file): Call write_ovld_static_init.
18158 2021-07-27 Bill Schmidt <wschmidt@linux.ibm.com>
18160 * config/rs6000/rs6000-gen-builtins.c (write_bif_static_init): New
18162 (write_init_file): Call write_bif_static_init.
18164 2021-07-27 Bill Schmidt <wschmidt@linux.ibm.com>
18166 * config/rs6000/rs6000-gen-builtins.c (typemap): New struct.
18167 (TYPE_MAP_SIZE): New macro.
18168 (type_map): New initialized variable.
18169 (typemap_cmp): New function.
18170 (write_type_node): Likewise.
18171 (write_fntype_init): Implement.
18173 2021-07-27 Martin Sebor <msebor@redhat.com>
18175 PR tree-optimization/101584
18176 * tree-ssa-uninit.c (builtin_call_nomodifying_p): New function.
18177 (check_defs): Call it.
18179 2021-07-27 Aldy Hernandez <aldyh@redhat.com>
18181 * tree-ssa-dom.c (dom_jump_threader_simplifier):
18182 Put avail_exprs_stack in the class, instead of passing it to
18183 jump_threader_simplifier.
18184 (dom_jump_threader_simplifier::simplify): Add state argument.
18185 (dom_opt_dom_walker): Add state.
18186 (pass_dominator::execute): Pass state to threader.
18187 (dom_opt_dom_walker::before_dom_children): Use state.
18188 * tree-ssa-threadedge.c (jump_threader::jump_threader): Replace
18189 arguments by state.
18190 (jump_threader::record_temporary_equivalences_from_phis):
18191 Register equivalences through the state variable.
18192 (jump_threader::record_temporary_equivalences_from_stmts_at_dest):
18193 Record ranges in a statement through the state variable.
18194 (jump_threader::simplify_control_stmt_condition): Pass state to
18196 (jump_threader::simplify_control_stmt_condition_1): Same.
18197 (jump_threader::thread_around_empty_blocks): Remove obsolete
18199 (jump_threader::thread_through_normal_block): Record equivalences
18200 on edge through the state variable.
18201 (jump_threader::thread_across_edge): Abstract state pushing.
18202 (jt_state::jt_state): New.
18203 (jt_state::push): New.
18204 (jt_state::pop): New.
18205 (jt_state::register_equiv): New.
18206 (jt_state::record_ranges_from_stmt): New.
18207 (jt_state::register_equivs_on_edge): New.
18208 (jump_threader_simplifier::jump_threader_simplifier): Move from
18210 (jump_threader_simplifier::simplify): Add state argument.
18211 * tree-ssa-threadedge.h (class jt_state): New.
18212 (class jump_threader): Add state to constructor.
18213 (class jump_threader_simplifier): Add state to simplify. Remove
18214 avail_exprs_stack from class.
18215 * tree-vrp.c (vrp_jump_threader_simplifier::simplify): Add state
18217 (vrp_jump_threader::vrp_jump_threader): Add state.
18218 (vrp_jump_threader::~vrp_jump_threader): Cleanup state.
18220 2021-07-27 Aldy Hernandez <aldyh@redhat.com>
18222 * Makefile.in (OBJS): Add gimple-range-path.o.
18223 * gimple-range-path.cc: New file.
18224 * gimple-range-path.h: New file.
18226 2021-07-27 Jonathan Wright <jonathan.wright@arm.com>
18228 * config/aarch64/aarch64-simd.md: Push sign/zero-extension
18229 inside vec_duplicate for all patterns.
18230 * simplify-rtx.c (simplify_context::simplify_unary_operation_1):
18231 Push sign/zero-extension inside vec_duplicate.
18233 2021-07-27 Richard Biener <rguenther@suse.de>
18235 PR tree-optimization/101573
18236 * tree-ssa-uninit.c (warn_uninit_phi_uses): New function
18237 looking at uninitialized PHI arg defs in some constrained cases.
18238 (warn_uninitialized_vars): Call it.
18239 (execute_early_warn_uninitialized): Calculate dominators.
18241 2021-07-27 Richard Biener <rguenther@suse.de>
18243 PR tree-optimization/39821
18244 * tree-vect-stmts.c (vect_model_promotion_demotion_cost): Use
18245 vector_stmt for widening arithmetic.
18246 (vectorizable_conversion): Adjust.
18248 2021-07-27 Martin Jambor <mjambor@suse.cz>
18250 * cgraph.h (ipa_replace_map): New field force_load_ref.
18251 * ipa-prop.h (ipa_param_descriptor): Reduce precision of move_cost,
18252 aded new flag load_dereferenced, adjusted comments.
18253 (ipa_get_param_dereferenced): New function.
18254 (ipa_set_param_dereferenced): Likewise.
18255 * cgraphclones.c (cgraph_node::create_virtual_clone): Follow it.
18256 * ipa-cp.c: Include gimple.h.
18257 (ipcp_discover_new_direct_edges): Take into account dereferenced flag.
18258 (get_replacement_map): New parameter force_load_ref, set the
18259 appropriate flag in ipa_replace_map if set.
18260 (struct symbol_and_index_together): New type.
18261 (adjust_refs_in_act_callers): New function.
18262 (adjust_references_in_caller): Likewise.
18263 (create_specialized_node): When appropriate, call
18264 adjust_references_in_caller and force only load references.
18265 * ipa-prop.c (load_from_dereferenced_name): New function.
18266 (ipa_analyze_controlled_uses): Also detect loads from a
18267 dereference, harden testing of call statements.
18268 (ipa_write_node_info): Stream the dereferenced flag.
18269 (ipa_read_node_info): Likewise.
18270 (ipa_set_jf_constant): Also create refdesc when jump function
18271 references a variable.
18272 (cgraph_node_for_jfunc): Rename to symtab_node_for_jfunc, work
18273 also on references of variables and return a symtab_node. Adjust
18275 (propagate_controlled_uses): Also remove references to VAR_DECLs.
18277 2021-07-27 Jakub Jelinek <jakub@redhat.com>
18279 PR middle-end/101586
18280 * gimple-fold.c (clear_padding_type): Ignore FIELD_DECLs with byte
18281 positions above or equal to sz except for diagnostics of flexible
18284 2021-07-26 Andrew MacLeod <amacleod@redhat.com>
18286 PR tree-optimization/78888
18287 * gimple-range-fold.cc (get_letter_range): New.
18288 (fold_using_range::range_of_builtin_call): Call get_letter_range.
18290 2021-07-26 Andrew MacLeod <amacleod@redhat.com>
18292 PR tree-optimization/78888
18293 * gimple-range-fold.cc (fold_using_range::range_of_builtin_call): Add cases
18294 for CFN_BUILT_IN_TOUPPER and CFN_BUILT_IN_TOLOWER.
18296 2021-07-26 Roger Sayle <roger@nextmovesoftware.com>
18297 Marc Glisse <marc.glisse@inria.fr>
18299 * match.pd (rotate): Simplify equality/inequality of rotations.
18300 (bswap): Simplify equality/inequality tests of byte swapping.
18302 2021-07-26 Aldy Hernandez <aldyh@redhat.com>
18304 * range-op.cc (operator_bitwise_xor::op1_op2_relation_effect):
18307 2021-07-26 Aldy Hernandez <aldyh@redhat.com>
18309 * range-op.cc (operator_lshift::fold_range): Pass rel to
18310 base class fold_range.
18311 (operator_rshift::fold_range): Same.
18313 2021-07-26 Ashimida <ashimida@linux.alibaba.com>
18316 * toplev.h (min_align_loops_log): Remove declaration.
18317 (min_align_jumps_log, min_align_labels_log): Likewise.
18318 (min_align_functions_log): Likewise.
18320 2021-07-26 Aldy Hernandez <aldyh@redhat.com>
18322 * tree-vrp.c (vrp_simplify_cond_using_ranges): Rename vr_values
18324 (execute_vrp): Abstract out simplification of conditionals...
18325 (simplify_casted_conds): ...here.
18327 2021-07-26 Aldy Hernandez <aldyh@redhat.com>
18329 * gimple-array-bounds.cc (array_bounds_checker::get_value_range):
18330 Add gimple argument.
18331 (array_bounds_checker::check_array_ref): Same.
18332 (array_bounds_checker::check_addr_expr): Same.
18333 (array_bounds_checker::check_array_bounds): Pass statement to
18334 check_array_bounds and check_addr_expr.
18335 * gimple-array-bounds.h (check_array_bounds): Add gimple argument.
18336 (check_addr_expr): Same.
18337 (get_value_range): Same.
18339 2021-07-26 Tamar Christina <tamar.christina@arm.com>
18341 * config/aarch64/aarch64-simd-builtins.def (sdot, udot): Rename to..
18342 (sdot_prod, udot_prod): ... This.
18343 * config/aarch64/aarch64-simd.md (aarch64_<sur>dot<vsi2qi>): Merged
18345 (<sur>dot_prod<vsi2qi>): ... this.
18346 (aarch64_<sur>dot_lane<vsi2qi>, aarch64_<sur>dot_laneq<vsi2qi>):
18347 Change operands order.
18348 (<sur>sadv16qi): Use new operands order.
18349 * config/aarch64/arm_neon.h (vdot_u32, vdotq_u32, vdot_s32,
18350 vdotq_s32): Use new RTL ordering.
18352 2021-07-26 Tamar Christina <tamar.christina@arm.com>
18354 * config/aarch64/aarch64-builtins.c (TYPES_TERNOP_SUSS,
18355 aarch64_types_ternop_suss_qualifiers): New.
18356 * config/aarch64/aarch64-simd-builtins.def (usdot_prod): Use it.
18357 * config/aarch64/aarch64-simd.md (usdot_prod<vsi2qi>): Re-organize RTL.
18358 * config/aarch64/arm_neon.h (vusdot_s32, vusdotq_s32): Use it.
18360 2021-07-23 Jakub Jelinek <jakub@redhat.com>
18362 PR rtl-optimization/101562
18363 * expmed.c (store_integral_bit_field): Only use movstrict_optab
18364 if the operand isn't paradoxical.
18366 2021-07-23 Aldy Hernandez <aldyh@redhat.com>
18368 * gimple-array-bounds.h (class array_bounds_checker): Change
18369 ranges type to range_query.
18371 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
18373 * config/aarch64/arm_neon.h (vst1_s64_x2): Use
18374 __builtin_memcpy instead of constructing
18375 __builtin_aarch64_simd_oi one vector at a time.
18376 (vst1_u64_x2): Likewise.
18377 (vst1_f64_x2): Likewise.
18378 (vst1_s8_x2): Likewise.
18379 (vst1_p8_x2): Likewise.
18380 (vst1_s16_x2): Likewise.
18381 (vst1_p16_x2): Likewise.
18382 (vst1_s32_x2): Likewise.
18383 (vst1_u8_x2): Likewise.
18384 (vst1_u16_x2): Likewise.
18385 (vst1_u32_x2): Likewise.
18386 (vst1_f16_x2): Likewise.
18387 (vst1_f32_x2): Likewise.
18388 (vst1_p64_x2): Likewise.
18389 (vst1q_s8_x2): Likewise.
18390 (vst1q_p8_x2): Likewise.
18391 (vst1q_s16_x2): Likewise.
18392 (vst1q_p16_x2): Likewise.
18393 (vst1q_s32_x2): Likewise.
18394 (vst1q_s64_x2): Likewise.
18395 (vst1q_u8_x2): Likewise.
18396 (vst1q_u16_x2): Likewise.
18397 (vst1q_u32_x2): Likewise.
18398 (vst1q_u64_x2): Likewise.
18399 (vst1q_f16_x2): Likewise.
18400 (vst1q_f32_x2): Likewise.
18401 (vst1q_f64_x2): Likewise.
18402 (vst1q_p64_x2): Likewise.
18404 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
18406 * config/aarch64/arm_neon.h (vst1_s64_x3): Use
18407 __builtin_memcpy instead of constructing
18408 __builtin_aarch64_simd_ci one vector at a time.
18409 (vst1_u64_x3): Likewise.
18410 (vst1_f64_x3): Likewise.
18411 (vst1_s8_x3): Likewise.
18412 (vst1_p8_x3): Likewise.
18413 (vst1_s16_x3): Likewise.
18414 (vst1_p16_x3): Likewise.
18415 (vst1_s32_x3): Likewise.
18416 (vst1_u8_x3): Likewise.
18417 (vst1_u16_x3): Likewise.
18418 (vst1_u32_x3): Likewise.
18419 (vst1_f16_x3): Likewise.
18420 (vst1_f32_x3): Likewise.
18421 (vst1_p64_x3): Likewise.
18422 (vst1q_s8_x3): Likewise.
18423 (vst1q_p8_x3): Likewise.
18424 (vst1q_s16_x3): Likewise.
18425 (vst1q_p16_x3): Likewise.
18426 (vst1q_s32_x3): Likewise.
18427 (vst1q_s64_x3): Likewise.
18428 (vst1q_u8_x3): Likewise.
18429 (vst1q_u16_x3): Likewise.
18430 (vst1q_u32_x3): Likewise.
18431 (vst1q_u64_x3): Likewise.
18432 (vst1q_f16_x3): Likewise.
18433 (vst1q_f32_x3): Likewise.
18434 (vst1q_f64_x3): Likewise.
18435 (vst1q_p64_x3): Likewise.
18437 2021-07-23 H.J. Lu <hjl.tools@gmail.com>
18440 * config/i386/i386.c (ix86_gen_scratch_sse_rtx): Don't return
18441 hard register when LRA is in progress.
18443 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
18445 * config/aarch64/arm_neon.h (vst1_s8_x4): Use
18446 __builtin_memcpy instead of using a union.
18447 (vst1q_s8_x4): Likewise.
18448 (vst1_s16_x4): Likewise.
18449 (vst1q_s16_x4): Likewise.
18450 (vst1_s32_x4): Likewise.
18451 (vst1q_s32_x4): Likewise.
18452 (vst1_u8_x4): Likewise.
18453 (vst1q_u8_x4): Likewise.
18454 (vst1_u16_x4): Likewise.
18455 (vst1q_u16_x4): Likewise.
18456 (vst1_u32_x4): Likewise.
18457 (vst1q_u32_x4): Likewise.
18458 (vst1_f16_x4): Likewise.
18459 (vst1q_f16_x4): Likewise.
18460 (vst1_f32_x4): Likewise.
18461 (vst1q_f32_x4): Likewise.
18462 (vst1_p8_x4): Likewise.
18463 (vst1q_p8_x4): Likewise.
18464 (vst1_p16_x4): Likewise.
18465 (vst1q_p16_x4): Likewise.
18466 (vst1_s64_x4): Likewise.
18467 (vst1_u64_x4): Likewise.
18468 (vst1_p64_x4): Likewise.
18469 (vst1q_s64_x4): Likewise.
18470 (vst1q_u64_x4): Likewise.
18471 (vst1q_p64_x4): Likewise.
18472 (vst1_f64_x4): Likewise.
18473 (vst1q_f64_x4): Likewise.
18475 2021-07-23 Jonathan Wrightt <jonathan.wright@arm.com>
18477 * config/aarch64/arm_neon.h (vst2_s64): Use __builtin_memcpy
18478 instead of constructing __builtin_aarch64_simd_oi one vector
18480 (vst2_u64): Likewise.
18481 (vst2_f64): Likewise.
18482 (vst2_s8): Likewise.
18483 (vst2_p8): Likewise.
18484 (vst2_s16): Likewise.
18485 (vst2_p16): Likewise.
18486 (vst2_s32): Likewise.
18487 (vst2_u8): Likewise.
18488 (vst2_u16): Likewise.
18489 (vst2_u32): Likewise.
18490 (vst2_f16): Likewise.
18491 (vst2_f32): Likewise.
18492 (vst2_p64): Likewise.
18493 (vst2q_s8): Likewise.
18494 (vst2q_p8): Likewise.
18495 (vst2q_s16): Likewise.
18496 (vst2q_p16): Likewise.
18497 (vst2q_s32): Likewise.
18498 (vst2q_s64): Likewise.
18499 (vst2q_u8): Likewise.
18500 (vst2q_u16): Likewise.
18501 (vst2q_u32): Likewise.
18502 (vst2q_u64): Likewise.
18503 (vst2q_f16): Likewise.
18504 (vst2q_f32): Likewise.
18505 (vst2q_f64): Likewise.
18506 (vst2q_p64): Likewise.
18508 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
18510 * config/aarch64/arm_neon.h (vst3_s64): Use __builtin_memcpy
18511 instead of constructing __builtin_aarch64_simd_ci one vector
18513 (vst3_u64): Likewise.
18514 (vst3_f64): Likewise.
18515 (vst3_s8): Likewise.
18516 (vst3_p8): Likewise.
18517 (vst3_s16): Likewise.
18518 (vst3_p16): Likewise.
18519 (vst3_s32): Likewise.
18520 (vst3_u8): Likewise.
18521 (vst3_u16): Likewise.
18522 (vst3_u32): Likewise.
18523 (vst3_f16): Likewise.
18524 (vst3_f32): Likewise.
18525 (vst3_p64): Likewise.
18526 (vst3q_s8): Likewise.
18527 (vst3q_p8): Likewise.
18528 (vst3q_s16): Likewise.
18529 (vst3q_p16): Likewise.
18530 (vst3q_s32): Likewise.
18531 (vst3q_s64): Likewise.
18532 (vst3q_u8): Likewise.
18533 (vst3q_u16): Likewise.
18534 (vst3q_u32): Likewise.
18535 (vst3q_u64): Likewise.
18536 (vst3q_f16): Likewise.
18537 (vst3q_f32): Likewise.
18538 (vst3q_f64): Likewise.
18539 (vst3q_p64): Likewise.
18541 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
18543 * config/aarch64/arm_neon.h (vst4_s64): Use __builtin_memcpy
18544 instead of constructing __builtin_aarch64_simd_xi one vector
18546 (vst4_u64): Likewise.
18547 (vst4_f64): Likewise.
18548 (vst4_s8): Likewise.
18549 (vst4_p8): Likewise.
18550 (vst4_s16): Likewise.
18551 (vst4_p16): Likewise.
18552 (vst4_s32): Likewise.
18553 (vst4_u8): Likewise.
18554 (vst4_u16): Likewise.
18555 (vst4_u32): Likewise.
18556 (vst4_f16): Likewise.
18557 (vst4_f32): Likewise.
18558 (vst4_p64): Likewise.
18559 (vst4q_s8): Likewise.
18560 (vst4q_p8): Likewise.
18561 (vst4q_s16): Likewise.
18562 (vst4q_p16): Likewise.
18563 (vst4q_s32): Likewise.
18564 (vst4q_s64): Likewise.
18565 (vst4q_u8): Likewise.
18566 (vst4q_u16): Likewise.
18567 (vst4q_u32): Likewise.
18568 (vst4q_u64): Likewise.
18569 (vst4q_f16): Likewise.
18570 (vst4q_f32): Likewise.
18571 (vst4q_f64): Likewise.
18572 (vst4q_p64): Likewise.
18574 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
18576 * config/aarch64/arm_neon.h (vtbx4_s8): Use __builtin_memcpy
18577 instead of constructing __builtin_aarch64_simd_oi one vector
18579 (vtbx4_u8): Likewise.
18580 (vtbx4_p8): Likewise.
18582 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
18584 * config/aarch64/arm_neon.h (vtbl3_s8): Use __builtin_memcpy
18585 instead of constructing __builtin_aarch64_simd_oi one vector
18587 (vtbl3_u8): Likewise.
18588 (vtbl3_p8): Likewise.
18589 (vtbl4_s8): Likewise.
18590 (vtbl4_u8): Likewise.
18591 (vtbl4_p8): Likewise.
18593 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
18595 * config/aarch64/arm_neon.h (vqtbx2_s8): Use __builtin_memcpy
18596 instead of constructing __builtin_aarch64_simd_oi one vector
18598 (vqtbx2_u8): Likewise.
18599 (vqtbx2_p8): Likewise.
18600 (vqtbx2q_s8): Likewise.
18601 (vqtbx2q_u8): Likewise.
18602 (vqtbx2q_p8): Likewise.
18603 (vqtbx3_s8): Use __builtin_memcpy instead of constructing
18604 __builtin_aarch64_simd_ci one vector at a time.
18605 (vqtbx3_u8): Likewise.
18606 (vqtbx3_p8): Likewise.
18607 (vqtbx3q_s8): Likewise.
18608 (vqtbx3q_u8): Likewise.
18609 (vqtbx3q_p8): Likewise.
18610 (vqtbx4_s8): Use __builtin_memcpy instead of constructing
18611 __builtin_aarch64_simd_xi one vector at a time.
18612 (vqtbx4_u8): Likewise.
18613 (vqtbx4_p8): Likewise.
18614 (vqtbx4q_s8): Likewise.
18615 (vqtbx4q_u8): Likewise.
18616 (vqtbx4q_p8): Likewise.
18618 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
18620 * config/aarch64/arm_neon.h (vqtbl2_s8): Use __builtin_memcpy
18621 instead of constructing __builtin_aarch64_simd_oi one vector
18623 (vqtbl2_u8): Likewise.
18624 (vqtbl2_p8): Likewise.
18625 (vqtbl2q_s8): Likewise.
18626 (vqtbl2q_u8): Likewise.
18627 (vqtbl2q_p8): Likewise.
18628 (vqtbl3_s8): Use __builtin_memcpy instead of constructing
18629 __builtin_aarch64_simd_ci one vector at a time.
18630 (vqtbl3_u8): Likewise.
18631 (vqtbl3_p8): Likewise.
18632 (vqtbl3q_s8): Likewise.
18633 (vqtbl3q_u8): Likewise.
18634 (vqtbl3q_p8): Likewise.
18635 (vqtbl4_s8): Use __builtin_memcpy instead of constructing
18636 __builtin_aarch64_simd_xi one vector at a time.
18637 (vqtbl4_u8): Likewise.
18638 (vqtbl4_p8): Likewise.
18639 (vqtbl4q_s8): Likewise.
18640 (vqtbl4q_u8): Likewise.
18641 (vqtbl4q_p8): Likewise.
18643 2021-07-23 Haochen Gui <guihaoc@gcc.gnu.org>
18646 * config/rs6000/rs6000.md (cstore<mode>4): Fix wrong fall through.
18648 2021-07-22 Andrew Pinski <apinski@marvell.com>
18650 PR tree-optimization/10153
18651 * tree-tailcall.c (create_tailcall_accumulator):
18652 Don't call fold_convert as the type should be correct already.
18653 (tree_optimize_tail_calls_1): Use build_{one,zero}_cst instead
18654 of integer_{one,zero}_node for the call of create_tailcall_accumulator.
18656 2021-07-22 Aldy Hernandez <aldyh@redhat.com>
18658 * gimple-range-cache.cc (non_null_ref::adjust_range): Replace
18659 varying_p check for null/non-null check.
18661 2021-07-22 Andrew MacLeod <amacleod@redhat.com>
18663 PR tree-optimization/101511
18664 * value-relation.cc (relation_oracle::query_relation): Check if ssa1
18665 is in ssa2's equiv set, and don't trap if so.
18667 2021-07-22 Andrew MacLeod <amacleod@redhat.com>
18669 PR tree-optimization/101497
18670 * gimple-range-fold.cc (fold_using_range::range_of_cond_expr): Check
18673 2021-07-22 Andrew MacLeod <amacleod@redhat.com>
18675 PR tree-optimization/101496
18676 * vr-values.c (simplify_using_ranges::fold_cond): Call range_of_stmt
18677 first, then vrp_visit_cond_Stmt.
18679 2021-07-22 liuhongt <hongtao.liu@intel.com>
18681 * config/i386/i386-expand.c
18682 (ix86_broadcast_from_integer_constant): Rename to ..
18683 (ix86_broadcast_from_constant): .. this, and extend it to
18685 (ix86_expand_vector_move): Extend to float mode.
18686 * config/i386/i386-features.c
18687 (replace_constant_pool_with_broadcast): Remove.
18688 (remove_partial_avx_dependency_gate): Ditto.
18689 (constant_pool_broadcast): Ditto.
18690 (class pass_constant_pool_broadcast): Ditto.
18691 (make_pass_constant_pool_broadcast): Ditto.
18692 (remove_partial_avx_dependency): Adjust gate.
18693 * config/i386/i386-passes.def: Remove pass_constant_pool_broadcast.
18694 * config/i386/i386-protos.h
18695 (make_pass_constant_pool_broadcast): Remove.
18697 2021-07-22 liuhongt <hongtao.liu@intel.com>
18699 * config/i386/constraints.md (Wb): New constraint.
18701 * config/i386/i386.md (*ashlhi3_1): Extend to avx512 mask
18703 (*ashlqi3_1): Ditto.
18704 (*<insn><mode>3_1): Split to ..
18705 (*ashr<mode>3_1): this, ...
18706 (*lshr<mode>3_1): and this, also extend this pattern to avx512
18708 (*<insn><mode>3_1): Split to ..
18709 (*ashr<mode>3_1): this, ...
18710 (*lshrqi3_1): and this, also extend this pattern to avx512
18712 (*lshrhi3_1): And this, also extend this pattern to avx512
18714 * config/i386/sse.md (k<code><mode>): New define_split after
18715 it to convert generic shift pattern to mask shift ones.
18717 2021-07-21 Thomas Schwinge <thomas@codesourcery.com>
18718 Joseph Myers <joseph@codesourcery.com>
18719 Cesar Philippidis <cesar@codesourcery.com>
18721 * tree-core.h (omp_clause_code): Add 'OMP_CLAUSE_NOHOST'.
18722 * tree.c (omp_clause_num_ops, omp_clause_code_name, walk_tree_1):
18724 * tree-pretty-print.c (dump_omp_clause): Likewise.
18725 * omp-general.c (oacc_verify_routine_clauses): Likewise.
18726 * gimplify.c (gimplify_scan_omp_clauses)
18727 (gimplify_adjust_omp_clauses): Likewise.
18728 * tree-nested.c (convert_nonlocal_omp_clauses)
18729 (convert_local_omp_clauses): Likewise.
18730 * omp-low.c (scan_sharing_clauses): Likewise.
18731 * omp-offload.c (execute_oacc_device_lower): Update.
18733 2021-07-21 Martin Sebor <msebor@redhat.com>
18735 * tree-ssa-alias.c (walk_aliased_vdefs_1): Fix typos in a comment.
18737 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
18739 * config/rs6000/rs6000-gen-builtins.c (write_init_bif_table):
18742 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
18744 * config/rs6000/rs6000-gen-builtins.c (write_fntype): New
18746 (write_fntype_init): New stub function.
18747 (write_init_bif_table): Likewise.
18748 (write_init_ovld_table): New function.
18749 (write_init_file): Implement.
18751 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
18753 * config/rs6000/rs6000-gen-builtins.c
18754 (write_autogenerated_header): New function.
18755 (write_decls): Likewise.
18756 (write_extern_fntype): New callback function.
18757 (write_header_file): Implement.
18759 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
18761 * config/rs6000/rs6000-gen-builtins.c (write_defines_file):
18764 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
18766 * config/rs6000/rs6000-gen-builtins.c (complete_vector_type): New
18768 (complete_base_type): Likewise.
18769 (construct_fntype_id): Likewise.
18770 (parse_bif_entry): Call contruct_fntype_id.
18771 (parse_ovld_entry): Likewise.
18773 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
18775 * config/rs6000/rs6000-gen-builtins.c (ovld_stanza): New struct.
18776 (MAXOVLDSTANZAS): New macro.
18777 (ovld_stanzas): New variable.
18778 (curr_ovld_stanza): Likewise.
18779 (MAXOVLDS): New macro.
18780 (ovlddata): New struct.
18781 (ovlds): New variable.
18782 (curr_ovld): Likewise.
18783 (max_ovld_args): Likewise.
18784 (parse_ovld_entry): New function.
18785 (parse_ovld_stanza): Likewise.
18786 (parse_ovld): Implement.
18788 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
18790 * config/rs6000/rs6000-gen-builtins.c (parse_bif_attrs):
18793 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
18795 * config/rs6000/rs6000-gen-builtins.c (parse_args): New function.
18796 (parse_prototype): Implement.
18798 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
18800 * config/rs6000/rs6000-gen-builtins.c (bif_stanza): New enum.
18801 (curr_bif_stanza): New variable.
18802 (stanza_entry): New struct.
18803 (stanza_map): New initialized variable.
18804 (enable_string): Likewise.
18805 (fnkinds): New enum.
18806 (typelist): New struct.
18807 (attrinfo): Likewise.
18808 (MAXRESTROPNDS): New macro.
18809 (prototype): New struct.
18810 (MAXBIFS): New macro.
18811 (bifdata): New struct.
18812 (bifs): New variable.
18813 (curr_bif): Likewise.
18814 (bif_order): Likewise.
18815 (bif_index): Likewise.
18816 (fatal): New function.
18817 (stanza_name_to_stanza): Likewise.
18818 (parse_bif_attrs): New stub function.
18819 (parse_prototype): Likewise.
18820 (parse_bif_entry): New function.
18821 (parse_bif_stanza): Likewise.
18822 (parse_bif): Implement.
18823 (set_bif_order): New function.
18824 (create_bif_order): Implement.
18826 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
18828 * config/rs6000/rs6000-gen-builtins.c (rbtree.h): New #include.
18829 (num_bifs): New variable.
18830 (num_ovld_stanzas): Likewise.
18831 (num_ovlds): Likewise.
18832 (parse_codes): New enum.
18833 (bif_rbt): New variable.
18834 (ovld_rbt): Likewise.
18835 (fntype_rbt): Likewise.
18836 (bifo_rbt): Likewise.
18837 (parse_bif): New stub function.
18838 (create_bif_order): Likewise.
18839 (parse_ovld): Likewise.
18840 (write_header_file): Likewise.
18841 (write_init_file): Likewise.
18842 (write_defines_file): Likewise.
18843 (delete_output_files): New function.
18846 2021-07-21 H.J. Lu <hjl.tools@gmail.com>
18849 * config/i386/i386-builtin.def: Remove OPTION_MASK_ISA_SSE4_2
18850 from CRC32 _builtin functions.
18852 2021-07-21 Sebastian Huber <sebastian.huber@embedded-brains.de>
18854 * coverage.c (build_gcov_info_var_registration): Mark the object placed
18855 in the linker set as referenced so that it does not get optimized away.
18857 2021-07-21 Kito Cheng <kito.cheng@sifive.com>
18860 2021-07-20 Kito Cheng <kito.cheng@sifive.com>
18862 * config.gcc (riscv*-*-*): Detect which python is available.
18864 2021-07-21 Jakub Jelinek <jakub@redhat.com>
18866 PR middle-end/101535
18867 * gimplify.c (omp_check_private): Properly skip ORT_TARGET_DATA
18868 contexts in which decl isn't privatized and for ORT_TARGET return
18869 false if decl is mapped.
18871 2021-07-21 Richard Sandiford <richard.sandiford@arm.com>
18873 * gimple-loop-jam.c: Include tree-ssa-sccvn.h.
18874 (tree_loop_unroll_and_jam): Run value-numbering on a loop that
18875 has been successfully unrolled.
18877 2021-07-21 Richard Sandiford <richard.sandiford@arm.com>
18879 * tree-ssa-loop-manip.c (determine_exit_conditions): Return a null
18880 exit condition if no tail loop is needed, and if the original exit
18881 condition should therefore be kept as-is.
18882 (tree_transform_and_unroll_loop): Handle that case here too.
18884 2021-07-21 Kewen Lin <linkw@linux.ibm.com>
18886 * tree-data-ref.c (free_dependence_relations): Adjust to pass vec
18888 (free_data_refs): Likewise.
18889 * tree-data-ref.h (free_dependence_relations): Likewise.
18890 (free_data_refs): Likewise.
18891 * tree-predcom.c (struct chain): Use auto_vec instead of vec for
18893 (struct component): Likewise.
18894 (pcom_worker::pcom_worker): Adjust for auto_vec and renaming changes.
18895 (pcom_worker::~pcom_worker): Likewise.
18896 (pcom_worker::release_chain): Adjust as auto_vec changes.
18897 (pcom_worker::loop): Rename to ...
18898 (pcom_worker::m_loop): ... this.
18899 (pcom_worker::datarefs): Rename to ...
18900 (pcom_worker::m_datarefs): ... this. Use auto_vec instead of vec.
18901 (pcom_worker::dependences): Rename to ...
18902 (pcom_worker::m_dependences): ... this. Use auto_vec instead of vec.
18903 (pcom_worker::chains): Rename to ...
18904 (pcom_worker::m_chains): ... this. Use auto_vec instead of vec.
18905 (pcom_worker::looparound_phis): Rename to ...
18906 (pcom_worker::m_looparound_phis): ... this. Use auto_vec instead of
18908 (pcom_worker::cache): Rename to ...
18909 (pcom_worker::m_cache): ... this. Use auto_vec instead of vec.
18910 (pcom_worker::release_chain): Adjust for auto_vec changes.
18911 (pcom_worker::release_chains): Adjust for auto_vec and renaming
18913 (release_component): Remove.
18914 (release_components): Adjust for release_component removal.
18915 (component_of): Adjust to use vec.
18916 (merge_comps): Likewise.
18917 (pcom_worker::aff_combination_dr_offset): Adjust for renaming changes.
18918 (pcom_worker::determine_offset): Likewise.
18919 (class comp_ptrs): Remove.
18920 (pcom_worker::split_data_refs_to_components): Adjust for renaming
18921 changes, for comp_ptrs removal with auto_vec.
18922 (pcom_worker::suitable_component_p): Adjust for renaming changes.
18923 (pcom_worker::filter_suitable_components): Adjust for release_component
18925 (pcom_worker::valid_initializer_p): Adjust for renaming changes.
18926 (pcom_worker::find_looparound_phi): Likewise.
18927 (pcom_worker::add_looparound_copies): Likewise.
18928 (pcom_worker::determine_roots_comp): Likewise.
18929 (pcom_worker::single_nonlooparound_use): Likewise.
18930 (pcom_worker::execute_pred_commoning_chain): Likewise.
18931 (pcom_worker::execute_pred_commoning): Likewise.
18932 (pcom_worker::try_combine_chains): Likewise.
18933 (pcom_worker::prepare_initializers_chain): Likewise.
18934 (pcom_worker::prepare_initializers): Likewise.
18935 (pcom_worker::prepare_finalizers_chain): Likewise.
18936 (pcom_worker::prepare_finalizers): Likewise.
18937 (pcom_worker::tree_predictive_commoning_loop): Likewise.
18939 2021-07-20 Martin Sebor <msebor@redhat.com>
18941 PR middle-end/101397
18942 * builtins.c (gimple_call_return_array): Add argument. Correct
18943 offsets for memchr, mempcpy, stpcpy, and stpncpy.
18944 (compute_objsize_r): Adjust offset computation for argument returning
18947 2021-07-20 Martin Sebor <msebor@redhat.com>
18949 PR middle-end/101300
18950 * tree-ssa-uninit.c (check_defs): Handle UBSAN built-ins.
18952 2021-07-20 Jeff Law <jlaw@localhost.localdomain>
18954 * function.c (assign_parm_setup_block): Use adjust_address instead
18955 of change_address to preserve MEM_EXPR and friends.
18957 2021-07-20 Martin Sebor <msebor@redhat.com>
18959 * cfgloop.h (single_likely_exit): Adjust by-value argument to
18960 by-const-reference.
18961 * cfgloopanal.c (single_likely_exit): Same.
18962 * cgraph.h (struct cgraph_node): Same.
18963 * cgraphclones.c (cgraph_node::create_virtual_clone): Same.
18964 * genautomata.c (merge_states): Same.
18965 * genextract.c (VEC_char_to_string): Same.
18966 * genmatch.c (dt_node::gen_kids_1): Same.
18967 (walk_captures): Adjust by-value argument to by-reference.
18968 * gimple-ssa-store-merging.c (check_no_overlap): Adjust by-value argument
18969 to by-const-reference.
18970 * gimple.c (gimple_build_call_vec): Same.
18971 (gimple_build_call_internal_vec): Same.
18972 (gimple_build_switch): Same.
18973 (sort_case_labels): Same.
18974 (preprocess_case_label_vec_for_gimple): Adjust by-value argument to
18976 * gimple.h (gimple_build_call_vec): Adjust by-value argument to
18977 by-const-reference.
18978 (gimple_build_call_internal_vec): Same.
18979 (gimple_build_switch): Same.
18980 (sort_case_labels): Same.
18981 (preprocess_case_label_vec_for_gimple): Adjust by-value argument to
18983 * haifa-sched.c (calc_priorities): Adjust by-value argument to
18984 by-const-reference.
18985 (sched_init_luids): Same.
18986 (haifa_init_h_i_d): Same.
18987 * ipa-cp.c (ipa_get_indirect_edge_target_1): Same.
18988 (adjust_callers_for_value_intersection): Adjust by-value argument to
18990 (find_more_scalar_values_for_callers_subset): Adjust by-value argument to
18991 by-const-reference.
18992 (find_more_contexts_for_caller_subset): Same.
18993 (find_aggregate_values_for_callers_subset): Same.
18994 (copy_useful_known_contexts): Same.
18995 * ipa-fnsummary.c (remap_edge_summaries): Same.
18996 (remap_freqcounting_predicate): Same.
18997 * ipa-inline.c (add_new_edges_to_heap): Adjust by-value argument to
18999 * ipa-predicate.c (predicate::remap_after_inlining): Adjust by-value argument
19000 to by-const-reference.
19001 * ipa-predicate.h (predicate::remap_after_inlining): Same.
19002 * ipa-prop.c (ipa_find_agg_cst_for_param): Same.
19003 * ipa-prop.h (ipa_find_agg_cst_for_param): Same.
19004 * ira-build.c (ira_loop_tree_body_rev_postorder): Same.
19005 * read-rtl.c (add_overload_instance): Same.
19006 * rtl.h (native_decode_rtx): Same.
19007 (native_decode_vector_rtx): Same.
19008 * sched-int.h (sched_init_luids): Same.
19009 (haifa_init_h_i_d): Same.
19010 * simplify-rtx.c (native_decode_vector_rtx): Same.
19011 (native_decode_rtx): Same.
19012 * tree-call-cdce.c (gen_shrink_wrap_conditions): Same.
19013 (shrink_wrap_one_built_in_call_with_conds): Same.
19014 (shrink_wrap_conditional_dead_built_in_calls): Same.
19015 * tree-data-ref.c (create_runtime_alias_checks): Same.
19016 (compute_all_dependences): Same.
19017 * tree-data-ref.h (compute_all_dependences): Same.
19018 (create_runtime_alias_checks): Same.
19019 (index_in_loop_nest): Same.
19020 * tree-if-conv.c (mask_exists): Same.
19021 * tree-loop-distribution.c (class loop_distribution): Same.
19022 (loop_distribution::create_rdg_vertices): Same.
19023 (dump_rdg_partitions): Same.
19024 (debug_rdg_partitions): Same.
19025 (partition_contains_all_rw): Same.
19026 (loop_distribution::distribute_loop): Same.
19027 * tree-parloops.c (oacc_entry_exit_ok_1): Same.
19028 (oacc_entry_exit_single_gang): Same.
19029 * tree-ssa-loop-im.c (hoist_memory_references): Same.
19030 (loop_suitable_for_sm): Same.
19031 * tree-ssa-loop-niter.c (bound_index): Same.
19032 * tree-ssa-reassoc.c (update_ops): Same.
19033 (swap_ops_for_binary_stmt): Same.
19034 (rewrite_expr_tree): Same.
19035 (rewrite_expr_tree_parallel): Same.
19036 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Same.
19037 * tree-ssa-sccvn.h (ao_ref_init_from_vn_reference): Same.
19038 * tree-ssa-structalias.c (process_all_all_constraints): Same.
19039 (make_constraints_to): Same.
19040 (handle_lhs_call): Same.
19041 (find_func_aliases_for_builtin_call): Same.
19042 (sort_fieldstack): Same.
19043 (check_for_overlaps): Same.
19044 * tree-vect-loop-manip.c (vect_create_cond_for_align_checks): Same.
19045 (vect_create_cond_for_unequal_addrs): Same.
19046 (vect_create_cond_for_lower_bounds): Same.
19047 (vect_create_cond_for_alias_checks): Same.
19048 * tree-vect-slp-patterns.c (vect_validate_multiplication): Same.
19049 * tree-vect-slp.c (vect_analyze_slp_instance): Same.
19050 (vect_make_slp_decision): Same.
19051 (vect_slp_bbs): Same.
19052 (duplicate_and_interleave): Same.
19053 (vect_transform_slp_perm_load): Same.
19054 (vect_schedule_slp): Same.
19055 * tree-vectorizer.h (vect_transform_slp_perm_load): Same.
19056 (vect_schedule_slp): Same.
19057 (duplicate_and_interleave): Same.
19058 * tree.c (build_vector_from_ctor): Same.
19059 (build_vector): Same.
19060 (check_vector_cst): Same.
19061 (check_vector_cst_duplicate): Same.
19062 (check_vector_cst_fill): Same.
19063 (check_vector_cst_stepped): Same.
19064 * tree.h (build_vector_from_ctor): Same.
19066 2021-07-20 Jakub Jelinek <jakub@redhat.com>
19069 * config/rs6000/rs6000-protos.h (easy_altivec_constant): Change return
19070 type from bool to int.
19071 * config/rs6000/rs6000.c (vspltis_constant): Fix up handling the
19072 EASY_VECTOR_MSB case if either step or copies is not 1.
19073 (vspltis_shifted): Fix comment typo.
19074 (easy_altivec_constant): Change return type from bool to int, instead
19075 of returning true return byte size of the element mode that should be
19076 used to synthetize the constant.
19077 * config/rs6000/predicates.md (easy_vector_constant_msb): Require
19078 that vspltis_shifted is 0, handle the case where easy_altivec_constant
19079 assumes using different vector mode from CONST_VECTOR's mode.
19080 * config/rs6000/altivec.md (easy_vector_constant_msb splitter): Use
19081 easy_altivec_constant to determine mode in which -1 >> -1 should be
19082 performed, use rs6000_expand_vector_init instead of gen_vec_initv4sisi.
19084 2021-07-20 Richard Biener <rguenther@suse.de>
19087 * dwarf2out.h (dwarf_file_data): Add key member.
19088 * dwarf2out.c (dwarf_file_hasher::equal): Compare key.
19089 (dwarf_file_hasher::hash): Hash key.
19090 (lookup_filename): Remap the filename and store it in the
19091 filename member of dwarf_file_data when creating a new
19093 (file_name_acquire): Do not remap the filename again.
19094 (maybe_emit_file): Likewise.
19096 2021-07-20 Jonathan Wright <jonathan.wright@arm.com>
19098 * config/aarch64/aarch64-simd-builtins.def: Use two variant
19099 generators for all TBL/TBX intrinsics and rename to
19100 consistent forms: qtbl[1234] or qtbx[1234].
19101 * config/aarch64/aarch64-simd.md (aarch64_tbl1<mode>):
19103 (aarch64_qtbl1<mode>): This.
19104 (aarch64_tbx1<mode>): Rename to...
19105 (aarch64_qtbx1<mode>): This.
19106 (aarch64_tbl2v16qi): Delete.
19107 (aarch64_tbl3<mode>): Rename to...
19108 (aarch64_qtbl2<mode>): This.
19109 (aarch64_tbx4<mode>): Rename to...
19110 (aarch64_qtbx2<mode>): This.
19111 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_1): Use
19112 renamed qtbl1 and qtbl2 RTL patterns.
19113 * config/aarch64/arm_neon.h (vqtbl1_p8): Use renamed qtbl1
19115 (vqtbl1_s8): Likewise.
19116 (vqtbl1_u8): Likewise.
19117 (vqtbl1q_p8): Likewise.
19118 (vqtbl1q_s8): Likewise.
19119 (vqtbl1q_u8): Likewise.
19120 (vqtbx1_s8): Use renamed qtbx1 RTL pattern.
19121 (vqtbx1_u8): Likewise.
19122 (vqtbx1_p8): Likewise.
19123 (vqtbx1q_s8): Likewise.
19124 (vqtbx1q_u8): Likewise.
19125 (vqtbx1q_p8): Likewise.
19126 (vtbl1_s8): Use renamed qtbl1 RTL pattern.
19127 (vtbl1_u8): Likewise.
19128 (vtbl1_p8): Likewise.
19129 (vtbl2_s8): Likewise
19130 (vtbl2_u8): Likewise.
19131 (vtbl2_p8): Likewise.
19132 (vtbl3_s8): Use renamed qtbl2 RTL pattern.
19133 (vtbl3_u8): Likewise.
19134 (vtbl3_p8): Likewise.
19135 (vtbl4_s8): Likewise.
19136 (vtbl4_u8): Likewise.
19137 (vtbl4_p8): Likewise.
19138 (vtbx2_s8): Use renamed qtbx2 RTL pattern.
19139 (vtbx2_u8): Likewise.
19140 (vtbx2_p8): Likewise.
19141 (vqtbl2_s8): Use renamed qtbl2 RTL pattern.
19142 (vqtbl2_u8): Likewise.
19143 (vqtbl2_p8): Likewise.
19144 (vqtbl2q_s8): Likewise.
19145 (vqtbl2q_u8): Likewise.
19146 (vqtbl2q_p8): Likewise.
19147 (vqtbx2_s8): Use renamed qtbx2 RTL pattern.
19148 (vqtbx2_u8): Likewise.
19149 (vqtbx2_p8): Likewise.
19150 (vqtbx2q_s8): Likewise.
19151 (vqtbx2q_u8): Likewise.
19152 (vqtbx2q_p8): Likewise.
19153 (vtbx4_s8): Likewise.
19154 (vtbx4_u8): Likewise.
19155 (vtbx4_p8): Likewise.
19157 2021-07-20 Uroš Bizjak <ubizjak@gmail.com>
19160 * config/i386/sync.md (define_peephole2 atomic_storedi_fpu):
19162 (define_peephole2 atomic_loaddi_fpu): Ditto.
19164 2021-07-20 Kito Cheng <kito.cheng@sifive.com>
19166 * config.gcc (riscv*-*-*): Detect which python is available.
19168 2021-07-20 Kewen Lin <linkw@linux.ibm.com>
19170 * config/rs6000/vsx.md (mulhs_<mode>): Rename to...
19171 (smul<mode>3_highpart): ... this.
19172 (mulhu_<mode>): Rename to...
19173 (umul<mode>3_highpart): ... this.
19174 * config/rs6000/rs6000-builtin.def (MULHS_V2DI, MULHS_V4SI,
19175 MULHU_V2DI, MULHU_V4SI): Adjust.
19177 2021-07-20 Kewen Lin <linkw@linux.ibm.com>
19179 PR tree-optimization/100696
19180 * internal-fn.c (first_commutative_argument): Add info for IFN_MULH.
19181 * internal-fn.def (IFN_MULH): New internal function.
19182 * tree-vect-patterns.c (vect_recog_mulhs_pattern): Add support to
19183 recog normal multiply highpart as IFN_MULH.
19184 * config/i386/i386.c (ix86_add_stmt_cost): Adjust for combined
19187 2021-07-19 Indu Bhagat <indu.bhagat@oracle.com>
19189 * config/elfos.h (CTF_DEBUGGING_INFO): New definition.
19190 (BTF_DEBUGGING_INFO): Likewise.
19191 * doc/tm.texi.in: Document the new macros.
19192 * doc/tm.texi: Regenerated.
19193 * toplev.c: Guard initialization of debug hooks.
19195 2021-07-19 Indu Bhagat <indu.bhagat@oracle.com>
19197 * flags.h (ctf_debuginfo_p): New function declaration.
19198 * opts.c (ctf_debuginfo_p): New function definition.
19200 2021-07-19 Andrew Stubbs <ams@codesourcery.com>
19203 * config/gcn/gcn-hsa.h (DRIVER_SELF_SPECS): New.
19204 (ASM_SPEC): Set -mattr for xnack and sram-ecc.
19205 * config/gcn/gcn-opts.h (enum sram_ecc_type): New.
19206 * config/gcn/gcn-valu.md: Add a warning comment.
19207 * config/gcn/gcn.c (gcn_option_override): Add "sorry" for -mxnack.
19208 (output_file_start): Add xnack and sram-ecc state to ".amdgcn_target".
19209 * config/gcn/gcn.md: Add a warning comment.
19210 * config/gcn/gcn.opt: Add -mxnack and -msram-ecc.
19211 * config/gcn/mkoffload.c (EF_AMDGPU_MACH_AMDGCN_GFX908): Remove
19213 (EF_AMDGPU_XNACK): New.
19214 (EF_AMDGPU_SRAM_ECC): New.
19216 (copy_early_debug_info): Use elf_flags.
19217 (main): Handle -mxnack and -msram-ecc options.
19218 * doc/invoke.texi: Document -mxnack and -msram-ecc.
19220 2021-07-19 Andrew Pinski <apinski@marvell.com>
19223 * config/aarch64/aarch64.md (csneg3_uxtw_insn): Rename to ...
19224 (*cs<neg_not_cs>3_uxtw_insn4): and extend to NEG_NOT.
19226 2021-07-19 Richard Biener <rguenther@suse.de>
19228 PR tree-optimization/101505
19229 * tree-vect-patterns.c (vect_determine_precisions): Walk
19230 PHIs also for loop vectorization.
19232 2021-07-19 Richard Biener <rguenther@suse.de>
19234 * gimple.h (gimple_expr_type): Remove.
19235 * doc/gimple.texi: Remove gimple_expr_type documentation.
19237 2021-07-19 Richard Biener <rguenther@suse.de>
19239 * tree-ssa-sccvn.c (vn_reference_eq): Handle NULL vr->type.
19240 (ao_ref_init_from_vn_reference): Likewise.
19241 (fully_constant_reference): Likewise.
19242 (vn_reference_lookup_call): Do not set vr->type to random
19244 * tree-ssa-pre.c (compute_avail): Do not try to PRE calls
19246 * tree-vect-generic.c (expand_vector_piecewise): Pass in
19247 whether we expanded parallel.
19248 (expand_vector_parallel): Adjust.
19249 (expand_vector_addition): Likewise.
19250 (expand_vector_comparison): Likewise.
19251 (expand_vector_operation): Likewise.
19252 (expand_vector_scalar_condition): Likewise.
19253 (expand_vector_conversion): Likewise.
19255 2021-07-19 Richard Biener <rguenther@suse.de>
19257 * tree-vrp.c (register_edge_assert_for_2): Use the
19259 (vrp_folder::fold_predicate_in): Likewise.
19260 * vr-values.c (gimple_assign_nonzero_p): Likewise.
19261 (vr_values::extract_range_from_comparison): Likewise.
19262 (vr_values::extract_range_from_ubsan_builtin): Use the
19263 type of the first operand.
19264 (vr_values::extract_range_basic): Push down type
19265 computation, use the appropriate LHS.
19266 (vr_values::extract_range_from_assignment): Use the
19269 2021-07-18 H.J. Lu <hjl.tools@gmail.com>
19272 * common/config/i386/i386-common.c (ix86_handle_option): For
19273 -mgeneral-regs-only, enable the GPR only instructions which are
19274 enabled implicitly by SSE ISAs unless they have been disabled
19277 2021-07-18 H.J. Lu <hjl.tools@gmail.com>
19280 * config/i386/i386.c (ix86_check_avx_upper_stores): Moved before
19281 ix86_avx_u128_mode_needed.
19282 (ix86_avx_u128_mode_needed): Return AVX_U128_DIRTY if callee
19283 returns AVX register.
19285 2021-07-17 Jan Hubicka <hubicka@ucw.cz>
19287 * tree-ssa-structalias.c (handle_rhs_call): Support EAF_NOT_RETURNED.
19288 (handle_const_call): Liekise
19289 (handle_pure_call): Liekise
19291 2021-07-17 Andrew MacLeod <amacleod@redhat.com>
19293 PR tree-optimization/96542
19294 * range-op.cc (range_operator::wi_fold_in_parts): New.
19295 (range_operator::fold_range): Call wi_fold_in_parts.
19296 (operator_lshift::wi_fold): Fix broken lshift by [0,0].
19297 * range-op.h (wi_fold_in_parts): Add prototype.
19299 2021-07-16 David Malcolm <dmalcolm@redhat.com>
19301 * doc/analyzer.texi: Add __analyzer_dump_state.
19303 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
19305 * config/rs6000/rbtree.c: New file.
19306 * config/rs6000/rbtree.h: New file.
19308 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
19310 * config/rs6000/rs6000-gen-builtins.c (restriction): New enum.
19311 (typeinfo): Add restr field.
19312 (match_bracketed_pair): New function.
19313 (match_const_restriction): Implement.
19315 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
19317 * config/rs6000/rs6000-gen-builtins.c (match_basetype): Implement.
19319 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
19321 * config/rs6000/rs6000-gen-builtins.c (void_status): New enum.
19322 (basetype): Likewise.
19323 (typeinfo): Likewise.
19324 (handle_pointer): New function.
19325 (match_basetype): New stub function.
19326 (match_const_restriction): Likewise.
19327 (match_type): New function.
19329 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
19331 * config/rs6000/rs6000-gen-builtins.c (consume_whitespace): New
19333 (advance_line): Likewise.
19334 (safe_inc_pos): Likewise.
19335 (match_identifier): Likewise.
19336 (match_integer): Likewise.
19337 (match_to_right_bracket): Likewise.
19339 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
19341 * config/rs6000/rs6000-gen-builtins.c (bif_file): New variable.
19342 (ovld_file): Likewise.
19343 (header_file): Likewise.
19344 (init_file): Likewise.
19345 (defines_file): Likewise.
19346 (pgm_path): Likewise.
19347 (bif_path): Likewise.
19348 (ovld_path): Likewise.
19349 (header_path): Likewise.
19350 (init_path): Likewise.
19351 (defines_path): Likewise.
19352 (LINELEN): New macro.
19353 (linebuf): New variable.
19357 (bif_diag): New function.
19358 (ovld_diag): Likewise.
19360 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
19362 * config/rs6000/rs6000-builtin-new.def: New.
19363 * config/rs6000/rs6000-overload.def: New.
19365 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
19367 * config/rs6000/rs6000-gen-builtins.c: New.
19369 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
19371 * Makefile.in (EXTRA_GTYPE_DEPS): New variable.
19372 (s-gtype): Depend on EXTRA_GTYPE_DEPS.
19373 * gengtype-state.c (state_writer::write_state_file_list): Add a
19374 parameter to the fileslist expression for the number of build
19376 (read_state_files_list): Detect build headers and strip the
19377 initial "./" or ".\" from their names.
19378 * gengtype.c (build_headers): New global variable.
19379 (num_build_headers): Likewise.
19380 (open_base_files): Emit #include for each build header.
19381 (main): Detect and count build headers.
19382 * gengtype.h (build_headers): New extern variable.
19383 (num_build_headers): Likewise.
19385 2021-07-16 Richard Biener <rguenther@suse.de>
19387 * gimple-ssa-store-merging.c (verify_symbolic_number_p): Use
19388 the type of the LHS.
19389 (find_bswap_or_nop_1): Likewise.
19390 (find_bswap_or_nop): Likewise.
19391 * tree-vectorizer.h (vect_get_smallest_scalar_type): Adjust
19393 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
19394 Remove unused parameters, pass in the scalar type. Fix
19395 internal store function handling.
19396 * tree-vect-stmts.c (vect_analyze_stmt): Remove assert.
19397 (vect_get_vector_types_for_stmt): Move down check for
19398 existing vector stmt after we've determined a scalar type.
19399 Pass down the used scalar type to vect_get_smallest_scalar_type.
19400 * tree-vect-generic.c (expand_vector_condition): Use
19401 the type of the LHS.
19402 (expand_vector_scalar_condition): Likewise.
19403 (expand_vector_operations_1): Likewise.
19404 * tree-vect-patterns.c (vect_widened_op_tree): Likewise.
19405 (vect_recog_dot_prod_pattern): Likewise.
19406 (vect_recog_sad_pattern): Likewise.
19407 (vect_recog_widen_op_pattern): Likewise.
19408 (vect_recog_widen_sum_pattern): Likewise.
19409 (vect_recog_mixed_size_cond_pattern): Likewise.
19411 2021-07-16 Jan Hubicka <hubicka@ucw.cz>
19413 * ipa-modref.c (struct escape_entry): Use eaf_fleags_t.
19414 (dump_eaf_flags): Dump EAF_NOT_RETURNED
19415 (eaf_flags_useful_p): Use eaf_fleags_t; handle const functions
19416 and EAF_NOT_RETURNED.
19417 (modref_summary::useful_p): Likewise.
19418 (modref_summary_lto::useful_p): Likewise.
19419 (struct) modref_summary_lto: Use eaf_fleags_t.
19420 (deref_flags): Handle EAF_NOT_RETURNED.
19421 (struct escape_point): Use min_flags.
19422 (modref_lattice::init): Add EAF_NOT_RETURNED.
19423 (merge_call_lhs_flags): Ignore EAF_NOT_RETURNED functions
19424 (analyze_ssa_name_flags): Clear EAF_NOT_RETURNED on return;
19426 (analyze_parms): Also analyze const functions; update conition on
19428 (modref_write): Update streaming.
19429 (read_section): Update streaming.
19430 (remap_arg_flags): Use eaf_flags_t.
19431 (modref_merge_call_site_flags): Hanlde EAF_NOT_RETURNED.
19432 * ipa-modref.h: (eaf_flags_t): New typedef.
19433 (struct modref_summary): Use eaf_flags_t.
19434 * tree-core.h (EAF_NOT_RETURNED): New constant.
19436 2021-07-16 Richard Biener <rguenther@suse.de>
19438 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Use
19439 the type of the LHS.
19440 (gimple_assign_nonnegative_warnv_p): Likewise.
19441 (gimple_call_nonnegative_warnv_p): Likewise. Return false
19442 if the call has no LHS.
19443 * gimple.c (gimple_could_trap_p_1): Use the type of the LHS.
19444 * tree-eh.c (stmt_could_throw_1_p): Likewise.
19445 * tree-inline.c (insert_init_stmt): Likewise.
19446 * tree-ssa-loop-niter.c (get_val_for): Likewise.
19447 * tree-outof-ssa.c (ssa_is_replaceable_p): Use the type of
19449 * tree-ssa-sccvn.c (init_vn_nary_op_from_stmt): Take a
19450 gassign *. Use the type of the lhs.
19451 (vn_nary_op_lookup_stmt): Adjust.
19452 (vn_nary_op_insert_stmt): Likewise.
19454 2021-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
19456 * config/s390/predicates.md (bras_sym_operand): Accept all
19457 functions in 64-bit mode, use UNSPEC_PLT31.
19458 (larl_operand): Use UNSPEC_PLT31.
19459 * config/s390/s390.c (s390_loadrelative_operand_p): Likewise.
19460 (legitimize_pic_address): Likewise.
19461 (s390_emit_tls_call_insn): Mark __tls_get_offset as function,
19463 (s390_delegitimize_address): Use UNSPEC_PLT31.
19464 (s390_output_addr_const_extra): Likewise.
19465 (print_operand): Add @PLT to TLS calls, handle %K.
19466 (s390_function_profiler): Mark __fentry__/_mcount as function,
19467 use %K, use UNSPEC_PLT31.
19468 (s390_output_mi_thunk): Use only UNSPEC_GOT, use %K.
19469 (s390_emit_call): Use UNSPEC_PLT31.
19470 (s390_emit_tpf_eh_return): Mark __tpf_eh_return as function.
19471 * config/s390/s390.md (UNSPEC_PLT31): Rename from UNSPEC_PLT.
19472 (*movdi_64): Use %K.
19473 (reload_base_64): Likewise.
19474 (*sibcall_brc): Likewise.
19475 (*sibcall_brcl): Likewise.
19476 (*sibcall_value_brc): Likewise.
19477 (*sibcall_value_brcl): Likewise.
19479 (*brasl): Likewise.
19480 (*bras_r): Likewise.
19481 (*brasl_r): Likewise.
19482 (*bras_tls): Likewise.
19483 (*brasl_tls): Likewise.
19484 (main_base_64): Likewise.
19485 (reload_base_64): Likewise.
19486 (@split_stack_call<mode>): Likewise.
19488 2021-07-16 Richard Biener <rguenther@suse.de>
19490 PR tree-optimization/101467
19491 * tree-vect-stmts.c (vect_gen_while): Properly guard
19492 make_temp_ssa_name usage.
19494 2021-07-16 Cooper Qu <cooper.qu@linux.alibaba.com>
19496 * config.gcc: Don't use forked print-sysroot-suffix.sh and
19497 t-sysroot-suffix for C-SKY.
19498 * config/csky/print-sysroot-suffix.sh: Delete.
19499 * config/csky/t-csky-linux: Delete.
19500 * config/csky/t-sysroot-suffix: Define MULTILIB_DIRNAMES
19501 instead of CSKY_MULTILIB_DIRNAMES.
19503 2021-07-16 Richard Biener <rguenther@suse.de>
19505 * tree-vect-loop.c (vect_transform_cycle_phi): Correct sign
19506 conversion issues with the partial reduction of the reused
19507 vector accumulator.
19509 2021-07-16 Richard Biener <rguenther@suse.de>
19511 * config/i386/i386-options.c (ix86_option_override_internal): Set
19512 param_vect_partial_vector_usage to zero if not set.
19514 2021-07-15 Uroš Bizjak <ubizjak@gmail.com>
19517 * config/i386/i386.h (VALID_SSE_REG_MODE): Add TDmode.
19518 (VALID_INT_MODE_P): Add SDmode and DDmode.
19519 Add TDmode for TARGET_64BIT.
19520 (VALID_DFP_MODE_P): Remove.
19521 * config/i386/i386.c (ix86_hard_regno_mode_ok):
19522 Do not use VALID_DFP_MODE_P.
19524 2021-07-15 Andrew MacLeod <amacleod@redhat.com>
19526 * gimple-range-fold.cc (adjust_pointer_diff_expr): Use
19528 (fold_using_range::fold_stmt): Ditto.
19529 (fold_using_range::range_of_range_op): Ditto.
19530 (fold_using_range::range_of_phi): Ditto.
19531 (fold_using_range::range_of_call): Ditto.
19532 (fold_using_range::range_of_builtin_ubsan_call): Ditto.
19533 (fold_using_range::range_of_builtin_call): Ditto.
19534 (fold_using_range::range_of_cond_expr): Ditto.
19535 * gimple-range-fold.h (gimple_range_type): New.
19537 2021-07-15 Martin Sebor <msebor@redhat.com>
19539 PR middle-end/97027
19540 * tree-ssa-strlen.c (handle_assign): New function.
19541 (maybe_warn_overflow): Add argument.
19542 (nonzero_bytes_for_type): New function.
19543 (count_nonzero_bytes): Handle more tree types. Call
19544 nonzero_bytes_for_tye.
19545 (count_nonzero_bytes): Handle types.
19546 (handle_store): Handle stores from function calls.
19547 (strlen_check_and_optimize_call): Move code to handle_assign. Call
19548 it for assignments from function calls.
19550 2021-07-15 David Malcolm <dmalcolm@redhat.com>
19555 * doc/invoke.texi: Add -Wanalyzer-use-of-uninitialized-value.
19557 2021-07-15 David Malcolm <dmalcolm@redhat.com>
19559 * doc/invoke.texi (-fdump-analyzer-exploded-paths): New.
19561 2021-07-15 Martin Sebor <msebor@redhat.com>
19565 * fold-const.c (operand_compare::operand_equal_p): Handle OEP_DECL_NAME.
19566 (operand_compare::verify_hash_value): Same.
19567 * tree-core.h (OEP_DECL_NAME): New.
19569 2021-07-15 Martin Jambor <mjambor@suse.cz>
19571 * profile-count.h (profile_count::value): Change the return type to
19573 * gimple-pretty-print.c (dump_gimple_bb_header): Adjust print
19575 * tree-cfg.c (dump_function_to_file): Likewise.
19577 2021-07-15 Bill Schmidt <wschmidt@linux.ibm.com>
19580 * config/rs6000/rs6000-p8swap.c (has_part_mult): New.
19581 (rs6000_analyze_swaps): Insns containing a subreg of a mult are
19584 2021-07-15 Richard Biener <rguenther@suse.de>
19586 * tree-vectorizer.h (vect_gen_while): Match up with
19587 vect_gen_while_not.
19588 * tree-vect-stmts.c (vect_gen_while): Adjust API to that
19589 of vect_gen_while_not.
19590 (vect_gen_while_not): Adjust.
19591 * tree-vect-loop-manip.c (vect_set_loop_controls_directly): Likewise.
19593 2021-07-15 Aldy Hernandez <aldyh@redhat.com>
19595 * gimple-range-cache.cc (non_null_ref::adjust_range): New.
19596 (ranger_cache::range_of_def): Call adjust_range.
19597 (ranger_cache::entry_range): Same.
19598 * gimple-range-cache.h (non_null_ref::adjust_range): New.
19599 * gimple-range.cc (gimple_ranger::range_of_expr): Call
19601 (gimple_ranger::range_on_entry): Same.
19603 2021-07-15 Tamar Christina <tamar.christina@arm.com>
19606 2021-07-14 Tamar Christina <tamar.christina@arm.com>
19608 * config/arm/neon.md (<sup>dot_prod<vsi2qi>): Drop statements.
19610 2021-07-15 Tamar Christina <tamar.christina@arm.com>
19613 2021-07-14 Tamar Christina <tamar.christina@arm.com>
19615 * config/aarch64/aarch64-simd-builtins.def (udot, sdot): Rename to...
19616 (sdot_prod, udot_prod): ...These.
19617 * config/aarch64/aarch64-simd.md (<sur>dot_prod<vsi2qi>): Remove.
19618 (aarch64_<sur>dot<vsi2qi>): Rename to...
19619 (<sur>dot_prod<vsi2qi>): ...This.
19620 * config/aarch64/arm_neon.h (vdot_u32, vdotq_u32, vdot_s32, vdotq_s32):
19623 2021-07-15 Jakub Jelinek <jakub@redhat.com>
19625 PR middle-end/101437
19626 * gimplify.c (gimplify_expr): Throw away volatile reads from empty
19627 types even if they have non-BLKmode TYPE_MODE.
19629 2021-07-15 Richard Biener <rguenther@suse.de>
19632 * gcc.c (process_command): Process -gtoggle like process_options
19633 would after parsing options.
19635 2021-07-15 Trevor Saunders <tbsaunde@tbsaunde.org>
19637 * cfgexpand.c (expand_asm_loc): Adjust.
19638 (expand_asm_stmt): Likewise.
19639 * config/arm/aarch-common-protos.h (arm_md_asm_adjust): Likewise.
19640 * config/arm/aarch-common.c (arm_md_asm_adjust): Likewise.
19641 * config/arm/arm.c (thumb1_md_asm_adjust): Likewise.
19642 * config/avr/avr.c (avr_md_asm_adjust): Likewise.
19643 * config/cris/cris.c (cris_md_asm_adjust): Likewise.
19644 * config/i386/i386.c (ix86_md_asm_adjust): Likewise.
19645 * config/mn10300/mn10300.c (mn10300_md_asm_adjust): Likewise.
19646 * config/nds32/nds32.c (nds32_md_asm_adjust): Likewise.
19647 * config/pdp11/pdp11.c (pdp11_md_asm_adjust): Likewise.
19648 * config/rs6000/rs6000.c (rs6000_md_asm_adjust): Likewise.
19649 * config/s390/s390.c (s390_md_asm_adjust): Likewise.
19650 * config/vax/vax.c (vax_md_asm_adjust): Likewise.
19651 * config/visium/visium.c (visium_md_asm_adjust): Likewise.
19652 * doc/tm.texi: Regenerate.
19653 * target.def: Add location argument to md_asm_adjust.
19655 2021-07-15 Trevor Saunders <tbsaunde@tbsaunde.org>
19657 * tree-diagnostic.c (diagnostic_report_current_function): Use the
19658 diagnostic's location, not input_location.
19660 2021-07-15 Trevor Saunders <tbsaunde@tbsaunde.org>
19662 * cfgexpand.c (tree_conflicts_with_clobbers_p): Pass location to
19664 (expand_asm_stmt): Likewise.
19666 2021-07-14 Peter Bergner <bergner@linux.ibm.com>
19668 * config/rs6000/rs6000.c (adjacent_mem_locations): Return the lower
19669 addressed memory rtx, if any.
19670 (rs6000_split_multireg_move): Fix code formatting.
19671 Handle MMA build built-ins with operands in adjacent memory locations.
19673 2021-07-14 Peter Bergner <bergner@linux.ibm.com>
19675 * config/rs6000/rs6000.c (rs6000_split_multireg_move): Move to later
19678 2021-07-14 Jason Merrill <jason@redhat.com>
19680 * sel-sched-ir.h (get_all_loop_exits): Use auto_vec.
19682 2021-07-14 Jason Merrill <jason@redhat.com>
19684 * doc/invoke.texi: -fdelete-dead-exceptions is on by default for
19687 2021-07-14 Tamar Christina <tamar.christina@arm.com>
19689 * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
19690 Remove erroneous line.
19692 2021-07-14 Andrew MacLeod <amacleod@redhat.com>
19694 * params.opt (param_evrp_mode): Change default.
19696 2021-07-14 Tamar Christina <tamar.christina@arm.com>
19698 * config/aarch64/aarch64-simd-builtins.def (udot, sdot): Rename to...
19699 (sdot_prod, udot_prod): ...These.
19700 * config/aarch64/aarch64-simd.md (<sur>dot_prod<vsi2qi>): Remove.
19701 (aarch64_<sur>dot<vsi2qi>): Rename to...
19702 (<sur>dot_prod<vsi2qi>): ...This.
19703 * config/aarch64/arm_neon.h (vdot_u32, vdotq_u32, vdot_s32, vdotq_s32):
19706 2021-07-14 Tamar Christina <tamar.christina@arm.com>
19708 * config/arm/neon.md (<sup>dot_prod<vsi2qi>): Drop statements.
19710 2021-07-14 Tamar Christina <tamar.christina@arm.com>
19712 * doc/sourcebuild.texi (arm_v8_2a_i8mm_neon_hw): Document.
19714 2021-07-14 Tamar Christina <tamar.christina@arm.com>
19716 * config/arm/neon.md (usdot_prod<vsi2qi>): New.
19718 2021-07-14 Tamar Christina <tamar.christina@arm.com>
19720 * config/aarch64/aarch64-simd.md (aarch64_usdot<vsi2qi>): Rename to...
19721 (usdot_prod<vsi2qi>): ... This.
19722 * config/aarch64/aarch64-simd-builtins.def (usdot): Rename to...
19723 (usdot_prod): ...This.
19724 * config/aarch64/arm_neon.h (vusdot_s32, vusdotq_s32): Likewise.
19725 * config/aarch64/aarch64-sve.md (@aarch64_<sur>dot_prod<vsi2qi>):
19727 (@<sur>dot_prod<vsi2qi>): ...This.
19728 * config/aarch64/aarch64-sve-builtins-base.cc
19729 (svusdot_impl::expand): Use it.
19731 2021-07-14 Tamar Christina <tamar.christina@arm.com>
19733 * optabs.def (usdot_prod_optab): New.
19734 * doc/md.texi: Document it and clarify other dot prod optabs.
19735 * optabs-tree.h (enum optab_subtype): Add optab_vector_mixed_sign.
19736 * optabs-tree.c (optab_for_tree_code): Support usdot_prod_optab.
19737 * optabs.c (expand_widen_pattern_expr): Likewise.
19738 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
19739 * tree-vect-loop.c (vectorizable_reduction): Query dot-product kind.
19740 * tree-vect-patterns.c (vect_supportable_direct_optab_p): Take optional
19742 (vect_widened_op_tree): Optionally ignore
19744 (vect_recog_dot_prod_pattern): Support usdot_prod_optab.
19746 2021-07-14 H.J. Lu <hjl.tools@gmail.com>
19749 * config/i386/driver-i386.c (host_detect_local_cpu): Check
19750 "arch [32|64]" and "tune [32|64]" for 32-bit and 64-bit codegen.
19751 Enable UINTR only for 64-bit codegen.
19752 * config/i386/i386-options.c
19753 (ix86_option_override_internal::DEF_PTA): Skip PTA_UINTR if not
19755 * config/i386/i386.h (ARCH_ARG): New.
19756 (CC1_CPU_SPEC): Pass "[arch|tune] 32" for 32-bit codegen and
19757 "[arch|tune] 64" for 64-bit codegen.
19759 2021-07-14 Richard Biener <rguenther@suse.de>
19761 PR tree-optimization/101445
19762 * tree-vect-stmts.c (vectorizable_load): Do the gap adjustment
19763 of the IV in the correct direction for negative stride
19766 2021-07-14 Jakub Jelinek <jakub@redhat.com>
19769 * godump.c (godump_str_hash): New type.
19770 (godump_container::pot_dummy_types): Use string_hash instead of
19771 ptr_hash in the hash_set.
19773 2021-07-14 Richard Biener <rguenther@suse.de>
19775 * tree-vect-loop.c (vect_find_reusable_accumulator): Handle
19776 vector types where the old vector type has a multiple of
19777 the new vector type elements.
19778 (vect_create_partial_epilog): New function, split out from...
19779 (vect_create_epilog_for_reduction): ... here.
19780 (vect_transform_cycle_phi): Reduce the re-used accumulator
19781 to the new vector type.
19783 2021-07-14 Alexandre Oliva <oliva@adacore.com>
19785 * tree-ssa-alias.c (attr_fnspec::verify): Fix index in
19786 non-'t'-sized arg check.
19788 2021-07-14 Alexandre Oliva <oliva@adacore.com>
19790 * tree-cfg.c (cleanup_dead_labels_eh): Update
19791 post_landing_pad label upon change of landing pad block's
19793 (cleanup_dead_labels): Check that a removed label is not that
19796 2021-07-13 Jonathan Wright <jonathan.wright@arm.com>
19798 * combine.c (combine_simplify_rtx): Add vec_select -> subreg
19800 * config/aarch64/aarch64.md (*zero_extend<SHORT:mode><GPI:mode>2_aarch64):
19801 Add Neon to general purpose register case for zero-extend
19803 * config/arm/vfp.md (*arm_movsi_vfp): Remove "*" from *t -> r
19804 case to prevent some cases opting to go through memory.
19805 * cse.c (fold_rtx): Add vec_select -> subreg simplification.
19806 * rtl.c (rtvec_series_p): Define predicate to determine
19807 whether a vector contains a linear series of integers.
19808 * rtl.h (rtvec_series_p): Define.
19809 * rtlanal.c (vec_series_lowpart_p): Define predicate to
19810 determine if a vector selection is equivalent to the low part
19812 * rtlanal.h (vec_series_lowpart_p): Define.
19813 * simplify-rtx.c (simplify_context::simplify_binary_operation_1):
19814 Add vec_select -> subreg simplification.
19816 2021-07-13 Paul A. Clarke <pc@us.ibm.com>
19818 * config/rs6000/smmintrin.h (_mm_testz_si128, _mm_testc_si128,
19819 _mm_testnzc_si128, _mm_test_all_ones, _mm_test_all_zeros,
19820 _mm_test_mix_ones_zeros): New.
19822 2021-07-13 Roger Sayle <roger@nextmovesoftware.com>
19823 Richard Biener <rguenther@suse.de>
19825 * gimple.c (gimple_could_trap_p_1): Make S argument a
19826 "const gimple*". Preserve constness in call to
19827 gimple_asm_volatile_p.
19828 (gimple_could_trap_p): Make S argument a "const gimple*".
19829 * gimple.h (gimple_could_trap_p_1, gimple_could_trap_p):
19830 Update function prototypes.
19832 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
19834 * tree-vectorizer.h (vect_reusable_accumulator): New structure.
19835 (_loop_vec_info::main_loop_edge): New field.
19836 (_loop_vec_info::skip_main_loop_edge): Likewise.
19837 (_loop_vec_info::skip_this_loop_edge): Likewise.
19838 (_loop_vec_info::reusable_accumulators): Likewise.
19839 (_stmt_vec_info::reduc_scalar_results): Likewise.
19840 (_stmt_vec_info::reused_accumulator): Likewise.
19841 (vect_get_main_loop_result): Declare.
19842 * tree-vectorizer.c (vec_info::new_stmt_vec_info): Initialize
19843 reduc_scalar_inputs.
19844 (vec_info::free_stmt_vec_info): Free reduc_scalar_inputs.
19845 * tree-vect-loop-manip.c (vect_get_main_loop_result): New function.
19846 (vect_do_peeling): Fill an epilogue loop's main_loop_edge,
19847 skip_main_loop_edge and skip_this_loop_edge fields.
19848 * tree-vect-loop.c (INCLUDE_ALGORITHM): Define.
19849 (vect_emit_reduction_init_stmts): New function.
19850 (get_initial_def_for_reduction): Use it.
19851 (get_initial_defs_for_reduction): Likewise. Change the vinfo
19852 parameter to a loop_vec_info.
19853 (vect_create_epilog_for_reduction): Store the scalar results
19854 in the reduc_info. If an epilogue loop is reusing an accumulator
19855 from the main loop, and if the epilogue loop can also be skipped,
19856 try to place the reduction code in the join block. Record
19857 accumulators that could potentially be reused by epilogue loops.
19858 (vect_transform_cycle_phi): When vectorizing epilogue loops,
19859 try to reuse accumulators from the main loop. Record the initial
19860 value in reduc_info for non-SLP reductions too.
19862 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
19864 * tree-vect-loop.c (get_initial_def_for_reduction): Remove
19865 adjustment handling. Take the neutral value as an argument,
19866 in place of the code argument.
19867 (vect_transform_cycle_phi): Update accordingly. Handle the
19868 initial values of cond reductions separately from code reductions.
19869 Choose the adjustment here rather than in
19870 get_initial_def_for_reduction. Sink the splat of vec_initial_def.
19872 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
19874 * tree-vect-loop.c (neutral_op_for_slp_reduction): Replace with...
19875 (neutral_op_for_reduction): ...this, providing a more general
19877 (vect_create_epilog_for_reduction): Update accordingly.
19878 (vectorizable_reduction): Likewise.
19879 (vect_transform_cycle_phi): Likewise.
19881 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
19883 * tree-vect-loop.c (get_initial_def_for_reduction): Take the
19884 reduc_info instead of the original stmt_vec_info.
19885 (vect_transform_cycle_phi): Update accordingly.
19887 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
19889 * tree-vect-loop.c (get_initial_defs_for_reduction): Take the
19890 reduc_info as an additional parameter.
19891 (vect_transform_cycle_phi): Update accordingly.
19893 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
19895 * tree-vectorizer.h: Include tree-ssa-operands.h.
19896 (vect_phi_initial_value): New function.
19897 * tree-vect-loop.c (neutral_op_for_slp_reduction): Use it.
19898 (get_initial_defs_for_reduction, info_for_reduction): Likewise.
19899 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
19900 (vect_transform_cycle_phi, vectorizable_induction): Likewise.
19902 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
19904 * tree-vect-loop.c (vect_create_epilog_for_reduction): Convert
19905 the phi results to vectype after creating them. Remove later
19906 conversion code that thus becomes redundant.
19908 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
19910 * tree-vect-loop.c (vect_create_epilog_for_reduction): Replace
19911 the new_phis vector with a reduc_inputs vector. Combine handling
19912 of reduction chains and ncopies > 1.
19914 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
19916 * tree-vect-loop.c (vect_create_epilog_for_reduction): Truncate
19917 scalar_results to group_size elements after reducing down from
19918 N*group_size elements. Construct an array_slice of the live-out
19919 stmts and assert that there is one stmt per scalar result.
19921 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
19923 * tree-vect-loop.c (vect_create_epilog_for_reduction): Remove
19924 nested_in_vect_loop and use double_reduc everywhere. Remove dead
19925 assignment to "loop".
19927 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
19929 * internal-fn.c (vectorized_internal_fn_supported_p): Handle
19930 vector types first. For scalar types, consider both the preferred
19931 vector mode and the alternative vector modes.
19932 * optabs-query.c (can_vec_mask_load_store_p): Use the same
19933 structure as above, in particular using related_vector_mode
19934 for modes provided by autovectorize_vector_modes.
19936 2021-07-13 Jakub Jelinek <jakub@redhat.com>
19937 Richard Biener <rguenther@suse.de>
19939 PR tree-optimization/101419
19940 * tree-pass.h (PROP_objsz): Define.
19941 (make_pass_early_object_sizes): Declare.
19942 * passes.def (pass_all_early_optimizations): Rename pass_object_sizes
19943 there to pass_early_object_sizes, drop parameter.
19944 (pass_all_optimizations): Move pass_object_sizes right after pass_ccp,
19945 drop parameter, move pass_post_ipa_warn right after that.
19946 * tree-object-size.c (pass_object_sizes::execute): Rename to...
19947 (object_sizes_execute): ... this. Add insert_min_max_p argument.
19948 (pass_data_object_sizes): Move after object_sizes_execute.
19949 (pass_object_sizes): Likewise. In execute method call
19950 object_sizes_execute, drop set_pass_param method and insert_min_max_p
19951 non-static data member and its initializer in the ctor.
19952 (pass_data_early_object_sizes, pass_early_object_sizes,
19953 make_pass_early_object_sizes): New.
19954 * tree-ssa-sccvn.c (copy_reference_ops_from_ref): Use
19955 (cfun->curr_properties & PROP_objsz) instead of cfun->after_inlining.
19957 2021-07-13 Kito Cheng <kito.cheng@sifive.com>
19960 * config/riscv/constraints.md ("S"): Update description and remove
19962 * doc/md.texi (Machine Constraints): Document the 'S' constraints
19965 2021-07-13 Richard Biener <rguenther@suse.de>
19968 2021-07-12 Richard Biener <rguenther@suse.de>
19970 * tree-vect-slp.c (vect_slp_region): Show the number of
19971 SLP graph entries in the optimization message.
19973 2021-07-13 Michael Meissner <meissner@linux.ibm.com>
19975 * config/rs6000/altivec.md (xxspltiw_v4sf): Change local variable
19977 * config/rs6000/rs6000-protos.h (rs6000_const_f32_to_i32): Change
19978 return type to long.
19979 * config/rs6000/rs6000.c (rs6000_const_f32_to_i32): Change return
19982 2021-07-12 Andrew MacLeod <amacleod@redhat.com>
19984 * gimple-range-fold.cc (fold_using_range::range_of_builtin_ubsan_call):
19985 Query relation between the 2 operands and use it.
19987 2021-07-12 Sergei Trofimovich <siarheit@google.com>
19989 * doc/cfg.texi: Fix s/ei_safe_safe/ei_safe_edge/ typo.
19991 2021-07-12 Uroš Bizjak <ubizjak@gmail.com>
19994 * config/i386/predicates.md (vec_setm_sse41_operand):
19995 Rename from vec_setm_operand.
19996 (vec_setm_avx2_operand): New predicate.
19997 * config/i386/sse.md (vec_set<V_128:mode>): Use V_128 mode iterator.
19998 Use vec_setm_sse41_operand as operand 2 predicate.
19999 (vec_set<V_256_512:mode): New expander.
20000 * config/i386/mmx.md (vec_setv2hi): Use vec_setm_sse41_operand
20001 as operand 2 predicate.
20003 2021-07-12 Andrew MacLeod <amacleod@redhat.com>
20005 PR tree-optimization/101335
20006 * range-op.cc (operator_cast::lhs_op1_relation): Delete.
20008 2021-07-12 Andrew Pinski <apinski@marvell.com>
20010 * tree-ssa-phiopt.c (match_simplify_replacement): Move
20011 insert of the sequence before the movement of the
20012 statement. Check if to see if the statement is used
20013 outside of the original phi to see if we should move it.
20015 2021-07-12 Richard Biener <rguenther@suse.de>
20017 * dump-context.h (debug_dump_context::debug_dump_context):
20018 Add FILE * parameter defaulted to stderr.
20019 * dumpfile.c (debug_dump_context::debug_dump_context): Adjust.
20020 * tree-vect-slp.c (dot_slp_tree): New functions.
20022 2021-07-12 Richard Biener <rguenther@suse.de>
20024 PR tree-optimization/101373
20025 * tree-ssa-pre.c (prune_clobbered_mems): Also prune trapping
20026 references when the BB may not return.
20027 (compute_avail): Pass in the function we're working on and
20028 replace cfun references with it. Externally throwing
20029 const calls also possibly terminate the function.
20030 (pass_pre::execute): Pass down the function we're working on.
20031 * gcse.c (compute_hash_table_work): Externally throwing
20032 const/pure calls also need record_last_mem_set_info.
20033 * postreload-gcse.c (record_opr_changes): Looping or externally
20034 throwing const/pure calls also need record_last_mem_set_info.
20036 2021-07-12 Uroš Bizjak <ubizjak@gmail.com>
20038 * recog.c (memory_address_addr_space_p): Change the type to bool.
20039 Return true/false instead of 1/0.
20040 (offsettable_memref_p): Ditto.
20041 (offsettable_nonstrict_memref_p): Ditto.
20042 (offsettable_address_addr_space_p): Ditto.
20043 Change the type of addressp indirect function to bool.
20044 * recog.h (memory_address_addr_space_p): Change the type to bool.
20045 (strict_memory_address_addr_space_p): Ditto.
20046 (offsettable_memref_p): Ditto.
20047 (offsettable_nonstrict_memref_p): Ditto.
20048 (offsettable_address_addr_space_p): Ditto.
20049 * reload.c (maybe_memory_address_addr_space_p): Ditto.
20050 (strict_memory_address_addr_space_p): Change the type to bool.
20051 Return true/false instead of 1/0.
20052 (maybe_memory_address_addr_space_p): Change the type to bool.
20054 2021-07-12 Richard Biener <rguenther@suse.de>
20056 * tree-vect-slp.c (vect_slp_region): Show the number of
20057 SLP graph entries in the optimization message.
20059 2021-07-12 Richard Biener <rguenther@suse.de>
20061 PR tree-optimization/101394
20062 * tree-ssa-pre.c (do_pre_regular_insertion): Avoid inserting
20063 copies from abnormals for a full redundancy.
20065 2021-07-12 Richard Biener <rguenther@suse.de>
20067 PR middle-end/101423
20068 * gimple.c (gimple_could_trap_p_1): Internal function calls
20070 * tree-eh.c (tree_could_trap_p): Likewise.
20072 2021-07-12 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
20075 * config/arm/arm_neon.h (vmul_n_u32): Replace call to builtin with
20077 (vmulq_n_u32): Likewise.
20078 (vmul_n_f32): Gate __a * __b on __FAST_MATH__.
20079 (vmulq_n_f32): Likewise.
20080 (vmul_n_f16): Likewise.
20081 (vmulq_n_f16): Likewise.
20083 2021-07-12 Martin Liska <mliska@suse.cz>
20085 PR sanitizer/101425
20086 * gcc.c (check_offload_target_name): Call
20087 candidates_list_and_hint only if we have a candidate.
20089 2021-07-12 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
20092 * config/arm/neon.md (vec_init): Move to ...
20093 * config/arm/vec-common.md (vec_init): ... here.
20094 Change the pattern's mode to VDQX and gate it on VALID_MVE_MODE.
20096 2021-07-12 Roger Sayle <roger@nextmovesoftware.com>
20098 PR tree-optimization/101403
20099 * match.pd ((T)bswap(X)>>C): Correctly handle cases where
20100 signedness of the shift is not the same as the signedness of
20101 the type extension.
20103 2021-07-09 Roger Sayle <roger@nextmovesoftware.com>
20104 Uroš Bizjak <ubizjak@gmail.com>
20106 * config/i386/i386.md (*divmodsi4_const): Optimize SImode
20107 divmod of a constant numerator with new define_insn_and_split.
20109 2021-07-09 Iain Sandoe <iain@sandoe.co.uk>
20112 * config/i386/i386-expand.c (ix86_expand_call): If a call is
20113 to a non-local-binding, or local but to a public symbol, then
20114 assume that it might be indirected via the lazy symbol binder.
20115 Mark R10 and R10 as clobbered in that case.
20117 2021-07-09 Eric Botcazou <ebotcazou@adacore.com>
20120 * gcc.c (ASM_DEBUG_DWARF_OPTION): Set again to --gdwarf2 in
20121 the case where HAVE_AS_WORKING_DWARF_N_FLAG is not defined
20122 and HAVE_LD_BROKEN_PE_DWARF5 is defined.
20124 2021-07-09 Uroš Bizjak <ubizjak@gmail.com>
20126 * config/i386/i386.md (*udivmodsi4_pow2_zext_1): Limit the
20127 log2 range of operands[3] to [1,31].
20128 (*udivmodsi4_pow2_zext_2): Ditto. Correct insn RTX pattern.
20130 2021-07-09 Sergei Trofimovich <siarheit@google.com>
20132 * doc/md.texi: Don't split @smallexample in multiple @groups.
20134 2021-07-09 Sergei Trofimovich <siarheit@google.com>
20136 * doc/md.texi: Add missing 'see' word.
20138 2021-07-09 Andrew Pinski <apinski@marvell.com>
20140 * tree-ssa-phiopt.c (phiopt_early_allow): Change arguments
20141 to take sequence and gimple_match_op. Accept the case where
20142 op is a SSA_NAME and one statement in the sequence.
20143 Also allow constants.
20144 (gimple_simplify_phiopt): Always pass a sequence to resimplify.
20145 Update call to phiopt_early_allow. Discard the sequence if not
20148 2021-07-09 Xi Ruoyao <xry111@mengyan1223.wang>
20153 * config/mips/mips.c (mips_const_insns): Use MSA_SUPPORTED_MODE_P
20154 instead of ISA_HAS_MSA.
20155 (mips_expand_vec_unpack): Likewise.
20156 (mips_expand_vector_init): Likewise.
20158 2021-07-09 Kewen Lin <linkw@linux.ibm.com>
20160 * config/rs6000/vsx.md (mods_<mode>): Rename to...
20161 (mod<mode>3): ... this.
20162 (modu_<mode>): Rename to...
20163 (umod<mode>3): ... this.
20164 * config/rs6000/rs6000-builtin.def (MODS_V2DI, MODS_V4SI, MODU_V2DI,
20165 MODU_V4SI): Adjust.
20167 2021-07-08 Jeff Law <jeffreyalaw@gmail.com>
20169 * config/h8300/shiftrotate.md (variable shifts): Expose condition
20170 code handling for the test before the loop.
20172 2021-07-08 Martin Jambor <mjambor@suse.cz>
20175 * ipa-sra.c (class isra_call_summary): New member
20176 m_before_any_store, initialize it in the constructor.
20177 (isra_call_summary::dump): Dump the new field.
20178 (ipa_sra_call_summaries::duplicate): Copy it.
20179 (process_scan_results): Set it.
20180 (isra_write_edge_summary): Stream it.
20181 (isra_read_edge_summary): Likewise.
20182 (param_splitting_across_edge): Only override
20183 safe_to_import_accesses if m_before_any_store is set.
20185 2021-07-08 Martin Sebor <msebor@redhat.com>
20187 PR bootstrap/101374
20188 * gimple-array-bounds.cc (array_bounds_checker::check_mem_ref):
20189 Use Object Size Type 0 instead of 1.
20191 2021-07-08 Richard Sandiford <richard.sandiford@arm.com>
20193 * tree-vect-loop.c (vectorizable_reduction): Remove always-true
20196 2021-07-08 Richard Sandiford <richard.sandiford@arm.com>
20198 * match.pd: Simplify an extend-operate-truncate sequence involving
20201 2021-07-08 Roger Sayle <roger@nextmovesoftware.com>
20202 Richard Biener <rguenther@suse.de>
20204 PR tree-optimization/40210
20205 * match.pd (bswap optimizations): Simplify (bswap(x)>>C1)&C2 as
20206 (x>>C3)&C2 when possible. Simplify bswap(x)>>C1 as ((T)x)>>C2
20207 when possible. Simplify bswap(x)&C1 as (x>>C2)&C1 when 0<=C1<=255.
20209 2021-07-08 Uroš Bizjak <ubizjak@gmail.com>
20212 * config/i386/i386-expand.c (ix86_expand_sse_unpack):
20214 * config/i386/mmx.md (V_32): New mode iterator.
20215 (mov<V_32:mode>): Use V_32 mode iterator.
20216 (*mov<V_32:mode>_internal): Ditto.
20217 (*push<V_32:mode>2_rex64): Ditto.
20218 (*push<V_32:mode>2): Ditto.
20219 (movmisalign<V_32:mode>): Ditto.
20220 (mmx_<any_shiftrt:insn>v1si3): New insn pattern.
20221 (sse4_1_<any_extend:code>v2qiv2hi2): Ditto.
20222 (vec_unpacks_lo_v4qi): New expander.
20223 (vec_unpacks_hi_v4qi): Ditto.
20224 (vec_unpacku_lo_v4qi): Ditto.
20225 (vec_unpacku_hi_v4qi): Ditto.
20226 * config/i386/i386.h (VALID_SSE2_REG_MODE): Add V1SImode.
20227 (VALID_INT_MODE_P): Ditto.
20229 2021-07-08 Michael Meissner <meissner@linux.ibm.com>
20232 * config/rs6000/rs6000.md (udivti3): New insn.
20233 (divti3): New insn.
20234 (umodti3): New insn.
20235 (modti3): New insn.
20237 2021-07-07 Martin Sebor <msebor@redhat.com>
20239 PR tree-optimization/100137
20240 PR tree-optimization/99121
20241 PR tree-optimization/97027
20242 * builtins.c (access_ref::access_ref): Also set offmax.
20243 (access_ref::offset_in_range): Define new function.
20244 (access_ref::add_offset): Set offmax.
20245 (access_ref::inform_access): Handle access_none.
20246 (handle_mem_ref): Clear ostype.
20247 (compute_objsize_r): Handle ASSERT_EXPR.
20248 * builtins.h (struct access_ref): Add offmax member.
20249 * gimple-array-bounds.cc (array_bounds_checker::check_mem_ref): Use
20250 compute_objsize() and simplify.
20252 2021-07-07 Peter Bergner <bergner@linux.ibm.com>
20254 * config/rs6000/rs6000-call.c (mma_init_builtins): Use VSX_BUILTIN_LXVP
20255 and VSX_BUILTIN_STXVP.
20257 2021-07-07 Martin Sebor <msebor@redhat.com>
20260 * config/aarch64/aarch64.c (aarch64_simd_lane_bounds): Remove
20261 a stray %K from error_at() missed in r12-2088.
20263 2021-07-07 Richard Biener <rguenther@suse.de>
20265 PR tree-optimization/99728
20266 * tree-ssa-loop-im.c (gather_mem_refs_stmt): Record
20268 (mem_refs_may_alias_p): Add assert we handled aggregate
20270 (sm_seq_valid_bb): Give up when running into aggregate copies.
20271 (ref_indep_loop_p): Handle aggregate copies as never
20272 being invariant themselves but allow other refs to be
20273 disambiguated against them.
20274 (can_sm_ref_p): Do not try to apply store-motion to aggregate
20277 2021-07-06 Indu Bhagat <indu.bhagat@oracle.com>
20280 * dwarf2ctf.c (ctf_get_AT_data_member_location): Multiply by 8 to get
20283 2021-07-06 Martin Sebor <msebor@redhat.com>
20285 * gimple-pretty-print.c (percent_G_format): Remove.
20286 * tree-diagnostic.c (default_tree_printer): Remove calls.
20287 * tree-pretty-print.c (percent_K_format): Remove.
20288 * tree-pretty-print.h (percent_K_format): Remove.
20290 2021-07-06 Martin Sebor <msebor@redhat.com>
20292 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_builtin):
20293 Remove %K and use error_at.
20294 (aarch64_expand_fcmla_builtin): Same.
20295 (aarch64_expand_builtin_tme): Same.
20296 (aarch64_expand_builtin_memtag): Same.
20297 * config/arm/arm-builtins.c (arm_expand_acle_builtin): Same.
20298 (arm_expand_builtin): Same.
20299 * config/arm/arm.c (bounds_check): Same.
20301 2021-07-06 Martin Sebor <msebor@redhat.com>
20303 * builtins.c (warn_string_no_nul): Remove %G.
20304 (maybe_warn_for_bound): Same.
20305 (warn_for_access): Same.
20306 (check_access): Same.
20307 (check_strncat_sizes): Same.
20308 (expand_builtin_strncat): Same.
20309 (expand_builtin_strncmp): Same.
20310 (expand_builtin): Same.
20311 (expand_builtin_object_size): Same.
20312 (warn_dealloc_offset): Same.
20313 (maybe_emit_free_warning): Same.
20314 * calls.c (maybe_warn_alloc_args_overflow): Same.
20315 (maybe_warn_nonstring_arg): Same.
20316 (maybe_warn_rdwr_sizes): Same.
20317 * expr.c (expand_expr_real_1): Remove %K.
20318 * gimple-fold.c (gimple_fold_builtin_strncpy): Remove %G.
20319 (gimple_fold_builtin_strncat): Same.
20320 * gimple-ssa-sprintf.c (format_directive): Same.
20321 (handle_printf_call): Same.
20322 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Same.
20323 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Same.
20324 (maybe_diag_access_bounds): Same. Call gimple_location.
20325 (check_bounds_or_overlap): Same.
20326 * trans-mem.c (ipa_tm_scan_irr_block): Remove %K. Simplify.
20327 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Remove %G.
20328 * tree-ssa-strlen.c (maybe_warn_overflow): Same.
20329 (maybe_diag_stxncpy_trunc): Same.
20330 (handle_builtin_stxncpy_strncat): Same.
20331 (maybe_warn_pointless_strcmp): Same.
20332 * tree-ssa-uninit.c (maybe_warn_operand): Same.
20334 2021-07-06 Uroš Bizjak <ubizjak@gmail.com>
20337 * config/i386/predicates.md (vec_setm_operand): Enable
20338 register_operand for TARGET_SSE4_1.
20339 * config/i386/mmx.md (vec_setv2hi): Use vec_setm_operand
20340 as operand 2 predicate. Call ix86_expand_vector_set_var
20341 for non-constant index operand.
20342 (vec_setv4qi): Use vec_setm_mmx_operand as operand 2 predicate.
20343 Call ix86_expand_vector_set_var for non-constant index operand.
20345 2021-07-06 Jeff Law <jeffreyalaw@gmail.com>
20347 * config/h8300/jumpcall.md (*branch): When possible, generate
20348 the comparison in CCZN mode.
20349 * config/h8300/predicates.md (simple_memory_operand): Reject all
20350 auto-increment addressing modes.
20352 2021-07-06 Iain Sandoe <iain@sandoe.co.uk>
20354 PR bootstrap/100246
20355 * config/i386/i386.h (struct stringop_algs): Define a CTOR for
20358 2021-07-06 Richard Biener <rguenther@suse.de>
20360 * doc/md.texi (vec_fmaddsub<mode>4): Document.
20361 (vec_fmsubadd<mode>4): Likewise.
20362 * optabs.def (vec_fmaddsub$a4): Add.
20363 (vec_fmsubadd$a4): Likewise.
20364 * internal-fn.def (IFN_VEC_FMADDSUB): Add.
20365 (IFN_VEC_FMSUBADD): Likewise.
20366 * tree-vect-slp-patterns.c (addsub_pattern::recognize):
20367 Refactor to handle IFN_VEC_FMADDSUB and IFN_VEC_FMSUBADD.
20368 (addsub_pattern::build): Likewise.
20369 * tree-vect-slp.c (vect_optimize_slp): CFN_VEC_FMADDSUB
20370 and CFN_VEC_FMSUBADD are not transparent for permutes.
20371 * config/i386/sse.md (vec_fmaddsub<mode>4): New expander.
20372 (vec_fmsubadd<mode>4): Likewise.
20374 2021-07-06 Richard Biener <rguenther@suse.de>
20376 * doc/invoke.texi (fmove-loop-stores): Document.
20377 * common.opt (fmove-loop-stores): New option.
20378 * opts.c (default_options_table): Enable -fmove-loop-stores
20379 at -O1 but not -Og.
20380 * tree-ssa-loop-im.c (pass_lim::execute): Pass
20381 flag_move_loop_stores instead of true to
20382 loop_invariant_motion_in_fun.
20384 2021-07-06 Iain Sandoe <iain@sandoe.co.uk>
20386 * doc/install.texi: Document --with-dsymutil.
20388 2021-07-06 Andrew Pinski <apinski@marvell.com>
20390 PR tree-optimization/101256
20391 * dbgcnt.def (phiopt_edge_range): New counter.
20392 * tree-ssa-phiopt.c (replace_phi_edge_with_variable):
20393 Check to make sure the new name is defined in the same
20394 bb as the conditional before duplicating range info.
20395 Also add debug counter.
20397 2021-07-06 Kewen Lin <linkw@linux.ibm.com>
20399 PR rtl-optimization/100328
20400 * config/i386/i386-options.c (ix86_option_override_internal):
20401 Set param_ira_consider_dup_in_all_alts to 0.
20403 2021-07-06 Kewen Lin <linkw@linux.ibm.com>
20405 PR rtl-optimization/100328
20406 * doc/invoke.texi (ira-consider-dup-in-all-alts): Document new
20408 * ira.c (ira_get_dup_out_num): Adjust as parameter
20409 param_ira_consider_dup_in_all_alts.
20410 * params.opt (ira-consider-dup-in-all-alts): New.
20411 * ira-conflicts.c (process_regs_for_copy): Add one parameter
20412 single_input_op_has_cstr_p.
20413 (get_freq_for_shuffle_copy): New function.
20414 (add_insn_allocno_copies): Adjust as single_input_op_has_cstr_p.
20415 * ira-int.h (ira_get_dup_out_num): Add one bool parameter.
20417 2021-07-05 Jeff Law <jeffreyalaw@gmail.com>
20419 * config/h8300/shiftrotate.md (shift-by-variable patterns): Update to
20420 generate condition code aware RTL directly.
20422 2021-07-05 Andrew Pinski <apinski@marvell.com>
20424 PR tree-optimization/101039
20425 * match.pd (A CMP 0 ? A : -A): New patterns.
20426 * tree-ssa-phiopt.c (abs_replacement): Delete function.
20427 (tree_ssa_phiopt_worker): Don't call abs_replacement.
20428 Update comment about abs_replacement.
20430 2021-07-05 Andrew Pinski <apinski@marvell.com>
20432 * tree-ssa-phiopt.c (gimple_simplify_phiopt):
20433 If "A ? B : C" fails to simplify, try "(!A) ? C : B".
20435 2021-07-05 Andrew Pinski <apinski@marvell.com>
20437 * tree-ssa-phiopt.c (match_simplify_replacement):
20438 Add early_p argument. Call gimple_simplify_phiopt
20439 instead of gimple_simplify.
20440 (tree_ssa_phiopt_worker): Update call to
20441 match_simplify_replacement and allow unconditionally.
20442 (phiopt_early_allow): New function.
20443 (gimple_simplify_phiopt): New function.
20445 2021-07-05 Andrew Pinski <apinski@marvell.com>
20447 PR middle-end/101237
20448 * fold-const.c (negate_expr_p): Remove call to element_mode
20449 and TREE_MODE/TREE_TYPE when calling HONOR_SIGNED_ZEROS,
20450 HONOR_SIGN_DEPENDENT_ROUNDING, and HONOR_SNANS.
20451 (fold_negate_expr_1): Likewise.
20452 (const_unop): Likewise.
20453 (fold_cond_expr_with_comparison): Likewise.
20454 (fold_binary_loc): Likewise.
20455 (fold_ternary_loc): Likewise.
20456 (tree_call_nonnegative_warnv_p): Likewise.
20457 * match.pd (-(A + B) -> (-B) - A): Likewise.
20459 2021-07-05 Iain Sandoe <iain@sandoe.co.uk>
20461 * configure.ac: Handle --with-dsymutil in the same way as we
20462 do for the assembler and linker. (DEFAULT_DSYMUTIL): New.
20463 Extract the type and version for the dsymutil configured or
20464 found by the default searches.
20465 * config.in: Regenerated.
20466 * configure: Regenerated.
20467 * collect2.c (do_dsymutil): Handle locating dsymutil in the
20468 same way as for the assembler and linker.
20469 * config/darwin.h (DSYMUTIL): Delete.
20470 * gcc.c: Report a configured dsymutil correctly.
20471 * exec-tool.in: Allow for dsymutil.
20473 2021-07-05 Uroš Bizjak <ubizjak@gmail.com>
20475 * config/i386/i386-expand.c (ix86_split_mmx_punpck):
20476 Handle V4QI and V2HI modes.
20477 (expand_vec_perm_blend): Allow 4-byte vector modes with TARGET_SSE4_1.
20478 Handle V4QI mode. Emit mmx_pblendvb32 for 4-byte modes.
20479 (expand_vec_perm_pshufb): Rewrite to use switch statemets.
20480 Handle 4-byte dual operands with TARGET_XOP and single operands
20481 with TARGET_SSSE3. Emit mmx_ppermv32 for TARGET_XOP and
20482 mmx_pshufbv4qi3 for TARGET_SSSE3.
20483 (expand_vec_perm_pblendv): Allow 4-byte vector modes with TARGET_SSE4_1.
20484 (expand_vec_perm_interleave2): Allow 4-byte vector modes.
20485 (expand_vec_perm_pshufb2): Allow 4-byte vector modes with TARGET_SSSE3.
20486 (expand_vec_perm_even_odd_1): Handle V4QI mode.
20487 (expand_vec_perm_broadcast_1): Handle V4QI mode.
20488 (ix86_vectorize_vec_perm_const): Handle V4QI mode.
20489 * config/i386/mmx.md (mmx_ppermv32): New insn pattern.
20490 (mmx_pshufbv4qi3): Ditto.
20491 (*mmx_pblendw32): Ditto.
20492 (*mmx_pblendw64): Rename from *mmx_pblendw.
20493 (mmx_punpckhbw_low): New insn_and_split pattern.
20494 (mmx_punpcklbw_low): Ditto.
20496 2021-07-05 Richard Biener <rguenther@suse.de>
20498 * tree-vect-loop-manip.c (vect_loop_versioning): Do not
20499 set LOOP_C_INFINITE on the vectorized loop.
20501 2021-07-05 Richard Biener <rguenther@suse.de>
20503 PR middle-end/101291
20504 * cfgloopmanip.c (loop_version): Set the loop copy of the
20505 versioned loop to the new loop.
20507 2021-07-04 Iain Sandoe <iain@sandoe.co.uk>
20510 * config.gcc: Ensure that Darwin biarch definitions are
20511 added before i386.h.
20512 * config/i386/darwin.h (TARGET_64BIT): Remove.
20513 (PR80556_WORKAROUND): New.
20514 (REAL_LIBGCC_SPEC): Amend to use PR80556_WORKAROUND.
20515 (DARWIN_SUBARCH_SPEC): New.
20516 * config/i386/darwin32-biarch.h (TARGET_64BIT_DEFAULT,
20517 TARGET_BI_ARCH, PR80556_WORKAROUND): New.
20518 (REAL_LIBGCC_SPEC): Remove.
20519 * config/i386/darwin64-biarch.h (TARGET_64BIT_DEFAULT,
20520 TARGET_BI_ARCH, PR80556_WORKAROUND): New.
20521 (REAL_LIBGCC_SPEC): Remove.
20523 2021-07-03 H.J. Lu <hjl.tools@gmail.com>
20525 PR middle-end/101294
20526 * expr.c (store_constructor): Don't use vec_duplicate on vector.
20528 2021-07-02 Martin Sebor <msebor@redhat.com>
20530 PR middle-end/98871
20531 PR middle-end/98512
20532 * diagnostic.c (get_any_inlining_info): New.
20533 (update_effective_level_from_pragmas): Handle inlining context.
20534 (diagnostic_enabled): Same.
20535 (diagnostic_report_diagnostic): Same.
20536 * diagnostic.h (struct diagnostic_info): Add ctor.
20537 (struct diagnostic_context): Add new member.
20538 * tree-diagnostic.c (set_inlining_locations): New.
20539 (tree_diagnostics_defaults): Set new callback pointer.
20541 2021-07-02 Peter Bergner <bergner@linux.ibm.com>
20543 * config/rs6000/rs6000-builtin.def (BU_MMA_PAIR_LD, BU_MMA_PAIR_ST):
20545 (__builtin_vsx_lxvp, __builtin_vsx_stxvp): New built-ins.
20546 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_mma_builtin): Expand
20547 lxvp and stxvp built-ins.
20548 (mma_init_builtins): Handle lxvp and stxvp built-ins.
20549 (builtin_function_type): Likewise.
20550 * doc/extend.texi (__builtin_vsx_lxvp, __builtin_mma_stxvp): Document.
20552 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
20554 * config/h8300/h8300-protos.h (compute_a_shift_cc): Accept
20555 additional argument for the code.
20556 * config/h8300/h8300.c (compute_a_shift_cc): Accept additional
20557 argument for the code. Just return if the ZN bits are useful or
20558 not rather than the old style CC_* enums.
20559 * config/h8300/shiftrotate.md (shiftqi_noscratch): Move before
20560 more generic shiftqi patterns.
20561 (shifthi_noscratch, shiftsi_noscratch): Similarly.
20562 (shiftqi_noscratch_set_flags): New pattern.
20563 (shifthi_noscratch_set_flags, shiftsi_noscratch_set_flags): Likewise.
20565 2021-07-02 Andrew MacLeod <amacleod@redhat.com>
20567 PR tree-optimization/101223
20568 * range-op.cc (build_lt): Add -1 for signed values.
20569 (built_gt): Subtract -1 for signed values.
20571 2021-07-02 David Faust <david.faust@oracle.com>
20573 * btfout.c (get_btf_kind): Support BTF_KIND_FLOAT.
20574 (btf_asm_type): Likewise.
20576 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
20578 * config/h8300/h8300-protos.h (output_a_shift): Make first argument
20579 an array of rtx rather than a pointer to rtx. Add code argument.
20580 (compute_a_shift_length): Similarly.
20581 * config/h8300/h8300.c (h8300_shift_costs): Adjust now that the
20582 shift itself isn't an operand. Create dummy operand[0] to carry
20583 a mode and pass a suitable rtx code to compute_a_shift_length.
20584 (get_shift_alg): Adjust operand number of clobber in output templates.
20585 (output_a_shift): Make first argument an array of rtx rather than
20586 a pointer to rtx. Add code argument for the type of shift.
20587 Adjust now that the shift itself is no longer an operand.
20588 (compute_a_shift_length): Similarly.
20589 * config/h8300/shiftrotate.md (shiftqi, shifthi, shiftsi): Use an
20590 iterator rather than nshift_operator.
20591 (shiftqi_noscratch, shifthi_noscratch, shiftsi_noscratch): Likewise.
20592 (shiftqi_clobber_flags): Adjust to API changes in output_a_shift
20593 and compute_a_shift_length.
20594 (shiftqi_noscratch_clobber_flags): Likewise.
20595 (shifthi_noscratch_clobber_flags): Likewise.
20596 (shiftsi_noscratch_clobber_flags): Likewise.
20598 2021-07-02 Iain Sandoe <iain@sandoe.co.uk>
20601 * config/darwin.h (DSYMUTIL_SPEC): Do not try to run
20602 dsymutil for BTF/CTF.
20604 2021-07-02 Iain Sandoe <iain@sandoe.co.uk>
20607 * config/darwin.h (CTF_INFO_SECTION_NAME): Update the
20608 segment to include BTF.
20609 (BTF_INFO_SECTION_NAME): New.
20611 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
20613 * config/m32r/m32r-protos.h (call_operand): Adjust return type.
20614 (small_data_operand, memreg_operand, small_insn_p): Likewise.
20615 * config/m32r/m32r.c (call_operand): Adjust return type.
20616 (small_data_operand, memreg_operand): Likewise.
20618 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
20620 * config/frv/frv-protos.h (integer_register_operand): Adjust return
20622 (frv_load_operand, gpr_or_fpr_operand, gpr_no_subreg_operand): Likewise.
20623 (fpr_or_int6_operand, gpr_or_int_operand); Likewise.
20624 (gpr_or_int12_operand, gpr_or_int10_operand); Likewise.
20625 (move_source_operand, move_destination_operand): Likewise.
20626 (condexec_source_operand, condexec_dest_operand): Likewise.
20627 (lr_operand, gpr_or_memory_operand, fpr_or_memory_operand): Likewise.
20628 (reg_or_0_operand, fcc_operand, icc_operand, cc_operand): Likewise.
20629 (fcr_operand, icr_operand, cr_operand, call_operand): Likewise.
20630 (fpr_operand, even_reg_operand, odd_reg_operand): Likewise.
20631 (even_gpr_operand, odd_gpr_operand, quad_fpr_operand): Likewise.
20632 (even_fpr_operand, odd_fpr_operand): Likewise.
20633 (dbl_memory_one_insn_operand, dbl_memory_two_insn_operand): Likewise.
20634 (int12_operand, int6_operand, int5_operand, uint5_operand): Likewise.
20635 (uint4_operand, uint1_operand, int_2word_operand): Likewise
20636 (upper_int16_operand, uint16_operand, symbolic_operand): Likewise.
20637 (relational_operator, float_relational_operator): Likewise.
20638 (ccr_eqne_operator, minmax_operator): Likewise.
20639 (condexec_si_binary_operator, condexec_si_media_operator): Likewise.
20640 (condexec_si_divide_operator, condexec_si_unary_operator): Likewise.
20641 (condexec_sf_conv_operator, condexec_sf_add_operator): Likewise.
20642 (intop_compare_operator, acc_operand, even_acc_operand): Likewise.
20643 (quad_acc_operand, accg_operand): Likewise.
20645 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
20647 * config/stormy16/stormy16-protos.h (xstormy16_below_100_symbol): Change
20648 return type to a bool.
20649 (nonimmediate_nonstack_operand): Likewise.
20650 (xstormy16_splittable_below100_operand): Likewise.
20651 * config/stormy16/stormy16.c (xstormy16_below_100_symbol): Fix
20653 (xstormy16_splittable_below100_operand): Likewise.
20655 2021-07-02 Richard Biener <rguenther@suse.de>
20657 PR tree-optimization/101293
20658 * tree-ssa-loop-im.c (mem_ref_hasher::equal): Compare MEM_REF bases
20659 with combined offsets.
20660 (gather_mem_refs_stmt): Hash MEM_REFs as if their offset were
20661 combined with the rest of the offset.
20663 2021-07-02 Eric Botcazou <ebotcazou@adacore.com>
20665 * config/i386/i386.c (asm_preferred_eh_data_format): Always use the
20666 PIC encodings for PE-COFF targets.
20668 2021-07-02 Jakub Jelinek <jakub@redhat.com>
20671 * config/i386/i386-expand.c (ix86_broadcast_from_integer_constant):
20672 Return nullptr for TImode inner mode.
20674 2021-07-02 Richard Biener <rguenther@suse.de>
20676 PR tree-optimization/101280
20677 PR tree-optimization/101173
20678 * gimple-loop-interchange.cc
20679 (tree_loop_interchange::valid_data_dependences): Properly
20680 guard all dependence checks with DDR_REVERSED_P or its
20683 2021-07-02 Hongyu Wang <hongyu.wang@intel.com>
20685 * config/i386/i386-expand.c (ix86_expand_builtin):
20686 Add branch to clear odata when ZF is set for asedecenc_expand
20687 and wideaesdecenc_expand.
20689 2021-07-02 Eugene Rozenfeld <erozen@microsoft.com>
20691 * config/i386/gcc-auto-profile: regenerate
20693 2021-07-02 liuhongt <hongtao.liu@intel.com>
20695 * config/i386/sse.md (trunc<mode><pmov_dst_4>2): Refined to ..
20696 (trunc<mode><pmov_dst_4_lower>2): this.
20698 2021-07-01 David Malcolm <dmalcolm@redhat.com>
20700 * diagnostic.h (diagnostic_context::m_file_cache): New field.
20701 * input.c (class fcache): Rename to...
20702 (class file_cache_slot): ...this, making most members private and
20703 prefixing fields with "m_".
20704 (file_cache_slot::get_file_path): New accessor.
20705 (file_cache_slot::get_use_count): New accessor.
20706 (file_cache_slot::missing_trailing_newline_p): New accessor.
20707 (file_cache_slot::inc_use_count): New.
20708 (fcache_buffer_size): Move to...
20709 (file_cache_slot::buffer_size): ...here.
20710 (fcache_line_record_size): Move to...
20711 (file_cache_slot::line_record_size): ...here.
20712 (fcache_tab): Delete, in favor of global_dc->m_file_cache.
20713 (fcache_tab_size): Move to file_cache::num_file_slots.
20714 (diagnostic_file_cache_init): Update for move of fcache_tab
20715 to global_dc->m_file_cache.
20716 (diagnostic_file_cache_fini): Likewise.
20717 (lookup_file_in_cache_tab): Convert to...
20718 (file_cache::lookup_file): ...this.
20719 (diagnostics_file_cache_forcibly_evict_file): Update for move of
20720 fcache_tab to global_dc->m_file_cache, moving most of
20721 implementation to...
20722 (file_cache::forcibly_evict_file): ...this new function and...
20723 (file_cache_slot::evict): ...this new function.
20724 (evicted_cache_tab_entry): Convert to...
20725 (file_cache::evicted_cache_tab_entry): ...this.
20726 (add_file_to_cache_tab): Convert to...
20727 (file_cache::add_file): ...this, moving bulk of implementation
20729 (file_cache_slot::create): ..this new function.
20730 (file_cache::file_cache): New.
20731 (file_cache::~file_cache): New.
20732 (lookup_or_add_file_to_cache_tab): Convert to...
20733 (file_cache::lookup_or_add_file): ..this new function.
20734 (fcache::fcache): Rename to...
20735 (file_cache_slot::file_cache_slot): ...this, adding "m_" prefixes
20737 (fcache::~fcache): Rename to...
20738 (file_cache_slot::~file_cache_slot): ...this, adding "m_" prefixes
20740 (needs_read): Convert to...
20741 (file_cache_slot::needs_read_p): ...this.
20742 (needs_grow): Convert to...
20743 (file_cache_slot::needs_grow_p): ...this.
20744 (maybe_grow): Convert to...
20745 (file_cache_slot::maybe_grow): ...this.
20746 (read_data): Convert to...
20747 (file_cache_slot::read_data): ...this.
20748 (maybe_read_data): Convert to...
20749 (file_cache_slot::maybe_read_data): ...this.
20750 (get_next_line): Convert to...
20751 (file_cache_slot::get_next_line): ...this.
20752 (goto_next_line): Convert to...
20753 (file_cache_slot::goto_next_line): ...this.
20754 (read_line_num): Convert to...
20755 (file_cache_slot::read_line_num): ...this.
20756 (location_get_source_line): Update for moving of globals to
20757 global_dc->m_file_cache.
20758 (location_missing_trailing_newline): Likewise.
20759 * input.h (class file_cache_slot): New forward decl.
20760 (class file_cache): New.
20762 2021-07-01 Michael Meissner <meissner@linux.ibm.com>
20764 * config/rs6000/rs6000.c (rs6000_maybe_emit_fp_cmove): Add IEEE
20765 128-bit floating point conditional move support.
20766 (have_compare_and_set_mask): Add IEEE 128-bit floating point
20768 * config/rs6000/rs6000.md (mov<mode>cc, IEEE128 iterator): New insn.
20769 (mov<mode>cc_p10, IEEE128 iterator): New insn.
20770 (mov<mode>cc_invert_p10, IEEE128 iterator): New insn.
20771 (fpmask<mode>, IEEE128 iterator): New insn.
20772 (xxsel<mode>, IEEE128 iterator): New insn.
20774 2021-07-01 Iain Sandoe <iain@sandoe.co.uk>
20777 * config/darwin.h (CTF_INFO_SECTION_NAME): New.
20779 2021-07-01 H.J. Lu <hjl.tools@gmail.com>
20781 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
20783 * config/i386/i386-protos.h (ix86_expand_vector_init_duplicate):
20785 * config/i386/sse.md (INT_BROADCAST_MODE): New mode iterator.
20786 (vec_duplicate<mode>): New expander.
20788 2021-07-01 H.J. Lu <hjl.tools@gmail.com>
20791 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
20793 (ix86_byte_broadcast): New function.
20794 (ix86_convert_const_wide_int_to_broadcast): Likewise.
20795 (ix86_expand_move): Convert CONST_WIDE_INT to broadcast if mode
20796 size is 16 bytes or bigger.
20797 (ix86_broadcast_from_integer_constant): New function.
20798 (ix86_expand_vector_move): Convert CONST_WIDE_INT and CONST_VECTOR
20799 to broadcast if mode size is 16 bytes or bigger.
20800 * config/i386/i386-protos.h (ix86_gen_scratch_sse_rtx): New
20802 * config/i386/i386.c (ix86_gen_scratch_sse_rtx): New function.
20804 2021-07-01 Uroš Bizjak <ubizjak@gmail.com>
20806 * config/i386/predicates.md (ix86_endbr_immediate_operand):
20807 Return true/false instead of 1/0.
20808 (movq_parallel): Ditto.
20810 2021-07-01 Uroš Bizjak <ubizjak@gmail.com>
20812 * recog.c (general_operand): Return true/false instead of 1/0.
20813 (register_operand): Ditto.
20814 (immediate_operand): Ditto.
20815 (const_int_operand): Ditto.
20816 (const_scalar_int_operand): Ditto.
20817 (const_double_operand): Ditto.
20818 (push_operand): Ditto.
20819 (pop_operand): Ditto.
20820 (memory_operand): Ditto.
20821 (indirect_operand): Ditto.
20823 2021-07-01 Uroš Bizjak <ubizjak@gmail.com>
20825 * genpreds.c (write_predicate_subfunction):
20826 Change the type of written subfunction to bool.
20827 (write_one_predicate_function):
20828 Change the type of written function to bool.
20829 (write_tm_preds_h): Ditto.
20830 * recog.h (*insn_operand_predicate_fn): Change the type to bool.
20831 * recog.c (general_operand): Change the type to bool.
20832 (address_operand): Ditto.
20833 (register_operand): Ditto.
20834 (pmode_register_operand): Ditto.
20835 (scratch_operand): Ditto.
20836 (immediate_operand): Ditto.
20837 (const_int_operand): Ditto.
20838 (const_scalar_int_operand): Ditto.
20839 (const_double_operand): Ditto.
20840 (nonimmediate_operand): Ditto.
20841 (nonmemory_operand): Ditto.
20842 (push_operand): Ditto.
20843 (pop_operand): Ditto.
20844 (memory_operand): Ditto.
20845 (indirect_operand): Ditto.
20846 (ordered_comparison_operator): Ditto.
20847 (comparison_operator): Ditto.
20848 * config/i386/i386-expand.c (ix86_expand_sse_cmp):
20849 Change the type of indirect predicate function to bool.
20850 * config/rs6000/rs6000.c (easy_vector_constant):
20851 Change the type to bool.
20852 * config/mips/mips-protos.h (m16_based_address_p):
20853 Change the type of operand 3 to bool.
20855 2021-07-01 Richard Biener <rguenther@suse.de>
20857 PR tree-optimization/101280
20858 PR tree-optimization/101173
20859 * gimple-loop-interchange.cc
20860 (tree_loop_interchange::valid_data_dependences): Revert
20861 previous change and instead correctly handle DDR_REVERSED_P
20864 2021-07-01 Richard Biener <rguenther@suse.de>
20866 PR tree-optimization/101278
20867 * tree-ssa-dse.c (dse_classify_store): First check for
20868 uses, then ignore stmt for chaining purposes.
20870 2021-07-01 Richard Biener <rguenther@suse.de>
20872 PR tree-optimization/100778
20873 * tree-vect-slp.c (vect_schedule_slp_node): Do not place trapping
20874 vectorized ops ahead of their scalar BB.
20876 2021-07-01 Uroš Bizjak <ubizjak@gmail.com>
20879 * config/i386/i386.md (*nabs<dwi>2_doubleword):
20880 New insn_and_split pattern.
20881 (*nabs<dwi>2_1): Ditto.
20882 * config/i386/i386-features.c
20883 (general_scalar_chain::compute_convert_gain):
20884 Handle (NEG (ABS (...))) RTX. Rewrite src code
20885 scanner as switch statement.
20886 (general_scalar_chain::convert_insn):
20887 Handle (NEG (ABS (...))) RTX.
20888 (general_scalar_to_vector_candidate_p):
20889 Detect (NEG (ABS (...))) RTX. Reorder case statements
20890 for (AND (NOT (...) ...)) fallthrough.
20892 2021-07-01 Richard Biener <rguenther@suse.de>
20894 PR tree-optimization/101178
20895 * tree-vect-slp.c (slpg_vertex::materialize): Remove.
20896 (slpg::perm_in): Add.
20897 (slpg::get_perm_in): Remove.
20898 (slpg::get_perm_materialized): Add.
20899 (vect_optimize_slp): Handle VEC_PERM nodes more optimally
20900 during permute propagation and materialization.
20902 2021-07-01 Jakub Jelinek <jakub@redhat.com>
20905 * dwarf2out.c (loc_list_from_tree_1): Handle COMPOUND_LITERAL_EXPR.
20907 2021-07-01 Jakub Jelinek <jakub@redhat.com>
20909 PR middle-end/94366
20910 * omp-low.c (lower_rec_input_clauses): Rename is_fp_and_or to
20911 is_truth_op, set it for TRUTH_*IF_EXPR regardless of new_var's type,
20912 use boolean_type_node instead of integer_type_node as NE_EXPR type.
20913 (lower_reduction_clauses): Likewise.
20915 2021-06-30 Hafiz Abid Qadeer <abidh@codesourcery.com>
20917 * config/gcn/gcn.c: Include dwarf2.h.
20918 (gcn_addr_space_debug): New function.
20919 (TARGET_ADDR_SPACE_DEBUG): New hook.
20921 2021-06-30 Hafiz Abid Qadeer <abidh@codesourcery.com>
20923 * common/config/gcn/gcn-common.c
20924 (gcn_option_optimization_table): Change OPT_fomit_frame_pointer to -O3.
20925 * config/gcn/gcn.c (gcn_expand_prologue): Prefer the frame pointer
20927 (gcn_expand_prologue): Prefer the frame pointer when emitting CFI.
20928 (gcn_frame_pointer_rqd): New function.
20929 (TARGET_FRAME_POINTER_REQUIRED): New hook.
20931 2021-06-30 Hafiz Abid Qadeer <abidh@codesourcery.com>
20933 * config/gcn/gcn.c (move_callee_saved_registers): Emit CFI notes for
20934 prologue register saves.
20935 (gcn_debug_unwind_info): Use UI_DWARF2.
20936 (gcn_dwarf_register_number): Map DWARF_LINK_REGISTER to DWARF PC.
20937 (gcn_dwarf_register_span): DWARF_LINK_REGISTER doesn't span.
20938 * config/gcn/gcn.h: (DWARF_FRAME_RETURN_COLUMN): New define.
20939 (DWARF_LINK_REGISTER): New define.
20940 (FIRST_PSEUDO_REGISTER): Increment.
20941 (FIXED_REGISTERS): Add entry for DWARF_LINK_REGISTER.
20942 (CALL_USED_REGISTERS): Likewise.
20943 (REGISTER_NAMES): Likewise.
20945 2021-06-30 Richard Biener <rguenther@suse.de>
20947 PR tree-optimization/101267
20948 * tree-vect-stmts.c (vect_check_scalar_mask): Adjust
20949 API and use SLP compatible interface of vect_is_simple_use.
20950 Reject not vectorized SLP defs for callers that do not support
20952 (vect_check_store_rhs): Handle masked stores and pass down
20953 the appropriate operator index.
20954 (vectorizable_call): Adjust.
20955 (vectorizable_store): Likewise.
20956 (vectorizable_load): Likewise. Handle SLP pecularity of
20958 (vect_is_simple_use): Remove special-casing of masked stores.
20960 2021-06-30 Tobias Burnus <tobias@codesourcery.com>
20962 * common.opt (foffload): Remove help as Driver only.
20963 * gcc.c (display_help): Add -foffload.
20965 2021-06-30 Tobias Burnus <tobias@codesourcery.com>
20967 * gcc.c (close_at_file, execute): Replace alloca by XALLOCAVEC.
20968 (check_offload_target_name): Fix splitting OFFLOAD_TARGETS into
20969 a candidate list; better inform no offload target is configured
20970 and fix hint extraction when passed target is not '\0' at [len].
20971 * common.opt (foffload): Add tailing '.'.
20972 (foffload-options): Likewise; fix flag name in the help string.
20974 2021-06-30 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
20977 * config/arm/arm_neon.h: Move vabs intrinsics before vcage_f32.
20978 (vcage_f32): Gate comparison on __FAST_MATH__.
20979 (vcageq_f32): Likewise.
20980 (vcale_f32): Likewise.
20981 (vcaleq_f32): Likewise.
20982 (vcagt_f32): Likewise.
20983 (vcagtq_f32): Likewise.
20984 (vcalt_f32): Likewise.
20985 (vcaltq_f32): Likewise.
20986 (vcage_f16): Likewise.
20987 (vcageq_f16): Likewise.
20988 (vcale_f16): Likewise.
20989 (vcaleq_f16): Likewise.
20990 (vcagt_f16): Likewise.
20991 (vcagtq_f16): Likewise.
20992 (vcalt_f16): Likewise.
20993 (vcaltq_f16): Likewise.
20995 2021-06-30 Richard Biener <rguenther@suse.de>
20997 PR tree-optimization/101264
20998 * tree-vect-slp.c (vect_optimize_slp): Propagate the
20999 computed perm_in to all "any" permute successors
21000 we cannot de-duplicate immediately.
21002 2021-06-30 liuhongt <hongtao.liu@intel.com>
21005 * config/i386/sse.md
21006 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>):
21008 (avx512f_sfixupimm<mode><maskz_scalar_name><round_saeonly_name>):
21010 (avx512f_sfixupimm<mode>_mask<round_saeonly_name>"): Refined.
21011 * config/i386/subst.md (maskz_scalar): New define_subst.
21012 (maskz_scalar_name): New subst_attr.
21013 (maskz_scalar_op5): Ditto.
21014 (round_saeonly_maskz_scalar_op5): Ditto.
21015 (round_saeonly_maskz_scalar_operand5): Ditto.
21017 2021-06-30 David Edelsohn <dje.gcc@gmail.com>
21019 * config/rs6000/rs6000.c (rs6000_xcoff_section_type_flags):
21020 Increase code CSECT alignment to at least 32 bytes.
21021 * config/rs6000/xcoff.h (TEXT_SECTION_ASM_OP): Add 32 byte
21022 alignment designation.
21024 2021-06-29 Sergei Trofimovich <siarheit@google.com>
21026 * doc/generic.texi: Fix s/net yet/not yet/ typo.
21028 2021-06-29 Andrew MacLeod <amacleod@redhat.com>
21030 PR tree-optimization/101254
21031 * range-op.cc (operator_minus::op1_op2_relation_effect): Check for
21032 wrapping/non-wrapping when setting the result range.
21034 2021-06-29 Andrew MacLeod <amacleod@redhat.com>
21036 * value-query.cc (gimple_range_global): Allow phis.
21038 2021-06-29 Andrew MacLeod <amacleod@redhat.com>
21040 * vr-values.c (vr_values::vrp_stmt_computes_nonzero): Use stmt.
21041 (simplify_using_ranges::op_with_boolean_value_range_p): Add a
21042 statement for location context.
21043 (check_for_binary_op_overflow): Ditto.
21044 (simplify_using_ranges::get_vr_for_comparison): Ditto.
21045 (simplify_using_ranges::compare_name_with_value): Ditto.
21046 (simplify_using_ranges::compare_names): Ditto.
21047 (vrp_evaluate_conditional_warnv_with_ops_using_ranges): Ditto.
21048 (simplify_using_ranges::simplify_truth_ops_using_ranges): Ditto.
21049 (simplify_using_ranges::simplify_min_or_max_using_ranges): Ditto.
21050 (simplify_using_ranges::simplify_internal_call_using_ranges): Ditto.
21051 (simplify_using_ranges::two_valued_val_range_p): Ditto.
21052 (simplify_using_ranges::simplify): Ditto.
21053 * vr-values.h: Adjust prototypes.
21055 2021-06-29 Uroš Bizjak <ubizjak@gmail.com>
21058 * config/i386/mmx.md (vec_addsubv2sf3): New insn pattern.
21060 2021-06-29 Julian Brown <julian@codesourcery.com>
21062 * config/gcn/gcn.c (gcn_init_libfuncs): New function.
21063 (TARGET_INIT_LIBFUNCS): Define target hook using above function.
21064 * config/gcn/gcn.h (UNITS_PER_WORD): Define to 8 for IN_LIBGCC2, 4
21066 (LIBGCC2_UNITS_PER_WORD, BITS_PER_WORD): Remove definitions.
21067 (MAX_FIXED_MODE_SIZE): Change to 128.
21069 2021-06-29 Julian Brown <julian@codesourcery.com>
21071 * config/gcn/gcn.md (UNSPEC_FLBIT_INT): New unspec constant.
21072 (s_mnemonic): Add clrsb.
21073 (gcn_flbit<mode>_int): Add insn pattern for SImode/DImode.
21074 (clrsb<mode>2): Add expander for SImode/DImode.
21076 2021-06-29 Julian Brown <julian@codesourcery.com>
21078 * config/gcn/gcn.md (<su>mulsidi3, <su>mulsidi3_reg, <su>mulsidi3_imm,
21079 muldi3): Add patterns.
21081 2021-06-29 Julian Brown <julian@codesourcery.com>
21083 * config/gcn/gcn.md (<su>mulsi3_highpart): Change to expander.
21084 (<su>mulsi3_highpart_reg, <su>mulsi3_highpart_imm): New patterns.
21086 2021-06-29 Julian Brown <julian@codesourcery.com>
21088 * config/gcn/gcn.md (mulsi3): Make s_mulk_i32 variant clobber SCC.
21090 2021-06-29 Joseph Myers <joseph@codesourcery.com>
21092 * btfout.c, ctfout.c: Include "memmodel.h".
21094 2021-06-29 Tobias Burnus <tobias@codesourcery.com>
21096 * gcc.c (check_offload_target_name): Cast len argument to
21097 %q.*s to 'int'; avoid -Wstringop-truncation warning.
21099 2021-06-29 Richard Biener <rguenther@suse.de>
21101 * tree-vect-slp.c (vect_optimize_slp): Forward propagate
21102 to "any" permute nodes and relax "any" permute proapgation
21103 during iterative backward propagation.
21105 2021-06-29 Tobias Burnus <tobias@codesourcery.com>
21108 * common.opt (-foffload=): Update description.
21109 (-foffload-options=): New.
21110 * doc/invoke.texi (C Language Options): Document
21111 -foffload and -foffload-options.
21112 * gcc.c (check_offload_target_name): New, split off from
21113 handle_foffload_option.
21114 (check_foffload_target_names): New.
21115 (handle_foffload_option): Handle -foffload=default.
21116 (driver_handle_option): Update for -foffload-options.
21117 * lto-opts.c (lto_write_options): Use -foffload-options
21118 instead of -foffload.
21119 * lto-wrapper.c (merge_and_complain, append_offload_options):
21121 * opts.c (common_handle_option): Likewise.
21123 2021-06-29 Tobias Burnus <tobias@codesourcery.com>
21125 * doc/invoke.texi (C Language Options): Sort options
21126 alphabetically in optlist and also the description itself.
21127 Remove leftover -fallow-single-precision from and add missing
21128 -fgnu-tm to the optlist.
21130 2021-06-29 Richard Biener <rguenther@suse.de>
21132 * tree-vect-slp.c (slpg_vertex::visited): Remove.
21133 (vect_slp_perms_eq): Handle -1 permutes.
21134 (vect_optimize_slp): Rewrite permute propagation.
21136 2021-06-29 Jakub Jelinek <jakub@redhat.com>
21139 * match.pd ((intptr_t)x eq/ne CST to x eq/ne (typeof x) CST): Don't
21140 perform the optimization in GENERIC when sanitizing and x has a
21143 2021-06-29 Richard Biener <rguenther@suse.de>
21145 PR tree-optimization/101242
21146 * tree-vect-slp.c (vect_slp_build_vertices): Force-add
21147 PHIs with not represented initial values as leafs.
21149 2021-06-29 Jan-Benedict Glaw <jbglaw@getslash.de>
21151 * config/pdp11/pdp11.h (ASM_OUTPUT_SKIP): Fix signedness warning.
21152 * config/pdp11/pdp11.c (pdp11_asm_print_operand_punct_valid_p): Remove
21153 "register" keyword.
21154 (pdp11_initial_elimination_offset) Remove unused variable.
21155 (pdp11_cmp_length) Ditto.
21156 (pdp11_insn_cost): Ditto, and fix signedness warning.
21158 2021-06-29 David Edelsohn <dje.gcc@gmail.com>
21160 * btfout.c: Include tm_p.h.
21163 2021-06-28 Indu Bhagat <indu.bhagat@oracle.com>
21165 * config/bpf/bpf.c (bpf_expand_prologue): Do not mark insns as
21167 (bpf_expand_epilogue): Likewise.
21168 * config/bpf/bpf.h (DWARF2_FRAME_INFO): Define to 0.
21169 Do not define DBX_DEBUGGING_INFO.
21171 2021-06-28 Indu Bhagat <indu.bhagat@oracle.com>
21173 * doc/invoke.texi: Document the CTF and BTF debug info options.
21175 2021-06-28 Indu Bhagat <indu.bhagat@oracle.com>
21176 David Faust <david.faust@oracle.com>
21177 Jose E. Marchesi <jose.marchesi@oracle.com>
21178 Weimin Pan <weimin.pan@oracle.com>
21180 * Makefile.in: Add ctfc.*, ctfout.c and btfout.c files to
21181 GTFILES. Add new object files.
21182 * common.opt: Add CTF and BTF debug info options.
21183 * btfout.c: New file.
21184 * ctfc.c: Likewise.
21185 * ctfc.h: Likewise.
21186 * ctfout.c: Likewise.
21187 * dwarf2ctf.c: Likewise.
21188 * dwarf2ctf.h: Likewise.
21189 * dwarf2cfi.c (dwarf2out_do_frame): Acknowledge CTF_DEBUG and
21191 * dwarf2out.c (dwarf2out_source_line): Likewise.
21192 (dwarf2out_finish): Skip emitting DWARF if CTF or BTF are to
21194 (debug_format_do_cu): New function.
21195 (dwarf2out_early_finish): Traverse DIEs and emit CTF/BTF for
21197 Include dwarf2ctf.c.
21198 * final.c (dwarf2_debug_info_emitted_p): Acknowledge DWARF-based debug
21200 * flag-types.h (enum debug_info_type): Add CTF_DEBUG and BTF_DEBUG.
21201 (CTF_DEBUG): New bitmask.
21202 (BTF_DEBUG): Likewise.
21203 (enum ctf_debug_info_levels): New enum.
21204 * gengtype.c (open_base_files): Handle ctfc.h.
21205 (main): Handle uint32_t type.
21206 * flags.h (btf_debuginfo_p): New definition.
21207 (dwarf_based_debuginfo_p): Likewise.
21208 * opts.c (debug_type_names): Add entries for CTF and BTF.
21209 (btf_debuginfo_p): New function.
21210 (dwarf_based_debuginfo_p): Likewise.
21211 (common_handle_option): Handle -gctfN and -gbtf options.
21212 (set_debug_level): Set CTF_DEBUG, BTF_DEBUG whenever appropriate.
21213 * toplev.c (process_options): Inform the user and ignore -gctfLEVEL if
21216 2021-06-28 Jose E. Marchesi <jose.marchesi@oracle.com>
21218 * dwarf2out.c (AT_class): Function is no longer static.
21219 (AT_int): Likewise.
21220 (AT_unsigned): Likewise.
21221 (AT_loc): Likewise.
21222 (get_AT): Likewise.
21223 (get_AT_string): Likewise.
21224 (get_AT_flag): Likewise.
21225 (get_AT_unsigned): Likewise.
21226 (get_AT_ref): Likewise.
21227 (new_die_raw): Likewise.
21228 (lookup_decl_die): Likewise.
21229 (base_type_die): Likewise.
21230 (add_name_attribute): Likewise.
21231 (add_AT_int): Likewise.
21232 (add_AT_unsigned): Likewise.
21233 (add_AT_loc): Likewise.
21234 (dw_get_die_tag): New function.
21235 (dw_get_die_child): Likewise.
21236 (dw_get_die_sib): Likewise.
21237 (struct dwarf_file_data): Move from here to dwarf2out.h
21238 (struct dw_attr_struct): Likewise.
21239 * dwarf2out.h: Analogous changes.
21241 2021-06-28 Martin Jambor <mjambor@suse.cz>
21244 * ipa-param-manipulation.h (class ipa_param_body_adjustments): New
21245 members m_dead_stmts and m_dead_ssas.
21246 * ipa-param-manipulation.c
21247 (ipa_param_body_adjustments::mark_dead_statements): New function.
21248 (ipa_param_body_adjustments::common_initialization): Call it on
21249 all removed but not split parameters.
21250 (ipa_param_body_adjustments::ipa_param_body_adjustments): Initialize
21252 (ipa_param_body_adjustments::modify_call_stmt): Remove arguments that
21254 * tree-inline.c (remap_gimple_stmt): Do not copy dead statements, reset
21255 dead debug statements.
21256 (copy_phis_for_bb): Do not copy dead PHI nodes.
21258 2021-06-28 Martin Jambor <mjambor@suse.cz>
21261 * symtab-clones.h (clone_info): Removed member param_adjustments.
21262 * ipa-param-manipulation.h: Adjust initial comment to reflect how we
21263 deal with pass-through splits now.
21264 (ipa_param_performed_split): Removed.
21265 (ipa_param_adjustments::modify_call): Adjusted parameters.
21266 (class ipa_param_body_adjustments): Adjusted parameters of
21267 register_replacement, modify_gimple_stmt and modify_call_stmt.
21268 (ipa_verify_edge_has_no_modifications): Declare.
21269 (ipa_edge_modifications_finalize): Declare.
21270 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Remove
21271 performed_splits processing, pas only edge to padjs->modify_call,
21272 check that call arguments were not modified if they should not have
21274 * cgraphclones.c (cgraph_node::create_clone): Do not copy performed
21276 * ipa-param-manipulation.c (struct pass_through_split_map): New type.
21277 (ipa_edge_modification_info): Likewise.
21278 (ipa_edge_modification_sum): Likewise.
21279 (ipa_edge_modifications): New edge summary.
21280 (ipa_verify_edge_has_no_modifications): New function.
21281 (transitive_split_p): Removed.
21282 (transitive_split_map): Likewise.
21283 (init_transitive_splits): Likewise.
21284 (ipa_param_adjustments::modify_call): Adjusted to use the new edge
21285 summary instead of performed_splits.
21286 (ipa_param_body_adjustments::register_replacement): Drop dummy
21287 parameter, set base_index of the created ipa_param_body_replacement.
21288 (phi_arg_will_live_p): New function.
21289 (ipa_param_body_adjustments::common_initialization): Do not create
21290 IPA_SRA dummy decls.
21291 (simple_tree_swap_info): Removed.
21292 (remap_split_decl_to_dummy): Likewise.
21293 (record_argument_state_1): New function.
21294 (record_argument_state): Likewise.
21295 (ipa_param_body_adjustments::modify_call_stmt): New parameter
21296 orig_stmt. Do not work with dummy decls, save necessary info about
21297 changes to ipa_edge_modifications.
21298 (ipa_param_body_adjustments::modify_gimple_stmt): New parameter
21299 orig_stmt, pass it to modify_call_stmt.
21300 (ipa_param_body_adjustments::modify_cfun_body): Adjust call to
21301 modify_gimple_stmt.
21302 (ipa_edge_modifications_finalize): New function.
21303 * tree-inline.c (remap_gimple_stmt): Pass original statement to
21304 modify_gimple_stmt.
21305 (copy_phis_for_bb): Do not copy dead PHI nodes.
21306 (expand_call_inline): Do not remap performed_splits.
21307 (update_clone_info): Likewise.
21308 * toplev.c: Include ipa-param-manipulation.h.
21309 (toplev::finalize): Call ipa_edge_modifications_finalize.
21311 2021-06-28 Andrew Pinski <apinski@marvell.com>
21313 * tree-ssa-phiopt.c (replace_phi_edge_with_variable): Duplicate range
21314 info if we're the only things setting the target PHI.
21315 (value_replacement): Don't duplicate range here.
21316 (minmax_replacement): Likewise.
21318 2021-06-28 Richard Biener <rguenther@suse.de>
21320 PR tree-optimization/101229
21321 * gimple-walk.c (gimple_walk_op): Handle PHIs.
21323 2021-06-28 Martin Liska <mliska@suse.cz>
21325 * config/v850/v850.c (construct_dispose_instruction): Allocate
21327 (construct_prepare_instruction): Likewise.
21329 2021-06-28 Martin Liska <mliska@suse.cz>
21331 * config/v850/v850.c (v850_option_override): Build default
21333 (v850_can_inline_p): New. Allow MASK_PROLOG_FUNCTION to be
21334 ignored for inlining.
21335 (TARGET_CAN_INLINE_P): New.
21337 2021-06-28 Richard Biener <rguenther@suse.de>
21339 PR tree-optimization/101207
21340 * tree-vect-slp.c (vect_optimize_slp): Do BB reduction
21341 permute eliding for load permutations properly.
21343 2021-06-28 Richard Biener <rguenther@suse.de>
21345 PR tree-optimization/101173
21346 * gimple-loop-interchange.cc
21347 (tree_loop_interchange::valid_data_dependences): Disallow outer
21348 loop dependence distance of zero.
21350 2021-06-28 liuhongt <hongtao.liu@intel.com>
21353 * config/i386/sse.md (*avx_cmp<mode>3_lt): New
21354 define_insn_and_split.
21355 (*avx_cmp<mode>3_ltint): Ditto.
21356 (*avx2_pcmp<mode>3_3): Ditto.
21357 (*avx2_pcmp<mode>3_4): Ditto.
21358 (*avx2_pcmp<mode>3_5): Ditto.
21360 2021-06-28 liuhongt <hongtao.liu@intel.com>
21362 * config/i386/i386-builtin.def (IX86_BUILTIN_BLENDVPD256,
21363 IX86_BUILTIN_BLENDVPS256, IX86_BUILTIN_PBLENDVB256,
21364 IX86_BUILTIN_BLENDVPD, IX86_BUILTIN_BLENDVPS,
21365 IX86_BUILTIN_PBLENDVB128): Replace icode with
21367 * config/i386/i386.c (ix86_gimple_fold_builtin): Fold blendv
21369 * config/i386/sse.md (*<sse4_1_avx2>_pblendvb_lt_subreg_not):
21370 New pre_reload splitter.
21372 2021-06-27 Andrew Pinski <apinski@marvell.com>
21374 PR middle-end/101230
21375 * fold-const.c (fold_ternary_loc): Check
21376 the return value of invert_tree_comparison.
21378 2021-06-27 David Edelsohn <dje.gcc@gmail.com>
21380 * config.gcc: Add SPDX License Identifier.
21381 (powerpc-ibm-aix789): Default to aix73.h.
21382 (powerpc-ibm-aix7.2.*.*): New stanza.
21383 * config/rs6000/aix72.h: Add SPDX License Identifier.
21384 * config/rs6000/aix73.h: New file.
21386 2021-06-26 Jason Merrill <jason@redhat.com>
21388 * except.c: #include "dwarf2.h" instead of "dwarf2out.h".
21390 2021-06-26 Andrew Pinski <apinski@marvell.com>
21392 * genmatch.c (lower_cond): Copy for_subst_vec
21393 for the simplify also.
21394 (lower): Swap the order for lower_for and lower_cond.
21396 2021-06-26 Andrew Pinski <apinski@marvell.com>
21398 * tree-ssa-phiopt.c (match_simplify_replacement): Reset
21399 flow senatitive info on the moved ssa set.
21401 2021-06-26 Andrew Pinski <apinski@marvell.com>
21403 * fold-const.c (fold_cond_expr_with_comparison):
21404 Exand arg0 into comp_code, arg00, and arg01.
21405 (fold_ternary_loc): Use invert_tree_comparison
21406 instead of fold_invert_truthvalue for the case
21407 where we have A CMP B ? C : A.
21409 2021-06-25 Martin Sebor <msebor@redhat.com>
21411 PR middle-end/101216
21412 * calls.c (maybe_warn_rdwr_sizes): Use the no_warning constant.
21414 2021-06-25 Jeff Law <jeffreyalaw@gmail.com>
21416 * config/h8300/h8300.c (select_cc_mode): Handle ASHIFTRT and LSHIFTRT.
21418 2021-06-25 Richard Biener <rguenther@suse.de>
21420 PR tree-optimization/101202
21421 * tree-vect-slp.c (vect_optimize_slp): Explicitely handle
21424 2021-06-25 Richard Biener <rguenther@suse.de>
21426 * tree-vect-slp-patterns.c (addsub_pattern::build): Copy
21427 STMT_VINFO_REDUC_DEF from the original representative.
21429 2021-06-25 Martin Sebor <msebor@redhat.com>
21431 * builtins.c (warn_string_no_nul): Replace uses of TREE_NO_WARNING,
21432 gimple_no_warning_p and gimple_set_no_warning with
21433 warning_suppressed_p, and suppress_warning.
21435 (maybe_warn_for_bound): Same.
21436 (warn_for_access): Same.
21437 (check_access): Same.
21438 (expand_builtin_strncmp): Same.
21439 (fold_builtin_varargs): Same.
21440 * calls.c (maybe_warn_nonstring_arg): Same.
21441 (maybe_warn_rdwr_sizes): Same.
21442 * cfgexpand.c (expand_call_stmt): Same.
21443 * cgraphunit.c (check_global_declaration): Same.
21444 * fold-const.c (fold_undefer_overflow_warnings): Same.
21445 (fold_truth_not_expr): Same.
21446 (fold_unary_loc): Same.
21447 (fold_checksum_tree): Same.
21448 * gimple-array-bounds.cc (array_bounds_checker::check_array_ref): Same.
21449 (array_bounds_checker::check_mem_ref): Same.
21450 (array_bounds_checker::check_addr_expr): Same.
21451 (array_bounds_checker::check_array_bounds): Same.
21452 * gimple-expr.c (copy_var_decl): Same.
21453 * gimple-fold.c (gimple_fold_builtin_strcpy): Same.
21454 (gimple_fold_builtin_strncat): Same.
21455 (gimple_fold_builtin_stxcpy_chk): Same.
21456 (gimple_fold_builtin_stpcpy): Same.
21457 (gimple_fold_builtin_sprintf): Same.
21458 (fold_stmt_1): Same.
21459 * gimple-ssa-isolate-paths.c (diag_returned_locals): Same.
21460 * gimple-ssa-nonnull-compare.c (do_warn_nonnull_compare): Same.
21461 * gimple-ssa-sprintf.c (handle_printf_call): Same.
21462 * gimple-ssa-store-merging.c (imm_store_chain_info::output_merged_store): Same.
21463 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Same.
21464 * gimple-ssa-warn-restrict.h: Adjust declarations.
21465 (maybe_diag_access_bounds): Replace uses of TREE_NO_WARNING,
21466 gimple_no_warning_p and gimple_set_no_warning with
21467 warning_suppressed_p, and suppress_warning.
21468 (check_call): Same.
21469 (check_bounds_or_overlap): Same.
21470 * gimple.c (gimple_build_call_from_tree): Same.
21471 * gimplify.c (gimplify_return_expr): Same.
21472 (gimplify_cond_expr): Same.
21473 (gimplify_modify_expr_complex_part): Same.
21474 (gimplify_modify_expr): Same.
21475 (gimple_push_cleanup): Same.
21476 (gimplify_expr): Same.
21477 * omp-expand.c (expand_omp_for_generic): Same.
21478 (expand_omp_taskloop_for_outer): Same.
21479 * omp-low.c (lower_rec_input_clauses): Same.
21480 (lower_lastprivate_clauses): Same.
21481 (lower_send_clauses): Same.
21482 (lower_omp_target): Same.
21483 * tree-cfg.c (pass_warn_function_return::execute): Same.
21484 * tree-complex.c (create_one_component_var): Same.
21485 * tree-inline.c (remap_gimple_op_r): Same.
21486 (copy_tree_body_r): Same.
21487 (declare_return_variable): Same.
21488 (expand_call_inline): Same.
21489 * tree-nested.c (lookup_field_for_decl): Same.
21490 * tree-sra.c (create_access_replacement): Same.
21491 (generate_subtree_copies): Same.
21492 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Same.
21493 * tree-ssa-forwprop.c (combine_cond_expr_cond): Same.
21494 * tree-ssa-loop-ch.c (ch_base::copy_headers): Same.
21495 * tree-ssa-loop-im.c (execute_sm): Same.
21496 * tree-ssa-phiopt.c (cond_store_replacement): Same.
21497 * tree-ssa-strlen.c (maybe_warn_overflow): Same.
21498 (handle_builtin_strcpy): Same.
21499 (maybe_diag_stxncpy_trunc): Same.
21500 (handle_builtin_stxncpy_strncat): Same.
21501 (handle_builtin_strcat): Same.
21502 * tree-ssa-uninit.c (get_no_uninit_warning): Same.
21503 (set_no_uninit_warning): Same.
21504 (uninit_undefined_value_p): Same.
21505 (warn_uninit): Same.
21506 (maybe_warn_operand): Same.
21507 * tree-vrp.c (compare_values_warnv): Same.
21508 * vr-values.c (vr_values::extract_range_for_var_from_comparison_expr): Same.
21509 (test_for_singularity): Same.
21510 * gimple.h (warning_suppressed_p): New function.
21511 (suppress_warning): Same.
21512 (copy_no_warning): Same.
21513 (gimple_set_block): Call gimple_set_location.
21514 (gimple_set_location): Call copy_warning.
21516 2021-06-25 Martin Sebor <msebor@redhat.com>
21518 * tree.h (warning_suppressed_at, copy_warning,
21519 warning_suppressed_p, suppress_warning): New functions.
21521 2021-06-25 Martin Sebor <msebor@redhat.com>
21523 * Makefile.in (OBJS-libcommon): Add diagnostic-spec.o.
21524 * gengtype.c (open_base_files): Add diagnostic-spec.h.
21525 * diagnostic-spec.c: New file.
21526 * diagnostic-spec.h: New file.
21527 * tree.h (no_warning, all_warnings, suppress_warning_at): New
21529 * warning-control.cc: New file.
21531 2021-06-25 liuhongt <hongtao.liu@intel.com>
21534 * config/i386/i386.c (x86_order_regs_for_local_alloc):
21537 2021-06-24 Andrew MacLeod <amacleod@redhat.com>
21539 PR tree-optimization/101189
21540 * gimple-range-fold.cc (fold_using_range::range_of_range_op): Pass
21541 LHS range of condition to postfold routine.
21542 (fold_using_range::postfold_gcond_edges): Only process the TRUE or
21543 FALSE edge if the LHS range supports it being taken.
21544 * gimple-range-fold.h (postfold_gcond_edges): Add range parameter.
21546 2021-06-24 Andrew MacLeod <amacleod@redhat.com>
21548 * value-relation.cc (equiv_oracle::dump): Do not dump NULL blocks.
21549 (relation_oracle::find_relation_block): Check correct bitmap.
21550 (relation_oracle::dump): Do not dump NULL blocks.
21552 2021-06-24 Andrew MacLeod <amacleod@redhat.com>
21554 * gimple-range-cache.cc (ranger_cache::propagate_cache): Call
21555 range_on_edge instead of manually calculating.
21557 2021-06-24 Andrew MacLeod <amacleod@redhat.com>
21559 * range-op.cc: Fix comment.
21561 2021-06-24 Uroš Bizjak <ubizjak@gmail.com>
21564 * config/i386/i386-expand.c (ix86_expand_sse_unpack):
21565 Handle V8QI and V4HI modes.
21566 * config/i386/mmx.md (sse4_1_<any_extend:code>v4qiv4hi2):
21568 (sse4_1_<any_extend:code>v4qiv4hi2): Ditto.
21569 (mmxpackmode): New mode attribute.
21570 (vec_pack_trunc_<mmxpackmode:mode>): New expander.
21571 (mmxunpackmode): New mode attribute.
21572 (vec_unpacks_lo_<mmxunpackmode:mode>): New expander.
21573 (vec_unpacks_hi_<mmxunpackmode:mode>): Ditto.
21574 (vec_unpacku_lo_<mmxunpackmode:mode>): Ditto.
21575 (vec_unpacku_hi_<mmxunpackmode:mode>): Ditto.
21576 * config/i386/i386.md (extsuffix): Move from ...
21577 * config/i386/sse.md: ... here.
21579 2021-06-24 Eric Botcazou <ebotcazou@adacore.com>
21581 * dwarf2out.c (dwarf2out_assembly_start): Emit .file 0 marker here..
21582 (dwarf2out_finish): ...instead of here.
21584 2021-06-24 Eric Botcazou <ebotcazou@adacore.com>
21586 * configure.ac (--gdwarf-5 option): Use objdump instead of readelf.
21587 (working --gdwarf-4/--gdwarf-5 for all sources): Likewise.
21588 (--gdwarf-4 not refusing generated .debug_line): Adjust for Windows.
21589 * configure: Regenerate.
21591 2021-06-24 Richard Biener <rguenther@suse.de>
21593 * config/i386/sse.md (vec_addsubv4df3, vec_addsubv2df3,
21594 vec_addsubv8sf3, vec_addsubv4sf3): Merge into ...
21595 (vec_addsub<mode>3): ... using a new addsub_cst mode attribute.
21597 2021-06-24 Richard Biener <rguenther@suse.de>
21599 * config/i386/sse.md (avx_addsubv4df3): Rename to
21601 (avx_addsubv8sf3): Rename to vec_addsubv8sf3.
21602 (sse3_addsubv2df3): Rename to vec_addsubv2df3.
21603 (sse3_addsubv4sf3): Rename to vec_addsubv4sf3.
21604 * config/i386/i386-builtin.def: Adjust.
21605 * internal-fn.def (VEC_ADDSUB): New internal optab fn.
21606 * optabs.def (vec_addsub_optab): New optab.
21607 * tree-vect-slp-patterns.c (class addsub_pattern): New.
21608 (slp_patterns): Add addsub_pattern.
21609 * tree-vect-slp.c (vect_optimize_slp): Disable propagation
21610 across CFN_VEC_ADDSUB.
21611 * tree-vectorizer.h (vect_pattern::vect_pattern): Make
21613 * doc/md.texi (vec_addsub<mode>3): Document.
21615 2021-06-24 Jakub Jelinek <jakub@redhat.com>
21617 PR middle-end/101170
21618 * df-scan.c (df_ref_record): For paradoxical big-endian SUBREGs
21619 where regno + subreg_regno_offset wraps around use 0 as starting
21622 2021-06-24 Jakub Jelinek <jakub@redhat.com>
21624 PR middle-end/101172
21625 * stor-layout.c (finish_bitfield_representative): If nextf has
21626 error_mark_node type, set repr type to error_mark_node too.
21628 2021-06-24 Ilya Leoshkevich <iii@linux.ibm.com>
21630 * config/s390/s390.c (s390_function_profiler): Ignore labelno
21632 * config/s390/s390.h (NO_PROFILE_COUNTERS): Define.
21634 2021-06-24 Richard Biener <rguenther@suse.de>
21636 * tree-vect-slp.c (vect_optimize_slp): Do not propagate
21637 across operations that have different semantics on different
21640 2021-06-24 Jakub Jelinek <jakub@redhat.com>
21642 * tree.h (OMP_CLAUSE_MAP_IN_REDUCTION): Document meaning for OpenMP.
21643 * gimplify.c (gimplify_scan_omp_clauses): For OpenMP map clauses
21644 with OMP_CLAUSE_MAP_IN_REDUCTION flag partially defer gimplification
21645 of non-decl OMP_CLAUSE_DECL. For OMP_CLAUSE_IN_REDUCTION on
21646 OMP_TARGET user outer_ctx instead of ctx for placeholders and
21647 initializer/combiner gimplification.
21648 * omp-low.c (scan_sharing_clauses): Handle OMP_CLAUSE_MAP_IN_REDUCTION
21649 on target constructs.
21650 (lower_rec_input_clauses): Likewise.
21651 (lower_omp_target): Likewise.
21652 * omp-expand.c (expand_omp_target): Temporarily ignore nowait clause
21653 on target if in_reduction is present.
21655 2021-06-24 Kewen Lin <linkw@linux.ibm.com>
21657 * tree-predcom.c (class pcom_worker): New class.
21658 (release_chain): Renamed to...
21659 (pcom_worker::release_chain): ...this.
21660 (release_chains): Renamed to...
21661 (pcom_worker::release_chains): ...this.
21662 (aff_combination_dr_offset): Renamed to...
21663 (pcom_worker::aff_combination_dr_offset): ...this.
21664 (determine_offset): Renamed to...
21665 (pcom_worker::determine_offset): ...this.
21666 (class comp_ptrs): New class.
21667 (split_data_refs_to_components): Renamed to...
21668 (pcom_worker::split_data_refs_to_components): ...this,
21669 and update with class comp_ptrs.
21670 (suitable_component_p): Renamed to...
21671 (pcom_worker::suitable_component_p): ...this.
21672 (filter_suitable_components): Renamed to...
21673 (pcom_worker::filter_suitable_components): ...this.
21674 (valid_initializer_p): Renamed to...
21675 (pcom_worker::valid_initializer_p): ...this.
21676 (find_looparound_phi): Renamed to...
21677 (pcom_worker::find_looparound_phi): ...this.
21678 (add_looparound_copies): Renamed to...
21679 (pcom_worker::add_looparound_copies): ...this.
21680 (determine_roots_comp): Renamed to...
21681 (pcom_worker::determine_roots_comp): ...this.
21682 (determine_roots): Renamed to...
21683 (pcom_worker::determine_roots): ...this.
21684 (single_nonlooparound_use): Renamed to...
21685 (pcom_worker::single_nonlooparound_use): ...this.
21686 (remove_stmt): Renamed to...
21687 (pcom_worker::remove_stmt): ...this.
21688 (execute_pred_commoning_chain): Renamed to...
21689 (pcom_worker::execute_pred_commoning_chain): ...this.
21690 (execute_pred_commoning): Renamed to...
21691 (pcom_worker::execute_pred_commoning): ...this.
21692 (struct epcc_data): New member worker.
21693 (execute_pred_commoning_cbck): Call execute_pred_commoning
21694 with pcom_worker pointer.
21695 (find_use_stmt): Renamed to...
21696 (pcom_worker::find_use_stmt): ...this.
21697 (find_associative_operation_root): Renamed to...
21698 (pcom_worker::find_associative_operation_root): ...this.
21699 (find_common_use_stmt): Renamed to...
21700 (pcom_worker::find_common_use_stmt): ...this.
21701 (combinable_refs_p): Renamed to...
21702 (pcom_worker::combinable_refs_p): ...this.
21703 (reassociate_to_the_same_stmt): Renamed to...
21704 (pcom_worker::reassociate_to_the_same_stmt): ...this.
21705 (stmt_combining_refs): Renamed to...
21706 (pcom_worker::stmt_combining_refs): ...this.
21707 (combine_chains): Renamed to...
21708 (pcom_worker::combine_chains): ...this.
21709 (try_combine_chains): Renamed to...
21710 (pcom_worker::try_combine_chains): ...this.
21711 (prepare_initializers_chain): Renamed to...
21712 (pcom_worker::prepare_initializers_chain): ...this.
21713 (prepare_initializers): Renamed to...
21714 (pcom_worker::prepare_initializers): ...this.
21715 (prepare_finalizers_chain): Renamed to...
21716 (pcom_worker::prepare_finalizers_chain): ...this.
21717 (prepare_finalizers): Renamed to...
21718 (pcom_worker::prepare_finalizers): ...this.
21719 (tree_predictive_commoning_loop): Renamed to...
21720 (pcom_worker::tree_predictive_commoning_loop): ...this, adjust
21721 some calls and remove some cleanup code.
21722 (tree_predictive_commoning): Adjusted to use pcom_worker instance.
21723 (static variable looparound_phis): Remove.
21724 (static variable name_expansions): Remove.
21726 2021-06-24 Richard Biener <rguenther@suse.de>
21728 * tree-vect-slp.c (slpg_vertex): New struct.
21729 (vect_slp_build_vertices): Adjust.
21730 (vect_optimize_slp): Likewise. Maintain an outgoing permute
21731 and a materialized one.
21733 2021-06-24 Richard Biener <rguenther@suse.de>
21735 PR tree-optimization/101105
21736 * tree-vect-data-refs.c (vect_prune_runtime_alias_test_list):
21737 Only ignore steps when they are equal or scalar order is preserved.
21739 2021-06-24 liuhongt <hongtao.liu@intel.com>
21742 * config/i386/i386-expand.c (ix86_expand_vec_interleave):
21743 Adjust comments for ix86_expand_vecop_qihi2.
21744 (ix86_expand_vecmul_qihi): Renamed to ..
21745 (ix86_expand_vecop_qihi2): Adjust function prototype to
21746 support shift operation, add static to definition.
21747 (ix86_expand_vec_shift_qihi_constant): Add static to definition.
21748 (ix86_expand_vecop_qihi): Call ix86_expand_vecop_qihi2 and
21749 ix86_expand_vec_shift_qihi_constant.
21750 * config/i386/i386-protos.h (ix86_expand_vecmul_qihi): Deleted.
21751 (ix86_expand_vec_shift_qihi_constant): Deleted.
21752 * config/i386/sse.md (VI12_256_512_AVX512VL): New mode
21754 (mulv8qi3): Call ix86_expand_vecop_qihi directly, add
21755 condition TARGET_64BIT.
21756 (mul<mode>3): Ditto.
21757 (<insn><mode>3): Ditto.
21758 (vlshr<mode>3): Extend to support avx512 vlshr.
21759 (v<insn><mode>3): New expander for
21761 (v<insn>v8qi3): Ditto.
21762 (vashrv8hi3<mask_name>): Renamed to ..
21763 (vashr<mode>3): And extend to support V16QImode for avx512.
21764 (vashrv16qi3): Deleted.
21765 (vashrv2di3<mask_name>): Extend expander to support avx512
21768 2021-06-23 Dimitar Dimitrov <dimitar@dinux.eu>
21770 * doc/lto.texi (Design Overview): Update that slim objects are
21773 2021-06-23 Aaron Sawdey <acsawdey@linux.ibm.com>
21775 * config/rs6000/rs6000-cpus.def: Take OPTION_MASK_PCREL_OPT out
21776 of OTHER_POWER10_MASKS so it will not be enabled by default.
21778 2021-06-23 Richard Biener <rguenther@suse.de>
21779 Martin Jambor <mjambor@suse.cz>
21781 * tree-inline.c (setup_one_parameter): Set TREE_READONLY of the
21782 param replacement unconditionally. Adjust comment.
21784 2021-06-23 Andrew MacLeod <amacleod@redhat.com>
21786 * Makefile.in (OBJS): Add gimple-range-fold.o
21787 * gimple-range-fold.cc: New.
21788 * gimple-range-fold.h: New.
21789 * gimple-range-gori.cc (gimple_range_calc_op1): Move to here.
21790 (gimple_range_calc_op2): Ditto.
21791 * gimple-range-gori.h: Move prototypes to here.
21792 * gimple-range.cc: Adjust include files.
21793 (fur_source:fur_source): Relocate to gimple-range-fold.cc.
21794 (fur_source::get_operand): Ditto.
21795 (fur_source::get_phi_operand): Ditto.
21796 (fur_source::query_relation): Ditto.
21797 (fur_source::register_relation): Ditto.
21798 (class fur_edge): Ditto.
21799 (fur_edge::fur_edge): Ditto.
21800 (fur_edge::get_operand): Ditto.
21801 (fur_edge::get_phi_operand): Ditto.
21802 (fur_stmt::fur_stmt): Ditto.
21803 (fur_stmt::get_operand): Ditto.
21804 (fur_stmt::get_phi_operand): Ditto.
21805 (fur_stmt::query_relation): Ditto.
21806 (class fur_depend): Relocate to gimple-range-fold.h.
21807 (fur_depend::fur_depend): Relocate to gimple-range-fold.cc.
21808 (fur_depend::register_relation): Ditto.
21809 (fur_depend::register_relation): Ditto.
21810 (class fur_list): Ditto.
21811 (fur_list::fur_list): Ditto.
21812 (fur_list::get_operand): Ditto.
21813 (fur_list::get_phi_operand): Ditto.
21814 (fold_range): Ditto.
21815 (adjust_pointer_diff_expr): Ditto.
21816 (gimple_range_adjustment): Ditto.
21817 (gimple_range_base_of_assignment): Ditto.
21818 (gimple_range_operand1): Ditto.
21819 (gimple_range_operand2): Ditto.
21820 (gimple_range_calc_op1): Relocate to gimple-range-gori.cc.
21821 (gimple_range_calc_op2): Ditto.
21822 (fold_using_range::fold_stmt): Relocate to gimple-range-fold.cc.
21823 (fold_using_range::range_of_range_op): Ditto.
21824 (fold_using_range::range_of_address): Ditto.
21825 (fold_using_range::range_of_phi): Ditto.
21826 (fold_using_range::range_of_call): Ditto.
21827 (fold_using_range::range_of_builtin_ubsan_call): Ditto.
21828 (fold_using_range::range_of_builtin_call): Ditto.
21829 (fold_using_range::range_of_cond_expr): Ditto.
21830 (fold_using_range::range_of_ssa_name_with_loop_info): Ditto.
21831 (fold_using_range::relation_fold_and_or): Ditto.
21832 (fold_using_range::postfold_gcond_edges): Ditto.
21833 * gimple-range.h: Add gimple-range-fold.h to include files. Change
21834 GIMPLE_RANGE_STMT_H to GIMPLE_RANGE_H.
21835 (gimple_range_handler): Relocate to gimple-range-fold.h.
21836 (gimple_range_ssa_p): Ditto.
21837 (range_compatible_p): Ditto.
21838 (class fur_source): Ditto.
21839 (class fur_stmt): Ditto.
21840 (class fold_using_range): Ditto.
21841 (gimple_range_calc_op1): Relocate to gimple-range-gori.h
21842 (gimple_range_calc_op2): Ditto.
21844 2021-06-23 Andrew MacLeod <amacleod@redhat.com>
21846 PR tree-optimization/101148
21847 PR tree-optimization/101014
21848 * gimple-range-cache.cc (ranger_cache::ranger_cache): Adjust.
21849 (ranger_cache::~ranger_cache): Adjust.
21850 (ranger_cache::block_range): Check if propagation disallowed.
21851 (ranger_cache::propagate_cache): Disallow propagation if new value
21852 can't be stored properly.
21853 * gimple-range-cache.h (ranger_cache::m_propfail): New member.
21855 2021-06-23 Andrew MacLeod <amacleod@redhat.com>
21857 * gimple-range-cache.cc (class ssa_block_ranges): Adjust prototype.
21858 (sbr_vector::set_bb_range): Return true.
21859 (class sbr_sparse_bitmap): Adjust.
21860 (sbr_sparse_bitmap::set_bb_range): Return value.
21861 (block_range_cache::set_bb_range): Return value.
21862 (ranger_cache::propagate_cache): Use return value to print msg.
21863 * gimple-range-cache.h (class block_range_cache): Adjust.
21865 2021-06-23 Andrew MacLeod <amacleod@redhat.com>
21867 * gimple-range.cc (dump_bb): Use range_on_edge from the cache.
21869 2021-06-23 Jeff Law <jeffreyalaw@gmail.com>
21871 * config/h8300/logical.md (<code><mode>3<ccnz>): Use <cczn>
21872 so this pattern can be used for test/compare removal. Pass
21873 current insn to compute_logical_op_length and output_logical_op.
21874 * config/h8300/h8300.c (compute_logical_op_cc): Remove.
21875 (h8300_and_costs): Add argument to compute_logical_op_length.
21876 (output_logical_op): Add new argument. Use it to determine if the
21877 condition codes are used and adjust the output accordingly.
21878 (compute_logical_op_length): Add new argument and update length
21879 computations when condition codes are used.
21880 * config/h8300/h8300-protos.h (compute_logical_op_length): Update
21882 (output_logical_op): Likewise.
21884 2021-06-23 Uroš Bizjak <ubizjak@gmail.com>
21887 * config/i386/i386-expand.c (expand_vec_perm_pshufb):
21888 Handle 64bit modes for TARGET_XOP. Use indirect gen_* functions.
21889 * config/i386/mmx.md (mmx_ppermv64): New insn pattern.
21890 * config/i386/i386.md (unspec): Move UNSPEC_XOP_PERMUTE from ...
21891 * config/i386/sse.md (unspec): ... here.
21893 2021-06-23 Martin Liska <mliska@suse.cz>
21896 * optc-save-gen.awk: Put back arm_fp16_format to
21899 2021-06-23 Uroš Bizjak <ubizjak@gmail.com>
21902 * config/i386/i386.md (bsr_rex64): Add zero-flag setting RTX.
21905 (clz<mode>2): Update RTX pattern for additions.
21907 2021-06-23 Jakub Jelinek <jakub@redhat.com>
21909 PR middle-end/101167
21910 * omp-low.c (lower_omp_regimplify_p): Regimplify also PARM_DECLs
21911 and RESULT_DECLs that have DECL_HAS_VALUE_EXPR_P set.
21913 2021-06-22 Sergei Trofimovich <siarheit@google.com>
21915 * doc/rtl.texi: drop unbalanced parenthesis.
21917 2021-06-22 Richard Biener <rguenther@suse.de>
21919 PR middle-end/101156
21920 * gimplify.c (gimplify_expr): Remove premature incorrect
21923 2021-06-22 Jakub Jelinek <jakub@redhat.com>
21925 PR tree-optimization/101159
21926 * tree-vect-patterns.c (vect_recog_popcount_pattern): Fix some
21929 2021-06-22 Jakub Jelinek <jakub@redhat.com>
21931 PR middle-end/101160
21932 * function.c (assign_parms): For decl_result with TYPE_EMPTY_P type
21933 clear crtl->return_rtx instead of keeping it referencing a pseudo.
21935 2021-06-22 Jakub Jelinek <jakub@redhat.com>
21936 Andrew Pinski <apinski@marvell.com>
21938 PR tree-optimization/101162
21939 * fold-const.c (range_check_type): Handle OFFSET_TYPE like pointer
21942 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
21944 * range-op.cc (range_relational_tests): New.
21945 (range_op_tests): Call range_relational_tests.
21947 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
21949 * range-op.cc (operator_cast::lhs_op1_relation): New.
21950 (operator_identity::lhs_op1_relation): Mew.
21952 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
21954 * range-op.cc (operator_minus::op1_op2_relation_effect): New.
21956 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
21958 * range-op.cc (operator_plus::lhs_op1_relation): New.
21959 (operator_plus::lhs_op2_relation): New.
21961 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
21963 * gimple-range-cache.cc (ranger_cache::ranger_cache): Create a
21964 relation_oracle if dominators exist.
21965 (ranger_cache::~ranger_cache): Dispose of oracle.
21966 (ranger_cache::dump_bb): Dump oracle.
21967 * gimple-range.cc (fur_source::fur_source): New.
21968 (fur_source::get_operand): Use mmeber query.
21969 (fur_source::get_phi_operand): Use member_query.
21970 (fur_source::query_relation): New.
21971 (fur_source::register_dependency): Delete.
21972 (fur_source::register_relation): New.
21973 (fur_edge::fur_edge): Adjust.
21974 (fur_edge::get_phi_operand): Fix comment.
21975 (fur_edge::query): Delete.
21976 (fur_stmt::fur_stmt): Adjust.
21977 (fur_stmt::query): Delete.
21978 (fur_depend::fur_depend): Adjust.
21979 (fur_depend::register_relation): New.
21980 (fur_depend::register_relation): New.
21981 (fur_list::fur_list): Adjust.
21982 (fur_list::get_operand): Use member query.
21983 (fold_using_range::range_of_range_op): Process and query relations.
21984 (fold_using_range::range_of_address): Adjust dependency call.
21985 (fold_using_range::range_of_phi): Ditto.
21986 (gimple_ranger::gimple_ranger): New. Use ranger_ache oracle.
21987 (fold_using_range::relation_fold_and_or): New.
21988 (fold_using_range::postfold_gcond_edges): New.
21989 * gimple-range.h (class gimple_ranger): Adjust.
21990 (class fur_source): Adjust members.
21991 (class fur_stmt): Ditto.
21992 (class fold_using_range): Ditto.
21994 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
21996 * range-op.cc (range_operator::wi_fold): Apply relation effect.
21997 (range_operator::fold_range): Adjust and apply relation effect.
21998 (*::fold_range): Add relation parameters.
21999 (*::op1_range): Ditto.
22000 (*::op2_range): Ditto.
22001 (range_operator::lhs_op1_relation): New.
22002 (range_operator::lhs_op2_relation): New.
22003 (range_operator::op1_op2_relation): New.
22004 (range_operator::op1_op2_relation_effect): New.
22005 (relop_early_resolve): New.
22006 (operator_equal::op1_op2_relation): New.
22007 (operator_equal::fold_range): Call relop_early_resolve.
22008 (operator_not_equal::op1_op2_relation): New.
22009 (operator_not_equal::fold_range): Call relop_early_resolve.
22010 (operator_lt::op1_op2_relation): New.
22011 (operator_lt::fold_range): Call relop_early_resolve.
22012 (operator_le::op1_op2_relation): New.
22013 (operator_le::fold_range): Call relop_early_resolve.
22014 (operator_gt::op1_op2_relation): New.
22015 (operator_gt::fold_range): Call relop_early_resolve.
22016 (operator_ge::op1_op2_relation): New.
22017 (operator_ge::fold_range): Call relop_early_resolve.
22018 * range-op.h (class range_operator): Adjust parameters and methods.
22020 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
22022 * Makefile.in (OBJS): Add value-relation.o.
22023 * gimple-range.h: Adjust include files.
22024 * tree-data-ref.c: Adjust include file order.
22025 * value-query.cc (range_query::get_value_range): Default to no oracle.
22026 (range_query::query_relation): New.
22027 (range_query::query_relation): New.
22028 * value-query.h (class range_query): Adjust.
22029 * value-relation.cc: New.
22030 * value-relation.h: New.
22032 2021-06-22 Richard Biener <rguenther@suse.de>
22034 PR tree-optimization/101151
22035 * tree-ssa-sink.c (statement_sink_location): Expand irreducible
22038 2021-06-22 Jojo R <rjiejie@linux.alibaba.com>
22040 * config/riscv/riscv.c (thead_c906_tune_info): New.
22041 (riscv_tune_info_table): Use new tune.
22043 2021-06-22 Richard Biener <rguenther@suse.de>
22045 PR tree-optimization/101158
22046 * tree-vect-slp.c (vect_build_slp_tree_1): Move same operand
22047 checking after checking for matching operation.
22049 2021-06-22 Richard Biener <rguenther@suse.de>
22051 PR tree-optimization/101159
22052 * tree-vect-patterns.c (vect_recog_popcount_pattern): Add
22053 missing NULL vectype check.
22055 2021-06-22 Richard Biener <rguenther@suse.de>
22057 PR tree-optimization/101154
22058 * tree-vect-slp.c (vect_build_slp_tree_2): Fix out-of-bound access.
22060 2021-06-22 Jakub Jelinek <jakub@redhat.com>
22063 * config/i386/i386-protos.h (ix86_last_zero_store_uid): Declare.
22064 * config/i386/i386-expand.c (ix86_last_zero_store_uid): New variable.
22065 * config/i386/i386.c (ix86_expand_prologue): Clear it.
22066 * config/i386/i386.md (peephole2s for 1/2/4 stores of const0_rtx):
22067 Remove "" from match_operand. Emit new insns using emit_move_insn and
22068 set ix86_last_zero_store_uid to INSN_UID of the last store.
22069 Add peephole2s for 1/2/4 stores of const0_rtx following previous
22072 2021-06-22 Martin Liska <mliska@suse.cz>
22074 * auto-profile.c (AUTO_PROFILE_VERSION): Bump as string format
22077 2021-06-22 Martin Liska <mliska@suse.cz>
22079 * gcov-io.h: Remove padding entries.
22081 2021-06-22 liuhongt <hongtao.liu@intel.com>
22083 PR tree-optimization/97770
22084 * tree-vect-patterns.c (vect_recog_popcount_pattern):
22086 (vect_recog_func vect_vect_recog_func_ptrs): Add new pattern.
22088 2021-06-22 liuhongt <hongtao.liu@intel.com>
22091 * config/i386/i386-builtin.def (BDESC): Adjust builtin name.
22092 * config/i386/sse.md (<avx512>_expand<mode>_mask): Rename to ..
22093 (expand<mode>_mask): this ..
22094 (*expand<mode>_mask): New pre_reload splitter to transform
22095 v{,p}expand* to vmov* when mask is zero, all ones, or has all
22096 ones in it's lower part, otherwise still generate
22099 2021-06-22 liuhongt <hongtao.liu@intel.com>
22102 * config/i386/i386-expand.c
22103 (ix86_expand_special_args_builtin): Keep constm1_operand only
22104 if it satisfies insn's operand predicate.
22106 2021-06-21 Jason Merrill <jason@redhat.com>
22109 * df-scan.c (df_ref_record): Check that regno < endregno.
22110 * function.c (assign_parms, expand_function_end): Do nothing with a
22111 TYPE_EMPTY_P result.
22113 2021-06-21 Richard Biener <rguenther@suse.de>
22115 PR tree-optimization/101120
22116 * tree-vect-data-refs.c (bump_vector_ptr): Fold the
22118 * tree-vect-slp.c (vect_transform_slp_perm_load): Add
22119 DR chain DCE capability.
22120 * tree-vectorizer.h (vect_transform_slp_perm_load): Adjust.
22121 * tree-vect-stmts.c (vectorizable_load): Remove unused
22122 loads in the DR chain for SLP.
22124 2021-06-21 Jakub Jelinek <jakub@redhat.com>
22126 PR inline-asm/100785
22127 * gimplify.c (gimplify_asm_expr): Don't diagnose errors if
22128 output or input operands were already error_mark_node.
22129 * cfgexpand.c (expand_asm_stmt): If errors are emitted,
22130 remove all inputs, outputs and clobbers from the asm and
22131 set template to "".
22133 2021-06-21 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
22135 * config/arm/arm_neon.h (vceq_s8): Replace builtin with __a == __b.
22136 (vceq_s16): Likewise.
22137 (vceq_s32): Likewise.
22138 (vceq_u8): Likewise.
22139 (vceq_u16): Likewise.
22140 (vceq_u32): Likewise.
22141 (vceq_p8): Likewise.
22142 (vceqq_s8): Likewise.
22143 (vceqq_s16): Likewise.
22144 (vceqq_s32): Likewise.
22145 (vceqq_u8): Likewise.
22146 (vceqq_u16): Likewise.
22147 (vceqq_u32): Likewise.
22148 (vceqq_p8): Likewise.
22149 (vceq_f32): Gate __a == __b on __FAST_MATH__.
22150 (vceqq_f32): Likewise.
22151 (vceq_f16): Likewise.
22152 (vceqq_f16): Likewise.
22154 2021-06-21 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
22157 * config/arm/iterators.md (NEON_VACMP): Remove.
22158 * config/arm/neon.md (neon_vca<cmp_op><mode>): Use GLTE instead of GTGE
22160 (neon_vca<cmp_op><mode>_insn): Likewise.
22161 (neon_vca<cmp_op_unsp><mode>_insn_unspec): Use NEON_VAGLTE instead of
22164 2021-06-21 Richard Biener <rguenther@suse.de>
22166 PR tree-optimization/101121
22167 * tree-vect-slp.c (vect_build_slp_tree_2): To not fail fatally
22168 when we just lack a stmt with the desired op when doing permutation.
22169 (vect_build_slp_tree): When caching a failed SLP build attempt
22170 assert that at least one lane is marked as not matching.
22172 2021-06-21 liuhongt <hongtao.liu@intel.com>
22175 * config/i386/i386.md: (*anddi_1): Disparage slightly the mask
22176 register alternative.
22177 (*and<mode>_1): Ditto.
22179 (*andn<mode>_1): Ditto.
22180 (*<code><mode>_1): Ditto.
22181 (*<code>qi_1): Ditto.
22182 (*one_cmpl<mode>2_1): Ditto.
22183 (*one_cmplsi2_1_zext): Ditto.
22184 (*one_cmplqi2_1): Ditto.
22185 * config/i386/i386.c (x86_order_regs_for_local_alloc): Change
22186 the order of mask registers to be before general registers.
22188 2021-06-21 Roger Sayle <roger@nextmovesoftware.com>
22191 * config/i386/i386.md: New define_peephole2s to shrink writing
22192 1, 2 or 4 consecutive zeros to memory when optimizing for size.
22194 2021-06-18 Jeff Law <jeffreyalaw@gmail.com>
22196 * config/h8300/h8300.c (h8300_select_cc_mode): Handle SYMBOL_REF.
22197 * config/h8300/logical.md (<code><mode>3 logcial expander): Generate
22198 more efficient code when the source can be trivially simplified.
22200 2021-06-18 Andrew MacLeod <amacleod@redhat.com>
22202 * gimple-range-cache.cc (ranger_cache::range_of_def): Calculate
22203 a range if global is not available.
22204 (ranger_cache::entry_range): Fallback to range_of_def.
22205 * gimple-range-cache.h (range_of_def): Adjust prototype.
22207 2021-06-18 Andrew MacLeod <amacleod@redhat.com>
22209 PR tree-optimization/101014
22210 * gimple-range-cache.cc (ranger_cache::ranger_cache): Remove poor
22212 (ranger_cache::~ranger_cache): Ditto.
22213 (ranger_cache::enable_new_values): Delete.
22214 (ranger_cache::push_poor_value): Delete.
22215 (ranger_cache::range_of_def): Remove poor value processing.
22216 (ranger_cache::entry_range): Ditto.
22217 (ranger_cache::fill_block_cache): Ditto.
22218 * gimple-range-cache.h (class ranger_cache): Remove poor value members.
22219 * gimple-range.cc (gimple_ranger::range_of_expr): Remove call.
22220 * gimple-range.h (class gimple_ranger): Adjust.
22222 2021-06-18 Srinath Parvathaneni <srinath.parvathaneni@arm.com>
22225 * common/config/arm/arm-common.c (arm_canon_arch_option_1): New function
22226 derived from arm_canon_arch.
22227 (arm_canon_arch_option): Call it.
22228 (arm_canon_arch_multilib_option): New function.
22229 * config/arm/arm-cpus.in (IGNORE_FOR_MULTILIB): New fgroup.
22230 * config/arm/arm.h (arm_canon_arch_multilib_option): New prototype.
22231 (CANON_ARCH_MULTILIB_SPEC_FUNCTION): New macro.
22232 (MULTILIB_ARCH_CANONICAL_SPECS): New macro.
22233 (DRIVER_SELF_SPECS): Add MULTILIB_ARCH_CANONICAL_SPECS.
22234 * config/arm/arm.opt (mlibarch): New option.
22235 * config/arm/t-rmprofile (MULTILIB_MATCHES): For armv8*-m, replace use
22236 of march on RHS with mlibarch.
22238 2021-06-18 Marcel Vollweiler <marcel@codesourcery.com>
22240 * config.in: Regenerate.
22241 * config/gcn/gcn.c (print_operand_address): Fix for global_load assembler
22243 * configure: Regenerate.
22244 * configure.ac: Fix for global_load assembler functions.
22246 2021-06-18 Richard Biener <rguenther@suse.de>
22248 PR tree-optimization/101112
22249 * tree-vect-slp.c (vect_slp_linearize_chain): Fix condition
22250 to lookup a pattern stmt def.
22252 2021-06-18 Jakub Jelinek <jakub@redhat.com>
22254 PR middle-end/101062
22255 * stor-layout.c (finish_bitfield_layout): Don't add bitfield
22256 representatives in QUAL_UNION_TYPE.
22258 2021-06-18 Andrew Pinski <apinski@marvell.com>
22260 * tree-ssa-phiopt.c (replace_phi_edge_with_variable):
22261 Add counting of how many times it is done.
22262 (factor_out_conditional_conversion): Likewise.
22263 (match_simplify_replacement): Likewise.
22264 (value_replacement): Likewise.
22265 (spaceship_replacement): Likewise.
22266 (cond_store_replacement): Likewise.
22267 (cond_if_else_store_replacement_1): Likewise.
22268 (hoist_adjacent_loads): Likewise.
22270 2021-06-18 Andrew Pinski <apinski@marvell.com>
22272 * tree-cfg.c (verify_gimple_assign_unary): Reject point and offset
22273 types on NEGATE_EXPR, ABS_EXPR, BIT_NOT_EXPR, PAREN_EXPR and CNONJ_EXPR.
22274 (verify_gimple_assign_binary): Reject point and offset types on
22275 MULT_EXPR, MULT_HIGHPART_EXPR, TRUNC_DIV_EXPR, CEIL_DIV_EXPR,
22276 FLOOR_DIV_EXPR, ROUND_DIV_EXPR, TRUNC_MOD_EXPR, CEIL_MOD_EXPR,
22277 FLOOR_MOD_EXPR, ROUND_MOD_EXPR, RDIV_EXPR, and EXACT_DIV_EXPR.
22279 2021-06-18 Michael Meissner <meissner@linux.ibm.com>
22281 * config/rs6000/rs6000.c (rs6000_emit_minmax): Add support for ISA
22282 3.1 IEEE 128-bit floating point xsmaxcqp/xsmincqp instructions.
22283 * config/rs6000/rs6000.md (s<minmax><mode>3, IEEE128 iterator):
22286 2021-06-17 Aaron Sawdey <acsawdey@linux.ibm.com>
22288 * config/rs6000/genfusion.pl (gen_logical_addsubf): Add
22289 earlyclobber to alts 0/1.
22290 (gen_addadd): Add earlyclobber to alts 0/1.
22291 * config/rs6000/fusion.md: Regenerate file.
22293 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
22295 * cfgloopanal.c (get_loop_hot_path): Make path an auto_vec.
22297 2021-06-17 Andrew MacLeod <amacleod@redhat.com>
22299 * gimple-range-cache.cc: Comment cleanups.
22300 * gimple-range-gori.cc: Comment cleanups.
22301 * gimple-range.cc: Comment/spacing cleanups
22302 * value-range.h: Comment cleanups.
22304 2021-06-17 H.J. Lu <hjl.tools@gmail.com>
22307 * calls.c (expand_call): Replace PUSH_ARGS with
22308 targetm.calls.push_argument (0).
22309 (emit_library_call_value_1): Likewise.
22310 * defaults.h (PUSH_ARGS): Removed.
22311 (PUSH_ARGS_REVERSED): Replace PUSH_ARGS with
22312 targetm.calls.push_argument (0).
22313 * expr.c (block_move_libcall_safe_for_call_parm): Likewise.
22314 (emit_push_insn): Pass the number bytes to push to
22315 targetm.calls.push_argument and pass 0 if ARGS_ADDR is 0.
22316 * hooks.c (hook_bool_uint_true): New.
22317 * hooks.h (hook_bool_uint_true): Likewise.
22318 * rtlanal.c (nonzero_bits1): Replace PUSH_ARGS with
22319 targetm.calls.push_argument (0).
22320 * target.def (push_argument): Add a targetm.calls hook.
22321 * targhooks.c (default_push_argument): New.
22322 * targhooks.h (default_push_argument): Likewise.
22323 * config/bpf/bpf.h (PUSH_ARGS): Removed.
22324 * config/cr16/cr16.c (TARGET_PUSH_ARGUMENT): New.
22325 * config/cr16/cr16.h (PUSH_ARGS): Removed.
22326 * config/i386/i386.c (ix86_push_argument): New.
22327 (TARGET_PUSH_ARGUMENT): Likewise.
22328 * config/i386/i386.h (PUSH_ARGS): Removed.
22329 * config/m32c/m32c.c (TARGET_PUSH_ARGUMENT): New.
22330 * config/m32c/m32c.h (PUSH_ARGS): Removed.
22331 * config/nios2/nios2.h (PUSH_ARGS): Likewise.
22332 * config/pru/pru.h (PUSH_ARGS): Likewise.
22333 * doc/tm.texi.in: Remove PUSH_ARGS documentation. Add
22334 TARGET_PUSH_ARGUMENT hook.
22335 * doc/tm.texi: Regenerated.
22337 2021-06-17 Uroš Bizjak <ubizjak@gmail.com>
22340 * config/i386/i386-expand.c (expand_vector_set_var):
22341 Handle V2FS mode remapping. Pass TARGET_MMX_WITH_SSE to
22342 ix86_expand_vector_init_duplicate.
22343 (ix86_expand_vector_init_duplicate): Emit insv_1 for
22344 QImode for !TARGET_PARTIAL_REG_STALL.
22345 * config/i386/predicates.md (vec_setm_mmx_operand): New predicate.
22346 * config/i386/mmx.md (vec_setv2sf): Use vec_setm_mmx_operand
22347 as operand 2 predicate. Call ix86_expand_vector_set_var
22348 for non-constant index operand.
22349 (vec_setv2si): Ditto.
22350 (vec_setv4hi): Ditto.
22351 (vec_setv8qi): ditto.
22353 2021-06-17 Aldy Hernandez <aldyh@redhat.com>
22355 PR tree-optimization/100790
22356 * gimple-range.cc (range_of_builtin_call): Cleanup clz and ctz
22359 2021-06-17 Martin Liska <mliska@suse.cz>
22361 * doc/invoke.texi: Use consistently -O1 instead of -O.
22363 2021-06-17 Martin Liska <mliska@suse.cz>
22365 * gcov-io.h: Update documentation entry about string format.
22367 2021-06-17 Marius Hillenbrand <mhillen@linux.ibm.com>
22370 * config/s390/vecintrin.h (vec_doublee): Fix to use
22371 __builtin_s390_vflls.
22372 (vec_floate): Fix to use __builtin_s390_vflrd.
22374 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
22376 * dominance.c (get_dominated_to_depth): Return auto_vec<basic_block>.
22377 * dominance.h (get_dominated_to_depth): Likewise.
22378 (get_all_dominated_blocks): Likewise.
22379 * cfgcleanup.c (delete_unreachable_blocks): Adjust.
22380 * gcse.c (hoist_code): Likewise.
22381 * tree-cfg.c (remove_edge_and_dominated_blocks): Likewise.
22382 * tree-parloops.c (oacc_entry_exit_ok): Likewise.
22383 * tree-ssa-dce.c (eliminate_unnecessary_stmts): Likewise.
22384 * tree-ssa-phiprop.c (pass_phiprop::execute): Likewise.
22386 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
22388 * dominance.c (get_dominated_by_region): Return auto_vec<basic_block>.
22389 * dominance.h (get_dominated_by_region): Likewise.
22390 * tree-cfg.c (gimple_duplicate_sese_region): Adjust.
22391 (gimple_duplicate_sese_tail): Likewise.
22392 (move_sese_region_to_fn): Likewise.
22394 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
22396 * dominance.c (get_dominated_by): Return auto_vec<basic_block>.
22397 * dominance.h (get_dominated_by): Likewise.
22398 * auto-profile.c (afdo_find_equiv_class): Adjust.
22399 * cfgloopmanip.c (duplicate_loop_to_header_edge): Likewise.
22400 * loop-unroll.c (unroll_loop_runtime_iterations): Likewise.
22401 * tree-cfg.c (test_linear_chain): Likewise.
22402 (test_diamond): Likewise.
22404 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
22406 * cfgloop.h (get_loop_hot_path): Return auto_vec<basic_block>.
22407 * cfgloopanal.c (get_loop_hot_path): Likewise.
22408 * tree-ssa-loop-ivcanon.c (tree_estimate_loop_size): Likewise.
22410 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
22412 * cgraph.c (cgraph_node::collect_callers): Return
22413 auto_vec<cgraph_edge *>.
22414 * cgraph.h (cgraph_node::collect_callers): Likewise.
22415 * ipa-cp.c (create_specialized_node): Adjust.
22416 (decide_about_value): Likewise.
22417 (decide_whether_version_node): Likewise.
22418 * ipa-sra.c (process_isra_node_results): Likewise.
22420 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
22422 * vec.h (vl_ptr>::using_auto_storage): Handle null m_vec.
22423 (auto_vec<T, 0>::auto_vec): Define move constructor, and delete copy
22425 (auto_vec<T, 0>::operator=): Define move assignment and delete copy
22428 2021-06-17 Aldy Hernandez <aldyh@redhat.com>
22430 * gimple-range.cc (debug_seed_ranger): New.
22431 (dump_ranger): New.
22432 (debug_ranger): New.
22434 2021-06-17 Richard Biener <rguenther@suse.de>
22436 PR tree-optimization/54400
22437 * tree-vectorizer.h (enum slp_instance_kind): Add
22438 slp_inst_kind_bb_reduc.
22439 (reduction_fn_for_scalar_code): Declare.
22440 * tree-vect-data-refs.c (vect_slp_analyze_instance_dependence):
22441 Check SLP_INSTANCE_KIND instead of looking at the
22443 (vect_slp_analyze_instance_alignment): Likewise.
22444 * tree-vect-loop.c (reduction_fn_for_scalar_code): Export.
22445 * tree-vect-slp.c (vect_slp_linearize_chain): Split out
22446 chain linearization from vect_build_slp_tree_2 and generalize
22447 for the use of BB reduction vectorization.
22448 (vect_build_slp_tree_2): Adjust accordingly.
22449 (vect_optimize_slp): Elide permutes at the root of BB reduction
22451 (vectorizable_bb_reduc_epilogue): New function.
22452 (vect_slp_prune_covered_roots): Likewise.
22453 (vect_slp_analyze_operations): Use them.
22454 (vect_slp_check_for_constructors): Recognize associatable
22455 chains for BB reduction vectorization.
22456 (vectorize_slp_instance_root_stmt): Generate code for the
22457 BB reduction epilogue.
22459 2021-06-17 Andrew MacLeod <amacleod@redhat.com>
22461 * gimple-range-gori.cc (gori_compute::has_edge_range_p): Check with
22463 (gori_compute::may_recompute_p): New.
22464 (gori_compute::outgoing_edge_range_p): Perform recomputations.
22465 * gimple-range-gori.h (class gori_compute): Add prototype.
22467 2021-06-17 Andrew MacLeod <amacleod@redhat.com>
22469 * gimple-range-cache.cc (ranger_cache::range_on_edge): Always return
22470 true when a range can be calculated.
22471 * gimple-range.cc (gimple_ranger::dump_bb): Check has_edge_range_p.
22473 2021-06-16 Martin Sebor <msebor@redhat.com>
22475 * doc/invoke.texi (-Wmismatched-dealloc, -Wmismatched-new-delete):
22476 Correct documented defaults.
22478 2021-06-16 Andrew MacLeod <amacleod@redhat.com>
22480 * gimple-range-cache.cc (ranger_cache::ranger_cache): Initialize
22481 m_new_value_p directly.
22483 2021-06-16 Uroš Bizjak <ubizjak@gmail.com>
22486 * config/i386/i386-expand.c (expand_vec_perm_2perm_pblendv):
22487 Handle 64bit modes for TARGET_SSE4_1.
22488 (expand_vec_perm_pshufb2): Handle 64bit modes for TARGET_SSSE3.
22489 (expand_vec_perm_even_odd_pack): Handle V4HI mode.
22490 (expand_vec_perm_even_odd_1) <case E_V4HImode>: Expand via
22491 expand_vec_perm_pshufb2 for TARGET_SSSE3 and via
22492 expand_vec_perm_even_odd_pack for TARGET_SSE4_1.
22493 * config/i386/mmx.md (mmx_packusdw): New insn pattern.
22495 2021-06-16 Jonathan Wright <jonathan.wright@arm.com>
22497 * config/aarch64/aarch64-simd.md (aarch64_<sur><addsub>hn<mode>):
22498 Change to an expander that emits the correct instruction
22499 depending on endianness.
22500 (aarch64_<sur><addsub>hn<mode>_insn_le): Define.
22501 (aarch64_<sur><addsub>hn<mode>_insn_be): Define.
22503 2021-06-16 Jonathan Wright <jonathan.wright@arm.com>
22505 * config/aarch64/aarch64-simd-builtins.def: Split generator
22506 for aarch64_<su>qmovn builtins into scalar and vector
22508 * config/aarch64/aarch64-simd.md (aarch64_<su>qmovn<mode>_insn_le):
22510 (aarch64_<su>qmovn<mode>_insn_be): Define.
22511 (aarch64_<su>qmovn<mode>): Split into scalar and vector
22512 variants. Change vector variant to an expander that emits the
22513 correct instruction depending on endianness.
22515 2021-06-16 Jonathan Wright <jonathan.wright@arm.com>
22517 * config/aarch64/aarch64-simd-builtins.def: Split generator
22518 for aarch64_sqmovun builtins into scalar and vector variants.
22519 * config/aarch64/aarch64-simd.md (aarch64_sqmovun<mode>):
22520 Split into scalar and vector variants. Change vector variant
22521 to an expander that emits the correct instruction depending
22523 (aarch64_sqmovun<mode>_insn_le): Define.
22524 (aarch64_sqmovun<mode>_insn_be): Define.
22526 2021-06-16 Jonathan Wright <jonathan.wright@arm.com>
22528 * config/aarch64/aarch64-simd.md (aarch64_xtn<mode>_insn_le):
22529 Define - modeling zero-high-half semantics.
22530 (aarch64_xtn<mode>): Change to an expander that emits the
22531 appropriate instruction depending on endianness.
22532 (aarch64_xtn<mode>_insn_be): Define - modeling zero-high-half
22534 (aarch64_xtn2<mode>_le): Rename to...
22535 (aarch64_xtn2<mode>_insn_le): This.
22536 (aarch64_xtn2<mode>_be): Rename to...
22537 (aarch64_xtn2<mode>_insn_be): This.
22538 (vec_pack_trunc_<mode>): Emit truncation instruction instead
22540 * config/aarch64/iterators.md (Vnarrowd): Add Vnarrowd mode
22541 attribute iterator.
22543 2021-06-16 Martin Jambor <mjambor@suse.cz>
22545 PR tree-optimization/100453
22546 * tree-sra.c (create_access): Disqualify any const candidates
22547 which are written to.
22548 (sra_modify_expr): Do not store sub-replacements back to a const base.
22549 (handle_unscalarized_data_in_subtree): Likewise.
22550 (sra_modify_assign): Likewise. Earlier, use TREE_READONLy test
22551 instead of constant_decl_p.
22553 2021-06-16 Jakub Jelinek <jakub@redhat.com>
22555 PR middle-end/101062
22556 * stor-layout.c (finish_bitfield_representative): For fields in unions
22557 assume nextf is always NULL.
22558 (finish_bitfield_layout): Compute bit field representatives also in
22559 unions, but handle it as if each bitfield was the only field in the
22562 2021-06-16 Richard Biener <rguenther@suse.de>
22564 PR tree-optimization/101088
22565 * tree-ssa-loop-im.c (sm_seq_valid_bb): Only look for
22566 supported refs on edges. Do not assert same ref but
22567 different kind stores are unsuported but mark them so.
22568 (hoist_memory_references): Only look for supported refs
22571 2021-06-16 Roger Sayle <roger@nextmovesoftware.com>
22573 PR rtl-optimization/46235
22574 * config/i386/i386.md: New define_split for bt followed by cmov.
22575 (*bt<mode>_setcqi): New define_insn_and_split for bt followed by setc.
22576 (*bt<mode>_setncqi): New define_insn_and_split for bt then setnc.
22577 (*bt<mode>_setnc<mode>): New define_insn_and_split for bt followed
22578 by setnc with zero extension.
22580 2021-06-16 Richard Biener <rguenther@suse.de>
22582 PR tree-optimization/101083
22583 * tree-vect-slp.c (vect_slp_build_two_operator_nodes): Get
22584 vectype as argument.
22585 (vect_build_slp_tree_2): Adjust.
22587 2021-06-15 Martin Sebor <msebor@redhat.com>
22589 PR middle-end/100876
22590 * builtins.c: (gimple_call_return_array): Account for size_t
22591 mangling as either unsigned int or unsigned long
22593 2021-06-15 Jeff Law <jeffreyalaw@gmail.com>
22595 * compare-elim.c (try_eliminate_compare): Run DCE to clean things
22596 up before eliminating comparisons.
22598 2021-06-15 Aldy Hernandez <aldyh@redhat.com>
22600 * range-op.cc (operator_bitwise_or::wi_fold): Make sure
22601 nonzero|X is nonzero.
22602 (range_op_bitwise_and_tests): Add tests for above.
22604 2021-06-15 Carl Love <cel@us.ibm.com>
22607 * config/rs6000/rs6000-builtin.def (VCMPEQUT): Fix the ICODE for the
22609 (VRLQ, VSLQ, VSRQ, VSRAQ): Remove unused BU_P10_OVERLOAD_2
22612 2021-06-15 Tobias Burnus <tobias@codesourcery.com>
22615 * gimplify.c (enum gimplify_defaultmap_kind): Add GDMK_SCALAR_TARGET.
22616 (struct gimplify_omp_ctx): Extend defaultmap array by one.
22617 (new_omp_context): Init defaultmap[GDMK_SCALAR_TARGET].
22618 (omp_notice_variable): Update type classification for Fortran.
22619 (gimplify_scan_omp_clauses): Update calls for new argument; handle
22620 GDMK_SCALAR_TARGET; for Fortran, GDMK_POINTER avoid GOVD_MAP_0LEN_ARRAY.
22621 * langhooks-def.h (lhd_omp_scalar_p): Add 'ptr_ok' argument.
22622 * langhooks.c (lhd_omp_scalar_p): Likewise.
22623 (LANG_HOOKS_OMP_ALLOCATABLE_P, LANG_HOOKS_OMP_SCALAR_TARGET_P): New.
22624 (LANG_HOOKS_DECLS): Add them.
22625 * langhooks.h (struct lang_hooks_for_decls): Add new hooks, update
22626 omp_scalar_p pointer type to include the new bool argument.
22628 2021-06-15 David Malcolm <dmalcolm@redhat.com>
22630 * doc/analyzer.texi
22631 (Special Functions for Debugging the Analyzer): Add
22632 __analyzer_dump_capacity.
22634 2021-06-15 Jakub Jelinek <jakub@redhat.com>
22637 * expr.c (expand_expr_real_2) <case VEC_PACK_FIX_TRUNC_EXPR,
22638 case VEC_PACK_TRUNC_EXPR>: Clear subtarget when changing mode.
22640 2021-06-15 Richard Biener <rguenther@suse.de>
22642 * cfgloopanal.c (mark_irreducible_loops): Use a dominance
22643 check to identify loop latches.
22644 * cfgloop.c (verify_loop_structure): Likewise.
22645 * loop-init.c (apply_loop_flags): Allow marked irreducible
22646 regions even with multiple latches.
22647 * predict.c (rebuild_frequencies): Simplify.
22649 2021-06-15 Richard Biener <rguenther@suse.de>
22651 * tree-ssa-threadupdate.c
22652 (jump_thread_path_registry::mark_threaded_blocks): Assert we
22653 have marked irreducible regions.
22655 2021-06-14 Martin Sebor <msebor@redhat.com>
22658 * builtins.c (gimple_call_return_array): Check for attribute fn spec.
22659 Handle calls to placement new.
22660 (ndecl_dealloc_argno): Avoid placement delete.
22662 2021-06-14 Peter Bergner <bergner@linux.ibm.com>
22665 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_mma_builtin): Use
22666 create_tmp_reg_or_ssa_name().
22668 2021-06-14 Andrew MacLeod <amacleod@redhat.com>
22670 * gimple-range-cache.cc (ranger_cache::ranger_cache): Adjust.
22671 (ranger_cache::enable_new_values): Set to specified value and
22672 return the old value.
22673 (ranger_cache::disable_new_values): Delete.
22674 (ranger_cache::fill_block_cache): Disable non 1st order derived
22676 * gimple-range-cache.h (ranger_cache): Adjust prototypes.
22677 * gimple-range.cc (gimple_ranger::range_of_expr): Adjust.
22679 2021-06-14 Uroš Bizjak <ubizjak@gmail.com>
22682 * config/i386/i386-expand.c (ix86_vectorize_vec_perm_const):
22683 Return true early when testing with V2HImode.
22684 * config/i386/mmx.md (*punpckwd): Split to sse2_pshuflw_1.
22686 2021-06-14 Christophe Lyon <christophe.lyon@linaro.org>
22688 * config/arm/mve.md (mve_vec_unpack<US>_lo_<mode>): New pattern.
22689 (mve_vec_unpack<US>_hi_<mode>): New pattern.
22690 (@mve_vec_pack_trunc_lo_<mode>): New pattern.
22691 (mve_vmovntq_<supf><mode>): Prefix with '@'.
22692 * config/arm/neon.md (vec_unpack<US>_hi_<mode>): Move to
22694 (vec_unpack<US>_lo_<mode>): Likewise.
22695 (vec_pack_trunc_<mode>): Rename to
22696 neon_quad_vec_pack_trunc_<mode>.
22697 * config/arm/vec-common.md (vec_unpack<US>_hi_<mode>): New
22699 (vec_unpack<US>_lo_<mode>): New.
22700 (vec_pack_trunc_<mode>): New.
22702 2021-06-14 Richard Biener <rguenther@suse.de>
22704 PR tree-optimization/100934
22705 * tree-ssa-dom.c (pass_dominator::execute): Properly
22706 mark irreducible regions.
22708 2021-06-14 Martin Liska <mliska@suse.cz>
22710 * doc/invoke.texi: Put r{...} on the same line as @item.
22712 2021-06-14 Martin Liska <mliska@suse.cz>
22714 * doc/invoke.texi: Add missing newline.
22716 2021-06-14 Martin Liska <mliska@suse.cz>
22718 * doc/invoke.texi: Remove '+' charasters.
22720 2021-06-14 Claudiu Zissulescu <claziss@synopsys.com>
22722 * config.gcc (arc): Add support for with_cpu option.
22723 * config/arc/arc.h (OPTION_DEFAULT_SPECS): Add fpu.
22725 2021-06-14 Richard Biener <rguenther@suse.de>
22727 PR tree-optimization/101031
22728 * tree-ssa-strlen.c (maybe_invalidate): Increment max_size
22729 instead of size when accounting for a possibly string
22732 2021-06-14 Martin Liska <mliska@suse.cz>
22734 * gimple-ssa-evrp.c (pointer_equiv_analyzer::~pointer_equiv_analyzer): Use delete[].
22736 2021-06-14 Aldy Hernandez <aldyh@redhat.com>
22738 * value-query.cc (gimple_range_global): Call get_range_global
22739 if called after inlining.
22741 2021-06-13 Uroš Bizjak <ubizjak@gmail.com>
22744 * config/i386/i386-expand.c (expand_vec_perm_pshufb):
22745 Emit constant permutation insn directly from here.
22747 2021-06-13 Trevor Saunders <tbsaunde@tbsaunde.org>
22749 * attribs.c (find_attribute_namespace): Iterate over vec<> with
22751 * auto-profile.c (afdo_find_equiv_class): Likewise.
22752 * gcc.c (do_specs_vec): Likewise.
22753 (do_spec_1): Likewise.
22754 (driver::set_up_specs): Likewise.
22755 * gimple-loop-jam.c (any_access_function_variant_p): Likewise.
22756 * gimple-ssa-store-merging.c (compatible_load_p): Likewise.
22757 (imm_store_chain_info::try_coalesce_bswap): Likewise.
22758 (imm_store_chain_info::coalesce_immediate_stores): Likewise.
22759 (get_location_for_stmts): Likewise.
22760 * graphite-poly.c (print_iteration_domains): Likewise.
22761 (free_poly_bb): Likewise.
22762 (remove_gbbs_in_scop): Likewise.
22763 (free_scop): Likewise.
22764 (dump_gbb_cases): Likewise.
22765 (dump_gbb_conditions): Likewise.
22766 (print_pdrs): Likewise.
22767 (print_scop): Likewise.
22768 * ifcvt.c (cond_move_process_if_block): Likewise.
22769 * lower-subreg.c (decompose_multiword_subregs): Likewise.
22770 * regcprop.c (pass_cprop_hardreg::execute): Likewise.
22771 * sanopt.c (sanitize_rewrite_addressable_params): Likewise.
22772 * sel-sched-dump.c (dump_insn_vector): Likewise.
22773 * store-motion.c (store_ops_ok): Likewise.
22774 (store_killed_in_insn): Likewise.
22775 * timevar.c (timer::named_items::print): Likewise.
22776 * tree-cfgcleanup.c (cleanup_control_flow_pre): Likewise.
22777 (cleanup_tree_cfg_noloop): Likewise.
22778 * tree-data-ref.c (dump_data_references): Likewise.
22779 (print_dir_vectors): Likewise.
22780 (print_dist_vectors): Likewise.
22781 (dump_data_dependence_relations): Likewise.
22782 (dump_dist_dir_vectors): Likewise.
22783 (dump_ddrs): Likewise.
22784 (create_runtime_alias_checks): Likewise.
22785 (free_subscripts): Likewise.
22786 (save_dist_v): Likewise.
22787 (save_dir_v): Likewise.
22788 (invariant_access_functions): Likewise.
22789 (same_access_functions): Likewise.
22790 (access_functions_are_affine_or_constant_p): Likewise.
22791 (find_data_references_in_stmt): Likewise.
22792 (graphite_find_data_references_in_stmt): Likewise.
22793 (free_dependence_relations): Likewise.
22794 (free_data_refs): Likewise.
22795 * tree-inline.c (copy_debug_stmts): Likewise.
22796 * tree-into-ssa.c (dump_currdefs): Likewise.
22797 (rewrite_update_phi_arguments): Likewise.
22798 * tree-ssa-propagate.c (clean_up_loop_closed_phi): Likewise.
22799 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
22801 (vect_slp_analyze_node_dependences): Likewise.
22802 (vect_slp_analyze_instance_dependence): Likewise.
22803 (vect_record_base_alignments): Likewise.
22804 (vect_get_peeling_costs_all_drs): Likewise.
22805 (vect_peeling_supportable): Likewise.
22806 * tree-vectorizer.c (vec_info::~vec_info): Likewise.
22807 (vec_info::free_stmt_vec_infos): Likewise.
22809 2021-06-13 Jeff Law <jeffreyalaw@gmail.com>
22811 * config/h8300/logical.md (<code>qi3_1<cczn>): New pattern.
22812 (andqi3_1<cczn>): Removed.
22813 (<ors>qi3_1): Do not split for IOR/XOR a single bit.
22814 (H8/SX bit logicals): Split out from other patterns.
22815 * config/h8300/multiply.md (mulqihi3_const<cczn>): Renamed from
22816 mulqihi3_const_clobber_flags.
22817 (mulqihi3<cczn>, mulhisi3_const<cczn>, mulhisi3<cczn>): Similarly
22819 2021-06-13 H.J. Lu <hjl.tools@gmail.com>
22822 * config/i386/i386.c (ix86_expand_prologue): Set red_zone_used
22823 to true if red zone is used.
22824 (ix86_output_indirect_jmp): Replace ix86_red_zone_size with
22825 ix86_red_zone_used.
22826 * config/i386/i386.h (machine_function): Add red_zone_used.
22827 (ix86_red_zone_size): Removed.
22828 (ix86_red_zone_used): New.
22829 * config/i386/i386.md (peephole2 patterns): Replace
22830 ix86_red_zone_size with ix86_red_zone_used.
22832 2021-06-12 Jason Merrill <jason@redhat.com>
22834 * doc/extend.texi (unused variable attribute): Applies to
22835 structure fields as well.
22837 2021-06-12 Eugene Rozenfeld <erozen@microsoft.com>
22839 * auto-profile.c (read_profile): fix a typo in an error string
22841 2021-06-11 Thomas Schwinge <thomas@codesourcery.com>
22843 * tree-pretty-print.h (dump_omp_clauses): Add 'bool = true'
22845 * tree-pretty-print.c (dump_omp_clauses): Update.
22846 (dump_generic_node) <OMP_CLAUSE>: Use it.
22848 2021-06-11 Srinath Parvathaneni <srinath.parvathaneni@arm.com>
22851 * config/arm/arm_mve.h (__arm_vld1q): Change __ARM_mve_coerce(p0,
22852 int8_t const *) to __ARM_mve_coerce1(p0, int8_t *) in the argument for
22853 the polymorphic variants matching code.
22854 (__arm_vld1q_z): Likewise.
22855 (__arm_vld2q): Likewise.
22856 (__arm_vld4q): Likewise.
22857 (__arm_vldrbq_gather_offset): Likewise.
22858 (__arm_vldrbq_gather_offset_z): Likewise.
22860 2021-06-11 Roger Sayle <roger@nextmovesoftware.com>
22862 PR tree-optimization/96392
22863 * fold-const.h (tree_expr_maybe_real_minus_zero_p): Fix prototype.
22865 2021-06-11 Roger Sayle <roger@nextmovesoftware.com>
22867 PR tree-optimization/96392
22868 * fold-const.c (fold_real_zero_addition_p): Take both arguments
22869 of the addition or subtraction, not just the zero. Use this
22870 other argument in tests for signaling NaNs and signed zeros.
22871 (tree_expr_maybe_real_minus_zero_p): New predicate.
22872 * fold-const.h (fold_real_zero_addition_p): Update prototype.
22873 (tree_expr_maybe_real_minus_zero_p): New function prototype.
22874 * match.pd: Update calls to fold_real_zero_addition_p.
22875 Replace HONOR_NANS with tree_expr_maybe_nan_p.
22876 Replace HONOR_SIGNED_ZEROS with tree_expr_maybe_real_minus_zero_p.
22877 Replace HONOR_SNANS with tree_expr_maybe_signaling_nan_p.
22878 * tree-ssa-reassoc.c (eliminate_using_constants): Update
22879 call to fold_real_zero_addition_p.
22881 2021-06-11 Richard Biener <rguenther@suse.de>
22883 PR tree-optimization/101025
22884 * tree-ssa-loop-im.c (sm_seq_valid_bb): Make sure to process
22885 all refs that require dependence checking.
22887 2021-06-11 Richard Biener <rguenther@suse.de>
22889 PR tree-optimization/101028
22890 * tree-vect-slp.c (vect_build_slp_tree_2): When SLP
22891 reassoc discovery fails fatally, mark appropriate lanes
22894 2021-06-11 Richard Biener <rguenther@suse.de>
22896 PR tree-optimization/101026
22897 * tree-vect-slp.c (vect_build_slp_tree_2): Make sure we
22898 have a representative for the associated chain nodes.
22900 2021-06-11 Jakub Jelinek <jakub@redhat.com>
22902 PR rtl-optimization/101008
22903 * simplify-rtx.c (relational_result): New function.
22904 (simplify_logical_relational_operation,
22905 simplify_relational_operation): Use it.
22907 2021-06-11 Jakub Jelinek <jakub@redhat.com>
22910 * config/i386/sse.md (*vec_concat<mode>_0_1): Require TARGET_SSE2.
22912 2021-06-11 Uroš Bizjak <ubizjak@gmail.com>
22915 * config/i386/i386-expand.c (expand_vec_perm_pshufb): Return
22916 false if the permutation can be implemented with constant
22917 permutation instruction in wider mode.
22918 (canonicalize_vector_int_perm): Move above expand_vec_perm_pshufb.
22919 Handle V8QImode and V4HImode.
22921 2021-06-11 Martin Liska <mliska@suse.cz>
22923 PR gcov-profile/100788
22924 * common.opt: Add new option.
22925 * coverage.c (coverage_begin_function): Emit warning instead on
22926 the internal compiler error.
22927 * doc/invoke.texi: Document the option.
22928 * toplev.c (process_options): Enable it by default.
22930 2021-06-11 Richard Biener <rguenther@suse.de>
22932 PR middle-end/101009
22933 * tree-data-ref.c (build_classic_dist_vector_1): Make sure
22934 to set *init_b to true when we encounter a constant equal
22936 (compute_affine_dependence): Also dump the actual DR_REF.
22938 2021-06-10 Aldy Hernandez <aldyh@redhat.com>
22940 PR tree-optimization/100984
22941 * gimple-ssa-evrp.c (ssa_equiv_stack): Use auto_vec for
22942 replacements table.
22943 (ssa_equiv_stack::~ssa_equiv_stack): Remove.
22945 2021-06-11 Kewen Lin <linkw@linux.ibm.com>
22947 * config/rs6000/rs6000.md
22948 (floatsi<SFDF:mode>2_lfiwax_<QHI:mode>_mem_zext): New
22949 define_insn_and_split.
22951 2021-06-11 Richard Biener <rguenther@suse.de>
22953 * tree-vect-slp.c (vect_build_slp_tree_2): Use stablesort
22954 to sort operands of the associative chain.
22956 2021-06-11 Richard Biener <rguenther@suse.de>
22958 * system.h (gcc_stablesort_r): Declare.
22959 * sort.cc (gcc_sort_r): Support stable sort.
22960 (gcc_stablesort_r): Define.
22961 * vec.h (vec<>::stablesort): Add.
22963 2021-06-10 Uroš Bizjak <ubizjak@gmail.com>
22966 * config/i386/i386-expand.c (ix86_split_mmx_punpck):
22967 Handle V2SF mode. Emit SHUFPS to fixup unpack-high for V2SF mode.
22968 (expand_vec_perm_blend): Handle 64bit modes for TARGET_SSE4_1.
22969 (expand_vec_perm_pshufb): Handle 64bit modes for TARGET_SSSE3.
22970 (expand_vec_perm_pblendv): Handle 64bit modes for TARGET_SSE4_1.
22971 (expand_vec_perm_interleave2): Handle 64bit modes.
22972 (expand_vec_perm_even_odd_pack): Handle V8QI mode.
22973 (expand_vec_perm_even_odd_1): Ditto.
22974 (ix86_vectorize_vec_perm_const): Ditto.
22975 * config/i386/i386.md (UNSPEC_PSHUFB): Move from ...
22976 * config/i386/sse.md: ... here.
22977 * config/i386/mmx.md (*vec_interleave_lowv2sf):
22978 New insn_and_split pattern.
22979 (*vec_interleave_highv2sf): Ditto.
22980 (mmx_pshufbv8qi3): New insn pattern.
22981 (*mmx_pblendw): Ditto.
22983 2021-06-10 Peter Bergner <bergner@linux.ibm.com>
22985 * config/rs6000/rs6000-builtin.def (build_pair): New built-in.
22986 (build_acc): Likewise.
22987 * config/rs6000/rs6000-call.c (mma_expand_builtin): Swap assemble
22988 source operands in little-endian mode.
22989 (rs6000_gimple_fold_mma_builtin): Handle VSX_BUILTIN_BUILD_PAIR.
22990 (mma_init_builtins): Likewise.
22991 * config/rs6000/rs6000.c (rs6000_split_multireg_move): Handle endianness
22992 ordering for the MMA assemble and build source operands.
22993 * doc/extend.texi (__builtin_vsx_build_acc, __builtin_mma_build_pair):
22995 (__builtin_mma_assemble_acc, __builtin_mma_assemble_pair): Remove
22998 2021-06-10 Jeff Law <jeffreyalaw@gmail.com>
23000 * config/h8300/h8300.c (select_cc_mode): Handle MEM. Use
23002 * config/h8300/extensions.md: Replace _clobber_flags patterns
23005 2021-06-10 Robin Dapp <rdapp@linux.ibm.com>
23007 * config/s390/vector.md (vcond_mask_<mode><mode>): Change to
23008 (vcond_mask_<mode><tointvec>): this.
23010 2021-06-10 Andrew Stubbs <ams@codesourcery.com>
23011 Thomas Schwinge <thomas@codesourcery.com>
23013 * omp-builtins.def (BUILT_IN_GOACC_ENTER_EXIT_DATA): Split into...
23014 (BUILT_IN_GOACC_ENTER_DATA, BUILT_IN_GOACC_EXIT_DATA): ... these.
23015 * gimple.h (enum gf_mask): Split
23016 'GF_OMP_TARGET_KIND_OACC_ENTER_EXIT_DATA' into
23017 'GF_OMP_TARGET_KIND_OACC_ENTER_DATA' and
23018 'GF_OMP_TARGET_KIND_OACC_EXIT_DATA'.
23019 (is_gimple_omp_oacc): Update.
23020 * gimple-pretty-print.c (dump_gimple_omp_target): Likewise.
23021 * gimplify.c (gimplify_omp_target_update): Likewise.
23022 * omp-expand.c (expand_omp_target, build_omp_regions_1)
23023 (omp_make_gimple_edges): Likewise.
23024 * omp-low.c (check_omp_nesting_restrictions, lower_omp_target):
23027 2021-06-10 Aldy Hernandez <aldyh@redhat.com>
23029 * value-query.cc (value_query::value_on_edge): Rename name to
23031 (range_query::range_on_edge): Same.
23032 (range_query::value_of_expr): Same.
23033 (range_query::value_on_edge): Same.
23034 * value-query.h (class value_query): Same.
23035 (class range_query): Same.
23037 2021-06-10 Richard Biener <rguenther@suse.de>
23039 PR tree-optimization/101003
23040 * tree-vect-slp.c (vect_build_slp_tree_2): Appropriately
23041 use the pattern stmt defs when linearizing a chain.
23043 2021-06-10 Jakub Jelinek <jakub@redhat.com>
23046 * ifcvt.c (noce_get_alt_condition, noce_try_abs): Use
23047 prev_nonnote_nondebug_insn instead of prev_nonnote_insn.
23049 2021-06-10 Clement Chigot <clement.chigot@atos.net>
23051 * config/rs6000/aix71.h (ASM_CPU_SPEC): Add Power10 directive.
23052 * config/rs6000/aix72.h (ASM_CPU_SPEC): Likewise.
23054 2021-06-09 Andrew Pinski <apinski@marvell.com>
23056 PR tree-optimization/100925
23057 * match.pd (a ? CST1 : CST2): Limit transformations
23058 that would produce a negative to integeral types only.
23059 Change !POINTER_TYPE_P to INTEGRAL_TYPE_P also.
23061 2021-06-09 Jeff Law <jeffreyalaw@gmail.com>
23064 2021-06-09 Jeff Law <jeffreyalaw@gmail.com>
23066 * doc/tm.texi: Correctly update.
23068 2021-06-09 Jeff Law <jeffreyalaw@gmail.com>
23070 * doc/tm.texi: Correctly update.
23072 2021-06-09 H.J. Lu <hjl.tools@gmail.com>
23075 * doc/tm.texi.in (Trampolines): Add a missing blank line.
23077 2021-06-09 Paul Eggert <eggert@cs.ucla.edu>
23080 * doc/invoke.texi (Code Gen Options); Document that -fno-trampolines
23081 and -ftrampolines work only with Ada.
23082 * doc/tm.texi.in (Trampolines): Likewise.
23083 * doc/tm.texi: Regenerated.
23085 2021-06-09 Carl Love <cel@us.ibm.com>
23087 * config/rs6000/altivec.h (vec_signextll, vec_signexti, vec_signextq):
23088 Add define for new builtins.
23089 * config/rs6000/altivec.md(altivec_vreveti2): Add define_expand.
23090 * config/rs6000/rs6000-builtin.def (VSIGNEXTI, VSIGNEXTLL): Add
23091 overloaded builtin definitions.
23092 (VSIGNEXTSB2W, VSIGNEXTSH2W, VSIGNEXTSB2D, VSIGNEXTSH2D,VSIGNEXTSW2D,
23093 VSIGNEXTSD2Q): Add builtin expansions.
23094 (SIGNEXT): Add P10 overload definition.
23095 * config/rs6000/rs6000-call.c (P9V_BUILTIN_VEC_VSIGNEXTI, P9V_BUILTIN_VEC_VSIGNEXTLL,
23096 P10_BUILTIN_VEC_SIGNEXT): Add overloaded argument definitions.
23097 * config/rs6000/vsx.md (vsx_sign_extend_v2di_v1ti): Add define_insn.
23098 (vsignextend_v2di_v1ti, vsignextend_qi_<mode>, vsignextend_hi_<mode>,
23099 vsignextend_si_v2di)[VIlong]: Add define_expand.
23100 Make define_insn vsx_sign_extend_si_v2di visible.
23101 * doc/extend.texi: Add documentation for the vec_signexti,
23102 vec_signextll builtins and vec_signextq.
23104 2021-06-09 Carl Love <cel@us.ibm.com>
23106 * config/rs6000/rs6000.c (__fixkfti, __fixunskfti, __floattikf,
23107 __floatuntikf): Names changed to __fixkfti_sw, __fixunskfti_sw,
23108 __floattikf_sw, __floatuntikf_sw respectively.
23109 * config/rs6000/rs6000.md (floatti<mode>2, floatunsti<mode>2,
23110 fix_trunc<mode>ti2, fixuns_trunc<mode>ti2): Add
23111 define_insn for mode IEEE 128.
23113 2021-06-09 Carl Love <cel@us.ibm.com>
23115 * config/rs6000/altivec.md (altivec_vslq, altivec_vsrq):
23116 Rename to altivec_vslq_<mode>, altivec_vsrq_<mode>, mode VEC_TI.
23117 * config/rs6000/vector.md (VEC_TI): Was named VSX_TI in vsx.md.
23118 (vashlv1ti3): Change to vashl<mode>3, mode VEC_TI.
23119 (vlshrv1ti3): Change to vlshr<mode>3, mode VEC_TI.
23120 * config/rs6000/vsx.md (VSX_TI): Remove define_mode_iterator. Update
23121 uses of VSX_TI to VEC_TI.
23123 2021-06-09 Carl Love <cel@us.ibm.com>
23125 * config/rs6000/dfp.md (floattitd2, fixtdti2): New define_insns.
23127 2021-06-09 Carl Love <cel@us.ibm.com>
23129 * config/rs6000/altivec.h (vec_dive, vec_mod): Add define for new
23131 * config/rs6000/altivec.md (UNSPEC_VMULEUD, UNSPEC_VMULESD,
23132 UNSPEC_VMULOUD, UNSPEC_VMULOSD): New unspecs.
23133 (altivec_eqv1ti, altivec_gtv1ti, altivec_gtuv1ti, altivec_vmuleud,
23134 altivec_vmuloud, altivec_vmulesd, altivec_vmulosd, altivec_vrlq,
23135 altivec_vrlqmi, altivec_vrlqmi_inst, altivec_vrlqnm,
23136 altivec_vrlqnm_inst, altivec_vslq, altivec_vsrq, altivec_vsraq,
23137 altivec_vcmpequt_p, altivec_vcmpgtst_p, altivec_vcmpgtut_p): New
23139 (vec_widen_umult_even_v2di, vec_widen_smult_even_v2di,
23140 vec_widen_umult_odd_v2di, vec_widen_smult_odd_v2di, altivec_vrlqmi,
23141 altivec_vrlqnm): New define_expands.
23142 * config/rs6000/rs6000-builtin.def (VCMPEQUT_P, VCMPGTST_P,
23143 VCMPGTUT_P): Add macro expansions.
23144 (BU_P10V_AV_P): Add builtin predicate definition.
23145 (VCMPGTUT, VCMPGTST, VCMPEQUT, CMPNET, CMPGE_1TI,
23146 CMPGE_U1TI, CMPLE_1TI, CMPLE_U1TI, VNOR_V1TI_UNS, VNOR_V1TI, VCMPNET_P,
23147 VCMPAET_P, VMULEUD, VMULESD, VMULOUD, VMULOSD, VRLQ,
23148 VSLQ, VSRQ, VSRAQ, VRLQNM, DIV_V1TI, UDIV_V1TI, DIVES_V1TI, DIVEU_V1TI,
23149 MODS_V1TI, MODU_V1TI, VRLQMI): New macro expansions.
23150 (VRLQ, VSLQ, VSRQ, VSRAQ, DIVE, MOD): New overload expansions.
23151 * config/rs6000/rs6000-call.c (P10_BUILTIN_VCMPEQUT,
23152 P10V_BUILTIN_CMPGE_1TI, P10V_BUILTIN_CMPGE_U1TI,
23153 P10V_BUILTIN_VCMPGTUT, P10V_BUILTIN_VCMPGTST,
23154 P10V_BUILTIN_CMPLE_1TI, P10V_BUILTIN_VCMPLE_U1TI,
23155 P10V_BUILTIN_DIV_V1TI, P10V_BUILTIN_UDIV_V1TI,
23156 P10V_BUILTIN_VMULESD, P10V_BUILTIN_VMULEUD,
23157 P10V_BUILTIN_VMULOSD, P10V_BUILTIN_VMULOUD,
23158 P10V_BUILTIN_VNOR_V1TI, P10V_BUILTIN_VNOR_V1TI_UNS,
23159 P10V_BUILTIN_VRLQ, P10V_BUILTIN_VRLQMI,
23160 P10V_BUILTIN_VRLQNM, P10V_BUILTIN_VSLQ,
23161 P10V_BUILTIN_VSRQ, P10V_BUILTIN_VSRAQ,
23162 P10V_BUILTIN_VCMPGTUT_P, P10V_BUILTIN_VCMPGTST_P,
23163 P10V_BUILTIN_VCMPEQUT_P, P10V_BUILTIN_VCMPGTUT_P,
23164 P10V_BUILTIN_VCMPGTST_P, P10V_BUILTIN_CMPNET,
23165 P10V_BUILTIN_VCMPNET_P, P10V_BUILTIN_VCMPAET_P,
23166 P10V_BUILTIN_DIVES_V1TI, P10V_BUILTIN_MODS_V1TI,
23167 P10V_BUILTIN_MODU_V1TI):
23168 New overloaded definitions.
23169 (rs6000_gimple_fold_builtin) [P10V_BUILTIN_VCMPEQUT,
23170 P10V_BUILTIN_CMPNET, P10V_BUILTIN_CMPGE_1TI,
23171 P10V_BUILTIN_CMPGE_U1TI, P10V_BUILTIN_VCMPGTUT,
23172 P10V_BUILTIN_VCMPGTST, P10V_BUILTIN_CMPLE_1TI,
23173 P10V_BUILTIN_CMPLE_U1TI]: New case statements.
23174 (rs6000_init_builtins) [bool_V1TI_type_node, int_ftype_int_v1ti_v1ti]:
23176 (altivec_init_builtins): New E_V1TImode case statement.
23177 (builtin_function_type)[P10_BUILTIN_128BIT_VMULEUD,
23178 P10_BUILTIN_128BIT_VMULOUD, P10_BUILTIN_128BIT_DIVEU_V1TI,
23179 P10_BUILTIN_128BIT_MODU_V1TI, P10_BUILTIN_CMPGE_U1TI,
23180 P10_BUILTIN_VCMPGTUT, P10_BUILTIN_VCMPEQUT]: New case statements.
23181 * config/rs6000/rs6000.c (rs6000_handle_altivec_attribute) [E_TImode,
23182 E_V1TImode]: New case statements.
23183 * config/rs6000/rs6000.h (rs6000_builtin_type_index): New enum
23184 value RS6000_BTI_bool_V1TI.
23185 * config/rs6000/vector.md (vector_gtv1ti,vector_nltv1ti,
23186 vector_gtuv1ti, vector_nltuv1ti, vector_ngtv1ti, vector_ngtuv1ti,
23187 vector_eq_v1ti_p, vector_ne_v1ti_p, vector_ae_v1ti_p,
23188 vector_gt_v1ti_p, vector_gtu_v1ti_p, vrotlv1ti3, vashlv1ti3,
23189 vlshrv1ti3, vashrv1ti3): New define_expands.
23190 * config/rs6000/vsx.md (UNSPEC_VSX_DIVSQ, UNSPEC_VSX_DIVUQ,
23191 UNSPEC_VSX_DIVESQ, UNSPEC_VSX_DIVEUQ, UNSPEC_VSX_MODSQ,
23192 UNSPEC_VSX_MODUQ): New unspecs.
23193 (mulv2di3, vsx_div_v1ti, vsx_udiv_v1ti, vsx_dives_v1ti,
23194 vsx_diveu_v1ti, vsx_mods_v1ti, vsx_modu_v1ti, xxswapd_v1ti): New
23196 (vcmpnet): New define_expand.
23197 * doc/extend.texi: Add documentation for the new builtins vec_rl,
23198 vec_rlmi, vec_rlnm, vec_sl, vec_sr, vec_sra, vec_mule, vec_mulo,
23199 vec_div, vec_dive, vec_mod, vec_cmpeq, vec_cmpne, vec_cmpgt, vec_cmplt,
23200 vec_cmpge, vec_cmple, vec_all_eq, vec_all_ne, vec_all_gt, vec_all_lt,
23201 vec_all_ge, vec_all_le, vec_any_eq, vec_any_ne, vec_any_gt, vec_any_lt,
23202 vec_any_ge, vec_any_le.
23204 2021-06-09 Carl Love <cel@us.ibm.com>
23206 * config/rs6000/altivec.md (altivec_vrl<VI_char>mi): Fix
23207 bug in argument generation.
23209 2021-06-09 Christophe Lyon <christophe.lyon@linaro.org>
23211 * config/arm/iterators.md (<supf>): Remove VCLZQ_U, VCLZQ_S.
23213 * config/arm/mve.md (mve_vclzq_<supf><mode>): Add '@' prefix,
23214 remove <supf> iterator.
23215 (mve_vclzq_u<mode>): New.
23216 * config/arm/neon.md (clz<mode>2): Rename to neon_vclz<mode>.
23217 (neon_vclz<mode): Move to ...
23218 * config/arm/unspecs.md (VCLZQ_U, VCLZQ_S): Remove.
23219 * config/arm/vec-common.md: ... here. Add support for MVE.
23221 2021-06-09 Christophe Lyon <christophe.lyon@linaro.org>
23223 * config/arm/mve.md (mve_vhaddq_<supf><mode>): Prefix with '@'.
23224 (@mve_vrhaddq_<supf><mode): Likewise.
23225 * config/arm/neon.md (neon_v<r>hadd<sup><mode>): Likewise.
23226 * config/arm/vec-common.md (avg<mode>3_floor, uavg<mode>3_floor)
23227 (avg<mode>3_ceil", uavg<mode>3_ceil): New patterns.
23229 2021-06-09 imba-tjd <109224573@qq.com>
23231 * doc/invoke.texi: Fix typo.
23233 2021-06-09 Roger Sayle <roger@nextmovesoftware.com>
23235 PR middle-end/53267
23236 * fold-const-call.c (fold_const_call_sss) [CASE_CFN_FMOD]:
23237 Support evaluation of fmod/fmodf/fmodl at compile-time.
23239 2021-06-09 Richard Biener <rguenther@suse.de>
23241 PR tree-optimization/100981
23242 * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
23243 gimple_get_lhs to also handle calls.
23244 * tree-vect-slp-patterns.c (complex_pattern::build): Transfer
23247 2021-06-09 Richard Biener <rguenther@suse.de>
23249 PR tree-optimization/97832
23250 * tree-vectorizer.h (_slp_tree::failed): New.
23251 * tree-vect-slp.c (_slp_tree::_slp_tree): Initialize
23253 (_slp_tree::~_slp_tree): Free failed.
23254 (vect_build_slp_tree): Retain failed nodes and record
23255 matches in them, copying that back out when running
23256 into a cached fail. Dump start and end of discovery.
23257 (dt_sort_cmp): New.
23258 (vect_build_slp_tree_2): Handle associatable chains
23259 together doing more aggressive operand swapping.
23261 2021-06-09 H.J. Lu <hjl.tools@gmail.com>
23264 * config.gcc (gcc_cv_initfini_array): Set to yes for Linux and
23266 * doc/install.texi: Require glibc 2.1 and binutils 2.12 for
23267 Linux and GNU targets.
23269 2021-06-09 Richard Biener <rguenther@suse.de>
23271 * tree-vect-stmts.c (vect_is_simple_use): Always get dt
23274 2021-06-09 Claudiu Zissulescu <claziss@synopsys.com>
23276 * config/arc/arc.md (loop_end): Change it to
23277 define_insn_and_split.
23279 2021-06-09 Claudiu Zissulescu <claziss@synopsys.com>
23281 * config/arc/arc.md (maddhisi4): Use VMAC2H instruction.
23282 (machi): New pattern.
23283 (umaddhisi4): Use VMAC2HU instruction.
23284 (umachi): New pattern.
23286 2021-06-09 Claudiu Zissulescu <claziss@synopsys.com>
23288 * config/arc/arc-protos.h (arc_split_move_p): New prototype.
23289 * config/arc/arc.c (arc_split_move_p): New function.
23290 (arc_split_move): Clean up.
23291 * config/arc/arc.md (movdi_insn): Clean up, use arc_split_move_p.
23292 (movdf_insn): Likewise.
23293 * config/arc/simdext.md (mov<VWH>_insn): Likewise.
23295 2021-06-09 Uroš Bizjak <ubizjak@gmail.com>
23298 * config/i386/i386.c (print_operand_address_as): Rename "no_rip"
23299 argument to "raw". Do not emit segment overrides when "raw" is true.
23301 2021-06-09 Martin Liska <mliska@suse.cz>
23303 * doc/gcov.texi: Create a proper JSON files.
23304 * doc/invoke.texi: Remove dots in order to make it a valid
23307 2021-06-09 Xionghu Luo <luoxhu@linux.ibm.com>
23309 * config/rs6000/rs6000-p8swap.c (pattern_is_rotate64): New.
23310 (insn_is_load_p): Use pattern_is_rotate64.
23311 (insn_is_swap_p): Likewise.
23312 (quad_aligned_load_p): Likewise.
23313 (const_load_sequence_p): Likewise.
23314 (replace_swapped_aligned_load): Likewise.
23315 (recombine_lvx_pattern): Likewise.
23316 (recombine_stvx_pattern): Likewise.
23318 2021-06-09 Andrew MacLeod <amacleod@redhat.com>
23320 * gimple-range-gori.cc (gori_compute::outgoing_edge_range_p): Use a
23321 fur_stmt source record.
23322 * gimple-range.cc (fur_source::get_operand): Generic range query.
23323 (fur_source::get_phi_operand): New.
23324 (fur_source::register_dependency): New.
23325 (fur_source::query): New.
23326 (class fur_edge): New. Edge source for operands.
23327 (fur_edge::fur_edge): New.
23328 (fur_edge::get_operand): New.
23329 (fur_edge::get_phi_operand): New.
23330 (fur_edge::query): New.
23331 (fur_stmt::fur_stmt): New.
23332 (fur_stmt::get_operand): New.
23333 (fur_stmt::get_phi_operand): New.
23334 (fur_stmt::query): New.
23335 (class fur_depend): New. Statement source and process dependencies.
23336 (fur_depend::fur_depend): New.
23337 (fur_depend::register_dependency): New.
23338 (class fur_list): New. List source for operands.
23339 (fur_list::fur_list): New.
23340 (fur_list::get_operand): New.
23341 (fur_list::get_phi_operand): New.
23342 (fold_range): New. Instantiate appropriate fur_source class and fold.
23343 (fold_using_range::range_of_range_op): Use new API.
23344 (fold_using_range::range_of_address): Ditto.
23345 (fold_using_range::range_of_phi): Ditto.
23346 (imple_ranger::fold_range_internal): Use fur_depend class.
23347 (fold_using_range::range_of_ssa_name_with_loop_info): Use new API.
23348 * gimple-range.h (class fur_source): Now a base class.
23349 (class fur_stmt): New.
23350 (fold_range): New prototypes.
23351 (fur_source::fur_source): Delete.
23353 2021-06-08 Andrew Pinski <apinski@marvell.com>
23355 PR tree-optimization/25290
23356 * tree-ssa-phiopt.c (xor_replacement): Delete.
23357 (tree_ssa_phiopt_worker): Delete use of xor_replacement.
23358 (match_simplify_replacement): Allow one cheap preparation
23359 statement that can be moved to before the if.
23361 2021-06-08 Pat Haugen <pthaugen@linux.ibm.com>
23363 * config/rs6000/power10.md (power10-fused-load, power10-fused-store,
23364 power10-fused_alu, power10-fused-vec, power10-fused-branch): New.
23366 2021-06-08 Jeff Law <jeffreyalaw@gmail.com>
23368 * config/h8300/logical.md (andqi3_1): Move BCLR case into define_insn_and_split.
23369 Create length attribute on define_insn_and_split. Only split for cases which we
23371 (andqi3_1<cczn>): Renamed from andqi3_1_clobber_flags. Only handle AND here and
23372 fix length computation.
23373 (b<code><mode>msx): Combine QImode and HImode H8/SX patterns using iterator.
23375 2021-06-08 Richard Biener <rguenther@suse.de>
23377 PR tree-optimization/100923
23378 * tree-ssa-sccvn.c (valueize_refs_1): Take a pointer to
23379 the operand vector to be valueized.
23380 (valueize_refs): Likewise.
23381 (valueize_shared_reference_ops_from_ref): Adjust.
23382 (valueize_shared_reference_ops_from_call): Likewise.
23383 (vn_reference_lookup_3): Likewise.
23384 (vn_reference_lookup_pieces): Likewise. Re-valueize
23385 with honoring availability when we are about to create
23386 the ao_ref and valueized before.
23387 (vn_reference_lookup): Likewise.
23388 (vn_reference_insert_pieces): Adjust.
23390 2021-06-08 Richard Biener <rguenther@suse.de>
23392 * tree-vectorizer.h (_slp_instance::root_stmt): Change to...
23393 (_slp_instance::root_stmts): ... a vector.
23394 (SLP_INSTANCE_ROOT_STMT): Rename to ...
23395 (SLP_INSTANCE_ROOT_STMTS): ... this.
23396 (slp_root::root): Change to...
23397 (slp_root::roots): ... a vector.
23398 (slp_root::slp_root): Adjust.
23399 * tree-vect-slp.c (_slp_instance::location): Adjust.
23400 (vect_free_slp_instance): Release the root stmt vector.
23401 (vect_build_slp_instance): Adjust.
23402 (vect_analyze_slp): Likewise.
23403 (_bb_vec_info::~_bb_vec_info): Likewise.
23404 (vect_slp_analyze_operations): Likewise.
23405 (vect_bb_vectorization_profitable_p): Likewise. Adjust
23406 costs for the root stmt.
23407 (vect_slp_check_for_constructors): Gather all BIT_INSERT_EXPRs
23409 (vect_slp_analyze_bb_1): Simplify by marking all root stmts
23411 (vectorize_slp_instance_root_stmt): Adjust.
23412 (vect_schedule_slp): Likewise.
23414 2021-06-08 Aldy Hernandez <aldyh@redhat.com>
23416 * gimple-ssa-evrp.c (class ssa_equiv_stack): New.
23417 (ssa_equiv_stack::ssa_equiv_stack): New.
23418 (ssa_equiv_stack::~ssa_equiv_stack): New.
23419 (ssa_equiv_stack::enter): New.
23420 (ssa_equiv_stack::leave): New.
23421 (ssa_equiv_stack::push_replacement): New.
23422 (ssa_equiv_stack::get_replacement): New.
23423 (is_pointer_ssa): New.
23424 (class pointer_equiv_analyzer): New.
23425 (pointer_equiv_analyzer::pointer_equiv_analyzer): New.
23426 (pointer_equiv_analyzer::~pointer_equiv_analyzer): New.
23427 (pointer_equiv_analyzer::set_global_equiv): New.
23428 (pointer_equiv_analyzer::set_cond_equiv): New.
23429 (pointer_equiv_analyzer::get_equiv): New.
23430 (pointer_equiv_analyzer::enter): New.
23431 (pointer_equiv_analyzer::leave): New.
23432 (pointer_equiv_analyzer::get_equiv_expr): New.
23433 (pta_valueize): New.
23434 (pointer_equiv_analyzer::visit_stmt): New.
23435 (pointer_equiv_analyzer::visit_edge): New.
23436 (hybrid_folder::value_of_expr): Call PTA.
23437 (hybrid_folder::value_on_edge): Same.
23438 (hybrid_folder::pre_fold_bb): New.
23439 (hybrid_folder::post_fold_bb): New.
23440 (hybrid_folder::pre_fold_stmt): New.
23441 (rvrp_folder::pre_fold_bb): New.
23442 (rvrp_folder::post_fold_bb): New.
23443 (rvrp_folder::pre_fold_stmt): New.
23444 (rvrp_folder::value_of_expr): Call PTA.
23445 (rvrp_folder::value_on_edge): Same.
23447 2021-06-08 Jakub Jelinek <jakub@redhat.com>
23450 * tree-inline.c (copy_tree_body_r): For OMP_CLAUSE_DEPEND don't
23451 check TREE_CODE if OMP_CLAUSE_DECL is NULL.
23453 2021-06-08 Richard Biener <rguenther@suse.de>
23455 PR middle-end/100951
23456 * tree-vect-generic.c (expand_vector_piecewise): Build a
23457 VECTOR_CST if all elements are constant.
23458 (expand_vector_condition): Likewise.
23459 (lower_vec_perm): Likewise.
23460 (expand_vector_conversion): Likewise.
23462 2021-06-08 Martin Liska <mliska@suse.cz>
23464 * doc/invoke.texi: Document new param evrp-sparse-threshold.
23466 2021-06-08 Martin Liska <mliska@suse.cz>
23468 * genautomata.c (create_automata): Fix typo.
23470 2021-06-08 Kewen Lin <linkw@linux.ibm.com>
23472 PR tree-optimization/100794
23473 * tree-predcom.c (tree_predictive_commoning_loop): Add parameter
23474 allow_unroll_p and only allow unrolling when it's true.
23475 (tree_predictive_commoning): Add parameter allow_unroll_p and
23477 (run_tree_predictive_commoning): Likewise.
23478 (pass_predcom::gate): Check flag_tree_loop_vectorize and
23479 global_options_set.x_flag_predictive_commoning.
23480 (pass_predcom::execute): Adjust for allow_unroll_p.
23482 2021-06-08 Kewen Lin <linkw@linux.ibm.com>
23484 * tree-predcom.c (execute_pred_commoning): Remove update_ssa call.
23485 (tree_predictive_commoning_loop): Factor some cleanup stuffs into
23486 lambda function cleanup, remove scev_reset call, and adjust return
23488 (tree_predictive_commoning): Adjust for different changed values,
23489 only set flag TODO_update_ssa_only_virtuals if changed.
23490 (pass_data pass_data_predcom): Remove TODO_update_ssa_only_virtuals
23491 from todo_flags_finish.
23493 2021-06-07 Andrew MacLeod <amacleod@redhat.com>
23495 * gimple-range-cache.cc (class sbr_sparse_bitmap): New.
23496 (sbr_sparse_bitmap::sbr_sparse_bitmap): New.
23497 (sbr_sparse_bitmap::bitmap_set_quad): New.
23498 (sbr_sparse_bitmap::bitmap_get_quad): New.
23499 (sbr_sparse_bitmap::set_bb_range): New.
23500 (sbr_sparse_bitmap::get_bb_range): New.
23501 (sbr_sparse_bitmap::bb_range_p): New.
23502 (block_range_cache::block_range_cache): initialize bitmap obstack.
23503 (block_range_cache::~block_range_cache): Destruct obstack.
23504 (block_range_cache::set_bb_range): Decide when to utilze the
23505 sparse on entry cache.
23506 * gimple-range-cache.h (block_range_cache): Add bitmap obstack.
23507 * params.opt (-param=evrp-sparse-threshold): New.
23509 2021-06-07 Andrew MacLeod <amacleod@redhat.com>
23511 * bitmap.c (bitmap_set_aligned_chunk): New.
23512 (bitmap_get_aligned_chunk): New.
23513 (test_aligned_chunk): New.
23514 (bitmap_c_tests): Call test_aligned_chunk.
23515 * bitmap.h (bitmap_set_aligned_chunk, bitmap_get_aligned_chunk): New.
23517 2021-06-07 Uroš Bizjak <ubizjak@gmail.com>
23520 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
23522 (ix86_expand_vector_init_one_nonzero): Ditto.
23523 (ix86_expand_vector_init_one_var): Ditto.
23524 (ix86_expand_vector_init_general): Ditto.
23525 * config/i386/mmx.md (vec_initv4qiqi): New expander.
23527 2021-06-07 Jeff Law <jeffreyalaw@gmail.com>
23529 * config/h8300/movepush.md: Change most _clobber_flags
23530 patterns to instead use <cczn> subst.
23531 (movsi_cczn): New pattern with usable CC cases split out.
23532 (movsi_h8sx_cczn): Likewise.
23534 2021-06-07 Martin Liska <mliska@suse.cz>
23536 * common/common-target.def: Split long lines and replace them
23538 * target.def: Likewise.
23539 * doc/tm.texi: Re-generated.
23541 2021-06-07 Jakub Jelinek <jakub@redhat.com>
23544 * fold-const.c (fold_read_from_vector): Return NULL if trying to
23545 read from a CONSTRUCTOR with vector type elements.
23547 2021-06-07 Jakub Jelinek <jakub@redhat.com>
23549 PR middle-end/100898
23550 * tree-inline.c (copy_bb): Only use gimple_call_arg_ptr if memcpy
23551 should copy any arguments. Don't call gimple_call_num_args
23552 on id->call_stmt or call_stmt more than once.
23554 2021-06-07 liuhongt <hongtao.liu@intel.com>
23557 * config/i386/sse.md (*sse4_1_zero_extendv8qiv8hi2_3): Refine
23559 (<insn>v4siv4di2): Delete constraints for define_expand.
23561 2021-06-07 liuhongt <hongtao.liu@intel.com>
23564 * config/i386/i386-expand.c (ix86_expand_builtin): Remove
23565 assignment of cfun->machine->has_explicit_vzeroupper.
23566 * config/i386/i386-features.c
23567 (ix86_add_reg_usage_to_vzerouppers): Delete.
23568 (ix86_add_reg_usage_to_vzeroupper): Ditto.
23569 (rest_of_handle_insert_vzeroupper): Remove
23570 ix86_add_reg_usage_to_vzerouppers, add df_analyze at the end
23572 (gate): Remove cfun->machine->has_explicit_vzeroupper.
23573 * config/i386/i386-protos.h (ix86_expand_avx_vzeroupper):
23575 * config/i386/i386.c (ix86_insn_callee_abi): New function.
23576 (ix86_initialize_callee_abi): Ditto.
23577 (ix86_expand_avx_vzeroupper): Ditto.
23578 (ix86_hard_regno_call_part_clobbered): Adjust for vzeroupper
23580 (TARGET_INSN_CALLEE_ABI): Define as ix86_insn_callee_abi.
23581 (ix86_emit_mode_set): Call ix86_expand_avx_vzeroupper
23583 * config/i386/i386.h (struct GTY(()) machine_function): Delete
23584 has_explicit_vzeroupper.
23585 * config/i386/i386.md (enum unspec): New member
23587 (ABI_DEFAULT,ABI_VZEROUPPER,ABI_UNKNOWN): New
23588 define_constants for insn callee abi index.
23589 * config/i386/predicates.md (vzeroupper_pattern): Adjust.
23590 * config/i386/sse.md (UNSPECV_VZEROUPPER): Deleted.
23591 (avx_vzeroupper): Call ix86_expand_avx_vzeroupper.
23592 (*avx_vzeroupper): Rename to ..
23593 (avx_vzeroupper_callee_abi): .. this, and adjust pattern as
23594 call_insn which has a special vzeroupper ABI.
23595 (*avx_vzeroupper_1): Deleted.
23597 2021-06-07 liuhongt <hongtao.liu@intel.com>
23600 * df-scan.c (df_get_call_refs): When call_insn is a fake call,
23601 it won't use stack pointer reg.
23602 * final.c (leaf_function_p): When call_insn is a fake call, it
23603 won't affect caller as a leaf function.
23604 * reg-stack.c (callee_clobbers_any_stack_reg): New.
23605 (subst_stack_regs): When call_insn doesn't clobber any stack
23606 reg, don't clear the arguments.
23607 * rtl.c (shallow_copy_rtx): Don't clear flag used when orig is
23609 * shrink-wrap.c (requires_stack_frame_p): No need for stack
23610 frame for a fake call.
23611 * rtl.h (FAKE_CALL_P): New macro.
23613 2021-06-06 Eric Botcazou <ebotcazou@adacore.com>
23615 * config/sparc/sparc-protos.h (order_regs_for_local_alloc): Rename
23617 (sparc_order_regs_for_local_alloc): ...this.
23618 (sparc_leaf_reg_remap): Declare.
23619 * config/sparc/sparc.h (ADJUST_REG_ALLOC_ORDER): Adjust.
23620 (LEAF_REG_REMAP): Reimplement as call to sparc_leaf_reg_remap.
23621 * config/sparc/sparc.c (leaf_reg_remap): Delete.
23622 (order_regs_for_local_alloc): Rename to...
23623 (sparc_order_regs_for_local_alloc): ...this.
23624 (sparc_leaf_reg_remap): New function.
23625 (sparc_conditional_register_usage): Do not modify leaf_reg_remap.
23627 2021-06-06 David Edelsohn <dje.gcc@gmail.com>
23629 * config/rs6000/rs6000.c (rs6000_xcoff_asm_output_aligned_decl_common):
23630 Use assemble_name to output BSS section name.
23632 2021-06-06 Uroš Bizjak <ubizjak@gmail.com>
23634 * config/i386/constraints.md (Bs):
23635 Remove boolean operators from match_test RTX.
23638 (M): Use "mode" variable instead of GET_MODE (op) in match_test RTX.
23641 2021-06-06 Martin Liska <mliska@suse.cz>
23643 * doc/extend.texi: Add missing @headitem.
23644 * doc/invoke.texi: Likewise.
23645 * doc/objc.texi: Likewise.
23647 2021-06-06 Martin Liska <mliska@suse.cz>
23649 * genhooks.c (emit_findices): Remove unused function.
23650 (emit_documentation): Do not call emit_findices
23651 and do not search for @Fcode directives.
23653 2021-06-06 Martin Liska <mliska@suse.cz>
23655 * doc/invoke.texi: Remove extra character.
23657 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
23659 * config/sh/sh.md (doloop_end_split): Fix empty split condition.
23661 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
23663 * config/sparc/sparc.md (*snedi<W:mode>_zero_vis3,
23664 *neg_snedi<W:mode>_zero_subxc, *plus_snedi<W:mode>_zero,
23665 *plus_plus_snedi<W:mode>_zero, *minus_snedi<W:mode>_zero,
23666 *minus_minus_snedi<W:mode>_zero): Fix empty split condition.
23668 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
23670 * config/or1k/or1k.md (*movdi): Fix empty split condition.
23672 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
23674 * config/mips/mips.md (<anonymous>, bswapsi2, bswapdi2): Fix empty
23677 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
23679 * config/m68k/m68k.md (*zero_extend_inc, *zero_extend_dec,
23680 *zero_extendsidi2): Fix empty split condition.
23682 2021-06-05 Jeff Law <jeffreyalaw@gmail.com>
23684 * config/h8300/addsub.md: Fix split condition in define_insn_and_split
23686 * config/h8300/bitfield.md: Likewise.
23687 * config/h8300/combiner.md: Likewise.
23688 * config/h8300/divmod.md: Likewise.
23689 * config/h8300/extensions.md: Likewise.
23690 * config/h8300/jumpcall.md: Likewise.
23691 * config/h8300/movepush.md: Likewise.
23692 * config/h8300/multiply.md: Likewise.
23693 * config/h8300/other.md: Likewise.
23694 * config/h8300/shiftrotate.md: Likewise.
23695 * config/h8300/logical.md: Likewise. Fix split pattern to use
23696 code iterator that somehow slipped through.
23698 2021-06-04 Tobias Burnus <tobias@codesourcery.com>
23700 PR middle-end/100905
23701 * tree-nested.c (convert_nonlocal_omp_clauses,
23702 convert_local_omp_clauses): Handle OMP_CLAUSE_BIND.
23704 2021-06-04 Martin Sebor <msebor@redhat.com>
23706 PR middle-end/100732
23707 * gimple-fold.c (gimple_fold_builtin_sprintf): Avoid folding calls
23708 with either source or destination argument of invalid type.
23709 * tree-ssa-uninit.c (maybe_warn_pass_by_reference): Avoid checking
23710 calls with arguments of invalid type.
23712 2021-06-04 Martin Sebor <msebor@redhat.com>
23714 * attribs.c (init_attr_rdwr_indices): Use VLA bounds in the expected
23716 (attr_access::vla_bounds): Also handle VLA bounds.
23718 2021-06-04 Uroš Bizjak <ubizjak@gmail.com>
23720 * config/i386/predicates.md (GOT_memory_operand):
23721 Implement using match_code RTXes.
23722 (GOT32_symbol_operand): Ditto.
23724 2021-06-04 Uroš Bizjak <ubizjak@gmail.com>
23727 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
23729 (ix86_expand_vector_init_general): Ditto.
23730 Use SImode instead of word_mode for logic operations
23731 when GET_MODE_SIZE (mode) < UNITS_PER_WORD.
23732 (expand_vec_perm_even_odd_1): Assert that V2HI mode should be
23733 implemented by expand_vec_perm_1.
23734 (expand_vec_perm_broadcast_1): Assert that V2HI and V4HI modes
23735 should be implemented using standard shuffle patterns.
23736 (ix86_vectorize_vec_perm_const): Handle V2HImode. Add V4HI and
23737 V2HI modes to modes, implementable with shuffle for one operand.
23738 * config/i386/mmx.md (*punpckwd): New insn_and_split pattern.
23739 (*pshufw_1): New insn pattern.
23740 (*vec_dupv2hi): Ditto.
23741 (vec_initv2hihi): New expander.
23743 2021-06-04 Kewen Lin <linkw@linux.ibm.com>
23745 * config/arm/vfp.md (no_literal_pool_df_immediate,
23746 no_literal_pool_sf_immediate): Fix empty split condition.
23748 2021-06-04 Kewen Lin <linkw@linux.ibm.com>
23750 * config/i386/i386.md (*load_tp_x32_zext, *add_tp_x32_zext,
23751 *tls_dynamic_gnu2_combine_32): Fix empty split condition.
23752 * config/i386/sse.md (*<sse2_avx2>_pmovmskb_lt,
23753 *<sse2_avx2>_pmovmskb_zext_lt, *sse2_pmovmskb_ext_lt,
23754 *<sse4_1_avx2>_pblendvb_lt): Likewise.
23756 2021-06-04 Jakub Jelinek <jakub@redhat.com>
23759 * config/i386/i386-expand.c (ix86_expand_vector_init): Handle
23760 concatenation from half-sized modes with TImode elements.
23762 2021-06-04 Claudiu Zissulescu <claziss@synopsys.com>
23764 * config/arc/arc.c (arc_override_options): Disable millicode
23765 thunks when RF16 is on.
23767 2021-06-04 Haochen Gui <guihaoc@gcc.gnu.org>
23769 * config/rs6000/rs6000.h (PROMOTE_MODE): Remove.
23771 2021-06-04 Haochen Gui <guihaoc@gcc.gnu.org>
23773 * config/rs6000/rs6000-call.c (rs6000_promote_function_mode):
23774 Replace PROMOTE_MODE marco with its content.
23776 2021-06-03 Kewen Lin <linkw@linux.ibm.com>
23778 * config/cris/cris.md (*addi_reload): Fix empty split condition.
23780 2021-06-03 Jim Wilson <jimw@sifive.com>
23782 * config.gcc (riscv*-*-*): If --with-riscv-attribute not used,
23783 turn it on for all riscv targets.
23785 2021-06-03 Uroš Bizjak <ubizjak@gmail.com>
23788 * config/i386/i386-expand.c (ix86_expand_vector_set):
23789 Handle V2HI and V4QI modes.
23790 (ix86_expand_vector_extract): Ditto.
23791 * config/i386/mmx.md (*pinsrw): New insn pattern.
23794 (*pextrw_zext): Ditto.
23796 (*pextrb_zext): Ditto.
23797 (vec_setv2hi): New expander.
23798 (vec_extractv2hihi): Ditto.
23799 (vec_setv4qi): Ditto.
23800 (vec_extractv4qiqi): Ditto.
23801 (vec_setv8qi): Enable only for TARGET_SSE4_1.
23802 (vec_extractv8qiqi): Ditto.
23804 2021-06-03 Aaron Sawdey <acsawdey@linux.ibm.com>
23806 * config/rs6000/genfusion.pl (gen_logical_addsubf): Fix input
23807 order to subf instruction.
23808 * config/rs6000/fusion.md: Regenerate.
23810 2021-06-03 Aldy Hernandez <aldyh@redhat.com>
23812 * calls.c (get_size_range): Use range_of_expr instead of
23813 determine_value_range.
23814 * tree-affine.c (expr_to_aff_combination): Same.
23815 * tree-data-ref.c (split_constant_offset): Same.
23816 * tree-vrp.c (determine_value_range_1): Remove.
23817 (determine_value_range): Remove.
23818 * tree-vrp.h (determine_value_range): Remove.
23820 2021-06-03 Aldy Hernandez <aldyh@redhat.com>
23822 * function-tests.c (test_ranges): Call gimple_range_tests.
23823 * gimple-range-cache.cc (ranger_cache::range_of_expr): Pass stmt
23825 * gimple-range.cc (fur_source::get_operand): Do not call
23826 get_tree_range or gimple_range_global.
23828 (get_tree_range): Move to value-query.cc.
23829 Call get_arith_expr_range.
23830 (gimple_ranger::range_of_expr): Add argument to get_tree_range.
23831 Include gimple-range-tests.cc.
23832 * gimple-range.h (fold_range): Add argument.
23833 (get_tree_range): Remove.
23834 * selftest.h (gimple_range_tests): New.
23835 * value-query.cc (global_range_query::range_of_expr): Add
23837 (range_query::get_tree_range): Move from gimple-range.cc.
23838 * value-query.h (class range_query): Add get_tree_range and
23839 get_arith_expr_range. Make fur_source a friend.
23840 * vr-values.c (vr_values::range_of_expr): Pass stmt to
23842 * gimple-range-tests.cc: New file.
23844 2021-06-03 Aldy Hernandez <aldyh@redhat.com>
23846 * gimple-range.cc (gimple_ranger::export_global_ranges): Call
23847 update_global_range.
23848 * value-query.cc (update_global_range): New.
23849 * value-query.h (update_global_range): New.
23851 2021-06-03 David Malcolm <dmalcolm@redhat.com>
23853 * diagnostic-show-locus.c (diagnostic_show_locus): Don't reject
23854 printing the same location twice if there are fix-it hints,
23855 multiple locations, or a label.
23857 2021-06-03 Andre Vieira <andre.simoesdiasvieira@arm.com>
23859 * tree-vect-loop.c (vect_transform_loop): Use main loop's various'
23860 thresholds to narrow the upper bound on epilogue iterations.
23862 2021-06-03 Christophe Lyon <christophe.lyon@linaro.org>
23864 * config/arm/mve.md (mve_vabsq_f<mode>): Use 'abs' instead of unspec.
23865 (mve_vabsq_s<mode>): Likewise.
23866 * config/arm/neon.md (abs<mode>2): Rename to neon_abs<mode>2.
23867 * config/arm/unspecs.md (VABSQ_F, VABSQ_S): Delete.
23868 * config/arm/vec-common.md (neg<mode>2): Rename to
23869 <absneg_str><mode>2.
23871 2021-06-03 Claudiu Zissulescu <claziss@synopsys.com>
23873 * common/config/arc/arc-common.c (arc_option_optimization_table):
23874 Remove malign-call.
23875 * config/arc/arc.c (arc_unalign_branch_p): Remove unused function.
23876 * config/arc/arc.h (TARGET_MIXED_CODE): Remove macro.
23877 (INDEX_REG_CLASS): Only refer to GENERAL_REGS.
23878 * config/arc/arc.md (abssi2_mixed): Remove pattern.
23879 * config/arc/arc.opt (munalign-prob-threshold): Mark it obsolete.
23880 (malign-call): Likewise.
23881 (mmixed-code): Likewise.
23882 * doc/invoke.texi (ARC): Update doc.
23884 2021-06-03 Martin Liska <mliska@suse.cz>
23886 * common.opt: Use proper Enum values.
23887 * opts.c (COVERAGE_SANITIZER_OPT): Remove.
23888 (parse_sanitizer_options): Handle only sanitizer_opts.
23889 (common_handle_option): Just assign value.
23891 2021-06-03 Eric Botcazou <ebotcazou@adacore.com>
23894 * tree-inline.c (inline_forbidden_p): Remove test on return type.
23896 2021-06-03 Eric Botcazou <ebotcazou@adacore.com>
23898 * dwarf2out.c (loc_list_from_tree_1) <FUNCTION_DECL>: Also generate
23899 DW_OP_GNU_variable_value referencing an existing DIE at file scope.
23900 (type_byte_size): Inline into...
23901 (add_byte_size_attribute): ...this and call add_scalar_info.
23903 2021-06-03 Eric Botcazou <ebotcazou@adacore.com>
23905 * dwarf2out.c (mem_loc_descriptor) <UDIV>: Fix typo.
23906 (typed_binop_from_tree): New function.
23907 (loc_list_from_tree_1) <EXACT_DIV_EXPR>: For an unsigned type,
23908 turn a divide by a power of 2 into a shift.
23909 <CEIL_DIV_EXPR>: For an unsigned type, use a signed divide if the
23910 size of the mode is lower than DWARF2_ADDR_SIZE; otherwise, do a
23911 typed divide by calling typed_binop_from_tree.
23913 2021-06-03 Eric Botcazou <ebotcazou@adacore.com>
23915 * dwarf2out.c (scompare_loc_descriptor): Fix head comment.
23916 (is_handled_procedure_type): Likewise.
23917 (struct loc_descr_context): Add strict_signedness field.
23918 (resolve_args_picking_1): Deal with DW_OP_[GNU_]deref_type,
23919 DW_OP_[GNU_]convert and DW_OP_[GNU_]reinterpret.
23920 (resolve_args_picking): Minor tweak.
23921 (function_to_dwarf_procedure): Initialize strict_signedness field.
23922 (type_byte_size): Likewise.
23923 (field_byte_offset): Likewise.
23924 (gen_descr_array_type_die): Likewise.
23925 (gen_variant_part): Likewise.
23926 (loc_list_from_tree_1) <CALL_EXPR>: Tidy up and set strict_signedness
23927 to true when a context is present before evaluating the arguments.
23928 <COND_EXPR>: Do not generate a useless comparison with zero.
23929 When dereferencing an address, if strict_signedness is true and the
23930 type is small and signed, use DW_OP_deref_type to do the dereference
23931 and then DW_OP_convert to convert back to the generic type.
23933 2021-06-03 Jakub Jelinek <jakub@redhat.com>
23936 * tree-inline.c (copy_tree_body_r): Handle iterators on
23937 OMP_CLAUSE_AFFINITY or OMP_CLAUSE_DEPEND.
23939 2021-06-03 Kewen Lin <linkw@linux.ibm.com>
23941 * config/arc/arc.md (*bbit_di): Remove.
23943 2021-06-02 Christoph Muellner <cmuellner@gcc.gnu.org>
23945 PR rtl-optimization/100264
23946 * ree.c (get_sub_rtx): Ignore SET expressions without register
23947 destinations and remove assertion, as it is not valid anymore
23948 with this new behaviour.
23949 (merge_def_and_ext): Eliminate destination check for register
23950 as such SET expressions can't occur anymore.
23951 (combine_reaching_defs): Likewise.
23953 2021-06-02 Jakub Jelinek <jakub@redhat.com>
23956 * config/xtensa/xtensa.h (LEAF_REG_REMAP): Cast REGNO to int to avoid
23957 -Wtype-limits warnings.
23958 (DWARF_FRAME_REGISTER): Rewrite into ternary operator with addition
23959 in operands to avoid -Wsign-compare warnings.
23961 2021-06-02 Pat Haugen <pthaugen@linux.ibm.com>
23963 * config/rs6000/rs6000-logue.c (rs6000_emit_prologue): Use
23966 2021-06-02 Vineet Gupta <vgupta@synopsys.com>
23968 * config/arc/arc.h (TARGET_CPU_DEFAULT): Change to hs38_linux.
23970 2021-06-02 Ilya Leoshkevich <iii@linux.ibm.com>
23972 * config/s390/s390.md(*ashrdi3_31<setcc><cconly>): Use a single
23974 * config/s390/subst.md(cconly_subst): Use a single constraint
23975 in (match_scratch).
23977 2021-06-02 Martin Liska <mliska@suse.cz>
23979 * ipa-icf.h: Use auto_vec for memory_access_types.
23981 2021-06-02 Jeff Law <jeffreyalaw@gmail.com>
23983 * config/h8300/h8300-protos.h (compute_a_shift_length): Drop unused
23984 argument from prototype.
23985 (output_logical_op): Add rtx_code argument.
23986 (compute_logical_op_length): Likewise.
23987 * config/h8300/h8300.c (h8300_and_costs): Pass additional argument
23988 to compute_a_shift_length.
23989 (output_logical_op); New argument with the rtx code rather than
23990 extracting it from an operand. Handle QImode too.
23991 (compute_logical_op_length): Similary.
23992 (compute_a_shift_length): Drop unused argument.
23993 * config/h8300/h8300.md (logicals): New code iterator.
23994 * config/h8300/logical.md (<code><mode>3 expander): Combine
23995 the "and" expander with the "ior"/"xor" expander.
23996 (bclr<mode>msx): Combine the QI/HI mode patterns.
23997 (<logical><mode>3 insns): Use code iterator rather than match_operator.
23998 Handle QImode as well. Update call to output_logical_op and
23999 compute_logical_op_length to pass in rtx_code
24000 Fix split condition on all define_insn_and_split patterns.
24001 (one_cmpl<mode>2<cczn>): Use <cczn> to support both clobbering
24002 the flags and setting ZN via existing define_subst.
24003 * config/h8300/shiftrotate.md: Drop unused argument from
24004 calls to compute_a_shift_length.
24005 Signed-off-by: Jeff Law <jeffreyalaw@gmail.com>
24007 2021-06-01 Andrew Pinski <apinski@marvell.com>
24009 PR tree-optimization/25290
24010 * tree-ssa-phiopt.c (match_simplify_replacement):
24012 (tree_ssa_phiopt_worker): Use match_simplify_replacement.
24013 (two_value_replacement): Change the comment about
24014 conditional_replacement.
24015 (conditional_replacement): Delete.
24017 2021-06-01 Andrew Pinski <apinski@marvell.com>
24019 PR tree-optimization/95481
24020 * tree-tailcall.c (find_tail_calls): Handle empty typed
24023 2021-06-01 Andrew Pinski <apinski@marvell.com>
24025 * gimplify.c (zero_sized_field_decl): Delete
24026 (zero_sized_type): Delete
24027 (gimplify_init_ctor_eval): Use is_empty_type instead
24028 of zero_sized_field_decl.
24029 (gimplify_modify_expr): Use is_empty_type instead of
24032 2021-06-01 Jason Merrill <jason@redhat.com>
24035 * tree.h (CALL_FROM_NEW_OR_DELETE_P): Adjust comment.
24037 2021-06-01 Jason Merrill <jason@redhat.com>
24040 * diagnostic.h (warning_enabled_at): Declare.
24041 * diagnostic.c (diagnostic_enabled): Factor out from...
24042 (diagnostic_report_diagnostic): ...here.
24043 (warning_enabled_at): New.
24045 2021-06-01 Aldy Hernandez <aldyh@redhat.com>
24047 * gimple-ssa-evrp.c: Enable exporting of global ranges.
24049 2021-06-01 Martin Liska <mliska@suse.cz>
24052 * doc/invoke.texi: Mention that -fgcse-after-reload
24053 is enabled with -O3.
24055 2021-06-01 liuhongt <hongtao.liu@intel.com>
24057 PR tree-optimization/98365
24058 * tree-if-conv.c (strip_nop_cond_scalar_reduction): New function.
24059 (is_cond_scalar_reduction): Handle nop_expr in cond scalar reduction.
24060 (convert_scalar_cond_reduction): Ditto.
24061 (predicate_scalar_phi): Ditto.
24063 2021-06-01 Andrew MacLeod <amacleod@redhat.com>
24065 PR tree-optimization/100781
24066 * gimple-range-cache.cc (ranger_cache::ranger_cache): Enable new
24067 value calculation by default.
24068 (ranger_cache::enable_new_values): New.
24069 (ranger_cache::disable_new_values): New.
24070 (ranger_cache::push_poor_value): Check if new values are allowed.
24071 * gimple-range-cache.h (class ranger_cache): New member/methods.
24072 * gimple-range.cc (gimple_ranger::range_of_expr): Check for debug
24073 statement, and disable/renable new value calculation.
24075 2021-06-01 Andrew MacLeod <amacleod@redhat.com>
24077 * gimple-range-cache.cc (ranger_cache::ssa_range_in_bb): Delete.
24078 (ranger_cache::range_of_def): New.
24079 (ranger_cache::entry_range): New.
24080 (ranger_cache::exit_range): New.
24081 (ranger_cache::range_of_expr): Adjust.
24082 (ranger_cache::range_on_edge): Adjust.
24083 (ranger_cache::propagate_cache): Call exit_range directly.
24084 * gimple-range-cache.h (class ranger_cache): Adjust.
24086 2021-06-01 Andrew MacLeod <amacleod@redhat.com>
24088 * gimple-range-cache.cc (ranger_cache::ranger_cache): Adjust for
24089 gori_compute being a member rather than base class.
24090 dervied call to member call.
24091 (ranger_cache::dump): No longer dump gori_map.
24092 (ranger_cache::dump_bb): New.
24093 (ranger_cache::get_non_stale_global_range): Adjust for gori_compute
24094 being a member rather than base class.
24095 (ranger_cache::set_global_range): Ditto.
24096 (ranger_cache::ssa_range_in_bb): Ditto.
24097 (ranger_cache::range_of_expr): New.
24098 (ranger_cache::range_on_edge): New.
24099 (ranger_cache::block_range): Adjust for gori_computes. Debug changes.
24100 (ranger_cache::propagate_cache): Adjust debugging output.
24101 (ranger_cache::fill_block_cache): Adjust for gori_computes. Debug
24103 * gimple-range-cache.h (class ranger_cache): Make gori_compute a
24104 member, and inherit from range_query instead.
24105 (ranger_cache::dump_bb): New. split from dump.
24106 * gimple-range-gori.cc (gori_compute::ssa_range_in_bb): Delete.
24107 (gori_compute::expr_range_at_stmt): Delete.
24108 (gori_compute::compute_name_range_op): Delete.
24109 (gori_compute::compute_operand_range_switch): Add fur_source.
24110 (gori_compute::compute_operand_range): Add fur_source param, inline
24111 old compute_name_range_op and optimize_logical_operands.
24112 (struct tf_range): Delete.
24113 (gori_compute::logical_combine): Adjust
24114 (gori_compute::optimize_logical_operands): Delete.
24115 (gori_compute::compute_logical_operands_in_chain): Delete.
24116 (gori_compute::compute_logical_operands): Adjust.
24117 (gori_compute::compute_operand1_range): Adjust to fur_source.
24118 (gori_compute::compute_operand2_range): Ditto.
24119 (gori_compute::compute_operand1_and_operand2_range): Ditto.
24120 (gori_compute::outgoing_edge_range_p): Add range_query parameter,
24121 and adjust to fur_source.
24122 * gimple-range-gori.h (class gori_compute): Simplify and adjust to
24123 range_query and fur_source.
24124 * gimple-range.cc (gimple_ranger::range_on_edge): Query range_on_edge
24125 from the ranger_cache..
24126 (gimple_ranger::fold_range_internal): Adjust to base class change of
24128 (gimple_ranger::dump_bb): Adjust dump.
24129 * gimple-range.h (gimple_ranger):export gori computes object.
24131 2021-06-01 Andrew MacLeod <amacleod@redhat.com>
24133 PR tree-optimization/100774
24134 * gimple-range-cache.cc (ranger_cache::get_non_stale_global_range):
24135 Constant values are also not stale.
24136 (ranger_cache::set_global_range): Range invariant values should also
24137 have the correct timestamp.
24139 2021-05-31 Martin Liska <mliska@suse.cz>
24141 * tree-streamer-in.c (unpack_ts_function_decl_value_fields):
24142 Unpack FUNCTION_DECL_DECL_TYPE.
24143 * tree-streamer-out.c (pack_ts_function_decl_value_fields):
24144 Stream FUNCTION_DECL_DECL_TYPE instead of
24145 DECL_IS_OPERATOR_NEW_P.
24146 * tree.h (set_function_decl_type): Use FUNCTION_DECL_DECL_TYPE
24148 (DECL_IS_OPERATOR_NEW_P): Likewise.
24149 (DECL_IS_OPERATOR_DELETE_P): Likewise.
24150 (DECL_LAMBDA_FUNCTION_P): Likewise.
24152 2021-05-31 Richard Biener <rguenther@suse.de>
24155 * internal-fn.c (expand_SHUFFLEVECTOR): Define.
24156 * internal-fn.def (SHUFFLEVECTOR): New.
24157 * internal-fn.h (expand_SHUFFLEVECTOR): Declare.
24158 * doc/extend.texi: Document __builtin_shufflevector.
24160 2021-05-31 Peter Bergner <bergner@linux.ibm.com>
24163 * config/rs6000/predicates.md(mma_assemble_input_operand): Allow
24164 indexed form addresses.
24166 2021-05-29 Jeff Law <jlaw@tachyum.com>
24168 * config/h8300/h8300.c (h8300_emit_stack_adjustment): Drop unused
24169 parameter. Call callers fixed.
24171 (output_plussi): Add FALLTHRU markers.
24172 (h8300_shift_needs_scratch_p): Add gcc_unreachable marker.
24174 2021-05-29 Jakub Jelinek <jakub@redhat.com>
24176 PR middle-end/99928
24177 * gimplify.c (gimplify_scan_omp_clauses): For taskloop simd
24178 combined with parallel, make sure to add shared clause to
24179 parallel for explicit linear clause.
24181 2021-05-29 Aldy Hernandez <aldyh@redhat.com>
24183 PR tree-optimization/100787
24184 * gimple-ssa-evrp.c: Disable exporting of global ranges.
24186 2021-05-28 Jason Merrill <jason@redhat.com>
24188 * tree-iterator.h (struct tree_stmt_iterator): Add operator++,
24189 operator--, operator*, operator==, and operator!=.
24190 (class tsi_range): New.
24192 2021-05-28 Richard Biener <rguenther@suse.de>
24194 PR tree-optimization/100778
24195 * tree-vect-slp.c (vect_build_slp_tree_1): Prevent possibly
24196 trapping ops in different BBs.
24198 2021-05-28 Richard Biener <rguenther@suse.de>
24201 * tree-inline.c (copy_bb): When processing __builtin_va_arg_pack
24202 copy fntype from original call.
24204 2021-05-28 Martin Liska <mliska@suse.cz>
24206 PR gcov-profile/100751
24207 * doc/gcov.texi: Revert partially a hunk that was wrong.
24209 2021-05-28 Cooper Qu <cooper.qu@linux.alibaba.com>
24211 * config/csky/csky-linux-elf.h (HAVE_sync_compare_and_swapqi):
24213 (HAVE_sync_compare_and_swaphi): Likewise.
24214 (HAVE_sync_compare_and_swapsi): Likewise.
24216 2021-05-28 Jakub Jelinek <jakub@redhat.com>
24218 PR middle-end/99928
24219 * tree.h (OMP_CLAUSE_MAP_IMPLICIT): Define.
24221 2021-05-28 Tobias Burnus <tobias@codesourcery.com>
24223 * gimplify.c (gimplify_omp_affinity): New.
24224 (gimplify_scan_omp_clauses): Call it; remove affinity clause afterwards.
24225 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_AFFINITY.
24226 * tree-pretty-print.c (dump_omp_clause): Handle OMP_CLAUSE_AFFINITY.
24227 * tree.c (omp_clause_num_ops, omp_clause_code_name): Add clause.
24228 (walk_tree_1): Handle OMP_CLAUSE_AFFINITY.
24230 2021-05-28 Joern Rennecke <joern.rennecke@riscy-ip.com>
24231 Richard Biener <rguenther@suse.de>
24233 * match.pd <popcount & / + pattern matching>:
24234 When generating popcount directly fails, try doing it in two halves.
24236 2021-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
24238 * Makefile.in (generated_files): Add gimple-match.c and
24241 2021-05-28 Joern Rennecke <joern.rennecke@embecosm.com>
24243 * gensupport.c (alter_predicate_for_insn): Handle MATCH_DUP.
24245 2021-05-28 Joern Rennecke <joern.rennecke@embecosm.com>
24247 * gensupport.c (alter_constraints): Add MATCH_SCRATCH case.
24249 2021-05-28 Kewen Lin <linkw@linux.ibm.com>
24251 PR tree-optimization/99398
24252 * tree-ssa-forwprop.c (simplify_permutation): Optimize some cases
24253 where the fed operands are CTOR/CST and propagated through
24254 VIEW_CONVERT_EXPR. Call vec_perm_indices::new_shrunk_vector.
24255 * vec-perm-indices.c (vec_perm_indices::new_shrunk_vector): New
24257 * vec-perm-indices.h (vec_perm_indices::new_shrunk_vector): New
24260 2021-05-27 Uroš Bizjak <ubizjak@gmail.com>
24262 * config/i386/mmx.md (addv2sf3): Do not call
24263 ix86_fixup_binary_operands_no_copy.
24266 (<smaxmin:code>v2sf3): Ditto.
24267 (<plusminus:insn><MMXMODEI:mode>3): Ditto.
24268 (<plusminus:insn><VI_32:mode>3): Remove expander.
24269 (<plusminus:insn><VI_32:mode>3): Rename from
24270 "*<plusminus:insn><VI_32:mode>3".
24271 (mulv4hi): Do not call ix86_fixup_binary_operands_no_copy.
24272 (mulv2hi3): Remove expander.
24273 (mulv2hi3): Rename from *mulv2hi3.
24274 (<s>mulv2hi3_highpart): Remove expander.
24275 (<s>mulv2hi3_highpart): Rename from *<s>mulv2hi3_highpart.
24276 (<smaxmin:code><MMXMODE14:mode>3): Rename from
24277 "*mmx_<smaxmin:code><MMXMODE14:mode>3".
24278 (<smaxmin:code><SMAXMIN_MMXMODEI:mode>3): Remove expander.
24279 (SMAXMIN_MMXMODEI): Remove mode iterator.
24280 (<smaxmin:code>v4hi3): New expander.
24281 (<smaxmin:code>v4qi3): Rename from *<smaxmin:code>v4qi3.
24282 (<smaxmin:code>v2hi3): Rename from *<smaxmin:code>v2hi3.
24283 (<smaxmin:code><SMAXMIN_VI_32:mode>3): Remove expander.
24284 (SMAXMIN_VI_32): Remove mode iterator.
24285 (<umaxmin:code><MMXMODE24:mode>3): Rename from
24286 "*mmx_<umaxmin:code><MMXMODE24:mode>3".
24287 (<umaxmin:code><UMAXMIN_MMXMODEI:mode>3): Remove expander.
24288 (UMAXMIN_MMXMODEI): Remove mode iterator.
24289 (<umaxmin:code>v8qi3): New expander.
24290 (<umaxmin:code>v4qi3): Rename from *<umaxmin:code>v4qi3.
24291 (<umaxmin:code>v2hi3): Rename from *<umaxmin:code>v2hi3.
24292 (<umaxmin:code><SMAXMIN_VI_32:mode>3): Remove expander.
24293 (UMAXMIN_VI_32): Remove mode iterator.
24294 (<any_shift:insn>v2hi3): Remove expander.
24295 (<any_shift:insn>v2hi3): Rename from *<any_shift:insn>v2hi3.
24296 (<any_logic:code><MMXMODEI:mode>3): Do not call
24297 ix86_fixup_binary_operands_no_copy.
24298 (<any_logic:code><VI_32:mode>3): Remove expander.
24299 (<any_logic:code><VI_32:mode>3): Rename from
24300 "*<any_logic:code><VI_32:mode>3".
24301 (uavg<mode>3_ceil): Do not call ix86_fixup_binary_operands_no_copy.
24302 * config/i386/sse.md (div<VF2:mode>3): Do not call
24303 ix86_fixup_binary_operands_no_copy.
24304 (div<VF1:mode>3): Ditto.
24305 (<maxmin:code><VI8_AVX2_AVX512F:mode>3): Ditto.
24306 (smulhrsv4hi3): Ditto.
24307 (smulhrsv2hi3): Ditto.
24309 2021-05-27 Martin Sebor <msebor@redhat.com>
24311 * ggc.h (gt_ggc_mx): Add overloads for all integers.
24313 * hash-map.h (class hash_map): Add pch_nx_helper overloads for all
24315 (hash_map::operator==): New function.
24317 2021-05-27 Uroš Bizjak <ubizjak@gmail.com>
24320 * config/i386/i386-expand.c (ix86_expand_int_sse_cmp):
24321 For TARGET_XOP bypass SSE comparisons for all supported vector modes.
24322 * config/i386/mmx.md (*xop_maskcmp<MMXMODEI:mode>3): New insn pattern.
24323 (*xop_maskcmp<VI_32:mode>3): Ditto.
24324 (*xop_maskcmp_uns<MMXMODEI:mode>3): Ditto.
24325 (*xop_maskcmp_uns<VI_32:mode>3): Ditto.
24327 2021-05-27 Richard Earnshaw <rearnsha@arm.com>
24330 * config/arm/arm.c (arm_configure_build_target): Remove parameter
24331 opts_set, directly check opts parameters for being non-null.
24332 (arm_option_restore): Update call to arm_configure_build_target.
24333 (arm_option_override): Likewise.
24334 (arm_can_inline_p): Likewise.
24335 (arm_valid_target_attribute_tree): Likewise.
24336 * config/arm/arm-c.c (arm_pragma_target_parse): Likewise.
24337 * config/arm/arm-protos.h (arm_configure_build_target): Adjust
24340 2021-05-27 Aldy Hernandez <aldyh@redhat.com>
24342 * vr-values.c (simplify_conversion_using_ranges): Use
24343 get_range_query instead of get_global_range_query.
24345 2021-05-27 Aldy Hernandez <aldyh@redhat.com>
24347 * gimple-range.cc (get_range_global): Move to value-query.cc.
24348 (gimple_range_global): Same.
24349 (get_global_range_query): Same.
24350 (global_range_query::range_of_expr): Same.
24351 * gimple-range.h (class global_range_query): Move to
24353 (gimple_range_global): Same.
24354 * tree-ssanames.c (get_range_info): Move to value-query.cc.
24355 (get_ptr_nonnull): Same.
24356 * tree-ssanames.h (get_range_info): Remove.
24357 (get_ptr_nonnull): Remove.
24358 * value-query.cc (get_ssa_name_range_info): Move from
24360 (get_ssa_name_ptr_info_nonnull): Same.
24361 (get_range_global): Move from gimple-range.cc.
24362 (gimple_range_global): Same.
24363 (get_global_range_query): Same.
24364 (global_range_query::range_of_expr): Same.
24365 * value-query.h (class global_range_query): Move from
24367 (gimple_range_global): Same.
24369 2021-05-27 Uroš Bizjak <ubizjak@gmail.com>
24372 * config/i386/mmx.md (uavgv4qi3_ceil): New insn pattern.
24373 (uavgv2hi3_ceil): Ditto.
24375 2021-05-26 Eric Botcazou <ebotcazou@adacore.com>
24378 * doc/extend.texi (scalar_storage_order): Rephrase slightly.
24380 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
24382 * tree-ssanames.c (get_range_info): Merge both copies of
24383 get_range_info into one that works with irange.
24384 * tree-ssanames.h (get_range_info): Remove version that works on
24387 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
24389 * builtins.c (check_nul_terminated_array): Convert to get_range_query.
24390 (expand_builtin_strnlen): Same.
24391 (determine_block_size): Same.
24392 * fold-const.c (expr_not_equal_to): Same.
24393 * gimple-fold.c (size_must_be_zero_p): Same.
24394 * gimple-match-head.c: Include gimple-range.h.
24395 * gimple-pretty-print.c (dump_ssaname_info): Convert to get_range_query.
24396 * gimple-ssa-warn-restrict.c
24397 (builtin_memref::extend_offset_range): Same.
24398 * graphite-sese-to-poly.c (add_param_constraints): Same.
24399 * internal-fn.c (get_min_precision): Same.
24400 * ipa-fnsummary.c (set_switch_stmt_execution_predicate): Same.
24401 * ipa-prop.c (ipa_compute_jump_functions_for_edge): Same.
24403 * tree-data-ref.c (split_constant_offset): Same.
24404 (dr_step_indicator): Same.
24405 * tree-dfa.c (get_ref_base_and_extent): Same.
24406 * tree-scalar-evolution.c (iv_can_overflow_p): Same.
24407 * tree-ssa-loop-niter.c (refine_value_range_using_guard): Same.
24408 (determine_value_range): Same.
24409 (record_nonwrapping_iv): Same.
24410 (infer_loop_bounds_from_signedness): Same.
24411 (scev_var_range_cant_overflow): Same.
24412 * tree-ssa-phiopt.c (two_value_replacement): Same.
24413 * tree-ssa-pre.c (insert_into_preds_of_block): Same.
24414 * tree-ssa-reassoc.c (optimize_range_tests_to_bit_test): Same.
24415 * tree-ssa-strlen.c (handle_builtin_stxncpy_strncat): Same.
24417 (dump_strlen_info): Same.
24418 (set_strlen_range): Same.
24419 (maybe_diag_stxncpy_trunc): Same.
24420 (get_len_or_size): Same.
24421 (handle_integral_assign): Same.
24422 * tree-ssa-structalias.c (find_what_p_points_to): Same.
24423 * tree-ssa-uninit.c (find_var_cmp_const): Same.
24424 * tree-switch-conversion.c (bit_test_cluster::emit): Same.
24425 * tree-vect-patterns.c (vect_get_range_info): Same.
24426 (vect_recog_divmod_pattern): Same.
24427 * tree-vrp.c (intersect_range_with_nonzero_bits): Same.
24428 (register_edge_assert_for_2): Same.
24429 (determine_value_range_1): Same.
24430 * tree.c (get_range_pos_neg): Same.
24431 * vr-values.c (vr_values::get_lattice_entry): Same.
24432 (vr_values::update_value_range): Same.
24433 (simplify_conversion_using_ranges): Same.
24435 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
24437 * gimple-ssa-warn-alloca.c (alloca_call_type): Use
24438 get_range_query instead of query argument.
24439 (pass_walloca::execute): Enable and disable global ranger.
24441 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
24443 * gimple-ssa-evrp.c (rvrp_folder::rvrp_folder): Call
24445 (rvrp_folder::~rvrp_folder): Call disable_ranger.
24446 (hybrid_folder::hybrid_folder): Call enable_ranger.
24447 (hybrid_folder::~hybrid_folder): Call disable_ranger.
24449 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
24451 * function.c (allocate_struct_function): Set cfun->x_range_query.
24452 * function.h (struct function): Declare x_range_query.
24453 (get_range_query): New.
24454 (get_global_range_query): New.
24455 * gimple-range-cache.cc (ssa_global_cache::ssa_global_cache):
24456 Remove call to safe_grow_cleared.
24457 * gimple-range.cc (get_range_global): New.
24458 (gimple_range_global): Move from gimple-range.h.
24459 (get_global_range_query): New.
24460 (global_range_query::range_of_expr): New.
24461 (enable_ranger): New.
24462 (disable_ranger): New.
24463 * gimple-range.h (gimple_range_global): Move to gimple-range.cc.
24464 (class global_range_query): New.
24465 (enable_ranger): New.
24466 (disable_ranger): New.
24467 * gimple-ssa-evrp.c (evrp_folder::~evrp_folder): Rename
24468 dump_all_value_ranges to dump.
24469 * tree-vrp.c (vrp_prop::finalize): Same.
24470 * value-query.cc (range_query::dump): New.
24471 * value-query.h (range_query::dump): New.
24472 * vr-values.c (vr_values::dump_all_value_ranges): Rename to...
24473 (vr_values::dump): ...this.
24474 * vr-values.h (class vr_values): Rename dump_all_value_ranges to
24475 dump and make virtual.
24477 2021-05-26 Uroš Bizjak <ubizjak@gmail.com>
24479 * config/i386/i386.c (ix86_autovectorize_vector_modes):
24480 Add V4QImode and V16QImode for TARGET_SSE2.
24481 * doc/sourcebuild.texi (Vector-specific attributes):
24482 Add vect64 and vect32 description.
24484 2021-05-26 Bernd Edlinger <bernd.edlinger@hotmail.de>
24486 * gimple-range-gori.cc (range_def_chain::register_dependency):
24487 Resize m_def_chain when needed.
24489 2021-05-26 Christophe Lyon <christophe.lyon@linaro.org>
24491 * config/arm/mve.md (mve_vaddvq_<supf><mode>): Prefix with '@'.
24492 * config/arm/neon.md (reduc_plus_scal_<mode>): Move to ..
24493 * config/arm/vec-common.md: .. here. Add support for MVE.
24495 2021-05-26 Jakub Jelinek <jakub@redhat.com>
24497 * config/epiphany/epiphany.c (epiphany_print_operand_address): Remove
24499 * config/microblaze/microblaze.c (microblaze_legitimize_address,
24501 microblaze_option_override, print_operand): Likewise.
24502 * config/microblaze/microblaze.md (call_internal_plt,
24503 call_value_intern_plt, call_value_intern): Likewise.
24504 * config/arm/aout.h (ASM_OUTPUT_ALIGN): Likewise.
24505 * config/iq2000/iq2000.md (call_internal1, call_value_internal1,
24506 call_value_multiple_internal1): Likewise.
24507 * config/bfin/bfin.c (symbolic_reference_mentioned_p): Likewise.
24509 2021-05-26 Jan-Benedict Glaw <jbglaw@lug-owl.de>
24511 * config/arc/arc.c (arc_address_cost, arc_print_operand_address,
24512 arc_ccfsm_advance, symbolic_reference_mentioned_p,
24513 arc_raw_symbolic_reference_mentioned_p): Remove register
24516 2021-05-26 Jakub Jelinek <jakub@redhat.com>
24519 * omp-low.c: Include omp-offload.h.
24520 (create_omp_child_function): If current_function_decl has
24521 "omp declare target" attribute and is_gimple_omp_offloaded,
24522 remove that attribute from the copy of attribute list and
24523 add "omp target entrypoint" attribute instead.
24524 (lower_omp_target): Mark .omp_data_sizes.* and .omp_data_kinds.*
24525 variables for offloading if in omp_maybe_offloaded_ctx.
24526 * omp-offload.c (pass_omp_target_link::execute): Nullify second
24527 argument to GOMP_target_data_ext in offloaded code.
24529 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
24531 * config/csky/csky.c (csky_can_change_mode_class): Delete.
24532 For csky, HF/SF mode use the low bits of VREGS.
24534 2021-05-26 Eric Botcazou <ebotcazou@adacore.com>
24536 * gimplify.c (gimplify_decl_expr): Do not clear TREE_READONLY on a
24537 DECL which is a reference for OMP.
24539 2021-05-26 Martin Liska <mliska@suse.cz>
24541 PR gcov-profile/100751
24542 * doc/gcov.texi: Document that __gcov_dump can be called just
24543 once and that __gcov_reset resets run-time counters.
24545 2021-05-26 Martin Liska <mliska@suse.cz>
24547 * doc/install.texi: Port relevant part from install-old.texi
24548 and re-generate list of CPUs and systems.
24550 2021-05-26 Martin Liska <mliska@suse.cz>
24552 * Makefile.in: Remove it.
24553 * doc/include/fdl.texi: Update next/previous chapters.
24554 * doc/install.texi: Likewise.
24555 * doc/install-old.texi: Removed.
24557 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
24559 * config/csky/csky.c (ck810_legitimate_index_p): Support
24560 "base + index" with DF mode.
24561 * config/csky/constraints.md ("Y"): New constraint for memory operands
24562 without index register.
24563 * config/csky/csky_insn_fpuv2.md (fpuv3_movdf): Use "Y" instead of "m"
24564 when mov between memory and general registers, and lower their priority.
24565 * config/csky/csky_insn_fpuv3.md (fpuv2_movdf): Likewise.
24567 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
24569 * config/csky/csky.c (TARGET_PROMOTE_PROTOTYPES): Delete.
24571 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
24573 * config/csky/csky.md (untyped_call): Emit clobber for return
24574 registers to mark them used.
24576 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
24578 * config/csky/csky.md (cskyv2_sextend_ldbs): New.
24580 2021-05-26 Andrew Pinski <apinski@marvell.com>
24582 * match.pd (x < 0 ? ~y : y): New patterns.
24584 2021-05-26 Andrew Pinski <apinski@marvell.com>
24586 * match.pd (A?CST1:CST2): Add simplifcations for A?0:+-1, A?+-1:0,
24587 A?POW2:0 and A?0:POW2.
24589 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
24591 * gimple-range-gori.cc (class logical_stmt_cache): Delete
24592 (logical_stmt_cache::logical_stmt_cache ): Delete.
24593 (logical_stmt_cache::~logical_stmt_cache): Delete.
24594 (logical_stmt_cache::cache_entry::dump): Delete.
24595 (logical_stmt_cache::get_range): Delete.
24596 (logical_stmt_cache::cached_name ): Delete.
24597 (logical_stmt_cache::same_cached_name): Delete.
24598 (logical_stmt_cache::cacheable_p): Delete.
24599 (logical_stmt_cache::slot_diagnostics ): Delete.
24600 (logical_stmt_cache::dump): Delete.
24601 (gori_compute_cache::gori_compute_cache): Delete.
24602 (gori_compute_cache::~gori_compute_cache): Delete.
24603 (gori_compute_cache::compute_operand_range): Delete.
24604 (gori_compute_cache::cache_stmt): Delete.
24605 * gimple-range-gori.h (gori_compute::compute_operand_range): Remove
24607 (class gori_compute_cache): Delete.
24609 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
24611 * gimple-range.cc (fold_using_range::range_of_range_op): Use m_gori
24613 (fold_using_range::range_of_address): Adjust.
24614 (fold_using_range::range_of_phi): Adjust.
24615 * gimple-range.h (class fur_source): Adjust.
24616 (fur_source::fur_source): Adjust.
24618 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
24620 * gimple-range-gori.cc (gori_compute::expr_range_at_stmt): Rename
24621 from expr_range_in_bb and adjust.
24622 (gori_compute::compute_name_range_op): Adjust.
24623 (gori_compute::optimize_logical_operands): Adjust.
24624 (gori_compute::compute_logical_operands_in_chain): Adjust.
24625 (gori_compute::compute_operand1_range): Adjust.
24626 (gori_compute::compute_operand2_range): Adjust.
24627 (ori_compute_cache::cache_stmt): Adjust.
24628 * gimple-range-gori.h (gori_compute): Rename prototype.
24630 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
24632 * gimple-range.cc (gimple_ranger::range_of_expr): Non-null should be
24633 checked only after range_of_stmt, not range_on_entry.
24634 (gimple_ranger::range_on_entry): Check for non-null in any
24635 predecessor block, if it is not already non-null.
24636 (gimple_ranger::range_on_exit): DOnt check for non-null after
24637 range on entry call.
24638 (gimple_ranger::dump_bb): New. Split from dump.
24639 (gimple_ranger::dump): Adjust.
24640 * gimple-range.h (class gimple_ranger): Adjust.
24642 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
24644 * gimple-range-cache.cc (struct range_timestamp): Delete.
24645 (class temporal_cache): Adjust.
24646 (temporal_cache::get_timestamp): Delete.
24647 (temporal_cache::set_dependency): Delete.
24648 (temporal_cache::temporal_value): Adjust.
24649 (temporal_cache::current_p): Take dependencies as params.
24650 (temporal_cache::set_timestamp): Adjust.
24651 (temporal_cache::set_always_current): Adjust.
24652 (ranger_cache::get_non_stale_global_range): Adjust.
24653 (ranger_cache::register_dependency): Delete.
24654 * gimple-range-cache.h (class range_cache): Adjust.
24656 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
24658 * gimple-range-gori.cc (range_def_chain::range_def_chain): init
24660 (range_def_chain::~range_def_chain): Dispose of obstack rather than
24661 each individual bitmap.
24662 (range_def_chain::set_import): New.
24663 (range_def_chain::get_imports): New.
24664 (range_def_chain::chain_import_p): New.
24665 (range_def_chain::register_dependency): Rename from build_def_chain
24667 (range_def_chain::def_chain_in_bitmap_p): New.
24668 (range_def_chain::add_def_chain_to_bitmap): New.
24669 (range_def_chain::has_def_chain): Just check first depenedence.
24670 (range_def_chain::get_def_chain): Process imports, use generic
24671 register_dependency routine.
24672 (range_def_chain::dump): New.
24673 (gori_map::gori_map): Allocate import list.
24674 (gori_map::~gori_map): Release imports.
24675 (gori_map::exports): Check for past allocated block size.
24676 (gori_map::imports): New.
24677 (gori_map::def_chain_in_export_p): Delete.
24678 (gori_map::is_import_p): New.
24679 (gori_map::maybe_add_gori): Handle imports.
24680 (gori_map::dump): Adjust output, add imports.
24681 (gori_compute::has_edge_range_p): Remove def_chain_in_export call.
24682 (gori_export_iterator::gori_export_iterator): New.
24683 (gori_export_iterator::next): New.
24684 (gori_export_iterator::get_name): New.
24685 * gimple-range-gori.h (range_def_chain): Add imports and direct
24686 dependecies via struct rdc.
24687 (range_def_chain::depend1): New.
24688 (range_def_chain::depend2): New.
24689 (class gori_map): Adjust.
24690 (FOR_EACH_GORI_IMPORT_NAME): New.
24691 (FOR_EACH_GORI_EXPORT_NAME): New.
24692 (class gori_export_iterator): New.
24694 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
24696 * gimple-range-cache.cc (ranger_cache::ranger_cache): Move initial
24697 export cache filling to here.
24698 * gimple-range-gori.cc (gori_compute::gori_compute) : From Here.
24700 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
24702 * gimple-range-gori.cc (range_def_chain): Move to gimple-range-gori.h.
24703 (gori_map): Move to gimple-range-gori.h.
24704 (gori_compute::gori_compute): Adjust.
24705 (gori_compute::~gori_compute): Delete.
24706 (gori_compute::compute_operand_range_switch): Adjust.
24707 (gori_compute::compute_operand_range): Adjust.
24708 (gori_compute::compute_logical_operands): Adjust.
24709 (gori_compute::has_edge_range_p ): Adjust.
24710 (gori_compute::set_range_invariant): Delete.
24711 (gori_compute::dump): Adjust.
24712 (gori_compute::outgoing_edge_range_p): Adjust.
24713 * gimple-range-gori.h (class range_def_chain): Relocate here.
24714 (class gori_map): Relocate here.
24715 (class gori_compute): Inherit from gori_map, and adjust.
24717 2021-05-25 Aldy Hernandez <aldyh@redhat.com>
24719 * value-range.cc (range_tests_legacy): Use
24720 build_nonstandard_integer_type instead of int and short.
24722 2021-05-25 Eric Botcazou <ebotcazou@adacore.com>
24724 * gimplify.c (gimplify_decl_expr): Clear TREE_READONLY on the DECL
24725 when really creating an initialization statement for it.
24727 2021-05-25 Eric Botcazou <ebotcazou@adacore.com>
24729 * tree-inline.c (setup_one_parameter): Fix thinko in new condition.
24731 2021-05-25 Kito Cheng <kito.cheng@sifive.com>
24733 * config/riscv/riscv.h (ASM_SPEC): Pass -mno-relax.
24735 2021-05-25 Martin Liska <mliska@suse.cz>
24737 PR tree-optimization/92860
24739 * optc-save-gen.awk: Remove exceptions.
24741 2021-05-25 Martin Liska <mliska@suse.cz>
24743 * asan.h (sanitize_coverage_p): New function.
24744 * doc/extend.texi: Document it.
24745 * fold-const.c (fold_range_test): Use sanitize_flags_p
24746 instead of flag_sanitize_coverage.
24747 (fold_truth_andor): Likewise.
24748 * sancov.c: Likewise.
24749 * tree-ssa-ifcombine.c (ifcombine_ifandif): Likewise.
24750 * ipa-inline.c (sanitize_attrs_match_for_inline_p): Handle
24751 -fsanitize-coverage when inlining.
24753 2021-05-25 Cooper Qu <cooper.qu@linux.alibaba.com>
24755 * config/csky/csky-modes.def : Fix copyright.
24757 2021-05-25 Cooper Qu <cooper.qu@linux.alibaba.com>
24759 * config/csky/csky-modes.def : Amend copyright.
24760 * config/csky/csky_insn_fpuv2.md : Likewise.
24761 * config/csky/csky_insn_fpuv3.md : Likewise.
24763 2021-05-25 Richard Biener <rguenther@suse.de>
24765 PR middle-end/100727
24766 * calls.c (initialize_argument_information): Explicitely test
24767 for WITH_SIZE_EXPR.
24768 * gimple-expr.c (mark_addressable): Skip outer WITH_SIZE_EXPR.
24770 2021-05-25 Geng Qi <gengqi@linux.alibaba.com>
24772 * config/csky/csky.h (FRAME_POINTER_REGNUM): Use
24773 HARD_FRAME_POINTER_REGNUM and FRAME_POINTER_REGNUM instead of
24774 the signle definition. The signle definition may not work well
24775 at simplify_subreg_regno().
24776 (HARD_FRAME_POINTER_REGNUM): New.
24777 (ELIMINABLE_REGS): Add for HARD_FRAME_POINTER_REGNUM.
24778 * config/csky/csky.c (get_csky_live_regs, csky_can_eliminate,
24779 csky_initial_elimination_offset, csky_expand_prologue,
24780 csky_expand_epilogue): Add for HARD_FRAME_POINTER_REGNUM.
24782 2021-05-25 Geng Qi <gengqi@linux.alibaba.com>
24784 * config/csky/csky.c (csky_option_override):
24785 Init csky_arch_isa_features[] in advance, so TARGET_DSP
24786 and TARGET_DIV can be set well.
24788 2021-05-25 Geng Qi <gengqi@linux.alibaba.com>
24790 * config/csky/constraints.md ("l", "h"): Delete.
24791 * config/csky/csky.h (reg_class, REG_CLASS_NAMES,
24792 REG_CLASS_CONTENTS): Delete LO_REGS and HI_REGS.
24793 * config/csky/csky.c (regno_reg_classm,
24794 csky_secondary_reload, csky_register_move_cost):
24795 Use HILO_REGS instead of LO_REGS and HI_REGS.
24797 2021-05-25 Geng Qi <gengqi@linux.alibaba.com>
24799 * config/csky/constraints.md ("W"): New constriant for mem operand
24800 with base reg, index register.
24801 ("Q"): Renamed and modified "csky_valid_fpuv2_mem_operand" to
24802 "csky_valid_mem_constraint_operand" to deal with both "Q" and "W"
24804 ("Dv"): New constraint for const double value that can be used at
24806 * config/csky/csky-modes.def (HFmode): New mode.
24807 * config/csky/csky-protos.h (csky_valid_fpuv2_mem_operand): Rename
24808 to "csky_valid_mem_constraint_operand" and support new constraint
24810 (csky_get_movedouble_length): New.
24811 (fpuv3_output_move): New.
24812 (fpuv3_const_double): New.
24813 * config/csky/csky.c (csky_option_override): New arch CK860 with fpv3.
24814 (decompose_csky_address): Refine.
24815 (csky_print_operand): New "CONST_DOUBLE" operand.
24816 (csky_output_move): Support fpv3 instructions.
24817 (csky_get_movedouble_length): New.
24818 (fpuv3_output_move): New.
24819 (fpuv3_const_double): New.
24820 (csky_emit_compare): Cover float comparsion.
24821 (csky_emit_compare_float): Refine.
24822 (csky_vaild_fpuv2_mem_operand): Rename to
24823 "csky_valid_mem_constraint_operand" and support new constraint "W".
24824 (ck860_rtx_costs): New.
24825 (csky_rtx_costs): Add the cost calculation of CK860.
24826 (regno_reg_class): New vregs for fpuv3.
24827 (csky_dbx_regno): Likewise.
24828 (csky_cpu_cpp_builtins): New builtin macro for fpuv3.
24829 (csky_conditional_register_usage): Suporrot fpuv3.
24830 (csky_dwarf_register_span): Suporrot fpuv3.
24831 (csky_init_builtins, csky_mangle_type): Support "__fp16" type.
24832 (ck810_legitimate_index_p): Support fp16.
24833 * config/csky/csky.h (TARGET_TLS): ADD CK860.
24834 (CSKY_VREG_P, CSKY_VREG_LO_P, CSKY_VREG_HI_P): Support fpuv3.
24835 (TARGET_SINGLE_FPU): Support fpuv3.
24836 (TARGET_SUPPORT_FPV3): New.
24837 (FIRST_PSEUDO_REGISTER): Change to 202 to hold the new fpuv3 registers.
24838 (FIXED_REGISTERS, CALL_REALLY_USED_REGISTERS, REGISTER_NAMES,
24839 REG_CLASS_CONTENTS): Support fpuv3.
24840 * config/csky/csky.md (movsf): Move to cksy_insn_fpu.md and refine.
24841 (csky_movsf_fpv2): Likewise.
24842 (ck801_movsf): Likewise.
24843 (csky_movsf): Likewise.
24845 (csky_movdf_fpv2): Likewise.
24846 (ck801_movdf): Likewise.
24847 (csky_movdf): Likewise.
24848 (movsicc): Refine. Use "comparison_operatior" instead of
24849 "ordered_comparison_operatior".
24850 (addsicc): Likewise.
24851 (CSKY_FIRST_VFP3_REGNUM, CSKY_LAST_VFP3_REGNUM): New constant.
24852 (call_value_internal_vh): New.
24853 * config/csky/csky_cores.def (CK860): New arch and cpu.
24858 * config/csky/csky_insn_fpu.md: Refactor. Separate all float patterns
24859 into emit-patterns and match-patterns, remain the emit-patterns here,
24860 and move the match-patterns to csky_insn_fpuv2.md or
24861 csky_insn_fpuv3.md.
24862 * config/csky/csky_insn_fpuv2.md: New file for fpuv2 instructions.
24863 * config/csky/csky_insn_fpuv3.md: New file and new patterns for fpuv3
24865 * config/csky/csky_isa.def (fcr): New.
24870 (CK860): New definition for ck860.
24871 * config/csky/csky_tables.opt (ck860): New processors ck860,
24872 ck860f. And new arch ck860.
24877 * config/csky/predicates.md (csky_float_comparsion_operator): Delete
24878 "geu", "gtu", "leu", "ltu", which will never appear at float comparison.
24879 * config/csky/t-csky-elf: Support 860.
24880 * config/csky/t-csky-linux: Likewise.
24881 * doc/md.texi: Add "Q" and "W" constraints for C-SKY.
24883 2021-05-24 Aaron Sawdey <acsawdey@linux.ibm.com>
24885 * config/rs6000/genfusion.pl (gen_logical_addsubf): Refactor to
24886 add generation of logical-add and add-logical fusion pairs.
24887 * config/rs6000/rs6000-cpus.def: Add new fusion to ISA 3.1 mask
24889 * config/rs6000/rs6000.c (rs6000_option_override_internal): Turn on
24890 logical-add and add-logical fusion by default.
24891 * config/rs6000/rs6000.opt: Add -mpower10-fusion-logical-add and
24892 -mpower10-fusion-add-logical options.
24893 * config/rs6000/fusion.md: Regenerate file.
24895 2021-05-24 Aldy Hernandez <aldyh@redhat.com>
24897 * value-range.cc (irange::legacy_equal_p): Check type when
24898 comparing VR_VARYING types.
24899 (range_tests_legacy): Test comparing VARYING ranges of different
24902 2021-05-24 Wilco Dijkstra <wdijkstr@arm.com>
24904 * config/aarch64/aarch64.c (neoversen1_tunings):
24905 Enable AARCH64_EXTRA_TUNE_CHEAP_SHIFT_EXTEND.
24907 2021-05-24 Wilco Dijkstra <wdijkstr@arm.com>
24909 * config/aarch64/aarch64.c (aarch64_classify_symbol): Use GOT for
24910 extern weak symbols. Limit symbol offsets for non-GOT symbols with
24913 2021-05-24 Christophe Lyon <christophe.lyon@linaro.org>
24915 * config/arm/neon.md (vec_load_lanesxi<mode>)
24916 (vec_store_lanexoi<mode>): Move ...
24917 * config/arm/vec-common.md: here.
24919 2021-05-24 Christophe Lyon <christophe.lyon@linaro.org>
24921 * config/arm/neon.md (vec_load_lanesoi<mode>)
24922 (vec_store_lanesoi<mode>): Move ...
24923 * config/arm/vec-common.md: here.
24925 2021-05-24 liuhongt <hongtao.liu@intel.com>
24928 * config/i386/i386.c (ix86_gimple_fold_builtin): Replacing
24929 stmt with GIMPLE_NOP when lhs doesn't exist.
24931 2021-05-23 Uroš Bizjak <ubizjak@gmail.com>
24934 * config/i386/mmx.md (*push<VI_32:mode>2_rex64):
24935 New instruction pattern.
24936 (*push<VI_32:mode>2): Ditto.
24937 (push splitter for SSE registers): New splitter.
24939 2021-05-23 Andrew Pinski <apinski@marvell.com>
24941 * match.pd ((A & C) != 0 ? D : 0): Limit to non pointer types.
24943 2021-05-22 Aaron Sawdey <acsawdey@linux.ibm.com>
24945 * config/rs6000/genfusion.pl (gen_addadd): Fix incorrect attr types.
24946 * config/rs6000/fusion.md: Regenerate file.
24948 2021-05-21 Aaron Sawdey <acsawdey@linux.ibm.com>
24950 * config/rs6000/genfusion.pl (gen_addadd): New function.
24951 * config/rs6000/fusion.md: Regenerate file.
24952 * config/rs6000/rs6000-cpus.def: Add
24953 OPTION_MASK_P10_FUSION_2ADD to masks.
24954 * config/rs6000/rs6000.c (rs6000_option_override_internal):
24955 Handle default value of OPTION_MASK_P10_FUSION_2ADD.
24956 * config/rs6000/rs6000.opt: Add -mpower10-fusion-2add.
24958 2021-05-21 Jakub Jelinek <jakub@redhat.com>
24960 PR middle-end/99928
24961 * tree.h (OMP_CLAUSE_FIRSTPRIVATE_IMPLICIT_TARGET): Define.
24962 * gimplify.c (enum gimplify_omp_var_data): Fix up
24963 GOVD_MAP_HAS_ATTACHMENTS value, add GOVD_FIRSTPRIVATE_IMPLICIT.
24964 (omp_lastprivate_for_combined_outer_constructs): If combined target
24965 has GOVD_FIRSTPRIVATE_IMPLICIT set for the decl, change it to
24966 GOVD_MAP | GOVD_SEEN.
24967 (gimplify_scan_omp_clauses): Set GOVD_FIRSTPRIVATE_IMPLICIT for
24968 firstprivate clauses with OMP_CLAUSE_FIRSTPRIVATE_IMPLICIT.
24969 (gimplify_adjust_omp_clauses): For firstprivate clauses with
24970 OMP_CLAUSE_FIRSTPRIVATE_IMPLICIT either clear that bit and
24971 OMP_CLAUSE_FIRSTPRIVATE_IMPLICIT_TARGET too, or remove it and
24972 let it be replaced by implicit map clause.
24974 2021-05-21 Jakub Jelinek <jakub@redhat.com>
24976 PR middle-end/99928
24977 * gimplify.c (omp_lastprivate_for_combined_outer_constructs): New
24979 (gimplify_scan_omp_clauses) <case OMP_CLAUSE_LASTPRIVATE>: Use it.
24980 (gimplify_omp_for): Likewise.
24982 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
24984 PR middle-end/90115
24985 * omp-low.c (oacc_privatization_candidate_p): Reject 'static',
24986 'external' in blocks.
24988 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
24990 PR middle-end/90115
24991 * flag-types.h (enum openacc_privatization): New.
24992 * params.opt (-param=openacc-privatization): New.
24993 * doc/invoke.texi (openacc-privatization): Document it.
24994 * omp-general.h (get_openacc_privatization_dump_flags): New
24996 * omp-low.c (oacc_privatization_candidate_p): Add diagnostics.
24997 * omp-offload.c (execute_oacc_device_lower)
24998 <IFN_UNIQUE_OACC_PRIVATE>: Re-work diagnostics.
24999 * target.def (goacc.adjust_private_decl): Add 'location_t'
25001 * doc/tm.texi: Regenerate.
25002 * config/gcn/gcn-protos.h (gcn_goacc_adjust_private_decl): Adjust.
25003 * config/gcn/gcn-tree.c (gcn_goacc_adjust_private_decl): Likewise.
25004 * config/nvptx/nvptx.c (nvptx_goacc_adjust_private_decl):
25005 Likewise. Preserve it for...
25006 (nvptx_goacc_expand_var_decl): ... use here.
25008 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
25010 * doc/sourcebuild.texi (Other attributes): Document '__OPTIMIZE__'
25013 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
25015 PR middle-end/90115
25016 * omp-low.c (oacc_privatization_candidate_p): New function.
25017 (oacc_privatization_scan_clause_chain)
25018 (oacc_privatization_scan_decl_chain): Use it. Also
25019 'gcc_checking_assert' that we're not seeing duplicates.
25021 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
25023 PR middle-end/90115
25024 * omp-offload.c (execute_oacc_device_lower): Skip processing if no
25027 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
25029 PR middle-end/90115
25030 * omp-offload.c (execute_oacc_device_lower): Explain.
25032 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
25034 PR middle-end/90115
25035 * omp-offload.c (execute_oacc_device_lower)
25036 <IFN_UNIQUE_OACC_PRIVATE>: Diagnose and handle for 'level == -1'
25038 * internal-fn.c (expand_UNIQUE): Don't expect
25039 'IFN_UNIQUE_OACC_PRIVATE'.
25041 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
25043 PR middle-end/90115
25044 * omp-low.c (lower_omp_for): Don't evaluate OpenMP 'for' clauses.
25046 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
25048 PR middle-end/90115
25049 * config/nvptx/nvptx.c (nvptx_goacc_adjust_private_decl)
25050 (nvptx_goacc_expand_var_decl): Tighten.
25052 2021-05-21 Julian Brown <julian@codesourcery.com>
25053 Chung-Lin Tang <cltang@codesourcery.com>
25054 Thomas Schwinge <thomas@codesourcery.com>
25056 PR middle-end/90115
25057 * doc/tm.texi.in (TARGET_GOACC_EXPAND_VAR_DECL)
25058 (TARGET_GOACC_ADJUST_PRIVATE_DECL): Add documentation hooks.
25059 * doc/tm.texi: Regenerate.
25060 * expr.c (expand_expr_real_1): Expand decls using the
25061 expand_var_decl OpenACC hook if defined.
25062 * internal-fn.c (expand_UNIQUE): Handle IFN_UNIQUE_OACC_PRIVATE.
25063 * internal-fn.h (IFN_UNIQUE_CODES): Add OACC_PRIVATE.
25064 * omp-low.c (omp_context): Add oacc_privatization_candidates
25066 (lower_oacc_reductions): Add PRIVATE_MARKER parameter. Insert
25068 (lower_oacc_head_tail): Add PRIVATE_MARKER parameter. Modify
25069 private marker's gimple call arguments, and pass it to
25070 lower_oacc_reductions.
25071 (oacc_privatization_scan_clause_chain)
25072 (oacc_privatization_scan_decl_chain, lower_oacc_private_marker):
25074 (lower_omp_for, lower_omp_target, lower_omp_1): Use these.
25075 * omp-offload.c (convert.h): Include.
25076 (oacc_loop_xform_head_tail): Treat private-variable markers like
25077 fork/join when transforming head/tail sequences.
25078 (struct var_decl_rewrite_info): Add struct.
25079 (oacc_rewrite_var_decl, is_sync_builtin_call): New functions.
25080 (execute_oacc_device_lower): Support rewriting gang-private
25081 variables using target hook, and fix up addr_expr and var_decl
25083 * target.def (adjust_private_decl, expand_var_decl): New hooks.
25084 * config/gcn/gcn-protos.h (gcn_goacc_adjust_gangprivate_decl):
25086 (gcn_goacc_adjust_private_decl): ...this.
25087 * config/gcn/gcn-tree.c (gcn_goacc_adjust_gangprivate_decl):
25089 (gcn_goacc_adjust_private_decl): ...this. Add LEVEL parameter.
25090 * config/gcn/gcn.c (TARGET_GOACC_ADJUST_GANGPRIVATE_DECL): Rename
25091 definition using gcn_goacc_adjust_gangprivate_decl...
25092 (TARGET_GOACC_ADJUST_PRIVATE_DECL): ...to this, using
25093 gcn_goacc_adjust_private_decl.
25094 * config/nvptx/nvptx.c (tree-pretty-print.h): Include.
25095 (gang_private_shared_size): New global variable.
25096 (gang_private_shared_align): Likewise.
25097 (gang_private_shared_sym): Likewise.
25098 (gang_private_shared_hmap): Likewise.
25099 (nvptx_option_override): Initialize these.
25100 (nvptx_file_end): Output gang_private_shared_sym.
25101 (nvptx_goacc_adjust_private_decl, nvptx_goacc_expand_var_decl):
25103 (nvptx_set_current_function): Clear gang_private_shared_hmap.
25104 (TARGET_GOACC_ADJUST_PRIVATE_DECL): Define hook.
25105 (TARGET_GOACC_EXPAND_VAR_DECL): Likewise.
25107 2021-05-21 H.J. Lu <hjl.tools@gmail.com>
25109 * config/i386/i386-modes.def (MAX_BITSIZE_MODE_ANY_INT): Removed.
25111 2021-05-21 Richard Biener <rguenther@suse.de>
25112 H.J. Lu <hjl.tools@gmail.com>
25114 PR middle-end/90773
25115 * expr.c (expand_constructor): Elide expand_constructor if
25116 move by pieces is preferred.
25118 2021-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
25120 * config/aarch64/aarch64-builtins.c (aarch64_call_properties):
25121 Take a flag and mode value as arguments.
25122 (aarch64_modifies_global_state_p): Likewise.
25123 (aarch64_reads_global_state_p): Likewise.
25124 (aarch64_could_trap_p): Likewise.
25125 (aarch64_get_attributes): Likewise.
25126 (aarch64_init_simd_builtins): Adjust callsite of above.
25127 (aarch64_init_fcmla_laneq_builtins): Use aarch64_get_attributes to get
25128 function attributes to apply to builtins.
25129 (aarch64_init_crc32_builtins): Likewise.
25130 (aarch64_init_builtin_rsqrt): Likewise.
25132 2021-05-21 Aaron Sawdey <acsawdey@linux.ibm.com>
25134 * config/rs6000/rs6000.md (define_attr "type"): Add types for fusion.
25135 * config/rs6000/genfusion.pl (gen_ld_cmpi_p10): Use new fusion types.
25136 (gen_2logical): Use new fusion types.
25137 * config/rs6000/fusion.md: Regenerate.
25139 2021-05-21 Uroš Bizjak <ubizjak@gmail.com>
25142 * config/i386/i386-expand.c (ix86_expand_sse_movcc):
25143 Handle V4QI and V2HI modes.
25144 (ix86_expand_sse_movcc): Ditto.
25145 * config/i386/mmx.md (*<sat_plusminus:insn><VI_32:mode>3):
25146 New instruction pattern.
25147 (*eq<VI_32:mode>3): Ditto.
25148 (*gt<VI_32:mode>3): Ditto.
25149 (*xop_pcmov_<VI_32:mode>): Ditto.
25150 (mmx_pblendvb32): Ditto.
25151 (mmx_pblendvb64): Rename from mmx_pblendvb.
25152 (vec_cmp<VI_32:mode><VI_32:mode>): New expander.
25153 (vec_cmpu<VI_32:mode><VI_32:mode>): Ditto.
25154 (vcond<VI_32:mode><VI_32:mode>): Ditto.
25155 (vcondu<VI_32:mode><VI_32:mode>): Ditto.
25156 (vcond_mask_<VI_32:mode><VI_32:mode>): Ditto.
25158 2021-05-21 Jakub Jelinek <jakub@redhat.com>
25160 PR tree-optimization/94589
25161 * tree-ssa-phiopt.c (spaceship_replacement): For integral rhs1 and
25162 rhs2, treat x <= 4 equivalently to x < 5 etc. In cmp1 and cmp2 (if
25163 not the same as cmp3) treat <= the same as < and >= the same as >.
25164 Don't require that cond2_phi_edge is true edge, instead take
25165 false/true edges into account based on cmp1/cmp2 comparison kinds.
25167 2021-05-21 Uroš Bizjak <ubizjak@gmail.com>
25170 * config/i386/mmx.md (SMAXMIN_MMXMODEI): New mode iterator.
25171 (<smaxmin:code><SMAXMIN_MMXMODEI:mode>3): Macroize expander
25172 from <smaxmin:code>v4hi3> and <smaxmin:code><MMXMODE14:mode>3
25173 using SMAXMIN_MMXMODEI mode iterator.
25174 (*<smaxmin:code>v4qi3): New insn pattern.
25175 (*<smaxmin:code>v2hi3): Ditto.
25176 (SMAXMIN_VI_32): New mode iterator.
25177 (<smaxmin:code><SMAXMIN_VI_32>mode3): New expander.
25178 (UMAXMIN_MMXMODEI): New mode iterator.
25179 (<umaxmin:code><UMAXMIN_MMXMODEI:mode>3): Macroize expander
25180 from <umaxmin:code>v8qi3> and <umaxmin:code><MMXMODE24:mode>3
25181 using UMAXMIN_MMXMODEI mode iterator.
25182 (*<umaxmin:code>v4qi3): New insn pattern.
25183 (*<umaxmin:code>v2hi3): Ditto.
25184 (UMAXMIN_VI_32): New mode iterator.
25185 (<umaxmin:code><UMAXMIN_VI_32>mode3): New expander.
25186 (abs<VI_32:mode>2): New insn pattern.
25187 (ssse3_abs<MMXMODEI:mode>2, abs<MMXMODEI:mode>2): Move from ...
25188 * config/i386/sse.md: ... here.
25190 2021-05-20 Clement Chigot <clement.chigot@atos.net>
25191 David Edelsohn <dje.gcc@gmail.com>
25193 * collect2.c (scan_prog_file): Issue non-fatal warning for
25196 2021-05-20 Jonathan Wakely <jwakely@redhat.com>
25198 * doc/invoke.texi (-Wno-c++11-extensions)
25199 (-Wno-c++14-extensions, -Wno-c++17-extensions)
25200 (-Wno-c++20-extensions, -Wno-c++23-extensions): Document
25203 2021-05-20 Indu Bhagat <indu.bhagat@oracle.com>
25205 * config/c6x/c6x.c (c6x_output_file_unwind): Use dwarf_debuginfo_p.
25206 * config/darwin.c (darwin_override_options): Likewise.
25207 * config/i386/cygming.h (DBX_REGISTER_NUMBER): Likewise.
25208 * config/i386/darwin.h (DBX_REGISTER_NUMBER): Likewise.
25209 (DWARF2_FRAME_REG_OUT): Likewise.
25210 * config/mips/mips.c (mips_output_filename): Likewise.
25211 * config/rs6000/rs6000.c (rs6000_xcoff_declare_function_name):
25213 (rs6000_dbx_register_number): Likewise.
25214 * dbxout.c: Include flags.h.
25215 * dwarf2cfi.c (cfi_label_required_p): Likewise.
25216 (dwarf2out_do_frame): Likewise.
25217 * except.c: Include flags.h.
25218 * final.c (dwarf2_debug_info_emitted_p): Likewise.
25219 (final_scan_insn_1): Likewise.
25220 * flags.h (dwarf_debuginfo_p): New function declaration.
25221 * opts.c (dwarf_debuginfo_p): New function definition.
25222 * targhooks.c (default_debug_unwind_info): Use dwarf_debuginfo_p.
25223 * toplev.c (process_options): Likewise.
25225 2021-05-20 Indu Bhagat <indu.bhagat@oracle.com>
25227 * common.opt: Change type to support bitmasks.
25228 * flag-types.h (enum debug_info_type): Rename enumerator constants.
25229 (NO_DEBUG): New bitmask.
25230 (DBX_DEBUG): Likewise.
25231 (DWARF2_DEBUG): Likewise.
25232 (XCOFF_DEBUG): Likewise.
25233 (VMS_DEBUG): Likewise.
25234 (VMS_AND_DWARF2_DEBUG): Likewise.
25235 * flags.h (debug_set_to_format): New function declaration.
25236 (debug_set_count): Likewise.
25237 (debug_set_names): Likewise.
25238 * opts.c (debug_type_masks): Array of bitmasks for debug formats.
25239 (debug_set_to_format): New function definition.
25240 (debug_set_count): Likewise.
25241 (debug_set_names): Likewise.
25242 (set_debug_level): Update access to debug_type_names.
25243 * toplev.c: Likewise.
25245 2021-05-20 Martin Sebor <msebor@redhat.com>
25247 PR middle-end/100684
25248 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Handle C++ lambda.
25250 2021-05-20 Uroš Bizjak <ubizjak@gmail.com>
25253 * config/i386/i386.md (isa): Remove x64_bmi.
25254 (enabled): Remove x64_bmi.
25255 * config/i386/mmx.md (mmx_andnot<MMXMODEI:mode>3):
25256 Remove general register alternative.
25257 (*andnot<VI_32:mode>3): Ditto.
25258 (*mmx_<any_logic:code><MMXMODEI:mode>3): Ditto.
25259 (*<any_logic:code><VI_32:mode>3): Ditto.
25261 2021-05-20 Kewen Lin <linkw@linux.ibm.com>
25263 * config/arm/arm.c: Include head files tree-vectorizer.h and
25266 2021-05-20 Uroš Bizjak <ubizjak@gmail.com>
25269 * config/i386/mmx.md (Yv_Yw): Revert adding V4QI and V2HI modes.
25270 (*<plusminus:insn><VI_32:mode>3): Use Yw instad of <Yv_Yw> constrint.
25271 (<s>mulv4hi3_highpart): New expander.
25272 (*<s>mulv2hi3_highpart): New insn pattern.
25273 (<s>mulv2hi3_higpart): New expander.
25274 (*<any_shift:insn>v2hi3): New insn pattern.
25275 (<any_shift:insn>v2hi3): New expander.
25276 * config/i386/sse.md (smulhrsv2hi3): New expander.
25277 (*smulhrsv2hi3): New insn pattern.
25279 2021-05-20 Kewen Lin <linkw@linux.ibm.com>
25281 * doc/invoke.texi (vect-inner-loop-cost-factor): Document new
25283 * params.opt (vect-inner-loop-cost-factor): New.
25284 * targhooks.c (default_add_stmt_cost): Replace hardcoded factor
25285 50 with LOOP_VINFO_INNER_LOOP_COST_FACTOR, include head file
25286 tree-vectorizer.h and its required ones.
25287 * config/aarch64/aarch64.c (aarch64_add_stmt_cost): Replace
25288 hardcoded factor 50 with LOOP_VINFO_INNER_LOOP_COST_FACTOR.
25289 * config/arm/arm.c (arm_add_stmt_cost): Likewise.
25290 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
25291 * config/rs6000/rs6000.c (rs6000_add_stmt_cost): Likewise.
25292 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
25294 (_loop_vec_info::_loop_vec_info): Init inner_loop_cost_factor.
25295 * tree-vectorizer.h (_loop_vec_info): Add inner_loop_cost_factor.
25296 (LOOP_VINFO_INNER_LOOP_COST_FACTOR): New macro.
25298 2021-05-20 Christophe Lyon <christophe.lyon@linaro.org>
25299 Torbjörn Svensson <torbjorn.svensson@st.com>
25302 * doc/cpp.texi (Common Predefined Macros): Document __FILE_NAME__.
25304 2021-05-20 Jakub Jelinek <jakub@redhat.com>
25306 PR middle-end/99928
25307 * gimplify.c (gimplify_scan_omp_clauses) <case OMP_CLAUSE_LINEAR>: For
25308 explicit linear clause when combined with target, make it map(tofrom:)
25309 instead of no clause or firstprivate.
25311 2021-05-20 Jakub Jelinek <jakub@redhat.com>
25313 PR tree-optimization/94589
25314 * match.pd ((X & Y) == X -> (X & ~Y) == 0): Simplify even in presence
25315 of integral conversions.
25317 2021-05-19 Andrew MacLeod <amacleod@redhat.com>
25319 * gimple-range.cc (fur_source::get_operand): New.
25320 (gimple_range_fold): Delete.
25321 (fold_using_range::fold_stmt): Move from gimple_ranger::calc_stmt.
25322 (fold_using_range::range_of_range_op): Move from gimple_ranger.
25323 (fold_using_range::range_of_address): Ditto.
25324 (fold_using_range::range_of_phi): Ditto.
25325 (fold_using_range::range_of_call): Ditto.
25326 (fold_using_range::range_of_builtin_ubsan_call): Move from
25327 range_of_builtin_ubsan_call.
25328 (fold_using_range::range_of_builtin_call): Move from
25329 range_of_builtin_call.
25330 (gimple_ranger::range_of_builtin_call): Delete.
25331 (fold_using_range::range_of_cond_expr): Move from gimple_ranger.
25332 (gimple_ranger::fold_range_internal): New.
25333 (gimple_ranger::range_of_stmt): Use new fold_using_range API.
25334 (fold_using_range::range_of_ssa_name_with_loop_info): Move from
25335 gimple_ranger. Improve ranges of SSA_NAMES when possible.
25336 * gimple-range.h (gimple_ranger): Remove various range_of routines.
25337 (class fur_source): New.
25338 (class fold_using_range): New.
25339 (fur_source::fur_source): New.
25341 * vr-values.c (vr_values::extract_range_basic): Use fold_using_range
25342 instead of range_of_builtin_call.
25344 2021-05-19 Jonathan Wakely <jwakely@redhat.com>
25346 * doc/cpp.texi (Common Predefined Macros): Update documentation
25347 for the __GXX_EXPERIMENTAL_CXX0X__ macro.
25349 2021-05-19 Alex Coplan <alex.coplan@arm.com>
25352 * config/arm/arm.md (nonsecure_call_internal): Always ensure
25353 callee's address is in a register.
25355 2021-05-19 Geng Qi <gengqi@linux.alibaba.com>
25357 * common/config/riscv/riscv-common.c
25358 (riscv_subset_list::parsing_subset_version): Properly parse the letter
25360 (riscv_subset_list::parse_std_ext,
25361 riscv_subset_list::parse_multiletter_ext): To handle errors generated
25362 in riscv_subset_list::parsing_subset_version.
25364 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
25366 * config/aarch64/aarch64-simd.md: Use "neon_move_narrow_q"
25367 type attribute in patterns generating XTN(2).
25369 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
25371 * config/aarch64/aarch64-simd.md (aarch64_simd_vec_pack_trunc_<mode>):
25372 Remove as duplicate of...
25373 (aarch64_xtn<mode>): This.
25374 (aarch64_xtn2<mode>_le): Move position in file.
25375 (aarch64_xtn2<mode>_be): Move position in file.
25376 (aarch64_xtn2<mode>): Move position in file.
25377 (vec_pack_trunc_<mode>): Define as an expander.
25379 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
25381 * config/aarch64/aarch64-simd-builtins.def: Split builtin
25382 generation for aarch64_<sur>q<r>shr<u>n_n<mode> pattern into
25383 separate scalar and vector generators.
25384 * config/aarch64/aarch64-simd.md
25385 (aarch64_<sur>q<r>shr<u>n_n<mode>): Define as an expander and
25387 (aarch64_<sur>q<r>shr<u>n_n<mode>_insn_le): This and...
25388 (aarch64_<sur>q<r>shr<u>n_n<mode>_insn_be): This.
25389 * config/aarch64/iterators.md: Define SD_HSDI iterator.
25391 2021-05-19 Jonathn Wright <jonathan.wright@arm.com>
25393 * config/aarch64/aarch64-simd.md: Use UNSPEC_SQXTUN instead
25395 * config/aarch64/iterators.md: Remove UNSPEC_SQXTUN2.
25397 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
25399 * config/aarch64/aarch64-simd.md (aarch64_<sur>q<r>shr<u>n2_n<mode>):
25400 Implement as an expand emitting a big/little endian
25401 instruction pattern.
25402 (aarch64_<sur>q<r>shr<u>n2_n<mode>_insn_le): Define.
25403 (aarch64_<sur>q<r>shr<u>n2_n<mode>_insn_be): Define.
25405 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
25407 * config/aarch64/aarch64-simd.md (aarch64_<sur><addsub>hn2<mode>):
25408 Implement as an expand emitting a big/little endian
25409 instruction pattern.
25410 (aarch64_<sur><addsub>hn2<mode>_insn_le): Define.
25411 (aarch64_<sur><addsub>hn2<mode>_insn_be): Define.
25412 * config/aarch64/iterators.md: Remove UNSPEC_[R]ADDHN2 and
25413 UNSPEC_[R]SUBHN2 unspecs and ADDSUBHN2 iterator.
25415 2021-05-19 Richard Biener <rguenther@suse.de>
25417 PR middle-end/100672
25418 * fold-const.c (fold_negate_expr_1): Use element_precision.
25419 (negate_expr_p): Likewise.
25421 2021-05-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
25423 * config/aarch64/iterators.md (SVE_PRED_LOAD): New iterator.
25424 (pred_load): New int attribute.
25425 * config/aarch64/aarch64-sve.md
25426 (aarch64_load_<ANY_EXTEND:optab><SVE_HSDI:mode><SVE_PARTIAL_I:mode>): Use
25427 SVE_PRED_LOAD enum iterator and corresponding pred_load attribute.
25428 * config/aarch64/aarch64-sve-builtins-base.cc (expand): Update call to
25429 code_for_aarch64_load.
25431 2021-05-19 Richard Biener <rguenther@suse.de>
25433 * cfgexpand.c (discover_nonconstant_array_refs_r): Make
25434 sure TARGET_MEM_REF bases are expanded as memory.
25435 * tree-ssa-operands.c (operands_scanner::get_tmr_operands):
25436 Do not mark TARGET_MEM_REF bases addressable.
25437 * tree-ssa.c (non_rewritable_mem_ref_base): Handle
25438 TARGET_MEM_REF bases as never rewritable.
25439 * gimple-walk.c (walk_stmt_load_store_addr_ops): Do not
25440 walk TARGET_MEM_REF bases as address-takens.
25441 * tree-ssa-dce.c (ref_may_be_aliased): Handle TARGET_MEM_REF.
25443 2021-05-19 Richard Biener <rguenther@suse.de>
25445 * builtins.c (get_object_alignment_1): Strip outer
25447 * tree-dfa.c (get_ref_base_and_extent): Handle outer
25448 WITH_SIZE_EXPR for size processing and process the
25450 * tree-ssa-alias.c (ao_ref_base_alias_set): Strip
25451 outer WITH_SIZE_EXPR.
25452 (ao_ref_base_alias_ptr_type): Likewise.
25453 (refs_may_alias_p_2): Allow WITH_SIZE_EXPR in ref->ref
25454 and handle that accordingly, stripping it for the
25455 core alias workers.
25456 * tree.c (get_base_address): Handle WITH_SIZE_EXPR by
25457 looking through it instead of returning NULL.
25459 2021-05-19 Jakub Jelinek <jakub@redhat.com>
25461 PR middle-end/100576
25462 * builtins.c (check_read_access): Convert bound to size_type_node if
25465 2021-05-19 Richard Biener <rguenther@suse.de>
25467 * tree-cfg.c (verify_types_in_gimple_min_lval): Inline...
25468 (verify_types_in_gimple_reference): ... here. Sanitize.
25469 (verify_gimple_call): Verify references in LHS and arguments.
25470 (verify_gimple_assign_single): Reject WITH_SIZE_EXPR.
25472 2021-05-19 Uroš Bizjak <ubizjak@gmail.com>
25474 * config/i386/i386.h (VALID_INT_MODE_P):
25475 Add V8QI, V4HI and V2SI modes for TARGET_64BIT.
25476 * config/i386/i386.md (isa): Add x64_bmi.
25477 (enabled): Handle x64_bmi.
25478 * config/i386/mmx.md (mmx_andnot<MMXMODEI:mode>3):
25479 Add alternative using 64bit general registers.
25480 (*mmx_<any_logic:code><MMXMODEI:mode>3): Ditto.
25482 2021-05-19 Jakub Jelinek <jakub@redhat.com>
25484 PR middle-end/99928
25485 * tree.h (OMP_MASTER_COMBINED): Define.
25486 * gimplify.c (gimplify_scan_omp_clauses): Rewrite lastprivate
25487 handling for outer combined/composite constructs to a loop.
25488 Handle lastprivate on combined target.
25489 (gimplify_expr): Formatting fix.
25491 2021-05-19 Xionghu Luo <luoxhu@linux.ibm.com>
25493 * passes.def: Add sink_code pass before store_merging.
25494 * tree-ssa-sink.c (pass_sink_code:clone): New.
25496 2021-05-18 Bill Schmidt <wschmidt@linux.ibm.com>
25498 * config/rs6000/freebsd64.h (ADJUST_FIELD_ALIGN): Remove call to
25499 rs6000_special_adjust_field_align_p.
25500 * config/rs6000/linux64.h (ADJUST_FIELD_ALIGN): Likewise.
25501 * config/rs6000/rs6000-call.c (rs6000_function_arg_boundary):
25502 Remove ABI warning.
25503 (rs6000_function_arg): Likewise.
25504 * config/rs6000/rs6000-protos.h
25505 (rs6000_special_adjust_field_align_p): Remove prototype.
25506 * config/rs6000/rs6000.c (rs6000_special_adjust_field_align_p):
25508 * config/rs6000/sysv4.h (ADJUST_FIELD_ALIGN): Remove call to
25509 rs6000_special_adjust_field_align_p.
25511 2021-05-18 Uroš Bizjak <ubizjak@gmail.com>
25514 * config/i386/i386.h (VALID_SSE2_REG_MODE):
25515 Add V4QI and V2HI modes.
25516 (VALID_INT_MODE_P): Ditto.
25517 * config/i386/mmx.md (VI_32): New mode iterator.
25518 (mmxvecsize): Handle V4QI and V2HI.
25520 (mov<VI_32:mode>): New expander.
25521 (*mov<mode>_internal): New insn pattern.
25522 (movmisalign<VI_32:mode>): New expander.
25523 (neg<VI_32:mode>): New expander.
25524 (<plusminus:insn><VI_32:mode>3): New expander.
25525 (*<plusminus:insn><VI_32:mode>3): New insn pattern.
25526 (mulv2hi3): New expander.
25527 (*mulv2hi3): New insn pattern.
25528 (one_cmpl<VI_32:mode>2): New expander.
25529 (*andnot<VI_32:mode>3): New insn pattern.
25530 (<any_logic:code><VI_32:mode>3): New expander.
25531 (*<any_logic:code><VI_32:mode>3): New insn pattern.
25533 2021-05-18 Uroš Bizjak <ubizjak@gmail.com>
25535 * config/i386/sse.md (<any_extend:insn>v4qiv4di2):
25536 Fix a mode mismatch with operand 1.
25538 2021-05-18 Uroš Bizjak <ubizjak@gmail.com>
25541 * config/i386/i386-expand.c (split_double_mode): Return
25542 temporary register when simplify_gen_subreg fails with
25543 the high half od the paradoxical subreg.
25545 2021-05-18 Richard Biener <rguenther@suse.de>
25547 * cfgexpand.c (expand_one_var): Pass in forced_stack_var
25548 and honor it when expanding.
25549 (expand_used_vars_for_block): Pass through forced_stack_var.
25550 (expand_used_vars): Likewise.
25551 (discover_nonconstant_array_refs_r): Set bits in
25552 forced_stack_vars instead of marking vars TREE_ADDRESSABLE.
25553 (avoid_type_punning_on_regs): Likewise.
25554 (discover_nonconstant_array_refs): Likewise.
25555 (pass_expand::execute): Create and pass down forced_stack_var
25556 bitmap. For parameters and returns temporarily set
25557 TREE_ADDRESSABLE when expand_function_start.
25559 2021-05-18 Thomas Schwinge <thomas@codesourcery.com>
25561 * doc/sourcebuild.texi: Document 'dg-note'.
25563 2021-05-18 Tobias Burnus <tobias@codesourcery.com>
25566 * configure: Regenerate.
25567 * configure.ac (BUILD_CFLAG, BUILD_CXXFLAGS): Add $(CFLAGS-$@).
25569 2021-05-18 Thomas Schwinge <thomas@codesourcery.com>
25571 * gimple.h (is_gimple_omp_oacc): Tighten.
25572 * omp-low.c (check_omp_nesting_restrictions): Adjust.
25574 2021-05-18 Richard Biener <rguenther@suse.de>
25576 * tree-ssa-operands.c (mark_address_taken): Simplify.
25578 2021-05-18 Martin Liska <mliska@suse.cz>
25580 * config/gcn/mkoffload.c (STR): Redefine.
25581 * config/i386/intelmic-mkoffload.c (STR): Likewise.
25582 * config/nvptx/mkoffload.c (STR): Likewise.
25584 2021-05-18 Martin Liska <mliska@suse.cz>
25586 * common/config/aarch64/aarch64-common.c (aarch64_parse_extension):
25587 Use startswith function instead of strncmp.
25588 * common/config/bfin/bfin-common.c (bfin_handle_option): Likewise.
25589 * common/config/riscv/riscv-common.c (riscv_subset_list::parse): Likewise.
25590 * config/aarch64/aarch64-sve-builtins-shapes.cc (parse_type): Likewise.
25591 * config/aarch64/aarch64.c (aarch64_process_one_target_attr): Likewise.
25592 * config/alpha/alpha.c (alpha_elf_section_type_flags): Likewise.
25593 * config/arm/aarch-common.c (arm_md_asm_adjust): Likewise.
25594 * config/arm/arm.c (arm_file_start): Likewise.
25595 (arm_valid_target_attribute_rec): Likewise.
25596 (thumb1_md_asm_adjust): Likewise.
25597 * config/arm/driver-arm.c (host_detect_local_cpu): Likewise.
25598 * config/avr/avr.c (STR_PREFIX_P): Likewise.
25599 (avr_set_current_function): Likewise.
25600 (avr_handle_addr_attribute): Likewise.
25601 (avr_asm_output_aligned_decl_common): Likewise.
25602 (avr_asm_named_section): Likewise.
25603 (avr_section_type_flags): Likewise.
25604 (avr_asm_select_section): Likewise.
25605 * config/c6x/c6x.c (c6x_in_small_data_p): Likewise.
25606 (c6x_section_type_flags): Likewise.
25607 * config/darwin-c.c (darwin_cfstring_ref_p): Likewise.
25608 (darwin_objc_declare_unresolved_class_reference): Likewise.
25609 (darwin_objc_declare_class_definition): Likewise.
25610 * config/darwin.c (indirect_data): Likewise.
25611 (darwin_encode_section_info): Likewise.
25612 (darwin_objc2_section): Likewise.
25613 (darwin_objc1_section): Likewise.
25614 (machopic_select_section): Likewise.
25615 (darwin_globalize_label): Likewise.
25616 (darwin_label_is_anonymous_local_objc_name): Likewise.
25617 (darwin_asm_named_section): Likewise.
25618 (darwin_asm_output_dwarf_offset): Likewise.
25619 * config/frv/frv.c (frv_string_begins_with): Likewise.
25620 (frv_in_small_data_p): Likewise.
25621 * config/gcn/mkoffload.c (STR): Likewise.
25623 * config/i386/i386-builtins.c (get_builtin_code_for_version): Likewise.
25624 * config/i386/i386-options.c (ix86_option_override_internal): Likewise.
25625 * config/i386/i386.c (x86_64_elf_section_type_flags): Likewise.
25626 (ix86_md_asm_adjust): Likewise.
25627 * config/i386/intelmic-mkoffload.c (STR): Likewise.
25628 * config/i386/winnt.c (i386_pe_asm_named_section): Likewise.
25629 (i386_pe_file_end): Likewise.
25630 * config/ia64/ia64.c (ia64_in_small_data_p): Likewise.
25631 (ia64_section_type_flags): Likewise.
25632 * config/mips/driver-native.c (host_detect_local_cpu): Likewise.
25633 * config/mips/mips.c (mips_handle_interrupt_attr): Likewise.
25634 (mips16_stub_function_p): Likewise.
25635 (mips_function_rodata_section): Likewise.
25636 * config/msp430/msp430.c (msp430_mcu_name): Likewise.
25637 (msp430_function_section): Likewise.
25638 (msp430_section_type_flags): Likewise.
25639 (msp430_expand_helper): Likewise.
25640 * config/nios2/nios2.c (nios2_small_section_name_p): Likewise.
25641 (nios2_valid_target_attribute_rec): Likewise.
25642 * config/nvptx/mkoffload.c (process): Likewise.
25644 * config/pa/som.h: Likewise.
25645 * config/pdp11/pdp11.c (pdp11_output_ident): Likewise.
25646 * config/riscv/riscv.c (riscv_elf_select_rtx_section): Likewise.
25647 * config/rs6000/rs6000.c (VTABLE_NAME_P): Likewise.
25648 (rs6000_inner_target_options): Likewise.
25649 * config/s390/driver-native.c (s390_host_detect_local_cpu): Likewise.
25650 * config/sparc/driver-sparc.c (host_detect_local_cpu): Likewise.
25651 * config/vax/vax.c (vax_output_int_move): Likewise.
25652 * config/vms/vms-ld.c (startswith): Likewise.
25653 (process_args): Likewise.
25655 * config/vms/vms.c: Likewise.
25657 2021-05-18 Jakub Jelinek <jakub@redhat.com>
25659 PR rtl-optimization/100590
25660 * regcprop.c (copyprop_hardreg_forward_1): Only DCE dead sets if
25661 they are NONJUMP_INSN_P.
25663 2021-05-18 Jakub Jelinek <jakub@redhat.com>
25666 * function.c (push_dummy_function): Set DECL_ARTIFICIAL and
25667 DECL_ASSEMBLER_NAME on the fn_decl.
25669 2021-05-18 Jakub Jelinek <jakub@redhat.com>
25671 PR tree-optimization/94589
25672 * tree-ssa-phiopt.c (spaceship_replacement): Pattern match
25673 phi result used in (res & ~1) == 0 comparison as res >= 0 as
25674 res == 2 would be UB with -ffinite-math-only.
25676 2021-05-18 Martin Liska <mliska@suse.cz>
25678 * Makefile.in: genversion.o should depend on DATESTAMP.
25680 2021-05-18 Claudiu Zissulescu <claziss@synopsys.com>
25682 * config/arc/simdext.md (negv2si2): Remove round bracket.
25684 2021-05-18 Andreas Krebbel <krebbel@linux.ibm.com>
25686 * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal): Define
25687 _Bool as macro expanding to _Bool.
25689 2021-05-18 Andreas Krebbel <krebbel@linux.ibm.com>
25692 * tree.c (build_reference_type_for_mode)
25693 (build_pointer_type_for_mode): Pick pointer mode if MODE argument
25695 (build_reference_type, build_pointer_type): Invoke
25696 build_*_type_for_mode with VOIDmode.
25698 2021-05-17 Andrew MacLeod <amacleod@redhat.com>
25700 PR tree-optimization/100512
25701 * gimple-range-cache.cc (ranger_cache::set_global_range): Mark const
25702 and non-zero pointer ranges as invariant.
25703 * gimple-range.cc (gimple_ranger::range_of_stmt): Remove pointer
25704 processing from here.
25706 2021-05-17 Tom de Vries <tdevries@suse.de>
25709 * config/nvptx/nvptx-protos.h (nvptx_output_atomic_insn): Declare
25710 * config/nvptx/nvptx.c (nvptx_output_barrier)
25711 (nvptx_output_atomic_insn): New function.
25712 (nvptx_print_operand): Add support for 'B'.
25713 * config/nvptx/nvptx.md: Use nvptx_output_atomic_insn for atomic
25716 2021-05-17 Aldy Hernandez <aldyh@redhat.com>
25718 PR tree-optimization/100349
25719 * vr-values.c (bounds_of_var_in_loop): Bail if scev returns
25722 2021-05-17 Tamar Christina <tamar.christina@arm.com>
25724 * config/aarch64/driver-aarch64.c (DEFAULT_ARCH): New.
25725 (host_detect_local_cpu): Use it.
25727 2021-05-17 Martin Liska <mliska@suse.cz>
25729 * doc/invoke.texi: Add 2 missing dots.
25731 2021-05-17 Marius Hillenbrand <mhillen@linux.ibm.com>
25733 PR bootstrap/100552
25734 * configure.ac: Replace pattern substitution with call to sed.
25735 * configure: Regenerate.
25737 2021-05-17 Richard Biener <rguenther@suse.de>
25739 PR middle-end/100582
25740 * tree.c (array_at_struct_end_p): Get to the base of the
25741 reference before looking for the underlying decl.
25743 2021-05-17 Joern Rennecke <joern.rennecke@embecosm.com>
25745 * genoutput.c (validate_insn_alternatives) Make "wrong number of
25746 alternatives" message more specific, and remove assumption on where
25749 2021-05-17 Christophe Lyon <christophe.lyon@linaro.org>
25751 * config/arm/iterators.md (V16): New iterator.
25752 (VH_cvtto): New iterator.
25753 (v_cmp_result): Added V4HF and V8HF support.
25754 * config/arm/vec-common.md (vec_cmp<mode><v_cmp_result>): Use VDQWH.
25755 (vcond<mode><mode>): Likewise.
25756 (vcond_mask_<mode><v_cmp_result>): Likewise.
25757 (vcond<VH_cvtto><mode>): New expander.
25759 2021-05-17 Christophe Lyon <christophe.lyon@linaro.org>
25761 * config/arm/arm-protos.h (arm_expand_vector_compare): Update
25763 * config/arm/arm.c (arm_expand_vector_compare): Add support for
25765 (arm_expand_vcond): Likewise.
25766 * config/arm/iterators.md (supf): Remove VCMPNEQ_S, VCMPEQQ_S,
25767 VCMPEQQ_N_S, VCMPNEQ_N_S.
25768 (VCMPNEQ, VCMPEQQ, VCMPEQQ_N, VCMPNEQ_N): Remove.
25769 * config/arm/mve.md (@mve_vcmp<mve_cmp_op>q_<mode>): Add '@' prefix.
25770 (@mve_vcmp<mve_cmp_op>q_f<mode>): Likewise.
25771 (@mve_vcmp<mve_cmp_op>q_n_f<mode>): Likewise.
25772 (@mve_vpselq_<supf><mode>): Likewise.
25773 (@mve_vpselq_f<mode>"): Likewise.
25774 * config/arm/neon.md (vec_cmp<mode><v_cmp_result): Enable for MVE
25775 and move to vec-common.md.
25776 (vec_cmpu<mode><mode>): Likewise.
25777 (vcond<mode><mode>): Likewise.
25778 (vcond<V_cvtto><mode>): Likewise.
25779 (vcondu<mode><v_cmp_result>): Likewise.
25780 (vcond_mask_<mode><v_cmp_result>): Likewise.
25781 * config/arm/unspecs.md (VCMPNEQ_U, VCMPNEQ_S, VCMPEQQ_S)
25782 (VCMPEQQ_N_S, VCMPNEQ_N_S, VCMPEQQ_U, CMPEQQ_N_U, VCMPNEQ_N_U)
25783 (VCMPGEQ_N_S, VCMPGEQ_S, VCMPGTQ_N_S, VCMPGTQ_S, VCMPLEQ_N_S)
25784 (VCMPLEQ_S, VCMPLTQ_N_S, VCMPLTQ_S, VCMPCSQ_N_U, VCMPCSQ_U)
25785 (VCMPHIQ_N_U, VCMPHIQ_U): Remove.
25786 * config/arm/vec-common.md (vec_cmp<mode><v_cmp_result): Moved
25788 (vec_cmpu<mode><mode>): Likewise.
25789 (vcond<mode><mode>): Likewise.
25790 (vcond<V_cvtto><mode>): Likewise.
25791 (vcondu<mode><v_cmp_result>): Likewise.
25792 (vcond_mask_<mode><v_cmp_result>): Likewise. Added unsafe math
25795 2021-05-17 liuhongt <hongtao.liu@intel.com>
25798 * config/i386/i386.c (ix86_gimple_fold_builtin): Use
25799 gsi_insert_seq_before instead.
25801 2021-05-17 Christophe Lyon <christophe.lyon@linaro.org>
25803 * doc/sourcebuild.texi (arm_qbit_ok): Rename into...
25804 (arm_sat_ok): ...this.
25806 2021-05-17 Martin Liska <mliska@suse.cz>
25808 * lto-wrapper.c (merge_flto_options): Factor out a new function.
25809 (merge_and_complain): Use it.
25810 (run_gcc): Merge also linker command line -flto=foo argument
25813 2021-05-16 Christophe Lyon <christophe.lyon@linaro.org>
25815 * config/arm/arm.h (CPP_SPEC): Remove error message about
25816 -mlittle-endian/-mbig-endian conflict.
25818 2021-05-15 Bill Schmidt <wschmidt@linux.ibm.com>
25820 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Define
25821 __ROP_PROTECT__ if -mrop-protect is selected.
25823 2021-05-15 Bill Schmidt <wschmidt@linux.ibm.com>
25825 * config/rs6000/rs6000-internal.h (rs6000_stack): Add
25826 rop_hash_save_offset and rop_hash_size.
25827 * config/rs6000/rs6000-logue.c (rs6000_stack_info): Compute
25828 rop_hash_size and rop_hash_save_offset.
25829 (debug_stack_info): Dump rop_hash_save_offset and rop_hash_size.
25830 (rs6000_emit_prologue): Emit hashst[p] in prologue.
25831 (rs6000_emit_epilogue): Emit hashchk[p] in epilogue.
25832 * config/rs6000/rs6000.md (unspec): Add UNSPEC_HASHST and
25834 (hashst): New define_insn.
25835 (hashchk): Likewise.
25837 2021-05-15 Bill Schmidt <wschmidt@linux.ibm.com>
25839 * config/rs6000/rs6000.c (rs6000_option_override_internal):
25840 Disable shrink wrap when inserting ROP-protect instructions.
25841 * config/rs6000/rs6000.opt (mrop-protect): New option.
25842 (mprivileged): Likewise.
25843 * doc/invoke.texi: Document mrop-protect and mprivileged.
25845 2021-05-15 Hans-Peter Nilsson <hp@axis.com>
25847 * reorg.c (fill_slots_from_thread): Reinstate code typoed out in
25850 2021-05-15 Martin Jambor <mjambor@suse.cz>
25853 2021-05-13 Martin Jambor <mjambor@suse.cz>
25855 PR tree-optimization/100453
25856 * tree-sra.c (sra_modify_assign): All const base accesses do not
25857 need refreshing, not just those from decl_pool.
25858 (sra_modify_assign): Do not refresh into a const base decl.
25860 2021-05-15 Jakub Jelinek <jakub@redhat.com>
25862 PR rtl-optimization/100342
25863 * regcprop.c (copy_value): When copying a source reg in a wider
25864 mode than it has recorded for the value, adjust recorded destination
25865 mode too or punt if !REG_CAN_CHANGE_MODE_P.
25867 2021-05-14 Jason Merrill <jason@redhat.com>
25869 * intl.h: Add comments.
25871 2021-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
25873 * config/aarch64/aarch64-simd.md
25874 (aarch64_sqdml<SBINQOPS:as>l2_lane<mode>_internal): Split into...
25875 (aarch64_sqdmlsl2_lane<mode>_internal): ... This...
25876 (aarch64_sqdmlal2_lane<mode>_internal): ... And this.
25877 (aarch64_sqdml<SBINQOPS:as>l2_laneq<mode>_internal): Split into ...
25878 (aarch64_sqdmlsl2_laneq<mode>_internal): ... This...
25879 (aarch64_sqdmlal2_laneq<mode>_internal): ... And this.
25880 (aarch64_sqdml<SBINQOPS:as>l2_n<mode>_internal): Split into...
25881 (aarch64_sqdmlsl2_n<mode>_internal): ... This...
25882 (aarch64_sqdmlal2_n<mode>_internal): ... And this.
25884 2021-05-14 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
25887 * config/arm/arm_neon.h (vtst_s8): Replace call to vtst builtin with it's
25888 boolean logic equivalent.
25889 (vtst_s16): Likewise.
25890 (vtst_s32): Likewise.
25891 (vtst_u8): Likewise.
25892 (vtst_u16): Likewise.
25893 (vtst_u32): Likewise.
25894 (vtst_p8): Likewise.
25895 (vtst_p16): Likewise.
25896 (vtstq_s8): Likewise.
25897 (vtstq_s16): Likewise.
25898 (vtstq_s32): Likewise.
25899 (vtstq_u8): Likewise.
25900 (vtstq_u16): Likewise.
25901 (vtstq_u32): Likewise.
25902 (vtstq_p8): Likewise.
25903 (vtstq_p16): Likewise.
25904 * config/arm/arm_neon_builtins.def: Remove entry for vtst.
25905 * config/arm/neon.md (neon_vtst<mode>): Remove pattern.
25907 2021-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
25909 * config/aarch64/aarch64-simd.md (aarch64_sqdmlal2<mode>): Merge into...
25910 (aarch64_sqdml<SBINQOPS:as>l2<mode>): ... This.
25911 (aarch64_sqdmlsl2<mode>): Delete.
25912 (aarch64_sqdmlal2_lane<mode>): Merge this...
25913 (aarch64_sqdmlsl2_lane<mode>): ... And this...
25914 (aarch64_sqdml<SBINQOPS:as>l2_lane<mode>): ... Into this.
25915 (aarch64_sqdmlal2_laneq<mode>): Merge this...
25916 (aarch64_sqdmlsl2_laneq<mode>): ... And this...
25917 (aarch64_sqdml<SBINQOPS:as>l2_laneq<mode>): ... Into this.
25918 (aarch64_sqdmlal2_n<mode>): Merge this...
25919 (aarch64_sqdmlsl2_n<mode>): ... And this...
25920 (aarch64_sqdml<SBINQOPS:as>l2_n<mode>): ... Into this.
25922 2021-05-13 Martin Sebor <msebor@redhat.com>
25924 PR middle-end/100574
25925 * builtins.c (access_ref::get_ref): Improve detection of PHIs with
25926 all null arguments.
25928 2021-05-13 Martin Sebor <msebor@redhat.com>
25930 PR tree-optimization/93100
25931 PR middle-end/98583
25932 * tree-ssa-uninit.c (check_defs): Exclude intrinsic functions that
25933 don't modify referenced objects.
25935 2021-05-13 Martin Jambor <mjambor@suse.cz>
25937 PR tree-optimization/100453
25938 * tree-sra.c (sra_modify_assign): All const base accesses do not
25939 need refreshing, not just those from decl_pool.
25940 (sra_modify_assign): Do not refresh into a const base decl.
25942 2021-05-13 Martin Liska <mliska@suse.cz>
25944 * tree-ssa-dom.c: Remove m_simplifier.
25946 2021-05-13 Richard Earnshaw <rearnsha@arm.com>
25949 * config/arm/arm.c (arm_canonicalize_comparison): Correctly
25950 canonicalize DImode inequality comparisons against the
25951 maximum integral value.
25953 2021-05-13 Jakub Jelinek <jakub@redhat.com>
25955 PR tree-optimization/98856
25956 * config/i386/i386.c (ix86_shift_rotate_cost): Add CODE argument.
25957 Expect V2DI and V4DI arithmetic right shifts to be emulated.
25958 (ix86_rtx_costs, ix86_add_stmt_cost): Adjust ix86_shift_rotate_cost
25960 * config/i386/i386-expand.c (expand_vec_perm_2perm_interleave,
25961 expand_vec_perm_2perm_pblendv): New functions.
25962 (ix86_expand_vec_perm_const_1): Use them.
25963 * config/i386/sse.md (ashr<mode>3<mask_name>): Rename to ...
25964 (<mask_codefor>ashr<mode>3<mask_name>): ... this.
25965 (ashr<mode>3): New define_expand with VI248_AVX512BW iterator.
25966 (ashrv4di3): New define_expand.
25967 (ashrv2di3): Change condition to TARGET_SSE2, handle !TARGET_XOP
25968 and !TARGET_AVX512VL expansion.
25970 2021-05-13 Uroš Bizjak <ubizjak@gmail.com>
25973 * config/i386/i386-expand.c (ix86_expand_sse_movcc): Force mode
25974 sizes < 16 to a register when constructing vpcmov pattern.
25975 * config/i386/mmx.md (*xop_pcmov_<mode>): Use MMXMODE124 mode.
25977 2021-05-13 Martin Liska <mliska@suse.cz>
25979 * gcov-io.c (gcov_write_block): Remove.
25980 (gcov_write_words): Likewise.
25981 (gcov_read_words): Re-implement using gcov_read_bytes.
25982 (gcov_allocate): Remove.
25983 (GCOV_BLOCK_SIZE): Likewise.
25984 (struct gcov_var): Remove most of the fields.
25985 (gcov_position): Implement with ftell.
25986 (gcov_rewrite): Remove setting of start and offset fields.
25987 (from_file): Re-format.
25988 (gcov_open): Remove setbuf call. It should not be needed.
25989 (gcov_close): Remove internal buffer handling.
25990 (gcov_magic): Use __builtin_bswap32.
25991 (gcov_write_counter): Use directly gcov_write_unsigned.
25992 (gcov_write_string): Use direct fwrite and do not round
25994 (gcov_seek): Use directly fseek.
25995 (gcov_write_tag): Use gcov_write_unsigned directly.
25996 (gcov_write_length): Likewise.
25997 (gcov_write_tag_length): Likewise.
25998 (gcov_read_bytes): Use directly fread.
25999 (gcov_read_unsigned): Use gcov_read_words.
26000 (gcov_read_counter): Likewise.
26001 (gcov_read_string): Use gcov_read_bytes.
26002 * gcov-io.h (GCOV_WORD_SIZE): Adjust to reflect
26003 that size is not in bytes, but words (4B).
26004 (GCOV_TAG_FUNCTION_LENGTH): Likewise.
26005 (GCOV_TAG_ARCS_LENGTH): Likewise.
26006 (GCOV_TAG_ARCS_NUM): Likewise.
26007 (GCOV_TAG_COUNTER_LENGTH): Likewise.
26008 (GCOV_TAG_COUNTER_NUM): Likewise.
26009 (GCOV_TAG_SUMMARY_LENGTH): Likewise.
26011 2021-05-13 liuhongt <hongtao.liu@intel.com>
26014 * config/i386/sse.md (ssedoublevecmode): Add attribute for
26015 V64QI/V32HI/V16SI/V4DI.
26016 (ssehalfvecmode): Add attribute for V2DI/V2DF.
26017 (*vec_concatv4si_0): Extend to VI124_128.
26018 (*vec_concat<mode>_0): New pre-reload splitter.
26019 * config/i386/predicates.md (movq_parallel): New predicate.
26021 2021-05-13 Alexandre Oliva <oliva@adacore.com>
26023 * targhooks.c (default_zero_call_used_regs): Retry using
26024 successfully-zeroed registers as sources.
26026 2021-05-12 Tobias Burnus <tobias@codesourcery.com>
26028 * omp-low.c (finish_taskreg_scan): Use the proper detach decl.
26030 2021-05-12 Aldy Hernandez <aldyh@redhat.com>
26033 * gimple-range.cc (range_of_builtin_call): Skip out on
26034 processing __builtin_clz when varying.
26036 2021-05-12 Tom de Vries <tdevries@suse.de>
26039 * config/nvptx/nvptx-opts.h (enum ptx_version): New enum.
26040 * config/nvptx/nvptx.c (nvptx_file_start): Print .version according
26041 to ptx_version_option.
26042 * config/nvptx/nvptx.h (TARGET_PTX_6_3): Define.
26043 * config/nvptx/nvptx.md (define_insn "nvptx_shuffle<mode>")
26044 (define_insn "nvptx_vote_ballot"): Use sync variant for
26046 * config/nvptx/nvptx.opt (ptx_version): Add enum.
26047 (mptx): Add option.
26048 * doc/invoke.texi (Nvidia PTX Options): Add mptx item.
26050 2021-05-12 Richard Biener <rguenther@suse.de>
26052 PR tree-optimization/100566
26053 * tree-ssa-sccvn.c (dominated_by_p_w_unex): Properly handle
26054 allow_back for all edge queries.
26056 2021-05-12 liuhongt <hongtao.liu@intel.com>
26059 * config/i386/sse.md (<sse4_1_avx2>_pblendvb): Add
26060 splitters for pblendvb of NOT mask register.
26062 2021-05-12 Richard Biener <rguenther@suse.de>
26064 PR tree-optimization/100519
26065 * tree-ssa-reassoc.c (can_associate_p): Split into...
26066 (can_associate_op_p): ... this
26067 (can_associate_type_p): ... and this.
26068 (is_reassociable_op): Call can_associate_op_p.
26069 (break_up_subtract_bb): Call the appropriate predicates.
26070 (reassociate_bb): Likewise.
26072 2021-05-12 Martin Liska <mliska@suse.cz>
26074 * lto-wrapper.c (merge_and_complain): Merge -flto=arg options.
26075 (run_gcc): Use -flto argument detection for merged
26078 2021-05-12 Martin Liska <mliska@suse.cz>
26080 * lto-wrapper.c (print_lto_docs_link): New function.
26081 (run_gcc): Print warning about missing job server detection
26082 after we know NR of partitions. Do the same for -flto{,=1}.
26083 * opts.c (get_option_html_page): Support -flto option.
26085 2021-05-12 Martin Liska <mliska@suse.cz>
26087 * lto-wrapper.c (get_options_from_collect_gcc_options): Change
26089 (append_option): Remove.
26090 (find_option): Rework to use the vector type.
26091 (remove_option): Remove.
26092 (merge_and_complain): Use vectors for cl_decoded_option data
26094 (append_compiler_options): Likewise.
26095 (append_diag_options): Likewise.
26096 (append_linker_options): Likewise.
26097 (append_offload_options): Likewise.
26098 (compile_offload_image): Likewise.
26099 (compile_images_for_offload_targets): Likewise.
26100 (find_and_merge_options): Likewise.
26101 (run_gcc): Likewise.
26103 2021-05-12 Bernd Edlinger <bernd.edlinger@hotmail.de>
26106 * dwarf2out.c (dwarf2out_finish): Set
26107 have_multiple_function_sections with multi-range text_section.
26109 2021-05-12 Martin Liska <mliska@suse.cz>
26111 PR bootstrap/100560
26112 * Makefile.in: Remove version.h from linker command line.
26114 2021-05-12 Richard Biener <rguenther@suse.de>
26116 PR middle-end/100547
26117 * rtl.h (rtvec_alloc): Make argument size_t.
26118 * rtl.c (rtvec_alloc): Verify the count is less than INT_MAX.
26120 2021-05-12 Jakub Jelinek <jakub@redhat.com>
26122 PR middle-end/100508
26123 * cfgexpand.c (expand_debug_expr): For DEBUG_EXPR_DECL with vector
26124 type, don't reuse DECL_RTL if it has different mode, instead force
26125 creation of a new DEBUG_EXPR.
26127 2021-05-12 Jakub Jelinek <jakub@redhat.com>
26128 Marc Glisse <marc.glisse@inria.fr>
26130 PR tree-optimization/94589
26131 * match.pd ((X & Y) == X -> (X & ~Y) == 0,
26132 (X | Y) == Y -> (X & ~Y) == 0): New GIMPLE simplifications.
26134 2021-05-12 Uroš Bizjak <ubizjak@gmail.com>
26137 * config/i386/i386-expand.c (ix86_expand_sse_movcc): Handle V2SF mode.
26138 * config/i386/mmx.md (MMXMODE124): New mode iterator.
26140 (mmxintvecmode): New mode attribute.
26141 (mmxintvecmodelower): Ditto.
26142 (*mmx_maskcmpv2sf3_comm): New insn pattern.
26143 (*mmx_maskcmpv2sf3): Ditto.
26144 (vec_cmpv2sfv2si): New expander.
26145 (vcond<V2FI:mode>v2si): Ditto.
26146 (mmx_vlendvps): New insn pattern.
26147 (vcond<MMXMODE124:mode><MMXMODEI:mode>): Also handle V2SFmode.
26148 (vcondu<MMXMODE124:mode><MMXMODEI:mode>): Ditto.
26149 (vcond_mask_<mode><mmxintvecmodelower>): Ditto.
26151 2021-05-11 Martin Sebor <msebor@redhat.com>
26153 PR middle-end/21433
26154 * expr.c (expand_expr_real_1): Replace unreachable code with an assert.
26156 2021-05-11 Richard Biener <rguenther@suse.de>
26158 * gimple-fold.c (gimple_fold_call): Do not call
26159 maybe_fold_reference on call arguments or the static chain.
26160 (fold_stmt_1): Do not call maybe_fold_reference on GIMPLE_ASM
26163 2021-05-11 Martin Liska <mliska@suse.cz>
26165 * builtins.def (DEF_HSAIL_BUILTIN): Remove.
26166 (DEF_HSAIL_ATOMIC_BUILTIN): Likewise.
26167 (DEF_HSAIL_SAT_BUILTIN): Likewise.
26168 (DEF_HSAIL_INTR_BUILTIN): Likewise.
26169 (DEF_HSAIL_CVT_ZEROI_SAT_BUILTIN): Likewise.
26170 * doc/frontends.texi: Remove BRIG.
26171 * doc/install.texi: Likewise.
26172 * doc/invoke.texi: Likewise.
26173 * doc/standards.texi: Likewise.
26174 * brig-builtins.def: Removed.
26175 * brig/ChangeLog: Removed.
26176 * brig/Make-lang.in: Removed.
26177 * brig/brig-builtins.h: Removed.
26178 * brig/brig-c.h: Removed.
26179 * brig/brig-lang.c: Removed.
26180 * brig/brigfrontend/brig-arg-block-handler.cc: Removed.
26181 * brig/brigfrontend/brig-atomic-inst-handler.cc: Removed.
26182 * brig/brigfrontend/brig-basic-inst-handler.cc: Removed.
26183 * brig/brigfrontend/brig-branch-inst-handler.cc: Removed.
26184 * brig/brigfrontend/brig-cmp-inst-handler.cc: Removed.
26185 * brig/brigfrontend/brig-code-entry-handler.cc: Removed.
26186 * brig/brigfrontend/brig-code-entry-handler.h: Removed.
26187 * brig/brigfrontend/brig-comment-handler.cc: Removed.
26188 * brig/brigfrontend/brig-control-handler.cc: Removed.
26189 * brig/brigfrontend/brig-copy-move-inst-handler.cc: Removed.
26190 * brig/brigfrontend/brig-cvt-inst-handler.cc: Removed.
26191 * brig/brigfrontend/brig-fbarrier-handler.cc: Removed.
26192 * brig/brigfrontend/brig-function-handler.cc: Removed.
26193 * brig/brigfrontend/brig-function.cc: Removed.
26194 * brig/brigfrontend/brig-function.h: Removed.
26195 * brig/brigfrontend/brig-inst-mod-handler.cc: Removed.
26196 * brig/brigfrontend/brig-label-handler.cc: Removed.
26197 * brig/brigfrontend/brig-lane-inst-handler.cc: Removed.
26198 * brig/brigfrontend/brig-machine.c: Removed.
26199 * brig/brigfrontend/brig-machine.h: Removed.
26200 * brig/brigfrontend/brig-mem-inst-handler.cc: Removed.
26201 * brig/brigfrontend/brig-module-handler.cc: Removed.
26202 * brig/brigfrontend/brig-queue-inst-handler.cc: Removed.
26203 * brig/brigfrontend/brig-seg-inst-handler.cc: Removed.
26204 * brig/brigfrontend/brig-signal-inst-handler.cc: Removed.
26205 * brig/brigfrontend/brig-to-generic.cc: Removed.
26206 * brig/brigfrontend/brig-to-generic.h: Removed.
26207 * brig/brigfrontend/brig-util.cc: Removed.
26208 * brig/brigfrontend/brig-util.h: Removed.
26209 * brig/brigfrontend/brig-variable-handler.cc: Removed.
26210 * brig/brigfrontend/hsa-brig-format.h: Removed.
26211 * brig/brigfrontend/phsa.h: Removed.
26212 * brig/brigspec.c: Removed.
26213 * brig/config-lang.in: Removed.
26214 * brig/gccbrig.texi: Removed.
26215 * brig/lang-specs.h: Removed.
26216 * brig/lang.opt: Removed.
26218 2021-05-11 Richard Biener <rguenther@suse.de>
26221 * ipa-param-manipulation.c
26222 (ipa_param_body_adjustments::modify_call_stmt): Avoid
26223 altering SSA_NAME_DEF_STMT by adjusting the calls LHS
26224 via gimple_call_lhs_ptr.
26226 2021-05-11 Alex Coplan <alex.coplan@arm.com>
26229 * config/arm/arm.c (cmse_nonsecure_call_inline_register_clear):
26230 Avoid emitting CFA adjusts on the sp if we have the fp.
26232 2021-05-11 Richard Sandiford <richard.sandiford@arm.com>
26234 * config/aarch64/iterators.md (VMUL_CHANGE_NLANES): Delete.
26235 (VMULD): New iterator.
26236 (VCOND): Handle V4HF and V8HF.
26237 (VCONQ): Fix entry for V2SF.
26238 * config/aarch64/aarch64-simd.md (mul_lane<mode>3): Use VMULD
26239 instead of VMUL. Use a 64-bit vector mode for the indexed operand.
26240 (*aarch64_mul3_elt_<vswap_width_name><mode>): Merge with...
26241 (mul_laneq<mode>3): ...this define_insn. Use VMUL instead of VDQSF.
26242 Use a 128-bit vector mode for the indexed operand. Use stype for
26243 the scheduling type.
26245 2021-05-11 Richard Biener <rguenther@suse.de>
26247 * gimple-fold.c (maybe_fold_reference): Only return
26248 is_gimple_min_invariant values.
26250 2021-05-11 Richard Biener <rguenther@suse.de>
26252 PR middle-end/100509
26253 * gimple-fold.c (fold_gimple_assign): Only call
26254 get_symbol_constant_value on register type symbols.
26256 2021-05-11 Srinath Parvathaneni <srinath.parvathaneni@arm.com>
26257 Joe Ramsay <joe.ramsay@arm.com>
26260 * config/arm/arm_mve.h (__arm_vstrwq_scatter_offset): Fix wrong arguments.
26261 (__arm_vcmpneq): Remove duplicate definition.
26262 (__arm_vstrwq_scatter_offset_p): Likewise.
26263 (__arm_vmaxq_x): Likewise.
26264 (__arm_vmlsdavaq): Likewise.
26265 (__arm_vmlsdavaxq): Likewise.
26266 (__arm_vmlsdavq_p): Likewise.
26267 (__arm_vmlsdavxq_p): Likewise.
26268 (__arm_vrmlaldavhaq): Likewise.
26269 (__arm_vstrbq_p): Likewise.
26270 (__arm_vstrbq_scatter_offset): Likewise.
26271 (__arm_vstrbq_scatter_offset_p): Likewise.
26272 (__arm_vstrdq_scatter_offset): Likewise.
26273 (__arm_vstrdq_scatter_offset_p): Likewise.
26274 (__arm_vstrdq_scatter_shifted_offset): Likewise.
26275 (__arm_vstrdq_scatter_shifted_offset_p): Likewise.
26277 2021-05-11 Jakub Jelinek <jakub@redhat.com>
26279 PR middle-end/100471
26280 * omp-low.c (lower_omp_task_reductions): For OMP_TASKLOOP, if data
26281 is 0, bypass the reduction loop including
26282 GOMP_taskgroup_reduction_unregister call.
26284 2021-05-11 Kewen Lin <linkw@linux.ibm.com>
26286 * config/rs6000/rs6000.c (struct rs6000_cost_data): New member
26287 costing_for_scalar.
26288 (rs6000_density_test): Early return if costing_for_scalar is true.
26289 (rs6000_init_cost): Init costing_for_scalar of rs6000_cost_data.
26291 2021-05-11 Kewen Lin <linkw@linux.ibm.com>
26293 * doc/tm.texi: Regenerated.
26294 * target.def (init_cost): Add new parameter costing_for_scalar.
26295 * targhooks.c (default_init_cost): Adjust for new parameter.
26296 * targhooks.h (default_init_cost): Likewise.
26297 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Likewise.
26298 (vect_compute_single_scalar_iteration_cost): Likewise.
26299 (vect_analyze_loop_2): Likewise.
26300 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Likewise.
26301 (vect_bb_vectorization_profitable_p): Likewise.
26302 * tree-vectorizer.h (init_cost): Likewise.
26303 * config/aarch64/aarch64.c (aarch64_init_cost): Likewise.
26304 * config/i386/i386.c (ix86_init_cost): Likewise.
26305 * config/rs6000/rs6000.c (rs6000_init_cost): Likewise.
26307 2021-05-11 Kewen Lin <linkw@linux.ibm.com>
26309 * config/rs6000/rs6000.c (rs6000_vect_nonmem): Renamed to
26310 vect_nonmem and moved into...
26311 (struct rs6000_cost_data): ...here.
26312 (rs6000_init_cost): Use vect_nonmem of cost_data instead.
26313 (rs6000_add_stmt_cost): Likewise.
26314 (rs6000_finish_cost): Likewise.
26316 2021-05-10 Eric Botcazou <ebotcazou@adacore.com>
26318 * range-op.cc (get_bool_state): Adjust head comment.
26319 (operator_not_equal::op1_range): Fix comment.
26320 (operator_bitwise_xor::op1_range): Remove call to gcc_unreachable.
26322 2021-05-10 Martin Sebor <msebor@redhat.com>
26324 PR middle-end/100425
26325 PR middle-end/100510
26326 * gimple-ssa-warn-alloca.c (pass_walloca::firast_time_p): Rename...
26327 (pass_walloca::xlimit_certain_p): ...to this.
26328 (pass_walloca::gate): Execute for any kind of handled warning.
26329 (pass_walloca::execute): Avoid issuing "maybe" and "unbounded"
26330 warnings when xlimit_certain_p is set.
26332 2021-05-10 Pat Haugen <pthaugen@linux.ibm.com>
26334 * config/rs6000/rs6000.c (rs6000_ira_change_pseudo_allocno_class):
26335 Return ALTIVEC_REGS if that is best_class.
26336 (rs6000_compute_pressure_classes): Add ALTIVEC_REGS.
26338 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
26340 * config/arm/arm.h (CPP_SPEC): Remove error message about
26343 2021-05-10 Martin Jambor <mjambor@suse.cz>
26345 * ipa-prop.h (IPA_NODE_REF): Removed.
26346 (IPA_NODE_REF_GET_CREATE): Likewise.
26347 (IPA_EDGE_REF): Likewise.
26348 (IPA_EDGE_REF_GET_CREATE): Likewise.
26349 (IS_VALID_JUMP_FUNC_INDEX): Likewise.
26350 * ipa-cp.c (print_all_lattices): Replaced IPA_NODE_REF with a direct
26351 use of ipa_node_params_sum.
26352 (ipcp_versionable_function_p): Likewise.
26353 (push_node_to_stack): Likewise.
26354 (pop_node_from_stack): Likewise.
26355 (set_single_call_flag): Replaced two IPA_NODE_REF with one single
26356 direct use of ipa_node_params_sum.
26357 (initialize_node_lattices): Replaced IPA_NODE_REF with a direct use of
26358 ipa_node_params_sum.
26359 (ipa_context_from_jfunc): Replaced IPA_EDGE_REF with a direct use of
26361 (ipcp_verify_propagated_values): Replaced IPA_NODE_REF with a direct
26362 use of ipa_node_params_sum.
26363 (self_recursively_generated_p): Likewise.
26364 (propagate_scalar_across_jump_function): Likewise.
26365 (propagate_context_across_jump_function): Replaced IPA_EDGE_REF with a
26366 direct use of ipa_edge_args_sum, moved the lookup after the early
26367 exit. Replaced IPA_NODE_REF with a direct use of ipa_node_params_sum.
26368 (propagate_bits_across_jump_function): Replaced IPA_NODE_REF with
26369 direct uses of ipa_node_params_sum.
26370 (propagate_vr_across_jump_function): Likewise.
26371 (propagate_aggregate_lattice): Likewise.
26372 (propagate_aggs_across_jump_function): Likewise.
26373 (propagate_constants_across_call): Likewise, also replaced
26374 IPA_EDGE_REF with a direct use of ipa_edge_args_sum.
26375 (good_cloning_opportunity_p): Replaced IPA_NODE_REF with a direct use
26376 of ipa_node_params_sum.
26377 (estimate_local_effects): Likewise.
26378 (add_all_node_vals_to_toposort): Likewise.
26379 (propagate_constants_topo): Likewise.
26380 (ipcp_propagate_stage): Likewise.
26381 (ipcp_discover_new_direct_edges): Likewise.
26382 (calls_same_node_or_its_all_contexts_clone_p): Likewise.
26383 (cgraph_edge_brings_value_p): Likewise (in both overloaded functions).
26384 (get_info_about_necessary_edges): Likewise.
26385 (want_remove_some_param_p): Likewise.
26386 (create_specialized_node): Likewise.
26387 (self_recursive_pass_through_p): Likewise.
26388 (self_recursive_agg_pass_through_p): Likewise.
26389 (find_more_scalar_values_for_callers_subset): Likewise and also
26390 replaced IPA_EDGE_REF with direct uses of ipa_edge_args_sum, in one
26391 case replacing two of those with a single query.
26392 (find_more_contexts_for_caller_subset): Likewise for the
26393 ipa_polymorphic_call_context overload.
26394 (intersect_aggregates_with_edge): Replaced IPA_EDGE_REF with a direct
26395 use of ipa_edge_args_sum. Replaced IPA_NODE_REF with direct uses of
26396 ipa_node_params_sum.
26397 (find_aggregate_values_for_callers_subset): Likewise, also reusing
26398 results of ipa_edge_args_sum->get.
26399 (cgraph_edge_brings_all_scalars_for_node): Replaced IPA_NODE_REF with
26400 direct uses of ipa_node_params_sum, replaced IPA_EDGE_REF with a
26401 direct use of ipa_edge_args_sum.
26402 (cgraph_edge_brings_all_agg_vals_for_node): Likewise, moved node
26403 summary query after the early exit and reused the result later.
26404 (decide_about_value): Replaced IPA_NODE_REF with a direct use of
26405 ipa_node_params_sum.
26406 (decide_whether_version_node): Likewise. Removed re-querying for
26407 summaries after cloning.
26408 (spread_undeadness): Replaced IPA_NODE_REF with a direct use of
26409 ipa_node_params_sum.
26410 (has_undead_caller_from_outside_scc_p): Likewise, reusing results of
26412 (identify_dead_nodes): Likewise.
26413 (ipcp_store_bits_results): Replaced IPA_NODE_REF with direct uses of
26414 ipa_node_params_sum.
26415 (ipcp_store_vr_results): Likewise.
26416 * ipa-fnsummary.c (evaluate_properties_for_edge): Likewise.
26417 (ipa_fn_summary_t::duplicate): Likewise.
26418 (analyze_function_body): Likewise.
26419 (estimate_calls_size_and_time): Likewise.
26420 (ipa_cached_call_context::duplicate_from): Likewise.
26421 (ipa_call_context::equal_to): Likewise.
26422 (remap_edge_params): Likewise.
26423 (ipa_merge_fn_summary_after_inlining): Likewise.
26424 (inline_read_section): Likewise.
26425 * ipa-icf.c (sem_function::param_used_p): Likewise.
26426 * ipa-modref.c (compute_parm_map): Likewise.
26427 (compute_parm_map): Replaced IPA_EDGE_REF with a direct use of
26429 (get_access_for_fnspec): Replaced IPA_NODE_REF with a direct use of
26430 ipa_node_params_sum and replaced IPA_EDGE_REF with a direct use of
26432 * ipa-profile.c (check_argument_count): Likewise.
26433 * ipa-prop.c (ipa_alloc_node_params): Replaced IPA_NODE_REF_GET_CREATE
26434 with a direct use of ipa_node_params_sum.
26435 (ipa_initialize_node_params): Likewise.
26436 (ipa_print_node_jump_functions_for_edge): Replaced IPA_EDGE_REF with a
26437 direct use of ipa_edge_args_sum and reused the query result.
26438 (ipa_compute_jump_functions_for_edge): Replaced IPA_NODE_REF with a
26439 direct use of ipa_node_params_sum and replaced IPA_EDGE_REF with a
26440 direct use of ipa_edge_args_sum.
26441 (ipa_note_param_call): Replaced IPA_NODE_REF with a direct use of
26442 ipa_node_params_sum and reused the result of the query.
26443 (ipa_analyze_node): Likewise.
26444 (ipa_analyze_controlled_uses): Replaced IPA_NODE_REF with a direct use
26445 of ipa_node_params_sum.
26446 (update_jump_functions_after_inlining): Replaced IPA_EDGE_REF with
26447 direct uses of ipa_edge_args_sum.
26448 (update_indirect_edges_after_inlining): Replaced IPA_NODE_REF with
26449 direct uses of ipa_node_params_sum and replaced IPA_EDGE_REF with a
26450 direct use of ipa_edge_args_sum. Removed superficial re-querying the
26452 (propagate_controlled_uses): Replaced IPA_NODE_REF with direct uses of
26453 ipa_node_params_sum and replaced IPA_EDGE_REF with a direct use of
26455 (ipa_propagate_indirect_call_infos): Replaced IPA_EDGE_REF with a
26456 direct use of ipa_edge_args_sum.
26457 (ipa_edge_args_sum_t::duplicate): Replaced IPA_NODE_REF with a direct
26458 use of ipa_node_params_sum.
26459 (ipa_print_node_params): Likewise.
26460 (ipa_write_node_info): Likewise and also replaced IPA_EDGE_REF with
26461 direct uses of ipa_edge_args_sum.
26462 (ipa_read_edge_info): Replaced IPA_EDGE_REF with a direct use of
26464 (ipa_read_node_info): Replaced IPA_NODE_REF with a direct use of
26465 ipa_node_params_sum.
26466 (ipa_prop_write_jump_functions): Likewise. Move variable node to the
26467 scopes where it is used.
26469 2021-05-10 Uroš Bizjak <ubizjak@gmail.com>
26471 * config/i386/i386-expand.c (ix86_expand_sse_movcc)
26472 <case E_V2SImode>: Force op_true to register.
26474 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
26476 * config/arm/iterators.md (MVE_FP_COMPARISONS): New.
26477 * config/arm/mve.md (mve_vcmp<mve_cmp_op>q_f<mode>)
26478 (mve_vcmp<mve_cmp_op>q_n_f<mode>): New, merge all vcmp_*f*
26480 (mve_vcmpeqq_f<mode>, mve_vcmpeqq_n_f<mode>, mve_vcmpgeq_f<mode>)
26481 (mve_vcmpgeq_n_f<mode>, mve_vcmpgtq_f<mode>)
26482 (mve_vcmpgtq_n_f<mode>, mve_vcmpleq_f<mode>)
26483 (mve_vcmpleq_n_f<mode>, mve_vcmpltq_f<mode>)
26484 (mve_vcmpltq_n_f<mode>, mve_vcmpneq_f<mode>)
26485 (mve_vcmpneq_n_f<mode>): Remove.
26486 * config/arm/unspecs.md (VCMPEQQ_F, VCMPEQQ_N_F, VCMPGEQ_F)
26487 (VCMPGEQ_N_F, VCMPGTQ_F, VCMPGTQ_N_F, VCMPLEQ_F, VCMPLEQ_N_F)
26488 (VCMPLTQ_F, VCMPLTQ_N_F, VCMPNEQ_F, VCMPNEQ_N_F): Remove.
26490 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
26492 * config/arm/iterators.md (MVE_COMPARISONS): New.
26494 (mve_cmp_type): New.
26495 * config/arm/mve.md (mve_vcmp<mve_cmp_op>q_<mode>): New, merge all
26497 (mve_vcmpneq_<mode>, mve_vcmpcsq_n_<mode>, mve_vcmpcsq_<mode>)
26498 (mve_vcmpeqq_n_<mode>, mve_vcmpeqq_<mode>, mve_vcmpgeq_n_<mode>)
26499 (mve_vcmpgeq_<mode>, mve_vcmpgtq_n_<mode>, mve_vcmpgtq_<mode>)
26500 (mve_vcmphiq_n_<mode>, mve_vcmphiq_<mode>, mve_vcmpleq_n_<mode>)
26501 (mve_vcmpleq_<mode>, mve_vcmpltq_n_<mode>, mve_vcmpltq_<mode>)
26502 (mve_vcmpneq_n_<mode>, mve_vcmpltq_n_<mode>, mve_vcmpltq_<mode>)
26503 (mve_vcmpneq_n_<mode>): Remove.
26505 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
26507 * config/arm/arm_mve.h (__arm_vcmp*): Remove 's' suffix.
26508 * config/arm/arm_mve_builtins.def (vcmp*): Remove 's' suffix.
26509 * config/arm/mve.md (mve_vcmp*): Remove 's' suffix in pattern
26512 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
26514 * config/arm/arm_mve_builtins.def (vcmpneq_u): Remove.
26515 (vcmpneq_n_u): Likewise.
26516 (vcmpeqq_u,): Likewise.
26517 (vcmpeqq_n_u): Likewise.
26518 * config/arm/iterators.md (supf): Remove VCMPNEQ_U, VCMPEQQ_U,
26519 VCMPEQQ_N_U and VCMPNEQ_N_U.
26520 * config/arm/mve.md (mve_vcmpneq): Remove <supf> iteration.
26521 (mve_vcmpeqq_n): Likewise.
26522 (mve_vcmpeqq): Likewise.
26523 (mve_vcmpneq_n): Likewise.
26525 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
26527 * config/arm/arm_mve.h (__arm_vcmpeq*u*, __arm_vcmpne*u*): Call
26528 the 's' version of the builtin.
26530 2021-05-10 Richard Biener <rguenther@suse.de>
26532 PR tree-optimization/100492
26533 * tree-loop-distribution.c (find_seed_stmts_for_distribution):
26534 Find nothing when the loop contains an irreducible region.
26536 2021-05-10 Richard Biener <rguenther@suse.de>
26538 PR middle-end/100464
26540 * gimple-fold.c (canonicalize_constructor_val): Do not set
26543 2021-05-10 Richard Biener <rguenther@suse.de>
26545 PR tree-optimization/100434
26546 * tree-ssa-dse.c (initialize_ao_ref_for_dse): Handle
26548 (dse_optimize_stmt): Handle call LHS by dropping the
26549 LHS or the whole call if it doesn't have other
26551 (pass_dse::execute): Adjust.
26553 2021-05-10 Martin Liska <mliska@suse.cz>
26555 * Makefile.in: Add missing genversion rule.
26557 2021-05-10 Alex Coplan <alex.coplan@arm.com>
26560 * config/arm/mve.md (*mve_mov<mode>): Simplify output code. Use
26561 vldrw.u32 and vstrw.32 for V2D[IF]mode loads and stores.
26563 2021-05-10 Martin Liska <mliska@suse.cz>
26565 * builtins.c (is_builtin_name): Use startswith
26566 function instead of strncmp.
26567 * collect2.c (main): Likewise.
26568 (has_lto_section): Likewise.
26569 (scan_libraries): Likewise.
26570 * coverage.c (coverage_checksum_string): Likewise.
26571 (coverage_init): Likewise.
26572 * dwarf2out.c (is_cxx): Likewise.
26573 (gen_compile_unit_die): Likewise.
26574 * gcc-ar.c (main): Likewise.
26575 * gcc.c (init_spec): Likewise.
26576 (read_specs): Likewise.
26577 (execute): Likewise.
26578 (check_live_switch): Likewise.
26579 * genattrtab.c (write_attr_case): Likewise.
26580 (IS_ATTR_GROUP): Likewise.
26581 * gencfn-macros.c (main): Likewise.
26582 * gengtype.c (type_for_name): Likewise.
26583 (gen_rtx_next): Likewise.
26584 (get_file_langdir): Likewise.
26585 (write_local): Likewise.
26586 * genmatch.c (get_operator): Likewise.
26587 (get_operand_type): Likewise.
26588 (expr::gen_transform): Likewise.
26589 * genoutput.c (validate_optab_operands): Likewise.
26590 * incpath.c (add_sysroot_to_chain): Likewise.
26591 * langhooks.c (lang_GNU_C): Likewise.
26592 (lang_GNU_CXX): Likewise.
26593 (lang_GNU_Fortran): Likewise.
26594 (lang_GNU_OBJC): Likewise.
26595 * lto-wrapper.c (run_gcc): Likewise.
26596 * omp-general.c (omp_max_simt_vf): Likewise.
26597 * omp-low.c (omp_runtime_api_call): Likewise.
26598 * opts-common.c (parse_options_from_collect_gcc_options): Likewise.
26599 * read-rtl-function.c (function_reader::read_rtx_operand_r): Likewise.
26600 * real.c (real_from_string): Likewise.
26601 * selftest.c (assert_str_startswith): Likewise.
26602 * timevar.c (timer::validate_phases): Likewise.
26603 * tree.c (get_file_function_name): Likewise.
26604 * ubsan.c (ubsan_use_new_style_p): Likewise.
26605 * varasm.c (default_function_rodata_section): Likewise.
26606 (incorporeal_function_p): Likewise.
26607 (default_section_type_flags): Likewise.
26608 * system.h (startswith): Define startswith.
26610 2021-05-10 Martin Liska <mliska@suse.cz>
26612 * bitmap.h (class auto_bitmap): Remove
26613 __cplusplus >= 201103.
26614 * config/aarch64/aarch64.c: Likewise.
26615 * gimple-ssa-store-merging.c (store_immediate_info::store_immediate_info):
26617 * sbitmap.h: Likewise.
26619 2021-05-10 Martin Liska <mliska@suse.cz>
26621 * Makefile.in: Rename gcov-iov to genversion and depend
26622 on version.h (instead of gcov-iov.h).
26623 * gcov-io.h: Include version.h instread of gcov-iov.h.
26624 * gengtype-state.c (read_state_version): Likewise.
26625 * gcov-iov.c: Moved to...
26626 * genversion.c: ...here.
26627 * lto-streamer.h (LTO_major_version): Define it with
26629 * version.c: Removed.
26630 * version.h: Removed.
26632 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
26634 * config/arc/arc.md (UNSPEC_ARC_DMPYWH): Define.
26635 * config/arc/simdext.md (VCT): Add predicates for iterator
26638 (voptab): Likewise.
26639 (vec_widen_<V_US>mult_hi_v4hi): Change pattern predicate.
26640 (<voptab>v2si3): New patterns.
26642 (reduc_plus_scal_v4hi): Likewise.
26643 (reduc_plus_scal_v2si): Likewise.
26644 (vec_duplicatev2si): Likewise.
26645 (vec_duplicatev4hi): Likewise.
26647 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
26649 * config/arc/simdext.md: Format and cleanup file.
26651 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
26653 * config/arc/simdext.md (movmisalignv2hi): Allow misaligned access
26654 only when munaligned-access option is on.
26655 (movmisalign<mode>): Likewise.
26657 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
26659 * common/config/arc/arc-common.c (arc_handle_option): Remove dot
26661 * config/arc/arc.c (arc_reorg): Remove underscore from string.
26663 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
26665 * config/arc/arc.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
26666 (CTZ_DEFINED_VALUE_AT_ZERO): Likewise.
26667 * config/arc/arc.md (clrsbsi2): Cleanup pattern.
26668 (norm_f): Likewise.
26671 (clzsi2): Use fls instruction when available.
26672 (arc_clzsi2): Likewise.
26674 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
26676 * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add r26 and r27.
26678 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
26680 * doc/extend.texi (__builtin_arc_sr): Swap arguments.
26682 2021-05-10 Bernd Edlinger <bernd.edlinger@hotmail.de>
26684 PR middle-end/100467
26685 * toplev.c (compile_file): Call insn_locations_init before
26686 targetm.asm_out.code_end.
26688 2021-05-07 Andrew Stubbs <ams@codesourcery.com>
26691 2021-05-07 Andrew Stubbs <ams@codesourcery.com>
26693 * config/gcn/gcn.c (gcn_scalar_mode_supported_p): Disable TImode.
26695 2021-05-07 Jakub Jelinek <jakub@redhat.com>
26696 Andrew Stubbs <amd@codesourcery.com>
26699 * builtins.c (try_store_by_multiple_pieces): Use force_operand for
26700 emit_move_insn operands.
26702 2021-05-07 Eric Botcazou <ebotcazou@adacore.com>
26704 * cfgexpand.c (expand_gimple_basic_block): Do not inherit a current
26705 location for the outgoing edges of an empty block.
26706 * dwarf2out.c (add_subscript_info): Retrieve the bounds and index
26707 type by means of the get_array_descr_info langhook, if it is set and
26708 returns true. Remove obsolete code dealing with unnamed subtypes.
26710 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
26712 * gimple-range-cache.cc (ssa_block_ranges): Virtualize.
26713 (sbr_vector): Renamed from ssa_block_cache.
26714 (sbr_vector::sbr_vector): Allocate from obstack abd initialize.
26715 (ssa_block_ranges::~ssa_block_ranges): Remove.
26716 (sbr_vector::set_bb_range): Use varying and undefined cached values.
26717 (ssa_block_ranges::set_bb_varying): Remove.
26718 (sbr_vector::get_bb_range): Adjust assert.
26719 (sbr_vector::bb_range_p): Adjust assert.
26720 (~block_range_cache): No freeing loop required.
26721 (block_range_cache::get_block_ranges): Remove.
26722 (block_range_cache::set_bb_range): Inline get_block_ranges.
26723 (block_range_cache::set_bb_varying): Remove.
26724 * gimple-range-cache.h (set_bb_varying): Remove prototype.
26725 * value-range.h (irange_allocator::get_memory): New.
26727 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
26729 * gimple-range-cache.cc (non_null_ref::non_null_deref_p): Search
26730 dominator tree is available and requested.
26731 (ranger_cache::ssa_range_in_bb): Don't search dom tree here.
26732 (ranger_cache::fill_block_cache): Don't search dom tree here either.
26733 * gimple-range-cache.h (non_null_deref_p): Add dom_search param.
26735 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
26737 * gimple-range.cc (gimple_ranger::range_on_exit): Handle block with
26738 only PHI nodes better.
26740 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
26742 * gimple-range-edge.h (gimple_outgoing_range): Rename from
26744 (gcond_edge_range): Export prototype.
26745 * gimple-range-edge.cc (gcond_edge_range): New.
26746 (gimple_outgoing_range::edge_range_p): Use gcond_edge_range.
26747 * gimple-range-gori.h (gori_compute): Use gimple_outgoing_range.
26749 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
26751 * gimple-range-edge.cc (outgoing_range::calc_switch_ranges): Compute
26752 default range into a temp and allocate only what is needed.
26754 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
26756 * range-op.cc (operator_trunc_mod::wi_fold): x % 0 is UNDEFINED.
26758 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
26760 * gimple-range.h (gimple_range_global): Pick up parameter initial
26761 values, and use-before defined locals are UNDEFINED.
26763 2021-05-07 Eric Botcazou <ebotcazou@adacore.com>
26765 * doc/extend.texi (scalar_storage_order): Mention effect on pointer
26767 * tree.h (reverse_storage_order_for_component_p): Return false if
26768 the type is a pointer.
26770 2021-05-07 Andrew Stubbs <ams@codesourcery.com>
26772 * config/gcn/gcn.c (gcn_scalar_mode_supported_p): Disable TImode.
26774 2021-05-07 Uroš Bizjak <ubizjak@gmail.com>
26777 * config/i386/i386-expand.c (ix86_expand_sse_movcc):
26778 Handle V8QI, V4HI and V2SI modes.
26779 * config/i386/mmx.md (mmx_pblendvb): New insn pattern.
26780 * config/i386/sse.md (unspec): Move UNSPEC_BLENDV ...
26781 * config/i386/i386.md (unspec): ... here.
26783 2021-05-07 Tobias Burnus <tobias@codesourcery.com>
26784 Tom de Vries <tdevries@suse.de>
26786 * omp-low.c (lower_rec_simd_input_clauses): Set max_vf = 1 if
26787 a truth_value_p reduction variable is nonintegral.
26789 2021-05-07 Uroš Bizjak <ubizjak@gmail.com>
26792 * config/i386/i386-expand.c (ix86_use_mask_cmp_p):
26793 Return false for mode sizes < 16.
26795 2021-05-07 Jakub Jelinek <jakub@redhat.com>
26798 * config/i386/mmx.md (*xop_pcmov_<mode>): New define_insn.
26800 2021-05-06 Martin Jambor <mjambor@suse.cz>
26802 * ipa-sra.c (ipa_sra_dump_all_summaries): Dump edge summaries even
26803 when there is no function summary.
26804 (ipa_sra_summarize_function): produce edge summaries even when
26807 2021-05-06 Tom Tromey <tom@tromey.com>
26809 * godump.c (string_hash_eq): Remove.
26810 (go_finish): Use htab_eq_string.
26812 2021-05-06 Tom Tromey <tom@tromey.com>
26814 * gengtype-state.c (read_state): Use htab_eq_string.
26815 (string_eq): Remove.
26817 2021-05-06 Tom Tromey <tom@tromey.com>
26819 * gensupport.c (htab_eq_string): Remove.
26821 2021-05-06 Bernd Edlinger <bernd.edlinger@hotmail.de>
26824 * debug.h (gcc_debug_hooks): Add set_ignored_loc function pointer.
26825 * dwarf2out.h (dw_fde_node::ignored_debug): New data item.
26826 * dbxout.c (dbx_debug_hooks, xcoff_debug_hooks): Add dummy
26827 set_ignored_loc callbacks.
26828 * debug.c (do_nothing_debug_hooks): Likewise.
26829 * vmsdbgout.c (vmsdbg_debug_hooks): Likewise.
26830 * dwarf2out.c (text_section_used, cold_text_section_used): Remove.
26831 (in_text_section_p, last_text_label, last_cold_label,
26832 switch_text_ranges, switch_cold_ranges): New data items.
26833 (dwarf2out_note_section_used): Remove.
26834 (dwarf2out_begin_prologue): Set fde->ignored_debug and
26836 (mark_ignored_debug_section): New helper function.
26837 (dwarf2out_end_epilogue, dwarf2out_switch_text_section): Call
26838 mark_ignored_debug_section.
26839 (dwarf2_debug_hooks): Use dwarf2out_set_ignored_loc.
26840 (dwarf2_lineno_debug_hooks): Use dummy for set_ignored_loc.
26841 (size_of_aranges): Adjust formula for multi-part text ranges size.
26842 (output_aranges): Output multi-part text ranges.
26843 (dwarf2out_set_ignored_loc): New callback function.
26844 (dwarf2out_finish): Output multi-part text ranges.
26845 (dwarf2out_c_finalize): Clear new data items.
26846 * final.c (final_start_function_1): Call set_ignored_loc callback.
26847 (final_scan_insn_1): Likewise.
26848 * ggc-page.c (gt_ggc_mx): New helper function.
26849 * stringpool.c (gt_pch_nx): Likewise.
26851 2021-05-06 Richard Biener <rguenther@suse.de>
26853 * timevar.def (TV_TREE_INSERT_PHI_NODES): Remove.
26854 (TV_TREE_SSA_REWRITE_BLOCKS): Likewise.
26855 (TV_TREE_INTO_SSA): New.
26856 * tree-into-ssa.c (insert_phi_nodes): Do not account separately.
26857 (rewrite_blocks): Likewise.
26858 (pass_data_build_ssa): Account to TV_TREE_INTO_SSA.
26860 2021-05-06 Jakub Jelinek <jakub@redhat.com>
26862 * tree-ssa-phiopt.c (value_replacement, minmax_replacement,
26863 abs_replacement, xor_replacement,
26864 cond_removal_in_popcount_clz_ctz_pattern,
26865 replace_phi_edge_with_variable): Change type of phi argument from
26866 gimple * to gphi *.
26868 2021-05-06 Richard Biener <rguenther@suse.de>
26870 * tree-ssa-loop-split.c (split_loop): Delay updating SSA form.
26871 Output an opt-info message.
26872 (do_split_loop_on_cond): Likewise.
26873 (tree_ssa_split_loops): Update SSA form here.
26875 2021-05-06 Richard Biener <rguenther@suse.de>
26877 * tree-inline.c (tree_function_versioning): Fix DECL_BY_REFERENCE
26878 return variable removal.
26880 2021-05-06 Marius Hillenbrand <mhillen@linux.ibm.com>
26882 * config/s390/s390-builtins.def (O_M5, O1_M5, ...): Remove unused macros.
26883 (s390_vec_permi_s64, s390_vec_permi_b64, s390_vec_permi_u64)
26884 (s390_vec_permi_dbl, s390_vpdi): Use the O3_U2 type for the immediate
26886 * config/s390/s390.c (s390_const_operand_ok): Remove unused
26889 2021-05-06 Jakub Jelinek <jakub@redhat.com>
26891 PR tree-optimization/94589
26892 * tree-ssa-phiopt.c (tree_ssa_phiopt_worker): Call
26893 spaceship_replacement.
26894 (cond_only_block_p, spaceship_replacement): New functions.
26896 2021-05-06 Richard Biener <rguenther@suse.de>
26899 * tree-emutls.c (gen_emutls_addr): Pass in whether we're
26900 dealing with a debug use and only query existing addresses
26902 (lower_emutls_1): Avoid splitting out addresses for debug
26903 stmts, reset the debug stmt when we fail to find existing
26905 (lower_emutls_phi_arg): Set wi.stmt.
26907 2021-05-06 Christoph Muellner <cmuellner@gcc.gnu.org>
26910 * config/riscv/riscv.c (riscv_block_move_loop): Use cbranch helper.
26911 * config/riscv/riscv.md (cbranch<mode>4): Generate helpers.
26912 (stack_protect_test): Use cbranch helper.
26914 2021-05-05 Eric Botcazou <ebotcazou@adacore.com>
26917 * config/i386/i386.c (ix86_compute_frame_layout): For a SEH target,
26918 always return the establisher frame for __builtin_frame_address (0).
26920 2021-05-05 Ivan Sorokin <vanyacpp@gmail.com>
26923 * config/i386/i386-builtins.c (ix86_cpu_model_type_node): New.
26924 (ix86_cpu_model_var): Likewise.
26925 (ix86_cpu_features2_type_node): Likewise.
26926 (ix86_cpu_features2_var): Likewise.
26927 (fold_builtin_cpu): Cache __cpu_model and __cpu_features2 with
26930 2021-05-05 Martin Sebor <msebor@redhat.com>
26932 * passes.def (pass_warn_printf): Run after SSA.
26934 2021-05-05 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
26936 * config/arm/neon.md (neon_vtst_combine<mode>): New pattern.
26937 * config/arm/predicates.md (minus_one_operand): New predicate.
26939 2021-05-05 Jeff Law <jlaw@tachyum.com>
26941 * config/avr/avr.md: Remove references to CC_STATUS_INIT.
26943 2021-05-05 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
26945 PR rtl-optimization/100263
26946 * postreload.c (move2add_valid_value_p): Ensure register can
26949 2021-05-05 Eric Botcazou <ebotcazou@adacore.com>
26951 PR rtl-optimization/100411
26952 * cfgcleanup.c (try_crossjump_to_edge): Also skip end of prologue
26953 and beginning of function markers.
26955 2021-05-05 Jeff Law <jlaw@tachyum.com>
26957 * config/cr16/cr16.h (NOTICE_UPDATE_CC): Remove.
26958 * config/cr16/cr16.c (notice_update_cc): Remove.
26959 * config/cr16/cr16-protos.h (notice_update_cc): Remove.
26961 2021-05-05 Uroš Bizjak <ubizjak@gmail.com>
26964 * config/i386/i386-expand.c (ix86_expand_int_sse_cmp):
26965 Handle V8QI, V4HI and V2SI modes.
26966 * config/i386/i386.c (ix86_build_const_vector): Handle V2SImode.
26967 (ix86_build_signbit_mask): Ditto.
26968 * config/i386/mmx.md (MMXMODE14): New mode iterator.
26969 (<smaxmin:code><MMXMODE14:mode>3): New expander.
26970 (*mmx_<smaxmin:code><MMXMODE14:mode>3): New insn pattern.
26971 (<umaxmin:code><MMXMODE24:mode>3): New expander.
26972 (*mmx_<umaxmin:code><MMXMODE24:mode>3): New insn pattern.
26973 (vec_cmp<MMXMODEI:mode><MMXMODEI:mode>): New expander.
26974 (vec_cmpu<MMXMODEI:mode><MMXMODEI:mode>): Ditto.
26975 (vcond<MMXMODEI:mode><MMXMODEI:mode>): Ditto.
26976 (vcondu<MMXMODEI:mode><MMXMODEI:mode>): Ditto.
26977 (vcond_mask_<MMXMODEI:mode><MMXMODEI:mode>): Ditto.
26979 2021-05-05 Eric Botcazou <ebotcazou@adacore.com>
26981 * dwarf2out.c (loc_list_from_tree_1) <DECL>: During early DWARF, do
26982 not expand the VALUE_EXPR of variables put in the non-local frame.
26983 * gimplify.c (gimplify_type_sizes) <RECORD_TYPE>: If the type is not
26984 to be ignored for debug info, ensure its variable offsets are not.
26986 2021-05-05 Richard Biener <rguenther@suse.de>
26988 PR tree-optimization/79333
26989 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_stmt):
26990 Fold stmt following SSA edges.
26992 2021-05-05 Richard Biener <rguenther@suse.de>
26994 PR middle-end/100394
26995 * calls.c (expand_call): Preserve possibly throwing calls.
26996 * cfgexpand.c (expand_call_stmt): When a call can throw signal
26997 RTL expansion there are side-effects.
26998 * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Simplify,
26999 mark all possibly throwing stmts necessary unless we can elide
27001 * tree-ssa-dse.c (pass_dse::execute): Preserve exceptions unless
27002 -fdelete-dead-exceptions.
27003 * tree.h (DECL_PURE_P): Add note about exceptions.
27005 2021-05-05 Alexandre Oliva <oliva@adacore.com>
27007 * config/i386/vxworks.h (DBX_REGISTER_NUMBER): Make it
27010 2021-05-04 David Edelsohn <dje.gcc@gmail.com>
27012 * config/rs6000/rs6000-call.c (rs6000_output_mi_thunk): Use
27013 get_fnname_from_decl for name of thunk.
27014 * config/rs6000/rs6000.c (rs6000_declare_alias): Use assemble_name
27015 and ASM_OUTPUT_LABEL.
27016 (rs6000_xcoff_declare_function_name): Use assemble_name and
27018 (rs6000_xcoff_declare_object_name): Use ASM_OUTPUT_LABEL.
27019 (rs6000_xcoff_encode_section_info): Don't add mapping class
27020 for aliases. Always add [DS] mapping class to primary
27022 (rs6000_asm_weaken_decl): Don't explicitly add [DS].
27024 2021-05-04 Martin Sebor <msebor@redhat.com>
27026 PR middle-end/100307
27027 * builtins.c (compute_objsize_r): Clear base0 for pointers.
27029 2021-05-04 Jeff Law <jlaw@tachyum.com>
27031 * config/bfin/bfin.h (NOTICE_UPDATE_CC): Remove.
27033 2021-05-04 Segher Boessenkool <segher@kernel.crashing.org>
27035 * caller-save.c: Remove CC0.
27036 * cfgcleanup.c: Remove CC0.
27037 * cfgrtl.c: Remove CC0.
27038 * combine.c: Remove CC0.
27039 * compare-elim.c: Remove CC0.
27040 * conditions.h: Remove CC0.
27041 * config/h8300/h8300.h: Remove CC0.
27042 * config/h8300/h8300-protos.h: Remove CC0.
27043 * config/h8300/peepholes.md: Remove CC0.
27044 * config/i386/x86-tune-sched.c: Remove CC0.
27045 * config/m68k/m68k.c: Remove CC0.
27046 * config/rl78/rl78.c: Remove CC0.
27047 * config/sparc/sparc.c: Remove CC0.
27048 * config/xtensa/xtensa.c: Remove CC0.
27049 (gen_conditional_move): Use pc_rtx instead of cc0_rtx in a piece of
27050 RTL where that is used as a placeholder only.
27051 * cprop.c: Remove CC0.
27052 * cse.c: Remove CC0.
27053 * cselib.c: Remove CC0.
27054 * df-problems.c: Remove CC0.
27055 * df-scan.c: Remove CC0.
27056 * doc/md.texi: Remove CC0. Adjust an example.
27057 * doc/rtl.texi: Remove CC0. Adjust an example.
27058 * doc/tm.texi: Regenerate.
27059 * doc/tm.texi.in: Remove CC0.
27060 * emit-rtl.c: Remove CC0.
27061 * final.c: Remove CC0.
27062 * fwprop.c: Remove CC0.
27063 * gcse-common.c: Remove CC0.
27064 * gcse.c: Remove CC0.
27065 * genattrtab.c: Remove CC0.
27066 * genconfig.c: Remove CC0.
27067 * genemit.c: Remove CC0.
27068 * genextract.c: Remove CC0.
27069 * gengenrtl.c: Remove CC0.
27070 * genrecog.c: Remove CC0.
27071 * haifa-sched.c: Remove CC0.
27072 * ifcvt.c: Remove CC0.
27073 * ira-costs.c: Remove CC0.
27074 * ira.c: Remove CC0.
27075 * jump.c: Remove CC0.
27076 * loop-invariant.c: Remove CC0.
27077 * lra-constraints.c: Remove CC0.
27078 * lra-eliminations.c: Remove CC0.
27079 * optabs.c: Remove CC0.
27080 * postreload-gcse.c: Remove CC0.
27081 * postreload.c: Remove CC0.
27082 * print-rtl.c: Remove CC0.
27083 * read-rtl-function.c: Remove CC0.
27084 * reg-notes.def: Remove CC0.
27085 * reg-stack.c: Remove CC0.
27086 * reginfo.c: Remove CC0.
27087 * regrename.c: Remove CC0.
27088 * reload.c: Remove CC0.
27089 * reload1.c: Remove CC0.
27090 * reorg.c: Remove CC0.
27091 * resource.c: Remove CC0.
27092 * rtl.c: Remove CC0.
27093 * rtl.def: Remove CC0.
27094 * rtl.h: Remove CC0.
27095 * rtlanal.c: Remove CC0.
27096 * sched-deps.c: Remove CC0.
27097 * sched-rgn.c: Remove CC0.
27098 * shrink-wrap.c: Remove CC0.
27099 * simplify-rtx.c: Remove CC0.
27100 * system.h: Remove CC0. Poison NOTICE_UPDATE_CC, CC_STATUS_MDEP_INIT,
27101 CC_STATUS_MDEP, and CC_STATUS.
27102 * target.def: Remove CC0.
27103 * valtrack.c: Remove CC0.
27104 * var-tracking.c: Remove CC0.
27106 2021-05-04 Richard Biener <rguenther@suse.de>
27108 PR tree-optimization/100414
27109 * tree-ssa-phiopt.c (get_non_trapping): Do not compute dominance
27111 (tree_ssa_phiopt_worker): But unconditionally here.
27113 2021-05-04 Tobias Burnus <tobias@codesourcery.com>
27115 * omp-low.c (lower_rec_input_clauses, lower_reduction_clauses): Handle
27116 && and || with floating-point and complex arguments.
27118 2021-05-04 Eric Botcazou <ebotcazou@adacore.com>
27120 * tree-inline.c (insert_debug_decl_map): Delete.
27121 (copy_debug_stmt): Minor tweak.
27122 (setup_one_parameter): Do not use a variable if the value is either
27123 a read-only DECL or a non-addressable local variable in the caller.
27124 In this case, insert the debug-only variable in the map manually.
27125 (expand_call_inline): Do not generate a CLOBBER for these values.
27126 * tree-inline.h (debug_map): Minor tweak.
27128 2021-05-04 Eric Botcazou <ebotcazou@adacore.com>
27130 * builtins.c (builtin_with_linkage_p): Return true for stp[n]cpy.
27131 * symtab.c (symtab_node::output_to_lto_symbol_table_p): Tidy up.
27133 2021-05-04 Richard Biener <rguenther@suse.de>
27135 PR tree-optimization/100329
27136 * tree-ssa-reassoc.c (can_reassociate_p): Do not reassociate
27138 (insert_stmt_after): Assert we're not running into asm goto.
27140 2021-05-04 Richard Biener <rguenther@suse.de>
27142 PR tree-optimization/100398
27143 * tree-ssa-dse.c (pass_dse::execute): Preserve control
27146 2021-05-04 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
27148 * builtins.c (try_store_by_multiple_pieces): Fix constfun's prototype.
27150 2021-05-04 Alexandre Oliva <oliva@adacore.com>
27152 * builtins.c (try_store_by_multiple_pieces): New.
27153 (expand_builtin_memset_args): Use it. If target_char_cast
27154 fails, proceed as for non-constant val. Pass len's ctz to...
27155 * expr.c (clear_storage_hints): ... this. Try store by
27156 multiple pieces after setmem.
27157 (clear_storage): Adjust.
27158 * expr.h (clear_storage_hints): Likewise.
27159 (try_store_by_multiple_pieces): Declare.
27160 * passes.def: Replace the last copy_prop with ccp.
27162 2021-05-03 Tom de Vries <tdevries@suse.de>
27165 * omp-low.c (lower_rec_input_clauses): Disable SIMT for user-defined
27168 2021-05-03 Richard Biener <rguenther@suse.de>
27170 * tree-ssa-dse.c (dse_classify_store): Track two PHI defs.
27172 2021-05-03 Richard Biener <rguenther@suse.de>
27174 * tree-ssa-dse.c: Do not include domwalk.h but cfganal.h.
27175 (dse_dom_walker): Remove.
27176 (dse_dom_walker::dse_optimize_stmt): Rename...
27177 (dse_optimize_stmt): ... to this, pass in live_bytes sbitmap.
27178 (dse_dom_walker::before_dom_children): Inline ...
27179 (pass_dse::execute): ... here. Perform a reverse program
27182 2021-05-03 H.J. Lu <hjl.tools@gmail.com>
27185 * configure: Regenerated.
27187 2021-05-03 Ilya Leoshkevich <iii@linux.ibm.com>
27190 * config/s390/s390.c (s390_hard_fp_reg_p): New function.
27191 (s390_md_asm_adjust): Handle hard registers.
27193 2021-05-03 Jakub Jelinek <jakub@redhat.com>
27195 PR tree-optimization/100382
27196 * tree-ssa-dse.c: Include tree-eh.h.
27197 (dse_dom_walker::before_dom_children): Don't remove stmts if
27198 stmt_unremovable_because_of_non_call_eh_p is true.
27200 2021-05-02 David Edelsohn <dje.gcc@gmail.com>
27202 * varasm.c (compute_reloc_for_var): Split out from...
27203 (get_variable_section): Use it.
27204 * output.h (compute_reloc_for_var): Declare.
27205 * config/rs6000/rs6000-protos.h
27206 (rs6000_xcoff_asm_output_aligned_decl_common): Change alignment to
27208 * config/rs6000/rs6000.c (rs6000_legitimize_tls_address_aix):
27209 Don't append storage mapping class to symbol.
27210 (rs6000_xcoff_asm_named_section): Add BS and UL mapping classes.
27211 Don't convert TLS BSS to common.
27212 (rs6000_xcoff_unique_section): Don't fall back to select_secton.
27213 (rs6000_xcoff_section_type_flags): Add SECTION_BSS if DECL is
27215 (rs6000_xcoff_asm_globalize_decl_name): Don't strip storage
27217 (rs6000_xcoff_asm_output_aligned_decl_common): Align is unsigned int.
27218 If align is 0 from TLS class, use the same rules as varasm.c
27219 If not common, switch to BSS section manually.
27220 If common, emit appropriate comm or lcomm directive.
27221 (rs6000_xcoff_encode_section_info): Add logic to append all
27222 storage mapping classes.
27223 (rs6000_asm_weaken_decl): Adjust for qualname symbols.
27224 * config/rs6000/xcoff.h (ASM_OUTPUT_ALIGNED_DECL_LOCAL): Use
27225 rs6000_xcoff_asm_output_aligned_decl_common.
27226 (ASM_OUTPUT_ALIGNED_DECL_LOCAL): Use
27227 rs6000_xcoff_asm_output_aligned_decl_common.
27228 (ASM_OUTPUT_TLS_COMMON): Use
27229 rs6000_xcoff_asm_output_aligned_decl_common.
27231 2021-05-02 Jakub Jelinek <jakub@redhat.com>
27234 * config/nvptx/nvptx.c (nvptx_sese_pseudo): Use nullptr instead of 0
27235 as first argument of pseudo_node_t constructors.
27237 2021-05-02 Jakub Jelinek <jakub@redhat.com>
27240 * config/i386/t-i386 (TM_H): Add $(srcdir)/config/i386/i386-isa.def.
27242 2021-05-01 Aldy Hernandez <aldyh@redhat.com>
27244 * value-range.cc (DEFINE_INT_RANGE_GC_STUBS): Remove.
27245 (gt_pch_nx (int_range<1> *&)): New.
27246 (gt_ggc_mx (int_range<1> *&)): New.
27247 * value-range.h (class irange): Add GTY support for
27250 2021-05-01 Geng Qi <gengqi@linux.alibaba.com>
27252 * doc/options.texi (Negative): Change either or to both and.
27254 2021-04-30 Jonathan Wright <jonathan.wright@arm.com>
27256 * config/aarch64/aarch64-simd-builtins.def: Add
27257 float_ml[as][q]_laneq builtin generator macros.
27258 * config/aarch64/aarch64-simd.md (mul_laneq<mode>3): Define.
27259 (aarch64_float_mla_laneq<mode>): Define.
27260 (aarch64_float_mls_laneq<mode>): Define.
27261 * config/aarch64/arm_neon.h (vmla_laneq_f32): Use RTL builtin
27262 instead of GCC vector extensions.
27263 (vmlaq_laneq_f32): Likewise.
27264 (vmls_laneq_f32): Likewise.
27265 (vmlsq_laneq_f32): Likewise.
27267 2021-04-30 Jonathan Wright <jonathan.wright@arm.com>
27269 * config/aarch64/aarch64-simd-builtins.def: Add
27270 float_ml[as]_lane builtin generator macros.
27271 * config/aarch64/aarch64-simd.md (*aarch64_mul3_elt<mode>):
27273 (mul_lane<mode>3): This, and re-order arguments.
27274 (aarch64_float_mla_lane<mode>): Define.
27275 (aarch64_float_mls_lane<mode>): Define.
27276 * config/aarch64/arm_neon.h (vmla_lane_f32): Use RTL builtin
27277 instead of GCC vector extensions.
27278 (vmlaq_lane_f32): Likewise.
27279 (vmls_lane_f32): Likewise.
27280 (vmlsq_lane_f32): Likewise.
27282 2021-04-30 Jonathan Wright <jonathan.wright@arm.com>
27284 * config/aarch64/aarch64-simd-builtins.def: Add float_ml[as]
27285 builtin generator macros.
27286 * config/aarch64/aarch64-simd.md (aarch64_float_mla<mode>):
27288 (aarch64_float_mls<mode>): Define.
27289 * config/aarch64/arm_neon.h (vmla_f32): Use RTL builtin
27290 instead of relying on GCC vector extensions.
27291 (vmla_f64): Likewise.
27292 (vmlaq_f32): Likewise.
27293 (vmlaq_f64): Likewise.
27294 (vmls_f32): Likewise.
27295 (vmls_f64): Likewise.
27296 (vmlsq_f32): Likewise.
27297 (vmlsq_f64): Likewise.
27298 * config/aarch64/iterators.md: Define VDQF_DF mode iterator.
27300 2021-04-30 Jonathan Wright <jonathan.wright@arm.com>
27302 * config/aarch64/aarch64-simd-builtins.def: Add
27303 float_ml[as]_n_builtin generator macros.
27304 * config/aarch64/aarch64-simd.md (*aarch64_mul3_elt_from_dup<mode>):
27306 (mul_n<mode>3): This, and re-order arguments.
27307 (aarch64_float_mla_n<mode>): Define.
27308 (aarch64_float_mls_n<mode>): Define.
27309 * config/aarch64/arm_neon.h (vmla_n_f32): Use RTL builtin
27310 instead of inline asm.
27311 (vmlaq_n_f32): Likewise.
27312 (vmls_n_f32): Likewise.
27313 (vmlsq_n_f32): Likewise.
27315 2021-04-30 Jonathan Wright <joanthan.wright@arm.com>
27317 * config/aarch64/aarch64-simd-builtins.def: Add pmull[2]
27318 builtin generator macros.
27319 * config/aarch64/aarch64-simd.md (aarch64_pmullv8qi): Define.
27320 (aarch64_pmull_hiv16qi_insn): Define.
27321 (aarch64_pmull_hiv16qi): Define.
27322 * config/aarch64/arm_neon.h (vmull_high_p8): Use RTL builtin
27323 instead of inline asm.
27324 (vmull_p8): Likewise.
27326 2021-04-30 Senthil Kumar Selvaraj <saaadhu@gcc.gnu.org>
27328 * config/avr/avr.md: Adjust peepholes to match and
27329 generate parallels with clobber of REG_CC.
27330 (mov<mode>_insn): Rename to mov<mode>_insn_split.
27331 (*mov<mode>_insn): Rename to mov<mode>_insn.
27333 2021-04-30 David Edelsohn <dje.gcc@gmail.com>
27335 * varasm.c (use_blocks_for_decl_p): Don't use section anchors
27336 for VAR_DECLs if -fdata-sections enabled.
27338 2021-04-30 Michael Meissner <meissner@linux.ibm.com>
27340 PR bootstrap/100327
27341 * config/rs6000/rs6000.c
27342 (TARGET_LIBGCC_FLOATING_MODE_SUPPORTED_P): Define.
27343 (rs6000_libgcc_floating_mode_supported_p): New target hook.
27345 2021-04-30 Aldy Hernandez <aldyh@redhat.com>
27347 * tree-ssa-threadbackward.c (class thread_jumps): Split out code
27349 (class back_threader_registry): ...to here...
27350 (class back_threader_profitability): ...and here...
27351 (thread_jumps::thread_through_all_blocks): Remove argument.
27352 (back_threader_registry::back_threader_registry): New.
27353 (back_threader_registry::~back_threader_registry): New.
27354 (back_threader_registry::thread_through_all_blocks): New.
27355 (thread_jumps::profitable_jump_thread_path): Move from here...
27356 (back_threader_profitability::profitable_path_p): ...to here.
27357 (thread_jumps::find_taken_edge): New.
27358 (thread_jumps::convert_and_register_current_path): Move...
27359 (back_threader_registry::register_path): ...to here.
27360 (thread_jumps::register_jump_thread_path_if_profitable): Move...
27361 (thread_jumps::maybe_register_path): ...to here.
27362 (thread_jumps::handle_phi): Call find_taken_edge and
27363 maybe_register_path.
27364 (thread_jumps::handle_assignment): Same.
27365 (thread_jumps::fsm_find_control_statement_thread_paths): Remove
27366 tree argument to handle_phi and handle_assignment.
27367 (thread_jumps::find_jump_threads_backwards): Set m_name. Remove
27368 set of m_speed_p and m_max_threaded_paths.
27369 (pass_thread_jumps::execute): Remove second argument from
27370 find_jump_threads_backwards.
27371 (pass_early_thread_jumps::execute): Same.
27373 2021-04-30 Aldy Hernandez <aldyh@redhat.com>
27375 * tree-ssa-dom.c (class dom_jump_threader_simplifier): New.
27376 (class dom_opt_dom_walker): Initialize some class variables.
27377 (pass_dominator::execute): Pass evrp_range_analyzer and
27378 dom_jump_threader_simplifier to dom_opt_dom_walker.
27379 Adjust for some functions moving into classes.
27380 (simplify_stmt_for_jump_threading): Adjust and move to...
27381 (jump_threader_simplifier::simplify): ...here.
27382 (dom_opt_dom_walker::before_dom_children): Adjust for
27383 m_evrp_range_analyzer.
27384 (dom_opt_dom_walker::after_dom_children): Remove x_vr_values hack.
27385 (test_for_singularity): Place in dom_opt_dom_walker class.
27386 (dom_opt_dom_walker::optimize_stmt): The argument
27387 evrp_range_analyzer is now a class field.
27388 * tree-ssa-threadbackward.c (class thread_jumps): Add m_registry.
27389 (thread_jumps::thread_through_all_blocks): New.
27390 (thread_jumps::convert_and_register_current_path): Use m_registry.
27391 (pass_thread_jumps::execute): Adjust for thread_through_all_blocks
27392 being in the threader class.
27393 (pass_early_thread_jumps::execute): Same.
27394 * tree-ssa-threadedge.c (threadedge_initialize_values): Move...
27395 (jump_threader::jump_threader): ...here.
27396 (threadedge_finalize_values): Move...
27397 (jump_threader::~jump_threader): ...here.
27398 (jump_threader::remove_jump_threads_including): New.
27399 (jump_threader::thread_through_all_blocks): New.
27400 (record_temporary_equivalences_from_phis): Move...
27401 (jump_threader::record_temporary_equivalences_from_phis): ...here.
27402 (record_temporary_equivalences_from_stmts_at_dest): Move...
27403 (jump_threader::record_temporary_equivalences_from_stmts_at_dest):
27405 (simplify_control_stmt_condition_1): Move to jump_threader class.
27406 (simplify_control_stmt_condition): Move...
27407 (jump_threader::simplify_control_stmt_condition): ...here.
27408 (thread_around_empty_blocks): Move...
27409 (jump_threader::thread_around_empty_blocks): ...here.
27410 (thread_through_normal_block): Move...
27411 (jump_threader::thread_through_normal_block): ...here.
27412 (thread_across_edge): Move...
27413 (jump_threader::thread_across_edge): ...here.
27414 (thread_outgoing_edges): Move...
27415 (jump_threader::thread_outgoing_edges): ...here.
27416 * tree-ssa-threadedge.h: Move externally facing functings...
27417 (class jump_threader): ...here...
27418 (class jump_threader_simplifier): ...and here.
27419 * tree-ssa-threadupdate.c (struct redirection_data): Remove comment.
27420 (jump_thread_path_allocator::jump_thread_path_allocator): New.
27421 (jump_thread_path_allocator::~jump_thread_path_allocator): New.
27422 (jump_thread_path_allocator::allocate_thread_edge): New.
27423 (jump_thread_path_allocator::allocate_thread_path): New.
27424 (jump_thread_path_registry::jump_thread_path_registry): New.
27425 (jump_thread_path_registry::~jump_thread_path_registry): New.
27426 (jump_thread_path_registry::allocate_thread_edge): New.
27427 (jump_thread_path_registry::allocate_thread_path): New.
27428 (dump_jump_thread_path): Make extern.
27429 (debug (const vec<jump_thread_edge *> &path)): New.
27430 (struct removed_edges): Move to tree-ssa-threadupdate.h.
27431 (struct thread_stats_d): Remove.
27432 (remove_ctrl_stmt_and_useless_edges): Make static.
27433 (lookup_redirection_data): Move...
27434 (jump_thread_path_registry::lookup_redirection_data): ...here.
27435 (ssa_redirect_edges): Make static.
27436 (thread_block_1): Move...
27437 (jump_thread_path_registry::thread_block_1): ...here.
27438 (thread_block): Move...
27439 (jump_thread_path_registry::thread_block): ...here.
27440 (thread_through_loop_header): Move...
27441 (jump_thread_path_registry::thread_through_loop_header): ...here.
27442 (mark_threaded_blocks): Move...
27443 (jump_thread_path_registry::mark_threaded_blocks): ...here.
27444 (debug_path): Move...
27445 (jump_thread_path_registry::debug_path): ...here.
27446 (debug_all_paths): Move...
27447 (jump_thread_path_registry::dump): ..here.
27448 (rewire_first_differing_edge): Move...
27449 (jump_thread_path_registry::rewire_first_differing_edge): ...here.
27450 (adjust_paths_after_duplication): Move...
27451 (jump_thread_path_registry::adjust_paths_after_duplication): ...here.
27452 (duplicate_thread_path): Move...
27453 (jump_thread_path_registry::duplicate_thread_path): ..here.
27454 (remove_jump_threads_including): Move...
27455 (jump_thread_path_registry::remove_jump_threads_including): ...here.
27456 (thread_through_all_blocks): Move to...
27457 (jump_thread_path_registry::thread_through_all_blocks): ...here.
27458 (delete_jump_thread_path): Remove.
27459 (register_jump_thread): Move...
27460 (jump_thread_path_registry::register_jump_thread): ...here.
27461 * tree-ssa-threadupdate.h: Move externally facing functions...
27462 (class jump_thread_path_allocator): ...here...
27463 (class jump_thread_path_registry): ...and here.
27464 (thread_through_all_blocks): Remove.
27465 (struct removed_edges): New.
27466 (register_jump_thread): Remove.
27467 (remove_jump_threads_including): Remove.
27468 (delete_jump_thread_path): Remove.
27469 (remove_ctrl_stmt_and_useless_edges): Remove.
27470 (free_dom_edge_info): New prototype.
27471 * tree-vrp.c: Remove x_vr_values hack.
27472 (class vrp_jump_threader_simplifier): New.
27473 (vrp_jump_threader_simplifier::simplify): New.
27474 (vrp_jump_threader::vrp_jump_threader): Adjust method signature.
27475 Remove m_dummy_cond.
27476 Instantiate m_simplifier and m_threader.
27477 (vrp_jump_threader::thread_through_all_blocks): New.
27478 (vrp_jump_threader::simplify_stmt): Remove.
27479 (vrp_jump_threader::after_dom_children): Do not set m_dummy_cond.
27480 Remove x_vr_values hack.
27481 (execute_vrp): Adjust for thread_through_all_blocks being in a
27484 2021-04-30 Christophe Lyon <christophe.lyon@linaro.org>
27486 * genflags.c (gen_insn): Print failed expansion string.
27488 2021-04-30 H.J. Lu <hjl.tools@gmail.com>
27490 * expr.c (alignment_for_piecewise_move): Call mode_for_size
27491 without limit to MAX_FIXED_MODE_SIZE.
27493 2021-04-30 H.J. Lu <hjl.tools@gmail.com>
27495 PR middle-end/90773
27496 * builtins.c (builtin_memset_gen_str): Don't use return from
27497 simplify_gen_subreg.
27499 2021-04-30 Uroš Bizjak <ubizjak@gmail.com>
27502 * config/i386/i386.md (*add<mode>3_carry_0r): New insn pattern.
27503 (*addsi3_carry_zext_0r): Ditto.
27504 (*sub<mode>3_carry_0): Ditto.
27505 (*subsi3_carry_zext_0r): Ditto.
27506 * config/i386/predicates.md (ix86_carry_flag_unset_operator):
27508 * config/i386/i386.c (ix86_rtx_costs) <case PLUS, case MINUS>:
27509 Also consider ix86_carry_flag_unset_operator to calculate
27510 the cost of adc/sbb insn.
27512 2021-04-30 Roman Zhuykov <zhroma@ispras.ru>
27514 PR rtl-optimization/100225
27515 PR rtl-optimization/84878
27516 * modulo-sched.c (sms_schedule): Use note_stores to skip loops
27517 where we have an instruction which touches (writes) any hard
27518 register from df->regular_block_artificial_uses set.
27519 Allow not-single-set instruction only right before basic block
27522 2021-04-30 Geng Qi <gengqi@linux.alibaba.com>
27524 * config/riscv/riscv.opt (march=,mabi=): Negative itself.
27526 2021-04-30 LevyHsu <admin@levyhsu.com>
27528 * config/riscv/riscv.c (riscv_min_arithmetic_precision): New.
27529 * config/riscv/riscv.h (TARGET_MIN_ARITHMETIC_PRECISION): New.
27530 * config/riscv/riscv.md (addv<mode>4, uaddv<mode>4): New.
27531 (subv<mode>4, usubv<mode>4, mulv<mode>4, umulv<mode>4): New.
27533 2021-04-29 Alexandre Oliva <oliva@adacore.com>
27535 * config.gcc: Merged x86 and x86_64 cpu_type-setting cases.
27537 2021-04-29 Alexandre Oliva <oliva@adacore.com>
27539 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Rename to...
27540 (ASM_OUTPUT_MAX_SKIP_ALIGN): ... this. Enclose in do/while(0).
27541 * config/i386/i386.c: Adjust.
27542 * config/i386/i386.md: Adjust.
27543 * config/i386/darwin.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Drop.
27544 * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
27545 * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
27546 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
27547 * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
27548 * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
27549 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
27550 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
27551 * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
27552 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
27553 (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
27555 2021-04-29 Uroš Bizjak <ubizjak@gmail.com>
27557 * config/i386/i386-expand.c (ix86_expand_int_compare):
27558 Swap operands of GTU and LEU comparison to emit carry flag comparison.
27559 * config/i386/i386.md (*add<mode>3_carry_0): Change insn
27560 predicate to allow more combine opportunities with memory operands.
27561 (*sub<mode>3_carry_0): Ditto.
27563 2021-04-29 Richard Sandiford <richard.sandiford@arm.com>
27565 PR rtl-optimization/100303
27566 * rtl-ssa/accesses.cc (function_info::make_use_available): Take a
27567 boolean that indicates whether the use will only be used in
27568 debug instructions. Treat it in the same way that existing
27569 cross-EBB debug references would be handled if so.
27570 (function_info::make_uses_available): Likewise.
27571 * rtl-ssa/functions.h (function_info::make_uses_available): Update
27572 prototype accordingly.
27573 (function_info::make_uses_available): Likewise.
27574 * fwprop.c (try_fwprop_subst): Update call accordingly.
27576 2021-04-29 Jeff Law <jlaw@tachyum.com>
27578 * config/nios2/nios2-protos.h (nios2_fpu_insn_enabled): Move outside
27581 2021-04-29 Uroš Bizjak <ubizjak@gmail.com>
27582 Richard Biener <rguenther@suse.de>
27585 * config/i386/i386-builtin.def (IX86_BUILTIN_MASKLOADPD)
27586 (IX86_BUILTIN_MASKLOADPS, IX86_BUILTIN_MASKLOADPD256)
27587 (IX86_BUILTIN_MASKLOADPS256, IX86_BUILTIN_MASKLOADD)
27588 (IX86_BUILTIN_MASKLOADQ, IX86_BUILTIN_MASKLOADD256)
27589 (IX86_BUILTIN_MASKLOADQ256): Move from SPECIAL_ARGS
27590 to PURE_ARGS category.
27591 * config/i386/i386-builtins.c (ix86_init_mmx_sse_builtins):
27592 Handle PURE_ARGS category.
27593 * config/i386/i386-expand.c (ix86_expand_builtin): Ditto.
27595 2021-04-29 Eric Botcazou <ebotcazou@adacore.com>
27597 * configure.ac: Check for the presence of sys/locking.h header and
27598 for whether _LK_LOCK is supported by _locking.
27599 * configure: Regenerate.
27600 * config.in: Likewise.
27601 * gcov-io.h: Define GCOV_LOCKED_WITH_LOCKING if HOST_HAS_LK_LOCK.
27602 * gcov-io.c (gcov_open): Add support for GCOV_LOCKED_WITH_LOCKING.
27603 * system.h: Include <sys/locking.h> if HAVE_SYS_LOCKING_H.
27605 2021-04-29 Uroš Bizjak <ubizjak@gmail.com>
27607 * config/i386/predicates.md (fcmov_comparison_operator):
27608 Do not check for trivial FP comparison operator.
27609 <case GEU, case LTU>: Allow CCGZmode.
27610 <case GTU, case LEU>: Do not allow CCCmode.
27611 (ix86_comparison_operator) <case GTU, case LEU>: Allow only CCmode.
27612 (ix86_carry_flag_operator): Match only LTU and UNLT code.
27613 Do not check for trivial FP comparison operator. Allow CCGZmode.
27615 2021-04-29 Tom de Vries <tdevries@suse.de>
27617 * omp-expand.c (expand_omp_simd): Add step_orig, and replace uses of
27618 fd->loop.step by either step or orig_step.
27620 2021-04-29 Eric Botcazou <ebotcazou@adacore.com>
27622 * config/sparc/sparc.c (gen_load_pcrel_sym): Delete.
27623 (load_got_register): Do the PIC dance here.
27624 (sparc_legitimize_tls_address): Simplify.
27625 (sparc_emit_probe_stack_range): Likewise.
27626 (sparc32_initialize_trampoline): Likewise.
27627 (sparc64_initialize_trampoline): Likewise.
27628 * config/sparc/sparc.md (load_pcrel_sym<P:mode>): Add @ marker.
27629 (probe_stack_range<P:mode>): Likewise.
27630 (flush<P:mode>): Likewise.
27631 (tgd_hi22<P:mode>): Likewise.
27632 (tgd_lo10<P:mode>): Likewise.
27633 (tgd_add<P:mode>): Likewise.
27634 (tgd_call<P:mode>): Likewise.
27635 (tldm_hi22<P:mode>): Likewise.
27636 (tldm_lo10<P:mode>): Likewise.
27637 (tldm_add<P:mode>): Likewise.
27638 (tldm_call<P:mode>): Likewise.
27639 (tldo_hix22<P:mode>): Likewise.
27640 (tldo_lox10<P:mode>): Likewise.
27641 (tldo_add<P:mode>): Likewise.
27642 (tie_hi22<P:mode>): Likewise.
27643 (tie_lo10<P:mode>): Likewise.
27644 (tie_add<P:mode>): Likewise.
27645 (tle_hix22<P:mode>): Likewise.
27646 (tle_lox10<P:mode>): Likewise.
27647 (stack_protect_setsi): Rename to...
27648 (stack_protect_set32): ...this.
27649 (stack_protect_setdi): Rename to...
27650 (stack_protect_set64): ...this.
27651 (stack_protect_set): Adjust calls to above.
27652 (stack_protect_testsi): Rename to...
27653 (stack_protect_test32): ...this.
27654 (stack_protect_testdi): Rename to...
27655 (stack_protect_test64): ...this.
27656 (stack_protect_test): Adjust calls to above.
27658 2021-04-29 H.J. Lu <hjl.tools@gmail.com>
27660 PR middle-end/90773
27661 * builtins.c (builtin_memcpy_read_str): Add a dummy argument.
27662 (builtin_strncpy_read_str): Likewise.
27663 (builtin_memset_read_str): Add an argument for the previous RTL
27664 information and generate the new RTL from the previous RTL info.
27665 (builtin_memset_gen_str): Likewise.
27666 * builtins.h (builtin_strncpy_read_str): Update the prototype.
27667 (builtin_memset_read_str): Likewise.
27668 * expr.c (by_pieces_ninsns): If targetm.overlap_op_by_pieces_p()
27669 returns true, round up size and alignment to the widest integer
27670 mode for maximum size.
27671 (pieces_addr::adjust): Add a pointer to by_pieces_prev argument
27672 and pass it to m_constfn.
27673 (op_by_pieces_d): Add m_push and m_overlap_op_by_pieces.
27674 (op_by_pieces_d::op_by_pieces_d): Add a bool argument to
27675 initialize m_push. Initialize m_overlap_op_by_pieces with
27676 targetm.overlap_op_by_pieces_p ().
27677 (op_by_pieces_d::run): Pass the previous RTL information to
27678 pieces_addr::adjust and generate overlapping operations if
27679 m_overlap_op_by_pieces is true.
27681 (move_by_pieces_d::move_by_pieces_d): Updated for op_by_pieces_d
27683 (store_by_pieces_d::store_by_pieces_d): Updated for op_by_pieces_d
27685 (can_store_by_pieces): Use by_pieces_constfn on constfun.
27686 (store_by_pieces): Use by_pieces_constfn on constfun. Updated
27687 for op_by_pieces_d change.
27688 (clear_by_pieces_1): Add a dummy argument.
27689 (clear_by_pieces): Updated for op_by_pieces_d change.
27690 (compare_by_pieces_d::compare_by_pieces_d): Likewise.
27691 (string_cst_read_str): Add a dummy argument.
27692 * expr.h (by_pieces_constfn): Add a dummy argument.
27693 (by_pieces_prev): New.
27694 * target.def (overlap_op_by_pieces_p): New target hook.
27695 * config/i386/i386.c (TARGET_OVERLAP_OP_BY_PIECES_P): New.
27696 * doc/tm.texi.in: Add TARGET_OVERLAP_OP_BY_PIECES_P.
27697 * doc/tm.texi: Regenerated.
27699 2021-04-29 Richard Biener <rguenther@suse.de>
27701 PR tree-optimization/100253
27702 * tree-vect-stmts.c (vectorizable_load): Do not assume
27703 element alignment when DR_MISALIGNMENT is -1.
27704 (vectorizable_store): Likewise.
27706 2021-04-29 Jakub Jelinek <jakub@redhat.com>
27709 * config/aarch64/aarch64.c (aarch64_add_offset_1_temporaries): Use
27710 absu_hwi instead of abs_hwi.
27712 2021-04-29 Richard Biener <rguenther@suse.de>
27714 PR middle-end/38474
27715 * tree-ssa-structalias.c (add_graph_edge): Avoid direct
27716 forwarding when indirect forwarding through ESCAPED
27719 2021-04-29 Tom de Vries <tdevries@suse.de>
27722 * internal-fn.c (expand_GOMP_SIMT_ENTER_ALLOC)
27723 (expand_GOMP_SIMT_LAST_LANE, expand_GOMP_SIMT_ORDERED_PRED)
27724 (expand_GOMP_SIMT_VOTE_ANY, expand_GOMP_SIMT_XCHG_BFLY)
27725 (expand_GOMP_SIMT_XCHG_IDX): Ensure target is assigned to.
27727 2021-04-29 Richard Biener <rguenther@suse.de>
27729 PR tree-optimization/99912
27730 * tree-ssa-dse.c (dse_dom_walker::m_need_cfg_cleanup): New.
27731 (dse_dom_walker::todo): Likewise.
27732 (dse_dom_walker::dse_optimize_stmt): Move VDEF check to the
27734 (dse_dom_walker::before_dom_children): Remove trivially
27735 dead SSA defs and schedule CFG cleanup if we removed all
27737 (pass_dse::execute): Get TODO as computed by the DOM walker
27738 and return it. Wipe dominator info earlier.
27740 2021-04-29 Richard Biener <rguenther@suse.de>
27743 * ipa-prop.c (ipcp_modif_dom_walker::before_dom_children):
27744 Track blocks to cleanup EH in new m_need_eh_cleanup.
27745 (ipcp_modif_dom_walker::cleanup_eh): New.
27746 (ipcp_transform_function): Release dominator info before
27749 2021-04-29 Martin Sebor <msebor@redhat.com>
27751 PR middle-end/100250
27752 * attribs.c (attr_access::array_as_string): Avoid dereferencing
27753 a pointer when it's null.
27755 2021-04-29 Martin Sebor <msebor@redhat.com>
27757 * Makefile.in (OBJS): Add ipa-free-lang-data.o.
27758 * ipa-free-lang-data.cc: New file.
27759 * tree.c: Move pass free_lang_data to file above.
27760 (build_array_type_1): Declare extern.
27761 * tree.h (build_array_type_1): Declare.
27763 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
27765 * config/aarch64/aarch64-simd-builtins.def: Modify comment to
27766 make consistent with updated RTL pattern.
27767 * config/aarch64/aarch64-simd.md (aarch64_<sur>qmovn<mode>):
27768 Implement using ss_truncate and us_truncate rather than
27770 * config/aarch64/iterators.md: Remove redundant unspecs and
27771 iterator: UNSPEC_[SU]QXTN and SUQMOVN respectively.
27773 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
27775 * config/aarch64/arm_acle.h (__attribute__): Make intrinsic
27776 attributes consistent with those defined in arm_neon.h.
27778 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
27780 * config/aarch64/arm_fp16.h (__attribute__): Make intrinsic
27781 attributes consistent with those defined in arm_neon.h.
27783 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
27785 * config/aarch64/aarch64-simd-builtins.def: Add
27786 float_trunc_rodd builtin generator macros.
27787 * config/aarch64/aarch64-simd.md (aarch64_float_trunc_rodd_df):
27789 (aarch64_float_trunc_rodd_lo_v2sf): Define.
27790 (aarch64_float_trunc_rodd_hi_v4sf_le): Define.
27791 (aarch64_float_trunc_rodd_hi_v4sf_be): Define.
27792 (aarch64_float_trunc_rodd_hi_v4sf): Define.
27793 * config/aarch64/arm_neon.h (vcvtx_f32_f64): Use RTL builtin
27794 instead of inline asm.
27795 (vcvtx_high_f32_f64): Likewise.
27796 (vcvtxd_f32_f64): Likewise.
27797 * config/aarch64/iterators.md: Add FCVTXN unspec.
27799 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
27801 * config/aarch64/aarch64-simd-builtins.def: Add tbx1 builtin
27803 * config/aarch64/aarch64-simd.md (aarch64_tbx1<mode>):
27805 * config/aarch64/arm_neon.h (vqtbx1_s8): USE RTL builtin
27806 instead of inline asm.
27807 (vqtbx1_u8): Likewise.
27808 (vqtbx1_p8): Likewise.
27809 (vqtbx1q_s8): Likewise.
27810 (vqtbx1q_u8): Likewise.
27811 (vqtbx1q_p8): Likewise.
27812 (vtbx2_s8): Likewise.
27813 (vtbx2_u8): Likewise.
27814 (vtbx2_p8): Likewise.
27816 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
27818 * config/aarch64/aarch64-simd-builtins.def: Add tbl1 builtin
27820 * config/aarch64/arm_neon.h (vqtbl1_p8): Use RTL builtin
27821 instead of inline asm.
27822 (vqtbl1_s8): Likewise.
27823 (vqtbl1_u8): Likewise.
27824 (vqtbl1q_p8): Likewise.
27825 (vqtbl1q_s8): Likewise.
27826 (vqtbl1q_u8): Likewise.
27827 (vtbl1_s8): Likewise.
27828 (vtbl1_u8): Likewise.
27829 (vtbl1_p8): Likewise.
27830 (vtbl2_s8): Likewise.
27831 (vtbl2_u8): Likewise.
27832 (vtbl2_p8): Likewise.
27834 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
27836 * config/aarch64/aarch64-simd-builtins.def: Add polynomial
27837 ssri_n buitin generator macro.
27838 * config/aarch64/arm_neon.h (vsri_n_p8): Use RTL builtin
27839 instead of inline asm.
27840 (vsri_n_p16): Likewise.
27841 (vsri_n_p64): Likewise.
27842 (vsriq_n_p8): Likewise.
27843 (vsriq_n_p16): Likewise.
27844 (vsriq_n_p64): Likewise.
27846 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
27848 * config/aarch64/aarch64-simd-builtins.def: Use VALLP mode
27849 iterator for polynomial ssli_n builtin generator macro.
27850 * config/aarch64/arm_neon.h (vsli_n_p8): Use RTL builtin
27851 instead of inline asm.
27852 (vsli_n_p16): Likewise.
27853 (vsliq_n_p8): Likewise.
27854 (vsliq_n_p16): Likewise.
27855 * config/aarch64/iterators.md: Define VALLP mode iterator.
27857 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
27859 * config/aarch64/aarch64-simd-builtins.def: Use VDQV_L
27860 iterator to generate [su]adalp RTL builtins.
27861 * config/aarch64/aarch64-simd.md: Use VDQV_L iterator in
27862 [su]adalp RTL pattern.
27863 * config/aarch64/arm_neon.h (vpadal_s32): Use RTL builtin
27864 instead of inline asm.
27865 (vpadal_u32): Likewise.
27867 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
27869 * config/aarch64/aarch64-simd-builtins.def: Add [su]addlp
27870 builtin generator macros.
27871 * config/aarch64/aarch64-simd.md (aarch64_<su>addlp<mode>):
27873 * config/aarch64/arm_neon.h (vpaddl_s8): Use RTL builtin
27874 instead of inline asm.
27875 (vpaddl_s16): Likewise.
27876 (vpaddl_s32): Likewise.
27877 (vpaddl_u8): Likewise.
27878 (vpaddl_u16): Likewise.
27879 (vpaddl_u32): Likewise.
27880 (vpaddlq_s8): Likewise.
27881 (vpaddlq_s16): Likewise.
27882 (vpaddlq_s32): Likewise.
27883 (vpaddlq_u8): Likewise.
27884 (vpaddlq_u16): Likewise.
27885 (vpaddlq_u32): Liwewise.
27886 * config/aarch64/iterators.md: Define [SU]ADDLP unspecs with
27887 appropriate attributes.
27889 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
27891 * config/aarch64/aarch64-simd-builtins.def: Use VDQ_I iterator
27892 for aarch64_addp<mode> builtin macro generator.
27893 * config/aarch64/aarch64-simd.md: Use VDQ_I iterator in
27894 aarch64_addp<mode> RTL pattern.
27895 * config/aarch64/arm_neon.h (vpaddq_s8): Use RTL builtin
27896 instead of inline asm.
27897 (vpaddq_s16): Likewise.
27898 (vpaddq_s32): Likewise.
27899 (vpaddq_s64): Likewise.
27900 (vpaddq_u8): Likewise.
27901 (vpaddq_u16): Likewise.
27902 (vpaddq_u32): Likewise.
27903 (vpaddq_u64): Likewise.
27905 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
27907 * config/aarch64/aarch64-simd-builtins.def: Add sq[r]dmulh_n
27908 builtin generator macros.
27909 * config/aarch64/aarch64-simd.md (aarch64_sq<r>dmulh_n<mode>):
27911 * config/aarch64/arm_neon.h (vqdmulh_n_s16): Use RTL builtin
27912 instead of inline asm.
27913 (vqdmulh_n_s32): Likewise.
27914 (vqdmulhq_n_s16): Likewise.
27915 (vqdmulhq_n_s32): Likewise.
27916 (vqrdmulh_n_s16): Likewise.
27917 (vqrdmulh_n_s32): Likewise.
27918 (vqrdmulhq_n_s16): Likewise.
27919 (vqrdmulhq_n_s32): Likewise.
27921 2021-04-28 Tobias Burnus <tobias@codesourcery.com>
27923 * doc/install.texi (--enable-offload-defaulted): Document.
27925 2021-04-28 Senthil Kumar Selvaraj <saaadhu@gcc.gnu.org>
27927 * config/avr/avr-dimode.md: Turn existing patterns into
27928 define_insn_and_split style patterns where the splitter
27929 adds a clobber of the condition code register. Drop "cc"
27930 attribute. Add new patterns to match output of
27932 * config/avr/avr-fixed.md: Likewise.
27933 * config/avr/avr.c (cc_reg_rtx): New.
27934 (avr_parallel_insn_from_insns): Adjust insn count
27935 for removal of set of cc0.
27936 (avr_is_casesi_sequence): Likewise.
27937 (avr_casei_sequence_check_operands): Likewise.
27938 (avr_optimize_casesi): Likewise. Also insert
27939 new insns after jump_insn.
27940 (avr_pass_casesi::avr_rest_of_handle_casesi): Adjust
27941 for removal of set of cc0.
27942 (avr_init_expanders): Initialize cc_reg_rtx.
27943 (avr_regno_reg_class): Handle REG_CC.
27944 (cond_string): Remove usage of CC_OVERFLOW_UNUSABLE.
27945 (avr_notice_update_cc): Remove function.
27946 (ret_cond_branch): Remove usage of CC_OVERFLOW_UNUSABLE.
27947 (compare_condition): Adjust for PARALLEL with
27949 (out_shift_with_cnt): Likewise.
27950 (ashlhi3_out): Likewise.
27951 (ashrhi3_out): Likewise.
27952 (lshrhi3_out): Likewise.
27953 (avr_class_max_nregs): Return single reg for REG_CC.
27954 (avr_compare_pattern): Check for REG_CC instead
27956 (avr_reorg_remove_redundant_compare): Likewise.
27957 (avr_reorg):Adjust for PARALLEL with REG_CC clobber.
27958 (avr_hard_regno_nregs): Return single reg for REG_CC.
27959 (avr_hard_regno_mode_ok): Allow only CCmode for REG_CC.
27960 (avr_md_asm_adjust): Clobber REG_CC.
27961 (TARGET_HARD_REGNO_NREGS): Define.
27962 (TARGET_CLASS_MAX_NREGS): Define.
27963 (TARGET_MD_ASM_ADJUST): Define.
27964 * config/avr/avr.h (FIRST_PSEUDO_REGISTER): Adjust
27966 (enum reg_class): Add CC_REG class.
27967 (NOTICE_UPDATE_CC): Remove.
27968 (CC_OVERFLOW_UNUSABLE): Remove.
27969 (CC_NO_CARRY): Remove.
27970 * config/avr/avr.md: Turn existing patterns into
27971 define_insn_and_split style patterns where the splitter
27972 adds a clobber of the condition code register. Drop "cc"
27973 attribute. Add new patterns to match output of
27975 (sez): Remove unused pattern.
27977 2021-04-28 Richard Earnshaw <rearnsha@arm.com>
27980 * config/arm/arm.c (arm_hard_regno_mode_ok): Only allow VPR to be
27983 2021-04-28 Richard Sandiford <richard.sandiford@arm.com>
27986 * config/aarch64/constraints.md (Utq): Require the address to
27987 be valid for both the element mode and for V2DImode.
27989 2021-04-28 Jakub Jelinek <jakub@redhat.com>
27990 Tobias Burnus <tobias@codesourcery.com>
27992 * configure.ac (OFFLOAD_DEFAULTED): AC_DEFINE if offload-defaulted.
27993 * gcc.c (process_command): New variable.
27994 (driver::maybe_putenv_OFFLOAD_TARGETS): If OFFLOAD_DEFAULTED,
27995 set it if -foffload is defaulted.
27996 * lto-wrapper.c (OFFLOAD_TARGET_DEFAULT_ENV): Define.
27997 (compile_offload_image): If OFFLOAD_DEFAULTED and
27998 OFFLOAD_TARGET_DEFAULT is in the environment, don't fail
27999 if corresponding mkoffload can't be found.
28000 (compile_images_for_offload_targets): Likewise. Free and clear
28001 offload_names if no valid offload is found.
28002 * config.in: Regenerate.
28003 * configure: Regenerate.
28005 2021-04-28 Richard Biener <rguenther@suse.de>
28007 PR tree-optimization/100292
28008 * tree-vect-generic.c (expand_vector_condition): Do not fold
28011 2021-04-27 David Edelsohn <dje.gcc@gmail.com>
28013 * config/rs6000/aix.h (SUBTARGET_DRIVER_SELF_SPECS): New.
28014 * config/rs6000/aix64.opt (m64): New.
28017 2021-04-27 Maciej W. Rozycki <macro@orcam.me.uk>
28019 * config/vax/vax.c (print_operand_address, vax_address_cost_1)
28020 (index_term_p): Handle ASHIFT too.
28022 2021-04-27 Maciej W. Rozycki <macro@orcam.me.uk>
28024 * config/vax/builtins.md (jbb<ccss>i<mode>): Remove operand #3.
28025 (sync_lock_test_and_set<mode>): Adjust accordingly.
28026 (sync_lock_release<mode>): Likewise.
28028 2021-04-27 Maciej W. Rozycki <macro@orcam.me.uk>
28030 * config/vax/vax-protos.h (adjacent_operands_p): Remove
28032 * config/vax/vax.c (adjacent_operands_p): Remove.
28034 2021-04-27 Maciej W. Rozycki <macro@linux-mips.org>
28036 * ifcvt.c (dead_or_predicable) [!IFCVT_MODIFY_TESTS]: Fall
28037 through to the non-conditional execution case if getting the
28038 condition for conditional execution has failed.
28040 2021-04-27 Richard Sandiford <richard.sandiford@arm.com>
28042 PR middle-end/100284
28043 * gimple.c (gimple_could_trap_p_1): Remove VEC_COND_EXPR test.
28044 * tree-eh.c (operation_could_trap_p): Handle VEC_COND_EXPR rather
28045 than asserting on it.
28047 2021-04-27 David Edelsohn <dje.gcc@gmail.com>
28049 * config/rs6000/rs6000.c (rs6000_aix_precompute_tls_p): Protect
28050 with TARGET_AIX_OS.
28052 2021-04-27 David Edelsohn <dje.gcc@gmail.com>
28055 * calls.c (precompute_register_parameters): Additionally test
28056 targetm.precompute_tls_p to pre-compute argument.
28057 * config/rs6000/aix.h (TARGET_PRECOMPUTE_TLS_P): Define.
28058 * config/rs6000/rs6000.c (rs6000_aix_precompute_tls_p): New.
28059 * target.def (precompute_tls_p): New.
28060 * doc/tm.texi.in (TARGET_PRECOMPUTE_TLS_P): Add hook documentation.
28061 * doc/tm.texi: Regenerated.
28063 2021-04-27 Jakub Jelinek <jakub@redhat.com>
28066 * config/aarch64/aarch64.c (aarch64_print_operand): Cast -UINTVAL
28067 back to HOST_WIDE_INT.
28069 2021-04-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
28072 * simplify-rtx.c (simplify_context::simplify_subreg): Check the
28073 memory alignment for the outer mode.
28075 2021-04-27 H.J. Lu <hjl.tools@gmail.com>
28077 PR middle-end/90773
28078 * expr.c (op_by_pieces_d::get_usable_mode): New member function.
28079 (op_by_pieces_d::run): Cange a while loop to a do-while loop.
28081 2021-04-27 Alex Coplan <alex.coplan@arm.com>
28084 * config/arm/arm.c (arm_split_compare_and_swap): Fix up codegen
28085 with negative immediates: ensure we expand cbranchsi4_scratch
28086 correctly and ensure we satisfy its constraints.
28087 * config/arm/sync.md
28088 (@atomic_compare_and_swap<CCSI:arch><NARROW:mode>_1): Don't
28089 attempt to tie two output operands together with constraints;
28090 collapse two alternatives.
28091 (@atomic_compare_and_swap<CCSI:arch><SIDI:mode>_1): Likewise.
28092 * config/arm/thumb1.md (cbranchsi4_neg_late): New.
28094 2021-04-27 Jakub Jelinek <jakub@redhat.com>
28097 * config/aarch64/predicates.md (aarch64_sub_immediate,
28098 aarch64_plus_immediate): Use -UINTVAL instead of -INTVAL.
28099 * config/aarch64/aarch64.md (casesi, rotl<mode>3): Likewise.
28100 * config/aarch64/aarch64.c (aarch64_print_operand,
28101 aarch64_split_atomic_op, aarch64_expand_subvti): Likewise.
28103 2021-04-27 Jakub Jelinek <jakub@redhat.com>
28105 PR tree-optimization/100239
28106 * tree-vect-generic.c (lower_vec_perm): Don't accept constant
28107 permutations with all indices from the first zero element as vec_shl.
28109 2021-04-27 Jakub Jelinek <jakub@redhat.com>
28111 PR rtl-optimization/100254
28112 * cfgcleanup.c (outgoing_edges_match): Check REG_EH_REGION on
28113 last1 and last2 insns rather than BB_END (bb1) and BB_END (bb2) insns.
28115 2021-04-27 Richard Biener <rguenther@suse.de>
28117 PR tree-optimization/99912
28118 * passes.def: Add comment about new TODO_remove_unused_locals.
28119 * tree-stdarg.c (pass_data_stdarg): Run TODO_remove_unused_locals
28122 2021-04-27 Richard Biener <rguenther@suse.de>
28124 PR tree-optimization/99912
28125 * passes.def (pass_all_optimizations): Add pass_dse before
28126 the first pass_dce, move the first pass_dse before the
28127 pass_dce following pass_pre.
28129 2021-04-27 Jakub Jelinek <jakub@redhat.com>
28131 PR tree-optimization/95527
28132 * generic-match-head.c: Include tm.h.
28133 * gimple-match-head.c: Include tm.h.
28134 * match.pd (CLZ == INTEGER_CST): Don't use
28135 #ifdef CLZ_DEFINED_VALUE_AT_ZERO, only test CLZ_DEFINED_VALUE_AT_ZERO
28136 if clz == CFN_CLZ. Add missing val declaration.
28137 (CTZ cmp CST): New simplifications.
28139 2021-04-27 Jakub Jelinek <jakub@redhat.com>
28141 PR tree-optimization/96696
28142 * expr.c (expand_expr_divmod): New function.
28143 (expand_expr_real_2) <case TRUNC_DIV_EXPR>: Use it for truncations and
28144 divisions. Formatting fixes.
28145 <case MULT_EXPR>: Optimize x / y * y as x - x % y if the latter is
28148 2021-04-27 Martin Jambor <mjambor@suse.cz>
28151 * ipa-param-manipulation.c (ipa_param_adjustments::modify_call):
28152 If removing a call statement LHS SSA name, release it.
28154 2021-04-27 Richard Earnshaw <rearnsha@arm.com>
28157 * config/arm/arm.c (THUMB2_WORK_REGS): Check PIC_OFFSET_TABLE_REGNUM
28158 is valid before including it in the mask.
28160 2021-04-27 Richard Sandiford <richard.sandiford@arm.com>
28163 * config/aarch64/aarch64.c (aarch64_comp_type_attributes): Handle
28166 2021-04-27 Richard Biener <rguenther@suse.de>
28168 PR tree-optimization/100051
28169 * tree-ssa-alias.c (indirect_ref_may_alias_decl_p): Add
28170 disambiguator based on access size vs. decl size.
28172 2021-04-27 Richard Biener <rguenther@suse.de>
28174 PR tree-optimization/100278
28175 * tree-ssa-pre.c (compute_avail): Give up when we cannot
28176 adjust TBAA beacuse of mismatching bases.
28178 2021-04-27 Jakub Jelinek <jakub@redhat.com>
28181 * config/i386/i386.md (*<insn><mode>3_mask, *<insn><mode>3_mask_1):
28182 For any_rotate define_insn_split and following splitters, use
28183 SWI iterator instead of SWI48.
28185 2021-04-27 Richard Biener <rguenther@suse.de>
28187 PR tree-optimization/99776
28188 * match.pd (bit_field_ref (ctor)): Relax element extract
28189 type compatibility checks.
28191 2021-04-27 Cui,Lili <lili.cui@intel.com>
28193 * common/config/i386/i386-common.c (processor_names):
28194 Sync processor_names with processor_type.
28195 * config/i386/i386-options.c (processor_cost_table):
28196 Sync processor_cost_table with processor_type.
28198 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
28200 * value-range.cc (irange::irange_set_1bit_anti_range): Add assert.
28201 (irange::set): Call irange_set_1bit_anti_range for handling all
28202 1-bit ranges. Fall through on ~[MIN,MAX].
28204 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
28206 * value-range.cc (irange::legacy_num_pairs): Remove.
28207 (irange::invert): Change gcc_assert to gcc_checking_assert.
28208 * value-range.h (irange::num_pairs): Adjust for a cached
28209 num_pairs(). Also, rename all gcc_assert's to
28210 gcc_checking_assert's.
28212 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
28214 * value-range.cc (irange::operator=): Set m_kind.
28215 (irange::copy_to_legacy): Handle varying and undefined sources
28216 as a legacy copy since they can be easily copied.
28217 (irange::irange_set): Set m_kind.
28218 (irange::irange_set_anti_range): Same.
28219 (irange::set): Rename normalize_min_max to normalize_kind.
28220 (irange::verify_range): Adjust for multi-ranges having the
28222 (irange::irange_union): Set m_kind.
28223 (irange::irange_intersect): Same.
28224 (irange::invert): Same.
28225 * value-range.h (irange::kind): Always return m_kind.
28226 (irange::varying_p): Rename to...
28227 (irange::varying_comptaible_p): ...this.
28228 (irange::undefined_p): Only look at m_kind.
28229 (irange::irange): Always set VR_UNDEFINED if applicable.
28230 (irange::set_undefined): Always set VR_UNDEFINED.
28231 (irange::set_varying): Always set m_kind to VR_VARYING.
28232 (irange::normalize_min_max): Rename to...
28233 (irange::normalize_kind): ...this.
28235 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
28237 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::set_ssa_range_info):
28238 Adjust for constant_p including varying_p.
28239 * tree-vrp.c (vrp_prop::finalize): Same.
28240 (determine_value_range): Same.
28241 * vr-values.c (vr_values::range_of_expr): Same.
28242 * value-range.cc (irange::symbolic_p): Do not check varying_p.
28243 (irange::constant_p): Same.
28245 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
28247 * value-range.cc (irange::legacy_lower_bound): Replace
28248 !undefined_p check with num_ranges > 0.
28249 (irange::legacy_upper_bound): Same.
28250 * value-range.h (irange::type): Same.
28251 (irange::lower_bound): Same.
28252 (irange::upper_bound): Same.
28254 2021-04-26 Richard Biener <rguenther@suse.de>
28256 PR tree-optimization/99956
28257 * gimple-loop-interchange.cc (compute_access_stride):
28258 Try instantiating the access in a shallower loop nest
28259 if instantiating failed.
28260 (compute_access_strides): Pass adjustable loop_nest
28261 to compute_access_stride.
28263 2021-04-26 Christophe Lyon <christophe.lyon@linaro.org>
28265 * doc/sourcebuild.texi (arm_cmse_hw): Document.
28267 2021-04-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
28269 * config/aarch64/iterators.md (vwcore): Handle V4BF, V8BF.
28271 2021-04-26 Thomas Schwinge <thomas@codesourcery.com>
28272 Nathan Sidwell <nathan@codesourcery.com>
28273 Tom de Vries <vries@codesourcery.com>
28274 Julian Brown <julian@codesourcery.com>
28275 Kwok Cheung Yeung <kcy@codesourcery.com>
28277 * omp-offload.c (oacc_validate_dims): Implement
28278 '-Wopenacc-parallelism'.
28279 * doc/invoke.texi (-Wopenacc-parallelism): Document.
28281 2021-04-26 Richard Biener <rguenther@suse.de>
28283 * tree-cfg.h (gimplify_build1): Remove.
28284 (gimplify_build2): Likewise.
28285 (gimplify_build3): Likewise.
28286 * tree-cfg.c (gimplify_build1): Move to tree-vect-generic.c.
28287 (gimplify_build2): Likewise.
28288 (gimplify_build3): Likewise.
28289 * tree-vect-generic.c (gimplify_build1): Move from tree-cfg.c.
28291 (gimplify_build2): Likewise.
28292 (gimplify_build3): Likewise.
28293 (tree_vec_extract): Use resimplify with following SSA edges.
28294 (expand_vector_parallel): Avoid passing NULL size/bitpos
28295 to tree_vec_extract.
28296 * expr.c (store_constructor): Deal with zero-element CTORs.
28297 * match.pd (bit_field_ref <vector CTOR>): Make sure to
28298 produce vector constants when possible.
28300 2021-04-26 Richard Biener <rguenther@suse.de>
28302 * tree-complex.c: Include gimple-fold.h.
28303 (expand_complex_addition): Use gimple_build.
28304 (expand_complex_multiplication_components): Likewise.
28305 (expand_complex_multiplication): Likewise.
28306 (expand_complex_div_straight): Likewise.
28307 (expand_complex_div_wide): Likewise.
28308 (expand_complex_division): Likewise.
28309 (expand_complex_conjugate): Likewise.
28310 (expand_complex_comparison): Likewise.
28312 2021-04-26 Richard Biener <rguenther@suse.de>
28314 * tree-ssa-phiopt.c (two_value_replacement): Remove use
28315 of legacy gimplify_buildN API.
28317 2021-04-26 Richard Biener <rguenther@suse.de>
28319 PR tree-optimization/99473
28320 * tree-ssa-phiopt.c (cond_store_replacement): Handle all
28323 2021-04-26 Richard Biener <rguenther@suse.de>
28325 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_builtin):
28326 Use replace_call_with_value.
28328 2021-04-26 Richard Biener <rguenther@suse.de>
28330 * tree-ssa-propagate.h (valid_gimple_rhs_p): Remove.
28331 (update_gimple_call): Likewise.
28332 (update_call_from_tree): Likewise.
28333 * tree-ssa-propagate.c (valid_gimple_rhs_p): Remove.
28334 (valid_gimple_call_p): Likewise.
28335 (move_ssa_defining_stmt_for_defs): Likewise.
28336 (finish_update_gimple_call): Likewise.
28337 (update_gimple_call): Likewise.
28338 (update_call_from_tree): Likewise.
28339 (propagate_tree_value_into_stmt): Use replace_call_with_value.
28340 * gimple-fold.h (update_gimple_call): Declare.
28341 * gimple-fold.c (valid_gimple_rhs_p): Move here from
28342 tree-ssa-propagate.c.
28343 (update_gimple_call): Likewise.
28344 (valid_gimple_call_p): Likewise.
28345 (finish_update_gimple_call): Likewise, and simplify.
28346 (gimplify_and_update_call_from_tree): Implement
28347 update_call_from_tree functionality, avoid excessive
28348 push/pop_gimplify_context.
28349 (gimple_fold_builtin): Use only gimplify_and_update_call_from_tree.
28350 (gimple_fold_call): Likewise.
28351 * gimple-ssa-sprintf.c (try_substitute_return_value): Likewise.
28352 * tree-ssa-ccp.c (ccp_folder::fold_stmt): Likewise.
28353 (pass_fold_builtins::execute): Likewise.
28354 (optimize_stack_restore): Use replace_call_with_value.
28355 * tree-cfg.c (fold_loop_internal_call): Likewise.
28356 * tree-ssa-dce.c (maybe_optimize_arith_overflow): Use
28357 only gimplify_and_update_call_from_tree.
28358 * tree-ssa-strlen.c (handle_builtin_strlen): Likewise.
28359 (handle_builtin_strchr): Likewise.
28360 * tsan.c: Include gimple-fold.h instead of tree-ssa-propagate.h.
28362 2021-04-26 Jakub Jelinek <jakub@redhat.com>
28365 * vmsdbgout.c (ASM_OUTPUT_DEBUG_STRING, vmsdbgout_begin_block,
28366 vmsdbgout_end_block, lookup_filename, vmsdbgout_source_line): Remove
28369 2021-04-25 liuhongt <hongtao.liu@intel.com>
28372 * config/i386/i386-builtin.def (BDESC): Change the icode of
28373 the following builtins to CODE_FOR_nothing.
28374 * config/i386/i386.c (ix86_gimple_fold_builtin): Fold
28375 IX86_BUILTIN_PCMPEQB128, IX86_BUILTIN_PCMPEQW128,
28376 IX86_BUILTIN_PCMPEQD128, IX86_BUILTIN_PCMPEQQ,
28377 IX86_BUILTIN_PCMPEQB256, IX86_BUILTIN_PCMPEQW256,
28378 IX86_BUILTIN_PCMPEQD256, IX86_BUILTIN_PCMPEQQ256,
28379 IX86_BUILTIN_PCMPGTB128, IX86_BUILTIN_PCMPGTW128,
28380 IX86_BUILTIN_PCMPGTD128, IX86_BUILTIN_PCMPGTQ,
28381 IX86_BUILTIN_PCMPGTB256, IX86_BUILTIN_PCMPGTW256,
28382 IX86_BUILTIN_PCMPGTD256, IX86_BUILTIN_PCMPGTQ256.
28383 * config/i386/sse.md (avx2_eq<mode>3): Deleted.
28384 (sse2_eq<mode>3): Ditto.
28385 (sse4_1_eqv2di3): Ditto.
28386 (sse2_gt<mode>3): Rename to ..
28387 (*sse2_gt<mode>3): .. this.
28389 2021-04-24 Iain Sandoe <iain@sandoe.co.uk>
28392 2021-04-24 Iain Sandoe <iain@sandoe.co.uk>
28395 * config/darwin.c (darwin_binds_local_p): Assume that any
28396 public symbol might be interposed for PIC code. Update function
28397 header comment to reflect current Darwin capability.
28399 2021-04-24 Iain Sandoe <iain@sandoe.co.uk>
28402 * config/darwin.c (darwin_binds_local_p): Assume that any
28403 public symbol might be interposed for PIC code. Update function
28404 header comment to reflect current Darwin capability.
28406 2021-04-24 Richard Sandiford <richard.sandiford@arm.com>
28408 * doc/sourcebuild.texi: Document no-opts and any-opts target
28411 2021-04-23 YiFei Zhu <zhuyifei1999@gmail.com>
28413 * config/bpf/bpf.h (ASM_OUTPUT_ALIGNED_BSS): Use .type and .lcomm.
28415 2021-04-23 YiFei Zhu <zhuyifei1999@gmail.com>
28417 * config/bpf/bpf.h (FUNCTION_BOUNDARY): Set to 64.
28419 2021-04-23 Uroš Bizjak <ubizjak@gmail.com>
28422 * config/i386/i386-options.c (ix86_option_override_internal):
28423 Error out when -m96bit-long-double is used with 64bit targets.
28424 * config/i386/i386.md (*pushxf_rounded): Remove pattern.
28426 2021-04-23 Martin Liska <mliska@suse.cz>
28428 * lto-wrapper.c: Remove FIXME about usage of
28429 hardware_concurrency. The function is not on par with
28432 2021-04-23 Uroš Bizjak <ubizjak@gmail.com>
28435 * config/i386/sync.md (FILD_ATOMIC/FIST_ATOMIC FP load peephole2):
28436 Copy operand 3 to operand 4. Use sse_reg_operand
28437 as operand 3 predicate.
28438 (FILD_ATOMIC/FIST_ATOMIC FP load peephole2 with mem blockage): Ditto.
28439 (LDX_ATOMIC/STX_ATOMIC FP load peephole2): Ditto.
28440 (LDX_ATOMIC/LDX_ATOMIC FP load peephole2 with mem blockage): Ditto.
28441 (FILD_ATOMIC/FIST_ATOMIC FP store peephole2):
28442 Copy operand 1 to operand 0.
28443 (FILD_ATOMIC/FIST_ATOMIC FP store peephole2 with mem blockage): Ditto.
28444 (LDX_ATOMIC/STX_ATOMIC FP store peephole2): Ditto.
28445 (LDX_ATOMIC/LDX_ATOMIC FP store peephole2 with mem blockage): Ditto.
28447 2021-04-23 Alex Coplan <alex.coplan@arm.com>
28449 PR rtl-optimization/100230
28450 * early-remat.c (early_remat::sort_candidates): Use delete[]
28451 instead of delete for array allocated with new[].
28453 2021-04-23 Richard Biener <rguenther@suse.de>
28455 * genmatch.c (lower_cond): Remove VEC_COND_EXPR special-casing.
28456 (capture_info::capture_info): Likewise.
28457 (capture_info::walk_match): Likewise.
28458 (expr::gen_transform): Likewise.
28459 (dt_simplify::gen_1): Likewise.
28460 * gimple-match-head.c (maybe_resimplify_conditional_op):
28461 Remove VEC_COND_EXPR special-casing.
28462 (gimple_simplify): Likewise.
28463 * gimple.c (gimple_could_trap_p_1): Adjust.
28464 * tree-ssa-pre.c (compute_avail): Allow VEC_COND_EXPR
28465 to participate in PRE.
28467 2021-04-23 Richard Biener <rguenther@suse.de>
28469 * cfganal.c (connect_infinite_loops_to_exit): First call
28470 add_noreturn_fake_exit_edges.
28471 * ipa-sra.c (process_scan_results): Do not call the now redundant
28472 add_noreturn_fake_exit_edges.
28473 * predict.c (tree_estimate_probability): Likewise.
28474 (rebuild_frequencies): Likewise.
28475 * store-motion.c (one_store_motion_pass): Likewise.
28477 2021-04-23 Richard Biener <rguenther@suse.de>
28479 PR tree-optimization/100222
28480 * predict.c (pass_profile::execute): Remove redundant call to
28481 mark_irreducible_loops.
28482 (report_predictor_hitrates): Likewise.
28484 2021-04-23 Richard Biener <rguenther@suse.de>
28486 * tree-ssa-loop-ivopts.c (rewrite_use_nonlinear_expr): Avoid
28487 valid_gimple_rhs_p by instead gimplifying to one.
28489 2021-04-23 Richard Biener <rguenther@suse.de>
28491 PR tree-optimization/99971
28492 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences):
28493 Always use TBAA for loads.
28495 2021-04-23 liuhongt <hongtao.liu@intel.com>
28498 * config/i386/i386-options.c (ix86_option_override_internal):
28499 Clear MASK_AVX256_SPLIT_UNALIGNED_LOAD/STORE in x_target_flags
28500 when X86_TUNE_AVX256_UNALIGNED_LOAD/STORE_OPTIMAL is enabled
28501 by target attribute.
28503 2021-04-23 David Edelsohn <dje.gcc@gmail.com>
28505 * config/rs6000/aix71.h (PREFERRED_DEBUGGING_TYPE): Change to
28507 * config/rs6000/aix72.h (PREFERRED_DEBUGGING_TYPE): Same.
28509 2021-04-22 David Edelsohn <dje.gcc@gmail.com>
28511 * config.gcc (powerpc-ibm-aix6.*): Remove.
28512 * config/rs6000/aix61.h: Delete.
28514 2021-04-22 Martin Liska <mliska@suse.cz>
28516 PR testsuite/100159
28517 PR testsuite/100192
28518 * builtins.c (expand_builtin): Fix typos and missing comments.
28519 * dwarf2out.c (gen_subprogram_die): Likewise.
28520 (gen_struct_or_union_type_die): Likewise.
28522 2021-04-22 Uroš Bizjak <ubizjak@gmail.com>
28525 * config/i386/i386-expand.c (ix86_expand_convert_uns_sidf_sse):
28526 Remove the sign with FE_DOWNWARD, where x - x = -0.0.
28528 2021-04-21 Iain Sandoe <iain@sandoe.co.uk>
28530 * config/i386/darwin.h (TARGET_64BIT): Remove definition
28531 based on TARGET_ISA_64BIT.
28532 (TARGET_64BIT_P): Remove definition based on
28533 TARGET_ISA_64BIT_P().
28535 2021-04-21 Martin Liska <mliska@suse.cz>
28538 2021-04-21 Martin Liska <mliska@suse.cz>
28540 * lto-wrapper.c (cpuset_popcount): Remove.
28541 (init_num_threads): Remove and use hardware_concurrency.
28543 2021-04-21 Martin Liska <mliska@suse.cz>
28546 * main.c (main): Call toplev::finalize in CHECKING_P mode.
28547 * ipa-modref.c (ipa_modref_c_finalize): summaries are NULL
28548 when incremental LTO linking happens.
28550 2021-04-21 Martin Liska <mliska@suse.cz>
28552 * lto-wrapper.c (run_gcc): When -flto=jobserver is used, but the
28553 makeserver cannot be detected, then use -flto=N fallback.
28555 2021-04-21 Richard Sandiford <richard.sandiford@arm.com>
28557 * acinclude.m4 (gcc_AC_INITFINI_ARRAY): When cross-compiling,
28558 default to yes for aarch64-linux-gnu.
28559 * configure: Regenerate.
28561 2021-04-21 Martin Liska <mliska@suse.cz>
28563 * lto-wrapper.c (cpuset_popcount): Remove.
28564 (init_num_threads): Remove and use hardware_concurrency.
28566 2021-04-21 Martin Liska <mliska@suse.cz>
28568 * config/i386/i386.c: Remove superfluous || TARGET_MACHO
28569 which remains to be '(... || 0)' and clang complains about it.
28570 * dwarf2out.c (AT_vms_delta): Declare conditionally.
28571 (add_AT_vms_delta): Likewise.
28572 * tree.c (fld_simplified_type): Use rather more common pattern
28573 for disabling of something (#if 0).
28574 (get_tree_code_name): Likewise.
28575 (verify_type_variant): Likewise.
28577 2021-04-21 Martin Liska <mliska@suse.cz>
28579 * config/i386/i386-expand.c (decide_alignment): Use newly named
28580 macro TARGET_CPU_P.
28581 * config/i386/i386.c (ix86_decompose_address): Likewise.
28582 (ix86_address_cost): Likewise.
28583 (ix86_lea_outperforms): Likewise.
28584 (ix86_avoid_lea_for_addr): Likewise.
28585 (ix86_add_stmt_cost): Likewise.
28586 * config/i386/i386.h (TARGET_*): Remove.
28587 (TARGET_CPU_P): New macro.
28588 * config/i386/i386.md: Use newly named macro TARGET_CPU_P.
28589 * config/i386/x86-tune-sched-atom.c (do_reorder_for_imul): Likewise.
28590 (swap_top_of_ready_list): Likewise.
28591 (ix86_atom_sched_reorder): Likewise.
28592 * config/i386/x86-tune-sched-bd.c (ix86_bd_has_dispatch): Likewise.
28593 * config/i386/x86-tune-sched.c (ix86_adjust_cost): Likewise.
28595 2021-04-21 Martin Liska <mliska@suse.cz>
28597 * config/i386/i386-options.c (TARGET_EXPLICIT_NO_SAHF_P):
28599 (SET_TARGET_NO_SAHF): Likewise.
28600 (TARGET_EXPLICIT_PREFETCH_SSE_P): Likewise.
28601 (SET_TARGET_PREFETCH_SSE): Likewise.
28602 (TARGET_EXPLICIT_NO_TUNE_P): Likewise.
28603 (SET_TARGET_NO_TUNE): Likewise.
28604 (TARGET_EXPLICIT_NO_80387_P): Likewise.
28605 (SET_TARGET_NO_80387): Likewise.
28607 * config/i386/i386.h (TARGET_*): Remove.
28608 * opth-gen.awk: Generate new used macros.
28610 2021-04-21 Martin Liska <mliska@suse.cz>
28612 * config/i386/i386.h (PTA_*): Remove.
28613 (enum pta_flag): New.
28614 (DEF_PTA): Generate PTA_* values from i386-isa.def.
28615 * config/i386/i386-isa.def: New file.
28617 2021-04-21 Alex Coplan <alex.coplan@arm.com>
28620 * config/aarch64/aarch64-bti-insert.c (aarch64_bti_j_insn_p): New.
28621 (rest_of_insert_bti): Avoid inserting duplicate bti j insns for
28622 jump table targets.
28624 2021-04-21 H.J. Lu <hjl.tools@gmail.com>
28626 * config.gcc: Install mwaitintrin.h for i[34567]86-*-* and
28627 x86_64-*-* targets.
28628 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_MWAIT_SET):
28630 (OPTION_MASK_ISA2_MWAIT_UNSET): Likewise.
28631 (ix86_handle_option): Handle -mmwait.
28632 * config/i386/i386-builtins.c (ix86_init_mmx_sse_builtins):
28633 Replace OPTION_MASK_ISA_SSE3 with OPTION_MASK_ISA2_MWAIT on
28634 __builtin_ia32_monitor and __builtin_ia32_mwait.
28635 * config/i386/i386-options.c (isa2_opts): Add -mmwait.
28636 (ix86_valid_target_attribute_inner_p): Likewise.
28637 (ix86_option_override_internal): Enable mwait/monitor
28638 instructions for -msse3.
28639 * config/i386/i386.h (TARGET_MWAIT): New.
28640 (TARGET_MWAIT_P): Likewise.
28641 * config/i386/i386.opt: Add -mmwait.
28642 * config/i386/mwaitintrin.h: New file.
28643 * config/i386/pmmintrin.h: Include <mwaitintrin.h>.
28644 * config/i386/sse.md (sse3_mwait): Replace TARGET_SSE3 with
28646 (@sse3_monitor_<mode>): Likewise.
28647 * config/i386/x86gprintrin.h: Include <mwaitintrin.h>.
28648 * doc/extend.texi: Document mwait target attribute.
28649 * doc/invoke.texi: Document -mmwait.
28651 2021-04-21 Martin Liska <mliska@suse.cz>
28653 * config/i386/i386-options.c (DEF_ENUM): Remove it.
28654 * config/i386/i386-opts.h (DEF_ENUM): Likewise.
28655 * config/i386/stringop.def (DEF_ENUM): Likewise.
28657 2021-04-21 Martin Liska <mliska@suse.cz>
28659 * tree-cfg.c (gimple_verify_flow_info): Use qD instead
28660 of print_generic_expr.
28662 2021-04-21 Jakub Jelinek <jakub@redhat.com>
28664 PR rtl-optimization/100148
28665 * cprop.c (constprop_register): Use next_nondebug_insn instead of
28668 2021-04-21 Martin Liska <mliska@suse.cz>
28671 * cgraphunit.c (cgraph_node::analyze): Remove duplicate
28672 free_dominance_info calls.
28674 2021-04-21 Richard Biener <rguenther@suse.de>
28676 * gimple-fold.c (maybe_fold_reference): Remove is_lhs
28677 parameter (and assume it to be false).
28678 (fold_gimple_assign): Adjust, remove all callers of
28679 maybe_fold_reference calling it with is_lhs true.
28680 (gimple_fold_call): Likewise.
28681 (fold_stmt_1): Likewise.
28683 2021-04-21 Richard Biener <rguenther@suse.de>
28685 * fold-const.c (pedantic_non_lvalue_loc): Remove.
28686 (fold_binary_loc): Adjust.
28687 (fold_ternary_loc): Likewise.
28689 2021-04-21 Richard Sandiford <richard.sandiford@arm.com>
28691 PR middle-end/100130
28692 * varasm.c (get_block_for_decl): Make sure that any use of the
28693 retain attribute matches the section's retain flag.
28694 (switch_to_section): Check for retain mismatches even when
28695 changing sections, but do not warn if the given decl is the
28696 section's named.decl.
28697 (output_object_block): Pass the first decl in the block (if any)
28698 to switch_to_section.
28700 2021-04-20 H.J. Lu <hjl.tools@gmail.com>
28702 * config/i386/i386-c.c (ix86_target_macros_internal): Define
28703 __CRC32__ for -mcrc32.
28704 * config/i386/i386-options.c (ix86_option_override_internal):
28705 Enable crc32 instruction for -msse4.2.
28706 * config/i386/i386.md (sse4_2_crc32<mode>): Remove TARGET_SSE4_2
28708 (sse4_2_crc32di): Likewise.
28709 * config/i386/ia32intrin.h: Use crc32 target option for CRC32
28712 2021-04-20 Segher Boessenkool <segher@kernel.crashing.org>
28715 * config/rs6000/rs6000.c (rs6000_machine_from_flags): Do not consider
28718 2021-04-20 Martin Liska <mliska@suse.cz>
28720 * doc/invoke.texi: Fix typo.
28721 * params.opt: Likewise.
28723 2021-04-20 Martin Liska <mliska@suse.cz>
28725 * doc/invoke.texi: Document new param.
28727 2021-04-19 Andrew MacLeod <amacleod@redhat.com>
28729 PR tree-optimization/100081
28730 * gimple-range-cache.h (ranger_cache): Inherit from gori_compute
28731 rather than gori_compute_cache.
28732 * gimple-range-gori.cc (is_gimple_logical_p): Move to top of file.
28733 (range_def_chain::m_logical_depth): New member.
28734 (range_def_chain::range_def_chain): Initialize m_logical_depth.
28735 (range_def_chain::get_def_chain): Don't build defchains through more
28736 than LOGICAL_LIMIT logical expressions.
28737 * params.opt (param_ranger_logical_depth): New.
28739 2021-04-19 Richard Earnshaw <rearnsha@arm.com>
28742 * config/arm/arm.c (arm_configure_build_target): Do not strip
28743 extended FPU/SIMD feature bits from the target ISA when -mfpu
28744 is specified (partial revert of r11-8168).
28746 2021-04-19 Thomas Schwinge <thomas@codesourcery.com>
28748 * params.opt (-param=openacc-kernels=): Add.
28749 * omp-oacc-kernels-decompose.cc
28750 (pass_omp_oacc_kernels_decompose::gate): Use it.
28751 * doc/invoke.texi (-fopenacc-kernels=@var{mode}): Move...
28752 (--param): ... here, 'openacc-kernels'.
28754 2021-04-19 Martin Liska <mliska@suse.cz>
28757 * gengtype.c (finish_root_table): Align function arguments
28758 in between declaration and definition.
28760 2021-04-19 Eric Botcazou <ebotcazou@adacore.com>
28762 * config/i386/winnt.c (i386_pe_seh_cold_init): Properly deal with
28763 frames larger than the SEH maximum frame size.
28765 2021-04-18 Segher Boessenkool <segher@kernel.crashing.org>
28767 PR rtl-optimization/99927
28768 * combine.c (distribute_notes) [REG_UNUSED]: If the register already
28769 is dead, just drop it.
28771 2021-04-17 Iain Buclaw <ibuclaw@gdcproject.org>
28774 * config/i386/winnt-d.c (TARGET_D_TEMPLATES_ALWAYS_COMDAT): Define.
28775 * doc/tm.texi: Regenerate.
28776 * doc/tm.texi.in (D language and ABI): Add @hook for
28777 TARGET_D_TEMPLATES_ALWAYS_COMDAT.
28779 2021-04-17 Iain Buclaw <ibuclaw@gdcproject.org>
28781 * config/darwin-d.c (darwin_d_handle_target_object_format): New
28783 (darwin_d_register_target_info): New function.
28784 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
28785 * config/dragonfly-d.c (dragonfly_d_handle_target_object_format): New
28787 (dragonfly_d_register_target_info): New function.
28788 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
28789 * config/freebsd-d.c (freebsd_d_handle_target_object_format): New
28791 (freebsd_d_register_target_info): New function.
28792 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
28793 * config/glibc-d.c (glibc_d_handle_target_object_format): New
28795 (glibc_d_register_target_info): New function.
28796 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
28797 * config/i386/i386-d.c (ix86_d_handle_target_object_format): New
28799 (ix86_d_register_target_info): Add ix86_d_handle_target_object_format
28800 as handler for objectFormat key.
28801 * config/i386/winnt-d.c (winnt_d_handle_target_object_format): New
28803 (winnt_d_register_target_info): New function.
28804 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
28805 * config/netbsd-d.c (netbsd_d_handle_target_object_format): New
28807 (netbsd_d_register_target_info): New function.
28808 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
28809 * config/openbsd-d.c (openbsd_d_handle_target_object_format): New
28811 (openbsd_d_register_target_info): New function.
28812 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
28813 * config/pa/pa-d.c (pa_d_handle_target_object_format): New function.
28814 (pa_d_register_target_info): Add pa_d_handle_target_object_format as
28815 handler for objectFormat key.
28816 * config/rs6000/rs6000-d.c (rs6000_d_handle_target_object_format): New
28818 (rs6000_d_register_target_info): Add
28819 rs6000_d_handle_target_object_format as handler for objectFormat key.
28820 * config/sol2-d.c (solaris_d_handle_target_object_format): New
28822 (solaris_d_register_target_info): New function.
28823 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
28825 2021-04-16 Jakub Jelinek <jakub@redhat.com>
28828 * config/aarch64/aarch64.c (aarch64_function_arg_alignment): Change
28829 abi_break argument from bool * to unsigned *, store there the pre-GCC 9
28831 (aarch64_layout_arg, aarch64_gimplify_va_arg_expr): Adjust callers.
28832 (aarch64_function_arg_regno_p): Likewise. Only emit -Wpsabi note if
28833 the old and new alignment after applying MIN/MAX to it is different.
28835 2021-04-16 Tamar Christina <tamar.christina@arm.com>
28838 * config/aarch64/aarch64-sve.md (@aarch64_sve_trn1_conv<mode>): New.
28839 * config/aarch64/aarch64.c (aarch64_expand_sve_const_pred_trn): Use new
28841 * config/aarch64/iterators.md (UNSPEC_TRN1_CONV): New.
28843 2021-04-16 Bill Schmidt <wschmidt@linux.ibm.com>
28845 * doc/extend.texi (PowerPC AltiVec/VSX Built-in Functions): Revise
28846 this section and its subsections.
28848 2021-04-16 Jakub Jelinek <jakub@redhat.com>
28851 * config/aarch64/aarch64.md (*neg_asr_si2_extr, *extrsi5_insn_di): New
28852 define_insn patterns.
28854 2021-04-16 Richard Sandiford <richard.sandiford@arm.com>
28856 PR rtl-optimization/98689
28857 * reg-notes.def (UNTYPED_CALL): New note.
28858 * combine.c (distribute_notes): Handle it.
28859 * emit-rtl.c (try_split): Likewise.
28860 * rtlanal.c (rtx_properties::try_to_add_insn): Likewise. Assume
28861 that calls with the note implicitly set all return value registers.
28862 * builtins.c (expand_builtin_apply): Add a REG_UNTYPED_CALL
28865 2021-04-16 Richard Sandiford <richard.sandiford@arm.com>
28867 PR rtl-optimization/99596
28868 * rtlanal.c (rtx_properties::try_to_add_insn): Don't add global
28869 register accesses for const calls. Assume that pure functions
28870 can only read from global registers. Ignore cases in which
28871 the stack pointer has been marked global.
28873 2021-04-16 Jakub Jelinek <jakub@redhat.com>
28876 * tree-vect-loop.c (vect_transform_loop): Don't remove just
28877 dead scalar .MASK_LOAD calls, but also dead .COND_* calls - replace
28878 them by their last argument.
28880 2021-04-15 Martin Liska <mliska@suse.cz>
28882 * doc/invoke.texi: Other params don't use it, remove it.
28884 2021-04-15 Richard Biener <rguenther@suse.de>
28886 * gimple-builder.h: Add deprecation note.
28888 2021-04-15 Richard Sandiford <richard.sandiford@arm.com>
28891 * attribs.h (restrict_type_identity_attributes_to): Declare.
28892 * attribs.c (restrict_type_identity_attributes_to): New function.
28894 2021-04-15 Richard Sandiford <richard.sandiford@arm.com>
28897 * attribs.h (affects_type_identity_attributes): Declare.
28898 * attribs.c (remove_attributes_matching): New function.
28899 (affects_type_identity_attributes): Likewise.
28901 2021-04-15 Jakub Jelinek <jakub@redhat.com>
28904 * config/aarch64/aarch64.md (*<LOGICAL:optab>_<SHIFT:optab><mode>3):
28905 Add combine splitters for *<LOGICAL:optab>_ashl<mode>3 with
28906 ZERO_EXTEND, SIGN_EXTEND or AND.
28908 2021-04-14 Richard Sandiford <richard.sandiford@arm.com>
28910 PR rtl-optimization/99929
28911 * rtl.h (same_vector_encodings_p): New function.
28912 * cse.c (exp_equiv_p): Check that CONST_VECTORs have the same encoding.
28913 * cselib.c (rtx_equal_for_cselib_1): Likewise.
28914 * jump.c (rtx_renumbered_equal_p): Likewise.
28915 * lra-constraints.c (operands_match_p): Likewise.
28916 * reload.c (operands_match_p): Likewise.
28917 * rtl.c (rtx_equal_p_cb, rtx_equal_p): Likewise.
28919 2021-04-14 Richard Sandiford <richard.sandiford@arm.com>
28921 * print-rtl.c (rtx_writer::print_rtx_operand_codes_E_and_V): Print
28922 more information about variable-length CONST_VECTORs.
28924 2021-04-14 Vladimir N. Makarov <vmakarov@redhat.com>
28926 PR rtl-optimization/100066
28927 * lra-constraints.c (split_reg): Check paradoxical_subreg_p for
28928 ordered modes when choosing splitting mode for hard reg.
28930 2021-04-14 Richard Sandiford <richard.sandiford@arm.com>
28933 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector_sel):
28935 (aarch64_expand_sve_const_vector): Use it for nelts_per_pattern==2.
28937 2021-04-14 Andreas Krebbel <krebbel@linux.ibm.com>
28939 * config/s390/s390-builtins.def (O_M5, O_M12, ...): Add new macros
28940 for mask operand types.
28941 (s390_vec_permi_s64, s390_vec_permi_b64, s390_vec_permi_u64)
28942 (s390_vec_permi_dbl, s390_vpdi): Use the M5 type for the immediate
28944 (s390_vec_msum_u128, s390_vmslg): Use the M12 type for the
28946 * config/s390/s390.c (s390_const_operand_ok): Check the new
28947 operand types and generate a list of valid values.
28949 2021-04-14 Iain Buclaw <ibuclaw@gdcproject.org>
28951 * doc/tm.texi: Regenerate.
28952 * doc/tm.texi.in (D language and ABI): Add @hook for
28953 TARGET_D_REGISTER_OS_TARGET_INFO.
28955 2021-04-14 Iain Buclaw <ibuclaw@gdcproject.org>
28957 * config/aarch64/aarch64-d.c (aarch64_d_handle_target_float_abi): New
28959 (aarch64_d_register_target_info): New function.
28960 * config/aarch64/aarch64-protos.h (aarch64_d_register_target_info):
28962 * config/aarch64/aarch64.h (TARGET_D_REGISTER_CPU_TARGET_INFO):
28964 * config/arm/arm-d.c (arm_d_handle_target_float_abi): New function.
28965 (arm_d_register_target_info): New function.
28966 * config/arm/arm-protos.h (arm_d_register_target_info): Declare.
28967 * config/arm/arm.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
28968 * config/i386/i386-d.c (ix86_d_handle_target_float_abi): New function.
28969 (ix86_d_register_target_info): New function.
28970 * config/i386/i386-protos.h (ix86_d_register_target_info): Declare.
28971 * config/i386/i386.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
28972 * config/mips/mips-d.c (mips_d_handle_target_float_abi): New function.
28973 (mips_d_register_target_info): New function.
28974 * config/mips/mips-protos.h (mips_d_register_target_info): Declare.
28975 * config/mips/mips.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
28976 * config/pa/pa-d.c (pa_d_handle_target_float_abi): New function.
28977 (pa_d_register_target_info): New function.
28978 * config/pa/pa-protos.h (pa_d_register_target_info): Declare.
28979 * config/pa/pa.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
28980 * config/riscv/riscv-d.c (riscv_d_handle_target_float_abi): New
28982 (riscv_d_register_target_info): New function.
28983 * config/riscv/riscv-protos.h (riscv_d_register_target_info): Declare.
28984 * config/riscv/riscv.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
28985 * config/rs6000/rs6000-d.c (rs6000_d_handle_target_float_abi): New
28987 (rs6000_d_register_target_info): New function.
28988 * config/rs6000/rs6000-protos.h (rs6000_d_register_target_info):
28990 * config/rs6000/rs6000.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
28991 * config/s390/s390-d.c (s390_d_handle_target_float_abi): New function.
28992 (s390_d_register_target_info): New function.
28993 * config/s390/s390-protos.h (s390_d_register_target_info): Declare.
28994 * config/s390/s390.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
28995 * config/sparc/sparc-d.c (sparc_d_handle_target_float_abi): New
28997 (sparc_d_register_target_info): New function.
28998 * config/sparc/sparc-protos.h (sparc_d_register_target_info): Declare.
28999 * config/sparc/sparc.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
29000 * doc/tm.texi: Regenerate.
29001 * doc/tm.texi.in (D language and ABI): Add @hook for
29002 TARGET_D_REGISTER_CPU_TARGET_INFO.
29004 2021-04-14 Iain Buclaw <ibuclaw@gdcproject.org>
29006 * config/i386/i386-d.c (ix86_d_has_stdcall_convention): New function.
29007 * config/i386/i386-protos.h (ix86_d_has_stdcall_convention): Declare.
29008 * config/i386/i386.h (TARGET_D_HAS_STDCALL_CONVENTION): Define.
29009 * doc/tm.texi: Regenerate.
29010 * doc/tm.texi.in (D language and ABI): Add @hook for
29011 TARGET_D_HAS_STDCALL_CONVENTION.
29013 2021-04-14 Richard Biener <rguenther@suse.de>
29015 * tree-cfg.c (verify_gimple_assign_ternary): Verify that
29016 VEC_COND_EXPRs have a gimple_val condition.
29017 * tree-ssa-propagate.c (valid_gimple_rhs_p): VEC_COND_EXPR
29018 can no longer have a GENERIC condition.
29020 2021-04-14 Richard Earnshaw <rearnsha@arm.com>
29023 * config/arm/arm.c (arm_configure_build_target): Strip isa_all_fpbits
29024 from the isa_delta when -mfpu has been used.
29025 (arm_options_perform_arch_sanity_checks): It's the architecture that
29026 lacks an FPU not the processor.
29028 2021-04-13 Richard Biener <rguenther@suse.de>
29030 PR tree-optimization/100053
29031 * tree-ssa-sccvn.c (vn_nary_op_get_predicated_value): Do
29032 not use optimistic dominance queries for backedges to validate
29034 (dominated_by_p_w_unex): Add parameter to ignore executable
29035 state on backedges.
29036 (rpo_elim::eliminate_avail): Adjust.
29038 2021-04-13 Jakub Jelinek <jakub@redhat.com>
29041 * config/aarch64/aarch64.md (*aarch64_bfxil<mode>_extr,
29042 *aarch64_bfxilsi_extrdi): New define_insn patterns.
29044 2021-04-13 Jakub Jelinek <jakub@redhat.com>
29047 * simplify-rtx.c (simplify_immed_subreg): For MODE_COMPOSITE_P
29048 outermode, return NULL if the result doesn't encode back to the
29049 original byte sequence.
29050 (simplify_gen_subreg): Don't create SUBREGs from constants to
29051 MODE_COMPOSITE_P outermode.
29053 2021-04-12 Jakub Jelinek <jakub@redhat.com>
29055 PR rtl-optimization/99905
29056 * combine.c (expand_compound_operation): If pos + len > modewidth,
29057 perform the right shift by pos in inner_mode and then convert to mode,
29058 instead of trying to simplify a shift of rtx with inner_mode by pos
29059 as if it was a shift in mode.
29061 2021-04-12 Jakub Jelinek <jakub@redhat.com>
29064 * combine.c (simplify_and_const_int_1): Don't optimize varop
29065 away if it has side-effects.
29067 2021-04-12 Martin Liska <mliska@suse.cz>
29069 * doc/extend.texi: Escape @smallexample content.
29071 2021-04-12 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
29073 * config/s390/s390.md ("*movdi_31", "*movdi_64"): Add
29074 alternative in order to load a DFP zero.
29076 2021-04-12 Martin Liska <mliska@suse.cz>
29078 * doc/extend.texi: Be more precise in documentation
29079 of symver attribute.
29081 2021-04-12 Martin Liska <mliska@suse.cz>
29084 * gimplify.c (gimplify_expr): Right now, we unpoison all
29085 variables before a goto <dest>. We should not do it if we are
29088 2021-04-12 Cui,Lili <lili.cui@intel.com>
29090 * common/config/i386/cpuinfo.h (get_intel_cpu): Handle
29092 * common/config/i386/i386-common.c (processor_names): Add
29094 (processor_alias_table): Add rocketlake.
29095 * common/config/i386/i386-cpuinfo.h (processor_subtypes): Add
29096 INTEL_COREI7_ROCKETLAKE.
29097 * config.gcc: Add -march=rocketlake.
29098 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
29100 * config/i386/i386-options.c (m_ROCKETLAKE) : Define.
29101 (processor_cost_table): Add rocketlake cost.
29102 * config/i386/i386.h (ix86_size_cost) : Define
29104 (processor_type) : Add PROCESSOR_ROCKETLAKE.
29105 (PTA_ROCKETLAKE): Ditto.
29106 * doc/extend.texi: Add rocketlake.
29107 * doc/invoke.texi: Add rocketlake.
29109 2021-04-12 Cui,Lili <lili.cui@intel.com>
29111 * config/i386/i386.h (PTA_ALDERLAKE): Change alderlake ISA list.
29112 * config/i386/i386-options.c (m_CORE_AVX2): Add m_ALDERLAKE.
29113 * common/config/i386/cpuinfo.h (get_intel_cpu): Add AlderLake model.
29114 * doc/invoke.texi: Change alderlake ISA list.
29116 2021-04-11 Hafiz Abid Qadeer <abidh@codesourcery.com>
29118 PR middle-end/98088
29119 * omp-expand.c (expand_oacc_collapse_init): Update condition in
29122 2021-04-10 H.J. Lu <hjl.tools@gmail.com>
29125 * config/i386/serializeintrin.h (_serialize): Defined as macro.
29127 2021-04-10 Jakub Jelinek <jakub@redhat.com>
29130 * expr.c (expand_expr_addr_expr_1): Test is_global_var rather than
29131 just TREE_STATIC on COMPOUND_LITERAL_EXPR_DECLs.
29133 2021-04-10 Jakub Jelinek <jakub@redhat.com>
29135 PR middle-end/99989
29136 * gimple-ssa-warn-alloca.c
29137 (alloca_type_and_limit::alloca_type_and_limit): Initialize limit to
29138 0 with integer precision unconditionally.
29140 2021-04-10 Jakub Jelinek <jakub@redhat.com>
29142 PR rtl-optimization/98601
29143 * rtlanal.c (rtx_addr_can_trap_p_1): Allow in assert unknown size
29144 not just for BLKmode, but also for VOIDmode. For STRICT_ALIGNMENT
29145 unaligned_mems handle VOIDmode like BLKmode.
29147 2021-04-10 Jan Hubicka <hubicka@ucw.cz>
29150 * tree.c (free_lang_data_in_decl): Do not release body of
29151 declare_variant_alt.
29153 2021-04-09 Richard Sandiford <richard.sandiford@arm.com>
29155 * config/aarch64/aarch64.c (aarch64_option_restore): If the
29156 architecture was specified explicitly and the tuning wasn't,
29157 tune for the architecture rather than the configured default CPU.
29159 2021-04-09 Richard Sandiford <richard.sandiford@arm.com>
29161 * config/aarch64/aarch64.md (tlsdesc_small_sve_<mode>): Use X30
29162 as the temporary register.
29164 2021-04-09 Martin Liska <mliska@suse.cz>
29166 * doc/extend.texi: Move non-target attributes on the top level.
29168 2021-04-09 Martin Liska <mliska@suse.cz>
29170 * doc/invoke.texi: Document minimum and maximum value of the
29171 argument for both supported compression algorithms.
29173 2021-04-08 David Edelsohn <dje.gcc@gmail.com>
29175 * config/rs6000/rs6000.c (rs6000_xcoff_select_section): Select
29176 TLS BSS before TLS data.
29177 * config/rs6000/xcoff.h (ASM_OUTPUT_TLS_COMMON): Use .comm.
29179 2021-04-08 Richard Sandiford <richard.sandiford@arm.com>
29181 * doc/sourcebuild.texi (stdint_types_mbig_endian): Document.
29183 2021-04-08 Richard Sandiford <richard.sandiford@arm.com>
29185 * match.pd: Extend vec_cond folds to handle shifts.
29187 2021-04-08 Maciej W. Rozycki <macro@orcam.me.uk>
29189 * config/vax/vax.md: Fix comment for `*bit<mode>' pattern's
29192 2021-04-08 Alex Coplan <alex.coplan@arm.com>
29195 * config/arm/iterators.md (MVE_vecs): New.
29196 (V_elem): Also handle V2DF.
29197 * config/arm/mve.md (*mve_mov<mode>): Rename to ...
29198 (*mve_vdup<mode>): ... this. Remove second alternative since
29199 vec_duplicate of const_int is not canonical RTL, and we don't
29200 want to match symbol_refs.
29201 (*mve_vec_duplicate<mode>): Delete (pattern is redundant).
29203 2021-04-08 Xionghu Luo <luoxhu@linux.ibm.com>
29205 * fold-const.c (fold_single_bit_test): Fix typo.
29206 * print-rtl.c (print_rtx_insn_vec): Call print_rtl_single
29209 2021-04-07 Richard Sandiford <richard.sandiford@arm.com>
29211 PR tree-optimization/97513
29212 * tree-vect-slp.c (vect_add_slp_permutation): New function,
29214 (vectorizable_slp_permutation): ...here. Detect cases in which
29215 all VEC_PERM_EXPRs are guaranteed to have the same stepped
29216 permute vector and only generate one permute vector for that case.
29217 Extend that case to handle variable-length vectors.
29219 2021-04-07 Richard Sandiford <richard.sandiford@arm.com>
29221 PR tree-optimization/99873
29222 * tree-vect-slp.c (vect_slp_prefer_store_lanes_p): New function.
29223 (vect_build_slp_instance): Don't split store groups that could
29224 use IFN_STORE_LANES.
29226 2021-04-07 Jakub Jelinek <jakub@redhat.com>
29229 * varasm.c (output_constant_pool_contents): Don't strip name encoding
29230 from XSTR (desc->sym, 0) or from label before passing those to
29233 2021-04-07 Richard Biener <rguenther@suse.de>
29235 PR tree-optimization/99954
29236 * tree-loop-distribution.c: Include tree-affine.h.
29237 (generate_memcpy_builtin): Try using tree-affine to prove
29239 (loop_distribution::classify_builtin_ldst): Always classify
29242 2021-04-07 Richard Biener <rguenther@suse.de>
29244 PR tree-optimization/99947
29245 * tree-vect-loop.c (vectorizable_induction): Pre-allocate
29246 steps vector to avoid pushing elements from the reallocated
29249 2021-04-07 Richard Biener <rguenther@suse.de>
29251 * tree-ssa-sccvn.h (print_vn_reference_ops): Declare.
29252 * tree-ssa-pre.c (print_pre_expr): Factor out VN reference operand
29254 * tree-ssa-sccvn.c (print_vn_reference_ops): ... into this new
29256 (debug_vn_reference_ops): New.
29258 2021-04-07 Bin Cheng <bin.cheng@linux.alibaba.com>
29260 PR tree-optimization/98736
29261 * tree-loop-distribution.c
29262 * (loop_distribution::bb_top_order_init):
29263 Compute RPO with programing order preserved by calling function
29264 rev_post_order_and_mark_dfs_back_seme.
29266 2021-04-06 Vladimir N. Makarov <vmakarov@redhat.com>
29269 * lra-constraints.c (split_reg): Don't check paradoxical_subreg_p.
29270 * lra-lives.c (clear_sparseset_regnos, regnos_in_sparseset_p): New
29272 (process_bb_lives): Don't update biggest mode of hard reg for
29273 implicit in multi-register group. Use the new functions for
29274 updating dead_set and unused_set by register notes.
29276 2021-04-06 Xianmiao Qu <xianmiao_qu@c-sky.com>
29278 * config/csky/csky_pipeline_ck802.md : Use insn reservation name
29281 2021-04-06 H.J. Lu <hjl.tools@gmail.com>
29283 * config/i386/x86-tune-costs.h (skylake_memcpy): Updated.
29284 (skylake_memset): Likewise.
29285 (skylake_cost): Change CLEAR_RATIO to 17.
29286 * config/i386/x86-tune.def (X86_TUNE_PREFER_KNOWN_REP_MOVSB_STOSB):
29287 Replace m_CANNONLAKE, m_ICELAKE_CLIENT, m_ICELAKE_SERVER,
29288 m_TIGERLAKE and m_SAPPHIRERAPIDS with m_SKYLAKE and m_CORE_AVX512.
29290 2021-04-06 Richard Biener <rguenther@suse.de>
29292 PR tree-optimization/99880
29293 * tree-vect-loop.c (maybe_set_vectorized_backedge_value): Only
29294 set vectorized defs of relevant PHIs.
29296 2021-04-06 Richard Biener <rguenther@suse.de>
29298 PR tree-optimization/99924
29299 * tree-vect-slp.c (vect_bb_partition_graph_r): Do not mark
29300 nodes w/o scalar stmts as visited.
29302 2021-04-06 Alex Coplan <alex.coplan@arm.com>
29305 * config/arm/arm.c (arm_libcall_uses_aapcs_base): Also use base
29306 PCS for [su]fix_optab.
29308 2021-04-03 Iain Sandoe <iain@sandoe.co.uk>
29310 * config/darwin.c (machopic_legitimize_pic_address): Check
29311 that the current pic register is one of the hard reg set
29312 before setting liveness.
29314 2021-04-03 Iain Sandoe <iain@sandoe.co.uk>
29316 * config/darwin.c (machopic_legitimize_pic_address): Fix
29317 whitespace, remove unused code.
29319 2021-04-03 Jakub Jelinek <jakub@redhat.com>
29321 PR tree-optimization/99882
29322 * gimple-ssa-store-merging.c (bswap_view_convert): Handle val with
29325 2021-04-03 Jakub Jelinek <jakub@redhat.com>
29327 PR rtl-optimization/99863
29328 * dse.c (replace_read): Drop regs_live argument. Instead of
29329 regs_live, use store_insn->fixed_regs_live if non-NULL,
29330 otherwise punt if insns sequence clobbers or sets any hard
29333 2021-04-03 Jakub Jelinek <jakub@redhat.com>
29336 * targhooks.h (default_print_patchable_function_entry_1): Declare.
29337 * targhooks.c (default_print_patchable_function_entry_1): New function,
29338 copied from default_print_patchable_function_entry with an added flags
29340 (default_print_patchable_function_entry): Rewritten into a small
29341 wrapper around default_print_patchable_function_entry_1.
29342 * config/rs6000/rs6000.c (TARGET_ASM_PRINT_PATCHABLE_FUNCTION_ENTRY):
29344 (rs6000_print_patchable_function_entry): New function.
29346 2021-04-02 Eric Botcazou <ebotcazou@adacore.com>
29348 * doc/invoke.texi (fdelete-dead-exceptions): Minor tweak.
29350 2021-04-01 Jason Merrill <jason@redhat.com>
29353 * common.opt: Document v15 and v16.
29355 2021-04-01 Richard Biener <rguenther@suse.de>
29357 PR tree-optimization/99863
29358 * gimplify.c (gimplify_init_constructor): Recompute vector
29361 2021-04-01 Jakub Jelinek <jakub@redhat.com>
29363 * doc/extend.texi (symver attribute): Fix up syntax errors
29366 2021-04-01 Jakub Jelinek <jakub@redhat.com>
29368 PR tree-optimization/96573
29369 * gimple-ssa-store-merging.c (init_symbolic_number): Handle
29370 also pointer types.
29372 2021-04-01 Richard Biener <rguenther@suse.de>
29374 PR tree-optimization/99856
29375 * tree-vect-patterns.c (vect_recog_over_widening_pattern): Promote
29376 precision to vector element precision.
29378 2021-04-01 Martin Jambor <mjambor@suse.cz>
29380 PR tree-optimization/97009
29381 * tree-sra.c (access_or_its_child_written): New function.
29382 (propagate_subaccesses_from_rhs): Use it instead of a simple grp_write
29385 2021-03-31 Jan Hubicka <hubicka@ucw.cz>
29388 * cif-code.def (USES_COMDAT_LOCAL): Make CIF_FINAL_NORMAL.
29390 2021-03-31 Pat Haugen <pthaugen@linux.ibm.com>
29393 * config/rs6000/altivec.md (xxspltiw_v4si, xxspltiw_v4sf_inst,
29394 xxspltidp_v2df_inst, xxsplti32dx_v4si_inst, xxsplti32dx_v4sf_inst,
29395 xxblend_<mode>, xxpermx_inst, xxeval): Mark prefixed.
29396 * config/rs6000/mma.md (mma_<vvi4i4i8>, mma_<avvi4i4i8>,
29397 mma_<vvi4i4i2>, mma_<avvi4i4i2>, mma_<vvi4i4>, mma_<avvi4i4>,
29398 mma_<pvi4i2>, mma_<apvi4i2>, mma_<vvi4i4i4>, mma_<avvi4i4i4>):
29400 * config/rs6000/rs6000.c (rs6000_final_prescan_insn): Adjust test.
29401 * config/rs6000/rs6000.md (define_attr "maybe_prefixed"): New.
29402 (define_attr "prefixed"): Update initializer.
29404 2021-03-31 Jakub Jelinek <jakub@redhat.com>
29407 * dwarf2out.c (debug_ranges_dwo_section): New variable.
29408 (DW_RANGES_IDX_SKELETON): Define.
29409 (struct dw_ranges): Add begin_entry and end_entry members.
29410 (DEBUG_DWO_RNGLISTS_SECTION): Define.
29411 (add_ranges_num): Adjust r initializer for addition of *_entry
29413 (add_ranges_by_labels): For -gsplit-dwarf and force_direct,
29414 set idx to DW_RANGES_IDX_SKELETON.
29415 (use_distinct_base_address_for_range): New function.
29416 (index_rnglists): Don't set r->idx if it is equal to
29417 DW_RANGES_IDX_SKELETON. Initialize r->begin_entry and
29418 r->end_entry for -gsplit-dwarf if those will be needed by
29420 (output_rnglists): Add DWO argument. If true, switch to
29421 debug_ranges_dwo_section rather than debug_ranges_section.
29422 Adjust l1/l2 label indexes. Only output the offset table when
29423 dwo is true and don't include in there the skeleton range
29424 entry if present. For -gsplit-dwarf, skip ranges that belong
29425 to the other rnglists section. Change return type from void
29426 to bool and return true if there are any range entries for
29427 the other section. For dwarf_split_debug_info use
29428 DW_RLE_startx_endx, DW_RLE_startx_length and DW_RLE_base_addressx
29429 entries instead of DW_RLE_start_end, DW_RLE_start_length and
29430 DW_RLE_base_address. Use use_distinct_base_address_for_range.
29431 (init_sections_and_labels): Initialize debug_ranges_dwo_section
29432 if -gsplit-dwarf and DWARF >= 5. Adjust ranges_section_label
29433 and range_base_label indexes.
29434 (dwarf2out_finish): Call index_rnglists earlier before finalizing
29435 .debug_addr. Never emit DW_AT_rnglists_base attribute. For
29436 -gsplit-dwarf and DWARF >= 5 call output_rnglists up to twice
29437 with different dwo arguments.
29438 (dwarf2out_c_finalize): Clear debug_ranges_dwo_section.
29440 2021-03-31 Richard Sandiford <richard.sandiford@arm.com>
29442 PR tree-optimization/98268
29443 * gimple-fold.c (maybe_canonicalize_mem_ref_addr): Call
29444 recompute_tree_invariant_for_addr_expr after successfully
29445 folding a TARGET_MEM_REF that occurs inside an ADDR_EXPR.
29447 2021-03-31 Richard Sandiford <richard.sandiford@arm.com>
29449 PR tree-optimization/99726
29450 * tree-data-ref.c (create_intersect_range_checks_index): Bail
29451 out if there is more than one access function SCEV for the loop
29454 2021-03-31 Richard Sandiford <richard.sandiford@arm.com>
29456 PR rtl-optimization/97141
29457 PR rtl-optimization/98726
29458 * emit-rtl.c (valid_for_const_vector_p): Return true for
29460 * rtx-vector-builder.h (rtx_vector_builder::step): Return a
29461 poly_wide_int instead of a wide_int.
29462 (rtx_vector_builder::apply_set): Take a poly_wide_int instead
29464 * rtx-vector-builder.c (rtx_vector_builder::apply_set): Likewise.
29465 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Return
29466 false for CONST_VECTORs that cannot be forced to memory.
29467 * config/aarch64/aarch64-simd.md (mov<mode>): If a CONST_VECTOR
29468 is too complex to force to memory, build it up from individual
29471 2021-03-31 Jan Hubicka <jh@suse.cz>
29474 * cgraph.c (cgraph_node::release_body): Fix overactive check.
29476 2021-03-31 Christophe Lyon <christophe.lyon@linaro.org>
29479 * config/arm/vec-common.md (mul<mode>3): Disable on iwMMXT, expect
29482 2021-03-31 H.J. Lu <hjl.tools@gmail.com>
29484 * config/i386/i386-expand.c (expand_set_or_cpymem_via_rep):
29485 For TARGET_PREFER_KNOWN_REP_MOVSB_STOSB, don't convert QImode
29487 (decide_alg): For TARGET_PREFER_KNOWN_REP_MOVSB_STOSB, use
29488 "rep movsb/stosb" only for known sizes.
29489 * config/i386/i386-options.c (processor_cost_table): Use Ice
29490 Lake cost for Cannon Lake, Ice Lake, Tiger Lake, Sapphire
29491 Rapids and Alder Lake.
29492 * config/i386/i386.h (TARGET_PREFER_KNOWN_REP_MOVSB_STOSB): New.
29493 * config/i386/x86-tune-costs.h (icelake_memcpy): New.
29494 (icelake_memset): Likewise.
29495 (icelake_cost): Likewise.
29496 * config/i386/x86-tune.def (X86_TUNE_PREFER_KNOWN_REP_MOVSB_STOSB):
29499 2021-03-31 Richard Sandiford <richard.sandiford@arm.com>
29502 * config/aarch64/aarch64.c
29503 (aarch64_vectorize_preferred_vector_alignment): Query the size
29504 of the provided SVE vector; do not assume that all SVE vectors
29505 have the same size.
29507 2021-03-31 Jan Hubicka <jh@suse.cz>
29510 * cgraph.c (cgraph_node::release_body): Remove all callers and
29512 * cgraphclones.c (cgraph_node::materialize_clone): Do not do it here.
29513 * cgraphunit.c (cgraph_node::expand): And here.
29515 2021-03-31 Martin Liska <mliska@suse.cz>
29517 * ipa-modref.c (analyze_ssa_name_flags): Fix coding style
29518 and one negated condition.
29520 2021-03-31 Jakub Jelinek <jakub@redhat.com>
29521 Richard Sandiford <richard.sandiford@arm.com>
29524 * config/aarch64/aarch64.md (*add<mode>3_poly_1): Swap Uai and Uav
29525 constraints on operands[2] and similarly 0 and rk constraints
29526 on operands[1] corresponding to that.
29528 2021-03-31 Jakub Jelinek <jakub@redhat.com>
29531 * configure.ac (HAVE_LD_BROKEN_PE_DWARF5): New AC_DEFINE if PECOFF
29532 linker doesn't support DWARF sections new in DWARF5.
29533 * config/i386/i386-options.c (ix86_option_override_internal): Default
29534 to dwarf_version 4 if HAVE_LD_BROKEN_PE_DWARF5 for TARGET_PECOFF
29536 * config.in: Regenerated.
29537 * configure: Regenerated.
29539 2021-03-30 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
29542 * config/aarch64/aarch64.c (aarch64_analyze_loop_vinfo): Check for
29543 available issue_info before using it.
29545 2021-03-30 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
29548 * config/aarch64/aarch64.md (sub<mode>3_compare1_imm): Do not allow zero
29551 2021-03-30 Xionghu Luo <luoxhu@linux.ibm.com>
29554 * config/rs6000/altivec.md (altivec_lvsl_reg): Change to ...
29555 (altivec_lvsl_reg_<mode>): ... this.
29556 (altivec_lvsr_reg): Change to ...
29557 (altivec_lvsr_reg_<mode>): ... this.
29558 * config/rs6000/predicates.md (vec_set_index_operand): New.
29559 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
29560 Enable 32bit variable vec_insert for all TARGET_VSX.
29561 * config/rs6000/rs6000.c (rs6000_expand_vector_set_var_p9):
29562 Enable 32bit variable vec_insert for p9 and above.
29563 (rs6000_expand_vector_set_var_p8): Rename to ...
29564 (rs6000_expand_vector_set_var_p7): ... this.
29565 (rs6000_expand_vector_set): Use TARGET_VSX and adjust assert
29567 * config/rs6000/vector.md (vec_set<mode>): Use vec_set_index_operand.
29568 * config/rs6000/vsx.md (xl_len_r): Use gen_altivec_lvsl_reg_di and
29569 gen_altivec_lvsr_reg_di.
29571 2021-03-30 H.J. Lu <hjl.tools@gmail.com>
29574 * config/i386/ia32intrin.h (__rdtsc): Defined as macro.
29575 (__rdtscp): Likewise.
29577 2021-03-30 Tamar Christina <tamar.christina@arm.com>
29579 PR tree-optimization/99825
29580 * tree-vect-slp-patterns.c (vect_check_evenodd_blend):
29581 Reject non-mult 2 lanes.
29583 2021-03-30 Richard Earnshaw <rearnsha@arm.com>
29586 * config/arm/arm.c (arm_file_start): Fix emission of
29587 Tag_ABI_VFP_args attribute.
29589 2021-03-30 Richard Biener <rguenther@suse.de>
29591 PR tree-optimization/99824
29592 * stor-layout.c (set_min_and_max_values_for_integral_type):
29593 Assert the precision is within the bounds of
29594 WIDE_INT_MAX_PRECISION.
29595 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
29596 the outermost component ref only to lower the access size
29597 and initialize that from the access type.
29599 2021-03-30 Richard Sandiford <richard.sandiford@arm.com>
29602 * config/aarch64/aarch64.md (mov<mode>): Pass multi-instruction
29603 CONST_INTs to aarch64_expand_mov_immediate when called after RA.
29605 2021-03-30 Mihailo Stojanovic <mihailo.stojanovic@typhoon-hil.com>
29607 * config/aarch64/aarch64.md
29608 (<optab>_trunc<fcvt_target><GPI:mode>2): Set the "arch"
29609 attribute to disambiguate between SIMD and FP variants of the
29612 2021-03-29 Jan Hubicka <hubicka@ucw.cz>
29614 * ipa-modref.c (merge_call_lhs_flags): Correct handling of deref.
29615 (analyze_ssa_name_flags): Fix typo in comment.
29617 2021-03-29 Alex Coplan <alex.coplan@arm.com>
29620 * config/aarch64/aarch64-sve-builtins.cc
29621 (function_builder::add_function): Add placeholder_p argument, use
29622 placeholder decls if this is set.
29623 (function_builder::add_unique_function): Instead of conditionally adding
29624 direct overloads, unconditionally add either a direct overload or a
29626 (function_builder::add_overloaded_function): Set placeholder_p if we're
29627 using C++ overloads. Use the obstack for string storage instead
29628 of relying on the tree nodes.
29629 (function_builder::add_overloaded_functions): Don't return early for
29630 m_direct_overloads: we need to add placeholders.
29631 * config/aarch64/aarch64-sve-builtins.h
29632 (function_builder::add_function): Add placeholder_p argument.
29634 2021-03-29 Richard Biener <rguenther@suse.de>
29636 PR tree-optimization/99807
29637 * tree-vect-slp.c (vect_slp_analyze_node_operations_1): Move
29638 assert below VEC_PERM handling.
29640 2021-03-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
29643 * config/aarch64/aarch64-simd.md (move_lo_quad_internal_<mode>): Use
29644 aarch64_simd_or_scalar_imm_zero to match zeroes. Remove pattern
29645 matching const_int 0.
29646 (move_lo_quad_internal_be_<mode>): Likewise.
29647 (move_lo_quad_<mode>): Update for the above.
29648 * config/aarch64/iterators.md (VQ_2E): Delete.
29650 2021-03-29 Jakub Jelinek <jakub@redhat.com>
29652 PR tree-optimization/99777
29653 * fold-const.c (extract_muldiv_1): For conversions, punt on casts from
29654 types other than scalar integral types.
29656 2021-03-28 David Edelsohn <dje.gcc@gmail.com>
29658 * config/rs6000/rs6000.c (rs6000_output_dwarf_dtprel): Do not add
29659 XCOFF TLS reloc decorations.
29661 2021-03-28 Gerald Pfeifer <gerald@pfeifer.com>
29663 * doc/analyzer.texi (Analyzer Internals): Update link to
29664 "A Memory Model for Static Analysis of C Programs".
29666 2021-03-26 David Edelsohn <dje.gcc@gmail.com>
29668 * config/rs6000/aix.h (ADJUST_FIELD_ALIGN): Call function.
29669 * config/rs6000/rs6000-protos.h (rs6000_special_adjust_field_align):
29671 * config/rs6000/rs6000.c (rs6000_special_adjust_field_align): New.
29672 (rs6000_special_round_type_align): Recursively check innermost first
29675 2021-03-26 Jakub Jelinek <jakub@redhat.com>
29678 * dwarf2out.h (struct dw_fde_node): Add rule18 member.
29679 * dwarf2cfi.c (dwarf2out_frame_debug_expr): When handling (set hfp sp)
29680 assignment with drap_reg active, queue reg save for hfp with offset 0
29681 and flush queued reg saves. When handling a push with rule18,
29682 defer queueing reg save for hfp and just assert the offset is 0.
29683 (scan_trace): Assert that fde->rule18 is false.
29685 2021-03-26 Vladimir Makarov <vmakarov@redhat.com>
29688 * ira-costs.c (record_reg_classes): Put case with
29689 CT_RELAXED_MEMORY adjacent to one with CT_MEMORY.
29690 * ira.c (ira_setup_alts): Ditto.
29691 * lra-constraints.c (process_alt_operands): Ditto.
29692 * recog.c (asm_operand_ok): Ditto.
29693 * reload.c (find_reloads): Ditto.
29695 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
29697 * config/aarch64/aarch64-protos.h
29698 (cpu_addrcost_table::post_modify_ld3_st3): New member variable.
29699 (cpu_addrcost_table::post_modify_ld4_st4): Likewise.
29700 * config/aarch64/aarch64.c (generic_addrcost_table): Update
29701 accordingly, using the same costs as for post_modify.
29702 (exynosm1_addrcost_table, xgene1_addrcost_table): Likewise.
29703 (thunderx2t99_addrcost_table, thunderx3t110_addrcost_table):
29704 (tsv110_addrcost_table, qdf24xx_addrcost_table): Likewise.
29705 (a64fx_addrcost_table): Likewise.
29706 (neoversev1_addrcost_table): New.
29707 (neoversev1_tunings): Use neoversev1_addrcost_table.
29708 (aarch64_address_cost): Use the new post_modify costs for CImode
29711 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
29713 * config/aarch64/aarch64.opt
29714 (-param=aarch64-loop-vect-issue-rate-niters=): New parameter.
29715 * doc/invoke.texi: Document it.
29716 * config/aarch64/aarch64-protos.h (aarch64_base_vec_issue_info)
29717 (aarch64_scalar_vec_issue_info, aarch64_simd_vec_issue_info)
29718 (aarch64_advsimd_vec_issue_info, aarch64_sve_vec_issue_info)
29719 (aarch64_vec_issue_info): New structures.
29720 (cpu_vector_cost): Write comments above the variables rather
29722 (cpu_vector_cost::issue_info): New member variable.
29723 * config/aarch64/aarch64.c: Include gimple-pretty-print.h
29724 and tree-ssa-loop-niter.h.
29725 (generic_vector_cost, a64fx_vector_cost, qdf24xx_vector_cost)
29726 (thunderx_vector_cost, tsv110_vector_cost, cortexa57_vector_cost)
29727 (exynosm1_vector_cost, xgene1_vector_cost, thunderx2t99_vector_cost)
29728 (thunderx3t110_vector_cost): Initialize issue_info to null.
29729 (neoversev1_scalar_issue_info, neoversev1_advsimd_issue_info)
29730 (neoversev1_sve_issue_info, neoversev1_vec_issue_info): New structures.
29731 (neoversev1_vector_cost): Use them.
29732 (aarch64_vec_op_count, aarch64_sve_op_count): New structures.
29733 (aarch64_vector_costs::saw_sve_only_op): New member variable.
29734 (aarch64_vector_costs::num_vector_iterations): Likewise.
29735 (aarch64_vector_costs::scalar_ops): Likewise.
29736 (aarch64_vector_costs::advsimd_ops): Likewise.
29737 (aarch64_vector_costs::sve_ops): Likewise.
29738 (aarch64_vector_costs::seen_loads): Likewise.
29739 (aarch64_simd_vec_costs_for_flags): New function.
29740 (aarch64_analyze_loop_vinfo): Initialize num_vector_iterations.
29741 Count the number of predicate operations required by SVE WHILE
29743 (aarch64_comparison_type, aarch64_multiply_add_p): New functions.
29744 (aarch64_sve_only_stmt_p, aarch64_in_loop_reduction_latency): Likewise.
29745 (aarch64_count_ops): Likewise.
29746 (aarch64_add_stmt_cost): Record whether see an SVE operation
29747 that cannot currently be implementing using Advanced SIMD.
29748 Record issue information about the scalar, Advanced SIMD
29749 and (where relevant) SVE versions of a loop.
29750 (aarch64_vec_op_count::dump): New function.
29751 (aarch64_sve_op_count::dump): Likewise.
29752 (aarch64_estimate_min_cycles_per_iter): Likewise.
29753 (aarch64_adjust_body_cost): If issue information is available,
29754 try to compare the issue rates of the various loop implementations
29755 and increase or decrease the vector body cost accordingly.
29757 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
29759 * config/aarch64/aarch64.c (aarch64_detect_vector_stmt_subtype):
29760 Assume a zero cost for induction phis.
29762 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
29764 * config/aarch64/aarch64.c (aarch64_embedded_comparison_type): New
29766 (aarch64_adjust_stmt_cost): Add the costs of embedded scalar and
29767 vector comparisons.
29769 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
29771 * config/aarch64/aarch64.c (aarch64_detect_scalar_stmt_subtype):
29773 (aarch64_add_stmt_cost): Call it.
29775 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
29777 * config/aarch64/aarch64-tuning-flags.def (matched_vector_throughput):
29778 New tuning parameter.
29779 * config/aarch64/aarch64.c (neoversev1_tunings): Use it.
29780 (aarch64_estimated_sve_vq): New function.
29781 (aarch64_vector_costs::analyzed_vinfo): New member variable.
29782 (aarch64_vector_costs::is_loop): Likewise.
29783 (aarch64_vector_costs::unrolled_advsimd_niters): Likewise.
29784 (aarch64_vector_costs::unrolled_advsimd_stmts): Likewise.
29785 (aarch64_record_potential_advsimd_unrolling): New function.
29786 (aarch64_analyze_loop_vinfo, aarch64_analyze_bb_vinfo): Likewise.
29787 (aarch64_add_stmt_cost): Call aarch64_analyze_loop_vinfo or
29788 aarch64_analyze_bb_vinfo on the first use of a costs structure.
29789 Detect whether we're vectorizing a loop for SVE that might be
29790 completely unrolled if it used Advanced SIMD instead.
29791 (aarch64_adjust_body_cost_for_latency): New function.
29792 (aarch64_finish_cost): Call it.
29794 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
29796 * config/aarch64/aarch64.c (aarch64_vector_costs): New structure.
29797 (aarch64_init_cost): New function.
29798 (aarch64_add_stmt_cost): Use aarch64_vector_costs instead of
29799 the default unsigned[3].
29800 (aarch64_finish_cost, aarch64_destroy_cost_data): New functions.
29801 (TARGET_VECTORIZE_INIT_COST): Override.
29802 (TARGET_VECTORIZE_FINISH_COST): Likewise.
29803 (TARGET_VECTORIZE_DESTROY_COST_DATA): Likewise.
29805 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
29807 * config/aarch64/aarch64.c (neoversev1_advsimd_vector_cost)
29808 (neoversev1_sve_vector_cost): New cost structures.
29809 (neoversev1_vector_cost): Likewise.
29810 (neoversev1_tunings): Use them. Enable use_new_vector_costs.
29812 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
29814 * config/aarch64/aarch64-protos.h
29815 (sve_vec_cost::scatter_store_elt_cost): New member variable.
29816 * config/aarch64/aarch64.c (generic_sve_vector_cost): Update
29817 accordingly, taking the cost from the cost of a scalar_store.
29818 (a64fx_sve_vector_cost): Likewise.
29819 (aarch64_detect_vector_stmt_subtype): Detect scatter stores.
29821 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
29823 * config/aarch64/aarch64-protos.h
29824 (simd_vec_cost::store_elt_extra_cost): New member variable.
29825 * config/aarch64/aarch64.c (generic_advsimd_vector_cost): Update
29826 accordingly, using the vec_to_scalar cost for the new field.
29827 (generic_sve_vector_cost, a64fx_advsimd_vector_cost): Likewise.
29828 (a64fx_sve_vector_cost, qdf24xx_advsimd_vector_cost): Likewise.
29829 (thunderx_advsimd_vector_cost, tsv110_advsimd_vector_cost): Likewise.
29830 (cortexa57_advsimd_vector_cost, exynosm1_advsimd_vector_cost)
29831 (xgene1_advsimd_vector_cost, thunderx2t99_advsimd_vector_cost)
29832 (thunderx3t110_advsimd_vector_cost): Likewise.
29833 (aarch64_detect_vector_stmt_subtype): Detect single-element stores.
29835 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
29837 * config/aarch64/aarch64-protos.h (simd_vec_cost::ld2_st2_permute_cost)
29838 (simd_vec_cost::ld3_st3_permute_cost): New member variables.
29839 (simd_vec_cost::ld4_st4_permute_cost): Likewise.
29840 * config/aarch64/aarch64.c (generic_advsimd_vector_cost): Update
29841 accordingly, using zero for the new costs.
29842 (generic_sve_vector_cost, a64fx_advsimd_vector_cost): Likewise.
29843 (a64fx_sve_vector_cost, qdf24xx_advsimd_vector_cost): Likewise.
29844 (thunderx_advsimd_vector_cost, tsv110_advsimd_vector_cost): Likewise.
29845 (cortexa57_advsimd_vector_cost, exynosm1_advsimd_vector_cost)
29846 (xgene1_advsimd_vector_cost, thunderx2t99_advsimd_vector_cost)
29847 (thunderx3t110_advsimd_vector_cost): Likewise.
29848 (aarch64_ld234_st234_vectors): New function.
29849 (aarch64_adjust_stmt_cost): Likewise.
29850 (aarch64_add_stmt_cost): Call aarch64_adjust_stmt_cost if using
29851 the new vector costs.
29853 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
29855 * config/aarch64/aarch64-protos.h (sve_vec_cost): Turn into a
29856 derived class of simd_vec_cost. Add information about CLAST[AB]
29857 and FADDA instructions.
29858 * config/aarch64/aarch64.c (generic_sve_vector_cost): Update
29859 accordingly, using the vec_to_scalar costs for the new fields.
29860 (a64fx_sve_vector_cost): Likewise.
29861 (aarch64_reduc_type): New function.
29862 (aarch64_sve_in_loop_reduction_latency): Likewise.
29863 (aarch64_detect_vector_stmt_subtype): Take a vinfo parameter.
29864 Use aarch64_sve_in_loop_reduction_latency to handle SVE reductions
29865 that occur in the loop body.
29866 (aarch64_add_stmt_cost): Update call accordingly.
29868 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
29870 * config/aarch64/aarch64-tuning-flags.def (use_new_vector_costs):
29872 * config/aarch64/aarch64-protos.h (simd_vec_cost): Put comments
29873 above the fields rather than to the right.
29874 (simd_vec_cost::reduc_i8_cost): New member variable.
29875 (simd_vec_cost::reduc_i16_cost): Likewise.
29876 (simd_vec_cost::reduc_i32_cost): Likewise.
29877 (simd_vec_cost::reduc_i64_cost): Likewise.
29878 (simd_vec_cost::reduc_f16_cost): Likewise.
29879 (simd_vec_cost::reduc_f32_cost): Likewise.
29880 (simd_vec_cost::reduc_f64_cost): Likewise.
29881 * config/aarch64/aarch64.c (generic_advsimd_vector_cost): Update
29882 accordingly, using the vec_to_scalar_cost for the new fields.
29883 (generic_sve_vector_cost, a64fx_advsimd_vector_cost): Likewise.
29884 (a64fx_sve_vector_cost, qdf24xx_advsimd_vector_cost): Likewise.
29885 (thunderx_advsimd_vector_cost, tsv110_advsimd_vector_cost): Likewise.
29886 (cortexa57_advsimd_vector_cost, exynosm1_advsimd_vector_cost)
29887 (xgene1_advsimd_vector_cost, thunderx2t99_advsimd_vector_cost)
29888 (thunderx3t110_advsimd_vector_cost): Likewise.
29889 (aarch64_use_new_vector_costs_p): New function.
29890 (aarch64_simd_vec_costs): New function, split out from...
29891 (aarch64_builtin_vectorization_cost): ...here.
29892 (aarch64_is_reduction): New function.
29893 (aarch64_detect_vector_stmt_subtype): Likewise.
29894 (aarch64_add_stmt_cost): Call aarch64_detect_vector_stmt_subtype if
29895 using the new vector costs.
29897 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
29900 * tree-emutls.c (get_emutls_init_templ_addr): Mark initializer of weak
29901 TLS declarations as public.
29903 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
29905 * config/aarch64/aarch64-d.c (IN_TARGET_CODE): Define.
29906 * config/arm/arm-d.c (IN_TARGET_CODE): Likewise.
29907 * config/i386/i386-d.c (IN_TARGET_CODE): Likewise.
29908 * config/mips/mips-d.c (IN_TARGET_CODE): Likewise.
29909 * config/pa/pa-d.c (IN_TARGET_CODE): Likewise.
29910 * config/riscv/riscv-d.c (IN_TARGET_CODE): Likewise.
29911 * config/rs6000/rs6000-d.c (IN_TARGET_CODE): Likewise.
29912 * config/s390/s390-d.c (IN_TARGET_CODE): Likewise.
29913 * config/sparc/sparc-d.c (IN_TARGET_CODE): Likewise.
29915 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
29918 * config.gcc (*-*-cygwin*): Add winnt-d.o
29919 (*-*-mingw*): Likewise.
29920 * config/i386/cygwin.h (EXTRA_TARGET_D_OS_VERSIONS): New macro.
29921 * config/i386/mingw32.h (EXTRA_TARGET_D_OS_VERSIONS): Likewise.
29922 * config/i386/t-cygming: Add winnt-d.o.
29923 * config/i386/winnt-d.c: New file.
29925 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
29927 * config/freebsd-d.c: Include memmodel.h.
29929 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
29932 * config.gcc (*-*-openbsd*): Add openbsd-d.o.
29933 * config/t-openbsd: Add openbsd-d.o.
29934 * config/openbsd-d.c: New file.
29936 2021-03-25 Stam Markianos-Wright <stam.markianos-wright@arm.com>
29938 PR tree-optimization/96974
29939 * tree-vect-stmts.c (vect_get_vector_types_for_stmt): Replace assert
29940 with graceful exit.
29942 2021-03-25 H.J. Lu <hjl.tools@gmail.com>
29945 2021-03-25 H.J. Lu <hjl.tools@gmail.com>
29949 * config/i386/i386.c (ix86_can_inline_p): Don't check ISA for
29950 always_inline in system headers.
29952 2021-03-25 Kewen Lin <linkw@linux.ibm.com>
29954 * tree-vect-loop.c (vect_model_reduction_cost): Init inside_cost.
29956 2021-03-25 Jakub Jelinek <jakub@redhat.com>
29959 * tree-core.h (enum operand_equal_flag): Add OEP_ADDRESS_OF_SAME_FIELD.
29960 * fold-const.c (operand_compare::operand_equal_p): Don't compare
29961 field offsets if OEP_ADDRESS_OF_SAME_FIELD.
29963 2021-03-25 H.J. Lu <hjl.tools@gmail.com>
29967 * config/i386/i386.c (ix86_can_inline_p): Don't check ISA for
29968 always_inline in system headers.
29970 2021-03-25 Richard Biener <rguenther@suse.de>
29972 PR tree-optimization/99746
29973 * tree-vect-slp-patterns.c (complex_pattern::build): Do not mark
29974 the scalar stmt as patterned. Instead set up required things
29977 2021-03-25 Xionghu Luo <luoxhu@linux.ibm.com>
29979 * config/rs6000/rs6000.c (power8_costs): Change l2 cache
29982 2021-03-24 Martin Liska <mliska@suse.cz>
29985 * common/config/i386/i386-common.c (ARRAY_SIZE): Fix off-by-one
29987 * config/i386/i386-options.c (ix86_option_override_internal):
29988 Add run-time assert.
29990 2021-03-24 Martin Jambor <mjambor@suse.cz>
29993 * ipa-cp.c (initialize_node_lattices): Mark as bottom all
29994 parameters with unknown type.
29995 (ipacp_value_safe_for_type): New function.
29996 (propagate_vals_across_arith_jfunc): Verify that the constant type
29997 can be used for a type of the formal parameter.
29998 (propagate_vals_across_ancestor): Likewise.
29999 (propagate_scalar_across_jump_function): Likewise. Pass the type
30000 also to propagate_vals_across_ancestor.
30002 2021-03-24 Christophe Lyon <christophe.lyon@linaro.org>
30005 * config/arm/mve.md (movmisalign<mode>_mve_store): Use Ux
30007 (movmisalign<mode>_mve_load): Likewise.
30009 2021-03-24 Jakub Jelinek <jakub@redhat.com>
30012 * config/arm/vec-common.md (one_cmpl<mode>2, neg<mode>2,
30013 movmisalign<mode>): Disable expanders for TARGET_REALLY_IWMMXT.
30015 2021-03-24 Alexandre Oliva <oliva@adacore.com>
30017 * doc/sourcebuild.texi (sysconf): New effective target.
30019 2021-03-24 Alexandre Oliva <oliva@adacore.com>
30021 * config/i386/predicates.md (reg_or_const_vec_operand): New.
30022 * config/i386/sse.md (ssse3_pshufbv8qi3): Add an expander for
30023 the now *-prefixed insn_and_split, turn the splitter const vec
30024 into an input for the insn, making it an ignored immediate for
30025 non-split cases, and loaded into the scratch register
30028 2021-03-23 Vladimir N. Makarov <vmakarov@redhat.com>
30031 * config/aarch64/constraints.md (Utq, UOb, UOh, UOw, UOd, UOty):
30032 Use define_relaxed_memory_constraint for them.
30034 2021-03-23 Iain Sandoe <iain@sandoe.co.uk>
30037 * config/host-darwin.c (darwin_gt_pch_use_address): Add a
30038 colon to the diagnostic message.
30040 2021-03-23 Ilya Leoshkevich <iii@linux.ibm.com>
30042 * fwprop.c (fwprop_propagation::fwprop_propagation): Look at
30044 (try_fwprop_subst_note): Use set_info instead of insn_info.
30045 (try_fwprop_subst_pattern): Likewise.
30046 (try_fwprop_subst_notes): Likewise.
30047 (try_fwprop_subst): Likewise.
30048 (forward_propagate_subreg): Likewise.
30049 (forward_propagate_and_simplify): Likewise.
30050 (forward_propagate_into): Likewise.
30051 * rtl-ssa/accesses.h (set_info::single_nondebug_use) New
30053 (set_info::single_nondebug_insn_use): Likewise.
30054 (set_info::single_phi_use): Likewise.
30055 * rtl-ssa/member-fns.inl (set_info::single_nondebug_use) New
30057 (set_info::single_nondebug_insn_use): Likewise.
30058 (set_info::single_phi_use): Likewise.
30060 2021-03-23 Christophe Lyon <christophe.lyon@linaro.org>
30062 * doc/sourcebuild.texi (arm_dsp_ok, arm_dsp): Document.
30064 2021-03-23 Jakub Jelinek <jakub@redhat.com>
30067 * config/aarch64/aarch64.c (aarch64_add_offset): Tell
30068 expand_mult to perform an unsigned rather than a signed
30071 2021-03-23 H.J. Lu <hjl.tools@gmail.com>
30074 * config/i386/cpuid.h (__cpuid): Add __volatile__.
30075 (__cpuid_count): Likewise.
30077 2021-03-23 Richard Biener <rguenther@suse.de>
30079 PR tree-optimization/99721
30080 * tree-vect-slp.c (vect_slp_analyze_node_operations):
30081 Make sure we can schedule the node.
30083 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
30085 * config/riscv/riscv.c (riscv_subword): Take endianness into
30086 account when calculating the byte offset.
30088 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
30090 * config/riscv/predicates.md (subreg_lowpart_operator): New predicate
30091 * config/riscv/riscv.md (*addsi3_extended2, *subsi3_extended2)
30092 (*negsi2_extended2, *mulsi3_extended2, *<optab>si3_mask)
30093 (*<optab>si3_mask_1, *<optab>di3_mask, *<optab>di3_mask_1)
30094 (*<optab>si3_extend_mask, *<optab>si3_extend_mask_1): Use
30095 new predicate "subreg_lowpart_operator"
30097 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
30099 * config/riscv/riscv.c (riscv_swap_instruction): New function
30100 to byteswap an SImode rtx containing an instruction.
30101 (riscv_trampoline_init): Byteswap the generated instructions
30104 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
30106 * common/config/riscv/riscv-common.c
30107 (TARGET_DEFAULT_TARGET_FLAGS): Set default endianness.
30108 * config.gcc (riscv32be-*, riscv64be-*): Set
30109 TARGET_BIG_ENDIAN_DEFAULT to 1.
30110 * config/riscv/elf.h (LINK_SPEC): Change -melf* value
30111 depending on default endianness.
30112 * config/riscv/freebsd.h (LINK_SPEC): Likewise.
30113 * config/riscv/linux.h (LINK_SPEC): Likewise.
30114 * config/riscv/riscv.c (TARGET_DEFAULT_TARGET_FLAGS): Set
30115 default endianness.
30116 * config/riscv/riscv.h (DEFAULT_ENDIAN_SPEC): New macro.
30118 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
30120 * config/riscv/elf.h (LINK_SPEC): Pass linker endianness flag.
30121 * config/riscv/freebsd.h (LINK_SPEC): Likewise.
30122 * config/riscv/linux.h (LINK_SPEC): Likewise.
30123 * config/riscv/riscv.h (ASM_SPEC): Pass -mbig-endian and
30125 (BYTES_BIG_ENDIAN): Handle big endian.
30126 (WORDS_BIG_ENDIAN): Define to BYTES_BIG_ENDIAN.
30127 * config/riscv/riscv.opt (-mbig-endian, -mlittle-endian): New
30129 * doc/invoke.texi (-mbig-endian, -mlittle-endian): Document.
30131 2021-03-23 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
30133 * regcprop.c (find_oldest_value_reg): Ask target whether
30134 different mode is fine for replacement register.
30136 2021-03-23 Aldy Hernandez <aldyh@redhat.com>
30138 PR tree-optimization/99296
30139 * value-range.cc (irange::irange_set_1bit_anti_range): New.
30140 (irange::irange_set_anti_range): Call irange_set_1bit_anti_range
30141 * value-range.h (irange::irange_set_1bit_anti_range): New.
30143 2021-03-22 Vladimir N. Makarov <vmakarov@redhat.com>
30146 * config/aarch64/constraints.md (UtQ): Use
30147 define_relaxed_memory_constraint for it.
30148 * doc/md.texi (define_relaxed_memory_constraint): Describe it.
30149 * genoutput.c (main): Process DEFINE_RELAXED_MEMORY_CONSTRAINT.
30150 * genpreds.c (constraint_data): Add bitfield is_relaxed_memory.
30151 (have_relaxed_memory_constraints): New static var.
30152 (relaxed_memory_start, relaxed_memory_end): Ditto.
30153 (add_constraint): Add arg is_relaxed_memory. Check name for
30154 relaxed memory. Set up is_relaxed_memory in constraint_data and
30155 have_relaxed_memory_constraints. Adjust calls.
30156 (choose_enum_order): Process relaxed memory.
30157 (write_tm_preds_h): Ditto.
30158 (main): Process DEFINE_RELAXED_MEMORY_CONSTRAINT.
30159 * gensupport.c (process_rtx): Process DEFINE_RELAXED_MEMORY_CONSTRAINT.
30160 * ira-costs.c (record_reg_classes): Process CT_RELAXED_MEMORY.
30161 * ira-lives.c (single_reg_class): Use
30162 insn_extra_relaxed_memory_constraint.
30163 * ira.c (ira_setup_alts): CT_RELAXED_MEMORY.
30164 * lra-constraints.c (valid_address_p): Use
30165 insn_extra_relaxed_memory_constraint instead of other memory
30167 (process_alt_operands): Process CT_RELAXED_MEMORY.
30168 (curr_insn_transform): Use insn_extra_relaxed_memory_constraint.
30169 * recog.c (asm_operand_ok, preprocess_constraints): Process
30171 * reload.c (find_reloads): Ditto.
30172 * rtl.def (DEFINE_RELAXED_MEMORY_CONSTRAINT): New.
30173 * stmt.c (parse_input_constraint): Use
30174 insn_extra_relaxed_memory_constraint.
30176 2021-03-22 Segher Boessenkool <segher@kernel.crashing.org>
30179 * ubsan.c (ubsan_instrument_float_cast): Don't test for unordered if
30182 2021-03-22 Alex Coplan <alex.coplan@arm.com>
30185 * config/arm/arm-protos.h (neon_make_constant): Add generate
30186 argument to guard emitting insns, default to true.
30187 * config/arm/arm.c (arm_legitimate_constant_p_1): Reject
30188 CONST_VECTORs which neon_make_constant can't handle.
30189 (neon_vdup_constant): Add generate argument, avoid emitting
30190 insns if it's not set.
30191 (neon_make_constant): Plumb new generate argument through.
30192 * config/arm/constraints.md (Ui): New. Use it...
30193 * config/arm/mve.md (*mve_mov<mode>): ... here.
30194 * config/arm/vec-common.md (movv8hf): Use neon_make_constant to
30195 synthesize constants.
30197 2021-03-22 Richard Biener <rguenther@suse.de>
30199 * debug.h: Add deprecation warning.
30201 2021-03-22 Richard Biener <rguenther@suse.de>
30203 PR tree-optimization/99694
30204 * tree-ssa-sccvn.c (visit_phi): Ignore edges with the
30207 2021-03-22 Kito Cheng <kito.cheng@sifive.com>
30210 * config/riscv/riscv.c (riscv_expand_block_move): Get RTL value
30211 after type checking.
30213 2021-03-22 Jakub Jelinek <jakub@redhat.com>
30217 * dwarf2out.c (get_full_len): Use get_precision rather than
30219 (add_const_value_attribute): Make sure add_AT_wide argument has
30220 precision prec rather than some very wide one.
30222 2021-03-22 Kewen Lin <linkw@linux.ibm.com>
30224 * config/rs6000/rs6000.md (*rotldi3_insert_sf,
30225 *mov<SFDF:mode><SFDF2:mode>cc_p9, floatsi<mode>2_lfiwax,
30226 floatsi<mode>2_lfiwax_mem, floatunssi<mode>2_lfiwzx,
30227 floatunssi<mode>2_lfiwzx_mem, *floatsidf2_internal,
30228 *floatunssidf2_internal, fix_trunc<mode>si2_stfiwx,
30229 fix_trunc<mode>si2_internal, fixuns_trunc<mode>si2_stfiwx,
30230 *round32<mode>2_fprs, *roundu32<mode>2_fprs,
30231 *fix_trunc<mode>si2_internal): Fix empty split condition.
30232 * config/rs6000/vsx.md (*vsx_le_undo_permute_<mode>,
30233 vsx_reduc_<VEC_reduc_name>_v2df, vsx_reduc_<VEC_reduc_name>_v4sf,
30234 *vsx_reduc_<VEC_reduc_name>_v2df_scalar,
30235 *vsx_reduc_<VEC_reduc_name>_v4sf_scalar): Likewise.
30237 2021-03-22 Xionghu Luo <luoxhu@linux.ibm.com>
30240 * config/rs6000/rs6000.c (rs6000_expand_vector_set_var_p9):
30241 Convert idx to DImode.
30242 (rs6000_expand_vector_set_var_p8): Likewise.
30244 2021-03-21 Jakub Jelinek <jakub@redhat.com>
30247 * dwarf2out.c (insert_float): Change return type from void to
30248 unsigned, handle GET_MODE_SIZE (mode) == 2 and return element size.
30249 (mem_loc_descriptor, loc_descriptor, add_const_value_attribute):
30252 2021-03-20 H.J. Lu <hjl.tools@gmail.com>
30255 * config/i386/i386.c (construct_container): Check cfun != NULL
30256 before accessing silent_p.
30258 2021-03-20 Ahamed Husni <ahamedhusni73@gmail.com>
30260 * asan.c: Fix typos in comments.
30262 2021-03-20 Vladimir N. Makarov <vmakarov@redhat.com>
30264 PR rtl-optimization/99680
30265 * lra-constraints.c (skip_contraint_modifiers): Rename to skip_constraint_modifiers.
30266 (process_address_1): Check empty constraint before using
30269 2021-03-19 Pat Haugen <pthaugen@linux.ibm.com>
30271 * config/rs6000/rs6000.c (power10_cost): New.
30272 (rs6000_option_override_internal): Set Power10 costs.
30273 (rs6000_issue_rate): Set Power10 issue rate.
30274 * config/rs6000/power10.md: Rewrite for Power10.
30276 2021-03-19 Vladimir N. Makarov <vmakarov@redhat.com>
30279 * lra-constraints.c (process_address_1): Don't use unknown
30280 constraint for address constraint.
30282 2021-03-19 Iain Sandoe <iain@sandoe.co.uk>
30285 * config.gcc (powerpc-*-darwin8): Delete the reference to
30286 the now removed darwin8.h.
30288 2021-03-19 Olivier Hainque <hainque@adacore.com>
30291 * config/vxworksae.h (VX_CPU_PREFIX): Define.
30293 2021-03-19 John David Anglin <danglin@gcc.gnu.org>
30295 * config/pa/pa.c (import_milli): Use memcpy instead of strncpy.
30297 2021-03-19 Tamar Christina <tamar.christina@arm.com>
30299 PR tree-optimization/99656
30300 * tree-vect-slp-patterns.c (linear_loads_p,
30301 complex_add_pattern::matches, is_eq_or_top,
30302 vect_validate_multiplication, complex_mul_pattern::matches,
30303 complex_fms_pattern::matches): Remove complex_perm_kinds_t.
30304 * tree-vectorizer.h: (complex_load_perm_t): Removed.
30305 (slp_tree_to_load_perm_map_t): Use complex_perm_kinds_t instead of
30306 complex_load_perm_t.
30308 2021-03-19 H.J. Lu <hjl.tools@gmail.com>
30311 * config/i386/i386-options.c (ix86_init_machine_status): Set
30313 * config/i386/i386.c (init_cumulative_args): Set silent_p to
30315 (construct_container): Return early for return and argument
30316 errors if silent_p is true.
30317 * config/i386/i386.h (machine_function): Add silent_p.
30319 2021-03-19 Jakub Jelinek <jakub@redhat.com>
30322 * config/arm/constraints.md (Ds): New constraint.
30323 * config/arm/vec-common.md (mve_vshlq_<supf><mode>): Use w,Ds
30324 constraint instead of w,Dm.
30326 2021-03-19 Andrew Stubbs <ams@codesourcery.com>
30328 * config/gcn/gcn.c (gcn_parse_amdgpu_hsa_kernel_attribute): Fix quotes
30331 2021-03-19 Eric Botcazou <ebotcazou@adacore.com>
30333 PR middle-end/99641
30334 * fold-const.c (native_encode_initializer) <CONSTRUCTOR>: For an
30335 array type, do the computation of the current position in sizetype.
30337 2021-03-18 Vladimir N. Makarov <vmakarov@redhat.com>
30340 * lra-constraints.c (process_address_1): Use lookup_constraint
30341 only for a single constraint.
30343 2021-03-18 Martin Sebor <msebor@redhat.com>
30345 PR middle-end/99502
30346 * gimple-array-bounds.cc (inbounds_vbase_memaccess_p): Rename...
30347 (inbounds_memaccess_p): ...to this. Check the ending offset of
30348 the accessed member.
30350 2021-03-18 Andrew Stubbs <ams@codesourcery.com>
30352 * config/gcn/gcn.c (gcn_parse_amdgpu_hsa_kernel_attribute): Add %< and
30353 %> quote markers to error messages.
30354 (gcn_goacc_validate_dims): Likewise.
30355 (gcn_conditional_register_usage): Remove exclaimation mark from error
30357 (gcn_vectorize_vec_perm_const): Ensure perm is fully uninitialized.
30359 2021-03-18 Jan Hubicka <hubicka@ucw.cz>
30361 * config/i386/x86-tune-costs.h (struct processor_costs): Fix costs of
30364 2021-03-18 Sinan Lin <sinan@isrc.iscas.ac.cn>
30365 Kito Cheng <kito.cheng@sifive.com>
30367 * config/riscv/riscv.c (riscv_block_move_straight): Change type
30368 to unsigned HOST_WIDE_INT for parameter and local variable with
30369 HOST_WIDE_INT type.
30370 (riscv_adjust_block_mem): Ditto.
30371 (riscv_block_move_loop): Ditto.
30372 (riscv_expand_block_move): Ditto.
30374 2021-03-18 Nick Clifton <nickc@redhat.com>
30376 * config/v850/v850.c (construct_restore_jr): Increase static
30378 (construct_save_jarl): Likewise.
30379 * config/v850/v850.h (DWARF2_DEBUGGING_INFO): Define.
30381 2021-03-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
30383 * config/aarch64/aarch64.c (aarch64_adjust_generic_arch_tuning): Define.
30384 (aarch64_override_options_internal): Use it.
30385 (generic_tunings): Add AARCH64_EXTRA_TUNE_CSE_SVE_VL_CONSTANTS to
30388 2021-03-17 Sandra Loosemore <sandra@codesourcery.com>
30390 * config/nios2/nios2.c (nios2_custom_check_insns): Clean up
30391 error message format issues.
30392 (nios2_option_override): Likewise.
30393 (nios2_expand_fpu_builtin): Likewise.
30394 (nios2_init_custom_builtins): Adjust to avoid bogus strncpy
30395 truncation warning.
30396 (nios2_expand_custom_builtin): More error message format fixes.
30397 (nios2_expand_rdwrctl_builtin): Likewise.
30398 (nios2_expand_rdprs_builtin): Likewise.
30399 (nios2_expand_eni_builtin): Likewise.
30400 (nios2_expand_builtin): Likewise.
30401 (nios2_register_custom_code): Likewise.
30402 (nios2_valid_target_attribute_rec): Likewise.
30403 (nios2_add_insn_asm): Fix uninitialized variable warning.
30405 2021-03-17 Jan Hubicka <jh@suse.cz>
30407 * config/i386/x86-tune-costs.h (struct processor_costs): Update costs
30408 of gather to match reality.
30409 * config/i386/x86-tune.def (X86_TUNE_USE_GATHER): Enable for znver3.
30411 2021-03-17 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
30413 * config/aarch64/aarch64-builtins.c (aarch64_expand_rng_builtin): Use EQ
30414 to compare against CC_REG rather than NE.
30416 2021-03-17 H.J. Lu <hjl.tools@gmail.com>
30419 * config/i386/i386.c (ix86_force_load_from_GOT_p): Support
30420 inline assembly statements.
30421 (ix86_print_operand): Update 'P' handling for -fno-plt.
30423 2021-03-17 Tamar Christina <tamar.christina@arm.com>
30426 * config/aarch64/aarch64.c
30427 (aarch64_simd_clone_compute_vecsize_and_simdlen): Remove unused var.
30429 2021-03-16 Segher Boessenkool <segher@kernel.crashing.org>
30432 * config/rs6000/predicates.md (branch_comparison_operator): Allow
30433 ordered and unordered for CCFPmode, if flag_finite_math_only.
30435 2021-03-16 Jakub Jelinek <jakub@redhat.com>
30438 * config/i386/i386-expand.c (ix86_split_lea_for_addr): Emit a MULT
30439 rather than ASHIFT.
30440 * config/i386/i386.md (mult by 1248 into ashift): New splitter.
30442 2021-03-16 Martin Liska <mliska@suse.cz>
30445 * optc-save-gen.awk: Add flag_ipa_ra to exceptions for
30446 cl_optimization_compare function.
30448 2021-03-16 Ilya Leoshkevich <iii@linux.ibm.com>
30450 * config/s390/s390.c (f_constraint_p): Treat "fv" constraints
30453 2021-03-16 Jakub Jelinek <jakub@redhat.com>
30456 * config/i386/i386.h (struct machine_function): Add
30457 has_explicit_vzeroupper bitfield.
30458 * config/i386/i386-expand.c (ix86_expand_builtin): Set
30459 cfun->machine->has_explicit_vzeroupper when expanding
30460 IX86_BUILTIN_VZEROUPPER.
30461 * config/i386/i386-features.c (rest_of_handle_insert_vzeroupper):
30462 Do the mode switching only when TARGET_VZEROUPPER, expensive
30463 optimizations turned on and not optimizing for size.
30464 (pass_insert_vzeroupper::gate): Enable even when
30465 cfun->machine->has_explicit_vzeroupper is set.
30467 2021-03-16 Jakub Jelinek <jakub@redhat.com>
30470 * config/aarch64/aarch64.c
30471 (aarch64_simd_clone_compute_vecsize_and_simdlen): If not a function
30472 definition, walk TYPE_ARG_TYPES list if non-NULL for argument types
30473 instead of DECL_ARGUMENTS. Ignore types for uniform arguments.
30475 2021-03-15 Richard Biener <rguenther@suse.de>
30477 PR tree-optimization/98834
30478 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle missing
30479 subsetting by truncating the access size.
30481 2021-03-15 Jan Hubicka <hubicka@ucw.cz>
30483 * config/i386/i386-options.c (processor_cost_table): Add znver3_cost.
30484 * config/i386/x86-tune-costs.h (znver3_cost): New gobal variable; copy
30487 2021-03-15 Martin Liska <mliska@suse.cz>
30489 * spellcheck.c: Add missing comma in initialization.
30491 2021-03-14 Uroš Bizjak <ubizjak@gmail.com>
30493 * config/i386/sse.md (*vec_extract<mode>): Merge alternative 0 with
30494 alternative 2 and alternative 1 with alternative 3 using
30495 YW register constraint.
30496 (*vec_extract<PEXTR_MODE12:mode>_zext): Merge alternatives
30497 using YW register constraint.
30498 (*vec_extractv16qi_zext): Ditto.
30499 (*vec_extractv4si): Merge alternatives 4 and 5
30500 using Yw register constraint.
30501 (*ssse3_palignr<mode>_perm): Use Yw instead of v for alternative 3.
30503 2021-03-13 Martin Sebor <msebor@redhat.com>
30505 PR tree-optimization/99489
30506 * builtins.c (gimple_call_alloc_size): Fail gracefully when argument
30507 is not a call statement.
30509 2021-03-13 Jakub Jelinek <jakub@redhat.com>
30511 PR tree-optimization/99544
30512 * match.pd (X + (X << C) -> X * (1 + (1 << C))): Don't simplify
30513 if for vector types multiplication can't be done in type's mode.
30515 2021-03-12 Eric Botcazou <ebotcazou@adacore.com>
30518 * config/sparc/constraints.md (w): Rename to...
30519 (W): ... this and ditch previous implementation.
30520 * config/sparc/sparc.md (*movdi_insn_sp64): Replace W with m.
30521 (*movdf_insn_sp64): Likewise.
30522 (*mov<VM64:mode>_insn_sp64): Likewise.
30523 * config/sparc/sync.md (*atomic_compare_and_swap<mode>_1): Replace
30525 (atomic_compare_and_swap_leon3_1): Likewise.
30526 (*atomic_compare_and_swapdi_v8plus): Likewise.
30527 * config/sparc/sparc.c (memory_ok_for_ldd): Remove useless test on
30528 architecture and add missing address validity check during LRA.
30530 2021-03-12 Tobias Burnus <tobias@codesourcery.com>
30533 * gimplify.c (omp_add_variable): Handle NULL_TREE as size
30534 occuring for assumed-size arrays in use_device_{ptr,addr}.
30536 2021-03-12 Jakub Jelinek <jakub@redhat.com>
30539 * config/i386/constraints.md (YW): New internal constraint.
30540 * config/i386/sse.md (v_Yw): Add V4TI, V2TI, V1TI and TI cases.
30541 (*<sse2_avx2>_<insn><mode>3<mask_name>,
30542 *<sse2_avx2>_uavg<mode>3<mask_name>, *abs<mode>2,
30543 *<s>mul<mode>3_highpart<mask_name>): Use <v_Yw> instead of v in
30545 (<sse2_avx2>_psadbw): Use YW instead of v in constraints.
30546 (*avx2_pmaddwd, *sse2_pmaddwd, *<code>v8hi3, *<code>v16qi3,
30547 avx2_pmaddubsw256, ssse3_pmaddubsw128): Merge last two alternatives
30548 into one, use Yw instead of former x,v.
30549 (ashr<mode>3, <insn><mode>3): Use <v_Yw> instead of x in constraints of
30550 the last alternative.
30551 (<sse2_avx2>_packsswb<mask_name>, <sse2_avx2>_packssdw<mask_name>,
30552 <sse2_avx2>_packuswb<mask_name>, <sse4_1_avx2>_packusdw<mask_name>,
30553 *<ssse3_avx2>_pmulhrsw<mode>3<mask_name>, <ssse3_avx2>_palignr<mode>,
30554 <ssse3_avx2>_pshufb<mode>3<mask_name>): Merge last two alternatives
30555 into one, use <v_Yw> instead of former x,v.
30556 (avx2_interleave_highv32qi<mask_name>,
30557 vec_interleave_highv16qi<mask_name>): Use Yw instead of v in
30558 constraints. Add && <mask_avx512bw_condition> to condition.
30559 (avx2_interleave_lowv32qi<mask_name>,
30560 vec_interleave_lowv16qi<mask_name>,
30561 avx2_interleave_highv16hi<mask_name>,
30562 vec_interleave_highv8hi<mask_name>,
30563 avx2_interleave_lowv16hi<mask_name>, vec_interleave_lowv8hi<mask_name>,
30564 avx2_pshuflw_1<mask_name>, sse2_pshuflw_1<mask_name>,
30565 avx2_pshufhw_1<mask_name>, sse2_pshufhw_1<mask_name>,
30566 avx2_<code>v16qiv16hi2<mask_name>, sse4_1_<code>v8qiv8hi2<mask_name>,
30567 *sse4_1_<code>v8qiv8hi2<mask_name>_1, <sse2_avx2>_<insn><mode>3): Use
30568 Yw instead of v in constraints.
30569 * config/i386/mmx.md (Yv_Yw): New define_mode_attr.
30570 (*mmx_<insn><mode>3, mmx_ashr<mode>3, mmx_<insn><mode>3): Use <Yv_Yw>
30571 instead of Yv in constraints.
30572 (*mmx_<insn><mode>3, *mmx_mulv4hi3, *mmx_smulv4hi3_highpart,
30573 *mmx_umulv4hi3_highpart, *mmx_pmaddwd, *mmx_<code>v4hi3,
30574 *mmx_<code>v8qi3, mmx_pack<s_trunsuffix>swb, mmx_packssdw,
30575 mmx_punpckhbw, mmx_punpcklbw, mmx_punpckhwd, mmx_punpcklwd,
30576 *mmx_uavgv8qi3, *mmx_uavgv4hi3, mmx_psadbw): Use Yw instead of Yv in
30578 (*mmx_pinsrw, *mmx_pinsrb, *mmx_pextrw, *mmx_pextrw_zext, *mmx_pextrb,
30579 *mmx_pextrb_zext): Use YW instead of Yv in constraints.
30580 (*mmx_eq<mode>3, mmx_gt<mode>3): Use x instead of Yv in constraints.
30581 (mmx_andnot<mode>3, *mmx_<code><mode>3): Split last alternative into
30582 two, one with just x, another isa avx512vl with v.
30584 2021-03-12 Martin Liska <mliska@suse.cz>
30586 * doc/invoke.texi: Add missing param documentation.
30588 2021-03-11 David Malcolm <dmalcolm@redhat.com>
30591 * Makefile.in (ANALYZER_OBJS): Add analyzer/feasible-graph.o and
30592 analyzer/trimmed-graph.o.
30593 * doc/analyzer.texi (Analyzer Paths): Rewrite description of
30594 feasibility checking to reflect new implementation.
30595 * doc/invoke.texi (-fdump-analyzer-feasibility): Document new
30597 * shortest-paths.h (shortest_paths::get_shortest_distance): New.
30599 2021-03-11 David Malcolm <dmalcolm@redhat.com>
30601 * digraph.cc (selftest::test_shortest_paths): Update
30602 shortest_paths init for new param. Add test of
30603 SPS_TO_GIVEN_TARGET.
30604 * shortest-paths.h (enum shortest_path_sense): New.
30605 (shortest_paths::shortest_paths): Add "sense" param.
30606 Update for renamings. Generalize to use "sense" param.
30607 (shortest_paths::get_shortest_path): Rename param.
30608 (shortest_paths::m_sense): New field.
30609 (shortest_paths::m_prev): Rename...
30610 (shortest_paths::m_best_edge): ...to this.
30611 (shortest_paths::get_shortest_path): Update for renamings.
30612 Conditionalize flipping of path on sense of traversal.
30614 2021-03-11 David Malcolm <dmalcolm@redhat.com>
30616 * digraph.cc (selftest::test_shortest_paths): Add test coverage
30617 for paths from B and C.
30618 * shortest-paths.h (shortest_paths::shortest_paths): Handle
30619 unreachable nodes, rather than asserting.
30621 2021-03-11 David Edelsohn <dje.gcc@gmail.com>
30624 * config/rs6000/rs6000.c (rs6000_xcoff_file_start): Don't create
30625 xcoff_tbss_section_name.
30626 * config/rs6000/xcoff.h (ASM_OUTPUT_TLS_COMMON): Use .lcomm.
30627 * xcoffout.c (xcoff_tbss_section_name): Delete.
30628 * xcoffout.h (xcoff_tbss_section_name): Delete.
30630 2021-03-11 Richard Biener <rguenther@suse.de>
30632 PR tree-optimization/99523
30633 * tree-cfg.c (dump_function_to_file): Dump SSA names
30634 w/o identifier to the decls section as well, not only those
30635 without a VAR_DECL.
30637 2021-03-11 Jakub Jelinek <jakub@redhat.com>
30640 * ipa-icf-gimple.c (func_checker::compare_gimple_call): For internal
30641 function calls with lhs fail if the lhs don't have compatible types.
30643 2021-03-11 Hans-Peter Nilsson <hp@axis.com>
30645 * config/cris/cris.h (HARD_FRAME_POINTER_REGNUM): Define.
30646 Change FRAME_POINTER_REGNUM to correspond to a new faked
30647 register faked_fp, part of GENNONACR_REGS like faked_ap.
30648 (CRIS_FAKED_REGS_CONTENTS): New helper macro.
30649 (FIRST_PSEUDO_REGISTER, FIXED_REGISTERS, CALL_USED_REGISTERS):
30650 (REG_ALLOC_ORDER, REG_CLASS_CONTENTS, REGNO_OK_FOR_BASE_P)
30651 (ELIMINABLE_REGS, REGISTER_NAMES): Adjust accordingly.
30652 * config/cris/cris.md (CRIS_FP_REGNUM): Renumber to new faked
30654 (CRIS_REAL_FP_REGNUM): New constant.
30655 * config/cris/cris.c (cris_reg_saved_in_regsave_area): Check
30656 for HARD_FRAME_POINTER_REGNUM instead of FRAME_POINTER_REGNUM.
30657 (cris_initial_elimination_offset): Handle elimination changes
30658 to HARD_FRAME_POINTER_REGNUM instead of FRAME_POINTER_REGNUM
30659 and add one from FRAME_POINTER_REGNUM to
30660 HARD_FRAME_POINTER_REGNUM.
30661 (cris_expand_prologue, cris_expand_epilogue): Emit code for
30662 hard_frame_pointer_rtx instead of frame_pointer_rtx.
30664 2021-03-10 David Edelsohn <dje.gcc@gmail.com>
30667 * config/rs6000/aix.h (ADJUST_FIELD_ALIGN): Add check for DCmode.
30668 * config/rs6000/rs6000.c (rs6000_special_round_type_align): Same.
30670 2021-03-10 Vladimir N. Makarov <vmakarov@redhat.com>
30673 * lra-constraints.c (process_address_1): Don't check unknown
30674 constraint, use X for empty constraint.
30676 2021-03-10 Alex Coplan <alex.coplan@arm.com>
30678 * config/aarch64/aarch64.c (aarch64_vfp_is_call_or_return_candidate):
30679 Fix typo in comment describing "is_ha" argument.
30681 2021-03-10 John David Anglin <danglin@gcc.gnu.org>
30683 * doc/sourcebuild.texi: Document LRA target selector.
30685 2021-03-10 David Malcolm <dmalcolm@redhat.com>
30687 * doc/ux.texi: Add subsection contrasting interactive versus
30688 batch usage of GCC.
30690 2021-03-10 Joel Hutton <joel.hutton@arm.com>
30693 * tree-vect-stmts.c (vectorizable_store): Fix scatter store mask
30695 (vectorizable_load): Fix gather load mask check condition.
30697 2021-03-10 Richard Biener <rguenther@suse.de>
30699 PR tree-optimization/99510
30700 * tree.c (check_aligned_type): Check that the candidate
30701 has TYPE_USER_ALIGN set instead of matching with the
30704 2021-03-10 Eric Botcazou <ebotcazou@adacore.com>
30706 * config/sparc/sparc.c (sparc_regmode_natural_size): Return 4 for
30707 float and vector integer modes only if the mode is not larger.
30709 2021-03-10 Hans-Peter Nilsson <hp@axis.com>
30711 * config/cris/cris.h (DWARF_FRAME_REGISTERS): Define.
30713 2021-03-09 Vladimir N. Makarov <vmakarov@redhat.com>
30715 * ira.c (ira_setup_alts, ira_get_dup_out_num): Process digital
30717 * ira-lives.c (single_reg_class): Ditto.
30719 2021-03-09 Sebastian Huber <sebastian.huber@embedded-brains.de>
30721 * config.gcc (aarch64-*-rtems*): Include general rtems.h after
30722 the architecture-specific rtems.h.
30723 (aarch64-*-rtems*): Likewise.
30724 (arm*-*-rtems*): Likewise.
30725 (epiphany-*-rtems*): Likewise.
30726 (riscv*-*-rtems*): Likewise.
30728 2021-03-09 Jakub Jelinek <jakub@redhat.com>
30730 PR tree-optimization/99305
30731 * tree-ssa-phiopt.c (conditional_replacement): Test integer_pow2p
30732 before integer_all_onesp instead of vice versa.
30734 2021-03-09 Richard Earnshaw <rearnsha@arm.com>
30736 * common/config/arm/arm-common.c (arm_config_default): Change type
30737 of 'i' to unsigned.
30739 2021-03-09 Vladimir N. Makarov <vmakarov@redhat.com>
30742 * lra-constraints.c (process_address_1): Process constraint 'g'
30743 separately and digital constraints containing more one digit.
30745 2021-03-09 Nick Clifton <nickc@redhat.com>
30747 * config/rx/rx.h (DBX_DEBUGGING_INFO): Define.
30748 (DWARF"_DEBUGGING_INFO): Define.
30750 2021-03-09 Eric Botcazou <ebotcazou@adacore.com>
30753 * calls.c (initialize_argument_information): When the argument
30754 is passed by reference, do not make a copy in a thunk only if
30755 the argument is already in memory. Remove redundant test for
30756 the case of callee copy.
30758 2021-03-09 Vladimir N. Makarov <vmakarov@redhat.com>
30761 * lra-constraints.c (process_address_1): Process 0..9 constraints
30762 in process_address_1.
30764 2021-03-09 Andreas Krebbel <krebbel@linux.ibm.com>
30766 * config/s390/s390.c (struct s390_processor processor_table):
30767 Binutils name string must not be empty.
30769 2021-03-09 Claudiu Zissulescu <claziss@synopsys.com>
30771 * config/arc/arc.c (arc_attr_type): Remove function.
30773 2021-03-09 Martin Liska <mliska@suse.cz>
30776 * config/i386/i386-options.c (ix86_option_override_internal):
30777 Set isa_flags for OPTS argument and not for the global
30780 2021-03-09 Aaron Sawdey <acsawdey@linux.ibm.com>
30782 * config/rs6000/predicates.md (ds_form_mem_operand): Check
30785 2021-03-09 Aaron Sawdey <acsawdey@linux.ibm.com>
30788 * config/rs6000/predicates.md (ds_form_mem_operand) New
30790 * config/rs6000/genfusion.pl (gen_ld_cmpi_p10) Use
30791 ds_form_mem_operand in ld/lwa patterns.
30792 * config/rs6000/fusion.md: Regenerate file.
30794 2021-03-08 Martin Sebor <msebor@redhat.com>
30796 PR middle-end/98266
30797 * gimple-array-bounds.cc (inbounds_vbase_memaccess_p): New function.
30798 (array_bounds_checker::check_array_bounds): Call it.
30800 2021-03-08 Martin Sebor <msebor@redhat.com>
30802 PR middle-end/97631
30803 * tree-ssa-strlen.c (maybe_warn_overflow): Test rawmem.
30804 (handle_builtin_stxncpy_strncat): Rename locals. Determine
30805 destination size from allocation calls. Issue a more appropriate
30807 (handle_builtin_memcpy): Pass true as rawmem to maybe_warn_overflow.
30808 (handle_builtin_memset): Same.
30810 2021-03-08 Peter Bergner <bergner@linux.ibm.com>
30813 * config/rs6000/rs6000.c (rs6000_emit_le_vsx_permute): Add an assert
30814 to ensure we do not have an Altivec style address.
30815 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): Disable if passed
30816 an Altivec style address.
30817 (*vsx_le_perm_store_<mode>): Likewise.
30818 (splitters after *vsx_le_perm_store_<mode>): Likewise.
30819 (vsx_load_<mode>): Disable special expander if passed an Altivec
30821 (vsx_store_<mode>): Likewise.
30823 2021-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
30826 * config/aarch64/predicates.md (aarch64_simd_shift_imm_vec_qi): Define.
30827 (aarch64_simd_shift_imm_vec_hi): Likewise.
30828 (aarch64_simd_shift_imm_vec_si): Likewise.
30829 (aarch64_simd_shift_imm_vec_di): Likewise.
30830 * config/aarch64/aarch64-simd.md (aarch64_shrn<mode>_insn_le): Use
30831 predicate from above.
30832 (aarch64_shrn<mode>_insn_be): Likewise.
30833 (aarch64_rshrn<mode>_insn_le): Likewise.
30834 (aarch64_rshrn<mode>_insn_be): Likewise.
30835 (aarch64_shrn2<mode>_insn_le): Likewise.
30836 (aarch64_shrn2<mode>_insn_be): Likewise.
30837 (aarch64_rshrn2<mode>_insn_le): Likewise.
30838 (aarch64_rshrn2<mode>_insn_be): Likewise.
30840 2021-03-08 Vladimir N. Makarov <vmakarov@redhat.com>
30843 * lra-constraints.c (skip_contraint_modifiers): New function.
30844 (process_address_1): Use it before lookup_constraint call.
30846 2021-03-08 Martin Liska <mliska@suse.cz>
30849 * config/i386/i386-options.c (ix86_option_override_internal):
30850 Enable UINTR and HRESET for -march that supports it.
30852 2021-03-08 Ilya Leoshkevich <iii@linux.ibm.com>
30854 * config/s390/s390.c (f_constraint_p): New function.
30855 (s390_md_asm_adjust): Implement TARGET_MD_ASM_ADJUST.
30856 (TARGET_MD_ASM_ADJUST): Likewise.
30858 2021-03-08 Tobias Burnus <tobias@codesourcery.com>
30861 * tree-nested.c (convert_local_reference_stmt): Avoid calling
30862 lookup_field_for_decl for Fortran module (= namespace context).
30864 2021-03-08 Andreas Krebbel <krebbel@linux.ibm.com>
30866 * config/s390/s390.c (s390_expand_vec_compare): Implement <0
30867 comparison with arithmetic right shift.
30868 (s390_expand_vcond): No need for a force_reg anymore.
30869 s390_vec_compare will do it.
30870 * config/s390/vector.md ("vec_cmp<mode><tointvec>"): Accept also
30871 immediate operands.
30873 2021-03-07 Jakub Jelinek <jakub@redhat.com>
30876 * config/i386/constraints.md (Yw): Use SSE_REGS if TARGET_SSE
30877 but TARGET_AVX512BW or TARGET_AVX512VL is not set. Adjust description
30879 * config/i386/sse.md (v_Yw): New define_mode_attr.
30880 (*<insn><mode>3, *mul<mode>3<mask_name>, *avx2_<code><mode>3,
30881 *sse4_1_<code><mode>3<mask_name>): Use <v_Yw> instead of v
30883 * config/i386/mmx.md (mmx_pshufw_1, *vec_dupv4hi): Use Yw instead of
30884 xYw in constraints.
30886 2021-03-06 Julian Brown <julian@codesourcery.com>
30888 * tree-pretty-print.c (dump_generic_node): Emit non-generic
30889 address space info for aggregates.
30891 2021-03-06 Hans-Peter Nilsson <hp@axis.com>
30893 * config/cris/cris.h (MAX_FIXED_MODE_SIZE): Don't define.
30895 2021-03-05 Jakub Jelinek <jakub@redhat.com>
30897 PR middle-end/99322
30898 * tree-cfg.c (bb_to_omp_idx): New variable.
30899 (execute_build_cfg): Release the bb_to_omp_idx vector after
30900 cleanup_tree_cfg returns.
30901 (handle_abnormal_edges): Remove bb_to_omp_idx argument, adjust
30902 for bb_to_omp_idx being a vec<int> instead of pointer to array
30904 (make_edges): Remove bb_to_omp_idx local variable, don't pass
30905 it to handle_abnormal_edges, adjust for bb_to_omp_idx being a
30906 vec<int> instead of pointer to array of ints and don't free/release
30908 (remove_bb): When removing a bb and placing forced label somewhere
30909 else, ensure it is put into the same OpenMP region during cfg
30910 pass if possible or to entry successor as fallback. Unregister
30911 bb from bb_to_omp_idx.
30913 2021-03-05 Vladimir N. Makarov <vmakarov@redhat.com>
30916 * lra-constraints.c (process_address_1): Skip decomposing address
30917 for asm insn operand with unknown constraint.
30919 2021-03-05 Martin Jambor <mjambor@suse.cz>
30922 * cgraph.c (cgraph_edge::set_call_stmt): Do not update all
30923 corresponding speculative edges if we are about to resolve
30924 sepculation. Make edge direct (and so resolve speculations) before
30925 removing it from call_site_hash.
30926 (cgraph_edge::make_direct): Relax the initial assert to allow calling
30927 the function on speculative direct edges.
30929 2021-03-05 Eric Botcazou <ebotcazou@adacore.com>
30931 PR rtl-optimization/99376
30932 * rtlanal.c (nonzero_bits1) <arithmetic operators>: If the number
30933 of low-order zero bits is too large, set the result to 0 directly.
30935 2021-03-04 Jakub Jelinek <jakub@redhat.com>
30937 PR middle-end/93235
30938 * expmed.c (store_bit_field_using_insv): Return false of xop0 is a
30939 SUBREG and a SUBREG to op_mode can't be created.
30941 2021-03-04 Alex Coplan <alex.coplan@arm.com>
30944 * config/aarch64/aarch64-sve-builtins.cc
30945 (function_resolver::require_vector_type): Handle error_mark_node.
30947 2021-03-04 Ilya Leoshkevich <iii@linux.ibm.com>
30949 * cfgexpand.c (expand_asm_loc): Pass new parameter.
30950 (expand_asm_stmt): Likewise.
30951 * config/arm/aarch-common-protos.h (arm_md_asm_adjust): Add new
30953 * config/arm/aarch-common.c (arm_md_asm_adjust): Likewise.
30954 * config/arm/arm.c (thumb1_md_asm_adjust): Likewise.
30955 * config/cris/cris.c (cris_md_asm_adjust): Likewise.
30956 * config/i386/i386.c (ix86_md_asm_adjust): Likewise.
30957 * config/mn10300/mn10300.c (mn10300_md_asm_adjust): Likewise.
30958 * config/nds32/nds32.c (nds32_md_asm_adjust): Likewise.
30959 * config/pdp11/pdp11.c (pdp11_md_asm_adjust): Likewise.
30960 * config/rs6000/rs6000.c (rs6000_md_asm_adjust): Likewise.
30961 * config/vax/vax.c (vax_md_asm_adjust): Likewise.
30962 * config/visium/visium.c (visium_md_asm_adjust): Likewise.
30963 * doc/tm.texi (md_asm_adjust): Likewise.
30964 * target.def (md_asm_adjust): Likewise.
30966 2021-03-04 Richard Biener <rguenther@suse.de>
30968 PR middle-end/97855
30969 * tree-pretty-print.c: Poison pp_printf.
30970 (dump_decl_name): Avoid use of pp_printf.
30971 (dump_block_node): Likewise.
30972 (dump_generic_node): Likewise.
30974 2021-03-04 Martin Sebor <msebor@redhat.com>
30976 PR middle-end/96963
30977 PR middle-end/94655
30978 * builtins.c (handle_array_ref): New helper.
30979 (handle_mem_ref): New helper.
30980 (compute_objsize_r): Factor out ARRAY_REF and MEM_REF handling
30981 into new helper functions. Correct a workaround for vectorized
30984 2021-03-03 Pat Haugen <pthaugen@linux.ibm.com>
30986 * config/rs6000/dfp.md (extendddtd2, trunctddd2, *cmp<mode>_internal1,
30987 floatditd2, ftrunc<mode>2, fix<mode>di2, dfp_ddedpd_<mode>,
30988 dfp_denbcd_<mode>, dfp_dxex_<mode>, dfp_diex_<mode>,
30989 *dfp_sgnfcnc_<mode>, dfp_dscli_<mode>, dfp_dscri_<mode>): Update size
30990 attribute for Power10.
30991 * config/rs6000/mma.md (*movoo): Likewise.
30992 * config/rs6000/rs6000.md (define_attr "size"): Add 256.
30993 (define_mode_attr bits): Add DD/TD modes.
30994 * config/rs6000/sync.md (load_quadpti, store_quadpti, load_lockedpti,
30995 store_conditionalpti): Update size attribute for Power10.
30997 2021-03-03 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
31000 * config/sparc/t-sparc (tree-ssanames.o-warn): Don't error for
31001 -Wuninitialized, -Wmaybe-uninitialized.
31002 (wide-int.o-warn): Likewise.
31004 2021-03-03 Richard Earnshaw <rearnsha@arm.com>
31006 * common/config/arm/arm-common.c: Include configargs.h.
31007 (arm_config_default): New function.
31008 (arm_target_mode): Renamed from arm_target_thumb_only. Handle
31009 processors that do not support Thumb. Take into account the
31010 --with-mode configuration setting for selecting the default.
31011 * config/arm/arm.h (OPTION_DEFAULT_SPECS): Remove entry for 'mode'.
31012 (TARGET_MODE_SPEC_FUNCTIONS): Update for function name change.
31014 2021-03-03 Martin Liska <mliska@suse.cz>
31016 PR gcov-profile/97461
31017 * gcov-io.h (GCOV_PREALLOCATED_KVP): Remove.
31019 2021-03-03 Eric Botcazou <ebotcazou@adacore.com>
31022 * config/i386/i386.c (ix86_compute_frame_layout): For a SEH target,
31023 point back the hard frame pointer to its default location when the
31024 frame is larger than SEH_MAX_FRAME_SIZE.
31026 2021-03-03 Jakub Jelinek <jakub@redhat.com>
31029 * config/i386/predicates.md (logic_operator): New define_predicate.
31030 * config/i386/i386.md (mov + mem using comm arith peephole2):
31031 Punt if operands[1] is EXT_REX_SSE_REGNO_P, AVX512BW is not enabled
31032 and the inner mode is [QH]Imode.
31034 2021-03-03 Jakub Jelinek <jakub@redhat.com>
31037 * dwarf2out.c (dw_loc_list_struct): Add end_entry member.
31038 (new_loc_list): Clear end_entry.
31039 (output_loc_list): Only use DW_LLE_startx_length for -gsplit-dwarf
31040 if HAVE_AS_LEB128, otherwise use DW_LLE_startx_endx. Fix comment
31042 (index_location_lists): For dwarf_version >= 5 without HAVE_AS_LEB128,
31043 initialize also end_entry.
31045 2021-03-03 Jakub Jelinek <jakub@redhat.com>
31048 * cfgrtl.c (fixup_partitions): When changing some bbs from hot to cold
31049 partitions, if in non-layout mode after reorder_blocks also move
31050 affected blocks to ensure a single partition transition.
31052 2021-03-03 Jason Merrill <jason@redhat.com>
31055 * cgraphunit.c (process_function_and_variable_attributes): Don't
31056 warn about flatten on an alias if the target also has it.
31057 * cgraph.h (symtab_node::get_alias_target_tree): New.
31059 2021-03-02 David Edelsohn <dje.gcc@gmail.com>
31061 * config/rs6000/rs6000.md (tls_get_tpointer_internal): Prepend
31062 period to symbol name.
31063 (tls_get_addr_internal<mode>): Same.
31065 2021-03-02 David Malcolm <dmalcolm@redhat.com>
31068 * diagnostic-show-locus.c
31069 (selftest::test_one_liner_many_fixits_2): Fix accidental usage of
31072 2021-03-02 Martin Sebor <msebor@redhat.com>
31074 PR middle-end/99276
31075 * builtins.c (warn_for_access): Remove stray warning text.
31077 2021-03-02 Martin Sebor <msebor@redhat.com>
31079 PR middle-end/99295
31080 * doc/extend.texi (attribute malloc): Reword and clarify nonaliasing
31083 2021-03-02 Jakub Jelinek <jakub@redhat.com>
31086 * dwarf2out.c (output_macinfo_op): Use DW_MACRO_*_str* even with
31087 -gdwarf-5 -gstrict-dwarf. For -gsplit-dwarf -gdwarf-5 use
31088 DW_MACRO_*_strx instead of DW_MACRO_*_strp. Handle
31089 DW_MACRO_define_strx and DW_MACRO_undef_strx.
31090 (save_macinfo_strings): Use DW_MACRO_*_str* even with
31091 -gdwarf-5 -gstrict-dwarf. Handle DW_MACRO_define_strx and
31092 DW_MACRO_undef_strx.
31094 2021-03-02 Andreas Krebbel <krebbel@linux.ibm.com>
31096 * config/s390/s390-builtin-types.def (BT_FN_V4SF_V8HI_UINT): New
31098 (BT_FN_V8HI_V8HI_UINT): Likewise.
31099 (BT_FN_V8HI_V4SF_V4SF_UINT): Likewise.
31100 * config/s390/s390-builtins.def (B_NNPA): New macro definition.
31101 (s390_vclfnhs, s390_vclfnls, s390_vcrnfs, s390_vcfn, s390_vcnf):
31102 New builtin definitions.
31103 * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal): Bump
31104 vector extension version.
31105 * config/s390/s390.c (s390_expand_builtin): Check if builtins are
31106 available with current -march level.
31107 * config/s390/s390.md (UNSPEC_NNPA_VCLFNHS_V8HI)
31108 (UNSPEC_NNPA_VCLFNLS_V8HI, UNSPEC_NNPA_VCRNFS_V8HI)
31109 (UNSPEC_NNPA_VCFN_V8HI, UNSPEC_NNPA_VCNF_V8HI): New constants.
31110 * config/s390/vecintrin.h (vec_extend_to_fp32_hi): New macro.
31111 (vec_extend_to_fp32_lo): Likewise.
31112 (vec_round_from_fp32): Likewise.
31113 (vec_convert_to_fp16): Likewise.
31114 (vec_convert_from_fp16): Likewise.
31115 * config/s390/vx-builtins.md (vclfnhs_v8hi): New insn pattern.
31116 (vclfnls_v8hi): Likewise.
31117 (vcrnfs_v8hi): Likewise.
31118 (vcfn_v8hi): Likewise.
31119 (vcnf_v8hi): Likewise.
31121 2021-03-02 Andreas Krebbel <krebbel@linux.ibm.com>
31123 * common/config/s390/s390-common.c (processor_flags_table): New entry.
31124 * config.gcc: Enable arch14 for --with-arch and --with-tune.
31125 * config/s390/driver-native.c (s390_host_detect_local_cpu): Pick
31126 arch14 for unknown CPU models.
31127 * config/s390/s390-opts.h (enum processor_type): Add PROCESSOR_ARCH14.
31128 * config/s390/s390.c (s390_issue_rate): Add case for PROCESSOR_ARCH14.
31129 (s390_get_sched_attrmask): Likewise.
31130 (s390_get_unit_mask): Likewise.
31131 * config/s390/s390.h (enum processor_flags): Add PF_NNPA and PF_ARCH14.
31132 (TARGET_CPU_ARCH14, TARGET_CPU_ARCH14_P, TARGET_CPU_NNPA)
31133 (TARGET_CPU_NNPA_P, TARGET_ARCH14, TARGET_ARCH14_P, TARGET_NNPA)
31134 (TARGET_NNPA_P): New macro definitions.
31135 * config/s390/s390.md ("cpu_facility", "enabled"): Add arch14 and nnpa.
31136 * config/s390/s390.opt: Add PROCESSOR_ARCH14.
31138 2021-03-02 Jakub Jelinek <jakub@redhat.com>
31140 PR middle-end/95757
31141 * tree-vrp.c (register_edge_assert_for): Remove superfluous ()s around
31142 condition. Call register_edge_assert_for_1 for == 0, != 0, == 1 and
31143 != 1 comparisons if name is lhs of a comparison.
31145 2021-03-01 Iain Sandoe <iain@sandoe.co.uk>
31149 * config/darwin-protos.h (darwin_should_restore_cfa_state): New.
31150 * config/darwin.c (darwin_should_restore_cfa_state): New.
31151 * config/darwin.h (TARGET_ASM_SHOULD_RESTORE_CFA_STATE): New.
31152 * doc/tm.texi: Regenerated.
31153 * doc/tm.texi.in: Document TARGET_ASM_SHOULD_RESTORE_CFA_STATE.
31154 * dwarf2cfi.c (connect_traces): If the target requests, restore
31155 the CFA expression after a DW_CFA_restore.
31156 * target.def (TARGET_ASM_SHOULD_RESTORE_CFA_STATE): New hook.
31158 2021-03-01 Martin Liska <mliska@suse.cz>
31161 * optc-save-gen.awk: Add 4 more exceptions.
31163 2021-03-01 Nathan Sidwell <nathan@acm.org>
31166 * tree.h (TYPE_ALIGN_RAW): New accessor.
31167 (TYPE_ALIGN): Use it.
31169 2021-03-01 Jan Hubicka <jh@suse.cz>
31172 * ipa-fnsummary.c (compute_fn_summary): Fix sanity check.
31174 2021-03-01 Eric Botcazou <ebotcazou@adacore.com>
31177 * config/i386/i386.c (ix86_compute_frame_layout): For a SEH target,
31178 point the hard frame pointer to the SSE register save area instead
31179 of the general register save area. Perform only minimal adjustment
31180 for small frames if it is initially not correctly aligned.
31181 (ix86_expand_prologue): Remove early saves for a SEH target.
31182 * config/i386/winnt.c (struct seh_frame_state): Document constraint.
31184 2021-02-28 Jakub Jelinek <jakub@redhat.com>
31187 * ipa.c (symbol_table::remove_unreachable_nodes): Fix a comment
31188 typo - referneced -> referenced.
31189 * tree.c (component_ref_size): Fix comment typo -
31190 refernce -> reference.
31191 * tree-ssa-alias.c (access_path_may_continue_p): Fix comment typo -
31192 traling -> trailing.
31193 (aliasing_component_refs_p): Fix comment typos -
31194 refernce -> reference and refernece -> reference and
31195 traling -> trailing.
31196 (nonoverlapping_refs_since_match_p): Fix comment typo -
31197 referneces -> references.
31198 * doc/invoke.texi (--param modref-max-bases): Fix a typo -
31199 referneces -> references.
31201 2021-02-27 Iain Sandoe <iain@sandoe.co.uk>
31203 * config/host-darwin.c (darwin_gt_pch_use_address): Modify
31204 diagnostic message to avoid use of a contraction and format
31207 2021-02-27 Jakub Jelinek <jakub@redhat.com>
31210 * gcse.c (gcse_or_cprop_is_too_expensive): Use %wu instead of
31211 HOST_WIDE_INT_PRINT_UNSIGNED in warning format string.
31212 * ipa-devirt.c (ipa_odr_read_section): Use %wd instead of
31213 HOST_WIDE_INT_PRINT_DEC in inform format string. Fix comment
31216 2021-02-26 Richard Biener <rguenther@suse.de>
31218 PR middle-end/99281
31219 * expr.c (store_field): For calls with return-slot optimization
31220 and addressable return type expand the store directly.
31222 2021-02-26 Richard Biener <rguenther@suse.de>
31225 * builtins.c (warn_string_no_nul): Fix diagnostic formatting.
31227 2021-02-26 Peter Bergner <bergner@linux.ibm.com>
31230 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Replace assert
31233 2021-02-26 Aaron Sawdey <acsawdey@linux.ibm.com>
31235 * config.gcc: Add rs6000-pcrel-opt.o.
31236 * config/rs6000/rs6000-pcrel-opt.c: New file.
31237 * config/rs6000/pcrel-opt.md: New file.
31238 * config/rs6000/predicates.md: Add d_form_memory predicate.
31239 * config/rs6000/rs6000-cpus.def: Add OPTION_MASK_PCREL_OPT.
31240 * config/rs6000/rs6000-passes.def: Add pass_pcrel_opt.
31241 * config/rs6000/rs6000-protos.h: Add reg_to_non_prefixed(),
31242 pcrel_opt_valid_mem_p(), output_pcrel_opt_reloc(),
31243 and make_pass_pcrel_opt().
31244 * config/rs6000/rs6000.c (reg_to_non_prefixed): Make global.
31245 (rs6000_option_override_internal): Add pcrel-opt.
31246 (rs6000_delegitimize_address): Support pcrel-opt.
31247 (rs6000_opt_masks): Add pcrel-opt.
31248 (pcrel_opt_valid_mem_p): New function.
31249 (reg_to_non_prefixed): Make global.
31250 (rs6000_asm_output_opcode): Reset prepend_p_to_next_insn.
31251 (output_pcrel_opt_reloc): New function.
31252 * config/rs6000/rs6000.md (loads_extern_addr): New attr.
31253 (pcrel_extern_addr): Set loads_extern_addr.
31254 Add include for pcrel-opt.md.
31255 * config/rs6000/rs6000.opt: Add -mpcrel-opt.
31256 * config/rs6000/t-rs6000: Add rules for pcrel-opt.c and
31259 2021-02-26 YunQiang Su <yunqiang.su@cipunited.com>
31262 * config/mips/mips.c (mips_expand_ext_as_unaligned_load):
31263 If TARGET_64BIT and dest is SUBREG, we check the width, if it
31264 equal to SImode, we use SImode operation, just like what we are
31267 2021-02-26 Marek Polacek <polacek@redhat.com>
31269 * builtins.c (warn_for_access): Fix typos.
31271 2021-02-25 Iain Sandoe <iain@sandoe.co.uk>
31273 * config/aarch64/aarch64.md (<optab>_rol<mode>3): Add a '#'
31274 mark in front of the immediate quantity.
31275 (<optab>_rolsi3_uxtw): Likewise.
31277 2021-02-25 Richard Earnshaw <rearnsha@arm.com>
31280 * config/arm/thumb2.md (nonsecure_call_reg_thumb2_fpcxt): New pattern.
31281 (nonsecure_call_value_reg_thumb2_fpcxt): Likewise.
31282 (nonsecure_call_reg_thumb2): Restrict to using r4 for the callee
31283 address and disable when the FPCXT is not available.
31284 (nonsecure_call_value_reg_thumb2): Likewise.
31286 2021-02-25 Nathan Sidwell <nathan@acm.org>
31289 * doc/invoke.texi (flang-info-module-cmi): Renamed option.
31291 2021-02-25 Tamar Christina <tamar.christina@arm.com>
31293 * tree-vect-slp.c (optimize_load_redistribution_1): Abort on NULL nodes.
31295 2021-02-25 Richard Biener <rguenther@suse.de>
31297 PR tree-optimization/99253
31298 * tree-vect-loop.c (check_reduction_path): First compute
31299 code, then verify out-of-loop uses.
31301 2021-02-25 Jakub Jelinek <jakub@redhat.com>
31304 * match.pd ((T)(A) + CST -> (T)(A + CST)): Add :s to convert.
31306 2021-02-25 Jakub Jelinek <jakub@redhat.com>
31308 PR tree-optimization/80635
31309 * tree-vrp.c (vrp_simplify_cond_using_ranges): Also handle
31310 VIEW_CONVERT_EXPR if modes are the same, innerop is integral and
31311 has mode precision.
31313 2021-02-25 Richard Biener <rguenther@suse.de>
31315 * tree-vect-slp.c (optimize_load_redistribution_1): Delay
31316 load_map population.
31317 (vect_match_slp_patterns_2): Revert part of last change.
31318 (vect_analyze_slp): Do not interleave optimize_load_redistribution
31319 with pattern detection but do it afterwards. Dump the
31320 whole SLP graph after pattern recognition and load
31321 redistribution optimization finished.
31323 2021-02-24 Jakub Jelinek <jakub@redhat.com>
31326 * omp-low.c (struct omp_context): Add teams_nested_p and
31327 nonteams_nested_p members.
31328 (scan_omp_target): Diagnose teams nested inside of target with other
31329 directives strictly nested inside of the same target.
31330 (check_omp_nesting_restrictions): Set ctx->teams_nested_p or
31331 ctx->nonteams_nested_p as needed.
31333 2021-02-24 Vladimir N. Makarov <vmakarov@redhat.com>
31335 PR inline-asm/99123
31336 * lra-constraints.c (uses_hard_regs_p): Don't use decompose_mem_address.
31338 2021-02-24 Hans-Peter Nilsson <hp@axis.com>
31340 * config/cris/cris.c (cris_expand_prologue): Set
31341 current_function_static_stack_size, if flag_stack_usage_info.
31343 2021-02-24 Pat Haugen <pthaugen@linux.ibm.com>
31345 * config/rs6000/rs6000.c (next_insn_prefixed_p): Rename.
31346 (rs6000_final_prescan_insn): Adjust.
31347 (rs6000_asm_output_opcode): Likewise.
31349 2021-02-24 Martin Sebor <msebor@redhat.com>
31351 PR middle-end/97172
31352 * attribs.c (attr_access::free_lang_data): Clear attribute arg spec
31353 from function arguments.
31355 2021-02-24 Tamar Christina <tamar.christina@arm.com>
31357 PR tree-optimization/99220
31358 * tree-vect-slp.c (optimize_load_redistribution_1): Remove
31359 node from cache when it's about to be deleted.
31361 2021-02-24 Jakub Jelinek <jakub@redhat.com>
31363 PR tree-optimization/99225
31364 * fold-const.c (fold_binary_loc) <case NE_EXPR>: In (x & (1 << y)) != 0
31365 to ((x >> y) & 1) != 0 simplifications use build_one_cst instead of
31366 build_int_cst (..., 1). Formatting fixes.
31368 2021-02-24 Tamar Christina <tamar.christina@arm.com>
31370 PR tree-optimization/99149
31371 * tree-vect-slp-patterns.c (vect_detect_pair_op): Don't recreate the
31373 (vect_slp_reset_pattern): Remove.
31374 (complex_fma_pattern::matches): Remove call to vect_slp_reset_pattern.
31375 (complex_mul_pattern::build, complex_fma_pattern::build,
31376 complex_fms_pattern::build): Fix ref counts.
31377 * tree-vect-slp.c (vect_free_slp_tree): Undo SLP only pattern relevancy
31378 when node is being deleted.
31379 (vect_match_slp_patterns_2): Correct result of cache hit on patterns.
31380 (vect_schedule_slp): Invalidate SLP_TREE_REPRESENTATIVE of removed
31382 * tree-vectorizer.c (vec_info::new_stmt_vec_info): Initialize value.
31384 2021-02-24 Matthias Klose <doko@ubuntu.com>
31387 2020-12-07 Matthias Klose <doko@ubuntu.com>
31389 * genextract.c (print_header): Undefine ENABLE_RTL_CHECKING
31390 and ENABLE_RTL_FLAG_CHECKING.
31392 2021-02-24 Richard Biener <rguenther@suse.de>
31395 * builtins.c (fold_builtin_next_arg): Avoid NULL arg.
31397 2021-02-23 Peter Bergner <bergner@linux.ibm.com>
31399 * config/rs6000/mma.md (mma_assemble_pair): Rename from this...
31400 (vsx_assemble_pair): ...to this.
31401 (*mma_assemble_pair): Rename from this...
31402 (*vsx_assemble_pair): ...to this.
31403 (mma_disassemble_pair): Rename from this...
31404 (vsx_disassemble_pair): ...to this.
31405 (*mma_disassemble_pair): Rename from this...
31406 (*vsx_disassemble_pair): ...to this.
31407 * config/rs6000/rs6000-builtin.def (BU_MMA_V2, BU_MMA_V3,
31408 BU_COMPAT): New macros.
31409 (mma_assemble_pair): Rename from this...
31410 (vsx_assemble_pair): ...to this.
31411 (mma_disassemble_pair): Rename from this...
31412 (vsx_disassemble_pair): ...to this.
31413 (mma_assemble_pair): New compatibility built-in.
31414 (mma_disassemble_pair): Likewise.
31415 * config/rs6000/rs6000-call.c (struct builtin_compatibility): New.
31416 (RS6000_BUILTIN_COMPAT): Define.
31417 (bdesc_compat): New.
31418 (mma_expand_builtin): Use VSX_BUILTIN_DISASSEMBLE_PAIR_INTERNAL.
31419 (rs6000_gimple_fold_mma_builtin): Use MMA_BUILTIN_DISASSEMBLE_PAIR
31420 and VSX_BUILTIN_ASSEMBLE_PAIR.
31421 (rs6000_init_builtins): Register compatibility built-ins.
31422 (mma_init_builtins): Use VSX_BUILTIN_ASSEMBLE_PAIR,
31423 VSX_BUILTIN_ASSEMBLE_PAIR_INTERNAL, VSX_BUILTIN_DISASSEMBLE_PAIR and
31424 VSX_BUILTIN_DISASSEMBLE_PAIR_INTERNAL.
31425 * doc/extend.texi (__builtin_mma_assemble_pair): Rename from this...
31426 (__builtin_vsx_assemble_pair): ...to this.
31427 (__builtin_mma_disassemble_pair): Rename from this...
31428 (__builtin_vsx_disassemble_pair): ...to this.
31430 2021-02-23 Martin Liska <mliska@suse.cz>
31433 * ipa-icf.c (sem_variable::merge): Do not merge 2 variables
31434 with different alignment. That leads to an invalid red zone
31435 size allocated in runtime.
31437 2021-02-23 Jakub Jelinek <jakub@redhat.com>
31439 PR tree-optimization/99204
31440 * fold-const.c (fold_read_from_constant_string): Check that
31441 tree_fits_uhwi_p (index) rather than just that index is INTEGER_CST.
31443 2021-02-23 Segher Boessenkool <segher@kernel.crashing.org>
31444 Kewen Lin <linkw@gcc.gnu.org>
31446 * config/rs6000/rs6000.md (*rotl<mode>3_insert_3): Renamed to...
31447 (rotl<mode>3_insert_3): ...this.
31448 (plus_ior_xor): New code_iterator.
31449 (define_split for GPR rl*imi): New splitter.
31450 * config/rs6000/vsx.md (vsx_init_v4si): Use gen_rotldi3_insert_3
31451 for integer merging.
31453 2021-02-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
31455 * config/aarch64/aarch64-tuning-flags.def (cse_sve_vl_constants):
31457 * config/aarch64/aarch64.md (add<mode>3): Force CONST_POLY_INT immediates
31458 into a register when the above is enabled.
31459 * config/aarch64/aarch64.c (neoversev1_tunings):
31460 AARCH64_EXTRA_TUNE_CSE_SVE_VL_CONSTANTS.
31461 (aarch64_rtx_costs): Use AARCH64_EXTRA_TUNE_CSE_SVE_VL_CONSTANTS.
31463 2021-02-22 Hans-Peter Nilsson <hp@axis.com>
31465 * config/cris/cris.c (cris_print_operand) <'T'>: Change
31466 valid operand from is now an addi mult-value to shift-value.
31467 * config/cris/cris.md (*addi): Change expression of scaled
31468 operand from mult to ashift.
31469 * config/cris/cris.md (*addi_reload): New insn_and_split.
31471 2021-02-22 John David Anglin <danglin@gcc.gnu.org>
31474 * config/pa/pa.c (TARGET_ASM_CAN_OUTPUT_MI_THUNK): Define as
31475 hook_bool_const_tree_hwi_hwi_const_tree_true.
31476 (pa_asm_output_mi_thunk): Add support for nonzero vcall_offset.
31478 2021-02-22 Andre Vieira <andre.simoesdiasvieira@arm.com>
31480 PR rtl-optimization/98791
31481 * ira-conflicts.c (process_regs_for_copy): Don't create allocno copies
31482 for unordered modes.
31484 2021-02-22 Martin Liska <mliska@suse.cz>
31486 * tree-inline.c (inline_forbidden_p): Set
31487 inline_forbidden_reason.
31489 2021-02-22 Richard Biener <rguenther@suse.de>
31491 * tree-vect-slp.c (vect_bb_vectorization_profitable_p): Dump
31494 2021-02-22 Richard Biener <rguenther@suse.de>
31496 PR tree-optimization/99165
31497 * gimple-ssa-store-merging.c (pass_store_merging::process_store):
31498 Accumulate changed to ret.
31500 2021-02-21 Uros Bizjak <ubizjak@gmail.com>
31503 2020-12-09 Uroš Bizjak <ubizjak@gmail.com>
31505 * config/i386/i386.h (REG_ALLOC_ORDER): Remove
31507 2021-02-20 Ilya Leoshkevich <iii@linux.ibm.com>
31510 * config/s390/vector.md (trunctf<DFP_ALL:mode>2_vr): New
31512 (trunctf<DFP_ALL:mode>2): Likewise.
31513 (trunctdtf2_vr): Likewise.
31514 (trunctdtf2): Likewise.
31515 (extend<DFP_ALL:mode>tf2_vr): Likewise.
31516 (extend<DFP_ALL:mode>tf2): Likewise.
31517 (extendtftd2_vr): Likewise.
31518 (extendtftd2): Likewise.
31520 2021-02-20 Ilya Leoshkevich <iii@linux.ibm.com>
31522 * config/s390/vector.md (*fprx2_to_tf): Rename to fprx2_to_tf,
31523 add memory alternative.
31524 (tf_to_fprx2): New pattern.
31526 2021-02-19 Martin Sebor <msebor@redhat.com>
31529 * attribs.c (init_attr_rdwr_indices): Guard vblist use.
31530 (attr_access::free_lang_data): Remove a spurious test.
31532 2021-02-19 Nathan Sidwell <nathan@acm.org>
31534 * doc/invoke.texi (flang-info-module-read): Document.
31536 2021-02-19 Martin Liska <mliska@suse.cz>
31538 PR translation/99167
31539 * params.opt: Fix typo.
31541 2021-02-19 Richard Biener <rguenther@suse.de>
31543 PR middle-end/99122
31544 * tree-inline.c (inline_forbidden_p): Do not inline functions
31545 with VLA arguments or return value.
31547 2021-02-19 Jakub Jelinek <jakub@redhat.com>
31550 * config/arm/arm.md (*stack_protect_combined_set_insn,
31551 *stack_protect_combined_test_insn): If force_const_mem result
31552 is not valid general operand, force its address into the destination
31555 2021-02-19 Jakub Jelinek <jakub@redhat.com>
31558 * tree-cfg.c (gimple_merge_blocks): If bb a starts with eh landing
31559 pad or non-local label, put FORCED_LABELs from bb b after that label
31560 rather than before it.
31562 2021-02-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
31565 * config/aarch64/aarch64-sve.md (<ASHIFT:optab><mode>3): Use
31566 expand_vector_broadcast' to emit the vec_duplicate operand.
31568 2021-02-18 Vladimir N. Makarov <vmakarov@redhat.com>
31570 PR rtl-optimization/96264
31571 * lra-remat.c (reg_overlap_for_remat_p): Check also output insn
31574 2021-02-18 H.J. Lu <hjl.tools@gmail.com>
31577 * varasm.c (get_section): Replace SUPPORTS_SHF_GNU_RETAIN with
31578 looking up the retain attribute.
31579 (resolve_unique_section): Likewise.
31580 (get_variable_section): Likewise.
31581 (switch_to_section): Likewise. Warn when a symbol without the
31582 retain attribute and a symbol with the retain attribute are
31583 placed in the section with the same name, instead of the used
31585 * doc/extend.texi: Document the "retain" attribute.
31587 2021-02-18 Nathan Sidwell <nathan@acm.org>
31590 * doc/invoke.texi (flang-info-include-translate): Document header
31593 2021-02-18 Richard Biener <rguenther@suse.de>
31595 PR middle-end/99122
31596 * ipa-fnsummary.c (analyze_function_body): Set
31597 CIF_FUNCTION_NOT_INLINABLE for VLA parameter calls.
31598 * tree-inline.c (insert_init_debug_bind): Pass NULL for
31599 error_mark_node values.
31600 (force_value_to_type): Do not build V_C_Es for WITH_SIZE_EXPR
31602 (setup_one_parameter): Delay force_value_to_type until when
31605 2021-02-18 Hans-Peter Nilsson <hp@axis.com>
31607 PR tree-optimization/99142
31608 * match.pd (clz cmp 0): Gate replacement on single_use of clz result.
31610 2021-02-18 Jakub Jelinek <jakub@redhat.com>
31612 * wide-int-bitmask.h (wide_int_bitmask::wide_int_bitmask (),
31613 wide_int_bitmask::wide_int_bitmask (uint64_t),
31614 wide_int_bitmask::wide_int_bitmask (uint64_t, uint64_t),
31615 wide_int_bitmask::operator ~ () const,
31616 wide_int_bitmask::operator | (wide_int_bitmask) const,
31617 wide_int_bitmask::operator & (wide_int_bitmask) const): Use constexpr
31619 * config/i386/i386.h (PTA_3DNOW, PTA_3DNOW_A, PTA_64BIT, PTA_ABM,
31620 PTA_AES, PTA_AVX, PTA_BMI, PTA_CX16, PTA_F16C, PTA_FMA, PTA_FMA4,
31621 PTA_FSGSBASE, PTA_LWP, PTA_LZCNT, PTA_MMX, PTA_MOVBE, PTA_NO_SAHF,
31622 PTA_PCLMUL, PTA_POPCNT, PTA_PREFETCH_SSE, PTA_RDRND, PTA_SSE, PTA_SSE2,
31623 PTA_SSE3, PTA_SSE4_1, PTA_SSE4_2, PTA_SSE4A, PTA_SSSE3, PTA_TBM,
31624 PTA_XOP, PTA_AVX2, PTA_BMI2, PTA_RTM, PTA_HLE, PTA_PRFCHW, PTA_RDSEED,
31625 PTA_ADX, PTA_FXSR, PTA_XSAVE, PTA_XSAVEOPT, PTA_AVX512F, PTA_AVX512ER,
31626 PTA_AVX512PF, PTA_AVX512CD, PTA_NO_TUNE, PTA_SHA, PTA_PREFETCHWT1,
31627 PTA_CLFLUSHOPT, PTA_XSAVEC, PTA_XSAVES, PTA_AVX512DQ, PTA_AVX512BW,
31628 PTA_AVX512VL, PTA_AVX512IFMA, PTA_AVX512VBMI, PTA_CLWB, PTA_MWAITX,
31629 PTA_CLZERO, PTA_NO_80387, PTA_PKU, PTA_AVX5124VNNIW, PTA_AVX5124FMAPS,
31630 PTA_AVX512VPOPCNTDQ, PTA_SGX, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES,
31631 PTA_AVX512VBMI2, PTA_VPCLMULQDQ, PTA_AVX512BITALG, PTA_RDPID,
31632 PTA_PCONFIG, PTA_WBNOINVD, PTA_AVX512VP2INTERSECT, PTA_PTWRITE,
31633 PTA_AVX512BF16, PTA_WAITPKG, PTA_MOVDIRI, PTA_MOVDIR64B, PTA_ENQCMD,
31634 PTA_CLDEMOTE, PTA_SERIALIZE, PTA_TSXLDTRK, PTA_AMX_TILE, PTA_AMX_INT8,
31635 PTA_AMX_BF16, PTA_UINTR, PTA_HRESET, PTA_KL, PTA_WIDEKL, PTA_AVXVNNI,
31636 PTA_X86_64_BASELINE, PTA_X86_64_V2, PTA_X86_64_V3, PTA_X86_64_V4,
31637 PTA_CORE2, PTA_NEHALEM, PTA_WESTMERE, PTA_SANDYBRIDGE, PTA_IVYBRIDGE,
31638 PTA_HASWELL, PTA_BROADWELL, PTA_SKYLAKE, PTA_SKYLAKE_AVX512,
31639 PTA_CASCADELAKE, PTA_COOPERLAKE, PTA_CANNONLAKE, PTA_ICELAKE_CLIENT,
31640 PTA_ICELAKE_SERVER, PTA_TIGERLAKE, PTA_SAPPHIRERAPIDS, PTA_ALDERLAKE,
31641 PTA_KNL, PTA_BONNELL, PTA_SILVERMONT, PTA_GOLDMONT, PTA_GOLDMONT_PLUS,
31642 PTA_TREMONT, PTA_KNM): Use constexpr instead of const.
31644 2021-02-18 Jakub Jelinek <jakub@redhat.com>
31646 PR middle-end/99109
31647 * gimple-array-bounds.cc (build_zero_elt_array_type): Rename to ...
31648 (build_printable_array_type): ... this. Add nelts argument. For
31649 overaligned eltype, use TYPE_MAIN_VARIANT (eltype) instead. If
31650 nelts, call build_array_type_nelts.
31651 (array_bounds_checker::check_mem_ref): Use build_printable_array_type
31652 instead of build_zero_elt_array_type and build_array_type_nelts.
31654 2021-02-18 Jakub Jelinek <jakub@redhat.com>
31657 * config/i386/i386.c (distance_non_agu_define): Don't call
31658 extract_insn_cached here.
31659 (ix86_lea_outperforms): Save and restore recog_data around call
31660 to distance_non_agu_define and distance_agu_use.
31661 (ix86_ok_to_clobber_flags): Remove.
31662 (ix86_avoid_lea_for_add): Don't call ix86_ok_to_clobber_flags.
31663 (ix86_avoid_lea_for_addr): Likewise. Adjust function comment.
31664 * config/i386/i386.md (*lea<mode>): Change from define_insn_and_split
31665 into define_insn. Move the splitting to define_peephole2 and
31666 check there using peep2_regno_dead_p if FLAGS_REG is dead.
31668 2021-02-17 Julian Brown <julian@codesourcery.com>
31670 * gimplify.c (gimplify_scan_omp_clauses): Handle ATTACH_DETACH
31673 2021-02-17 Xi Ruoyao <xry111@mengyan1223.wang>
31676 * config/mips/mips.c (mips_symbol_insns): Do not use
31677 MSA_SUPPORTED_MODE_P if mode is MAX_MACHINE_MODE.
31679 2021-02-16 Vladimir N. Makarov <vmakarov@redhat.com>
31681 PR inline-asm/98096
31682 * stmt.c (resolve_operand_name_1): Take inout operands into account
31683 for access to labels by names.
31684 * doc/extend.texi: Describe counting operands for accessing labels.
31686 2021-02-16 Richard Biener <rguenther@suse.de>
31688 PR tree-optimization/38474
31689 * tree-ssa-structalias.c (variable_info::address_taken): New.
31690 (new_var_info): Initialize address_taken.
31691 (process_constraint): Set address_taken.
31692 (solve_constraints): Use the new address_taken flag rather
31693 than is_reg_var for sorting variables.
31694 (dump_constraint): Dump the variable number if the name
31697 2021-02-16 Jakub Jelinek <jakub@redhat.com>
31700 * tree-vect-stmts.c (vectorizable_simd_clone_call): For num_calls != 1
31701 multiply by 4096 and for inbranch by 8192.
31702 * config/i386/i386.c (ix86_simd_clone_usable): For TARGET_AVX512F,
31703 return 3, 2 or 1 for mangle letters 'b', 'c' or 'd'.
31705 2021-02-15 Maya Rashish <coypu@sdf.org>
31707 * config/aarch64/aarch64.c (aarch64_init_builtins):
31708 Call SUBTARGET_INIT_BUILTINS.
31710 2021-02-15 Peter Bergner <bergner@linux.ibm.com>
31712 PR rtl-optimization/98872
31713 * init-regs.c (initialize_uninitialized_regs): Skip initialization
31714 if CONST0_RTX is NULL.
31716 2021-02-15 Richard Sandiford <richard.sandiford@arm.com>
31718 PR rtl-optimization/98863
31719 * rtl-ssa/functions.h (function_info::bb_live_out_info): Delete.
31720 (function_info::build_info): Turn into a declaration, moving the
31721 definition to internals.h.
31722 (function_info::bb_walker): Declare.
31723 (function_info::create_reg_use): Likewise.
31724 (function_info::calculate_potential_phi_regs): Take a build_info
31726 (function_info::place_phis, function_info::create_ebbs): Declare.
31727 (function_info::calculate_ebb_live_in_for_debug): Likewise.
31728 (function_info::populate_backedge_phis): Delete.
31729 (function_info::start_block, function_info::end_block): Declare.
31730 (function_info::populate_phi_inputs): Delete.
31731 (function_info::m_potential_phi_regs): Move information to build_info.
31732 * rtl-ssa/internals.h: New file.
31733 (function_info::bb_phi_info): New class.
31734 (function_info::build_info): Moved from functions.h.
31735 Add a constructor and destructor.
31736 (function_info::build_info::ebb_use): Delete.
31737 (function_info::build_info::ebb_def): Likewise.
31738 (function_info::build_info::bb_live_out): Likewise.
31739 (function_info::build_info::tmp_ebb_live_in_for_debug): New variable.
31740 (function_info::build_info::potential_phi_regs): Likewise.
31741 (function_info::build_info::potential_phi_regs_for_debug): Likewise.
31742 (function_info::build_info::ebb_def_regs): Likewise.
31743 (function_info::build_info::bb_phis): Likewise.
31744 (function_info::build_info::bb_mem_live_out): Likewise.
31745 (function_info::build_info::bb_to_rpo): Likewise.
31746 (function_info::build_info::def_stack): Likewise.
31747 (function_info::build_info::old_def_stack_limit): Likewise.
31748 * rtl-ssa/internals.inl (function_info::build_info::record_reg_def):
31749 Remove the regno argument. Push the previous definition onto the
31750 definition stack where necessary.
31751 * rtl-ssa/accesses.cc: Include internals.h.
31752 * rtl-ssa/changes.cc: Likewise.
31753 * rtl-ssa/blocks.cc: Likewise.
31754 (function_info::build_info::build_info): Define.
31755 (function_info::build_info::~build_info): Likewise.
31756 (function_info::bb_walker): New class.
31757 (function_info::bb_walker::bb_walker): Define.
31758 (function_info::add_live_out_use): Convert a logarithmic-complexity
31759 test into a linear one. Allow the same definition to be passed
31761 (function_info::calculate_potential_phi_regs): Moved from
31762 functions.cc. Take a build_info parameter and store the
31763 information there instead.
31764 (function_info::place_phis): New function.
31765 (function_info::add_entry_block_defs): Update call to record_reg_def.
31766 (function_info::calculate_ebb_live_in_for_debug): New function.
31767 (function_info::add_phi_nodes): Use bb_phis to decide which
31768 registers need phi nodes and initialize ebb_def_regs accordingly.
31769 Do not add degenerate phis here.
31770 (function_info::add_artificial_accesses): Use create_reg_use.
31771 Assert that all definitions are listed in the DF LR sets.
31772 Update call to record_reg_def.
31773 (function_info::record_block_live_out): Record live-out register
31774 values in the phis of successor blocks. Use the live-out set
31775 when processing the last block in an EBB, instead of always
31776 using the live-in sets of successor blocks. AND the live sets
31777 with the set of registers that have been defined in the EBB,
31778 rather than with all potential phi registers. Cope correctly
31779 with branches back to the start of the current EBB.
31780 (function_info::start_block): New function.
31781 (function_info::end_block): Likewise.
31782 (function_info::populate_phi_inputs): Likewise.
31783 (function_info::create_ebbs): Likewise.
31784 (function_info::process_all_blocks): Rewrite into a multi-phase
31786 * rtl-ssa/functions.cc: Include internals.h.
31787 (function_info::calculate_potential_phi_regs): Move to blocks.cc.
31788 (function_info::init_function_data): Remove caller.
31789 * rtl-ssa/insns.cc: Include internals.h
31790 (function_info::create_reg_use): New function. Lazily any
31791 degenerate phis needed by the linear RPO view.
31792 (function_info::record_use): Use create_reg_use. When processing
31793 debug uses, use potential_phi_regs and test it before checking
31794 whether the register is live on entry to the current EBB. Lazily
31795 calculate ebb_live_in_for_debug.
31796 (function_info::record_call_clobbers): Update call to record_reg_def.
31797 (function_info::record_def): Likewise.
31799 2021-02-15 Martin Liska <mliska@suse.cz>
31801 * toplev.c (init_asm_output): Free output of
31802 gen_command_line_string function.
31803 (process_options): Likewise.
31805 2021-02-15 Martin Liska <mliska@suse.cz>
31807 * params.opt: Add 2 missing Param keywords.
31809 2021-02-15 Eric Botcazou <ebotcazou@adacore.com>
31811 * df-core.c (df_worklist_dataflow_doublequeue): Use proper cast.
31813 2021-02-15 Jakub Jelinek <jakub@redhat.com>
31815 PR tree-optimization/99079
31816 * match.pd (A % (pow2pcst << N) -> A & ((pow2pcst << N) - 1)): Remove
31817 useless tree_nop_conversion_p (type, TREE_TYPE (@3)) check. Instead
31818 require both type and TREE_TYPE (@1) to be integral types and either
31819 type having smaller or equal precision, or TREE_TYPE (@1) being
31820 unsigned type, or type being signed type. If TREE_TYPE (@1)
31821 doesn't have wrapping overflow, perform the subtraction of one in
31824 2021-02-14 Jan Hubicka <hubicka@ucw.cz>
31825 Richard Biener <rguether@suse.de>
31828 * ipa-reference.c (ipa_init): Only conditinally initialize
31829 reference_vars_to_consider.
31830 (propagate): Conditionally deninitialize reference_vars_to_consider.
31831 (ipa_reference_write_optimization_summary): Sanity check that
31832 reference_vars_to_consider is not allocated.
31834 2021-02-13 Levy Hsu <admin@levyhsu.com>
31837 * config/riscv/riscv-shorten-memrefs.c (pass_shorten_memrefs): Add
31838 extend parameter to get_si_mem_base_reg declaration.
31839 (get_si_mem_base_reg): Add extend parameter. Set it.
31840 (analyze): Pass extend arg to get_si_mem_base_reg.
31841 (transform): Likewise. Use it when rewriting mems.
31842 * config/riscv/riscv.c (riscv_legitimize_move): Check for subword
31843 loads and emit sign/zero extending load followed by subreg move.
31845 2021-02-13 Jim Wilson <jimw@sifive.com>
31848 * config/riscv/riscv.c (riscv_compressed_lw_address_p): Drop early
31849 exit when !reload_completed. Only perform check for compressed reg
31850 if reload_completed.
31851 (riscv_rtx_costs): In MEM case, when optimizing for size and
31852 shorten memrefs, if not compressible, then increase cost.
31854 2021-02-13 Jakub Jelinek <jakub@redhat.com>
31856 PR rtl-optimization/98439
31857 * recog.c (pass_split_before_regstack::gate): Enable even when
31858 pass_split_before_sched2 is enabled if -fselective-scheduling2 is
31861 2021-02-13 Jakub Jelinek <jakub@redhat.com>
31864 * config/i386/mmx.md (*mmx_pshufd_1): Add a combine splitter for
31865 swap of V2SImode elements in memory into DImode memory rotate by 32.
31867 2021-02-12 Martin Sebor <msebor@redhat.com>
31869 * tree-pretty-print.c (print_generic_expr_to_str): Update comment.
31871 2021-02-12 Richard Sandiford <richard.sandiford@arm.com>
31873 * rtl-ssa/accesses.cc (function_info::make_use_available): Use
31874 m_temp_obstack rather than m_obstack to allocate the temporary use.
31876 2021-02-12 Richard Sandiford <richard.sandiford@arm.com>
31878 * df-problems.c (df_lr_bb_local_compute): Treat partial definitions
31879 as read-modify operations.
31881 2021-02-12 Richard Biener <rguenther@suse.de>
31883 PR middle-end/38474
31884 * ipa-fnsummary.c (unmodified_parm_1): Only walk when
31885 fbi->aa_walk_budget is bigger than zero. Update
31886 fbi->aa_walk_budget.
31887 (param_change_prob): Likewise.
31888 * ipa-prop.c (detect_type_change_from_memory_writes):
31889 Properly account walk_aliased_vdefs.
31890 (parm_preserved_before_stmt_p): Canonicalize updates.
31891 (parm_ref_data_preserved_p): Likewise.
31892 (parm_ref_data_pass_through_p): Likewise.
31893 (determine_known_aggregate_parts): Account own alias queries.
31895 2021-02-12 Martin Liska <mliska@suse.cz>
31897 * opts-common.c (decode_cmdline_option): Release werror_arg.
31898 * opts.c (gen_producer_string): Release output of
31899 gen_command_line_string.
31901 2021-02-12 Richard Biener <rguenther@suse.de>
31903 PR tree-optimization/38474
31904 * params.opt (-param=max-store-chains-to-track=): New param.
31905 (-param=max-stores-to-track=): Likewise.
31906 * doc/invoke.texi (max-store-chains-to-track): Document.
31907 (max-stores-to-track): Likewise.
31908 * gimple-ssa-store-merging.c (pass_store_merging::m_n_chains):
31910 (pass_store_merging::m_n_stores): Likewise.
31911 (pass_store_merging::terminate_and_process_chain): Update
31912 m_n_stores and m_n_chains.
31913 (pass_store_merging::process_store): Likewise. Terminate
31914 oldest chains if the number of stores or chains get too large.
31915 (imm_store_chain_info::terminate_and_process_chain): Dump
31918 2021-02-11 Eric Botcazou <ebotcazou@adacore.com>
31920 * config/i386/winnt.c (i386_pe_seh_unwind_emit): When switching to
31921 the cold section, emit a nop before the directive if the previous
31922 active instruction can throw.
31924 2021-02-11 Peter Bergner <bergner@linux.ibm.com>
31927 * config/rs6000/predicates.md (mma_assemble_input_operand): Restrict
31928 memory addresses that are legal for quad word accesses.
31930 2021-02-11 Andrea Corallo <andrea.corallo@arm.com>
31933 * config/arm/thumb2.md (*doloop_end_internal): Generate
31934 alternative sequence to handle long range branches.
31936 2021-02-11 Joel Hutton <joel.hutton@arm.com>
31938 PR tree-optimization/98772
31939 * optabs-tree.c (supportable_half_widening_operation): New function
31940 to check for supportable V8QI->V8HI widening patterns.
31941 * optabs-tree.h (supportable_half_widening_operation): New function.
31942 * tree-vect-stmts.c (vect_create_half_widening_stmts): New function
31943 to create promotion stmts for V8QI->V8HI widening patterns.
31944 (vectorizable_conversion): Add case for V8QI->V8HI.
31946 2021-02-11 Richard Biener <rguenther@suse.de>
31948 * sparseset.h (SPARSESET_ELT_BITS): Remove.
31949 (SPARSESET_ELT_TYPE): Use unsigned int.
31950 * fwprop.c: Do not include sparseset.h.
31952 2021-02-10 Jakub Jelinek <jakub@redhat.com>
31955 * varasm.c (declare_weak): For -fsyntax-only, allow even
31956 TREE_ASM_WRITTEN function decls.
31958 2021-02-10 Jakub Jelinek <jakub@redhat.com>
31961 * config/i386/sse.md (fix<fixunssuffix>_truncv2sfv2di2,
31962 <insn>v8qiv8hi2, <insn>v8qiv8si2, <insn>v4qiv4si2, <insn>v4hiv4si2,
31963 <insn>v8qiv8di2, <insn>v4qiv4di2, <insn>v2qiv2di2, <insn>v4hiv4di2,
31964 <insn>v2hiv2di2, <insn>v2siv2di2): Force operands[1] into REG before
31965 calling simplify_gen_subreg on it.
31967 2021-02-10 Martin Liska <mliska@suse.cz>
31969 * config/nvptx/nvptx.c (nvptx_option_override): Use
31970 flag_patchable_function_entry instead of the removed
31971 function_entry_patch_area_size.
31973 2021-02-10 Martin Liska <mliska@suse.cz>
31975 PR tree-optimization/99002
31976 PR tree-optimization/99026
31977 * gimple-if-to-switch.cc (if_chain::is_beneficial): Fix memory
31978 leak when adjacent cases are merged.
31979 * tree-switch-conversion.c (switch_decision_tree::analyze_switch_statement): Use
31981 (make_pass_lower_switch): Remove trailing whitespace.
31982 * tree-switch-conversion.h (release_clusters): New.
31984 2021-02-10 Richard Biener <rguenther@suse.de>
31986 PR rtl-optimization/99054
31987 * cfgrtl.c (rtl-optimization/99054): Return an auto_vec.
31988 (fixup_partitions): Adjust.
31989 (rtl_verify_edges): Likewise.
31991 2021-02-10 Jakub Jelinek <jakub@redhat.com>
31993 PR middle-end/99007
31994 * gimplify.c (gimplify_scan_omp_clauses): For MEM_REF on reductions,
31995 temporarily disable gimplify_ctxp->into_ssa around gimplify_expr
31998 2021-02-10 Richard Biener <rguenther@suse.de>
32001 * ipa-pure-const.c (propagate_malloc): Use an auto_vec<>
32004 2021-02-10 Richard Biener <rguenther@suse.de>
32006 PR tree-optimization/99024
32007 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Only
32008 clear loop->aux if it is associated with the destroyed loop_vinfo.
32010 2021-02-10 Martin Liska <mliska@suse.cz>
32012 PR tree-optimization/99002
32013 * gimple-if-to-switch.cc (find_conditions): Fix memory leak
32016 2021-02-10 Martin Liska <mliska@suse.cz>
32019 * ipa-icf.c (sem_item::add_reference): Fix memory leak when
32020 a reference exists.
32022 2021-02-10 Jakub Jelinek <jakub@redhat.com>
32025 * dwarf2out.c (prune_unused_types_walk): Mark DW_TAG_variable DIEs
32026 at class scope for DWARF5+.
32028 2021-02-09 Eric Botcazou <ebotcazou@adacore.com>
32030 PR rtl-optimization/96015
32031 * reorg.c (skip_consecutive_labels): Minor comment tweaks.
32032 (relax_delay_slots): When deleting a jump to the next active
32033 instruction over a barrier, first delete the barrier if the
32034 jump is the only way to reach the target label.
32036 2021-02-09 Andre Vieira <andre.simoesdiasvieira@arm.com>
32038 * config/aarch64/aarch64-cost-tables.h: Add entries for vect.mul.
32039 * config/aarch64/aarch64.c (aarch64_rtx_mult_cost): Use vect.mul for
32040 vector multiplies and vect.alu for SSRA.
32041 * config/arm/aarch-common-protos.h (struct vector_cost_table): Define
32042 vect.mul cost field.
32043 * config/arm/aarch-cost-tables.h: Add entries for vect.mul.
32044 * config/arm/arm.c: Likewise.
32046 2021-02-09 Richard Biener <rguenther@suse.de>
32048 PR tree-optimization/98863
32049 * tree-ssa-sccvn.h (vn_avail::next_undo): Add.
32050 * tree-ssa-sccvn.c (last_pushed_avail): New global.
32051 (rpo_elim::eliminate_push_avail): Chain pushed avails.
32052 (unwind_state::avail_top): Add.
32053 (do_unwind): Rewrite unwinding of avail entries.
32054 (do_rpo_vn): Initialize last_pushed_avail and
32055 avail_top of the undo state.
32057 2021-02-09 Jakub Jelinek <jakub@redhat.com>
32059 PR middle-end/99004
32060 * calls.c (maybe_warn_rdwr_sizes): Change s0 and s1 type from
32061 const char * to char * and free those pointers after use.
32063 2021-02-09 Richard Biener <rguenther@suse.de>
32065 PR tree-optimization/99017
32066 * tree-vect-slp.c (vect_bb_vectorization_profitable_p): Allow
32067 zero vector cost entries.
32069 2021-02-08 Andre Vieira <andre.simoesdiasvieira@arm.com>
32071 PR middle-end/98974
32072 * tree-vect-stmts.c (vectorizable_condition): Remove shadow vec_num
32073 parameter in vectorizable_condition.
32075 2021-02-08 Richard Biener <rguenther@suse.de>
32078 * tree.c (walk_tree_1): Walk VECTOR_CST elements.
32080 2021-02-08 Martin Liska <mliska@suse.cz>
32083 * cfgexpand.c (pass_expand::execute): Parse per-function option
32084 flag_patchable_function_entry and use it.
32085 * common.opt: Remove function_entry_patch_area_size and
32086 function_entry_patch_area_start global variables.
32087 * opts.c (parse_and_check_patch_area): New function.
32088 (common_handle_option): Use it.
32089 * opts.h (parse_and_check_patch_area): New function.
32090 * toplev.c (process_options): Parse and use
32091 function_entry_patch_area_size.
32093 2021-02-08 Martin Sebor <msebor@redhat.com>
32095 * doc/extend.texi (attribute malloc): Correct typos.
32097 2021-02-05 Nathan Sidwell <nathan@acm.org>
32100 * gcc.c (driver::maybe_run_linker): Check for input file
32101 accessibility if not linking.
32103 2021-02-05 Richard Biener <rguenther@suse.de>
32105 PR tree-optimization/98855
32106 * tree-vectorizer.h (add_stmt_cost): New overload.
32107 * tree-vect-slp.c (li_cost_vec_cmp): New.
32108 (vect_bb_slp_scalar_cost): Cost individual loop regions
32109 separately. Account for the scalar instance root stmt.
32111 2021-02-05 Tom de Vries <tdevries@suse.de>
32114 * tree-switch-conversion.c (jump_table_cluster::emit): Add loc
32116 (bit_test_cluster::emit): Reuse location_t for newly created
32118 (switch_decision_tree::try_switch_expansion): Preserve
32120 * tree-switch-conversion.h: Change function signatures.
32122 2021-02-05 Jakub Jelinek <jakub@redhat.com>
32125 * config/i386/i386-options.c (m_NONE, m_ALL): Define.
32126 * config/i386/x86-tune.def (X86_TUNE_BRANCH_PREDICTION_HINTS,
32127 X86_TUNE_PROMOTE_QI_REGS): Use m_NONE instead of 0U.
32128 (X86_TUNE_QIMODE_MATH): Use m_ALL instead of ~0U.
32130 2021-02-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
32132 * config/aarch64/aarch64-simd-builtins.def (get_high): Define builtin.
32133 * config/aarch64/aarch64-simd.md (aarch64_get_high<mode>): Define.
32134 * config/aarch64/arm_neon.h (__GET_HIGH): Delete.
32135 (vget_high_f16): Reimplement using new builtin.
32136 (vget_high_f32): Likewise.
32137 (vget_high_f64): Likewise.
32138 (vget_high_p8): Likewise.
32139 (vget_high_p16): Likewise.
32140 (vget_high_p64): Likewise.
32141 (vget_high_s8): Likewise.
32142 (vget_high_s16): Likewise.
32143 (vget_high_s32): Likewise.
32144 (vget_high_s64): Likewise.
32145 (vget_high_u8): Likewise.
32146 (vget_high_u16): Likewise.
32147 (vget_high_u32): Likewise.
32148 (vget_high_u64): Likewise.
32150 2021-02-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
32152 * config/aarch64/aarch64-simd-builtins.def (get_low): Define builtin.
32153 * config/aarch64/aarch64-simd.md (aarch64_get_low<mode>): Define.
32154 * config/aarch64/arm_neon.h (__GET_LOW): Delete.
32155 (vget_low_f16): Reimplement using new builtin.
32156 (vget_low_f32): Likewise.
32157 (vget_low_f64): Likewise.
32158 (vget_low_p8): Likewise.
32159 (vget_low_p16): Likewise.
32160 (vget_low_p64): Likewise.
32161 (vget_low_s8): Likewise.
32162 (vget_low_s16): Likewise.
32163 (vget_low_s32): Likewise.
32164 (vget_low_s64): Likewise.
32165 (vget_low_u8): Likewise.
32166 (vget_low_u16): Likewise.
32167 (vget_low_u32): Likewise.
32168 (vget_low_u64): Likewise.
32170 2021-02-05 Kito Cheng <kito.cheng@sifive.com>
32172 * gcc.c (print_multilib_info): Check all required argument is provided
32175 2021-02-05 liuhongt <hongtao.liu@intel.com>
32178 * config/i386/i386-expand.c (ix86_expand_sse_cmp): Don't
32179 generate integer mask comparison for 128/256-bits vector when
32180 op_true/op_false is NULL_RTX or CONSTM1_RTX/CONST0_RTX. Also
32181 delete redundant !maskcmp condition.
32182 (ix86_expand_int_vec_cmp): Ditto but no redundant deletion
32184 (ix86_expand_sse_movcc): Delete definition of maskcmp, add the
32185 condition directly to if (maskcmp), add extra check for
32186 cmpmode, it should be MODE_INT.
32187 (ix86_expand_fp_vec_cmp): Pass NULL to ix86_expand_sse_cmp's
32188 parameters op_true/op_false.
32189 (ix86_use_mask_cmp_p): New.
32191 2021-02-05 liuhongt <hongtao.liu@intel.com>
32194 * config/i386/x86-tune.def (X86_TUNE_AVX256_UNALIGNED_LOAD_OPTIMAL):
32195 Remove m_GENERIC from ~list.
32196 (X86_TUNE_AVX256_UNALIGNED_STORE_OPTIMAL): Ditto.
32198 2021-02-04 David Malcolm <dmalcolm@redhat.com>
32201 * diagnostic-show-locus.c (compatible_locations_p): Require
32202 locations in the same macro map to be either both from the
32203 macro definition, or both from the macro arguments.
32205 2021-02-04 Jonathan Wright <jonathan.wright@arm.com>
32207 * config/aarch64/aarch64-simd-builtins.def: Add
32208 [su]mull_hi_lane[q] builtin generator macros.
32209 * config/aarch64/aarch64-simd.md
32210 (aarch64_<su>mull_hi_lane<mode>_insn): Define.
32211 (aarch64_<su>mull_hi_lane<mode>): Define.
32212 (aarch64_<su>mull_hi_laneq<mode>_insn): Define.
32213 (aarch64_<su>mull_hi_laneq<mode>): Define.
32214 * config/aarch64/arm_neon.h (vmull_high_lane_s16): Use RTL
32215 builtin instead of inline asm.
32216 (vmull_high_lane_s32): Likewise.
32217 (vmull_high_lane_u16): Likewise.
32218 (vmull_high_lane_u32): Likewise.
32219 (vmull_high_laneq_s16): Likewise.
32220 (vmull_high_laneq_s32): Likewise.
32221 (vmull_high_laneq_u16): Likewise.
32222 (vmull_high_laneq_u32): Liekwise.
32224 2021-02-04 Jonathan Wright <jonathan.wright@arm.com>
32226 * config/aarch64/aarch64-simd-builtins.def: Add [su]mull_hi_n
32227 builtin generator macros.
32228 * config/aarch64/aarch64-simd.md
32229 (aarch64_<su>mull_hi_n<mode>_insn): Define.
32230 (aarch64_<su>mull_hi_n<mode>): Define.
32231 * config/aarch64/arm_neon.h (vmull_high_n_s16): Use RTL builtin
32232 instead of inline asm.
32233 (vmull_high_n_s32): Likewise.
32234 (vmull_high_n_u16): Likewise.
32235 (vmull_high_n_u32): Likewise.
32237 2021-02-04 Richard Biener <rguenther@suse.de>
32239 PR tree-optimization/98855
32240 * tree-vect-loop.c (vectorizable_phi): Do not cost
32241 single-argument PHIs.
32242 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Likewise.
32243 * tree-vect-stmts.c (vectorizable_bswap): Also perform
32244 costing for SLP operation.
32246 2021-02-04 Martin Liska <mliska@suse.cz>
32248 * doc/extend.texi: Mention -mprefer-vector-width in target
32251 2021-02-03 Martin Sebor <msebor@redhat.com>
32253 PR tree-optimization/98937
32254 * tree-ssa-strlen.c (strlen_dom_walker::~strlen_dom_walker): Define.
32255 Flush pointer_query cache.
32257 2021-02-03 Aaron Sawdey <acsawdey@linux.ibm.com>
32259 * config/rs6000/genfusion.pl (gen_2logical): Add missing
32260 fixes based on patch review.
32261 * config/rs6000/fusion.md: Regenerate file.
32263 2021-02-03 Aaron Sawdey <acsawdey@linux.ibm.com>
32265 * config/rs6000/t-rs6000: Comment out auto generation of
32268 2021-02-03 Andrew Stubbs <ams@codesourcery.com>
32270 * config/gcn/gcn-opts.h (enum processor_type): Add PROCESSOR_GFX908.
32271 * config/gcn/gcn.c (gcn_omp_device_kind_arch_isa): Add gfx908.
32272 (output_file_start): Add gfx908.
32273 * config/gcn/gcn.opt (gpu_type): Add gfx908.
32274 * config/gcn/t-gcn-hsa (MULTILIB_OPTIONS): Add march=gfx908.
32275 (MULTILIB_DIRNAMES): Add gfx908.
32276 * config/gcn/mkoffload.c (EF_AMDGPU_MACH_AMDGCN_GFX908): New define.
32277 (main): Recognize gfx908.
32278 * config/gcn/t-omp-device: Add gfx908.
32280 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
32282 * config/aarch64/aarch64-simd-builtins.def: Add
32283 [su]mlsl_hi_lane[q] builtin macro generators.
32284 * config/aarch64/aarch64-simd.md
32285 (aarch64_<su>mlsl_hi_lane<mode>_insn): Define.
32286 (aarch64_<su>mlsl_hi_lane<mode>): Define.
32287 (aarch64_<su>mlsl_hi_laneq<mode>_insn): Define.
32288 (aarch64_<su>mlsl_hi_laneq<mode>): Define.
32289 * config/aarch64/arm_neon.h (vmlsl_high_lane_s16): Use RTL
32290 builtin instead of inline asm.
32291 (vmlsl_high_lane_s32): Likewise.
32292 (vmlsl_high_lane_u16): Likewise.
32293 (vmlsl_high_lane_u32): Likewise.
32294 (vmlsl_high_laneq_s16): Likewise.
32295 (vmlsl_high_laneq_s32): Likewise.
32296 (vmlsl_high_laneq_u16): Likewise.
32297 (vmlsl_high_laneq_u32): Likewise.
32298 (vmlal_high_laneq_u32): Likewise.
32300 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
32302 * config/aarch64/aarch64-simd-builtins.def: Add
32303 [su]mlal_hi_lane[q] builtin generator macros.
32304 * config/aarch64/aarch64-simd.md
32305 (aarch64_<su>mlal_hi_lane<mode>_insn): Define.
32306 (aarch64_<su>mlal_hi_lane<mode>): Define.
32307 (aarch64_<su>mlal_hi_laneq<mode>_insn): Define.
32308 (aarch64_<su>mlal_hi_laneq<mode>): Define.
32309 * config/aarch64/arm_neon.h (vmlal_high_lane_s16): Use RTL
32310 builtin instead of inline asm.
32311 (vmlal_high_lane_s32): Likewise.
32312 (vmlal_high_lane_u16): Likewise.
32313 (vmlal_high_lane_u32): Likewise.
32314 (vmlal_high_laneq_s16): Likewise.
32315 (vmlal_high_laneq_s32): Likewise.
32316 (vmlal_high_laneq_u16): Likewise.
32317 (vmlal_high_laneq_u32): Likewise.
32319 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
32321 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlsl_hi_n
32322 builtin generator macros.
32323 * config/aarch64/aarch64-simd.md (aarch64_<su>mlsl_hi_n<mode>_insn):
32325 (aarch64_<su>mlsl_hi_n<mode>): Define.
32326 * config/aarch64/arm_neon.h (vmlsl_high_n_s16): Use RTL builtin
32327 instead of inline asm.
32328 (vmlsl_high_n_s32): Likewise.
32329 (vmlsl_high_n_u16): Likewise.
32330 (vmlsl_high_n_u32): Likewise.
32332 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
32334 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlal_hi_n
32335 builtin generator macros.
32336 * config/aarch64/aarch64-simd.md (aarch64_<su>mlal_hi_n<mode>_insn):
32338 (aarch64_<su>mlal_hi_n<mode>): Define.
32339 * config/aarch64/arm_neon.h (vmlal_high_n_s16): Use RTL builtin
32340 instead of inline asm.
32341 (vmlal_high_n_s32): Likewise.
32342 (vmlal_high_n_u16): Likewise.
32343 (vmlal_high_n_u32): Likewise.
32345 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
32347 * config/aarch64/aarch64-simd-builtins.def: Add RTL builtin
32349 * config/aarch64/aarch64-simd.md (*aarch64_<su>mlal_hi<mode>):
32351 (aarch64_<su>mlal_hi<mode>_insn): This.
32352 (aarch64_<su>mlal_hi<mode>): Define.
32353 * config/aarch64/arm_neon.h (vmlal_high_s8): Use RTL builtin
32354 instead of inline asm.
32355 (vmlal_high_s16): Likewise.
32356 (vmlal_high_s32): Likewise.
32357 (vmlal_high_u8): Likewise.
32358 (vmlal_high_u16): Likewise.
32359 (vmlal_high_u32): Likewise.
32361 2021-02-03 Ilya Leoshkevich <iii@linux.ibm.com>
32363 * lra-spills.c (remove_pseudos): Call lra_update_insn_recog_data()
32364 after calling alter_subreg() on a (mem).
32366 2021-02-03 Martin Liska <mliska@suse.cz>
32369 * lto-streamer-out.c (produce_lto_section): Fill up missing
32371 * lto-streamer.h (struct lto_section): Add _padding field.
32373 2021-02-03 Richard Biener <rguenther@suse.de>
32375 * lto-streamer.c (lto_get_section_name): Free temporary
32377 * tree-loop-distribution.c
32378 (loop_distribution::merge_dep_scc_partitions): Free edge data.
32380 2021-02-03 Jakub Jelinek <jakub@redhat.com>
32382 PR middle-end/97487
32383 * ifcvt.c (noce_can_force_operand): New function.
32384 (noce_emit_move_insn): Use it.
32385 (noce_try_sign_mask): Likewise. Formatting fix.
32387 2021-02-03 Jakub Jelinek <jakub@redhat.com>
32389 PR middle-end/97971
32390 * lra-constraints.c (process_alt_operands): For inline asm, don't call
32391 fatal_insn, but instead return false.
32393 2021-02-03 Jakub Jelinek <jakub@redhat.com>
32395 PR tree-optimization/98287
32396 * config/i386/mmx.md (<insn><mode>3): For shifts don't enable expander
32399 2021-02-03 Tamar Christina <tamar.christina@arm.com>
32401 PR tree-optimization/98928
32402 * tree-vect-loop.c (vect_analyze_loop_2): Change
32403 STMT_VINFO_SLP_VECT_ONLY to STMT_VINFO_SLP_VECT_ONLY_PATTERN.
32404 * tree-vect-slp-patterns.c (complex_pattern::build): Likewise.
32405 * tree-vectorizer.h (STMT_VINFO_SLP_VECT_ONLY_PATTERN): New.
32406 (class _stmt_vec_info): Add slp_vect_pattern_only_p.
32408 2021-02-02 Richard Biener <rguenther@suse.de>
32410 * gimple-loop-interchange.cc (prepare_data_references):
32412 * gimple-loop-jam.c (tree_loop_unroll_and_jam): Likewise.
32413 * tree-ssa-loop-im.c (hoist_memory_references): Likewise.
32414 * tree-vect-stmts.c (vectorizable_condition): Do not
32416 (vectorizable_comparison): Likewise.
32418 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
32420 * config/aarch64/aarch64-simd-builtins.def (ursqrte): Define builtin.
32421 * config/aarch64/aarch64-simd.md (aarch64_ursqrte<mode>): New pattern.
32422 * config/aarch64/arm_neon.h (vrsqrte_u32): Reimplement using builtin.
32423 (vrsqrteq_u32): Likewise.
32425 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
32427 * config/aarch64/aarch64-simd-builtins.def (sqxtun2): Define builtin.
32428 * config/aarch64/aarch64-simd.md (aarch64_sqxtun2<mode>_le): Define.
32429 (aarch64_sqxtun2<mode>_be): Likewise.
32430 (aarch64_sqxtun2<mode>): Likewise.
32431 * config/aarch64/arm_neon.h (vqmovun_high_s16): Reimplement using builtin.
32432 (vqmovun_high_s32): Likewise.
32433 (vqmovun_high_s64): Likewise.
32434 * config/aarch64/iterators.md (UNSPEC_SQXTUN2): Define.
32436 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
32438 * config/aarch64/aarch64-simd-builtins.def (bfdot_lane, bfdot_laneq): Use
32440 (bfmlalb_lane, bfmlalt_lane, bfmlalb_lane_q, bfmlalt_lane_q): Use FP flags.
32442 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
32444 * config/aarch64/aarch64-simd-builtins.def (fcmla_lane0, fcmla_lane90,
32445 fcmla_lane180, fcmla_lane270, fcmlaq_lane0, fcmlaq_lane90, fcmlaq_lane180,
32446 fcmlaq_lane270, scvtf, ucvtf, fcvtzs, fcvtzu, scvtfsi, scvtfdi, ucvtfsi,
32447 ucvtfdi, fcvtzshf, fcvtzuhf, fmlal_lane_low, fmlsl_lane_low,
32448 fmlal_laneq_low, fmlsl_laneq_low, fmlalq_lane_low, fmlslq_lane_low,
32449 fmlalq_laneq_low, fmlslq_laneq_low, fmlal_lane_high, fmlsl_lane_high,
32450 fmlal_laneq_high, fmlsl_laneq_high, fmlalq_lane_high, fmlslq_lane_high,
32451 fmlalq_laneq_high, fmlslq_laneq_high): Use FP flags.
32453 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
32455 * config/aarch64/aarch64-builtins.c (FLAG_LOAD): Define.
32456 * config/aarch64/aarch64-simd-builtins.def (ld1x2, ld2, ld3, ld4, ld2r,
32457 ld3r, ld4r, ld1, ld1x3, ld1x4): Use LOAD flags.
32459 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
32461 * config/aarch64/aarch64-simd-builtins.def (combine, zip1, zip2,
32462 uzp1, uzp2, trn1, trn2, simd_bsl): Use AUTO_FP flags.
32464 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
32466 * config/aarch64/aarch64-simd-builtins.def (clrsb, clz, ctz, popcount,
32467 vec_smult_lane_, vec_smlal_lane_, vec_smult_laneq_, vec_smlal_laneq_,
32468 vec_umult_lane_, vec_umlal_lane_, vec_umult_laneq_, vec_umlal_laneq_,
32469 ashl, sshl, ushl, srshl, urshl, sdot_lane, udot_lane, sdot_laneq,
32470 udot_laneq, usdot_lane, usdot_laneq, sudot_lane, sudot_laneq, ashr,
32471 ashr_simd, lshr, lshr_simd, srshr_n, urshr_n, ssra_n, usra_n, srsra_n,
32472 ursra_n, sshll_n, ushll_n, sshll2_n, ushll2_n, ssri_n, usri_n, ssli_n,
32473 ssli_n, usli_n, bswap, rbit, simd_bsl, eor3q, rax1q, xarq, bcaxq): Use
32474 NONE builtin flags.
32476 2021-02-02 Jakub Jelinek <jakub@redhat.com>
32478 PR tree-optimization/98848
32479 * tree-vect-patterns.c (vect_recog_over_widening_pattern): Punt if
32480 STMT_VINFO_DEF_TYPE (last_stmt_info) is vect_reduction_def.
32482 2021-02-02 Kito Cheng <kito.cheng@sifive.com>
32485 * expr.c: Check mode before calling store_expr.
32487 2021-02-02 Christophe Lyon <christophe.lyon@linaro.org>
32489 * config/arm/iterators.md (supf): Remove VORNQ_S and VORNQ_U.
32491 * config/arm/mve.md (mve_vornq_s<mode>): New entry for vorn
32492 instruction using expression ior.
32493 (mve_vornq_u<mode>): New expander.
32494 (mve_vornq_f<mode>): Use ior code instead of unspec.
32495 * config/arm/unspecs.md (VORNQ_S, VORNQ_U, VORNQ_F): Remove.
32497 2021-02-02 Alexandre Oliva <oliva@adacore.com>
32499 * tree-nested.c (convert_nonlocal_reference_op): Move
32500 current_function_decl restore after re-gimplification.
32501 (convert_local_reference_op): Likewise.
32503 2021-02-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
32505 * config/aarch64/aarch64-simd-builtins.def (rshrn, rshrn2):
32507 * config/aarch64/aarch64-simd.md (aarch64_rshrn<mode>_insn_le):
32509 (aarch64_rshrn<mode>_insn_be): Likewise.
32510 (aarch64_rshrn<mode>): Likewise.
32511 (aarch64_rshrn2<mode>_insn_le): Likewise.
32512 (aarch64_rshrn2<mode>_insn_be): Likewise.
32513 (aarch64_rshrn2<mode>): Likewise.
32514 * config/aarch64/aarch64.md (unspec): Add UNSPEC_RSHRN.
32515 * config/aarch64/arm_neon.h (vrshrn_high_n_s16): Reimplement
32517 (vrshrn_high_n_s32): Likewise.
32518 (vrshrn_high_n_s64): Likewise.
32519 (vrshrn_high_n_u16): Likewise.
32520 (vrshrn_high_n_u32): Likewise.
32521 (vrshrn_high_n_u64): Likewise.
32522 (vrshrn_n_s16): Likewise.
32523 (vrshrn_n_s32): Likewise.
32524 (vrshrn_n_s64): Likewise.
32525 (vrshrn_n_u16): Likewise.
32526 (vrshrn_n_u32): Likewise.
32527 (vrshrn_n_u64): Likewise.
32529 2021-02-01 Sergei Trofimovich <siarheit@google.com>
32531 PR tree-optimization/98499
32532 * ipa-modref.c (analyze_ssa_name_flags): treat RVO
32533 conservatively and assume all possible side-effects.
32535 2021-02-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
32537 * config/aarch64/aarch64-simd-builtins.def (vec_unpacks_hi,
32538 vec_unpacku_hi_): Define builtins.
32539 * config/aarch64/arm_neon.h (vmovl_high_s8): Reimplement using
32541 (vmovl_high_s16): Likewise.
32542 (vmovl_high_s32): Likewise.
32543 (vmovl_high_u8): Likewise.
32544 (vmovl_high_u16): Likewise.
32545 (vmovl_high_u32): Likewise.
32547 2021-02-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
32549 * config/aarch64/aarch64-simd-builtins.def (sabdl, uabdl):
32551 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl<mode>): New
32553 * config/aarch64/aarch64.md (unspec): Define UNSPEC_SABDL,
32555 * config/aarch64/arm_neon.h (vabdl_s8): Reimplemet using
32557 (vabdl_s16): Likewise.
32558 (vabdl_s32): Likewise.
32559 (vabdl_u8): Likewise.
32560 (vabdl_u16): Likewise.
32561 (vabdl_u32): Likewise.
32562 * config/aarch64/iterators.md (ABDL): New int iterator.
32563 (sur): Handle UNSPEC_SABDL, UNSPEC_UABDL.
32565 2021-02-01 Martin Sebor <msebor@redhat.com>
32567 * tree.h (BLOCK_VARS): Add comment.
32568 (BLOCK_SUBBLOCKS): Same.
32569 (BLOCK_SUPERCONTEXT): Same.
32570 (BLOCK_ABSTRACT_ORIGIN): Same.
32571 (inlined_function_outer_scope_p): Same.
32573 2021-02-01 Martin Sebor <msebor@redhat.com>
32575 PR middle-end/97172
32576 * attribs.c (attr_access::free_lang_data): Define new function.
32577 * attribs.h (attr_access::free_lang_data): Declare new function.
32579 2021-02-01 Richard Biener <rguenther@suse.de>
32581 * vec.h (auto_vec::auto_vec): Add memory stat parameters
32583 * bitmap.h (auto_bitmap::auto_bitmap): Likewise.
32585 2021-02-01 Tamar Christina <tamar.christina@arm.com>
32587 * config/aarch64/aarch64-simd.md (aarch64_<su>mlal_n<mode>,
32588 aarch64_<su>mlsl<mode>, aarch64_<su>mlsl_n<mode>): Flip mult operands.
32590 2021-02-01 Richard Biener <rguenther@suse.de>
32592 PR rtl-optimization/98863
32593 * config/i386/i386-features.c (convert_scalars_to_vector):
32594 Set DF_RD_PRUNE_DEAD_DEFS.
32596 2021-01-31 Eric Botcazou <ebotcazou@adacore.com>
32598 * system.h (SIZE_MAX): Define if not already defined.
32600 2021-01-30 Aaron Sawdey <acsawdey@linux.ibm.com>
32602 * config/rs6000/genfusion.pl (gen_2logical): New function to
32603 generate patterns for logical-logical fusion.
32604 * config/rs6000/fusion.md: Regenerated patterns.
32605 * config/rs6000/rs6000-cpus.def: Add
32606 OPTION_MASK_P10_FUSION_2LOGICAL.
32607 * config/rs6000/rs6000.c (rs6000_option_override_internal):
32608 Enable logical-logical fusion for p10.
32609 * config/rs6000/rs6000.opt: Add -mpower10-fusion-2logical.
32611 2021-01-30 David Edelsohn <dje.gcc@gmail.com>
32613 * config/rs6000/rs6000.opt: Add periods to new AIX options.
32615 2021-01-30 David Edelsohn <dje.gcc@gmail.com>
32617 * config/rs6000/rs6000.opt (mabi=vec-extabi): New.
32618 (mabi=vec-default): New.
32619 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Define
32620 __EXTABI__ for AIX Vector extended ABI.
32621 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Print AIX Vector
32623 (conditional_register_usage): If AIX vec_extabi enabled, vs20-vs31
32625 * doc/invoke.texi (PowerPC mabi): Add AIX vec-extabi and vec-default.
32627 2021-01-30 Jakub Jelinek <jakub@redhat.com>
32629 * config/i386/i386-features.c (remove_partial_avx_dependency): Clear
32630 DF_DEFER_INSN_RESCAN after calling df_process_deferred_rescans.
32632 2021-01-29 Vladimir N. Makarov <vmakarov@redhat.com>
32635 * lra-constraints.c (in_class_p): Don't narrow class only for REG
32638 2021-01-29 Will Schmidt <will_schmidt@vnet.ibm.com>
32640 * config/rs6000/rs6000-call.c (rs6000_expand_binup_builtin): Add
32641 clauses for CODE_FOR_vsx_xvcvuxddp_scale and
32642 CODE_FOR_vsx_xvcvsxddp_scale to the parameter checking code.
32644 2021-01-29 Andrew MacLeod <amacleod@redhat.com>
32646 PR tree-optimization/98866
32647 * gimple-range-gori.h (gori_compute:set_range_invariant): New.
32648 * gimple-range-gori.cc (gori_map::set_range_invariant): New.
32649 (gori_map::m_maybe_invariant): Rename from all_outgoing.
32650 (gori_map::gori_map): Rename all_outgoing to m_maybe_invariant.
32651 (gori_map::is_export_p): Ditto.
32652 (gori_map::calculate_gori): Ditto.
32653 (gori_compute::set_range_invariant): New.
32654 * gimple-range.cc (gimple_ranger::range_of_stmt): Set range
32655 invariant for pointers evaluating to [1, +INF].
32657 2021-01-29 Richard Biener <rguenther@suse.de>
32659 PR rtl-optimization/98863
32660 * config/i386/i386-features.c (remove_partial_avx_dependency):
32661 Do not perform DF analysis.
32662 (pass_data_remove_partial_avx_dependency): Remove
32665 2021-01-29 Jonathan Wright <jonathan.wright@arm.com>
32667 * config/aarch64/aarch64-simd-builtins.def: Add [su]mull_n
32668 builtin generator macros.
32669 * config/aarch64/aarch64-simd.md (aarch64_<su>mull_n<mode>):
32671 * config/aarch64/arm_neon.h (vmull_n_s16): Use RTL builtin
32672 instead of inline asm.
32673 (vmull_n_s32): Likewise.
32674 (vmull_n_u16): Likewise.
32675 (vmull_n_u32): Likewise.
32677 2021-01-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
32679 * config/aarch64/aarch64-simd-builtins.def (sabdl2, uabdl2):
32681 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
32683 (aarch64_<sur>abdl2<mode>): ... This.
32684 (<sur>sadv16qi): Adjust use of above.
32685 * config/aarch64/arm_neon.h (vabdl_high_s8): Reimplement using
32687 (vabdl_high_s16): Likewise.
32688 (vabdl_high_s32): Likewise.
32689 (vabdl_high_u8): Likewise.
32690 (vabdl_high_u16): Likewise.
32691 (vabdl_high_u32): Likewise.
32693 2021-01-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
32695 * config/aarch64/aarch64-simd-builtins.def (sabal2): Define
32697 (uabal2): Likewise.
32698 * config/aarch64/aarch64-simd.md (aarch64_<sur>abal2<mode>): New
32700 * config/aarch64/aarch64.md (unspec): Add UNSPEC_SABAL2 and
32702 * config/aarch64/arm_neon.h (vabal_high_s8): Reimplement using
32704 (vabal_high_s16): Likewise.
32705 (vabal_high_s32): Likewise.
32706 (vabal_high_u8): Likewise.
32707 (vabal_high_u16): Likewise.
32708 (vabal_high_u32): Likewise.
32709 * config/aarch64/iterators.md (ABAL2): New mode iterator.
32710 (sur): Handle UNSPEC_SABAL2, UNSPEC_UABAL2.
32712 2021-01-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
32714 * config/aarch64/aarch64-simd-builtins.def (sabal): Define
32717 * config/aarch64/aarch64-simd.md (aarch64_<sur>abal<mode>_4):
32719 (aarch64_<sur>abal<mode>): ... This
32720 (<sur>sadv16qi): Adust use of the above.
32721 * config/aarch64/arm_neon.h (vabal_s8): Reimplement using
32723 (vabal_s16): Likewise.
32724 (vabal_s32): Likewise.
32725 (vabal_u8): Likewise.
32726 (vabal_u16): Likewise.
32727 (vabal_u32): Likewise.
32729 2021-01-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
32731 * config/aarch64/aarch64-simd-builtins.def (saddlv, uaddlv):
32733 * config/aarch64/aarch64-simd.md (aarch64_<su>addlv<mode>):
32735 * config/aarch64/arm_neon.h (vaddlv_s8): Reimplement using
32737 (vaddlv_s16): Likewise.
32738 (vaddlv_u8): Likewise.
32739 (vaddlv_u16): Likewise.
32740 (vaddlvq_s8): Likewise.
32741 (vaddlvq_s16): Likewise.
32742 (vaddlvq_s32): Likewise.
32743 (vaddlvq_u8): Likewise.
32744 (vaddlvq_u16): Likewise.
32745 (vaddlvq_u32): Likewise.
32746 (vaddlv_s32): Likewise.
32747 (vaddlv_u32): Likewise.
32748 * config/aarch64/iterators.md (VDQV_L): New mode iterator.
32749 (unspec): Add UNSPEC_SADDLV, UNSPEC_UADDLV.
32750 (Vwstype): New mode attribute.
32752 (VWIDE_S): Likewise.
32753 (USADDLV): New int iterator.
32754 (su): Handle UNSPEC_SADDLV, UNSPEC_UADDLV.
32756 2021-01-29 Jonathan Wright <jonathan.wright@arm.com>
32758 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlsl_lane[q]
32759 builtin generator macros.
32760 * config/aarch64/aarch64-simd.md (aarch64_vec_<su>mlsl_lane<Qlane>):
32762 * config/aarch64/arm_neon.h (vmlsl_lane_s16): Use RTL builtin
32763 instead of inline asm.
32764 (vmlsl_lane_s32): Likewise.
32765 (vmlsl_lane_u16): Likewise.
32766 (vmlsl_lane_u32): Likewise.
32767 (vmlsl_laneq_s16): Likewise.
32768 (vmlsl_laneq_s32): Likewise.
32769 (vmlsl_laneq_u16): Likewise.
32770 (vmlsl_laneq_u32): Likewise.
32772 2021-01-29 Richard Biener <rguenther@suse.de>
32774 * doc/invoke.texi (--param max-gcse-memory): Document unit
32776 * gcse.c (gcse_or_cprop_is_too_expensive): Adjust.
32777 * params.opt (--param max-gcse-memory): Adjust default and
32778 document unit of size.
32780 2021-01-29 Richard Biener <rguenther@suse.de>
32782 PR rtl-optimization/98863
32783 * gcse.c (gcse_or_cprop_is_too_expensive): Use unsigned
32784 HOST_WIDE_INT for the memory estimate.
32786 2021-01-29 Bin Cheng <bin.cheng@linux.alibaba.com>
32787 Richard Biener <rguenther@suse.de>
32789 PR tree-optimization/97627
32790 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
32791 Do not analyze fake edges.
32793 2021-01-29 Richard Biener <rguenther@suse.de>
32795 PR rtl-optimization/98144
32796 * df.h (df_mir_bb_info): Add con_visited member.
32797 * df-problems.c (df_mir_alloc): Initialize con_visited,
32798 do not fully populate IN and OUT.
32799 (df_mir_reset): Likewise.
32800 (df_mir_confluence_0): Set con_visited.
32801 (df_mir_confluence_n): Properly handle implicitely
32802 fully populated IN and OUT as designated by con_visited
32803 and update con_visited accordingly.
32805 2021-01-29 Jakub Jelinek <jakub@redhat.com>
32808 * config/arm/vec-common.md (mve_vshlq_<supf><mode>,
32809 vashl<mode>3, vashr<mode>3, vlshr<mode>3): Add
32810 && !TARGET_REALLY_IWMMXT to conditions.
32812 2021-01-29 Jakub Jelinek <jakub@redhat.com>
32815 * cfgbuild.c (find_bb_boundaries): Reset debug_insn when seeing
32818 2021-01-28 Marek Polacek <polacek@redhat.com>
32821 * stor-layout.c (finalize_type_size): If we reset TYPE_USER_ALIGN in
32822 the main variant, maybe reset it in its variants too.
32823 * tree.c (check_base_type): Return true only if TYPE_USER_ALIGN match.
32824 (check_aligned_type): Check if TYPE_USER_ALIGN match.
32826 2021-01-28 Christophe Lyon <christophe.lyon@linaro.org>
32829 * config/arm/arm.c (arm_rtx_costs_internal): Adjust cost of vector
32830 of constant zero for comparisons.
32832 2021-01-28 Michael Meissner <meissner@linux.ibm.com>
32834 * config/rs6000/rs6000.c (rs6000_mangle_decl_assembler_name): Add
32835 support for mapping built-in function names for long double
32836 built-in functions if long double is IEEE 128-bit.
32838 2021-01-28 Jonathan Wright <jonathan.wright@arm.com>
32840 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlsl_n
32841 builtin generator macros.
32842 * config/aarch64/aarch64-simd.md (aarch64_<su>mlsl_n<mode>):
32844 * config/aarch64/arm_neon.h (vmlsl_n_s16): Use RTL builtin
32845 instead of inline asm.
32846 (vmlsl_n_s32): Likewise.
32847 (vmlsl_n_u16): Likewise.
32848 (vmlsl_n_u32): Likewise.
32850 2021-01-28 Jonathan Wright <jonathan.wright@arm.com>
32852 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlal_n
32853 builtin generator macros.
32854 * config/aarch64/aarch64-simd.md (aarch64_<su>mlal_n<mode>):
32856 * config/aarch64/arm_neon.h (vmlal_n_s16): Use RTL builtin
32857 instead of inline asm.
32858 (vmlal_n_s32): Likewise.
32859 (vmlal_n_u16): Likewise.
32860 (vmlal_n_u32): Likewise.
32862 2021-01-28 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
32864 * config/aarch64/aarch64-simd-builtins.def (shrn2): Define
32866 * config/aarch64/aarch64-simd.md (aarch64_shrn2<mode>_insn_le):
32868 (aarch64_shrn2<mode>_insn_be): Likewise.
32869 (aarch64_shrn2<mode>): Likewise.
32870 * config/aarch64/arm_neon.h (vshrn_high_n_s16): Reimlplement
32872 (vshrn_high_n_s32): Likewise.
32873 (vshrn_high_n_s64): Likewise.
32874 (vshrn_high_n_u16): Likewise.
32875 (vshrn_high_n_u32): Likewise.
32876 (vshrn_high_n_u64): Likewise.
32878 2021-01-28 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
32880 * config/aarch64/aarch64-simd-builtins.def (shrn): Define
32882 * config/aarch64/aarch64-simd.md (aarch64_shrn<mode>_insn_le):
32884 (aarch64_shrn<mode>_insn_be): Likewise.
32885 (aarch64_shrn<mode>): Likewise.
32886 * config/aarch64/arm_neon.h (vshrn_n_s16): Reimplement using
32888 (vshrn_n_s32): Likewise.
32889 (vshrn_n_s64): Likewise.
32890 (vshrn_n_u16): Likewise.
32891 (vshrn_n_u32): Likewise.
32892 (vshrn_n_u64): Likewise.
32893 * config/aarch64/iterators.md (vn_mode): New mode attribute.
32895 2021-01-28 Richard Biener <rguenther@suse.de>
32897 PR rtl-optimization/80960
32898 * dse.c (check_mem_read_rtx): Call get_addr on the
32901 2021-01-28 Xionghu Luo <luoxhu@linux.ibm.com>
32902 David Edelsohn <dje.gcc@gmail.com>
32905 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
32906 Don't generate VIEW_CONVERT_EXPR for fcode ALTIVEC_BUILTIN_VEC_INSERT
32908 * config/rs6000/rs6000-protos.h (rs6000_expand_vector_set_var):
32910 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Remove the
32911 wrapper call rs6000_expand_vector_set_var for cleanup. Call
32912 rs6000_expand_vector_set_var_p9 and rs6000_expand_vector_set_var_p8
32914 (rs6000_expand_vector_set_var): Delete.
32915 (rs6000_expand_vector_set_var_p9): Make static.
32916 (rs6000_expand_vector_set_var_p8): Make static.
32918 2021-01-28 Xing GUO <higuoxing@gmail.com>
32920 * common/config/riscv/riscv-common.c
32921 (riscv_subset_list::parsing_subset_version): Fix -march option parsing
32922 when `p` extension exists.
32924 2021-01-27 Vladimir N. Makarov <vmakarov@redhat.com>
32926 PR rtl-optimization/97684
32927 * ira.c (ira): Call ira_set_pseudo_classes before
32928 update_equiv_regs when it is necessary.
32930 2021-01-27 Jakub Jelinek <jakub@redhat.com>
32933 * config/aarch64/aarch64.md (*aarch64_bfxilsi_uxtw): Use
32934 %w0, %w1 and %2 instead of %0, %1 and %2.
32936 2021-01-27 Aaron Sawdey <acsawdey@linux.ibm.com>
32938 * config/rs6000/genfusion.pl: New script to generate
32939 define_insn_and_split patterns so combine can arrange fused
32940 instructions next to each other.
32941 * config/rs6000/fusion.md: New file, generated fused instruction
32942 patterns for combine.
32943 * config/rs6000/predicates.md (const_m1_to_1_operand): New predicate.
32944 (non_update_memory_operand): New predicate.
32945 * config/rs6000/rs6000-cpus.def: Add OPTION_MASK_P10_FUSION and
32946 OPTION_MASK_P10_FUSION_LD_CMPI to ISA_3_1_MASKS_SERVER and
32948 * config/rs6000/rs6000-protos.h (address_is_non_pfx_d_or_x): Add
32950 * config/rs6000/rs6000.c (rs6000_option_override_internal):
32951 Automatically set OPTION_MASK_P10_FUSION and
32952 OPTION_MASK_P10_FUSION_LD_CMPI if target is power10.
32953 (rs600_opt_masks): Allow -mpower10-fusion
32954 in function attributes.
32955 (address_is_non_pfx_d_or_x): New function.
32956 * config/rs6000/rs6000.h: Add MASK_P10_FUSION.
32957 * config/rs6000/rs6000.md: Include fusion.md.
32958 * config/rs6000/rs6000.opt: Add -mpower10-fusion
32959 and -mpower10-fusion-ld-cmpi.
32960 * config/rs6000/t-rs6000: Add dependencies involving fusion.md.
32962 2021-01-27 Jonathan Wright <jonathan.wright@arm.com>
32964 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlal
32965 builtin generator macros.
32966 * config/aarch64/aarch64-simd.md (*aarch64_<su>mlal<mode>):
32968 (aarch64_<su>mlal<mode>): This.
32969 * config/aarch64/arm_neon.h (vmlal_s8): Use RTL builtin
32970 instead of inline asm.
32971 (vmlal_s16): Likewise.
32972 (vmlal_s32): Likewise.
32973 (vmlal_u8): Likewise.
32974 (vmlal_u16): Likewise.
32975 (vmlal_u32): Likewise.
32977 2021-01-27 Richard Biener <rguenther@suse.de>
32979 PR tree-optimization/98854
32980 * tree-vect-slp.c (vect_build_slp_tree_2): Also build
32981 PHIs from scalars when the number of CTORs matches the
32982 number of children.
32984 2021-01-27 Jonathan Wright <jonathan.wright@arm.com>
32986 * config/aarch64/aarch64-simd-builtins.def: Add mls_n builtin
32988 * config/aarch64/aarch64-simd.md (*aarch64_mls_elt_merge<mode>):
32990 (aarch64_mls_n<mode>): This.
32991 * config/aarch64/arm_neon.h (vmls_n_s16): Use RTL builtin
32993 (vmls_n_s32): Likewise.
32994 (vmls_n_u16): Likewise.
32995 (vmls_n_u32): Likewise.
32996 (vmlsq_n_s16): Likewise.
32997 (vmlsq_n_s32): Likewise.
32998 (vmlsq_n_u16): Likewise.
32999 (vmlsq_n_u32): Likewise.
33001 2021-01-27 Jonathan Wright <jonathan.wright@arm.com>
33003 * config/aarch64/aarch64-simd-builtins.def: Add mls builtin
33005 * config/aarch64/arm_neon.h (vmls_s8): Use RTL builtin rather
33007 (vmls_s16): Likewise.
33008 (vmls_s32): Likewise.
33009 (vmls_u8): Likewise.
33010 (vmls_u16): Likewise.
33011 (vmls_u32): Likewise.
33012 (vmlsq_s8): Likewise.
33013 (vmlsq_s16): Likewise.
33014 (vmlsq_s32): Likewise.
33015 (vmlsq_u8): Likewise.
33016 (vmlsq_u16): Likewise.
33017 (vmlsq_u32): Likewise.
33019 2021-01-27 Jonathan Wright <jonathan.wright@arm.com>
33021 * config/aarch64/aarch64-simd-builtins.def: Add mla_n builtin
33023 * config/aarch64/aarch64-simd.md (*aarch64_mla_elt_merge<mode>):
33025 (aarch64_mla_n<mode>): This.
33026 * config/aarch64/arm_neon.h (vmla_n_s16): Use RTL builtin
33028 (vmla_n_s32): Likewise.
33029 (vmla_n_u16): Likewise.
33030 (vmla_n_u32): Likewise.
33031 (vmlaq_n_s16): Likewise.
33032 (vmlaq_n_s32): Likewise.
33033 (vmlaq_n_u16): Likewise.
33034 (vmlaq_n_u32): Likewise.
33036 2021-01-27 liuhongt <hongtao.liu@intel.com>
33039 * config/i386/sse.md (sse2_gt<mode>3): Drop !TARGET_XOP in condition.
33040 (*sse2_eq<mode>3): Ditto.
33042 2021-01-27 Jakub Jelinek <jakub@redhat.com>
33044 * tree-pass.h (PROP_trees): Rename to ...
33045 (PROP_gimple): ... this.
33046 * cfgexpand.c (pass_data_expand): Replace PROP_trees with PROP_gimple.
33047 * passes.c (execute_function_dump, execute_function_todo,
33048 execute_one_ipa_transform_pass, execute_one_pass): Likewise.
33049 * varpool.c (ctor_for_folding): Likewise.
33051 2021-01-27 Jakub Jelinek <jakub@redhat.com>
33053 PR tree-optimization/97260
33054 * varpool.c: Include tree-pass.h.
33055 (ctor_for_folding): In GENERIC return DECL_INITIAL for TREE_READONLY
33056 non-TREE_SIDE_EFFECTS automatic variables.
33058 2021-01-26 Paul Fee <paul.f.fee@gmail.com>
33060 * doc/cpp.texi (__cplusplus): Document value for -std=c++23
33062 * doc/invoke.texi: Document -std=c++23 and -std=gnu++23.
33063 * dwarf2out.c (highest_c_language): Recognise C++20 and C++23.
33064 (gen_compile_unit_die): Recognise C++23.
33066 2021-01-26 Jakub Jelinek <jakub@redhat.com>
33069 * dwarf2asm.c (dw2_assemble_integer): Cast DWARF2_ADDR_SIZE to int
33072 2021-01-26 Jakub Jelinek <jakub@redhat.com>
33075 * config/aarch64/aarch64.c (aarch64_mask_and_shift_for_ubfiz_p):
33076 Use UINTVAL (shft_amnt) and UINTVAL (mask) instead of INTVAL (shft_amnt)
33077 and INTVAL (mask). Add && INTVAL (mask) > 0 condition.
33079 2021-01-26 Richard Biener <rguenther@suse.de>
33081 * gimple-pretty-print.c (dump_binary_rhs): Handle
33082 VEC_WIDEN_{PLUS,MINUS}_{LO,HI}_EXPR.
33084 2021-01-26 Richard Biener <rguenther@suse.de>
33086 PR middle-end/98726
33087 * tree.h (vector_cst_int_elt): Remove.
33088 * tree.c (vector_cst_int_elt): Use poly_wide_int for computations,
33091 2021-01-26 Andrew Stubbs <ams@codesourcery.com>
33093 * config/gcn/gcn.c (gcn_expand_reduc_scalar): Use move instructions
33094 for V64DFmode min/max reductions.
33096 2021-01-26 Jakub Jelinek <jakub@redhat.com>
33098 * dwarf2asm.c (dw2_assemble_integer): Handle size twice as large
33099 as DWARF2_ADDR_SIZE if x is not a scalar int by emitting it as
33100 two halves, one with x and the other with const0_rtx, ordered
33101 depending on endianity.
33103 2021-01-26 Alexandre Oliva <oliva@adacore.com>
33105 * gimplify.c (gimplify_decl_expr): Skip asan marking calls for
33106 temporaries not seen in binding block, and not about to be
33107 added as gimple variables.
33109 2021-01-25 Martin Sebor <msebor@redhat.com>
33112 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Adjust warning text.
33114 2021-01-25 Martin Liska <mliska@suse.cz>
33116 * value-prof.c (get_nth_most_common_value): Use %s instead
33119 2021-01-25 Jakub Jelinek <jakub@redhat.com>
33122 * configure.ac (HAVE_AS_GDWARF_5_DEBUG_FLAG): Only define if
33123 readelf -wi is able to read the emitted .debug_info back.
33124 * configure: Regenerated.
33126 2021-01-25 Martin Liska <mliska@suse.cz>
33128 PR gcov-profile/98739
33129 * common.opt: Add missing sign symbol.
33130 * value-prof.c (get_nth_most_common_value): Restore handling
33131 of PROFILE_REPRODUCIBILITY_PARALLEL_RUNS and
33132 PROFILE_REPRODUCIBILITY_MULTITHREADED.
33134 2021-01-25 Richard Biener <rguenther@suse.de>
33136 PR middle-end/98807
33137 * tree.c (vector_element_bits): Always use precision of
33138 the element type for boolean vectors.
33140 2021-01-25 Sebastian Huber <sebastian.huber@embedded-brains.de>
33142 * config/rtems.h (STARTFILE_SPEC): Remove qnolinkcmds.
33143 (ENDFILE_SPEC): Evaluate qnolinkcmds.
33145 2021-01-25 Sebastian Huber <sebastian.huber@embedded-brains.de>
33147 * config/rtems.h (STARTFILE_SPEC): Remove nostdlib and
33148 nostartfiles handling since this is already done by
33149 LINK_COMMAND_SPEC. Evaluate qnolinkcmds.
33150 (ENDFILE_SPEC): Remove nostdlib and nostartfiles handling since this
33151 is already done by LINK_COMMAND_SPEC.
33152 (LIB_SPECS): Remove nostdlib and nodefaultlibs handling since
33153 this is already done by LINK_COMMAND_SPEC. Remove qnolinkcmds
33156 2021-01-25 Jakub Jelinek <jakub@redhat.com>
33159 * fold-const-call.c (host_size_t_cst_p): Renamed to ...
33160 (size_t_cst_p): ... this. Check and store unsigned HOST_WIDE_INT
33161 value rather than host size_t.
33162 (fold_const_call): Change type of s2 from size_t to
33163 unsigned HOST_WIDE_INT. Use size_t_cst_p instead of
33164 host_size_t_cst_p. For strncmp calls, pass MIN (s2, SIZE_MAX)
33165 instead of s2 as last argument.
33167 2021-01-25 Tamar Christina <tamar.christina@arm.com>
33169 * config/arm/iterators.md (rotsplit1, rotsplit2, conj_op, fcmac1,
33170 VCMLA_OP, VCMUL_OP): New.
33171 * config/arm/mve.md (mve_vcmlaq<mve_rot><mode>): Support vec_dup 0.
33172 * config/arm/neon.md (cmul<conj_op><mode>3): New.
33173 * config/arm/unspecs.md (UNSPEC_VCMLA_CONJ, UNSPEC_VCMLA180_CONJ,
33174 UNSPEC_VCMUL_CONJ): New.
33175 * config/arm/vec-common.md (cmul<conj_op><mode>3, arm_vcmla<rot><mode>,
33176 cml<fcmac1><conj_op><mode>4): New.
33178 2021-01-23 Jakub Jelinek <jakub@redhat.com>
33181 * config/rs6000/mmintrin.h (__m64): Add __may_alias__ attribute.
33183 2021-01-22 Jonathan Wright <jonathan.wright@arm.com>
33185 * config/aarch64/aarch64-simd-builtins.def: Add mla builtin
33187 * config/aarch64/arm_neon.h (vmla_s8): Use RTL builtin rather
33189 (vmla_s16): Likewise.
33190 (vmla_s32): Likewise.
33191 (vmla_u8): Likewise.
33192 (vmla_u16): Likewise.
33193 (vmla_u32): Likewise.
33194 (vmlaq_s8): Likewise.
33195 (vmlaq_s16): Likewise.
33196 (vmlaq_s32): Likewise.
33197 (vmlaq_u8): Likewise.
33198 (vmlaq_u16): Likewise.
33199 (vmlaq_u32): Likewise.
33201 2021-01-22 David Malcolm <dmalcolm@redhat.com>
33203 * doc/invoke.texi (GCC_EXTRA_DIAGNOSTIC_OUTPUT): Add @findex
33206 2021-01-22 Jakub Jelinek <jakub@redhat.com>
33209 * dwarf2out.c (output_file_names): For -gdwarf-5, if there are no
33210 filenames to emit, still emit the required 0 index directory and
33211 filename entries that match DW_AT_comp_dir and DW_AT_name of the
33214 2021-01-22 Marek Polacek <polacek@redhat.com>
33217 * doc/invoke.texi: Update C++ ABI Version 15 description.
33219 2021-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
33221 PR tree-optimization/98766
33222 * tree-ssa-math-opts.c (convert_mult_to_fma): Use maybe_le when
33223 comparing against type size with param_avoid_fma_max_bits.
33225 2021-01-22 Richard Biener <rguenther@suse.de>
33227 PR middle-end/98793
33228 * tree.c (vector_element_bits): Key single-bit bool vector on
33229 integer mode rather than not vector mode.
33231 2021-01-22 Xionghu Luo <luoxhu@linux.ibm.com>
33234 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
33235 Generate ARRAY_REF(VIEW_CONVERT_EXPR) for P8 and later
33237 * config/rs6000/rs6000.c (rs6000_expand_vector_set_var): Update
33238 to call different path for P8 and P9.
33239 (rs6000_expand_vector_set_var_p9): New function.
33240 (rs6000_expand_vector_set_var_p8): New function.
33242 2021-01-22 Xionghu Luo <luoxhu@linux.ibm.com>
33246 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
33247 Ajdust variable index vec_insert from address dereference to
33248 ARRAY_REF(VIEW_CONVERT_EXPR) tree expression.
33249 * config/rs6000/rs6000-protos.h (rs6000_expand_vector_set_var):
33251 * config/rs6000/rs6000.c (rs6000_expand_vector_set_var): New function.
33253 2021-01-22 Martin Liska <mliska@suse.cz>
33255 PR gcov-profile/98739
33256 * profile.c (compute_value_histograms): Drop time profile for
33257 -fprofile-reproducible=multithreaded.
33259 2021-01-22 Nathan Sidwell <nathan@acm.org>
33261 * gcc.c (process_command): Don't check OPT_SPECIAL_input_file
33264 2021-01-22 Richard Biener <rguenther@suse.de>
33266 PR middle-end/98773
33267 * tree-data-ref.c (initalize_matrix_A): Revert previous
33268 change, retaining failing on HOST_WIDE_INT_MIN CHREC_RIGHT.
33270 2021-01-22 Jakub Jelinek <jakub@redhat.com>
33272 PR tree-optimization/90248
33273 * match.pd (X cmp 0.0 ? 1.0 : -1.0 -> copysign(1, +-X),
33274 X cmp 0.0 ? -1.0 : +1.0 -> copysign(1, -+X)): Remove
33276 (X * (X cmp 0.0 ? 1.0 : -1.0) -> +-abs(X),
33277 X * (X cmp 0.0 ? -1.0 : 1.0) -> +-abs(X)): New simplifications.
33279 2021-01-22 Jakub Jelinek <jakub@redhat.com>
33281 PR tree-optimization/98255
33282 * tree-dfa.c (get_ref_base_and_extent): For ARRAY_REFs, sign
33283 extend index - low_bound from sizetype's precision rather than index
33285 (get_addr_base_and_unit_offset_1): Likewise.
33286 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Likewise.
33287 * gimple-fold.c (fold_const_aggregate_ref_1): Likewise.
33289 2021-01-22 Richard Biener <rguenther@suse.de>
33291 PR tree-optimization/98786
33292 * tree-ssa-phiopt.c (factor_out_conditional_conversion): Avoid
33293 adding new uses of abnormals. Verify we deal with a conditional
33296 2021-01-22 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
33299 * optc-save-gen.awk: Add arm_fp16_format to checked_options.
33301 2021-01-22 liuhongt <hongtao.liu@intel.com>
33305 * config/i386/sse.md (VI_128_256): New mode iterator.
33306 (*avx_cmp<mode>3_1, *avx_cmp<mode>3_2, *avx_cmp<mode>3_3,
33307 *avx_cmp<mode>3_4, *avx2_eq<mode>3, *avx2_pcmp<mode>3_1,
33308 *avx2_pcmp<mode>3_2, *avx2_gt<mode>3): New
33309 define_insn_and_split to lower avx512 vector comparison to avx
33310 version when dest is vector.
33311 (*<avx512>_cmp<mode>3,*<avx512>_cmp<mode>3,*<avx512>_ucmp<mode>3):
33312 define_insn_and_split for negating the comparison result.
33313 * config/i386/predicates.md (float_vector_all_ones_operand):
33315 * config/i386/i386-expand.c (ix86_expand_sse_movcc): Use
33316 general NOT operator without UNSPEC_MASKOP.
33318 2021-01-21 Vladimir N. Makarov <vmakarov@redhat.com>
33320 PR rtl-optimization/98777
33321 * lra-int.h (lra_pmode_pseudo): New extern.
33322 * lra.c (lra_pmode_pseudo): New global.
33324 * lra-eliminations.c (eliminate_regs_in_insn): Use it.
33326 2021-01-21 Ilya Leoshkevich <iii@linux.ibm.com>
33328 * fwprop.c (fwprop_propagation::classify_result): Allow
33329 (subreg (mem)) simplifications.
33331 2021-01-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
33333 * config/aarch64/aarch64-simd.md (aarch64_sqdml<SBINQOPS:as>l<mode>):
33335 (aarch64_sqdmlal<mode>): ... This...
33336 (aarch64_sqdmlsl<mode>): ... And this.
33337 (aarch64_sqdml<SBINQOPS:as>l_lane<mode>): Split into...
33338 (aarch64_sqdmlal_lane<mode>): ... This...
33339 (aarch64_sqdmlsl_lane<mode>): ... And this.
33340 (aarch64_sqdml<SBINQOPS:as>l_laneq<mode>): Split into...
33341 (aarch64_sqdmlsl_laneq<mode>): ... This...
33342 (aarch64_sqdmlal_laneq<mode>): ... And this.
33343 (aarch64_sqdml<SBINQOPS:as>l_n<mode>): Split into...
33344 (aarch64_sqdmlsl_n<mode>): ... This...
33345 (aarch64_sqdmlal_n<mode>): ... And this.
33346 (aarch64_sqdml<SBINQOPS:as>l2<mode>_internal): Split into...
33347 (aarch64_sqdmlal2<mode>_internal): ... This...
33348 (aarch64_sqdmlsl2<mode>_internal): ... And this.
33350 2021-01-21 Christophe Lyon <christophe.lyon@linaro.org>
33352 * config/arm/arm_mve.h (__arm_vcmpneq_s8): Fix return type.
33354 2021-01-21 Andrea Corallo <andrea.corallo@arm.com>
33357 * doc/sourcebuild.texi (arm_thumb2_no_arm_v8_1_lob): Document.
33359 2021-01-21 liuhongt <hongtao.liu@intel.com>
33361 PR rtl-optimization/98694
33362 * regcprop.c (copy_value): If SRC had been assigned a mode
33363 narrower than the copy, we can't link DEST into the chain even
33364 they have same hard_regno_nregs(i.e. HImode/SImode in i386
33367 2021-01-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
33369 * config/aarch64/aarch64-simd.md (aarch64_get_lane<mode>):
33370 Convert to define_insn_and_split. Split into simple move when moving
33373 2021-01-20 Segher Boessenkool <segher@kernel.crashing.org>
33375 * config/rs6000/rs6000.c (rs6000_emit_le_vsx_store): Change assert.
33376 Adjust comment. Simplify code.
33378 2021-01-20 Jakub Jelinek <jakub@redhat.com>
33381 * dwarf2out.c (reset_indirect_string): Also reset indirect strings
33382 with DW_FORM_line_strp form.
33383 (prune_unused_types_update_strings): Don't add into debug_str_hash
33384 indirect strings with DW_FORM_line_strp form.
33385 (adjust_name_comp_dir): New function.
33386 (dwarf2out_finish): Call it on CU DIEs after resetting
33387 debug_line_str_hash.
33389 2021-01-20 Vladimir N. Makarov <vmakarov@redhat.com>
33391 PR rtl-optimization/98722
33392 * lra-eliminations.c (eliminate_regs_in_insn): Check that target
33393 has no 3-op add insn to transform insns containing two pluses.
33395 2021-01-20 Richard Biener <rguenther@suse.de>
33397 * hwint.h (add_hwi): New function.
33398 (mul_hwi): Likewise.
33399 * tree-data-ref.c (initialize_matrix_A): Properly translate
33400 tree constants and avoid HOST_WIDE_INT_MIN.
33401 (lambda_matrix_row_add): Avoid undefined integer overflow
33402 and return true on such overflow.
33403 (lambda_matrix_right_hermite): Handle overflow from
33404 lambda_matrix_row_add gracefully. Simplify previous fix.
33405 (analyze_subscript_affine_affine): Likewise.
33407 2021-01-20 Eugene Rozenfeld <erozen@microsoft.com>
33409 PR tree-optimization/96674
33410 * match.pd: New patterns: x < y || y == XXX_MIN --> x <= y - 1
33411 x >= y && y != XXX_MIN --> x > y - 1
33413 2021-01-20 Richard Sandiford <richard.sandiford@arm.com>
33415 PR tree-optimization/98535
33416 * tree-vect-slp.c (duplicate_and_interleave): Use quick_grow_cleared.
33417 If the high and low permutes are the same, remove the high permutes
33418 from the working set and only continue with the low ones.
33420 2021-01-20 Jakub Jelinek <jakub@redhat.com>
33422 PR tree-optimization/98721
33423 * builtins.c (access_ref::inform_access): Don't assume
33424 SSA_NAME_IDENTIFIER must be non-NULL. Print messages about
33425 object whenever allocfn is NULL, rather than only when DECL_P
33426 is true. Use %qE instead of %qD for that. Formatting fixes.
33428 2021-01-20 Richard Biener <rguenther@suse.de>
33430 PR tree-optimization/98758
33431 * tree-data-ref.c (int_divides_p): Use lambda_int arguments.
33432 (lambda_matrix_right_hermite): Avoid undefinedness with
33433 signed integer abs and multiplication.
33434 (analyze_subscript_affine_affine): Use lambda_int.
33436 2021-01-20 David Malcolm <dmalcolm@redhat.com>
33439 * dwarf2out.c (output_line_info): Rename static variable
33440 "generation", moving it out of the function to...
33441 (output_line_info_generation): New.
33442 (init_sections_and_labels): Likewise, renaming the variable to...
33443 (init_sections_and_labels_generation): New.
33444 (dwarf2out_c_finalize): Reset the new variables.
33446 2021-01-19 Martin Sebor <msebor@redhat.com>
33448 PR middle-end/98664
33449 * tree-ssa-live.c (remove_unused_scope_block_p): Keep scopes for
33450 all functions, even if they're not declared artificial or inline.
33451 * tree.c (tree_inlined_location): Use macro expansion location
33452 only if scope traversal fails to expose one.
33454 2021-01-19 Richard Sandiford <richard.sandiford@arm.com>
33456 PR rtl-optimization/92294
33457 * alias.c (compare_base_symbol_refs): Take an extra parameter
33458 and add the distance between two symbols to it. Enshrine in
33459 comments that -1 means "either 0 or 1, but we can't tell
33460 which at compile time".
33461 (memrefs_conflict_p): Update call accordingly.
33462 (rtx_equal_for_memref_p): Likewise. Take the distance between symbols
33465 2021-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
33467 * config/aarch64/aarch64-simd-builtins.def (sqshl, uqshl,
33468 sqrshl, uqrshl, sqadd, uqadd, sqsub, uqsub, suqadd, usqadd, sqmovn,
33469 uqmovn, sqxtn2, uqxtn2, sqabs, sqneg, sqdmlal, sqdmlsl, sqdmlal_lane,
33470 sqdmlsl_lane, sqdmlal_laneq, sqdmlsl_laneq, sqdmlal_n, sqdmlsl_n,
33471 sqdmlal2, sqdmlsl2, sqdmlal2_lane, sqdmlsl2_lane, sqdmlal2_laneq,
33472 sqdmlsl2_laneq, sqdmlal2_n, sqdmlsl2_n, sqdmull, sqdmull_lane,
33473 sqdmull_laneq, sqdmull_n, sqdmull2, sqdmull2_lane, sqdmull2_laneq,
33474 sqdmull2_n, sqdmulh, sqrdmulh, sqdmulh_lane, sqdmulh_laneq,
33475 sqrdmulh_lane, sqrdmulh_laneq, sqshrun_n, sqrshrun_n, sqshrn_n,
33476 uqshrn_n, sqrshrn_n, uqrshrn_n, sqshlu_n, sqshl_n, uqshl_n, sqrdmlah,
33477 sqrdmlsh, sqrdmlah_lane, sqrdmlsh_lane, sqrdmlah_laneq, sqrdmlsh_laneq,
33478 sqmovun): Use NONE flags.
33480 2021-01-19 Richard Biener <rguenther@suse.de>
33483 * ipa-modref.c (analyze_stmt): Only record a summary for a
33486 2021-01-19 Richard Biener <rguenther@suse.de>
33488 PR middle-end/98638
33489 * tree-ssanames.c (fini_ssanames): Zero SSA_NAME_DEF_STMT.
33491 2021-01-19 Daniel Hellstrom <daniel@gaisler.com>
33493 * config/sparc/rtemself.h (TARGET_OS_CPP_BUILTINS): Add
33494 built-in define __FIX_LEON3FT_TN0018.
33496 2021-01-19 Richard Biener <rguenther@suse.de>
33499 * tree-inline.c (tree_function_versioning): Set input_location
33500 to UNKNOWN_LOCATION throughout the function.
33502 2021-01-19 Tobias Burnus <tobias@codesourcery.com>
33505 * omp-low.c (lower_omp_target): Handle nonpointer is_device_ptr.
33507 2021-01-19 Martin Jambor <mjambor@suse.cz>
33510 * ipa-sra.c (ssa_name_only_returned_p): New parameter fun. Check
33511 whether non-call exceptions allow removal of a statement.
33512 (isra_analyze_call): Pass the appropriate function to
33513 ssa_name_only_returned_p.
33515 2021-01-19 Geng Qi <gengqi@linux.alibaba.com>
33517 * config/riscv/arch-canonicalize (longext_sort): New function for
33518 sorting 'multi-letter'.
33519 * config/riscv/multilib-generator: Adjusting the loop of 'alt' in
33520 'alts'. The 'arch' may not be the first of 'alts'.
33521 (_expand_combination): Add underline for the 'ext' without '*'.
33522 This is because, a single-letter extension can always be treated well
33523 with a '_' prefix, but it cannot be separated out if it is appended
33526 2021-01-18 Vladimir N. Makarov <vmakarov@redhat.com>
33529 * ira.c (ira): Skip abnormal critical edge splitting.
33531 2021-01-18 Jakub Jelinek <jakub@redhat.com>
33533 PR tree-optimization/98727
33534 * tree-ssa-math-opts.c (match_arith_overflow): Fix up computation of
33535 second .MUL_OVERFLOW operand for signed multiplication with overflow
33536 checking if the second operand of multiplication is not constant.
33538 2021-01-18 David Edelsohn <dje.gcc@gmail.com>
33540 * doc/invoke.texi (-gdwarf): TPF defaults to version 2 and AIX
33541 defaults to version 4.
33543 2021-01-18 David Malcolm <dmalcolm@redhat.com>
33545 * attribs.h (fndecl_dealloc_argno): New decl.
33546 * builtins.c (call_dealloc_argno): Split out second half of
33548 (fndecl_dealloc_argno): New.
33549 * doc/extend.texi (Common Function Attributes): Document the
33550 interaction between the analyzer and the malloc attribute.
33551 * doc/invoke.texi (Static Analyzer Options): Likewise.
33553 2021-01-17 David Edelsohn <dje.gcc@gmail.com>
33555 * config/rs6000/aix71.h (SUBTARGET_OVERRIDE_OPTIONS): Override
33556 dwarf_version to 4.
33557 * config/rs6000/aix72.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
33559 2021-01-17 Martin Jambor <mjambor@suse.cz>
33562 * cgraph.c (clone_of_p): Check also former_clone_of as we climb
33565 2021-01-17 Mark Wielaard <mark@klomp.org>
33567 * common.opt (gdwarf-): Init(5).
33568 * doc/invoke.texi (-gdwarf): Document default to 5.
33570 2021-01-16 Kwok Cheung Yeung <kcy@codesourcery.com>
33572 * builtin-types.def
33573 (BT_FN_VOID_OMPFN_PTR_OMPCPYFN_LONG_LONG_BOOL_UINT_PTR_INT): Rename
33575 (BT_FN_VOID_OMPFN_PTR_OMPCPYFN_LONG_LONG_BOOL_UINT_PTR_INT_PTR):
33576 ...this. Add extra argument.
33577 * gimplify.c (omp_default_clause): Ensure that event handle is
33578 firstprivate in a task region.
33579 (gimplify_scan_omp_clauses): Handle OMP_CLAUSE_DETACH.
33580 (gimplify_adjust_omp_clauses): Likewise.
33581 * omp-builtins.def (BUILT_IN_GOMP_TASK): Change function type to
33582 BT_FN_VOID_OMPFN_PTR_OMPCPYFN_LONG_LONG_BOOL_UINT_PTR_INT_PTR.
33583 * omp-expand.c (expand_task_call): Add GOMP_TASK_FLAG_DETACH to flags
33584 if detach clause specified. Add detach argument when generating
33586 * omp-low.c (scan_sharing_clauses): Setup data environment for detach
33588 (finish_taskreg_scan): Move field for variable containing the event
33589 handle to the front of the struct.
33590 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_DETACH. Fix
33592 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
33593 OMP_CLAUSE_DETACH clause.
33594 (convert_local_omp_clauses): Handle OMP_CLAUSE_DETACH clause.
33595 * tree-pretty-print.c (dump_omp_clause): Handle OMP_CLAUSE_DETACH.
33596 * tree.c (omp_clause_num_ops): Add entry for OMP_CLAUSE_DETACH.
33598 (omp_clause_code_name): Add entry for OMP_CLAUSE_DETACH. Fix
33600 (walk_tree_1): Handle OMP_CLAUSE_DETACH.
33602 2021-01-16 Sebastian Huber <sebastian.huber@embedded-brains.de>
33604 * config/nios2/t-rtems: Reset all MULTILIB_* variables. Shorten
33605 multilib directory names. Use MULTILIB_REQUIRED instead of
33606 MULTILIB_EXCEPTIONS. Add -mhw-mul -mhw-mulx -mhw-div
33607 -mcustom-fpu-cfg=fph2 multilib.
33609 2021-01-16 Sebastian Huber <sebastian.huber@embedded-brains.de>
33611 * config/nios2/nios2.c (NIOS2_FPU_CONFIG_NUM): Adjust value.
33612 (nios2_init_fpu_configs): Provide register values for new
33613 -mcustom-fpu-cfg=fph2 option variant.
33614 * doc/invoke.texi (-mcustom-fpu-cfg=fph2): Document new option
33617 2021-01-16 Sebastian Huber <sebastian.huber@embedded-brains.de>
33619 * config/nios2/nios2.c (nios2_custom_check_insns): Remove
33620 custom instruction warnings.
33622 2021-01-16 Jakub Jelinek <jakub@redhat.com>
33624 PR tree-optimization/96669
33625 * match.pd ((CST << x) & 1 -> x == 0): New simplification.
33627 2021-01-16 Jakub Jelinek <jakub@redhat.com>
33629 PR tree-optimization/96271
33630 * passes.def: Pass false argument to first two pass_cd_dce
33631 instances and true to last instance. Add comment that
33632 last instance rewrites no longer addressed locals.
33633 * tree-ssa-dce.c (pass_cd_dce): Add update_address_taken_p member and
33635 (pass_cd_dce::set_pass_param): New method.
33636 (pass_cd_dce::execute): Return TODO_update_address_taken from
33637 last cd_dce instance.
33639 2021-01-15 Carl Love <cel@us.ibm.com>
33641 * config/rs6000/altivec.h (vec_mulh, vec_div, vec_dive, vec_mod):
33643 * config/rs6000/altivec.md (VIlong): Move define to file vsx.md.
33644 * config/rs6000/rs6000-builtin.def (DIVES_V4SI, DIVES_V2DI,
33645 DIVEU_V4SI, DIVEU_V2DI, DIVS_V4SI, DIVS_V2DI, DIVU_V4SI,
33646 DIVU_V2DI, MODS_V2DI, MODS_V4SI, MODU_V2DI, MODU_V4SI,
33647 MULHS_V2DI, MULHS_V4SI, MULHU_V2DI, MULHU_V4SI, MULLD_V2DI):
33648 Add builtin define.
33649 (MULH, DIVE, MOD): Add new BU_P10_OVERLOAD_2 definitions.
33650 * config/rs6000/rs6000-call.c (VSX_BUILTIN_VEC_DIV,
33651 VSX_BUILTIN_VEC_DIVE, P10_BUILTIN_VEC_MOD, P10_BUILTIN_VEC_MULH):
33652 New overloaded definitions.
33653 (builtin_function_type) [P10V_BUILTIN_DIVEU_V4SI,
33654 P10V_BUILTIN_DIVEU_V2DI, P10V_BUILTIN_DIVU_V4SI,
33655 P10V_BUILTIN_DIVU_V2DI, P10V_BUILTIN_MODU_V2DI,
33656 P10V_BUILTIN_MODU_V4SI, P10V_BUILTIN_MULHU_V2DI,
33657 P10V_BUILTIN_MULHU_V4SI]: Add case
33658 statement for builtins.
33659 * config/rs6000/rs6000.md (bits): Add new attribute sizes V4SI, V2DI.
33660 * config/rs6000/vsx.md (VIlong): Moved from config/rs6000/altivec.md.
33661 (UNSPEC_VDIVES, UNSPEC_VDIVEU): New unspec definitions.
33662 (vsx_mul_v2di): Add if TARGET_POWER10 statement.
33663 (vsx_udiv_v2di): Add if TARGET_POWER10 statement.
33664 (dives_<mode>, diveu_<mode>, div<mode>3, uvdiv<mode>3,
33665 mods_<mode>, modu_<mode>, mulhs_<mode>, mulhu_<mode>, mulv2di3):
33666 Add define_insn, mode is VIlong.
33667 * doc/extend.texi (vec_mulh, vec_mul, vec_div, vec_dive, vec_mod):
33668 Add builtin descriptions.
33670 2021-01-15 Eric Botcazou <ebotcazou@adacore.com>
33672 * final.c (final_start_function_1): Reset force_source_line.
33674 2021-01-15 Jakub Jelinek <jakub@redhat.com>
33676 PR tree-optimization/96669
33677 * match.pd (((1 << A) & 1) != 0 -> A == 0,
33678 ((1 << A) & 1) == 0 -> A != 0): Generalize for 1s replaced by
33679 possibly different power of two constants and to right shift too.
33681 2021-01-15 Jakub Jelinek <jakub@redhat.com>
33683 PR tree-optimization/96681
33684 * match.pd ((x < 0) ^ (y < 0) to (x ^ y) < 0): New simplification.
33685 ((x >= 0) ^ (y >= 0) to (x ^ y) < 0): Likewise.
33686 ((x < 0) ^ (y >= 0) to (x ^ y) >= 0): Likewise.
33687 ((x >= 0) ^ (y < 0) to (x ^ y) >= 0): Likewise.
33689 2021-01-15 Alexandre Oliva <oliva@adacore.com>
33691 * opts.c (gen_command_line_string): Exclude -dumpbase-ext.
33693 2021-01-15 Tamar Christina <tamar.christina@arm.com>
33695 * config/aarch64/aarch64-simd.md (cml<fcmac1><conj_op><mode>4,
33696 cmul<conj_op><mode>3): New.
33697 * config/aarch64/iterators.md (UNSPEC_FCMUL,
33698 UNSPEC_FCMUL180, UNSPEC_FCMLA_CONJ, UNSPEC_FCMLA180_CONJ,
33699 UNSPEC_CMLA_CONJ, UNSPEC_CMLA180_CONJ, UNSPEC_CMUL, UNSPEC_CMUL180,
33700 FCMLA_OP, FCMUL_OP, conj_op, rotsplit1, rotsplit2, fcmac1, sve_rot1,
33701 sve_rot2, SVE2_INT_CMLA_OP, SVE2_INT_CMUL_OP, SVE2_INT_CADD_OP): New.
33702 (rot): Add UNSPEC_FCMUL, UNSPEC_FCMUL180.
33703 (rot_op): Renamed to conj_op.
33704 * config/aarch64/aarch64-sve.md (cml<fcmac1><conj_op><mode>4,
33705 cmul<conj_op><mode>3): New.
33706 * config/aarch64/aarch64-sve2.md (cml<fcmac1><conj_op><mode>4,
33707 cmul<conj_op><mode>3): New.
33709 2021-01-15 David Malcolm <dmalcolm@redhat.com>
33713 (selftest::test_print_parseable_fixits_bytes_vs_display_columns):
33714 Escape the tempfile name when constructing the expected output.
33716 2021-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
33718 * config/aarch64/aarch64-simd.md (*aarch64_<su>mlsl_hi<mode>):
33720 (aarch64_<su>mlsl_hi<mode>): ... This.
33721 (aarch64_<su>mlsl_hi<mode>): Define.
33722 (*aarch64_<su>mlsl<mode): Rename to...
33723 (aarch64_<su>mlsl<mode): ... This.
33724 * config/aarch64/aarch64-simd-builtins.def (smlsl, umlsl,
33725 smlsl_hi, umlsl_hi): Define builtins.
33726 * config/aarch64/arm_neon.h (vmlsl_high_s8, vmlsl_high_s16,
33727 vmlsl_high_s32, vmlsl_high_u8, vmlsl_high_u16, vmlsl_high_u32,
33728 vmlsl_s8, vmlsl_s16, vmlsl_s32, vmlsl_u8,
33729 vmlsl_u16, vmlsl_u32): Reimplement with builtins.
33731 2021-01-15 Uroš Bizjak <ubizjak@gmail.com>
33733 * config/i386/i386-c.c (ix86_target_macros):
33734 Use cpp_define_formatted for __SIZEOF_FLOAT80__ definition.
33736 2021-01-15 Richard Sandiford <richard.sandiford@arm.com>
33739 * config.gcc (aarch64*-*-*): Add aarch64-cc-fusion.o to extra_objs.
33740 * Makefile.in (RTL_SSA_H): New variable.
33741 * config/aarch64/t-aarch64 (aarch64-cc-fusion.o): New rule.
33742 * config/aarch64/aarch64-protos.h (make_pass_cc_fusion): Declare.
33743 * config/aarch64/aarch64-passes.def: Add pass_cc_fusion after
33745 * config/aarch64/aarch64-cc-fusion.cc: New file.
33747 2021-01-15 Richard Sandiford <richard.sandiford@arm.com>
33749 * recog.h (insn_change_watermark::~insn_change_watermark): Avoid
33750 calling cancel_changes for changes that no longer exist.
33752 2021-01-15 Richard Sandiford <richard.sandiford@arm.com>
33754 * rtl-ssa/functions.h (function_info::ref_defs): Rename to...
33755 (function_info::reg_defs): ...this.
33756 * rtl-ssa/member-fns.inl (function_info::ref_defs): Rename to...
33757 (function_info::reg_defs): ...this.
33759 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
33762 * config/arm/arm_neon.h (vceqz_p64, vceqq_p64, vceqzq_p64): New.
33764 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
33767 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
33770 * config/arm/arm_neon.h (vceqz_p64, vceqq_p64, vceqzq_p64): New.
33772 2021-01-15 Richard Biener <rguenther@suse.de>
33774 PR tree-optimization/96376
33775 * tree-vect-stmts.c (get_load_store_type): Disregard alignment
33776 for VMAT_INVARIANT.
33778 2021-01-15 Martin Liska <mliska@suse.cz>
33780 * doc/install.texi: Document that some tests need pytest module.
33781 * doc/sourcebuild.texi: Likewise.
33783 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
33786 * config/arm/arm_neon.h (vceqz_p64, vceqq_p64, vceqzq_p64): New.
33788 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
33790 * config/arm/mve.md (mve_vshrq_n_s<mode>_imm): New entry.
33791 (mve_vshrq_n_u<mode>_imm): Likewise.
33792 * config/arm/neon.md (vashr<mode>3, vlshr<mode>3): Move to ...
33793 * config/arm/vec-common.md: ... here.
33795 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
33797 * config/arm/mve.md (mve_vshlq_<supf><mode>): Move to
33799 * config/arm/neon.md (vashl<mode>3): Delete.
33800 * config/arm/vec-common.md (mve_vshlq_<supf><mode>): New.
33801 (vasl<mode>3): New expander.
33803 2021-01-15 Richard Biener <rguenther@suse.de>
33805 PR tree-optimization/98685
33806 * tree-vect-slp.c (vect_schedule_slp_node): Refactor handling
33807 of vector extern defs.
33809 2021-01-14 David Malcolm <dmalcolm@redhat.com>
33812 * diagnostic.c (diagnostic_kind_text): Break out this array
33814 (diagnostic_build_prefix): ...here.
33815 (fancy_abort): Detect when diagnostic_initialize has not yet been
33816 called and fall back to a minimal implementation of printing the
33817 ICE, rather than segfaulting in internal_error.
33819 2021-01-14 David Malcolm <dmalcolm@redhat.com>
33821 * diagnostic.c (diagnostic_initialize): Eliminate
33822 parseable_fixits_p in favor of initializing extra_output_kind from
33823 GCC_EXTRA_DIAGNOSTIC_OUTPUT.
33824 (convert_column_unit): New function, split out from...
33825 (diagnostic_converted_column): ...this.
33826 (print_parseable_fixits): Add "column_unit" and "tabstop" params.
33827 Use them to call convert_column_unit on the column values.
33828 (diagnostic_report_diagnostic): Eliminate conditional on
33829 parseable_fixits_p in favor of a switch statement on
33830 extra_output_kind, passing the appropriate values to the new
33831 params of print_parseable_fixits.
33832 (selftest::test_print_parseable_fixits_none): Update for new
33833 params of print_parseable_fixits.
33834 (selftest::test_print_parseable_fixits_insert): Likewise.
33835 (selftest::test_print_parseable_fixits_remove): Likewise.
33836 (selftest::test_print_parseable_fixits_replace): Likewise.
33837 (selftest::test_print_parseable_fixits_bytes_vs_display_columns):
33839 (selftest::diagnostic_c_tests): Call it.
33840 * diagnostic.h (enum diagnostics_extra_output_kind): New.
33841 (diagnostic_context::parseable_fixits_p): Delete field in favor
33843 (diagnostic_context::extra_output_kind): ...this new field.
33844 * doc/invoke.texi (Environment Variables): Add
33845 GCC_EXTRA_DIAGNOSTIC_OUTPUT.
33846 * opts.c (common_handle_option): Update handling of
33847 OPT_fdiagnostics_parseable_fixits for change to diagnostic_context
33850 2021-01-14 Tamar Christina <tamar.christina@arm.com>
33852 * tree-vect-slp-patterns.c (class complex_operations_pattern,
33853 complex_operations_pattern::matches,
33854 complex_operations_pattern::recognize,
33855 complex_operations_pattern::build): New.
33856 (slp_patterns): Use it.
33858 2021-01-14 Tamar Christina <tamar.christina@arm.com>
33860 * internal-fn.def (COMPLEX_FMS, COMPLEX_FMS_CONJ): New.
33861 * optabs.def (cmls_optab, cmls_conj_optab): New.
33862 * doc/md.texi: Document them.
33863 * tree-vect-slp-patterns.c (class complex_fms_pattern,
33864 complex_fms_pattern::matches, complex_fms_pattern::recognize,
33865 complex_fms_pattern::build): New.
33867 2021-01-14 Tamar Christina <tamar.christina@arm.com>
33869 * internal-fn.def (COMPLEX_FMA, COMPLEX_FMA_CONJ): New.
33870 * optabs.def (cmla_optab, cmla_conj_optab): New.
33871 * doc/md.texi: Document them.
33872 * tree-vect-slp-patterns.c (vect_match_call_p,
33873 class complex_fma_pattern, vect_slp_reset_pattern,
33874 complex_fma_pattern::matches, complex_fma_pattern::recognize,
33875 complex_fma_pattern::build): New.
33877 2021-01-14 Tamar Christina <tamar.christina@arm.com>
33879 * internal-fn.def (COMPLEX_MUL, COMPLEX_MUL_CONJ): New.
33880 * optabs.def (cmul_optab, cmul_conj_optab): New.
33881 * doc/md.texi: Document them.
33882 * tree-vect-slp-patterns.c (vect_match_call_complex_mla,
33883 vect_normalize_conj_loc, is_eq_or_top, vect_validate_multiplication,
33884 vect_build_combine_node, class complex_mul_pattern,
33885 complex_mul_pattern::matches, complex_mul_pattern::recognize,
33886 complex_mul_pattern::build): New.
33888 2021-01-14 Tamar Christina <tamar.christina@arm.com>
33890 * tree-vect-slp.c (optimize_load_redistribution_1): New.
33891 (optimize_load_redistribution, vect_is_slp_load_node): New.
33892 (vect_match_slp_patterns): Use it.
33894 2021-01-14 Tamar Christina <tamar.christina@arm.com>
33896 * tree-vect-slp-patterns.c (complex_add_pattern::build):
33899 2021-01-14 Thomas Schwinge <thomas@codesourcery.com>
33901 * config/gcn/mkoffload.c (main): Create an offload image only in
33902 64-bit configurations.
33904 2021-01-14 H.J. Lu <hjl.tools@gmail.com>
33907 * config/i386/i386-options.c (ix86_option_override_internal):
33908 Issue an error for -fcf-protection with CF_BRANCH when compiling
33909 for 32-bit non-TARGET_CMOV targets.
33911 2021-01-14 Uroš Bizjak <ubizjak@gmail.com>
33914 * config/i386/i386-options.c (ix86_valid_target_attribute_inner_p):
33915 Remove declaration and initialization of shadow variable "ret".
33916 (ix86_option_override_internal): Remove delcaration of
33917 shadow variable "i". Redeclare shadowed variable to unsigned.
33918 * common/config/i386/i386-common.c (pta_size): Redeclare to unsigned.
33919 * config/i386/i386-builtins.c (get_builtin_code_for_version):
33920 Update for redeclaration.
33921 * config/i386/i386.h (pta_size): Ditto.
33923 2021-01-14 Richard Biener <rguenther@suse.de>
33925 PR tree-optimization/98674
33926 * tree-data-ref.c (base_supports_access_fn_components_p): New.
33927 (initialize_data_dependence_relation): For two bases without
33928 possible access fns resort to type size equality when determining
33929 shape compatibility.
33931 2021-01-14 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
33934 * config/arm/arm_neon.h: Replace calls to __builtin_vcge* by
33935 <=, >= operators in vcle and vcge intrinsics respectively.
33936 * config/arm/arm_neon_builtins.def: Remove entry for
33939 2021-01-14 Uroš Bizjak <ubizjak@gmail.com>
33942 * config/i386/i386-options.c (ix86_function_specific_save):
33943 Remove redundant assignment to opts->x_ix86_branch_cost.
33944 * config/i386/i386.c (ix86_prefetch_sse):
33945 Rename from x86_prefetch_sse. Update all uses.
33946 * config/i386/i386.h: Update for rename.
33947 * config/i386/i386-options.h: Ditto.
33949 2021-01-14 Jakub Jelinek <jakub@redhat.com>
33952 * config/i386/sse.md (*sse4_1_zero_extendv8qiv8hi2_3,
33953 *sse4_1_zero_extendv4hiv4si2_3, *sse4_1_zero_extendv2siv2di2_3):
33954 Use Bm instead of m for non-avx. Add isa attribute.
33956 2021-01-14 Jakub Jelinek <jakub@redhat.com>
33958 PR tree-optimization/96688
33959 * match.pd (~(X >> Y) -> ~X >> Y): New simplification if
33960 ~X can be simplified.
33962 2021-01-14 Richard Sandiford <richard.sandiford@arm.com>
33964 * tree-vect-stmts.c (vect_model_load_cost): Account for unused
33965 IFN_LOAD_LANES results.
33967 2021-01-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
33969 * config/aarch64/aarch64-simd.md (aarch64_<su>xtl<mode>):
33971 (aarch64_xtn<mode>): Likewise.
33972 * config/aarch64/aarch64-simd-builtins.def (sxtl, uxtl, xtn):
33975 * config/aarch64/arm_neon.h (vmovl_s8): Reimplement using
33977 (vmovl_s16): Likewise.
33978 (vmovl_s32): Likewise.
33979 (vmovl_u8): Likewise.
33980 (vmovl_u16): Likewise.
33981 (vmovl_u32): Likewise.
33982 (vmovn_s16): Likewise.
33983 (vmovn_s32): Likewise.
33984 (vmovn_s64): Likewise.
33985 (vmovn_u16): Likewise.
33986 (vmovn_u32): Likewise.
33987 (vmovn_u64): Likewise.
33989 2021-01-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
33991 * config/aarch64/aarch64-simd.md (aarch64_<su>qxtn2<mode>_le):
33993 (aarch64_<su>qxtn2<mode>_be): Likewise.
33994 (aarch64_<su>qxtn2<mode>): Likewise.
33995 * config/aarch64/aarch64-simd-builtins.def (sqxtn2, uqxtn2):
33997 * config/aarch64/iterators.md (SAT_TRUNC): Define code_iterator.
33998 (su): Handle ss_truncate and us_truncate.
33999 * config/aarch64/arm_neon.h (vqmovn_high_s16): Reimplement using
34001 (vqmovn_high_s32): Likewise.
34002 (vqmovn_high_s64): Likewise.
34003 (vqmovn_high_u16): Likewise.
34004 (vqmovn_high_u32): Likewise.
34005 (vqmovn_high_u64): Likewise.
34007 2021-01-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
34009 * config/aarch64/aarch64-simd.md (aarch64_xtn2<mode>_le):
34011 (aarch64_xtn2<mode>_be): Likewise.
34012 (aarch64_xtn2<mode>): Likewise.
34013 * config/aarch64/aarch64-simd-builtins.def (xtn2): Define
34015 * config/aarch64/arm_neon.h (vmovn_high_s16): Reimplement using
34017 (vmovn_high_s32): Likewise.
34018 (vmovn_high_s64): Likewise.
34019 (vmovn_high_u16): Likewise.
34020 (vmovn_high_u32): Likewise.
34021 (vmovn_high_u64): Likewise.
34023 2021-01-13 Stafford Horne <shorne@gmail.com>
34025 * config/or1k/or1k.h (ASM_PREFERRED_EH_DATA_FORMAT): New macro.
34027 2021-01-13 Stafford Horne <shorne@gmail.com>
34029 * config/or1k/linux.h (TARGET_ASM_FILE_END): Define macro.
34031 2021-01-13 Stafford Horne <shorne@gmail.com>
34033 * config/or1k/or1k.h (TARGET_CPU_CPP_BUILTINS): Add builtin
34034 define for __or1k_hard_float__.
34036 2021-01-13 Stafford Horne <shorne@gmail.com>
34038 * config/or1k/or1k.h (NO_PROFILE_COUNTERS): Define as 1.
34039 (PROFILE_HOOK): Define to call _mcount.
34040 (FUNCTION_PROFILER): Change from abort to no-op.
34042 2021-01-13 Jakub Jelinek <jakub@redhat.com>
34044 PR tree-optimization/96691
34045 * match.pd ((~X | C) ^ D -> (X | C) ^ (~D ^ C),
34046 (~X & C) ^ D -> (X & C) ^ (D ^ C)): New simplifications if
34047 (~D ^ C) or (D ^ C) can be simplified.
34049 2021-01-13 Richard Biener <rguenther@suse.de>
34051 PR tree-optimization/92645
34052 * match.pd (BIT_FIELD_REF to conversion): Delay canonicalization
34053 until after vector lowering.
34055 2021-01-13 Richard Sandiford <richard.sandiford@arm.com>
34057 * config/aarch64/aarch64-sve.md (fnma<mode>4): Extend from SVE_FULL_I
34059 (@aarch64_pred_fnma<mode>, cond_fnma<mode>, *cond_fnma<mode>_2)
34060 (*cond_fnma<mode>_4, *cond_fnma<mode>_any): Likewise.
34062 2021-01-13 Richard Sandiford <richard.sandiford@arm.com>
34064 * config/aarch64/aarch64-sve.md (fma<mode>4): Extend from SVE_FULL_I
34066 (@aarch64_pred_fma<mode>, cond_fma<mode>, *cond_fma<mode>_2)
34067 (*cond_fma<mode>_4, *cond_fma<mode>_any): Likewise.
34069 2021-01-13 Richard Biener <rguenther@suse.de>
34071 PR tree-optimization/92645
34072 * tree-vect-slp.c (vect_build_slp_tree_1): Relax supported
34073 BIT_FIELD_REF argument.
34074 (vect_build_slp_tree_2): Record the desired vector type
34075 on the external vector def.
34076 (vectorizable_slp_permutation): Handle required punning
34077 of existing vector defs.
34079 2021-01-13 Richard Sandiford <richard.sandiford@arm.com>
34081 * rtl-ssa/accesses.h (def_lookup): Fix order of comparison results.
34083 2021-01-13 Richard Sandiford <richard.sandiford@arm.com>
34085 * config/sh/sh.md (movsf_ie): Remove operands[2] test.
34087 2021-01-13 Samuel Thibault <samuel.thibault@ens-lyon.org>
34089 * config.gcc [$target == *-*-gnu*]: Enable
34090 'default_gnu_indirect_function'.
34092 2021-01-13 Jakub Jelinek <jakub@redhat.com>
34095 * optabs.c (expand_vec_perm_const): Don't force v0 and v1 into
34096 registers before calling targetm.vectorize.vec_perm_const, only after
34098 * config/i386/i386-expand.c (ix86_vectorize_vec_perm_const): Handle
34099 two argument permutation when one operand is zero vector and only
34100 after that force operands into registers.
34101 * config/i386/sse.md (*avx2_zero_extendv16qiv16hi2_1): New
34102 define_insn_and_split pattern.
34103 (*avx512bw_zero_extendv32qiv32hi2_1): Likewise.
34104 (*avx512f_zero_extendv16hiv16si2_1): Likewise.
34105 (*avx2_zero_extendv8hiv8si2_1): Likewise.
34106 (*avx512f_zero_extendv8siv8di2_1): Likewise.
34107 (*avx2_zero_extendv4siv4di2_1): Likewise.
34108 * config/mips/mips.c (mips_vectorize_vec_perm_const): Force operands
34110 * config/arm/arm.c (arm_vectorize_vec_perm_const): Likewise.
34111 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Likewise.
34112 * config/ia64/ia64.c (ia64_vectorize_vec_perm_const): Likewise.
34113 * config/aarch64/aarch64.c (aarch64_vectorize_vec_perm_const): Likewise.
34114 * config/rs6000/rs6000.c (rs6000_vectorize_vec_perm_const): Likewise.
34115 * config/gcn/gcn.c (gcn_vectorize_vec_perm_const): Likewise. Use std::swap.
34117 2021-01-13 Martin Liska <mliska@suse.cz>
34119 PR tree-optimization/98455
34120 * gimple-if-to-switch.cc (condition_info::record_phi_mapping):
34121 Record also virtual PHIs.
34122 (pass_if_to_switch::execute): Return TODO_cleanup_cfg only
34125 2021-01-13 Jonathan Wakely <jwakely@redhat.com>
34127 * doc/invoke.texi (C++ Modules): Fix typos.
34129 2021-01-13 Richard Biener <rguenther@suse.de>
34131 PR tree-optimization/98640
34132 * tree-ssa-sccvn.c (visit_nary_op): Do not try to
34133 handle plus or minus from a truncated operand to be
34136 2021-01-13 Jakub Jelinek <jakub@redhat.com>
34139 * config/i386/i386.md (*btr<mode>_1, *btr<mode>_2): New
34140 define_insn_and_split patterns.
34141 (splitter after *btr<mode>_2): New splitter.
34143 2021-01-13 Martin Liska <mliska@suse.cz>
34146 * cgraphunit.c (analyze_functions): Remove dead code.
34148 2021-01-13 Qian Jianhua <qianjh@cn.fujitsu.com>
34150 * config/aarch64/aarch64-cost-tables.h (a64fx_extra_costs): New.
34151 * config/aarch64/aarch64.c (a64fx_addrcost_table): New.
34152 (a64fx_regmove_cost, a64fx_vector_cost): New.
34153 (a64fx_tunings): Use the new added cost tables.
34155 2021-01-13 Jakub Jelinek <jakub@redhat.com>
34158 * config/i386/predicates.md (pmovzx_parallel): New predicate.
34159 * config/i386/sse.md (*sse4_1_zero_extendv8qiv8hi2_3): New
34160 define_insn_and_split pattern.
34161 (*sse4_1_zero_extendv4hiv4si2_3): Likewise.
34162 (*sse4_1_zero_extendv2siv2di2_3): Likewise.
34164 2021-01-13 Julian Brown <julian@codesourcery.com>
34166 * config/gcn/gcn.c (gcn_conditional_register_usage): Remove dead code
34167 to fix v0 register.
34169 2021-01-13 Julian Brown <julian@codesourcery.com>
34171 * config/gcn/gcn.c (gcn_md_reorg): Fix case where EXEC reg is live
34174 2021-01-13 Julian Brown <julian@codesourcery.com>
34176 * config/gcn/gcn-valu.md (recip<mode>2<exec>, recip<mode>2): Use unspec
34177 for reciprocal-approximation instructions.
34178 (div<mode>3): Use fused multiply-accumulate operations for reciprocal
34179 refinement and division result.
34180 * config/gcn/gcn.md (UNSPEC_RCP): New unspec constant.
34182 2021-01-13 Julian Brown <julian@codesourcery.com>
34184 * config/gcn/gcn-valu.md (subdf): Rename to...
34187 2021-01-12 Martin Liska <mliska@suse.cz>
34189 * gcov.c (source_info::debug): Fix printf format for 32-bit hosts.
34191 2021-01-12 Andrea Corallo <andrea.corallo@arm.com>
34193 * function-abi.h: Fix typo.
34195 2021-01-12 Christophe Lyon <christophe.lyon@linaro.org>
34199 * config/arm/arm.h (ARM_HAVE_NEON_V8QI_LDST): New macro.
34200 (ARM_HAVE_NEON_V16QI_LDST, ARM_HAVE_NEON_V4HI_LDST): Likewise.
34201 (ARM_HAVE_NEON_V8HI_LDST, ARM_HAVE_NEON_V2SI_LDST): Likewise.
34202 (ARM_HAVE_NEON_V4SI_LDST, ARM_HAVE_NEON_V4HF_LDST): Likewise.
34203 (ARM_HAVE_NEON_V8HF_LDST, ARM_HAVE_NEON_V4BF_LDST): Likewise.
34204 (ARM_HAVE_NEON_V8BF_LDST, ARM_HAVE_NEON_V2SF_LDST): Likewise.
34205 (ARM_HAVE_NEON_V4SF_LDST, ARM_HAVE_NEON_DI_LDST): Likewise.
34206 (ARM_HAVE_NEON_V2DI_LDST): Likewise.
34207 (ARM_HAVE_V8QI_LDST, ARM_HAVE_V16QI_LDST): Likewise.
34208 (ARM_HAVE_V4HI_LDST, ARM_HAVE_V8HI_LDST): Likewise.
34209 (ARM_HAVE_V2SI_LDST, ARM_HAVE_V4SI_LDST, ARM_HAVE_V4HF_LDST): Likewise.
34210 (ARM_HAVE_V8HF_LDST, ARM_HAVE_V4BF_LDST, ARM_HAVE_V8BF_LDST): Likewise.
34211 (ARM_HAVE_V2SF_LDST, ARM_HAVE_V4SF_LDST, ARM_HAVE_DI_LDST): Likewise.
34212 (ARM_HAVE_V2DI_LDST): Likewise.
34213 * config/arm/mve.md (*movmisalign<mode>_mve_store): New pattern.
34214 (*movmisalign<mode>_mve_load): New pattern.
34215 * config/arm/neon.md (movmisalign<mode>): Move to ...
34216 * config/arm/vec-common.md: ... here.
34218 2021-01-12 Vladimir N. Makarov <vmakarov@redhat.com>
34221 * lra-eliminations.c (eliminate_regs_in_insn): Add transformation
34222 of pattern 'plus (plus (hard reg, const), pseudo)'.
34224 2021-01-12 Richard Biener <rguenther@suse.de>
34226 PR tree-optimization/98550
34227 * tree-vect-slp.c (vect_record_max_nunits): Check whether
34228 the group size is a multiple of the vector element count.
34229 (vect_build_slp_tree_1): When we need to fail because
34230 the vector type choosen causes unrolling do so lazily
34231 without affecting matches only at the end to guide group splitting.
34233 2021-01-12 Martin Liska <mliska@suse.cz>
34236 * optc-save-gen.awk: Compare also n_target_save vars with
34239 2021-01-12 Martin Liska <mliska@suse.cz>
34241 * gcov.c (source_info::debug): New.
34242 (print_usage): Add --debug (-D) option.
34243 (process_args): Likewise.
34244 (generate_results): Call src->debug after
34245 accumulate_line_counts.
34246 (read_graph_file): Properly assign id for EXIT_BLOCK.
34247 * profile.c (branch_prob): Dump function body before it is
34250 2021-01-12 Jakub Jelinek <jakub@redhat.com>
34252 PR tree-optimization/98629
34253 * tree-ssa-math-opts.c (arith_overflow_check_p): Don't update use_stmt
34254 unless returning non-zero.
34256 2021-01-12 Jakub Jelinek <jakub@redhat.com>
34258 PR tree-optimization/95731
34259 * tree-ssa-reassoc.c (optimize_range_tests_cmp_bitwise): Also optimize
34260 x < 0 && y < 0 && z < 0 into (x | y | z) < 0 for signed x, y, z.
34261 (optimize_range_tests): Call optimize_range_tests_cmp_bitwise
34262 only after optimize_range_tests_var_bound.
34264 2021-01-12 Jakub Jelinek <jakub@redhat.com>
34266 * configure.ac: Ensure c/Make-lang.in comes first in @all_lang_makefrags@.
34267 * configure: Regenerated.
34269 2021-01-12 liuhongt <hongtao.liu@intel.com>
34272 * config/i386/i386-builtins.h (BUILTIN_DESC_SWAP_OPERANDS):
34274 * config/i386/i386-expand.c (ix86_expand_sse_comi): Delete
34277 2021-01-12 Alexandre Oliva <oliva@adacore.com>
34279 * ssa-iterators.h (end_imm_use_stmt_traverse): Forward
34281 (auto_end_imm_use_stmt_traverse): New struct.
34282 (FOR_EACH_IMM_USE_STMT): Use it.
34283 (BREAK_FROM_IMM_USE_STMT, RETURN_FROM_IMM_USE_STMT): Remove,
34285 * gimple-ssa-strength-reduction.c: ... here, ...
34286 * graphite-scop-detection.c: ... here, ...
34287 * ipa-modref.c, ipa-pure-const.c, ipa-sra.c: ... here, ...
34288 * tree-predcom.c, tree-ssa-ccp.c: ... here, ...
34289 * tree-ssa-dce.c, tree-ssa-dse.c: ... here, ...
34290 * tree-ssa-loop-ivopts.c, tree-ssa-math-opts.c: ... here, ...
34291 * tree-ssa-phiprop.c, tree-ssa.c: ... here, ...
34292 * tree-vect-slp.c: ... and here, ...
34293 * doc/tree-ssa.texi: ... and the example here.
34295 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
34297 * config/aarch64/aarch64-sve.md (sdiv_pow2<mode>3): Extend from
34298 SVE_FULL_I to SVE_I. Generate an UNSPEC_PRED_X.
34299 (*sdiv_pow2<mode>3): New pattern.
34300 (@cond_<sve_int_op><mode>): Extend from SVE_FULL_I to SVE_I.
34301 Wrap the ASRD in an UNSPEC_PRED_X.
34302 (*cond_<sve_int_op><mode>_2): Likewise. Replace the UNSPEC_PRED_X
34303 predicate with a constant PTRUE, if it isn't already.
34304 (*cond_<sve_int_op><mode>_z): Replace with...
34305 (*cond_<sve_int_op><mode>_any): ...this new pattern.
34307 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
34309 * config/aarch64/aarch64-sve.md (*cond_bic<mode>_2): Extend from
34310 SVE_FULL_I to SVE_I.
34311 (*cond_bic<mode>_any): Likewise.
34313 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
34315 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart)
34316 (@aarch64_pred_<MUL_HIGHPART:optab><mode>): Extend from SVE_FULL_I
34319 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
34321 * config/aarch64/aarch64-sve.md (<su>abd<mode>_3): Extend from
34322 SVE_FULL_I to SVE_I.
34323 (*aarch64_cond_<su>abd<mode>_2): Likewise.
34324 (*aarch64_cond_<su>abd<mode>_any): Likewise.
34325 (@aarch64_pred_<su>abd<mode>): Likewise. Use UNSPEC_PRED_X
34326 for the max and min but not for the minus.
34327 (*aarch64_cond_<su>abd<mode>_3): New pattern.
34329 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
34331 * config/aarch64/iterators.md (SVE_24I): New iterator.
34332 * config/aarch64/aarch64-sve.md (*aarch64_adr<mode>_shift): Extend from
34333 SVE_FULL_SDI to SVE_24I. Use containers rather than elements.
34335 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
34337 * config/aarch64/aarch64-sve.md (@cond_<SVE_INT_BINARY:optab><mode>)
34338 (*cond_<SVE_INT_BINARY:optab><mode>_2): Extend from SVE_FULL_I
34340 (*cond_<SVE_INT_BINARY:optab><mode>_3): Likewise.
34341 (*cond_<SVE_INT_BINARY:optab><mode>_any): Likewise.
34342 (*cond_<SVE_INT_BINARY:optab><mode>_2_const): Likewise.
34343 (*cond_<SVE_INT_BINARY:optab><mode>_any_const): Likewise.
34345 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
34347 * config/aarch64/aarch64-sve.md (<SVE_INT_BINARY_IMM:optab><mode>3)
34348 (@aarch64_pred_<SVE_INT_BINARY_IMM:optab><mode>)
34349 (*post_ra_<SVE_INT_BINARY_IMM:optab><mode>3): Extend from SVE_FULL_I
34352 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
34354 * config/aarch64/aarch64-sve.md (<ASHIFT:optab><mode>3)
34355 (v<ASHIFT:optab><mode>3, @aarch64_pred_<optab><mode>)
34356 (*post_ra_v<ASHIFT:optab><mode>3): Extend from SVE_FULL_I to SVE_I.
34358 2021-01-11 Martin Liska <mliska@suse.cz>
34361 * symtab-clones.h (clone_info::release): Release
34362 symtab::m_clones with ggc_delete as it's a GGC memory.
34364 2021-01-11 Matthias Klose <doko@ubuntu.com>
34366 * Makefile.in (LINK_PROGRESS): Show the link target.
34368 2021-01-11 Richard Biener <rguenther@suse.de>
34370 PR tree-optimization/91403
34371 * tree-vect-data-refs.c (vect_analyze_group_access_1): Cap
34372 single-element interleaving group size at 4096 elements.
34374 2021-01-11 Richard Biener <rguenther@suse.de>
34376 PR tree-optimization/98526
34377 * tree-vect-loop.c (vect_model_reduction_cost): Remove costing
34378 of the actual reduction op for the regular case.
34379 (vectorizable_reduction): Cost the stmts
34380 vect_transform_reduction produces here.
34382 2021-01-11 Andreas Krebbel <krebbel@linux.ibm.com>
34384 * tree-ssa-forwprop.c (simplify_vector_constructor): For
34385 big-endian, use UNPACK[_FLOAT]_HI.
34387 2021-01-11 Tamar Christina <tamar.christina@arm.com>
34389 * tree-vect-slp-patterns.c (class complex_pattern,
34390 class complex_add_pattern): Add parameters to matches.
34391 (complex_add_pattern::build): Free memory.
34392 (complex_add_pattern::matches): Move validation end of match.
34393 (complex_add_pattern::recognize): Likewise.
34395 2021-01-11 Tamar Christina <tamar.christina@arm.com>
34397 * tree-vect-slp-patterns.c (linear_loads_p): Fix externals.
34399 2021-01-11 Tamar Christina <tamar.christina@arm.com>
34401 * tree-vect-slp-patterns.c (is_linear_load_p): Fix ambiguity.
34403 2021-01-11 Jakub Jelinek <jakub@redhat.com>
34405 PR tree-optimization/95867
34406 * tree-ssa-math-opts.h: New header.
34407 * tree-ssa-math-opts.c: Include tree-ssa-math-opts.h.
34408 (powi_as_mults): No longer static. Use build_one_cst instead of
34409 build_real. Formatting fix.
34410 * tree-ssa-reassoc.c: Include tree-ssa-math-opts.h.
34411 (attempt_builtin_powi): Handle multiplication reassociation without
34412 powi_fndecl using powi_as_mults.
34413 (reassociate_bb): For integral types don't require
34414 -funsafe-math-optimizations to call attempt_builtin_powi.
34416 2021-01-11 Jakub Jelinek <jakub@redhat.com>
34418 PR tree-optimization/95852
34419 * tree-ssa-math-opts.c (maybe_optimize_guarding_check): Change
34420 mul_stmts parameter type to vec<gimple *> &. Before cond_stmt
34421 allow in the bb any of the stmts in that vector, div_stmt and
34422 up to 3 cast stmts.
34423 (arith_cast_equal_p): New function.
34424 (arith_overflow_check_p): Add cast_stmt argument, handle signed
34425 multiply overflow checks.
34426 (match_arith_overflow): Adjust caller. Handle signed multiply
34429 2021-01-11 Jakub Jelinek <jakub@redhat.com>
34431 PR tree-optimization/95852
34432 * tree-ssa-math-opts.c (maybe_optimize_guarding_check): New function.
34433 (uaddsub_overflow_check_p): Renamed to ...
34434 (arith_overflow_check_p): ... this. Handle also multiplication
34435 with overflow check.
34436 (match_uaddsub_overflow): Renamed to ...
34437 (match_arith_overflow): ... this. Add cfg_changed argument. Handle
34438 also multiplication with overflow check. Adjust function comment.
34439 (math_opts_dom_walker::after_dom_children): Adjust callers. Call
34440 match_arith_overflow also for MULT_EXPR.
34442 2021-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
34444 * config/aarch64/arm_neon.h (vmovl_s8): Reimplement using
34445 __builtin_convertvector.
34446 (vmovl_s16): Likewise.
34447 (vmovl_s32): Likewise.
34448 (vmovl_u8): Likewise.
34449 (vmovl_u16): Likewise.
34450 (vmovl_u32): Likewise.
34451 (vmovn_s16): Likewise.
34452 (vmovn_s32): Likewise.
34453 (vmovn_s64): Likewise.
34454 (vmovn_u16): Likewise.
34455 (vmovn_u32): Likewise.
34456 (vmovn_u64): Likewise.
34458 2021-01-11 Martin Liska <mliska@suse.cz>
34460 * gimple-if-to-switch.cc (struct condition_info): Use auto_var.
34461 (if_chain::is_beneficial): Delete clusters
34462 (find_conditions): Make second argument of conditions_in_bbs a
34463 pointer so that we control over it's lifetime.
34464 (pass_if_to_switch::execute): Delete them.
34466 2021-01-11 Kewen Lin <linkw@linux.ibm.com>
34468 * ira.c (move_unallocated_pseudos): Check other_reg and skip if
34471 2021-01-09 Maciej W. Rozycki <macro@linux-mips.org>
34473 * config/vax/vax.md (cc): Remove mode attribute.
34474 (subst_<cc>, subst_f<cc>): Rename to...
34475 (subst_<mode>, subst_f<VAXccnz:mode>): ... these respectively.
34476 (*cbranch<VAXint:mode>4_<VAXcc:mode>): Update for `cc' removal.
34477 (*cbranch<VAXfp:mode>4_<VAXccnz:mode>): Likewise.
34478 (*branch_<mode>, *branch_<mode>_reversed): Likewise.
34480 2021-01-09 Maciej W. Rozycki <macro@linux-mips.org>
34482 * config/vax/vax.md (subst_f<cc>): Add mode to operands and
34483 `const_double_zero'.
34485 2021-01-09 Maciej W. Rozycki <macro@linux-mips.org>
34487 * config/pdp11/pdp11.md (PDPfp): New mode iterator.
34488 (fcc_cc, fcc_ccnz): Use it. Add mode to `const_double_zero' and
34491 2021-01-09 Maciej W. Rozycki <macro@linux-mips.org>
34493 * genemit.c (gen_exp) <CONST_DOUBLE>: Handle `const_double_zero'
34495 * read-rtl.c (rtx_reader::read_rtx_code): Handle machine mode
34496 with `const_double_zero'.
34497 * doc/rtl.texi (Constant Expression Types): Document it.
34499 2021-01-09 Jakub Jelinek <jakub@redhat.com>
34502 * tree-cfg.c (verify_gimple_assign_binary): Allow lhs of
34503 POINTER_DIFF_EXPR to be any integral type.
34505 2021-01-09 Jakub Jelinek <jakub@redhat.com>
34507 PR rtl-optimization/98603
34508 * function.c (instantiate_virtual_regs_in_insn): For asm goto
34509 with impossible constraints, drop all SETs, CLOBBERs, drop PARALLEL
34510 if any, set ASM_OPERANDS mode to VOIDmode and change
34511 ASM_OPERANDS_OUTPUT_CONSTRAINT and ASM_OPERANDS_OUTPUT_IDX.
34513 2021-01-09 Alexandre Oliva <oliva@gnu.org>
34516 * final.c (notice_source_line): Narrow down the condition to
34517 skip a line-0 marker.
34519 2021-01-08 Sergei Trofimovich <siarheit@google.com>
34521 * ipa-modref.c (merge_call_side_effects): Fix
34522 linebreak split by reordering two print calls.
34524 2021-01-08 Ilya Leoshkevich <iii@linux.ibm.com>
34526 * config/s390/vector.md (*tf_to_fprx2_0): Rename from
34527 "*mov_tf_to_fprx2_0" for consistency, fix constraint.
34528 (*tf_to_fprx2_1): Rename from "*mov_tf_to_fprx2_1" for
34529 consistency, fix constraint.
34531 2021-01-08 Ilya Leoshkevich <iii@linux.ibm.com>
34533 * config/s390/s390-c.c (s390_def_or_undef_macro): Accept
34534 callables instead of mask values.
34535 (struct target_flag_set_p): New predicate.
34536 (s390_cpu_cpp_builtins_internal): Define or undefine
34537 __LONG_DOUBLE_VX__ macro.
34539 2021-01-08 H.J. Lu <hjl.tools@gmail.com>
34542 * config/i386/i386.c (x86_function_profiler): Use R10 and R11
34543 to call mcount in large model with PIC for NO_PROFILE_COUNTERS
34546 2021-01-08 Richard Biener <rguenther@suse.de>
34548 * tree-ssa-sccvn.c (pass_fre::execute): Reset the SCEV hash table.
34550 2021-01-08 Richard Biener <rguenther@suse.de>
34552 * tree-vect-slp.c (scalar_stmts_to_slp_tree_map_t): Fix.
34553 (vect_build_slp_tree): On cache hit release the matched
34554 scalar stmts vector.
34555 * tree-vect-stmts.c (vectorizable_store): Properly free
34556 vec_oprnds before possibly gathering them again.
34558 2021-01-08 Richard Biener <rguenther@suse.de>
34560 PR tree-optimization/98544
34561 * tree-vect-slp.c (vect_optimize_slp): Always materialize
34562 permutes at a permute node.
34564 2021-01-08 H.J. Lu <hjl.tools@gmail.com>
34567 * config/i386/i386.c (x86_function_profiler): Use R10 to call
34568 mcount in large model. Sorry for large model with PIC.
34570 2021-01-08 Jakub Jelinek <jakub@redhat.com>
34573 * config/i386/i386.opt (ix86_cmodel, ix86_incoming_stack_boundary_arg,
34574 ix86_pmode, ix86_preferred_stack_boundary_arg, ix86_regparm,
34575 ix86_veclibabi_type): Remove x_ prefix, use TargetVariable instead of
34576 TargetSave and initialize for variables with enum types.
34577 (mfentry, mstack-protector-guard-reg=, mstack-protector-guard-offset=,
34578 mstack-protector-guard-symbol=): Add Save.
34579 * config/i386/i386-options.c (ix86_function_specific_save,
34580 ix86_function_specific_restore): Don't save or restore x_ix86_cmodel,
34581 x_ix86_incoming_stack_boundary_arg, x_ix86_pmode,
34582 x_ix86_preferred_stack_boundary_arg, x_ix86_regparm,
34583 x_ix86_veclibabi_type.
34585 2021-01-08 Richard Sandiford <richard.sandiford@arm.com>
34587 * config/aarch64/aarch64-sve.md (*cnot<mode>): Extend from
34588 SVE_FULL_I to SVE_I.
34589 (*cond_cnot<mode>_2, *cond_cnot<mode>_any): Likewise.
34591 2021-01-08 Richard Sandiford <richard.sandiford@arm.com>
34593 * config/aarch64/aarch64-sve.md (*cond_uxt<mode>_2): Extend from
34594 SVE_FULL_I to SVE_I.
34595 (*cond_uxt<mode>_any): Likewise.
34597 2021-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
34599 * config/aarch64/iterators.md (Vwhalf): New iterator.
34600 * config/aarch64/aarch64-simd.md (aarch64_<sur>adalp<mode>_3):
34602 (aarch64_<sur>adalp<mode>): ... This. Make more
34604 (<sur>sadv16qi): Adjust callsite of the above.
34605 * config/aarch64/aarch64-simd-builtins.def (sadalp, uadalp): New
34607 * config/aarch64/arm_neon.h (vpadal_s8): Reimplement using
34609 (vpadal_s16): Likewise.
34610 (vpadal_u8): Likewise.
34611 (vpadal_u16): Likewise.
34612 (vpadalq_s8): Likewise.
34613 (vpadalq_s16): Likewise.
34614 (vpadalq_s32): Likewise.
34615 (vpadalq_u8): Likewise.
34616 (vpadalq_u16): Likewise.
34617 (vpadalq_u32): Likewise.
34619 2021-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
34621 * config/aarch64/aarch64-simd.md (aarch64_<su>abd<mode>_3):
34623 (aarch64_<su>abd<mode>): ... This.
34624 (<sur>sadv16qi): Adjust callsite of the above.
34625 * config/aarch64/aarch64-simd-builtins.def (sabd, uabd): Define
34627 * config/aarch64/arm_neon.h (vabd_s8): Reimplement using
34629 (vabd_s16): Likewise.
34630 (vabd_s32): Likewise.
34631 (vabd_u8): Likewise.
34632 (vabd_u16): Likewise.
34633 (vabd_u32): Likewise.
34634 (vabdq_s8): Likewise.
34635 (vabdq_s16): Likewise.
34636 (vabdq_s32): Likewise.
34637 (vabdq_u8): Likewise.
34638 (vabdq_u16): Likewise.
34639 (vabdq_u32): Likewise.
34641 2021-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
34643 * config/aarch64/aarch64-simd-builtins.def (saba, uaba): Define
34645 * config/aarch64/arm_neon.h (vaba_s8): Implement using builtin.
34646 (vaba_s16): Likewise.
34647 (vaba_s32): Likewise.
34648 (vaba_u8): Likewise.
34649 (vaba_u16): Likewise.
34650 (vaba_u32): Likewise.
34651 (vabaq_s8): Likewise.
34652 (vabaq_s16): Likewise.
34653 (vabaq_s32): Likewise.
34654 (vabaq_u8): Likewise.
34655 (vabaq_u16): Likewise.
34656 (vabaq_u32): Likewise.
34658 2021-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
34660 * config/aarch64/aarch64-simd.md (aba<mode>_3): Rename to...
34661 (aarch64_<su>aba<mode>): ... This. Handle uaba as well.
34662 Change RTL pattern to match.
34664 2021-01-08 Kito Cheng <kito.cheng@sifive.com>
34666 * common/config/riscv/riscv-common.c (riscv_current_subset_list): New.
34667 * config/riscv/riscv-c.c (riscv-subset.h): New.
34668 (INCLUDE_STRING): Define.
34669 (riscv_cpu_cpp_builtins): Add new style architecture extension
34671 * config/riscv/riscv-subset.h (riscv_subset_list::begin): New.
34672 (riscv_subset_list::end): New.
34673 (riscv_current_subset_list): New.
34675 2021-01-08 Kito Cheng <kito.cheng@sifive.com>
34677 * common/config/riscv/riscv-common.c (RISCV_DONT_CARE_VERSION):
34678 Move to riscv-subset.h.
34679 (struct riscv_subset_t): Ditto.
34680 (class riscv_subset_list): Ditto.
34681 * config/riscv/riscv-subset.h (RISCV_DONT_CARE_VERSION): Move
34682 from riscv-common.c.
34683 (struct riscv_subset_t): Ditto.
34684 (class riscv_subset_list): Ditto.
34685 * config/riscv/t-riscv ($(common_out_file)): Add file
34688 2021-01-07 Jakub Jelinek <jakub@redhat.com>
34691 * config/i386/i386.md (*bmi_blsi_<mode>_cmp, *bmi_blsi_<mode>_ccno):
34692 New define_insn patterns.
34694 2021-01-07 Richard Sandiford <richard.sandiford@arm.com>
34696 * config/aarch64/aarch64-sve.md (@cond_<SVE_INT_UNARY:optab><mode>)
34697 (*cond_<SVE_INT_UNARY:optab><mode>_2): Extend from SVE_FULL_I to SVE_I.
34698 (*cond_<SVE_INT_UNARY:optab><mode>_any): Likewise.
34700 2021-01-07 Richard Sandiford <richard.sandiford@arm.com>
34702 PR tree-optimization/98560
34703 * internal-fn.def (IFN_VCONDU, IFN_VCONDEQ): Use type vec_cond.
34704 * internal-fn.c (vec_cond_mask_direct): Get the data mode from
34706 (vec_cond_direct): Likewise argument 2.
34707 (vec_condu_direct, vec_condeq_direct): Delete.
34708 (expand_vect_cond_optab_fn): Rename to...
34709 (expand_vec_cond_optab_fn): ...this, replacing old macro.
34710 (expand_vec_condu_optab_fn, expand_vec_condeq_optab_fn): Delete.
34711 (expand_vect_cond_mask_optab_fn): Rename to...
34712 (expand_vec_cond_mask_optab_fn): ...this, replacing old macro.
34713 (direct_vec_cond_mask_optab_supported_p): Treat the optab as a
34715 (direct_vec_cond_optab_supported_p): Likewise.
34716 (direct_vec_condu_optab_supported_p): Delete.
34717 (direct_vec_condeq_optab_supported_p): Delete.
34718 * gimple-isel.cc: Include internal-fn.h.
34719 (gimple_expand_vec_cond_expr): Check that IFN_VCONDEQ is supported
34722 2021-01-07 Richard Sandiford <richard.sandiford@arm.com>
34724 PR tree-optimization/98560
34725 * gimple-isel.cc (gimple_expand_vec_cond_expr): If we fail to use
34726 IFN_VCOND{,U,EQ}, fall back on IFN_VCOND_MASK.
34728 2021-01-07 Uroš Bizjak <ubizjak@gmail.com>
34730 * config/i386/i386.md (insn): Merge from plusminus_insn, shift_insn,
34731 rotate_insn and optab code attributes.
34732 Update all uses to merged code attribute.
34733 * config/i386/sse.md: Update all uses to merged code attribute.
34734 * config/i386/mmx.md: Update all uses to merged code attribute.
34736 2021-01-07 Jakub Jelinek <jakub@redhat.com>
34738 PR tree-optimization/98568
34739 * gimple-ssa-store-merging.c (bswap_view_convert): New function.
34740 (bswap_replace): Use it.
34742 2021-01-06 Vladimir N. Makarov <vmakarov@redhat.com>
34744 PR rtl-optimization/97978
34745 * lra-int.h (lra_hard_reg_split_p): New external.
34746 * lra.c (lra_hard_reg_split_p): New global.
34747 (lra): Set up lra_hard_reg_split_p after splitting a hard reg.
34748 * lra-assigns.c (lra_assign): Don't check allocation correctness
34749 after hard reg splitting.
34751 2021-01-06 Martin Sebor <msebor@redhat.com>
34754 * builtins.c (new_delete_mismatch_p): New overload.
34755 (new_delete_mismatch_p (tree, tree)): Call it.
34757 2021-01-06 Alexandre Oliva <oliva@adacore.com>
34759 * Makefile.in (T_GLIMITS_H): New.
34760 (stmp-int-hdrs): Depend on it, use it.
34761 * config/t-vxworks (T_GLIMITS_H): Override it.
34762 (vxw-glimits.h): New.
34764 2021-01-06 Richard Biener <rguenther@suse.de>
34766 PR tree-optimization/98513
34767 * value-range.cc (intersect_ranges): Compare the upper bounds
34768 for the expected relation.
34770 2021-01-06 Gerald Pfeifer <gerald@pfeifer.com>
34773 2020-12-28 Gerald Pfeifer <gerald@pfeifer.com>
34775 * doc/standards.texi (HSAIL): Remove section.
34777 2021-01-05 Samuel Thibault <samuel.thibault@ens-lyon.org>
34779 * configure: Re-generate.
34781 2021-01-05 Jakub Jelinek <jakub@redhat.com>
34783 * doc/invoke.texi (-std=c++20): Adjust for the publication of
34784 ISO 14882:2020 standard.
34785 * doc/standards.texi: Likewise.
34787 2021-01-05 Jakub Jelinek <jakub@redhat.com>
34789 PR tree-optimization/94802
34790 * expr.h (maybe_optimize_sub_cmp_0): Declare.
34791 * expr.c: Include tree-pretty-print.h and flags.h.
34792 (maybe_optimize_sub_cmp_0): New function.
34793 (do_store_flag): Use it.
34794 * cfgexpand.c (expand_gimple_cond): Likewise.
34796 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
34798 * mux-utils.h (pointer_mux::m_ptr): Tweak description of contents.
34799 * rtlanal.c (simple_regno_set): Tweak description to clarify the
34802 2021-01-05 Richard Biener <rguenther@suse.de>
34804 PR tree-optimization/98516
34805 * tree-vect-slp.c (vect_optimize_slp): Permute the incoming
34806 lanes when materializing on a VEC_PERM node.
34807 (vectorizable_slp_permutation): Dump the permute properly.
34809 2021-01-05 Richard Biener <rguenther@suse.de>
34811 * tree-vect-slp.c (vect_slp_region): Move debug counter
34812 to cover individual subgraphs.
34814 2021-01-05 Richard Biener <rguenther@suse.de>
34816 PR tree-optimization/98428
34817 * tree-vect-slp.c (vect_build_slp_tree_1): Properly reject
34818 vector lane extracts for loop vectorization.
34820 2021-01-05 Jakub Jelinek <jakub@redhat.com>
34822 PR tree-optimization/98514
34823 * tree-ssa-reassoc.c (bb_rank): Change type from long * to
34825 (operand_rank): Change type from hash_map<tree, long> to
34826 hash_map<tree, int64_t>.
34827 (phi_rank): Change return type from long to int64_t.
34828 (loop_carried_phi): Change block_rank variable type from long to
34830 (propagate_rank): Change return type, rank parameter type and
34831 op_rank variable type from long to int64_t.
34832 (find_operand_rank): Change return type from long to int64_t
34833 and change slot variable type from long * to int64_t *.
34834 (insert_operand_rank): Change rank parameter type from long to
34836 (get_rank): Change return type and rank variable type from long to
34837 int64_t. Use PRId64 instead of ld to print the rank.
34838 (init_reassoc): Change rank variable type from long to int64_t
34839 and adjust correspondingly bb_rank and operand_rank initialization.
34841 2021-01-05 Jakub Jelinek <jakub@redhat.com>
34843 PR tree-optimization/96928
34844 * tree-ssa-phiopt.c (xor_replacement): New function.
34845 (tree_ssa_phiopt_worker): Call it.
34847 2021-01-05 Jakub Jelinek <jakub@redhat.com>
34849 PR tree-optimization/96930
34850 * match.pd ((A / (1 << B)) -> (A >> B)): If A is extended
34851 from narrower value which has the same type as 1 << B, perform
34852 the right shift on the narrower value followed by extension.
34854 2021-01-05 Jakub Jelinek <jakub@redhat.com>
34856 PR tree-optimization/96239
34857 * gimple-ssa-store-merging.c (maybe_optimize_vector_constructor): New
34859 (get_status_for_store_merging): Don't return BB_INVALID for blocks
34860 with potential bswap optimizable CONSTRUCTORs.
34861 (pass_store_merging::execute): Optimize vector CONSTRUCTORs with bswap
34864 2021-01-05 Richard Biener <rguenther@suse.de>
34866 PR tree-optimization/98381
34867 * tree.c (vector_element_bits): Properly compute bool vector
34869 * tree-vect-loop.c (vectorizable_live_operation): Properly
34870 compute the last lane bit offset.
34872 2021-01-05 Uroš Bizjak <ubizjak@gmail.com>
34875 * config/i386/sse.md (sse_cvtps2pi): Redefine as define_insn_and_split.
34876 Clear the top 64 bytes of the input XMM register.
34877 (sse_cvttps2pi): Ditto.
34879 2021-01-05 Uroš Bizjak <ubizjak@gmail.com>
34882 * config/i386/xopintrin.h (_mm256_cmov_si256): New.
34884 2021-01-05 H.J. Lu <hjl.tools@gmail.com>
34887 * config/i386/xmmintrin.h (_mm_extract_pi16): Cast to unsigned
34890 2021-01-05 Claudiu Zissulescu <claziss@synopsys.com>
34892 * config/arc/arc.md (maddsidi4_split): Use ACC_REG_FIRST.
34893 (umaddsidi4_split): Likewise.
34895 2021-01-05 liuhongt <hongtao.liu@intel.com>
34898 * config/i386/sse.md (*sse2_pmovskb_zexthisi): New
34899 define_insn_and_split for zero_extend of subreg HI of pmovskb
34901 (*sse2_pmovskb_zexthisi): Add new combine splitters for
34902 zero_extend of not of subreg HI of pmovskb result.
34904 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
34907 * explow.c (convert_memory_address_addr_space_1): Handle UNSPECs
34909 * config/aarch64/aarch64.c (aarch64_expand_mov_immediate): Use
34910 convert_memory_address to convert symbolic immediates to ptr_mode
34911 before forcing them to memory.
34913 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
34915 PR rtl-optimization/97144
34916 * recog.c (constrain_operands): Initialize matching_operand
34917 for each alternative, rather than only doing it once.
34919 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
34921 PR rtl-optimization/98403
34922 * rtl-ssa/changes.cc (function_info::finalize_new_accesses): Explain
34923 why we don't remove call clobbers.
34924 (function_info::apply_changes_to_insn): Don't attempt to add
34925 call clobbers here.
34927 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
34929 PR tree-optimization/98371
34930 * tree-vect-loop.c (vect_reanalyze_as_main_loop): New function.
34931 (vect_analyze_loop): If an epilogue loop appears to be cheaper
34932 than the main loop, re-analyze it as a main loop before adopting
34935 2021-01-05 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
34938 * configure.ac (NETLIBS): Determine using AX_LIB_SOCKET_NSL.
34939 * aclocal.m4, configure: Regenerate.
34940 * Makefile.in (NETLIBS): Define.
34941 (BACKEND): Remove $(CODYLIB).
34943 2021-01-05 Jakub Jelinek <jakub@redhat.com>
34945 PR rtl-optimization/98334
34946 * simplify-rtx.c (simplify_context::simplify_binary_operation_1):
34947 Optimize (X - 1) * Y + Y to X * Y or (X + 1) * Y - Y to X * Y.
34949 2021-01-05 Bernd Edlinger <bernd.edlinger@hotmail.de>
34951 * tree-inline.c (expand_call_inline): Restore input_location.
34952 Return result from recursive call.
34954 2021-01-04 Richard Sandiford <richard.sandiford@arm.com>
34956 PR tree-optimization/95401
34957 * config/aarch64/aarch64-sve-builtins.cc
34958 (gimple_folder::load_store_cookie): Use bits rather than bytes
34959 for the alignment argument to IFN_MASK_LOAD and IFN_MASK_STORE.
34960 * gimple-fold.c (gimple_fold_mask_load_store_mem_ref): Likewise.
34961 * tree-vect-stmts.c (vectorizable_store): Likewise.
34962 (vectorizable_load): Likewise.
34964 2021-01-04 Richard Biener <rguenther@suse.de>
34966 PR tree-optimization/98308
34967 * tree-vect-stmts.c (vectorizable_load): Set invariant mask
34970 2021-01-04 Jakub Jelinek <jakub@redhat.com>
34972 PR tree-optimization/95771
34973 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Handle types
34974 with precision smaller than int's precision and types with precision
34975 twice as large as long long. Formatting fixes.
34977 2021-01-04 Richard Biener <rguenther@suse.de>
34979 PR tree-optimization/98464
34980 * tree-ssa-sccvn.c (vn_valueize_for_srt): Rename from ...
34981 (vn_valueize_wrapper): ... this. Temporarily adjust vn_context_bb.
34982 (process_bb): Adjust.
34984 2021-01-04 Matthew Malcomson <matthew.malcomson@arm.com>
34987 * doc/invoke.texi (-fsanitize=address): Fix wording describing
34988 clash with -fsanitize=hwaddress.
34990 2021-01-04 Richard Biener <rguenther@suse.de>
34992 PR tree-optimization/98282
34993 * tree-ssa-sccvn.c (vn_get_stmt_kind): Classify tcc_reference on
34994 invariants as VN_NARY.
34996 2021-01-04 Richard Sandiford <richard.sandiford@arm.com>
34999 * config/aarch64/aarch64-simd.md (aarch64_combine<mode>): Accept
35000 aarch64_simd_reg_or_zero for operand 2. Use the combinez patterns
35001 to handle zero operands.
35003 2021-01-04 Richard Sandiford <richard.sandiford@arm.com>
35005 * config/aarch64/aarch64.c (offset_6bit_signed_scaled_p): New function.
35006 (offset_6bit_unsigned_scaled_p): Fix typo in comment.
35007 (aarch64_sve_prefetch_operand_p): Accept MUL VLs in the range
35010 2021-01-04 Richard Biener <rguenther@suse.de>
35012 PR tree-optimization/98393
35013 * tree-vect-slp.c (vect_build_slp_tree): Properly zero matches
35014 when hitting the limit.
35016 2021-01-04 Richard Biener <rguenther@suse.de>
35018 PR tree-optimization/98291
35019 * tree-vect-loop.c (vectorizable_reduction): Bypass
35020 associativity check for SLP reductions with VF 1.
35022 2021-01-04 Jakub Jelinek <jakub@redhat.com>
35024 PR tree-optimization/96782
35025 * match.pd (x == ~x -> false, x != ~x -> true): New simplifications.
35027 2021-01-04 Bernd Edlinger <bernd.edlinger@hotmail.de>
35029 * collect-utils.c (collect_execute): Check dumppfx.
35030 * collect2.c (maybe_run_lto_and_relink, do_link): Pass atsuffix
35031 to collect_execute.
35032 (do_link): Add new parameter atsuffix.
35033 (main): Handle -dumpdir option. Skip one argument for
35034 -o, -isystem and -B options.
35035 * gcc.c (make_at_file): New helper function.
35036 (close_at_file): Use it.
35038 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
35040 * config/darwin.h (MIN_LD64_NO_COAL_SECTS): Adjust.
35041 Amend handling for LD64_VERSION fallback defaults.
35043 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
35045 * config.gcc: Compute default version information
35046 from the configured target. Likewise defaults for
35048 * config/darwin10.h: Removed.
35049 * config/darwin12.h: Removed.
35050 * config/darwin9.h: Removed.
35051 * config/rs6000/darwin8.h: Removed.
35053 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
35055 * config/darwin9.h (ASM_OUTPUT_ALIGNED_COMMON): Delete.
35057 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
35059 * config/darwin9.h (STACK_CHECK_STATIC_BUILTIN): Move from here..
35060 * config/darwin.h (STACK_CHECK_STATIC_BUILTIN): .. to here.
35062 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
35064 * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Move from
35066 * config/darwin.h (LINK_GCC_C_SEQUENCE_SPEC): ... to here.
35068 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
35070 * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Move the spec
35071 for the Darwin10 unwinder stub from here ...
35072 * config/darwin.h (LINK_COMMAND_SPEC_A): ... to here.
35074 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
35076 * config/darwin.h (DSYMUTIL_SPEC): Default to DWARF
35077 (ASM_DEBUG_SPEC):Only define if the assembler supports
35079 (PREFERRED_DEBUGGING_TYPE): Default to DWARF.
35080 (DARWIN_PREFER_DWARF): Define.
35081 * config/darwin9.h (PREFERRED_DEBUGGING_TYPE): Remove.
35082 (DARWIN_PREFER_DWARF): Likewise
35083 (DSYMUTIL_SPEC): Likewise.
35084 (COLLECT_RUN_DSYMUTIL): Likewise.
35085 (ASM_DEBUG_SPEC): Likewise.
35086 (ASM_DEBUG_OPTION_SPEC): Likewise.
35088 2021-01-02 Jan Hubicka <jh@suse.cz>
35090 * cfg.c (free_block): ggc_free bb.
35092 2021-01-01 Jakub Jelinek <jakub@redhat.com>
35094 * gcc.c (process_command): Update copyright notice dates.
35095 * gcov-dump.c (print_version): Ditto.
35096 * gcov.c (print_version): Ditto.
35097 * gcov-tool.c (print_version): Ditto.
35098 * gengtype.c (create_file): Ditto.
35099 * doc/cpp.texi: Bump @copying's copyright year.
35100 * doc/cppinternals.texi: Ditto.
35101 * doc/gcc.texi: Ditto.
35102 * doc/gccint.texi: Ditto.
35103 * doc/gcov.texi: Ditto.
35104 * doc/install.texi: Ditto.
35105 * doc/invoke.texi: Ditto.
35107 2021-01-01 Jakub Jelinek <jakub@redhat.com>
35109 * ChangeLog-2020: Rotate ChangeLog. New file.
35112 Copyright (C) 2021 Free Software Foundation, Inc.
35114 Copying and distribution of this file, with or without modification,
35115 are permitted in any medium without royalty provided the copyright
35116 notice and this notice are preserved.