1 2020-03-06 David Edelsohn <dje.gcc@gmail.com>
4 * config/rs6000/aix61.h (TARGET_NO_SUM_IN_TOC): Set to 1 for
6 (TARGET_NO_FP_IN_TOC): Same.
7 * config/rs6000/aix71.h: Same.
8 * config/rs6000/aix72.h: Same.
10 2020-03-06 Andrew Pinski <apinski@marvell.com>
11 Jeff Law <law@redhat.com>
13 PR rtl-optimization/93996
14 * haifa-sched.c (remove_notes): Be more careful when adding
17 2020-03-06 Delia Burduv <delia.burduv@arm.com>
19 * config/arm/arm_neon.h (vld2_bf16): New.
26 (vld2q_dup_bf16): New.
28 (vld3q_dup_bf16): New.
30 (vld4q_dup_bf16): New.
31 * config/arm/arm_neon_builtins.def
32 (vld2): Changed to VAR13 and added v4bf, v8bf
33 (vld2_dup): Changed to VAR8 and added v4bf, v8bf
34 (vld3): Changed to VAR13 and added v4bf, v8bf
35 (vld3_dup): Changed to VAR8 and added v4bf, v8bf
36 (vld4): Changed to VAR13 and added v4bf, v8bf
37 (vld4_dup): Changed to VAR8 and added v4bf, v8bf
38 * config/arm/iterators.md (VDXBF2): New iterator.
39 *config/arm/neon.md (neon_vld2): Use new iterators.
40 (neon_vld2_dup<mode): Use new iterators.
41 (neon_vld3<mode>): Likewise.
42 (neon_vld3qa<mode>): Likewise.
43 (neon_vld3qb<mode>): Likewise.
44 (neon_vld3_dup<mode>): Likewise.
45 (neon_vld4<mode>): Likewise.
46 (neon_vld4qa<mode>): Likewise.
47 (neon_vld4qb<mode>): Likewise.
48 (neon_vld4_dup<mode>): Likewise.
49 (neon_vld2_dupv8bf): New.
50 (neon_vld3_dupv8bf): Likewise.
51 (neon_vld4_dupv8bf): Likewise.
53 2020-03-06 Delia Burduv <delia.burduv@arm.com>
55 * config/arm/arm_neon.h (bfloat16x4x2_t): New typedef.
56 (bfloat16x8x2_t): New typedef.
57 (bfloat16x4x3_t): New typedef.
58 (bfloat16x8x3_t): New typedef.
59 (bfloat16x4x4_t): New typedef.
60 (bfloat16x8x4_t): New typedef.
67 * config/arm/arm-builtins.c (v2bf_UP): Define.
69 (arm_init_simd_builtin_types): Init Bfloat16x2_t eltype.
70 * config/arm/arm-modes.def (V2BF): New mode.
71 * config/arm/arm-simd-builtin-types.def
72 (Bfloat16x2_t): New entry.
73 * config/arm/arm_neon_builtins.def
74 (vst2): Changed to VAR13 and added v4bf, v8bf
75 (vst3): Changed to VAR13 and added v4bf, v8bf
76 (vst4): Changed to VAR13 and added v4bf, v8bf
77 * config/arm/iterators.md (VDXBF): New iterator.
78 (VQ2BF): New iterator.
79 *config/arm/neon.md (neon_vst2<mode>): Used new iterators.
80 (neon_vst2<mode>): Used new iterators.
81 (neon_vst3<mode>): Used new iterators.
82 (neon_vst3<mode>): Used new iterators.
83 (neon_vst3qa<mode>): Used new iterators.
84 (neon_vst3qb<mode>): Used new iterators.
85 (neon_vst4<mode>): Used new iterators.
86 (neon_vst4<mode>): Used new iterators.
87 (neon_vst4qa<mode>): Used new iterators.
88 (neon_vst4qb<mode>): Used new iterators.
90 2020-03-06 Delia Burduv <delia.burduv@arm.com>
92 * config/aarch64/aarch64-simd-builtins.def
93 (bfcvtn): New built-in function.
94 (bfcvtn_q): New built-in function.
95 (bfcvtn2): New built-in function.
96 (bfcvt): New built-in function.
97 * config/aarch64/aarch64-simd.md
98 (aarch64_bfcvtn<q><mode>): New pattern.
99 (aarch64_bfcvtn2v8bf): New pattern.
100 (aarch64_bfcvtbf): New pattern.
101 * config/aarch64/arm_bf16.h (float32_t): New typedef.
102 (vcvth_bf16_f32): New intrinsic.
103 * config/aarch64/arm_bf16.h (vcvt_bf16_f32): New intrinsic.
104 (vcvtq_low_bf16_f32): New intrinsic.
105 (vcvtq_high_bf16_f32): New intrinsic.
106 * config/aarch64/iterators.md (V4SF_TO_BF): New mode iterator.
107 (UNSPEC_BFCVTN): New UNSPEC.
108 (UNSPEC_BFCVTN2): New UNSPEC.
109 (UNSPEC_BFCVT): New UNSPEC.
110 * config/arm/types.md (bf_cvt): New type.
112 2020-03-06 Andreas Krebbel <krebbel@linux.ibm.com>
114 * config/s390/s390.md ("tabort"): Get rid of two consecutive
115 blanks in format string.
117 2020-03-05 H.J. Lu <hongjiu.lu@intel.com>
121 * config/i386/i386-protos.h (ix86_output_ssemov): New prototype.
122 * config/i386/i386.c (ix86_get_ssemov): New function.
123 (ix86_output_ssemov): Likewise.
124 * config/i386/sse.md (VMOVE:mov<mode>_internal): Call
125 ix86_output_ssemov for TYPE_SSEMOV. Remove TARGET_AVX512VL
127 (*movxi_internal_avx512f): Call ix86_output_ssemov for TYPE_SSEMOV.
128 (*movoi_internal_avx): Call ix86_output_ssemov for TYPE_SSEMOV.
129 Remove ext_sse_reg_operand and TARGET_AVX512VL check.
130 (*movti_internal): Likewise.
131 (*movtf_internal): Call ix86_output_ssemov for TYPE_SSEMOV.
133 2020-03-05 Jeff Law <law@redhat.com>
135 PR tree-optimization/91890
136 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Remove LOC argument.
137 Use gimple_or_expr_nonartificial_location.
138 (check_bounds_overlap): Drop LOC argument to maybe_diag_access_bounds.
139 Use gimple_or_expr_nonartificial_location.
140 * gimple.c (gimple_or_expr_nonartificial_location): New function.
141 * gimple.h (gimple_or_expr_nonartificial_location): Declare it.
142 * tree-ssa-strlen.c (maybe_warn_overflow): Use
143 gimple_or_expr_nonartificial_location.
144 (maybe_diag_stxncpy_trunc, handle_builtin_stxncpy_strncat): Likewise.
145 (maybe_warn_pointless_strcmp): Likewise.
147 2020-03-05 Jakub Jelinek <jakub@redhat.com>
150 * config/i386/avx2intrin.h (_mm_mask_i32gather_ps): Fix first cast of
151 SRC and MASK arguments to __m128 from __m128d.
152 (_mm256_mask_i32gather_ps): Fix first cast of MASK argument to __m256
154 (_mm_mask_i64gather_ps): Fix first cast of MASK argument to __m128
156 * config/i386/xopintrin.h (_mm_permute2_pd): Fix first cast of C
157 argument to __m128i from __m128d.
158 (_mm256_permute2_pd): Fix first cast of C argument to __m256i from
160 (_mm_permute2_ps): Fix first cast of C argument to __m128i from __m128.
161 (_mm256_permute2_ps): Fix first cast of C argument to __m256i from
164 2020-03-05 Delia Burduv <delia.burduv@arm.com>
166 * config/arm/arm_neon.h (vbfmmlaq_f32): New.
167 (vbfmlalbq_f32): New.
168 (vbfmlaltq_f32): New.
169 (vbfmlalbq_lane_f32): New.
170 (vbfmlaltq_lane_f32): New.
171 (vbfmlalbq_laneq_f32): New.
172 (vbfmlaltq_laneq_f32): New.
173 * config/arm/arm_neon_builtins.def (vmmla): New.
180 * config/arm/iterators.md (BF_MA): New int iterator.
181 (bt): New int attribute.
182 (VQXBF): Copy of VQX with V8BF.
183 * config/arm/neon.md (neon_vmmlav8bf): New insn.
184 (neon_vfma<bt>v8bf): New insn.
185 (neon_vfma<bt>_lanev8bf): New insn.
186 (neon_vfma<bt>_laneqv8bf): New expand.
187 (neon_vget_high<mode>): Changed iterator to VQXBF.
188 * config/arm/unspecs.md (UNSPEC_BFMMLA): New UNSPEC.
189 (UNSPEC_BFMAB): New UNSPEC.
190 (UNSPEC_BFMAT): New UNSPEC.
192 2020-03-05 Jakub Jelinek <jakub@redhat.com>
195 * tree-pretty-print.h (pretty_print_string): Declare.
196 * tree-pretty-print.c (pretty_print_string): Remove forward
197 declaration, no longer static. Change nbytes parameter type
198 from unsigned to size_t.
199 * print-rtl.c (print_value) <case CONST_STRING>: Use
200 pretty_print_string and for shrink way too long strings.
202 2020-03-05 Richard Biener <rguenther@suse.de>
203 Jakub Jelinek <jakub@redhat.com>
205 PR tree-optimization/93582
206 * tree-ssa-sccvn.c (vn_reference_lookup_3): Treat POINTER_PLUS_EXPR
207 last operand as signed when looking for memset offset. Formatting
210 2020-03-04 Andrew Pinski <apinski@marvell.com>
213 * value-prof.c (dump_histogram_value): Use std::abs.
215 2020-03-04 Martin Sebor <msebor@redhat.com>
217 PR tree-optimization/93986
218 * tree-ssa-strlen.c (maybe_warn_overflow): Convert all wide_int
219 operands to the same precision widest_int to avoid ICEs.
221 2020-03-04 Bill Schmidt <wschmidt@linux.ibm.com>
224 * rs6000-cpus.def (OTHER_ALTIVEC_MASKS): New #define.
225 * rs6000.c (rs6000_disable_incompatible_switches): Add table entry
226 for OPTION_MASK_ALTIVEC.
228 2020-03-04 Andreas Krebbel <krebbel@linux.ibm.com>
230 * config.gcc: Include the glibc-stdint.h header for zTPF.
232 2020-03-04 Andreas Krebbel <krebbel@linux.ibm.com>
234 * config/s390/s390.c (s390_secondary_memory_needed): Disallow
235 direct FPR-GPR copies.
236 (s390_register_info_gprtofpr): Disallow GPR content to be saved in
239 2020-03-04 Andreas Krebbel <krebbel@linux.ibm.com>
241 * config/s390/s390.c (s390_emit_prologue): Specify the 2 new
242 operands to the prologue_tpf expander.
243 (s390_emit_epilogue): Likewise.
244 (s390_option_override_internal): Do error checking and setup for
246 * config/s390/tpf.h (TPF_TRACE_PROLOGUE_CHECK)
247 (TPF_TRACE_EPILOGUE_CHECK, TPF_TRACE_PROLOGUE_TARGET)
248 (TPF_TRACE_EPILOGUE_TARGET, TPF_TRACE_PROLOGUE_SKIP_TARGET)
249 (TPF_TRACE_EPILOGUE_SKIP_TARGET): New macro definitions.
250 * config/s390/tpf.md ("prologue_tpf", "epilogue_tpf"): Add two new
251 operands for the check flag and the branch target.
252 * config/s390/tpf.opt ("mtpf-trace-hook-prologue-check")
253 ("mtpf-trace-hook-prologue-target")
254 ("mtpf-trace-hook-epilogue-check")
255 ("mtpf-trace-hook-epilogue-target", "mtpf-trace-skip"): New
257 * doc/invoke.texi: Document -mtpf-trace-skip option. The other
258 options are for debugging purposes and will not be documented
261 2020-03-04 Jakub Jelinek <jakub@redhat.com>
264 * tree-inline.c (copy_decl_to_var): Copy DECL_BY_REFERENCE flag.
266 * tree-ssa-sccvn.c (vn_walk_cb_data::push_partial_def): Add offseti
267 argument. Change pd argument so that it can be modified. Turn
268 constant non-CONSTRUCTOR store into non-constant if it is too large.
269 Adjust offset and size of CONSTRUCTOR or non-constant store to avoid
271 (vn_walk_cb_data::vn_walk_cb_data, vn_reference_lookup_3): Adjust
274 2020-02-04 Richard Biener <rguenther@suse.de>
276 PR tree-optimization/93964
277 * graphite-isl-ast-to-gimple.c
278 (gcc_expression_from_isl_ast_expr_id): Add intermediate
279 conversion for pointer to integer converts.
280 * graphite-scop-detection.c (assign_parameter_index_in_region):
283 2020-03-04 Martin Liska <mliska@suse.cz>
287 * doc/invoke.texi: Clarify --help=language and --help=common
290 2020-03-04 Jakub Jelinek <jakub@redhat.com>
292 PR tree-optimization/94001
293 * tree-tailcall.c (process_assignment): Before comparing op1 to
294 *ass_var, verify *ass_var is non-NULL.
296 2020-03-04 Kito Cheng <kito.cheng@sifive.com>
299 * config/riscv/riscv.c (riscv_emit_float_compare): Using NE to compare
302 2020-03-03 Dennis Zhang <dennis.zhang@arm.com>
304 * config/arm/arm_bf16.h (vcvtah_f32_bf16, vcvth_bf16_f32): New.
305 * config/arm/arm_neon.h (vcvt_f32_bf16, vcvtq_low_f32_bf16): New.
306 (vcvtq_high_f32_bf16, vcvt_bf16_f32): New.
307 (vcvtq_low_bf16_f32, vcvtq_high_bf16_f32): New.
308 * config/arm/arm_neon_builtins.def (vbfcvt, vbfcvt_high): New entries.
309 (vbfcvtv4sf, vbfcvtv4sf_high): Likewise.
310 * config/arm/iterators.md (VBFCVT, VBFCVTM): New mode iterators.
311 (V_bf_low, V_bf_cvt_m): New mode attributes.
312 * config/arm/neon.md (neon_vbfcvtv4sf<VBFCVT:mode>): New.
313 (neon_vbfcvtv4sf_highv8bf, neon_vbfcvtsf): New.
314 (neon_vbfcvt<VBFCVT:mode>, neon_vbfcvt_highv8bf): New.
315 (neon_vbfcvtbf_cvtmode<mode>, neon_vbfcvtbf): New
316 * config/arm/unspecs.md (UNSPEC_BFCVT, UNSPEC_BFCVT_HIG): New.
318 2020-03-03 Jakub Jelinek <jakub@redhat.com>
320 PR tree-optimization/93582
321 * tree-ssa-sccvn.h (vn_reference_lookup): Add mask argument.
322 * tree-ssa-sccvn.c (struct vn_walk_cb_data): Add mask and masked_result
323 members, initialize them in the constructor and if mask is non-NULL,
324 artificially push_partial_def {} for the portions of the mask that
326 (vn_walk_cb_data::finish): If mask is non-NULL, set masked_result to
327 val and return (void *)-1. Formatting fix.
328 (vn_reference_lookup_pieces): Adjust vn_walk_cb_data initialization.
330 (vn_reference_lookup): Add mask argument. If non-NULL, don't call
331 fully_constant_vn_reference_p nor vn_reference_lookup_1 and return
333 (visit_nary_op): Handle BIT_AND_EXPR of a memory load and INTEGER_CST
335 (visit_stmt): Formatting fix.
337 2020-03-03 Richard Biener <rguenther@suse.de>
339 PR tree-optimization/93946
340 * alias.h (refs_same_for_tbaa_p): Declare.
341 * alias.c (refs_same_for_tbaa_p): New function.
342 * tree-ssa-alias.c (ao_ref_alias_set): For a NULL ref return
344 * tree-ssa-scopedtables.h
345 (avail_exprs_stack::lookup_avail_expr): Add output argument
346 giving access to the hashtable entry.
347 * tree-ssa-scopedtables.c (avail_exprs_stack::lookup_avail_expr):
349 * tree-ssa-dom.c: Include alias.h.
350 (dom_opt_dom_walker::optimize_stmt): Validate TBAA state before
351 removing redundant store.
352 * tree-ssa-sccvn.h (vn_reference_s::base_set): New member.
353 (ao_ref_init_from_vn_reference): Adjust prototype.
354 (vn_reference_lookup_pieces): Likewise.
355 (vn_reference_insert_pieces): Likewise.
356 * tree-ssa-sccvn.c: Track base alias set in addition to alias
358 (eliminate_dom_walker::eliminate_stmt): Also check base alias
359 set when removing redundant stores.
360 (visit_reference_op_store): Likewise.
361 * dse.c (record_store): Adjust valdity check for redundant
364 2020-03-03 Jakub Jelinek <jakub@redhat.com>
367 * config/s390/s390.h (OPTION_DEFAULT_SPECS): Reorder.
369 PR rtl-optimization/94002
370 * explow.c (plus_constant): Punt if cst has VOIDmode and
371 get_pool_mode is different from mode.
373 2020-03-03 Claudiu Zissulescu <claziss@synopsys.com>
375 * config/arc/arc.c (leigitimate_small_data_address_p): Check if an
376 address has an offset which fits the scalling constraint for a
377 load/store operation.
378 (legitimate_scaled_address_p): Update use
379 leigitimate_small_data_address_p.
380 (arc_print_operand): Likewise.
381 (arc_legitimate_address_p): Likewise.
382 (legitimate_small_data_address_p): Likewise.
384 2020-03-03 Claudiu Zissulescu <claziss@synopsys.com>
386 * config/arc/arc.md (fmasf4_fpu): Use accl_operand predicate.
387 (fnmasf4_fpu): Likewise.
389 2020-03-03 Claudiu Zissulescu <claziss@synopsys.com>
391 * config/arc/arc.md (adddi3): Early expand the 64bit operation into
394 (adddi3_i): Remove pattern.
395 (subdi3_i): Likewise.
397 2020-03-03 Claudiu Zissulescu <claziss@synopsys.com>
399 * config/arc/arc.md (eh_return): Add length info.
401 2020-03-02 David Malcolm <dmalcolm@redhat.com>
403 * doc/invoke.texi (-fanalyzer-show-duplicate-count): New.
405 2020-03-02 David Malcolm <dmalcolm@redhat.com>
407 * doc/invoke.texi (Static Analyzer Options): Add
408 -Wanalyzer-stale-setjmp-buffer to the list of options enabled
411 2020-03-02 Uroš Bizjak <ubizjak@gmail.com>
414 * config/i386/i386.md (movstrict<mode>): Allow only
415 registers with VALID_INT_MODE_P modes.
417 2020-03-02 Andrew Stubbs <ams@codesourcery.com>
419 * config/gcn/gcn-valu.md (dpp_move<mode>): New.
420 (reduc_insn): Use 'U' and 'B' operand codes.
421 (reduc_<reduc_op>_scal_<mode>): Allow all types.
422 (reduc_<reduc_op>_scal_v64di): Delete.
423 (*<reduc_op>_dpp_shr_<mode>): Allow all 1reg types.
424 (*plus_carry_dpp_shr_v64si): Change to ...
425 (*plus_carry_dpp_shr_<mode>): ... this and allow all 1reg int types.
426 (mov_from_lane63_v64di): Change to ...
427 (mov_from_lane63_<mode>): ... this, and allow all 64-bit modes.
428 * config/gcn/gcn.c (gcn_expand_dpp_shr_insn): Increase buffer size.
429 Support UNSPEC_MOV_DPP_SHR output formats.
430 (gcn_expand_reduc_scalar): Add "use_moves" reductions.
431 Add "use_extends" reductions.
432 (print_operand_address): Add 'I' and 'U' codes.
433 * config/gcn/gcn.md (unspec): Add UNSPEC_MOV_DPP_SHR.
435 2020-03-02 Martin Liska <mliska@suse.cz>
437 * lto-wrapper.c: Fix typo in comment about
438 C++ standard version.
440 2020-03-01 Martin Sebor <msebor@redhat.com>
443 * calls.c (init_attr_rdwr_indices): Correctly handle attribute.
445 2020-03-01 Martin Sebor <msebor@redhat.com>
448 * tree-ssa-strlen.c (count_nonzero_bytes): Set the size to that
449 of a pointer in the outermost ADDR_EXPRs.
451 2020-02-28 Jeff Law <law@redhat.com>
453 * config/v850/v850.h (STATIC_CHAIN_REGNUM): Change to r19.
454 * config/v850/v850.c (v850_asm_trampoline_template): Update
457 2020-02-28 Michael Meissner <meissner@linux.ibm.com>
460 * config/rs6000/vsx.md (vsx_extract_<mode>_<VS_scalar>mode_var):
463 2020-02-28 Martin Liska <mliska@suse.cz>
466 * configure.ac: Improve detection of ld_date by requiring
467 either two dashes or none.
468 * configure: Regenerate.
470 2020-02-28 Vladimir Makarov <vmakarov@redhat.com>
472 PR rtl-optimization/93564
473 * ira-color.c (assign_hard_reg): Prefer smaller hard regno when we
474 do not honor reg alloc order.
476 2020-02-27 Joel Hutton <Joel.Hutton@arm.com>
479 * config/aarch64/aarch64.c (aarch64_override_options): Fix
480 misleading warning string.
482 2020-02-27 Martin Sebor <msebor@redhat.com>
484 * doc/invoke.texi (-Wbuiltin-declaration-mismatch): Fix a typo.
486 2020-02-27 Michael Meissner <meissner@linux.ibm.com>
489 * config/rs6000/vsx.md (vsx_extract_<mode>_var, VSX_D iterator):
490 Split the insn into two parts. This insn only does variable
491 extract from a register.
492 (vsx_extract_<mode>_var_load, VSX_D iterator): New insn, do
493 variable extract from memory.
494 (vsx_extract_v4sf_var): Split the insn into two parts. This insn
495 only does variable extract from a register.
496 (vsx_extract_v4sf_var_load): New insn, do variable extract from
498 (vsx_extract_<mode>_var, VSX_EXTRACT_I iterator): Split the insn
499 into two parts. This insn only does variable extract from a
501 (vsx_extract_<mode>_var_load, VSX_EXTRACT_I iterator): New insn,
502 do variable extract from memory.
504 2020-02-27 Martin Jambor <mjambor@suse.cz>
505 Feng Xue <fxue@os.amperecomputing.com>
508 * ipa-cp.c (same_node_or_its_all_contexts_clone_p): Replaced with
509 new function calls_same_node_or_its_all_contexts_clone_p.
510 (cgraph_edge_brings_value_p): Use it.
511 (cgraph_edge_brings_value_p): Likewise.
512 (self_recursive_pass_through_p): Return false if caller is a clone.
513 (self_recursive_agg_pass_through_p): Likewise.
515 2020-02-27 Jan Hubicka <hubicka@ucw.cz>
518 * alias.c (ends_tbaa_access_path_p): Break out from ...
519 (component_uses_parent_alias_set_from): ... here.
520 * alias.h (ends_tbaa_access_path_p): Declare.
521 * tree-ssa-alias.c (access_path_may_continue_p): Break out from ...;
522 handle trailing arrays past end of tbaa access path.
523 (aliasing_component_refs_p): ... here; likewise.
524 (nonoverlapping_refs_since_match_p): Track TBAA segment of the access
525 path; disambiguate also past end of it.
526 (nonoverlapping_component_refs_p): Use only TBAA segment of the access
529 2020-02-27 Mihail Ionescu <mihail.ionescu@arm.com>
531 * (__ARM_NUM_LANES, __arm_lane, __arm_lane_q): Move to the
532 beginning of the file.
533 (vcreate_bf16, vcombine_bf16): New.
534 (vdup_n_bf16, vdupq_n_bf16): New.
535 (vdup_lane_bf16, vdup_laneq_bf16): New.
536 (vdupq_lane_bf16, vdupq_laneq_bf16): New.
537 (vduph_lane_bf16, vduph_laneq_bf16): New.
538 (vset_lane_bf16, vsetq_lane_bf16): New.
539 (vget_lane_bf16, vgetq_lane_bf16): New.
540 (vget_high_bf16, vget_low_bf16): New.
541 (vreinterpret_bf16_u8, vreinterpretq_bf16_u8): New.
542 (vreinterpret_bf16_u16, vreinterpretq_bf16_u16): New.
543 (vreinterpret_bf16_u32, vreinterpretq_bf16_u32): New.
544 (vreinterpret_bf16_u64, vreinterpretq_bf16_u64): New.
545 (vreinterpret_bf16_s8, vreinterpretq_bf16_s8): New.
546 (vreinterpret_bf16_s16, vreinterpretq_bf16_s16): New.
547 (vreinterpret_bf16_s32, vreinterpretq_bf16_s32): New.
548 (vreinterpret_bf16_s64, vreinterpretq_bf16_s64): New.
549 (vreinterpret_bf16_p8, vreinterpretq_bf16_p8): New.
550 (vreinterpret_bf16_p16, vreinterpretq_bf16_p16): New.
551 (vreinterpret_bf16_p64, vreinterpretq_bf16_p64): New.
552 (vreinterpret_bf16_f32, vreinterpretq_bf16_f32): New.
553 (vreinterpret_bf16_f64, vreinterpretq_bf16_f64): New.
554 (vreinterpretq_bf16_p128): New.
555 (vreinterpret_s8_bf16, vreinterpretq_s8_bf16): New.
556 (vreinterpret_s16_bf16, vreinterpretq_s16_bf16): New.
557 (vreinterpret_s32_bf16, vreinterpretq_s32_bf16): New.
558 (vreinterpret_s64_bf16, vreinterpretq_s64_bf16): New.
559 (vreinterpret_u8_bf16, vreinterpretq_u8_bf16): New.
560 (vreinterpret_u16_bf16, vreinterpretq_u16_bf16): New.
561 (vreinterpret_u32_bf16, vreinterpretq_u32_bf16): New.
562 (vreinterpret_u64_bf16, vreinterpretq_u64_bf16): New.
563 (vreinterpret_p8_bf16, vreinterpretq_p8_bf16): New.
564 (vreinterpret_p16_bf16, vreinterpretq_p16_bf16): New.
565 (vreinterpret_p64_bf16, vreinterpretq_p64_bf16): New.
566 (vreinterpret_f32_bf16, vreinterpretq_f32_bf16): New.
567 (vreinterpretq_p128_bf16): New.
568 * config/arm/arm_neon_builtins.def (VDX): Add V4BF.
570 (V_elem_l): Likewise.
573 (V_DOUBLE): Likewise.
574 (VDQX): Add V4BF and V8BF.
575 (V_two_elem, V_three_elem, V_four_elem): Likewise.
578 (V_double_vector_mode): Likewise.
579 (V_cmp_result): Likewise.
580 (V_uf_sclr): Likewise.
581 (V_sz_elem): Likewise.
582 (Is_d_reg): Likewise.
583 (V_mode_nunits): Likewise.
584 * config/arm/neon.md (neon_vdup_lane): Enable for BFloat16.
586 2020-02-27 Andrew Stubbs <ams@codesourcery.com>
588 * config/gcn/gcn-valu.md (VEC_SUBDWORD_MODE): New mode iterator.
589 (<expander><mode>2<exec>): Change modes to VEC_ALL1REG_INT_MODE.
590 (<expander><mode>3<exec>): Likewise.
591 (<expander><mode>3): New.
592 (v<expander><mode>3): New.
593 (<expander><mode>3): New.
594 (<expander><mode>3<exec>): Rename to ...
595 (<expander>v64si3<exec>): ... this, and change modes to V64SI.
596 * config/gcn/gcn.md (mnemonic): Use '%B' for not.
598 2020-02-27 Alexandre Oliva <oliva@adacore.com>
600 * config/vx-common.h (NO_DOLLAR_IN_LABEL, NO_DOT_IN_LABEL): Leave
603 2020-02-27 Richard Biener <rguenther@suse.de>
605 PR tree-optimization/93508
606 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle _CHK like
607 non-_CHK variants. Valueize their length arguments.
609 2020-02-27 Richard Biener <rguenther@suse.de>
611 PR tree-optimization/93953
612 * tree-vect-slp.c (slp_copy_subtree): Avoid keeping a reference
613 to the hash-map entry.
615 2020-02-27 Andrew Stubbs <ams@codesourcery.com>
617 * config/gcn/gcn.md (mov<mode>): Add transformations for BI subregs.
619 2020-02-27 Mark Williams <mwilliams@fb.com>
621 * dwarf2out.c (file_name_acquire): Call remap_debug_filename.
622 * lto-opts.c (lto_write_options): Drop -fdebug-prefix-map,
623 -ffile-prefix-map and -fmacro-prefix-map.
624 * lto-streamer-out.c: Include file-prefix-map.h.
625 (lto_output_location): Remap the file part of locations.
627 2020-02-27 Jakub Jelinek <jakub@redhat.com>
630 * gimplify.c (gimplify_init_constructor): Don't promote readonly
631 DECL_REGISTER variables to TREE_STATIC.
633 PR tree-optimization/93582
634 PR tree-optimization/93945
635 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle memset with
636 non-zero INTEGER_CST second argument and ref->offset or ref->size
637 not a multiple of BITS_PER_UNIT.
639 2020-02-27 Jonathan Wakely <jwakely@redhat.com>
641 * doc/install.texi (Binaries): Update description of BullFreeware.
643 2020-02-26 Sandra Loosemore <sandra@codesourcery.com>
647 * doc/invoke.texi (Option Summary): Re-alphabetize warnings in
648 C++ Language Options, Warning Options, and Static Analyzer
649 Options lists. Document negative form of options enabled by
650 default. Move some things around to more accurately sort
651 warnings by category.
652 (C++ Dialect Options, Warning Options, Static Analyzer
653 Options): Document negative form of options when enabled by
654 default. Move some things around to more accurately sort
655 warnings by category. Add some missing index entries.
658 2020-02-26 Carl Love <cel@us.ibm.com>
661 * doc/extend.texi (PowerPC AltiVec Built-in Functions available on
662 ISA 2.07): The builtin-function name __builtin_crypto_vpmsumb is only
663 for the vector unsigned short arguments. It is also listed as the
664 name of the built-in for arguments vector unsigned short,
665 vector unsigned int and vector unsigned long long built-ins. The
666 name of the builtins for these arguments should be:
667 __builtin_crypto_vpmsumh, __builtin_crypto_vpmsumw and
668 __builtin_crypto_vpmsumd respectively.
670 2020-02-26 Richard Biener <rguenther@suse.de>
672 * tree-vect-slp.c (vect_print_slp_tree): Also dump ref count
673 and load permutation.
675 2020-02-26 Richard Sandiford <richard.sandiford@arm.com>
678 * optabs-tree.c (supportable_convert_operation): Reject types with
681 2020-02-26 David Malcolm <dmalcolm@redhat.com>
683 * Makefile.in (ANALYZER_OBJS): Add analyzer/bar-chart.o.
685 2020-02-26 Jakub Jelinek <jakub@redhat.com>
687 PR tree-optimization/93820
688 * gimple-ssa-store-merging.c (check_no_overlap): Change RHS_CODE
689 argument to ALL_INTEGER_CST_P boolean.
690 (imm_store_chain_info::try_coalesce_bswap): Adjust caller.
691 (imm_store_chain_info::coalesce_immediate_stores): Likewise. Handle
692 adjacent INTEGER_CST store into merged_store->only_constants like
695 2020-02-25 Jakub Jelinek <jakub@redhat.com>
698 * config/sh/sh.c (expand_cbranchdi4): Fix comment typo, probablity
700 * cfghooks.c (verify_flow_info): Likewise.
701 * predict.c (combine_predictions_for_bb): Likewise.
702 * bb-reorder.c (connect_better_edge_p): Likewise. Fix comment typo,
703 sucessor -> successor.
704 (find_traces_1_round): Fix comment typo, destinarion -> destination.
705 * omp-expand.c (expand_oacc_for): Fix comment typo, sucessors ->
707 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): Fix dump
708 message typo, sucessors -> successors.
710 2020-02-25 Martin Sebor <msebor@redhat.com>
712 * doc/extend.texi (attribute access): Correct an example.
714 2020-02-25 Mihail Ionescu <mihail.ionescu@arm.com>
716 * config/aarch64/aarch64-builtins.c (aarch64_scalar_builtin_types):
718 (aarch64_init_simd_builtin_scalar_types): Register simd_bf.
720 * config/aarch64/iterators.md (VALLDIF): Enable for V4BF and V8BF.
721 (VD): Enable for V4BF.
723 (VQ): Enable for V8BF.
726 (VDBL, Vdbl): Add V4BF.
727 (V_INT_EQUIV, v_int_equiv): Add V4BF and V8BF.
728 * config/aarch64/arm_neon.h (bfloat16x4x2_t): New typedef.
729 (bfloat16x8x2_t): Likewise.
730 (bfloat16x4x3_t): Likewise.
731 (bfloat16x8x3_t): Likewise.
732 (bfloat16x4x4_t): Likewise.
733 (bfloat16x8x4_t): Likewise.
734 (vcombine_bf16): New.
735 (vld1_bf16, vld1_bf16_x2): New.
736 (vld1_bf16_x3, vld1_bf16_x4): New.
737 (vld1q_bf16, vld1q_bf16_x2): New.
738 (vld1q_bf16_x3, vld1q_bf16_x4): New.
739 (vld1_lane_bf16): New.
740 (vld1q_lane_bf16): New.
741 (vld1_dup_bf16): New.
742 (vld1q_dup_bf16): New.
745 (vld2_dup_bf16): New.
746 (vld2q_dup_bf16): New.
749 (vld3_dup_bf16): New.
750 (vld3q_dup_bf16): New.
753 (vld4_dup_bf16): New.
754 (vld4q_dup_bf16): New.
755 (vst1_bf16, vst1_bf16_x2): New.
756 (vst1_bf16_x3, vst1_bf16_x4): New.
757 (vst1q_bf16, vst1q_bf16_x2): New.
758 (vst1q_bf16_x3, vst1q_bf16_x4): New.
759 (vst1_lane_bf16): New.
760 (vst1q_lane_bf16): New.
768 2020-02-25 Mihail Ionescu <mihail.ionescu@arm.com>
770 * config/aarch64/iterators.md (VDQF_F16) Add V4BF and V8BF.
771 (VALL_F16): Likewise.
772 (VALLDI_F16): Likewise.
775 (vswap_width_name): Likewise.
776 (VSWAP_WIDTH): Likewise.
780 * config/aarch64/arm_neon.h (vset_lane_bf16, vsetq_lane_bf16): New.
781 (vget_lane_bf16, vgetq_lane_bf16): New.
783 (vdup_n_bf16, vdupq_n_bf16): New.
784 (vdup_lane_bf16, vdup_laneq_bf16): New.
785 (vdupq_lane_bf16, vdupq_laneq_bf16): New.
786 (vduph_lane_bf16, vduph_laneq_bf16): New.
787 (vreinterpret_bf16_u8, vreinterpretq_bf16_u8): New.
788 (vreinterpret_bf16_u16, vreinterpretq_bf16_u16): New.
789 (vreinterpret_bf16_u32, vreinterpretq_bf16_u32): New.
790 (vreinterpret_bf16_u64, vreinterpretq_bf16_u64): New.
791 (vreinterpret_bf16_s8, vreinterpretq_bf16_s8): New.
792 (vreinterpret_bf16_s16, vreinterpretq_bf16_s16): New.
793 (vreinterpret_bf16_s32, vreinterpretq_bf16_s32): New.
794 (vreinterpret_bf16_s64, vreinterpretq_bf16_s64): New.
795 (vreinterpret_bf16_p8, vreinterpretq_bf16_p8): New.
796 (vreinterpret_bf16_p16, vreinterpretq_bf16_p16): New.
797 (vreinterpret_bf16_p64, vreinterpretq_bf16_p64): New
798 (vreinterpret_bf16_f16, vreinterpretq_bf16_f16): New
799 (vreinterpret_bf16_f32, vreinterpretq_bf16_f32): New.
800 (vreinterpret_bf16_f64, vreinterpretq_bf16_f64): New.
801 (vreinterpretq_bf16_p128): New.
802 (vreinterpret_s8_bf16, vreinterpretq_s8_bf16): New.
803 (vreinterpret_s16_bf16, vreinterpretq_s16_bf16): New.
804 (vreinterpret_s32_bf16, vreinterpretq_s32_bf16): New.
805 (vreinterpret_s64_bf16, vreinterpretq_s64_bf16): New.
806 (vreinterpret_u8_bf16, vreinterpretq_u8_bf16): New.
807 (vreinterpret_u16_bf16, vreinterpretq_u16_bf16): New.
808 (vreinterpret_u32_bf16, vreinterpretq_u32_bf16): New.
809 (vreinterpret_u64_bf16, vreinterpretq_u64_bf16): New.
810 (vreinterpret_p8_bf16, vreinterpretq_p8_bf16): New.
811 (vreinterpret_p16_bf16, vreinterpretq_p16_bf16): New.
812 (vreinterpret_p64_bf16, vreinterpretq_p64_bf16): New.
813 (vreinterpret_f32_bf16, vreinterpretq_f32_bf16): New.
814 (vreinterpret_f64_bf16,vreinterpretq_f64_bf16): New.
815 (vreinterpret_f16_bf16,vreinterpretq_f16_bf16): New.
816 (vreinterpretq_p128_bf16): New.
818 2020-02-25 Dennis Zhang <dennis.zhang@arm.com>
820 * config/arm/arm_neon.h (vbfdot_f32, vbfdotq_f32): New
821 (vbfdot_lane_f32, vbfdotq_laneq_f32): New.
822 (vbfdot_laneq_f32, vbfdotq_lane_f32): New.
823 * config/arm/arm_neon_builtins.def (vbfdot): New entry.
824 (vbfdot_lanev4bf, vbfdot_lanev8bf): Likewise.
825 * config/arm/iterators.md (VSF2BF): New attribute.
826 * config/arm/neon.md (neon_vbfdot<VCVTF:mode>): New entry.
827 (neon_vbfdot_lanev4bf<VCVTF:mode>): Likewise.
828 (neon_vbfdot_lanev8bf<VCVTF:mode>): Likewise.
830 2020-02-25 Christophe Lyon <christophe.lyon@linaro.org>
832 * config/arm/arm.md (required_for_purecode): New attribute.
833 (enabled): Handle required_for_purecode.
834 * config/arm/thumb1.md (thumb1_movsi_insn): Add alternative to
835 work with -mpure-code.
837 2020-02-25 Jakub Jelinek <jakub@redhat.com>
839 PR rtl-optimization/93908
840 * combine.c (find_split_point): For store into ZERO_EXTRACT, and src
843 2019-02-25 Eric Botcazou <ebotcazou@adacore.com>
845 * dwarf2out.c (dwarf2out_size_function): Run in early-DWARF mode.
847 2020-02-25 Roman Zhuykov <zhroma@ispras.ru>
849 * doc/install.texi (--enable-checking): Adjust wording.
851 2020-02-25 Richard Biener <rguenther@suse.de>
853 PR tree-optimization/93868
854 * tree-vect-slp.c (slp_copy_subtree): New function.
855 (vect_attempt_slp_rearrange_stmts): Copy the SLP tree before
856 re-arranging stmts in it.
858 2020-02-25 Jakub Jelinek <jakub@redhat.com>
861 * passes.c (pass_manager::dump_passes): Create a cgraph node for the
862 dummy function and remove it at the end.
865 * config/lm32/lm32.c (lm32_setup_incoming_varargs): Fix comment typo
866 paramter -> parameter.
867 * config/aarch64/aarch64.c (aarch64_is_extend_from_extract): Likewise.
868 * ipa-prop.h (struct ipa_agg_replacement_value): Likewise.
870 2020-02-24 Roman Zhuykov <zhroma@ispras.ru>
872 * doc/install.texi (--enable-checking): Properly document current
874 (--enable-stage1-checking): Minor clarification about bootstrap.
876 2020-02-24 David Malcolm <dmalcolm@redhat.com>
879 * doc/invoke.texi (-Wnanalyzer-tainted-array-index): Note that
880 -fanalyzer-checker=taint is also required.
881 (-fanalyzer-checker=): Note that providing this option enables the
882 given checker, and doing so may be required for checkers that are
885 2020-02-24 David Malcolm <dmalcolm@redhat.com>
887 * doc/invoke.texi (-fanalyzer-verbosity=): "2" only shows
888 significant control flow events; add a "3" which shows all
889 control flow events; the old "3" becomes "4".
891 2020-02-24 Jakub Jelinek <jakub@redhat.com>
893 PR tree-optimization/93582
894 * tree-ssa-sccvn.c (vn_walk_cb_data::push_partial_def): Consider
895 pd.offset and pd.size to be counted in bits rather than bytes, add
896 support for maxsizei that is not a multiple of BITS_PER_UNIT and
897 handle bitfield stores and loads.
898 (vn_reference_lookup_3): Don't call ranges_known_overlap_p with
899 uncomparable quantities - bytes vs. bits. Allow push_partial_def
900 on offsets/sizes that aren't multiple of BITS_PER_UNIT and adjust
901 pd.offset/pd.size to be counted in bits rather than bytes.
902 Formatting fix. Rename shadowed len variable to buflen.
904 2020-02-24 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
905 Kugan Vivekandarajah <kugan.vivekanandarajah@linaro.org>
908 * gcc.c (putenv_COLLECT_AS_OPTIONS): New function.
909 (driver::main): Call putenv_COLLECT_AS_OPTIONS.
910 * opts-common.c (parse_options_from_collect_gcc_options): New function.
911 (prepend_xassembler_to_collect_as_options): Likewise.
912 * opts.h (parse_options_from_collect_gcc_options): Declare prototype.
913 (prepend_xassembler_to_collect_as_options): Likewise.
914 * lto-opts.c (lto_write_options): Stream assembler options
915 in COLLECT_AS_OPTIONS.
916 * lto-wrapper.c (xassembler_options_error): New static variable.
917 (get_options_from_collect_gcc_options): Move parsing options code to
918 parse_options_from_collect_gcc_options and call it.
919 (merge_and_complain): Validate -Xassembler options.
920 (append_compiler_options): Handle OPT_Xassembler.
921 (run_gcc): Append command line -Xassembler options to
923 * doc/invoke.texi: Add documentation about using Xassembler
926 2020-02-24 Kito Cheng <kito.cheng@sifive.com>
928 * config/riscv/riscv.c (riscv_emit_float_compare): Change the code gen
930 (riscv_rtx_costs): Update cost model for LTGT.
932 2020-02-23 Vladimir Makarov <vmakarov@redhat.com>
934 PR rtl-optimization/93564
935 * ira-color.c (struct update_cost_queue_elem): New member start.
936 (queue_update_cost, get_next_update_cost): Add new arg start.
937 (allocnos_conflict_p): New function.
938 (update_costs_from_allocno): Add new arg conflict_cost_update_p.
939 Add checking conflicts with allocnos_conflict_p.
940 (update_costs_from_prefs, restore_costs_from_copies): Adjust
941 update_costs_from_allocno calls.
942 (update_conflict_hard_regno_costs): Add checking conflicts with
943 allocnos_conflict_p. Adjust calls of queue_update_cost and
944 get_next_update_cost.
945 (assign_hard_reg): Adjust calls of queue_update_cost. Add
947 (bucket_allocno_compare_func): Restore previous version.
949 2020-02-21 John David Anglin <danglin@gcc.gnu.org>
951 * gcc/config/pa/pa.c (pa_function_value): Fix check for word and
952 double-word size when handling aggregate return values.
953 * gcc/config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Fix to indicate
954 that homogeneous SFmode and DFmode aggregates are passed and returned
955 in general registers.
957 2020-02-21 Jakub Jelinek <jakub@redhat.com>
960 * opts.c (print_filtered_help): Translate help before appending
961 messages to it rather than after that.
963 2020-02-19 Richard Sandiford <richard.sandiford@arm.com>
965 PR rtl-optimization/PR92989
966 * lra-lives.c (process_bb_lives): Restore the original order
967 of the bb liveness update. Call make_hard_regno_dead for each
968 register clobbered at the start of an EH receiver.
970 2020-02-18 Feng Xue <fxue@os.amperecomputing.com>
973 * ipa-cp.c (self_recursively_generated_p): Mark self-dependent value as
974 self-recursively generated.
976 2020-02-21 Iain Sandoe <iain@sandoe.co.uk>
979 * config/darwin-c.c (pop_field_alignment): Adjust quoting of
982 2020-02-21 Mihail Ionescu <mihail.ionescu@arm.com>
984 * doc/sourcebuild.texi (arm_v8_1m_mve_ok):
985 Document new target supports option.
987 2020-02-21 Dennis Zhang <dennis.zhang@arm.com>
989 * config/arm/arm_neon.h (vmmlaq_s32, vmmlaq_u32, vusmmlaq_s32): New.
990 * config/arm/arm_neon_builtins.def (smmla, ummla, usmmla): New.
991 * config/arm/iterators.md (MATMUL): New iterator.
992 (sup): Add UNSPEC_MATMUL_S, UNSPEC_MATMUL_U, and UNSPEC_MATMUL_US.
993 (mmla_sfx): New attribute.
994 * config/arm/neon.md (neon_<sup>mmlav16qi): New.
995 * config/arm/unspecs.md (UNSPEC_MATMUL_S, UNSPEC_MATMUL_U): New.
996 (UNSPEC_MATMUL_US): New.
998 2020-02-21 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1000 * config/arm/arm.md: Prevent scalar shifts from being used when big
1003 2020-02-21 Jan Hubicka <hubicka@ucw.cz>
1004 Richard Biener <rguenther@suse.de>
1006 PR tree-optimization/93586
1007 * tree-ssa-alias.c (nonoverlapping_array_refs_p): Finish array walk
1008 after mismatched array refs; do not sure type size information to
1009 recover from unmatched referneces with !flag_strict_aliasing_p.
1011 2020-02-21 Andrew Stubbs <ams@codesourcery.com>
1013 * config/gcn/gcn-valu.md (gather_load<mode>): Rename to ...
1014 (gather_load<mode>v64si): ... this and set operand 2 to V64SI.
1015 (scatter_store<mode>): Rename to ...
1016 (scatter_store<mode>v64si): ... this and set operand 1 to V64SI.
1017 (scatter<mode>_exec): Delete. Move contents ...
1018 (mask_scatter_store<mode>): ... here, and rename that to ...
1019 (mask_gather_load<mode>v64si): ... this. Set operand 2 to V64SI.
1020 Remove mode conversion.
1021 (mask_gather_load<mode>): Rename to ...
1022 (mask_scatter_store<mode>v64si): ... this. Set operand 1 to V64SI.
1023 Remove mode conversion.
1024 * config/gcn/gcn.c (gcn_expand_scaled_offsets): Remove mode conversion.
1026 2020-02-21 Martin Jambor <mjambor@suse.cz>
1028 PR tree-optimization/93845
1029 * tree-sra.c (verify_sra_access_forest): Only test access size of
1032 2020-02-21 Andrew Stubbs <ams@codesourcery.com>
1034 * config/gcn/gcn.c (gcn_hard_regno_mode_ok): Align VGPR pairs.
1035 * config/gcn/gcn-valu.md (addv64di3): Remove early-clobber.
1036 (addv64di3_exec): Likewise.
1037 (subv64di3): Likewise.
1038 (subv64di3_exec): Likewise.
1039 (addv64di3_zext): Likewise.
1040 (addv64di3_zext_exec): Likewise.
1041 (addv64di3_zext_dup): Likewise.
1042 (addv64di3_zext_dup_exec): Likewise.
1043 (addv64di3_zext_dup2): Likewise.
1044 (addv64di3_zext_dup2_exec): Likewise.
1045 (addv64di3_sext_dup2): Likewise.
1046 (addv64di3_sext_dup2_exec): Likewise.
1047 (<expander>v64di3): Likewise.
1048 (<expander>v64di3_exec): Likewise.
1049 (*<reduc_op>_dpp_shr_v64di): Likewise.
1050 (*plus_carry_dpp_shr_v64di): Likewise.
1051 * config/gcn/gcn.md (adddi3): Likewise.
1052 (addptrdi3): Likewise.
1053 (<expander>di3): Likewise.
1055 2020-02-21 Andrew Stubbs <ams@codesourcery.com>
1057 * config/gcn/gcn-valu.md (vec_seriesv64di): Use gen_vec_duplicatev64di.
1059 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
1061 * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Add SVE
1062 support. Use aarch64_emit_mult instead of emitting multiplication
1063 instructions directly.
1064 * config/aarch64/aarch64-sve.md (sqrt<mode>2, rsqrt<mode>2)
1065 (@aarch64_rsqrte<mode>, @aarch64_rsqrts<mode>): New expanders.
1067 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
1069 * config/aarch64/aarch64.c (aarch64_emit_mult): New function.
1070 (aarch64_emit_approx_div): Add SVE support. Use aarch64_emit_mult
1071 instead of emitting multiplication instructions directly.
1072 * config/aarch64/iterators.md (SVE_COND_FP_BINARY_OPTAB): New iterator.
1073 * config/aarch64/aarch64-sve.md (div<mode>3, @aarch64_frecpe<mode>)
1074 (@aarch64_frecps<mode>): New expanders.
1076 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
1078 * config/aarch64/aarch64-protos.h (AARCH64_APPROX_MODE): Operate
1079 on and produce uint64_ts rather than ints.
1080 (AARCH64_APPROX_NONE, AARCH64_APPROX_ALL): Change to uint64_ts.
1081 (cpu_approx_modes): Change the fields from unsigned int to uint64_t.
1083 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
1085 * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Don't create
1086 an unused xmsk register when handling approximate rsqrt.
1088 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
1090 * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Fix inverted
1091 flag_finite_math_only condition.
1093 2020-02-20 Uroš Bizjak <ubizjak@gmail.com>
1096 * config/i386/mmx.md (*vec_extractv2sf_1): Match source operand
1097 to destination operand for shufps alternative.
1098 (*vec_extractv2si_1): Ditto.
1100 2020-02-20 Peter Bergner <bergner@linux.ibm.com>
1103 * config/rs6000/rs6000.c (rs6000_legitimate_address_p): Handle VSX
1106 2020-02-20 Martin Liska <mliska@suse.cz>
1108 PR translation/93831
1109 * config/darwin.c (darwin_override_options): Change 64b to 64-bit mode.
1111 2020-02-20 Martin Liska <mliska@suse.cz>
1113 PR translation/93830
1114 * common/config/avr/avr-common.c: Remote trailing "|".
1116 2020-02-19 Bernd Edlinger <bernd.edlinger@hotmail.de>
1118 * collect2.c (maybe_run_lto_and_relink): Fix typo in
1121 2020-02-19 Richard Sandiford <richard.sandiford@arm.com>
1123 PR tree-optimization/93767
1124 * tree-vect-data-refs.c (vect_compile_time_alias): Remove the
1125 access-size bias from the offset calculations for negative strides.
1127 2020-02-19 Bernd Edlinger <bernd.edlinger@hotmail.de>
1129 * collect2.c (c_file, o_file): Make const again.
1130 (ldout,lderrout, dump_ld_file): Remove.
1131 (tool_cleanup): Avoid calling not signal-safe functions.
1132 (maybe_run_lto_and_relink): Avoid possible signal handler
1133 access to unintialzed memory (lto_o_files).
1134 (main): Avoid leaking temp files in $TMPDIR.
1135 Initialize c_file/o_file with concat, which avoids exposing
1136 uninitialized memory to signal handler, which calls unlink(!).
1137 Avoid calling maybe_unlink when the main function returns,
1138 since the atexit handler is already doing this.
1139 * collect2.h (dump_ld_file, ldout, lderrout): Remove.
1141 2020-02-19 Martin Jambor <mjambor@suse.cz>
1143 PR tree-optimization/93776
1144 * tree-sra.c (create_access): Do not create zero size accesses.
1145 (get_access_for_expr): Do not search for zero sized accesses.
1147 2020-02-19 Martin Jambor <mjambor@suse.cz>
1149 PR tree-optimization/93667
1150 * tree-sra.c (scalarizable_type_p): Return false if record fields
1151 do not follow wach other.
1153 2020-01-21 Kito Cheng <kito.cheng@sifive.com>
1155 * config/riscv/riscv.c (riscv_output_move) Using fmv.x.w/fmv.w.x
1156 rather than fmv.x.s/fmv.s.x.
1158 2020-02-18 James Greenhalgh <james.greenhalgh@arm.com>
1160 * config/aarch64/aarch64-simd-builtins.def
1161 (intrinsic_vec_smult_lo_): New.
1162 (intrinsic_vec_umult_lo_): Likewise.
1163 (vec_widen_smult_hi_): Likewise.
1164 (vec_widen_umult_hi_): Likewise.
1165 * config/aarch64/aarch64-simd.md
1166 (aarch64_intrinsic_vec_<su>mult_lo_<mode>): New.
1167 * config/aarch64/arm_neon.h (vmull_high_s8): Use intrinsics.
1168 (vmull_high_s16): Likewise.
1169 (vmull_high_s32): Likewise.
1170 (vmull_high_u8): Likewise.
1171 (vmull_high_u16): Likewise.
1172 (vmull_high_u32): Likewise.
1173 (vmull_s8): Likewise.
1174 (vmull_s16): Likewise.
1175 (vmull_s32): Likewise.
1176 (vmull_u8): Likewise.
1177 (vmull_u16): Likewise.
1178 (vmull_u32): Likewise.
1180 2020-02-18 Martin Liska <mliska@suse.cz>
1182 * value-prof.c (stream_out_histogram_value): Restore LTO PGO
1183 bootstrap by missing removal of invalid sanity check.
1185 2020-02-18 Martin Liska <mliska@suse.cz>
1188 * ipa-icf-gimple.c (func_checker::compare_gimple_assign):
1189 Always compare LHS of gimple_assign.
1191 2020-02-18 Martin Liska <mliska@suse.cz>
1194 * cgraph.c (cgraph_node::verify_node): Verify MALLOC attribute
1195 and return type of functions.
1196 * ipa-param-manipulation.c (ipa_param_adjustments::adjust_decl):
1197 Drop MALLOC attribute for void functions.
1198 * ipa-pure-const.c (funct_state_summary_t::duplicate): Drop
1199 malloc_state for a new VOID clone.
1201 2020-02-18 Martin Liska <mliska@suse.cz>
1204 * common.opt: Add -fprofile-reproducibility.
1205 * doc/invoke.texi: Document it.
1206 * value-prof.c (dump_histogram_value):
1207 Document and support behavior for counters[0]
1208 being a negative value.
1209 (get_nth_most_common_value): Handle negative
1210 counters[0] in respect to flag_profile_reproducible.
1212 2020-02-18 Jakub Jelinek <jakub@redhat.com>
1215 * cgraph.c (verify_speculative_call): Use speculative_id instead of
1216 speculative_uid in messages. Remove trailing whitespace from error
1217 message. Use num_speculative_call_targets instead of
1218 num_speculative_targets in a message.
1219 (cgraph_node::verify_node): Use call_stmt instead of cal_stmt in
1220 edge messages and stmt instead of cal_stmt in reference message.
1222 PR tree-optimization/93780
1223 * tree-ssa.c (non_rewritable_lvalue_p): Check valid_vector_subparts_p
1224 before calling build_vector_type.
1225 (execute_update_addresses_taken): Likewise.
1228 * params.opt (-param=ipa-max-switch-predicate-bounds=): Fix help
1229 typo, functoin -> function.
1230 * tree.c (free_lang_data_in_decl): Fix comment typo,
1231 functoin -> function.
1232 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
1234 2020-02-17 David Malcolm <dmalcolm@redhat.com>
1236 * diagnostic.c (print_any_cwe): Don't call get_cwe_url if URLs
1238 (print_option_information): Don't call get_option_url if URLs
1241 2020-02-17 Alexandre Oliva <oliva@adacore.com>
1243 * tree-emutls.c (new_emutls_decl, emutls_common_1): Complete
1244 handling of register_common-less targets.
1246 2020-02-17 Martin Liska <mliska@suse.cz>
1249 * ipa-devirt.c (odr_types_equivalent_p): Fix grammar.
1251 2020-02-17 Martin Liska <mliska@suse.cz>
1253 PR translation/93755
1254 * config/rs6000/rs6000.c (rs6000_option_override_internal):
1257 2020-02-17 Martin Liska <mliska@suse.cz>
1260 * config/rx/elf.opt: Fix typo.
1262 2020-02-17 Richard Biener <rguenther@suse.de>
1265 * opts-global.c (print_ignored_options): Use inform and
1268 2020-02-17 Jiufu Guo <guojiufu@linux.ibm.com>
1271 * config/rs6000/rs6000.md (untyped_call): Add emit_clobber.
1273 2020-02-16 Uroš Bizjak <ubizjak@gmail.com>
1276 * config/i386/i386.md (atan2xf3): Swap operands 1 and 2.
1277 (atan2<mode>3): Update operand order in the call to gen_atan2xf3.
1279 2020-02-15 Jason Merrill <jason@redhat.com>
1281 * doc/invoke.texi (C Dialect Options): Add -std=c++20.
1283 2020-02-15 Jakub Jelinek <jakub@redhat.com>
1285 PR tree-optimization/93744
1286 * match.pd (((m1 >/</>=/<= m2) * d -> (m1 >/</>=/<= m2) ? d : 0,
1287 A - ((A - B) & -(C cmp D)) -> (C cmp D) ? B : A,
1288 A + ((B - A) & -(C cmp D)) -> (C cmp D) ? B : A): For GENERIC, make
1289 sure @2 in the first and @1 in the other patterns has no side-effects.
1291 2020-02-15 David Malcolm <dmalcolm@redhat.com>
1292 Bernd Edlinger <bernd.edlinger@hotmail.de>
1296 * config.in (DIAGNOSTICS_URLS_DEFAULT): New define.
1297 * configure.ac (--with-diagnostics-urls): New configuration
1298 option, based on --with-diagnostics-color.
1299 (DIAGNOSTICS_URLS_DEFAULT): New define.
1300 * config.h: Regenerate.
1301 * configure: Regenerate.
1302 * diagnostic.c (diagnostic_urls_init): Handle -1 for
1303 DIAGNOSTICS_URLS_DEFAULT from configure-time
1304 --with-diagnostics-urls=auto-if-env by querying for a GCC_URLS
1305 and TERM_URLS environment variable.
1306 * diagnostic-url.h (diagnostic_url_format): New enum type.
1307 (diagnostic_urls_enabled_p): rename to...
1308 (determine_url_format): ... this, and change return type.
1309 * diagnostic-color.c (parse_env_vars_for_urls): New helper function.
1310 (auto_enable_urls): Disable URLs on xfce4-terminal, gnome-terminal,
1311 the linux console, and mingw.
1312 (diagnostic_urls_enabled_p): rename to...
1313 (determine_url_format): ... this, and adjust.
1314 * pretty-print.h (pretty_printer::show_urls): rename to...
1315 (pretty_printer::url_format): ... this, and change to enum.
1316 * pretty-print.c (pretty_printer::pretty_printer,
1317 pp_begin_url, pp_end_url, test_urls): Adjust.
1318 * doc/install.texi (--with-diagnostics-urls): Document the new
1319 configuration option.
1320 (--with-diagnostics-color): Document the existing interaction
1321 with GCC_COLORS better.
1322 * doc/invoke.texi (-fdiagnostics-urls): Add GCC_URLS and TERM_URLS
1323 vindex reference. Update description of defaults based on the above.
1324 (-fdiagnostics-color): Update description of how -fdiagnostics-color
1325 interacts with GCC_COLORS.
1327 2020-02-14 Eric Botcazou <ebotcazou@adacore.com>
1330 * config/sparc/sparc.c (eligible_for_call_delay): Test HAVE_GNU_LD in
1331 conjunction with TARGET_GNU_TLS in early return.
1333 2020-02-14 Alexander Monakov <amonakov@ispras.ru>
1335 * rtlanal.c (rtx_cost): Handle a SET up front. Avoid division if
1336 the mode is not wider than UNITS_PER_WORD.
1338 2020-02-14 Martin Jambor <mjambor@suse.cz>
1340 PR tree-optimization/93516
1341 * tree-sra.c (propagate_subaccesses_from_rhs): Do not create
1342 access of the same type as the parent.
1343 (propagate_subaccesses_from_lhs): Likewise.
1345 2020-02-14 Hongtao Liu <hongtao.liu@intel.com>
1348 * config/i386/avx512vbmi2intrin.h
1349 (_mm512_shrdi_epi16, _mm512_mask_shrdi_epi16,
1350 _mm512_maskz_shrdi_epi16, _mm512_shrdi_epi32,
1351 _mm512_mask_shrdi_epi32, _mm512_maskz_shrdi_epi32,
1352 _m512_shrdi_epi64, _m512_mask_shrdi_epi64,
1353 _m512_maskz_shrdi_epi64, _mm512_shldi_epi16,
1354 _mm512_mask_shldi_epi16, _mm512_maskz_shldi_epi16,
1355 _mm512_shldi_epi32, _mm512_mask_shldi_epi32,
1356 _mm512_maskz_shldi_epi32, _mm512_shldi_epi64,
1357 _mm512_mask_shldi_epi64, _mm512_maskz_shldi_epi64): Fix typo
1358 of lacking a closing parenthesis.
1359 * config/i386/avx512vbmi2vlintrin.h
1360 (_mm256_shrdi_epi16, _mm256_mask_shrdi_epi16,
1361 _mm256_maskz_shrdi_epi16, _mm256_shrdi_epi32,
1362 _mm256_mask_shrdi_epi32, _mm256_maskz_shrdi_epi32,
1363 _m256_shrdi_epi64, _m256_mask_shrdi_epi64,
1364 _m256_maskz_shrdi_epi64, _mm256_shldi_epi16,
1365 _mm256_mask_shldi_epi16, _mm256_maskz_shldi_epi16,
1366 _mm256_shldi_epi32, _mm256_mask_shldi_epi32,
1367 _mm256_maskz_shldi_epi32, _mm256_shldi_epi64,
1368 _mm256_mask_shldi_epi64, _mm256_maskz_shldi_epi64,
1369 _mm_shrdi_epi16, _mm_mask_shrdi_epi16,
1370 _mm_maskz_shrdi_epi16, _mm_shrdi_epi32,
1371 _mm_mask_shrdi_epi32, _mm_maskz_shrdi_epi32,
1372 _mm_shrdi_epi64, _mm_mask_shrdi_epi64,
1373 _m_maskz_shrdi_epi64, _mm_shldi_epi16,
1374 _mm_mask_shldi_epi16, _mm_maskz_shldi_epi16,
1375 _mm_shldi_epi32, _mm_mask_shldi_epi32,
1376 _mm_maskz_shldi_epi32, _mm_shldi_epi64,
1377 _mm_mask_shldi_epi64, _mm_maskz_shldi_epi64): Ditto.
1379 2020-02-13 H.J. Lu <hongjiu.lu@intel.com>
1382 * config/i386/i386.c (ix86_trampoline_init): Skip ENDBR32 at
1383 the target function entry.
1385 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
1387 * common/config/arc/arc-common.c (arc_option_optimization_table):
1388 Disable if-conversion step when optimized for size.
1390 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
1392 * config/arc/arc.c (arc_conditional_register_usage): R0-R3 and
1393 R12-R15 are always in ARCOMPACT16_REGS register class.
1394 * config/arc/arc.opt (mq-class): Deprecate.
1395 * config/arc/constraint.md ("q"): Remove dependency on mq-class
1397 * doc/invoke.texi (mq-class): Update text.
1398 * common/config/arc/arc-common.c (arc_option_optimization_table):
1401 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
1403 * config/arc/arc.c (arc_insn_cost): New function.
1404 (TARGET_INSN_COST): Define.
1405 * config/arc/arc.md (cost): New attribute.
1406 (add_n): Use arc_nonmemory_operand.
1407 (ashlsi3_insn): Likewise, also update constraints.
1408 (ashrsi3_insn): Likewise.
1409 (rotrsi3): Likewise.
1410 (add_shift): Likewise.
1411 * config/arc/predicates.md (arc_nonmemory_operand): New predicate.
1413 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
1415 * config/arc/arc.md (mulsidi_600): Correctly select mlo/mhi
1417 (umulsidi_600): Likewise.
1419 2020-02-13 Jakub Jelinek <jakub@redhat.com>
1422 * config/i386/avx512bitalgintrin.h (_mm512_mask_popcnt_epi8,
1423 _mm512_mask_popcnt_epi16, _mm256_mask_popcnt_epi8,
1424 _mm256_mask_popcnt_epi16, _mm_mask_popcnt_epi8,
1425 _mm_mask_popcnt_epi16): Rename __B argument to __A and __A to __W,
1426 pass __A to the builtin followed by __W instead of __A followed by
1428 * config/i386/avx512vpopcntdqintrin.h (_mm512_mask_popcnt_epi32,
1429 _mm512_mask_popcnt_epi64): Likewise.
1430 * config/i386/avx512vpopcntdqvlintrin.h (_mm_mask_popcnt_epi32,
1431 _mm256_mask_popcnt_epi32, _mm_mask_popcnt_epi64,
1432 _mm256_mask_popcnt_epi64): Likewise.
1434 PR tree-optimization/93582
1435 * fold-const.h (shift_bytes_in_array_left,
1436 shift_bytes_in_array_right): Declare.
1437 * fold-const.c (shift_bytes_in_array_left,
1438 shift_bytes_in_array_right): New function, moved from
1439 gimple-ssa-store-merging.c, no longer static.
1440 * gimple-ssa-store-merging.c (shift_bytes_in_array): Move
1441 to gimple-ssa-store-merging.c and rename to shift_bytes_in_array_left.
1442 (shift_bytes_in_array_right): Move to gimple-ssa-store-merging.c.
1443 (encode_tree_to_bitpos): Use shift_bytes_in_array_left instead of
1444 shift_bytes_in_array.
1445 (verify_shift_bytes_in_array): Rename to ...
1446 (verify_shift_bytes_in_array_left): ... this. Use
1447 shift_bytes_in_array_left instead of shift_bytes_in_array.
1448 (store_merging_c_tests): Call verify_shift_bytes_in_array_left
1449 instead of verify_shift_bytes_in_array.
1450 * tree-ssa-sccvn.c (vn_reference_lookup_3): For native_encode_expr
1451 / native_interpret_expr where the store covers all needed bits,
1452 punt on PDP-endian, otherwise allow all involved offsets and sizes
1453 not to be byte-aligned.
1456 * config/i386/sse.md (k<code><mode>): Drop mode from last operand and
1457 use const_0_to_255_operand predicate instead of immediate_operand.
1458 (avx512dq_fpclass<mode><mask_scalar_merge_name>,
1459 avx512dq_vmfpclass<mode><mask_scalar_merge_name>,
1460 vgf2p8affineinvqb_<mode><mask_name>,
1461 vgf2p8affineqb_<mode><mask_name>): Drop mode from
1462 const_0_to_255_operand predicated operands.
1464 2020-02-12 Jeff Law <law@redhat.com>
1466 * config/h8300/h8300.md (comparison shortening peepholes): Use
1467 a mode iterator to merge the HImode and SImode peepholes.
1469 2020-02-12 Jakub Jelinek <jakub@redhat.com>
1472 * real.c (is_even): Make static. Function comment fix.
1473 (is_halfway_below): Make static, don't assert R is not inf/nan,
1474 instead return false for those. Small formatting fixes.
1476 2020-02-12 Martin Sebor <msebor@redhat.com>
1479 * tree-ssa-strlen.c (handle_builtin_stxncpy): Rename...
1480 (handle_builtin_stxncpy_strncat): ...to this. Change first argument.
1481 Issue only -Wstringop-overflow strncat, never -Wstringop-truncation.
1482 (strlen_check_and_optimize_call): Adjust callee name.
1484 2020-02-12 Jeff Law <law@redhat.com>
1486 * config/h8300/h8300.md (comparison shortening peepholes): Drop
1487 (and (xor)) variant. Combine other two into single peephole.
1489 2020-02-12 Wilco Dijkstra <wdijkstr@arm.com>
1491 PR rtl-optimization/93565
1492 * config/aarch64/aarch64.c (aarch64_rtx_costs): Add CTZ costs.
1494 2020-02-12 Wilco Dijkstra <wdijkstr@arm.com>
1496 * config/aarch64/aarch64-simd.md
1497 (aarch64_zero_extend<GPI:mode>_reduc_plus_<VDQV_E:mode>): New pattern.
1498 * config/aarch64/aarch64.md (popcount<mode>2): Use it instead of
1499 generating separate ADDV and zero_extend patterns.
1500 * config/aarch64/iterators.md (VDQV_E): New iterator.
1502 2020-02-12 Jeff Law <law@redhat.com>
1504 * config/h8300/h8300.md (cpymemsi, movmd): Remove dead patterns,
1505 expanders, splits, etc.
1506 (movmd_internal_<mode>, movmd splitter, movstr, movsd): Likewise.
1507 (stpcpy_internal_<mode>, stpcpy splitter): Likewise.
1508 (peepholes to convert QI/HI mode pushes to SI mode pushes): Likewise.
1509 * config/h8300/h8300.c (h8300_swap_into_er6): Remove unused function.
1510 (h8300_swap_out_of_er6, h8sx_emit_movmd): Likewise
1511 * config/h8300/h8300-protos.h (h8300_swap_into_er6): Remove unused
1513 (h8300_swap_out_of_er6, h8sx_emit_movmd): Likewise.
1515 2020-02-12 Jakub Jelinek <jakub@redhat.com>
1518 * config/i386/sse.md (VI48F_256_DQ): New mode iterator.
1519 (avx512vl_vextractf128<mode>): Use it instead of VI48F_256. Remove
1520 TARGET_AVX512DQ from condition.
1521 (vec_extract_lo_<mode><mask_name>): Use <mask_avx512dq_condition>
1522 instead of <mask_mode512bit_condition> in condition. If
1523 TARGET_AVX512DQ is false, emit vextract*64x4 instead of
1525 (vec_extract_lo_<mode><mask_name>): Drop <mask_avx512dq_condition>
1528 2020-02-12 Kewen Lin <linkw@gcc.gnu.org>
1531 * ira.c (combine_and_move_insns): Skip multiple_sets def_insn.
1533 2020-02-12 Segher Boessenkool <segher@kernel.crashing.org>
1535 * config/rs6000/rs6000.c (rs6000_debug_print_mode): Don't use sizeof
1536 where strlen is more legible.
1537 (rs6000_builtin_vectorized_libmass): Ditto.
1538 (rs6000_print_options_internal): Ditto.
1540 2020-02-11 Martin Sebor <msebor@redhat.com>
1542 PR tree-optimization/93683
1543 * tree-ssa-alias.c (stmt_kills_ref_p): Avoid using LHS when not set.
1545 2020-02-11 Michael Meissner <meissner@linux.ibm.com>
1547 * config/rs6000/predicates.md (cint34_operand): Rename the
1548 -mprefixed-addr option to be -mprefixed.
1549 * config/rs6000/rs6000-cpus.def (ISA_FUTURE_MASKS_SERVER): Rename
1550 the -mprefixed-addr option to be -mprefixed.
1551 (OTHER_FUTURE_MASKS): Likewise.
1552 (POWERPC_MASKS): Likewise.
1553 * config/rs6000/rs6000.c (rs6000_option_override_internal): Rename
1554 the -mprefixed-addr option to be -mprefixed. Change error
1555 messages to refer to -mprefixed.
1556 (num_insns_constant_gpr): Rename the -mprefixed-addr option to be
1558 (rs6000_legitimate_offset_address_p): Likewise.
1559 (rs6000_mode_dependent_address): Likewise.
1560 (rs6000_opt_masks): Change the spelling of "-mprefixed-addr" to be
1561 "-mprefixed" for target attributes and pragmas.
1562 (address_to_insn_form): Rename the -mprefixed-addr option to be
1564 (rs6000_adjust_insn_length): Likewise.
1565 * config/rs6000/rs6000.h (FINAL_PRESCAN_INSN): Rename the
1566 -mprefixed-addr option to be -mprefixed.
1567 (ASM_OUTPUT_OPCODE): Likewise.
1568 * config/rs6000/rs6000.md (prefixed insn attribute): Rename the
1569 -mprefixed-addr option to be -mprefixed.
1570 * config/rs6000/rs6000.opt (-mprefixed): Rename the
1571 -mprefixed-addr option to be prefixed. Change the option from
1572 being undocumented to being documented.
1573 * doc/invoke.texi (RS/6000 and PowerPC Options): Document the
1574 -mprefixed option. Update the -mpcrel documentation to mention
1577 2020-02-11 Hans-Peter Nilsson <hp@axis.com>
1579 * ira-conflicts.c (print_hard_reg_set): Correct output for sets
1580 including FIRST_PSEUDO_REGISTER - 1.
1581 * ira-color.c (print_hard_reg_set): Ditto.
1583 2020-02-11 Stam Markianos-Wright <stam.markianos-wright@arm.com>
1585 * config/arm/arm-builtins.c (enum arm_type_qualifiers):
1586 (USTERNOP_QUALIFIERS): New define.
1587 (USMAC_LANE_QUADTUP_QUALIFIERS): New define.
1588 (SUMAC_LANE_QUADTUP_QUALIFIERS): New define.
1589 (arm_expand_builtin_args): Add case ARG_BUILTIN_LANE_QUADTUP_INDEX.
1590 (arm_expand_builtin_1): Add qualifier_lane_quadtup_index.
1591 * config/arm/arm_neon.h (vusdot_s32): New.
1592 (vusdot_lane_s32): New.
1593 (vusdotq_lane_s32): New.
1594 (vsudot_lane_s32): New.
1595 (vsudotq_lane_s32): New.
1596 * config/arm/arm_neon_builtins.def (usdot, usdot_lane,sudot_lane): New.
1597 * config/arm/iterators.md (DOTPROD_I8MM): New.
1598 (sup, opsuffix): Add <us/su>.
1599 * config/arm/neon.md (neon_usdot, <us/su>dot_lane: New.
1600 * config/arm/unspecs.md (UNSPEC_DOT_US, UNSPEC_DOT_SU): New.
1602 2020-02-11 Richard Biener <rguenther@suse.de>
1604 PR tree-optimization/93661
1605 PR tree-optimization/93662
1606 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly guard
1608 * tree-sra.c (get_access_for_expr): Likewise.
1610 2020-02-10 Jakub Jelinek <jakub@redhat.com>
1613 * config/i386/sse.md (VI_256_AVX2): New mode iterator.
1614 (vcond_mask_<mode><sseintvecmodelower>): Use it instead of VI_256.
1615 Change condition from TARGET_AVX2 to TARGET_AVX.
1617 2020-02-10 Iain Sandoe <iain@sandoe.co.uk>
1620 * config/darwin-c.c (darwin_cfstring_ref_p): Fix up last
1621 argument of strncmp.
1623 2020-02-10 Hans-Peter Nilsson <hp@axis.com>
1625 Try to generate zero-based comparisons.
1626 * config/cris/cris.c (cris_reduce_compare): New function.
1627 * config/cris/cris-protos.h (cris_reduce_compare): Add prototype.
1628 * config/cris/cris.md ("cbranch<mode>4", "cbranchdi4", "cstoredi4")
1629 (cstore<mode>4"): Apply cris_reduce_compare in expanders.
1631 2020-02-10 Richard Earnshaw <rearnsha@arm.com>
1634 * config/arm/arm.md (movsi_compare0): Allow SP as a source register
1635 in Thumb state and also as a destination in Arm state. Add T16
1638 2020-02-10 Hans-Peter Nilsson <hp@axis.com>
1640 * md.texi (Define Subst): Match closing paren in example.
1642 2020-02-10 Jakub Jelinek <jakub@redhat.com>
1646 * config/i386/i386.c (x86_64_elf_section_type_flags): Fix up last
1647 arguments of strncmp.
1649 2020-02-10 Feng Xue <fxue@os.amperecomputing.com>
1652 * ipa-cp.c (ipcp_lattice::add_value): Add source with same call edge
1653 but different source value.
1654 (adjust_callers_for_value_intersection): New function.
1655 (gather_edges_for_value): Adjust order of callers to let a
1656 non-self-recursive caller be the first element.
1657 (self_recursive_pass_through_p): Add a new parameter "simple", and
1658 check generalized self-recursive pass-through jump function.
1659 (self_recursive_agg_pass_through_p): Likewise.
1660 (find_more_scalar_values_for_callers_subset): Compute value from
1661 pass-through jump function for self-recursive.
1662 (intersect_with_plats): Cleanup previous implementation code for value
1663 itersection with self-recursive call edge.
1664 (intersect_with_agg_replacements): Likewise.
1665 (intersect_aggregates_with_edge): Deduce value from pass-through jump
1666 function for self-recursive call edge. Cleanup previous implementation
1667 code for value intersection with self-recursive call edge.
1668 (decide_whether_version_node): Remove dead callers and adjust order
1669 to let a non-self-recursive caller be the first element.
1671 2020-02-09 Uroš Bizjak <ubizjak@gmail.com>
1673 * recog.c: Move pass_split_before_sched2 code in front of
1674 pass_split_before_regstack.
1675 (pass_data_split_before_sched2): Rename pass to split3 from split4.
1676 (pass_data_split_before_regstack): Rename pass to split4 from split3.
1677 (rest_of_handle_split_before_sched2): Remove.
1678 (pass_split_before_sched2::execute): Unconditionally call
1680 (enable_split_before_sched2): New function.
1681 (pass_split_before_sched2::gate): Use enable_split_before_sched2.
1682 (pass_split_before_regstack::gate): Ditto.
1683 * config/nds32/nds32.c (nds32_split_double_word_load_store_p):
1684 Update name check for renamed split4 pass.
1685 * config/sh/sh.c (register_sh_passes): Update pass insertion
1686 point for renamed split4 pass.
1688 2020-02-09 Jakub Jelinek <jakub@redhat.com>
1690 * gimplify.c (gimplify_adjust_omp_clauses_1): Promote
1691 DECL_IN_CONSTANT_POOL variables into "omp declare target" to avoid
1692 copying them around between host and target.
1694 2020-02-08 Andrew Pinski <apinski@marvell.com>
1697 * config/aarch64/aarch64-simd.md (movmisalign<mode>): Check
1698 STRICT_ALIGNMENT also.
1700 2020-02-08 Jim Wilson <jimw@sifive.com>
1703 * config/riscv/riscv.h (HARD_REGNO_CALLER_SAVE_MODE): Define.
1705 2020-02-08 Uroš Bizjak <ubizjak@gmail.com>
1706 Jakub Jelinek <jakub@redhat.com>
1709 * config/i386/i386.h (CALL_USED_REGISTERS): Make
1710 xmm16-xmm31 call-used even in 64-bit ms-abi.
1712 2020-02-07 Dennis Zhang <dennis.zhang@arm.com>
1714 * config/aarch64/aarch64-simd-builtins.def (simd_smmla): New entry.
1715 (simd_ummla, simd_usmmla): Likewise.
1716 * config/aarch64/aarch64-simd.md (aarch64_simd_<sur>mmlav16qi): New.
1717 * config/aarch64/arm_neon.h (vmmlaq_s32, vmmlaq_u32): New.
1718 (vusmmlaq_s32): New.
1720 2020-02-07 Richard Biener <rguenther@suse.de>
1723 * tree-inline.c (fold_marked_statements): Do a PRE walk,
1724 skipping unreachable regions.
1725 (optimize_inline_calls): Skip folding stmts when we didn't
1728 2020-02-07 H.J. Lu <hongjiu.lu@intel.com>
1731 * config/i386/i386.c (function_arg_ms_64): Add a type argument.
1732 Don't return aggregates with only SFmode and DFmode in SSE
1734 (ix86_function_arg): Pass arg.type to function_arg_ms_64.
1736 2020-02-07 Jakub Jelinek <jakub@redhat.com>
1739 * config/rs6000/rs6000-logue.c
1740 (rs6000_emit_probe_stack_range_stack_clash): Always use gen_add3_insn,
1741 if it fails, move rs into end_addr and retry. Add
1742 REG_FRAME_RELATED_EXPR note whenever it returns more than one insn or
1743 the insn pattern doesn't describe well what exactly happens to
1747 * config/i386/predicates.md (avx_identity_operand): Remove.
1748 * config/i386/sse.md (*avx_vec_concat<mode>_1): Remove.
1749 (avx_<castmode><avxsizesuffix>_<castmode>,
1750 avx512f_<castmode><avxsizesuffix>_256<castmode>): Change patterns to
1751 a VEC_CONCAT of the operand and UNSPEC_CAST.
1752 (avx512f_<castmode><avxsizesuffix>_<castmode>): Change pattern to
1753 a VEC_CONCAT of VEC_CONCAT of the operand and UNSPEC_CAST with
1757 * config/i386/i386.c (ix86_lea_outperforms): Make sure to clear
1758 recog_data.insn if distance_non_agu_define changed it.
1760 2020-02-06 Michael Meissner <meissner@linux.ibm.com>
1763 * config/rs6000/rs6000.c (reg_to_non_prefixed): Before ISA 3.0
1764 we only had X-FORM (reg+reg) addressing for vectors. Also before
1765 ISA 3.0, we only had X-FORM addressing for scalars in the
1766 traditional Altivec registers.
1768 2020-02-06 <zhongyunde@huawei.com>
1769 Vladimir Makarov <vmakarov@redhat.com>
1771 PR rtl-optimization/93561
1772 * lra-assigns.c (spill_for): Check that tested hard regno is not out of
1773 hard register range.
1775 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
1777 * config/aarch64/aarch64.md (aarch64_movk<mode>): Add a type
1780 2020-02-06 Segher Boessenkool <segher@kernel.crashing.org>
1782 * config/rs6000/rs6000.c (rs6000_emit_set_long_const): Handle the case
1783 where the low and the high 32 bits are equal to each other specially,
1784 with an rldimi instruction.
1786 2020-02-06 Mihail Ionescu <mihail.ionescu@arm.com>
1788 * config/arm/arm-cpus.in: Set profile M for armv8.1-m.main.
1790 2020-02-06 Mihail Ionescu <mihail.ionescu@arm.com>
1792 * config/arm/arm-tables.opt: Regenerate.
1794 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
1797 * config/aarch64/aarch64-protos.h (aarch64_movk_shift): Declare.
1798 * config/aarch64/aarch64.c (aarch64_movk_shift): New function.
1799 * config/aarch64/aarch64.md (aarch64_movk<mode>): New pattern.
1801 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
1803 PR rtl-optimization/87763
1804 * config/aarch64/aarch64.md (*ashiftsi_extvdi_bfiz): New pattern.
1806 2020-02-06 Delia Burduv <delia.burduv@arm.com>
1808 * config/aarch64/aarch64-simd-builtins.def
1809 (bfmlaq): New built-in function.
1810 (bfmlalb): New built-in function.
1811 (bfmlalt): New built-in function.
1812 (bfmlalb_lane): New built-in function.
1813 (bfmlalt_lane): New built-in function.
1814 * config/aarch64/aarch64-simd.md
1815 (aarch64_bfmmlaqv4sf): New pattern.
1816 (aarch64_bfmlal<bt>v4sf): New pattern.
1817 (aarch64_bfmlal<bt>_lane<q>v4sf): New pattern.
1818 * config/aarch64/arm_neon.h (vbfmmlaq_f32): New intrinsic.
1819 (vbfmlalbq_f32): New intrinsic.
1820 (vbfmlaltq_f32): New intrinsic.
1821 (vbfmlalbq_lane_f32): New intrinsic.
1822 (vbfmlaltq_lane_f32): New intrinsic.
1823 (vbfmlalbq_laneq_f32): New intrinsic.
1824 (vbfmlaltq_laneq_f32): New intrinsic.
1825 * config/aarch64/iterators.md (BF_MLA): New int iterator.
1826 (bt): New int attribute.
1828 2020-02-06 Uroš Bizjak <ubizjak@gmail.com>
1830 * config/i386/i386.md (*pushtf): Emit "#" instead of
1831 calling gcc_unreachable in insn output.
1834 (*pushsf_rex64): Ditto for alternatives other than 1.
1835 (*pushsf): Ditto for alternatives other than 1.
1837 2020-02-06 Martin Liska <mliska@suse.cz>
1839 PR gcov-profile/91971
1840 PR gcov-profile/93466
1841 * coverage.c (coverage_init): Revert mangling of
1842 path into filename. It can lead to huge filename length.
1843 Creation of subfolders seem more natural.
1845 2020-02-06 Stam Markianos-Wright <stam.markianos-wright@arm.com>
1848 * config/arm/arm.c (arm_block_arith_comp_libfuncs_for_mode): New.
1849 (arm_init_libfuncs): Add BFmode support to block spurious BF libfuncs.
1850 Use arm_block_arith_comp_libfuncs_for_mode for HFmode.
1852 2020-02-06 Jakub Jelinek <jakub@redhat.com>
1855 * config/i386/predicates.md (avx_identity_operand): New predicate.
1856 * config/i386/sse.md (*avx_vec_concat<mode>_1): New
1857 define_insn_and_split.
1860 * omp-low.c (use_pointer_for_field): For nested constructs, also
1861 look for map clauses on target construct.
1862 (scan_omp_1_stmt) <case GIMPLE_OMP_TARGET>: Bump temporarily
1863 taskreg_nesting_level.
1866 * gimplify.c (gimplify_scan_omp_clauses) <do_notice>: If adding
1867 shared clause, call omp_notice_variable on outer context if any.
1869 2020-02-05 Jason Merrill <jason@redhat.com>
1872 * symtab.c (symtab_node::nonzero_address): A DECL_COMDAT decl has
1873 non-zero address even if weak and not yet defined.
1875 2020-02-05 Martin Sebor <msebor@redhat.com>
1877 PR tree-optimization/92765
1878 * gimple-fold.c (get_range_strlen_tree): Handle MEM_REF and PARM_DECL.
1879 * tree-ssa-strlen.c (compute_string_length): Remove.
1880 (determine_min_objsize): Remove.
1881 (get_len_or_size): Add an argument. Call get_range_strlen_dynamic.
1882 Avoid using type size as the upper bound on string length.
1883 (handle_builtin_string_cmp): Add an argument. Adjust.
1884 (strlen_check_and_optimize_call): Pass additional argument to
1885 handle_builtin_string_cmp.
1887 2020-02-05 Uroš Bizjak <ubizjak@gmail.com>
1889 * config/i386/i386.md (*pushdi2_rex64 peephole2): Remove.
1890 (*pushdi2_rex64 peephole2): Unconditionally split after
1892 (*ashl<mode>3_doubleword): Ditto.
1893 (*<shift_insn><mode>3_doubleword): Ditto.
1895 2020-02-05 Michael Meissner <meissner@linux.ibm.com>
1898 * config/rs6000/rs6000.c (get_vector_offset): Fix
1900 2020-02-05 Andrew Stubbs <ams@codesourcery.com>
1902 * config/gcn/t-gcn-hsa (MULTILIB_OPTIONS): Use / not space.
1904 2020-02-05 David Malcolm <dmalcolm@redhat.com>
1907 (Special Functions for Debugging the Analyzer): Update description
1908 of __analyzer_dump_exploded_nodes.
1910 2020-02-05 Jakub Jelinek <jakub@redhat.com>
1913 * config/i386/i386-features.c (ix86_add_reg_usage_to_vzeroupper): Only
1914 include sets and not clobbers in the vzeroupper pattern.
1915 * config/i386/sse.md (*avx_vzeroupper): Require in insn condition that
1916 the parallel has 17 (64-bit) or 9 (32-bit) elts.
1917 (*avx_vzeroupper_1): New define_insn_and_split.
1920 * recog.c (pass_split_after_reload::gate): For STACK_REGS targets,
1921 don't run when !optimize.
1922 (pass_split_before_regstack::gate): For STACK_REGS targets, run even
1925 2020-02-05 Richard Biener <rguenther@suse.de>
1928 * genmatch.c (dt_node::gen_kids_1): Emit number of argument
1929 checks before matching calls.
1931 2020-02-05 Jakub Jelinek <jakub@redhat.com>
1933 * tree-ssa-alias.c (aliasing_matching_component_refs_p): Fix up
1934 function comment typo.
1937 * omp-simd-clone.c (expand_simd_clones): If simd_clone_mangle or
1938 simd_clone_create failed when i == 0, adjust clone->nargs by
1941 2020-02-05 Martin Liska <mliska@suse.cz>
1944 * doc/invoke.texi: Document that one should
1945 not combine ASLR and -fpch.
1947 2020-02-04 Richard Biener <rguenther@suse.de>
1949 PR tree-optimization/93538
1950 * match.pd (addr EQ/NE ptr): Amend to handle &ptr->x EQ/NE ptr.
1952 2020-02-04 Richard Biener <rguenther@suse.de>
1954 PR tree-optimization/91123
1955 * tree-ssa-sccvn.c (vn_walk_cb_data::finish): New method.
1956 (vn_walk_cb_data::last_vuse): New member.
1957 (vn_walk_cb_data::saved_operands): Likewsie.
1958 (vn_walk_cb_data::~vn_walk_cb_data): Release saved_operands.
1959 (vn_walk_cb_data::push_partial_def): Use finish.
1960 (vn_reference_lookup_2): Update last_vuse and use finish if
1961 we've saved operands.
1962 (vn_reference_lookup_3): Use finish and update calls to
1963 push_partial_defs everywhere. When translating through
1964 memcpy or aggregate copies save off operands and alias-set.
1965 (eliminate_dom_walker::eliminate_stmt): Restore VN_WALKREWRITE
1966 operation for redundant store removal.
1968 2020-02-04 Richard Biener <rguenther@suse.de>
1970 PR tree-optimization/92819
1971 * tree-ssa-forwprop.c (simplify_vector_constructor): Avoid
1972 generating more stmts than before.
1974 2020-02-04 Martin Liska <mliska@suse.cz>
1976 * config/arm/arm.c (arm_gen_far_branch): Move the function
1977 outside of selftests.
1979 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
1981 * config/rs6000/rs6000.c (adjust_vec_address_pcrel): New helper
1982 function to adjust PC-relative vector addresses.
1983 (rs6000_adjust_vec_address): Call adjust_vec_address_pcrel to
1984 handle vectors with PC-relative addresses.
1986 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
1988 * config/rs6000/rs6000.c (reg_to_non_prefixed): Add forward
1990 (hard_reg_and_mode_to_addr_mask): Delete.
1991 (rs6000_adjust_vec_address): If the original vector address
1992 was REG+REG or REG+OFFSET and the element is not zero, do the add
1993 of the elements in the original address before adding the offset
1994 for the vector element. Use address_to_insn_form to validate the
1995 address using the register being loaded, rather than guessing
1996 whether the address is a DS-FORM or DQ-FORM address.
1998 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
2000 * config/rs6000/rs6000.c (get_vector_offset): New helper function
2001 to calculate the offset in memory from the start of a vector of a
2002 particular element. Add code to keep the element number in
2003 bounds if the element number is variable.
2004 (rs6000_adjust_vec_address): Move calculation of offset of the
2005 vector element to get_vector_offset.
2006 (rs6000_split_vec_extract_var): Do not do the initial AND of
2007 element here, move the code to get_vector_offset.
2009 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
2011 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add some
2014 2020-02-03 Segher Boessenkool <segher@kernel.crashing.org>
2016 * config/rs6000/constraints.md: Improve documentation.
2018 2020-02-03 Richard Earnshaw <rearnsha@arm.com>
2021 * config/arm/t-arm: ($(srcdir)/config/arm/arm-tune.md)
2022 ($(srcdir)/config/arm/arm-tables.opt): Use move-if-change.
2024 2020-02-03 Andrew Stubbs <ams@codesourcery.com>
2026 * config.gcc: Remove "carrizo" support.
2027 * config/gcn/gcn-opts.h (processor_type): Likewise.
2028 * config/gcn/gcn.c (gcn_omp_device_kind_arch_isa): Likewise.
2029 * config/gcn/gcn.opt (gpu_type): Likewise.
2030 * config/gcn/t-omp-device: Likewise.
2032 2020-02-03 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2035 * config/arm/arm-protos.h: New function arm_gen_far_branch prototype.
2036 * config/arm/arm.c (arm_gen_far_branch): New function
2038 * config/arm/arm.md: Update b<cond> for Thumb2 range checks.
2040 2020-02-03 Julian Brown <julian@codesourcery.com>
2041 Tobias Burnus <tobias@codesourcery.com>
2043 * doc/invoke.texi: Update mention of OpenACC version to 2.6.
2045 2020-02-03 Jakub Jelinek <jakub@redhat.com>
2048 * config/s390/s390.md (popcounthi2_z196): Fix up expander to emit
2049 valid RTL to sum up the lowest and second lowest bytes of the popcnt
2052 2020-02-02 Vladimir Makarov <vmakarov@redhat.com>
2054 PR rtl-optimization/91333
2055 * ira-color.c (struct allocno_color_data): Add member
2057 (init_allocno_threads): Set the member up.
2058 (bucket_allocno_compare_func): Add compare hard reg
2061 2020-01-31 Sandra Loosemore <sandra@codesourcery.com>
2063 nios2: Support for GOT-relative DW_EH_PE_datarel encoding.
2065 * configure.ac [nios2-*-*]: Check HAVE_AS_NIOS2_GOTOFF_RELOCATION.
2066 * config.in: Regenerated.
2067 * configure: Regenerated.
2068 * config/nios2/nios2.h (ASM_PREFERRED_EH_DATA_FORMAT): Fix handling
2069 for PIC when HAVE_AS_NIOS2_GOTOFF_RELOCATION.
2070 (ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX): New.
2072 2020-02-01 Andrew Burgess <andrew.burgess@embecosm.com>
2074 * configure: Regenerate.
2076 2020-01-31 Vladimir Makarov <vmakarov@redhat.com>
2078 PR rtl-optimization/91333
2079 * ira-color.c (bucket_allocno_compare_func): Move conflict hard
2080 reg preferences comparison up.
2082 2020-01-31 Richard Sandiford <richard.sandiford@arm.com>
2084 * config/aarch64/aarch64.h (TARGET_SVE_BF16): New macro.
2085 * config/aarch64/aarch64-sve-builtins-sve2.h (svcvtnt): Move to
2086 aarch64-sve-builtins-base.h.
2087 * config/aarch64/aarch64-sve-builtins-sve2.cc (svcvtnt): Move to
2088 aarch64-sve-builtins-base.cc.
2089 * config/aarch64/aarch64-sve-builtins-base.h (svbfdot, svbfdot_lane)
2090 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
2092 * config/aarch64/aarch64-sve-builtins-base.cc (svbfdot, svbfdot_lane)
2093 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
2094 (svcvtnt): New functions.
2095 * config/aarch64/aarch64-sve-builtins-base.def (svbfdot, svbfdot_lane)
2096 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
2097 (svcvtnt): New functions.
2098 (svcvt): Add a form that converts f32 to bf16.
2099 * config/aarch64/aarch64-sve-builtins-shapes.h (ternary_bfloat)
2100 (ternary_bfloat_lane, ternary_bfloat_lanex2, ternary_bfloat_opt_n):
2102 * config/aarch64/aarch64-sve-builtins-shapes.cc (parse_element_type):
2103 Treat B as bfloat16_t.
2104 (ternary_bfloat_lane_base): New class.
2105 (ternary_bfloat_def): Likewise.
2106 (ternary_bfloat): New shape.
2107 (ternary_bfloat_lane_def): New class.
2108 (ternary_bfloat_lane): New shape.
2109 (ternary_bfloat_lanex2_def): New class.
2110 (ternary_bfloat_lanex2): New shape.
2111 (ternary_bfloat_opt_n_def): New class.
2112 (ternary_bfloat_opt_n): New shape.
2113 * config/aarch64/aarch64-sve-builtins.cc (TYPES_cvt_bfloat): New macro.
2114 * config/aarch64/aarch64-sve.md (@aarch64_sve_<sve_fp_op>vnx4sf)
2115 (@aarch64_sve_<sve_fp_op>_lanevnx4sf): New patterns.
2116 (@aarch64_sve_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>)
2117 (@cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
2118 (*cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
2119 (@aarch64_sve_cvtnt<VNx8BF_ONLY:mode>): Likewise.
2120 * config/aarch64/aarch64-sve2.md (@aarch64_sve2_cvtnt<mode>): Key
2121 the pattern off the narrow mode instead of the wider one.
2122 * config/aarch64/iterators.md (VNx8BF_ONLY): New mode iterator.
2123 (UNSPEC_BFMLALB, UNSPEC_BFMLALT, UNSPEC_BFMMLA): New unspecs.
2124 (sve_fp_op): Handle them.
2125 (SVE_BFLOAT_TERNARY_LONG): New int itertor.
2126 (SVE_BFLOAT_TERNARY_LONG_LANE): Likewise.
2128 2020-01-31 Richard Sandiford <richard.sandiford@arm.com>
2130 * config/aarch64/arm_sve.h: Include arm_bf16.h.
2131 * config/aarch64/aarch64-modes.def (BF): Move definition before
2132 VECTOR_MODES. Remove separate VECTOR_MODES for V4BF and V8BF.
2133 (SVE_MODES): Handle BF modes.
2134 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
2136 (aarch64_full_sve_mode): Likewise.
2137 * config/aarch64/iterators.md (SVE_STRUCT): Add VNx16BF, VNx24BF
2139 (SVE_FULL, SVE_FULL_HSD, SVE_ALL): Add VNx8BF.
2140 (Vetype, Vesize, Vctype, VEL, Vel, VEL_INT, V128, v128, vwcore)
2141 (V_INT_EQUIV, v_int_equiv, V_FP_EQUIV, v_fp_equiv, vector_count)
2142 (insn_length, VSINGLE, vsingle, VPRED, vpred, VDOUBLE): Handle the
2144 * config/aarch64/aarch64-sve-builtins.h (TYPE_bfloat): New
2146 * config/aarch64/aarch64-sve-builtins.cc (TYPES_all_arith): New macro.
2147 (TYPES_all_data): Add bf16.
2148 (TYPES_reinterpret1, TYPES_reinterpret): Likewise.
2149 (register_tuple_type): Increase buffer size.
2150 * config/aarch64/aarch64-sve-builtins.def (svbfloat16_t): New type.
2151 (bf16): New type suffix.
2152 * config/aarch64/aarch64-sve-builtins-base.def (svabd, svadd, svaddv)
2153 (svcmpeq, svcmpge, svcmpgt, svcmple, svcmplt, svcmpne, svmad, svmax)
2154 (svmaxv, svmin, svminv, svmla, svmls, svmsb, svmul, svsub, svsubr):
2155 Change type from all_data to all_arith.
2156 * config/aarch64/aarch64-sve-builtins-sve2.def (svaddp, svmaxp)
2159 2020-01-31 Dennis Zhang <dennis.zhang@arm.com>
2160 Matthew Malcomson <matthew.malcomson@arm.com>
2161 Richard Sandiford <richard.sandiford@arm.com>
2163 * doc/invoke.texi (f32mm): Document new AArch64 -march= extension.
2164 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Define
2165 __ARM_FEATURE_SVE_MATMUL_INT8, __ARM_FEATURE_SVE_MATMUL_FP32 and
2166 __ARM_FEATURE_SVE_MATMUL_FP64 as appropriate. Don't define
2167 __ARM_FEATURE_MATMUL_FP64.
2168 * config/aarch64/aarch64-option-extensions.def (fp, simd, fp16)
2169 (sve): Add AARCH64_FL_F32MM to the list of extensions that should
2170 be disabled at the same time.
2171 (f32mm): New extension.
2172 * config/aarch64/aarch64.h (AARCH64_FL_F32MM): New macro.
2173 (AARCH64_FL_F64MM): Bump to the next bit up.
2174 (AARCH64_ISA_F32MM, TARGET_SVE_I8MM, TARGET_F32MM, TARGET_SVE_F32MM)
2175 (TARGET_SVE_F64MM): New macros.
2176 * config/aarch64/iterators.md (SVE_MATMULF): New mode iterator.
2177 (UNSPEC_FMMLA, UNSPEC_SMATMUL, UNSPEC_UMATMUL, UNSPEC_USMATMUL)
2178 (UNSPEC_TRN1Q, UNSPEC_TRN2Q, UNSPEC_UZP1Q, UNSPEC_UZP2Q, UNSPEC_ZIP1Q)
2179 (UNSPEC_ZIP2Q): New unspeccs.
2180 (DOTPROD_US_ONLY, PERMUTEQ, MATMUL, FMMLA): New int iterators.
2181 (optab, sur, perm_insn): Handle the new unspecs.
2182 (sve_fp_op): Handle UNSPEC_FMMLA. Resort.
2183 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use
2184 TARGET_SVE_F64MM instead of separate tests.
2185 (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod<vsi2qi>): New pattern.
2186 (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod_lane<vsi2qi>): Likewise.
2187 (@aarch64_sve_add_<MATMUL:optab><vsi2qi>): Likewise.
2188 (@aarch64_sve_<FMMLA:sve_fp_op><mode>): Likewise.
2189 (@aarch64_sve_<PERMUTEQ:optab><mode>): Likewise.
2190 * config/aarch64/aarch64-sve-builtins.cc (TYPES_s_float): New macro.
2191 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): Use it.
2192 (TYPES_s_signed): New macro.
2193 (TYPES_s_integer): Use it.
2194 (TYPES_d_float): New macro.
2195 (TYPES_d_data): Use it.
2196 * config/aarch64/aarch64-sve-builtins-shapes.h (mmla): Declare.
2197 (ternary_intq_uintq_lane, ternary_intq_uintq_opt_n, ternary_uintq_intq)
2198 (ternary_uintq_intq_lane, ternary_uintq_intq_opt_n): Likewise.
2199 * config/aarch64/aarch64-sve-builtins-shapes.cc (mmla_def): New class.
2200 (svmmla): New shape.
2201 (ternary_resize2_opt_n_base): Add TYPE_CLASS2 and TYPE_CLASS3
2202 template parameters.
2203 (ternary_resize2_lane_base): Likewise.
2204 (ternary_resize2_base): New class.
2205 (ternary_qq_lane_base): Likewise.
2206 (ternary_intq_uintq_lane_def): Likewise.
2207 (ternary_intq_uintq_lane): New shape.
2208 (ternary_intq_uintq_opt_n_def): New class
2209 (ternary_intq_uintq_opt_n): New shape.
2210 (ternary_qq_lane_def): Inherit from ternary_qq_lane_base.
2211 (ternary_uintq_intq_def): New class.
2212 (ternary_uintq_intq): New shape.
2213 (ternary_uintq_intq_lane_def): New class.
2214 (ternary_uintq_intq_lane): New shape.
2215 (ternary_uintq_intq_opt_n_def): New class.
2216 (ternary_uintq_intq_opt_n): New shape.
2217 * config/aarch64/aarch64-sve-builtins-base.h (svmmla, svsudot)
2218 (svsudot_lane, svtrn1q, svtrn2q, svusdot, svusdot_lane, svusmmla)
2219 (svuzp1q, svuzp2q, svzip1q, svzip2q): Declare.
2220 * config/aarch64/aarch64-sve-builtins-base.cc (svdot_lane_impl):
2222 (svdotprod_lane_impl): ...this new class.
2223 (svmmla_impl, svusdot_impl): New classes.
2224 (svdot_lane): Update to use svdotprod_lane_impl.
2225 (svmmla, svsudot, svsudot_lane, svtrn1q, svtrn2q, svusdot)
2226 (svusdot_lane, svusmmla, svuzp1q, svuzp2q, svzip1q, svzip2q): New
2228 * config/aarch64/aarch64-sve-builtins-base.def (svmmla): New base
2229 function, with no types defined.
2230 (svmmla, svusmmla, svsudot, svsudot_lane, svusdot, svusdot_lane): New
2231 AARCH64_FL_I8MM functions.
2232 (svmmla): New AARCH64_FL_F32MM function.
2233 (svld1ro): Depend only on AARCH64_FL_F64MM, not on AARCH64_FL_V8_6.
2234 (svmmla, svtrn1q, svtrn2q, svuz1q, svuz2q, svzip1q, svzip2q): New
2235 AARCH64_FL_F64MM function.
2236 (REQUIRED_EXTENSIONS):
2238 2020-01-31 Andrew Stubbs <ams@codesourcery.com>
2240 * config/gcn/gcn-valu.md (addv64di3_exec): Allow one '0' in each
2243 2020-01-31 Uroš Bizjak <ubizjak@gmail.com>
2245 * config/i386/i386.md (*movoi_internal_avx): Do not check for
2246 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL. Remove MODE_V8SF handling.
2247 (*movti_internal): Do not check for
2248 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
2249 (*movtf_internal): Move check for TARGET_SSE2 and size optimization
2250 just after check for TARGET_AVX.
2251 (*movdf_internal): Ditto.
2252 * config/i386/mmx.md (*mov<mode>_internal): Do not check for
2253 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
2254 * config/i386/sse.md (mov<mode>_internal): Only check
2255 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL with V2DFmode. Move check
2256 for TARGET_SSE2 and size optimization just after check for TARGET_AVX.
2257 (<sse>_andnot<mode>3<mask_name>): Move check for
2258 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL after check for TARGET_AVX.
2259 (<code><mode>3<mask_name>): Ditto.
2260 (*andnot<mode>3): Ditto.
2261 (*andnottf3): Ditto.
2262 (*<code><mode>3): Ditto.
2263 (*<code>tf3): Ditto.
2264 (*andnot<VI:mode>3): Remove
2265 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL handling.
2266 (<mask_codefor><code><VI48_AVX_AVX512F:mode>3<mask_name>): Ditto.
2267 (*<code><VI12_AVX_AVX512F:mode>3): Ditto.
2268 (sse4_1_blendv<ssemodesuffix>): Ditto.
2269 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL):
2270 Explain that tune applies to 128bit instructions only.
2272 2020-01-31 Kwok Cheung Yeung <kcy@codesourcery.com>
2274 * config/gcn/mkoffload.c (process_asm): Add sgpr_count and vgpr_count
2275 to definition of hsa_kernel_description. Parse assembly to find SGPR
2276 and VGPR count of kernel and store in hsa_kernel_description.
2278 2020-01-31 Tamar Christina <tamar.christina@arm.com>
2280 PR rtl-optimization/91838
2281 * simplify-rtx.c (simplify_binary_operation_1): Update LSHIFTRT case
2282 to truncate if allowed or reject combination.
2284 2020-01-31 Andrew Stubbs <ams@codesourcery.com>
2286 * tree-ssa-loop-ivopts.c (get_iv): Use sizetype for zero-step.
2287 (find_inv_vars_cb): Likewise.
2289 2020-01-31 David Malcolm <dmalcolm@redhat.com>
2291 * calls.c (special_function_p): Split out the check for DECL_NAME
2292 being non-NULL and fndecl being extern at file scope into a
2293 new maybe_special_function_p and call it. Drop check for fndecl
2294 being non-NULL that was after a usage of DECL_NAME (fndecl).
2295 * tree.h (maybe_special_function_p): New inline function.
2297 2020-01-30 Andrew Stubbs <ams@codesourcery.com>
2299 * config/gcn/gcn-valu.md (gather<mode>_exec): Move contents ...
2300 (mask_gather_load<mode>): ... here, and zero-initialize the
2302 (maskload<mode>di): Zero-initialize the destination.
2305 2020-01-30 David Malcolm <dmalcolm@redhat.com>
2308 * doc/analyzer.texi (Limitations): Note that constraints on
2309 floating-point values are currently ignored.
2311 2020-01-30 Jakub Jelinek <jakub@redhat.com>
2314 * symtab.c (symtab_node::noninterposable_alias): If localalias
2315 already exists, but is not usable, append numbers after it until
2316 a unique name is found. Formatting fix.
2319 * combine.c (simplify_comparison) <case ROTATE>: Punt on out of range
2322 2020-01-30 Andrew Stubbs <ams@codesourcery.com>
2324 * config/gcn/gcn.c (print_operand): Handle LTGT.
2325 * config/gcn/predicates.md (gcn_fp_compare_operator): Allow ltgt.
2327 2020-01-30 Richard Biener <rguenther@suse.de>
2329 * tree-pretty-print.c (dump_generic_node): Wrap VECTOR_CST
2330 and CONSTRUCTOR in _Literal (type) with TDF_GIMPLE.
2332 2020-01-30 John David Anglin <danglin@gcc.gnu.org>
2334 * config/pa/pa.c (pa_elf_select_rtx_section): Place function pointers
2335 without a DECL in .data.rel.ro.local.
2337 2020-01-30 Jakub Jelinek <jakub@redhat.com>
2340 * config/arm/arm.md (uaddvdi4): Actually emit what gen_uaddvsi4
2344 * config/i386/sse.md
2345 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext): Renamed to ...
2346 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext): ... this. Use
2347 any_extend code iterator instead of always zero_extend.
2348 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_lt): Renamed to ...
2349 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_lt): ... this.
2350 Use any_extend code iterator instead of always zero_extend.
2351 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_shift): Renamed to ...
2352 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_shift): ... this.
2353 Use any_extend code iterator instead of always zero_extend.
2354 (*sse2_pmovmskb_ext): New define_insn.
2355 (*sse2_pmovmskb_ext_lt): New define_insn_and_split.
2358 * config/i386/i386.md (*popcountsi2_zext): New define_insn_and_split.
2359 (*popcountsi2_zext_falsedep): New define_insn.
2361 2020-01-30 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
2363 * config.in: Regenerated.
2364 * configure: Regenerated.
2366 2020-01-29 Tobias Burnus <tobias@codesourcery.com>
2369 * config/gcn/gcn-hsa.h (ASM_SPEC): Add -mattr=-code-object-v3 as
2370 LLVM's assembler changed the default in version 9.
2372 2020-01-24 Jeff Law <law@redhat.com>
2374 PR tree-optimization/89689
2375 * builtins.def (BUILT_IN_OBJECT_SIZE): Make it const rather than pure.
2377 2020-01-29 Richard Sandiford <richard.sandiford@arm.com>
2381 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
2383 PR rtl-optimization/87763
2384 * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
2385 simplification to handle subregs as well as bare regs.
2386 * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
2388 2020-01-29 Joel Hutton <Joel.Hutton@arm.com>
2391 * ira.c (ira): Revert use of simplified LRA algorithm.
2393 2020-01-29 Martin Jambor <mjambor@suse.cz>
2395 PR tree-optimization/92706
2396 * tree-sra.c (struct access): Fields first_link, last_link,
2397 next_queued and grp_queued renamed to first_rhs_link, last_rhs_link,
2398 next_rhs_queued and grp_rhs_queued respectively, new fields
2399 first_lhs_link, last_lhs_link, next_lhs_queued and grp_lhs_queued.
2400 (struct assign_link): Field next renamed to next_rhs, new field
2401 next_lhs. Updated comment.
2402 (work_queue_head): Renamed to rhs_work_queue_head.
2403 (lhs_work_queue_head): New variable.
2404 (add_link_to_lhs): New function.
2405 (relink_to_new_repr): Also relink LHS lists.
2406 (add_access_to_work_queue): Renamed to add_access_to_rhs_work_queue.
2407 (add_access_to_lhs_work_queue): New function.
2408 (pop_access_from_work_queue): Renamed to
2409 pop_access_from_rhs_work_queue.
2410 (pop_access_from_lhs_work_queue): New function.
2411 (build_accesses_from_assign): Also add links to LHS lists and to LHS
2413 (child_would_conflict_in_lacc): Renamed to
2414 child_would_conflict_in_acc. Adjusted parameter names.
2415 (create_artificial_child_access): New parameter set_grp_read, use it.
2416 (subtree_mark_written_and_enqueue): Renamed to
2417 subtree_mark_written_and_rhs_enqueue.
2418 (propagate_subaccesses_across_link): Renamed to
2419 propagate_subaccesses_from_rhs.
2420 (propagate_subaccesses_from_lhs): New function.
2421 (propagate_all_subaccesses): Also propagate subaccesses from LHSs to
2424 2020-01-29 Martin Jambor <mjambor@suse.cz>
2426 PR tree-optimization/92706
2427 * tree-sra.c (struct access): Adjust comment of
2428 grp_total_scalarization.
2429 (find_access_in_subtree): Look for single children spanning an entire
2431 (scalarizable_type_p): Allow register accesses, adjust callers.
2432 (completely_scalarize): Remove function.
2433 (scalarize_elem): Likewise.
2434 (create_total_scalarization_access): Likewise.
2435 (sort_and_splice_var_accesses): Do not track total scalarization
2437 (analyze_access_subtree): New parameter totally, adjust to new meaning
2438 of grp_total_scalarization.
2439 (analyze_access_trees): Pass new parameter to analyze_access_subtree.
2440 (can_totally_scalarize_forest_p): New function.
2441 (create_total_scalarization_access): Likewise.
2442 (create_total_access_and_reshape): Likewise.
2443 (total_should_skip_creating_access): Likewise.
2444 (totally_scalarize_subtree): Likewise.
2445 (analyze_all_variable_accesses): Perform total scalarization after
2446 subaccess propagation using the new functions above.
2447 (initialize_constant_pool_replacements): Output initializers by
2448 traversing the access tree.
2450 2020-01-29 Martin Jambor <mjambor@suse.cz>
2452 * tree-sra.c (verify_sra_access_forest): New function.
2453 (verify_all_sra_access_forests): Likewise.
2454 (create_artificial_child_access): Set parent.
2455 (analyze_all_variable_accesses): Call the verifier.
2457 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
2459 * cgraph.c (cgraph_edge::resolve_speculation): Only lookup direct edge
2460 if called on indirect edge.
2461 (cgraph_edge::redirect_call_stmt_to_callee): Lookup indirect edge of
2462 speculative call if needed.
2464 2020-01-29 Richard Biener <rguenther@suse.de>
2466 PR tree-optimization/93428
2467 * tree-vect-slp.c (vect_build_slp_tree_2): Compute the load
2468 permutation when the load node is created.
2469 (vect_analyze_slp_instance): Re-use it here.
2471 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
2473 * ipa-prop.c (update_indirect_edges_after_inlining): Fix warning.
2475 2020-01-28 Vladimir Makarov <vmakarov@redhat.com>
2477 PR rtl-optimization/93272
2478 * ira-lives.c (process_out_of_region_eh_regs): New function.
2479 (process_bb_node_lives): Call it.
2481 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
2483 * coverage.c (read_counts_file): Make error message lowercase.
2485 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
2487 * profile-count.c (profile_quality_display_names): Fix ordering.
2489 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
2492 * cgraph.c (cgraph_add_edge_to_call_site_hash): Update call site
2493 hash only when edge is first within the sequence.
2494 (cgraph_edge::set_call_stmt): Update handling of speculative calls.
2495 (symbol_table::create_edge): Do not set target_prob.
2496 (cgraph_edge::remove_caller): Watch for speculative calls when updating
2498 (cgraph_edge::make_speculative): Drop target_prob parameter.
2499 (cgraph_edge::speculative_call_info): Remove.
2500 (cgraph_edge::first_speculative_call_target): New member function.
2501 (update_call_stmt_hash_for_removing_direct_edge): New function.
2502 (cgraph_edge::resolve_speculation): Rewrite to new API.
2503 (cgraph_edge::speculative_call_for_target): New member function.
2504 (cgraph_edge::make_direct): Rewrite to new API; fix handling of
2505 multiple speculation targets.
2506 (cgraph_edge::redirect_call_stmt_to_callee): Likewise; fix updating
2508 (verify_speculative_call): Verify that targets form an interval.
2509 * cgraph.h (cgraph_edge::speculative_call_info): Remove.
2510 (cgraph_edge::first_speculative_call_target): New member function.
2511 (cgraph_edge::next_speculative_call_target): New member function.
2512 (cgraph_edge::speculative_call_target_ref): New member function.
2513 (cgraph_edge;:speculative_call_indirect_edge): New member funtion.
2514 (cgraph_edge): Remove target_prob.
2515 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
2516 Fix handling of speculative calls.
2517 * ipa-devirt.c (ipa_devirt): Fix handling of speculative cals.
2518 * ipa-fnsummary.c (analyze_function_body): Likewise.
2519 * ipa-inline.c (speculation_useful_p): Use new speculative call API.
2520 * ipa-profile.c (dump_histogram): Fix formating.
2521 (ipa_profile_generate_summary): Watch for overflows.
2522 (ipa_profile): Do not require probablity to be 1/2; update to new API.
2523 * ipa-prop.c (ipa_make_edge_direct_to_target): Update to new API.
2524 (update_indirect_edges_after_inlining): Update to new API.
2525 * ipa-utils.c (ipa_merge_profiles): Rewrite merging of speculative call
2527 * profile-count.h: (profile_probability::adjusted): New.
2528 * tree-inline.c (copy_bb): Update to new speculative call API; fix
2529 updating of profile.
2530 * value-prof.c (gimple_ic_transform): Rename to ...
2531 (dump_ic_profile): ... this one; update dumping.
2532 (stream_in_histogram_value): Fix formating.
2533 (gimple_value_profile_transformations): Update.
2535 2020-01-28 H.J. Lu <hongjiu.lu@intel.com>
2538 * config/i386/i386.md (*movoi_internal_avx): Remove
2539 TARGET_SSE_TYPELESS_STORES check.
2540 (*movti_internal): Prefer TARGET_AVX over
2541 TARGET_SSE_TYPELESS_STORES.
2542 (*movtf_internal): Likewise.
2543 * config/i386/sse.md (mov<mode>_internal): Prefer TARGET_AVX over
2544 TARGET_SSE_TYPELESS_STORES. Remove "<MODE_SIZE> == 16" check
2545 from TARGET_SSE_TYPELESS_STORES.
2547 2020-01-28 David Malcolm <dmalcolm@redhat.com>
2549 * diagnostic-core.h (warning_at): Rename overload to...
2550 (warning_meta): ...this.
2551 (emit_diagnostic_valist): Delete decl of overload taking
2552 diagnostic_metadata.
2553 * diagnostic.c (emit_diagnostic_valist): Likewise for defn.
2554 (warning_at): Rename overload taking diagnostic_metadata to...
2555 (warning_meta): ...this.
2557 2020-01-28 Richard Biener <rguenther@suse.de>
2559 PR tree-optimization/93439
2560 * tree-parloops.c (create_loop_fn): Move clique bookkeeping...
2561 * tree-cfg.c (move_sese_region_to_fn): ... here.
2562 (verify_types_in_gimple_reference): Verify used cliques are
2565 2020-01-28 H.J. Lu <hongjiu.lu@intel.com>
2568 * config/i386/i386-options.c (set_ix86_tune_features): Add an
2569 argument of a pointer to struct gcc_options and pass it to
2570 parse_mtune_ctrl_str.
2571 (ix86_function_specific_restore): Pass opts to
2572 set_ix86_tune_features.
2573 (ix86_option_override_internal): Likewise.
2574 (parse_mtune_ctrl_str): Add an argument of a pointer to struct
2575 gcc_options and use it for x_ix86_tune_ctrl_string.
2577 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
2579 PR rtl-optimization/87763
2580 * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
2581 simplification to handle subregs as well as bare regs.
2582 * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
2584 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
2586 * tree-vect-loop.c (vectorizable_reduction): Fail gracefully
2587 for reduction chains that (now) include a call.
2589 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
2591 PR tree-optimization/92822
2592 * tree-ssa-forwprop.c (simplify_vector_constructor): When filling
2593 out the don't-care elements of a vector whose significant elements
2594 are duplicates, make the don't-care elements duplicates too.
2596 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
2598 PR tree-optimization/93434
2599 * tree-predcom.c (split_data_refs_to_components): Record which
2600 components have had aliasing loads removed. Prevent store-store
2601 commoning for all such components.
2603 2020-01-28 Jakub Jelinek <jakub@redhat.com>
2606 * config/i386/i386.c (ix86_fold_builtin) <do_shift>: If mask is not
2607 -1 or is_vshift is true, use new_vector with number of elts npatterns
2608 rather than new_unary_operation.
2610 PR tree-optimization/93454
2611 * gimple-fold.c (fold_array_ctor_reference): Perform
2612 elt_size.to_uhwi () just once, instead of calling it in every
2613 iteration. Punt if that value is above size of the temporary
2614 buffer. Decrease third native_encode_expr argument when
2615 bufoff + elt_sz is above size of buf.
2617 2020-01-27 Joseph Myers <joseph@codesourcery.com>
2619 * config/mips/mips.c (mips_declare_object_name)
2620 [USE_GNU_UNIQUE_OBJECT]: Support use of gnu_unique_object.
2622 2020-01-27 Martin Liska <mliska@suse.cz>
2624 PR gcov-profile/93403
2625 * tree-profile.c (gimple_init_gcov_profiler): Generate
2626 both __gcov_indirect_call_profiler_v4 and
2627 __gcov_indirect_call_profiler_v4_atomic.
2629 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
2632 * config/aarch64/aarch64-simd.md (aarch64_get_half<mode>): New
2634 (@aarch64_split_simd_mov<mode>): Use it.
2635 (aarch64_simd_mov_from_<mode>low): Add a GPR alternative.
2636 Leave the vec_extract patterns to handle 2-element vectors.
2637 (aarch64_simd_mov_from_<mode>high): Likewise.
2638 (vec_extract<VQMOV_NO2E:mode><Vhalf>): New expander.
2639 (vec_extractv2dfv1df): Likewise.
2641 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
2643 * config/aarch64/aarch64.c (aarch64_if_then_else_costs): Match
2644 jump conditions for *compare_condjump<GPI:mode>.
2646 2020-01-27 David Malcolm <dmalcolm@redhat.com>
2649 * digraph.cc (test_edge::test_edge): Specify template for base
2652 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
2654 * config/arc/arc.c (arc_rtx_costs): Update mul64 cost.
2656 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
2658 * config/arc/arc-protos.h (gen_mlo): Remove.
2659 (gen_mhi): Likewise.
2660 * config/arc/arc.c (AUX_MULHI): Define.
2661 (arc_must_save_reister): Special handling for r58/59.
2662 (arc_compute_frame_size): Consider mlo/mhi registers.
2663 (arc_save_callee_saves): Emit fp/sp move only when emit_move
2665 (arc_conditional_register_usage): Remove TARGET_BIG_ENDIAN from
2666 mlo/mhi name selection.
2667 (arc_restore_callee_saves): Don't early restore blink when ISR.
2668 (arc_expand_prologue): Add mlo/mhi saving.
2669 (arc_expand_epilogue): Add mlo/mhi restoring.
2672 * config/arc/arc.h (DBX_REGISTER_NUMBER): Correct register
2673 numbering when MUL64 option is used.
2674 (DWARF2_FRAME_REG_OUT): Define.
2675 * config/arc/arc.md (arc600_stall): New pattern.
2676 (VUNSPEC_ARC_ARC600_STALL): Define.
2677 (mulsi64): Use correct mlo/mhi registers.
2678 (mulsi_600): Clean it up.
2679 * config/arc/predicates.md (mlo_operand): Remove any dependency on
2681 (mhi_operand): Likewise.
2683 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
2684 Petro Karashchenko <petro.karashchenko@ring.com>
2686 * config/arc/arc.c (arc_is_uncached_mem_p): Check struct
2687 attributes if needed.
2688 (prepare_move_operands): Generate special unspec instruction for
2690 (arc_isuncached_mem_p): Propagate uncached attribute to each
2692 * config/arc/arc.md (VUNSPEC_ARC_LDDI): Define.
2693 (VUNSPEC_ARC_STDI): Likewise.
2694 (ALLI): New mode iterator.
2695 (mALLI): New mode attribute.
2696 (lddi): New instruction pattern.
2698 (stdidi_split): Split instruction for architectures which are not
2699 supporting ll64 option.
2700 (lddidi_split): Likewise.
2702 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
2704 PR rtl-optimization/92989
2705 * lra-lives.c (process_bb_lives): Update the live-in set before
2706 processing additional clobbers.
2708 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
2710 PR rtl-optimization/93170
2711 * cselib.c (cselib_invalidate_regno_val): New function, split out
2713 (cselib_invalidate_regno): ...here.
2714 (cselib_invalidated_by_call_p): New function.
2715 (cselib_process_insn): Iterate over all the hard-register entries in
2716 REG_VALUES and invalidate any that cross call-clobbered registers.
2718 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
2720 * dojump.c (split_comparison): Use HONOR_NANS rather than
2721 HONOR_SNANS when splitting LTGT.
2723 2020-01-27 Martin Liska <mliska@suse.cz>
2726 * opts.c (print_filtered_help): Exclude language-specific
2727 options from --help=common unless enabled in all FEs.
2729 2020-01-27 Martin Liska <mliska@suse.cz>
2731 * opts.c (print_help): Exclude params from
2732 all except --help=param.
2734 2020-01-27 Martin Liska <mliska@suse.cz>
2737 * config/i386/i386-features.c (make_resolver_func):
2738 Align the code with ppc64 target implementation.
2739 Do not generate a unique name for resolver function.
2741 2020-01-27 Richard Biener <rguenther@suse.de>
2743 PR tree-optimization/93397
2744 * tree-vect-slp.c (vect_analyze_slp_instance): Delay
2745 converted reduction chain SLP graph adjustment.
2747 2020-01-26 Marek Polacek <polacek@redhat.com>
2750 * sanopt.c (sanitize_rewrite_addressable_params): Avoid crash on
2753 2020-01-26 Jason Merrill <jason@redhat.com>
2756 * tree.c (verify_type_variant): Only verify TYPE_NEEDS_CONSTRUCTING
2759 2020-01-26 Darius Galis <darius.galis@cyberthorstudios.com>
2761 * config/rx/rx.md (setmemsi): Added rx_allow_string_insns constraint
2762 (rx_setmem): Likewise.
2764 2020-01-26 Jakub Jelinek <jakub@redhat.com>
2767 * config/i386/i386.md (*addv<dwi>4_doubleword, *subv<dwi>4_doubleword):
2768 Use nonimmediate_operand instead of x86_64_hilo_general_operand and
2769 drop <di> from constraint of last operand.
2772 * config/i386/sse.md (*avx_vperm_broadcast_<mode>): Disallow for
2773 TARGET_AVX2 and V4DFmode not in the split condition, but in the
2774 pattern condition, though allow { 0, 0, 0, 0 } broadcast always.
2776 2020-01-25 Feng Xue <fxue@os.amperecomputing.com>
2779 * ipa-cp.c (get_info_about_necessary_edges): Remove value
2782 2020-01-24 Jeff Law <law@redhat.com>
2784 PR tree-optimization/92788
2785 * tree-ssa-threadedge.c (thread_across_edge): Check EDGE_COMPLEX
2788 2020-01-24 Jakub Jelinek <jakub@redhat.com>
2791 * config/i386/sse.md (*avx_vperm_broadcast_v4sf,
2792 *avx_vperm_broadcast_<mode>,
2793 <sse2_avx_avx512f>_vpermil<mode><mask_name>,
2794 *<sse2_avx_avx512f>_vpermilp<mode><mask_name>):
2795 Move before avx2_perm<mode>/avx512f_perm<mode>.
2798 * simplify-rtx.c (simplify_const_unary_operation,
2799 simplify_const_binary_operation): Punt for mode precision above
2800 MAX_BITSIZE_MODE_ANY_INT.
2802 2020-01-24 Andrew Pinski <apinski@marvell.com>
2804 * config/arm/aarch-cost-tables.h (cortexa57_extra_costs): Change
2807 2020-01-24 Jeff Law <law@redhat.com>
2810 * config/h8300/h8300.c (h8300_print_operand): Only call byte_reg
2811 for REGs. Call output_operand_lossage to get more reasonable
2814 2020-01-24 Andrew Stubbs <ams@codesourcery.com>
2816 * config/gcn/gcn-valu.md (vec_cmp<mode>di): Use
2817 gcn_fp_compare_operator.
2818 (vec_cmpu<mode>di): Use gcn_compare_operator.
2819 (vec_cmp<u>v64qidi): Use gcn_compare_operator.
2820 (vec_cmp<mode>di_exec): Use gcn_fp_compare_operator.
2821 (vec_cmpu<mode>di_exec): Use gcn_compare_operator.
2822 (vec_cmp<u>v64qidi_exec): Use gcn_compare_operator.
2823 (vec_cmp<mode>di_dup): Use gcn_fp_compare_operator.
2824 (vec_cmp<mode>di_dup_exec): Use gcn_fp_compare_operator.
2825 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): Use
2826 gcn_fp_compare_operator.
2827 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): Use
2828 gcn_fp_compare_operator.
2829 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): Use
2830 gcn_fp_compare_operator.
2831 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): Use
2832 gcn_fp_compare_operator.
2834 2020-01-24 Maciej W. Rozycki <macro@wdc.com>
2836 * doc/install.texi (Cross-Compiler-Specific Options): Document
2837 `--with-toolexeclibdir' option.
2839 2020-01-24 Hans-Peter Nilsson <hp@axis.com>
2841 * target.def (flags_regnum): Also mention effect on delay slot filling.
2842 * doc/tm.texi: Regenerate.
2844 2020-01-23 Jeff Law <law@redhat.com>
2846 PR translation/90162
2847 * config/h8300/h8300.c (h8300_option_override): Fix diagnostic text.
2849 2020-01-23 Mikael Tillenius <mti-1@tillenius.com>
2852 * config/h8300/h8300.h (FUNCTION_PROFILER): Fix emission of
2855 2020-01-23 Jakub Jelinek <jakub@redhat.com>
2857 PR rtl-optimization/93402
2858 * postreload.c (reload_combine_recognize_pattern): Don't try to adjust
2861 2020-01-23 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
2863 * config.in: Regenerated.
2864 * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to 1
2865 for TARGET_LIBC_GNUSTACK.
2866 * configure: Regenerated.
2867 * configure.ac: Define TARGET_LIBC_GNUSTACK if glibc version is
2868 found to be 2.31 or greater.
2870 2020-01-23 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
2872 * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to
2874 * config/mips/mips.c (TARGET_ASM_FILE_END): Define to ...
2875 (mips_asm_file_end): New function. Delegate to
2876 file_end_indicate_exec_stack if NEED_INDICATE_EXEC_STACK is true.
2877 * config/mips/mips.h (NEED_INDICATE_EXEC_STACK): Define to 0.
2879 2020-01-23 Jakub Jelinek <jakub@redhat.com>
2882 * config/i386/i386-modes.def (POImode): New mode.
2883 (MAX_BITSIZE_MODE_ANY_INT): Change from 128 to 160.
2884 * config/i386/i386.md (DPWI): New mode attribute.
2885 (addv<mode>4, subv<mode>4): Use <DPWI> instead of <DWI>.
2887 (QPWI): ... this. Use POI instead of OI for TImode.
2888 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1,
2889 *subv<dwi>4_doubleword, *subv<dwi>4_doubleword_1): Use <QPWI>
2892 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
2895 * config/aarch64/aarch64.md (UNSPEC_SPECULATION_TRACKER_REV): New
2897 (speculation_tracker_rev): New pattern.
2898 * config/aarch64/aarch64-speculation.cc (aarch64_do_track_speculation):
2899 Use speculation_tracker_rev to track the inverse condition.
2901 2020-01-23 Richard Biener <rguenther@suse.de>
2903 PR tree-optimization/93381
2904 * tree-ssa-sccvn.c (vn_walk_cb_data::push_partial_def): Take
2905 alias-set of the def as argument and record the first one.
2906 (vn_walk_cb_data::first_set): New member.
2907 (vn_reference_lookup_3): Pass the alias-set of the current def
2908 to push_partial_def. Fix alias-set used in the aggregate copy
2910 (vn_reference_lookup): Consistently set *last_vuse_ptr.
2911 * real.c (clear_significand_below): Fix out-of-bound access.
2913 2020-01-23 Jakub Jelinek <jakub@redhat.com>
2916 * config/i386/i386.md (*bmi2_bzhi_<mode>3_2, *bmi2_bzhi_<mode>3_3):
2917 New define_insn patterns.
2919 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
2921 * doc/sourcebuild.texi (check-function-bodies): Add an
2922 optional target/xfail selector.
2924 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
2926 PR rtl-optimization/93124
2927 * auto-inc-dec.c (merge_in_block): Don't add auto inc/decs to
2928 bare USE and CLOBBER insns.
2930 2020-01-22 Andrew Pinski <apinski@marvell.com>
2932 * config/arc/arc.c (output_short_suffix): Check insn for nullness.
2934 2020-01-22 David Malcolm <dmalcolm@redhat.com>
2937 * gdbinit.in (break-on-saved-diagnostic): Update for move of
2938 diagnostic_manager into "ana" namespace.
2939 * selftest-run-tests.c (selftest::run_tests): Update for move of
2940 selftest::run_analyzer_selftests to
2941 ana::selftest::run_analyzer_selftests.
2943 2020-01-22 Richard Sandiford <richard.sandiford@arm.com>
2945 * cfgexpand.c (union_stack_vars): Update the size.
2947 2020-01-22 Richard Biener <rguenther@suse.de>
2949 PR tree-optimization/93381
2950 * tree-ssa-structalias.c (find_func_aliases): Assume offsetting
2951 throughout, handle all conversions the same.
2953 2020-01-22 Jakub Jelinek <jakub@redhat.com>
2956 * config/aarch64/aarch64.c (aarch64_expand_subvti): Only use
2957 gen_subdi3_compare1_imm if low_in2 satisfies aarch64_plus_immediate
2958 predicate, not whenever it is CONST_INT. Otherwise, force_reg it.
2959 Call force_reg on high_in2 unconditionally.
2961 2020-01-22 Martin Liska <mliska@suse.cz>
2963 PR tree-optimization/92924
2964 * profile.c (compute_value_histograms): Divide
2967 2020-01-22 Jakub Jelinek <jakub@redhat.com>
2970 * output.h (assemble_name_resolve): Declare.
2971 * varasm.c (assemble_name_resolve): New function.
2972 (assemble_name): Use it.
2973 * config/i386/i386.h (ASM_OUTPUT_SYMBOL_REF): Define.
2975 2020-01-22 Joseph Myers <joseph@codesourcery.com>
2977 * doc/sourcebuild.texi (Texinfo Manuals, Front End): Refer to
2978 update_web_docs_git instead of update_web_docs_svn.
2980 2020-01-21 Andrew Pinski <apinski@marvell.com>
2983 * config/aarch64/aarch64.md (tlsgd_small_<mode>): Have operand 0
2984 as PTR mode. Have operand 1 as being modeless, it can be P mode.
2985 (*tlsgd_small_<mode>): Likewise.
2986 * config/aarch64/aarch64.c (aarch64_load_symref_appropriately)
2987 <case SYMBOL_SMALL_TLSGD>: Call gen_tlsgd_small_* with a ptr_mode
2988 register. Convert that register back to dest using convert_mode.
2990 2020-01-21 Jim Wilson <jimw@sifive.com>
2992 * config/riscv/riscv-sr.c (riscv_sr_match_prologue): Use INTVAL
2995 2020-01-21 H.J. Lu <hongjiu.lu@intel.com>
2996 Uros Bizjak <ubizjak@gmail.com>
2999 * config/i386/i386.c (ix86_tls_module_base): Replace Pmode
3001 (legitimize_tls_address): Do GNU2 TLS address computation in
3002 ptr_mode and zero-extend result to Pmode.
3003 * config/i386/i386.md (@tls_dynamic_gnu2_64_<mode>): Replace
3004 :P with :PTR and Pmode with ptr_mode.
3005 (*tls_dynamic_gnu2_lea_64_<mode>): Likewise.
3006 (*tls_dynamic_gnu2_call_64_<mode>): Likewise.
3007 (*tls_dynamic_gnu2_combine_64_<mode>): Likewise.
3009 2020-01-21 Jakub Jelinek <jakub@redhat.com>
3012 * config/riscv/riscv.c (riscv_rtx_costs) <case ZERO_EXTRACT>: Verify
3013 the last two operands are CONST_INT_P before using them as such.
3015 2020-01-21 Richard Sandiford <richard.sandiford@arm.com>
3017 * config/aarch64/aarch64-sve-builtins.def: Use get_typenode_from_name
3018 to get the integer element types.
3020 2020-01-21 Richard Sandiford <richard.sandiford@arm.com>
3022 * config/aarch64/aarch64-sve-builtins.h
3023 (function_expander::convert_to_pmode): Declare.
3024 * config/aarch64/aarch64-sve-builtins.cc
3025 (function_expander::convert_to_pmode): New function.
3026 (function_expander::get_contiguous_base): Use it.
3027 (function_expander::prepare_gather_address_operands): Likewise.
3028 * config/aarch64/aarch64-sve-builtins-sve2.cc
3029 (svwhilerw_svwhilewr_impl::expand): Likewise.
3031 2020-01-21 Szabolcs Nagy <szabolcs.nagy@arm.com>
3034 * config/aarch64/aarch64.c (aarch64_declare_function_name): Set
3035 cfun->machine->label_is_assembled.
3036 (aarch64_print_patchable_function_entry): New.
3037 (TARGET_ASM_PRINT_PATCHABLE_FUNCTION_ENTRY): Define.
3038 * config/aarch64/aarch64.h (struct machine_function): New field,
3041 2020-01-21 David Malcolm <dmalcolm@redhat.com>
3044 * ipa-profile.c (ipa_profile): Delete call_sums and set it to
3047 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
3050 * cgraph.c (cgraph_edge::resolve_speculation,
3051 cgraph_edge::redirect_call_stmt_to_callee): Fix update of
3052 call_stmt_site_hash.
3054 2020-01-21 Martin Liska <mliska@suse.cz>
3056 * config/rs6000/rs6000.c (common_mode_defined): Remove
3059 2020-01-21 Richard Biener <rguenther@suse.de>
3061 PR tree-optimization/92328
3062 * tree-ssa-sccvn.c (vn_reference_lookup_3): Preserve
3063 type when value-numbering same-sized store by inserting a
3065 (eliminate_dom_walker::eliminate_stmt): When eliminating
3066 a redundant store handle bit-reinterpretation of the same value.
3068 2020-01-21 Andrew Pinski <apinski@marvel.com>
3071 * tree-into-ssa.c (prepare_block_for_update_1): Split out
3073 (prepare_block_for_update): This. Use a worklist instead of
3076 2020-01-21 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3078 * gcc/config/arm/arm.c (clear_operation_p):
3079 Initialise last_regno, skip first iteration
3080 based on the first_set value and use ints instead
3081 of the unnecessary HOST_WIDE_INTs.
3083 2020-01-21 Jakub Jelinek <jakub@redhat.com>
3086 * config/rs6000/rs6000.c (rs6000_emit_cmove): If using fsel, punt for
3087 compare_mode other than SFmode or DFmode.
3089 2020-01-21 Kito Cheng <kito.cheng@sifive.com>
3092 * config/riscv/riscv-protos.h (riscv_hard_regno_rename_ok): New.
3093 * config/riscv/riscv.c (riscv_hard_regno_rename_ok): New.
3094 * config/riscv/riscv.h (HARD_REGNO_RENAME_OK): Defined.
3096 2020-01-20 Wilco Dijkstra <wdijkstr@arm.com>
3098 * config/aarch64/aarch64.c (neoversen1_tunings): Set jump_align to 4.
3100 2020-01-20 Andrew Pinski <apinski@marvell.com>
3103 * targhooks.c (default_print_patchable_function_entry): Use
3104 output_asm_insn to emit the nop instruction.
3106 2020-01-20 Fangrui Song <maskray@google.com>
3109 * targhooks.c (default_print_patchable_function_entry): Align to
3112 2020-01-20 H.J. Lu <hongjiu.lu@intel.com>
3115 * config/i386/i386.c (legitimize_tls_address): Pass Pmode to
3116 gen_tls_dynamic_gnu2_64. Compute GNU2 TLS address in ptr_mode.
3117 * config/i386/i386.md (tls_dynamic_gnu2_64): Renamed to ...
3118 (@tls_dynamic_gnu2_64_<mode>): This. Replace DI with P.
3119 (*tls_dynamic_gnu2_lea_64): Renamed to ...
3120 (*tls_dynamic_gnu2_lea_64_<mode>): This. Replace DI with P.
3121 Remove the {q} suffix from lea.
3122 (*tls_dynamic_gnu2_call_64): Renamed to ...
3123 (*tls_dynamic_gnu2_call_64_<mode>): This. Replace DI with P.
3124 (*tls_dynamic_gnu2_combine_64): Renamed to ...
3125 (*tls_dynamic_gnu2_combine_64_<mode>): This. Replace DI with P.
3126 Pass Pmode to gen_tls_dynamic_gnu2_64.
3128 2020-01-20 Wilco Dijkstra <wdijkstr@arm.com>
3130 * config/aarch64/aarch64.h (SLOW_BYTE_ACCESS): Set to 1.
3132 2020-01-20 Richard Sandiford <richard.sandiford@arm.com>
3134 * config/aarch64/aarch64-sve-builtins-base.cc
3135 (svld1ro_impl::memory_vector_mode): Remove parameter name.
3137 2020-01-20 Richard Biener <rguenther@suse.de>
3140 * dwarf2out.c (prune_unused_types): Unconditionally mark
3141 called function DIEs.
3143 2020-01-20 Martin Liska <mliska@suse.cz>
3145 PR tree-optimization/93199
3146 * tree-eh.c (struct leh_state): Add
3147 new field outer_non_cleanup.
3148 (cleanup_is_dead_in): Pass leh_state instead
3149 of eh_region. Add a checking that state->outer_non_cleanup
3150 points to outer non-clean up region.
3151 (lower_try_finally): Record outer_non_cleanup
3153 (lower_catch): Likewise.
3154 (lower_eh_filter): Likewise.
3155 (lower_eh_must_not_throw): Likewise.
3156 (lower_cleanup): Likewise.
3158 2020-01-20 Richard Biener <rguenther@suse.de>
3160 PR tree-optimization/93094
3161 * tree-vectorizer.h (vect_loop_versioning): Adjust.
3162 (vect_transform_loop): Likewise.
3163 * tree-vectorizer.c (try_vectorize_loop_1): Pass down
3164 loop_vectorized_call to vect_transform_loop.
3165 * tree-vect-loop.c (vect_transform_loop): Pass down
3166 loop_vectorized_call to vect_loop_versioning.
3167 * tree-vect-loop-manip.c (vect_loop_versioning): Use
3168 the earlier discovered loop_vectorized_call.
3170 2020-01-19 Eric S. Raymond <esr@thyrsus.com>
3172 * doc/contribute.texi: Update for SVN -> Git transition.
3173 * doc/install.texi: Likewise.
3175 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
3178 * cgraph.c (cgraph_edge::make_speculative): Increase number of
3179 speculative targets.
3180 (verify_speculative_call): New function
3181 (cgraph_node::verify_node): Use it.
3182 * ipa-profile.c (ipa_profile): Fix formating; do not set number of
3185 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
3188 * cgraph.c (cgraph_edge::resolve_speculation): Fix foramting.
3189 (cgraph_edge::make_direct): Remove all indirect targets.
3190 (cgraph_edge::redirect_call_stmt_to_callee): Use make_direct..
3191 (cgraph_node::verify_node): Verify that only one call_stmt or
3192 lto_stmt_uid is set.
3193 * cgraphclones.c (cgraph_edge::clone): Set only one call_stmt or
3195 * lto-cgraph.c (lto_output_edge): Simplify streaming of stmt.
3196 (lto_output_ref): Simplify streaming of stmt.
3197 * lto-streamer-in.c (fixup_call_stmt_edges_1): Clear lto_stmt_uid.
3199 2020-01-18 Tamar Christina <tamar.christina@arm.com>
3201 * config/aarch64/aarch64-sve-builtins-base.cc (memory_vector_mode):
3202 Mark parameter unused.
3204 2020-01-18 Hans-Peter Nilsson <hp@axis.com>
3206 * config.gcc <obsolete targets>: Add crisv32-*-* and cris-*-linux*
3208 2019-01-18 Gerald Pfeifer <gerald@pfeifer.com>
3210 * varpool.c (ctor_useable_for_folding_p): Fix grammar.
3212 2020-01-18 Iain Sandoe <iain@sandoe.co.uk>
3214 * Makefile.in: Add coroutine-passes.o.
3215 * builtin-types.def (BT_CONST_SIZE): New.
3216 (BT_FN_BOOL_PTR): New.
3217 (BT_FN_PTR_PTR_CONST_SIZE_BOOL): New.
3218 * builtins.def (DEF_COROUTINE_BUILTIN): New.
3219 * coroutine-builtins.def: New file.
3220 * coroutine-passes.cc: New file.
3221 * function.h (struct GTY function): Add a bit to indicate that the
3222 function is a coroutine component.
3223 * internal-fn.c (expand_CO_FRAME): New.
3224 (expand_CO_YIELD): New.
3225 (expand_CO_SUSPN): New.
3226 (expand_CO_ACTOR): New.
3227 * internal-fn.def (CO_ACTOR): New.
3231 * passes.def: Add pass_coroutine_lower_builtins,
3232 pass_coroutine_early_expand_ifns.
3233 * tree-pass.h (make_pass_coroutine_lower_builtins): New.
3234 (make_pass_coroutine_early_expand_ifns): New.
3235 * doc/invoke.texi: Document the fcoroutines command line
3238 2020-01-18 Jakub Jelinek <jakub@redhat.com>
3240 * config/arm/vfp.md (*clear_vfp_multiple): Remove unused variable.
3243 * config/arm/arm.c (clear_operation_p): Don't use REGNO until
3244 after checking the argument is a REG. Don't use REGNO (reg)
3245 again to set last_regno, reuse regno variable instead.
3247 2020-01-17 David Malcolm <dmalcolm@redhat.com>
3249 * doc/analyzer.texi (Limitations): Add note about NaN.
3251 2020-01-17 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3252 Sudakshina Das <sudi.das@arm.com>
3254 * config/arm/arm.md (ashldi3): Generate thumb2_lsll for both reg
3255 and valid immediate.
3256 (ashrdi3): Generate thumb2_asrl for both reg and valid immediate.
3257 (lshrdi3): Generate thumb2_lsrl for valid immediates.
3258 * config/arm/constraints.md (Pg): New.
3259 * config/arm/predicates.md (long_shift_imm): New.
3260 (arm_reg_or_long_shift_imm): Likewise.
3261 * config/arm/thumb2.md (thumb2_asrl): New immediate alternative.
3262 (thumb2_lsll): Likewise.
3265 2020-01-17 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3266 Sudakshina Das <sudi.das@arm.com>
3268 * config/arm/arm.md (ashldi3): Generate thumb2_lsll for TARGET_HAVE_MVE.
3269 (ashrdi3): Generate thumb2_asrl for TARGET_HAVE_MVE.
3270 * config/arm/arm.c (arm_hard_regno_mode_ok): Allocate even odd
3271 register pairs for doubleword quantities for ARMv8.1M-Mainline.
3272 * config/arm/thumb2.md (thumb2_asrl): New.
3273 (thumb2_lsll): Likewise.
3275 2020-01-17 Jakub Jelinek <jakub@redhat.com>
3277 * config/arm/arm.c (cmse_nonsecure_call_inline_register_clear): Remove
3280 2020-01-17 Alexander Monakov <amonakov@ispras.ru>
3282 * gdbinit.in (help-gcc-hooks): New command.
3283 (pp, pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, ptc, pdn, ptn, pdd, prc,
3284 pi, pbm, pel, trt): Take $arg0 instead of $ if supplied. Update
3287 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
3289 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use the
3290 correct target macro.
3292 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
3294 * config/aarch64/aarch64-protos.h
3295 (aarch64_sve_ld1ro_operand_p): New.
3296 * config/aarch64/aarch64-sve-builtins-base.cc
3297 (class load_replicate): New.
3298 (class svld1ro_impl): New.
3299 (class svld1rq_impl): Change to inherit from load_replicate.
3300 (svld1ro): New sve intrinsic function base.
3301 * config/aarch64/aarch64-sve-builtins-base.def (svld1ro):
3302 New DEF_SVE_FUNCTION.
3303 * config/aarch64/aarch64-sve-builtins-base.h
3304 (svld1ro): New decl.
3305 * config/aarch64/aarch64-sve-builtins.cc
3306 (function_expander::add_mem_operand): Modify assert to allow
3308 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): New
3310 * config/aarch64/aarch64.c
3311 (aarch64_sve_ld1rq_operand_p): Implement in terms of ...
3312 (aarch64_sve_ld1rq_ld1ro_operand_p): This.
3313 (aarch64_sve_ld1ro_operand_p): New.
3314 * config/aarch64/aarch64.md (UNSPEC_LD1RO): New unspec.
3315 * config/aarch64/constraints.md (UOb,UOh,UOw,UOd): New.
3316 * config/aarch64/predicates.md
3317 (aarch64_sve_ld1ro_operand_{b,h,w,d}): New.
3319 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
3321 * config/aarch64/aarch64-c.c (_ARM_FEATURE_MATMUL_FLOAT64):
3322 Introduce this ACLE specified predefined macro.
3323 * config/aarch64/aarch64-option-extensions.def (f64mm): New.
3324 (fp): Disabling this disables f64mm.
3325 (simd): Disabling this disables f64mm.
3326 (fp16): Disabling this disables f64mm.
3327 (sve): Disabling this disables f64mm.
3328 * config/aarch64/aarch64.h (AARCH64_FL_F64MM): New.
3329 (AARCH64_ISA_F64MM): New.
3330 (TARGET_F64MM): New.
3331 * doc/invoke.texi (f64mm): Document new option.
3333 2020-01-17 Wilco Dijkstra <wdijkstr@arm.com>
3335 * config/aarch64/aarch64.c (generic_tunings): Add branch fusion.
3336 (neoversen1_tunings): Likewise.
3338 2020-01-17 Wilco Dijkstra <wdijkstr@arm.com>
3341 * config/aarch64/aarch64.c (aarch64_split_compare_and_swap)
3342 Add assert to ensure prolog has been emitted.
3343 (aarch64_split_atomic_op): Likewise.
3344 * config/aarch64/atomics.md (aarch64_compare_and_swap<mode>)
3345 Use epilogue_completed rather than reload_completed.
3346 (aarch64_atomic_exchange<mode>): Likewise.
3347 (aarch64_atomic_<atomic_optab><mode>): Likewise.
3348 (atomic_nand<mode>): Likewise.
3349 (aarch64_atomic_fetch_<atomic_optab><mode>): Likewise.
3350 (atomic_fetch_nand<mode>): Likewise.
3351 (aarch64_atomic_<atomic_optab>_fetch<mode>): Likewise.
3352 (atomic_nand_fetch<mode>): Likewise.
3354 2020-01-17 Richard Sandiford <richard.sandiford@arm.com>
3357 * config/aarch64/aarch64.h (REVERSIBLE_CC_MODE): Return false
3359 (REVERSE_CONDITION): Delete.
3360 * config/aarch64/iterators.md (CC_ONLY): New mode iterator.
3361 (CCFP_CCFPE): Likewise.
3362 (e): New mode attribute.
3363 * config/aarch64/aarch64.md (ccmp<GPI:mode>): Rename to...
3364 (@ccmp<CC_ONLY:mode><GPI:mode>): ...this, using CC_ONLY instead of CC.
3365 (fccmp<GPF:mode>, fccmpe<GPF:mode>): Merge into...
3366 (@ccmp<CCFP_CCFPE:mode><GPF:mode>): ...this combined pattern.
3367 (@ccmp<CC_ONLY:mode><GPI:mode>_rev): New pattern.
3368 (@ccmp<CCFP_CCFPE:mode><GPF:mode>_rev): Likewise.
3369 * config/aarch64/aarch64.c (aarch64_gen_compare_reg): Update
3370 name of generator from gen_ccmpdi to gen_ccmpccdi.
3371 (aarch64_gen_ccmp_next): Use code_for_ccmp. If we want to reverse
3372 the previous comparison but aren't able to, use the new ccmp_rev
3375 2020-01-17 Richard Sandiford <richard.sandiford@arm.com>
3377 * gimplify.c (gimplify_return_expr): Use poly_int_tree_p rather
3378 than testing directly for INTEGER_CST.
3379 (gimplify_target_expr, gimplify_omp_depend): Likewise.
3381 2020-01-17 Jakub Jelinek <jakub@redhat.com>
3383 PR tree-optimization/93292
3384 * tree-vect-stmts.c (vectorizable_comparison): Punt also if
3385 get_vectype_for_scalar_type returns NULL.
3387 2020-01-16 Jan Hubicka <hubicka@ucw.cz>
3389 * params.opt (-param=max-predicted-iterations): Increase range from 0.
3390 * predict.c (estimate_loops): Add 1 to param_max_predicted_iterations.
3392 2020-01-16 Jan Hubicka <hubicka@ucw.cz>
3394 * ipa-fnsummary.c (estimate_calls_size_and_time): Fix formating of
3396 * params.opt: (max-predicted-iterations): Set bounds.
3397 * predict.c (real_almost_one, real_br_prob_base,
3398 real_inv_br_prob_base, real_one_half, real_bb_freq_max): Remove.
3399 (propagate_freq): Add max_cyclic_prob parameter; cap cyclic
3400 probabilities; do not truncate to reg_br_prob_bases.
3401 (estimate_loops_at_level): Pass max_cyclic_prob.
3402 (estimate_loops): Compute max_cyclic_prob.
3403 (estimate_bb_frequencies): Do not initialize real_*; update calculation
3405 * profile-count.c (profile_probability::to_sreal): New.
3406 * profile-count.h (class sreal): Move up in file.
3407 (profile_probability::to_sreal): Declare.
3409 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
3412 (arm_invalid_conversion): New function for target hook.
3413 (arm_invalid_unary_op): New function for target hook.
3414 (arm_invalid_binary_op): New function for target hook.
3416 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
3418 * config.gcc: Add arm_bf16.h.
3419 * config/arm/arm-builtins.c (arm_mangle_builtin_type): Fix comment.
3420 (arm_simd_builtin_std_type): Add BFmode.
3421 (arm_init_simd_builtin_types): Define element types for vector types.
3422 (arm_init_bf16_types): New function.
3423 (arm_init_builtins): Add arm_init_bf16_types function call.
3424 * config/arm/arm-modes.def: Add BFmode and V4BF, V8BF vector modes.
3425 * config/arm/arm-simd-builtin-types.def: Add V4BF, V8BF.
3426 * config/arm/arm.c (aapcs_vfp_sub_candidate): Add BFmode.
3427 (arm_hard_regno_mode_ok): Add BFmode and tidy up statements.
3428 (arm_vector_mode_supported_p): Add V4BF, V8BF.
3429 (arm_mangle_type): Add __bf16.
3430 * config/arm/arm.h: Add V4BF, V8BF to VALID_NEON_DREG_MODE,
3431 VALID_NEON_QREG_MODE respectively. Add export arm_bf16_type_node,
3432 arm_bf16_ptr_type_node.
3433 * config/arm/arm.md: Add BFmode to movhf expand, mov pattern and
3434 define_split between ARM registers.
3435 * config/arm/arm_bf16.h: New file.
3436 * config/arm/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
3437 * config/arm/iterators.md: (ANY64_BF, VDXMOV, VHFBF, HFBF, fporbf): New.
3439 * config/arm/neon.md: Add BF vector types to movhf NEON move patterns.
3440 * config/arm/vfp.md: Add BFmode to movhf patterns.
3442 2020-01-16 Mihail Ionescu <mihail.ionescu@arm.com>
3443 Andre Vieira <andre.simoesdiasvieira@arm.com>
3445 * config/arm/arm-cpus.in (mve, mve_float): New features.
3446 (dsp, mve, mve.fp): New options.
3447 * config/arm/arm.h (TARGET_HAVE_MVE, TARGET_HAVE_MVE_FLOAT): Define.
3448 * config/arm/t-rmprofile: Map v8.1-M multilibs to v8-M.
3449 * doc/invoke.texi: Document the armv8.1-m mve and dps options.
3451 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3452 Thomas Preud'homme <thomas.preudhomme@arm.com>
3454 * config/arm/arm-cpus.in (ARMv8_1m_main): Redefine as an extension to
3456 * config/arm/arm.c (arm_options_perform_arch_sanity_checks): Remove
3457 error for using -mcmse when targeting Armv8.1-M Mainline.
3459 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3460 Thomas Preud'homme <thomas.preudhomme@arm.com>
3462 * config/arm/arm.md (nonsecure_call_internal): Do not force memory
3463 address in r4 when targeting Armv8.1-M Mainline.
3464 (nonsecure_call_value_internal): Likewise.
3465 * config/arm/thumb2.md (nonsecure_call_reg_thumb2): Make memory address
3466 a register match_operand again. Emit BLXNS when targeting
3468 (nonsecure_call_value_reg_thumb2): Likewise.
3470 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3471 Thomas Preud'homme <thomas.preudhomme@arm.com>
3473 * config/arm/arm.c (arm_add_cfa_adjust_cfa_note): Declare early.
3474 (cmse_nonsecure_call_inline_register_clear): Define new lazy_fpclear
3475 variable as true when floating-point ABI is not hard. Replace
3476 check against TARGET_HARD_FLOAT_ABI by checks against lazy_fpclear.
3477 Generate VLSTM and VLLDM instruction respectively before and
3478 after a function call to cmse_nonsecure_call function.
3479 * config/arm/unspecs.md (VUNSPEC_VLSTM): Define unspec.
3480 (VUNSPEC_VLLDM): Likewise.
3481 * config/arm/vfp.md (lazy_store_multiple_insn): New define_insn.
3482 (lazy_load_multiple_insn): Likewise.
3484 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3485 Thomas Preud'homme <thomas.preudhomme@arm.com>
3487 * config/arm/arm.c (vfp_emit_fstmd): Declare early.
3488 (arm_emit_vfp_multi_reg_pop): Likewise.
3489 (cmse_nonsecure_call_inline_register_clear): Abstract number of VFP
3490 registers to clear in max_fp_regno. Emit VPUSH and VPOP to save and
3491 restore callee-saved VFP registers.
3493 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3494 Thomas Preud'homme <thomas.preudhomme@arm.com>
3496 * config/arm/arm.c (arm_emit_multi_reg_pop): Declare early.
3497 (cmse_nonsecure_call_clear_caller_saved): Rename into ...
3498 (cmse_nonsecure_call_inline_register_clear): This. Save and clear
3499 callee-saved GPRs as well as clear ip register before doing a nonsecure
3500 call then restore callee-saved GPRs after it when targeting
3502 (arm_reorg): Adapt to function rename.
3504 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3505 Thomas Preud'homme <thomas.preudhomme@arm.com>
3507 * config/arm/arm-protos.h (clear_operation_p): Adapt prototype.
3508 * config/arm/arm.c (clear_operation_p): Extend to be able to check a
3509 clear_vfp_multiple pattern based on a new vfp parameter.
3510 (cmse_clear_registers): Generate VSCCLRM to clear VFP registers when
3511 targeting Armv8.1-M Mainline.
3512 (cmse_nonsecure_entry_clear_before_return): Clear VFP registers
3513 unconditionally when targeting Armv8.1-M Mainline architecture. Check
3514 whether VFP registers are available before looking call_used_regs for a
3516 * config/arm/predicates.md (clear_multiple_operation): Adapt to change
3517 of prototype of clear_operation_p.
3518 (clear_vfp_multiple_operation): New predicate.
3519 * config/arm/unspecs.md (VUNSPEC_VSCCLRM_VPR): New volatile unspec.
3520 * config/arm/vfp.md (clear_vfp_multiple): New define_insn.
3522 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3523 Thomas Preud'homme <thomas.preudhomme@arm.com>
3525 * config/arm/arm-protos.h (clear_operation_p): Declare.
3526 * config/arm/arm.c (clear_operation_p): New function.
3527 (cmse_clear_registers): Generate clear_multiple instruction pattern if
3528 targeting Armv8.1-M Mainline or successor.
3529 (output_return_instruction): Only output APSR register clearing if
3530 Armv8.1-M Mainline instructions not available.
3531 (thumb_exit): Likewise.
3532 * config/arm/predicates.md (clear_multiple_operation): New predicate.
3533 * config/arm/thumb2.md (clear_apsr): New define_insn.
3534 (clear_multiple): Likewise.
3535 * config/arm/unspecs.md (VUNSPEC_CLRM_APSR): New volatile unspec.
3537 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3538 Thomas Preud'homme <thomas.preudhomme@arm.com>
3540 * config/arm/arm.c (fp_sysreg_names): Declare and define.
3541 (use_return_insn): Also return false for Armv8.1-M Mainline.
3542 (output_return_instruction): Skip FPSCR clearing if Armv8.1-M
3543 Mainline instructions are available.
3544 (arm_compute_frame_layout): Allocate space in frame for FPCXTNS
3545 when targeting Armv8.1-M Mainline Security Extensions.
3546 (arm_expand_prologue): Save FPCXTNS if this is an Armv8.1-M
3547 Mainline entry function.
3548 (cmse_nonsecure_entry_clear_before_return): Clear IP and r4 if
3549 targeting Armv8.1-M Mainline or successor.
3550 (arm_expand_epilogue): Fix indentation of caller-saved register
3551 clearing. Restore FPCXTNS if this is an Armv8.1-M Mainline
3553 * config/arm/arm.h (TARGET_HAVE_FP_CMSE): New macro.
3554 (FP_SYSREGS): Likewise.
3555 (enum vfp_sysregs_encoding): Define enum.
3556 (fp_sysreg_names): Declare.
3557 * config/arm/unspecs.md (VUNSPEC_VSTR_VLDR): New volatile unspec.
3558 * config/arm/vfp.md (push_fpsysreg_insn): New define_insn.
3559 (pop_fpsysreg_insn): Likewise.
3561 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3562 Thomas Preud'homme <thomas.preudhomme@arm.com>
3564 * config/arm/arm-cpus.in (armv8_1m_main): New feature.
3565 (ARMv4, ARMv4t, ARMv5t, ARMv5te, ARMv5tej, ARMv6, ARMv6j, ARMv6k,
3566 ARMv6z, ARMv6kz, ARMv6zk, ARMv6t2, ARMv6m, ARMv7, ARMv7a, ARMv7ve,
3567 ARMv7r, ARMv7m, ARMv7em, ARMv8a, ARMv8_1a, ARMv8_2a, ARMv8_3a,
3568 ARMv8_4a, ARMv8_5a, ARMv8m_base, ARMv8m_main, ARMv8r): Reindent.
3569 (ARMv8_1m_main): New feature group.
3570 (armv8.1-m.main): New architecture.
3571 * config/arm/arm-tables.opt: Regenerate.
3572 * config/arm/arm.c (arm_arch8_1m_main): Define and default initialize.
3573 (arm_option_reconfigure_globals): Initialize arm_arch8_1m_main.
3574 (arm_options_perform_arch_sanity_checks): Error out when targeting
3575 Armv8.1-M Mainline Security Extensions.
3576 * config/arm/arm.h (arm_arch8_1m_main): Declare.
3578 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
3580 * config/aarch64/aarch64-simd-builtins.def (aarch64_bfdot,
3581 aarch64_bfdot_lane, aarch64_bfdot_laneq): New.
3582 * config/aarch64/aarch64-simd.md (aarch64_bfdot, aarch64_bfdot_lane,
3583 aarch64_bfdot_laneq): New.
3584 * config/aarch64/arm_bf16.h (vbfdot_f32, vbfdotq_f32,
3585 vbfdot_lane_f32, vbfdotq_lane_f32, vbfdot_laneq_f32,
3586 vbfdotq_laneq_f32): New.
3587 * config/aarch64/iterators.md (UNSPEC_BFDOT, Vbfdottype,
3588 VBFMLA_W, VBF): New.
3589 (isquadop): Add V4BF, V8BF.
3591 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
3593 * config/aarch64/aarch64-builtins.c: (enum aarch64_type_qualifiers):
3594 New qualifier_lane_quadtup_index, TYPES_TERNOP_SSUS,
3595 TYPES_QUADOPSSUS_LANE_QUADTUP, TYPES_QUADOPSSSU_LANE_QUADTUP.
3596 (aarch64_simd_expand_args): Add case SIMD_ARG_LANE_QUADTUP_INDEX.
3597 (aarch64_simd_expand_builtin): Add qualifier_lane_quadtup_index.
3598 * config/aarch64/aarch64-simd-builtins.def (usdot, usdot_lane,
3599 usdot_laneq, sudot_lane,sudot_laneq): New.
3600 * config/aarch64/aarch64-simd.md (aarch64_usdot): New.
3601 (aarch64_<sur>dot_lane): New.
3602 * config/aarch64/arm_neon.h (vusdot_s32): New.
3604 (vusdot_lane_s32): New.
3605 (vsudot_lane_s32): New.
3606 * config/aarch64/iterators.md (DOTPROD_I8MM): New iterator.
3607 (UNSPEC_USDOT, UNSPEC_SUDOT): New unspecs.
3609 2020-01-16 Martin Liska <mliska@suse.cz>
3611 * value-prof.c (dump_histogram_value): Fix
3612 obvious spacing issue.
3614 2020-01-16 Andrew Pinski <apinski@marvell.com>
3616 * tree-ssa-sccvn.c(vn_reference_lookup_3): Check lhs for
3617 !storage_order_barrier_p.
3619 2020-01-16 Andrew Pinski <apinski@marvell.com>
3621 * sched-int.h (_dep): Add unused bit-field field for the padding.
3622 * sched-deps.c (init_dep_1): Init unused field.
3624 2020-01-16 Andrew Pinski <apinski@marvell.com>
3626 * optabs.h (create_expand_operand): Initialize target field also.
3628 2020-01-16 Andre Vieira <andre.simoesdiasvieira@arm.com>
3630 PR tree-optimization/92429
3631 * tree-ssa-loop-niter.h (simplify_replace_tree): Add parameter.
3632 * tree-ssa-loop-niter.c (simplify_replace_tree): Add parameter to
3634 * tree-vect-loop.c (update_epilogue_vinfo): Do not fold when replacing
3637 2020-01-16 Richard Sandiford <richard.sandiford@arm.com>
3639 * config/aarch64/aarch64.c (aarch64_split_sve_subreg_move): Apply
3640 aarch64_sve_int_mode to each mode.
3642 2020-01-15 David Malcolm <dmalcolm@redhat.com>
3644 * doc/analyzer.texi (Overview): Add note about
3645 -fdump-ipa-analyzer.
3647 2020-01-15 Wilco Dijkstra <wdijkstr@arm.com>
3649 PR tree-optimization/93231
3650 * tree-ssa-forwprop.c (optimize_count_trailing_zeroes): Check
3651 input_type is unsigned. Use tree_to_shwi for shift constant.
3652 Check CST_STRING element size is CHAR_TYPE_SIZE bits.
3653 (simplify_count_trailing_zeroes): Add test to handle known non-zero
3654 inputs more efficiently.
3656 2020-01-15 Uroš Bizjak <ubizjak@gmail.com>
3658 * config/i386/i386.md (*movsf_internal): Do not require
3659 SSE2 ISA for alternatives 14 and 15.
3661 2020-01-15 Richard Biener <rguenther@suse.de>
3664 * tree-eh.c (sink_clobbers): If we already visited the destination
3665 block do not defer insertion.
3666 (pass_lower_eh_dispatch::execute): Maintain BB_VISITED for
3667 the purpose of defered insertion.
3669 2020-01-15 Jakub Jelinek <jakub@redhat.com>
3671 * BASE-VER: Bump to 10.0.1.
3673 2020-01-15 Richard Sandiford <richard.sandiford@arm.com>
3675 PR tree-optimization/93247
3676 * tree-vect-loop.c (update_epilogue_loop_vinfo): Check the access
3677 type of the stmt that we're going to vectorize.
3679 2020-01-15 Richard Sandiford <richard.sandiford@arm.com>
3681 * tree-vect-slp.c (vectorize_slp_instance_root_stmt): Use a
3682 VIEW_CONVERT_EXPR if the vectorized constructor has a diffeent
3685 2020-01-15 Martin Liska <mliska@suse.cz>
3687 * ipa-profile.c (ipa_profile_read_edge_summary): Do not allow
3688 2 calls of streamer_read_hwi in a function call.
3690 2020-01-15 Richard Biener <rguenther@suse.de>
3692 * alias.c (record_alias_subset): Avoid redundant work when
3693 subset is already recorded.
3695 2020-01-14 David Malcolm <dmalcolm@redhat.com>
3697 * doc/invoke.texi (-fdiagnostics-show-cwe): Add note that some of
3698 the analyzer options provide CWE identifiers.
3700 2020-01-14 David Malcolm <dmalcolm@redhat.com>
3702 * tree-diagnostic-path.cc (path_summary::event_range::print):
3703 When testing for UNKNOWN_LOCATION, look through ad-hoc wrappers
3704 using get_pure_location.
3706 2020-01-15 Jakub Jelinek <jakub@redhat.com>
3708 PR tree-optimization/93262
3709 * tree-ssa-dse.c (maybe_trim_memstar_call): For *_chk builtins,
3710 perform head trimming only if the last argument is constant,
3711 either all ones, or larger or equal to head trim, in the latter
3712 case decrease the last argument by head_trim.
3714 PR tree-optimization/93249
3715 * tree-ssa-dse.c: Include builtins.h and gimple-fold.h.
3716 (maybe_trim_memstar_call): Move head_trim and tail_trim vars to
3717 function body scope, reindent. For BUILTIN_IN_STRNCPY*, don't
3718 perform head trim unless we can prove there are no '\0' chars
3719 from the source among the first head_trim chars.
3721 2020-01-14 David Malcolm <dmalcolm@redhat.com>
3723 * Makefile.in (ANALYZER_OBJS): Add analyzer/function-set.o.
3725 2020-01-15 Jakub Jelinek <jakub@redhat.com>
3728 * config/i386/sse.md
3729 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1,
3730 *<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1,
3731 *<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1,
3732 *<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1): Use
3733 just a single alternative instead of two, make operands 1 and 2
3736 2020-01-14 Jan Hubicka <hubicka@ucw.cz>
3739 * ipa-devirt.c (odr_types_equivalent_p): Compare TREE_ADDRESSABLE and
3742 2020-01-14 David Malcolm <dmalcolm@redhat.com>
3744 * Makefile.in (lang_opt_files): Add analyzer.opt.
3745 (ANALYZER_OBJS): New.
3746 (OBJS): Add digraph.o, graphviz.o, ordered-hash-map-tests.o,
3747 tristate.o and ANALYZER_OBJS.
3748 (TEXI_GCCINT_FILES): Add analyzer.texi.
3749 * common.opt (-fanalyzer): New driver option.
3750 * config.in: Regenerate.
3751 * configure: Regenerate.
3752 * configure.ac (--disable-analyzer, ENABLE_ANALYZER): New option.
3753 (gccdepdir): Also create depdir for "analyzer" subdir.
3754 * digraph.cc: New file.
3755 * digraph.h: New file.
3756 * doc/analyzer.texi: New file.
3757 * doc/gccint.texi ("Static Analyzer") New menu item.
3758 (analyzer.texi): Include it.
3759 * doc/invoke.texi ("Static Analyzer Options"): New list and new section.
3760 ("Warning Options"): Add static analysis warnings to the list.
3761 (-Wno-analyzer-double-fclose): New option.
3762 (-Wno-analyzer-double-free): New option.
3763 (-Wno-analyzer-exposure-through-output-file): New option.
3764 (-Wno-analyzer-file-leak): New option.
3765 (-Wno-analyzer-free-of-non-heap): New option.
3766 (-Wno-analyzer-malloc-leak): New option.
3767 (-Wno-analyzer-possible-null-argument): New option.
3768 (-Wno-analyzer-possible-null-dereference): New option.
3769 (-Wno-analyzer-null-argument): New option.
3770 (-Wno-analyzer-null-dereference): New option.
3771 (-Wno-analyzer-stale-setjmp-buffer): New option.
3772 (-Wno-analyzer-tainted-array-index): New option.
3773 (-Wno-analyzer-use-after-free): New option.
3774 (-Wno-analyzer-use-of-pointer-in-stale-stack-frame): New option.
3775 (-Wno-analyzer-use-of-uninitialized-value): New option.
3776 (-Wanalyzer-too-complex): New option.
3777 (-fanalyzer-call-summaries): New warning.
3778 (-fanalyzer-checker=): New warning.
3779 (-fanalyzer-fine-grained): New warning.
3780 (-fno-analyzer-state-merge): New warning.
3781 (-fno-analyzer-state-purge): New warning.
3782 (-fanalyzer-transitivity): New warning.
3783 (-fanalyzer-verbose-edges): New warning.
3784 (-fanalyzer-verbose-state-changes): New warning.
3785 (-fanalyzer-verbosity=): New warning.
3786 (-fdump-analyzer): New warning.
3787 (-fdump-analyzer-callgraph): New warning.
3788 (-fdump-analyzer-exploded-graph): New warning.
3789 (-fdump-analyzer-exploded-nodes): New warning.
3790 (-fdump-analyzer-exploded-nodes-2): New warning.
3791 (-fdump-analyzer-exploded-nodes-3): New warning.
3792 (-fdump-analyzer-supergraph): New warning.
3793 * doc/sourcebuild.texi (dg-require-dot): New.
3794 (dg-check-dot): New.
3795 * gdbinit.in (break-on-saved-diagnostic): New command.
3796 * graphviz.cc: New file.
3797 * graphviz.h: New file.
3798 * ordered-hash-map-tests.cc: New file.
3799 * ordered-hash-map.h: New file.
3800 * passes.def (pass_analyzer): Add before
3801 pass_ipa_whole_program_visibility.
3802 * selftest-run-tests.c (selftest::run_tests): Call
3803 selftest::ordered_hash_map_tests_cc_tests.
3804 * selftest.h (selftest::ordered_hash_map_tests_cc_tests): New
3806 * shortest-paths.h: New file.
3807 * timevar.def (TV_ANALYZER): New timevar.
3808 (TV_ANALYZER_SUPERGRAPH): Likewise.
3809 (TV_ANALYZER_STATE_PURGE): Likewise.
3810 (TV_ANALYZER_PLAN): Likewise.
3811 (TV_ANALYZER_SCC): Likewise.
3812 (TV_ANALYZER_WORKLIST): Likewise.
3813 (TV_ANALYZER_DUMP): Likewise.
3814 (TV_ANALYZER_DIAGNOSTICS): Likewise.
3815 (TV_ANALYZER_SHORTEST_PATHS): Likewise.
3816 * tree-pass.h (make_pass_analyzer): New decl.
3817 * tristate.cc: New file.
3818 * tristate.h: New file.
3820 2020-01-14 Uroš Bizjak <ubizjak@gmail.com>
3823 * config/i386/i386.md (*movsf_internal): Require SSE2 ISA for
3824 alternatives 9 and 10.
3826 2020-01-14 David Malcolm <dmalcolm@redhat.com>
3828 * attribs.c (excl_hash_traits::empty_zero_p): New static constant.
3829 * gcov.c (function_start_pair_hash::empty_zero_p): Likewise.
3830 * graphite.c (struct sese_scev_hash::empty_zero_p): Likewise.
3831 * hash-map-tests.c (selftest::test_nonzero_empty_key): New selftest.
3832 (selftest::hash_map_tests_c_tests): Call it.
3833 * hash-map-traits.h (simple_hashmap_traits::empty_zero_p):
3834 New static constant, using the value of = H::empty_zero_p.
3835 (unbounded_hashmap_traits::empty_zero_p): Likewise, using the value
3836 from default_hash_traits <Value>.
3837 * hash-map.h (hash_map::empty_zero_p): Likewise, using the value
3839 * hash-set-tests.c (value_hash_traits::empty_zero_p): Likewise.
3840 * hash-table.h (hash_table::alloc_entries): Guard the loop of
3841 calls to mark_empty with !Descriptor::empty_zero_p.
3842 (hash_table::empty_slow): Conditionalize the memset call with a
3843 check that Descriptor::empty_zero_p; otherwise, loop through the
3844 entries calling mark_empty on them.
3845 * hash-traits.h (int_hash::empty_zero_p): New static constant.
3846 (pointer_hash::empty_zero_p): Likewise.
3847 (pair_hash::empty_zero_p): Likewise.
3848 * ipa-devirt.c (default_hash_traits <type_pair>::empty_zero_p):
3850 * ipa-prop.c (ipa_bit_ggc_hash_traits::empty_zero_p): Likewise.
3851 (ipa_vr_ggc_hash_traits::empty_zero_p): Likewise.
3852 * profile.c (location_triplet_hash::empty_zero_p): Likewise.
3853 * sanopt.c (sanopt_tree_triplet_hash::empty_zero_p): Likewise.
3854 (sanopt_tree_couple_hash::empty_zero_p): Likewise.
3855 * tree-hasher.h (int_tree_hasher::empty_zero_p): Likewise.
3856 * tree-ssa-sccvn.c (vn_ssa_aux_hasher::empty_zero_p): Likewise.
3857 * tree-vect-slp.c (bst_traits::empty_zero_p): Likewise.
3859 (default_hash_traits<scalar_cond_masked_key>::empty_zero_p):
3862 2020-01-14 Kewen Lin <linkw@gcc.gnu.org>
3864 * cfgloopanal.c (average_num_loop_insns): Free bbs when early return,
3865 fix typo on return value.
3867 2020-01-14 Xiong Hu Luo <luoxhu@linux.ibm.com>
3870 * cgraph.c (symbol_table::create_edge): Init speculative_id and
3872 (cgraph_edge::make_speculative): Add param for setting speculative_id
3874 (cgraph_edge::speculative_call_info): Update comments and find reference
3875 by speculative_id for multiple indirect targets.
3876 (cgraph_edge::resolve_speculation): Decrease the speculations
3877 for indirect edge, drop it's speculative if not direct target
3878 left. Update comments.
3879 (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
3880 (cgraph_node::dump): Print num_speculative_call_targets.
3881 (cgraph_node::verify_node): Don't report error if speculative
3882 edge not include statement.
3883 (cgraph_edge::num_speculative_call_targets_p): New function.
3884 * cgraph.h (int common_target_id): Remove.
3885 (int common_target_probability): Remove.
3886 (num_speculative_call_targets): New variable.
3887 (make_speculative): Add param for setting speculative_id.
3888 (cgraph_edge::num_speculative_call_targets_p): New declare.
3889 (target_prob): New variable.
3890 (speculative_id): New variable.
3891 * ipa-fnsummary.c (analyze_function_body): Create and duplicate
3892 call summaries for multiple speculative call targets.
3893 * cgraphclones.c (cgraph_node::create_clone): Clone speculative_id.
3894 * ipa-profile.c (struct speculative_call_target): New struct.
3895 (class speculative_call_summary): New class.
3896 (class speculative_call_summaries): New class.
3897 (call_sums): New variable.
3898 (ipa_profile_generate_summary): Generate indirect multiple targets summaries.
3899 (ipa_profile_write_edge_summary): New function.
3900 (ipa_profile_write_summary): Stream out indirect multiple targets summaries.
3901 (ipa_profile_dump_all_summaries): New function.
3902 (ipa_profile_read_edge_summary): New function.
3903 (ipa_profile_read_summary_section): New function.
3904 (ipa_profile_read_summary): Stream in indirect multiple targets summaries.
3905 (ipa_profile): Generate num_speculative_call_targets from
3907 * ipa-ref.h (speculative_id): New variable.
3908 * ipa-utils.c (ipa_merge_profiles): Update with target_prob.
3909 * lto-cgraph.c (lto_output_edge): Remove indirect common_target_id and
3910 common_target_probability. Stream out speculative_id and
3911 num_speculative_call_targets.
3912 (input_edge): Likewise.
3913 * predict.c (dump_prediction): Remove edges count assert to be
3915 * symtab.c (symtab_node::create_reference): Init speculative_id.
3916 (symtab_node::clone_references): Clone speculative_id.
3917 (symtab_node::clone_referring): Clone speculative_id.
3918 (symtab_node::clone_reference): Clone speculative_id.
3919 (symtab_node::clear_stmts_in_references): Clear speculative_id.
3920 * tree-inline.c (copy_bb): Duplicate all the speculative edges
3921 if indirect call contains multiple speculative targets.
3922 * value-prof.h (check_ic_target): Remove.
3923 * value-prof.c (gimple_value_profile_transformations):
3924 Use void function gimple_ic_transform.
3925 * value-prof.c (gimple_ic_transform): Handle topn case.
3926 Fix comment typos. Change it to a void function.
3928 2020-01-13 Andrew Pinski <apinski@marvell.com>
3930 * config/aarch64/aarch64-cores.def (octeontx2): New define.
3931 (octeontx2t98): New define.
3932 (octeontx2t96): New define.
3933 (octeontx2t93): New define.
3934 (octeontx2f95): New define.
3935 (octeontx2f95n): New define.
3936 (octeontx2f95mm): New define.
3937 * config/aarch64/aarch64-tune.md: Regenerate.
3938 * doc/invoke.texi (-mcpu=): Document the new cpu types.
3940 2020-01-13 Jason Merrill <jason@redhat.com>
3942 PR c++/33799 - destroy return value if local cleanup throws.
3943 * gimplify.c (gimplify_return_expr): Handle COMPOUND_EXPR.
3945 2020-01-13 Martin Liska <mliska@suse.cz>
3947 * ipa-cp.c (get_max_overall_size): Use newly
3948 renamed param param_ipa_cp_unit_growth.
3949 * params.opt: Remove legacy param name.
3951 2020-01-13 Martin Sebor <msebor@redhat.com>
3953 PR tree-optimization/93213
3954 * tree-ssa-strlen.c (handle_store): Only allow single-byte nul-over-nul
3955 stores to be eliminated.
3957 2020-01-13 Martin Liska <mliska@suse.cz>
3959 * opts.c (print_help): Do not print CL_PARAM
3960 and CL_WARNING for CL_OPTIMIZATION.
3962 2020-01-13 Jonathan Wakely <jwakely@redhat.com>
3965 * doc/invoke.texi (Warning Options): Add caveat about some warnings
3966 depending on optimization settings.
3968 2020-01-13 Jakub Jelinek <jakub@redhat.com>
3970 PR tree-optimization/90838
3971 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
3972 SCALAR_INT_TYPE_MODE directly in CTZ_DEFINED_VALUE_AT_ZERO macro
3973 argument rather than to initialize temporary for targets that
3974 don't use the mode argument at all. Initialize ctzval to avoid
3977 2020-01-10 Thomas Schwinge <thomas@codesourcery.com>
3979 * tree.h (OMP_CLAUSE_USE_DEVICE_PTR_IF_PRESENT): New definition.
3980 * tree-core.h: Document it.
3981 * gimplify.c (gimplify_omp_workshare): Set it.
3982 * omp-low.c (lower_omp_target): Use it.
3983 * tree-pretty-print.c (dump_omp_clause): Print it.
3985 * omp-low.c (lower_omp_target) <OMP_CLAUSE_USE_DEVICE_PTR etc.>:
3986 Assert that for OpenACC we always have 'GOMP_MAP_USE_DEVICE_PTR'.
3988 2020-01-10 David Malcolm <dmalcolm@redhat.com>
3990 * Makefile.in (OBJS): Add tree-diagnostic-path.o.
3991 * common.opt (fdiagnostics-path-format=): New option.
3992 (diagnostic_path_format): New enum.
3993 (fdiagnostics-show-path-depths): New option.
3994 * coretypes.h (diagnostic_event_id_t): New forward decl.
3995 * diagnostic-color.c (color_dict): Add "path".
3996 * diagnostic-event-id.h: New file.
3997 * diagnostic-format-json.cc (json_from_expanded_location): Make
3999 (json_end_diagnostic): Call context->make_json_for_path if it
4000 exists and the diagnostic has a path.
4001 (diagnostic_output_format_init): Clear context->print_path.
4002 * diagnostic-path.h: New file.
4003 * diagnostic-show-locus.c (colorizer::set_range): Special-case
4004 when printing a run of events in a diagnostic_path so that they
4005 all get the same color.
4006 (layout::m_diagnostic_path_p): New field.
4007 (layout::layout): Initialize it.
4008 (layout::print_any_labels): Don't colorize the label text for an
4009 event in a diagnostic_path.
4010 (gcc_rich_location::add_location_if_nearby): Add
4011 "restrict_to_current_line_spans" and "label" params. Pass the
4012 former to layout.maybe_add_location_range; pass the latter
4013 when calling add_range.
4014 * diagnostic.c: Include "diagnostic-path.h".
4015 (diagnostic_initialize): Initialize context->path_format and
4016 context->show_path_depths.
4017 (diagnostic_show_any_path): New function.
4018 (diagnostic_path::interprocedural_p): New function.
4019 (diagnostic_report_diagnostic): Call diagnostic_show_any_path.
4020 (simple_diagnostic_path::num_events): New function.
4021 (simple_diagnostic_path::get_event): New function.
4022 (simple_diagnostic_path::add_event): New function.
4023 (simple_diagnostic_event::simple_diagnostic_event): New ctor.
4024 (simple_diagnostic_event::~simple_diagnostic_event): New dtor.
4025 (debug): New overload taking a diagnostic_path *.
4026 * diagnostic.def (DK_DIAGNOSTIC_PATH): New.
4027 * diagnostic.h (enum diagnostic_path_format): New enum.
4028 (json::value): New forward decl.
4029 (diagnostic_context::path_format): New field.
4030 (diagnostic_context::show_path_depths): New field.
4031 (diagnostic_context::print_path): New callback field.
4032 (diagnostic_context::make_json_for_path): New callback field.
4033 (diagnostic_show_any_path): New decl.
4034 (json_from_expanded_location): New decl.
4035 * doc/invoke.texi (-fdiagnostics-path-format=): New option.
4036 (-fdiagnostics-show-path-depths): New option.
4037 (-fdiagnostics-color): Add "path" to description of default
4038 GCC_COLORS; describe it.
4039 (-fdiagnostics-format=json): Document how diagnostic paths are
4040 represented in the JSON output format.
4041 * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
4042 Add optional params "restrict_to_current_line_spans" and "label".
4043 * opts.c (common_handle_option): Handle
4044 OPT_fdiagnostics_path_format_ and
4045 OPT_fdiagnostics_show_path_depths.
4046 * pretty-print.c: Include "diagnostic-event-id.h".
4047 (pp_format): Implement "%@" format code for printing
4048 diagnostic_event_id_t *.
4049 (selftest::test_pp_format): Add tests for "%@".
4050 * selftest-run-tests.c (selftest::run_tests): Call
4051 selftest::tree_diagnostic_path_cc_tests.
4052 * selftest.h (selftest::tree_diagnostic_path_cc_tests): New decl.
4053 * toplev.c (general_init): Initialize global_dc->path_format and
4054 global_dc->show_path_depths.
4055 * tree-diagnostic-path.cc: New file.
4056 * tree-diagnostic.c (maybe_unwind_expanded_macro_loc): Make
4057 non-static. Drop "diagnostic" param in favor of storing the
4058 original value of "where" and re-using it.
4059 (virt_loc_aware_diagnostic_finalizer): Update for dropped param of
4060 maybe_unwind_expanded_macro_loc.
4061 (tree_diagnostics_defaults): Initialize context->print_path and
4062 context->make_json_for_path.
4063 * tree-diagnostic.h (default_tree_diagnostic_path_printer): New
4065 (default_tree_make_json_for_path): New decl.
4066 (maybe_unwind_expanded_macro_loc): New decl.
4068 2020-01-10 Jakub Jelinek <jakub@redhat.com>
4070 PR tree-optimization/93210
4071 * fold-const.h (native_encode_initializer,
4072 can_native_interpret_type_p): Declare.
4073 * fold-const.c (native_encode_string): Fix up handling with off != -1,
4075 (native_encode_initializer): New function, moved from dwarf2out.c.
4076 Adjust to native_encode_expr compatible arguments, including dry-run
4077 and partial extraction modes. Don't handle STRING_CST.
4078 (can_native_interpret_type_p): No longer static.
4079 * gimple-fold.c (fold_ctor_reference): For native_encode_expr, verify
4080 offset / BITS_PER_UNIT fits into int and don't call it if
4081 can_native_interpret_type_p fails. If suboff is NULL and for
4082 CONSTRUCTOR fold_{,non}array_ctor_reference returns NULL, retry with
4083 native_encode_initializer.
4084 (fold_const_aggregate_ref_1): Formatting fix.
4085 * dwarf2out.c (native_encode_initializer): Moved to fold-const.c.
4086 (tree_add_const_value_attribute): Adjust caller.
4088 PR tree-optimization/90838
4089 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
4090 SCALAR_INT_TYPE_MODE instead of TYPE_MODE as operand of
4091 CTZ_DEFINED_VALUE_AT_ZERO.
4093 2020-01-10 Vladimir Makarov <vmakarov@redhat.com>
4096 * lra-constraints.c (match_reload): Permit input operands have the
4097 same mode as output while other input operands have a different
4100 2020-01-10 Wilco Dijkstra <wdijkstr@arm.com>
4102 PR tree-optimization/90838
4103 * tree-ssa-forwprop.c (check_ctz_array): Add new function.
4104 (check_ctz_string): Likewise.
4105 (optimize_count_trailing_zeroes): Likewise.
4106 (simplify_count_trailing_zeroes): Likewise.
4107 (pass_forwprop::execute): Try ctz simplification.
4108 * match.pd: Add matching for ctz idioms.
4110 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
4112 * config/aarch64/aarch64.c (aarch64_invalid_conversion): New function
4114 (aarch64_invalid_unary_op): New function for target hook.
4115 (aarch64_invalid_binary_op): New function for target hook.
4117 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
4119 * config.gcc: Add arm_bf16.h.
4120 * config/aarch64/aarch64-builtins.c
4121 (aarch64_simd_builtin_std_type): Add BFmode.
4122 (aarch64_init_simd_builtin_types): Define element types for vector
4124 (aarch64_init_bf16_types): New function.
4125 (aarch64_general_init_builtins): Add arm_init_bf16_types function call.
4126 * config/aarch64/aarch64-modes.def: Add BFmode and V4BF, V8BF vector
4128 * config/aarch64/aarch64-simd-builtin-types.def: Add BF SIMD types.
4129 * config/aarch64/aarch64-simd.md: Add BF vector types to NEON move
4131 * config/aarch64/aarch64.h (AARCH64_VALID_SIMD_DREG_MODE): Add V4BF.
4132 (AARCH64_VALID_SIMD_QREG_MODE): Add V8BF.
4133 * config/aarch64/aarch64.c
4134 (aarch64_classify_vector_mode): Add support for BF types.
4135 (aarch64_gimplify_va_arg_expr): Add support for BF types.
4136 (aarch64_vq_mode): Add support for BF types.
4137 (aarch64_simd_container_mode): Add support for BF types.
4138 (aarch64_mangle_type): Add support for BF scalar type.
4139 * config/aarch64/aarch64.md: Add BFmode to movhf pattern.
4140 * config/aarch64/arm_bf16.h: New file.
4141 * config/aarch64/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
4142 * config/aarch64/iterators.md: Add BF types to mode attributes.
4143 (HFBF, GPF_TF_F16_MOV, VDMOV, VQMOV, VQMOV_NO2Em VALL_F16MOV): New.
4145 2020-01-10 Jason Merrill <jason@redhat.com>
4147 PR c++/93173 - incorrect tree sharing.
4148 * gimplify.c (copy_if_shared): No longer static.
4149 * gimplify.h: Declare it.
4151 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
4153 * doc/invoke.texi (-msve-vector-bits=): Document that
4154 -msve-vector-bits=128 now generates VL-specific code for
4155 little-endian targets.
4156 * config/aarch64/aarch64-sve-builtins.cc (register_builtin_types): Use
4157 build_vector_type_for_mode to construct the data vector types.
4158 * config/aarch64/aarch64.c (aarch64_convert_sve_vector_bits): Generate
4159 VL-specific code for -msve-vector-bits=128 on little-endian targets.
4160 (aarch64_simd_container_mode): Always prefer Advanced SIMD modes
4161 for 128-bit vectors.
4163 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
4165 * config/aarch64/aarch64.c (aarch64_evpc_sel): Fix gen_vcond_mask
4168 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
4170 * config/aarch64/aarch64-builtins.c
4171 (aarch64_builtin_vectorized_function): Check for specific vector modes,
4172 rather than checking the number of elements and the element mode.
4174 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
4176 * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
4177 get_related_vectype_for_scalar_type rather than build_vector_type
4178 to create the index type for a conditional reduction.
4180 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
4182 * tree-vect-loop.c (update_epilogue_loop_vinfo): Update DR_REF
4183 for any type of gather or scatter, including strided accesses.
4185 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
4187 * tree-vectorizer.h (get_dr_vinfo_offset): Add missing function
4190 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
4192 * tree-vect-data-refs.c (vect_create_addr_base_for_vector_ref): Use
4194 * tree-vect-loop.c (update_epilogue_loop_vinfo): Remove orig_drs_init
4195 parameter and its use to reset DR_OFFSET's.
4196 (vect_transform_loop): Remove orig_drs_init argument.
4197 * tree-vect-loop-manip.c (vect_update_init_of_dr): Update the offset
4198 member of dr_vec_info rather than the offset of the associated
4199 data_reference's innermost_loop_behavior.
4200 (vect_update_init_of_dr): Pass dr_vec_info instead of data_reference.
4201 (vect_do_peeling): Remove orig_drs_init parameter and its construction.
4202 * tree-vect-stmts.c (check_scan_store): Replace use of DR_OFFSET with
4203 get_dr_vinfo_offset.
4204 (vectorizable_store): Likewise.
4205 (vectorizable_load): Likewise.
4207 2020-01-10 Richard Biener <rguenther@suse.de>
4209 * gimple-ssa-store-merging
4210 (pass_store_merging::terminate_all_aliasing_chains): Cache alias info.
4212 2020-01-10 Martin Liska <mliska@suse.cz>
4215 * ipa-inline-analysis.c (offline_size): Make proper parenthesis
4216 encapsulation that was there before r280040.
4218 2020-01-10 Richard Biener <rguenther@suse.de>
4221 * tree-eh.c (sink_clobbers): Move clobbers to out-of-IL
4222 sequences to avoid walking them again for secondary opportunities.
4223 (pass_lower_eh_dispatch::execute): Instead actually insert
4226 2020-01-10 Richard Biener <rguenther@suse.de>
4229 * tree-eh.c (redirect_eh_edge_1): Avoid some work if possible.
4230 (cleanup_all_empty_eh): Walk landing pads in reverse order to
4231 avoid quadraticness.
4233 2020-01-10 Martin Jambor <mjambor@suse.cz>
4235 * params.opt (param_ipa_sra_max_replacements): Mark as Optimization.
4236 * ipa-sra.c (pull_accesses_from_callee): New parameter caller, use it
4237 to get param_ipa_sra_max_replacements.
4238 (param_splitting_across_edge): Pass the caller to
4239 pull_accesses_from_callee.
4241 2020-01-10 Martin Jambor <mjambor@suse.cz>
4243 * params.opt (param_ipcp_unit_growth): Mark as Optimization.
4244 * ipa-cp.c (max_new_size): Removed.
4245 (orig_overall_size): New variable.
4246 (get_max_overall_size): New function.
4247 (estimate_local_effects): Use it. Adjust dump.
4248 (decide_about_value): Likewise.
4249 (ipcp_propagate_stage): Do not calculate max_new_size, just store
4250 orig_overall_size. Adjust dump.
4251 (ipa_cp_c_finalize): Clear orig_overall_size instead of max_new_size.
4253 2020-01-10 Martin Jambor <mjambor@suse.cz>
4255 * params.opt (param_ipa_max_agg_items): Mark as Optimization
4256 * ipa-cp.c (merge_agg_lats_step): New parameter max_agg_items, use
4257 instead of param_ipa_max_agg_items.
4258 (merge_aggregate_lattices): Extract param_ipa_max_agg_items from
4259 optimization info for the callee.
4261 2020-01-09 Kwok Cheung Yeung <kcy@codesourcery.com>
4263 * lto-streamer-in.c (input_function): Remove streamed-in inline debug
4264 markers if debug_inline_points is false.
4266 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4268 * config.gcc (aarch64*-*-*): Add aarch64-sve-builtins-sve2.o to
4270 * config/aarch64/t-aarch64 (aarch64-sve-builtins.o): Depend on
4271 aarch64-sve-builtins-base.def, aarch64-sve-builtins-sve2.def and
4272 aarch64-sve-builtins-sve2.h.
4273 (aarch64-sve-builtins-sve2.o): New rule.
4274 * config/aarch64/aarch64.h (AARCH64_ISA_SVE2_AES): New macro.
4275 (AARCH64_ISA_SVE2_BITPERM, AARCH64_ISA_SVE2_SHA3): Likewise.
4276 (AARCH64_ISA_SVE2_SM4, TARGET_SVE2_AES, TARGET_SVE2_BITPERM): Likewise.
4277 (TARGET_SVE2_SHA, TARGET_SVE2_SM4): Likewise.
4278 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
4279 TARGET_SVE2_AES, TARGET_SVE2_BITPERM, TARGET_SVE2_SHA3 and
4281 * config/aarch64/aarch64-sve.md: Update comments with SVE2
4282 instructions that are handled here.
4283 (@cond_asrd<mode>): Generalize to...
4284 (@cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>): ...this.
4285 (*cond_asrd<mode>_2): Generalize to...
4286 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_2): ...this.
4287 (*cond_asrd<mode>_z): Generalize to...
4288 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_z): ...this.
4289 * config/aarch64/aarch64.md (UNSPEC_LDNT1_GATHER): New unspec.
4290 (UNSPEC_STNT1_SCATTER, UNSPEC_WHILEGE, UNSPEC_WHILEGT): Likewise.
4291 (UNSPEC_WHILEHI, UNSPEC_WHILEHS): Likewise.
4292 * config/aarch64/aarch64-sve2.md (@aarch64_gather_ldnt<mode>): New
4294 (@aarch64_gather_ldnt_<ANY_EXTEND:optab><SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
4295 (@aarch64_scatter_stnt<mode>): Likewise.
4296 (@aarch64_scatter_stnt_<SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
4297 (@aarch64_mul_lane_<mode>): Likewise.
4298 (@aarch64_sve_suqadd<mode>_const): Likewise.
4299 (*<sur>h<addsub><mode>): Generalize to...
4300 (@aarch64_pred_<SVE2_COND_INT_BINARY_REV:sve_int_op><mode>): ...this
4302 (@cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>): New expander.
4303 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_2): New pattern.
4304 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_3): Likewise.
4305 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_any): Likewise.
4306 (*cond_<SVE2_COND_INT_BINARY_NOREV:sve_int_op><mode>_z): Likewise.
4307 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op><mode>):: Likewise.
4308 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op>_lane_<mode>): Likewise.
4309 (@aarch64_pred_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): Likewise.
4310 (@cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): New expander.
4311 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_2): New pattern.
4312 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_3): Likewise.
4313 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_any): Likewise.
4314 (@aarch64_sve_<SVE2_INT_TERNARY:sve_int_op><mode>): Likewise.
4315 (@aarch64_sve_<SVE2_INT_TERNARY_LANE:sve_int_op>_lane_<mode>)
4316 (@aarch64_sve_add_mul_lane_<mode>): Likewise.
4317 (@aarch64_sve_sub_mul_lane_<mode>): Likewise.
4318 (@aarch64_sve2_xar<mode>): Likewise.
4319 (@aarch64_sve2_bcax<mode>): Likewise.
4320 (*aarch64_sve2_eor3<mode>): Rename to...
4321 (@aarch64_sve2_eor3<mode>): ...this.
4322 (@aarch64_sve2_bsl<mode>): New expander.
4323 (@aarch64_sve2_nbsl<mode>): Likewise.
4324 (@aarch64_sve2_bsl1n<mode>): Likewise.
4325 (@aarch64_sve2_bsl2n<mode>): Likewise.
4326 (@aarch64_sve_add_<SHIFTRT:sve_int_op><mode>): Likewise.
4327 (*aarch64_sve2_sra<mode>): Add MOVPRFX support.
4328 (@aarch64_sve_add_<VRSHR_N:sve_int_op><mode>): New pattern.
4329 (@aarch64_sve_<SVE2_INT_SHIFT_INSERT:sve_int_op><mode>): Likewise.
4330 (@aarch64_sve2_<USMAX:su>aba<mode>): New expander.
4331 (*aarch64_sve2_<USMAX:su>aba<mode>): New pattern.
4332 (@aarch64_sve_<SVE2_INT_BINARY_WIDE:sve_int_op><mode>): Likewise.
4333 (<su>mull<bt><Vwide>): Generalize to...
4334 (@aarch64_sve_<SVE2_INT_BINARY_LONG:sve_int_op><mode>): ...this new
4336 (@aarch64_sve_<SVE2_INT_BINARY_LONG_lANE:sve_int_op>_lane_<mode>)
4337 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_LONG:sve_int_op><mode>)
4338 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG:sve_int_op><mode>)
4339 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
4340 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG:sve_int_op><mode>)
4341 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
4342 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG:sve_int_op><mode>)
4343 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
4344 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG:sve_int_op><mode>)
4345 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
4346 (@aarch64_sve_<SVE2_FP_TERNARY_LONG:sve_fp_op><mode>): New patterns.
4347 (@aarch64_<SVE2_FP_TERNARY_LONG_LANE:sve_fp_op>_lane_<mode>)
4348 (@aarch64_sve_<SVE2_INT_UNARY_NARROWB:sve_int_op><mode>): Likewise.
4349 (@aarch64_sve_<SVE2_INT_UNARY_NARROWT:sve_int_op><mode>): Likewise.
4350 (@aarch64_sve_<SVE2_INT_BINARY_NARROWB:sve_int_op><mode>): Likewise.
4351 (@aarch64_sve_<SVE2_INT_BINARY_NARROWT:sve_int_op><mode>): Likewise.
4352 (<SHRNB:r>shrnb<mode>): Generalize to...
4353 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWB:sve_int_op><mode>): ...this
4355 (<SHRNT:r>shrnt<mode>): Generalize to...
4356 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWT:sve_int_op><mode>): ...this
4358 (@aarch64_pred_<SVE2_INT_BINARY_PAIR:sve_int_op><mode>): New pattern.
4359 (@aarch64_pred_<SVE2_FP_BINARY_PAIR:sve_fp_op><mode>): Likewise.
4360 (@cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>): New expander.
4361 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_2): New pattern.
4362 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_z): Likewise.
4363 (@aarch64_sve_<SVE2_INT_CADD:optab><mode>): Likewise.
4364 (@aarch64_sve_<SVE2_INT_CMLA:optab><mode>): Likewise.
4365 (@aarch64_<SVE2_INT_CMLA:optab>_lane_<mode>): Likewise.
4366 (@aarch64_sve_<SVE2_INT_CDOT:optab><mode>): Likewise.
4367 (@aarch64_<SVE2_INT_CDOT:optab>_lane_<mode>): Likewise.
4368 (@aarch64_pred_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): Likewise.
4369 (@cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New expander.
4370 (*cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New pattern.
4371 (@aarch64_sve2_cvtnt<mode>): Likewise.
4372 (@aarch64_pred_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): Likewise.
4373 (@cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): New expander.
4374 (*cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>_any): New pattern.
4375 (@aarch64_sve2_cvtxnt<mode>): Likewise.
4376 (@aarch64_pred_<SVE2_U32_UNARY:sve_int_op><mode>): Likewise.
4377 (@cond_<SVE2_U32_UNARY:sve_int_op><mode>): New expander.
4378 (*cond_<SVE2_U32_UNARY:sve_int_op><mode>): New pattern.
4379 (@aarch64_pred_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): Likewise.
4380 (@cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New expander.
4381 (*cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New pattern.
4382 (@aarch64_sve2_pmul<mode>): Likewise.
4383 (@aarch64_sve_<SVE2_PMULL:optab><mode>): Likewise.
4384 (@aarch64_sve_<SVE2_PMULL_PAIR:optab><mode>): Likewise.
4385 (@aarch64_sve2_tbl2<mode>): Likewise.
4386 (@aarch64_sve2_tbx<mode>): Likewise.
4387 (@aarch64_sve_<SVE2_INT_BITPERM:sve_int_op><mode>): Likewise.
4388 (@aarch64_sve2_histcnt<mode>): Likewise.
4389 (@aarch64_sve2_histseg<mode>): Likewise.
4390 (@aarch64_pred_<SVE2_MATCH:sve_int_op><mode>): Likewise.
4391 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_cc): Likewise.
4392 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_ptest): Likewise.
4393 (aarch64_sve2_aes<CRYPTO_AES:aes_op>): Likewise.
4394 (aarch64_sve2_aes<CRYPTO_AESMC:aesmc_op>): Likewise.
4395 (*aarch64_sve2_aese_fused, *aarch64_sve2_aesd_fused): Likewise.
4396 (aarch64_sve2_rax1, aarch64_sve2_sm4e, aarch64_sve2_sm4ekey): Likewise.
4397 (<su>mulh<r>s<mode>3): Update after above pattern name changes.
4398 * config/aarch64/iterators.md (VNx16QI_ONLY, VNx4SF_ONLY)
4399 (SVE_STRUCT2, SVE_FULL_BHI, SVE_FULL_HSI, SVE_FULL_HDI)
4400 (SVE2_PMULL_PAIR_I): New mode iterators.
4401 (UNSPEC_ADCLB, UNSPEC_ADCLT, UNSPEC_ADDHNB, UNSPEC_ADDHNT, UNSPEC_BDEP)
4402 (UNSPEC_BEXT, UNSPEC_BGRP, UNSPEC_CADD90, UNSPEC_CADD270, UNSPEC_CDOT)
4403 (UNSPEC_CDOT90, UNSPEC_CDOT180, UNSPEC_CDOT270, UNSPEC_CMLA)
4404 (UNSPEC_CMLA90, UNSPEC_CMLA180, UNSPEC_CMLA270, UNSPEC_COND_FCVTLT)
4405 (UNSPEC_COND_FCVTNT, UNSPEC_COND_FCVTX, UNSPEC_COND_FCVTXNT)
4406 (UNSPEC_COND_FLOGB, UNSPEC_EORBT, UNSPEC_EORTB, UNSPEC_FADDP)
4407 (UNSPEC_FMAXP, UNSPEC_FMAXNMP, UNSPEC_FMLALB, UNSPEC_FMLALT)
4408 (UNSPEC_FMLSLB, UNSPEC_FMLSLT, UNSPEC_FMINP, UNSPEC_FMINNMP)
4409 (UNSPEC_HISTCNT, UNSPEC_HISTSEG, UNSPEC_MATCH, UNSPEC_NMATCH)
4410 (UNSPEC_PMULLB, UNSPEC_PMULLB_PAIR, UNSPEC_PMULLT, UNSPEC_PMULLT_PAIR)
4411 (UNSPEC_RADDHNB, UNSPEC_RADDHNT, UNSPEC_RSUBHNB, UNSPEC_RSUBHNT)
4412 (UNSPEC_SLI, UNSPEC_SRI, UNSPEC_SABDLB, UNSPEC_SABDLT, UNSPEC_SADDLB)
4413 (UNSPEC_SADDLBT, UNSPEC_SADDLT, UNSPEC_SADDWB, UNSPEC_SADDWT)
4414 (UNSPEC_SBCLB, UNSPEC_SBCLT, UNSPEC_SMAXP, UNSPEC_SMINP)
4415 (UNSPEC_SQCADD90, UNSPEC_SQCADD270, UNSPEC_SQDMULLB, UNSPEC_SQDMULLBT)
4416 (UNSPEC_SQDMULLT, UNSPEC_SQRDCMLAH, UNSPEC_SQRDCMLAH90)
4417 (UNSPEC_SQRDCMLAH180, UNSPEC_SQRDCMLAH270, UNSPEC_SQRSHRNB)
4418 (UNSPEC_SQRSHRNT, UNSPEC_SQRSHRUNB, UNSPEC_SQRSHRUNT, UNSPEC_SQSHRNB)
4419 (UNSPEC_SQSHRNT, UNSPEC_SQSHRUNB, UNSPEC_SQSHRUNT, UNSPEC_SQXTNB)
4420 (UNSPEC_SQXTNT, UNSPEC_SQXTUNB, UNSPEC_SQXTUNT, UNSPEC_SSHLLB)
4421 (UNSPEC_SSHLLT, UNSPEC_SSUBLB, UNSPEC_SSUBLBT, UNSPEC_SSUBLT)
4422 (UNSPEC_SSUBLTB, UNSPEC_SSUBWB, UNSPEC_SSUBWT, UNSPEC_SUBHNB)
4423 (UNSPEC_SUBHNT, UNSPEC_TBL2, UNSPEC_UABDLB, UNSPEC_UABDLT)
4424 (UNSPEC_UADDLB, UNSPEC_UADDLT, UNSPEC_UADDWB, UNSPEC_UADDWT)
4425 (UNSPEC_UMAXP, UNSPEC_UMINP, UNSPEC_UQRSHRNB, UNSPEC_UQRSHRNT)
4426 (UNSPEC_UQSHRNB, UNSPEC_UQSHRNT, UNSPEC_UQXTNB, UNSPEC_UQXTNT)
4427 (UNSPEC_USHLLB, UNSPEC_USHLLT, UNSPEC_USUBLB, UNSPEC_USUBLT)
4428 (UNSPEC_USUBWB, UNSPEC_USUBWT): New unspecs.
4429 (UNSPEC_SMULLB, UNSPEC_SMULLT, UNSPEC_UMULLB, UNSPEC_UMULLT)
4430 (UNSPEC_SMULHS, UNSPEC_SMULHRS, UNSPEC_UMULHS, UNSPEC_UMULHRS)
4431 (UNSPEC_RSHRNB, UNSPEC_RSHRNT, UNSPEC_SHRNB, UNSPEC_SHRNT): Move
4433 (VNARROW, Ventype): New mode attributes.
4434 (Vewtype): Handle VNx2DI. Fix typo in comment.
4435 (VDOUBLE): New mode attribute.
4436 (sve_lane_con): Handle VNx8HI.
4437 (SVE_INT_UNARY): Include ss_abs and ss_neg for TARGET_SVE2.
4438 (SVE_INT_BINARY): Likewise ss_plus, us_plus, ss_minus and us_minus.
4439 (sve_int_op, sve_int_op_rev): Handle the above codes.
4440 (sve_pred_int_rhs2_operand): Likewise.
4441 (MULLBT, SHRNB, SHRNT): Delete.
4442 (SVE_INT_SHIFT_IMM): New int iterator.
4443 (SVE_WHILE): Add UNSPEC_WHILEGE, UNSPEC_WHILEGT, UNSPEC_WHILEHI
4444 and UNSPEC_WHILEHS for TARGET_SVE2.
4445 (SVE2_U32_UNARY, SVE2_INT_UNARY_NARROWB, SVE2_INT_UNARY_NARROWT)
4446 (SVE2_INT_BINARY, SVE2_INT_BINARY_LANE, SVE2_INT_BINARY_LONG)
4447 (SVE2_INT_BINARY_LONG_LANE, SVE2_INT_BINARY_NARROWB)
4448 (SVE2_INT_BINARY_NARROWT, SVE2_INT_BINARY_PAIR, SVE2_FP_BINARY_PAIR)
4449 (SVE2_INT_BINARY_PAIR_LONG, SVE2_INT_BINARY_WIDE): New int iterators.
4450 (SVE2_INT_SHIFT_IMM_LONG, SVE2_INT_SHIFT_IMM_NARROWB): Likewise.
4451 (SVE2_INT_SHIFT_IMM_NARROWT, SVE2_INT_SHIFT_INSERT, SVE2_INT_CADD)
4452 (SVE2_INT_BITPERM, SVE2_INT_TERNARY, SVE2_INT_TERNARY_LANE): Likewise.
4453 (SVE2_FP_TERNARY_LONG, SVE2_FP_TERNARY_LONG_LANE, SVE2_INT_CMLA)
4454 (SVE2_INT_CDOT, SVE2_INT_ADD_BINARY_LONG, SVE2_INT_QADD_BINARY_LONG)
4455 (SVE2_INT_SUB_BINARY_LONG, SVE2_INT_QSUB_BINARY_LONG): Likewise.
4456 (SVE2_INT_ADD_BINARY_LONG_LANE, SVE2_INT_QADD_BINARY_LONG_LANE)
4457 (SVE2_INT_SUB_BINARY_LONG_LANE, SVE2_INT_QSUB_BINARY_LONG_LANE)
4458 (SVE2_COND_INT_UNARY_FP, SVE2_COND_FP_UNARY_LONG): Likewise.
4459 (SVE2_COND_FP_UNARY_NARROWB, SVE2_COND_INT_BINARY): Likewise.
4460 (SVE2_COND_INT_BINARY_NOREV, SVE2_COND_INT_BINARY_REV): Likewise.
4461 (SVE2_COND_INT_SHIFT, SVE2_MATCH, SVE2_PMULL): Likewise.
4462 (optab): Handle the new unspecs.
4463 (su, r): Remove entries for UNSPEC_SHRNB, UNSPEC_SHRNT, UNSPEC_RSHRNB
4465 (lr): Handle the new unspecs.
4467 (cmp_op, while_optab_cmp, sve_int_op): Handle the new unspecs.
4468 (sve_int_op_rev, sve_int_add_op, sve_int_qadd_op, sve_int_sub_op)
4469 (sve_int_qsub_op): New int attributes.
4470 (sve_fp_op, rot): Handle the new unspecs.
4471 * config/aarch64/aarch64-sve-builtins.h
4472 (function_resolver::require_matching_pointer_type): Declare.
4473 (function_resolver::resolve_unary): Add an optional boolean argument.
4474 (function_resolver::finish_opt_n_resolution): Add an optional
4475 type_suffix_index argument.
4476 (gimple_folder::redirect_call): Declare.
4477 (gimple_expander::prepare_gather_address_operands): Add an optional
4479 * config/aarch64/aarch64-sve-builtins.cc: Include
4480 aarch64-sve-builtins-sve2.h.
4481 (TYPES_b_unsigned, TYPES_b_integer, TYPES_bh_integer): New macros.
4482 (TYPES_bs_unsigned, TYPES_hs_signed, TYPES_hs_integer): Likewise.
4483 (TYPES_hd_unsigned, TYPES_hsd_signed): Likewise.
4484 (TYPES_hsd_integer): Use TYPES_hsd_signed.
4485 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): New macros.
4486 (TYPES_s_unsigned): Likewise.
4487 (TYPES_s_integer): Use TYPES_s_unsigned.
4488 (TYPES_sd_signed, TYPES_sd_unsigned): New macros.
4489 (TYPES_sd_integer): Use them.
4490 (TYPES_d_unsigned): New macro.
4491 (TYPES_d_integer): Use it.
4492 (TYPES_d_data, TYPES_cvt_long, TYPES_cvt_narrow_s): New macros.
4493 (TYPES_cvt_narrow): Likewise.
4494 (DEF_SVE_TYPES_ARRAY): Include the new types macros above.
4495 (preds_mx): New variable.
4496 (function_builder::add_overloaded_function): Allow the new feature
4497 set to be more restrictive than the original one.
4498 (function_resolver::infer_pointer_type): Remove qualifiers from
4499 the pointer type before printing it.
4500 (function_resolver::require_matching_pointer_type): New function.
4501 (function_resolver::resolve_sv_displacement): Handle functions
4502 that don't support 32-bit vector indices or svint32_t vector offsets.
4503 (function_resolver::finish_opt_n_resolution): Take the inferred type
4504 as a separate argument.
4505 (function_resolver::resolve_unary): Optionally treat all forms in
4506 the same way as normal merging functions.
4507 (gimple_folder::redirect_call): New function.
4508 (function_expander::prepare_gather_address_operands): Add an argument
4509 that says whether scaled forms are available. If they aren't,
4510 handle scaling of vector indices and don't add the extension and
4512 (function_expander::map_to_unspecs): If aarch64_sve isn't available,
4513 fall back to using cond_* instead.
4514 * config/aarch64/aarch64-sve-builtins-functions.h (rtx_code_function):
4515 Split out the member variables into...
4516 (rtx_code_function_base): ...this new base class.
4517 (rtx_code_function_rotated): Inherit rtx_code_function_base.
4518 (unspec_based_function): Split out the member variables into...
4519 (unspec_based_function_base): ...this new base class.
4520 (unspec_based_function_rotated): Inherit unspec_based_function_base.
4521 (unspec_based_function_exact_insn): New class.
4522 (unspec_based_add_function, unspec_based_add_lane_function)
4523 (unspec_based_lane_function, unspec_based_pred_function)
4524 (unspec_based_qadd_function, unspec_based_qadd_lane_function)
4525 (unspec_based_qsub_function, unspec_based_qsub_lane_function)
4526 (unspec_based_sub_function, unspec_based_sub_lane_function): New
4528 (unspec_based_fused_function): New class.
4529 (unspec_based_mla_function, unspec_based_mls_function): New typedefs.
4530 (unspec_based_fused_lane_function): New class.
4531 (unspec_based_mla_lane_function, unspec_based_mls_lane_function): New
4533 (CODE_FOR_MODE1): New macro.
4534 (fixed_insn_function): New class.
4535 (while_comparison): Likewise.
4536 * config/aarch64/aarch64-sve-builtins-shapes.h (binary_long_lane)
4537 (binary_long_opt_n, binary_narrowb_opt_n, binary_narrowt_opt_n)
4538 (binary_to_uint, binary_wide, binary_wide_opt_n, compare, compare_ptr)
4539 (load_ext_gather_index_restricted, load_ext_gather_offset_restricted)
4540 (load_gather_sv_restricted, shift_left_imm_long): Declare.
4541 (shift_left_imm_to_uint, shift_right_imm_narrowb): Likewise.
4542 (shift_right_imm_narrowt, shift_right_imm_narrowb_to_uint): Likewise.
4543 (shift_right_imm_narrowt_to_uint, store_scatter_index_restricted)
4544 (store_scatter_offset_restricted, tbl_tuple, ternary_long_lane)
4545 (ternary_long_opt_n, ternary_qq_lane_rotate, ternary_qq_rotate)
4546 (ternary_shift_left_imm, ternary_shift_right_imm, ternary_uint)
4547 (unary_convert_narrowt, unary_long, unary_narrowb, unary_narrowt)
4548 (unary_narrowb_to_uint, unary_narrowt_to_uint, unary_to_int): Likewise.
4549 * config/aarch64/aarch64-sve-builtins-shapes.cc (apply_predication):
4550 Also add an initial argument for unary_convert_narrowt, regardless
4551 of the predication type.
4552 (build_32_64): Allow loads and stores to specify MODE_none.
4553 (build_sv_index64, build_sv_uint_offset): New functions.
4554 (long_type_suffix): New function.
4555 (binary_imm_narrowb_base, binary_imm_narrowt_base): New classes.
4556 (binary_imm_long_base, load_gather_sv_base): Likewise.
4557 (shift_right_imm_narrow_wrapper, ternary_shift_imm_base): Likewise.
4558 (ternary_resize2_opt_n_base, ternary_resize2_lane_base): Likewise.
4559 (unary_narrowb_base, unary_narrowt_base): Likewise.
4560 (binary_long_lane_def, binary_long_lane): New shape.
4561 (binary_long_opt_n_def, binary_long_opt_n): Likewise.
4562 (binary_narrowb_opt_n_def, binary_narrowb_opt_n): Likewise.
4563 (binary_narrowt_opt_n_def, binary_narrowt_opt_n): Likewise.
4564 (binary_to_uint_def, binary_to_uint): Likewise.
4565 (binary_wide_def, binary_wide): Likewise.
4566 (binary_wide_opt_n_def, binary_wide_opt_n): Likewise.
4567 (compare_def, compare): Likewise.
4568 (compare_ptr_def, compare_ptr): Likewise.
4569 (load_ext_gather_index_restricted_def,
4570 load_ext_gather_index_restricted): Likewise.
4571 (load_ext_gather_offset_restricted_def,
4572 load_ext_gather_offset_restricted): Likewise.
4573 (load_gather_sv_def): Inherit from load_gather_sv_base.
4574 (load_gather_sv_restricted_def, load_gather_sv_restricted): New shape.
4575 (shift_left_imm_def, shift_left_imm): Likewise.
4576 (shift_left_imm_long_def, shift_left_imm_long): Likewise.
4577 (shift_left_imm_to_uint_def, shift_left_imm_to_uint): Likewise.
4578 (store_scatter_index_restricted_def,
4579 store_scatter_index_restricted): Likewise.
4580 (store_scatter_offset_restricted_def,
4581 store_scatter_offset_restricted): Likewise.
4582 (tbl_tuple_def, tbl_tuple): Likewise.
4583 (ternary_long_lane_def, ternary_long_lane): Likewise.
4584 (ternary_long_opt_n_def, ternary_long_opt_n): Likewise.
4585 (ternary_qq_lane_def): Inherit from ternary_resize2_lane_base.
4586 (ternary_qq_lane_rotate_def, ternary_qq_lane_rotate): New shape
4587 (ternary_qq_opt_n_def): Inherit from ternary_resize2_opt_n_base.
4588 (ternary_qq_rotate_def, ternary_qq_rotate): New shape.
4589 (ternary_shift_left_imm_def, ternary_shift_left_imm): Likewise.
4590 (ternary_shift_right_imm_def, ternary_shift_right_imm): Likewise.
4591 (ternary_uint_def, ternary_uint): Likewise.
4592 (unary_convert): Fix typo in comment.
4593 (unary_convert_narrowt_def, unary_convert_narrowt): New shape.
4594 (unary_long_def, unary_long): Likewise.
4595 (unary_narrowb_def, unary_narrowb): Likewise.
4596 (unary_narrowt_def, unary_narrowt): Likewise.
4597 (unary_narrowb_to_uint_def, unary_narrowb_to_uint): Likewise.
4598 (unary_narrowt_to_uint_def, unary_narrowt_to_uint): Likewise.
4599 (unary_to_int_def, unary_to_int): Likewise.
4600 * config/aarch64/aarch64-sve-builtins-base.cc (unspec_cmla)
4601 (unspec_fcmla, unspec_cond_fcmla, expand_mla_mls_lane): New functions.
4602 (svasrd_impl): Delete.
4603 (svcadd_impl::expand): Handle integer operations too.
4604 (svcmla_impl::expand, svcmla_lane::expand): Likewise, using the
4605 new functions to derive the unspec numbers.
4606 (svmla_svmls_lane_impl): Replace with...
4607 (svmla_lane_impl, svmls_lane_impl): ...these new classes. Handle
4608 integer operations too.
4609 (svwhile_impl): Rename to...
4610 (svwhilelx_impl): ...this and inherit from while_comparison.
4611 (svasrd): Use unspec_based_function.
4612 (svmla_lane): Use svmla_lane_impl.
4613 (svmls_lane): Use svmls_lane_impl.
4614 (svrecpe, svrsqrte): Handle unsigned integer operations too.
4615 (svwhilele, svwhilelt): Use svwhilelx_impl.
4616 * config/aarch64/aarch64-sve-builtins-sve2.h: New file.
4617 * config/aarch64/aarch64-sve-builtins-sve2.cc: Likewise.
4618 * config/aarch64/aarch64-sve-builtins-sve2.def: Likewise.
4619 * config/aarch64/aarch64-sve-builtins.def: Include
4620 aarch64-sve-builtins-sve2.def.
4622 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4624 * config/aarch64/aarch64-protos.h (aarch64_sve_arith_immediate_p)
4625 (aarch64_sve_sqadd_sqsub_immediate_p): Add a machine_mode argument.
4626 * config/aarch64/aarch64.c (aarch64_sve_arith_immediate_p)
4627 (aarch64_sve_sqadd_sqsub_immediate_p): Likewise. Handle scalar
4628 immediates as well as vector ones.
4629 * config/aarch64/predicates.md (aarch64_sve_arith_immediate)
4630 (aarch64_sve_sub_arith_immediate, aarch64_sve_qadd_immediate)
4631 (aarch64_sve_qsub_immediate): Update calls accordingly.
4633 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4635 * config/aarch64/aarch64-sve2.md: Add banner comments.
4636 (<su>mulh<r>s<mode>3): Move further up file.
4637 (<su>mull<bt><Vwide>, <r>shrnb<mode>, <r>shrnt<mode>)
4638 (*aarch64_sve2_sra<mode>): Move further down file.
4639 * config/aarch64/t-aarch64 (s-check-sve-md): Check aarch64-sve2.md too.
4641 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4643 * config/aarch64/iterators.md (SVE_WHILE): Add UNSPEC_WHILERW
4645 (while_optab_cmp): Handle them.
4646 * config/aarch64/aarch64-sve.md
4647 (*while_<while_optab_cmp><GPI:mode><PRED_ALL:mode>_ptest): Make public
4648 and add a "@" marker.
4649 * config/aarch64/aarch64-sve2.md (check_<raw_war>_ptrs<mode>): Use it
4650 instead of gen_aarch64_sve2_while_ptest.
4651 (@aarch64_sve2_while<cmp_op><GPI:mode><PRED_ALL:mode>_ptest): Delete.
4653 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4655 * config/aarch64/aarch64.md (UNSPEC_WHILE_LE): Rename to...
4656 (UNSPEC_WHILELE): ...this.
4657 (UNSPEC_WHILE_LO): Rename to...
4658 (UNSPEC_WHILELO): ...this.
4659 (UNSPEC_WHILE_LS): Rename to...
4660 (UNSPEC_WHILELS): ...this.
4661 (UNSPEC_WHILE_LT): Rename to...
4662 (UNSPEC_WHILELT): ...this.
4663 * config/aarch64/iterators.md (SVE_WHILE): Update accordingly.
4664 (cmp_op, while_optab_cmp): Likewise.
4665 * config/aarch64/aarch64.c (aarch64_sve_move_pred_via_while): Likewise.
4666 * config/aarch64/aarch64-sve-builtins-base.cc (svwhilele): Likewise.
4667 (svwhilelt): Likewise.
4669 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4671 * config/aarch64/aarch64-sve-builtins-shapes.h (unary_count): Delete.
4672 (unary_to_uint): Define.
4673 * config/aarch64/aarch64-sve-builtins-shapes.cc (unary_count_def)
4674 (unary_count): Rename to...
4675 (unary_to_uint_def, unary_to_uint): ...this.
4676 * config/aarch64/aarch64-sve-builtins-base.def: Update accordingly.
4678 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4680 * config/aarch64/aarch64-sve-builtins-functions.h
4681 (code_for_mode_function): New class.
4682 (CODE_FOR_MODE0, QUIET_CODE_FOR_MODE0): New macros.
4683 * config/aarch64/aarch64-sve-builtins-base.cc (svcompact_impl)
4684 (svext_impl, svmul_lane_impl, svsplice_impl, svtmad_impl): Delete.
4685 (svcompact, svext, svsplice): Use QUIET_CODE_FOR_MODE0.
4686 (svmul_lane, svtmad): Use CODE_FOR_MODE0.
4688 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4690 * config/aarch64/iterators.md (addsub): New code attribute.
4691 * config/aarch64/aarch64-simd.md (aarch64_<su_optab><optab><mode>):
4693 (aarch64_<su_optab>q<addsub><mode>): ...this, making the same change
4694 in the asm string and attributes. Fix indentation.
4695 * config/aarch64/aarch64-sve.md (@aarch64_<su_optab><optab><mode>):
4697 (@aarch64_sve_<optab><mode>): ...this.
4698 * config/aarch64/aarch64-sve-builtins.h
4699 (function_expander::expand_signed_unpred_op): Delete.
4700 * config/aarch64/aarch64-sve-builtins.cc
4701 (function_expander::expand_signed_unpred_op): Likewise.
4702 (function_expander::map_to_rtx_codes): If the optab isn't defined,
4703 try using code_for_aarch64_sve instead.
4704 * config/aarch64/aarch64-sve-builtins-base.cc (svqadd_impl): Delete.
4705 (svqsub_impl): Likewise.
4706 (svqadd, svqsub): Use rtx_code_function instead.
4708 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4710 * config/aarch64/iterators.md (SRHSUB, URHSUB): Delete.
4711 (HADDSUB, sur, addsub): Remove them.
4713 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4715 * tree-nrv.c (pass_return_slot::execute): Handle all internal
4716 functions the same way, rather than singling out those that
4717 aren't mapped directly to optabs.
4719 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4721 * target.def (compatible_vector_types_p): New target hook.
4722 * hooks.h (hook_bool_const_tree_const_tree_true): Declare.
4723 * hooks.c (hook_bool_const_tree_const_tree_true): New function.
4724 * doc/tm.texi.in (TARGET_COMPATIBLE_VECTOR_TYPES_P): New hook.
4725 * doc/tm.texi: Regenerate.
4726 * gimple-expr.c: Include target.h.
4727 (useless_type_conversion_p): Use targetm.compatible_vector_types_p.
4728 * config/aarch64/aarch64.c (aarch64_compatible_vector_types_p): New
4730 (TARGET_COMPATIBLE_VECTOR_TYPES_P): Define.
4731 * config/aarch64/aarch64-sve-builtins.cc (gimple_folder::convert_pred):
4732 Use the original predicate if it already has a suitable type.
4734 2020-01-09 Martin Jambor <mjambor@suse.cz>
4736 * cgraph.h (cgraph_edge): Make remove, set_call_stmt, make_direct,
4737 resolve_speculation and redirect_call_stmt_to_callee static. Change
4738 return type of set_call_stmt to cgraph_edge *.
4739 * auto-profile.c (afdo_indirect_call): Adjust call to
4740 redirect_call_stmt_to_callee.
4741 * cgraph.c (cgraph_edge::set_call_stmt): Make return cgraph-edge *,
4742 make the this pointer explicit, adjust self-recursive calls and the
4743 call top make_direct. Return the resulting edge.
4744 (cgraph_edge::remove): Make this pointer explicit.
4745 (cgraph_edge::resolve_speculation): Likewise, adjust call to remove.
4746 (cgraph_edge::make_direct): Likewise, adjust call to
4747 resolve_speculation.
4748 (cgraph_edge::redirect_call_stmt_to_callee): Likewise, also adjust
4749 call to set_call_stmt.
4750 (cgraph_update_edges_for_call_stmt_node): Update call to
4751 set_call_stmt and remove.
4752 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
4753 Renamed edge to master_edge. Adjusted calls to set_call_stmt.
4754 (cgraph_node::create_edge_including_clones): Moved "first" definition
4755 of edge to the block where it was used. Adjusted calls to
4757 (cgraph_node::remove_symbol_and_inline_clones): Adjust call to
4758 cgraph_edge::remove.
4759 * cgraphunit.c (walk_polymorphic_call_targets): Adjusted calls to
4760 make_direct and redirect_call_stmt_to_callee.
4761 * ipa-fnsummary.c (redirect_to_unreachable): Adjust calls to
4762 resolve_speculation and make_direct.
4763 * ipa-inline-transform.c (inline_transform): Adjust call to
4764 redirect_call_stmt_to_callee.
4765 (check_speculations_1):: Adjust call to resolve_speculation.
4766 * ipa-inline.c (resolve_noninline_speculation): Adjust call to
4767 resolve-speculation.
4768 (inline_small_functions): Adjust call to resolve_speculation.
4769 (ipa_inline): Likewise.
4770 * ipa-prop.c (ipa_make_edge_direct_to_target): Adjust call to
4772 * ipa-visibility.c (function_and_variable_visibility): Make iteration
4773 safe with regards to edge removal, adjust calls to
4774 redirect_call_stmt_to_callee.
4775 * ipa.c (walk_polymorphic_call_targets): Adjust calls to make_direct
4776 and redirect_call_stmt_to_callee.
4777 * multiple_target.c (create_dispatcher_calls): Adjust call to
4778 redirect_call_stmt_to_callee
4779 (redirect_to_specific_clone): Likewise.
4780 * tree-cfgcleanup.c (delete_unreachable_blocks_update_callgraph):
4781 Adjust calls to cgraph_edge::remove.
4782 * tree-inline.c (copy_bb): Adjust call to set_call_stmt.
4783 (redirect_all_calls): Adjust call to redirect_call_stmt_to_callee.
4784 (expand_call_inline): Adjust call to cgraph_edge::remove.
4786 2020-01-09 Martin Liska <mliska@suse.cz>
4788 * params.opt: Set Optimization for
4789 param_max_speculative_devirt_maydefs.
4791 2020-01-09 Martin Sebor <msebor@redhat.com>
4795 * builtins.c (compute_objsize): Avoid handling MEM_REFs of vector type.
4797 2020-01-09 Martin Liska <mliska@suse.cz>
4799 * auto-profile.c (auto_profile): Use opt_for_fn
4801 * ipa-cp.c (ipcp_lattice::add_value): Likewise.
4802 (propagate_vals_across_arith_jfunc): Likewise.
4803 (hint_time_bonus): Likewise.
4804 (incorporate_penalties): Likewise.
4805 (good_cloning_opportunity_p): Likewise.
4806 (perform_estimation_of_a_value): Likewise.
4807 (estimate_local_effects): Likewise.
4808 (ipcp_propagate_stage): Likewise.
4809 * ipa-fnsummary.c (decompose_param_expr): Likewise.
4810 (set_switch_stmt_execution_predicate): Likewise.
4811 (analyze_function_body): Likewise.
4812 * ipa-inline-analysis.c (offline_size): Likewise.
4813 * ipa-inline.c (early_inliner): Likewise.
4814 * ipa-prop.c (ipa_analyze_node): Likewise.
4815 (ipcp_transform_function): Likewise.
4816 * ipa-sra.c (process_scan_results): Likewise.
4817 (ipa_sra_summarize_function): Likewise.
4818 * params.opt: Rename ipcp-unit-growth to
4819 ipa-cp-unit-growth. Add Optimization for various
4820 IPA-related parameters.
4822 2020-01-09 Richard Biener <rguenther@suse.de>
4825 * gimplify.c (gimplify_expr): Deal with NOP definitions.
4827 2020-01-09 Richard Biener <rguenther@suse.de>
4829 PR tree-optimization/93040
4830 * gimple-ssa-store-merging.c (find_bswap_or_nop): Raise search limit.
4832 2020-01-09 Georg-Johann Lay <avr@gjlay.de>
4834 * common/config/avr/avr-common.c (avr_option_optimization_table)
4835 [OPT_LEVELS_1_PLUS]: Set -fsplit-wide-types-early.
4837 2020-01-09 Martin Liska <mliska@suse.cz>
4839 * cgraphclones.c (symbol_table::materialize_all_clones):
4840 Use cgraph_node::dump_name.
4842 2020-01-09 Jakub Jelinek <jakub@redhat.com>
4845 * config/riscv/riscv.c (riscv_print_operand_reloc): Use
4846 output_operand_lossage instead of gcc_unreachable.
4847 * doc/md.texi (riscv f constraint): Fix typo.
4850 * config/i386/i386.md (subv<mode>4): Use SWIDWI iterator instead of
4851 SWI. Use <general_hilo_operand> instead of <general_operand>. Use
4852 CONST_SCALAR_INT_P instead of CONST_INT_P.
4853 (*subv<mode>4_1): Rename to ...
4854 (subv<mode>4_1): ... this.
4855 (*subv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
4856 define_insn_and_split patterns.
4857 (*subv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
4860 2020-01-08 David Malcolm <dmalcolm@redhat.com>
4862 * vec.c (class selftest::count_dtor): New class.
4863 (selftest::test_auto_delete_vec): New test.
4864 (selftest::vec_c_tests): Call it.
4865 * vec.h (class auto_delete_vec): New class template.
4866 (auto_delete_vec<T>::~auto_delete_vec): New dtor.
4868 2020-01-08 David Malcolm <dmalcolm@redhat.com>
4870 * sbitmap.h (auto_sbitmap): Add operator const_sbitmap.
4872 2020-01-08 Jim Wilson <jimw@sifive.com>
4874 * config/riscv/riscv.c (riscv_legitimize_tls_address): Ifdef out
4875 use of TLS_MODEL_LOCAL_EXEC when not pic.
4877 2020-01-08 David Malcolm <dmalcolm@redhat.com>
4879 * hash-map-tests.c (selftest::test_map_of_strings_to_int): Fix
4882 2020-01-08 Jakub Jelinek <jakub@redhat.com>
4885 * config/i386/i386.md (*stack_protect_set_2_<mode> peephole2,
4886 *stack_protect_set_3 peephole2): Also check that the second
4887 insns source is general_operand.
4890 * config/i386/i386.md (addcarry<mode>_0): Use nonimmediate_operand
4891 predicate for output operand instead of register_operand.
4892 (addcarry<mode>, addcarry<mode>_1): Likewise. Add alternative with
4893 memory destination and non-memory operands[2].
4895 2020-01-08 Martin Liska <mliska@suse.cz>
4897 * cgraph.c (cgraph_node::dump): Use ::dump_name or
4898 ::dump_asm_name instead of (::name or ::asm_name).
4899 * cgraphclones.c (symbol_table::materialize_all_clones): Likewise.
4900 * cgraphunit.c (walk_polymorphic_call_targets): Likewise.
4901 (analyze_functions): Likewise.
4902 (expand_all_functions): Likewise.
4903 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
4904 (propagate_bits_across_jump_function): Likewise.
4905 (dump_profile_updates): Likewise.
4906 (ipcp_store_bits_results): Likewise.
4907 (ipcp_store_vr_results): Likewise.
4908 * ipa-devirt.c (dump_targets): Likewise.
4909 * ipa-fnsummary.c (analyze_function_body): Likewise.
4910 * ipa-hsa.c (check_warn_node_versionable): Likewise.
4911 (process_hsa_functions): Likewise.
4912 * ipa-icf.c (sem_item_optimizer::merge_classes): Likewise.
4913 (set_alias_uids): Likewise.
4914 * ipa-inline-transform.c (save_inline_function_body): Likewise.
4915 * ipa-inline.c (recursive_inlining): Likewise.
4916 (inline_to_all_callers_1): Likewise.
4917 (ipa_inline): Likewise.
4918 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
4919 (ipa_propagate_frequency): Likewise.
4920 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
4921 (remove_described_reference): Likewise.
4922 * ipa-pure-const.c (worse_state): Likewise.
4923 (check_retval_uses): Likewise.
4924 (analyze_function): Likewise.
4925 (propagate_pure_const): Likewise.
4926 (propagate_nothrow): Likewise.
4927 (dump_malloc_lattice): Likewise.
4928 (propagate_malloc): Likewise.
4929 (pass_local_pure_const::execute): Likewise.
4930 * ipa-visibility.c (optimize_weakref): Likewise.
4931 (function_and_variable_visibility): Likewise.
4932 * ipa.c (symbol_table::remove_unreachable_nodes): Likewise.
4933 (ipa_discover_variable_flags): Likewise.
4934 * lto-streamer-out.c (output_function): Likewise.
4935 (output_constructor): Likewise.
4936 * tree-inline.c (copy_bb): Likewise.
4937 * tree-ssa-structalias.c (ipa_pta_execute): Likewise.
4938 * varpool.c (symbol_table::remove_unreferenced_decls): Likewise.
4940 2020-01-08 Richard Biener <rguenther@suse.de>
4943 * tree-eh.c (sink_clobbers): Update virtual operands for
4944 the first and last stmt only. Add a dry-run capability.
4945 (pass_lower_eh_dispatch::execute): Perform clobber sinking
4946 after CFG manipulations and in RPO order to catch all
4947 secondary opportunities reliably.
4949 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
4952 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
4954 2019-01-08 Richard Biener <rguenther@suse.de>
4957 * gimple-fold.c (rewrite_to_defined_overflow): Mark stmt modified.
4958 * tree-ssa-loop-im.c (move_computations_worker): Properly adjust
4959 virtual operand, also updating SSA use.
4960 * gimple-loop-interchange.cc (loop_cand::undo_simple_reduction):
4961 Update stmt after resetting virtual operand.
4962 (tree_loop_interchange::move_code_to_inner_loop): Likewise.
4963 * gimple-iterator.c (gsi_remove): When not removing the stmt
4964 permanently do not delink immediate uses or mark the stmt modified.
4966 2020-01-08 Martin Liska <mliska@suse.cz>
4968 * ipa-fnsummary.c (dump_ipa_call_summary): Use symtab_node::dump_name.
4969 (ipa_call_context::estimate_size_and_time): Likewise.
4970 (inline_analyze_function): Likewise.
4972 2020-01-08 Martin Liska <mliska@suse.cz>
4974 * cgraph.c (cgraph_node::dump): Use systematically
4977 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
4979 Add -nodevicespecs option for avr.
4982 * config/avr/avr.opt (-nodevicespecs): New driver option.
4983 * config/avr/driver-avr.c (avr_devicespecs_file): Only issue
4984 "-specs=device-specs/..." if that option is not set.
4985 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
4987 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
4989 Implement 64-bit double functions for avr.
4992 * config.gcc (tm_defines) [target=avr]: Support --with-libf7,
4993 --with-double-comparison.
4994 * doc/install.texi: Document them.
4995 * config/avr/avr-c.c (avr_cpu_cpp_builtins)
4996 <WITH_LIBF7_LIBGCC, WITH_LIBF7_MATH, WITH_LIBF7_MATH_SYMBOLS>
4997 <WITH_DOUBLE_COMPARISON>: New built-in defines.
4998 * doc/invoke.texi (AVR Built-in Macros): Document them.
4999 * config/avr/avr-protos.h (avr_float_lib_compare_returns_bool): New.
5000 * config/avr/avr.c (avr_float_lib_compare_returns_bool): New function.
5001 * config/avr/avr.h (FLOAT_LIB_COMPARE_RETURNS_BOOL): New macro.
5003 2020-01-08 Richard Earnshaw <rearnsha@arm.com>
5006 * config/arm/t-multilib (MULTILIB_MATCHES): Add rules to match
5007 armv7-a{+mp,+sec,+mp+sec} to appropriate armv7 multilib variants
5008 when only building rm-profile multilibs.
5010 2020-01-08 Feng Xue <fxue@os.amperecomputing.com>
5013 * ipa-cp.c (self_recursively_generated_p): Find matched aggregate
5014 lattice for a value to check.
5015 (propagate_vals_across_arith_jfunc): Add an assertion to ensure
5016 finite propagation in self-recursive scc.
5018 2020-01-08 Luo Xiong Hu <luoxhu@linux.ibm.com>
5020 * ipa-inline.c (caller_growth_limits): Restore the AND.
5022 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
5024 * config/gcn/gcn-valu.md (VEC_1REG_INT_ALT): Delete iterator.
5025 (VEC_ALLREG_ALT): New iterator.
5026 (VEC_ALLREG_INT_MODE): New iterator.
5027 (VCMP_MODE): New iterator.
5028 (VCMP_MODE_INT): New iterator.
5029 (vec_cmpu<mode>di): Use VCMP_MODE_INT.
5030 (vec_cmp<u>v64qidi): New define_expand.
5031 (vec_cmp<mode>di_exec): Use VCMP_MODE.
5032 (vec_cmpu<mode>di_exec): New define_expand.
5033 (vec_cmp<u>v64qidi_exec): New define_expand.
5034 (vec_cmp<mode>di_dup): Use VCMP_MODE.
5035 (vec_cmp<mode>di_dup_exec): Use VCMP_MODE.
5036 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>): Rename ...
5037 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): ... to this.
5038 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>_exec): Rename ...
5039 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): ... to this.
5040 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>): Rename ...
5041 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): ... to this.
5042 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>_exec): Rename ...
5043 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): ... to
5045 * config/gcn/gcn.c (print_operand): Fix 8 and 16 bit suffixes.
5046 * config/gcn/gcn.md (expander): Add sign_extend and zero_extend.
5048 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
5050 * config/gcn/constraints.md (DA): Update description and match.
5052 (Db): New constraint.
5053 * config/gcn/gcn-protos.h (gcn_inline_constant64_p): Add second
5055 * config/gcn/gcn.c (gcn_inline_constant64_p): Add 'mixed' parameter.
5056 Implement 'Db' mixed immediate type.
5057 * config/gcn/gcn-valu.md (addcv64si3<exec_vcc>): Rework constraints.
5058 (addcv64si3_dup<exec_vcc>): Delete.
5059 (subcv64si3<exec_vcc>): Rework constraints.
5060 (addv64di3): Rework constraints.
5061 (addv64di3_exec): Rework constraints.
5062 (subv64di3): Rework constraints.
5063 (addv64di3_dup): Delete.
5064 (addv64di3_dup_exec): Delete.
5065 (addv64di3_zext): Rework constraints.
5066 (addv64di3_zext_exec): Rework constraints.
5067 (addv64di3_zext_dup): Rework constraints.
5068 (addv64di3_zext_dup_exec): Rework constraints.
5069 (addv64di3_zext_dup2): Rework constraints.
5070 (addv64di3_zext_dup2_exec): Rework constraints.
5071 (addv64di3_sext_dup2): Rework constraints.
5072 (addv64di3_sext_dup2_exec): Rework constraints.
5074 2020-01-07 Andre Vieira <andre.simoesdiasvieira@arm.com>
5076 * doc/sourcebuild.texi (arm_little_endian, arm_nothumb): Documented
5077 existing target checks.
5079 2020-01-07 Richard Biener <rguenther@suse.de>
5081 * doc/install.texi: Bump minimal supported MPC version.
5083 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
5085 * langhooks-def.h (lhd_simulate_enum_decl): Declare.
5086 (LANG_HOOKS_SIMULATE_ENUM_DECL): Use it.
5087 * langhooks.c: Include stor-layout.h.
5088 (lhd_simulate_enum_decl): New function.
5089 * config/aarch64/aarch64-sve-builtins.cc (init_builtins): Call
5090 handle_arm_sve_h for the LTO frontend.
5091 (register_vector_type): Cope with null returns from pushdecl.
5093 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
5095 * config/aarch64/aarch64-protos.h (aarch64_sve::svbool_type_p)
5096 (aarch64_sve::nvectors_if_data_type): Replace with...
5097 (aarch64_sve::builtin_type_p): ...this.
5098 * config/aarch64/aarch64-sve-builtins.cc: Include attribs.h.
5099 (find_vector_type): Delete.
5100 (add_sve_type_attribute): New function.
5101 (lookup_sve_type_attribute): Likewise.
5102 (register_builtin_types): Add an "SVE type" attribute to each type.
5103 (register_tuple_type): Likewise.
5104 (svbool_type_p, nvectors_if_data_type): Delete.
5105 (mangle_builtin_type): Use lookup_sve_type_attribute.
5106 (builtin_type_p): Likewise. Add an overload that returns the
5107 number of constituent vector and predicate registers.
5108 * config/aarch64/aarch64.c (aarch64_sve_argument_p): Delete.
5109 (aarch64_returns_value_in_sve_regs_p): Use aarch64_sve::builtin_type_p
5110 instead of aarch64_sve_argument_p.
5111 (aarch64_takes_arguments_in_sve_regs_p): Likewise.
5112 (aarch64_pass_by_reference): Likewise.
5113 (aarch64_function_value_1): Likewise.
5114 (aarch64_return_in_memory): Likewise.
5115 (aarch64_layout_arg): Likewise.
5117 2020-01-07 Jakub Jelinek <jakub@redhat.com>
5119 PR tree-optimization/93156
5120 * tree-ssa-ccp.c (bit_value_binop): For x * x note that the second
5121 least significant bit is always clear.
5123 PR tree-optimization/93118
5124 * match.pd ((x >> c) << c -> x & (-1<<c)): Add nop_convert?. Add new
5125 simplifier with two intermediate conversions.
5127 2020-01-07 Martin Liska <mliska@suse.cz>
5129 * params.opt: Add Optimization for various parameters.
5131 2020-01-07 Martin Liska <mliska@suse.cz>
5134 * doc/extend.texi: Explain cloning for target_clone
5137 2020-01-07 Martin Liska <mliska@suse.cz>
5139 PR tree-optimization/92860
5140 * common.opt: Make in Optimization option
5141 as it is affected by -O0, which is an Optimization
5143 * tree-inline.c (tree_inlinable_function_p):
5144 Use opt_for_fn for warn_inline.
5145 (expand_call_inline): Likewise.
5147 2020-01-07 Martin Liska <mliska@suse.cz>
5149 PR tree-optimization/92860
5150 * common.opt: Make flag_ree as optimization
5153 2020-01-07 Martin Liska <mliska@suse.cz>
5155 PR optimization/92860
5156 * params.opt: Mark param_min_crossjump_insns with Optimization
5159 2020-01-07 Luo Xiong Hu <luoxhu@linux.ibm.com>
5161 * ipa-inline-analysis.c (estimate_growth): Fix typo.
5162 * ipa-inline.c (caller_growth_limits): Use OR instead of AND.
5164 2020-01-06 Michael Meissner <meissner@linux.ibm.com>
5166 * config/rs6000/rs6000.c (hard_reg_and_mode_to_addr_mask): New
5167 helper function to return the valid addressing formats for a given
5168 hard register and mode.
5169 (rs6000_adjust_vec_address): Call hard_reg_and_mode_to_addr_mask.
5171 * config/rs6000/constraints.md (Q constraint): Update
5173 * doc/md.texi (RS/6000 constraints): Update 'Q' cosntraint
5176 * config/rs6000/vsx.md (vsx_extract_<mode>_var, VSX_D iterator):
5177 Use 'Q' for doing vector extract from memory.
5178 (vsx_extract_v4sf_var): Use 'Q' for doing vector extract from
5180 (vsx_extract_<mode>_var, VSX_EXTRACT_I iterator): Use 'Q' for
5181 doing vector extract from memory.
5182 (vsx_extract_<mode>_<VS_scalar>mode_var): Use 'Q' for doing vector
5183 extract from memory.
5185 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add support
5186 for the offset being 34-bits when -mcpu=future is used.
5188 2020-01-06 John David Anglin <danglin@gcc.gnu.org>
5190 * config/pa/pa.md: Revert change to use ordered_comparison_operator
5191 instead of cmpib_comparison_operator in cmpib patterns.
5192 * config/pa/predicates.md (cmpib_comparison_operator): Revert removal
5193 of cmpib_comparison_operator. Revise comment.
5195 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
5197 * tree-vect-slp.c (vect_build_slp_tree_1): Require all shifts
5198 in an IFN_DIV_POW2 node to be equal.
5200 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
5202 * tree-vect-stmts.c (vect_check_load_store_mask): Rename to...
5203 (vect_check_scalar_mask): ...this.
5204 (vectorizable_store, vectorizable_load): Update call accordingly.
5205 (vectorizable_call): Use vect_check_scalar_mask to check the mask
5206 argument in calls to conditional internal functions.
5208 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
5210 * config/gcn/gcn-valu.md (subv64di3): Use separate alternatives for
5211 '0' matching inputs.
5212 (subv64di3_exec): Likewise.
5214 2020-01-06 Bryan Stenson <bryan@siliconvortex.com>
5216 * config/mips/mips.c (vr4130_align_insns): Fix typo.
5217 * doc/md.texi (movstr): Likewise.
5219 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
5221 * config/gcn/gcn-valu.md (vec_extract<mode><scalar_mode>): Add early
5224 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
5226 * config/aarch64/t-aarch64 ($(srcdir)/config/aarch64/aarch64-tune.md):
5228 (s-aarch64-tune-md): ...this new stamp file. Pipe the new contents
5229 to a temporary file and use move-if-change to update the real
5230 file where necessary.
5232 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
5234 * config/aarch64/aarch64-sve.md (@aarch64_sel_dup<mode>): Use Upl
5235 rather than Upa for CPY /M.
5237 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
5239 * config/gcn/gcn.c (gcn_inline_constant_p): Allow 64 as an inline
5242 2020-01-06 Martin Liska <mliska@suse.cz>
5244 PR tree-optimization/92860
5245 * params.opt: Mark param_max_combine_insns with Optimization
5248 2020-01-05 Jakub Jelinek <jakub@redhat.com>
5251 * config/i386/i386.md (SWIDWI): New mode iterator.
5252 (DWI, dwi): Add TImode variants.
5253 (addv<mode>4): Use SWIDWI iterator instead of SWI. Use
5254 <general_hilo_operand> instead of <general_operand>. Use
5255 CONST_SCALAR_INT_P instead of CONST_INT_P.
5256 (*addv<mode>4_1): Rename to ...
5257 (addv<mode>4_1): ... this.
5258 (QWI): New mode attribute.
5259 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
5260 define_insn_and_split patterns.
5261 (*addv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
5263 (uaddv<mode>4): Use SWIDWI iterator instead of SWI. Use
5264 <general_hilo_operand> instead of <general_operand>.
5265 (*addcarry<mode>_1): New define_insn.
5266 (*add<dwi>3_doubleword_cc_overflow_1): New define_insn_and_split.
5268 2020-01-03 Konstantin Kharlamov <Hi-Angel@yandex.ru>
5270 * gdbinit.in (pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, pdd, pbs, pbm):
5271 Use "call" instead of "set".
5273 2020-01-03 Martin Jambor <mjambor@suse.cz>
5276 * ipa-cp.c (print_all_lattices): Skip functions without info.
5278 2020-01-03 Jakub Jelinek <jakub@redhat.com>
5281 * config/i386/i386-options.c (ix86_simd_clone_adjust): If
5282 TARGET_PREFER_AVX128, use prefer-vector-width=256 for 'c' and 'd'
5283 simd clones. If TARGET_PREFER_AVX256, use prefer-vector-width=512
5284 for 'e' simd clones.
5287 * config/i386/i386.opt (x_prefer_vector_width_type): Remove TargetSave
5289 (mprefer-vector-width=): Add Save.
5290 * config/i386/i386-options.c (ix86_target_string): Add PVW argument, print
5291 -mprefer-vector-width= if non-zero. Fix up -mfpmath= comment.
5292 (ix86_debug_options, ix86_function_specific_print): Adjust
5293 ix86_target_string callers.
5294 (ix86_valid_target_attribute_inner_p): Handle prefer-vector-width=.
5295 (ix86_valid_target_attribute_tree): Likewise.
5296 * config/i386/i386-options.h (ix86_target_string): Add PVW argument.
5297 * config/i386/i386-expand.c (ix86_expand_builtin): Adjust
5298 ix86_target_string caller.
5301 * config/i386/i386.md (abs<mode>2): Use expand_simple_binop instead of
5302 emitting ASHIFTRT, XOR and MINUS by hand. Use gen_int_mode with QImode
5303 instead of gen_int_shift_amount + convert_modes.
5305 PR rtl-optimization/93088
5306 * loop-iv.c (find_single_def_src): Punt after looking through
5307 128 reg copies for regs with single definitions. Move definitions
5310 2020-01-02 Dennis Zhang <dennis.zhang@arm.com>
5312 * config/arm/arm-c.c (arm_cpu_builtins): Define
5313 __ARM_FEATURE_MATMUL_INT8, __ARM_FEATURE_BF16_VECTOR_ARITHMETIC,
5314 __ARM_FEATURE_BF16_SCALAR_ARITHMETIC, and
5315 __ARM_BF16_FORMAT_ALTERNATIVE when enabled.
5316 * config/arm/arm-cpus.in (armv8_6, i8mm, bf16): New features.
5317 * config/arm/arm-tables.opt: Regenerated.
5318 * config/arm/arm.c (arm_option_reconfigure_globals): Initialize
5319 arm_arch_i8mm and arm_arch_bf16 when enabled.
5320 * config/arm/arm.h (TARGET_I8MM): New macro.
5321 (TARGET_BF16_FP, TARGET_BF16_SIMD): Likewise.
5322 * config/arm/t-aprofile: Add matching rules for -march=armv8.6-a.
5323 * config/arm/t-arm-elf (all_v8_archs): Add armv8.6-a.
5324 * config/arm/t-multilib: Add matching rules for -march=armv8.6-a.
5325 (v8_6_a_simd_variants): New.
5326 (v8_*_a_simd_variants): Add i8mm and bf16.
5327 * doc/invoke.texi (armv8.6-a, i8mm, bf16): Document new options.
5329 2020-01-02 Jakub Jelinek <jakub@redhat.com>
5332 * predict.c (compute_function_frequency): Don't call
5333 warn_function_cold on functions that already have cold attribute.
5335 2020-01-01 John David Anglin <danglin@gcc.gnu.org>
5338 * config/pa/pa.c (pa_elf_select_rtx_section): New. Put references to
5339 COMDAT group function labels in .data.rel.ro.local section.
5340 * config/pa/pa32-linux.h (TARGET_ASM_SELECT_RTX_SECTION): Define.
5343 * config/pa/pa.md (scc): Use ordered_comparison_operator instead of
5344 comparison_operator in B and S integer comparisons. Likewise, use
5345 ordered_comparison_operator instead of cmpib_comparison_operator in
5347 * config/pa/predicates.md (cmpib_comparison_operator): Remove.
5349 2020-01-01 Jakub Jelinek <jakub@redhat.com>
5351 Update copyright years.
5353 * gcc.c (process_command): Update copyright notice dates.
5354 * gcov-dump.c (print_version): Ditto.
5355 * gcov.c (print_version): Ditto.
5356 * gcov-tool.c (print_version): Ditto.
5357 * gengtype.c (create_file): Ditto.
5358 * doc/cpp.texi: Bump @copying's copyright year.
5359 * doc/cppinternals.texi: Ditto.
5360 * doc/gcc.texi: Ditto.
5361 * doc/gccint.texi: Ditto.
5362 * doc/gcov.texi: Ditto.
5363 * doc/install.texi: Ditto.
5364 * doc/invoke.texi: Ditto.
5366 2020-01-01 Jan Hubicka <hubicka@ucw.cz>
5368 * ipa.c (walk_polymorphic_call_targets): Fix updating of overall
5371 2020-01-01 Jakub Jelinek <jakub@redhat.com>
5373 PR tree-optimization/93098
5374 * match.pd (popcount): For shift amounts, use integer_onep
5375 or wi::to_widest () == cst instead of tree_to_uhwi () == cst
5376 tests. Make sure that precision is power of two larger than or equal
5377 to 16. Ensure shift is never negative. Use HOST_WIDE_INT_UC macro
5378 instead of ULL suffixed constants. Formatting fixes.
5380 Copyright (C) 2020 Free Software Foundation, Inc.
5382 Copying and distribution of this file, with or without modification,
5383 are permitted in any medium without royalty provided the copyright
5384 notice and this notice are preserved.