1 2020-03-06 Wilco Dijkstra <wdijkstr@arm.com>
4 * config/aarch64/aarch64-builtins.c (TYPES_TERNOPU_LANE): Add define.
5 * config/aarch64/aarch64-simd.md
6 (aarch64_vec_<su>mult_lane<Qlane>): Add new insn for widening lane mul.
7 (aarch64_vec_<su>mlal_lane<Qlane>): Likewise.
8 * config/aarch64/aarch64-simd-builtins.def: Add intrinsics.
9 * config/aarch64/arm_neon.h:
10 (vmlal_lane_s16): Expand using intrinsics rather than inline asm.
11 (vmlal_lane_u16): Likewise.
12 (vmlal_lane_s32): Likewise.
13 (vmlal_lane_u32): Likewise.
14 (vmlal_laneq_s16): Likewise.
15 (vmlal_laneq_u16): Likewise.
16 (vmlal_laneq_s32): Likewise.
17 (vmlal_laneq_u32): Likewise.
18 (vmull_lane_s16): Likewise.
19 (vmull_lane_u16): Likewise.
20 (vmull_lane_s32): Likewise.
21 (vmull_lane_u32): Likewise.
22 (vmull_laneq_s16): Likewise.
23 (vmull_laneq_u16): Likewise.
24 (vmull_laneq_s32): Likewise.
25 (vmull_laneq_u32): Likewise.
26 * config/aarch64/iterators.md (Vcondtype): New iterator for lane mul.
29 2020-03-06 Wilco Dijkstra <wdijkstr@arm.com>
31 * aarch64/aarch64-simd.md (aarch64_mla_elt<mode>): Correct lane syntax.
32 (aarch64_mla_elt_<vswap_width_name><mode>): Likewise.
33 (aarch64_mls_elt<mode>): Likewise.
34 (aarch64_mls_elt_<vswap_width_name><mode>): Likewise.
35 (aarch64_fma4_elt<mode>): Likewise.
36 (aarch64_fma4_elt_<vswap_width_name><mode>): Likewise.
37 (aarch64_fma4_elt_to_64v2df): Likewise.
38 (aarch64_fnma4_elt<mode>): Likewise.
39 (aarch64_fnma4_elt_<vswap_width_name><mode>): Likewise.
40 (aarch64_fnma4_elt_to_64v2df): Likewise.
42 2020-03-06 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
44 * config/aarch64/aarch64-sve2.md (@aarch64_sve_<sve_int_op><mode>:
45 Specify movprfx attribute.
46 (@aarch64_sve_<sve_int_op>_lane_<mode>): Likewise.
48 2020-03-06 David Edelsohn <dje.gcc@gmail.com>
51 * config/rs6000/aix61.h (TARGET_NO_SUM_IN_TOC): Set to 1 for
53 (TARGET_NO_FP_IN_TOC): Same.
54 * config/rs6000/aix71.h: Same.
55 * config/rs6000/aix72.h: Same.
57 2020-03-06 Andrew Pinski <apinski@marvell.com>
58 Jeff Law <law@redhat.com>
60 PR rtl-optimization/93996
61 * haifa-sched.c (remove_notes): Be more careful when adding
64 2020-03-06 Delia Burduv <delia.burduv@arm.com>
66 * config/arm/arm_neon.h (vld2_bf16): New.
73 (vld2q_dup_bf16): New.
75 (vld3q_dup_bf16): New.
77 (vld4q_dup_bf16): New.
78 * config/arm/arm_neon_builtins.def
79 (vld2): Changed to VAR13 and added v4bf, v8bf
80 (vld2_dup): Changed to VAR8 and added v4bf, v8bf
81 (vld3): Changed to VAR13 and added v4bf, v8bf
82 (vld3_dup): Changed to VAR8 and added v4bf, v8bf
83 (vld4): Changed to VAR13 and added v4bf, v8bf
84 (vld4_dup): Changed to VAR8 and added v4bf, v8bf
85 * config/arm/iterators.md (VDXBF2): New iterator.
86 *config/arm/neon.md (neon_vld2): Use new iterators.
87 (neon_vld2_dup<mode): Use new iterators.
88 (neon_vld3<mode>): Likewise.
89 (neon_vld3qa<mode>): Likewise.
90 (neon_vld3qb<mode>): Likewise.
91 (neon_vld3_dup<mode>): Likewise.
92 (neon_vld4<mode>): Likewise.
93 (neon_vld4qa<mode>): Likewise.
94 (neon_vld4qb<mode>): Likewise.
95 (neon_vld4_dup<mode>): Likewise.
96 (neon_vld2_dupv8bf): New.
97 (neon_vld3_dupv8bf): Likewise.
98 (neon_vld4_dupv8bf): Likewise.
100 2020-03-06 Delia Burduv <delia.burduv@arm.com>
102 * config/arm/arm_neon.h (bfloat16x4x2_t): New typedef.
103 (bfloat16x8x2_t): New typedef.
104 (bfloat16x4x3_t): New typedef.
105 (bfloat16x8x3_t): New typedef.
106 (bfloat16x4x4_t): New typedef.
107 (bfloat16x8x4_t): New typedef.
114 * config/arm/arm-builtins.c (v2bf_UP): Define.
116 (arm_init_simd_builtin_types): Init Bfloat16x2_t eltype.
117 * config/arm/arm-modes.def (V2BF): New mode.
118 * config/arm/arm-simd-builtin-types.def
119 (Bfloat16x2_t): New entry.
120 * config/arm/arm_neon_builtins.def
121 (vst2): Changed to VAR13 and added v4bf, v8bf
122 (vst3): Changed to VAR13 and added v4bf, v8bf
123 (vst4): Changed to VAR13 and added v4bf, v8bf
124 * config/arm/iterators.md (VDXBF): New iterator.
125 (VQ2BF): New iterator.
126 *config/arm/neon.md (neon_vst2<mode>): Used new iterators.
127 (neon_vst2<mode>): Used new iterators.
128 (neon_vst3<mode>): Used new iterators.
129 (neon_vst3<mode>): Used new iterators.
130 (neon_vst3qa<mode>): Used new iterators.
131 (neon_vst3qb<mode>): Used new iterators.
132 (neon_vst4<mode>): Used new iterators.
133 (neon_vst4<mode>): Used new iterators.
134 (neon_vst4qa<mode>): Used new iterators.
135 (neon_vst4qb<mode>): Used new iterators.
137 2020-03-06 Delia Burduv <delia.burduv@arm.com>
139 * config/aarch64/aarch64-simd-builtins.def
140 (bfcvtn): New built-in function.
141 (bfcvtn_q): New built-in function.
142 (bfcvtn2): New built-in function.
143 (bfcvt): New built-in function.
144 * config/aarch64/aarch64-simd.md
145 (aarch64_bfcvtn<q><mode>): New pattern.
146 (aarch64_bfcvtn2v8bf): New pattern.
147 (aarch64_bfcvtbf): New pattern.
148 * config/aarch64/arm_bf16.h (float32_t): New typedef.
149 (vcvth_bf16_f32): New intrinsic.
150 * config/aarch64/arm_bf16.h (vcvt_bf16_f32): New intrinsic.
151 (vcvtq_low_bf16_f32): New intrinsic.
152 (vcvtq_high_bf16_f32): New intrinsic.
153 * config/aarch64/iterators.md (V4SF_TO_BF): New mode iterator.
154 (UNSPEC_BFCVTN): New UNSPEC.
155 (UNSPEC_BFCVTN2): New UNSPEC.
156 (UNSPEC_BFCVT): New UNSPEC.
157 * config/arm/types.md (bf_cvt): New type.
159 2020-03-06 Andreas Krebbel <krebbel@linux.ibm.com>
161 * config/s390/s390.md ("tabort"): Get rid of two consecutive
162 blanks in format string.
164 2020-03-05 H.J. Lu <hongjiu.lu@intel.com>
168 * config/i386/i386-protos.h (ix86_output_ssemov): New prototype.
169 * config/i386/i386.c (ix86_get_ssemov): New function.
170 (ix86_output_ssemov): Likewise.
171 * config/i386/sse.md (VMOVE:mov<mode>_internal): Call
172 ix86_output_ssemov for TYPE_SSEMOV. Remove TARGET_AVX512VL
174 (*movxi_internal_avx512f): Call ix86_output_ssemov for TYPE_SSEMOV.
175 (*movoi_internal_avx): Call ix86_output_ssemov for TYPE_SSEMOV.
176 Remove ext_sse_reg_operand and TARGET_AVX512VL check.
177 (*movti_internal): Likewise.
178 (*movtf_internal): Call ix86_output_ssemov for TYPE_SSEMOV.
180 2020-03-05 Jeff Law <law@redhat.com>
182 PR tree-optimization/91890
183 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Remove LOC argument.
184 Use gimple_or_expr_nonartificial_location.
185 (check_bounds_overlap): Drop LOC argument to maybe_diag_access_bounds.
186 Use gimple_or_expr_nonartificial_location.
187 * gimple.c (gimple_or_expr_nonartificial_location): New function.
188 * gimple.h (gimple_or_expr_nonartificial_location): Declare it.
189 * tree-ssa-strlen.c (maybe_warn_overflow): Use
190 gimple_or_expr_nonartificial_location.
191 (maybe_diag_stxncpy_trunc, handle_builtin_stxncpy_strncat): Likewise.
192 (maybe_warn_pointless_strcmp): Likewise.
194 2020-03-05 Jakub Jelinek <jakub@redhat.com>
197 * config/i386/avx2intrin.h (_mm_mask_i32gather_ps): Fix first cast of
198 SRC and MASK arguments to __m128 from __m128d.
199 (_mm256_mask_i32gather_ps): Fix first cast of MASK argument to __m256
201 (_mm_mask_i64gather_ps): Fix first cast of MASK argument to __m128
203 * config/i386/xopintrin.h (_mm_permute2_pd): Fix first cast of C
204 argument to __m128i from __m128d.
205 (_mm256_permute2_pd): Fix first cast of C argument to __m256i from
207 (_mm_permute2_ps): Fix first cast of C argument to __m128i from __m128.
208 (_mm256_permute2_ps): Fix first cast of C argument to __m256i from
211 2020-03-05 Delia Burduv <delia.burduv@arm.com>
213 * config/arm/arm_neon.h (vbfmmlaq_f32): New.
214 (vbfmlalbq_f32): New.
215 (vbfmlaltq_f32): New.
216 (vbfmlalbq_lane_f32): New.
217 (vbfmlaltq_lane_f32): New.
218 (vbfmlalbq_laneq_f32): New.
219 (vbfmlaltq_laneq_f32): New.
220 * config/arm/arm_neon_builtins.def (vmmla): New.
227 * config/arm/iterators.md (BF_MA): New int iterator.
228 (bt): New int attribute.
229 (VQXBF): Copy of VQX with V8BF.
230 * config/arm/neon.md (neon_vmmlav8bf): New insn.
231 (neon_vfma<bt>v8bf): New insn.
232 (neon_vfma<bt>_lanev8bf): New insn.
233 (neon_vfma<bt>_laneqv8bf): New expand.
234 (neon_vget_high<mode>): Changed iterator to VQXBF.
235 * config/arm/unspecs.md (UNSPEC_BFMMLA): New UNSPEC.
236 (UNSPEC_BFMAB): New UNSPEC.
237 (UNSPEC_BFMAT): New UNSPEC.
239 2020-03-05 Jakub Jelinek <jakub@redhat.com>
242 * tree-pretty-print.h (pretty_print_string): Declare.
243 * tree-pretty-print.c (pretty_print_string): Remove forward
244 declaration, no longer static. Change nbytes parameter type
245 from unsigned to size_t.
246 * print-rtl.c (print_value) <case CONST_STRING>: Use
247 pretty_print_string and for shrink way too long strings.
249 2020-03-05 Richard Biener <rguenther@suse.de>
250 Jakub Jelinek <jakub@redhat.com>
252 PR tree-optimization/93582
253 * tree-ssa-sccvn.c (vn_reference_lookup_3): Treat POINTER_PLUS_EXPR
254 last operand as signed when looking for memset offset. Formatting
257 2020-03-04 Andrew Pinski <apinski@marvell.com>
260 * value-prof.c (dump_histogram_value): Use std::abs.
262 2020-03-04 Martin Sebor <msebor@redhat.com>
264 PR tree-optimization/93986
265 * tree-ssa-strlen.c (maybe_warn_overflow): Convert all wide_int
266 operands to the same precision widest_int to avoid ICEs.
268 2020-03-04 Bill Schmidt <wschmidt@linux.ibm.com>
271 * rs6000-cpus.def (OTHER_ALTIVEC_MASKS): New #define.
272 * rs6000.c (rs6000_disable_incompatible_switches): Add table entry
273 for OPTION_MASK_ALTIVEC.
275 2020-03-04 Andreas Krebbel <krebbel@linux.ibm.com>
277 * config.gcc: Include the glibc-stdint.h header for zTPF.
279 2020-03-04 Andreas Krebbel <krebbel@linux.ibm.com>
281 * config/s390/s390.c (s390_secondary_memory_needed): Disallow
282 direct FPR-GPR copies.
283 (s390_register_info_gprtofpr): Disallow GPR content to be saved in
286 2020-03-04 Andreas Krebbel <krebbel@linux.ibm.com>
288 * config/s390/s390.c (s390_emit_prologue): Specify the 2 new
289 operands to the prologue_tpf expander.
290 (s390_emit_epilogue): Likewise.
291 (s390_option_override_internal): Do error checking and setup for
293 * config/s390/tpf.h (TPF_TRACE_PROLOGUE_CHECK)
294 (TPF_TRACE_EPILOGUE_CHECK, TPF_TRACE_PROLOGUE_TARGET)
295 (TPF_TRACE_EPILOGUE_TARGET, TPF_TRACE_PROLOGUE_SKIP_TARGET)
296 (TPF_TRACE_EPILOGUE_SKIP_TARGET): New macro definitions.
297 * config/s390/tpf.md ("prologue_tpf", "epilogue_tpf"): Add two new
298 operands for the check flag and the branch target.
299 * config/s390/tpf.opt ("mtpf-trace-hook-prologue-check")
300 ("mtpf-trace-hook-prologue-target")
301 ("mtpf-trace-hook-epilogue-check")
302 ("mtpf-trace-hook-epilogue-target", "mtpf-trace-skip"): New
304 * doc/invoke.texi: Document -mtpf-trace-skip option. The other
305 options are for debugging purposes and will not be documented
308 2020-03-04 Jakub Jelinek <jakub@redhat.com>
311 * tree-inline.c (copy_decl_to_var): Copy DECL_BY_REFERENCE flag.
313 * tree-ssa-sccvn.c (vn_walk_cb_data::push_partial_def): Add offseti
314 argument. Change pd argument so that it can be modified. Turn
315 constant non-CONSTRUCTOR store into non-constant if it is too large.
316 Adjust offset and size of CONSTRUCTOR or non-constant store to avoid
318 (vn_walk_cb_data::vn_walk_cb_data, vn_reference_lookup_3): Adjust
321 2020-02-04 Richard Biener <rguenther@suse.de>
323 PR tree-optimization/93964
324 * graphite-isl-ast-to-gimple.c
325 (gcc_expression_from_isl_ast_expr_id): Add intermediate
326 conversion for pointer to integer converts.
327 * graphite-scop-detection.c (assign_parameter_index_in_region):
330 2020-03-04 Martin Liska <mliska@suse.cz>
334 * doc/invoke.texi: Clarify --help=language and --help=common
337 2020-03-04 Jakub Jelinek <jakub@redhat.com>
339 PR tree-optimization/94001
340 * tree-tailcall.c (process_assignment): Before comparing op1 to
341 *ass_var, verify *ass_var is non-NULL.
343 2020-03-04 Kito Cheng <kito.cheng@sifive.com>
346 * config/riscv/riscv.c (riscv_emit_float_compare): Using NE to compare
349 2020-03-03 Dennis Zhang <dennis.zhang@arm.com>
351 * config/arm/arm_bf16.h (vcvtah_f32_bf16, vcvth_bf16_f32): New.
352 * config/arm/arm_neon.h (vcvt_f32_bf16, vcvtq_low_f32_bf16): New.
353 (vcvtq_high_f32_bf16, vcvt_bf16_f32): New.
354 (vcvtq_low_bf16_f32, vcvtq_high_bf16_f32): New.
355 * config/arm/arm_neon_builtins.def (vbfcvt, vbfcvt_high): New entries.
356 (vbfcvtv4sf, vbfcvtv4sf_high): Likewise.
357 * config/arm/iterators.md (VBFCVT, VBFCVTM): New mode iterators.
358 (V_bf_low, V_bf_cvt_m): New mode attributes.
359 * config/arm/neon.md (neon_vbfcvtv4sf<VBFCVT:mode>): New.
360 (neon_vbfcvtv4sf_highv8bf, neon_vbfcvtsf): New.
361 (neon_vbfcvt<VBFCVT:mode>, neon_vbfcvt_highv8bf): New.
362 (neon_vbfcvtbf_cvtmode<mode>, neon_vbfcvtbf): New
363 * config/arm/unspecs.md (UNSPEC_BFCVT, UNSPEC_BFCVT_HIG): New.
365 2020-03-03 Jakub Jelinek <jakub@redhat.com>
367 PR tree-optimization/93582
368 * tree-ssa-sccvn.h (vn_reference_lookup): Add mask argument.
369 * tree-ssa-sccvn.c (struct vn_walk_cb_data): Add mask and masked_result
370 members, initialize them in the constructor and if mask is non-NULL,
371 artificially push_partial_def {} for the portions of the mask that
373 (vn_walk_cb_data::finish): If mask is non-NULL, set masked_result to
374 val and return (void *)-1. Formatting fix.
375 (vn_reference_lookup_pieces): Adjust vn_walk_cb_data initialization.
377 (vn_reference_lookup): Add mask argument. If non-NULL, don't call
378 fully_constant_vn_reference_p nor vn_reference_lookup_1 and return
380 (visit_nary_op): Handle BIT_AND_EXPR of a memory load and INTEGER_CST
382 (visit_stmt): Formatting fix.
384 2020-03-03 Richard Biener <rguenther@suse.de>
386 PR tree-optimization/93946
387 * alias.h (refs_same_for_tbaa_p): Declare.
388 * alias.c (refs_same_for_tbaa_p): New function.
389 * tree-ssa-alias.c (ao_ref_alias_set): For a NULL ref return
391 * tree-ssa-scopedtables.h
392 (avail_exprs_stack::lookup_avail_expr): Add output argument
393 giving access to the hashtable entry.
394 * tree-ssa-scopedtables.c (avail_exprs_stack::lookup_avail_expr):
396 * tree-ssa-dom.c: Include alias.h.
397 (dom_opt_dom_walker::optimize_stmt): Validate TBAA state before
398 removing redundant store.
399 * tree-ssa-sccvn.h (vn_reference_s::base_set): New member.
400 (ao_ref_init_from_vn_reference): Adjust prototype.
401 (vn_reference_lookup_pieces): Likewise.
402 (vn_reference_insert_pieces): Likewise.
403 * tree-ssa-sccvn.c: Track base alias set in addition to alias
405 (eliminate_dom_walker::eliminate_stmt): Also check base alias
406 set when removing redundant stores.
407 (visit_reference_op_store): Likewise.
408 * dse.c (record_store): Adjust valdity check for redundant
411 2020-03-03 Jakub Jelinek <jakub@redhat.com>
414 * config/s390/s390.h (OPTION_DEFAULT_SPECS): Reorder.
416 PR rtl-optimization/94002
417 * explow.c (plus_constant): Punt if cst has VOIDmode and
418 get_pool_mode is different from mode.
420 2020-03-03 Claudiu Zissulescu <claziss@synopsys.com>
422 * config/arc/arc.c (leigitimate_small_data_address_p): Check if an
423 address has an offset which fits the scalling constraint for a
424 load/store operation.
425 (legitimate_scaled_address_p): Update use
426 leigitimate_small_data_address_p.
427 (arc_print_operand): Likewise.
428 (arc_legitimate_address_p): Likewise.
429 (legitimate_small_data_address_p): Likewise.
431 2020-03-03 Claudiu Zissulescu <claziss@synopsys.com>
433 * config/arc/arc.md (fmasf4_fpu): Use accl_operand predicate.
434 (fnmasf4_fpu): Likewise.
436 2020-03-03 Claudiu Zissulescu <claziss@synopsys.com>
438 * config/arc/arc.md (adddi3): Early expand the 64bit operation into
441 (adddi3_i): Remove pattern.
442 (subdi3_i): Likewise.
444 2020-03-03 Claudiu Zissulescu <claziss@synopsys.com>
446 * config/arc/arc.md (eh_return): Add length info.
448 2020-03-02 David Malcolm <dmalcolm@redhat.com>
450 * doc/invoke.texi (-fanalyzer-show-duplicate-count): New.
452 2020-03-02 David Malcolm <dmalcolm@redhat.com>
454 * doc/invoke.texi (Static Analyzer Options): Add
455 -Wanalyzer-stale-setjmp-buffer to the list of options enabled
458 2020-03-02 Uroš Bizjak <ubizjak@gmail.com>
461 * config/i386/i386.md (movstrict<mode>): Allow only
462 registers with VALID_INT_MODE_P modes.
464 2020-03-02 Andrew Stubbs <ams@codesourcery.com>
466 * config/gcn/gcn-valu.md (dpp_move<mode>): New.
467 (reduc_insn): Use 'U' and 'B' operand codes.
468 (reduc_<reduc_op>_scal_<mode>): Allow all types.
469 (reduc_<reduc_op>_scal_v64di): Delete.
470 (*<reduc_op>_dpp_shr_<mode>): Allow all 1reg types.
471 (*plus_carry_dpp_shr_v64si): Change to ...
472 (*plus_carry_dpp_shr_<mode>): ... this and allow all 1reg int types.
473 (mov_from_lane63_v64di): Change to ...
474 (mov_from_lane63_<mode>): ... this, and allow all 64-bit modes.
475 * config/gcn/gcn.c (gcn_expand_dpp_shr_insn): Increase buffer size.
476 Support UNSPEC_MOV_DPP_SHR output formats.
477 (gcn_expand_reduc_scalar): Add "use_moves" reductions.
478 Add "use_extends" reductions.
479 (print_operand_address): Add 'I' and 'U' codes.
480 * config/gcn/gcn.md (unspec): Add UNSPEC_MOV_DPP_SHR.
482 2020-03-02 Martin Liska <mliska@suse.cz>
484 * lto-wrapper.c: Fix typo in comment about
485 C++ standard version.
487 2020-03-01 Martin Sebor <msebor@redhat.com>
490 * calls.c (init_attr_rdwr_indices): Correctly handle attribute.
492 2020-03-01 Martin Sebor <msebor@redhat.com>
495 * tree-ssa-strlen.c (count_nonzero_bytes): Set the size to that
496 of a pointer in the outermost ADDR_EXPRs.
498 2020-02-28 Jeff Law <law@redhat.com>
500 * config/v850/v850.h (STATIC_CHAIN_REGNUM): Change to r19.
501 * config/v850/v850.c (v850_asm_trampoline_template): Update
504 2020-02-28 Michael Meissner <meissner@linux.ibm.com>
507 * config/rs6000/vsx.md (vsx_extract_<mode>_<VS_scalar>mode_var):
510 2020-02-28 Martin Liska <mliska@suse.cz>
513 * configure.ac: Improve detection of ld_date by requiring
514 either two dashes or none.
515 * configure: Regenerate.
517 2020-02-28 Vladimir Makarov <vmakarov@redhat.com>
519 PR rtl-optimization/93564
520 * ira-color.c (assign_hard_reg): Prefer smaller hard regno when we
521 do not honor reg alloc order.
523 2020-02-27 Joel Hutton <Joel.Hutton@arm.com>
526 * config/aarch64/aarch64.c (aarch64_override_options): Fix
527 misleading warning string.
529 2020-02-27 Martin Sebor <msebor@redhat.com>
531 * doc/invoke.texi (-Wbuiltin-declaration-mismatch): Fix a typo.
533 2020-02-27 Michael Meissner <meissner@linux.ibm.com>
536 * config/rs6000/vsx.md (vsx_extract_<mode>_var, VSX_D iterator):
537 Split the insn into two parts. This insn only does variable
538 extract from a register.
539 (vsx_extract_<mode>_var_load, VSX_D iterator): New insn, do
540 variable extract from memory.
541 (vsx_extract_v4sf_var): Split the insn into two parts. This insn
542 only does variable extract from a register.
543 (vsx_extract_v4sf_var_load): New insn, do variable extract from
545 (vsx_extract_<mode>_var, VSX_EXTRACT_I iterator): Split the insn
546 into two parts. This insn only does variable extract from a
548 (vsx_extract_<mode>_var_load, VSX_EXTRACT_I iterator): New insn,
549 do variable extract from memory.
551 2020-02-27 Martin Jambor <mjambor@suse.cz>
552 Feng Xue <fxue@os.amperecomputing.com>
555 * ipa-cp.c (same_node_or_its_all_contexts_clone_p): Replaced with
556 new function calls_same_node_or_its_all_contexts_clone_p.
557 (cgraph_edge_brings_value_p): Use it.
558 (cgraph_edge_brings_value_p): Likewise.
559 (self_recursive_pass_through_p): Return false if caller is a clone.
560 (self_recursive_agg_pass_through_p): Likewise.
562 2020-02-27 Jan Hubicka <hubicka@ucw.cz>
565 * alias.c (ends_tbaa_access_path_p): Break out from ...
566 (component_uses_parent_alias_set_from): ... here.
567 * alias.h (ends_tbaa_access_path_p): Declare.
568 * tree-ssa-alias.c (access_path_may_continue_p): Break out from ...;
569 handle trailing arrays past end of tbaa access path.
570 (aliasing_component_refs_p): ... here; likewise.
571 (nonoverlapping_refs_since_match_p): Track TBAA segment of the access
572 path; disambiguate also past end of it.
573 (nonoverlapping_component_refs_p): Use only TBAA segment of the access
576 2020-02-27 Mihail Ionescu <mihail.ionescu@arm.com>
578 * (__ARM_NUM_LANES, __arm_lane, __arm_lane_q): Move to the
579 beginning of the file.
580 (vcreate_bf16, vcombine_bf16): New.
581 (vdup_n_bf16, vdupq_n_bf16): New.
582 (vdup_lane_bf16, vdup_laneq_bf16): New.
583 (vdupq_lane_bf16, vdupq_laneq_bf16): New.
584 (vduph_lane_bf16, vduph_laneq_bf16): New.
585 (vset_lane_bf16, vsetq_lane_bf16): New.
586 (vget_lane_bf16, vgetq_lane_bf16): New.
587 (vget_high_bf16, vget_low_bf16): New.
588 (vreinterpret_bf16_u8, vreinterpretq_bf16_u8): New.
589 (vreinterpret_bf16_u16, vreinterpretq_bf16_u16): New.
590 (vreinterpret_bf16_u32, vreinterpretq_bf16_u32): New.
591 (vreinterpret_bf16_u64, vreinterpretq_bf16_u64): New.
592 (vreinterpret_bf16_s8, vreinterpretq_bf16_s8): New.
593 (vreinterpret_bf16_s16, vreinterpretq_bf16_s16): New.
594 (vreinterpret_bf16_s32, vreinterpretq_bf16_s32): New.
595 (vreinterpret_bf16_s64, vreinterpretq_bf16_s64): New.
596 (vreinterpret_bf16_p8, vreinterpretq_bf16_p8): New.
597 (vreinterpret_bf16_p16, vreinterpretq_bf16_p16): New.
598 (vreinterpret_bf16_p64, vreinterpretq_bf16_p64): New.
599 (vreinterpret_bf16_f32, vreinterpretq_bf16_f32): New.
600 (vreinterpret_bf16_f64, vreinterpretq_bf16_f64): New.
601 (vreinterpretq_bf16_p128): New.
602 (vreinterpret_s8_bf16, vreinterpretq_s8_bf16): New.
603 (vreinterpret_s16_bf16, vreinterpretq_s16_bf16): New.
604 (vreinterpret_s32_bf16, vreinterpretq_s32_bf16): New.
605 (vreinterpret_s64_bf16, vreinterpretq_s64_bf16): New.
606 (vreinterpret_u8_bf16, vreinterpretq_u8_bf16): New.
607 (vreinterpret_u16_bf16, vreinterpretq_u16_bf16): New.
608 (vreinterpret_u32_bf16, vreinterpretq_u32_bf16): New.
609 (vreinterpret_u64_bf16, vreinterpretq_u64_bf16): New.
610 (vreinterpret_p8_bf16, vreinterpretq_p8_bf16): New.
611 (vreinterpret_p16_bf16, vreinterpretq_p16_bf16): New.
612 (vreinterpret_p64_bf16, vreinterpretq_p64_bf16): New.
613 (vreinterpret_f32_bf16, vreinterpretq_f32_bf16): New.
614 (vreinterpretq_p128_bf16): New.
615 * config/arm/arm_neon_builtins.def (VDX): Add V4BF.
617 (V_elem_l): Likewise.
620 (V_DOUBLE): Likewise.
621 (VDQX): Add V4BF and V8BF.
622 (V_two_elem, V_three_elem, V_four_elem): Likewise.
625 (V_double_vector_mode): Likewise.
626 (V_cmp_result): Likewise.
627 (V_uf_sclr): Likewise.
628 (V_sz_elem): Likewise.
629 (Is_d_reg): Likewise.
630 (V_mode_nunits): Likewise.
631 * config/arm/neon.md (neon_vdup_lane): Enable for BFloat16.
633 2020-02-27 Andrew Stubbs <ams@codesourcery.com>
635 * config/gcn/gcn-valu.md (VEC_SUBDWORD_MODE): New mode iterator.
636 (<expander><mode>2<exec>): Change modes to VEC_ALL1REG_INT_MODE.
637 (<expander><mode>3<exec>): Likewise.
638 (<expander><mode>3): New.
639 (v<expander><mode>3): New.
640 (<expander><mode>3): New.
641 (<expander><mode>3<exec>): Rename to ...
642 (<expander>v64si3<exec>): ... this, and change modes to V64SI.
643 * config/gcn/gcn.md (mnemonic): Use '%B' for not.
645 2020-02-27 Alexandre Oliva <oliva@adacore.com>
647 * config/vx-common.h (NO_DOLLAR_IN_LABEL, NO_DOT_IN_LABEL): Leave
650 2020-02-27 Richard Biener <rguenther@suse.de>
652 PR tree-optimization/93508
653 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle _CHK like
654 non-_CHK variants. Valueize their length arguments.
656 2020-02-27 Richard Biener <rguenther@suse.de>
658 PR tree-optimization/93953
659 * tree-vect-slp.c (slp_copy_subtree): Avoid keeping a reference
660 to the hash-map entry.
662 2020-02-27 Andrew Stubbs <ams@codesourcery.com>
664 * config/gcn/gcn.md (mov<mode>): Add transformations for BI subregs.
666 2020-02-27 Mark Williams <mwilliams@fb.com>
668 * dwarf2out.c (file_name_acquire): Call remap_debug_filename.
669 * lto-opts.c (lto_write_options): Drop -fdebug-prefix-map,
670 -ffile-prefix-map and -fmacro-prefix-map.
671 * lto-streamer-out.c: Include file-prefix-map.h.
672 (lto_output_location): Remap the file part of locations.
674 2020-02-27 Jakub Jelinek <jakub@redhat.com>
677 * gimplify.c (gimplify_init_constructor): Don't promote readonly
678 DECL_REGISTER variables to TREE_STATIC.
680 PR tree-optimization/93582
681 PR tree-optimization/93945
682 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle memset with
683 non-zero INTEGER_CST second argument and ref->offset or ref->size
684 not a multiple of BITS_PER_UNIT.
686 2020-02-27 Jonathan Wakely <jwakely@redhat.com>
688 * doc/install.texi (Binaries): Update description of BullFreeware.
690 2020-02-26 Sandra Loosemore <sandra@codesourcery.com>
694 * doc/invoke.texi (Option Summary): Re-alphabetize warnings in
695 C++ Language Options, Warning Options, and Static Analyzer
696 Options lists. Document negative form of options enabled by
697 default. Move some things around to more accurately sort
698 warnings by category.
699 (C++ Dialect Options, Warning Options, Static Analyzer
700 Options): Document negative form of options when enabled by
701 default. Move some things around to more accurately sort
702 warnings by category. Add some missing index entries.
705 2020-02-26 Carl Love <cel@us.ibm.com>
708 * doc/extend.texi (PowerPC AltiVec Built-in Functions available on
709 ISA 2.07): The builtin-function name __builtin_crypto_vpmsumb is only
710 for the vector unsigned short arguments. It is also listed as the
711 name of the built-in for arguments vector unsigned short,
712 vector unsigned int and vector unsigned long long built-ins. The
713 name of the builtins for these arguments should be:
714 __builtin_crypto_vpmsumh, __builtin_crypto_vpmsumw and
715 __builtin_crypto_vpmsumd respectively.
717 2020-02-26 Richard Biener <rguenther@suse.de>
719 * tree-vect-slp.c (vect_print_slp_tree): Also dump ref count
720 and load permutation.
722 2020-02-26 Richard Sandiford <richard.sandiford@arm.com>
725 * optabs-tree.c (supportable_convert_operation): Reject types with
728 2020-02-26 David Malcolm <dmalcolm@redhat.com>
730 * Makefile.in (ANALYZER_OBJS): Add analyzer/bar-chart.o.
732 2020-02-26 Jakub Jelinek <jakub@redhat.com>
734 PR tree-optimization/93820
735 * gimple-ssa-store-merging.c (check_no_overlap): Change RHS_CODE
736 argument to ALL_INTEGER_CST_P boolean.
737 (imm_store_chain_info::try_coalesce_bswap): Adjust caller.
738 (imm_store_chain_info::coalesce_immediate_stores): Likewise. Handle
739 adjacent INTEGER_CST store into merged_store->only_constants like
742 2020-02-25 Jakub Jelinek <jakub@redhat.com>
745 * config/sh/sh.c (expand_cbranchdi4): Fix comment typo, probablity
747 * cfghooks.c (verify_flow_info): Likewise.
748 * predict.c (combine_predictions_for_bb): Likewise.
749 * bb-reorder.c (connect_better_edge_p): Likewise. Fix comment typo,
750 sucessor -> successor.
751 (find_traces_1_round): Fix comment typo, destinarion -> destination.
752 * omp-expand.c (expand_oacc_for): Fix comment typo, sucessors ->
754 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): Fix dump
755 message typo, sucessors -> successors.
757 2020-02-25 Martin Sebor <msebor@redhat.com>
759 * doc/extend.texi (attribute access): Correct an example.
761 2020-02-25 Mihail Ionescu <mihail.ionescu@arm.com>
763 * config/aarch64/aarch64-builtins.c (aarch64_scalar_builtin_types):
765 (aarch64_init_simd_builtin_scalar_types): Register simd_bf.
767 * config/aarch64/iterators.md (VALLDIF): Enable for V4BF and V8BF.
768 (VD): Enable for V4BF.
770 (VQ): Enable for V8BF.
773 (VDBL, Vdbl): Add V4BF.
774 (V_INT_EQUIV, v_int_equiv): Add V4BF and V8BF.
775 * config/aarch64/arm_neon.h (bfloat16x4x2_t): New typedef.
776 (bfloat16x8x2_t): Likewise.
777 (bfloat16x4x3_t): Likewise.
778 (bfloat16x8x3_t): Likewise.
779 (bfloat16x4x4_t): Likewise.
780 (bfloat16x8x4_t): Likewise.
781 (vcombine_bf16): New.
782 (vld1_bf16, vld1_bf16_x2): New.
783 (vld1_bf16_x3, vld1_bf16_x4): New.
784 (vld1q_bf16, vld1q_bf16_x2): New.
785 (vld1q_bf16_x3, vld1q_bf16_x4): New.
786 (vld1_lane_bf16): New.
787 (vld1q_lane_bf16): New.
788 (vld1_dup_bf16): New.
789 (vld1q_dup_bf16): New.
792 (vld2_dup_bf16): New.
793 (vld2q_dup_bf16): New.
796 (vld3_dup_bf16): New.
797 (vld3q_dup_bf16): New.
800 (vld4_dup_bf16): New.
801 (vld4q_dup_bf16): New.
802 (vst1_bf16, vst1_bf16_x2): New.
803 (vst1_bf16_x3, vst1_bf16_x4): New.
804 (vst1q_bf16, vst1q_bf16_x2): New.
805 (vst1q_bf16_x3, vst1q_bf16_x4): New.
806 (vst1_lane_bf16): New.
807 (vst1q_lane_bf16): New.
815 2020-02-25 Mihail Ionescu <mihail.ionescu@arm.com>
817 * config/aarch64/iterators.md (VDQF_F16) Add V4BF and V8BF.
818 (VALL_F16): Likewise.
819 (VALLDI_F16): Likewise.
822 (vswap_width_name): Likewise.
823 (VSWAP_WIDTH): Likewise.
827 * config/aarch64/arm_neon.h (vset_lane_bf16, vsetq_lane_bf16): New.
828 (vget_lane_bf16, vgetq_lane_bf16): New.
830 (vdup_n_bf16, vdupq_n_bf16): New.
831 (vdup_lane_bf16, vdup_laneq_bf16): New.
832 (vdupq_lane_bf16, vdupq_laneq_bf16): New.
833 (vduph_lane_bf16, vduph_laneq_bf16): New.
834 (vreinterpret_bf16_u8, vreinterpretq_bf16_u8): New.
835 (vreinterpret_bf16_u16, vreinterpretq_bf16_u16): New.
836 (vreinterpret_bf16_u32, vreinterpretq_bf16_u32): New.
837 (vreinterpret_bf16_u64, vreinterpretq_bf16_u64): New.
838 (vreinterpret_bf16_s8, vreinterpretq_bf16_s8): New.
839 (vreinterpret_bf16_s16, vreinterpretq_bf16_s16): New.
840 (vreinterpret_bf16_s32, vreinterpretq_bf16_s32): New.
841 (vreinterpret_bf16_s64, vreinterpretq_bf16_s64): New.
842 (vreinterpret_bf16_p8, vreinterpretq_bf16_p8): New.
843 (vreinterpret_bf16_p16, vreinterpretq_bf16_p16): New.
844 (vreinterpret_bf16_p64, vreinterpretq_bf16_p64): New
845 (vreinterpret_bf16_f16, vreinterpretq_bf16_f16): New
846 (vreinterpret_bf16_f32, vreinterpretq_bf16_f32): New.
847 (vreinterpret_bf16_f64, vreinterpretq_bf16_f64): New.
848 (vreinterpretq_bf16_p128): New.
849 (vreinterpret_s8_bf16, vreinterpretq_s8_bf16): New.
850 (vreinterpret_s16_bf16, vreinterpretq_s16_bf16): New.
851 (vreinterpret_s32_bf16, vreinterpretq_s32_bf16): New.
852 (vreinterpret_s64_bf16, vreinterpretq_s64_bf16): New.
853 (vreinterpret_u8_bf16, vreinterpretq_u8_bf16): New.
854 (vreinterpret_u16_bf16, vreinterpretq_u16_bf16): New.
855 (vreinterpret_u32_bf16, vreinterpretq_u32_bf16): New.
856 (vreinterpret_u64_bf16, vreinterpretq_u64_bf16): New.
857 (vreinterpret_p8_bf16, vreinterpretq_p8_bf16): New.
858 (vreinterpret_p16_bf16, vreinterpretq_p16_bf16): New.
859 (vreinterpret_p64_bf16, vreinterpretq_p64_bf16): New.
860 (vreinterpret_f32_bf16, vreinterpretq_f32_bf16): New.
861 (vreinterpret_f64_bf16,vreinterpretq_f64_bf16): New.
862 (vreinterpret_f16_bf16,vreinterpretq_f16_bf16): New.
863 (vreinterpretq_p128_bf16): New.
865 2020-02-25 Dennis Zhang <dennis.zhang@arm.com>
867 * config/arm/arm_neon.h (vbfdot_f32, vbfdotq_f32): New
868 (vbfdot_lane_f32, vbfdotq_laneq_f32): New.
869 (vbfdot_laneq_f32, vbfdotq_lane_f32): New.
870 * config/arm/arm_neon_builtins.def (vbfdot): New entry.
871 (vbfdot_lanev4bf, vbfdot_lanev8bf): Likewise.
872 * config/arm/iterators.md (VSF2BF): New attribute.
873 * config/arm/neon.md (neon_vbfdot<VCVTF:mode>): New entry.
874 (neon_vbfdot_lanev4bf<VCVTF:mode>): Likewise.
875 (neon_vbfdot_lanev8bf<VCVTF:mode>): Likewise.
877 2020-02-25 Christophe Lyon <christophe.lyon@linaro.org>
879 * config/arm/arm.md (required_for_purecode): New attribute.
880 (enabled): Handle required_for_purecode.
881 * config/arm/thumb1.md (thumb1_movsi_insn): Add alternative to
882 work with -mpure-code.
884 2020-02-25 Jakub Jelinek <jakub@redhat.com>
886 PR rtl-optimization/93908
887 * combine.c (find_split_point): For store into ZERO_EXTRACT, and src
890 2019-02-25 Eric Botcazou <ebotcazou@adacore.com>
892 * dwarf2out.c (dwarf2out_size_function): Run in early-DWARF mode.
894 2020-02-25 Roman Zhuykov <zhroma@ispras.ru>
896 * doc/install.texi (--enable-checking): Adjust wording.
898 2020-02-25 Richard Biener <rguenther@suse.de>
900 PR tree-optimization/93868
901 * tree-vect-slp.c (slp_copy_subtree): New function.
902 (vect_attempt_slp_rearrange_stmts): Copy the SLP tree before
903 re-arranging stmts in it.
905 2020-02-25 Jakub Jelinek <jakub@redhat.com>
908 * passes.c (pass_manager::dump_passes): Create a cgraph node for the
909 dummy function and remove it at the end.
912 * config/lm32/lm32.c (lm32_setup_incoming_varargs): Fix comment typo
913 paramter -> parameter.
914 * config/aarch64/aarch64.c (aarch64_is_extend_from_extract): Likewise.
915 * ipa-prop.h (struct ipa_agg_replacement_value): Likewise.
917 2020-02-24 Roman Zhuykov <zhroma@ispras.ru>
919 * doc/install.texi (--enable-checking): Properly document current
921 (--enable-stage1-checking): Minor clarification about bootstrap.
923 2020-02-24 David Malcolm <dmalcolm@redhat.com>
926 * doc/invoke.texi (-Wnanalyzer-tainted-array-index): Note that
927 -fanalyzer-checker=taint is also required.
928 (-fanalyzer-checker=): Note that providing this option enables the
929 given checker, and doing so may be required for checkers that are
932 2020-02-24 David Malcolm <dmalcolm@redhat.com>
934 * doc/invoke.texi (-fanalyzer-verbosity=): "2" only shows
935 significant control flow events; add a "3" which shows all
936 control flow events; the old "3" becomes "4".
938 2020-02-24 Jakub Jelinek <jakub@redhat.com>
940 PR tree-optimization/93582
941 * tree-ssa-sccvn.c (vn_walk_cb_data::push_partial_def): Consider
942 pd.offset and pd.size to be counted in bits rather than bytes, add
943 support for maxsizei that is not a multiple of BITS_PER_UNIT and
944 handle bitfield stores and loads.
945 (vn_reference_lookup_3): Don't call ranges_known_overlap_p with
946 uncomparable quantities - bytes vs. bits. Allow push_partial_def
947 on offsets/sizes that aren't multiple of BITS_PER_UNIT and adjust
948 pd.offset/pd.size to be counted in bits rather than bytes.
949 Formatting fix. Rename shadowed len variable to buflen.
951 2020-02-24 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
952 Kugan Vivekandarajah <kugan.vivekanandarajah@linaro.org>
955 * gcc.c (putenv_COLLECT_AS_OPTIONS): New function.
956 (driver::main): Call putenv_COLLECT_AS_OPTIONS.
957 * opts-common.c (parse_options_from_collect_gcc_options): New function.
958 (prepend_xassembler_to_collect_as_options): Likewise.
959 * opts.h (parse_options_from_collect_gcc_options): Declare prototype.
960 (prepend_xassembler_to_collect_as_options): Likewise.
961 * lto-opts.c (lto_write_options): Stream assembler options
962 in COLLECT_AS_OPTIONS.
963 * lto-wrapper.c (xassembler_options_error): New static variable.
964 (get_options_from_collect_gcc_options): Move parsing options code to
965 parse_options_from_collect_gcc_options and call it.
966 (merge_and_complain): Validate -Xassembler options.
967 (append_compiler_options): Handle OPT_Xassembler.
968 (run_gcc): Append command line -Xassembler options to
970 * doc/invoke.texi: Add documentation about using Xassembler
973 2020-02-24 Kito Cheng <kito.cheng@sifive.com>
975 * config/riscv/riscv.c (riscv_emit_float_compare): Change the code gen
977 (riscv_rtx_costs): Update cost model for LTGT.
979 2020-02-23 Vladimir Makarov <vmakarov@redhat.com>
981 PR rtl-optimization/93564
982 * ira-color.c (struct update_cost_queue_elem): New member start.
983 (queue_update_cost, get_next_update_cost): Add new arg start.
984 (allocnos_conflict_p): New function.
985 (update_costs_from_allocno): Add new arg conflict_cost_update_p.
986 Add checking conflicts with allocnos_conflict_p.
987 (update_costs_from_prefs, restore_costs_from_copies): Adjust
988 update_costs_from_allocno calls.
989 (update_conflict_hard_regno_costs): Add checking conflicts with
990 allocnos_conflict_p. Adjust calls of queue_update_cost and
991 get_next_update_cost.
992 (assign_hard_reg): Adjust calls of queue_update_cost. Add
994 (bucket_allocno_compare_func): Restore previous version.
996 2020-02-21 John David Anglin <danglin@gcc.gnu.org>
998 * gcc/config/pa/pa.c (pa_function_value): Fix check for word and
999 double-word size when handling aggregate return values.
1000 * gcc/config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Fix to indicate
1001 that homogeneous SFmode and DFmode aggregates are passed and returned
1002 in general registers.
1004 2020-02-21 Jakub Jelinek <jakub@redhat.com>
1006 PR translation/93759
1007 * opts.c (print_filtered_help): Translate help before appending
1008 messages to it rather than after that.
1010 2020-02-19 Richard Sandiford <richard.sandiford@arm.com>
1012 PR rtl-optimization/PR92989
1013 * lra-lives.c (process_bb_lives): Restore the original order
1014 of the bb liveness update. Call make_hard_regno_dead for each
1015 register clobbered at the start of an EH receiver.
1017 2020-02-18 Feng Xue <fxue@os.amperecomputing.com>
1020 * ipa-cp.c (self_recursively_generated_p): Mark self-dependent value as
1021 self-recursively generated.
1023 2020-02-21 Iain Sandoe <iain@sandoe.co.uk>
1026 * config/darwin-c.c (pop_field_alignment): Adjust quoting of
1029 2020-02-21 Mihail Ionescu <mihail.ionescu@arm.com>
1031 * doc/sourcebuild.texi (arm_v8_1m_mve_ok):
1032 Document new target supports option.
1034 2020-02-21 Dennis Zhang <dennis.zhang@arm.com>
1036 * config/arm/arm_neon.h (vmmlaq_s32, vmmlaq_u32, vusmmlaq_s32): New.
1037 * config/arm/arm_neon_builtins.def (smmla, ummla, usmmla): New.
1038 * config/arm/iterators.md (MATMUL): New iterator.
1039 (sup): Add UNSPEC_MATMUL_S, UNSPEC_MATMUL_U, and UNSPEC_MATMUL_US.
1040 (mmla_sfx): New attribute.
1041 * config/arm/neon.md (neon_<sup>mmlav16qi): New.
1042 * config/arm/unspecs.md (UNSPEC_MATMUL_S, UNSPEC_MATMUL_U): New.
1043 (UNSPEC_MATMUL_US): New.
1045 2020-02-21 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1047 * config/arm/arm.md: Prevent scalar shifts from being used when big
1050 2020-02-21 Jan Hubicka <hubicka@ucw.cz>
1051 Richard Biener <rguenther@suse.de>
1053 PR tree-optimization/93586
1054 * tree-ssa-alias.c (nonoverlapping_array_refs_p): Finish array walk
1055 after mismatched array refs; do not sure type size information to
1056 recover from unmatched referneces with !flag_strict_aliasing_p.
1058 2020-02-21 Andrew Stubbs <ams@codesourcery.com>
1060 * config/gcn/gcn-valu.md (gather_load<mode>): Rename to ...
1061 (gather_load<mode>v64si): ... this and set operand 2 to V64SI.
1062 (scatter_store<mode>): Rename to ...
1063 (scatter_store<mode>v64si): ... this and set operand 1 to V64SI.
1064 (scatter<mode>_exec): Delete. Move contents ...
1065 (mask_scatter_store<mode>): ... here, and rename that to ...
1066 (mask_gather_load<mode>v64si): ... this. Set operand 2 to V64SI.
1067 Remove mode conversion.
1068 (mask_gather_load<mode>): Rename to ...
1069 (mask_scatter_store<mode>v64si): ... this. Set operand 1 to V64SI.
1070 Remove mode conversion.
1071 * config/gcn/gcn.c (gcn_expand_scaled_offsets): Remove mode conversion.
1073 2020-02-21 Martin Jambor <mjambor@suse.cz>
1075 PR tree-optimization/93845
1076 * tree-sra.c (verify_sra_access_forest): Only test access size of
1079 2020-02-21 Andrew Stubbs <ams@codesourcery.com>
1081 * config/gcn/gcn.c (gcn_hard_regno_mode_ok): Align VGPR pairs.
1082 * config/gcn/gcn-valu.md (addv64di3): Remove early-clobber.
1083 (addv64di3_exec): Likewise.
1084 (subv64di3): Likewise.
1085 (subv64di3_exec): Likewise.
1086 (addv64di3_zext): Likewise.
1087 (addv64di3_zext_exec): Likewise.
1088 (addv64di3_zext_dup): Likewise.
1089 (addv64di3_zext_dup_exec): Likewise.
1090 (addv64di3_zext_dup2): Likewise.
1091 (addv64di3_zext_dup2_exec): Likewise.
1092 (addv64di3_sext_dup2): Likewise.
1093 (addv64di3_sext_dup2_exec): Likewise.
1094 (<expander>v64di3): Likewise.
1095 (<expander>v64di3_exec): Likewise.
1096 (*<reduc_op>_dpp_shr_v64di): Likewise.
1097 (*plus_carry_dpp_shr_v64di): Likewise.
1098 * config/gcn/gcn.md (adddi3): Likewise.
1099 (addptrdi3): Likewise.
1100 (<expander>di3): Likewise.
1102 2020-02-21 Andrew Stubbs <ams@codesourcery.com>
1104 * config/gcn/gcn-valu.md (vec_seriesv64di): Use gen_vec_duplicatev64di.
1106 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
1108 * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Add SVE
1109 support. Use aarch64_emit_mult instead of emitting multiplication
1110 instructions directly.
1111 * config/aarch64/aarch64-sve.md (sqrt<mode>2, rsqrt<mode>2)
1112 (@aarch64_rsqrte<mode>, @aarch64_rsqrts<mode>): New expanders.
1114 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
1116 * config/aarch64/aarch64.c (aarch64_emit_mult): New function.
1117 (aarch64_emit_approx_div): Add SVE support. Use aarch64_emit_mult
1118 instead of emitting multiplication instructions directly.
1119 * config/aarch64/iterators.md (SVE_COND_FP_BINARY_OPTAB): New iterator.
1120 * config/aarch64/aarch64-sve.md (div<mode>3, @aarch64_frecpe<mode>)
1121 (@aarch64_frecps<mode>): New expanders.
1123 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
1125 * config/aarch64/aarch64-protos.h (AARCH64_APPROX_MODE): Operate
1126 on and produce uint64_ts rather than ints.
1127 (AARCH64_APPROX_NONE, AARCH64_APPROX_ALL): Change to uint64_ts.
1128 (cpu_approx_modes): Change the fields from unsigned int to uint64_t.
1130 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
1132 * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Don't create
1133 an unused xmsk register when handling approximate rsqrt.
1135 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
1137 * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Fix inverted
1138 flag_finite_math_only condition.
1140 2020-02-20 Uroš Bizjak <ubizjak@gmail.com>
1143 * config/i386/mmx.md (*vec_extractv2sf_1): Match source operand
1144 to destination operand for shufps alternative.
1145 (*vec_extractv2si_1): Ditto.
1147 2020-02-20 Peter Bergner <bergner@linux.ibm.com>
1150 * config/rs6000/rs6000.c (rs6000_legitimate_address_p): Handle VSX
1153 2020-02-20 Martin Liska <mliska@suse.cz>
1155 PR translation/93831
1156 * config/darwin.c (darwin_override_options): Change 64b to 64-bit mode.
1158 2020-02-20 Martin Liska <mliska@suse.cz>
1160 PR translation/93830
1161 * common/config/avr/avr-common.c: Remote trailing "|".
1163 2020-02-19 Bernd Edlinger <bernd.edlinger@hotmail.de>
1165 * collect2.c (maybe_run_lto_and_relink): Fix typo in
1168 2020-02-19 Richard Sandiford <richard.sandiford@arm.com>
1170 PR tree-optimization/93767
1171 * tree-vect-data-refs.c (vect_compile_time_alias): Remove the
1172 access-size bias from the offset calculations for negative strides.
1174 2020-02-19 Bernd Edlinger <bernd.edlinger@hotmail.de>
1176 * collect2.c (c_file, o_file): Make const again.
1177 (ldout,lderrout, dump_ld_file): Remove.
1178 (tool_cleanup): Avoid calling not signal-safe functions.
1179 (maybe_run_lto_and_relink): Avoid possible signal handler
1180 access to unintialzed memory (lto_o_files).
1181 (main): Avoid leaking temp files in $TMPDIR.
1182 Initialize c_file/o_file with concat, which avoids exposing
1183 uninitialized memory to signal handler, which calls unlink(!).
1184 Avoid calling maybe_unlink when the main function returns,
1185 since the atexit handler is already doing this.
1186 * collect2.h (dump_ld_file, ldout, lderrout): Remove.
1188 2020-02-19 Martin Jambor <mjambor@suse.cz>
1190 PR tree-optimization/93776
1191 * tree-sra.c (create_access): Do not create zero size accesses.
1192 (get_access_for_expr): Do not search for zero sized accesses.
1194 2020-02-19 Martin Jambor <mjambor@suse.cz>
1196 PR tree-optimization/93667
1197 * tree-sra.c (scalarizable_type_p): Return false if record fields
1198 do not follow wach other.
1200 2020-01-21 Kito Cheng <kito.cheng@sifive.com>
1202 * config/riscv/riscv.c (riscv_output_move) Using fmv.x.w/fmv.w.x
1203 rather than fmv.x.s/fmv.s.x.
1205 2020-02-18 James Greenhalgh <james.greenhalgh@arm.com>
1207 * config/aarch64/aarch64-simd-builtins.def
1208 (intrinsic_vec_smult_lo_): New.
1209 (intrinsic_vec_umult_lo_): Likewise.
1210 (vec_widen_smult_hi_): Likewise.
1211 (vec_widen_umult_hi_): Likewise.
1212 * config/aarch64/aarch64-simd.md
1213 (aarch64_intrinsic_vec_<su>mult_lo_<mode>): New.
1214 * config/aarch64/arm_neon.h (vmull_high_s8): Use intrinsics.
1215 (vmull_high_s16): Likewise.
1216 (vmull_high_s32): Likewise.
1217 (vmull_high_u8): Likewise.
1218 (vmull_high_u16): Likewise.
1219 (vmull_high_u32): Likewise.
1220 (vmull_s8): Likewise.
1221 (vmull_s16): Likewise.
1222 (vmull_s32): Likewise.
1223 (vmull_u8): Likewise.
1224 (vmull_u16): Likewise.
1225 (vmull_u32): Likewise.
1227 2020-02-18 Martin Liska <mliska@suse.cz>
1229 * value-prof.c (stream_out_histogram_value): Restore LTO PGO
1230 bootstrap by missing removal of invalid sanity check.
1232 2020-02-18 Martin Liska <mliska@suse.cz>
1235 * ipa-icf-gimple.c (func_checker::compare_gimple_assign):
1236 Always compare LHS of gimple_assign.
1238 2020-02-18 Martin Liska <mliska@suse.cz>
1241 * cgraph.c (cgraph_node::verify_node): Verify MALLOC attribute
1242 and return type of functions.
1243 * ipa-param-manipulation.c (ipa_param_adjustments::adjust_decl):
1244 Drop MALLOC attribute for void functions.
1245 * ipa-pure-const.c (funct_state_summary_t::duplicate): Drop
1246 malloc_state for a new VOID clone.
1248 2020-02-18 Martin Liska <mliska@suse.cz>
1251 * common.opt: Add -fprofile-reproducibility.
1252 * doc/invoke.texi: Document it.
1253 * value-prof.c (dump_histogram_value):
1254 Document and support behavior for counters[0]
1255 being a negative value.
1256 (get_nth_most_common_value): Handle negative
1257 counters[0] in respect to flag_profile_reproducible.
1259 2020-02-18 Jakub Jelinek <jakub@redhat.com>
1262 * cgraph.c (verify_speculative_call): Use speculative_id instead of
1263 speculative_uid in messages. Remove trailing whitespace from error
1264 message. Use num_speculative_call_targets instead of
1265 num_speculative_targets in a message.
1266 (cgraph_node::verify_node): Use call_stmt instead of cal_stmt in
1267 edge messages and stmt instead of cal_stmt in reference message.
1269 PR tree-optimization/93780
1270 * tree-ssa.c (non_rewritable_lvalue_p): Check valid_vector_subparts_p
1271 before calling build_vector_type.
1272 (execute_update_addresses_taken): Likewise.
1275 * params.opt (-param=ipa-max-switch-predicate-bounds=): Fix help
1276 typo, functoin -> function.
1277 * tree.c (free_lang_data_in_decl): Fix comment typo,
1278 functoin -> function.
1279 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
1281 2020-02-17 David Malcolm <dmalcolm@redhat.com>
1283 * diagnostic.c (print_any_cwe): Don't call get_cwe_url if URLs
1285 (print_option_information): Don't call get_option_url if URLs
1288 2020-02-17 Alexandre Oliva <oliva@adacore.com>
1290 * tree-emutls.c (new_emutls_decl, emutls_common_1): Complete
1291 handling of register_common-less targets.
1293 2020-02-17 Martin Liska <mliska@suse.cz>
1296 * ipa-devirt.c (odr_types_equivalent_p): Fix grammar.
1298 2020-02-17 Martin Liska <mliska@suse.cz>
1300 PR translation/93755
1301 * config/rs6000/rs6000.c (rs6000_option_override_internal):
1304 2020-02-17 Martin Liska <mliska@suse.cz>
1307 * config/rx/elf.opt: Fix typo.
1309 2020-02-17 Richard Biener <rguenther@suse.de>
1312 * opts-global.c (print_ignored_options): Use inform and
1315 2020-02-17 Jiufu Guo <guojiufu@linux.ibm.com>
1318 * config/rs6000/rs6000.md (untyped_call): Add emit_clobber.
1320 2020-02-16 Uroš Bizjak <ubizjak@gmail.com>
1323 * config/i386/i386.md (atan2xf3): Swap operands 1 and 2.
1324 (atan2<mode>3): Update operand order in the call to gen_atan2xf3.
1326 2020-02-15 Jason Merrill <jason@redhat.com>
1328 * doc/invoke.texi (C Dialect Options): Add -std=c++20.
1330 2020-02-15 Jakub Jelinek <jakub@redhat.com>
1332 PR tree-optimization/93744
1333 * match.pd (((m1 >/</>=/<= m2) * d -> (m1 >/</>=/<= m2) ? d : 0,
1334 A - ((A - B) & -(C cmp D)) -> (C cmp D) ? B : A,
1335 A + ((B - A) & -(C cmp D)) -> (C cmp D) ? B : A): For GENERIC, make
1336 sure @2 in the first and @1 in the other patterns has no side-effects.
1338 2020-02-15 David Malcolm <dmalcolm@redhat.com>
1339 Bernd Edlinger <bernd.edlinger@hotmail.de>
1343 * config.in (DIAGNOSTICS_URLS_DEFAULT): New define.
1344 * configure.ac (--with-diagnostics-urls): New configuration
1345 option, based on --with-diagnostics-color.
1346 (DIAGNOSTICS_URLS_DEFAULT): New define.
1347 * config.h: Regenerate.
1348 * configure: Regenerate.
1349 * diagnostic.c (diagnostic_urls_init): Handle -1 for
1350 DIAGNOSTICS_URLS_DEFAULT from configure-time
1351 --with-diagnostics-urls=auto-if-env by querying for a GCC_URLS
1352 and TERM_URLS environment variable.
1353 * diagnostic-url.h (diagnostic_url_format): New enum type.
1354 (diagnostic_urls_enabled_p): rename to...
1355 (determine_url_format): ... this, and change return type.
1356 * diagnostic-color.c (parse_env_vars_for_urls): New helper function.
1357 (auto_enable_urls): Disable URLs on xfce4-terminal, gnome-terminal,
1358 the linux console, and mingw.
1359 (diagnostic_urls_enabled_p): rename to...
1360 (determine_url_format): ... this, and adjust.
1361 * pretty-print.h (pretty_printer::show_urls): rename to...
1362 (pretty_printer::url_format): ... this, and change to enum.
1363 * pretty-print.c (pretty_printer::pretty_printer,
1364 pp_begin_url, pp_end_url, test_urls): Adjust.
1365 * doc/install.texi (--with-diagnostics-urls): Document the new
1366 configuration option.
1367 (--with-diagnostics-color): Document the existing interaction
1368 with GCC_COLORS better.
1369 * doc/invoke.texi (-fdiagnostics-urls): Add GCC_URLS and TERM_URLS
1370 vindex reference. Update description of defaults based on the above.
1371 (-fdiagnostics-color): Update description of how -fdiagnostics-color
1372 interacts with GCC_COLORS.
1374 2020-02-14 Eric Botcazou <ebotcazou@adacore.com>
1377 * config/sparc/sparc.c (eligible_for_call_delay): Test HAVE_GNU_LD in
1378 conjunction with TARGET_GNU_TLS in early return.
1380 2020-02-14 Alexander Monakov <amonakov@ispras.ru>
1382 * rtlanal.c (rtx_cost): Handle a SET up front. Avoid division if
1383 the mode is not wider than UNITS_PER_WORD.
1385 2020-02-14 Martin Jambor <mjambor@suse.cz>
1387 PR tree-optimization/93516
1388 * tree-sra.c (propagate_subaccesses_from_rhs): Do not create
1389 access of the same type as the parent.
1390 (propagate_subaccesses_from_lhs): Likewise.
1392 2020-02-14 Hongtao Liu <hongtao.liu@intel.com>
1395 * config/i386/avx512vbmi2intrin.h
1396 (_mm512_shrdi_epi16, _mm512_mask_shrdi_epi16,
1397 _mm512_maskz_shrdi_epi16, _mm512_shrdi_epi32,
1398 _mm512_mask_shrdi_epi32, _mm512_maskz_shrdi_epi32,
1399 _m512_shrdi_epi64, _m512_mask_shrdi_epi64,
1400 _m512_maskz_shrdi_epi64, _mm512_shldi_epi16,
1401 _mm512_mask_shldi_epi16, _mm512_maskz_shldi_epi16,
1402 _mm512_shldi_epi32, _mm512_mask_shldi_epi32,
1403 _mm512_maskz_shldi_epi32, _mm512_shldi_epi64,
1404 _mm512_mask_shldi_epi64, _mm512_maskz_shldi_epi64): Fix typo
1405 of lacking a closing parenthesis.
1406 * config/i386/avx512vbmi2vlintrin.h
1407 (_mm256_shrdi_epi16, _mm256_mask_shrdi_epi16,
1408 _mm256_maskz_shrdi_epi16, _mm256_shrdi_epi32,
1409 _mm256_mask_shrdi_epi32, _mm256_maskz_shrdi_epi32,
1410 _m256_shrdi_epi64, _m256_mask_shrdi_epi64,
1411 _m256_maskz_shrdi_epi64, _mm256_shldi_epi16,
1412 _mm256_mask_shldi_epi16, _mm256_maskz_shldi_epi16,
1413 _mm256_shldi_epi32, _mm256_mask_shldi_epi32,
1414 _mm256_maskz_shldi_epi32, _mm256_shldi_epi64,
1415 _mm256_mask_shldi_epi64, _mm256_maskz_shldi_epi64,
1416 _mm_shrdi_epi16, _mm_mask_shrdi_epi16,
1417 _mm_maskz_shrdi_epi16, _mm_shrdi_epi32,
1418 _mm_mask_shrdi_epi32, _mm_maskz_shrdi_epi32,
1419 _mm_shrdi_epi64, _mm_mask_shrdi_epi64,
1420 _m_maskz_shrdi_epi64, _mm_shldi_epi16,
1421 _mm_mask_shldi_epi16, _mm_maskz_shldi_epi16,
1422 _mm_shldi_epi32, _mm_mask_shldi_epi32,
1423 _mm_maskz_shldi_epi32, _mm_shldi_epi64,
1424 _mm_mask_shldi_epi64, _mm_maskz_shldi_epi64): Ditto.
1426 2020-02-13 H.J. Lu <hongjiu.lu@intel.com>
1429 * config/i386/i386.c (ix86_trampoline_init): Skip ENDBR32 at
1430 the target function entry.
1432 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
1434 * common/config/arc/arc-common.c (arc_option_optimization_table):
1435 Disable if-conversion step when optimized for size.
1437 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
1439 * config/arc/arc.c (arc_conditional_register_usage): R0-R3 and
1440 R12-R15 are always in ARCOMPACT16_REGS register class.
1441 * config/arc/arc.opt (mq-class): Deprecate.
1442 * config/arc/constraint.md ("q"): Remove dependency on mq-class
1444 * doc/invoke.texi (mq-class): Update text.
1445 * common/config/arc/arc-common.c (arc_option_optimization_table):
1448 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
1450 * config/arc/arc.c (arc_insn_cost): New function.
1451 (TARGET_INSN_COST): Define.
1452 * config/arc/arc.md (cost): New attribute.
1453 (add_n): Use arc_nonmemory_operand.
1454 (ashlsi3_insn): Likewise, also update constraints.
1455 (ashrsi3_insn): Likewise.
1456 (rotrsi3): Likewise.
1457 (add_shift): Likewise.
1458 * config/arc/predicates.md (arc_nonmemory_operand): New predicate.
1460 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
1462 * config/arc/arc.md (mulsidi_600): Correctly select mlo/mhi
1464 (umulsidi_600): Likewise.
1466 2020-02-13 Jakub Jelinek <jakub@redhat.com>
1469 * config/i386/avx512bitalgintrin.h (_mm512_mask_popcnt_epi8,
1470 _mm512_mask_popcnt_epi16, _mm256_mask_popcnt_epi8,
1471 _mm256_mask_popcnt_epi16, _mm_mask_popcnt_epi8,
1472 _mm_mask_popcnt_epi16): Rename __B argument to __A and __A to __W,
1473 pass __A to the builtin followed by __W instead of __A followed by
1475 * config/i386/avx512vpopcntdqintrin.h (_mm512_mask_popcnt_epi32,
1476 _mm512_mask_popcnt_epi64): Likewise.
1477 * config/i386/avx512vpopcntdqvlintrin.h (_mm_mask_popcnt_epi32,
1478 _mm256_mask_popcnt_epi32, _mm_mask_popcnt_epi64,
1479 _mm256_mask_popcnt_epi64): Likewise.
1481 PR tree-optimization/93582
1482 * fold-const.h (shift_bytes_in_array_left,
1483 shift_bytes_in_array_right): Declare.
1484 * fold-const.c (shift_bytes_in_array_left,
1485 shift_bytes_in_array_right): New function, moved from
1486 gimple-ssa-store-merging.c, no longer static.
1487 * gimple-ssa-store-merging.c (shift_bytes_in_array): Move
1488 to gimple-ssa-store-merging.c and rename to shift_bytes_in_array_left.
1489 (shift_bytes_in_array_right): Move to gimple-ssa-store-merging.c.
1490 (encode_tree_to_bitpos): Use shift_bytes_in_array_left instead of
1491 shift_bytes_in_array.
1492 (verify_shift_bytes_in_array): Rename to ...
1493 (verify_shift_bytes_in_array_left): ... this. Use
1494 shift_bytes_in_array_left instead of shift_bytes_in_array.
1495 (store_merging_c_tests): Call verify_shift_bytes_in_array_left
1496 instead of verify_shift_bytes_in_array.
1497 * tree-ssa-sccvn.c (vn_reference_lookup_3): For native_encode_expr
1498 / native_interpret_expr where the store covers all needed bits,
1499 punt on PDP-endian, otherwise allow all involved offsets and sizes
1500 not to be byte-aligned.
1503 * config/i386/sse.md (k<code><mode>): Drop mode from last operand and
1504 use const_0_to_255_operand predicate instead of immediate_operand.
1505 (avx512dq_fpclass<mode><mask_scalar_merge_name>,
1506 avx512dq_vmfpclass<mode><mask_scalar_merge_name>,
1507 vgf2p8affineinvqb_<mode><mask_name>,
1508 vgf2p8affineqb_<mode><mask_name>): Drop mode from
1509 const_0_to_255_operand predicated operands.
1511 2020-02-12 Jeff Law <law@redhat.com>
1513 * config/h8300/h8300.md (comparison shortening peepholes): Use
1514 a mode iterator to merge the HImode and SImode peepholes.
1516 2020-02-12 Jakub Jelinek <jakub@redhat.com>
1519 * real.c (is_even): Make static. Function comment fix.
1520 (is_halfway_below): Make static, don't assert R is not inf/nan,
1521 instead return false for those. Small formatting fixes.
1523 2020-02-12 Martin Sebor <msebor@redhat.com>
1526 * tree-ssa-strlen.c (handle_builtin_stxncpy): Rename...
1527 (handle_builtin_stxncpy_strncat): ...to this. Change first argument.
1528 Issue only -Wstringop-overflow strncat, never -Wstringop-truncation.
1529 (strlen_check_and_optimize_call): Adjust callee name.
1531 2020-02-12 Jeff Law <law@redhat.com>
1533 * config/h8300/h8300.md (comparison shortening peepholes): Drop
1534 (and (xor)) variant. Combine other two into single peephole.
1536 2020-02-12 Wilco Dijkstra <wdijkstr@arm.com>
1538 PR rtl-optimization/93565
1539 * config/aarch64/aarch64.c (aarch64_rtx_costs): Add CTZ costs.
1541 2020-02-12 Wilco Dijkstra <wdijkstr@arm.com>
1543 * config/aarch64/aarch64-simd.md
1544 (aarch64_zero_extend<GPI:mode>_reduc_plus_<VDQV_E:mode>): New pattern.
1545 * config/aarch64/aarch64.md (popcount<mode>2): Use it instead of
1546 generating separate ADDV and zero_extend patterns.
1547 * config/aarch64/iterators.md (VDQV_E): New iterator.
1549 2020-02-12 Jeff Law <law@redhat.com>
1551 * config/h8300/h8300.md (cpymemsi, movmd): Remove dead patterns,
1552 expanders, splits, etc.
1553 (movmd_internal_<mode>, movmd splitter, movstr, movsd): Likewise.
1554 (stpcpy_internal_<mode>, stpcpy splitter): Likewise.
1555 (peepholes to convert QI/HI mode pushes to SI mode pushes): Likewise.
1556 * config/h8300/h8300.c (h8300_swap_into_er6): Remove unused function.
1557 (h8300_swap_out_of_er6, h8sx_emit_movmd): Likewise
1558 * config/h8300/h8300-protos.h (h8300_swap_into_er6): Remove unused
1560 (h8300_swap_out_of_er6, h8sx_emit_movmd): Likewise.
1562 2020-02-12 Jakub Jelinek <jakub@redhat.com>
1565 * config/i386/sse.md (VI48F_256_DQ): New mode iterator.
1566 (avx512vl_vextractf128<mode>): Use it instead of VI48F_256. Remove
1567 TARGET_AVX512DQ from condition.
1568 (vec_extract_lo_<mode><mask_name>): Use <mask_avx512dq_condition>
1569 instead of <mask_mode512bit_condition> in condition. If
1570 TARGET_AVX512DQ is false, emit vextract*64x4 instead of
1572 (vec_extract_lo_<mode><mask_name>): Drop <mask_avx512dq_condition>
1575 2020-02-12 Kewen Lin <linkw@gcc.gnu.org>
1578 * ira.c (combine_and_move_insns): Skip multiple_sets def_insn.
1580 2020-02-12 Segher Boessenkool <segher@kernel.crashing.org>
1582 * config/rs6000/rs6000.c (rs6000_debug_print_mode): Don't use sizeof
1583 where strlen is more legible.
1584 (rs6000_builtin_vectorized_libmass): Ditto.
1585 (rs6000_print_options_internal): Ditto.
1587 2020-02-11 Martin Sebor <msebor@redhat.com>
1589 PR tree-optimization/93683
1590 * tree-ssa-alias.c (stmt_kills_ref_p): Avoid using LHS when not set.
1592 2020-02-11 Michael Meissner <meissner@linux.ibm.com>
1594 * config/rs6000/predicates.md (cint34_operand): Rename the
1595 -mprefixed-addr option to be -mprefixed.
1596 * config/rs6000/rs6000-cpus.def (ISA_FUTURE_MASKS_SERVER): Rename
1597 the -mprefixed-addr option to be -mprefixed.
1598 (OTHER_FUTURE_MASKS): Likewise.
1599 (POWERPC_MASKS): Likewise.
1600 * config/rs6000/rs6000.c (rs6000_option_override_internal): Rename
1601 the -mprefixed-addr option to be -mprefixed. Change error
1602 messages to refer to -mprefixed.
1603 (num_insns_constant_gpr): Rename the -mprefixed-addr option to be
1605 (rs6000_legitimate_offset_address_p): Likewise.
1606 (rs6000_mode_dependent_address): Likewise.
1607 (rs6000_opt_masks): Change the spelling of "-mprefixed-addr" to be
1608 "-mprefixed" for target attributes and pragmas.
1609 (address_to_insn_form): Rename the -mprefixed-addr option to be
1611 (rs6000_adjust_insn_length): Likewise.
1612 * config/rs6000/rs6000.h (FINAL_PRESCAN_INSN): Rename the
1613 -mprefixed-addr option to be -mprefixed.
1614 (ASM_OUTPUT_OPCODE): Likewise.
1615 * config/rs6000/rs6000.md (prefixed insn attribute): Rename the
1616 -mprefixed-addr option to be -mprefixed.
1617 * config/rs6000/rs6000.opt (-mprefixed): Rename the
1618 -mprefixed-addr option to be prefixed. Change the option from
1619 being undocumented to being documented.
1620 * doc/invoke.texi (RS/6000 and PowerPC Options): Document the
1621 -mprefixed option. Update the -mpcrel documentation to mention
1624 2020-02-11 Hans-Peter Nilsson <hp@axis.com>
1626 * ira-conflicts.c (print_hard_reg_set): Correct output for sets
1627 including FIRST_PSEUDO_REGISTER - 1.
1628 * ira-color.c (print_hard_reg_set): Ditto.
1630 2020-02-11 Stam Markianos-Wright <stam.markianos-wright@arm.com>
1632 * config/arm/arm-builtins.c (enum arm_type_qualifiers):
1633 (USTERNOP_QUALIFIERS): New define.
1634 (USMAC_LANE_QUADTUP_QUALIFIERS): New define.
1635 (SUMAC_LANE_QUADTUP_QUALIFIERS): New define.
1636 (arm_expand_builtin_args): Add case ARG_BUILTIN_LANE_QUADTUP_INDEX.
1637 (arm_expand_builtin_1): Add qualifier_lane_quadtup_index.
1638 * config/arm/arm_neon.h (vusdot_s32): New.
1639 (vusdot_lane_s32): New.
1640 (vusdotq_lane_s32): New.
1641 (vsudot_lane_s32): New.
1642 (vsudotq_lane_s32): New.
1643 * config/arm/arm_neon_builtins.def (usdot, usdot_lane,sudot_lane): New.
1644 * config/arm/iterators.md (DOTPROD_I8MM): New.
1645 (sup, opsuffix): Add <us/su>.
1646 * config/arm/neon.md (neon_usdot, <us/su>dot_lane: New.
1647 * config/arm/unspecs.md (UNSPEC_DOT_US, UNSPEC_DOT_SU): New.
1649 2020-02-11 Richard Biener <rguenther@suse.de>
1651 PR tree-optimization/93661
1652 PR tree-optimization/93662
1653 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly guard
1655 * tree-sra.c (get_access_for_expr): Likewise.
1657 2020-02-10 Jakub Jelinek <jakub@redhat.com>
1660 * config/i386/sse.md (VI_256_AVX2): New mode iterator.
1661 (vcond_mask_<mode><sseintvecmodelower>): Use it instead of VI_256.
1662 Change condition from TARGET_AVX2 to TARGET_AVX.
1664 2020-02-10 Iain Sandoe <iain@sandoe.co.uk>
1667 * config/darwin-c.c (darwin_cfstring_ref_p): Fix up last
1668 argument of strncmp.
1670 2020-02-10 Hans-Peter Nilsson <hp@axis.com>
1672 Try to generate zero-based comparisons.
1673 * config/cris/cris.c (cris_reduce_compare): New function.
1674 * config/cris/cris-protos.h (cris_reduce_compare): Add prototype.
1675 * config/cris/cris.md ("cbranch<mode>4", "cbranchdi4", "cstoredi4")
1676 (cstore<mode>4"): Apply cris_reduce_compare in expanders.
1678 2020-02-10 Richard Earnshaw <rearnsha@arm.com>
1681 * config/arm/arm.md (movsi_compare0): Allow SP as a source register
1682 in Thumb state and also as a destination in Arm state. Add T16
1685 2020-02-10 Hans-Peter Nilsson <hp@axis.com>
1687 * md.texi (Define Subst): Match closing paren in example.
1689 2020-02-10 Jakub Jelinek <jakub@redhat.com>
1693 * config/i386/i386.c (x86_64_elf_section_type_flags): Fix up last
1694 arguments of strncmp.
1696 2020-02-10 Feng Xue <fxue@os.amperecomputing.com>
1699 * ipa-cp.c (ipcp_lattice::add_value): Add source with same call edge
1700 but different source value.
1701 (adjust_callers_for_value_intersection): New function.
1702 (gather_edges_for_value): Adjust order of callers to let a
1703 non-self-recursive caller be the first element.
1704 (self_recursive_pass_through_p): Add a new parameter "simple", and
1705 check generalized self-recursive pass-through jump function.
1706 (self_recursive_agg_pass_through_p): Likewise.
1707 (find_more_scalar_values_for_callers_subset): Compute value from
1708 pass-through jump function for self-recursive.
1709 (intersect_with_plats): Cleanup previous implementation code for value
1710 itersection with self-recursive call edge.
1711 (intersect_with_agg_replacements): Likewise.
1712 (intersect_aggregates_with_edge): Deduce value from pass-through jump
1713 function for self-recursive call edge. Cleanup previous implementation
1714 code for value intersection with self-recursive call edge.
1715 (decide_whether_version_node): Remove dead callers and adjust order
1716 to let a non-self-recursive caller be the first element.
1718 2020-02-09 Uroš Bizjak <ubizjak@gmail.com>
1720 * recog.c: Move pass_split_before_sched2 code in front of
1721 pass_split_before_regstack.
1722 (pass_data_split_before_sched2): Rename pass to split3 from split4.
1723 (pass_data_split_before_regstack): Rename pass to split4 from split3.
1724 (rest_of_handle_split_before_sched2): Remove.
1725 (pass_split_before_sched2::execute): Unconditionally call
1727 (enable_split_before_sched2): New function.
1728 (pass_split_before_sched2::gate): Use enable_split_before_sched2.
1729 (pass_split_before_regstack::gate): Ditto.
1730 * config/nds32/nds32.c (nds32_split_double_word_load_store_p):
1731 Update name check for renamed split4 pass.
1732 * config/sh/sh.c (register_sh_passes): Update pass insertion
1733 point for renamed split4 pass.
1735 2020-02-09 Jakub Jelinek <jakub@redhat.com>
1737 * gimplify.c (gimplify_adjust_omp_clauses_1): Promote
1738 DECL_IN_CONSTANT_POOL variables into "omp declare target" to avoid
1739 copying them around between host and target.
1741 2020-02-08 Andrew Pinski <apinski@marvell.com>
1744 * config/aarch64/aarch64-simd.md (movmisalign<mode>): Check
1745 STRICT_ALIGNMENT also.
1747 2020-02-08 Jim Wilson <jimw@sifive.com>
1750 * config/riscv/riscv.h (HARD_REGNO_CALLER_SAVE_MODE): Define.
1752 2020-02-08 Uroš Bizjak <ubizjak@gmail.com>
1753 Jakub Jelinek <jakub@redhat.com>
1756 * config/i386/i386.h (CALL_USED_REGISTERS): Make
1757 xmm16-xmm31 call-used even in 64-bit ms-abi.
1759 2020-02-07 Dennis Zhang <dennis.zhang@arm.com>
1761 * config/aarch64/aarch64-simd-builtins.def (simd_smmla): New entry.
1762 (simd_ummla, simd_usmmla): Likewise.
1763 * config/aarch64/aarch64-simd.md (aarch64_simd_<sur>mmlav16qi): New.
1764 * config/aarch64/arm_neon.h (vmmlaq_s32, vmmlaq_u32): New.
1765 (vusmmlaq_s32): New.
1767 2020-02-07 Richard Biener <rguenther@suse.de>
1770 * tree-inline.c (fold_marked_statements): Do a PRE walk,
1771 skipping unreachable regions.
1772 (optimize_inline_calls): Skip folding stmts when we didn't
1775 2020-02-07 H.J. Lu <hongjiu.lu@intel.com>
1778 * config/i386/i386.c (function_arg_ms_64): Add a type argument.
1779 Don't return aggregates with only SFmode and DFmode in SSE
1781 (ix86_function_arg): Pass arg.type to function_arg_ms_64.
1783 2020-02-07 Jakub Jelinek <jakub@redhat.com>
1786 * config/rs6000/rs6000-logue.c
1787 (rs6000_emit_probe_stack_range_stack_clash): Always use gen_add3_insn,
1788 if it fails, move rs into end_addr and retry. Add
1789 REG_FRAME_RELATED_EXPR note whenever it returns more than one insn or
1790 the insn pattern doesn't describe well what exactly happens to
1794 * config/i386/predicates.md (avx_identity_operand): Remove.
1795 * config/i386/sse.md (*avx_vec_concat<mode>_1): Remove.
1796 (avx_<castmode><avxsizesuffix>_<castmode>,
1797 avx512f_<castmode><avxsizesuffix>_256<castmode>): Change patterns to
1798 a VEC_CONCAT of the operand and UNSPEC_CAST.
1799 (avx512f_<castmode><avxsizesuffix>_<castmode>): Change pattern to
1800 a VEC_CONCAT of VEC_CONCAT of the operand and UNSPEC_CAST with
1804 * config/i386/i386.c (ix86_lea_outperforms): Make sure to clear
1805 recog_data.insn if distance_non_agu_define changed it.
1807 2020-02-06 Michael Meissner <meissner@linux.ibm.com>
1810 * config/rs6000/rs6000.c (reg_to_non_prefixed): Before ISA 3.0
1811 we only had X-FORM (reg+reg) addressing for vectors. Also before
1812 ISA 3.0, we only had X-FORM addressing for scalars in the
1813 traditional Altivec registers.
1815 2020-02-06 <zhongyunde@huawei.com>
1816 Vladimir Makarov <vmakarov@redhat.com>
1818 PR rtl-optimization/93561
1819 * lra-assigns.c (spill_for): Check that tested hard regno is not out of
1820 hard register range.
1822 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
1824 * config/aarch64/aarch64.md (aarch64_movk<mode>): Add a type
1827 2020-02-06 Segher Boessenkool <segher@kernel.crashing.org>
1829 * config/rs6000/rs6000.c (rs6000_emit_set_long_const): Handle the case
1830 where the low and the high 32 bits are equal to each other specially,
1831 with an rldimi instruction.
1833 2020-02-06 Mihail Ionescu <mihail.ionescu@arm.com>
1835 * config/arm/arm-cpus.in: Set profile M for armv8.1-m.main.
1837 2020-02-06 Mihail Ionescu <mihail.ionescu@arm.com>
1839 * config/arm/arm-tables.opt: Regenerate.
1841 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
1844 * config/aarch64/aarch64-protos.h (aarch64_movk_shift): Declare.
1845 * config/aarch64/aarch64.c (aarch64_movk_shift): New function.
1846 * config/aarch64/aarch64.md (aarch64_movk<mode>): New pattern.
1848 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
1850 PR rtl-optimization/87763
1851 * config/aarch64/aarch64.md (*ashiftsi_extvdi_bfiz): New pattern.
1853 2020-02-06 Delia Burduv <delia.burduv@arm.com>
1855 * config/aarch64/aarch64-simd-builtins.def
1856 (bfmlaq): New built-in function.
1857 (bfmlalb): New built-in function.
1858 (bfmlalt): New built-in function.
1859 (bfmlalb_lane): New built-in function.
1860 (bfmlalt_lane): New built-in function.
1861 * config/aarch64/aarch64-simd.md
1862 (aarch64_bfmmlaqv4sf): New pattern.
1863 (aarch64_bfmlal<bt>v4sf): New pattern.
1864 (aarch64_bfmlal<bt>_lane<q>v4sf): New pattern.
1865 * config/aarch64/arm_neon.h (vbfmmlaq_f32): New intrinsic.
1866 (vbfmlalbq_f32): New intrinsic.
1867 (vbfmlaltq_f32): New intrinsic.
1868 (vbfmlalbq_lane_f32): New intrinsic.
1869 (vbfmlaltq_lane_f32): New intrinsic.
1870 (vbfmlalbq_laneq_f32): New intrinsic.
1871 (vbfmlaltq_laneq_f32): New intrinsic.
1872 * config/aarch64/iterators.md (BF_MLA): New int iterator.
1873 (bt): New int attribute.
1875 2020-02-06 Uroš Bizjak <ubizjak@gmail.com>
1877 * config/i386/i386.md (*pushtf): Emit "#" instead of
1878 calling gcc_unreachable in insn output.
1881 (*pushsf_rex64): Ditto for alternatives other than 1.
1882 (*pushsf): Ditto for alternatives other than 1.
1884 2020-02-06 Martin Liska <mliska@suse.cz>
1886 PR gcov-profile/91971
1887 PR gcov-profile/93466
1888 * coverage.c (coverage_init): Revert mangling of
1889 path into filename. It can lead to huge filename length.
1890 Creation of subfolders seem more natural.
1892 2020-02-06 Stam Markianos-Wright <stam.markianos-wright@arm.com>
1895 * config/arm/arm.c (arm_block_arith_comp_libfuncs_for_mode): New.
1896 (arm_init_libfuncs): Add BFmode support to block spurious BF libfuncs.
1897 Use arm_block_arith_comp_libfuncs_for_mode for HFmode.
1899 2020-02-06 Jakub Jelinek <jakub@redhat.com>
1902 * config/i386/predicates.md (avx_identity_operand): New predicate.
1903 * config/i386/sse.md (*avx_vec_concat<mode>_1): New
1904 define_insn_and_split.
1907 * omp-low.c (use_pointer_for_field): For nested constructs, also
1908 look for map clauses on target construct.
1909 (scan_omp_1_stmt) <case GIMPLE_OMP_TARGET>: Bump temporarily
1910 taskreg_nesting_level.
1913 * gimplify.c (gimplify_scan_omp_clauses) <do_notice>: If adding
1914 shared clause, call omp_notice_variable on outer context if any.
1916 2020-02-05 Jason Merrill <jason@redhat.com>
1919 * symtab.c (symtab_node::nonzero_address): A DECL_COMDAT decl has
1920 non-zero address even if weak and not yet defined.
1922 2020-02-05 Martin Sebor <msebor@redhat.com>
1924 PR tree-optimization/92765
1925 * gimple-fold.c (get_range_strlen_tree): Handle MEM_REF and PARM_DECL.
1926 * tree-ssa-strlen.c (compute_string_length): Remove.
1927 (determine_min_objsize): Remove.
1928 (get_len_or_size): Add an argument. Call get_range_strlen_dynamic.
1929 Avoid using type size as the upper bound on string length.
1930 (handle_builtin_string_cmp): Add an argument. Adjust.
1931 (strlen_check_and_optimize_call): Pass additional argument to
1932 handle_builtin_string_cmp.
1934 2020-02-05 Uroš Bizjak <ubizjak@gmail.com>
1936 * config/i386/i386.md (*pushdi2_rex64 peephole2): Remove.
1937 (*pushdi2_rex64 peephole2): Unconditionally split after
1939 (*ashl<mode>3_doubleword): Ditto.
1940 (*<shift_insn><mode>3_doubleword): Ditto.
1942 2020-02-05 Michael Meissner <meissner@linux.ibm.com>
1945 * config/rs6000/rs6000.c (get_vector_offset): Fix
1947 2020-02-05 Andrew Stubbs <ams@codesourcery.com>
1949 * config/gcn/t-gcn-hsa (MULTILIB_OPTIONS): Use / not space.
1951 2020-02-05 David Malcolm <dmalcolm@redhat.com>
1954 (Special Functions for Debugging the Analyzer): Update description
1955 of __analyzer_dump_exploded_nodes.
1957 2020-02-05 Jakub Jelinek <jakub@redhat.com>
1960 * config/i386/i386-features.c (ix86_add_reg_usage_to_vzeroupper): Only
1961 include sets and not clobbers in the vzeroupper pattern.
1962 * config/i386/sse.md (*avx_vzeroupper): Require in insn condition that
1963 the parallel has 17 (64-bit) or 9 (32-bit) elts.
1964 (*avx_vzeroupper_1): New define_insn_and_split.
1967 * recog.c (pass_split_after_reload::gate): For STACK_REGS targets,
1968 don't run when !optimize.
1969 (pass_split_before_regstack::gate): For STACK_REGS targets, run even
1972 2020-02-05 Richard Biener <rguenther@suse.de>
1975 * genmatch.c (dt_node::gen_kids_1): Emit number of argument
1976 checks before matching calls.
1978 2020-02-05 Jakub Jelinek <jakub@redhat.com>
1980 * tree-ssa-alias.c (aliasing_matching_component_refs_p): Fix up
1981 function comment typo.
1984 * omp-simd-clone.c (expand_simd_clones): If simd_clone_mangle or
1985 simd_clone_create failed when i == 0, adjust clone->nargs by
1988 2020-02-05 Martin Liska <mliska@suse.cz>
1991 * doc/invoke.texi: Document that one should
1992 not combine ASLR and -fpch.
1994 2020-02-04 Richard Biener <rguenther@suse.de>
1996 PR tree-optimization/93538
1997 * match.pd (addr EQ/NE ptr): Amend to handle &ptr->x EQ/NE ptr.
1999 2020-02-04 Richard Biener <rguenther@suse.de>
2001 PR tree-optimization/91123
2002 * tree-ssa-sccvn.c (vn_walk_cb_data::finish): New method.
2003 (vn_walk_cb_data::last_vuse): New member.
2004 (vn_walk_cb_data::saved_operands): Likewsie.
2005 (vn_walk_cb_data::~vn_walk_cb_data): Release saved_operands.
2006 (vn_walk_cb_data::push_partial_def): Use finish.
2007 (vn_reference_lookup_2): Update last_vuse and use finish if
2008 we've saved operands.
2009 (vn_reference_lookup_3): Use finish and update calls to
2010 push_partial_defs everywhere. When translating through
2011 memcpy or aggregate copies save off operands and alias-set.
2012 (eliminate_dom_walker::eliminate_stmt): Restore VN_WALKREWRITE
2013 operation for redundant store removal.
2015 2020-02-04 Richard Biener <rguenther@suse.de>
2017 PR tree-optimization/92819
2018 * tree-ssa-forwprop.c (simplify_vector_constructor): Avoid
2019 generating more stmts than before.
2021 2020-02-04 Martin Liska <mliska@suse.cz>
2023 * config/arm/arm.c (arm_gen_far_branch): Move the function
2024 outside of selftests.
2026 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
2028 * config/rs6000/rs6000.c (adjust_vec_address_pcrel): New helper
2029 function to adjust PC-relative vector addresses.
2030 (rs6000_adjust_vec_address): Call adjust_vec_address_pcrel to
2031 handle vectors with PC-relative addresses.
2033 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
2035 * config/rs6000/rs6000.c (reg_to_non_prefixed): Add forward
2037 (hard_reg_and_mode_to_addr_mask): Delete.
2038 (rs6000_adjust_vec_address): If the original vector address
2039 was REG+REG or REG+OFFSET and the element is not zero, do the add
2040 of the elements in the original address before adding the offset
2041 for the vector element. Use address_to_insn_form to validate the
2042 address using the register being loaded, rather than guessing
2043 whether the address is a DS-FORM or DQ-FORM address.
2045 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
2047 * config/rs6000/rs6000.c (get_vector_offset): New helper function
2048 to calculate the offset in memory from the start of a vector of a
2049 particular element. Add code to keep the element number in
2050 bounds if the element number is variable.
2051 (rs6000_adjust_vec_address): Move calculation of offset of the
2052 vector element to get_vector_offset.
2053 (rs6000_split_vec_extract_var): Do not do the initial AND of
2054 element here, move the code to get_vector_offset.
2056 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
2058 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add some
2061 2020-02-03 Segher Boessenkool <segher@kernel.crashing.org>
2063 * config/rs6000/constraints.md: Improve documentation.
2065 2020-02-03 Richard Earnshaw <rearnsha@arm.com>
2068 * config/arm/t-arm: ($(srcdir)/config/arm/arm-tune.md)
2069 ($(srcdir)/config/arm/arm-tables.opt): Use move-if-change.
2071 2020-02-03 Andrew Stubbs <ams@codesourcery.com>
2073 * config.gcc: Remove "carrizo" support.
2074 * config/gcn/gcn-opts.h (processor_type): Likewise.
2075 * config/gcn/gcn.c (gcn_omp_device_kind_arch_isa): Likewise.
2076 * config/gcn/gcn.opt (gpu_type): Likewise.
2077 * config/gcn/t-omp-device: Likewise.
2079 2020-02-03 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2082 * config/arm/arm-protos.h: New function arm_gen_far_branch prototype.
2083 * config/arm/arm.c (arm_gen_far_branch): New function
2085 * config/arm/arm.md: Update b<cond> for Thumb2 range checks.
2087 2020-02-03 Julian Brown <julian@codesourcery.com>
2088 Tobias Burnus <tobias@codesourcery.com>
2090 * doc/invoke.texi: Update mention of OpenACC version to 2.6.
2092 2020-02-03 Jakub Jelinek <jakub@redhat.com>
2095 * config/s390/s390.md (popcounthi2_z196): Fix up expander to emit
2096 valid RTL to sum up the lowest and second lowest bytes of the popcnt
2099 2020-02-02 Vladimir Makarov <vmakarov@redhat.com>
2101 PR rtl-optimization/91333
2102 * ira-color.c (struct allocno_color_data): Add member
2104 (init_allocno_threads): Set the member up.
2105 (bucket_allocno_compare_func): Add compare hard reg
2108 2020-01-31 Sandra Loosemore <sandra@codesourcery.com>
2110 nios2: Support for GOT-relative DW_EH_PE_datarel encoding.
2112 * configure.ac [nios2-*-*]: Check HAVE_AS_NIOS2_GOTOFF_RELOCATION.
2113 * config.in: Regenerated.
2114 * configure: Regenerated.
2115 * config/nios2/nios2.h (ASM_PREFERRED_EH_DATA_FORMAT): Fix handling
2116 for PIC when HAVE_AS_NIOS2_GOTOFF_RELOCATION.
2117 (ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX): New.
2119 2020-02-01 Andrew Burgess <andrew.burgess@embecosm.com>
2121 * configure: Regenerate.
2123 2020-01-31 Vladimir Makarov <vmakarov@redhat.com>
2125 PR rtl-optimization/91333
2126 * ira-color.c (bucket_allocno_compare_func): Move conflict hard
2127 reg preferences comparison up.
2129 2020-01-31 Richard Sandiford <richard.sandiford@arm.com>
2131 * config/aarch64/aarch64.h (TARGET_SVE_BF16): New macro.
2132 * config/aarch64/aarch64-sve-builtins-sve2.h (svcvtnt): Move to
2133 aarch64-sve-builtins-base.h.
2134 * config/aarch64/aarch64-sve-builtins-sve2.cc (svcvtnt): Move to
2135 aarch64-sve-builtins-base.cc.
2136 * config/aarch64/aarch64-sve-builtins-base.h (svbfdot, svbfdot_lane)
2137 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
2139 * config/aarch64/aarch64-sve-builtins-base.cc (svbfdot, svbfdot_lane)
2140 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
2141 (svcvtnt): New functions.
2142 * config/aarch64/aarch64-sve-builtins-base.def (svbfdot, svbfdot_lane)
2143 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
2144 (svcvtnt): New functions.
2145 (svcvt): Add a form that converts f32 to bf16.
2146 * config/aarch64/aarch64-sve-builtins-shapes.h (ternary_bfloat)
2147 (ternary_bfloat_lane, ternary_bfloat_lanex2, ternary_bfloat_opt_n):
2149 * config/aarch64/aarch64-sve-builtins-shapes.cc (parse_element_type):
2150 Treat B as bfloat16_t.
2151 (ternary_bfloat_lane_base): New class.
2152 (ternary_bfloat_def): Likewise.
2153 (ternary_bfloat): New shape.
2154 (ternary_bfloat_lane_def): New class.
2155 (ternary_bfloat_lane): New shape.
2156 (ternary_bfloat_lanex2_def): New class.
2157 (ternary_bfloat_lanex2): New shape.
2158 (ternary_bfloat_opt_n_def): New class.
2159 (ternary_bfloat_opt_n): New shape.
2160 * config/aarch64/aarch64-sve-builtins.cc (TYPES_cvt_bfloat): New macro.
2161 * config/aarch64/aarch64-sve.md (@aarch64_sve_<sve_fp_op>vnx4sf)
2162 (@aarch64_sve_<sve_fp_op>_lanevnx4sf): New patterns.
2163 (@aarch64_sve_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>)
2164 (@cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
2165 (*cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
2166 (@aarch64_sve_cvtnt<VNx8BF_ONLY:mode>): Likewise.
2167 * config/aarch64/aarch64-sve2.md (@aarch64_sve2_cvtnt<mode>): Key
2168 the pattern off the narrow mode instead of the wider one.
2169 * config/aarch64/iterators.md (VNx8BF_ONLY): New mode iterator.
2170 (UNSPEC_BFMLALB, UNSPEC_BFMLALT, UNSPEC_BFMMLA): New unspecs.
2171 (sve_fp_op): Handle them.
2172 (SVE_BFLOAT_TERNARY_LONG): New int itertor.
2173 (SVE_BFLOAT_TERNARY_LONG_LANE): Likewise.
2175 2020-01-31 Richard Sandiford <richard.sandiford@arm.com>
2177 * config/aarch64/arm_sve.h: Include arm_bf16.h.
2178 * config/aarch64/aarch64-modes.def (BF): Move definition before
2179 VECTOR_MODES. Remove separate VECTOR_MODES for V4BF and V8BF.
2180 (SVE_MODES): Handle BF modes.
2181 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
2183 (aarch64_full_sve_mode): Likewise.
2184 * config/aarch64/iterators.md (SVE_STRUCT): Add VNx16BF, VNx24BF
2186 (SVE_FULL, SVE_FULL_HSD, SVE_ALL): Add VNx8BF.
2187 (Vetype, Vesize, Vctype, VEL, Vel, VEL_INT, V128, v128, vwcore)
2188 (V_INT_EQUIV, v_int_equiv, V_FP_EQUIV, v_fp_equiv, vector_count)
2189 (insn_length, VSINGLE, vsingle, VPRED, vpred, VDOUBLE): Handle the
2191 * config/aarch64/aarch64-sve-builtins.h (TYPE_bfloat): New
2193 * config/aarch64/aarch64-sve-builtins.cc (TYPES_all_arith): New macro.
2194 (TYPES_all_data): Add bf16.
2195 (TYPES_reinterpret1, TYPES_reinterpret): Likewise.
2196 (register_tuple_type): Increase buffer size.
2197 * config/aarch64/aarch64-sve-builtins.def (svbfloat16_t): New type.
2198 (bf16): New type suffix.
2199 * config/aarch64/aarch64-sve-builtins-base.def (svabd, svadd, svaddv)
2200 (svcmpeq, svcmpge, svcmpgt, svcmple, svcmplt, svcmpne, svmad, svmax)
2201 (svmaxv, svmin, svminv, svmla, svmls, svmsb, svmul, svsub, svsubr):
2202 Change type from all_data to all_arith.
2203 * config/aarch64/aarch64-sve-builtins-sve2.def (svaddp, svmaxp)
2206 2020-01-31 Dennis Zhang <dennis.zhang@arm.com>
2207 Matthew Malcomson <matthew.malcomson@arm.com>
2208 Richard Sandiford <richard.sandiford@arm.com>
2210 * doc/invoke.texi (f32mm): Document new AArch64 -march= extension.
2211 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Define
2212 __ARM_FEATURE_SVE_MATMUL_INT8, __ARM_FEATURE_SVE_MATMUL_FP32 and
2213 __ARM_FEATURE_SVE_MATMUL_FP64 as appropriate. Don't define
2214 __ARM_FEATURE_MATMUL_FP64.
2215 * config/aarch64/aarch64-option-extensions.def (fp, simd, fp16)
2216 (sve): Add AARCH64_FL_F32MM to the list of extensions that should
2217 be disabled at the same time.
2218 (f32mm): New extension.
2219 * config/aarch64/aarch64.h (AARCH64_FL_F32MM): New macro.
2220 (AARCH64_FL_F64MM): Bump to the next bit up.
2221 (AARCH64_ISA_F32MM, TARGET_SVE_I8MM, TARGET_F32MM, TARGET_SVE_F32MM)
2222 (TARGET_SVE_F64MM): New macros.
2223 * config/aarch64/iterators.md (SVE_MATMULF): New mode iterator.
2224 (UNSPEC_FMMLA, UNSPEC_SMATMUL, UNSPEC_UMATMUL, UNSPEC_USMATMUL)
2225 (UNSPEC_TRN1Q, UNSPEC_TRN2Q, UNSPEC_UZP1Q, UNSPEC_UZP2Q, UNSPEC_ZIP1Q)
2226 (UNSPEC_ZIP2Q): New unspeccs.
2227 (DOTPROD_US_ONLY, PERMUTEQ, MATMUL, FMMLA): New int iterators.
2228 (optab, sur, perm_insn): Handle the new unspecs.
2229 (sve_fp_op): Handle UNSPEC_FMMLA. Resort.
2230 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use
2231 TARGET_SVE_F64MM instead of separate tests.
2232 (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod<vsi2qi>): New pattern.
2233 (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod_lane<vsi2qi>): Likewise.
2234 (@aarch64_sve_add_<MATMUL:optab><vsi2qi>): Likewise.
2235 (@aarch64_sve_<FMMLA:sve_fp_op><mode>): Likewise.
2236 (@aarch64_sve_<PERMUTEQ:optab><mode>): Likewise.
2237 * config/aarch64/aarch64-sve-builtins.cc (TYPES_s_float): New macro.
2238 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): Use it.
2239 (TYPES_s_signed): New macro.
2240 (TYPES_s_integer): Use it.
2241 (TYPES_d_float): New macro.
2242 (TYPES_d_data): Use it.
2243 * config/aarch64/aarch64-sve-builtins-shapes.h (mmla): Declare.
2244 (ternary_intq_uintq_lane, ternary_intq_uintq_opt_n, ternary_uintq_intq)
2245 (ternary_uintq_intq_lane, ternary_uintq_intq_opt_n): Likewise.
2246 * config/aarch64/aarch64-sve-builtins-shapes.cc (mmla_def): New class.
2247 (svmmla): New shape.
2248 (ternary_resize2_opt_n_base): Add TYPE_CLASS2 and TYPE_CLASS3
2249 template parameters.
2250 (ternary_resize2_lane_base): Likewise.
2251 (ternary_resize2_base): New class.
2252 (ternary_qq_lane_base): Likewise.
2253 (ternary_intq_uintq_lane_def): Likewise.
2254 (ternary_intq_uintq_lane): New shape.
2255 (ternary_intq_uintq_opt_n_def): New class
2256 (ternary_intq_uintq_opt_n): New shape.
2257 (ternary_qq_lane_def): Inherit from ternary_qq_lane_base.
2258 (ternary_uintq_intq_def): New class.
2259 (ternary_uintq_intq): New shape.
2260 (ternary_uintq_intq_lane_def): New class.
2261 (ternary_uintq_intq_lane): New shape.
2262 (ternary_uintq_intq_opt_n_def): New class.
2263 (ternary_uintq_intq_opt_n): New shape.
2264 * config/aarch64/aarch64-sve-builtins-base.h (svmmla, svsudot)
2265 (svsudot_lane, svtrn1q, svtrn2q, svusdot, svusdot_lane, svusmmla)
2266 (svuzp1q, svuzp2q, svzip1q, svzip2q): Declare.
2267 * config/aarch64/aarch64-sve-builtins-base.cc (svdot_lane_impl):
2269 (svdotprod_lane_impl): ...this new class.
2270 (svmmla_impl, svusdot_impl): New classes.
2271 (svdot_lane): Update to use svdotprod_lane_impl.
2272 (svmmla, svsudot, svsudot_lane, svtrn1q, svtrn2q, svusdot)
2273 (svusdot_lane, svusmmla, svuzp1q, svuzp2q, svzip1q, svzip2q): New
2275 * config/aarch64/aarch64-sve-builtins-base.def (svmmla): New base
2276 function, with no types defined.
2277 (svmmla, svusmmla, svsudot, svsudot_lane, svusdot, svusdot_lane): New
2278 AARCH64_FL_I8MM functions.
2279 (svmmla): New AARCH64_FL_F32MM function.
2280 (svld1ro): Depend only on AARCH64_FL_F64MM, not on AARCH64_FL_V8_6.
2281 (svmmla, svtrn1q, svtrn2q, svuz1q, svuz2q, svzip1q, svzip2q): New
2282 AARCH64_FL_F64MM function.
2283 (REQUIRED_EXTENSIONS):
2285 2020-01-31 Andrew Stubbs <ams@codesourcery.com>
2287 * config/gcn/gcn-valu.md (addv64di3_exec): Allow one '0' in each
2290 2020-01-31 Uroš Bizjak <ubizjak@gmail.com>
2292 * config/i386/i386.md (*movoi_internal_avx): Do not check for
2293 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL. Remove MODE_V8SF handling.
2294 (*movti_internal): Do not check for
2295 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
2296 (*movtf_internal): Move check for TARGET_SSE2 and size optimization
2297 just after check for TARGET_AVX.
2298 (*movdf_internal): Ditto.
2299 * config/i386/mmx.md (*mov<mode>_internal): Do not check for
2300 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
2301 * config/i386/sse.md (mov<mode>_internal): Only check
2302 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL with V2DFmode. Move check
2303 for TARGET_SSE2 and size optimization just after check for TARGET_AVX.
2304 (<sse>_andnot<mode>3<mask_name>): Move check for
2305 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL after check for TARGET_AVX.
2306 (<code><mode>3<mask_name>): Ditto.
2307 (*andnot<mode>3): Ditto.
2308 (*andnottf3): Ditto.
2309 (*<code><mode>3): Ditto.
2310 (*<code>tf3): Ditto.
2311 (*andnot<VI:mode>3): Remove
2312 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL handling.
2313 (<mask_codefor><code><VI48_AVX_AVX512F:mode>3<mask_name>): Ditto.
2314 (*<code><VI12_AVX_AVX512F:mode>3): Ditto.
2315 (sse4_1_blendv<ssemodesuffix>): Ditto.
2316 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL):
2317 Explain that tune applies to 128bit instructions only.
2319 2020-01-31 Kwok Cheung Yeung <kcy@codesourcery.com>
2321 * config/gcn/mkoffload.c (process_asm): Add sgpr_count and vgpr_count
2322 to definition of hsa_kernel_description. Parse assembly to find SGPR
2323 and VGPR count of kernel and store in hsa_kernel_description.
2325 2020-01-31 Tamar Christina <tamar.christina@arm.com>
2327 PR rtl-optimization/91838
2328 * simplify-rtx.c (simplify_binary_operation_1): Update LSHIFTRT case
2329 to truncate if allowed or reject combination.
2331 2020-01-31 Andrew Stubbs <ams@codesourcery.com>
2333 * tree-ssa-loop-ivopts.c (get_iv): Use sizetype for zero-step.
2334 (find_inv_vars_cb): Likewise.
2336 2020-01-31 David Malcolm <dmalcolm@redhat.com>
2338 * calls.c (special_function_p): Split out the check for DECL_NAME
2339 being non-NULL and fndecl being extern at file scope into a
2340 new maybe_special_function_p and call it. Drop check for fndecl
2341 being non-NULL that was after a usage of DECL_NAME (fndecl).
2342 * tree.h (maybe_special_function_p): New inline function.
2344 2020-01-30 Andrew Stubbs <ams@codesourcery.com>
2346 * config/gcn/gcn-valu.md (gather<mode>_exec): Move contents ...
2347 (mask_gather_load<mode>): ... here, and zero-initialize the
2349 (maskload<mode>di): Zero-initialize the destination.
2352 2020-01-30 David Malcolm <dmalcolm@redhat.com>
2355 * doc/analyzer.texi (Limitations): Note that constraints on
2356 floating-point values are currently ignored.
2358 2020-01-30 Jakub Jelinek <jakub@redhat.com>
2361 * symtab.c (symtab_node::noninterposable_alias): If localalias
2362 already exists, but is not usable, append numbers after it until
2363 a unique name is found. Formatting fix.
2366 * combine.c (simplify_comparison) <case ROTATE>: Punt on out of range
2369 2020-01-30 Andrew Stubbs <ams@codesourcery.com>
2371 * config/gcn/gcn.c (print_operand): Handle LTGT.
2372 * config/gcn/predicates.md (gcn_fp_compare_operator): Allow ltgt.
2374 2020-01-30 Richard Biener <rguenther@suse.de>
2376 * tree-pretty-print.c (dump_generic_node): Wrap VECTOR_CST
2377 and CONSTRUCTOR in _Literal (type) with TDF_GIMPLE.
2379 2020-01-30 John David Anglin <danglin@gcc.gnu.org>
2381 * config/pa/pa.c (pa_elf_select_rtx_section): Place function pointers
2382 without a DECL in .data.rel.ro.local.
2384 2020-01-30 Jakub Jelinek <jakub@redhat.com>
2387 * config/arm/arm.md (uaddvdi4): Actually emit what gen_uaddvsi4
2391 * config/i386/sse.md
2392 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext): Renamed to ...
2393 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext): ... this. Use
2394 any_extend code iterator instead of always zero_extend.
2395 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_lt): Renamed to ...
2396 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_lt): ... this.
2397 Use any_extend code iterator instead of always zero_extend.
2398 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_shift): Renamed to ...
2399 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_shift): ... this.
2400 Use any_extend code iterator instead of always zero_extend.
2401 (*sse2_pmovmskb_ext): New define_insn.
2402 (*sse2_pmovmskb_ext_lt): New define_insn_and_split.
2405 * config/i386/i386.md (*popcountsi2_zext): New define_insn_and_split.
2406 (*popcountsi2_zext_falsedep): New define_insn.
2408 2020-01-30 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
2410 * config.in: Regenerated.
2411 * configure: Regenerated.
2413 2020-01-29 Tobias Burnus <tobias@codesourcery.com>
2416 * config/gcn/gcn-hsa.h (ASM_SPEC): Add -mattr=-code-object-v3 as
2417 LLVM's assembler changed the default in version 9.
2419 2020-01-24 Jeff Law <law@redhat.com>
2421 PR tree-optimization/89689
2422 * builtins.def (BUILT_IN_OBJECT_SIZE): Make it const rather than pure.
2424 2020-01-29 Richard Sandiford <richard.sandiford@arm.com>
2428 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
2430 PR rtl-optimization/87763
2431 * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
2432 simplification to handle subregs as well as bare regs.
2433 * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
2435 2020-01-29 Joel Hutton <Joel.Hutton@arm.com>
2438 * ira.c (ira): Revert use of simplified LRA algorithm.
2440 2020-01-29 Martin Jambor <mjambor@suse.cz>
2442 PR tree-optimization/92706
2443 * tree-sra.c (struct access): Fields first_link, last_link,
2444 next_queued and grp_queued renamed to first_rhs_link, last_rhs_link,
2445 next_rhs_queued and grp_rhs_queued respectively, new fields
2446 first_lhs_link, last_lhs_link, next_lhs_queued and grp_lhs_queued.
2447 (struct assign_link): Field next renamed to next_rhs, new field
2448 next_lhs. Updated comment.
2449 (work_queue_head): Renamed to rhs_work_queue_head.
2450 (lhs_work_queue_head): New variable.
2451 (add_link_to_lhs): New function.
2452 (relink_to_new_repr): Also relink LHS lists.
2453 (add_access_to_work_queue): Renamed to add_access_to_rhs_work_queue.
2454 (add_access_to_lhs_work_queue): New function.
2455 (pop_access_from_work_queue): Renamed to
2456 pop_access_from_rhs_work_queue.
2457 (pop_access_from_lhs_work_queue): New function.
2458 (build_accesses_from_assign): Also add links to LHS lists and to LHS
2460 (child_would_conflict_in_lacc): Renamed to
2461 child_would_conflict_in_acc. Adjusted parameter names.
2462 (create_artificial_child_access): New parameter set_grp_read, use it.
2463 (subtree_mark_written_and_enqueue): Renamed to
2464 subtree_mark_written_and_rhs_enqueue.
2465 (propagate_subaccesses_across_link): Renamed to
2466 propagate_subaccesses_from_rhs.
2467 (propagate_subaccesses_from_lhs): New function.
2468 (propagate_all_subaccesses): Also propagate subaccesses from LHSs to
2471 2020-01-29 Martin Jambor <mjambor@suse.cz>
2473 PR tree-optimization/92706
2474 * tree-sra.c (struct access): Adjust comment of
2475 grp_total_scalarization.
2476 (find_access_in_subtree): Look for single children spanning an entire
2478 (scalarizable_type_p): Allow register accesses, adjust callers.
2479 (completely_scalarize): Remove function.
2480 (scalarize_elem): Likewise.
2481 (create_total_scalarization_access): Likewise.
2482 (sort_and_splice_var_accesses): Do not track total scalarization
2484 (analyze_access_subtree): New parameter totally, adjust to new meaning
2485 of grp_total_scalarization.
2486 (analyze_access_trees): Pass new parameter to analyze_access_subtree.
2487 (can_totally_scalarize_forest_p): New function.
2488 (create_total_scalarization_access): Likewise.
2489 (create_total_access_and_reshape): Likewise.
2490 (total_should_skip_creating_access): Likewise.
2491 (totally_scalarize_subtree): Likewise.
2492 (analyze_all_variable_accesses): Perform total scalarization after
2493 subaccess propagation using the new functions above.
2494 (initialize_constant_pool_replacements): Output initializers by
2495 traversing the access tree.
2497 2020-01-29 Martin Jambor <mjambor@suse.cz>
2499 * tree-sra.c (verify_sra_access_forest): New function.
2500 (verify_all_sra_access_forests): Likewise.
2501 (create_artificial_child_access): Set parent.
2502 (analyze_all_variable_accesses): Call the verifier.
2504 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
2506 * cgraph.c (cgraph_edge::resolve_speculation): Only lookup direct edge
2507 if called on indirect edge.
2508 (cgraph_edge::redirect_call_stmt_to_callee): Lookup indirect edge of
2509 speculative call if needed.
2511 2020-01-29 Richard Biener <rguenther@suse.de>
2513 PR tree-optimization/93428
2514 * tree-vect-slp.c (vect_build_slp_tree_2): Compute the load
2515 permutation when the load node is created.
2516 (vect_analyze_slp_instance): Re-use it here.
2518 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
2520 * ipa-prop.c (update_indirect_edges_after_inlining): Fix warning.
2522 2020-01-28 Vladimir Makarov <vmakarov@redhat.com>
2524 PR rtl-optimization/93272
2525 * ira-lives.c (process_out_of_region_eh_regs): New function.
2526 (process_bb_node_lives): Call it.
2528 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
2530 * coverage.c (read_counts_file): Make error message lowercase.
2532 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
2534 * profile-count.c (profile_quality_display_names): Fix ordering.
2536 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
2539 * cgraph.c (cgraph_add_edge_to_call_site_hash): Update call site
2540 hash only when edge is first within the sequence.
2541 (cgraph_edge::set_call_stmt): Update handling of speculative calls.
2542 (symbol_table::create_edge): Do not set target_prob.
2543 (cgraph_edge::remove_caller): Watch for speculative calls when updating
2545 (cgraph_edge::make_speculative): Drop target_prob parameter.
2546 (cgraph_edge::speculative_call_info): Remove.
2547 (cgraph_edge::first_speculative_call_target): New member function.
2548 (update_call_stmt_hash_for_removing_direct_edge): New function.
2549 (cgraph_edge::resolve_speculation): Rewrite to new API.
2550 (cgraph_edge::speculative_call_for_target): New member function.
2551 (cgraph_edge::make_direct): Rewrite to new API; fix handling of
2552 multiple speculation targets.
2553 (cgraph_edge::redirect_call_stmt_to_callee): Likewise; fix updating
2555 (verify_speculative_call): Verify that targets form an interval.
2556 * cgraph.h (cgraph_edge::speculative_call_info): Remove.
2557 (cgraph_edge::first_speculative_call_target): New member function.
2558 (cgraph_edge::next_speculative_call_target): New member function.
2559 (cgraph_edge::speculative_call_target_ref): New member function.
2560 (cgraph_edge;:speculative_call_indirect_edge): New member funtion.
2561 (cgraph_edge): Remove target_prob.
2562 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
2563 Fix handling of speculative calls.
2564 * ipa-devirt.c (ipa_devirt): Fix handling of speculative cals.
2565 * ipa-fnsummary.c (analyze_function_body): Likewise.
2566 * ipa-inline.c (speculation_useful_p): Use new speculative call API.
2567 * ipa-profile.c (dump_histogram): Fix formating.
2568 (ipa_profile_generate_summary): Watch for overflows.
2569 (ipa_profile): Do not require probablity to be 1/2; update to new API.
2570 * ipa-prop.c (ipa_make_edge_direct_to_target): Update to new API.
2571 (update_indirect_edges_after_inlining): Update to new API.
2572 * ipa-utils.c (ipa_merge_profiles): Rewrite merging of speculative call
2574 * profile-count.h: (profile_probability::adjusted): New.
2575 * tree-inline.c (copy_bb): Update to new speculative call API; fix
2576 updating of profile.
2577 * value-prof.c (gimple_ic_transform): Rename to ...
2578 (dump_ic_profile): ... this one; update dumping.
2579 (stream_in_histogram_value): Fix formating.
2580 (gimple_value_profile_transformations): Update.
2582 2020-01-28 H.J. Lu <hongjiu.lu@intel.com>
2585 * config/i386/i386.md (*movoi_internal_avx): Remove
2586 TARGET_SSE_TYPELESS_STORES check.
2587 (*movti_internal): Prefer TARGET_AVX over
2588 TARGET_SSE_TYPELESS_STORES.
2589 (*movtf_internal): Likewise.
2590 * config/i386/sse.md (mov<mode>_internal): Prefer TARGET_AVX over
2591 TARGET_SSE_TYPELESS_STORES. Remove "<MODE_SIZE> == 16" check
2592 from TARGET_SSE_TYPELESS_STORES.
2594 2020-01-28 David Malcolm <dmalcolm@redhat.com>
2596 * diagnostic-core.h (warning_at): Rename overload to...
2597 (warning_meta): ...this.
2598 (emit_diagnostic_valist): Delete decl of overload taking
2599 diagnostic_metadata.
2600 * diagnostic.c (emit_diagnostic_valist): Likewise for defn.
2601 (warning_at): Rename overload taking diagnostic_metadata to...
2602 (warning_meta): ...this.
2604 2020-01-28 Richard Biener <rguenther@suse.de>
2606 PR tree-optimization/93439
2607 * tree-parloops.c (create_loop_fn): Move clique bookkeeping...
2608 * tree-cfg.c (move_sese_region_to_fn): ... here.
2609 (verify_types_in_gimple_reference): Verify used cliques are
2612 2020-01-28 H.J. Lu <hongjiu.lu@intel.com>
2615 * config/i386/i386-options.c (set_ix86_tune_features): Add an
2616 argument of a pointer to struct gcc_options and pass it to
2617 parse_mtune_ctrl_str.
2618 (ix86_function_specific_restore): Pass opts to
2619 set_ix86_tune_features.
2620 (ix86_option_override_internal): Likewise.
2621 (parse_mtune_ctrl_str): Add an argument of a pointer to struct
2622 gcc_options and use it for x_ix86_tune_ctrl_string.
2624 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
2626 PR rtl-optimization/87763
2627 * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
2628 simplification to handle subregs as well as bare regs.
2629 * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
2631 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
2633 * tree-vect-loop.c (vectorizable_reduction): Fail gracefully
2634 for reduction chains that (now) include a call.
2636 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
2638 PR tree-optimization/92822
2639 * tree-ssa-forwprop.c (simplify_vector_constructor): When filling
2640 out the don't-care elements of a vector whose significant elements
2641 are duplicates, make the don't-care elements duplicates too.
2643 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
2645 PR tree-optimization/93434
2646 * tree-predcom.c (split_data_refs_to_components): Record which
2647 components have had aliasing loads removed. Prevent store-store
2648 commoning for all such components.
2650 2020-01-28 Jakub Jelinek <jakub@redhat.com>
2653 * config/i386/i386.c (ix86_fold_builtin) <do_shift>: If mask is not
2654 -1 or is_vshift is true, use new_vector with number of elts npatterns
2655 rather than new_unary_operation.
2657 PR tree-optimization/93454
2658 * gimple-fold.c (fold_array_ctor_reference): Perform
2659 elt_size.to_uhwi () just once, instead of calling it in every
2660 iteration. Punt if that value is above size of the temporary
2661 buffer. Decrease third native_encode_expr argument when
2662 bufoff + elt_sz is above size of buf.
2664 2020-01-27 Joseph Myers <joseph@codesourcery.com>
2666 * config/mips/mips.c (mips_declare_object_name)
2667 [USE_GNU_UNIQUE_OBJECT]: Support use of gnu_unique_object.
2669 2020-01-27 Martin Liska <mliska@suse.cz>
2671 PR gcov-profile/93403
2672 * tree-profile.c (gimple_init_gcov_profiler): Generate
2673 both __gcov_indirect_call_profiler_v4 and
2674 __gcov_indirect_call_profiler_v4_atomic.
2676 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
2679 * config/aarch64/aarch64-simd.md (aarch64_get_half<mode>): New
2681 (@aarch64_split_simd_mov<mode>): Use it.
2682 (aarch64_simd_mov_from_<mode>low): Add a GPR alternative.
2683 Leave the vec_extract patterns to handle 2-element vectors.
2684 (aarch64_simd_mov_from_<mode>high): Likewise.
2685 (vec_extract<VQMOV_NO2E:mode><Vhalf>): New expander.
2686 (vec_extractv2dfv1df): Likewise.
2688 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
2690 * config/aarch64/aarch64.c (aarch64_if_then_else_costs): Match
2691 jump conditions for *compare_condjump<GPI:mode>.
2693 2020-01-27 David Malcolm <dmalcolm@redhat.com>
2696 * digraph.cc (test_edge::test_edge): Specify template for base
2699 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
2701 * config/arc/arc.c (arc_rtx_costs): Update mul64 cost.
2703 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
2705 * config/arc/arc-protos.h (gen_mlo): Remove.
2706 (gen_mhi): Likewise.
2707 * config/arc/arc.c (AUX_MULHI): Define.
2708 (arc_must_save_reister): Special handling for r58/59.
2709 (arc_compute_frame_size): Consider mlo/mhi registers.
2710 (arc_save_callee_saves): Emit fp/sp move only when emit_move
2712 (arc_conditional_register_usage): Remove TARGET_BIG_ENDIAN from
2713 mlo/mhi name selection.
2714 (arc_restore_callee_saves): Don't early restore blink when ISR.
2715 (arc_expand_prologue): Add mlo/mhi saving.
2716 (arc_expand_epilogue): Add mlo/mhi restoring.
2719 * config/arc/arc.h (DBX_REGISTER_NUMBER): Correct register
2720 numbering when MUL64 option is used.
2721 (DWARF2_FRAME_REG_OUT): Define.
2722 * config/arc/arc.md (arc600_stall): New pattern.
2723 (VUNSPEC_ARC_ARC600_STALL): Define.
2724 (mulsi64): Use correct mlo/mhi registers.
2725 (mulsi_600): Clean it up.
2726 * config/arc/predicates.md (mlo_operand): Remove any dependency on
2728 (mhi_operand): Likewise.
2730 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
2731 Petro Karashchenko <petro.karashchenko@ring.com>
2733 * config/arc/arc.c (arc_is_uncached_mem_p): Check struct
2734 attributes if needed.
2735 (prepare_move_operands): Generate special unspec instruction for
2737 (arc_isuncached_mem_p): Propagate uncached attribute to each
2739 * config/arc/arc.md (VUNSPEC_ARC_LDDI): Define.
2740 (VUNSPEC_ARC_STDI): Likewise.
2741 (ALLI): New mode iterator.
2742 (mALLI): New mode attribute.
2743 (lddi): New instruction pattern.
2745 (stdidi_split): Split instruction for architectures which are not
2746 supporting ll64 option.
2747 (lddidi_split): Likewise.
2749 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
2751 PR rtl-optimization/92989
2752 * lra-lives.c (process_bb_lives): Update the live-in set before
2753 processing additional clobbers.
2755 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
2757 PR rtl-optimization/93170
2758 * cselib.c (cselib_invalidate_regno_val): New function, split out
2760 (cselib_invalidate_regno): ...here.
2761 (cselib_invalidated_by_call_p): New function.
2762 (cselib_process_insn): Iterate over all the hard-register entries in
2763 REG_VALUES and invalidate any that cross call-clobbered registers.
2765 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
2767 * dojump.c (split_comparison): Use HONOR_NANS rather than
2768 HONOR_SNANS when splitting LTGT.
2770 2020-01-27 Martin Liska <mliska@suse.cz>
2773 * opts.c (print_filtered_help): Exclude language-specific
2774 options from --help=common unless enabled in all FEs.
2776 2020-01-27 Martin Liska <mliska@suse.cz>
2778 * opts.c (print_help): Exclude params from
2779 all except --help=param.
2781 2020-01-27 Martin Liska <mliska@suse.cz>
2784 * config/i386/i386-features.c (make_resolver_func):
2785 Align the code with ppc64 target implementation.
2786 Do not generate a unique name for resolver function.
2788 2020-01-27 Richard Biener <rguenther@suse.de>
2790 PR tree-optimization/93397
2791 * tree-vect-slp.c (vect_analyze_slp_instance): Delay
2792 converted reduction chain SLP graph adjustment.
2794 2020-01-26 Marek Polacek <polacek@redhat.com>
2797 * sanopt.c (sanitize_rewrite_addressable_params): Avoid crash on
2800 2020-01-26 Jason Merrill <jason@redhat.com>
2803 * tree.c (verify_type_variant): Only verify TYPE_NEEDS_CONSTRUCTING
2806 2020-01-26 Darius Galis <darius.galis@cyberthorstudios.com>
2808 * config/rx/rx.md (setmemsi): Added rx_allow_string_insns constraint
2809 (rx_setmem): Likewise.
2811 2020-01-26 Jakub Jelinek <jakub@redhat.com>
2814 * config/i386/i386.md (*addv<dwi>4_doubleword, *subv<dwi>4_doubleword):
2815 Use nonimmediate_operand instead of x86_64_hilo_general_operand and
2816 drop <di> from constraint of last operand.
2819 * config/i386/sse.md (*avx_vperm_broadcast_<mode>): Disallow for
2820 TARGET_AVX2 and V4DFmode not in the split condition, but in the
2821 pattern condition, though allow { 0, 0, 0, 0 } broadcast always.
2823 2020-01-25 Feng Xue <fxue@os.amperecomputing.com>
2826 * ipa-cp.c (get_info_about_necessary_edges): Remove value
2829 2020-01-24 Jeff Law <law@redhat.com>
2831 PR tree-optimization/92788
2832 * tree-ssa-threadedge.c (thread_across_edge): Check EDGE_COMPLEX
2835 2020-01-24 Jakub Jelinek <jakub@redhat.com>
2838 * config/i386/sse.md (*avx_vperm_broadcast_v4sf,
2839 *avx_vperm_broadcast_<mode>,
2840 <sse2_avx_avx512f>_vpermil<mode><mask_name>,
2841 *<sse2_avx_avx512f>_vpermilp<mode><mask_name>):
2842 Move before avx2_perm<mode>/avx512f_perm<mode>.
2845 * simplify-rtx.c (simplify_const_unary_operation,
2846 simplify_const_binary_operation): Punt for mode precision above
2847 MAX_BITSIZE_MODE_ANY_INT.
2849 2020-01-24 Andrew Pinski <apinski@marvell.com>
2851 * config/arm/aarch-cost-tables.h (cortexa57_extra_costs): Change
2854 2020-01-24 Jeff Law <law@redhat.com>
2857 * config/h8300/h8300.c (h8300_print_operand): Only call byte_reg
2858 for REGs. Call output_operand_lossage to get more reasonable
2861 2020-01-24 Andrew Stubbs <ams@codesourcery.com>
2863 * config/gcn/gcn-valu.md (vec_cmp<mode>di): Use
2864 gcn_fp_compare_operator.
2865 (vec_cmpu<mode>di): Use gcn_compare_operator.
2866 (vec_cmp<u>v64qidi): Use gcn_compare_operator.
2867 (vec_cmp<mode>di_exec): Use gcn_fp_compare_operator.
2868 (vec_cmpu<mode>di_exec): Use gcn_compare_operator.
2869 (vec_cmp<u>v64qidi_exec): Use gcn_compare_operator.
2870 (vec_cmp<mode>di_dup): Use gcn_fp_compare_operator.
2871 (vec_cmp<mode>di_dup_exec): Use gcn_fp_compare_operator.
2872 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): Use
2873 gcn_fp_compare_operator.
2874 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): Use
2875 gcn_fp_compare_operator.
2876 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): Use
2877 gcn_fp_compare_operator.
2878 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): Use
2879 gcn_fp_compare_operator.
2881 2020-01-24 Maciej W. Rozycki <macro@wdc.com>
2883 * doc/install.texi (Cross-Compiler-Specific Options): Document
2884 `--with-toolexeclibdir' option.
2886 2020-01-24 Hans-Peter Nilsson <hp@axis.com>
2888 * target.def (flags_regnum): Also mention effect on delay slot filling.
2889 * doc/tm.texi: Regenerate.
2891 2020-01-23 Jeff Law <law@redhat.com>
2893 PR translation/90162
2894 * config/h8300/h8300.c (h8300_option_override): Fix diagnostic text.
2896 2020-01-23 Mikael Tillenius <mti-1@tillenius.com>
2899 * config/h8300/h8300.h (FUNCTION_PROFILER): Fix emission of
2902 2020-01-23 Jakub Jelinek <jakub@redhat.com>
2904 PR rtl-optimization/93402
2905 * postreload.c (reload_combine_recognize_pattern): Don't try to adjust
2908 2020-01-23 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
2910 * config.in: Regenerated.
2911 * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to 1
2912 for TARGET_LIBC_GNUSTACK.
2913 * configure: Regenerated.
2914 * configure.ac: Define TARGET_LIBC_GNUSTACK if glibc version is
2915 found to be 2.31 or greater.
2917 2020-01-23 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
2919 * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to
2921 * config/mips/mips.c (TARGET_ASM_FILE_END): Define to ...
2922 (mips_asm_file_end): New function. Delegate to
2923 file_end_indicate_exec_stack if NEED_INDICATE_EXEC_STACK is true.
2924 * config/mips/mips.h (NEED_INDICATE_EXEC_STACK): Define to 0.
2926 2020-01-23 Jakub Jelinek <jakub@redhat.com>
2929 * config/i386/i386-modes.def (POImode): New mode.
2930 (MAX_BITSIZE_MODE_ANY_INT): Change from 128 to 160.
2931 * config/i386/i386.md (DPWI): New mode attribute.
2932 (addv<mode>4, subv<mode>4): Use <DPWI> instead of <DWI>.
2934 (QPWI): ... this. Use POI instead of OI for TImode.
2935 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1,
2936 *subv<dwi>4_doubleword, *subv<dwi>4_doubleword_1): Use <QPWI>
2939 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
2942 * config/aarch64/aarch64.md (UNSPEC_SPECULATION_TRACKER_REV): New
2944 (speculation_tracker_rev): New pattern.
2945 * config/aarch64/aarch64-speculation.cc (aarch64_do_track_speculation):
2946 Use speculation_tracker_rev to track the inverse condition.
2948 2020-01-23 Richard Biener <rguenther@suse.de>
2950 PR tree-optimization/93381
2951 * tree-ssa-sccvn.c (vn_walk_cb_data::push_partial_def): Take
2952 alias-set of the def as argument and record the first one.
2953 (vn_walk_cb_data::first_set): New member.
2954 (vn_reference_lookup_3): Pass the alias-set of the current def
2955 to push_partial_def. Fix alias-set used in the aggregate copy
2957 (vn_reference_lookup): Consistently set *last_vuse_ptr.
2958 * real.c (clear_significand_below): Fix out-of-bound access.
2960 2020-01-23 Jakub Jelinek <jakub@redhat.com>
2963 * config/i386/i386.md (*bmi2_bzhi_<mode>3_2, *bmi2_bzhi_<mode>3_3):
2964 New define_insn patterns.
2966 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
2968 * doc/sourcebuild.texi (check-function-bodies): Add an
2969 optional target/xfail selector.
2971 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
2973 PR rtl-optimization/93124
2974 * auto-inc-dec.c (merge_in_block): Don't add auto inc/decs to
2975 bare USE and CLOBBER insns.
2977 2020-01-22 Andrew Pinski <apinski@marvell.com>
2979 * config/arc/arc.c (output_short_suffix): Check insn for nullness.
2981 2020-01-22 David Malcolm <dmalcolm@redhat.com>
2984 * gdbinit.in (break-on-saved-diagnostic): Update for move of
2985 diagnostic_manager into "ana" namespace.
2986 * selftest-run-tests.c (selftest::run_tests): Update for move of
2987 selftest::run_analyzer_selftests to
2988 ana::selftest::run_analyzer_selftests.
2990 2020-01-22 Richard Sandiford <richard.sandiford@arm.com>
2992 * cfgexpand.c (union_stack_vars): Update the size.
2994 2020-01-22 Richard Biener <rguenther@suse.de>
2996 PR tree-optimization/93381
2997 * tree-ssa-structalias.c (find_func_aliases): Assume offsetting
2998 throughout, handle all conversions the same.
3000 2020-01-22 Jakub Jelinek <jakub@redhat.com>
3003 * config/aarch64/aarch64.c (aarch64_expand_subvti): Only use
3004 gen_subdi3_compare1_imm if low_in2 satisfies aarch64_plus_immediate
3005 predicate, not whenever it is CONST_INT. Otherwise, force_reg it.
3006 Call force_reg on high_in2 unconditionally.
3008 2020-01-22 Martin Liska <mliska@suse.cz>
3010 PR tree-optimization/92924
3011 * profile.c (compute_value_histograms): Divide
3014 2020-01-22 Jakub Jelinek <jakub@redhat.com>
3017 * output.h (assemble_name_resolve): Declare.
3018 * varasm.c (assemble_name_resolve): New function.
3019 (assemble_name): Use it.
3020 * config/i386/i386.h (ASM_OUTPUT_SYMBOL_REF): Define.
3022 2020-01-22 Joseph Myers <joseph@codesourcery.com>
3024 * doc/sourcebuild.texi (Texinfo Manuals, Front End): Refer to
3025 update_web_docs_git instead of update_web_docs_svn.
3027 2020-01-21 Andrew Pinski <apinski@marvell.com>
3030 * config/aarch64/aarch64.md (tlsgd_small_<mode>): Have operand 0
3031 as PTR mode. Have operand 1 as being modeless, it can be P mode.
3032 (*tlsgd_small_<mode>): Likewise.
3033 * config/aarch64/aarch64.c (aarch64_load_symref_appropriately)
3034 <case SYMBOL_SMALL_TLSGD>: Call gen_tlsgd_small_* with a ptr_mode
3035 register. Convert that register back to dest using convert_mode.
3037 2020-01-21 Jim Wilson <jimw@sifive.com>
3039 * config/riscv/riscv-sr.c (riscv_sr_match_prologue): Use INTVAL
3042 2020-01-21 H.J. Lu <hongjiu.lu@intel.com>
3043 Uros Bizjak <ubizjak@gmail.com>
3046 * config/i386/i386.c (ix86_tls_module_base): Replace Pmode
3048 (legitimize_tls_address): Do GNU2 TLS address computation in
3049 ptr_mode and zero-extend result to Pmode.
3050 * config/i386/i386.md (@tls_dynamic_gnu2_64_<mode>): Replace
3051 :P with :PTR and Pmode with ptr_mode.
3052 (*tls_dynamic_gnu2_lea_64_<mode>): Likewise.
3053 (*tls_dynamic_gnu2_call_64_<mode>): Likewise.
3054 (*tls_dynamic_gnu2_combine_64_<mode>): Likewise.
3056 2020-01-21 Jakub Jelinek <jakub@redhat.com>
3059 * config/riscv/riscv.c (riscv_rtx_costs) <case ZERO_EXTRACT>: Verify
3060 the last two operands are CONST_INT_P before using them as such.
3062 2020-01-21 Richard Sandiford <richard.sandiford@arm.com>
3064 * config/aarch64/aarch64-sve-builtins.def: Use get_typenode_from_name
3065 to get the integer element types.
3067 2020-01-21 Richard Sandiford <richard.sandiford@arm.com>
3069 * config/aarch64/aarch64-sve-builtins.h
3070 (function_expander::convert_to_pmode): Declare.
3071 * config/aarch64/aarch64-sve-builtins.cc
3072 (function_expander::convert_to_pmode): New function.
3073 (function_expander::get_contiguous_base): Use it.
3074 (function_expander::prepare_gather_address_operands): Likewise.
3075 * config/aarch64/aarch64-sve-builtins-sve2.cc
3076 (svwhilerw_svwhilewr_impl::expand): Likewise.
3078 2020-01-21 Szabolcs Nagy <szabolcs.nagy@arm.com>
3081 * config/aarch64/aarch64.c (aarch64_declare_function_name): Set
3082 cfun->machine->label_is_assembled.
3083 (aarch64_print_patchable_function_entry): New.
3084 (TARGET_ASM_PRINT_PATCHABLE_FUNCTION_ENTRY): Define.
3085 * config/aarch64/aarch64.h (struct machine_function): New field,
3088 2020-01-21 David Malcolm <dmalcolm@redhat.com>
3091 * ipa-profile.c (ipa_profile): Delete call_sums and set it to
3094 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
3097 * cgraph.c (cgraph_edge::resolve_speculation,
3098 cgraph_edge::redirect_call_stmt_to_callee): Fix update of
3099 call_stmt_site_hash.
3101 2020-01-21 Martin Liska <mliska@suse.cz>
3103 * config/rs6000/rs6000.c (common_mode_defined): Remove
3106 2020-01-21 Richard Biener <rguenther@suse.de>
3108 PR tree-optimization/92328
3109 * tree-ssa-sccvn.c (vn_reference_lookup_3): Preserve
3110 type when value-numbering same-sized store by inserting a
3112 (eliminate_dom_walker::eliminate_stmt): When eliminating
3113 a redundant store handle bit-reinterpretation of the same value.
3115 2020-01-21 Andrew Pinski <apinski@marvel.com>
3118 * tree-into-ssa.c (prepare_block_for_update_1): Split out
3120 (prepare_block_for_update): This. Use a worklist instead of
3123 2020-01-21 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3125 * gcc/config/arm/arm.c (clear_operation_p):
3126 Initialise last_regno, skip first iteration
3127 based on the first_set value and use ints instead
3128 of the unnecessary HOST_WIDE_INTs.
3130 2020-01-21 Jakub Jelinek <jakub@redhat.com>
3133 * config/rs6000/rs6000.c (rs6000_emit_cmove): If using fsel, punt for
3134 compare_mode other than SFmode or DFmode.
3136 2020-01-21 Kito Cheng <kito.cheng@sifive.com>
3139 * config/riscv/riscv-protos.h (riscv_hard_regno_rename_ok): New.
3140 * config/riscv/riscv.c (riscv_hard_regno_rename_ok): New.
3141 * config/riscv/riscv.h (HARD_REGNO_RENAME_OK): Defined.
3143 2020-01-20 Wilco Dijkstra <wdijkstr@arm.com>
3145 * config/aarch64/aarch64.c (neoversen1_tunings): Set jump_align to 4.
3147 2020-01-20 Andrew Pinski <apinski@marvell.com>
3150 * targhooks.c (default_print_patchable_function_entry): Use
3151 output_asm_insn to emit the nop instruction.
3153 2020-01-20 Fangrui Song <maskray@google.com>
3156 * targhooks.c (default_print_patchable_function_entry): Align to
3159 2020-01-20 H.J. Lu <hongjiu.lu@intel.com>
3162 * config/i386/i386.c (legitimize_tls_address): Pass Pmode to
3163 gen_tls_dynamic_gnu2_64. Compute GNU2 TLS address in ptr_mode.
3164 * config/i386/i386.md (tls_dynamic_gnu2_64): Renamed to ...
3165 (@tls_dynamic_gnu2_64_<mode>): This. Replace DI with P.
3166 (*tls_dynamic_gnu2_lea_64): Renamed to ...
3167 (*tls_dynamic_gnu2_lea_64_<mode>): This. Replace DI with P.
3168 Remove the {q} suffix from lea.
3169 (*tls_dynamic_gnu2_call_64): Renamed to ...
3170 (*tls_dynamic_gnu2_call_64_<mode>): This. Replace DI with P.
3171 (*tls_dynamic_gnu2_combine_64): Renamed to ...
3172 (*tls_dynamic_gnu2_combine_64_<mode>): This. Replace DI with P.
3173 Pass Pmode to gen_tls_dynamic_gnu2_64.
3175 2020-01-20 Wilco Dijkstra <wdijkstr@arm.com>
3177 * config/aarch64/aarch64.h (SLOW_BYTE_ACCESS): Set to 1.
3179 2020-01-20 Richard Sandiford <richard.sandiford@arm.com>
3181 * config/aarch64/aarch64-sve-builtins-base.cc
3182 (svld1ro_impl::memory_vector_mode): Remove parameter name.
3184 2020-01-20 Richard Biener <rguenther@suse.de>
3187 * dwarf2out.c (prune_unused_types): Unconditionally mark
3188 called function DIEs.
3190 2020-01-20 Martin Liska <mliska@suse.cz>
3192 PR tree-optimization/93199
3193 * tree-eh.c (struct leh_state): Add
3194 new field outer_non_cleanup.
3195 (cleanup_is_dead_in): Pass leh_state instead
3196 of eh_region. Add a checking that state->outer_non_cleanup
3197 points to outer non-clean up region.
3198 (lower_try_finally): Record outer_non_cleanup
3200 (lower_catch): Likewise.
3201 (lower_eh_filter): Likewise.
3202 (lower_eh_must_not_throw): Likewise.
3203 (lower_cleanup): Likewise.
3205 2020-01-20 Richard Biener <rguenther@suse.de>
3207 PR tree-optimization/93094
3208 * tree-vectorizer.h (vect_loop_versioning): Adjust.
3209 (vect_transform_loop): Likewise.
3210 * tree-vectorizer.c (try_vectorize_loop_1): Pass down
3211 loop_vectorized_call to vect_transform_loop.
3212 * tree-vect-loop.c (vect_transform_loop): Pass down
3213 loop_vectorized_call to vect_loop_versioning.
3214 * tree-vect-loop-manip.c (vect_loop_versioning): Use
3215 the earlier discovered loop_vectorized_call.
3217 2020-01-19 Eric S. Raymond <esr@thyrsus.com>
3219 * doc/contribute.texi: Update for SVN -> Git transition.
3220 * doc/install.texi: Likewise.
3222 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
3225 * cgraph.c (cgraph_edge::make_speculative): Increase number of
3226 speculative targets.
3227 (verify_speculative_call): New function
3228 (cgraph_node::verify_node): Use it.
3229 * ipa-profile.c (ipa_profile): Fix formating; do not set number of
3232 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
3235 * cgraph.c (cgraph_edge::resolve_speculation): Fix foramting.
3236 (cgraph_edge::make_direct): Remove all indirect targets.
3237 (cgraph_edge::redirect_call_stmt_to_callee): Use make_direct..
3238 (cgraph_node::verify_node): Verify that only one call_stmt or
3239 lto_stmt_uid is set.
3240 * cgraphclones.c (cgraph_edge::clone): Set only one call_stmt or
3242 * lto-cgraph.c (lto_output_edge): Simplify streaming of stmt.
3243 (lto_output_ref): Simplify streaming of stmt.
3244 * lto-streamer-in.c (fixup_call_stmt_edges_1): Clear lto_stmt_uid.
3246 2020-01-18 Tamar Christina <tamar.christina@arm.com>
3248 * config/aarch64/aarch64-sve-builtins-base.cc (memory_vector_mode):
3249 Mark parameter unused.
3251 2020-01-18 Hans-Peter Nilsson <hp@axis.com>
3253 * config.gcc <obsolete targets>: Add crisv32-*-* and cris-*-linux*
3255 2019-01-18 Gerald Pfeifer <gerald@pfeifer.com>
3257 * varpool.c (ctor_useable_for_folding_p): Fix grammar.
3259 2020-01-18 Iain Sandoe <iain@sandoe.co.uk>
3261 * Makefile.in: Add coroutine-passes.o.
3262 * builtin-types.def (BT_CONST_SIZE): New.
3263 (BT_FN_BOOL_PTR): New.
3264 (BT_FN_PTR_PTR_CONST_SIZE_BOOL): New.
3265 * builtins.def (DEF_COROUTINE_BUILTIN): New.
3266 * coroutine-builtins.def: New file.
3267 * coroutine-passes.cc: New file.
3268 * function.h (struct GTY function): Add a bit to indicate that the
3269 function is a coroutine component.
3270 * internal-fn.c (expand_CO_FRAME): New.
3271 (expand_CO_YIELD): New.
3272 (expand_CO_SUSPN): New.
3273 (expand_CO_ACTOR): New.
3274 * internal-fn.def (CO_ACTOR): New.
3278 * passes.def: Add pass_coroutine_lower_builtins,
3279 pass_coroutine_early_expand_ifns.
3280 * tree-pass.h (make_pass_coroutine_lower_builtins): New.
3281 (make_pass_coroutine_early_expand_ifns): New.
3282 * doc/invoke.texi: Document the fcoroutines command line
3285 2020-01-18 Jakub Jelinek <jakub@redhat.com>
3287 * config/arm/vfp.md (*clear_vfp_multiple): Remove unused variable.
3290 * config/arm/arm.c (clear_operation_p): Don't use REGNO until
3291 after checking the argument is a REG. Don't use REGNO (reg)
3292 again to set last_regno, reuse regno variable instead.
3294 2020-01-17 David Malcolm <dmalcolm@redhat.com>
3296 * doc/analyzer.texi (Limitations): Add note about NaN.
3298 2020-01-17 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3299 Sudakshina Das <sudi.das@arm.com>
3301 * config/arm/arm.md (ashldi3): Generate thumb2_lsll for both reg
3302 and valid immediate.
3303 (ashrdi3): Generate thumb2_asrl for both reg and valid immediate.
3304 (lshrdi3): Generate thumb2_lsrl for valid immediates.
3305 * config/arm/constraints.md (Pg): New.
3306 * config/arm/predicates.md (long_shift_imm): New.
3307 (arm_reg_or_long_shift_imm): Likewise.
3308 * config/arm/thumb2.md (thumb2_asrl): New immediate alternative.
3309 (thumb2_lsll): Likewise.
3312 2020-01-17 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3313 Sudakshina Das <sudi.das@arm.com>
3315 * config/arm/arm.md (ashldi3): Generate thumb2_lsll for TARGET_HAVE_MVE.
3316 (ashrdi3): Generate thumb2_asrl for TARGET_HAVE_MVE.
3317 * config/arm/arm.c (arm_hard_regno_mode_ok): Allocate even odd
3318 register pairs for doubleword quantities for ARMv8.1M-Mainline.
3319 * config/arm/thumb2.md (thumb2_asrl): New.
3320 (thumb2_lsll): Likewise.
3322 2020-01-17 Jakub Jelinek <jakub@redhat.com>
3324 * config/arm/arm.c (cmse_nonsecure_call_inline_register_clear): Remove
3327 2020-01-17 Alexander Monakov <amonakov@ispras.ru>
3329 * gdbinit.in (help-gcc-hooks): New command.
3330 (pp, pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, ptc, pdn, ptn, pdd, prc,
3331 pi, pbm, pel, trt): Take $arg0 instead of $ if supplied. Update
3334 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
3336 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use the
3337 correct target macro.
3339 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
3341 * config/aarch64/aarch64-protos.h
3342 (aarch64_sve_ld1ro_operand_p): New.
3343 * config/aarch64/aarch64-sve-builtins-base.cc
3344 (class load_replicate): New.
3345 (class svld1ro_impl): New.
3346 (class svld1rq_impl): Change to inherit from load_replicate.
3347 (svld1ro): New sve intrinsic function base.
3348 * config/aarch64/aarch64-sve-builtins-base.def (svld1ro):
3349 New DEF_SVE_FUNCTION.
3350 * config/aarch64/aarch64-sve-builtins-base.h
3351 (svld1ro): New decl.
3352 * config/aarch64/aarch64-sve-builtins.cc
3353 (function_expander::add_mem_operand): Modify assert to allow
3355 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): New
3357 * config/aarch64/aarch64.c
3358 (aarch64_sve_ld1rq_operand_p): Implement in terms of ...
3359 (aarch64_sve_ld1rq_ld1ro_operand_p): This.
3360 (aarch64_sve_ld1ro_operand_p): New.
3361 * config/aarch64/aarch64.md (UNSPEC_LD1RO): New unspec.
3362 * config/aarch64/constraints.md (UOb,UOh,UOw,UOd): New.
3363 * config/aarch64/predicates.md
3364 (aarch64_sve_ld1ro_operand_{b,h,w,d}): New.
3366 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
3368 * config/aarch64/aarch64-c.c (_ARM_FEATURE_MATMUL_FLOAT64):
3369 Introduce this ACLE specified predefined macro.
3370 * config/aarch64/aarch64-option-extensions.def (f64mm): New.
3371 (fp): Disabling this disables f64mm.
3372 (simd): Disabling this disables f64mm.
3373 (fp16): Disabling this disables f64mm.
3374 (sve): Disabling this disables f64mm.
3375 * config/aarch64/aarch64.h (AARCH64_FL_F64MM): New.
3376 (AARCH64_ISA_F64MM): New.
3377 (TARGET_F64MM): New.
3378 * doc/invoke.texi (f64mm): Document new option.
3380 2020-01-17 Wilco Dijkstra <wdijkstr@arm.com>
3382 * config/aarch64/aarch64.c (generic_tunings): Add branch fusion.
3383 (neoversen1_tunings): Likewise.
3385 2020-01-17 Wilco Dijkstra <wdijkstr@arm.com>
3388 * config/aarch64/aarch64.c (aarch64_split_compare_and_swap)
3389 Add assert to ensure prolog has been emitted.
3390 (aarch64_split_atomic_op): Likewise.
3391 * config/aarch64/atomics.md (aarch64_compare_and_swap<mode>)
3392 Use epilogue_completed rather than reload_completed.
3393 (aarch64_atomic_exchange<mode>): Likewise.
3394 (aarch64_atomic_<atomic_optab><mode>): Likewise.
3395 (atomic_nand<mode>): Likewise.
3396 (aarch64_atomic_fetch_<atomic_optab><mode>): Likewise.
3397 (atomic_fetch_nand<mode>): Likewise.
3398 (aarch64_atomic_<atomic_optab>_fetch<mode>): Likewise.
3399 (atomic_nand_fetch<mode>): Likewise.
3401 2020-01-17 Richard Sandiford <richard.sandiford@arm.com>
3404 * config/aarch64/aarch64.h (REVERSIBLE_CC_MODE): Return false
3406 (REVERSE_CONDITION): Delete.
3407 * config/aarch64/iterators.md (CC_ONLY): New mode iterator.
3408 (CCFP_CCFPE): Likewise.
3409 (e): New mode attribute.
3410 * config/aarch64/aarch64.md (ccmp<GPI:mode>): Rename to...
3411 (@ccmp<CC_ONLY:mode><GPI:mode>): ...this, using CC_ONLY instead of CC.
3412 (fccmp<GPF:mode>, fccmpe<GPF:mode>): Merge into...
3413 (@ccmp<CCFP_CCFPE:mode><GPF:mode>): ...this combined pattern.
3414 (@ccmp<CC_ONLY:mode><GPI:mode>_rev): New pattern.
3415 (@ccmp<CCFP_CCFPE:mode><GPF:mode>_rev): Likewise.
3416 * config/aarch64/aarch64.c (aarch64_gen_compare_reg): Update
3417 name of generator from gen_ccmpdi to gen_ccmpccdi.
3418 (aarch64_gen_ccmp_next): Use code_for_ccmp. If we want to reverse
3419 the previous comparison but aren't able to, use the new ccmp_rev
3422 2020-01-17 Richard Sandiford <richard.sandiford@arm.com>
3424 * gimplify.c (gimplify_return_expr): Use poly_int_tree_p rather
3425 than testing directly for INTEGER_CST.
3426 (gimplify_target_expr, gimplify_omp_depend): Likewise.
3428 2020-01-17 Jakub Jelinek <jakub@redhat.com>
3430 PR tree-optimization/93292
3431 * tree-vect-stmts.c (vectorizable_comparison): Punt also if
3432 get_vectype_for_scalar_type returns NULL.
3434 2020-01-16 Jan Hubicka <hubicka@ucw.cz>
3436 * params.opt (-param=max-predicted-iterations): Increase range from 0.
3437 * predict.c (estimate_loops): Add 1 to param_max_predicted_iterations.
3439 2020-01-16 Jan Hubicka <hubicka@ucw.cz>
3441 * ipa-fnsummary.c (estimate_calls_size_and_time): Fix formating of
3443 * params.opt: (max-predicted-iterations): Set bounds.
3444 * predict.c (real_almost_one, real_br_prob_base,
3445 real_inv_br_prob_base, real_one_half, real_bb_freq_max): Remove.
3446 (propagate_freq): Add max_cyclic_prob parameter; cap cyclic
3447 probabilities; do not truncate to reg_br_prob_bases.
3448 (estimate_loops_at_level): Pass max_cyclic_prob.
3449 (estimate_loops): Compute max_cyclic_prob.
3450 (estimate_bb_frequencies): Do not initialize real_*; update calculation
3452 * profile-count.c (profile_probability::to_sreal): New.
3453 * profile-count.h (class sreal): Move up in file.
3454 (profile_probability::to_sreal): Declare.
3456 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
3459 (arm_invalid_conversion): New function for target hook.
3460 (arm_invalid_unary_op): New function for target hook.
3461 (arm_invalid_binary_op): New function for target hook.
3463 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
3465 * config.gcc: Add arm_bf16.h.
3466 * config/arm/arm-builtins.c (arm_mangle_builtin_type): Fix comment.
3467 (arm_simd_builtin_std_type): Add BFmode.
3468 (arm_init_simd_builtin_types): Define element types for vector types.
3469 (arm_init_bf16_types): New function.
3470 (arm_init_builtins): Add arm_init_bf16_types function call.
3471 * config/arm/arm-modes.def: Add BFmode and V4BF, V8BF vector modes.
3472 * config/arm/arm-simd-builtin-types.def: Add V4BF, V8BF.
3473 * config/arm/arm.c (aapcs_vfp_sub_candidate): Add BFmode.
3474 (arm_hard_regno_mode_ok): Add BFmode and tidy up statements.
3475 (arm_vector_mode_supported_p): Add V4BF, V8BF.
3476 (arm_mangle_type): Add __bf16.
3477 * config/arm/arm.h: Add V4BF, V8BF to VALID_NEON_DREG_MODE,
3478 VALID_NEON_QREG_MODE respectively. Add export arm_bf16_type_node,
3479 arm_bf16_ptr_type_node.
3480 * config/arm/arm.md: Add BFmode to movhf expand, mov pattern and
3481 define_split between ARM registers.
3482 * config/arm/arm_bf16.h: New file.
3483 * config/arm/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
3484 * config/arm/iterators.md: (ANY64_BF, VDXMOV, VHFBF, HFBF, fporbf): New.
3486 * config/arm/neon.md: Add BF vector types to movhf NEON move patterns.
3487 * config/arm/vfp.md: Add BFmode to movhf patterns.
3489 2020-01-16 Mihail Ionescu <mihail.ionescu@arm.com>
3490 Andre Vieira <andre.simoesdiasvieira@arm.com>
3492 * config/arm/arm-cpus.in (mve, mve_float): New features.
3493 (dsp, mve, mve.fp): New options.
3494 * config/arm/arm.h (TARGET_HAVE_MVE, TARGET_HAVE_MVE_FLOAT): Define.
3495 * config/arm/t-rmprofile: Map v8.1-M multilibs to v8-M.
3496 * doc/invoke.texi: Document the armv8.1-m mve and dps options.
3498 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3499 Thomas Preud'homme <thomas.preudhomme@arm.com>
3501 * config/arm/arm-cpus.in (ARMv8_1m_main): Redefine as an extension to
3503 * config/arm/arm.c (arm_options_perform_arch_sanity_checks): Remove
3504 error for using -mcmse when targeting Armv8.1-M Mainline.
3506 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3507 Thomas Preud'homme <thomas.preudhomme@arm.com>
3509 * config/arm/arm.md (nonsecure_call_internal): Do not force memory
3510 address in r4 when targeting Armv8.1-M Mainline.
3511 (nonsecure_call_value_internal): Likewise.
3512 * config/arm/thumb2.md (nonsecure_call_reg_thumb2): Make memory address
3513 a register match_operand again. Emit BLXNS when targeting
3515 (nonsecure_call_value_reg_thumb2): Likewise.
3517 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3518 Thomas Preud'homme <thomas.preudhomme@arm.com>
3520 * config/arm/arm.c (arm_add_cfa_adjust_cfa_note): Declare early.
3521 (cmse_nonsecure_call_inline_register_clear): Define new lazy_fpclear
3522 variable as true when floating-point ABI is not hard. Replace
3523 check against TARGET_HARD_FLOAT_ABI by checks against lazy_fpclear.
3524 Generate VLSTM and VLLDM instruction respectively before and
3525 after a function call to cmse_nonsecure_call function.
3526 * config/arm/unspecs.md (VUNSPEC_VLSTM): Define unspec.
3527 (VUNSPEC_VLLDM): Likewise.
3528 * config/arm/vfp.md (lazy_store_multiple_insn): New define_insn.
3529 (lazy_load_multiple_insn): Likewise.
3531 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3532 Thomas Preud'homme <thomas.preudhomme@arm.com>
3534 * config/arm/arm.c (vfp_emit_fstmd): Declare early.
3535 (arm_emit_vfp_multi_reg_pop): Likewise.
3536 (cmse_nonsecure_call_inline_register_clear): Abstract number of VFP
3537 registers to clear in max_fp_regno. Emit VPUSH and VPOP to save and
3538 restore callee-saved VFP registers.
3540 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3541 Thomas Preud'homme <thomas.preudhomme@arm.com>
3543 * config/arm/arm.c (arm_emit_multi_reg_pop): Declare early.
3544 (cmse_nonsecure_call_clear_caller_saved): Rename into ...
3545 (cmse_nonsecure_call_inline_register_clear): This. Save and clear
3546 callee-saved GPRs as well as clear ip register before doing a nonsecure
3547 call then restore callee-saved GPRs after it when targeting
3549 (arm_reorg): Adapt to function rename.
3551 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3552 Thomas Preud'homme <thomas.preudhomme@arm.com>
3554 * config/arm/arm-protos.h (clear_operation_p): Adapt prototype.
3555 * config/arm/arm.c (clear_operation_p): Extend to be able to check a
3556 clear_vfp_multiple pattern based on a new vfp parameter.
3557 (cmse_clear_registers): Generate VSCCLRM to clear VFP registers when
3558 targeting Armv8.1-M Mainline.
3559 (cmse_nonsecure_entry_clear_before_return): Clear VFP registers
3560 unconditionally when targeting Armv8.1-M Mainline architecture. Check
3561 whether VFP registers are available before looking call_used_regs for a
3563 * config/arm/predicates.md (clear_multiple_operation): Adapt to change
3564 of prototype of clear_operation_p.
3565 (clear_vfp_multiple_operation): New predicate.
3566 * config/arm/unspecs.md (VUNSPEC_VSCCLRM_VPR): New volatile unspec.
3567 * config/arm/vfp.md (clear_vfp_multiple): New define_insn.
3569 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3570 Thomas Preud'homme <thomas.preudhomme@arm.com>
3572 * config/arm/arm-protos.h (clear_operation_p): Declare.
3573 * config/arm/arm.c (clear_operation_p): New function.
3574 (cmse_clear_registers): Generate clear_multiple instruction pattern if
3575 targeting Armv8.1-M Mainline or successor.
3576 (output_return_instruction): Only output APSR register clearing if
3577 Armv8.1-M Mainline instructions not available.
3578 (thumb_exit): Likewise.
3579 * config/arm/predicates.md (clear_multiple_operation): New predicate.
3580 * config/arm/thumb2.md (clear_apsr): New define_insn.
3581 (clear_multiple): Likewise.
3582 * config/arm/unspecs.md (VUNSPEC_CLRM_APSR): New volatile unspec.
3584 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3585 Thomas Preud'homme <thomas.preudhomme@arm.com>
3587 * config/arm/arm.c (fp_sysreg_names): Declare and define.
3588 (use_return_insn): Also return false for Armv8.1-M Mainline.
3589 (output_return_instruction): Skip FPSCR clearing if Armv8.1-M
3590 Mainline instructions are available.
3591 (arm_compute_frame_layout): Allocate space in frame for FPCXTNS
3592 when targeting Armv8.1-M Mainline Security Extensions.
3593 (arm_expand_prologue): Save FPCXTNS if this is an Armv8.1-M
3594 Mainline entry function.
3595 (cmse_nonsecure_entry_clear_before_return): Clear IP and r4 if
3596 targeting Armv8.1-M Mainline or successor.
3597 (arm_expand_epilogue): Fix indentation of caller-saved register
3598 clearing. Restore FPCXTNS if this is an Armv8.1-M Mainline
3600 * config/arm/arm.h (TARGET_HAVE_FP_CMSE): New macro.
3601 (FP_SYSREGS): Likewise.
3602 (enum vfp_sysregs_encoding): Define enum.
3603 (fp_sysreg_names): Declare.
3604 * config/arm/unspecs.md (VUNSPEC_VSTR_VLDR): New volatile unspec.
3605 * config/arm/vfp.md (push_fpsysreg_insn): New define_insn.
3606 (pop_fpsysreg_insn): Likewise.
3608 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3609 Thomas Preud'homme <thomas.preudhomme@arm.com>
3611 * config/arm/arm-cpus.in (armv8_1m_main): New feature.
3612 (ARMv4, ARMv4t, ARMv5t, ARMv5te, ARMv5tej, ARMv6, ARMv6j, ARMv6k,
3613 ARMv6z, ARMv6kz, ARMv6zk, ARMv6t2, ARMv6m, ARMv7, ARMv7a, ARMv7ve,
3614 ARMv7r, ARMv7m, ARMv7em, ARMv8a, ARMv8_1a, ARMv8_2a, ARMv8_3a,
3615 ARMv8_4a, ARMv8_5a, ARMv8m_base, ARMv8m_main, ARMv8r): Reindent.
3616 (ARMv8_1m_main): New feature group.
3617 (armv8.1-m.main): New architecture.
3618 * config/arm/arm-tables.opt: Regenerate.
3619 * config/arm/arm.c (arm_arch8_1m_main): Define and default initialize.
3620 (arm_option_reconfigure_globals): Initialize arm_arch8_1m_main.
3621 (arm_options_perform_arch_sanity_checks): Error out when targeting
3622 Armv8.1-M Mainline Security Extensions.
3623 * config/arm/arm.h (arm_arch8_1m_main): Declare.
3625 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
3627 * config/aarch64/aarch64-simd-builtins.def (aarch64_bfdot,
3628 aarch64_bfdot_lane, aarch64_bfdot_laneq): New.
3629 * config/aarch64/aarch64-simd.md (aarch64_bfdot, aarch64_bfdot_lane,
3630 aarch64_bfdot_laneq): New.
3631 * config/aarch64/arm_bf16.h (vbfdot_f32, vbfdotq_f32,
3632 vbfdot_lane_f32, vbfdotq_lane_f32, vbfdot_laneq_f32,
3633 vbfdotq_laneq_f32): New.
3634 * config/aarch64/iterators.md (UNSPEC_BFDOT, Vbfdottype,
3635 VBFMLA_W, VBF): New.
3636 (isquadop): Add V4BF, V8BF.
3638 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
3640 * config/aarch64/aarch64-builtins.c: (enum aarch64_type_qualifiers):
3641 New qualifier_lane_quadtup_index, TYPES_TERNOP_SSUS,
3642 TYPES_QUADOPSSUS_LANE_QUADTUP, TYPES_QUADOPSSSU_LANE_QUADTUP.
3643 (aarch64_simd_expand_args): Add case SIMD_ARG_LANE_QUADTUP_INDEX.
3644 (aarch64_simd_expand_builtin): Add qualifier_lane_quadtup_index.
3645 * config/aarch64/aarch64-simd-builtins.def (usdot, usdot_lane,
3646 usdot_laneq, sudot_lane,sudot_laneq): New.
3647 * config/aarch64/aarch64-simd.md (aarch64_usdot): New.
3648 (aarch64_<sur>dot_lane): New.
3649 * config/aarch64/arm_neon.h (vusdot_s32): New.
3651 (vusdot_lane_s32): New.
3652 (vsudot_lane_s32): New.
3653 * config/aarch64/iterators.md (DOTPROD_I8MM): New iterator.
3654 (UNSPEC_USDOT, UNSPEC_SUDOT): New unspecs.
3656 2020-01-16 Martin Liska <mliska@suse.cz>
3658 * value-prof.c (dump_histogram_value): Fix
3659 obvious spacing issue.
3661 2020-01-16 Andrew Pinski <apinski@marvell.com>
3663 * tree-ssa-sccvn.c(vn_reference_lookup_3): Check lhs for
3664 !storage_order_barrier_p.
3666 2020-01-16 Andrew Pinski <apinski@marvell.com>
3668 * sched-int.h (_dep): Add unused bit-field field for the padding.
3669 * sched-deps.c (init_dep_1): Init unused field.
3671 2020-01-16 Andrew Pinski <apinski@marvell.com>
3673 * optabs.h (create_expand_operand): Initialize target field also.
3675 2020-01-16 Andre Vieira <andre.simoesdiasvieira@arm.com>
3677 PR tree-optimization/92429
3678 * tree-ssa-loop-niter.h (simplify_replace_tree): Add parameter.
3679 * tree-ssa-loop-niter.c (simplify_replace_tree): Add parameter to
3681 * tree-vect-loop.c (update_epilogue_vinfo): Do not fold when replacing
3684 2020-01-16 Richard Sandiford <richard.sandiford@arm.com>
3686 * config/aarch64/aarch64.c (aarch64_split_sve_subreg_move): Apply
3687 aarch64_sve_int_mode to each mode.
3689 2020-01-15 David Malcolm <dmalcolm@redhat.com>
3691 * doc/analyzer.texi (Overview): Add note about
3692 -fdump-ipa-analyzer.
3694 2020-01-15 Wilco Dijkstra <wdijkstr@arm.com>
3696 PR tree-optimization/93231
3697 * tree-ssa-forwprop.c (optimize_count_trailing_zeroes): Check
3698 input_type is unsigned. Use tree_to_shwi for shift constant.
3699 Check CST_STRING element size is CHAR_TYPE_SIZE bits.
3700 (simplify_count_trailing_zeroes): Add test to handle known non-zero
3701 inputs more efficiently.
3703 2020-01-15 Uroš Bizjak <ubizjak@gmail.com>
3705 * config/i386/i386.md (*movsf_internal): Do not require
3706 SSE2 ISA for alternatives 14 and 15.
3708 2020-01-15 Richard Biener <rguenther@suse.de>
3711 * tree-eh.c (sink_clobbers): If we already visited the destination
3712 block do not defer insertion.
3713 (pass_lower_eh_dispatch::execute): Maintain BB_VISITED for
3714 the purpose of defered insertion.
3716 2020-01-15 Jakub Jelinek <jakub@redhat.com>
3718 * BASE-VER: Bump to 10.0.1.
3720 2020-01-15 Richard Sandiford <richard.sandiford@arm.com>
3722 PR tree-optimization/93247
3723 * tree-vect-loop.c (update_epilogue_loop_vinfo): Check the access
3724 type of the stmt that we're going to vectorize.
3726 2020-01-15 Richard Sandiford <richard.sandiford@arm.com>
3728 * tree-vect-slp.c (vectorize_slp_instance_root_stmt): Use a
3729 VIEW_CONVERT_EXPR if the vectorized constructor has a diffeent
3732 2020-01-15 Martin Liska <mliska@suse.cz>
3734 * ipa-profile.c (ipa_profile_read_edge_summary): Do not allow
3735 2 calls of streamer_read_hwi in a function call.
3737 2020-01-15 Richard Biener <rguenther@suse.de>
3739 * alias.c (record_alias_subset): Avoid redundant work when
3740 subset is already recorded.
3742 2020-01-14 David Malcolm <dmalcolm@redhat.com>
3744 * doc/invoke.texi (-fdiagnostics-show-cwe): Add note that some of
3745 the analyzer options provide CWE identifiers.
3747 2020-01-14 David Malcolm <dmalcolm@redhat.com>
3749 * tree-diagnostic-path.cc (path_summary::event_range::print):
3750 When testing for UNKNOWN_LOCATION, look through ad-hoc wrappers
3751 using get_pure_location.
3753 2020-01-15 Jakub Jelinek <jakub@redhat.com>
3755 PR tree-optimization/93262
3756 * tree-ssa-dse.c (maybe_trim_memstar_call): For *_chk builtins,
3757 perform head trimming only if the last argument is constant,
3758 either all ones, or larger or equal to head trim, in the latter
3759 case decrease the last argument by head_trim.
3761 PR tree-optimization/93249
3762 * tree-ssa-dse.c: Include builtins.h and gimple-fold.h.
3763 (maybe_trim_memstar_call): Move head_trim and tail_trim vars to
3764 function body scope, reindent. For BUILTIN_IN_STRNCPY*, don't
3765 perform head trim unless we can prove there are no '\0' chars
3766 from the source among the first head_trim chars.
3768 2020-01-14 David Malcolm <dmalcolm@redhat.com>
3770 * Makefile.in (ANALYZER_OBJS): Add analyzer/function-set.o.
3772 2020-01-15 Jakub Jelinek <jakub@redhat.com>
3775 * config/i386/sse.md
3776 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1,
3777 *<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1,
3778 *<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1,
3779 *<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1): Use
3780 just a single alternative instead of two, make operands 1 and 2
3783 2020-01-14 Jan Hubicka <hubicka@ucw.cz>
3786 * ipa-devirt.c (odr_types_equivalent_p): Compare TREE_ADDRESSABLE and
3789 2020-01-14 David Malcolm <dmalcolm@redhat.com>
3791 * Makefile.in (lang_opt_files): Add analyzer.opt.
3792 (ANALYZER_OBJS): New.
3793 (OBJS): Add digraph.o, graphviz.o, ordered-hash-map-tests.o,
3794 tristate.o and ANALYZER_OBJS.
3795 (TEXI_GCCINT_FILES): Add analyzer.texi.
3796 * common.opt (-fanalyzer): New driver option.
3797 * config.in: Regenerate.
3798 * configure: Regenerate.
3799 * configure.ac (--disable-analyzer, ENABLE_ANALYZER): New option.
3800 (gccdepdir): Also create depdir for "analyzer" subdir.
3801 * digraph.cc: New file.
3802 * digraph.h: New file.
3803 * doc/analyzer.texi: New file.
3804 * doc/gccint.texi ("Static Analyzer") New menu item.
3805 (analyzer.texi): Include it.
3806 * doc/invoke.texi ("Static Analyzer Options"): New list and new section.
3807 ("Warning Options"): Add static analysis warnings to the list.
3808 (-Wno-analyzer-double-fclose): New option.
3809 (-Wno-analyzer-double-free): New option.
3810 (-Wno-analyzer-exposure-through-output-file): New option.
3811 (-Wno-analyzer-file-leak): New option.
3812 (-Wno-analyzer-free-of-non-heap): New option.
3813 (-Wno-analyzer-malloc-leak): New option.
3814 (-Wno-analyzer-possible-null-argument): New option.
3815 (-Wno-analyzer-possible-null-dereference): New option.
3816 (-Wno-analyzer-null-argument): New option.
3817 (-Wno-analyzer-null-dereference): New option.
3818 (-Wno-analyzer-stale-setjmp-buffer): New option.
3819 (-Wno-analyzer-tainted-array-index): New option.
3820 (-Wno-analyzer-use-after-free): New option.
3821 (-Wno-analyzer-use-of-pointer-in-stale-stack-frame): New option.
3822 (-Wno-analyzer-use-of-uninitialized-value): New option.
3823 (-Wanalyzer-too-complex): New option.
3824 (-fanalyzer-call-summaries): New warning.
3825 (-fanalyzer-checker=): New warning.
3826 (-fanalyzer-fine-grained): New warning.
3827 (-fno-analyzer-state-merge): New warning.
3828 (-fno-analyzer-state-purge): New warning.
3829 (-fanalyzer-transitivity): New warning.
3830 (-fanalyzer-verbose-edges): New warning.
3831 (-fanalyzer-verbose-state-changes): New warning.
3832 (-fanalyzer-verbosity=): New warning.
3833 (-fdump-analyzer): New warning.
3834 (-fdump-analyzer-callgraph): New warning.
3835 (-fdump-analyzer-exploded-graph): New warning.
3836 (-fdump-analyzer-exploded-nodes): New warning.
3837 (-fdump-analyzer-exploded-nodes-2): New warning.
3838 (-fdump-analyzer-exploded-nodes-3): New warning.
3839 (-fdump-analyzer-supergraph): New warning.
3840 * doc/sourcebuild.texi (dg-require-dot): New.
3841 (dg-check-dot): New.
3842 * gdbinit.in (break-on-saved-diagnostic): New command.
3843 * graphviz.cc: New file.
3844 * graphviz.h: New file.
3845 * ordered-hash-map-tests.cc: New file.
3846 * ordered-hash-map.h: New file.
3847 * passes.def (pass_analyzer): Add before
3848 pass_ipa_whole_program_visibility.
3849 * selftest-run-tests.c (selftest::run_tests): Call
3850 selftest::ordered_hash_map_tests_cc_tests.
3851 * selftest.h (selftest::ordered_hash_map_tests_cc_tests): New
3853 * shortest-paths.h: New file.
3854 * timevar.def (TV_ANALYZER): New timevar.
3855 (TV_ANALYZER_SUPERGRAPH): Likewise.
3856 (TV_ANALYZER_STATE_PURGE): Likewise.
3857 (TV_ANALYZER_PLAN): Likewise.
3858 (TV_ANALYZER_SCC): Likewise.
3859 (TV_ANALYZER_WORKLIST): Likewise.
3860 (TV_ANALYZER_DUMP): Likewise.
3861 (TV_ANALYZER_DIAGNOSTICS): Likewise.
3862 (TV_ANALYZER_SHORTEST_PATHS): Likewise.
3863 * tree-pass.h (make_pass_analyzer): New decl.
3864 * tristate.cc: New file.
3865 * tristate.h: New file.
3867 2020-01-14 Uroš Bizjak <ubizjak@gmail.com>
3870 * config/i386/i386.md (*movsf_internal): Require SSE2 ISA for
3871 alternatives 9 and 10.
3873 2020-01-14 David Malcolm <dmalcolm@redhat.com>
3875 * attribs.c (excl_hash_traits::empty_zero_p): New static constant.
3876 * gcov.c (function_start_pair_hash::empty_zero_p): Likewise.
3877 * graphite.c (struct sese_scev_hash::empty_zero_p): Likewise.
3878 * hash-map-tests.c (selftest::test_nonzero_empty_key): New selftest.
3879 (selftest::hash_map_tests_c_tests): Call it.
3880 * hash-map-traits.h (simple_hashmap_traits::empty_zero_p):
3881 New static constant, using the value of = H::empty_zero_p.
3882 (unbounded_hashmap_traits::empty_zero_p): Likewise, using the value
3883 from default_hash_traits <Value>.
3884 * hash-map.h (hash_map::empty_zero_p): Likewise, using the value
3886 * hash-set-tests.c (value_hash_traits::empty_zero_p): Likewise.
3887 * hash-table.h (hash_table::alloc_entries): Guard the loop of
3888 calls to mark_empty with !Descriptor::empty_zero_p.
3889 (hash_table::empty_slow): Conditionalize the memset call with a
3890 check that Descriptor::empty_zero_p; otherwise, loop through the
3891 entries calling mark_empty on them.
3892 * hash-traits.h (int_hash::empty_zero_p): New static constant.
3893 (pointer_hash::empty_zero_p): Likewise.
3894 (pair_hash::empty_zero_p): Likewise.
3895 * ipa-devirt.c (default_hash_traits <type_pair>::empty_zero_p):
3897 * ipa-prop.c (ipa_bit_ggc_hash_traits::empty_zero_p): Likewise.
3898 (ipa_vr_ggc_hash_traits::empty_zero_p): Likewise.
3899 * profile.c (location_triplet_hash::empty_zero_p): Likewise.
3900 * sanopt.c (sanopt_tree_triplet_hash::empty_zero_p): Likewise.
3901 (sanopt_tree_couple_hash::empty_zero_p): Likewise.
3902 * tree-hasher.h (int_tree_hasher::empty_zero_p): Likewise.
3903 * tree-ssa-sccvn.c (vn_ssa_aux_hasher::empty_zero_p): Likewise.
3904 * tree-vect-slp.c (bst_traits::empty_zero_p): Likewise.
3906 (default_hash_traits<scalar_cond_masked_key>::empty_zero_p):
3909 2020-01-14 Kewen Lin <linkw@gcc.gnu.org>
3911 * cfgloopanal.c (average_num_loop_insns): Free bbs when early return,
3912 fix typo on return value.
3914 2020-01-14 Xiong Hu Luo <luoxhu@linux.ibm.com>
3917 * cgraph.c (symbol_table::create_edge): Init speculative_id and
3919 (cgraph_edge::make_speculative): Add param for setting speculative_id
3921 (cgraph_edge::speculative_call_info): Update comments and find reference
3922 by speculative_id for multiple indirect targets.
3923 (cgraph_edge::resolve_speculation): Decrease the speculations
3924 for indirect edge, drop it's speculative if not direct target
3925 left. Update comments.
3926 (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
3927 (cgraph_node::dump): Print num_speculative_call_targets.
3928 (cgraph_node::verify_node): Don't report error if speculative
3929 edge not include statement.
3930 (cgraph_edge::num_speculative_call_targets_p): New function.
3931 * cgraph.h (int common_target_id): Remove.
3932 (int common_target_probability): Remove.
3933 (num_speculative_call_targets): New variable.
3934 (make_speculative): Add param for setting speculative_id.
3935 (cgraph_edge::num_speculative_call_targets_p): New declare.
3936 (target_prob): New variable.
3937 (speculative_id): New variable.
3938 * ipa-fnsummary.c (analyze_function_body): Create and duplicate
3939 call summaries for multiple speculative call targets.
3940 * cgraphclones.c (cgraph_node::create_clone): Clone speculative_id.
3941 * ipa-profile.c (struct speculative_call_target): New struct.
3942 (class speculative_call_summary): New class.
3943 (class speculative_call_summaries): New class.
3944 (call_sums): New variable.
3945 (ipa_profile_generate_summary): Generate indirect multiple targets summaries.
3946 (ipa_profile_write_edge_summary): New function.
3947 (ipa_profile_write_summary): Stream out indirect multiple targets summaries.
3948 (ipa_profile_dump_all_summaries): New function.
3949 (ipa_profile_read_edge_summary): New function.
3950 (ipa_profile_read_summary_section): New function.
3951 (ipa_profile_read_summary): Stream in indirect multiple targets summaries.
3952 (ipa_profile): Generate num_speculative_call_targets from
3954 * ipa-ref.h (speculative_id): New variable.
3955 * ipa-utils.c (ipa_merge_profiles): Update with target_prob.
3956 * lto-cgraph.c (lto_output_edge): Remove indirect common_target_id and
3957 common_target_probability. Stream out speculative_id and
3958 num_speculative_call_targets.
3959 (input_edge): Likewise.
3960 * predict.c (dump_prediction): Remove edges count assert to be
3962 * symtab.c (symtab_node::create_reference): Init speculative_id.
3963 (symtab_node::clone_references): Clone speculative_id.
3964 (symtab_node::clone_referring): Clone speculative_id.
3965 (symtab_node::clone_reference): Clone speculative_id.
3966 (symtab_node::clear_stmts_in_references): Clear speculative_id.
3967 * tree-inline.c (copy_bb): Duplicate all the speculative edges
3968 if indirect call contains multiple speculative targets.
3969 * value-prof.h (check_ic_target): Remove.
3970 * value-prof.c (gimple_value_profile_transformations):
3971 Use void function gimple_ic_transform.
3972 * value-prof.c (gimple_ic_transform): Handle topn case.
3973 Fix comment typos. Change it to a void function.
3975 2020-01-13 Andrew Pinski <apinski@marvell.com>
3977 * config/aarch64/aarch64-cores.def (octeontx2): New define.
3978 (octeontx2t98): New define.
3979 (octeontx2t96): New define.
3980 (octeontx2t93): New define.
3981 (octeontx2f95): New define.
3982 (octeontx2f95n): New define.
3983 (octeontx2f95mm): New define.
3984 * config/aarch64/aarch64-tune.md: Regenerate.
3985 * doc/invoke.texi (-mcpu=): Document the new cpu types.
3987 2020-01-13 Jason Merrill <jason@redhat.com>
3989 PR c++/33799 - destroy return value if local cleanup throws.
3990 * gimplify.c (gimplify_return_expr): Handle COMPOUND_EXPR.
3992 2020-01-13 Martin Liska <mliska@suse.cz>
3994 * ipa-cp.c (get_max_overall_size): Use newly
3995 renamed param param_ipa_cp_unit_growth.
3996 * params.opt: Remove legacy param name.
3998 2020-01-13 Martin Sebor <msebor@redhat.com>
4000 PR tree-optimization/93213
4001 * tree-ssa-strlen.c (handle_store): Only allow single-byte nul-over-nul
4002 stores to be eliminated.
4004 2020-01-13 Martin Liska <mliska@suse.cz>
4006 * opts.c (print_help): Do not print CL_PARAM
4007 and CL_WARNING for CL_OPTIMIZATION.
4009 2020-01-13 Jonathan Wakely <jwakely@redhat.com>
4012 * doc/invoke.texi (Warning Options): Add caveat about some warnings
4013 depending on optimization settings.
4015 2020-01-13 Jakub Jelinek <jakub@redhat.com>
4017 PR tree-optimization/90838
4018 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
4019 SCALAR_INT_TYPE_MODE directly in CTZ_DEFINED_VALUE_AT_ZERO macro
4020 argument rather than to initialize temporary for targets that
4021 don't use the mode argument at all. Initialize ctzval to avoid
4024 2020-01-10 Thomas Schwinge <thomas@codesourcery.com>
4026 * tree.h (OMP_CLAUSE_USE_DEVICE_PTR_IF_PRESENT): New definition.
4027 * tree-core.h: Document it.
4028 * gimplify.c (gimplify_omp_workshare): Set it.
4029 * omp-low.c (lower_omp_target): Use it.
4030 * tree-pretty-print.c (dump_omp_clause): Print it.
4032 * omp-low.c (lower_omp_target) <OMP_CLAUSE_USE_DEVICE_PTR etc.>:
4033 Assert that for OpenACC we always have 'GOMP_MAP_USE_DEVICE_PTR'.
4035 2020-01-10 David Malcolm <dmalcolm@redhat.com>
4037 * Makefile.in (OBJS): Add tree-diagnostic-path.o.
4038 * common.opt (fdiagnostics-path-format=): New option.
4039 (diagnostic_path_format): New enum.
4040 (fdiagnostics-show-path-depths): New option.
4041 * coretypes.h (diagnostic_event_id_t): New forward decl.
4042 * diagnostic-color.c (color_dict): Add "path".
4043 * diagnostic-event-id.h: New file.
4044 * diagnostic-format-json.cc (json_from_expanded_location): Make
4046 (json_end_diagnostic): Call context->make_json_for_path if it
4047 exists and the diagnostic has a path.
4048 (diagnostic_output_format_init): Clear context->print_path.
4049 * diagnostic-path.h: New file.
4050 * diagnostic-show-locus.c (colorizer::set_range): Special-case
4051 when printing a run of events in a diagnostic_path so that they
4052 all get the same color.
4053 (layout::m_diagnostic_path_p): New field.
4054 (layout::layout): Initialize it.
4055 (layout::print_any_labels): Don't colorize the label text for an
4056 event in a diagnostic_path.
4057 (gcc_rich_location::add_location_if_nearby): Add
4058 "restrict_to_current_line_spans" and "label" params. Pass the
4059 former to layout.maybe_add_location_range; pass the latter
4060 when calling add_range.
4061 * diagnostic.c: Include "diagnostic-path.h".
4062 (diagnostic_initialize): Initialize context->path_format and
4063 context->show_path_depths.
4064 (diagnostic_show_any_path): New function.
4065 (diagnostic_path::interprocedural_p): New function.
4066 (diagnostic_report_diagnostic): Call diagnostic_show_any_path.
4067 (simple_diagnostic_path::num_events): New function.
4068 (simple_diagnostic_path::get_event): New function.
4069 (simple_diagnostic_path::add_event): New function.
4070 (simple_diagnostic_event::simple_diagnostic_event): New ctor.
4071 (simple_diagnostic_event::~simple_diagnostic_event): New dtor.
4072 (debug): New overload taking a diagnostic_path *.
4073 * diagnostic.def (DK_DIAGNOSTIC_PATH): New.
4074 * diagnostic.h (enum diagnostic_path_format): New enum.
4075 (json::value): New forward decl.
4076 (diagnostic_context::path_format): New field.
4077 (diagnostic_context::show_path_depths): New field.
4078 (diagnostic_context::print_path): New callback field.
4079 (diagnostic_context::make_json_for_path): New callback field.
4080 (diagnostic_show_any_path): New decl.
4081 (json_from_expanded_location): New decl.
4082 * doc/invoke.texi (-fdiagnostics-path-format=): New option.
4083 (-fdiagnostics-show-path-depths): New option.
4084 (-fdiagnostics-color): Add "path" to description of default
4085 GCC_COLORS; describe it.
4086 (-fdiagnostics-format=json): Document how diagnostic paths are
4087 represented in the JSON output format.
4088 * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
4089 Add optional params "restrict_to_current_line_spans" and "label".
4090 * opts.c (common_handle_option): Handle
4091 OPT_fdiagnostics_path_format_ and
4092 OPT_fdiagnostics_show_path_depths.
4093 * pretty-print.c: Include "diagnostic-event-id.h".
4094 (pp_format): Implement "%@" format code for printing
4095 diagnostic_event_id_t *.
4096 (selftest::test_pp_format): Add tests for "%@".
4097 * selftest-run-tests.c (selftest::run_tests): Call
4098 selftest::tree_diagnostic_path_cc_tests.
4099 * selftest.h (selftest::tree_diagnostic_path_cc_tests): New decl.
4100 * toplev.c (general_init): Initialize global_dc->path_format and
4101 global_dc->show_path_depths.
4102 * tree-diagnostic-path.cc: New file.
4103 * tree-diagnostic.c (maybe_unwind_expanded_macro_loc): Make
4104 non-static. Drop "diagnostic" param in favor of storing the
4105 original value of "where" and re-using it.
4106 (virt_loc_aware_diagnostic_finalizer): Update for dropped param of
4107 maybe_unwind_expanded_macro_loc.
4108 (tree_diagnostics_defaults): Initialize context->print_path and
4109 context->make_json_for_path.
4110 * tree-diagnostic.h (default_tree_diagnostic_path_printer): New
4112 (default_tree_make_json_for_path): New decl.
4113 (maybe_unwind_expanded_macro_loc): New decl.
4115 2020-01-10 Jakub Jelinek <jakub@redhat.com>
4117 PR tree-optimization/93210
4118 * fold-const.h (native_encode_initializer,
4119 can_native_interpret_type_p): Declare.
4120 * fold-const.c (native_encode_string): Fix up handling with off != -1,
4122 (native_encode_initializer): New function, moved from dwarf2out.c.
4123 Adjust to native_encode_expr compatible arguments, including dry-run
4124 and partial extraction modes. Don't handle STRING_CST.
4125 (can_native_interpret_type_p): No longer static.
4126 * gimple-fold.c (fold_ctor_reference): For native_encode_expr, verify
4127 offset / BITS_PER_UNIT fits into int and don't call it if
4128 can_native_interpret_type_p fails. If suboff is NULL and for
4129 CONSTRUCTOR fold_{,non}array_ctor_reference returns NULL, retry with
4130 native_encode_initializer.
4131 (fold_const_aggregate_ref_1): Formatting fix.
4132 * dwarf2out.c (native_encode_initializer): Moved to fold-const.c.
4133 (tree_add_const_value_attribute): Adjust caller.
4135 PR tree-optimization/90838
4136 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
4137 SCALAR_INT_TYPE_MODE instead of TYPE_MODE as operand of
4138 CTZ_DEFINED_VALUE_AT_ZERO.
4140 2020-01-10 Vladimir Makarov <vmakarov@redhat.com>
4143 * lra-constraints.c (match_reload): Permit input operands have the
4144 same mode as output while other input operands have a different
4147 2020-01-10 Wilco Dijkstra <wdijkstr@arm.com>
4149 PR tree-optimization/90838
4150 * tree-ssa-forwprop.c (check_ctz_array): Add new function.
4151 (check_ctz_string): Likewise.
4152 (optimize_count_trailing_zeroes): Likewise.
4153 (simplify_count_trailing_zeroes): Likewise.
4154 (pass_forwprop::execute): Try ctz simplification.
4155 * match.pd: Add matching for ctz idioms.
4157 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
4159 * config/aarch64/aarch64.c (aarch64_invalid_conversion): New function
4161 (aarch64_invalid_unary_op): New function for target hook.
4162 (aarch64_invalid_binary_op): New function for target hook.
4164 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
4166 * config.gcc: Add arm_bf16.h.
4167 * config/aarch64/aarch64-builtins.c
4168 (aarch64_simd_builtin_std_type): Add BFmode.
4169 (aarch64_init_simd_builtin_types): Define element types for vector
4171 (aarch64_init_bf16_types): New function.
4172 (aarch64_general_init_builtins): Add arm_init_bf16_types function call.
4173 * config/aarch64/aarch64-modes.def: Add BFmode and V4BF, V8BF vector
4175 * config/aarch64/aarch64-simd-builtin-types.def: Add BF SIMD types.
4176 * config/aarch64/aarch64-simd.md: Add BF vector types to NEON move
4178 * config/aarch64/aarch64.h (AARCH64_VALID_SIMD_DREG_MODE): Add V4BF.
4179 (AARCH64_VALID_SIMD_QREG_MODE): Add V8BF.
4180 * config/aarch64/aarch64.c
4181 (aarch64_classify_vector_mode): Add support for BF types.
4182 (aarch64_gimplify_va_arg_expr): Add support for BF types.
4183 (aarch64_vq_mode): Add support for BF types.
4184 (aarch64_simd_container_mode): Add support for BF types.
4185 (aarch64_mangle_type): Add support for BF scalar type.
4186 * config/aarch64/aarch64.md: Add BFmode to movhf pattern.
4187 * config/aarch64/arm_bf16.h: New file.
4188 * config/aarch64/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
4189 * config/aarch64/iterators.md: Add BF types to mode attributes.
4190 (HFBF, GPF_TF_F16_MOV, VDMOV, VQMOV, VQMOV_NO2Em VALL_F16MOV): New.
4192 2020-01-10 Jason Merrill <jason@redhat.com>
4194 PR c++/93173 - incorrect tree sharing.
4195 * gimplify.c (copy_if_shared): No longer static.
4196 * gimplify.h: Declare it.
4198 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
4200 * doc/invoke.texi (-msve-vector-bits=): Document that
4201 -msve-vector-bits=128 now generates VL-specific code for
4202 little-endian targets.
4203 * config/aarch64/aarch64-sve-builtins.cc (register_builtin_types): Use
4204 build_vector_type_for_mode to construct the data vector types.
4205 * config/aarch64/aarch64.c (aarch64_convert_sve_vector_bits): Generate
4206 VL-specific code for -msve-vector-bits=128 on little-endian targets.
4207 (aarch64_simd_container_mode): Always prefer Advanced SIMD modes
4208 for 128-bit vectors.
4210 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
4212 * config/aarch64/aarch64.c (aarch64_evpc_sel): Fix gen_vcond_mask
4215 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
4217 * config/aarch64/aarch64-builtins.c
4218 (aarch64_builtin_vectorized_function): Check for specific vector modes,
4219 rather than checking the number of elements and the element mode.
4221 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
4223 * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
4224 get_related_vectype_for_scalar_type rather than build_vector_type
4225 to create the index type for a conditional reduction.
4227 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
4229 * tree-vect-loop.c (update_epilogue_loop_vinfo): Update DR_REF
4230 for any type of gather or scatter, including strided accesses.
4232 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
4234 * tree-vectorizer.h (get_dr_vinfo_offset): Add missing function
4237 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
4239 * tree-vect-data-refs.c (vect_create_addr_base_for_vector_ref): Use
4241 * tree-vect-loop.c (update_epilogue_loop_vinfo): Remove orig_drs_init
4242 parameter and its use to reset DR_OFFSET's.
4243 (vect_transform_loop): Remove orig_drs_init argument.
4244 * tree-vect-loop-manip.c (vect_update_init_of_dr): Update the offset
4245 member of dr_vec_info rather than the offset of the associated
4246 data_reference's innermost_loop_behavior.
4247 (vect_update_init_of_dr): Pass dr_vec_info instead of data_reference.
4248 (vect_do_peeling): Remove orig_drs_init parameter and its construction.
4249 * tree-vect-stmts.c (check_scan_store): Replace use of DR_OFFSET with
4250 get_dr_vinfo_offset.
4251 (vectorizable_store): Likewise.
4252 (vectorizable_load): Likewise.
4254 2020-01-10 Richard Biener <rguenther@suse.de>
4256 * gimple-ssa-store-merging
4257 (pass_store_merging::terminate_all_aliasing_chains): Cache alias info.
4259 2020-01-10 Martin Liska <mliska@suse.cz>
4262 * ipa-inline-analysis.c (offline_size): Make proper parenthesis
4263 encapsulation that was there before r280040.
4265 2020-01-10 Richard Biener <rguenther@suse.de>
4268 * tree-eh.c (sink_clobbers): Move clobbers to out-of-IL
4269 sequences to avoid walking them again for secondary opportunities.
4270 (pass_lower_eh_dispatch::execute): Instead actually insert
4273 2020-01-10 Richard Biener <rguenther@suse.de>
4276 * tree-eh.c (redirect_eh_edge_1): Avoid some work if possible.
4277 (cleanup_all_empty_eh): Walk landing pads in reverse order to
4278 avoid quadraticness.
4280 2020-01-10 Martin Jambor <mjambor@suse.cz>
4282 * params.opt (param_ipa_sra_max_replacements): Mark as Optimization.
4283 * ipa-sra.c (pull_accesses_from_callee): New parameter caller, use it
4284 to get param_ipa_sra_max_replacements.
4285 (param_splitting_across_edge): Pass the caller to
4286 pull_accesses_from_callee.
4288 2020-01-10 Martin Jambor <mjambor@suse.cz>
4290 * params.opt (param_ipcp_unit_growth): Mark as Optimization.
4291 * ipa-cp.c (max_new_size): Removed.
4292 (orig_overall_size): New variable.
4293 (get_max_overall_size): New function.
4294 (estimate_local_effects): Use it. Adjust dump.
4295 (decide_about_value): Likewise.
4296 (ipcp_propagate_stage): Do not calculate max_new_size, just store
4297 orig_overall_size. Adjust dump.
4298 (ipa_cp_c_finalize): Clear orig_overall_size instead of max_new_size.
4300 2020-01-10 Martin Jambor <mjambor@suse.cz>
4302 * params.opt (param_ipa_max_agg_items): Mark as Optimization
4303 * ipa-cp.c (merge_agg_lats_step): New parameter max_agg_items, use
4304 instead of param_ipa_max_agg_items.
4305 (merge_aggregate_lattices): Extract param_ipa_max_agg_items from
4306 optimization info for the callee.
4308 2020-01-09 Kwok Cheung Yeung <kcy@codesourcery.com>
4310 * lto-streamer-in.c (input_function): Remove streamed-in inline debug
4311 markers if debug_inline_points is false.
4313 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4315 * config.gcc (aarch64*-*-*): Add aarch64-sve-builtins-sve2.o to
4317 * config/aarch64/t-aarch64 (aarch64-sve-builtins.o): Depend on
4318 aarch64-sve-builtins-base.def, aarch64-sve-builtins-sve2.def and
4319 aarch64-sve-builtins-sve2.h.
4320 (aarch64-sve-builtins-sve2.o): New rule.
4321 * config/aarch64/aarch64.h (AARCH64_ISA_SVE2_AES): New macro.
4322 (AARCH64_ISA_SVE2_BITPERM, AARCH64_ISA_SVE2_SHA3): Likewise.
4323 (AARCH64_ISA_SVE2_SM4, TARGET_SVE2_AES, TARGET_SVE2_BITPERM): Likewise.
4324 (TARGET_SVE2_SHA, TARGET_SVE2_SM4): Likewise.
4325 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
4326 TARGET_SVE2_AES, TARGET_SVE2_BITPERM, TARGET_SVE2_SHA3 and
4328 * config/aarch64/aarch64-sve.md: Update comments with SVE2
4329 instructions that are handled here.
4330 (@cond_asrd<mode>): Generalize to...
4331 (@cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>): ...this.
4332 (*cond_asrd<mode>_2): Generalize to...
4333 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_2): ...this.
4334 (*cond_asrd<mode>_z): Generalize to...
4335 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_z): ...this.
4336 * config/aarch64/aarch64.md (UNSPEC_LDNT1_GATHER): New unspec.
4337 (UNSPEC_STNT1_SCATTER, UNSPEC_WHILEGE, UNSPEC_WHILEGT): Likewise.
4338 (UNSPEC_WHILEHI, UNSPEC_WHILEHS): Likewise.
4339 * config/aarch64/aarch64-sve2.md (@aarch64_gather_ldnt<mode>): New
4341 (@aarch64_gather_ldnt_<ANY_EXTEND:optab><SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
4342 (@aarch64_scatter_stnt<mode>): Likewise.
4343 (@aarch64_scatter_stnt_<SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
4344 (@aarch64_mul_lane_<mode>): Likewise.
4345 (@aarch64_sve_suqadd<mode>_const): Likewise.
4346 (*<sur>h<addsub><mode>): Generalize to...
4347 (@aarch64_pred_<SVE2_COND_INT_BINARY_REV:sve_int_op><mode>): ...this
4349 (@cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>): New expander.
4350 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_2): New pattern.
4351 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_3): Likewise.
4352 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_any): Likewise.
4353 (*cond_<SVE2_COND_INT_BINARY_NOREV:sve_int_op><mode>_z): Likewise.
4354 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op><mode>):: Likewise.
4355 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op>_lane_<mode>): Likewise.
4356 (@aarch64_pred_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): Likewise.
4357 (@cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): New expander.
4358 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_2): New pattern.
4359 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_3): Likewise.
4360 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_any): Likewise.
4361 (@aarch64_sve_<SVE2_INT_TERNARY:sve_int_op><mode>): Likewise.
4362 (@aarch64_sve_<SVE2_INT_TERNARY_LANE:sve_int_op>_lane_<mode>)
4363 (@aarch64_sve_add_mul_lane_<mode>): Likewise.
4364 (@aarch64_sve_sub_mul_lane_<mode>): Likewise.
4365 (@aarch64_sve2_xar<mode>): Likewise.
4366 (@aarch64_sve2_bcax<mode>): Likewise.
4367 (*aarch64_sve2_eor3<mode>): Rename to...
4368 (@aarch64_sve2_eor3<mode>): ...this.
4369 (@aarch64_sve2_bsl<mode>): New expander.
4370 (@aarch64_sve2_nbsl<mode>): Likewise.
4371 (@aarch64_sve2_bsl1n<mode>): Likewise.
4372 (@aarch64_sve2_bsl2n<mode>): Likewise.
4373 (@aarch64_sve_add_<SHIFTRT:sve_int_op><mode>): Likewise.
4374 (*aarch64_sve2_sra<mode>): Add MOVPRFX support.
4375 (@aarch64_sve_add_<VRSHR_N:sve_int_op><mode>): New pattern.
4376 (@aarch64_sve_<SVE2_INT_SHIFT_INSERT:sve_int_op><mode>): Likewise.
4377 (@aarch64_sve2_<USMAX:su>aba<mode>): New expander.
4378 (*aarch64_sve2_<USMAX:su>aba<mode>): New pattern.
4379 (@aarch64_sve_<SVE2_INT_BINARY_WIDE:sve_int_op><mode>): Likewise.
4380 (<su>mull<bt><Vwide>): Generalize to...
4381 (@aarch64_sve_<SVE2_INT_BINARY_LONG:sve_int_op><mode>): ...this new
4383 (@aarch64_sve_<SVE2_INT_BINARY_LONG_lANE:sve_int_op>_lane_<mode>)
4384 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_LONG:sve_int_op><mode>)
4385 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG:sve_int_op><mode>)
4386 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
4387 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG:sve_int_op><mode>)
4388 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
4389 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG:sve_int_op><mode>)
4390 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
4391 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG:sve_int_op><mode>)
4392 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
4393 (@aarch64_sve_<SVE2_FP_TERNARY_LONG:sve_fp_op><mode>): New patterns.
4394 (@aarch64_<SVE2_FP_TERNARY_LONG_LANE:sve_fp_op>_lane_<mode>)
4395 (@aarch64_sve_<SVE2_INT_UNARY_NARROWB:sve_int_op><mode>): Likewise.
4396 (@aarch64_sve_<SVE2_INT_UNARY_NARROWT:sve_int_op><mode>): Likewise.
4397 (@aarch64_sve_<SVE2_INT_BINARY_NARROWB:sve_int_op><mode>): Likewise.
4398 (@aarch64_sve_<SVE2_INT_BINARY_NARROWT:sve_int_op><mode>): Likewise.
4399 (<SHRNB:r>shrnb<mode>): Generalize to...
4400 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWB:sve_int_op><mode>): ...this
4402 (<SHRNT:r>shrnt<mode>): Generalize to...
4403 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWT:sve_int_op><mode>): ...this
4405 (@aarch64_pred_<SVE2_INT_BINARY_PAIR:sve_int_op><mode>): New pattern.
4406 (@aarch64_pred_<SVE2_FP_BINARY_PAIR:sve_fp_op><mode>): Likewise.
4407 (@cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>): New expander.
4408 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_2): New pattern.
4409 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_z): Likewise.
4410 (@aarch64_sve_<SVE2_INT_CADD:optab><mode>): Likewise.
4411 (@aarch64_sve_<SVE2_INT_CMLA:optab><mode>): Likewise.
4412 (@aarch64_<SVE2_INT_CMLA:optab>_lane_<mode>): Likewise.
4413 (@aarch64_sve_<SVE2_INT_CDOT:optab><mode>): Likewise.
4414 (@aarch64_<SVE2_INT_CDOT:optab>_lane_<mode>): Likewise.
4415 (@aarch64_pred_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): Likewise.
4416 (@cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New expander.
4417 (*cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New pattern.
4418 (@aarch64_sve2_cvtnt<mode>): Likewise.
4419 (@aarch64_pred_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): Likewise.
4420 (@cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): New expander.
4421 (*cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>_any): New pattern.
4422 (@aarch64_sve2_cvtxnt<mode>): Likewise.
4423 (@aarch64_pred_<SVE2_U32_UNARY:sve_int_op><mode>): Likewise.
4424 (@cond_<SVE2_U32_UNARY:sve_int_op><mode>): New expander.
4425 (*cond_<SVE2_U32_UNARY:sve_int_op><mode>): New pattern.
4426 (@aarch64_pred_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): Likewise.
4427 (@cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New expander.
4428 (*cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New pattern.
4429 (@aarch64_sve2_pmul<mode>): Likewise.
4430 (@aarch64_sve_<SVE2_PMULL:optab><mode>): Likewise.
4431 (@aarch64_sve_<SVE2_PMULL_PAIR:optab><mode>): Likewise.
4432 (@aarch64_sve2_tbl2<mode>): Likewise.
4433 (@aarch64_sve2_tbx<mode>): Likewise.
4434 (@aarch64_sve_<SVE2_INT_BITPERM:sve_int_op><mode>): Likewise.
4435 (@aarch64_sve2_histcnt<mode>): Likewise.
4436 (@aarch64_sve2_histseg<mode>): Likewise.
4437 (@aarch64_pred_<SVE2_MATCH:sve_int_op><mode>): Likewise.
4438 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_cc): Likewise.
4439 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_ptest): Likewise.
4440 (aarch64_sve2_aes<CRYPTO_AES:aes_op>): Likewise.
4441 (aarch64_sve2_aes<CRYPTO_AESMC:aesmc_op>): Likewise.
4442 (*aarch64_sve2_aese_fused, *aarch64_sve2_aesd_fused): Likewise.
4443 (aarch64_sve2_rax1, aarch64_sve2_sm4e, aarch64_sve2_sm4ekey): Likewise.
4444 (<su>mulh<r>s<mode>3): Update after above pattern name changes.
4445 * config/aarch64/iterators.md (VNx16QI_ONLY, VNx4SF_ONLY)
4446 (SVE_STRUCT2, SVE_FULL_BHI, SVE_FULL_HSI, SVE_FULL_HDI)
4447 (SVE2_PMULL_PAIR_I): New mode iterators.
4448 (UNSPEC_ADCLB, UNSPEC_ADCLT, UNSPEC_ADDHNB, UNSPEC_ADDHNT, UNSPEC_BDEP)
4449 (UNSPEC_BEXT, UNSPEC_BGRP, UNSPEC_CADD90, UNSPEC_CADD270, UNSPEC_CDOT)
4450 (UNSPEC_CDOT90, UNSPEC_CDOT180, UNSPEC_CDOT270, UNSPEC_CMLA)
4451 (UNSPEC_CMLA90, UNSPEC_CMLA180, UNSPEC_CMLA270, UNSPEC_COND_FCVTLT)
4452 (UNSPEC_COND_FCVTNT, UNSPEC_COND_FCVTX, UNSPEC_COND_FCVTXNT)
4453 (UNSPEC_COND_FLOGB, UNSPEC_EORBT, UNSPEC_EORTB, UNSPEC_FADDP)
4454 (UNSPEC_FMAXP, UNSPEC_FMAXNMP, UNSPEC_FMLALB, UNSPEC_FMLALT)
4455 (UNSPEC_FMLSLB, UNSPEC_FMLSLT, UNSPEC_FMINP, UNSPEC_FMINNMP)
4456 (UNSPEC_HISTCNT, UNSPEC_HISTSEG, UNSPEC_MATCH, UNSPEC_NMATCH)
4457 (UNSPEC_PMULLB, UNSPEC_PMULLB_PAIR, UNSPEC_PMULLT, UNSPEC_PMULLT_PAIR)
4458 (UNSPEC_RADDHNB, UNSPEC_RADDHNT, UNSPEC_RSUBHNB, UNSPEC_RSUBHNT)
4459 (UNSPEC_SLI, UNSPEC_SRI, UNSPEC_SABDLB, UNSPEC_SABDLT, UNSPEC_SADDLB)
4460 (UNSPEC_SADDLBT, UNSPEC_SADDLT, UNSPEC_SADDWB, UNSPEC_SADDWT)
4461 (UNSPEC_SBCLB, UNSPEC_SBCLT, UNSPEC_SMAXP, UNSPEC_SMINP)
4462 (UNSPEC_SQCADD90, UNSPEC_SQCADD270, UNSPEC_SQDMULLB, UNSPEC_SQDMULLBT)
4463 (UNSPEC_SQDMULLT, UNSPEC_SQRDCMLAH, UNSPEC_SQRDCMLAH90)
4464 (UNSPEC_SQRDCMLAH180, UNSPEC_SQRDCMLAH270, UNSPEC_SQRSHRNB)
4465 (UNSPEC_SQRSHRNT, UNSPEC_SQRSHRUNB, UNSPEC_SQRSHRUNT, UNSPEC_SQSHRNB)
4466 (UNSPEC_SQSHRNT, UNSPEC_SQSHRUNB, UNSPEC_SQSHRUNT, UNSPEC_SQXTNB)
4467 (UNSPEC_SQXTNT, UNSPEC_SQXTUNB, UNSPEC_SQXTUNT, UNSPEC_SSHLLB)
4468 (UNSPEC_SSHLLT, UNSPEC_SSUBLB, UNSPEC_SSUBLBT, UNSPEC_SSUBLT)
4469 (UNSPEC_SSUBLTB, UNSPEC_SSUBWB, UNSPEC_SSUBWT, UNSPEC_SUBHNB)
4470 (UNSPEC_SUBHNT, UNSPEC_TBL2, UNSPEC_UABDLB, UNSPEC_UABDLT)
4471 (UNSPEC_UADDLB, UNSPEC_UADDLT, UNSPEC_UADDWB, UNSPEC_UADDWT)
4472 (UNSPEC_UMAXP, UNSPEC_UMINP, UNSPEC_UQRSHRNB, UNSPEC_UQRSHRNT)
4473 (UNSPEC_UQSHRNB, UNSPEC_UQSHRNT, UNSPEC_UQXTNB, UNSPEC_UQXTNT)
4474 (UNSPEC_USHLLB, UNSPEC_USHLLT, UNSPEC_USUBLB, UNSPEC_USUBLT)
4475 (UNSPEC_USUBWB, UNSPEC_USUBWT): New unspecs.
4476 (UNSPEC_SMULLB, UNSPEC_SMULLT, UNSPEC_UMULLB, UNSPEC_UMULLT)
4477 (UNSPEC_SMULHS, UNSPEC_SMULHRS, UNSPEC_UMULHS, UNSPEC_UMULHRS)
4478 (UNSPEC_RSHRNB, UNSPEC_RSHRNT, UNSPEC_SHRNB, UNSPEC_SHRNT): Move
4480 (VNARROW, Ventype): New mode attributes.
4481 (Vewtype): Handle VNx2DI. Fix typo in comment.
4482 (VDOUBLE): New mode attribute.
4483 (sve_lane_con): Handle VNx8HI.
4484 (SVE_INT_UNARY): Include ss_abs and ss_neg for TARGET_SVE2.
4485 (SVE_INT_BINARY): Likewise ss_plus, us_plus, ss_minus and us_minus.
4486 (sve_int_op, sve_int_op_rev): Handle the above codes.
4487 (sve_pred_int_rhs2_operand): Likewise.
4488 (MULLBT, SHRNB, SHRNT): Delete.
4489 (SVE_INT_SHIFT_IMM): New int iterator.
4490 (SVE_WHILE): Add UNSPEC_WHILEGE, UNSPEC_WHILEGT, UNSPEC_WHILEHI
4491 and UNSPEC_WHILEHS for TARGET_SVE2.
4492 (SVE2_U32_UNARY, SVE2_INT_UNARY_NARROWB, SVE2_INT_UNARY_NARROWT)
4493 (SVE2_INT_BINARY, SVE2_INT_BINARY_LANE, SVE2_INT_BINARY_LONG)
4494 (SVE2_INT_BINARY_LONG_LANE, SVE2_INT_BINARY_NARROWB)
4495 (SVE2_INT_BINARY_NARROWT, SVE2_INT_BINARY_PAIR, SVE2_FP_BINARY_PAIR)
4496 (SVE2_INT_BINARY_PAIR_LONG, SVE2_INT_BINARY_WIDE): New int iterators.
4497 (SVE2_INT_SHIFT_IMM_LONG, SVE2_INT_SHIFT_IMM_NARROWB): Likewise.
4498 (SVE2_INT_SHIFT_IMM_NARROWT, SVE2_INT_SHIFT_INSERT, SVE2_INT_CADD)
4499 (SVE2_INT_BITPERM, SVE2_INT_TERNARY, SVE2_INT_TERNARY_LANE): Likewise.
4500 (SVE2_FP_TERNARY_LONG, SVE2_FP_TERNARY_LONG_LANE, SVE2_INT_CMLA)
4501 (SVE2_INT_CDOT, SVE2_INT_ADD_BINARY_LONG, SVE2_INT_QADD_BINARY_LONG)
4502 (SVE2_INT_SUB_BINARY_LONG, SVE2_INT_QSUB_BINARY_LONG): Likewise.
4503 (SVE2_INT_ADD_BINARY_LONG_LANE, SVE2_INT_QADD_BINARY_LONG_LANE)
4504 (SVE2_INT_SUB_BINARY_LONG_LANE, SVE2_INT_QSUB_BINARY_LONG_LANE)
4505 (SVE2_COND_INT_UNARY_FP, SVE2_COND_FP_UNARY_LONG): Likewise.
4506 (SVE2_COND_FP_UNARY_NARROWB, SVE2_COND_INT_BINARY): Likewise.
4507 (SVE2_COND_INT_BINARY_NOREV, SVE2_COND_INT_BINARY_REV): Likewise.
4508 (SVE2_COND_INT_SHIFT, SVE2_MATCH, SVE2_PMULL): Likewise.
4509 (optab): Handle the new unspecs.
4510 (su, r): Remove entries for UNSPEC_SHRNB, UNSPEC_SHRNT, UNSPEC_RSHRNB
4512 (lr): Handle the new unspecs.
4514 (cmp_op, while_optab_cmp, sve_int_op): Handle the new unspecs.
4515 (sve_int_op_rev, sve_int_add_op, sve_int_qadd_op, sve_int_sub_op)
4516 (sve_int_qsub_op): New int attributes.
4517 (sve_fp_op, rot): Handle the new unspecs.
4518 * config/aarch64/aarch64-sve-builtins.h
4519 (function_resolver::require_matching_pointer_type): Declare.
4520 (function_resolver::resolve_unary): Add an optional boolean argument.
4521 (function_resolver::finish_opt_n_resolution): Add an optional
4522 type_suffix_index argument.
4523 (gimple_folder::redirect_call): Declare.
4524 (gimple_expander::prepare_gather_address_operands): Add an optional
4526 * config/aarch64/aarch64-sve-builtins.cc: Include
4527 aarch64-sve-builtins-sve2.h.
4528 (TYPES_b_unsigned, TYPES_b_integer, TYPES_bh_integer): New macros.
4529 (TYPES_bs_unsigned, TYPES_hs_signed, TYPES_hs_integer): Likewise.
4530 (TYPES_hd_unsigned, TYPES_hsd_signed): Likewise.
4531 (TYPES_hsd_integer): Use TYPES_hsd_signed.
4532 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): New macros.
4533 (TYPES_s_unsigned): Likewise.
4534 (TYPES_s_integer): Use TYPES_s_unsigned.
4535 (TYPES_sd_signed, TYPES_sd_unsigned): New macros.
4536 (TYPES_sd_integer): Use them.
4537 (TYPES_d_unsigned): New macro.
4538 (TYPES_d_integer): Use it.
4539 (TYPES_d_data, TYPES_cvt_long, TYPES_cvt_narrow_s): New macros.
4540 (TYPES_cvt_narrow): Likewise.
4541 (DEF_SVE_TYPES_ARRAY): Include the new types macros above.
4542 (preds_mx): New variable.
4543 (function_builder::add_overloaded_function): Allow the new feature
4544 set to be more restrictive than the original one.
4545 (function_resolver::infer_pointer_type): Remove qualifiers from
4546 the pointer type before printing it.
4547 (function_resolver::require_matching_pointer_type): New function.
4548 (function_resolver::resolve_sv_displacement): Handle functions
4549 that don't support 32-bit vector indices or svint32_t vector offsets.
4550 (function_resolver::finish_opt_n_resolution): Take the inferred type
4551 as a separate argument.
4552 (function_resolver::resolve_unary): Optionally treat all forms in
4553 the same way as normal merging functions.
4554 (gimple_folder::redirect_call): New function.
4555 (function_expander::prepare_gather_address_operands): Add an argument
4556 that says whether scaled forms are available. If they aren't,
4557 handle scaling of vector indices and don't add the extension and
4559 (function_expander::map_to_unspecs): If aarch64_sve isn't available,
4560 fall back to using cond_* instead.
4561 * config/aarch64/aarch64-sve-builtins-functions.h (rtx_code_function):
4562 Split out the member variables into...
4563 (rtx_code_function_base): ...this new base class.
4564 (rtx_code_function_rotated): Inherit rtx_code_function_base.
4565 (unspec_based_function): Split out the member variables into...
4566 (unspec_based_function_base): ...this new base class.
4567 (unspec_based_function_rotated): Inherit unspec_based_function_base.
4568 (unspec_based_function_exact_insn): New class.
4569 (unspec_based_add_function, unspec_based_add_lane_function)
4570 (unspec_based_lane_function, unspec_based_pred_function)
4571 (unspec_based_qadd_function, unspec_based_qadd_lane_function)
4572 (unspec_based_qsub_function, unspec_based_qsub_lane_function)
4573 (unspec_based_sub_function, unspec_based_sub_lane_function): New
4575 (unspec_based_fused_function): New class.
4576 (unspec_based_mla_function, unspec_based_mls_function): New typedefs.
4577 (unspec_based_fused_lane_function): New class.
4578 (unspec_based_mla_lane_function, unspec_based_mls_lane_function): New
4580 (CODE_FOR_MODE1): New macro.
4581 (fixed_insn_function): New class.
4582 (while_comparison): Likewise.
4583 * config/aarch64/aarch64-sve-builtins-shapes.h (binary_long_lane)
4584 (binary_long_opt_n, binary_narrowb_opt_n, binary_narrowt_opt_n)
4585 (binary_to_uint, binary_wide, binary_wide_opt_n, compare, compare_ptr)
4586 (load_ext_gather_index_restricted, load_ext_gather_offset_restricted)
4587 (load_gather_sv_restricted, shift_left_imm_long): Declare.
4588 (shift_left_imm_to_uint, shift_right_imm_narrowb): Likewise.
4589 (shift_right_imm_narrowt, shift_right_imm_narrowb_to_uint): Likewise.
4590 (shift_right_imm_narrowt_to_uint, store_scatter_index_restricted)
4591 (store_scatter_offset_restricted, tbl_tuple, ternary_long_lane)
4592 (ternary_long_opt_n, ternary_qq_lane_rotate, ternary_qq_rotate)
4593 (ternary_shift_left_imm, ternary_shift_right_imm, ternary_uint)
4594 (unary_convert_narrowt, unary_long, unary_narrowb, unary_narrowt)
4595 (unary_narrowb_to_uint, unary_narrowt_to_uint, unary_to_int): Likewise.
4596 * config/aarch64/aarch64-sve-builtins-shapes.cc (apply_predication):
4597 Also add an initial argument for unary_convert_narrowt, regardless
4598 of the predication type.
4599 (build_32_64): Allow loads and stores to specify MODE_none.
4600 (build_sv_index64, build_sv_uint_offset): New functions.
4601 (long_type_suffix): New function.
4602 (binary_imm_narrowb_base, binary_imm_narrowt_base): New classes.
4603 (binary_imm_long_base, load_gather_sv_base): Likewise.
4604 (shift_right_imm_narrow_wrapper, ternary_shift_imm_base): Likewise.
4605 (ternary_resize2_opt_n_base, ternary_resize2_lane_base): Likewise.
4606 (unary_narrowb_base, unary_narrowt_base): Likewise.
4607 (binary_long_lane_def, binary_long_lane): New shape.
4608 (binary_long_opt_n_def, binary_long_opt_n): Likewise.
4609 (binary_narrowb_opt_n_def, binary_narrowb_opt_n): Likewise.
4610 (binary_narrowt_opt_n_def, binary_narrowt_opt_n): Likewise.
4611 (binary_to_uint_def, binary_to_uint): Likewise.
4612 (binary_wide_def, binary_wide): Likewise.
4613 (binary_wide_opt_n_def, binary_wide_opt_n): Likewise.
4614 (compare_def, compare): Likewise.
4615 (compare_ptr_def, compare_ptr): Likewise.
4616 (load_ext_gather_index_restricted_def,
4617 load_ext_gather_index_restricted): Likewise.
4618 (load_ext_gather_offset_restricted_def,
4619 load_ext_gather_offset_restricted): Likewise.
4620 (load_gather_sv_def): Inherit from load_gather_sv_base.
4621 (load_gather_sv_restricted_def, load_gather_sv_restricted): New shape.
4622 (shift_left_imm_def, shift_left_imm): Likewise.
4623 (shift_left_imm_long_def, shift_left_imm_long): Likewise.
4624 (shift_left_imm_to_uint_def, shift_left_imm_to_uint): Likewise.
4625 (store_scatter_index_restricted_def,
4626 store_scatter_index_restricted): Likewise.
4627 (store_scatter_offset_restricted_def,
4628 store_scatter_offset_restricted): Likewise.
4629 (tbl_tuple_def, tbl_tuple): Likewise.
4630 (ternary_long_lane_def, ternary_long_lane): Likewise.
4631 (ternary_long_opt_n_def, ternary_long_opt_n): Likewise.
4632 (ternary_qq_lane_def): Inherit from ternary_resize2_lane_base.
4633 (ternary_qq_lane_rotate_def, ternary_qq_lane_rotate): New shape
4634 (ternary_qq_opt_n_def): Inherit from ternary_resize2_opt_n_base.
4635 (ternary_qq_rotate_def, ternary_qq_rotate): New shape.
4636 (ternary_shift_left_imm_def, ternary_shift_left_imm): Likewise.
4637 (ternary_shift_right_imm_def, ternary_shift_right_imm): Likewise.
4638 (ternary_uint_def, ternary_uint): Likewise.
4639 (unary_convert): Fix typo in comment.
4640 (unary_convert_narrowt_def, unary_convert_narrowt): New shape.
4641 (unary_long_def, unary_long): Likewise.
4642 (unary_narrowb_def, unary_narrowb): Likewise.
4643 (unary_narrowt_def, unary_narrowt): Likewise.
4644 (unary_narrowb_to_uint_def, unary_narrowb_to_uint): Likewise.
4645 (unary_narrowt_to_uint_def, unary_narrowt_to_uint): Likewise.
4646 (unary_to_int_def, unary_to_int): Likewise.
4647 * config/aarch64/aarch64-sve-builtins-base.cc (unspec_cmla)
4648 (unspec_fcmla, unspec_cond_fcmla, expand_mla_mls_lane): New functions.
4649 (svasrd_impl): Delete.
4650 (svcadd_impl::expand): Handle integer operations too.
4651 (svcmla_impl::expand, svcmla_lane::expand): Likewise, using the
4652 new functions to derive the unspec numbers.
4653 (svmla_svmls_lane_impl): Replace with...
4654 (svmla_lane_impl, svmls_lane_impl): ...these new classes. Handle
4655 integer operations too.
4656 (svwhile_impl): Rename to...
4657 (svwhilelx_impl): ...this and inherit from while_comparison.
4658 (svasrd): Use unspec_based_function.
4659 (svmla_lane): Use svmla_lane_impl.
4660 (svmls_lane): Use svmls_lane_impl.
4661 (svrecpe, svrsqrte): Handle unsigned integer operations too.
4662 (svwhilele, svwhilelt): Use svwhilelx_impl.
4663 * config/aarch64/aarch64-sve-builtins-sve2.h: New file.
4664 * config/aarch64/aarch64-sve-builtins-sve2.cc: Likewise.
4665 * config/aarch64/aarch64-sve-builtins-sve2.def: Likewise.
4666 * config/aarch64/aarch64-sve-builtins.def: Include
4667 aarch64-sve-builtins-sve2.def.
4669 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4671 * config/aarch64/aarch64-protos.h (aarch64_sve_arith_immediate_p)
4672 (aarch64_sve_sqadd_sqsub_immediate_p): Add a machine_mode argument.
4673 * config/aarch64/aarch64.c (aarch64_sve_arith_immediate_p)
4674 (aarch64_sve_sqadd_sqsub_immediate_p): Likewise. Handle scalar
4675 immediates as well as vector ones.
4676 * config/aarch64/predicates.md (aarch64_sve_arith_immediate)
4677 (aarch64_sve_sub_arith_immediate, aarch64_sve_qadd_immediate)
4678 (aarch64_sve_qsub_immediate): Update calls accordingly.
4680 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4682 * config/aarch64/aarch64-sve2.md: Add banner comments.
4683 (<su>mulh<r>s<mode>3): Move further up file.
4684 (<su>mull<bt><Vwide>, <r>shrnb<mode>, <r>shrnt<mode>)
4685 (*aarch64_sve2_sra<mode>): Move further down file.
4686 * config/aarch64/t-aarch64 (s-check-sve-md): Check aarch64-sve2.md too.
4688 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4690 * config/aarch64/iterators.md (SVE_WHILE): Add UNSPEC_WHILERW
4692 (while_optab_cmp): Handle them.
4693 * config/aarch64/aarch64-sve.md
4694 (*while_<while_optab_cmp><GPI:mode><PRED_ALL:mode>_ptest): Make public
4695 and add a "@" marker.
4696 * config/aarch64/aarch64-sve2.md (check_<raw_war>_ptrs<mode>): Use it
4697 instead of gen_aarch64_sve2_while_ptest.
4698 (@aarch64_sve2_while<cmp_op><GPI:mode><PRED_ALL:mode>_ptest): Delete.
4700 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4702 * config/aarch64/aarch64.md (UNSPEC_WHILE_LE): Rename to...
4703 (UNSPEC_WHILELE): ...this.
4704 (UNSPEC_WHILE_LO): Rename to...
4705 (UNSPEC_WHILELO): ...this.
4706 (UNSPEC_WHILE_LS): Rename to...
4707 (UNSPEC_WHILELS): ...this.
4708 (UNSPEC_WHILE_LT): Rename to...
4709 (UNSPEC_WHILELT): ...this.
4710 * config/aarch64/iterators.md (SVE_WHILE): Update accordingly.
4711 (cmp_op, while_optab_cmp): Likewise.
4712 * config/aarch64/aarch64.c (aarch64_sve_move_pred_via_while): Likewise.
4713 * config/aarch64/aarch64-sve-builtins-base.cc (svwhilele): Likewise.
4714 (svwhilelt): Likewise.
4716 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4718 * config/aarch64/aarch64-sve-builtins-shapes.h (unary_count): Delete.
4719 (unary_to_uint): Define.
4720 * config/aarch64/aarch64-sve-builtins-shapes.cc (unary_count_def)
4721 (unary_count): Rename to...
4722 (unary_to_uint_def, unary_to_uint): ...this.
4723 * config/aarch64/aarch64-sve-builtins-base.def: Update accordingly.
4725 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4727 * config/aarch64/aarch64-sve-builtins-functions.h
4728 (code_for_mode_function): New class.
4729 (CODE_FOR_MODE0, QUIET_CODE_FOR_MODE0): New macros.
4730 * config/aarch64/aarch64-sve-builtins-base.cc (svcompact_impl)
4731 (svext_impl, svmul_lane_impl, svsplice_impl, svtmad_impl): Delete.
4732 (svcompact, svext, svsplice): Use QUIET_CODE_FOR_MODE0.
4733 (svmul_lane, svtmad): Use CODE_FOR_MODE0.
4735 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4737 * config/aarch64/iterators.md (addsub): New code attribute.
4738 * config/aarch64/aarch64-simd.md (aarch64_<su_optab><optab><mode>):
4740 (aarch64_<su_optab>q<addsub><mode>): ...this, making the same change
4741 in the asm string and attributes. Fix indentation.
4742 * config/aarch64/aarch64-sve.md (@aarch64_<su_optab><optab><mode>):
4744 (@aarch64_sve_<optab><mode>): ...this.
4745 * config/aarch64/aarch64-sve-builtins.h
4746 (function_expander::expand_signed_unpred_op): Delete.
4747 * config/aarch64/aarch64-sve-builtins.cc
4748 (function_expander::expand_signed_unpred_op): Likewise.
4749 (function_expander::map_to_rtx_codes): If the optab isn't defined,
4750 try using code_for_aarch64_sve instead.
4751 * config/aarch64/aarch64-sve-builtins-base.cc (svqadd_impl): Delete.
4752 (svqsub_impl): Likewise.
4753 (svqadd, svqsub): Use rtx_code_function instead.
4755 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4757 * config/aarch64/iterators.md (SRHSUB, URHSUB): Delete.
4758 (HADDSUB, sur, addsub): Remove them.
4760 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4762 * tree-nrv.c (pass_return_slot::execute): Handle all internal
4763 functions the same way, rather than singling out those that
4764 aren't mapped directly to optabs.
4766 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4768 * target.def (compatible_vector_types_p): New target hook.
4769 * hooks.h (hook_bool_const_tree_const_tree_true): Declare.
4770 * hooks.c (hook_bool_const_tree_const_tree_true): New function.
4771 * doc/tm.texi.in (TARGET_COMPATIBLE_VECTOR_TYPES_P): New hook.
4772 * doc/tm.texi: Regenerate.
4773 * gimple-expr.c: Include target.h.
4774 (useless_type_conversion_p): Use targetm.compatible_vector_types_p.
4775 * config/aarch64/aarch64.c (aarch64_compatible_vector_types_p): New
4777 (TARGET_COMPATIBLE_VECTOR_TYPES_P): Define.
4778 * config/aarch64/aarch64-sve-builtins.cc (gimple_folder::convert_pred):
4779 Use the original predicate if it already has a suitable type.
4781 2020-01-09 Martin Jambor <mjambor@suse.cz>
4783 * cgraph.h (cgraph_edge): Make remove, set_call_stmt, make_direct,
4784 resolve_speculation and redirect_call_stmt_to_callee static. Change
4785 return type of set_call_stmt to cgraph_edge *.
4786 * auto-profile.c (afdo_indirect_call): Adjust call to
4787 redirect_call_stmt_to_callee.
4788 * cgraph.c (cgraph_edge::set_call_stmt): Make return cgraph-edge *,
4789 make the this pointer explicit, adjust self-recursive calls and the
4790 call top make_direct. Return the resulting edge.
4791 (cgraph_edge::remove): Make this pointer explicit.
4792 (cgraph_edge::resolve_speculation): Likewise, adjust call to remove.
4793 (cgraph_edge::make_direct): Likewise, adjust call to
4794 resolve_speculation.
4795 (cgraph_edge::redirect_call_stmt_to_callee): Likewise, also adjust
4796 call to set_call_stmt.
4797 (cgraph_update_edges_for_call_stmt_node): Update call to
4798 set_call_stmt and remove.
4799 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
4800 Renamed edge to master_edge. Adjusted calls to set_call_stmt.
4801 (cgraph_node::create_edge_including_clones): Moved "first" definition
4802 of edge to the block where it was used. Adjusted calls to
4804 (cgraph_node::remove_symbol_and_inline_clones): Adjust call to
4805 cgraph_edge::remove.
4806 * cgraphunit.c (walk_polymorphic_call_targets): Adjusted calls to
4807 make_direct and redirect_call_stmt_to_callee.
4808 * ipa-fnsummary.c (redirect_to_unreachable): Adjust calls to
4809 resolve_speculation and make_direct.
4810 * ipa-inline-transform.c (inline_transform): Adjust call to
4811 redirect_call_stmt_to_callee.
4812 (check_speculations_1):: Adjust call to resolve_speculation.
4813 * ipa-inline.c (resolve_noninline_speculation): Adjust call to
4814 resolve-speculation.
4815 (inline_small_functions): Adjust call to resolve_speculation.
4816 (ipa_inline): Likewise.
4817 * ipa-prop.c (ipa_make_edge_direct_to_target): Adjust call to
4819 * ipa-visibility.c (function_and_variable_visibility): Make iteration
4820 safe with regards to edge removal, adjust calls to
4821 redirect_call_stmt_to_callee.
4822 * ipa.c (walk_polymorphic_call_targets): Adjust calls to make_direct
4823 and redirect_call_stmt_to_callee.
4824 * multiple_target.c (create_dispatcher_calls): Adjust call to
4825 redirect_call_stmt_to_callee
4826 (redirect_to_specific_clone): Likewise.
4827 * tree-cfgcleanup.c (delete_unreachable_blocks_update_callgraph):
4828 Adjust calls to cgraph_edge::remove.
4829 * tree-inline.c (copy_bb): Adjust call to set_call_stmt.
4830 (redirect_all_calls): Adjust call to redirect_call_stmt_to_callee.
4831 (expand_call_inline): Adjust call to cgraph_edge::remove.
4833 2020-01-09 Martin Liska <mliska@suse.cz>
4835 * params.opt: Set Optimization for
4836 param_max_speculative_devirt_maydefs.
4838 2020-01-09 Martin Sebor <msebor@redhat.com>
4842 * builtins.c (compute_objsize): Avoid handling MEM_REFs of vector type.
4844 2020-01-09 Martin Liska <mliska@suse.cz>
4846 * auto-profile.c (auto_profile): Use opt_for_fn
4848 * ipa-cp.c (ipcp_lattice::add_value): Likewise.
4849 (propagate_vals_across_arith_jfunc): Likewise.
4850 (hint_time_bonus): Likewise.
4851 (incorporate_penalties): Likewise.
4852 (good_cloning_opportunity_p): Likewise.
4853 (perform_estimation_of_a_value): Likewise.
4854 (estimate_local_effects): Likewise.
4855 (ipcp_propagate_stage): Likewise.
4856 * ipa-fnsummary.c (decompose_param_expr): Likewise.
4857 (set_switch_stmt_execution_predicate): Likewise.
4858 (analyze_function_body): Likewise.
4859 * ipa-inline-analysis.c (offline_size): Likewise.
4860 * ipa-inline.c (early_inliner): Likewise.
4861 * ipa-prop.c (ipa_analyze_node): Likewise.
4862 (ipcp_transform_function): Likewise.
4863 * ipa-sra.c (process_scan_results): Likewise.
4864 (ipa_sra_summarize_function): Likewise.
4865 * params.opt: Rename ipcp-unit-growth to
4866 ipa-cp-unit-growth. Add Optimization for various
4867 IPA-related parameters.
4869 2020-01-09 Richard Biener <rguenther@suse.de>
4872 * gimplify.c (gimplify_expr): Deal with NOP definitions.
4874 2020-01-09 Richard Biener <rguenther@suse.de>
4876 PR tree-optimization/93040
4877 * gimple-ssa-store-merging.c (find_bswap_or_nop): Raise search limit.
4879 2020-01-09 Georg-Johann Lay <avr@gjlay.de>
4881 * common/config/avr/avr-common.c (avr_option_optimization_table)
4882 [OPT_LEVELS_1_PLUS]: Set -fsplit-wide-types-early.
4884 2020-01-09 Martin Liska <mliska@suse.cz>
4886 * cgraphclones.c (symbol_table::materialize_all_clones):
4887 Use cgraph_node::dump_name.
4889 2020-01-09 Jakub Jelinek <jakub@redhat.com>
4892 * config/riscv/riscv.c (riscv_print_operand_reloc): Use
4893 output_operand_lossage instead of gcc_unreachable.
4894 * doc/md.texi (riscv f constraint): Fix typo.
4897 * config/i386/i386.md (subv<mode>4): Use SWIDWI iterator instead of
4898 SWI. Use <general_hilo_operand> instead of <general_operand>. Use
4899 CONST_SCALAR_INT_P instead of CONST_INT_P.
4900 (*subv<mode>4_1): Rename to ...
4901 (subv<mode>4_1): ... this.
4902 (*subv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
4903 define_insn_and_split patterns.
4904 (*subv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
4907 2020-01-08 David Malcolm <dmalcolm@redhat.com>
4909 * vec.c (class selftest::count_dtor): New class.
4910 (selftest::test_auto_delete_vec): New test.
4911 (selftest::vec_c_tests): Call it.
4912 * vec.h (class auto_delete_vec): New class template.
4913 (auto_delete_vec<T>::~auto_delete_vec): New dtor.
4915 2020-01-08 David Malcolm <dmalcolm@redhat.com>
4917 * sbitmap.h (auto_sbitmap): Add operator const_sbitmap.
4919 2020-01-08 Jim Wilson <jimw@sifive.com>
4921 * config/riscv/riscv.c (riscv_legitimize_tls_address): Ifdef out
4922 use of TLS_MODEL_LOCAL_EXEC when not pic.
4924 2020-01-08 David Malcolm <dmalcolm@redhat.com>
4926 * hash-map-tests.c (selftest::test_map_of_strings_to_int): Fix
4929 2020-01-08 Jakub Jelinek <jakub@redhat.com>
4932 * config/i386/i386.md (*stack_protect_set_2_<mode> peephole2,
4933 *stack_protect_set_3 peephole2): Also check that the second
4934 insns source is general_operand.
4937 * config/i386/i386.md (addcarry<mode>_0): Use nonimmediate_operand
4938 predicate for output operand instead of register_operand.
4939 (addcarry<mode>, addcarry<mode>_1): Likewise. Add alternative with
4940 memory destination and non-memory operands[2].
4942 2020-01-08 Martin Liska <mliska@suse.cz>
4944 * cgraph.c (cgraph_node::dump): Use ::dump_name or
4945 ::dump_asm_name instead of (::name or ::asm_name).
4946 * cgraphclones.c (symbol_table::materialize_all_clones): Likewise.
4947 * cgraphunit.c (walk_polymorphic_call_targets): Likewise.
4948 (analyze_functions): Likewise.
4949 (expand_all_functions): Likewise.
4950 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
4951 (propagate_bits_across_jump_function): Likewise.
4952 (dump_profile_updates): Likewise.
4953 (ipcp_store_bits_results): Likewise.
4954 (ipcp_store_vr_results): Likewise.
4955 * ipa-devirt.c (dump_targets): Likewise.
4956 * ipa-fnsummary.c (analyze_function_body): Likewise.
4957 * ipa-hsa.c (check_warn_node_versionable): Likewise.
4958 (process_hsa_functions): Likewise.
4959 * ipa-icf.c (sem_item_optimizer::merge_classes): Likewise.
4960 (set_alias_uids): Likewise.
4961 * ipa-inline-transform.c (save_inline_function_body): Likewise.
4962 * ipa-inline.c (recursive_inlining): Likewise.
4963 (inline_to_all_callers_1): Likewise.
4964 (ipa_inline): Likewise.
4965 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
4966 (ipa_propagate_frequency): Likewise.
4967 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
4968 (remove_described_reference): Likewise.
4969 * ipa-pure-const.c (worse_state): Likewise.
4970 (check_retval_uses): Likewise.
4971 (analyze_function): Likewise.
4972 (propagate_pure_const): Likewise.
4973 (propagate_nothrow): Likewise.
4974 (dump_malloc_lattice): Likewise.
4975 (propagate_malloc): Likewise.
4976 (pass_local_pure_const::execute): Likewise.
4977 * ipa-visibility.c (optimize_weakref): Likewise.
4978 (function_and_variable_visibility): Likewise.
4979 * ipa.c (symbol_table::remove_unreachable_nodes): Likewise.
4980 (ipa_discover_variable_flags): Likewise.
4981 * lto-streamer-out.c (output_function): Likewise.
4982 (output_constructor): Likewise.
4983 * tree-inline.c (copy_bb): Likewise.
4984 * tree-ssa-structalias.c (ipa_pta_execute): Likewise.
4985 * varpool.c (symbol_table::remove_unreferenced_decls): Likewise.
4987 2020-01-08 Richard Biener <rguenther@suse.de>
4990 * tree-eh.c (sink_clobbers): Update virtual operands for
4991 the first and last stmt only. Add a dry-run capability.
4992 (pass_lower_eh_dispatch::execute): Perform clobber sinking
4993 after CFG manipulations and in RPO order to catch all
4994 secondary opportunities reliably.
4996 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
4999 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
5001 2019-01-08 Richard Biener <rguenther@suse.de>
5004 * gimple-fold.c (rewrite_to_defined_overflow): Mark stmt modified.
5005 * tree-ssa-loop-im.c (move_computations_worker): Properly adjust
5006 virtual operand, also updating SSA use.
5007 * gimple-loop-interchange.cc (loop_cand::undo_simple_reduction):
5008 Update stmt after resetting virtual operand.
5009 (tree_loop_interchange::move_code_to_inner_loop): Likewise.
5010 * gimple-iterator.c (gsi_remove): When not removing the stmt
5011 permanently do not delink immediate uses or mark the stmt modified.
5013 2020-01-08 Martin Liska <mliska@suse.cz>
5015 * ipa-fnsummary.c (dump_ipa_call_summary): Use symtab_node::dump_name.
5016 (ipa_call_context::estimate_size_and_time): Likewise.
5017 (inline_analyze_function): Likewise.
5019 2020-01-08 Martin Liska <mliska@suse.cz>
5021 * cgraph.c (cgraph_node::dump): Use systematically
5024 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
5026 Add -nodevicespecs option for avr.
5029 * config/avr/avr.opt (-nodevicespecs): New driver option.
5030 * config/avr/driver-avr.c (avr_devicespecs_file): Only issue
5031 "-specs=device-specs/..." if that option is not set.
5032 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
5034 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
5036 Implement 64-bit double functions for avr.
5039 * config.gcc (tm_defines) [target=avr]: Support --with-libf7,
5040 --with-double-comparison.
5041 * doc/install.texi: Document them.
5042 * config/avr/avr-c.c (avr_cpu_cpp_builtins)
5043 <WITH_LIBF7_LIBGCC, WITH_LIBF7_MATH, WITH_LIBF7_MATH_SYMBOLS>
5044 <WITH_DOUBLE_COMPARISON>: New built-in defines.
5045 * doc/invoke.texi (AVR Built-in Macros): Document them.
5046 * config/avr/avr-protos.h (avr_float_lib_compare_returns_bool): New.
5047 * config/avr/avr.c (avr_float_lib_compare_returns_bool): New function.
5048 * config/avr/avr.h (FLOAT_LIB_COMPARE_RETURNS_BOOL): New macro.
5050 2020-01-08 Richard Earnshaw <rearnsha@arm.com>
5053 * config/arm/t-multilib (MULTILIB_MATCHES): Add rules to match
5054 armv7-a{+mp,+sec,+mp+sec} to appropriate armv7 multilib variants
5055 when only building rm-profile multilibs.
5057 2020-01-08 Feng Xue <fxue@os.amperecomputing.com>
5060 * ipa-cp.c (self_recursively_generated_p): Find matched aggregate
5061 lattice for a value to check.
5062 (propagate_vals_across_arith_jfunc): Add an assertion to ensure
5063 finite propagation in self-recursive scc.
5065 2020-01-08 Luo Xiong Hu <luoxhu@linux.ibm.com>
5067 * ipa-inline.c (caller_growth_limits): Restore the AND.
5069 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
5071 * config/gcn/gcn-valu.md (VEC_1REG_INT_ALT): Delete iterator.
5072 (VEC_ALLREG_ALT): New iterator.
5073 (VEC_ALLREG_INT_MODE): New iterator.
5074 (VCMP_MODE): New iterator.
5075 (VCMP_MODE_INT): New iterator.
5076 (vec_cmpu<mode>di): Use VCMP_MODE_INT.
5077 (vec_cmp<u>v64qidi): New define_expand.
5078 (vec_cmp<mode>di_exec): Use VCMP_MODE.
5079 (vec_cmpu<mode>di_exec): New define_expand.
5080 (vec_cmp<u>v64qidi_exec): New define_expand.
5081 (vec_cmp<mode>di_dup): Use VCMP_MODE.
5082 (vec_cmp<mode>di_dup_exec): Use VCMP_MODE.
5083 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>): Rename ...
5084 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): ... to this.
5085 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>_exec): Rename ...
5086 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): ... to this.
5087 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>): Rename ...
5088 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): ... to this.
5089 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>_exec): Rename ...
5090 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): ... to
5092 * config/gcn/gcn.c (print_operand): Fix 8 and 16 bit suffixes.
5093 * config/gcn/gcn.md (expander): Add sign_extend and zero_extend.
5095 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
5097 * config/gcn/constraints.md (DA): Update description and match.
5099 (Db): New constraint.
5100 * config/gcn/gcn-protos.h (gcn_inline_constant64_p): Add second
5102 * config/gcn/gcn.c (gcn_inline_constant64_p): Add 'mixed' parameter.
5103 Implement 'Db' mixed immediate type.
5104 * config/gcn/gcn-valu.md (addcv64si3<exec_vcc>): Rework constraints.
5105 (addcv64si3_dup<exec_vcc>): Delete.
5106 (subcv64si3<exec_vcc>): Rework constraints.
5107 (addv64di3): Rework constraints.
5108 (addv64di3_exec): Rework constraints.
5109 (subv64di3): Rework constraints.
5110 (addv64di3_dup): Delete.
5111 (addv64di3_dup_exec): Delete.
5112 (addv64di3_zext): Rework constraints.
5113 (addv64di3_zext_exec): Rework constraints.
5114 (addv64di3_zext_dup): Rework constraints.
5115 (addv64di3_zext_dup_exec): Rework constraints.
5116 (addv64di3_zext_dup2): Rework constraints.
5117 (addv64di3_zext_dup2_exec): Rework constraints.
5118 (addv64di3_sext_dup2): Rework constraints.
5119 (addv64di3_sext_dup2_exec): Rework constraints.
5121 2020-01-07 Andre Vieira <andre.simoesdiasvieira@arm.com>
5123 * doc/sourcebuild.texi (arm_little_endian, arm_nothumb): Documented
5124 existing target checks.
5126 2020-01-07 Richard Biener <rguenther@suse.de>
5128 * doc/install.texi: Bump minimal supported MPC version.
5130 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
5132 * langhooks-def.h (lhd_simulate_enum_decl): Declare.
5133 (LANG_HOOKS_SIMULATE_ENUM_DECL): Use it.
5134 * langhooks.c: Include stor-layout.h.
5135 (lhd_simulate_enum_decl): New function.
5136 * config/aarch64/aarch64-sve-builtins.cc (init_builtins): Call
5137 handle_arm_sve_h for the LTO frontend.
5138 (register_vector_type): Cope with null returns from pushdecl.
5140 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
5142 * config/aarch64/aarch64-protos.h (aarch64_sve::svbool_type_p)
5143 (aarch64_sve::nvectors_if_data_type): Replace with...
5144 (aarch64_sve::builtin_type_p): ...this.
5145 * config/aarch64/aarch64-sve-builtins.cc: Include attribs.h.
5146 (find_vector_type): Delete.
5147 (add_sve_type_attribute): New function.
5148 (lookup_sve_type_attribute): Likewise.
5149 (register_builtin_types): Add an "SVE type" attribute to each type.
5150 (register_tuple_type): Likewise.
5151 (svbool_type_p, nvectors_if_data_type): Delete.
5152 (mangle_builtin_type): Use lookup_sve_type_attribute.
5153 (builtin_type_p): Likewise. Add an overload that returns the
5154 number of constituent vector and predicate registers.
5155 * config/aarch64/aarch64.c (aarch64_sve_argument_p): Delete.
5156 (aarch64_returns_value_in_sve_regs_p): Use aarch64_sve::builtin_type_p
5157 instead of aarch64_sve_argument_p.
5158 (aarch64_takes_arguments_in_sve_regs_p): Likewise.
5159 (aarch64_pass_by_reference): Likewise.
5160 (aarch64_function_value_1): Likewise.
5161 (aarch64_return_in_memory): Likewise.
5162 (aarch64_layout_arg): Likewise.
5164 2020-01-07 Jakub Jelinek <jakub@redhat.com>
5166 PR tree-optimization/93156
5167 * tree-ssa-ccp.c (bit_value_binop): For x * x note that the second
5168 least significant bit is always clear.
5170 PR tree-optimization/93118
5171 * match.pd ((x >> c) << c -> x & (-1<<c)): Add nop_convert?. Add new
5172 simplifier with two intermediate conversions.
5174 2020-01-07 Martin Liska <mliska@suse.cz>
5176 * params.opt: Add Optimization for various parameters.
5178 2020-01-07 Martin Liska <mliska@suse.cz>
5181 * doc/extend.texi: Explain cloning for target_clone
5184 2020-01-07 Martin Liska <mliska@suse.cz>
5186 PR tree-optimization/92860
5187 * common.opt: Make in Optimization option
5188 as it is affected by -O0, which is an Optimization
5190 * tree-inline.c (tree_inlinable_function_p):
5191 Use opt_for_fn for warn_inline.
5192 (expand_call_inline): Likewise.
5194 2020-01-07 Martin Liska <mliska@suse.cz>
5196 PR tree-optimization/92860
5197 * common.opt: Make flag_ree as optimization
5200 2020-01-07 Martin Liska <mliska@suse.cz>
5202 PR optimization/92860
5203 * params.opt: Mark param_min_crossjump_insns with Optimization
5206 2020-01-07 Luo Xiong Hu <luoxhu@linux.ibm.com>
5208 * ipa-inline-analysis.c (estimate_growth): Fix typo.
5209 * ipa-inline.c (caller_growth_limits): Use OR instead of AND.
5211 2020-01-06 Michael Meissner <meissner@linux.ibm.com>
5213 * config/rs6000/rs6000.c (hard_reg_and_mode_to_addr_mask): New
5214 helper function to return the valid addressing formats for a given
5215 hard register and mode.
5216 (rs6000_adjust_vec_address): Call hard_reg_and_mode_to_addr_mask.
5218 * config/rs6000/constraints.md (Q constraint): Update
5220 * doc/md.texi (RS/6000 constraints): Update 'Q' cosntraint
5223 * config/rs6000/vsx.md (vsx_extract_<mode>_var, VSX_D iterator):
5224 Use 'Q' for doing vector extract from memory.
5225 (vsx_extract_v4sf_var): Use 'Q' for doing vector extract from
5227 (vsx_extract_<mode>_var, VSX_EXTRACT_I iterator): Use 'Q' for
5228 doing vector extract from memory.
5229 (vsx_extract_<mode>_<VS_scalar>mode_var): Use 'Q' for doing vector
5230 extract from memory.
5232 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add support
5233 for the offset being 34-bits when -mcpu=future is used.
5235 2020-01-06 John David Anglin <danglin@gcc.gnu.org>
5237 * config/pa/pa.md: Revert change to use ordered_comparison_operator
5238 instead of cmpib_comparison_operator in cmpib patterns.
5239 * config/pa/predicates.md (cmpib_comparison_operator): Revert removal
5240 of cmpib_comparison_operator. Revise comment.
5242 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
5244 * tree-vect-slp.c (vect_build_slp_tree_1): Require all shifts
5245 in an IFN_DIV_POW2 node to be equal.
5247 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
5249 * tree-vect-stmts.c (vect_check_load_store_mask): Rename to...
5250 (vect_check_scalar_mask): ...this.
5251 (vectorizable_store, vectorizable_load): Update call accordingly.
5252 (vectorizable_call): Use vect_check_scalar_mask to check the mask
5253 argument in calls to conditional internal functions.
5255 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
5257 * config/gcn/gcn-valu.md (subv64di3): Use separate alternatives for
5258 '0' matching inputs.
5259 (subv64di3_exec): Likewise.
5261 2020-01-06 Bryan Stenson <bryan@siliconvortex.com>
5263 * config/mips/mips.c (vr4130_align_insns): Fix typo.
5264 * doc/md.texi (movstr): Likewise.
5266 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
5268 * config/gcn/gcn-valu.md (vec_extract<mode><scalar_mode>): Add early
5271 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
5273 * config/aarch64/t-aarch64 ($(srcdir)/config/aarch64/aarch64-tune.md):
5275 (s-aarch64-tune-md): ...this new stamp file. Pipe the new contents
5276 to a temporary file and use move-if-change to update the real
5277 file where necessary.
5279 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
5281 * config/aarch64/aarch64-sve.md (@aarch64_sel_dup<mode>): Use Upl
5282 rather than Upa for CPY /M.
5284 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
5286 * config/gcn/gcn.c (gcn_inline_constant_p): Allow 64 as an inline
5289 2020-01-06 Martin Liska <mliska@suse.cz>
5291 PR tree-optimization/92860
5292 * params.opt: Mark param_max_combine_insns with Optimization
5295 2020-01-05 Jakub Jelinek <jakub@redhat.com>
5298 * config/i386/i386.md (SWIDWI): New mode iterator.
5299 (DWI, dwi): Add TImode variants.
5300 (addv<mode>4): Use SWIDWI iterator instead of SWI. Use
5301 <general_hilo_operand> instead of <general_operand>. Use
5302 CONST_SCALAR_INT_P instead of CONST_INT_P.
5303 (*addv<mode>4_1): Rename to ...
5304 (addv<mode>4_1): ... this.
5305 (QWI): New mode attribute.
5306 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
5307 define_insn_and_split patterns.
5308 (*addv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
5310 (uaddv<mode>4): Use SWIDWI iterator instead of SWI. Use
5311 <general_hilo_operand> instead of <general_operand>.
5312 (*addcarry<mode>_1): New define_insn.
5313 (*add<dwi>3_doubleword_cc_overflow_1): New define_insn_and_split.
5315 2020-01-03 Konstantin Kharlamov <Hi-Angel@yandex.ru>
5317 * gdbinit.in (pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, pdd, pbs, pbm):
5318 Use "call" instead of "set".
5320 2020-01-03 Martin Jambor <mjambor@suse.cz>
5323 * ipa-cp.c (print_all_lattices): Skip functions without info.
5325 2020-01-03 Jakub Jelinek <jakub@redhat.com>
5328 * config/i386/i386-options.c (ix86_simd_clone_adjust): If
5329 TARGET_PREFER_AVX128, use prefer-vector-width=256 for 'c' and 'd'
5330 simd clones. If TARGET_PREFER_AVX256, use prefer-vector-width=512
5331 for 'e' simd clones.
5334 * config/i386/i386.opt (x_prefer_vector_width_type): Remove TargetSave
5336 (mprefer-vector-width=): Add Save.
5337 * config/i386/i386-options.c (ix86_target_string): Add PVW argument, print
5338 -mprefer-vector-width= if non-zero. Fix up -mfpmath= comment.
5339 (ix86_debug_options, ix86_function_specific_print): Adjust
5340 ix86_target_string callers.
5341 (ix86_valid_target_attribute_inner_p): Handle prefer-vector-width=.
5342 (ix86_valid_target_attribute_tree): Likewise.
5343 * config/i386/i386-options.h (ix86_target_string): Add PVW argument.
5344 * config/i386/i386-expand.c (ix86_expand_builtin): Adjust
5345 ix86_target_string caller.
5348 * config/i386/i386.md (abs<mode>2): Use expand_simple_binop instead of
5349 emitting ASHIFTRT, XOR and MINUS by hand. Use gen_int_mode with QImode
5350 instead of gen_int_shift_amount + convert_modes.
5352 PR rtl-optimization/93088
5353 * loop-iv.c (find_single_def_src): Punt after looking through
5354 128 reg copies for regs with single definitions. Move definitions
5357 2020-01-02 Dennis Zhang <dennis.zhang@arm.com>
5359 * config/arm/arm-c.c (arm_cpu_builtins): Define
5360 __ARM_FEATURE_MATMUL_INT8, __ARM_FEATURE_BF16_VECTOR_ARITHMETIC,
5361 __ARM_FEATURE_BF16_SCALAR_ARITHMETIC, and
5362 __ARM_BF16_FORMAT_ALTERNATIVE when enabled.
5363 * config/arm/arm-cpus.in (armv8_6, i8mm, bf16): New features.
5364 * config/arm/arm-tables.opt: Regenerated.
5365 * config/arm/arm.c (arm_option_reconfigure_globals): Initialize
5366 arm_arch_i8mm and arm_arch_bf16 when enabled.
5367 * config/arm/arm.h (TARGET_I8MM): New macro.
5368 (TARGET_BF16_FP, TARGET_BF16_SIMD): Likewise.
5369 * config/arm/t-aprofile: Add matching rules for -march=armv8.6-a.
5370 * config/arm/t-arm-elf (all_v8_archs): Add armv8.6-a.
5371 * config/arm/t-multilib: Add matching rules for -march=armv8.6-a.
5372 (v8_6_a_simd_variants): New.
5373 (v8_*_a_simd_variants): Add i8mm and bf16.
5374 * doc/invoke.texi (armv8.6-a, i8mm, bf16): Document new options.
5376 2020-01-02 Jakub Jelinek <jakub@redhat.com>
5379 * predict.c (compute_function_frequency): Don't call
5380 warn_function_cold on functions that already have cold attribute.
5382 2020-01-01 John David Anglin <danglin@gcc.gnu.org>
5385 * config/pa/pa.c (pa_elf_select_rtx_section): New. Put references to
5386 COMDAT group function labels in .data.rel.ro.local section.
5387 * config/pa/pa32-linux.h (TARGET_ASM_SELECT_RTX_SECTION): Define.
5390 * config/pa/pa.md (scc): Use ordered_comparison_operator instead of
5391 comparison_operator in B and S integer comparisons. Likewise, use
5392 ordered_comparison_operator instead of cmpib_comparison_operator in
5394 * config/pa/predicates.md (cmpib_comparison_operator): Remove.
5396 2020-01-01 Jakub Jelinek <jakub@redhat.com>
5398 Update copyright years.
5400 * gcc.c (process_command): Update copyright notice dates.
5401 * gcov-dump.c (print_version): Ditto.
5402 * gcov.c (print_version): Ditto.
5403 * gcov-tool.c (print_version): Ditto.
5404 * gengtype.c (create_file): Ditto.
5405 * doc/cpp.texi: Bump @copying's copyright year.
5406 * doc/cppinternals.texi: Ditto.
5407 * doc/gcc.texi: Ditto.
5408 * doc/gccint.texi: Ditto.
5409 * doc/gcov.texi: Ditto.
5410 * doc/install.texi: Ditto.
5411 * doc/invoke.texi: Ditto.
5413 2020-01-01 Jan Hubicka <hubicka@ucw.cz>
5415 * ipa.c (walk_polymorphic_call_targets): Fix updating of overall
5418 2020-01-01 Jakub Jelinek <jakub@redhat.com>
5420 PR tree-optimization/93098
5421 * match.pd (popcount): For shift amounts, use integer_onep
5422 or wi::to_widest () == cst instead of tree_to_uhwi () == cst
5423 tests. Make sure that precision is power of two larger than or equal
5424 to 16. Ensure shift is never negative. Use HOST_WIDE_INT_UC macro
5425 instead of ULL suffixed constants. Formatting fixes.
5427 Copyright (C) 2020 Free Software Foundation, Inc.
5429 Copying and distribution of this file, with or without modification,
5430 are permitted in any medium without royalty provided the copyright
5431 notice and this notice are preserved.