1 2020-03-10 Will Schmidt <will_schmidt@vnet.ibm.com>
4 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Add
5 clause to handle P9V_BUILTIN_VEC_LXVL with const arguments.
7 2020-03-10 Roman Zhuykov <zhroma@ispras.ru>
9 * loop-iv.c (find_simple_exit): Make it static.
10 * cfgloop.h: Remove the corresponding prototype.
12 2020-03-10 Roman Zhuykov <zhroma@ispras.ru>
14 * ddg.c (create_ddg): Fix intendation.
15 (set_recurrence_length): Likewise.
16 (create_ddg_all_sccs): Likewise.
18 2020-03-10 Jakub Jelinek <jakub@redhat.com>
21 * config/i386/i386.md (*testqi_ext_3): Call ix86_match_ccmode with
22 CCZmode instead of CCNOmode if operands[2] has DImode and pos + len
25 2020-03-09 Jason Merrill <jason@redhat.com>
27 * gdbinit.in (pgs): Fix typo in documentation.
29 2020-03-09 Vladimir Makarov <vmakarov@redhat.com>
33 2020-02-28 Vladimir Makarov <vmakarov@redhat.com>
35 PR rtl-optimization/93564
36 * ira-color.c (assign_hard_reg): Prefer smaller hard regno when we
37 do not honor reg alloc order.
39 2020-03-09 Andrew Pinski <apinski@marvell.com>
42 * doc/extend.texi (x86 Operand Modifiers): Fix column
45 2020-03-09 Martin Liska <mliska@suse.cz>
48 * config/rs6000/rs6000.c (rs6000_option_override_internal):
49 Remove set of str_align_loops and str_align_jumps as these
50 should be set in previous 2 conditions in the function.
52 2020-03-09 Jakub Jelinek <jakub@redhat.com>
54 PR rtl-optimization/94045
55 * params.opt (-param=max-find-base-term-values=): New option.
56 * alias.c (find_base_term): Add cut-off for number of visited VALUEs
57 in a single toplevel find_base_term call.
59 2020-03-06 Wilco Dijkstra <wdijkstr@arm.com>
62 * config/aarch64/aarch64-builtins.c (TYPES_TERNOPU_LANE): Add define.
63 * config/aarch64/aarch64-simd.md
64 (aarch64_vec_<su>mult_lane<Qlane>): Add new insn for widening lane mul.
65 (aarch64_vec_<su>mlal_lane<Qlane>): Likewise.
66 * config/aarch64/aarch64-simd-builtins.def: Add intrinsics.
67 * config/aarch64/arm_neon.h:
68 (vmlal_lane_s16): Expand using intrinsics rather than inline asm.
69 (vmlal_lane_u16): Likewise.
70 (vmlal_lane_s32): Likewise.
71 (vmlal_lane_u32): Likewise.
72 (vmlal_laneq_s16): Likewise.
73 (vmlal_laneq_u16): Likewise.
74 (vmlal_laneq_s32): Likewise.
75 (vmlal_laneq_u32): Likewise.
76 (vmull_lane_s16): Likewise.
77 (vmull_lane_u16): Likewise.
78 (vmull_lane_s32): Likewise.
79 (vmull_lane_u32): Likewise.
80 (vmull_laneq_s16): Likewise.
81 (vmull_laneq_u16): Likewise.
82 (vmull_laneq_s32): Likewise.
83 (vmull_laneq_u32): Likewise.
84 * config/aarch64/iterators.md (Vcondtype): New iterator for lane mul.
87 2020-03-06 Wilco Dijkstra <wdijkstr@arm.com>
89 * aarch64/aarch64-simd.md (aarch64_mla_elt<mode>): Correct lane syntax.
90 (aarch64_mla_elt_<vswap_width_name><mode>): Likewise.
91 (aarch64_mls_elt<mode>): Likewise.
92 (aarch64_mls_elt_<vswap_width_name><mode>): Likewise.
93 (aarch64_fma4_elt<mode>): Likewise.
94 (aarch64_fma4_elt_<vswap_width_name><mode>): Likewise.
95 (aarch64_fma4_elt_to_64v2df): Likewise.
96 (aarch64_fnma4_elt<mode>): Likewise.
97 (aarch64_fnma4_elt_<vswap_width_name><mode>): Likewise.
98 (aarch64_fnma4_elt_to_64v2df): Likewise.
100 2020-03-06 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
102 * config/aarch64/aarch64-sve2.md (@aarch64_sve_<sve_int_op><mode>:
103 Specify movprfx attribute.
104 (@aarch64_sve_<sve_int_op>_lane_<mode>): Likewise.
106 2020-03-06 David Edelsohn <dje.gcc@gmail.com>
109 * config/rs6000/aix61.h (TARGET_NO_SUM_IN_TOC): Set to 1 for
111 (TARGET_NO_FP_IN_TOC): Same.
112 * config/rs6000/aix71.h: Same.
113 * config/rs6000/aix72.h: Same.
115 2020-03-06 Andrew Pinski <apinski@marvell.com>
116 Jeff Law <law@redhat.com>
118 PR rtl-optimization/93996
119 * haifa-sched.c (remove_notes): Be more careful when adding
122 2020-03-06 Delia Burduv <delia.burduv@arm.com>
124 * config/arm/arm_neon.h (vld2_bf16): New.
130 (vld2_dup_bf16): New.
131 (vld2q_dup_bf16): New.
132 (vld3_dup_bf16): New.
133 (vld3q_dup_bf16): New.
134 (vld4_dup_bf16): New.
135 (vld4q_dup_bf16): New.
136 * config/arm/arm_neon_builtins.def
137 (vld2): Changed to VAR13 and added v4bf, v8bf
138 (vld2_dup): Changed to VAR8 and added v4bf, v8bf
139 (vld3): Changed to VAR13 and added v4bf, v8bf
140 (vld3_dup): Changed to VAR8 and added v4bf, v8bf
141 (vld4): Changed to VAR13 and added v4bf, v8bf
142 (vld4_dup): Changed to VAR8 and added v4bf, v8bf
143 * config/arm/iterators.md (VDXBF2): New iterator.
144 *config/arm/neon.md (neon_vld2): Use new iterators.
145 (neon_vld2_dup<mode): Use new iterators.
146 (neon_vld3<mode>): Likewise.
147 (neon_vld3qa<mode>): Likewise.
148 (neon_vld3qb<mode>): Likewise.
149 (neon_vld3_dup<mode>): Likewise.
150 (neon_vld4<mode>): Likewise.
151 (neon_vld4qa<mode>): Likewise.
152 (neon_vld4qb<mode>): Likewise.
153 (neon_vld4_dup<mode>): Likewise.
154 (neon_vld2_dupv8bf): New.
155 (neon_vld3_dupv8bf): Likewise.
156 (neon_vld4_dupv8bf): Likewise.
158 2020-03-06 Delia Burduv <delia.burduv@arm.com>
160 * config/arm/arm_neon.h (bfloat16x4x2_t): New typedef.
161 (bfloat16x8x2_t): New typedef.
162 (bfloat16x4x3_t): New typedef.
163 (bfloat16x8x3_t): New typedef.
164 (bfloat16x4x4_t): New typedef.
165 (bfloat16x8x4_t): New typedef.
172 * config/arm/arm-builtins.c (v2bf_UP): Define.
174 (arm_init_simd_builtin_types): Init Bfloat16x2_t eltype.
175 * config/arm/arm-modes.def (V2BF): New mode.
176 * config/arm/arm-simd-builtin-types.def
177 (Bfloat16x2_t): New entry.
178 * config/arm/arm_neon_builtins.def
179 (vst2): Changed to VAR13 and added v4bf, v8bf
180 (vst3): Changed to VAR13 and added v4bf, v8bf
181 (vst4): Changed to VAR13 and added v4bf, v8bf
182 * config/arm/iterators.md (VDXBF): New iterator.
183 (VQ2BF): New iterator.
184 *config/arm/neon.md (neon_vst2<mode>): Used new iterators.
185 (neon_vst2<mode>): Used new iterators.
186 (neon_vst3<mode>): Used new iterators.
187 (neon_vst3<mode>): Used new iterators.
188 (neon_vst3qa<mode>): Used new iterators.
189 (neon_vst3qb<mode>): Used new iterators.
190 (neon_vst4<mode>): Used new iterators.
191 (neon_vst4<mode>): Used new iterators.
192 (neon_vst4qa<mode>): Used new iterators.
193 (neon_vst4qb<mode>): Used new iterators.
195 2020-03-06 Delia Burduv <delia.burduv@arm.com>
197 * config/aarch64/aarch64-simd-builtins.def
198 (bfcvtn): New built-in function.
199 (bfcvtn_q): New built-in function.
200 (bfcvtn2): New built-in function.
201 (bfcvt): New built-in function.
202 * config/aarch64/aarch64-simd.md
203 (aarch64_bfcvtn<q><mode>): New pattern.
204 (aarch64_bfcvtn2v8bf): New pattern.
205 (aarch64_bfcvtbf): New pattern.
206 * config/aarch64/arm_bf16.h (float32_t): New typedef.
207 (vcvth_bf16_f32): New intrinsic.
208 * config/aarch64/arm_bf16.h (vcvt_bf16_f32): New intrinsic.
209 (vcvtq_low_bf16_f32): New intrinsic.
210 (vcvtq_high_bf16_f32): New intrinsic.
211 * config/aarch64/iterators.md (V4SF_TO_BF): New mode iterator.
212 (UNSPEC_BFCVTN): New UNSPEC.
213 (UNSPEC_BFCVTN2): New UNSPEC.
214 (UNSPEC_BFCVT): New UNSPEC.
215 * config/arm/types.md (bf_cvt): New type.
217 2020-03-06 Andreas Krebbel <krebbel@linux.ibm.com>
219 * config/s390/s390.md ("tabort"): Get rid of two consecutive
220 blanks in format string.
222 2020-03-05 H.J. Lu <hongjiu.lu@intel.com>
226 * config/i386/i386-protos.h (ix86_output_ssemov): New prototype.
227 * config/i386/i386.c (ix86_get_ssemov): New function.
228 (ix86_output_ssemov): Likewise.
229 * config/i386/sse.md (VMOVE:mov<mode>_internal): Call
230 ix86_output_ssemov for TYPE_SSEMOV. Remove TARGET_AVX512VL
232 (*movxi_internal_avx512f): Call ix86_output_ssemov for TYPE_SSEMOV.
233 (*movoi_internal_avx): Call ix86_output_ssemov for TYPE_SSEMOV.
234 Remove ext_sse_reg_operand and TARGET_AVX512VL check.
235 (*movti_internal): Likewise.
236 (*movtf_internal): Call ix86_output_ssemov for TYPE_SSEMOV.
238 2020-03-05 Jeff Law <law@redhat.com>
240 PR tree-optimization/91890
241 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Remove LOC argument.
242 Use gimple_or_expr_nonartificial_location.
243 (check_bounds_overlap): Drop LOC argument to maybe_diag_access_bounds.
244 Use gimple_or_expr_nonartificial_location.
245 * gimple.c (gimple_or_expr_nonartificial_location): New function.
246 * gimple.h (gimple_or_expr_nonartificial_location): Declare it.
247 * tree-ssa-strlen.c (maybe_warn_overflow): Use
248 gimple_or_expr_nonartificial_location.
249 (maybe_diag_stxncpy_trunc, handle_builtin_stxncpy_strncat): Likewise.
250 (maybe_warn_pointless_strcmp): Likewise.
252 2020-03-05 Jakub Jelinek <jakub@redhat.com>
255 * config/i386/avx2intrin.h (_mm_mask_i32gather_ps): Fix first cast of
256 SRC and MASK arguments to __m128 from __m128d.
257 (_mm256_mask_i32gather_ps): Fix first cast of MASK argument to __m256
259 (_mm_mask_i64gather_ps): Fix first cast of MASK argument to __m128
261 * config/i386/xopintrin.h (_mm_permute2_pd): Fix first cast of C
262 argument to __m128i from __m128d.
263 (_mm256_permute2_pd): Fix first cast of C argument to __m256i from
265 (_mm_permute2_ps): Fix first cast of C argument to __m128i from __m128.
266 (_mm256_permute2_ps): Fix first cast of C argument to __m256i from
269 2020-03-05 Delia Burduv <delia.burduv@arm.com>
271 * config/arm/arm_neon.h (vbfmmlaq_f32): New.
272 (vbfmlalbq_f32): New.
273 (vbfmlaltq_f32): New.
274 (vbfmlalbq_lane_f32): New.
275 (vbfmlaltq_lane_f32): New.
276 (vbfmlalbq_laneq_f32): New.
277 (vbfmlaltq_laneq_f32): New.
278 * config/arm/arm_neon_builtins.def (vmmla): New.
285 * config/arm/iterators.md (BF_MA): New int iterator.
286 (bt): New int attribute.
287 (VQXBF): Copy of VQX with V8BF.
288 * config/arm/neon.md (neon_vmmlav8bf): New insn.
289 (neon_vfma<bt>v8bf): New insn.
290 (neon_vfma<bt>_lanev8bf): New insn.
291 (neon_vfma<bt>_laneqv8bf): New expand.
292 (neon_vget_high<mode>): Changed iterator to VQXBF.
293 * config/arm/unspecs.md (UNSPEC_BFMMLA): New UNSPEC.
294 (UNSPEC_BFMAB): New UNSPEC.
295 (UNSPEC_BFMAT): New UNSPEC.
297 2020-03-05 Jakub Jelinek <jakub@redhat.com>
300 * tree-pretty-print.h (pretty_print_string): Declare.
301 * tree-pretty-print.c (pretty_print_string): Remove forward
302 declaration, no longer static. Change nbytes parameter type
303 from unsigned to size_t.
304 * print-rtl.c (print_value) <case CONST_STRING>: Use
305 pretty_print_string and for shrink way too long strings.
307 2020-03-05 Richard Biener <rguenther@suse.de>
308 Jakub Jelinek <jakub@redhat.com>
310 PR tree-optimization/93582
311 * tree-ssa-sccvn.c (vn_reference_lookup_3): Treat POINTER_PLUS_EXPR
312 last operand as signed when looking for memset offset. Formatting
315 2020-03-04 Andrew Pinski <apinski@marvell.com>
318 * value-prof.c (dump_histogram_value): Use std::abs.
320 2020-03-04 Martin Sebor <msebor@redhat.com>
322 PR tree-optimization/93986
323 * tree-ssa-strlen.c (maybe_warn_overflow): Convert all wide_int
324 operands to the same precision widest_int to avoid ICEs.
326 2020-03-04 Bill Schmidt <wschmidt@linux.ibm.com>
329 * rs6000-cpus.def (OTHER_ALTIVEC_MASKS): New #define.
330 * rs6000.c (rs6000_disable_incompatible_switches): Add table entry
331 for OPTION_MASK_ALTIVEC.
333 2020-03-04 Andreas Krebbel <krebbel@linux.ibm.com>
335 * config.gcc: Include the glibc-stdint.h header for zTPF.
337 2020-03-04 Andreas Krebbel <krebbel@linux.ibm.com>
339 * config/s390/s390.c (s390_secondary_memory_needed): Disallow
340 direct FPR-GPR copies.
341 (s390_register_info_gprtofpr): Disallow GPR content to be saved in
344 2020-03-04 Andreas Krebbel <krebbel@linux.ibm.com>
346 * config/s390/s390.c (s390_emit_prologue): Specify the 2 new
347 operands to the prologue_tpf expander.
348 (s390_emit_epilogue): Likewise.
349 (s390_option_override_internal): Do error checking and setup for
351 * config/s390/tpf.h (TPF_TRACE_PROLOGUE_CHECK)
352 (TPF_TRACE_EPILOGUE_CHECK, TPF_TRACE_PROLOGUE_TARGET)
353 (TPF_TRACE_EPILOGUE_TARGET, TPF_TRACE_PROLOGUE_SKIP_TARGET)
354 (TPF_TRACE_EPILOGUE_SKIP_TARGET): New macro definitions.
355 * config/s390/tpf.md ("prologue_tpf", "epilogue_tpf"): Add two new
356 operands for the check flag and the branch target.
357 * config/s390/tpf.opt ("mtpf-trace-hook-prologue-check")
358 ("mtpf-trace-hook-prologue-target")
359 ("mtpf-trace-hook-epilogue-check")
360 ("mtpf-trace-hook-epilogue-target", "mtpf-trace-skip"): New
362 * doc/invoke.texi: Document -mtpf-trace-skip option. The other
363 options are for debugging purposes and will not be documented
366 2020-03-04 Jakub Jelinek <jakub@redhat.com>
369 * tree-inline.c (copy_decl_to_var): Copy DECL_BY_REFERENCE flag.
371 * tree-ssa-sccvn.c (vn_walk_cb_data::push_partial_def): Add offseti
372 argument. Change pd argument so that it can be modified. Turn
373 constant non-CONSTRUCTOR store into non-constant if it is too large.
374 Adjust offset and size of CONSTRUCTOR or non-constant store to avoid
376 (vn_walk_cb_data::vn_walk_cb_data, vn_reference_lookup_3): Adjust
379 2020-02-04 Richard Biener <rguenther@suse.de>
381 PR tree-optimization/93964
382 * graphite-isl-ast-to-gimple.c
383 (gcc_expression_from_isl_ast_expr_id): Add intermediate
384 conversion for pointer to integer converts.
385 * graphite-scop-detection.c (assign_parameter_index_in_region):
388 2020-03-04 Martin Liska <mliska@suse.cz>
392 * doc/invoke.texi: Clarify --help=language and --help=common
395 2020-03-04 Jakub Jelinek <jakub@redhat.com>
397 PR tree-optimization/94001
398 * tree-tailcall.c (process_assignment): Before comparing op1 to
399 *ass_var, verify *ass_var is non-NULL.
401 2020-03-04 Kito Cheng <kito.cheng@sifive.com>
404 * config/riscv/riscv.c (riscv_emit_float_compare): Using NE to compare
407 2020-03-03 Dennis Zhang <dennis.zhang@arm.com>
409 * config/arm/arm_bf16.h (vcvtah_f32_bf16, vcvth_bf16_f32): New.
410 * config/arm/arm_neon.h (vcvt_f32_bf16, vcvtq_low_f32_bf16): New.
411 (vcvtq_high_f32_bf16, vcvt_bf16_f32): New.
412 (vcvtq_low_bf16_f32, vcvtq_high_bf16_f32): New.
413 * config/arm/arm_neon_builtins.def (vbfcvt, vbfcvt_high): New entries.
414 (vbfcvtv4sf, vbfcvtv4sf_high): Likewise.
415 * config/arm/iterators.md (VBFCVT, VBFCVTM): New mode iterators.
416 (V_bf_low, V_bf_cvt_m): New mode attributes.
417 * config/arm/neon.md (neon_vbfcvtv4sf<VBFCVT:mode>): New.
418 (neon_vbfcvtv4sf_highv8bf, neon_vbfcvtsf): New.
419 (neon_vbfcvt<VBFCVT:mode>, neon_vbfcvt_highv8bf): New.
420 (neon_vbfcvtbf_cvtmode<mode>, neon_vbfcvtbf): New
421 * config/arm/unspecs.md (UNSPEC_BFCVT, UNSPEC_BFCVT_HIG): New.
423 2020-03-03 Jakub Jelinek <jakub@redhat.com>
425 PR tree-optimization/93582
426 * tree-ssa-sccvn.h (vn_reference_lookup): Add mask argument.
427 * tree-ssa-sccvn.c (struct vn_walk_cb_data): Add mask and masked_result
428 members, initialize them in the constructor and if mask is non-NULL,
429 artificially push_partial_def {} for the portions of the mask that
431 (vn_walk_cb_data::finish): If mask is non-NULL, set masked_result to
432 val and return (void *)-1. Formatting fix.
433 (vn_reference_lookup_pieces): Adjust vn_walk_cb_data initialization.
435 (vn_reference_lookup): Add mask argument. If non-NULL, don't call
436 fully_constant_vn_reference_p nor vn_reference_lookup_1 and return
438 (visit_nary_op): Handle BIT_AND_EXPR of a memory load and INTEGER_CST
440 (visit_stmt): Formatting fix.
442 2020-03-03 Richard Biener <rguenther@suse.de>
444 PR tree-optimization/93946
445 * alias.h (refs_same_for_tbaa_p): Declare.
446 * alias.c (refs_same_for_tbaa_p): New function.
447 * tree-ssa-alias.c (ao_ref_alias_set): For a NULL ref return
449 * tree-ssa-scopedtables.h
450 (avail_exprs_stack::lookup_avail_expr): Add output argument
451 giving access to the hashtable entry.
452 * tree-ssa-scopedtables.c (avail_exprs_stack::lookup_avail_expr):
454 * tree-ssa-dom.c: Include alias.h.
455 (dom_opt_dom_walker::optimize_stmt): Validate TBAA state before
456 removing redundant store.
457 * tree-ssa-sccvn.h (vn_reference_s::base_set): New member.
458 (ao_ref_init_from_vn_reference): Adjust prototype.
459 (vn_reference_lookup_pieces): Likewise.
460 (vn_reference_insert_pieces): Likewise.
461 * tree-ssa-sccvn.c: Track base alias set in addition to alias
463 (eliminate_dom_walker::eliminate_stmt): Also check base alias
464 set when removing redundant stores.
465 (visit_reference_op_store): Likewise.
466 * dse.c (record_store): Adjust valdity check for redundant
469 2020-03-03 Jakub Jelinek <jakub@redhat.com>
472 * config/s390/s390.h (OPTION_DEFAULT_SPECS): Reorder.
474 PR rtl-optimization/94002
475 * explow.c (plus_constant): Punt if cst has VOIDmode and
476 get_pool_mode is different from mode.
478 2020-03-03 Claudiu Zissulescu <claziss@synopsys.com>
480 * config/arc/arc.c (leigitimate_small_data_address_p): Check if an
481 address has an offset which fits the scalling constraint for a
482 load/store operation.
483 (legitimate_scaled_address_p): Update use
484 leigitimate_small_data_address_p.
485 (arc_print_operand): Likewise.
486 (arc_legitimate_address_p): Likewise.
487 (legitimate_small_data_address_p): Likewise.
489 2020-03-03 Claudiu Zissulescu <claziss@synopsys.com>
491 * config/arc/arc.md (fmasf4_fpu): Use accl_operand predicate.
492 (fnmasf4_fpu): Likewise.
494 2020-03-03 Claudiu Zissulescu <claziss@synopsys.com>
496 * config/arc/arc.md (adddi3): Early expand the 64bit operation into
499 (adddi3_i): Remove pattern.
500 (subdi3_i): Likewise.
502 2020-03-03 Claudiu Zissulescu <claziss@synopsys.com>
504 * config/arc/arc.md (eh_return): Add length info.
506 2020-03-02 David Malcolm <dmalcolm@redhat.com>
508 * doc/invoke.texi (-fanalyzer-show-duplicate-count): New.
510 2020-03-02 David Malcolm <dmalcolm@redhat.com>
512 * doc/invoke.texi (Static Analyzer Options): Add
513 -Wanalyzer-stale-setjmp-buffer to the list of options enabled
516 2020-03-02 Uroš Bizjak <ubizjak@gmail.com>
519 * config/i386/i386.md (movstrict<mode>): Allow only
520 registers with VALID_INT_MODE_P modes.
522 2020-03-02 Andrew Stubbs <ams@codesourcery.com>
524 * config/gcn/gcn-valu.md (dpp_move<mode>): New.
525 (reduc_insn): Use 'U' and 'B' operand codes.
526 (reduc_<reduc_op>_scal_<mode>): Allow all types.
527 (reduc_<reduc_op>_scal_v64di): Delete.
528 (*<reduc_op>_dpp_shr_<mode>): Allow all 1reg types.
529 (*plus_carry_dpp_shr_v64si): Change to ...
530 (*plus_carry_dpp_shr_<mode>): ... this and allow all 1reg int types.
531 (mov_from_lane63_v64di): Change to ...
532 (mov_from_lane63_<mode>): ... this, and allow all 64-bit modes.
533 * config/gcn/gcn.c (gcn_expand_dpp_shr_insn): Increase buffer size.
534 Support UNSPEC_MOV_DPP_SHR output formats.
535 (gcn_expand_reduc_scalar): Add "use_moves" reductions.
536 Add "use_extends" reductions.
537 (print_operand_address): Add 'I' and 'U' codes.
538 * config/gcn/gcn.md (unspec): Add UNSPEC_MOV_DPP_SHR.
540 2020-03-02 Martin Liska <mliska@suse.cz>
542 * lto-wrapper.c: Fix typo in comment about
543 C++ standard version.
545 2020-03-01 Martin Sebor <msebor@redhat.com>
548 * calls.c (init_attr_rdwr_indices): Correctly handle attribute.
550 2020-03-01 Martin Sebor <msebor@redhat.com>
553 * tree-ssa-strlen.c (count_nonzero_bytes): Set the size to that
554 of a pointer in the outermost ADDR_EXPRs.
556 2020-02-28 Jeff Law <law@redhat.com>
558 * config/v850/v850.h (STATIC_CHAIN_REGNUM): Change to r19.
559 * config/v850/v850.c (v850_asm_trampoline_template): Update
562 2020-02-28 Michael Meissner <meissner@linux.ibm.com>
565 * config/rs6000/vsx.md (vsx_extract_<mode>_<VS_scalar>mode_var):
568 2020-02-28 Martin Liska <mliska@suse.cz>
571 * configure.ac: Improve detection of ld_date by requiring
572 either two dashes or none.
573 * configure: Regenerate.
575 2020-02-28 Vladimir Makarov <vmakarov@redhat.com>
577 PR rtl-optimization/93564
578 * ira-color.c (assign_hard_reg): Prefer smaller hard regno when we
579 do not honor reg alloc order.
581 2020-02-27 Joel Hutton <Joel.Hutton@arm.com>
584 * config/aarch64/aarch64.c (aarch64_override_options): Fix
585 misleading warning string.
587 2020-02-27 Martin Sebor <msebor@redhat.com>
589 * doc/invoke.texi (-Wbuiltin-declaration-mismatch): Fix a typo.
591 2020-02-27 Michael Meissner <meissner@linux.ibm.com>
594 * config/rs6000/vsx.md (vsx_extract_<mode>_var, VSX_D iterator):
595 Split the insn into two parts. This insn only does variable
596 extract from a register.
597 (vsx_extract_<mode>_var_load, VSX_D iterator): New insn, do
598 variable extract from memory.
599 (vsx_extract_v4sf_var): Split the insn into two parts. This insn
600 only does variable extract from a register.
601 (vsx_extract_v4sf_var_load): New insn, do variable extract from
603 (vsx_extract_<mode>_var, VSX_EXTRACT_I iterator): Split the insn
604 into two parts. This insn only does variable extract from a
606 (vsx_extract_<mode>_var_load, VSX_EXTRACT_I iterator): New insn,
607 do variable extract from memory.
609 2020-02-27 Martin Jambor <mjambor@suse.cz>
610 Feng Xue <fxue@os.amperecomputing.com>
613 * ipa-cp.c (same_node_or_its_all_contexts_clone_p): Replaced with
614 new function calls_same_node_or_its_all_contexts_clone_p.
615 (cgraph_edge_brings_value_p): Use it.
616 (cgraph_edge_brings_value_p): Likewise.
617 (self_recursive_pass_through_p): Return false if caller is a clone.
618 (self_recursive_agg_pass_through_p): Likewise.
620 2020-02-27 Jan Hubicka <hubicka@ucw.cz>
623 * alias.c (ends_tbaa_access_path_p): Break out from ...
624 (component_uses_parent_alias_set_from): ... here.
625 * alias.h (ends_tbaa_access_path_p): Declare.
626 * tree-ssa-alias.c (access_path_may_continue_p): Break out from ...;
627 handle trailing arrays past end of tbaa access path.
628 (aliasing_component_refs_p): ... here; likewise.
629 (nonoverlapping_refs_since_match_p): Track TBAA segment of the access
630 path; disambiguate also past end of it.
631 (nonoverlapping_component_refs_p): Use only TBAA segment of the access
634 2020-02-27 Mihail Ionescu <mihail.ionescu@arm.com>
636 * (__ARM_NUM_LANES, __arm_lane, __arm_lane_q): Move to the
637 beginning of the file.
638 (vcreate_bf16, vcombine_bf16): New.
639 (vdup_n_bf16, vdupq_n_bf16): New.
640 (vdup_lane_bf16, vdup_laneq_bf16): New.
641 (vdupq_lane_bf16, vdupq_laneq_bf16): New.
642 (vduph_lane_bf16, vduph_laneq_bf16): New.
643 (vset_lane_bf16, vsetq_lane_bf16): New.
644 (vget_lane_bf16, vgetq_lane_bf16): New.
645 (vget_high_bf16, vget_low_bf16): New.
646 (vreinterpret_bf16_u8, vreinterpretq_bf16_u8): New.
647 (vreinterpret_bf16_u16, vreinterpretq_bf16_u16): New.
648 (vreinterpret_bf16_u32, vreinterpretq_bf16_u32): New.
649 (vreinterpret_bf16_u64, vreinterpretq_bf16_u64): New.
650 (vreinterpret_bf16_s8, vreinterpretq_bf16_s8): New.
651 (vreinterpret_bf16_s16, vreinterpretq_bf16_s16): New.
652 (vreinterpret_bf16_s32, vreinterpretq_bf16_s32): New.
653 (vreinterpret_bf16_s64, vreinterpretq_bf16_s64): New.
654 (vreinterpret_bf16_p8, vreinterpretq_bf16_p8): New.
655 (vreinterpret_bf16_p16, vreinterpretq_bf16_p16): New.
656 (vreinterpret_bf16_p64, vreinterpretq_bf16_p64): New.
657 (vreinterpret_bf16_f32, vreinterpretq_bf16_f32): New.
658 (vreinterpret_bf16_f64, vreinterpretq_bf16_f64): New.
659 (vreinterpretq_bf16_p128): New.
660 (vreinterpret_s8_bf16, vreinterpretq_s8_bf16): New.
661 (vreinterpret_s16_bf16, vreinterpretq_s16_bf16): New.
662 (vreinterpret_s32_bf16, vreinterpretq_s32_bf16): New.
663 (vreinterpret_s64_bf16, vreinterpretq_s64_bf16): New.
664 (vreinterpret_u8_bf16, vreinterpretq_u8_bf16): New.
665 (vreinterpret_u16_bf16, vreinterpretq_u16_bf16): New.
666 (vreinterpret_u32_bf16, vreinterpretq_u32_bf16): New.
667 (vreinterpret_u64_bf16, vreinterpretq_u64_bf16): New.
668 (vreinterpret_p8_bf16, vreinterpretq_p8_bf16): New.
669 (vreinterpret_p16_bf16, vreinterpretq_p16_bf16): New.
670 (vreinterpret_p64_bf16, vreinterpretq_p64_bf16): New.
671 (vreinterpret_f32_bf16, vreinterpretq_f32_bf16): New.
672 (vreinterpretq_p128_bf16): New.
673 * config/arm/arm_neon_builtins.def (VDX): Add V4BF.
675 (V_elem_l): Likewise.
678 (V_DOUBLE): Likewise.
679 (VDQX): Add V4BF and V8BF.
680 (V_two_elem, V_three_elem, V_four_elem): Likewise.
683 (V_double_vector_mode): Likewise.
684 (V_cmp_result): Likewise.
685 (V_uf_sclr): Likewise.
686 (V_sz_elem): Likewise.
687 (Is_d_reg): Likewise.
688 (V_mode_nunits): Likewise.
689 * config/arm/neon.md (neon_vdup_lane): Enable for BFloat16.
691 2020-02-27 Andrew Stubbs <ams@codesourcery.com>
693 * config/gcn/gcn-valu.md (VEC_SUBDWORD_MODE): New mode iterator.
694 (<expander><mode>2<exec>): Change modes to VEC_ALL1REG_INT_MODE.
695 (<expander><mode>3<exec>): Likewise.
696 (<expander><mode>3): New.
697 (v<expander><mode>3): New.
698 (<expander><mode>3): New.
699 (<expander><mode>3<exec>): Rename to ...
700 (<expander>v64si3<exec>): ... this, and change modes to V64SI.
701 * config/gcn/gcn.md (mnemonic): Use '%B' for not.
703 2020-02-27 Alexandre Oliva <oliva@adacore.com>
705 * config/vx-common.h (NO_DOLLAR_IN_LABEL, NO_DOT_IN_LABEL): Leave
708 2020-02-27 Richard Biener <rguenther@suse.de>
710 PR tree-optimization/93508
711 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle _CHK like
712 non-_CHK variants. Valueize their length arguments.
714 2020-02-27 Richard Biener <rguenther@suse.de>
716 PR tree-optimization/93953
717 * tree-vect-slp.c (slp_copy_subtree): Avoid keeping a reference
718 to the hash-map entry.
720 2020-02-27 Andrew Stubbs <ams@codesourcery.com>
722 * config/gcn/gcn.md (mov<mode>): Add transformations for BI subregs.
724 2020-02-27 Mark Williams <mwilliams@fb.com>
726 * dwarf2out.c (file_name_acquire): Call remap_debug_filename.
727 * lto-opts.c (lto_write_options): Drop -fdebug-prefix-map,
728 -ffile-prefix-map and -fmacro-prefix-map.
729 * lto-streamer-out.c: Include file-prefix-map.h.
730 (lto_output_location): Remap the file part of locations.
732 2020-02-27 Jakub Jelinek <jakub@redhat.com>
735 * gimplify.c (gimplify_init_constructor): Don't promote readonly
736 DECL_REGISTER variables to TREE_STATIC.
738 PR tree-optimization/93582
739 PR tree-optimization/93945
740 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle memset with
741 non-zero INTEGER_CST second argument and ref->offset or ref->size
742 not a multiple of BITS_PER_UNIT.
744 2020-02-27 Jonathan Wakely <jwakely@redhat.com>
746 * doc/install.texi (Binaries): Update description of BullFreeware.
748 2020-02-26 Sandra Loosemore <sandra@codesourcery.com>
752 * doc/invoke.texi (Option Summary): Re-alphabetize warnings in
753 C++ Language Options, Warning Options, and Static Analyzer
754 Options lists. Document negative form of options enabled by
755 default. Move some things around to more accurately sort
756 warnings by category.
757 (C++ Dialect Options, Warning Options, Static Analyzer
758 Options): Document negative form of options when enabled by
759 default. Move some things around to more accurately sort
760 warnings by category. Add some missing index entries.
763 2020-02-26 Carl Love <cel@us.ibm.com>
766 * doc/extend.texi (PowerPC AltiVec Built-in Functions available on
767 ISA 2.07): The builtin-function name __builtin_crypto_vpmsumb is only
768 for the vector unsigned short arguments. It is also listed as the
769 name of the built-in for arguments vector unsigned short,
770 vector unsigned int and vector unsigned long long built-ins. The
771 name of the builtins for these arguments should be:
772 __builtin_crypto_vpmsumh, __builtin_crypto_vpmsumw and
773 __builtin_crypto_vpmsumd respectively.
775 2020-02-26 Richard Biener <rguenther@suse.de>
777 * tree-vect-slp.c (vect_print_slp_tree): Also dump ref count
778 and load permutation.
780 2020-02-26 Richard Sandiford <richard.sandiford@arm.com>
783 * optabs-tree.c (supportable_convert_operation): Reject types with
786 2020-02-26 David Malcolm <dmalcolm@redhat.com>
788 * Makefile.in (ANALYZER_OBJS): Add analyzer/bar-chart.o.
790 2020-02-26 Jakub Jelinek <jakub@redhat.com>
792 PR tree-optimization/93820
793 * gimple-ssa-store-merging.c (check_no_overlap): Change RHS_CODE
794 argument to ALL_INTEGER_CST_P boolean.
795 (imm_store_chain_info::try_coalesce_bswap): Adjust caller.
796 (imm_store_chain_info::coalesce_immediate_stores): Likewise. Handle
797 adjacent INTEGER_CST store into merged_store->only_constants like
800 2020-02-25 Jakub Jelinek <jakub@redhat.com>
803 * config/sh/sh.c (expand_cbranchdi4): Fix comment typo, probablity
805 * cfghooks.c (verify_flow_info): Likewise.
806 * predict.c (combine_predictions_for_bb): Likewise.
807 * bb-reorder.c (connect_better_edge_p): Likewise. Fix comment typo,
808 sucessor -> successor.
809 (find_traces_1_round): Fix comment typo, destinarion -> destination.
810 * omp-expand.c (expand_oacc_for): Fix comment typo, sucessors ->
812 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): Fix dump
813 message typo, sucessors -> successors.
815 2020-02-25 Martin Sebor <msebor@redhat.com>
817 * doc/extend.texi (attribute access): Correct an example.
819 2020-02-25 Mihail Ionescu <mihail.ionescu@arm.com>
821 * config/aarch64/aarch64-builtins.c (aarch64_scalar_builtin_types):
823 (aarch64_init_simd_builtin_scalar_types): Register simd_bf.
825 * config/aarch64/iterators.md (VALLDIF): Enable for V4BF and V8BF.
826 (VD): Enable for V4BF.
828 (VQ): Enable for V8BF.
831 (VDBL, Vdbl): Add V4BF.
832 (V_INT_EQUIV, v_int_equiv): Add V4BF and V8BF.
833 * config/aarch64/arm_neon.h (bfloat16x4x2_t): New typedef.
834 (bfloat16x8x2_t): Likewise.
835 (bfloat16x4x3_t): Likewise.
836 (bfloat16x8x3_t): Likewise.
837 (bfloat16x4x4_t): Likewise.
838 (bfloat16x8x4_t): Likewise.
839 (vcombine_bf16): New.
840 (vld1_bf16, vld1_bf16_x2): New.
841 (vld1_bf16_x3, vld1_bf16_x4): New.
842 (vld1q_bf16, vld1q_bf16_x2): New.
843 (vld1q_bf16_x3, vld1q_bf16_x4): New.
844 (vld1_lane_bf16): New.
845 (vld1q_lane_bf16): New.
846 (vld1_dup_bf16): New.
847 (vld1q_dup_bf16): New.
850 (vld2_dup_bf16): New.
851 (vld2q_dup_bf16): New.
854 (vld3_dup_bf16): New.
855 (vld3q_dup_bf16): New.
858 (vld4_dup_bf16): New.
859 (vld4q_dup_bf16): New.
860 (vst1_bf16, vst1_bf16_x2): New.
861 (vst1_bf16_x3, vst1_bf16_x4): New.
862 (vst1q_bf16, vst1q_bf16_x2): New.
863 (vst1q_bf16_x3, vst1q_bf16_x4): New.
864 (vst1_lane_bf16): New.
865 (vst1q_lane_bf16): New.
873 2020-02-25 Mihail Ionescu <mihail.ionescu@arm.com>
875 * config/aarch64/iterators.md (VDQF_F16) Add V4BF and V8BF.
876 (VALL_F16): Likewise.
877 (VALLDI_F16): Likewise.
880 (vswap_width_name): Likewise.
881 (VSWAP_WIDTH): Likewise.
885 * config/aarch64/arm_neon.h (vset_lane_bf16, vsetq_lane_bf16): New.
886 (vget_lane_bf16, vgetq_lane_bf16): New.
888 (vdup_n_bf16, vdupq_n_bf16): New.
889 (vdup_lane_bf16, vdup_laneq_bf16): New.
890 (vdupq_lane_bf16, vdupq_laneq_bf16): New.
891 (vduph_lane_bf16, vduph_laneq_bf16): New.
892 (vreinterpret_bf16_u8, vreinterpretq_bf16_u8): New.
893 (vreinterpret_bf16_u16, vreinterpretq_bf16_u16): New.
894 (vreinterpret_bf16_u32, vreinterpretq_bf16_u32): New.
895 (vreinterpret_bf16_u64, vreinterpretq_bf16_u64): New.
896 (vreinterpret_bf16_s8, vreinterpretq_bf16_s8): New.
897 (vreinterpret_bf16_s16, vreinterpretq_bf16_s16): New.
898 (vreinterpret_bf16_s32, vreinterpretq_bf16_s32): New.
899 (vreinterpret_bf16_s64, vreinterpretq_bf16_s64): New.
900 (vreinterpret_bf16_p8, vreinterpretq_bf16_p8): New.
901 (vreinterpret_bf16_p16, vreinterpretq_bf16_p16): New.
902 (vreinterpret_bf16_p64, vreinterpretq_bf16_p64): New
903 (vreinterpret_bf16_f16, vreinterpretq_bf16_f16): New
904 (vreinterpret_bf16_f32, vreinterpretq_bf16_f32): New.
905 (vreinterpret_bf16_f64, vreinterpretq_bf16_f64): New.
906 (vreinterpretq_bf16_p128): New.
907 (vreinterpret_s8_bf16, vreinterpretq_s8_bf16): New.
908 (vreinterpret_s16_bf16, vreinterpretq_s16_bf16): New.
909 (vreinterpret_s32_bf16, vreinterpretq_s32_bf16): New.
910 (vreinterpret_s64_bf16, vreinterpretq_s64_bf16): New.
911 (vreinterpret_u8_bf16, vreinterpretq_u8_bf16): New.
912 (vreinterpret_u16_bf16, vreinterpretq_u16_bf16): New.
913 (vreinterpret_u32_bf16, vreinterpretq_u32_bf16): New.
914 (vreinterpret_u64_bf16, vreinterpretq_u64_bf16): New.
915 (vreinterpret_p8_bf16, vreinterpretq_p8_bf16): New.
916 (vreinterpret_p16_bf16, vreinterpretq_p16_bf16): New.
917 (vreinterpret_p64_bf16, vreinterpretq_p64_bf16): New.
918 (vreinterpret_f32_bf16, vreinterpretq_f32_bf16): New.
919 (vreinterpret_f64_bf16,vreinterpretq_f64_bf16): New.
920 (vreinterpret_f16_bf16,vreinterpretq_f16_bf16): New.
921 (vreinterpretq_p128_bf16): New.
923 2020-02-25 Dennis Zhang <dennis.zhang@arm.com>
925 * config/arm/arm_neon.h (vbfdot_f32, vbfdotq_f32): New
926 (vbfdot_lane_f32, vbfdotq_laneq_f32): New.
927 (vbfdot_laneq_f32, vbfdotq_lane_f32): New.
928 * config/arm/arm_neon_builtins.def (vbfdot): New entry.
929 (vbfdot_lanev4bf, vbfdot_lanev8bf): Likewise.
930 * config/arm/iterators.md (VSF2BF): New attribute.
931 * config/arm/neon.md (neon_vbfdot<VCVTF:mode>): New entry.
932 (neon_vbfdot_lanev4bf<VCVTF:mode>): Likewise.
933 (neon_vbfdot_lanev8bf<VCVTF:mode>): Likewise.
935 2020-02-25 Christophe Lyon <christophe.lyon@linaro.org>
937 * config/arm/arm.md (required_for_purecode): New attribute.
938 (enabled): Handle required_for_purecode.
939 * config/arm/thumb1.md (thumb1_movsi_insn): Add alternative to
940 work with -mpure-code.
942 2020-02-25 Jakub Jelinek <jakub@redhat.com>
944 PR rtl-optimization/93908
945 * combine.c (find_split_point): For store into ZERO_EXTRACT, and src
948 2019-02-25 Eric Botcazou <ebotcazou@adacore.com>
950 * dwarf2out.c (dwarf2out_size_function): Run in early-DWARF mode.
952 2020-02-25 Roman Zhuykov <zhroma@ispras.ru>
954 * doc/install.texi (--enable-checking): Adjust wording.
956 2020-02-25 Richard Biener <rguenther@suse.de>
958 PR tree-optimization/93868
959 * tree-vect-slp.c (slp_copy_subtree): New function.
960 (vect_attempt_slp_rearrange_stmts): Copy the SLP tree before
961 re-arranging stmts in it.
963 2020-02-25 Jakub Jelinek <jakub@redhat.com>
966 * passes.c (pass_manager::dump_passes): Create a cgraph node for the
967 dummy function and remove it at the end.
970 * config/lm32/lm32.c (lm32_setup_incoming_varargs): Fix comment typo
971 paramter -> parameter.
972 * config/aarch64/aarch64.c (aarch64_is_extend_from_extract): Likewise.
973 * ipa-prop.h (struct ipa_agg_replacement_value): Likewise.
975 2020-02-24 Roman Zhuykov <zhroma@ispras.ru>
977 * doc/install.texi (--enable-checking): Properly document current
979 (--enable-stage1-checking): Minor clarification about bootstrap.
981 2020-02-24 David Malcolm <dmalcolm@redhat.com>
984 * doc/invoke.texi (-Wnanalyzer-tainted-array-index): Note that
985 -fanalyzer-checker=taint is also required.
986 (-fanalyzer-checker=): Note that providing this option enables the
987 given checker, and doing so may be required for checkers that are
990 2020-02-24 David Malcolm <dmalcolm@redhat.com>
992 * doc/invoke.texi (-fanalyzer-verbosity=): "2" only shows
993 significant control flow events; add a "3" which shows all
994 control flow events; the old "3" becomes "4".
996 2020-02-24 Jakub Jelinek <jakub@redhat.com>
998 PR tree-optimization/93582
999 * tree-ssa-sccvn.c (vn_walk_cb_data::push_partial_def): Consider
1000 pd.offset and pd.size to be counted in bits rather than bytes, add
1001 support for maxsizei that is not a multiple of BITS_PER_UNIT and
1002 handle bitfield stores and loads.
1003 (vn_reference_lookup_3): Don't call ranges_known_overlap_p with
1004 uncomparable quantities - bytes vs. bits. Allow push_partial_def
1005 on offsets/sizes that aren't multiple of BITS_PER_UNIT and adjust
1006 pd.offset/pd.size to be counted in bits rather than bytes.
1007 Formatting fix. Rename shadowed len variable to buflen.
1009 2020-02-24 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1010 Kugan Vivekandarajah <kugan.vivekanandarajah@linaro.org>
1013 * gcc.c (putenv_COLLECT_AS_OPTIONS): New function.
1014 (driver::main): Call putenv_COLLECT_AS_OPTIONS.
1015 * opts-common.c (parse_options_from_collect_gcc_options): New function.
1016 (prepend_xassembler_to_collect_as_options): Likewise.
1017 * opts.h (parse_options_from_collect_gcc_options): Declare prototype.
1018 (prepend_xassembler_to_collect_as_options): Likewise.
1019 * lto-opts.c (lto_write_options): Stream assembler options
1020 in COLLECT_AS_OPTIONS.
1021 * lto-wrapper.c (xassembler_options_error): New static variable.
1022 (get_options_from_collect_gcc_options): Move parsing options code to
1023 parse_options_from_collect_gcc_options and call it.
1024 (merge_and_complain): Validate -Xassembler options.
1025 (append_compiler_options): Handle OPT_Xassembler.
1026 (run_gcc): Append command line -Xassembler options to
1027 collect_gcc_options.
1028 * doc/invoke.texi: Add documentation about using Xassembler
1031 2020-02-24 Kito Cheng <kito.cheng@sifive.com>
1033 * config/riscv/riscv.c (riscv_emit_float_compare): Change the code gen
1035 (riscv_rtx_costs): Update cost model for LTGT.
1037 2020-02-23 Vladimir Makarov <vmakarov@redhat.com>
1039 PR rtl-optimization/93564
1040 * ira-color.c (struct update_cost_queue_elem): New member start.
1041 (queue_update_cost, get_next_update_cost): Add new arg start.
1042 (allocnos_conflict_p): New function.
1043 (update_costs_from_allocno): Add new arg conflict_cost_update_p.
1044 Add checking conflicts with allocnos_conflict_p.
1045 (update_costs_from_prefs, restore_costs_from_copies): Adjust
1046 update_costs_from_allocno calls.
1047 (update_conflict_hard_regno_costs): Add checking conflicts with
1048 allocnos_conflict_p. Adjust calls of queue_update_cost and
1049 get_next_update_cost.
1050 (assign_hard_reg): Adjust calls of queue_update_cost. Add
1052 (bucket_allocno_compare_func): Restore previous version.
1054 2020-02-21 John David Anglin <danglin@gcc.gnu.org>
1056 * gcc/config/pa/pa.c (pa_function_value): Fix check for word and
1057 double-word size when handling aggregate return values.
1058 * gcc/config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Fix to indicate
1059 that homogeneous SFmode and DFmode aggregates are passed and returned
1060 in general registers.
1062 2020-02-21 Jakub Jelinek <jakub@redhat.com>
1064 PR translation/93759
1065 * opts.c (print_filtered_help): Translate help before appending
1066 messages to it rather than after that.
1068 2020-02-19 Richard Sandiford <richard.sandiford@arm.com>
1070 PR rtl-optimization/PR92989
1071 * lra-lives.c (process_bb_lives): Restore the original order
1072 of the bb liveness update. Call make_hard_regno_dead for each
1073 register clobbered at the start of an EH receiver.
1075 2020-02-18 Feng Xue <fxue@os.amperecomputing.com>
1078 * ipa-cp.c (self_recursively_generated_p): Mark self-dependent value as
1079 self-recursively generated.
1081 2020-02-21 Iain Sandoe <iain@sandoe.co.uk>
1084 * config/darwin-c.c (pop_field_alignment): Adjust quoting of
1087 2020-02-21 Mihail Ionescu <mihail.ionescu@arm.com>
1089 * doc/sourcebuild.texi (arm_v8_1m_mve_ok):
1090 Document new target supports option.
1092 2020-02-21 Dennis Zhang <dennis.zhang@arm.com>
1094 * config/arm/arm_neon.h (vmmlaq_s32, vmmlaq_u32, vusmmlaq_s32): New.
1095 * config/arm/arm_neon_builtins.def (smmla, ummla, usmmla): New.
1096 * config/arm/iterators.md (MATMUL): New iterator.
1097 (sup): Add UNSPEC_MATMUL_S, UNSPEC_MATMUL_U, and UNSPEC_MATMUL_US.
1098 (mmla_sfx): New attribute.
1099 * config/arm/neon.md (neon_<sup>mmlav16qi): New.
1100 * config/arm/unspecs.md (UNSPEC_MATMUL_S, UNSPEC_MATMUL_U): New.
1101 (UNSPEC_MATMUL_US): New.
1103 2020-02-21 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1105 * config/arm/arm.md: Prevent scalar shifts from being used when big
1108 2020-02-21 Jan Hubicka <hubicka@ucw.cz>
1109 Richard Biener <rguenther@suse.de>
1111 PR tree-optimization/93586
1112 * tree-ssa-alias.c (nonoverlapping_array_refs_p): Finish array walk
1113 after mismatched array refs; do not sure type size information to
1114 recover from unmatched referneces with !flag_strict_aliasing_p.
1116 2020-02-21 Andrew Stubbs <ams@codesourcery.com>
1118 * config/gcn/gcn-valu.md (gather_load<mode>): Rename to ...
1119 (gather_load<mode>v64si): ... this and set operand 2 to V64SI.
1120 (scatter_store<mode>): Rename to ...
1121 (scatter_store<mode>v64si): ... this and set operand 1 to V64SI.
1122 (scatter<mode>_exec): Delete. Move contents ...
1123 (mask_scatter_store<mode>): ... here, and rename that to ...
1124 (mask_gather_load<mode>v64si): ... this. Set operand 2 to V64SI.
1125 Remove mode conversion.
1126 (mask_gather_load<mode>): Rename to ...
1127 (mask_scatter_store<mode>v64si): ... this. Set operand 1 to V64SI.
1128 Remove mode conversion.
1129 * config/gcn/gcn.c (gcn_expand_scaled_offsets): Remove mode conversion.
1131 2020-02-21 Martin Jambor <mjambor@suse.cz>
1133 PR tree-optimization/93845
1134 * tree-sra.c (verify_sra_access_forest): Only test access size of
1137 2020-02-21 Andrew Stubbs <ams@codesourcery.com>
1139 * config/gcn/gcn.c (gcn_hard_regno_mode_ok): Align VGPR pairs.
1140 * config/gcn/gcn-valu.md (addv64di3): Remove early-clobber.
1141 (addv64di3_exec): Likewise.
1142 (subv64di3): Likewise.
1143 (subv64di3_exec): Likewise.
1144 (addv64di3_zext): Likewise.
1145 (addv64di3_zext_exec): Likewise.
1146 (addv64di3_zext_dup): Likewise.
1147 (addv64di3_zext_dup_exec): Likewise.
1148 (addv64di3_zext_dup2): Likewise.
1149 (addv64di3_zext_dup2_exec): Likewise.
1150 (addv64di3_sext_dup2): Likewise.
1151 (addv64di3_sext_dup2_exec): Likewise.
1152 (<expander>v64di3): Likewise.
1153 (<expander>v64di3_exec): Likewise.
1154 (*<reduc_op>_dpp_shr_v64di): Likewise.
1155 (*plus_carry_dpp_shr_v64di): Likewise.
1156 * config/gcn/gcn.md (adddi3): Likewise.
1157 (addptrdi3): Likewise.
1158 (<expander>di3): Likewise.
1160 2020-02-21 Andrew Stubbs <ams@codesourcery.com>
1162 * config/gcn/gcn-valu.md (vec_seriesv64di): Use gen_vec_duplicatev64di.
1164 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
1166 * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Add SVE
1167 support. Use aarch64_emit_mult instead of emitting multiplication
1168 instructions directly.
1169 * config/aarch64/aarch64-sve.md (sqrt<mode>2, rsqrt<mode>2)
1170 (@aarch64_rsqrte<mode>, @aarch64_rsqrts<mode>): New expanders.
1172 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
1174 * config/aarch64/aarch64.c (aarch64_emit_mult): New function.
1175 (aarch64_emit_approx_div): Add SVE support. Use aarch64_emit_mult
1176 instead of emitting multiplication instructions directly.
1177 * config/aarch64/iterators.md (SVE_COND_FP_BINARY_OPTAB): New iterator.
1178 * config/aarch64/aarch64-sve.md (div<mode>3, @aarch64_frecpe<mode>)
1179 (@aarch64_frecps<mode>): New expanders.
1181 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
1183 * config/aarch64/aarch64-protos.h (AARCH64_APPROX_MODE): Operate
1184 on and produce uint64_ts rather than ints.
1185 (AARCH64_APPROX_NONE, AARCH64_APPROX_ALL): Change to uint64_ts.
1186 (cpu_approx_modes): Change the fields from unsigned int to uint64_t.
1188 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
1190 * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Don't create
1191 an unused xmsk register when handling approximate rsqrt.
1193 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
1195 * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Fix inverted
1196 flag_finite_math_only condition.
1198 2020-02-20 Uroš Bizjak <ubizjak@gmail.com>
1201 * config/i386/mmx.md (*vec_extractv2sf_1): Match source operand
1202 to destination operand for shufps alternative.
1203 (*vec_extractv2si_1): Ditto.
1205 2020-02-20 Peter Bergner <bergner@linux.ibm.com>
1208 * config/rs6000/rs6000.c (rs6000_legitimate_address_p): Handle VSX
1211 2020-02-20 Martin Liska <mliska@suse.cz>
1213 PR translation/93831
1214 * config/darwin.c (darwin_override_options): Change 64b to 64-bit mode.
1216 2020-02-20 Martin Liska <mliska@suse.cz>
1218 PR translation/93830
1219 * common/config/avr/avr-common.c: Remote trailing "|".
1221 2020-02-19 Bernd Edlinger <bernd.edlinger@hotmail.de>
1223 * collect2.c (maybe_run_lto_and_relink): Fix typo in
1226 2020-02-19 Richard Sandiford <richard.sandiford@arm.com>
1228 PR tree-optimization/93767
1229 * tree-vect-data-refs.c (vect_compile_time_alias): Remove the
1230 access-size bias from the offset calculations for negative strides.
1232 2020-02-19 Bernd Edlinger <bernd.edlinger@hotmail.de>
1234 * collect2.c (c_file, o_file): Make const again.
1235 (ldout,lderrout, dump_ld_file): Remove.
1236 (tool_cleanup): Avoid calling not signal-safe functions.
1237 (maybe_run_lto_and_relink): Avoid possible signal handler
1238 access to unintialzed memory (lto_o_files).
1239 (main): Avoid leaking temp files in $TMPDIR.
1240 Initialize c_file/o_file with concat, which avoids exposing
1241 uninitialized memory to signal handler, which calls unlink(!).
1242 Avoid calling maybe_unlink when the main function returns,
1243 since the atexit handler is already doing this.
1244 * collect2.h (dump_ld_file, ldout, lderrout): Remove.
1246 2020-02-19 Martin Jambor <mjambor@suse.cz>
1248 PR tree-optimization/93776
1249 * tree-sra.c (create_access): Do not create zero size accesses.
1250 (get_access_for_expr): Do not search for zero sized accesses.
1252 2020-02-19 Martin Jambor <mjambor@suse.cz>
1254 PR tree-optimization/93667
1255 * tree-sra.c (scalarizable_type_p): Return false if record fields
1256 do not follow wach other.
1258 2020-01-21 Kito Cheng <kito.cheng@sifive.com>
1260 * config/riscv/riscv.c (riscv_output_move) Using fmv.x.w/fmv.w.x
1261 rather than fmv.x.s/fmv.s.x.
1263 2020-02-18 James Greenhalgh <james.greenhalgh@arm.com>
1265 * config/aarch64/aarch64-simd-builtins.def
1266 (intrinsic_vec_smult_lo_): New.
1267 (intrinsic_vec_umult_lo_): Likewise.
1268 (vec_widen_smult_hi_): Likewise.
1269 (vec_widen_umult_hi_): Likewise.
1270 * config/aarch64/aarch64-simd.md
1271 (aarch64_intrinsic_vec_<su>mult_lo_<mode>): New.
1272 * config/aarch64/arm_neon.h (vmull_high_s8): Use intrinsics.
1273 (vmull_high_s16): Likewise.
1274 (vmull_high_s32): Likewise.
1275 (vmull_high_u8): Likewise.
1276 (vmull_high_u16): Likewise.
1277 (vmull_high_u32): Likewise.
1278 (vmull_s8): Likewise.
1279 (vmull_s16): Likewise.
1280 (vmull_s32): Likewise.
1281 (vmull_u8): Likewise.
1282 (vmull_u16): Likewise.
1283 (vmull_u32): Likewise.
1285 2020-02-18 Martin Liska <mliska@suse.cz>
1287 * value-prof.c (stream_out_histogram_value): Restore LTO PGO
1288 bootstrap by missing removal of invalid sanity check.
1290 2020-02-18 Martin Liska <mliska@suse.cz>
1293 * ipa-icf-gimple.c (func_checker::compare_gimple_assign):
1294 Always compare LHS of gimple_assign.
1296 2020-02-18 Martin Liska <mliska@suse.cz>
1299 * cgraph.c (cgraph_node::verify_node): Verify MALLOC attribute
1300 and return type of functions.
1301 * ipa-param-manipulation.c (ipa_param_adjustments::adjust_decl):
1302 Drop MALLOC attribute for void functions.
1303 * ipa-pure-const.c (funct_state_summary_t::duplicate): Drop
1304 malloc_state for a new VOID clone.
1306 2020-02-18 Martin Liska <mliska@suse.cz>
1309 * common.opt: Add -fprofile-reproducibility.
1310 * doc/invoke.texi: Document it.
1311 * value-prof.c (dump_histogram_value):
1312 Document and support behavior for counters[0]
1313 being a negative value.
1314 (get_nth_most_common_value): Handle negative
1315 counters[0] in respect to flag_profile_reproducible.
1317 2020-02-18 Jakub Jelinek <jakub@redhat.com>
1320 * cgraph.c (verify_speculative_call): Use speculative_id instead of
1321 speculative_uid in messages. Remove trailing whitespace from error
1322 message. Use num_speculative_call_targets instead of
1323 num_speculative_targets in a message.
1324 (cgraph_node::verify_node): Use call_stmt instead of cal_stmt in
1325 edge messages and stmt instead of cal_stmt in reference message.
1327 PR tree-optimization/93780
1328 * tree-ssa.c (non_rewritable_lvalue_p): Check valid_vector_subparts_p
1329 before calling build_vector_type.
1330 (execute_update_addresses_taken): Likewise.
1333 * params.opt (-param=ipa-max-switch-predicate-bounds=): Fix help
1334 typo, functoin -> function.
1335 * tree.c (free_lang_data_in_decl): Fix comment typo,
1336 functoin -> function.
1337 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
1339 2020-02-17 David Malcolm <dmalcolm@redhat.com>
1341 * diagnostic.c (print_any_cwe): Don't call get_cwe_url if URLs
1343 (print_option_information): Don't call get_option_url if URLs
1346 2020-02-17 Alexandre Oliva <oliva@adacore.com>
1348 * tree-emutls.c (new_emutls_decl, emutls_common_1): Complete
1349 handling of register_common-less targets.
1351 2020-02-17 Martin Liska <mliska@suse.cz>
1354 * ipa-devirt.c (odr_types_equivalent_p): Fix grammar.
1356 2020-02-17 Martin Liska <mliska@suse.cz>
1358 PR translation/93755
1359 * config/rs6000/rs6000.c (rs6000_option_override_internal):
1362 2020-02-17 Martin Liska <mliska@suse.cz>
1365 * config/rx/elf.opt: Fix typo.
1367 2020-02-17 Richard Biener <rguenther@suse.de>
1370 * opts-global.c (print_ignored_options): Use inform and
1373 2020-02-17 Jiufu Guo <guojiufu@linux.ibm.com>
1376 * config/rs6000/rs6000.md (untyped_call): Add emit_clobber.
1378 2020-02-16 Uroš Bizjak <ubizjak@gmail.com>
1381 * config/i386/i386.md (atan2xf3): Swap operands 1 and 2.
1382 (atan2<mode>3): Update operand order in the call to gen_atan2xf3.
1384 2020-02-15 Jason Merrill <jason@redhat.com>
1386 * doc/invoke.texi (C Dialect Options): Add -std=c++20.
1388 2020-02-15 Jakub Jelinek <jakub@redhat.com>
1390 PR tree-optimization/93744
1391 * match.pd (((m1 >/</>=/<= m2) * d -> (m1 >/</>=/<= m2) ? d : 0,
1392 A - ((A - B) & -(C cmp D)) -> (C cmp D) ? B : A,
1393 A + ((B - A) & -(C cmp D)) -> (C cmp D) ? B : A): For GENERIC, make
1394 sure @2 in the first and @1 in the other patterns has no side-effects.
1396 2020-02-15 David Malcolm <dmalcolm@redhat.com>
1397 Bernd Edlinger <bernd.edlinger@hotmail.de>
1401 * config.in (DIAGNOSTICS_URLS_DEFAULT): New define.
1402 * configure.ac (--with-diagnostics-urls): New configuration
1403 option, based on --with-diagnostics-color.
1404 (DIAGNOSTICS_URLS_DEFAULT): New define.
1405 * config.h: Regenerate.
1406 * configure: Regenerate.
1407 * diagnostic.c (diagnostic_urls_init): Handle -1 for
1408 DIAGNOSTICS_URLS_DEFAULT from configure-time
1409 --with-diagnostics-urls=auto-if-env by querying for a GCC_URLS
1410 and TERM_URLS environment variable.
1411 * diagnostic-url.h (diagnostic_url_format): New enum type.
1412 (diagnostic_urls_enabled_p): rename to...
1413 (determine_url_format): ... this, and change return type.
1414 * diagnostic-color.c (parse_env_vars_for_urls): New helper function.
1415 (auto_enable_urls): Disable URLs on xfce4-terminal, gnome-terminal,
1416 the linux console, and mingw.
1417 (diagnostic_urls_enabled_p): rename to...
1418 (determine_url_format): ... this, and adjust.
1419 * pretty-print.h (pretty_printer::show_urls): rename to...
1420 (pretty_printer::url_format): ... this, and change to enum.
1421 * pretty-print.c (pretty_printer::pretty_printer,
1422 pp_begin_url, pp_end_url, test_urls): Adjust.
1423 * doc/install.texi (--with-diagnostics-urls): Document the new
1424 configuration option.
1425 (--with-diagnostics-color): Document the existing interaction
1426 with GCC_COLORS better.
1427 * doc/invoke.texi (-fdiagnostics-urls): Add GCC_URLS and TERM_URLS
1428 vindex reference. Update description of defaults based on the above.
1429 (-fdiagnostics-color): Update description of how -fdiagnostics-color
1430 interacts with GCC_COLORS.
1432 2020-02-14 Eric Botcazou <ebotcazou@adacore.com>
1435 * config/sparc/sparc.c (eligible_for_call_delay): Test HAVE_GNU_LD in
1436 conjunction with TARGET_GNU_TLS in early return.
1438 2020-02-14 Alexander Monakov <amonakov@ispras.ru>
1440 * rtlanal.c (rtx_cost): Handle a SET up front. Avoid division if
1441 the mode is not wider than UNITS_PER_WORD.
1443 2020-02-14 Martin Jambor <mjambor@suse.cz>
1445 PR tree-optimization/93516
1446 * tree-sra.c (propagate_subaccesses_from_rhs): Do not create
1447 access of the same type as the parent.
1448 (propagate_subaccesses_from_lhs): Likewise.
1450 2020-02-14 Hongtao Liu <hongtao.liu@intel.com>
1453 * config/i386/avx512vbmi2intrin.h
1454 (_mm512_shrdi_epi16, _mm512_mask_shrdi_epi16,
1455 _mm512_maskz_shrdi_epi16, _mm512_shrdi_epi32,
1456 _mm512_mask_shrdi_epi32, _mm512_maskz_shrdi_epi32,
1457 _m512_shrdi_epi64, _m512_mask_shrdi_epi64,
1458 _m512_maskz_shrdi_epi64, _mm512_shldi_epi16,
1459 _mm512_mask_shldi_epi16, _mm512_maskz_shldi_epi16,
1460 _mm512_shldi_epi32, _mm512_mask_shldi_epi32,
1461 _mm512_maskz_shldi_epi32, _mm512_shldi_epi64,
1462 _mm512_mask_shldi_epi64, _mm512_maskz_shldi_epi64): Fix typo
1463 of lacking a closing parenthesis.
1464 * config/i386/avx512vbmi2vlintrin.h
1465 (_mm256_shrdi_epi16, _mm256_mask_shrdi_epi16,
1466 _mm256_maskz_shrdi_epi16, _mm256_shrdi_epi32,
1467 _mm256_mask_shrdi_epi32, _mm256_maskz_shrdi_epi32,
1468 _m256_shrdi_epi64, _m256_mask_shrdi_epi64,
1469 _m256_maskz_shrdi_epi64, _mm256_shldi_epi16,
1470 _mm256_mask_shldi_epi16, _mm256_maskz_shldi_epi16,
1471 _mm256_shldi_epi32, _mm256_mask_shldi_epi32,
1472 _mm256_maskz_shldi_epi32, _mm256_shldi_epi64,
1473 _mm256_mask_shldi_epi64, _mm256_maskz_shldi_epi64,
1474 _mm_shrdi_epi16, _mm_mask_shrdi_epi16,
1475 _mm_maskz_shrdi_epi16, _mm_shrdi_epi32,
1476 _mm_mask_shrdi_epi32, _mm_maskz_shrdi_epi32,
1477 _mm_shrdi_epi64, _mm_mask_shrdi_epi64,
1478 _m_maskz_shrdi_epi64, _mm_shldi_epi16,
1479 _mm_mask_shldi_epi16, _mm_maskz_shldi_epi16,
1480 _mm_shldi_epi32, _mm_mask_shldi_epi32,
1481 _mm_maskz_shldi_epi32, _mm_shldi_epi64,
1482 _mm_mask_shldi_epi64, _mm_maskz_shldi_epi64): Ditto.
1484 2020-02-13 H.J. Lu <hongjiu.lu@intel.com>
1487 * config/i386/i386.c (ix86_trampoline_init): Skip ENDBR32 at
1488 the target function entry.
1490 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
1492 * common/config/arc/arc-common.c (arc_option_optimization_table):
1493 Disable if-conversion step when optimized for size.
1495 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
1497 * config/arc/arc.c (arc_conditional_register_usage): R0-R3 and
1498 R12-R15 are always in ARCOMPACT16_REGS register class.
1499 * config/arc/arc.opt (mq-class): Deprecate.
1500 * config/arc/constraint.md ("q"): Remove dependency on mq-class
1502 * doc/invoke.texi (mq-class): Update text.
1503 * common/config/arc/arc-common.c (arc_option_optimization_table):
1506 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
1508 * config/arc/arc.c (arc_insn_cost): New function.
1509 (TARGET_INSN_COST): Define.
1510 * config/arc/arc.md (cost): New attribute.
1511 (add_n): Use arc_nonmemory_operand.
1512 (ashlsi3_insn): Likewise, also update constraints.
1513 (ashrsi3_insn): Likewise.
1514 (rotrsi3): Likewise.
1515 (add_shift): Likewise.
1516 * config/arc/predicates.md (arc_nonmemory_operand): New predicate.
1518 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
1520 * config/arc/arc.md (mulsidi_600): Correctly select mlo/mhi
1522 (umulsidi_600): Likewise.
1524 2020-02-13 Jakub Jelinek <jakub@redhat.com>
1527 * config/i386/avx512bitalgintrin.h (_mm512_mask_popcnt_epi8,
1528 _mm512_mask_popcnt_epi16, _mm256_mask_popcnt_epi8,
1529 _mm256_mask_popcnt_epi16, _mm_mask_popcnt_epi8,
1530 _mm_mask_popcnt_epi16): Rename __B argument to __A and __A to __W,
1531 pass __A to the builtin followed by __W instead of __A followed by
1533 * config/i386/avx512vpopcntdqintrin.h (_mm512_mask_popcnt_epi32,
1534 _mm512_mask_popcnt_epi64): Likewise.
1535 * config/i386/avx512vpopcntdqvlintrin.h (_mm_mask_popcnt_epi32,
1536 _mm256_mask_popcnt_epi32, _mm_mask_popcnt_epi64,
1537 _mm256_mask_popcnt_epi64): Likewise.
1539 PR tree-optimization/93582
1540 * fold-const.h (shift_bytes_in_array_left,
1541 shift_bytes_in_array_right): Declare.
1542 * fold-const.c (shift_bytes_in_array_left,
1543 shift_bytes_in_array_right): New function, moved from
1544 gimple-ssa-store-merging.c, no longer static.
1545 * gimple-ssa-store-merging.c (shift_bytes_in_array): Move
1546 to gimple-ssa-store-merging.c and rename to shift_bytes_in_array_left.
1547 (shift_bytes_in_array_right): Move to gimple-ssa-store-merging.c.
1548 (encode_tree_to_bitpos): Use shift_bytes_in_array_left instead of
1549 shift_bytes_in_array.
1550 (verify_shift_bytes_in_array): Rename to ...
1551 (verify_shift_bytes_in_array_left): ... this. Use
1552 shift_bytes_in_array_left instead of shift_bytes_in_array.
1553 (store_merging_c_tests): Call verify_shift_bytes_in_array_left
1554 instead of verify_shift_bytes_in_array.
1555 * tree-ssa-sccvn.c (vn_reference_lookup_3): For native_encode_expr
1556 / native_interpret_expr where the store covers all needed bits,
1557 punt on PDP-endian, otherwise allow all involved offsets and sizes
1558 not to be byte-aligned.
1561 * config/i386/sse.md (k<code><mode>): Drop mode from last operand and
1562 use const_0_to_255_operand predicate instead of immediate_operand.
1563 (avx512dq_fpclass<mode><mask_scalar_merge_name>,
1564 avx512dq_vmfpclass<mode><mask_scalar_merge_name>,
1565 vgf2p8affineinvqb_<mode><mask_name>,
1566 vgf2p8affineqb_<mode><mask_name>): Drop mode from
1567 const_0_to_255_operand predicated operands.
1569 2020-02-12 Jeff Law <law@redhat.com>
1571 * config/h8300/h8300.md (comparison shortening peepholes): Use
1572 a mode iterator to merge the HImode and SImode peepholes.
1574 2020-02-12 Jakub Jelinek <jakub@redhat.com>
1577 * real.c (is_even): Make static. Function comment fix.
1578 (is_halfway_below): Make static, don't assert R is not inf/nan,
1579 instead return false for those. Small formatting fixes.
1581 2020-02-12 Martin Sebor <msebor@redhat.com>
1584 * tree-ssa-strlen.c (handle_builtin_stxncpy): Rename...
1585 (handle_builtin_stxncpy_strncat): ...to this. Change first argument.
1586 Issue only -Wstringop-overflow strncat, never -Wstringop-truncation.
1587 (strlen_check_and_optimize_call): Adjust callee name.
1589 2020-02-12 Jeff Law <law@redhat.com>
1591 * config/h8300/h8300.md (comparison shortening peepholes): Drop
1592 (and (xor)) variant. Combine other two into single peephole.
1594 2020-02-12 Wilco Dijkstra <wdijkstr@arm.com>
1596 PR rtl-optimization/93565
1597 * config/aarch64/aarch64.c (aarch64_rtx_costs): Add CTZ costs.
1599 2020-02-12 Wilco Dijkstra <wdijkstr@arm.com>
1601 * config/aarch64/aarch64-simd.md
1602 (aarch64_zero_extend<GPI:mode>_reduc_plus_<VDQV_E:mode>): New pattern.
1603 * config/aarch64/aarch64.md (popcount<mode>2): Use it instead of
1604 generating separate ADDV and zero_extend patterns.
1605 * config/aarch64/iterators.md (VDQV_E): New iterator.
1607 2020-02-12 Jeff Law <law@redhat.com>
1609 * config/h8300/h8300.md (cpymemsi, movmd): Remove dead patterns,
1610 expanders, splits, etc.
1611 (movmd_internal_<mode>, movmd splitter, movstr, movsd): Likewise.
1612 (stpcpy_internal_<mode>, stpcpy splitter): Likewise.
1613 (peepholes to convert QI/HI mode pushes to SI mode pushes): Likewise.
1614 * config/h8300/h8300.c (h8300_swap_into_er6): Remove unused function.
1615 (h8300_swap_out_of_er6, h8sx_emit_movmd): Likewise
1616 * config/h8300/h8300-protos.h (h8300_swap_into_er6): Remove unused
1618 (h8300_swap_out_of_er6, h8sx_emit_movmd): Likewise.
1620 2020-02-12 Jakub Jelinek <jakub@redhat.com>
1623 * config/i386/sse.md (VI48F_256_DQ): New mode iterator.
1624 (avx512vl_vextractf128<mode>): Use it instead of VI48F_256. Remove
1625 TARGET_AVX512DQ from condition.
1626 (vec_extract_lo_<mode><mask_name>): Use <mask_avx512dq_condition>
1627 instead of <mask_mode512bit_condition> in condition. If
1628 TARGET_AVX512DQ is false, emit vextract*64x4 instead of
1630 (vec_extract_lo_<mode><mask_name>): Drop <mask_avx512dq_condition>
1633 2020-02-12 Kewen Lin <linkw@gcc.gnu.org>
1636 * ira.c (combine_and_move_insns): Skip multiple_sets def_insn.
1638 2020-02-12 Segher Boessenkool <segher@kernel.crashing.org>
1640 * config/rs6000/rs6000.c (rs6000_debug_print_mode): Don't use sizeof
1641 where strlen is more legible.
1642 (rs6000_builtin_vectorized_libmass): Ditto.
1643 (rs6000_print_options_internal): Ditto.
1645 2020-02-11 Martin Sebor <msebor@redhat.com>
1647 PR tree-optimization/93683
1648 * tree-ssa-alias.c (stmt_kills_ref_p): Avoid using LHS when not set.
1650 2020-02-11 Michael Meissner <meissner@linux.ibm.com>
1652 * config/rs6000/predicates.md (cint34_operand): Rename the
1653 -mprefixed-addr option to be -mprefixed.
1654 * config/rs6000/rs6000-cpus.def (ISA_FUTURE_MASKS_SERVER): Rename
1655 the -mprefixed-addr option to be -mprefixed.
1656 (OTHER_FUTURE_MASKS): Likewise.
1657 (POWERPC_MASKS): Likewise.
1658 * config/rs6000/rs6000.c (rs6000_option_override_internal): Rename
1659 the -mprefixed-addr option to be -mprefixed. Change error
1660 messages to refer to -mprefixed.
1661 (num_insns_constant_gpr): Rename the -mprefixed-addr option to be
1663 (rs6000_legitimate_offset_address_p): Likewise.
1664 (rs6000_mode_dependent_address): Likewise.
1665 (rs6000_opt_masks): Change the spelling of "-mprefixed-addr" to be
1666 "-mprefixed" for target attributes and pragmas.
1667 (address_to_insn_form): Rename the -mprefixed-addr option to be
1669 (rs6000_adjust_insn_length): Likewise.
1670 * config/rs6000/rs6000.h (FINAL_PRESCAN_INSN): Rename the
1671 -mprefixed-addr option to be -mprefixed.
1672 (ASM_OUTPUT_OPCODE): Likewise.
1673 * config/rs6000/rs6000.md (prefixed insn attribute): Rename the
1674 -mprefixed-addr option to be -mprefixed.
1675 * config/rs6000/rs6000.opt (-mprefixed): Rename the
1676 -mprefixed-addr option to be prefixed. Change the option from
1677 being undocumented to being documented.
1678 * doc/invoke.texi (RS/6000 and PowerPC Options): Document the
1679 -mprefixed option. Update the -mpcrel documentation to mention
1682 2020-02-11 Hans-Peter Nilsson <hp@axis.com>
1684 * ira-conflicts.c (print_hard_reg_set): Correct output for sets
1685 including FIRST_PSEUDO_REGISTER - 1.
1686 * ira-color.c (print_hard_reg_set): Ditto.
1688 2020-02-11 Stam Markianos-Wright <stam.markianos-wright@arm.com>
1690 * config/arm/arm-builtins.c (enum arm_type_qualifiers):
1691 (USTERNOP_QUALIFIERS): New define.
1692 (USMAC_LANE_QUADTUP_QUALIFIERS): New define.
1693 (SUMAC_LANE_QUADTUP_QUALIFIERS): New define.
1694 (arm_expand_builtin_args): Add case ARG_BUILTIN_LANE_QUADTUP_INDEX.
1695 (arm_expand_builtin_1): Add qualifier_lane_quadtup_index.
1696 * config/arm/arm_neon.h (vusdot_s32): New.
1697 (vusdot_lane_s32): New.
1698 (vusdotq_lane_s32): New.
1699 (vsudot_lane_s32): New.
1700 (vsudotq_lane_s32): New.
1701 * config/arm/arm_neon_builtins.def (usdot, usdot_lane,sudot_lane): New.
1702 * config/arm/iterators.md (DOTPROD_I8MM): New.
1703 (sup, opsuffix): Add <us/su>.
1704 * config/arm/neon.md (neon_usdot, <us/su>dot_lane: New.
1705 * config/arm/unspecs.md (UNSPEC_DOT_US, UNSPEC_DOT_SU): New.
1707 2020-02-11 Richard Biener <rguenther@suse.de>
1709 PR tree-optimization/93661
1710 PR tree-optimization/93662
1711 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly guard
1713 * tree-sra.c (get_access_for_expr): Likewise.
1715 2020-02-10 Jakub Jelinek <jakub@redhat.com>
1718 * config/i386/sse.md (VI_256_AVX2): New mode iterator.
1719 (vcond_mask_<mode><sseintvecmodelower>): Use it instead of VI_256.
1720 Change condition from TARGET_AVX2 to TARGET_AVX.
1722 2020-02-10 Iain Sandoe <iain@sandoe.co.uk>
1725 * config/darwin-c.c (darwin_cfstring_ref_p): Fix up last
1726 argument of strncmp.
1728 2020-02-10 Hans-Peter Nilsson <hp@axis.com>
1730 Try to generate zero-based comparisons.
1731 * config/cris/cris.c (cris_reduce_compare): New function.
1732 * config/cris/cris-protos.h (cris_reduce_compare): Add prototype.
1733 * config/cris/cris.md ("cbranch<mode>4", "cbranchdi4", "cstoredi4")
1734 (cstore<mode>4"): Apply cris_reduce_compare in expanders.
1736 2020-02-10 Richard Earnshaw <rearnsha@arm.com>
1739 * config/arm/arm.md (movsi_compare0): Allow SP as a source register
1740 in Thumb state and also as a destination in Arm state. Add T16
1743 2020-02-10 Hans-Peter Nilsson <hp@axis.com>
1745 * md.texi (Define Subst): Match closing paren in example.
1747 2020-02-10 Jakub Jelinek <jakub@redhat.com>
1751 * config/i386/i386.c (x86_64_elf_section_type_flags): Fix up last
1752 arguments of strncmp.
1754 2020-02-10 Feng Xue <fxue@os.amperecomputing.com>
1757 * ipa-cp.c (ipcp_lattice::add_value): Add source with same call edge
1758 but different source value.
1759 (adjust_callers_for_value_intersection): New function.
1760 (gather_edges_for_value): Adjust order of callers to let a
1761 non-self-recursive caller be the first element.
1762 (self_recursive_pass_through_p): Add a new parameter "simple", and
1763 check generalized self-recursive pass-through jump function.
1764 (self_recursive_agg_pass_through_p): Likewise.
1765 (find_more_scalar_values_for_callers_subset): Compute value from
1766 pass-through jump function for self-recursive.
1767 (intersect_with_plats): Cleanup previous implementation code for value
1768 itersection with self-recursive call edge.
1769 (intersect_with_agg_replacements): Likewise.
1770 (intersect_aggregates_with_edge): Deduce value from pass-through jump
1771 function for self-recursive call edge. Cleanup previous implementation
1772 code for value intersection with self-recursive call edge.
1773 (decide_whether_version_node): Remove dead callers and adjust order
1774 to let a non-self-recursive caller be the first element.
1776 2020-02-09 Uroš Bizjak <ubizjak@gmail.com>
1778 * recog.c: Move pass_split_before_sched2 code in front of
1779 pass_split_before_regstack.
1780 (pass_data_split_before_sched2): Rename pass to split3 from split4.
1781 (pass_data_split_before_regstack): Rename pass to split4 from split3.
1782 (rest_of_handle_split_before_sched2): Remove.
1783 (pass_split_before_sched2::execute): Unconditionally call
1785 (enable_split_before_sched2): New function.
1786 (pass_split_before_sched2::gate): Use enable_split_before_sched2.
1787 (pass_split_before_regstack::gate): Ditto.
1788 * config/nds32/nds32.c (nds32_split_double_word_load_store_p):
1789 Update name check for renamed split4 pass.
1790 * config/sh/sh.c (register_sh_passes): Update pass insertion
1791 point for renamed split4 pass.
1793 2020-02-09 Jakub Jelinek <jakub@redhat.com>
1795 * gimplify.c (gimplify_adjust_omp_clauses_1): Promote
1796 DECL_IN_CONSTANT_POOL variables into "omp declare target" to avoid
1797 copying them around between host and target.
1799 2020-02-08 Andrew Pinski <apinski@marvell.com>
1802 * config/aarch64/aarch64-simd.md (movmisalign<mode>): Check
1803 STRICT_ALIGNMENT also.
1805 2020-02-08 Jim Wilson <jimw@sifive.com>
1808 * config/riscv/riscv.h (HARD_REGNO_CALLER_SAVE_MODE): Define.
1810 2020-02-08 Uroš Bizjak <ubizjak@gmail.com>
1811 Jakub Jelinek <jakub@redhat.com>
1814 * config/i386/i386.h (CALL_USED_REGISTERS): Make
1815 xmm16-xmm31 call-used even in 64-bit ms-abi.
1817 2020-02-07 Dennis Zhang <dennis.zhang@arm.com>
1819 * config/aarch64/aarch64-simd-builtins.def (simd_smmla): New entry.
1820 (simd_ummla, simd_usmmla): Likewise.
1821 * config/aarch64/aarch64-simd.md (aarch64_simd_<sur>mmlav16qi): New.
1822 * config/aarch64/arm_neon.h (vmmlaq_s32, vmmlaq_u32): New.
1823 (vusmmlaq_s32): New.
1825 2020-02-07 Richard Biener <rguenther@suse.de>
1828 * tree-inline.c (fold_marked_statements): Do a PRE walk,
1829 skipping unreachable regions.
1830 (optimize_inline_calls): Skip folding stmts when we didn't
1833 2020-02-07 H.J. Lu <hongjiu.lu@intel.com>
1836 * config/i386/i386.c (function_arg_ms_64): Add a type argument.
1837 Don't return aggregates with only SFmode and DFmode in SSE
1839 (ix86_function_arg): Pass arg.type to function_arg_ms_64.
1841 2020-02-07 Jakub Jelinek <jakub@redhat.com>
1844 * config/rs6000/rs6000-logue.c
1845 (rs6000_emit_probe_stack_range_stack_clash): Always use gen_add3_insn,
1846 if it fails, move rs into end_addr and retry. Add
1847 REG_FRAME_RELATED_EXPR note whenever it returns more than one insn or
1848 the insn pattern doesn't describe well what exactly happens to
1852 * config/i386/predicates.md (avx_identity_operand): Remove.
1853 * config/i386/sse.md (*avx_vec_concat<mode>_1): Remove.
1854 (avx_<castmode><avxsizesuffix>_<castmode>,
1855 avx512f_<castmode><avxsizesuffix>_256<castmode>): Change patterns to
1856 a VEC_CONCAT of the operand and UNSPEC_CAST.
1857 (avx512f_<castmode><avxsizesuffix>_<castmode>): Change pattern to
1858 a VEC_CONCAT of VEC_CONCAT of the operand and UNSPEC_CAST with
1862 * config/i386/i386.c (ix86_lea_outperforms): Make sure to clear
1863 recog_data.insn if distance_non_agu_define changed it.
1865 2020-02-06 Michael Meissner <meissner@linux.ibm.com>
1868 * config/rs6000/rs6000.c (reg_to_non_prefixed): Before ISA 3.0
1869 we only had X-FORM (reg+reg) addressing for vectors. Also before
1870 ISA 3.0, we only had X-FORM addressing for scalars in the
1871 traditional Altivec registers.
1873 2020-02-06 <zhongyunde@huawei.com>
1874 Vladimir Makarov <vmakarov@redhat.com>
1876 PR rtl-optimization/93561
1877 * lra-assigns.c (spill_for): Check that tested hard regno is not out of
1878 hard register range.
1880 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
1882 * config/aarch64/aarch64.md (aarch64_movk<mode>): Add a type
1885 2020-02-06 Segher Boessenkool <segher@kernel.crashing.org>
1887 * config/rs6000/rs6000.c (rs6000_emit_set_long_const): Handle the case
1888 where the low and the high 32 bits are equal to each other specially,
1889 with an rldimi instruction.
1891 2020-02-06 Mihail Ionescu <mihail.ionescu@arm.com>
1893 * config/arm/arm-cpus.in: Set profile M for armv8.1-m.main.
1895 2020-02-06 Mihail Ionescu <mihail.ionescu@arm.com>
1897 * config/arm/arm-tables.opt: Regenerate.
1899 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
1902 * config/aarch64/aarch64-protos.h (aarch64_movk_shift): Declare.
1903 * config/aarch64/aarch64.c (aarch64_movk_shift): New function.
1904 * config/aarch64/aarch64.md (aarch64_movk<mode>): New pattern.
1906 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
1908 PR rtl-optimization/87763
1909 * config/aarch64/aarch64.md (*ashiftsi_extvdi_bfiz): New pattern.
1911 2020-02-06 Delia Burduv <delia.burduv@arm.com>
1913 * config/aarch64/aarch64-simd-builtins.def
1914 (bfmlaq): New built-in function.
1915 (bfmlalb): New built-in function.
1916 (bfmlalt): New built-in function.
1917 (bfmlalb_lane): New built-in function.
1918 (bfmlalt_lane): New built-in function.
1919 * config/aarch64/aarch64-simd.md
1920 (aarch64_bfmmlaqv4sf): New pattern.
1921 (aarch64_bfmlal<bt>v4sf): New pattern.
1922 (aarch64_bfmlal<bt>_lane<q>v4sf): New pattern.
1923 * config/aarch64/arm_neon.h (vbfmmlaq_f32): New intrinsic.
1924 (vbfmlalbq_f32): New intrinsic.
1925 (vbfmlaltq_f32): New intrinsic.
1926 (vbfmlalbq_lane_f32): New intrinsic.
1927 (vbfmlaltq_lane_f32): New intrinsic.
1928 (vbfmlalbq_laneq_f32): New intrinsic.
1929 (vbfmlaltq_laneq_f32): New intrinsic.
1930 * config/aarch64/iterators.md (BF_MLA): New int iterator.
1931 (bt): New int attribute.
1933 2020-02-06 Uroš Bizjak <ubizjak@gmail.com>
1935 * config/i386/i386.md (*pushtf): Emit "#" instead of
1936 calling gcc_unreachable in insn output.
1939 (*pushsf_rex64): Ditto for alternatives other than 1.
1940 (*pushsf): Ditto for alternatives other than 1.
1942 2020-02-06 Martin Liska <mliska@suse.cz>
1944 PR gcov-profile/91971
1945 PR gcov-profile/93466
1946 * coverage.c (coverage_init): Revert mangling of
1947 path into filename. It can lead to huge filename length.
1948 Creation of subfolders seem more natural.
1950 2020-02-06 Stam Markianos-Wright <stam.markianos-wright@arm.com>
1953 * config/arm/arm.c (arm_block_arith_comp_libfuncs_for_mode): New.
1954 (arm_init_libfuncs): Add BFmode support to block spurious BF libfuncs.
1955 Use arm_block_arith_comp_libfuncs_for_mode for HFmode.
1957 2020-02-06 Jakub Jelinek <jakub@redhat.com>
1960 * config/i386/predicates.md (avx_identity_operand): New predicate.
1961 * config/i386/sse.md (*avx_vec_concat<mode>_1): New
1962 define_insn_and_split.
1965 * omp-low.c (use_pointer_for_field): For nested constructs, also
1966 look for map clauses on target construct.
1967 (scan_omp_1_stmt) <case GIMPLE_OMP_TARGET>: Bump temporarily
1968 taskreg_nesting_level.
1971 * gimplify.c (gimplify_scan_omp_clauses) <do_notice>: If adding
1972 shared clause, call omp_notice_variable on outer context if any.
1974 2020-02-05 Jason Merrill <jason@redhat.com>
1977 * symtab.c (symtab_node::nonzero_address): A DECL_COMDAT decl has
1978 non-zero address even if weak and not yet defined.
1980 2020-02-05 Martin Sebor <msebor@redhat.com>
1982 PR tree-optimization/92765
1983 * gimple-fold.c (get_range_strlen_tree): Handle MEM_REF and PARM_DECL.
1984 * tree-ssa-strlen.c (compute_string_length): Remove.
1985 (determine_min_objsize): Remove.
1986 (get_len_or_size): Add an argument. Call get_range_strlen_dynamic.
1987 Avoid using type size as the upper bound on string length.
1988 (handle_builtin_string_cmp): Add an argument. Adjust.
1989 (strlen_check_and_optimize_call): Pass additional argument to
1990 handle_builtin_string_cmp.
1992 2020-02-05 Uroš Bizjak <ubizjak@gmail.com>
1994 * config/i386/i386.md (*pushdi2_rex64 peephole2): Remove.
1995 (*pushdi2_rex64 peephole2): Unconditionally split after
1997 (*ashl<mode>3_doubleword): Ditto.
1998 (*<shift_insn><mode>3_doubleword): Ditto.
2000 2020-02-05 Michael Meissner <meissner@linux.ibm.com>
2003 * config/rs6000/rs6000.c (get_vector_offset): Fix
2005 2020-02-05 Andrew Stubbs <ams@codesourcery.com>
2007 * config/gcn/t-gcn-hsa (MULTILIB_OPTIONS): Use / not space.
2009 2020-02-05 David Malcolm <dmalcolm@redhat.com>
2012 (Special Functions for Debugging the Analyzer): Update description
2013 of __analyzer_dump_exploded_nodes.
2015 2020-02-05 Jakub Jelinek <jakub@redhat.com>
2018 * config/i386/i386-features.c (ix86_add_reg_usage_to_vzeroupper): Only
2019 include sets and not clobbers in the vzeroupper pattern.
2020 * config/i386/sse.md (*avx_vzeroupper): Require in insn condition that
2021 the parallel has 17 (64-bit) or 9 (32-bit) elts.
2022 (*avx_vzeroupper_1): New define_insn_and_split.
2025 * recog.c (pass_split_after_reload::gate): For STACK_REGS targets,
2026 don't run when !optimize.
2027 (pass_split_before_regstack::gate): For STACK_REGS targets, run even
2030 2020-02-05 Richard Biener <rguenther@suse.de>
2033 * genmatch.c (dt_node::gen_kids_1): Emit number of argument
2034 checks before matching calls.
2036 2020-02-05 Jakub Jelinek <jakub@redhat.com>
2038 * tree-ssa-alias.c (aliasing_matching_component_refs_p): Fix up
2039 function comment typo.
2042 * omp-simd-clone.c (expand_simd_clones): If simd_clone_mangle or
2043 simd_clone_create failed when i == 0, adjust clone->nargs by
2046 2020-02-05 Martin Liska <mliska@suse.cz>
2049 * doc/invoke.texi: Document that one should
2050 not combine ASLR and -fpch.
2052 2020-02-04 Richard Biener <rguenther@suse.de>
2054 PR tree-optimization/93538
2055 * match.pd (addr EQ/NE ptr): Amend to handle &ptr->x EQ/NE ptr.
2057 2020-02-04 Richard Biener <rguenther@suse.de>
2059 PR tree-optimization/91123
2060 * tree-ssa-sccvn.c (vn_walk_cb_data::finish): New method.
2061 (vn_walk_cb_data::last_vuse): New member.
2062 (vn_walk_cb_data::saved_operands): Likewsie.
2063 (vn_walk_cb_data::~vn_walk_cb_data): Release saved_operands.
2064 (vn_walk_cb_data::push_partial_def): Use finish.
2065 (vn_reference_lookup_2): Update last_vuse and use finish if
2066 we've saved operands.
2067 (vn_reference_lookup_3): Use finish and update calls to
2068 push_partial_defs everywhere. When translating through
2069 memcpy or aggregate copies save off operands and alias-set.
2070 (eliminate_dom_walker::eliminate_stmt): Restore VN_WALKREWRITE
2071 operation for redundant store removal.
2073 2020-02-04 Richard Biener <rguenther@suse.de>
2075 PR tree-optimization/92819
2076 * tree-ssa-forwprop.c (simplify_vector_constructor): Avoid
2077 generating more stmts than before.
2079 2020-02-04 Martin Liska <mliska@suse.cz>
2081 * config/arm/arm.c (arm_gen_far_branch): Move the function
2082 outside of selftests.
2084 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
2086 * config/rs6000/rs6000.c (adjust_vec_address_pcrel): New helper
2087 function to adjust PC-relative vector addresses.
2088 (rs6000_adjust_vec_address): Call adjust_vec_address_pcrel to
2089 handle vectors with PC-relative addresses.
2091 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
2093 * config/rs6000/rs6000.c (reg_to_non_prefixed): Add forward
2095 (hard_reg_and_mode_to_addr_mask): Delete.
2096 (rs6000_adjust_vec_address): If the original vector address
2097 was REG+REG or REG+OFFSET and the element is not zero, do the add
2098 of the elements in the original address before adding the offset
2099 for the vector element. Use address_to_insn_form to validate the
2100 address using the register being loaded, rather than guessing
2101 whether the address is a DS-FORM or DQ-FORM address.
2103 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
2105 * config/rs6000/rs6000.c (get_vector_offset): New helper function
2106 to calculate the offset in memory from the start of a vector of a
2107 particular element. Add code to keep the element number in
2108 bounds if the element number is variable.
2109 (rs6000_adjust_vec_address): Move calculation of offset of the
2110 vector element to get_vector_offset.
2111 (rs6000_split_vec_extract_var): Do not do the initial AND of
2112 element here, move the code to get_vector_offset.
2114 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
2116 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add some
2119 2020-02-03 Segher Boessenkool <segher@kernel.crashing.org>
2121 * config/rs6000/constraints.md: Improve documentation.
2123 2020-02-03 Richard Earnshaw <rearnsha@arm.com>
2126 * config/arm/t-arm: ($(srcdir)/config/arm/arm-tune.md)
2127 ($(srcdir)/config/arm/arm-tables.opt): Use move-if-change.
2129 2020-02-03 Andrew Stubbs <ams@codesourcery.com>
2131 * config.gcc: Remove "carrizo" support.
2132 * config/gcn/gcn-opts.h (processor_type): Likewise.
2133 * config/gcn/gcn.c (gcn_omp_device_kind_arch_isa): Likewise.
2134 * config/gcn/gcn.opt (gpu_type): Likewise.
2135 * config/gcn/t-omp-device: Likewise.
2137 2020-02-03 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2140 * config/arm/arm-protos.h: New function arm_gen_far_branch prototype.
2141 * config/arm/arm.c (arm_gen_far_branch): New function
2143 * config/arm/arm.md: Update b<cond> for Thumb2 range checks.
2145 2020-02-03 Julian Brown <julian@codesourcery.com>
2146 Tobias Burnus <tobias@codesourcery.com>
2148 * doc/invoke.texi: Update mention of OpenACC version to 2.6.
2150 2020-02-03 Jakub Jelinek <jakub@redhat.com>
2153 * config/s390/s390.md (popcounthi2_z196): Fix up expander to emit
2154 valid RTL to sum up the lowest and second lowest bytes of the popcnt
2157 2020-02-02 Vladimir Makarov <vmakarov@redhat.com>
2159 PR rtl-optimization/91333
2160 * ira-color.c (struct allocno_color_data): Add member
2162 (init_allocno_threads): Set the member up.
2163 (bucket_allocno_compare_func): Add compare hard reg
2166 2020-01-31 Sandra Loosemore <sandra@codesourcery.com>
2168 nios2: Support for GOT-relative DW_EH_PE_datarel encoding.
2170 * configure.ac [nios2-*-*]: Check HAVE_AS_NIOS2_GOTOFF_RELOCATION.
2171 * config.in: Regenerated.
2172 * configure: Regenerated.
2173 * config/nios2/nios2.h (ASM_PREFERRED_EH_DATA_FORMAT): Fix handling
2174 for PIC when HAVE_AS_NIOS2_GOTOFF_RELOCATION.
2175 (ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX): New.
2177 2020-02-01 Andrew Burgess <andrew.burgess@embecosm.com>
2179 * configure: Regenerate.
2181 2020-01-31 Vladimir Makarov <vmakarov@redhat.com>
2183 PR rtl-optimization/91333
2184 * ira-color.c (bucket_allocno_compare_func): Move conflict hard
2185 reg preferences comparison up.
2187 2020-01-31 Richard Sandiford <richard.sandiford@arm.com>
2189 * config/aarch64/aarch64.h (TARGET_SVE_BF16): New macro.
2190 * config/aarch64/aarch64-sve-builtins-sve2.h (svcvtnt): Move to
2191 aarch64-sve-builtins-base.h.
2192 * config/aarch64/aarch64-sve-builtins-sve2.cc (svcvtnt): Move to
2193 aarch64-sve-builtins-base.cc.
2194 * config/aarch64/aarch64-sve-builtins-base.h (svbfdot, svbfdot_lane)
2195 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
2197 * config/aarch64/aarch64-sve-builtins-base.cc (svbfdot, svbfdot_lane)
2198 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
2199 (svcvtnt): New functions.
2200 * config/aarch64/aarch64-sve-builtins-base.def (svbfdot, svbfdot_lane)
2201 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
2202 (svcvtnt): New functions.
2203 (svcvt): Add a form that converts f32 to bf16.
2204 * config/aarch64/aarch64-sve-builtins-shapes.h (ternary_bfloat)
2205 (ternary_bfloat_lane, ternary_bfloat_lanex2, ternary_bfloat_opt_n):
2207 * config/aarch64/aarch64-sve-builtins-shapes.cc (parse_element_type):
2208 Treat B as bfloat16_t.
2209 (ternary_bfloat_lane_base): New class.
2210 (ternary_bfloat_def): Likewise.
2211 (ternary_bfloat): New shape.
2212 (ternary_bfloat_lane_def): New class.
2213 (ternary_bfloat_lane): New shape.
2214 (ternary_bfloat_lanex2_def): New class.
2215 (ternary_bfloat_lanex2): New shape.
2216 (ternary_bfloat_opt_n_def): New class.
2217 (ternary_bfloat_opt_n): New shape.
2218 * config/aarch64/aarch64-sve-builtins.cc (TYPES_cvt_bfloat): New macro.
2219 * config/aarch64/aarch64-sve.md (@aarch64_sve_<sve_fp_op>vnx4sf)
2220 (@aarch64_sve_<sve_fp_op>_lanevnx4sf): New patterns.
2221 (@aarch64_sve_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>)
2222 (@cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
2223 (*cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
2224 (@aarch64_sve_cvtnt<VNx8BF_ONLY:mode>): Likewise.
2225 * config/aarch64/aarch64-sve2.md (@aarch64_sve2_cvtnt<mode>): Key
2226 the pattern off the narrow mode instead of the wider one.
2227 * config/aarch64/iterators.md (VNx8BF_ONLY): New mode iterator.
2228 (UNSPEC_BFMLALB, UNSPEC_BFMLALT, UNSPEC_BFMMLA): New unspecs.
2229 (sve_fp_op): Handle them.
2230 (SVE_BFLOAT_TERNARY_LONG): New int itertor.
2231 (SVE_BFLOAT_TERNARY_LONG_LANE): Likewise.
2233 2020-01-31 Richard Sandiford <richard.sandiford@arm.com>
2235 * config/aarch64/arm_sve.h: Include arm_bf16.h.
2236 * config/aarch64/aarch64-modes.def (BF): Move definition before
2237 VECTOR_MODES. Remove separate VECTOR_MODES for V4BF and V8BF.
2238 (SVE_MODES): Handle BF modes.
2239 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
2241 (aarch64_full_sve_mode): Likewise.
2242 * config/aarch64/iterators.md (SVE_STRUCT): Add VNx16BF, VNx24BF
2244 (SVE_FULL, SVE_FULL_HSD, SVE_ALL): Add VNx8BF.
2245 (Vetype, Vesize, Vctype, VEL, Vel, VEL_INT, V128, v128, vwcore)
2246 (V_INT_EQUIV, v_int_equiv, V_FP_EQUIV, v_fp_equiv, vector_count)
2247 (insn_length, VSINGLE, vsingle, VPRED, vpred, VDOUBLE): Handle the
2249 * config/aarch64/aarch64-sve-builtins.h (TYPE_bfloat): New
2251 * config/aarch64/aarch64-sve-builtins.cc (TYPES_all_arith): New macro.
2252 (TYPES_all_data): Add bf16.
2253 (TYPES_reinterpret1, TYPES_reinterpret): Likewise.
2254 (register_tuple_type): Increase buffer size.
2255 * config/aarch64/aarch64-sve-builtins.def (svbfloat16_t): New type.
2256 (bf16): New type suffix.
2257 * config/aarch64/aarch64-sve-builtins-base.def (svabd, svadd, svaddv)
2258 (svcmpeq, svcmpge, svcmpgt, svcmple, svcmplt, svcmpne, svmad, svmax)
2259 (svmaxv, svmin, svminv, svmla, svmls, svmsb, svmul, svsub, svsubr):
2260 Change type from all_data to all_arith.
2261 * config/aarch64/aarch64-sve-builtins-sve2.def (svaddp, svmaxp)
2264 2020-01-31 Dennis Zhang <dennis.zhang@arm.com>
2265 Matthew Malcomson <matthew.malcomson@arm.com>
2266 Richard Sandiford <richard.sandiford@arm.com>
2268 * doc/invoke.texi (f32mm): Document new AArch64 -march= extension.
2269 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Define
2270 __ARM_FEATURE_SVE_MATMUL_INT8, __ARM_FEATURE_SVE_MATMUL_FP32 and
2271 __ARM_FEATURE_SVE_MATMUL_FP64 as appropriate. Don't define
2272 __ARM_FEATURE_MATMUL_FP64.
2273 * config/aarch64/aarch64-option-extensions.def (fp, simd, fp16)
2274 (sve): Add AARCH64_FL_F32MM to the list of extensions that should
2275 be disabled at the same time.
2276 (f32mm): New extension.
2277 * config/aarch64/aarch64.h (AARCH64_FL_F32MM): New macro.
2278 (AARCH64_FL_F64MM): Bump to the next bit up.
2279 (AARCH64_ISA_F32MM, TARGET_SVE_I8MM, TARGET_F32MM, TARGET_SVE_F32MM)
2280 (TARGET_SVE_F64MM): New macros.
2281 * config/aarch64/iterators.md (SVE_MATMULF): New mode iterator.
2282 (UNSPEC_FMMLA, UNSPEC_SMATMUL, UNSPEC_UMATMUL, UNSPEC_USMATMUL)
2283 (UNSPEC_TRN1Q, UNSPEC_TRN2Q, UNSPEC_UZP1Q, UNSPEC_UZP2Q, UNSPEC_ZIP1Q)
2284 (UNSPEC_ZIP2Q): New unspeccs.
2285 (DOTPROD_US_ONLY, PERMUTEQ, MATMUL, FMMLA): New int iterators.
2286 (optab, sur, perm_insn): Handle the new unspecs.
2287 (sve_fp_op): Handle UNSPEC_FMMLA. Resort.
2288 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use
2289 TARGET_SVE_F64MM instead of separate tests.
2290 (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod<vsi2qi>): New pattern.
2291 (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod_lane<vsi2qi>): Likewise.
2292 (@aarch64_sve_add_<MATMUL:optab><vsi2qi>): Likewise.
2293 (@aarch64_sve_<FMMLA:sve_fp_op><mode>): Likewise.
2294 (@aarch64_sve_<PERMUTEQ:optab><mode>): Likewise.
2295 * config/aarch64/aarch64-sve-builtins.cc (TYPES_s_float): New macro.
2296 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): Use it.
2297 (TYPES_s_signed): New macro.
2298 (TYPES_s_integer): Use it.
2299 (TYPES_d_float): New macro.
2300 (TYPES_d_data): Use it.
2301 * config/aarch64/aarch64-sve-builtins-shapes.h (mmla): Declare.
2302 (ternary_intq_uintq_lane, ternary_intq_uintq_opt_n, ternary_uintq_intq)
2303 (ternary_uintq_intq_lane, ternary_uintq_intq_opt_n): Likewise.
2304 * config/aarch64/aarch64-sve-builtins-shapes.cc (mmla_def): New class.
2305 (svmmla): New shape.
2306 (ternary_resize2_opt_n_base): Add TYPE_CLASS2 and TYPE_CLASS3
2307 template parameters.
2308 (ternary_resize2_lane_base): Likewise.
2309 (ternary_resize2_base): New class.
2310 (ternary_qq_lane_base): Likewise.
2311 (ternary_intq_uintq_lane_def): Likewise.
2312 (ternary_intq_uintq_lane): New shape.
2313 (ternary_intq_uintq_opt_n_def): New class
2314 (ternary_intq_uintq_opt_n): New shape.
2315 (ternary_qq_lane_def): Inherit from ternary_qq_lane_base.
2316 (ternary_uintq_intq_def): New class.
2317 (ternary_uintq_intq): New shape.
2318 (ternary_uintq_intq_lane_def): New class.
2319 (ternary_uintq_intq_lane): New shape.
2320 (ternary_uintq_intq_opt_n_def): New class.
2321 (ternary_uintq_intq_opt_n): New shape.
2322 * config/aarch64/aarch64-sve-builtins-base.h (svmmla, svsudot)
2323 (svsudot_lane, svtrn1q, svtrn2q, svusdot, svusdot_lane, svusmmla)
2324 (svuzp1q, svuzp2q, svzip1q, svzip2q): Declare.
2325 * config/aarch64/aarch64-sve-builtins-base.cc (svdot_lane_impl):
2327 (svdotprod_lane_impl): ...this new class.
2328 (svmmla_impl, svusdot_impl): New classes.
2329 (svdot_lane): Update to use svdotprod_lane_impl.
2330 (svmmla, svsudot, svsudot_lane, svtrn1q, svtrn2q, svusdot)
2331 (svusdot_lane, svusmmla, svuzp1q, svuzp2q, svzip1q, svzip2q): New
2333 * config/aarch64/aarch64-sve-builtins-base.def (svmmla): New base
2334 function, with no types defined.
2335 (svmmla, svusmmla, svsudot, svsudot_lane, svusdot, svusdot_lane): New
2336 AARCH64_FL_I8MM functions.
2337 (svmmla): New AARCH64_FL_F32MM function.
2338 (svld1ro): Depend only on AARCH64_FL_F64MM, not on AARCH64_FL_V8_6.
2339 (svmmla, svtrn1q, svtrn2q, svuz1q, svuz2q, svzip1q, svzip2q): New
2340 AARCH64_FL_F64MM function.
2341 (REQUIRED_EXTENSIONS):
2343 2020-01-31 Andrew Stubbs <ams@codesourcery.com>
2345 * config/gcn/gcn-valu.md (addv64di3_exec): Allow one '0' in each
2348 2020-01-31 Uroš Bizjak <ubizjak@gmail.com>
2350 * config/i386/i386.md (*movoi_internal_avx): Do not check for
2351 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL. Remove MODE_V8SF handling.
2352 (*movti_internal): Do not check for
2353 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
2354 (*movtf_internal): Move check for TARGET_SSE2 and size optimization
2355 just after check for TARGET_AVX.
2356 (*movdf_internal): Ditto.
2357 * config/i386/mmx.md (*mov<mode>_internal): Do not check for
2358 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
2359 * config/i386/sse.md (mov<mode>_internal): Only check
2360 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL with V2DFmode. Move check
2361 for TARGET_SSE2 and size optimization just after check for TARGET_AVX.
2362 (<sse>_andnot<mode>3<mask_name>): Move check for
2363 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL after check for TARGET_AVX.
2364 (<code><mode>3<mask_name>): Ditto.
2365 (*andnot<mode>3): Ditto.
2366 (*andnottf3): Ditto.
2367 (*<code><mode>3): Ditto.
2368 (*<code>tf3): Ditto.
2369 (*andnot<VI:mode>3): Remove
2370 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL handling.
2371 (<mask_codefor><code><VI48_AVX_AVX512F:mode>3<mask_name>): Ditto.
2372 (*<code><VI12_AVX_AVX512F:mode>3): Ditto.
2373 (sse4_1_blendv<ssemodesuffix>): Ditto.
2374 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL):
2375 Explain that tune applies to 128bit instructions only.
2377 2020-01-31 Kwok Cheung Yeung <kcy@codesourcery.com>
2379 * config/gcn/mkoffload.c (process_asm): Add sgpr_count and vgpr_count
2380 to definition of hsa_kernel_description. Parse assembly to find SGPR
2381 and VGPR count of kernel and store in hsa_kernel_description.
2383 2020-01-31 Tamar Christina <tamar.christina@arm.com>
2385 PR rtl-optimization/91838
2386 * simplify-rtx.c (simplify_binary_operation_1): Update LSHIFTRT case
2387 to truncate if allowed or reject combination.
2389 2020-01-31 Andrew Stubbs <ams@codesourcery.com>
2391 * tree-ssa-loop-ivopts.c (get_iv): Use sizetype for zero-step.
2392 (find_inv_vars_cb): Likewise.
2394 2020-01-31 David Malcolm <dmalcolm@redhat.com>
2396 * calls.c (special_function_p): Split out the check for DECL_NAME
2397 being non-NULL and fndecl being extern at file scope into a
2398 new maybe_special_function_p and call it. Drop check for fndecl
2399 being non-NULL that was after a usage of DECL_NAME (fndecl).
2400 * tree.h (maybe_special_function_p): New inline function.
2402 2020-01-30 Andrew Stubbs <ams@codesourcery.com>
2404 * config/gcn/gcn-valu.md (gather<mode>_exec): Move contents ...
2405 (mask_gather_load<mode>): ... here, and zero-initialize the
2407 (maskload<mode>di): Zero-initialize the destination.
2410 2020-01-30 David Malcolm <dmalcolm@redhat.com>
2413 * doc/analyzer.texi (Limitations): Note that constraints on
2414 floating-point values are currently ignored.
2416 2020-01-30 Jakub Jelinek <jakub@redhat.com>
2419 * symtab.c (symtab_node::noninterposable_alias): If localalias
2420 already exists, but is not usable, append numbers after it until
2421 a unique name is found. Formatting fix.
2424 * combine.c (simplify_comparison) <case ROTATE>: Punt on out of range
2427 2020-01-30 Andrew Stubbs <ams@codesourcery.com>
2429 * config/gcn/gcn.c (print_operand): Handle LTGT.
2430 * config/gcn/predicates.md (gcn_fp_compare_operator): Allow ltgt.
2432 2020-01-30 Richard Biener <rguenther@suse.de>
2434 * tree-pretty-print.c (dump_generic_node): Wrap VECTOR_CST
2435 and CONSTRUCTOR in _Literal (type) with TDF_GIMPLE.
2437 2020-01-30 John David Anglin <danglin@gcc.gnu.org>
2439 * config/pa/pa.c (pa_elf_select_rtx_section): Place function pointers
2440 without a DECL in .data.rel.ro.local.
2442 2020-01-30 Jakub Jelinek <jakub@redhat.com>
2445 * config/arm/arm.md (uaddvdi4): Actually emit what gen_uaddvsi4
2449 * config/i386/sse.md
2450 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext): Renamed to ...
2451 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext): ... this. Use
2452 any_extend code iterator instead of always zero_extend.
2453 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_lt): Renamed to ...
2454 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_lt): ... this.
2455 Use any_extend code iterator instead of always zero_extend.
2456 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_shift): Renamed to ...
2457 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_shift): ... this.
2458 Use any_extend code iterator instead of always zero_extend.
2459 (*sse2_pmovmskb_ext): New define_insn.
2460 (*sse2_pmovmskb_ext_lt): New define_insn_and_split.
2463 * config/i386/i386.md (*popcountsi2_zext): New define_insn_and_split.
2464 (*popcountsi2_zext_falsedep): New define_insn.
2466 2020-01-30 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
2468 * config.in: Regenerated.
2469 * configure: Regenerated.
2471 2020-01-29 Tobias Burnus <tobias@codesourcery.com>
2474 * config/gcn/gcn-hsa.h (ASM_SPEC): Add -mattr=-code-object-v3 as
2475 LLVM's assembler changed the default in version 9.
2477 2020-01-24 Jeff Law <law@redhat.com>
2479 PR tree-optimization/89689
2480 * builtins.def (BUILT_IN_OBJECT_SIZE): Make it const rather than pure.
2482 2020-01-29 Richard Sandiford <richard.sandiford@arm.com>
2486 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
2488 PR rtl-optimization/87763
2489 * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
2490 simplification to handle subregs as well as bare regs.
2491 * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
2493 2020-01-29 Joel Hutton <Joel.Hutton@arm.com>
2496 * ira.c (ira): Revert use of simplified LRA algorithm.
2498 2020-01-29 Martin Jambor <mjambor@suse.cz>
2500 PR tree-optimization/92706
2501 * tree-sra.c (struct access): Fields first_link, last_link,
2502 next_queued and grp_queued renamed to first_rhs_link, last_rhs_link,
2503 next_rhs_queued and grp_rhs_queued respectively, new fields
2504 first_lhs_link, last_lhs_link, next_lhs_queued and grp_lhs_queued.
2505 (struct assign_link): Field next renamed to next_rhs, new field
2506 next_lhs. Updated comment.
2507 (work_queue_head): Renamed to rhs_work_queue_head.
2508 (lhs_work_queue_head): New variable.
2509 (add_link_to_lhs): New function.
2510 (relink_to_new_repr): Also relink LHS lists.
2511 (add_access_to_work_queue): Renamed to add_access_to_rhs_work_queue.
2512 (add_access_to_lhs_work_queue): New function.
2513 (pop_access_from_work_queue): Renamed to
2514 pop_access_from_rhs_work_queue.
2515 (pop_access_from_lhs_work_queue): New function.
2516 (build_accesses_from_assign): Also add links to LHS lists and to LHS
2518 (child_would_conflict_in_lacc): Renamed to
2519 child_would_conflict_in_acc. Adjusted parameter names.
2520 (create_artificial_child_access): New parameter set_grp_read, use it.
2521 (subtree_mark_written_and_enqueue): Renamed to
2522 subtree_mark_written_and_rhs_enqueue.
2523 (propagate_subaccesses_across_link): Renamed to
2524 propagate_subaccesses_from_rhs.
2525 (propagate_subaccesses_from_lhs): New function.
2526 (propagate_all_subaccesses): Also propagate subaccesses from LHSs to
2529 2020-01-29 Martin Jambor <mjambor@suse.cz>
2531 PR tree-optimization/92706
2532 * tree-sra.c (struct access): Adjust comment of
2533 grp_total_scalarization.
2534 (find_access_in_subtree): Look for single children spanning an entire
2536 (scalarizable_type_p): Allow register accesses, adjust callers.
2537 (completely_scalarize): Remove function.
2538 (scalarize_elem): Likewise.
2539 (create_total_scalarization_access): Likewise.
2540 (sort_and_splice_var_accesses): Do not track total scalarization
2542 (analyze_access_subtree): New parameter totally, adjust to new meaning
2543 of grp_total_scalarization.
2544 (analyze_access_trees): Pass new parameter to analyze_access_subtree.
2545 (can_totally_scalarize_forest_p): New function.
2546 (create_total_scalarization_access): Likewise.
2547 (create_total_access_and_reshape): Likewise.
2548 (total_should_skip_creating_access): Likewise.
2549 (totally_scalarize_subtree): Likewise.
2550 (analyze_all_variable_accesses): Perform total scalarization after
2551 subaccess propagation using the new functions above.
2552 (initialize_constant_pool_replacements): Output initializers by
2553 traversing the access tree.
2555 2020-01-29 Martin Jambor <mjambor@suse.cz>
2557 * tree-sra.c (verify_sra_access_forest): New function.
2558 (verify_all_sra_access_forests): Likewise.
2559 (create_artificial_child_access): Set parent.
2560 (analyze_all_variable_accesses): Call the verifier.
2562 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
2564 * cgraph.c (cgraph_edge::resolve_speculation): Only lookup direct edge
2565 if called on indirect edge.
2566 (cgraph_edge::redirect_call_stmt_to_callee): Lookup indirect edge of
2567 speculative call if needed.
2569 2020-01-29 Richard Biener <rguenther@suse.de>
2571 PR tree-optimization/93428
2572 * tree-vect-slp.c (vect_build_slp_tree_2): Compute the load
2573 permutation when the load node is created.
2574 (vect_analyze_slp_instance): Re-use it here.
2576 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
2578 * ipa-prop.c (update_indirect_edges_after_inlining): Fix warning.
2580 2020-01-28 Vladimir Makarov <vmakarov@redhat.com>
2582 PR rtl-optimization/93272
2583 * ira-lives.c (process_out_of_region_eh_regs): New function.
2584 (process_bb_node_lives): Call it.
2586 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
2588 * coverage.c (read_counts_file): Make error message lowercase.
2590 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
2592 * profile-count.c (profile_quality_display_names): Fix ordering.
2594 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
2597 * cgraph.c (cgraph_add_edge_to_call_site_hash): Update call site
2598 hash only when edge is first within the sequence.
2599 (cgraph_edge::set_call_stmt): Update handling of speculative calls.
2600 (symbol_table::create_edge): Do not set target_prob.
2601 (cgraph_edge::remove_caller): Watch for speculative calls when updating
2603 (cgraph_edge::make_speculative): Drop target_prob parameter.
2604 (cgraph_edge::speculative_call_info): Remove.
2605 (cgraph_edge::first_speculative_call_target): New member function.
2606 (update_call_stmt_hash_for_removing_direct_edge): New function.
2607 (cgraph_edge::resolve_speculation): Rewrite to new API.
2608 (cgraph_edge::speculative_call_for_target): New member function.
2609 (cgraph_edge::make_direct): Rewrite to new API; fix handling of
2610 multiple speculation targets.
2611 (cgraph_edge::redirect_call_stmt_to_callee): Likewise; fix updating
2613 (verify_speculative_call): Verify that targets form an interval.
2614 * cgraph.h (cgraph_edge::speculative_call_info): Remove.
2615 (cgraph_edge::first_speculative_call_target): New member function.
2616 (cgraph_edge::next_speculative_call_target): New member function.
2617 (cgraph_edge::speculative_call_target_ref): New member function.
2618 (cgraph_edge;:speculative_call_indirect_edge): New member funtion.
2619 (cgraph_edge): Remove target_prob.
2620 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
2621 Fix handling of speculative calls.
2622 * ipa-devirt.c (ipa_devirt): Fix handling of speculative cals.
2623 * ipa-fnsummary.c (analyze_function_body): Likewise.
2624 * ipa-inline.c (speculation_useful_p): Use new speculative call API.
2625 * ipa-profile.c (dump_histogram): Fix formating.
2626 (ipa_profile_generate_summary): Watch for overflows.
2627 (ipa_profile): Do not require probablity to be 1/2; update to new API.
2628 * ipa-prop.c (ipa_make_edge_direct_to_target): Update to new API.
2629 (update_indirect_edges_after_inlining): Update to new API.
2630 * ipa-utils.c (ipa_merge_profiles): Rewrite merging of speculative call
2632 * profile-count.h: (profile_probability::adjusted): New.
2633 * tree-inline.c (copy_bb): Update to new speculative call API; fix
2634 updating of profile.
2635 * value-prof.c (gimple_ic_transform): Rename to ...
2636 (dump_ic_profile): ... this one; update dumping.
2637 (stream_in_histogram_value): Fix formating.
2638 (gimple_value_profile_transformations): Update.
2640 2020-01-28 H.J. Lu <hongjiu.lu@intel.com>
2643 * config/i386/i386.md (*movoi_internal_avx): Remove
2644 TARGET_SSE_TYPELESS_STORES check.
2645 (*movti_internal): Prefer TARGET_AVX over
2646 TARGET_SSE_TYPELESS_STORES.
2647 (*movtf_internal): Likewise.
2648 * config/i386/sse.md (mov<mode>_internal): Prefer TARGET_AVX over
2649 TARGET_SSE_TYPELESS_STORES. Remove "<MODE_SIZE> == 16" check
2650 from TARGET_SSE_TYPELESS_STORES.
2652 2020-01-28 David Malcolm <dmalcolm@redhat.com>
2654 * diagnostic-core.h (warning_at): Rename overload to...
2655 (warning_meta): ...this.
2656 (emit_diagnostic_valist): Delete decl of overload taking
2657 diagnostic_metadata.
2658 * diagnostic.c (emit_diagnostic_valist): Likewise for defn.
2659 (warning_at): Rename overload taking diagnostic_metadata to...
2660 (warning_meta): ...this.
2662 2020-01-28 Richard Biener <rguenther@suse.de>
2664 PR tree-optimization/93439
2665 * tree-parloops.c (create_loop_fn): Move clique bookkeeping...
2666 * tree-cfg.c (move_sese_region_to_fn): ... here.
2667 (verify_types_in_gimple_reference): Verify used cliques are
2670 2020-01-28 H.J. Lu <hongjiu.lu@intel.com>
2673 * config/i386/i386-options.c (set_ix86_tune_features): Add an
2674 argument of a pointer to struct gcc_options and pass it to
2675 parse_mtune_ctrl_str.
2676 (ix86_function_specific_restore): Pass opts to
2677 set_ix86_tune_features.
2678 (ix86_option_override_internal): Likewise.
2679 (parse_mtune_ctrl_str): Add an argument of a pointer to struct
2680 gcc_options and use it for x_ix86_tune_ctrl_string.
2682 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
2684 PR rtl-optimization/87763
2685 * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
2686 simplification to handle subregs as well as bare regs.
2687 * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
2689 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
2691 * tree-vect-loop.c (vectorizable_reduction): Fail gracefully
2692 for reduction chains that (now) include a call.
2694 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
2696 PR tree-optimization/92822
2697 * tree-ssa-forwprop.c (simplify_vector_constructor): When filling
2698 out the don't-care elements of a vector whose significant elements
2699 are duplicates, make the don't-care elements duplicates too.
2701 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
2703 PR tree-optimization/93434
2704 * tree-predcom.c (split_data_refs_to_components): Record which
2705 components have had aliasing loads removed. Prevent store-store
2706 commoning for all such components.
2708 2020-01-28 Jakub Jelinek <jakub@redhat.com>
2711 * config/i386/i386.c (ix86_fold_builtin) <do_shift>: If mask is not
2712 -1 or is_vshift is true, use new_vector with number of elts npatterns
2713 rather than new_unary_operation.
2715 PR tree-optimization/93454
2716 * gimple-fold.c (fold_array_ctor_reference): Perform
2717 elt_size.to_uhwi () just once, instead of calling it in every
2718 iteration. Punt if that value is above size of the temporary
2719 buffer. Decrease third native_encode_expr argument when
2720 bufoff + elt_sz is above size of buf.
2722 2020-01-27 Joseph Myers <joseph@codesourcery.com>
2724 * config/mips/mips.c (mips_declare_object_name)
2725 [USE_GNU_UNIQUE_OBJECT]: Support use of gnu_unique_object.
2727 2020-01-27 Martin Liska <mliska@suse.cz>
2729 PR gcov-profile/93403
2730 * tree-profile.c (gimple_init_gcov_profiler): Generate
2731 both __gcov_indirect_call_profiler_v4 and
2732 __gcov_indirect_call_profiler_v4_atomic.
2734 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
2737 * config/aarch64/aarch64-simd.md (aarch64_get_half<mode>): New
2739 (@aarch64_split_simd_mov<mode>): Use it.
2740 (aarch64_simd_mov_from_<mode>low): Add a GPR alternative.
2741 Leave the vec_extract patterns to handle 2-element vectors.
2742 (aarch64_simd_mov_from_<mode>high): Likewise.
2743 (vec_extract<VQMOV_NO2E:mode><Vhalf>): New expander.
2744 (vec_extractv2dfv1df): Likewise.
2746 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
2748 * config/aarch64/aarch64.c (aarch64_if_then_else_costs): Match
2749 jump conditions for *compare_condjump<GPI:mode>.
2751 2020-01-27 David Malcolm <dmalcolm@redhat.com>
2754 * digraph.cc (test_edge::test_edge): Specify template for base
2757 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
2759 * config/arc/arc.c (arc_rtx_costs): Update mul64 cost.
2761 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
2763 * config/arc/arc-protos.h (gen_mlo): Remove.
2764 (gen_mhi): Likewise.
2765 * config/arc/arc.c (AUX_MULHI): Define.
2766 (arc_must_save_reister): Special handling for r58/59.
2767 (arc_compute_frame_size): Consider mlo/mhi registers.
2768 (arc_save_callee_saves): Emit fp/sp move only when emit_move
2770 (arc_conditional_register_usage): Remove TARGET_BIG_ENDIAN from
2771 mlo/mhi name selection.
2772 (arc_restore_callee_saves): Don't early restore blink when ISR.
2773 (arc_expand_prologue): Add mlo/mhi saving.
2774 (arc_expand_epilogue): Add mlo/mhi restoring.
2777 * config/arc/arc.h (DBX_REGISTER_NUMBER): Correct register
2778 numbering when MUL64 option is used.
2779 (DWARF2_FRAME_REG_OUT): Define.
2780 * config/arc/arc.md (arc600_stall): New pattern.
2781 (VUNSPEC_ARC_ARC600_STALL): Define.
2782 (mulsi64): Use correct mlo/mhi registers.
2783 (mulsi_600): Clean it up.
2784 * config/arc/predicates.md (mlo_operand): Remove any dependency on
2786 (mhi_operand): Likewise.
2788 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
2789 Petro Karashchenko <petro.karashchenko@ring.com>
2791 * config/arc/arc.c (arc_is_uncached_mem_p): Check struct
2792 attributes if needed.
2793 (prepare_move_operands): Generate special unspec instruction for
2795 (arc_isuncached_mem_p): Propagate uncached attribute to each
2797 * config/arc/arc.md (VUNSPEC_ARC_LDDI): Define.
2798 (VUNSPEC_ARC_STDI): Likewise.
2799 (ALLI): New mode iterator.
2800 (mALLI): New mode attribute.
2801 (lddi): New instruction pattern.
2803 (stdidi_split): Split instruction for architectures which are not
2804 supporting ll64 option.
2805 (lddidi_split): Likewise.
2807 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
2809 PR rtl-optimization/92989
2810 * lra-lives.c (process_bb_lives): Update the live-in set before
2811 processing additional clobbers.
2813 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
2815 PR rtl-optimization/93170
2816 * cselib.c (cselib_invalidate_regno_val): New function, split out
2818 (cselib_invalidate_regno): ...here.
2819 (cselib_invalidated_by_call_p): New function.
2820 (cselib_process_insn): Iterate over all the hard-register entries in
2821 REG_VALUES and invalidate any that cross call-clobbered registers.
2823 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
2825 * dojump.c (split_comparison): Use HONOR_NANS rather than
2826 HONOR_SNANS when splitting LTGT.
2828 2020-01-27 Martin Liska <mliska@suse.cz>
2831 * opts.c (print_filtered_help): Exclude language-specific
2832 options from --help=common unless enabled in all FEs.
2834 2020-01-27 Martin Liska <mliska@suse.cz>
2836 * opts.c (print_help): Exclude params from
2837 all except --help=param.
2839 2020-01-27 Martin Liska <mliska@suse.cz>
2842 * config/i386/i386-features.c (make_resolver_func):
2843 Align the code with ppc64 target implementation.
2844 Do not generate a unique name for resolver function.
2846 2020-01-27 Richard Biener <rguenther@suse.de>
2848 PR tree-optimization/93397
2849 * tree-vect-slp.c (vect_analyze_slp_instance): Delay
2850 converted reduction chain SLP graph adjustment.
2852 2020-01-26 Marek Polacek <polacek@redhat.com>
2855 * sanopt.c (sanitize_rewrite_addressable_params): Avoid crash on
2858 2020-01-26 Jason Merrill <jason@redhat.com>
2861 * tree.c (verify_type_variant): Only verify TYPE_NEEDS_CONSTRUCTING
2864 2020-01-26 Darius Galis <darius.galis@cyberthorstudios.com>
2866 * config/rx/rx.md (setmemsi): Added rx_allow_string_insns constraint
2867 (rx_setmem): Likewise.
2869 2020-01-26 Jakub Jelinek <jakub@redhat.com>
2872 * config/i386/i386.md (*addv<dwi>4_doubleword, *subv<dwi>4_doubleword):
2873 Use nonimmediate_operand instead of x86_64_hilo_general_operand and
2874 drop <di> from constraint of last operand.
2877 * config/i386/sse.md (*avx_vperm_broadcast_<mode>): Disallow for
2878 TARGET_AVX2 and V4DFmode not in the split condition, but in the
2879 pattern condition, though allow { 0, 0, 0, 0 } broadcast always.
2881 2020-01-25 Feng Xue <fxue@os.amperecomputing.com>
2884 * ipa-cp.c (get_info_about_necessary_edges): Remove value
2887 2020-01-24 Jeff Law <law@redhat.com>
2889 PR tree-optimization/92788
2890 * tree-ssa-threadedge.c (thread_across_edge): Check EDGE_COMPLEX
2893 2020-01-24 Jakub Jelinek <jakub@redhat.com>
2896 * config/i386/sse.md (*avx_vperm_broadcast_v4sf,
2897 *avx_vperm_broadcast_<mode>,
2898 <sse2_avx_avx512f>_vpermil<mode><mask_name>,
2899 *<sse2_avx_avx512f>_vpermilp<mode><mask_name>):
2900 Move before avx2_perm<mode>/avx512f_perm<mode>.
2903 * simplify-rtx.c (simplify_const_unary_operation,
2904 simplify_const_binary_operation): Punt for mode precision above
2905 MAX_BITSIZE_MODE_ANY_INT.
2907 2020-01-24 Andrew Pinski <apinski@marvell.com>
2909 * config/arm/aarch-cost-tables.h (cortexa57_extra_costs): Change
2912 2020-01-24 Jeff Law <law@redhat.com>
2915 * config/h8300/h8300.c (h8300_print_operand): Only call byte_reg
2916 for REGs. Call output_operand_lossage to get more reasonable
2919 2020-01-24 Andrew Stubbs <ams@codesourcery.com>
2921 * config/gcn/gcn-valu.md (vec_cmp<mode>di): Use
2922 gcn_fp_compare_operator.
2923 (vec_cmpu<mode>di): Use gcn_compare_operator.
2924 (vec_cmp<u>v64qidi): Use gcn_compare_operator.
2925 (vec_cmp<mode>di_exec): Use gcn_fp_compare_operator.
2926 (vec_cmpu<mode>di_exec): Use gcn_compare_operator.
2927 (vec_cmp<u>v64qidi_exec): Use gcn_compare_operator.
2928 (vec_cmp<mode>di_dup): Use gcn_fp_compare_operator.
2929 (vec_cmp<mode>di_dup_exec): Use gcn_fp_compare_operator.
2930 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): Use
2931 gcn_fp_compare_operator.
2932 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): Use
2933 gcn_fp_compare_operator.
2934 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): Use
2935 gcn_fp_compare_operator.
2936 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): Use
2937 gcn_fp_compare_operator.
2939 2020-01-24 Maciej W. Rozycki <macro@wdc.com>
2941 * doc/install.texi (Cross-Compiler-Specific Options): Document
2942 `--with-toolexeclibdir' option.
2944 2020-01-24 Hans-Peter Nilsson <hp@axis.com>
2946 * target.def (flags_regnum): Also mention effect on delay slot filling.
2947 * doc/tm.texi: Regenerate.
2949 2020-01-23 Jeff Law <law@redhat.com>
2951 PR translation/90162
2952 * config/h8300/h8300.c (h8300_option_override): Fix diagnostic text.
2954 2020-01-23 Mikael Tillenius <mti-1@tillenius.com>
2957 * config/h8300/h8300.h (FUNCTION_PROFILER): Fix emission of
2960 2020-01-23 Jakub Jelinek <jakub@redhat.com>
2962 PR rtl-optimization/93402
2963 * postreload.c (reload_combine_recognize_pattern): Don't try to adjust
2966 2020-01-23 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
2968 * config.in: Regenerated.
2969 * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to 1
2970 for TARGET_LIBC_GNUSTACK.
2971 * configure: Regenerated.
2972 * configure.ac: Define TARGET_LIBC_GNUSTACK if glibc version is
2973 found to be 2.31 or greater.
2975 2020-01-23 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
2977 * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to
2979 * config/mips/mips.c (TARGET_ASM_FILE_END): Define to ...
2980 (mips_asm_file_end): New function. Delegate to
2981 file_end_indicate_exec_stack if NEED_INDICATE_EXEC_STACK is true.
2982 * config/mips/mips.h (NEED_INDICATE_EXEC_STACK): Define to 0.
2984 2020-01-23 Jakub Jelinek <jakub@redhat.com>
2987 * config/i386/i386-modes.def (POImode): New mode.
2988 (MAX_BITSIZE_MODE_ANY_INT): Change from 128 to 160.
2989 * config/i386/i386.md (DPWI): New mode attribute.
2990 (addv<mode>4, subv<mode>4): Use <DPWI> instead of <DWI>.
2992 (QPWI): ... this. Use POI instead of OI for TImode.
2993 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1,
2994 *subv<dwi>4_doubleword, *subv<dwi>4_doubleword_1): Use <QPWI>
2997 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
3000 * config/aarch64/aarch64.md (UNSPEC_SPECULATION_TRACKER_REV): New
3002 (speculation_tracker_rev): New pattern.
3003 * config/aarch64/aarch64-speculation.cc (aarch64_do_track_speculation):
3004 Use speculation_tracker_rev to track the inverse condition.
3006 2020-01-23 Richard Biener <rguenther@suse.de>
3008 PR tree-optimization/93381
3009 * tree-ssa-sccvn.c (vn_walk_cb_data::push_partial_def): Take
3010 alias-set of the def as argument and record the first one.
3011 (vn_walk_cb_data::first_set): New member.
3012 (vn_reference_lookup_3): Pass the alias-set of the current def
3013 to push_partial_def. Fix alias-set used in the aggregate copy
3015 (vn_reference_lookup): Consistently set *last_vuse_ptr.
3016 * real.c (clear_significand_below): Fix out-of-bound access.
3018 2020-01-23 Jakub Jelinek <jakub@redhat.com>
3021 * config/i386/i386.md (*bmi2_bzhi_<mode>3_2, *bmi2_bzhi_<mode>3_3):
3022 New define_insn patterns.
3024 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
3026 * doc/sourcebuild.texi (check-function-bodies): Add an
3027 optional target/xfail selector.
3029 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
3031 PR rtl-optimization/93124
3032 * auto-inc-dec.c (merge_in_block): Don't add auto inc/decs to
3033 bare USE and CLOBBER insns.
3035 2020-01-22 Andrew Pinski <apinski@marvell.com>
3037 * config/arc/arc.c (output_short_suffix): Check insn for nullness.
3039 2020-01-22 David Malcolm <dmalcolm@redhat.com>
3042 * gdbinit.in (break-on-saved-diagnostic): Update for move of
3043 diagnostic_manager into "ana" namespace.
3044 * selftest-run-tests.c (selftest::run_tests): Update for move of
3045 selftest::run_analyzer_selftests to
3046 ana::selftest::run_analyzer_selftests.
3048 2020-01-22 Richard Sandiford <richard.sandiford@arm.com>
3050 * cfgexpand.c (union_stack_vars): Update the size.
3052 2020-01-22 Richard Biener <rguenther@suse.de>
3054 PR tree-optimization/93381
3055 * tree-ssa-structalias.c (find_func_aliases): Assume offsetting
3056 throughout, handle all conversions the same.
3058 2020-01-22 Jakub Jelinek <jakub@redhat.com>
3061 * config/aarch64/aarch64.c (aarch64_expand_subvti): Only use
3062 gen_subdi3_compare1_imm if low_in2 satisfies aarch64_plus_immediate
3063 predicate, not whenever it is CONST_INT. Otherwise, force_reg it.
3064 Call force_reg on high_in2 unconditionally.
3066 2020-01-22 Martin Liska <mliska@suse.cz>
3068 PR tree-optimization/92924
3069 * profile.c (compute_value_histograms): Divide
3072 2020-01-22 Jakub Jelinek <jakub@redhat.com>
3075 * output.h (assemble_name_resolve): Declare.
3076 * varasm.c (assemble_name_resolve): New function.
3077 (assemble_name): Use it.
3078 * config/i386/i386.h (ASM_OUTPUT_SYMBOL_REF): Define.
3080 2020-01-22 Joseph Myers <joseph@codesourcery.com>
3082 * doc/sourcebuild.texi (Texinfo Manuals, Front End): Refer to
3083 update_web_docs_git instead of update_web_docs_svn.
3085 2020-01-21 Andrew Pinski <apinski@marvell.com>
3088 * config/aarch64/aarch64.md (tlsgd_small_<mode>): Have operand 0
3089 as PTR mode. Have operand 1 as being modeless, it can be P mode.
3090 (*tlsgd_small_<mode>): Likewise.
3091 * config/aarch64/aarch64.c (aarch64_load_symref_appropriately)
3092 <case SYMBOL_SMALL_TLSGD>: Call gen_tlsgd_small_* with a ptr_mode
3093 register. Convert that register back to dest using convert_mode.
3095 2020-01-21 Jim Wilson <jimw@sifive.com>
3097 * config/riscv/riscv-sr.c (riscv_sr_match_prologue): Use INTVAL
3100 2020-01-21 H.J. Lu <hongjiu.lu@intel.com>
3101 Uros Bizjak <ubizjak@gmail.com>
3104 * config/i386/i386.c (ix86_tls_module_base): Replace Pmode
3106 (legitimize_tls_address): Do GNU2 TLS address computation in
3107 ptr_mode and zero-extend result to Pmode.
3108 * config/i386/i386.md (@tls_dynamic_gnu2_64_<mode>): Replace
3109 :P with :PTR and Pmode with ptr_mode.
3110 (*tls_dynamic_gnu2_lea_64_<mode>): Likewise.
3111 (*tls_dynamic_gnu2_call_64_<mode>): Likewise.
3112 (*tls_dynamic_gnu2_combine_64_<mode>): Likewise.
3114 2020-01-21 Jakub Jelinek <jakub@redhat.com>
3117 * config/riscv/riscv.c (riscv_rtx_costs) <case ZERO_EXTRACT>: Verify
3118 the last two operands are CONST_INT_P before using them as such.
3120 2020-01-21 Richard Sandiford <richard.sandiford@arm.com>
3122 * config/aarch64/aarch64-sve-builtins.def: Use get_typenode_from_name
3123 to get the integer element types.
3125 2020-01-21 Richard Sandiford <richard.sandiford@arm.com>
3127 * config/aarch64/aarch64-sve-builtins.h
3128 (function_expander::convert_to_pmode): Declare.
3129 * config/aarch64/aarch64-sve-builtins.cc
3130 (function_expander::convert_to_pmode): New function.
3131 (function_expander::get_contiguous_base): Use it.
3132 (function_expander::prepare_gather_address_operands): Likewise.
3133 * config/aarch64/aarch64-sve-builtins-sve2.cc
3134 (svwhilerw_svwhilewr_impl::expand): Likewise.
3136 2020-01-21 Szabolcs Nagy <szabolcs.nagy@arm.com>
3139 * config/aarch64/aarch64.c (aarch64_declare_function_name): Set
3140 cfun->machine->label_is_assembled.
3141 (aarch64_print_patchable_function_entry): New.
3142 (TARGET_ASM_PRINT_PATCHABLE_FUNCTION_ENTRY): Define.
3143 * config/aarch64/aarch64.h (struct machine_function): New field,
3146 2020-01-21 David Malcolm <dmalcolm@redhat.com>
3149 * ipa-profile.c (ipa_profile): Delete call_sums and set it to
3152 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
3155 * cgraph.c (cgraph_edge::resolve_speculation,
3156 cgraph_edge::redirect_call_stmt_to_callee): Fix update of
3157 call_stmt_site_hash.
3159 2020-01-21 Martin Liska <mliska@suse.cz>
3161 * config/rs6000/rs6000.c (common_mode_defined): Remove
3164 2020-01-21 Richard Biener <rguenther@suse.de>
3166 PR tree-optimization/92328
3167 * tree-ssa-sccvn.c (vn_reference_lookup_3): Preserve
3168 type when value-numbering same-sized store by inserting a
3170 (eliminate_dom_walker::eliminate_stmt): When eliminating
3171 a redundant store handle bit-reinterpretation of the same value.
3173 2020-01-21 Andrew Pinski <apinski@marvel.com>
3176 * tree-into-ssa.c (prepare_block_for_update_1): Split out
3178 (prepare_block_for_update): This. Use a worklist instead of
3181 2020-01-21 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3183 * gcc/config/arm/arm.c (clear_operation_p):
3184 Initialise last_regno, skip first iteration
3185 based on the first_set value and use ints instead
3186 of the unnecessary HOST_WIDE_INTs.
3188 2020-01-21 Jakub Jelinek <jakub@redhat.com>
3191 * config/rs6000/rs6000.c (rs6000_emit_cmove): If using fsel, punt for
3192 compare_mode other than SFmode or DFmode.
3194 2020-01-21 Kito Cheng <kito.cheng@sifive.com>
3197 * config/riscv/riscv-protos.h (riscv_hard_regno_rename_ok): New.
3198 * config/riscv/riscv.c (riscv_hard_regno_rename_ok): New.
3199 * config/riscv/riscv.h (HARD_REGNO_RENAME_OK): Defined.
3201 2020-01-20 Wilco Dijkstra <wdijkstr@arm.com>
3203 * config/aarch64/aarch64.c (neoversen1_tunings): Set jump_align to 4.
3205 2020-01-20 Andrew Pinski <apinski@marvell.com>
3208 * targhooks.c (default_print_patchable_function_entry): Use
3209 output_asm_insn to emit the nop instruction.
3211 2020-01-20 Fangrui Song <maskray@google.com>
3214 * targhooks.c (default_print_patchable_function_entry): Align to
3217 2020-01-20 H.J. Lu <hongjiu.lu@intel.com>
3220 * config/i386/i386.c (legitimize_tls_address): Pass Pmode to
3221 gen_tls_dynamic_gnu2_64. Compute GNU2 TLS address in ptr_mode.
3222 * config/i386/i386.md (tls_dynamic_gnu2_64): Renamed to ...
3223 (@tls_dynamic_gnu2_64_<mode>): This. Replace DI with P.
3224 (*tls_dynamic_gnu2_lea_64): Renamed to ...
3225 (*tls_dynamic_gnu2_lea_64_<mode>): This. Replace DI with P.
3226 Remove the {q} suffix from lea.
3227 (*tls_dynamic_gnu2_call_64): Renamed to ...
3228 (*tls_dynamic_gnu2_call_64_<mode>): This. Replace DI with P.
3229 (*tls_dynamic_gnu2_combine_64): Renamed to ...
3230 (*tls_dynamic_gnu2_combine_64_<mode>): This. Replace DI with P.
3231 Pass Pmode to gen_tls_dynamic_gnu2_64.
3233 2020-01-20 Wilco Dijkstra <wdijkstr@arm.com>
3235 * config/aarch64/aarch64.h (SLOW_BYTE_ACCESS): Set to 1.
3237 2020-01-20 Richard Sandiford <richard.sandiford@arm.com>
3239 * config/aarch64/aarch64-sve-builtins-base.cc
3240 (svld1ro_impl::memory_vector_mode): Remove parameter name.
3242 2020-01-20 Richard Biener <rguenther@suse.de>
3245 * dwarf2out.c (prune_unused_types): Unconditionally mark
3246 called function DIEs.
3248 2020-01-20 Martin Liska <mliska@suse.cz>
3250 PR tree-optimization/93199
3251 * tree-eh.c (struct leh_state): Add
3252 new field outer_non_cleanup.
3253 (cleanup_is_dead_in): Pass leh_state instead
3254 of eh_region. Add a checking that state->outer_non_cleanup
3255 points to outer non-clean up region.
3256 (lower_try_finally): Record outer_non_cleanup
3258 (lower_catch): Likewise.
3259 (lower_eh_filter): Likewise.
3260 (lower_eh_must_not_throw): Likewise.
3261 (lower_cleanup): Likewise.
3263 2020-01-20 Richard Biener <rguenther@suse.de>
3265 PR tree-optimization/93094
3266 * tree-vectorizer.h (vect_loop_versioning): Adjust.
3267 (vect_transform_loop): Likewise.
3268 * tree-vectorizer.c (try_vectorize_loop_1): Pass down
3269 loop_vectorized_call to vect_transform_loop.
3270 * tree-vect-loop.c (vect_transform_loop): Pass down
3271 loop_vectorized_call to vect_loop_versioning.
3272 * tree-vect-loop-manip.c (vect_loop_versioning): Use
3273 the earlier discovered loop_vectorized_call.
3275 2020-01-19 Eric S. Raymond <esr@thyrsus.com>
3277 * doc/contribute.texi: Update for SVN -> Git transition.
3278 * doc/install.texi: Likewise.
3280 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
3283 * cgraph.c (cgraph_edge::make_speculative): Increase number of
3284 speculative targets.
3285 (verify_speculative_call): New function
3286 (cgraph_node::verify_node): Use it.
3287 * ipa-profile.c (ipa_profile): Fix formating; do not set number of
3290 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
3293 * cgraph.c (cgraph_edge::resolve_speculation): Fix foramting.
3294 (cgraph_edge::make_direct): Remove all indirect targets.
3295 (cgraph_edge::redirect_call_stmt_to_callee): Use make_direct..
3296 (cgraph_node::verify_node): Verify that only one call_stmt or
3297 lto_stmt_uid is set.
3298 * cgraphclones.c (cgraph_edge::clone): Set only one call_stmt or
3300 * lto-cgraph.c (lto_output_edge): Simplify streaming of stmt.
3301 (lto_output_ref): Simplify streaming of stmt.
3302 * lto-streamer-in.c (fixup_call_stmt_edges_1): Clear lto_stmt_uid.
3304 2020-01-18 Tamar Christina <tamar.christina@arm.com>
3306 * config/aarch64/aarch64-sve-builtins-base.cc (memory_vector_mode):
3307 Mark parameter unused.
3309 2020-01-18 Hans-Peter Nilsson <hp@axis.com>
3311 * config.gcc <obsolete targets>: Add crisv32-*-* and cris-*-linux*
3313 2019-01-18 Gerald Pfeifer <gerald@pfeifer.com>
3315 * varpool.c (ctor_useable_for_folding_p): Fix grammar.
3317 2020-01-18 Iain Sandoe <iain@sandoe.co.uk>
3319 * Makefile.in: Add coroutine-passes.o.
3320 * builtin-types.def (BT_CONST_SIZE): New.
3321 (BT_FN_BOOL_PTR): New.
3322 (BT_FN_PTR_PTR_CONST_SIZE_BOOL): New.
3323 * builtins.def (DEF_COROUTINE_BUILTIN): New.
3324 * coroutine-builtins.def: New file.
3325 * coroutine-passes.cc: New file.
3326 * function.h (struct GTY function): Add a bit to indicate that the
3327 function is a coroutine component.
3328 * internal-fn.c (expand_CO_FRAME): New.
3329 (expand_CO_YIELD): New.
3330 (expand_CO_SUSPN): New.
3331 (expand_CO_ACTOR): New.
3332 * internal-fn.def (CO_ACTOR): New.
3336 * passes.def: Add pass_coroutine_lower_builtins,
3337 pass_coroutine_early_expand_ifns.
3338 * tree-pass.h (make_pass_coroutine_lower_builtins): New.
3339 (make_pass_coroutine_early_expand_ifns): New.
3340 * doc/invoke.texi: Document the fcoroutines command line
3343 2020-01-18 Jakub Jelinek <jakub@redhat.com>
3345 * config/arm/vfp.md (*clear_vfp_multiple): Remove unused variable.
3348 * config/arm/arm.c (clear_operation_p): Don't use REGNO until
3349 after checking the argument is a REG. Don't use REGNO (reg)
3350 again to set last_regno, reuse regno variable instead.
3352 2020-01-17 David Malcolm <dmalcolm@redhat.com>
3354 * doc/analyzer.texi (Limitations): Add note about NaN.
3356 2020-01-17 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3357 Sudakshina Das <sudi.das@arm.com>
3359 * config/arm/arm.md (ashldi3): Generate thumb2_lsll for both reg
3360 and valid immediate.
3361 (ashrdi3): Generate thumb2_asrl for both reg and valid immediate.
3362 (lshrdi3): Generate thumb2_lsrl for valid immediates.
3363 * config/arm/constraints.md (Pg): New.
3364 * config/arm/predicates.md (long_shift_imm): New.
3365 (arm_reg_or_long_shift_imm): Likewise.
3366 * config/arm/thumb2.md (thumb2_asrl): New immediate alternative.
3367 (thumb2_lsll): Likewise.
3370 2020-01-17 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3371 Sudakshina Das <sudi.das@arm.com>
3373 * config/arm/arm.md (ashldi3): Generate thumb2_lsll for TARGET_HAVE_MVE.
3374 (ashrdi3): Generate thumb2_asrl for TARGET_HAVE_MVE.
3375 * config/arm/arm.c (arm_hard_regno_mode_ok): Allocate even odd
3376 register pairs for doubleword quantities for ARMv8.1M-Mainline.
3377 * config/arm/thumb2.md (thumb2_asrl): New.
3378 (thumb2_lsll): Likewise.
3380 2020-01-17 Jakub Jelinek <jakub@redhat.com>
3382 * config/arm/arm.c (cmse_nonsecure_call_inline_register_clear): Remove
3385 2020-01-17 Alexander Monakov <amonakov@ispras.ru>
3387 * gdbinit.in (help-gcc-hooks): New command.
3388 (pp, pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, ptc, pdn, ptn, pdd, prc,
3389 pi, pbm, pel, trt): Take $arg0 instead of $ if supplied. Update
3392 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
3394 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use the
3395 correct target macro.
3397 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
3399 * config/aarch64/aarch64-protos.h
3400 (aarch64_sve_ld1ro_operand_p): New.
3401 * config/aarch64/aarch64-sve-builtins-base.cc
3402 (class load_replicate): New.
3403 (class svld1ro_impl): New.
3404 (class svld1rq_impl): Change to inherit from load_replicate.
3405 (svld1ro): New sve intrinsic function base.
3406 * config/aarch64/aarch64-sve-builtins-base.def (svld1ro):
3407 New DEF_SVE_FUNCTION.
3408 * config/aarch64/aarch64-sve-builtins-base.h
3409 (svld1ro): New decl.
3410 * config/aarch64/aarch64-sve-builtins.cc
3411 (function_expander::add_mem_operand): Modify assert to allow
3413 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): New
3415 * config/aarch64/aarch64.c
3416 (aarch64_sve_ld1rq_operand_p): Implement in terms of ...
3417 (aarch64_sve_ld1rq_ld1ro_operand_p): This.
3418 (aarch64_sve_ld1ro_operand_p): New.
3419 * config/aarch64/aarch64.md (UNSPEC_LD1RO): New unspec.
3420 * config/aarch64/constraints.md (UOb,UOh,UOw,UOd): New.
3421 * config/aarch64/predicates.md
3422 (aarch64_sve_ld1ro_operand_{b,h,w,d}): New.
3424 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
3426 * config/aarch64/aarch64-c.c (_ARM_FEATURE_MATMUL_FLOAT64):
3427 Introduce this ACLE specified predefined macro.
3428 * config/aarch64/aarch64-option-extensions.def (f64mm): New.
3429 (fp): Disabling this disables f64mm.
3430 (simd): Disabling this disables f64mm.
3431 (fp16): Disabling this disables f64mm.
3432 (sve): Disabling this disables f64mm.
3433 * config/aarch64/aarch64.h (AARCH64_FL_F64MM): New.
3434 (AARCH64_ISA_F64MM): New.
3435 (TARGET_F64MM): New.
3436 * doc/invoke.texi (f64mm): Document new option.
3438 2020-01-17 Wilco Dijkstra <wdijkstr@arm.com>
3440 * config/aarch64/aarch64.c (generic_tunings): Add branch fusion.
3441 (neoversen1_tunings): Likewise.
3443 2020-01-17 Wilco Dijkstra <wdijkstr@arm.com>
3446 * config/aarch64/aarch64.c (aarch64_split_compare_and_swap)
3447 Add assert to ensure prolog has been emitted.
3448 (aarch64_split_atomic_op): Likewise.
3449 * config/aarch64/atomics.md (aarch64_compare_and_swap<mode>)
3450 Use epilogue_completed rather than reload_completed.
3451 (aarch64_atomic_exchange<mode>): Likewise.
3452 (aarch64_atomic_<atomic_optab><mode>): Likewise.
3453 (atomic_nand<mode>): Likewise.
3454 (aarch64_atomic_fetch_<atomic_optab><mode>): Likewise.
3455 (atomic_fetch_nand<mode>): Likewise.
3456 (aarch64_atomic_<atomic_optab>_fetch<mode>): Likewise.
3457 (atomic_nand_fetch<mode>): Likewise.
3459 2020-01-17 Richard Sandiford <richard.sandiford@arm.com>
3462 * config/aarch64/aarch64.h (REVERSIBLE_CC_MODE): Return false
3464 (REVERSE_CONDITION): Delete.
3465 * config/aarch64/iterators.md (CC_ONLY): New mode iterator.
3466 (CCFP_CCFPE): Likewise.
3467 (e): New mode attribute.
3468 * config/aarch64/aarch64.md (ccmp<GPI:mode>): Rename to...
3469 (@ccmp<CC_ONLY:mode><GPI:mode>): ...this, using CC_ONLY instead of CC.
3470 (fccmp<GPF:mode>, fccmpe<GPF:mode>): Merge into...
3471 (@ccmp<CCFP_CCFPE:mode><GPF:mode>): ...this combined pattern.
3472 (@ccmp<CC_ONLY:mode><GPI:mode>_rev): New pattern.
3473 (@ccmp<CCFP_CCFPE:mode><GPF:mode>_rev): Likewise.
3474 * config/aarch64/aarch64.c (aarch64_gen_compare_reg): Update
3475 name of generator from gen_ccmpdi to gen_ccmpccdi.
3476 (aarch64_gen_ccmp_next): Use code_for_ccmp. If we want to reverse
3477 the previous comparison but aren't able to, use the new ccmp_rev
3480 2020-01-17 Richard Sandiford <richard.sandiford@arm.com>
3482 * gimplify.c (gimplify_return_expr): Use poly_int_tree_p rather
3483 than testing directly for INTEGER_CST.
3484 (gimplify_target_expr, gimplify_omp_depend): Likewise.
3486 2020-01-17 Jakub Jelinek <jakub@redhat.com>
3488 PR tree-optimization/93292
3489 * tree-vect-stmts.c (vectorizable_comparison): Punt also if
3490 get_vectype_for_scalar_type returns NULL.
3492 2020-01-16 Jan Hubicka <hubicka@ucw.cz>
3494 * params.opt (-param=max-predicted-iterations): Increase range from 0.
3495 * predict.c (estimate_loops): Add 1 to param_max_predicted_iterations.
3497 2020-01-16 Jan Hubicka <hubicka@ucw.cz>
3499 * ipa-fnsummary.c (estimate_calls_size_and_time): Fix formating of
3501 * params.opt: (max-predicted-iterations): Set bounds.
3502 * predict.c (real_almost_one, real_br_prob_base,
3503 real_inv_br_prob_base, real_one_half, real_bb_freq_max): Remove.
3504 (propagate_freq): Add max_cyclic_prob parameter; cap cyclic
3505 probabilities; do not truncate to reg_br_prob_bases.
3506 (estimate_loops_at_level): Pass max_cyclic_prob.
3507 (estimate_loops): Compute max_cyclic_prob.
3508 (estimate_bb_frequencies): Do not initialize real_*; update calculation
3510 * profile-count.c (profile_probability::to_sreal): New.
3511 * profile-count.h (class sreal): Move up in file.
3512 (profile_probability::to_sreal): Declare.
3514 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
3517 (arm_invalid_conversion): New function for target hook.
3518 (arm_invalid_unary_op): New function for target hook.
3519 (arm_invalid_binary_op): New function for target hook.
3521 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
3523 * config.gcc: Add arm_bf16.h.
3524 * config/arm/arm-builtins.c (arm_mangle_builtin_type): Fix comment.
3525 (arm_simd_builtin_std_type): Add BFmode.
3526 (arm_init_simd_builtin_types): Define element types for vector types.
3527 (arm_init_bf16_types): New function.
3528 (arm_init_builtins): Add arm_init_bf16_types function call.
3529 * config/arm/arm-modes.def: Add BFmode and V4BF, V8BF vector modes.
3530 * config/arm/arm-simd-builtin-types.def: Add V4BF, V8BF.
3531 * config/arm/arm.c (aapcs_vfp_sub_candidate): Add BFmode.
3532 (arm_hard_regno_mode_ok): Add BFmode and tidy up statements.
3533 (arm_vector_mode_supported_p): Add V4BF, V8BF.
3534 (arm_mangle_type): Add __bf16.
3535 * config/arm/arm.h: Add V4BF, V8BF to VALID_NEON_DREG_MODE,
3536 VALID_NEON_QREG_MODE respectively. Add export arm_bf16_type_node,
3537 arm_bf16_ptr_type_node.
3538 * config/arm/arm.md: Add BFmode to movhf expand, mov pattern and
3539 define_split between ARM registers.
3540 * config/arm/arm_bf16.h: New file.
3541 * config/arm/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
3542 * config/arm/iterators.md: (ANY64_BF, VDXMOV, VHFBF, HFBF, fporbf): New.
3544 * config/arm/neon.md: Add BF vector types to movhf NEON move patterns.
3545 * config/arm/vfp.md: Add BFmode to movhf patterns.
3547 2020-01-16 Mihail Ionescu <mihail.ionescu@arm.com>
3548 Andre Vieira <andre.simoesdiasvieira@arm.com>
3550 * config/arm/arm-cpus.in (mve, mve_float): New features.
3551 (dsp, mve, mve.fp): New options.
3552 * config/arm/arm.h (TARGET_HAVE_MVE, TARGET_HAVE_MVE_FLOAT): Define.
3553 * config/arm/t-rmprofile: Map v8.1-M multilibs to v8-M.
3554 * doc/invoke.texi: Document the armv8.1-m mve and dps options.
3556 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3557 Thomas Preud'homme <thomas.preudhomme@arm.com>
3559 * config/arm/arm-cpus.in (ARMv8_1m_main): Redefine as an extension to
3561 * config/arm/arm.c (arm_options_perform_arch_sanity_checks): Remove
3562 error for using -mcmse when targeting Armv8.1-M Mainline.
3564 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3565 Thomas Preud'homme <thomas.preudhomme@arm.com>
3567 * config/arm/arm.md (nonsecure_call_internal): Do not force memory
3568 address in r4 when targeting Armv8.1-M Mainline.
3569 (nonsecure_call_value_internal): Likewise.
3570 * config/arm/thumb2.md (nonsecure_call_reg_thumb2): Make memory address
3571 a register match_operand again. Emit BLXNS when targeting
3573 (nonsecure_call_value_reg_thumb2): Likewise.
3575 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3576 Thomas Preud'homme <thomas.preudhomme@arm.com>
3578 * config/arm/arm.c (arm_add_cfa_adjust_cfa_note): Declare early.
3579 (cmse_nonsecure_call_inline_register_clear): Define new lazy_fpclear
3580 variable as true when floating-point ABI is not hard. Replace
3581 check against TARGET_HARD_FLOAT_ABI by checks against lazy_fpclear.
3582 Generate VLSTM and VLLDM instruction respectively before and
3583 after a function call to cmse_nonsecure_call function.
3584 * config/arm/unspecs.md (VUNSPEC_VLSTM): Define unspec.
3585 (VUNSPEC_VLLDM): Likewise.
3586 * config/arm/vfp.md (lazy_store_multiple_insn): New define_insn.
3587 (lazy_load_multiple_insn): Likewise.
3589 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3590 Thomas Preud'homme <thomas.preudhomme@arm.com>
3592 * config/arm/arm.c (vfp_emit_fstmd): Declare early.
3593 (arm_emit_vfp_multi_reg_pop): Likewise.
3594 (cmse_nonsecure_call_inline_register_clear): Abstract number of VFP
3595 registers to clear in max_fp_regno. Emit VPUSH and VPOP to save and
3596 restore callee-saved VFP registers.
3598 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3599 Thomas Preud'homme <thomas.preudhomme@arm.com>
3601 * config/arm/arm.c (arm_emit_multi_reg_pop): Declare early.
3602 (cmse_nonsecure_call_clear_caller_saved): Rename into ...
3603 (cmse_nonsecure_call_inline_register_clear): This. Save and clear
3604 callee-saved GPRs as well as clear ip register before doing a nonsecure
3605 call then restore callee-saved GPRs after it when targeting
3607 (arm_reorg): Adapt to function rename.
3609 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3610 Thomas Preud'homme <thomas.preudhomme@arm.com>
3612 * config/arm/arm-protos.h (clear_operation_p): Adapt prototype.
3613 * config/arm/arm.c (clear_operation_p): Extend to be able to check a
3614 clear_vfp_multiple pattern based on a new vfp parameter.
3615 (cmse_clear_registers): Generate VSCCLRM to clear VFP registers when
3616 targeting Armv8.1-M Mainline.
3617 (cmse_nonsecure_entry_clear_before_return): Clear VFP registers
3618 unconditionally when targeting Armv8.1-M Mainline architecture. Check
3619 whether VFP registers are available before looking call_used_regs for a
3621 * config/arm/predicates.md (clear_multiple_operation): Adapt to change
3622 of prototype of clear_operation_p.
3623 (clear_vfp_multiple_operation): New predicate.
3624 * config/arm/unspecs.md (VUNSPEC_VSCCLRM_VPR): New volatile unspec.
3625 * config/arm/vfp.md (clear_vfp_multiple): New define_insn.
3627 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3628 Thomas Preud'homme <thomas.preudhomme@arm.com>
3630 * config/arm/arm-protos.h (clear_operation_p): Declare.
3631 * config/arm/arm.c (clear_operation_p): New function.
3632 (cmse_clear_registers): Generate clear_multiple instruction pattern if
3633 targeting Armv8.1-M Mainline or successor.
3634 (output_return_instruction): Only output APSR register clearing if
3635 Armv8.1-M Mainline instructions not available.
3636 (thumb_exit): Likewise.
3637 * config/arm/predicates.md (clear_multiple_operation): New predicate.
3638 * config/arm/thumb2.md (clear_apsr): New define_insn.
3639 (clear_multiple): Likewise.
3640 * config/arm/unspecs.md (VUNSPEC_CLRM_APSR): New volatile unspec.
3642 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3643 Thomas Preud'homme <thomas.preudhomme@arm.com>
3645 * config/arm/arm.c (fp_sysreg_names): Declare and define.
3646 (use_return_insn): Also return false for Armv8.1-M Mainline.
3647 (output_return_instruction): Skip FPSCR clearing if Armv8.1-M
3648 Mainline instructions are available.
3649 (arm_compute_frame_layout): Allocate space in frame for FPCXTNS
3650 when targeting Armv8.1-M Mainline Security Extensions.
3651 (arm_expand_prologue): Save FPCXTNS if this is an Armv8.1-M
3652 Mainline entry function.
3653 (cmse_nonsecure_entry_clear_before_return): Clear IP and r4 if
3654 targeting Armv8.1-M Mainline or successor.
3655 (arm_expand_epilogue): Fix indentation of caller-saved register
3656 clearing. Restore FPCXTNS if this is an Armv8.1-M Mainline
3658 * config/arm/arm.h (TARGET_HAVE_FP_CMSE): New macro.
3659 (FP_SYSREGS): Likewise.
3660 (enum vfp_sysregs_encoding): Define enum.
3661 (fp_sysreg_names): Declare.
3662 * config/arm/unspecs.md (VUNSPEC_VSTR_VLDR): New volatile unspec.
3663 * config/arm/vfp.md (push_fpsysreg_insn): New define_insn.
3664 (pop_fpsysreg_insn): Likewise.
3666 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3667 Thomas Preud'homme <thomas.preudhomme@arm.com>
3669 * config/arm/arm-cpus.in (armv8_1m_main): New feature.
3670 (ARMv4, ARMv4t, ARMv5t, ARMv5te, ARMv5tej, ARMv6, ARMv6j, ARMv6k,
3671 ARMv6z, ARMv6kz, ARMv6zk, ARMv6t2, ARMv6m, ARMv7, ARMv7a, ARMv7ve,
3672 ARMv7r, ARMv7m, ARMv7em, ARMv8a, ARMv8_1a, ARMv8_2a, ARMv8_3a,
3673 ARMv8_4a, ARMv8_5a, ARMv8m_base, ARMv8m_main, ARMv8r): Reindent.
3674 (ARMv8_1m_main): New feature group.
3675 (armv8.1-m.main): New architecture.
3676 * config/arm/arm-tables.opt: Regenerate.
3677 * config/arm/arm.c (arm_arch8_1m_main): Define and default initialize.
3678 (arm_option_reconfigure_globals): Initialize arm_arch8_1m_main.
3679 (arm_options_perform_arch_sanity_checks): Error out when targeting
3680 Armv8.1-M Mainline Security Extensions.
3681 * config/arm/arm.h (arm_arch8_1m_main): Declare.
3683 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
3685 * config/aarch64/aarch64-simd-builtins.def (aarch64_bfdot,
3686 aarch64_bfdot_lane, aarch64_bfdot_laneq): New.
3687 * config/aarch64/aarch64-simd.md (aarch64_bfdot, aarch64_bfdot_lane,
3688 aarch64_bfdot_laneq): New.
3689 * config/aarch64/arm_bf16.h (vbfdot_f32, vbfdotq_f32,
3690 vbfdot_lane_f32, vbfdotq_lane_f32, vbfdot_laneq_f32,
3691 vbfdotq_laneq_f32): New.
3692 * config/aarch64/iterators.md (UNSPEC_BFDOT, Vbfdottype,
3693 VBFMLA_W, VBF): New.
3694 (isquadop): Add V4BF, V8BF.
3696 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
3698 * config/aarch64/aarch64-builtins.c: (enum aarch64_type_qualifiers):
3699 New qualifier_lane_quadtup_index, TYPES_TERNOP_SSUS,
3700 TYPES_QUADOPSSUS_LANE_QUADTUP, TYPES_QUADOPSSSU_LANE_QUADTUP.
3701 (aarch64_simd_expand_args): Add case SIMD_ARG_LANE_QUADTUP_INDEX.
3702 (aarch64_simd_expand_builtin): Add qualifier_lane_quadtup_index.
3703 * config/aarch64/aarch64-simd-builtins.def (usdot, usdot_lane,
3704 usdot_laneq, sudot_lane,sudot_laneq): New.
3705 * config/aarch64/aarch64-simd.md (aarch64_usdot): New.
3706 (aarch64_<sur>dot_lane): New.
3707 * config/aarch64/arm_neon.h (vusdot_s32): New.
3709 (vusdot_lane_s32): New.
3710 (vsudot_lane_s32): New.
3711 * config/aarch64/iterators.md (DOTPROD_I8MM): New iterator.
3712 (UNSPEC_USDOT, UNSPEC_SUDOT): New unspecs.
3714 2020-01-16 Martin Liska <mliska@suse.cz>
3716 * value-prof.c (dump_histogram_value): Fix
3717 obvious spacing issue.
3719 2020-01-16 Andrew Pinski <apinski@marvell.com>
3721 * tree-ssa-sccvn.c(vn_reference_lookup_3): Check lhs for
3722 !storage_order_barrier_p.
3724 2020-01-16 Andrew Pinski <apinski@marvell.com>
3726 * sched-int.h (_dep): Add unused bit-field field for the padding.
3727 * sched-deps.c (init_dep_1): Init unused field.
3729 2020-01-16 Andrew Pinski <apinski@marvell.com>
3731 * optabs.h (create_expand_operand): Initialize target field also.
3733 2020-01-16 Andre Vieira <andre.simoesdiasvieira@arm.com>
3735 PR tree-optimization/92429
3736 * tree-ssa-loop-niter.h (simplify_replace_tree): Add parameter.
3737 * tree-ssa-loop-niter.c (simplify_replace_tree): Add parameter to
3739 * tree-vect-loop.c (update_epilogue_vinfo): Do not fold when replacing
3742 2020-01-16 Richard Sandiford <richard.sandiford@arm.com>
3744 * config/aarch64/aarch64.c (aarch64_split_sve_subreg_move): Apply
3745 aarch64_sve_int_mode to each mode.
3747 2020-01-15 David Malcolm <dmalcolm@redhat.com>
3749 * doc/analyzer.texi (Overview): Add note about
3750 -fdump-ipa-analyzer.
3752 2020-01-15 Wilco Dijkstra <wdijkstr@arm.com>
3754 PR tree-optimization/93231
3755 * tree-ssa-forwprop.c (optimize_count_trailing_zeroes): Check
3756 input_type is unsigned. Use tree_to_shwi for shift constant.
3757 Check CST_STRING element size is CHAR_TYPE_SIZE bits.
3758 (simplify_count_trailing_zeroes): Add test to handle known non-zero
3759 inputs more efficiently.
3761 2020-01-15 Uroš Bizjak <ubizjak@gmail.com>
3763 * config/i386/i386.md (*movsf_internal): Do not require
3764 SSE2 ISA for alternatives 14 and 15.
3766 2020-01-15 Richard Biener <rguenther@suse.de>
3769 * tree-eh.c (sink_clobbers): If we already visited the destination
3770 block do not defer insertion.
3771 (pass_lower_eh_dispatch::execute): Maintain BB_VISITED for
3772 the purpose of defered insertion.
3774 2020-01-15 Jakub Jelinek <jakub@redhat.com>
3776 * BASE-VER: Bump to 10.0.1.
3778 2020-01-15 Richard Sandiford <richard.sandiford@arm.com>
3780 PR tree-optimization/93247
3781 * tree-vect-loop.c (update_epilogue_loop_vinfo): Check the access
3782 type of the stmt that we're going to vectorize.
3784 2020-01-15 Richard Sandiford <richard.sandiford@arm.com>
3786 * tree-vect-slp.c (vectorize_slp_instance_root_stmt): Use a
3787 VIEW_CONVERT_EXPR if the vectorized constructor has a diffeent
3790 2020-01-15 Martin Liska <mliska@suse.cz>
3792 * ipa-profile.c (ipa_profile_read_edge_summary): Do not allow
3793 2 calls of streamer_read_hwi in a function call.
3795 2020-01-15 Richard Biener <rguenther@suse.de>
3797 * alias.c (record_alias_subset): Avoid redundant work when
3798 subset is already recorded.
3800 2020-01-14 David Malcolm <dmalcolm@redhat.com>
3802 * doc/invoke.texi (-fdiagnostics-show-cwe): Add note that some of
3803 the analyzer options provide CWE identifiers.
3805 2020-01-14 David Malcolm <dmalcolm@redhat.com>
3807 * tree-diagnostic-path.cc (path_summary::event_range::print):
3808 When testing for UNKNOWN_LOCATION, look through ad-hoc wrappers
3809 using get_pure_location.
3811 2020-01-15 Jakub Jelinek <jakub@redhat.com>
3813 PR tree-optimization/93262
3814 * tree-ssa-dse.c (maybe_trim_memstar_call): For *_chk builtins,
3815 perform head trimming only if the last argument is constant,
3816 either all ones, or larger or equal to head trim, in the latter
3817 case decrease the last argument by head_trim.
3819 PR tree-optimization/93249
3820 * tree-ssa-dse.c: Include builtins.h and gimple-fold.h.
3821 (maybe_trim_memstar_call): Move head_trim and tail_trim vars to
3822 function body scope, reindent. For BUILTIN_IN_STRNCPY*, don't
3823 perform head trim unless we can prove there are no '\0' chars
3824 from the source among the first head_trim chars.
3826 2020-01-14 David Malcolm <dmalcolm@redhat.com>
3828 * Makefile.in (ANALYZER_OBJS): Add analyzer/function-set.o.
3830 2020-01-15 Jakub Jelinek <jakub@redhat.com>
3833 * config/i386/sse.md
3834 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1,
3835 *<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1,
3836 *<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1,
3837 *<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1): Use
3838 just a single alternative instead of two, make operands 1 and 2
3841 2020-01-14 Jan Hubicka <hubicka@ucw.cz>
3844 * ipa-devirt.c (odr_types_equivalent_p): Compare TREE_ADDRESSABLE and
3847 2020-01-14 David Malcolm <dmalcolm@redhat.com>
3849 * Makefile.in (lang_opt_files): Add analyzer.opt.
3850 (ANALYZER_OBJS): New.
3851 (OBJS): Add digraph.o, graphviz.o, ordered-hash-map-tests.o,
3852 tristate.o and ANALYZER_OBJS.
3853 (TEXI_GCCINT_FILES): Add analyzer.texi.
3854 * common.opt (-fanalyzer): New driver option.
3855 * config.in: Regenerate.
3856 * configure: Regenerate.
3857 * configure.ac (--disable-analyzer, ENABLE_ANALYZER): New option.
3858 (gccdepdir): Also create depdir for "analyzer" subdir.
3859 * digraph.cc: New file.
3860 * digraph.h: New file.
3861 * doc/analyzer.texi: New file.
3862 * doc/gccint.texi ("Static Analyzer") New menu item.
3863 (analyzer.texi): Include it.
3864 * doc/invoke.texi ("Static Analyzer Options"): New list and new section.
3865 ("Warning Options"): Add static analysis warnings to the list.
3866 (-Wno-analyzer-double-fclose): New option.
3867 (-Wno-analyzer-double-free): New option.
3868 (-Wno-analyzer-exposure-through-output-file): New option.
3869 (-Wno-analyzer-file-leak): New option.
3870 (-Wno-analyzer-free-of-non-heap): New option.
3871 (-Wno-analyzer-malloc-leak): New option.
3872 (-Wno-analyzer-possible-null-argument): New option.
3873 (-Wno-analyzer-possible-null-dereference): New option.
3874 (-Wno-analyzer-null-argument): New option.
3875 (-Wno-analyzer-null-dereference): New option.
3876 (-Wno-analyzer-stale-setjmp-buffer): New option.
3877 (-Wno-analyzer-tainted-array-index): New option.
3878 (-Wno-analyzer-use-after-free): New option.
3879 (-Wno-analyzer-use-of-pointer-in-stale-stack-frame): New option.
3880 (-Wno-analyzer-use-of-uninitialized-value): New option.
3881 (-Wanalyzer-too-complex): New option.
3882 (-fanalyzer-call-summaries): New warning.
3883 (-fanalyzer-checker=): New warning.
3884 (-fanalyzer-fine-grained): New warning.
3885 (-fno-analyzer-state-merge): New warning.
3886 (-fno-analyzer-state-purge): New warning.
3887 (-fanalyzer-transitivity): New warning.
3888 (-fanalyzer-verbose-edges): New warning.
3889 (-fanalyzer-verbose-state-changes): New warning.
3890 (-fanalyzer-verbosity=): New warning.
3891 (-fdump-analyzer): New warning.
3892 (-fdump-analyzer-callgraph): New warning.
3893 (-fdump-analyzer-exploded-graph): New warning.
3894 (-fdump-analyzer-exploded-nodes): New warning.
3895 (-fdump-analyzer-exploded-nodes-2): New warning.
3896 (-fdump-analyzer-exploded-nodes-3): New warning.
3897 (-fdump-analyzer-supergraph): New warning.
3898 * doc/sourcebuild.texi (dg-require-dot): New.
3899 (dg-check-dot): New.
3900 * gdbinit.in (break-on-saved-diagnostic): New command.
3901 * graphviz.cc: New file.
3902 * graphviz.h: New file.
3903 * ordered-hash-map-tests.cc: New file.
3904 * ordered-hash-map.h: New file.
3905 * passes.def (pass_analyzer): Add before
3906 pass_ipa_whole_program_visibility.
3907 * selftest-run-tests.c (selftest::run_tests): Call
3908 selftest::ordered_hash_map_tests_cc_tests.
3909 * selftest.h (selftest::ordered_hash_map_tests_cc_tests): New
3911 * shortest-paths.h: New file.
3912 * timevar.def (TV_ANALYZER): New timevar.
3913 (TV_ANALYZER_SUPERGRAPH): Likewise.
3914 (TV_ANALYZER_STATE_PURGE): Likewise.
3915 (TV_ANALYZER_PLAN): Likewise.
3916 (TV_ANALYZER_SCC): Likewise.
3917 (TV_ANALYZER_WORKLIST): Likewise.
3918 (TV_ANALYZER_DUMP): Likewise.
3919 (TV_ANALYZER_DIAGNOSTICS): Likewise.
3920 (TV_ANALYZER_SHORTEST_PATHS): Likewise.
3921 * tree-pass.h (make_pass_analyzer): New decl.
3922 * tristate.cc: New file.
3923 * tristate.h: New file.
3925 2020-01-14 Uroš Bizjak <ubizjak@gmail.com>
3928 * config/i386/i386.md (*movsf_internal): Require SSE2 ISA for
3929 alternatives 9 and 10.
3931 2020-01-14 David Malcolm <dmalcolm@redhat.com>
3933 * attribs.c (excl_hash_traits::empty_zero_p): New static constant.
3934 * gcov.c (function_start_pair_hash::empty_zero_p): Likewise.
3935 * graphite.c (struct sese_scev_hash::empty_zero_p): Likewise.
3936 * hash-map-tests.c (selftest::test_nonzero_empty_key): New selftest.
3937 (selftest::hash_map_tests_c_tests): Call it.
3938 * hash-map-traits.h (simple_hashmap_traits::empty_zero_p):
3939 New static constant, using the value of = H::empty_zero_p.
3940 (unbounded_hashmap_traits::empty_zero_p): Likewise, using the value
3941 from default_hash_traits <Value>.
3942 * hash-map.h (hash_map::empty_zero_p): Likewise, using the value
3944 * hash-set-tests.c (value_hash_traits::empty_zero_p): Likewise.
3945 * hash-table.h (hash_table::alloc_entries): Guard the loop of
3946 calls to mark_empty with !Descriptor::empty_zero_p.
3947 (hash_table::empty_slow): Conditionalize the memset call with a
3948 check that Descriptor::empty_zero_p; otherwise, loop through the
3949 entries calling mark_empty on them.
3950 * hash-traits.h (int_hash::empty_zero_p): New static constant.
3951 (pointer_hash::empty_zero_p): Likewise.
3952 (pair_hash::empty_zero_p): Likewise.
3953 * ipa-devirt.c (default_hash_traits <type_pair>::empty_zero_p):
3955 * ipa-prop.c (ipa_bit_ggc_hash_traits::empty_zero_p): Likewise.
3956 (ipa_vr_ggc_hash_traits::empty_zero_p): Likewise.
3957 * profile.c (location_triplet_hash::empty_zero_p): Likewise.
3958 * sanopt.c (sanopt_tree_triplet_hash::empty_zero_p): Likewise.
3959 (sanopt_tree_couple_hash::empty_zero_p): Likewise.
3960 * tree-hasher.h (int_tree_hasher::empty_zero_p): Likewise.
3961 * tree-ssa-sccvn.c (vn_ssa_aux_hasher::empty_zero_p): Likewise.
3962 * tree-vect-slp.c (bst_traits::empty_zero_p): Likewise.
3964 (default_hash_traits<scalar_cond_masked_key>::empty_zero_p):
3967 2020-01-14 Kewen Lin <linkw@gcc.gnu.org>
3969 * cfgloopanal.c (average_num_loop_insns): Free bbs when early return,
3970 fix typo on return value.
3972 2020-01-14 Xiong Hu Luo <luoxhu@linux.ibm.com>
3975 * cgraph.c (symbol_table::create_edge): Init speculative_id and
3977 (cgraph_edge::make_speculative): Add param for setting speculative_id
3979 (cgraph_edge::speculative_call_info): Update comments and find reference
3980 by speculative_id for multiple indirect targets.
3981 (cgraph_edge::resolve_speculation): Decrease the speculations
3982 for indirect edge, drop it's speculative if not direct target
3983 left. Update comments.
3984 (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
3985 (cgraph_node::dump): Print num_speculative_call_targets.
3986 (cgraph_node::verify_node): Don't report error if speculative
3987 edge not include statement.
3988 (cgraph_edge::num_speculative_call_targets_p): New function.
3989 * cgraph.h (int common_target_id): Remove.
3990 (int common_target_probability): Remove.
3991 (num_speculative_call_targets): New variable.
3992 (make_speculative): Add param for setting speculative_id.
3993 (cgraph_edge::num_speculative_call_targets_p): New declare.
3994 (target_prob): New variable.
3995 (speculative_id): New variable.
3996 * ipa-fnsummary.c (analyze_function_body): Create and duplicate
3997 call summaries for multiple speculative call targets.
3998 * cgraphclones.c (cgraph_node::create_clone): Clone speculative_id.
3999 * ipa-profile.c (struct speculative_call_target): New struct.
4000 (class speculative_call_summary): New class.
4001 (class speculative_call_summaries): New class.
4002 (call_sums): New variable.
4003 (ipa_profile_generate_summary): Generate indirect multiple targets summaries.
4004 (ipa_profile_write_edge_summary): New function.
4005 (ipa_profile_write_summary): Stream out indirect multiple targets summaries.
4006 (ipa_profile_dump_all_summaries): New function.
4007 (ipa_profile_read_edge_summary): New function.
4008 (ipa_profile_read_summary_section): New function.
4009 (ipa_profile_read_summary): Stream in indirect multiple targets summaries.
4010 (ipa_profile): Generate num_speculative_call_targets from
4012 * ipa-ref.h (speculative_id): New variable.
4013 * ipa-utils.c (ipa_merge_profiles): Update with target_prob.
4014 * lto-cgraph.c (lto_output_edge): Remove indirect common_target_id and
4015 common_target_probability. Stream out speculative_id and
4016 num_speculative_call_targets.
4017 (input_edge): Likewise.
4018 * predict.c (dump_prediction): Remove edges count assert to be
4020 * symtab.c (symtab_node::create_reference): Init speculative_id.
4021 (symtab_node::clone_references): Clone speculative_id.
4022 (symtab_node::clone_referring): Clone speculative_id.
4023 (symtab_node::clone_reference): Clone speculative_id.
4024 (symtab_node::clear_stmts_in_references): Clear speculative_id.
4025 * tree-inline.c (copy_bb): Duplicate all the speculative edges
4026 if indirect call contains multiple speculative targets.
4027 * value-prof.h (check_ic_target): Remove.
4028 * value-prof.c (gimple_value_profile_transformations):
4029 Use void function gimple_ic_transform.
4030 * value-prof.c (gimple_ic_transform): Handle topn case.
4031 Fix comment typos. Change it to a void function.
4033 2020-01-13 Andrew Pinski <apinski@marvell.com>
4035 * config/aarch64/aarch64-cores.def (octeontx2): New define.
4036 (octeontx2t98): New define.
4037 (octeontx2t96): New define.
4038 (octeontx2t93): New define.
4039 (octeontx2f95): New define.
4040 (octeontx2f95n): New define.
4041 (octeontx2f95mm): New define.
4042 * config/aarch64/aarch64-tune.md: Regenerate.
4043 * doc/invoke.texi (-mcpu=): Document the new cpu types.
4045 2020-01-13 Jason Merrill <jason@redhat.com>
4047 PR c++/33799 - destroy return value if local cleanup throws.
4048 * gimplify.c (gimplify_return_expr): Handle COMPOUND_EXPR.
4050 2020-01-13 Martin Liska <mliska@suse.cz>
4052 * ipa-cp.c (get_max_overall_size): Use newly
4053 renamed param param_ipa_cp_unit_growth.
4054 * params.opt: Remove legacy param name.
4056 2020-01-13 Martin Sebor <msebor@redhat.com>
4058 PR tree-optimization/93213
4059 * tree-ssa-strlen.c (handle_store): Only allow single-byte nul-over-nul
4060 stores to be eliminated.
4062 2020-01-13 Martin Liska <mliska@suse.cz>
4064 * opts.c (print_help): Do not print CL_PARAM
4065 and CL_WARNING for CL_OPTIMIZATION.
4067 2020-01-13 Jonathan Wakely <jwakely@redhat.com>
4070 * doc/invoke.texi (Warning Options): Add caveat about some warnings
4071 depending on optimization settings.
4073 2020-01-13 Jakub Jelinek <jakub@redhat.com>
4075 PR tree-optimization/90838
4076 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
4077 SCALAR_INT_TYPE_MODE directly in CTZ_DEFINED_VALUE_AT_ZERO macro
4078 argument rather than to initialize temporary for targets that
4079 don't use the mode argument at all. Initialize ctzval to avoid
4082 2020-01-10 Thomas Schwinge <thomas@codesourcery.com>
4084 * tree.h (OMP_CLAUSE_USE_DEVICE_PTR_IF_PRESENT): New definition.
4085 * tree-core.h: Document it.
4086 * gimplify.c (gimplify_omp_workshare): Set it.
4087 * omp-low.c (lower_omp_target): Use it.
4088 * tree-pretty-print.c (dump_omp_clause): Print it.
4090 * omp-low.c (lower_omp_target) <OMP_CLAUSE_USE_DEVICE_PTR etc.>:
4091 Assert that for OpenACC we always have 'GOMP_MAP_USE_DEVICE_PTR'.
4093 2020-01-10 David Malcolm <dmalcolm@redhat.com>
4095 * Makefile.in (OBJS): Add tree-diagnostic-path.o.
4096 * common.opt (fdiagnostics-path-format=): New option.
4097 (diagnostic_path_format): New enum.
4098 (fdiagnostics-show-path-depths): New option.
4099 * coretypes.h (diagnostic_event_id_t): New forward decl.
4100 * diagnostic-color.c (color_dict): Add "path".
4101 * diagnostic-event-id.h: New file.
4102 * diagnostic-format-json.cc (json_from_expanded_location): Make
4104 (json_end_diagnostic): Call context->make_json_for_path if it
4105 exists and the diagnostic has a path.
4106 (diagnostic_output_format_init): Clear context->print_path.
4107 * diagnostic-path.h: New file.
4108 * diagnostic-show-locus.c (colorizer::set_range): Special-case
4109 when printing a run of events in a diagnostic_path so that they
4110 all get the same color.
4111 (layout::m_diagnostic_path_p): New field.
4112 (layout::layout): Initialize it.
4113 (layout::print_any_labels): Don't colorize the label text for an
4114 event in a diagnostic_path.
4115 (gcc_rich_location::add_location_if_nearby): Add
4116 "restrict_to_current_line_spans" and "label" params. Pass the
4117 former to layout.maybe_add_location_range; pass the latter
4118 when calling add_range.
4119 * diagnostic.c: Include "diagnostic-path.h".
4120 (diagnostic_initialize): Initialize context->path_format and
4121 context->show_path_depths.
4122 (diagnostic_show_any_path): New function.
4123 (diagnostic_path::interprocedural_p): New function.
4124 (diagnostic_report_diagnostic): Call diagnostic_show_any_path.
4125 (simple_diagnostic_path::num_events): New function.
4126 (simple_diagnostic_path::get_event): New function.
4127 (simple_diagnostic_path::add_event): New function.
4128 (simple_diagnostic_event::simple_diagnostic_event): New ctor.
4129 (simple_diagnostic_event::~simple_diagnostic_event): New dtor.
4130 (debug): New overload taking a diagnostic_path *.
4131 * diagnostic.def (DK_DIAGNOSTIC_PATH): New.
4132 * diagnostic.h (enum diagnostic_path_format): New enum.
4133 (json::value): New forward decl.
4134 (diagnostic_context::path_format): New field.
4135 (diagnostic_context::show_path_depths): New field.
4136 (diagnostic_context::print_path): New callback field.
4137 (diagnostic_context::make_json_for_path): New callback field.
4138 (diagnostic_show_any_path): New decl.
4139 (json_from_expanded_location): New decl.
4140 * doc/invoke.texi (-fdiagnostics-path-format=): New option.
4141 (-fdiagnostics-show-path-depths): New option.
4142 (-fdiagnostics-color): Add "path" to description of default
4143 GCC_COLORS; describe it.
4144 (-fdiagnostics-format=json): Document how diagnostic paths are
4145 represented in the JSON output format.
4146 * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
4147 Add optional params "restrict_to_current_line_spans" and "label".
4148 * opts.c (common_handle_option): Handle
4149 OPT_fdiagnostics_path_format_ and
4150 OPT_fdiagnostics_show_path_depths.
4151 * pretty-print.c: Include "diagnostic-event-id.h".
4152 (pp_format): Implement "%@" format code for printing
4153 diagnostic_event_id_t *.
4154 (selftest::test_pp_format): Add tests for "%@".
4155 * selftest-run-tests.c (selftest::run_tests): Call
4156 selftest::tree_diagnostic_path_cc_tests.
4157 * selftest.h (selftest::tree_diagnostic_path_cc_tests): New decl.
4158 * toplev.c (general_init): Initialize global_dc->path_format and
4159 global_dc->show_path_depths.
4160 * tree-diagnostic-path.cc: New file.
4161 * tree-diagnostic.c (maybe_unwind_expanded_macro_loc): Make
4162 non-static. Drop "diagnostic" param in favor of storing the
4163 original value of "where" and re-using it.
4164 (virt_loc_aware_diagnostic_finalizer): Update for dropped param of
4165 maybe_unwind_expanded_macro_loc.
4166 (tree_diagnostics_defaults): Initialize context->print_path and
4167 context->make_json_for_path.
4168 * tree-diagnostic.h (default_tree_diagnostic_path_printer): New
4170 (default_tree_make_json_for_path): New decl.
4171 (maybe_unwind_expanded_macro_loc): New decl.
4173 2020-01-10 Jakub Jelinek <jakub@redhat.com>
4175 PR tree-optimization/93210
4176 * fold-const.h (native_encode_initializer,
4177 can_native_interpret_type_p): Declare.
4178 * fold-const.c (native_encode_string): Fix up handling with off != -1,
4180 (native_encode_initializer): New function, moved from dwarf2out.c.
4181 Adjust to native_encode_expr compatible arguments, including dry-run
4182 and partial extraction modes. Don't handle STRING_CST.
4183 (can_native_interpret_type_p): No longer static.
4184 * gimple-fold.c (fold_ctor_reference): For native_encode_expr, verify
4185 offset / BITS_PER_UNIT fits into int and don't call it if
4186 can_native_interpret_type_p fails. If suboff is NULL and for
4187 CONSTRUCTOR fold_{,non}array_ctor_reference returns NULL, retry with
4188 native_encode_initializer.
4189 (fold_const_aggregate_ref_1): Formatting fix.
4190 * dwarf2out.c (native_encode_initializer): Moved to fold-const.c.
4191 (tree_add_const_value_attribute): Adjust caller.
4193 PR tree-optimization/90838
4194 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
4195 SCALAR_INT_TYPE_MODE instead of TYPE_MODE as operand of
4196 CTZ_DEFINED_VALUE_AT_ZERO.
4198 2020-01-10 Vladimir Makarov <vmakarov@redhat.com>
4201 * lra-constraints.c (match_reload): Permit input operands have the
4202 same mode as output while other input operands have a different
4205 2020-01-10 Wilco Dijkstra <wdijkstr@arm.com>
4207 PR tree-optimization/90838
4208 * tree-ssa-forwprop.c (check_ctz_array): Add new function.
4209 (check_ctz_string): Likewise.
4210 (optimize_count_trailing_zeroes): Likewise.
4211 (simplify_count_trailing_zeroes): Likewise.
4212 (pass_forwprop::execute): Try ctz simplification.
4213 * match.pd: Add matching for ctz idioms.
4215 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
4217 * config/aarch64/aarch64.c (aarch64_invalid_conversion): New function
4219 (aarch64_invalid_unary_op): New function for target hook.
4220 (aarch64_invalid_binary_op): New function for target hook.
4222 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
4224 * config.gcc: Add arm_bf16.h.
4225 * config/aarch64/aarch64-builtins.c
4226 (aarch64_simd_builtin_std_type): Add BFmode.
4227 (aarch64_init_simd_builtin_types): Define element types for vector
4229 (aarch64_init_bf16_types): New function.
4230 (aarch64_general_init_builtins): Add arm_init_bf16_types function call.
4231 * config/aarch64/aarch64-modes.def: Add BFmode and V4BF, V8BF vector
4233 * config/aarch64/aarch64-simd-builtin-types.def: Add BF SIMD types.
4234 * config/aarch64/aarch64-simd.md: Add BF vector types to NEON move
4236 * config/aarch64/aarch64.h (AARCH64_VALID_SIMD_DREG_MODE): Add V4BF.
4237 (AARCH64_VALID_SIMD_QREG_MODE): Add V8BF.
4238 * config/aarch64/aarch64.c
4239 (aarch64_classify_vector_mode): Add support for BF types.
4240 (aarch64_gimplify_va_arg_expr): Add support for BF types.
4241 (aarch64_vq_mode): Add support for BF types.
4242 (aarch64_simd_container_mode): Add support for BF types.
4243 (aarch64_mangle_type): Add support for BF scalar type.
4244 * config/aarch64/aarch64.md: Add BFmode to movhf pattern.
4245 * config/aarch64/arm_bf16.h: New file.
4246 * config/aarch64/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
4247 * config/aarch64/iterators.md: Add BF types to mode attributes.
4248 (HFBF, GPF_TF_F16_MOV, VDMOV, VQMOV, VQMOV_NO2Em VALL_F16MOV): New.
4250 2020-01-10 Jason Merrill <jason@redhat.com>
4252 PR c++/93173 - incorrect tree sharing.
4253 * gimplify.c (copy_if_shared): No longer static.
4254 * gimplify.h: Declare it.
4256 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
4258 * doc/invoke.texi (-msve-vector-bits=): Document that
4259 -msve-vector-bits=128 now generates VL-specific code for
4260 little-endian targets.
4261 * config/aarch64/aarch64-sve-builtins.cc (register_builtin_types): Use
4262 build_vector_type_for_mode to construct the data vector types.
4263 * config/aarch64/aarch64.c (aarch64_convert_sve_vector_bits): Generate
4264 VL-specific code for -msve-vector-bits=128 on little-endian targets.
4265 (aarch64_simd_container_mode): Always prefer Advanced SIMD modes
4266 for 128-bit vectors.
4268 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
4270 * config/aarch64/aarch64.c (aarch64_evpc_sel): Fix gen_vcond_mask
4273 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
4275 * config/aarch64/aarch64-builtins.c
4276 (aarch64_builtin_vectorized_function): Check for specific vector modes,
4277 rather than checking the number of elements and the element mode.
4279 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
4281 * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
4282 get_related_vectype_for_scalar_type rather than build_vector_type
4283 to create the index type for a conditional reduction.
4285 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
4287 * tree-vect-loop.c (update_epilogue_loop_vinfo): Update DR_REF
4288 for any type of gather or scatter, including strided accesses.
4290 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
4292 * tree-vectorizer.h (get_dr_vinfo_offset): Add missing function
4295 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
4297 * tree-vect-data-refs.c (vect_create_addr_base_for_vector_ref): Use
4299 * tree-vect-loop.c (update_epilogue_loop_vinfo): Remove orig_drs_init
4300 parameter and its use to reset DR_OFFSET's.
4301 (vect_transform_loop): Remove orig_drs_init argument.
4302 * tree-vect-loop-manip.c (vect_update_init_of_dr): Update the offset
4303 member of dr_vec_info rather than the offset of the associated
4304 data_reference's innermost_loop_behavior.
4305 (vect_update_init_of_dr): Pass dr_vec_info instead of data_reference.
4306 (vect_do_peeling): Remove orig_drs_init parameter and its construction.
4307 * tree-vect-stmts.c (check_scan_store): Replace use of DR_OFFSET with
4308 get_dr_vinfo_offset.
4309 (vectorizable_store): Likewise.
4310 (vectorizable_load): Likewise.
4312 2020-01-10 Richard Biener <rguenther@suse.de>
4314 * gimple-ssa-store-merging
4315 (pass_store_merging::terminate_all_aliasing_chains): Cache alias info.
4317 2020-01-10 Martin Liska <mliska@suse.cz>
4320 * ipa-inline-analysis.c (offline_size): Make proper parenthesis
4321 encapsulation that was there before r280040.
4323 2020-01-10 Richard Biener <rguenther@suse.de>
4326 * tree-eh.c (sink_clobbers): Move clobbers to out-of-IL
4327 sequences to avoid walking them again for secondary opportunities.
4328 (pass_lower_eh_dispatch::execute): Instead actually insert
4331 2020-01-10 Richard Biener <rguenther@suse.de>
4334 * tree-eh.c (redirect_eh_edge_1): Avoid some work if possible.
4335 (cleanup_all_empty_eh): Walk landing pads in reverse order to
4336 avoid quadraticness.
4338 2020-01-10 Martin Jambor <mjambor@suse.cz>
4340 * params.opt (param_ipa_sra_max_replacements): Mark as Optimization.
4341 * ipa-sra.c (pull_accesses_from_callee): New parameter caller, use it
4342 to get param_ipa_sra_max_replacements.
4343 (param_splitting_across_edge): Pass the caller to
4344 pull_accesses_from_callee.
4346 2020-01-10 Martin Jambor <mjambor@suse.cz>
4348 * params.opt (param_ipcp_unit_growth): Mark as Optimization.
4349 * ipa-cp.c (max_new_size): Removed.
4350 (orig_overall_size): New variable.
4351 (get_max_overall_size): New function.
4352 (estimate_local_effects): Use it. Adjust dump.
4353 (decide_about_value): Likewise.
4354 (ipcp_propagate_stage): Do not calculate max_new_size, just store
4355 orig_overall_size. Adjust dump.
4356 (ipa_cp_c_finalize): Clear orig_overall_size instead of max_new_size.
4358 2020-01-10 Martin Jambor <mjambor@suse.cz>
4360 * params.opt (param_ipa_max_agg_items): Mark as Optimization
4361 * ipa-cp.c (merge_agg_lats_step): New parameter max_agg_items, use
4362 instead of param_ipa_max_agg_items.
4363 (merge_aggregate_lattices): Extract param_ipa_max_agg_items from
4364 optimization info for the callee.
4366 2020-01-09 Kwok Cheung Yeung <kcy@codesourcery.com>
4368 * lto-streamer-in.c (input_function): Remove streamed-in inline debug
4369 markers if debug_inline_points is false.
4371 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4373 * config.gcc (aarch64*-*-*): Add aarch64-sve-builtins-sve2.o to
4375 * config/aarch64/t-aarch64 (aarch64-sve-builtins.o): Depend on
4376 aarch64-sve-builtins-base.def, aarch64-sve-builtins-sve2.def and
4377 aarch64-sve-builtins-sve2.h.
4378 (aarch64-sve-builtins-sve2.o): New rule.
4379 * config/aarch64/aarch64.h (AARCH64_ISA_SVE2_AES): New macro.
4380 (AARCH64_ISA_SVE2_BITPERM, AARCH64_ISA_SVE2_SHA3): Likewise.
4381 (AARCH64_ISA_SVE2_SM4, TARGET_SVE2_AES, TARGET_SVE2_BITPERM): Likewise.
4382 (TARGET_SVE2_SHA, TARGET_SVE2_SM4): Likewise.
4383 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
4384 TARGET_SVE2_AES, TARGET_SVE2_BITPERM, TARGET_SVE2_SHA3 and
4386 * config/aarch64/aarch64-sve.md: Update comments with SVE2
4387 instructions that are handled here.
4388 (@cond_asrd<mode>): Generalize to...
4389 (@cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>): ...this.
4390 (*cond_asrd<mode>_2): Generalize to...
4391 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_2): ...this.
4392 (*cond_asrd<mode>_z): Generalize to...
4393 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_z): ...this.
4394 * config/aarch64/aarch64.md (UNSPEC_LDNT1_GATHER): New unspec.
4395 (UNSPEC_STNT1_SCATTER, UNSPEC_WHILEGE, UNSPEC_WHILEGT): Likewise.
4396 (UNSPEC_WHILEHI, UNSPEC_WHILEHS): Likewise.
4397 * config/aarch64/aarch64-sve2.md (@aarch64_gather_ldnt<mode>): New
4399 (@aarch64_gather_ldnt_<ANY_EXTEND:optab><SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
4400 (@aarch64_scatter_stnt<mode>): Likewise.
4401 (@aarch64_scatter_stnt_<SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
4402 (@aarch64_mul_lane_<mode>): Likewise.
4403 (@aarch64_sve_suqadd<mode>_const): Likewise.
4404 (*<sur>h<addsub><mode>): Generalize to...
4405 (@aarch64_pred_<SVE2_COND_INT_BINARY_REV:sve_int_op><mode>): ...this
4407 (@cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>): New expander.
4408 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_2): New pattern.
4409 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_3): Likewise.
4410 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_any): Likewise.
4411 (*cond_<SVE2_COND_INT_BINARY_NOREV:sve_int_op><mode>_z): Likewise.
4412 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op><mode>):: Likewise.
4413 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op>_lane_<mode>): Likewise.
4414 (@aarch64_pred_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): Likewise.
4415 (@cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): New expander.
4416 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_2): New pattern.
4417 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_3): Likewise.
4418 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_any): Likewise.
4419 (@aarch64_sve_<SVE2_INT_TERNARY:sve_int_op><mode>): Likewise.
4420 (@aarch64_sve_<SVE2_INT_TERNARY_LANE:sve_int_op>_lane_<mode>)
4421 (@aarch64_sve_add_mul_lane_<mode>): Likewise.
4422 (@aarch64_sve_sub_mul_lane_<mode>): Likewise.
4423 (@aarch64_sve2_xar<mode>): Likewise.
4424 (@aarch64_sve2_bcax<mode>): Likewise.
4425 (*aarch64_sve2_eor3<mode>): Rename to...
4426 (@aarch64_sve2_eor3<mode>): ...this.
4427 (@aarch64_sve2_bsl<mode>): New expander.
4428 (@aarch64_sve2_nbsl<mode>): Likewise.
4429 (@aarch64_sve2_bsl1n<mode>): Likewise.
4430 (@aarch64_sve2_bsl2n<mode>): Likewise.
4431 (@aarch64_sve_add_<SHIFTRT:sve_int_op><mode>): Likewise.
4432 (*aarch64_sve2_sra<mode>): Add MOVPRFX support.
4433 (@aarch64_sve_add_<VRSHR_N:sve_int_op><mode>): New pattern.
4434 (@aarch64_sve_<SVE2_INT_SHIFT_INSERT:sve_int_op><mode>): Likewise.
4435 (@aarch64_sve2_<USMAX:su>aba<mode>): New expander.
4436 (*aarch64_sve2_<USMAX:su>aba<mode>): New pattern.
4437 (@aarch64_sve_<SVE2_INT_BINARY_WIDE:sve_int_op><mode>): Likewise.
4438 (<su>mull<bt><Vwide>): Generalize to...
4439 (@aarch64_sve_<SVE2_INT_BINARY_LONG:sve_int_op><mode>): ...this new
4441 (@aarch64_sve_<SVE2_INT_BINARY_LONG_lANE:sve_int_op>_lane_<mode>)
4442 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_LONG:sve_int_op><mode>)
4443 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG:sve_int_op><mode>)
4444 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
4445 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG:sve_int_op><mode>)
4446 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
4447 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG:sve_int_op><mode>)
4448 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
4449 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG:sve_int_op><mode>)
4450 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
4451 (@aarch64_sve_<SVE2_FP_TERNARY_LONG:sve_fp_op><mode>): New patterns.
4452 (@aarch64_<SVE2_FP_TERNARY_LONG_LANE:sve_fp_op>_lane_<mode>)
4453 (@aarch64_sve_<SVE2_INT_UNARY_NARROWB:sve_int_op><mode>): Likewise.
4454 (@aarch64_sve_<SVE2_INT_UNARY_NARROWT:sve_int_op><mode>): Likewise.
4455 (@aarch64_sve_<SVE2_INT_BINARY_NARROWB:sve_int_op><mode>): Likewise.
4456 (@aarch64_sve_<SVE2_INT_BINARY_NARROWT:sve_int_op><mode>): Likewise.
4457 (<SHRNB:r>shrnb<mode>): Generalize to...
4458 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWB:sve_int_op><mode>): ...this
4460 (<SHRNT:r>shrnt<mode>): Generalize to...
4461 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWT:sve_int_op><mode>): ...this
4463 (@aarch64_pred_<SVE2_INT_BINARY_PAIR:sve_int_op><mode>): New pattern.
4464 (@aarch64_pred_<SVE2_FP_BINARY_PAIR:sve_fp_op><mode>): Likewise.
4465 (@cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>): New expander.
4466 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_2): New pattern.
4467 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_z): Likewise.
4468 (@aarch64_sve_<SVE2_INT_CADD:optab><mode>): Likewise.
4469 (@aarch64_sve_<SVE2_INT_CMLA:optab><mode>): Likewise.
4470 (@aarch64_<SVE2_INT_CMLA:optab>_lane_<mode>): Likewise.
4471 (@aarch64_sve_<SVE2_INT_CDOT:optab><mode>): Likewise.
4472 (@aarch64_<SVE2_INT_CDOT:optab>_lane_<mode>): Likewise.
4473 (@aarch64_pred_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): Likewise.
4474 (@cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New expander.
4475 (*cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New pattern.
4476 (@aarch64_sve2_cvtnt<mode>): Likewise.
4477 (@aarch64_pred_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): Likewise.
4478 (@cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): New expander.
4479 (*cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>_any): New pattern.
4480 (@aarch64_sve2_cvtxnt<mode>): Likewise.
4481 (@aarch64_pred_<SVE2_U32_UNARY:sve_int_op><mode>): Likewise.
4482 (@cond_<SVE2_U32_UNARY:sve_int_op><mode>): New expander.
4483 (*cond_<SVE2_U32_UNARY:sve_int_op><mode>): New pattern.
4484 (@aarch64_pred_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): Likewise.
4485 (@cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New expander.
4486 (*cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New pattern.
4487 (@aarch64_sve2_pmul<mode>): Likewise.
4488 (@aarch64_sve_<SVE2_PMULL:optab><mode>): Likewise.
4489 (@aarch64_sve_<SVE2_PMULL_PAIR:optab><mode>): Likewise.
4490 (@aarch64_sve2_tbl2<mode>): Likewise.
4491 (@aarch64_sve2_tbx<mode>): Likewise.
4492 (@aarch64_sve_<SVE2_INT_BITPERM:sve_int_op><mode>): Likewise.
4493 (@aarch64_sve2_histcnt<mode>): Likewise.
4494 (@aarch64_sve2_histseg<mode>): Likewise.
4495 (@aarch64_pred_<SVE2_MATCH:sve_int_op><mode>): Likewise.
4496 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_cc): Likewise.
4497 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_ptest): Likewise.
4498 (aarch64_sve2_aes<CRYPTO_AES:aes_op>): Likewise.
4499 (aarch64_sve2_aes<CRYPTO_AESMC:aesmc_op>): Likewise.
4500 (*aarch64_sve2_aese_fused, *aarch64_sve2_aesd_fused): Likewise.
4501 (aarch64_sve2_rax1, aarch64_sve2_sm4e, aarch64_sve2_sm4ekey): Likewise.
4502 (<su>mulh<r>s<mode>3): Update after above pattern name changes.
4503 * config/aarch64/iterators.md (VNx16QI_ONLY, VNx4SF_ONLY)
4504 (SVE_STRUCT2, SVE_FULL_BHI, SVE_FULL_HSI, SVE_FULL_HDI)
4505 (SVE2_PMULL_PAIR_I): New mode iterators.
4506 (UNSPEC_ADCLB, UNSPEC_ADCLT, UNSPEC_ADDHNB, UNSPEC_ADDHNT, UNSPEC_BDEP)
4507 (UNSPEC_BEXT, UNSPEC_BGRP, UNSPEC_CADD90, UNSPEC_CADD270, UNSPEC_CDOT)
4508 (UNSPEC_CDOT90, UNSPEC_CDOT180, UNSPEC_CDOT270, UNSPEC_CMLA)
4509 (UNSPEC_CMLA90, UNSPEC_CMLA180, UNSPEC_CMLA270, UNSPEC_COND_FCVTLT)
4510 (UNSPEC_COND_FCVTNT, UNSPEC_COND_FCVTX, UNSPEC_COND_FCVTXNT)
4511 (UNSPEC_COND_FLOGB, UNSPEC_EORBT, UNSPEC_EORTB, UNSPEC_FADDP)
4512 (UNSPEC_FMAXP, UNSPEC_FMAXNMP, UNSPEC_FMLALB, UNSPEC_FMLALT)
4513 (UNSPEC_FMLSLB, UNSPEC_FMLSLT, UNSPEC_FMINP, UNSPEC_FMINNMP)
4514 (UNSPEC_HISTCNT, UNSPEC_HISTSEG, UNSPEC_MATCH, UNSPEC_NMATCH)
4515 (UNSPEC_PMULLB, UNSPEC_PMULLB_PAIR, UNSPEC_PMULLT, UNSPEC_PMULLT_PAIR)
4516 (UNSPEC_RADDHNB, UNSPEC_RADDHNT, UNSPEC_RSUBHNB, UNSPEC_RSUBHNT)
4517 (UNSPEC_SLI, UNSPEC_SRI, UNSPEC_SABDLB, UNSPEC_SABDLT, UNSPEC_SADDLB)
4518 (UNSPEC_SADDLBT, UNSPEC_SADDLT, UNSPEC_SADDWB, UNSPEC_SADDWT)
4519 (UNSPEC_SBCLB, UNSPEC_SBCLT, UNSPEC_SMAXP, UNSPEC_SMINP)
4520 (UNSPEC_SQCADD90, UNSPEC_SQCADD270, UNSPEC_SQDMULLB, UNSPEC_SQDMULLBT)
4521 (UNSPEC_SQDMULLT, UNSPEC_SQRDCMLAH, UNSPEC_SQRDCMLAH90)
4522 (UNSPEC_SQRDCMLAH180, UNSPEC_SQRDCMLAH270, UNSPEC_SQRSHRNB)
4523 (UNSPEC_SQRSHRNT, UNSPEC_SQRSHRUNB, UNSPEC_SQRSHRUNT, UNSPEC_SQSHRNB)
4524 (UNSPEC_SQSHRNT, UNSPEC_SQSHRUNB, UNSPEC_SQSHRUNT, UNSPEC_SQXTNB)
4525 (UNSPEC_SQXTNT, UNSPEC_SQXTUNB, UNSPEC_SQXTUNT, UNSPEC_SSHLLB)
4526 (UNSPEC_SSHLLT, UNSPEC_SSUBLB, UNSPEC_SSUBLBT, UNSPEC_SSUBLT)
4527 (UNSPEC_SSUBLTB, UNSPEC_SSUBWB, UNSPEC_SSUBWT, UNSPEC_SUBHNB)
4528 (UNSPEC_SUBHNT, UNSPEC_TBL2, UNSPEC_UABDLB, UNSPEC_UABDLT)
4529 (UNSPEC_UADDLB, UNSPEC_UADDLT, UNSPEC_UADDWB, UNSPEC_UADDWT)
4530 (UNSPEC_UMAXP, UNSPEC_UMINP, UNSPEC_UQRSHRNB, UNSPEC_UQRSHRNT)
4531 (UNSPEC_UQSHRNB, UNSPEC_UQSHRNT, UNSPEC_UQXTNB, UNSPEC_UQXTNT)
4532 (UNSPEC_USHLLB, UNSPEC_USHLLT, UNSPEC_USUBLB, UNSPEC_USUBLT)
4533 (UNSPEC_USUBWB, UNSPEC_USUBWT): New unspecs.
4534 (UNSPEC_SMULLB, UNSPEC_SMULLT, UNSPEC_UMULLB, UNSPEC_UMULLT)
4535 (UNSPEC_SMULHS, UNSPEC_SMULHRS, UNSPEC_UMULHS, UNSPEC_UMULHRS)
4536 (UNSPEC_RSHRNB, UNSPEC_RSHRNT, UNSPEC_SHRNB, UNSPEC_SHRNT): Move
4538 (VNARROW, Ventype): New mode attributes.
4539 (Vewtype): Handle VNx2DI. Fix typo in comment.
4540 (VDOUBLE): New mode attribute.
4541 (sve_lane_con): Handle VNx8HI.
4542 (SVE_INT_UNARY): Include ss_abs and ss_neg for TARGET_SVE2.
4543 (SVE_INT_BINARY): Likewise ss_plus, us_plus, ss_minus and us_minus.
4544 (sve_int_op, sve_int_op_rev): Handle the above codes.
4545 (sve_pred_int_rhs2_operand): Likewise.
4546 (MULLBT, SHRNB, SHRNT): Delete.
4547 (SVE_INT_SHIFT_IMM): New int iterator.
4548 (SVE_WHILE): Add UNSPEC_WHILEGE, UNSPEC_WHILEGT, UNSPEC_WHILEHI
4549 and UNSPEC_WHILEHS for TARGET_SVE2.
4550 (SVE2_U32_UNARY, SVE2_INT_UNARY_NARROWB, SVE2_INT_UNARY_NARROWT)
4551 (SVE2_INT_BINARY, SVE2_INT_BINARY_LANE, SVE2_INT_BINARY_LONG)
4552 (SVE2_INT_BINARY_LONG_LANE, SVE2_INT_BINARY_NARROWB)
4553 (SVE2_INT_BINARY_NARROWT, SVE2_INT_BINARY_PAIR, SVE2_FP_BINARY_PAIR)
4554 (SVE2_INT_BINARY_PAIR_LONG, SVE2_INT_BINARY_WIDE): New int iterators.
4555 (SVE2_INT_SHIFT_IMM_LONG, SVE2_INT_SHIFT_IMM_NARROWB): Likewise.
4556 (SVE2_INT_SHIFT_IMM_NARROWT, SVE2_INT_SHIFT_INSERT, SVE2_INT_CADD)
4557 (SVE2_INT_BITPERM, SVE2_INT_TERNARY, SVE2_INT_TERNARY_LANE): Likewise.
4558 (SVE2_FP_TERNARY_LONG, SVE2_FP_TERNARY_LONG_LANE, SVE2_INT_CMLA)
4559 (SVE2_INT_CDOT, SVE2_INT_ADD_BINARY_LONG, SVE2_INT_QADD_BINARY_LONG)
4560 (SVE2_INT_SUB_BINARY_LONG, SVE2_INT_QSUB_BINARY_LONG): Likewise.
4561 (SVE2_INT_ADD_BINARY_LONG_LANE, SVE2_INT_QADD_BINARY_LONG_LANE)
4562 (SVE2_INT_SUB_BINARY_LONG_LANE, SVE2_INT_QSUB_BINARY_LONG_LANE)
4563 (SVE2_COND_INT_UNARY_FP, SVE2_COND_FP_UNARY_LONG): Likewise.
4564 (SVE2_COND_FP_UNARY_NARROWB, SVE2_COND_INT_BINARY): Likewise.
4565 (SVE2_COND_INT_BINARY_NOREV, SVE2_COND_INT_BINARY_REV): Likewise.
4566 (SVE2_COND_INT_SHIFT, SVE2_MATCH, SVE2_PMULL): Likewise.
4567 (optab): Handle the new unspecs.
4568 (su, r): Remove entries for UNSPEC_SHRNB, UNSPEC_SHRNT, UNSPEC_RSHRNB
4570 (lr): Handle the new unspecs.
4572 (cmp_op, while_optab_cmp, sve_int_op): Handle the new unspecs.
4573 (sve_int_op_rev, sve_int_add_op, sve_int_qadd_op, sve_int_sub_op)
4574 (sve_int_qsub_op): New int attributes.
4575 (sve_fp_op, rot): Handle the new unspecs.
4576 * config/aarch64/aarch64-sve-builtins.h
4577 (function_resolver::require_matching_pointer_type): Declare.
4578 (function_resolver::resolve_unary): Add an optional boolean argument.
4579 (function_resolver::finish_opt_n_resolution): Add an optional
4580 type_suffix_index argument.
4581 (gimple_folder::redirect_call): Declare.
4582 (gimple_expander::prepare_gather_address_operands): Add an optional
4584 * config/aarch64/aarch64-sve-builtins.cc: Include
4585 aarch64-sve-builtins-sve2.h.
4586 (TYPES_b_unsigned, TYPES_b_integer, TYPES_bh_integer): New macros.
4587 (TYPES_bs_unsigned, TYPES_hs_signed, TYPES_hs_integer): Likewise.
4588 (TYPES_hd_unsigned, TYPES_hsd_signed): Likewise.
4589 (TYPES_hsd_integer): Use TYPES_hsd_signed.
4590 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): New macros.
4591 (TYPES_s_unsigned): Likewise.
4592 (TYPES_s_integer): Use TYPES_s_unsigned.
4593 (TYPES_sd_signed, TYPES_sd_unsigned): New macros.
4594 (TYPES_sd_integer): Use them.
4595 (TYPES_d_unsigned): New macro.
4596 (TYPES_d_integer): Use it.
4597 (TYPES_d_data, TYPES_cvt_long, TYPES_cvt_narrow_s): New macros.
4598 (TYPES_cvt_narrow): Likewise.
4599 (DEF_SVE_TYPES_ARRAY): Include the new types macros above.
4600 (preds_mx): New variable.
4601 (function_builder::add_overloaded_function): Allow the new feature
4602 set to be more restrictive than the original one.
4603 (function_resolver::infer_pointer_type): Remove qualifiers from
4604 the pointer type before printing it.
4605 (function_resolver::require_matching_pointer_type): New function.
4606 (function_resolver::resolve_sv_displacement): Handle functions
4607 that don't support 32-bit vector indices or svint32_t vector offsets.
4608 (function_resolver::finish_opt_n_resolution): Take the inferred type
4609 as a separate argument.
4610 (function_resolver::resolve_unary): Optionally treat all forms in
4611 the same way as normal merging functions.
4612 (gimple_folder::redirect_call): New function.
4613 (function_expander::prepare_gather_address_operands): Add an argument
4614 that says whether scaled forms are available. If they aren't,
4615 handle scaling of vector indices and don't add the extension and
4617 (function_expander::map_to_unspecs): If aarch64_sve isn't available,
4618 fall back to using cond_* instead.
4619 * config/aarch64/aarch64-sve-builtins-functions.h (rtx_code_function):
4620 Split out the member variables into...
4621 (rtx_code_function_base): ...this new base class.
4622 (rtx_code_function_rotated): Inherit rtx_code_function_base.
4623 (unspec_based_function): Split out the member variables into...
4624 (unspec_based_function_base): ...this new base class.
4625 (unspec_based_function_rotated): Inherit unspec_based_function_base.
4626 (unspec_based_function_exact_insn): New class.
4627 (unspec_based_add_function, unspec_based_add_lane_function)
4628 (unspec_based_lane_function, unspec_based_pred_function)
4629 (unspec_based_qadd_function, unspec_based_qadd_lane_function)
4630 (unspec_based_qsub_function, unspec_based_qsub_lane_function)
4631 (unspec_based_sub_function, unspec_based_sub_lane_function): New
4633 (unspec_based_fused_function): New class.
4634 (unspec_based_mla_function, unspec_based_mls_function): New typedefs.
4635 (unspec_based_fused_lane_function): New class.
4636 (unspec_based_mla_lane_function, unspec_based_mls_lane_function): New
4638 (CODE_FOR_MODE1): New macro.
4639 (fixed_insn_function): New class.
4640 (while_comparison): Likewise.
4641 * config/aarch64/aarch64-sve-builtins-shapes.h (binary_long_lane)
4642 (binary_long_opt_n, binary_narrowb_opt_n, binary_narrowt_opt_n)
4643 (binary_to_uint, binary_wide, binary_wide_opt_n, compare, compare_ptr)
4644 (load_ext_gather_index_restricted, load_ext_gather_offset_restricted)
4645 (load_gather_sv_restricted, shift_left_imm_long): Declare.
4646 (shift_left_imm_to_uint, shift_right_imm_narrowb): Likewise.
4647 (shift_right_imm_narrowt, shift_right_imm_narrowb_to_uint): Likewise.
4648 (shift_right_imm_narrowt_to_uint, store_scatter_index_restricted)
4649 (store_scatter_offset_restricted, tbl_tuple, ternary_long_lane)
4650 (ternary_long_opt_n, ternary_qq_lane_rotate, ternary_qq_rotate)
4651 (ternary_shift_left_imm, ternary_shift_right_imm, ternary_uint)
4652 (unary_convert_narrowt, unary_long, unary_narrowb, unary_narrowt)
4653 (unary_narrowb_to_uint, unary_narrowt_to_uint, unary_to_int): Likewise.
4654 * config/aarch64/aarch64-sve-builtins-shapes.cc (apply_predication):
4655 Also add an initial argument for unary_convert_narrowt, regardless
4656 of the predication type.
4657 (build_32_64): Allow loads and stores to specify MODE_none.
4658 (build_sv_index64, build_sv_uint_offset): New functions.
4659 (long_type_suffix): New function.
4660 (binary_imm_narrowb_base, binary_imm_narrowt_base): New classes.
4661 (binary_imm_long_base, load_gather_sv_base): Likewise.
4662 (shift_right_imm_narrow_wrapper, ternary_shift_imm_base): Likewise.
4663 (ternary_resize2_opt_n_base, ternary_resize2_lane_base): Likewise.
4664 (unary_narrowb_base, unary_narrowt_base): Likewise.
4665 (binary_long_lane_def, binary_long_lane): New shape.
4666 (binary_long_opt_n_def, binary_long_opt_n): Likewise.
4667 (binary_narrowb_opt_n_def, binary_narrowb_opt_n): Likewise.
4668 (binary_narrowt_opt_n_def, binary_narrowt_opt_n): Likewise.
4669 (binary_to_uint_def, binary_to_uint): Likewise.
4670 (binary_wide_def, binary_wide): Likewise.
4671 (binary_wide_opt_n_def, binary_wide_opt_n): Likewise.
4672 (compare_def, compare): Likewise.
4673 (compare_ptr_def, compare_ptr): Likewise.
4674 (load_ext_gather_index_restricted_def,
4675 load_ext_gather_index_restricted): Likewise.
4676 (load_ext_gather_offset_restricted_def,
4677 load_ext_gather_offset_restricted): Likewise.
4678 (load_gather_sv_def): Inherit from load_gather_sv_base.
4679 (load_gather_sv_restricted_def, load_gather_sv_restricted): New shape.
4680 (shift_left_imm_def, shift_left_imm): Likewise.
4681 (shift_left_imm_long_def, shift_left_imm_long): Likewise.
4682 (shift_left_imm_to_uint_def, shift_left_imm_to_uint): Likewise.
4683 (store_scatter_index_restricted_def,
4684 store_scatter_index_restricted): Likewise.
4685 (store_scatter_offset_restricted_def,
4686 store_scatter_offset_restricted): Likewise.
4687 (tbl_tuple_def, tbl_tuple): Likewise.
4688 (ternary_long_lane_def, ternary_long_lane): Likewise.
4689 (ternary_long_opt_n_def, ternary_long_opt_n): Likewise.
4690 (ternary_qq_lane_def): Inherit from ternary_resize2_lane_base.
4691 (ternary_qq_lane_rotate_def, ternary_qq_lane_rotate): New shape
4692 (ternary_qq_opt_n_def): Inherit from ternary_resize2_opt_n_base.
4693 (ternary_qq_rotate_def, ternary_qq_rotate): New shape.
4694 (ternary_shift_left_imm_def, ternary_shift_left_imm): Likewise.
4695 (ternary_shift_right_imm_def, ternary_shift_right_imm): Likewise.
4696 (ternary_uint_def, ternary_uint): Likewise.
4697 (unary_convert): Fix typo in comment.
4698 (unary_convert_narrowt_def, unary_convert_narrowt): New shape.
4699 (unary_long_def, unary_long): Likewise.
4700 (unary_narrowb_def, unary_narrowb): Likewise.
4701 (unary_narrowt_def, unary_narrowt): Likewise.
4702 (unary_narrowb_to_uint_def, unary_narrowb_to_uint): Likewise.
4703 (unary_narrowt_to_uint_def, unary_narrowt_to_uint): Likewise.
4704 (unary_to_int_def, unary_to_int): Likewise.
4705 * config/aarch64/aarch64-sve-builtins-base.cc (unspec_cmla)
4706 (unspec_fcmla, unspec_cond_fcmla, expand_mla_mls_lane): New functions.
4707 (svasrd_impl): Delete.
4708 (svcadd_impl::expand): Handle integer operations too.
4709 (svcmla_impl::expand, svcmla_lane::expand): Likewise, using the
4710 new functions to derive the unspec numbers.
4711 (svmla_svmls_lane_impl): Replace with...
4712 (svmla_lane_impl, svmls_lane_impl): ...these new classes. Handle
4713 integer operations too.
4714 (svwhile_impl): Rename to...
4715 (svwhilelx_impl): ...this and inherit from while_comparison.
4716 (svasrd): Use unspec_based_function.
4717 (svmla_lane): Use svmla_lane_impl.
4718 (svmls_lane): Use svmls_lane_impl.
4719 (svrecpe, svrsqrte): Handle unsigned integer operations too.
4720 (svwhilele, svwhilelt): Use svwhilelx_impl.
4721 * config/aarch64/aarch64-sve-builtins-sve2.h: New file.
4722 * config/aarch64/aarch64-sve-builtins-sve2.cc: Likewise.
4723 * config/aarch64/aarch64-sve-builtins-sve2.def: Likewise.
4724 * config/aarch64/aarch64-sve-builtins.def: Include
4725 aarch64-sve-builtins-sve2.def.
4727 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4729 * config/aarch64/aarch64-protos.h (aarch64_sve_arith_immediate_p)
4730 (aarch64_sve_sqadd_sqsub_immediate_p): Add a machine_mode argument.
4731 * config/aarch64/aarch64.c (aarch64_sve_arith_immediate_p)
4732 (aarch64_sve_sqadd_sqsub_immediate_p): Likewise. Handle scalar
4733 immediates as well as vector ones.
4734 * config/aarch64/predicates.md (aarch64_sve_arith_immediate)
4735 (aarch64_sve_sub_arith_immediate, aarch64_sve_qadd_immediate)
4736 (aarch64_sve_qsub_immediate): Update calls accordingly.
4738 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4740 * config/aarch64/aarch64-sve2.md: Add banner comments.
4741 (<su>mulh<r>s<mode>3): Move further up file.
4742 (<su>mull<bt><Vwide>, <r>shrnb<mode>, <r>shrnt<mode>)
4743 (*aarch64_sve2_sra<mode>): Move further down file.
4744 * config/aarch64/t-aarch64 (s-check-sve-md): Check aarch64-sve2.md too.
4746 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4748 * config/aarch64/iterators.md (SVE_WHILE): Add UNSPEC_WHILERW
4750 (while_optab_cmp): Handle them.
4751 * config/aarch64/aarch64-sve.md
4752 (*while_<while_optab_cmp><GPI:mode><PRED_ALL:mode>_ptest): Make public
4753 and add a "@" marker.
4754 * config/aarch64/aarch64-sve2.md (check_<raw_war>_ptrs<mode>): Use it
4755 instead of gen_aarch64_sve2_while_ptest.
4756 (@aarch64_sve2_while<cmp_op><GPI:mode><PRED_ALL:mode>_ptest): Delete.
4758 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4760 * config/aarch64/aarch64.md (UNSPEC_WHILE_LE): Rename to...
4761 (UNSPEC_WHILELE): ...this.
4762 (UNSPEC_WHILE_LO): Rename to...
4763 (UNSPEC_WHILELO): ...this.
4764 (UNSPEC_WHILE_LS): Rename to...
4765 (UNSPEC_WHILELS): ...this.
4766 (UNSPEC_WHILE_LT): Rename to...
4767 (UNSPEC_WHILELT): ...this.
4768 * config/aarch64/iterators.md (SVE_WHILE): Update accordingly.
4769 (cmp_op, while_optab_cmp): Likewise.
4770 * config/aarch64/aarch64.c (aarch64_sve_move_pred_via_while): Likewise.
4771 * config/aarch64/aarch64-sve-builtins-base.cc (svwhilele): Likewise.
4772 (svwhilelt): Likewise.
4774 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4776 * config/aarch64/aarch64-sve-builtins-shapes.h (unary_count): Delete.
4777 (unary_to_uint): Define.
4778 * config/aarch64/aarch64-sve-builtins-shapes.cc (unary_count_def)
4779 (unary_count): Rename to...
4780 (unary_to_uint_def, unary_to_uint): ...this.
4781 * config/aarch64/aarch64-sve-builtins-base.def: Update accordingly.
4783 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4785 * config/aarch64/aarch64-sve-builtins-functions.h
4786 (code_for_mode_function): New class.
4787 (CODE_FOR_MODE0, QUIET_CODE_FOR_MODE0): New macros.
4788 * config/aarch64/aarch64-sve-builtins-base.cc (svcompact_impl)
4789 (svext_impl, svmul_lane_impl, svsplice_impl, svtmad_impl): Delete.
4790 (svcompact, svext, svsplice): Use QUIET_CODE_FOR_MODE0.
4791 (svmul_lane, svtmad): Use CODE_FOR_MODE0.
4793 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4795 * config/aarch64/iterators.md (addsub): New code attribute.
4796 * config/aarch64/aarch64-simd.md (aarch64_<su_optab><optab><mode>):
4798 (aarch64_<su_optab>q<addsub><mode>): ...this, making the same change
4799 in the asm string and attributes. Fix indentation.
4800 * config/aarch64/aarch64-sve.md (@aarch64_<su_optab><optab><mode>):
4802 (@aarch64_sve_<optab><mode>): ...this.
4803 * config/aarch64/aarch64-sve-builtins.h
4804 (function_expander::expand_signed_unpred_op): Delete.
4805 * config/aarch64/aarch64-sve-builtins.cc
4806 (function_expander::expand_signed_unpred_op): Likewise.
4807 (function_expander::map_to_rtx_codes): If the optab isn't defined,
4808 try using code_for_aarch64_sve instead.
4809 * config/aarch64/aarch64-sve-builtins-base.cc (svqadd_impl): Delete.
4810 (svqsub_impl): Likewise.
4811 (svqadd, svqsub): Use rtx_code_function instead.
4813 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4815 * config/aarch64/iterators.md (SRHSUB, URHSUB): Delete.
4816 (HADDSUB, sur, addsub): Remove them.
4818 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4820 * tree-nrv.c (pass_return_slot::execute): Handle all internal
4821 functions the same way, rather than singling out those that
4822 aren't mapped directly to optabs.
4824 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4826 * target.def (compatible_vector_types_p): New target hook.
4827 * hooks.h (hook_bool_const_tree_const_tree_true): Declare.
4828 * hooks.c (hook_bool_const_tree_const_tree_true): New function.
4829 * doc/tm.texi.in (TARGET_COMPATIBLE_VECTOR_TYPES_P): New hook.
4830 * doc/tm.texi: Regenerate.
4831 * gimple-expr.c: Include target.h.
4832 (useless_type_conversion_p): Use targetm.compatible_vector_types_p.
4833 * config/aarch64/aarch64.c (aarch64_compatible_vector_types_p): New
4835 (TARGET_COMPATIBLE_VECTOR_TYPES_P): Define.
4836 * config/aarch64/aarch64-sve-builtins.cc (gimple_folder::convert_pred):
4837 Use the original predicate if it already has a suitable type.
4839 2020-01-09 Martin Jambor <mjambor@suse.cz>
4841 * cgraph.h (cgraph_edge): Make remove, set_call_stmt, make_direct,
4842 resolve_speculation and redirect_call_stmt_to_callee static. Change
4843 return type of set_call_stmt to cgraph_edge *.
4844 * auto-profile.c (afdo_indirect_call): Adjust call to
4845 redirect_call_stmt_to_callee.
4846 * cgraph.c (cgraph_edge::set_call_stmt): Make return cgraph-edge *,
4847 make the this pointer explicit, adjust self-recursive calls and the
4848 call top make_direct. Return the resulting edge.
4849 (cgraph_edge::remove): Make this pointer explicit.
4850 (cgraph_edge::resolve_speculation): Likewise, adjust call to remove.
4851 (cgraph_edge::make_direct): Likewise, adjust call to
4852 resolve_speculation.
4853 (cgraph_edge::redirect_call_stmt_to_callee): Likewise, also adjust
4854 call to set_call_stmt.
4855 (cgraph_update_edges_for_call_stmt_node): Update call to
4856 set_call_stmt and remove.
4857 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
4858 Renamed edge to master_edge. Adjusted calls to set_call_stmt.
4859 (cgraph_node::create_edge_including_clones): Moved "first" definition
4860 of edge to the block where it was used. Adjusted calls to
4862 (cgraph_node::remove_symbol_and_inline_clones): Adjust call to
4863 cgraph_edge::remove.
4864 * cgraphunit.c (walk_polymorphic_call_targets): Adjusted calls to
4865 make_direct and redirect_call_stmt_to_callee.
4866 * ipa-fnsummary.c (redirect_to_unreachable): Adjust calls to
4867 resolve_speculation and make_direct.
4868 * ipa-inline-transform.c (inline_transform): Adjust call to
4869 redirect_call_stmt_to_callee.
4870 (check_speculations_1):: Adjust call to resolve_speculation.
4871 * ipa-inline.c (resolve_noninline_speculation): Adjust call to
4872 resolve-speculation.
4873 (inline_small_functions): Adjust call to resolve_speculation.
4874 (ipa_inline): Likewise.
4875 * ipa-prop.c (ipa_make_edge_direct_to_target): Adjust call to
4877 * ipa-visibility.c (function_and_variable_visibility): Make iteration
4878 safe with regards to edge removal, adjust calls to
4879 redirect_call_stmt_to_callee.
4880 * ipa.c (walk_polymorphic_call_targets): Adjust calls to make_direct
4881 and redirect_call_stmt_to_callee.
4882 * multiple_target.c (create_dispatcher_calls): Adjust call to
4883 redirect_call_stmt_to_callee
4884 (redirect_to_specific_clone): Likewise.
4885 * tree-cfgcleanup.c (delete_unreachable_blocks_update_callgraph):
4886 Adjust calls to cgraph_edge::remove.
4887 * tree-inline.c (copy_bb): Adjust call to set_call_stmt.
4888 (redirect_all_calls): Adjust call to redirect_call_stmt_to_callee.
4889 (expand_call_inline): Adjust call to cgraph_edge::remove.
4891 2020-01-09 Martin Liska <mliska@suse.cz>
4893 * params.opt: Set Optimization for
4894 param_max_speculative_devirt_maydefs.
4896 2020-01-09 Martin Sebor <msebor@redhat.com>
4900 * builtins.c (compute_objsize): Avoid handling MEM_REFs of vector type.
4902 2020-01-09 Martin Liska <mliska@suse.cz>
4904 * auto-profile.c (auto_profile): Use opt_for_fn
4906 * ipa-cp.c (ipcp_lattice::add_value): Likewise.
4907 (propagate_vals_across_arith_jfunc): Likewise.
4908 (hint_time_bonus): Likewise.
4909 (incorporate_penalties): Likewise.
4910 (good_cloning_opportunity_p): Likewise.
4911 (perform_estimation_of_a_value): Likewise.
4912 (estimate_local_effects): Likewise.
4913 (ipcp_propagate_stage): Likewise.
4914 * ipa-fnsummary.c (decompose_param_expr): Likewise.
4915 (set_switch_stmt_execution_predicate): Likewise.
4916 (analyze_function_body): Likewise.
4917 * ipa-inline-analysis.c (offline_size): Likewise.
4918 * ipa-inline.c (early_inliner): Likewise.
4919 * ipa-prop.c (ipa_analyze_node): Likewise.
4920 (ipcp_transform_function): Likewise.
4921 * ipa-sra.c (process_scan_results): Likewise.
4922 (ipa_sra_summarize_function): Likewise.
4923 * params.opt: Rename ipcp-unit-growth to
4924 ipa-cp-unit-growth. Add Optimization for various
4925 IPA-related parameters.
4927 2020-01-09 Richard Biener <rguenther@suse.de>
4930 * gimplify.c (gimplify_expr): Deal with NOP definitions.
4932 2020-01-09 Richard Biener <rguenther@suse.de>
4934 PR tree-optimization/93040
4935 * gimple-ssa-store-merging.c (find_bswap_or_nop): Raise search limit.
4937 2020-01-09 Georg-Johann Lay <avr@gjlay.de>
4939 * common/config/avr/avr-common.c (avr_option_optimization_table)
4940 [OPT_LEVELS_1_PLUS]: Set -fsplit-wide-types-early.
4942 2020-01-09 Martin Liska <mliska@suse.cz>
4944 * cgraphclones.c (symbol_table::materialize_all_clones):
4945 Use cgraph_node::dump_name.
4947 2020-01-09 Jakub Jelinek <jakub@redhat.com>
4950 * config/riscv/riscv.c (riscv_print_operand_reloc): Use
4951 output_operand_lossage instead of gcc_unreachable.
4952 * doc/md.texi (riscv f constraint): Fix typo.
4955 * config/i386/i386.md (subv<mode>4): Use SWIDWI iterator instead of
4956 SWI. Use <general_hilo_operand> instead of <general_operand>. Use
4957 CONST_SCALAR_INT_P instead of CONST_INT_P.
4958 (*subv<mode>4_1): Rename to ...
4959 (subv<mode>4_1): ... this.
4960 (*subv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
4961 define_insn_and_split patterns.
4962 (*subv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
4965 2020-01-08 David Malcolm <dmalcolm@redhat.com>
4967 * vec.c (class selftest::count_dtor): New class.
4968 (selftest::test_auto_delete_vec): New test.
4969 (selftest::vec_c_tests): Call it.
4970 * vec.h (class auto_delete_vec): New class template.
4971 (auto_delete_vec<T>::~auto_delete_vec): New dtor.
4973 2020-01-08 David Malcolm <dmalcolm@redhat.com>
4975 * sbitmap.h (auto_sbitmap): Add operator const_sbitmap.
4977 2020-01-08 Jim Wilson <jimw@sifive.com>
4979 * config/riscv/riscv.c (riscv_legitimize_tls_address): Ifdef out
4980 use of TLS_MODEL_LOCAL_EXEC when not pic.
4982 2020-01-08 David Malcolm <dmalcolm@redhat.com>
4984 * hash-map-tests.c (selftest::test_map_of_strings_to_int): Fix
4987 2020-01-08 Jakub Jelinek <jakub@redhat.com>
4990 * config/i386/i386.md (*stack_protect_set_2_<mode> peephole2,
4991 *stack_protect_set_3 peephole2): Also check that the second
4992 insns source is general_operand.
4995 * config/i386/i386.md (addcarry<mode>_0): Use nonimmediate_operand
4996 predicate for output operand instead of register_operand.
4997 (addcarry<mode>, addcarry<mode>_1): Likewise. Add alternative with
4998 memory destination and non-memory operands[2].
5000 2020-01-08 Martin Liska <mliska@suse.cz>
5002 * cgraph.c (cgraph_node::dump): Use ::dump_name or
5003 ::dump_asm_name instead of (::name or ::asm_name).
5004 * cgraphclones.c (symbol_table::materialize_all_clones): Likewise.
5005 * cgraphunit.c (walk_polymorphic_call_targets): Likewise.
5006 (analyze_functions): Likewise.
5007 (expand_all_functions): Likewise.
5008 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
5009 (propagate_bits_across_jump_function): Likewise.
5010 (dump_profile_updates): Likewise.
5011 (ipcp_store_bits_results): Likewise.
5012 (ipcp_store_vr_results): Likewise.
5013 * ipa-devirt.c (dump_targets): Likewise.
5014 * ipa-fnsummary.c (analyze_function_body): Likewise.
5015 * ipa-hsa.c (check_warn_node_versionable): Likewise.
5016 (process_hsa_functions): Likewise.
5017 * ipa-icf.c (sem_item_optimizer::merge_classes): Likewise.
5018 (set_alias_uids): Likewise.
5019 * ipa-inline-transform.c (save_inline_function_body): Likewise.
5020 * ipa-inline.c (recursive_inlining): Likewise.
5021 (inline_to_all_callers_1): Likewise.
5022 (ipa_inline): Likewise.
5023 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
5024 (ipa_propagate_frequency): Likewise.
5025 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
5026 (remove_described_reference): Likewise.
5027 * ipa-pure-const.c (worse_state): Likewise.
5028 (check_retval_uses): Likewise.
5029 (analyze_function): Likewise.
5030 (propagate_pure_const): Likewise.
5031 (propagate_nothrow): Likewise.
5032 (dump_malloc_lattice): Likewise.
5033 (propagate_malloc): Likewise.
5034 (pass_local_pure_const::execute): Likewise.
5035 * ipa-visibility.c (optimize_weakref): Likewise.
5036 (function_and_variable_visibility): Likewise.
5037 * ipa.c (symbol_table::remove_unreachable_nodes): Likewise.
5038 (ipa_discover_variable_flags): Likewise.
5039 * lto-streamer-out.c (output_function): Likewise.
5040 (output_constructor): Likewise.
5041 * tree-inline.c (copy_bb): Likewise.
5042 * tree-ssa-structalias.c (ipa_pta_execute): Likewise.
5043 * varpool.c (symbol_table::remove_unreferenced_decls): Likewise.
5045 2020-01-08 Richard Biener <rguenther@suse.de>
5048 * tree-eh.c (sink_clobbers): Update virtual operands for
5049 the first and last stmt only. Add a dry-run capability.
5050 (pass_lower_eh_dispatch::execute): Perform clobber sinking
5051 after CFG manipulations and in RPO order to catch all
5052 secondary opportunities reliably.
5054 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
5057 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
5059 2019-01-08 Richard Biener <rguenther@suse.de>
5062 * gimple-fold.c (rewrite_to_defined_overflow): Mark stmt modified.
5063 * tree-ssa-loop-im.c (move_computations_worker): Properly adjust
5064 virtual operand, also updating SSA use.
5065 * gimple-loop-interchange.cc (loop_cand::undo_simple_reduction):
5066 Update stmt after resetting virtual operand.
5067 (tree_loop_interchange::move_code_to_inner_loop): Likewise.
5068 * gimple-iterator.c (gsi_remove): When not removing the stmt
5069 permanently do not delink immediate uses or mark the stmt modified.
5071 2020-01-08 Martin Liska <mliska@suse.cz>
5073 * ipa-fnsummary.c (dump_ipa_call_summary): Use symtab_node::dump_name.
5074 (ipa_call_context::estimate_size_and_time): Likewise.
5075 (inline_analyze_function): Likewise.
5077 2020-01-08 Martin Liska <mliska@suse.cz>
5079 * cgraph.c (cgraph_node::dump): Use systematically
5082 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
5084 Add -nodevicespecs option for avr.
5087 * config/avr/avr.opt (-nodevicespecs): New driver option.
5088 * config/avr/driver-avr.c (avr_devicespecs_file): Only issue
5089 "-specs=device-specs/..." if that option is not set.
5090 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
5092 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
5094 Implement 64-bit double functions for avr.
5097 * config.gcc (tm_defines) [target=avr]: Support --with-libf7,
5098 --with-double-comparison.
5099 * doc/install.texi: Document them.
5100 * config/avr/avr-c.c (avr_cpu_cpp_builtins)
5101 <WITH_LIBF7_LIBGCC, WITH_LIBF7_MATH, WITH_LIBF7_MATH_SYMBOLS>
5102 <WITH_DOUBLE_COMPARISON>: New built-in defines.
5103 * doc/invoke.texi (AVR Built-in Macros): Document them.
5104 * config/avr/avr-protos.h (avr_float_lib_compare_returns_bool): New.
5105 * config/avr/avr.c (avr_float_lib_compare_returns_bool): New function.
5106 * config/avr/avr.h (FLOAT_LIB_COMPARE_RETURNS_BOOL): New macro.
5108 2020-01-08 Richard Earnshaw <rearnsha@arm.com>
5111 * config/arm/t-multilib (MULTILIB_MATCHES): Add rules to match
5112 armv7-a{+mp,+sec,+mp+sec} to appropriate armv7 multilib variants
5113 when only building rm-profile multilibs.
5115 2020-01-08 Feng Xue <fxue@os.amperecomputing.com>
5118 * ipa-cp.c (self_recursively_generated_p): Find matched aggregate
5119 lattice for a value to check.
5120 (propagate_vals_across_arith_jfunc): Add an assertion to ensure
5121 finite propagation in self-recursive scc.
5123 2020-01-08 Luo Xiong Hu <luoxhu@linux.ibm.com>
5125 * ipa-inline.c (caller_growth_limits): Restore the AND.
5127 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
5129 * config/gcn/gcn-valu.md (VEC_1REG_INT_ALT): Delete iterator.
5130 (VEC_ALLREG_ALT): New iterator.
5131 (VEC_ALLREG_INT_MODE): New iterator.
5132 (VCMP_MODE): New iterator.
5133 (VCMP_MODE_INT): New iterator.
5134 (vec_cmpu<mode>di): Use VCMP_MODE_INT.
5135 (vec_cmp<u>v64qidi): New define_expand.
5136 (vec_cmp<mode>di_exec): Use VCMP_MODE.
5137 (vec_cmpu<mode>di_exec): New define_expand.
5138 (vec_cmp<u>v64qidi_exec): New define_expand.
5139 (vec_cmp<mode>di_dup): Use VCMP_MODE.
5140 (vec_cmp<mode>di_dup_exec): Use VCMP_MODE.
5141 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>): Rename ...
5142 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): ... to this.
5143 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>_exec): Rename ...
5144 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): ... to this.
5145 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>): Rename ...
5146 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): ... to this.
5147 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>_exec): Rename ...
5148 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): ... to
5150 * config/gcn/gcn.c (print_operand): Fix 8 and 16 bit suffixes.
5151 * config/gcn/gcn.md (expander): Add sign_extend and zero_extend.
5153 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
5155 * config/gcn/constraints.md (DA): Update description and match.
5157 (Db): New constraint.
5158 * config/gcn/gcn-protos.h (gcn_inline_constant64_p): Add second
5160 * config/gcn/gcn.c (gcn_inline_constant64_p): Add 'mixed' parameter.
5161 Implement 'Db' mixed immediate type.
5162 * config/gcn/gcn-valu.md (addcv64si3<exec_vcc>): Rework constraints.
5163 (addcv64si3_dup<exec_vcc>): Delete.
5164 (subcv64si3<exec_vcc>): Rework constraints.
5165 (addv64di3): Rework constraints.
5166 (addv64di3_exec): Rework constraints.
5167 (subv64di3): Rework constraints.
5168 (addv64di3_dup): Delete.
5169 (addv64di3_dup_exec): Delete.
5170 (addv64di3_zext): Rework constraints.
5171 (addv64di3_zext_exec): Rework constraints.
5172 (addv64di3_zext_dup): Rework constraints.
5173 (addv64di3_zext_dup_exec): Rework constraints.
5174 (addv64di3_zext_dup2): Rework constraints.
5175 (addv64di3_zext_dup2_exec): Rework constraints.
5176 (addv64di3_sext_dup2): Rework constraints.
5177 (addv64di3_sext_dup2_exec): Rework constraints.
5179 2020-01-07 Andre Vieira <andre.simoesdiasvieira@arm.com>
5181 * doc/sourcebuild.texi (arm_little_endian, arm_nothumb): Documented
5182 existing target checks.
5184 2020-01-07 Richard Biener <rguenther@suse.de>
5186 * doc/install.texi: Bump minimal supported MPC version.
5188 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
5190 * langhooks-def.h (lhd_simulate_enum_decl): Declare.
5191 (LANG_HOOKS_SIMULATE_ENUM_DECL): Use it.
5192 * langhooks.c: Include stor-layout.h.
5193 (lhd_simulate_enum_decl): New function.
5194 * config/aarch64/aarch64-sve-builtins.cc (init_builtins): Call
5195 handle_arm_sve_h for the LTO frontend.
5196 (register_vector_type): Cope with null returns from pushdecl.
5198 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
5200 * config/aarch64/aarch64-protos.h (aarch64_sve::svbool_type_p)
5201 (aarch64_sve::nvectors_if_data_type): Replace with...
5202 (aarch64_sve::builtin_type_p): ...this.
5203 * config/aarch64/aarch64-sve-builtins.cc: Include attribs.h.
5204 (find_vector_type): Delete.
5205 (add_sve_type_attribute): New function.
5206 (lookup_sve_type_attribute): Likewise.
5207 (register_builtin_types): Add an "SVE type" attribute to each type.
5208 (register_tuple_type): Likewise.
5209 (svbool_type_p, nvectors_if_data_type): Delete.
5210 (mangle_builtin_type): Use lookup_sve_type_attribute.
5211 (builtin_type_p): Likewise. Add an overload that returns the
5212 number of constituent vector and predicate registers.
5213 * config/aarch64/aarch64.c (aarch64_sve_argument_p): Delete.
5214 (aarch64_returns_value_in_sve_regs_p): Use aarch64_sve::builtin_type_p
5215 instead of aarch64_sve_argument_p.
5216 (aarch64_takes_arguments_in_sve_regs_p): Likewise.
5217 (aarch64_pass_by_reference): Likewise.
5218 (aarch64_function_value_1): Likewise.
5219 (aarch64_return_in_memory): Likewise.
5220 (aarch64_layout_arg): Likewise.
5222 2020-01-07 Jakub Jelinek <jakub@redhat.com>
5224 PR tree-optimization/93156
5225 * tree-ssa-ccp.c (bit_value_binop): For x * x note that the second
5226 least significant bit is always clear.
5228 PR tree-optimization/93118
5229 * match.pd ((x >> c) << c -> x & (-1<<c)): Add nop_convert?. Add new
5230 simplifier with two intermediate conversions.
5232 2020-01-07 Martin Liska <mliska@suse.cz>
5234 * params.opt: Add Optimization for various parameters.
5236 2020-01-07 Martin Liska <mliska@suse.cz>
5239 * doc/extend.texi: Explain cloning for target_clone
5242 2020-01-07 Martin Liska <mliska@suse.cz>
5244 PR tree-optimization/92860
5245 * common.opt: Make in Optimization option
5246 as it is affected by -O0, which is an Optimization
5248 * tree-inline.c (tree_inlinable_function_p):
5249 Use opt_for_fn for warn_inline.
5250 (expand_call_inline): Likewise.
5252 2020-01-07 Martin Liska <mliska@suse.cz>
5254 PR tree-optimization/92860
5255 * common.opt: Make flag_ree as optimization
5258 2020-01-07 Martin Liska <mliska@suse.cz>
5260 PR optimization/92860
5261 * params.opt: Mark param_min_crossjump_insns with Optimization
5264 2020-01-07 Luo Xiong Hu <luoxhu@linux.ibm.com>
5266 * ipa-inline-analysis.c (estimate_growth): Fix typo.
5267 * ipa-inline.c (caller_growth_limits): Use OR instead of AND.
5269 2020-01-06 Michael Meissner <meissner@linux.ibm.com>
5271 * config/rs6000/rs6000.c (hard_reg_and_mode_to_addr_mask): New
5272 helper function to return the valid addressing formats for a given
5273 hard register and mode.
5274 (rs6000_adjust_vec_address): Call hard_reg_and_mode_to_addr_mask.
5276 * config/rs6000/constraints.md (Q constraint): Update
5278 * doc/md.texi (RS/6000 constraints): Update 'Q' cosntraint
5281 * config/rs6000/vsx.md (vsx_extract_<mode>_var, VSX_D iterator):
5282 Use 'Q' for doing vector extract from memory.
5283 (vsx_extract_v4sf_var): Use 'Q' for doing vector extract from
5285 (vsx_extract_<mode>_var, VSX_EXTRACT_I iterator): Use 'Q' for
5286 doing vector extract from memory.
5287 (vsx_extract_<mode>_<VS_scalar>mode_var): Use 'Q' for doing vector
5288 extract from memory.
5290 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add support
5291 for the offset being 34-bits when -mcpu=future is used.
5293 2020-01-06 John David Anglin <danglin@gcc.gnu.org>
5295 * config/pa/pa.md: Revert change to use ordered_comparison_operator
5296 instead of cmpib_comparison_operator in cmpib patterns.
5297 * config/pa/predicates.md (cmpib_comparison_operator): Revert removal
5298 of cmpib_comparison_operator. Revise comment.
5300 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
5302 * tree-vect-slp.c (vect_build_slp_tree_1): Require all shifts
5303 in an IFN_DIV_POW2 node to be equal.
5305 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
5307 * tree-vect-stmts.c (vect_check_load_store_mask): Rename to...
5308 (vect_check_scalar_mask): ...this.
5309 (vectorizable_store, vectorizable_load): Update call accordingly.
5310 (vectorizable_call): Use vect_check_scalar_mask to check the mask
5311 argument in calls to conditional internal functions.
5313 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
5315 * config/gcn/gcn-valu.md (subv64di3): Use separate alternatives for
5316 '0' matching inputs.
5317 (subv64di3_exec): Likewise.
5319 2020-01-06 Bryan Stenson <bryan@siliconvortex.com>
5321 * config/mips/mips.c (vr4130_align_insns): Fix typo.
5322 * doc/md.texi (movstr): Likewise.
5324 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
5326 * config/gcn/gcn-valu.md (vec_extract<mode><scalar_mode>): Add early
5329 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
5331 * config/aarch64/t-aarch64 ($(srcdir)/config/aarch64/aarch64-tune.md):
5333 (s-aarch64-tune-md): ...this new stamp file. Pipe the new contents
5334 to a temporary file and use move-if-change to update the real
5335 file where necessary.
5337 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
5339 * config/aarch64/aarch64-sve.md (@aarch64_sel_dup<mode>): Use Upl
5340 rather than Upa for CPY /M.
5342 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
5344 * config/gcn/gcn.c (gcn_inline_constant_p): Allow 64 as an inline
5347 2020-01-06 Martin Liska <mliska@suse.cz>
5349 PR tree-optimization/92860
5350 * params.opt: Mark param_max_combine_insns with Optimization
5353 2020-01-05 Jakub Jelinek <jakub@redhat.com>
5356 * config/i386/i386.md (SWIDWI): New mode iterator.
5357 (DWI, dwi): Add TImode variants.
5358 (addv<mode>4): Use SWIDWI iterator instead of SWI. Use
5359 <general_hilo_operand> instead of <general_operand>. Use
5360 CONST_SCALAR_INT_P instead of CONST_INT_P.
5361 (*addv<mode>4_1): Rename to ...
5362 (addv<mode>4_1): ... this.
5363 (QWI): New mode attribute.
5364 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
5365 define_insn_and_split patterns.
5366 (*addv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
5368 (uaddv<mode>4): Use SWIDWI iterator instead of SWI. Use
5369 <general_hilo_operand> instead of <general_operand>.
5370 (*addcarry<mode>_1): New define_insn.
5371 (*add<dwi>3_doubleword_cc_overflow_1): New define_insn_and_split.
5373 2020-01-03 Konstantin Kharlamov <Hi-Angel@yandex.ru>
5375 * gdbinit.in (pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, pdd, pbs, pbm):
5376 Use "call" instead of "set".
5378 2020-01-03 Martin Jambor <mjambor@suse.cz>
5381 * ipa-cp.c (print_all_lattices): Skip functions without info.
5383 2020-01-03 Jakub Jelinek <jakub@redhat.com>
5386 * config/i386/i386-options.c (ix86_simd_clone_adjust): If
5387 TARGET_PREFER_AVX128, use prefer-vector-width=256 for 'c' and 'd'
5388 simd clones. If TARGET_PREFER_AVX256, use prefer-vector-width=512
5389 for 'e' simd clones.
5392 * config/i386/i386.opt (x_prefer_vector_width_type): Remove TargetSave
5394 (mprefer-vector-width=): Add Save.
5395 * config/i386/i386-options.c (ix86_target_string): Add PVW argument, print
5396 -mprefer-vector-width= if non-zero. Fix up -mfpmath= comment.
5397 (ix86_debug_options, ix86_function_specific_print): Adjust
5398 ix86_target_string callers.
5399 (ix86_valid_target_attribute_inner_p): Handle prefer-vector-width=.
5400 (ix86_valid_target_attribute_tree): Likewise.
5401 * config/i386/i386-options.h (ix86_target_string): Add PVW argument.
5402 * config/i386/i386-expand.c (ix86_expand_builtin): Adjust
5403 ix86_target_string caller.
5406 * config/i386/i386.md (abs<mode>2): Use expand_simple_binop instead of
5407 emitting ASHIFTRT, XOR and MINUS by hand. Use gen_int_mode with QImode
5408 instead of gen_int_shift_amount + convert_modes.
5410 PR rtl-optimization/93088
5411 * loop-iv.c (find_single_def_src): Punt after looking through
5412 128 reg copies for regs with single definitions. Move definitions
5415 2020-01-02 Dennis Zhang <dennis.zhang@arm.com>
5417 * config/arm/arm-c.c (arm_cpu_builtins): Define
5418 __ARM_FEATURE_MATMUL_INT8, __ARM_FEATURE_BF16_VECTOR_ARITHMETIC,
5419 __ARM_FEATURE_BF16_SCALAR_ARITHMETIC, and
5420 __ARM_BF16_FORMAT_ALTERNATIVE when enabled.
5421 * config/arm/arm-cpus.in (armv8_6, i8mm, bf16): New features.
5422 * config/arm/arm-tables.opt: Regenerated.
5423 * config/arm/arm.c (arm_option_reconfigure_globals): Initialize
5424 arm_arch_i8mm and arm_arch_bf16 when enabled.
5425 * config/arm/arm.h (TARGET_I8MM): New macro.
5426 (TARGET_BF16_FP, TARGET_BF16_SIMD): Likewise.
5427 * config/arm/t-aprofile: Add matching rules for -march=armv8.6-a.
5428 * config/arm/t-arm-elf (all_v8_archs): Add armv8.6-a.
5429 * config/arm/t-multilib: Add matching rules for -march=armv8.6-a.
5430 (v8_6_a_simd_variants): New.
5431 (v8_*_a_simd_variants): Add i8mm and bf16.
5432 * doc/invoke.texi (armv8.6-a, i8mm, bf16): Document new options.
5434 2020-01-02 Jakub Jelinek <jakub@redhat.com>
5437 * predict.c (compute_function_frequency): Don't call
5438 warn_function_cold on functions that already have cold attribute.
5440 2020-01-01 John David Anglin <danglin@gcc.gnu.org>
5443 * config/pa/pa.c (pa_elf_select_rtx_section): New. Put references to
5444 COMDAT group function labels in .data.rel.ro.local section.
5445 * config/pa/pa32-linux.h (TARGET_ASM_SELECT_RTX_SECTION): Define.
5448 * config/pa/pa.md (scc): Use ordered_comparison_operator instead of
5449 comparison_operator in B and S integer comparisons. Likewise, use
5450 ordered_comparison_operator instead of cmpib_comparison_operator in
5452 * config/pa/predicates.md (cmpib_comparison_operator): Remove.
5454 2020-01-01 Jakub Jelinek <jakub@redhat.com>
5456 Update copyright years.
5458 * gcc.c (process_command): Update copyright notice dates.
5459 * gcov-dump.c (print_version): Ditto.
5460 * gcov.c (print_version): Ditto.
5461 * gcov-tool.c (print_version): Ditto.
5462 * gengtype.c (create_file): Ditto.
5463 * doc/cpp.texi: Bump @copying's copyright year.
5464 * doc/cppinternals.texi: Ditto.
5465 * doc/gcc.texi: Ditto.
5466 * doc/gccint.texi: Ditto.
5467 * doc/gcov.texi: Ditto.
5468 * doc/install.texi: Ditto.
5469 * doc/invoke.texi: Ditto.
5471 2020-01-01 Jan Hubicka <hubicka@ucw.cz>
5473 * ipa.c (walk_polymorphic_call_targets): Fix updating of overall
5476 2020-01-01 Jakub Jelinek <jakub@redhat.com>
5478 PR tree-optimization/93098
5479 * match.pd (popcount): For shift amounts, use integer_onep
5480 or wi::to_widest () == cst instead of tree_to_uhwi () == cst
5481 tests. Make sure that precision is power of two larger than or equal
5482 to 16. Ensure shift is never negative. Use HOST_WIDE_INT_UC macro
5483 instead of ULL suffixed constants. Formatting fixes.
5485 Copyright (C) 2020 Free Software Foundation, Inc.
5487 Copying and distribution of this file, with or without modification,
5488 are permitted in any medium without royalty provided the copyright
5489 notice and this notice are preserved.