1 2020-03-11 Eric Botcazou <ebotcazou@adacore.com>
4 * tree.c (variably_modified_type_p) <RECORD_TYPE>: Recurse into fields
5 whose type is a qualified union.
7 2020-03-11 Jakub Jelinek <jakub@redhat.com>
10 * config/aarch64/aarch64.c (aarch64_add_offset_1): Use absu_hwi
11 instead of abs_hwi, change moffset type to unsigned HOST_WIDE_INT.
14 * value-prof.c (dump_histogram_value): Use abs_hwi instead of
16 (get_nth_most_common_value): Use abs_hwi instead of abs.
19 * dfp.c (decimal_to_binary): Only use decimal128ToString if from->cl
20 is rvc_normal, otherwise use real_to_decimal to print the number to
23 PR tree-optimization/94114
24 * tree-loop-distribution.c (generate_memset_builtin): Call
25 rewrite_to_non_trapping_overflow even on mem.
26 (generate_memcpy_builtin): Call rewrite_to_non_trapping_overflow even
29 2020-03-10 Jeff Law <law@redhat.com>
31 * config/bfin/bfin.md (movsi_insv): Add length attribute.
33 2020-03-10 Jiufu Guo <guojiufu@linux.ibm.com>
36 * gcc/config/rs6000/rs6000.c (rs6000_emit_p9_fp_minmax): Check
37 NAN and SIGNED_ZEROR for smax/smin.
39 2020-03-10 Will Schmidt <will_schmidt@vnet.ibm.com>
42 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Add
43 clause to handle P9V_BUILTIN_VEC_LXVL with const arguments.
45 2020-03-10 Roman Zhuykov <zhroma@ispras.ru>
47 * loop-iv.c (find_simple_exit): Make it static.
48 * cfgloop.h: Remove the corresponding prototype.
50 2020-03-10 Roman Zhuykov <zhroma@ispras.ru>
52 * ddg.c (create_ddg): Fix intendation.
53 (set_recurrence_length): Likewise.
54 (create_ddg_all_sccs): Likewise.
56 2020-03-10 Jakub Jelinek <jakub@redhat.com>
59 * config/i386/i386.md (*testqi_ext_3): Call ix86_match_ccmode with
60 CCZmode instead of CCNOmode if operands[2] has DImode and pos + len
63 2020-03-09 Jason Merrill <jason@redhat.com>
65 * gdbinit.in (pgs): Fix typo in documentation.
67 2020-03-09 Vladimir Makarov <vmakarov@redhat.com>
71 2020-02-28 Vladimir Makarov <vmakarov@redhat.com>
73 PR rtl-optimization/93564
74 * ira-color.c (assign_hard_reg): Prefer smaller hard regno when we
75 do not honor reg alloc order.
77 2020-03-09 Andrew Pinski <apinski@marvell.com>
80 * doc/extend.texi (x86 Operand Modifiers): Fix column
83 2020-03-09 Martin Liska <mliska@suse.cz>
86 * config/rs6000/rs6000.c (rs6000_option_override_internal):
87 Remove set of str_align_loops and str_align_jumps as these
88 should be set in previous 2 conditions in the function.
90 2020-03-09 Jakub Jelinek <jakub@redhat.com>
92 PR rtl-optimization/94045
93 * params.opt (-param=max-find-base-term-values=): New option.
94 * alias.c (find_base_term): Add cut-off for number of visited VALUEs
95 in a single toplevel find_base_term call.
97 2020-03-06 Wilco Dijkstra <wdijkstr@arm.com>
100 * config/aarch64/aarch64-builtins.c (TYPES_TERNOPU_LANE): Add define.
101 * config/aarch64/aarch64-simd.md
102 (aarch64_vec_<su>mult_lane<Qlane>): Add new insn for widening lane mul.
103 (aarch64_vec_<su>mlal_lane<Qlane>): Likewise.
104 * config/aarch64/aarch64-simd-builtins.def: Add intrinsics.
105 * config/aarch64/arm_neon.h:
106 (vmlal_lane_s16): Expand using intrinsics rather than inline asm.
107 (vmlal_lane_u16): Likewise.
108 (vmlal_lane_s32): Likewise.
109 (vmlal_lane_u32): Likewise.
110 (vmlal_laneq_s16): Likewise.
111 (vmlal_laneq_u16): Likewise.
112 (vmlal_laneq_s32): Likewise.
113 (vmlal_laneq_u32): Likewise.
114 (vmull_lane_s16): Likewise.
115 (vmull_lane_u16): Likewise.
116 (vmull_lane_s32): Likewise.
117 (vmull_lane_u32): Likewise.
118 (vmull_laneq_s16): Likewise.
119 (vmull_laneq_u16): Likewise.
120 (vmull_laneq_s32): Likewise.
121 (vmull_laneq_u32): Likewise.
122 * config/aarch64/iterators.md (Vcondtype): New iterator for lane mul.
125 2020-03-06 Wilco Dijkstra <wdijkstr@arm.com>
127 * aarch64/aarch64-simd.md (aarch64_mla_elt<mode>): Correct lane syntax.
128 (aarch64_mla_elt_<vswap_width_name><mode>): Likewise.
129 (aarch64_mls_elt<mode>): Likewise.
130 (aarch64_mls_elt_<vswap_width_name><mode>): Likewise.
131 (aarch64_fma4_elt<mode>): Likewise.
132 (aarch64_fma4_elt_<vswap_width_name><mode>): Likewise.
133 (aarch64_fma4_elt_to_64v2df): Likewise.
134 (aarch64_fnma4_elt<mode>): Likewise.
135 (aarch64_fnma4_elt_<vswap_width_name><mode>): Likewise.
136 (aarch64_fnma4_elt_to_64v2df): Likewise.
138 2020-03-06 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
140 * config/aarch64/aarch64-sve2.md (@aarch64_sve_<sve_int_op><mode>:
141 Specify movprfx attribute.
142 (@aarch64_sve_<sve_int_op>_lane_<mode>): Likewise.
144 2020-03-06 David Edelsohn <dje.gcc@gmail.com>
147 * config/rs6000/aix61.h (TARGET_NO_SUM_IN_TOC): Set to 1 for
149 (TARGET_NO_FP_IN_TOC): Same.
150 * config/rs6000/aix71.h: Same.
151 * config/rs6000/aix72.h: Same.
153 2020-03-06 Andrew Pinski <apinski@marvell.com>
154 Jeff Law <law@redhat.com>
156 PR rtl-optimization/93996
157 * haifa-sched.c (remove_notes): Be more careful when adding
160 2020-03-06 Delia Burduv <delia.burduv@arm.com>
162 * config/arm/arm_neon.h (vld2_bf16): New.
168 (vld2_dup_bf16): New.
169 (vld2q_dup_bf16): New.
170 (vld3_dup_bf16): New.
171 (vld3q_dup_bf16): New.
172 (vld4_dup_bf16): New.
173 (vld4q_dup_bf16): New.
174 * config/arm/arm_neon_builtins.def
175 (vld2): Changed to VAR13 and added v4bf, v8bf
176 (vld2_dup): Changed to VAR8 and added v4bf, v8bf
177 (vld3): Changed to VAR13 and added v4bf, v8bf
178 (vld3_dup): Changed to VAR8 and added v4bf, v8bf
179 (vld4): Changed to VAR13 and added v4bf, v8bf
180 (vld4_dup): Changed to VAR8 and added v4bf, v8bf
181 * config/arm/iterators.md (VDXBF2): New iterator.
182 *config/arm/neon.md (neon_vld2): Use new iterators.
183 (neon_vld2_dup<mode): Use new iterators.
184 (neon_vld3<mode>): Likewise.
185 (neon_vld3qa<mode>): Likewise.
186 (neon_vld3qb<mode>): Likewise.
187 (neon_vld3_dup<mode>): Likewise.
188 (neon_vld4<mode>): Likewise.
189 (neon_vld4qa<mode>): Likewise.
190 (neon_vld4qb<mode>): Likewise.
191 (neon_vld4_dup<mode>): Likewise.
192 (neon_vld2_dupv8bf): New.
193 (neon_vld3_dupv8bf): Likewise.
194 (neon_vld4_dupv8bf): Likewise.
196 2020-03-06 Delia Burduv <delia.burduv@arm.com>
198 * config/arm/arm_neon.h (bfloat16x4x2_t): New typedef.
199 (bfloat16x8x2_t): New typedef.
200 (bfloat16x4x3_t): New typedef.
201 (bfloat16x8x3_t): New typedef.
202 (bfloat16x4x4_t): New typedef.
203 (bfloat16x8x4_t): New typedef.
210 * config/arm/arm-builtins.c (v2bf_UP): Define.
212 (arm_init_simd_builtin_types): Init Bfloat16x2_t eltype.
213 * config/arm/arm-modes.def (V2BF): New mode.
214 * config/arm/arm-simd-builtin-types.def
215 (Bfloat16x2_t): New entry.
216 * config/arm/arm_neon_builtins.def
217 (vst2): Changed to VAR13 and added v4bf, v8bf
218 (vst3): Changed to VAR13 and added v4bf, v8bf
219 (vst4): Changed to VAR13 and added v4bf, v8bf
220 * config/arm/iterators.md (VDXBF): New iterator.
221 (VQ2BF): New iterator.
222 *config/arm/neon.md (neon_vst2<mode>): Used new iterators.
223 (neon_vst2<mode>): Used new iterators.
224 (neon_vst3<mode>): Used new iterators.
225 (neon_vst3<mode>): Used new iterators.
226 (neon_vst3qa<mode>): Used new iterators.
227 (neon_vst3qb<mode>): Used new iterators.
228 (neon_vst4<mode>): Used new iterators.
229 (neon_vst4<mode>): Used new iterators.
230 (neon_vst4qa<mode>): Used new iterators.
231 (neon_vst4qb<mode>): Used new iterators.
233 2020-03-06 Delia Burduv <delia.burduv@arm.com>
235 * config/aarch64/aarch64-simd-builtins.def
236 (bfcvtn): New built-in function.
237 (bfcvtn_q): New built-in function.
238 (bfcvtn2): New built-in function.
239 (bfcvt): New built-in function.
240 * config/aarch64/aarch64-simd.md
241 (aarch64_bfcvtn<q><mode>): New pattern.
242 (aarch64_bfcvtn2v8bf): New pattern.
243 (aarch64_bfcvtbf): New pattern.
244 * config/aarch64/arm_bf16.h (float32_t): New typedef.
245 (vcvth_bf16_f32): New intrinsic.
246 * config/aarch64/arm_bf16.h (vcvt_bf16_f32): New intrinsic.
247 (vcvtq_low_bf16_f32): New intrinsic.
248 (vcvtq_high_bf16_f32): New intrinsic.
249 * config/aarch64/iterators.md (V4SF_TO_BF): New mode iterator.
250 (UNSPEC_BFCVTN): New UNSPEC.
251 (UNSPEC_BFCVTN2): New UNSPEC.
252 (UNSPEC_BFCVT): New UNSPEC.
253 * config/arm/types.md (bf_cvt): New type.
255 2020-03-06 Andreas Krebbel <krebbel@linux.ibm.com>
257 * config/s390/s390.md ("tabort"): Get rid of two consecutive
258 blanks in format string.
260 2020-03-05 H.J. Lu <hongjiu.lu@intel.com>
264 * config/i386/i386-protos.h (ix86_output_ssemov): New prototype.
265 * config/i386/i386.c (ix86_get_ssemov): New function.
266 (ix86_output_ssemov): Likewise.
267 * config/i386/sse.md (VMOVE:mov<mode>_internal): Call
268 ix86_output_ssemov for TYPE_SSEMOV. Remove TARGET_AVX512VL
270 (*movxi_internal_avx512f): Call ix86_output_ssemov for TYPE_SSEMOV.
271 (*movoi_internal_avx): Call ix86_output_ssemov for TYPE_SSEMOV.
272 Remove ext_sse_reg_operand and TARGET_AVX512VL check.
273 (*movti_internal): Likewise.
274 (*movtf_internal): Call ix86_output_ssemov for TYPE_SSEMOV.
276 2020-03-05 Jeff Law <law@redhat.com>
278 PR tree-optimization/91890
279 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Remove LOC argument.
280 Use gimple_or_expr_nonartificial_location.
281 (check_bounds_overlap): Drop LOC argument to maybe_diag_access_bounds.
282 Use gimple_or_expr_nonartificial_location.
283 * gimple.c (gimple_or_expr_nonartificial_location): New function.
284 * gimple.h (gimple_or_expr_nonartificial_location): Declare it.
285 * tree-ssa-strlen.c (maybe_warn_overflow): Use
286 gimple_or_expr_nonartificial_location.
287 (maybe_diag_stxncpy_trunc, handle_builtin_stxncpy_strncat): Likewise.
288 (maybe_warn_pointless_strcmp): Likewise.
290 2020-03-05 Jakub Jelinek <jakub@redhat.com>
293 * config/i386/avx2intrin.h (_mm_mask_i32gather_ps): Fix first cast of
294 SRC and MASK arguments to __m128 from __m128d.
295 (_mm256_mask_i32gather_ps): Fix first cast of MASK argument to __m256
297 (_mm_mask_i64gather_ps): Fix first cast of MASK argument to __m128
299 * config/i386/xopintrin.h (_mm_permute2_pd): Fix first cast of C
300 argument to __m128i from __m128d.
301 (_mm256_permute2_pd): Fix first cast of C argument to __m256i from
303 (_mm_permute2_ps): Fix first cast of C argument to __m128i from __m128.
304 (_mm256_permute2_ps): Fix first cast of C argument to __m256i from
307 2020-03-05 Delia Burduv <delia.burduv@arm.com>
309 * config/arm/arm_neon.h (vbfmmlaq_f32): New.
310 (vbfmlalbq_f32): New.
311 (vbfmlaltq_f32): New.
312 (vbfmlalbq_lane_f32): New.
313 (vbfmlaltq_lane_f32): New.
314 (vbfmlalbq_laneq_f32): New.
315 (vbfmlaltq_laneq_f32): New.
316 * config/arm/arm_neon_builtins.def (vmmla): New.
323 * config/arm/iterators.md (BF_MA): New int iterator.
324 (bt): New int attribute.
325 (VQXBF): Copy of VQX with V8BF.
326 * config/arm/neon.md (neon_vmmlav8bf): New insn.
327 (neon_vfma<bt>v8bf): New insn.
328 (neon_vfma<bt>_lanev8bf): New insn.
329 (neon_vfma<bt>_laneqv8bf): New expand.
330 (neon_vget_high<mode>): Changed iterator to VQXBF.
331 * config/arm/unspecs.md (UNSPEC_BFMMLA): New UNSPEC.
332 (UNSPEC_BFMAB): New UNSPEC.
333 (UNSPEC_BFMAT): New UNSPEC.
335 2020-03-05 Jakub Jelinek <jakub@redhat.com>
338 * tree-pretty-print.h (pretty_print_string): Declare.
339 * tree-pretty-print.c (pretty_print_string): Remove forward
340 declaration, no longer static. Change nbytes parameter type
341 from unsigned to size_t.
342 * print-rtl.c (print_value) <case CONST_STRING>: Use
343 pretty_print_string and for shrink way too long strings.
345 2020-03-05 Richard Biener <rguenther@suse.de>
346 Jakub Jelinek <jakub@redhat.com>
348 PR tree-optimization/93582
349 * tree-ssa-sccvn.c (vn_reference_lookup_3): Treat POINTER_PLUS_EXPR
350 last operand as signed when looking for memset offset. Formatting
353 2020-03-04 Andrew Pinski <apinski@marvell.com>
356 * value-prof.c (dump_histogram_value): Use std::abs.
358 2020-03-04 Martin Sebor <msebor@redhat.com>
360 PR tree-optimization/93986
361 * tree-ssa-strlen.c (maybe_warn_overflow): Convert all wide_int
362 operands to the same precision widest_int to avoid ICEs.
364 2020-03-04 Bill Schmidt <wschmidt@linux.ibm.com>
367 * rs6000-cpus.def (OTHER_ALTIVEC_MASKS): New #define.
368 * rs6000.c (rs6000_disable_incompatible_switches): Add table entry
369 for OPTION_MASK_ALTIVEC.
371 2020-03-04 Andreas Krebbel <krebbel@linux.ibm.com>
373 * config.gcc: Include the glibc-stdint.h header for zTPF.
375 2020-03-04 Andreas Krebbel <krebbel@linux.ibm.com>
377 * config/s390/s390.c (s390_secondary_memory_needed): Disallow
378 direct FPR-GPR copies.
379 (s390_register_info_gprtofpr): Disallow GPR content to be saved in
382 2020-03-04 Andreas Krebbel <krebbel@linux.ibm.com>
384 * config/s390/s390.c (s390_emit_prologue): Specify the 2 new
385 operands to the prologue_tpf expander.
386 (s390_emit_epilogue): Likewise.
387 (s390_option_override_internal): Do error checking and setup for
389 * config/s390/tpf.h (TPF_TRACE_PROLOGUE_CHECK)
390 (TPF_TRACE_EPILOGUE_CHECK, TPF_TRACE_PROLOGUE_TARGET)
391 (TPF_TRACE_EPILOGUE_TARGET, TPF_TRACE_PROLOGUE_SKIP_TARGET)
392 (TPF_TRACE_EPILOGUE_SKIP_TARGET): New macro definitions.
393 * config/s390/tpf.md ("prologue_tpf", "epilogue_tpf"): Add two new
394 operands for the check flag and the branch target.
395 * config/s390/tpf.opt ("mtpf-trace-hook-prologue-check")
396 ("mtpf-trace-hook-prologue-target")
397 ("mtpf-trace-hook-epilogue-check")
398 ("mtpf-trace-hook-epilogue-target", "mtpf-trace-skip"): New
400 * doc/invoke.texi: Document -mtpf-trace-skip option. The other
401 options are for debugging purposes and will not be documented
404 2020-03-04 Jakub Jelinek <jakub@redhat.com>
407 * tree-inline.c (copy_decl_to_var): Copy DECL_BY_REFERENCE flag.
409 * tree-ssa-sccvn.c (vn_walk_cb_data::push_partial_def): Add offseti
410 argument. Change pd argument so that it can be modified. Turn
411 constant non-CONSTRUCTOR store into non-constant if it is too large.
412 Adjust offset and size of CONSTRUCTOR or non-constant store to avoid
414 (vn_walk_cb_data::vn_walk_cb_data, vn_reference_lookup_3): Adjust
417 2020-02-04 Richard Biener <rguenther@suse.de>
419 PR tree-optimization/93964
420 * graphite-isl-ast-to-gimple.c
421 (gcc_expression_from_isl_ast_expr_id): Add intermediate
422 conversion for pointer to integer converts.
423 * graphite-scop-detection.c (assign_parameter_index_in_region):
426 2020-03-04 Martin Liska <mliska@suse.cz>
430 * doc/invoke.texi: Clarify --help=language and --help=common
433 2020-03-04 Jakub Jelinek <jakub@redhat.com>
435 PR tree-optimization/94001
436 * tree-tailcall.c (process_assignment): Before comparing op1 to
437 *ass_var, verify *ass_var is non-NULL.
439 2020-03-04 Kito Cheng <kito.cheng@sifive.com>
442 * config/riscv/riscv.c (riscv_emit_float_compare): Using NE to compare
445 2020-03-03 Dennis Zhang <dennis.zhang@arm.com>
447 * config/arm/arm_bf16.h (vcvtah_f32_bf16, vcvth_bf16_f32): New.
448 * config/arm/arm_neon.h (vcvt_f32_bf16, vcvtq_low_f32_bf16): New.
449 (vcvtq_high_f32_bf16, vcvt_bf16_f32): New.
450 (vcvtq_low_bf16_f32, vcvtq_high_bf16_f32): New.
451 * config/arm/arm_neon_builtins.def (vbfcvt, vbfcvt_high): New entries.
452 (vbfcvtv4sf, vbfcvtv4sf_high): Likewise.
453 * config/arm/iterators.md (VBFCVT, VBFCVTM): New mode iterators.
454 (V_bf_low, V_bf_cvt_m): New mode attributes.
455 * config/arm/neon.md (neon_vbfcvtv4sf<VBFCVT:mode>): New.
456 (neon_vbfcvtv4sf_highv8bf, neon_vbfcvtsf): New.
457 (neon_vbfcvt<VBFCVT:mode>, neon_vbfcvt_highv8bf): New.
458 (neon_vbfcvtbf_cvtmode<mode>, neon_vbfcvtbf): New
459 * config/arm/unspecs.md (UNSPEC_BFCVT, UNSPEC_BFCVT_HIG): New.
461 2020-03-03 Jakub Jelinek <jakub@redhat.com>
463 PR tree-optimization/93582
464 * tree-ssa-sccvn.h (vn_reference_lookup): Add mask argument.
465 * tree-ssa-sccvn.c (struct vn_walk_cb_data): Add mask and masked_result
466 members, initialize them in the constructor and if mask is non-NULL,
467 artificially push_partial_def {} for the portions of the mask that
469 (vn_walk_cb_data::finish): If mask is non-NULL, set masked_result to
470 val and return (void *)-1. Formatting fix.
471 (vn_reference_lookup_pieces): Adjust vn_walk_cb_data initialization.
473 (vn_reference_lookup): Add mask argument. If non-NULL, don't call
474 fully_constant_vn_reference_p nor vn_reference_lookup_1 and return
476 (visit_nary_op): Handle BIT_AND_EXPR of a memory load and INTEGER_CST
478 (visit_stmt): Formatting fix.
480 2020-03-03 Richard Biener <rguenther@suse.de>
482 PR tree-optimization/93946
483 * alias.h (refs_same_for_tbaa_p): Declare.
484 * alias.c (refs_same_for_tbaa_p): New function.
485 * tree-ssa-alias.c (ao_ref_alias_set): For a NULL ref return
487 * tree-ssa-scopedtables.h
488 (avail_exprs_stack::lookup_avail_expr): Add output argument
489 giving access to the hashtable entry.
490 * tree-ssa-scopedtables.c (avail_exprs_stack::lookup_avail_expr):
492 * tree-ssa-dom.c: Include alias.h.
493 (dom_opt_dom_walker::optimize_stmt): Validate TBAA state before
494 removing redundant store.
495 * tree-ssa-sccvn.h (vn_reference_s::base_set): New member.
496 (ao_ref_init_from_vn_reference): Adjust prototype.
497 (vn_reference_lookup_pieces): Likewise.
498 (vn_reference_insert_pieces): Likewise.
499 * tree-ssa-sccvn.c: Track base alias set in addition to alias
501 (eliminate_dom_walker::eliminate_stmt): Also check base alias
502 set when removing redundant stores.
503 (visit_reference_op_store): Likewise.
504 * dse.c (record_store): Adjust valdity check for redundant
507 2020-03-03 Jakub Jelinek <jakub@redhat.com>
510 * config/s390/s390.h (OPTION_DEFAULT_SPECS): Reorder.
512 PR rtl-optimization/94002
513 * explow.c (plus_constant): Punt if cst has VOIDmode and
514 get_pool_mode is different from mode.
516 2020-03-03 Claudiu Zissulescu <claziss@synopsys.com>
518 * config/arc/arc.c (leigitimate_small_data_address_p): Check if an
519 address has an offset which fits the scalling constraint for a
520 load/store operation.
521 (legitimate_scaled_address_p): Update use
522 leigitimate_small_data_address_p.
523 (arc_print_operand): Likewise.
524 (arc_legitimate_address_p): Likewise.
525 (legitimate_small_data_address_p): Likewise.
527 2020-03-03 Claudiu Zissulescu <claziss@synopsys.com>
529 * config/arc/arc.md (fmasf4_fpu): Use accl_operand predicate.
530 (fnmasf4_fpu): Likewise.
532 2020-03-03 Claudiu Zissulescu <claziss@synopsys.com>
534 * config/arc/arc.md (adddi3): Early expand the 64bit operation into
537 (adddi3_i): Remove pattern.
538 (subdi3_i): Likewise.
540 2020-03-03 Claudiu Zissulescu <claziss@synopsys.com>
542 * config/arc/arc.md (eh_return): Add length info.
544 2020-03-02 David Malcolm <dmalcolm@redhat.com>
546 * doc/invoke.texi (-fanalyzer-show-duplicate-count): New.
548 2020-03-02 David Malcolm <dmalcolm@redhat.com>
550 * doc/invoke.texi (Static Analyzer Options): Add
551 -Wanalyzer-stale-setjmp-buffer to the list of options enabled
554 2020-03-02 Uroš Bizjak <ubizjak@gmail.com>
557 * config/i386/i386.md (movstrict<mode>): Allow only
558 registers with VALID_INT_MODE_P modes.
560 2020-03-02 Andrew Stubbs <ams@codesourcery.com>
562 * config/gcn/gcn-valu.md (dpp_move<mode>): New.
563 (reduc_insn): Use 'U' and 'B' operand codes.
564 (reduc_<reduc_op>_scal_<mode>): Allow all types.
565 (reduc_<reduc_op>_scal_v64di): Delete.
566 (*<reduc_op>_dpp_shr_<mode>): Allow all 1reg types.
567 (*plus_carry_dpp_shr_v64si): Change to ...
568 (*plus_carry_dpp_shr_<mode>): ... this and allow all 1reg int types.
569 (mov_from_lane63_v64di): Change to ...
570 (mov_from_lane63_<mode>): ... this, and allow all 64-bit modes.
571 * config/gcn/gcn.c (gcn_expand_dpp_shr_insn): Increase buffer size.
572 Support UNSPEC_MOV_DPP_SHR output formats.
573 (gcn_expand_reduc_scalar): Add "use_moves" reductions.
574 Add "use_extends" reductions.
575 (print_operand_address): Add 'I' and 'U' codes.
576 * config/gcn/gcn.md (unspec): Add UNSPEC_MOV_DPP_SHR.
578 2020-03-02 Martin Liska <mliska@suse.cz>
580 * lto-wrapper.c: Fix typo in comment about
581 C++ standard version.
583 2020-03-01 Martin Sebor <msebor@redhat.com>
586 * calls.c (init_attr_rdwr_indices): Correctly handle attribute.
588 2020-03-01 Martin Sebor <msebor@redhat.com>
591 * tree-ssa-strlen.c (count_nonzero_bytes): Set the size to that
592 of a pointer in the outermost ADDR_EXPRs.
594 2020-02-28 Jeff Law <law@redhat.com>
596 * config/v850/v850.h (STATIC_CHAIN_REGNUM): Change to r19.
597 * config/v850/v850.c (v850_asm_trampoline_template): Update
600 2020-02-28 Michael Meissner <meissner@linux.ibm.com>
603 * config/rs6000/vsx.md (vsx_extract_<mode>_<VS_scalar>mode_var):
606 2020-02-28 Martin Liska <mliska@suse.cz>
609 * configure.ac: Improve detection of ld_date by requiring
610 either two dashes or none.
611 * configure: Regenerate.
613 2020-02-28 Vladimir Makarov <vmakarov@redhat.com>
615 PR rtl-optimization/93564
616 * ira-color.c (assign_hard_reg): Prefer smaller hard regno when we
617 do not honor reg alloc order.
619 2020-02-27 Joel Hutton <Joel.Hutton@arm.com>
622 * config/aarch64/aarch64.c (aarch64_override_options): Fix
623 misleading warning string.
625 2020-02-27 Martin Sebor <msebor@redhat.com>
627 * doc/invoke.texi (-Wbuiltin-declaration-mismatch): Fix a typo.
629 2020-02-27 Michael Meissner <meissner@linux.ibm.com>
632 * config/rs6000/vsx.md (vsx_extract_<mode>_var, VSX_D iterator):
633 Split the insn into two parts. This insn only does variable
634 extract from a register.
635 (vsx_extract_<mode>_var_load, VSX_D iterator): New insn, do
636 variable extract from memory.
637 (vsx_extract_v4sf_var): Split the insn into two parts. This insn
638 only does variable extract from a register.
639 (vsx_extract_v4sf_var_load): New insn, do variable extract from
641 (vsx_extract_<mode>_var, VSX_EXTRACT_I iterator): Split the insn
642 into two parts. This insn only does variable extract from a
644 (vsx_extract_<mode>_var_load, VSX_EXTRACT_I iterator): New insn,
645 do variable extract from memory.
647 2020-02-27 Martin Jambor <mjambor@suse.cz>
648 Feng Xue <fxue@os.amperecomputing.com>
651 * ipa-cp.c (same_node_or_its_all_contexts_clone_p): Replaced with
652 new function calls_same_node_or_its_all_contexts_clone_p.
653 (cgraph_edge_brings_value_p): Use it.
654 (cgraph_edge_brings_value_p): Likewise.
655 (self_recursive_pass_through_p): Return false if caller is a clone.
656 (self_recursive_agg_pass_through_p): Likewise.
658 2020-02-27 Jan Hubicka <hubicka@ucw.cz>
661 * alias.c (ends_tbaa_access_path_p): Break out from ...
662 (component_uses_parent_alias_set_from): ... here.
663 * alias.h (ends_tbaa_access_path_p): Declare.
664 * tree-ssa-alias.c (access_path_may_continue_p): Break out from ...;
665 handle trailing arrays past end of tbaa access path.
666 (aliasing_component_refs_p): ... here; likewise.
667 (nonoverlapping_refs_since_match_p): Track TBAA segment of the access
668 path; disambiguate also past end of it.
669 (nonoverlapping_component_refs_p): Use only TBAA segment of the access
672 2020-02-27 Mihail Ionescu <mihail.ionescu@arm.com>
674 * (__ARM_NUM_LANES, __arm_lane, __arm_lane_q): Move to the
675 beginning of the file.
676 (vcreate_bf16, vcombine_bf16): New.
677 (vdup_n_bf16, vdupq_n_bf16): New.
678 (vdup_lane_bf16, vdup_laneq_bf16): New.
679 (vdupq_lane_bf16, vdupq_laneq_bf16): New.
680 (vduph_lane_bf16, vduph_laneq_bf16): New.
681 (vset_lane_bf16, vsetq_lane_bf16): New.
682 (vget_lane_bf16, vgetq_lane_bf16): New.
683 (vget_high_bf16, vget_low_bf16): New.
684 (vreinterpret_bf16_u8, vreinterpretq_bf16_u8): New.
685 (vreinterpret_bf16_u16, vreinterpretq_bf16_u16): New.
686 (vreinterpret_bf16_u32, vreinterpretq_bf16_u32): New.
687 (vreinterpret_bf16_u64, vreinterpretq_bf16_u64): New.
688 (vreinterpret_bf16_s8, vreinterpretq_bf16_s8): New.
689 (vreinterpret_bf16_s16, vreinterpretq_bf16_s16): New.
690 (vreinterpret_bf16_s32, vreinterpretq_bf16_s32): New.
691 (vreinterpret_bf16_s64, vreinterpretq_bf16_s64): New.
692 (vreinterpret_bf16_p8, vreinterpretq_bf16_p8): New.
693 (vreinterpret_bf16_p16, vreinterpretq_bf16_p16): New.
694 (vreinterpret_bf16_p64, vreinterpretq_bf16_p64): New.
695 (vreinterpret_bf16_f32, vreinterpretq_bf16_f32): New.
696 (vreinterpret_bf16_f64, vreinterpretq_bf16_f64): New.
697 (vreinterpretq_bf16_p128): New.
698 (vreinterpret_s8_bf16, vreinterpretq_s8_bf16): New.
699 (vreinterpret_s16_bf16, vreinterpretq_s16_bf16): New.
700 (vreinterpret_s32_bf16, vreinterpretq_s32_bf16): New.
701 (vreinterpret_s64_bf16, vreinterpretq_s64_bf16): New.
702 (vreinterpret_u8_bf16, vreinterpretq_u8_bf16): New.
703 (vreinterpret_u16_bf16, vreinterpretq_u16_bf16): New.
704 (vreinterpret_u32_bf16, vreinterpretq_u32_bf16): New.
705 (vreinterpret_u64_bf16, vreinterpretq_u64_bf16): New.
706 (vreinterpret_p8_bf16, vreinterpretq_p8_bf16): New.
707 (vreinterpret_p16_bf16, vreinterpretq_p16_bf16): New.
708 (vreinterpret_p64_bf16, vreinterpretq_p64_bf16): New.
709 (vreinterpret_f32_bf16, vreinterpretq_f32_bf16): New.
710 (vreinterpretq_p128_bf16): New.
711 * config/arm/arm_neon_builtins.def (VDX): Add V4BF.
713 (V_elem_l): Likewise.
716 (V_DOUBLE): Likewise.
717 (VDQX): Add V4BF and V8BF.
718 (V_two_elem, V_three_elem, V_four_elem): Likewise.
721 (V_double_vector_mode): Likewise.
722 (V_cmp_result): Likewise.
723 (V_uf_sclr): Likewise.
724 (V_sz_elem): Likewise.
725 (Is_d_reg): Likewise.
726 (V_mode_nunits): Likewise.
727 * config/arm/neon.md (neon_vdup_lane): Enable for BFloat16.
729 2020-02-27 Andrew Stubbs <ams@codesourcery.com>
731 * config/gcn/gcn-valu.md (VEC_SUBDWORD_MODE): New mode iterator.
732 (<expander><mode>2<exec>): Change modes to VEC_ALL1REG_INT_MODE.
733 (<expander><mode>3<exec>): Likewise.
734 (<expander><mode>3): New.
735 (v<expander><mode>3): New.
736 (<expander><mode>3): New.
737 (<expander><mode>3<exec>): Rename to ...
738 (<expander>v64si3<exec>): ... this, and change modes to V64SI.
739 * config/gcn/gcn.md (mnemonic): Use '%B' for not.
741 2020-02-27 Alexandre Oliva <oliva@adacore.com>
743 * config/vx-common.h (NO_DOLLAR_IN_LABEL, NO_DOT_IN_LABEL): Leave
746 2020-02-27 Richard Biener <rguenther@suse.de>
748 PR tree-optimization/93508
749 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle _CHK like
750 non-_CHK variants. Valueize their length arguments.
752 2020-02-27 Richard Biener <rguenther@suse.de>
754 PR tree-optimization/93953
755 * tree-vect-slp.c (slp_copy_subtree): Avoid keeping a reference
756 to the hash-map entry.
758 2020-02-27 Andrew Stubbs <ams@codesourcery.com>
760 * config/gcn/gcn.md (mov<mode>): Add transformations for BI subregs.
762 2020-02-27 Mark Williams <mwilliams@fb.com>
764 * dwarf2out.c (file_name_acquire): Call remap_debug_filename.
765 * lto-opts.c (lto_write_options): Drop -fdebug-prefix-map,
766 -ffile-prefix-map and -fmacro-prefix-map.
767 * lto-streamer-out.c: Include file-prefix-map.h.
768 (lto_output_location): Remap the file part of locations.
770 2020-02-27 Jakub Jelinek <jakub@redhat.com>
773 * gimplify.c (gimplify_init_constructor): Don't promote readonly
774 DECL_REGISTER variables to TREE_STATIC.
776 PR tree-optimization/93582
777 PR tree-optimization/93945
778 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle memset with
779 non-zero INTEGER_CST second argument and ref->offset or ref->size
780 not a multiple of BITS_PER_UNIT.
782 2020-02-27 Jonathan Wakely <jwakely@redhat.com>
784 * doc/install.texi (Binaries): Update description of BullFreeware.
786 2020-02-26 Sandra Loosemore <sandra@codesourcery.com>
790 * doc/invoke.texi (Option Summary): Re-alphabetize warnings in
791 C++ Language Options, Warning Options, and Static Analyzer
792 Options lists. Document negative form of options enabled by
793 default. Move some things around to more accurately sort
794 warnings by category.
795 (C++ Dialect Options, Warning Options, Static Analyzer
796 Options): Document negative form of options when enabled by
797 default. Move some things around to more accurately sort
798 warnings by category. Add some missing index entries.
801 2020-02-26 Carl Love <cel@us.ibm.com>
804 * doc/extend.texi (PowerPC AltiVec Built-in Functions available on
805 ISA 2.07): The builtin-function name __builtin_crypto_vpmsumb is only
806 for the vector unsigned short arguments. It is also listed as the
807 name of the built-in for arguments vector unsigned short,
808 vector unsigned int and vector unsigned long long built-ins. The
809 name of the builtins for these arguments should be:
810 __builtin_crypto_vpmsumh, __builtin_crypto_vpmsumw and
811 __builtin_crypto_vpmsumd respectively.
813 2020-02-26 Richard Biener <rguenther@suse.de>
815 * tree-vect-slp.c (vect_print_slp_tree): Also dump ref count
816 and load permutation.
818 2020-02-26 Richard Sandiford <richard.sandiford@arm.com>
821 * optabs-tree.c (supportable_convert_operation): Reject types with
824 2020-02-26 David Malcolm <dmalcolm@redhat.com>
826 * Makefile.in (ANALYZER_OBJS): Add analyzer/bar-chart.o.
828 2020-02-26 Jakub Jelinek <jakub@redhat.com>
830 PR tree-optimization/93820
831 * gimple-ssa-store-merging.c (check_no_overlap): Change RHS_CODE
832 argument to ALL_INTEGER_CST_P boolean.
833 (imm_store_chain_info::try_coalesce_bswap): Adjust caller.
834 (imm_store_chain_info::coalesce_immediate_stores): Likewise. Handle
835 adjacent INTEGER_CST store into merged_store->only_constants like
838 2020-02-25 Jakub Jelinek <jakub@redhat.com>
841 * config/sh/sh.c (expand_cbranchdi4): Fix comment typo, probablity
843 * cfghooks.c (verify_flow_info): Likewise.
844 * predict.c (combine_predictions_for_bb): Likewise.
845 * bb-reorder.c (connect_better_edge_p): Likewise. Fix comment typo,
846 sucessor -> successor.
847 (find_traces_1_round): Fix comment typo, destinarion -> destination.
848 * omp-expand.c (expand_oacc_for): Fix comment typo, sucessors ->
850 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): Fix dump
851 message typo, sucessors -> successors.
853 2020-02-25 Martin Sebor <msebor@redhat.com>
855 * doc/extend.texi (attribute access): Correct an example.
857 2020-02-25 Mihail Ionescu <mihail.ionescu@arm.com>
859 * config/aarch64/aarch64-builtins.c (aarch64_scalar_builtin_types):
861 (aarch64_init_simd_builtin_scalar_types): Register simd_bf.
863 * config/aarch64/iterators.md (VALLDIF): Enable for V4BF and V8BF.
864 (VD): Enable for V4BF.
866 (VQ): Enable for V8BF.
869 (VDBL, Vdbl): Add V4BF.
870 (V_INT_EQUIV, v_int_equiv): Add V4BF and V8BF.
871 * config/aarch64/arm_neon.h (bfloat16x4x2_t): New typedef.
872 (bfloat16x8x2_t): Likewise.
873 (bfloat16x4x3_t): Likewise.
874 (bfloat16x8x3_t): Likewise.
875 (bfloat16x4x4_t): Likewise.
876 (bfloat16x8x4_t): Likewise.
877 (vcombine_bf16): New.
878 (vld1_bf16, vld1_bf16_x2): New.
879 (vld1_bf16_x3, vld1_bf16_x4): New.
880 (vld1q_bf16, vld1q_bf16_x2): New.
881 (vld1q_bf16_x3, vld1q_bf16_x4): New.
882 (vld1_lane_bf16): New.
883 (vld1q_lane_bf16): New.
884 (vld1_dup_bf16): New.
885 (vld1q_dup_bf16): New.
888 (vld2_dup_bf16): New.
889 (vld2q_dup_bf16): New.
892 (vld3_dup_bf16): New.
893 (vld3q_dup_bf16): New.
896 (vld4_dup_bf16): New.
897 (vld4q_dup_bf16): New.
898 (vst1_bf16, vst1_bf16_x2): New.
899 (vst1_bf16_x3, vst1_bf16_x4): New.
900 (vst1q_bf16, vst1q_bf16_x2): New.
901 (vst1q_bf16_x3, vst1q_bf16_x4): New.
902 (vst1_lane_bf16): New.
903 (vst1q_lane_bf16): New.
911 2020-02-25 Mihail Ionescu <mihail.ionescu@arm.com>
913 * config/aarch64/iterators.md (VDQF_F16) Add V4BF and V8BF.
914 (VALL_F16): Likewise.
915 (VALLDI_F16): Likewise.
918 (vswap_width_name): Likewise.
919 (VSWAP_WIDTH): Likewise.
923 * config/aarch64/arm_neon.h (vset_lane_bf16, vsetq_lane_bf16): New.
924 (vget_lane_bf16, vgetq_lane_bf16): New.
926 (vdup_n_bf16, vdupq_n_bf16): New.
927 (vdup_lane_bf16, vdup_laneq_bf16): New.
928 (vdupq_lane_bf16, vdupq_laneq_bf16): New.
929 (vduph_lane_bf16, vduph_laneq_bf16): New.
930 (vreinterpret_bf16_u8, vreinterpretq_bf16_u8): New.
931 (vreinterpret_bf16_u16, vreinterpretq_bf16_u16): New.
932 (vreinterpret_bf16_u32, vreinterpretq_bf16_u32): New.
933 (vreinterpret_bf16_u64, vreinterpretq_bf16_u64): New.
934 (vreinterpret_bf16_s8, vreinterpretq_bf16_s8): New.
935 (vreinterpret_bf16_s16, vreinterpretq_bf16_s16): New.
936 (vreinterpret_bf16_s32, vreinterpretq_bf16_s32): New.
937 (vreinterpret_bf16_s64, vreinterpretq_bf16_s64): New.
938 (vreinterpret_bf16_p8, vreinterpretq_bf16_p8): New.
939 (vreinterpret_bf16_p16, vreinterpretq_bf16_p16): New.
940 (vreinterpret_bf16_p64, vreinterpretq_bf16_p64): New
941 (vreinterpret_bf16_f16, vreinterpretq_bf16_f16): New
942 (vreinterpret_bf16_f32, vreinterpretq_bf16_f32): New.
943 (vreinterpret_bf16_f64, vreinterpretq_bf16_f64): New.
944 (vreinterpretq_bf16_p128): New.
945 (vreinterpret_s8_bf16, vreinterpretq_s8_bf16): New.
946 (vreinterpret_s16_bf16, vreinterpretq_s16_bf16): New.
947 (vreinterpret_s32_bf16, vreinterpretq_s32_bf16): New.
948 (vreinterpret_s64_bf16, vreinterpretq_s64_bf16): New.
949 (vreinterpret_u8_bf16, vreinterpretq_u8_bf16): New.
950 (vreinterpret_u16_bf16, vreinterpretq_u16_bf16): New.
951 (vreinterpret_u32_bf16, vreinterpretq_u32_bf16): New.
952 (vreinterpret_u64_bf16, vreinterpretq_u64_bf16): New.
953 (vreinterpret_p8_bf16, vreinterpretq_p8_bf16): New.
954 (vreinterpret_p16_bf16, vreinterpretq_p16_bf16): New.
955 (vreinterpret_p64_bf16, vreinterpretq_p64_bf16): New.
956 (vreinterpret_f32_bf16, vreinterpretq_f32_bf16): New.
957 (vreinterpret_f64_bf16,vreinterpretq_f64_bf16): New.
958 (vreinterpret_f16_bf16,vreinterpretq_f16_bf16): New.
959 (vreinterpretq_p128_bf16): New.
961 2020-02-25 Dennis Zhang <dennis.zhang@arm.com>
963 * config/arm/arm_neon.h (vbfdot_f32, vbfdotq_f32): New
964 (vbfdot_lane_f32, vbfdotq_laneq_f32): New.
965 (vbfdot_laneq_f32, vbfdotq_lane_f32): New.
966 * config/arm/arm_neon_builtins.def (vbfdot): New entry.
967 (vbfdot_lanev4bf, vbfdot_lanev8bf): Likewise.
968 * config/arm/iterators.md (VSF2BF): New attribute.
969 * config/arm/neon.md (neon_vbfdot<VCVTF:mode>): New entry.
970 (neon_vbfdot_lanev4bf<VCVTF:mode>): Likewise.
971 (neon_vbfdot_lanev8bf<VCVTF:mode>): Likewise.
973 2020-02-25 Christophe Lyon <christophe.lyon@linaro.org>
975 * config/arm/arm.md (required_for_purecode): New attribute.
976 (enabled): Handle required_for_purecode.
977 * config/arm/thumb1.md (thumb1_movsi_insn): Add alternative to
978 work with -mpure-code.
980 2020-02-25 Jakub Jelinek <jakub@redhat.com>
982 PR rtl-optimization/93908
983 * combine.c (find_split_point): For store into ZERO_EXTRACT, and src
986 2019-02-25 Eric Botcazou <ebotcazou@adacore.com>
988 * dwarf2out.c (dwarf2out_size_function): Run in early-DWARF mode.
990 2020-02-25 Roman Zhuykov <zhroma@ispras.ru>
992 * doc/install.texi (--enable-checking): Adjust wording.
994 2020-02-25 Richard Biener <rguenther@suse.de>
996 PR tree-optimization/93868
997 * tree-vect-slp.c (slp_copy_subtree): New function.
998 (vect_attempt_slp_rearrange_stmts): Copy the SLP tree before
999 re-arranging stmts in it.
1001 2020-02-25 Jakub Jelinek <jakub@redhat.com>
1004 * passes.c (pass_manager::dump_passes): Create a cgraph node for the
1005 dummy function and remove it at the end.
1007 PR translation/93864
1008 * config/lm32/lm32.c (lm32_setup_incoming_varargs): Fix comment typo
1009 paramter -> parameter.
1010 * config/aarch64/aarch64.c (aarch64_is_extend_from_extract): Likewise.
1011 * ipa-prop.h (struct ipa_agg_replacement_value): Likewise.
1013 2020-02-24 Roman Zhuykov <zhroma@ispras.ru>
1015 * doc/install.texi (--enable-checking): Properly document current
1017 (--enable-stage1-checking): Minor clarification about bootstrap.
1019 2020-02-24 David Malcolm <dmalcolm@redhat.com>
1022 * doc/invoke.texi (-Wnanalyzer-tainted-array-index): Note that
1023 -fanalyzer-checker=taint is also required.
1024 (-fanalyzer-checker=): Note that providing this option enables the
1025 given checker, and doing so may be required for checkers that are
1026 disabled by default.
1028 2020-02-24 David Malcolm <dmalcolm@redhat.com>
1030 * doc/invoke.texi (-fanalyzer-verbosity=): "2" only shows
1031 significant control flow events; add a "3" which shows all
1032 control flow events; the old "3" becomes "4".
1034 2020-02-24 Jakub Jelinek <jakub@redhat.com>
1036 PR tree-optimization/93582
1037 * tree-ssa-sccvn.c (vn_walk_cb_data::push_partial_def): Consider
1038 pd.offset and pd.size to be counted in bits rather than bytes, add
1039 support for maxsizei that is not a multiple of BITS_PER_UNIT and
1040 handle bitfield stores and loads.
1041 (vn_reference_lookup_3): Don't call ranges_known_overlap_p with
1042 uncomparable quantities - bytes vs. bits. Allow push_partial_def
1043 on offsets/sizes that aren't multiple of BITS_PER_UNIT and adjust
1044 pd.offset/pd.size to be counted in bits rather than bytes.
1045 Formatting fix. Rename shadowed len variable to buflen.
1047 2020-02-24 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1048 Kugan Vivekandarajah <kugan.vivekanandarajah@linaro.org>
1051 * gcc.c (putenv_COLLECT_AS_OPTIONS): New function.
1052 (driver::main): Call putenv_COLLECT_AS_OPTIONS.
1053 * opts-common.c (parse_options_from_collect_gcc_options): New function.
1054 (prepend_xassembler_to_collect_as_options): Likewise.
1055 * opts.h (parse_options_from_collect_gcc_options): Declare prototype.
1056 (prepend_xassembler_to_collect_as_options): Likewise.
1057 * lto-opts.c (lto_write_options): Stream assembler options
1058 in COLLECT_AS_OPTIONS.
1059 * lto-wrapper.c (xassembler_options_error): New static variable.
1060 (get_options_from_collect_gcc_options): Move parsing options code to
1061 parse_options_from_collect_gcc_options and call it.
1062 (merge_and_complain): Validate -Xassembler options.
1063 (append_compiler_options): Handle OPT_Xassembler.
1064 (run_gcc): Append command line -Xassembler options to
1065 collect_gcc_options.
1066 * doc/invoke.texi: Add documentation about using Xassembler
1069 2020-02-24 Kito Cheng <kito.cheng@sifive.com>
1071 * config/riscv/riscv.c (riscv_emit_float_compare): Change the code gen
1073 (riscv_rtx_costs): Update cost model for LTGT.
1075 2020-02-23 Vladimir Makarov <vmakarov@redhat.com>
1077 PR rtl-optimization/93564
1078 * ira-color.c (struct update_cost_queue_elem): New member start.
1079 (queue_update_cost, get_next_update_cost): Add new arg start.
1080 (allocnos_conflict_p): New function.
1081 (update_costs_from_allocno): Add new arg conflict_cost_update_p.
1082 Add checking conflicts with allocnos_conflict_p.
1083 (update_costs_from_prefs, restore_costs_from_copies): Adjust
1084 update_costs_from_allocno calls.
1085 (update_conflict_hard_regno_costs): Add checking conflicts with
1086 allocnos_conflict_p. Adjust calls of queue_update_cost and
1087 get_next_update_cost.
1088 (assign_hard_reg): Adjust calls of queue_update_cost. Add
1090 (bucket_allocno_compare_func): Restore previous version.
1092 2020-02-21 John David Anglin <danglin@gcc.gnu.org>
1094 * gcc/config/pa/pa.c (pa_function_value): Fix check for word and
1095 double-word size when handling aggregate return values.
1096 * gcc/config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Fix to indicate
1097 that homogeneous SFmode and DFmode aggregates are passed and returned
1098 in general registers.
1100 2020-02-21 Jakub Jelinek <jakub@redhat.com>
1102 PR translation/93759
1103 * opts.c (print_filtered_help): Translate help before appending
1104 messages to it rather than after that.
1106 2020-02-19 Richard Sandiford <richard.sandiford@arm.com>
1108 PR rtl-optimization/PR92989
1109 * lra-lives.c (process_bb_lives): Restore the original order
1110 of the bb liveness update. Call make_hard_regno_dead for each
1111 register clobbered at the start of an EH receiver.
1113 2020-02-18 Feng Xue <fxue@os.amperecomputing.com>
1116 * ipa-cp.c (self_recursively_generated_p): Mark self-dependent value as
1117 self-recursively generated.
1119 2020-02-21 Iain Sandoe <iain@sandoe.co.uk>
1122 * config/darwin-c.c (pop_field_alignment): Adjust quoting of
1125 2020-02-21 Mihail Ionescu <mihail.ionescu@arm.com>
1127 * doc/sourcebuild.texi (arm_v8_1m_mve_ok):
1128 Document new target supports option.
1130 2020-02-21 Dennis Zhang <dennis.zhang@arm.com>
1132 * config/arm/arm_neon.h (vmmlaq_s32, vmmlaq_u32, vusmmlaq_s32): New.
1133 * config/arm/arm_neon_builtins.def (smmla, ummla, usmmla): New.
1134 * config/arm/iterators.md (MATMUL): New iterator.
1135 (sup): Add UNSPEC_MATMUL_S, UNSPEC_MATMUL_U, and UNSPEC_MATMUL_US.
1136 (mmla_sfx): New attribute.
1137 * config/arm/neon.md (neon_<sup>mmlav16qi): New.
1138 * config/arm/unspecs.md (UNSPEC_MATMUL_S, UNSPEC_MATMUL_U): New.
1139 (UNSPEC_MATMUL_US): New.
1141 2020-02-21 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1143 * config/arm/arm.md: Prevent scalar shifts from being used when big
1146 2020-02-21 Jan Hubicka <hubicka@ucw.cz>
1147 Richard Biener <rguenther@suse.de>
1149 PR tree-optimization/93586
1150 * tree-ssa-alias.c (nonoverlapping_array_refs_p): Finish array walk
1151 after mismatched array refs; do not sure type size information to
1152 recover from unmatched referneces with !flag_strict_aliasing_p.
1154 2020-02-21 Andrew Stubbs <ams@codesourcery.com>
1156 * config/gcn/gcn-valu.md (gather_load<mode>): Rename to ...
1157 (gather_load<mode>v64si): ... this and set operand 2 to V64SI.
1158 (scatter_store<mode>): Rename to ...
1159 (scatter_store<mode>v64si): ... this and set operand 1 to V64SI.
1160 (scatter<mode>_exec): Delete. Move contents ...
1161 (mask_scatter_store<mode>): ... here, and rename that to ...
1162 (mask_gather_load<mode>v64si): ... this. Set operand 2 to V64SI.
1163 Remove mode conversion.
1164 (mask_gather_load<mode>): Rename to ...
1165 (mask_scatter_store<mode>v64si): ... this. Set operand 1 to V64SI.
1166 Remove mode conversion.
1167 * config/gcn/gcn.c (gcn_expand_scaled_offsets): Remove mode conversion.
1169 2020-02-21 Martin Jambor <mjambor@suse.cz>
1171 PR tree-optimization/93845
1172 * tree-sra.c (verify_sra_access_forest): Only test access size of
1175 2020-02-21 Andrew Stubbs <ams@codesourcery.com>
1177 * config/gcn/gcn.c (gcn_hard_regno_mode_ok): Align VGPR pairs.
1178 * config/gcn/gcn-valu.md (addv64di3): Remove early-clobber.
1179 (addv64di3_exec): Likewise.
1180 (subv64di3): Likewise.
1181 (subv64di3_exec): Likewise.
1182 (addv64di3_zext): Likewise.
1183 (addv64di3_zext_exec): Likewise.
1184 (addv64di3_zext_dup): Likewise.
1185 (addv64di3_zext_dup_exec): Likewise.
1186 (addv64di3_zext_dup2): Likewise.
1187 (addv64di3_zext_dup2_exec): Likewise.
1188 (addv64di3_sext_dup2): Likewise.
1189 (addv64di3_sext_dup2_exec): Likewise.
1190 (<expander>v64di3): Likewise.
1191 (<expander>v64di3_exec): Likewise.
1192 (*<reduc_op>_dpp_shr_v64di): Likewise.
1193 (*plus_carry_dpp_shr_v64di): Likewise.
1194 * config/gcn/gcn.md (adddi3): Likewise.
1195 (addptrdi3): Likewise.
1196 (<expander>di3): Likewise.
1198 2020-02-21 Andrew Stubbs <ams@codesourcery.com>
1200 * config/gcn/gcn-valu.md (vec_seriesv64di): Use gen_vec_duplicatev64di.
1202 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
1204 * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Add SVE
1205 support. Use aarch64_emit_mult instead of emitting multiplication
1206 instructions directly.
1207 * config/aarch64/aarch64-sve.md (sqrt<mode>2, rsqrt<mode>2)
1208 (@aarch64_rsqrte<mode>, @aarch64_rsqrts<mode>): New expanders.
1210 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
1212 * config/aarch64/aarch64.c (aarch64_emit_mult): New function.
1213 (aarch64_emit_approx_div): Add SVE support. Use aarch64_emit_mult
1214 instead of emitting multiplication instructions directly.
1215 * config/aarch64/iterators.md (SVE_COND_FP_BINARY_OPTAB): New iterator.
1216 * config/aarch64/aarch64-sve.md (div<mode>3, @aarch64_frecpe<mode>)
1217 (@aarch64_frecps<mode>): New expanders.
1219 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
1221 * config/aarch64/aarch64-protos.h (AARCH64_APPROX_MODE): Operate
1222 on and produce uint64_ts rather than ints.
1223 (AARCH64_APPROX_NONE, AARCH64_APPROX_ALL): Change to uint64_ts.
1224 (cpu_approx_modes): Change the fields from unsigned int to uint64_t.
1226 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
1228 * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Don't create
1229 an unused xmsk register when handling approximate rsqrt.
1231 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
1233 * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Fix inverted
1234 flag_finite_math_only condition.
1236 2020-02-20 Uroš Bizjak <ubizjak@gmail.com>
1239 * config/i386/mmx.md (*vec_extractv2sf_1): Match source operand
1240 to destination operand for shufps alternative.
1241 (*vec_extractv2si_1): Ditto.
1243 2020-02-20 Peter Bergner <bergner@linux.ibm.com>
1246 * config/rs6000/rs6000.c (rs6000_legitimate_address_p): Handle VSX
1249 2020-02-20 Martin Liska <mliska@suse.cz>
1251 PR translation/93831
1252 * config/darwin.c (darwin_override_options): Change 64b to 64-bit mode.
1254 2020-02-20 Martin Liska <mliska@suse.cz>
1256 PR translation/93830
1257 * common/config/avr/avr-common.c: Remote trailing "|".
1259 2020-02-19 Bernd Edlinger <bernd.edlinger@hotmail.de>
1261 * collect2.c (maybe_run_lto_and_relink): Fix typo in
1264 2020-02-19 Richard Sandiford <richard.sandiford@arm.com>
1266 PR tree-optimization/93767
1267 * tree-vect-data-refs.c (vect_compile_time_alias): Remove the
1268 access-size bias from the offset calculations for negative strides.
1270 2020-02-19 Bernd Edlinger <bernd.edlinger@hotmail.de>
1272 * collect2.c (c_file, o_file): Make const again.
1273 (ldout,lderrout, dump_ld_file): Remove.
1274 (tool_cleanup): Avoid calling not signal-safe functions.
1275 (maybe_run_lto_and_relink): Avoid possible signal handler
1276 access to unintialzed memory (lto_o_files).
1277 (main): Avoid leaking temp files in $TMPDIR.
1278 Initialize c_file/o_file with concat, which avoids exposing
1279 uninitialized memory to signal handler, which calls unlink(!).
1280 Avoid calling maybe_unlink when the main function returns,
1281 since the atexit handler is already doing this.
1282 * collect2.h (dump_ld_file, ldout, lderrout): Remove.
1284 2020-02-19 Martin Jambor <mjambor@suse.cz>
1286 PR tree-optimization/93776
1287 * tree-sra.c (create_access): Do not create zero size accesses.
1288 (get_access_for_expr): Do not search for zero sized accesses.
1290 2020-02-19 Martin Jambor <mjambor@suse.cz>
1292 PR tree-optimization/93667
1293 * tree-sra.c (scalarizable_type_p): Return false if record fields
1294 do not follow wach other.
1296 2020-01-21 Kito Cheng <kito.cheng@sifive.com>
1298 * config/riscv/riscv.c (riscv_output_move) Using fmv.x.w/fmv.w.x
1299 rather than fmv.x.s/fmv.s.x.
1301 2020-02-18 James Greenhalgh <james.greenhalgh@arm.com>
1303 * config/aarch64/aarch64-simd-builtins.def
1304 (intrinsic_vec_smult_lo_): New.
1305 (intrinsic_vec_umult_lo_): Likewise.
1306 (vec_widen_smult_hi_): Likewise.
1307 (vec_widen_umult_hi_): Likewise.
1308 * config/aarch64/aarch64-simd.md
1309 (aarch64_intrinsic_vec_<su>mult_lo_<mode>): New.
1310 * config/aarch64/arm_neon.h (vmull_high_s8): Use intrinsics.
1311 (vmull_high_s16): Likewise.
1312 (vmull_high_s32): Likewise.
1313 (vmull_high_u8): Likewise.
1314 (vmull_high_u16): Likewise.
1315 (vmull_high_u32): Likewise.
1316 (vmull_s8): Likewise.
1317 (vmull_s16): Likewise.
1318 (vmull_s32): Likewise.
1319 (vmull_u8): Likewise.
1320 (vmull_u16): Likewise.
1321 (vmull_u32): Likewise.
1323 2020-02-18 Martin Liska <mliska@suse.cz>
1325 * value-prof.c (stream_out_histogram_value): Restore LTO PGO
1326 bootstrap by missing removal of invalid sanity check.
1328 2020-02-18 Martin Liska <mliska@suse.cz>
1331 * ipa-icf-gimple.c (func_checker::compare_gimple_assign):
1332 Always compare LHS of gimple_assign.
1334 2020-02-18 Martin Liska <mliska@suse.cz>
1337 * cgraph.c (cgraph_node::verify_node): Verify MALLOC attribute
1338 and return type of functions.
1339 * ipa-param-manipulation.c (ipa_param_adjustments::adjust_decl):
1340 Drop MALLOC attribute for void functions.
1341 * ipa-pure-const.c (funct_state_summary_t::duplicate): Drop
1342 malloc_state for a new VOID clone.
1344 2020-02-18 Martin Liska <mliska@suse.cz>
1347 * common.opt: Add -fprofile-reproducibility.
1348 * doc/invoke.texi: Document it.
1349 * value-prof.c (dump_histogram_value):
1350 Document and support behavior for counters[0]
1351 being a negative value.
1352 (get_nth_most_common_value): Handle negative
1353 counters[0] in respect to flag_profile_reproducible.
1355 2020-02-18 Jakub Jelinek <jakub@redhat.com>
1358 * cgraph.c (verify_speculative_call): Use speculative_id instead of
1359 speculative_uid in messages. Remove trailing whitespace from error
1360 message. Use num_speculative_call_targets instead of
1361 num_speculative_targets in a message.
1362 (cgraph_node::verify_node): Use call_stmt instead of cal_stmt in
1363 edge messages and stmt instead of cal_stmt in reference message.
1365 PR tree-optimization/93780
1366 * tree-ssa.c (non_rewritable_lvalue_p): Check valid_vector_subparts_p
1367 before calling build_vector_type.
1368 (execute_update_addresses_taken): Likewise.
1371 * params.opt (-param=ipa-max-switch-predicate-bounds=): Fix help
1372 typo, functoin -> function.
1373 * tree.c (free_lang_data_in_decl): Fix comment typo,
1374 functoin -> function.
1375 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
1377 2020-02-17 David Malcolm <dmalcolm@redhat.com>
1379 * diagnostic.c (print_any_cwe): Don't call get_cwe_url if URLs
1381 (print_option_information): Don't call get_option_url if URLs
1384 2020-02-17 Alexandre Oliva <oliva@adacore.com>
1386 * tree-emutls.c (new_emutls_decl, emutls_common_1): Complete
1387 handling of register_common-less targets.
1389 2020-02-17 Martin Liska <mliska@suse.cz>
1392 * ipa-devirt.c (odr_types_equivalent_p): Fix grammar.
1394 2020-02-17 Martin Liska <mliska@suse.cz>
1396 PR translation/93755
1397 * config/rs6000/rs6000.c (rs6000_option_override_internal):
1400 2020-02-17 Martin Liska <mliska@suse.cz>
1403 * config/rx/elf.opt: Fix typo.
1405 2020-02-17 Richard Biener <rguenther@suse.de>
1408 * opts-global.c (print_ignored_options): Use inform and
1411 2020-02-17 Jiufu Guo <guojiufu@linux.ibm.com>
1414 * config/rs6000/rs6000.md (untyped_call): Add emit_clobber.
1416 2020-02-16 Uroš Bizjak <ubizjak@gmail.com>
1419 * config/i386/i386.md (atan2xf3): Swap operands 1 and 2.
1420 (atan2<mode>3): Update operand order in the call to gen_atan2xf3.
1422 2020-02-15 Jason Merrill <jason@redhat.com>
1424 * doc/invoke.texi (C Dialect Options): Add -std=c++20.
1426 2020-02-15 Jakub Jelinek <jakub@redhat.com>
1428 PR tree-optimization/93744
1429 * match.pd (((m1 >/</>=/<= m2) * d -> (m1 >/</>=/<= m2) ? d : 0,
1430 A - ((A - B) & -(C cmp D)) -> (C cmp D) ? B : A,
1431 A + ((B - A) & -(C cmp D)) -> (C cmp D) ? B : A): For GENERIC, make
1432 sure @2 in the first and @1 in the other patterns has no side-effects.
1434 2020-02-15 David Malcolm <dmalcolm@redhat.com>
1435 Bernd Edlinger <bernd.edlinger@hotmail.de>
1439 * config.in (DIAGNOSTICS_URLS_DEFAULT): New define.
1440 * configure.ac (--with-diagnostics-urls): New configuration
1441 option, based on --with-diagnostics-color.
1442 (DIAGNOSTICS_URLS_DEFAULT): New define.
1443 * config.h: Regenerate.
1444 * configure: Regenerate.
1445 * diagnostic.c (diagnostic_urls_init): Handle -1 for
1446 DIAGNOSTICS_URLS_DEFAULT from configure-time
1447 --with-diagnostics-urls=auto-if-env by querying for a GCC_URLS
1448 and TERM_URLS environment variable.
1449 * diagnostic-url.h (diagnostic_url_format): New enum type.
1450 (diagnostic_urls_enabled_p): rename to...
1451 (determine_url_format): ... this, and change return type.
1452 * diagnostic-color.c (parse_env_vars_for_urls): New helper function.
1453 (auto_enable_urls): Disable URLs on xfce4-terminal, gnome-terminal,
1454 the linux console, and mingw.
1455 (diagnostic_urls_enabled_p): rename to...
1456 (determine_url_format): ... this, and adjust.
1457 * pretty-print.h (pretty_printer::show_urls): rename to...
1458 (pretty_printer::url_format): ... this, and change to enum.
1459 * pretty-print.c (pretty_printer::pretty_printer,
1460 pp_begin_url, pp_end_url, test_urls): Adjust.
1461 * doc/install.texi (--with-diagnostics-urls): Document the new
1462 configuration option.
1463 (--with-diagnostics-color): Document the existing interaction
1464 with GCC_COLORS better.
1465 * doc/invoke.texi (-fdiagnostics-urls): Add GCC_URLS and TERM_URLS
1466 vindex reference. Update description of defaults based on the above.
1467 (-fdiagnostics-color): Update description of how -fdiagnostics-color
1468 interacts with GCC_COLORS.
1470 2020-02-14 Eric Botcazou <ebotcazou@adacore.com>
1473 * config/sparc/sparc.c (eligible_for_call_delay): Test HAVE_GNU_LD in
1474 conjunction with TARGET_GNU_TLS in early return.
1476 2020-02-14 Alexander Monakov <amonakov@ispras.ru>
1478 * rtlanal.c (rtx_cost): Handle a SET up front. Avoid division if
1479 the mode is not wider than UNITS_PER_WORD.
1481 2020-02-14 Martin Jambor <mjambor@suse.cz>
1483 PR tree-optimization/93516
1484 * tree-sra.c (propagate_subaccesses_from_rhs): Do not create
1485 access of the same type as the parent.
1486 (propagate_subaccesses_from_lhs): Likewise.
1488 2020-02-14 Hongtao Liu <hongtao.liu@intel.com>
1491 * config/i386/avx512vbmi2intrin.h
1492 (_mm512_shrdi_epi16, _mm512_mask_shrdi_epi16,
1493 _mm512_maskz_shrdi_epi16, _mm512_shrdi_epi32,
1494 _mm512_mask_shrdi_epi32, _mm512_maskz_shrdi_epi32,
1495 _m512_shrdi_epi64, _m512_mask_shrdi_epi64,
1496 _m512_maskz_shrdi_epi64, _mm512_shldi_epi16,
1497 _mm512_mask_shldi_epi16, _mm512_maskz_shldi_epi16,
1498 _mm512_shldi_epi32, _mm512_mask_shldi_epi32,
1499 _mm512_maskz_shldi_epi32, _mm512_shldi_epi64,
1500 _mm512_mask_shldi_epi64, _mm512_maskz_shldi_epi64): Fix typo
1501 of lacking a closing parenthesis.
1502 * config/i386/avx512vbmi2vlintrin.h
1503 (_mm256_shrdi_epi16, _mm256_mask_shrdi_epi16,
1504 _mm256_maskz_shrdi_epi16, _mm256_shrdi_epi32,
1505 _mm256_mask_shrdi_epi32, _mm256_maskz_shrdi_epi32,
1506 _m256_shrdi_epi64, _m256_mask_shrdi_epi64,
1507 _m256_maskz_shrdi_epi64, _mm256_shldi_epi16,
1508 _mm256_mask_shldi_epi16, _mm256_maskz_shldi_epi16,
1509 _mm256_shldi_epi32, _mm256_mask_shldi_epi32,
1510 _mm256_maskz_shldi_epi32, _mm256_shldi_epi64,
1511 _mm256_mask_shldi_epi64, _mm256_maskz_shldi_epi64,
1512 _mm_shrdi_epi16, _mm_mask_shrdi_epi16,
1513 _mm_maskz_shrdi_epi16, _mm_shrdi_epi32,
1514 _mm_mask_shrdi_epi32, _mm_maskz_shrdi_epi32,
1515 _mm_shrdi_epi64, _mm_mask_shrdi_epi64,
1516 _m_maskz_shrdi_epi64, _mm_shldi_epi16,
1517 _mm_mask_shldi_epi16, _mm_maskz_shldi_epi16,
1518 _mm_shldi_epi32, _mm_mask_shldi_epi32,
1519 _mm_maskz_shldi_epi32, _mm_shldi_epi64,
1520 _mm_mask_shldi_epi64, _mm_maskz_shldi_epi64): Ditto.
1522 2020-02-13 H.J. Lu <hongjiu.lu@intel.com>
1525 * config/i386/i386.c (ix86_trampoline_init): Skip ENDBR32 at
1526 the target function entry.
1528 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
1530 * common/config/arc/arc-common.c (arc_option_optimization_table):
1531 Disable if-conversion step when optimized for size.
1533 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
1535 * config/arc/arc.c (arc_conditional_register_usage): R0-R3 and
1536 R12-R15 are always in ARCOMPACT16_REGS register class.
1537 * config/arc/arc.opt (mq-class): Deprecate.
1538 * config/arc/constraint.md ("q"): Remove dependency on mq-class
1540 * doc/invoke.texi (mq-class): Update text.
1541 * common/config/arc/arc-common.c (arc_option_optimization_table):
1544 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
1546 * config/arc/arc.c (arc_insn_cost): New function.
1547 (TARGET_INSN_COST): Define.
1548 * config/arc/arc.md (cost): New attribute.
1549 (add_n): Use arc_nonmemory_operand.
1550 (ashlsi3_insn): Likewise, also update constraints.
1551 (ashrsi3_insn): Likewise.
1552 (rotrsi3): Likewise.
1553 (add_shift): Likewise.
1554 * config/arc/predicates.md (arc_nonmemory_operand): New predicate.
1556 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
1558 * config/arc/arc.md (mulsidi_600): Correctly select mlo/mhi
1560 (umulsidi_600): Likewise.
1562 2020-02-13 Jakub Jelinek <jakub@redhat.com>
1565 * config/i386/avx512bitalgintrin.h (_mm512_mask_popcnt_epi8,
1566 _mm512_mask_popcnt_epi16, _mm256_mask_popcnt_epi8,
1567 _mm256_mask_popcnt_epi16, _mm_mask_popcnt_epi8,
1568 _mm_mask_popcnt_epi16): Rename __B argument to __A and __A to __W,
1569 pass __A to the builtin followed by __W instead of __A followed by
1571 * config/i386/avx512vpopcntdqintrin.h (_mm512_mask_popcnt_epi32,
1572 _mm512_mask_popcnt_epi64): Likewise.
1573 * config/i386/avx512vpopcntdqvlintrin.h (_mm_mask_popcnt_epi32,
1574 _mm256_mask_popcnt_epi32, _mm_mask_popcnt_epi64,
1575 _mm256_mask_popcnt_epi64): Likewise.
1577 PR tree-optimization/93582
1578 * fold-const.h (shift_bytes_in_array_left,
1579 shift_bytes_in_array_right): Declare.
1580 * fold-const.c (shift_bytes_in_array_left,
1581 shift_bytes_in_array_right): New function, moved from
1582 gimple-ssa-store-merging.c, no longer static.
1583 * gimple-ssa-store-merging.c (shift_bytes_in_array): Move
1584 to gimple-ssa-store-merging.c and rename to shift_bytes_in_array_left.
1585 (shift_bytes_in_array_right): Move to gimple-ssa-store-merging.c.
1586 (encode_tree_to_bitpos): Use shift_bytes_in_array_left instead of
1587 shift_bytes_in_array.
1588 (verify_shift_bytes_in_array): Rename to ...
1589 (verify_shift_bytes_in_array_left): ... this. Use
1590 shift_bytes_in_array_left instead of shift_bytes_in_array.
1591 (store_merging_c_tests): Call verify_shift_bytes_in_array_left
1592 instead of verify_shift_bytes_in_array.
1593 * tree-ssa-sccvn.c (vn_reference_lookup_3): For native_encode_expr
1594 / native_interpret_expr where the store covers all needed bits,
1595 punt on PDP-endian, otherwise allow all involved offsets and sizes
1596 not to be byte-aligned.
1599 * config/i386/sse.md (k<code><mode>): Drop mode from last operand and
1600 use const_0_to_255_operand predicate instead of immediate_operand.
1601 (avx512dq_fpclass<mode><mask_scalar_merge_name>,
1602 avx512dq_vmfpclass<mode><mask_scalar_merge_name>,
1603 vgf2p8affineinvqb_<mode><mask_name>,
1604 vgf2p8affineqb_<mode><mask_name>): Drop mode from
1605 const_0_to_255_operand predicated operands.
1607 2020-02-12 Jeff Law <law@redhat.com>
1609 * config/h8300/h8300.md (comparison shortening peepholes): Use
1610 a mode iterator to merge the HImode and SImode peepholes.
1612 2020-02-12 Jakub Jelinek <jakub@redhat.com>
1615 * real.c (is_even): Make static. Function comment fix.
1616 (is_halfway_below): Make static, don't assert R is not inf/nan,
1617 instead return false for those. Small formatting fixes.
1619 2020-02-12 Martin Sebor <msebor@redhat.com>
1622 * tree-ssa-strlen.c (handle_builtin_stxncpy): Rename...
1623 (handle_builtin_stxncpy_strncat): ...to this. Change first argument.
1624 Issue only -Wstringop-overflow strncat, never -Wstringop-truncation.
1625 (strlen_check_and_optimize_call): Adjust callee name.
1627 2020-02-12 Jeff Law <law@redhat.com>
1629 * config/h8300/h8300.md (comparison shortening peepholes): Drop
1630 (and (xor)) variant. Combine other two into single peephole.
1632 2020-02-12 Wilco Dijkstra <wdijkstr@arm.com>
1634 PR rtl-optimization/93565
1635 * config/aarch64/aarch64.c (aarch64_rtx_costs): Add CTZ costs.
1637 2020-02-12 Wilco Dijkstra <wdijkstr@arm.com>
1639 * config/aarch64/aarch64-simd.md
1640 (aarch64_zero_extend<GPI:mode>_reduc_plus_<VDQV_E:mode>): New pattern.
1641 * config/aarch64/aarch64.md (popcount<mode>2): Use it instead of
1642 generating separate ADDV and zero_extend patterns.
1643 * config/aarch64/iterators.md (VDQV_E): New iterator.
1645 2020-02-12 Jeff Law <law@redhat.com>
1647 * config/h8300/h8300.md (cpymemsi, movmd): Remove dead patterns,
1648 expanders, splits, etc.
1649 (movmd_internal_<mode>, movmd splitter, movstr, movsd): Likewise.
1650 (stpcpy_internal_<mode>, stpcpy splitter): Likewise.
1651 (peepholes to convert QI/HI mode pushes to SI mode pushes): Likewise.
1652 * config/h8300/h8300.c (h8300_swap_into_er6): Remove unused function.
1653 (h8300_swap_out_of_er6, h8sx_emit_movmd): Likewise
1654 * config/h8300/h8300-protos.h (h8300_swap_into_er6): Remove unused
1656 (h8300_swap_out_of_er6, h8sx_emit_movmd): Likewise.
1658 2020-02-12 Jakub Jelinek <jakub@redhat.com>
1661 * config/i386/sse.md (VI48F_256_DQ): New mode iterator.
1662 (avx512vl_vextractf128<mode>): Use it instead of VI48F_256. Remove
1663 TARGET_AVX512DQ from condition.
1664 (vec_extract_lo_<mode><mask_name>): Use <mask_avx512dq_condition>
1665 instead of <mask_mode512bit_condition> in condition. If
1666 TARGET_AVX512DQ is false, emit vextract*64x4 instead of
1668 (vec_extract_lo_<mode><mask_name>): Drop <mask_avx512dq_condition>
1671 2020-02-12 Kewen Lin <linkw@gcc.gnu.org>
1674 * ira.c (combine_and_move_insns): Skip multiple_sets def_insn.
1676 2020-02-12 Segher Boessenkool <segher@kernel.crashing.org>
1678 * config/rs6000/rs6000.c (rs6000_debug_print_mode): Don't use sizeof
1679 where strlen is more legible.
1680 (rs6000_builtin_vectorized_libmass): Ditto.
1681 (rs6000_print_options_internal): Ditto.
1683 2020-02-11 Martin Sebor <msebor@redhat.com>
1685 PR tree-optimization/93683
1686 * tree-ssa-alias.c (stmt_kills_ref_p): Avoid using LHS when not set.
1688 2020-02-11 Michael Meissner <meissner@linux.ibm.com>
1690 * config/rs6000/predicates.md (cint34_operand): Rename the
1691 -mprefixed-addr option to be -mprefixed.
1692 * config/rs6000/rs6000-cpus.def (ISA_FUTURE_MASKS_SERVER): Rename
1693 the -mprefixed-addr option to be -mprefixed.
1694 (OTHER_FUTURE_MASKS): Likewise.
1695 (POWERPC_MASKS): Likewise.
1696 * config/rs6000/rs6000.c (rs6000_option_override_internal): Rename
1697 the -mprefixed-addr option to be -mprefixed. Change error
1698 messages to refer to -mprefixed.
1699 (num_insns_constant_gpr): Rename the -mprefixed-addr option to be
1701 (rs6000_legitimate_offset_address_p): Likewise.
1702 (rs6000_mode_dependent_address): Likewise.
1703 (rs6000_opt_masks): Change the spelling of "-mprefixed-addr" to be
1704 "-mprefixed" for target attributes and pragmas.
1705 (address_to_insn_form): Rename the -mprefixed-addr option to be
1707 (rs6000_adjust_insn_length): Likewise.
1708 * config/rs6000/rs6000.h (FINAL_PRESCAN_INSN): Rename the
1709 -mprefixed-addr option to be -mprefixed.
1710 (ASM_OUTPUT_OPCODE): Likewise.
1711 * config/rs6000/rs6000.md (prefixed insn attribute): Rename the
1712 -mprefixed-addr option to be -mprefixed.
1713 * config/rs6000/rs6000.opt (-mprefixed): Rename the
1714 -mprefixed-addr option to be prefixed. Change the option from
1715 being undocumented to being documented.
1716 * doc/invoke.texi (RS/6000 and PowerPC Options): Document the
1717 -mprefixed option. Update the -mpcrel documentation to mention
1720 2020-02-11 Hans-Peter Nilsson <hp@axis.com>
1722 * ira-conflicts.c (print_hard_reg_set): Correct output for sets
1723 including FIRST_PSEUDO_REGISTER - 1.
1724 * ira-color.c (print_hard_reg_set): Ditto.
1726 2020-02-11 Stam Markianos-Wright <stam.markianos-wright@arm.com>
1728 * config/arm/arm-builtins.c (enum arm_type_qualifiers):
1729 (USTERNOP_QUALIFIERS): New define.
1730 (USMAC_LANE_QUADTUP_QUALIFIERS): New define.
1731 (SUMAC_LANE_QUADTUP_QUALIFIERS): New define.
1732 (arm_expand_builtin_args): Add case ARG_BUILTIN_LANE_QUADTUP_INDEX.
1733 (arm_expand_builtin_1): Add qualifier_lane_quadtup_index.
1734 * config/arm/arm_neon.h (vusdot_s32): New.
1735 (vusdot_lane_s32): New.
1736 (vusdotq_lane_s32): New.
1737 (vsudot_lane_s32): New.
1738 (vsudotq_lane_s32): New.
1739 * config/arm/arm_neon_builtins.def (usdot, usdot_lane,sudot_lane): New.
1740 * config/arm/iterators.md (DOTPROD_I8MM): New.
1741 (sup, opsuffix): Add <us/su>.
1742 * config/arm/neon.md (neon_usdot, <us/su>dot_lane: New.
1743 * config/arm/unspecs.md (UNSPEC_DOT_US, UNSPEC_DOT_SU): New.
1745 2020-02-11 Richard Biener <rguenther@suse.de>
1747 PR tree-optimization/93661
1748 PR tree-optimization/93662
1749 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly guard
1751 * tree-sra.c (get_access_for_expr): Likewise.
1753 2020-02-10 Jakub Jelinek <jakub@redhat.com>
1756 * config/i386/sse.md (VI_256_AVX2): New mode iterator.
1757 (vcond_mask_<mode><sseintvecmodelower>): Use it instead of VI_256.
1758 Change condition from TARGET_AVX2 to TARGET_AVX.
1760 2020-02-10 Iain Sandoe <iain@sandoe.co.uk>
1763 * config/darwin-c.c (darwin_cfstring_ref_p): Fix up last
1764 argument of strncmp.
1766 2020-02-10 Hans-Peter Nilsson <hp@axis.com>
1768 Try to generate zero-based comparisons.
1769 * config/cris/cris.c (cris_reduce_compare): New function.
1770 * config/cris/cris-protos.h (cris_reduce_compare): Add prototype.
1771 * config/cris/cris.md ("cbranch<mode>4", "cbranchdi4", "cstoredi4")
1772 (cstore<mode>4"): Apply cris_reduce_compare in expanders.
1774 2020-02-10 Richard Earnshaw <rearnsha@arm.com>
1777 * config/arm/arm.md (movsi_compare0): Allow SP as a source register
1778 in Thumb state and also as a destination in Arm state. Add T16
1781 2020-02-10 Hans-Peter Nilsson <hp@axis.com>
1783 * md.texi (Define Subst): Match closing paren in example.
1785 2020-02-10 Jakub Jelinek <jakub@redhat.com>
1789 * config/i386/i386.c (x86_64_elf_section_type_flags): Fix up last
1790 arguments of strncmp.
1792 2020-02-10 Feng Xue <fxue@os.amperecomputing.com>
1795 * ipa-cp.c (ipcp_lattice::add_value): Add source with same call edge
1796 but different source value.
1797 (adjust_callers_for_value_intersection): New function.
1798 (gather_edges_for_value): Adjust order of callers to let a
1799 non-self-recursive caller be the first element.
1800 (self_recursive_pass_through_p): Add a new parameter "simple", and
1801 check generalized self-recursive pass-through jump function.
1802 (self_recursive_agg_pass_through_p): Likewise.
1803 (find_more_scalar_values_for_callers_subset): Compute value from
1804 pass-through jump function for self-recursive.
1805 (intersect_with_plats): Cleanup previous implementation code for value
1806 itersection with self-recursive call edge.
1807 (intersect_with_agg_replacements): Likewise.
1808 (intersect_aggregates_with_edge): Deduce value from pass-through jump
1809 function for self-recursive call edge. Cleanup previous implementation
1810 code for value intersection with self-recursive call edge.
1811 (decide_whether_version_node): Remove dead callers and adjust order
1812 to let a non-self-recursive caller be the first element.
1814 2020-02-09 Uroš Bizjak <ubizjak@gmail.com>
1816 * recog.c: Move pass_split_before_sched2 code in front of
1817 pass_split_before_regstack.
1818 (pass_data_split_before_sched2): Rename pass to split3 from split4.
1819 (pass_data_split_before_regstack): Rename pass to split4 from split3.
1820 (rest_of_handle_split_before_sched2): Remove.
1821 (pass_split_before_sched2::execute): Unconditionally call
1823 (enable_split_before_sched2): New function.
1824 (pass_split_before_sched2::gate): Use enable_split_before_sched2.
1825 (pass_split_before_regstack::gate): Ditto.
1826 * config/nds32/nds32.c (nds32_split_double_word_load_store_p):
1827 Update name check for renamed split4 pass.
1828 * config/sh/sh.c (register_sh_passes): Update pass insertion
1829 point for renamed split4 pass.
1831 2020-02-09 Jakub Jelinek <jakub@redhat.com>
1833 * gimplify.c (gimplify_adjust_omp_clauses_1): Promote
1834 DECL_IN_CONSTANT_POOL variables into "omp declare target" to avoid
1835 copying them around between host and target.
1837 2020-02-08 Andrew Pinski <apinski@marvell.com>
1840 * config/aarch64/aarch64-simd.md (movmisalign<mode>): Check
1841 STRICT_ALIGNMENT also.
1843 2020-02-08 Jim Wilson <jimw@sifive.com>
1846 * config/riscv/riscv.h (HARD_REGNO_CALLER_SAVE_MODE): Define.
1848 2020-02-08 Uroš Bizjak <ubizjak@gmail.com>
1849 Jakub Jelinek <jakub@redhat.com>
1852 * config/i386/i386.h (CALL_USED_REGISTERS): Make
1853 xmm16-xmm31 call-used even in 64-bit ms-abi.
1855 2020-02-07 Dennis Zhang <dennis.zhang@arm.com>
1857 * config/aarch64/aarch64-simd-builtins.def (simd_smmla): New entry.
1858 (simd_ummla, simd_usmmla): Likewise.
1859 * config/aarch64/aarch64-simd.md (aarch64_simd_<sur>mmlav16qi): New.
1860 * config/aarch64/arm_neon.h (vmmlaq_s32, vmmlaq_u32): New.
1861 (vusmmlaq_s32): New.
1863 2020-02-07 Richard Biener <rguenther@suse.de>
1866 * tree-inline.c (fold_marked_statements): Do a PRE walk,
1867 skipping unreachable regions.
1868 (optimize_inline_calls): Skip folding stmts when we didn't
1871 2020-02-07 H.J. Lu <hongjiu.lu@intel.com>
1874 * config/i386/i386.c (function_arg_ms_64): Add a type argument.
1875 Don't return aggregates with only SFmode and DFmode in SSE
1877 (ix86_function_arg): Pass arg.type to function_arg_ms_64.
1879 2020-02-07 Jakub Jelinek <jakub@redhat.com>
1882 * config/rs6000/rs6000-logue.c
1883 (rs6000_emit_probe_stack_range_stack_clash): Always use gen_add3_insn,
1884 if it fails, move rs into end_addr and retry. Add
1885 REG_FRAME_RELATED_EXPR note whenever it returns more than one insn or
1886 the insn pattern doesn't describe well what exactly happens to
1890 * config/i386/predicates.md (avx_identity_operand): Remove.
1891 * config/i386/sse.md (*avx_vec_concat<mode>_1): Remove.
1892 (avx_<castmode><avxsizesuffix>_<castmode>,
1893 avx512f_<castmode><avxsizesuffix>_256<castmode>): Change patterns to
1894 a VEC_CONCAT of the operand and UNSPEC_CAST.
1895 (avx512f_<castmode><avxsizesuffix>_<castmode>): Change pattern to
1896 a VEC_CONCAT of VEC_CONCAT of the operand and UNSPEC_CAST with
1900 * config/i386/i386.c (ix86_lea_outperforms): Make sure to clear
1901 recog_data.insn if distance_non_agu_define changed it.
1903 2020-02-06 Michael Meissner <meissner@linux.ibm.com>
1906 * config/rs6000/rs6000.c (reg_to_non_prefixed): Before ISA 3.0
1907 we only had X-FORM (reg+reg) addressing for vectors. Also before
1908 ISA 3.0, we only had X-FORM addressing for scalars in the
1909 traditional Altivec registers.
1911 2020-02-06 <zhongyunde@huawei.com>
1912 Vladimir Makarov <vmakarov@redhat.com>
1914 PR rtl-optimization/93561
1915 * lra-assigns.c (spill_for): Check that tested hard regno is not out of
1916 hard register range.
1918 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
1920 * config/aarch64/aarch64.md (aarch64_movk<mode>): Add a type
1923 2020-02-06 Segher Boessenkool <segher@kernel.crashing.org>
1925 * config/rs6000/rs6000.c (rs6000_emit_set_long_const): Handle the case
1926 where the low and the high 32 bits are equal to each other specially,
1927 with an rldimi instruction.
1929 2020-02-06 Mihail Ionescu <mihail.ionescu@arm.com>
1931 * config/arm/arm-cpus.in: Set profile M for armv8.1-m.main.
1933 2020-02-06 Mihail Ionescu <mihail.ionescu@arm.com>
1935 * config/arm/arm-tables.opt: Regenerate.
1937 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
1940 * config/aarch64/aarch64-protos.h (aarch64_movk_shift): Declare.
1941 * config/aarch64/aarch64.c (aarch64_movk_shift): New function.
1942 * config/aarch64/aarch64.md (aarch64_movk<mode>): New pattern.
1944 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
1946 PR rtl-optimization/87763
1947 * config/aarch64/aarch64.md (*ashiftsi_extvdi_bfiz): New pattern.
1949 2020-02-06 Delia Burduv <delia.burduv@arm.com>
1951 * config/aarch64/aarch64-simd-builtins.def
1952 (bfmlaq): New built-in function.
1953 (bfmlalb): New built-in function.
1954 (bfmlalt): New built-in function.
1955 (bfmlalb_lane): New built-in function.
1956 (bfmlalt_lane): New built-in function.
1957 * config/aarch64/aarch64-simd.md
1958 (aarch64_bfmmlaqv4sf): New pattern.
1959 (aarch64_bfmlal<bt>v4sf): New pattern.
1960 (aarch64_bfmlal<bt>_lane<q>v4sf): New pattern.
1961 * config/aarch64/arm_neon.h (vbfmmlaq_f32): New intrinsic.
1962 (vbfmlalbq_f32): New intrinsic.
1963 (vbfmlaltq_f32): New intrinsic.
1964 (vbfmlalbq_lane_f32): New intrinsic.
1965 (vbfmlaltq_lane_f32): New intrinsic.
1966 (vbfmlalbq_laneq_f32): New intrinsic.
1967 (vbfmlaltq_laneq_f32): New intrinsic.
1968 * config/aarch64/iterators.md (BF_MLA): New int iterator.
1969 (bt): New int attribute.
1971 2020-02-06 Uroš Bizjak <ubizjak@gmail.com>
1973 * config/i386/i386.md (*pushtf): Emit "#" instead of
1974 calling gcc_unreachable in insn output.
1977 (*pushsf_rex64): Ditto for alternatives other than 1.
1978 (*pushsf): Ditto for alternatives other than 1.
1980 2020-02-06 Martin Liska <mliska@suse.cz>
1982 PR gcov-profile/91971
1983 PR gcov-profile/93466
1984 * coverage.c (coverage_init): Revert mangling of
1985 path into filename. It can lead to huge filename length.
1986 Creation of subfolders seem more natural.
1988 2020-02-06 Stam Markianos-Wright <stam.markianos-wright@arm.com>
1991 * config/arm/arm.c (arm_block_arith_comp_libfuncs_for_mode): New.
1992 (arm_init_libfuncs): Add BFmode support to block spurious BF libfuncs.
1993 Use arm_block_arith_comp_libfuncs_for_mode for HFmode.
1995 2020-02-06 Jakub Jelinek <jakub@redhat.com>
1998 * config/i386/predicates.md (avx_identity_operand): New predicate.
1999 * config/i386/sse.md (*avx_vec_concat<mode>_1): New
2000 define_insn_and_split.
2003 * omp-low.c (use_pointer_for_field): For nested constructs, also
2004 look for map clauses on target construct.
2005 (scan_omp_1_stmt) <case GIMPLE_OMP_TARGET>: Bump temporarily
2006 taskreg_nesting_level.
2009 * gimplify.c (gimplify_scan_omp_clauses) <do_notice>: If adding
2010 shared clause, call omp_notice_variable on outer context if any.
2012 2020-02-05 Jason Merrill <jason@redhat.com>
2015 * symtab.c (symtab_node::nonzero_address): A DECL_COMDAT decl has
2016 non-zero address even if weak and not yet defined.
2018 2020-02-05 Martin Sebor <msebor@redhat.com>
2020 PR tree-optimization/92765
2021 * gimple-fold.c (get_range_strlen_tree): Handle MEM_REF and PARM_DECL.
2022 * tree-ssa-strlen.c (compute_string_length): Remove.
2023 (determine_min_objsize): Remove.
2024 (get_len_or_size): Add an argument. Call get_range_strlen_dynamic.
2025 Avoid using type size as the upper bound on string length.
2026 (handle_builtin_string_cmp): Add an argument. Adjust.
2027 (strlen_check_and_optimize_call): Pass additional argument to
2028 handle_builtin_string_cmp.
2030 2020-02-05 Uroš Bizjak <ubizjak@gmail.com>
2032 * config/i386/i386.md (*pushdi2_rex64 peephole2): Remove.
2033 (*pushdi2_rex64 peephole2): Unconditionally split after
2035 (*ashl<mode>3_doubleword): Ditto.
2036 (*<shift_insn><mode>3_doubleword): Ditto.
2038 2020-02-05 Michael Meissner <meissner@linux.ibm.com>
2041 * config/rs6000/rs6000.c (get_vector_offset): Fix
2043 2020-02-05 Andrew Stubbs <ams@codesourcery.com>
2045 * config/gcn/t-gcn-hsa (MULTILIB_OPTIONS): Use / not space.
2047 2020-02-05 David Malcolm <dmalcolm@redhat.com>
2050 (Special Functions for Debugging the Analyzer): Update description
2051 of __analyzer_dump_exploded_nodes.
2053 2020-02-05 Jakub Jelinek <jakub@redhat.com>
2056 * config/i386/i386-features.c (ix86_add_reg_usage_to_vzeroupper): Only
2057 include sets and not clobbers in the vzeroupper pattern.
2058 * config/i386/sse.md (*avx_vzeroupper): Require in insn condition that
2059 the parallel has 17 (64-bit) or 9 (32-bit) elts.
2060 (*avx_vzeroupper_1): New define_insn_and_split.
2063 * recog.c (pass_split_after_reload::gate): For STACK_REGS targets,
2064 don't run when !optimize.
2065 (pass_split_before_regstack::gate): For STACK_REGS targets, run even
2068 2020-02-05 Richard Biener <rguenther@suse.de>
2071 * genmatch.c (dt_node::gen_kids_1): Emit number of argument
2072 checks before matching calls.
2074 2020-02-05 Jakub Jelinek <jakub@redhat.com>
2076 * tree-ssa-alias.c (aliasing_matching_component_refs_p): Fix up
2077 function comment typo.
2080 * omp-simd-clone.c (expand_simd_clones): If simd_clone_mangle or
2081 simd_clone_create failed when i == 0, adjust clone->nargs by
2084 2020-02-05 Martin Liska <mliska@suse.cz>
2087 * doc/invoke.texi: Document that one should
2088 not combine ASLR and -fpch.
2090 2020-02-04 Richard Biener <rguenther@suse.de>
2092 PR tree-optimization/93538
2093 * match.pd (addr EQ/NE ptr): Amend to handle &ptr->x EQ/NE ptr.
2095 2020-02-04 Richard Biener <rguenther@suse.de>
2097 PR tree-optimization/91123
2098 * tree-ssa-sccvn.c (vn_walk_cb_data::finish): New method.
2099 (vn_walk_cb_data::last_vuse): New member.
2100 (vn_walk_cb_data::saved_operands): Likewsie.
2101 (vn_walk_cb_data::~vn_walk_cb_data): Release saved_operands.
2102 (vn_walk_cb_data::push_partial_def): Use finish.
2103 (vn_reference_lookup_2): Update last_vuse and use finish if
2104 we've saved operands.
2105 (vn_reference_lookup_3): Use finish and update calls to
2106 push_partial_defs everywhere. When translating through
2107 memcpy or aggregate copies save off operands and alias-set.
2108 (eliminate_dom_walker::eliminate_stmt): Restore VN_WALKREWRITE
2109 operation for redundant store removal.
2111 2020-02-04 Richard Biener <rguenther@suse.de>
2113 PR tree-optimization/92819
2114 * tree-ssa-forwprop.c (simplify_vector_constructor): Avoid
2115 generating more stmts than before.
2117 2020-02-04 Martin Liska <mliska@suse.cz>
2119 * config/arm/arm.c (arm_gen_far_branch): Move the function
2120 outside of selftests.
2122 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
2124 * config/rs6000/rs6000.c (adjust_vec_address_pcrel): New helper
2125 function to adjust PC-relative vector addresses.
2126 (rs6000_adjust_vec_address): Call adjust_vec_address_pcrel to
2127 handle vectors with PC-relative addresses.
2129 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
2131 * config/rs6000/rs6000.c (reg_to_non_prefixed): Add forward
2133 (hard_reg_and_mode_to_addr_mask): Delete.
2134 (rs6000_adjust_vec_address): If the original vector address
2135 was REG+REG or REG+OFFSET and the element is not zero, do the add
2136 of the elements in the original address before adding the offset
2137 for the vector element. Use address_to_insn_form to validate the
2138 address using the register being loaded, rather than guessing
2139 whether the address is a DS-FORM or DQ-FORM address.
2141 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
2143 * config/rs6000/rs6000.c (get_vector_offset): New helper function
2144 to calculate the offset in memory from the start of a vector of a
2145 particular element. Add code to keep the element number in
2146 bounds if the element number is variable.
2147 (rs6000_adjust_vec_address): Move calculation of offset of the
2148 vector element to get_vector_offset.
2149 (rs6000_split_vec_extract_var): Do not do the initial AND of
2150 element here, move the code to get_vector_offset.
2152 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
2154 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add some
2157 2020-02-03 Segher Boessenkool <segher@kernel.crashing.org>
2159 * config/rs6000/constraints.md: Improve documentation.
2161 2020-02-03 Richard Earnshaw <rearnsha@arm.com>
2164 * config/arm/t-arm: ($(srcdir)/config/arm/arm-tune.md)
2165 ($(srcdir)/config/arm/arm-tables.opt): Use move-if-change.
2167 2020-02-03 Andrew Stubbs <ams@codesourcery.com>
2169 * config.gcc: Remove "carrizo" support.
2170 * config/gcn/gcn-opts.h (processor_type): Likewise.
2171 * config/gcn/gcn.c (gcn_omp_device_kind_arch_isa): Likewise.
2172 * config/gcn/gcn.opt (gpu_type): Likewise.
2173 * config/gcn/t-omp-device: Likewise.
2175 2020-02-03 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2178 * config/arm/arm-protos.h: New function arm_gen_far_branch prototype.
2179 * config/arm/arm.c (arm_gen_far_branch): New function
2181 * config/arm/arm.md: Update b<cond> for Thumb2 range checks.
2183 2020-02-03 Julian Brown <julian@codesourcery.com>
2184 Tobias Burnus <tobias@codesourcery.com>
2186 * doc/invoke.texi: Update mention of OpenACC version to 2.6.
2188 2020-02-03 Jakub Jelinek <jakub@redhat.com>
2191 * config/s390/s390.md (popcounthi2_z196): Fix up expander to emit
2192 valid RTL to sum up the lowest and second lowest bytes of the popcnt
2195 2020-02-02 Vladimir Makarov <vmakarov@redhat.com>
2197 PR rtl-optimization/91333
2198 * ira-color.c (struct allocno_color_data): Add member
2200 (init_allocno_threads): Set the member up.
2201 (bucket_allocno_compare_func): Add compare hard reg
2204 2020-01-31 Sandra Loosemore <sandra@codesourcery.com>
2206 nios2: Support for GOT-relative DW_EH_PE_datarel encoding.
2208 * configure.ac [nios2-*-*]: Check HAVE_AS_NIOS2_GOTOFF_RELOCATION.
2209 * config.in: Regenerated.
2210 * configure: Regenerated.
2211 * config/nios2/nios2.h (ASM_PREFERRED_EH_DATA_FORMAT): Fix handling
2212 for PIC when HAVE_AS_NIOS2_GOTOFF_RELOCATION.
2213 (ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX): New.
2215 2020-02-01 Andrew Burgess <andrew.burgess@embecosm.com>
2217 * configure: Regenerate.
2219 2020-01-31 Vladimir Makarov <vmakarov@redhat.com>
2221 PR rtl-optimization/91333
2222 * ira-color.c (bucket_allocno_compare_func): Move conflict hard
2223 reg preferences comparison up.
2225 2020-01-31 Richard Sandiford <richard.sandiford@arm.com>
2227 * config/aarch64/aarch64.h (TARGET_SVE_BF16): New macro.
2228 * config/aarch64/aarch64-sve-builtins-sve2.h (svcvtnt): Move to
2229 aarch64-sve-builtins-base.h.
2230 * config/aarch64/aarch64-sve-builtins-sve2.cc (svcvtnt): Move to
2231 aarch64-sve-builtins-base.cc.
2232 * config/aarch64/aarch64-sve-builtins-base.h (svbfdot, svbfdot_lane)
2233 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
2235 * config/aarch64/aarch64-sve-builtins-base.cc (svbfdot, svbfdot_lane)
2236 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
2237 (svcvtnt): New functions.
2238 * config/aarch64/aarch64-sve-builtins-base.def (svbfdot, svbfdot_lane)
2239 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
2240 (svcvtnt): New functions.
2241 (svcvt): Add a form that converts f32 to bf16.
2242 * config/aarch64/aarch64-sve-builtins-shapes.h (ternary_bfloat)
2243 (ternary_bfloat_lane, ternary_bfloat_lanex2, ternary_bfloat_opt_n):
2245 * config/aarch64/aarch64-sve-builtins-shapes.cc (parse_element_type):
2246 Treat B as bfloat16_t.
2247 (ternary_bfloat_lane_base): New class.
2248 (ternary_bfloat_def): Likewise.
2249 (ternary_bfloat): New shape.
2250 (ternary_bfloat_lane_def): New class.
2251 (ternary_bfloat_lane): New shape.
2252 (ternary_bfloat_lanex2_def): New class.
2253 (ternary_bfloat_lanex2): New shape.
2254 (ternary_bfloat_opt_n_def): New class.
2255 (ternary_bfloat_opt_n): New shape.
2256 * config/aarch64/aarch64-sve-builtins.cc (TYPES_cvt_bfloat): New macro.
2257 * config/aarch64/aarch64-sve.md (@aarch64_sve_<sve_fp_op>vnx4sf)
2258 (@aarch64_sve_<sve_fp_op>_lanevnx4sf): New patterns.
2259 (@aarch64_sve_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>)
2260 (@cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
2261 (*cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
2262 (@aarch64_sve_cvtnt<VNx8BF_ONLY:mode>): Likewise.
2263 * config/aarch64/aarch64-sve2.md (@aarch64_sve2_cvtnt<mode>): Key
2264 the pattern off the narrow mode instead of the wider one.
2265 * config/aarch64/iterators.md (VNx8BF_ONLY): New mode iterator.
2266 (UNSPEC_BFMLALB, UNSPEC_BFMLALT, UNSPEC_BFMMLA): New unspecs.
2267 (sve_fp_op): Handle them.
2268 (SVE_BFLOAT_TERNARY_LONG): New int itertor.
2269 (SVE_BFLOAT_TERNARY_LONG_LANE): Likewise.
2271 2020-01-31 Richard Sandiford <richard.sandiford@arm.com>
2273 * config/aarch64/arm_sve.h: Include arm_bf16.h.
2274 * config/aarch64/aarch64-modes.def (BF): Move definition before
2275 VECTOR_MODES. Remove separate VECTOR_MODES for V4BF and V8BF.
2276 (SVE_MODES): Handle BF modes.
2277 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
2279 (aarch64_full_sve_mode): Likewise.
2280 * config/aarch64/iterators.md (SVE_STRUCT): Add VNx16BF, VNx24BF
2282 (SVE_FULL, SVE_FULL_HSD, SVE_ALL): Add VNx8BF.
2283 (Vetype, Vesize, Vctype, VEL, Vel, VEL_INT, V128, v128, vwcore)
2284 (V_INT_EQUIV, v_int_equiv, V_FP_EQUIV, v_fp_equiv, vector_count)
2285 (insn_length, VSINGLE, vsingle, VPRED, vpred, VDOUBLE): Handle the
2287 * config/aarch64/aarch64-sve-builtins.h (TYPE_bfloat): New
2289 * config/aarch64/aarch64-sve-builtins.cc (TYPES_all_arith): New macro.
2290 (TYPES_all_data): Add bf16.
2291 (TYPES_reinterpret1, TYPES_reinterpret): Likewise.
2292 (register_tuple_type): Increase buffer size.
2293 * config/aarch64/aarch64-sve-builtins.def (svbfloat16_t): New type.
2294 (bf16): New type suffix.
2295 * config/aarch64/aarch64-sve-builtins-base.def (svabd, svadd, svaddv)
2296 (svcmpeq, svcmpge, svcmpgt, svcmple, svcmplt, svcmpne, svmad, svmax)
2297 (svmaxv, svmin, svminv, svmla, svmls, svmsb, svmul, svsub, svsubr):
2298 Change type from all_data to all_arith.
2299 * config/aarch64/aarch64-sve-builtins-sve2.def (svaddp, svmaxp)
2302 2020-01-31 Dennis Zhang <dennis.zhang@arm.com>
2303 Matthew Malcomson <matthew.malcomson@arm.com>
2304 Richard Sandiford <richard.sandiford@arm.com>
2306 * doc/invoke.texi (f32mm): Document new AArch64 -march= extension.
2307 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Define
2308 __ARM_FEATURE_SVE_MATMUL_INT8, __ARM_FEATURE_SVE_MATMUL_FP32 and
2309 __ARM_FEATURE_SVE_MATMUL_FP64 as appropriate. Don't define
2310 __ARM_FEATURE_MATMUL_FP64.
2311 * config/aarch64/aarch64-option-extensions.def (fp, simd, fp16)
2312 (sve): Add AARCH64_FL_F32MM to the list of extensions that should
2313 be disabled at the same time.
2314 (f32mm): New extension.
2315 * config/aarch64/aarch64.h (AARCH64_FL_F32MM): New macro.
2316 (AARCH64_FL_F64MM): Bump to the next bit up.
2317 (AARCH64_ISA_F32MM, TARGET_SVE_I8MM, TARGET_F32MM, TARGET_SVE_F32MM)
2318 (TARGET_SVE_F64MM): New macros.
2319 * config/aarch64/iterators.md (SVE_MATMULF): New mode iterator.
2320 (UNSPEC_FMMLA, UNSPEC_SMATMUL, UNSPEC_UMATMUL, UNSPEC_USMATMUL)
2321 (UNSPEC_TRN1Q, UNSPEC_TRN2Q, UNSPEC_UZP1Q, UNSPEC_UZP2Q, UNSPEC_ZIP1Q)
2322 (UNSPEC_ZIP2Q): New unspeccs.
2323 (DOTPROD_US_ONLY, PERMUTEQ, MATMUL, FMMLA): New int iterators.
2324 (optab, sur, perm_insn): Handle the new unspecs.
2325 (sve_fp_op): Handle UNSPEC_FMMLA. Resort.
2326 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use
2327 TARGET_SVE_F64MM instead of separate tests.
2328 (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod<vsi2qi>): New pattern.
2329 (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod_lane<vsi2qi>): Likewise.
2330 (@aarch64_sve_add_<MATMUL:optab><vsi2qi>): Likewise.
2331 (@aarch64_sve_<FMMLA:sve_fp_op><mode>): Likewise.
2332 (@aarch64_sve_<PERMUTEQ:optab><mode>): Likewise.
2333 * config/aarch64/aarch64-sve-builtins.cc (TYPES_s_float): New macro.
2334 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): Use it.
2335 (TYPES_s_signed): New macro.
2336 (TYPES_s_integer): Use it.
2337 (TYPES_d_float): New macro.
2338 (TYPES_d_data): Use it.
2339 * config/aarch64/aarch64-sve-builtins-shapes.h (mmla): Declare.
2340 (ternary_intq_uintq_lane, ternary_intq_uintq_opt_n, ternary_uintq_intq)
2341 (ternary_uintq_intq_lane, ternary_uintq_intq_opt_n): Likewise.
2342 * config/aarch64/aarch64-sve-builtins-shapes.cc (mmla_def): New class.
2343 (svmmla): New shape.
2344 (ternary_resize2_opt_n_base): Add TYPE_CLASS2 and TYPE_CLASS3
2345 template parameters.
2346 (ternary_resize2_lane_base): Likewise.
2347 (ternary_resize2_base): New class.
2348 (ternary_qq_lane_base): Likewise.
2349 (ternary_intq_uintq_lane_def): Likewise.
2350 (ternary_intq_uintq_lane): New shape.
2351 (ternary_intq_uintq_opt_n_def): New class
2352 (ternary_intq_uintq_opt_n): New shape.
2353 (ternary_qq_lane_def): Inherit from ternary_qq_lane_base.
2354 (ternary_uintq_intq_def): New class.
2355 (ternary_uintq_intq): New shape.
2356 (ternary_uintq_intq_lane_def): New class.
2357 (ternary_uintq_intq_lane): New shape.
2358 (ternary_uintq_intq_opt_n_def): New class.
2359 (ternary_uintq_intq_opt_n): New shape.
2360 * config/aarch64/aarch64-sve-builtins-base.h (svmmla, svsudot)
2361 (svsudot_lane, svtrn1q, svtrn2q, svusdot, svusdot_lane, svusmmla)
2362 (svuzp1q, svuzp2q, svzip1q, svzip2q): Declare.
2363 * config/aarch64/aarch64-sve-builtins-base.cc (svdot_lane_impl):
2365 (svdotprod_lane_impl): ...this new class.
2366 (svmmla_impl, svusdot_impl): New classes.
2367 (svdot_lane): Update to use svdotprod_lane_impl.
2368 (svmmla, svsudot, svsudot_lane, svtrn1q, svtrn2q, svusdot)
2369 (svusdot_lane, svusmmla, svuzp1q, svuzp2q, svzip1q, svzip2q): New
2371 * config/aarch64/aarch64-sve-builtins-base.def (svmmla): New base
2372 function, with no types defined.
2373 (svmmla, svusmmla, svsudot, svsudot_lane, svusdot, svusdot_lane): New
2374 AARCH64_FL_I8MM functions.
2375 (svmmla): New AARCH64_FL_F32MM function.
2376 (svld1ro): Depend only on AARCH64_FL_F64MM, not on AARCH64_FL_V8_6.
2377 (svmmla, svtrn1q, svtrn2q, svuz1q, svuz2q, svzip1q, svzip2q): New
2378 AARCH64_FL_F64MM function.
2379 (REQUIRED_EXTENSIONS):
2381 2020-01-31 Andrew Stubbs <ams@codesourcery.com>
2383 * config/gcn/gcn-valu.md (addv64di3_exec): Allow one '0' in each
2386 2020-01-31 Uroš Bizjak <ubizjak@gmail.com>
2388 * config/i386/i386.md (*movoi_internal_avx): Do not check for
2389 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL. Remove MODE_V8SF handling.
2390 (*movti_internal): Do not check for
2391 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
2392 (*movtf_internal): Move check for TARGET_SSE2 and size optimization
2393 just after check for TARGET_AVX.
2394 (*movdf_internal): Ditto.
2395 * config/i386/mmx.md (*mov<mode>_internal): Do not check for
2396 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
2397 * config/i386/sse.md (mov<mode>_internal): Only check
2398 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL with V2DFmode. Move check
2399 for TARGET_SSE2 and size optimization just after check for TARGET_AVX.
2400 (<sse>_andnot<mode>3<mask_name>): Move check for
2401 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL after check for TARGET_AVX.
2402 (<code><mode>3<mask_name>): Ditto.
2403 (*andnot<mode>3): Ditto.
2404 (*andnottf3): Ditto.
2405 (*<code><mode>3): Ditto.
2406 (*<code>tf3): Ditto.
2407 (*andnot<VI:mode>3): Remove
2408 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL handling.
2409 (<mask_codefor><code><VI48_AVX_AVX512F:mode>3<mask_name>): Ditto.
2410 (*<code><VI12_AVX_AVX512F:mode>3): Ditto.
2411 (sse4_1_blendv<ssemodesuffix>): Ditto.
2412 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL):
2413 Explain that tune applies to 128bit instructions only.
2415 2020-01-31 Kwok Cheung Yeung <kcy@codesourcery.com>
2417 * config/gcn/mkoffload.c (process_asm): Add sgpr_count and vgpr_count
2418 to definition of hsa_kernel_description. Parse assembly to find SGPR
2419 and VGPR count of kernel and store in hsa_kernel_description.
2421 2020-01-31 Tamar Christina <tamar.christina@arm.com>
2423 PR rtl-optimization/91838
2424 * simplify-rtx.c (simplify_binary_operation_1): Update LSHIFTRT case
2425 to truncate if allowed or reject combination.
2427 2020-01-31 Andrew Stubbs <ams@codesourcery.com>
2429 * tree-ssa-loop-ivopts.c (get_iv): Use sizetype for zero-step.
2430 (find_inv_vars_cb): Likewise.
2432 2020-01-31 David Malcolm <dmalcolm@redhat.com>
2434 * calls.c (special_function_p): Split out the check for DECL_NAME
2435 being non-NULL and fndecl being extern at file scope into a
2436 new maybe_special_function_p and call it. Drop check for fndecl
2437 being non-NULL that was after a usage of DECL_NAME (fndecl).
2438 * tree.h (maybe_special_function_p): New inline function.
2440 2020-01-30 Andrew Stubbs <ams@codesourcery.com>
2442 * config/gcn/gcn-valu.md (gather<mode>_exec): Move contents ...
2443 (mask_gather_load<mode>): ... here, and zero-initialize the
2445 (maskload<mode>di): Zero-initialize the destination.
2448 2020-01-30 David Malcolm <dmalcolm@redhat.com>
2451 * doc/analyzer.texi (Limitations): Note that constraints on
2452 floating-point values are currently ignored.
2454 2020-01-30 Jakub Jelinek <jakub@redhat.com>
2457 * symtab.c (symtab_node::noninterposable_alias): If localalias
2458 already exists, but is not usable, append numbers after it until
2459 a unique name is found. Formatting fix.
2462 * combine.c (simplify_comparison) <case ROTATE>: Punt on out of range
2465 2020-01-30 Andrew Stubbs <ams@codesourcery.com>
2467 * config/gcn/gcn.c (print_operand): Handle LTGT.
2468 * config/gcn/predicates.md (gcn_fp_compare_operator): Allow ltgt.
2470 2020-01-30 Richard Biener <rguenther@suse.de>
2472 * tree-pretty-print.c (dump_generic_node): Wrap VECTOR_CST
2473 and CONSTRUCTOR in _Literal (type) with TDF_GIMPLE.
2475 2020-01-30 John David Anglin <danglin@gcc.gnu.org>
2477 * config/pa/pa.c (pa_elf_select_rtx_section): Place function pointers
2478 without a DECL in .data.rel.ro.local.
2480 2020-01-30 Jakub Jelinek <jakub@redhat.com>
2483 * config/arm/arm.md (uaddvdi4): Actually emit what gen_uaddvsi4
2487 * config/i386/sse.md
2488 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext): Renamed to ...
2489 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext): ... this. Use
2490 any_extend code iterator instead of always zero_extend.
2491 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_lt): Renamed to ...
2492 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_lt): ... this.
2493 Use any_extend code iterator instead of always zero_extend.
2494 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_shift): Renamed to ...
2495 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_shift): ... this.
2496 Use any_extend code iterator instead of always zero_extend.
2497 (*sse2_pmovmskb_ext): New define_insn.
2498 (*sse2_pmovmskb_ext_lt): New define_insn_and_split.
2501 * config/i386/i386.md (*popcountsi2_zext): New define_insn_and_split.
2502 (*popcountsi2_zext_falsedep): New define_insn.
2504 2020-01-30 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
2506 * config.in: Regenerated.
2507 * configure: Regenerated.
2509 2020-01-29 Tobias Burnus <tobias@codesourcery.com>
2512 * config/gcn/gcn-hsa.h (ASM_SPEC): Add -mattr=-code-object-v3 as
2513 LLVM's assembler changed the default in version 9.
2515 2020-01-24 Jeff Law <law@redhat.com>
2517 PR tree-optimization/89689
2518 * builtins.def (BUILT_IN_OBJECT_SIZE): Make it const rather than pure.
2520 2020-01-29 Richard Sandiford <richard.sandiford@arm.com>
2524 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
2526 PR rtl-optimization/87763
2527 * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
2528 simplification to handle subregs as well as bare regs.
2529 * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
2531 2020-01-29 Joel Hutton <Joel.Hutton@arm.com>
2534 * ira.c (ira): Revert use of simplified LRA algorithm.
2536 2020-01-29 Martin Jambor <mjambor@suse.cz>
2538 PR tree-optimization/92706
2539 * tree-sra.c (struct access): Fields first_link, last_link,
2540 next_queued and grp_queued renamed to first_rhs_link, last_rhs_link,
2541 next_rhs_queued and grp_rhs_queued respectively, new fields
2542 first_lhs_link, last_lhs_link, next_lhs_queued and grp_lhs_queued.
2543 (struct assign_link): Field next renamed to next_rhs, new field
2544 next_lhs. Updated comment.
2545 (work_queue_head): Renamed to rhs_work_queue_head.
2546 (lhs_work_queue_head): New variable.
2547 (add_link_to_lhs): New function.
2548 (relink_to_new_repr): Also relink LHS lists.
2549 (add_access_to_work_queue): Renamed to add_access_to_rhs_work_queue.
2550 (add_access_to_lhs_work_queue): New function.
2551 (pop_access_from_work_queue): Renamed to
2552 pop_access_from_rhs_work_queue.
2553 (pop_access_from_lhs_work_queue): New function.
2554 (build_accesses_from_assign): Also add links to LHS lists and to LHS
2556 (child_would_conflict_in_lacc): Renamed to
2557 child_would_conflict_in_acc. Adjusted parameter names.
2558 (create_artificial_child_access): New parameter set_grp_read, use it.
2559 (subtree_mark_written_and_enqueue): Renamed to
2560 subtree_mark_written_and_rhs_enqueue.
2561 (propagate_subaccesses_across_link): Renamed to
2562 propagate_subaccesses_from_rhs.
2563 (propagate_subaccesses_from_lhs): New function.
2564 (propagate_all_subaccesses): Also propagate subaccesses from LHSs to
2567 2020-01-29 Martin Jambor <mjambor@suse.cz>
2569 PR tree-optimization/92706
2570 * tree-sra.c (struct access): Adjust comment of
2571 grp_total_scalarization.
2572 (find_access_in_subtree): Look for single children spanning an entire
2574 (scalarizable_type_p): Allow register accesses, adjust callers.
2575 (completely_scalarize): Remove function.
2576 (scalarize_elem): Likewise.
2577 (create_total_scalarization_access): Likewise.
2578 (sort_and_splice_var_accesses): Do not track total scalarization
2580 (analyze_access_subtree): New parameter totally, adjust to new meaning
2581 of grp_total_scalarization.
2582 (analyze_access_trees): Pass new parameter to analyze_access_subtree.
2583 (can_totally_scalarize_forest_p): New function.
2584 (create_total_scalarization_access): Likewise.
2585 (create_total_access_and_reshape): Likewise.
2586 (total_should_skip_creating_access): Likewise.
2587 (totally_scalarize_subtree): Likewise.
2588 (analyze_all_variable_accesses): Perform total scalarization after
2589 subaccess propagation using the new functions above.
2590 (initialize_constant_pool_replacements): Output initializers by
2591 traversing the access tree.
2593 2020-01-29 Martin Jambor <mjambor@suse.cz>
2595 * tree-sra.c (verify_sra_access_forest): New function.
2596 (verify_all_sra_access_forests): Likewise.
2597 (create_artificial_child_access): Set parent.
2598 (analyze_all_variable_accesses): Call the verifier.
2600 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
2602 * cgraph.c (cgraph_edge::resolve_speculation): Only lookup direct edge
2603 if called on indirect edge.
2604 (cgraph_edge::redirect_call_stmt_to_callee): Lookup indirect edge of
2605 speculative call if needed.
2607 2020-01-29 Richard Biener <rguenther@suse.de>
2609 PR tree-optimization/93428
2610 * tree-vect-slp.c (vect_build_slp_tree_2): Compute the load
2611 permutation when the load node is created.
2612 (vect_analyze_slp_instance): Re-use it here.
2614 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
2616 * ipa-prop.c (update_indirect_edges_after_inlining): Fix warning.
2618 2020-01-28 Vladimir Makarov <vmakarov@redhat.com>
2620 PR rtl-optimization/93272
2621 * ira-lives.c (process_out_of_region_eh_regs): New function.
2622 (process_bb_node_lives): Call it.
2624 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
2626 * coverage.c (read_counts_file): Make error message lowercase.
2628 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
2630 * profile-count.c (profile_quality_display_names): Fix ordering.
2632 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
2635 * cgraph.c (cgraph_add_edge_to_call_site_hash): Update call site
2636 hash only when edge is first within the sequence.
2637 (cgraph_edge::set_call_stmt): Update handling of speculative calls.
2638 (symbol_table::create_edge): Do not set target_prob.
2639 (cgraph_edge::remove_caller): Watch for speculative calls when updating
2641 (cgraph_edge::make_speculative): Drop target_prob parameter.
2642 (cgraph_edge::speculative_call_info): Remove.
2643 (cgraph_edge::first_speculative_call_target): New member function.
2644 (update_call_stmt_hash_for_removing_direct_edge): New function.
2645 (cgraph_edge::resolve_speculation): Rewrite to new API.
2646 (cgraph_edge::speculative_call_for_target): New member function.
2647 (cgraph_edge::make_direct): Rewrite to new API; fix handling of
2648 multiple speculation targets.
2649 (cgraph_edge::redirect_call_stmt_to_callee): Likewise; fix updating
2651 (verify_speculative_call): Verify that targets form an interval.
2652 * cgraph.h (cgraph_edge::speculative_call_info): Remove.
2653 (cgraph_edge::first_speculative_call_target): New member function.
2654 (cgraph_edge::next_speculative_call_target): New member function.
2655 (cgraph_edge::speculative_call_target_ref): New member function.
2656 (cgraph_edge;:speculative_call_indirect_edge): New member funtion.
2657 (cgraph_edge): Remove target_prob.
2658 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
2659 Fix handling of speculative calls.
2660 * ipa-devirt.c (ipa_devirt): Fix handling of speculative cals.
2661 * ipa-fnsummary.c (analyze_function_body): Likewise.
2662 * ipa-inline.c (speculation_useful_p): Use new speculative call API.
2663 * ipa-profile.c (dump_histogram): Fix formating.
2664 (ipa_profile_generate_summary): Watch for overflows.
2665 (ipa_profile): Do not require probablity to be 1/2; update to new API.
2666 * ipa-prop.c (ipa_make_edge_direct_to_target): Update to new API.
2667 (update_indirect_edges_after_inlining): Update to new API.
2668 * ipa-utils.c (ipa_merge_profiles): Rewrite merging of speculative call
2670 * profile-count.h: (profile_probability::adjusted): New.
2671 * tree-inline.c (copy_bb): Update to new speculative call API; fix
2672 updating of profile.
2673 * value-prof.c (gimple_ic_transform): Rename to ...
2674 (dump_ic_profile): ... this one; update dumping.
2675 (stream_in_histogram_value): Fix formating.
2676 (gimple_value_profile_transformations): Update.
2678 2020-01-28 H.J. Lu <hongjiu.lu@intel.com>
2681 * config/i386/i386.md (*movoi_internal_avx): Remove
2682 TARGET_SSE_TYPELESS_STORES check.
2683 (*movti_internal): Prefer TARGET_AVX over
2684 TARGET_SSE_TYPELESS_STORES.
2685 (*movtf_internal): Likewise.
2686 * config/i386/sse.md (mov<mode>_internal): Prefer TARGET_AVX over
2687 TARGET_SSE_TYPELESS_STORES. Remove "<MODE_SIZE> == 16" check
2688 from TARGET_SSE_TYPELESS_STORES.
2690 2020-01-28 David Malcolm <dmalcolm@redhat.com>
2692 * diagnostic-core.h (warning_at): Rename overload to...
2693 (warning_meta): ...this.
2694 (emit_diagnostic_valist): Delete decl of overload taking
2695 diagnostic_metadata.
2696 * diagnostic.c (emit_diagnostic_valist): Likewise for defn.
2697 (warning_at): Rename overload taking diagnostic_metadata to...
2698 (warning_meta): ...this.
2700 2020-01-28 Richard Biener <rguenther@suse.de>
2702 PR tree-optimization/93439
2703 * tree-parloops.c (create_loop_fn): Move clique bookkeeping...
2704 * tree-cfg.c (move_sese_region_to_fn): ... here.
2705 (verify_types_in_gimple_reference): Verify used cliques are
2708 2020-01-28 H.J. Lu <hongjiu.lu@intel.com>
2711 * config/i386/i386-options.c (set_ix86_tune_features): Add an
2712 argument of a pointer to struct gcc_options and pass it to
2713 parse_mtune_ctrl_str.
2714 (ix86_function_specific_restore): Pass opts to
2715 set_ix86_tune_features.
2716 (ix86_option_override_internal): Likewise.
2717 (parse_mtune_ctrl_str): Add an argument of a pointer to struct
2718 gcc_options and use it for x_ix86_tune_ctrl_string.
2720 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
2722 PR rtl-optimization/87763
2723 * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
2724 simplification to handle subregs as well as bare regs.
2725 * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
2727 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
2729 * tree-vect-loop.c (vectorizable_reduction): Fail gracefully
2730 for reduction chains that (now) include a call.
2732 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
2734 PR tree-optimization/92822
2735 * tree-ssa-forwprop.c (simplify_vector_constructor): When filling
2736 out the don't-care elements of a vector whose significant elements
2737 are duplicates, make the don't-care elements duplicates too.
2739 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
2741 PR tree-optimization/93434
2742 * tree-predcom.c (split_data_refs_to_components): Record which
2743 components have had aliasing loads removed. Prevent store-store
2744 commoning for all such components.
2746 2020-01-28 Jakub Jelinek <jakub@redhat.com>
2749 * config/i386/i386.c (ix86_fold_builtin) <do_shift>: If mask is not
2750 -1 or is_vshift is true, use new_vector with number of elts npatterns
2751 rather than new_unary_operation.
2753 PR tree-optimization/93454
2754 * gimple-fold.c (fold_array_ctor_reference): Perform
2755 elt_size.to_uhwi () just once, instead of calling it in every
2756 iteration. Punt if that value is above size of the temporary
2757 buffer. Decrease third native_encode_expr argument when
2758 bufoff + elt_sz is above size of buf.
2760 2020-01-27 Joseph Myers <joseph@codesourcery.com>
2762 * config/mips/mips.c (mips_declare_object_name)
2763 [USE_GNU_UNIQUE_OBJECT]: Support use of gnu_unique_object.
2765 2020-01-27 Martin Liska <mliska@suse.cz>
2767 PR gcov-profile/93403
2768 * tree-profile.c (gimple_init_gcov_profiler): Generate
2769 both __gcov_indirect_call_profiler_v4 and
2770 __gcov_indirect_call_profiler_v4_atomic.
2772 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
2775 * config/aarch64/aarch64-simd.md (aarch64_get_half<mode>): New
2777 (@aarch64_split_simd_mov<mode>): Use it.
2778 (aarch64_simd_mov_from_<mode>low): Add a GPR alternative.
2779 Leave the vec_extract patterns to handle 2-element vectors.
2780 (aarch64_simd_mov_from_<mode>high): Likewise.
2781 (vec_extract<VQMOV_NO2E:mode><Vhalf>): New expander.
2782 (vec_extractv2dfv1df): Likewise.
2784 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
2786 * config/aarch64/aarch64.c (aarch64_if_then_else_costs): Match
2787 jump conditions for *compare_condjump<GPI:mode>.
2789 2020-01-27 David Malcolm <dmalcolm@redhat.com>
2792 * digraph.cc (test_edge::test_edge): Specify template for base
2795 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
2797 * config/arc/arc.c (arc_rtx_costs): Update mul64 cost.
2799 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
2801 * config/arc/arc-protos.h (gen_mlo): Remove.
2802 (gen_mhi): Likewise.
2803 * config/arc/arc.c (AUX_MULHI): Define.
2804 (arc_must_save_reister): Special handling for r58/59.
2805 (arc_compute_frame_size): Consider mlo/mhi registers.
2806 (arc_save_callee_saves): Emit fp/sp move only when emit_move
2808 (arc_conditional_register_usage): Remove TARGET_BIG_ENDIAN from
2809 mlo/mhi name selection.
2810 (arc_restore_callee_saves): Don't early restore blink when ISR.
2811 (arc_expand_prologue): Add mlo/mhi saving.
2812 (arc_expand_epilogue): Add mlo/mhi restoring.
2815 * config/arc/arc.h (DBX_REGISTER_NUMBER): Correct register
2816 numbering when MUL64 option is used.
2817 (DWARF2_FRAME_REG_OUT): Define.
2818 * config/arc/arc.md (arc600_stall): New pattern.
2819 (VUNSPEC_ARC_ARC600_STALL): Define.
2820 (mulsi64): Use correct mlo/mhi registers.
2821 (mulsi_600): Clean it up.
2822 * config/arc/predicates.md (mlo_operand): Remove any dependency on
2824 (mhi_operand): Likewise.
2826 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
2827 Petro Karashchenko <petro.karashchenko@ring.com>
2829 * config/arc/arc.c (arc_is_uncached_mem_p): Check struct
2830 attributes if needed.
2831 (prepare_move_operands): Generate special unspec instruction for
2833 (arc_isuncached_mem_p): Propagate uncached attribute to each
2835 * config/arc/arc.md (VUNSPEC_ARC_LDDI): Define.
2836 (VUNSPEC_ARC_STDI): Likewise.
2837 (ALLI): New mode iterator.
2838 (mALLI): New mode attribute.
2839 (lddi): New instruction pattern.
2841 (stdidi_split): Split instruction for architectures which are not
2842 supporting ll64 option.
2843 (lddidi_split): Likewise.
2845 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
2847 PR rtl-optimization/92989
2848 * lra-lives.c (process_bb_lives): Update the live-in set before
2849 processing additional clobbers.
2851 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
2853 PR rtl-optimization/93170
2854 * cselib.c (cselib_invalidate_regno_val): New function, split out
2856 (cselib_invalidate_regno): ...here.
2857 (cselib_invalidated_by_call_p): New function.
2858 (cselib_process_insn): Iterate over all the hard-register entries in
2859 REG_VALUES and invalidate any that cross call-clobbered registers.
2861 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
2863 * dojump.c (split_comparison): Use HONOR_NANS rather than
2864 HONOR_SNANS when splitting LTGT.
2866 2020-01-27 Martin Liska <mliska@suse.cz>
2869 * opts.c (print_filtered_help): Exclude language-specific
2870 options from --help=common unless enabled in all FEs.
2872 2020-01-27 Martin Liska <mliska@suse.cz>
2874 * opts.c (print_help): Exclude params from
2875 all except --help=param.
2877 2020-01-27 Martin Liska <mliska@suse.cz>
2880 * config/i386/i386-features.c (make_resolver_func):
2881 Align the code with ppc64 target implementation.
2882 Do not generate a unique name for resolver function.
2884 2020-01-27 Richard Biener <rguenther@suse.de>
2886 PR tree-optimization/93397
2887 * tree-vect-slp.c (vect_analyze_slp_instance): Delay
2888 converted reduction chain SLP graph adjustment.
2890 2020-01-26 Marek Polacek <polacek@redhat.com>
2893 * sanopt.c (sanitize_rewrite_addressable_params): Avoid crash on
2896 2020-01-26 Jason Merrill <jason@redhat.com>
2899 * tree.c (verify_type_variant): Only verify TYPE_NEEDS_CONSTRUCTING
2902 2020-01-26 Darius Galis <darius.galis@cyberthorstudios.com>
2904 * config/rx/rx.md (setmemsi): Added rx_allow_string_insns constraint
2905 (rx_setmem): Likewise.
2907 2020-01-26 Jakub Jelinek <jakub@redhat.com>
2910 * config/i386/i386.md (*addv<dwi>4_doubleword, *subv<dwi>4_doubleword):
2911 Use nonimmediate_operand instead of x86_64_hilo_general_operand and
2912 drop <di> from constraint of last operand.
2915 * config/i386/sse.md (*avx_vperm_broadcast_<mode>): Disallow for
2916 TARGET_AVX2 and V4DFmode not in the split condition, but in the
2917 pattern condition, though allow { 0, 0, 0, 0 } broadcast always.
2919 2020-01-25 Feng Xue <fxue@os.amperecomputing.com>
2922 * ipa-cp.c (get_info_about_necessary_edges): Remove value
2925 2020-01-24 Jeff Law <law@redhat.com>
2927 PR tree-optimization/92788
2928 * tree-ssa-threadedge.c (thread_across_edge): Check EDGE_COMPLEX
2931 2020-01-24 Jakub Jelinek <jakub@redhat.com>
2934 * config/i386/sse.md (*avx_vperm_broadcast_v4sf,
2935 *avx_vperm_broadcast_<mode>,
2936 <sse2_avx_avx512f>_vpermil<mode><mask_name>,
2937 *<sse2_avx_avx512f>_vpermilp<mode><mask_name>):
2938 Move before avx2_perm<mode>/avx512f_perm<mode>.
2941 * simplify-rtx.c (simplify_const_unary_operation,
2942 simplify_const_binary_operation): Punt for mode precision above
2943 MAX_BITSIZE_MODE_ANY_INT.
2945 2020-01-24 Andrew Pinski <apinski@marvell.com>
2947 * config/arm/aarch-cost-tables.h (cortexa57_extra_costs): Change
2950 2020-01-24 Jeff Law <law@redhat.com>
2953 * config/h8300/h8300.c (h8300_print_operand): Only call byte_reg
2954 for REGs. Call output_operand_lossage to get more reasonable
2957 2020-01-24 Andrew Stubbs <ams@codesourcery.com>
2959 * config/gcn/gcn-valu.md (vec_cmp<mode>di): Use
2960 gcn_fp_compare_operator.
2961 (vec_cmpu<mode>di): Use gcn_compare_operator.
2962 (vec_cmp<u>v64qidi): Use gcn_compare_operator.
2963 (vec_cmp<mode>di_exec): Use gcn_fp_compare_operator.
2964 (vec_cmpu<mode>di_exec): Use gcn_compare_operator.
2965 (vec_cmp<u>v64qidi_exec): Use gcn_compare_operator.
2966 (vec_cmp<mode>di_dup): Use gcn_fp_compare_operator.
2967 (vec_cmp<mode>di_dup_exec): Use gcn_fp_compare_operator.
2968 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): Use
2969 gcn_fp_compare_operator.
2970 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): Use
2971 gcn_fp_compare_operator.
2972 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): Use
2973 gcn_fp_compare_operator.
2974 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): Use
2975 gcn_fp_compare_operator.
2977 2020-01-24 Maciej W. Rozycki <macro@wdc.com>
2979 * doc/install.texi (Cross-Compiler-Specific Options): Document
2980 `--with-toolexeclibdir' option.
2982 2020-01-24 Hans-Peter Nilsson <hp@axis.com>
2984 * target.def (flags_regnum): Also mention effect on delay slot filling.
2985 * doc/tm.texi: Regenerate.
2987 2020-01-23 Jeff Law <law@redhat.com>
2989 PR translation/90162
2990 * config/h8300/h8300.c (h8300_option_override): Fix diagnostic text.
2992 2020-01-23 Mikael Tillenius <mti-1@tillenius.com>
2995 * config/h8300/h8300.h (FUNCTION_PROFILER): Fix emission of
2998 2020-01-23 Jakub Jelinek <jakub@redhat.com>
3000 PR rtl-optimization/93402
3001 * postreload.c (reload_combine_recognize_pattern): Don't try to adjust
3004 2020-01-23 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
3006 * config.in: Regenerated.
3007 * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to 1
3008 for TARGET_LIBC_GNUSTACK.
3009 * configure: Regenerated.
3010 * configure.ac: Define TARGET_LIBC_GNUSTACK if glibc version is
3011 found to be 2.31 or greater.
3013 2020-01-23 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
3015 * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to
3017 * config/mips/mips.c (TARGET_ASM_FILE_END): Define to ...
3018 (mips_asm_file_end): New function. Delegate to
3019 file_end_indicate_exec_stack if NEED_INDICATE_EXEC_STACK is true.
3020 * config/mips/mips.h (NEED_INDICATE_EXEC_STACK): Define to 0.
3022 2020-01-23 Jakub Jelinek <jakub@redhat.com>
3025 * config/i386/i386-modes.def (POImode): New mode.
3026 (MAX_BITSIZE_MODE_ANY_INT): Change from 128 to 160.
3027 * config/i386/i386.md (DPWI): New mode attribute.
3028 (addv<mode>4, subv<mode>4): Use <DPWI> instead of <DWI>.
3030 (QPWI): ... this. Use POI instead of OI for TImode.
3031 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1,
3032 *subv<dwi>4_doubleword, *subv<dwi>4_doubleword_1): Use <QPWI>
3035 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
3038 * config/aarch64/aarch64.md (UNSPEC_SPECULATION_TRACKER_REV): New
3040 (speculation_tracker_rev): New pattern.
3041 * config/aarch64/aarch64-speculation.cc (aarch64_do_track_speculation):
3042 Use speculation_tracker_rev to track the inverse condition.
3044 2020-01-23 Richard Biener <rguenther@suse.de>
3046 PR tree-optimization/93381
3047 * tree-ssa-sccvn.c (vn_walk_cb_data::push_partial_def): Take
3048 alias-set of the def as argument and record the first one.
3049 (vn_walk_cb_data::first_set): New member.
3050 (vn_reference_lookup_3): Pass the alias-set of the current def
3051 to push_partial_def. Fix alias-set used in the aggregate copy
3053 (vn_reference_lookup): Consistently set *last_vuse_ptr.
3054 * real.c (clear_significand_below): Fix out-of-bound access.
3056 2020-01-23 Jakub Jelinek <jakub@redhat.com>
3059 * config/i386/i386.md (*bmi2_bzhi_<mode>3_2, *bmi2_bzhi_<mode>3_3):
3060 New define_insn patterns.
3062 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
3064 * doc/sourcebuild.texi (check-function-bodies): Add an
3065 optional target/xfail selector.
3067 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
3069 PR rtl-optimization/93124
3070 * auto-inc-dec.c (merge_in_block): Don't add auto inc/decs to
3071 bare USE and CLOBBER insns.
3073 2020-01-22 Andrew Pinski <apinski@marvell.com>
3075 * config/arc/arc.c (output_short_suffix): Check insn for nullness.
3077 2020-01-22 David Malcolm <dmalcolm@redhat.com>
3080 * gdbinit.in (break-on-saved-diagnostic): Update for move of
3081 diagnostic_manager into "ana" namespace.
3082 * selftest-run-tests.c (selftest::run_tests): Update for move of
3083 selftest::run_analyzer_selftests to
3084 ana::selftest::run_analyzer_selftests.
3086 2020-01-22 Richard Sandiford <richard.sandiford@arm.com>
3088 * cfgexpand.c (union_stack_vars): Update the size.
3090 2020-01-22 Richard Biener <rguenther@suse.de>
3092 PR tree-optimization/93381
3093 * tree-ssa-structalias.c (find_func_aliases): Assume offsetting
3094 throughout, handle all conversions the same.
3096 2020-01-22 Jakub Jelinek <jakub@redhat.com>
3099 * config/aarch64/aarch64.c (aarch64_expand_subvti): Only use
3100 gen_subdi3_compare1_imm if low_in2 satisfies aarch64_plus_immediate
3101 predicate, not whenever it is CONST_INT. Otherwise, force_reg it.
3102 Call force_reg on high_in2 unconditionally.
3104 2020-01-22 Martin Liska <mliska@suse.cz>
3106 PR tree-optimization/92924
3107 * profile.c (compute_value_histograms): Divide
3110 2020-01-22 Jakub Jelinek <jakub@redhat.com>
3113 * output.h (assemble_name_resolve): Declare.
3114 * varasm.c (assemble_name_resolve): New function.
3115 (assemble_name): Use it.
3116 * config/i386/i386.h (ASM_OUTPUT_SYMBOL_REF): Define.
3118 2020-01-22 Joseph Myers <joseph@codesourcery.com>
3120 * doc/sourcebuild.texi (Texinfo Manuals, Front End): Refer to
3121 update_web_docs_git instead of update_web_docs_svn.
3123 2020-01-21 Andrew Pinski <apinski@marvell.com>
3126 * config/aarch64/aarch64.md (tlsgd_small_<mode>): Have operand 0
3127 as PTR mode. Have operand 1 as being modeless, it can be P mode.
3128 (*tlsgd_small_<mode>): Likewise.
3129 * config/aarch64/aarch64.c (aarch64_load_symref_appropriately)
3130 <case SYMBOL_SMALL_TLSGD>: Call gen_tlsgd_small_* with a ptr_mode
3131 register. Convert that register back to dest using convert_mode.
3133 2020-01-21 Jim Wilson <jimw@sifive.com>
3135 * config/riscv/riscv-sr.c (riscv_sr_match_prologue): Use INTVAL
3138 2020-01-21 H.J. Lu <hongjiu.lu@intel.com>
3139 Uros Bizjak <ubizjak@gmail.com>
3142 * config/i386/i386.c (ix86_tls_module_base): Replace Pmode
3144 (legitimize_tls_address): Do GNU2 TLS address computation in
3145 ptr_mode and zero-extend result to Pmode.
3146 * config/i386/i386.md (@tls_dynamic_gnu2_64_<mode>): Replace
3147 :P with :PTR and Pmode with ptr_mode.
3148 (*tls_dynamic_gnu2_lea_64_<mode>): Likewise.
3149 (*tls_dynamic_gnu2_call_64_<mode>): Likewise.
3150 (*tls_dynamic_gnu2_combine_64_<mode>): Likewise.
3152 2020-01-21 Jakub Jelinek <jakub@redhat.com>
3155 * config/riscv/riscv.c (riscv_rtx_costs) <case ZERO_EXTRACT>: Verify
3156 the last two operands are CONST_INT_P before using them as such.
3158 2020-01-21 Richard Sandiford <richard.sandiford@arm.com>
3160 * config/aarch64/aarch64-sve-builtins.def: Use get_typenode_from_name
3161 to get the integer element types.
3163 2020-01-21 Richard Sandiford <richard.sandiford@arm.com>
3165 * config/aarch64/aarch64-sve-builtins.h
3166 (function_expander::convert_to_pmode): Declare.
3167 * config/aarch64/aarch64-sve-builtins.cc
3168 (function_expander::convert_to_pmode): New function.
3169 (function_expander::get_contiguous_base): Use it.
3170 (function_expander::prepare_gather_address_operands): Likewise.
3171 * config/aarch64/aarch64-sve-builtins-sve2.cc
3172 (svwhilerw_svwhilewr_impl::expand): Likewise.
3174 2020-01-21 Szabolcs Nagy <szabolcs.nagy@arm.com>
3177 * config/aarch64/aarch64.c (aarch64_declare_function_name): Set
3178 cfun->machine->label_is_assembled.
3179 (aarch64_print_patchable_function_entry): New.
3180 (TARGET_ASM_PRINT_PATCHABLE_FUNCTION_ENTRY): Define.
3181 * config/aarch64/aarch64.h (struct machine_function): New field,
3184 2020-01-21 David Malcolm <dmalcolm@redhat.com>
3187 * ipa-profile.c (ipa_profile): Delete call_sums and set it to
3190 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
3193 * cgraph.c (cgraph_edge::resolve_speculation,
3194 cgraph_edge::redirect_call_stmt_to_callee): Fix update of
3195 call_stmt_site_hash.
3197 2020-01-21 Martin Liska <mliska@suse.cz>
3199 * config/rs6000/rs6000.c (common_mode_defined): Remove
3202 2020-01-21 Richard Biener <rguenther@suse.de>
3204 PR tree-optimization/92328
3205 * tree-ssa-sccvn.c (vn_reference_lookup_3): Preserve
3206 type when value-numbering same-sized store by inserting a
3208 (eliminate_dom_walker::eliminate_stmt): When eliminating
3209 a redundant store handle bit-reinterpretation of the same value.
3211 2020-01-21 Andrew Pinski <apinski@marvel.com>
3214 * tree-into-ssa.c (prepare_block_for_update_1): Split out
3216 (prepare_block_for_update): This. Use a worklist instead of
3219 2020-01-21 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3221 * gcc/config/arm/arm.c (clear_operation_p):
3222 Initialise last_regno, skip first iteration
3223 based on the first_set value and use ints instead
3224 of the unnecessary HOST_WIDE_INTs.
3226 2020-01-21 Jakub Jelinek <jakub@redhat.com>
3229 * config/rs6000/rs6000.c (rs6000_emit_cmove): If using fsel, punt for
3230 compare_mode other than SFmode or DFmode.
3232 2020-01-21 Kito Cheng <kito.cheng@sifive.com>
3235 * config/riscv/riscv-protos.h (riscv_hard_regno_rename_ok): New.
3236 * config/riscv/riscv.c (riscv_hard_regno_rename_ok): New.
3237 * config/riscv/riscv.h (HARD_REGNO_RENAME_OK): Defined.
3239 2020-01-20 Wilco Dijkstra <wdijkstr@arm.com>
3241 * config/aarch64/aarch64.c (neoversen1_tunings): Set jump_align to 4.
3243 2020-01-20 Andrew Pinski <apinski@marvell.com>
3246 * targhooks.c (default_print_patchable_function_entry): Use
3247 output_asm_insn to emit the nop instruction.
3249 2020-01-20 Fangrui Song <maskray@google.com>
3252 * targhooks.c (default_print_patchable_function_entry): Align to
3255 2020-01-20 H.J. Lu <hongjiu.lu@intel.com>
3258 * config/i386/i386.c (legitimize_tls_address): Pass Pmode to
3259 gen_tls_dynamic_gnu2_64. Compute GNU2 TLS address in ptr_mode.
3260 * config/i386/i386.md (tls_dynamic_gnu2_64): Renamed to ...
3261 (@tls_dynamic_gnu2_64_<mode>): This. Replace DI with P.
3262 (*tls_dynamic_gnu2_lea_64): Renamed to ...
3263 (*tls_dynamic_gnu2_lea_64_<mode>): This. Replace DI with P.
3264 Remove the {q} suffix from lea.
3265 (*tls_dynamic_gnu2_call_64): Renamed to ...
3266 (*tls_dynamic_gnu2_call_64_<mode>): This. Replace DI with P.
3267 (*tls_dynamic_gnu2_combine_64): Renamed to ...
3268 (*tls_dynamic_gnu2_combine_64_<mode>): This. Replace DI with P.
3269 Pass Pmode to gen_tls_dynamic_gnu2_64.
3271 2020-01-20 Wilco Dijkstra <wdijkstr@arm.com>
3273 * config/aarch64/aarch64.h (SLOW_BYTE_ACCESS): Set to 1.
3275 2020-01-20 Richard Sandiford <richard.sandiford@arm.com>
3277 * config/aarch64/aarch64-sve-builtins-base.cc
3278 (svld1ro_impl::memory_vector_mode): Remove parameter name.
3280 2020-01-20 Richard Biener <rguenther@suse.de>
3283 * dwarf2out.c (prune_unused_types): Unconditionally mark
3284 called function DIEs.
3286 2020-01-20 Martin Liska <mliska@suse.cz>
3288 PR tree-optimization/93199
3289 * tree-eh.c (struct leh_state): Add
3290 new field outer_non_cleanup.
3291 (cleanup_is_dead_in): Pass leh_state instead
3292 of eh_region. Add a checking that state->outer_non_cleanup
3293 points to outer non-clean up region.
3294 (lower_try_finally): Record outer_non_cleanup
3296 (lower_catch): Likewise.
3297 (lower_eh_filter): Likewise.
3298 (lower_eh_must_not_throw): Likewise.
3299 (lower_cleanup): Likewise.
3301 2020-01-20 Richard Biener <rguenther@suse.de>
3303 PR tree-optimization/93094
3304 * tree-vectorizer.h (vect_loop_versioning): Adjust.
3305 (vect_transform_loop): Likewise.
3306 * tree-vectorizer.c (try_vectorize_loop_1): Pass down
3307 loop_vectorized_call to vect_transform_loop.
3308 * tree-vect-loop.c (vect_transform_loop): Pass down
3309 loop_vectorized_call to vect_loop_versioning.
3310 * tree-vect-loop-manip.c (vect_loop_versioning): Use
3311 the earlier discovered loop_vectorized_call.
3313 2020-01-19 Eric S. Raymond <esr@thyrsus.com>
3315 * doc/contribute.texi: Update for SVN -> Git transition.
3316 * doc/install.texi: Likewise.
3318 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
3321 * cgraph.c (cgraph_edge::make_speculative): Increase number of
3322 speculative targets.
3323 (verify_speculative_call): New function
3324 (cgraph_node::verify_node): Use it.
3325 * ipa-profile.c (ipa_profile): Fix formating; do not set number of
3328 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
3331 * cgraph.c (cgraph_edge::resolve_speculation): Fix foramting.
3332 (cgraph_edge::make_direct): Remove all indirect targets.
3333 (cgraph_edge::redirect_call_stmt_to_callee): Use make_direct..
3334 (cgraph_node::verify_node): Verify that only one call_stmt or
3335 lto_stmt_uid is set.
3336 * cgraphclones.c (cgraph_edge::clone): Set only one call_stmt or
3338 * lto-cgraph.c (lto_output_edge): Simplify streaming of stmt.
3339 (lto_output_ref): Simplify streaming of stmt.
3340 * lto-streamer-in.c (fixup_call_stmt_edges_1): Clear lto_stmt_uid.
3342 2020-01-18 Tamar Christina <tamar.christina@arm.com>
3344 * config/aarch64/aarch64-sve-builtins-base.cc (memory_vector_mode):
3345 Mark parameter unused.
3347 2020-01-18 Hans-Peter Nilsson <hp@axis.com>
3349 * config.gcc <obsolete targets>: Add crisv32-*-* and cris-*-linux*
3351 2019-01-18 Gerald Pfeifer <gerald@pfeifer.com>
3353 * varpool.c (ctor_useable_for_folding_p): Fix grammar.
3355 2020-01-18 Iain Sandoe <iain@sandoe.co.uk>
3357 * Makefile.in: Add coroutine-passes.o.
3358 * builtin-types.def (BT_CONST_SIZE): New.
3359 (BT_FN_BOOL_PTR): New.
3360 (BT_FN_PTR_PTR_CONST_SIZE_BOOL): New.
3361 * builtins.def (DEF_COROUTINE_BUILTIN): New.
3362 * coroutine-builtins.def: New file.
3363 * coroutine-passes.cc: New file.
3364 * function.h (struct GTY function): Add a bit to indicate that the
3365 function is a coroutine component.
3366 * internal-fn.c (expand_CO_FRAME): New.
3367 (expand_CO_YIELD): New.
3368 (expand_CO_SUSPN): New.
3369 (expand_CO_ACTOR): New.
3370 * internal-fn.def (CO_ACTOR): New.
3374 * passes.def: Add pass_coroutine_lower_builtins,
3375 pass_coroutine_early_expand_ifns.
3376 * tree-pass.h (make_pass_coroutine_lower_builtins): New.
3377 (make_pass_coroutine_early_expand_ifns): New.
3378 * doc/invoke.texi: Document the fcoroutines command line
3381 2020-01-18 Jakub Jelinek <jakub@redhat.com>
3383 * config/arm/vfp.md (*clear_vfp_multiple): Remove unused variable.
3386 * config/arm/arm.c (clear_operation_p): Don't use REGNO until
3387 after checking the argument is a REG. Don't use REGNO (reg)
3388 again to set last_regno, reuse regno variable instead.
3390 2020-01-17 David Malcolm <dmalcolm@redhat.com>
3392 * doc/analyzer.texi (Limitations): Add note about NaN.
3394 2020-01-17 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3395 Sudakshina Das <sudi.das@arm.com>
3397 * config/arm/arm.md (ashldi3): Generate thumb2_lsll for both reg
3398 and valid immediate.
3399 (ashrdi3): Generate thumb2_asrl for both reg and valid immediate.
3400 (lshrdi3): Generate thumb2_lsrl for valid immediates.
3401 * config/arm/constraints.md (Pg): New.
3402 * config/arm/predicates.md (long_shift_imm): New.
3403 (arm_reg_or_long_shift_imm): Likewise.
3404 * config/arm/thumb2.md (thumb2_asrl): New immediate alternative.
3405 (thumb2_lsll): Likewise.
3408 2020-01-17 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3409 Sudakshina Das <sudi.das@arm.com>
3411 * config/arm/arm.md (ashldi3): Generate thumb2_lsll for TARGET_HAVE_MVE.
3412 (ashrdi3): Generate thumb2_asrl for TARGET_HAVE_MVE.
3413 * config/arm/arm.c (arm_hard_regno_mode_ok): Allocate even odd
3414 register pairs for doubleword quantities for ARMv8.1M-Mainline.
3415 * config/arm/thumb2.md (thumb2_asrl): New.
3416 (thumb2_lsll): Likewise.
3418 2020-01-17 Jakub Jelinek <jakub@redhat.com>
3420 * config/arm/arm.c (cmse_nonsecure_call_inline_register_clear): Remove
3423 2020-01-17 Alexander Monakov <amonakov@ispras.ru>
3425 * gdbinit.in (help-gcc-hooks): New command.
3426 (pp, pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, ptc, pdn, ptn, pdd, prc,
3427 pi, pbm, pel, trt): Take $arg0 instead of $ if supplied. Update
3430 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
3432 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use the
3433 correct target macro.
3435 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
3437 * config/aarch64/aarch64-protos.h
3438 (aarch64_sve_ld1ro_operand_p): New.
3439 * config/aarch64/aarch64-sve-builtins-base.cc
3440 (class load_replicate): New.
3441 (class svld1ro_impl): New.
3442 (class svld1rq_impl): Change to inherit from load_replicate.
3443 (svld1ro): New sve intrinsic function base.
3444 * config/aarch64/aarch64-sve-builtins-base.def (svld1ro):
3445 New DEF_SVE_FUNCTION.
3446 * config/aarch64/aarch64-sve-builtins-base.h
3447 (svld1ro): New decl.
3448 * config/aarch64/aarch64-sve-builtins.cc
3449 (function_expander::add_mem_operand): Modify assert to allow
3451 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): New
3453 * config/aarch64/aarch64.c
3454 (aarch64_sve_ld1rq_operand_p): Implement in terms of ...
3455 (aarch64_sve_ld1rq_ld1ro_operand_p): This.
3456 (aarch64_sve_ld1ro_operand_p): New.
3457 * config/aarch64/aarch64.md (UNSPEC_LD1RO): New unspec.
3458 * config/aarch64/constraints.md (UOb,UOh,UOw,UOd): New.
3459 * config/aarch64/predicates.md
3460 (aarch64_sve_ld1ro_operand_{b,h,w,d}): New.
3462 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
3464 * config/aarch64/aarch64-c.c (_ARM_FEATURE_MATMUL_FLOAT64):
3465 Introduce this ACLE specified predefined macro.
3466 * config/aarch64/aarch64-option-extensions.def (f64mm): New.
3467 (fp): Disabling this disables f64mm.
3468 (simd): Disabling this disables f64mm.
3469 (fp16): Disabling this disables f64mm.
3470 (sve): Disabling this disables f64mm.
3471 * config/aarch64/aarch64.h (AARCH64_FL_F64MM): New.
3472 (AARCH64_ISA_F64MM): New.
3473 (TARGET_F64MM): New.
3474 * doc/invoke.texi (f64mm): Document new option.
3476 2020-01-17 Wilco Dijkstra <wdijkstr@arm.com>
3478 * config/aarch64/aarch64.c (generic_tunings): Add branch fusion.
3479 (neoversen1_tunings): Likewise.
3481 2020-01-17 Wilco Dijkstra <wdijkstr@arm.com>
3484 * config/aarch64/aarch64.c (aarch64_split_compare_and_swap)
3485 Add assert to ensure prolog has been emitted.
3486 (aarch64_split_atomic_op): Likewise.
3487 * config/aarch64/atomics.md (aarch64_compare_and_swap<mode>)
3488 Use epilogue_completed rather than reload_completed.
3489 (aarch64_atomic_exchange<mode>): Likewise.
3490 (aarch64_atomic_<atomic_optab><mode>): Likewise.
3491 (atomic_nand<mode>): Likewise.
3492 (aarch64_atomic_fetch_<atomic_optab><mode>): Likewise.
3493 (atomic_fetch_nand<mode>): Likewise.
3494 (aarch64_atomic_<atomic_optab>_fetch<mode>): Likewise.
3495 (atomic_nand_fetch<mode>): Likewise.
3497 2020-01-17 Richard Sandiford <richard.sandiford@arm.com>
3500 * config/aarch64/aarch64.h (REVERSIBLE_CC_MODE): Return false
3502 (REVERSE_CONDITION): Delete.
3503 * config/aarch64/iterators.md (CC_ONLY): New mode iterator.
3504 (CCFP_CCFPE): Likewise.
3505 (e): New mode attribute.
3506 * config/aarch64/aarch64.md (ccmp<GPI:mode>): Rename to...
3507 (@ccmp<CC_ONLY:mode><GPI:mode>): ...this, using CC_ONLY instead of CC.
3508 (fccmp<GPF:mode>, fccmpe<GPF:mode>): Merge into...
3509 (@ccmp<CCFP_CCFPE:mode><GPF:mode>): ...this combined pattern.
3510 (@ccmp<CC_ONLY:mode><GPI:mode>_rev): New pattern.
3511 (@ccmp<CCFP_CCFPE:mode><GPF:mode>_rev): Likewise.
3512 * config/aarch64/aarch64.c (aarch64_gen_compare_reg): Update
3513 name of generator from gen_ccmpdi to gen_ccmpccdi.
3514 (aarch64_gen_ccmp_next): Use code_for_ccmp. If we want to reverse
3515 the previous comparison but aren't able to, use the new ccmp_rev
3518 2020-01-17 Richard Sandiford <richard.sandiford@arm.com>
3520 * gimplify.c (gimplify_return_expr): Use poly_int_tree_p rather
3521 than testing directly for INTEGER_CST.
3522 (gimplify_target_expr, gimplify_omp_depend): Likewise.
3524 2020-01-17 Jakub Jelinek <jakub@redhat.com>
3526 PR tree-optimization/93292
3527 * tree-vect-stmts.c (vectorizable_comparison): Punt also if
3528 get_vectype_for_scalar_type returns NULL.
3530 2020-01-16 Jan Hubicka <hubicka@ucw.cz>
3532 * params.opt (-param=max-predicted-iterations): Increase range from 0.
3533 * predict.c (estimate_loops): Add 1 to param_max_predicted_iterations.
3535 2020-01-16 Jan Hubicka <hubicka@ucw.cz>
3537 * ipa-fnsummary.c (estimate_calls_size_and_time): Fix formating of
3539 * params.opt: (max-predicted-iterations): Set bounds.
3540 * predict.c (real_almost_one, real_br_prob_base,
3541 real_inv_br_prob_base, real_one_half, real_bb_freq_max): Remove.
3542 (propagate_freq): Add max_cyclic_prob parameter; cap cyclic
3543 probabilities; do not truncate to reg_br_prob_bases.
3544 (estimate_loops_at_level): Pass max_cyclic_prob.
3545 (estimate_loops): Compute max_cyclic_prob.
3546 (estimate_bb_frequencies): Do not initialize real_*; update calculation
3548 * profile-count.c (profile_probability::to_sreal): New.
3549 * profile-count.h (class sreal): Move up in file.
3550 (profile_probability::to_sreal): Declare.
3552 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
3555 (arm_invalid_conversion): New function for target hook.
3556 (arm_invalid_unary_op): New function for target hook.
3557 (arm_invalid_binary_op): New function for target hook.
3559 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
3561 * config.gcc: Add arm_bf16.h.
3562 * config/arm/arm-builtins.c (arm_mangle_builtin_type): Fix comment.
3563 (arm_simd_builtin_std_type): Add BFmode.
3564 (arm_init_simd_builtin_types): Define element types for vector types.
3565 (arm_init_bf16_types): New function.
3566 (arm_init_builtins): Add arm_init_bf16_types function call.
3567 * config/arm/arm-modes.def: Add BFmode and V4BF, V8BF vector modes.
3568 * config/arm/arm-simd-builtin-types.def: Add V4BF, V8BF.
3569 * config/arm/arm.c (aapcs_vfp_sub_candidate): Add BFmode.
3570 (arm_hard_regno_mode_ok): Add BFmode and tidy up statements.
3571 (arm_vector_mode_supported_p): Add V4BF, V8BF.
3572 (arm_mangle_type): Add __bf16.
3573 * config/arm/arm.h: Add V4BF, V8BF to VALID_NEON_DREG_MODE,
3574 VALID_NEON_QREG_MODE respectively. Add export arm_bf16_type_node,
3575 arm_bf16_ptr_type_node.
3576 * config/arm/arm.md: Add BFmode to movhf expand, mov pattern and
3577 define_split between ARM registers.
3578 * config/arm/arm_bf16.h: New file.
3579 * config/arm/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
3580 * config/arm/iterators.md: (ANY64_BF, VDXMOV, VHFBF, HFBF, fporbf): New.
3582 * config/arm/neon.md: Add BF vector types to movhf NEON move patterns.
3583 * config/arm/vfp.md: Add BFmode to movhf patterns.
3585 2020-01-16 Mihail Ionescu <mihail.ionescu@arm.com>
3586 Andre Vieira <andre.simoesdiasvieira@arm.com>
3588 * config/arm/arm-cpus.in (mve, mve_float): New features.
3589 (dsp, mve, mve.fp): New options.
3590 * config/arm/arm.h (TARGET_HAVE_MVE, TARGET_HAVE_MVE_FLOAT): Define.
3591 * config/arm/t-rmprofile: Map v8.1-M multilibs to v8-M.
3592 * doc/invoke.texi: Document the armv8.1-m mve and dps options.
3594 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3595 Thomas Preud'homme <thomas.preudhomme@arm.com>
3597 * config/arm/arm-cpus.in (ARMv8_1m_main): Redefine as an extension to
3599 * config/arm/arm.c (arm_options_perform_arch_sanity_checks): Remove
3600 error for using -mcmse when targeting Armv8.1-M Mainline.
3602 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3603 Thomas Preud'homme <thomas.preudhomme@arm.com>
3605 * config/arm/arm.md (nonsecure_call_internal): Do not force memory
3606 address in r4 when targeting Armv8.1-M Mainline.
3607 (nonsecure_call_value_internal): Likewise.
3608 * config/arm/thumb2.md (nonsecure_call_reg_thumb2): Make memory address
3609 a register match_operand again. Emit BLXNS when targeting
3611 (nonsecure_call_value_reg_thumb2): Likewise.
3613 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3614 Thomas Preud'homme <thomas.preudhomme@arm.com>
3616 * config/arm/arm.c (arm_add_cfa_adjust_cfa_note): Declare early.
3617 (cmse_nonsecure_call_inline_register_clear): Define new lazy_fpclear
3618 variable as true when floating-point ABI is not hard. Replace
3619 check against TARGET_HARD_FLOAT_ABI by checks against lazy_fpclear.
3620 Generate VLSTM and VLLDM instruction respectively before and
3621 after a function call to cmse_nonsecure_call function.
3622 * config/arm/unspecs.md (VUNSPEC_VLSTM): Define unspec.
3623 (VUNSPEC_VLLDM): Likewise.
3624 * config/arm/vfp.md (lazy_store_multiple_insn): New define_insn.
3625 (lazy_load_multiple_insn): Likewise.
3627 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3628 Thomas Preud'homme <thomas.preudhomme@arm.com>
3630 * config/arm/arm.c (vfp_emit_fstmd): Declare early.
3631 (arm_emit_vfp_multi_reg_pop): Likewise.
3632 (cmse_nonsecure_call_inline_register_clear): Abstract number of VFP
3633 registers to clear in max_fp_regno. Emit VPUSH and VPOP to save and
3634 restore callee-saved VFP registers.
3636 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3637 Thomas Preud'homme <thomas.preudhomme@arm.com>
3639 * config/arm/arm.c (arm_emit_multi_reg_pop): Declare early.
3640 (cmse_nonsecure_call_clear_caller_saved): Rename into ...
3641 (cmse_nonsecure_call_inline_register_clear): This. Save and clear
3642 callee-saved GPRs as well as clear ip register before doing a nonsecure
3643 call then restore callee-saved GPRs after it when targeting
3645 (arm_reorg): Adapt to function rename.
3647 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3648 Thomas Preud'homme <thomas.preudhomme@arm.com>
3650 * config/arm/arm-protos.h (clear_operation_p): Adapt prototype.
3651 * config/arm/arm.c (clear_operation_p): Extend to be able to check a
3652 clear_vfp_multiple pattern based on a new vfp parameter.
3653 (cmse_clear_registers): Generate VSCCLRM to clear VFP registers when
3654 targeting Armv8.1-M Mainline.
3655 (cmse_nonsecure_entry_clear_before_return): Clear VFP registers
3656 unconditionally when targeting Armv8.1-M Mainline architecture. Check
3657 whether VFP registers are available before looking call_used_regs for a
3659 * config/arm/predicates.md (clear_multiple_operation): Adapt to change
3660 of prototype of clear_operation_p.
3661 (clear_vfp_multiple_operation): New predicate.
3662 * config/arm/unspecs.md (VUNSPEC_VSCCLRM_VPR): New volatile unspec.
3663 * config/arm/vfp.md (clear_vfp_multiple): New define_insn.
3665 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3666 Thomas Preud'homme <thomas.preudhomme@arm.com>
3668 * config/arm/arm-protos.h (clear_operation_p): Declare.
3669 * config/arm/arm.c (clear_operation_p): New function.
3670 (cmse_clear_registers): Generate clear_multiple instruction pattern if
3671 targeting Armv8.1-M Mainline or successor.
3672 (output_return_instruction): Only output APSR register clearing if
3673 Armv8.1-M Mainline instructions not available.
3674 (thumb_exit): Likewise.
3675 * config/arm/predicates.md (clear_multiple_operation): New predicate.
3676 * config/arm/thumb2.md (clear_apsr): New define_insn.
3677 (clear_multiple): Likewise.
3678 * config/arm/unspecs.md (VUNSPEC_CLRM_APSR): New volatile unspec.
3680 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3681 Thomas Preud'homme <thomas.preudhomme@arm.com>
3683 * config/arm/arm.c (fp_sysreg_names): Declare and define.
3684 (use_return_insn): Also return false for Armv8.1-M Mainline.
3685 (output_return_instruction): Skip FPSCR clearing if Armv8.1-M
3686 Mainline instructions are available.
3687 (arm_compute_frame_layout): Allocate space in frame for FPCXTNS
3688 when targeting Armv8.1-M Mainline Security Extensions.
3689 (arm_expand_prologue): Save FPCXTNS if this is an Armv8.1-M
3690 Mainline entry function.
3691 (cmse_nonsecure_entry_clear_before_return): Clear IP and r4 if
3692 targeting Armv8.1-M Mainline or successor.
3693 (arm_expand_epilogue): Fix indentation of caller-saved register
3694 clearing. Restore FPCXTNS if this is an Armv8.1-M Mainline
3696 * config/arm/arm.h (TARGET_HAVE_FP_CMSE): New macro.
3697 (FP_SYSREGS): Likewise.
3698 (enum vfp_sysregs_encoding): Define enum.
3699 (fp_sysreg_names): Declare.
3700 * config/arm/unspecs.md (VUNSPEC_VSTR_VLDR): New volatile unspec.
3701 * config/arm/vfp.md (push_fpsysreg_insn): New define_insn.
3702 (pop_fpsysreg_insn): Likewise.
3704 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3705 Thomas Preud'homme <thomas.preudhomme@arm.com>
3707 * config/arm/arm-cpus.in (armv8_1m_main): New feature.
3708 (ARMv4, ARMv4t, ARMv5t, ARMv5te, ARMv5tej, ARMv6, ARMv6j, ARMv6k,
3709 ARMv6z, ARMv6kz, ARMv6zk, ARMv6t2, ARMv6m, ARMv7, ARMv7a, ARMv7ve,
3710 ARMv7r, ARMv7m, ARMv7em, ARMv8a, ARMv8_1a, ARMv8_2a, ARMv8_3a,
3711 ARMv8_4a, ARMv8_5a, ARMv8m_base, ARMv8m_main, ARMv8r): Reindent.
3712 (ARMv8_1m_main): New feature group.
3713 (armv8.1-m.main): New architecture.
3714 * config/arm/arm-tables.opt: Regenerate.
3715 * config/arm/arm.c (arm_arch8_1m_main): Define and default initialize.
3716 (arm_option_reconfigure_globals): Initialize arm_arch8_1m_main.
3717 (arm_options_perform_arch_sanity_checks): Error out when targeting
3718 Armv8.1-M Mainline Security Extensions.
3719 * config/arm/arm.h (arm_arch8_1m_main): Declare.
3721 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
3723 * config/aarch64/aarch64-simd-builtins.def (aarch64_bfdot,
3724 aarch64_bfdot_lane, aarch64_bfdot_laneq): New.
3725 * config/aarch64/aarch64-simd.md (aarch64_bfdot, aarch64_bfdot_lane,
3726 aarch64_bfdot_laneq): New.
3727 * config/aarch64/arm_bf16.h (vbfdot_f32, vbfdotq_f32,
3728 vbfdot_lane_f32, vbfdotq_lane_f32, vbfdot_laneq_f32,
3729 vbfdotq_laneq_f32): New.
3730 * config/aarch64/iterators.md (UNSPEC_BFDOT, Vbfdottype,
3731 VBFMLA_W, VBF): New.
3732 (isquadop): Add V4BF, V8BF.
3734 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
3736 * config/aarch64/aarch64-builtins.c: (enum aarch64_type_qualifiers):
3737 New qualifier_lane_quadtup_index, TYPES_TERNOP_SSUS,
3738 TYPES_QUADOPSSUS_LANE_QUADTUP, TYPES_QUADOPSSSU_LANE_QUADTUP.
3739 (aarch64_simd_expand_args): Add case SIMD_ARG_LANE_QUADTUP_INDEX.
3740 (aarch64_simd_expand_builtin): Add qualifier_lane_quadtup_index.
3741 * config/aarch64/aarch64-simd-builtins.def (usdot, usdot_lane,
3742 usdot_laneq, sudot_lane,sudot_laneq): New.
3743 * config/aarch64/aarch64-simd.md (aarch64_usdot): New.
3744 (aarch64_<sur>dot_lane): New.
3745 * config/aarch64/arm_neon.h (vusdot_s32): New.
3747 (vusdot_lane_s32): New.
3748 (vsudot_lane_s32): New.
3749 * config/aarch64/iterators.md (DOTPROD_I8MM): New iterator.
3750 (UNSPEC_USDOT, UNSPEC_SUDOT): New unspecs.
3752 2020-01-16 Martin Liska <mliska@suse.cz>
3754 * value-prof.c (dump_histogram_value): Fix
3755 obvious spacing issue.
3757 2020-01-16 Andrew Pinski <apinski@marvell.com>
3759 * tree-ssa-sccvn.c(vn_reference_lookup_3): Check lhs for
3760 !storage_order_barrier_p.
3762 2020-01-16 Andrew Pinski <apinski@marvell.com>
3764 * sched-int.h (_dep): Add unused bit-field field for the padding.
3765 * sched-deps.c (init_dep_1): Init unused field.
3767 2020-01-16 Andrew Pinski <apinski@marvell.com>
3769 * optabs.h (create_expand_operand): Initialize target field also.
3771 2020-01-16 Andre Vieira <andre.simoesdiasvieira@arm.com>
3773 PR tree-optimization/92429
3774 * tree-ssa-loop-niter.h (simplify_replace_tree): Add parameter.
3775 * tree-ssa-loop-niter.c (simplify_replace_tree): Add parameter to
3777 * tree-vect-loop.c (update_epilogue_vinfo): Do not fold when replacing
3780 2020-01-16 Richard Sandiford <richard.sandiford@arm.com>
3782 * config/aarch64/aarch64.c (aarch64_split_sve_subreg_move): Apply
3783 aarch64_sve_int_mode to each mode.
3785 2020-01-15 David Malcolm <dmalcolm@redhat.com>
3787 * doc/analyzer.texi (Overview): Add note about
3788 -fdump-ipa-analyzer.
3790 2020-01-15 Wilco Dijkstra <wdijkstr@arm.com>
3792 PR tree-optimization/93231
3793 * tree-ssa-forwprop.c (optimize_count_trailing_zeroes): Check
3794 input_type is unsigned. Use tree_to_shwi for shift constant.
3795 Check CST_STRING element size is CHAR_TYPE_SIZE bits.
3796 (simplify_count_trailing_zeroes): Add test to handle known non-zero
3797 inputs more efficiently.
3799 2020-01-15 Uroš Bizjak <ubizjak@gmail.com>
3801 * config/i386/i386.md (*movsf_internal): Do not require
3802 SSE2 ISA for alternatives 14 and 15.
3804 2020-01-15 Richard Biener <rguenther@suse.de>
3807 * tree-eh.c (sink_clobbers): If we already visited the destination
3808 block do not defer insertion.
3809 (pass_lower_eh_dispatch::execute): Maintain BB_VISITED for
3810 the purpose of defered insertion.
3812 2020-01-15 Jakub Jelinek <jakub@redhat.com>
3814 * BASE-VER: Bump to 10.0.1.
3816 2020-01-15 Richard Sandiford <richard.sandiford@arm.com>
3818 PR tree-optimization/93247
3819 * tree-vect-loop.c (update_epilogue_loop_vinfo): Check the access
3820 type of the stmt that we're going to vectorize.
3822 2020-01-15 Richard Sandiford <richard.sandiford@arm.com>
3824 * tree-vect-slp.c (vectorize_slp_instance_root_stmt): Use a
3825 VIEW_CONVERT_EXPR if the vectorized constructor has a diffeent
3828 2020-01-15 Martin Liska <mliska@suse.cz>
3830 * ipa-profile.c (ipa_profile_read_edge_summary): Do not allow
3831 2 calls of streamer_read_hwi in a function call.
3833 2020-01-15 Richard Biener <rguenther@suse.de>
3835 * alias.c (record_alias_subset): Avoid redundant work when
3836 subset is already recorded.
3838 2020-01-14 David Malcolm <dmalcolm@redhat.com>
3840 * doc/invoke.texi (-fdiagnostics-show-cwe): Add note that some of
3841 the analyzer options provide CWE identifiers.
3843 2020-01-14 David Malcolm <dmalcolm@redhat.com>
3845 * tree-diagnostic-path.cc (path_summary::event_range::print):
3846 When testing for UNKNOWN_LOCATION, look through ad-hoc wrappers
3847 using get_pure_location.
3849 2020-01-15 Jakub Jelinek <jakub@redhat.com>
3851 PR tree-optimization/93262
3852 * tree-ssa-dse.c (maybe_trim_memstar_call): For *_chk builtins,
3853 perform head trimming only if the last argument is constant,
3854 either all ones, or larger or equal to head trim, in the latter
3855 case decrease the last argument by head_trim.
3857 PR tree-optimization/93249
3858 * tree-ssa-dse.c: Include builtins.h and gimple-fold.h.
3859 (maybe_trim_memstar_call): Move head_trim and tail_trim vars to
3860 function body scope, reindent. For BUILTIN_IN_STRNCPY*, don't
3861 perform head trim unless we can prove there are no '\0' chars
3862 from the source among the first head_trim chars.
3864 2020-01-14 David Malcolm <dmalcolm@redhat.com>
3866 * Makefile.in (ANALYZER_OBJS): Add analyzer/function-set.o.
3868 2020-01-15 Jakub Jelinek <jakub@redhat.com>
3871 * config/i386/sse.md
3872 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1,
3873 *<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1,
3874 *<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1,
3875 *<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1): Use
3876 just a single alternative instead of two, make operands 1 and 2
3879 2020-01-14 Jan Hubicka <hubicka@ucw.cz>
3882 * ipa-devirt.c (odr_types_equivalent_p): Compare TREE_ADDRESSABLE and
3885 2020-01-14 David Malcolm <dmalcolm@redhat.com>
3887 * Makefile.in (lang_opt_files): Add analyzer.opt.
3888 (ANALYZER_OBJS): New.
3889 (OBJS): Add digraph.o, graphviz.o, ordered-hash-map-tests.o,
3890 tristate.o and ANALYZER_OBJS.
3891 (TEXI_GCCINT_FILES): Add analyzer.texi.
3892 * common.opt (-fanalyzer): New driver option.
3893 * config.in: Regenerate.
3894 * configure: Regenerate.
3895 * configure.ac (--disable-analyzer, ENABLE_ANALYZER): New option.
3896 (gccdepdir): Also create depdir for "analyzer" subdir.
3897 * digraph.cc: New file.
3898 * digraph.h: New file.
3899 * doc/analyzer.texi: New file.
3900 * doc/gccint.texi ("Static Analyzer") New menu item.
3901 (analyzer.texi): Include it.
3902 * doc/invoke.texi ("Static Analyzer Options"): New list and new section.
3903 ("Warning Options"): Add static analysis warnings to the list.
3904 (-Wno-analyzer-double-fclose): New option.
3905 (-Wno-analyzer-double-free): New option.
3906 (-Wno-analyzer-exposure-through-output-file): New option.
3907 (-Wno-analyzer-file-leak): New option.
3908 (-Wno-analyzer-free-of-non-heap): New option.
3909 (-Wno-analyzer-malloc-leak): New option.
3910 (-Wno-analyzer-possible-null-argument): New option.
3911 (-Wno-analyzer-possible-null-dereference): New option.
3912 (-Wno-analyzer-null-argument): New option.
3913 (-Wno-analyzer-null-dereference): New option.
3914 (-Wno-analyzer-stale-setjmp-buffer): New option.
3915 (-Wno-analyzer-tainted-array-index): New option.
3916 (-Wno-analyzer-use-after-free): New option.
3917 (-Wno-analyzer-use-of-pointer-in-stale-stack-frame): New option.
3918 (-Wno-analyzer-use-of-uninitialized-value): New option.
3919 (-Wanalyzer-too-complex): New option.
3920 (-fanalyzer-call-summaries): New warning.
3921 (-fanalyzer-checker=): New warning.
3922 (-fanalyzer-fine-grained): New warning.
3923 (-fno-analyzer-state-merge): New warning.
3924 (-fno-analyzer-state-purge): New warning.
3925 (-fanalyzer-transitivity): New warning.
3926 (-fanalyzer-verbose-edges): New warning.
3927 (-fanalyzer-verbose-state-changes): New warning.
3928 (-fanalyzer-verbosity=): New warning.
3929 (-fdump-analyzer): New warning.
3930 (-fdump-analyzer-callgraph): New warning.
3931 (-fdump-analyzer-exploded-graph): New warning.
3932 (-fdump-analyzer-exploded-nodes): New warning.
3933 (-fdump-analyzer-exploded-nodes-2): New warning.
3934 (-fdump-analyzer-exploded-nodes-3): New warning.
3935 (-fdump-analyzer-supergraph): New warning.
3936 * doc/sourcebuild.texi (dg-require-dot): New.
3937 (dg-check-dot): New.
3938 * gdbinit.in (break-on-saved-diagnostic): New command.
3939 * graphviz.cc: New file.
3940 * graphviz.h: New file.
3941 * ordered-hash-map-tests.cc: New file.
3942 * ordered-hash-map.h: New file.
3943 * passes.def (pass_analyzer): Add before
3944 pass_ipa_whole_program_visibility.
3945 * selftest-run-tests.c (selftest::run_tests): Call
3946 selftest::ordered_hash_map_tests_cc_tests.
3947 * selftest.h (selftest::ordered_hash_map_tests_cc_tests): New
3949 * shortest-paths.h: New file.
3950 * timevar.def (TV_ANALYZER): New timevar.
3951 (TV_ANALYZER_SUPERGRAPH): Likewise.
3952 (TV_ANALYZER_STATE_PURGE): Likewise.
3953 (TV_ANALYZER_PLAN): Likewise.
3954 (TV_ANALYZER_SCC): Likewise.
3955 (TV_ANALYZER_WORKLIST): Likewise.
3956 (TV_ANALYZER_DUMP): Likewise.
3957 (TV_ANALYZER_DIAGNOSTICS): Likewise.
3958 (TV_ANALYZER_SHORTEST_PATHS): Likewise.
3959 * tree-pass.h (make_pass_analyzer): New decl.
3960 * tristate.cc: New file.
3961 * tristate.h: New file.
3963 2020-01-14 Uroš Bizjak <ubizjak@gmail.com>
3966 * config/i386/i386.md (*movsf_internal): Require SSE2 ISA for
3967 alternatives 9 and 10.
3969 2020-01-14 David Malcolm <dmalcolm@redhat.com>
3971 * attribs.c (excl_hash_traits::empty_zero_p): New static constant.
3972 * gcov.c (function_start_pair_hash::empty_zero_p): Likewise.
3973 * graphite.c (struct sese_scev_hash::empty_zero_p): Likewise.
3974 * hash-map-tests.c (selftest::test_nonzero_empty_key): New selftest.
3975 (selftest::hash_map_tests_c_tests): Call it.
3976 * hash-map-traits.h (simple_hashmap_traits::empty_zero_p):
3977 New static constant, using the value of = H::empty_zero_p.
3978 (unbounded_hashmap_traits::empty_zero_p): Likewise, using the value
3979 from default_hash_traits <Value>.
3980 * hash-map.h (hash_map::empty_zero_p): Likewise, using the value
3982 * hash-set-tests.c (value_hash_traits::empty_zero_p): Likewise.
3983 * hash-table.h (hash_table::alloc_entries): Guard the loop of
3984 calls to mark_empty with !Descriptor::empty_zero_p.
3985 (hash_table::empty_slow): Conditionalize the memset call with a
3986 check that Descriptor::empty_zero_p; otherwise, loop through the
3987 entries calling mark_empty on them.
3988 * hash-traits.h (int_hash::empty_zero_p): New static constant.
3989 (pointer_hash::empty_zero_p): Likewise.
3990 (pair_hash::empty_zero_p): Likewise.
3991 * ipa-devirt.c (default_hash_traits <type_pair>::empty_zero_p):
3993 * ipa-prop.c (ipa_bit_ggc_hash_traits::empty_zero_p): Likewise.
3994 (ipa_vr_ggc_hash_traits::empty_zero_p): Likewise.
3995 * profile.c (location_triplet_hash::empty_zero_p): Likewise.
3996 * sanopt.c (sanopt_tree_triplet_hash::empty_zero_p): Likewise.
3997 (sanopt_tree_couple_hash::empty_zero_p): Likewise.
3998 * tree-hasher.h (int_tree_hasher::empty_zero_p): Likewise.
3999 * tree-ssa-sccvn.c (vn_ssa_aux_hasher::empty_zero_p): Likewise.
4000 * tree-vect-slp.c (bst_traits::empty_zero_p): Likewise.
4002 (default_hash_traits<scalar_cond_masked_key>::empty_zero_p):
4005 2020-01-14 Kewen Lin <linkw@gcc.gnu.org>
4007 * cfgloopanal.c (average_num_loop_insns): Free bbs when early return,
4008 fix typo on return value.
4010 2020-01-14 Xiong Hu Luo <luoxhu@linux.ibm.com>
4013 * cgraph.c (symbol_table::create_edge): Init speculative_id and
4015 (cgraph_edge::make_speculative): Add param for setting speculative_id
4017 (cgraph_edge::speculative_call_info): Update comments and find reference
4018 by speculative_id for multiple indirect targets.
4019 (cgraph_edge::resolve_speculation): Decrease the speculations
4020 for indirect edge, drop it's speculative if not direct target
4021 left. Update comments.
4022 (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
4023 (cgraph_node::dump): Print num_speculative_call_targets.
4024 (cgraph_node::verify_node): Don't report error if speculative
4025 edge not include statement.
4026 (cgraph_edge::num_speculative_call_targets_p): New function.
4027 * cgraph.h (int common_target_id): Remove.
4028 (int common_target_probability): Remove.
4029 (num_speculative_call_targets): New variable.
4030 (make_speculative): Add param for setting speculative_id.
4031 (cgraph_edge::num_speculative_call_targets_p): New declare.
4032 (target_prob): New variable.
4033 (speculative_id): New variable.
4034 * ipa-fnsummary.c (analyze_function_body): Create and duplicate
4035 call summaries for multiple speculative call targets.
4036 * cgraphclones.c (cgraph_node::create_clone): Clone speculative_id.
4037 * ipa-profile.c (struct speculative_call_target): New struct.
4038 (class speculative_call_summary): New class.
4039 (class speculative_call_summaries): New class.
4040 (call_sums): New variable.
4041 (ipa_profile_generate_summary): Generate indirect multiple targets summaries.
4042 (ipa_profile_write_edge_summary): New function.
4043 (ipa_profile_write_summary): Stream out indirect multiple targets summaries.
4044 (ipa_profile_dump_all_summaries): New function.
4045 (ipa_profile_read_edge_summary): New function.
4046 (ipa_profile_read_summary_section): New function.
4047 (ipa_profile_read_summary): Stream in indirect multiple targets summaries.
4048 (ipa_profile): Generate num_speculative_call_targets from
4050 * ipa-ref.h (speculative_id): New variable.
4051 * ipa-utils.c (ipa_merge_profiles): Update with target_prob.
4052 * lto-cgraph.c (lto_output_edge): Remove indirect common_target_id and
4053 common_target_probability. Stream out speculative_id and
4054 num_speculative_call_targets.
4055 (input_edge): Likewise.
4056 * predict.c (dump_prediction): Remove edges count assert to be
4058 * symtab.c (symtab_node::create_reference): Init speculative_id.
4059 (symtab_node::clone_references): Clone speculative_id.
4060 (symtab_node::clone_referring): Clone speculative_id.
4061 (symtab_node::clone_reference): Clone speculative_id.
4062 (symtab_node::clear_stmts_in_references): Clear speculative_id.
4063 * tree-inline.c (copy_bb): Duplicate all the speculative edges
4064 if indirect call contains multiple speculative targets.
4065 * value-prof.h (check_ic_target): Remove.
4066 * value-prof.c (gimple_value_profile_transformations):
4067 Use void function gimple_ic_transform.
4068 * value-prof.c (gimple_ic_transform): Handle topn case.
4069 Fix comment typos. Change it to a void function.
4071 2020-01-13 Andrew Pinski <apinski@marvell.com>
4073 * config/aarch64/aarch64-cores.def (octeontx2): New define.
4074 (octeontx2t98): New define.
4075 (octeontx2t96): New define.
4076 (octeontx2t93): New define.
4077 (octeontx2f95): New define.
4078 (octeontx2f95n): New define.
4079 (octeontx2f95mm): New define.
4080 * config/aarch64/aarch64-tune.md: Regenerate.
4081 * doc/invoke.texi (-mcpu=): Document the new cpu types.
4083 2020-01-13 Jason Merrill <jason@redhat.com>
4085 PR c++/33799 - destroy return value if local cleanup throws.
4086 * gimplify.c (gimplify_return_expr): Handle COMPOUND_EXPR.
4088 2020-01-13 Martin Liska <mliska@suse.cz>
4090 * ipa-cp.c (get_max_overall_size): Use newly
4091 renamed param param_ipa_cp_unit_growth.
4092 * params.opt: Remove legacy param name.
4094 2020-01-13 Martin Sebor <msebor@redhat.com>
4096 PR tree-optimization/93213
4097 * tree-ssa-strlen.c (handle_store): Only allow single-byte nul-over-nul
4098 stores to be eliminated.
4100 2020-01-13 Martin Liska <mliska@suse.cz>
4102 * opts.c (print_help): Do not print CL_PARAM
4103 and CL_WARNING for CL_OPTIMIZATION.
4105 2020-01-13 Jonathan Wakely <jwakely@redhat.com>
4108 * doc/invoke.texi (Warning Options): Add caveat about some warnings
4109 depending on optimization settings.
4111 2020-01-13 Jakub Jelinek <jakub@redhat.com>
4113 PR tree-optimization/90838
4114 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
4115 SCALAR_INT_TYPE_MODE directly in CTZ_DEFINED_VALUE_AT_ZERO macro
4116 argument rather than to initialize temporary for targets that
4117 don't use the mode argument at all. Initialize ctzval to avoid
4120 2020-01-10 Thomas Schwinge <thomas@codesourcery.com>
4122 * tree.h (OMP_CLAUSE_USE_DEVICE_PTR_IF_PRESENT): New definition.
4123 * tree-core.h: Document it.
4124 * gimplify.c (gimplify_omp_workshare): Set it.
4125 * omp-low.c (lower_omp_target): Use it.
4126 * tree-pretty-print.c (dump_omp_clause): Print it.
4128 * omp-low.c (lower_omp_target) <OMP_CLAUSE_USE_DEVICE_PTR etc.>:
4129 Assert that for OpenACC we always have 'GOMP_MAP_USE_DEVICE_PTR'.
4131 2020-01-10 David Malcolm <dmalcolm@redhat.com>
4133 * Makefile.in (OBJS): Add tree-diagnostic-path.o.
4134 * common.opt (fdiagnostics-path-format=): New option.
4135 (diagnostic_path_format): New enum.
4136 (fdiagnostics-show-path-depths): New option.
4137 * coretypes.h (diagnostic_event_id_t): New forward decl.
4138 * diagnostic-color.c (color_dict): Add "path".
4139 * diagnostic-event-id.h: New file.
4140 * diagnostic-format-json.cc (json_from_expanded_location): Make
4142 (json_end_diagnostic): Call context->make_json_for_path if it
4143 exists and the diagnostic has a path.
4144 (diagnostic_output_format_init): Clear context->print_path.
4145 * diagnostic-path.h: New file.
4146 * diagnostic-show-locus.c (colorizer::set_range): Special-case
4147 when printing a run of events in a diagnostic_path so that they
4148 all get the same color.
4149 (layout::m_diagnostic_path_p): New field.
4150 (layout::layout): Initialize it.
4151 (layout::print_any_labels): Don't colorize the label text for an
4152 event in a diagnostic_path.
4153 (gcc_rich_location::add_location_if_nearby): Add
4154 "restrict_to_current_line_spans" and "label" params. Pass the
4155 former to layout.maybe_add_location_range; pass the latter
4156 when calling add_range.
4157 * diagnostic.c: Include "diagnostic-path.h".
4158 (diagnostic_initialize): Initialize context->path_format and
4159 context->show_path_depths.
4160 (diagnostic_show_any_path): New function.
4161 (diagnostic_path::interprocedural_p): New function.
4162 (diagnostic_report_diagnostic): Call diagnostic_show_any_path.
4163 (simple_diagnostic_path::num_events): New function.
4164 (simple_diagnostic_path::get_event): New function.
4165 (simple_diagnostic_path::add_event): New function.
4166 (simple_diagnostic_event::simple_diagnostic_event): New ctor.
4167 (simple_diagnostic_event::~simple_diagnostic_event): New dtor.
4168 (debug): New overload taking a diagnostic_path *.
4169 * diagnostic.def (DK_DIAGNOSTIC_PATH): New.
4170 * diagnostic.h (enum diagnostic_path_format): New enum.
4171 (json::value): New forward decl.
4172 (diagnostic_context::path_format): New field.
4173 (diagnostic_context::show_path_depths): New field.
4174 (diagnostic_context::print_path): New callback field.
4175 (diagnostic_context::make_json_for_path): New callback field.
4176 (diagnostic_show_any_path): New decl.
4177 (json_from_expanded_location): New decl.
4178 * doc/invoke.texi (-fdiagnostics-path-format=): New option.
4179 (-fdiagnostics-show-path-depths): New option.
4180 (-fdiagnostics-color): Add "path" to description of default
4181 GCC_COLORS; describe it.
4182 (-fdiagnostics-format=json): Document how diagnostic paths are
4183 represented in the JSON output format.
4184 * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
4185 Add optional params "restrict_to_current_line_spans" and "label".
4186 * opts.c (common_handle_option): Handle
4187 OPT_fdiagnostics_path_format_ and
4188 OPT_fdiagnostics_show_path_depths.
4189 * pretty-print.c: Include "diagnostic-event-id.h".
4190 (pp_format): Implement "%@" format code for printing
4191 diagnostic_event_id_t *.
4192 (selftest::test_pp_format): Add tests for "%@".
4193 * selftest-run-tests.c (selftest::run_tests): Call
4194 selftest::tree_diagnostic_path_cc_tests.
4195 * selftest.h (selftest::tree_diagnostic_path_cc_tests): New decl.
4196 * toplev.c (general_init): Initialize global_dc->path_format and
4197 global_dc->show_path_depths.
4198 * tree-diagnostic-path.cc: New file.
4199 * tree-diagnostic.c (maybe_unwind_expanded_macro_loc): Make
4200 non-static. Drop "diagnostic" param in favor of storing the
4201 original value of "where" and re-using it.
4202 (virt_loc_aware_diagnostic_finalizer): Update for dropped param of
4203 maybe_unwind_expanded_macro_loc.
4204 (tree_diagnostics_defaults): Initialize context->print_path and
4205 context->make_json_for_path.
4206 * tree-diagnostic.h (default_tree_diagnostic_path_printer): New
4208 (default_tree_make_json_for_path): New decl.
4209 (maybe_unwind_expanded_macro_loc): New decl.
4211 2020-01-10 Jakub Jelinek <jakub@redhat.com>
4213 PR tree-optimization/93210
4214 * fold-const.h (native_encode_initializer,
4215 can_native_interpret_type_p): Declare.
4216 * fold-const.c (native_encode_string): Fix up handling with off != -1,
4218 (native_encode_initializer): New function, moved from dwarf2out.c.
4219 Adjust to native_encode_expr compatible arguments, including dry-run
4220 and partial extraction modes. Don't handle STRING_CST.
4221 (can_native_interpret_type_p): No longer static.
4222 * gimple-fold.c (fold_ctor_reference): For native_encode_expr, verify
4223 offset / BITS_PER_UNIT fits into int and don't call it if
4224 can_native_interpret_type_p fails. If suboff is NULL and for
4225 CONSTRUCTOR fold_{,non}array_ctor_reference returns NULL, retry with
4226 native_encode_initializer.
4227 (fold_const_aggregate_ref_1): Formatting fix.
4228 * dwarf2out.c (native_encode_initializer): Moved to fold-const.c.
4229 (tree_add_const_value_attribute): Adjust caller.
4231 PR tree-optimization/90838
4232 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
4233 SCALAR_INT_TYPE_MODE instead of TYPE_MODE as operand of
4234 CTZ_DEFINED_VALUE_AT_ZERO.
4236 2020-01-10 Vladimir Makarov <vmakarov@redhat.com>
4239 * lra-constraints.c (match_reload): Permit input operands have the
4240 same mode as output while other input operands have a different
4243 2020-01-10 Wilco Dijkstra <wdijkstr@arm.com>
4245 PR tree-optimization/90838
4246 * tree-ssa-forwprop.c (check_ctz_array): Add new function.
4247 (check_ctz_string): Likewise.
4248 (optimize_count_trailing_zeroes): Likewise.
4249 (simplify_count_trailing_zeroes): Likewise.
4250 (pass_forwprop::execute): Try ctz simplification.
4251 * match.pd: Add matching for ctz idioms.
4253 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
4255 * config/aarch64/aarch64.c (aarch64_invalid_conversion): New function
4257 (aarch64_invalid_unary_op): New function for target hook.
4258 (aarch64_invalid_binary_op): New function for target hook.
4260 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
4262 * config.gcc: Add arm_bf16.h.
4263 * config/aarch64/aarch64-builtins.c
4264 (aarch64_simd_builtin_std_type): Add BFmode.
4265 (aarch64_init_simd_builtin_types): Define element types for vector
4267 (aarch64_init_bf16_types): New function.
4268 (aarch64_general_init_builtins): Add arm_init_bf16_types function call.
4269 * config/aarch64/aarch64-modes.def: Add BFmode and V4BF, V8BF vector
4271 * config/aarch64/aarch64-simd-builtin-types.def: Add BF SIMD types.
4272 * config/aarch64/aarch64-simd.md: Add BF vector types to NEON move
4274 * config/aarch64/aarch64.h (AARCH64_VALID_SIMD_DREG_MODE): Add V4BF.
4275 (AARCH64_VALID_SIMD_QREG_MODE): Add V8BF.
4276 * config/aarch64/aarch64.c
4277 (aarch64_classify_vector_mode): Add support for BF types.
4278 (aarch64_gimplify_va_arg_expr): Add support for BF types.
4279 (aarch64_vq_mode): Add support for BF types.
4280 (aarch64_simd_container_mode): Add support for BF types.
4281 (aarch64_mangle_type): Add support for BF scalar type.
4282 * config/aarch64/aarch64.md: Add BFmode to movhf pattern.
4283 * config/aarch64/arm_bf16.h: New file.
4284 * config/aarch64/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
4285 * config/aarch64/iterators.md: Add BF types to mode attributes.
4286 (HFBF, GPF_TF_F16_MOV, VDMOV, VQMOV, VQMOV_NO2Em VALL_F16MOV): New.
4288 2020-01-10 Jason Merrill <jason@redhat.com>
4290 PR c++/93173 - incorrect tree sharing.
4291 * gimplify.c (copy_if_shared): No longer static.
4292 * gimplify.h: Declare it.
4294 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
4296 * doc/invoke.texi (-msve-vector-bits=): Document that
4297 -msve-vector-bits=128 now generates VL-specific code for
4298 little-endian targets.
4299 * config/aarch64/aarch64-sve-builtins.cc (register_builtin_types): Use
4300 build_vector_type_for_mode to construct the data vector types.
4301 * config/aarch64/aarch64.c (aarch64_convert_sve_vector_bits): Generate
4302 VL-specific code for -msve-vector-bits=128 on little-endian targets.
4303 (aarch64_simd_container_mode): Always prefer Advanced SIMD modes
4304 for 128-bit vectors.
4306 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
4308 * config/aarch64/aarch64.c (aarch64_evpc_sel): Fix gen_vcond_mask
4311 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
4313 * config/aarch64/aarch64-builtins.c
4314 (aarch64_builtin_vectorized_function): Check for specific vector modes,
4315 rather than checking the number of elements and the element mode.
4317 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
4319 * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
4320 get_related_vectype_for_scalar_type rather than build_vector_type
4321 to create the index type for a conditional reduction.
4323 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
4325 * tree-vect-loop.c (update_epilogue_loop_vinfo): Update DR_REF
4326 for any type of gather or scatter, including strided accesses.
4328 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
4330 * tree-vectorizer.h (get_dr_vinfo_offset): Add missing function
4333 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
4335 * tree-vect-data-refs.c (vect_create_addr_base_for_vector_ref): Use
4337 * tree-vect-loop.c (update_epilogue_loop_vinfo): Remove orig_drs_init
4338 parameter and its use to reset DR_OFFSET's.
4339 (vect_transform_loop): Remove orig_drs_init argument.
4340 * tree-vect-loop-manip.c (vect_update_init_of_dr): Update the offset
4341 member of dr_vec_info rather than the offset of the associated
4342 data_reference's innermost_loop_behavior.
4343 (vect_update_init_of_dr): Pass dr_vec_info instead of data_reference.
4344 (vect_do_peeling): Remove orig_drs_init parameter and its construction.
4345 * tree-vect-stmts.c (check_scan_store): Replace use of DR_OFFSET with
4346 get_dr_vinfo_offset.
4347 (vectorizable_store): Likewise.
4348 (vectorizable_load): Likewise.
4350 2020-01-10 Richard Biener <rguenther@suse.de>
4352 * gimple-ssa-store-merging
4353 (pass_store_merging::terminate_all_aliasing_chains): Cache alias info.
4355 2020-01-10 Martin Liska <mliska@suse.cz>
4358 * ipa-inline-analysis.c (offline_size): Make proper parenthesis
4359 encapsulation that was there before r280040.
4361 2020-01-10 Richard Biener <rguenther@suse.de>
4364 * tree-eh.c (sink_clobbers): Move clobbers to out-of-IL
4365 sequences to avoid walking them again for secondary opportunities.
4366 (pass_lower_eh_dispatch::execute): Instead actually insert
4369 2020-01-10 Richard Biener <rguenther@suse.de>
4372 * tree-eh.c (redirect_eh_edge_1): Avoid some work if possible.
4373 (cleanup_all_empty_eh): Walk landing pads in reverse order to
4374 avoid quadraticness.
4376 2020-01-10 Martin Jambor <mjambor@suse.cz>
4378 * params.opt (param_ipa_sra_max_replacements): Mark as Optimization.
4379 * ipa-sra.c (pull_accesses_from_callee): New parameter caller, use it
4380 to get param_ipa_sra_max_replacements.
4381 (param_splitting_across_edge): Pass the caller to
4382 pull_accesses_from_callee.
4384 2020-01-10 Martin Jambor <mjambor@suse.cz>
4386 * params.opt (param_ipcp_unit_growth): Mark as Optimization.
4387 * ipa-cp.c (max_new_size): Removed.
4388 (orig_overall_size): New variable.
4389 (get_max_overall_size): New function.
4390 (estimate_local_effects): Use it. Adjust dump.
4391 (decide_about_value): Likewise.
4392 (ipcp_propagate_stage): Do not calculate max_new_size, just store
4393 orig_overall_size. Adjust dump.
4394 (ipa_cp_c_finalize): Clear orig_overall_size instead of max_new_size.
4396 2020-01-10 Martin Jambor <mjambor@suse.cz>
4398 * params.opt (param_ipa_max_agg_items): Mark as Optimization
4399 * ipa-cp.c (merge_agg_lats_step): New parameter max_agg_items, use
4400 instead of param_ipa_max_agg_items.
4401 (merge_aggregate_lattices): Extract param_ipa_max_agg_items from
4402 optimization info for the callee.
4404 2020-01-09 Kwok Cheung Yeung <kcy@codesourcery.com>
4406 * lto-streamer-in.c (input_function): Remove streamed-in inline debug
4407 markers if debug_inline_points is false.
4409 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4411 * config.gcc (aarch64*-*-*): Add aarch64-sve-builtins-sve2.o to
4413 * config/aarch64/t-aarch64 (aarch64-sve-builtins.o): Depend on
4414 aarch64-sve-builtins-base.def, aarch64-sve-builtins-sve2.def and
4415 aarch64-sve-builtins-sve2.h.
4416 (aarch64-sve-builtins-sve2.o): New rule.
4417 * config/aarch64/aarch64.h (AARCH64_ISA_SVE2_AES): New macro.
4418 (AARCH64_ISA_SVE2_BITPERM, AARCH64_ISA_SVE2_SHA3): Likewise.
4419 (AARCH64_ISA_SVE2_SM4, TARGET_SVE2_AES, TARGET_SVE2_BITPERM): Likewise.
4420 (TARGET_SVE2_SHA, TARGET_SVE2_SM4): Likewise.
4421 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
4422 TARGET_SVE2_AES, TARGET_SVE2_BITPERM, TARGET_SVE2_SHA3 and
4424 * config/aarch64/aarch64-sve.md: Update comments with SVE2
4425 instructions that are handled here.
4426 (@cond_asrd<mode>): Generalize to...
4427 (@cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>): ...this.
4428 (*cond_asrd<mode>_2): Generalize to...
4429 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_2): ...this.
4430 (*cond_asrd<mode>_z): Generalize to...
4431 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_z): ...this.
4432 * config/aarch64/aarch64.md (UNSPEC_LDNT1_GATHER): New unspec.
4433 (UNSPEC_STNT1_SCATTER, UNSPEC_WHILEGE, UNSPEC_WHILEGT): Likewise.
4434 (UNSPEC_WHILEHI, UNSPEC_WHILEHS): Likewise.
4435 * config/aarch64/aarch64-sve2.md (@aarch64_gather_ldnt<mode>): New
4437 (@aarch64_gather_ldnt_<ANY_EXTEND:optab><SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
4438 (@aarch64_scatter_stnt<mode>): Likewise.
4439 (@aarch64_scatter_stnt_<SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
4440 (@aarch64_mul_lane_<mode>): Likewise.
4441 (@aarch64_sve_suqadd<mode>_const): Likewise.
4442 (*<sur>h<addsub><mode>): Generalize to...
4443 (@aarch64_pred_<SVE2_COND_INT_BINARY_REV:sve_int_op><mode>): ...this
4445 (@cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>): New expander.
4446 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_2): New pattern.
4447 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_3): Likewise.
4448 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_any): Likewise.
4449 (*cond_<SVE2_COND_INT_BINARY_NOREV:sve_int_op><mode>_z): Likewise.
4450 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op><mode>):: Likewise.
4451 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op>_lane_<mode>): Likewise.
4452 (@aarch64_pred_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): Likewise.
4453 (@cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): New expander.
4454 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_2): New pattern.
4455 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_3): Likewise.
4456 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_any): Likewise.
4457 (@aarch64_sve_<SVE2_INT_TERNARY:sve_int_op><mode>): Likewise.
4458 (@aarch64_sve_<SVE2_INT_TERNARY_LANE:sve_int_op>_lane_<mode>)
4459 (@aarch64_sve_add_mul_lane_<mode>): Likewise.
4460 (@aarch64_sve_sub_mul_lane_<mode>): Likewise.
4461 (@aarch64_sve2_xar<mode>): Likewise.
4462 (@aarch64_sve2_bcax<mode>): Likewise.
4463 (*aarch64_sve2_eor3<mode>): Rename to...
4464 (@aarch64_sve2_eor3<mode>): ...this.
4465 (@aarch64_sve2_bsl<mode>): New expander.
4466 (@aarch64_sve2_nbsl<mode>): Likewise.
4467 (@aarch64_sve2_bsl1n<mode>): Likewise.
4468 (@aarch64_sve2_bsl2n<mode>): Likewise.
4469 (@aarch64_sve_add_<SHIFTRT:sve_int_op><mode>): Likewise.
4470 (*aarch64_sve2_sra<mode>): Add MOVPRFX support.
4471 (@aarch64_sve_add_<VRSHR_N:sve_int_op><mode>): New pattern.
4472 (@aarch64_sve_<SVE2_INT_SHIFT_INSERT:sve_int_op><mode>): Likewise.
4473 (@aarch64_sve2_<USMAX:su>aba<mode>): New expander.
4474 (*aarch64_sve2_<USMAX:su>aba<mode>): New pattern.
4475 (@aarch64_sve_<SVE2_INT_BINARY_WIDE:sve_int_op><mode>): Likewise.
4476 (<su>mull<bt><Vwide>): Generalize to...
4477 (@aarch64_sve_<SVE2_INT_BINARY_LONG:sve_int_op><mode>): ...this new
4479 (@aarch64_sve_<SVE2_INT_BINARY_LONG_lANE:sve_int_op>_lane_<mode>)
4480 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_LONG:sve_int_op><mode>)
4481 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG:sve_int_op><mode>)
4482 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
4483 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG:sve_int_op><mode>)
4484 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
4485 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG:sve_int_op><mode>)
4486 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
4487 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG:sve_int_op><mode>)
4488 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
4489 (@aarch64_sve_<SVE2_FP_TERNARY_LONG:sve_fp_op><mode>): New patterns.
4490 (@aarch64_<SVE2_FP_TERNARY_LONG_LANE:sve_fp_op>_lane_<mode>)
4491 (@aarch64_sve_<SVE2_INT_UNARY_NARROWB:sve_int_op><mode>): Likewise.
4492 (@aarch64_sve_<SVE2_INT_UNARY_NARROWT:sve_int_op><mode>): Likewise.
4493 (@aarch64_sve_<SVE2_INT_BINARY_NARROWB:sve_int_op><mode>): Likewise.
4494 (@aarch64_sve_<SVE2_INT_BINARY_NARROWT:sve_int_op><mode>): Likewise.
4495 (<SHRNB:r>shrnb<mode>): Generalize to...
4496 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWB:sve_int_op><mode>): ...this
4498 (<SHRNT:r>shrnt<mode>): Generalize to...
4499 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWT:sve_int_op><mode>): ...this
4501 (@aarch64_pred_<SVE2_INT_BINARY_PAIR:sve_int_op><mode>): New pattern.
4502 (@aarch64_pred_<SVE2_FP_BINARY_PAIR:sve_fp_op><mode>): Likewise.
4503 (@cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>): New expander.
4504 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_2): New pattern.
4505 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_z): Likewise.
4506 (@aarch64_sve_<SVE2_INT_CADD:optab><mode>): Likewise.
4507 (@aarch64_sve_<SVE2_INT_CMLA:optab><mode>): Likewise.
4508 (@aarch64_<SVE2_INT_CMLA:optab>_lane_<mode>): Likewise.
4509 (@aarch64_sve_<SVE2_INT_CDOT:optab><mode>): Likewise.
4510 (@aarch64_<SVE2_INT_CDOT:optab>_lane_<mode>): Likewise.
4511 (@aarch64_pred_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): Likewise.
4512 (@cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New expander.
4513 (*cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New pattern.
4514 (@aarch64_sve2_cvtnt<mode>): Likewise.
4515 (@aarch64_pred_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): Likewise.
4516 (@cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): New expander.
4517 (*cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>_any): New pattern.
4518 (@aarch64_sve2_cvtxnt<mode>): Likewise.
4519 (@aarch64_pred_<SVE2_U32_UNARY:sve_int_op><mode>): Likewise.
4520 (@cond_<SVE2_U32_UNARY:sve_int_op><mode>): New expander.
4521 (*cond_<SVE2_U32_UNARY:sve_int_op><mode>): New pattern.
4522 (@aarch64_pred_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): Likewise.
4523 (@cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New expander.
4524 (*cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New pattern.
4525 (@aarch64_sve2_pmul<mode>): Likewise.
4526 (@aarch64_sve_<SVE2_PMULL:optab><mode>): Likewise.
4527 (@aarch64_sve_<SVE2_PMULL_PAIR:optab><mode>): Likewise.
4528 (@aarch64_sve2_tbl2<mode>): Likewise.
4529 (@aarch64_sve2_tbx<mode>): Likewise.
4530 (@aarch64_sve_<SVE2_INT_BITPERM:sve_int_op><mode>): Likewise.
4531 (@aarch64_sve2_histcnt<mode>): Likewise.
4532 (@aarch64_sve2_histseg<mode>): Likewise.
4533 (@aarch64_pred_<SVE2_MATCH:sve_int_op><mode>): Likewise.
4534 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_cc): Likewise.
4535 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_ptest): Likewise.
4536 (aarch64_sve2_aes<CRYPTO_AES:aes_op>): Likewise.
4537 (aarch64_sve2_aes<CRYPTO_AESMC:aesmc_op>): Likewise.
4538 (*aarch64_sve2_aese_fused, *aarch64_sve2_aesd_fused): Likewise.
4539 (aarch64_sve2_rax1, aarch64_sve2_sm4e, aarch64_sve2_sm4ekey): Likewise.
4540 (<su>mulh<r>s<mode>3): Update after above pattern name changes.
4541 * config/aarch64/iterators.md (VNx16QI_ONLY, VNx4SF_ONLY)
4542 (SVE_STRUCT2, SVE_FULL_BHI, SVE_FULL_HSI, SVE_FULL_HDI)
4543 (SVE2_PMULL_PAIR_I): New mode iterators.
4544 (UNSPEC_ADCLB, UNSPEC_ADCLT, UNSPEC_ADDHNB, UNSPEC_ADDHNT, UNSPEC_BDEP)
4545 (UNSPEC_BEXT, UNSPEC_BGRP, UNSPEC_CADD90, UNSPEC_CADD270, UNSPEC_CDOT)
4546 (UNSPEC_CDOT90, UNSPEC_CDOT180, UNSPEC_CDOT270, UNSPEC_CMLA)
4547 (UNSPEC_CMLA90, UNSPEC_CMLA180, UNSPEC_CMLA270, UNSPEC_COND_FCVTLT)
4548 (UNSPEC_COND_FCVTNT, UNSPEC_COND_FCVTX, UNSPEC_COND_FCVTXNT)
4549 (UNSPEC_COND_FLOGB, UNSPEC_EORBT, UNSPEC_EORTB, UNSPEC_FADDP)
4550 (UNSPEC_FMAXP, UNSPEC_FMAXNMP, UNSPEC_FMLALB, UNSPEC_FMLALT)
4551 (UNSPEC_FMLSLB, UNSPEC_FMLSLT, UNSPEC_FMINP, UNSPEC_FMINNMP)
4552 (UNSPEC_HISTCNT, UNSPEC_HISTSEG, UNSPEC_MATCH, UNSPEC_NMATCH)
4553 (UNSPEC_PMULLB, UNSPEC_PMULLB_PAIR, UNSPEC_PMULLT, UNSPEC_PMULLT_PAIR)
4554 (UNSPEC_RADDHNB, UNSPEC_RADDHNT, UNSPEC_RSUBHNB, UNSPEC_RSUBHNT)
4555 (UNSPEC_SLI, UNSPEC_SRI, UNSPEC_SABDLB, UNSPEC_SABDLT, UNSPEC_SADDLB)
4556 (UNSPEC_SADDLBT, UNSPEC_SADDLT, UNSPEC_SADDWB, UNSPEC_SADDWT)
4557 (UNSPEC_SBCLB, UNSPEC_SBCLT, UNSPEC_SMAXP, UNSPEC_SMINP)
4558 (UNSPEC_SQCADD90, UNSPEC_SQCADD270, UNSPEC_SQDMULLB, UNSPEC_SQDMULLBT)
4559 (UNSPEC_SQDMULLT, UNSPEC_SQRDCMLAH, UNSPEC_SQRDCMLAH90)
4560 (UNSPEC_SQRDCMLAH180, UNSPEC_SQRDCMLAH270, UNSPEC_SQRSHRNB)
4561 (UNSPEC_SQRSHRNT, UNSPEC_SQRSHRUNB, UNSPEC_SQRSHRUNT, UNSPEC_SQSHRNB)
4562 (UNSPEC_SQSHRNT, UNSPEC_SQSHRUNB, UNSPEC_SQSHRUNT, UNSPEC_SQXTNB)
4563 (UNSPEC_SQXTNT, UNSPEC_SQXTUNB, UNSPEC_SQXTUNT, UNSPEC_SSHLLB)
4564 (UNSPEC_SSHLLT, UNSPEC_SSUBLB, UNSPEC_SSUBLBT, UNSPEC_SSUBLT)
4565 (UNSPEC_SSUBLTB, UNSPEC_SSUBWB, UNSPEC_SSUBWT, UNSPEC_SUBHNB)
4566 (UNSPEC_SUBHNT, UNSPEC_TBL2, UNSPEC_UABDLB, UNSPEC_UABDLT)
4567 (UNSPEC_UADDLB, UNSPEC_UADDLT, UNSPEC_UADDWB, UNSPEC_UADDWT)
4568 (UNSPEC_UMAXP, UNSPEC_UMINP, UNSPEC_UQRSHRNB, UNSPEC_UQRSHRNT)
4569 (UNSPEC_UQSHRNB, UNSPEC_UQSHRNT, UNSPEC_UQXTNB, UNSPEC_UQXTNT)
4570 (UNSPEC_USHLLB, UNSPEC_USHLLT, UNSPEC_USUBLB, UNSPEC_USUBLT)
4571 (UNSPEC_USUBWB, UNSPEC_USUBWT): New unspecs.
4572 (UNSPEC_SMULLB, UNSPEC_SMULLT, UNSPEC_UMULLB, UNSPEC_UMULLT)
4573 (UNSPEC_SMULHS, UNSPEC_SMULHRS, UNSPEC_UMULHS, UNSPEC_UMULHRS)
4574 (UNSPEC_RSHRNB, UNSPEC_RSHRNT, UNSPEC_SHRNB, UNSPEC_SHRNT): Move
4576 (VNARROW, Ventype): New mode attributes.
4577 (Vewtype): Handle VNx2DI. Fix typo in comment.
4578 (VDOUBLE): New mode attribute.
4579 (sve_lane_con): Handle VNx8HI.
4580 (SVE_INT_UNARY): Include ss_abs and ss_neg for TARGET_SVE2.
4581 (SVE_INT_BINARY): Likewise ss_plus, us_plus, ss_minus and us_minus.
4582 (sve_int_op, sve_int_op_rev): Handle the above codes.
4583 (sve_pred_int_rhs2_operand): Likewise.
4584 (MULLBT, SHRNB, SHRNT): Delete.
4585 (SVE_INT_SHIFT_IMM): New int iterator.
4586 (SVE_WHILE): Add UNSPEC_WHILEGE, UNSPEC_WHILEGT, UNSPEC_WHILEHI
4587 and UNSPEC_WHILEHS for TARGET_SVE2.
4588 (SVE2_U32_UNARY, SVE2_INT_UNARY_NARROWB, SVE2_INT_UNARY_NARROWT)
4589 (SVE2_INT_BINARY, SVE2_INT_BINARY_LANE, SVE2_INT_BINARY_LONG)
4590 (SVE2_INT_BINARY_LONG_LANE, SVE2_INT_BINARY_NARROWB)
4591 (SVE2_INT_BINARY_NARROWT, SVE2_INT_BINARY_PAIR, SVE2_FP_BINARY_PAIR)
4592 (SVE2_INT_BINARY_PAIR_LONG, SVE2_INT_BINARY_WIDE): New int iterators.
4593 (SVE2_INT_SHIFT_IMM_LONG, SVE2_INT_SHIFT_IMM_NARROWB): Likewise.
4594 (SVE2_INT_SHIFT_IMM_NARROWT, SVE2_INT_SHIFT_INSERT, SVE2_INT_CADD)
4595 (SVE2_INT_BITPERM, SVE2_INT_TERNARY, SVE2_INT_TERNARY_LANE): Likewise.
4596 (SVE2_FP_TERNARY_LONG, SVE2_FP_TERNARY_LONG_LANE, SVE2_INT_CMLA)
4597 (SVE2_INT_CDOT, SVE2_INT_ADD_BINARY_LONG, SVE2_INT_QADD_BINARY_LONG)
4598 (SVE2_INT_SUB_BINARY_LONG, SVE2_INT_QSUB_BINARY_LONG): Likewise.
4599 (SVE2_INT_ADD_BINARY_LONG_LANE, SVE2_INT_QADD_BINARY_LONG_LANE)
4600 (SVE2_INT_SUB_BINARY_LONG_LANE, SVE2_INT_QSUB_BINARY_LONG_LANE)
4601 (SVE2_COND_INT_UNARY_FP, SVE2_COND_FP_UNARY_LONG): Likewise.
4602 (SVE2_COND_FP_UNARY_NARROWB, SVE2_COND_INT_BINARY): Likewise.
4603 (SVE2_COND_INT_BINARY_NOREV, SVE2_COND_INT_BINARY_REV): Likewise.
4604 (SVE2_COND_INT_SHIFT, SVE2_MATCH, SVE2_PMULL): Likewise.
4605 (optab): Handle the new unspecs.
4606 (su, r): Remove entries for UNSPEC_SHRNB, UNSPEC_SHRNT, UNSPEC_RSHRNB
4608 (lr): Handle the new unspecs.
4610 (cmp_op, while_optab_cmp, sve_int_op): Handle the new unspecs.
4611 (sve_int_op_rev, sve_int_add_op, sve_int_qadd_op, sve_int_sub_op)
4612 (sve_int_qsub_op): New int attributes.
4613 (sve_fp_op, rot): Handle the new unspecs.
4614 * config/aarch64/aarch64-sve-builtins.h
4615 (function_resolver::require_matching_pointer_type): Declare.
4616 (function_resolver::resolve_unary): Add an optional boolean argument.
4617 (function_resolver::finish_opt_n_resolution): Add an optional
4618 type_suffix_index argument.
4619 (gimple_folder::redirect_call): Declare.
4620 (gimple_expander::prepare_gather_address_operands): Add an optional
4622 * config/aarch64/aarch64-sve-builtins.cc: Include
4623 aarch64-sve-builtins-sve2.h.
4624 (TYPES_b_unsigned, TYPES_b_integer, TYPES_bh_integer): New macros.
4625 (TYPES_bs_unsigned, TYPES_hs_signed, TYPES_hs_integer): Likewise.
4626 (TYPES_hd_unsigned, TYPES_hsd_signed): Likewise.
4627 (TYPES_hsd_integer): Use TYPES_hsd_signed.
4628 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): New macros.
4629 (TYPES_s_unsigned): Likewise.
4630 (TYPES_s_integer): Use TYPES_s_unsigned.
4631 (TYPES_sd_signed, TYPES_sd_unsigned): New macros.
4632 (TYPES_sd_integer): Use them.
4633 (TYPES_d_unsigned): New macro.
4634 (TYPES_d_integer): Use it.
4635 (TYPES_d_data, TYPES_cvt_long, TYPES_cvt_narrow_s): New macros.
4636 (TYPES_cvt_narrow): Likewise.
4637 (DEF_SVE_TYPES_ARRAY): Include the new types macros above.
4638 (preds_mx): New variable.
4639 (function_builder::add_overloaded_function): Allow the new feature
4640 set to be more restrictive than the original one.
4641 (function_resolver::infer_pointer_type): Remove qualifiers from
4642 the pointer type before printing it.
4643 (function_resolver::require_matching_pointer_type): New function.
4644 (function_resolver::resolve_sv_displacement): Handle functions
4645 that don't support 32-bit vector indices or svint32_t vector offsets.
4646 (function_resolver::finish_opt_n_resolution): Take the inferred type
4647 as a separate argument.
4648 (function_resolver::resolve_unary): Optionally treat all forms in
4649 the same way as normal merging functions.
4650 (gimple_folder::redirect_call): New function.
4651 (function_expander::prepare_gather_address_operands): Add an argument
4652 that says whether scaled forms are available. If they aren't,
4653 handle scaling of vector indices and don't add the extension and
4655 (function_expander::map_to_unspecs): If aarch64_sve isn't available,
4656 fall back to using cond_* instead.
4657 * config/aarch64/aarch64-sve-builtins-functions.h (rtx_code_function):
4658 Split out the member variables into...
4659 (rtx_code_function_base): ...this new base class.
4660 (rtx_code_function_rotated): Inherit rtx_code_function_base.
4661 (unspec_based_function): Split out the member variables into...
4662 (unspec_based_function_base): ...this new base class.
4663 (unspec_based_function_rotated): Inherit unspec_based_function_base.
4664 (unspec_based_function_exact_insn): New class.
4665 (unspec_based_add_function, unspec_based_add_lane_function)
4666 (unspec_based_lane_function, unspec_based_pred_function)
4667 (unspec_based_qadd_function, unspec_based_qadd_lane_function)
4668 (unspec_based_qsub_function, unspec_based_qsub_lane_function)
4669 (unspec_based_sub_function, unspec_based_sub_lane_function): New
4671 (unspec_based_fused_function): New class.
4672 (unspec_based_mla_function, unspec_based_mls_function): New typedefs.
4673 (unspec_based_fused_lane_function): New class.
4674 (unspec_based_mla_lane_function, unspec_based_mls_lane_function): New
4676 (CODE_FOR_MODE1): New macro.
4677 (fixed_insn_function): New class.
4678 (while_comparison): Likewise.
4679 * config/aarch64/aarch64-sve-builtins-shapes.h (binary_long_lane)
4680 (binary_long_opt_n, binary_narrowb_opt_n, binary_narrowt_opt_n)
4681 (binary_to_uint, binary_wide, binary_wide_opt_n, compare, compare_ptr)
4682 (load_ext_gather_index_restricted, load_ext_gather_offset_restricted)
4683 (load_gather_sv_restricted, shift_left_imm_long): Declare.
4684 (shift_left_imm_to_uint, shift_right_imm_narrowb): Likewise.
4685 (shift_right_imm_narrowt, shift_right_imm_narrowb_to_uint): Likewise.
4686 (shift_right_imm_narrowt_to_uint, store_scatter_index_restricted)
4687 (store_scatter_offset_restricted, tbl_tuple, ternary_long_lane)
4688 (ternary_long_opt_n, ternary_qq_lane_rotate, ternary_qq_rotate)
4689 (ternary_shift_left_imm, ternary_shift_right_imm, ternary_uint)
4690 (unary_convert_narrowt, unary_long, unary_narrowb, unary_narrowt)
4691 (unary_narrowb_to_uint, unary_narrowt_to_uint, unary_to_int): Likewise.
4692 * config/aarch64/aarch64-sve-builtins-shapes.cc (apply_predication):
4693 Also add an initial argument for unary_convert_narrowt, regardless
4694 of the predication type.
4695 (build_32_64): Allow loads and stores to specify MODE_none.
4696 (build_sv_index64, build_sv_uint_offset): New functions.
4697 (long_type_suffix): New function.
4698 (binary_imm_narrowb_base, binary_imm_narrowt_base): New classes.
4699 (binary_imm_long_base, load_gather_sv_base): Likewise.
4700 (shift_right_imm_narrow_wrapper, ternary_shift_imm_base): Likewise.
4701 (ternary_resize2_opt_n_base, ternary_resize2_lane_base): Likewise.
4702 (unary_narrowb_base, unary_narrowt_base): Likewise.
4703 (binary_long_lane_def, binary_long_lane): New shape.
4704 (binary_long_opt_n_def, binary_long_opt_n): Likewise.
4705 (binary_narrowb_opt_n_def, binary_narrowb_opt_n): Likewise.
4706 (binary_narrowt_opt_n_def, binary_narrowt_opt_n): Likewise.
4707 (binary_to_uint_def, binary_to_uint): Likewise.
4708 (binary_wide_def, binary_wide): Likewise.
4709 (binary_wide_opt_n_def, binary_wide_opt_n): Likewise.
4710 (compare_def, compare): Likewise.
4711 (compare_ptr_def, compare_ptr): Likewise.
4712 (load_ext_gather_index_restricted_def,
4713 load_ext_gather_index_restricted): Likewise.
4714 (load_ext_gather_offset_restricted_def,
4715 load_ext_gather_offset_restricted): Likewise.
4716 (load_gather_sv_def): Inherit from load_gather_sv_base.
4717 (load_gather_sv_restricted_def, load_gather_sv_restricted): New shape.
4718 (shift_left_imm_def, shift_left_imm): Likewise.
4719 (shift_left_imm_long_def, shift_left_imm_long): Likewise.
4720 (shift_left_imm_to_uint_def, shift_left_imm_to_uint): Likewise.
4721 (store_scatter_index_restricted_def,
4722 store_scatter_index_restricted): Likewise.
4723 (store_scatter_offset_restricted_def,
4724 store_scatter_offset_restricted): Likewise.
4725 (tbl_tuple_def, tbl_tuple): Likewise.
4726 (ternary_long_lane_def, ternary_long_lane): Likewise.
4727 (ternary_long_opt_n_def, ternary_long_opt_n): Likewise.
4728 (ternary_qq_lane_def): Inherit from ternary_resize2_lane_base.
4729 (ternary_qq_lane_rotate_def, ternary_qq_lane_rotate): New shape
4730 (ternary_qq_opt_n_def): Inherit from ternary_resize2_opt_n_base.
4731 (ternary_qq_rotate_def, ternary_qq_rotate): New shape.
4732 (ternary_shift_left_imm_def, ternary_shift_left_imm): Likewise.
4733 (ternary_shift_right_imm_def, ternary_shift_right_imm): Likewise.
4734 (ternary_uint_def, ternary_uint): Likewise.
4735 (unary_convert): Fix typo in comment.
4736 (unary_convert_narrowt_def, unary_convert_narrowt): New shape.
4737 (unary_long_def, unary_long): Likewise.
4738 (unary_narrowb_def, unary_narrowb): Likewise.
4739 (unary_narrowt_def, unary_narrowt): Likewise.
4740 (unary_narrowb_to_uint_def, unary_narrowb_to_uint): Likewise.
4741 (unary_narrowt_to_uint_def, unary_narrowt_to_uint): Likewise.
4742 (unary_to_int_def, unary_to_int): Likewise.
4743 * config/aarch64/aarch64-sve-builtins-base.cc (unspec_cmla)
4744 (unspec_fcmla, unspec_cond_fcmla, expand_mla_mls_lane): New functions.
4745 (svasrd_impl): Delete.
4746 (svcadd_impl::expand): Handle integer operations too.
4747 (svcmla_impl::expand, svcmla_lane::expand): Likewise, using the
4748 new functions to derive the unspec numbers.
4749 (svmla_svmls_lane_impl): Replace with...
4750 (svmla_lane_impl, svmls_lane_impl): ...these new classes. Handle
4751 integer operations too.
4752 (svwhile_impl): Rename to...
4753 (svwhilelx_impl): ...this and inherit from while_comparison.
4754 (svasrd): Use unspec_based_function.
4755 (svmla_lane): Use svmla_lane_impl.
4756 (svmls_lane): Use svmls_lane_impl.
4757 (svrecpe, svrsqrte): Handle unsigned integer operations too.
4758 (svwhilele, svwhilelt): Use svwhilelx_impl.
4759 * config/aarch64/aarch64-sve-builtins-sve2.h: New file.
4760 * config/aarch64/aarch64-sve-builtins-sve2.cc: Likewise.
4761 * config/aarch64/aarch64-sve-builtins-sve2.def: Likewise.
4762 * config/aarch64/aarch64-sve-builtins.def: Include
4763 aarch64-sve-builtins-sve2.def.
4765 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4767 * config/aarch64/aarch64-protos.h (aarch64_sve_arith_immediate_p)
4768 (aarch64_sve_sqadd_sqsub_immediate_p): Add a machine_mode argument.
4769 * config/aarch64/aarch64.c (aarch64_sve_arith_immediate_p)
4770 (aarch64_sve_sqadd_sqsub_immediate_p): Likewise. Handle scalar
4771 immediates as well as vector ones.
4772 * config/aarch64/predicates.md (aarch64_sve_arith_immediate)
4773 (aarch64_sve_sub_arith_immediate, aarch64_sve_qadd_immediate)
4774 (aarch64_sve_qsub_immediate): Update calls accordingly.
4776 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4778 * config/aarch64/aarch64-sve2.md: Add banner comments.
4779 (<su>mulh<r>s<mode>3): Move further up file.
4780 (<su>mull<bt><Vwide>, <r>shrnb<mode>, <r>shrnt<mode>)
4781 (*aarch64_sve2_sra<mode>): Move further down file.
4782 * config/aarch64/t-aarch64 (s-check-sve-md): Check aarch64-sve2.md too.
4784 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4786 * config/aarch64/iterators.md (SVE_WHILE): Add UNSPEC_WHILERW
4788 (while_optab_cmp): Handle them.
4789 * config/aarch64/aarch64-sve.md
4790 (*while_<while_optab_cmp><GPI:mode><PRED_ALL:mode>_ptest): Make public
4791 and add a "@" marker.
4792 * config/aarch64/aarch64-sve2.md (check_<raw_war>_ptrs<mode>): Use it
4793 instead of gen_aarch64_sve2_while_ptest.
4794 (@aarch64_sve2_while<cmp_op><GPI:mode><PRED_ALL:mode>_ptest): Delete.
4796 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4798 * config/aarch64/aarch64.md (UNSPEC_WHILE_LE): Rename to...
4799 (UNSPEC_WHILELE): ...this.
4800 (UNSPEC_WHILE_LO): Rename to...
4801 (UNSPEC_WHILELO): ...this.
4802 (UNSPEC_WHILE_LS): Rename to...
4803 (UNSPEC_WHILELS): ...this.
4804 (UNSPEC_WHILE_LT): Rename to...
4805 (UNSPEC_WHILELT): ...this.
4806 * config/aarch64/iterators.md (SVE_WHILE): Update accordingly.
4807 (cmp_op, while_optab_cmp): Likewise.
4808 * config/aarch64/aarch64.c (aarch64_sve_move_pred_via_while): Likewise.
4809 * config/aarch64/aarch64-sve-builtins-base.cc (svwhilele): Likewise.
4810 (svwhilelt): Likewise.
4812 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4814 * config/aarch64/aarch64-sve-builtins-shapes.h (unary_count): Delete.
4815 (unary_to_uint): Define.
4816 * config/aarch64/aarch64-sve-builtins-shapes.cc (unary_count_def)
4817 (unary_count): Rename to...
4818 (unary_to_uint_def, unary_to_uint): ...this.
4819 * config/aarch64/aarch64-sve-builtins-base.def: Update accordingly.
4821 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4823 * config/aarch64/aarch64-sve-builtins-functions.h
4824 (code_for_mode_function): New class.
4825 (CODE_FOR_MODE0, QUIET_CODE_FOR_MODE0): New macros.
4826 * config/aarch64/aarch64-sve-builtins-base.cc (svcompact_impl)
4827 (svext_impl, svmul_lane_impl, svsplice_impl, svtmad_impl): Delete.
4828 (svcompact, svext, svsplice): Use QUIET_CODE_FOR_MODE0.
4829 (svmul_lane, svtmad): Use CODE_FOR_MODE0.
4831 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4833 * config/aarch64/iterators.md (addsub): New code attribute.
4834 * config/aarch64/aarch64-simd.md (aarch64_<su_optab><optab><mode>):
4836 (aarch64_<su_optab>q<addsub><mode>): ...this, making the same change
4837 in the asm string and attributes. Fix indentation.
4838 * config/aarch64/aarch64-sve.md (@aarch64_<su_optab><optab><mode>):
4840 (@aarch64_sve_<optab><mode>): ...this.
4841 * config/aarch64/aarch64-sve-builtins.h
4842 (function_expander::expand_signed_unpred_op): Delete.
4843 * config/aarch64/aarch64-sve-builtins.cc
4844 (function_expander::expand_signed_unpred_op): Likewise.
4845 (function_expander::map_to_rtx_codes): If the optab isn't defined,
4846 try using code_for_aarch64_sve instead.
4847 * config/aarch64/aarch64-sve-builtins-base.cc (svqadd_impl): Delete.
4848 (svqsub_impl): Likewise.
4849 (svqadd, svqsub): Use rtx_code_function instead.
4851 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4853 * config/aarch64/iterators.md (SRHSUB, URHSUB): Delete.
4854 (HADDSUB, sur, addsub): Remove them.
4856 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4858 * tree-nrv.c (pass_return_slot::execute): Handle all internal
4859 functions the same way, rather than singling out those that
4860 aren't mapped directly to optabs.
4862 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4864 * target.def (compatible_vector_types_p): New target hook.
4865 * hooks.h (hook_bool_const_tree_const_tree_true): Declare.
4866 * hooks.c (hook_bool_const_tree_const_tree_true): New function.
4867 * doc/tm.texi.in (TARGET_COMPATIBLE_VECTOR_TYPES_P): New hook.
4868 * doc/tm.texi: Regenerate.
4869 * gimple-expr.c: Include target.h.
4870 (useless_type_conversion_p): Use targetm.compatible_vector_types_p.
4871 * config/aarch64/aarch64.c (aarch64_compatible_vector_types_p): New
4873 (TARGET_COMPATIBLE_VECTOR_TYPES_P): Define.
4874 * config/aarch64/aarch64-sve-builtins.cc (gimple_folder::convert_pred):
4875 Use the original predicate if it already has a suitable type.
4877 2020-01-09 Martin Jambor <mjambor@suse.cz>
4879 * cgraph.h (cgraph_edge): Make remove, set_call_stmt, make_direct,
4880 resolve_speculation and redirect_call_stmt_to_callee static. Change
4881 return type of set_call_stmt to cgraph_edge *.
4882 * auto-profile.c (afdo_indirect_call): Adjust call to
4883 redirect_call_stmt_to_callee.
4884 * cgraph.c (cgraph_edge::set_call_stmt): Make return cgraph-edge *,
4885 make the this pointer explicit, adjust self-recursive calls and the
4886 call top make_direct. Return the resulting edge.
4887 (cgraph_edge::remove): Make this pointer explicit.
4888 (cgraph_edge::resolve_speculation): Likewise, adjust call to remove.
4889 (cgraph_edge::make_direct): Likewise, adjust call to
4890 resolve_speculation.
4891 (cgraph_edge::redirect_call_stmt_to_callee): Likewise, also adjust
4892 call to set_call_stmt.
4893 (cgraph_update_edges_for_call_stmt_node): Update call to
4894 set_call_stmt and remove.
4895 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
4896 Renamed edge to master_edge. Adjusted calls to set_call_stmt.
4897 (cgraph_node::create_edge_including_clones): Moved "first" definition
4898 of edge to the block where it was used. Adjusted calls to
4900 (cgraph_node::remove_symbol_and_inline_clones): Adjust call to
4901 cgraph_edge::remove.
4902 * cgraphunit.c (walk_polymorphic_call_targets): Adjusted calls to
4903 make_direct and redirect_call_stmt_to_callee.
4904 * ipa-fnsummary.c (redirect_to_unreachable): Adjust calls to
4905 resolve_speculation and make_direct.
4906 * ipa-inline-transform.c (inline_transform): Adjust call to
4907 redirect_call_stmt_to_callee.
4908 (check_speculations_1):: Adjust call to resolve_speculation.
4909 * ipa-inline.c (resolve_noninline_speculation): Adjust call to
4910 resolve-speculation.
4911 (inline_small_functions): Adjust call to resolve_speculation.
4912 (ipa_inline): Likewise.
4913 * ipa-prop.c (ipa_make_edge_direct_to_target): Adjust call to
4915 * ipa-visibility.c (function_and_variable_visibility): Make iteration
4916 safe with regards to edge removal, adjust calls to
4917 redirect_call_stmt_to_callee.
4918 * ipa.c (walk_polymorphic_call_targets): Adjust calls to make_direct
4919 and redirect_call_stmt_to_callee.
4920 * multiple_target.c (create_dispatcher_calls): Adjust call to
4921 redirect_call_stmt_to_callee
4922 (redirect_to_specific_clone): Likewise.
4923 * tree-cfgcleanup.c (delete_unreachable_blocks_update_callgraph):
4924 Adjust calls to cgraph_edge::remove.
4925 * tree-inline.c (copy_bb): Adjust call to set_call_stmt.
4926 (redirect_all_calls): Adjust call to redirect_call_stmt_to_callee.
4927 (expand_call_inline): Adjust call to cgraph_edge::remove.
4929 2020-01-09 Martin Liska <mliska@suse.cz>
4931 * params.opt: Set Optimization for
4932 param_max_speculative_devirt_maydefs.
4934 2020-01-09 Martin Sebor <msebor@redhat.com>
4938 * builtins.c (compute_objsize): Avoid handling MEM_REFs of vector type.
4940 2020-01-09 Martin Liska <mliska@suse.cz>
4942 * auto-profile.c (auto_profile): Use opt_for_fn
4944 * ipa-cp.c (ipcp_lattice::add_value): Likewise.
4945 (propagate_vals_across_arith_jfunc): Likewise.
4946 (hint_time_bonus): Likewise.
4947 (incorporate_penalties): Likewise.
4948 (good_cloning_opportunity_p): Likewise.
4949 (perform_estimation_of_a_value): Likewise.
4950 (estimate_local_effects): Likewise.
4951 (ipcp_propagate_stage): Likewise.
4952 * ipa-fnsummary.c (decompose_param_expr): Likewise.
4953 (set_switch_stmt_execution_predicate): Likewise.
4954 (analyze_function_body): Likewise.
4955 * ipa-inline-analysis.c (offline_size): Likewise.
4956 * ipa-inline.c (early_inliner): Likewise.
4957 * ipa-prop.c (ipa_analyze_node): Likewise.
4958 (ipcp_transform_function): Likewise.
4959 * ipa-sra.c (process_scan_results): Likewise.
4960 (ipa_sra_summarize_function): Likewise.
4961 * params.opt: Rename ipcp-unit-growth to
4962 ipa-cp-unit-growth. Add Optimization for various
4963 IPA-related parameters.
4965 2020-01-09 Richard Biener <rguenther@suse.de>
4968 * gimplify.c (gimplify_expr): Deal with NOP definitions.
4970 2020-01-09 Richard Biener <rguenther@suse.de>
4972 PR tree-optimization/93040
4973 * gimple-ssa-store-merging.c (find_bswap_or_nop): Raise search limit.
4975 2020-01-09 Georg-Johann Lay <avr@gjlay.de>
4977 * common/config/avr/avr-common.c (avr_option_optimization_table)
4978 [OPT_LEVELS_1_PLUS]: Set -fsplit-wide-types-early.
4980 2020-01-09 Martin Liska <mliska@suse.cz>
4982 * cgraphclones.c (symbol_table::materialize_all_clones):
4983 Use cgraph_node::dump_name.
4985 2020-01-09 Jakub Jelinek <jakub@redhat.com>
4988 * config/riscv/riscv.c (riscv_print_operand_reloc): Use
4989 output_operand_lossage instead of gcc_unreachable.
4990 * doc/md.texi (riscv f constraint): Fix typo.
4993 * config/i386/i386.md (subv<mode>4): Use SWIDWI iterator instead of
4994 SWI. Use <general_hilo_operand> instead of <general_operand>. Use
4995 CONST_SCALAR_INT_P instead of CONST_INT_P.
4996 (*subv<mode>4_1): Rename to ...
4997 (subv<mode>4_1): ... this.
4998 (*subv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
4999 define_insn_and_split patterns.
5000 (*subv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
5003 2020-01-08 David Malcolm <dmalcolm@redhat.com>
5005 * vec.c (class selftest::count_dtor): New class.
5006 (selftest::test_auto_delete_vec): New test.
5007 (selftest::vec_c_tests): Call it.
5008 * vec.h (class auto_delete_vec): New class template.
5009 (auto_delete_vec<T>::~auto_delete_vec): New dtor.
5011 2020-01-08 David Malcolm <dmalcolm@redhat.com>
5013 * sbitmap.h (auto_sbitmap): Add operator const_sbitmap.
5015 2020-01-08 Jim Wilson <jimw@sifive.com>
5017 * config/riscv/riscv.c (riscv_legitimize_tls_address): Ifdef out
5018 use of TLS_MODEL_LOCAL_EXEC when not pic.
5020 2020-01-08 David Malcolm <dmalcolm@redhat.com>
5022 * hash-map-tests.c (selftest::test_map_of_strings_to_int): Fix
5025 2020-01-08 Jakub Jelinek <jakub@redhat.com>
5028 * config/i386/i386.md (*stack_protect_set_2_<mode> peephole2,
5029 *stack_protect_set_3 peephole2): Also check that the second
5030 insns source is general_operand.
5033 * config/i386/i386.md (addcarry<mode>_0): Use nonimmediate_operand
5034 predicate for output operand instead of register_operand.
5035 (addcarry<mode>, addcarry<mode>_1): Likewise. Add alternative with
5036 memory destination and non-memory operands[2].
5038 2020-01-08 Martin Liska <mliska@suse.cz>
5040 * cgraph.c (cgraph_node::dump): Use ::dump_name or
5041 ::dump_asm_name instead of (::name or ::asm_name).
5042 * cgraphclones.c (symbol_table::materialize_all_clones): Likewise.
5043 * cgraphunit.c (walk_polymorphic_call_targets): Likewise.
5044 (analyze_functions): Likewise.
5045 (expand_all_functions): Likewise.
5046 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
5047 (propagate_bits_across_jump_function): Likewise.
5048 (dump_profile_updates): Likewise.
5049 (ipcp_store_bits_results): Likewise.
5050 (ipcp_store_vr_results): Likewise.
5051 * ipa-devirt.c (dump_targets): Likewise.
5052 * ipa-fnsummary.c (analyze_function_body): Likewise.
5053 * ipa-hsa.c (check_warn_node_versionable): Likewise.
5054 (process_hsa_functions): Likewise.
5055 * ipa-icf.c (sem_item_optimizer::merge_classes): Likewise.
5056 (set_alias_uids): Likewise.
5057 * ipa-inline-transform.c (save_inline_function_body): Likewise.
5058 * ipa-inline.c (recursive_inlining): Likewise.
5059 (inline_to_all_callers_1): Likewise.
5060 (ipa_inline): Likewise.
5061 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
5062 (ipa_propagate_frequency): Likewise.
5063 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
5064 (remove_described_reference): Likewise.
5065 * ipa-pure-const.c (worse_state): Likewise.
5066 (check_retval_uses): Likewise.
5067 (analyze_function): Likewise.
5068 (propagate_pure_const): Likewise.
5069 (propagate_nothrow): Likewise.
5070 (dump_malloc_lattice): Likewise.
5071 (propagate_malloc): Likewise.
5072 (pass_local_pure_const::execute): Likewise.
5073 * ipa-visibility.c (optimize_weakref): Likewise.
5074 (function_and_variable_visibility): Likewise.
5075 * ipa.c (symbol_table::remove_unreachable_nodes): Likewise.
5076 (ipa_discover_variable_flags): Likewise.
5077 * lto-streamer-out.c (output_function): Likewise.
5078 (output_constructor): Likewise.
5079 * tree-inline.c (copy_bb): Likewise.
5080 * tree-ssa-structalias.c (ipa_pta_execute): Likewise.
5081 * varpool.c (symbol_table::remove_unreferenced_decls): Likewise.
5083 2020-01-08 Richard Biener <rguenther@suse.de>
5086 * tree-eh.c (sink_clobbers): Update virtual operands for
5087 the first and last stmt only. Add a dry-run capability.
5088 (pass_lower_eh_dispatch::execute): Perform clobber sinking
5089 after CFG manipulations and in RPO order to catch all
5090 secondary opportunities reliably.
5092 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
5095 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
5097 2019-01-08 Richard Biener <rguenther@suse.de>
5100 * gimple-fold.c (rewrite_to_defined_overflow): Mark stmt modified.
5101 * tree-ssa-loop-im.c (move_computations_worker): Properly adjust
5102 virtual operand, also updating SSA use.
5103 * gimple-loop-interchange.cc (loop_cand::undo_simple_reduction):
5104 Update stmt after resetting virtual operand.
5105 (tree_loop_interchange::move_code_to_inner_loop): Likewise.
5106 * gimple-iterator.c (gsi_remove): When not removing the stmt
5107 permanently do not delink immediate uses or mark the stmt modified.
5109 2020-01-08 Martin Liska <mliska@suse.cz>
5111 * ipa-fnsummary.c (dump_ipa_call_summary): Use symtab_node::dump_name.
5112 (ipa_call_context::estimate_size_and_time): Likewise.
5113 (inline_analyze_function): Likewise.
5115 2020-01-08 Martin Liska <mliska@suse.cz>
5117 * cgraph.c (cgraph_node::dump): Use systematically
5120 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
5122 Add -nodevicespecs option for avr.
5125 * config/avr/avr.opt (-nodevicespecs): New driver option.
5126 * config/avr/driver-avr.c (avr_devicespecs_file): Only issue
5127 "-specs=device-specs/..." if that option is not set.
5128 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
5130 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
5132 Implement 64-bit double functions for avr.
5135 * config.gcc (tm_defines) [target=avr]: Support --with-libf7,
5136 --with-double-comparison.
5137 * doc/install.texi: Document them.
5138 * config/avr/avr-c.c (avr_cpu_cpp_builtins)
5139 <WITH_LIBF7_LIBGCC, WITH_LIBF7_MATH, WITH_LIBF7_MATH_SYMBOLS>
5140 <WITH_DOUBLE_COMPARISON>: New built-in defines.
5141 * doc/invoke.texi (AVR Built-in Macros): Document them.
5142 * config/avr/avr-protos.h (avr_float_lib_compare_returns_bool): New.
5143 * config/avr/avr.c (avr_float_lib_compare_returns_bool): New function.
5144 * config/avr/avr.h (FLOAT_LIB_COMPARE_RETURNS_BOOL): New macro.
5146 2020-01-08 Richard Earnshaw <rearnsha@arm.com>
5149 * config/arm/t-multilib (MULTILIB_MATCHES): Add rules to match
5150 armv7-a{+mp,+sec,+mp+sec} to appropriate armv7 multilib variants
5151 when only building rm-profile multilibs.
5153 2020-01-08 Feng Xue <fxue@os.amperecomputing.com>
5156 * ipa-cp.c (self_recursively_generated_p): Find matched aggregate
5157 lattice for a value to check.
5158 (propagate_vals_across_arith_jfunc): Add an assertion to ensure
5159 finite propagation in self-recursive scc.
5161 2020-01-08 Luo Xiong Hu <luoxhu@linux.ibm.com>
5163 * ipa-inline.c (caller_growth_limits): Restore the AND.
5165 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
5167 * config/gcn/gcn-valu.md (VEC_1REG_INT_ALT): Delete iterator.
5168 (VEC_ALLREG_ALT): New iterator.
5169 (VEC_ALLREG_INT_MODE): New iterator.
5170 (VCMP_MODE): New iterator.
5171 (VCMP_MODE_INT): New iterator.
5172 (vec_cmpu<mode>di): Use VCMP_MODE_INT.
5173 (vec_cmp<u>v64qidi): New define_expand.
5174 (vec_cmp<mode>di_exec): Use VCMP_MODE.
5175 (vec_cmpu<mode>di_exec): New define_expand.
5176 (vec_cmp<u>v64qidi_exec): New define_expand.
5177 (vec_cmp<mode>di_dup): Use VCMP_MODE.
5178 (vec_cmp<mode>di_dup_exec): Use VCMP_MODE.
5179 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>): Rename ...
5180 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): ... to this.
5181 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>_exec): Rename ...
5182 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): ... to this.
5183 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>): Rename ...
5184 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): ... to this.
5185 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>_exec): Rename ...
5186 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): ... to
5188 * config/gcn/gcn.c (print_operand): Fix 8 and 16 bit suffixes.
5189 * config/gcn/gcn.md (expander): Add sign_extend and zero_extend.
5191 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
5193 * config/gcn/constraints.md (DA): Update description and match.
5195 (Db): New constraint.
5196 * config/gcn/gcn-protos.h (gcn_inline_constant64_p): Add second
5198 * config/gcn/gcn.c (gcn_inline_constant64_p): Add 'mixed' parameter.
5199 Implement 'Db' mixed immediate type.
5200 * config/gcn/gcn-valu.md (addcv64si3<exec_vcc>): Rework constraints.
5201 (addcv64si3_dup<exec_vcc>): Delete.
5202 (subcv64si3<exec_vcc>): Rework constraints.
5203 (addv64di3): Rework constraints.
5204 (addv64di3_exec): Rework constraints.
5205 (subv64di3): Rework constraints.
5206 (addv64di3_dup): Delete.
5207 (addv64di3_dup_exec): Delete.
5208 (addv64di3_zext): Rework constraints.
5209 (addv64di3_zext_exec): Rework constraints.
5210 (addv64di3_zext_dup): Rework constraints.
5211 (addv64di3_zext_dup_exec): Rework constraints.
5212 (addv64di3_zext_dup2): Rework constraints.
5213 (addv64di3_zext_dup2_exec): Rework constraints.
5214 (addv64di3_sext_dup2): Rework constraints.
5215 (addv64di3_sext_dup2_exec): Rework constraints.
5217 2020-01-07 Andre Vieira <andre.simoesdiasvieira@arm.com>
5219 * doc/sourcebuild.texi (arm_little_endian, arm_nothumb): Documented
5220 existing target checks.
5222 2020-01-07 Richard Biener <rguenther@suse.de>
5224 * doc/install.texi: Bump minimal supported MPC version.
5226 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
5228 * langhooks-def.h (lhd_simulate_enum_decl): Declare.
5229 (LANG_HOOKS_SIMULATE_ENUM_DECL): Use it.
5230 * langhooks.c: Include stor-layout.h.
5231 (lhd_simulate_enum_decl): New function.
5232 * config/aarch64/aarch64-sve-builtins.cc (init_builtins): Call
5233 handle_arm_sve_h for the LTO frontend.
5234 (register_vector_type): Cope with null returns from pushdecl.
5236 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
5238 * config/aarch64/aarch64-protos.h (aarch64_sve::svbool_type_p)
5239 (aarch64_sve::nvectors_if_data_type): Replace with...
5240 (aarch64_sve::builtin_type_p): ...this.
5241 * config/aarch64/aarch64-sve-builtins.cc: Include attribs.h.
5242 (find_vector_type): Delete.
5243 (add_sve_type_attribute): New function.
5244 (lookup_sve_type_attribute): Likewise.
5245 (register_builtin_types): Add an "SVE type" attribute to each type.
5246 (register_tuple_type): Likewise.
5247 (svbool_type_p, nvectors_if_data_type): Delete.
5248 (mangle_builtin_type): Use lookup_sve_type_attribute.
5249 (builtin_type_p): Likewise. Add an overload that returns the
5250 number of constituent vector and predicate registers.
5251 * config/aarch64/aarch64.c (aarch64_sve_argument_p): Delete.
5252 (aarch64_returns_value_in_sve_regs_p): Use aarch64_sve::builtin_type_p
5253 instead of aarch64_sve_argument_p.
5254 (aarch64_takes_arguments_in_sve_regs_p): Likewise.
5255 (aarch64_pass_by_reference): Likewise.
5256 (aarch64_function_value_1): Likewise.
5257 (aarch64_return_in_memory): Likewise.
5258 (aarch64_layout_arg): Likewise.
5260 2020-01-07 Jakub Jelinek <jakub@redhat.com>
5262 PR tree-optimization/93156
5263 * tree-ssa-ccp.c (bit_value_binop): For x * x note that the second
5264 least significant bit is always clear.
5266 PR tree-optimization/93118
5267 * match.pd ((x >> c) << c -> x & (-1<<c)): Add nop_convert?. Add new
5268 simplifier with two intermediate conversions.
5270 2020-01-07 Martin Liska <mliska@suse.cz>
5272 * params.opt: Add Optimization for various parameters.
5274 2020-01-07 Martin Liska <mliska@suse.cz>
5277 * doc/extend.texi: Explain cloning for target_clone
5280 2020-01-07 Martin Liska <mliska@suse.cz>
5282 PR tree-optimization/92860
5283 * common.opt: Make in Optimization option
5284 as it is affected by -O0, which is an Optimization
5286 * tree-inline.c (tree_inlinable_function_p):
5287 Use opt_for_fn for warn_inline.
5288 (expand_call_inline): Likewise.
5290 2020-01-07 Martin Liska <mliska@suse.cz>
5292 PR tree-optimization/92860
5293 * common.opt: Make flag_ree as optimization
5296 2020-01-07 Martin Liska <mliska@suse.cz>
5298 PR optimization/92860
5299 * params.opt: Mark param_min_crossjump_insns with Optimization
5302 2020-01-07 Luo Xiong Hu <luoxhu@linux.ibm.com>
5304 * ipa-inline-analysis.c (estimate_growth): Fix typo.
5305 * ipa-inline.c (caller_growth_limits): Use OR instead of AND.
5307 2020-01-06 Michael Meissner <meissner@linux.ibm.com>
5309 * config/rs6000/rs6000.c (hard_reg_and_mode_to_addr_mask): New
5310 helper function to return the valid addressing formats for a given
5311 hard register and mode.
5312 (rs6000_adjust_vec_address): Call hard_reg_and_mode_to_addr_mask.
5314 * config/rs6000/constraints.md (Q constraint): Update
5316 * doc/md.texi (RS/6000 constraints): Update 'Q' cosntraint
5319 * config/rs6000/vsx.md (vsx_extract_<mode>_var, VSX_D iterator):
5320 Use 'Q' for doing vector extract from memory.
5321 (vsx_extract_v4sf_var): Use 'Q' for doing vector extract from
5323 (vsx_extract_<mode>_var, VSX_EXTRACT_I iterator): Use 'Q' for
5324 doing vector extract from memory.
5325 (vsx_extract_<mode>_<VS_scalar>mode_var): Use 'Q' for doing vector
5326 extract from memory.
5328 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add support
5329 for the offset being 34-bits when -mcpu=future is used.
5331 2020-01-06 John David Anglin <danglin@gcc.gnu.org>
5333 * config/pa/pa.md: Revert change to use ordered_comparison_operator
5334 instead of cmpib_comparison_operator in cmpib patterns.
5335 * config/pa/predicates.md (cmpib_comparison_operator): Revert removal
5336 of cmpib_comparison_operator. Revise comment.
5338 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
5340 * tree-vect-slp.c (vect_build_slp_tree_1): Require all shifts
5341 in an IFN_DIV_POW2 node to be equal.
5343 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
5345 * tree-vect-stmts.c (vect_check_load_store_mask): Rename to...
5346 (vect_check_scalar_mask): ...this.
5347 (vectorizable_store, vectorizable_load): Update call accordingly.
5348 (vectorizable_call): Use vect_check_scalar_mask to check the mask
5349 argument in calls to conditional internal functions.
5351 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
5353 * config/gcn/gcn-valu.md (subv64di3): Use separate alternatives for
5354 '0' matching inputs.
5355 (subv64di3_exec): Likewise.
5357 2020-01-06 Bryan Stenson <bryan@siliconvortex.com>
5359 * config/mips/mips.c (vr4130_align_insns): Fix typo.
5360 * doc/md.texi (movstr): Likewise.
5362 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
5364 * config/gcn/gcn-valu.md (vec_extract<mode><scalar_mode>): Add early
5367 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
5369 * config/aarch64/t-aarch64 ($(srcdir)/config/aarch64/aarch64-tune.md):
5371 (s-aarch64-tune-md): ...this new stamp file. Pipe the new contents
5372 to a temporary file and use move-if-change to update the real
5373 file where necessary.
5375 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
5377 * config/aarch64/aarch64-sve.md (@aarch64_sel_dup<mode>): Use Upl
5378 rather than Upa for CPY /M.
5380 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
5382 * config/gcn/gcn.c (gcn_inline_constant_p): Allow 64 as an inline
5385 2020-01-06 Martin Liska <mliska@suse.cz>
5387 PR tree-optimization/92860
5388 * params.opt: Mark param_max_combine_insns with Optimization
5391 2020-01-05 Jakub Jelinek <jakub@redhat.com>
5394 * config/i386/i386.md (SWIDWI): New mode iterator.
5395 (DWI, dwi): Add TImode variants.
5396 (addv<mode>4): Use SWIDWI iterator instead of SWI. Use
5397 <general_hilo_operand> instead of <general_operand>. Use
5398 CONST_SCALAR_INT_P instead of CONST_INT_P.
5399 (*addv<mode>4_1): Rename to ...
5400 (addv<mode>4_1): ... this.
5401 (QWI): New mode attribute.
5402 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
5403 define_insn_and_split patterns.
5404 (*addv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
5406 (uaddv<mode>4): Use SWIDWI iterator instead of SWI. Use
5407 <general_hilo_operand> instead of <general_operand>.
5408 (*addcarry<mode>_1): New define_insn.
5409 (*add<dwi>3_doubleword_cc_overflow_1): New define_insn_and_split.
5411 2020-01-03 Konstantin Kharlamov <Hi-Angel@yandex.ru>
5413 * gdbinit.in (pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, pdd, pbs, pbm):
5414 Use "call" instead of "set".
5416 2020-01-03 Martin Jambor <mjambor@suse.cz>
5419 * ipa-cp.c (print_all_lattices): Skip functions without info.
5421 2020-01-03 Jakub Jelinek <jakub@redhat.com>
5424 * config/i386/i386-options.c (ix86_simd_clone_adjust): If
5425 TARGET_PREFER_AVX128, use prefer-vector-width=256 for 'c' and 'd'
5426 simd clones. If TARGET_PREFER_AVX256, use prefer-vector-width=512
5427 for 'e' simd clones.
5430 * config/i386/i386.opt (x_prefer_vector_width_type): Remove TargetSave
5432 (mprefer-vector-width=): Add Save.
5433 * config/i386/i386-options.c (ix86_target_string): Add PVW argument, print
5434 -mprefer-vector-width= if non-zero. Fix up -mfpmath= comment.
5435 (ix86_debug_options, ix86_function_specific_print): Adjust
5436 ix86_target_string callers.
5437 (ix86_valid_target_attribute_inner_p): Handle prefer-vector-width=.
5438 (ix86_valid_target_attribute_tree): Likewise.
5439 * config/i386/i386-options.h (ix86_target_string): Add PVW argument.
5440 * config/i386/i386-expand.c (ix86_expand_builtin): Adjust
5441 ix86_target_string caller.
5444 * config/i386/i386.md (abs<mode>2): Use expand_simple_binop instead of
5445 emitting ASHIFTRT, XOR and MINUS by hand. Use gen_int_mode with QImode
5446 instead of gen_int_shift_amount + convert_modes.
5448 PR rtl-optimization/93088
5449 * loop-iv.c (find_single_def_src): Punt after looking through
5450 128 reg copies for regs with single definitions. Move definitions
5453 2020-01-02 Dennis Zhang <dennis.zhang@arm.com>
5455 * config/arm/arm-c.c (arm_cpu_builtins): Define
5456 __ARM_FEATURE_MATMUL_INT8, __ARM_FEATURE_BF16_VECTOR_ARITHMETIC,
5457 __ARM_FEATURE_BF16_SCALAR_ARITHMETIC, and
5458 __ARM_BF16_FORMAT_ALTERNATIVE when enabled.
5459 * config/arm/arm-cpus.in (armv8_6, i8mm, bf16): New features.
5460 * config/arm/arm-tables.opt: Regenerated.
5461 * config/arm/arm.c (arm_option_reconfigure_globals): Initialize
5462 arm_arch_i8mm and arm_arch_bf16 when enabled.
5463 * config/arm/arm.h (TARGET_I8MM): New macro.
5464 (TARGET_BF16_FP, TARGET_BF16_SIMD): Likewise.
5465 * config/arm/t-aprofile: Add matching rules for -march=armv8.6-a.
5466 * config/arm/t-arm-elf (all_v8_archs): Add armv8.6-a.
5467 * config/arm/t-multilib: Add matching rules for -march=armv8.6-a.
5468 (v8_6_a_simd_variants): New.
5469 (v8_*_a_simd_variants): Add i8mm and bf16.
5470 * doc/invoke.texi (armv8.6-a, i8mm, bf16): Document new options.
5472 2020-01-02 Jakub Jelinek <jakub@redhat.com>
5475 * predict.c (compute_function_frequency): Don't call
5476 warn_function_cold on functions that already have cold attribute.
5478 2020-01-01 John David Anglin <danglin@gcc.gnu.org>
5481 * config/pa/pa.c (pa_elf_select_rtx_section): New. Put references to
5482 COMDAT group function labels in .data.rel.ro.local section.
5483 * config/pa/pa32-linux.h (TARGET_ASM_SELECT_RTX_SECTION): Define.
5486 * config/pa/pa.md (scc): Use ordered_comparison_operator instead of
5487 comparison_operator in B and S integer comparisons. Likewise, use
5488 ordered_comparison_operator instead of cmpib_comparison_operator in
5490 * config/pa/predicates.md (cmpib_comparison_operator): Remove.
5492 2020-01-01 Jakub Jelinek <jakub@redhat.com>
5494 Update copyright years.
5496 * gcc.c (process_command): Update copyright notice dates.
5497 * gcov-dump.c (print_version): Ditto.
5498 * gcov.c (print_version): Ditto.
5499 * gcov-tool.c (print_version): Ditto.
5500 * gengtype.c (create_file): Ditto.
5501 * doc/cpp.texi: Bump @copying's copyright year.
5502 * doc/cppinternals.texi: Ditto.
5503 * doc/gcc.texi: Ditto.
5504 * doc/gccint.texi: Ditto.
5505 * doc/gcov.texi: Ditto.
5506 * doc/install.texi: Ditto.
5507 * doc/invoke.texi: Ditto.
5509 2020-01-01 Jan Hubicka <hubicka@ucw.cz>
5511 * ipa.c (walk_polymorphic_call_targets): Fix updating of overall
5514 2020-01-01 Jakub Jelinek <jakub@redhat.com>
5516 PR tree-optimization/93098
5517 * match.pd (popcount): For shift amounts, use integer_onep
5518 or wi::to_widest () == cst instead of tree_to_uhwi () == cst
5519 tests. Make sure that precision is power of two larger than or equal
5520 to 16. Ensure shift is never negative. Use HOST_WIDE_INT_UC macro
5521 instead of ULL suffixed constants. Formatting fixes.
5523 Copyright (C) 2020 Free Software Foundation, Inc.
5525 Copying and distribution of this file, with or without modification,
5526 are permitted in any medium without royalty provided the copyright
5527 notice and this notice are preserved.