1 2020-01-31 Vladimir Makarov <vmakarov@redhat.com>
3 PR rtl-optimization/91333
4 * ira-color.c (bucket_allocno_compare_func): Move conflict hard
5 reg preferences comparison up.
7 2020-01-31 Richard Sandiford <richard.sandiford@arm.com>
9 * config/aarch64/aarch64.h (TARGET_SVE_BF16): New macro.
10 * config/aarch64/aarch64-sve-builtins-sve2.h (svcvtnt): Move to
11 aarch64-sve-builtins-base.h.
12 * config/aarch64/aarch64-sve-builtins-sve2.cc (svcvtnt): Move to
13 aarch64-sve-builtins-base.cc.
14 * config/aarch64/aarch64-sve-builtins-base.h (svbfdot, svbfdot_lane)
15 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
17 * config/aarch64/aarch64-sve-builtins-base.cc (svbfdot, svbfdot_lane)
18 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
19 (svcvtnt): New functions.
20 * config/aarch64/aarch64-sve-builtins-base.def (svbfdot, svbfdot_lane)
21 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
22 (svcvtnt): New functions.
23 (svcvt): Add a form that converts f32 to bf16.
24 * config/aarch64/aarch64-sve-builtins-shapes.h (ternary_bfloat)
25 (ternary_bfloat_lane, ternary_bfloat_lanex2, ternary_bfloat_opt_n):
27 * config/aarch64/aarch64-sve-builtins-shapes.cc (parse_element_type):
28 Treat B as bfloat16_t.
29 (ternary_bfloat_lane_base): New class.
30 (ternary_bfloat_def): Likewise.
31 (ternary_bfloat): New shape.
32 (ternary_bfloat_lane_def): New class.
33 (ternary_bfloat_lane): New shape.
34 (ternary_bfloat_lanex2_def): New class.
35 (ternary_bfloat_lanex2): New shape.
36 (ternary_bfloat_opt_n_def): New class.
37 (ternary_bfloat_opt_n): New shape.
38 * config/aarch64/aarch64-sve-builtins.cc (TYPES_cvt_bfloat): New macro.
39 * config/aarch64/aarch64-sve.md (@aarch64_sve_<sve_fp_op>vnx4sf)
40 (@aarch64_sve_<sve_fp_op>_lanevnx4sf): New patterns.
41 (@aarch64_sve_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>)
42 (@cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
43 (*cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
44 (@aarch64_sve_cvtnt<VNx8BF_ONLY:mode>): Likewise.
45 * config/aarch64/aarch64-sve2.md (@aarch64_sve2_cvtnt<mode>): Key
46 the pattern off the narrow mode instead of the wider one.
47 * config/aarch64/iterators.md (VNx8BF_ONLY): New mode iterator.
48 (UNSPEC_BFMLALB, UNSPEC_BFMLALT, UNSPEC_BFMMLA): New unspecs.
49 (sve_fp_op): Handle them.
50 (SVE_BFLOAT_TERNARY_LONG): New int itertor.
51 (SVE_BFLOAT_TERNARY_LONG_LANE): Likewise.
53 2020-01-31 Richard Sandiford <richard.sandiford@arm.com>
55 * config/aarch64/arm_sve.h: Include arm_bf16.h.
56 * config/aarch64/aarch64-modes.def (BF): Move definition before
57 VECTOR_MODES. Remove separate VECTOR_MODES for V4BF and V8BF.
58 (SVE_MODES): Handle BF modes.
59 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
61 (aarch64_full_sve_mode): Likewise.
62 * config/aarch64/iterators.md (SVE_STRUCT): Add VNx16BF, VNx24BF
64 (SVE_FULL, SVE_FULL_HSD, SVE_ALL): Add VNx8BF.
65 (Vetype, Vesize, Vctype, VEL, Vel, VEL_INT, V128, v128, vwcore)
66 (V_INT_EQUIV, v_int_equiv, V_FP_EQUIV, v_fp_equiv, vector_count)
67 (insn_length, VSINGLE, vsingle, VPRED, vpred, VDOUBLE): Handle the
69 * config/aarch64/aarch64-sve-builtins.h (TYPE_bfloat): New
71 * config/aarch64/aarch64-sve-builtins.cc (TYPES_all_arith): New macro.
72 (TYPES_all_data): Add bf16.
73 (TYPES_reinterpret1, TYPES_reinterpret): Likewise.
74 (register_tuple_type): Increase buffer size.
75 * config/aarch64/aarch64-sve-builtins.def (svbfloat16_t): New type.
76 (bf16): New type suffix.
77 * config/aarch64/aarch64-sve-builtins-base.def (svabd, svadd, svaddv)
78 (svcmpeq, svcmpge, svcmpgt, svcmple, svcmplt, svcmpne, svmad, svmax)
79 (svmaxv, svmin, svminv, svmla, svmls, svmsb, svmul, svsub, svsubr):
80 Change type from all_data to all_arith.
81 * config/aarch64/aarch64-sve-builtins-sve2.def (svaddp, svmaxp)
84 2020-01-31 Dennis Zhang <dennis.zhang@arm.com>
85 Matthew Malcomson <matthew.malcomson@arm.com>
86 Richard Sandiford <richard.sandiford@arm.com>
88 * doc/invoke.texi (f32mm): Document new AArch64 -march= extension.
89 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Define
90 __ARM_FEATURE_SVE_MATMUL_INT8, __ARM_FEATURE_SVE_MATMUL_FP32 and
91 __ARM_FEATURE_SVE_MATMUL_FP64 as appropriate. Don't define
92 __ARM_FEATURE_MATMUL_FP64.
93 * config/aarch64/aarch64-option-extensions.def (fp, simd, fp16)
94 (sve): Add AARCH64_FL_F32MM to the list of extensions that should
95 be disabled at the same time.
96 (f32mm): New extension.
97 * config/aarch64/aarch64.h (AARCH64_FL_F32MM): New macro.
98 (AARCH64_FL_F64MM): Bump to the next bit up.
99 (AARCH64_ISA_F32MM, TARGET_SVE_I8MM, TARGET_F32MM, TARGET_SVE_F32MM)
100 (TARGET_SVE_F64MM): New macros.
101 * config/aarch64/iterators.md (SVE_MATMULF): New mode iterator.
102 (UNSPEC_FMMLA, UNSPEC_SMATMUL, UNSPEC_UMATMUL, UNSPEC_USMATMUL)
103 (UNSPEC_TRN1Q, UNSPEC_TRN2Q, UNSPEC_UZP1Q, UNSPEC_UZP2Q, UNSPEC_ZIP1Q)
104 (UNSPEC_ZIP2Q): New unspeccs.
105 (DOTPROD_US_ONLY, PERMUTEQ, MATMUL, FMMLA): New int iterators.
106 (optab, sur, perm_insn): Handle the new unspecs.
107 (sve_fp_op): Handle UNSPEC_FMMLA. Resort.
108 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use
109 TARGET_SVE_F64MM instead of separate tests.
110 (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod<vsi2qi>): New pattern.
111 (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod_lane<vsi2qi>): Likewise.
112 (@aarch64_sve_add_<MATMUL:optab><vsi2qi>): Likewise.
113 (@aarch64_sve_<FMMLA:sve_fp_op><mode>): Likewise.
114 (@aarch64_sve_<PERMUTEQ:optab><mode>): Likewise.
115 * config/aarch64/aarch64-sve-builtins.cc (TYPES_s_float): New macro.
116 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): Use it.
117 (TYPES_s_signed): New macro.
118 (TYPES_s_integer): Use it.
119 (TYPES_d_float): New macro.
120 (TYPES_d_data): Use it.
121 * config/aarch64/aarch64-sve-builtins-shapes.h (mmla): Declare.
122 (ternary_intq_uintq_lane, ternary_intq_uintq_opt_n, ternary_uintq_intq)
123 (ternary_uintq_intq_lane, ternary_uintq_intq_opt_n): Likewise.
124 * config/aarch64/aarch64-sve-builtins-shapes.cc (mmla_def): New class.
126 (ternary_resize2_opt_n_base): Add TYPE_CLASS2 and TYPE_CLASS3
128 (ternary_resize2_lane_base): Likewise.
129 (ternary_resize2_base): New class.
130 (ternary_qq_lane_base): Likewise.
131 (ternary_intq_uintq_lane_def): Likewise.
132 (ternary_intq_uintq_lane): New shape.
133 (ternary_intq_uintq_opt_n_def): New class
134 (ternary_intq_uintq_opt_n): New shape.
135 (ternary_qq_lane_def): Inherit from ternary_qq_lane_base.
136 (ternary_uintq_intq_def): New class.
137 (ternary_uintq_intq): New shape.
138 (ternary_uintq_intq_lane_def): New class.
139 (ternary_uintq_intq_lane): New shape.
140 (ternary_uintq_intq_opt_n_def): New class.
141 (ternary_uintq_intq_opt_n): New shape.
142 * config/aarch64/aarch64-sve-builtins-base.h (svmmla, svsudot)
143 (svsudot_lane, svtrn1q, svtrn2q, svusdot, svusdot_lane, svusmmla)
144 (svuzp1q, svuzp2q, svzip1q, svzip2q): Declare.
145 * config/aarch64/aarch64-sve-builtins-base.cc (svdot_lane_impl):
147 (svdotprod_lane_impl): ...this new class.
148 (svmmla_impl, svusdot_impl): New classes.
149 (svdot_lane): Update to use svdotprod_lane_impl.
150 (svmmla, svsudot, svsudot_lane, svtrn1q, svtrn2q, svusdot)
151 (svusdot_lane, svusmmla, svuzp1q, svuzp2q, svzip1q, svzip2q): New
153 * config/aarch64/aarch64-sve-builtins-base.def (svmmla): New base
154 function, with no types defined.
155 (svmmla, svusmmla, svsudot, svsudot_lane, svusdot, svusdot_lane): New
156 AARCH64_FL_I8MM functions.
157 (svmmla): New AARCH64_FL_F32MM function.
158 (svld1ro): Depend only on AARCH64_FL_F64MM, not on AARCH64_FL_V8_6.
159 (svmmla, svtrn1q, svtrn2q, svuz1q, svuz2q, svzip1q, svzip2q): New
160 AARCH64_FL_F64MM function.
161 (REQUIRED_EXTENSIONS):
163 2020-01-31 Andrew Stubbs <ams@codesourcery.com>
165 * config/gcn/gcn-valu.md (addv64di3_exec): Allow one '0' in each
168 2020-01-31 Uroš Bizjak <ubizjak@gmail.com>
170 * config/i386/i386.md (*movoi_internal_avx): Do not check for
171 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL. Remove MODE_V8SF handling.
172 (*movti_internal): Do not check for
173 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
174 (*movtf_internal): Move check for TARGET_SSE2 and size optimization
175 just after check for TARGET_AVX.
176 (*movdf_internal): Ditto.
177 * config/i386/mmx.md (*mov<mode>_internal): Do not check for
178 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
179 * config/i386/sse.md (mov<mode>_internal): Only check
180 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL with V2DFmode. Move check
181 for TARGET_SSE2 and size optimization just after check for TARGET_AVX.
182 (<sse>_andnot<mode>3<mask_name>): Move check for
183 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL after check for TARGET_AVX.
184 (<code><mode>3<mask_name>): Ditto.
185 (*andnot<mode>3): Ditto.
187 (*<code><mode>3): Ditto.
189 (*andnot<VI:mode>3): Remove
190 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL handling.
191 (<mask_codefor><code><VI48_AVX_AVX512F:mode>3<mask_name>): Ditto.
192 (*<code><VI12_AVX_AVX512F:mode>3): Ditto.
193 (sse4_1_blendv<ssemodesuffix>): Ditto.
194 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL):
195 Explain that tune applies to 128bit instructions only.
197 2020-01-31 Kwok Cheung Yeung <kcy@codesourcery.com>
199 * config/gcn/mkoffload.c (process_asm): Add sgpr_count and vgpr_count
200 to definition of hsa_kernel_description. Parse assembly to find SGPR
201 and VGPR count of kernel and store in hsa_kernel_description.
203 2020-01-31 Tamar Christina <tamar.christina@arm.com>
205 PR rtl-optimization/91838
206 * simplify-rtx.c (simplify_binary_operation_1): Update LSHIFTRT case
207 to truncate if allowed or reject combination.
209 2020-01-31 Andrew Stubbs <ams@codesourcery.com>
211 * tree-ssa-loop-ivopts.c (get_iv): Use sizetype for zero-step.
212 (find_inv_vars_cb): Likewise.
214 2020-01-31 David Malcolm <dmalcolm@redhat.com>
216 * calls.c (special_function_p): Split out the check for DECL_NAME
217 being non-NULL and fndecl being extern at file scope into a
218 new maybe_special_function_p and call it. Drop check for fndecl
219 being non-NULL that was after a usage of DECL_NAME (fndecl).
220 * tree.h (maybe_special_function_p): New inline function.
222 2020-01-30 Andrew Stubbs <ams@codesourcery.com>
224 * config/gcn/gcn-valu.md (gather<mode>_exec): Move contents ...
225 (mask_gather_load<mode>): ... here, and zero-initialize the
227 (maskload<mode>di): Zero-initialize the destination.
230 2020-01-30 David Malcolm <dmalcolm@redhat.com>
233 * doc/analyzer.texi (Limitations): Note that constraints on
234 floating-point values are currently ignored.
236 2020-01-30 Jakub Jelinek <jakub@redhat.com>
239 * symtab.c (symtab_node::noninterposable_alias): If localalias
240 already exists, but is not usable, append numbers after it until
241 a unique name is found. Formatting fix.
244 * combine.c (simplify_comparison) <case ROTATE>: Punt on out of range
247 2020-01-30 Andrew Stubbs <ams@codesourcery.com>
249 * config/gcn/gcn.c (print_operand): Handle LTGT.
250 * config/gcn/predicates.md (gcn_fp_compare_operator): Allow ltgt.
252 2020-01-30 Richard Biener <rguenther@suse.de>
254 * tree-pretty-print.c (dump_generic_node): Wrap VECTOR_CST
255 and CONSTRUCTOR in _Literal (type) with TDF_GIMPLE.
257 2020-01-30 John David Anglin <danglin@gcc.gnu.org>
259 * config/pa/pa.c (pa_elf_select_rtx_section): Place function pointers
260 without a DECL in .data.rel.ro.local.
262 2020-01-30 Jakub Jelinek <jakub@redhat.com>
265 * config/arm/arm.md (uaddvdi4): Actually emit what gen_uaddvsi4
270 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext): Renamed to ...
271 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext): ... this. Use
272 any_extend code iterator instead of always zero_extend.
273 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_lt): Renamed to ...
274 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_lt): ... this.
275 Use any_extend code iterator instead of always zero_extend.
276 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_shift): Renamed to ...
277 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_shift): ... this.
278 Use any_extend code iterator instead of always zero_extend.
279 (*sse2_pmovmskb_ext): New define_insn.
280 (*sse2_pmovmskb_ext_lt): New define_insn_and_split.
283 * config/i386/i386.md (*popcountsi2_zext): New define_insn_and_split.
284 (*popcountsi2_zext_falsedep): New define_insn.
286 2020-01-30 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
288 * config.in: Regenerated.
289 * configure: Regenerated.
291 2020-01-29 Tobias Burnus <tobias@codesourcery.com>
294 * config/gcn/gcn-hsa.h (ASM_SPEC): Add -mattr=-code-object-v3 as
295 LLVM's assembler changed the default in version 9.
297 2020-01-24 Jeff Law <law@redhat.com>
299 PR tree-optimization/89689
300 * builtins.def (BUILT_IN_OBJECT_SIZE): Make it const rather than pure.
302 2020-01-29 Richard Sandiford <richard.sandiford@arm.com>
306 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
308 PR rtl-optimization/87763
309 * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
310 simplification to handle subregs as well as bare regs.
311 * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
313 2020-01-29 Joel Hutton <Joel.Hutton@arm.com>
316 * ira.c (ira): Revert use of simplified LRA algorithm.
318 2020-01-29 Martin Jambor <mjambor@suse.cz>
320 PR tree-optimization/92706
321 * tree-sra.c (struct access): Fields first_link, last_link,
322 next_queued and grp_queued renamed to first_rhs_link, last_rhs_link,
323 next_rhs_queued and grp_rhs_queued respectively, new fields
324 first_lhs_link, last_lhs_link, next_lhs_queued and grp_lhs_queued.
325 (struct assign_link): Field next renamed to next_rhs, new field
326 next_lhs. Updated comment.
327 (work_queue_head): Renamed to rhs_work_queue_head.
328 (lhs_work_queue_head): New variable.
329 (add_link_to_lhs): New function.
330 (relink_to_new_repr): Also relink LHS lists.
331 (add_access_to_work_queue): Renamed to add_access_to_rhs_work_queue.
332 (add_access_to_lhs_work_queue): New function.
333 (pop_access_from_work_queue): Renamed to
334 pop_access_from_rhs_work_queue.
335 (pop_access_from_lhs_work_queue): New function.
336 (build_accesses_from_assign): Also add links to LHS lists and to LHS
338 (child_would_conflict_in_lacc): Renamed to
339 child_would_conflict_in_acc. Adjusted parameter names.
340 (create_artificial_child_access): New parameter set_grp_read, use it.
341 (subtree_mark_written_and_enqueue): Renamed to
342 subtree_mark_written_and_rhs_enqueue.
343 (propagate_subaccesses_across_link): Renamed to
344 propagate_subaccesses_from_rhs.
345 (propagate_subaccesses_from_lhs): New function.
346 (propagate_all_subaccesses): Also propagate subaccesses from LHSs to
349 2020-01-29 Martin Jambor <mjambor@suse.cz>
351 PR tree-optimization/92706
352 * tree-sra.c (struct access): Adjust comment of
353 grp_total_scalarization.
354 (find_access_in_subtree): Look for single children spanning an entire
356 (scalarizable_type_p): Allow register accesses, adjust callers.
357 (completely_scalarize): Remove function.
358 (scalarize_elem): Likewise.
359 (create_total_scalarization_access): Likewise.
360 (sort_and_splice_var_accesses): Do not track total scalarization
362 (analyze_access_subtree): New parameter totally, adjust to new meaning
363 of grp_total_scalarization.
364 (analyze_access_trees): Pass new parameter to analyze_access_subtree.
365 (can_totally_scalarize_forest_p): New function.
366 (create_total_scalarization_access): Likewise.
367 (create_total_access_and_reshape): Likewise.
368 (total_should_skip_creating_access): Likewise.
369 (totally_scalarize_subtree): Likewise.
370 (analyze_all_variable_accesses): Perform total scalarization after
371 subaccess propagation using the new functions above.
372 (initialize_constant_pool_replacements): Output initializers by
373 traversing the access tree.
375 2020-01-29 Martin Jambor <mjambor@suse.cz>
377 * tree-sra.c (verify_sra_access_forest): New function.
378 (verify_all_sra_access_forests): Likewise.
379 (create_artificial_child_access): Set parent.
380 (analyze_all_variable_accesses): Call the verifier.
382 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
384 * cgraph.c (cgraph_edge::resolve_speculation): Only lookup direct edge
385 if called on indirect edge.
386 (cgraph_edge::redirect_call_stmt_to_callee): Lookup indirect edge of
387 speculative call if needed.
389 2020-01-29 Richard Biener <rguenther@suse.de>
391 PR tree-optimization/93428
392 * tree-vect-slp.c (vect_build_slp_tree_2): Compute the load
393 permutation when the load node is created.
394 (vect_analyze_slp_instance): Re-use it here.
396 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
398 * ipa-prop.c (update_indirect_edges_after_inlining): Fix warning.
400 2020-01-28 Vladimir Makarov <vmakarov@redhat.com>
402 PR rtl-optimization/93272
403 * ira-lives.c (process_out_of_region_eh_regs): New function.
404 (process_bb_node_lives): Call it.
406 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
408 * coverage.c (read_counts_file): Make error message lowercase.
410 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
412 * profile-count.c (profile_quality_display_names): Fix ordering.
414 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
417 * cgraph.c (cgraph_add_edge_to_call_site_hash): Update call site
418 hash only when edge is first within the sequence.
419 (cgraph_edge::set_call_stmt): Update handling of speculative calls.
420 (symbol_table::create_edge): Do not set target_prob.
421 (cgraph_edge::remove_caller): Watch for speculative calls when updating
423 (cgraph_edge::make_speculative): Drop target_prob parameter.
424 (cgraph_edge::speculative_call_info): Remove.
425 (cgraph_edge::first_speculative_call_target): New member function.
426 (update_call_stmt_hash_for_removing_direct_edge): New function.
427 (cgraph_edge::resolve_speculation): Rewrite to new API.
428 (cgraph_edge::speculative_call_for_target): New member function.
429 (cgraph_edge::make_direct): Rewrite to new API; fix handling of
430 multiple speculation targets.
431 (cgraph_edge::redirect_call_stmt_to_callee): Likewise; fix updating
433 (verify_speculative_call): Verify that targets form an interval.
434 * cgraph.h (cgraph_edge::speculative_call_info): Remove.
435 (cgraph_edge::first_speculative_call_target): New member function.
436 (cgraph_edge::next_speculative_call_target): New member function.
437 (cgraph_edge::speculative_call_target_ref): New member function.
438 (cgraph_edge;:speculative_call_indirect_edge): New member funtion.
439 (cgraph_edge): Remove target_prob.
440 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
441 Fix handling of speculative calls.
442 * ipa-devirt.c (ipa_devirt): Fix handling of speculative cals.
443 * ipa-fnsummary.c (analyze_function_body): Likewise.
444 * ipa-inline.c (speculation_useful_p): Use new speculative call API.
445 * ipa-profile.c (dump_histogram): Fix formating.
446 (ipa_profile_generate_summary): Watch for overflows.
447 (ipa_profile): Do not require probablity to be 1/2; update to new API.
448 * ipa-prop.c (ipa_make_edge_direct_to_target): Update to new API.
449 (update_indirect_edges_after_inlining): Update to new API.
450 * ipa-utils.c (ipa_merge_profiles): Rewrite merging of speculative call
452 * profile-count.h: (profile_probability::adjusted): New.
453 * tree-inline.c (copy_bb): Update to new speculative call API; fix
455 * value-prof.c (gimple_ic_transform): Rename to ...
456 (dump_ic_profile): ... this one; update dumping.
457 (stream_in_histogram_value): Fix formating.
458 (gimple_value_profile_transformations): Update.
460 2020-01-28 H.J. Lu <hongjiu.lu@intel.com>
463 * config/i386/i386.md (*movoi_internal_avx): Remove
464 TARGET_SSE_TYPELESS_STORES check.
465 (*movti_internal): Prefer TARGET_AVX over
466 TARGET_SSE_TYPELESS_STORES.
467 (*movtf_internal): Likewise.
468 * config/i386/sse.md (mov<mode>_internal): Prefer TARGET_AVX over
469 TARGET_SSE_TYPELESS_STORES. Remove "<MODE_SIZE> == 16" check
470 from TARGET_SSE_TYPELESS_STORES.
472 2020-01-28 David Malcolm <dmalcolm@redhat.com>
474 * diagnostic-core.h (warning_at): Rename overload to...
475 (warning_meta): ...this.
476 (emit_diagnostic_valist): Delete decl of overload taking
478 * diagnostic.c (emit_diagnostic_valist): Likewise for defn.
479 (warning_at): Rename overload taking diagnostic_metadata to...
480 (warning_meta): ...this.
482 2020-01-28 Richard Biener <rguenther@suse.de>
484 PR tree-optimization/93439
485 * tree-parloops.c (create_loop_fn): Move clique bookkeeping...
486 * tree-cfg.c (move_sese_region_to_fn): ... here.
487 (verify_types_in_gimple_reference): Verify used cliques are
490 2020-01-28 H.J. Lu <hongjiu.lu@intel.com>
493 * config/i386/i386-options.c (set_ix86_tune_features): Add an
494 argument of a pointer to struct gcc_options and pass it to
495 parse_mtune_ctrl_str.
496 (ix86_function_specific_restore): Pass opts to
497 set_ix86_tune_features.
498 (ix86_option_override_internal): Likewise.
499 (parse_mtune_ctrl_str): Add an argument of a pointer to struct
500 gcc_options and use it for x_ix86_tune_ctrl_string.
502 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
504 PR rtl-optimization/87763
505 * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
506 simplification to handle subregs as well as bare regs.
507 * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
509 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
511 * tree-vect-loop.c (vectorizable_reduction): Fail gracefully
512 for reduction chains that (now) include a call.
514 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
516 PR tree-optimization/92822
517 * tree-ssa-forwprop.c (simplify_vector_constructor): When filling
518 out the don't-care elements of a vector whose significant elements
519 are duplicates, make the don't-care elements duplicates too.
521 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
523 PR tree-optimization/93434
524 * tree-predcom.c (split_data_refs_to_components): Record which
525 components have had aliasing loads removed. Prevent store-store
526 commoning for all such components.
528 2020-01-28 Jakub Jelinek <jakub@redhat.com>
531 * config/i386/i386.c (ix86_fold_builtin) <do_shift>: If mask is not
532 -1 or is_vshift is true, use new_vector with number of elts npatterns
533 rather than new_unary_operation.
535 PR tree-optimization/93454
536 * gimple-fold.c (fold_array_ctor_reference): Perform
537 elt_size.to_uhwi () just once, instead of calling it in every
538 iteration. Punt if that value is above size of the temporary
539 buffer. Decrease third native_encode_expr argument when
540 bufoff + elt_sz is above size of buf.
542 2020-01-27 Joseph Myers <joseph@codesourcery.com>
544 * config/mips/mips.c (mips_declare_object_name)
545 [USE_GNU_UNIQUE_OBJECT]: Support use of gnu_unique_object.
547 2020-01-27 Martin Liska <mliska@suse.cz>
549 PR gcov-profile/93403
550 * tree-profile.c (gimple_init_gcov_profiler): Generate
551 both __gcov_indirect_call_profiler_v4 and
552 __gcov_indirect_call_profiler_v4_atomic.
554 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
557 * config/aarch64/aarch64-simd.md (aarch64_get_half<mode>): New
559 (@aarch64_split_simd_mov<mode>): Use it.
560 (aarch64_simd_mov_from_<mode>low): Add a GPR alternative.
561 Leave the vec_extract patterns to handle 2-element vectors.
562 (aarch64_simd_mov_from_<mode>high): Likewise.
563 (vec_extract<VQMOV_NO2E:mode><Vhalf>): New expander.
564 (vec_extractv2dfv1df): Likewise.
566 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
568 * config/aarch64/aarch64.c (aarch64_if_then_else_costs): Match
569 jump conditions for *compare_condjump<GPI:mode>.
571 2020-01-27 David Malcolm <dmalcolm@redhat.com>
574 * digraph.cc (test_edge::test_edge): Specify template for base
577 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
579 * config/arc/arc.c (arc_rtx_costs): Update mul64 cost.
581 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
583 * config/arc/arc-protos.h (gen_mlo): Remove.
585 * config/arc/arc.c (AUX_MULHI): Define.
586 (arc_must_save_reister): Special handling for r58/59.
587 (arc_compute_frame_size): Consider mlo/mhi registers.
588 (arc_save_callee_saves): Emit fp/sp move only when emit_move
590 (arc_conditional_register_usage): Remove TARGET_BIG_ENDIAN from
591 mlo/mhi name selection.
592 (arc_restore_callee_saves): Don't early restore blink when ISR.
593 (arc_expand_prologue): Add mlo/mhi saving.
594 (arc_expand_epilogue): Add mlo/mhi restoring.
597 * config/arc/arc.h (DBX_REGISTER_NUMBER): Correct register
598 numbering when MUL64 option is used.
599 (DWARF2_FRAME_REG_OUT): Define.
600 * config/arc/arc.md (arc600_stall): New pattern.
601 (VUNSPEC_ARC_ARC600_STALL): Define.
602 (mulsi64): Use correct mlo/mhi registers.
603 (mulsi_600): Clean it up.
604 * config/arc/predicates.md (mlo_operand): Remove any dependency on
606 (mhi_operand): Likewise.
608 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
609 Petro Karashchenko <petro.karashchenko@ring.com>
611 * config/arc/arc.c (arc_is_uncached_mem_p): Check struct
612 attributes if needed.
613 (prepare_move_operands): Generate special unspec instruction for
615 (arc_isuncached_mem_p): Propagate uncached attribute to each
617 * config/arc/arc.md (VUNSPEC_ARC_LDDI): Define.
618 (VUNSPEC_ARC_STDI): Likewise.
619 (ALLI): New mode iterator.
620 (mALLI): New mode attribute.
621 (lddi): New instruction pattern.
623 (stdidi_split): Split instruction for architectures which are not
624 supporting ll64 option.
625 (lddidi_split): Likewise.
627 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
629 PR rtl-optimization/92989
630 * lra-lives.c (process_bb_lives): Update the live-in set before
631 processing additional clobbers.
633 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
635 PR rtl-optimization/93170
636 * cselib.c (cselib_invalidate_regno_val): New function, split out
638 (cselib_invalidate_regno): ...here.
639 (cselib_invalidated_by_call_p): New function.
640 (cselib_process_insn): Iterate over all the hard-register entries in
641 REG_VALUES and invalidate any that cross call-clobbered registers.
643 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
645 * dojump.c (split_comparison): Use HONOR_NANS rather than
646 HONOR_SNANS when splitting LTGT.
648 2020-01-27 Martin Liska <mliska@suse.cz>
651 * opts.c (print_filtered_help): Exclude language-specific
652 options from --help=common unless enabled in all FEs.
654 2020-01-27 Martin Liska <mliska@suse.cz>
656 * opts.c (print_help): Exclude params from
657 all except --help=param.
659 2020-01-27 Martin Liska <mliska@suse.cz>
662 * config/i386/i386-features.c (make_resolver_func):
663 Align the code with ppc64 target implementation.
664 Do not generate a unique name for resolver function.
666 2020-01-27 Richard Biener <rguenther@suse.de>
668 PR tree-optimization/93397
669 * tree-vect-slp.c (vect_analyze_slp_instance): Delay
670 converted reduction chain SLP graph adjustment.
672 2020-01-26 Marek Polacek <polacek@redhat.com>
675 * sanopt.c (sanitize_rewrite_addressable_params): Avoid crash on
678 2020-01-26 Jason Merrill <jason@redhat.com>
681 * tree.c (verify_type_variant): Only verify TYPE_NEEDS_CONSTRUCTING
684 2020-01-26 Darius Galis <darius.galis@cyberthorstudios.com>
686 * config/rx/rx.md (setmemsi): Added rx_allow_string_insns constraint
687 (rx_setmem): Likewise.
689 2020-01-26 Jakub Jelinek <jakub@redhat.com>
692 * config/i386/i386.md (*addv<dwi>4_doubleword, *subv<dwi>4_doubleword):
693 Use nonimmediate_operand instead of x86_64_hilo_general_operand and
694 drop <di> from constraint of last operand.
697 * config/i386/sse.md (*avx_vperm_broadcast_<mode>): Disallow for
698 TARGET_AVX2 and V4DFmode not in the split condition, but in the
699 pattern condition, though allow { 0, 0, 0, 0 } broadcast always.
701 2020-01-25 Feng Xue <fxue@os.amperecomputing.com>
704 * ipa-cp.c (get_info_about_necessary_edges): Remove value
707 2020-01-24 Jeff Law <law@redhat.com>
709 PR tree-optimization/92788
710 * tree-ssa-threadedge.c (thread_across_edge): Check EDGE_COMPLEX
713 2020-01-24 Jakub Jelinek <jakub@redhat.com>
716 * config/i386/sse.md (*avx_vperm_broadcast_v4sf,
717 *avx_vperm_broadcast_<mode>,
718 <sse2_avx_avx512f>_vpermil<mode><mask_name>,
719 *<sse2_avx_avx512f>_vpermilp<mode><mask_name>):
720 Move before avx2_perm<mode>/avx512f_perm<mode>.
723 * simplify-rtx.c (simplify_const_unary_operation,
724 simplify_const_binary_operation): Punt for mode precision above
725 MAX_BITSIZE_MODE_ANY_INT.
727 2020-01-24 Andrew Pinski <apinski@marvell.com>
729 * config/arm/aarch-cost-tables.h (cortexa57_extra_costs): Change
732 2020-01-24 Jeff Law <law@redhat.com>
735 * config/h8300/h8300.c (h8300_print_operand): Only call byte_reg
736 for REGs. Call output_operand_lossage to get more reasonable
739 2020-01-24 Andrew Stubbs <ams@codesourcery.com>
741 * config/gcn/gcn-valu.md (vec_cmp<mode>di): Use
742 gcn_fp_compare_operator.
743 (vec_cmpu<mode>di): Use gcn_compare_operator.
744 (vec_cmp<u>v64qidi): Use gcn_compare_operator.
745 (vec_cmp<mode>di_exec): Use gcn_fp_compare_operator.
746 (vec_cmpu<mode>di_exec): Use gcn_compare_operator.
747 (vec_cmp<u>v64qidi_exec): Use gcn_compare_operator.
748 (vec_cmp<mode>di_dup): Use gcn_fp_compare_operator.
749 (vec_cmp<mode>di_dup_exec): Use gcn_fp_compare_operator.
750 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): Use
751 gcn_fp_compare_operator.
752 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): Use
753 gcn_fp_compare_operator.
754 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): Use
755 gcn_fp_compare_operator.
756 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): Use
757 gcn_fp_compare_operator.
759 2020-01-24 Maciej W. Rozycki <macro@wdc.com>
761 * doc/install.texi (Cross-Compiler-Specific Options): Document
762 `--with-toolexeclibdir' option.
764 2020-01-24 Hans-Peter Nilsson <hp@axis.com>
766 * target.def (flags_regnum): Also mention effect on delay slot filling.
767 * doc/tm.texi: Regenerate.
769 2020-01-23 Jeff Law <law@redhat.com>
772 * config/h8300/h8300.c (h8300_option_override): Fix diagnostic text.
774 2020-01-23 Mikael Tillenius <mti-1@tillenius.com>
777 * config/h8300/h8300.h (FUNCTION_PROFILER): Fix emission of
780 2020-01-23 Jakub Jelinek <jakub@redhat.com>
782 PR rtl-optimization/93402
783 * postreload.c (reload_combine_recognize_pattern): Don't try to adjust
786 2020-01-23 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
788 * config.in: Regenerated.
789 * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to 1
790 for TARGET_LIBC_GNUSTACK.
791 * configure: Regenerated.
792 * configure.ac: Define TARGET_LIBC_GNUSTACK if glibc version is
793 found to be 2.31 or greater.
795 2020-01-23 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
797 * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to
799 * config/mips/mips.c (TARGET_ASM_FILE_END): Define to ...
800 (mips_asm_file_end): New function. Delegate to
801 file_end_indicate_exec_stack if NEED_INDICATE_EXEC_STACK is true.
802 * config/mips/mips.h (NEED_INDICATE_EXEC_STACK): Define to 0.
804 2020-01-23 Jakub Jelinek <jakub@redhat.com>
807 * config/i386/i386-modes.def (POImode): New mode.
808 (MAX_BITSIZE_MODE_ANY_INT): Change from 128 to 160.
809 * config/i386/i386.md (DPWI): New mode attribute.
810 (addv<mode>4, subv<mode>4): Use <DPWI> instead of <DWI>.
812 (QPWI): ... this. Use POI instead of OI for TImode.
813 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1,
814 *subv<dwi>4_doubleword, *subv<dwi>4_doubleword_1): Use <QPWI>
817 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
820 * config/aarch64/aarch64.md (UNSPEC_SPECULATION_TRACKER_REV): New
822 (speculation_tracker_rev): New pattern.
823 * config/aarch64/aarch64-speculation.cc (aarch64_do_track_speculation):
824 Use speculation_tracker_rev to track the inverse condition.
826 2020-01-23 Richard Biener <rguenther@suse.de>
828 PR tree-optimization/93381
829 * tree-ssa-sccvn.c (vn_walk_cb_data::push_partial_def): Take
830 alias-set of the def as argument and record the first one.
831 (vn_walk_cb_data::first_set): New member.
832 (vn_reference_lookup_3): Pass the alias-set of the current def
833 to push_partial_def. Fix alias-set used in the aggregate copy
835 (vn_reference_lookup): Consistently set *last_vuse_ptr.
836 * real.c (clear_significand_below): Fix out-of-bound access.
838 2020-01-23 Jakub Jelinek <jakub@redhat.com>
841 * config/i386/i386.md (*bmi2_bzhi_<mode>3_2, *bmi2_bzhi_<mode>3_3):
842 New define_insn patterns.
844 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
846 * doc/sourcebuild.texi (check-function-bodies): Add an
847 optional target/xfail selector.
849 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
851 PR rtl-optimization/93124
852 * auto-inc-dec.c (merge_in_block): Don't add auto inc/decs to
853 bare USE and CLOBBER insns.
855 2020-01-22 Andrew Pinski <apinski@marvell.com>
857 * config/arc/arc.c (output_short_suffix): Check insn for nullness.
859 2020-01-22 David Malcolm <dmalcolm@redhat.com>
862 * gdbinit.in (break-on-saved-diagnostic): Update for move of
863 diagnostic_manager into "ana" namespace.
864 * selftest-run-tests.c (selftest::run_tests): Update for move of
865 selftest::run_analyzer_selftests to
866 ana::selftest::run_analyzer_selftests.
868 2020-01-22 Richard Sandiford <richard.sandiford@arm.com>
870 * cfgexpand.c (union_stack_vars): Update the size.
872 2020-01-22 Richard Biener <rguenther@suse.de>
874 PR tree-optimization/93381
875 * tree-ssa-structalias.c (find_func_aliases): Assume offsetting
876 throughout, handle all conversions the same.
878 2020-01-22 Jakub Jelinek <jakub@redhat.com>
881 * config/aarch64/aarch64.c (aarch64_expand_subvti): Only use
882 gen_subdi3_compare1_imm if low_in2 satisfies aarch64_plus_immediate
883 predicate, not whenever it is CONST_INT. Otherwise, force_reg it.
884 Call force_reg on high_in2 unconditionally.
886 2020-01-22 Martin Liska <mliska@suse.cz>
888 PR tree-optimization/92924
889 * profile.c (compute_value_histograms): Divide
892 2020-01-22 Jakub Jelinek <jakub@redhat.com>
895 * output.h (assemble_name_resolve): Declare.
896 * varasm.c (assemble_name_resolve): New function.
897 (assemble_name): Use it.
898 * config/i386/i386.h (ASM_OUTPUT_SYMBOL_REF): Define.
900 2020-01-22 Joseph Myers <joseph@codesourcery.com>
902 * doc/sourcebuild.texi (Texinfo Manuals, Front End): Refer to
903 update_web_docs_git instead of update_web_docs_svn.
905 2020-01-21 Andrew Pinski <apinski@marvell.com>
908 * config/aarch64/aarch64.md (tlsgd_small_<mode>): Have operand 0
909 as PTR mode. Have operand 1 as being modeless, it can be P mode.
910 (*tlsgd_small_<mode>): Likewise.
911 * config/aarch64/aarch64.c (aarch64_load_symref_appropriately)
912 <case SYMBOL_SMALL_TLSGD>: Call gen_tlsgd_small_* with a ptr_mode
913 register. Convert that register back to dest using convert_mode.
915 2020-01-21 Jim Wilson <jimw@sifive.com>
917 * config/riscv/riscv-sr.c (riscv_sr_match_prologue): Use INTVAL
920 2020-01-21 H.J. Lu <hongjiu.lu@intel.com>
921 Uros Bizjak <ubizjak@gmail.com>
924 * config/i386/i386.c (ix86_tls_module_base): Replace Pmode
926 (legitimize_tls_address): Do GNU2 TLS address computation in
927 ptr_mode and zero-extend result to Pmode.
928 * config/i386/i386.md (@tls_dynamic_gnu2_64_<mode>): Replace
929 :P with :PTR and Pmode with ptr_mode.
930 (*tls_dynamic_gnu2_lea_64_<mode>): Likewise.
931 (*tls_dynamic_gnu2_call_64_<mode>): Likewise.
932 (*tls_dynamic_gnu2_combine_64_<mode>): Likewise.
934 2020-01-21 Jakub Jelinek <jakub@redhat.com>
937 * config/riscv/riscv.c (riscv_rtx_costs) <case ZERO_EXTRACT>: Verify
938 the last two operands are CONST_INT_P before using them as such.
940 2020-01-21 Richard Sandiford <richard.sandiford@arm.com>
942 * config/aarch64/aarch64-sve-builtins.def: Use get_typenode_from_name
943 to get the integer element types.
945 2020-01-21 Richard Sandiford <richard.sandiford@arm.com>
947 * config/aarch64/aarch64-sve-builtins.h
948 (function_expander::convert_to_pmode): Declare.
949 * config/aarch64/aarch64-sve-builtins.cc
950 (function_expander::convert_to_pmode): New function.
951 (function_expander::get_contiguous_base): Use it.
952 (function_expander::prepare_gather_address_operands): Likewise.
953 * config/aarch64/aarch64-sve-builtins-sve2.cc
954 (svwhilerw_svwhilewr_impl::expand): Likewise.
956 2020-01-21 Szabolcs Nagy <szabolcs.nagy@arm.com>
959 * config/aarch64/aarch64.c (aarch64_declare_function_name): Set
960 cfun->machine->label_is_assembled.
961 (aarch64_print_patchable_function_entry): New.
962 (TARGET_ASM_PRINT_PATCHABLE_FUNCTION_ENTRY): Define.
963 * config/aarch64/aarch64.h (struct machine_function): New field,
966 2020-01-21 David Malcolm <dmalcolm@redhat.com>
969 * ipa-profile.c (ipa_profile): Delete call_sums and set it to
972 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
975 * cgraph.c (cgraph_edge::resolve_speculation,
976 cgraph_edge::redirect_call_stmt_to_callee): Fix update of
979 2020-01-21 Martin Liska <mliska@suse.cz>
981 * config/rs6000/rs6000.c (common_mode_defined): Remove
984 2020-01-21 Richard Biener <rguenther@suse.de>
986 PR tree-optimization/92328
987 * tree-ssa-sccvn.c (vn_reference_lookup_3): Preserve
988 type when value-numbering same-sized store by inserting a
990 (eliminate_dom_walker::eliminate_stmt): When eliminating
991 a redundant store handle bit-reinterpretation of the same value.
993 2020-01-21 Andrew Pinski <apinski@marvel.com>
996 * tree-into-ssa.c (prepare_block_for_update_1): Split out
998 (prepare_block_for_update): This. Use a worklist instead of
1001 2020-01-21 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1003 * gcc/config/arm/arm.c (clear_operation_p):
1004 Initialise last_regno, skip first iteration
1005 based on the first_set value and use ints instead
1006 of the unnecessary HOST_WIDE_INTs.
1008 2020-01-21 Jakub Jelinek <jakub@redhat.com>
1011 * config/rs6000/rs6000.c (rs6000_emit_cmove): If using fsel, punt for
1012 compare_mode other than SFmode or DFmode.
1014 2020-01-21 Kito Cheng <kito.cheng@sifive.com>
1017 * config/riscv/riscv-protos.h (riscv_hard_regno_rename_ok): New.
1018 * config/riscv/riscv.c (riscv_hard_regno_rename_ok): New.
1019 * config/riscv/riscv.h (HARD_REGNO_RENAME_OK): Defined.
1021 2020-01-20 Wilco Dijkstra <wdijkstr@arm.com>
1023 * config/aarch64/aarch64.c (neoversen1_tunings): Set jump_align to 4.
1025 2020-01-20 Andrew Pinski <apinski@marvell.com>
1028 * targhooks.c (default_print_patchable_function_entry): Use
1029 output_asm_insn to emit the nop instruction.
1031 2020-01-20 Fangrui Song <maskray@google.com>
1034 * targhooks.c (default_print_patchable_function_entry): Align to
1037 2020-01-20 H.J. Lu <hongjiu.lu@intel.com>
1040 * config/i386/i386.c (legitimize_tls_address): Pass Pmode to
1041 gen_tls_dynamic_gnu2_64. Compute GNU2 TLS address in ptr_mode.
1042 * config/i386/i386.md (tls_dynamic_gnu2_64): Renamed to ...
1043 (@tls_dynamic_gnu2_64_<mode>): This. Replace DI with P.
1044 (*tls_dynamic_gnu2_lea_64): Renamed to ...
1045 (*tls_dynamic_gnu2_lea_64_<mode>): This. Replace DI with P.
1046 Remove the {q} suffix from lea.
1047 (*tls_dynamic_gnu2_call_64): Renamed to ...
1048 (*tls_dynamic_gnu2_call_64_<mode>): This. Replace DI with P.
1049 (*tls_dynamic_gnu2_combine_64): Renamed to ...
1050 (*tls_dynamic_gnu2_combine_64_<mode>): This. Replace DI with P.
1051 Pass Pmode to gen_tls_dynamic_gnu2_64.
1053 2020-01-20 Wilco Dijkstra <wdijkstr@arm.com>
1055 * config/aarch64/aarch64.h (SLOW_BYTE_ACCESS): Set to 1.
1057 2020-01-20 Richard Sandiford <richard.sandiford@arm.com>
1059 * config/aarch64/aarch64-sve-builtins-base.cc
1060 (svld1ro_impl::memory_vector_mode): Remove parameter name.
1062 2020-01-20 Richard Biener <rguenther@suse.de>
1065 * dwarf2out.c (prune_unused_types): Unconditionally mark
1066 called function DIEs.
1068 2020-01-20 Martin Liska <mliska@suse.cz>
1070 PR tree-optimization/93199
1071 * tree-eh.c (struct leh_state): Add
1072 new field outer_non_cleanup.
1073 (cleanup_is_dead_in): Pass leh_state instead
1074 of eh_region. Add a checking that state->outer_non_cleanup
1075 points to outer non-clean up region.
1076 (lower_try_finally): Record outer_non_cleanup
1078 (lower_catch): Likewise.
1079 (lower_eh_filter): Likewise.
1080 (lower_eh_must_not_throw): Likewise.
1081 (lower_cleanup): Likewise.
1083 2020-01-20 Richard Biener <rguenther@suse.de>
1085 PR tree-optimization/93094
1086 * tree-vectorizer.h (vect_loop_versioning): Adjust.
1087 (vect_transform_loop): Likewise.
1088 * tree-vectorizer.c (try_vectorize_loop_1): Pass down
1089 loop_vectorized_call to vect_transform_loop.
1090 * tree-vect-loop.c (vect_transform_loop): Pass down
1091 loop_vectorized_call to vect_loop_versioning.
1092 * tree-vect-loop-manip.c (vect_loop_versioning): Use
1093 the earlier discovered loop_vectorized_call.
1095 2020-01-19 Eric S. Raymond <esr@thyrsus.com>
1097 * doc/contribute.texi: Update for SVN -> Git transition.
1098 * doc/install.texi: Likewise.
1100 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
1103 * cgraph.c (cgraph_edge::make_speculative): Increase number of
1104 speculative targets.
1105 (verify_speculative_call): New function
1106 (cgraph_node::verify_node): Use it.
1107 * ipa-profile.c (ipa_profile): Fix formating; do not set number of
1110 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
1113 * cgraph.c (cgraph_edge::resolve_speculation): Fix foramting.
1114 (cgraph_edge::make_direct): Remove all indirect targets.
1115 (cgraph_edge::redirect_call_stmt_to_callee): Use make_direct..
1116 (cgraph_node::verify_node): Verify that only one call_stmt or
1117 lto_stmt_uid is set.
1118 * cgraphclones.c (cgraph_edge::clone): Set only one call_stmt or
1120 * lto-cgraph.c (lto_output_edge): Simplify streaming of stmt.
1121 (lto_output_ref): Simplify streaming of stmt.
1122 * lto-streamer-in.c (fixup_call_stmt_edges_1): Clear lto_stmt_uid.
1124 2020-01-18 Tamar Christina <tamar.christina@arm.com>
1126 * config/aarch64/aarch64-sve-builtins-base.cc (memory_vector_mode):
1127 Mark parameter unused.
1129 2020-01-18 Hans-Peter Nilsson <hp@axis.com>
1131 * config.gcc <obsolete targets>: Add crisv32-*-* and cris-*-linux*
1133 2019-01-18 Gerald Pfeifer <gerald@pfeifer.com>
1135 * varpool.c (ctor_useable_for_folding_p): Fix grammar.
1137 2020-01-18 Iain Sandoe <iain@sandoe.co.uk>
1139 * Makefile.in: Add coroutine-passes.o.
1140 * builtin-types.def (BT_CONST_SIZE): New.
1141 (BT_FN_BOOL_PTR): New.
1142 (BT_FN_PTR_PTR_CONST_SIZE_BOOL): New.
1143 * builtins.def (DEF_COROUTINE_BUILTIN): New.
1144 * coroutine-builtins.def: New file.
1145 * coroutine-passes.cc: New file.
1146 * function.h (struct GTY function): Add a bit to indicate that the
1147 function is a coroutine component.
1148 * internal-fn.c (expand_CO_FRAME): New.
1149 (expand_CO_YIELD): New.
1150 (expand_CO_SUSPN): New.
1151 (expand_CO_ACTOR): New.
1152 * internal-fn.def (CO_ACTOR): New.
1156 * passes.def: Add pass_coroutine_lower_builtins,
1157 pass_coroutine_early_expand_ifns.
1158 * tree-pass.h (make_pass_coroutine_lower_builtins): New.
1159 (make_pass_coroutine_early_expand_ifns): New.
1160 * doc/invoke.texi: Document the fcoroutines command line
1163 2020-01-18 Jakub Jelinek <jakub@redhat.com>
1165 * config/arm/vfp.md (*clear_vfp_multiple): Remove unused variable.
1168 * config/arm/arm.c (clear_operation_p): Don't use REGNO until
1169 after checking the argument is a REG. Don't use REGNO (reg)
1170 again to set last_regno, reuse regno variable instead.
1172 2020-01-17 David Malcolm <dmalcolm@redhat.com>
1174 * doc/analyzer.texi (Limitations): Add note about NaN.
1176 2020-01-17 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1177 Sudakshina Das <sudi.das@arm.com>
1179 * config/arm/arm.md (ashldi3): Generate thumb2_lsll for both reg
1180 and valid immediate.
1181 (ashrdi3): Generate thumb2_asrl for both reg and valid immediate.
1182 (lshrdi3): Generate thumb2_lsrl for valid immediates.
1183 * config/arm/constraints.md (Pg): New.
1184 * config/arm/predicates.md (long_shift_imm): New.
1185 (arm_reg_or_long_shift_imm): Likewise.
1186 * config/arm/thumb2.md (thumb2_asrl): New immediate alternative.
1187 (thumb2_lsll): Likewise.
1190 2020-01-17 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1191 Sudakshina Das <sudi.das@arm.com>
1193 * config/arm/arm.md (ashldi3): Generate thumb2_lsll for TARGET_HAVE_MVE.
1194 (ashrdi3): Generate thumb2_asrl for TARGET_HAVE_MVE.
1195 * config/arm/arm.c (arm_hard_regno_mode_ok): Allocate even odd
1196 register pairs for doubleword quantities for ARMv8.1M-Mainline.
1197 * config/arm/thumb2.md (thumb2_asrl): New.
1198 (thumb2_lsll): Likewise.
1200 2020-01-17 Jakub Jelinek <jakub@redhat.com>
1202 * config/arm/arm.c (cmse_nonsecure_call_inline_register_clear): Remove
1205 2020-01-17 Alexander Monakov <amonakov@ispras.ru>
1207 * gdbinit.in (help-gcc-hooks): New command.
1208 (pp, pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, ptc, pdn, ptn, pdd, prc,
1209 pi, pbm, pel, trt): Take $arg0 instead of $ if supplied. Update
1212 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
1214 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use the
1215 correct target macro.
1217 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
1219 * config/aarch64/aarch64-protos.h
1220 (aarch64_sve_ld1ro_operand_p): New.
1221 * config/aarch64/aarch64-sve-builtins-base.cc
1222 (class load_replicate): New.
1223 (class svld1ro_impl): New.
1224 (class svld1rq_impl): Change to inherit from load_replicate.
1225 (svld1ro): New sve intrinsic function base.
1226 * config/aarch64/aarch64-sve-builtins-base.def (svld1ro):
1227 New DEF_SVE_FUNCTION.
1228 * config/aarch64/aarch64-sve-builtins-base.h
1229 (svld1ro): New decl.
1230 * config/aarch64/aarch64-sve-builtins.cc
1231 (function_expander::add_mem_operand): Modify assert to allow
1233 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): New
1235 * config/aarch64/aarch64.c
1236 (aarch64_sve_ld1rq_operand_p): Implement in terms of ...
1237 (aarch64_sve_ld1rq_ld1ro_operand_p): This.
1238 (aarch64_sve_ld1ro_operand_p): New.
1239 * config/aarch64/aarch64.md (UNSPEC_LD1RO): New unspec.
1240 * config/aarch64/constraints.md (UOb,UOh,UOw,UOd): New.
1241 * config/aarch64/predicates.md
1242 (aarch64_sve_ld1ro_operand_{b,h,w,d}): New.
1244 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
1246 * config/aarch64/aarch64-c.c (_ARM_FEATURE_MATMUL_FLOAT64):
1247 Introduce this ACLE specified predefined macro.
1248 * config/aarch64/aarch64-option-extensions.def (f64mm): New.
1249 (fp): Disabling this disables f64mm.
1250 (simd): Disabling this disables f64mm.
1251 (fp16): Disabling this disables f64mm.
1252 (sve): Disabling this disables f64mm.
1253 * config/aarch64/aarch64.h (AARCH64_FL_F64MM): New.
1254 (AARCH64_ISA_F64MM): New.
1255 (TARGET_F64MM): New.
1256 * doc/invoke.texi (f64mm): Document new option.
1258 2020-01-17 Wilco Dijkstra <wdijkstr@arm.com>
1260 * config/aarch64/aarch64.c (generic_tunings): Add branch fusion.
1261 (neoversen1_tunings): Likewise.
1263 2020-01-17 Wilco Dijkstra <wdijkstr@arm.com>
1266 * config/aarch64/aarch64.c (aarch64_split_compare_and_swap)
1267 Add assert to ensure prolog has been emitted.
1268 (aarch64_split_atomic_op): Likewise.
1269 * config/aarch64/atomics.md (aarch64_compare_and_swap<mode>)
1270 Use epilogue_completed rather than reload_completed.
1271 (aarch64_atomic_exchange<mode>): Likewise.
1272 (aarch64_atomic_<atomic_optab><mode>): Likewise.
1273 (atomic_nand<mode>): Likewise.
1274 (aarch64_atomic_fetch_<atomic_optab><mode>): Likewise.
1275 (atomic_fetch_nand<mode>): Likewise.
1276 (aarch64_atomic_<atomic_optab>_fetch<mode>): Likewise.
1277 (atomic_nand_fetch<mode>): Likewise.
1279 2020-01-17 Richard Sandiford <richard.sandiford@arm.com>
1282 * config/aarch64/aarch64.h (REVERSIBLE_CC_MODE): Return false
1284 (REVERSE_CONDITION): Delete.
1285 * config/aarch64/iterators.md (CC_ONLY): New mode iterator.
1286 (CCFP_CCFPE): Likewise.
1287 (e): New mode attribute.
1288 * config/aarch64/aarch64.md (ccmp<GPI:mode>): Rename to...
1289 (@ccmp<CC_ONLY:mode><GPI:mode>): ...this, using CC_ONLY instead of CC.
1290 (fccmp<GPF:mode>, fccmpe<GPF:mode>): Merge into...
1291 (@ccmp<CCFP_CCFPE:mode><GPF:mode>): ...this combined pattern.
1292 (@ccmp<CC_ONLY:mode><GPI:mode>_rev): New pattern.
1293 (@ccmp<CCFP_CCFPE:mode><GPF:mode>_rev): Likewise.
1294 * config/aarch64/aarch64.c (aarch64_gen_compare_reg): Update
1295 name of generator from gen_ccmpdi to gen_ccmpccdi.
1296 (aarch64_gen_ccmp_next): Use code_for_ccmp. If we want to reverse
1297 the previous comparison but aren't able to, use the new ccmp_rev
1300 2020-01-17 Richard Sandiford <richard.sandiford@arm.com>
1302 * gimplify.c (gimplify_return_expr): Use poly_int_tree_p rather
1303 than testing directly for INTEGER_CST.
1304 (gimplify_target_expr, gimplify_omp_depend): Likewise.
1306 2020-01-17 Jakub Jelinek <jakub@redhat.com>
1308 PR tree-optimization/93292
1309 * tree-vect-stmts.c (vectorizable_comparison): Punt also if
1310 get_vectype_for_scalar_type returns NULL.
1312 2020-01-16 Jan Hubicka <hubicka@ucw.cz>
1314 * params.opt (-param=max-predicted-iterations): Increase range from 0.
1315 * predict.c (estimate_loops): Add 1 to param_max_predicted_iterations.
1317 2020-01-16 Jan Hubicka <hubicka@ucw.cz>
1319 * ipa-fnsummary.c (estimate_calls_size_and_time): Fix formating of
1321 * params.opt: (max-predicted-iterations): Set bounds.
1322 * predict.c (real_almost_one, real_br_prob_base,
1323 real_inv_br_prob_base, real_one_half, real_bb_freq_max): Remove.
1324 (propagate_freq): Add max_cyclic_prob parameter; cap cyclic
1325 probabilities; do not truncate to reg_br_prob_bases.
1326 (estimate_loops_at_level): Pass max_cyclic_prob.
1327 (estimate_loops): Compute max_cyclic_prob.
1328 (estimate_bb_frequencies): Do not initialize real_*; update calculation
1330 * profile-count.c (profile_probability::to_sreal): New.
1331 * profile-count.h (class sreal): Move up in file.
1332 (profile_probability::to_sreal): Declare.
1334 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
1337 (arm_invalid_conversion): New function for target hook.
1338 (arm_invalid_unary_op): New function for target hook.
1339 (arm_invalid_binary_op): New function for target hook.
1341 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
1343 * config.gcc: Add arm_bf16.h.
1344 * config/arm/arm-builtins.c (arm_mangle_builtin_type): Fix comment.
1345 (arm_simd_builtin_std_type): Add BFmode.
1346 (arm_init_simd_builtin_types): Define element types for vector types.
1347 (arm_init_bf16_types): New function.
1348 (arm_init_builtins): Add arm_init_bf16_types function call.
1349 * config/arm/arm-modes.def: Add BFmode and V4BF, V8BF vector modes.
1350 * config/arm/arm-simd-builtin-types.def: Add V4BF, V8BF.
1351 * config/arm/arm.c (aapcs_vfp_sub_candidate): Add BFmode.
1352 (arm_hard_regno_mode_ok): Add BFmode and tidy up statements.
1353 (arm_vector_mode_supported_p): Add V4BF, V8BF.
1354 (arm_mangle_type): Add __bf16.
1355 * config/arm/arm.h: Add V4BF, V8BF to VALID_NEON_DREG_MODE,
1356 VALID_NEON_QREG_MODE respectively. Add export arm_bf16_type_node,
1357 arm_bf16_ptr_type_node.
1358 * config/arm/arm.md: Add BFmode to movhf expand, mov pattern and
1359 define_split between ARM registers.
1360 * config/arm/arm_bf16.h: New file.
1361 * config/arm/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
1362 * config/arm/iterators.md: (ANY64_BF, VDXMOV, VHFBF, HFBF, fporbf): New.
1364 * config/arm/neon.md: Add BF vector types to movhf NEON move patterns.
1365 * config/arm/vfp.md: Add BFmode to movhf patterns.
1367 2020-01-16 Mihail Ionescu <mihail.ionescu@arm.com>
1368 Andre Vieira <andre.simoesdiasvieira@arm.com>
1370 * config/arm/arm-cpus.in (mve, mve_float): New features.
1371 (dsp, mve, mve.fp): New options.
1372 * config/arm/arm.h (TARGET_HAVE_MVE, TARGET_HAVE_MVE_FLOAT): Define.
1373 * config/arm/t-rmprofile: Map v8.1-M multilibs to v8-M.
1374 * doc/invoke.texi: Document the armv8.1-m mve and dps options.
1376 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1377 Thomas Preud'homme <thomas.preudhomme@arm.com>
1379 * config/arm/arm-cpus.in (ARMv8_1m_main): Redefine as an extension to
1381 * config/arm/arm.c (arm_options_perform_arch_sanity_checks): Remove
1382 error for using -mcmse when targeting Armv8.1-M Mainline.
1384 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1385 Thomas Preud'homme <thomas.preudhomme@arm.com>
1387 * config/arm/arm.md (nonsecure_call_internal): Do not force memory
1388 address in r4 when targeting Armv8.1-M Mainline.
1389 (nonsecure_call_value_internal): Likewise.
1390 * config/arm/thumb2.md (nonsecure_call_reg_thumb2): Make memory address
1391 a register match_operand again. Emit BLXNS when targeting
1393 (nonsecure_call_value_reg_thumb2): Likewise.
1395 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1396 Thomas Preud'homme <thomas.preudhomme@arm.com>
1398 * config/arm/arm.c (arm_add_cfa_adjust_cfa_note): Declare early.
1399 (cmse_nonsecure_call_inline_register_clear): Define new lazy_fpclear
1400 variable as true when floating-point ABI is not hard. Replace
1401 check against TARGET_HARD_FLOAT_ABI by checks against lazy_fpclear.
1402 Generate VLSTM and VLLDM instruction respectively before and
1403 after a function call to cmse_nonsecure_call function.
1404 * config/arm/unspecs.md (VUNSPEC_VLSTM): Define unspec.
1405 (VUNSPEC_VLLDM): Likewise.
1406 * config/arm/vfp.md (lazy_store_multiple_insn): New define_insn.
1407 (lazy_load_multiple_insn): Likewise.
1409 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1410 Thomas Preud'homme <thomas.preudhomme@arm.com>
1412 * config/arm/arm.c (vfp_emit_fstmd): Declare early.
1413 (arm_emit_vfp_multi_reg_pop): Likewise.
1414 (cmse_nonsecure_call_inline_register_clear): Abstract number of VFP
1415 registers to clear in max_fp_regno. Emit VPUSH and VPOP to save and
1416 restore callee-saved VFP registers.
1418 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1419 Thomas Preud'homme <thomas.preudhomme@arm.com>
1421 * config/arm/arm.c (arm_emit_multi_reg_pop): Declare early.
1422 (cmse_nonsecure_call_clear_caller_saved): Rename into ...
1423 (cmse_nonsecure_call_inline_register_clear): This. Save and clear
1424 callee-saved GPRs as well as clear ip register before doing a nonsecure
1425 call then restore callee-saved GPRs after it when targeting
1427 (arm_reorg): Adapt to function rename.
1429 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1430 Thomas Preud'homme <thomas.preudhomme@arm.com>
1432 * config/arm/arm-protos.h (clear_operation_p): Adapt prototype.
1433 * config/arm/arm.c (clear_operation_p): Extend to be able to check a
1434 clear_vfp_multiple pattern based on a new vfp parameter.
1435 (cmse_clear_registers): Generate VSCCLRM to clear VFP registers when
1436 targeting Armv8.1-M Mainline.
1437 (cmse_nonsecure_entry_clear_before_return): Clear VFP registers
1438 unconditionally when targeting Armv8.1-M Mainline architecture. Check
1439 whether VFP registers are available before looking call_used_regs for a
1441 * config/arm/predicates.md (clear_multiple_operation): Adapt to change
1442 of prototype of clear_operation_p.
1443 (clear_vfp_multiple_operation): New predicate.
1444 * config/arm/unspecs.md (VUNSPEC_VSCCLRM_VPR): New volatile unspec.
1445 * config/arm/vfp.md (clear_vfp_multiple): New define_insn.
1447 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1448 Thomas Preud'homme <thomas.preudhomme@arm.com>
1450 * config/arm/arm-protos.h (clear_operation_p): Declare.
1451 * config/arm/arm.c (clear_operation_p): New function.
1452 (cmse_clear_registers): Generate clear_multiple instruction pattern if
1453 targeting Armv8.1-M Mainline or successor.
1454 (output_return_instruction): Only output APSR register clearing if
1455 Armv8.1-M Mainline instructions not available.
1456 (thumb_exit): Likewise.
1457 * config/arm/predicates.md (clear_multiple_operation): New predicate.
1458 * config/arm/thumb2.md (clear_apsr): New define_insn.
1459 (clear_multiple): Likewise.
1460 * config/arm/unspecs.md (VUNSPEC_CLRM_APSR): New volatile unspec.
1462 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1463 Thomas Preud'homme <thomas.preudhomme@arm.com>
1465 * config/arm/arm.c (fp_sysreg_names): Declare and define.
1466 (use_return_insn): Also return false for Armv8.1-M Mainline.
1467 (output_return_instruction): Skip FPSCR clearing if Armv8.1-M
1468 Mainline instructions are available.
1469 (arm_compute_frame_layout): Allocate space in frame for FPCXTNS
1470 when targeting Armv8.1-M Mainline Security Extensions.
1471 (arm_expand_prologue): Save FPCXTNS if this is an Armv8.1-M
1472 Mainline entry function.
1473 (cmse_nonsecure_entry_clear_before_return): Clear IP and r4 if
1474 targeting Armv8.1-M Mainline or successor.
1475 (arm_expand_epilogue): Fix indentation of caller-saved register
1476 clearing. Restore FPCXTNS if this is an Armv8.1-M Mainline
1478 * config/arm/arm.h (TARGET_HAVE_FP_CMSE): New macro.
1479 (FP_SYSREGS): Likewise.
1480 (enum vfp_sysregs_encoding): Define enum.
1481 (fp_sysreg_names): Declare.
1482 * config/arm/unspecs.md (VUNSPEC_VSTR_VLDR): New volatile unspec.
1483 * config/arm/vfp.md (push_fpsysreg_insn): New define_insn.
1484 (pop_fpsysreg_insn): Likewise.
1486 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1487 Thomas Preud'homme <thomas.preudhomme@arm.com>
1489 * config/arm/arm-cpus.in (armv8_1m_main): New feature.
1490 (ARMv4, ARMv4t, ARMv5t, ARMv5te, ARMv5tej, ARMv6, ARMv6j, ARMv6k,
1491 ARMv6z, ARMv6kz, ARMv6zk, ARMv6t2, ARMv6m, ARMv7, ARMv7a, ARMv7ve,
1492 ARMv7r, ARMv7m, ARMv7em, ARMv8a, ARMv8_1a, ARMv8_2a, ARMv8_3a,
1493 ARMv8_4a, ARMv8_5a, ARMv8m_base, ARMv8m_main, ARMv8r): Reindent.
1494 (ARMv8_1m_main): New feature group.
1495 (armv8.1-m.main): New architecture.
1496 * config/arm/arm-tables.opt: Regenerate.
1497 * config/arm/arm.c (arm_arch8_1m_main): Define and default initialize.
1498 (arm_option_reconfigure_globals): Initialize arm_arch8_1m_main.
1499 (arm_options_perform_arch_sanity_checks): Error out when targeting
1500 Armv8.1-M Mainline Security Extensions.
1501 * config/arm/arm.h (arm_arch8_1m_main): Declare.
1503 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
1505 * config/aarch64/aarch64-simd-builtins.def (aarch64_bfdot,
1506 aarch64_bfdot_lane, aarch64_bfdot_laneq): New.
1507 * config/aarch64/aarch64-simd.md (aarch64_bfdot, aarch64_bfdot_lane,
1508 aarch64_bfdot_laneq): New.
1509 * config/aarch64/arm_bf16.h (vbfdot_f32, vbfdotq_f32,
1510 vbfdot_lane_f32, vbfdotq_lane_f32, vbfdot_laneq_f32,
1511 vbfdotq_laneq_f32): New.
1512 * config/aarch64/iterators.md (UNSPEC_BFDOT, Vbfdottype,
1513 VBFMLA_W, VBF): New.
1514 (isquadop): Add V4BF, V8BF.
1516 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
1518 * config/aarch64/aarch64-builtins.c: (enum aarch64_type_qualifiers):
1519 New qualifier_lane_quadtup_index, TYPES_TERNOP_SSUS,
1520 TYPES_QUADOPSSUS_LANE_QUADTUP, TYPES_QUADOPSSSU_LANE_QUADTUP.
1521 (aarch64_simd_expand_args): Add case SIMD_ARG_LANE_QUADTUP_INDEX.
1522 (aarch64_simd_expand_builtin): Add qualifier_lane_quadtup_index.
1523 * config/aarch64/aarch64-simd-builtins.def (usdot, usdot_lane,
1524 usdot_laneq, sudot_lane,sudot_laneq): New.
1525 * config/aarch64/aarch64-simd.md (aarch64_usdot): New.
1526 (aarch64_<sur>dot_lane): New.
1527 * config/aarch64/arm_neon.h (vusdot_s32): New.
1529 (vusdot_lane_s32): New.
1530 (vsudot_lane_s32): New.
1531 * config/aarch64/iterators.md (DOTPROD_I8MM): New iterator.
1532 (UNSPEC_USDOT, UNSPEC_SUDOT): New unspecs.
1534 2020-01-16 Martin Liska <mliska@suse.cz>
1536 * value-prof.c (dump_histogram_value): Fix
1537 obvious spacing issue.
1539 2020-01-16 Andrew Pinski <apinski@marvell.com>
1541 * tree-ssa-sccvn.c(vn_reference_lookup_3): Check lhs for
1542 !storage_order_barrier_p.
1544 2020-01-16 Andrew Pinski <apinski@marvell.com>
1546 * sched-int.h (_dep): Add unused bit-field field for the padding.
1547 * sched-deps.c (init_dep_1): Init unused field.
1549 2020-01-16 Andrew Pinski <apinski@marvell.com>
1551 * optabs.h (create_expand_operand): Initialize target field also.
1553 2020-01-16 Andre Vieira <andre.simoesdiasvieira@arm.com>
1555 PR tree-optimization/92429
1556 * tree-ssa-loop-niter.h (simplify_replace_tree): Add parameter.
1557 * tree-ssa-loop-niter.c (simplify_replace_tree): Add parameter to
1559 * tree-vect-loop.c (update_epilogue_vinfo): Do not fold when replacing
1562 2020-01-16 Richard Sandiford <richard.sandiford@arm.com>
1564 * config/aarch64/aarch64.c (aarch64_split_sve_subreg_move): Apply
1565 aarch64_sve_int_mode to each mode.
1567 2020-01-15 David Malcolm <dmalcolm@redhat.com>
1569 * doc/analyzer.texi (Overview): Add note about
1570 -fdump-ipa-analyzer.
1572 2020-01-15 Wilco Dijkstra <wdijkstr@arm.com>
1574 PR tree-optimization/93231
1575 * tree-ssa-forwprop.c (optimize_count_trailing_zeroes): Check
1576 input_type is unsigned. Use tree_to_shwi for shift constant.
1577 Check CST_STRING element size is CHAR_TYPE_SIZE bits.
1578 (simplify_count_trailing_zeroes): Add test to handle known non-zero
1579 inputs more efficiently.
1581 2020-01-15 Uroš Bizjak <ubizjak@gmail.com>
1583 * config/i386/i386.md (*movsf_internal): Do not require
1584 SSE2 ISA for alternatives 14 and 15.
1586 2020-01-15 Richard Biener <rguenther@suse.de>
1589 * tree-eh.c (sink_clobbers): If we already visited the destination
1590 block do not defer insertion.
1591 (pass_lower_eh_dispatch::execute): Maintain BB_VISITED for
1592 the purpose of defered insertion.
1594 2020-01-15 Jakub Jelinek <jakub@redhat.com>
1596 * BASE-VER: Bump to 10.0.1.
1598 2020-01-15 Richard Sandiford <richard.sandiford@arm.com>
1600 PR tree-optimization/93247
1601 * tree-vect-loop.c (update_epilogue_loop_vinfo): Check the access
1602 type of the stmt that we're going to vectorize.
1604 2020-01-15 Richard Sandiford <richard.sandiford@arm.com>
1606 * tree-vect-slp.c (vectorize_slp_instance_root_stmt): Use a
1607 VIEW_CONVERT_EXPR if the vectorized constructor has a diffeent
1610 2020-01-15 Martin Liska <mliska@suse.cz>
1612 * ipa-profile.c (ipa_profile_read_edge_summary): Do not allow
1613 2 calls of streamer_read_hwi in a function call.
1615 2020-01-15 Richard Biener <rguenther@suse.de>
1617 * alias.c (record_alias_subset): Avoid redundant work when
1618 subset is already recorded.
1620 2020-01-14 David Malcolm <dmalcolm@redhat.com>
1622 * doc/invoke.texi (-fdiagnostics-show-cwe): Add note that some of
1623 the analyzer options provide CWE identifiers.
1625 2020-01-14 David Malcolm <dmalcolm@redhat.com>
1627 * tree-diagnostic-path.cc (path_summary::event_range::print):
1628 When testing for UNKNOWN_LOCATION, look through ad-hoc wrappers
1629 using get_pure_location.
1631 2020-01-15 Jakub Jelinek <jakub@redhat.com>
1633 PR tree-optimization/93262
1634 * tree-ssa-dse.c (maybe_trim_memstar_call): For *_chk builtins,
1635 perform head trimming only if the last argument is constant,
1636 either all ones, or larger or equal to head trim, in the latter
1637 case decrease the last argument by head_trim.
1639 PR tree-optimization/93249
1640 * tree-ssa-dse.c: Include builtins.h and gimple-fold.h.
1641 (maybe_trim_memstar_call): Move head_trim and tail_trim vars to
1642 function body scope, reindent. For BUILTIN_IN_STRNCPY*, don't
1643 perform head trim unless we can prove there are no '\0' chars
1644 from the source among the first head_trim chars.
1646 2020-01-14 David Malcolm <dmalcolm@redhat.com>
1648 * Makefile.in (ANALYZER_OBJS): Add analyzer/function-set.o.
1650 2020-01-15 Jakub Jelinek <jakub@redhat.com>
1653 * config/i386/sse.md
1654 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1,
1655 *<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1,
1656 *<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1,
1657 *<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1): Use
1658 just a single alternative instead of two, make operands 1 and 2
1661 2020-01-14 Jan Hubicka <hubicka@ucw.cz>
1664 * ipa-devirt.c (odr_types_equivalent_p): Compare TREE_ADDRESSABLE and
1667 2020-01-14 David Malcolm <dmalcolm@redhat.com>
1669 * Makefile.in (lang_opt_files): Add analyzer.opt.
1670 (ANALYZER_OBJS): New.
1671 (OBJS): Add digraph.o, graphviz.o, ordered-hash-map-tests.o,
1672 tristate.o and ANALYZER_OBJS.
1673 (TEXI_GCCINT_FILES): Add analyzer.texi.
1674 * common.opt (-fanalyzer): New driver option.
1675 * config.in: Regenerate.
1676 * configure: Regenerate.
1677 * configure.ac (--disable-analyzer, ENABLE_ANALYZER): New option.
1678 (gccdepdir): Also create depdir for "analyzer" subdir.
1679 * digraph.cc: New file.
1680 * digraph.h: New file.
1681 * doc/analyzer.texi: New file.
1682 * doc/gccint.texi ("Static Analyzer") New menu item.
1683 (analyzer.texi): Include it.
1684 * doc/invoke.texi ("Static Analyzer Options"): New list and new section.
1685 ("Warning Options"): Add static analysis warnings to the list.
1686 (-Wno-analyzer-double-fclose): New option.
1687 (-Wno-analyzer-double-free): New option.
1688 (-Wno-analyzer-exposure-through-output-file): New option.
1689 (-Wno-analyzer-file-leak): New option.
1690 (-Wno-analyzer-free-of-non-heap): New option.
1691 (-Wno-analyzer-malloc-leak): New option.
1692 (-Wno-analyzer-possible-null-argument): New option.
1693 (-Wno-analyzer-possible-null-dereference): New option.
1694 (-Wno-analyzer-null-argument): New option.
1695 (-Wno-analyzer-null-dereference): New option.
1696 (-Wno-analyzer-stale-setjmp-buffer): New option.
1697 (-Wno-analyzer-tainted-array-index): New option.
1698 (-Wno-analyzer-use-after-free): New option.
1699 (-Wno-analyzer-use-of-pointer-in-stale-stack-frame): New option.
1700 (-Wno-analyzer-use-of-uninitialized-value): New option.
1701 (-Wanalyzer-too-complex): New option.
1702 (-fanalyzer-call-summaries): New warning.
1703 (-fanalyzer-checker=): New warning.
1704 (-fanalyzer-fine-grained): New warning.
1705 (-fno-analyzer-state-merge): New warning.
1706 (-fno-analyzer-state-purge): New warning.
1707 (-fanalyzer-transitivity): New warning.
1708 (-fanalyzer-verbose-edges): New warning.
1709 (-fanalyzer-verbose-state-changes): New warning.
1710 (-fanalyzer-verbosity=): New warning.
1711 (-fdump-analyzer): New warning.
1712 (-fdump-analyzer-callgraph): New warning.
1713 (-fdump-analyzer-exploded-graph): New warning.
1714 (-fdump-analyzer-exploded-nodes): New warning.
1715 (-fdump-analyzer-exploded-nodes-2): New warning.
1716 (-fdump-analyzer-exploded-nodes-3): New warning.
1717 (-fdump-analyzer-supergraph): New warning.
1718 * doc/sourcebuild.texi (dg-require-dot): New.
1719 (dg-check-dot): New.
1720 * gdbinit.in (break-on-saved-diagnostic): New command.
1721 * graphviz.cc: New file.
1722 * graphviz.h: New file.
1723 * ordered-hash-map-tests.cc: New file.
1724 * ordered-hash-map.h: New file.
1725 * passes.def (pass_analyzer): Add before
1726 pass_ipa_whole_program_visibility.
1727 * selftest-run-tests.c (selftest::run_tests): Call
1728 selftest::ordered_hash_map_tests_cc_tests.
1729 * selftest.h (selftest::ordered_hash_map_tests_cc_tests): New
1731 * shortest-paths.h: New file.
1732 * timevar.def (TV_ANALYZER): New timevar.
1733 (TV_ANALYZER_SUPERGRAPH): Likewise.
1734 (TV_ANALYZER_STATE_PURGE): Likewise.
1735 (TV_ANALYZER_PLAN): Likewise.
1736 (TV_ANALYZER_SCC): Likewise.
1737 (TV_ANALYZER_WORKLIST): Likewise.
1738 (TV_ANALYZER_DUMP): Likewise.
1739 (TV_ANALYZER_DIAGNOSTICS): Likewise.
1740 (TV_ANALYZER_SHORTEST_PATHS): Likewise.
1741 * tree-pass.h (make_pass_analyzer): New decl.
1742 * tristate.cc: New file.
1743 * tristate.h: New file.
1745 2020-01-14 Uroš Bizjak <ubizjak@gmail.com>
1748 * config/i386/i386.md (*movsf_internal): Require SSE2 ISA for
1749 alternatives 9 and 10.
1751 2020-01-14 David Malcolm <dmalcolm@redhat.com>
1753 * attribs.c (excl_hash_traits::empty_zero_p): New static constant.
1754 * gcov.c (function_start_pair_hash::empty_zero_p): Likewise.
1755 * graphite.c (struct sese_scev_hash::empty_zero_p): Likewise.
1756 * hash-map-tests.c (selftest::test_nonzero_empty_key): New selftest.
1757 (selftest::hash_map_tests_c_tests): Call it.
1758 * hash-map-traits.h (simple_hashmap_traits::empty_zero_p):
1759 New static constant, using the value of = H::empty_zero_p.
1760 (unbounded_hashmap_traits::empty_zero_p): Likewise, using the value
1761 from default_hash_traits <Value>.
1762 * hash-map.h (hash_map::empty_zero_p): Likewise, using the value
1764 * hash-set-tests.c (value_hash_traits::empty_zero_p): Likewise.
1765 * hash-table.h (hash_table::alloc_entries): Guard the loop of
1766 calls to mark_empty with !Descriptor::empty_zero_p.
1767 (hash_table::empty_slow): Conditionalize the memset call with a
1768 check that Descriptor::empty_zero_p; otherwise, loop through the
1769 entries calling mark_empty on them.
1770 * hash-traits.h (int_hash::empty_zero_p): New static constant.
1771 (pointer_hash::empty_zero_p): Likewise.
1772 (pair_hash::empty_zero_p): Likewise.
1773 * ipa-devirt.c (default_hash_traits <type_pair>::empty_zero_p):
1775 * ipa-prop.c (ipa_bit_ggc_hash_traits::empty_zero_p): Likewise.
1776 (ipa_vr_ggc_hash_traits::empty_zero_p): Likewise.
1777 * profile.c (location_triplet_hash::empty_zero_p): Likewise.
1778 * sanopt.c (sanopt_tree_triplet_hash::empty_zero_p): Likewise.
1779 (sanopt_tree_couple_hash::empty_zero_p): Likewise.
1780 * tree-hasher.h (int_tree_hasher::empty_zero_p): Likewise.
1781 * tree-ssa-sccvn.c (vn_ssa_aux_hasher::empty_zero_p): Likewise.
1782 * tree-vect-slp.c (bst_traits::empty_zero_p): Likewise.
1784 (default_hash_traits<scalar_cond_masked_key>::empty_zero_p):
1787 2020-01-14 Kewen Lin <linkw@gcc.gnu.org>
1789 * cfgloopanal.c (average_num_loop_insns): Free bbs when early return,
1790 fix typo on return value.
1792 2020-01-14 Xiong Hu Luo <luoxhu@linux.ibm.com>
1795 * cgraph.c (symbol_table::create_edge): Init speculative_id and
1797 (cgraph_edge::make_speculative): Add param for setting speculative_id
1799 (cgraph_edge::speculative_call_info): Update comments and find reference
1800 by speculative_id for multiple indirect targets.
1801 (cgraph_edge::resolve_speculation): Decrease the speculations
1802 for indirect edge, drop it's speculative if not direct target
1803 left. Update comments.
1804 (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
1805 (cgraph_node::dump): Print num_speculative_call_targets.
1806 (cgraph_node::verify_node): Don't report error if speculative
1807 edge not include statement.
1808 (cgraph_edge::num_speculative_call_targets_p): New function.
1809 * cgraph.h (int common_target_id): Remove.
1810 (int common_target_probability): Remove.
1811 (num_speculative_call_targets): New variable.
1812 (make_speculative): Add param for setting speculative_id.
1813 (cgraph_edge::num_speculative_call_targets_p): New declare.
1814 (target_prob): New variable.
1815 (speculative_id): New variable.
1816 * ipa-fnsummary.c (analyze_function_body): Create and duplicate
1817 call summaries for multiple speculative call targets.
1818 * cgraphclones.c (cgraph_node::create_clone): Clone speculative_id.
1819 * ipa-profile.c (struct speculative_call_target): New struct.
1820 (class speculative_call_summary): New class.
1821 (class speculative_call_summaries): New class.
1822 (call_sums): New variable.
1823 (ipa_profile_generate_summary): Generate indirect multiple targets summaries.
1824 (ipa_profile_write_edge_summary): New function.
1825 (ipa_profile_write_summary): Stream out indirect multiple targets summaries.
1826 (ipa_profile_dump_all_summaries): New function.
1827 (ipa_profile_read_edge_summary): New function.
1828 (ipa_profile_read_summary_section): New function.
1829 (ipa_profile_read_summary): Stream in indirect multiple targets summaries.
1830 (ipa_profile): Generate num_speculative_call_targets from
1832 * ipa-ref.h (speculative_id): New variable.
1833 * ipa-utils.c (ipa_merge_profiles): Update with target_prob.
1834 * lto-cgraph.c (lto_output_edge): Remove indirect common_target_id and
1835 common_target_probability. Stream out speculative_id and
1836 num_speculative_call_targets.
1837 (input_edge): Likewise.
1838 * predict.c (dump_prediction): Remove edges count assert to be
1840 * symtab.c (symtab_node::create_reference): Init speculative_id.
1841 (symtab_node::clone_references): Clone speculative_id.
1842 (symtab_node::clone_referring): Clone speculative_id.
1843 (symtab_node::clone_reference): Clone speculative_id.
1844 (symtab_node::clear_stmts_in_references): Clear speculative_id.
1845 * tree-inline.c (copy_bb): Duplicate all the speculative edges
1846 if indirect call contains multiple speculative targets.
1847 * value-prof.h (check_ic_target): Remove.
1848 * value-prof.c (gimple_value_profile_transformations):
1849 Use void function gimple_ic_transform.
1850 * value-prof.c (gimple_ic_transform): Handle topn case.
1851 Fix comment typos. Change it to a void function.
1853 2020-01-13 Andrew Pinski <apinski@marvell.com>
1855 * config/aarch64/aarch64-cores.def (octeontx2): New define.
1856 (octeontx2t98): New define.
1857 (octeontx2t96): New define.
1858 (octeontx2t93): New define.
1859 (octeontx2f95): New define.
1860 (octeontx2f95n): New define.
1861 (octeontx2f95mm): New define.
1862 * config/aarch64/aarch64-tune.md: Regenerate.
1863 * doc/invoke.texi (-mcpu=): Document the new cpu types.
1865 2020-01-13 Jason Merrill <jason@redhat.com>
1867 PR c++/33799 - destroy return value if local cleanup throws.
1868 * gimplify.c (gimplify_return_expr): Handle COMPOUND_EXPR.
1870 2020-01-13 Martin Liska <mliska@suse.cz>
1872 * ipa-cp.c (get_max_overall_size): Use newly
1873 renamed param param_ipa_cp_unit_growth.
1874 * params.opt: Remove legacy param name.
1876 2020-01-13 Martin Sebor <msebor@redhat.com>
1878 PR tree-optimization/93213
1879 * tree-ssa-strlen.c (handle_store): Only allow single-byte nul-over-nul
1880 stores to be eliminated.
1882 2020-01-13 Martin Liska <mliska@suse.cz>
1884 * opts.c (print_help): Do not print CL_PARAM
1885 and CL_WARNING for CL_OPTIMIZATION.
1887 2020-01-13 Jonathan Wakely <jwakely@redhat.com>
1890 * doc/invoke.texi (Warning Options): Add caveat about some warnings
1891 depending on optimization settings.
1893 2020-01-13 Jakub Jelinek <jakub@redhat.com>
1895 PR tree-optimization/90838
1896 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
1897 SCALAR_INT_TYPE_MODE directly in CTZ_DEFINED_VALUE_AT_ZERO macro
1898 argument rather than to initialize temporary for targets that
1899 don't use the mode argument at all. Initialize ctzval to avoid
1902 2020-01-10 Thomas Schwinge <thomas@codesourcery.com>
1904 * tree.h (OMP_CLAUSE_USE_DEVICE_PTR_IF_PRESENT): New definition.
1905 * tree-core.h: Document it.
1906 * gimplify.c (gimplify_omp_workshare): Set it.
1907 * omp-low.c (lower_omp_target): Use it.
1908 * tree-pretty-print.c (dump_omp_clause): Print it.
1910 * omp-low.c (lower_omp_target) <OMP_CLAUSE_USE_DEVICE_PTR etc.>:
1911 Assert that for OpenACC we always have 'GOMP_MAP_USE_DEVICE_PTR'.
1913 2020-01-10 David Malcolm <dmalcolm@redhat.com>
1915 * Makefile.in (OBJS): Add tree-diagnostic-path.o.
1916 * common.opt (fdiagnostics-path-format=): New option.
1917 (diagnostic_path_format): New enum.
1918 (fdiagnostics-show-path-depths): New option.
1919 * coretypes.h (diagnostic_event_id_t): New forward decl.
1920 * diagnostic-color.c (color_dict): Add "path".
1921 * diagnostic-event-id.h: New file.
1922 * diagnostic-format-json.cc (json_from_expanded_location): Make
1924 (json_end_diagnostic): Call context->make_json_for_path if it
1925 exists and the diagnostic has a path.
1926 (diagnostic_output_format_init): Clear context->print_path.
1927 * diagnostic-path.h: New file.
1928 * diagnostic-show-locus.c (colorizer::set_range): Special-case
1929 when printing a run of events in a diagnostic_path so that they
1930 all get the same color.
1931 (layout::m_diagnostic_path_p): New field.
1932 (layout::layout): Initialize it.
1933 (layout::print_any_labels): Don't colorize the label text for an
1934 event in a diagnostic_path.
1935 (gcc_rich_location::add_location_if_nearby): Add
1936 "restrict_to_current_line_spans" and "label" params. Pass the
1937 former to layout.maybe_add_location_range; pass the latter
1938 when calling add_range.
1939 * diagnostic.c: Include "diagnostic-path.h".
1940 (diagnostic_initialize): Initialize context->path_format and
1941 context->show_path_depths.
1942 (diagnostic_show_any_path): New function.
1943 (diagnostic_path::interprocedural_p): New function.
1944 (diagnostic_report_diagnostic): Call diagnostic_show_any_path.
1945 (simple_diagnostic_path::num_events): New function.
1946 (simple_diagnostic_path::get_event): New function.
1947 (simple_diagnostic_path::add_event): New function.
1948 (simple_diagnostic_event::simple_diagnostic_event): New ctor.
1949 (simple_diagnostic_event::~simple_diagnostic_event): New dtor.
1950 (debug): New overload taking a diagnostic_path *.
1951 * diagnostic.def (DK_DIAGNOSTIC_PATH): New.
1952 * diagnostic.h (enum diagnostic_path_format): New enum.
1953 (json::value): New forward decl.
1954 (diagnostic_context::path_format): New field.
1955 (diagnostic_context::show_path_depths): New field.
1956 (diagnostic_context::print_path): New callback field.
1957 (diagnostic_context::make_json_for_path): New callback field.
1958 (diagnostic_show_any_path): New decl.
1959 (json_from_expanded_location): New decl.
1960 * doc/invoke.texi (-fdiagnostics-path-format=): New option.
1961 (-fdiagnostics-show-path-depths): New option.
1962 (-fdiagnostics-color): Add "path" to description of default
1963 GCC_COLORS; describe it.
1964 (-fdiagnostics-format=json): Document how diagnostic paths are
1965 represented in the JSON output format.
1966 * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
1967 Add optional params "restrict_to_current_line_spans" and "label".
1968 * opts.c (common_handle_option): Handle
1969 OPT_fdiagnostics_path_format_ and
1970 OPT_fdiagnostics_show_path_depths.
1971 * pretty-print.c: Include "diagnostic-event-id.h".
1972 (pp_format): Implement "%@" format code for printing
1973 diagnostic_event_id_t *.
1974 (selftest::test_pp_format): Add tests for "%@".
1975 * selftest-run-tests.c (selftest::run_tests): Call
1976 selftest::tree_diagnostic_path_cc_tests.
1977 * selftest.h (selftest::tree_diagnostic_path_cc_tests): New decl.
1978 * toplev.c (general_init): Initialize global_dc->path_format and
1979 global_dc->show_path_depths.
1980 * tree-diagnostic-path.cc: New file.
1981 * tree-diagnostic.c (maybe_unwind_expanded_macro_loc): Make
1982 non-static. Drop "diagnostic" param in favor of storing the
1983 original value of "where" and re-using it.
1984 (virt_loc_aware_diagnostic_finalizer): Update for dropped param of
1985 maybe_unwind_expanded_macro_loc.
1986 (tree_diagnostics_defaults): Initialize context->print_path and
1987 context->make_json_for_path.
1988 * tree-diagnostic.h (default_tree_diagnostic_path_printer): New
1990 (default_tree_make_json_for_path): New decl.
1991 (maybe_unwind_expanded_macro_loc): New decl.
1993 2020-01-10 Jakub Jelinek <jakub@redhat.com>
1995 PR tree-optimization/93210
1996 * fold-const.h (native_encode_initializer,
1997 can_native_interpret_type_p): Declare.
1998 * fold-const.c (native_encode_string): Fix up handling with off != -1,
2000 (native_encode_initializer): New function, moved from dwarf2out.c.
2001 Adjust to native_encode_expr compatible arguments, including dry-run
2002 and partial extraction modes. Don't handle STRING_CST.
2003 (can_native_interpret_type_p): No longer static.
2004 * gimple-fold.c (fold_ctor_reference): For native_encode_expr, verify
2005 offset / BITS_PER_UNIT fits into int and don't call it if
2006 can_native_interpret_type_p fails. If suboff is NULL and for
2007 CONSTRUCTOR fold_{,non}array_ctor_reference returns NULL, retry with
2008 native_encode_initializer.
2009 (fold_const_aggregate_ref_1): Formatting fix.
2010 * dwarf2out.c (native_encode_initializer): Moved to fold-const.c.
2011 (tree_add_const_value_attribute): Adjust caller.
2013 PR tree-optimization/90838
2014 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
2015 SCALAR_INT_TYPE_MODE instead of TYPE_MODE as operand of
2016 CTZ_DEFINED_VALUE_AT_ZERO.
2018 2020-01-10 Vladimir Makarov <vmakarov@redhat.com>
2021 * lra-constraints.c (match_reload): Permit input operands have the
2022 same mode as output while other input operands have a different
2025 2020-01-10 Wilco Dijkstra <wdijkstr@arm.com>
2027 PR tree-optimization/90838
2028 * tree-ssa-forwprop.c (check_ctz_array): Add new function.
2029 (check_ctz_string): Likewise.
2030 (optimize_count_trailing_zeroes): Likewise.
2031 (simplify_count_trailing_zeroes): Likewise.
2032 (pass_forwprop::execute): Try ctz simplification.
2033 * match.pd: Add matching for ctz idioms.
2035 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2037 * config/aarch64/aarch64.c (aarch64_invalid_conversion): New function
2039 (aarch64_invalid_unary_op): New function for target hook.
2040 (aarch64_invalid_binary_op): New function for target hook.
2042 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2044 * config.gcc: Add arm_bf16.h.
2045 * config/aarch64/aarch64-builtins.c
2046 (aarch64_simd_builtin_std_type): Add BFmode.
2047 (aarch64_init_simd_builtin_types): Define element types for vector
2049 (aarch64_init_bf16_types): New function.
2050 (aarch64_general_init_builtins): Add arm_init_bf16_types function call.
2051 * config/aarch64/aarch64-modes.def: Add BFmode and V4BF, V8BF vector
2053 * config/aarch64/aarch64-simd-builtin-types.def: Add BF SIMD types.
2054 * config/aarch64/aarch64-simd.md: Add BF vector types to NEON move
2056 * config/aarch64/aarch64.h (AARCH64_VALID_SIMD_DREG_MODE): Add V4BF.
2057 (AARCH64_VALID_SIMD_QREG_MODE): Add V8BF.
2058 * config/aarch64/aarch64.c
2059 (aarch64_classify_vector_mode): Add support for BF types.
2060 (aarch64_gimplify_va_arg_expr): Add support for BF types.
2061 (aarch64_vq_mode): Add support for BF types.
2062 (aarch64_simd_container_mode): Add support for BF types.
2063 (aarch64_mangle_type): Add support for BF scalar type.
2064 * config/aarch64/aarch64.md: Add BFmode to movhf pattern.
2065 * config/aarch64/arm_bf16.h: New file.
2066 * config/aarch64/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
2067 * config/aarch64/iterators.md: Add BF types to mode attributes.
2068 (HFBF, GPF_TF_F16_MOV, VDMOV, VQMOV, VQMOV_NO2Em VALL_F16MOV): New.
2070 2020-01-10 Jason Merrill <jason@redhat.com>
2072 PR c++/93173 - incorrect tree sharing.
2073 * gimplify.c (copy_if_shared): No longer static.
2074 * gimplify.h: Declare it.
2076 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
2078 * doc/invoke.texi (-msve-vector-bits=): Document that
2079 -msve-vector-bits=128 now generates VL-specific code for
2080 little-endian targets.
2081 * config/aarch64/aarch64-sve-builtins.cc (register_builtin_types): Use
2082 build_vector_type_for_mode to construct the data vector types.
2083 * config/aarch64/aarch64.c (aarch64_convert_sve_vector_bits): Generate
2084 VL-specific code for -msve-vector-bits=128 on little-endian targets.
2085 (aarch64_simd_container_mode): Always prefer Advanced SIMD modes
2086 for 128-bit vectors.
2088 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
2090 * config/aarch64/aarch64.c (aarch64_evpc_sel): Fix gen_vcond_mask
2093 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
2095 * config/aarch64/aarch64-builtins.c
2096 (aarch64_builtin_vectorized_function): Check for specific vector modes,
2097 rather than checking the number of elements and the element mode.
2099 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
2101 * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
2102 get_related_vectype_for_scalar_type rather than build_vector_type
2103 to create the index type for a conditional reduction.
2105 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
2107 * tree-vect-loop.c (update_epilogue_loop_vinfo): Update DR_REF
2108 for any type of gather or scatter, including strided accesses.
2110 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
2112 * tree-vectorizer.h (get_dr_vinfo_offset): Add missing function
2115 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
2117 * tree-vect-data-refs.c (vect_create_addr_base_for_vector_ref): Use
2119 * tree-vect-loop.c (update_epilogue_loop_vinfo): Remove orig_drs_init
2120 parameter and its use to reset DR_OFFSET's.
2121 (vect_transform_loop): Remove orig_drs_init argument.
2122 * tree-vect-loop-manip.c (vect_update_init_of_dr): Update the offset
2123 member of dr_vec_info rather than the offset of the associated
2124 data_reference's innermost_loop_behavior.
2125 (vect_update_init_of_dr): Pass dr_vec_info instead of data_reference.
2126 (vect_do_peeling): Remove orig_drs_init parameter and its construction.
2127 * tree-vect-stmts.c (check_scan_store): Replace use of DR_OFFSET with
2128 get_dr_vinfo_offset.
2129 (vectorizable_store): Likewise.
2130 (vectorizable_load): Likewise.
2132 2020-01-10 Richard Biener <rguenther@suse.de>
2134 * gimple-ssa-store-merging
2135 (pass_store_merging::terminate_all_aliasing_chains): Cache alias info.
2137 2020-01-10 Martin Liska <mliska@suse.cz>
2140 * ipa-inline-analysis.c (offline_size): Make proper parenthesis
2141 encapsulation that was there before r280040.
2143 2020-01-10 Richard Biener <rguenther@suse.de>
2146 * tree-eh.c (sink_clobbers): Move clobbers to out-of-IL
2147 sequences to avoid walking them again for secondary opportunities.
2148 (pass_lower_eh_dispatch::execute): Instead actually insert
2151 2020-01-10 Richard Biener <rguenther@suse.de>
2154 * tree-eh.c (redirect_eh_edge_1): Avoid some work if possible.
2155 (cleanup_all_empty_eh): Walk landing pads in reverse order to
2156 avoid quadraticness.
2158 2020-01-10 Martin Jambor <mjambor@suse.cz>
2160 * params.opt (param_ipa_sra_max_replacements): Mark as Optimization.
2161 * ipa-sra.c (pull_accesses_from_callee): New parameter caller, use it
2162 to get param_ipa_sra_max_replacements.
2163 (param_splitting_across_edge): Pass the caller to
2164 pull_accesses_from_callee.
2166 2020-01-10 Martin Jambor <mjambor@suse.cz>
2168 * params.opt (param_ipcp_unit_growth): Mark as Optimization.
2169 * ipa-cp.c (max_new_size): Removed.
2170 (orig_overall_size): New variable.
2171 (get_max_overall_size): New function.
2172 (estimate_local_effects): Use it. Adjust dump.
2173 (decide_about_value): Likewise.
2174 (ipcp_propagate_stage): Do not calculate max_new_size, just store
2175 orig_overall_size. Adjust dump.
2176 (ipa_cp_c_finalize): Clear orig_overall_size instead of max_new_size.
2178 2020-01-10 Martin Jambor <mjambor@suse.cz>
2180 * params.opt (param_ipa_max_agg_items): Mark as Optimization
2181 * ipa-cp.c (merge_agg_lats_step): New parameter max_agg_items, use
2182 instead of param_ipa_max_agg_items.
2183 (merge_aggregate_lattices): Extract param_ipa_max_agg_items from
2184 optimization info for the callee.
2186 2020-01-09 Kwok Cheung Yeung <kcy@codesourcery.com>
2188 * lto-streamer-in.c (input_function): Remove streamed-in inline debug
2189 markers if debug_inline_points is false.
2191 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
2193 * config.gcc (aarch64*-*-*): Add aarch64-sve-builtins-sve2.o to
2195 * config/aarch64/t-aarch64 (aarch64-sve-builtins.o): Depend on
2196 aarch64-sve-builtins-base.def, aarch64-sve-builtins-sve2.def and
2197 aarch64-sve-builtins-sve2.h.
2198 (aarch64-sve-builtins-sve2.o): New rule.
2199 * config/aarch64/aarch64.h (AARCH64_ISA_SVE2_AES): New macro.
2200 (AARCH64_ISA_SVE2_BITPERM, AARCH64_ISA_SVE2_SHA3): Likewise.
2201 (AARCH64_ISA_SVE2_SM4, TARGET_SVE2_AES, TARGET_SVE2_BITPERM): Likewise.
2202 (TARGET_SVE2_SHA, TARGET_SVE2_SM4): Likewise.
2203 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
2204 TARGET_SVE2_AES, TARGET_SVE2_BITPERM, TARGET_SVE2_SHA3 and
2206 * config/aarch64/aarch64-sve.md: Update comments with SVE2
2207 instructions that are handled here.
2208 (@cond_asrd<mode>): Generalize to...
2209 (@cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>): ...this.
2210 (*cond_asrd<mode>_2): Generalize to...
2211 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_2): ...this.
2212 (*cond_asrd<mode>_z): Generalize to...
2213 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_z): ...this.
2214 * config/aarch64/aarch64.md (UNSPEC_LDNT1_GATHER): New unspec.
2215 (UNSPEC_STNT1_SCATTER, UNSPEC_WHILEGE, UNSPEC_WHILEGT): Likewise.
2216 (UNSPEC_WHILEHI, UNSPEC_WHILEHS): Likewise.
2217 * config/aarch64/aarch64-sve2.md (@aarch64_gather_ldnt<mode>): New
2219 (@aarch64_gather_ldnt_<ANY_EXTEND:optab><SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
2220 (@aarch64_scatter_stnt<mode>): Likewise.
2221 (@aarch64_scatter_stnt_<SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
2222 (@aarch64_mul_lane_<mode>): Likewise.
2223 (@aarch64_sve_suqadd<mode>_const): Likewise.
2224 (*<sur>h<addsub><mode>): Generalize to...
2225 (@aarch64_pred_<SVE2_COND_INT_BINARY_REV:sve_int_op><mode>): ...this
2227 (@cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>): New expander.
2228 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_2): New pattern.
2229 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_3): Likewise.
2230 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_any): Likewise.
2231 (*cond_<SVE2_COND_INT_BINARY_NOREV:sve_int_op><mode>_z): Likewise.
2232 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op><mode>):: Likewise.
2233 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op>_lane_<mode>): Likewise.
2234 (@aarch64_pred_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): Likewise.
2235 (@cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): New expander.
2236 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_2): New pattern.
2237 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_3): Likewise.
2238 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_any): Likewise.
2239 (@aarch64_sve_<SVE2_INT_TERNARY:sve_int_op><mode>): Likewise.
2240 (@aarch64_sve_<SVE2_INT_TERNARY_LANE:sve_int_op>_lane_<mode>)
2241 (@aarch64_sve_add_mul_lane_<mode>): Likewise.
2242 (@aarch64_sve_sub_mul_lane_<mode>): Likewise.
2243 (@aarch64_sve2_xar<mode>): Likewise.
2244 (@aarch64_sve2_bcax<mode>): Likewise.
2245 (*aarch64_sve2_eor3<mode>): Rename to...
2246 (@aarch64_sve2_eor3<mode>): ...this.
2247 (@aarch64_sve2_bsl<mode>): New expander.
2248 (@aarch64_sve2_nbsl<mode>): Likewise.
2249 (@aarch64_sve2_bsl1n<mode>): Likewise.
2250 (@aarch64_sve2_bsl2n<mode>): Likewise.
2251 (@aarch64_sve_add_<SHIFTRT:sve_int_op><mode>): Likewise.
2252 (*aarch64_sve2_sra<mode>): Add MOVPRFX support.
2253 (@aarch64_sve_add_<VRSHR_N:sve_int_op><mode>): New pattern.
2254 (@aarch64_sve_<SVE2_INT_SHIFT_INSERT:sve_int_op><mode>): Likewise.
2255 (@aarch64_sve2_<USMAX:su>aba<mode>): New expander.
2256 (*aarch64_sve2_<USMAX:su>aba<mode>): New pattern.
2257 (@aarch64_sve_<SVE2_INT_BINARY_WIDE:sve_int_op><mode>): Likewise.
2258 (<su>mull<bt><Vwide>): Generalize to...
2259 (@aarch64_sve_<SVE2_INT_BINARY_LONG:sve_int_op><mode>): ...this new
2261 (@aarch64_sve_<SVE2_INT_BINARY_LONG_lANE:sve_int_op>_lane_<mode>)
2262 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_LONG:sve_int_op><mode>)
2263 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG:sve_int_op><mode>)
2264 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
2265 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG:sve_int_op><mode>)
2266 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
2267 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG:sve_int_op><mode>)
2268 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
2269 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG:sve_int_op><mode>)
2270 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
2271 (@aarch64_sve_<SVE2_FP_TERNARY_LONG:sve_fp_op><mode>): New patterns.
2272 (@aarch64_<SVE2_FP_TERNARY_LONG_LANE:sve_fp_op>_lane_<mode>)
2273 (@aarch64_sve_<SVE2_INT_UNARY_NARROWB:sve_int_op><mode>): Likewise.
2274 (@aarch64_sve_<SVE2_INT_UNARY_NARROWT:sve_int_op><mode>): Likewise.
2275 (@aarch64_sve_<SVE2_INT_BINARY_NARROWB:sve_int_op><mode>): Likewise.
2276 (@aarch64_sve_<SVE2_INT_BINARY_NARROWT:sve_int_op><mode>): Likewise.
2277 (<SHRNB:r>shrnb<mode>): Generalize to...
2278 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWB:sve_int_op><mode>): ...this
2280 (<SHRNT:r>shrnt<mode>): Generalize to...
2281 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWT:sve_int_op><mode>): ...this
2283 (@aarch64_pred_<SVE2_INT_BINARY_PAIR:sve_int_op><mode>): New pattern.
2284 (@aarch64_pred_<SVE2_FP_BINARY_PAIR:sve_fp_op><mode>): Likewise.
2285 (@cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>): New expander.
2286 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_2): New pattern.
2287 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_z): Likewise.
2288 (@aarch64_sve_<SVE2_INT_CADD:optab><mode>): Likewise.
2289 (@aarch64_sve_<SVE2_INT_CMLA:optab><mode>): Likewise.
2290 (@aarch64_<SVE2_INT_CMLA:optab>_lane_<mode>): Likewise.
2291 (@aarch64_sve_<SVE2_INT_CDOT:optab><mode>): Likewise.
2292 (@aarch64_<SVE2_INT_CDOT:optab>_lane_<mode>): Likewise.
2293 (@aarch64_pred_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): Likewise.
2294 (@cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New expander.
2295 (*cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New pattern.
2296 (@aarch64_sve2_cvtnt<mode>): Likewise.
2297 (@aarch64_pred_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): Likewise.
2298 (@cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): New expander.
2299 (*cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>_any): New pattern.
2300 (@aarch64_sve2_cvtxnt<mode>): Likewise.
2301 (@aarch64_pred_<SVE2_U32_UNARY:sve_int_op><mode>): Likewise.
2302 (@cond_<SVE2_U32_UNARY:sve_int_op><mode>): New expander.
2303 (*cond_<SVE2_U32_UNARY:sve_int_op><mode>): New pattern.
2304 (@aarch64_pred_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): Likewise.
2305 (@cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New expander.
2306 (*cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New pattern.
2307 (@aarch64_sve2_pmul<mode>): Likewise.
2308 (@aarch64_sve_<SVE2_PMULL:optab><mode>): Likewise.
2309 (@aarch64_sve_<SVE2_PMULL_PAIR:optab><mode>): Likewise.
2310 (@aarch64_sve2_tbl2<mode>): Likewise.
2311 (@aarch64_sve2_tbx<mode>): Likewise.
2312 (@aarch64_sve_<SVE2_INT_BITPERM:sve_int_op><mode>): Likewise.
2313 (@aarch64_sve2_histcnt<mode>): Likewise.
2314 (@aarch64_sve2_histseg<mode>): Likewise.
2315 (@aarch64_pred_<SVE2_MATCH:sve_int_op><mode>): Likewise.
2316 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_cc): Likewise.
2317 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_ptest): Likewise.
2318 (aarch64_sve2_aes<CRYPTO_AES:aes_op>): Likewise.
2319 (aarch64_sve2_aes<CRYPTO_AESMC:aesmc_op>): Likewise.
2320 (*aarch64_sve2_aese_fused, *aarch64_sve2_aesd_fused): Likewise.
2321 (aarch64_sve2_rax1, aarch64_sve2_sm4e, aarch64_sve2_sm4ekey): Likewise.
2322 (<su>mulh<r>s<mode>3): Update after above pattern name changes.
2323 * config/aarch64/iterators.md (VNx16QI_ONLY, VNx4SF_ONLY)
2324 (SVE_STRUCT2, SVE_FULL_BHI, SVE_FULL_HSI, SVE_FULL_HDI)
2325 (SVE2_PMULL_PAIR_I): New mode iterators.
2326 (UNSPEC_ADCLB, UNSPEC_ADCLT, UNSPEC_ADDHNB, UNSPEC_ADDHNT, UNSPEC_BDEP)
2327 (UNSPEC_BEXT, UNSPEC_BGRP, UNSPEC_CADD90, UNSPEC_CADD270, UNSPEC_CDOT)
2328 (UNSPEC_CDOT90, UNSPEC_CDOT180, UNSPEC_CDOT270, UNSPEC_CMLA)
2329 (UNSPEC_CMLA90, UNSPEC_CMLA180, UNSPEC_CMLA270, UNSPEC_COND_FCVTLT)
2330 (UNSPEC_COND_FCVTNT, UNSPEC_COND_FCVTX, UNSPEC_COND_FCVTXNT)
2331 (UNSPEC_COND_FLOGB, UNSPEC_EORBT, UNSPEC_EORTB, UNSPEC_FADDP)
2332 (UNSPEC_FMAXP, UNSPEC_FMAXNMP, UNSPEC_FMLALB, UNSPEC_FMLALT)
2333 (UNSPEC_FMLSLB, UNSPEC_FMLSLT, UNSPEC_FMINP, UNSPEC_FMINNMP)
2334 (UNSPEC_HISTCNT, UNSPEC_HISTSEG, UNSPEC_MATCH, UNSPEC_NMATCH)
2335 (UNSPEC_PMULLB, UNSPEC_PMULLB_PAIR, UNSPEC_PMULLT, UNSPEC_PMULLT_PAIR)
2336 (UNSPEC_RADDHNB, UNSPEC_RADDHNT, UNSPEC_RSUBHNB, UNSPEC_RSUBHNT)
2337 (UNSPEC_SLI, UNSPEC_SRI, UNSPEC_SABDLB, UNSPEC_SABDLT, UNSPEC_SADDLB)
2338 (UNSPEC_SADDLBT, UNSPEC_SADDLT, UNSPEC_SADDWB, UNSPEC_SADDWT)
2339 (UNSPEC_SBCLB, UNSPEC_SBCLT, UNSPEC_SMAXP, UNSPEC_SMINP)
2340 (UNSPEC_SQCADD90, UNSPEC_SQCADD270, UNSPEC_SQDMULLB, UNSPEC_SQDMULLBT)
2341 (UNSPEC_SQDMULLT, UNSPEC_SQRDCMLAH, UNSPEC_SQRDCMLAH90)
2342 (UNSPEC_SQRDCMLAH180, UNSPEC_SQRDCMLAH270, UNSPEC_SQRSHRNB)
2343 (UNSPEC_SQRSHRNT, UNSPEC_SQRSHRUNB, UNSPEC_SQRSHRUNT, UNSPEC_SQSHRNB)
2344 (UNSPEC_SQSHRNT, UNSPEC_SQSHRUNB, UNSPEC_SQSHRUNT, UNSPEC_SQXTNB)
2345 (UNSPEC_SQXTNT, UNSPEC_SQXTUNB, UNSPEC_SQXTUNT, UNSPEC_SSHLLB)
2346 (UNSPEC_SSHLLT, UNSPEC_SSUBLB, UNSPEC_SSUBLBT, UNSPEC_SSUBLT)
2347 (UNSPEC_SSUBLTB, UNSPEC_SSUBWB, UNSPEC_SSUBWT, UNSPEC_SUBHNB)
2348 (UNSPEC_SUBHNT, UNSPEC_TBL2, UNSPEC_UABDLB, UNSPEC_UABDLT)
2349 (UNSPEC_UADDLB, UNSPEC_UADDLT, UNSPEC_UADDWB, UNSPEC_UADDWT)
2350 (UNSPEC_UMAXP, UNSPEC_UMINP, UNSPEC_UQRSHRNB, UNSPEC_UQRSHRNT)
2351 (UNSPEC_UQSHRNB, UNSPEC_UQSHRNT, UNSPEC_UQXTNB, UNSPEC_UQXTNT)
2352 (UNSPEC_USHLLB, UNSPEC_USHLLT, UNSPEC_USUBLB, UNSPEC_USUBLT)
2353 (UNSPEC_USUBWB, UNSPEC_USUBWT): New unspecs.
2354 (UNSPEC_SMULLB, UNSPEC_SMULLT, UNSPEC_UMULLB, UNSPEC_UMULLT)
2355 (UNSPEC_SMULHS, UNSPEC_SMULHRS, UNSPEC_UMULHS, UNSPEC_UMULHRS)
2356 (UNSPEC_RSHRNB, UNSPEC_RSHRNT, UNSPEC_SHRNB, UNSPEC_SHRNT): Move
2358 (VNARROW, Ventype): New mode attributes.
2359 (Vewtype): Handle VNx2DI. Fix typo in comment.
2360 (VDOUBLE): New mode attribute.
2361 (sve_lane_con): Handle VNx8HI.
2362 (SVE_INT_UNARY): Include ss_abs and ss_neg for TARGET_SVE2.
2363 (SVE_INT_BINARY): Likewise ss_plus, us_plus, ss_minus and us_minus.
2364 (sve_int_op, sve_int_op_rev): Handle the above codes.
2365 (sve_pred_int_rhs2_operand): Likewise.
2366 (MULLBT, SHRNB, SHRNT): Delete.
2367 (SVE_INT_SHIFT_IMM): New int iterator.
2368 (SVE_WHILE): Add UNSPEC_WHILEGE, UNSPEC_WHILEGT, UNSPEC_WHILEHI
2369 and UNSPEC_WHILEHS for TARGET_SVE2.
2370 (SVE2_U32_UNARY, SVE2_INT_UNARY_NARROWB, SVE2_INT_UNARY_NARROWT)
2371 (SVE2_INT_BINARY, SVE2_INT_BINARY_LANE, SVE2_INT_BINARY_LONG)
2372 (SVE2_INT_BINARY_LONG_LANE, SVE2_INT_BINARY_NARROWB)
2373 (SVE2_INT_BINARY_NARROWT, SVE2_INT_BINARY_PAIR, SVE2_FP_BINARY_PAIR)
2374 (SVE2_INT_BINARY_PAIR_LONG, SVE2_INT_BINARY_WIDE): New int iterators.
2375 (SVE2_INT_SHIFT_IMM_LONG, SVE2_INT_SHIFT_IMM_NARROWB): Likewise.
2376 (SVE2_INT_SHIFT_IMM_NARROWT, SVE2_INT_SHIFT_INSERT, SVE2_INT_CADD)
2377 (SVE2_INT_BITPERM, SVE2_INT_TERNARY, SVE2_INT_TERNARY_LANE): Likewise.
2378 (SVE2_FP_TERNARY_LONG, SVE2_FP_TERNARY_LONG_LANE, SVE2_INT_CMLA)
2379 (SVE2_INT_CDOT, SVE2_INT_ADD_BINARY_LONG, SVE2_INT_QADD_BINARY_LONG)
2380 (SVE2_INT_SUB_BINARY_LONG, SVE2_INT_QSUB_BINARY_LONG): Likewise.
2381 (SVE2_INT_ADD_BINARY_LONG_LANE, SVE2_INT_QADD_BINARY_LONG_LANE)
2382 (SVE2_INT_SUB_BINARY_LONG_LANE, SVE2_INT_QSUB_BINARY_LONG_LANE)
2383 (SVE2_COND_INT_UNARY_FP, SVE2_COND_FP_UNARY_LONG): Likewise.
2384 (SVE2_COND_FP_UNARY_NARROWB, SVE2_COND_INT_BINARY): Likewise.
2385 (SVE2_COND_INT_BINARY_NOREV, SVE2_COND_INT_BINARY_REV): Likewise.
2386 (SVE2_COND_INT_SHIFT, SVE2_MATCH, SVE2_PMULL): Likewise.
2387 (optab): Handle the new unspecs.
2388 (su, r): Remove entries for UNSPEC_SHRNB, UNSPEC_SHRNT, UNSPEC_RSHRNB
2390 (lr): Handle the new unspecs.
2392 (cmp_op, while_optab_cmp, sve_int_op): Handle the new unspecs.
2393 (sve_int_op_rev, sve_int_add_op, sve_int_qadd_op, sve_int_sub_op)
2394 (sve_int_qsub_op): New int attributes.
2395 (sve_fp_op, rot): Handle the new unspecs.
2396 * config/aarch64/aarch64-sve-builtins.h
2397 (function_resolver::require_matching_pointer_type): Declare.
2398 (function_resolver::resolve_unary): Add an optional boolean argument.
2399 (function_resolver::finish_opt_n_resolution): Add an optional
2400 type_suffix_index argument.
2401 (gimple_folder::redirect_call): Declare.
2402 (gimple_expander::prepare_gather_address_operands): Add an optional
2404 * config/aarch64/aarch64-sve-builtins.cc: Include
2405 aarch64-sve-builtins-sve2.h.
2406 (TYPES_b_unsigned, TYPES_b_integer, TYPES_bh_integer): New macros.
2407 (TYPES_bs_unsigned, TYPES_hs_signed, TYPES_hs_integer): Likewise.
2408 (TYPES_hd_unsigned, TYPES_hsd_signed): Likewise.
2409 (TYPES_hsd_integer): Use TYPES_hsd_signed.
2410 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): New macros.
2411 (TYPES_s_unsigned): Likewise.
2412 (TYPES_s_integer): Use TYPES_s_unsigned.
2413 (TYPES_sd_signed, TYPES_sd_unsigned): New macros.
2414 (TYPES_sd_integer): Use them.
2415 (TYPES_d_unsigned): New macro.
2416 (TYPES_d_integer): Use it.
2417 (TYPES_d_data, TYPES_cvt_long, TYPES_cvt_narrow_s): New macros.
2418 (TYPES_cvt_narrow): Likewise.
2419 (DEF_SVE_TYPES_ARRAY): Include the new types macros above.
2420 (preds_mx): New variable.
2421 (function_builder::add_overloaded_function): Allow the new feature
2422 set to be more restrictive than the original one.
2423 (function_resolver::infer_pointer_type): Remove qualifiers from
2424 the pointer type before printing it.
2425 (function_resolver::require_matching_pointer_type): New function.
2426 (function_resolver::resolve_sv_displacement): Handle functions
2427 that don't support 32-bit vector indices or svint32_t vector offsets.
2428 (function_resolver::finish_opt_n_resolution): Take the inferred type
2429 as a separate argument.
2430 (function_resolver::resolve_unary): Optionally treat all forms in
2431 the same way as normal merging functions.
2432 (gimple_folder::redirect_call): New function.
2433 (function_expander::prepare_gather_address_operands): Add an argument
2434 that says whether scaled forms are available. If they aren't,
2435 handle scaling of vector indices and don't add the extension and
2437 (function_expander::map_to_unspecs): If aarch64_sve isn't available,
2438 fall back to using cond_* instead.
2439 * config/aarch64/aarch64-sve-builtins-functions.h (rtx_code_function):
2440 Split out the member variables into...
2441 (rtx_code_function_base): ...this new base class.
2442 (rtx_code_function_rotated): Inherit rtx_code_function_base.
2443 (unspec_based_function): Split out the member variables into...
2444 (unspec_based_function_base): ...this new base class.
2445 (unspec_based_function_rotated): Inherit unspec_based_function_base.
2446 (unspec_based_function_exact_insn): New class.
2447 (unspec_based_add_function, unspec_based_add_lane_function)
2448 (unspec_based_lane_function, unspec_based_pred_function)
2449 (unspec_based_qadd_function, unspec_based_qadd_lane_function)
2450 (unspec_based_qsub_function, unspec_based_qsub_lane_function)
2451 (unspec_based_sub_function, unspec_based_sub_lane_function): New
2453 (unspec_based_fused_function): New class.
2454 (unspec_based_mla_function, unspec_based_mls_function): New typedefs.
2455 (unspec_based_fused_lane_function): New class.
2456 (unspec_based_mla_lane_function, unspec_based_mls_lane_function): New
2458 (CODE_FOR_MODE1): New macro.
2459 (fixed_insn_function): New class.
2460 (while_comparison): Likewise.
2461 * config/aarch64/aarch64-sve-builtins-shapes.h (binary_long_lane)
2462 (binary_long_opt_n, binary_narrowb_opt_n, binary_narrowt_opt_n)
2463 (binary_to_uint, binary_wide, binary_wide_opt_n, compare, compare_ptr)
2464 (load_ext_gather_index_restricted, load_ext_gather_offset_restricted)
2465 (load_gather_sv_restricted, shift_left_imm_long): Declare.
2466 (shift_left_imm_to_uint, shift_right_imm_narrowb): Likewise.
2467 (shift_right_imm_narrowt, shift_right_imm_narrowb_to_uint): Likewise.
2468 (shift_right_imm_narrowt_to_uint, store_scatter_index_restricted)
2469 (store_scatter_offset_restricted, tbl_tuple, ternary_long_lane)
2470 (ternary_long_opt_n, ternary_qq_lane_rotate, ternary_qq_rotate)
2471 (ternary_shift_left_imm, ternary_shift_right_imm, ternary_uint)
2472 (unary_convert_narrowt, unary_long, unary_narrowb, unary_narrowt)
2473 (unary_narrowb_to_uint, unary_narrowt_to_uint, unary_to_int): Likewise.
2474 * config/aarch64/aarch64-sve-builtins-shapes.cc (apply_predication):
2475 Also add an initial argument for unary_convert_narrowt, regardless
2476 of the predication type.
2477 (build_32_64): Allow loads and stores to specify MODE_none.
2478 (build_sv_index64, build_sv_uint_offset): New functions.
2479 (long_type_suffix): New function.
2480 (binary_imm_narrowb_base, binary_imm_narrowt_base): New classes.
2481 (binary_imm_long_base, load_gather_sv_base): Likewise.
2482 (shift_right_imm_narrow_wrapper, ternary_shift_imm_base): Likewise.
2483 (ternary_resize2_opt_n_base, ternary_resize2_lane_base): Likewise.
2484 (unary_narrowb_base, unary_narrowt_base): Likewise.
2485 (binary_long_lane_def, binary_long_lane): New shape.
2486 (binary_long_opt_n_def, binary_long_opt_n): Likewise.
2487 (binary_narrowb_opt_n_def, binary_narrowb_opt_n): Likewise.
2488 (binary_narrowt_opt_n_def, binary_narrowt_opt_n): Likewise.
2489 (binary_to_uint_def, binary_to_uint): Likewise.
2490 (binary_wide_def, binary_wide): Likewise.
2491 (binary_wide_opt_n_def, binary_wide_opt_n): Likewise.
2492 (compare_def, compare): Likewise.
2493 (compare_ptr_def, compare_ptr): Likewise.
2494 (load_ext_gather_index_restricted_def,
2495 load_ext_gather_index_restricted): Likewise.
2496 (load_ext_gather_offset_restricted_def,
2497 load_ext_gather_offset_restricted): Likewise.
2498 (load_gather_sv_def): Inherit from load_gather_sv_base.
2499 (load_gather_sv_restricted_def, load_gather_sv_restricted): New shape.
2500 (shift_left_imm_def, shift_left_imm): Likewise.
2501 (shift_left_imm_long_def, shift_left_imm_long): Likewise.
2502 (shift_left_imm_to_uint_def, shift_left_imm_to_uint): Likewise.
2503 (store_scatter_index_restricted_def,
2504 store_scatter_index_restricted): Likewise.
2505 (store_scatter_offset_restricted_def,
2506 store_scatter_offset_restricted): Likewise.
2507 (tbl_tuple_def, tbl_tuple): Likewise.
2508 (ternary_long_lane_def, ternary_long_lane): Likewise.
2509 (ternary_long_opt_n_def, ternary_long_opt_n): Likewise.
2510 (ternary_qq_lane_def): Inherit from ternary_resize2_lane_base.
2511 (ternary_qq_lane_rotate_def, ternary_qq_lane_rotate): New shape
2512 (ternary_qq_opt_n_def): Inherit from ternary_resize2_opt_n_base.
2513 (ternary_qq_rotate_def, ternary_qq_rotate): New shape.
2514 (ternary_shift_left_imm_def, ternary_shift_left_imm): Likewise.
2515 (ternary_shift_right_imm_def, ternary_shift_right_imm): Likewise.
2516 (ternary_uint_def, ternary_uint): Likewise.
2517 (unary_convert): Fix typo in comment.
2518 (unary_convert_narrowt_def, unary_convert_narrowt): New shape.
2519 (unary_long_def, unary_long): Likewise.
2520 (unary_narrowb_def, unary_narrowb): Likewise.
2521 (unary_narrowt_def, unary_narrowt): Likewise.
2522 (unary_narrowb_to_uint_def, unary_narrowb_to_uint): Likewise.
2523 (unary_narrowt_to_uint_def, unary_narrowt_to_uint): Likewise.
2524 (unary_to_int_def, unary_to_int): Likewise.
2525 * config/aarch64/aarch64-sve-builtins-base.cc (unspec_cmla)
2526 (unspec_fcmla, unspec_cond_fcmla, expand_mla_mls_lane): New functions.
2527 (svasrd_impl): Delete.
2528 (svcadd_impl::expand): Handle integer operations too.
2529 (svcmla_impl::expand, svcmla_lane::expand): Likewise, using the
2530 new functions to derive the unspec numbers.
2531 (svmla_svmls_lane_impl): Replace with...
2532 (svmla_lane_impl, svmls_lane_impl): ...these new classes. Handle
2533 integer operations too.
2534 (svwhile_impl): Rename to...
2535 (svwhilelx_impl): ...this and inherit from while_comparison.
2536 (svasrd): Use unspec_based_function.
2537 (svmla_lane): Use svmla_lane_impl.
2538 (svmls_lane): Use svmls_lane_impl.
2539 (svrecpe, svrsqrte): Handle unsigned integer operations too.
2540 (svwhilele, svwhilelt): Use svwhilelx_impl.
2541 * config/aarch64/aarch64-sve-builtins-sve2.h: New file.
2542 * config/aarch64/aarch64-sve-builtins-sve2.cc: Likewise.
2543 * config/aarch64/aarch64-sve-builtins-sve2.def: Likewise.
2544 * config/aarch64/aarch64-sve-builtins.def: Include
2545 aarch64-sve-builtins-sve2.def.
2547 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
2549 * config/aarch64/aarch64-protos.h (aarch64_sve_arith_immediate_p)
2550 (aarch64_sve_sqadd_sqsub_immediate_p): Add a machine_mode argument.
2551 * config/aarch64/aarch64.c (aarch64_sve_arith_immediate_p)
2552 (aarch64_sve_sqadd_sqsub_immediate_p): Likewise. Handle scalar
2553 immediates as well as vector ones.
2554 * config/aarch64/predicates.md (aarch64_sve_arith_immediate)
2555 (aarch64_sve_sub_arith_immediate, aarch64_sve_qadd_immediate)
2556 (aarch64_sve_qsub_immediate): Update calls accordingly.
2558 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
2560 * config/aarch64/aarch64-sve2.md: Add banner comments.
2561 (<su>mulh<r>s<mode>3): Move further up file.
2562 (<su>mull<bt><Vwide>, <r>shrnb<mode>, <r>shrnt<mode>)
2563 (*aarch64_sve2_sra<mode>): Move further down file.
2564 * config/aarch64/t-aarch64 (s-check-sve-md): Check aarch64-sve2.md too.
2566 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
2568 * config/aarch64/iterators.md (SVE_WHILE): Add UNSPEC_WHILERW
2570 (while_optab_cmp): Handle them.
2571 * config/aarch64/aarch64-sve.md
2572 (*while_<while_optab_cmp><GPI:mode><PRED_ALL:mode>_ptest): Make public
2573 and add a "@" marker.
2574 * config/aarch64/aarch64-sve2.md (check_<raw_war>_ptrs<mode>): Use it
2575 instead of gen_aarch64_sve2_while_ptest.
2576 (@aarch64_sve2_while<cmp_op><GPI:mode><PRED_ALL:mode>_ptest): Delete.
2578 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
2580 * config/aarch64/aarch64.md (UNSPEC_WHILE_LE): Rename to...
2581 (UNSPEC_WHILELE): ...this.
2582 (UNSPEC_WHILE_LO): Rename to...
2583 (UNSPEC_WHILELO): ...this.
2584 (UNSPEC_WHILE_LS): Rename to...
2585 (UNSPEC_WHILELS): ...this.
2586 (UNSPEC_WHILE_LT): Rename to...
2587 (UNSPEC_WHILELT): ...this.
2588 * config/aarch64/iterators.md (SVE_WHILE): Update accordingly.
2589 (cmp_op, while_optab_cmp): Likewise.
2590 * config/aarch64/aarch64.c (aarch64_sve_move_pred_via_while): Likewise.
2591 * config/aarch64/aarch64-sve-builtins-base.cc (svwhilele): Likewise.
2592 (svwhilelt): Likewise.
2594 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
2596 * config/aarch64/aarch64-sve-builtins-shapes.h (unary_count): Delete.
2597 (unary_to_uint): Define.
2598 * config/aarch64/aarch64-sve-builtins-shapes.cc (unary_count_def)
2599 (unary_count): Rename to...
2600 (unary_to_uint_def, unary_to_uint): ...this.
2601 * config/aarch64/aarch64-sve-builtins-base.def: Update accordingly.
2603 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
2605 * config/aarch64/aarch64-sve-builtins-functions.h
2606 (code_for_mode_function): New class.
2607 (CODE_FOR_MODE0, QUIET_CODE_FOR_MODE0): New macros.
2608 * config/aarch64/aarch64-sve-builtins-base.cc (svcompact_impl)
2609 (svext_impl, svmul_lane_impl, svsplice_impl, svtmad_impl): Delete.
2610 (svcompact, svext, svsplice): Use QUIET_CODE_FOR_MODE0.
2611 (svmul_lane, svtmad): Use CODE_FOR_MODE0.
2613 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
2615 * config/aarch64/iterators.md (addsub): New code attribute.
2616 * config/aarch64/aarch64-simd.md (aarch64_<su_optab><optab><mode>):
2618 (aarch64_<su_optab>q<addsub><mode>): ...this, making the same change
2619 in the asm string and attributes. Fix indentation.
2620 * config/aarch64/aarch64-sve.md (@aarch64_<su_optab><optab><mode>):
2622 (@aarch64_sve_<optab><mode>): ...this.
2623 * config/aarch64/aarch64-sve-builtins.h
2624 (function_expander::expand_signed_unpred_op): Delete.
2625 * config/aarch64/aarch64-sve-builtins.cc
2626 (function_expander::expand_signed_unpred_op): Likewise.
2627 (function_expander::map_to_rtx_codes): If the optab isn't defined,
2628 try using code_for_aarch64_sve instead.
2629 * config/aarch64/aarch64-sve-builtins-base.cc (svqadd_impl): Delete.
2630 (svqsub_impl): Likewise.
2631 (svqadd, svqsub): Use rtx_code_function instead.
2633 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
2635 * config/aarch64/iterators.md (SRHSUB, URHSUB): Delete.
2636 (HADDSUB, sur, addsub): Remove them.
2638 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
2640 * tree-nrv.c (pass_return_slot::execute): Handle all internal
2641 functions the same way, rather than singling out those that
2642 aren't mapped directly to optabs.
2644 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
2646 * target.def (compatible_vector_types_p): New target hook.
2647 * hooks.h (hook_bool_const_tree_const_tree_true): Declare.
2648 * hooks.c (hook_bool_const_tree_const_tree_true): New function.
2649 * doc/tm.texi.in (TARGET_COMPATIBLE_VECTOR_TYPES_P): New hook.
2650 * doc/tm.texi: Regenerate.
2651 * gimple-expr.c: Include target.h.
2652 (useless_type_conversion_p): Use targetm.compatible_vector_types_p.
2653 * config/aarch64/aarch64.c (aarch64_compatible_vector_types_p): New
2655 (TARGET_COMPATIBLE_VECTOR_TYPES_P): Define.
2656 * config/aarch64/aarch64-sve-builtins.cc (gimple_folder::convert_pred):
2657 Use the original predicate if it already has a suitable type.
2659 2020-01-09 Martin Jambor <mjambor@suse.cz>
2661 * cgraph.h (cgraph_edge): Make remove, set_call_stmt, make_direct,
2662 resolve_speculation and redirect_call_stmt_to_callee static. Change
2663 return type of set_call_stmt to cgraph_edge *.
2664 * auto-profile.c (afdo_indirect_call): Adjust call to
2665 redirect_call_stmt_to_callee.
2666 * cgraph.c (cgraph_edge::set_call_stmt): Make return cgraph-edge *,
2667 make the this pointer explicit, adjust self-recursive calls and the
2668 call top make_direct. Return the resulting edge.
2669 (cgraph_edge::remove): Make this pointer explicit.
2670 (cgraph_edge::resolve_speculation): Likewise, adjust call to remove.
2671 (cgraph_edge::make_direct): Likewise, adjust call to
2672 resolve_speculation.
2673 (cgraph_edge::redirect_call_stmt_to_callee): Likewise, also adjust
2674 call to set_call_stmt.
2675 (cgraph_update_edges_for_call_stmt_node): Update call to
2676 set_call_stmt and remove.
2677 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
2678 Renamed edge to master_edge. Adjusted calls to set_call_stmt.
2679 (cgraph_node::create_edge_including_clones): Moved "first" definition
2680 of edge to the block where it was used. Adjusted calls to
2682 (cgraph_node::remove_symbol_and_inline_clones): Adjust call to
2683 cgraph_edge::remove.
2684 * cgraphunit.c (walk_polymorphic_call_targets): Adjusted calls to
2685 make_direct and redirect_call_stmt_to_callee.
2686 * ipa-fnsummary.c (redirect_to_unreachable): Adjust calls to
2687 resolve_speculation and make_direct.
2688 * ipa-inline-transform.c (inline_transform): Adjust call to
2689 redirect_call_stmt_to_callee.
2690 (check_speculations_1):: Adjust call to resolve_speculation.
2691 * ipa-inline.c (resolve_noninline_speculation): Adjust call to
2692 resolve-speculation.
2693 (inline_small_functions): Adjust call to resolve_speculation.
2694 (ipa_inline): Likewise.
2695 * ipa-prop.c (ipa_make_edge_direct_to_target): Adjust call to
2697 * ipa-visibility.c (function_and_variable_visibility): Make iteration
2698 safe with regards to edge removal, adjust calls to
2699 redirect_call_stmt_to_callee.
2700 * ipa.c (walk_polymorphic_call_targets): Adjust calls to make_direct
2701 and redirect_call_stmt_to_callee.
2702 * multiple_target.c (create_dispatcher_calls): Adjust call to
2703 redirect_call_stmt_to_callee
2704 (redirect_to_specific_clone): Likewise.
2705 * tree-cfgcleanup.c (delete_unreachable_blocks_update_callgraph):
2706 Adjust calls to cgraph_edge::remove.
2707 * tree-inline.c (copy_bb): Adjust call to set_call_stmt.
2708 (redirect_all_calls): Adjust call to redirect_call_stmt_to_callee.
2709 (expand_call_inline): Adjust call to cgraph_edge::remove.
2711 2020-01-09 Martin Liska <mliska@suse.cz>
2713 * params.opt: Set Optimization for
2714 param_max_speculative_devirt_maydefs.
2716 2020-01-09 Martin Sebor <msebor@redhat.com>
2720 * builtins.c (compute_objsize): Avoid handling MEM_REFs of vector type.
2722 2020-01-09 Martin Liska <mliska@suse.cz>
2724 * auto-profile.c (auto_profile): Use opt_for_fn
2726 * ipa-cp.c (ipcp_lattice::add_value): Likewise.
2727 (propagate_vals_across_arith_jfunc): Likewise.
2728 (hint_time_bonus): Likewise.
2729 (incorporate_penalties): Likewise.
2730 (good_cloning_opportunity_p): Likewise.
2731 (perform_estimation_of_a_value): Likewise.
2732 (estimate_local_effects): Likewise.
2733 (ipcp_propagate_stage): Likewise.
2734 * ipa-fnsummary.c (decompose_param_expr): Likewise.
2735 (set_switch_stmt_execution_predicate): Likewise.
2736 (analyze_function_body): Likewise.
2737 * ipa-inline-analysis.c (offline_size): Likewise.
2738 * ipa-inline.c (early_inliner): Likewise.
2739 * ipa-prop.c (ipa_analyze_node): Likewise.
2740 (ipcp_transform_function): Likewise.
2741 * ipa-sra.c (process_scan_results): Likewise.
2742 (ipa_sra_summarize_function): Likewise.
2743 * params.opt: Rename ipcp-unit-growth to
2744 ipa-cp-unit-growth. Add Optimization for various
2745 IPA-related parameters.
2747 2020-01-09 Richard Biener <rguenther@suse.de>
2750 * gimplify.c (gimplify_expr): Deal with NOP definitions.
2752 2020-01-09 Richard Biener <rguenther@suse.de>
2754 PR tree-optimization/93040
2755 * gimple-ssa-store-merging.c (find_bswap_or_nop): Raise search limit.
2757 2020-01-09 Georg-Johann Lay <avr@gjlay.de>
2759 * common/config/avr/avr-common.c (avr_option_optimization_table)
2760 [OPT_LEVELS_1_PLUS]: Set -fsplit-wide-types-early.
2762 2020-01-09 Martin Liska <mliska@suse.cz>
2764 * cgraphclones.c (symbol_table::materialize_all_clones):
2765 Use cgraph_node::dump_name.
2767 2020-01-09 Jakub Jelinek <jakub@redhat.com>
2770 * config/riscv/riscv.c (riscv_print_operand_reloc): Use
2771 output_operand_lossage instead of gcc_unreachable.
2772 * doc/md.texi (riscv f constraint): Fix typo.
2775 * config/i386/i386.md (subv<mode>4): Use SWIDWI iterator instead of
2776 SWI. Use <general_hilo_operand> instead of <general_operand>. Use
2777 CONST_SCALAR_INT_P instead of CONST_INT_P.
2778 (*subv<mode>4_1): Rename to ...
2779 (subv<mode>4_1): ... this.
2780 (*subv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
2781 define_insn_and_split patterns.
2782 (*subv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
2785 2020-01-08 David Malcolm <dmalcolm@redhat.com>
2787 * vec.c (class selftest::count_dtor): New class.
2788 (selftest::test_auto_delete_vec): New test.
2789 (selftest::vec_c_tests): Call it.
2790 * vec.h (class auto_delete_vec): New class template.
2791 (auto_delete_vec<T>::~auto_delete_vec): New dtor.
2793 2020-01-08 David Malcolm <dmalcolm@redhat.com>
2795 * sbitmap.h (auto_sbitmap): Add operator const_sbitmap.
2797 2020-01-08 Jim Wilson <jimw@sifive.com>
2799 * config/riscv/riscv.c (riscv_legitimize_tls_address): Ifdef out
2800 use of TLS_MODEL_LOCAL_EXEC when not pic.
2802 2020-01-08 David Malcolm <dmalcolm@redhat.com>
2804 * hash-map-tests.c (selftest::test_map_of_strings_to_int): Fix
2807 2020-01-08 Jakub Jelinek <jakub@redhat.com>
2810 * config/i386/i386.md (*stack_protect_set_2_<mode> peephole2,
2811 *stack_protect_set_3 peephole2): Also check that the second
2812 insns source is general_operand.
2815 * config/i386/i386.md (addcarry<mode>_0): Use nonimmediate_operand
2816 predicate for output operand instead of register_operand.
2817 (addcarry<mode>, addcarry<mode>_1): Likewise. Add alternative with
2818 memory destination and non-memory operands[2].
2820 2020-01-08 Martin Liska <mliska@suse.cz>
2822 * cgraph.c (cgraph_node::dump): Use ::dump_name or
2823 ::dump_asm_name instead of (::name or ::asm_name).
2824 * cgraphclones.c (symbol_table::materialize_all_clones): Likewise.
2825 * cgraphunit.c (walk_polymorphic_call_targets): Likewise.
2826 (analyze_functions): Likewise.
2827 (expand_all_functions): Likewise.
2828 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
2829 (propagate_bits_across_jump_function): Likewise.
2830 (dump_profile_updates): Likewise.
2831 (ipcp_store_bits_results): Likewise.
2832 (ipcp_store_vr_results): Likewise.
2833 * ipa-devirt.c (dump_targets): Likewise.
2834 * ipa-fnsummary.c (analyze_function_body): Likewise.
2835 * ipa-hsa.c (check_warn_node_versionable): Likewise.
2836 (process_hsa_functions): Likewise.
2837 * ipa-icf.c (sem_item_optimizer::merge_classes): Likewise.
2838 (set_alias_uids): Likewise.
2839 * ipa-inline-transform.c (save_inline_function_body): Likewise.
2840 * ipa-inline.c (recursive_inlining): Likewise.
2841 (inline_to_all_callers_1): Likewise.
2842 (ipa_inline): Likewise.
2843 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
2844 (ipa_propagate_frequency): Likewise.
2845 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
2846 (remove_described_reference): Likewise.
2847 * ipa-pure-const.c (worse_state): Likewise.
2848 (check_retval_uses): Likewise.
2849 (analyze_function): Likewise.
2850 (propagate_pure_const): Likewise.
2851 (propagate_nothrow): Likewise.
2852 (dump_malloc_lattice): Likewise.
2853 (propagate_malloc): Likewise.
2854 (pass_local_pure_const::execute): Likewise.
2855 * ipa-visibility.c (optimize_weakref): Likewise.
2856 (function_and_variable_visibility): Likewise.
2857 * ipa.c (symbol_table::remove_unreachable_nodes): Likewise.
2858 (ipa_discover_variable_flags): Likewise.
2859 * lto-streamer-out.c (output_function): Likewise.
2860 (output_constructor): Likewise.
2861 * tree-inline.c (copy_bb): Likewise.
2862 * tree-ssa-structalias.c (ipa_pta_execute): Likewise.
2863 * varpool.c (symbol_table::remove_unreferenced_decls): Likewise.
2865 2020-01-08 Richard Biener <rguenther@suse.de>
2868 * tree-eh.c (sink_clobbers): Update virtual operands for
2869 the first and last stmt only. Add a dry-run capability.
2870 (pass_lower_eh_dispatch::execute): Perform clobber sinking
2871 after CFG manipulations and in RPO order to catch all
2872 secondary opportunities reliably.
2874 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
2877 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
2879 2019-01-08 Richard Biener <rguenther@suse.de>
2882 * gimple-fold.c (rewrite_to_defined_overflow): Mark stmt modified.
2883 * tree-ssa-loop-im.c (move_computations_worker): Properly adjust
2884 virtual operand, also updating SSA use.
2885 * gimple-loop-interchange.cc (loop_cand::undo_simple_reduction):
2886 Update stmt after resetting virtual operand.
2887 (tree_loop_interchange::move_code_to_inner_loop): Likewise.
2888 * gimple-iterator.c (gsi_remove): When not removing the stmt
2889 permanently do not delink immediate uses or mark the stmt modified.
2891 2020-01-08 Martin Liska <mliska@suse.cz>
2893 * ipa-fnsummary.c (dump_ipa_call_summary): Use symtab_node::dump_name.
2894 (ipa_call_context::estimate_size_and_time): Likewise.
2895 (inline_analyze_function): Likewise.
2897 2020-01-08 Martin Liska <mliska@suse.cz>
2899 * cgraph.c (cgraph_node::dump): Use systematically
2902 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
2904 Add -nodevicespecs option for avr.
2907 * config/avr/avr.opt (-nodevicespecs): New driver option.
2908 * config/avr/driver-avr.c (avr_devicespecs_file): Only issue
2909 "-specs=device-specs/..." if that option is not set.
2910 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
2912 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
2914 Implement 64-bit double functions for avr.
2917 * config.gcc (tm_defines) [target=avr]: Support --with-libf7,
2918 --with-double-comparison.
2919 * doc/install.texi: Document them.
2920 * config/avr/avr-c.c (avr_cpu_cpp_builtins)
2921 <WITH_LIBF7_LIBGCC, WITH_LIBF7_MATH, WITH_LIBF7_MATH_SYMBOLS>
2922 <WITH_DOUBLE_COMPARISON>: New built-in defines.
2923 * doc/invoke.texi (AVR Built-in Macros): Document them.
2924 * config/avr/avr-protos.h (avr_float_lib_compare_returns_bool): New.
2925 * config/avr/avr.c (avr_float_lib_compare_returns_bool): New function.
2926 * config/avr/avr.h (FLOAT_LIB_COMPARE_RETURNS_BOOL): New macro.
2928 2020-01-08 Richard Earnshaw <rearnsha@arm.com>
2931 * config/arm/t-multilib (MULTILIB_MATCHES): Add rules to match
2932 armv7-a{+mp,+sec,+mp+sec} to appropriate armv7 multilib variants
2933 when only building rm-profile multilibs.
2935 2020-01-08 Feng Xue <fxue@os.amperecomputing.com>
2938 * ipa-cp.c (self_recursively_generated_p): Find matched aggregate
2939 lattice for a value to check.
2940 (propagate_vals_across_arith_jfunc): Add an assertion to ensure
2941 finite propagation in self-recursive scc.
2943 2020-01-08 Luo Xiong Hu <luoxhu@linux.ibm.com>
2945 * ipa-inline.c (caller_growth_limits): Restore the AND.
2947 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
2949 * config/gcn/gcn-valu.md (VEC_1REG_INT_ALT): Delete iterator.
2950 (VEC_ALLREG_ALT): New iterator.
2951 (VEC_ALLREG_INT_MODE): New iterator.
2952 (VCMP_MODE): New iterator.
2953 (VCMP_MODE_INT): New iterator.
2954 (vec_cmpu<mode>di): Use VCMP_MODE_INT.
2955 (vec_cmp<u>v64qidi): New define_expand.
2956 (vec_cmp<mode>di_exec): Use VCMP_MODE.
2957 (vec_cmpu<mode>di_exec): New define_expand.
2958 (vec_cmp<u>v64qidi_exec): New define_expand.
2959 (vec_cmp<mode>di_dup): Use VCMP_MODE.
2960 (vec_cmp<mode>di_dup_exec): Use VCMP_MODE.
2961 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>): Rename ...
2962 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): ... to this.
2963 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>_exec): Rename ...
2964 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): ... to this.
2965 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>): Rename ...
2966 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): ... to this.
2967 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>_exec): Rename ...
2968 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): ... to
2970 * config/gcn/gcn.c (print_operand): Fix 8 and 16 bit suffixes.
2971 * config/gcn/gcn.md (expander): Add sign_extend and zero_extend.
2973 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
2975 * config/gcn/constraints.md (DA): Update description and match.
2977 (Db): New constraint.
2978 * config/gcn/gcn-protos.h (gcn_inline_constant64_p): Add second
2980 * config/gcn/gcn.c (gcn_inline_constant64_p): Add 'mixed' parameter.
2981 Implement 'Db' mixed immediate type.
2982 * config/gcn/gcn-valu.md (addcv64si3<exec_vcc>): Rework constraints.
2983 (addcv64si3_dup<exec_vcc>): Delete.
2984 (subcv64si3<exec_vcc>): Rework constraints.
2985 (addv64di3): Rework constraints.
2986 (addv64di3_exec): Rework constraints.
2987 (subv64di3): Rework constraints.
2988 (addv64di3_dup): Delete.
2989 (addv64di3_dup_exec): Delete.
2990 (addv64di3_zext): Rework constraints.
2991 (addv64di3_zext_exec): Rework constraints.
2992 (addv64di3_zext_dup): Rework constraints.
2993 (addv64di3_zext_dup_exec): Rework constraints.
2994 (addv64di3_zext_dup2): Rework constraints.
2995 (addv64di3_zext_dup2_exec): Rework constraints.
2996 (addv64di3_sext_dup2): Rework constraints.
2997 (addv64di3_sext_dup2_exec): Rework constraints.
2999 2020-01-07 Andre Vieira <andre.simoesdiasvieira@arm.com>
3001 * doc/sourcebuild.texi (arm_little_endian, arm_nothumb): Documented
3002 existing target checks.
3004 2020-01-07 Richard Biener <rguenther@suse.de>
3006 * doc/install.texi: Bump minimal supported MPC version.
3008 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
3010 * langhooks-def.h (lhd_simulate_enum_decl): Declare.
3011 (LANG_HOOKS_SIMULATE_ENUM_DECL): Use it.
3012 * langhooks.c: Include stor-layout.h.
3013 (lhd_simulate_enum_decl): New function.
3014 * config/aarch64/aarch64-sve-builtins.cc (init_builtins): Call
3015 handle_arm_sve_h for the LTO frontend.
3016 (register_vector_type): Cope with null returns from pushdecl.
3018 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
3020 * config/aarch64/aarch64-protos.h (aarch64_sve::svbool_type_p)
3021 (aarch64_sve::nvectors_if_data_type): Replace with...
3022 (aarch64_sve::builtin_type_p): ...this.
3023 * config/aarch64/aarch64-sve-builtins.cc: Include attribs.h.
3024 (find_vector_type): Delete.
3025 (add_sve_type_attribute): New function.
3026 (lookup_sve_type_attribute): Likewise.
3027 (register_builtin_types): Add an "SVE type" attribute to each type.
3028 (register_tuple_type): Likewise.
3029 (svbool_type_p, nvectors_if_data_type): Delete.
3030 (mangle_builtin_type): Use lookup_sve_type_attribute.
3031 (builtin_type_p): Likewise. Add an overload that returns the
3032 number of constituent vector and predicate registers.
3033 * config/aarch64/aarch64.c (aarch64_sve_argument_p): Delete.
3034 (aarch64_returns_value_in_sve_regs_p): Use aarch64_sve::builtin_type_p
3035 instead of aarch64_sve_argument_p.
3036 (aarch64_takes_arguments_in_sve_regs_p): Likewise.
3037 (aarch64_pass_by_reference): Likewise.
3038 (aarch64_function_value_1): Likewise.
3039 (aarch64_return_in_memory): Likewise.
3040 (aarch64_layout_arg): Likewise.
3042 2020-01-07 Jakub Jelinek <jakub@redhat.com>
3044 PR tree-optimization/93156
3045 * tree-ssa-ccp.c (bit_value_binop): For x * x note that the second
3046 least significant bit is always clear.
3048 PR tree-optimization/93118
3049 * match.pd ((x >> c) << c -> x & (-1<<c)): Add nop_convert?. Add new
3050 simplifier with two intermediate conversions.
3052 2020-01-07 Martin Liska <mliska@suse.cz>
3054 * params.opt: Add Optimization for various parameters.
3056 2020-01-07 Martin Liska <mliska@suse.cz>
3059 * doc/extend.texi: Explain cloning for target_clone
3062 2020-01-07 Martin Liska <mliska@suse.cz>
3064 PR tree-optimization/92860
3065 * common.opt: Make in Optimization option
3066 as it is affected by -O0, which is an Optimization
3068 * tree-inline.c (tree_inlinable_function_p):
3069 Use opt_for_fn for warn_inline.
3070 (expand_call_inline): Likewise.
3072 2020-01-07 Martin Liska <mliska@suse.cz>
3074 PR tree-optimization/92860
3075 * common.opt: Make flag_ree as optimization
3078 2020-01-07 Martin Liska <mliska@suse.cz>
3080 PR optimization/92860
3081 * params.opt: Mark param_min_crossjump_insns with Optimization
3084 2020-01-07 Luo Xiong Hu <luoxhu@linux.ibm.com>
3086 * ipa-inline-analysis.c (estimate_growth): Fix typo.
3087 * ipa-inline.c (caller_growth_limits): Use OR instead of AND.
3089 2020-01-06 Michael Meissner <meissner@linux.ibm.com>
3091 * config/rs6000/rs6000.c (hard_reg_and_mode_to_addr_mask): New
3092 helper function to return the valid addressing formats for a given
3093 hard register and mode.
3094 (rs6000_adjust_vec_address): Call hard_reg_and_mode_to_addr_mask.
3096 * config/rs6000/constraints.md (Q constraint): Update
3098 * doc/md.texi (RS/6000 constraints): Update 'Q' cosntraint
3101 * config/rs6000/vsx.md (vsx_extract_<mode>_var, VSX_D iterator):
3102 Use 'Q' for doing vector extract from memory.
3103 (vsx_extract_v4sf_var): Use 'Q' for doing vector extract from
3105 (vsx_extract_<mode>_var, VSX_EXTRACT_I iterator): Use 'Q' for
3106 doing vector extract from memory.
3107 (vsx_extract_<mode>_<VS_scalar>mode_var): Use 'Q' for doing vector
3108 extract from memory.
3110 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add support
3111 for the offset being 34-bits when -mcpu=future is used.
3113 2020-01-06 John David Anglin <danglin@gcc.gnu.org>
3115 * config/pa/pa.md: Revert change to use ordered_comparison_operator
3116 instead of cmpib_comparison_operator in cmpib patterns.
3117 * config/pa/predicates.md (cmpib_comparison_operator): Revert removal
3118 of cmpib_comparison_operator. Revise comment.
3120 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
3122 * tree-vect-slp.c (vect_build_slp_tree_1): Require all shifts
3123 in an IFN_DIV_POW2 node to be equal.
3125 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
3127 * tree-vect-stmts.c (vect_check_load_store_mask): Rename to...
3128 (vect_check_scalar_mask): ...this.
3129 (vectorizable_store, vectorizable_load): Update call accordingly.
3130 (vectorizable_call): Use vect_check_scalar_mask to check the mask
3131 argument in calls to conditional internal functions.
3133 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
3135 * config/gcn/gcn-valu.md (subv64di3): Use separate alternatives for
3136 '0' matching inputs.
3137 (subv64di3_exec): Likewise.
3139 2020-01-06 Bryan Stenson <bryan@siliconvortex.com>
3141 * config/mips/mips.c (vr4130_align_insns): Fix typo.
3142 * doc/md.texi (movstr): Likewise.
3144 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
3146 * config/gcn/gcn-valu.md (vec_extract<mode><scalar_mode>): Add early
3149 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
3151 * config/aarch64/t-aarch64 ($(srcdir)/config/aarch64/aarch64-tune.md):
3153 (s-aarch64-tune-md): ...this new stamp file. Pipe the new contents
3154 to a temporary file and use move-if-change to update the real
3155 file where necessary.
3157 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
3159 * config/aarch64/aarch64-sve.md (@aarch64_sel_dup<mode>): Use Upl
3160 rather than Upa for CPY /M.
3162 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
3164 * config/gcn/gcn.c (gcn_inline_constant_p): Allow 64 as an inline
3167 2020-01-06 Martin Liska <mliska@suse.cz>
3169 PR tree-optimization/92860
3170 * params.opt: Mark param_max_combine_insns with Optimization
3173 2020-01-05 Jakub Jelinek <jakub@redhat.com>
3176 * config/i386/i386.md (SWIDWI): New mode iterator.
3177 (DWI, dwi): Add TImode variants.
3178 (addv<mode>4): Use SWIDWI iterator instead of SWI. Use
3179 <general_hilo_operand> instead of <general_operand>. Use
3180 CONST_SCALAR_INT_P instead of CONST_INT_P.
3181 (*addv<mode>4_1): Rename to ...
3182 (addv<mode>4_1): ... this.
3183 (QWI): New mode attribute.
3184 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
3185 define_insn_and_split patterns.
3186 (*addv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
3188 (uaddv<mode>4): Use SWIDWI iterator instead of SWI. Use
3189 <general_hilo_operand> instead of <general_operand>.
3190 (*addcarry<mode>_1): New define_insn.
3191 (*add<dwi>3_doubleword_cc_overflow_1): New define_insn_and_split.
3193 2020-01-03 Konstantin Kharlamov <Hi-Angel@yandex.ru>
3195 * gdbinit.in (pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, pdd, pbs, pbm):
3196 Use "call" instead of "set".
3198 2020-01-03 Martin Jambor <mjambor@suse.cz>
3201 * ipa-cp.c (print_all_lattices): Skip functions without info.
3203 2020-01-03 Jakub Jelinek <jakub@redhat.com>
3206 * config/i386/i386-options.c (ix86_simd_clone_adjust): If
3207 TARGET_PREFER_AVX128, use prefer-vector-width=256 for 'c' and 'd'
3208 simd clones. If TARGET_PREFER_AVX256, use prefer-vector-width=512
3209 for 'e' simd clones.
3212 * config/i386/i386.opt (x_prefer_vector_width_type): Remove TargetSave
3214 (mprefer-vector-width=): Add Save.
3215 * config/i386/i386-options.c (ix86_target_string): Add PVW argument, print
3216 -mprefer-vector-width= if non-zero. Fix up -mfpmath= comment.
3217 (ix86_debug_options, ix86_function_specific_print): Adjust
3218 ix86_target_string callers.
3219 (ix86_valid_target_attribute_inner_p): Handle prefer-vector-width=.
3220 (ix86_valid_target_attribute_tree): Likewise.
3221 * config/i386/i386-options.h (ix86_target_string): Add PVW argument.
3222 * config/i386/i386-expand.c (ix86_expand_builtin): Adjust
3223 ix86_target_string caller.
3226 * config/i386/i386.md (abs<mode>2): Use expand_simple_binop instead of
3227 emitting ASHIFTRT, XOR and MINUS by hand. Use gen_int_mode with QImode
3228 instead of gen_int_shift_amount + convert_modes.
3230 PR rtl-optimization/93088
3231 * loop-iv.c (find_single_def_src): Punt after looking through
3232 128 reg copies for regs with single definitions. Move definitions
3235 2020-01-02 Dennis Zhang <dennis.zhang@arm.com>
3237 * config/arm/arm-c.c (arm_cpu_builtins): Define
3238 __ARM_FEATURE_MATMUL_INT8, __ARM_FEATURE_BF16_VECTOR_ARITHMETIC,
3239 __ARM_FEATURE_BF16_SCALAR_ARITHMETIC, and
3240 __ARM_BF16_FORMAT_ALTERNATIVE when enabled.
3241 * config/arm/arm-cpus.in (armv8_6, i8mm, bf16): New features.
3242 * config/arm/arm-tables.opt: Regenerated.
3243 * config/arm/arm.c (arm_option_reconfigure_globals): Initialize
3244 arm_arch_i8mm and arm_arch_bf16 when enabled.
3245 * config/arm/arm.h (TARGET_I8MM): New macro.
3246 (TARGET_BF16_FP, TARGET_BF16_SIMD): Likewise.
3247 * config/arm/t-aprofile: Add matching rules for -march=armv8.6-a.
3248 * config/arm/t-arm-elf (all_v8_archs): Add armv8.6-a.
3249 * config/arm/t-multilib: Add matching rules for -march=armv8.6-a.
3250 (v8_6_a_simd_variants): New.
3251 (v8_*_a_simd_variants): Add i8mm and bf16.
3252 * doc/invoke.texi (armv8.6-a, i8mm, bf16): Document new options.
3254 2020-01-02 Jakub Jelinek <jakub@redhat.com>
3257 * predict.c (compute_function_frequency): Don't call
3258 warn_function_cold on functions that already have cold attribute.
3260 2020-01-01 John David Anglin <danglin@gcc.gnu.org>
3263 * config/pa/pa.c (pa_elf_select_rtx_section): New. Put references to
3264 COMDAT group function labels in .data.rel.ro.local section.
3265 * config/pa/pa32-linux.h (TARGET_ASM_SELECT_RTX_SECTION): Define.
3268 * config/pa/pa.md (scc): Use ordered_comparison_operator instead of
3269 comparison_operator in B and S integer comparisons. Likewise, use
3270 ordered_comparison_operator instead of cmpib_comparison_operator in
3272 * config/pa/predicates.md (cmpib_comparison_operator): Remove.
3274 2020-01-01 Jakub Jelinek <jakub@redhat.com>
3276 Update copyright years.
3278 * gcc.c (process_command): Update copyright notice dates.
3279 * gcov-dump.c (print_version): Ditto.
3280 * gcov.c (print_version): Ditto.
3281 * gcov-tool.c (print_version): Ditto.
3282 * gengtype.c (create_file): Ditto.
3283 * doc/cpp.texi: Bump @copying's copyright year.
3284 * doc/cppinternals.texi: Ditto.
3285 * doc/gcc.texi: Ditto.
3286 * doc/gccint.texi: Ditto.
3287 * doc/gcov.texi: Ditto.
3288 * doc/install.texi: Ditto.
3289 * doc/invoke.texi: Ditto.
3291 2020-01-01 Jan Hubicka <hubicka@ucw.cz>
3293 * ipa.c (walk_polymorphic_call_targets): Fix updating of overall
3296 2020-01-01 Jakub Jelinek <jakub@redhat.com>
3298 PR tree-optimization/93098
3299 * match.pd (popcount): For shift amounts, use integer_onep
3300 or wi::to_widest () == cst instead of tree_to_uhwi () == cst
3301 tests. Make sure that precision is power of two larger than or equal
3302 to 16. Ensure shift is never negative. Use HOST_WIDE_INT_UC macro
3303 instead of ULL suffixed constants. Formatting fixes.
3305 Copyright (C) 2020 Free Software Foundation, Inc.
3307 Copying and distribution of this file, with or without modification,
3308 are permitted in any medium without royalty provided the copyright
3309 notice and this notice are preserved.