1 2020-01-14 Uroš Bizjak <ubizjak@gmail.com>
4 * config/i386/i386.md (*movsf_internal): Require SSE2 ISA for
7 2020-01-14 David Malcolm <dmalcolm@redhat.com>
9 * attribs.c (excl_hash_traits::empty_zero_p): New static constant.
10 * gcov.c (function_start_pair_hash::empty_zero_p): Likewise.
11 * graphite.c (struct sese_scev_hash::empty_zero_p): Likewise.
12 * hash-map-tests.c (selftest::test_nonzero_empty_key): New selftest.
13 (selftest::hash_map_tests_c_tests): Call it.
14 * hash-map-traits.h (simple_hashmap_traits::empty_zero_p):
15 New static constant, using the value of = H::empty_zero_p.
16 (unbounded_hashmap_traits::empty_zero_p): Likewise, using the value
17 from default_hash_traits <Value>.
18 * hash-map.h (hash_map::empty_zero_p): Likewise, using the value
20 * hash-set-tests.c (value_hash_traits::empty_zero_p): Likewise.
21 * hash-table.h (hash_table::alloc_entries): Guard the loop of
22 calls to mark_empty with !Descriptor::empty_zero_p.
23 (hash_table::empty_slow): Conditionalize the memset call with a
24 check that Descriptor::empty_zero_p; otherwise, loop through the
25 entries calling mark_empty on them.
26 * hash-traits.h (int_hash::empty_zero_p): New static constant.
27 (pointer_hash::empty_zero_p): Likewise.
28 (pair_hash::empty_zero_p): Likewise.
29 * ipa-devirt.c (default_hash_traits <type_pair>::empty_zero_p):
31 * ipa-prop.c (ipa_bit_ggc_hash_traits::empty_zero_p): Likewise.
32 (ipa_vr_ggc_hash_traits::empty_zero_p): Likewise.
33 * profile.c (location_triplet_hash::empty_zero_p): Likewise.
34 * sanopt.c (sanopt_tree_triplet_hash::empty_zero_p): Likewise.
35 (sanopt_tree_couple_hash::empty_zero_p): Likewise.
36 * tree-hasher.h (int_tree_hasher::empty_zero_p): Likewise.
37 * tree-ssa-sccvn.c (vn_ssa_aux_hasher::empty_zero_p): Likewise.
38 * tree-vect-slp.c (bst_traits::empty_zero_p): Likewise.
40 (default_hash_traits<scalar_cond_masked_key>::empty_zero_p):
43 2020-01-14 Kewen Lin <linkw@gcc.gnu.org>
45 * cfgloopanal.c (average_num_loop_insns): Free bbs when early return,
46 fix typo on return value.
48 2020-01-14 Xiong Hu Luo <luoxhu@linux.ibm.com>
51 * cgraph.c (symbol_table::create_edge): Init speculative_id and
53 (cgraph_edge::make_speculative): Add param for setting speculative_id
55 (cgraph_edge::speculative_call_info): Update comments and find reference
56 by speculative_id for multiple indirect targets.
57 (cgraph_edge::resolve_speculation): Decrease the speculations
58 for indirect edge, drop it's speculative if not direct target
59 left. Update comments.
60 (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
61 (cgraph_node::dump): Print num_speculative_call_targets.
62 (cgraph_node::verify_node): Don't report error if speculative
63 edge not include statement.
64 (cgraph_edge::num_speculative_call_targets_p): New function.
65 * cgraph.h (int common_target_id): Remove.
66 (int common_target_probability): Remove.
67 (num_speculative_call_targets): New variable.
68 (make_speculative): Add param for setting speculative_id.
69 (cgraph_edge::num_speculative_call_targets_p): New declare.
70 (target_prob): New variable.
71 (speculative_id): New variable.
72 * ipa-fnsummary.c (analyze_function_body): Create and duplicate
73 call summaries for multiple speculative call targets.
74 * cgraphclones.c (cgraph_node::create_clone): Clone speculative_id.
75 * ipa-profile.c (struct speculative_call_target): New struct.
76 (class speculative_call_summary): New class.
77 (class speculative_call_summaries): New class.
78 (call_sums): New variable.
79 (ipa_profile_generate_summary): Generate indirect multiple targets summaries.
80 (ipa_profile_write_edge_summary): New function.
81 (ipa_profile_write_summary): Stream out indirect multiple targets summaries.
82 (ipa_profile_dump_all_summaries): New function.
83 (ipa_profile_read_edge_summary): New function.
84 (ipa_profile_read_summary_section): New function.
85 (ipa_profile_read_summary): Stream in indirect multiple targets summaries.
86 (ipa_profile): Generate num_speculative_call_targets from
88 * ipa-ref.h (speculative_id): New variable.
89 * ipa-utils.c (ipa_merge_profiles): Update with target_prob.
90 * lto-cgraph.c (lto_output_edge): Remove indirect common_target_id and
91 common_target_probability. Stream out speculative_id and
92 num_speculative_call_targets.
93 (input_edge): Likewise.
94 * predict.c (dump_prediction): Remove edges count assert to be
96 * symtab.c (symtab_node::create_reference): Init speculative_id.
97 (symtab_node::clone_references): Clone speculative_id.
98 (symtab_node::clone_referring): Clone speculative_id.
99 (symtab_node::clone_reference): Clone speculative_id.
100 (symtab_node::clear_stmts_in_references): Clear speculative_id.
101 * tree-inline.c (copy_bb): Duplicate all the speculative edges
102 if indirect call contains multiple speculative targets.
103 * value-prof.h (check_ic_target): Remove.
104 * value-prof.c (gimple_value_profile_transformations):
105 Use void function gimple_ic_transform.
106 * value-prof.c (gimple_ic_transform): Handle topn case.
107 Fix comment typos. Change it to a void function.
109 2020-01-13 Andrew Pinski <apinski@marvell.com>
111 * config/aarch64/aarch64-cores.def (octeontx2): New define.
112 (octeontx2t98): New define.
113 (octeontx2t96): New define.
114 (octeontx2t93): New define.
115 (octeontx2f95): New define.
116 (octeontx2f95n): New define.
117 (octeontx2f95mm): New define.
118 * config/aarch64/aarch64-tune.md: Regenerate.
119 * doc/invoke.texi (-mcpu=): Document the new cpu types.
121 2020-01-13 Jason Merrill <jason@redhat.com>
123 PR c++/33799 - destroy return value if local cleanup throws.
124 * gimplify.c (gimplify_return_expr): Handle COMPOUND_EXPR.
126 2020-01-13 Martin Liska <mliska@suse.cz>
128 * ipa-cp.c (get_max_overall_size): Use newly
129 renamed param param_ipa_cp_unit_growth.
130 * params.opt: Remove legacy param name.
132 2020-01-13 Martin Sebor <msebor@redhat.com>
134 PR tree-optimization/93213
135 * tree-ssa-strlen.c (handle_store): Only allow single-byte nul-over-nul
136 stores to be eliminated.
138 2020-01-13 Martin Liska <mliska@suse.cz>
140 * opts.c (print_help): Do not print CL_PARAM
141 and CL_WARNING for CL_OPTIMIZATION.
143 2020-01-13 Jonathan Wakely <jwakely@redhat.com>
146 * doc/invoke.texi (Warning Options): Add caveat about some warnings
147 depending on optimization settings.
149 2020-01-13 Jakub Jelinek <jakub@redhat.com>
151 PR tree-optimization/90838
152 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
153 SCALAR_INT_TYPE_MODE directly in CTZ_DEFINED_VALUE_AT_ZERO macro
154 argument rather than to initialize temporary for targets that
155 don't use the mode argument at all. Initialize ctzval to avoid
158 2020-01-10 Thomas Schwinge <thomas@codesourcery.com>
160 * tree.h (OMP_CLAUSE_USE_DEVICE_PTR_IF_PRESENT): New definition.
161 * tree-core.h: Document it.
162 * gimplify.c (gimplify_omp_workshare): Set it.
163 * omp-low.c (lower_omp_target): Use it.
164 * tree-pretty-print.c (dump_omp_clause): Print it.
166 * omp-low.c (lower_omp_target) <OMP_CLAUSE_USE_DEVICE_PTR etc.>:
167 Assert that for OpenACC we always have 'GOMP_MAP_USE_DEVICE_PTR'.
169 2020-01-10 David Malcolm <dmalcolm@redhat.com>
171 * Makefile.in (OBJS): Add tree-diagnostic-path.o.
172 * common.opt (fdiagnostics-path-format=): New option.
173 (diagnostic_path_format): New enum.
174 (fdiagnostics-show-path-depths): New option.
175 * coretypes.h (diagnostic_event_id_t): New forward decl.
176 * diagnostic-color.c (color_dict): Add "path".
177 * diagnostic-event-id.h: New file.
178 * diagnostic-format-json.cc (json_from_expanded_location): Make
180 (json_end_diagnostic): Call context->make_json_for_path if it
181 exists and the diagnostic has a path.
182 (diagnostic_output_format_init): Clear context->print_path.
183 * diagnostic-path.h: New file.
184 * diagnostic-show-locus.c (colorizer::set_range): Special-case
185 when printing a run of events in a diagnostic_path so that they
186 all get the same color.
187 (layout::m_diagnostic_path_p): New field.
188 (layout::layout): Initialize it.
189 (layout::print_any_labels): Don't colorize the label text for an
190 event in a diagnostic_path.
191 (gcc_rich_location::add_location_if_nearby): Add
192 "restrict_to_current_line_spans" and "label" params. Pass the
193 former to layout.maybe_add_location_range; pass the latter
194 when calling add_range.
195 * diagnostic.c: Include "diagnostic-path.h".
196 (diagnostic_initialize): Initialize context->path_format and
197 context->show_path_depths.
198 (diagnostic_show_any_path): New function.
199 (diagnostic_path::interprocedural_p): New function.
200 (diagnostic_report_diagnostic): Call diagnostic_show_any_path.
201 (simple_diagnostic_path::num_events): New function.
202 (simple_diagnostic_path::get_event): New function.
203 (simple_diagnostic_path::add_event): New function.
204 (simple_diagnostic_event::simple_diagnostic_event): New ctor.
205 (simple_diagnostic_event::~simple_diagnostic_event): New dtor.
206 (debug): New overload taking a diagnostic_path *.
207 * diagnostic.def (DK_DIAGNOSTIC_PATH): New.
208 * diagnostic.h (enum diagnostic_path_format): New enum.
209 (json::value): New forward decl.
210 (diagnostic_context::path_format): New field.
211 (diagnostic_context::show_path_depths): New field.
212 (diagnostic_context::print_path): New callback field.
213 (diagnostic_context::make_json_for_path): New callback field.
214 (diagnostic_show_any_path): New decl.
215 (json_from_expanded_location): New decl.
216 * doc/invoke.texi (-fdiagnostics-path-format=): New option.
217 (-fdiagnostics-show-path-depths): New option.
218 (-fdiagnostics-color): Add "path" to description of default
219 GCC_COLORS; describe it.
220 (-fdiagnostics-format=json): Document how diagnostic paths are
221 represented in the JSON output format.
222 * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
223 Add optional params "restrict_to_current_line_spans" and "label".
224 * opts.c (common_handle_option): Handle
225 OPT_fdiagnostics_path_format_ and
226 OPT_fdiagnostics_show_path_depths.
227 * pretty-print.c: Include "diagnostic-event-id.h".
228 (pp_format): Implement "%@" format code for printing
229 diagnostic_event_id_t *.
230 (selftest::test_pp_format): Add tests for "%@".
231 * selftest-run-tests.c (selftest::run_tests): Call
232 selftest::tree_diagnostic_path_cc_tests.
233 * selftest.h (selftest::tree_diagnostic_path_cc_tests): New decl.
234 * toplev.c (general_init): Initialize global_dc->path_format and
235 global_dc->show_path_depths.
236 * tree-diagnostic-path.cc: New file.
237 * tree-diagnostic.c (maybe_unwind_expanded_macro_loc): Make
238 non-static. Drop "diagnostic" param in favor of storing the
239 original value of "where" and re-using it.
240 (virt_loc_aware_diagnostic_finalizer): Update for dropped param of
241 maybe_unwind_expanded_macro_loc.
242 (tree_diagnostics_defaults): Initialize context->print_path and
243 context->make_json_for_path.
244 * tree-diagnostic.h (default_tree_diagnostic_path_printer): New
246 (default_tree_make_json_for_path): New decl.
247 (maybe_unwind_expanded_macro_loc): New decl.
249 2020-01-10 Jakub Jelinek <jakub@redhat.com>
251 PR tree-optimization/93210
252 * fold-const.h (native_encode_initializer,
253 can_native_interpret_type_p): Declare.
254 * fold-const.c (native_encode_string): Fix up handling with off != -1,
256 (native_encode_initializer): New function, moved from dwarf2out.c.
257 Adjust to native_encode_expr compatible arguments, including dry-run
258 and partial extraction modes. Don't handle STRING_CST.
259 (can_native_interpret_type_p): No longer static.
260 * gimple-fold.c (fold_ctor_reference): For native_encode_expr, verify
261 offset / BITS_PER_UNIT fits into int and don't call it if
262 can_native_interpret_type_p fails. If suboff is NULL and for
263 CONSTRUCTOR fold_{,non}array_ctor_reference returns NULL, retry with
264 native_encode_initializer.
265 (fold_const_aggregate_ref_1): Formatting fix.
266 * dwarf2out.c (native_encode_initializer): Moved to fold-const.c.
267 (tree_add_const_value_attribute): Adjust caller.
269 PR tree-optimization/90838
270 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
271 SCALAR_INT_TYPE_MODE instead of TYPE_MODE as operand of
272 CTZ_DEFINED_VALUE_AT_ZERO.
274 2020-01-10 Vladimir Makarov <vmakarov@redhat.com>
277 * lra-constraints.c (match_reload): Permit input operands have the
278 same mode as output while other input operands have a different
281 2020-01-10 Wilco Dijkstra <wdijkstr@arm.com>
283 PR tree-optimization/90838
284 * tree-ssa-forwprop.c (check_ctz_array): Add new function.
285 (check_ctz_string): Likewise.
286 (optimize_count_trailing_zeroes): Likewise.
287 (simplify_count_trailing_zeroes): Likewise.
288 (pass_forwprop::execute): Try ctz simplification.
289 * match.pd: Add matching for ctz idioms.
291 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
293 * config/aarch64/aarch64.c (aarch64_invalid_conversion): New function
295 (aarch64_invalid_unary_op): New function for target hook.
296 (aarch64_invalid_binary_op): New function for target hook.
298 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
300 * config.gcc: Add arm_bf16.h.
301 * config/aarch64/aarch64-builtins.c
302 (aarch64_simd_builtin_std_type): Add BFmode.
303 (aarch64_init_simd_builtin_types): Define element types for vector
305 (aarch64_init_bf16_types): New function.
306 (aarch64_general_init_builtins): Add arm_init_bf16_types function call.
307 * config/aarch64/aarch64-modes.def: Add BFmode and V4BF, V8BF vector
309 * config/aarch64/aarch64-simd-builtin-types.def: Add BF SIMD types.
310 * config/aarch64/aarch64-simd.md: Add BF vector types to NEON move
312 * config/aarch64/aarch64.h (AARCH64_VALID_SIMD_DREG_MODE): Add V4BF.
313 (AARCH64_VALID_SIMD_QREG_MODE): Add V8BF.
314 * config/aarch64/aarch64.c
315 (aarch64_classify_vector_mode): Add support for BF types.
316 (aarch64_gimplify_va_arg_expr): Add support for BF types.
317 (aarch64_vq_mode): Add support for BF types.
318 (aarch64_simd_container_mode): Add support for BF types.
319 (aarch64_mangle_type): Add support for BF scalar type.
320 * config/aarch64/aarch64.md: Add BFmode to movhf pattern.
321 * config/aarch64/arm_bf16.h: New file.
322 * config/aarch64/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
323 * config/aarch64/iterators.md: Add BF types to mode attributes.
324 (HFBF, GPF_TF_F16_MOV, VDMOV, VQMOV, VQMOV_NO2Em VALL_F16MOV): New.
326 2020-01-10 Jason Merrill <jason@redhat.com>
328 PR c++/93173 - incorrect tree sharing.
329 * gimplify.c (copy_if_shared): No longer static.
330 * gimplify.h: Declare it.
332 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
334 * doc/invoke.texi (-msve-vector-bits=): Document that
335 -msve-vector-bits=128 now generates VL-specific code for
336 little-endian targets.
337 * config/aarch64/aarch64-sve-builtins.cc (register_builtin_types): Use
338 build_vector_type_for_mode to construct the data vector types.
339 * config/aarch64/aarch64.c (aarch64_convert_sve_vector_bits): Generate
340 VL-specific code for -msve-vector-bits=128 on little-endian targets.
341 (aarch64_simd_container_mode): Always prefer Advanced SIMD modes
344 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
346 * config/aarch64/aarch64.c (aarch64_evpc_sel): Fix gen_vcond_mask
349 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
351 * config/aarch64/aarch64-builtins.c
352 (aarch64_builtin_vectorized_function): Check for specific vector modes,
353 rather than checking the number of elements and the element mode.
355 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
357 * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
358 get_related_vectype_for_scalar_type rather than build_vector_type
359 to create the index type for a conditional reduction.
361 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
363 * tree-vect-loop.c (update_epilogue_loop_vinfo): Update DR_REF
364 for any type of gather or scatter, including strided accesses.
366 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
368 * tree-vectorizer.h (get_dr_vinfo_offset): Add missing function
371 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
373 * tree-vect-data-refs.c (vect_create_addr_base_for_vector_ref): Use
375 * tree-vect-loop.c (update_epilogue_loop_vinfo): Remove orig_drs_init
376 parameter and its use to reset DR_OFFSET's.
377 (vect_transform_loop): Remove orig_drs_init argument.
378 * tree-vect-loop-manip.c (vect_update_init_of_dr): Update the offset
379 member of dr_vec_info rather than the offset of the associated
380 data_reference's innermost_loop_behavior.
381 (vect_update_init_of_dr): Pass dr_vec_info instead of data_reference.
382 (vect_do_peeling): Remove orig_drs_init parameter and its construction.
383 * tree-vect-stmts.c (check_scan_store): Replace use of DR_OFFSET with
385 (vectorizable_store): Likewise.
386 (vectorizable_load): Likewise.
388 2020-01-10 Richard Biener <rguenther@suse.de>
390 * gimple-ssa-store-merging
391 (pass_store_merging::terminate_all_aliasing_chains): Cache alias info.
393 2020-01-10 Martin Liska <mliska@suse.cz>
396 * ipa-inline-analysis.c (offline_size): Make proper parenthesis
397 encapsulation that was there before r280040.
399 2020-01-10 Richard Biener <rguenther@suse.de>
402 * tree-eh.c (sink_clobbers): Move clobbers to out-of-IL
403 sequences to avoid walking them again for secondary opportunities.
404 (pass_lower_eh_dispatch::execute): Instead actually insert
407 2020-01-10 Richard Biener <rguenther@suse.de>
410 * tree-eh.c (redirect_eh_edge_1): Avoid some work if possible.
411 (cleanup_all_empty_eh): Walk landing pads in reverse order to
414 2020-01-10 Martin Jambor <mjambor@suse.cz>
416 * params.opt (param_ipa_sra_max_replacements): Mark as Optimization.
417 * ipa-sra.c (pull_accesses_from_callee): New parameter caller, use it
418 to get param_ipa_sra_max_replacements.
419 (param_splitting_across_edge): Pass the caller to
420 pull_accesses_from_callee.
422 2020-01-10 Martin Jambor <mjambor@suse.cz>
424 * params.opt (param_ipcp_unit_growth): Mark as Optimization.
425 * ipa-cp.c (max_new_size): Removed.
426 (orig_overall_size): New variable.
427 (get_max_overall_size): New function.
428 (estimate_local_effects): Use it. Adjust dump.
429 (decide_about_value): Likewise.
430 (ipcp_propagate_stage): Do not calculate max_new_size, just store
431 orig_overall_size. Adjust dump.
432 (ipa_cp_c_finalize): Clear orig_overall_size instead of max_new_size.
434 2020-01-10 Martin Jambor <mjambor@suse.cz>
436 * params.opt (param_ipa_max_agg_items): Mark as Optimization
437 * ipa-cp.c (merge_agg_lats_step): New parameter max_agg_items, use
438 instead of param_ipa_max_agg_items.
439 (merge_aggregate_lattices): Extract param_ipa_max_agg_items from
440 optimization info for the callee.
442 2020-01-09 Kwok Cheung Yeung <kcy@codesourcery.com>
444 * lto-streamer-in.c (input_function): Remove streamed-in inline debug
445 markers if debug_inline_points is false.
447 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
449 * config.gcc (aarch64*-*-*): Add aarch64-sve-builtins-sve2.o to
451 * config/aarch64/t-aarch64 (aarch64-sve-builtins.o): Depend on
452 aarch64-sve-builtins-base.def, aarch64-sve-builtins-sve2.def and
453 aarch64-sve-builtins-sve2.h.
454 (aarch64-sve-builtins-sve2.o): New rule.
455 * config/aarch64/aarch64.h (AARCH64_ISA_SVE2_AES): New macro.
456 (AARCH64_ISA_SVE2_BITPERM, AARCH64_ISA_SVE2_SHA3): Likewise.
457 (AARCH64_ISA_SVE2_SM4, TARGET_SVE2_AES, TARGET_SVE2_BITPERM): Likewise.
458 (TARGET_SVE2_SHA, TARGET_SVE2_SM4): Likewise.
459 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
460 TARGET_SVE2_AES, TARGET_SVE2_BITPERM, TARGET_SVE2_SHA3 and
462 * config/aarch64/aarch64-sve.md: Update comments with SVE2
463 instructions that are handled here.
464 (@cond_asrd<mode>): Generalize to...
465 (@cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>): ...this.
466 (*cond_asrd<mode>_2): Generalize to...
467 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_2): ...this.
468 (*cond_asrd<mode>_z): Generalize to...
469 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_z): ...this.
470 * config/aarch64/aarch64.md (UNSPEC_LDNT1_GATHER): New unspec.
471 (UNSPEC_STNT1_SCATTER, UNSPEC_WHILEGE, UNSPEC_WHILEGT): Likewise.
472 (UNSPEC_WHILEHI, UNSPEC_WHILEHS): Likewise.
473 * config/aarch64/aarch64-sve2.md (@aarch64_gather_ldnt<mode>): New
475 (@aarch64_gather_ldnt_<ANY_EXTEND:optab><SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
476 (@aarch64_scatter_stnt<mode>): Likewise.
477 (@aarch64_scatter_stnt_<SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
478 (@aarch64_mul_lane_<mode>): Likewise.
479 (@aarch64_sve_suqadd<mode>_const): Likewise.
480 (*<sur>h<addsub><mode>): Generalize to...
481 (@aarch64_pred_<SVE2_COND_INT_BINARY_REV:sve_int_op><mode>): ...this
483 (@cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>): New expander.
484 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_2): New pattern.
485 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_3): Likewise.
486 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_any): Likewise.
487 (*cond_<SVE2_COND_INT_BINARY_NOREV:sve_int_op><mode>_z): Likewise.
488 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op><mode>):: Likewise.
489 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op>_lane_<mode>): Likewise.
490 (@aarch64_pred_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): Likewise.
491 (@cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): New expander.
492 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_2): New pattern.
493 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_3): Likewise.
494 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_any): Likewise.
495 (@aarch64_sve_<SVE2_INT_TERNARY:sve_int_op><mode>): Likewise.
496 (@aarch64_sve_<SVE2_INT_TERNARY_LANE:sve_int_op>_lane_<mode>)
497 (@aarch64_sve_add_mul_lane_<mode>): Likewise.
498 (@aarch64_sve_sub_mul_lane_<mode>): Likewise.
499 (@aarch64_sve2_xar<mode>): Likewise.
500 (@aarch64_sve2_bcax<mode>): Likewise.
501 (*aarch64_sve2_eor3<mode>): Rename to...
502 (@aarch64_sve2_eor3<mode>): ...this.
503 (@aarch64_sve2_bsl<mode>): New expander.
504 (@aarch64_sve2_nbsl<mode>): Likewise.
505 (@aarch64_sve2_bsl1n<mode>): Likewise.
506 (@aarch64_sve2_bsl2n<mode>): Likewise.
507 (@aarch64_sve_add_<SHIFTRT:sve_int_op><mode>): Likewise.
508 (*aarch64_sve2_sra<mode>): Add MOVPRFX support.
509 (@aarch64_sve_add_<VRSHR_N:sve_int_op><mode>): New pattern.
510 (@aarch64_sve_<SVE2_INT_SHIFT_INSERT:sve_int_op><mode>): Likewise.
511 (@aarch64_sve2_<USMAX:su>aba<mode>): New expander.
512 (*aarch64_sve2_<USMAX:su>aba<mode>): New pattern.
513 (@aarch64_sve_<SVE2_INT_BINARY_WIDE:sve_int_op><mode>): Likewise.
514 (<su>mull<bt><Vwide>): Generalize to...
515 (@aarch64_sve_<SVE2_INT_BINARY_LONG:sve_int_op><mode>): ...this new
517 (@aarch64_sve_<SVE2_INT_BINARY_LONG_lANE:sve_int_op>_lane_<mode>)
518 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_LONG:sve_int_op><mode>)
519 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG:sve_int_op><mode>)
520 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
521 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG:sve_int_op><mode>)
522 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
523 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG:sve_int_op><mode>)
524 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
525 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG:sve_int_op><mode>)
526 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
527 (@aarch64_sve_<SVE2_FP_TERNARY_LONG:sve_fp_op><mode>): New patterns.
528 (@aarch64_<SVE2_FP_TERNARY_LONG_LANE:sve_fp_op>_lane_<mode>)
529 (@aarch64_sve_<SVE2_INT_UNARY_NARROWB:sve_int_op><mode>): Likewise.
530 (@aarch64_sve_<SVE2_INT_UNARY_NARROWT:sve_int_op><mode>): Likewise.
531 (@aarch64_sve_<SVE2_INT_BINARY_NARROWB:sve_int_op><mode>): Likewise.
532 (@aarch64_sve_<SVE2_INT_BINARY_NARROWT:sve_int_op><mode>): Likewise.
533 (<SHRNB:r>shrnb<mode>): Generalize to...
534 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWB:sve_int_op><mode>): ...this
536 (<SHRNT:r>shrnt<mode>): Generalize to...
537 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWT:sve_int_op><mode>): ...this
539 (@aarch64_pred_<SVE2_INT_BINARY_PAIR:sve_int_op><mode>): New pattern.
540 (@aarch64_pred_<SVE2_FP_BINARY_PAIR:sve_fp_op><mode>): Likewise.
541 (@cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>): New expander.
542 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_2): New pattern.
543 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_z): Likewise.
544 (@aarch64_sve_<SVE2_INT_CADD:optab><mode>): Likewise.
545 (@aarch64_sve_<SVE2_INT_CMLA:optab><mode>): Likewise.
546 (@aarch64_<SVE2_INT_CMLA:optab>_lane_<mode>): Likewise.
547 (@aarch64_sve_<SVE2_INT_CDOT:optab><mode>): Likewise.
548 (@aarch64_<SVE2_INT_CDOT:optab>_lane_<mode>): Likewise.
549 (@aarch64_pred_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): Likewise.
550 (@cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New expander.
551 (*cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New pattern.
552 (@aarch64_sve2_cvtnt<mode>): Likewise.
553 (@aarch64_pred_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): Likewise.
554 (@cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): New expander.
555 (*cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>_any): New pattern.
556 (@aarch64_sve2_cvtxnt<mode>): Likewise.
557 (@aarch64_pred_<SVE2_U32_UNARY:sve_int_op><mode>): Likewise.
558 (@cond_<SVE2_U32_UNARY:sve_int_op><mode>): New expander.
559 (*cond_<SVE2_U32_UNARY:sve_int_op><mode>): New pattern.
560 (@aarch64_pred_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): Likewise.
561 (@cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New expander.
562 (*cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New pattern.
563 (@aarch64_sve2_pmul<mode>): Likewise.
564 (@aarch64_sve_<SVE2_PMULL:optab><mode>): Likewise.
565 (@aarch64_sve_<SVE2_PMULL_PAIR:optab><mode>): Likewise.
566 (@aarch64_sve2_tbl2<mode>): Likewise.
567 (@aarch64_sve2_tbx<mode>): Likewise.
568 (@aarch64_sve_<SVE2_INT_BITPERM:sve_int_op><mode>): Likewise.
569 (@aarch64_sve2_histcnt<mode>): Likewise.
570 (@aarch64_sve2_histseg<mode>): Likewise.
571 (@aarch64_pred_<SVE2_MATCH:sve_int_op><mode>): Likewise.
572 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_cc): Likewise.
573 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_ptest): Likewise.
574 (aarch64_sve2_aes<CRYPTO_AES:aes_op>): Likewise.
575 (aarch64_sve2_aes<CRYPTO_AESMC:aesmc_op>): Likewise.
576 (*aarch64_sve2_aese_fused, *aarch64_sve2_aesd_fused): Likewise.
577 (aarch64_sve2_rax1, aarch64_sve2_sm4e, aarch64_sve2_sm4ekey): Likewise.
578 (<su>mulh<r>s<mode>3): Update after above pattern name changes.
579 * config/aarch64/iterators.md (VNx16QI_ONLY, VNx4SF_ONLY)
580 (SVE_STRUCT2, SVE_FULL_BHI, SVE_FULL_HSI, SVE_FULL_HDI)
581 (SVE2_PMULL_PAIR_I): New mode iterators.
582 (UNSPEC_ADCLB, UNSPEC_ADCLT, UNSPEC_ADDHNB, UNSPEC_ADDHNT, UNSPEC_BDEP)
583 (UNSPEC_BEXT, UNSPEC_BGRP, UNSPEC_CADD90, UNSPEC_CADD270, UNSPEC_CDOT)
584 (UNSPEC_CDOT90, UNSPEC_CDOT180, UNSPEC_CDOT270, UNSPEC_CMLA)
585 (UNSPEC_CMLA90, UNSPEC_CMLA180, UNSPEC_CMLA270, UNSPEC_COND_FCVTLT)
586 (UNSPEC_COND_FCVTNT, UNSPEC_COND_FCVTX, UNSPEC_COND_FCVTXNT)
587 (UNSPEC_COND_FLOGB, UNSPEC_EORBT, UNSPEC_EORTB, UNSPEC_FADDP)
588 (UNSPEC_FMAXP, UNSPEC_FMAXNMP, UNSPEC_FMLALB, UNSPEC_FMLALT)
589 (UNSPEC_FMLSLB, UNSPEC_FMLSLT, UNSPEC_FMINP, UNSPEC_FMINNMP)
590 (UNSPEC_HISTCNT, UNSPEC_HISTSEG, UNSPEC_MATCH, UNSPEC_NMATCH)
591 (UNSPEC_PMULLB, UNSPEC_PMULLB_PAIR, UNSPEC_PMULLT, UNSPEC_PMULLT_PAIR)
592 (UNSPEC_RADDHNB, UNSPEC_RADDHNT, UNSPEC_RSUBHNB, UNSPEC_RSUBHNT)
593 (UNSPEC_SLI, UNSPEC_SRI, UNSPEC_SABDLB, UNSPEC_SABDLT, UNSPEC_SADDLB)
594 (UNSPEC_SADDLBT, UNSPEC_SADDLT, UNSPEC_SADDWB, UNSPEC_SADDWT)
595 (UNSPEC_SBCLB, UNSPEC_SBCLT, UNSPEC_SMAXP, UNSPEC_SMINP)
596 (UNSPEC_SQCADD90, UNSPEC_SQCADD270, UNSPEC_SQDMULLB, UNSPEC_SQDMULLBT)
597 (UNSPEC_SQDMULLT, UNSPEC_SQRDCMLAH, UNSPEC_SQRDCMLAH90)
598 (UNSPEC_SQRDCMLAH180, UNSPEC_SQRDCMLAH270, UNSPEC_SQRSHRNB)
599 (UNSPEC_SQRSHRNT, UNSPEC_SQRSHRUNB, UNSPEC_SQRSHRUNT, UNSPEC_SQSHRNB)
600 (UNSPEC_SQSHRNT, UNSPEC_SQSHRUNB, UNSPEC_SQSHRUNT, UNSPEC_SQXTNB)
601 (UNSPEC_SQXTNT, UNSPEC_SQXTUNB, UNSPEC_SQXTUNT, UNSPEC_SSHLLB)
602 (UNSPEC_SSHLLT, UNSPEC_SSUBLB, UNSPEC_SSUBLBT, UNSPEC_SSUBLT)
603 (UNSPEC_SSUBLTB, UNSPEC_SSUBWB, UNSPEC_SSUBWT, UNSPEC_SUBHNB)
604 (UNSPEC_SUBHNT, UNSPEC_TBL2, UNSPEC_UABDLB, UNSPEC_UABDLT)
605 (UNSPEC_UADDLB, UNSPEC_UADDLT, UNSPEC_UADDWB, UNSPEC_UADDWT)
606 (UNSPEC_UMAXP, UNSPEC_UMINP, UNSPEC_UQRSHRNB, UNSPEC_UQRSHRNT)
607 (UNSPEC_UQSHRNB, UNSPEC_UQSHRNT, UNSPEC_UQXTNB, UNSPEC_UQXTNT)
608 (UNSPEC_USHLLB, UNSPEC_USHLLT, UNSPEC_USUBLB, UNSPEC_USUBLT)
609 (UNSPEC_USUBWB, UNSPEC_USUBWT): New unspecs.
610 (UNSPEC_SMULLB, UNSPEC_SMULLT, UNSPEC_UMULLB, UNSPEC_UMULLT)
611 (UNSPEC_SMULHS, UNSPEC_SMULHRS, UNSPEC_UMULHS, UNSPEC_UMULHRS)
612 (UNSPEC_RSHRNB, UNSPEC_RSHRNT, UNSPEC_SHRNB, UNSPEC_SHRNT): Move
614 (VNARROW, Ventype): New mode attributes.
615 (Vewtype): Handle VNx2DI. Fix typo in comment.
616 (VDOUBLE): New mode attribute.
617 (sve_lane_con): Handle VNx8HI.
618 (SVE_INT_UNARY): Include ss_abs and ss_neg for TARGET_SVE2.
619 (SVE_INT_BINARY): Likewise ss_plus, us_plus, ss_minus and us_minus.
620 (sve_int_op, sve_int_op_rev): Handle the above codes.
621 (sve_pred_int_rhs2_operand): Likewise.
622 (MULLBT, SHRNB, SHRNT): Delete.
623 (SVE_INT_SHIFT_IMM): New int iterator.
624 (SVE_WHILE): Add UNSPEC_WHILEGE, UNSPEC_WHILEGT, UNSPEC_WHILEHI
625 and UNSPEC_WHILEHS for TARGET_SVE2.
626 (SVE2_U32_UNARY, SVE2_INT_UNARY_NARROWB, SVE2_INT_UNARY_NARROWT)
627 (SVE2_INT_BINARY, SVE2_INT_BINARY_LANE, SVE2_INT_BINARY_LONG)
628 (SVE2_INT_BINARY_LONG_LANE, SVE2_INT_BINARY_NARROWB)
629 (SVE2_INT_BINARY_NARROWT, SVE2_INT_BINARY_PAIR, SVE2_FP_BINARY_PAIR)
630 (SVE2_INT_BINARY_PAIR_LONG, SVE2_INT_BINARY_WIDE): New int iterators.
631 (SVE2_INT_SHIFT_IMM_LONG, SVE2_INT_SHIFT_IMM_NARROWB): Likewise.
632 (SVE2_INT_SHIFT_IMM_NARROWT, SVE2_INT_SHIFT_INSERT, SVE2_INT_CADD)
633 (SVE2_INT_BITPERM, SVE2_INT_TERNARY, SVE2_INT_TERNARY_LANE): Likewise.
634 (SVE2_FP_TERNARY_LONG, SVE2_FP_TERNARY_LONG_LANE, SVE2_INT_CMLA)
635 (SVE2_INT_CDOT, SVE2_INT_ADD_BINARY_LONG, SVE2_INT_QADD_BINARY_LONG)
636 (SVE2_INT_SUB_BINARY_LONG, SVE2_INT_QSUB_BINARY_LONG): Likewise.
637 (SVE2_INT_ADD_BINARY_LONG_LANE, SVE2_INT_QADD_BINARY_LONG_LANE)
638 (SVE2_INT_SUB_BINARY_LONG_LANE, SVE2_INT_QSUB_BINARY_LONG_LANE)
639 (SVE2_COND_INT_UNARY_FP, SVE2_COND_FP_UNARY_LONG): Likewise.
640 (SVE2_COND_FP_UNARY_NARROWB, SVE2_COND_INT_BINARY): Likewise.
641 (SVE2_COND_INT_BINARY_NOREV, SVE2_COND_INT_BINARY_REV): Likewise.
642 (SVE2_COND_INT_SHIFT, SVE2_MATCH, SVE2_PMULL): Likewise.
643 (optab): Handle the new unspecs.
644 (su, r): Remove entries for UNSPEC_SHRNB, UNSPEC_SHRNT, UNSPEC_RSHRNB
646 (lr): Handle the new unspecs.
648 (cmp_op, while_optab_cmp, sve_int_op): Handle the new unspecs.
649 (sve_int_op_rev, sve_int_add_op, sve_int_qadd_op, sve_int_sub_op)
650 (sve_int_qsub_op): New int attributes.
651 (sve_fp_op, rot): Handle the new unspecs.
652 * config/aarch64/aarch64-sve-builtins.h
653 (function_resolver::require_matching_pointer_type): Declare.
654 (function_resolver::resolve_unary): Add an optional boolean argument.
655 (function_resolver::finish_opt_n_resolution): Add an optional
656 type_suffix_index argument.
657 (gimple_folder::redirect_call): Declare.
658 (gimple_expander::prepare_gather_address_operands): Add an optional
660 * config/aarch64/aarch64-sve-builtins.cc: Include
661 aarch64-sve-builtins-sve2.h.
662 (TYPES_b_unsigned, TYPES_b_integer, TYPES_bh_integer): New macros.
663 (TYPES_bs_unsigned, TYPES_hs_signed, TYPES_hs_integer): Likewise.
664 (TYPES_hd_unsigned, TYPES_hsd_signed): Likewise.
665 (TYPES_hsd_integer): Use TYPES_hsd_signed.
666 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): New macros.
667 (TYPES_s_unsigned): Likewise.
668 (TYPES_s_integer): Use TYPES_s_unsigned.
669 (TYPES_sd_signed, TYPES_sd_unsigned): New macros.
670 (TYPES_sd_integer): Use them.
671 (TYPES_d_unsigned): New macro.
672 (TYPES_d_integer): Use it.
673 (TYPES_d_data, TYPES_cvt_long, TYPES_cvt_narrow_s): New macros.
674 (TYPES_cvt_narrow): Likewise.
675 (DEF_SVE_TYPES_ARRAY): Include the new types macros above.
676 (preds_mx): New variable.
677 (function_builder::add_overloaded_function): Allow the new feature
678 set to be more restrictive than the original one.
679 (function_resolver::infer_pointer_type): Remove qualifiers from
680 the pointer type before printing it.
681 (function_resolver::require_matching_pointer_type): New function.
682 (function_resolver::resolve_sv_displacement): Handle functions
683 that don't support 32-bit vector indices or svint32_t vector offsets.
684 (function_resolver::finish_opt_n_resolution): Take the inferred type
685 as a separate argument.
686 (function_resolver::resolve_unary): Optionally treat all forms in
687 the same way as normal merging functions.
688 (gimple_folder::redirect_call): New function.
689 (function_expander::prepare_gather_address_operands): Add an argument
690 that says whether scaled forms are available. If they aren't,
691 handle scaling of vector indices and don't add the extension and
693 (function_expander::map_to_unspecs): If aarch64_sve isn't available,
694 fall back to using cond_* instead.
695 * config/aarch64/aarch64-sve-builtins-functions.h (rtx_code_function):
696 Split out the member variables into...
697 (rtx_code_function_base): ...this new base class.
698 (rtx_code_function_rotated): Inherit rtx_code_function_base.
699 (unspec_based_function): Split out the member variables into...
700 (unspec_based_function_base): ...this new base class.
701 (unspec_based_function_rotated): Inherit unspec_based_function_base.
702 (unspec_based_function_exact_insn): New class.
703 (unspec_based_add_function, unspec_based_add_lane_function)
704 (unspec_based_lane_function, unspec_based_pred_function)
705 (unspec_based_qadd_function, unspec_based_qadd_lane_function)
706 (unspec_based_qsub_function, unspec_based_qsub_lane_function)
707 (unspec_based_sub_function, unspec_based_sub_lane_function): New
709 (unspec_based_fused_function): New class.
710 (unspec_based_mla_function, unspec_based_mls_function): New typedefs.
711 (unspec_based_fused_lane_function): New class.
712 (unspec_based_mla_lane_function, unspec_based_mls_lane_function): New
714 (CODE_FOR_MODE1): New macro.
715 (fixed_insn_function): New class.
716 (while_comparison): Likewise.
717 * config/aarch64/aarch64-sve-builtins-shapes.h (binary_long_lane)
718 (binary_long_opt_n, binary_narrowb_opt_n, binary_narrowt_opt_n)
719 (binary_to_uint, binary_wide, binary_wide_opt_n, compare, compare_ptr)
720 (load_ext_gather_index_restricted, load_ext_gather_offset_restricted)
721 (load_gather_sv_restricted, shift_left_imm_long): Declare.
722 (shift_left_imm_to_uint, shift_right_imm_narrowb): Likewise.
723 (shift_right_imm_narrowt, shift_right_imm_narrowb_to_uint): Likewise.
724 (shift_right_imm_narrowt_to_uint, store_scatter_index_restricted)
725 (store_scatter_offset_restricted, tbl_tuple, ternary_long_lane)
726 (ternary_long_opt_n, ternary_qq_lane_rotate, ternary_qq_rotate)
727 (ternary_shift_left_imm, ternary_shift_right_imm, ternary_uint)
728 (unary_convert_narrowt, unary_long, unary_narrowb, unary_narrowt)
729 (unary_narrowb_to_uint, unary_narrowt_to_uint, unary_to_int): Likewise.
730 * config/aarch64/aarch64-sve-builtins-shapes.cc (apply_predication):
731 Also add an initial argument for unary_convert_narrowt, regardless
732 of the predication type.
733 (build_32_64): Allow loads and stores to specify MODE_none.
734 (build_sv_index64, build_sv_uint_offset): New functions.
735 (long_type_suffix): New function.
736 (binary_imm_narrowb_base, binary_imm_narrowt_base): New classes.
737 (binary_imm_long_base, load_gather_sv_base): Likewise.
738 (shift_right_imm_narrow_wrapper, ternary_shift_imm_base): Likewise.
739 (ternary_resize2_opt_n_base, ternary_resize2_lane_base): Likewise.
740 (unary_narrowb_base, unary_narrowt_base): Likewise.
741 (binary_long_lane_def, binary_long_lane): New shape.
742 (binary_long_opt_n_def, binary_long_opt_n): Likewise.
743 (binary_narrowb_opt_n_def, binary_narrowb_opt_n): Likewise.
744 (binary_narrowt_opt_n_def, binary_narrowt_opt_n): Likewise.
745 (binary_to_uint_def, binary_to_uint): Likewise.
746 (binary_wide_def, binary_wide): Likewise.
747 (binary_wide_opt_n_def, binary_wide_opt_n): Likewise.
748 (compare_def, compare): Likewise.
749 (compare_ptr_def, compare_ptr): Likewise.
750 (load_ext_gather_index_restricted_def,
751 load_ext_gather_index_restricted): Likewise.
752 (load_ext_gather_offset_restricted_def,
753 load_ext_gather_offset_restricted): Likewise.
754 (load_gather_sv_def): Inherit from load_gather_sv_base.
755 (load_gather_sv_restricted_def, load_gather_sv_restricted): New shape.
756 (shift_left_imm_def, shift_left_imm): Likewise.
757 (shift_left_imm_long_def, shift_left_imm_long): Likewise.
758 (shift_left_imm_to_uint_def, shift_left_imm_to_uint): Likewise.
759 (store_scatter_index_restricted_def,
760 store_scatter_index_restricted): Likewise.
761 (store_scatter_offset_restricted_def,
762 store_scatter_offset_restricted): Likewise.
763 (tbl_tuple_def, tbl_tuple): Likewise.
764 (ternary_long_lane_def, ternary_long_lane): Likewise.
765 (ternary_long_opt_n_def, ternary_long_opt_n): Likewise.
766 (ternary_qq_lane_def): Inherit from ternary_resize2_lane_base.
767 (ternary_qq_lane_rotate_def, ternary_qq_lane_rotate): New shape
768 (ternary_qq_opt_n_def): Inherit from ternary_resize2_opt_n_base.
769 (ternary_qq_rotate_def, ternary_qq_rotate): New shape.
770 (ternary_shift_left_imm_def, ternary_shift_left_imm): Likewise.
771 (ternary_shift_right_imm_def, ternary_shift_right_imm): Likewise.
772 (ternary_uint_def, ternary_uint): Likewise.
773 (unary_convert): Fix typo in comment.
774 (unary_convert_narrowt_def, unary_convert_narrowt): New shape.
775 (unary_long_def, unary_long): Likewise.
776 (unary_narrowb_def, unary_narrowb): Likewise.
777 (unary_narrowt_def, unary_narrowt): Likewise.
778 (unary_narrowb_to_uint_def, unary_narrowb_to_uint): Likewise.
779 (unary_narrowt_to_uint_def, unary_narrowt_to_uint): Likewise.
780 (unary_to_int_def, unary_to_int): Likewise.
781 * config/aarch64/aarch64-sve-builtins-base.cc (unspec_cmla)
782 (unspec_fcmla, unspec_cond_fcmla, expand_mla_mls_lane): New functions.
783 (svasrd_impl): Delete.
784 (svcadd_impl::expand): Handle integer operations too.
785 (svcmla_impl::expand, svcmla_lane::expand): Likewise, using the
786 new functions to derive the unspec numbers.
787 (svmla_svmls_lane_impl): Replace with...
788 (svmla_lane_impl, svmls_lane_impl): ...these new classes. Handle
789 integer operations too.
790 (svwhile_impl): Rename to...
791 (svwhilelx_impl): ...this and inherit from while_comparison.
792 (svasrd): Use unspec_based_function.
793 (svmla_lane): Use svmla_lane_impl.
794 (svmls_lane): Use svmls_lane_impl.
795 (svrecpe, svrsqrte): Handle unsigned integer operations too.
796 (svwhilele, svwhilelt): Use svwhilelx_impl.
797 * config/aarch64/aarch64-sve-builtins-sve2.h: New file.
798 * config/aarch64/aarch64-sve-builtins-sve2.cc: Likewise.
799 * config/aarch64/aarch64-sve-builtins-sve2.def: Likewise.
800 * config/aarch64/aarch64-sve-builtins.def: Include
801 aarch64-sve-builtins-sve2.def.
803 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
805 * config/aarch64/aarch64-protos.h (aarch64_sve_arith_immediate_p)
806 (aarch64_sve_sqadd_sqsub_immediate_p): Add a machine_mode argument.
807 * config/aarch64/aarch64.c (aarch64_sve_arith_immediate_p)
808 (aarch64_sve_sqadd_sqsub_immediate_p): Likewise. Handle scalar
809 immediates as well as vector ones.
810 * config/aarch64/predicates.md (aarch64_sve_arith_immediate)
811 (aarch64_sve_sub_arith_immediate, aarch64_sve_qadd_immediate)
812 (aarch64_sve_qsub_immediate): Update calls accordingly.
814 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
816 * config/aarch64/aarch64-sve2.md: Add banner comments.
817 (<su>mulh<r>s<mode>3): Move further up file.
818 (<su>mull<bt><Vwide>, <r>shrnb<mode>, <r>shrnt<mode>)
819 (*aarch64_sve2_sra<mode>): Move further down file.
820 * config/aarch64/t-aarch64 (s-check-sve-md): Check aarch64-sve2.md too.
822 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
824 * config/aarch64/iterators.md (SVE_WHILE): Add UNSPEC_WHILERW
826 (while_optab_cmp): Handle them.
827 * config/aarch64/aarch64-sve.md
828 (*while_<while_optab_cmp><GPI:mode><PRED_ALL:mode>_ptest): Make public
829 and add a "@" marker.
830 * config/aarch64/aarch64-sve2.md (check_<raw_war>_ptrs<mode>): Use it
831 instead of gen_aarch64_sve2_while_ptest.
832 (@aarch64_sve2_while<cmp_op><GPI:mode><PRED_ALL:mode>_ptest): Delete.
834 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
836 * config/aarch64/aarch64.md (UNSPEC_WHILE_LE): Rename to...
837 (UNSPEC_WHILELE): ...this.
838 (UNSPEC_WHILE_LO): Rename to...
839 (UNSPEC_WHILELO): ...this.
840 (UNSPEC_WHILE_LS): Rename to...
841 (UNSPEC_WHILELS): ...this.
842 (UNSPEC_WHILE_LT): Rename to...
843 (UNSPEC_WHILELT): ...this.
844 * config/aarch64/iterators.md (SVE_WHILE): Update accordingly.
845 (cmp_op, while_optab_cmp): Likewise.
846 * config/aarch64/aarch64.c (aarch64_sve_move_pred_via_while): Likewise.
847 * config/aarch64/aarch64-sve-builtins-base.cc (svwhilele): Likewise.
848 (svwhilelt): Likewise.
850 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
852 * config/aarch64/aarch64-sve-builtins-shapes.h (unary_count): Delete.
853 (unary_to_uint): Define.
854 * config/aarch64/aarch64-sve-builtins-shapes.cc (unary_count_def)
855 (unary_count): Rename to...
856 (unary_to_uint_def, unary_to_uint): ...this.
857 * config/aarch64/aarch64-sve-builtins-base.def: Update accordingly.
859 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
861 * config/aarch64/aarch64-sve-builtins-functions.h
862 (code_for_mode_function): New class.
863 (CODE_FOR_MODE0, QUIET_CODE_FOR_MODE0): New macros.
864 * config/aarch64/aarch64-sve-builtins-base.cc (svcompact_impl)
865 (svext_impl, svmul_lane_impl, svsplice_impl, svtmad_impl): Delete.
866 (svcompact, svext, svsplice): Use QUIET_CODE_FOR_MODE0.
867 (svmul_lane, svtmad): Use CODE_FOR_MODE0.
869 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
871 * config/aarch64/iterators.md (addsub): New code attribute.
872 * config/aarch64/aarch64-simd.md (aarch64_<su_optab><optab><mode>):
874 (aarch64_<su_optab>q<addsub><mode>): ...this, making the same change
875 in the asm string and attributes. Fix indentation.
876 * config/aarch64/aarch64-sve.md (@aarch64_<su_optab><optab><mode>):
878 (@aarch64_sve_<optab><mode>): ...this.
879 * config/aarch64/aarch64-sve-builtins.h
880 (function_expander::expand_signed_unpred_op): Delete.
881 * config/aarch64/aarch64-sve-builtins.cc
882 (function_expander::expand_signed_unpred_op): Likewise.
883 (function_expander::map_to_rtx_codes): If the optab isn't defined,
884 try using code_for_aarch64_sve instead.
885 * config/aarch64/aarch64-sve-builtins-base.cc (svqadd_impl): Delete.
886 (svqsub_impl): Likewise.
887 (svqadd, svqsub): Use rtx_code_function instead.
889 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
891 * config/aarch64/iterators.md (SRHSUB, URHSUB): Delete.
892 (HADDSUB, sur, addsub): Remove them.
894 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
896 * tree-nrv.c (pass_return_slot::execute): Handle all internal
897 functions the same way, rather than singling out those that
898 aren't mapped directly to optabs.
900 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
902 * target.def (compatible_vector_types_p): New target hook.
903 * hooks.h (hook_bool_const_tree_const_tree_true): Declare.
904 * hooks.c (hook_bool_const_tree_const_tree_true): New function.
905 * doc/tm.texi.in (TARGET_COMPATIBLE_VECTOR_TYPES_P): New hook.
906 * doc/tm.texi: Regenerate.
907 * gimple-expr.c: Include target.h.
908 (useless_type_conversion_p): Use targetm.compatible_vector_types_p.
909 * config/aarch64/aarch64.c (aarch64_compatible_vector_types_p): New
911 (TARGET_COMPATIBLE_VECTOR_TYPES_P): Define.
912 * config/aarch64/aarch64-sve-builtins.cc (gimple_folder::convert_pred):
913 Use the original predicate if it already has a suitable type.
915 2020-01-09 Martin Jambor <mjambor@suse.cz>
917 * cgraph.h (cgraph_edge): Make remove, set_call_stmt, make_direct,
918 resolve_speculation and redirect_call_stmt_to_callee static. Change
919 return type of set_call_stmt to cgraph_edge *.
920 * auto-profile.c (afdo_indirect_call): Adjust call to
921 redirect_call_stmt_to_callee.
922 * cgraph.c (cgraph_edge::set_call_stmt): Make return cgraph-edge *,
923 make the this pointer explicit, adjust self-recursive calls and the
924 call top make_direct. Return the resulting edge.
925 (cgraph_edge::remove): Make this pointer explicit.
926 (cgraph_edge::resolve_speculation): Likewise, adjust call to remove.
927 (cgraph_edge::make_direct): Likewise, adjust call to
929 (cgraph_edge::redirect_call_stmt_to_callee): Likewise, also adjust
930 call to set_call_stmt.
931 (cgraph_update_edges_for_call_stmt_node): Update call to
932 set_call_stmt and remove.
933 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
934 Renamed edge to master_edge. Adjusted calls to set_call_stmt.
935 (cgraph_node::create_edge_including_clones): Moved "first" definition
936 of edge to the block where it was used. Adjusted calls to
938 (cgraph_node::remove_symbol_and_inline_clones): Adjust call to
940 * cgraphunit.c (walk_polymorphic_call_targets): Adjusted calls to
941 make_direct and redirect_call_stmt_to_callee.
942 * ipa-fnsummary.c (redirect_to_unreachable): Adjust calls to
943 resolve_speculation and make_direct.
944 * ipa-inline-transform.c (inline_transform): Adjust call to
945 redirect_call_stmt_to_callee.
946 (check_speculations_1):: Adjust call to resolve_speculation.
947 * ipa-inline.c (resolve_noninline_speculation): Adjust call to
949 (inline_small_functions): Adjust call to resolve_speculation.
950 (ipa_inline): Likewise.
951 * ipa-prop.c (ipa_make_edge_direct_to_target): Adjust call to
953 * ipa-visibility.c (function_and_variable_visibility): Make iteration
954 safe with regards to edge removal, adjust calls to
955 redirect_call_stmt_to_callee.
956 * ipa.c (walk_polymorphic_call_targets): Adjust calls to make_direct
957 and redirect_call_stmt_to_callee.
958 * multiple_target.c (create_dispatcher_calls): Adjust call to
959 redirect_call_stmt_to_callee
960 (redirect_to_specific_clone): Likewise.
961 * tree-cfgcleanup.c (delete_unreachable_blocks_update_callgraph):
962 Adjust calls to cgraph_edge::remove.
963 * tree-inline.c (copy_bb): Adjust call to set_call_stmt.
964 (redirect_all_calls): Adjust call to redirect_call_stmt_to_callee.
965 (expand_call_inline): Adjust call to cgraph_edge::remove.
967 2020-01-09 Martin Liska <mliska@suse.cz>
969 * params.opt: Set Optimization for
970 param_max_speculative_devirt_maydefs.
972 2020-01-09 Martin Sebor <msebor@redhat.com>
976 * builtins.c (compute_objsize): Avoid handling MEM_REFs of vector type.
978 2020-01-09 Martin Liska <mliska@suse.cz>
980 * auto-profile.c (auto_profile): Use opt_for_fn
982 * ipa-cp.c (ipcp_lattice::add_value): Likewise.
983 (propagate_vals_across_arith_jfunc): Likewise.
984 (hint_time_bonus): Likewise.
985 (incorporate_penalties): Likewise.
986 (good_cloning_opportunity_p): Likewise.
987 (perform_estimation_of_a_value): Likewise.
988 (estimate_local_effects): Likewise.
989 (ipcp_propagate_stage): Likewise.
990 * ipa-fnsummary.c (decompose_param_expr): Likewise.
991 (set_switch_stmt_execution_predicate): Likewise.
992 (analyze_function_body): Likewise.
993 * ipa-inline-analysis.c (offline_size): Likewise.
994 * ipa-inline.c (early_inliner): Likewise.
995 * ipa-prop.c (ipa_analyze_node): Likewise.
996 (ipcp_transform_function): Likewise.
997 * ipa-sra.c (process_scan_results): Likewise.
998 (ipa_sra_summarize_function): Likewise.
999 * params.opt: Rename ipcp-unit-growth to
1000 ipa-cp-unit-growth. Add Optimization for various
1001 IPA-related parameters.
1003 2020-01-09 Richard Biener <rguenther@suse.de>
1006 * gimplify.c (gimplify_expr): Deal with NOP definitions.
1008 2020-01-09 Richard Biener <rguenther@suse.de>
1010 PR tree-optimization/93040
1011 * gimple-ssa-store-merging.c (find_bswap_or_nop): Raise search limit.
1013 2020-01-09 Georg-Johann Lay <avr@gjlay.de>
1015 * common/config/avr/avr-common.c (avr_option_optimization_table)
1016 [OPT_LEVELS_1_PLUS]: Set -fsplit-wide-types-early.
1018 2020-01-09 Martin Liska <mliska@suse.cz>
1020 * cgraphclones.c (symbol_table::materialize_all_clones):
1021 Use cgraph_node::dump_name.
1023 2020-01-09 Jakub Jelinek <jakub@redhat.com>
1026 * config/riscv/riscv.c (riscv_print_operand_reloc): Use
1027 output_operand_lossage instead of gcc_unreachable.
1028 * doc/md.texi (riscv f constraint): Fix typo.
1031 * config/i386/i386.md (subv<mode>4): Use SWIDWI iterator instead of
1032 SWI. Use <general_hilo_operand> instead of <general_operand>. Use
1033 CONST_SCALAR_INT_P instead of CONST_INT_P.
1034 (*subv<mode>4_1): Rename to ...
1035 (subv<mode>4_1): ... this.
1036 (*subv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
1037 define_insn_and_split patterns.
1038 (*subv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
1041 2020-01-08 David Malcolm <dmalcolm@redhat.com>
1043 * vec.c (class selftest::count_dtor): New class.
1044 (selftest::test_auto_delete_vec): New test.
1045 (selftest::vec_c_tests): Call it.
1046 * vec.h (class auto_delete_vec): New class template.
1047 (auto_delete_vec<T>::~auto_delete_vec): New dtor.
1049 2020-01-08 David Malcolm <dmalcolm@redhat.com>
1051 * sbitmap.h (auto_sbitmap): Add operator const_sbitmap.
1053 2020-01-08 Jim Wilson <jimw@sifive.com>
1055 * config/riscv/riscv.c (riscv_legitimize_tls_address): Ifdef out
1056 use of TLS_MODEL_LOCAL_EXEC when not pic.
1058 2020-01-08 David Malcolm <dmalcolm@redhat.com>
1060 * hash-map-tests.c (selftest::test_map_of_strings_to_int): Fix
1063 2020-01-08 Jakub Jelinek <jakub@redhat.com>
1066 * config/i386/i386.md (*stack_protect_set_2_<mode> peephole2,
1067 *stack_protect_set_3 peephole2): Also check that the second
1068 insns source is general_operand.
1071 * config/i386/i386.md (addcarry<mode>_0): Use nonimmediate_operand
1072 predicate for output operand instead of register_operand.
1073 (addcarry<mode>, addcarry<mode>_1): Likewise. Add alternative with
1074 memory destination and non-memory operands[2].
1076 2020-01-08 Martin Liska <mliska@suse.cz>
1078 * cgraph.c (cgraph_node::dump): Use ::dump_name or
1079 ::dump_asm_name instead of (::name or ::asm_name).
1080 * cgraphclones.c (symbol_table::materialize_all_clones): Likewise.
1081 * cgraphunit.c (walk_polymorphic_call_targets): Likewise.
1082 (analyze_functions): Likewise.
1083 (expand_all_functions): Likewise.
1084 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
1085 (propagate_bits_across_jump_function): Likewise.
1086 (dump_profile_updates): Likewise.
1087 (ipcp_store_bits_results): Likewise.
1088 (ipcp_store_vr_results): Likewise.
1089 * ipa-devirt.c (dump_targets): Likewise.
1090 * ipa-fnsummary.c (analyze_function_body): Likewise.
1091 * ipa-hsa.c (check_warn_node_versionable): Likewise.
1092 (process_hsa_functions): Likewise.
1093 * ipa-icf.c (sem_item_optimizer::merge_classes): Likewise.
1094 (set_alias_uids): Likewise.
1095 * ipa-inline-transform.c (save_inline_function_body): Likewise.
1096 * ipa-inline.c (recursive_inlining): Likewise.
1097 (inline_to_all_callers_1): Likewise.
1098 (ipa_inline): Likewise.
1099 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
1100 (ipa_propagate_frequency): Likewise.
1101 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
1102 (remove_described_reference): Likewise.
1103 * ipa-pure-const.c (worse_state): Likewise.
1104 (check_retval_uses): Likewise.
1105 (analyze_function): Likewise.
1106 (propagate_pure_const): Likewise.
1107 (propagate_nothrow): Likewise.
1108 (dump_malloc_lattice): Likewise.
1109 (propagate_malloc): Likewise.
1110 (pass_local_pure_const::execute): Likewise.
1111 * ipa-visibility.c (optimize_weakref): Likewise.
1112 (function_and_variable_visibility): Likewise.
1113 * ipa.c (symbol_table::remove_unreachable_nodes): Likewise.
1114 (ipa_discover_variable_flags): Likewise.
1115 * lto-streamer-out.c (output_function): Likewise.
1116 (output_constructor): Likewise.
1117 * tree-inline.c (copy_bb): Likewise.
1118 * tree-ssa-structalias.c (ipa_pta_execute): Likewise.
1119 * varpool.c (symbol_table::remove_unreferenced_decls): Likewise.
1121 2020-01-08 Richard Biener <rguenther@suse.de>
1124 * tree-eh.c (sink_clobbers): Update virtual operands for
1125 the first and last stmt only. Add a dry-run capability.
1126 (pass_lower_eh_dispatch::execute): Perform clobber sinking
1127 after CFG manipulations and in RPO order to catch all
1128 secondary opportunities reliably.
1130 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
1133 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
1135 2019-01-08 Richard Biener <rguenther@suse.de>
1138 * gimple-fold.c (rewrite_to_defined_overflow): Mark stmt modified.
1139 * tree-ssa-loop-im.c (move_computations_worker): Properly adjust
1140 virtual operand, also updating SSA use.
1141 * gimple-loop-interchange.cc (loop_cand::undo_simple_reduction):
1142 Update stmt after resetting virtual operand.
1143 (tree_loop_interchange::move_code_to_inner_loop): Likewise.
1144 * gimple-iterator.c (gsi_remove): When not removing the stmt
1145 permanently do not delink immediate uses or mark the stmt modified.
1147 2020-01-08 Martin Liska <mliska@suse.cz>
1149 * ipa-fnsummary.c (dump_ipa_call_summary): Use symtab_node::dump_name.
1150 (ipa_call_context::estimate_size_and_time): Likewise.
1151 (inline_analyze_function): Likewise.
1153 2020-01-08 Martin Liska <mliska@suse.cz>
1155 * cgraph.c (cgraph_node::dump): Use systematically
1158 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
1160 Add -nodevicespecs option for avr.
1163 * config/avr/avr.opt (-nodevicespecs): New driver option.
1164 * config/avr/driver-avr.c (avr_devicespecs_file): Only issue
1165 "-specs=device-specs/..." if that option is not set.
1166 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
1168 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
1170 Implement 64-bit double functions for avr.
1173 * config.gcc (tm_defines) [target=avr]: Support --with-libf7,
1174 --with-double-comparison.
1175 * doc/install.texi: Document them.
1176 * config/avr/avr-c.c (avr_cpu_cpp_builtins)
1177 <WITH_LIBF7_LIBGCC, WITH_LIBF7_MATH, WITH_LIBF7_MATH_SYMBOLS>
1178 <WITH_DOUBLE_COMPARISON>: New built-in defines.
1179 * doc/invoke.texi (AVR Built-in Macros): Document them.
1180 * config/avr/avr-protos.h (avr_float_lib_compare_returns_bool): New.
1181 * config/avr/avr.c (avr_float_lib_compare_returns_bool): New function.
1182 * config/avr/avr.h (FLOAT_LIB_COMPARE_RETURNS_BOOL): New macro.
1184 2020-01-08 Richard Earnshaw <rearnsha@arm.com>
1187 * config/arm/t-multilib (MULTILIB_MATCHES): Add rules to match
1188 armv7-a{+mp,+sec,+mp+sec} to appropriate armv7 multilib variants
1189 when only building rm-profile multilibs.
1191 2020-01-08 Feng Xue <fxue@os.amperecomputing.com>
1194 * ipa-cp.c (self_recursively_generated_p): Find matched aggregate
1195 lattice for a value to check.
1196 (propagate_vals_across_arith_jfunc): Add an assertion to ensure
1197 finite propagation in self-recursive scc.
1199 2020-01-08 Luo Xiong Hu <luoxhu@linux.ibm.com>
1201 * ipa-inline.c (caller_growth_limits): Restore the AND.
1203 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
1205 * config/gcn/gcn-valu.md (VEC_1REG_INT_ALT): Delete iterator.
1206 (VEC_ALLREG_ALT): New iterator.
1207 (VEC_ALLREG_INT_MODE): New iterator.
1208 (VCMP_MODE): New iterator.
1209 (VCMP_MODE_INT): New iterator.
1210 (vec_cmpu<mode>di): Use VCMP_MODE_INT.
1211 (vec_cmp<u>v64qidi): New define_expand.
1212 (vec_cmp<mode>di_exec): Use VCMP_MODE.
1213 (vec_cmpu<mode>di_exec): New define_expand.
1214 (vec_cmp<u>v64qidi_exec): New define_expand.
1215 (vec_cmp<mode>di_dup): Use VCMP_MODE.
1216 (vec_cmp<mode>di_dup_exec): Use VCMP_MODE.
1217 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>): Rename ...
1218 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): ... to this.
1219 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>_exec): Rename ...
1220 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): ... to this.
1221 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>): Rename ...
1222 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): ... to this.
1223 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>_exec): Rename ...
1224 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): ... to
1226 * config/gcn/gcn.c (print_operand): Fix 8 and 16 bit suffixes.
1227 * config/gcn/gcn.md (expander): Add sign_extend and zero_extend.
1229 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
1231 * config/gcn/constraints.md (DA): Update description and match.
1233 (Db): New constraint.
1234 * config/gcn/gcn-protos.h (gcn_inline_constant64_p): Add second
1236 * config/gcn/gcn.c (gcn_inline_constant64_p): Add 'mixed' parameter.
1237 Implement 'Db' mixed immediate type.
1238 * config/gcn/gcn-valu.md (addcv64si3<exec_vcc>): Rework constraints.
1239 (addcv64si3_dup<exec_vcc>): Delete.
1240 (subcv64si3<exec_vcc>): Rework constraints.
1241 (addv64di3): Rework constraints.
1242 (addv64di3_exec): Rework constraints.
1243 (subv64di3): Rework constraints.
1244 (addv64di3_dup): Delete.
1245 (addv64di3_dup_exec): Delete.
1246 (addv64di3_zext): Rework constraints.
1247 (addv64di3_zext_exec): Rework constraints.
1248 (addv64di3_zext_dup): Rework constraints.
1249 (addv64di3_zext_dup_exec): Rework constraints.
1250 (addv64di3_zext_dup2): Rework constraints.
1251 (addv64di3_zext_dup2_exec): Rework constraints.
1252 (addv64di3_sext_dup2): Rework constraints.
1253 (addv64di3_sext_dup2_exec): Rework constraints.
1255 2020-01-07 Andre Vieira <andre.simoesdiasvieira@arm.com>
1257 * doc/sourcebuild.texi (arm_little_endian, arm_nothumb): Documented
1258 existing target checks.
1260 2020-01-07 Richard Biener <rguenther@suse.de>
1262 * doc/install.texi: Bump minimal supported MPC version.
1264 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
1266 * langhooks-def.h (lhd_simulate_enum_decl): Declare.
1267 (LANG_HOOKS_SIMULATE_ENUM_DECL): Use it.
1268 * langhooks.c: Include stor-layout.h.
1269 (lhd_simulate_enum_decl): New function.
1270 * config/aarch64/aarch64-sve-builtins.cc (init_builtins): Call
1271 handle_arm_sve_h for the LTO frontend.
1272 (register_vector_type): Cope with null returns from pushdecl.
1274 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
1276 * config/aarch64/aarch64-protos.h (aarch64_sve::svbool_type_p)
1277 (aarch64_sve::nvectors_if_data_type): Replace with...
1278 (aarch64_sve::builtin_type_p): ...this.
1279 * config/aarch64/aarch64-sve-builtins.cc: Include attribs.h.
1280 (find_vector_type): Delete.
1281 (add_sve_type_attribute): New function.
1282 (lookup_sve_type_attribute): Likewise.
1283 (register_builtin_types): Add an "SVE type" attribute to each type.
1284 (register_tuple_type): Likewise.
1285 (svbool_type_p, nvectors_if_data_type): Delete.
1286 (mangle_builtin_type): Use lookup_sve_type_attribute.
1287 (builtin_type_p): Likewise. Add an overload that returns the
1288 number of constituent vector and predicate registers.
1289 * config/aarch64/aarch64.c (aarch64_sve_argument_p): Delete.
1290 (aarch64_returns_value_in_sve_regs_p): Use aarch64_sve::builtin_type_p
1291 instead of aarch64_sve_argument_p.
1292 (aarch64_takes_arguments_in_sve_regs_p): Likewise.
1293 (aarch64_pass_by_reference): Likewise.
1294 (aarch64_function_value_1): Likewise.
1295 (aarch64_return_in_memory): Likewise.
1296 (aarch64_layout_arg): Likewise.
1298 2020-01-07 Jakub Jelinek <jakub@redhat.com>
1300 PR tree-optimization/93156
1301 * tree-ssa-ccp.c (bit_value_binop): For x * x note that the second
1302 least significant bit is always clear.
1304 PR tree-optimization/93118
1305 * match.pd ((x >> c) << c -> x & (-1<<c)): Add nop_convert?. Add new
1306 simplifier with two intermediate conversions.
1308 2020-01-07 Martin Liska <mliska@suse.cz>
1310 * params.opt: Add Optimization for various parameters.
1312 2020-01-07 Martin Liska <mliska@suse.cz>
1315 * doc/extend.texi: Explain cloning for target_clone
1318 2020-01-07 Martin Liska <mliska@suse.cz>
1320 PR tree-optimization/92860
1321 * common.opt: Make in Optimization option
1322 as it is affected by -O0, which is an Optimization
1324 * tree-inline.c (tree_inlinable_function_p):
1325 Use opt_for_fn for warn_inline.
1326 (expand_call_inline): Likewise.
1328 2020-01-07 Martin Liska <mliska@suse.cz>
1330 PR tree-optimization/92860
1331 * common.opt: Make flag_ree as optimization
1334 2020-01-07 Martin Liska <mliska@suse.cz>
1336 PR optimization/92860
1337 * params.opt: Mark param_min_crossjump_insns with Optimization
1340 2020-01-07 Luo Xiong Hu <luoxhu@linux.ibm.com>
1342 * ipa-inline-analysis.c (estimate_growth): Fix typo.
1343 * ipa-inline.c (caller_growth_limits): Use OR instead of AND.
1345 2020-01-06 Michael Meissner <meissner@linux.ibm.com>
1347 * config/rs6000/rs6000.c (hard_reg_and_mode_to_addr_mask): New
1348 helper function to return the valid addressing formats for a given
1349 hard register and mode.
1350 (rs6000_adjust_vec_address): Call hard_reg_and_mode_to_addr_mask.
1352 * config/rs6000/constraints.md (Q constraint): Update
1354 * doc/md.texi (RS/6000 constraints): Update 'Q' cosntraint
1357 * config/rs6000/vsx.md (vsx_extract_<mode>_var, VSX_D iterator):
1358 Use 'Q' for doing vector extract from memory.
1359 (vsx_extract_v4sf_var): Use 'Q' for doing vector extract from
1361 (vsx_extract_<mode>_var, VSX_EXTRACT_I iterator): Use 'Q' for
1362 doing vector extract from memory.
1363 (vsx_extract_<mode>_<VS_scalar>mode_var): Use 'Q' for doing vector
1364 extract from memory.
1366 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add support
1367 for the offset being 34-bits when -mcpu=future is used.
1369 2020-01-06 John David Anglin <danglin@gcc.gnu.org>
1371 * config/pa/pa.md: Revert change to use ordered_comparison_operator
1372 instead of cmpib_comparison_operator in cmpib patterns.
1373 * config/pa/predicates.md (cmpib_comparison_operator): Revert removal
1374 of cmpib_comparison_operator. Revise comment.
1376 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
1378 * tree-vect-slp.c (vect_build_slp_tree_1): Require all shifts
1379 in an IFN_DIV_POW2 node to be equal.
1381 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
1383 * tree-vect-stmts.c (vect_check_load_store_mask): Rename to...
1384 (vect_check_scalar_mask): ...this.
1385 (vectorizable_store, vectorizable_load): Update call accordingly.
1386 (vectorizable_call): Use vect_check_scalar_mask to check the mask
1387 argument in calls to conditional internal functions.
1389 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
1391 * config/gcn/gcn-valu.md (subv64di3): Use separate alternatives for
1392 '0' matching inputs.
1393 (subv64di3_exec): Likewise.
1395 2020-01-06 Bryan Stenson <bryan@siliconvortex.com>
1397 * config/mips/mips.c (vr4130_align_insns): Fix typo.
1398 * doc/md.texi (movstr): Likewise.
1400 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
1402 * config/gcn/gcn-valu.md (vec_extract<mode><scalar_mode>): Add early
1405 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
1407 * config/aarch64/t-aarch64 ($(srcdir)/config/aarch64/aarch64-tune.md):
1409 (s-aarch64-tune-md): ...this new stamp file. Pipe the new contents
1410 to a temporary file and use move-if-change to update the real
1411 file where necessary.
1413 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
1415 * config/aarch64/aarch64-sve.md (@aarch64_sel_dup<mode>): Use Upl
1416 rather than Upa for CPY /M.
1418 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
1420 * config/gcn/gcn.c (gcn_inline_constant_p): Allow 64 as an inline
1423 2020-01-06 Martin Liska <mliska@suse.cz>
1425 PR tree-optimization/92860
1426 * params.opt: Mark param_max_combine_insns with Optimization
1429 2020-01-05 Jakub Jelinek <jakub@redhat.com>
1432 * config/i386/i386.md (SWIDWI): New mode iterator.
1433 (DWI, dwi): Add TImode variants.
1434 (addv<mode>4): Use SWIDWI iterator instead of SWI. Use
1435 <general_hilo_operand> instead of <general_operand>. Use
1436 CONST_SCALAR_INT_P instead of CONST_INT_P.
1437 (*addv<mode>4_1): Rename to ...
1438 (addv<mode>4_1): ... this.
1439 (QWI): New mode attribute.
1440 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
1441 define_insn_and_split patterns.
1442 (*addv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
1444 (uaddv<mode>4): Use SWIDWI iterator instead of SWI. Use
1445 <general_hilo_operand> instead of <general_operand>.
1446 (*addcarry<mode>_1): New define_insn.
1447 (*add<dwi>3_doubleword_cc_overflow_1): New define_insn_and_split.
1449 2020-01-03 Konstantin Kharlamov <Hi-Angel@yandex.ru>
1451 * gdbinit.in (pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, pdd, pbs, pbm):
1452 Use "call" instead of "set".
1454 2020-01-03 Martin Jambor <mjambor@suse.cz>
1457 * ipa-cp.c (print_all_lattices): Skip functions without info.
1459 2020-01-03 Jakub Jelinek <jakub@redhat.com>
1462 * config/i386/i386-options.c (ix86_simd_clone_adjust): If
1463 TARGET_PREFER_AVX128, use prefer-vector-width=256 for 'c' and 'd'
1464 simd clones. If TARGET_PREFER_AVX256, use prefer-vector-width=512
1465 for 'e' simd clones.
1468 * config/i386/i386.opt (x_prefer_vector_width_type): Remove TargetSave
1470 (mprefer-vector-width=): Add Save.
1471 * config/i386/i386-options.c (ix86_target_string): Add PVW argument, print
1472 -mprefer-vector-width= if non-zero. Fix up -mfpmath= comment.
1473 (ix86_debug_options, ix86_function_specific_print): Adjust
1474 ix86_target_string callers.
1475 (ix86_valid_target_attribute_inner_p): Handle prefer-vector-width=.
1476 (ix86_valid_target_attribute_tree): Likewise.
1477 * config/i386/i386-options.h (ix86_target_string): Add PVW argument.
1478 * config/i386/i386-expand.c (ix86_expand_builtin): Adjust
1479 ix86_target_string caller.
1482 * config/i386/i386.md (abs<mode>2): Use expand_simple_binop instead of
1483 emitting ASHIFTRT, XOR and MINUS by hand. Use gen_int_mode with QImode
1484 instead of gen_int_shift_amount + convert_modes.
1486 PR rtl-optimization/93088
1487 * loop-iv.c (find_single_def_src): Punt after looking through
1488 128 reg copies for regs with single definitions. Move definitions
1491 2020-01-02 Dennis Zhang <dennis.zhang@arm.com>
1493 * config/arm/arm-c.c (arm_cpu_builtins): Define
1494 __ARM_FEATURE_MATMUL_INT8, __ARM_FEATURE_BF16_VECTOR_ARITHMETIC,
1495 __ARM_FEATURE_BF16_SCALAR_ARITHMETIC, and
1496 __ARM_BF16_FORMAT_ALTERNATIVE when enabled.
1497 * config/arm/arm-cpus.in (armv8_6, i8mm, bf16): New features.
1498 * config/arm/arm-tables.opt: Regenerated.
1499 * config/arm/arm.c (arm_option_reconfigure_globals): Initialize
1500 arm_arch_i8mm and arm_arch_bf16 when enabled.
1501 * config/arm/arm.h (TARGET_I8MM): New macro.
1502 (TARGET_BF16_FP, TARGET_BF16_SIMD): Likewise.
1503 * config/arm/t-aprofile: Add matching rules for -march=armv8.6-a.
1504 * config/arm/t-arm-elf (all_v8_archs): Add armv8.6-a.
1505 * config/arm/t-multilib: Add matching rules for -march=armv8.6-a.
1506 (v8_6_a_simd_variants): New.
1507 (v8_*_a_simd_variants): Add i8mm and bf16.
1508 * doc/invoke.texi (armv8.6-a, i8mm, bf16): Document new options.
1510 2020-01-02 Jakub Jelinek <jakub@redhat.com>
1513 * predict.c (compute_function_frequency): Don't call
1514 warn_function_cold on functions that already have cold attribute.
1516 2020-01-01 John David Anglin <danglin@gcc.gnu.org>
1519 * config/pa/pa.c (pa_elf_select_rtx_section): New. Put references to
1520 COMDAT group function labels in .data.rel.ro.local section.
1521 * config/pa/pa32-linux.h (TARGET_ASM_SELECT_RTX_SECTION): Define.
1524 * config/pa/pa.md (scc): Use ordered_comparison_operator instead of
1525 comparison_operator in B and S integer comparisons. Likewise, use
1526 ordered_comparison_operator instead of cmpib_comparison_operator in
1528 * config/pa/predicates.md (cmpib_comparison_operator): Remove.
1530 2020-01-01 Jakub Jelinek <jakub@redhat.com>
1532 Update copyright years.
1534 * gcc.c (process_command): Update copyright notice dates.
1535 * gcov-dump.c (print_version): Ditto.
1536 * gcov.c (print_version): Ditto.
1537 * gcov-tool.c (print_version): Ditto.
1538 * gengtype.c (create_file): Ditto.
1539 * doc/cpp.texi: Bump @copying's copyright year.
1540 * doc/cppinternals.texi: Ditto.
1541 * doc/gcc.texi: Ditto.
1542 * doc/gccint.texi: Ditto.
1543 * doc/gcov.texi: Ditto.
1544 * doc/install.texi: Ditto.
1545 * doc/invoke.texi: Ditto.
1547 2020-01-01 Jan Hubicka <hubicka@ucw.cz>
1549 * ipa.c (walk_polymorphic_call_targets): Fix updating of overall
1552 2020-01-01 Jakub Jelinek <jakub@redhat.com>
1554 PR tree-optimization/93098
1555 * match.pd (popcount): For shift amounts, use integer_onep
1556 or wi::to_widest () == cst instead of tree_to_uhwi () == cst
1557 tests. Make sure that precision is power of two larger than or equal
1558 to 16. Ensure shift is never negative. Use HOST_WIDE_INT_UC macro
1559 instead of ULL suffixed constants. Formatting fixes.
1561 Copyright (C) 2020 Free Software Foundation, Inc.
1563 Copying and distribution of this file, with or without modification,
1564 are permitted in any medium without royalty provided the copyright
1565 notice and this notice are preserved.