1 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
3 * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
4 unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
6 2018-06-22 Maya Rashish <coypu@sdf.org>
8 * doc/invoke.texi (mno-fancy-math-387): Update for changes
9 made to OpenBSD and NetBSD through the years.
11 2018-06-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
13 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
14 behavior of vec_pack (vector double, vector double) to match
15 behavior of vec_float2 (vector double, vector double).
17 2018-06-22 Olivier Hainque <hainque@adacore.com>
19 * gimplify.c (gimplify_function_tree): Prevent creation
20 of a trampoline for the address of the current function
21 passed to entry/exit instrumentation hooks.
23 2018-06-22 Aaron Sawdey <acsawdey@linux.ibm.com>
26 * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
29 2018-06-22 Martin Liska <mliska@suse.cz>
31 PR tree-optimization/86263
32 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
33 Bail out if is_enabled is false.
34 * tree-switch-conversion.h (jump_table_cluster::is_enabled):
36 (jump_table_cluster::is_enabled): New function.
38 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
40 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
41 BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
42 * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
43 (lto_input_ts_binfo_tree_pointers): Likewise.
44 * tree-streamer-out.c (streamer_write_tree_bitfields,
45 write_ts_binfo_tree_pointers): Likewise.
46 * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
48 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
50 * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
52 2018-06-22 Martin Liska <mliska@suse.cz>
54 * symbol-summary.h (get): Make it pure and inline move
55 functionality from ::get function.
56 (get): Remove and inline into ::get and ::get_create.
57 (get_create): Move code from ::get function.
59 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
62 * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
63 -x assembler-with-cpp.
65 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
67 * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
68 _FILE_OFFSET_BITS=64 for C++.
70 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
72 * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
73 conversion insn that shows up when pr85657-3.c is compiled using
74 IEEE 128-bit long double.
75 (neg<mode>2_internal): Use the correct mode to check whether the
77 * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
78 multiply and divide external functions from being created more
81 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
83 * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
85 (rtl_merge_blocks): Likewise. Do not emit a NOP if the location of
86 the edge can be forwarded.
87 (cfg_layout_merge_blocks): Likewise.
89 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
91 * except.c (finish_eh_generation): Commit edge insertions only after
92 the EH edges have been redirected from post-landing to landing pads.
94 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
96 * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
97 create_tmp_var_for to create the FRAME decl.
98 (finalize_nesting_tree_1): Do not unchain the FRAME decl.
100 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
102 * tree-inline.c (copy_edges_for_bb): Minor tweak.
103 (maybe_move_debug_stmts_to_successors): Also reset the locus of the
104 debug statement when resetting its value.
105 (expand_call_inline): Copy the locus of the call onto the assignment
106 of the return value, if any. Use local variable in more cases.
108 2018-06-21 Martin Liska <mliska@suse.cz>
110 * ipa-pure-const.c (propagate_nothrow): Use
111 funct_state_summaries->get.
112 (dump_malloc_lattice): Likewise.
113 (propagate_malloc): Likewise.
115 2018-06-21 Richard Biener <rguenther@suse.de>
117 * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
118 comment. Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
119 * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
121 * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
122 BLOCK_ABSTRACT_ORIGIN unconditionally.
124 2018-06-21 David Malcolm <dmalcolm@redhat.com>
126 * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
128 * ipa-reference.c (ipa_reference_c_finalize): Delete
129 ipa_ref_opt_sum_summaries and set it to NULL.
131 2018-06-21 Tom de Vries <tdevries@suse.de>
133 PR tree-optimization/85859
134 * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
135 test with comment from bb_no_side_effects_p.
137 2018-06-21 Richard Biener <rguenther@suse.de>
139 PR tree-optimization/86232
140 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
141 max for constant niter.
143 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
145 * config/aarch64/aarch64-simd.md
146 (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
148 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
150 * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
151 Make opernads of the unspec commutative.
153 2018-06-21 Richard Biener <rguenther@suse.de>
155 * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
156 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
157 Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
158 (vect_analyze_data_ref_dependence): Re-order checks to deal with
160 (vect_record_base_alignments): Do not record base alignment
161 for gathers or scatters.
162 (vect_compute_data_ref_alignment): Drop return value that is always
163 true. Bail out early for gathers or scatters.
164 (vect_enhance_data_refs_alignment): Bail out early for gathers
166 (vect_find_same_alignment_drs): Likewise.
167 (vect_analyze_data_refs_alignment): Remove dead code.
168 (vect_slp_analyze_and_verify_node_alignment): Likewise.
169 (vect_analyze_data_refs): For possible gathers or scatters do
170 not create an alternate DR, just check their possible validity
171 and mark them. Adjust DECL_NONALIASED handling to not rely
173 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
174 update inits of gathers or scatters.
175 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
176 Also copy gather/scatter flag to pattern vinfo.
178 2018-06-20 Kelvin Nilsen <kelvin@gcc.gnu.org>
180 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
181 behavior of vec_packsu (vector unsigned long long, vector unsigned
182 long long) to match behavior of vec_packs with same signature.
184 2018-06-20 Chung-Lin Tang <cltang@codesourcery.com>
185 Thomas Schwinge <thomas@codesourcery.com>
186 Cesar Philippidis <cesar@codesourcery.com>
188 * gimplify.c (gimplify_scan_omp_clauses): Add support for
189 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
190 (gimplify_adjust_omp_clauses): Likewise.
191 (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
192 support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
193 (gimplify_omp_target_update): Update handling of acc update and
195 * omp-low.c (install_var_field): Remove unused parameter
196 base_pointers_restrict.
197 (scan_sharing_clauses): Remove base_pointers_restrict parameter.
198 Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
200 (omp_target_base_pointers_restrict_p): Delete.
201 (scan_omp_target): Update call to scan_sharing_clauses.
202 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
204 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
205 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
206 (convert_local_omp_clauses): Likewise.
207 * tree-pretty-print.c (dump_omp_clause): Likewise.
208 * tree.c (omp_clause_num_ops): Add entries for OMP_CLAUSE_{IF_PRESENT,
210 (omp_clause_code_name): Likewise.
212 2018-06-20 Jakub Jelinek <jakub@redhat.com>
215 * var-tracking.c (use_narrower_mode_test): Check if shift amount can
218 PR tree-optimization/86231
219 * tree-vrp.c (union_ranges): For ( [ ) ] or ( )[ ] range and
220 anti-range don't overwrite *vr0min before using it to compute *vr0max.
222 2018-06-20 Tom de Vries <tdevries@suse.de>
224 PR tree-optimization/86097
225 * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
226 iv type if signedness of iv type is not the same as that of *nit.
228 2018-06-20 Jakub Jelinek <jakub@redhat.com>
230 * cfgrtl.c (rtl_verify_edges): Formatting fix. If bb->preds has any
231 EDGE_EH edges, verify they are all EDGE_EH.
233 2018-06-20 Maya Rashish <coypu@sdf.org>
235 * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
237 2018-06-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
239 * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
240 * config/aarch64/aarch64.c (xgene1_tunings): Add
241 AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
242 (aarch64_mode_valid_for_sched_fusion_p):
244 (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
245 * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
247 * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
249 (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
250 * config/aarch64/iterators.md (VQ2): New mode iterator.
252 2018-06-20 Martin Liska <mliska@suse.cz>
254 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
255 Change default ratio from 10 to 8.
257 2018-06-20 Martin Liska <mliska@suse.cz>
259 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
261 (bit_test_cluster::find_bit_tests): Likewise.
262 (switch_decision_tree::analyze_switch_statement): Find clusters.
263 * tree-switch-conversion.h (struct jump_table_cluster): Document
266 2018-06-20 Martin Liska <mliska@suse.cz>
268 * tree-switch-conversion.c (switch_conversion::collect):
269 Record m_uniq property.
270 (switch_conversion::expand): Bail out for special conditions.
271 (group_cluster::~group_cluster): New.
272 (group_cluster::group_cluster): Likewise.
273 (group_cluster::dump): Likewise.
274 (jump_table_cluster::emit): New.
275 (switch_decision_tree::fix_phi_operands_for_edges): New.
276 (struct case_node): Remove struct.
277 (jump_table_cluster::can_be_handled): New.
278 (case_values_threshold): Moved to header.
279 (reset_out_edges_aux): Likewise.
280 (jump_table_cluster::is_beneficial): New.
281 (bit_test_cluster::can_be_handled): Likewise.
282 (add_case_node): Remove.
283 (bit_test_cluster::is_beneficial): New.
284 (case_bit_test::cmp): New.
285 (bit_test_cluster::emit): New.
286 (expand_switch_as_decision_tree_p): Remove.
287 (bit_test_cluster::hoist_edge_and_branch_if_true): New.
288 (fix_phi_operands_for_edge): Likewise.
289 (switch_decision_tree::analyze_switch_statement): New.
290 (compute_cases_per_edge): Move ...
291 (switch_decision_tree::compute_cases_per_edge): ... here.
292 (try_switch_expansion): Likewise.
293 (switch_decision_tree::try_switch_expansion): Likewise.
294 (record_phi_operand_mapping): Likewise.
295 (switch_decision_tree::record_phi_operand_mapping): Likewise.
296 (emit_case_decision_tree): Likewise.
297 (switch_decision_tree::emit): Likewise.
298 (balance_case_nodes): Likewise.
299 (switch_decision_tree::balance_case_nodes): Likewise.
300 (dump_case_nodes): Likewise.
301 (switch_decision_tree::dump_case_nodes): Likewise.
302 (emit_jump): Likewise.
303 (switch_decision_tree::emit_jump): Likewise.
304 (emit_cmp_and_jump_insns): Likewise.
305 (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
306 (emit_case_nodes): Likewise.
307 (switch_decision_tree::emit_case_nodes): Likewise.
308 (conditional_probability): Remove.
309 * tree-switch-conversion.h (enum cluster_type): New.
311 (struct cluster): Likewise.
312 (cluster::cluster): Likewise.
313 (struct simple_cluster): Likewise.
314 (simple_cluster::simple_cluster): Likewise.
315 (struct group_cluster): Likewise.
316 (struct jump_table_cluster): Likewise.
317 (struct bit_test_cluster): Likewise.
318 (struct min_cluster_item): Likewise.
319 (struct case_tree_node): Likewise.
320 (case_tree_node::case_tree_node): Likewise.
321 (jump_table_cluster::case_values_threshold): Likewise.
322 (struct case_bit_test): Likewise.
323 (struct switch_decision_tree): Likewise.
324 (struct switch_conversion): Likewise.
325 (switch_decision_tree::reset_out_edges_aux): Likewise.
327 2018-06-20 Martin Liska <mliska@suse.cz>
329 * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
330 (hoist_edge_and_branch_if_true): Likewise.
331 (expand_switch_using_bit_tests_p): Likewise.
332 (struct case_bit_test): Likewise.
333 (case_bit_test_cmp): Likewise.
334 (emit_case_bit_tests): Likewise.
335 (switch_conversion::switch_conversion): New class.
336 (struct switch_conv_info): Remove old struct.
337 (collect_switch_conv_info): More to ...
338 (switch_conversion::collect): ... this.
339 (check_range): Likewise.
340 (switch_conversion::check_range): Likewise.
341 (check_all_empty_except_final): Likewise.
342 (switch_conversion::check_all_empty_except_final): Likewise.
343 (check_final_bb): Likewise.
344 (switch_conversion::check_final_bb): Likewise.
345 (create_temp_arrays): Likewise.
346 (switch_conversion::create_temp_arrays): Likewise.
347 (free_temp_arrays): Likewise.
348 (gather_default_values): Likewise.
349 (switch_conversion::gather_default_values): Likewise.
350 (build_constructors): Likewise.
351 (switch_conversion::build_constructors): Likewise.
352 (constructor_contains_same_values_p): Likewise.
353 (switch_conversion::contains_same_values_p): Likewise.
354 (array_value_type): Likewise.
355 (switch_conversion::array_value_type): Likewise.
356 (build_one_array): Likewise.
357 (switch_conversion::build_one_array): Likewise.
358 (build_arrays): Likewise.
359 (switch_conversion::build_arrays): Likewise.
360 (gen_def_assigns): Likewise.
361 (switch_conversion::gen_def_assigns): Likewise.
362 (prune_bbs): Likewise.
363 (switch_conversion::prune_bbs): Likewise.
364 (fix_phi_nodes): Likewise.
365 (switch_conversion::fix_phi_nodes): Likewise.
366 (gen_inbound_check): Likewise.
367 (switch_conversion::gen_inbound_check): Likewise.
368 (process_switch): Use the newly created class.
369 (switch_conversion::expand): New.
370 (switch_conversion::~switch_conversion): New.
371 * tree-switch-conversion.h: New file.
373 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
375 * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
376 tree-vect-patterns.c.
377 * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
378 (vect_recog_dot_prod_pattern): Use it. Remove the type_in argument.
379 (vect_recog_sad_pattern): Likewise.
380 (vect_recog_widen_sum_pattern): Likewise.
381 (vect_recog_pow_pattern): Likewise. Check for a null vectype.
382 (vect_recog_widen_shift_pattern): Remove the type_in argument.
383 (vect_recog_rotate_pattern): Likewise.
384 (vect_recog_mult_pattern): Likewise.
385 (vect_recog_vector_vector_shift_pattern): Likewise.
386 (vect_recog_divmod_pattern): Likewise.
387 (vect_recog_mixed_size_cond_pattern): Likewise.
388 (vect_recog_bool_pattern): Likewise.
389 (vect_recog_mask_conversion_pattern): Likewise.
390 (vect_try_gather_scatter_pattern): Likewise.
391 (vect_recog_widen_mult_pattern): Likewise. Check for a null vectype.
392 (vect_recog_over_widening_pattern): Likewise.
393 (vect_recog_gather_scatter_pattern): Likewise.
394 (vect_recog_func_ptr): Move from tree-vectorizer.h
395 (vect_vect_recog_func_ptrs): Move further down the file.
396 (vect_recog_func): Likewise. Remove the third argument.
397 (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
398 (vect_pattern_recog_1): Expect the pattern function to do any
399 necessary target tests. Also expect it to provide a vector type.
400 Remove the type_in handling.
402 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
404 * tree-vect-patterns.c (vect_pattern_detected): New function.
405 (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
406 (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
407 (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
408 (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
409 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
410 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
411 (vect_recog_mask_conversion_pattern)
412 (vect_try_gather_scatter_pattern): Likewise.
414 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
416 * tree-vect-patterns.c (vect_get_internal_def): New function.
417 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
418 (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
419 (search_type_for_mask_1): Use it.
421 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
423 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
424 redundant WIDEN_SUM_EXPR handling.
425 (vect_recog_sad_pattern): Likewise.
427 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
429 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
430 redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
431 (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
432 (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
433 (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
435 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
437 * tree-vect-stmts.c (vectorizable_call): Make sure that we
438 use the stmt_vec_info of the original bb statement for the
439 new zero assignment, even if the call is part of a pattern.
441 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
443 * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
444 that the sequence is attached to the original statement rather
445 than the pattern statement.
446 * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
447 PATTERN_DEF_SEQ from the original statement rather than
448 the main pattern statement.
449 * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
450 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
451 (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
453 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
455 * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
456 definition statements before the early exit for statements that aren't
458 * tree-vect-loop.c (vect_transform_loop_stmt): New function,
460 (vect_transform_loop): ...here. Process pattern definition
461 statements without first checking whether the main pattern
462 statement is live or relevant.
464 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
466 * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
467 -O0 if the locus represent UNKNOWN_LOCATION but have different values.
469 2018-06-19 Aaron Sawdey <acsawdey@linux.ibm.com>
471 * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
472 TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
473 (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
474 (expand_block_compare): Change select_block_compare_mode call.
475 (expand_strncmp_align_check): Use new functions, fix comment.
476 (emit_final_str_compare_gpr): New function.
477 (expand_strn_compare): Refactor and clean up code.
478 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
480 2018-06-19 Tony Reix <tony.reix@atos.com>
481 Damien Bergamini <damien.bergamini@atos.com>
482 David Edelsohn <dje.gcc@gmail.com>
484 * collect2.c (static_obj): New variable.
485 (static_libs): New variable.
486 (is_in_list): Uncomment declaration.
487 (main): Track AIX libraries linked statically.
488 (is_in_list): Uncomment definition.
489 (scan_prog_file): Don't add AIX shared libraries initializer
490 to constructor list if linking statically.
492 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
494 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
496 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
498 2018-06-19 Jan Hubicka <hubicka@ucw.cz>
500 * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
503 2018-06-19 Martin Liska <mliska@suse.cz>
505 * config/i386/i386.c (ix86_can_inline_p): Do not use
506 ipa_fn_summaries::get_create.
507 * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
509 (devirtualization_time_bonus): Likewise.
510 (ipcp_propagate_stage): Likewise.
511 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
512 (edge_set_predicate): Likewise.
513 (evaluate_conditions_for_known_args): Likewise.
514 (evaluate_properties_for_edge): Likewise.
515 (ipa_call_summary::reset): Tranform to ...
516 (ipa_call_summary::~ipa_call_summary): ... this.
517 (ipa_fn_summary::reset): Transform to ...
518 (ipa_fn_summary::~ipa_fn_summary): ... this.
519 (ipa_fn_summary_t::remove): Rename to ...
520 (ipa_fn_summary_t::remove_callees): ... this.
521 (ipa_fn_summary_t::duplicate): Use placement new
522 instead of memory copy.
523 (ipa_call_summary_t::duplicate): Likewise.
524 (ipa_call_summary_t::remove): Remove.
525 (dump_ipa_call_summary): Change get_create to get.
526 (ipa_dump_fn_summary): Dump only when summary exists.
527 (analyze_function_body): Use symbol_summary::get instead
529 (compute_fn_summary): Likewise.
530 (estimate_edge_devirt_benefit): Likewise.
531 (estimate_edge_size_and_time): Likewise.
532 (inline_update_callee_summaries): Likewise.
533 (remap_edge_change_prob): Likewise.
534 (remap_edge_summaries): Likewise.
535 (ipa_merge_fn_summary_after_inlining): Likewise.
536 (write_ipa_call_summary): Likewise.
537 (ipa_fn_summary_write): Likewise.
538 (ipa_free_fn_summary): Likewise.
539 * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
540 (struct ipa_call_summary): Likewise.
541 * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
543 * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
544 (estimate_size_after_inlining): Likewise.
545 (estimate_growth): Likewise.
546 (growth_likely_positive): Likewise.
547 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
548 (inline_call): Likewise.
549 * ipa-inline.c (caller_growth_limits): Likewise.
550 (can_inline_edge_p): Likewise.
551 (can_inline_edge_by_limits_p): Likewise.
552 (compute_uninlined_call_time): Likewise.
553 (compute_inlined_call_time): Likewise.
554 (want_inline_small_function_p): Likewise.
555 (edge_badness): Likewise.
556 (update_caller_keys): Likewise.
557 (update_callee_keys): Likewise.
558 (inline_small_functions): Likewise.
559 (inline_to_all_callers_1): Likewise.
560 (dump_overall_stats): Likewise.
561 (early_inline_small_functions): Likewise.
562 (early_inliner): Likewise.
563 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
564 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
565 * ipa-pure-const.c (malloc_candidate_p): Likewise.
566 * ipa-split.c (execute_split_functions): Likewise.
567 * symbol-summary.h: Likewise.
568 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
570 2018-06-19 Richard Biener <rguenther@suse.de>
572 * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
573 (vectorize_loops): ... here. Fix dbgcnt handling.
574 (try_vectorize_loop): Wrap try_vectorize_loop_1.
576 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
579 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
580 ieee128 argument takes up only one (vector) register, not two (floating
583 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
585 * gimplify.c (gimplify_init_constructor): Really never clear for an
586 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
588 2018-06-19 Richard Biener <rguenther@suse.de>
590 PR tree-optimization/86179
591 * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
592 after failed recognition.
594 2018-06-18 Martin Sebor <msebor@redhat.com>
597 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
598 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
599 Handle integer subtraction.
600 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
601 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
603 2018-06-18 David Malcolm <dmalcolm@redhat.com>
605 * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
606 param from rtx to rtx_insn *.
607 * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
609 (frv_ifcvt_modify_insn): Likwise.
610 (frv_ifcvt_modify_final): Likwise for local "existing_insn",
611 adding an as_a <rtx_insn *> cast. Likewise for local "insn".
612 * config/mips/mips.c (r10k_insert_cache_barriers): Add an
613 as_a <rtx_insn *> cast to local "unprotected_region" once
614 it's been established that it's not NULL or pc_rtx.
615 * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
616 param "sethi" from rtx to rtx_insn *.
617 (nds32_group_float_insns): Likewise for param "insn".
618 * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
620 (vax_output_int_subtract): Likewise.
621 * config/vax/vax.c (vax_output_int_add): Likewise for param
623 (vax_output_int_subtract): Likewise.
624 * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
625 (emit_pattern_after): Likewise for param "after".
626 (emit_insn_after): Likewise.
627 (emit_jump_insn_after): Likewise.
628 (emit_call_insn_after): Likewise.
629 (emit_debug_insn_after): Likewise.
630 (emit_pattern_before): Likewise for param "before".
631 (emit_insn_before): Likewise.
632 (emit_jump_insn_before): Likewise.
633 * final.c (get_insn_template): Likewise for param "insn", removing
635 * output.h (get_insn_template): Likewise for 2nd param.
636 * rtl.h (emit_insn_before): Likewise.
637 (emit_jump_insn_before): Likewise.
638 (emit_debug_insn_before_noloc): Likewise.
639 (emit_insn_after): Likewise.
640 (emit_jump_insn_after): Likewise.
641 (emit_call_insn_after): Likewise.
642 (emit_debug_insn_after): Likewise.
643 (set_insn_deleted): Likewise for param.
645 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
648 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
649 floating point modes, so that IFmode is numerically greater than
650 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
651 to declare the ordering. This prevents IFmode from being
652 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
653 machine. Include rs6000-modes.h to share the fractional values
654 between genmodes* and the rest of the compiler.
658 * config/rs6000/rs6000-modes.h: New file.
659 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
660 meaning of rs6000_long_double_size so that 126..128 selects an
661 appropriate 128-bit floating point type.
662 (rs6000_option_override_internal): Likewise.
663 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
664 (TARGET_LONG_DOUBLE_128): Change the meaning of
665 rs6000_long_double_size so that 126..128 selects an appropriate
666 128-bit floating point type.
667 (LONG_DOUBLE_TYPE_SIZE): Update comment.
668 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
669 source and destination to match the standard usage.
670 (truncifkf2): Likewise.
671 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
672 ISA 2.07 to use an explicit clobber, instead of passing in a
674 (copysign<mode>3_soft): Likewise.
676 2018-06-18 David Malcolm <dmalcolm@redhat.com>
678 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
679 Replace dump_printf_loc call with DUMP_VECT_SCOPE.
680 (vect_slp_analyze_instance_dependence): Likewise.
681 (vect_enhance_data_refs_alignment): Likewise.
682 (vect_analyze_data_refs_alignment): Likewise.
683 (vect_slp_analyze_and_verify_instance_alignment
684 (vect_analyze_data_ref_accesses): Likewise.
685 (vect_prune_runtime_alias_test_list): Likewise.
686 (vect_analyze_data_refs): Likewise.
687 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
688 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
689 (vect_analyze_scalar_cycles_1): Likewise.
690 (vect_get_loop_niters): Likewise.
691 (vect_analyze_loop_form_1): Likewise.
692 (vect_update_vf_for_slp): Likewise.
693 (vect_analyze_loop_operations): Likewise.
694 (vect_analyze_loop): Likewise.
695 (vectorizable_induction): Likewise.
696 (vect_transform_loop): Likewise.
697 * tree-vect-patterns.c (vect_pattern_recog): Likewise.
698 * tree-vect-slp.c (vect_analyze_slp): Likewise.
699 (vect_make_slp_decision): Likewise.
700 (vect_detect_hybrid_slp): Likewise.
701 (vect_slp_analyze_operations): Likewise.
702 (vect_slp_bb): Likewise.
703 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
704 (vectorizable_bswap): Likewise.
705 (vectorizable_call): Likewise.
706 (vectorizable_simd_clone_call): Likewise.
707 (vectorizable_conversion): Likewise.
708 (vectorizable_assignment): Likewise.
709 (vectorizable_shift): Likewise.
710 (vectorizable_operation): Likewise.
711 * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
713 2018-06-18 Martin Sebor <msebor@redhat.com>
715 PR tree-optimization/81384
716 * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
717 * builtins.c (expand_builtin_strnlen): New function.
718 (expand_builtin): Call it.
719 (fold_builtin_n): Avoid setting TREE_NO_WARNING.
720 * builtins.def (BUILT_IN_STRNLEN): New.
721 * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
722 Warn for bounds in excess of maximum object size.
723 * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
724 single-value ranges. Handle strnlen.
725 (handle_builtin_strlen): Handle strnlen.
726 (strlen_check_and_optimize_stmt): Same.
727 * doc/extend.texi (Other Builtins): Document strnlen.
729 2018-06-18 Maya Rashish <coypu@sdf.org>
731 * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
732 (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
733 (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
735 * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
737 * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
739 2018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
741 * tree.c (escaped_string::escape): Replace cast to char * by
742 const_cast<char *> (unescaped).
744 2018-06-18 Nick Clifton <nickc@redhat.com>
747 * tree.c (escaped_string): New class. Converts an unescaped
748 string into its escaped equivalent.
749 (warn_deprecated_use): Use the new class to convert the
750 deprecation message, if present.
751 (test_escaped_strings): New self test.
752 (test_c_tests): Add test_escaped_strings.
753 * doc/extend.texi (deprecated): Add a note that the
754 deprecation message is affected by the -fmessage-length
755 option, and that control characters will be escaped.
756 (#pragma GCC error): Document this pragma.
757 (#pragma GCC warning): Likewise.
758 * doc/invoke.texi (-fmessage-length): Document this option's
759 effect on the #warning and #error preprocessor directives and
760 the deprecated attribute.
762 2018-06-18 Eric Botcazou <ebotcazou@adacore.com>
764 * tree.c (decl_value_expr_lookup): Revert latest change.
765 (decl_value_expr_insert): Likewise.
767 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
769 * gimplify.c (nonlocal_vlas): Delete.
770 (nonlocal_vla_vars): Likewise.
771 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
773 (gimplify_body): Do not create and destroy nonlocal_vlas.
774 * tree-nested.c: Include diagnostic.h.
775 (use_pointer_in_frame): Tweak.
776 (lookup_field_for_decl): Add assertion and declare the transformation.
777 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
778 internal error when the reference is in a wrong context. Do not
779 create a debug decl by default.
780 (note_nonlocal_block_vlas): Delete.
781 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
782 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
783 create a debug decl by default.
784 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
785 call is in a wrong context.
786 (fixup_vla_decls): New function.
787 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
788 debug variables were created.
789 * tree.c (decl_value_expr_lookup): Add checking assertion.
790 (decl_value_expr_insert): Likewise.
792 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
795 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
796 * tree-scalar-evolution.c (interpret_expr): Likewise.
797 (expression_expensive_p): Likewise.
798 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
799 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
800 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
801 (ssa_defined_by_minus_one_stmt_p): New.
803 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
806 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
807 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
808 * dojump.c (do_jump): Likewise.
809 * expr.c (expand_expr_real_2): Check operand type's sign.
810 * fold-const.c (const_unop): Handle ABSU_EXPR.
811 (fold_abs_const): Likewise.
812 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
813 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
814 (strip_sign_op_1): Likesise.
815 * match.pd: Add new pattern to generate ABSU_EXPR.
816 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
817 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
818 * tree-eh.c (operation_could_trap_helper_p): Likewise.
819 * tree-inline.c (estimate_operator_cost): Likewise.
820 * tree-pretty-print.c (dump_generic_node): Likewise.
821 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
822 * tree.def (ABSU_EXPR): New.
824 2018-06-16 Jakub Jelinek <jakub@redhat.com>
827 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
828 documented as preserved for backward compatibility only.
829 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
831 PR rtl-optimization/86108
832 * bb-reorder.c (create_forwarder_block): Renamed to ...
833 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
834 jump from new landing pad to the second part.
835 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
838 2018-06-15 Jakub Jelinek <jakub@redhat.com>
841 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
842 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
843 Only call store_expr for halves if the mode is the same.
846 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
847 Fix up comment formatting.
849 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
851 * typed-splay-tree.h (typed_splay_tree::remove): New function.
852 (typed_splay_tree::closure,
853 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
854 (typed_splay_tree::typed_splay_tree,
855 typed_splay_tree::operator =): Declared private.
856 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
857 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
858 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
859 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
860 typed_splay_tree::splay_tree_splay,
861 typed_splay_tree::splay_tree_foreach_helper,
862 typed_splay_tree::splay_tree_insert,
863 typed_splay_tree::splay_tree_remove,
864 typed_splay_tree::splay_tree_lookup,
865 typed_splay_tree::splay_tree_predecessor,
866 typed_splay_tree::splay_tree_successor,
867 typed_splay_tree::splay_tree_min,
868 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
869 (typed_splay_tree::root, typed_splay_tree::comp,
870 typed_splay_tree::delete_key,
871 typed_splay_tree::delete_value): New data members.
872 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
873 typed_splay_tree::remove.
875 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
877 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
878 -mginv and -mno-ginv to the assembler.
879 * config/mips/mips.opt (-mcrc): New option.
881 * doc/invoke.text (-mcrc): Document.
884 2018-06-15 Nick Clifton <nickc@redhat.com>
887 * tree.c (escaped_string): New class. Converts an unescaped
888 string into its escaped equivalent.
889 (warn_deprecated_use): Use the new class to convert the
890 deprecation message, if present.
891 (test_escaped_strings): New self test.
892 (test_c_tests): Add test_escaped_strings.
893 * doc/extend.texi (deprecated): Add a note that the
894 deprecation message is affected by the -fmessage-length
895 option, and that control characters will be escaped.
896 (#pragma GCC error): Document this pragma.
897 (#pragma GCC warning): Likewise.
898 * doc/invoke.texi (-fmessage-length): Document this option's
899 effect on the #warning and #error preprocessor directives and
900 the deprecated attribute.
902 2018-06-15 Richard Biener <rguenther@suse.de>
904 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
905 here, also noting vector size used.
906 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
907 size used in MSG_OPTIMIZED_LOCATIONS dump.
908 (pass_slp_vectorize::execute): Adjust.
910 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
913 * config/arc/arc.c (arc_return_address_register): Fix
916 2018-06-15 Richard Biener <rguenther@suse.de>
919 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
920 leave useless conversion stripping to force_gimple_operand_gsi.
921 (gimplify_build2): Likewise.
922 (gimplify_build1): Likewise.
924 2018-06-15 Richard Biener <rguenther@suse.de>
927 * tree-cfg.c (move_stmt_op): unshare invariant addresses
928 before adjusting their block.
930 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
932 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
933 multilibs for *-*-rtems*.
934 * config/riscv/t-rtems: New file.
936 2018-06-14 Jakub Jelinek <jakub@redhat.com>
939 * match.pd ((A +- CST1) +- CST2): Punt if last resort
940 unsigned_type_for returns NULL.
943 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
944 subregs of multi-word pseudos unless the float mode has word size.
946 2018-06-14 Richard Biener <rguenther@suse.de>
949 * tree-vect-generic.c (build_word_mode_vector_type): Remove
950 duplicate and harmful type_hash_canon.
951 * tree.c (type_hash_canon): Assert we didn't find ourselves.
953 2018-06-14 Richard Biener <rguenther@suse.de>
956 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
959 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
961 * config/rtems.h (STDINT_LONG32): Define.
963 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
964 Prachi Godbole <prachi.godbole@imgtec.com>
966 * config/mips/mips-cpus.def: Define P6600.
967 * config/mips/mips-tables.opt: Regenerate.
968 * config/mips/mips.c (mips_ucbranch_type): New enum.
969 (mips_rtx_cost_data): Add support for P6600.
970 (mips_issue_rate): Likewise.
971 (mips_multipass_dfa_lookahead): Likewise.
972 (mips_avoid_hazard): Likewise.
973 (mips_reorg_process_insns): Likewise.
974 (mips_classify_branch_p6600): New function.
975 * config/mips/mips.h (TUNE_P6600): New define.
976 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
977 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
978 * config/mips/mips.md: Include p6600.md.
979 (processor): Add p6600.
980 * config/mips/p6600.md: New file.
981 * doc/invoke.texi: Add p6600 to supported architectures.
983 2018-06-13 Martin Sebor <msebor@redhat.com>
985 PR tree-optimization/86114
986 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
988 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
990 2018-06-13 Richard Biener <rguenther@suse.de>
992 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
993 Properly set vector type of the intermediate stmt.
994 * tree-vect-stmts.c (vectorizable_operation): The destination
995 var always has vectype_out type.
997 2018-06-13 Jeff Law <law@redhat.com>
999 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
1000 integer 0 for argument to print_rtl_with_bb.
1001 (rl78_reorg): Likewise.
1003 2018-06-13 David Malcolm <dmalcolm@redhat.com>
1005 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
1006 from rtx to rtx_insn *.
1007 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
1009 (add_sched_insns_for_speculation): Likewise for local "target",
1010 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
1011 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
1012 from rtx_insn ** to rtx_code_label **.
1013 (reorg_emit_nops): Likewise.
1014 (c6x_reorg): Likewise for local "call_labels".
1015 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
1017 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
1018 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
1019 the loops over LABEL_REFS.
1020 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
1022 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
1023 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
1024 (split_branches): Strengthen local "olabel" from rtx to
1025 rtx_insn *, adding a safe_as_a cast.
1026 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
1028 (add_insn_after): Likewise for first two params.
1029 (add_insn_before): Likewise.
1030 (remove_insn): Likewise for param.
1031 (emit_pattern_before_noloc): Likewise for second and third params.
1032 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
1033 (emit_call_insn_before_noloc): Likewise.
1034 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
1036 (emit_barrier_before): Likewise.
1037 (emit_label_before): Strengthen "label" param from "rtx" to
1038 "rtx_code_label *". Strengthen "before" param from "rtx" to
1040 (emit_insn_after_1): Strengthen "after" param from "rtx" to
1042 (emit_pattern_after_noloc): Likewise.
1043 (emit_insn_after_noloc): Likewise.
1044 (emit_jump_insn_after_noloc): Likewise.
1045 (emit_call_insn_after_noloc): Likewise.
1046 (emit_debug_insn_after_noloc): Likewise.
1047 (emit_barrier_after): Likewise.
1048 (emit_label_after): Likewise for both params.
1049 (emit_pattern_after_setloc): Likewise for "after" param. Convert
1050 "loc" param from "int" to "location_t".
1051 (emit_insn_after_setloc): Likewise.
1052 (emit_jump_insn_after_setloc): Likewise.
1053 (emit_call_insn_after_setloc): Likewise.
1054 (emit_debug_insn_after_setloc): Likewise.
1055 (emit_pattern_before_setloc): Likewise for "before" param. Convert
1056 "loc" param from "int" to "location_t".
1057 (emit_pattern_before): Convert NULL_RTX to NULL.
1058 (emit_insn_before_setloc): Convert "loc" param from "int" to
1060 (emit_jump_insn_before_setloc): Likewise.
1061 (emit_call_insn_before_setloc): Likewise.
1062 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
1063 rtx_insn *. Convert "loc" param from "int" to "location_t".
1064 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
1065 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
1066 Convert 3rd param from "int" to "location_t".
1067 (emit_barrier_before, emit_barrier_after, next_real_insn):
1068 Strengthen param from rtx to rtx_insn *.
1069 (emit_label_before): Strengthen 1st param from "rtx" to
1070 "rtx_code_label *". Strengthen 2nd param from "rtx" to
1072 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
1073 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
1074 Strengthen 2nd param from "rtx" to "rtx_insn *".
1075 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
1076 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
1077 Likewise. Convert 3rd param from "int" to "location_t".
1078 (emit_label_after): Strengthen 1st param from "rtx" to
1080 (next_real_insn, remove_insn): Strengthen param from "rtx" to
1082 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
1083 from "rtx" to "rtx_insn *".
1085 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
1087 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
1088 bodies streamed in with -Q.
1089 * dumpfile.c (dump_files): Add lto-stream-out dump file.
1090 * dumpfile.h (tree_dump_index): Add lto_stream_out.
1091 * gimple-streamer-out.c: Include gimple-pretty-print.h
1092 (output_bb): Dump stmts streamed.
1093 * lto-section-out.c: Include print-tree.h
1094 (lto_begin_section): Dump sections created.
1095 (lto_output_decl_index): Dump decl encoded.
1096 * lto-streamer-out.c: Include print-tree.h
1097 (create_output_block): Dump output block created.
1098 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
1099 (output_function): Dump function output.
1100 (output_constructor): Dump constructor streamed.
1101 (write_global_stream): Output indexes encoded.
1102 (produce_asm_for_decls): Dump streams encoded.
1103 * lto-streamer.c (streamer_dump_file): New global var.
1104 * lto-streamer.h (streamer_dump_file): Declare.
1105 * passes.c (ipa_write_summaries): Initialize streamer dump.
1106 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
1109 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
1112 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
1113 offsets for register save directives. Emit a second batch of save
1114 directives, if need be, when the function accesses prior frames.
1116 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1118 * config/arc/fpu.md (fmasf4): Force operand to register.
1119 (fnmasf4): Likewise.
1121 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1123 * config/arc/arc-protos.h (arc_pad_return): Remove.
1124 * config/arc/arc.c (machine_function): Remove force_short_suffix
1126 (arc_print_operand): Adjust printing of '&'.
1127 (arc_verify_short): Remove conditional printing of short suffix.
1128 (arc_final_prescan_insn): Remove reference to size_reason.
1129 (pad_return): New function.
1130 (arc_reorg): Call pad_return.
1131 (arc_pad_return): Remove.
1132 (arc_init_machine_status): Remove reference to force_short_suffix.
1133 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
1134 (attr length): When attribute iscompact is true force to 2
1135 regardless; in the case of maybe check if we want to force the
1136 instruction to have 4 bytes length.
1137 (nopv): Change it to generate 4 byte long nop as well.
1138 (blockage): New pattern.
1139 (simple_return): Remove call to arc_pad_return.
1140 (p_return_i): Likewise.
1142 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1144 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
1146 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1148 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
1151 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1153 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
1154 for ARC700 and ARCv2.
1156 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
1159 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
1160 operands[2] instead of operands[1].
1163 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
1165 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
1166 case, check whether the outer register overlaps an unallocatable
1167 register, not just whether it fits the required class.
1169 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
1171 * poly-int.h (can_div_trunc_p): Add new overload in which all values
1173 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
1174 (memrefs_conflict_p): Likewise.
1175 (init_alias_analysis): Likewise.
1176 * cfgexpand.c (expand_debug_expr): Likewise.
1177 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
1178 * cse.c (fold_rtx): Likewise.
1179 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
1180 * expr.c (emit_block_move_hints): Likewise.
1181 (clear_storage_hints, push_block, emit_push_insn): Likewise.
1182 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
1183 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
1184 (emit_group_store): Likewise.
1185 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
1186 to read the PRE/POST_MODIFY increment.
1187 * calls.c (store_one_arg): Use strip_offset.
1188 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
1190 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
1192 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
1193 (simplify_binary_operation_1): Extend CONST_INT handling to
1195 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
1196 than a HOST_WIDE_INT.
1197 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
1199 (adjust_mems, add_stores): Update accodingly.
1200 (vt_canonicalize_addr): Track polynomial offsets.
1201 (emit_note_insn_var_location): Likewise.
1202 (vt_add_function_parameter): Likewise.
1203 (vt_initialize): Likewise.
1205 2018-06-12 Jeff Law <law@redhat.com>
1207 * config.gcc (alpha*-*-freebsd*): Remove.
1208 * config/alpha/freebsd.h: Remove.
1210 2018-06-12 David Malcolm <dmalcolm@redhat.com>
1213 * spellcheck-tree.c (levenshtein_distance): Rename to...
1214 (get_edit_distance): ...this, and update for underlying renaming.
1215 * spellcheck-tree.h (levenshtein_distance): Rename to...
1216 (get_edit_distance): ...this.
1217 * spellcheck.c (levenshtein_distance): Rename to...
1218 (get_edit_distance): ...this. Convert from Levenshtein distance
1219 to Damerau-Levenshtein distance by supporting transpositions of
1220 adjacent characters. Rename "v1" to "v_next" and "v0" to
1222 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
1223 (selftest::test_edit_distance_unit_test_oneway): ...this, and
1224 update for underlying renaming.
1225 (selftest::levenshtein_distance_unit_test): Rename to...
1226 (selftest::test_get_edit_distance_unit): ...this, and update for
1227 underlying renaming.
1228 (selftest::test_find_closest_string): Add example from PR 69968
1229 where transposition helps
1230 (selftest::test_metric_conditions): Update for renaming.
1231 (selftest::test_metric_conditions): Likewise.
1232 (selftest::spellcheck_c_tests): Likewise.
1233 * spellcheck.h (levenshtein_distance): Rename both overloads to...
1234 (get_edit_distance): ...this.
1235 (best_match::consider): Update for renaming.
1237 2018-06-12 Martin Sebor <msebor@redhat.com>
1239 PR tree-optimization/85259
1240 * builtins.c (compute_objsize): Handle constant offsets.
1241 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
1242 true iff a warning has been issued.
1243 * gimple.h (gimple_nonartificial_location): New function.
1244 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
1245 gimple_nonartificial_location and handle -Wno-system-headers.
1246 (handle_builtin_stxncpy): Same.
1248 2018-06-12 Martin Sebor <msebor@redhat.com>
1251 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
1253 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
1255 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
1256 BUILTIN_VEC_XST entries for pointer to double and long long.
1258 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
1261 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
1263 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
1266 2018-06-12 Martin Liska <mliska@suse.cz>
1268 * doc/options.texi: Document IntegerRange.
1270 2018-06-12 Martin Liska <mliska@suse.cz>
1272 * config/i386/i386.opt: Make MPX-related options as Deprecated.
1273 * opt-functions.awk: Handle Deprecated flag.
1274 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
1276 (read_cmdline_option): Report warning for a deprecated option.
1277 * opts.h (struct cl_option): Add new field cl_deprecated.
1278 (CL_ERR_DEPRECATED): New.
1280 2018-06-12 Martin Liska <mliska@suse.cz>
1282 * doc/options.texi: Document Deprecated option flag.
1284 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1286 * config/arc/arc-arch.h (arc_extras): New enum.
1287 (arc_cpu_t):Add field extra.
1288 (arc_cpu_types): Consider the extras.
1289 * config/arc/arc-cpus.def: Add extras info.
1290 * config/arc/arc-opts.h (processor_type): Consider extra field.
1291 * config/arc/arc.c (arc_override_options): Handle extra field.
1293 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1295 * config/arc/arc-arch.h: Update ARC_OPTX macro.
1296 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
1298 * config/arc/arc.c (arc_init): Update pic warning.
1299 (irq_range): Update irq range parsing warnings.
1300 (arc_override_options): Update various warning messages.
1301 (arc_handle_aux_attribute): Likewise.
1303 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
1305 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
1307 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
1309 * doc/sourcebuild.texi: Document usage of line number 0 in verify
1310 compiler messages directives.
1312 2018-06-12 Matthew Fortune <mfortune@gmail.com>
1314 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
1315 * config/mips/mips-tables.opt: Regenerate.
1316 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
1318 * doc/invoke.texi: Document -march=i6500.
1320 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
1322 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
1323 (i6400_gpmul): Add cpu_unit.
1324 (i6400_gpdiv): Likewise.
1325 (i6400_msa_add_d): Update reservations.
1326 (i6400_msa_int_add) Likewise.
1327 (i6400_msa_short_logic3) Likewise.
1328 (i6400_msa_short_logic2) Likewise.
1329 (i6400_msa_short_logic) Likewise.
1330 (i6400_msa_move) Likewise.
1331 (i6400_msa_cmp) Likewise.
1332 (i6400_msa_short_float2) Likewise.
1333 (i6400_msa_div_d) Likewise.
1334 (i6400_msa_long_logic1) Likewise.
1335 (i6400_msa_long_logic2) Likewise.
1336 (i6400_msa_mult) Likewise.
1337 (i6400_msa_long_float2) Likewise.
1338 (i6400_msa_long_float4) Likewise.
1339 (i6400_msa_long_float5) Likewise.
1340 (i6400_msa_long_float8) Likewise.
1341 (i6400_fpu_fadd): Include frint type.
1342 (i6400_fpu_store): New define_insn_reservation.
1343 (i6400_fpu_load): Likewise.
1344 (i6400_fpu_move): Likewise.
1345 (i6400_fpu_fcmp): Likewise.
1346 (i6400_fpu_fmadd): Likewise.
1347 (i6400_int_mult): Include imul3nc type and update reservation.
1348 (i6400_int_div): Include idiv3 type and update reservation.
1349 (i6400_int_load): Update to check type not move_type.
1350 (i6400_int_store): Likewise.
1351 (i6400_int_prefetch): Set zero latency.
1353 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
1355 * gcc.c: Document new %@{...} sequence.
1356 (LINK_COMMAND_SPEC): Use it for the -L switches.
1357 (cpp_unique_options): Use it for the -I switches.
1358 (at_file_argbuf): New global variable.
1359 (in_at_file): Likewise.
1360 (alloc_args): Create at_file_argbuf.
1361 (clear_args): Truncate at_file_argbuf.
1362 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
1363 (open_at_file): New function.
1364 (close_at_file): Likewise.
1365 (create_at_file): Delete.
1366 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
1369 (validate_switches_from_spec): Deal with %@{...} sequence.
1370 (validate_switches): Likewise.
1371 (driver::finalize): Call clear_args.
1373 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
1375 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
1377 2018-06-11 Martin Sebor <msebor@redhat.com>
1379 * doc/invoke.texi (-Wall): List -Wc++17-compat.
1380 (Wno-class-memaccess): Add @opindex.
1381 (Wno-templates, Wno-multiple-inheritance): Same.
1382 (Wno-virtual-inheritance, Wno-namespaces): Same.
1383 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
1384 (Wno-format-overflow, Wno-format-truncation): Same.
1385 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
1386 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
1387 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
1388 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
1389 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
1390 (Wno-misspelled-isr): Same.
1392 2018-06-11 Martin Sebor <msebor@redhat.com>
1394 * PR tree-optimization/86083
1395 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
1397 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
1399 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
1401 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
1404 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
1405 on the correct operand.
1406 (*movdi_internal64): Ditto.
1408 2018-06-11 Martin Liska <mliska@suse.cz>
1410 PR tree-optimization/86089
1411 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
1413 2018-06-11 Julia Koval <julia.koval@intel.com>
1415 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
1416 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
1417 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
1419 2018-06-11 Olivier Hainque <hainque@adacore.com>
1421 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
1422 for Ada with strict dwarf2.
1424 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
1427 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
1430 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
1432 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
1434 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
1436 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
1439 2018-06-08 Martin Liska <mliska@suse.cz>
1441 * tree-cfg.h (debug_function): Fix argument type to match
1444 2018-06-08 Martin Liska <mliska@suse.cz>
1446 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
1447 Remove usage of MPX-related (and removed) fields.
1448 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
1450 2018-06-08 David Malcolm <dmalcolm@redhat.com>
1452 * cfg.c (debug): Use TDF_NONE rather than 0.
1453 * cfghooks.c (debug): Likewise.
1454 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
1455 (struct dump_option_value_info): Convert to...
1456 (struct kv_pair): ...this template type.
1457 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
1459 (optinfo_verbosity_options): Likewise.
1460 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
1462 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
1463 than int for "optgroup_flags" param.
1464 (dump_generic_expr_loc): Use dump_flags_t rather than int for
1466 (dump_dec): Likewise.
1467 (dump_finish): Use TDF_NONE rather than 0.
1468 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
1469 rather than int for "optgroup_flags" param. Use TDF_NONE rather
1470 than 0. Update for change to option_ptr.
1471 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
1472 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
1473 0. Update for changes to optinfo_verbosity_options and
1475 (opt_info_switch_p): Convert optgroup_flags from int to
1477 (dump_basic_block): Use dump_flags_t rather than int
1478 for "dump_kind" param.
1479 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
1480 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
1481 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
1482 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
1483 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
1484 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
1485 TDF_NONE): Convert from macros to...
1486 (enum dump_flag): ...this new enum.
1487 (dump_flags_t): Update to use enum.
1488 (operator|, operator&, operator~, operator|=, operator&=):
1489 Implement for dump_flags_t.
1490 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
1491 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
1492 Convert from macros to...
1493 (enum optgroup_flag): ...this new enum.
1494 (optgroup_flags_t): New typedef.
1495 (operator|, operator|=): Implement for optgroup_flags_t.
1496 (struct dump_file_info): Convert field "alt_flags" to
1497 dump_flags_t. Convert field "optgroup_flags" to
1499 (dump_basic_block): Use dump_flags_t rather than int for param.
1500 (dump_generic_expr_loc): Likewise.
1501 (dump_dec): Likewise.
1502 (dump_register): Convert param "optgroup_flags" to
1504 (opt_info_enable_passes): Likewise.
1505 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
1507 * gimple-pretty-print.c (debug): Likewise.
1508 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
1509 (merged_store_group::apply_stores): Likewise.
1510 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
1511 * gimple.c (verify_gimple_pp): Likewise.
1512 * graphite-poly.c (print_pbb_body): Likewise.
1513 * passes.c (pass_manager::register_one_dump_file): Convert
1514 local "optgroup_flags" to optgroup_flags_t.
1515 * print-tree.c (print_node): Use TDF_NONE rather than 0.
1517 (debug_body): Likewise.
1518 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
1519 to optgroup_flags_t.
1520 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
1522 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
1523 (convert_mult_to_fma): Likewise.
1524 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
1525 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
1526 * tree-vect-data-refs.c (dump_lower_bound): Convert param
1527 "dump_kind" to dump_flags_t.
1529 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
1531 * config/rs6000/rs6000.c (min, max): Delete.
1533 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
1535 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
1536 -mabi=spe and -mabi=no-spe.
1538 2018-06-08 Martin Liska <mliska@suse.cz>
1540 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
1541 where we expect an existing summary.
1543 2018-06-08 Martin Liska <mliska@suse.cz>
1545 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
1546 * ipa-inline.h (estimate_edge_growth): Likewise.
1548 2018-06-08 Martin Liska <mliska@suse.cz>
1550 * cgraph.c (function_version_hasher::hash): Use
1551 cgraph_node::get_uid ().
1552 (function_version_hasher::equal):
1553 * cgraph.h (cgraph_node::get_uid): New method.
1554 * ipa-inline.c (update_caller_keys): Use
1555 cgraph_node::get_uid ().
1556 (update_callee_keys): Likewise.
1557 * ipa-utils.c (searchc): Likewise.
1558 (ipa_reduced_postorder): Likewise.
1559 * lto-cgraph.c (input_node): Likewise.
1560 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
1561 * symbol-summary.h (symtab_insertion): Likewise.
1562 (symtab_removal): Likewise.
1563 (symtab_duplication): Likewise.
1564 * tree-pretty-print.c (dump_function_header): Likewise.
1565 * tree-sra.c (convert_callers_for_node): Likewise.
1567 2018-06-08 Martin Liska <mliska@suse.cz>
1569 * cgraph.c (symbol_table::create_edge): Always assign a new
1571 (symbol_table::free_edge): Do not recycle numbers.
1572 * cgraph.h (cgraph_edge::get): New method.
1573 * symbol-summary.h (symtab_removal): Use it.
1574 (symtab_duplication): Likewise.
1575 (call_summary::hashable_uid): Remove.
1577 2018-06-08 Martin Liska <mliska@suse.cz>
1579 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
1580 (initialize_growth_caches): Remove.
1581 (free_growth_caches): Likewise.
1582 (do_estimate_edge_time): Use edge_growth_cache.
1583 (do_estimate_edge_size): Likewise.
1584 (do_estimate_edge_hints): Likewise.
1585 * ipa-inline.c (reset_edge_caches): Likewise.
1586 (recursive_inlining): Likewise.
1587 (inline_small_functions): Likewise.
1588 * ipa-inline.h (initialize_growth_caches): Remove.
1589 (estimate_edge_size): Likewise.
1590 (estimate_edge_time): Likewise.
1591 (estimate_edge_hints): Likewise.
1592 (reset_edge_growth_cache): Likewise.
1593 * symbol-summary.h (call_summary::remove): New method.
1595 2018-06-08 Martin Liska <mliska@suse.cz>
1597 * ipa-cp.c (class edge_clone_summary): New summary.
1598 (grow_edge_clone_vectors): Remove.
1599 (ipcp_edge_duplication_hook): Remove.
1600 (class edge_clone_summary_t): New call_summary class.
1601 (ipcp_edge_removal_hook): Remove.
1602 (edge_clone_summary_t::duplicate): New function.
1603 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
1604 (create_specialized_node): Likewise.
1605 (ipcp_driver): Initialize edge_clone_summaries and do not
1608 2018-06-08 Martin Liska <mliska@suse.cz>
1610 * symbol-summary.h (get): New function.
1611 (call_summary::m_initialize_when_cloning): New class member.
1613 2018-06-08 Martin Liska <mliska@suse.cz>
1615 * cgraph.c (cgraph_node::remove): Do not recycle uid.
1616 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
1617 (symbol_table::allocate_cgraph_symbol): Do not set uid.
1618 * passes.c (uid_hash_t): Record removed_nodes by their uids.
1619 (remove_cgraph_node_from_order): Use the removed_nodes set.
1620 (do_per_function_toporder): Likwise.
1621 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
1622 instead of summary_uid.
1623 (symtab_removal): Likewise.
1624 (symtab_duplication): Likewise.
1626 2018-06-08 Martin Liska <mliska@suse.cz>
1628 * ipa-cp.c (ipcp_store_bits_results): Use
1629 ipcp_transformation_sum.
1630 (ipcp_store_vr_results): Likewise.
1631 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
1633 (ipcp_transformation_initialize): ... this.
1634 (ipa_set_node_agg_value_chain):
1635 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
1636 (write_ipcp_transformation_info): Likewise.
1637 (read_ipcp_transformation_info): Likewise.
1638 (ipcp_update_bits): Likewise.
1639 (ipcp_update_vr): Likewise.
1640 (ipcp_transform_function): Likewise.
1641 * ipa-prop.h: Rename ipcp_transformation_summary to
1642 ipcp_transformation.
1643 (class ipcp_transformation_t): New function summary.
1644 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
1645 (ipa_get_agg_replacements_for_node): Likewise.
1647 2018-06-08 Martin Liska <mliska@suse.cz>
1649 * ipa-pure-const.c (struct funct_state_d): Do it class instead
1651 (class funct_state_summary_t): New function_summary class.
1652 (has_function_state): Remove.
1653 (get_function_state): Likewise.
1654 (set_function_state): Likewise.
1655 (add_new_function): Likewise.
1656 (funct_state_summary_t::insert): New function.
1657 (duplicate_node_data): Remove.
1658 (remove_node_data): Remove.
1659 (funct_state_summary_t::duplicate): New function.
1660 (register_hooks): Create new funct_state_summaries.
1661 (pure_const_generate_summary): Use it.
1662 (pure_const_write_summary): Likewise.
1663 (pure_const_read_summary): Likewise.
1664 (propagate_pure_const): Likewise.
1665 (propagate_nothrow): Likewise.
1666 (dump_malloc_lattice): Likewise.
1667 (propagate_malloc): Likewise.
1668 (execute): Do not register hooks, just remove summary
1670 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
1673 2018-06-08 Martin Liska <mliska@suse.cz>
1675 * ipa-reference.c (remove_node_data): Remove.
1676 (duplicate_node_data): Likewise.
1677 (class ipa_ref_var_info_summary_t): New class.
1678 (class ipa_ref_opt_summary_t): Likewise.
1679 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
1680 (get_reference_optimization_summary): Use
1681 ipa_ref_opt_sum_summaries.
1682 (set_reference_vars_info): Remove.
1683 (set_reference_optimization_summary): Likewise.
1684 (ipa_init): Create summaries.
1685 (init_function_info): Use function summary.
1686 (ipa_ref_opt_summary_t::duplicate): New function.
1687 (ipa_ref_opt_summary_t::remove): New function.
1688 (get_read_write_all_from_node): Fix GNU coding style.
1689 (propagate): Use function summary.
1690 (write_node_summary_p): Fix GNU coding style.
1691 (stream_out_bitmap): Likewise.
1692 (ipa_reference_read_optimization_summary): Use function summary.
1693 (ipa_reference_c_finalize): Do not release hooks.
1695 2018-06-08 Martin Liska <mliska@suse.cz>
1697 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
1698 (analyze_function_body): Extract multiple calls of get_create.
1699 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
1700 * ipa-inline.c (recursive_inlining): Use ::get method.
1701 * ipa-inline.h (estimate_edge_growth): Likewise.
1703 2018-06-08 Martin Liska <mliska@suse.cz>
1705 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
1707 (hsa_function_summary::hsa_function_summary): Use the new enum
1709 (hsa_gpu_implementation_p): Use hsa_summaries::get.
1710 * hsa-gen.c (hsa_get_host_function): Likewise.
1711 (get_brig_function_name): Likewise.
1712 * ipa-hsa.c (process_hsa_functions): Likewise.
1713 (ipa_hsa_write_summary): Likewise.
1714 * symbol-summary.h (symtab_duplication): Use ::get function/
1715 (get): New function.
1717 2018-06-08 Martin Liska <mliska@suse.cz>
1719 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
1721 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
1722 (hsa_register_kernel): Likewise.
1723 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
1724 * hsa-gen.c (hsa_get_host_function): Likewise.
1725 (get_brig_function_name): Likewise.
1726 (generate_hsa): Likewise.
1727 (pass_gen_hsail::execute): Likewise.
1728 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
1729 (devirtualization_time_bonus): Likewise.
1730 (ipcp_propagate_stage): Likewise.
1731 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
1732 (edge_set_predicate): Likewise.
1733 (evaluate_conditions_for_known_args): Likewise.
1734 (evaluate_properties_for_edge): Likewise.
1735 (ipa_fn_summary::reset): Likewise.
1736 (ipa_fn_summary_t::duplicate): Likewise.
1737 (dump_ipa_call_summary): Likewise.
1738 (ipa_dump_fn_summary): Likewise.
1739 (analyze_function_body): Likewise.
1740 (compute_fn_summary): Likewise.
1741 (estimate_edge_devirt_benefit): Likewise.
1742 (estimate_edge_size_and_time): Likewise.
1743 (estimate_calls_size_and_time): Likewise.
1744 (estimate_node_size_and_time): Likewise.
1745 (inline_update_callee_summaries): Likewise.
1746 (remap_edge_change_prob): Likewise.
1747 (remap_edge_summaries): Likewise.
1748 (ipa_merge_fn_summary_after_inlining): Likewise.
1749 (ipa_update_overall_fn_summary): Likewise.
1750 (read_ipa_call_summary): Likewise.
1751 (inline_read_section): Likewise.
1752 (write_ipa_call_summary): Likewise.
1753 (ipa_fn_summary_write): Likewise.
1754 (ipa_free_fn_summary): Likewise.
1755 * ipa-hsa.c (process_hsa_functions): Likewise.
1756 (ipa_hsa_write_summary): Likewise.
1757 (ipa_hsa_read_section): Likewise.
1758 * ipa-icf.c (sem_function::merge): Likewise.
1759 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
1760 (do_estimate_edge_time): Likewise.
1761 (estimate_size_after_inlining): Likewise.
1762 (estimate_growth): Likewise.
1763 (growth_likely_positive): Likewise.
1764 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
1765 (inline_call): Likewise.
1766 * ipa-inline.c (caller_growth_limits): Likewise.
1767 (can_inline_edge_p): Likewise.
1768 (can_inline_edge_by_limits_p): Likewise.
1769 (compute_uninlined_call_time): Likewise.
1770 (compute_inlined_call_time): Likewise.
1771 (want_inline_small_function_p): Likewise.
1772 (edge_badness): Likewise.
1773 (update_caller_keys): Likewise.
1774 (update_callee_keys): Likewise.
1775 (recursive_inlining): Likewise.
1776 (inline_small_functions): Likewise.
1777 (inline_to_all_callers_1): Likewise.
1778 (dump_overall_stats): Likewise.
1779 (early_inline_small_functions): Likewise.
1780 (early_inliner): Likewise.
1781 * ipa-inline.h (estimate_edge_growth): Likewise.
1782 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
1783 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
1784 * ipa-prop.h (IPA_NODE_REF): Likewise.
1785 (IPA_EDGE_REF): Likewise.
1786 * ipa-pure-const.c (malloc_candidate_p): Likewise.
1787 (propagate_malloc): Likewise.
1788 * ipa-split.c (execute_split_functions): Likewise.
1789 * symbol-summary.h: Rename get to get_create.
1791 (get_create): Likewise.
1792 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
1794 2018-06-08 Martin Liska <mliska@suse.cz>
1796 * symbol-summary.h (release): Move definition out of class
1798 (symtab_removal): Likewise.
1799 (symtab_duplication): Likewise.
1801 2018-06-08 Martin Liska <mliska@suse.cz>
1803 * symbol-summary.h (function_summary): Move constructor
1804 implementation out of class declaration.
1805 (release): Likewise.
1806 (symtab_insertion): Likewise.
1807 (symtab_removal): Likewise.
1808 (symtab_duplication): Likewise.
1811 2018-06-08 Martin Liska <mliska@suse.cz>
1813 * Makefile.in: Remove support for MPX (macros, related functions,
1814 fields in cgraph_node, ...).
1815 * builtin-types.def (BT_BND): Likewise.
1816 (BT_FN_BND_CONST_PTR): Likewise.
1817 (BT_FN_CONST_PTR_BND): Likewise.
1818 (BT_FN_VOID_PTR_BND): Likewise.
1819 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
1820 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
1821 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
1822 (expand_builtin_mempcpy_with_bounds): Likewise.
1823 (expand_builtin_memset_with_bounds): Likewise.
1824 (expand_builtin_memset_args): Likewise.
1825 (std_expand_builtin_va_start): Likewise.
1826 (expand_builtin): Likewise.
1827 (expand_builtin_with_bounds): Likewise.
1828 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
1829 (DEF_LIB_BUILTIN_CHKP): Likewise.
1830 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
1831 (DEF_CHKP_BUILTIN): Likewise.
1832 (BUILT_IN_MEMCPY): Likewise.
1833 (BUILT_IN_MEMMOVE): Likewise.
1834 (BUILT_IN_MEMPCPY): Likewise.
1835 (BUILT_IN_MEMSET): Likewise.
1836 (BUILT_IN_STPCPY): Likewise.
1837 (BUILT_IN_STRCAT): Likewise.
1838 (BUILT_IN_STRCHR): Likewise.
1839 (BUILT_IN_STRCPY): Likewise.
1840 (BUILT_IN_STRLEN): Likewise.
1841 (BUILT_IN_MEMCPY_CHK): Likewise.
1842 (BUILT_IN_MEMMOVE_CHK): Likewise.
1843 (BUILT_IN_MEMPCPY_CHK): Likewise.
1844 (BUILT_IN_MEMSET_CHK): Likewise.
1845 (BUILT_IN_STPCPY_CHK): Likewise.
1846 (BUILT_IN_STRCAT_CHK): Likewise.
1847 (BUILT_IN_STRCPY_CHK): Likewise.
1848 * calls.c (store_bounds): Likewise.
1849 (emit_call_1): Likewise.
1850 (special_function_p): Likewise.
1851 (maybe_warn_nonstring_arg): Likewise.
1852 (initialize_argument_information): Likewise.
1853 (finalize_must_preallocate): Likewise.
1854 (compute_argument_addresses): Likewise.
1855 (expand_call): Likewise.
1856 * cfgexpand.c (expand_call_stmt): Likewise.
1857 (expand_return): Likewise.
1858 (expand_gimple_stmt_1): Likewise.
1859 (pass_expand::execute): Likewise.
1860 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
1861 (cgraph_node::remove): Likewise.
1862 (cgraph_node::dump): Likewise.
1863 (cgraph_node::verify_node): Likewise.
1864 * cgraph.h (chkp_function_instrumented_p): Likewise.
1865 (symtab_node::get_alias_target): Likewise.
1866 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
1867 (cgraph_local_p): Likewise.
1868 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
1869 (cgraph_edge::rebuild_references): Likewise.
1870 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
1871 (walk_polymorphic_call_targets): Likewise.
1872 (cgraph_node::expand_thunk): Likewise.
1873 (symbol_table::output_weakrefs): Likewise.
1874 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
1875 (ix86_handle_option): Likewise.
1876 * config/i386/constraints.md: Likewise.
1877 * config/i386/i386-builtin-types.def (BND): Likewise.
1881 * config/i386/i386-builtin.def (BDESC_END): Likewise.
1882 (BDESC_FIRST): Likewise.
1884 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
1885 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
1886 * config/i386/i386.c (enum reg_class): Likewise.
1887 (ix86_target_string): Likewise.
1888 (ix86_option_override_internal): Likewise.
1889 (ix86_conditional_register_usage): Likewise.
1890 (ix86_valid_target_attribute_inner_p): Likewise.
1891 (ix86_set_indirect_branch_type): Likewise.
1892 (ix86_set_current_function): Likewise.
1893 (ix86_function_arg_regno_p): Likewise.
1894 (init_cumulative_args): Likewise.
1895 (ix86_function_arg_advance): Likewise.
1896 (ix86_function_arg): Likewise.
1897 (ix86_pass_by_reference): Likewise.
1898 (ix86_function_value_regno_p): Likewise.
1899 (ix86_function_value_1): Likewise.
1900 (ix86_function_value_bounds): Likewise.
1901 (ix86_return_in_memory): Likewise.
1902 (ix86_setup_incoming_vararg_bounds): Likewise.
1903 (ix86_va_start): Likewise.
1904 (indirect_thunk_need_prefix): Likewise.
1905 (print_reg): Likewise.
1906 (ix86_print_operand): Likewise.
1907 (ix86_expand_call): Likewise.
1908 (ix86_output_function_return): Likewise.
1909 (reg_encoded_number): Likewise.
1910 (BDESC_VERIFYS): Likewise.
1911 (ix86_init_mpx_builtins): Likewise.
1912 (ix86_init_builtins): Likewise.
1913 (ix86_emit_cmove): Likewise.
1914 (ix86_emit_move_max): Likewise.
1915 (ix86_expand_builtin): Likewise.
1916 (ix86_builtin_mpx_function): Likewise.
1917 (ix86_get_arg_address_for_bt): Likewise.
1918 (ix86_load_bounds): Likewise.
1919 (ix86_store_bounds): Likewise.
1920 (ix86_load_returned_bounds): Likewise.
1921 (ix86_store_returned_bounds): Likewise.
1922 (ix86_class_likely_spilled_p): Likewise.
1923 (ix86_hard_regno_mode_ok): Likewise.
1924 (x86_order_regs_for_local_alloc): Likewise.
1925 (ix86_mitigate_rop): Likewise.
1926 (ix86_bnd_prefixed_insn_p): Likewise.
1927 (ix86_mpx_bound_mode): Likewise.
1928 (ix86_make_bounds_constant): Likewise.
1929 (ix86_initialize_bounds): Likewise.
1930 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
1931 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
1932 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
1933 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
1934 (TARGET_CHKP_BOUND_MODE): Likewise.
1935 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
1936 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
1937 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
1938 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
1939 * config/i386/i386.h (TARGET_MPX): Likewise.
1940 (TARGET_MPX_P): Likewise.
1941 (VALID_BND_REG_MODE): Likewise.
1942 (FIRST_BND_REG): Likewise.
1943 (LAST_BND_REG): Likewise.
1944 (enum reg_class): Likewise.
1945 (BND_REG_P): Likewise.
1946 (BND_REGNO_P): Likewise.
1947 (BNDmode): Likewise.
1948 (ADJUST_INSN_LENGTH): Likewise.
1949 * config/i386/i386.md: Likewise.
1950 * config/i386/i386.opt: Likewise.
1951 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
1952 (defined): Likewise.
1953 (LINK_MPX): Likewise.
1954 (MPX_SPEC): Likewise.
1955 (LIBMPX_SPEC): Likewise.
1956 (LIBMPXWRAPPERS_SPEC): Likewise.
1957 (CHKP_SPEC): Likewise.
1958 * config/i386/predicates.md: Likewise.
1959 * dbxout.c (dbxout_type): Likewise.
1960 * doc/extend.texi: Likewise.
1961 * doc/invoke.texi: Likewise.
1962 * doc/md.texi: Likewise.
1963 * doc/tm.texi: Likewise.
1964 * doc/tm.texi.in: Likewise.
1965 * dwarf2out.c (is_base_type): Likewise.
1966 (gen_formal_types_die): Likewise.
1967 (gen_subprogram_die): Likewise.
1968 (gen_type_die_with_usage): Likewise.
1969 (gen_decl_die): Likewise.
1970 (dwarf2out_late_global_decl): Likewise.
1971 * expr.c (expand_assignment): Likewise.
1972 (emit_storent_insn): Likewise.
1973 (store_expr_with_bounds): Likewise.
1974 (store_expr): Likewise.
1975 (expand_expr_real_1): Likewise.
1976 * expr.h (store_expr_with_bounds): Likewise.
1977 * function.c (use_register_for_decl): Likewise.
1978 (struct bounds_parm_data): Likewise.
1979 (assign_parms_augmented_arg_list): Likewise.
1980 (assign_parm_find_entry_rtl): Likewise.
1981 (assign_parm_is_stack_parm): Likewise.
1982 (assign_parm_load_bounds): Likewise.
1983 (assign_bounds): Likewise.
1984 (assign_parms): Likewise.
1985 (expand_function_start): Likewise.
1986 * gcc.c (CHKP_SPEC): Likewise.
1987 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
1988 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
1989 (wrestrict_dom_walker::check_call): Likewise.
1990 * gimple.c (gimple_build_call_from_tree): Likewise.
1991 * gimple.h (enum gf_mask): Likewise.
1992 (gimple_call_with_bounds_p): Likewise.
1993 (gimple_call_set_with_bounds): Likewise.
1994 * gimplify.c (gimplify_init_constructor): Likewise.
1995 * ipa-cp.c (initialize_node_lattices): Likewise.
1996 (propagate_constants_across_call): Likewise.
1997 (find_more_scalar_values_for_callers_subset): Likewise.
1998 * ipa-hsa.c (process_hsa_functions): Likewise.
1999 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
2000 * ipa-icf.c (sem_function::merge): Likewise.
2001 * ipa-inline.c (early_inliner): Likewise.
2002 * ipa-pure-const.c (warn_function_noreturn): Likewise.
2003 (warn_function_cold): Likewise.
2004 (propagate_pure_const): Likewise.
2005 * ipa-ref.h (enum GTY): Likewise.
2006 * ipa-split.c (find_retbnd): Likewise.
2007 (consider_split): Likewise.
2008 (split_function): Likewise.
2009 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
2010 * ipa.c (walk_polymorphic_call_targets): Likewise.
2011 (symbol_table::remove_unreachable_nodes): Likewise.
2012 (process_references): Likewise.
2013 (cgraph_build_static_cdtor_1): Likewise.
2014 * lto-cgraph.c (lto_output_node): Likewise.
2015 (output_refs): Likewise.
2016 (compute_ltrans_boundary): Likewise.
2017 (input_overwrite_node): Likewise.
2018 (input_node): Likewise.
2019 (input_cgraph_1): Likewise.
2020 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
2021 * passes.c (pass_manager::execute_early_local_passes): Likewise.
2022 (class pass_chkp_instrumentation_passes): Likewise.
2023 (make_pass_chkp_instrumentation_passes): Likewise.
2024 * passes.def: Likewise.
2025 * rtl.h (struct GTY): Likewise.
2026 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
2027 * stor-layout.c (layout_type): Likewise.
2028 * symtab.c: Likewise.
2029 * target.def: Likewise.
2030 * targhooks.c (default_chkp_bound_type): Likewise.
2031 (default_chkp_bound_mode): Likewise.
2032 (default_builtin_chkp_function): Likewise.
2033 (default_chkp_function_value_bounds): Likewise.
2034 (default_chkp_make_bounds_constant): Likewise.
2035 (default_chkp_initialize_bounds): Likewise.
2036 * targhooks.h (default_chkp_bound_type): Likewise.
2037 (default_chkp_bound_mode): Likewise.
2038 (default_builtin_chkp_function): Likewise.
2039 (default_chkp_function_value_bounds): Likewise.
2040 (default_chkp_make_bounds_constant): Likewise.
2041 (default_chkp_initialize_bounds): Likewise.
2042 * toplev.c (compile_file): Likewise.
2043 (process_options): Likewise.
2044 * tree-core.h (DEF_BUILTIN): Likewise.
2045 (DEF_BUILTIN_CHKP): Likewise.
2046 * tree-inline.c (declare_return_variable): Likewise.
2047 (remap_gimple_stmt): Likewise.
2048 (copy_bb): Likewise.
2049 (initialize_inlined_parameters): Likewise.
2050 (expand_call_inline): Likewise.
2051 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
2052 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
2053 (make_pass_ipa_chkp_produce_thunks): Likewise.
2054 (make_pass_chkp): Likewise.
2055 (make_pass_chkp_opt): Likewise.
2056 (make_pass_chkp_instrumentation_passes): Likewise.
2057 * tree-pretty-print.c (dump_generic_node): Likewise.
2058 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
2059 * tree-ssa-dce.c (propagate_necessity): Likewise.
2060 (eliminate_unnecessary_stmts): Likewise.
2061 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
2062 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
2063 * tree-ssa-sccvn.h: Likewise.
2064 * tree-ssa-strlen.c (get_string_length): Likewise.
2065 (valid_builtin_call): Likewise.
2066 (adjust_last_stmt): Likewise.
2067 (handle_builtin_strchr): Likewise.
2068 (handle_builtin_strcpy): Likewise.
2069 (handle_builtin_stxncpy): Likewise.
2070 (handle_builtin_memcpy): Likewise.
2071 (handle_builtin_strcat): Likewise.
2072 (strlen_check_and_optimize_stmt): Likewise.
2073 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
2074 * tree-streamer-in.c: Likewise.
2075 * tree-streamer.c (record_common_node): Likewise.
2076 * tree.c (tree_code_size): Likewise.
2077 (wide_int_to_tree_1): Likewise.
2078 (type_contains_placeholder_1): Likewise.
2079 (build_common_tree_nodes): Likewise.
2080 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
2081 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
2082 (POINTER_BOUNDS_P): Likewise.
2083 (BOUNDED_TYPE_P): Likewise.
2084 (BOUNDED_P): Likewise.
2085 (CALL_WITH_BOUNDS_P): Likewise.
2086 (pointer_bounds_type_node): Likewise.
2087 * value-prof.c (gimple_ic): Likewise.
2088 * var-tracking.c (vt_add_function_parameters): Likewise.
2089 * varasm.c (make_decl_rtl): Likewise.
2090 (assemble_start_function): Likewise.
2091 (output_constant): Likewise.
2092 (maybe_assemble_visibility): Likewise.
2093 * varpool.c (ctor_for_folding): Likewise.
2094 * chkp-builtins.def: Remove.
2095 * ipa-chkp.c: Remove.
2096 * ipa-chkp.h: Remove.
2097 * rtl-chkp.c: Remove.
2098 * rtl-chkp.h: Remove.
2099 * tree-chkp-opt.c: Remove.
2100 * tree-chkp.c: Remove.
2101 * tree-chkp.h: Remove.
2103 2018-06-07 Carl Love <cel@us.ibm.com>
2105 * gcc/config/rs6000/vsx.md (vextract_fp_from_shorth,
2106 vextract_fp_from_shortl): Add BE support.
2108 2018-06-07 Paul Koning <ni1d@arrl.net>
2110 * compare-elim.c (try_merge_compare): Don't merge compare if
2111 address contains a side effect.
2112 (try_eliminate_compare): Likewise.
2114 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
2116 * config.gcc: Support "tremont".
2117 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
2118 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2120 * config/i386/i386.c (m_TREMONT): Define.
2121 (processor_target_table): Add "tremont".
2122 (PTA_TREMONT): Define.
2123 (ix86_lea_outperforms): Add TARGET_TREMONT.
2124 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
2125 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
2126 and M_INTEL_GOLDMONT_PLUS.
2127 (fold_builtin_cpu): Add "tremont".
2128 (ix86_add_stmt_cost): Add TARGET_TREMONT.
2129 (ix86_option_override_internal): Add "tremont".
2130 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
2131 (processor_type): Add PROCESSOR_TREMONT.
2132 * config/i386/x86-tune.def: Add m_TREMONT.
2133 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
2135 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
2137 * gcc/config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
2138 symbol defined for msp430i* devices to be lower case.
2140 gcc/testsuite/gcc.target/msp430/
2141 * msp430i-device-symbol.c: New test.
2142 * msp430f-device-symbol.c: New test.
2143 * msp430.h: New test header file.
2145 2018-06-07 Richard Biener <rguenther@suse.de>
2147 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
2148 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
2149 Properly wrap signed arithmetic if overflow wraps.
2151 2018-06-07 Jakub Jelinek <jakub@redhat.com>
2153 PR tree-optimization/69615
2154 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
2155 of a cast from a same precision integral SSA_NAME in a bb dominated
2156 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
2157 cast to utype if rhs2 has already a compatible type.
2159 2018-06-07 Richard Biener <rguenther@suse.de>
2161 PR tree-optimization/85935
2162 * graphite-scop-detection.c (find_params_in_bb): Analyze
2163 condition operands with respect to the correct loop. Assert
2164 the analysis doesn't fail.
2166 2018-06-04 Carl Love <cel@us.ibm.com>
2168 * gcc/config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
2169 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
2170 as it is slightly cheaper.
2171 (first_match_or_eos_index_<mode>):
2172 Calculate index using natural element order.
2173 (first_match_index_<mode>):
2174 Calculate index using natural element order.
2175 (first_match_or_eos_index_<mode>):
2176 Calculate index using natural order.
2177 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
2178 for BE and LE modes.
2179 * gcc/config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
2180 P9V_BUILTIN_VCLZLSBB_V16QI.
2181 * gcc/config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
2184 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
2186 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
2187 indentation and line wrap for many prototypes. Add missing
2188 @smallexample directives around block of prototypes for vec_xl and
2191 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
2193 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
2194 track if we pass or return IEEE 128-bit floating point.
2195 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
2196 C++ mangling that is compatible with GCC 8.1.
2197 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
2198 (init_cumulative_args): Note if we pass or return IEEE 128-bit
2199 floating point types.
2200 (rs6000_function_arg_advance_1): Likewise.
2201 (rs6000_mangle_type): Optionally generate mangled names that match
2202 what GCC 8.1 generated for IEEE 128-bit floating point types.
2203 (rs6000_globalize_decl_name): If we have an external function that
2204 passes or returns IEEE 128-bit types, generate a weak reference
2205 from the mangled name used in GCC 8.1 to the current mangled
2207 (rs6000_init_builtins): Make __ibm128 use the long double type if
2208 long double is IBM extended double. Make __float128 use the long
2209 double type if long double is IEEE 128-bit.
2212 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
2213 macro for __ibm128 built-in functions.
2214 (PACK_IF): Add __ibm128 pack/unpack functions.
2215 (UNPACK_IF): Likewise.
2216 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
2217 enable long double built-in functions if long double is IEEE
2218 128-bit floating point.
2219 (rs6000_invalid_builtin): Update long double built-in function
2221 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
2222 functions, adjust the built-in function to use the long double
2223 built-in function if __ibm128 and long double are the same type.
2224 * doc/extend.texi (PowerPC builtins): Update documention for
2225 __builtin_{,un}pack_longdouble. Add documentation for
2226 __builtin_{,un}pack_ibm128.
2228 2018-06-06 Jim Wilson <jimw@sifive.com>
2230 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
2231 (struct machine_function): New field interrupt_mode.
2232 (riscv_handle_type_attribute): New function. Add forward declaration.
2233 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
2234 (riscv_expand_epilogue): Check interrupt_mode field.
2235 (riscv_set_current_function): Check interrupt attribute args and
2236 set interrupt_mode field.
2237 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
2238 (riscv_sret, riscv_uret): New.
2239 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
2240 new arguments to interrupt attribute.
2242 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
2245 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
2246 Don't handle -mcpu=power8 if -mpower9-vector is also used.
2248 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
2250 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
2251 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
2252 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
2253 several redundant entries.
2255 2018-06-06 David Malcolm <dmalcolm@redhat.com>
2257 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
2258 type from "rtx" to "rtx_insn *".
2259 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
2260 for local "call_insn", removing cast.
2261 (ix86_expand_call): Likewise, introducing a "call_insn" local.
2263 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
2265 PR tree-optimization/86066
2266 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
2267 for BIT_INSERT_EXPR stores.
2269 2018-06-06 Richard Biener <rguenther@suse.de>
2271 PR tree-optimization/86062
2272 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
2273 component refs ontop
2274 of to be offsetted base.
2276 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
2278 * gcc/config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
2279 to be static and remove check on interrupt attribute name.
2281 gcc/testsuite/gcc.target/msp430/
2282 * function-attributes-4.c: New test.
2283 * static-interrupts.c: New test.
2285 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
2287 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
2288 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
2290 2018-06-05 Steve Ellcey <sellcey@cavium.com>
2293 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
2295 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
2296 Remove second argument, change how error is called.
2297 (aarch64_layout_arg): Remove second argument from
2298 aarch64_err_no_fpadvsimd call.
2299 (aarch64_init_cumulative_args): Ditto.
2300 (aarch64_gimplify_va_arg_expr): Ditto.
2301 * config/aarch64/aarch64.md (mov<mode>): Ditto.
2303 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
2305 * config/i386/i386.md (simple_return_indirect_internal): New expander.
2306 (*simple_return_indirect_internal<mode>): Rename from
2307 simple_return_indirect_internal. Use W mode iterator.
2308 (rstorssp): New expander.
2309 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
2310 (clrssbsy): New expander.
2311 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
2313 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
2315 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
2317 (cmse_check_pointed_object): Likewise.
2319 2018-06-05 Martin Liska <mliska@suse.cz>
2321 PR gcov-profile/47618
2322 * doc/invoke.texi: Document how -fprofile-dir format
2325 2018-06-05 Richard Biener <rguenther@suse.de>
2327 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
2328 removal pretend DOM info isn't available so we do not update
2329 it and only remove edges, not dominated blocks. Actually free
2330 DOM info in case we removed something. Remove unreachable blocks.
2331 (mfb_keep_latches): Work with either DOM info or marked backedges.
2332 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
2333 first. Mark backedges if DOM info isn't available.
2334 (Re-)compute DOM info after cleanup_control_flow_pre.
2336 2018-06-05 Richard Biener <rguenther@suse.de>
2338 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
2339 (locus_discrim_hasher::hash): Adjust.
2340 (locus_discrim_hasher::equal): Likewise.
2341 (next_discriminator_for_locus): Work on line directly.
2342 (same_line_p): Pass in expanded locus1 as well.
2343 (assign_discriminators): Avoid redundant location expansions.
2345 2018-06-05 Richard Biener <rguenther@suse.de>
2347 PR tree-optimization/86046
2348 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
2349 if required after clearing TREE_ADDRESSABLE.
2351 2018-06-05 Richard Biener <rguenther@suse.de>
2353 PR tree-optimization/86047
2354 * tree-ssa-loop.c (for_each_index): Glob handling of all
2355 decls and constants and really handle all of them.
2357 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2360 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
2361 qualifier_void_pointer and qualifier_const_void_pointer.
2362 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
2363 (arm_init_builtins): Handle the above.
2364 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
2365 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
2366 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
2369 2018-06-05 Martin Liska <mliska@suse.cz>
2371 * auto-profile.c (read_autofdo_file): Do not use
2372 gcov_ctr_summary struct.
2373 (afdo_callsite_hot_enough_for_early_inline): Likewise.
2374 * coverage.c (struct counts_entry): Likewise.
2375 (read_counts_file): Read just single summary entry.
2376 (get_coverage_counts): Use gcov_summary struct.
2377 * coverage.h (get_coverage_counts): Likewise.
2378 * gcov-dump.c (dump_working_sets): Likewise.
2379 (tag_summary): Dump just single summary.
2380 * gcov-io.c (gcov_write_summary): Write just histogram
2382 (gcov_read_summary): Read just single summary.
2383 (compute_working_sets): Use gcov_summary struct.
2384 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
2385 of GCOV_COUNTERS_SUMMABLE.
2386 (GCOV_COUNTERS_SUMMABLE): Remove.
2387 (GCOV_FIRST_VALUE_COUNTER): Replace with
2388 GCOV_COUNTER_V_INTERVAL.
2389 (struct gcov_ctr_summary): Remove.
2390 (struct gcov_summary): Directly use fields of former
2392 (compute_working_sets): Use gcov_summary struct.
2393 * gcov.c (read_count_file): Do not use ctrs fields.
2394 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
2396 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
2398 * profile.c: Likewise.
2399 * profile.h: Likewise.
2401 2018-06-05 Martin Liska <mliska@suse.cz>
2403 PR gcov-profile/84846
2404 * gcov.c (output_lines): Print working directory only
2405 in intermediate format.
2407 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
2409 * config/s390/s390-builtin-types.def: Add void function type.
2410 * config/s390/s390-builtins.def: Use the function type for the
2413 2018-06-04 Jim Wilson <jimw@sifive.com>
2415 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
2417 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
2418 and maybe_eh_return. Change regno to unsigned int. Use new args to
2419 handle EH_RETURN_DATA_REGNO registers properly.
2420 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
2421 (riscv_expand_epilogue): Update comment. Change argument name and
2422 type. Update code to use new name and type. Pass new args to
2423 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
2425 * config/riscv/riscv.md (NORMAL_RETURN): New.
2426 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
2427 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
2428 (eh_return): Call gen_eh_return_internal and emit barrier.
2429 (eh_return_internal): Call riscv_expand_epilogue.
2431 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
2433 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
2434 bit_insertion field and declare can_be_merged_into method.
2435 (merged_store_group::can_be_merged_into): New method.
2436 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
2437 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
2438 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
2440 2018-06-04 Richard Biener <rguenther@suse.de>
2442 PR tree-optimization/85955
2443 * builtins.c (fold_builtin_sincos): Convert pointers to
2444 destination to appropriate type before dereferencing.
2446 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
2448 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
2450 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
2452 * expr.c (expand_expr_real_1): Force the operand into memory if
2453 its TYPE_MODE is BLKmode and if there is no integer mode for
2454 the number of bits being extracted.
2456 2018-06-04 Jakub Jelinek <jakub@redhat.com>
2460 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
2461 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
2463 2018-06-04 Richard Biener <rguenther@suse.de>
2465 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
2466 (cleanup_tree_cfg_noloop): ... single caller. Do
2467 start_recording_case_labels later.
2469 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
2471 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
2472 to _IMMINTRIN_H_INCLUDED.
2473 * config/i386/pconfigintrin.h: Ditto.
2474 * config/i386/waitpkgintrin.h: Ditto.
2475 * config/i386/immintrin.h: Add includes for sgxintrin.h,
2476 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
2477 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
2478 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
2479 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
2480 waitpkgintrin.h and cldemoteintrin.h.
2482 2018-06-04 Richard Biener <rguenther@suse.de>
2484 PR tree-optimization/86038
2485 * tracer.c (find_best_successor): Check probability for
2486 being initialized, bail out if not.
2488 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
2491 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
2492 of bits to ignore when comparing architectures.
2494 2018-06-04 Jakub Jelinek <jakub@redhat.com>
2496 PR tree-optimization/69615
2497 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
2498 maximum or minimum of the type, try to merge it also as if
2499 range1 is + [-, x - 1] or + [x + 1, -].
2502 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
2504 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
2506 PR tree-optimization/86034
2507 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
2508 the unsigned bitfield type in a bit insertion sequence if it does not
2509 have a larger precision than the bitfield size.
2510 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
2512 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
2514 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
2516 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
2518 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
2519 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
2520 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
2521 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
2523 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
2525 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
2526 Disable -fdelete-null-pointer-checks for ELF toolchain.
2528 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
2529 Kito Cheng <kito.cheng@gmail.com>
2531 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
2532 (nds32le-*-*, nds32be-*-*): Integrate checking process.
2533 (nds32*-*-*): Add glibc and uclibc conditions.
2534 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
2535 (TARGET_EXCEPT_UNWIND_INFO): Define.
2536 * config/nds32/elf.h: New file.
2537 * config/nds32/linux.h: New file.
2538 * config/nds32/nds32-elf.opt: New file.
2539 * config/nds32/nds32-linux.opt: New file.
2540 * config/nds32/nds32-fp-as-gp.c
2541 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
2542 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
2544 (nds32_asm_file_end): Ditto.
2545 (nds32_print_operand): Ditto.
2546 (nds32_insert_attributes): Ditto.
2547 (nds32_init_libfuncs): New function.
2548 (TARGET_HAVE_TLS): Define.
2549 (TARGET_INIT_LIBFUNCS): Define.
2550 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
2552 (TARGET_ELF): Apply different mcmodel setting.
2553 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
2554 been migrated into elf.h and linux.h files.
2555 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
2556 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
2557 (mcmodel): The content has been migrated into nds32-elf.opt and
2558 nds32-linux.opt files.
2559 * config/nds32/t-elf: New file.
2560 * config/nds32/t-linux: New file.
2562 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
2563 Shiva Chen <shiva0217@gmail.com>
2565 * config/nds32/constants.md (unspec_volatile_element): Add
2566 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
2567 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
2569 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
2570 (make_pass_nds32_fp_as_gp): Declare.
2571 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
2573 (nds32_asm_function_end_prologue): Remove unused asm output.
2574 (nds32_asm_function_begin_epilogue): Remove unused asm output.
2575 (nds32_asm_file_start): Output necessary fp_as_gp information.
2576 (nds32_option_override): Adjust register usage.
2577 (nds32_expand_prologue): Consider fp_as_gp situation.
2578 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
2579 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
2582 (simple_return): Ditto.
2583 (omit_fp_begin): Output special directive for fp_as_gp.
2584 (omit_fp_end): Output special directive for fp_as_gp.
2585 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
2586 mforbid-fp-as-gp): New options.
2588 2018-06-01 Mark Wielaard <mark@klomp.org>
2590 * dwarf2out.c (dwarf2out_finish): Remove generation of
2591 DW_AT_loclists_base.
2593 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
2595 * gimple-ssa-store-merging.c: Include gimple-fold.h.
2596 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
2597 (struct merged_store_group): Add bit_insertion field.
2598 (dump_char_array): Use standard hexadecimal format.
2599 (merged_store_group::merged_store_group): Set bit_insertion to false.
2600 (merged_store_group::apply_stores): Use optimal buffer size. Deal
2601 with BIT_INSERT_EXPR stores. Move up code updating the mask and
2602 also print the mask in the dump file.
2603 (pass_store_merging::gate): Minor tweak.
2604 (imm_store_chain_info::coalesce_immediate): Fix wrong association
2605 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
2606 stores with INTEGER_CST stores.
2607 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
2608 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
2609 and use it throughout. Generate bit insertion sequences if need be.
2610 (pass_store_merging::process_store): Remove redundant condition.
2611 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
2613 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
2615 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
2616 the 128-bit floating point types. Fix function comment.
2618 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2620 * config/aarch64/aarch64-simd.md
2621 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
2623 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
2626 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
2628 PR tree-optimization/85989
2629 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
2631 (backprop::intersect_uses): Check it when deciding whether this
2632 is a backedge reference.
2633 (backprop::process_block): Add each phi to m_visited_phis
2634 after visiting it, then clear it at the end.
2636 2018-06-01 Richard Biener <rguenther@suse.de>
2638 * tree-vectorizer.h (vect_dr_stmt): New function.
2639 (vect_get_load_cost): Adjust.
2640 (vect_get_store_cost): Likewise.
2641 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
2642 Use vect_dr_stmt instead of DR_SMTT.
2643 (vect_record_base_alignments): Likewise.
2644 (vect_calculate_target_alignment): Likewise.
2645 (vect_compute_data_ref_alignment): Likewise and make static.
2646 (vect_update_misalignment_for_peel): Likewise.
2647 (vect_verify_datarefs_alignment): Likewise.
2648 (vector_alignment_reachable_p): Likewise.
2649 (vect_get_data_access_cost): Likewise. Pass down
2650 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
2651 (vect_get_peeling_costs_all_drs): Likewise.
2652 (vect_peeling_hash_get_lowest_cost): Likewise.
2653 (vect_enhance_data_refs_alignment): Likewise.
2654 (vect_find_same_alignment_drs): Likewise.
2655 (vect_analyze_data_refs_alignment): Likewise.
2656 (vect_analyze_group_access_1): Likewise.
2657 (vect_analyze_group_access): Likewise.
2658 (vect_analyze_data_ref_access): Likewise.
2659 (vect_analyze_data_ref_accesses): Likewise.
2660 (vect_vfa_segment_size): Likewise.
2661 (vect_small_gap_p): Likewise.
2662 (vectorizable_with_step_bound_p): Likewise.
2663 (vect_prune_runtime_alias_test_list): Likewise.
2664 (vect_analyze_data_refs): Likewise.
2665 (vect_supportable_dr_alignment): Likewise.
2666 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
2667 (vect_gen_prolog_loop_niters): Likewise.
2668 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
2669 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
2671 (vect_recog_mask_conversion_pattern): Likewise.
2672 (vect_try_gather_scatter_pattern): Likewise.
2673 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
2674 to vect_get_store_cost.
2675 (vect_get_store_cost): Get stmt_info instead of DR.
2676 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
2677 (vect_get_load_cost): Get stmt_info instead of DR.
2679 2018-06-01 Richard Biener <rguenther@suse.de>
2682 * gimple-fold.c (var_decl_component_p): Also allow offsetted
2683 vars wrapped in MEM_REFs.
2685 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
2687 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
2688 Fix subreg tests so that we only return a choice between
2689 GENERAL_REGS and FP_REGS if the original classes included both.
2691 2018-06-01 Richard Biener <rguenther@suse.de>
2694 * tree-ssa-structalias.c (get_function_part_constraint):
2695 Handle NULL fi->decl.
2696 (find_func_aliases_for_call): Properly handle indirect
2697 fi from direct call.
2698 (find_func_clobbers): Likewise.
2699 (ipa_pta_execute): Likewise.
2700 (create_variable_info_for): For functions that are ifunc_resolver
2701 resolve to a varinfo that contains the result of the resolver call.
2702 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
2705 2018-05-31 Michael Collison <michael.collison@arm.com>
2707 * config/aarch64/aarch64.md:
2708 (*fix_to_zero_extenddfdi2): New pattern.
2709 * gcc.target/aarch64/fix_extend1.c: New testcase.
2711 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
2715 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
2716 and BUILT_IN_STRNCMP_EQ.
2717 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
2718 BUILT_IN_STRNCMP_EQ.
2719 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
2720 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
2721 (gimple_fold_builtin): Likewise.
2722 * tree-ssa-strlen.c (compute_string_length): New function.
2723 (determine_min_obsize): New function.
2724 (handle_builtin_string_cmp): New function to handle calls to
2725 string compare functions.
2726 (strlen_optimize_stmt): Add handling to builtin string compare
2728 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
2729 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
2730 * tree.c (build_common_builtin_nodes): Add new defines of
2731 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
2733 2018-05-31 Jakub Jelinek <jakub@redhat.com>
2736 * bb-reorder.c (pass_partition_blocks::gate): Return false for
2737 functions with naked attribute.
2739 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
2741 * config/i386/sse.md (avx_vec_concat<mode>):
2742 Substitute concat_tg_mode mode attribute with xtg_mode.
2743 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
2744 (concat_tg_mode): Remove mode attribute.
2746 2018-05-31 Martin Sebor <msebor@redhat.com>
2749 * calls.c (alloc_max_size): Correct a logic error/typo.
2750 Treat excessive arguments as infinite. Warn for invalid arguments.
2751 * doc/invoke.texi (-Walloc-size-larger-than): Update.
2753 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
2756 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
2757 and movx for Haswell.
2759 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
2760 Cesar Philippidis <cesar@codesourcery.com>
2763 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
2764 when emitting error on private/firstprivate reductions.
2765 * omp-low.c (lower_omp_target): Avoid reference-type processing
2766 on pointers for firstprivate clause.
2768 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
2770 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
2773 * config/aarch64/aarch64-simd.md
2774 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
2775 (aarch64_ld1_x3_<mode>): Likewise
2776 (aarch64_st1x2<VALLDIF:mode>): Likewise
2777 (aarch64_st1_x2_<mode>): Likewise
2778 (aarch64_st1x3<VALLDIF:mode>): Likewise
2779 (aarch64_st1_x3_<mode>): Likewise
2780 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
2781 (vld1_s8_x3): Likewise.
2782 (vld1_u16_x3): Likewise.
2783 (vld1_s16_x3): Likewise.
2784 (vld1_u32_x3): Likewise.
2785 (vld1_s32_x3): Likewise.
2786 (vld1_u64_x3): Likewise.
2787 (vld1_s64_x3): Likewise.
2788 (vld1_f16_x3): Likewise.
2789 (vld1_f32_x3): Likewise.
2790 (vld1_f64_x3): Likewise.
2791 (vld1_p8_x3): Likewise.
2792 (vld1_p16_x3): Likewise.
2793 (vld1_p64_x3): Likewise.
2794 (vld1q_u8_x3): Likewise.
2795 (vld1q_s8_x3): Likewise.
2796 (vld1q_u16_x3): Likewise.
2797 (vld1q_s16_x3): Likewise.
2798 (vld1q_u32_x3): Likewise.
2799 (vld1q_s32_x3): Likewise.
2800 (vld1q_u64_x3): Likewise.
2801 (vld1q_s64_x3): Likewise.
2802 (vld1q_f16_x3): Likewise.
2803 (vld1q_f32_x3): Likewise.
2804 (vld1q_f64_x3): Likewise.
2805 (vld1q_p8_x3): Likewise.
2806 (vld1q_p16_x3): Likewise.
2807 (vld1q_p64_x3): Likewise.
2808 (vst1_s64_x2): Likewise.
2809 (vst1_u64_x2): Likewise.
2810 (vst1_f64_x2): Likewise.
2811 (vst1_s8_x2): Likewise.
2812 (vst1_p8_x2): Likewise.
2813 (vst1_s16_x2): Likewise.
2814 (vst1_p16_x2): Likewise.
2815 (vst1_s32_x2): Likewise.
2816 (vst1_u8_x2): Likewise.
2817 (vst1_u16_x2): Likewise.
2818 (vst1_u32_x2): Likewise.
2819 (vst1_f16_x2): Likewise.
2820 (vst1_f32_x2): Likewise.
2821 (vst1_p64_x2): Likewise.
2822 (vst1q_s8_x2): Likewise.
2823 (vst1q_p8_x2): Likewise.
2824 (vst1q_s16_x2): Likewise.
2825 (vst1q_p16_x2): Likewise.
2826 (vst1q_s32_x2): Likewise.
2827 (vst1q_s64_x2): Likewise.
2828 (vst1q_u8_x2): Likewise.
2829 (vst1q_u16_x2): Likewise.
2830 (vst1q_u32_x2): Likewise.
2831 (vst1q_u64_x2): Likewise.
2832 (vst1q_f16_x2): Likewise.
2833 (vst1q_f32_x2): Likewise.
2834 (vst1q_f64_x2): Likewise.
2835 (vst1q_p64_x2): Likewise.
2836 (vst1_s64_x3): Likewise.
2837 (vst1_u64_x3): Likewise.
2838 (vst1_f64_x3): Likewise.
2839 (vst1_s8_x3): Likewise.
2840 (vst1_p8_x3): Likewise.
2841 (vst1_s16_x3): Likewise.
2842 (vst1_p16_x3): Likewise.
2843 (vst1_s32_x3): Likewise.
2844 (vst1_u8_x3): Likewise.
2845 (vst1_u16_x3): Likewise.
2846 (vst1_u32_x3): Likewise.
2847 (vst1_f16_x3): Likewise.
2848 (vst1_f32_x3): Likewise.
2849 (vst1_p64_x3): Likewise.
2850 (vst1q_s8_x3): Likewise.
2851 (vst1q_p8_x3): Likewise.
2852 (vst1q_s16_x3): Likewise.
2853 (vst1q_p16_x3): Likewise.
2854 (vst1q_s32_x3): Likewise.
2855 (vst1q_s64_x3): Likewise.
2856 (vst1q_u8_x3): Likewise.
2857 (vst1q_u16_x3): Likewise.
2858 (vst1q_u32_x3): Likewise.
2859 (vst1q_u64_x3): Likewise.
2860 (vst1q_f16_x3): Likewise.
2861 (vst1q_f32_x3): Likewise.
2862 (vst1q_f64_x3): Likewise.
2863 (vst1q_p64_x3): Likewise.
2865 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
2867 * config/msp430/msp430.c (msp430_output_labelref): Prepend
2868 user_label_prefix to name.
2870 * tree-core.h: Update comment about the format of NAME string
2871 passed to handler in attribute_spec.
2873 * config/msp430/msp430.md: Remove erroneous subreg expression from
2874 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
2875 zero_extend{q,h}isi2.
2877 2018-05-30 Borislav Petkov <bp@suse.de>
2879 * doc/extend.texi: Document some architecture specific
2880 constraints and sort entries.
2882 2018-05-30 Martin Sebor <msebor@redhat.com>
2885 * builtins.c (expand_builtin_stpcpy_1): New function.
2886 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
2887 only if the former succeeds.
2889 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
2891 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
2894 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
2896 * doc/invoke.texi (-flinker-output): Document
2898 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
2900 * passes.c (ipa_write_summaries): Only modify statements if body
2902 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
2903 incrementally linking.
2904 (ipa_passes): Likewise.
2905 * lto-cgraph.c (lto_output_node): When incrementally linking do not
2906 pass down resolution info.
2907 * common.opt (flag_incremental_link): Update info.
2908 * gcc.c (plugin specs): Turn flinker-output=* to
2909 -plugin-opt=-linker-output-known
2910 * toplev.c (compile_file): Also cut compilation when doing incremental
2912 * flag-types. (enum lto_partition_model): Add
2913 LTO_LINKER_OUTPUT_NOLTOREL.
2914 (invoke.texi): Add -flinker-output docs.
2915 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
2916 link same way as WPA; do not stream in dead initializers.
2918 * dwarf2out.c (dwarf2out_die_ref_for_decl,
2919 darf2out_register_external_decl): Support incremental link.
2921 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
2923 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
2925 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
2927 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
2928 it down to simple_object_copy_lto_debug_sections.
2929 (run_gcc): Determine incremental LTO link time and configure
2930 lto1 into non-wpa mode, disable renaming of debug sections.
2932 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
2934 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
2935 descriptions of various incorrectly documented functions.
2937 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
2940 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
2941 address check not strict.
2943 2018-05-30 Richard Biener <rguenther@suse.de>
2945 PR tree-optimization/85964
2946 * tracer.c (better_p): Drop initialized count check, we only
2947 call the function with initialized counts now.
2948 (find_best_successor): Do find a best edge if one
2949 has uninitialized count.
2950 (find_best_predecessor): Likewise. Do BB frequency check only
2951 if count is initialized.
2953 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
2955 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
2956 (aarch64_ldrstr_offset_compare): New.
2957 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
2958 load/store orderings.
2959 (aarch64_gen_adjusted_ldpstp): Likewise.
2961 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
2963 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
2964 Check for subset of GENERAL_REGS and FP_REGS.
2965 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
2968 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
2970 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
2971 and wi::to_poly_offset. Add the current offset and then check
2972 whether the sum fits, rather than using an unchecked addition of
2973 a checked term. Check for a shwi rather than a uhwi.
2974 * expr.c (get_bit_range): Use tree_to_poly_uint64.
2975 (store_constructor): Use poly_int_tree_p.
2976 (expand_expr_real_1): Likewise.
2977 * function.c (assign_temp): Likewise.
2978 * fold-const.c (const_binop): Use poly_int_tree_p and
2980 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
2982 * ipa-icf-gimple.c (func_checker::compare_operand): Use
2983 to_poly_offset for MEM offsets.
2984 * ipa-icf.c (sem_variable::equals): Likewise.
2985 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
2986 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
2987 wi::to_poly_offset for BIT_FIELD_REF offsets.
2988 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
2990 * var-tracking.c (emit_note_insn_var_location): Use
2991 tree_to_poly_uint64.
2993 2018-05-29 Jim Wilson <jimw@sifive.com>
2995 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
2997 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
3000 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
3001 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
3003 (sse4_1_round<mode>2): Use nonimmediate_operand
3004 for operand 1 predicate.
3006 2018-05-29 Martin Sebor <msebor@redhat.com>
3007 Richard Biener <rguenther@suse.de>
3010 * calls.c (get_size_range): Call determine_value_range instead
3011 of get_value_range..
3012 * tree-vrp.h (determine_value_range): Declared new function.
3013 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
3015 2018-05-29 Richard Biener <rguenther@suse.de>
3017 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
3018 sure to use non-pattern stmts for get_earlier_stmt arguments.
3019 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
3020 called on pattern stmts.
3021 (get_later_stmt): Likewise.
3023 2018-05-29 Martin Liska <mliska@suse.cz>
3025 PR gcov-profile/85759
3026 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
3029 2018-05-29 Jakub Jelinek <jakub@redhat.com>
3031 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
3033 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
3037 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
3038 VEC_PACK_FLOAT_EXPR): New tree codes.
3039 * tree-pretty-print.c (op_code_prio): Handle
3040 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
3041 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
3042 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
3043 * tree-inline.c (estimate_operator_cost): Likewise.
3044 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
3045 * fold-const.c (const_binop): Likewise.
3046 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
3047 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
3048 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
3049 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
3050 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
3051 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
3052 * expr.c (expand_expr_real_2): Likewise.
3053 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
3054 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
3055 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
3057 * optabs.c (expand_widen_pattern_expr): For
3058 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
3059 sign from result type rather than operand's type.
3060 (expand_binop_directly): For vec_packu_float_optab and
3061 vec_packs_float_optab allow result type to be different from operand's
3063 * optabs-tree.c (optab_for_tree_code): Handle
3064 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
3065 VEC_PACK_FLOAT_EXPR. Formatting fixes.
3066 * tree-vect-generic.c (expand_vector_operations_1): Handle
3067 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
3068 VEC_PACK_FLOAT_EXPR.
3069 * tree-vect-stmts.c (supportable_widening_operation): Handle
3071 (supportable_narrowing_operation): Handle FLOAT_EXPR.
3072 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
3073 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
3074 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
3075 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
3077 (vec_pack<floatprefix>_float_<mode>): New expander.
3078 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
3080 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
3081 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
3082 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
3083 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
3084 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
3086 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
3087 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
3088 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
3089 VEC_PACK_FLOAT_EXPR): Document.
3091 2018-05-29 Richard Biener <rguenther@suse.de>
3093 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
3095 (stmt_vec_info_vec): Make pointer.
3096 (init_stmt_vec_info_vec): Remove.
3097 (free_stmt_vec_info_vec): Likewise.
3098 (set_stmt_vec_info_vec): New function.
3099 (free_stmt_vec_infos): Likewise.
3100 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
3101 (set_vinfo_for_stmt): Likewise.
3102 (get_earlier_stmt): Likewise.
3103 (get_later_stmt): Likewise.
3104 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
3105 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
3106 (vec_info::~vec_info): Free stmt_vec_infos.
3107 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
3108 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
3109 (pass_slp_vectorize::execute): Likewise.
3110 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
3111 (free_stmt_vec_info_vec): Likewise.
3112 (set_stmt_vec_info_vec): New function.
3113 (free_stmt_vec_infos): Likewise.
3114 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
3115 the global stmt_vec_info_vec.
3116 * tree-parloops.c (gather_scalar_reductions): Use
3117 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
3120 2018-05-29 Richard Biener <rguenther@suse.de>
3122 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
3124 2018-05-29 Martin Liska <mliska@suse.cz>
3125 David Malcolm <dmalcolm@redhat.com>
3127 * vec.c (test_reverse): New.
3128 (vec_c_tests): Add new test.
3129 * vec.h (vl_ptr>::reverse): New function.
3131 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
3133 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
3135 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
3138 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
3140 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
3142 2018-05-28 Richard Biener <rguenther@suse.de>
3144 PR tree-optimization/85933
3145 * tree-vect-data-refs.c (vect_record_base_alignments): Only
3146 look at stmts marked as vectorizable.
3148 2018-05-28 Richard Biener <rguenther@suse.de>
3150 PR tree-optimization/85934
3151 * tree-vect-generic.c (expand_vector_operations_1): Hoist
3152 vector boolean check before scalar optimization.
3154 2018-05-28 Jakub Jelinek <jakub@redhat.com>
3156 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
3159 2018-05-28 Mark Wielaard <mark@klomp.org>
3161 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
3162 if it is an expression containing a minus sign.
3164 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
3166 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
3168 2018-05-27 Paul Koning <ni1d@arrl.net>
3170 * config/pdp11/pdp11.md (truncsihi2): Remove.
3172 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
3173 Chung-Ju Wu <jasonwucj@gmail.com>
3175 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
3177 (unaligned_store_dw): Ditto.
3178 * config/nds32/nds32-memory-manipulation.c
3179 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
3180 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
3181 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
3182 (emit_setmem_word_loop): Rename to ...
3183 (emit_setmem_doubleword_loop): ... this.
3184 (nds32_gen_dup_4_byte_to_word_value): New function.
3185 (nds32_gen_dup_8_byte_to_double_word_value): New function.
3186 (nds32_expand_setmem_loop): Refine implementation.
3187 (nds32_expand_setmem_loop_v3m): Ditto.
3188 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
3191 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
3193 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
3195 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
3197 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
3198 (nds32_init_machine_status): Initialize machine->attr_naked_p and
3199 machine->attr_no_prologue_p.
3200 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
3201 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
3202 (nds32_expand_epilogue): Consider attr_naked_p.
3203 (nds32_expand_epilogue_v3pop): Likewise.
3204 (nds32_can_use_return_insn): Likewise.
3205 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
3206 attr_no_prologue_p fields.
3207 * config/nds32/nds32.opt (mret-in-naked-func): New option.
3209 2018-05-27 Jakub Jelinek <jakub@redhat.com>
3212 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
3214 * config/i386/sse.md
3215 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
3217 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
3219 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
3221 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
3223 (*<floatsuffix>floatv2div2sf2): Rename to ...
3224 (*float<floatunssuffix>v2div2sf2): ... this.
3225 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
3226 (float<floatunssuffix>v2div2sf2_mask): ... this.
3227 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
3228 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
3229 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
3231 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
3233 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
3235 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
3237 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
3239 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
3241 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
3242 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
3243 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
3244 gen_ufix_truncv8dfv8si2.
3245 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
3246 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
3247 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
3248 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
3249 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
3250 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
3251 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
3252 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
3254 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
3258 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
3260 2018-05-25 Jim Wilson <jimw@sifive.com>
3262 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
3263 * config/riscv/riscv.c (struct machine_function): Add
3264 interrupt_handler_p and attribute_checked_p fields.
3265 (riscv_attribute_table): Add interrupt.
3266 (riscv_interrupt_type_p): New.
3267 (riscv_save_reg_p): Save extra regs for interrupt handler.
3268 (riscv_use_save_libcall): Return false for interrupt handler.
3269 (riscv_first_stack_step): Add forward declaration.
3270 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
3271 for interrupt handler with large frame. Use it for saved reg list.
3272 (riscv_expand_prologue): Move flag_stack_usage_info support to
3273 eliminate duplication.
3274 (riscv_expand_epilogue): Generate mret for interrupt handler.
3275 (riscv_epilogue_uses): New.
3276 (riscv_can_use_return_insn): Return false for interrupt handler.
3277 (riscv_function_ok_for_sibcall): Likewise.
3278 (riscv_set_current_function): Add interrupt handler support.
3279 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
3280 * config/riscv/riscv.md (UNSPECV_MRET): New.
3282 (riscv_frflags, riscv_fsflags): Use tab after opcode.
3284 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
3286 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
3288 PR tree-optimization/85712
3289 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
3290 this candidate has already been replaced in-situ by a copy.
3292 2018-05-25 Jason Merrill <jason@redhat.com>
3294 PR c++/80485 - inline function non-zero address.
3295 * symtab.c (nonzero_address): Check DECL_COMDAT.
3297 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
3300 * config/alpha/alpha.md (ashlsi3): New insn pattern.
3301 (*ashlsi_se): Rename from *ashldi_se. Define as sign
3302 extension of SImode operation. Use const123_operand predicate.
3303 (*saddsi_1): Remove.
3304 (*saddl_se_1): Ditto.
3306 (*ssubl_se_1): Ditto.
3307 * config/alpha/predicates.md (const123_operand): New predicate.
3308 * config/alpha/constraints.md (P): Use IN_RANGE.
3310 2018-05-25 Richard Biener <rguenther@suse.de>
3312 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
3314 (ref_maybe_used_by_stmt_p): Likewise.
3315 (stmt_may_clobber_ref_p): Likewise.
3316 (stmt_may_clobber_ref_p_1): Likewise.
3317 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
3319 (ref_maybe_used_by_stmt_p): Likewise.
3320 (stmt_may_clobber_ref_p): Likewise.
3321 (stmt_may_clobber_ref_p_1): Likewise.
3322 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
3323 the alias oracle to disambiguate DRs with stmts DR analysis
3325 (vect_analyze_data_refs): Do not give up on not analyzable
3326 DRs for BB vectorization. Remove code truncating the dataref
3329 2018-05-25 Jakub Jelinek <jakub@redhat.com>
3332 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
3333 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
3334 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
3336 2018-05-25 Richard Biener <rguenther@suse.de>
3338 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
3339 function, combining stmt data ref gathering and fatal analysis
3341 (vect_analyze_data_refs): Remove now redudnant code and simplify.
3342 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
3343 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
3344 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
3345 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
3347 2018-05-25 Bin Cheng <bin.cheng@arm.com>
3349 PR tree-optimization/85720
3350 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
3351 SCC if all partitions are builtins.
3352 (version_loop_by_alias_check): New parameter. Generate cancelable
3353 runtime alias check if all partitions are builtins.
3354 (distribute_loop): Update call to above function.
3356 2018-05-25 Bin Cheng <bin.cheng@arm.com>
3358 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
3359 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
3360 (parm_default_def_partition_arg): Ditto.
3361 (set_parm_default_def_partition): Ditto.
3362 (get_parm_default_def_partitions): Ditto and make it static.
3363 (get_undefined_value_partitions): Ditto and make it static.
3364 (remove_ssa_form): Refactor call to init_var_map here.
3365 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
3366 computation for loop region.
3367 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
3368 (register_default_def): Delete.
3369 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
3370 (parm_default_def_partition_arg): Ditto.
3371 (set_parm_default_def_partition): Ditto.
3372 (get_parm_default_def_partitions): Ditto and make it static.
3373 (get_undefined_value_partitions): Ditto and make it static.
3374 (coalesce_with_default, coalesce_with_default): Update comment.
3375 (create_coalesce_list_for_region): New func factored out from
3376 create_outofssa_var_map.
3377 (populate_coalesce_list_for_outofssa): New func factored out from
3378 create_outofssa_var_map and coalesce_ssa_name.
3379 (create_outofssa_var_map): Delete.
3380 (coalesce_ssa_name): Refactor to support live range computation.
3381 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
3382 (get_parm_default_def_partitions): Delete.
3383 (get_undefined_value_partitions): Ditto.
3384 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
3385 computation for loop region.
3386 (new_tree_live_info, loe_visit_block): Ditto.
3387 (live_worklist, set_var_live_on_entry): Ditto.
3388 (calculate_live_on_exit, verify_live_on_entry): Ditto.
3389 * tree-ssa-live.h (struct _var_map): New fields.
3390 (init_var_map): Change decl.
3391 (region_contains_p): New.
3393 2018-05-25 Bin Cheng <bin.cheng@arm.com>
3395 * tree-ssa-live.h (live_merge_and_clear): Delete.
3397 2018-05-25 Richard Biener <rguenther@suse.de>
3400 * tree-dump.c (dequeue_and_dump): Remove access to removed
3401 operand 2 of a SWITCH_EXPR.
3403 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3405 * doc/sourcebuild.texi (vect_double_cond_arith): Include
3406 multiplication and division.
3407 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
3408 (cond_udiv@var{m}, cond_umod@var{m}): Document.
3409 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
3410 (cond_udiv_optab, cond_umod_optab): New optabs.
3411 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
3412 (IFN_COND_RDIV): New internal functions.
3413 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
3414 TRUNC_MOD_EXPR and RDIV_EXPR.
3415 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
3416 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
3418 (SVE_INT_BINARY): Include mult.
3419 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
3420 (optab, sve_int_op): Handle mult.
3421 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
3423 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
3424 for SVE_INT_BINARY_SD.
3426 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3428 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
3429 (optab, sve_int_op): Handle div and udiv.
3430 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
3431 for SVE_INT_BINARY_SD.
3432 (*<optab><mode>3): New insn for the same.
3434 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3436 * tree-vect-patterns.c: Include predict.h.
3437 (vect_recog_divmod_pattern): Restrict check for division support
3438 to when optimizing for size.
3440 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3442 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
3443 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
3444 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
3445 (gimple_match_op::set_op): Likewise.
3446 (gimple_resimplify4): Declare.
3447 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
3448 (expr::gen_transform): Likewise.
3449 (decision_tree::gen): Generate a simplification routine for 4 operands.
3450 * gimple-match-head.c (gimple_simplify): Add an overload for
3451 4 operands. In the top-level function, handle up to 4 call
3452 arguments and call gimple_resimplify4.
3453 (gimple_resimplify4): New function.
3454 (build_call_internal): Pass a fourth operand.
3455 (maybe_push_to_seq): Likewise.
3456 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
3457 Fold VEC_COND_EXPRs of an operation and a default value into
3458 an IFN_COND_* function if possible.
3459 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
3461 (SVE_COND_FP_BINARY): Include them.
3462 (optab, sve_fp_op): Handle them.
3463 (SVE_INT_BINARY_REV): New code iterator.
3464 (SVE_COND_FP_BINARY_REV): New int iterator.
3465 (commutative): New int attribute.
3466 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
3468 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
3470 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
3471 (*cond_<optab><mode>): New patterns for reversed operands.
3473 2018-05-25 Richard Biener <rguenther@suse.de>
3475 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
3476 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
3477 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
3478 (STMT_VINFO_GROUPED_ACCESS): Adjust.
3479 * tree-vect-data-refs.c (everywhere): Adjust users.
3480 * tree-vect-loop.c (everywhere): Likewise.
3481 * tree-vect-slp.c (everywhere): Likewise.
3482 * tree-vect-stmts.c (everywhere): Likewise.
3483 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
3485 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3487 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
3489 (gcc_cv_as_section_exclude): ... this.
3490 Try Solaris as #exclude syntax.
3491 * configure: Regenerate.
3492 * config.in: Regenerate.
3493 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
3495 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
3496 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
3498 * varasm.c (default_elf_asm_named_section): Don't check if
3499 HAVE_GAS_SECTION_EXCLUDE is defined.
3501 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3503 * doc/md.texi: Update the documentation of the cond_* optabs
3504 to mention the new final operand. Fix GET_MODE_NUNITS call.
3505 Describe the scalar case too.
3506 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
3507 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
3509 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
3510 (get_conditional_internal_fn): Update comment.
3511 * tree-vect-loop.c (vectorizable_reduction): Pass the original
3512 accumulator value as a final argument to conditional functions.
3513 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
3514 a define_expand and add an "else" operand. Assert for now that
3515 the else operand is equal to operand 2. Use SVE_INT_BINARY and
3516 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
3517 (*cond_<optab><mode>): New patterns.
3518 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
3519 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
3520 (UNSPEC_COND_EOR): Delete.
3521 (optab): Remove associated mappings.
3522 (SVE_INT_BINARY): New code iterator.
3523 (sve_int_op): Remove int attribute and add "minus" to the code
3525 (SVE_COND_INT_OP): Delete.
3526 (SVE_COND_FP_OP): Rename to...
3527 (SVE_COND_FP_BINARY): ...this.
3529 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3531 * optabs.c (can_reuse_operands_p): New function.
3532 (maybe_legitimize_operands): Try to reuse the results for
3535 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
3537 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
3538 Add {q} suffix to insn mnemonic.
3540 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
3542 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
3543 (msp430_warn_func_return): New.
3545 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
3547 * fold-const.c (tree_nonzero_bits): New function.
3548 * fold-const.h (tree_nonzero_bits): Likewise.
3549 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
3550 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
3552 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
3556 * varasm.c (assemble_alias): Check ifunc_resolver only on
3559 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
3562 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
3563 when memory input operand is handled.
3565 2018-05-24 Luis Machado <luis.machado@linaro.org>
3567 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
3569 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
3571 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
3573 * match.pd: Delay FMA folds until after vectorization.
3575 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
3578 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
3579 address check not strict.
3581 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
3583 * gimple-match.h (gimple_match_op): New class.
3584 (mprts_hook): Replace parameters with a gimple_match_op *.
3585 (maybe_build_generic_op): Likewise.
3586 (gimple_simplified_result_is_gimple_val): Replace parameters with
3587 a const gimple_match_op *.
3588 (gimple_simplify): Replace code_helper * and tree * parameters with
3589 a gimple_match_op * parameter.
3590 (gimple_resimplify1): Replace code_helper *, tree and tree *
3591 parameters with a gimple_match_op * parameter.
3592 (gimple_resimplify2): Likewise.
3593 (gimple_resimplify3): Likewise.
3594 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
3595 parameters with a gimple_match_op * parameter.
3596 * gimple-match-head.c (gimple_simplify): Change prototypes of
3597 auto-generated functions to take a gimple_match_op * instead of
3598 separate code_helper * and tree * parameters. Make the same
3599 change in the top-level overload and update calls to the
3600 gimple_resimplify routines. Update calls to the auto-generated
3601 functions and to maybe_push_res_to_seq in the publicly-facing
3602 operation-specific gimple_simplify overloads.
3603 (gimple_match_op::MAX_NUM_OPS): Define.
3604 (gimple_resimplify1): Replace rcode and ops with a single res_op
3605 parameter. Update call to gimple_simplify.
3606 (gimple_resimplify2): Likewise.
3607 (gimple_resimplify3): Likewise.
3608 (mprts_hook): Replace parameters with a gimple_match_op *.
3609 (maybe_build_generic_op): Likewise.
3610 (build_call_internal): Replace type, nargs and ops with
3611 a gimple_match_op *.
3612 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
3613 with a single gimple_match_op *. Update calls to mprts_hook,
3614 build_call_internal and gimple_simplified_result_is_gimple_val.
3615 Factor out code that is common to the tree_code and combined_fn cases.
3616 * genmatch.c (expr::gen_transform): Replace tem_code and
3617 tem_ops with a gimple_match_op called tem_op. Update calls
3618 to the gimple_resimplify functions and maybe_push_res_to_seq.
3619 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
3620 res_ops. Update call to the gimple_resimplify functions.
3621 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
3622 (decision_tree::gen): Make the functions take a gimple_match_op *
3623 called res_op instead of separate res_code and res_ops parameters.
3624 Update call accordingly.
3625 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
3626 and ops with a single res_op parameter. Update calls to
3627 maybe_build_generic_op and maybe_push_res_to_seq.
3628 (fold_stmt_1): Update calls to gimple_simplify and
3629 replace_stmt_with_simplification.
3630 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
3631 and gimple_simplified_result_is_gimple_val.
3632 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
3634 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
3635 with a gimple_match_op *.
3636 (vn_nary_build_or_lookup): Likewise. Update call to
3637 vn_nary_build_or_lookup_1.
3638 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
3639 gimple_match_op *. Update calls to the gimple_resimplify routines
3640 and to gimple_simplified_result_is_gimple_val.
3641 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
3642 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
3643 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
3644 (visit_nary_op): Likewise.
3645 (visit_reference_op_load): Likewise.
3647 2018-05-23 Luis Machado <luis.machado@linaro.org>
3649 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
3650 modifier for printing the step amount.
3652 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
3655 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
3658 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
3660 * doc/sourcebuild.texi (Endianness): New subsubsection.
3662 2018-05-23 Luis Machado <luis.machado@linaro.org>
3664 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3665 <prefetch_dynamic_strides>: New const bool field.
3666 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3667 prefetch_dynamic_strides.
3668 (exynosm1_prefetch_tune): Likewise.
3669 (thunderxt88_prefetch_tune): Likewise.
3670 (thunderx_prefetch_tune): Likewise.
3671 (thunderx2t99_prefetch_tune): Likewise.
3672 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
3674 (aarch64_override_options_internal): Update to set
3675 PARAM_PREFETCH_DYNAMIC_STRIDES.
3676 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
3677 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
3678 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
3679 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
3680 prefetch-dynamic-strides setting.
3682 2018-05-23 Luis Machado <luis.machado@linaro.org>
3684 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3685 <minimum_stride>: New const int field.
3686 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3687 minimum_stride field defaulting to -1.
3688 (exynosm1_prefetch_tune): Likewise.
3689 (thunderxt88_prefetch_tune): Likewise.
3690 (thunderx_prefetch_tune): Likewise.
3691 (thunderx2t99_prefetch_tune): Likewise.
3692 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
3693 <default_opt_level>: Set to 3.
3694 (aarch64_override_options_internal): Update to set
3695 PARAM_PREFETCH_MINIMUM_STRIDE.
3696 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
3697 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
3698 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
3699 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
3700 stride is constant and is below the minimum stride threshold.
3702 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3704 * config/arm/arm-cpus.in (mode26): Delete.
3705 (armv4): Delete mode26 reference.
3706 * config/arm/arm.c (arm_configure_build_target): Delete use of
3709 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
3711 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
3713 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
3714 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
3716 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
3717 (floatunsdidf2): Ditto.
3719 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
3721 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
3722 (fixuns_trunc<mode>si2_avx512f): Ditto.
3723 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
3724 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
3725 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
3727 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
3729 PR rtl-optimization/79985
3730 * df-scan.c (df_insn_refs_collect): Remove special case for
3731 global registers and asm statements.
3733 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
3735 * extend.texi (Global Register Variables): Rewrite the bullet list.
3736 Note that the register is available for allocation. Note that access
3737 via inline asm must use constraints. Add note about async-signal
3738 handlers. Remove paragraph about automagic register selection.
3740 2018-05-23 Richard Biener <rguenther@suse.de>
3742 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
3743 of fixed offset from memset VN.
3745 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
3747 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
3749 (alloc_cand_and_find_basis): Initialize first_interp field.
3750 (slsr_process_mul): Modify first_interp field.
3751 (slsr_process_add): Likewise.
3752 (slsr_process_cast): Modify first_interp field for each new
3754 (slsr_process_copy): Likewise.
3755 (dump_candidate): Dump first_interp field.
3756 (replace_mult_candidate): Process all interpretations, not just
3758 (replace_rhs_if_not_dup): Likewise.
3759 (replace_one_candidate): Likewise.
3761 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
3763 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
3765 (aarch64_needs_frame_chain): New function.
3766 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
3768 2018-05-23 Sudakshina Das <sudi.das@arm.com>
3771 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
3772 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
3773 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
3774 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
3775 as true for strict-align.
3776 (aarch64_can_inline_p): Perform checks even when callee has no
3777 attributes to check for strict alignment.
3778 * doc/extend.texi (AArch64 Function Attributes): Document
3780 * doc/invoke.texi: (AArch64 Options): Likewise.
3782 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
3784 PR tree-optimization/85853
3785 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
3786 the handling of the root of the node to...
3787 (vect_slp_analyze_node_operations_1): ...this new function,
3788 and run the whole thing with the child nodes' def types
3789 set according to their SLP node's def type.
3791 2018-05-23 Richard Biener <rguenther@suse.de>
3794 * tree-data-ref.c (create_runtime_alias_checks): Defer
3795 and ignore overflow warnings.
3797 2018-05-23 Yury Gribov <tetra2005@gmail.com>
3799 PR tree-optimization/85822
3800 * tree-vrp.c (is_masked_range_test): Fix handling of negative
3803 2018-05-23 Richard Biener <rguenther@suse.de>
3805 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
3806 memset constants via native_interpret_expr.
3808 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
3811 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
3813 (cgraph_node::create_alias): Likewise.
3814 (cgraph_node::get_availability): Check ifunc_resolver instead
3815 of looking up ifunc attribute.
3816 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
3817 * varasm.c (do_assemble_alias): Likewise.
3818 (assemble_alias): Likewise.
3819 (default_binds_local_p_3): Likewise.
3820 * cgraph.h (cgraph_node): Add ifunc_resolver.
3821 (cgraph_node::only_called_directly_or_aliased_p): Return false
3823 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
3825 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
3826 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
3827 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
3828 instead of looking up ifunc attribute.
3830 2018-05-22 Luis Machado <luis.machado@linaro.org>
3832 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
3834 2018-05-22 Martin Sebor <msebor@redhat.com>
3837 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
3838 only when expasion succeeds.
3839 (expand_builtin_strcmp): Same.
3840 (expand_builtin_strncmp): Same.
3842 2018-05-22 Martin Sebor <msebor@redhat.com>
3844 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
3846 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
3847 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3849 * config/aarch64/aarch64-ldpstp.md: Replace uses of
3850 aarch64_mem_pair_operand with memory_operand and delete operand swapping
3852 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
3853 Add check for legitimate_address.
3854 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
3855 (aarch64_swap_ldrstr_operands): New.
3856 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
3859 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
3860 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3862 * config/aarch64/aarch64.md: New patterns to generate stp
3864 (store_pair_sw, store_pair_dw): New patterns to generate stp for
3865 single words and double words.
3866 (load_pair_sw, load_pair_dw): Likewise.
3867 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
3869 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
3871 * config/aarch64/aarch64-ldpstp.md: Modify peephole
3872 for different mode ldpstp and add peephole for merged zero stores.
3874 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
3876 (aarch64_gen_store_pair): Rename calls to match new patterns.
3877 (aarch64_gen_load_pair): Rename calls to match new patterns.
3878 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
3879 (load_pair<DREG:mode><DREG2:mode>): ... This.
3880 (store_pair<mode>): Rename to...
3881 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
3882 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
3884 (V_INT_EQUIV): Handle SImode.
3885 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
3888 2018-05-22 Martin Sebor <msebor@redhat.com>
3891 * calls.c (maybe_warn_nonstring_arg): Use string length to set
3892 or ajust the presumed bound on an operation to avoid unnecessary
3895 2018-05-22 Martin Sebor <msebor@redhat.com>
3897 PR tree-optimization/85826
3898 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
3899 assuming that a DECL necesarily has a constant size.
3901 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
3904 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
3906 2018-05-22 Richard Biener <rguenther@suse.de>
3908 PR tree-optimization/85834
3909 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
3910 non-constant and non-zero memset arguments.
3912 2018-05-22 Martin Liska <mliska@suse.cz>
3915 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
3917 2018-05-22 Richard Biener <rguenther@suse.de>
3919 PR tree-optimization/85863
3920 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
3921 comparisons when vectype is specified.
3922 (vectorizable_condition): Do not specify vectype for
3923 vect_is_simple_cond when SLP vectorizing.
3925 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
3928 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
3929 define __ibm128 as long double.
3930 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
3931 as a distinct type when IEEE 128-bit support is enabled.
3932 (init_float128_ieee): Fix up conversions between IFmode and IEEE
3933 128-bit types to use the correct functions.
3934 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
3935 convert between 128-bit floating point types that have different
3936 modes but the same representation, instead of using gen_lowpart to
3938 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
3940 (IFKF_reg): New attributes to give the register constraints for
3942 (extend<mode>tf2_internal): New insns to mark an explicit
3943 conversion between 128-bit floating point types that have a
3944 different mode but share the same representation.
3946 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
3948 PR tree-optimization/85814
3949 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
3950 a null return from get_strinfo when unsharing the next
3951 strinfo in the chain.
3953 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
3956 * varasm.c (weak_finish): Clean up weak_decls.
3958 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3960 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
3961 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
3962 UNSPEC_UADALP values.
3963 * config/aarch64/iterators.md (ABAL): New int iterator.
3966 (sur): Add mappings for the above.
3967 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
3969 (aarch64_<sur>abal<mode>_4): Likewise.
3970 (aarch64_<sur>adalp<mode>_3): Likewise.
3971 (<sur>sadv16qi): New define_expand.
3973 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
3975 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
3976 (*movdf_internal): Ditto.
3977 (*rcpsf2_sse): Ditto.
3978 (*rsqrtsf2_sse): Ditto.
3979 (*sqrt<mode>2_sse): Ditto.
3981 2018-05-21 Tamar Christina <tamar.christina@arm.com>
3983 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
3985 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
3986 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
3987 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
3988 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
3990 * config/aarch64/arm_neon.h: Likewise.
3991 * config/aarch64/iterators.md (VQ_I): New.
3993 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
3995 * config.gcc: Add arc/t-multilib-linux to tmake_file for
3997 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
4000 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
4002 * config/nds32/constraints.md (S): New constraint.
4003 * config/nds32/nds32.md (call_internal): Use constraint S.
4004 (call_value_internal): Likewise.
4005 (sibcall_internal): Likewise.
4006 (sibcall_value_internal): Likewise.
4008 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
4009 Chung-Ju Wu <jasonwucj@gmail.com>
4011 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
4014 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
4015 Chung-Ju Wu <jasonwucj@gmail.com>
4017 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
4018 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
4019 (nds32_rtx_costs_impl): Simplify.
4020 (nds32_address_cost_impl): Simplify.
4021 (nds32_init_rtx_costs): New function.
4022 (nds32_rtx_costs_speed_prefer): Likewise.
4023 (nds32_rtx_costs_size_prefer): Likewise.
4024 (nds32_address_cost_speed_prefer): Likewise.
4025 (nds32_address_cost_speed_fwprop): Likewise.
4026 (nds32_address_cost_size_prefer): Likewise.
4027 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
4028 * config/nds32/nds32.c (nds32_option_override): Use
4029 nds32_init_rtx_costs function.
4031 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
4033 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
4034 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
4035 (TARGET_PIPELINE_N8): Likewise.
4036 (TARGET_PIPELINE_N10): Likewise.
4037 (TARGET_PIPELINE_N13): Likewise.
4038 (TARGET_PIPELINE_GRAYWOLF): Likewise.
4040 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
4042 * config/nds32/nds32-fpu.md: Update copyright year.
4044 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
4046 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
4048 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
4050 * config/nds32/nds32.c
4051 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
4052 * config/nds32/nds32.opt (minline-asm-r15): New option.
4054 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
4056 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
4058 * config/nds32/nds32.md (abssi2): New pattern.
4060 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
4062 * config/i386/i386.md (rex64namesuffix): New mode attribute.
4063 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
4064 Merge insn pattern from sse_cvtsi2ss<round_name> and
4065 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
4066 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
4067 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
4068 using SWI48 mode iterator.
4069 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
4070 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
4071 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
4072 pattern from sse_cvttss2si<round_saeonly_name>
4073 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
4074 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
4075 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
4076 using SWI48 mode iterator.
4077 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
4078 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
4079 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
4080 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
4081 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
4082 using SWI48 mode iterator.
4083 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
4084 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
4085 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
4086 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
4087 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
4088 SWI48 mode iterator.
4089 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
4090 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
4091 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
4092 pattern from sse_cvttsd2si<round_saeonly_name>
4093 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
4095 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
4097 * config/nds32/nds32-md-auxiliary.c
4098 (nds32_valid_smw_lwm_base_p): Refine.
4099 (nds32_output_smw_single_word): Refine.
4100 (nds32_output_smw_double_word): New.
4101 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
4103 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
4105 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
4106 (nds32_output_stack_pop): Refine.
4107 (nds32_expand_unaligned_load): Refine.
4108 (nds32_expand_unaligned_store): Refine.
4110 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
4111 Chung-Ju Wu <jasonwucj@gmail.com>
4113 * config/nds32/constants.md: Add TP_REGNUM constant.
4114 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
4115 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
4117 * config/nds32/nds32-doubleword.md: Consider flag_pic.
4118 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
4119 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
4120 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
4121 and PIC code generation.
4122 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
4124 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
4126 * config/nds32/nds32.md: Support TLS and PIC.
4127 * config/nds32/nds32.c: Support TLS and PIC.
4128 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
4129 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
4132 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
4134 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
4135 mode with E_ prefix.
4137 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
4138 Chung-Ju Wu <jasonwucj@gmail.com>
4140 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
4141 * config/nds32/nds32-md-auxiliary.c
4142 (symbolic_reference_mentioned_p): New.
4143 (nds32_legitimize_ict_address): New.
4144 (nds32_expand_ict_move): New.
4145 (nds32_indirect_call_referenced_p): New.
4146 (nds32_symbol_binds_local_p): Delete.
4147 (nds32_long_call_p): Modify.
4148 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
4149 * config/nds32/nds32-protos.h
4150 (symbolic_reference_mentioned_p): Declare.
4151 (nds32_legitimize_ict_address): Declare.
4152 (nds32_expand_ict_move): Declare.
4153 (nds32_indirect_call_referenced_p): Declare.
4154 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
4155 (nds32_relax_group): Use nds32_ict_const_p as condition.
4156 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
4157 (nds32_asm_file_start): Output ict_model directive in asm code.
4158 (nds32_legitimate_address_p): Consider indirect call.
4159 (nds32_print_operand): Consider indirect call.
4160 (nds32_print_operand_address): Consider indirect call.
4161 (nds32_insert_attributes): Handle "indirect_call" attribute.
4162 (TARGET_LEGITIMATE_ADDRESS_P): Define.
4163 (TARGET_LEGITIMATE_CONSTANT_P): Define.
4164 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
4165 (TARGET_DELEGITIMIZE_ADDRESS): Define.
4166 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
4167 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
4168 (TARGET_ICT_MODEL_SMALL): Define.
4169 (TARGET_ICT_MODEL_LARGE): Define.
4170 * config/nds32/nds32.md (movsi): Consider ict model.
4171 (call, call_value): Consider ict model.
4172 (sibcall, sibcall_value): Consider ict model.
4173 * config/nds32/nds32.opt (mict-model): New option.
4174 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
4177 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
4178 Monk Chiang <sh.chiang04@gmail.com>
4179 Jim Wilson <jimw@sifive.com>
4181 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
4182 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
4183 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
4184 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
4185 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
4186 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
4187 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
4188 compute save_libcall_adjustment properly.
4189 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
4190 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
4191 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
4192 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
4193 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
4194 (ABI_SPEC): Handle mabi=ilp32e.
4195 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
4196 (RVE): Add RVE mask.
4197 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
4198 <-march>: Add rv32e as an example.
4200 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
4203 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
4204 (intra_create_variable_infos): Handle C++ constructors.
4206 2018-05-18 Martin Liska <mliska@suse.cz>
4208 * passes.def: Remove a redundant pass.
4210 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
4213 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
4215 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4217 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
4219 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
4221 (armv2, armv2a, armv3, armv3m): Delete architectures.
4222 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
4223 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
4224 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
4226 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
4227 (*mulsidi3adddi): Likewise.
4228 (mulsidi3): Likewise.
4229 (*mulsidi3_nov6): Likewise.
4230 (umulsidi3): Likewise.
4231 (umulsidi3_nov6): Likewise.
4232 (umaddsidi4): Likewise.
4233 (*umulsidi3adddi): Likewise.
4234 (smulsi3_highpart): Likewise.
4235 (*smulsi3_highpart_nov6): Likewise.
4236 (umulsi3_highpart): Likewise.
4237 (*umulsi3_highpart_nov6): Likewise.
4238 * config/arm/arm.h (arm_arch3m): Delete.
4239 * config/arm/arm.c (arm_arch3m): Delete.
4240 (arm_option_override_internal): Update armv3-related comment.
4241 (arm_configure_build_target): Delete use of isa_bit_mode32.
4242 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
4243 (arm_rtx_costs_internal): Delete check of arm_arch3m.
4244 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
4246 (mulusa3): Likewise.
4247 * config/arm/arm-protos.h (arm_arch3m): Delete.
4248 * config/arm/arm-tables.opt: Regenerate.
4249 * config/arm/arm-tune.md: Likewise.
4250 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
4251 deleted architectures.
4253 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4255 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
4256 (armv5t, armv5te): New features.
4257 (ARMv5, ARMv5e): Delete fgroups.
4258 (ARMv5t, ARMv5te): Adjust for above changes.
4260 (armv5, armv5e): Delete arches.
4261 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
4263 (*call_reg_arm): Likewise.
4264 (*call_value_reg_armv5): Likewise.
4265 (*call_value_reg_arm): Likewise.
4266 (*call_symbol): Likewise.
4267 (*call_value_symbol): Likewise.
4268 (*sibcall_insn): Likewise.
4269 (*sibcall_value_insn): Likewise.
4271 (prefetch): Likewise.
4272 (define_split and define_peephole2 dependent on arm_arch5):
4274 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
4276 (TARGET_ARM_QBIT): Likewise.
4277 (TARGET_DSP_MULTIPLY): Likewise.
4278 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
4279 (arm_arch5, arm_arch5e): Delete.
4280 (arm_arch5t, arm_arch5te): Declare.
4281 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
4282 (arm_arch5t): Declare.
4283 (arm_option_reconfigure_globals): Update for the above.
4284 (arm_options_perform_arch_sanity_checks): Update comment, replace
4285 use of arm_arch5 with arm_arch5t.
4286 (use_return_insn): Likewise.
4287 (arm_emit_call_insn): Likewise.
4288 (output_return_instruction): Likewise.
4289 (arm_final_prescan_insn): Likewise.
4290 (arm_coproc_builtin_available): Likewise.
4291 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
4292 arm_arch5e with arm_arch5t and arm_arch5te.
4293 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
4294 (arm_arch5t, arm_arch5te): Declare.
4295 * config/arm/arm-tables.opt: Regenerate.
4296 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
4297 * config/arm/t-multilib: Likewise.
4298 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
4299 instead of arm_arch5.
4300 (*call_reg_thumb1): Likewise.
4301 (*call_value_reg_thumb1_v5): Likewise.
4302 (*call_value_reg_thumb1): Likewise.
4303 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
4305 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
4307 2018-05-18 Martin Liska <mliska@suse.cz>
4309 PR gcov-profile/84846
4310 * doc/gcov.texi: Document -t option of gcov tool.
4312 2018-05-18 Martin Liska <mliska@suse.cz>
4314 PR gcov-profile/84846
4315 * gcov.c (print_usage): Add new -t option.
4316 (process_args): Handle the option.
4317 (generate_results): Use stdout as output when requested by
4320 2018-05-18 Martin Liska <mliska@suse.cz>
4322 PR gcov-profile/84846
4323 * coverage.c (coverage_init): Write PWD to .gcno file.
4324 * doc/gcov.texi: Document how working directory is printed.
4325 * gcov-dump.c (dump_gcov_file): Print PWD.
4326 * gcov.c (output_intermediate_file): Likewise.
4327 (read_graph_file): Read PWD string.
4328 (output_lines): Print PWD.
4330 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
4333 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
4334 for retval and return false if all args to phi are zero.
4336 2018-05-18 Richard Biener <rguenther@suse.de>
4338 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
4340 (evrp_dom_walker::before_dom_children): Call it.
4342 2018-05-18 Richard Biener <rguenther@suse.de>
4344 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
4345 results when processing array refs with variable index.
4347 2018-05-18 Toon Moene <toon@moene.org>
4349 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
4350 directly after that of -floop-interchange. Indicate that both
4351 options are enabled by default when specifying -O3.
4353 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4355 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
4356 iterator. Delete separate integer-mode vec_set<mode> expander.
4357 (aarch64_simd_vec_setv2di): Delete.
4358 (vec_setv2di): Delete.
4359 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
4360 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
4361 the "w, r" alternative.
4363 2018-05-18 Martin Liska <mliska@suse.cz>
4365 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
4366 * tree-pass.h (make_pass_lower_switch_O0): New function.
4367 * tree-switch-conversion.c (node_has_low_bound): Remove.
4368 (node_has_high_bound): Likewise.
4369 (node_is_bounded): Likewise.
4370 (class pass_lower_switch): Make it a template type and create
4372 (pass_lower_switch::execute): Add template argument.
4373 (make_pass_lower_switch): New function.
4374 (make_pass_lower_switch_O0): New function.
4375 (do_jump_if_equal): Remove.
4376 (emit_case_nodes): Simplify to just handle all 3 cases and leave
4377 all the hard work to tree optimization passes.
4379 2018-05-18 Martin Liska <mliska@suse.cz>
4381 * dbgcnt.c (limit_low): Renamed from limit.
4382 (limit_high): New variable.
4383 (dbg_cnt_is_enabled): Check for upper limit.
4384 (dbg_cnt): Adjust dumping.
4385 (dbg_cnt_set_limit_by_index): Add new argument for high
4387 (dbg_cnt_set_limit_by_name): Likewise.
4388 (dbg_cnt_process_single_pair): Parse new format.
4389 (dbg_cnt_process_opt): Use strtok.
4390 (dbg_cnt_list_all_counters): Remove 'value' and add
4392 * doc/invoke.texi: Document changes.
4394 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
4396 * doc/sourcebuild.texi (scalar_all_fma): Document.
4397 * tree.def (FMA_EXPR): Delete.
4398 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
4399 * internal-fn.c (ternary_direct): New macro.
4400 (expand_ternary_optab_fn): Likewise.
4401 (direct_ternary_optab_supported_p): Likewise.
4402 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
4403 * builtins.c (fold_builtin_fma): Delete.
4404 (fold_builtin_3): Don't call it.
4405 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
4406 * expr.c (expand_expr_real_2): Likewise.
4407 * fold-const.c (operand_equal_p): Likewise.
4408 (fold_ternary_loc): Likewise.
4409 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
4410 * gimple.c (DEFTREECODE): Likewise.
4411 * gimplify.c (gimplify_expr): Likewise.
4412 * optabs-tree.c (optab_for_tree_code): Likewise.
4413 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
4414 * tree-eh.c (operation_could_trap_p): Likewise.
4415 (stmt_could_throw_1_p): Likewise.
4416 * tree-inline.c (estimate_operator_cost): Likewise.
4417 * tree-pretty-print.c (dump_generic_node): Likewise.
4418 (op_code_prio): Likewise.
4419 * tree-ssa-loop-im.c (stmt_cost): Likewise.
4420 * tree-ssa-operands.c (get_expr_operands): Likewise.
4421 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
4422 * fold-const-call.h (fold_fma): Delete.
4423 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
4424 CFN_FNMA and CFN_FNMS.
4426 * genmatch.c (combined_fn): New enum.
4427 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
4428 (commutative_op): New function.
4429 (commutate): Use it. Handle more than 2 operands.
4430 (dt_operand::gen_gimple_expr): Use commutative_op.
4431 (parser::parse_expr): Allow :c to be used with non-binary
4432 operators if the commutative operand is known.
4433 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
4434 CFN_FMS, CFN_FNMA and CFN_FNMS.
4435 (backprop::process_assign_use): Remove FMA_EXPR handling.
4436 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
4437 (gen_hsa_fma): New function.
4438 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
4439 IFN_FNMA and IFN_FNMS.
4440 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
4441 * gimple-fold.h (follow_all_ssa_edges): Declare.
4442 * gimple-fold.c (follow_all_ssa_edges): New function.
4443 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
4444 gimple_build interface and use follow_all_ssa_edges to fold the result.
4445 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
4446 instead of checking for optabs directly.
4447 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
4448 rather than FMA_EXPRs.
4449 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
4450 call to IFN_FMA instead of an FMA_EXPR.
4452 2018-05-17 Jim Wilson <jimw@sifive.com>
4454 * expr.c (do_tablejump): When converting index to Pmode, if we have a
4455 sign extended promoted subreg, and the range does not have the sign bit
4456 set, then do a sign extend.
4458 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
4459 test, check for sign extended subreg and/or constant operands, and
4460 do a sign extend in that case.
4462 2018-05-17 Steve Ellcey <sellcey@cavium.com>
4464 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
4465 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
4467 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
4468 Change logics_shift_reg to logics_shift_imm.
4469 (thunderx2t99_fp_loadpair_basic): Delete.
4470 (thunderx2t99_fp_storepair_basic): Delete.
4471 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
4472 (thunderx2t99_asimd_polynomial): Delete.
4473 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
4474 and neon_fp_mul_d_scalar_q.
4475 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
4476 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
4477 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
4478 (thunderx2t99_asimd_lut): Add missing tbl types.
4479 (thunderx2t99_asimd_ext): Delete.
4480 (thunderx2t99_asimd_load1_1_mult): Delete.
4481 (thunderx2t99_asimd_load1_2_mult): Delete.
4482 (thunderx2t99_asimd_load1_ldp): New.
4483 (thunderx2t99_asimd_load1): New.
4484 (thunderx2t99_asimd_load2): Add missing *load2* types.
4485 (thunderx2t99_asimd_load3): New.
4486 (thunderx2t99_asimd_load4): New.
4487 (thunderx2t99_asimd_store1_1_mult): Delete.
4488 (thunderx2t99_asimd_store1_2_mult): Delete.
4489 (thunderx2t99_asimd_store2_mult): Delete.
4490 (thunderx2t99_asimd_store2_onelane): Delete.
4491 (thunderx2t99_asimd_store_stp): New.
4492 (thunderx2t99_asimd_store1): New.
4493 (thunderx2t99_asimd_store2): New.
4494 (thunderx2t99_asimd_store3): New.
4495 (thunderx2t99_asimd_store4): New.
4497 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
4499 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
4500 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
4502 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
4503 Segher Boessenkool <segher@kernel.crashing.org>
4506 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
4509 2018-05-17 Richard Biener <rguenther@suse.de>
4511 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
4512 for pruning loop and prune defs feeding only already visited PHIs.
4514 2018-05-17 Richard Biener <rguenther@suse.de>
4516 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
4518 2018-05-17 Bin Cheng <bin.cheng@arm.com>
4519 Richard Biener <rguenther@suse.de>
4521 PR tree-optimization/85793
4522 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
4523 for VMAT_ELEMENTWISE.
4525 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
4527 * internal-fn.h (lookup_internal_fn): Declare
4528 * internal-fn.c (lookup_internal_fn): New function.
4529 * gimple.c (gimple_build_call_from_tree): Handle calls to
4531 * gimple-pretty-print.c (dump_gimple_call): Print "." before
4532 internal function names.
4533 * tree-pretty-print.c (dump_generic_node): Likewise.
4534 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
4536 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
4538 * gimple-fold.h (gimple_build): Make the function forms take
4539 combined_fn rather than built_in_function.
4540 (gimple_simplify): Likewise.
4541 * gimple-match-head.c (gimple_simplify): Likewise.
4542 * gimple-fold.c (gimple_build): Likewise.
4543 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
4544 rather than gimple_build_call_internal.
4545 (get_initial_defs_for_reduction): Likewise.
4546 (vect_create_epilog_for_reduction): Likewise.
4547 (vectorizable_live_operation): Likewise.
4549 2018-05-17 Martin Liska <mliska@suse.cz>
4551 * gimple-ssa-sprintf.c (format_directive): Do not use
4552 space in between 'G_' and '('.
4554 2018-05-17 Jakub Jelinek <jakub@redhat.com>
4557 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
4558 even if the mask is not all ones.
4561 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
4563 (ix86_gimple_fold_builtin): Likewise.
4566 * config/i386/i386.c: Include tree-vector-builder.h.
4567 (ix86_vector_shift_count): New function.
4568 (ix86_fold_builtin): Fold shift builtins by scalar count.
4569 (ix86_gimple_fold_builtin): Likewise.
4571 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
4572 _mm512_setzero): New intrinsics.
4574 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
4575 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4577 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
4578 code generation for cases where splatting a value is not useful.
4579 * simplify-rtx.c (simplify_ternary_operation): Simplify
4580 vec_merge across a vec_duplicate and a paradoxical subreg forming
4581 a vector mode to a vec_concat.
4583 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
4585 * config.gcc: Support "goldmont-plus".
4586 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
4588 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
4589 PROCESSOR_GOLDMONT_PLUS.
4590 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
4591 (processor_target_table): Add "goldmont-plus".
4592 (PTA_GOLDMONT_PLUS): Define.
4593 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
4594 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
4595 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
4596 (fold_builtin_cpu): Add "goldmont-plus".
4597 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
4598 (ix86_option_override_internal): Add "goldmont-plus".
4599 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
4600 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
4601 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
4602 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
4604 2018-05-17 Richard Biener <rguenther@suse.de>
4606 PR tree-optimization/85757
4607 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
4608 remove defs that only feed that PHI from further processing.
4610 2018-05-16 Jim Wilson <jimw@sifive.com>
4612 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
4614 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
4616 2018-05-16 Mark Wielaard <mark@klomp.org>
4618 * dwarf2out.c (count_index_strings): New function.
4619 (output_indirect_strings): Call count_index_strings and generate
4620 header for dwarf_version >= 5.
4622 2018-05-16 Mark Wielaard <mark@klomp.org>
4624 * dwarf2out.c (dwarf_FORM): New function.
4625 (set_indirect_string): Use dwarf_FORM.
4626 (reset_indirect_string): Likewise.
4627 (size_of_die): Likewise.
4628 (value_format): Likewise.
4629 (output_die): Likewise.
4630 (add_skeleton_AT_string): Likewise.
4631 (output_macinfo_op): Likewise.
4632 (index_string): Likewise.
4633 (output_index_string_offset): Likewise.
4634 (output_index_string): Likewise.
4635 (count_index_strings): Likewise.
4637 2018-05-16 Carl Love <cel@us.ibm.com>
4639 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
4640 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
4642 2018-05-16 Martin Jambor <mjambor@suse.cz>
4644 * ipa-prop.c (ipa_free_all_edge_args): Remove.
4645 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
4647 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
4649 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
4650 (fnma<mode>4): Likewise.
4651 (fms<mode>4): Likewise.
4652 (fnms<mode>4): Likewise.
4653 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
4654 (aarch64_fnma<mode>4): Likewise.
4655 (aarch64_fms<mode>4): Likewise.
4656 (aarch64_fnms<mode>4): Likewise.
4657 (aarch64_fnmadd<mode>4): Likewise.
4659 2018-05-16 Jason Merrill <jason@redhat.com>
4661 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
4663 2018-05-16 Richard Biener <rguenther@suse.de>
4665 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
4666 (dump_stmt_cost): Declare.
4667 (add_stmt_cost): Dump cost we add.
4668 (add_stmt_costs): New function.
4669 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
4671 (vect_analyze_stmt): Adjust prototype.
4672 (vectorizable_condition): Likewise.
4673 (vectorizable_live_operation): Likewise.
4674 (vectorizable_reduction): Likewise.
4675 (vectorizable_induction): Likewise.
4676 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
4677 cost vector to pass to vectorizable_ and record afterwards.
4678 (vect_model_reduction_cost): Take cost vector argument and adjust.
4679 (vect_model_induction_cost): Likewise.
4680 (vectorizable_reduction): Likewise.
4681 (vectorizable_induction): Likewise.
4682 (vectorizable_live_operation): Likewise.
4683 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
4684 SLP_TREE_NUMBER_OF_VEC_STMTS.
4685 (vect_analyze_slp_cost_1): Remove.
4686 (vect_analyze_slp_cost): Likewise.
4687 (vect_slp_analyze_node_operations): Take visited args and
4688 a target cost vector. Avoid processing already visited stmt sets.
4689 (vect_slp_analyze_operations): Use a local cost vector to gather
4690 costs and register those of non-discarded instances.
4691 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
4692 (vect_schedule_slp_instance): Remove copying of
4693 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
4695 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
4696 adding cost. Record cost entry location.
4697 (vect_prologue_cost_for_slp_op): Function to compute cost of
4698 a constant or invariant generated for SLP vect in the prologue,
4699 split out from vect_analyze_slp_cost_1.
4700 (vect_model_simple_cost): Make static. Adjust for SLP costing.
4701 (vect_model_promotion_demotion_cost): Likewise.
4702 (vect_model_store_cost): Likewise, make static.
4703 (vect_model_load_cost): Likewise.
4704 (vectorizable_bswap): Add cost vector arg and adjust.
4705 (vectorizable_call): Likewise.
4706 (vectorizable_simd_clone_call): Likewise.
4707 (vectorizable_conversion): Likewise.
4708 (vectorizable_assignment): Likewise.
4709 (vectorizable_shift): Likewise.
4710 (vectorizable_operation): Likewise.
4711 (vectorizable_store): Likewise.
4712 (vectorizable_load): Likewise.
4713 (vectorizable_condition): Likewise.
4714 (vectorizable_comparison): Likewise.
4715 (can_vectorize_live_stmts): Likewise.
4716 (vect_analyze_stmt): Likewise.
4717 (vect_transform_stmt): Adjust calls to vectorizable_*.
4718 * tree-vectorizer.c: Include gimple-pretty-print.h.
4719 (dump_stmt_cost): New function.
4721 2018-05-16 Richard Biener <rguenther@suse.de>
4723 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
4724 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
4725 * tree-ssa-dse.c: Include tree-ssa-loop.h.
4726 (check_name): New callback.
4727 (dse_classify_store): Track cycles via a visited bitmap of PHI
4728 defs and simplify handling of in-loop and across loop dead stores
4729 and properly fail for loop-variant refs. Handle byte-tracking with
4730 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
4733 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
4735 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
4736 (vect_get_mask_type_for_stmt): Likewise.
4737 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
4739 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
4740 to determine the statement's vector type and the vector type that
4741 should be used for calculating nunits. Deal with cases in which
4742 the type has to be deferred.
4743 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
4744 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
4745 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
4746 (vect_determine_vf_for_stmt): New functions, split out from...
4747 (vect_determine_vectorization_factor): ...here.
4748 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
4749 (vect_get_mask_type_for_stmt): New functions, split out from
4750 vect_determine_vectorization_factor.
4752 2018-05-16 Richard Biener <rguenther@suse.de>
4754 * tree-cfg.c (verify_gimple_assign_ternary): Properly
4755 verify the [VEC_]COND_EXPR embedded comparison.
4757 2018-05-15 Martin Sebor <msebor@redhat.com>
4759 PR tree-optimization/85753
4760 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
4761 RECORD_TYPE in addition to ARRAY_TYPE.
4763 2018-05-15 Martin Sebor <msebor@redhat.com>
4766 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
4768 2018-05-15 Richard Biener <rguenther@suse.de>
4770 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
4771 add by_clobber_p one. Change algorithm to collect all defs
4772 representing uses we need to walk and try reducing them to
4773 a single one before failing.
4774 (dse_dom_walker::dse_optimize_stmt): Adjust.
4776 2018-05-13 Mark Wielaard <mark@klomp.org>
4778 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
4779 (size_of_loc_descr): Likewise.
4780 (output_loc_operands): Likewise.
4781 (output_loc_operands_raw): Likewise.
4782 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
4783 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
4784 (hash_loc_operands): Likewise.
4785 (compare_loc_operands): Likewise.
4787 2018-05-14 Mark Wielaard <mark@klomp.org>
4789 * dwarf2out.c (count_index_addrs): New function.
4790 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
4792 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
4794 PR tree-optimization/83648
4795 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
4796 return value as malloc candidate.
4798 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
4801 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
4802 param as true in call to suggest_attribute.
4804 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
4806 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
4807 -mreadonly-in-sdata.
4809 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4811 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
4813 (aarch64_crypto_aesd_fused): Likewise.
4815 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
4817 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
4818 (movsi_aarch64): Likewise.
4819 (load_pairsi): Likewise.
4820 (load_pairdi): Likewise.
4821 (store_pairsi): Likewise.
4822 (store_pairdi): Likewise.
4823 (load_pairsf): Likewise.
4824 (load_pairdf): Likewise.
4825 (store_pairsf): Likewise.
4826 (store_pairdf): Likewise.
4827 (zero_extend): Likewise.
4828 (trunc): Swap alternatives.
4829 (fcvt_target): Add '?' to prefer w over r.
4831 2018-05-14 Jakub Jelinek <jakub@redhat.com>
4834 * config/i386/i386.md: Disallow non-commutative arithmetics in
4835 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
4836 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
4837 in the peephole2 before it.
4839 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
4841 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
4842 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
4843 (ix86_handle_option): Handle -mcldemote.
4844 * config.gcc: New header.
4845 * config/i386/cldemoteintrin.h: New file.
4846 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
4847 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
4849 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
4850 OPTION_MASK_ISA_CLDEMOTE.
4851 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
4852 (ix86_valid_target_attribute_inner_p): Ditto.
4853 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
4854 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
4855 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
4856 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
4857 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
4859 * config/i386/i386.opt: Add -mcldemote.
4860 * config/i386/x86intrin.h: New header.
4861 * doc/invoke.texi: Add -mcldemote.
4863 2018-05-14 Richard Biener <rguenther@suse.de>
4865 * doc/match-and-simplify.texi: Adjust :s documentation.
4867 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
4869 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
4870 intended memcpy size.
4871 (REORDER_45): Likewise.
4873 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
4875 * sort.cc: New file.
4876 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
4877 * vec.c (qsort_chk): Use gcc_qsort.
4878 * Makefile.in (OBJS-libcommon): Add sort.o.
4879 (build/sort.o): New target. Use it...
4880 (BUILD_RTL): ... here, and...
4881 (build/gencfn-macros): ... here, and...
4882 (build/genmatch): ... here.
4884 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
4885 Chung-Ju Wu <jasonwucj@gmail.com>
4887 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
4888 * config/nds32/nds32-graywolf.md: New file.
4889 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
4890 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
4892 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
4893 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
4894 * config/nds32/nds32.md (pipeline_model): Add graywolf.
4895 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
4896 * config/nds32/pipelines.md: Include n15 settings.
4898 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
4899 Chung-Ju Wu <jasonwucj@gmail.com>
4901 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
4902 * config/nds32/nds32-n13.md: New file.
4903 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
4904 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
4906 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
4907 * config/nds32/nds32.md (pipeline_model): Add n13.
4908 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
4909 * config/nds32/pipelines.md: Include n13 settings.
4911 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
4912 Chung-Ju Wu <jasonwucj@gmail.com>
4914 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
4915 * config/nds32/nds32-n10.md: New file.
4916 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
4917 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
4919 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
4920 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
4921 * config/nds32/nds32.md (pipeline_model): Add n10.
4922 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
4923 * config/nds32/pipelines.md: Include n10 settings.
4925 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
4926 Kito Cheng <kito.cheng@gmail.com>
4927 Chung-Ju Wu <jasonwucj@gmail.com>
4929 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
4930 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
4931 Add enum values for DSP extension instructions.
4932 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
4934 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
4935 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
4937 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
4938 * config/nds32/nds32-dspext.md: New file for DSP implementation.
4939 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
4940 * config/nds32/nds32-intrinsic.md: Likewise.
4941 * config/nds32/nds32_intrinsic.h: Likewise.
4942 * config/nds32/nds32-md-auxiliary.c: Likewise.
4943 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
4944 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
4945 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
4946 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
4947 * config/nds32/nds32-protos.h: New declarations for DSP extension.
4948 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
4949 TYPE_DMAC in switch statement.
4950 * config/nds32/nds32.c: New checking and implementation for DSP
4951 extension instructions.
4952 * config/nds32/nds32.h: Likewise.
4953 * config/nds32/nds32.md: Likewise.
4954 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
4955 * config/nds32/predicates.md: Implement new predicates for DSP
4958 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
4960 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
4961 Reformat alternatives and attributes so it is easier to identify
4962 which constraints/attributes go with which instruction.
4963 (mov<mode>_hardfloat32, FMOVE64): Likewise.
4964 (mov<mode>_softfloat32, FMOVE64): Likewise.
4965 (mov<mode>_hardfloat64, FMOVE64): Likewise.
4966 (mov<mode>_softfloat64, FMOVE64): Likewise.
4968 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
4970 * doc/extend.texi (PowerPC Built-in Functions): Rename this
4972 (Basic PowerPC Built-in Functions): The new name of the
4973 subsection previously known as "PowerPC Built-in Functions".
4974 (Basic PowerPC Built-in Functions Available on all Configurations):
4976 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
4977 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
4978 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
4979 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
4981 2018-05-11 Martin Jambor <mjambor@suse.cz>
4984 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
4987 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
4990 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
4992 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
4994 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
4995 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
4996 (ix86_handle_option): Handle -mwaitpkg.
4997 * config.gcc: New header.
4998 * config/i386/cpuid.h (bit_WAITPKG): New bit.
4999 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
5000 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
5002 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
5003 OPTION_MASK_ISA_WAITPKG.
5004 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
5005 (ix86_option_override_internal): Add PTA_WAITPKG.
5006 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
5007 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
5008 IX86_BUILTIN_TPAUSE.
5009 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
5010 __builtin_ia32_umwait and __builtin_ia32_tpause.
5011 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
5012 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
5013 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
5014 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
5015 UNSPECV_TPAUSE): New.
5016 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
5017 * config/i386/i386.opt: Add -mwaitpkg.
5018 * config/i386/waitpkgintrin.h: New file.
5019 * config/i386/x86intrin.h: New header.
5020 * doc/invoke.texi: Add -mwaitpkg.
5022 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
5025 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
5027 (cortex-m0): Use armv6s-m isa.
5028 (cortex-m0plus): Likewise.
5029 (cortex-m1): Likewise.
5030 (cortex-m0.small-multiply): Likewise.
5031 (cortex-m0plus.small-multiply): Likewise.
5032 (cortex-m1.small-multiply): Likewise.
5034 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
5035 Jakub Jelinek <jakub@redhat.com>
5037 PR tree-optimization/85692
5038 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
5039 source permute as well.
5041 2018-05-11 Martin Liska <mliska@suse.cz>
5044 * doc/extend.texi: Document LLVM style format for no_sanitize
5047 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
5049 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
5050 mode_supports_vsx_dform_quad to mode_supports_dq_form.
5051 (mode_supports_vsx_dform_quad): Likewise.
5052 (mode_supports_vmx_dform): Move these functions to be next to the
5053 other mode_supports functions.
5054 (mode_supports_dq_form): Likewise.
5055 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
5056 mode_supports_dq_form.
5057 (reg_offset_addressing_ok_p): Likewise.
5058 (offsettable_ok_by_alignment): Likewise.
5059 (rs6000_legitimate_offset_address_p): Likewise.
5060 (legitimate_lo_sum_address_p): Likewise.
5061 (rs6000_legitimize_address): Likewise.
5062 (rs6000_legitimize_reload_address): Likewise.
5063 (rs6000_secondary_reload_inner): Likewise.
5064 (rs6000_preferred_reload_class): Likewise.
5065 (rs6000_output_move_128bit): Likewise.
5067 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
5069 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
5070 Generate SImode target register for null target.
5071 <case IX86_BUILTIN_XGETBV>: Ditto.
5072 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
5073 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
5075 2018-05-10 Carl Love <cel@us.ibm.com>
5077 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
5078 dcbtt and dcbtstt if operands[2] is 0.
5080 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
5083 * config/i386/sse.md (usadv64qi): New expander.
5085 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
5087 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
5088 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
5089 -maltivec=be support.
5090 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
5091 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
5092 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
5093 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
5094 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
5095 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
5096 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
5097 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
5098 altivec_vsumsws): Adjust.
5099 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
5100 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
5101 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
5103 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
5104 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
5105 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
5106 (altivec_lve<VI_char>x): Delete expand.
5107 (*altivec_lve<VI_char>x_internal): Rename to...
5108 (altivec_lve<VI_char>x): ... this.
5109 (altivec_lvxl_<mode>): Delete expand.
5110 (*altivec_lvxl_<mode>_internal): Rename to ...
5111 (altivec_lvxl_<mode>): ... this.
5112 (altivec_stvxl_<mode>): Delete expand.
5113 (*altivec_stvxl_<mode>_internal): Rename to ...
5114 (altivec_stvxl_<mode>): ... this.
5115 (altivec_stve<VI_char>x): Delete expand.
5116 (*altivec_stve<VI_char>x_internal): Rename to ...
5117 (altivec_stve<VI_char>x): ... this.
5118 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
5119 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
5120 reduc_plus_scal_<mode>): Adjust.
5121 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
5123 (rs6000_cpu_cpp_builtins): Adjust.
5124 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
5125 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
5126 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
5127 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
5128 -maltivec=be support.
5129 (rs6000_split_vec_extract_var): Adjust.
5130 (rs6000_split_v4si_init): Adjust.
5131 (swap_selector_for_mode): Delete.
5132 (altivec_expand_lvx_be, altivec_expand_stvx_be,
5133 altivec_expand_stvex_be): Delete.
5134 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
5135 -maltivec=be support.
5136 (rs6000_gimple_fold_builtin): Ditto.
5137 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
5139 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
5140 (TARGET_DIRECT_MOVE_64BIT): Adjust.
5141 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
5142 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
5143 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
5144 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
5145 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
5146 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
5147 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
5148 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
5149 anonymous split): Adjust.
5150 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
5151 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
5153 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
5155 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
5156 when --with-gxx-include-dir is also specified.
5157 * configure: Regenerate.
5159 2018-05-09 Jim Wilson <jimw@sifive.com>
5162 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
5163 * config/riscv/t-withmultilib: New.
5164 * config/riscv/withmultilib.h: New.
5165 * doc/install.texi: Document RISC-V --with-multilib-list support.
5167 2018-05-09 Richard Biener <rguenther@suse.de>
5169 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
5171 (vect_bb_vectorization_profitable_p): Adjust. Compute
5172 actual scalar cost using the cost vector and the add_stmt_cost
5175 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
5177 PR rtl-optimization/85645
5178 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
5179 in the REG_CFA_REGISTER note for LR, don't leave it empty.
5181 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
5183 PR rtl-optimization/85645
5184 * shrink-wrap.c (spread_components): Return a boolean saying if
5185 anything was changed.
5186 (try_shrink_wrapping_separate): Iterate spread_components until
5187 nothing changes anymore.
5189 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
5191 PR rtl-optimization/85645
5192 * regrename.c (build_def_use): Also kill the chains that include the
5193 destination of a REG_CFA_REGISTER note.
5195 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
5197 PR rtl-optimization/85645
5198 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
5199 insn that has a REG_CFA_REGISTER note.
5201 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
5203 * cfgexpand.c (expand_clobber): New function.
5204 (expand_gimple_stmt_1): Use it.
5205 * tree-vect-stmts.c (vect_clobber_variable): New function,
5207 (vectorizable_simd_clone_call): ...here.
5208 (vectorizable_store): Emit a clobber either side of an
5209 IFN_STORE_LANES sequence.
5210 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
5212 2018-05-09 Tom de Vries <tom@codesourcery.com>
5215 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
5216 (define_insn "trap_if_false"): Add exit after trap.
5218 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
5220 PR rtl-optimization/85638
5221 * bb-reorder.c: Include common/common-target.h.
5222 (create_forwarder_block): New function extracted from...
5223 (fix_up_crossing_landing_pad): ...here. Rename into...
5224 (dw2_fix_up_crossing_landing_pad): ...this.
5225 (sjlj_fix_up_crossing_landing_pad): New function.
5226 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
5227 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
5228 from both partitions and exit the loop after one iteration.
5230 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
5233 * doc/extend.texi (PowerPC Built-in Functions): Rename this
5235 (Basic PowerPC Built-in Functions): The new name of the
5236 subsection previously known as "PowerPC Built-in Functions".
5237 (Basic PowerPC Built-in Functions Available on all Configurations):
5239 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
5241 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
5243 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
5245 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
5248 2018-05-08 Jim Wilson <jimw@sifive.com>
5250 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
5251 (LD_EMUL_SUFFIX): New.
5252 (LINK_SPEC): Use it.
5254 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
5256 * doc/extend.texi (PowerPC Built-in Functions): Rename this
5258 (Basic PowerPC Built-in Functions): The new name of the
5259 subsection previously known as "PowerPC Built-in Functions".
5260 (Basic PowerPC Built-in Functions Available on all Configurations):
5262 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
5264 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
5266 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
5268 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
5271 2018-05-08 Jakub Jelinek <jakub@redhat.com>
5274 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
5275 after cmpelim optimization.
5277 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
5279 * config.gcc: Support "goldmont".
5280 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
5281 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
5283 * config/i386/i386.c (m_GOLDMONT): Define.
5284 (processor_target_table): Add "goldmont".
5285 (PTA_GOLDMONT): Define.
5286 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
5287 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
5288 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
5289 (fold_builtin_cpu): Add "goldmont".
5290 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
5291 (ix86_option_override_internal): Add "goldmont".
5292 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
5293 (processor_type): Add PROCESSOR_GOLDMONT.
5294 * config/i386/i386.md: Add CPU "glm".
5295 * config/i386/glm.md: New file.
5296 * config/i386/x86-tune.def: Add m_GOLDMONT.
5297 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
5299 2018-05-08 Jakub Jelinek <jakub@redhat.com>
5302 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
5304 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
5305 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
5306 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
5309 * config/i386/i386.c (ix86_fold_builtin): Handle
5310 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
5313 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
5314 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
5316 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
5319 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
5320 (check_arch): Likewise.
5321 (check_fpu): Return the result rather than printing it.
5322 (end arch): Fix operator precedence.
5323 (end cpu): Likewise.
5324 (END): Print the result from check_fpu.
5326 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
5327 Alan Hayward <alan.hayward@arm.com>
5328 David Sherwood <david.sherwood@arm.com>
5330 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
5331 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
5332 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
5333 (*fcmuo<mode>_and): New patterns.
5335 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
5337 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
5338 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
5339 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
5340 (cmp_op, sve_imm_con): New code attributes.
5341 (SVE_COND_INT_CMP, imm_con): Delete.
5342 (cmp_op): Remove above unspecs from int attribute.
5343 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
5345 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
5346 comparison-specific unspecs.
5347 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
5348 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
5349 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
5350 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
5351 (*vec_fcm<cmp_op><mode>): Rename to...
5352 (*fcm<cmp_op><mode>): ...this and adjust likewise.
5353 (*vec_fcmuo<mode>): Rename to...
5354 (*fcmuo<mode>): ...this and adjust likewise.
5355 (*pred_fcm<cmp_op><mode>): New pattern.
5356 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
5357 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
5359 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
5361 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
5362 (aarch64_emit_sve_predicated_cond): New function.
5363 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
5364 (aarch64_emit_unspec_cond_or): Replace with...
5365 (aarch64_emit_sve_or_conds): ...this new function. Use
5366 aarch64_emit_sve_ptrue_op for the individual comparisons and
5367 aarch64_emit_binop to OR them together.
5368 (aarch64_emit_inverted_unspec_cond): Replace with...
5369 (aarch64_emit_sve_inverted_cond): ...this new function. Use
5370 aarch64_emit_sve_ptrue_op for the comparison and
5371 aarch64_emit_unop to invert the result.
5372 (aarch64_expand_sve_vec_cmp_float): Update after the above
5373 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
5375 2018-05-07 Nathan Sidwell <nathan@acm.org>
5377 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
5378 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
5379 (Backwards Compatibility): Likewise.
5381 2018-05-07 Luis Machado <luis.machado@linaro.org>
5385 2018-05-07 Luis Machado <luis.machado@linaro.org>
5387 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5388 <prefetch_dynamic_strides>: New const bool field.
5389 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5390 prefetch_dynamic_strides.
5391 (exynosm1_prefetch_tune): Likewise.
5392 (thunderxt88_prefetch_tune): Likewise.
5393 (thunderx_prefetch_tune): Likewise.
5394 (thunderx2t99_prefetch_tune): Likewise.
5395 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
5397 (aarch64_override_options_internal): Update to set
5398 PARAM_PREFETCH_DYNAMIC_STRIDES.
5399 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
5400 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
5401 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
5402 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
5403 prefetch-dynamic-strides setting.
5405 2018-05-07 Luis Machado <luis.machado@linaro.org>
5407 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5408 <minimum_stride>: New const int field.
5409 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5410 minimum_stride field.
5411 (exynosm1_prefetch_tune): Likewise.
5412 (thunderxt88_prefetch_tune): Likewise.
5413 (thunderx_prefetch_tune): Likewise.
5414 (thunderx2t99_prefetch_tune): Likewise.
5415 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
5416 (aarch64_override_options_internal): Update to set
5417 PARAM_PREFETCH_MINIMUM_STRIDE.
5418 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
5419 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
5420 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
5421 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
5422 stride is constant and is below the minimum stride threshold.
5424 2018-05-07 Luis Machado <luis.machado@linaro.org>
5426 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
5429 2018-05-07 Luis Machado <luis.machado@linaro.org>
5431 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5432 <prefetch_dynamic_strides>: New const bool field.
5433 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5434 prefetch_dynamic_strides.
5435 (exynosm1_prefetch_tune): Likewise.
5436 (thunderxt88_prefetch_tune): Likewise.
5437 (thunderx_prefetch_tune): Likewise.
5438 (thunderx2t99_prefetch_tune): Likewise.
5439 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
5441 (aarch64_override_options_internal): Update to set
5442 PARAM_PREFETCH_DYNAMIC_STRIDES.
5443 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
5444 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
5445 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
5446 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
5447 prefetch-dynamic-strides setting.
5449 2018-05-07 Luis Machado <luis.machado@linaro.org>
5451 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5452 <minimum_stride>: New const int field.
5453 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5454 minimum_stride field.
5455 (exynosm1_prefetch_tune): Likewise.
5456 (thunderxt88_prefetch_tune): Likewise.
5457 (thunderx_prefetch_tune): Likewise.
5458 (thunderx2t99_prefetch_tune): Likewise.
5459 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
5460 (aarch64_override_options_internal): Update to set
5461 PARAM_PREFETCH_MINIMUM_STRIDE.
5462 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
5463 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
5464 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
5465 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
5466 stride is constant and is below the minimum stride threshold.
5468 2018-05-06 Jakub Jelinek <jakub@redhat.com>
5471 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
5472 the type is addressable. Don't force op into register if it has
5475 2018-05-05 Roland McGrath <mcgrathr@google.com>
5478 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
5479 any section for which we don't know a specific type it should have,
5480 regardless of name. Previously this was done only for the exact
5481 names ".init_array", ".fini_array", and ".preinit_array".
5482 (default_elf_asm_named_section): Add comment about
5483 relationship with default_section_type_flags and SECTION_NOTYPE.
5484 (get_section): Don't consider it a type conflict if one side has
5485 SECTION_NOTYPE and the other doesn't, as long as neither has the
5486 SECTION_BSS et al used in the default_section_type_flags logic.
5488 2018-05-05 Tom de Vries <tom@codesourcery.com>
5491 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
5492 (workaround_barsyncs): New function.
5493 (nvptx_reorg): Use workaround_barsyncs.
5494 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
5495 (define_expand "nvptx_membar_cta"): New define_expand.
5496 (define_insn "*nvptx_membar_cta"): New insn.
5498 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
5500 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
5501 To improve optimization opportunities.
5502 * builtin-types.def: The new needed builtin types for the above.
5504 2018-05-04 Richard Biener <rguenther@suse.de>
5506 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
5507 * gimple-ssa-store-merging.c
5508 (imm_store_chain_info::output_merged_store): Remove redundant create,
5509 release split_store vector contents on failure.
5510 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
5511 scalar stmt vector on cache hit.
5513 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
5515 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
5517 * config.gcc (powerpc-xilinx-eabi*): Remove.
5518 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
5520 (fusion_addis_mem_combo_load): Ditto.
5521 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
5523 (rs6000_cpu_cpp_builtins): Ditto.
5524 * config/rs6000/rs6000-linux.c
5525 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
5526 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
5527 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
5529 (rs6000_setup_reg_addr_masks): Ditto.
5530 (rs6000_init_hard_regno_mode_ok): Ditto.
5531 (rs6000_option_override_internal): Ditto.
5532 (legitimate_lo_sum_address_p): Ditto.
5533 (rs6000_legitimize_address): Ditto.
5534 (rs6000_legitimize_reload_address): Ditto.
5535 (rs6000_legitimate_address_p): Ditto.
5536 (abi_v4_pass_in_fpr): Ditto.
5537 (setup_incoming_varargs): Ditto.
5538 (rs6000_gimplify_va_arg): Ditto.
5539 (rs6000_split_multireg_move): Ditto.
5540 (rs6000_savres_strategy): Ditto.
5541 (rs6000_emit_prologue_components): Ditto.
5542 (rs6000_emit_epilogue_components): Ditto.
5543 (rs6000_emit_prologue): Ditto.
5544 (rs6000_emit_epilogue): Ditto.
5545 (rs6000_elf_file_end): Ditto.
5546 (rs6000_function_value): Ditto.
5547 (rs6000_libcall_value): Ditto.
5548 * config/rs6000/rs6000.h: Ditto.
5549 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
5550 (TARGET_MINMAX): ... this. New.
5551 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
5552 * config/rs6000/rs6000.md: Remove Xilinx FP support.
5553 (*movsi_internal1_single): Delete.
5554 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
5555 mfpu=, mxilinx-fpu): Delete.
5556 * config/rs6000/singlefp.h: Delete.
5557 * config/rs6000/sysv4.h: Remove Xilinx FP support.
5558 * config/rs6000/t-rs6000: Ditto.
5559 * config/rs6000/t-xilinx: Delete.
5560 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
5561 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
5562 (VStype_simple): Delete.
5563 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
5564 * config/rs6000/xfpu.h: Delete.
5565 * config/rs6000/xfpu.md: Delete.
5566 * config/rs6000/xilinx.h: Delete.
5567 * config/rs6000/xilinx.opt: Delete.
5568 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
5569 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
5571 2018-05-04 Tom de Vries <tom@codesourcery.com>
5574 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
5577 2018-05-04 Richard Biener <rguenther@suse.de>
5580 * tree-complex.c (update_complex_assignment): We are always in SSA form.
5581 (expand_complex_div_wide): Likewise.
5582 (expand_complex_operations_1): Likewise.
5583 (expand_complex_libcall): Preserve EH info of the original stmt.
5584 (tree_lower_complex): Handle removed blocks.
5585 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
5586 on complex multiplication and division libcall builtins.
5588 2018-05-04 Richard Biener <rguenther@suse.de>
5591 * fold-const.c (negate_expr_p): Restrict negation of operand
5592 zero of a division to when we know that can happen without
5594 (fold_negate_expr_1): Likewise.
5596 2018-05-04 Jakub Jelinek <jakub@redhat.com>
5599 * real.h (real_nextafter): Declare.
5600 * real.c (real_nextafter): New function.
5601 * fold-const-call.c (fold_const_nextafter): New function.
5602 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
5603 CASE_CFN_NEXTTOWARD.
5604 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
5605 even when arg1_mode is different from arg0_mode.
5607 2018-05-03 Nathan Sidwell <nathan@acm.org>
5609 * doc/extend.texi (Deprecated Features): Remove
5611 (Backwards Compatibility): Likewise.
5612 * doc/invoke.texi (C++ Language Options): Likewise.
5613 (C++ Dialect Options): Likewise.
5615 2018-05-03 Jakub Jelinek <jakub@redhat.com>
5618 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
5619 _mm512_mask_mullox_epi64): New intrinsics.
5621 2018-05-03 Tom de Vries <tom@codesourcery.com>
5624 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
5625 dump files): Add offload-tree.
5627 2018-05-03 Richard Biener <rguenther@suse.de>
5629 PR tree-optimization/85615
5630 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
5631 to loops not nested in BBs loop father to avoid creating multi-entry
5634 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5636 PR tree-optimization/70291
5637 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
5638 arguments. Change return type to tree. Emit libcall as a new
5639 statement rather than replacing existing one when inplace_p is true.
5640 (expand_complex_multiplication_components): New function.
5641 (expand_complex_multiplication): Expand floating-point complex
5642 multiplication using the above.
5643 (expand_complex_division): Rename inner_type parameter to type.
5644 Update expand_complex_libcall call-site.
5645 (expand_complex_operations_1): Update expand_complex_multiplication
5646 and expand_complex_division call-sites.
5648 2018-05-02 Jakub Jelinek <jakub@redhat.com>
5651 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
5652 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
5653 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
5654 the highest significant bit of the shift count mask is clear. In
5655 check whether and[sq]i3 is needed verify that all significant bits
5656 of the shift count other than the highest are set.
5658 2018-05-02 Tom de Vries <tom@codesourcery.com>
5661 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
5662 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
5663 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
5664 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
5665 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
5666 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
5667 __builtin_goacc_parlevel_size.
5669 2018-05-02 Richard Biener <rguenther@suse.de>
5671 PR tree-optimization/85597
5672 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
5673 do not use split vect_get_vec_defs call but call vect_get_slp_defs
5676 2018-05-02 Tom de Vries <tom@codesourcery.com>
5679 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
5680 dump files): Add ltrans-tree.
5682 2018-05-02 Tom de Vries <tom@codesourcery.com>
5685 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
5686 dump files): Add wpa-ipa.
5688 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
5690 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
5691 powerpc*-*-linux*paired* target.
5692 * config/rs6000/750cl.h: Delete.
5693 * config/rs6000/paired.h: Delete.
5694 * config/rs6000/paired.md: Delete.
5695 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
5697 * config/rs6000/rs6000-builtin.def: Remove paired float support.
5698 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
5699 comment. Remove paired float support.
5700 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
5701 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
5703 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
5704 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
5706 * config/rs6000/rs6000.c: Remove paired float support.
5707 (paired_expand_vector_init, paired_expand_vector_move,
5708 paired_emit_vector_compare, paired_emit_vector_cond_expr,
5709 (paired_expand_lv_builtin, paired_expand_stv_builtin,
5710 paired_expand_builtin, paired_expand_predicate_builtin,
5711 paired_init_builtins): Delete.
5712 * config/rs6000/rs6000.h: Remove paired float support.
5713 * config/rs6000/rs6000.md: Remove paired float support.
5714 (move_from_CR_ov_bit): Delete.
5715 * config/rs6000/rs6000.opt (mpaired): Delete.
5716 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
5717 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
5719 2018-05-02 Richard Biener <rguenther@suse.de>
5722 * gimplify.c (gimplify_save_expr): When in SSA form allow
5723 SAVE_EXPRs to compute to SSA vars.
5725 2018-05-02 Jakub Jelinek <jakub@redhat.com>
5728 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
5729 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
5730 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
5731 clobber operands[2], instead use a new pseudo. Formatting fixes.
5733 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
5735 PR tree-optimization/85586
5736 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
5737 exit early for statements in the same group if the accesses are
5740 2018-05-02 Tom de Vries <tom@codesourcery.com>
5743 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
5746 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
5748 PR tree-optimization/85143
5749 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
5751 2018-05-01 Tom de Vries <tom@codesourcery.com>
5754 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
5755 not found" error message.
5757 2018-05-01 Tom de Vries <tom@codesourcery.com>
5760 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
5761 * vec.c (test_ordered_remove_if): New function.
5762 (vec_c_tests): Call test_ordered_remove_if.
5763 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
5764 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
5765 * tree-vect-patterns.c (vect_pattern_recog_1): Use
5766 VEC_ORDERED_REMOVE_IF.
5768 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5770 PR tree-optimization/82665
5771 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
5772 pointer subtraction where arguments come from a memchr call.
5774 2018-05-01 Jakub Jelinek <jakub@redhat.com>
5776 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
5777 --push-state --as-needed and --pop-state instead of --as-needed and
5778 --no-as-needed if ld supports it.
5779 * configure: Regenerated.
5782 * doc/install.texi2html: Replace _002d with - and _002a with * in
5783 generated html files using sed.
5785 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5788 * gcc-rich-location.c (blank_line_before_p): New function.
5789 (use_new_line): New function.
5790 (gcc_rich_location::add_fixit_insert_formatted): New function.
5791 * gcc-rich-location.h
5792 (gcc_rich_location::add_fixit_insert_formatted): New function.
5794 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5796 * selftest.c (assert_streq): Rename "expected" and "actual" to
5797 "val1" and "val2". Extend NULL-handling to cover both inputs
5798 symmetrically, while still requiring both to be non-NULL for a pass.
5799 * selftest.h (assert_streq): Rename "expected" and "actual" to
5801 (ASSERT_EQ): Likewise.
5802 (ASSERT_EQ_AT): Likewise.
5803 (ASSERT_KNOWN_EQ): Likewise.
5804 (ASSERT_KNOWN_EQ_AT): Likewise.
5805 (ASSERT_NE): Likewise.
5806 (ASSERT_MAYBE_NE): Likewise.
5807 (ASSERT_MAYBE_NE_AT): Likewise.
5808 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
5809 the assertion to pass.
5810 (ASSERT_STREQ_AT): Likewise.
5812 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
5814 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
5815 interaction with -pie.
5817 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5819 * selftest.h: Fix alphabetization of per-source-file selftest
5822 2018-04-30 Jason Merrill <jason@redhat.com>
5824 PR c++/61982 - dead stores to destroyed objects.
5825 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
5828 2018-04-30 Jason Merrill <jason@redhat.com>
5830 * tree.c (build_clobber): New.
5831 * tree.h: Declare it.
5832 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
5834 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5836 * diagnostic-show-locus.c (layout::layout): Update for
5837 location_get_source_line returning a char_span.
5838 (struct char_span): Move to input.h.
5839 (struct correction): Update for fields in char_span becoming
5841 (struct source_line): Update for location_get_source_line
5842 returning a char_span.
5843 (layout::print_line): Likewise.
5844 * edit-context.c (edited_file::print_content): Likewise.
5845 (edited_file::print_diff_hunk): Likewise.
5846 (edited_file::print_run_of_changed_lines): Likewise.
5847 (edited_file::get_num_lines): Likewise.
5848 (edited_line::edited_line): Likewise.
5849 * final.c (asm_show_source): Likewise.
5850 * input.c (location_get_source_line): Convert return type
5851 from const char * to char_span, losing the final "line_len"
5853 (dump_location_info): Update for the above.
5854 (get_substring_ranges_for_loc): Likewise. Use a char_span
5855 when handling the literal within the line.
5856 (test_reading_source_line): Update for location_get_source_line
5857 returning a char_span.
5858 * input.h (class char_span): Move here from
5859 diagnostic-show-locus.c, converting from a struct to a class.
5860 Make data members private.
5861 (char_span::operator bool): New.
5862 (char_span::length): New.
5863 (char_span::get_buffer): New.
5864 (char_span::operator[]): New.
5865 (char_span::subspan): Make const.
5866 (char_span::xstrdup): New.
5867 (location_get_source_line): Convert return type from const char *
5868 to char_span, losing the final "line_size" param.
5870 2018-04-30 Jan Hubicka <jh@suse.cz>
5872 * lto-wrapper.c (ltrans_priorities): New static var.
5873 (cmp_priority): New.
5874 (run_gcc): Read priorities and if doing parallel build order
5875 the Makefile by them.
5877 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5879 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
5881 2018-04-30 Richard Biener <rguenther@suse.de>
5883 * tree-cfg.c (verify_address): Remove base argument, add
5884 flag whether to check TREE_ADDRESSABLE and do that.
5885 (verify_expr): Remove.
5886 (verify_types_in_gimple_reference): Add pieces from verify_expr.
5887 (verify_gimple_assign_single): Likewise.
5888 (verify_gimple_switch): Likewise.
5889 (verify_expr_location_1): Dereference tp once. Add (disabled)
5890 piece from verify_expr.
5891 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
5893 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
5895 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
5897 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
5899 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
5900 (small_data_pattern): Likewise.
5901 (arc_rewrite_small_data): Likewise.
5902 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
5903 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
5904 (get_symbol_alignment): New function.
5905 (legitimate_small_data_address_p): Likewise.
5906 (legitimate_scaled_address): Update, call
5907 legitimate_small_data_address_p.
5908 (output_sdata): New static variable.
5909 (arc_print_operand): Update how we handle small data operands.
5910 (arc_print_operand_address): Likewise.
5911 (arc_legitimate_address_p): Update, use
5912 legitimate_small_data_address_p.
5913 (arc_rewrite_small_data_p): Remove.
5914 (arc_rewrite_small_data_1): Likewise.
5915 (arc_rewrite_small_data): Likewise.
5916 (small_data_pattern): Likewise.
5917 (compact_sda_memory_operand): Update to use
5918 legitimate_small_data_address_p and get_symbol_alignment.
5919 (prepare_move_operands): Don't rewite sdata pattern.
5920 (prepare_extend_operands): Remove.
5921 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
5923 (zero_extendqisi2): Likewise.
5924 (zero_extendhisi2): Likewise.
5925 (extendqihi2): Likewise.
5926 (extendqisi2): Likewise.
5927 (extendhisi2): Likewise.
5931 * config/arc/constraints.md (Usd): Change it to memory constraint.
5933 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
5935 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
5936 as source of std instructions.
5937 * config/arc/arc.md (movsi_insn): Update pattern predicate to
5938 allow 6-bit constants as source for store instructions.
5939 (movdi_insn): Update instruction pattern to allow 6-bit constants
5940 as source for store instructions.
5942 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
5944 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
5946 2018-04-30 Nathan Sidwell <nathan@acm.org>
5947 Sandra Loosemore <sandra@codesourcery.com>
5949 * dumpfile.c (dump_open): Allow '-' for stdout.
5950 * doc/invoke.texi (Developer Options): Document dump filename
5951 determination early. Document stdin/stdout selection.
5953 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
5955 Microblaze Target: PIC data text relative
5957 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
5958 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
5960 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
5961 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
5962 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
5963 New addressing mode for data-text relative position indepenedent code.
5964 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
5965 'ADDRESS_SYMBOLIC_TXT_REL'.
5966 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
5967 (microblaze_legitimate_pic_operand): Exclude function calls from
5968 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
5969 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
5971 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
5972 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
5973 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
5974 for 'address + offset'.
5975 (microblaze_expand_prologue): Add new function prologue call for
5977 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
5978 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
5979 table in case of TARGET_PIC_DATA_TEXT_REL.
5980 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
5981 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
5982 Add new macros 'UNSPEC_TEXT',
5983 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
5984 + exclude function calls from 'UNSPEC_PLT' in case of data text
5986 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
5987 new target hook for generating address diff vector tables in case of
5989 * doc/tm.texi : Regenerate.
5990 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
5991 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
5992 of addr diff vector generation.
5993 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
5994 target hook definition.
5995 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
5996 Add default function for generate_pic_addr_diff_vec -> flag_pic.
5997 * doc/invoke.texi (Add new pic option): Add new microblaze pic
5998 option for data text relative.
6000 2018-04-30 Richard Biener <rguenther@suse.de>
6002 * tree-chrec.h (evolution_function_is_constant_p): Remove
6004 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
6006 2018-04-30 Richard Biener <rguenther@suse.de>
6009 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
6011 2018-04-30 Richard Biener <rguenther@suse.de>
6013 PR tree-optimization/28364
6014 PR tree-optimization/85275
6015 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
6016 copying first exit test.
6018 2018-04-28 Mark Wielaard <mark@klomp.org>
6020 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
6022 (dwarf_AT): Handle DW_AT_addr_base.
6023 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
6025 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
6028 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
6029 (*ashl<dwi>3_doubleword_mask_1): Ditto.
6030 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
6031 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
6033 2018-04-28 Richard Biener <rguenther@suse.de>
6035 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
6036 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
6037 to reflect use. Only add interesting stmts.
6039 2018-04-27 Martin Jambor <mjambor@suse.cz>
6042 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
6043 the jump function allows for passing through aggregate values.
6045 2018-04-27 David Malcolm <dmalcolm@redhat.com>
6047 * input.h (in_system_header_at): Convert from macro to inline
6049 (from_macro_expansion_at): Likewise.
6050 (from_macro_definition_at): Likewise.
6052 2018-04-27 Jeff Law <law@redhat.com>
6054 * config.gcc: Mark tile* targets as deprecated/obsolete.
6056 2018-04-27 Richard Biener <rguenther@suse.de>
6058 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
6061 2018-04-27 Richard Biener <rguenther@suse.de>
6063 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
6065 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
6067 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
6068 with Yd constraint. Set "preferred_for_speed" attribute from
6069 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
6071 (*movdi_internal): Ditto.
6072 (movti_interunit splitters): Remove
6073 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
6074 (movdi_interunit splitters): Ditto.
6075 * config/i386/constraints.md (Ye): Remove.
6076 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
6078 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6081 * config/aarch64/constraints.md (Usg): Limit to 31.
6084 2018-04-27 Jakub Jelinek <jakub@redhat.com>
6086 PR tree-optimization/85529
6087 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
6088 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
6089 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
6090 zero extension or masking of the MSB bit.
6091 (optimize_range_tests): Add FIRST_BB argument, pass it through
6092 to optimize_range_tests_var_bound.
6093 (maybe_optimize_range_tests, reassociate_bb): Adjust
6094 optimize_range_tests callers.
6096 2018-04-26 Richard Biener <rguenther@suse.de>
6097 Jakub Jelinek <jakub@redhat.com>
6099 * cgraph.h (symbol_table): Just declare debug method here.
6100 * symtab.c (symbol_table::debug): Define.
6102 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
6104 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
6106 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
6108 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
6109 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
6110 (*movdi_internal): Substitute Yi and Yj constraint with x
6111 and Ym and Yn constraint with y constraint. Update "isa"
6112 attribute and set "preferred_for_speed" attribute from
6113 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
6114 (*movsi_internal): Ditto.
6115 (*movdf_internal): Ditto.
6116 (*movsf_internal): Ditto.
6117 (*zero_extendsidi2): Ditto.
6118 * config/i386/sse.md (vec_set<mode>_0): Ditto.
6119 (sse2_loadld): Ditto.
6120 (*vec_extract<ssevecmodelower>_0): Ditto.
6121 (*vec_extractv4si_0_zext_sse4): Ditto.
6122 (vec_concatv2di): Ditto.
6123 (*vec_dup<mode>): Ditto.
6124 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
6125 * config/i386/constraints.md (Yi): Remove.
6130 2018-04-26 Nathan Sidwell <nathan@acm.org>
6132 * dumpfile.c (dump_open): New.
6133 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
6134 (dump_finish): Detect stdio/stderr by value not name.
6136 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
6138 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
6140 2018-04-26 Tom de Vries <tom@codesourcery.com>
6143 * config/nvptx/nvptx.c (verify_neutering_jumps)
6144 (verify_neutering_labels): New function
6145 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
6147 2018-04-26 Tom de Vries <tom@codesourcery.com>
6150 * config/nvptx/nvptx.c (needs_neutering_p): New function.
6151 (nvptx_single): Use needs_neutering_p to skip over insns that do not
6154 2018-04-26 Richard Biener <rguenther@suse.de>
6155 Tom de Vries <tom@codesourcery.com>
6158 * lto-streamer-out.c (output_function): Fixup loops if required to match
6159 discovery done in the reader.
6161 2018-04-26 Richard Biener <rguenther@suse.de>
6163 PR tree-optimization/85116
6164 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
6165 have a loop exit from the single latch predecessor. Remove
6166 case of header with just condition.
6167 (ch_base::copy_headers): Exclude infinite loops from any
6169 (pass_ch::execute): Record exits.
6171 2018-04-26 Richard Biener <rguenther@suse.de>
6173 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
6174 prologue cost vector and pass it to vect_get_load_cost.
6175 (vect_get_peeling_costs_all_drs): Likewise.
6176 (vect_peeling_hash_get_lowest_cost): Likewise.
6177 (vect_enhance_data_refs_alignment): Likewise.
6179 2018-04-26 Richard Biener <rguenther@suse.de>
6182 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
6183 checking of integer<->pointer conversions.
6184 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
6185 sign-/zero-extending pointer types.
6186 (expand_omp_for_static_chunk): Likewise.
6188 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
6189 Jean Lee <xiaoyur347@gmail.com>
6191 * config/mips/mips.c (mips_asan_shadow_offset): New function.
6192 (TARGET_ASAN_SHADOW_OFFSET): Define.
6193 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
6194 true for -fsanitize=address.
6196 2018-04-25 Mark Wielaard <mark@klomp.org>
6198 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
6201 2018-04-25 Jakub Jelinek <jakub@redhat.com>
6203 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
6204 than "alu", remove explicit "memory" and "imm_disp" attributes.
6205 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
6208 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
6209 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
6210 gen_lowpart_no_emit.
6212 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
6215 * config/i386/i386.c (ix86_expand_builtin): Change memory
6216 operand to XI, extend p0 to Pmode.
6217 * config/i386/i386.md: Change unspec volatile and operand
6218 1 mode to XI, change operand 0 mode to P.
6220 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
6222 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
6223 GET_MODE_MASK before any checking.
6224 (nds32_can_use_bset_p): Likewise.
6225 (nds32_can_use_btgl_p): Likewise.
6227 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
6229 * config/nds32/nds32-doubleword.md: New define_split pattern for
6230 illegal register number.
6232 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
6234 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
6236 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
6238 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
6240 2018-04-25 Richard Biener <rguenther@suse.de>
6242 * lto-streamer.h (LTO_major_version): Bump to 8.
6244 2018-04-25 Jakub Jelinek <jakub@redhat.com>
6246 * BASE-VER: Set to 9.0.0.
6248 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
6250 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
6251 in __abskf2 and __powikf2.
6253 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6256 * config/aarch64/constraints.md (Usg, Usj): New constraints.
6257 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
6258 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
6259 Use the above on operand 2. Reindent.
6260 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
6262 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
6265 * common/config/i386/i386-common.c (ix86_handle_option): Don't
6267 * config/i386/i386.opt (mcet): Removed.
6268 * doc/install.texi: Remove -mcet documentation.
6269 * doc/invoke.texi: Likewise.
6271 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
6274 * doc/install.texi: Remove -mcet from bootstrap-cet.
6276 2018-04-24 Jakub Jelinek <jakub@redhat.com>
6279 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
6280 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
6284 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
6285 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
6286 containing a CONST_VECTOR.
6288 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
6290 * doc/install.texi: Update newlib dependency for nvptx.
6292 2018-04-24 Jakub Jelinek <jakub@redhat.com>
6295 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
6296 instead of INTVAL when shifting x left.
6298 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
6300 PR tree-optimization/85478
6301 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
6302 vect_grouped_store_supported for single element vectors.
6304 2018-04-24 Richard Biener <rguenther@suse.de>
6307 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
6308 load cost increase to the case of non-constant step.
6310 2018-04-24 Jakub Jelinek <jakub@redhat.com>
6313 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
6314 destination if any_malformed_asm.
6316 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
6319 * expr.c (store_field): In the bitfield case, if the value comes from
6320 a function call and is returned in registers by means of a PARALLEL,
6321 do not change the mode of the temporary unless BLKmode and VOIDmode.
6323 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
6325 PR rtl-optimization/85423
6326 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
6327 dependencies to debug insns when the previous insn is non-debug.
6329 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
6331 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
6332 enums into a single definition.
6333 (fls): Fix predicates and printing.
6336 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
6338 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
6339 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
6340 and short u6 immediate.
6341 (check_if_valid_sleep_operand): Remove.
6342 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
6344 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
6346 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
6347 flag_always_save_lp condition.
6348 * config/nds32/nds32.opt (malways-save-lp): New option.
6350 2018-04-22 Shiva Chen <shiva0217@gmail.com>
6352 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
6353 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
6354 * config/nds32/nds32.h
6355 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
6356 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
6358 2018-04-22 Shiva Chen <shiva0217@gmail.com>
6360 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
6361 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
6363 2018-04-22 Shiva Chen <shiva0217@gmail.com>
6364 Chung-Ju Wu <jasonwucj@gmail.com>
6366 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
6368 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
6369 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
6371 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
6373 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
6375 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
6377 * config/nds32/nds32-protos.h (nds32_data_alignment,
6378 nds32_local_alignment): Declare.
6379 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
6380 nds32_local_alignment): New functions.
6381 (TARGET_CONSTANT_ALIGNMENT): Define.
6382 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
6384 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
6386 * config/nds32/nds32.c
6387 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
6388 (TARGET_MODES_TIEABLE_P): Likewise.
6390 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
6392 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
6395 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
6396 Chung-Ju Wu <jasonwucj@gmail.com>
6398 * config/nds32/constants.md (unspec_volatile_element): Add enum values
6399 for unaligned access.
6400 * config/nds32/nds32-intrinsic.c: Implementation of expanding
6402 * config/nds32/nds32-intrinsic.md: Likewise.
6403 * config/nds32/nds32_intrinsic.h: Likewise.
6404 * config/nds32/nds32.h (nds32_builtins): Likewise.
6405 * config/nds32/nds32.opt (munaligned-access): New option.
6406 * config/nds32/nds32.c (nds32_asm_file_start): Display
6407 flag_unaligned_access status.
6409 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
6411 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
6412 -mno-relax is present.
6413 * config/riscv/linux.h (LINK_SPEC): Ditto.
6415 2018-04-20 Martin Sebor <msebor@redhat.com>
6418 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
6420 (gimple_fold_builtin_stxcpy_chk): Same.
6421 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
6423 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
6426 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
6427 __powikf2 when long double is IEEE 128-bit.
6429 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
6431 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
6432 step to make sure stack always aligned.
6434 2018-04-20 Carl Love <cel@us.ibm.com>
6437 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
6438 size check for arg0.
6440 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
6441 Tom de Vries <tom@codesourcery.com>
6444 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
6445 Emit insns for calls too.
6446 (nvptx_find_par): Always look for worker-level predecessor insn.
6447 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
6449 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
6450 (nvptx_process_pars): Propagate frames for calls.
6452 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
6455 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
6457 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
6458 (ix86_handle_option): Don't handle OPT_mibt.
6459 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
6461 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
6463 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
6464 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
6465 (ix86_target_macros): Define __CET__ with flag_cf_protection
6466 for -fcf-protection.
6467 * config/i386/i386.c (isa2_opts): Remove -mibt.
6468 * config/i386/i386.h (TARGET_IBT): Removed.
6469 (TARGET_IBT_P): Likewise.
6470 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
6471 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
6472 * config/i386/i386.opt (mcet): Update help message.
6475 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
6476 -mcet as an alias for -mshstk.
6478 2018-04-20 Richard Biener <rguenther@suse.de>
6481 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
6482 complexity by forcing a single use of the multiply operand.
6484 2018-04-20 Martin Jambor <mjambor@suse.cz>
6487 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
6488 recursion dependency to only apply to non-clones.
6490 2018-04-20 Martin Jambor <mjambor@suse.cz>
6493 * ipa-cp.c (create_specialized_node): Check that clones of
6494 self-recursive edges exist during IPA-CP.
6496 2018-04-19 Toon Moene <toon@moene.org>
6498 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
6501 2018-04-19 Jakub Jelinek <jakub@redhat.com>
6503 PR tree-optimization/85467
6504 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
6505 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
6506 VECTOR_CST element to type.
6508 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
6511 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
6512 * config/i386/i386.md (builtin_setjmp_setup): Removed.
6513 (builtin_longjmp): Likewise.
6514 (save_stack_nonlocal): New pattern.
6515 (restore_stack_nonlocal): Likewise.
6517 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
6520 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
6521 Replace ASM_OUTPUT_LABEL with fprintf.
6523 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
6526 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
6527 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
6528 * config/i386/i386-c.c (ix86_target_macros_internal): Also
6529 define __IBT__ and __SHSTK__ for -fcf-protection.
6530 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
6532 (ix86_trampoline_init): Likewise.
6533 (x86_output_mi_thunk): Likewise.
6534 (ix86_notrack_prefixed_insn_p): Likewise.
6535 (ix86_option_override_internal): Don't disallow -fcf-protection.
6536 * config/i386/i386.md (rdssp<mode>): Also enable for
6538 (incssp<mode>): Likewise.
6539 (nop_endbr): Likewise.
6540 * config/i386/i386.opt (mcet): Change help message to built-in
6544 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
6545 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
6546 enable CET built-in functions.
6548 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
6550 * common/config/i386/i386-common.c
6551 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
6552 OPTION_MASK_ISA_MOVDIRI_UNSET,
6553 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
6554 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
6555 * config.gcc (movdirintrin.h): New header.
6556 * config/i386/cpuid.h (bit_MOVDIRI,
6557 bit_MOVDIR64B): New bits.
6558 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
6560 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
6561 (VOID, PVOID, PCVOID)): New function types.
6562 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
6563 __builtin_ia32_directstoreu_u64,
6564 __builtin_ia32_movdir64b): New builtins.
6565 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
6566 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
6568 (ix86_valid_target_attribute_inner_p): Ditto.
6569 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
6570 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
6571 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
6572 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
6573 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
6574 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
6575 (movdiri<mode>, movdir64b_<mode>): New.
6576 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
6577 * config/i386/immintrin.h: Include movdirintrin.h.
6578 * config/i386/movdirintrin.h: New file.
6579 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
6581 2018-04-19 Richard Biener <rguenther@suse.de>
6584 * cfg.c (clear_bb_flags): When loop state says we have
6585 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
6587 2018-04-19 Richard Biener <rguenther@suse.de>
6589 PR tree-optimization/84737
6590 * tree-vect-data-refs.c (vect_copy_ref_info): New function
6591 copying restrict info.
6592 (vect_setup_realignment): Use it.
6593 * tree-vectorizer.h (vect_copy_ref_info): Declare.
6594 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
6595 the first DR to all generated stores.
6596 (vectorizable_load): Likewise for loads.
6598 2018-04-19 Jakub Jelinek <jakub@redhat.com>
6600 PR tree-optimization/85446
6601 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
6602 the integral and pointer types to have the same precision.
6604 * doc/install.texi: Document --disable-cet being the default and
6607 2018-04-18 Martin Liska <mliska@suse.cz>
6609 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
6612 2018-04-18 Martin Liska <mliska@suse.cz>
6615 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
6618 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
6619 arguments if they are comparable.
6621 2018-04-18 Martin Liska <mliska@suse.cz>
6624 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
6627 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
6630 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
6633 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
6634 ENDBR after calling __morestack.
6636 2018-04-18 David Malcolm <dmalcolm@redhat.com>
6639 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
6640 by using gcc_base_ver to generate a gcc_driver_version, and use
6641 it when generating GCC_DRIVER_NAME.
6642 * configure: Regenerate.
6644 2018-04-18 Jakub Jelinek <jakub@redhat.com>
6647 * config.gcc: Obsolete powerpc*-*-*spe*.
6649 2018-04-17 Jakub Jelinek <jakub@redhat.com>
6652 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
6653 (stabstr_D): Change type of unum from unsigned int to
6654 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
6657 2018-04-17 Jim Wilson <jimw@sifive.com>
6660 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
6661 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
6662 Set arg_pointer_offset after using pretend_args_size.
6664 2018-04-17 Jakub Jelinek <jakub@redhat.com>
6666 PR rtl-optimization/85431
6667 * dse.c (record_store): Ignore zero width stores.
6670 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
6671 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
6672 __builtin_stack_restore rather than after it.
6673 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
6674 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
6675 argument instead of virtual_dynamic_stack_rtx.
6677 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
6679 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
6681 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
6682 Add note to error message to explain internal mapping of overloaded
6683 built-in function name to non-overloaded built-in function name.
6684 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
6687 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
6690 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
6691 where the inputs overlap with the output.
6693 2018-04-17 Jakub Jelinek <jakub@redhat.com>
6695 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
6696 (=v, v) alternative and explicit "memory" attribute.
6697 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
6698 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
6700 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
6701 "sselog1" type instead of "sselog".
6702 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
6703 "sselog". Remove explicit "memory" attribute.
6704 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
6705 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
6707 (vec_extract_hi_v32hi): Merge all alternatives into one, use
6708 "sselog1" type instead of "sselog". Remove explicit "memory"
6710 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
6711 use "sselog1" type instead of "sselog". Remove explicit "memory"
6713 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
6714 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
6716 (vec_extract_hi_v64qi): Merge all alternatives into one, use
6717 "sselog1" type instead of "sselog". Remove explicit "memory"
6719 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
6720 use "sselog1" type instead of "sselog". Remove explicit "memory"
6724 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
6727 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
6730 2018-04-17 Martin Jambor <mjambor@suse.cz>
6733 * ipa-cp.c (create_specialized_node): Call
6734 expand_all_artificial_thunks if necessary.
6736 2018-04-17 Martin Liska <mliska@suse.cz>
6739 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
6740 in message, remote space in between '_G' and '('.
6742 2018-04-17 Jakub Jelinek <jakub@redhat.com>
6745 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
6746 avx512f_vmcmp<mode>3<round_saeonly_name>,
6747 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
6748 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
6749 avx512f_rndscale<mode><round_saeonly_name>,
6750 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
6751 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
6752 Use %<iptr>2 instead of %2 for -masm=intel.
6753 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
6754 avx512f_vcvttss2usi<round_saeonly_name>,
6755 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
6757 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
6758 avx512f_vcvttsd2usi<round_saeonly_name>,
6759 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
6760 Use %q1 instead of %1 for -masm=intel.
6761 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
6762 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
6763 of %3 for -masm=intel.
6764 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
6766 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
6768 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
6770 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
6771 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
6773 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
6775 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
6776 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
6777 %g1 and one with %0 and %1.
6778 (avx512er_vmrcp28<mode><round_saeonly_name>,
6779 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
6781 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
6782 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
6783 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
6784 of %0 and %{%4%} for -masm=intel.
6785 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
6786 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
6787 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
6788 order of %0 and %{%5%}%{z%} for -masm=intel.
6790 2018-04-17 Jan Hubicka <jh@suse.cz>
6793 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
6795 2018-04-17 Martin Liska <mliska@suse.cz>
6798 * multiple_target.c (create_dispatcher_calls): Set apostrophes
6799 for target_clone error message. Make default implementation
6800 clone to be a local declaration.
6801 (separate_attrs): Add new argument and check for an empty
6803 (expand_target_clones): Handle it.
6804 (ipa_target_clone): Make redirection just for target_clones
6807 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
6808 Tom de Vries <tom@codesourcery.com>
6811 * omp-expand.c (expand_oacc_for): Add dummy false branch for
6812 tiled basic blocks without omp continue statements.
6814 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
6817 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
6818 vec_extract expression as having side effects to make sure it gets
6821 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
6824 * config/i386/i386.c (get_builtin_code_for_version): Check
6827 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
6830 * gcc/config.gcc: Support "skylake".
6831 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
6833 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
6834 (processor_target_table): Add "skylake".
6835 (ix86_option_override_internal): Add "skylake".
6836 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
6837 PROCESSOR_CANNONLAKE.
6838 (get_builtin_code_for_version): Fix priority for
6839 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
6840 PROCESSOR_SKYLAKE-AVX512.
6841 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
6842 (processor_type): Add PROCESSOR_SKYLAKE.
6844 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
6845 Jason Merrill <jason@redhat.com>
6848 * convert.c (convert_to_integer_1): Use direct recursion for
6849 enumeral types and types with a precision less than the number
6850 of bits in their mode.
6852 2018-04-16 Julia Koval <julia.koval@intel.com>
6855 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
6856 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
6858 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
6861 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
6862 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
6863 and -mno-direct-move.
6865 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
6868 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
6869 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
6870 Ensure negative shifts result in {0}.
6872 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
6874 PR rtl-optimization/79916
6875 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
6876 regs (if any) to define how to gnerate SD moves when LRA is in
6879 2018-04-13 Jakub Jelinek <jakub@redhat.com>
6881 PR rtl-optimization/85393
6882 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
6883 * except.c (expand_dw2_landing_pad_for_region): Make static.
6884 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
6885 a label and unconditional jump to old_bb, rather than
6886 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
6889 PR rtl-optimization/85376
6890 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
6891 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
6892 instead of a specific value.
6894 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
6895 Bin Cheng <bin.cheng@arm.com>
6897 PR tree-optimization/82965
6898 PR tree-optimization/83991
6899 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
6900 by_profile_only parameter.
6901 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
6902 information if the loop was predicted to iterate too many times.
6903 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
6905 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
6908 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
6911 2018-04-13 Martin Liska <mliska@suse.cz>
6912 Jakub Jelinek <jakub@redhat.com>
6915 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
6916 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
6917 * builtins.c (expand_builtin_memory_copy_args): Use
6918 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
6919 handle dest_addr == pc_rtx.
6921 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
6924 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
6925 asked to not generate direct moves.
6926 (fix_trunc<mode>si2_stfiwx): Similar.
6927 (fix_trunc<mode>si2_internal): Similar.
6929 2018-04-12 Jakub Jelinek <jakub@redhat.com>
6932 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
6933 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
6934 lookup if dest in some wider mode is known to be const0_rtx and
6935 if so, record permanent equivalence for it to be ZERO_EXTEND of
6936 the narrower mode destination.
6938 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
6940 * lto-streamer-out.c (output_function): Revert 259346.
6941 * omp-expand.c (expand_oacc_for): Likewise.
6943 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
6945 PR rtl-optimization/85354
6946 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
6947 * sel-sched.c (sel_global_init): ... here.
6949 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
6952 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
6953 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
6954 mode for PE-COFF targets.
6955 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
6956 (i386_pe_asm_lto_end): Likewise.
6957 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
6958 (TARGET_ASM_LTO_END): Likewise.
6959 * config/i386/winnt.c (saved_debug_info_level): New static variable.
6960 (i386_pe_asm_lto_start): New function.
6961 (i386_pe_asm_lto_end): Likewise.
6963 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
6964 Richard Biener <rguenther@suse.de>
6967 * lto-streamer-out.c (output_function): Fix CFG loop state before
6969 * omp-expand.c (expand_oacc_for): Handle calls to internal
6970 functions like regular functions.
6972 2018-04-12 Richard Biener <rguenther@suse.de>
6975 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
6976 for the early LTO debug to properly generate references to it
6977 during DIE emission. Do not re-use that for the skeleton for
6979 (dwarf2out_early_finish): Likewise.
6981 2018-04-12 Jakub Jelinek <jakub@redhat.com>
6984 * config/i386/sse.md
6985 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
6986 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
6987 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
6988 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
6989 and output is a reg, avoid creating invalid lowpart subreg, but
6990 instead split into a 512-bit move. Don't split if not AVX512VL,
6991 input is xmm16+ reg and output is a mem.
6992 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
6993 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
6994 xmm16+ reg and output is a mem.
6996 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6998 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
6999 also for flag_dwarf2_cfi_asm.
7001 2018-04-12 Jakub Jelinek <jakub@redhat.com>
7003 PR rtl-optimization/85342
7004 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
7005 a bool scalar var inside of the loop instead. Don't try to update
7006 recog_data.operand after failed apply_change_group.
7008 2018-04-12 Tom de Vries <tom@codesourcery.com>
7011 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
7012 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
7013 array with flexible array member as array without given dimension.
7014 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
7015 argument for undefined param to true.
7017 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
7020 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
7021 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
7022 from PowerPC section.
7023 * config/rs6000/sysv4.opt (mcall-): Improve help text.
7024 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
7025 help text that is too long.
7026 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
7027 help text that is too long.
7028 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
7029 help text that is too long.
7031 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
7033 * config/alpha/alpha.md (stack_probe_internal): Rename
7034 from "probe_stack". Update all callers.
7036 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
7038 PR rtl-optimization/84566
7039 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
7040 sched_macro_fuse_insns.
7042 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
7045 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
7046 (compute_block_dependences): ... from here.
7048 2018-04-11 Jakub Jelinek <jakub@redhat.com>
7050 PR tree-optimization/85331
7051 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
7052 from int to HOST_WIDE_INT.
7054 2018-04-11 Martin Jambor <mjambor@suse.cz>
7057 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
7058 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
7059 not the same as the source val.
7060 (cgraph_edge_brings_value_p): New parameter.
7061 (gather_edges_for_value): Pass destination value to
7062 cgraph_edge_brings_value_p.
7063 (perhaps_add_new_callers): Likewise.
7064 (get_info_about_necessary_edges): Likewise and exclude values brought
7065 only by self-recursive edges.
7066 (create_specialized_node): Redirect only clones of self-calling edges.
7067 (+self_recursive_pass_through_p): New function.
7068 (find_more_scalar_values_for_callers_subset): Use it.
7069 (find_aggregate_values_for_callers_subset): Likewise.
7070 (known_aggs_to_agg_replacement_list): Removed.
7071 (decide_whether_version_node): Re-calculate known constants for all
7072 remaining context clones.
7074 2018-04-11 Richard Biener <rguenther@suse.de>
7077 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
7078 from early DWARF output.
7079 (dwarf2out_early_finish): Output line info unconditionally into
7080 early DWARF and add reference to it.
7082 2018-04-11 Jakub Jelinek <jakub@redhat.com>
7085 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
7086 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
7087 other than V2DFmode using iptr mode attribute.
7088 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
7090 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
7092 PR rtl-optimization/84659
7093 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
7095 2018-04-11 Jakub Jelinek <jakub@redhat.com>
7098 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
7100 (output_loc_list): Pass address of a dummy size variable even in the
7101 locview handling loop.
7102 (index_location_lists): Add comment on why skip_loc_list_entry can't
7105 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
7108 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
7111 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
7114 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
7115 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
7116 and -mstring-compare-inline-limit.
7118 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
7121 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
7122 for stack clash protection in a register whenever we need it to be in
7125 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
7127 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
7128 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
7130 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
7133 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
7135 (mlong-double-): Ditto.
7136 * config/rs6000/sysv4.opt (msdata=): Ditto.
7137 (mtls-size=): Ditto.
7139 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
7141 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
7142 erroneous entries for
7143 "vector int vec_ldl (int, long int *)", and
7144 "vector unsigned int vec_ldl (int, unsigned long int *)".
7145 Add comments and entries for
7146 "vector bool char vec_ldl (int, bool char *)",
7147 "vector bool short vec_ldl (int, bool short *)",
7148 "vector bool int vec_ldl (int, bool int *)",
7149 "vector bool long long vec_ldl (int, bool long long *)",
7150 "vector pixel vec_ldl (int, pixel *)",
7151 "vector long long vec_ldl (int, long long *)",
7152 "vector unsigned long long vec_ldl (int, unsigned long long *)".
7153 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
7154 type tree bool_long_long_type_node and correct definition of
7155 bool_V2DI_type_node to make reference to this new type tree.
7156 (rs6000_mangle_type): Replace erroneous reference to
7157 bool_long_type_node with bool_long_long_type_node.
7158 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
7159 comments to emphasize sign distinctions for char and int types and
7160 replace RS6000_BTI_bool_long constant with
7161 RS6000_BTI_bool_long_long constant. Also add comment to restrict
7162 use of RS6000_BTI_pixel.
7163 (bool_long_type_node): Remove this macro definition.
7164 (bool_long_long_type_node): New macro definition
7166 2018-04-10 Jakub Jelinek <jakub@redhat.com>
7168 PR rtl-optimization/85300
7169 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
7170 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
7171 simplify_unary_operation fails.
7173 2018-04-10 Martin Liska <mliska@suse.cz>
7175 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
7176 cgraph_edge and ipa_ref.
7178 2018-04-10 Jakub Jelinek <jakub@redhat.com>
7182 * config/i386/sse.md
7183 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
7184 computation of the VEC_MERGE selector from mask.
7185 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
7186 Fix decoding of the VEC_MERGE selector into mask.
7188 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
7190 PR tree-optimization/85286
7191 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
7193 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
7195 * final.c (final_1): Set insn_last_address as well as
7196 insn_current_address.
7198 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7201 * explow.c (emit_stack_probe): Call validize_mem on memory location
7202 before passing it to gen_probe_stack. Create address operand and
7203 legitimize it for the probe_stack_address case.
7205 2018-04-09 Jan Hubicka <jh@suse.cz>
7208 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
7209 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
7210 * tree.c (free_lang_data_in_type): Fix handling of binfos;
7212 (free_lang_data): Rebuild type inheritance graph.
7214 2018-04-09 Martin Sebor <msebor@redhat.com>
7216 * invoke.texi (-finline-small-functions): Mention other optimization
7218 (-findirect-inlining, -fpartial-inlining): Same.
7219 (-finline-functions-called-once): Same.
7220 (-freorder-blocks-and-partition): Same.
7222 2018-04-09 Jan Hubicka <jh@suse.cz>
7225 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
7226 jumps; choose last target that matches the criteria (i.e.
7227 no partition changes for non-crossing jumps).
7228 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
7229 support for redirecting crossing jumps to non-crossing.
7231 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
7233 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
7234 also for naked functions.
7236 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
7238 * config/arc/arc.md (add_shift): New pattern.
7239 (add_shift2): Likewise.
7240 (sub_shift): Likewise.
7241 (sub_shift_cmp0_noout): Likewise.
7242 (compare_si_ashiftsi): Likewise.
7243 (xbfu_cmp0_noout): New combine pattern.
7244 (xbfu_cmp0"): Likewise.
7245 (movsi_set_cc_insn): Place the predicable variant first.
7246 (commutative_binary_cmp0_noout): Remove clobber.
7247 (commutative_binary_cmp0): New pattern.
7248 (noncommutative_binary_cmp0): Likewise.
7249 (noncommutative_binary_cmp0_noout): Likewise.
7250 (noncommutative_binary_comparison_result_used): Removed.
7251 (rsub_cmp0): New pattern.
7252 (rsub_cmp0_noout): Likewise.
7253 (extzvsi): Changed, keep only meaningful variants.
7254 (SQH, SEZ): New iterators.
7255 (SQH_postfix): New mode attribute.
7256 (SEZ_prefix): New code attribute.
7257 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
7258 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
7259 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
7261 (noncommutative_operator): Check the availability of barrel
7264 2018-04-09 Richard Biener <rguenther@suse.de>
7266 PR tree-optimization/85284
7267 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
7268 Only use the niter constraining form of simple_iv when the exit
7271 2018-04-09 Tom de Vries <tom@codesourcery.com>
7274 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
7275 (define_expand "*memory_barrier"): New define_expand.
7276 (define_insn "memory_barrier"): New insn.
7278 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
7280 PR rtl-optimization/80463
7281 PR rtl-optimization/83972
7282 PR rtl-optimization/83480
7284 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
7285 correct producer for the insn.
7286 (tidy_control_flow): Fixup seqnos in case of debug insns.
7288 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
7290 PR rtl-optimization/83913
7292 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
7293 different sched-times when merging exprs.
7295 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
7297 PR rtl-optimization/83962
7299 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
7300 tidy_fallthru_edge and tidy_control_flow.
7302 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
7304 PR rtl-optimization/83530
7306 * sel-sched.c (force_next_insn): New global variable.
7307 (remove_insn_for_debug): When force_next_insn is true, also leave only
7308 next insn in the ready list.
7309 (sel_sched_region): When the region wasn't scheduled, make another pass
7310 over it with force_next_insn set to 1.
7312 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
7314 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
7316 * config/nds32/constants.md (unspec_volatile_element): Add enum values
7317 for interrupt control.
7318 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
7319 functions for interrupt control.
7320 * config/nds32/nds32-intrinsic.md: Likewise.
7321 * config/nds32/nds32_intrinsic.h: Likewise.
7322 * config/nds32/nds32.h (nds32_builtins): Likewise.
7324 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
7326 * config/nds32/nds32.c (nds32_init_machine_status,
7327 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
7328 strict_aligned_p field.
7329 (nds32_expand_to_rtl_hook): New function.
7330 (TARGET_EXPAND_TO_RTL_HOOK): Define.
7331 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
7333 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
7334 Chung-Ju Wu <jasonwucj@gmail.com>
7336 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
7337 * config/nds32/nds32-n7.md: New file.
7338 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
7339 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
7341 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
7342 * config/nds32/nds32.md (pipeline_model): Add n7.
7343 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
7344 * config/nds32/pipelines.md: Include n7 settings.
7346 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
7347 Chung-Ju Wu <jasonwucj@gmail.com>
7349 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
7350 * config/nds32/nds32-e8.md: New file.
7351 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
7352 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
7354 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
7355 * config/nds32/nds32.md (pipeline_model): Add e8.
7356 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
7357 * config/nds32/pipelines.md: Include e8 settings.
7359 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
7360 Chung-Ju Wu <jasonwucj@gmail.com>
7362 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
7363 * config/nds32/nds32-n8.md: New file.
7364 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
7365 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
7367 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
7368 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
7369 * config/nds32/nds32.md (pipeline_model): Add n8.
7370 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
7371 * config/nds32/pipelines.md: Include n8 settings.
7373 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
7374 Chung-Ju Wu <jasonwucj@gmail.com>
7376 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
7377 * config/nds32/nds32-n9-2r1w.md: New file.
7378 * config/nds32/nds32-n9-3r2w.md: New file.
7379 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
7380 nds32_register_ports): New or modify for cpu n9.
7381 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
7383 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
7384 * config/nds32/nds32-utils.c: New file.
7385 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
7386 TARGET_MUL_SLOW): Define.
7387 * config/nds32/nds32.md (pipeline_model): New attribute.
7388 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
7389 New options that support cpu n9.
7390 * config/nds32/pipelines.md: Include n9 settings.
7391 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
7393 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
7395 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
7396 information if necessary.
7397 (output_cond_branch_compare_zero): Likewise.
7398 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
7399 (nds32_target_alignment): Refine for alignment.
7400 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
7401 (FUNCTION_BOUNDARY): Modify.
7402 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
7404 * config/nds32/nds32.opt (malways-align, malign-functions): New.
7406 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
7408 * config/nds32/constants.md (unspec_volatile_element): Add values for
7409 TLB operation and data prefetch.
7410 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
7411 functions for TLB operation and data prefetch.
7412 * config/nds32/nds32-intrinsic.md: Likewise.
7413 * config/nds32/nds32_intrinsic.h: Likewise.
7414 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
7415 (nds32_print_operand): Likewise.
7416 * config/nds32/nds32.h (nds32_builtins): Likewise.
7418 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
7419 Andrew Pinski <pinsika@gcc.gnu.org>
7422 * match.pd: Use constant_boolean_node of correct type instead of
7423 boolean_true_node or boolean_false_node for simplifying
7424 pointer comparisons to zero.
7426 2018-04-07 Jakub Jelinek <jakub@redhat.com>
7428 PR tree-optimization/80021
7429 * tree.c (verify_type_variant): Make error call in verify_variant_match
7430 translatable and remove final full stop.
7432 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
7434 * config/nds32/constants.md (unspec_volatile_element): Add
7435 UNSPEC_VOLATILE_EH_RETURN.
7436 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
7437 nds32_output_stack_pop): Support dwarf exception handling process.
7438 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
7439 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
7440 exception handling process.
7441 (nds32_compute_stack_frame): Likewise.
7442 (nds32_return_addr_rtx): Likewise.
7443 (nds32_initial_elimination_offset): Likewise.
7444 (nds32_expand_prologue): Likewise.
7445 (nds32_expand_epilogue): Likewise.
7446 (nds32_dynamic_chain_address): New function.
7447 * config/nds32/nds32.h (machine_function): Add fields for dwarf
7449 (DYNAMIC_CHAIN_ADDRESS): Define.
7450 (EH_RETURN_DATA_REGNO): Define.
7451 (EH_RETURN_STACKADJ_RTX): Define.
7452 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
7453 patterns for dwarf exception handling.
7455 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
7457 * config/nds32/nds32.h: Clean up obsolete macros.
7459 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
7461 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
7462 Add enum values for particular instructions.
7463 * config/nds32/nds32-intrinsic.c: Implementation of expanding
7464 particular intrinsic functions.
7465 * config/nds32/nds32-intrinsic.md: Likewise.
7466 * config/nds32/nds32_intrinsic.h: Likewise.
7467 * config/nds32/nds32.h (nds32_builtins): Likewise.
7468 * config/nds32/nds32.md (type): Add pbsad and pbsada.
7469 (btst, ave): New patterns for particular instructions.
7471 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
7473 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
7474 Add enum values for atomic load/store and memory sync.
7475 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
7477 * config/nds32/nds32-intrinsic.md: Likewise.
7478 * config/nds32/nds32_intrinsic.h: Likewise.
7479 * config/nds32/nds32.h (nds32_builtins): Likewise.
7481 2018-04-07 Jakub Jelinek <jakub@redhat.com>
7483 PR tree-optimization/85257
7484 * fold-const.c (native_encode_vector): If not all elts could fit
7485 and off is -1, return 0 rather than offset.
7486 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
7487 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
7488 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
7489 adjust buffer in native_interpret_expr call.
7491 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
7493 * config/nds32/constants.md (unspec_volatile_element): Add cache
7494 control enum values.
7495 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
7496 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
7497 * config/nds32/nds32.c (nds32_cctl_names): New.
7498 (nds32_print_operand): Handle cache control register names.
7499 * config/nds32/nds32.h (nds32_builtins): New enum values.
7500 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
7502 * config/nds32/nds32.md (type): Add mmu.
7503 * config/nds32/pipelines.md (simple_insn): Add mmu.
7505 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
7507 * config/nds32/nds32.md (type): Remove call.
7508 * config/nds32/pipelines.md (simple_insn): Likewise.
7510 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
7512 * config/nds32/constants.md (unspec_volatile_element): Add
7513 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
7514 UNSPEC_VOLATILE_FMFCFG.
7515 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
7516 description for fmfcfg and fmfcsr.
7517 (bdesc_1arg): Add fmtcsr.
7518 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
7519 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
7520 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
7521 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
7522 unspec_fmfcfg): New patterns.
7523 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
7524 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
7525 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
7526 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
7527 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
7528 __nds32__fmfcfg): Define.
7530 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
7532 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
7533 intrinsic register names.
7534 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
7535 intrinsic register enum values and macros.
7537 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
7539 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
7540 for load/store addressing form.
7541 (nds32_print_operand_address): Likewise.
7543 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
7546 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
7547 based on LABEL_REF. Remove useless assertion.
7548 (pic_address_needs_scratch): Fix formatting.
7549 (sparc_legitimize_pic_address): Minor tweaks.
7550 (sparc_delegitimize_address): Adjust assertion accordingly.
7551 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
7552 into symbolic_operand.
7553 (movsi_high_pic_label_ref): Likewise.
7554 (movsi_lo_sum_pic_label_ref): Likewise.
7555 (movdi_pic_label_ref): Likewise.
7556 (movdi_high_pic_label_ref): Likewise.
7557 (movdi_lo_sum_pic_label_ref): Likewise.
7559 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
7561 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
7562 custom LIB_SPEC setup.
7564 2018-04-06 Ruslan Bukin <br@bsdpad.com>
7565 Kito Cheng <kito.cheng@gmail.com>
7567 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
7568 * config/riscv/freebsd.h: New.
7570 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
7572 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
7573 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
7576 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
7577 Kito Cheng <kito.cheng@gmail.com>
7579 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
7580 nds32_output_call, nds32_symbol_binds_local_p): New functions.
7581 * config/nds32/nds32-protos.h (nds32_output_call,
7582 nds32_output_return): Declare.
7583 * config/nds32/nds32.md: Refine all the call and return patterns.
7585 2018-04-06 Jakub Jelinek <jakub@redhat.com>
7588 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
7589 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
7591 PR rtl-optimization/84872
7592 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
7593 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
7596 2018-04-06 Tamar Christina <tamar.christina@arm.com>
7598 * expr.c (copy_blkmode_to_reg): Revert 254862.
7599 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
7601 2018-04-06 Richard Biener <rguenther@suse.de>
7604 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
7605 after seeing a component reference with an adjacent field. Treat
7606 refs to arrays at struct end of external decls similar to
7607 refs to unconstrained commons.
7609 2018-04-06 Jakub Jelinek <jakub@redhat.com>
7612 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
7613 look through SAVE_EXPRs with non-side-effects argument. Adjust
7615 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
7618 2018-04-06 Richard Biener <rguenther@suse.de>
7621 * alias.c (find_base_term): New wrapper around find_base_term
7622 unwinding CSELIB_VAL_PTR changes.
7623 (find_base_term): Do not restore CSELIB_VAL_PTR during the
7626 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7628 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
7630 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
7631 constant definitions.
7632 ("nop"): lr 0,0 -> nopr r0
7633 ("nop_lr0", "nop_lr1"): New insn definitions.
7635 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
7637 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
7638 NDS32_V3PUSH_AVAILABLE_P macro.
7640 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
7641 Chung-Ju Wu <jasonwucj@gmail.com>
7643 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
7644 (nds32*-*-*): Add float and fpu_config into supported_defaults.
7645 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
7646 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
7647 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
7648 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
7649 * config/nds32/constraints.md: New constraints and checking for hard
7650 float configuration.
7651 * config/nds32/iterators.md: New mode iterator and attribute for hard
7652 float configuration.
7653 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
7655 * config/nds32/nds32-fpu.md: New file.
7656 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
7657 deal with hard float code generation.
7658 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
7660 (abi_type, float_reg_number): New enum type.
7661 * config/nds32/nds32-predicates.c: New predicates for hard float.
7662 * config/nds32/nds32-protos.h: Declare functions for hard float.
7663 * config/nds32/nds32.c: Implementation for hard float configuration.
7664 * config/nds32/nds32.h: Definitions for hard float configuration.
7665 * config/nds32/nds32.md: Include hard float machine description and
7666 modify patterns for hard float configuration.
7667 * config/nds32/nds32.opt: New options for hard float configuration.
7668 * config/nds32/predicates.md: New predicates for hard float
7671 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
7673 * common/config/nds32/nds32-common.c
7674 (nds32_option_optimization_table): Enable -mreleax-hint by default.
7676 2018-04-05 Jakub Jelinek <jakub@redhat.com>
7679 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
7680 CONSTRUCTOR_ELT (ctor, ...)->value.
7682 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
7685 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
7687 2018-04-05 Tom de Vries <tom@codesourcery.com>
7690 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
7693 2018-04-05 Shiva Chen <shiva0217@gmail.com>
7694 Kito Cheng <kito.cheng@gmail.com>
7696 * config/nds32/constraints.md (U33): Fine-tune checking condition.
7697 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
7698 * config/nds32/nds32.h (nds32_16bit_address_type): Add
7699 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
7701 2018-04-05 Shiva Chen <shiva0217@gmail.com>
7702 Kito Cheng <kito.cheng@gmail.com>
7704 * config/nds32/constraints.md (Ufe): New memory constraint.
7705 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
7706 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
7707 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
7709 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
7710 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
7712 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7714 * config/nds32/nds32.md: Use optimize_size in the condition for
7715 alu-shift instructions.
7717 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7719 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
7721 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7723 * config/nds32/nds32.md (negsi2): Refine pattern.
7725 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
7726 Chung-Ju Wu <jasonwucj@gmail.com>
7728 * config/nds32/iterators.md (shift_rotate): New code iterator.
7729 (shift): New code attribute.
7730 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
7731 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
7732 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
7733 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
7734 bit-wise operations.
7735 (andsi3, *andsi3): Ditto.
7736 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
7737 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
7738 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
7739 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
7740 nds32_ior_operand, nds32_xor_operand): New predicates.
7742 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7744 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
7745 (addsi3, subsi3): ... this.
7747 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7749 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
7751 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7753 * config/nds32/nds32.md: Adjust indention.
7755 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
7757 * config/nds32/nds32.md (feature): New attribute.
7759 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7761 * config/nds32/nds32.md (subtype): New attribute.
7763 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
7766 * config/arm/arm-builtins.c (arm_expand_builtin): Change
7767 expansion to perform a bitwise AND of the argument followed by a
7768 boolean negation of the result.
7770 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
7772 PR rtl-optimization/84878
7773 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
7774 the basic block. Assert the use reference is not artificial and that
7775 it has an associated insn.
7777 2018-04-04 Michael Matz <matz@suse.de>
7779 * builtins.c (compute_objsize): Pass correct operand
7780 to array_at_struct_end_p.
7782 2018-04-04 Richard Biener <rguenther@suse.de>
7785 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
7786 from contexts for DINFO_LEVEL_TERSE and below.
7788 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
7790 * config/nds32/nds32-doubleword.md (move_<mode>): Require
7791 resiter_operand condition.
7792 * config/nds32/nds32.md (*move<mode>): Ditto.
7794 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
7795 Monk Chiang <sh.chiang04@gmail.com>
7797 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
7799 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
7801 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
7803 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
7804 Kito Cheng <kito.cheng@gmail.com>
7806 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
7807 nds32_cond_code_str, output_cond_branch,
7808 output_cond_branch_compare_zero, nds32_expand_cbranch,
7809 nds32_expand_cstore, nds32_expand_movcc,
7810 nds32_output_cbranchsi4_equality_zero,
7811 nds32_output_cbranchsi4_equality_reg,
7812 nds32_output_cbranchsi4_equality_reg_or_const_int,
7813 nds32_output_cbranchsi4_greater_less_zero: New functions.
7814 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
7815 nds32_expand_cstore, nds32_expand_movcc,
7816 nds32_output_cbranchsi4_equality_zero,
7817 nds32_output_cbranchsi4_equality_reg,
7818 nds32_output_cbranchsi4_equality_reg_or_const_int,
7819 nds32_output_cbranchsi4_greater_less_zero): Declare.
7820 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
7821 nds32_rimm11s_operand): New predicates.
7822 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
7823 * config/nds32/nds32.md: Rewrite all the branch and conditional move
7826 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
7828 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
7829 * config/nds32/nds32.md: Ditto.
7830 * config/nds32/pipelines.md: Ditto.
7832 2018-04-04 Richard Biener <rguenther@suse.de>
7834 PR tree-optimization/85168
7835 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
7836 propagating abnormals.
7838 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
7840 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
7842 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
7843 Kito Cheng <kito.cheng@gmail.com>
7845 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
7846 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
7847 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
7848 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
7849 * config/nds32/nds32.md (sibcall_internal): New.
7850 (sibcall_register): Remove.
7851 (sibcall_immediate): Remove.
7852 (sibcall_value_internal): New.
7853 (sibcall_value_register): Remove.
7854 (sibcall_value_immediate): Remove.
7855 * config/nds32/predicates.md (nds32_general_register_operand): New.
7856 (nds32_call_address_operand): New.
7858 2018-04-03 Jakub Jelinek <jakub@redhat.com>
7860 PR rtl-optimization/85167
7861 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
7862 bb_defs if *split_p, instead preinitialize it to NULL.
7864 PR tree-optimization/85156
7865 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
7866 evaluating the argument multiple times.
7868 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
7870 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
7872 (_mm_cvtpd_ps): Likewise.
7873 (_mm_cvttpd_epi32): Likewise.
7874 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
7875 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
7876 vector, pixel, and bool following altivec.h include.
7878 2018-04-03 Martin Sebor <msebor@redhat.com>
7880 * doc/extend.texi (Common Function Attributes): Clarify.
7881 (const attribute): Likewise.
7882 (pure attribute): Likewise.
7884 2018-04-03 Jakub Jelinek <jakub@redhat.com>
7887 * config/i386/i386.c (ix86_expand_vector_set): Use
7888 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
7890 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
7892 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
7893 instructions when changing rounding bits to preserve precision bits
7894 in the x87 control word.
7896 2018-04-03 Martin Liska <mliska@suse.cz>
7898 PR tree-optimization/82491
7899 * rtl.h (strip_offset_and_add): Replace += suboffset with
7900 poly_uint64 () + suboffset.
7902 2018-03-29 Martin Liska <mliska@suse.cz>
7903 Martin Jambor <mjambor@suse.cz>
7906 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
7907 param_type is not an integral or pointer type.
7909 2018-04-03 Richard Biener <rguenther@suse.de>
7911 * sese.h (recompute_all_dominators): Remove.
7913 2018-04-02 Martin Sebor <msebor@redhat.com>
7915 * doc/invoke.texi (-Wrestrict): Fix typos.
7917 2018-04-02 Jim Wilson <jimw@sifive.com>
7919 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
7920 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
7921 (<optab>di3, <optab>si3_extend): Likewise.
7922 (<optab>si3_mask, <optab>si3_mask_1): New.
7923 (<optab>di3_mask, <optab>di3_mask_1): New.
7924 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
7925 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
7926 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
7928 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
7930 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
7933 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
7935 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
7936 (nds32_canonicalize_comparison): New function.
7938 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
7939 Kito Cheng <kito.cheng@gmail.com>
7940 Kuan-Lin Chen <kuanlinchentw@gmail.com>
7942 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
7943 * config/nds32/constants.md (unspec_volatile_element): Add
7944 UNSPEC_VOLATILE_RELAX_GROUP.
7945 * config/nds32/nds32-relax-opt.c: New file.
7946 * config/nds32/nds32-predicates.c
7947 (nds32_symbol_load_store_p): New function.
7948 * config/nds32/nds32-protos.h
7949 (nds32_symbol_load_store_p): Declare function.
7950 (make_pass_nds32_relax_opt): Declare new rtl pass function.
7951 * config/nds32/nds32.c
7952 (nds32_register_pass): New function to register pass.
7953 (nds32_register_passes): New function to register passes.
7954 * config/nds32/nds32.md (relax_group): New pattern.
7955 * config/nds32/nds32.opt (mrelax-hint): New option.
7956 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
7958 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
7960 * config/nds32/t-nds32: Modify files dependency.
7962 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
7964 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
7965 (PROFILE_HOOK): Define its implementation.
7967 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
7969 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
7970 type and 32-bit size.
7972 2018-04-01 Jakub Jelinek <jakub@redhat.com>
7975 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
7976 (V_128_256): New mode iterator.
7977 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
7978 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
7979 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
7981 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
7982 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
7984 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
7987 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
7988 NaN inputs correctly.
7990 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
7993 * config/rs6000/vsx.md (??r): New mode attribute.
7994 (*vsx_mov<mode>_64bit): Use it.
7995 (*vsx_mov<mode>_32bit): Likewise.
7997 2018-03-30 Martin Sebor <msebor@redhat.com>
7999 PR tree-optimization/84818
8000 * builtins.c (check_access): Use warning_n.
8002 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8005 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
8007 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
8010 2018-03-30 Julia Koval <julia.koval@intel.com>
8013 * x86-tune.def (movx, partial_reg_dependency): Enable for
8016 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
8019 * lra-constraints.c (process_alt_operands): Move setting
8020 this_alternative_matches below.
8022 2018-03-29 Martin Liska <mliska@suse.cz>
8025 * doc/invoke.texi: Document how LTO works with debug info.
8026 Describe auto-load support of binutils. Mention 'x86-64'
8027 as valid option value of -march option.
8029 2018-03-29 Jakub Jelinek <jakub@redhat.com>
8031 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
8034 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
8035 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
8036 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
8039 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
8042 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
8043 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
8044 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
8045 for RS6000_BTM_POWERPC64.
8046 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
8047 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
8048 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
8053 2018-03-28 Carl Love <cel@us.ibm.com>
8056 2017-09-27 Carl Love <cel@us.ibm.com>
8058 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
8059 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
8060 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
8063 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8065 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
8066 instead of __vector bool.
8067 (_mm_max_pu8): Likewise.
8068 (_mm_min_pi16): Likewise.
8070 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
8073 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
8074 (DIVWEUO): Likewise.
8076 (DIVDEUO): Likewise.
8077 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
8078 DIVWEUO and DIVDEUO.
8079 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
8080 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
8081 (div_extend): Likewise.
8082 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
8084 (__builtin_divweuo): Likewise.
8085 (__builtin_divdeo): Likewise.
8086 (__builtin_divdeuo): Likewise.
8088 2018-03-28 Jakub Jelinek <jakub@redhat.com>
8091 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
8092 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
8094 PR tree-optimization/82004
8095 * gimple-match-head.c (optimize_pow_to_exp): New function.
8096 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
8097 Don't fold to exp if optimize_pow_to_exp is false.
8099 2018-03-28 Martin Liska <mliska@suse.cz>
8102 * calls.c (initialize_argument_information): Fix trailing space.
8103 * common.opt: Fix typo and provide better explanation for
8104 -fsanitize-coverage option.
8105 * config/i386/i386.opt: Fix typo.
8107 2018-03-28 Jakub Jelinek <jakub@redhat.com>
8108 Martin Liska <mliska@suse.cz>
8111 * gimplify.c (asan_poison_variable): Don't do the check for
8112 gimplify_omp_ctxp here.
8113 (gimplify_decl_expr): Do it here.
8114 (gimplify_target_expr): Likewise.
8116 2018-03-28 Martin Liska <mliska@suse.cz>
8119 * config/i386/i386.c (ix86_function_arg_advance): Do not call
8120 chkp_type_bounds_count if MPX is not enabled.
8122 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
8124 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
8126 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
8129 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
8130 function to create the function decl for complex long double
8131 multiply and divide for -mabi=ieeelongdouble.
8132 (init_float128_ieee): Call it.
8134 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
8137 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
8138 -fcf-protection=branch -mibt.
8139 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
8141 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
8144 * config/arm/arm.c (arm_valid_symbolic_address): Handle
8145 arm_word_relocations.
8147 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
8150 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
8151 extern array declarations.
8153 2018-03-27 Richard Biener <rguenther@suse.de>
8156 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
8157 explicit single_use checks.
8159 2018-03-27 Richard Biener <rguenther@suse.de>
8161 PR tree-optimization/85082
8162 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
8165 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
8167 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
8168 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
8169 Turn on fasynchronous-unwind-tables and funwind-tables.
8171 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
8174 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
8175 (*bmi_blsr_<mode>_ccz): Ditto.
8177 2018-03-26 Tom de Vries <tom@codesourcery.com>
8179 PR tree-optimization/85063
8180 * omp-general.c (offloading_function_p): New function. Factor out
8182 * omp-offload.c (pass_omp_target_link::gate): ... here.
8183 * omp-general.h (offloading_function_p): Declare.
8184 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
8185 with attribute omp declare target for offloading functions.
8187 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
8189 PR tree-optimization/84005
8190 * tree-data-ref.h (get_base_for_alignment): Declare.
8191 * tree-data-ref.c (get_base_for_alignment_1): New function.
8192 (get_base_for_alignment): Likewise.
8193 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
8194 get_base_for_alignment to find a suitable base object, instead
8195 of always using drb->base_address.
8197 2018-03-23 Jakub Jelinek <jakub@redhat.com>
8200 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
8201 known size by default.
8203 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
8206 * lra-constraints.c (process_alt_operands): Don't match BLKmode
8207 and non BLKmode operands.
8209 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8212 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
8213 Clean up attributes.
8215 2018-03-23 Richard Biener <rguenther@suse.de>
8218 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
8219 we are going to emit early debug for LTO.
8221 2018-03-23 Jakub Jelinek <jakub@redhat.com>
8224 * function.c (match_asm_constraints_1): Don't optimize if input
8225 doesn't satisfy general_operand predicate for output's mode.
8228 * alias.c (write_dependence_p): Don't require for x_canonicalized
8229 non-VOIDmode if x has VOIDmode.
8232 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
8233 just don't try to optimize it rather than assert it never happens.
8235 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
8237 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
8238 macro expansions for definition of ST_INTERNAL_<mode> and
8239 LD_INTERNAL_<mode> builtins.
8240 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
8242 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
8244 (altivec_expand_st_builtin): Likewise.
8245 (altivec_expand_builtin): Remove calls to deleted functions.
8246 (rs6000_address_for_altivec): Delete this function.
8247 * config/rs6000/vector.md: Remove expands for
8248 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
8250 2018-03-22 Sudakshina Das <sudi.das@arm.com>
8253 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
8254 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
8255 re-computing once computed.
8256 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
8257 (arm_init_machine_status): Initialize
8258 machine->static_chain_stack_bytes.
8260 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
8263 * doc/extend.texi: Add four new prototypes for vec_ld.
8264 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
8265 definitions for more logical presentation.
8266 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
8267 entries for V1TI variants of __builtin_altivec_ld builtin.
8268 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
8269 handling of V1TI variant of LVX icode pattern.
8270 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
8271 (rs6000_gimple_fold_builtin): Likewise.
8272 (altivec_init_builtins): Add code to define
8273 __builtin_altivec_lvx_v1ti function.
8275 2018-03-22 Jakub Jelinek <jakub@redhat.com>
8278 * function.c (match_asm_constraints_1): Don't do the optimization
8279 if input isn't a REG, SUBREG, MEM or constant.
8281 2018-03-22 Tom de Vries <tom@codesourcery.com>
8283 PR tree-optimization/84956
8284 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
8285 bb_has_abnormal_pred.
8287 2018-03-22 Jakub Jelinek <jakub@redhat.com>
8290 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
8291 DECL_INITIAL (decl) to decl at the end.
8292 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
8295 2018-03-21 Joseph Myers <joseph@codesourcery.com>
8297 * doc/extend.texi (__builtin_tgmath): Document when complex
8298 integer types are treated as _Complex _Float64.
8300 2018-03-21 Tom de Vries <tom@codesourcery.com>
8302 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
8304 2018-03-21 Jakub Jelinek <jakub@redhat.com>
8306 PR tree-optimization/84960
8307 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
8308 if it is ENTRY block, move them into single succ of ENTRY in that case.
8310 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
8312 PR tree-optimization/84811
8313 * poly-int.h (poly_span_traits): Remove the T3 parameter and
8314 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
8315 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
8316 (known_subrange_p): Update accordingly. Cast each value involved
8317 in the size comparison, rather than casting the result of the
8320 2018-03-21 Jakub Jelinek <jakub@redhat.com>
8322 PR tree-optimization/84982
8323 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
8324 by flipping the least significant bit rather than all bits from
8325 bitpos to bitpos + bitsize - 1.
8327 2018-03-21 Nathan Sidwell <nathan@acm.org>
8329 * doc/extend.texi (Deprecated Features): Remove mention of
8330 long-deleted deprecations.
8332 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8335 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
8336 * configure: Regenerate.
8338 2018-03-21 Tom de Vries <tom@codesourcery.com>
8340 PR tree-optimization/83126
8341 * tree-parloops.c (num_phis): New function.
8342 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
8344 2018-03-21 Nathan Sidwell <nathan@acm.org>
8346 * doc/extend.texi (Deprecated Features): Update deprecated flags,
8347 mention anon-struct/union members and trailing attributes.
8349 2018-03-21 Bin Cheng <bin.cheng@arm.com>
8351 PR tree-optimization/84969
8352 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
8353 builtin memset partitions if they set different rhs values.
8355 2018-03-21 Jakub Jelinek <jakub@redhat.com>
8357 PR rtl-optimization/84989
8358 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
8359 VEC_DUPLICATE with scalar result mode.
8361 2018-03-21 Martin Liska <mliska@suse.cz>
8364 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
8365 not intended return statement.
8367 2018-03-21 Martin Liska <mliska@suse.cz>
8370 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
8371 (chkp_find_bound_slots_1): Limit number of iterations.
8373 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
8376 * Minor grammar fixes for x86 options.
8378 2018-03-20 Jakub Jelinek <jakub@redhat.com>
8381 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
8382 holding REG_CFA_RESTORE notes, instead turn them into a USE.
8384 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
8387 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
8388 (altivec_lvx_<mode>_1op): Likewise.
8389 (altivec_stvx_<mode>_2op): Likewise.
8390 (altivec_stvx_<mode>_1op): Likewise.
8391 (altivec_lvx_<VM2:mode>): New define_expand.
8392 (altivec_stvx_<VM2:mode>): Likewise.
8393 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
8394 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
8395 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
8396 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
8397 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
8398 (rs6000_gen_lvx): Likewise.
8399 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
8400 (altivec_expand_stv_builtin): Likewise.
8401 (altivec_expand_builtin): Likewise.
8402 * config/rs6000/vector.md: Likewise.
8404 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8407 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
8410 2018-03-20 Richard Biener <rguenther@suse.de>
8413 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
8414 sign-conversions as zero, fall back to standard scalar_stmt
8417 2018-03-20 Martin Liska <mliska@suse.cz>
8420 * predict.c (rebuild_frequencies): Handle case when we have
8421 PROFILE_ABSENT, but flag_guess_branch_prob is false.
8423 2018-03-20 Jakub Jelinek <jakub@redhat.com>
8426 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
8427 flag_section_anchors.
8428 * varasm.c (use_blocks_for_decl_p): Remove hack for
8429 dw2_force_const_mem.
8432 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
8434 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
8435 be created, use lowpart_subreg of operands[0] rather than operands[0]
8437 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
8438 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
8439 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
8440 and n constraint instead of aarch64_shift_imm_di and Usd.
8441 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
8442 (*aarch64_<optab>_reg_minus<mode>3): ... this.
8444 2018-03-20 Sudakshina Das <sudi.das@arm.com>
8447 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
8448 to favor GPR over NEON registers.
8449 (<shift>di3_neon): Likewise.
8451 2018-03-20 Tom de Vries <tom@codesourcery.com>
8454 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
8455 (nvptx_process_pars): Emit bar.sync asap and alap.
8457 2018-03-20 Tom de Vries <tom@codesourcery.com>
8460 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
8461 seen_label if seen_label is already set.
8463 2018-03-20 Jakub Jelinek <jakub@redhat.com>
8466 * config/i386/i386.c (fold_builtin_cpu): For features above 31
8467 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
8468 Use 1U instead of 1. Formatting fixes.
8471 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
8472 instead of TREE_TYPE (s1) for the return value.
8474 2018-03-19 Jakub Jelinek <jakub@redhat.com>
8476 PR tree-optimization/84946
8477 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
8478 bitsize + bitsize in poly_uint64 rather than poly_int64.
8481 * dwarf2asm.c: Include fold-const.c.
8482 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
8483 of decl rather than decl itself.
8485 PR rtl-optimization/84643
8486 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
8488 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
8491 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
8492 calling assemble_variable.
8494 2018-03-19 Sudakshina Das <sudi.das@arm.com>
8497 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
8498 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
8500 2018-03-19 Jim Wilson <jimw@sifive.com>
8503 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
8504 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
8505 (riscv_first_stack_step): Likewise.
8506 (riscv_option_override): Use STACK_BOUNDARY instead of
8508 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
8510 (BIGGEST_ALIGNMENT): Set to 128.
8511 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
8512 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
8515 2018-03-19 Richard Biener <rguenther@suse.de>
8517 PR tree-optimization/84933
8518 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
8519 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
8521 2018-03-19 Richard Biener <rguenther@suse.de>
8523 PR tree-optimization/84859
8524 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
8525 (cond_if_else_store_replacement): Perform sinking operation on
8526 single-store BBs regardless of MAX_STORES_TO_SINK setting.
8527 Generalize what a BB with a single eligible store is.
8529 2018-03-19 Richard Biener <rguenther@suse.de>
8531 PR tree-optimization/84929
8532 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
8533 chrec_is_positive against non-chrec arg.
8535 2018-03-19 Tamar Christina <tamar.christina@arm.com>
8538 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
8540 2018-03-18 Martin Liska <mliska@suse.cz>
8542 PR rtl-optimization/84635
8543 * regrename.c (build_def_use): Use matches_mode only when
8546 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
8548 PR tree-optimization/84913
8549 * tree-vect-loop.c (vectorizable_reduction): Don't try to
8550 vectorize chains of COND_EXPRs.
8552 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
8554 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
8556 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
8558 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
8560 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
8562 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
8564 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
8565 Kito Cheng <kito.cheng@gmail.com>
8567 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
8568 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
8569 (nds32_adjust_reg_alloc_order): New function.
8570 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
8572 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
8574 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
8575 nds32_print_operand, nds32_print_operand_address): Use
8576 HOST_WIDE_INT_PRINT_DEC instead.
8578 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
8580 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
8582 2018-03-17 Jakub Jelinek <jakub@redhat.com>
8585 * config/i386/i386.c (initial_ix86_tune_features,
8586 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
8588 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
8589 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
8590 rather than 1u << ix86_tune. Formatting fix.
8591 (ix86_option_override_internal): Change ix86_arch_mask from
8592 unsigned int to unsigned HOST_WIDE_INT, initialize to
8593 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
8594 (ix86_function_specific_restore): Likewise.
8596 2018-03-16 Jakub Jelinek <jakub@redhat.com>
8599 * postreload.c (reload_combine_recognize_pattern): Perform
8600 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
8601 truncate_int_for_mode the result for the destination's mode.
8604 * hsa-gen.c (mem_type_for_type): Fix comment typo.
8605 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
8607 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
8610 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
8613 * lra-assigns.c (lra_split_hard_reg_for): Don't use
8614 regno_allocno_class_array and sorted_pseudos.
8615 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
8616 insns where regno is used.
8618 2018-03-16 Martin Liska <mliska@suse.cz>
8621 * multiple_target.c (create_dispatcher_calls): Redirect
8622 reference in the symbol table.
8624 2018-03-16 Martin Liska <mliska@suse.cz>
8627 * multiple_target.c (create_dispatcher_calls): Redirect also
8630 2018-03-16 Jakub Jelinek <jakub@redhat.com>
8634 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
8635 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
8636 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
8638 2018-03-16 Julia Koval <julia.koval@intel.com>
8640 * doc/invoke.texi (Skylake Server): Add CLWB.
8641 Cannonlake): Remove CLWB.
8643 2018-03-16 Jakub Jelinek <jakub@redhat.com>
8645 PR tree-optimization/84841
8646 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
8648 (FLOAT_ONE_CONST_TYPE): Define.
8649 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
8650 (sort_by_operand_rank): Put entries with higher constant_type last
8651 rather than first to match comments.
8653 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
8655 * config/nios2/nios2.md (movsi_internal): Fix thinko in
8658 2018-03-15 Jakub Jelinek <jakub@redhat.com>
8661 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
8662 check and use address of target always.
8664 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
8667 * config/i386/i386.c (indirect_thunk_needed): Update comments.
8668 (indirect_thunk_bnd_needed): Likewise.
8669 (indirect_thunks_used): Likewise.
8670 (indirect_thunks_bnd_used): Likewise.
8671 (indirect_return_needed): New.
8672 (indirect_return_bnd_needed): Likewise.
8673 (output_indirect_thunk_function): Add a bool argument for
8675 (output_indirect_thunk_function): Don't generate alias for
8676 function return thunk.
8677 (ix86_code_end): Call output_indirect_thunk_function to generate
8678 function return thunks.
8679 (ix86_output_function_return): Set indirect_return_bnd_needed
8680 and indirect_return_needed instead of indirect_thunk_bnd_needed
8681 and indirect_thunk_needed.
8683 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
8685 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
8686 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
8687 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
8689 2018-03-15 David Malcolm <dmalcolm@redhat.com>
8690 Paul Hua <paul.hua.gm@gmail.com>
8693 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
8695 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
8697 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
8698 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
8701 2018-03-15 Tamar Christina <tamar.christina@arm.com>
8704 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
8705 instead of GET_MODE_SIZE when comparing Units.
8707 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
8710 * varasm.c (hash_section): Return an unchangeble hash value
8711 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
8712 Return !aarch64_can_use_per_function_literal_pools_p ().
8714 2018-03-15 Jakub Jelinek <jakub@redhat.com>
8717 * optabs.c (emit_conditional_move): Pass address of cmode's copy
8718 rather than address of cmode as last argument to prepare_cmp_insn.
8720 2018-03-15 Julia Koval <julia.koval@intel.com>
8722 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
8723 F_AVX512VNNI, F_AVX512BITALG): New.
8725 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
8728 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
8729 insn for floating-point loads and stores.
8731 2018-03-14 Carl Love <cel@us.ibm.com>
8733 * config/rs6000/rs6000-c.c: Add macro definitions for
8734 ALTIVEC_BUILTIN_VEC_PERMXOR.
8735 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
8736 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
8737 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
8738 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
8740 * config/doc/extend.texi: Add prototypes for vec_permxor.
8742 2018-03-14 David Malcolm <dmalcolm@redhat.com>
8745 * diagnostic-show-locus.c (class layout_point): Convert m_line
8746 from int to linenum_type.
8747 (line_span::comparator): Use linenum "compare" function when
8748 comparing line numbers.
8749 (test_line_span): New function.
8750 (layout_range::contains_point): Convert param "row" from int to
8752 (layout_range::intersects_line_p): Likewise.
8753 (layout::will_show_line_p): Likewise.
8754 (layout::print_source_line): Likewise.
8755 (layout::should_print_annotation_line_p): Likewise.
8756 (layout::print_annotation_line): Likewise.
8757 (layout::print_leading_fixits): Likewise.
8758 (layout::annotation_line_showed_range_p): Likewise.
8759 (struct line_corrections): Likewise for field m_row.
8760 (line_corrections::line_corrections): Likewise for param "row".
8761 (layout::print_trailing_fixits): Likewise.
8762 (layout::get_state_at_point): Likewise.
8763 (layout::get_x_bound_for_row): Likewise.
8764 (layout::print_line): Likewise.
8765 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
8766 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
8767 * input.c (selftest::test_linenum_comparisons): New function.
8768 (selftest::input_c_tests): Call it.
8769 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
8770 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
8771 * selftest.h (ASSERT_GT): New macro.
8772 (ASSERT_GT_AT): New macro.
8773 (ASSERT_LT): New macro.
8774 (ASSERT_LT_AT): New macro.
8776 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
8778 PR rtl-optimization/84780
8779 * combine.c (distribute_links): Don't make a link based on pc_rtx.
8781 2018-03-14 Martin Liska <mliska@suse.cz>
8783 * tree.c (record_node_allocation_statistics): Use
8784 get_stats_node_kind.
8785 (get_stats_node_kind): New function extracted from
8786 record_node_allocation_statistics.
8787 (free_node): Use get_stats_node_kind.
8789 2018-03-14 Richard Biener <rguenther@suse.de>
8791 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
8792 that the value-set of ANTIC_IN doesn't grow.
8795 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
8797 (BB_VISITED_WITH_VISITED_SUCCS): New define.
8798 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
8800 2018-03-14 Julia Koval <julia.koval@intel.com>
8802 * config.gcc (icelake-client, icelake-server): New.
8804 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
8805 (initial_ix86_arch_features): Ditto.
8806 (PTA_SKYLAKE): Add SGX.
8807 (PTA_ICELAKE): Remove.
8808 (PTA_ICELAKE_CLIENT): New.
8809 (PTA_ICELAKE_SERVER): New.
8810 (ix86_option_override_internal): Split up icelake on icelake client and
8812 (get_builtin_code_for_version): Ditto.
8813 (fold_builtin_cpu): Ditto.
8814 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
8815 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
8816 * config/i386/i386.h (processor_type): Ditto.
8817 * doc/invoke.texi: Ditto.
8819 2018-03-14 Jakub Jelinek <jakub@redhat.com>
8822 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
8823 INTEGER_CST offset, add it together with bitpos / 8 and
8824 sign extend based on POINTER_SIZE.
8828 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
8831 * config/i386/constraints.md (Yc): New register constraint.
8832 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
8833 Use Yc constraint for alternative 2 of operand 0. Remove
8834 preferred_for_speed attribute.
8836 2018-03-14 Richard Biener <rguenther@suse.de>
8838 PR tree-optimization/84830
8839 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
8840 with the old one to avoid oscillations.
8842 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
8845 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
8847 (assign_by_spills): Return a flag of reload assignment failure.
8848 Do not process the reload assignment failures. Do not spill other
8849 reload pseudos if they has the same reg class. Update n if
8851 (lra_assign): Add a return arg. Set up from the result of
8852 assign_by_spills call.
8853 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
8854 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
8855 usage_insns if it is not NULL.
8856 (spill_hard_reg_in_range): New function.
8857 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
8858 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
8859 function prototypes.
8860 (lra_assign): Change prototype.
8861 * lra.c (lra): Add code to deal with fails by splitting hard reg
8864 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
8866 * config/riscv/riscv.opt (mrelax): New option.
8867 * config/riscv/riscv.c (riscv_file_start): Emit ".option
8868 "norelax" when riscv_mrelax is disabled.
8869 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
8871 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8874 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
8875 reassociation for int modes.
8877 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
8879 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
8880 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
8882 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
8883 * config/aarch64/aarch64-sve.md
8884 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
8885 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
8886 (*extend<mode><Vwide>2): Rename to...
8887 (aarch64_sve_extend<mode><Vwide>2): ...this.
8888 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
8889 renaming the old pattern to...
8890 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
8892 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
8893 define_expand, renaming the old pattern to...
8894 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
8895 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
8896 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
8897 account when deciding which SVE instruction the optab should use.
8898 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
8900 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
8902 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
8903 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
8904 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
8905 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
8906 (tlsdesc_small_<mode>): Turn a define_expand and use
8907 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
8908 (tlsdesc_small_advsimd_<mode>): ...this.
8909 (tlsdesc_small_sve_<mode>): New pattern.
8911 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
8913 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
8914 (UNSPEC_UMUL_HIGHPART): New constants.
8915 (MUL_HIGHPART): New int iteraor.
8916 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
8917 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
8919 (*<su>mul<mode>3_highpart): New define_insn.
8921 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
8924 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
8927 2018-03-13 Martin Liska <mliska@suse.cz>
8930 * (sem_item_optimizer::sem_item_optimizer): Initialize new
8932 (sem_item_optimizer::~sem_item_optimizer): Release it.
8933 (sem_item_optimizer::merge_classes): Register variable aliases.
8934 (sem_item_optimizer::fixup_pt_set): New function.
8935 (sem_item_optimizer::fixup_points_to_sets): Likewise.
8936 * ipa-icf.h: Declare new variables and functions.
8938 2018-03-13 Jakub Jelinek <jakub@redhat.com>
8941 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
8942 integer_pow2p@2 and test integer_pow2p in condition.
8943 (A < 0 ? C : 0): Similarly for @1.
8946 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
8947 characters starting at p contain '\0' character, don't look beyond
8951 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
8952 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
8955 * reg-stack.c (change_stack): Change update_end var from int to
8956 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
8957 also call set_block_for_insn on the newly added insns and rescan.
8960 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
8961 on the last operand.
8964 * tree.c (stabilize_reference_1): Return save_expr (e) for
8965 STATEMENT_LIST even if it doesn't have side-effects.
8967 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
8969 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
8971 2018-03-12 Renlin Li <renlin.li@arm.com>
8973 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
8974 aarch64_output_scalar_simd_mov_immediate.
8976 2018-03-12 Martin Sebor <msebor@redhat.com>
8978 PR tree-optimization/83456
8979 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
8980 for perfectly overlapping calls to memcpy.
8981 (gimple_fold_builtin_memory_chk): Same.
8982 (gimple_fold_builtin_strcpy): Handle no-warning.
8983 (gimple_fold_builtin_stxcpy_chk): Same.
8984 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
8986 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
8988 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
8989 parameter. Use it for SFmode.
8990 (rs6000_function_arg_advance_1): Adjust.
8991 (rs6000_function_arg): Adjust.
8992 (rs6000_gimplify_va_arg): Pass false for that new parameter.
8994 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
8996 PR rtl-optimization/84169
8997 PR rtl-optimization/84780
8998 * combine.c (can_combine_p): Check for a 2-insn combination whether
8999 the destination register is used between the two insns, too.
9001 2018-03-12 Richard Biener <rguenther@suse.de>
9003 PR tree-optimization/84803
9004 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
9005 for refs DR analysis didn't process.
9007 2018-03-12 Richard Biener <rguenther@suse.de>
9009 PR tree-optimization/84777
9010 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
9011 force-vectorize loops ignore whether we are optimizing for size.
9013 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
9015 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
9016 (TARGET_MD_ASM_ADJUST): Define.
9018 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
9019 Kito Cheng <kito.cheng@gmail.com>
9020 Chung-Ju Wu <jasonwucj@gmail.com>
9022 * config/nds32/nds32.c (nds32_compute_stack_frame,
9023 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
9024 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
9025 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
9026 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
9027 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
9028 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
9029 * config/nds32/nds32.md (prologue, epilogue): Use macro
9030 NDS32_V3PUSH_AVAILABLE_P to do checking.
9032 2018-03-11 Jakub Jelinek <jakub@redhat.com>
9035 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
9036 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
9037 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
9038 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
9039 addition of most attributes on !orig_type_die or the attribute not
9040 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
9042 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
9043 Chung-Ju Wu <jasonwucj@gmail.com>
9045 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
9047 * config/nds32/nds32.opt (mvh): New option.
9049 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
9050 Chung-Ju Wu <jasonwucj@gmail.com>
9052 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
9054 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
9055 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
9058 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
9059 Chung-Ju Wu <jasonwucj@gmail.com>
9061 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
9063 * config/nds32/nds32-multiple.md (strlensi): New pattern.
9064 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
9066 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
9067 Kito Cheng <kito.cheng@gmail.com>
9068 Chung-Ju Wu <jasonwucj@gmail.com>
9070 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
9071 UNSPEC_FFMISM and UNSPEC_FLMISM.
9072 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
9073 for ffb, ffmism and flmism.
9074 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
9075 (unspec_ffmism): Ditto.
9076 (unspec_flmism): Ditto.
9077 (nds32_expand_builtin_impl): Check if string extension is available.
9078 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
9079 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
9081 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
9084 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
9087 * lra-assigns.c (assign_by_spills): Return a flag of reload
9088 assignment failure. Do not process the reload assignment
9089 failures. Do not spill other reload pseudos if they has the same
9091 (lra_assign): Add a return arg. Set up from the result of
9092 assign_by_spills call.
9093 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
9094 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
9095 usage_insns if it is not NULL.
9096 (spill_hard_reg_in_range): New function.
9097 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
9098 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
9099 function prototypes.
9100 (lra_assign): Change prototype.
9101 * lra.c (lra): Add code to deal with fails by splitting hard reg
9104 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
9107 * config/i386/i386.opt: Replace Enforcment with Enforcement.
9109 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
9112 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
9113 (dw_val_node): Add val_symbolic_view.
9114 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
9115 (symview_upper_bound): New.
9116 (new_line_info_table): Initialize symviews_since_reset.
9117 (dwarf2out_source_line): Count symviews_since_reset and set
9118 symview_upper_bound.
9119 (dw_val_equal_p): Handle symview.
9120 (add_AT_symview): New.
9121 (print_dw_val): Handle symview.
9122 (attr_checksum, attr_checksum_ordered): Likewise.
9123 (same_dw_val_p, size_of_die): Likewise.
9124 (value_format, output_die): Likewise.
9125 (add_high_low_attributes): Use add_AT_symview for entry_view.
9126 (dwarf2out_finish): Reset symview_upper_bound, clear
9129 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
9132 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
9133 Add strict argument and use it.
9134 (rs6000_split_multireg_move): Update for new strict argument.
9135 (mem_operand_gpr): Disallow all non-offsettable addresses.
9136 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
9138 2018-03-09 Jakub Jelinek <jakub@redhat.com>
9141 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
9142 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
9143 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
9146 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
9147 decl, use remap_type if we want to use the type.
9149 2018-03-09 Martin Sebor <msebor@redhat.com>
9151 PR tree-optimization/84526
9152 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
9154 (builtin_access::generic_overlap): Be prepared to handle non-array
9157 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
9159 PR rtl-optimization/84682
9160 * lra-constraints.c (process_address_1): Check is_address flag
9161 for address constraints.
9162 (process_alt_operands): Likewise.
9163 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
9164 preprocess_constraints.
9165 * recog.h (preprocess_constraints): Add oploc parameter.
9167 * recog.c (preprocess_constraints): Test address_operand for
9168 CT_ADDRESS constraints.
9170 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
9173 * lra-assigns.c (assign_by_spills): Return a flag of reload
9174 assignment failure. Do not process the reload assignment
9175 failures. Do not spill other reload pseudos if they has the same
9177 (lra_assign): Add a return arg. Set up from the result of
9178 assign_by_spills call.
9179 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
9180 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
9181 usage_insns if it is not NULL.
9182 (spill_hard_reg_in_range): New function.
9183 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
9184 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
9185 function prototypes.
9186 (lra_assign): Change prototype.
9187 * lra.c (lra): Add code to deal with fails by splitting hard reg
9190 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9193 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
9194 Accept complain bool parameter. Only emit errors if it is true.
9195 (arm_parse_cpu_option_name): Likewise.
9196 (arm_target_thumb_only): Adjust callers of the above.
9197 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
9198 prototype to take a default true bool parameter.
9199 (arm_parse_arch_option_name): Likewise.
9201 2018-03-09 David Malcolm <dmalcolm@redhat.com>
9202 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
9206 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
9207 * configure: Regenerate.
9208 * configure.ac ("linker --version-script option"): New.
9209 ("linker soname option"): New.
9211 2018-03-09 Richard Biener <rguenther@suse.de>
9213 PR tree-optimization/84775
9214 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
9215 immediate uses of predicate stmts and mark them modified.
9218 PR tree-optimization/84178
9219 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
9221 (version_loop_for_if_conversion): Delay update_ssa call.
9222 (tree_if_conversion): Delay update_ssa until after predicate
9225 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
9228 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
9229 when the function accesses prior frames.
9231 2018-03-08 Jakub Jelinek <jakub@redhat.com>
9234 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
9235 gen_llsym, otherwise call maybe_gen_llsym.
9238 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
9239 has ',' character inside of it.
9241 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9244 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
9245 as clobbering CC_REGNUM.
9247 2018-03-08 Richard Biener <rguenther@suse.de>
9250 * tree-scalar-evolution.c: Include tree-into-ssa.h.
9251 (follow_copies_to_constant): Do not follow SSA names registered
9254 2018-03-08 Richard Biener <rguenther@suse.de>
9256 PR tree-optimization/84178
9257 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
9259 (version_loop_for_if_conversion): Delay update_ssa call.
9260 (tree_if_conversion): Delay update_ssa until after predicate
9263 2018-03-08 David Malcolm <dmalcolm@redhat.com>
9265 PR tree-optimization/84178
9266 * tree-if-conv.c (release_bb_predicate): Remove the
9267 the assertion that the stmts have NULL use_ops.
9268 Discard the statements, asserting that they haven't
9269 yet been added to a BB.
9271 2018-03-08 Richard Biener <rguenther@suse.de>
9273 PR tree-optimization/84746
9274 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
9275 (phi_translate): Pass in destination ANTIC_OUT set.
9276 (phi_translate_1): Likewise. For a simplified result lookup
9277 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
9278 (phi_translate_set): Adjust.
9279 (do_pre_regular_insertion): Likewise.
9280 (do_pre_partial_partial_insertion): Likewise.
9282 2018-03-08 Martin Liska <mliska@suse.cz>
9284 PR gcov-profile/84735
9285 * doc/gcov.texi: Document usage of profile files.
9286 * gcov-io.h: Document changes in the format.
9288 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
9292 * dwarf2out.c (struct dw_line_info_table): Update comments for
9294 (FORCE_RESET_NEXT_VIEW): New.
9295 (FORCE_RESETTING_VIEW_P): New.
9296 (RESETTING_VIEW_P): Check for -1 too.
9297 (ZERO_VIEW_P): Likewise.
9298 (new_line_info_table): Force-reset next view.
9299 (dwarf2out_begin_function): Likewise.
9300 (dwarf2out_source_line): Simplify zero_view_p initialization.
9301 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
9302 view directly. Omit view when omitting .loc at line 0.
9304 2018-03-08 Jakub Jelinek <jakub@redhat.com>
9306 PR tree-optimization/84740
9307 * tree-switch-conversion.c (process_switch): Call build_constructors
9308 only if info.phi_count is non-zero.
9310 PR tree-optimization/84739
9311 * tree-tailcall.c (find_tail_calls): Check call arguments against
9312 DECL_ARGUMENTS (current_function_decl) rather than
9313 DECL_ARGUMENTS (func) when checking for tail recursion.
9315 2018-03-07 Jakub Jelinek <jakub@redhat.com>
9317 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
9318 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
9319 Volker Reichelt's entry and add entries for people that perform
9320 GCC fuzzy testing and report numerous bugs.
9322 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
9325 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
9326 readonly data in sdata, if that is disabled.
9327 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
9328 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
9329 -mreadonly-in-sdata option.
9331 2018-03-07 Martin Sebor <msebor@redhat.com>
9333 PR tree-optimization/84468
9334 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
9335 basic block when looking for nul assignment.
9337 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
9340 * except.h (output_function_exception_table): Adjust prototype.
9341 * except.c (output_function_exception_table): Remove FNNAME parameter
9342 and add SECTION parameter. Ouput one part of the table at a time.
9343 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
9344 the first part of the exception table and emit unwind directives.
9345 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
9346 (i386_pe_seh_cold_init): Likewise.
9347 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
9348 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
9349 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
9350 (ix86_output_call_insn): Emit a nop in one more case for SEH.
9351 * config/i386/winnt.c: Include except.h.
9352 (struct seh_frame_state): Add reg_offset, after_prologue and
9353 in_cold_section fields.
9354 (i386_pe_seh_end_prologue): Set seh->after_prologue.
9355 (i386_pe_seh_cold_init): New function.
9356 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
9357 to seh->in_cold_section.
9358 (seh_emit_push): Record the offset of the push.
9359 (seh_emit_save): Record the offet of the save.
9360 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
9361 Test seh->after_prologue to disregard the epilogue.
9362 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
9363 (i386_pe_end_cold_function): New function.
9365 2018-03-07 Jakub Jelinek <jakub@redhat.com>
9368 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
9369 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
9372 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
9374 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
9375 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
9378 * multiple_target.c: Include tree-inline.h and intl.h.
9379 (expand_target_clones): Diagnose and fail if node->definition and
9380 !tree_versionable_function_p (node->decl).
9382 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
9384 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
9386 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
9387 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
9388 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
9390 2018-03-06 Jakub Jelinek <jakub@redhat.com>
9393 * combine.c (try_combine): Use reg_or_subregno instead of handling
9394 just paradoxical SUBREGs and REGs.
9396 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
9398 * config/arc/arc.c (arc_finalize_pic): Remove function.
9399 (arc_must_save_register): We use single base PIC register, remove
9400 checks to save/restore the PIC register.
9401 (arc_expand_prologue): Likewise.
9402 * config/arc/arc-protos.h (arc_set_default_type_attributes):
9404 (arc_verify_short): Likewise.
9405 (arc_attr_type): Likewise.
9406 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
9407 (walk_stores): Likewise.
9408 (arc_address_cost): Make it static.
9409 (arc_verify_short): Likewise.
9410 (branch_dest): Likewise.
9411 (arc_attr_type): Likewise.
9412 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
9413 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
9414 (arc_final_prescan_insn): Remove inserting the nops due to
9415 hardware hazards. It is done in reorg step.
9416 (insn_length_variant_t): Remove.
9417 (insn_length_parameters_t): Likewise.
9418 (arc_insn_length_parameters): Likewise.
9419 (arc_get_insn_variants): Likewise.
9420 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
9422 2018-03-06 Jakub Jelinek <jakub@redhat.com>
9425 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
9428 PR tree-optimization/84687
9429 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
9431 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
9433 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9435 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
9436 Rename to ppc_speculation_barrier.
9437 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
9438 __builtin_ppc_speculation_barrier.
9440 2018-03-05 Jakub Jelinek <jakub@redhat.com>
9443 * combine.c (combine_simplify_rtx): Don't try to simplify if
9444 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
9447 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
9449 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
9450 to 32 bytes when compiling for POWER9.
9452 2018-03-05 Jakub Jelinek <jakub@redhat.com>
9455 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
9456 regparm >= 3 with no arg reg available also for calls with
9457 flag_force_indirect_call. Pass decl to ix86_function_regparm.
9460 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
9462 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
9464 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
9467 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
9469 2018-03-05 Richard Biener <rguenther@suse.de>
9471 PR tree-optimization/84486
9472 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
9473 When inserting a __builtin_assume_aligned call set the LHS
9474 SSA name alignment info accordingly.
9476 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
9478 PR tree-optimization/84114
9479 * config/aarch64/aarch64.c (aarch64_reassociation_width)
9480 Avoid reassociation of FLOAT_MODE addition.
9482 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
9484 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
9485 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
9486 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
9487 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
9488 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
9489 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
9490 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
9492 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
9493 __builtin_ia32_wbinvd): New builtins.
9494 (SPECIAL_ARGS2): New.
9495 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
9496 (SPECIAL_ARGS2): New.
9497 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
9498 (ix86_valid_target_attribute_inner_p): Ditto.
9499 (ix86_init_mmx_sse_builtins): Add special_args2.
9500 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
9501 TARGET_WBNOINVD_P): New.
9502 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
9503 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
9504 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
9505 * config/i386/immintrin.h (_wbinvd): New intrinsic.
9506 * config/i386/pconfigintrin.h: New file.
9507 * config/i386/wbnoinvdintrin.h: Ditto.
9508 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
9510 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
9512 2018-03-05 Richard Biener <rguenther@suse.de>
9514 PR tree-optimization/84670
9515 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
9517 (BB_VISITED_WITH_VISITED_SUCCS): New define.
9518 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
9519 (compute_antic_aux): Only assert the number of values in ANTIC_IN
9520 doesn't grow if all successors (recursively) were visited at least
9523 2018-03-05 Richard Biener <rguenther@suse.de>
9525 PR tree-optimization/84650
9526 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
9527 if executed in the loop pipeline.
9529 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
9531 * doc/configfiles.texi (Configuration Files): Move info about
9532 conditionalizing $target-protos.h to...
9533 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
9534 differs from $target-protos.h.
9536 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
9537 Chung-Ju Wu <jasonwucj@gmail.com>
9539 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
9540 * config/nds32/nds32-multiple.md (setmemsi): Define.
9541 * config/nds32/nds32-memory-manipulation.c
9542 (nds32_gen_dup_4_byte_to_word_value): New.
9543 (emit_setmem_word_loop): New.
9544 (emit_setmem_byte_loop): New.
9545 (nds32_expand_setmem_loop): New.
9546 (nds32_expand_setmem_loop_v3m): New.
9547 (nds32_expand_setmem_unroll): New.
9548 (nds32_expand_setmem): New.
9550 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
9551 Chung-Ju Wu <jasonwucj@gmail.com>
9553 * config/nds32/nds32-memory-manipulation.c
9554 (nds32_emit_load_store): New.
9555 (nds32_emit_post_inc_load_store): New.
9556 (nds32_emit_mem_move): New.
9557 (nds32_emit_mem_move_block): New.
9558 (nds32_expand_movmemsi_loop_unknown_size): New.
9559 (nds32_expand_movmemsi_loop_known_size): New.
9560 (nds32_expand_movmemsi_loop): New.
9561 (nds32_expand_movmemsi_unroll): New.
9562 (nds32_expand_movmemqi): Rename ...
9563 (nds32_expand_movmemsi): ... to this.
9564 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
9565 (movmemsi): ... to this.
9566 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
9567 (nds32_expand_movmemsi): ... to this.
9569 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
9570 Monk Chiang <sh.chiang04@gmail.com>
9571 Chung-Ju Wu <jasonwucj@gmail.com>
9573 * config/nds32/nds32-protos.h
9574 (nds32_expand_load_multiple): New arguments.
9575 (nds32_expand_store_multiple): Ditto.
9576 (nds32_valid_multiple_load_store): Rename ...
9577 (nds32_valid_multiple_load_store_p): ... to this.
9578 * config/nds32/nds32-memory-manipulation.c
9579 (nds32_expand_load_multiple): Refine implementation.
9580 (nds32_expand_store_multiple): Ditto.
9581 * config/nds32/nds32-multiple.md
9582 (load_multiple): Update nds32_expand_load_multiple interface.
9583 (store_multiple): Update nds32_expand_store_multiple interface.
9584 * config/nds32/nds32-predicates.c
9585 (nds32_valid_multiple_load_store): Rename ...
9586 (nds32_valid_multiple_load_store_p): ... to this and refine
9588 * config/nds32/predicates.md
9589 (nds32_load_multiple_and_update_address_operation): New predicate.
9590 (nds32_store_multiple_and_update_address_operation): New predicate.
9592 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
9593 Chung-Ju Wu <jasonwucj@gmail.com>
9595 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
9596 (combo): New attribute.
9597 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
9599 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
9601 * config/nds32/nds32.opt: Change -mcmodel= default value.
9603 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
9604 Monk Chiang <sh.chiang04@gmail.com>
9605 Chung-Ju Wu <jasonwucj@gmail.com>
9607 * config/nds32/constants.md (unspec_element): New enum.
9608 * config/nds32/constraints.md (Umw): New constraint.
9609 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
9610 * config/nds32/nds32-intrinsic.md: Likewise.
9611 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
9612 (nds32_valid_smw_lwm_base_p): New.
9613 (nds32_output_smw_single_word): New.
9614 (nds32_output_lmw_single_word): New.
9615 (nds32_expand_unaligned_load): New.
9616 (nds32_expand_unaligned_store): New.
9617 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
9618 (nds32_output_smw_single_word): Declare.
9619 (nds32_output_lmw_single_word): Declare.
9620 (nds32_expand_unaligned_load): Declare.
9621 (nds32_expand_unaligned_store): Declare.
9622 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
9623 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
9624 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
9625 NDS32_BUILTIN_UASTORE_DW.
9626 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
9629 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
9630 Kito Cheng <kito.cheng@gmail.com>
9631 Chung-Ju Wu <jasonwucj@gmail.com>
9633 * config/nds32/nds32-intrinsic.c
9634 (nds32_expand_builtin_null_ftype_reg): Delete.
9635 (nds32_expand_builtin_reg_ftype_imm): Ditto.
9636 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
9637 (nds32_read_argument): New.
9638 (nds32_legitimize_target): Ditto.
9639 (nds32_legitimize_argument): Ditto.
9640 (nds32_check_constant_argument): Ditto.
9641 (nds32_expand_unop_builtin): Ditto.
9642 (nds32_expand_unopimm_builtin): Ditto.
9643 (nds32_expand_binop_builtin): Ditto.
9644 (nds32_builtin_decl_impl): Ditto.
9645 (builtin_description): Ditto.
9646 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
9647 (nds32_init_builtins_impl): Ditto.
9648 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
9649 (nds32_builtin_decl): New.
9650 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
9651 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
9653 2018-03-02 Jeff Law <law@redhat.com>
9655 * reorg.c (stop_search_p): Handle DEBUG_INSN.
9656 (redundant_insn, fill_simple_delay_slots): Likewise.
9657 (fill_slots_from_thread): Likewise.
9658 * resource.c (mark_referenced_resources): Likewise.
9659 (mark_set_resources, find_dead_or_set_registers): Likewise.
9661 2018-03-02 Jakub Jelinek <jakub@redhat.com>
9663 * substring-locations.h (format_warning_va): Formatting fix for
9665 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
9667 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
9668 * substring-locations.c: Include intl.h.
9669 (format_warning_va): Turned into small wrapper around
9670 format_warning_n_va, renamed to ...
9671 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
9672 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
9674 (format_warning_at_substring_n): New function.
9675 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
9676 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
9677 format_warning_at_substring with just a shorter name instead of
9678 const function pointer.
9679 (fmtwarn_n): New function.
9680 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
9681 appropriate, get rid of all the fmtstr temporaries, move conditionals
9682 with G_() wrapped string literals directly into fmtwarn arguments,
9683 cast dir.len to (int), formatting fixes.
9685 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
9687 * doc/invoke.texi: Remove "Cilk Plus" references.
9689 2018-03-02 Jakub Jelinek <jakub@redhat.com>
9690 Richard Biener <rguenther@suse.de>
9693 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
9694 for error or warning attributes if CALL_FROM_THUNK_P is set.
9697 2018-03-02 Jakub Jelinek <jakub@redhat.com>
9700 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
9701 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
9704 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
9705 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
9707 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
9708 instead of -1U in last predictors element's probability member.
9710 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
9713 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
9714 arguments if they are comparable.
9716 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
9718 PR tree-optimization/84634
9719 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
9720 masks and masked_loop_p with a single loop_masks, making sure it's
9721 null for bb vectorization.
9723 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
9725 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
9726 (vect_analyze_data_ref_access): Use loop->safe_len rather than
9727 loop->force_vectorize to check whether there is no alias.
9729 2018-03-02 Jakub Jelinek <jakub@redhat.com>
9732 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
9734 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
9736 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
9737 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
9738 instead of a loop around prev_real_insn.
9739 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
9743 * config/i386/i386.c (ix86_print_operand): Use conditional
9744 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
9747 2018-03-02 Richard Biener <rguenther@suse.de>
9749 PR tree-optimization/84427
9750 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
9751 (bitmap_set_subtract_values): Rewrite to handle multiple
9754 (prune_clobbered_mems): Likewise.
9755 (phi_translate): Take edge instead of pred/phiblock.
9756 (phi_translate_1): Likewise.
9757 (phi_translate_set): Likewise. Insert all translated
9758 exprs for a value into the set, keeping possibly multiple
9759 expressions per value.
9760 (compute_antic_aux): Adjust for phi_translate changes.
9761 When intersecting union the expressions and prune those
9762 not in the final value set, keeping possibly multiple
9763 expressions per value. Do not use value-insertion
9764 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
9765 all expressions. Add verification that the value-sets
9766 only shrink during iteration.
9767 (compute_partial_antic_aux): Adjust for the phi_translate changes.
9768 (do_pre_regular_insertion): Likewise.
9769 (do_pre_partial_partial_insertion): Likewise.
9771 2018-03-02 Richard Biener <rguenther@suse.de>
9774 * config/darwin.c (saved_debug_info_level): New static global.
9775 (darwin_asm_lto_start): Disable debug info generation for LTO out.
9776 (darwin_asm_lto_end): Restore debug info generation settings.
9778 2018-03-01 Martin Liska <mliska@suse.cz>
9781 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
9784 2018-03-01 Richard Biener <rguenther@suse.de>
9787 * dwarf2out.c (gen_variable_die): Properly handle late VLA
9788 type annotation with LTO when debug was disabled at compile-time.
9790 2018-03-01 Matthew Fortune <mfortune@gmail.com>
9792 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
9794 (mips_final_postscan_insn): Likewise.
9796 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
9798 PR rtl-optimization/84528
9799 * alias.c (init_alias_target): Add commentary.
9800 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
9801 a unique base value if the frame pointer is not eliminated
9802 to the stack pointer.
9804 2018-03-01 Tom de Vries <tom@codesourcery.com>
9806 PR rtl-optimization/83327
9807 * lra-int.h (hard_regs_spilled_into): Declare.
9808 * lra.c (hard_regs_spilled_into): Define.
9809 (init_reg_info): Init hard_regs_spilled_into.
9810 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
9811 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
9812 (process_bb_lives): Handle hard_regs_spilled_into.
9813 (lra_create_live_ranges_1): Before doing liveness propagation, clear
9814 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
9816 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
9818 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
9819 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
9820 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
9821 * config/rs6000/aix72.h: New file.
9823 2018-02-28 Jakub Jelinek <jakub@redhat.com>
9825 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
9826 instead of warning_at with conditional singular and plural messages
9830 * stor-layout.c (update_alignment_for_field): For
9831 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
9832 && !DECL_PACKED (field), do the alignment update, just use
9833 only desired_align instead of MAX (type_align, desired_align)
9835 (place_field): Don't do known_align < desired_align handling
9836 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
9837 is non-NULL, instead do it after rli->prev_field handling and
9838 only if not within a bitfield word. For DECL_PACKED (field)
9839 use type_align of BITS_PER_UNIT.
9841 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
9843 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
9844 superfluous parentheses and trailing spaces.
9846 2018-02-28 Richard Biener <rguenther@suse.de>
9848 PR tree-optimization/84584
9849 * graphite-scop-detection.c (scop_detection::add_scop): Discard
9850 SCoPs with fake exit edge.
9852 2018-02-28 Martin Liska <mliska@suse.cz>
9855 * timevar.c (timer::print): Fix format to properly print 100%
9858 2018-02-28 Richard Biener <rguenther@suse.de>
9861 * genmatch.c (capture_info::walk_match): Do not mark
9862 captured expressions without operands as expr_p given
9863 they act more like predicates and should be subject to
9864 "lost tail" side-effect preserving.
9866 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
9868 PR rtl-optimization/81611
9869 * auto-inc-dec.c (attempt_change): Move dead note from
9870 mem_insn if it's the next use of regno
9871 (find_address): Take address use of reg holding
9872 non-incremented value. Add parm to limit search to the named
9874 (merge_in_block): Attempt to use a mem insn that is the next
9875 use of the original regno.
9877 2018-02-27 Martin Sebor <msebor@redhat.com>
9880 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
9881 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
9883 2018-02-27 Martin Sebor <msebor@redhat.com>
9885 PR translation/84207
9886 * diagnostic-core.h (warning_n, error_n, inform_n): Change
9887 n argument to unsigned HOST_WIDE_INT.
9888 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
9889 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
9890 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
9891 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
9893 2018-02-27 Richard Biener <rguenther@suse.de>
9895 PR tree-optimization/84512
9896 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
9897 Do not use the estimate returned from record_stmt_cost for
9898 the scalar iteration cost but sum properly using add_stmt_cost.
9900 2018-02-27 Richard Biener <rguenther@suse.de>
9902 PR tree-optimization/84466
9903 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
9904 Adjust last change to less strictly validate use operands.
9906 2018-02-27 Martin Liska <mliska@suse.cz>
9908 PR gcov-profile/84548
9909 * gcov.c (process_file): Allow partial overlap and consider it
9910 also as group functions.
9911 (output_lines): Properly calculate range of lines for a group.
9913 2018-02-27 Martin Liska <mliska@suse.cz>
9915 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
9916 'ggc' suffixes. Change first column width.
9917 (timer::print): Fix formatting of the column.
9919 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
9921 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
9922 preserve inline entry blocks for the sake of debug inline
9923 entry point markers alone.
9924 (remove_unused_locals): Suggest in comments a better place to
9925 force the preservation of inline entry blocks that are
9926 otherwise unused, but do not preserve them.
9928 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
9930 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
9932 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
9935 * config/i386/constraints.md (Bs): Replace
9936 ix86_indirect_branch_register with
9937 TARGET_INDIRECT_BRANCH_REGISTER.
9939 * config/i386/i386.md (indirect_jump): Likewise.
9940 (tablejump): Likewise.
9941 (*sibcall_memory): Likewise.
9942 (*sibcall_value_memory): Likewise.
9943 Peepholes of indirect call and jump via memory: Likewise.
9944 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
9945 (*sibcall_value_GOT_32): Likewise.
9946 * config/i386/predicates.md (indirect_branch_operand): Likewise.
9947 (GOT_memory_operand): Likewise.
9948 (call_insn_operand): Likewise.
9949 (sibcall_insn_operand): Likewise.
9950 (GOT32_symbol_operand): Likewise.
9951 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
9953 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
9955 PR rtl-optimization/83496
9956 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
9957 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
9958 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
9959 redundant insn, if any.
9960 (relax_delay_slots): Likewise.
9961 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
9963 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
9965 PR tree-optimization/83965
9966 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
9967 that grouped statements are part of a reduction chain. Return
9968 true if the statement is not marked as a reduction itself but
9970 (vect_recog_dot_prod_pattern): Don't check whether the statement
9971 is part of a group here.
9972 (vect_recog_sad_pattern): Likewise.
9973 (vect_recog_widen_sum_pattern): Likewise.
9975 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
9978 * final.c (rest_of_clean_state): Also look for calls inside sequences.
9980 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
9983 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
9985 (ix86_output_indirect_function_return): New prototype.
9986 (ix86_split_simple_return_pop_internal): Likewise.
9987 * config/i386/i386.c (indirect_return_via_cx): New.
9988 (indirect_return_via_cx_bnd): Likewise.
9989 (indirect_thunk_name): Handle return va CX_REG.
9990 (output_indirect_thunk_function): Create alias for
9991 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
9992 (ix86_output_indirect_jmp): Remove the bool argument.
9993 (ix86_output_indirect_function_return): New function.
9994 (ix86_split_simple_return_pop_internal): Likewise.
9995 * config/i386/i386.md (*indirect_jump): Don't pass false
9996 to ix86_output_indirect_jmp.
9997 (*tablejump_1): Likewise.
9998 (simple_return_pop_internal): Change it to define_insn_and_split.
9999 Call ix86_split_simple_return_pop_internal to split it for
10000 -mfunction-return=.
10001 (simple_return_indirect_internal): Call
10002 ix86_output_indirect_function_return instead of
10003 ix86_output_indirect_jmp.
10005 2018-02-26 Jakub Jelinek <jakub@redhat.com>
10008 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
10009 memset and value initialization afterwards.
10011 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
10013 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
10015 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
10018 * common/config/aarch64/aarch64-common.c
10019 (aarch_option_optimization_table[]): Switch
10020 off fomit-frame-pointer
10022 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
10023 Chung-Ju Wu <jasonwucj@gmail.com>
10025 * config/nds32/nds32-multiple.md (load_multiple): Disallow
10027 (store_multiple): Ditto.
10029 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
10031 * config.gcc: Add --with-cpu support for nds32 target.
10032 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
10033 * config/nds32/nds32.opt: Add -mcpu= option.
10035 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
10037 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
10038 isel=yes): Warn for these deprecated options.
10040 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
10042 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
10043 ISA_2_5_MASKS_EMBEDDED.
10045 2018-02-23 Jakub Jelinek <jakub@redhat.com>
10047 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
10048 p->max as pointers rather than using iterative_hash_expr.
10050 2018-02-23 Carl Love <cel@us.ibm.com>
10052 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
10053 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
10055 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
10056 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
10057 P8V_BUILTIN_VEC_VUNSIGNED2.
10059 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
10062 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
10063 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
10065 * lra-constraints.c (curr_insn_transform): Set up
10066 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
10068 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
10069 * lra-eliminations.c (spill_pseudos): Ditto.
10070 (process_insn_for_elimination): Ditto.
10071 * lra-lives.c (reg_early_clobber_p): Use the new macros.
10072 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
10073 LRA_NON_CLOBBERED_ALT.
10075 2018-02-22 Martin Sebor <msebor@redhat.com>
10077 PR tree-optimization/84480
10078 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
10079 to maybe_diag_stxncpy_trunc. Call it.
10080 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
10081 from gimple_fold_builtin_strcpy. Print inlining stack.
10082 (handle_builtin_stxncpy): Print inlining stack.
10083 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
10085 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
10088 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
10089 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
10090 and -fcheck-pointer-bounds are used together.
10091 (indirect_thunk_prefix): New enum.
10092 (indirect_thunk_need_prefix): New function.
10093 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
10094 "_nt" instead of "_bnd" for NOTRACK prefix.
10095 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
10096 (output_indirect_thunk_function): Likewise.
10098 (ix86_code_end): Update output_indirect_thunk_function calls.
10099 (ix86_output_indirect_branch_via_reg): Replace
10100 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
10101 (ix86_output_indirect_branch_via_push): Likewise.
10102 (ix86_output_function_return): Likewise.
10103 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
10104 incompatible with -fcf-protection=branch and
10105 -fcheck-pointer-bounds.
10107 2018-02-22 Steve Ellcey <sellcey@cavium.com>
10110 * config/aarch64/aarch64.c (aarch64_print_address_internal):
10111 Change gcc_assert call to output_operand_lossage.
10113 2018-02-22 Steve Ellcey <sellcey@cavium.com>
10115 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
10117 2018-02-22 DJ Delorie <dj@redhat.com>
10118 Sebastian Perta <sebastian.perta@renesas.com>
10119 Oleg Endo <olegendo@gcc.gnu.org>
10121 * config/rx/rx.c (rx_rtx_costs): New function.
10122 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
10124 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
10126 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
10128 2018-02-22 Martin Liska <mliska@suse.cz>
10131 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
10132 Add "native" as a possible value.
10134 2018-02-22 Martin Liska <mliska@suse.cz>
10137 * config/i386/i386.c (ix86_option_override_internal):
10138 Add "native" as a possible value for -march and -mtune.
10140 2018-02-22 Jakub Jelinek <jakub@redhat.com>
10143 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
10144 to all type variants.
10146 PR tree-optimization/84503
10147 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
10148 width as info->bitpos + info->bitsize - start.
10149 (merged_store_group::merge_overlapping): Simplify width computation.
10150 (check_no_overlap): New function.
10151 (imm_store_chain_info::try_coalesce_bswap): Compute expected
10152 start + width and last_order of the group, fail if check_no_overlap
10154 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
10155 to group if check_no_overlap fails.
10157 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
10159 * config/rs6000/altivec.md: Delete contraint arguments to
10160 define_expand, define_split, and define_peephole2, and in
10161 define_insn_and_split if always unused.
10162 * config/rs6000/darwin.md: Ditto.
10163 * config/rs6000/dfp.md: Ditto.
10164 * config/rs6000/rs6000.md: Ditto.
10165 * config/rs6000/sync.md: Ditto.
10166 * config/rs6000/vector.md: Ditto.
10167 * config/rs6000/vsx.md: Ditto.
10169 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
10171 * config/rs6000/altivec.md: Write output control strings as braced
10172 blocks instead of double-quoted strings.
10173 * config/rs6000/darwin.md: Ditto.
10174 * config/rs6000/rs6000.md: Ditto.
10175 * config/rs6000/vector.md: Ditto.
10176 * config/rs6000/vsx.md: Ditto.
10178 2018-02-21 Jason Merrill <jason@redhat.com>
10180 PR c++/84314 - ICE with templates and fastcall attribute.
10181 * attribs.c (build_type_attribute_qual_variant): Remove assert.
10183 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
10185 * ipa-cp.c (determine_versionability): Fix comment typos.
10187 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
10190 * ipa-cp.c (determine_versionability): Do not version functions caling
10193 2018-02-21 Martin Liska <mliska@suse.cz>
10196 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
10197 Add "native" as a possible value.
10198 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
10199 the macro when native cpu detection is available.
10201 2018-02-21 Martin Liska <mliska@suse.cz>
10204 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
10205 Add "native" as a possible value.
10206 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
10207 when native cpu detection is available.
10209 2018-02-21 Jakub Jelinek <jakub@redhat.com>
10210 Martin Sebor <msebor@redhat.com>
10212 PR tree-optimization/84478
10213 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
10215 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
10216 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
10217 support which is conservatively correct, for 2 only stay conservative
10218 for maxlen. Formatting and comment capitalization fixes. Add STRICT
10219 argument to the 2 argument get_range_strlen, adjust 6 arg
10220 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
10222 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
10223 (gimple_fold_builtin_strlen): Pass true as last argument to
10226 2018-02-20 Martin Sebor <msebor@redhat.com>
10228 PR middle-end/84095
10229 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
10230 (builtin_memref::set_base_and_offset): Same. Handle inner references.
10231 (builtin_memref::builtin_memref): Factor out parts into
10232 set_base_and_offset and call it.
10234 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
10236 PR middle-end/84406
10237 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
10238 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
10239 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
10240 search at the associated MODE_INT.
10242 2018-02-20 Jeff Law <law@redhat.com>
10244 PR middle-end/82123
10245 PR tree-optimization/81592
10246 PR middle-end/79257
10247 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
10248 for range data rather than using global data.
10249 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
10250 range data rather than using global data.
10251 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
10252 pass it to children as needed.
10253 (struct directive::fmtresult): Similarly.
10254 (struct directive::set_width): Similarly.
10255 (struct directive::set_precision): Similarly.
10256 (format_integer, format_directive, parse_directive): Similarly.
10257 (format_none): Accept unnamed vr_values parameter.
10258 (format_percent, format_floating, format_character): Similarly.
10259 (format_string, format_plain): Similarly.
10260 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
10261 the EVRP range analyzer for range data rather than using global data.
10262 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
10263 gimple-ssa-evrp-analyze.h
10264 (class sprintf_dom_walker): Add after_dom_children member function.
10265 Add evrp_range_analyzer member.
10266 (sprintf_dom_walker::before_dom_children): Call into the EVRP
10267 range analyzer as needed.
10268 (sprintf_dom_walker::after_dom_children): New member function.
10269 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
10271 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
10272 (evrp_range_analyzer::pop_to_marker): Likewise.
10274 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
10276 PR tree-optimization/84419
10277 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
10278 with the required type if its current type is compatible but
10281 2018-02-20 Jakub Jelinek <jakub@redhat.com>
10283 PR middle-end/82004
10284 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
10285 after vectorization.
10287 2018-02-20 Martin Liska <mliska@suse.cz>
10290 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
10291 possible values if we don't have a hint.
10293 2018-02-20 Martin Liska <mliska@suse.cz>
10297 * final.c (shorten_branches): Build align_tab array with one
10299 * opts.c (finish_options): Add alignment option limit check.
10300 (MAX_CODE_ALIGN): Likewise.
10301 (MAX_CODE_ALIGN_VALUE): Likewise.
10302 * doc/invoke.texi: Document maximum allowed option value for
10303 all -falign-* options.
10305 2018-02-19 Jakub Jelinek <jakub@redhat.com>
10308 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
10309 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
10310 * var-tracking.c (emit_note_insn_var_location): Remove all references
10311 to NOTE_INSN_CALL_ARG_LOCATION.
10312 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
10313 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
10314 Use copy_rtx_if_shared.
10315 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
10316 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
10317 (dwarf2out_var_location): Remove handling of
10318 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
10320 * final.c (final_scan_insn): Remove all references to
10321 NOTE_INSN_CALL_ARG_LOCATION.
10322 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
10323 before dumping final insns.
10324 * except.c (emit_note_eh_region_end): Remove all references to
10325 NOTE_INSN_CALL_ARG_LOCATION.
10326 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
10327 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
10328 * config/arc/arc.c (hwloop_optimize): Likewise.
10329 * config/arm/arm.c (create_fix_barrier): Likewise.
10330 * config/s390/s390.c (s390_chunkify_start): Likewise.
10331 * config/sh/sh.c (find_barrier): Likewise.
10332 * config/i386/i386.c (rest_of_insert_endbranch,
10333 ix86_seh_fixup_eh_fallthru): Likewise.
10334 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
10335 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
10336 * config/frv/frv.c (frv_function_prologue): Likewise.
10337 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
10339 (note_outside_basic_block_p): Remove all references to
10340 NOTE_INSN_CALL_ARG_LOCATION.
10341 * gengtype.c (adjust_field_rtx_def): Likewise.
10342 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
10344 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
10345 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
10348 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
10351 PR tree-optimization/84452
10352 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
10353 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
10356 2018-02-19 Martin Liska <mliska@suse.cz>
10359 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
10361 2018-02-19 Martin Liska <mliska@suse.cz>
10362 Richard Sandiford <richard.sandiford@linaro.org>
10364 PR tree-optimization/82491
10365 * gimple-fold.c (get_base_constructor): Make earlier bail out
10368 2018-02-19 Carl Love <cel@us.ibm.com>
10370 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
10371 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
10373 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
10374 P8V_BUILTIN_VEC_NEG.
10376 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
10378 * config/rl78/rl78.md (movdf): New define expand.
10380 2018-02-19 Martin Liska <mliska@suse.cz>
10383 * doc/invoke.texi: Fix typo.
10384 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
10386 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
10388 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
10389 handle rs6000_single_float and rs6000_double_float specially for
10392 2018-02-16 Jeff Law <law@redhat.com>
10394 * config/rx/rx.c (add_pop_cfi_notes): New function.;
10395 (pop_regs): Use it.
10397 2018-02-16 Jakub Jelinek <jakub@redhat.com>
10400 * ipa-inline.c (inline_small_functions): Fix a typo.
10402 2018-02-16 Nathan Sidwell <nathan@acm.org>
10404 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
10406 2018-02-16 Carl Love <cel@us.ibm.com>
10408 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
10409 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
10410 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
10411 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
10412 expansion to P8V_BUILTIN_VEC_FLOAT2.
10414 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
10416 PR rtl-optimization/70023
10417 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
10418 src_regno into account.
10420 2018-02-16 Carl Love <cel@us.ibm.com>
10422 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
10423 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
10424 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
10425 * config/rs6000/rs6000.c: Remove case statements for
10426 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
10427 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
10428 and P9V_BUILTIN_VEC_VINSERT4B.
10429 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
10430 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
10431 * config/rs6000/vsx.md:
10432 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
10435 2018-02-16 Carl Love <cel@us.ibm.com>
10437 * config/rs6000/altivec.h: Add builtin names vec_extract4b
10439 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
10441 * config/rs6000/rs6000-c.c: Add the definitions for
10442 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
10443 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
10444 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
10445 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
10446 definition for insert4b and define insn *insert3b_internal.
10447 * doc/extend.texi: Add documentation for vec_extract4b.
10449 2018-02-16 Nathan Sidwell <nathan@acm.org>
10451 * doc/extend.texi (Backwards Compatibility): Mention friend
10452 injection. Note for-scope is deprecated.
10453 * doc/invoke.texi (-ffriend-injection): Deprecate.
10455 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
10457 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
10458 that moved to I2, also allow destinations that are a paradoxical
10459 subreg (instead of a normal reg).
10461 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
10464 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
10467 2018-02-16 Richard Biener <rguenther@suse.de>
10469 PR tree-optimization/84037
10470 PR tree-optimization/84016
10472 * config/i386/i386.c (ix86_builtin_vectorization_cost):
10473 Adjust vec_construct for the fact we need additional higher latency
10474 128bit inserts for AVX256 and AVX512 vector builds.
10475 (ix86_add_stmt_cost): Scale vector construction cost for
10478 2018-02-16 Richard Biener <rguenther@suse.de>
10480 PR tree-optimization/84417
10481 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
10482 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
10483 (non_rewritable_lvalue_p): Likewise, use poly-ints.
10485 2018-02-16 Martin Liska <mliska@suse.cz>
10488 * internal-fn.def (ASAN_CHECK): Set proper flags.
10489 (ASAN_MARK): Likewise.
10491 2018-02-16 Julia Koval <julia.koval@intel.com>
10493 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
10494 from PTA_CANNONLAKE.
10496 2018-02-16 Jakub Jelinek <jakub@redhat.com>
10499 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
10500 Use ++iter rather than iter++ for std::list iterators.
10501 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
10502 defer deleting them until all nodes in the forest are processed. Do
10503 free even leaf nodes. Change to_process into auto_vec.
10506 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
10507 * vec.h (vec_default_construct): Use memset instead of placement new
10508 if BROKEN_VALUE_INITIALIZATION is defined.
10509 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
10510 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
10513 PR rtl-optimization/83723
10514 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
10515 * lra.c (lra_substitute_pseudo): Likewise. If true, use
10516 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
10518 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
10520 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
10522 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
10524 PR rtl-optimization/81443
10525 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
10526 from inner REGs to paradoxical SUBREGs.
10528 2018-02-16 Richard Biener <rguenther@suse.de>
10530 PR tree-optimization/84399
10531 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
10532 For operands we can analyze at their definition make sure we can
10533 analyze them at each use as well.
10535 2018-02-16 Richard Biener <rguenther@suse.de>
10537 PR tree-optimization/84190
10538 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
10539 volatile accesses if the decl isn't volatile.
10541 2018-02-15 Jason Merrill <jason@redhat.com>
10543 PR c++/84314 - ICE with templates and fastcall attribute.
10544 * attribs.c (build_type_attribute_qual_variant): Don't clobber
10545 TYPE_CANONICAL on an existing type.
10547 2018-02-15 Jakub Jelinek <jakub@redhat.com>
10549 PR tree-optimization/84383
10550 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
10551 dstoff nor call operand_equal_p if dstbase is NULL.
10553 PR tree-optimization/84334
10554 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
10555 also a CONSTANT_CLASS_P, punt.
10557 2018-02-14 Jim Wilson <jimw@sifive.com>
10559 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
10560 first SMALL_OPERAND check. New local min_second_step. Move assert
10561 to where locals are set. Add TARGET_RVC support.
10562 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
10564 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
10566 * doc/invoke.texi: Correct -Wformat-overflow code sample.
10568 2018-02-14 Martin Sebor <msebor@redhat.com>
10570 PR tree-optimization/83698
10571 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
10572 arrays constrain the offset range to their bounds.
10573 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
10574 (builtin_access::overlap): Avoid setting the size of overlap if it's
10576 (maybe_diag_overlap): Also consider arrays when deciding what values
10577 of offsets to include in diagnostics.
10579 2018-02-14 Martin Sebor <msebor@redhat.com>
10582 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
10583 that correspond to the kind of a declaration.
10585 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
10588 * config/pa/pa.md: Load address of PIC label using the linkage table
10589 if the label is nonlocal.
10591 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
10593 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
10594 warning message if user requests -maltivec=be.
10595 * doc/invoke.texi: Document deprecation of -maltivec=be.
10597 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
10600 * config/rs6000/rs6000-c.c: Update definitions for
10601 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
10602 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
10604 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
10607 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
10608 add _get_ssp intrinsics. Remove argument from
10609 __builtin_ia32_rdssp[d|q].
10610 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
10611 * config/i386/i386-builtin.def: Remove argument from
10612 __builtin_ia32_rdssp[d|q].
10613 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
10614 ix86_expand_special_args_builtin for _rdssp[d|q].
10615 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
10616 Clear register before usage.
10617 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
10618 Add documentation for new _get_ssp and _inc_ssp intrinsics.
10620 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
10622 PR tree-optimization/84357
10623 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
10624 operand 1 of an ARRAY_REF too.
10626 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
10629 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
10630 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
10632 (set_of_reg): New struct.
10633 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
10634 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
10635 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
10637 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
10638 Split into bitclr, bitset, bitinvert patterns if appropriate.
10639 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
10640 use rx_fuse_in_memory_bitop.
10641 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
10642 to named insn, correct maximum insn length.
10644 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
10647 * machmode.def: Define a complex mode for PARTIAL_INT.
10648 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
10650 * doc/rtl.texi: Document CSPImode.
10651 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
10653 (msp430_hard_regno_nregs_with_padding): Likewise.
10655 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
10658 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
10660 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
10662 PR rtl-optimization/84169
10663 * combine.c (try_combine): New variable split_i2i3. Set it to true if
10664 we generated a parallel as new i3 and we split that to new i2 and i3
10665 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
10666 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
10667 those to i2, not i1. Partially rewrite this scan code.
10669 2018-02-13 Jakub Jelinek <jakub@redhat.com>
10672 * stor-layout.c (place_field): For variable length fields, adjust
10673 offset_align afterwards not just based on the field's alignment,
10674 but also on the size.
10676 PR middle-end/84309
10677 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
10678 of exps and logs in the use_exp2 case.
10680 2018-02-13 Jeff Law <law@redhat.com>
10682 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
10683 entry for "vector".
10685 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
10688 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
10692 * common.opt (gas-loc-support, gas-locview-support): New.
10693 (ginline-points, ginternal-reset-location-views): New.
10694 * doc/invoke.texi: Document them. Use @itemx where intended.
10695 (gvariable-location-views): Adjust.
10696 * target.def (reset_location_view): New.
10697 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
10698 (TARGET_RESET_LOCATION_VIEW): New.
10699 * doc/tm.texi: Rebuilt.
10700 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
10701 (dwarf2out_default_as_locview_support): New.
10702 (output_asm_line_debug_info): Use option variables.
10703 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
10704 (output_loc_list): Likewise.
10705 (add_high_low_attributes): Check option variables.
10706 Don't output entry view attribute in strict mode.
10707 (gen_inlined_subroutine_die): Check option variables.
10708 (dwarf2out_inline_entry): Likewise.
10709 (init_sections_and_labels): Likewise.
10710 (dwarf2out_early_finish): Likewise.
10711 (maybe_reset_location_view): New, from...
10712 (dwarf2out_var_location): ... here. Call it.
10713 * debug.h (dwarf2out_default_as_loc_support): Declare.
10714 (dwarf2out_default_as_locview_support): Declare.
10715 * hooks.c (hook_int_rtx_insn_0): New.
10716 * hooks.h (hook_int_rtx_insn_0): Declare.
10717 * toplev.c (process_options): Take -gas-loc-support and
10718 -gas-locview-support from dwarf2out. Enable
10719 -gvariable-location-views by default only with locview
10720 assembler support. Enable -ginternal-reset-location-views by
10721 default only if the target defines the corresponding hook.
10722 Enable -ginline-points by default if location views are
10723 enabled; force it disabled if statement frontiers are
10725 * tree-inline.c (expand_call_inline): Check option variables.
10726 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
10728 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
10730 PR tree-optimization/84321
10731 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
10732 handling. Also check whether the anti-range contains any values
10733 that satisfy the mask; switch to a VR_RANGE if not.
10735 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
10738 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
10740 2018-02-13 Martin Jambor <mjambor@suse.cz>
10743 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
10744 of call statements, also set location of a load to a temporary.
10746 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
10748 * config/rl78/rl78.c (add_vector_labels): New function.
10749 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
10750 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
10751 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
10752 which checks that no arguments are passed.
10753 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
10754 * doc/extend.texi: Documentation for the new attribute.
10756 2018-02-13 Andreas Schwab <schwab@suse.de>
10758 * config/riscv/linux.h (CPP_SPEC): Define.
10760 2018-02-13 Jakub Jelinek <jakub@redhat.com>
10763 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
10764 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
10765 OPTION_MASK_ISA_AES as first argument to def_builtin_const
10766 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
10767 instead of OPTION_MASK_ISA_PCLMUL as first argument to
10768 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
10769 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
10770 temporarily for AES and PCLMUL builtins.
10772 PR tree-optimization/84339
10773 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
10774 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
10777 PR middle-end/84309
10778 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
10779 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
10780 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
10782 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
10784 * omp-simd-clone.h: New file.
10785 * omp-simd-clone.c: Include omp-simd-clone.h.
10786 (expand_simd_clones): No longer static.
10787 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
10788 cgraph.h and omp-simd-clone.h.
10789 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
10790 (vect_recog_widen_shift_pattern): Formatting fix.
10791 (vect_pattern_recog_1): Don't check optab for calls.
10794 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
10795 operands[2] into a REG before using gen_lowpart on it.
10797 2018-02-12 Jeff Law <law@redhat.com>
10800 * config/sh/sh.c (find_barrier): Consider a sibling call
10803 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
10804 successfully back substituting a reg.
10806 2018-02-12 Richard Biener <rguenther@suse.de>
10808 PR tree-optimization/84037
10809 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
10810 parameter, move visited init to caller.
10811 (vect_slp_analyze_operations): Separate cost from validity
10812 check, initialize visited once for all instances.
10813 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
10815 * tree-vect-stmts.c (vect_model_simple_cost): Make early
10817 (vect_model_promotion_demotion_cost): Likewise.
10818 (vectorizable_bswap): Guard cost modeling with !slp_node
10819 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
10821 (vectorizable_call): Likewise.
10822 (vectorizable_conversion): Likewise.
10823 (vectorizable_assignment): Likewise.
10824 (vectorizable_shift): Likewise.
10825 (vectorizable_operation): Likewise.
10826 (vectorizable_store): Likewise.
10827 (vectorizable_load): Likewise.
10828 (vectorizable_condition): Likewise.
10829 (vectorizable_comparison): Likewise.
10831 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
10834 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
10835 (ASAN_MARK): Fix fnspec to account for return value, change pointer
10836 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
10838 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
10840 PR middle-end/83665
10841 * params.def (inline-min-speedup): Increase from 8 to 15.
10842 (max-inline-insns-auto): Decrease from 40 to 30.
10843 * ipa-split.c (consider_split): Add some buffer for function to
10844 be considered inlining candidate.
10845 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
10848 2018-02-12 Richard Biener <rguenther@suse.de>
10850 PR tree-optimization/84037
10851 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
10852 matched stmts if we cannot swap the non-matched ones.
10854 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
10856 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
10857 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
10858 _mm_maskz_scalef_round_ss): New intrinsics.
10859 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
10860 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
10861 __builtin_ia32_scalefss_round): Remove.
10862 (__builtin_ia32_scalefsd_mask_round,
10863 __builtin_ia32_scalefss_mask_round): New intrinsics.
10864 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
10865 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
10866 ((match_operand:VF_128 2 "<round_nimm_predicate>"
10867 "<round_constraint>")): Changed to ...
10868 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
10869 "<round_scalar_constraint>")): ... this.
10870 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
10871 %0, %1, %2<round_op3>}"): Changed to ...
10872 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
10873 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
10874 %2<round_scalar_mask_op3>}"): ... this.
10875 * config/i386/subst.md (round_scalar_nimm_predicate): New.
10877 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
10879 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
10880 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
10881 (_mm_maskz_sqrt_round_ss): New intrinsics.
10882 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
10883 (__builtin_ia32_sqrtsd_mask_round)
10884 (__builtin_ia32_sqrtss_mask_round): New builtins.
10885 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
10886 (__builtin_ia32_sqrtss_round): Remove.
10887 (__builtin_ia32_sqrtsd_mask_round)
10888 (__builtin_ia32_sqrtss_mask_round): New builtins.
10889 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
10890 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
10891 ((match_operand:VF_128 1 "vector_operand"
10892 "xBm,<round_constraint>")): Changed to ...
10893 ((match_operand:VF_128 1 "vector_operand"
10894 "xBm,<round_scalar_constraint>")): ... this.
10895 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
10896 %0, %2, %<iptr>1<round_op3>}): Changed to ...
10897 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
10898 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
10899 %<iptr>1<round_scalar_mask_op3>}): ... this.
10900 ((set_attr "prefix" "<round_prefix>")): Changed to ...
10901 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
10903 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
10906 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
10907 Cast vec_cmpeq result to correct type.
10908 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
10909 Cast vec_cmpgt result to correct type.
10911 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
10913 * final.c (final_scan_insn_1): Renamed from...
10914 (final_scan_insn): ... this. New wrapper, to recover
10915 seen from the outermost call in recursive ones.
10916 * config/sparc/sparc.c (output_return): Drop seen from call.
10917 (output_sibcall): Likewise.
10918 * config/visium/visium.c (output_branch): Likewise.
10920 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
10922 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
10925 2018-02-10 Alan Modra <amodra@gmail.com>
10928 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
10929 Specify LR as an input.
10931 2018-02-10 Jakub Jelinek <jakub@redhat.com>
10934 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
10935 remove_member_access_dummy_vars): New functions.
10936 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
10937 lower_omp_1, execute_lower_omp): Use them.
10939 PR rtl-optimization/84308
10940 * shrink-wrap.c (spread_components): Release todo vector.
10942 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
10944 PR rtl-optimization/57193
10945 * ira-color.c (struct allocno_color_data): Add member
10946 conflict_allocno_hard_prefs.
10947 (update_conflict_allocno_hard_prefs): New.
10948 (bucket_allocno_compare_func): Add a preference based on
10949 conflict_allocno_hard_prefs.
10950 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
10951 (color_allocnos): Remove a dead code. Initiate
10952 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
10954 2018-02-09 Jakub Jelinek <jakub@redhat.com>
10957 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
10958 constraint from =wa to wa. Avoid a subreg on the output operand,
10959 instead use a pseudo and subreg it in a move.
10960 (p9_xxbrd_<mode>): Changed to ...
10961 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
10962 (p9_xxbrd_v2df): New expander.
10963 (p9_xxbrw_<mode>): Changed to ...
10964 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
10965 (p9_xxbrw_v4sf): New expander.
10967 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
10969 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
10971 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
10974 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
10975 multiply in 32-bit mode.
10976 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
10977 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
10980 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
10982 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
10983 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
10984 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
10985 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
10987 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
10990 * dwarf2out.c (is_trivial_indirect_ref): New function.
10991 (dwarf2out_late_global_decl): Do not generate a location
10992 attribute for variables that have a non-trivial DECL_VALUE_EXPR
10993 and that are not defined in the current unit.
10995 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
10997 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
10998 instead of a libcall for UNORDERED.
11000 2018-02-09 Tamar Christina <tamar.christina@arm.com>
11003 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
11004 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
11006 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
11009 * config/s390/s390.c (s390_set_current_function): Invoke
11010 s390_indirect_branch_settings also if fndecl didn't change.
11012 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
11014 * config/rs6000/rs6000.md (blockage): Set length to zero.
11016 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
11018 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
11020 2018-02-09 Jakub Jelinek <jakub@redhat.com>
11023 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
11024 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
11028 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
11029 PARALLEL incoming that failed vt_get_decl_and_offset check.
11031 PR middle-end/84237
11032 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
11033 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
11035 (get_variable_section): For decls in named .bss* sections pass true as
11036 second argument to bss_initializer_p.
11038 2018-02-09 Marek Polacek <polacek@redhat.com>
11039 Jakub Jelinek <jakub@redhat.com>
11042 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
11043 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
11044 Sync some changes from cxx_fold_indirect_ref.
11046 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
11048 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
11050 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
11051 (BLOCK_INLINE_ENTRY_LABEL): New.
11052 (dwarf2out_var_location): Disregard inline entry markers.
11053 (inline_entry_data): New struct.
11054 (inline_entry_data_hasher): New hashtable type.
11055 (inline_entry_data_hasher::hash): New.
11056 (inline_entry_data_hasher::equal): New.
11057 (inline_entry_data_table): New variable.
11058 (add_high_low_attributes): Add DW_AT_entry_pc and
11059 DW_AT_GNU_entry_view attributes if a pending entry is found
11060 in inline_entry_data_table. Add old entry_pc attribute only
11061 if debug nonbinding markers are disabled.
11062 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
11063 markers are enabled.
11064 (block_within_block_p, dwarf2out_inline_entry): New.
11065 (dwarf2out_finish): Check that no entries remained in
11066 inline_entry_data_table.
11067 * final.c (reemit_insn_block_notes): Handle inline entry notes.
11068 (final_scan_insn, notice_source_line): Likewise.
11069 (rest_of_clean_state): Skip inline entry markers.
11070 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
11072 * gimple.c (gimple_build_debug_inline_entry): New.
11073 * gimple.h (enum gimple_debug_subcode): Add
11074 GIMPLE_DEBUG_INLINE_ENTRY.
11075 (gimple_build_debug_inline_entry): Declare.
11076 (gimple_debug_inline_entry_p): New.
11077 (gimple_debug_nonbind_marker_p): Adjust.
11078 * insn-notes.def (INLINE_ENTRY): New.
11079 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
11080 inline entry marker notes.
11081 (print_insn): Likewise.
11082 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
11083 (INSN_DEBUG_MARKER_KIND): Likewise.
11084 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
11085 * tree-inline.c (expand_call_inline): Build and insert
11086 debug_inline_entry stmt.
11087 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
11088 inline entry blocks early, if nonbind markers are enabled.
11089 (dump_scope_block): Dump fragment info.
11090 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
11091 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
11092 (gimple_build_debug_inline_entry): New.
11093 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
11094 Enable/disable inline entry points too.
11095 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
11096 (DEBUG_INSN): Describe inline entry markers.
11098 * common.opt (gvariable-location-views): New.
11099 (gvariable-location-views=incompat5): New.
11100 * config.in: Rebuilt.
11101 * configure: Rebuilt.
11102 * configure.ac: Test assembler for view support.
11103 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
11104 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
11105 * dwarf2out.c (var_loc_view): New typedef.
11106 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
11107 (dwarf2out_locviews_in_attribute): New.
11108 (dwarf2out_locviews_in_loclist): New.
11109 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
11110 (enum dw_line_info_opcode): Add LI_adv_address.
11111 (struct dw_line_info_table): Add view.
11112 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
11113 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
11114 (zero_view_p): New variable.
11115 (ZERO_VIEW_P): New macro.
11116 (output_asm_line_debug_info): New.
11117 (struct var_loc_node): Add view.
11118 (add_AT_view_list, AT_loc_list): New.
11119 (add_var_loc_to_decl): Add view param. Test it against last.
11120 (new_loc_list): Add view params. Record them.
11121 (AT_loc_list_ptr): Handle loc and view lists.
11122 (view_list_to_loc_list_val_node): New.
11123 (print_dw_val): Handle dw_val_class_view_list.
11124 (size_of_die): Likewise.
11125 (value_format): Likewise.
11126 (loc_list_has_views): New.
11127 (gen_llsym): Set vl_symbol too.
11128 (maybe_gen_llsym, skip_loc_list_entry): New.
11129 (dwarf2out_maybe_output_loclist_view_pair): New.
11130 (output_loc_list): Output view list or entries too.
11131 (output_view_list_offset): New.
11132 (output_die): Handle dw_val_class_view_list.
11133 (output_dwarf_version): New.
11134 (output_compilation_unit_header): Use it.
11135 (output_skeleton_debug_sections): Likewise.
11136 (output_rnglists, output_line_info): Likewise.
11137 (output_pubnames, output_aranges): Update version comments.
11138 (output_one_line_info_table): Output view numbers in asm comments.
11139 (dw_loc_list): Determine current endview, pass it to new_loc_list.
11140 Call maybe_gen_llsym.
11141 (loc_list_from_tree_1): Adjust.
11142 (add_AT_location_description): Create view list attribute if
11143 needed, check it's absent otherwise.
11144 (convert_cfa_to_fb_loc_list): Adjust.
11145 (maybe_emit_file): Call output_asm_line_debug_info for test.
11146 (dwarf2out_var_location): Reset views as needed. Precompute
11147 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
11148 attribute. Set view.
11149 (new_line_info_table): Reset next view.
11150 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
11151 (dwarf2out_source_line): Likewise. Output view resets and labels to
11152 the assembler, or select appropriate line info opcodes.
11153 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
11154 (optimize_string_length): Catch it. Adjust.
11155 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
11156 dw_val_class_view_list, and remove it if no longer needed.
11157 (hash_loc_list): Hash view numbers.
11158 (loc_list_hasher::equal): Compare them.
11159 (optimize_location_lists): Check whether a view list symbol is
11160 needed, and whether the locview attribute is present, and
11161 whether they match. Remove the locview attribute if no longer
11163 (index_location_lists): Call skip_loc_list_entry for test.
11164 (dwarf2out_finish): Call output_asm_line_debug_info for test.
11165 Use output_dwarf_version.
11166 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
11167 (struct dw_val_node): Add val_view_list.
11168 * final.c (SEEN_NEXT_VIEW): New.
11169 (set_next_view_needed): New.
11170 (clear_next_view_needed): New.
11171 (maybe_output_next_view): New.
11172 (final_start_function): Rename to...
11173 (final_start_function_1): ... this. Take pointer to FIRST,
11174 add SEEN parameter. Emit param bindings in the initial view.
11175 (final_start_function): Reintroduce SEEN-less interface.
11176 (final): Rename to...
11177 (final_1): ... this. Take SEEN parameter. Output final pending
11178 next view at the end.
11179 (final): Reintroduce seen-less interface.
11180 (final_scan_insn): Output pending next view before switching
11181 sections or ending a block. Mark the next view as needed when
11182 outputting variable locations. Notify debug backend of section
11183 changes, and of location view changes.
11184 (rest_of_handle_final): Adjust.
11185 * toplev.c (process_options): Autodetect value for debug variable
11186 location views option. Warn on incompat5 without -gdwarf-5.
11187 * doc/invoke.texi (gvariable-location-views): New.
11188 (gvariable-location-views=incompat5): New.
11189 (gno-variable-location-views): New.
11191 2018-02-08 David Malcolm <dmalcolm@redhat.com>
11193 PR tree-optimization/84136
11194 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
11195 that the result of find_edge is non-NULL.
11197 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
11200 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
11201 storing integer register in SImode. Fix cost of 256 and 512
11202 byte aligned SSE register store.
11204 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
11206 * config/i386/i386.c (ix86_multiplication_cost): Fix
11207 multiplication cost for TARGET_AVX512DQ.
11209 2018-02-08 Marek Polacek <polacek@redhat.com>
11211 PR tree-optimization/84238
11212 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
11215 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
11217 PR tree-optimization/84265
11218 * tree-vect-stmts.c (vectorizable_store): Don't treat
11219 VMAT_CONTIGUOUS accesses as grouped.
11220 (vectorizable_load): Likewise.
11222 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
11224 PR tree-optimization/81635
11225 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
11226 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
11227 (test_round_for_mask): New functions.
11228 (wide_int_cc_tests): Call test_round_for_mask.
11229 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
11230 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
11231 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
11232 range returned by get_range_info.
11234 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
11237 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
11238 * symtab.c: Include builtins.h
11239 (symtab_node::output_to_lto_symbol_table_p): Move here
11240 from lto-streamer-out.c:output_symbol_p.
11241 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
11242 (output_symbol_p): Move all logic to symtab.c
11243 (produce_symtab): Update.
11245 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
11247 * config/s390/s390-opts.h (enum indirect_branch): Define.
11248 * config/s390/s390-protos.h (s390_return_addr_from_memory)
11249 (s390_indirect_branch_via_thunk)
11250 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
11251 (enum s390_indirect_branch_type): Define.
11252 * config/s390/s390.c (struct s390_frame_layout, struct
11253 machine_function): Remove.
11254 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
11255 (indirect_branch_table_label_no, indirect_branch_table_name):
11257 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
11258 (enum s390_indirect_branch_option): Define.
11259 (s390_return_addr_from_memory): New function.
11260 (s390_handle_string_attribute): New function.
11261 (s390_attribute_table): Add new attribute handler.
11262 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
11263 (s390_indirect_branch_via_thunk): New function.
11264 (s390_indirect_branch_via_inline_thunk): New function.
11265 (s390_function_ok_for_sibcall): When jumping via thunk disallow
11266 sibling call optimization for non z10 compiles.
11267 (s390_emit_call): Force indirect branch target to be a single
11268 register. Add r1 clobber for non-z10 compiles.
11269 (s390_emit_epilogue): Emit return jump via return_use expander.
11270 (s390_reorg): Handle JUMP_INSNs as execute targets.
11271 (s390_option_override_internal): Perform validity checks for the
11272 new command line options.
11273 (s390_indirect_branch_attrvalue): New function.
11274 (s390_indirect_branch_settings): New function.
11275 (s390_set_current_function): Invoke s390_indirect_branch_settings.
11276 (s390_output_indirect_thunk_function): New function.
11277 (s390_code_end): Implement target hook.
11278 (s390_case_values_threshold): Implement target hook.
11279 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
11281 * config/s390/s390.h (struct s390_frame_layout)
11282 (struct machine_function): Move here from s390.c.
11283 (TARGET_INDIRECT_BRANCH_NOBP_RET)
11284 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
11285 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
11286 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
11287 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
11288 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
11289 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
11290 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
11291 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
11292 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
11293 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
11294 (mnemonic attribute): Add values which aren't recognized
11296 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
11297 pattern for branch conversion. Fix mnemonic attribute.
11298 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
11299 indirect branch via thunk if requested.
11300 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
11301 ("*indirect_jump"): Disable for branch conversion using out of
11303 ("indirect_jump_via_thunk<mode>_z10")
11304 ("indirect_jump_via_thunk<mode>")
11305 ("indirect_jump_via_inlinethunk<mode>_z10")
11306 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
11307 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
11308 ("casesi_jump_via_inlinethunk<mode>_z10")
11309 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
11310 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
11311 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
11312 ("*indirect2_jump"): Disable for branch conversion.
11313 ("casesi_jump"): Turn into expander and expand patterns for branch
11315 ("return_use"): New expander.
11316 ("*return"): Emit return via thunk and rename it to ...
11317 ("*return<mode>"): ... this one.
11318 * config/s390/s390.opt: Add new options and and enum for the
11321 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
11323 * lra-constraints.c (match_reload): Unconditionally use
11324 gen_lowpart_SUBREG, rather than selecting between that
11325 and equivalent gen_rtx_SUBREG code.
11327 2018-02-08 Richard Biener <rguenther@suse.de>
11329 PR tree-optimization/84233
11330 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
11331 changed flag instead of boguously re-using phi_inserted.
11333 2018-02-08 Martin Jambor <mjambor@suse.cz>
11335 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
11336 static local variables.
11338 2018-02-08 Richard Biener <rguenther@suse.de>
11340 PR tree-optimization/84278
11341 * tree-vect-stmts.c (vectorizable_store): When looking for
11342 smaller vector types to perform grouped strided loads/stores
11343 make sure the mode is supported by the target.
11344 (vectorizable_load): Likewise.
11346 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
11348 * config/aarch64/aarch64.c (aarch64_components_for_bb):
11349 Increase LDP/STP opportunities by adding adjacent callee-saves.
11351 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
11353 PR rtl-optimization/84068
11354 PR rtl-optimization/83459
11355 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
11357 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
11359 PR tree-optimization/84224
11360 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
11361 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
11362 non-zero arguments.
11364 2018-02-07 Iain Sandoe <iain@codesourcery.com>
11367 * config/rs6000/altivec.md (*restore_world): Remove LR use.
11368 * config/rs6000/predicates.md (restore_world_operation): Adjust op
11369 count, remove one USE.
11371 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
11373 * doc/install.texi (Configuration): Document the
11374 --with-long-double-format={ibm,ieee} PowerPC configuration
11378 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
11379 Convert from define_expand to be define_insn_and_split. Rework
11380 float/double/_Float128 conversions to QI/HI/SImode to work with
11381 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
11382 conversions to QI/HImode types did a store and then a load to
11383 truncate the value. For conversions to VSX registers, don't split
11384 the insn, instead emit the code directly. Use the code iterator
11385 any_fix to combine signed and unsigned conversions.
11386 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
11387 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
11388 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
11389 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
11390 (fix_<mode>di2_hw): Likewise.
11391 (fixuns_<mode>di2_hw): Likewise.
11392 (fix_<mode>si2_hw): Likewise.
11393 (fixuns_<mode>si2_hw): Likewise.
11394 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
11395 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
11396 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
11397 fix<uns>_trunc<SFDF:mode>si2_p8.
11398 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
11399 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
11400 (fix<uns>_<mode>_mem): Likewise.
11401 (fctiw<u>z_<mode>_mem): Likewise.
11402 (fix<uns>_<mode>_mem): Likewise.
11403 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
11404 the register allocator from doing a direct move to the GPRs to do
11405 a store, and instead use the ISA 3.0 store byte/half-word from
11406 vector register instruction. For IEEE 128-bit floating point,
11407 also optimize stores of 32-bit ints.
11408 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
11410 2018-02-07 Alan Hayward <alan.hayward@arm.com>
11412 * genextract.c (push_pathstr_operand): New function to support
11414 (walk_rtx): Call push_pathstr_operand.
11415 (print_path): Support [a-zA-Z].
11417 2018-02-07 Richard Biener <rguenther@suse.de>
11419 PR tree-optimization/84037
11420 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
11421 (cse_and_gimplify_to_preheader): Declare.
11422 (vect_get_place_in_interleaving_chain): Likewise.
11423 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
11425 (_loop_vec_info::~_loop_vec_info): Delete it.
11426 (cse_and_gimplify_to_preheader): New function.
11427 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
11428 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
11429 (vectorizable_load): Likewise. For grouped stores always base
11430 the IV on the first element.
11431 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
11432 condition before gimplifying.
11434 2018-02-07 Jakub Jelinek <jakub@redhat.com>
11436 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
11437 *DIV_EXPR and *MOD_EXPR.
11439 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
11442 * config/i386/i386.c (ix86_option_override_internal): Mask out
11443 the CF_SET bit when checking -fcf-protection.
11445 2018-02-07 Tom de Vries <tom@codesourcery.com>
11448 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
11451 2018-02-07 Richard Biener <rguenther@suse.de>
11453 PR tree-optimization/84204
11454 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
11457 PR tree-optimization/84205
11458 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
11459 special-case isl_ast_op_zdiv_r.
11461 PR tree-optimization/84223
11462 * graphite-scop-detection.c (gather_bbs::before_dom_children):
11463 Only add conditions from within the region.
11464 (gather_bbs::after_dom_children): Adjust.
11466 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
11469 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
11470 * config/avr/avr.md: Only post-reload split REG-REG moves if
11471 either register is GENERAL_REG_P.
11473 2018-02-07 Jakub Jelinek <jakub@redhat.com>
11475 PR tree-optimization/84235
11476 * tree-ssa-scopedtables.c
11477 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
11478 if the subtraction is performed in floating point type where NaNs are
11479 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
11480 build 1. Formatting fix.
11482 2018-02-06 Jakub Jelinek <jakub@redhat.com>
11485 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
11486 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
11487 and skip it regardless of bb boundaries. Use CALL_P macro,
11488 don't test INSN_P (insn) together with CALL_P or JUMP_P check
11489 unnecessarily, formatting fix.
11491 2018-02-06 Michael Collison <michael.collison@arm.com>
11493 * config/arm/thumb2.md:
11494 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
11495 (*thumb_mov_notscc): Ditto.
11497 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
11500 * config/rs6000/rs6000.md (su code attribute): Use "u" for
11501 unsigned_fix, not "s".
11503 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11505 * configure.ac (gcc_fn_eh_frame_ro): New function.
11506 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
11507 correct .eh_frame permissions.
11508 * configure: Regenerate.
11510 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
11512 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
11513 irrelevant options.
11515 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11517 * config/rs6000/rs6000.c (rs6000_option_override_internal):
11518 Display warning message for -mno-speculate-indirect-jumps.
11520 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
11522 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
11524 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
11526 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
11528 PR tree-optimization/84225
11529 * tree-eh.c (find_trapping_overflow): Only call
11530 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
11532 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
11535 * config/i386/i386.c: Reimplement the check of possible options
11536 -mibt/-mshstk conbination. Change error messages.
11537 * doc/invoke.texi: Fix a typo: remove extra '='.
11539 2018-02-06 Marek Polacek <polacek@redhat.com>
11541 PR tree-optimization/84228
11542 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
11544 2018-02-06 Tamar Christina <tamar.christina@arm.com>
11547 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
11548 emitted arch directives.
11549 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
11550 __ARM_FEATURE_COPROC before changing architectures.
11552 2018-02-06 Richard Biener <rguenther@suse.de>
11554 * config/i386/i386.c (print_reg): Fix typo.
11555 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
11557 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
11559 * configure: Regenerate.
11561 2018-02-05 Martin Sebor <msebor@redhat.com>
11563 PR tree-optimization/83369
11564 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
11567 2018-02-05 Martin Liska <mliska@suse.cz>
11569 * doc/invoke.texi: Cherry-pick upstream r323995.
11571 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
11573 * ira.c (ira_init_register_move_cost): Adjust comment.
11575 2018-02-05 Martin Liska <mliska@suse.cz>
11577 PR gcov-profile/84137
11578 * doc/gcov.texi: Fix typo in documentation.
11580 2018-02-05 Martin Liska <mliska@suse.cz>
11582 PR gcov-profile/83879
11583 * doc/gcov.texi: Document necessity of --dynamic-list-data when
11584 using dlopen functionality.
11586 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
11588 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
11589 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
11590 _mm_maskz_range_ss, _mm_mask_range_round_ss,
11591 _mm_maskz_range_round_ss): New intrinsics.
11592 (__builtin_ia32_rangesd128_round)
11593 (__builtin_ia32_rangess128_round): Remove.
11594 (__builtin_ia32_rangesd128_mask_round,
11595 __builtin_ia32_rangess128_mask_round): New builtins.
11596 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
11597 __builtin_ia32_rangess128_round): Remove.
11598 (__builtin_ia32_rangesd128_mask_round,
11599 __builtin_ia32_rangess128_mask_round): New builtins.
11600 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
11601 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
11602 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
11603 "<round_saeonly_constraint>")): Changed to ...
11604 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
11605 "<round_saeonly_scalar_constraint>")): ... this.
11606 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
11607 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
11608 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
11609 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
11610 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
11612 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
11614 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
11616 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
11617 Remove all values except native, 8540 and 8548.
11619 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
11621 * config/i386/i386.c (ix86_output_function_return): Pass
11622 INVALID_REGNUM, instead of -1, as invalid register number to
11623 indirect_thunk_name and output_indirect_thunk.
11625 2018-02-02 Julia Koval <julia.koval@intel.com>
11627 * config.gcc: Add -march=icelake.
11628 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
11629 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
11630 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
11631 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
11632 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
11633 (processor_target_table): Add icelake.
11634 (ix86_option_override_internal): Handle new PTAs.
11635 (get_builtin_code_for_version): Handle icelake.
11636 (M_INTEL_COREI7_ICELAKE): New.
11637 (fold_builtin_cpu): Handle icelake.
11638 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
11639 * doc/invoke.texi: Add -march=icelake.
11641 2018-02-02 Julia Koval <julia.koval@intel.com>
11643 * config/i386/i386.c (ix86_option_override_internal): Change flags type
11644 to wide_int_bitmask.
11645 * wide-int-bitmask.h: New.
11647 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
11650 * config/i386/i386.md: Replace Pmode with word_mode in
11651 builtin_setjmp_setup and builtin_longjmp to support x32.
11653 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
11657 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
11659 (rs6000_supported_cpu_names): New static variable.
11660 (linux_cpu_translation_table): Likewise.
11661 (elf_platform) <cpu>: Define new static variable and use it.
11662 Translate kernel AT_PLATFORM name to canonical name if needed.
11663 Error if platform name is unknown.
11665 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
11668 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
11670 2018-02-01 Jeff Law <law@redhat.com>
11673 * config/i386/i386.c (release_scratch_register_on_entry): Add new
11674 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
11675 the scratch if RELEASE_VIA_POP is false.
11676 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
11677 If we have to save a temporary register, decrement SIZE appropriately.
11678 Pass new arguments to release_scratch_register_on_entry.
11679 (ix86_adjust_stack_and_probe): Likewise.
11680 (ix86_emit_probe_stack_range): Pass new arguments to
11681 release_scratch_register_on_entry.
11683 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
11685 PR rtl-optimization/84157
11686 * combine.c (change_zero_ext): Use REG_P predicate in
11687 front of HARD_REGISTER_P predicate.
11689 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
11691 * config/avr/avr.c (avr_option_override): Move disabling of
11692 -fdelete-null-pointer-checks to...
11693 * common/config/avr/avr-common.c (avr_option_optimization_table):
11696 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11698 PR tree-optimization/81635
11699 * tree-data-ref.c (split_constant_offset_1): For types that
11700 wrap on overflow, try to use range info to prove that wrapping
11703 2018-02-01 Renlin Li <renlin.li@arm.com>
11706 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
11707 TAILCALL_ADDR_REGS.
11708 (aarch64_register_move_cost): Likewise.
11709 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
11710 TAILCALL_ADDR_REGS.
11711 (REG_CLASS_NAMES): Likewise.
11712 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
11713 TAILCALL_ADDR_REGS. Remove IP registers.
11714 * config/aarch64/aarch64.md (Ucs): Update register constraint.
11716 2018-02-01 Richard Biener <rguenther@suse.de>
11718 * domwalk.h (dom_walker::dom_walker): Add additional constructor
11719 for specifying RPO order and allow NULL for that.
11720 * domwalk.c (dom_walker::dom_walker): Likewise.
11721 (dom_walker::walk): Handle NULL RPO order.
11722 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
11724 (rewrite_update_dom_walker): Likewise.
11725 (mark_def_dom_walker): Likewise.
11727 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11729 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
11730 (aarch64_maybe_expand_sve_subreg_move): Declare.
11731 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
11732 * config/aarch64/predicates.md (aarch64_any_register_operand): New
11734 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
11735 that are semantically a reverse operation.
11736 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
11737 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
11738 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
11740 (aarch64_can_change_mode_class): For big-endian, forbid changes
11741 between two SVE modes if they have different element sizes.
11743 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11745 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
11746 the TImode handling for big-endian targets.
11748 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11750 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
11751 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
11753 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
11754 Remove BSWAP handing for big-endian targets and use the form of
11755 LD1RQ appropariate for the mode.
11757 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11759 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
11760 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
11761 duplicated element.
11763 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11766 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
11767 check for operands that need to go through aarch64_sve_reload_be.
11769 2018-02-01 Jakub Jelinek <jakub@redhat.com>
11771 PR tree-optimization/81661
11772 PR tree-optimization/84117
11773 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
11774 * tree-eh.c: Include gimplify.h.
11775 (find_trapping_overflow, replace_trapping_overflow,
11776 rewrite_to_non_trapping_overflow): New functions.
11777 * tree-vect-loop.c: Include tree-eh.h.
11778 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
11779 * tree-data-ref.c: Include tree-eh.h.
11780 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
11782 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
11784 PR rtl-optimization/84123
11785 * combine.c (change_zero_ext): Check if hard register satisfies
11786 can_change_dest_mode before calling gen_lowpart_SUBREG.
11788 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
11791 * ira.c (ira_init_register_move_cost): Remove assert.
11793 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
11795 PR rtl-optimization/84071
11796 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
11797 * doc/tm.texi: Regenerate.
11799 2018-01-31 Richard Biener <rguenther@suse.de>
11801 PR tree-optimization/84132
11802 * tree-data-ref.c (analyze_miv_subscript): Properly
11803 check whether evolution_function_is_affine_multivariate_p
11804 before calling gcd_of_steps_may_divide_p.
11806 2018-01-31 Julia Koval <julia.koval@intel.com>
11809 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
11810 * config/i386/i386.md (rdpid_rex64) New.
11811 (rdpid): Make 32bit only.
11813 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
11816 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
11817 an IDENTIFIER_NODE for FUNCTION_TYPE's.
11819 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
11822 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
11824 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
11826 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
11828 PR rtl-optimization/84071
11829 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
11830 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
11832 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
11834 * config/arc/arc.c (arc_handle_aux_attribute): New function.
11835 (arc_attribute_table): Add 'aux' attribute.
11836 (arc_in_small_data_p): Consider aux like variables.
11837 (arc_is_aux_reg_p): New function.
11838 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
11839 (arc_get_aux_arg): New function.
11840 (prepare_move_operands): Handle aux-register access.
11841 (arc_handle_aux_attribute): New function.
11842 * doc/extend.texi (ARC Variable attributes): Add subsection.
11844 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
11846 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
11847 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
11848 (arc_attribute_table): Add 'uncached' attribute.
11849 (arc_print_operand): Print '.di' flag for uncached memory
11851 (arc_in_small_data_p): Do not consider for small data the uncached
11853 (arc_is_uncached_mem_p): New function.
11854 * config/arc/predicates.md (compact_store_memory_operand): Check
11855 for uncached memory accesses.
11856 (nonvol_nonimm_operand): Likewise.
11857 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
11859 2018-01-31 Jakub Jelinek <jakub@redhat.com>
11862 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
11863 falign-loops=): Add Optimization flag.
11865 2018-01-30 Jeff Law <law@redhat.com>
11868 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
11869 INT_REGISTERS_SAVED. Check it prior to calling
11870 get_scratch_register_on_entry.
11871 (ix86_adjust_stack_and_probe): Similarly.
11872 (ix86_emit_probe_stack_range): Similarly.
11873 (ix86_expand_prologue): Corresponding changes.
11875 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11878 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
11879 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
11881 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
11884 * lra-constraints.c (curr_insn_transform): Process AND in the
11887 2018-01-30 Jakub Jelinek <jakub@redhat.com>
11889 PR rtl-optimization/83986
11890 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
11891 dependence against last_pending_memory_flush in addition to
11892 pending_jump_insns.
11894 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
11896 PR tree-optimization/81611
11897 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
11900 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11903 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
11906 2018-01-30 Richard Biener <rguenther@suse.de>
11907 Jakub Jelinek <jakub@redhat.com>
11909 PR tree-optimization/84111
11910 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
11911 inner loops added during recursion, as they don't have up-to-date
11914 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
11917 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
11918 (can_inline_edge_by_limits_p): ... here.
11919 (can_early_inline_edge_p, check_callers,
11920 update_caller_keys, update_callee_keys, recursive_inlining,
11921 add_new_edges_to_heap, speculation_useful_p,
11922 inline_small_functions,
11923 inline_small_functions, flatten_function,
11924 inline_to_all_callers_1): Update.
11926 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
11928 * profile-count.c (profile_count::combine_with_ipa_count): Handle
11931 2018-01-30 Richard Biener <rguenther@suse.de>
11933 PR tree-optimization/83008
11934 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
11935 invariant and constant vector uses in stmts when they need
11936 more than one stmt.
11938 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11941 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
11942 * configure: Regenerate.
11944 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
11946 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
11948 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
11949 Use gen_rtx_REG rather than gen_lowpart.
11951 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
11953 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
11954 rather than 0 when creating partial subregs.
11956 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
11958 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
11961 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
11964 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
11965 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
11966 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
11967 flags. This restores the settings used before the 2017-07-24.
11968 Turning off pre increment/decrement/modify allows IVOPTS to
11969 optimize DF/SF loops where the index is an int.
11971 2018-01-29 Richard Biener <rguenther@suse.de>
11972 Kelvin Nilsen <kelvin@gcc.gnu.org>
11975 * tree-vect-stmts.c (vectorizable_call): Don't call
11976 targetm.vectorize_builtin_md_vectorized_function if callee is
11979 2018-01-22 Carl Love <cel@us.ibm.com>
11981 * doc/extend.tex: Fix typo in second arg in
11982 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
11984 2018-01-29 Richard Biener <rguenther@suse.de>
11986 PR tree-optimization/84086
11987 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
11988 (flush_ssaname_freelist): When SSA names were released reset
11989 the SCEV hash table.
11991 2018-01-29 Richard Biener <rguenther@suse.de>
11993 PR tree-optimization/84057
11994 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
11995 removed paths when removing edges.
11997 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
11999 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
12000 -mfunction-return=@var{choice}.
12002 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
12004 PR diagnostic/84034
12005 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
12006 Handle CR like TAB.
12007 (layout::print_source_line): Likewise.
12008 (test_get_line_width_without_trailing_whitespace): Add test cases.
12010 2018-01-27 Jakub Jelinek <jakub@redhat.com>
12012 PR middle-end/84040
12013 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
12016 2018-01-26 Jim Wilson <jimw@sifive.com>
12018 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
12020 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
12023 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12025 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
12026 and CMP + SUB-immediate -> SUBS.
12028 2018-01-26 Martin Sebor <msebor@redhat.com>
12030 PR tree-optimization/83896
12031 * tree-ssa-strlen.c (get_string_len): Rename...
12032 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
12033 Avoid assuming length is constant.
12034 (handle_char_store): Use HOST_WIDE_INT for string length.
12036 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
12039 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
12040 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
12042 2018-01-26 Richard Biener <rguenther@suse.de>
12044 PR rtl-optimization/84003
12045 * dse.c (record_store): Only record redundant stores when
12046 the earlier store aliases at least all accesses the later one does.
12048 2018-01-26 Jakub Jelinek <jakub@redhat.com>
12050 PR rtl-optimization/83985
12051 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
12052 REG_CFA_RESTORE insns.
12053 (delete_unmarked_insns): Don't ignore separate shrink wrapping
12054 REG_CFA_RESTORE insns here.
12057 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
12058 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
12060 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
12062 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
12063 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
12064 (arc_init): Likewise.
12065 (arc_override_options): Likewise.
12066 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
12068 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
12070 * config/arc/arc.h (TARGET_DBNZ): Define.
12071 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
12072 properly set the tune attribute.
12073 (dbnz): Use TARGET_DBNZ guard.
12074 * config/arc/arc.opt (mtune): Add core3 option.
12076 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
12078 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
12079 recognize new pic like addresses.
12080 (arc_delegitimize_address): Clean up.
12082 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
12084 * config/arc/arc-arches.def: Option mrf16 valid for all
12086 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
12087 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
12088 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
12089 * config/arc/arc-tables.opt: Regenerate.
12090 * config/arc/arc.c (arc_conditional_register_usage): Handle
12091 reduced register file case.
12092 (arc_file_start): Set must have build attributes.
12093 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
12094 mrf16 option value.
12095 * config/arc/arc.opt (mrf16): Add new option.
12096 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
12097 * config/arc/genmultilib.awk: Handle new mrf16 option.
12098 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
12099 * config/arc/t-multilib: Regenerate.
12100 * doc/invoke.texi (ARC Options): Document mrf16 option.
12102 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
12104 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
12105 * config/arc/arc.c (arc_handle_secure_attribute): New function.
12106 (arc_attribute_table): Add 'secure_call' attribute.
12107 (arc_print_operand): Print secure call operand.
12108 (arc_function_ok_for_sibcall): Don't optimize tail calls when
12110 (arc_is_secure_call_p): New function. * config/arc/arc.md
12111 (call_i): Add support for sjli instruction.
12112 (call_value_i): Likewise.
12113 * config/arc/constraints.md (Csc): New constraint.
12115 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
12116 John Eric Martin <John.Martin@emmicro-us.com>
12118 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
12119 * config/arc/arc.c (_arc_jli_section): New struct.
12120 (arc_jli_section): New type.
12121 (rc_jli_sections): New static variable.
12122 (arc_handle_jli_attribute): New function.
12123 (arc_attribute_table): Add jli_always and jli_fixed attribute.
12124 (arc_file_end): New function.
12125 (TARGET_ASM_FILE_END): Define.
12126 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
12127 (arc_add_jli_section): New function.
12128 (jli_call_scan): Likewise.
12129 (arc_reorg): Call jli_call_scan.
12130 (arc_output_addsi): Remove 'S' from printing asm operand.
12131 (arc_is_jli_call_p): New function.
12132 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
12134 (movhi_insn): Likewise.
12135 (movsi_insn): Likewise.
12136 (movsi_set_cc_insn): Likewise.
12137 (loadqi_update): Likewise.
12138 (load_zeroextendqisi_update): Likewise.
12139 (load_signextendqisi_update): Likewise.
12140 (loadhi_update): Likewise.
12141 (load_zeroextendhisi_update): Likewise.
12142 (load_signextendhisi_update): Likewise.
12143 (loadsi_update): Likewise.
12144 (loadsf_update): Likewise.
12145 (movsicc_insn): Likewise.
12146 (bset_insn): Likewise.
12147 (bxor_insn): Likewise.
12148 (bclr_insn): Likewise.
12149 (bmsk_insn): Likewise.
12150 (bicsi3_insn): Likewise.
12151 (cmpsi_cc_c_insn): Likewise.
12152 (movsi_ne): Likewise.
12153 (movsi_cond_exec): Likewise.
12154 (clrsbsi2): Likewise.
12155 (norm_f): Likewise.
12165 (call_i): Remove 'S' asm letter, add jli instruction.
12166 (call_value_i): Likewise.
12167 * config/arc/arc.op (mjli-always): New option.
12168 * config/arc/constraints.md (Cji): New constraint.
12169 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
12171 (subsf3_fpx): Likewise.
12172 (mulsf3_fpx): Likewise.
12173 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
12175 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
12176 function attrbutes.
12177 * doc/invoke.texi (ARC): Document mjli-always option.
12179 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
12181 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
12182 avoid addition with 0 and use incw and decw where possible.
12184 2018-01-26 Richard Biener <rguenther@suse.de>
12186 PR tree-optimization/81082
12187 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
12188 association if it requires casting to unsigned.
12189 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
12190 from fold_plusminus_mult_expr to catch important cases late when
12191 range info is available.
12193 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12195 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
12196 * configure.ac (hidden_linkonce): New test.
12197 * configure: Regenerate.
12198 * config.in: Regenerate.
12200 2018-01-26 Julia Koval <julia.koval@intel.com>
12202 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
12203 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
12204 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
12205 _mm_mask_bitshuffle_epi64_mask): Fix type.
12206 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
12207 USI_FTYPE_V4DI_V4DI_USI): Remove.
12208 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
12209 __builtin_ia32_vpshufbitqmb256_mask,
12210 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
12211 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
12212 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
12214 2018-01-26 Alan Modra <amodra@gmail.com>
12217 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
12218 UNSPEC_VBPERMQ. Sort other unspecs.
12220 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
12222 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
12224 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
12226 PR middle-end/83055
12227 * predict.c (drop_profile): Do not push/pop cfun; update also
12229 (handle_missing_profiles): Fix logic looking for zero profiles.
12231 2018-01-25 Jakub Jelinek <jakub@redhat.com>
12233 PR middle-end/83977
12234 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
12235 on functions with #pragma omp declare simd or functions with simd
12237 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
12238 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
12239 Remove trailing \n from warning_at calls.
12241 2018-01-25 Tom de Vries <tom@codesourcery.com>
12244 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
12245 for neutered workers.
12247 2018-01-24 Joseph Myers <joseph@codesourcery.com>
12250 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
12251 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
12253 2018-01-24 Jeff Law <law@redhat.com>
12256 * i386.c (get_probe_interval): Move to earlier point.
12257 (ix86_compute_frame_layout): If -fstack-clash-protection and
12258 the frame is larger than the probe interval, then use pushes
12259 to save registers rather than reg->mem moves.
12260 (ix86_expand_prologue): Remove conditional for int_registers_saved
12263 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
12266 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
12267 min/max for never referenced object.
12269 2018-01-24 Jakub Jelinek <jakub@redhat.com>
12271 PR middle-end/83977
12272 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
12274 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
12275 attributes from DECL_ATTRIBUTES (decl) without affecting
12276 DECL_ATTRIBUTES (current_function_decl).
12277 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
12278 functions with non-NULL DECL_ABSTRACT_ORIGIN.
12280 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
12282 PR tree-optimization/83979
12283 * fold-const.c (fold_comparison): Use constant_boolean_node
12284 instead of boolean_{true,false}_node.
12286 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
12288 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
12291 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12293 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
12294 Simplify the clause that sets the length attribute.
12295 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
12296 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
12297 clause that sets the length attribute.
12298 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
12300 2018-01-24 Tom de Vries <tom@codesourcery.com>
12303 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
12304 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
12305 Add strict parameter.
12306 (prevent_branch_around_nothing): Insert dummy insn between branch to
12307 label and label with no ptx insn inbetween.
12308 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
12310 2018-01-24 Tom de Vries <tom@codesourcery.com>
12313 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
12314 for neutered threads in warp.
12315 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
12317 2018-01-24 Richard Biener <rguenther@suse.de>
12319 PR tree-optimization/83176
12320 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
12323 2018-01-24 Richard Biener <rguenther@suse.de>
12325 PR tree-optimization/82819
12326 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
12327 code generating pluses that are no-ops in the target precision.
12329 2018-01-24 Richard Biener <rguenther@suse.de>
12331 PR middle-end/84000
12332 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
12334 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
12336 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
12337 to merge probabilities.
12338 * predict.c (probably_never_executed): Also mark as cold functions
12339 with global 0 profile and guessed local profile.
12340 * profile-count.c (profile_probability::combine_with_count): New
12342 * profile-count.h (profile_probability::operator*,
12343 profile_probability::operator*=, profile_probability::operator/,
12344 profile_probability::operator/=): Reduce precision to adjusted
12345 and set value to guessed on contradictory divisions.
12346 (profile_probability::combine_with_freq): Remove.
12347 (profile_probability::combine_wiht_count): Declare.
12348 (profile_count::force_nonzero):: Set to adjusted.
12349 (profile_count::probability_in):: Set quality to adjusted.
12350 * tree-ssa-tail-merge.c (replace_block_by): Use
12351 combine_with_count.
12353 2018-01-23 Andrew Waterman <andrew@sifive.com>
12354 Jim Wilson <jimw@sifive.com>
12356 * config/riscv/riscv.c (riscv_stack_boundary): New.
12357 (riscv_option_override): Set riscv_stack_boundary. Handle
12358 riscv_preferred_stack_boundary_arg.
12359 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
12360 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
12361 (STACK_BOUNDARY): Set to riscv_stack_boundary.
12362 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
12363 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
12364 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
12366 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
12369 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
12370 of struct ix86_frame.
12371 (ix86_expand_epilogue): Likewise. Add a local variable for
12372 the reg_save_offset field in struct ix86_frame.
12374 2018-01-23 Bin Cheng <bin.cheng@arm.com>
12376 PR tree-optimization/82604
12377 * tree-loop-distribution.c (enum partition_kind): New enum item
12378 PKIND_PARTIAL_MEMSET.
12379 (partition_builtin_p): Support above new enum item.
12380 (generate_code_for_partition): Ditto.
12381 (compute_access_range): Differentiate cases that equality can be
12382 proven at all loops, the innermost loops or no loops.
12383 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
12384 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
12385 (finalize_partitions, distribute_loop): Don't fuse partition of
12386 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
12387 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
12388 parloop is enabled.
12390 2018-01-23 Martin Liska <mliska@suse.cz>
12392 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
12393 order to ignore the predictor.
12394 (PRED_POLYMORPHIC_CALL): Likewise.
12395 (PRED_RECURSIVE_CALL): Likewise.
12397 2018-01-23 Martin Liska <mliska@suse.cz>
12399 * tree-profile.c (tree_profiling): Print function header to
12400 aware reader which function we are working on.
12401 * value-prof.c (gimple_find_values_to_profile): Do not print
12402 not interesting value histograms.
12404 2018-01-23 Martin Liska <mliska@suse.cz>
12406 * profile-count.h (enum profile_quality): Add
12407 profile_uninitialized as the first value. Do not number values
12408 as they are zero based.
12409 (profile_count::verify): Update sanity check.
12410 (profile_probability::verify): Likewise.
12412 2018-01-23 Nathan Sidwell <nathan@acm.org>
12414 * doc/invoke.texi (ffor-scope): Deprecate.
12416 2018-01-23 David Malcolm <dmalcolm@redhat.com>
12418 PR tree-optimization/83510
12419 * domwalk.c (set_all_edges_as_executable): New function.
12420 (dom_walker::dom_walker): Convert bool param
12421 "skip_unreachable_blocks" to enum reachability. Move setup of
12422 edge flags to set_all_edges_as_executable and only do it when
12423 reachability is REACHABLE_BLOCKS.
12424 * domwalk.h (enum dom_walker::reachability): New enum.
12425 (dom_walker::dom_walker): Convert bool param
12426 "skip_unreachable_blocks" to enum reachability.
12427 (set_all_edges_as_executable): New decl.
12428 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
12429 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
12431 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
12432 but converting true to REACHABLE_BLOCKS.
12433 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
12435 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
12436 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
12437 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
12439 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
12440 if check_all_array_refs will be called.
12442 2018-01-23 David Malcolm <dmalcolm@redhat.com>
12444 * tree.c (selftest::test_location_wrappers): Add more test
12447 2018-01-23 David Malcolm <dmalcolm@redhat.com>
12449 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
12450 (selftest::test_bit_in_range): Likewise.
12452 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
12455 * doc/sourcebuild.texi (vect_float): Say that the selector
12456 only describes the situation when -funsafe-math-optimizations is on.
12457 (vect_float_strict): Document.
12459 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
12461 PR tree-optimization/83965
12462 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
12463 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
12464 instead of checking only for a reduction.
12465 (vect_recog_widen_sum_pattern): Likewise.
12467 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
12469 * predict.c (probably_never_executed): Only use precise profile info.
12470 (compute_function_frequency): Skip after inlining hack since we now
12471 have quality checking.
12473 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
12475 * profile-count.h (profile_probability::very_unlikely,
12476 profile_probability::unlikely, profile_probability::even): Set
12477 precision to guessed.
12479 2018-01-23 Richard Biener <rguenther@suse.de>
12481 PR tree-optimization/83963
12482 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
12483 Properly terminate dominator walk when crossing the exit edge not
12484 when visiting its source block.
12486 2018-01-23 Jakub Jelinek <jakub@redhat.com>
12489 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
12490 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
12492 2018-01-22 Jakub Jelinek <jakub@redhat.com>
12494 PR tree-optimization/83957
12495 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
12496 semicolon after for body surrounded by braces.
12498 PR tree-optimization/83081
12499 * profile-count.h (profile_probability::split): New method.
12500 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
12501 Use profile_probability::split.
12502 (do_compare_rtx_and_jump): Fix adjustment of probabilities
12503 when splitting a single conditional jump into 2.
12505 2018-01-22 David Malcolm <dmalcolm@redhat.com>
12507 PR tree-optimization/69452
12508 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
12511 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12513 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
12514 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
12515 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
12517 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12519 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
12521 * config/rl78/rl78.md (movdi): New define_expand.
12522 * config/rl78/rl78.c (rl78_split_movdi): New function.
12524 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
12527 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
12529 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
12530 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
12531 128-bit to produce an UNSPEC move to get the double word with the
12532 signbit and then a shift directly to do signbit.
12533 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
12534 implementation with a new version that just does either a direct
12535 move or a regular move. Move memory interface to separate insns.
12536 Move insns so they are next to the expander.
12537 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
12538 with signbit move. Split big and little endian case.
12539 (signbit<mode>2_dm_mem_le): Likewise.
12540 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
12541 (signbit<mode>2_dm2): Likewise.
12543 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12545 * config/rl78/rl78.md (anddi3): New define_expand.
12547 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12549 * config/rl78/rl78.md (umindi3): New define_expand.
12551 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12553 * config/rl78/rl78.md (smindi3): New define_expand.
12555 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12557 * config/rl78/rl78.md (smaxdi3): New define_expand.
12559 2018-01-22 Carl Love <cel@us.ibm.com>
12561 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
12562 LVX_V1TI): Add macro expansion.
12563 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
12564 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
12565 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
12566 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
12567 Change check to determine if the instruction is a byte reversing
12568 entry. Fix typo in comment.
12569 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
12570 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
12571 Add def_builtin calls for new builtins.
12572 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
12573 Add define_insn expansion.
12575 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12577 * config/rl78/rl78.md (umaxdi3): New define_expand.
12579 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12581 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
12582 for non-QImode registers.
12584 2018-01-22 Richard Biener <rguenther@suse.de>
12586 PR tree-optimization/83963
12587 * graphite-scop-detection.c (scop_detection::get_sese): Delay
12588 including the loop exit block.
12589 (scop_detection::merge_sese): Likewise.
12590 (scop_detection::add_scop): Do it here instead.
12592 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12594 * doc/sourcebuild.texi (arm_softfloat): Document.
12596 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
12599 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
12600 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
12601 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
12603 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12604 David Edelsohn <dje.gcc@gmail.com>
12607 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
12608 Change "crset eq" to "crset 2".
12609 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
12610 (*call_indirect_aix<mode>_nospec): Likewise.
12611 (*call_value_indirect_aix<mode>_nospec): Likewise.
12612 (*call_indirect_elfv2<mode>_nospec): Likewise.
12613 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
12614 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
12615 change assembly output from . to $.
12616 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
12617 (indirect_jump<mode>_nospec): Change assembly output from . to $.
12618 (*tablejump<mode>_internal1_nospec): Likewise.
12620 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
12623 * config/sh/sh_optimize_sett_clrt.cc:
12624 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
12626 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
12628 PR tree-optimization/83940
12629 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
12630 offset_dt to vect_constant_def rather than vect_unknown_def_type.
12631 (vect_check_load_store_mask): Add a mask_dt_out parameter and
12632 use it to pass back the definition type.
12633 (vect_check_store_rhs): Likewise rhs_dt_out.
12634 (vect_build_gather_load_calls): Add a mask_dt argument and use
12635 it instead of a call to vect_is_simple_use.
12636 (vectorizable_store): Update calls to vect_check_load_store_mask
12637 and vect_check_store_rhs. Use the dt returned by the latter instead
12638 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
12639 instead of calls to vect_is_simple_use. Pass the scalar rather
12640 than the vector operand to vect_is_simple_use when handling
12641 second and subsequent copies of an rhs value.
12642 (vectorizable_load): Update calls to vect_check_load_store_mask
12643 and vect_build_gather_load_calls. Use the cached mask_dt and
12644 gs_info.offset_dt instead of calls to vect_is_simple_use.
12646 2018-01-20 Jakub Jelinek <jakub@redhat.com>
12648 PR middle-end/83945
12649 * tree-emutls.c: Include gimplify.h.
12650 (lower_emutls_2): New function.
12651 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
12652 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
12653 it before further processing.
12656 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
12657 UINTVAL (trueop1) instead of INTVAL (op1).
12659 2018-01-19 Jakub Jelinek <jakub@redhat.com>
12663 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
12664 INCOMING_FRAME_SP_OFFSET if not defined.
12665 (scan_trace): Add ENTRY argument. If true and
12666 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
12667 emit a note to adjust the CFA offset.
12668 (create_cfi_notes): Adjust scan_trace callers.
12669 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
12670 INCOMING_FRAME_SP_OFFSET in the CIE.
12671 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
12672 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
12674 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
12675 * doc/tm.texi: Regenerated.
12677 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
12679 PR rtl-optimization/83147
12680 * lra-constraints.c (remove_inheritance_pseudos): Use
12681 lra_substitute_pseudo_within_insn.
12683 2018-01-19 Tom de Vries <tom@codesourcery.com>
12684 Cesar Philippidis <cesar@codesourcery.com>
12687 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
12689 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
12692 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
12693 spaces for function labels.
12695 2018-01-19 Martin Liska <mliska@suse.cz>
12697 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
12698 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
12699 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
12700 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
12701 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
12702 (PRED_CONST_RETURN): Change from 69 to 65.
12703 (PRED_NULL_RETURN): Change from 91 to 71.
12704 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
12705 (PRED_LOOP_GUARD): Change from 66 to 73.
12707 2018-01-19 Martin Liska <mliska@suse.cz>
12709 * predict.c (predict_insn_def): Add new assert.
12710 (struct branch_predictor): Change type to signed integer.
12711 (test_prediction_value_range): Amend test to cover
12712 PROB_UNINITIALIZED.
12713 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
12714 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
12715 (PRED_LOOP_ITERATIONS_MAX): Likewise.
12716 (PRED_LOOP_IV_COMPARE): Likewise.
12717 * predict.h (PROB_UNINITIALIZED): Define new constant.
12719 2018-01-19 Martin Liska <mliska@suse.cz>
12721 * predict.c (dump_prediction): Add new format for
12722 analyze_brprob.py script which is enabled with -details
12724 * profile-count.h (precise_p): New function.
12726 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
12728 PR tree-optimization/83922
12729 * tree-vect-loop.c (vect_verify_full_masking): Return false if
12730 there are no statements that need masking.
12731 (vect_active_double_reduction_p): New function.
12732 (vect_analyze_loop_operations): Use it when handling phis that
12733 are not in the loop header.
12735 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
12737 PR tree-optimization/83914
12738 * tree-vect-loop.c (vectorizable_induction): Don't convert
12739 init_expr or apply the peeling adjustment for inductions
12740 that are nested within the vectorized loop.
12742 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12744 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
12747 2018-01-18 Jakub Jelinek <jakub@redhat.com>
12751 * function.h (gimplify_parameters): Add gimple_seq * argument.
12752 * function.c: Include gimple.h and options.h.
12753 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
12754 for the added local temporaries if needed.
12755 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
12756 if there are any parameter cleanups, wrap whole body into a
12757 try/finally with the cleanups.
12759 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
12762 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
12763 Use GET_MODE_CLASS for scalar floating point.
12765 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
12769 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
12770 Fix call of call_cgraph_insertion_hooks.
12772 2018-01-18 Martin Sebor <msebor@redhat.com>
12774 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
12776 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
12779 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
12782 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
12785 * common.opt: (-ffile-prefix-map): New option.
12786 * opts.c (common_handle_option): Defer it.
12787 * opts-global.c (handle_common_deferred_options): Handle it.
12788 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
12789 * file-prefix-map.h: New file.
12790 (remap_debug_filename, add_debug_prefix_map): ...here.
12791 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
12792 * final.c (debug_prefix_map, add_debug_prefix_map
12793 remap_debug_filename): Move to...
12794 * file-prefix-map.c: New file.
12795 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
12796 generalize, get rid of alloca(), use strrchr() instead of strchr().
12797 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
12798 Implement in terms of add_prefix_map().
12799 (remap_macro_filename, remap_debug_filename): Implement in term of
12801 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
12802 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
12803 * dbxout.c: Include file-prefix-map.h.
12804 * varasm.c: Likewise.
12805 * vmsdbgout.c: Likewise.
12806 * xcoffout.c: Likewise.
12807 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
12808 * doc/cppopts.texi (-fmacro-prefix-map): Document.
12809 * doc/invoke.texi (-ffile-prefix-map): Document.
12810 (-fdebug-prefix-map): Update description.
12812 2018-01-18 Martin Liska <mliska@suse.cz>
12814 * config/i386/i386.c (indirect_thunk_name): Document that also
12816 (output_indirect_thunk): Document why both instructions
12817 (pause and lfence) are generated.
12819 2018-01-18 Richard Biener <rguenther@suse.de>
12821 PR tree-optimization/83887
12822 * graphite-scop-detection.c
12823 (scop_detection::get_nearest_dom_with_single_entry): Remove.
12824 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
12825 (scop_detection::merge_sese): Re-implement with a flood-fill
12826 algorithm that properly finds a SESE region if it exists.
12828 2018-01-18 Jakub Jelinek <jakub@redhat.com>
12831 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
12832 pointer_diff optimizations use view_convert instead of convert.
12834 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12836 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
12837 Generate different code for -mno-speculate-indirect-jumps.
12838 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
12839 (*call_indirect_aix<mode>): Disable for
12840 -mno-speculate-indirect-jumps.
12841 (*call_indirect_aix<mode>_nospec): New define_insn.
12842 (*call_value_indirect_aix<mode>): Disable for
12843 -mno-speculate-indirect-jumps.
12844 (*call_value_indirect_aix<mode>_nospec): New define_insn.
12845 (*sibcall_nonlocal_sysv<mode>): Generate different code for
12846 -mno-speculate-indirect-jumps.
12847 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
12849 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
12851 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
12852 long double type, set the flags for noting the default long double
12853 type, even if we don't pass or return a long double type.
12855 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
12858 * ipa-inline.c (flatten_function): Do not overwrite final inlining
12861 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
12863 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
12864 support for merge[hl].
12865 (fold_mergehl_helper): New helper function.
12866 (tree-vector-builder.h): New #include for tree_vector_builder usage.
12867 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
12868 (altivec_vmrglw_direct): Add xxmrglw insn.
12870 2018-01-17 Andrew Waterman <andrew@sifive.com>
12872 * config/riscv/riscv.c (riscv_conditional_register_usage): If
12873 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
12875 2018-01-17 David Malcolm <dmalcolm@redhat.com>
12878 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
12879 call the lto_location_cache before reading the
12880 DECL_SOURCE_LOCATION of the types.
12882 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
12883 Richard Sandiford <richard.sandiford@linaro.org>
12885 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
12886 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
12887 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
12888 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
12889 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
12891 * config/aarch64/constraints.md (aarch64_movti_operand):
12893 * config/aarch64/predicates.md (Uti): Add new constraint.
12895 2018-01-17 Carl Love <cel@us.ibm.com>
12897 * config/rs6000/vsx.md (define_expand xl_len_r,
12898 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
12899 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
12901 (define_expand, define_insn): Move the shift left from the
12902 define_insn to the define_expand for lxvl and stxvl instructions.
12903 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
12904 and XL_LEN_R definitions to PURE.
12906 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
12908 * config/i386/i386.c (indirect_thunk_name): Declare regno
12909 as unsigned int. Compare regno with INVALID_REGNUM.
12910 (output_indirect_thunk): Ditto.
12911 (output_indirect_thunk_function): Ditto.
12912 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
12913 in the call to output_indirect_thunk_function.
12915 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
12917 PR middle-end/83884
12918 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
12919 rather than the size of inner_type to determine the stack slot size
12920 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
12922 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
12925 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
12928 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
12930 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
12931 endian Linux systems to optionally enable multilibs for selecting
12932 the long double type if the user configured an explicit type.
12933 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
12934 have no long double multilibs if not defined.
12935 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
12936 warn if the user used -mabi={ieee,ibm}longdouble and we built
12937 multilibs for long double.
12938 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
12939 appropriate multilib option.
12940 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
12942 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
12943 for building long double multilibs.
12944 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
12946 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
12948 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
12951 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
12953 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
12956 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
12959 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
12962 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
12964 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
12965 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
12967 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
12969 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
12970 different rtl trees depending on TARGET_64BIT.
12971 (rs6000_gen_lvx): Likewise.
12973 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
12975 * config/visium/visium.md (nop): Tweak comment.
12976 (hazard_nop): Likewise.
12978 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12980 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
12981 -mspeculate-indirect-jumps.
12982 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
12983 for -mno-speculate-indirect-jumps.
12984 (*call_indirect_elfv2<mode>_nospec): New define_insn.
12985 (*call_value_indirect_elfv2<mode>): Disable for
12986 -mno-speculate-indirect-jumps.
12987 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
12988 (indirect_jump): Emit different RTL for
12989 -mno-speculate-indirect-jumps.
12990 (*indirect_jump<mode>): Disable for
12991 -mno-speculate-indirect-jumps.
12992 (*indirect_jump<mode>_nospec): New define_insn.
12993 (tablejump): Emit different RTL for
12994 -mno-speculate-indirect-jumps.
12995 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
12996 (tablejumpsi_nospec): New define_expand.
12997 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
12998 (tablejumpdi_nospec): New define_expand.
12999 (*tablejump<mode>_internal1): Disable for
13000 -mno-speculate-indirect-jumps.
13001 (*tablejump<mode>_internal1_nospec): New define_insn.
13002 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
13005 2018-01-16 Artyom Skrobov tyomitch@gmail.com
13007 * caller-save.c (insert_save): Drop unnecessary parameter. All
13010 2018-01-16 Jakub Jelinek <jakub@redhat.com>
13011 Richard Biener <rguenth@suse.de>
13014 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
13015 return early, inline manually is_gimple_sizepos. Make sure if we
13016 call gimplify_expr we don't end up with a gimple constant.
13017 * tree.c (variably_modified_type_p): Don't return true for
13018 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
13019 * gimplify.h (is_gimple_sizepos): Remove.
13021 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
13023 PR tree-optimization/83857
13024 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
13025 vectorizable_live_operation for pure SLP statements.
13026 (vectorizable_live_operation): Handle PHIs.
13028 2018-01-16 Richard Biener <rguenther@suse.de>
13030 PR tree-optimization/83867
13031 * tree-vect-stmts.c (vect_transform_stmt): Precompute
13032 nested_in_vect_loop_p since the scalar stmt may get invalidated.
13034 2018-01-16 Jakub Jelinek <jakub@redhat.com>
13037 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
13038 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
13039 If off is not INTEGER_CST, issue a may not be aligned warning
13040 rather than isn't aligned. Use isn%'t rather than isn't.
13041 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
13043 <case MULT_EXPR>: Improve the case when bottom and one of the
13044 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
13045 operand, in that case check if the other operand is multiple of
13046 bottom divided by the INTEGER_CST operand.
13048 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
13051 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
13052 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
13053 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
13054 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
13055 * config/pa/pa.c (pa_function_arg_advance): Likewise.
13056 (pa_function_arg, pa_arg_partial_bytes): Likewise.
13057 (pa_function_arg_size): New function.
13059 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
13061 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
13062 in a separate statement.
13064 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
13066 PR tree-optimization/83847
13067 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
13068 group gathers and scatters.
13070 2018-01-16 Jakub Jelinek <jakub@redhat.com>
13072 PR rtl-optimization/86620
13073 * params.def (max-sched-ready-insns): Bump minimum value to 1.
13075 PR rtl-optimization/83213
13076 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
13077 to last if both are JUMP_INSNs.
13079 PR tree-optimization/83843
13080 * gimple-ssa-store-merging.c
13081 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
13082 store_immediate_info for bswap/nop orig_stores.
13084 2018-01-15 Andrew Waterman <andrew@sifive.com>
13086 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
13088 <UDIV>: Increase cost if !TARGET_DIV.
13090 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
13092 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
13093 (define_attr "cr_logical_3op"): New.
13094 (cceq_ior_compare): Adjust.
13095 (cceq_ior_compare_complement): Adjust.
13096 (*cceq_rev_compare): Adjust.
13097 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
13098 (is_cracked_insn): Adjust.
13099 (insn_must_be_first_in_group): Adjust.
13100 * config/rs6000/40x.md: Adjust.
13101 * config/rs6000/440.md: Adjust.
13102 * config/rs6000/476.md: Adjust.
13103 * config/rs6000/601.md: Adjust.
13104 * config/rs6000/603.md: Adjust.
13105 * config/rs6000/6xx.md: Adjust.
13106 * config/rs6000/7450.md: Adjust.
13107 * config/rs6000/7xx.md: Adjust.
13108 * config/rs6000/8540.md: Adjust.
13109 * config/rs6000/cell.md: Adjust.
13110 * config/rs6000/e300c2c3.md: Adjust.
13111 * config/rs6000/e500mc.md: Adjust.
13112 * config/rs6000/e500mc64.md: Adjust.
13113 * config/rs6000/e5500.md: Adjust.
13114 * config/rs6000/e6500.md: Adjust.
13115 * config/rs6000/mpc.md: Adjust.
13116 * config/rs6000/power4.md: Adjust.
13117 * config/rs6000/power5.md: Adjust.
13118 * config/rs6000/power6.md: Adjust.
13119 * config/rs6000/power7.md: Adjust.
13120 * config/rs6000/power8.md: Adjust.
13121 * config/rs6000/power9.md: Adjust.
13122 * config/rs6000/rs64.md: Adjust.
13123 * config/rs6000/titan.md: Adjust.
13125 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
13127 * config/i386/predicates.md (indirect_branch_operand): Rewrite
13128 ix86_indirect_branch_register logic.
13130 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
13132 * config/i386/constraints.md (Bs): Update
13133 ix86_indirect_branch_register check. Don't check
13134 ix86_indirect_branch_register with GOT_memory_operand.
13136 * config/i386/predicates.md (GOT_memory_operand): Don't check
13137 ix86_indirect_branch_register here.
13138 (GOT32_symbol_operand): Likewise.
13140 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
13142 * config/i386/predicates.md (constant_call_address_operand):
13143 Rewrite ix86_indirect_branch_register logic.
13144 (sibcall_insn_operand): Likewise.
13146 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
13148 * config/i386/constraints.md (Bs): Replace
13149 ix86_indirect_branch_thunk_register with
13150 ix86_indirect_branch_register.
13152 * config/i386/i386.md (indirect_jump): Likewise.
13153 (tablejump): Likewise.
13154 (*sibcall_memory): Likewise.
13155 (*sibcall_value_memory): Likewise.
13156 Peepholes of indirect call and jump via memory: Likewise.
13157 * config/i386/i386.opt: Likewise.
13158 * config/i386/predicates.md (indirect_branch_operand): Likewise.
13159 (GOT_memory_operand): Likewise.
13160 (call_insn_operand): Likewise.
13161 (sibcall_insn_operand): Likewise.
13162 (GOT32_symbol_operand): Likewise.
13164 2018-01-15 Jakub Jelinek <jakub@redhat.com>
13166 PR middle-end/83837
13167 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
13168 type rather than type addr's type points to.
13169 (expand_omp_atomic_mutex): Likewise.
13170 (expand_omp_atomic): Likewise.
13172 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
13175 * config/i386/i386.c (output_indirect_thunk_function): Use
13176 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
13177 for __x86_return_thunk.
13179 2018-01-15 Richard Biener <rguenther@suse.de>
13181 PR middle-end/83850
13182 * expmed.c (extract_bit_field_1): Fix typo.
13184 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13187 * config/arm/iterators.md (VF): New mode iterator.
13188 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
13189 Remove integer-related logic from pattern.
13190 (neon_vabd<mode>_3): Likewise.
13192 2018-01-15 Jakub Jelinek <jakub@redhat.com>
13194 PR middle-end/82694
13195 * common.opt (fstrict-overflow): No longer an alias.
13196 (fwrapv-pointer): New option.
13197 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
13198 also for pointer types based on flag_wrapv_pointer.
13199 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
13200 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
13201 opts->x_flag_wrapv got set.
13202 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
13203 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
13204 POINTER_TYPE_OVERFLOW_UNDEFINED.
13205 * match.pd: Likewise in address comparison pattern.
13206 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
13208 2018-01-15 Richard Biener <rguenther@suse.de>
13211 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
13212 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
13213 Reset type names to their identifier if their TYPE_DECL doesn't
13214 have linkage (and thus is used for ODR and devirt).
13215 (save_debug_info_for_decl): Remove.
13216 (save_debug_info_for_type): Likewise.
13217 (add_tree_to_fld_list): Adjust.
13218 * tree-pretty-print.c (dump_generic_node): Make dumping of
13219 type names more robust.
13221 2018-01-15 Richard Biener <rguenther@suse.de>
13223 * BASE-VER: Bump to 8.0.1.
13225 2018-01-14 Martin Sebor <msebor@redhat.com>
13228 * builtins.c (check_access): Avoid warning when the no-warning bit
13231 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
13233 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
13234 * ira-color (allocno_hard_regs_compare): Likewise.
13236 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
13239 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
13240 Use .pushsection/.popsection.
13242 2018-01-14 Martin Sebor <msebor@redhat.com>
13245 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
13247 2018-01-14 Jakub Jelinek <jakub@redhat.com>
13249 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
13250 entry from extra_headers.
13251 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
13252 extra_headers, make the list bitwise identical to the i?86-*-* one.
13254 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
13256 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
13257 -mcmodel=large with -mindirect-branch=thunk,
13258 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
13259 -mfunction-return=thunk-extern.
13260 * doc/invoke.texi: Document -mcmodel=large is incompatible with
13261 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
13262 -mfunction-return=thunk and -mfunction-return=thunk-extern.
13264 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
13266 * config/i386/i386.c (print_reg): Print the name of the full
13267 integer register without '%'.
13268 (ix86_print_operand): Handle 'V'.
13269 * doc/extend.texi: Document 'V' modifier.
13271 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
13273 * config/i386/constraints.md (Bs): Disallow memory operand for
13274 -mindirect-branch-register.
13276 * config/i386/predicates.md (indirect_branch_operand): Likewise.
13277 (GOT_memory_operand): Likewise.
13278 (call_insn_operand): Likewise.
13279 (sibcall_insn_operand): Likewise.
13280 (GOT32_symbol_operand): Likewise.
13281 * config/i386/i386.md (indirect_jump): Call convert_memory_address
13282 for -mindirect-branch-register.
13283 (tablejump): Likewise.
13284 (*sibcall_memory): Likewise.
13285 (*sibcall_value_memory): Likewise.
13286 Disallow peepholes of indirect call and jump via memory for
13287 -mindirect-branch-register.
13288 (*call_pop): Replace m with Bw.
13289 (*call_value_pop): Likewise.
13290 (*sibcall_pop_memory): Replace m with Bs.
13291 * config/i386/i386.opt (mindirect-branch-register): New option.
13292 * doc/invoke.texi: Document -mindirect-branch-register option.
13294 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
13296 * config/i386/i386-protos.h (ix86_output_function_return): New.
13297 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
13298 set function_return_type.
13299 (indirect_thunk_name): Add ret_p to indicate thunk for function
13301 (output_indirect_thunk_function): Pass false to
13302 indirect_thunk_name.
13303 (ix86_output_indirect_branch_via_reg): Likewise.
13304 (ix86_output_indirect_branch_via_push): Likewise.
13305 (output_indirect_thunk_function): Create alias for function
13306 return thunk if regno < 0.
13307 (ix86_output_function_return): New function.
13308 (ix86_handle_fndecl_attribute): Handle function_return.
13309 (ix86_attribute_table): Add function_return.
13310 * config/i386/i386.h (machine_function): Add
13311 function_return_type.
13312 * config/i386/i386.md (simple_return_internal): Use
13313 ix86_output_function_return.
13314 (simple_return_internal_long): Likewise.
13315 * config/i386/i386.opt (mfunction-return=): New option.
13316 (indirect_branch): Mention -mfunction-return=.
13317 * doc/extend.texi: Document function_return function attribute.
13318 * doc/invoke.texi: Document -mfunction-return= option.
13320 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
13322 * config/i386/i386-opts.h (indirect_branch): New.
13323 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
13324 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
13325 with local indirect jump when converting indirect call and jump.
13326 (ix86_set_indirect_branch_type): New.
13327 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
13328 (indirectlabelno): New.
13329 (indirect_thunk_needed): Likewise.
13330 (indirect_thunk_bnd_needed): Likewise.
13331 (indirect_thunks_used): Likewise.
13332 (indirect_thunks_bnd_used): Likewise.
13333 (INDIRECT_LABEL): Likewise.
13334 (indirect_thunk_name): Likewise.
13335 (output_indirect_thunk): Likewise.
13336 (output_indirect_thunk_function): Likewise.
13337 (ix86_output_indirect_branch_via_reg): Likewise.
13338 (ix86_output_indirect_branch_via_push): Likewise.
13339 (ix86_output_indirect_branch): Likewise.
13340 (ix86_output_indirect_jmp): Likewise.
13341 (ix86_code_end): Call output_indirect_thunk_function if needed.
13342 (ix86_output_call_insn): Call ix86_output_indirect_branch if
13344 (ix86_handle_fndecl_attribute): Handle indirect_branch.
13345 (ix86_attribute_table): Add indirect_branch.
13346 * config/i386/i386.h (machine_function): Add indirect_branch_type
13347 and has_local_indirect_jump.
13348 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
13350 (tablejump): Likewise.
13351 (*indirect_jump): Use ix86_output_indirect_jmp.
13352 (*tablejump_1): Likewise.
13353 (simple_return_indirect_internal): Likewise.
13354 * config/i386/i386.opt (mindirect-branch=): New option.
13355 (indirect_branch): New.
13358 (thunk-inline): Likewise.
13359 (thunk-extern): Likewise.
13360 * doc/extend.texi: Document indirect_branch function attribute.
13361 * doc/invoke.texi: Document -mindirect-branch= option.
13363 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
13366 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
13368 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
13370 * ipa-inline.c (want_inline_small_function_p): Return false if
13371 inlining has already failed with CIF_FINAL_ERROR.
13372 (update_caller_keys): Call want_inline_small_function_p before
13374 (update_callee_keys): Likewise.
13376 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
13378 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
13380 (rs6000_quadword_masked_address_p): Likewise.
13381 (quad_aligned_load_p): Likewise.
13382 (quad_aligned_store_p): Likewise.
13383 (const_load_sequence_p): Add comment to describe the outer-most loop.
13384 (mimic_memory_attributes_and_flags): New function.
13385 (rs6000_gen_stvx): Likewise.
13386 (replace_swapped_aligned_store): Likewise.
13387 (rs6000_gen_lvx): Likewise.
13388 (replace_swapped_aligned_load): Likewise.
13389 (replace_swapped_load_constant): Capitalize argument name in
13390 comment describing this function.
13391 (rs6000_analyze_swaps): Add a third pass to search for vector loads
13392 and stores that access quad-word aligned addresses and replace
13393 with stvx or lvx instructions when appropriate.
13394 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
13395 New function prototype.
13396 (rs6000_quadword_masked_address_p): Likewise.
13397 (rs6000_gen_lvx): Likewise.
13398 (rs6000_gen_stvx): Likewise.
13399 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
13400 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
13401 when memory address is aligned.
13402 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
13403 this split to select lvx instruction when memory address is aligned.
13404 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
13405 instruction when memory address is aligned.
13406 (*vsx_le_perm_load_v16qi): Likewise.
13407 (four unnamed splitters): Modify to select the stvx instruction
13408 when memory is aligned.
13410 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
13412 * predict.c (determine_unlikely_bbs): Handle correctly BBs
13413 which appears in the queue multiple times.
13415 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13416 Alan Hayward <alan.hayward@arm.com>
13417 David Sherwood <david.sherwood@arm.com>
13419 * tree-vectorizer.h (vec_lower_bound): New structure.
13420 (_loop_vec_info): Add check_nonzero and lower_bounds.
13421 (LOOP_VINFO_CHECK_NONZERO): New macro.
13422 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
13423 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
13424 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
13425 fields. Make seg_len the distance travelled, not including the
13427 (dr_direction_indicator): Declare.
13428 (dr_zero_step_indicator): Likewise.
13429 (dr_known_forward_stride_p): Likewise.
13430 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
13432 (runtime_alias_check_p): Allow runtime alias checks with
13434 (operator ==): Compare access_size and align.
13435 (prune_runtime_alias_test_list): Rework for new distinction between
13436 the access_size and seg_len.
13437 (create_intersect_range_checks_index): Likewise. Cope with polynomial
13439 (get_segment_min_max): New function.
13440 (create_intersect_range_checks): Use it.
13441 (dr_step_indicator): New function.
13442 (dr_direction_indicator): Likewise.
13443 (dr_zero_step_indicator): Likewise.
13444 (dr_known_forward_stride_p): Likewise.
13445 * tree-loop-distribution.c (data_ref_segment_size): Return
13446 DR_STEP * (niters - 1).
13447 (compute_alias_check_pairs): Update call to the dr_with_seg_len
13449 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
13450 (vect_preserves_scalar_order_p): New function, split out from...
13451 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
13452 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
13453 (vect_vfa_access_size): New function.
13454 (vect_vfa_align): Likewise.
13455 (vect_compile_time_alias): Take access_size_a and access_b arguments.
13456 (dump_lower_bound): New function.
13457 (vect_check_lower_bound): Likewise.
13458 (vect_small_gap_p): Likewise.
13459 (vectorizable_with_step_bound_p): Likewise.
13460 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
13461 depencies if the vectorization factor is 1. Convert the checks
13462 for nonzero steps into checks on the bounds of DR_STEP. Try using
13463 a bunds check for variable steps if the minimum required step is
13464 relatively small. Update calls to the dr_with_seg_len
13465 constructor and to vect_compile_time_alias.
13466 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
13468 (vect_loop_versioning): Call it.
13469 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
13471 (vect_estimate_min_profitable_iters): Account for any bounds checks.
13473 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13474 Alan Hayward <alan.hayward@arm.com>
13475 David Sherwood <david.sherwood@arm.com>
13477 * doc/sourcebuild.texi (vect_scatter_store): Document.
13478 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
13480 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
13482 * genopinit.c (main): Add supports_vec_scatter_store and
13483 supports_vec_scatter_store_cached to target_optabs.
13484 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
13485 IFN_MASK_SCATTER_STORE.
13486 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
13488 * internal-fn.h (internal_store_fn_p): Declare.
13489 (internal_fn_stored_value_index): Likewise.
13490 * internal-fn.c (scatter_store_direct): New macro.
13491 (expand_scatter_store_optab_fn): New function.
13492 (direct_scatter_store_optab_supported_p): New macro.
13493 (internal_store_fn_p): New function.
13494 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
13495 IFN_MASK_SCATTER_STORE.
13496 (internal_fn_mask_index): Likewise.
13497 (internal_fn_stored_value_index): New function.
13498 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
13499 for scatter stores.
13500 * optabs-query.h (supports_vec_scatter_store_p): Declare.
13501 * optabs-query.c (supports_vec_scatter_store_p): New function.
13502 * tree-vectorizer.h (vect_get_store_rhs): Declare.
13503 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
13504 true for scatter stores.
13505 (vect_gather_scatter_fn_p): Handle scatter stores too.
13506 (vect_check_gather_scatter): Consider using scatter stores if
13507 supports_vec_scatter_store_p.
13508 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
13509 scatter stores too.
13510 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
13511 internal_fn_stored_value_index.
13512 (check_load_store_masking): Handle scatter stores too.
13513 (vect_get_store_rhs): Make public.
13514 (vectorizable_call): Use internal_store_fn_p.
13515 (vectorizable_store): Handle scatter store internal functions.
13516 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
13517 when deciding whether the end of the group has been reached.
13518 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
13519 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
13520 (mask_scatter_store<mode>): New insns.
13522 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13523 Alan Hayward <alan.hayward@arm.com>
13524 David Sherwood <david.sherwood@arm.com>
13526 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
13527 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
13528 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
13530 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
13531 Use vect_truncate_gather_scatter_offset if we can't treat the
13532 operation as a normal gather load or scatter store.
13533 (get_group_load_store_type): Take the gather_scatter_info
13534 as argument. Try using a gather load or scatter store for
13535 single-element groups.
13536 (get_load_store_type): Update calls to get_group_load_store_type
13537 and vect_use_strided_gather_scatters_p.
13539 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13540 Alan Hayward <alan.hayward@arm.com>
13541 David Sherwood <david.sherwood@arm.com>
13543 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
13544 optional tree argument.
13545 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
13547 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
13548 but continue to use the current value as a fallback.
13549 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
13550 to compare the updates.
13551 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
13552 (get_load_store_type): Use it when handling a strided access.
13553 (vect_get_strided_load_store_ops): New function.
13554 (vect_get_data_ptr_increment): Likewise.
13555 (vectorizable_load): Handle strided gather loads. Always pass
13556 a step to vect_create_data_ref_ptr and bump_vector_ptr.
13558 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13559 Alan Hayward <alan.hayward@arm.com>
13560 David Sherwood <david.sherwood@arm.com>
13562 * doc/md.texi (gather_load@var{m}): Document.
13563 (mask_gather_load@var{m}): Likewise.
13564 * genopinit.c (main): Add supports_vec_gather_load and
13565 supports_vec_gather_load_cached to target_optabs.
13566 * optabs-tree.c (init_tree_optimization_optabs): Use
13567 ggc_cleared_alloc to allocate target_optabs.
13568 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
13569 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
13571 * internal-fn.h (internal_load_fn_p): Declare.
13572 (internal_gather_scatter_fn_p): Likewise.
13573 (internal_fn_mask_index): Likewise.
13574 (internal_gather_scatter_fn_supported_p): Likewise.
13575 * internal-fn.c (gather_load_direct): New macro.
13576 (expand_gather_load_optab_fn): New function.
13577 (direct_gather_load_optab_supported_p): New macro.
13578 (direct_internal_fn_optab): New function.
13579 (internal_load_fn_p): Likewise.
13580 (internal_gather_scatter_fn_p): Likewise.
13581 (internal_fn_mask_index): Likewise.
13582 (internal_gather_scatter_fn_supported_p): Likewise.
13583 * optabs-query.c (supports_at_least_one_mode_p): New function.
13584 (supports_vec_gather_load_p): Likewise.
13585 * optabs-query.h (supports_vec_gather_load_p): Declare.
13586 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
13587 and memory_type field.
13588 (NUM_PATTERNS): Bump to 15.
13589 * tree-vect-data-refs.c: Include internal-fn.h.
13590 (vect_gather_scatter_fn_p): New function.
13591 (vect_describe_gather_scatter_call): Likewise.
13592 (vect_check_gather_scatter): Try using internal functions for
13593 gather loads. Recognize existing calls to a gather load function.
13594 (vect_analyze_data_refs): Consider using gather loads if
13595 supports_vec_gather_load_p.
13596 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
13597 (vect_get_gather_scatter_offset_type): Likewise.
13598 (vect_convert_mask_for_vectype): Likewise.
13599 (vect_add_conversion_to_patterm): Likewise.
13600 (vect_try_gather_scatter_pattern): Likewise.
13601 (vect_recog_gather_scatter_pattern): New pattern recognizer.
13602 (vect_vect_recog_func_ptrs): Add it.
13603 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
13604 internal_fn_mask_index and internal_gather_scatter_fn_p.
13605 (check_load_store_masking): Take the gather_scatter_info as an
13606 argument and handle gather loads.
13607 (vect_get_gather_scatter_ops): New function.
13608 (vectorizable_call): Check internal_load_fn_p.
13609 (vectorizable_load): Likewise. Handle gather load internal
13611 (vectorizable_store): Update call to check_load_store_masking.
13612 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
13613 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
13614 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
13615 (aarch64_gather_scale_operand_d): New predicates.
13616 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
13617 (mask_gather_load<mode>): New insns.
13619 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13620 Alan Hayward <alan.hayward@arm.com>
13621 David Sherwood <david.sherwood@arm.com>
13623 * optabs.def (fold_left_plus_optab): New optab.
13624 * doc/md.texi (fold_left_plus_@var{m}): Document.
13625 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
13626 * internal-fn.c (fold_left_direct): Define.
13627 (expand_fold_left_optab_fn): Likewise.
13628 (direct_fold_left_optab_supported_p): Likewise.
13629 * fold-const-call.c (fold_const_fold_left): New function.
13630 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
13631 * tree-parloops.c (valid_reduction_p): New function.
13632 (gather_scalar_reductions): Use it.
13633 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
13634 (vect_finish_replace_stmt): Declare.
13635 * tree-vect-loop.c (fold_left_reduction_fn): New function.
13636 (needs_fold_left_reduction_p): New function, split out from...
13637 (vect_is_simple_reduction): ...here. Accept reductions that
13638 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
13639 (vect_force_simple_reduction): Also store the reduction type in
13640 the assignment's STMT_VINFO_REDUC_TYPE.
13641 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
13642 (merge_with_identity): New function.
13643 (vect_expand_fold_left): Likewise.
13644 (vectorize_fold_left_reduction): Likewise.
13645 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
13646 scalar phi in place for it. Check for target support and reject
13647 cases that would reassociate the operation. Defer the transform
13648 phase to vectorize_fold_left_reduction.
13649 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
13650 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
13651 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
13653 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13655 * tree-if-conv.c (predicate_mem_writes): Remove redundant
13656 call to ifc_temp_var.
13658 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13659 Alan Hayward <alan.hayward@arm.com>
13660 David Sherwood <david.sherwood@arm.com>
13662 * target.def (legitimize_address_displacement): Take the original
13663 offset as a poly_int.
13664 * targhooks.h (default_legitimize_address_displacement): Update
13666 * targhooks.c (default_legitimize_address_displacement): Likewise.
13667 * doc/tm.texi: Regenerate.
13668 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
13669 as an argument, moving assert of ad->disp == ad->disp_term to...
13670 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
13671 Try calling targetm.legitimize_address_displacement before expanding
13672 the address rather than afterwards, and adjust for the new interface.
13673 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
13674 Match the new hook interface. Handle SVE addresses.
13675 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
13676 new hook interface.
13678 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13680 * Makefile.in (OBJS): Add early-remat.o.
13681 * target.def (select_early_remat_modes): New hook.
13682 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
13683 * doc/tm.texi: Regenerate.
13684 * targhooks.h (default_select_early_remat_modes): Declare.
13685 * targhooks.c (default_select_early_remat_modes): New function.
13686 * timevar.def (TV_EARLY_REMAT): New timevar.
13687 * passes.def (pass_early_remat): New pass.
13688 * tree-pass.h (make_pass_early_remat): Declare.
13689 * early-remat.c: New file.
13690 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
13692 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
13694 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13695 Alan Hayward <alan.hayward@arm.com>
13696 David Sherwood <david.sherwood@arm.com>
13698 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
13699 vfm1 with a bound_epilog parameter.
13700 (vect_do_peeling): Update calls accordingly, and move the prologue
13701 call earlier in the function. Treat the base bound_epilog as 0 for
13702 fully-masked loops and retain vf - 1 for other loops. Add 1 to
13703 this base when peeling for gaps.
13704 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
13705 with fully-masked loops.
13706 (vect_estimate_min_profitable_iters): Handle the single peeled
13707 iteration in that case.
13709 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13710 Alan Hayward <alan.hayward@arm.com>
13711 David Sherwood <david.sherwood@arm.com>
13713 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
13714 single-element interleaving even if the size is not a power of 2.
13715 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
13716 accesses for single-element interleaving if the group size is
13719 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13720 Alan Hayward <alan.hayward@arm.com>
13721 David Sherwood <david.sherwood@arm.com>
13723 * doc/md.texi (fold_extract_last_@var{m}): Document.
13724 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
13725 * optabs.def (fold_extract_last_optab): New optab.
13726 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
13727 * internal-fn.c (fold_extract_direct): New macro.
13728 (expand_fold_extract_optab_fn): Likewise.
13729 (direct_fold_extract_optab_supported_p): Likewise.
13730 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
13731 * tree-vect-loop.c (vect_model_reduction_cost): Handle
13732 EXTRACT_LAST_REDUCTION.
13733 (get_initial_def_for_reduction): Do not create an initial vector
13734 for EXTRACT_LAST_REDUCTION reductions.
13735 (vectorizable_reduction): Leave the scalar phi in place for
13736 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
13737 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
13738 epilogue code for EXTRACT_LAST_REDUCTION and defer the
13739 transform phase to vectorizable_condition.
13740 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
13742 (vect_finish_stmt_generation): ...here.
13743 (vect_finish_replace_stmt): New function.
13744 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
13745 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
13747 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
13749 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13750 Alan Hayward <alan.hayward@arm.com>
13751 David Sherwood <david.sherwood@arm.com>
13753 * doc/md.texi (extract_last_@var{m}): Document.
13754 * optabs.def (extract_last_optab): New optab.
13755 * internal-fn.def (EXTRACT_LAST): New internal function.
13756 * internal-fn.c (cond_unary_direct): New macro.
13757 (expand_cond_unary_optab_fn): Likewise.
13758 (direct_cond_unary_optab_supported_p): Likewise.
13759 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
13760 loops using EXTRACT_LAST.
13761 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
13762 (extract_last_<mode>): ...this optab.
13763 (vec_extract<mode><Vel>): Update accordingly.
13765 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13766 Alan Hayward <alan.hayward@arm.com>
13767 David Sherwood <david.sherwood@arm.com>
13769 * target.def (empty_mask_is_expensive): New hook.
13770 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
13771 * doc/tm.texi: Regenerate.
13772 * targhooks.h (default_empty_mask_is_expensive): Declare.
13773 * targhooks.c (default_empty_mask_is_expensive): New function.
13774 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
13775 if the target says that empty masks are expensive.
13776 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
13778 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
13780 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13781 Alan Hayward <alan.hayward@arm.com>
13782 David Sherwood <david.sherwood@arm.com>
13784 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
13785 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
13786 (vect_use_loop_mask_for_alignment_p): New function.
13787 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
13788 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
13789 niters_skip argument. Make sure that the first niters_skip elements
13790 of the first iteration are inactive.
13791 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
13792 Update call to vect_set_loop_masks_directly.
13793 (get_misalign_in_elems): New function, split out from...
13794 (vect_gen_prolog_loop_niters): ...here.
13795 (vect_update_init_of_dr): Take a code argument that specifies whether
13796 the adjustment should be added or subtracted.
13797 (vect_update_init_of_drs): Likewise.
13798 (vect_prepare_for_masked_peels): New function.
13799 (vect_do_peeling): Skip prologue peeling if we're using a mask
13800 instead. Update call to vect_update_inits_of_drs.
13801 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
13803 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
13804 alignment. Do not include the number of peeled iterations in
13805 the minimum threshold in that case.
13806 (vectorizable_induction): Adjust the start value down by
13807 LOOP_VINFO_MASK_SKIP_NITERS iterations.
13808 (vect_transform_loop): Call vect_prepare_for_masked_peels.
13809 Take the number of skipped iterations into account when calculating
13811 * tree-vect-stmts.c (vect_gen_while_not): New function.
13813 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13814 Alan Hayward <alan.hayward@arm.com>
13815 David Sherwood <david.sherwood@arm.com>
13817 * doc/sourcebuild.texi (vect_fully_masked): Document.
13818 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
13819 default value to 0.
13820 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
13822 (vect_analyze_loop_2): ...here. Don't check the vectorization
13823 factor against the number of loop iterations if the loop is
13826 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13827 Alan Hayward <alan.hayward@arm.com>
13828 David Sherwood <david.sherwood@arm.com>
13830 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
13831 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
13832 (dump_groups): Update accordingly.
13833 (iv_use::mem_type): New member variable.
13834 (address_p): New function.
13835 (record_use): Add a mem_type argument and initialize the new
13837 (record_group_use): Add a mem_type argument. Use address_p.
13838 Remove obsolete null checks of base_object. Update call to record_use.
13839 (find_interesting_uses_op): Update call to record_group_use.
13840 (find_interesting_uses_cond): Likewise.
13841 (find_interesting_uses_address): Likewise.
13842 (get_mem_type_for_internal_fn): New function.
13843 (find_address_like_use): Likewise.
13844 (find_interesting_uses_stmt): Try find_address_like_use before
13845 calling find_interesting_uses_op.
13846 (addr_offset_valid_p): Use the iv mem_type field as the type
13847 of the addressed memory.
13848 (add_autoinc_candidates): Likewise.
13849 (get_address_cost): Likewise.
13850 (split_small_address_groups_p): Use address_p.
13851 (split_address_groups): Likewise.
13852 (add_iv_candidate_for_use): Likewise.
13853 (autoinc_possible_for_pair): Likewise.
13854 (rewrite_groups): Likewise.
13855 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
13856 (determine_group_iv_cost): Update after split of USE_ADDRESS.
13857 (get_alias_ptr_type_for_ptr_address): New function.
13858 (rewrite_use_address): Rewrite address uses in calls that were
13859 identified by find_address_like_use.
13861 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13862 Alan Hayward <alan.hayward@arm.com>
13863 David Sherwood <david.sherwood@arm.com>
13865 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
13867 * gimple-expr.h (is_gimple_addressable: Likewise.
13868 * gimple-expr.c (is_gimple_address): Likewise.
13869 * internal-fn.c (expand_call_mem_ref): New function.
13870 (expand_mask_load_optab_fn): Use it.
13871 (expand_mask_store_optab_fn): Likewise.
13873 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13874 Alan Hayward <alan.hayward@arm.com>
13875 David Sherwood <david.sherwood@arm.com>
13877 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
13878 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
13879 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
13880 (cond_umax@var{mode}): Document.
13881 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
13882 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
13883 (cond_umin_optab, cond_umax_optab): New optabs.
13884 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
13885 (COND_IOR, COND_XOR): New internal functions.
13886 * internal-fn.h (get_conditional_internal_fn): Declare.
13887 * internal-fn.c (cond_binary_direct): New macro.
13888 (expand_cond_binary_optab_fn): Likewise.
13889 (direct_cond_binary_optab_supported_p): Likewise.
13890 (get_conditional_internal_fn): New function.
13891 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
13892 Cope with reduction statements that are vectorized as calls rather
13894 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
13895 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
13896 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
13897 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
13898 (UNSPEC_COND_EOR): New unspecs.
13899 (optab): Add mappings for them.
13900 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
13901 (sve_int_op, sve_fp_op): New int attributes.
13903 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13904 Alan Hayward <alan.hayward@arm.com>
13905 David Sherwood <david.sherwood@arm.com>
13907 * optabs.def (while_ult_optab): New optab.
13908 * doc/md.texi (while_ult@var{m}@var{n}): Document.
13909 * internal-fn.def (WHILE_ULT): New internal function.
13910 * internal-fn.h (direct_internal_fn_supported_p): New override
13911 that takes two types as argument.
13912 * internal-fn.c (while_direct): New macro.
13913 (expand_while_optab_fn): New function.
13914 (convert_optab_supported_p): Likewise.
13915 (direct_while_optab_supported_p): New macro.
13916 * wide-int.h (wi::udiv_ceil): New function.
13917 * tree-vectorizer.h (rgroup_masks): New structure.
13918 (vec_loop_masks): New typedef.
13919 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
13920 and fully_masked_p.
13921 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
13922 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
13923 (vect_max_vf): New function.
13924 (slpeel_make_loop_iterate_ntimes): Delete.
13925 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
13926 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
13927 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
13928 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
13929 internal-fn.h, stor-layout.h and optabs-query.h.
13930 (vect_set_loop_mask): New function.
13931 (add_preheader_seq): Likewise.
13932 (add_header_seq): Likewise.
13933 (interleave_supported_p): Likewise.
13934 (vect_maybe_permute_loop_masks): Likewise.
13935 (vect_set_loop_masks_directly): Likewise.
13936 (vect_set_loop_condition_masked): Likewise.
13937 (vect_set_loop_condition_unmasked): New function, split out from
13938 slpeel_make_loop_iterate_ntimes.
13939 (slpeel_make_loop_iterate_ntimes): Rename to..
13940 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
13941 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
13942 (vect_do_peeling): Update call accordingly.
13943 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
13945 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
13946 mask_compare_type, can_fully_mask_p and fully_masked_p.
13947 (release_vec_loop_masks): New function.
13948 (_loop_vec_info): Use it to free the loop masks.
13949 (can_produce_all_loop_masks_p): New function.
13950 (vect_get_max_nscalars_per_iter): Likewise.
13951 (vect_verify_full_masking): Likewise.
13952 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
13953 retries, and free the mask rgroups before retrying. Check loop-wide
13954 reasons for disallowing fully-masked loops. Make the final decision
13955 about whether use a fully-masked loop or not.
13956 (vect_estimate_min_profitable_iters): Do not assume that peeling
13957 for the number of iterations will be needed for fully-masked loops.
13958 (vectorizable_reduction): Disable fully-masked loops.
13959 (vectorizable_live_operation): Likewise.
13960 (vect_halve_mask_nunits): New function.
13961 (vect_double_mask_nunits): Likewise.
13962 (vect_record_loop_mask): Likewise.
13963 (vect_get_loop_mask): Likewise.
13964 (vect_transform_loop): Handle the case in which the final loop
13965 iteration might handle a partial vector. Call vect_set_loop_condition
13966 instead of slpeel_make_loop_iterate_ntimes.
13967 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
13968 (check_load_store_masking): New function.
13969 (prepare_load_store_mask): Likewise.
13970 (vectorizable_store): Handle fully-masked loops.
13971 (vectorizable_load): Likewise.
13972 (supportable_widening_operation): Use vect_halve_mask_nunits for
13974 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
13975 (vect_gen_while): New function.
13976 * config/aarch64/aarch64.md (umax<mode>3): New expander.
13977 (aarch64_uqdec<mode>): New insn.
13979 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13980 Alan Hayward <alan.hayward@arm.com>
13981 David Sherwood <david.sherwood@arm.com>
13983 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
13984 (reduc_xor_scal_optab): New optabs.
13985 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
13986 (reduc_xor_scal_@var{m}): Document.
13987 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
13988 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
13989 internal functions.
13990 * fold-const-call.c (fold_const_call): Handle them.
13991 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
13992 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
13993 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
13994 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
13995 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
13996 (UNSPEC_XORV): New unspecs.
13997 (optab): Add entries for them.
13998 (BITWISEV): New int iterator.
13999 (bit_reduc_op): New int attributes.
14001 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14002 Alan Hayward <alan.hayward@arm.com>
14003 David Sherwood <david.sherwood@arm.com>
14005 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
14006 * internal-fn.def (VEC_SHL_INSERT): New internal function.
14007 * optabs.def (vec_shl_insert_optab): New optab.
14008 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
14009 (duplicate_and_interleave): Likewise.
14010 * tree-vect-loop.c: Include internal-fn.h.
14011 (neutral_op_for_slp_reduction): New function, split out from
14012 get_initial_defs_for_reduction.
14013 (get_initial_def_for_reduction): Handle option 2 for variable-length
14014 vectors by loading the neutral value into a vector and then shifting
14015 the initial value into element 0.
14016 (get_initial_defs_for_reduction): Replace the code argument with
14017 the neutral value calculated by neutral_op_for_slp_reduction.
14018 Use gimple_build_vector for constant-length vectors.
14019 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
14020 but the first group_size elements have a neutral value.
14021 Use duplicate_and_interleave otherwise.
14022 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
14023 Update call to get_initial_defs_for_reduction. Handle SLP
14024 reductions for variable-length vectors by creating one vector
14025 result for each scalar result, with the elements associated
14026 with other scalar results stubbed out with the neutral value.
14027 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
14028 Require IFN_VEC_SHL_INSERT for double reductions on
14029 variable-length vectors, or SLP reductions that have
14030 a neutral value. Require can_duplicate_and_interleave_p
14031 support for variable-length unchained SLP reductions if there
14032 is no neutral value, such as for MIN/MAX reductions. Also require
14033 the number of vector elements to be a multiple of the number of
14034 SLP statements when doing variable-length unchained SLP reductions.
14035 Update call to vect_create_epilog_for_reduction.
14036 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
14037 and remove initial values.
14038 (duplicate_and_interleave): Make public.
14039 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
14040 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
14042 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14043 Alan Hayward <alan.hayward@arm.com>
14044 David Sherwood <david.sherwood@arm.com>
14046 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
14047 (can_duplicate_and_interleave_p): New function.
14048 (vect_get_and_check_slp_defs): Take the vector of statements
14049 rather than just the current one. Remove excess parentheses.
14050 Restriction rejectinon of vect_constant_def and vect_external_def
14051 for variable-length vectors to boolean types, or types for which
14052 can_duplicate_and_interleave_p is false.
14053 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
14054 (duplicate_and_interleave): New function.
14055 (vect_get_constant_vectors): Use gimple_build_vector for
14056 constant-length vectors and suitable variable-length constant
14057 vectors. Use duplicate_and_interleave for other variable-length
14058 vectors. Don't defer the update when inserting new statements.
14060 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14061 Alan Hayward <alan.hayward@arm.com>
14062 David Sherwood <david.sherwood@arm.com>
14064 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
14065 min_profitable_iters doesn't go negative.
14067 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14068 Alan Hayward <alan.hayward@arm.com>
14069 David Sherwood <david.sherwood@arm.com>
14071 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
14072 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
14073 * optabs.def (vec_mask_load_lanes_optab): New optab.
14074 (vec_mask_store_lanes_optab): Likewise.
14075 * internal-fn.def (MASK_LOAD_LANES): New internal function.
14076 (MASK_STORE_LANES): Likewise.
14077 * internal-fn.c (mask_load_lanes_direct): New macro.
14078 (mask_store_lanes_direct): Likewise.
14079 (expand_mask_load_optab_fn): Handle masked operations.
14080 (expand_mask_load_lanes_optab_fn): New macro.
14081 (expand_mask_store_optab_fn): Handle masked operations.
14082 (expand_mask_store_lanes_optab_fn): New macro.
14083 (direct_mask_load_lanes_optab_supported_p): Likewise.
14084 (direct_mask_store_lanes_optab_supported_p): Likewise.
14085 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
14087 (vect_load_lanes_supported): Likewise.
14088 * tree-vect-data-refs.c (strip_conversion): New function.
14089 (can_group_stmts_p): Likewise.
14090 (vect_analyze_data_ref_accesses): Use it instead of checking
14091 for a pair of assignments.
14092 (vect_store_lanes_supported): Take a masked_p parameter.
14093 (vect_load_lanes_supported): Likewise.
14094 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
14095 vect_store_lanes_supported and vect_load_lanes_supported.
14096 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
14097 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
14098 parameter. Don't allow gaps for masked accesses.
14099 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
14100 and vect_load_lanes_supported.
14101 (get_load_store_type): Take a masked_p parameter and update
14102 call to get_group_load_store_type.
14103 (vectorizable_store): Update call to get_load_store_type.
14104 Handle IFN_MASK_STORE_LANES.
14105 (vectorizable_load): Update call to get_load_store_type.
14106 Handle IFN_MASK_LOAD_LANES.
14108 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14109 Alan Hayward <alan.hayward@arm.com>
14110 David Sherwood <david.sherwood@arm.com>
14112 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
14114 * config/aarch64/aarch64-protos.h
14115 (aarch64_sve_struct_memory_operand_p): Declare.
14116 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
14117 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
14118 (VPRED, vpred): Handle SVE structure modes.
14119 * config/aarch64/constraints.md (Utx): New constraint.
14120 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
14121 (aarch64_sve_struct_nonimmediate_operand): New predicates.
14122 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
14123 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
14124 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
14125 structure modes. Split into pieces after RA.
14126 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
14127 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
14129 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
14130 SVE structure modes.
14131 (aarch64_classify_address): Likewise.
14132 (sizetochar): Move earlier in file.
14133 (aarch64_print_operand): Handle SVE register lists.
14134 (aarch64_array_mode): New function.
14135 (aarch64_sve_struct_memory_operand_p): Likewise.
14136 (TARGET_ARRAY_MODE): Redefine.
14138 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14139 Alan Hayward <alan.hayward@arm.com>
14140 David Sherwood <david.sherwood@arm.com>
14142 * target.def (array_mode): New target hook.
14143 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
14144 * doc/tm.texi: Regenerate.
14145 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
14146 * hooks.c (hook_optmode_mode_uhwi_none): New function.
14147 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
14148 targetm.array_mode.
14149 * stor-layout.c (mode_for_array): Likewise. Support polynomial
14152 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14153 Alan Hayward <alan.hayward@arm.com>
14154 David Sherwood <david.sherwood@arm.com>
14156 * fold-const.c (fold_binary_loc): Check the argument types
14157 rather than the result type when testing for a vector operation.
14159 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14161 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
14162 * doc/tm.texi: Regenerate.
14164 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14165 Alan Hayward <alan.hayward@arm.com>
14166 David Sherwood <david.sherwood@arm.com>
14168 * doc/invoke.texi (-msve-vector-bits=): Document new option.
14169 (sve): Document new AArch64 extension.
14170 * doc/md.texi (w): Extend the description of the AArch64
14171 constraint to include SVE vectors.
14172 (Upl, Upa): Document new AArch64 predicate constraints.
14173 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
14175 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
14176 (msve-vector-bits=): New option.
14177 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
14178 SVE when these are disabled.
14179 (sve): New extension.
14180 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
14181 modes. Adjust their number of units based on aarch64_sve_vg.
14182 (MAX_BITSIZE_MODE_ANY_MODE): Define.
14183 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
14184 aarch64_addr_query_type.
14185 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
14186 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
14187 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
14188 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
14189 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
14190 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
14191 (aarch64_simd_imm_zero_p): Delete.
14192 (aarch64_check_zero_based_sve_index_immediate): Declare.
14193 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
14194 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
14195 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
14196 (aarch64_sve_float_mul_immediate_p): Likewise.
14197 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
14198 rather than an rtx.
14199 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
14200 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
14201 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
14202 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
14203 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
14204 (aarch64_regmode_natural_size): Likewise.
14205 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
14206 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
14208 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
14209 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
14210 for VG and the SVE predicate registers.
14211 (V_ALIASES): Add a "z"-prefixed alias.
14212 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
14213 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
14214 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
14215 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
14216 (REG_CLASS_NAMES): Add entries for them.
14217 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
14218 and the predicate registers.
14219 (aarch64_sve_vg): Declare.
14220 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
14221 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
14222 (REGMODE_NATURAL_SIZE): Define.
14223 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
14225 * config/aarch64/aarch64.c: Include cfgrtl.h.
14226 (simd_immediate_info): Add a constructor for series vectors,
14227 and an associated step field.
14228 (aarch64_sve_vg): New variable.
14229 (aarch64_dbx_register_number): Handle VG and the predicate registers.
14230 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
14231 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
14232 (VEC_ANY_DATA, VEC_STRUCT): New constants.
14233 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
14234 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
14235 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
14236 (aarch64_get_mask_mode): New functions.
14237 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
14238 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
14239 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
14240 predicate modes and predicate registers. Explicitly restrict
14241 GPRs to modes of 16 bytes or smaller. Only allow FP registers
14242 to store a vector mode if it is recognized by
14243 aarch64_classify_vector_mode.
14244 (aarch64_regmode_natural_size): New function.
14245 (aarch64_hard_regno_caller_save_mode): Return the original mode
14247 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
14248 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
14249 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
14250 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
14252 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
14253 does not overlap dest if the function is frame-related. Handle
14255 (aarch64_split_add_offset): New function.
14256 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
14257 them aarch64_add_offset.
14258 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
14259 and update call to aarch64_sub_sp.
14260 (aarch64_add_cfa_expression): New function.
14261 (aarch64_expand_prologue): Pass extra temporary registers to the
14262 functions above. Handle the case in which we need to emit new
14263 DW_CFA_expressions for registers that were originally saved
14264 relative to the stack pointer, but now have to be expressed
14265 relative to the frame pointer.
14266 (aarch64_output_mi_thunk): Pass extra temporary registers to the
14268 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
14269 IP0 and IP1 values for SVE frames.
14270 (aarch64_expand_vec_series): New function.
14271 (aarch64_expand_sve_widened_duplicate): Likewise.
14272 (aarch64_expand_sve_const_vector): Likewise.
14273 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
14274 Handle SVE constants. Use emit_move_insn to move a force_const_mem
14275 into the register, rather than emitting a SET directly.
14276 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
14277 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
14278 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
14279 (offset_9bit_signed_scaled_p): New functions.
14280 (aarch64_replicate_bitmask_imm): New function.
14281 (aarch64_bitmask_imm): Use it.
14282 (aarch64_cannot_force_const_mem): Reject expressions involving
14283 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
14284 (aarch64_classify_index): Handle SVE indices, by requiring
14285 a plain register index with a scale that matches the element size.
14286 (aarch64_classify_address): Handle SVE addresses. Assert that
14287 the mode of the address is VOIDmode or an integer mode.
14288 Update call to aarch64_classify_symbol.
14289 (aarch64_classify_symbolic_expression): Update call to
14290 aarch64_classify_symbol.
14291 (aarch64_const_vec_all_in_range_p): New function.
14292 (aarch64_print_vector_float_operand): Likewise.
14293 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
14294 "vN" for FP registers with SVE modes. Handle (const ...) vectors
14295 and the FP immediates 1.0 and 0.5.
14296 (aarch64_print_address_internal): Handle SVE addresses.
14297 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
14298 (aarch64_regno_regclass): Handle predicate registers.
14299 (aarch64_secondary_reload): Handle big-endian reloads of SVE
14301 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
14302 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
14303 (aarch64_convert_sve_vector_bits): New function.
14304 (aarch64_override_options): Use it to handle -msve-vector-bits=.
14305 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
14306 rather than an rtx.
14307 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
14308 Handle SVE vector and predicate modes. Accept VL-based constants
14309 that need only one temporary register, and VL offsets that require
14310 no temporary registers.
14311 (aarch64_conditional_register_usage): Mark the predicate registers
14312 as fixed if SVE isn't available.
14313 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
14314 Return true for SVE vector and predicate modes.
14315 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
14316 rather than an unsigned int. Handle SVE modes.
14317 (aarch64_preferred_simd_mode): Update call accordingly. Handle
14319 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
14321 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
14322 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
14323 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
14324 (aarch64_sve_float_mul_immediate_p): New functions.
14325 (aarch64_sve_valid_immediate): New function.
14326 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
14327 Explicitly reject structure modes. Check for INDEX constants.
14328 Handle PTRUE and PFALSE constants.
14329 (aarch64_check_zero_based_sve_index_immediate): New function.
14330 (aarch64_simd_imm_zero_p): Delete.
14331 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
14332 vector modes. Accept constants in the range of CNT[BHWD].
14333 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
14334 ask for an Advanced SIMD mode.
14335 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
14336 (aarch64_simd_vector_alignment): Handle SVE predicates.
14337 (aarch64_vectorize_preferred_vector_alignment): New function.
14338 (aarch64_simd_vector_alignment_reachable): Use it instead of
14340 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
14341 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
14343 (MAX_VECT_LEN): Delete.
14344 (expand_vec_perm_d): Add a vec_flags field.
14345 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
14346 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
14347 (aarch64_evpc_ext): Don't apply a big-endian lane correction
14349 (aarch64_evpc_rev): Rename to...
14350 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
14351 (aarch64_evpc_rev_global): New function.
14352 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
14353 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
14355 (aarch64_evpc_sve_tbl): New function.
14356 (aarch64_expand_vec_perm_const_1): Update after rename of
14357 aarch64_evpc_rev. Handle SVE permutes too, trying
14358 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
14359 than aarch64_evpc_tbl.
14360 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
14361 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
14362 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
14363 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
14364 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
14365 (aarch64_expand_sve_vcond): New functions.
14366 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
14367 of aarch64_vector_mode_p.
14368 (aarch64_dwarf_poly_indeterminate_value): New function.
14369 (aarch64_compute_pressure_classes): Likewise.
14370 (aarch64_can_change_mode_class): Likewise.
14371 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
14372 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
14373 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
14374 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
14375 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
14376 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
14377 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
14378 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
14380 (Dn, Dl, Dr): Accept const as well as const_vector.
14381 (Dz): Likewise. Compare against CONST0_RTX.
14382 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
14383 of "vector" where appropriate.
14384 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
14385 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
14386 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
14387 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
14388 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
14389 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
14390 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
14391 (v_int_equiv): Extend to SVE modes.
14392 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
14394 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
14395 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
14396 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
14397 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
14398 (SVE_COND_FP_CMP): New int iterators.
14399 (perm_hilo): Handle the new unpack unspecs.
14400 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
14402 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
14403 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
14404 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
14405 (aarch64_equality_operator, aarch64_constant_vector_operand)
14406 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
14407 (aarch64_sve_nonimmediate_operand): Likewise.
14408 (aarch64_sve_general_operand): Likewise.
14409 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
14410 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
14411 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
14412 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
14413 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
14414 (aarch64_sve_float_arith_immediate): Likewise.
14415 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
14416 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
14417 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
14418 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
14419 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
14420 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
14421 (aarch64_sve_float_arith_operand): Likewise.
14422 (aarch64_sve_float_arith_with_sub_operand): Likewise.
14423 (aarch64_sve_float_mul_operand): Likewise.
14424 (aarch64_sve_vec_perm_operand): Likewise.
14425 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
14426 (aarch64_mov_operand): Accept const_poly_int and const_vector.
14427 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
14428 as well as const_vector.
14429 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
14430 in file. Use CONST0_RTX and CONSTM1_RTX.
14431 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
14432 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
14433 Use aarch64_simd_imm_zero.
14434 * config/aarch64/aarch64-sve.md: New file.
14435 * config/aarch64/aarch64.md: Include it.
14436 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
14437 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
14438 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
14439 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
14440 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
14441 (sve): New attribute.
14442 (enabled): Disable instructions with the sve attribute unless
14444 (movqi, movhi): Pass CONST_POLY_INT operaneds through
14445 aarch64_expand_mov_immediate.
14446 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
14447 CNT[BHSD] immediates.
14448 (movti): Split CONST_POLY_INT moves into two halves.
14449 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
14450 Split additions that need a temporary here if the destination
14451 is the stack pointer.
14452 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
14453 (*add<mode>3_poly_1): New instruction.
14454 (set_clobber_cc): New expander.
14456 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14458 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
14459 parameter and use it instead of GET_MODE_SIZE (innermode). Use
14460 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
14461 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
14462 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
14463 Change innermode from fixed_mode_size to machine_mode.
14464 (simplify_subreg): Update call accordingly. Handle a constant-sized
14465 subreg of a variable-length CONST_VECTOR.
14467 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14468 Alan Hayward <alan.hayward@arm.com>
14469 David Sherwood <david.sherwood@arm.com>
14471 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
14472 (add_offset_to_base): New function, split out from...
14473 (create_mem_ref): ...here. When handling a scale other than 1,
14474 check first whether the address is valid without the offset.
14475 Add it into the base if so, leaving the index and scale as-is.
14477 2018-01-12 Jakub Jelinek <jakub@redhat.com>
14480 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
14481 fold_for_warn before checking if arg2 is INTEGER_CST.
14483 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
14485 * config/rs6000/predicates.md (load_multiple_operation): Delete.
14486 (store_multiple_operation): Delete.
14487 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
14488 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
14489 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
14490 guarded by TARGET_STRING.
14491 (rs6000_output_load_multiple): Delete.
14492 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
14493 OPTION_MASK_STRING / TARGET_STRING handling.
14494 (print_operand) <'N', 'O'>: Add comment that these are unused now.
14495 (const rs6000_opt_masks) <"string">: Change mask to 0.
14496 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
14497 (MASK_STRING): Delete.
14498 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
14500 (load_multiple): Delete.
14507 (store_multiple): Delete.
14514 (movmemsi_8reg): Delete.
14515 (corresponding unnamed define_insn): Delete.
14516 (movmemsi_6reg): Delete.
14517 (corresponding unnamed define_insn): Delete.
14518 (movmemsi_4reg): Delete.
14519 (corresponding unnamed define_insn): Delete.
14520 (movmemsi_2reg): Delete.
14521 (corresponding unnamed define_insn): Delete.
14522 (movmemsi_1reg): Delete.
14523 (corresponding unnamed define_insn): Delete.
14524 * config/rs6000/rs6000.opt (mno-string): New.
14525 (mstring): Replace by deprecation warning stub.
14526 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
14528 2018-01-12 Jakub Jelinek <jakub@redhat.com>
14530 * regrename.c (regrename_do_replace): If replacing the same
14531 reg multiple times, try to reuse last created gen_raw_REG.
14534 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
14535 main to workaround a bug in GDB.
14537 2018-01-12 Tom de Vries <tom@codesourcery.com>
14540 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
14542 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
14544 PR rtl-optimization/80481
14545 * ira-color.c (get_cap_member): New function.
14546 (allocnos_conflict_by_live_ranges_p): Use it.
14547 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
14548 (setup_slot_coalesced_allocno_live_ranges): Ditto.
14550 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
14553 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
14554 (*saddl_se_1): Ditto.
14555 (*ssubsi_1): Ditto.
14556 (*ssubl_se_1): Ditto.
14558 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
14560 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
14561 rather than wi::to_widest for DR_INITs.
14562 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
14563 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
14564 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
14566 (vect_analyze_group_access_1): Note that here.
14568 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
14570 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
14571 polynomial type sizes.
14573 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
14575 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
14576 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
14577 (gimple_add_tmp_var): Likewise.
14579 2018-01-12 Martin Liska <mliska@suse.cz>
14581 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
14582 (gimple_alloc_sizes): Likewise.
14583 (dump_gimple_statistics): Use PRIu64 in printf format.
14584 * gimple.h: Change uint64_t to int.
14586 2018-01-12 Martin Liska <mliska@suse.cz>
14588 * tree-core.h: Use uint64_t instead of int.
14589 * tree.c (tree_node_counts): Likewise.
14590 (tree_node_sizes): Likewise.
14591 (dump_tree_statistics): Use PRIu64 in printf format.
14593 2018-01-12 Martin Liska <mliska@suse.cz>
14595 * Makefile.in: As qsort_chk is implemented in vec.c, add
14596 vec.o to linkage of gencfn-macros.
14597 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
14598 passing the info to record_node_allocation_statistics.
14599 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
14601 * ggc-common.c (struct ggc_usage): Add operator== and use
14602 it in operator< and compare function.
14603 * mem-stats.h (struct mem_usage): Likewise.
14604 * vec.c (struct vec_usage): Remove operator< and compare
14605 function. Can be simply inherited.
14607 2018-01-12 Martin Jambor <mjambor@suse.cz>
14610 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
14611 * tree-ssa-math-opts.c: Include domwalk.h.
14612 (convert_mult_to_fma_1): New function.
14613 (fma_transformation_info): New type.
14614 (fma_deferring_state): Likewise.
14615 (cancel_fma_deferring): New function.
14616 (result_of_phi): Likewise.
14617 (last_fma_candidate_feeds_initial_phi): Likewise.
14618 (convert_mult_to_fma): Added deferring logic, split actual
14619 transformation to convert_mult_to_fma_1.
14620 (math_opts_dom_walker): New type.
14621 (math_opts_dom_walker::after_dom_children): New method, body moved
14622 here from pass_optimize_widening_mul::execute, added deferring logic
14624 (pass_optimize_widening_mul::execute): Moved most of code to
14625 math_opts_dom_walker::after_dom_children.
14626 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
14627 * config/i386/i386.c (ix86_option_override_internal): Added
14628 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
14630 2018-01-12 Richard Biener <rguenther@suse.de>
14633 * dwarf2out.c (gen_variable_die): Do not reset old_die for
14634 inline instance vars.
14636 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
14639 * config/rx/rx.c (rx_is_restricted_memory_address):
14640 Handle SUBREG case.
14642 2018-01-12 Richard Biener <rguenther@suse.de>
14644 PR tree-optimization/80846
14645 * target.def (split_reduction): New target hook.
14646 * targhooks.c (default_split_reduction): New function.
14647 * targhooks.h (default_split_reduction): Declare.
14648 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
14649 target requests first reduce vectors by combining low and high
14651 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
14652 (get_vectype_for_scalar_type_and_size): Export.
14653 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
14654 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
14655 * doc/tm.texi: Regenerate.
14656 * config/i386/i386.c (ix86_split_reduction): Implement
14657 TARGET_VECTORIZE_SPLIT_REDUCTION.
14659 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
14662 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
14663 in PIC mode except for TARGET_VXWORKS_RTP.
14664 * config/sparc/sparc.c: Include cfgrtl.h.
14665 (TARGET_INIT_PIC_REG): Define.
14666 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
14667 (sparc_pic_register_p): New predicate.
14668 (sparc_legitimate_address_p): Use it.
14669 (sparc_legitimize_pic_address): Likewise.
14670 (sparc_delegitimize_address): Likewise.
14671 (sparc_mode_dependent_address_p): Likewise.
14672 (gen_load_pcrel_sym): Remove 4th parameter.
14673 (load_got_register): Adjust call to above. Remove obsolete stuff.
14674 (sparc_expand_prologue): Do not call load_got_register here.
14675 (sparc_flat_expand_prologue): Likewise.
14676 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
14677 (sparc_use_pseudo_pic_reg): New function.
14678 (sparc_init_pic_reg): Likewise.
14679 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
14680 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
14682 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
14684 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
14685 Add item for branch_cost.
14687 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
14689 PR rtl-optimization/83565
14690 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
14691 not extend the result to a larger mode for rotate operations.
14692 (num_sign_bit_copies1): Likewise.
14694 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
14697 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
14699 Use values-Xc.o for -pedantic.
14700 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
14702 2018-01-12 Martin Liska <mliska@suse.cz>
14705 * ipa-devirt.c (final_warning_record::grow_type_warnings):
14707 (possible_polymorphic_call_targets): Use it.
14708 (ipa_devirt): Likewise.
14710 2018-01-12 Martin Liska <mliska@suse.cz>
14712 * profile-count.h (enum profile_quality): Use 0 as invalid
14713 enum value of profile_quality.
14715 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
14717 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
14718 -mext-string options.
14720 2018-01-12 Richard Biener <rguenther@suse.de>
14722 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
14723 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
14724 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
14726 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
14728 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
14730 * configure.ac (--with-long-double-format): Add support for the
14731 configuration option to change the default long double format on
14733 * config.gcc (powerpc*-linux*-*): Likewise.
14734 * configure: Regenerate.
14735 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
14736 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
14737 used without modification.
14739 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14741 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
14742 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
14743 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
14744 MISC_BUILTIN_SPEC_BARRIER.
14745 (rs6000_init_builtins): Likewise.
14746 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
14748 (speculation_barrier): New define_insn.
14749 * doc/extend.texi: Document __builtin_speculation_barrier.
14751 2018-01-11 Jakub Jelinek <jakub@redhat.com>
14754 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
14755 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
14756 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
14758 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
14759 integral modes instead of "ss" and "sd".
14760 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
14761 vectors with 32-bit and 64-bit elements.
14762 (vecdupssescalarmodesuffix): New mode attribute.
14763 (vec_dup<mode>): Use it.
14765 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
14768 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
14769 frame if argument is passed on stack.
14771 2018-01-11 Jakub Jelinek <jakub@redhat.com>
14774 * ree.c (combine_reaching_defs): Optimize also
14775 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
14776 reg2=any_extend(exp); reg1=reg2;, formatting fix.
14778 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
14780 PR middle-end/83189
14781 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
14783 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
14785 PR middle-end/83718
14786 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
14787 after they are computed.
14789 2018-01-11 Bin Cheng <bin.cheng@arm.com>
14791 PR tree-optimization/83695
14792 * gimple-loop-linterchange.cc
14793 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
14794 reset cached scev information after interchange.
14795 (pass_linterchange::execute): Remove call to scev_reset_htab.
14797 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14799 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
14800 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
14801 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
14802 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
14803 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
14804 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
14805 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
14806 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
14807 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
14808 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
14809 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
14810 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
14811 (V_lane_reg): Likewise.
14812 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
14814 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
14815 (vfmal_lane_low<mode>_intrinsic,
14816 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
14817 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
14818 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
14819 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
14820 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
14821 vfmsl_lane_high<mode>_intrinsic): New define_insns.
14823 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14825 * config/arm/arm-cpus.in (fp16fml): New feature.
14826 (ALL_SIMD): Add fp16fml.
14827 (armv8.2-a): Add fp16fml as an option.
14828 (armv8.3-a): Likewise.
14829 (armv8.4-a): Add fp16fml as part of fp16.
14830 * config/arm/arm.h (TARGET_FP16FML): Define.
14831 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
14833 * config/arm/arm-modes.def (V2HF): Define.
14834 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
14835 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
14836 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
14837 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
14838 vfmsl_low, vfmsl_high): New set of builtins.
14839 * config/arm/iterators.md (PLUSMINUS): New code iterator.
14840 (vfml_op): New code attribute.
14841 (VFMLHALVES): New int iterator.
14842 (VFML, VFMLSEL): New mode attributes.
14843 (V_reg): Define mapping for V2HF.
14844 (V_hi, V_lo): New mode attributes.
14845 (VF_constraint): Likewise.
14846 (vfml_half, vfml_half_selector): New int attributes.
14847 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
14849 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
14850 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
14852 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
14853 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
14854 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
14855 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
14857 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
14858 Document new effective target and option set.
14860 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14862 * config/arm/arm-cpus.in (armv8_4): New feature.
14863 (ARMv8_4a): New fgroup.
14864 (armv8.4-a): New arch.
14865 * config/arm/arm-tables.opt: Regenerate.
14866 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
14867 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
14868 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
14869 Add matching rules for -march=armv8.4-a and extensions.
14870 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
14872 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
14875 * config/rx/rx.md (BW): New mode attribute.
14876 (sync_lock_test_and_setsi): Add mode suffix to insn output.
14878 2018-01-11 Richard Biener <rguenther@suse.de>
14880 PR tree-optimization/83435
14881 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
14882 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
14883 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
14885 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14886 Alan Hayward <alan.hayward@arm.com>
14887 David Sherwood <david.sherwood@arm.com>
14889 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
14891 (aarch64_classify_address): Initialize it. Track polynomial offsets.
14892 (aarch64_print_address_internal): Use it to check for a zero offset.
14894 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14895 Alan Hayward <alan.hayward@arm.com>
14896 David Sherwood <david.sherwood@arm.com>
14898 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
14899 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
14900 Return a poly_int64 rather than a HOST_WIDE_INT.
14901 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
14902 rather than a HOST_WIDE_INT.
14903 * config/aarch64/aarch64.h (aarch64_frame): Protect with
14904 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
14905 hard_fp_offset, frame_size, initial_adjust, callee_offset and
14906 final_offset from HOST_WIDE_INT to poly_int64.
14907 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
14908 to_constant when getting the number of units in an Advanced SIMD
14910 (aarch64_builtin_vectorized_function): Check for a constant number
14912 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
14914 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
14915 attribute instead of GET_MODE_NUNITS.
14916 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
14917 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
14918 GET_MODE_SIZE for fixed-size registers.
14919 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
14920 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
14921 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
14922 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
14923 (aarch64_print_operand, aarch64_print_address_internal)
14924 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
14925 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
14926 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
14927 Handle polynomial GET_MODE_SIZE.
14928 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
14929 wider than SImode without modification.
14930 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
14931 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
14932 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
14933 passing and returning SVE modes.
14934 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
14935 rather than GEN_INT.
14936 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
14937 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
14938 (aarch64_allocate_and_probe_stack_space): Likewise.
14939 (aarch64_layout_frame): Cope with polynomial offsets.
14940 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
14941 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
14942 polynomial offsets.
14943 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
14944 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
14945 poly_int64 rather than a HOST_WIDE_INT.
14946 (aarch64_get_separate_components, aarch64_process_components)
14947 (aarch64_expand_prologue, aarch64_expand_epilogue)
14948 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
14949 (aarch64_anchor_offset): New function, split out from...
14950 (aarch64_legitimize_address): ...here.
14951 (aarch64_builtin_vectorization_cost): Handle polynomial
14952 TYPE_VECTOR_SUBPARTS.
14953 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
14955 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
14956 number of elements from the PARALLEL rather than the mode.
14957 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
14958 rather than GET_MODE_BITSIZE.
14959 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
14960 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
14961 (aarch64_expand_vec_perm_const_1): Handle polynomial
14962 d->perm.length () and d->perm elements.
14963 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
14964 Apply to_constant to d->perm elements.
14965 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
14966 polynomial CONST_VECTOR_NUNITS.
14967 (aarch64_move_pointer): Take amount as a poly_int64 rather
14969 (aarch64_progress_pointer): Avoid temporary variable.
14970 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
14971 the mode attribute instead of GET_MODE.
14973 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14974 Alan Hayward <alan.hayward@arm.com>
14975 David Sherwood <david.sherwood@arm.com>
14977 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
14978 x exists before using it.
14979 (aarch64_add_constant_internal): Rename to...
14980 (aarch64_add_offset_1): ...this. Replace regnum with separate
14981 src and dest rtxes. Handle the case in which they're different,
14982 including when the offset is zero. Replace scratchreg with an rtx.
14983 Use 2 additions if there is no spare register into which we can
14984 move a 16-bit constant.
14985 (aarch64_add_constant): Delete.
14986 (aarch64_add_offset): Replace reg with separate src and dest
14987 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
14988 Use aarch64_add_offset_1.
14989 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
14990 an rtx rather than an int. Take the delta as a poly_int64
14991 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
14992 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
14993 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
14994 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
14995 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
14996 and aarch64_add_sp.
14997 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
14998 aarch64_add_constant.
15000 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
15002 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
15003 Use scalar_float_mode.
15005 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
15007 * config/aarch64/aarch64-simd.md
15008 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
15009 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
15010 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
15011 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
15012 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
15013 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
15014 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
15015 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
15016 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
15017 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
15019 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
15022 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
15023 targ_options->x_arm_arch_string is non NULL.
15025 2018-01-11 Tamar Christina <tamar.christina@arm.com>
15027 * config/aarch64/aarch64.h
15028 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
15030 2018-01-11 Sudakshina Das <sudi.das@arm.com>
15033 * expmed.c (emit_store_flag_force): Swap if const op0
15034 and change VOIDmode to mode of op0.
15036 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
15038 PR rtl-optimization/83761
15039 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
15040 than bytes to mode_for_size.
15042 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
15044 PR middle-end/83189
15045 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
15046 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
15049 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
15051 PR middle-end/83575
15052 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
15053 when in layout mode.
15054 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
15055 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
15058 2018-01-10 Michael Collison <michael.collison@arm.com>
15060 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
15061 * config/aarch64/aarch64-option-extension.def: Add
15062 AARCH64_OPT_EXTENSION of 'fp16fml'.
15063 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
15064 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
15065 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
15066 * config/aarch64/constraints.md (Ui7): New constraint.
15067 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
15068 (VFMLA_SEL_W): Ditto.
15071 (VFMLA16_LOW): New int iterator.
15072 (VFMLA16_HIGH): Ditto.
15073 (UNSPEC_FMLAL): New unspec.
15074 (UNSPEC_FMLSL): Ditto.
15075 (UNSPEC_FMLAL2): Ditto.
15076 (UNSPEC_FMLSL2): Ditto.
15077 (f16mac): New code attribute.
15078 * config/aarch64/aarch64-simd-builtins.def
15079 (aarch64_fmlal_lowv2sf): Ditto.
15080 (aarch64_fmlsl_lowv2sf): Ditto.
15081 (aarch64_fmlalq_lowv4sf): Ditto.
15082 (aarch64_fmlslq_lowv4sf): Ditto.
15083 (aarch64_fmlal_highv2sf): Ditto.
15084 (aarch64_fmlsl_highv2sf): Ditto.
15085 (aarch64_fmlalq_highv4sf): Ditto.
15086 (aarch64_fmlslq_highv4sf): Ditto.
15087 (aarch64_fmlal_lane_lowv2sf): Ditto.
15088 (aarch64_fmlsl_lane_lowv2sf): Ditto.
15089 (aarch64_fmlal_laneq_lowv2sf): Ditto.
15090 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
15091 (aarch64_fmlalq_lane_lowv4sf): Ditto.
15092 (aarch64_fmlsl_lane_lowv4sf): Ditto.
15093 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
15094 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
15095 (aarch64_fmlal_lane_highv2sf): Ditto.
15096 (aarch64_fmlsl_lane_highv2sf): Ditto.
15097 (aarch64_fmlal_laneq_highv2sf): Ditto.
15098 (aarch64_fmlsl_laneq_highv2sf): Ditto.
15099 (aarch64_fmlalq_lane_highv4sf): Ditto.
15100 (aarch64_fmlsl_lane_highv4sf): Ditto.
15101 (aarch64_fmlalq_laneq_highv4sf): Ditto.
15102 (aarch64_fmlsl_laneq_highv4sf): Ditto.
15103 * config/aarch64/aarch64-simd.md:
15104 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
15105 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
15106 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
15107 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
15108 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
15109 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
15110 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
15111 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
15112 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
15113 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
15114 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
15115 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
15116 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
15117 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
15118 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
15119 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
15120 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
15121 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
15122 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
15123 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
15124 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
15125 (vfmlsl_low_u32): Ditto.
15126 (vfmlalq_low_u32): Ditto.
15127 (vfmlslq_low_u32): Ditto.
15128 (vfmlal_high_u32): Ditto.
15129 (vfmlsl_high_u32): Ditto.
15130 (vfmlalq_high_u32): Ditto.
15131 (vfmlslq_high_u32): Ditto.
15132 (vfmlal_lane_low_u32): Ditto.
15133 (vfmlsl_lane_low_u32): Ditto.
15134 (vfmlal_laneq_low_u32): Ditto.
15135 (vfmlsl_laneq_low_u32): Ditto.
15136 (vfmlalq_lane_low_u32): Ditto.
15137 (vfmlslq_lane_low_u32): Ditto.
15138 (vfmlalq_laneq_low_u32): Ditto.
15139 (vfmlslq_laneq_low_u32): Ditto.
15140 (vfmlal_lane_high_u32): Ditto.
15141 (vfmlsl_lane_high_u32): Ditto.
15142 (vfmlal_laneq_high_u32): Ditto.
15143 (vfmlsl_laneq_high_u32): Ditto.
15144 (vfmlalq_lane_high_u32): Ditto.
15145 (vfmlslq_lane_high_u32): Ditto.
15146 (vfmlalq_laneq_high_u32): Ditto.
15147 (vfmlslq_laneq_high_u32): Ditto.
15148 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
15149 (AARCH64_FL_FOR_ARCH8_4): New.
15150 (AARCH64_ISA_F16FML): New ISA flag.
15151 (TARGET_F16FML): New feature flag for fp16fml.
15152 (doc/invoke.texi): Document new fp16fml option.
15154 2018-01-10 Michael Collison <michael.collison@arm.com>
15156 * config/aarch64/aarch64-builtins.c:
15157 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
15158 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
15159 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
15160 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
15161 (AARCH64_ISA_SHA3): New ISA flag.
15162 (TARGET_SHA3): New feature flag for sha3.
15163 * config/aarch64/iterators.md (sha512_op): New int attribute.
15164 (CRYPTO_SHA512): New int iterator.
15165 (UNSPEC_SHA512H): New unspec.
15166 (UNSPEC_SHA512H2): Ditto.
15167 (UNSPEC_SHA512SU0): Ditto.
15168 (UNSPEC_SHA512SU1): Ditto.
15169 * config/aarch64/aarch64-simd-builtins.def
15170 (aarch64_crypto_sha512hqv2di): New builtin.
15171 (aarch64_crypto_sha512h2qv2di): Ditto.
15172 (aarch64_crypto_sha512su0qv2di): Ditto.
15173 (aarch64_crypto_sha512su1qv2di): Ditto.
15174 (aarch64_eor3qv8hi): Ditto.
15175 (aarch64_rax1qv2di): Ditto.
15176 (aarch64_xarqv2di): Ditto.
15177 (aarch64_bcaxqv8hi): Ditto.
15178 * config/aarch64/aarch64-simd.md:
15179 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
15180 (aarch64_crypto_sha512su0qv2di): Ditto.
15181 (aarch64_crypto_sha512su1qv2di): Ditto.
15182 (aarch64_eor3qv8hi): Ditto.
15183 (aarch64_rax1qv2di): Ditto.
15184 (aarch64_xarqv2di): Ditto.
15185 (aarch64_bcaxqv8hi): Ditto.
15186 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
15187 (vsha512h2q_u64): Ditto.
15188 (vsha512su0q_u64): Ditto.
15189 (vsha512su1q_u64): Ditto.
15190 (veor3q_u16): Ditto.
15191 (vrax1q_u64): Ditto.
15192 (vxarq_u64): Ditto.
15193 (vbcaxq_u16): Ditto.
15194 * config/arm/types.md (crypto_sha512): New type attribute.
15195 (crypto_sha3): Ditto.
15196 (doc/invoke.texi): Document new sha3 option.
15198 2018-01-10 Michael Collison <michael.collison@arm.com>
15200 * config/aarch64/aarch64-builtins.c:
15201 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
15202 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
15203 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
15204 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
15205 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
15206 (AARCH64_ISA_SM4): New ISA flag.
15207 (TARGET_SM4): New feature flag for sm4.
15208 * config/aarch64/aarch64-simd-builtins.def
15209 (aarch64_sm3ss1qv4si): Ditto.
15210 (aarch64_sm3tt1aq4si): Ditto.
15211 (aarch64_sm3tt1bq4si): Ditto.
15212 (aarch64_sm3tt2aq4si): Ditto.
15213 (aarch64_sm3tt2bq4si): Ditto.
15214 (aarch64_sm3partw1qv4si): Ditto.
15215 (aarch64_sm3partw2qv4si): Ditto.
15216 (aarch64_sm4eqv4si): Ditto.
15217 (aarch64_sm4ekeyqv4si): Ditto.
15218 * config/aarch64/aarch64-simd.md:
15219 (aarch64_sm3ss1qv4si): Ditto.
15220 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
15221 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
15222 (aarch64_sm4eqv4si): Ditto.
15223 (aarch64_sm4ekeyqv4si): Ditto.
15224 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
15225 (sm3part_op): Ditto.
15226 (CRYPTO_SM3TT): Ditto.
15227 (CRYPTO_SM3PART): Ditto.
15228 (UNSPEC_SM3SS1): New unspec.
15229 (UNSPEC_SM3TT1A): Ditto.
15230 (UNSPEC_SM3TT1B): Ditto.
15231 (UNSPEC_SM3TT2A): Ditto.
15232 (UNSPEC_SM3TT2B): Ditto.
15233 (UNSPEC_SM3PARTW1): Ditto.
15234 (UNSPEC_SM3PARTW2): Ditto.
15235 (UNSPEC_SM4E): Ditto.
15236 (UNSPEC_SM4EKEY): Ditto.
15237 * config/aarch64/constraints.md (Ui2): New constraint.
15238 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
15239 * config/arm/types.md (crypto_sm3): New type attribute.
15240 (crypto_sm4): Ditto.
15241 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
15242 (vsm3tt1aq_u32): Ditto.
15243 (vsm3tt1bq_u32): Ditto.
15244 (vsm3tt2aq_u32): Ditto.
15245 (vsm3tt2bq_u32): Ditto.
15246 (vsm3partw1q_u32): Ditto.
15247 (vsm3partw2q_u32): Ditto.
15248 (vsm4eq_u32): Ditto.
15249 (vsm4ekeyq_u32): Ditto.
15250 (doc/invoke.texi): Document new sm4 option.
15252 2018-01-10 Michael Collison <michael.collison@arm.com>
15254 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
15255 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
15256 (AARCH64_FL_FOR_ARCH8_4): New.
15257 (AARCH64_FL_V8_4): New flag.
15258 (doc/invoke.texi): Document new armv8.4-a option.
15260 2018-01-10 Michael Collison <michael.collison@arm.com>
15262 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
15263 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
15264 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
15265 * config/aarch64/aarch64-option-extension.def: Add
15266 AARCH64_OPT_EXTENSION of 'sha2'.
15267 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
15268 (crypto): Disable sha2 and aes if crypto disabled.
15269 (crypto): Enable aes and sha2 if enabled.
15270 (simd): Disable sha2 and aes if simd disabled.
15271 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
15273 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
15274 (TARGET_SHA2): New feature flag for sha2.
15275 (TARGET_AES): New feature flag for aes.
15276 * config/aarch64/aarch64-simd.md:
15277 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
15278 conditional on TARGET_AES.
15279 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
15280 (aarch64_crypto_sha1hsi): Make pattern conditional
15282 (aarch64_crypto_sha1hv4si): Ditto.
15283 (aarch64_be_crypto_sha1hv4si): Ditto.
15284 (aarch64_crypto_sha1su1v4si): Ditto.
15285 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
15286 (aarch64_crypto_sha1su0v4si): Ditto.
15287 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
15288 (aarch64_crypto_sha256su0v4si): Ditto.
15289 (aarch64_crypto_sha256su1v4si): Ditto.
15290 (doc/invoke.texi): Document new aes and sha2 options.
15292 2018-01-10 Martin Sebor <msebor@redhat.com>
15294 PR tree-optimization/83781
15295 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
15298 2018-01-11 Martin Sebor <msebor@gmail.com>
15299 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
15301 PR tree-optimization/83501
15302 PR tree-optimization/81703
15304 * tree-ssa-strlen.c (get_string_cst): Rename...
15305 (get_string_len): ...to this. Handle global constants.
15306 (handle_char_store): Adjust.
15308 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
15309 Jim Wilson <jimw@sifive.com>
15311 * config/riscv/riscv-protos.h (riscv_output_return): New.
15312 * config/riscv/riscv.c (struct machine_function): New naked_p field.
15313 (riscv_attribute_table, riscv_output_return),
15314 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
15315 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
15316 (riscv_compute_frame_info): Only compute frame->mask if not a naked
15318 (riscv_expand_prologue): Add early return for naked function.
15319 (riscv_expand_epilogue): Likewise.
15320 (riscv_function_ok_for_sibcall): Return false for naked function.
15321 (riscv_set_current_function): New.
15322 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
15323 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
15324 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
15325 * doc/extend.texi (RISC-V Function Attributes): New.
15327 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
15329 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
15330 check for 128-bit long double before checking TCmode.
15331 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
15332 128-bit long doubles before checking TFmode or TCmode.
15333 (FLOAT128_IBM_P): Likewise.
15335 2018-01-10 Martin Sebor <msebor@redhat.com>
15337 PR tree-optimization/83671
15338 * builtins.c (c_strlen): Unconditionally return zero for the empty
15340 Use -Warray-bounds for warnings.
15341 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
15342 for non-constant array indices with COMPONENT_REF, arrays of
15343 arrays, and pointers to arrays.
15344 (gimple_fold_builtin_strlen): Determine and set length range for
15345 non-constant character arrays.
15347 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
15349 PR middle-end/81897
15350 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
15353 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
15355 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
15357 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
15360 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
15361 VECTOR_MEM_ALTIVEC_OR_VSX_P.
15362 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
15363 indexed_or_indirect_operand predicate.
15364 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
15365 (*vsx_le_perm_load_v8hi): Likewise.
15366 (*vsx_le_perm_load_v16qi): Likewise.
15367 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
15368 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
15369 (*vsx_le_perm_store_v8hi): Likewise.
15370 (*vsx_le_perm_store_v16qi): Likewise.
15371 (eight unnamed splitters): Likewise.
15373 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
15375 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
15376 * config/rs6000/emmintrin.h: Likewise.
15377 * config/rs6000/mmintrin.h: Likewise.
15378 * config/rs6000/xmmintrin.h: Likewise.
15380 2018-01-10 David Malcolm <dmalcolm@redhat.com>
15383 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
15385 * tree.c (tree_nop_conversion): Return true for location wrapper
15387 (maybe_wrap_with_location): New function.
15388 (selftest::check_strip_nops): New function.
15389 (selftest::test_location_wrappers): New function.
15390 (selftest::tree_c_tests): Call it.
15391 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
15392 (maybe_wrap_with_location): New decl.
15393 (EXPR_LOCATION_WRAPPER_P): New macro.
15394 (location_wrapper_p): New inline function.
15395 (tree_strip_any_location_wrapper): New inline function.
15397 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
15400 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
15401 stack_realign_offset for the largest alignment of stack slot
15403 (ix86_find_max_used_stack_alignment): New function.
15404 (ix86_finalize_stack_frame_flags): Use it. Set
15405 max_used_stack_alignment if we don't realign stack.
15406 * config/i386/i386.h (machine_function): Add
15407 max_used_stack_alignment.
15409 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
15411 * config/arm/arm.opt (-mbranch-cost): New option.
15412 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
15415 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
15418 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
15419 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
15421 2018-01-10 Richard Biener <rguenther@suse.de>
15424 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
15425 early out so it also covers the case where we have a non-NULL
15428 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
15430 PR tree-optimization/83753
15431 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
15432 for non-strided grouped accesses if the number of elements is 1.
15434 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
15437 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
15438 * i386.h (TARGET_USE_GATHER): Define.
15439 * x86-tune.def (X86_TUNE_USE_GATHER): New.
15441 2018-01-10 Martin Liska <mliska@suse.cz>
15444 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
15445 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
15447 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
15448 CLEANUP_NO_PARTITIONING is not set.
15450 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
15452 * doc/rtl.texi: Remove documentation of (const ...) wrappers
15453 for vectors, as a partial revert of r254296.
15454 * rtl.h (const_vec_p): Delete.
15455 (const_vec_duplicate_p): Don't test for vector CONSTs.
15456 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
15457 * expmed.c (make_tree): Likewise.
15460 * common.md (E, F): Use CONSTANT_P instead of checking for
15462 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
15463 checking for CONST_VECTOR.
15465 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
15467 PR middle-end/83575
15468 * predict.c (force_edge_cold): Handle in more sane way edges
15469 with no prediction.
15471 2018-01-09 Carl Love <cel@us.ibm.com>
15473 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
15475 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
15476 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
15477 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
15478 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
15479 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
15480 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
15481 * config/rs6000/rs6000-protos.h: Add extern defition for
15482 rs6000_generate_float2_double_code.
15483 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
15485 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
15486 (float2_v2df): Add define_expand.
15488 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
15491 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
15492 op_mode in the force_to_mode call.
15494 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
15496 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
15497 instead of checking each element individually.
15498 (aarch64_evpc_uzp): Likewise.
15499 (aarch64_evpc_zip): Likewise.
15500 (aarch64_evpc_ext): Likewise.
15501 (aarch64_evpc_rev): Likewise.
15502 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
15503 instead of checking each element individually. Return true without
15505 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
15506 whether all selected elements come from the same input, instead of
15507 checking each element individually. Remove calls to gen_rtx_REG,
15508 start_sequence and end_sequence and instead assert that no rtl is
15511 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
15513 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
15514 order of HIGH and CONST checks.
15516 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
15518 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
15519 if the destination isn't an SSA_NAME.
15521 2018-01-09 Richard Biener <rguenther@suse.de>
15523 PR tree-optimization/83668
15524 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
15526 (canonicalize_loop_form): ... here, renamed from ...
15527 (canonicalize_loop_closed_ssa_form): ... this and amended to
15528 swap successor edges for loop exit blocks to make us use
15529 the RPO order we need for initial schedule generation.
15531 2018-01-09 Joseph Myers <joseph@codesourcery.com>
15533 PR tree-optimization/64811
15534 * match.pd: When optimizing comparisons with Inf, avoid
15535 introducing or losing exceptions from comparisons with NaN.
15537 2018-01-09 Martin Liska <mliska@suse.cz>
15540 * asan.c (shadow_mem_size): Add gcc_assert.
15542 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
15544 Don't save registers in main().
15547 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
15548 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
15549 * config/avr/avr.c (avr_set_current_function): Don't error if
15550 naked, OS_task or OS_main are specified at the same time.
15551 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
15553 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
15555 * common/config/avr/avr-common.c (avr_option_optimization_table):
15556 Switch on -mmain-is-OS_task for optimizing compilations.
15558 2018-01-09 Richard Biener <rguenther@suse.de>
15560 PR tree-optimization/83572
15561 * graphite.c: Include cfganal.h.
15562 (graphite_transform_loops): Connect infinite loops to exit
15563 and remove fake edges at the end.
15565 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
15567 * ipa-inline.c (edge_badness): Revert accidental checkin.
15569 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
15572 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
15573 symbols; not inline clones.
15575 2018-01-09 Jakub Jelinek <jakub@redhat.com>
15578 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
15579 hard registers. Formatting fixes.
15581 PR preprocessor/83722
15582 * gcc.c (try_generate_repro): Pass
15583 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
15584 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
15587 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
15588 Kito Cheng <kito.cheng@gmail.com>
15590 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
15591 (riscv_leaf_function_p): Delete.
15592 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
15594 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
15596 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
15598 (do_ifelse): New function.
15599 (do_isel): New function.
15600 (do_sub3): New function.
15601 (do_add3): New function.
15602 (do_load_mask_compare): New function.
15603 (do_overlap_load_compare): New function.
15604 (expand_compare_loop): New function.
15605 (expand_block_compare): Call expand_compare_loop() when appropriate.
15606 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
15607 option description.
15608 (-mblock-compare-inline-loop-limit): New option.
15610 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
15613 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
15614 Reverse order of second and third operands in first alternative.
15615 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
15616 of first and second elements in UNSPEC_VPERMR vector.
15617 (altivec_expand_vec_perm_le): Likewise.
15619 2018-01-08 Jeff Law <law@redhat.com>
15621 PR rtl-optimizatin/81308
15622 * tree-switch-conversion.c (cfg_altered): New file scoped static.
15623 (process_switch): If group_case_labels makes a change, then set
15625 (pass_convert_switch::execute): If a switch is converted, then
15626 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
15628 PR rtl-optimization/81308
15629 * recog.c (split_all_insns): Conditionally cleanup the CFG after
15632 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
15634 PR target/83663 - Revert r255946
15635 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
15636 generation for cases where splatting a value is not useful.
15637 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
15638 across a vec_duplicate and a paradoxical subreg forming a vector
15639 mode to a vec_concat.
15641 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15643 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
15644 -march=armv8.3-a variants.
15645 * config/arm/t-multilib: Likewise.
15646 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
15648 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
15650 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
15652 (cceq_ior_compare_complement): Give it a name so I can use it, and
15653 change boolean_or_operator predicate to boolean_operator so it can
15654 be used to generate a crand.
15655 (eqne): New code iterator.
15656 (bd/bd_neg): New code_attrs.
15657 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
15658 a single define_insn.
15659 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
15660 decrement (bdnzt/bdnzf/bdzt/bdzf).
15661 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
15662 with the new names of the branch decrement patterns, and added the
15663 names of the branch decrement conditional patterns.
15665 2018-01-08 Richard Biener <rguenther@suse.de>
15667 PR tree-optimization/83563
15668 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
15671 2018-01-08 Richard Biener <rguenther@suse.de>
15673 PR middle-end/83713
15674 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
15676 2018-01-08 Richard Biener <rguenther@suse.de>
15678 PR tree-optimization/83685
15679 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
15680 references to abnormals.
15682 2018-01-08 Richard Biener <rguenther@suse.de>
15685 * dwarf2out.c (output_indirect_strings): Handle empty
15686 skeleton_debug_str_hash.
15687 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
15689 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
15691 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
15692 (emit_store_direct): Likewise.
15693 (arc_trampoline_adjust_address): Likewise.
15694 (arc_asm_trampoline_template): New function.
15695 (arc_initialize_trampoline): Use asm_trampoline_template.
15696 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
15697 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
15698 * config/arc/arc.md (flush_icache): Delete pattern.
15700 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
15702 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
15703 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
15706 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
15709 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
15710 by not USED_FOR_TARGET.
15711 (make_pass_resolve_sw_modes): Likewise.
15713 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
15715 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
15718 2018-01-08 Richard Biener <rguenther@suse.de>
15720 PR middle-end/83580
15721 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
15723 2018-01-08 Richard Biener <rguenther@suse.de>
15725 PR middle-end/83517
15726 * match.pd ((t * 2) / 2) -> t): Add missing :c.
15728 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
15730 PR middle-end/81897
15731 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
15732 basic blocks with a small number of successors.
15733 (convert_control_dep_chain_into_preds): Improve handling of
15735 (dump_predicates): Split apart into...
15736 (dump_pred_chain): ...here...
15737 (dump_pred_info): ...and here.
15738 (can_one_predicate_be_invalidated_p): Add debugging printfs.
15739 (can_chain_union_be_invalidated_p): Improve check for invalidation
15741 (uninit_uses_cannot_happen): Avoid unnecessary if
15742 convert_control_dep_chain_into_preds yielded nothing.
15744 2018-01-06 Martin Sebor <msebor@redhat.com>
15746 PR tree-optimization/83640
15747 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
15748 subtracting negative offset from size.
15749 (builtin_access::overlap): Adjust offset bounds of the access to fall
15750 within the size of the object if possible.
15752 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
15754 PR rtl-optimization/83699
15755 * expmed.c (extract_bit_field_1): Restrict the vector usage of
15756 extract_bit_field_as_subreg to cases in which the extracted
15757 value is also a vector.
15759 * lra-constraints.c (process_alt_operands): Test for the equivalence
15760 substitutions when detecting a possible reload cycle.
15762 2018-01-06 Jakub Jelinek <jakub@redhat.com>
15765 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
15766 by default if flag_selective_schedling{,2}. Formatting fixes.
15768 PR rtl-optimization/83682
15769 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
15770 if it has non-VECTOR_MODE element mode.
15771 (vec_duplicate_p): Likewise.
15773 PR middle-end/83694
15774 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
15775 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
15777 2018-01-05 Jakub Jelinek <jakub@redhat.com>
15780 * config/i386/i386-builtin.def
15781 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
15782 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
15783 Require also OPTION_MASK_ISA_AVX512F in addition to
15784 OPTION_MASK_ISA_GFNI.
15785 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
15786 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
15787 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
15788 to OPTION_MASK_ISA_GFNI.
15789 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
15790 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
15791 OPTION_MASK_ISA_AVX512BW.
15792 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
15793 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
15794 addition to OPTION_MASK_ISA_GFNI.
15795 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
15796 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
15797 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
15798 to OPTION_MASK_ISA_GFNI.
15799 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
15800 a requirement for all ISAs rather than any of them with a few
15802 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
15804 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
15805 bitmasks to be enabled with 3 exceptions, instead of requiring any
15806 enabled ISA with lots of exceptions.
15807 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
15808 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
15809 Change avx512bw in isa attribute to avx512f.
15810 * config/i386/sgxintrin.h: Add license boilerplate.
15811 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
15812 to __AVX512F__ and __AVX512VL to __AVX512VL__.
15813 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
15814 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
15816 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
15817 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
15818 temporarily sse2 rather than sse if not enabled already.
15821 * config/i386/sse.md (VI248_VLBW): Rename to ...
15822 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
15823 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
15824 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
15825 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
15826 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
15827 mode iterator instead of VI248_VLBW.
15829 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
15831 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
15832 (record_modified): Skip clobbers; add debug output.
15833 (param_change_prob): Use sreal frequencies.
15835 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
15837 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
15838 punt for user-aligned variables.
15840 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
15842 * tree-chrec.c (chrec_contains_symbols): Return true for
15845 2018-01-05 Sudakshina Das <sudi.das@arm.com>
15848 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
15849 of (x|y) == x for BICS pattern.
15851 2018-01-05 Jakub Jelinek <jakub@redhat.com>
15853 PR tree-optimization/83605
15854 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
15855 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
15858 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
15860 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
15861 * config/epiphany/rtems.h: New file.
15863 2018-01-04 Jakub Jelinek <jakub@redhat.com>
15864 Uros Bizjak <ubizjak@gmail.com>
15867 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
15868 QIreg_operand instead of register_operand predicate.
15869 * config/i386/i386.c (ix86_rop_should_change_byte_p,
15870 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
15871 comments instead of -fmitigate[-_]rop.
15873 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
15876 * cgraphunit.c (symbol_table::compile): Switch to text_section
15877 before calling assembly_start debug hook.
15878 * run-rtl-passes.c (run_rtl_passes): Likewise.
15881 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15883 * tree-vrp.c (extract_range_from_binary_expr_1): Check
15884 range_int_cst_p rather than !symbolic_range_p before calling
15885 extract_range_from_multiplicative_op_1.
15887 2018-01-04 Jeff Law <law@redhat.com>
15889 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
15890 redundant test in assertion.
15892 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15894 * doc/rtl.texi: Document machine_mode wrapper classes.
15896 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15898 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
15899 using tree_to_uhwi.
15901 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15903 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
15904 the VEC_PERM_EXPR fold to fail.
15906 2018-01-04 Jakub Jelinek <jakub@redhat.com>
15909 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
15910 to switched_sections.
15912 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15915 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
15916 test for d.testing.
15918 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
15921 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
15922 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
15924 2018-01-04 Jakub Jelinek <jakub@redhat.com>
15927 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
15928 is BLKmode and bitpos not zero or mode change is needed.
15930 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15933 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
15936 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
15939 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
15940 instead of MULT rtx. Update all corresponding splitters.
15941 (*saddl_se): Ditto.
15942 (*ssub<modesuffix>): Ditto.
15943 (*ssubl_se): Ditto.
15944 (*cmp_sadd_di): Update split patterns.
15945 (*cmp_sadd_si): Ditto.
15946 (*cmp_sadd_sidi): Ditto.
15947 (*cmp_ssub_di): Ditto.
15948 (*cmp_ssub_si): Ditto.
15949 (*cmp_ssub_sidi): Ditto.
15950 * config/alpha/predicates.md (const23_operand): New predicate.
15951 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
15952 Look for ASHIFT, not MULT inner operand.
15953 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
15955 2018-01-04 Martin Liska <mliska@suse.cz>
15957 PR gcov-profile/83669
15958 * gcov.c (output_intermediate_file): Add version to intermediate
15960 * doc/gcov.texi: Document new field 'version' in intermediate
15961 file format. Fix location of '-k' option of gcov command.
15963 2018-01-04 Martin Liska <mliska@suse.cz>
15966 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
15968 2018-01-04 Jakub Jelinek <jakub@redhat.com>
15970 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
15972 2018-01-03 Martin Sebor <msebor@redhat.com>
15974 PR tree-optimization/83655
15975 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
15976 checking calls with invalid arguments.
15978 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15980 * tree-vect-stmts.c (vect_get_store_rhs): New function.
15981 (vectorizable_mask_load_store): Delete.
15982 (vectorizable_call): Return false for masked loads and stores.
15983 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
15984 instead of gimple_assign_rhs1.
15985 (vectorizable_load): Handle IFN_MASK_LOAD.
15986 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
15988 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15990 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
15992 (vectorizable_mask_load_store): ...here.
15993 (vectorizable_load): ...and here.
15995 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15997 * tree-vect-stmts.c (vect_build_all_ones_mask)
15998 (vect_build_zero_merge_argument): New functions, split out from...
15999 (vectorizable_load): ...here.
16001 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16003 * tree-vect-stmts.c (vect_check_store_rhs): New function,
16005 (vectorizable_mask_load_store): ...here.
16006 (vectorizable_store): ...and here.
16008 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16010 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
16012 (vectorizable_mask_load_store): ...here.
16014 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16016 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
16017 (vect_model_store_cost): Take a vec_load_store_type instead of a
16019 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
16020 (vect_model_store_cost): Take a vec_load_store_type instead of a
16022 (vectorizable_mask_load_store): Update accordingly.
16023 (vectorizable_store): Likewise.
16024 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
16026 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16028 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
16029 IFN_MASK_LOAD calls here rather than...
16030 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
16032 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16033 Alan Hayward <alan.hayward@arm.com>
16034 David Sherwood <david.sherwood@arm.com>
16036 * expmed.c (extract_bit_field_1): For vector extracts,
16037 fall back to extract_bit_field_as_subreg if vec_extract
16040 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16041 Alan Hayward <alan.hayward@arm.com>
16042 David Sherwood <david.sherwood@arm.com>
16044 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
16045 they are variable or constant sized.
16046 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
16047 slots for constant-sized data.
16049 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16050 Alan Hayward <alan.hayward@arm.com>
16051 David Sherwood <david.sherwood@arm.com>
16053 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
16054 handling COND_EXPRs with boolean comparisons, try to find a better
16055 basis for the mask type than the boolean itself.
16057 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16059 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
16060 is calculated and how it can be overridden.
16061 * genmodes.c (max_bitsize_mode_any_mode): New variable.
16062 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
16064 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
16067 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16068 Alan Hayward <alan.hayward@arm.com>
16069 David Sherwood <david.sherwood@arm.com>
16071 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
16072 Remove the mode argument.
16073 (aarch64_simd_valid_immediate): Remove the mode and inverse
16075 * config/aarch64/iterators.md (bitsize): New iterator.
16076 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
16077 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
16078 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
16079 aarch64_simd_valid_immediate.
16080 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
16081 (aarch64_reg_or_bic_imm): Likewise.
16082 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
16083 with an insn_type enum and msl with a modifier_type enum.
16084 Replace element_width with a scalar_mode. Change the shift
16085 to unsigned int. Add constructors for scalar_float_mode and
16086 scalar_int_mode elements.
16087 (aarch64_vect_float_const_representable_p): Delete.
16088 (aarch64_can_const_movi_rtx_p)
16089 (aarch64_simd_scalar_immediate_valid_for_move)
16090 (aarch64_simd_make_constant): Update call to
16091 aarch64_simd_valid_immediate.
16092 (aarch64_advsimd_valid_immediate_hs): New function.
16093 (aarch64_advsimd_valid_immediate): Likewise.
16094 (aarch64_simd_valid_immediate): Remove mode and inverse
16095 arguments. Rewrite to use the above. Use const_vec_duplicate_p
16096 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
16097 and aarch64_float_const_representable_p on the result.
16098 (aarch64_output_simd_mov_immediate): Remove mode argument.
16099 Update call to aarch64_simd_valid_immediate and use of
16100 simd_immediate_info.
16101 (aarch64_output_scalar_simd_mov_immediate): Update call
16104 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16105 Alan Hayward <alan.hayward@arm.com>
16106 David Sherwood <david.sherwood@arm.com>
16108 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
16109 (mode_nunits): Likewise CONST_MODE_NUNITS.
16110 * machmode.def (ADJUST_NUNITS): Document.
16111 * genmodes.c (mode_data::need_nunits_adj): New field.
16112 (blank_mode): Update accordingly.
16113 (adj_nunits): New variable.
16114 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
16116 (emit_mode_size_inline): Set need_bytesize_adj for all modes
16117 listed in adj_nunits.
16118 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
16119 listed in adj_nunits. Don't emit case statements for such modes.
16120 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
16121 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
16122 nothing if adj_nunits is nonnull.
16123 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
16124 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
16125 (emit_mode_fbit): Update use of print_maybe_const_decl.
16126 (emit_move_size): Likewise. Treat the array as non-const
16128 (emit_mode_adjustments): Handle adj_nunits.
16130 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16132 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
16133 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
16134 (VECTOR_MODES): Use it.
16135 (make_vector_modes): Take the prefix as an argument.
16137 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16138 Alan Hayward <alan.hayward@arm.com>
16139 David Sherwood <david.sherwood@arm.com>
16141 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
16142 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
16143 for MODE_VECTOR_BOOL.
16144 * machmode.def (VECTOR_BOOL_MODE): Document.
16145 * genmodes.c (VECTOR_BOOL_MODE): New macro.
16146 (make_vector_bool_mode): New function.
16147 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
16149 * lto-streamer-in.c (lto_input_mode_table): Likewise.
16150 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
16152 * stor-layout.c (int_mode_for_mode): Likewise.
16153 * tree.c (build_vector_type_for_mode): Likewise.
16154 * varasm.c (output_constant_pool_2): Likewise.
16155 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
16156 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
16157 for MODE_VECTOR_BOOL.
16158 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
16159 of mode class checks.
16160 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
16161 instead of a list of mode class checks.
16162 (expand_vector_scalar_condition): Likewise.
16163 (type_for_widest_vector_mode): Handle BImode as an inner mode.
16165 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16166 Alan Hayward <alan.hayward@arm.com>
16167 David Sherwood <david.sherwood@arm.com>
16169 * machmode.h (mode_size): Change from unsigned short to
16171 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
16172 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
16173 or if measurement_type is not polynomial.
16174 (fixed_size_mode::includes_p): Check for constant-sized modes.
16175 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
16176 return a poly_uint16 rather than an unsigned short.
16177 (emit_mode_size): Change the type of mode_size from unsigned short
16178 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
16179 (emit_mode_adjustments): Cope with polynomial vector sizes.
16180 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
16182 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
16184 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
16185 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
16186 * caller-save.c (setup_save_areas): Likewise.
16187 (replace_reg_with_saved_mem): Likewise.
16188 * calls.c (emit_library_call_value_1): Likewise.
16189 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
16190 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
16191 (gen_lowpart_for_combine): Likewise.
16192 * convert.c (convert_to_integer_1): Likewise.
16193 * cse.c (equiv_constant, cse_insn): Likewise.
16194 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
16195 (cselib_subst_to_values): Likewise.
16196 * dce.c (word_dce_process_block): Likewise.
16197 * df-problems.c (df_word_lr_mark_ref): Likewise.
16198 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
16199 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
16200 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
16201 (rtl_for_decl_location): Likewise.
16202 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
16203 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
16204 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
16205 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
16206 (expand_expr_real_1): Likewise.
16207 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
16208 (pad_below): Likewise.
16209 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
16210 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
16211 * ira.c (get_subreg_tracking_sizes): Likewise.
16212 * ira-build.c (ira_create_allocno_objects): Likewise.
16213 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
16214 (ira_sort_regnos_for_alter_reg): Likewise.
16215 * ira-costs.c (record_operand_costs): Likewise.
16216 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
16217 (resolve_simple_move): Likewise.
16218 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
16219 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
16220 (lra_constraints): Likewise.
16221 (CONST_POOL_OK_P): Reject variable-sized modes.
16222 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
16223 (add_pseudo_to_slot, lra_spill): Likewise.
16224 * omp-low.c (omp_clause_aligned_alignment): Likewise.
16225 * optabs-query.c (get_best_extraction_insn): Likewise.
16226 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
16227 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
16228 (expand_mult_highpart, valid_multiword_target_p): Likewise.
16229 * recog.c (offsettable_address_addr_space_p): Likewise.
16230 * regcprop.c (maybe_mode_change): Likewise.
16231 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
16232 * regrename.c (build_def_use): Likewise.
16233 * regstat.c (dump_reg_info): Likewise.
16234 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
16235 (find_reloads, find_reloads_subreg_address): Likewise.
16236 * reload1.c (eliminate_regs_1): Likewise.
16237 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
16238 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
16239 (simplify_binary_operation_1, simplify_subreg): Likewise.
16240 * targhooks.c (default_function_arg_padding): Likewise.
16241 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
16242 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
16243 (verify_gimple_assign_ternary): Likewise.
16244 * tree-inline.c (estimate_move_cost): Likewise.
16245 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
16246 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
16247 (get_address_cost_ainc): Likewise.
16248 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
16249 (vect_supportable_dr_alignment): Likewise.
16250 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
16251 (vectorizable_reduction): Likewise.
16252 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
16253 (vectorizable_operation, vectorizable_load): Likewise.
16254 * tree.c (build_same_sized_truth_vector_type): Likewise.
16255 * valtrack.c (cleanup_auto_inc_dec): Likewise.
16256 * var-tracking.c (emit_note_insn_var_location): Likewise.
16257 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
16258 (ADDR_VEC_ALIGN): Likewise.
16260 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16261 Alan Hayward <alan.hayward@arm.com>
16262 David Sherwood <david.sherwood@arm.com>
16264 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
16266 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
16267 or if measurement_type is polynomial.
16268 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
16269 * combine.c (make_extraction): Likewise.
16270 * dse.c (find_shift_sequence): Likewise.
16271 * dwarf2out.c (mem_loc_descriptor): Likewise.
16272 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
16273 (extract_bit_field, extract_low_bits): Likewise.
16274 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
16275 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
16276 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
16277 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
16278 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
16279 * reload.c (find_reloads): Likewise.
16280 * reload1.c (alter_reg): Likewise.
16281 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
16282 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
16283 * tree-if-conv.c (predicate_mem_writes): Likewise.
16284 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
16285 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
16286 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
16287 * valtrack.c (dead_debug_insert_temp): Likewise.
16288 * varasm.c (mergeable_constant_section): Likewise.
16289 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
16291 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16292 Alan Hayward <alan.hayward@arm.com>
16293 David Sherwood <david.sherwood@arm.com>
16295 * expr.c (expand_assignment): Cope with polynomial mode sizes
16296 when assigning to a CONCAT.
16298 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16299 Alan Hayward <alan.hayward@arm.com>
16300 David Sherwood <david.sherwood@arm.com>
16302 * machmode.h (mode_precision): Change from unsigned short to
16304 (mode_to_precision): Return a poly_uint16 rather than an unsigned
16306 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
16307 or if measurement_type is not polynomial.
16308 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
16309 in which the mode is already known to be a scalar_int_mode.
16310 * genmodes.c (emit_mode_precision): Change the type of mode_precision
16311 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
16313 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
16314 for GET_MODE_PRECISION.
16315 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
16316 for GET_MODE_PRECISION.
16317 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
16319 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
16320 (expand_field_assignment, make_extraction): Likewise.
16321 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
16322 (get_last_value): Likewise.
16323 * convert.c (convert_to_integer_1): Likewise.
16324 * cse.c (cse_insn): Likewise.
16325 * expr.c (expand_expr_real_1): Likewise.
16326 * lra-constraints.c (simplify_operand_subreg): Likewise.
16327 * optabs-query.c (can_atomic_load_p): Likewise.
16328 * optabs.c (expand_atomic_load): Likewise.
16329 (expand_atomic_store): Likewise.
16330 * ree.c (combine_reaching_defs): Likewise.
16331 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
16332 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
16333 * tree.h (type_has_mode_precision_p): Likewise.
16334 * ubsan.c (instrument_si_overflow): Likewise.
16336 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16337 Alan Hayward <alan.hayward@arm.com>
16338 David Sherwood <david.sherwood@arm.com>
16340 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
16341 polynomial numbers of units.
16342 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
16343 (valid_vector_subparts_p): New function.
16344 (build_vector_type): Remove temporary shim and take the number
16345 of units as a poly_uint64 rather than an int.
16346 (build_opaque_vector_type): Take the number of units as a
16347 poly_uint64 rather than an int.
16348 * tree.c (build_vector_from_ctor): Handle polynomial
16349 TYPE_VECTOR_SUBPARTS.
16350 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
16351 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
16352 (build_vector_from_val): If the number of units is variable,
16353 use build_vec_duplicate_cst for constant operands and
16354 VEC_DUPLICATE_EXPR otherwise.
16355 (make_vector_type): Remove temporary is_constant ().
16356 (build_vector_type, build_opaque_vector_type): Take the number of
16357 units as a poly_uint64 rather than an int.
16358 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
16360 * cfgexpand.c (expand_debug_expr): Likewise.
16361 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
16362 (store_constructor, expand_expr_real_1): Likewise.
16363 (const_scalar_mask_from_tree): Likewise.
16364 * fold-const-call.c (fold_const_reduction): Likewise.
16365 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
16366 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
16367 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
16368 (fold_relational_const): Likewise.
16369 (native_interpret_vector): Likewise. Change the size from an
16370 int to an unsigned int.
16371 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
16372 TYPE_VECTOR_SUBPARTS.
16373 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
16374 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
16375 duplicating a non-constant operand into a variable-length vector.
16376 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
16377 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
16378 * ipa-icf.c (sem_variable::equals): Likewise.
16379 * match.pd: Likewise.
16380 * omp-simd-clone.c (simd_clone_subparts): Likewise.
16381 * print-tree.c (print_node): Likewise.
16382 * stor-layout.c (layout_type): Likewise.
16383 * targhooks.c (default_builtin_vectorization_cost): Likewise.
16384 * tree-cfg.c (verify_gimple_comparison): Likewise.
16385 (verify_gimple_assign_binary): Likewise.
16386 (verify_gimple_assign_ternary): Likewise.
16387 (verify_gimple_assign_single): Likewise.
16388 * tree-pretty-print.c (dump_generic_node): Likewise.
16389 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
16390 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
16391 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
16392 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
16393 (vect_shift_permute_load_chain): Likewise.
16394 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
16395 (expand_vector_condition, optimize_vector_constructor): Likewise.
16396 (lower_vec_perm, get_compute_type): Likewise.
16397 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
16398 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
16399 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
16400 (vect_recog_mask_conversion_pattern): Likewise.
16401 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
16402 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
16403 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
16404 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
16405 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
16406 (vectorizable_shift, vectorizable_operation, vectorizable_store)
16407 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
16408 (supportable_widening_operation): Likewise.
16409 (supportable_narrowing_operation): Likewise.
16410 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
16412 * varasm.c (output_constant): Likewise.
16414 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16415 Alan Hayward <alan.hayward@arm.com>
16416 David Sherwood <david.sherwood@arm.com>
16418 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
16419 so that both the length == 3 and length != 3 cases set up their
16420 own permute vectors. Add comments explaining why we know the
16421 number of elements is constant.
16422 (vect_permute_load_chain): Likewise.
16424 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16425 Alan Hayward <alan.hayward@arm.com>
16426 David Sherwood <david.sherwood@arm.com>
16428 * machmode.h (mode_nunits): Change from unsigned char to
16430 (ONLY_FIXED_SIZE_MODES): New macro.
16431 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
16432 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
16433 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
16435 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
16436 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
16437 or if measurement_type is not polynomial.
16438 * genmodes.c (ZERO_COEFFS): New macro.
16439 (emit_mode_nunits_inline): Make mode_nunits_inline return a
16441 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
16442 Use ZERO_COEFFS when emitting initializers.
16443 * data-streamer.h (bp_pack_poly_value): New function.
16444 (bp_unpack_poly_value): Likewise.
16445 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
16446 for GET_MODE_NUNITS.
16447 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
16448 for GET_MODE_NUNITS.
16449 * tree.c (make_vector_type): Remove temporary shim and make
16450 the real function take the number of units as a poly_uint64
16451 rather than an int.
16452 (build_vector_type_for_mode): Handle polynomial nunits.
16453 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
16454 * emit-rtl.c (const_vec_series_p_1): Likewise.
16455 (gen_rtx_CONST_VECTOR): Likewise.
16456 * fold-const.c (test_vec_duplicate_folding): Likewise.
16457 * genrecog.c (validate_pattern): Likewise.
16458 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
16459 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
16460 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
16461 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
16462 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
16463 * rtlanal.c (subreg_get_info): Likewise.
16464 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
16465 (vect_grouped_load_supported): Likewise.
16466 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
16467 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
16468 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
16469 (simplify_const_unary_operation, simplify_binary_operation_1)
16470 (simplify_const_binary_operation, simplify_ternary_operation)
16471 (test_vector_ops_duplicate, test_vector_ops): Likewise.
16472 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
16473 instead of CONST_VECTOR_NUNITS.
16474 * varasm.c (output_constant_pool_2): Likewise.
16475 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
16476 explicit-encoded elements in the XVEC for variable-length vectors.
16478 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16480 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
16482 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16483 Alan Hayward <alan.hayward@arm.com>
16484 David Sherwood <david.sherwood@arm.com>
16486 * coretypes.h (fixed_size_mode): Declare.
16487 (fixed_size_mode_pod): New typedef.
16488 * builtins.h (target_builtins::x_apply_args_mode)
16489 (target_builtins::x_apply_result_mode): Change type to
16490 fixed_size_mode_pod.
16491 * builtins.c (apply_args_size, apply_result_size, result_vector)
16492 (expand_builtin_apply_args_1, expand_builtin_apply)
16493 (expand_builtin_return): Update accordingly.
16495 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16497 * cse.c (hash_rtx_cb): Hash only the encoded elements.
16498 * cselib.c (cselib_hash_rtx): Likewise.
16499 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
16500 CONST_VECTOR encoding.
16502 2018-01-03 Jakub Jelinek <jakub@redhat.com>
16503 Jeff Law <law@redhat.com>
16506 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
16507 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
16508 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
16509 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
16512 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
16513 explicitly probe *sp in a noreturn function if there were any callee
16514 register saves or frame pointer is needed.
16516 2018-01-03 Jakub Jelinek <jakub@redhat.com>
16519 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
16520 BLKmode for ternary, binary or unary expressions.
16523 * var-tracking.c (delete_vta_debug_insn): New inline function.
16524 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
16525 insns from get_insns () to NULL instead of each bb separately.
16526 Use delete_vta_debug_insn. No longer static.
16527 (vt_debug_insns_local, variable_tracking_main_1): Adjust
16528 delete_vta_debug_insns callers.
16529 * rtl.h (delete_vta_debug_insns): Declare.
16530 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
16531 instead of variable_tracking_main.
16533 2018-01-03 Martin Sebor <msebor@redhat.com>
16535 PR tree-optimization/83603
16536 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
16537 arguments past the endof the argument list in functions declared
16538 without a prototype.
16539 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
16540 Avoid checking when arguments are null.
16542 2018-01-03 Martin Sebor <msebor@redhat.com>
16545 * doc/extend.texi (attribute const): Fix a typo.
16546 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
16547 issuing -Wsuggest-attribute for void functions.
16549 2018-01-03 Martin Sebor <msebor@redhat.com>
16551 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
16552 offset_int::from instead of wide_int::to_shwi.
16553 (maybe_diag_overlap): Remove assertion.
16554 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
16555 * gimple-ssa-sprintf.c (format_directive): Same.
16556 (parse_directive): Same.
16557 (sprintf_dom_walker::compute_format_length): Same.
16558 (try_substitute_return_value): Same.
16560 2018-01-03 Jeff Law <law@redhat.com>
16562 PR middle-end/83654
16563 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
16564 non-constant residual for zero at runtime and avoid probing in
16565 that case. Reorganize code for trailing problem to mirror handling
16568 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
16570 PR tree-optimization/83501
16571 * tree-ssa-strlen.c (get_string_cst): New.
16572 (handle_char_store): Call get_string_cst.
16574 2018-01-03 Martin Liska <mliska@suse.cz>
16576 PR tree-optimization/83593
16577 * tree-ssa-strlen.c: Include tree-cfg.h.
16578 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
16579 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
16580 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
16582 (strlen_dom_walker::before_dom_children): Call
16583 gimple_purge_dead_eh_edges. Dump tranformation with details
16585 (strlen_dom_walker::before_dom_children): Update call by adding
16586 new argument cleanup_eh.
16587 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
16589 2018-01-03 Martin Liska <mliska@suse.cz>
16592 * cif-code.def (VARIADIC_THUNK): New enum value.
16593 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
16596 2018-01-03 Jan Beulich <jbeulich@suse.com>
16598 * sse.md (mov<mode>_internal): Tighten condition for when to use
16599 vmovdqu<ssescalarsize> for TI and OI modes.
16601 2018-01-03 Jakub Jelinek <jakub@redhat.com>
16603 Update copyright years.
16605 2018-01-03 Martin Liska <mliska@suse.cz>
16608 * ipa-visibility.c (function_and_variable_visibility): Skip
16609 functions with noipa attribure.
16611 2018-01-03 Jakub Jelinek <jakub@redhat.com>
16613 * gcc.c (process_command): Update copyright notice dates.
16614 * gcov-dump.c (print_version): Ditto.
16615 * gcov.c (print_version): Ditto.
16616 * gcov-tool.c (print_version): Ditto.
16617 * gengtype.c (create_file): Ditto.
16618 * doc/cpp.texi: Bump @copying's copyright year.
16619 * doc/cppinternals.texi: Ditto.
16620 * doc/gcc.texi: Ditto.
16621 * doc/gccint.texi: Ditto.
16622 * doc/gcov.texi: Ditto.
16623 * doc/install.texi: Ditto.
16624 * doc/invoke.texi: Ditto.
16626 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16628 * vector-builder.h (vector_builder::m_full_nelts): Change from
16629 unsigned int to poly_uint64.
16630 (vector_builder::full_nelts): Update prototype accordingly.
16631 (vector_builder::new_vector): Likewise.
16632 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
16633 (vector_builder::operator ==): Likewise.
16634 (vector_builder::finalize): Likewise.
16635 * int-vector-builder.h (int_vector_builder::int_vector_builder):
16636 Take the number of elements as a poly_uint64 rather than an
16638 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
16639 from unsigned int to poly_uint64.
16640 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
16641 (vec_perm_indices::new_vector): Likewise.
16642 (vec_perm_indices::length): Likewise.
16643 (vec_perm_indices::nelts_per_input): Likewise.
16644 (vec_perm_indices::input_nelts): Likewise.
16645 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
16646 number of elements per input as a poly_uint64 rather than an
16647 unsigned int. Use the original encoding for variable-length
16648 vectors, rather than clamping each individual element.
16649 For the second and subsequent elements in each pattern,
16650 clamp the step and base before clamping their sum.
16651 (vec_perm_indices::series_p): Handle polynomial element counts.
16652 (vec_perm_indices::all_in_range_p): Likewise.
16653 (vec_perm_indices_to_tree): Likewise.
16654 (vec_perm_indices_to_rtx): Likewise.
16655 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
16656 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
16657 (tree_vector_builder::new_binary_operation): Handle polynomial
16658 element counts. Return false if we need to know the number
16659 of elements at compile time.
16660 * fold-const.c (fold_vec_perm): Punt if the number of elements
16661 isn't known at compile time.
16663 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16665 * vec-perm-indices.h (vec_perm_builder): Change element type
16666 from HOST_WIDE_INT to poly_int64.
16667 (vec_perm_indices::element_type): Update accordingly.
16668 (vec_perm_indices::clamp): Handle polynomial element_types.
16669 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
16670 (vec_perm_indices::all_in_range_p): Likewise.
16671 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
16673 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
16674 polynomial vec_perm_indices element types.
16675 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
16676 * fold-const.c (fold_vec_perm): Likewise.
16677 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
16678 * tree-vect-generic.c (lower_vec_perm): Likewise.
16679 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
16680 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
16681 element type to HOST_WIDE_INT.
16683 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16684 Alan Hayward <alan.hayward@arm.com>
16685 David Sherwood <david.sherwood@arm.com>
16687 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
16688 rather than an int. Use plus_constant.
16689 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
16690 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
16692 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16693 Alan Hayward <alan.hayward@arm.com>
16694 David Sherwood <david.sherwood@arm.com>
16696 * calls.c (emit_call_1, expand_call): Change struct_value_size from
16697 a HOST_WIDE_INT to a poly_int64.
16699 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16700 Alan Hayward <alan.hayward@arm.com>
16701 David Sherwood <david.sherwood@arm.com>
16703 * calls.c (load_register_parameters): Cope with polynomial
16704 mode sizes. Require a constant size for BLKmode parameters
16705 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
16706 forces a parameter to be padded at the lsb end in order to
16707 fill a complete number of words, require the parameter size
16708 to be ordered wrt UNITS_PER_WORD.
16710 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16711 Alan Hayward <alan.hayward@arm.com>
16712 David Sherwood <david.sherwood@arm.com>
16714 * reload1.c (spill_stack_slot_width): Change element type
16715 from unsigned int to poly_uint64_pod.
16716 (alter_reg): Treat mode sizes as polynomial.
16718 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16719 Alan Hayward <alan.hayward@arm.com>
16720 David Sherwood <david.sherwood@arm.com>
16722 * reload.c (complex_word_subreg_p): New function.
16723 (reload_inner_reg_of_subreg, push_reload): Use it.
16725 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16726 Alan Hayward <alan.hayward@arm.com>
16727 David Sherwood <david.sherwood@arm.com>
16729 * lra-constraints.c (process_alt_operands): Reject matched
16730 operands whose sizes aren't ordered.
16731 (match_reload): Refer to this check here.
16733 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16734 Alan Hayward <alan.hayward@arm.com>
16735 David Sherwood <david.sherwood@arm.com>
16737 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
16738 that the mode size is in the set {1, 2, 4, 8, 16}.
16740 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16741 Alan Hayward <alan.hayward@arm.com>
16742 David Sherwood <david.sherwood@arm.com>
16744 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
16745 Use plus_constant instead of gen_rtx_PLUS.
16747 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16748 Alan Hayward <alan.hayward@arm.com>
16749 David Sherwood <david.sherwood@arm.com>
16751 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
16752 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
16753 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
16754 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
16755 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
16756 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
16757 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
16758 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
16759 * config/i386/i386.c (ix86_push_rounding): ...this new function.
16760 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
16762 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
16763 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
16764 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
16765 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
16766 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
16767 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
16768 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
16769 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
16770 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
16771 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
16773 * expr.c (emit_move_resolve_push): Treat the input and result
16774 of PUSH_ROUNDING as a poly_int64.
16775 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
16776 (emit_push_insn): Likewise.
16777 * lra-eliminations.c (mark_not_eliminable): Likewise.
16778 * recog.c (push_operand): Likewise.
16779 * reload1.c (elimination_effects): Likewise.
16780 * rtlanal.c (nonzero_bits1): Likewise.
16781 * calls.c (store_one_arg): Likewise. Require the padding to be
16782 known at compile time.
16784 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16785 Alan Hayward <alan.hayward@arm.com>
16786 David Sherwood <david.sherwood@arm.com>
16788 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
16789 Use plus_constant instead of gen_rtx_PLUS.
16791 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16792 Alan Hayward <alan.hayward@arm.com>
16793 David Sherwood <david.sherwood@arm.com>
16795 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
16796 rather than an int.
16798 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16799 Alan Hayward <alan.hayward@arm.com>
16800 David Sherwood <david.sherwood@arm.com>
16802 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
16803 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
16804 via stack temporaries. Treat the mode size as polynomial too.
16806 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16807 Alan Hayward <alan.hayward@arm.com>
16808 David Sherwood <david.sherwood@arm.com>
16810 * expr.c (expand_expr_real_2): When handling conversions involving
16811 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
16812 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
16813 as a poly_uint64 too.
16815 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16816 Alan Hayward <alan.hayward@arm.com>
16817 David Sherwood <david.sherwood@arm.com>
16819 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
16821 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16822 Alan Hayward <alan.hayward@arm.com>
16823 David Sherwood <david.sherwood@arm.com>
16825 * combine.c (can_change_dest_mode): Handle polynomial
16826 REGMODE_NATURAL_SIZE.
16827 * expmed.c (store_bit_field_1): Likewise.
16828 * expr.c (store_constructor): Likewise.
16829 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
16830 and polynomial REGMODE_NATURAL_SIZE.
16831 (gen_lowpart_common): Likewise.
16832 * reginfo.c (record_subregs_of_mode): Likewise.
16833 * rtlanal.c (read_modify_subreg_p): Likewise.
16835 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16836 Alan Hayward <alan.hayward@arm.com>
16837 David Sherwood <david.sherwood@arm.com>
16839 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
16840 numbers of elements.
16842 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16843 Alan Hayward <alan.hayward@arm.com>
16844 David Sherwood <david.sherwood@arm.com>
16846 * match.pd: Cope with polynomial numbers of vector elements.
16848 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16849 Alan Hayward <alan.hayward@arm.com>
16850 David Sherwood <david.sherwood@arm.com>
16852 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
16853 in a POINTER_PLUS_EXPR.
16855 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16856 Alan Hayward <alan.hayward@arm.com>
16857 David Sherwood <david.sherwood@arm.com>
16859 * omp-simd-clone.c (simd_clone_subparts): New function.
16860 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
16861 (ipa_simd_modify_function_body): Likewise.
16863 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16864 Alan Hayward <alan.hayward@arm.com>
16865 David Sherwood <david.sherwood@arm.com>
16867 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
16868 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
16869 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
16870 (expand_vector_condition, vector_element): Likewise.
16871 (subparts_gt): New function.
16872 (get_compute_type): Use subparts_gt.
16873 (count_type_subparts): Delete.
16874 (expand_vector_operations_1): Use subparts_gt instead of
16875 count_type_subparts.
16877 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16878 Alan Hayward <alan.hayward@arm.com>
16879 David Sherwood <david.sherwood@arm.com>
16881 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
16882 (vect_compile_time_alias): ...this new function. Do the calculation
16883 on poly_ints rather than trees.
16884 (vect_prune_runtime_alias_test_list): Update call accordingly.
16886 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16887 Alan Hayward <alan.hayward@arm.com>
16888 David Sherwood <david.sherwood@arm.com>
16890 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
16892 (vect_schedule_slp_instance): Likewise.
16894 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16895 Alan Hayward <alan.hayward@arm.com>
16896 David Sherwood <david.sherwood@arm.com>
16898 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
16899 constant and extern definitions for variable-length vectors.
16900 (vect_get_constant_vectors): Note that the number of units
16901 is known to be constant.
16903 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16904 Alan Hayward <alan.hayward@arm.com>
16905 David Sherwood <david.sherwood@arm.com>
16907 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
16908 of units as polynomial. Choose between WIDE and NARROW based
16911 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16912 Alan Hayward <alan.hayward@arm.com>
16913 David Sherwood <david.sherwood@arm.com>
16915 * tree-vect-stmts.c (simd_clone_subparts): New function.
16916 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
16918 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16919 Alan Hayward <alan.hayward@arm.com>
16920 David Sherwood <david.sherwood@arm.com>
16922 * tree-vect-stmts.c (vectorizable_call): Treat the number of
16923 vectors as polynomial. Use build_index_vector for
16924 IFN_GOMP_SIMD_LANE.
16926 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16927 Alan Hayward <alan.hayward@arm.com>
16928 David Sherwood <david.sherwood@arm.com>
16930 * tree-vect-stmts.c (get_load_store_type): Treat the number of
16931 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
16932 for variable-length vectors.
16933 (vectorizable_mask_load_store): Treat the number of units as
16934 polynomial, asserting that it is constant if the condition has
16935 already been enforced.
16936 (vectorizable_store, vectorizable_load): Likewise.
16938 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16939 Alan Hayward <alan.hayward@arm.com>
16940 David Sherwood <david.sherwood@arm.com>
16942 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
16943 of units as polynomial. Punt if we can't tell at compile time
16944 which vector contains the final result.
16946 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16947 Alan Hayward <alan.hayward@arm.com>
16948 David Sherwood <david.sherwood@arm.com>
16950 * tree-vect-loop.c (vectorizable_induction): Treat the number
16951 of units as polynomial. Punt on SLP inductions. Use an integer
16952 VEC_SERIES_EXPR for variable-length integer reductions. Use a
16953 cast of such a series for variable-length floating-point
16956 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16957 Alan Hayward <alan.hayward@arm.com>
16958 David Sherwood <david.sherwood@arm.com>
16960 * tree.h (build_index_vector): Declare.
16961 * tree.c (build_index_vector): New function.
16962 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
16963 of units as polynomial, forcibly converting it to a constant if
16964 vectorizable_reduction has already enforced the condition.
16965 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
16966 to create a {1,2,3,...} vector.
16967 (vectorizable_reduction): Treat the number of units as polynomial.
16968 Choose vectype_in based on the largest scalar element size rather
16969 than the smallest number of units. Enforce the restrictions
16972 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16973 Alan Hayward <alan.hayward@arm.com>
16974 David Sherwood <david.sherwood@arm.com>
16976 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
16977 number of units as polynomial.
16979 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16980 Alan Hayward <alan.hayward@arm.com>
16981 David Sherwood <david.sherwood@arm.com>
16983 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
16984 * target.def (autovectorize_vector_sizes): Return the vector sizes
16985 by pointer, using vector_sizes rather than a bitmask.
16986 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
16987 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
16988 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
16990 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
16991 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
16992 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
16993 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
16994 * omp-general.c (omp_max_vf): Likewise.
16995 * omp-low.c (omp_clause_aligned_alignment): Likewise.
16996 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
16997 * tree-vect-loop.c (vect_analyze_loop): Likewise.
16998 * tree-vect-slp.c (vect_slp_bb): Likewise.
16999 * doc/tm.texi: Regenerate.
17000 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
17002 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
17003 the vector size as a poly_uint64 rather than an unsigned int.
17004 (current_vector_size): Change from an unsigned int to a poly_uint64.
17005 (get_vectype_for_scalar_type): Update accordingly.
17006 * tree.h (build_truth_vector_type): Take the size and number of
17007 units as a poly_uint64 rather than an unsigned int.
17008 (build_vector_type): Add a temporary overload that takes
17009 the number of units as a poly_uint64 rather than an unsigned int.
17010 * tree.c (make_vector_type): Likewise.
17011 (build_truth_vector_type): Take the number of units as a poly_uint64
17012 rather than an unsigned int.
17014 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17015 Alan Hayward <alan.hayward@arm.com>
17016 David Sherwood <david.sherwood@arm.com>
17018 * target.def (get_mask_mode): Take the number of units and length
17019 as poly_uint64s rather than unsigned ints.
17020 * targhooks.h (default_get_mask_mode): Update accordingly.
17021 * targhooks.c (default_get_mask_mode): Likewise.
17022 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
17023 * doc/tm.texi: Regenerate.
17025 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17026 Alan Hayward <alan.hayward@arm.com>
17027 David Sherwood <david.sherwood@arm.com>
17029 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
17030 * omp-general.c (omp_max_vf): Likewise.
17031 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
17032 (expand_omp_simd): Handle polynomial safelen.
17033 * omp-low.c (omplow_simd_context): Add a default constructor.
17034 (omplow_simd_context::max_vf): Change from int to poly_uint64.
17035 (lower_rec_simd_input_clauses): Update accordingly.
17036 (lower_rec_input_clauses): Likewise.
17038 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17039 Alan Hayward <alan.hayward@arm.com>
17040 David Sherwood <david.sherwood@arm.com>
17042 * tree-vectorizer.h (vect_nunits_for_cost): New function.
17043 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
17044 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
17045 (vect_analyze_slp_cost): Likewise.
17046 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
17047 (vect_model_load_cost): Likewise.
17049 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17050 Alan Hayward <alan.hayward@arm.com>
17051 David Sherwood <david.sherwood@arm.com>
17053 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
17054 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
17055 from an unsigned int * to a poly_uint64_pod *.
17056 (calculate_unrolling_factor): New function.
17057 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
17059 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17060 Alan Hayward <alan.hayward@arm.com>
17061 David Sherwood <david.sherwood@arm.com>
17063 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
17064 from an unsigned int to a poly_uint64.
17065 (_loop_vec_info::slp_unrolling_factor): Likewise.
17066 (_loop_vec_info::vectorization_factor): Change from an int
17068 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
17069 (vect_get_num_vectors): New function.
17070 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
17071 (vect_get_num_copies): Use vect_get_num_vectors.
17072 (vect_analyze_data_ref_dependences): Change max_vf from an int *
17073 to an unsigned int *.
17074 (vect_analyze_data_refs): Change min_vf from an int * to a
17076 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
17077 than an unsigned HOST_WIDE_INT.
17078 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
17079 (vect_analyze_data_ref_dependence): Change max_vf from an int *
17080 to an unsigned int *.
17081 (vect_analyze_data_ref_dependences): Likewise.
17082 (vect_compute_data_ref_alignment): Handle polynomial vf.
17083 (vect_enhance_data_refs_alignment): Likewise.
17084 (vect_prune_runtime_alias_test_list): Likewise.
17085 (vect_shift_permute_load_chain): Likewise.
17086 (vect_supportable_dr_alignment): Likewise.
17087 (dependence_distance_ge_vf): Take the vectorization factor as a
17088 poly_uint64 rather than an unsigned HOST_WIDE_INT.
17089 (vect_analyze_data_refs): Change min_vf from an int * to a
17091 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
17092 vfm1 as a poly_uint64 rather than an int. Make the same change
17093 for the returned bound_scalar.
17094 (vect_gen_vector_loop_niters): Handle polynomial vf.
17095 (vect_do_peeling): Likewise. Update call to
17096 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
17097 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
17099 * tree-vect-loop.c (vect_determine_vectorization_factor)
17100 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
17101 (vect_get_known_peeling_cost): Likewise.
17102 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
17103 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
17104 (vect_transform_loop): Likewise. Use the lowest possible VF when
17105 updating the upper bounds of the loop.
17106 (vect_min_worthwhile_factor): Make static. Return an unsigned int
17107 rather than an int.
17108 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
17109 polynomial unroll factors.
17110 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
17111 (vect_make_slp_decision): Likewise.
17112 (vect_supported_load_permutation_p): Likewise, and polynomial
17114 (vect_analyze_slp_cost): Handle polynomial vf.
17115 (vect_slp_analyze_node_operations): Likewise.
17116 (vect_slp_analyze_bb_1): Likewise.
17117 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
17118 than an unsigned HOST_WIDE_INT.
17119 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
17120 (vectorizable_load): Handle polynomial vf.
17121 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
17123 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
17125 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17126 Alan Hayward <alan.hayward@arm.com>
17127 David Sherwood <david.sherwood@arm.com>
17129 * match.pd: Handle bit operations involving three constants
17130 and try to fold one pair.
17132 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17134 * tree-vect-loop-manip.c: Include gimple-fold.h.
17135 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
17136 niters_maybe_zero parameters. Handle other cases besides a step of 1.
17137 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
17138 Add a path that uses a step of VF instead of 1, but disable it
17140 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
17141 and niters_no_overflow parameters. Update calls to
17142 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
17143 Create a new SSA name if the latter choses to use a ste other
17144 than zero, and return it via niters_vector_mult_vf_var.
17145 * tree-vect-loop.c (vect_transform_loop): Update calls to
17146 vect_do_peeling, vect_gen_vector_loop_niters and
17147 slpeel_make_loop_iterate_ntimes.
17148 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
17149 (vect_gen_vector_loop_niters): Update declarations after above changes.
17151 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
17153 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
17154 128-bit round to integer instructions.
17155 (ceil<mode>2): Likewise.
17156 (btrunc<mode>2): Likewise.
17157 (round<mode>2): Likewise.
17159 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
17161 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
17162 unaligned VSX load/store on P8/P9.
17163 (expand_block_clear): Allow the use of unaligned VSX
17164 load/store on P8/P9.
17166 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
17168 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
17170 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
17171 swap associated with both a load and a store.
17173 2018-01-02 Andrew Waterman <andrew@sifive.com>
17175 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
17176 * config/riscv/riscv.md (clear_cache): Use it.
17178 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
17180 * web.c: Remove out-of-date comment.
17182 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17184 * expr.c (fixup_args_size_notes): Check that any existing
17185 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
17186 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
17187 (emit_single_push_insn): ...here.
17189 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17191 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
17192 (const_vector_encoded_nelts): New function.
17193 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
17194 (const_vector_int_elt, const_vector_elt): Declare.
17195 * emit-rtl.c (const_vector_int_elt_1): New function.
17196 (const_vector_elt): Likewise.
17197 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
17198 of CONST_VECTOR_ELT.
17200 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17202 * expr.c: Include rtx-vector-builder.h.
17203 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
17204 directly on the tree encoding.
17205 (const_vector_from_tree): Likewise.
17206 * optabs.c: Include rtx-vector-builder.h.
17207 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
17208 sequence of "u" values.
17209 * vec-perm-indices.c: Include rtx-vector-builder.h.
17210 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
17211 directly on the vec_perm_indices encoding.
17213 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17215 * doc/rtl.texi (const_vector): Describe new encoding scheme.
17216 * Makefile.in (OBJS): Add rtx-vector-builder.o.
17217 * rtx-vector-builder.h: New file.
17218 * rtx-vector-builder.c: Likewise.
17219 * rtl.h (rtx_def::u2): Add a const_vector field.
17220 (CONST_VECTOR_NPATTERNS): New macro.
17221 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
17222 (CONST_VECTOR_DUPLICATE_P): Likewise.
17223 (CONST_VECTOR_STEPPED_P): Likewise.
17224 (CONST_VECTOR_ENCODED_ELT): Likewise.
17225 (const_vec_duplicate_p): Check for a duplicated vector encoding.
17226 (unwrap_const_vec_duplicate): Likewise.
17227 (const_vec_series_p): Check for a non-duplicated vector encoding.
17228 Say that the function only returns true for integer vectors.
17229 * emit-rtl.c: Include rtx-vector-builder.h.
17230 (gen_const_vec_duplicate_1): Delete.
17231 (gen_const_vector): Call gen_const_vec_duplicate instead of
17232 gen_const_vec_duplicate_1.
17233 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
17234 (gen_const_vec_duplicate): Use rtx_vector_builder.
17235 (gen_const_vec_series): Likewise.
17236 (gen_rtx_CONST_VECTOR): Likewise.
17237 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
17238 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
17239 Build a new vector rather than modifying a CONST_VECTOR in-place.
17240 (handle_special_swappables): Update call accordingly.
17241 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
17242 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
17243 Build a new vector rather than modifying a CONST_VECTOR in-place.
17244 (handle_special_swappables): Update call accordingly.
17246 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17248 * simplify-rtx.c (simplify_const_binary_operation): Use
17249 CONST_VECTOR_ELT instead of XVECEXP.
17251 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17253 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
17254 the selector elements to be different from the data elements
17255 if the selector is a VECTOR_CST.
17256 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
17257 ssizetype for the selector.
17259 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17261 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
17262 before testing each element individually.
17263 * tree-vect-generic.c (lower_vec_perm): Likewise.
17265 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17267 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
17268 * selftest-run-tests.c (selftest::run_tests): Call it.
17269 * vector-builder.h (vector_builder::operator ==): New function.
17270 (vector_builder::operator !=): Likewise.
17271 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
17272 (vec_perm_indices::all_from_input_p): New function.
17273 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
17274 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
17275 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
17276 instead of reading the VECTOR_CST directly. Detect whether both
17277 vector inputs are the same before constructing the vec_perm_indices,
17278 and update the number of inputs argument accordingly. Use the
17279 utility functions added above. Only construct sel2 if we need to.
17281 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17283 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
17284 the broadcast of the low byte.
17285 (expand_mult_highpart): Use an explicit encoding for the permutes.
17286 * optabs-query.c (can_mult_highpart_p): Likewise.
17287 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
17288 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
17289 (vectorizable_bswap): Likewise.
17290 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
17291 explicit encoding for the power-of-2 permutes.
17292 (vect_permute_store_chain): Likewise.
17293 (vect_grouped_load_supported): Likewise.
17294 (vect_permute_load_chain): Likewise.
17296 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17298 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
17299 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
17300 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
17301 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
17302 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
17303 (vect_gen_perm_mask_any): Likewise.
17305 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17307 * int-vector-builder.h: New file.
17308 * vec-perm-indices.h: Include int-vector-builder.h.
17309 (vec_perm_indices): Redefine as an int_vector_builder.
17310 (auto_vec_perm_indices): Delete.
17311 (vec_perm_builder): Redefine as a stand-alone class.
17312 (vec_perm_indices::vec_perm_indices): New function.
17313 (vec_perm_indices::clamp): Likewise.
17314 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
17315 (vec_perm_indices::new_vector): New function.
17316 (vec_perm_indices::new_expanded_vector): Update for new
17317 vec_perm_indices class.
17318 (vec_perm_indices::rotate_inputs): New function.
17319 (vec_perm_indices::all_in_range_p): Operate directly on the
17320 encoded form, without computing elided elements.
17321 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
17322 encoding. Update for new vec_perm_indices class.
17323 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
17324 the given vec_perm_builder.
17325 (expand_vec_perm_var): Update vec_perm_builder constructor.
17326 (expand_mult_highpart): Use vec_perm_builder instead of
17327 auto_vec_perm_indices.
17328 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
17329 vec_perm_indices instead of auto_vec_perm_indices. Use a single
17330 or double series encoding as appropriate.
17331 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
17332 vec_perm_indices instead of auto_vec_perm_indices.
17333 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
17334 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
17335 (vect_permute_store_chain): Likewise.
17336 (vect_grouped_load_supported): Likewise.
17337 (vect_permute_load_chain): Likewise.
17338 (vect_shift_permute_load_chain): Likewise.
17339 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
17340 (vect_transform_slp_perm_load): Likewise.
17341 (vect_schedule_slp_instance): Likewise.
17342 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
17343 (vectorizable_mask_load_store): Likewise.
17344 (vectorizable_bswap): Likewise.
17345 (vectorizable_store): Likewise.
17346 (vectorizable_load): Likewise.
17347 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
17348 vec_perm_indices instead of auto_vec_perm_indices. Use
17349 tree_to_vec_perm_builder to read the vector from a tree.
17350 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
17351 vec_perm_builder instead of a vec_perm_indices.
17352 (have_whole_vector_shift): Use vec_perm_builder and
17353 vec_perm_indices instead of auto_vec_perm_indices. Leave the
17354 truncation to calc_vec_perm_mask_for_shift.
17355 (vect_create_epilog_for_reduction): Likewise.
17356 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
17357 from auto_vec_perm_indices to vec_perm_indices.
17358 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
17359 instead of changing individual elements.
17360 (aarch64_vectorize_vec_perm_const): Use new_vector to install
17361 the vector in d.perm.
17362 * config/arm/arm.c (expand_vec_perm_d::perm): Change
17363 from auto_vec_perm_indices to vec_perm_indices.
17364 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
17365 instead of changing individual elements.
17366 (arm_vectorize_vec_perm_const): Use new_vector to install
17367 the vector in d.perm.
17368 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
17369 Update vec_perm_builder constructor.
17370 (rs6000_expand_interleave): Likewise.
17371 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
17372 (rs6000_expand_interleave): Likewise.
17374 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17376 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
17377 to qimode could truncate the indices.
17378 * optabs.c (expand_vec_perm_var): Likewise.
17380 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17382 * Makefile.in (OBJS): Add vec-perm-indices.o.
17383 * vec-perm-indices.h: New file.
17384 * vec-perm-indices.c: Likewise.
17385 * target.h (vec_perm_indices): Replace with a forward class
17387 (auto_vec_perm_indices): Move to vec-perm-indices.h.
17388 * optabs.h: Include vec-perm-indices.h.
17389 (expand_vec_perm): Delete.
17390 (selector_fits_mode_p, expand_vec_perm_var): Declare.
17391 (expand_vec_perm_const): Declare.
17392 * target.def (vec_perm_const_ok): Replace with...
17393 (vec_perm_const): ...this new hook.
17394 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
17395 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
17396 * doc/tm.texi: Regenerate.
17397 * optabs.def (vec_perm_const): Delete.
17398 * doc/md.texi (vec_perm_const): Likewise.
17399 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
17400 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
17401 expand_vec_perm for constant permutation vectors. Assert that
17402 the mode of variable permutation vectors is the integer equivalent
17403 of the mode that is being permuted.
17404 * optabs-query.h (selector_fits_mode_p): Declare.
17405 * optabs-query.c: Include vec-perm-indices.h.
17406 (selector_fits_mode_p): New function.
17407 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
17408 is defined, instead of checking whether the vec_perm_const_optab
17409 exists. Use targetm.vectorize.vec_perm_const instead of
17410 targetm.vectorize.vec_perm_const_ok. Check whether the indices
17411 fit in the vector mode before using a variable permute.
17412 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
17413 vec_perm_indices instead of an rtx.
17414 (expand_vec_perm): Replace with...
17415 (expand_vec_perm_const): ...this new function. Take the selector
17416 as a vec_perm_indices rather than an rtx. Also take the mode of
17417 the selector. Update call to shift_amt_for_vec_perm_mask.
17418 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
17419 Use vec_perm_indices::new_expanded_vector to expand the original
17420 selector into bytes. Check whether the indices fit in the vector
17421 mode before using a variable permute.
17422 (expand_vec_perm_var): Make global.
17423 (expand_mult_highpart): Use expand_vec_perm_const.
17424 * fold-const.c: Includes vec-perm-indices.h.
17425 * tree-ssa-forwprop.c: Likewise.
17426 * tree-vect-data-refs.c: Likewise.
17427 * tree-vect-generic.c: Likewise.
17428 * tree-vect-loop.c: Likewise.
17429 * tree-vect-slp.c: Likewise.
17430 * tree-vect-stmts.c: Likewise.
17431 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
17433 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
17434 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
17435 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
17436 (aarch64_vectorize_vec_perm_const): ...this new function.
17437 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
17438 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
17439 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
17440 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
17441 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
17442 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
17443 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
17445 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
17446 check for NEON modes.
17447 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
17448 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
17449 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
17450 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
17452 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
17453 the old VEC_PERM_CONST conditions.
17454 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
17455 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
17456 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
17457 (ia64_vectorize_vec_perm_const_ok): Merge into...
17458 (ia64_vectorize_vec_perm_const): ...this new function.
17459 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
17460 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
17461 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
17462 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
17463 * config/mips/mips.c (mips_expand_vec_perm_const)
17464 (mips_vectorize_vec_perm_const_ok): Merge into...
17465 (mips_vectorize_vec_perm_const): ...this new function.
17466 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
17467 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
17468 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
17469 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
17470 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
17471 (rs6000_expand_vec_perm_const): Delete.
17472 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
17474 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
17475 (altivec_expand_vec_perm_const_le): Take each operand individually.
17476 Operate on constant selectors rather than rtxes.
17477 (altivec_expand_vec_perm_const): Likewise. Update call to
17478 altivec_expand_vec_perm_const_le.
17479 (rs6000_expand_vec_perm_const): Delete.
17480 (rs6000_vectorize_vec_perm_const_ok): Delete.
17481 (rs6000_vectorize_vec_perm_const): New function.
17482 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
17483 an element count and rtx array.
17484 (rs6000_expand_extract_even): Update call accordingly.
17485 (rs6000_expand_interleave): Likewise.
17486 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
17487 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
17488 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
17489 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
17490 (rs6000_expand_vec_perm_const): Delete.
17491 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
17492 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
17493 (altivec_expand_vec_perm_const_le): Take each operand individually.
17494 Operate on constant selectors rather than rtxes.
17495 (altivec_expand_vec_perm_const): Likewise. Update call to
17496 altivec_expand_vec_perm_const_le.
17497 (rs6000_expand_vec_perm_const): Delete.
17498 (rs6000_vectorize_vec_perm_const_ok): Delete.
17499 (rs6000_vectorize_vec_perm_const): New function. Remove stray
17500 reference to the SPE evmerge intructions.
17501 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
17502 an element count and rtx array.
17503 (rs6000_expand_extract_even): Update call accordingly.
17504 (rs6000_expand_interleave): Likewise.
17505 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
17506 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
17508 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
17510 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17512 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
17513 vector mode and that that mode matches the mode of the data
17515 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
17516 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
17517 directly using expand_vec_perm_1 when forcing selectors into
17519 (expand_vec_perm_var): New function, split out from expand_vec_perm.
17521 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17523 * optabs-query.h (can_vec_perm_p): Delete.
17524 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
17525 * optabs-query.c (can_vec_perm_p): Split into...
17526 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
17527 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
17528 particular selector is valid.
17529 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
17530 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
17531 (vect_grouped_load_supported): Likewise.
17532 (vect_shift_permute_load_chain): Likewise.
17533 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
17534 (vect_transform_slp_perm_load): Likewise.
17535 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
17536 (vectorizable_bswap): Likewise.
17537 (vect_gen_perm_mask_checked): Likewise.
17538 * fold-const.c (fold_ternary_loc): Likewise. Don't take
17539 implementations of variable permutation vectors into account
17540 when deciding which selector to use.
17541 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
17542 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
17543 with a false third argument.
17544 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
17545 to test whether the constant selector is valid and can_vec_perm_var_p
17546 to test whether a variable selector is valid.
17548 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17550 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
17551 * optabs-query.c (can_vec_perm_p): Likewise.
17552 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
17553 instead of vec_perm_indices.
17554 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
17555 (vect_gen_perm_mask_checked): Likewise,
17556 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
17557 (vect_gen_perm_mask_checked): Likewise,
17559 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17561 * optabs-query.h (qimode_for_vec_perm): Declare.
17562 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
17563 (qimode_for_vec_perm): ...this new function.
17564 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
17566 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
17568 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
17569 does not have a conditional at the top.
17571 2018-01-02 Richard Biener <rguenther@suse.de>
17573 * ipa-inline.c (big_speedup_p): Fix expression.
17575 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
17578 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
17581 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
17585 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
17586 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
17587 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
17588 cond_taken_branch_cost 3->4.
17590 2018-01-01 Jakub Jelinek <jakub@redhat.com>
17592 PR tree-optimization/83581
17593 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
17594 TODO_cleanup_cfg if any changes have been made.
17596 PR middle-end/83608
17597 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
17598 convert_modes if target mode has the right side, but different mode
17601 PR middle-end/83609
17602 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
17603 last argument when extracting from CONCAT. If either from_real or
17604 from_imag is NULL, use expansion through memory. If result is not
17605 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
17606 the parts directly to inner mode, if even that fails, use expansion
17609 PR middle-end/83623
17610 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
17611 check for bswap in mode rather than HImode and use that in expand_unop
17614 Copyright (C) 2018 Free Software Foundation, Inc.
17616 Copying and distribution of this file, with or without modification,
17617 are permitted in any medium without royalty provided the copyright
17618 notice and this notice are preserved.