1 2018-06-25 Martin Sebor <msebor@redhat.com>
3 * doc/extend.texi (Zero-length arrays): Update and clarify.
5 2018-06-25 Michael Meissner <meissner@linux.ibm.com>
7 * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
8 added IEEE/IBM long double multilib support on PowerPC little
10 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
11 (MULTILIB_DEFAULTS): Likewise.
12 * config/rs6000/rs6000.c (rs6000_option_override_internal):
14 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
15 * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
16 * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
18 2018-06-25 Alexander Monakov <amonakov@ispras.ru>
21 * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
22 (REORDER_45): Likewise.
24 2018-06-25 Jeff Law <law@redhat.com>
26 * config/v850/v850.md (divmodhi4): Make sure to sign extend the
27 dividend to 32 bits. Adjust length.
28 (udivmodhi4): Cleanup output template. Fix length.
30 2018-06-25 Carl Love <cel@us.ibm.com>
32 * config/rs6000/vsx.md: Change word selector to prefered location.
34 2018-06-25 Richard Biener <rguenther@suse.de>
36 PR tree-optimization/86304
37 * tree-vectorizer.c (vectorize_loops): Walk over new possibly
38 epilogue-if-converted loops as well.
40 2018-06-25 Jan Hubicka <hubicka@ucw.cz>
42 * lto-section-out.c (lto_begin_section): Do not print section
43 name for noaddr and unnumbered dumps.
45 2018-06-25 Richard Biener <rguenther@suse.de>
47 * tree-vectorizer.h (struct vec_info_shared): New structure
48 with parts split out from struct vec_info and loop_nest from
49 struct _loop_vec_info.
50 (struct vec_info): Adjust accordingly.
51 (struct _loop_vec_info): Likewise.
52 (LOOP_VINFO_LOOP_NEST): Adjust.
53 (LOOP_VINFO_DATAREFS): Likewise.
54 (LOOP_VINFO_DDRS): Likewise.
55 (struct _bb_vec_info): Likewise.
56 (BB_VINFO_DATAREFS): Likewise.
57 (BB_VINFO_DDRS): Likewise.
58 (struct _stmt_vec_info): Add dr_aux member.
59 (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
60 (DR_MISALIGNMENT_UNINITIALIZED): New.
61 (set_dr_misalignment): Adjust.
62 (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
63 (vect_analyze_loop): Adjust prototype.
64 (vect_analyze_loop_form): Likewise.
65 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
66 Compute dependences lazily.
67 (vect_record_base_alignments): Use shared datarefs/ddrs.
68 (vect_verify_datarefs_alignment): Likewise.
69 (vect_analyze_data_refs_alignment): Likewise.
70 (vect_analyze_data_ref_accesses): Likewise.
71 (vect_analyze_data_refs): Likewise.
72 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
73 constructor parameter for shared part.
74 (vect_analyze_loop_form): Pass in shared part and adjust.
75 (vect_analyze_loop_2): Pass in storage for the number of
76 stmts. Move loop nest finding to the caller. Compute
78 (vect_analyze_loop): Pass in shared part.
79 (vect_transform_loop): Verify shared datarefs are unchanged.
80 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
81 constructor parameter for shared part.
82 (vect_slp_analyze_bb_1): Pass in shared part and adjust.
83 (vect_slp_bb): Verify shared datarefs are unchanged before
85 * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
87 (new_stmt_vec_info): Initialize DR_AUX misalignment to
88 DR_MISALIGNMENT_UNINITIALIZED.
89 * tree-vectorizer.c (vec_info::vec_info): Add constructor
90 parameter for shared part.
91 (vec_info::~vec_info): Adjust.
92 (vec_info_shared::vec_info_shared): New.
93 (vec_info_shared::~vec_info_shared): Likewise.
94 (vec_info_shared::save_datarefs): Likewise.
95 (vec_info_shared::check_datarefs): Likewise.
96 (try_vectorize_loop_1): Construct shared part live for analyses
97 of a single loop for multiple vector sizes.
98 * tree-parloops.c (gather_scalar_reductions): Adjust.
100 2018-06-25 Richard Biener <rguenther@suse.de>
102 * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
103 DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
104 (vect_analyze_data_refs): Remove similar code from here and
105 simplify accordingly.
107 2018-06-25 Richard Biener <rguenther@suse.de>
109 * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
110 for reverse storage order accesses rather than asserting
111 they cannot happen here.
113 2018-06-25 Tom de Vries <tdevries@suse.de>
116 * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
117 Use data16 instead of .byte for insn prefix.
119 2018-06-25 Andreas Krebbel <krebbel@linux.ibm.com>
122 * parser.c (make_char_string_pack): Pass this literal chars
123 through cpp_interpret_string.
124 (cp_parser_userdef_numeric_literal): Check the result of
125 make_char_string_pack.
127 2018-06-24 Maya Rashish <coypu@sdf.org>
129 * ginclude/stddef.h: Simplify conditions around avoiding
130 re-definition of __size_t.
132 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
134 * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
135 unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
137 2018-06-22 Maya Rashish <coypu@sdf.org>
139 * doc/invoke.texi (mno-fancy-math-387): Update for changes
140 made to OpenBSD and NetBSD through the years.
142 2018-06-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
144 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
145 behavior of vec_pack (vector double, vector double) to match
146 behavior of vec_float2 (vector double, vector double).
148 2018-06-22 Olivier Hainque <hainque@adacore.com>
150 * gimplify.c (gimplify_function_tree): Prevent creation
151 of a trampoline for the address of the current function
152 passed to entry/exit instrumentation hooks.
154 2018-06-22 Aaron Sawdey <acsawdey@linux.ibm.com>
157 * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
160 2018-06-22 Martin Liska <mliska@suse.cz>
162 PR tree-optimization/86263
163 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
164 Bail out if is_enabled is false.
165 * tree-switch-conversion.h (jump_table_cluster::is_enabled):
167 (jump_table_cluster::is_enabled): New function.
169 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
171 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
172 BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
173 * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
174 (lto_input_ts_binfo_tree_pointers): Likewise.
175 * tree-streamer-out.c (streamer_write_tree_bitfields,
176 write_ts_binfo_tree_pointers): Likewise.
177 * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
179 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
181 * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
183 2018-06-22 Martin Liska <mliska@suse.cz>
185 * symbol-summary.h (get): Make it pure and inline move
186 functionality from ::get function.
187 (get): Remove and inline into ::get and ::get_create.
188 (get_create): Move code from ::get function.
190 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
193 * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
194 -x assembler-with-cpp.
196 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
198 * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
199 _FILE_OFFSET_BITS=64 for C++.
201 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
203 * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
204 conversion insn that shows up when pr85657-3.c is compiled using
205 IEEE 128-bit long double.
206 (neg<mode>2_internal): Use the correct mode to check whether the
207 mode is IBM extended.
208 * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
209 multiply and divide external functions from being created more
212 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
214 * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
216 (rtl_merge_blocks): Likewise. Do not emit a NOP if the location of
217 the edge can be forwarded.
218 (cfg_layout_merge_blocks): Likewise.
220 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
222 * except.c (finish_eh_generation): Commit edge insertions only after
223 the EH edges have been redirected from post-landing to landing pads.
225 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
227 * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
228 create_tmp_var_for to create the FRAME decl.
229 (finalize_nesting_tree_1): Do not unchain the FRAME decl.
231 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
233 * tree-inline.c (copy_edges_for_bb): Minor tweak.
234 (maybe_move_debug_stmts_to_successors): Also reset the locus of the
235 debug statement when resetting its value.
236 (expand_call_inline): Copy the locus of the call onto the assignment
237 of the return value, if any. Use local variable in more cases.
239 2018-06-21 Martin Liska <mliska@suse.cz>
241 * ipa-pure-const.c (propagate_nothrow): Use
242 funct_state_summaries->get.
243 (dump_malloc_lattice): Likewise.
244 (propagate_malloc): Likewise.
246 2018-06-21 Richard Biener <rguenther@suse.de>
248 * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
249 comment. Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
250 * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
252 * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
253 BLOCK_ABSTRACT_ORIGIN unconditionally.
255 2018-06-21 David Malcolm <dmalcolm@redhat.com>
257 * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
259 * ipa-reference.c (ipa_reference_c_finalize): Delete
260 ipa_ref_opt_sum_summaries and set it to NULL.
262 2018-06-21 Tom de Vries <tdevries@suse.de>
264 PR tree-optimization/85859
265 * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
266 test with comment from bb_no_side_effects_p.
268 2018-06-21 Richard Biener <rguenther@suse.de>
270 PR tree-optimization/86232
271 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
272 max for constant niter.
274 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
276 * config/aarch64/aarch64-simd.md
277 (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
279 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
281 * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
282 Make opernads of the unspec commutative.
284 2018-06-21 Richard Biener <rguenther@suse.de>
286 * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
287 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
288 Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
289 (vect_analyze_data_ref_dependence): Re-order checks to deal with
291 (vect_record_base_alignments): Do not record base alignment
292 for gathers or scatters.
293 (vect_compute_data_ref_alignment): Drop return value that is always
294 true. Bail out early for gathers or scatters.
295 (vect_enhance_data_refs_alignment): Bail out early for gathers
297 (vect_find_same_alignment_drs): Likewise.
298 (vect_analyze_data_refs_alignment): Remove dead code.
299 (vect_slp_analyze_and_verify_node_alignment): Likewise.
300 (vect_analyze_data_refs): For possible gathers or scatters do
301 not create an alternate DR, just check their possible validity
302 and mark them. Adjust DECL_NONALIASED handling to not rely
304 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
305 update inits of gathers or scatters.
306 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
307 Also copy gather/scatter flag to pattern vinfo.
309 2018-06-20 Kelvin Nilsen <kelvin@gcc.gnu.org>
311 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
312 behavior of vec_packsu (vector unsigned long long, vector unsigned
313 long long) to match behavior of vec_packs with same signature.
315 2018-06-20 Chung-Lin Tang <cltang@codesourcery.com>
316 Thomas Schwinge <thomas@codesourcery.com>
317 Cesar Philippidis <cesar@codesourcery.com>
319 * gimplify.c (gimplify_scan_omp_clauses): Add support for
320 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
321 (gimplify_adjust_omp_clauses): Likewise.
322 (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
323 support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
324 (gimplify_omp_target_update): Update handling of acc update and
326 * omp-low.c (install_var_field): Remove unused parameter
327 base_pointers_restrict.
328 (scan_sharing_clauses): Remove base_pointers_restrict parameter.
329 Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
331 (omp_target_base_pointers_restrict_p): Delete.
332 (scan_omp_target): Update call to scan_sharing_clauses.
333 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
335 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
336 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
337 (convert_local_omp_clauses): Likewise.
338 * tree-pretty-print.c (dump_omp_clause): Likewise.
339 * tree.c (omp_clause_num_ops): Add entries for OMP_CLAUSE_{IF_PRESENT,
341 (omp_clause_code_name): Likewise.
343 2018-06-20 Jakub Jelinek <jakub@redhat.com>
346 * var-tracking.c (use_narrower_mode_test): Check if shift amount can
349 PR tree-optimization/86231
350 * tree-vrp.c (union_ranges): For ( [ ) ] or ( )[ ] range and
351 anti-range don't overwrite *vr0min before using it to compute *vr0max.
353 2018-06-20 Tom de Vries <tdevries@suse.de>
355 PR tree-optimization/86097
356 * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
357 iv type if signedness of iv type is not the same as that of *nit.
359 2018-06-20 Jakub Jelinek <jakub@redhat.com>
361 * cfgrtl.c (rtl_verify_edges): Formatting fix. If bb->preds has any
362 EDGE_EH edges, verify they are all EDGE_EH.
364 2018-06-20 Maya Rashish <coypu@sdf.org>
366 * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
368 2018-06-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
370 * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
371 * config/aarch64/aarch64.c (xgene1_tunings): Add
372 AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
373 (aarch64_mode_valid_for_sched_fusion_p):
375 (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
376 * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
378 * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
380 (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
381 * config/aarch64/iterators.md (VQ2): New mode iterator.
383 2018-06-20 Martin Liska <mliska@suse.cz>
385 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
386 Change default ratio from 10 to 8.
388 2018-06-20 Martin Liska <mliska@suse.cz>
390 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
392 (bit_test_cluster::find_bit_tests): Likewise.
393 (switch_decision_tree::analyze_switch_statement): Find clusters.
394 * tree-switch-conversion.h (struct jump_table_cluster): Document
397 2018-06-20 Martin Liska <mliska@suse.cz>
399 * tree-switch-conversion.c (switch_conversion::collect):
400 Record m_uniq property.
401 (switch_conversion::expand): Bail out for special conditions.
402 (group_cluster::~group_cluster): New.
403 (group_cluster::group_cluster): Likewise.
404 (group_cluster::dump): Likewise.
405 (jump_table_cluster::emit): New.
406 (switch_decision_tree::fix_phi_operands_for_edges): New.
407 (struct case_node): Remove struct.
408 (jump_table_cluster::can_be_handled): New.
409 (case_values_threshold): Moved to header.
410 (reset_out_edges_aux): Likewise.
411 (jump_table_cluster::is_beneficial): New.
412 (bit_test_cluster::can_be_handled): Likewise.
413 (add_case_node): Remove.
414 (bit_test_cluster::is_beneficial): New.
415 (case_bit_test::cmp): New.
416 (bit_test_cluster::emit): New.
417 (expand_switch_as_decision_tree_p): Remove.
418 (bit_test_cluster::hoist_edge_and_branch_if_true): New.
419 (fix_phi_operands_for_edge): Likewise.
420 (switch_decision_tree::analyze_switch_statement): New.
421 (compute_cases_per_edge): Move ...
422 (switch_decision_tree::compute_cases_per_edge): ... here.
423 (try_switch_expansion): Likewise.
424 (switch_decision_tree::try_switch_expansion): Likewise.
425 (record_phi_operand_mapping): Likewise.
426 (switch_decision_tree::record_phi_operand_mapping): Likewise.
427 (emit_case_decision_tree): Likewise.
428 (switch_decision_tree::emit): Likewise.
429 (balance_case_nodes): Likewise.
430 (switch_decision_tree::balance_case_nodes): Likewise.
431 (dump_case_nodes): Likewise.
432 (switch_decision_tree::dump_case_nodes): Likewise.
433 (emit_jump): Likewise.
434 (switch_decision_tree::emit_jump): Likewise.
435 (emit_cmp_and_jump_insns): Likewise.
436 (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
437 (emit_case_nodes): Likewise.
438 (switch_decision_tree::emit_case_nodes): Likewise.
439 (conditional_probability): Remove.
440 * tree-switch-conversion.h (enum cluster_type): New.
442 (struct cluster): Likewise.
443 (cluster::cluster): Likewise.
444 (struct simple_cluster): Likewise.
445 (simple_cluster::simple_cluster): Likewise.
446 (struct group_cluster): Likewise.
447 (struct jump_table_cluster): Likewise.
448 (struct bit_test_cluster): Likewise.
449 (struct min_cluster_item): Likewise.
450 (struct case_tree_node): Likewise.
451 (case_tree_node::case_tree_node): Likewise.
452 (jump_table_cluster::case_values_threshold): Likewise.
453 (struct case_bit_test): Likewise.
454 (struct switch_decision_tree): Likewise.
455 (struct switch_conversion): Likewise.
456 (switch_decision_tree::reset_out_edges_aux): Likewise.
458 2018-06-20 Martin Liska <mliska@suse.cz>
460 * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
461 (hoist_edge_and_branch_if_true): Likewise.
462 (expand_switch_using_bit_tests_p): Likewise.
463 (struct case_bit_test): Likewise.
464 (case_bit_test_cmp): Likewise.
465 (emit_case_bit_tests): Likewise.
466 (switch_conversion::switch_conversion): New class.
467 (struct switch_conv_info): Remove old struct.
468 (collect_switch_conv_info): More to ...
469 (switch_conversion::collect): ... this.
470 (check_range): Likewise.
471 (switch_conversion::check_range): Likewise.
472 (check_all_empty_except_final): Likewise.
473 (switch_conversion::check_all_empty_except_final): Likewise.
474 (check_final_bb): Likewise.
475 (switch_conversion::check_final_bb): Likewise.
476 (create_temp_arrays): Likewise.
477 (switch_conversion::create_temp_arrays): Likewise.
478 (free_temp_arrays): Likewise.
479 (gather_default_values): Likewise.
480 (switch_conversion::gather_default_values): Likewise.
481 (build_constructors): Likewise.
482 (switch_conversion::build_constructors): Likewise.
483 (constructor_contains_same_values_p): Likewise.
484 (switch_conversion::contains_same_values_p): Likewise.
485 (array_value_type): Likewise.
486 (switch_conversion::array_value_type): Likewise.
487 (build_one_array): Likewise.
488 (switch_conversion::build_one_array): Likewise.
489 (build_arrays): Likewise.
490 (switch_conversion::build_arrays): Likewise.
491 (gen_def_assigns): Likewise.
492 (switch_conversion::gen_def_assigns): Likewise.
493 (prune_bbs): Likewise.
494 (switch_conversion::prune_bbs): Likewise.
495 (fix_phi_nodes): Likewise.
496 (switch_conversion::fix_phi_nodes): Likewise.
497 (gen_inbound_check): Likewise.
498 (switch_conversion::gen_inbound_check): Likewise.
499 (process_switch): Use the newly created class.
500 (switch_conversion::expand): New.
501 (switch_conversion::~switch_conversion): New.
502 * tree-switch-conversion.h: New file.
504 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
506 * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
507 tree-vect-patterns.c.
508 * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
509 (vect_recog_dot_prod_pattern): Use it. Remove the type_in argument.
510 (vect_recog_sad_pattern): Likewise.
511 (vect_recog_widen_sum_pattern): Likewise.
512 (vect_recog_pow_pattern): Likewise. Check for a null vectype.
513 (vect_recog_widen_shift_pattern): Remove the type_in argument.
514 (vect_recog_rotate_pattern): Likewise.
515 (vect_recog_mult_pattern): Likewise.
516 (vect_recog_vector_vector_shift_pattern): Likewise.
517 (vect_recog_divmod_pattern): Likewise.
518 (vect_recog_mixed_size_cond_pattern): Likewise.
519 (vect_recog_bool_pattern): Likewise.
520 (vect_recog_mask_conversion_pattern): Likewise.
521 (vect_try_gather_scatter_pattern): Likewise.
522 (vect_recog_widen_mult_pattern): Likewise. Check for a null vectype.
523 (vect_recog_over_widening_pattern): Likewise.
524 (vect_recog_gather_scatter_pattern): Likewise.
525 (vect_recog_func_ptr): Move from tree-vectorizer.h
526 (vect_vect_recog_func_ptrs): Move further down the file.
527 (vect_recog_func): Likewise. Remove the third argument.
528 (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
529 (vect_pattern_recog_1): Expect the pattern function to do any
530 necessary target tests. Also expect it to provide a vector type.
531 Remove the type_in handling.
533 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
535 * tree-vect-patterns.c (vect_pattern_detected): New function.
536 (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
537 (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
538 (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
539 (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
540 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
541 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
542 (vect_recog_mask_conversion_pattern)
543 (vect_try_gather_scatter_pattern): Likewise.
545 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
547 * tree-vect-patterns.c (vect_get_internal_def): New function.
548 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
549 (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
550 (search_type_for_mask_1): Use it.
552 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
554 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
555 redundant WIDEN_SUM_EXPR handling.
556 (vect_recog_sad_pattern): Likewise.
558 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
560 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
561 redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
562 (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
563 (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
564 (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
566 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
568 * tree-vect-stmts.c (vectorizable_call): Make sure that we
569 use the stmt_vec_info of the original bb statement for the
570 new zero assignment, even if the call is part of a pattern.
572 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
574 * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
575 that the sequence is attached to the original statement rather
576 than the pattern statement.
577 * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
578 PATTERN_DEF_SEQ from the original statement rather than
579 the main pattern statement.
580 * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
581 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
582 (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
584 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
586 * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
587 definition statements before the early exit for statements that aren't
589 * tree-vect-loop.c (vect_transform_loop_stmt): New function,
591 (vect_transform_loop): ...here. Process pattern definition
592 statements without first checking whether the main pattern
593 statement is live or relevant.
595 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
597 * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
598 -O0 if the locus represent UNKNOWN_LOCATION but have different values.
600 2018-06-19 Aaron Sawdey <acsawdey@linux.ibm.com>
602 * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
603 TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
604 (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
605 (expand_block_compare): Change select_block_compare_mode call.
606 (expand_strncmp_align_check): Use new functions, fix comment.
607 (emit_final_str_compare_gpr): New function.
608 (expand_strn_compare): Refactor and clean up code.
609 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
611 2018-06-19 Tony Reix <tony.reix@atos.com>
612 Damien Bergamini <damien.bergamini@atos.com>
613 David Edelsohn <dje.gcc@gmail.com>
615 * collect2.c (static_obj): New variable.
616 (static_libs): New variable.
617 (is_in_list): Uncomment declaration.
618 (main): Track AIX libraries linked statically.
619 (is_in_list): Uncomment definition.
620 (scan_prog_file): Don't add AIX shared libraries initializer
621 to constructor list if linking statically.
623 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
625 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
627 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
629 2018-06-19 Jan Hubicka <hubicka@ucw.cz>
631 * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
634 2018-06-19 Martin Liska <mliska@suse.cz>
636 * config/i386/i386.c (ix86_can_inline_p): Do not use
637 ipa_fn_summaries::get_create.
638 * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
640 (devirtualization_time_bonus): Likewise.
641 (ipcp_propagate_stage): Likewise.
642 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
643 (edge_set_predicate): Likewise.
644 (evaluate_conditions_for_known_args): Likewise.
645 (evaluate_properties_for_edge): Likewise.
646 (ipa_call_summary::reset): Tranform to ...
647 (ipa_call_summary::~ipa_call_summary): ... this.
648 (ipa_fn_summary::reset): Transform to ...
649 (ipa_fn_summary::~ipa_fn_summary): ... this.
650 (ipa_fn_summary_t::remove): Rename to ...
651 (ipa_fn_summary_t::remove_callees): ... this.
652 (ipa_fn_summary_t::duplicate): Use placement new
653 instead of memory copy.
654 (ipa_call_summary_t::duplicate): Likewise.
655 (ipa_call_summary_t::remove): Remove.
656 (dump_ipa_call_summary): Change get_create to get.
657 (ipa_dump_fn_summary): Dump only when summary exists.
658 (analyze_function_body): Use symbol_summary::get instead
660 (compute_fn_summary): Likewise.
661 (estimate_edge_devirt_benefit): Likewise.
662 (estimate_edge_size_and_time): Likewise.
663 (inline_update_callee_summaries): Likewise.
664 (remap_edge_change_prob): Likewise.
665 (remap_edge_summaries): Likewise.
666 (ipa_merge_fn_summary_after_inlining): Likewise.
667 (write_ipa_call_summary): Likewise.
668 (ipa_fn_summary_write): Likewise.
669 (ipa_free_fn_summary): Likewise.
670 * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
671 (struct ipa_call_summary): Likewise.
672 * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
674 * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
675 (estimate_size_after_inlining): Likewise.
676 (estimate_growth): Likewise.
677 (growth_likely_positive): Likewise.
678 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
679 (inline_call): Likewise.
680 * ipa-inline.c (caller_growth_limits): Likewise.
681 (can_inline_edge_p): Likewise.
682 (can_inline_edge_by_limits_p): Likewise.
683 (compute_uninlined_call_time): Likewise.
684 (compute_inlined_call_time): Likewise.
685 (want_inline_small_function_p): Likewise.
686 (edge_badness): Likewise.
687 (update_caller_keys): Likewise.
688 (update_callee_keys): Likewise.
689 (inline_small_functions): Likewise.
690 (inline_to_all_callers_1): Likewise.
691 (dump_overall_stats): Likewise.
692 (early_inline_small_functions): Likewise.
693 (early_inliner): Likewise.
694 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
695 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
696 * ipa-pure-const.c (malloc_candidate_p): Likewise.
697 * ipa-split.c (execute_split_functions): Likewise.
698 * symbol-summary.h: Likewise.
699 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
701 2018-06-19 Richard Biener <rguenther@suse.de>
703 * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
704 (vectorize_loops): ... here. Fix dbgcnt handling.
705 (try_vectorize_loop): Wrap try_vectorize_loop_1.
707 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
710 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
711 ieee128 argument takes up only one (vector) register, not two (floating
714 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
716 * gimplify.c (gimplify_init_constructor): Really never clear for an
717 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
719 2018-06-19 Richard Biener <rguenther@suse.de>
721 PR tree-optimization/86179
722 * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
723 after failed recognition.
725 2018-06-18 Martin Sebor <msebor@redhat.com>
728 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
729 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
730 Handle integer subtraction.
731 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
732 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
734 2018-06-18 David Malcolm <dmalcolm@redhat.com>
736 * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
737 param from rtx to rtx_insn *.
738 * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
740 (frv_ifcvt_modify_insn): Likwise.
741 (frv_ifcvt_modify_final): Likwise for local "existing_insn",
742 adding an as_a <rtx_insn *> cast. Likewise for local "insn".
743 * config/mips/mips.c (r10k_insert_cache_barriers): Add an
744 as_a <rtx_insn *> cast to local "unprotected_region" once
745 it's been established that it's not NULL or pc_rtx.
746 * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
747 param "sethi" from rtx to rtx_insn *.
748 (nds32_group_float_insns): Likewise for param "insn".
749 * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
751 (vax_output_int_subtract): Likewise.
752 * config/vax/vax.c (vax_output_int_add): Likewise for param
754 (vax_output_int_subtract): Likewise.
755 * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
756 (emit_pattern_after): Likewise for param "after".
757 (emit_insn_after): Likewise.
758 (emit_jump_insn_after): Likewise.
759 (emit_call_insn_after): Likewise.
760 (emit_debug_insn_after): Likewise.
761 (emit_pattern_before): Likewise for param "before".
762 (emit_insn_before): Likewise.
763 (emit_jump_insn_before): Likewise.
764 * final.c (get_insn_template): Likewise for param "insn", removing
766 * output.h (get_insn_template): Likewise for 2nd param.
767 * rtl.h (emit_insn_before): Likewise.
768 (emit_jump_insn_before): Likewise.
769 (emit_debug_insn_before_noloc): Likewise.
770 (emit_insn_after): Likewise.
771 (emit_jump_insn_after): Likewise.
772 (emit_call_insn_after): Likewise.
773 (emit_debug_insn_after): Likewise.
774 (set_insn_deleted): Likewise for param.
776 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
779 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
780 floating point modes, so that IFmode is numerically greater than
781 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
782 to declare the ordering. This prevents IFmode from being
783 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
784 machine. Include rs6000-modes.h to share the fractional values
785 between genmodes* and the rest of the compiler.
789 * config/rs6000/rs6000-modes.h: New file.
790 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
791 meaning of rs6000_long_double_size so that 126..128 selects an
792 appropriate 128-bit floating point type.
793 (rs6000_option_override_internal): Likewise.
794 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
795 (TARGET_LONG_DOUBLE_128): Change the meaning of
796 rs6000_long_double_size so that 126..128 selects an appropriate
797 128-bit floating point type.
798 (LONG_DOUBLE_TYPE_SIZE): Update comment.
799 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
800 source and destination to match the standard usage.
801 (truncifkf2): Likewise.
802 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
803 ISA 2.07 to use an explicit clobber, instead of passing in a
805 (copysign<mode>3_soft): Likewise.
807 2018-06-18 David Malcolm <dmalcolm@redhat.com>
809 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
810 Replace dump_printf_loc call with DUMP_VECT_SCOPE.
811 (vect_slp_analyze_instance_dependence): Likewise.
812 (vect_enhance_data_refs_alignment): Likewise.
813 (vect_analyze_data_refs_alignment): Likewise.
814 (vect_slp_analyze_and_verify_instance_alignment
815 (vect_analyze_data_ref_accesses): Likewise.
816 (vect_prune_runtime_alias_test_list): Likewise.
817 (vect_analyze_data_refs): Likewise.
818 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
819 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
820 (vect_analyze_scalar_cycles_1): Likewise.
821 (vect_get_loop_niters): Likewise.
822 (vect_analyze_loop_form_1): Likewise.
823 (vect_update_vf_for_slp): Likewise.
824 (vect_analyze_loop_operations): Likewise.
825 (vect_analyze_loop): Likewise.
826 (vectorizable_induction): Likewise.
827 (vect_transform_loop): Likewise.
828 * tree-vect-patterns.c (vect_pattern_recog): Likewise.
829 * tree-vect-slp.c (vect_analyze_slp): Likewise.
830 (vect_make_slp_decision): Likewise.
831 (vect_detect_hybrid_slp): Likewise.
832 (vect_slp_analyze_operations): Likewise.
833 (vect_slp_bb): Likewise.
834 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
835 (vectorizable_bswap): Likewise.
836 (vectorizable_call): Likewise.
837 (vectorizable_simd_clone_call): Likewise.
838 (vectorizable_conversion): Likewise.
839 (vectorizable_assignment): Likewise.
840 (vectorizable_shift): Likewise.
841 (vectorizable_operation): Likewise.
842 * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
844 2018-06-18 Martin Sebor <msebor@redhat.com>
846 PR tree-optimization/81384
847 * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
848 * builtins.c (expand_builtin_strnlen): New function.
849 (expand_builtin): Call it.
850 (fold_builtin_n): Avoid setting TREE_NO_WARNING.
851 * builtins.def (BUILT_IN_STRNLEN): New.
852 * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
853 Warn for bounds in excess of maximum object size.
854 * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
855 single-value ranges. Handle strnlen.
856 (handle_builtin_strlen): Handle strnlen.
857 (strlen_check_and_optimize_stmt): Same.
858 * doc/extend.texi (Other Builtins): Document strnlen.
860 2018-06-18 Maya Rashish <coypu@sdf.org>
862 * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
863 (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
864 (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
866 * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
868 * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
870 2018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
872 * tree.c (escaped_string::escape): Replace cast to char * by
873 const_cast<char *> (unescaped).
875 2018-06-18 Nick Clifton <nickc@redhat.com>
878 * tree.c (escaped_string): New class. Converts an unescaped
879 string into its escaped equivalent.
880 (warn_deprecated_use): Use the new class to convert the
881 deprecation message, if present.
882 (test_escaped_strings): New self test.
883 (test_c_tests): Add test_escaped_strings.
884 * doc/extend.texi (deprecated): Add a note that the
885 deprecation message is affected by the -fmessage-length
886 option, and that control characters will be escaped.
887 (#pragma GCC error): Document this pragma.
888 (#pragma GCC warning): Likewise.
889 * doc/invoke.texi (-fmessage-length): Document this option's
890 effect on the #warning and #error preprocessor directives and
891 the deprecated attribute.
893 2018-06-18 Eric Botcazou <ebotcazou@adacore.com>
895 * tree.c (decl_value_expr_lookup): Revert latest change.
896 (decl_value_expr_insert): Likewise.
898 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
900 * gimplify.c (nonlocal_vlas): Delete.
901 (nonlocal_vla_vars): Likewise.
902 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
904 (gimplify_body): Do not create and destroy nonlocal_vlas.
905 * tree-nested.c: Include diagnostic.h.
906 (use_pointer_in_frame): Tweak.
907 (lookup_field_for_decl): Add assertion and declare the transformation.
908 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
909 internal error when the reference is in a wrong context. Do not
910 create a debug decl by default.
911 (note_nonlocal_block_vlas): Delete.
912 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
913 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
914 create a debug decl by default.
915 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
916 call is in a wrong context.
917 (fixup_vla_decls): New function.
918 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
919 debug variables were created.
920 * tree.c (decl_value_expr_lookup): Add checking assertion.
921 (decl_value_expr_insert): Likewise.
923 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
926 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
927 * tree-scalar-evolution.c (interpret_expr): Likewise.
928 (expression_expensive_p): Likewise.
929 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
930 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
931 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
932 (ssa_defined_by_minus_one_stmt_p): New.
934 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
937 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
938 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
939 * dojump.c (do_jump): Likewise.
940 * expr.c (expand_expr_real_2): Check operand type's sign.
941 * fold-const.c (const_unop): Handle ABSU_EXPR.
942 (fold_abs_const): Likewise.
943 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
944 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
945 (strip_sign_op_1): Likesise.
946 * match.pd: Add new pattern to generate ABSU_EXPR.
947 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
948 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
949 * tree-eh.c (operation_could_trap_helper_p): Likewise.
950 * tree-inline.c (estimate_operator_cost): Likewise.
951 * tree-pretty-print.c (dump_generic_node): Likewise.
952 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
953 * tree.def (ABSU_EXPR): New.
955 2018-06-16 Jakub Jelinek <jakub@redhat.com>
958 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
959 documented as preserved for backward compatibility only.
960 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
962 PR rtl-optimization/86108
963 * bb-reorder.c (create_forwarder_block): Renamed to ...
964 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
965 jump from new landing pad to the second part.
966 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
969 2018-06-15 Jakub Jelinek <jakub@redhat.com>
972 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
973 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
974 Only call store_expr for halves if the mode is the same.
977 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
978 Fix up comment formatting.
980 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
982 * typed-splay-tree.h (typed_splay_tree::remove): New function.
983 (typed_splay_tree::closure,
984 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
985 (typed_splay_tree::typed_splay_tree,
986 typed_splay_tree::operator =): Declared private.
987 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
988 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
989 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
990 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
991 typed_splay_tree::splay_tree_splay,
992 typed_splay_tree::splay_tree_foreach_helper,
993 typed_splay_tree::splay_tree_insert,
994 typed_splay_tree::splay_tree_remove,
995 typed_splay_tree::splay_tree_lookup,
996 typed_splay_tree::splay_tree_predecessor,
997 typed_splay_tree::splay_tree_successor,
998 typed_splay_tree::splay_tree_min,
999 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
1000 (typed_splay_tree::root, typed_splay_tree::comp,
1001 typed_splay_tree::delete_key,
1002 typed_splay_tree::delete_value): New data members.
1003 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
1004 typed_splay_tree::remove.
1006 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
1008 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
1009 -mginv and -mno-ginv to the assembler.
1010 * config/mips/mips.opt (-mcrc): New option.
1012 * doc/invoke.text (-mcrc): Document.
1015 2018-06-15 Nick Clifton <nickc@redhat.com>
1018 * tree.c (escaped_string): New class. Converts an unescaped
1019 string into its escaped equivalent.
1020 (warn_deprecated_use): Use the new class to convert the
1021 deprecation message, if present.
1022 (test_escaped_strings): New self test.
1023 (test_c_tests): Add test_escaped_strings.
1024 * doc/extend.texi (deprecated): Add a note that the
1025 deprecation message is affected by the -fmessage-length
1026 option, and that control characters will be escaped.
1027 (#pragma GCC error): Document this pragma.
1028 (#pragma GCC warning): Likewise.
1029 * doc/invoke.texi (-fmessage-length): Document this option's
1030 effect on the #warning and #error preprocessor directives and
1031 the deprecated attribute.
1033 2018-06-15 Richard Biener <rguenther@suse.de>
1035 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
1036 here, also noting vector size used.
1037 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
1038 size used in MSG_OPTIMIZED_LOCATIONS dump.
1039 (pass_slp_vectorize::execute): Adjust.
1041 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
1044 * config/arc/arc.c (arc_return_address_register): Fix
1047 2018-06-15 Richard Biener <rguenther@suse.de>
1050 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
1051 leave useless conversion stripping to force_gimple_operand_gsi.
1052 (gimplify_build2): Likewise.
1053 (gimplify_build1): Likewise.
1055 2018-06-15 Richard Biener <rguenther@suse.de>
1058 * tree-cfg.c (move_stmt_op): unshare invariant addresses
1059 before adjusting their block.
1061 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
1063 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
1064 multilibs for *-*-rtems*.
1065 * config/riscv/t-rtems: New file.
1067 2018-06-14 Jakub Jelinek <jakub@redhat.com>
1070 * match.pd ((A +- CST1) +- CST2): Punt if last resort
1071 unsigned_type_for returns NULL.
1074 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
1075 subregs of multi-word pseudos unless the float mode has word size.
1077 2018-06-14 Richard Biener <rguenther@suse.de>
1080 * tree-vect-generic.c (build_word_mode_vector_type): Remove
1081 duplicate and harmful type_hash_canon.
1082 * tree.c (type_hash_canon): Assert we didn't find ourselves.
1084 2018-06-14 Richard Biener <rguenther@suse.de>
1087 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
1090 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
1092 * config/rtems.h (STDINT_LONG32): Define.
1094 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
1095 Prachi Godbole <prachi.godbole@imgtec.com>
1097 * config/mips/mips-cpus.def: Define P6600.
1098 * config/mips/mips-tables.opt: Regenerate.
1099 * config/mips/mips.c (mips_ucbranch_type): New enum.
1100 (mips_rtx_cost_data): Add support for P6600.
1101 (mips_issue_rate): Likewise.
1102 (mips_multipass_dfa_lookahead): Likewise.
1103 (mips_avoid_hazard): Likewise.
1104 (mips_reorg_process_insns): Likewise.
1105 (mips_classify_branch_p6600): New function.
1106 * config/mips/mips.h (TUNE_P6600): New define.
1107 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
1108 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
1109 * config/mips/mips.md: Include p6600.md.
1110 (processor): Add p6600.
1111 * config/mips/p6600.md: New file.
1112 * doc/invoke.texi: Add p6600 to supported architectures.
1114 2018-06-13 Martin Sebor <msebor@redhat.com>
1116 PR tree-optimization/86114
1117 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
1119 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
1121 2018-06-13 Richard Biener <rguenther@suse.de>
1123 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
1124 Properly set vector type of the intermediate stmt.
1125 * tree-vect-stmts.c (vectorizable_operation): The destination
1126 var always has vectype_out type.
1128 2018-06-13 Jeff Law <law@redhat.com>
1130 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
1131 integer 0 for argument to print_rtl_with_bb.
1132 (rl78_reorg): Likewise.
1134 2018-06-13 David Malcolm <dmalcolm@redhat.com>
1136 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
1137 from rtx to rtx_insn *.
1138 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
1140 (add_sched_insns_for_speculation): Likewise for local "target",
1141 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
1142 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
1143 from rtx_insn ** to rtx_code_label **.
1144 (reorg_emit_nops): Likewise.
1145 (c6x_reorg): Likewise for local "call_labels".
1146 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
1148 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
1149 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
1150 the loops over LABEL_REFS.
1151 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
1153 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
1154 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
1155 (split_branches): Strengthen local "olabel" from rtx to
1156 rtx_insn *, adding a safe_as_a cast.
1157 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
1159 (add_insn_after): Likewise for first two params.
1160 (add_insn_before): Likewise.
1161 (remove_insn): Likewise for param.
1162 (emit_pattern_before_noloc): Likewise for second and third params.
1163 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
1164 (emit_call_insn_before_noloc): Likewise.
1165 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
1167 (emit_barrier_before): Likewise.
1168 (emit_label_before): Strengthen "label" param from "rtx" to
1169 "rtx_code_label *". Strengthen "before" param from "rtx" to
1171 (emit_insn_after_1): Strengthen "after" param from "rtx" to
1173 (emit_pattern_after_noloc): Likewise.
1174 (emit_insn_after_noloc): Likewise.
1175 (emit_jump_insn_after_noloc): Likewise.
1176 (emit_call_insn_after_noloc): Likewise.
1177 (emit_debug_insn_after_noloc): Likewise.
1178 (emit_barrier_after): Likewise.
1179 (emit_label_after): Likewise for both params.
1180 (emit_pattern_after_setloc): Likewise for "after" param. Convert
1181 "loc" param from "int" to "location_t".
1182 (emit_insn_after_setloc): Likewise.
1183 (emit_jump_insn_after_setloc): Likewise.
1184 (emit_call_insn_after_setloc): Likewise.
1185 (emit_debug_insn_after_setloc): Likewise.
1186 (emit_pattern_before_setloc): Likewise for "before" param. Convert
1187 "loc" param from "int" to "location_t".
1188 (emit_pattern_before): Convert NULL_RTX to NULL.
1189 (emit_insn_before_setloc): Convert "loc" param from "int" to
1191 (emit_jump_insn_before_setloc): Likewise.
1192 (emit_call_insn_before_setloc): Likewise.
1193 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
1194 rtx_insn *. Convert "loc" param from "int" to "location_t".
1195 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
1196 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
1197 Convert 3rd param from "int" to "location_t".
1198 (emit_barrier_before, emit_barrier_after, next_real_insn):
1199 Strengthen param from rtx to rtx_insn *.
1200 (emit_label_before): Strengthen 1st param from "rtx" to
1201 "rtx_code_label *". Strengthen 2nd param from "rtx" to
1203 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
1204 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
1205 Strengthen 2nd param from "rtx" to "rtx_insn *".
1206 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
1207 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
1208 Likewise. Convert 3rd param from "int" to "location_t".
1209 (emit_label_after): Strengthen 1st param from "rtx" to
1211 (next_real_insn, remove_insn): Strengthen param from "rtx" to
1213 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
1214 from "rtx" to "rtx_insn *".
1216 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
1218 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
1219 bodies streamed in with -Q.
1220 * dumpfile.c (dump_files): Add lto-stream-out dump file.
1221 * dumpfile.h (tree_dump_index): Add lto_stream_out.
1222 * gimple-streamer-out.c: Include gimple-pretty-print.h
1223 (output_bb): Dump stmts streamed.
1224 * lto-section-out.c: Include print-tree.h
1225 (lto_begin_section): Dump sections created.
1226 (lto_output_decl_index): Dump decl encoded.
1227 * lto-streamer-out.c: Include print-tree.h
1228 (create_output_block): Dump output block created.
1229 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
1230 (output_function): Dump function output.
1231 (output_constructor): Dump constructor streamed.
1232 (write_global_stream): Output indexes encoded.
1233 (produce_asm_for_decls): Dump streams encoded.
1234 * lto-streamer.c (streamer_dump_file): New global var.
1235 * lto-streamer.h (streamer_dump_file): Declare.
1236 * passes.c (ipa_write_summaries): Initialize streamer dump.
1237 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
1240 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
1243 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
1244 offsets for register save directives. Emit a second batch of save
1245 directives, if need be, when the function accesses prior frames.
1247 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1249 * config/arc/fpu.md (fmasf4): Force operand to register.
1250 (fnmasf4): Likewise.
1252 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1254 * config/arc/arc-protos.h (arc_pad_return): Remove.
1255 * config/arc/arc.c (machine_function): Remove force_short_suffix
1257 (arc_print_operand): Adjust printing of '&'.
1258 (arc_verify_short): Remove conditional printing of short suffix.
1259 (arc_final_prescan_insn): Remove reference to size_reason.
1260 (pad_return): New function.
1261 (arc_reorg): Call pad_return.
1262 (arc_pad_return): Remove.
1263 (arc_init_machine_status): Remove reference to force_short_suffix.
1264 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
1265 (attr length): When attribute iscompact is true force to 2
1266 regardless; in the case of maybe check if we want to force the
1267 instruction to have 4 bytes length.
1268 (nopv): Change it to generate 4 byte long nop as well.
1269 (blockage): New pattern.
1270 (simple_return): Remove call to arc_pad_return.
1271 (p_return_i): Likewise.
1273 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1275 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
1277 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1279 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
1282 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1284 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
1285 for ARC700 and ARCv2.
1287 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
1290 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
1291 operands[2] instead of operands[1].
1294 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
1296 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
1297 case, check whether the outer register overlaps an unallocatable
1298 register, not just whether it fits the required class.
1300 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
1302 * poly-int.h (can_div_trunc_p): Add new overload in which all values
1304 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
1305 (memrefs_conflict_p): Likewise.
1306 (init_alias_analysis): Likewise.
1307 * cfgexpand.c (expand_debug_expr): Likewise.
1308 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
1309 * cse.c (fold_rtx): Likewise.
1310 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
1311 * expr.c (emit_block_move_hints): Likewise.
1312 (clear_storage_hints, push_block, emit_push_insn): Likewise.
1313 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
1314 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
1315 (emit_group_store): Likewise.
1316 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
1317 to read the PRE/POST_MODIFY increment.
1318 * calls.c (store_one_arg): Use strip_offset.
1319 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
1321 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
1323 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
1324 (simplify_binary_operation_1): Extend CONST_INT handling to
1326 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
1327 than a HOST_WIDE_INT.
1328 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
1330 (adjust_mems, add_stores): Update accodingly.
1331 (vt_canonicalize_addr): Track polynomial offsets.
1332 (emit_note_insn_var_location): Likewise.
1333 (vt_add_function_parameter): Likewise.
1334 (vt_initialize): Likewise.
1336 2018-06-12 Jeff Law <law@redhat.com>
1338 * config.gcc (alpha*-*-freebsd*): Remove.
1339 * config/alpha/freebsd.h: Remove.
1341 2018-06-12 David Malcolm <dmalcolm@redhat.com>
1344 * spellcheck-tree.c (levenshtein_distance): Rename to...
1345 (get_edit_distance): ...this, and update for underlying renaming.
1346 * spellcheck-tree.h (levenshtein_distance): Rename to...
1347 (get_edit_distance): ...this.
1348 * spellcheck.c (levenshtein_distance): Rename to...
1349 (get_edit_distance): ...this. Convert from Levenshtein distance
1350 to Damerau-Levenshtein distance by supporting transpositions of
1351 adjacent characters. Rename "v1" to "v_next" and "v0" to
1353 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
1354 (selftest::test_edit_distance_unit_test_oneway): ...this, and
1355 update for underlying renaming.
1356 (selftest::levenshtein_distance_unit_test): Rename to...
1357 (selftest::test_get_edit_distance_unit): ...this, and update for
1358 underlying renaming.
1359 (selftest::test_find_closest_string): Add example from PR 69968
1360 where transposition helps
1361 (selftest::test_metric_conditions): Update for renaming.
1362 (selftest::test_metric_conditions): Likewise.
1363 (selftest::spellcheck_c_tests): Likewise.
1364 * spellcheck.h (levenshtein_distance): Rename both overloads to...
1365 (get_edit_distance): ...this.
1366 (best_match::consider): Update for renaming.
1368 2018-06-12 Martin Sebor <msebor@redhat.com>
1370 PR tree-optimization/85259
1371 * builtins.c (compute_objsize): Handle constant offsets.
1372 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
1373 true iff a warning has been issued.
1374 * gimple.h (gimple_nonartificial_location): New function.
1375 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
1376 gimple_nonartificial_location and handle -Wno-system-headers.
1377 (handle_builtin_stxncpy): Same.
1379 2018-06-12 Martin Sebor <msebor@redhat.com>
1382 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
1384 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
1386 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
1387 BUILTIN_VEC_XST entries for pointer to double and long long.
1389 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
1392 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
1394 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
1397 2018-06-12 Martin Liska <mliska@suse.cz>
1399 * doc/options.texi: Document IntegerRange.
1401 2018-06-12 Martin Liska <mliska@suse.cz>
1403 * config/i386/i386.opt: Make MPX-related options as Deprecated.
1404 * opt-functions.awk: Handle Deprecated flag.
1405 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
1407 (read_cmdline_option): Report warning for a deprecated option.
1408 * opts.h (struct cl_option): Add new field cl_deprecated.
1409 (CL_ERR_DEPRECATED): New.
1411 2018-06-12 Martin Liska <mliska@suse.cz>
1413 * doc/options.texi: Document Deprecated option flag.
1415 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1417 * config/arc/arc-arch.h (arc_extras): New enum.
1418 (arc_cpu_t):Add field extra.
1419 (arc_cpu_types): Consider the extras.
1420 * config/arc/arc-cpus.def: Add extras info.
1421 * config/arc/arc-opts.h (processor_type): Consider extra field.
1422 * config/arc/arc.c (arc_override_options): Handle extra field.
1424 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1426 * config/arc/arc-arch.h: Update ARC_OPTX macro.
1427 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
1429 * config/arc/arc.c (arc_init): Update pic warning.
1430 (irq_range): Update irq range parsing warnings.
1431 (arc_override_options): Update various warning messages.
1432 (arc_handle_aux_attribute): Likewise.
1434 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
1436 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
1438 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
1440 * doc/sourcebuild.texi: Document usage of line number 0 in verify
1441 compiler messages directives.
1443 2018-06-12 Matthew Fortune <mfortune@gmail.com>
1445 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
1446 * config/mips/mips-tables.opt: Regenerate.
1447 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
1449 * doc/invoke.texi: Document -march=i6500.
1451 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
1453 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
1454 (i6400_gpmul): Add cpu_unit.
1455 (i6400_gpdiv): Likewise.
1456 (i6400_msa_add_d): Update reservations.
1457 (i6400_msa_int_add) Likewise.
1458 (i6400_msa_short_logic3) Likewise.
1459 (i6400_msa_short_logic2) Likewise.
1460 (i6400_msa_short_logic) Likewise.
1461 (i6400_msa_move) Likewise.
1462 (i6400_msa_cmp) Likewise.
1463 (i6400_msa_short_float2) Likewise.
1464 (i6400_msa_div_d) Likewise.
1465 (i6400_msa_long_logic1) Likewise.
1466 (i6400_msa_long_logic2) Likewise.
1467 (i6400_msa_mult) Likewise.
1468 (i6400_msa_long_float2) Likewise.
1469 (i6400_msa_long_float4) Likewise.
1470 (i6400_msa_long_float5) Likewise.
1471 (i6400_msa_long_float8) Likewise.
1472 (i6400_fpu_fadd): Include frint type.
1473 (i6400_fpu_store): New define_insn_reservation.
1474 (i6400_fpu_load): Likewise.
1475 (i6400_fpu_move): Likewise.
1476 (i6400_fpu_fcmp): Likewise.
1477 (i6400_fpu_fmadd): Likewise.
1478 (i6400_int_mult): Include imul3nc type and update reservation.
1479 (i6400_int_div): Include idiv3 type and update reservation.
1480 (i6400_int_load): Update to check type not move_type.
1481 (i6400_int_store): Likewise.
1482 (i6400_int_prefetch): Set zero latency.
1484 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
1486 * gcc.c: Document new %@{...} sequence.
1487 (LINK_COMMAND_SPEC): Use it for the -L switches.
1488 (cpp_unique_options): Use it for the -I switches.
1489 (at_file_argbuf): New global variable.
1490 (in_at_file): Likewise.
1491 (alloc_args): Create at_file_argbuf.
1492 (clear_args): Truncate at_file_argbuf.
1493 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
1494 (open_at_file): New function.
1495 (close_at_file): Likewise.
1496 (create_at_file): Delete.
1497 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
1500 (validate_switches_from_spec): Deal with %@{...} sequence.
1501 (validate_switches): Likewise.
1502 (driver::finalize): Call clear_args.
1504 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
1506 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
1508 2018-06-11 Martin Sebor <msebor@redhat.com>
1510 * doc/invoke.texi (-Wall): List -Wc++17-compat.
1511 (Wno-class-memaccess): Add @opindex.
1512 (Wno-templates, Wno-multiple-inheritance): Same.
1513 (Wno-virtual-inheritance, Wno-namespaces): Same.
1514 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
1515 (Wno-format-overflow, Wno-format-truncation): Same.
1516 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
1517 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
1518 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
1519 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
1520 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
1521 (Wno-misspelled-isr): Same.
1523 2018-06-11 Martin Sebor <msebor@redhat.com>
1525 * PR tree-optimization/86083
1526 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
1528 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
1530 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
1532 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
1535 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
1536 on the correct operand.
1537 (*movdi_internal64): Ditto.
1539 2018-06-11 Martin Liska <mliska@suse.cz>
1541 PR tree-optimization/86089
1542 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
1544 2018-06-11 Julia Koval <julia.koval@intel.com>
1546 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
1547 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
1548 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
1550 2018-06-11 Olivier Hainque <hainque@adacore.com>
1552 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
1553 for Ada with strict dwarf2.
1555 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
1558 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
1561 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
1563 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
1565 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
1567 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
1570 2018-06-08 Martin Liska <mliska@suse.cz>
1572 * tree-cfg.h (debug_function): Fix argument type to match
1575 2018-06-08 Martin Liska <mliska@suse.cz>
1577 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
1578 Remove usage of MPX-related (and removed) fields.
1579 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
1581 2018-06-08 David Malcolm <dmalcolm@redhat.com>
1583 * cfg.c (debug): Use TDF_NONE rather than 0.
1584 * cfghooks.c (debug): Likewise.
1585 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
1586 (struct dump_option_value_info): Convert to...
1587 (struct kv_pair): ...this template type.
1588 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
1590 (optinfo_verbosity_options): Likewise.
1591 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
1593 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
1594 than int for "optgroup_flags" param.
1595 (dump_generic_expr_loc): Use dump_flags_t rather than int for
1597 (dump_dec): Likewise.
1598 (dump_finish): Use TDF_NONE rather than 0.
1599 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
1600 rather than int for "optgroup_flags" param. Use TDF_NONE rather
1601 than 0. Update for change to option_ptr.
1602 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
1603 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
1604 0. Update for changes to optinfo_verbosity_options and
1606 (opt_info_switch_p): Convert optgroup_flags from int to
1608 (dump_basic_block): Use dump_flags_t rather than int
1609 for "dump_kind" param.
1610 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
1611 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
1612 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
1613 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
1614 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
1615 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
1616 TDF_NONE): Convert from macros to...
1617 (enum dump_flag): ...this new enum.
1618 (dump_flags_t): Update to use enum.
1619 (operator|, operator&, operator~, operator|=, operator&=):
1620 Implement for dump_flags_t.
1621 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
1622 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
1623 Convert from macros to...
1624 (enum optgroup_flag): ...this new enum.
1625 (optgroup_flags_t): New typedef.
1626 (operator|, operator|=): Implement for optgroup_flags_t.
1627 (struct dump_file_info): Convert field "alt_flags" to
1628 dump_flags_t. Convert field "optgroup_flags" to
1630 (dump_basic_block): Use dump_flags_t rather than int for param.
1631 (dump_generic_expr_loc): Likewise.
1632 (dump_dec): Likewise.
1633 (dump_register): Convert param "optgroup_flags" to
1635 (opt_info_enable_passes): Likewise.
1636 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
1638 * gimple-pretty-print.c (debug): Likewise.
1639 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
1640 (merged_store_group::apply_stores): Likewise.
1641 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
1642 * gimple.c (verify_gimple_pp): Likewise.
1643 * graphite-poly.c (print_pbb_body): Likewise.
1644 * passes.c (pass_manager::register_one_dump_file): Convert
1645 local "optgroup_flags" to optgroup_flags_t.
1646 * print-tree.c (print_node): Use TDF_NONE rather than 0.
1648 (debug_body): Likewise.
1649 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
1650 to optgroup_flags_t.
1651 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
1653 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
1654 (convert_mult_to_fma): Likewise.
1655 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
1656 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
1657 * tree-vect-data-refs.c (dump_lower_bound): Convert param
1658 "dump_kind" to dump_flags_t.
1660 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
1662 * config/rs6000/rs6000.c (min, max): Delete.
1664 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
1666 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
1667 -mabi=spe and -mabi=no-spe.
1669 2018-06-08 Martin Liska <mliska@suse.cz>
1671 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
1672 where we expect an existing summary.
1674 2018-06-08 Martin Liska <mliska@suse.cz>
1676 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
1677 * ipa-inline.h (estimate_edge_growth): Likewise.
1679 2018-06-08 Martin Liska <mliska@suse.cz>
1681 * cgraph.c (function_version_hasher::hash): Use
1682 cgraph_node::get_uid ().
1683 (function_version_hasher::equal):
1684 * cgraph.h (cgraph_node::get_uid): New method.
1685 * ipa-inline.c (update_caller_keys): Use
1686 cgraph_node::get_uid ().
1687 (update_callee_keys): Likewise.
1688 * ipa-utils.c (searchc): Likewise.
1689 (ipa_reduced_postorder): Likewise.
1690 * lto-cgraph.c (input_node): Likewise.
1691 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
1692 * symbol-summary.h (symtab_insertion): Likewise.
1693 (symtab_removal): Likewise.
1694 (symtab_duplication): Likewise.
1695 * tree-pretty-print.c (dump_function_header): Likewise.
1696 * tree-sra.c (convert_callers_for_node): Likewise.
1698 2018-06-08 Martin Liska <mliska@suse.cz>
1700 * cgraph.c (symbol_table::create_edge): Always assign a new
1702 (symbol_table::free_edge): Do not recycle numbers.
1703 * cgraph.h (cgraph_edge::get): New method.
1704 * symbol-summary.h (symtab_removal): Use it.
1705 (symtab_duplication): Likewise.
1706 (call_summary::hashable_uid): Remove.
1708 2018-06-08 Martin Liska <mliska@suse.cz>
1710 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
1711 (initialize_growth_caches): Remove.
1712 (free_growth_caches): Likewise.
1713 (do_estimate_edge_time): Use edge_growth_cache.
1714 (do_estimate_edge_size): Likewise.
1715 (do_estimate_edge_hints): Likewise.
1716 * ipa-inline.c (reset_edge_caches): Likewise.
1717 (recursive_inlining): Likewise.
1718 (inline_small_functions): Likewise.
1719 * ipa-inline.h (initialize_growth_caches): Remove.
1720 (estimate_edge_size): Likewise.
1721 (estimate_edge_time): Likewise.
1722 (estimate_edge_hints): Likewise.
1723 (reset_edge_growth_cache): Likewise.
1724 * symbol-summary.h (call_summary::remove): New method.
1726 2018-06-08 Martin Liska <mliska@suse.cz>
1728 * ipa-cp.c (class edge_clone_summary): New summary.
1729 (grow_edge_clone_vectors): Remove.
1730 (ipcp_edge_duplication_hook): Remove.
1731 (class edge_clone_summary_t): New call_summary class.
1732 (ipcp_edge_removal_hook): Remove.
1733 (edge_clone_summary_t::duplicate): New function.
1734 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
1735 (create_specialized_node): Likewise.
1736 (ipcp_driver): Initialize edge_clone_summaries and do not
1739 2018-06-08 Martin Liska <mliska@suse.cz>
1741 * symbol-summary.h (get): New function.
1742 (call_summary::m_initialize_when_cloning): New class member.
1744 2018-06-08 Martin Liska <mliska@suse.cz>
1746 * cgraph.c (cgraph_node::remove): Do not recycle uid.
1747 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
1748 (symbol_table::allocate_cgraph_symbol): Do not set uid.
1749 * passes.c (uid_hash_t): Record removed_nodes by their uids.
1750 (remove_cgraph_node_from_order): Use the removed_nodes set.
1751 (do_per_function_toporder): Likwise.
1752 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
1753 instead of summary_uid.
1754 (symtab_removal): Likewise.
1755 (symtab_duplication): Likewise.
1757 2018-06-08 Martin Liska <mliska@suse.cz>
1759 * ipa-cp.c (ipcp_store_bits_results): Use
1760 ipcp_transformation_sum.
1761 (ipcp_store_vr_results): Likewise.
1762 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
1764 (ipcp_transformation_initialize): ... this.
1765 (ipa_set_node_agg_value_chain):
1766 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
1767 (write_ipcp_transformation_info): Likewise.
1768 (read_ipcp_transformation_info): Likewise.
1769 (ipcp_update_bits): Likewise.
1770 (ipcp_update_vr): Likewise.
1771 (ipcp_transform_function): Likewise.
1772 * ipa-prop.h: Rename ipcp_transformation_summary to
1773 ipcp_transformation.
1774 (class ipcp_transformation_t): New function summary.
1775 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
1776 (ipa_get_agg_replacements_for_node): Likewise.
1778 2018-06-08 Martin Liska <mliska@suse.cz>
1780 * ipa-pure-const.c (struct funct_state_d): Do it class instead
1782 (class funct_state_summary_t): New function_summary class.
1783 (has_function_state): Remove.
1784 (get_function_state): Likewise.
1785 (set_function_state): Likewise.
1786 (add_new_function): Likewise.
1787 (funct_state_summary_t::insert): New function.
1788 (duplicate_node_data): Remove.
1789 (remove_node_data): Remove.
1790 (funct_state_summary_t::duplicate): New function.
1791 (register_hooks): Create new funct_state_summaries.
1792 (pure_const_generate_summary): Use it.
1793 (pure_const_write_summary): Likewise.
1794 (pure_const_read_summary): Likewise.
1795 (propagate_pure_const): Likewise.
1796 (propagate_nothrow): Likewise.
1797 (dump_malloc_lattice): Likewise.
1798 (propagate_malloc): Likewise.
1799 (execute): Do not register hooks, just remove summary
1801 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
1804 2018-06-08 Martin Liska <mliska@suse.cz>
1806 * ipa-reference.c (remove_node_data): Remove.
1807 (duplicate_node_data): Likewise.
1808 (class ipa_ref_var_info_summary_t): New class.
1809 (class ipa_ref_opt_summary_t): Likewise.
1810 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
1811 (get_reference_optimization_summary): Use
1812 ipa_ref_opt_sum_summaries.
1813 (set_reference_vars_info): Remove.
1814 (set_reference_optimization_summary): Likewise.
1815 (ipa_init): Create summaries.
1816 (init_function_info): Use function summary.
1817 (ipa_ref_opt_summary_t::duplicate): New function.
1818 (ipa_ref_opt_summary_t::remove): New function.
1819 (get_read_write_all_from_node): Fix GNU coding style.
1820 (propagate): Use function summary.
1821 (write_node_summary_p): Fix GNU coding style.
1822 (stream_out_bitmap): Likewise.
1823 (ipa_reference_read_optimization_summary): Use function summary.
1824 (ipa_reference_c_finalize): Do not release hooks.
1826 2018-06-08 Martin Liska <mliska@suse.cz>
1828 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
1829 (analyze_function_body): Extract multiple calls of get_create.
1830 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
1831 * ipa-inline.c (recursive_inlining): Use ::get method.
1832 * ipa-inline.h (estimate_edge_growth): Likewise.
1834 2018-06-08 Martin Liska <mliska@suse.cz>
1836 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
1838 (hsa_function_summary::hsa_function_summary): Use the new enum
1840 (hsa_gpu_implementation_p): Use hsa_summaries::get.
1841 * hsa-gen.c (hsa_get_host_function): Likewise.
1842 (get_brig_function_name): Likewise.
1843 * ipa-hsa.c (process_hsa_functions): Likewise.
1844 (ipa_hsa_write_summary): Likewise.
1845 * symbol-summary.h (symtab_duplication): Use ::get function/
1846 (get): New function.
1848 2018-06-08 Martin Liska <mliska@suse.cz>
1850 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
1852 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
1853 (hsa_register_kernel): Likewise.
1854 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
1855 * hsa-gen.c (hsa_get_host_function): Likewise.
1856 (get_brig_function_name): Likewise.
1857 (generate_hsa): Likewise.
1858 (pass_gen_hsail::execute): Likewise.
1859 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
1860 (devirtualization_time_bonus): Likewise.
1861 (ipcp_propagate_stage): Likewise.
1862 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
1863 (edge_set_predicate): Likewise.
1864 (evaluate_conditions_for_known_args): Likewise.
1865 (evaluate_properties_for_edge): Likewise.
1866 (ipa_fn_summary::reset): Likewise.
1867 (ipa_fn_summary_t::duplicate): Likewise.
1868 (dump_ipa_call_summary): Likewise.
1869 (ipa_dump_fn_summary): Likewise.
1870 (analyze_function_body): Likewise.
1871 (compute_fn_summary): Likewise.
1872 (estimate_edge_devirt_benefit): Likewise.
1873 (estimate_edge_size_and_time): Likewise.
1874 (estimate_calls_size_and_time): Likewise.
1875 (estimate_node_size_and_time): Likewise.
1876 (inline_update_callee_summaries): Likewise.
1877 (remap_edge_change_prob): Likewise.
1878 (remap_edge_summaries): Likewise.
1879 (ipa_merge_fn_summary_after_inlining): Likewise.
1880 (ipa_update_overall_fn_summary): Likewise.
1881 (read_ipa_call_summary): Likewise.
1882 (inline_read_section): Likewise.
1883 (write_ipa_call_summary): Likewise.
1884 (ipa_fn_summary_write): Likewise.
1885 (ipa_free_fn_summary): Likewise.
1886 * ipa-hsa.c (process_hsa_functions): Likewise.
1887 (ipa_hsa_write_summary): Likewise.
1888 (ipa_hsa_read_section): Likewise.
1889 * ipa-icf.c (sem_function::merge): Likewise.
1890 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
1891 (do_estimate_edge_time): Likewise.
1892 (estimate_size_after_inlining): Likewise.
1893 (estimate_growth): Likewise.
1894 (growth_likely_positive): Likewise.
1895 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
1896 (inline_call): Likewise.
1897 * ipa-inline.c (caller_growth_limits): Likewise.
1898 (can_inline_edge_p): Likewise.
1899 (can_inline_edge_by_limits_p): Likewise.
1900 (compute_uninlined_call_time): Likewise.
1901 (compute_inlined_call_time): Likewise.
1902 (want_inline_small_function_p): Likewise.
1903 (edge_badness): Likewise.
1904 (update_caller_keys): Likewise.
1905 (update_callee_keys): Likewise.
1906 (recursive_inlining): Likewise.
1907 (inline_small_functions): Likewise.
1908 (inline_to_all_callers_1): Likewise.
1909 (dump_overall_stats): Likewise.
1910 (early_inline_small_functions): Likewise.
1911 (early_inliner): Likewise.
1912 * ipa-inline.h (estimate_edge_growth): Likewise.
1913 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
1914 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
1915 * ipa-prop.h (IPA_NODE_REF): Likewise.
1916 (IPA_EDGE_REF): Likewise.
1917 * ipa-pure-const.c (malloc_candidate_p): Likewise.
1918 (propagate_malloc): Likewise.
1919 * ipa-split.c (execute_split_functions): Likewise.
1920 * symbol-summary.h: Rename get to get_create.
1922 (get_create): Likewise.
1923 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
1925 2018-06-08 Martin Liska <mliska@suse.cz>
1927 * symbol-summary.h (release): Move definition out of class
1929 (symtab_removal): Likewise.
1930 (symtab_duplication): Likewise.
1932 2018-06-08 Martin Liska <mliska@suse.cz>
1934 * symbol-summary.h (function_summary): Move constructor
1935 implementation out of class declaration.
1936 (release): Likewise.
1937 (symtab_insertion): Likewise.
1938 (symtab_removal): Likewise.
1939 (symtab_duplication): Likewise.
1942 2018-06-08 Martin Liska <mliska@suse.cz>
1944 * Makefile.in: Remove support for MPX (macros, related functions,
1945 fields in cgraph_node, ...).
1946 * builtin-types.def (BT_BND): Likewise.
1947 (BT_FN_BND_CONST_PTR): Likewise.
1948 (BT_FN_CONST_PTR_BND): Likewise.
1949 (BT_FN_VOID_PTR_BND): Likewise.
1950 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
1951 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
1952 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
1953 (expand_builtin_mempcpy_with_bounds): Likewise.
1954 (expand_builtin_memset_with_bounds): Likewise.
1955 (expand_builtin_memset_args): Likewise.
1956 (std_expand_builtin_va_start): Likewise.
1957 (expand_builtin): Likewise.
1958 (expand_builtin_with_bounds): Likewise.
1959 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
1960 (DEF_LIB_BUILTIN_CHKP): Likewise.
1961 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
1962 (DEF_CHKP_BUILTIN): Likewise.
1963 (BUILT_IN_MEMCPY): Likewise.
1964 (BUILT_IN_MEMMOVE): Likewise.
1965 (BUILT_IN_MEMPCPY): Likewise.
1966 (BUILT_IN_MEMSET): Likewise.
1967 (BUILT_IN_STPCPY): Likewise.
1968 (BUILT_IN_STRCAT): Likewise.
1969 (BUILT_IN_STRCHR): Likewise.
1970 (BUILT_IN_STRCPY): Likewise.
1971 (BUILT_IN_STRLEN): Likewise.
1972 (BUILT_IN_MEMCPY_CHK): Likewise.
1973 (BUILT_IN_MEMMOVE_CHK): Likewise.
1974 (BUILT_IN_MEMPCPY_CHK): Likewise.
1975 (BUILT_IN_MEMSET_CHK): Likewise.
1976 (BUILT_IN_STPCPY_CHK): Likewise.
1977 (BUILT_IN_STRCAT_CHK): Likewise.
1978 (BUILT_IN_STRCPY_CHK): Likewise.
1979 * calls.c (store_bounds): Likewise.
1980 (emit_call_1): Likewise.
1981 (special_function_p): Likewise.
1982 (maybe_warn_nonstring_arg): Likewise.
1983 (initialize_argument_information): Likewise.
1984 (finalize_must_preallocate): Likewise.
1985 (compute_argument_addresses): Likewise.
1986 (expand_call): Likewise.
1987 * cfgexpand.c (expand_call_stmt): Likewise.
1988 (expand_return): Likewise.
1989 (expand_gimple_stmt_1): Likewise.
1990 (pass_expand::execute): Likewise.
1991 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
1992 (cgraph_node::remove): Likewise.
1993 (cgraph_node::dump): Likewise.
1994 (cgraph_node::verify_node): Likewise.
1995 * cgraph.h (chkp_function_instrumented_p): Likewise.
1996 (symtab_node::get_alias_target): Likewise.
1997 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
1998 (cgraph_local_p): Likewise.
1999 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
2000 (cgraph_edge::rebuild_references): Likewise.
2001 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
2002 (walk_polymorphic_call_targets): Likewise.
2003 (cgraph_node::expand_thunk): Likewise.
2004 (symbol_table::output_weakrefs): Likewise.
2005 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
2006 (ix86_handle_option): Likewise.
2007 * config/i386/constraints.md: Likewise.
2008 * config/i386/i386-builtin-types.def (BND): Likewise.
2012 * config/i386/i386-builtin.def (BDESC_END): Likewise.
2013 (BDESC_FIRST): Likewise.
2015 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
2016 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
2017 * config/i386/i386.c (enum reg_class): Likewise.
2018 (ix86_target_string): Likewise.
2019 (ix86_option_override_internal): Likewise.
2020 (ix86_conditional_register_usage): Likewise.
2021 (ix86_valid_target_attribute_inner_p): Likewise.
2022 (ix86_set_indirect_branch_type): Likewise.
2023 (ix86_set_current_function): Likewise.
2024 (ix86_function_arg_regno_p): Likewise.
2025 (init_cumulative_args): Likewise.
2026 (ix86_function_arg_advance): Likewise.
2027 (ix86_function_arg): Likewise.
2028 (ix86_pass_by_reference): Likewise.
2029 (ix86_function_value_regno_p): Likewise.
2030 (ix86_function_value_1): Likewise.
2031 (ix86_function_value_bounds): Likewise.
2032 (ix86_return_in_memory): Likewise.
2033 (ix86_setup_incoming_vararg_bounds): Likewise.
2034 (ix86_va_start): Likewise.
2035 (indirect_thunk_need_prefix): Likewise.
2036 (print_reg): Likewise.
2037 (ix86_print_operand): Likewise.
2038 (ix86_expand_call): Likewise.
2039 (ix86_output_function_return): Likewise.
2040 (reg_encoded_number): Likewise.
2041 (BDESC_VERIFYS): Likewise.
2042 (ix86_init_mpx_builtins): Likewise.
2043 (ix86_init_builtins): Likewise.
2044 (ix86_emit_cmove): Likewise.
2045 (ix86_emit_move_max): Likewise.
2046 (ix86_expand_builtin): Likewise.
2047 (ix86_builtin_mpx_function): Likewise.
2048 (ix86_get_arg_address_for_bt): Likewise.
2049 (ix86_load_bounds): Likewise.
2050 (ix86_store_bounds): Likewise.
2051 (ix86_load_returned_bounds): Likewise.
2052 (ix86_store_returned_bounds): Likewise.
2053 (ix86_class_likely_spilled_p): Likewise.
2054 (ix86_hard_regno_mode_ok): Likewise.
2055 (x86_order_regs_for_local_alloc): Likewise.
2056 (ix86_mitigate_rop): Likewise.
2057 (ix86_bnd_prefixed_insn_p): Likewise.
2058 (ix86_mpx_bound_mode): Likewise.
2059 (ix86_make_bounds_constant): Likewise.
2060 (ix86_initialize_bounds): Likewise.
2061 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
2062 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
2063 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
2064 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
2065 (TARGET_CHKP_BOUND_MODE): Likewise.
2066 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
2067 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
2068 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
2069 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
2070 * config/i386/i386.h (TARGET_MPX): Likewise.
2071 (TARGET_MPX_P): Likewise.
2072 (VALID_BND_REG_MODE): Likewise.
2073 (FIRST_BND_REG): Likewise.
2074 (LAST_BND_REG): Likewise.
2075 (enum reg_class): Likewise.
2076 (BND_REG_P): Likewise.
2077 (BND_REGNO_P): Likewise.
2078 (BNDmode): Likewise.
2079 (ADJUST_INSN_LENGTH): Likewise.
2080 * config/i386/i386.md: Likewise.
2081 * config/i386/i386.opt: Likewise.
2082 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
2083 (defined): Likewise.
2084 (LINK_MPX): Likewise.
2085 (MPX_SPEC): Likewise.
2086 (LIBMPX_SPEC): Likewise.
2087 (LIBMPXWRAPPERS_SPEC): Likewise.
2088 (CHKP_SPEC): Likewise.
2089 * config/i386/predicates.md: Likewise.
2090 * dbxout.c (dbxout_type): Likewise.
2091 * doc/extend.texi: Likewise.
2092 * doc/invoke.texi: Likewise.
2093 * doc/md.texi: Likewise.
2094 * doc/tm.texi: Likewise.
2095 * doc/tm.texi.in: Likewise.
2096 * dwarf2out.c (is_base_type): Likewise.
2097 (gen_formal_types_die): Likewise.
2098 (gen_subprogram_die): Likewise.
2099 (gen_type_die_with_usage): Likewise.
2100 (gen_decl_die): Likewise.
2101 (dwarf2out_late_global_decl): Likewise.
2102 * expr.c (expand_assignment): Likewise.
2103 (emit_storent_insn): Likewise.
2104 (store_expr_with_bounds): Likewise.
2105 (store_expr): Likewise.
2106 (expand_expr_real_1): Likewise.
2107 * expr.h (store_expr_with_bounds): Likewise.
2108 * function.c (use_register_for_decl): Likewise.
2109 (struct bounds_parm_data): Likewise.
2110 (assign_parms_augmented_arg_list): Likewise.
2111 (assign_parm_find_entry_rtl): Likewise.
2112 (assign_parm_is_stack_parm): Likewise.
2113 (assign_parm_load_bounds): Likewise.
2114 (assign_bounds): Likewise.
2115 (assign_parms): Likewise.
2116 (expand_function_start): Likewise.
2117 * gcc.c (CHKP_SPEC): Likewise.
2118 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
2119 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
2120 (wrestrict_dom_walker::check_call): Likewise.
2121 * gimple.c (gimple_build_call_from_tree): Likewise.
2122 * gimple.h (enum gf_mask): Likewise.
2123 (gimple_call_with_bounds_p): Likewise.
2124 (gimple_call_set_with_bounds): Likewise.
2125 * gimplify.c (gimplify_init_constructor): Likewise.
2126 * ipa-cp.c (initialize_node_lattices): Likewise.
2127 (propagate_constants_across_call): Likewise.
2128 (find_more_scalar_values_for_callers_subset): Likewise.
2129 * ipa-hsa.c (process_hsa_functions): Likewise.
2130 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
2131 * ipa-icf.c (sem_function::merge): Likewise.
2132 * ipa-inline.c (early_inliner): Likewise.
2133 * ipa-pure-const.c (warn_function_noreturn): Likewise.
2134 (warn_function_cold): Likewise.
2135 (propagate_pure_const): Likewise.
2136 * ipa-ref.h (enum GTY): Likewise.
2137 * ipa-split.c (find_retbnd): Likewise.
2138 (consider_split): Likewise.
2139 (split_function): Likewise.
2140 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
2141 * ipa.c (walk_polymorphic_call_targets): Likewise.
2142 (symbol_table::remove_unreachable_nodes): Likewise.
2143 (process_references): Likewise.
2144 (cgraph_build_static_cdtor_1): Likewise.
2145 * lto-cgraph.c (lto_output_node): Likewise.
2146 (output_refs): Likewise.
2147 (compute_ltrans_boundary): Likewise.
2148 (input_overwrite_node): Likewise.
2149 (input_node): Likewise.
2150 (input_cgraph_1): Likewise.
2151 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
2152 * passes.c (pass_manager::execute_early_local_passes): Likewise.
2153 (class pass_chkp_instrumentation_passes): Likewise.
2154 (make_pass_chkp_instrumentation_passes): Likewise.
2155 * passes.def: Likewise.
2156 * rtl.h (struct GTY): Likewise.
2157 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
2158 * stor-layout.c (layout_type): Likewise.
2159 * symtab.c: Likewise.
2160 * target.def: Likewise.
2161 * targhooks.c (default_chkp_bound_type): Likewise.
2162 (default_chkp_bound_mode): Likewise.
2163 (default_builtin_chkp_function): Likewise.
2164 (default_chkp_function_value_bounds): Likewise.
2165 (default_chkp_make_bounds_constant): Likewise.
2166 (default_chkp_initialize_bounds): Likewise.
2167 * targhooks.h (default_chkp_bound_type): Likewise.
2168 (default_chkp_bound_mode): Likewise.
2169 (default_builtin_chkp_function): Likewise.
2170 (default_chkp_function_value_bounds): Likewise.
2171 (default_chkp_make_bounds_constant): Likewise.
2172 (default_chkp_initialize_bounds): Likewise.
2173 * toplev.c (compile_file): Likewise.
2174 (process_options): Likewise.
2175 * tree-core.h (DEF_BUILTIN): Likewise.
2176 (DEF_BUILTIN_CHKP): Likewise.
2177 * tree-inline.c (declare_return_variable): Likewise.
2178 (remap_gimple_stmt): Likewise.
2179 (copy_bb): Likewise.
2180 (initialize_inlined_parameters): Likewise.
2181 (expand_call_inline): Likewise.
2182 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
2183 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
2184 (make_pass_ipa_chkp_produce_thunks): Likewise.
2185 (make_pass_chkp): Likewise.
2186 (make_pass_chkp_opt): Likewise.
2187 (make_pass_chkp_instrumentation_passes): Likewise.
2188 * tree-pretty-print.c (dump_generic_node): Likewise.
2189 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
2190 * tree-ssa-dce.c (propagate_necessity): Likewise.
2191 (eliminate_unnecessary_stmts): Likewise.
2192 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
2193 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
2194 * tree-ssa-sccvn.h: Likewise.
2195 * tree-ssa-strlen.c (get_string_length): Likewise.
2196 (valid_builtin_call): Likewise.
2197 (adjust_last_stmt): Likewise.
2198 (handle_builtin_strchr): Likewise.
2199 (handle_builtin_strcpy): Likewise.
2200 (handle_builtin_stxncpy): Likewise.
2201 (handle_builtin_memcpy): Likewise.
2202 (handle_builtin_strcat): Likewise.
2203 (strlen_check_and_optimize_stmt): Likewise.
2204 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
2205 * tree-streamer-in.c: Likewise.
2206 * tree-streamer.c (record_common_node): Likewise.
2207 * tree.c (tree_code_size): Likewise.
2208 (wide_int_to_tree_1): Likewise.
2209 (type_contains_placeholder_1): Likewise.
2210 (build_common_tree_nodes): Likewise.
2211 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
2212 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
2213 (POINTER_BOUNDS_P): Likewise.
2214 (BOUNDED_TYPE_P): Likewise.
2215 (BOUNDED_P): Likewise.
2216 (CALL_WITH_BOUNDS_P): Likewise.
2217 (pointer_bounds_type_node): Likewise.
2218 * value-prof.c (gimple_ic): Likewise.
2219 * var-tracking.c (vt_add_function_parameters): Likewise.
2220 * varasm.c (make_decl_rtl): Likewise.
2221 (assemble_start_function): Likewise.
2222 (output_constant): Likewise.
2223 (maybe_assemble_visibility): Likewise.
2224 * varpool.c (ctor_for_folding): Likewise.
2225 * chkp-builtins.def: Remove.
2226 * ipa-chkp.c: Remove.
2227 * ipa-chkp.h: Remove.
2228 * rtl-chkp.c: Remove.
2229 * rtl-chkp.h: Remove.
2230 * tree-chkp-opt.c: Remove.
2231 * tree-chkp.c: Remove.
2232 * tree-chkp.h: Remove.
2234 2018-06-07 Carl Love <cel@us.ibm.com>
2236 * gcc/config/rs6000/vsx.md (vextract_fp_from_shorth,
2237 vextract_fp_from_shortl): Add BE support.
2239 2018-06-07 Paul Koning <ni1d@arrl.net>
2241 * compare-elim.c (try_merge_compare): Don't merge compare if
2242 address contains a side effect.
2243 (try_eliminate_compare): Likewise.
2245 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
2247 * config.gcc: Support "tremont".
2248 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
2249 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2251 * config/i386/i386.c (m_TREMONT): Define.
2252 (processor_target_table): Add "tremont".
2253 (PTA_TREMONT): Define.
2254 (ix86_lea_outperforms): Add TARGET_TREMONT.
2255 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
2256 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
2257 and M_INTEL_GOLDMONT_PLUS.
2258 (fold_builtin_cpu): Add "tremont".
2259 (ix86_add_stmt_cost): Add TARGET_TREMONT.
2260 (ix86_option_override_internal): Add "tremont".
2261 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
2262 (processor_type): Add PROCESSOR_TREMONT.
2263 * config/i386/x86-tune.def: Add m_TREMONT.
2264 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
2266 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
2268 * gcc/config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
2269 symbol defined for msp430i* devices to be lower case.
2271 gcc/testsuite/gcc.target/msp430/
2272 * msp430i-device-symbol.c: New test.
2273 * msp430f-device-symbol.c: New test.
2274 * msp430.h: New test header file.
2276 2018-06-07 Richard Biener <rguenther@suse.de>
2278 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
2279 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
2280 Properly wrap signed arithmetic if overflow wraps.
2282 2018-06-07 Jakub Jelinek <jakub@redhat.com>
2284 PR tree-optimization/69615
2285 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
2286 of a cast from a same precision integral SSA_NAME in a bb dominated
2287 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
2288 cast to utype if rhs2 has already a compatible type.
2290 2018-06-07 Richard Biener <rguenther@suse.de>
2292 PR tree-optimization/85935
2293 * graphite-scop-detection.c (find_params_in_bb): Analyze
2294 condition operands with respect to the correct loop. Assert
2295 the analysis doesn't fail.
2297 2018-06-04 Carl Love <cel@us.ibm.com>
2299 * gcc/config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
2300 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
2301 as it is slightly cheaper.
2302 (first_match_or_eos_index_<mode>):
2303 Calculate index using natural element order.
2304 (first_match_index_<mode>):
2305 Calculate index using natural element order.
2306 (first_match_or_eos_index_<mode>):
2307 Calculate index using natural order.
2308 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
2309 for BE and LE modes.
2310 * gcc/config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
2311 P9V_BUILTIN_VCLZLSBB_V16QI.
2312 * gcc/config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
2315 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
2317 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
2318 indentation and line wrap for many prototypes. Add missing
2319 @smallexample directives around block of prototypes for vec_xl and
2322 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
2324 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
2325 track if we pass or return IEEE 128-bit floating point.
2326 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
2327 C++ mangling that is compatible with GCC 8.1.
2328 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
2329 (init_cumulative_args): Note if we pass or return IEEE 128-bit
2330 floating point types.
2331 (rs6000_function_arg_advance_1): Likewise.
2332 (rs6000_mangle_type): Optionally generate mangled names that match
2333 what GCC 8.1 generated for IEEE 128-bit floating point types.
2334 (rs6000_globalize_decl_name): If we have an external function that
2335 passes or returns IEEE 128-bit types, generate a weak reference
2336 from the mangled name used in GCC 8.1 to the current mangled
2338 (rs6000_init_builtins): Make __ibm128 use the long double type if
2339 long double is IBM extended double. Make __float128 use the long
2340 double type if long double is IEEE 128-bit.
2343 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
2344 macro for __ibm128 built-in functions.
2345 (PACK_IF): Add __ibm128 pack/unpack functions.
2346 (UNPACK_IF): Likewise.
2347 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
2348 enable long double built-in functions if long double is IEEE
2349 128-bit floating point.
2350 (rs6000_invalid_builtin): Update long double built-in function
2352 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
2353 functions, adjust the built-in function to use the long double
2354 built-in function if __ibm128 and long double are the same type.
2355 * doc/extend.texi (PowerPC builtins): Update documention for
2356 __builtin_{,un}pack_longdouble. Add documentation for
2357 __builtin_{,un}pack_ibm128.
2359 2018-06-06 Jim Wilson <jimw@sifive.com>
2361 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
2362 (struct machine_function): New field interrupt_mode.
2363 (riscv_handle_type_attribute): New function. Add forward declaration.
2364 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
2365 (riscv_expand_epilogue): Check interrupt_mode field.
2366 (riscv_set_current_function): Check interrupt attribute args and
2367 set interrupt_mode field.
2368 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
2369 (riscv_sret, riscv_uret): New.
2370 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
2371 new arguments to interrupt attribute.
2373 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
2376 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
2377 Don't handle -mcpu=power8 if -mpower9-vector is also used.
2379 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
2381 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
2382 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
2383 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
2384 several redundant entries.
2386 2018-06-06 David Malcolm <dmalcolm@redhat.com>
2388 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
2389 type from "rtx" to "rtx_insn *".
2390 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
2391 for local "call_insn", removing cast.
2392 (ix86_expand_call): Likewise, introducing a "call_insn" local.
2394 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
2396 PR tree-optimization/86066
2397 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
2398 for BIT_INSERT_EXPR stores.
2400 2018-06-06 Richard Biener <rguenther@suse.de>
2402 PR tree-optimization/86062
2403 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
2404 component refs ontop
2405 of to be offsetted base.
2407 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
2409 * gcc/config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
2410 to be static and remove check on interrupt attribute name.
2412 gcc/testsuite/gcc.target/msp430/
2413 * function-attributes-4.c: New test.
2414 * static-interrupts.c: New test.
2416 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
2418 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
2419 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
2421 2018-06-05 Steve Ellcey <sellcey@cavium.com>
2424 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
2426 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
2427 Remove second argument, change how error is called.
2428 (aarch64_layout_arg): Remove second argument from
2429 aarch64_err_no_fpadvsimd call.
2430 (aarch64_init_cumulative_args): Ditto.
2431 (aarch64_gimplify_va_arg_expr): Ditto.
2432 * config/aarch64/aarch64.md (mov<mode>): Ditto.
2434 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
2436 * config/i386/i386.md (simple_return_indirect_internal): New expander.
2437 (*simple_return_indirect_internal<mode>): Rename from
2438 simple_return_indirect_internal. Use W mode iterator.
2439 (rstorssp): New expander.
2440 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
2441 (clrssbsy): New expander.
2442 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
2444 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
2446 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
2448 (cmse_check_pointed_object): Likewise.
2450 2018-06-05 Martin Liska <mliska@suse.cz>
2452 PR gcov-profile/47618
2453 * doc/invoke.texi: Document how -fprofile-dir format
2456 2018-06-05 Richard Biener <rguenther@suse.de>
2458 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
2459 removal pretend DOM info isn't available so we do not update
2460 it and only remove edges, not dominated blocks. Actually free
2461 DOM info in case we removed something. Remove unreachable blocks.
2462 (mfb_keep_latches): Work with either DOM info or marked backedges.
2463 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
2464 first. Mark backedges if DOM info isn't available.
2465 (Re-)compute DOM info after cleanup_control_flow_pre.
2467 2018-06-05 Richard Biener <rguenther@suse.de>
2469 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
2470 (locus_discrim_hasher::hash): Adjust.
2471 (locus_discrim_hasher::equal): Likewise.
2472 (next_discriminator_for_locus): Work on line directly.
2473 (same_line_p): Pass in expanded locus1 as well.
2474 (assign_discriminators): Avoid redundant location expansions.
2476 2018-06-05 Richard Biener <rguenther@suse.de>
2478 PR tree-optimization/86046
2479 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
2480 if required after clearing TREE_ADDRESSABLE.
2482 2018-06-05 Richard Biener <rguenther@suse.de>
2484 PR tree-optimization/86047
2485 * tree-ssa-loop.c (for_each_index): Glob handling of all
2486 decls and constants and really handle all of them.
2488 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2491 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
2492 qualifier_void_pointer and qualifier_const_void_pointer.
2493 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
2494 (arm_init_builtins): Handle the above.
2495 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
2496 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
2497 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
2500 2018-06-05 Martin Liska <mliska@suse.cz>
2502 * auto-profile.c (read_autofdo_file): Do not use
2503 gcov_ctr_summary struct.
2504 (afdo_callsite_hot_enough_for_early_inline): Likewise.
2505 * coverage.c (struct counts_entry): Likewise.
2506 (read_counts_file): Read just single summary entry.
2507 (get_coverage_counts): Use gcov_summary struct.
2508 * coverage.h (get_coverage_counts): Likewise.
2509 * gcov-dump.c (dump_working_sets): Likewise.
2510 (tag_summary): Dump just single summary.
2511 * gcov-io.c (gcov_write_summary): Write just histogram
2513 (gcov_read_summary): Read just single summary.
2514 (compute_working_sets): Use gcov_summary struct.
2515 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
2516 of GCOV_COUNTERS_SUMMABLE.
2517 (GCOV_COUNTERS_SUMMABLE): Remove.
2518 (GCOV_FIRST_VALUE_COUNTER): Replace with
2519 GCOV_COUNTER_V_INTERVAL.
2520 (struct gcov_ctr_summary): Remove.
2521 (struct gcov_summary): Directly use fields of former
2523 (compute_working_sets): Use gcov_summary struct.
2524 * gcov.c (read_count_file): Do not use ctrs fields.
2525 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
2527 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
2529 * profile.c: Likewise.
2530 * profile.h: Likewise.
2532 2018-06-05 Martin Liska <mliska@suse.cz>
2534 PR gcov-profile/84846
2535 * gcov.c (output_lines): Print working directory only
2536 in intermediate format.
2538 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
2540 * config/s390/s390-builtin-types.def: Add void function type.
2541 * config/s390/s390-builtins.def: Use the function type for the
2544 2018-06-04 Jim Wilson <jimw@sifive.com>
2546 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
2548 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
2549 and maybe_eh_return. Change regno to unsigned int. Use new args to
2550 handle EH_RETURN_DATA_REGNO registers properly.
2551 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
2552 (riscv_expand_epilogue): Update comment. Change argument name and
2553 type. Update code to use new name and type. Pass new args to
2554 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
2556 * config/riscv/riscv.md (NORMAL_RETURN): New.
2557 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
2558 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
2559 (eh_return): Call gen_eh_return_internal and emit barrier.
2560 (eh_return_internal): Call riscv_expand_epilogue.
2562 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
2564 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
2565 bit_insertion field and declare can_be_merged_into method.
2566 (merged_store_group::can_be_merged_into): New method.
2567 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
2568 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
2569 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
2571 2018-06-04 Richard Biener <rguenther@suse.de>
2573 PR tree-optimization/85955
2574 * builtins.c (fold_builtin_sincos): Convert pointers to
2575 destination to appropriate type before dereferencing.
2577 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
2579 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
2581 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
2583 * expr.c (expand_expr_real_1): Force the operand into memory if
2584 its TYPE_MODE is BLKmode and if there is no integer mode for
2585 the number of bits being extracted.
2587 2018-06-04 Jakub Jelinek <jakub@redhat.com>
2591 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
2592 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
2594 2018-06-04 Richard Biener <rguenther@suse.de>
2596 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
2597 (cleanup_tree_cfg_noloop): ... single caller. Do
2598 start_recording_case_labels later.
2600 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
2602 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
2603 to _IMMINTRIN_H_INCLUDED.
2604 * config/i386/pconfigintrin.h: Ditto.
2605 * config/i386/waitpkgintrin.h: Ditto.
2606 * config/i386/immintrin.h: Add includes for sgxintrin.h,
2607 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
2608 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
2609 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
2610 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
2611 waitpkgintrin.h and cldemoteintrin.h.
2613 2018-06-04 Richard Biener <rguenther@suse.de>
2615 PR tree-optimization/86038
2616 * tracer.c (find_best_successor): Check probability for
2617 being initialized, bail out if not.
2619 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
2622 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
2623 of bits to ignore when comparing architectures.
2625 2018-06-04 Jakub Jelinek <jakub@redhat.com>
2627 PR tree-optimization/69615
2628 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
2629 maximum or minimum of the type, try to merge it also as if
2630 range1 is + [-, x - 1] or + [x + 1, -].
2633 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
2635 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
2637 PR tree-optimization/86034
2638 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
2639 the unsigned bitfield type in a bit insertion sequence if it does not
2640 have a larger precision than the bitfield size.
2641 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
2643 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
2645 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
2647 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
2649 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
2650 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
2651 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
2652 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
2654 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
2656 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
2657 Disable -fdelete-null-pointer-checks for ELF toolchain.
2659 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
2660 Kito Cheng <kito.cheng@gmail.com>
2662 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
2663 (nds32le-*-*, nds32be-*-*): Integrate checking process.
2664 (nds32*-*-*): Add glibc and uclibc conditions.
2665 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
2666 (TARGET_EXCEPT_UNWIND_INFO): Define.
2667 * config/nds32/elf.h: New file.
2668 * config/nds32/linux.h: New file.
2669 * config/nds32/nds32-elf.opt: New file.
2670 * config/nds32/nds32-linux.opt: New file.
2671 * config/nds32/nds32-fp-as-gp.c
2672 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
2673 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
2675 (nds32_asm_file_end): Ditto.
2676 (nds32_print_operand): Ditto.
2677 (nds32_insert_attributes): Ditto.
2678 (nds32_init_libfuncs): New function.
2679 (TARGET_HAVE_TLS): Define.
2680 (TARGET_INIT_LIBFUNCS): Define.
2681 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
2683 (TARGET_ELF): Apply different mcmodel setting.
2684 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
2685 been migrated into elf.h and linux.h files.
2686 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
2687 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
2688 (mcmodel): The content has been migrated into nds32-elf.opt and
2689 nds32-linux.opt files.
2690 * config/nds32/t-elf: New file.
2691 * config/nds32/t-linux: New file.
2693 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
2694 Shiva Chen <shiva0217@gmail.com>
2696 * config/nds32/constants.md (unspec_volatile_element): Add
2697 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
2698 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
2700 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
2701 (make_pass_nds32_fp_as_gp): Declare.
2702 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
2704 (nds32_asm_function_end_prologue): Remove unused asm output.
2705 (nds32_asm_function_begin_epilogue): Remove unused asm output.
2706 (nds32_asm_file_start): Output necessary fp_as_gp information.
2707 (nds32_option_override): Adjust register usage.
2708 (nds32_expand_prologue): Consider fp_as_gp situation.
2709 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
2710 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
2713 (simple_return): Ditto.
2714 (omit_fp_begin): Output special directive for fp_as_gp.
2715 (omit_fp_end): Output special directive for fp_as_gp.
2716 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
2717 mforbid-fp-as-gp): New options.
2719 2018-06-01 Mark Wielaard <mark@klomp.org>
2721 * dwarf2out.c (dwarf2out_finish): Remove generation of
2722 DW_AT_loclists_base.
2724 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
2726 * gimple-ssa-store-merging.c: Include gimple-fold.h.
2727 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
2728 (struct merged_store_group): Add bit_insertion field.
2729 (dump_char_array): Use standard hexadecimal format.
2730 (merged_store_group::merged_store_group): Set bit_insertion to false.
2731 (merged_store_group::apply_stores): Use optimal buffer size. Deal
2732 with BIT_INSERT_EXPR stores. Move up code updating the mask and
2733 also print the mask in the dump file.
2734 (pass_store_merging::gate): Minor tweak.
2735 (imm_store_chain_info::coalesce_immediate): Fix wrong association
2736 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
2737 stores with INTEGER_CST stores.
2738 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
2739 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
2740 and use it throughout. Generate bit insertion sequences if need be.
2741 (pass_store_merging::process_store): Remove redundant condition.
2742 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
2744 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
2746 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
2747 the 128-bit floating point types. Fix function comment.
2749 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2751 * config/aarch64/aarch64-simd.md
2752 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
2754 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
2757 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
2759 PR tree-optimization/85989
2760 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
2762 (backprop::intersect_uses): Check it when deciding whether this
2763 is a backedge reference.
2764 (backprop::process_block): Add each phi to m_visited_phis
2765 after visiting it, then clear it at the end.
2767 2018-06-01 Richard Biener <rguenther@suse.de>
2769 * tree-vectorizer.h (vect_dr_stmt): New function.
2770 (vect_get_load_cost): Adjust.
2771 (vect_get_store_cost): Likewise.
2772 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
2773 Use vect_dr_stmt instead of DR_SMTT.
2774 (vect_record_base_alignments): Likewise.
2775 (vect_calculate_target_alignment): Likewise.
2776 (vect_compute_data_ref_alignment): Likewise and make static.
2777 (vect_update_misalignment_for_peel): Likewise.
2778 (vect_verify_datarefs_alignment): Likewise.
2779 (vector_alignment_reachable_p): Likewise.
2780 (vect_get_data_access_cost): Likewise. Pass down
2781 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
2782 (vect_get_peeling_costs_all_drs): Likewise.
2783 (vect_peeling_hash_get_lowest_cost): Likewise.
2784 (vect_enhance_data_refs_alignment): Likewise.
2785 (vect_find_same_alignment_drs): Likewise.
2786 (vect_analyze_data_refs_alignment): Likewise.
2787 (vect_analyze_group_access_1): Likewise.
2788 (vect_analyze_group_access): Likewise.
2789 (vect_analyze_data_ref_access): Likewise.
2790 (vect_analyze_data_ref_accesses): Likewise.
2791 (vect_vfa_segment_size): Likewise.
2792 (vect_small_gap_p): Likewise.
2793 (vectorizable_with_step_bound_p): Likewise.
2794 (vect_prune_runtime_alias_test_list): Likewise.
2795 (vect_analyze_data_refs): Likewise.
2796 (vect_supportable_dr_alignment): Likewise.
2797 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
2798 (vect_gen_prolog_loop_niters): Likewise.
2799 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
2800 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
2802 (vect_recog_mask_conversion_pattern): Likewise.
2803 (vect_try_gather_scatter_pattern): Likewise.
2804 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
2805 to vect_get_store_cost.
2806 (vect_get_store_cost): Get stmt_info instead of DR.
2807 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
2808 (vect_get_load_cost): Get stmt_info instead of DR.
2810 2018-06-01 Richard Biener <rguenther@suse.de>
2813 * gimple-fold.c (var_decl_component_p): Also allow offsetted
2814 vars wrapped in MEM_REFs.
2816 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
2818 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
2819 Fix subreg tests so that we only return a choice between
2820 GENERAL_REGS and FP_REGS if the original classes included both.
2822 2018-06-01 Richard Biener <rguenther@suse.de>
2825 * tree-ssa-structalias.c (get_function_part_constraint):
2826 Handle NULL fi->decl.
2827 (find_func_aliases_for_call): Properly handle indirect
2828 fi from direct call.
2829 (find_func_clobbers): Likewise.
2830 (ipa_pta_execute): Likewise.
2831 (create_variable_info_for): For functions that are ifunc_resolver
2832 resolve to a varinfo that contains the result of the resolver call.
2833 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
2836 2018-05-31 Michael Collison <michael.collison@arm.com>
2838 * config/aarch64/aarch64.md:
2839 (*fix_to_zero_extenddfdi2): New pattern.
2840 * gcc.target/aarch64/fix_extend1.c: New testcase.
2842 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
2846 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
2847 and BUILT_IN_STRNCMP_EQ.
2848 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
2849 BUILT_IN_STRNCMP_EQ.
2850 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
2851 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
2852 (gimple_fold_builtin): Likewise.
2853 * tree-ssa-strlen.c (compute_string_length): New function.
2854 (determine_min_obsize): New function.
2855 (handle_builtin_string_cmp): New function to handle calls to
2856 string compare functions.
2857 (strlen_optimize_stmt): Add handling to builtin string compare
2859 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
2860 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
2861 * tree.c (build_common_builtin_nodes): Add new defines of
2862 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
2864 2018-05-31 Jakub Jelinek <jakub@redhat.com>
2867 * bb-reorder.c (pass_partition_blocks::gate): Return false for
2868 functions with naked attribute.
2870 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
2872 * config/i386/sse.md (avx_vec_concat<mode>):
2873 Substitute concat_tg_mode mode attribute with xtg_mode.
2874 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
2875 (concat_tg_mode): Remove mode attribute.
2877 2018-05-31 Martin Sebor <msebor@redhat.com>
2880 * calls.c (alloc_max_size): Correct a logic error/typo.
2881 Treat excessive arguments as infinite. Warn for invalid arguments.
2882 * doc/invoke.texi (-Walloc-size-larger-than): Update.
2884 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
2887 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
2888 and movx for Haswell.
2890 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
2891 Cesar Philippidis <cesar@codesourcery.com>
2894 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
2895 when emitting error on private/firstprivate reductions.
2896 * omp-low.c (lower_omp_target): Avoid reference-type processing
2897 on pointers for firstprivate clause.
2899 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
2901 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
2904 * config/aarch64/aarch64-simd.md
2905 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
2906 (aarch64_ld1_x3_<mode>): Likewise
2907 (aarch64_st1x2<VALLDIF:mode>): Likewise
2908 (aarch64_st1_x2_<mode>): Likewise
2909 (aarch64_st1x3<VALLDIF:mode>): Likewise
2910 (aarch64_st1_x3_<mode>): Likewise
2911 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
2912 (vld1_s8_x3): Likewise.
2913 (vld1_u16_x3): Likewise.
2914 (vld1_s16_x3): Likewise.
2915 (vld1_u32_x3): Likewise.
2916 (vld1_s32_x3): Likewise.
2917 (vld1_u64_x3): Likewise.
2918 (vld1_s64_x3): Likewise.
2919 (vld1_f16_x3): Likewise.
2920 (vld1_f32_x3): Likewise.
2921 (vld1_f64_x3): Likewise.
2922 (vld1_p8_x3): Likewise.
2923 (vld1_p16_x3): Likewise.
2924 (vld1_p64_x3): Likewise.
2925 (vld1q_u8_x3): Likewise.
2926 (vld1q_s8_x3): Likewise.
2927 (vld1q_u16_x3): Likewise.
2928 (vld1q_s16_x3): Likewise.
2929 (vld1q_u32_x3): Likewise.
2930 (vld1q_s32_x3): Likewise.
2931 (vld1q_u64_x3): Likewise.
2932 (vld1q_s64_x3): Likewise.
2933 (vld1q_f16_x3): Likewise.
2934 (vld1q_f32_x3): Likewise.
2935 (vld1q_f64_x3): Likewise.
2936 (vld1q_p8_x3): Likewise.
2937 (vld1q_p16_x3): Likewise.
2938 (vld1q_p64_x3): Likewise.
2939 (vst1_s64_x2): Likewise.
2940 (vst1_u64_x2): Likewise.
2941 (vst1_f64_x2): Likewise.
2942 (vst1_s8_x2): Likewise.
2943 (vst1_p8_x2): Likewise.
2944 (vst1_s16_x2): Likewise.
2945 (vst1_p16_x2): Likewise.
2946 (vst1_s32_x2): Likewise.
2947 (vst1_u8_x2): Likewise.
2948 (vst1_u16_x2): Likewise.
2949 (vst1_u32_x2): Likewise.
2950 (vst1_f16_x2): Likewise.
2951 (vst1_f32_x2): Likewise.
2952 (vst1_p64_x2): Likewise.
2953 (vst1q_s8_x2): Likewise.
2954 (vst1q_p8_x2): Likewise.
2955 (vst1q_s16_x2): Likewise.
2956 (vst1q_p16_x2): Likewise.
2957 (vst1q_s32_x2): Likewise.
2958 (vst1q_s64_x2): Likewise.
2959 (vst1q_u8_x2): Likewise.
2960 (vst1q_u16_x2): Likewise.
2961 (vst1q_u32_x2): Likewise.
2962 (vst1q_u64_x2): Likewise.
2963 (vst1q_f16_x2): Likewise.
2964 (vst1q_f32_x2): Likewise.
2965 (vst1q_f64_x2): Likewise.
2966 (vst1q_p64_x2): Likewise.
2967 (vst1_s64_x3): Likewise.
2968 (vst1_u64_x3): Likewise.
2969 (vst1_f64_x3): Likewise.
2970 (vst1_s8_x3): Likewise.
2971 (vst1_p8_x3): Likewise.
2972 (vst1_s16_x3): Likewise.
2973 (vst1_p16_x3): Likewise.
2974 (vst1_s32_x3): Likewise.
2975 (vst1_u8_x3): Likewise.
2976 (vst1_u16_x3): Likewise.
2977 (vst1_u32_x3): Likewise.
2978 (vst1_f16_x3): Likewise.
2979 (vst1_f32_x3): Likewise.
2980 (vst1_p64_x3): Likewise.
2981 (vst1q_s8_x3): Likewise.
2982 (vst1q_p8_x3): Likewise.
2983 (vst1q_s16_x3): Likewise.
2984 (vst1q_p16_x3): Likewise.
2985 (vst1q_s32_x3): Likewise.
2986 (vst1q_s64_x3): Likewise.
2987 (vst1q_u8_x3): Likewise.
2988 (vst1q_u16_x3): Likewise.
2989 (vst1q_u32_x3): Likewise.
2990 (vst1q_u64_x3): Likewise.
2991 (vst1q_f16_x3): Likewise.
2992 (vst1q_f32_x3): Likewise.
2993 (vst1q_f64_x3): Likewise.
2994 (vst1q_p64_x3): Likewise.
2996 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
2998 * config/msp430/msp430.c (msp430_output_labelref): Prepend
2999 user_label_prefix to name.
3001 * tree-core.h: Update comment about the format of NAME string
3002 passed to handler in attribute_spec.
3004 * config/msp430/msp430.md: Remove erroneous subreg expression from
3005 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
3006 zero_extend{q,h}isi2.
3008 2018-05-30 Borislav Petkov <bp@suse.de>
3010 * doc/extend.texi: Document some architecture specific
3011 constraints and sort entries.
3013 2018-05-30 Martin Sebor <msebor@redhat.com>
3016 * builtins.c (expand_builtin_stpcpy_1): New function.
3017 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
3018 only if the former succeeds.
3020 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
3022 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
3025 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
3027 * doc/invoke.texi (-flinker-output): Document
3029 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
3031 * passes.c (ipa_write_summaries): Only modify statements if body
3033 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
3034 incrementally linking.
3035 (ipa_passes): Likewise.
3036 * lto-cgraph.c (lto_output_node): When incrementally linking do not
3037 pass down resolution info.
3038 * common.opt (flag_incremental_link): Update info.
3039 * gcc.c (plugin specs): Turn flinker-output=* to
3040 -plugin-opt=-linker-output-known
3041 * toplev.c (compile_file): Also cut compilation when doing incremental
3043 * flag-types. (enum lto_partition_model): Add
3044 LTO_LINKER_OUTPUT_NOLTOREL.
3045 (invoke.texi): Add -flinker-output docs.
3046 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
3047 link same way as WPA; do not stream in dead initializers.
3049 * dwarf2out.c (dwarf2out_die_ref_for_decl,
3050 darf2out_register_external_decl): Support incremental link.
3052 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
3054 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
3056 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
3058 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
3059 it down to simple_object_copy_lto_debug_sections.
3060 (run_gcc): Determine incremental LTO link time and configure
3061 lto1 into non-wpa mode, disable renaming of debug sections.
3063 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
3065 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
3066 descriptions of various incorrectly documented functions.
3068 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
3071 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
3072 address check not strict.
3074 2018-05-30 Richard Biener <rguenther@suse.de>
3076 PR tree-optimization/85964
3077 * tracer.c (better_p): Drop initialized count check, we only
3078 call the function with initialized counts now.
3079 (find_best_successor): Do find a best edge if one
3080 has uninitialized count.
3081 (find_best_predecessor): Likewise. Do BB frequency check only
3082 if count is initialized.
3084 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
3086 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
3087 (aarch64_ldrstr_offset_compare): New.
3088 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
3089 load/store orderings.
3090 (aarch64_gen_adjusted_ldpstp): Likewise.
3092 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
3094 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
3095 Check for subset of GENERAL_REGS and FP_REGS.
3096 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
3099 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
3101 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
3102 and wi::to_poly_offset. Add the current offset and then check
3103 whether the sum fits, rather than using an unchecked addition of
3104 a checked term. Check for a shwi rather than a uhwi.
3105 * expr.c (get_bit_range): Use tree_to_poly_uint64.
3106 (store_constructor): Use poly_int_tree_p.
3107 (expand_expr_real_1): Likewise.
3108 * function.c (assign_temp): Likewise.
3109 * fold-const.c (const_binop): Use poly_int_tree_p and
3111 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
3113 * ipa-icf-gimple.c (func_checker::compare_operand): Use
3114 to_poly_offset for MEM offsets.
3115 * ipa-icf.c (sem_variable::equals): Likewise.
3116 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
3117 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
3118 wi::to_poly_offset for BIT_FIELD_REF offsets.
3119 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
3121 * var-tracking.c (emit_note_insn_var_location): Use
3122 tree_to_poly_uint64.
3124 2018-05-29 Jim Wilson <jimw@sifive.com>
3126 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
3128 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
3131 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
3132 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
3134 (sse4_1_round<mode>2): Use nonimmediate_operand
3135 for operand 1 predicate.
3137 2018-05-29 Martin Sebor <msebor@redhat.com>
3138 Richard Biener <rguenther@suse.de>
3141 * calls.c (get_size_range): Call determine_value_range instead
3142 of get_value_range..
3143 * tree-vrp.h (determine_value_range): Declared new function.
3144 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
3146 2018-05-29 Richard Biener <rguenther@suse.de>
3148 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
3149 sure to use non-pattern stmts for get_earlier_stmt arguments.
3150 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
3151 called on pattern stmts.
3152 (get_later_stmt): Likewise.
3154 2018-05-29 Martin Liska <mliska@suse.cz>
3156 PR gcov-profile/85759
3157 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
3160 2018-05-29 Jakub Jelinek <jakub@redhat.com>
3162 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
3164 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
3168 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
3169 VEC_PACK_FLOAT_EXPR): New tree codes.
3170 * tree-pretty-print.c (op_code_prio): Handle
3171 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
3172 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
3173 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
3174 * tree-inline.c (estimate_operator_cost): Likewise.
3175 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
3176 * fold-const.c (const_binop): Likewise.
3177 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
3178 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
3179 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
3180 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
3181 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
3182 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
3183 * expr.c (expand_expr_real_2): Likewise.
3184 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
3185 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
3186 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
3188 * optabs.c (expand_widen_pattern_expr): For
3189 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
3190 sign from result type rather than operand's type.
3191 (expand_binop_directly): For vec_packu_float_optab and
3192 vec_packs_float_optab allow result type to be different from operand's
3194 * optabs-tree.c (optab_for_tree_code): Handle
3195 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
3196 VEC_PACK_FLOAT_EXPR. Formatting fixes.
3197 * tree-vect-generic.c (expand_vector_operations_1): Handle
3198 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
3199 VEC_PACK_FLOAT_EXPR.
3200 * tree-vect-stmts.c (supportable_widening_operation): Handle
3202 (supportable_narrowing_operation): Handle FLOAT_EXPR.
3203 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
3204 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
3205 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
3206 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
3208 (vec_pack<floatprefix>_float_<mode>): New expander.
3209 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
3211 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
3212 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
3213 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
3214 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
3215 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
3217 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
3218 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
3219 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
3220 VEC_PACK_FLOAT_EXPR): Document.
3222 2018-05-29 Richard Biener <rguenther@suse.de>
3224 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
3226 (stmt_vec_info_vec): Make pointer.
3227 (init_stmt_vec_info_vec): Remove.
3228 (free_stmt_vec_info_vec): Likewise.
3229 (set_stmt_vec_info_vec): New function.
3230 (free_stmt_vec_infos): Likewise.
3231 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
3232 (set_vinfo_for_stmt): Likewise.
3233 (get_earlier_stmt): Likewise.
3234 (get_later_stmt): Likewise.
3235 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
3236 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
3237 (vec_info::~vec_info): Free stmt_vec_infos.
3238 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
3239 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
3240 (pass_slp_vectorize::execute): Likewise.
3241 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
3242 (free_stmt_vec_info_vec): Likewise.
3243 (set_stmt_vec_info_vec): New function.
3244 (free_stmt_vec_infos): Likewise.
3245 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
3246 the global stmt_vec_info_vec.
3247 * tree-parloops.c (gather_scalar_reductions): Use
3248 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
3251 2018-05-29 Richard Biener <rguenther@suse.de>
3253 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
3255 2018-05-29 Martin Liska <mliska@suse.cz>
3256 David Malcolm <dmalcolm@redhat.com>
3258 * vec.c (test_reverse): New.
3259 (vec_c_tests): Add new test.
3260 * vec.h (vl_ptr>::reverse): New function.
3262 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
3264 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
3266 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
3269 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
3271 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
3273 2018-05-28 Richard Biener <rguenther@suse.de>
3275 PR tree-optimization/85933
3276 * tree-vect-data-refs.c (vect_record_base_alignments): Only
3277 look at stmts marked as vectorizable.
3279 2018-05-28 Richard Biener <rguenther@suse.de>
3281 PR tree-optimization/85934
3282 * tree-vect-generic.c (expand_vector_operations_1): Hoist
3283 vector boolean check before scalar optimization.
3285 2018-05-28 Jakub Jelinek <jakub@redhat.com>
3287 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
3290 2018-05-28 Mark Wielaard <mark@klomp.org>
3292 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
3293 if it is an expression containing a minus sign.
3295 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
3297 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
3299 2018-05-27 Paul Koning <ni1d@arrl.net>
3301 * config/pdp11/pdp11.md (truncsihi2): Remove.
3303 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
3304 Chung-Ju Wu <jasonwucj@gmail.com>
3306 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
3308 (unaligned_store_dw): Ditto.
3309 * config/nds32/nds32-memory-manipulation.c
3310 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
3311 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
3312 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
3313 (emit_setmem_word_loop): Rename to ...
3314 (emit_setmem_doubleword_loop): ... this.
3315 (nds32_gen_dup_4_byte_to_word_value): New function.
3316 (nds32_gen_dup_8_byte_to_double_word_value): New function.
3317 (nds32_expand_setmem_loop): Refine implementation.
3318 (nds32_expand_setmem_loop_v3m): Ditto.
3319 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
3322 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
3324 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
3326 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
3328 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
3329 (nds32_init_machine_status): Initialize machine->attr_naked_p and
3330 machine->attr_no_prologue_p.
3331 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
3332 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
3333 (nds32_expand_epilogue): Consider attr_naked_p.
3334 (nds32_expand_epilogue_v3pop): Likewise.
3335 (nds32_can_use_return_insn): Likewise.
3336 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
3337 attr_no_prologue_p fields.
3338 * config/nds32/nds32.opt (mret-in-naked-func): New option.
3340 2018-05-27 Jakub Jelinek <jakub@redhat.com>
3343 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
3345 * config/i386/sse.md
3346 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
3348 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
3350 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
3352 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
3354 (*<floatsuffix>floatv2div2sf2): Rename to ...
3355 (*float<floatunssuffix>v2div2sf2): ... this.
3356 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
3357 (float<floatunssuffix>v2div2sf2_mask): ... this.
3358 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
3359 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
3360 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
3362 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
3364 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
3366 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
3368 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
3370 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
3372 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
3373 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
3374 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
3375 gen_ufix_truncv8dfv8si2.
3376 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
3377 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
3378 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
3379 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
3380 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
3381 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
3382 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
3383 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
3385 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
3389 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
3391 2018-05-25 Jim Wilson <jimw@sifive.com>
3393 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
3394 * config/riscv/riscv.c (struct machine_function): Add
3395 interrupt_handler_p and attribute_checked_p fields.
3396 (riscv_attribute_table): Add interrupt.
3397 (riscv_interrupt_type_p): New.
3398 (riscv_save_reg_p): Save extra regs for interrupt handler.
3399 (riscv_use_save_libcall): Return false for interrupt handler.
3400 (riscv_first_stack_step): Add forward declaration.
3401 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
3402 for interrupt handler with large frame. Use it for saved reg list.
3403 (riscv_expand_prologue): Move flag_stack_usage_info support to
3404 eliminate duplication.
3405 (riscv_expand_epilogue): Generate mret for interrupt handler.
3406 (riscv_epilogue_uses): New.
3407 (riscv_can_use_return_insn): Return false for interrupt handler.
3408 (riscv_function_ok_for_sibcall): Likewise.
3409 (riscv_set_current_function): Add interrupt handler support.
3410 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
3411 * config/riscv/riscv.md (UNSPECV_MRET): New.
3413 (riscv_frflags, riscv_fsflags): Use tab after opcode.
3415 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
3417 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
3419 PR tree-optimization/85712
3420 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
3421 this candidate has already been replaced in-situ by a copy.
3423 2018-05-25 Jason Merrill <jason@redhat.com>
3425 PR c++/80485 - inline function non-zero address.
3426 * symtab.c (nonzero_address): Check DECL_COMDAT.
3428 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
3431 * config/alpha/alpha.md (ashlsi3): New insn pattern.
3432 (*ashlsi_se): Rename from *ashldi_se. Define as sign
3433 extension of SImode operation. Use const123_operand predicate.
3434 (*saddsi_1): Remove.
3435 (*saddl_se_1): Ditto.
3437 (*ssubl_se_1): Ditto.
3438 * config/alpha/predicates.md (const123_operand): New predicate.
3439 * config/alpha/constraints.md (P): Use IN_RANGE.
3441 2018-05-25 Richard Biener <rguenther@suse.de>
3443 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
3445 (ref_maybe_used_by_stmt_p): Likewise.
3446 (stmt_may_clobber_ref_p): Likewise.
3447 (stmt_may_clobber_ref_p_1): Likewise.
3448 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
3450 (ref_maybe_used_by_stmt_p): Likewise.
3451 (stmt_may_clobber_ref_p): Likewise.
3452 (stmt_may_clobber_ref_p_1): Likewise.
3453 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
3454 the alias oracle to disambiguate DRs with stmts DR analysis
3456 (vect_analyze_data_refs): Do not give up on not analyzable
3457 DRs for BB vectorization. Remove code truncating the dataref
3460 2018-05-25 Jakub Jelinek <jakub@redhat.com>
3463 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
3464 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
3465 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
3467 2018-05-25 Richard Biener <rguenther@suse.de>
3469 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
3470 function, combining stmt data ref gathering and fatal analysis
3472 (vect_analyze_data_refs): Remove now redudnant code and simplify.
3473 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
3474 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
3475 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
3476 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
3478 2018-05-25 Bin Cheng <bin.cheng@arm.com>
3480 PR tree-optimization/85720
3481 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
3482 SCC if all partitions are builtins.
3483 (version_loop_by_alias_check): New parameter. Generate cancelable
3484 runtime alias check if all partitions are builtins.
3485 (distribute_loop): Update call to above function.
3487 2018-05-25 Bin Cheng <bin.cheng@arm.com>
3489 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
3490 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
3491 (parm_default_def_partition_arg): Ditto.
3492 (set_parm_default_def_partition): Ditto.
3493 (get_parm_default_def_partitions): Ditto and make it static.
3494 (get_undefined_value_partitions): Ditto and make it static.
3495 (remove_ssa_form): Refactor call to init_var_map here.
3496 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
3497 computation for loop region.
3498 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
3499 (register_default_def): Delete.
3500 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
3501 (parm_default_def_partition_arg): Ditto.
3502 (set_parm_default_def_partition): Ditto.
3503 (get_parm_default_def_partitions): Ditto and make it static.
3504 (get_undefined_value_partitions): Ditto and make it static.
3505 (coalesce_with_default, coalesce_with_default): Update comment.
3506 (create_coalesce_list_for_region): New func factored out from
3507 create_outofssa_var_map.
3508 (populate_coalesce_list_for_outofssa): New func factored out from
3509 create_outofssa_var_map and coalesce_ssa_name.
3510 (create_outofssa_var_map): Delete.
3511 (coalesce_ssa_name): Refactor to support live range computation.
3512 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
3513 (get_parm_default_def_partitions): Delete.
3514 (get_undefined_value_partitions): Ditto.
3515 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
3516 computation for loop region.
3517 (new_tree_live_info, loe_visit_block): Ditto.
3518 (live_worklist, set_var_live_on_entry): Ditto.
3519 (calculate_live_on_exit, verify_live_on_entry): Ditto.
3520 * tree-ssa-live.h (struct _var_map): New fields.
3521 (init_var_map): Change decl.
3522 (region_contains_p): New.
3524 2018-05-25 Bin Cheng <bin.cheng@arm.com>
3526 * tree-ssa-live.h (live_merge_and_clear): Delete.
3528 2018-05-25 Richard Biener <rguenther@suse.de>
3531 * tree-dump.c (dequeue_and_dump): Remove access to removed
3532 operand 2 of a SWITCH_EXPR.
3534 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3536 * doc/sourcebuild.texi (vect_double_cond_arith): Include
3537 multiplication and division.
3538 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
3539 (cond_udiv@var{m}, cond_umod@var{m}): Document.
3540 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
3541 (cond_udiv_optab, cond_umod_optab): New optabs.
3542 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
3543 (IFN_COND_RDIV): New internal functions.
3544 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
3545 TRUNC_MOD_EXPR and RDIV_EXPR.
3546 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
3547 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
3549 (SVE_INT_BINARY): Include mult.
3550 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
3551 (optab, sve_int_op): Handle mult.
3552 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
3554 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
3555 for SVE_INT_BINARY_SD.
3557 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3559 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
3560 (optab, sve_int_op): Handle div and udiv.
3561 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
3562 for SVE_INT_BINARY_SD.
3563 (*<optab><mode>3): New insn for the same.
3565 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3567 * tree-vect-patterns.c: Include predict.h.
3568 (vect_recog_divmod_pattern): Restrict check for division support
3569 to when optimizing for size.
3571 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3573 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
3574 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
3575 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
3576 (gimple_match_op::set_op): Likewise.
3577 (gimple_resimplify4): Declare.
3578 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
3579 (expr::gen_transform): Likewise.
3580 (decision_tree::gen): Generate a simplification routine for 4 operands.
3581 * gimple-match-head.c (gimple_simplify): Add an overload for
3582 4 operands. In the top-level function, handle up to 4 call
3583 arguments and call gimple_resimplify4.
3584 (gimple_resimplify4): New function.
3585 (build_call_internal): Pass a fourth operand.
3586 (maybe_push_to_seq): Likewise.
3587 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
3588 Fold VEC_COND_EXPRs of an operation and a default value into
3589 an IFN_COND_* function if possible.
3590 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
3592 (SVE_COND_FP_BINARY): Include them.
3593 (optab, sve_fp_op): Handle them.
3594 (SVE_INT_BINARY_REV): New code iterator.
3595 (SVE_COND_FP_BINARY_REV): New int iterator.
3596 (commutative): New int attribute.
3597 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
3599 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
3601 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
3602 (*cond_<optab><mode>): New patterns for reversed operands.
3604 2018-05-25 Richard Biener <rguenther@suse.de>
3606 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
3607 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
3608 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
3609 (STMT_VINFO_GROUPED_ACCESS): Adjust.
3610 * tree-vect-data-refs.c (everywhere): Adjust users.
3611 * tree-vect-loop.c (everywhere): Likewise.
3612 * tree-vect-slp.c (everywhere): Likewise.
3613 * tree-vect-stmts.c (everywhere): Likewise.
3614 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
3616 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3618 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
3620 (gcc_cv_as_section_exclude): ... this.
3621 Try Solaris as #exclude syntax.
3622 * configure: Regenerate.
3623 * config.in: Regenerate.
3624 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
3626 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
3627 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
3629 * varasm.c (default_elf_asm_named_section): Don't check if
3630 HAVE_GAS_SECTION_EXCLUDE is defined.
3632 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3634 * doc/md.texi: Update the documentation of the cond_* optabs
3635 to mention the new final operand. Fix GET_MODE_NUNITS call.
3636 Describe the scalar case too.
3637 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
3638 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
3640 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
3641 (get_conditional_internal_fn): Update comment.
3642 * tree-vect-loop.c (vectorizable_reduction): Pass the original
3643 accumulator value as a final argument to conditional functions.
3644 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
3645 a define_expand and add an "else" operand. Assert for now that
3646 the else operand is equal to operand 2. Use SVE_INT_BINARY and
3647 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
3648 (*cond_<optab><mode>): New patterns.
3649 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
3650 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
3651 (UNSPEC_COND_EOR): Delete.
3652 (optab): Remove associated mappings.
3653 (SVE_INT_BINARY): New code iterator.
3654 (sve_int_op): Remove int attribute and add "minus" to the code
3656 (SVE_COND_INT_OP): Delete.
3657 (SVE_COND_FP_OP): Rename to...
3658 (SVE_COND_FP_BINARY): ...this.
3660 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3662 * optabs.c (can_reuse_operands_p): New function.
3663 (maybe_legitimize_operands): Try to reuse the results for
3666 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
3668 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
3669 Add {q} suffix to insn mnemonic.
3671 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
3673 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
3674 (msp430_warn_func_return): New.
3676 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
3678 * fold-const.c (tree_nonzero_bits): New function.
3679 * fold-const.h (tree_nonzero_bits): Likewise.
3680 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
3681 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
3683 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
3687 * varasm.c (assemble_alias): Check ifunc_resolver only on
3690 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
3693 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
3694 when memory input operand is handled.
3696 2018-05-24 Luis Machado <luis.machado@linaro.org>
3698 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
3700 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
3702 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
3704 * match.pd: Delay FMA folds until after vectorization.
3706 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
3709 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
3710 address check not strict.
3712 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
3714 * gimple-match.h (gimple_match_op): New class.
3715 (mprts_hook): Replace parameters with a gimple_match_op *.
3716 (maybe_build_generic_op): Likewise.
3717 (gimple_simplified_result_is_gimple_val): Replace parameters with
3718 a const gimple_match_op *.
3719 (gimple_simplify): Replace code_helper * and tree * parameters with
3720 a gimple_match_op * parameter.
3721 (gimple_resimplify1): Replace code_helper *, tree and tree *
3722 parameters with a gimple_match_op * parameter.
3723 (gimple_resimplify2): Likewise.
3724 (gimple_resimplify3): Likewise.
3725 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
3726 parameters with a gimple_match_op * parameter.
3727 * gimple-match-head.c (gimple_simplify): Change prototypes of
3728 auto-generated functions to take a gimple_match_op * instead of
3729 separate code_helper * and tree * parameters. Make the same
3730 change in the top-level overload and update calls to the
3731 gimple_resimplify routines. Update calls to the auto-generated
3732 functions and to maybe_push_res_to_seq in the publicly-facing
3733 operation-specific gimple_simplify overloads.
3734 (gimple_match_op::MAX_NUM_OPS): Define.
3735 (gimple_resimplify1): Replace rcode and ops with a single res_op
3736 parameter. Update call to gimple_simplify.
3737 (gimple_resimplify2): Likewise.
3738 (gimple_resimplify3): Likewise.
3739 (mprts_hook): Replace parameters with a gimple_match_op *.
3740 (maybe_build_generic_op): Likewise.
3741 (build_call_internal): Replace type, nargs and ops with
3742 a gimple_match_op *.
3743 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
3744 with a single gimple_match_op *. Update calls to mprts_hook,
3745 build_call_internal and gimple_simplified_result_is_gimple_val.
3746 Factor out code that is common to the tree_code and combined_fn cases.
3747 * genmatch.c (expr::gen_transform): Replace tem_code and
3748 tem_ops with a gimple_match_op called tem_op. Update calls
3749 to the gimple_resimplify functions and maybe_push_res_to_seq.
3750 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
3751 res_ops. Update call to the gimple_resimplify functions.
3752 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
3753 (decision_tree::gen): Make the functions take a gimple_match_op *
3754 called res_op instead of separate res_code and res_ops parameters.
3755 Update call accordingly.
3756 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
3757 and ops with a single res_op parameter. Update calls to
3758 maybe_build_generic_op and maybe_push_res_to_seq.
3759 (fold_stmt_1): Update calls to gimple_simplify and
3760 replace_stmt_with_simplification.
3761 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
3762 and gimple_simplified_result_is_gimple_val.
3763 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
3765 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
3766 with a gimple_match_op *.
3767 (vn_nary_build_or_lookup): Likewise. Update call to
3768 vn_nary_build_or_lookup_1.
3769 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
3770 gimple_match_op *. Update calls to the gimple_resimplify routines
3771 and to gimple_simplified_result_is_gimple_val.
3772 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
3773 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
3774 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
3775 (visit_nary_op): Likewise.
3776 (visit_reference_op_load): Likewise.
3778 2018-05-23 Luis Machado <luis.machado@linaro.org>
3780 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
3781 modifier for printing the step amount.
3783 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
3786 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
3789 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
3791 * doc/sourcebuild.texi (Endianness): New subsubsection.
3793 2018-05-23 Luis Machado <luis.machado@linaro.org>
3795 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3796 <prefetch_dynamic_strides>: New const bool field.
3797 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3798 prefetch_dynamic_strides.
3799 (exynosm1_prefetch_tune): Likewise.
3800 (thunderxt88_prefetch_tune): Likewise.
3801 (thunderx_prefetch_tune): Likewise.
3802 (thunderx2t99_prefetch_tune): Likewise.
3803 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
3805 (aarch64_override_options_internal): Update to set
3806 PARAM_PREFETCH_DYNAMIC_STRIDES.
3807 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
3808 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
3809 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
3810 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
3811 prefetch-dynamic-strides setting.
3813 2018-05-23 Luis Machado <luis.machado@linaro.org>
3815 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3816 <minimum_stride>: New const int field.
3817 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3818 minimum_stride field defaulting to -1.
3819 (exynosm1_prefetch_tune): Likewise.
3820 (thunderxt88_prefetch_tune): Likewise.
3821 (thunderx_prefetch_tune): Likewise.
3822 (thunderx2t99_prefetch_tune): Likewise.
3823 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
3824 <default_opt_level>: Set to 3.
3825 (aarch64_override_options_internal): Update to set
3826 PARAM_PREFETCH_MINIMUM_STRIDE.
3827 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
3828 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
3829 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
3830 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
3831 stride is constant and is below the minimum stride threshold.
3833 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3835 * config/arm/arm-cpus.in (mode26): Delete.
3836 (armv4): Delete mode26 reference.
3837 * config/arm/arm.c (arm_configure_build_target): Delete use of
3840 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
3842 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
3844 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
3845 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
3847 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
3848 (floatunsdidf2): Ditto.
3850 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
3852 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
3853 (fixuns_trunc<mode>si2_avx512f): Ditto.
3854 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
3855 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
3856 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
3858 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
3860 PR rtl-optimization/79985
3861 * df-scan.c (df_insn_refs_collect): Remove special case for
3862 global registers and asm statements.
3864 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
3866 * extend.texi (Global Register Variables): Rewrite the bullet list.
3867 Note that the register is available for allocation. Note that access
3868 via inline asm must use constraints. Add note about async-signal
3869 handlers. Remove paragraph about automagic register selection.
3871 2018-05-23 Richard Biener <rguenther@suse.de>
3873 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
3874 of fixed offset from memset VN.
3876 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
3878 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
3880 (alloc_cand_and_find_basis): Initialize first_interp field.
3881 (slsr_process_mul): Modify first_interp field.
3882 (slsr_process_add): Likewise.
3883 (slsr_process_cast): Modify first_interp field for each new
3885 (slsr_process_copy): Likewise.
3886 (dump_candidate): Dump first_interp field.
3887 (replace_mult_candidate): Process all interpretations, not just
3889 (replace_rhs_if_not_dup): Likewise.
3890 (replace_one_candidate): Likewise.
3892 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
3894 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
3896 (aarch64_needs_frame_chain): New function.
3897 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
3899 2018-05-23 Sudakshina Das <sudi.das@arm.com>
3902 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
3903 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
3904 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
3905 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
3906 as true for strict-align.
3907 (aarch64_can_inline_p): Perform checks even when callee has no
3908 attributes to check for strict alignment.
3909 * doc/extend.texi (AArch64 Function Attributes): Document
3911 * doc/invoke.texi: (AArch64 Options): Likewise.
3913 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
3915 PR tree-optimization/85853
3916 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
3917 the handling of the root of the node to...
3918 (vect_slp_analyze_node_operations_1): ...this new function,
3919 and run the whole thing with the child nodes' def types
3920 set according to their SLP node's def type.
3922 2018-05-23 Richard Biener <rguenther@suse.de>
3925 * tree-data-ref.c (create_runtime_alias_checks): Defer
3926 and ignore overflow warnings.
3928 2018-05-23 Yury Gribov <tetra2005@gmail.com>
3930 PR tree-optimization/85822
3931 * tree-vrp.c (is_masked_range_test): Fix handling of negative
3934 2018-05-23 Richard Biener <rguenther@suse.de>
3936 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
3937 memset constants via native_interpret_expr.
3939 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
3942 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
3944 (cgraph_node::create_alias): Likewise.
3945 (cgraph_node::get_availability): Check ifunc_resolver instead
3946 of looking up ifunc attribute.
3947 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
3948 * varasm.c (do_assemble_alias): Likewise.
3949 (assemble_alias): Likewise.
3950 (default_binds_local_p_3): Likewise.
3951 * cgraph.h (cgraph_node): Add ifunc_resolver.
3952 (cgraph_node::only_called_directly_or_aliased_p): Return false
3954 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
3956 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
3957 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
3958 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
3959 instead of looking up ifunc attribute.
3961 2018-05-22 Luis Machado <luis.machado@linaro.org>
3963 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
3965 2018-05-22 Martin Sebor <msebor@redhat.com>
3968 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
3969 only when expasion succeeds.
3970 (expand_builtin_strcmp): Same.
3971 (expand_builtin_strncmp): Same.
3973 2018-05-22 Martin Sebor <msebor@redhat.com>
3975 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
3977 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
3978 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3980 * config/aarch64/aarch64-ldpstp.md: Replace uses of
3981 aarch64_mem_pair_operand with memory_operand and delete operand swapping
3983 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
3984 Add check for legitimate_address.
3985 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
3986 (aarch64_swap_ldrstr_operands): New.
3987 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
3990 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
3991 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3993 * config/aarch64/aarch64.md: New patterns to generate stp
3995 (store_pair_sw, store_pair_dw): New patterns to generate stp for
3996 single words and double words.
3997 (load_pair_sw, load_pair_dw): Likewise.
3998 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
4000 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
4002 * config/aarch64/aarch64-ldpstp.md: Modify peephole
4003 for different mode ldpstp and add peephole for merged zero stores.
4005 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
4007 (aarch64_gen_store_pair): Rename calls to match new patterns.
4008 (aarch64_gen_load_pair): Rename calls to match new patterns.
4009 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
4010 (load_pair<DREG:mode><DREG2:mode>): ... This.
4011 (store_pair<mode>): Rename to...
4012 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
4013 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
4015 (V_INT_EQUIV): Handle SImode.
4016 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
4019 2018-05-22 Martin Sebor <msebor@redhat.com>
4022 * calls.c (maybe_warn_nonstring_arg): Use string length to set
4023 or ajust the presumed bound on an operation to avoid unnecessary
4026 2018-05-22 Martin Sebor <msebor@redhat.com>
4028 PR tree-optimization/85826
4029 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
4030 assuming that a DECL necesarily has a constant size.
4032 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
4035 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
4037 2018-05-22 Richard Biener <rguenther@suse.de>
4039 PR tree-optimization/85834
4040 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
4041 non-constant and non-zero memset arguments.
4043 2018-05-22 Martin Liska <mliska@suse.cz>
4046 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
4048 2018-05-22 Richard Biener <rguenther@suse.de>
4050 PR tree-optimization/85863
4051 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
4052 comparisons when vectype is specified.
4053 (vectorizable_condition): Do not specify vectype for
4054 vect_is_simple_cond when SLP vectorizing.
4056 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
4059 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
4060 define __ibm128 as long double.
4061 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
4062 as a distinct type when IEEE 128-bit support is enabled.
4063 (init_float128_ieee): Fix up conversions between IFmode and IEEE
4064 128-bit types to use the correct functions.
4065 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
4066 convert between 128-bit floating point types that have different
4067 modes but the same representation, instead of using gen_lowpart to
4069 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
4071 (IFKF_reg): New attributes to give the register constraints for
4073 (extend<mode>tf2_internal): New insns to mark an explicit
4074 conversion between 128-bit floating point types that have a
4075 different mode but share the same representation.
4077 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
4079 PR tree-optimization/85814
4080 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
4081 a null return from get_strinfo when unsharing the next
4082 strinfo in the chain.
4084 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
4087 * varasm.c (weak_finish): Clean up weak_decls.
4089 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4091 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
4092 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
4093 UNSPEC_UADALP values.
4094 * config/aarch64/iterators.md (ABAL): New int iterator.
4097 (sur): Add mappings for the above.
4098 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
4100 (aarch64_<sur>abal<mode>_4): Likewise.
4101 (aarch64_<sur>adalp<mode>_3): Likewise.
4102 (<sur>sadv16qi): New define_expand.
4104 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
4106 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
4107 (*movdf_internal): Ditto.
4108 (*rcpsf2_sse): Ditto.
4109 (*rsqrtsf2_sse): Ditto.
4110 (*sqrt<mode>2_sse): Ditto.
4112 2018-05-21 Tamar Christina <tamar.christina@arm.com>
4114 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
4116 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
4117 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
4118 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
4119 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
4121 * config/aarch64/arm_neon.h: Likewise.
4122 * config/aarch64/iterators.md (VQ_I): New.
4124 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
4126 * config.gcc: Add arc/t-multilib-linux to tmake_file for
4128 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
4131 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
4133 * config/nds32/constraints.md (S): New constraint.
4134 * config/nds32/nds32.md (call_internal): Use constraint S.
4135 (call_value_internal): Likewise.
4136 (sibcall_internal): Likewise.
4137 (sibcall_value_internal): Likewise.
4139 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
4140 Chung-Ju Wu <jasonwucj@gmail.com>
4142 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
4145 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
4146 Chung-Ju Wu <jasonwucj@gmail.com>
4148 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
4149 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
4150 (nds32_rtx_costs_impl): Simplify.
4151 (nds32_address_cost_impl): Simplify.
4152 (nds32_init_rtx_costs): New function.
4153 (nds32_rtx_costs_speed_prefer): Likewise.
4154 (nds32_rtx_costs_size_prefer): Likewise.
4155 (nds32_address_cost_speed_prefer): Likewise.
4156 (nds32_address_cost_speed_fwprop): Likewise.
4157 (nds32_address_cost_size_prefer): Likewise.
4158 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
4159 * config/nds32/nds32.c (nds32_option_override): Use
4160 nds32_init_rtx_costs function.
4162 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
4164 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
4165 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
4166 (TARGET_PIPELINE_N8): Likewise.
4167 (TARGET_PIPELINE_N10): Likewise.
4168 (TARGET_PIPELINE_N13): Likewise.
4169 (TARGET_PIPELINE_GRAYWOLF): Likewise.
4171 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
4173 * config/nds32/nds32-fpu.md: Update copyright year.
4175 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
4177 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
4179 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
4181 * config/nds32/nds32.c
4182 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
4183 * config/nds32/nds32.opt (minline-asm-r15): New option.
4185 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
4187 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
4189 * config/nds32/nds32.md (abssi2): New pattern.
4191 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
4193 * config/i386/i386.md (rex64namesuffix): New mode attribute.
4194 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
4195 Merge insn pattern from sse_cvtsi2ss<round_name> and
4196 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
4197 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
4198 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
4199 using SWI48 mode iterator.
4200 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
4201 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
4202 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
4203 pattern from sse_cvttss2si<round_saeonly_name>
4204 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
4205 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
4206 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
4207 using SWI48 mode iterator.
4208 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
4209 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
4210 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
4211 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
4212 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
4213 using SWI48 mode iterator.
4214 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
4215 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
4216 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
4217 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
4218 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
4219 SWI48 mode iterator.
4220 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
4221 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
4222 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
4223 pattern from sse_cvttsd2si<round_saeonly_name>
4224 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
4226 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
4228 * config/nds32/nds32-md-auxiliary.c
4229 (nds32_valid_smw_lwm_base_p): Refine.
4230 (nds32_output_smw_single_word): Refine.
4231 (nds32_output_smw_double_word): New.
4232 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
4234 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
4236 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
4237 (nds32_output_stack_pop): Refine.
4238 (nds32_expand_unaligned_load): Refine.
4239 (nds32_expand_unaligned_store): Refine.
4241 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
4242 Chung-Ju Wu <jasonwucj@gmail.com>
4244 * config/nds32/constants.md: Add TP_REGNUM constant.
4245 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
4246 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
4248 * config/nds32/nds32-doubleword.md: Consider flag_pic.
4249 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
4250 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
4251 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
4252 and PIC code generation.
4253 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
4255 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
4257 * config/nds32/nds32.md: Support TLS and PIC.
4258 * config/nds32/nds32.c: Support TLS and PIC.
4259 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
4260 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
4263 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
4265 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
4266 mode with E_ prefix.
4268 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
4269 Chung-Ju Wu <jasonwucj@gmail.com>
4271 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
4272 * config/nds32/nds32-md-auxiliary.c
4273 (symbolic_reference_mentioned_p): New.
4274 (nds32_legitimize_ict_address): New.
4275 (nds32_expand_ict_move): New.
4276 (nds32_indirect_call_referenced_p): New.
4277 (nds32_symbol_binds_local_p): Delete.
4278 (nds32_long_call_p): Modify.
4279 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
4280 * config/nds32/nds32-protos.h
4281 (symbolic_reference_mentioned_p): Declare.
4282 (nds32_legitimize_ict_address): Declare.
4283 (nds32_expand_ict_move): Declare.
4284 (nds32_indirect_call_referenced_p): Declare.
4285 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
4286 (nds32_relax_group): Use nds32_ict_const_p as condition.
4287 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
4288 (nds32_asm_file_start): Output ict_model directive in asm code.
4289 (nds32_legitimate_address_p): Consider indirect call.
4290 (nds32_print_operand): Consider indirect call.
4291 (nds32_print_operand_address): Consider indirect call.
4292 (nds32_insert_attributes): Handle "indirect_call" attribute.
4293 (TARGET_LEGITIMATE_ADDRESS_P): Define.
4294 (TARGET_LEGITIMATE_CONSTANT_P): Define.
4295 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
4296 (TARGET_DELEGITIMIZE_ADDRESS): Define.
4297 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
4298 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
4299 (TARGET_ICT_MODEL_SMALL): Define.
4300 (TARGET_ICT_MODEL_LARGE): Define.
4301 * config/nds32/nds32.md (movsi): Consider ict model.
4302 (call, call_value): Consider ict model.
4303 (sibcall, sibcall_value): Consider ict model.
4304 * config/nds32/nds32.opt (mict-model): New option.
4305 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
4308 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
4309 Monk Chiang <sh.chiang04@gmail.com>
4310 Jim Wilson <jimw@sifive.com>
4312 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
4313 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
4314 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
4315 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
4316 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
4317 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
4318 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
4319 compute save_libcall_adjustment properly.
4320 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
4321 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
4322 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
4323 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
4324 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
4325 (ABI_SPEC): Handle mabi=ilp32e.
4326 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
4327 (RVE): Add RVE mask.
4328 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
4329 <-march>: Add rv32e as an example.
4331 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
4334 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
4335 (intra_create_variable_infos): Handle C++ constructors.
4337 2018-05-18 Martin Liska <mliska@suse.cz>
4339 * passes.def: Remove a redundant pass.
4341 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
4344 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
4346 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4348 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
4350 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
4352 (armv2, armv2a, armv3, armv3m): Delete architectures.
4353 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
4354 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
4355 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
4357 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
4358 (*mulsidi3adddi): Likewise.
4359 (mulsidi3): Likewise.
4360 (*mulsidi3_nov6): Likewise.
4361 (umulsidi3): Likewise.
4362 (umulsidi3_nov6): Likewise.
4363 (umaddsidi4): Likewise.
4364 (*umulsidi3adddi): Likewise.
4365 (smulsi3_highpart): Likewise.
4366 (*smulsi3_highpart_nov6): Likewise.
4367 (umulsi3_highpart): Likewise.
4368 (*umulsi3_highpart_nov6): Likewise.
4369 * config/arm/arm.h (arm_arch3m): Delete.
4370 * config/arm/arm.c (arm_arch3m): Delete.
4371 (arm_option_override_internal): Update armv3-related comment.
4372 (arm_configure_build_target): Delete use of isa_bit_mode32.
4373 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
4374 (arm_rtx_costs_internal): Delete check of arm_arch3m.
4375 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
4377 (mulusa3): Likewise.
4378 * config/arm/arm-protos.h (arm_arch3m): Delete.
4379 * config/arm/arm-tables.opt: Regenerate.
4380 * config/arm/arm-tune.md: Likewise.
4381 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
4382 deleted architectures.
4384 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4386 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
4387 (armv5t, armv5te): New features.
4388 (ARMv5, ARMv5e): Delete fgroups.
4389 (ARMv5t, ARMv5te): Adjust for above changes.
4391 (armv5, armv5e): Delete arches.
4392 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
4394 (*call_reg_arm): Likewise.
4395 (*call_value_reg_armv5): Likewise.
4396 (*call_value_reg_arm): Likewise.
4397 (*call_symbol): Likewise.
4398 (*call_value_symbol): Likewise.
4399 (*sibcall_insn): Likewise.
4400 (*sibcall_value_insn): Likewise.
4402 (prefetch): Likewise.
4403 (define_split and define_peephole2 dependent on arm_arch5):
4405 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
4407 (TARGET_ARM_QBIT): Likewise.
4408 (TARGET_DSP_MULTIPLY): Likewise.
4409 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
4410 (arm_arch5, arm_arch5e): Delete.
4411 (arm_arch5t, arm_arch5te): Declare.
4412 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
4413 (arm_arch5t): Declare.
4414 (arm_option_reconfigure_globals): Update for the above.
4415 (arm_options_perform_arch_sanity_checks): Update comment, replace
4416 use of arm_arch5 with arm_arch5t.
4417 (use_return_insn): Likewise.
4418 (arm_emit_call_insn): Likewise.
4419 (output_return_instruction): Likewise.
4420 (arm_final_prescan_insn): Likewise.
4421 (arm_coproc_builtin_available): Likewise.
4422 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
4423 arm_arch5e with arm_arch5t and arm_arch5te.
4424 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
4425 (arm_arch5t, arm_arch5te): Declare.
4426 * config/arm/arm-tables.opt: Regenerate.
4427 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
4428 * config/arm/t-multilib: Likewise.
4429 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
4430 instead of arm_arch5.
4431 (*call_reg_thumb1): Likewise.
4432 (*call_value_reg_thumb1_v5): Likewise.
4433 (*call_value_reg_thumb1): Likewise.
4434 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
4436 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
4438 2018-05-18 Martin Liska <mliska@suse.cz>
4440 PR gcov-profile/84846
4441 * doc/gcov.texi: Document -t option of gcov tool.
4443 2018-05-18 Martin Liska <mliska@suse.cz>
4445 PR gcov-profile/84846
4446 * gcov.c (print_usage): Add new -t option.
4447 (process_args): Handle the option.
4448 (generate_results): Use stdout as output when requested by
4451 2018-05-18 Martin Liska <mliska@suse.cz>
4453 PR gcov-profile/84846
4454 * coverage.c (coverage_init): Write PWD to .gcno file.
4455 * doc/gcov.texi: Document how working directory is printed.
4456 * gcov-dump.c (dump_gcov_file): Print PWD.
4457 * gcov.c (output_intermediate_file): Likewise.
4458 (read_graph_file): Read PWD string.
4459 (output_lines): Print PWD.
4461 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
4464 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
4465 for retval and return false if all args to phi are zero.
4467 2018-05-18 Richard Biener <rguenther@suse.de>
4469 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
4471 (evrp_dom_walker::before_dom_children): Call it.
4473 2018-05-18 Richard Biener <rguenther@suse.de>
4475 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
4476 results when processing array refs with variable index.
4478 2018-05-18 Toon Moene <toon@moene.org>
4480 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
4481 directly after that of -floop-interchange. Indicate that both
4482 options are enabled by default when specifying -O3.
4484 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4486 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
4487 iterator. Delete separate integer-mode vec_set<mode> expander.
4488 (aarch64_simd_vec_setv2di): Delete.
4489 (vec_setv2di): Delete.
4490 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
4491 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
4492 the "w, r" alternative.
4494 2018-05-18 Martin Liska <mliska@suse.cz>
4496 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
4497 * tree-pass.h (make_pass_lower_switch_O0): New function.
4498 * tree-switch-conversion.c (node_has_low_bound): Remove.
4499 (node_has_high_bound): Likewise.
4500 (node_is_bounded): Likewise.
4501 (class pass_lower_switch): Make it a template type and create
4503 (pass_lower_switch::execute): Add template argument.
4504 (make_pass_lower_switch): New function.
4505 (make_pass_lower_switch_O0): New function.
4506 (do_jump_if_equal): Remove.
4507 (emit_case_nodes): Simplify to just handle all 3 cases and leave
4508 all the hard work to tree optimization passes.
4510 2018-05-18 Martin Liska <mliska@suse.cz>
4512 * dbgcnt.c (limit_low): Renamed from limit.
4513 (limit_high): New variable.
4514 (dbg_cnt_is_enabled): Check for upper limit.
4515 (dbg_cnt): Adjust dumping.
4516 (dbg_cnt_set_limit_by_index): Add new argument for high
4518 (dbg_cnt_set_limit_by_name): Likewise.
4519 (dbg_cnt_process_single_pair): Parse new format.
4520 (dbg_cnt_process_opt): Use strtok.
4521 (dbg_cnt_list_all_counters): Remove 'value' and add
4523 * doc/invoke.texi: Document changes.
4525 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
4527 * doc/sourcebuild.texi (scalar_all_fma): Document.
4528 * tree.def (FMA_EXPR): Delete.
4529 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
4530 * internal-fn.c (ternary_direct): New macro.
4531 (expand_ternary_optab_fn): Likewise.
4532 (direct_ternary_optab_supported_p): Likewise.
4533 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
4534 * builtins.c (fold_builtin_fma): Delete.
4535 (fold_builtin_3): Don't call it.
4536 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
4537 * expr.c (expand_expr_real_2): Likewise.
4538 * fold-const.c (operand_equal_p): Likewise.
4539 (fold_ternary_loc): Likewise.
4540 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
4541 * gimple.c (DEFTREECODE): Likewise.
4542 * gimplify.c (gimplify_expr): Likewise.
4543 * optabs-tree.c (optab_for_tree_code): Likewise.
4544 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
4545 * tree-eh.c (operation_could_trap_p): Likewise.
4546 (stmt_could_throw_1_p): Likewise.
4547 * tree-inline.c (estimate_operator_cost): Likewise.
4548 * tree-pretty-print.c (dump_generic_node): Likewise.
4549 (op_code_prio): Likewise.
4550 * tree-ssa-loop-im.c (stmt_cost): Likewise.
4551 * tree-ssa-operands.c (get_expr_operands): Likewise.
4552 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
4553 * fold-const-call.h (fold_fma): Delete.
4554 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
4555 CFN_FNMA and CFN_FNMS.
4557 * genmatch.c (combined_fn): New enum.
4558 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
4559 (commutative_op): New function.
4560 (commutate): Use it. Handle more than 2 operands.
4561 (dt_operand::gen_gimple_expr): Use commutative_op.
4562 (parser::parse_expr): Allow :c to be used with non-binary
4563 operators if the commutative operand is known.
4564 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
4565 CFN_FMS, CFN_FNMA and CFN_FNMS.
4566 (backprop::process_assign_use): Remove FMA_EXPR handling.
4567 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
4568 (gen_hsa_fma): New function.
4569 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
4570 IFN_FNMA and IFN_FNMS.
4571 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
4572 * gimple-fold.h (follow_all_ssa_edges): Declare.
4573 * gimple-fold.c (follow_all_ssa_edges): New function.
4574 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
4575 gimple_build interface and use follow_all_ssa_edges to fold the result.
4576 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
4577 instead of checking for optabs directly.
4578 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
4579 rather than FMA_EXPRs.
4580 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
4581 call to IFN_FMA instead of an FMA_EXPR.
4583 2018-05-17 Jim Wilson <jimw@sifive.com>
4585 * expr.c (do_tablejump): When converting index to Pmode, if we have a
4586 sign extended promoted subreg, and the range does not have the sign bit
4587 set, then do a sign extend.
4589 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
4590 test, check for sign extended subreg and/or constant operands, and
4591 do a sign extend in that case.
4593 2018-05-17 Steve Ellcey <sellcey@cavium.com>
4595 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
4596 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
4598 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
4599 Change logics_shift_reg to logics_shift_imm.
4600 (thunderx2t99_fp_loadpair_basic): Delete.
4601 (thunderx2t99_fp_storepair_basic): Delete.
4602 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
4603 (thunderx2t99_asimd_polynomial): Delete.
4604 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
4605 and neon_fp_mul_d_scalar_q.
4606 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
4607 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
4608 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
4609 (thunderx2t99_asimd_lut): Add missing tbl types.
4610 (thunderx2t99_asimd_ext): Delete.
4611 (thunderx2t99_asimd_load1_1_mult): Delete.
4612 (thunderx2t99_asimd_load1_2_mult): Delete.
4613 (thunderx2t99_asimd_load1_ldp): New.
4614 (thunderx2t99_asimd_load1): New.
4615 (thunderx2t99_asimd_load2): Add missing *load2* types.
4616 (thunderx2t99_asimd_load3): New.
4617 (thunderx2t99_asimd_load4): New.
4618 (thunderx2t99_asimd_store1_1_mult): Delete.
4619 (thunderx2t99_asimd_store1_2_mult): Delete.
4620 (thunderx2t99_asimd_store2_mult): Delete.
4621 (thunderx2t99_asimd_store2_onelane): Delete.
4622 (thunderx2t99_asimd_store_stp): New.
4623 (thunderx2t99_asimd_store1): New.
4624 (thunderx2t99_asimd_store2): New.
4625 (thunderx2t99_asimd_store3): New.
4626 (thunderx2t99_asimd_store4): New.
4628 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
4630 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
4631 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
4633 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
4634 Segher Boessenkool <segher@kernel.crashing.org>
4637 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
4640 2018-05-17 Richard Biener <rguenther@suse.de>
4642 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
4643 for pruning loop and prune defs feeding only already visited PHIs.
4645 2018-05-17 Richard Biener <rguenther@suse.de>
4647 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
4649 2018-05-17 Bin Cheng <bin.cheng@arm.com>
4650 Richard Biener <rguenther@suse.de>
4652 PR tree-optimization/85793
4653 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
4654 for VMAT_ELEMENTWISE.
4656 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
4658 * internal-fn.h (lookup_internal_fn): Declare
4659 * internal-fn.c (lookup_internal_fn): New function.
4660 * gimple.c (gimple_build_call_from_tree): Handle calls to
4662 * gimple-pretty-print.c (dump_gimple_call): Print "." before
4663 internal function names.
4664 * tree-pretty-print.c (dump_generic_node): Likewise.
4665 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
4667 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
4669 * gimple-fold.h (gimple_build): Make the function forms take
4670 combined_fn rather than built_in_function.
4671 (gimple_simplify): Likewise.
4672 * gimple-match-head.c (gimple_simplify): Likewise.
4673 * gimple-fold.c (gimple_build): Likewise.
4674 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
4675 rather than gimple_build_call_internal.
4676 (get_initial_defs_for_reduction): Likewise.
4677 (vect_create_epilog_for_reduction): Likewise.
4678 (vectorizable_live_operation): Likewise.
4680 2018-05-17 Martin Liska <mliska@suse.cz>
4682 * gimple-ssa-sprintf.c (format_directive): Do not use
4683 space in between 'G_' and '('.
4685 2018-05-17 Jakub Jelinek <jakub@redhat.com>
4688 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
4689 even if the mask is not all ones.
4692 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
4694 (ix86_gimple_fold_builtin): Likewise.
4697 * config/i386/i386.c: Include tree-vector-builder.h.
4698 (ix86_vector_shift_count): New function.
4699 (ix86_fold_builtin): Fold shift builtins by scalar count.
4700 (ix86_gimple_fold_builtin): Likewise.
4702 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
4703 _mm512_setzero): New intrinsics.
4705 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
4706 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4708 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
4709 code generation for cases where splatting a value is not useful.
4710 * simplify-rtx.c (simplify_ternary_operation): Simplify
4711 vec_merge across a vec_duplicate and a paradoxical subreg forming
4712 a vector mode to a vec_concat.
4714 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
4716 * config.gcc: Support "goldmont-plus".
4717 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
4719 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
4720 PROCESSOR_GOLDMONT_PLUS.
4721 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
4722 (processor_target_table): Add "goldmont-plus".
4723 (PTA_GOLDMONT_PLUS): Define.
4724 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
4725 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
4726 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
4727 (fold_builtin_cpu): Add "goldmont-plus".
4728 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
4729 (ix86_option_override_internal): Add "goldmont-plus".
4730 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
4731 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
4732 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
4733 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
4735 2018-05-17 Richard Biener <rguenther@suse.de>
4737 PR tree-optimization/85757
4738 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
4739 remove defs that only feed that PHI from further processing.
4741 2018-05-16 Jim Wilson <jimw@sifive.com>
4743 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
4745 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
4747 2018-05-16 Mark Wielaard <mark@klomp.org>
4749 * dwarf2out.c (count_index_strings): New function.
4750 (output_indirect_strings): Call count_index_strings and generate
4751 header for dwarf_version >= 5.
4753 2018-05-16 Mark Wielaard <mark@klomp.org>
4755 * dwarf2out.c (dwarf_FORM): New function.
4756 (set_indirect_string): Use dwarf_FORM.
4757 (reset_indirect_string): Likewise.
4758 (size_of_die): Likewise.
4759 (value_format): Likewise.
4760 (output_die): Likewise.
4761 (add_skeleton_AT_string): Likewise.
4762 (output_macinfo_op): Likewise.
4763 (index_string): Likewise.
4764 (output_index_string_offset): Likewise.
4765 (output_index_string): Likewise.
4766 (count_index_strings): Likewise.
4768 2018-05-16 Carl Love <cel@us.ibm.com>
4770 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
4771 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
4773 2018-05-16 Martin Jambor <mjambor@suse.cz>
4775 * ipa-prop.c (ipa_free_all_edge_args): Remove.
4776 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
4778 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
4780 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
4781 (fnma<mode>4): Likewise.
4782 (fms<mode>4): Likewise.
4783 (fnms<mode>4): Likewise.
4784 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
4785 (aarch64_fnma<mode>4): Likewise.
4786 (aarch64_fms<mode>4): Likewise.
4787 (aarch64_fnms<mode>4): Likewise.
4788 (aarch64_fnmadd<mode>4): Likewise.
4790 2018-05-16 Jason Merrill <jason@redhat.com>
4792 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
4794 2018-05-16 Richard Biener <rguenther@suse.de>
4796 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
4797 (dump_stmt_cost): Declare.
4798 (add_stmt_cost): Dump cost we add.
4799 (add_stmt_costs): New function.
4800 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
4802 (vect_analyze_stmt): Adjust prototype.
4803 (vectorizable_condition): Likewise.
4804 (vectorizable_live_operation): Likewise.
4805 (vectorizable_reduction): Likewise.
4806 (vectorizable_induction): Likewise.
4807 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
4808 cost vector to pass to vectorizable_ and record afterwards.
4809 (vect_model_reduction_cost): Take cost vector argument and adjust.
4810 (vect_model_induction_cost): Likewise.
4811 (vectorizable_reduction): Likewise.
4812 (vectorizable_induction): Likewise.
4813 (vectorizable_live_operation): Likewise.
4814 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
4815 SLP_TREE_NUMBER_OF_VEC_STMTS.
4816 (vect_analyze_slp_cost_1): Remove.
4817 (vect_analyze_slp_cost): Likewise.
4818 (vect_slp_analyze_node_operations): Take visited args and
4819 a target cost vector. Avoid processing already visited stmt sets.
4820 (vect_slp_analyze_operations): Use a local cost vector to gather
4821 costs and register those of non-discarded instances.
4822 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
4823 (vect_schedule_slp_instance): Remove copying of
4824 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
4826 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
4827 adding cost. Record cost entry location.
4828 (vect_prologue_cost_for_slp_op): Function to compute cost of
4829 a constant or invariant generated for SLP vect in the prologue,
4830 split out from vect_analyze_slp_cost_1.
4831 (vect_model_simple_cost): Make static. Adjust for SLP costing.
4832 (vect_model_promotion_demotion_cost): Likewise.
4833 (vect_model_store_cost): Likewise, make static.
4834 (vect_model_load_cost): Likewise.
4835 (vectorizable_bswap): Add cost vector arg and adjust.
4836 (vectorizable_call): Likewise.
4837 (vectorizable_simd_clone_call): Likewise.
4838 (vectorizable_conversion): Likewise.
4839 (vectorizable_assignment): Likewise.
4840 (vectorizable_shift): Likewise.
4841 (vectorizable_operation): Likewise.
4842 (vectorizable_store): Likewise.
4843 (vectorizable_load): Likewise.
4844 (vectorizable_condition): Likewise.
4845 (vectorizable_comparison): Likewise.
4846 (can_vectorize_live_stmts): Likewise.
4847 (vect_analyze_stmt): Likewise.
4848 (vect_transform_stmt): Adjust calls to vectorizable_*.
4849 * tree-vectorizer.c: Include gimple-pretty-print.h.
4850 (dump_stmt_cost): New function.
4852 2018-05-16 Richard Biener <rguenther@suse.de>
4854 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
4855 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
4856 * tree-ssa-dse.c: Include tree-ssa-loop.h.
4857 (check_name): New callback.
4858 (dse_classify_store): Track cycles via a visited bitmap of PHI
4859 defs and simplify handling of in-loop and across loop dead stores
4860 and properly fail for loop-variant refs. Handle byte-tracking with
4861 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
4864 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
4866 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
4867 (vect_get_mask_type_for_stmt): Likewise.
4868 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
4870 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
4871 to determine the statement's vector type and the vector type that
4872 should be used for calculating nunits. Deal with cases in which
4873 the type has to be deferred.
4874 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
4875 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
4876 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
4877 (vect_determine_vf_for_stmt): New functions, split out from...
4878 (vect_determine_vectorization_factor): ...here.
4879 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
4880 (vect_get_mask_type_for_stmt): New functions, split out from
4881 vect_determine_vectorization_factor.
4883 2018-05-16 Richard Biener <rguenther@suse.de>
4885 * tree-cfg.c (verify_gimple_assign_ternary): Properly
4886 verify the [VEC_]COND_EXPR embedded comparison.
4888 2018-05-15 Martin Sebor <msebor@redhat.com>
4890 PR tree-optimization/85753
4891 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
4892 RECORD_TYPE in addition to ARRAY_TYPE.
4894 2018-05-15 Martin Sebor <msebor@redhat.com>
4897 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
4899 2018-05-15 Richard Biener <rguenther@suse.de>
4901 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
4902 add by_clobber_p one. Change algorithm to collect all defs
4903 representing uses we need to walk and try reducing them to
4904 a single one before failing.
4905 (dse_dom_walker::dse_optimize_stmt): Adjust.
4907 2018-05-13 Mark Wielaard <mark@klomp.org>
4909 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
4910 (size_of_loc_descr): Likewise.
4911 (output_loc_operands): Likewise.
4912 (output_loc_operands_raw): Likewise.
4913 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
4914 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
4915 (hash_loc_operands): Likewise.
4916 (compare_loc_operands): Likewise.
4918 2018-05-14 Mark Wielaard <mark@klomp.org>
4920 * dwarf2out.c (count_index_addrs): New function.
4921 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
4923 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
4925 PR tree-optimization/83648
4926 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
4927 return value as malloc candidate.
4929 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
4932 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
4933 param as true in call to suggest_attribute.
4935 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
4937 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
4938 -mreadonly-in-sdata.
4940 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4942 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
4944 (aarch64_crypto_aesd_fused): Likewise.
4946 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
4948 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
4949 (movsi_aarch64): Likewise.
4950 (load_pairsi): Likewise.
4951 (load_pairdi): Likewise.
4952 (store_pairsi): Likewise.
4953 (store_pairdi): Likewise.
4954 (load_pairsf): Likewise.
4955 (load_pairdf): Likewise.
4956 (store_pairsf): Likewise.
4957 (store_pairdf): Likewise.
4958 (zero_extend): Likewise.
4959 (trunc): Swap alternatives.
4960 (fcvt_target): Add '?' to prefer w over r.
4962 2018-05-14 Jakub Jelinek <jakub@redhat.com>
4965 * config/i386/i386.md: Disallow non-commutative arithmetics in
4966 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
4967 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
4968 in the peephole2 before it.
4970 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
4972 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
4973 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
4974 (ix86_handle_option): Handle -mcldemote.
4975 * config.gcc: New header.
4976 * config/i386/cldemoteintrin.h: New file.
4977 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
4978 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
4980 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
4981 OPTION_MASK_ISA_CLDEMOTE.
4982 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
4983 (ix86_valid_target_attribute_inner_p): Ditto.
4984 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
4985 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
4986 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
4987 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
4988 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
4990 * config/i386/i386.opt: Add -mcldemote.
4991 * config/i386/x86intrin.h: New header.
4992 * doc/invoke.texi: Add -mcldemote.
4994 2018-05-14 Richard Biener <rguenther@suse.de>
4996 * doc/match-and-simplify.texi: Adjust :s documentation.
4998 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
5000 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
5001 intended memcpy size.
5002 (REORDER_45): Likewise.
5004 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
5006 * sort.cc: New file.
5007 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
5008 * vec.c (qsort_chk): Use gcc_qsort.
5009 * Makefile.in (OBJS-libcommon): Add sort.o.
5010 (build/sort.o): New target. Use it...
5011 (BUILD_RTL): ... here, and...
5012 (build/gencfn-macros): ... here, and...
5013 (build/genmatch): ... here.
5015 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
5016 Chung-Ju Wu <jasonwucj@gmail.com>
5018 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
5019 * config/nds32/nds32-graywolf.md: New file.
5020 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
5021 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
5023 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
5024 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
5025 * config/nds32/nds32.md (pipeline_model): Add graywolf.
5026 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
5027 * config/nds32/pipelines.md: Include n15 settings.
5029 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
5030 Chung-Ju Wu <jasonwucj@gmail.com>
5032 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
5033 * config/nds32/nds32-n13.md: New file.
5034 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
5035 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
5037 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
5038 * config/nds32/nds32.md (pipeline_model): Add n13.
5039 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
5040 * config/nds32/pipelines.md: Include n13 settings.
5042 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
5043 Chung-Ju Wu <jasonwucj@gmail.com>
5045 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
5046 * config/nds32/nds32-n10.md: New file.
5047 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
5048 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
5050 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
5051 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
5052 * config/nds32/nds32.md (pipeline_model): Add n10.
5053 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
5054 * config/nds32/pipelines.md: Include n10 settings.
5056 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
5057 Kito Cheng <kito.cheng@gmail.com>
5058 Chung-Ju Wu <jasonwucj@gmail.com>
5060 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
5061 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
5062 Add enum values for DSP extension instructions.
5063 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
5065 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
5066 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
5068 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
5069 * config/nds32/nds32-dspext.md: New file for DSP implementation.
5070 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
5071 * config/nds32/nds32-intrinsic.md: Likewise.
5072 * config/nds32/nds32_intrinsic.h: Likewise.
5073 * config/nds32/nds32-md-auxiliary.c: Likewise.
5074 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
5075 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
5076 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
5077 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
5078 * config/nds32/nds32-protos.h: New declarations for DSP extension.
5079 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
5080 TYPE_DMAC in switch statement.
5081 * config/nds32/nds32.c: New checking and implementation for DSP
5082 extension instructions.
5083 * config/nds32/nds32.h: Likewise.
5084 * config/nds32/nds32.md: Likewise.
5085 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
5086 * config/nds32/predicates.md: Implement new predicates for DSP
5089 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
5091 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
5092 Reformat alternatives and attributes so it is easier to identify
5093 which constraints/attributes go with which instruction.
5094 (mov<mode>_hardfloat32, FMOVE64): Likewise.
5095 (mov<mode>_softfloat32, FMOVE64): Likewise.
5096 (mov<mode>_hardfloat64, FMOVE64): Likewise.
5097 (mov<mode>_softfloat64, FMOVE64): Likewise.
5099 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
5101 * doc/extend.texi (PowerPC Built-in Functions): Rename this
5103 (Basic PowerPC Built-in Functions): The new name of the
5104 subsection previously known as "PowerPC Built-in Functions".
5105 (Basic PowerPC Built-in Functions Available on all Configurations):
5107 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
5108 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
5109 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
5110 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
5112 2018-05-11 Martin Jambor <mjambor@suse.cz>
5115 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
5118 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
5121 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
5123 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
5125 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
5126 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
5127 (ix86_handle_option): Handle -mwaitpkg.
5128 * config.gcc: New header.
5129 * config/i386/cpuid.h (bit_WAITPKG): New bit.
5130 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
5131 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
5133 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
5134 OPTION_MASK_ISA_WAITPKG.
5135 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
5136 (ix86_option_override_internal): Add PTA_WAITPKG.
5137 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
5138 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
5139 IX86_BUILTIN_TPAUSE.
5140 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
5141 __builtin_ia32_umwait and __builtin_ia32_tpause.
5142 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
5143 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
5144 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
5145 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
5146 UNSPECV_TPAUSE): New.
5147 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
5148 * config/i386/i386.opt: Add -mwaitpkg.
5149 * config/i386/waitpkgintrin.h: New file.
5150 * config/i386/x86intrin.h: New header.
5151 * doc/invoke.texi: Add -mwaitpkg.
5153 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
5156 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
5158 (cortex-m0): Use armv6s-m isa.
5159 (cortex-m0plus): Likewise.
5160 (cortex-m1): Likewise.
5161 (cortex-m0.small-multiply): Likewise.
5162 (cortex-m0plus.small-multiply): Likewise.
5163 (cortex-m1.small-multiply): Likewise.
5165 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
5166 Jakub Jelinek <jakub@redhat.com>
5168 PR tree-optimization/85692
5169 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
5170 source permute as well.
5172 2018-05-11 Martin Liska <mliska@suse.cz>
5175 * doc/extend.texi: Document LLVM style format for no_sanitize
5178 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
5180 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
5181 mode_supports_vsx_dform_quad to mode_supports_dq_form.
5182 (mode_supports_vsx_dform_quad): Likewise.
5183 (mode_supports_vmx_dform): Move these functions to be next to the
5184 other mode_supports functions.
5185 (mode_supports_dq_form): Likewise.
5186 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
5187 mode_supports_dq_form.
5188 (reg_offset_addressing_ok_p): Likewise.
5189 (offsettable_ok_by_alignment): Likewise.
5190 (rs6000_legitimate_offset_address_p): Likewise.
5191 (legitimate_lo_sum_address_p): Likewise.
5192 (rs6000_legitimize_address): Likewise.
5193 (rs6000_legitimize_reload_address): Likewise.
5194 (rs6000_secondary_reload_inner): Likewise.
5195 (rs6000_preferred_reload_class): Likewise.
5196 (rs6000_output_move_128bit): Likewise.
5198 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
5200 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
5201 Generate SImode target register for null target.
5202 <case IX86_BUILTIN_XGETBV>: Ditto.
5203 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
5204 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
5206 2018-05-10 Carl Love <cel@us.ibm.com>
5208 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
5209 dcbtt and dcbtstt if operands[2] is 0.
5211 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
5214 * config/i386/sse.md (usadv64qi): New expander.
5216 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
5218 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
5219 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
5220 -maltivec=be support.
5221 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
5222 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
5223 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
5224 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
5225 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
5226 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
5227 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
5228 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
5229 altivec_vsumsws): Adjust.
5230 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
5231 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
5232 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
5234 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
5235 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
5236 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
5237 (altivec_lve<VI_char>x): Delete expand.
5238 (*altivec_lve<VI_char>x_internal): Rename to...
5239 (altivec_lve<VI_char>x): ... this.
5240 (altivec_lvxl_<mode>): Delete expand.
5241 (*altivec_lvxl_<mode>_internal): Rename to ...
5242 (altivec_lvxl_<mode>): ... this.
5243 (altivec_stvxl_<mode>): Delete expand.
5244 (*altivec_stvxl_<mode>_internal): Rename to ...
5245 (altivec_stvxl_<mode>): ... this.
5246 (altivec_stve<VI_char>x): Delete expand.
5247 (*altivec_stve<VI_char>x_internal): Rename to ...
5248 (altivec_stve<VI_char>x): ... this.
5249 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
5250 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
5251 reduc_plus_scal_<mode>): Adjust.
5252 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
5254 (rs6000_cpu_cpp_builtins): Adjust.
5255 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
5256 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
5257 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
5258 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
5259 -maltivec=be support.
5260 (rs6000_split_vec_extract_var): Adjust.
5261 (rs6000_split_v4si_init): Adjust.
5262 (swap_selector_for_mode): Delete.
5263 (altivec_expand_lvx_be, altivec_expand_stvx_be,
5264 altivec_expand_stvex_be): Delete.
5265 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
5266 -maltivec=be support.
5267 (rs6000_gimple_fold_builtin): Ditto.
5268 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
5270 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
5271 (TARGET_DIRECT_MOVE_64BIT): Adjust.
5272 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
5273 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
5274 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
5275 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
5276 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
5277 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
5278 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
5279 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
5280 anonymous split): Adjust.
5281 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
5282 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
5284 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
5286 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
5287 when --with-gxx-include-dir is also specified.
5288 * configure: Regenerate.
5290 2018-05-09 Jim Wilson <jimw@sifive.com>
5293 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
5294 * config/riscv/t-withmultilib: New.
5295 * config/riscv/withmultilib.h: New.
5296 * doc/install.texi: Document RISC-V --with-multilib-list support.
5298 2018-05-09 Richard Biener <rguenther@suse.de>
5300 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
5302 (vect_bb_vectorization_profitable_p): Adjust. Compute
5303 actual scalar cost using the cost vector and the add_stmt_cost
5306 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
5308 PR rtl-optimization/85645
5309 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
5310 in the REG_CFA_REGISTER note for LR, don't leave it empty.
5312 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
5314 PR rtl-optimization/85645
5315 * shrink-wrap.c (spread_components): Return a boolean saying if
5316 anything was changed.
5317 (try_shrink_wrapping_separate): Iterate spread_components until
5318 nothing changes anymore.
5320 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
5322 PR rtl-optimization/85645
5323 * regrename.c (build_def_use): Also kill the chains that include the
5324 destination of a REG_CFA_REGISTER note.
5326 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
5328 PR rtl-optimization/85645
5329 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
5330 insn that has a REG_CFA_REGISTER note.
5332 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
5334 * cfgexpand.c (expand_clobber): New function.
5335 (expand_gimple_stmt_1): Use it.
5336 * tree-vect-stmts.c (vect_clobber_variable): New function,
5338 (vectorizable_simd_clone_call): ...here.
5339 (vectorizable_store): Emit a clobber either side of an
5340 IFN_STORE_LANES sequence.
5341 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
5343 2018-05-09 Tom de Vries <tom@codesourcery.com>
5346 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
5347 (define_insn "trap_if_false"): Add exit after trap.
5349 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
5351 PR rtl-optimization/85638
5352 * bb-reorder.c: Include common/common-target.h.
5353 (create_forwarder_block): New function extracted from...
5354 (fix_up_crossing_landing_pad): ...here. Rename into...
5355 (dw2_fix_up_crossing_landing_pad): ...this.
5356 (sjlj_fix_up_crossing_landing_pad): New function.
5357 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
5358 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
5359 from both partitions and exit the loop after one iteration.
5361 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
5364 * doc/extend.texi (PowerPC Built-in Functions): Rename this
5366 (Basic PowerPC Built-in Functions): The new name of the
5367 subsection previously known as "PowerPC Built-in Functions".
5368 (Basic PowerPC Built-in Functions Available on all Configurations):
5370 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
5372 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
5374 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
5376 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
5379 2018-05-08 Jim Wilson <jimw@sifive.com>
5381 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
5382 (LD_EMUL_SUFFIX): New.
5383 (LINK_SPEC): Use it.
5385 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
5387 * doc/extend.texi (PowerPC Built-in Functions): Rename this
5389 (Basic PowerPC Built-in Functions): The new name of the
5390 subsection previously known as "PowerPC Built-in Functions".
5391 (Basic PowerPC Built-in Functions Available on all Configurations):
5393 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
5395 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
5397 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
5399 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
5402 2018-05-08 Jakub Jelinek <jakub@redhat.com>
5405 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
5406 after cmpelim optimization.
5408 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
5410 * config.gcc: Support "goldmont".
5411 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
5412 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
5414 * config/i386/i386.c (m_GOLDMONT): Define.
5415 (processor_target_table): Add "goldmont".
5416 (PTA_GOLDMONT): Define.
5417 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
5418 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
5419 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
5420 (fold_builtin_cpu): Add "goldmont".
5421 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
5422 (ix86_option_override_internal): Add "goldmont".
5423 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
5424 (processor_type): Add PROCESSOR_GOLDMONT.
5425 * config/i386/i386.md: Add CPU "glm".
5426 * config/i386/glm.md: New file.
5427 * config/i386/x86-tune.def: Add m_GOLDMONT.
5428 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
5430 2018-05-08 Jakub Jelinek <jakub@redhat.com>
5433 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
5435 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
5436 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
5437 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
5440 * config/i386/i386.c (ix86_fold_builtin): Handle
5441 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
5444 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
5445 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
5447 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
5450 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
5451 (check_arch): Likewise.
5452 (check_fpu): Return the result rather than printing it.
5453 (end arch): Fix operator precedence.
5454 (end cpu): Likewise.
5455 (END): Print the result from check_fpu.
5457 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
5458 Alan Hayward <alan.hayward@arm.com>
5459 David Sherwood <david.sherwood@arm.com>
5461 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
5462 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
5463 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
5464 (*fcmuo<mode>_and): New patterns.
5466 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
5468 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
5469 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
5470 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
5471 (cmp_op, sve_imm_con): New code attributes.
5472 (SVE_COND_INT_CMP, imm_con): Delete.
5473 (cmp_op): Remove above unspecs from int attribute.
5474 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
5476 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
5477 comparison-specific unspecs.
5478 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
5479 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
5480 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
5481 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
5482 (*vec_fcm<cmp_op><mode>): Rename to...
5483 (*fcm<cmp_op><mode>): ...this and adjust likewise.
5484 (*vec_fcmuo<mode>): Rename to...
5485 (*fcmuo<mode>): ...this and adjust likewise.
5486 (*pred_fcm<cmp_op><mode>): New pattern.
5487 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
5488 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
5490 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
5492 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
5493 (aarch64_emit_sve_predicated_cond): New function.
5494 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
5495 (aarch64_emit_unspec_cond_or): Replace with...
5496 (aarch64_emit_sve_or_conds): ...this new function. Use
5497 aarch64_emit_sve_ptrue_op for the individual comparisons and
5498 aarch64_emit_binop to OR them together.
5499 (aarch64_emit_inverted_unspec_cond): Replace with...
5500 (aarch64_emit_sve_inverted_cond): ...this new function. Use
5501 aarch64_emit_sve_ptrue_op for the comparison and
5502 aarch64_emit_unop to invert the result.
5503 (aarch64_expand_sve_vec_cmp_float): Update after the above
5504 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
5506 2018-05-07 Nathan Sidwell <nathan@acm.org>
5508 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
5509 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
5510 (Backwards Compatibility): Likewise.
5512 2018-05-07 Luis Machado <luis.machado@linaro.org>
5516 2018-05-07 Luis Machado <luis.machado@linaro.org>
5518 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5519 <prefetch_dynamic_strides>: New const bool field.
5520 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5521 prefetch_dynamic_strides.
5522 (exynosm1_prefetch_tune): Likewise.
5523 (thunderxt88_prefetch_tune): Likewise.
5524 (thunderx_prefetch_tune): Likewise.
5525 (thunderx2t99_prefetch_tune): Likewise.
5526 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
5528 (aarch64_override_options_internal): Update to set
5529 PARAM_PREFETCH_DYNAMIC_STRIDES.
5530 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
5531 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
5532 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
5533 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
5534 prefetch-dynamic-strides setting.
5536 2018-05-07 Luis Machado <luis.machado@linaro.org>
5538 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5539 <minimum_stride>: New const int field.
5540 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5541 minimum_stride field.
5542 (exynosm1_prefetch_tune): Likewise.
5543 (thunderxt88_prefetch_tune): Likewise.
5544 (thunderx_prefetch_tune): Likewise.
5545 (thunderx2t99_prefetch_tune): Likewise.
5546 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
5547 (aarch64_override_options_internal): Update to set
5548 PARAM_PREFETCH_MINIMUM_STRIDE.
5549 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
5550 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
5551 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
5552 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
5553 stride is constant and is below the minimum stride threshold.
5555 2018-05-07 Luis Machado <luis.machado@linaro.org>
5557 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
5560 2018-05-07 Luis Machado <luis.machado@linaro.org>
5562 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5563 <prefetch_dynamic_strides>: New const bool field.
5564 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5565 prefetch_dynamic_strides.
5566 (exynosm1_prefetch_tune): Likewise.
5567 (thunderxt88_prefetch_tune): Likewise.
5568 (thunderx_prefetch_tune): Likewise.
5569 (thunderx2t99_prefetch_tune): Likewise.
5570 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
5572 (aarch64_override_options_internal): Update to set
5573 PARAM_PREFETCH_DYNAMIC_STRIDES.
5574 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
5575 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
5576 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
5577 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
5578 prefetch-dynamic-strides setting.
5580 2018-05-07 Luis Machado <luis.machado@linaro.org>
5582 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5583 <minimum_stride>: New const int field.
5584 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5585 minimum_stride field.
5586 (exynosm1_prefetch_tune): Likewise.
5587 (thunderxt88_prefetch_tune): Likewise.
5588 (thunderx_prefetch_tune): Likewise.
5589 (thunderx2t99_prefetch_tune): Likewise.
5590 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
5591 (aarch64_override_options_internal): Update to set
5592 PARAM_PREFETCH_MINIMUM_STRIDE.
5593 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
5594 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
5595 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
5596 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
5597 stride is constant and is below the minimum stride threshold.
5599 2018-05-06 Jakub Jelinek <jakub@redhat.com>
5602 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
5603 the type is addressable. Don't force op into register if it has
5606 2018-05-05 Roland McGrath <mcgrathr@google.com>
5609 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
5610 any section for which we don't know a specific type it should have,
5611 regardless of name. Previously this was done only for the exact
5612 names ".init_array", ".fini_array", and ".preinit_array".
5613 (default_elf_asm_named_section): Add comment about
5614 relationship with default_section_type_flags and SECTION_NOTYPE.
5615 (get_section): Don't consider it a type conflict if one side has
5616 SECTION_NOTYPE and the other doesn't, as long as neither has the
5617 SECTION_BSS et al used in the default_section_type_flags logic.
5619 2018-05-05 Tom de Vries <tom@codesourcery.com>
5622 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
5623 (workaround_barsyncs): New function.
5624 (nvptx_reorg): Use workaround_barsyncs.
5625 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
5626 (define_expand "nvptx_membar_cta"): New define_expand.
5627 (define_insn "*nvptx_membar_cta"): New insn.
5629 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
5631 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
5632 To improve optimization opportunities.
5633 * builtin-types.def: The new needed builtin types for the above.
5635 2018-05-04 Richard Biener <rguenther@suse.de>
5637 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
5638 * gimple-ssa-store-merging.c
5639 (imm_store_chain_info::output_merged_store): Remove redundant create,
5640 release split_store vector contents on failure.
5641 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
5642 scalar stmt vector on cache hit.
5644 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
5646 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
5648 * config.gcc (powerpc-xilinx-eabi*): Remove.
5649 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
5651 (fusion_addis_mem_combo_load): Ditto.
5652 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
5654 (rs6000_cpu_cpp_builtins): Ditto.
5655 * config/rs6000/rs6000-linux.c
5656 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
5657 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
5658 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
5660 (rs6000_setup_reg_addr_masks): Ditto.
5661 (rs6000_init_hard_regno_mode_ok): Ditto.
5662 (rs6000_option_override_internal): Ditto.
5663 (legitimate_lo_sum_address_p): Ditto.
5664 (rs6000_legitimize_address): Ditto.
5665 (rs6000_legitimize_reload_address): Ditto.
5666 (rs6000_legitimate_address_p): Ditto.
5667 (abi_v4_pass_in_fpr): Ditto.
5668 (setup_incoming_varargs): Ditto.
5669 (rs6000_gimplify_va_arg): Ditto.
5670 (rs6000_split_multireg_move): Ditto.
5671 (rs6000_savres_strategy): Ditto.
5672 (rs6000_emit_prologue_components): Ditto.
5673 (rs6000_emit_epilogue_components): Ditto.
5674 (rs6000_emit_prologue): Ditto.
5675 (rs6000_emit_epilogue): Ditto.
5676 (rs6000_elf_file_end): Ditto.
5677 (rs6000_function_value): Ditto.
5678 (rs6000_libcall_value): Ditto.
5679 * config/rs6000/rs6000.h: Ditto.
5680 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
5681 (TARGET_MINMAX): ... this. New.
5682 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
5683 * config/rs6000/rs6000.md: Remove Xilinx FP support.
5684 (*movsi_internal1_single): Delete.
5685 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
5686 mfpu=, mxilinx-fpu): Delete.
5687 * config/rs6000/singlefp.h: Delete.
5688 * config/rs6000/sysv4.h: Remove Xilinx FP support.
5689 * config/rs6000/t-rs6000: Ditto.
5690 * config/rs6000/t-xilinx: Delete.
5691 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
5692 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
5693 (VStype_simple): Delete.
5694 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
5695 * config/rs6000/xfpu.h: Delete.
5696 * config/rs6000/xfpu.md: Delete.
5697 * config/rs6000/xilinx.h: Delete.
5698 * config/rs6000/xilinx.opt: Delete.
5699 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
5700 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
5702 2018-05-04 Tom de Vries <tom@codesourcery.com>
5705 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
5708 2018-05-04 Richard Biener <rguenther@suse.de>
5711 * tree-complex.c (update_complex_assignment): We are always in SSA form.
5712 (expand_complex_div_wide): Likewise.
5713 (expand_complex_operations_1): Likewise.
5714 (expand_complex_libcall): Preserve EH info of the original stmt.
5715 (tree_lower_complex): Handle removed blocks.
5716 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
5717 on complex multiplication and division libcall builtins.
5719 2018-05-04 Richard Biener <rguenther@suse.de>
5722 * fold-const.c (negate_expr_p): Restrict negation of operand
5723 zero of a division to when we know that can happen without
5725 (fold_negate_expr_1): Likewise.
5727 2018-05-04 Jakub Jelinek <jakub@redhat.com>
5730 * real.h (real_nextafter): Declare.
5731 * real.c (real_nextafter): New function.
5732 * fold-const-call.c (fold_const_nextafter): New function.
5733 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
5734 CASE_CFN_NEXTTOWARD.
5735 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
5736 even when arg1_mode is different from arg0_mode.
5738 2018-05-03 Nathan Sidwell <nathan@acm.org>
5740 * doc/extend.texi (Deprecated Features): Remove
5742 (Backwards Compatibility): Likewise.
5743 * doc/invoke.texi (C++ Language Options): Likewise.
5744 (C++ Dialect Options): Likewise.
5746 2018-05-03 Jakub Jelinek <jakub@redhat.com>
5749 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
5750 _mm512_mask_mullox_epi64): New intrinsics.
5752 2018-05-03 Tom de Vries <tom@codesourcery.com>
5755 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
5756 dump files): Add offload-tree.
5758 2018-05-03 Richard Biener <rguenther@suse.de>
5760 PR tree-optimization/85615
5761 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
5762 to loops not nested in BBs loop father to avoid creating multi-entry
5765 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5767 PR tree-optimization/70291
5768 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
5769 arguments. Change return type to tree. Emit libcall as a new
5770 statement rather than replacing existing one when inplace_p is true.
5771 (expand_complex_multiplication_components): New function.
5772 (expand_complex_multiplication): Expand floating-point complex
5773 multiplication using the above.
5774 (expand_complex_division): Rename inner_type parameter to type.
5775 Update expand_complex_libcall call-site.
5776 (expand_complex_operations_1): Update expand_complex_multiplication
5777 and expand_complex_division call-sites.
5779 2018-05-02 Jakub Jelinek <jakub@redhat.com>
5782 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
5783 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
5784 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
5785 the highest significant bit of the shift count mask is clear. In
5786 check whether and[sq]i3 is needed verify that all significant bits
5787 of the shift count other than the highest are set.
5789 2018-05-02 Tom de Vries <tom@codesourcery.com>
5792 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
5793 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
5794 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
5795 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
5796 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
5797 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
5798 __builtin_goacc_parlevel_size.
5800 2018-05-02 Richard Biener <rguenther@suse.de>
5802 PR tree-optimization/85597
5803 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
5804 do not use split vect_get_vec_defs call but call vect_get_slp_defs
5807 2018-05-02 Tom de Vries <tom@codesourcery.com>
5810 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
5811 dump files): Add ltrans-tree.
5813 2018-05-02 Tom de Vries <tom@codesourcery.com>
5816 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
5817 dump files): Add wpa-ipa.
5819 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
5821 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
5822 powerpc*-*-linux*paired* target.
5823 * config/rs6000/750cl.h: Delete.
5824 * config/rs6000/paired.h: Delete.
5825 * config/rs6000/paired.md: Delete.
5826 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
5828 * config/rs6000/rs6000-builtin.def: Remove paired float support.
5829 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
5830 comment. Remove paired float support.
5831 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
5832 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
5834 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
5835 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
5837 * config/rs6000/rs6000.c: Remove paired float support.
5838 (paired_expand_vector_init, paired_expand_vector_move,
5839 paired_emit_vector_compare, paired_emit_vector_cond_expr,
5840 (paired_expand_lv_builtin, paired_expand_stv_builtin,
5841 paired_expand_builtin, paired_expand_predicate_builtin,
5842 paired_init_builtins): Delete.
5843 * config/rs6000/rs6000.h: Remove paired float support.
5844 * config/rs6000/rs6000.md: Remove paired float support.
5845 (move_from_CR_ov_bit): Delete.
5846 * config/rs6000/rs6000.opt (mpaired): Delete.
5847 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
5848 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
5850 2018-05-02 Richard Biener <rguenther@suse.de>
5853 * gimplify.c (gimplify_save_expr): When in SSA form allow
5854 SAVE_EXPRs to compute to SSA vars.
5856 2018-05-02 Jakub Jelinek <jakub@redhat.com>
5859 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
5860 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
5861 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
5862 clobber operands[2], instead use a new pseudo. Formatting fixes.
5864 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
5866 PR tree-optimization/85586
5867 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
5868 exit early for statements in the same group if the accesses are
5871 2018-05-02 Tom de Vries <tom@codesourcery.com>
5874 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
5877 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
5879 PR tree-optimization/85143
5880 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
5882 2018-05-01 Tom de Vries <tom@codesourcery.com>
5885 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
5886 not found" error message.
5888 2018-05-01 Tom de Vries <tom@codesourcery.com>
5891 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
5892 * vec.c (test_ordered_remove_if): New function.
5893 (vec_c_tests): Call test_ordered_remove_if.
5894 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
5895 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
5896 * tree-vect-patterns.c (vect_pattern_recog_1): Use
5897 VEC_ORDERED_REMOVE_IF.
5899 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5901 PR tree-optimization/82665
5902 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
5903 pointer subtraction where arguments come from a memchr call.
5905 2018-05-01 Jakub Jelinek <jakub@redhat.com>
5907 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
5908 --push-state --as-needed and --pop-state instead of --as-needed and
5909 --no-as-needed if ld supports it.
5910 * configure: Regenerated.
5913 * doc/install.texi2html: Replace _002d with - and _002a with * in
5914 generated html files using sed.
5916 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5919 * gcc-rich-location.c (blank_line_before_p): New function.
5920 (use_new_line): New function.
5921 (gcc_rich_location::add_fixit_insert_formatted): New function.
5922 * gcc-rich-location.h
5923 (gcc_rich_location::add_fixit_insert_formatted): New function.
5925 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5927 * selftest.c (assert_streq): Rename "expected" and "actual" to
5928 "val1" and "val2". Extend NULL-handling to cover both inputs
5929 symmetrically, while still requiring both to be non-NULL for a pass.
5930 * selftest.h (assert_streq): Rename "expected" and "actual" to
5932 (ASSERT_EQ): Likewise.
5933 (ASSERT_EQ_AT): Likewise.
5934 (ASSERT_KNOWN_EQ): Likewise.
5935 (ASSERT_KNOWN_EQ_AT): Likewise.
5936 (ASSERT_NE): Likewise.
5937 (ASSERT_MAYBE_NE): Likewise.
5938 (ASSERT_MAYBE_NE_AT): Likewise.
5939 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
5940 the assertion to pass.
5941 (ASSERT_STREQ_AT): Likewise.
5943 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
5945 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
5946 interaction with -pie.
5948 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5950 * selftest.h: Fix alphabetization of per-source-file selftest
5953 2018-04-30 Jason Merrill <jason@redhat.com>
5955 PR c++/61982 - dead stores to destroyed objects.
5956 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
5959 2018-04-30 Jason Merrill <jason@redhat.com>
5961 * tree.c (build_clobber): New.
5962 * tree.h: Declare it.
5963 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
5965 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5967 * diagnostic-show-locus.c (layout::layout): Update for
5968 location_get_source_line returning a char_span.
5969 (struct char_span): Move to input.h.
5970 (struct correction): Update for fields in char_span becoming
5972 (struct source_line): Update for location_get_source_line
5973 returning a char_span.
5974 (layout::print_line): Likewise.
5975 * edit-context.c (edited_file::print_content): Likewise.
5976 (edited_file::print_diff_hunk): Likewise.
5977 (edited_file::print_run_of_changed_lines): Likewise.
5978 (edited_file::get_num_lines): Likewise.
5979 (edited_line::edited_line): Likewise.
5980 * final.c (asm_show_source): Likewise.
5981 * input.c (location_get_source_line): Convert return type
5982 from const char * to char_span, losing the final "line_len"
5984 (dump_location_info): Update for the above.
5985 (get_substring_ranges_for_loc): Likewise. Use a char_span
5986 when handling the literal within the line.
5987 (test_reading_source_line): Update for location_get_source_line
5988 returning a char_span.
5989 * input.h (class char_span): Move here from
5990 diagnostic-show-locus.c, converting from a struct to a class.
5991 Make data members private.
5992 (char_span::operator bool): New.
5993 (char_span::length): New.
5994 (char_span::get_buffer): New.
5995 (char_span::operator[]): New.
5996 (char_span::subspan): Make const.
5997 (char_span::xstrdup): New.
5998 (location_get_source_line): Convert return type from const char *
5999 to char_span, losing the final "line_size" param.
6001 2018-04-30 Jan Hubicka <jh@suse.cz>
6003 * lto-wrapper.c (ltrans_priorities): New static var.
6004 (cmp_priority): New.
6005 (run_gcc): Read priorities and if doing parallel build order
6006 the Makefile by them.
6008 2018-04-30 David Malcolm <dmalcolm@redhat.com>
6010 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
6012 2018-04-30 Richard Biener <rguenther@suse.de>
6014 * tree-cfg.c (verify_address): Remove base argument, add
6015 flag whether to check TREE_ADDRESSABLE and do that.
6016 (verify_expr): Remove.
6017 (verify_types_in_gimple_reference): Add pieces from verify_expr.
6018 (verify_gimple_assign_single): Likewise.
6019 (verify_gimple_switch): Likewise.
6020 (verify_expr_location_1): Dereference tp once. Add (disabled)
6021 piece from verify_expr.
6022 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
6024 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
6026 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
6028 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
6030 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
6031 (small_data_pattern): Likewise.
6032 (arc_rewrite_small_data): Likewise.
6033 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
6034 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
6035 (get_symbol_alignment): New function.
6036 (legitimate_small_data_address_p): Likewise.
6037 (legitimate_scaled_address): Update, call
6038 legitimate_small_data_address_p.
6039 (output_sdata): New static variable.
6040 (arc_print_operand): Update how we handle small data operands.
6041 (arc_print_operand_address): Likewise.
6042 (arc_legitimate_address_p): Update, use
6043 legitimate_small_data_address_p.
6044 (arc_rewrite_small_data_p): Remove.
6045 (arc_rewrite_small_data_1): Likewise.
6046 (arc_rewrite_small_data): Likewise.
6047 (small_data_pattern): Likewise.
6048 (compact_sda_memory_operand): Update to use
6049 legitimate_small_data_address_p and get_symbol_alignment.
6050 (prepare_move_operands): Don't rewite sdata pattern.
6051 (prepare_extend_operands): Remove.
6052 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
6054 (zero_extendqisi2): Likewise.
6055 (zero_extendhisi2): Likewise.
6056 (extendqihi2): Likewise.
6057 (extendqisi2): Likewise.
6058 (extendhisi2): Likewise.
6062 * config/arc/constraints.md (Usd): Change it to memory constraint.
6064 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
6066 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
6067 as source of std instructions.
6068 * config/arc/arc.md (movsi_insn): Update pattern predicate to
6069 allow 6-bit constants as source for store instructions.
6070 (movdi_insn): Update instruction pattern to allow 6-bit constants
6071 as source for store instructions.
6073 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
6075 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
6077 2018-04-30 Nathan Sidwell <nathan@acm.org>
6078 Sandra Loosemore <sandra@codesourcery.com>
6080 * dumpfile.c (dump_open): Allow '-' for stdout.
6081 * doc/invoke.texi (Developer Options): Document dump filename
6082 determination early. Document stdin/stdout selection.
6084 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
6086 Microblaze Target: PIC data text relative
6088 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
6089 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
6091 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
6092 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
6093 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
6094 New addressing mode for data-text relative position indepenedent code.
6095 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
6096 'ADDRESS_SYMBOLIC_TXT_REL'.
6097 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
6098 (microblaze_legitimate_pic_operand): Exclude function calls from
6099 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
6100 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
6102 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
6103 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
6104 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
6105 for 'address + offset'.
6106 (microblaze_expand_prologue): Add new function prologue call for
6108 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
6109 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
6110 table in case of TARGET_PIC_DATA_TEXT_REL.
6111 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
6112 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
6113 Add new macros 'UNSPEC_TEXT',
6114 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
6115 + exclude function calls from 'UNSPEC_PLT' in case of data text
6117 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
6118 new target hook for generating address diff vector tables in case of
6120 * doc/tm.texi : Regenerate.
6121 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
6122 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
6123 of addr diff vector generation.
6124 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
6125 target hook definition.
6126 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
6127 Add default function for generate_pic_addr_diff_vec -> flag_pic.
6128 * doc/invoke.texi (Add new pic option): Add new microblaze pic
6129 option for data text relative.
6131 2018-04-30 Richard Biener <rguenther@suse.de>
6133 * tree-chrec.h (evolution_function_is_constant_p): Remove
6135 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
6137 2018-04-30 Richard Biener <rguenther@suse.de>
6140 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
6142 2018-04-30 Richard Biener <rguenther@suse.de>
6144 PR tree-optimization/28364
6145 PR tree-optimization/85275
6146 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
6147 copying first exit test.
6149 2018-04-28 Mark Wielaard <mark@klomp.org>
6151 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
6153 (dwarf_AT): Handle DW_AT_addr_base.
6154 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
6156 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
6159 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
6160 (*ashl<dwi>3_doubleword_mask_1): Ditto.
6161 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
6162 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
6164 2018-04-28 Richard Biener <rguenther@suse.de>
6166 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
6167 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
6168 to reflect use. Only add interesting stmts.
6170 2018-04-27 Martin Jambor <mjambor@suse.cz>
6173 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
6174 the jump function allows for passing through aggregate values.
6176 2018-04-27 David Malcolm <dmalcolm@redhat.com>
6178 * input.h (in_system_header_at): Convert from macro to inline
6180 (from_macro_expansion_at): Likewise.
6181 (from_macro_definition_at): Likewise.
6183 2018-04-27 Jeff Law <law@redhat.com>
6185 * config.gcc: Mark tile* targets as deprecated/obsolete.
6187 2018-04-27 Richard Biener <rguenther@suse.de>
6189 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
6192 2018-04-27 Richard Biener <rguenther@suse.de>
6194 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
6196 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
6198 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
6199 with Yd constraint. Set "preferred_for_speed" attribute from
6200 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
6202 (*movdi_internal): Ditto.
6203 (movti_interunit splitters): Remove
6204 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
6205 (movdi_interunit splitters): Ditto.
6206 * config/i386/constraints.md (Ye): Remove.
6207 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
6209 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6212 * config/aarch64/constraints.md (Usg): Limit to 31.
6215 2018-04-27 Jakub Jelinek <jakub@redhat.com>
6217 PR tree-optimization/85529
6218 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
6219 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
6220 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
6221 zero extension or masking of the MSB bit.
6222 (optimize_range_tests): Add FIRST_BB argument, pass it through
6223 to optimize_range_tests_var_bound.
6224 (maybe_optimize_range_tests, reassociate_bb): Adjust
6225 optimize_range_tests callers.
6227 2018-04-26 Richard Biener <rguenther@suse.de>
6228 Jakub Jelinek <jakub@redhat.com>
6230 * cgraph.h (symbol_table): Just declare debug method here.
6231 * symtab.c (symbol_table::debug): Define.
6233 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
6235 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
6237 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
6239 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
6240 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
6241 (*movdi_internal): Substitute Yi and Yj constraint with x
6242 and Ym and Yn constraint with y constraint. Update "isa"
6243 attribute and set "preferred_for_speed" attribute from
6244 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
6245 (*movsi_internal): Ditto.
6246 (*movdf_internal): Ditto.
6247 (*movsf_internal): Ditto.
6248 (*zero_extendsidi2): Ditto.
6249 * config/i386/sse.md (vec_set<mode>_0): Ditto.
6250 (sse2_loadld): Ditto.
6251 (*vec_extract<ssevecmodelower>_0): Ditto.
6252 (*vec_extractv4si_0_zext_sse4): Ditto.
6253 (vec_concatv2di): Ditto.
6254 (*vec_dup<mode>): Ditto.
6255 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
6256 * config/i386/constraints.md (Yi): Remove.
6261 2018-04-26 Nathan Sidwell <nathan@acm.org>
6263 * dumpfile.c (dump_open): New.
6264 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
6265 (dump_finish): Detect stdio/stderr by value not name.
6267 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
6269 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
6271 2018-04-26 Tom de Vries <tom@codesourcery.com>
6274 * config/nvptx/nvptx.c (verify_neutering_jumps)
6275 (verify_neutering_labels): New function
6276 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
6278 2018-04-26 Tom de Vries <tom@codesourcery.com>
6281 * config/nvptx/nvptx.c (needs_neutering_p): New function.
6282 (nvptx_single): Use needs_neutering_p to skip over insns that do not
6285 2018-04-26 Richard Biener <rguenther@suse.de>
6286 Tom de Vries <tom@codesourcery.com>
6289 * lto-streamer-out.c (output_function): Fixup loops if required to match
6290 discovery done in the reader.
6292 2018-04-26 Richard Biener <rguenther@suse.de>
6294 PR tree-optimization/85116
6295 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
6296 have a loop exit from the single latch predecessor. Remove
6297 case of header with just condition.
6298 (ch_base::copy_headers): Exclude infinite loops from any
6300 (pass_ch::execute): Record exits.
6302 2018-04-26 Richard Biener <rguenther@suse.de>
6304 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
6305 prologue cost vector and pass it to vect_get_load_cost.
6306 (vect_get_peeling_costs_all_drs): Likewise.
6307 (vect_peeling_hash_get_lowest_cost): Likewise.
6308 (vect_enhance_data_refs_alignment): Likewise.
6310 2018-04-26 Richard Biener <rguenther@suse.de>
6313 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
6314 checking of integer<->pointer conversions.
6315 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
6316 sign-/zero-extending pointer types.
6317 (expand_omp_for_static_chunk): Likewise.
6319 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
6320 Jean Lee <xiaoyur347@gmail.com>
6322 * config/mips/mips.c (mips_asan_shadow_offset): New function.
6323 (TARGET_ASAN_SHADOW_OFFSET): Define.
6324 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
6325 true for -fsanitize=address.
6327 2018-04-25 Mark Wielaard <mark@klomp.org>
6329 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
6332 2018-04-25 Jakub Jelinek <jakub@redhat.com>
6334 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
6335 than "alu", remove explicit "memory" and "imm_disp" attributes.
6336 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
6339 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
6340 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
6341 gen_lowpart_no_emit.
6343 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
6346 * config/i386/i386.c (ix86_expand_builtin): Change memory
6347 operand to XI, extend p0 to Pmode.
6348 * config/i386/i386.md: Change unspec volatile and operand
6349 1 mode to XI, change operand 0 mode to P.
6351 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
6353 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
6354 GET_MODE_MASK before any checking.
6355 (nds32_can_use_bset_p): Likewise.
6356 (nds32_can_use_btgl_p): Likewise.
6358 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
6360 * config/nds32/nds32-doubleword.md: New define_split pattern for
6361 illegal register number.
6363 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
6365 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
6367 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
6369 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
6371 2018-04-25 Richard Biener <rguenther@suse.de>
6373 * lto-streamer.h (LTO_major_version): Bump to 8.
6375 2018-04-25 Jakub Jelinek <jakub@redhat.com>
6377 * BASE-VER: Set to 9.0.0.
6379 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
6381 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
6382 in __abskf2 and __powikf2.
6384 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6387 * config/aarch64/constraints.md (Usg, Usj): New constraints.
6388 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
6389 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
6390 Use the above on operand 2. Reindent.
6391 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
6393 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
6396 * common/config/i386/i386-common.c (ix86_handle_option): Don't
6398 * config/i386/i386.opt (mcet): Removed.
6399 * doc/install.texi: Remove -mcet documentation.
6400 * doc/invoke.texi: Likewise.
6402 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
6405 * doc/install.texi: Remove -mcet from bootstrap-cet.
6407 2018-04-24 Jakub Jelinek <jakub@redhat.com>
6410 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
6411 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
6415 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
6416 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
6417 containing a CONST_VECTOR.
6419 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
6421 * doc/install.texi: Update newlib dependency for nvptx.
6423 2018-04-24 Jakub Jelinek <jakub@redhat.com>
6426 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
6427 instead of INTVAL when shifting x left.
6429 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
6431 PR tree-optimization/85478
6432 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
6433 vect_grouped_store_supported for single element vectors.
6435 2018-04-24 Richard Biener <rguenther@suse.de>
6438 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
6439 load cost increase to the case of non-constant step.
6441 2018-04-24 Jakub Jelinek <jakub@redhat.com>
6444 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
6445 destination if any_malformed_asm.
6447 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
6450 * expr.c (store_field): In the bitfield case, if the value comes from
6451 a function call and is returned in registers by means of a PARALLEL,
6452 do not change the mode of the temporary unless BLKmode and VOIDmode.
6454 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
6456 PR rtl-optimization/85423
6457 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
6458 dependencies to debug insns when the previous insn is non-debug.
6460 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
6462 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
6463 enums into a single definition.
6464 (fls): Fix predicates and printing.
6467 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
6469 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
6470 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
6471 and short u6 immediate.
6472 (check_if_valid_sleep_operand): Remove.
6473 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
6475 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
6477 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
6478 flag_always_save_lp condition.
6479 * config/nds32/nds32.opt (malways-save-lp): New option.
6481 2018-04-22 Shiva Chen <shiva0217@gmail.com>
6483 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
6484 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
6485 * config/nds32/nds32.h
6486 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
6487 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
6489 2018-04-22 Shiva Chen <shiva0217@gmail.com>
6491 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
6492 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
6494 2018-04-22 Shiva Chen <shiva0217@gmail.com>
6495 Chung-Ju Wu <jasonwucj@gmail.com>
6497 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
6499 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
6500 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
6502 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
6504 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
6506 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
6508 * config/nds32/nds32-protos.h (nds32_data_alignment,
6509 nds32_local_alignment): Declare.
6510 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
6511 nds32_local_alignment): New functions.
6512 (TARGET_CONSTANT_ALIGNMENT): Define.
6513 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
6515 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
6517 * config/nds32/nds32.c
6518 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
6519 (TARGET_MODES_TIEABLE_P): Likewise.
6521 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
6523 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
6526 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
6527 Chung-Ju Wu <jasonwucj@gmail.com>
6529 * config/nds32/constants.md (unspec_volatile_element): Add enum values
6530 for unaligned access.
6531 * config/nds32/nds32-intrinsic.c: Implementation of expanding
6533 * config/nds32/nds32-intrinsic.md: Likewise.
6534 * config/nds32/nds32_intrinsic.h: Likewise.
6535 * config/nds32/nds32.h (nds32_builtins): Likewise.
6536 * config/nds32/nds32.opt (munaligned-access): New option.
6537 * config/nds32/nds32.c (nds32_asm_file_start): Display
6538 flag_unaligned_access status.
6540 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
6542 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
6543 -mno-relax is present.
6544 * config/riscv/linux.h (LINK_SPEC): Ditto.
6546 2018-04-20 Martin Sebor <msebor@redhat.com>
6549 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
6551 (gimple_fold_builtin_stxcpy_chk): Same.
6552 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
6554 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
6557 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
6558 __powikf2 when long double is IEEE 128-bit.
6560 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
6562 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
6563 step to make sure stack always aligned.
6565 2018-04-20 Carl Love <cel@us.ibm.com>
6568 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
6569 size check for arg0.
6571 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
6572 Tom de Vries <tom@codesourcery.com>
6575 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
6576 Emit insns for calls too.
6577 (nvptx_find_par): Always look for worker-level predecessor insn.
6578 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
6580 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
6581 (nvptx_process_pars): Propagate frames for calls.
6583 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
6586 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
6588 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
6589 (ix86_handle_option): Don't handle OPT_mibt.
6590 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
6592 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
6594 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
6595 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
6596 (ix86_target_macros): Define __CET__ with flag_cf_protection
6597 for -fcf-protection.
6598 * config/i386/i386.c (isa2_opts): Remove -mibt.
6599 * config/i386/i386.h (TARGET_IBT): Removed.
6600 (TARGET_IBT_P): Likewise.
6601 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
6602 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
6603 * config/i386/i386.opt (mcet): Update help message.
6606 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
6607 -mcet as an alias for -mshstk.
6609 2018-04-20 Richard Biener <rguenther@suse.de>
6612 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
6613 complexity by forcing a single use of the multiply operand.
6615 2018-04-20 Martin Jambor <mjambor@suse.cz>
6618 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
6619 recursion dependency to only apply to non-clones.
6621 2018-04-20 Martin Jambor <mjambor@suse.cz>
6624 * ipa-cp.c (create_specialized_node): Check that clones of
6625 self-recursive edges exist during IPA-CP.
6627 2018-04-19 Toon Moene <toon@moene.org>
6629 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
6632 2018-04-19 Jakub Jelinek <jakub@redhat.com>
6634 PR tree-optimization/85467
6635 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
6636 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
6637 VECTOR_CST element to type.
6639 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
6642 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
6643 * config/i386/i386.md (builtin_setjmp_setup): Removed.
6644 (builtin_longjmp): Likewise.
6645 (save_stack_nonlocal): New pattern.
6646 (restore_stack_nonlocal): Likewise.
6648 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
6651 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
6652 Replace ASM_OUTPUT_LABEL with fprintf.
6654 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
6657 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
6658 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
6659 * config/i386/i386-c.c (ix86_target_macros_internal): Also
6660 define __IBT__ and __SHSTK__ for -fcf-protection.
6661 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
6663 (ix86_trampoline_init): Likewise.
6664 (x86_output_mi_thunk): Likewise.
6665 (ix86_notrack_prefixed_insn_p): Likewise.
6666 (ix86_option_override_internal): Don't disallow -fcf-protection.
6667 * config/i386/i386.md (rdssp<mode>): Also enable for
6669 (incssp<mode>): Likewise.
6670 (nop_endbr): Likewise.
6671 * config/i386/i386.opt (mcet): Change help message to built-in
6675 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
6676 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
6677 enable CET built-in functions.
6679 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
6681 * common/config/i386/i386-common.c
6682 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
6683 OPTION_MASK_ISA_MOVDIRI_UNSET,
6684 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
6685 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
6686 * config.gcc (movdirintrin.h): New header.
6687 * config/i386/cpuid.h (bit_MOVDIRI,
6688 bit_MOVDIR64B): New bits.
6689 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
6691 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
6692 (VOID, PVOID, PCVOID)): New function types.
6693 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
6694 __builtin_ia32_directstoreu_u64,
6695 __builtin_ia32_movdir64b): New builtins.
6696 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
6697 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
6699 (ix86_valid_target_attribute_inner_p): Ditto.
6700 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
6701 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
6702 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
6703 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
6704 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
6705 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
6706 (movdiri<mode>, movdir64b_<mode>): New.
6707 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
6708 * config/i386/immintrin.h: Include movdirintrin.h.
6709 * config/i386/movdirintrin.h: New file.
6710 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
6712 2018-04-19 Richard Biener <rguenther@suse.de>
6715 * cfg.c (clear_bb_flags): When loop state says we have
6716 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
6718 2018-04-19 Richard Biener <rguenther@suse.de>
6720 PR tree-optimization/84737
6721 * tree-vect-data-refs.c (vect_copy_ref_info): New function
6722 copying restrict info.
6723 (vect_setup_realignment): Use it.
6724 * tree-vectorizer.h (vect_copy_ref_info): Declare.
6725 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
6726 the first DR to all generated stores.
6727 (vectorizable_load): Likewise for loads.
6729 2018-04-19 Jakub Jelinek <jakub@redhat.com>
6731 PR tree-optimization/85446
6732 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
6733 the integral and pointer types to have the same precision.
6735 * doc/install.texi: Document --disable-cet being the default and
6738 2018-04-18 Martin Liska <mliska@suse.cz>
6740 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
6743 2018-04-18 Martin Liska <mliska@suse.cz>
6746 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
6749 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
6750 arguments if they are comparable.
6752 2018-04-18 Martin Liska <mliska@suse.cz>
6755 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
6758 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
6761 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
6764 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
6765 ENDBR after calling __morestack.
6767 2018-04-18 David Malcolm <dmalcolm@redhat.com>
6770 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
6771 by using gcc_base_ver to generate a gcc_driver_version, and use
6772 it when generating GCC_DRIVER_NAME.
6773 * configure: Regenerate.
6775 2018-04-18 Jakub Jelinek <jakub@redhat.com>
6778 * config.gcc: Obsolete powerpc*-*-*spe*.
6780 2018-04-17 Jakub Jelinek <jakub@redhat.com>
6783 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
6784 (stabstr_D): Change type of unum from unsigned int to
6785 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
6788 2018-04-17 Jim Wilson <jimw@sifive.com>
6791 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
6792 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
6793 Set arg_pointer_offset after using pretend_args_size.
6795 2018-04-17 Jakub Jelinek <jakub@redhat.com>
6797 PR rtl-optimization/85431
6798 * dse.c (record_store): Ignore zero width stores.
6801 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
6802 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
6803 __builtin_stack_restore rather than after it.
6804 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
6805 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
6806 argument instead of virtual_dynamic_stack_rtx.
6808 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
6810 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
6812 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
6813 Add note to error message to explain internal mapping of overloaded
6814 built-in function name to non-overloaded built-in function name.
6815 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
6818 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
6821 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
6822 where the inputs overlap with the output.
6824 2018-04-17 Jakub Jelinek <jakub@redhat.com>
6826 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
6827 (=v, v) alternative and explicit "memory" attribute.
6828 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
6829 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
6831 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
6832 "sselog1" type instead of "sselog".
6833 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
6834 "sselog". Remove explicit "memory" attribute.
6835 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
6836 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
6838 (vec_extract_hi_v32hi): Merge all alternatives into one, use
6839 "sselog1" type instead of "sselog". Remove explicit "memory"
6841 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
6842 use "sselog1" type instead of "sselog". Remove explicit "memory"
6844 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
6845 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
6847 (vec_extract_hi_v64qi): Merge all alternatives into one, use
6848 "sselog1" type instead of "sselog". Remove explicit "memory"
6850 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
6851 use "sselog1" type instead of "sselog". Remove explicit "memory"
6855 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
6858 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
6861 2018-04-17 Martin Jambor <mjambor@suse.cz>
6864 * ipa-cp.c (create_specialized_node): Call
6865 expand_all_artificial_thunks if necessary.
6867 2018-04-17 Martin Liska <mliska@suse.cz>
6870 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
6871 in message, remote space in between '_G' and '('.
6873 2018-04-17 Jakub Jelinek <jakub@redhat.com>
6876 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
6877 avx512f_vmcmp<mode>3<round_saeonly_name>,
6878 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
6879 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
6880 avx512f_rndscale<mode><round_saeonly_name>,
6881 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
6882 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
6883 Use %<iptr>2 instead of %2 for -masm=intel.
6884 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
6885 avx512f_vcvttss2usi<round_saeonly_name>,
6886 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
6888 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
6889 avx512f_vcvttsd2usi<round_saeonly_name>,
6890 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
6891 Use %q1 instead of %1 for -masm=intel.
6892 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
6893 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
6894 of %3 for -masm=intel.
6895 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
6897 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
6899 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
6901 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
6902 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
6904 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
6906 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
6907 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
6908 %g1 and one with %0 and %1.
6909 (avx512er_vmrcp28<mode><round_saeonly_name>,
6910 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
6912 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
6913 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
6914 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
6915 of %0 and %{%4%} for -masm=intel.
6916 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
6917 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
6918 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
6919 order of %0 and %{%5%}%{z%} for -masm=intel.
6921 2018-04-17 Jan Hubicka <jh@suse.cz>
6924 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
6926 2018-04-17 Martin Liska <mliska@suse.cz>
6929 * multiple_target.c (create_dispatcher_calls): Set apostrophes
6930 for target_clone error message. Make default implementation
6931 clone to be a local declaration.
6932 (separate_attrs): Add new argument and check for an empty
6934 (expand_target_clones): Handle it.
6935 (ipa_target_clone): Make redirection just for target_clones
6938 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
6939 Tom de Vries <tom@codesourcery.com>
6942 * omp-expand.c (expand_oacc_for): Add dummy false branch for
6943 tiled basic blocks without omp continue statements.
6945 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
6948 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
6949 vec_extract expression as having side effects to make sure it gets
6952 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
6955 * config/i386/i386.c (get_builtin_code_for_version): Check
6958 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
6961 * gcc/config.gcc: Support "skylake".
6962 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
6964 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
6965 (processor_target_table): Add "skylake".
6966 (ix86_option_override_internal): Add "skylake".
6967 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
6968 PROCESSOR_CANNONLAKE.
6969 (get_builtin_code_for_version): Fix priority for
6970 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
6971 PROCESSOR_SKYLAKE-AVX512.
6972 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
6973 (processor_type): Add PROCESSOR_SKYLAKE.
6975 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
6976 Jason Merrill <jason@redhat.com>
6979 * convert.c (convert_to_integer_1): Use direct recursion for
6980 enumeral types and types with a precision less than the number
6981 of bits in their mode.
6983 2018-04-16 Julia Koval <julia.koval@intel.com>
6986 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
6987 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
6989 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
6992 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
6993 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
6994 and -mno-direct-move.
6996 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
6999 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
7000 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
7001 Ensure negative shifts result in {0}.
7003 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
7005 PR rtl-optimization/79916
7006 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
7007 regs (if any) to define how to gnerate SD moves when LRA is in
7010 2018-04-13 Jakub Jelinek <jakub@redhat.com>
7012 PR rtl-optimization/85393
7013 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
7014 * except.c (expand_dw2_landing_pad_for_region): Make static.
7015 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
7016 a label and unconditional jump to old_bb, rather than
7017 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
7020 PR rtl-optimization/85376
7021 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
7022 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
7023 instead of a specific value.
7025 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
7026 Bin Cheng <bin.cheng@arm.com>
7028 PR tree-optimization/82965
7029 PR tree-optimization/83991
7030 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
7031 by_profile_only parameter.
7032 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
7033 information if the loop was predicted to iterate too many times.
7034 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
7036 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
7039 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
7042 2018-04-13 Martin Liska <mliska@suse.cz>
7043 Jakub Jelinek <jakub@redhat.com>
7046 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
7047 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
7048 * builtins.c (expand_builtin_memory_copy_args): Use
7049 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
7050 handle dest_addr == pc_rtx.
7052 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
7055 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
7056 asked to not generate direct moves.
7057 (fix_trunc<mode>si2_stfiwx): Similar.
7058 (fix_trunc<mode>si2_internal): Similar.
7060 2018-04-12 Jakub Jelinek <jakub@redhat.com>
7063 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
7064 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
7065 lookup if dest in some wider mode is known to be const0_rtx and
7066 if so, record permanent equivalence for it to be ZERO_EXTEND of
7067 the narrower mode destination.
7069 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
7071 * lto-streamer-out.c (output_function): Revert 259346.
7072 * omp-expand.c (expand_oacc_for): Likewise.
7074 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
7076 PR rtl-optimization/85354
7077 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
7078 * sel-sched.c (sel_global_init): ... here.
7080 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
7083 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
7084 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
7085 mode for PE-COFF targets.
7086 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
7087 (i386_pe_asm_lto_end): Likewise.
7088 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
7089 (TARGET_ASM_LTO_END): Likewise.
7090 * config/i386/winnt.c (saved_debug_info_level): New static variable.
7091 (i386_pe_asm_lto_start): New function.
7092 (i386_pe_asm_lto_end): Likewise.
7094 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
7095 Richard Biener <rguenther@suse.de>
7098 * lto-streamer-out.c (output_function): Fix CFG loop state before
7100 * omp-expand.c (expand_oacc_for): Handle calls to internal
7101 functions like regular functions.
7103 2018-04-12 Richard Biener <rguenther@suse.de>
7106 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
7107 for the early LTO debug to properly generate references to it
7108 during DIE emission. Do not re-use that for the skeleton for
7110 (dwarf2out_early_finish): Likewise.
7112 2018-04-12 Jakub Jelinek <jakub@redhat.com>
7115 * config/i386/sse.md
7116 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
7117 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
7118 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
7119 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
7120 and output is a reg, avoid creating invalid lowpart subreg, but
7121 instead split into a 512-bit move. Don't split if not AVX512VL,
7122 input is xmm16+ reg and output is a mem.
7123 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
7124 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
7125 xmm16+ reg and output is a mem.
7127 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7129 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
7130 also for flag_dwarf2_cfi_asm.
7132 2018-04-12 Jakub Jelinek <jakub@redhat.com>
7134 PR rtl-optimization/85342
7135 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
7136 a bool scalar var inside of the loop instead. Don't try to update
7137 recog_data.operand after failed apply_change_group.
7139 2018-04-12 Tom de Vries <tom@codesourcery.com>
7142 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
7143 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
7144 array with flexible array member as array without given dimension.
7145 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
7146 argument for undefined param to true.
7148 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
7151 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
7152 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
7153 from PowerPC section.
7154 * config/rs6000/sysv4.opt (mcall-): Improve help text.
7155 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
7156 help text that is too long.
7157 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
7158 help text that is too long.
7159 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
7160 help text that is too long.
7162 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
7164 * config/alpha/alpha.md (stack_probe_internal): Rename
7165 from "probe_stack". Update all callers.
7167 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
7169 PR rtl-optimization/84566
7170 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
7171 sched_macro_fuse_insns.
7173 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
7176 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
7177 (compute_block_dependences): ... from here.
7179 2018-04-11 Jakub Jelinek <jakub@redhat.com>
7181 PR tree-optimization/85331
7182 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
7183 from int to HOST_WIDE_INT.
7185 2018-04-11 Martin Jambor <mjambor@suse.cz>
7188 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
7189 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
7190 not the same as the source val.
7191 (cgraph_edge_brings_value_p): New parameter.
7192 (gather_edges_for_value): Pass destination value to
7193 cgraph_edge_brings_value_p.
7194 (perhaps_add_new_callers): Likewise.
7195 (get_info_about_necessary_edges): Likewise and exclude values brought
7196 only by self-recursive edges.
7197 (create_specialized_node): Redirect only clones of self-calling edges.
7198 (+self_recursive_pass_through_p): New function.
7199 (find_more_scalar_values_for_callers_subset): Use it.
7200 (find_aggregate_values_for_callers_subset): Likewise.
7201 (known_aggs_to_agg_replacement_list): Removed.
7202 (decide_whether_version_node): Re-calculate known constants for all
7203 remaining context clones.
7205 2018-04-11 Richard Biener <rguenther@suse.de>
7208 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
7209 from early DWARF output.
7210 (dwarf2out_early_finish): Output line info unconditionally into
7211 early DWARF and add reference to it.
7213 2018-04-11 Jakub Jelinek <jakub@redhat.com>
7216 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
7217 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
7218 other than V2DFmode using iptr mode attribute.
7219 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
7221 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
7223 PR rtl-optimization/84659
7224 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
7226 2018-04-11 Jakub Jelinek <jakub@redhat.com>
7229 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
7231 (output_loc_list): Pass address of a dummy size variable even in the
7232 locview handling loop.
7233 (index_location_lists): Add comment on why skip_loc_list_entry can't
7236 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
7239 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
7242 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
7245 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
7246 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
7247 and -mstring-compare-inline-limit.
7249 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
7252 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
7253 for stack clash protection in a register whenever we need it to be in
7256 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
7258 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
7259 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
7261 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
7264 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
7266 (mlong-double-): Ditto.
7267 * config/rs6000/sysv4.opt (msdata=): Ditto.
7268 (mtls-size=): Ditto.
7270 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
7272 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
7273 erroneous entries for
7274 "vector int vec_ldl (int, long int *)", and
7275 "vector unsigned int vec_ldl (int, unsigned long int *)".
7276 Add comments and entries for
7277 "vector bool char vec_ldl (int, bool char *)",
7278 "vector bool short vec_ldl (int, bool short *)",
7279 "vector bool int vec_ldl (int, bool int *)",
7280 "vector bool long long vec_ldl (int, bool long long *)",
7281 "vector pixel vec_ldl (int, pixel *)",
7282 "vector long long vec_ldl (int, long long *)",
7283 "vector unsigned long long vec_ldl (int, unsigned long long *)".
7284 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
7285 type tree bool_long_long_type_node and correct definition of
7286 bool_V2DI_type_node to make reference to this new type tree.
7287 (rs6000_mangle_type): Replace erroneous reference to
7288 bool_long_type_node with bool_long_long_type_node.
7289 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
7290 comments to emphasize sign distinctions for char and int types and
7291 replace RS6000_BTI_bool_long constant with
7292 RS6000_BTI_bool_long_long constant. Also add comment to restrict
7293 use of RS6000_BTI_pixel.
7294 (bool_long_type_node): Remove this macro definition.
7295 (bool_long_long_type_node): New macro definition
7297 2018-04-10 Jakub Jelinek <jakub@redhat.com>
7299 PR rtl-optimization/85300
7300 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
7301 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
7302 simplify_unary_operation fails.
7304 2018-04-10 Martin Liska <mliska@suse.cz>
7306 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
7307 cgraph_edge and ipa_ref.
7309 2018-04-10 Jakub Jelinek <jakub@redhat.com>
7313 * config/i386/sse.md
7314 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
7315 computation of the VEC_MERGE selector from mask.
7316 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
7317 Fix decoding of the VEC_MERGE selector into mask.
7319 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
7321 PR tree-optimization/85286
7322 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
7324 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
7326 * final.c (final_1): Set insn_last_address as well as
7327 insn_current_address.
7329 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7332 * explow.c (emit_stack_probe): Call validize_mem on memory location
7333 before passing it to gen_probe_stack. Create address operand and
7334 legitimize it for the probe_stack_address case.
7336 2018-04-09 Jan Hubicka <jh@suse.cz>
7339 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
7340 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
7341 * tree.c (free_lang_data_in_type): Fix handling of binfos;
7343 (free_lang_data): Rebuild type inheritance graph.
7345 2018-04-09 Martin Sebor <msebor@redhat.com>
7347 * invoke.texi (-finline-small-functions): Mention other optimization
7349 (-findirect-inlining, -fpartial-inlining): Same.
7350 (-finline-functions-called-once): Same.
7351 (-freorder-blocks-and-partition): Same.
7353 2018-04-09 Jan Hubicka <jh@suse.cz>
7356 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
7357 jumps; choose last target that matches the criteria (i.e.
7358 no partition changes for non-crossing jumps).
7359 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
7360 support for redirecting crossing jumps to non-crossing.
7362 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
7364 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
7365 also for naked functions.
7367 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
7369 * config/arc/arc.md (add_shift): New pattern.
7370 (add_shift2): Likewise.
7371 (sub_shift): Likewise.
7372 (sub_shift_cmp0_noout): Likewise.
7373 (compare_si_ashiftsi): Likewise.
7374 (xbfu_cmp0_noout): New combine pattern.
7375 (xbfu_cmp0"): Likewise.
7376 (movsi_set_cc_insn): Place the predicable variant first.
7377 (commutative_binary_cmp0_noout): Remove clobber.
7378 (commutative_binary_cmp0): New pattern.
7379 (noncommutative_binary_cmp0): Likewise.
7380 (noncommutative_binary_cmp0_noout): Likewise.
7381 (noncommutative_binary_comparison_result_used): Removed.
7382 (rsub_cmp0): New pattern.
7383 (rsub_cmp0_noout): Likewise.
7384 (extzvsi): Changed, keep only meaningful variants.
7385 (SQH, SEZ): New iterators.
7386 (SQH_postfix): New mode attribute.
7387 (SEZ_prefix): New code attribute.
7388 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
7389 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
7390 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
7392 (noncommutative_operator): Check the availability of barrel
7395 2018-04-09 Richard Biener <rguenther@suse.de>
7397 PR tree-optimization/85284
7398 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
7399 Only use the niter constraining form of simple_iv when the exit
7402 2018-04-09 Tom de Vries <tom@codesourcery.com>
7405 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
7406 (define_expand "*memory_barrier"): New define_expand.
7407 (define_insn "memory_barrier"): New insn.
7409 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
7411 PR rtl-optimization/80463
7412 PR rtl-optimization/83972
7413 PR rtl-optimization/83480
7415 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
7416 correct producer for the insn.
7417 (tidy_control_flow): Fixup seqnos in case of debug insns.
7419 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
7421 PR rtl-optimization/83913
7423 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
7424 different sched-times when merging exprs.
7426 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
7428 PR rtl-optimization/83962
7430 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
7431 tidy_fallthru_edge and tidy_control_flow.
7433 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
7435 PR rtl-optimization/83530
7437 * sel-sched.c (force_next_insn): New global variable.
7438 (remove_insn_for_debug): When force_next_insn is true, also leave only
7439 next insn in the ready list.
7440 (sel_sched_region): When the region wasn't scheduled, make another pass
7441 over it with force_next_insn set to 1.
7443 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
7445 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
7447 * config/nds32/constants.md (unspec_volatile_element): Add enum values
7448 for interrupt control.
7449 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
7450 functions for interrupt control.
7451 * config/nds32/nds32-intrinsic.md: Likewise.
7452 * config/nds32/nds32_intrinsic.h: Likewise.
7453 * config/nds32/nds32.h (nds32_builtins): Likewise.
7455 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
7457 * config/nds32/nds32.c (nds32_init_machine_status,
7458 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
7459 strict_aligned_p field.
7460 (nds32_expand_to_rtl_hook): New function.
7461 (TARGET_EXPAND_TO_RTL_HOOK): Define.
7462 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
7464 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
7465 Chung-Ju Wu <jasonwucj@gmail.com>
7467 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
7468 * config/nds32/nds32-n7.md: New file.
7469 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
7470 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
7472 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
7473 * config/nds32/nds32.md (pipeline_model): Add n7.
7474 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
7475 * config/nds32/pipelines.md: Include n7 settings.
7477 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
7478 Chung-Ju Wu <jasonwucj@gmail.com>
7480 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
7481 * config/nds32/nds32-e8.md: New file.
7482 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
7483 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
7485 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
7486 * config/nds32/nds32.md (pipeline_model): Add e8.
7487 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
7488 * config/nds32/pipelines.md: Include e8 settings.
7490 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
7491 Chung-Ju Wu <jasonwucj@gmail.com>
7493 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
7494 * config/nds32/nds32-n8.md: New file.
7495 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
7496 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
7498 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
7499 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
7500 * config/nds32/nds32.md (pipeline_model): Add n8.
7501 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
7502 * config/nds32/pipelines.md: Include n8 settings.
7504 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
7505 Chung-Ju Wu <jasonwucj@gmail.com>
7507 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
7508 * config/nds32/nds32-n9-2r1w.md: New file.
7509 * config/nds32/nds32-n9-3r2w.md: New file.
7510 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
7511 nds32_register_ports): New or modify for cpu n9.
7512 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
7514 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
7515 * config/nds32/nds32-utils.c: New file.
7516 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
7517 TARGET_MUL_SLOW): Define.
7518 * config/nds32/nds32.md (pipeline_model): New attribute.
7519 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
7520 New options that support cpu n9.
7521 * config/nds32/pipelines.md: Include n9 settings.
7522 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
7524 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
7526 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
7527 information if necessary.
7528 (output_cond_branch_compare_zero): Likewise.
7529 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
7530 (nds32_target_alignment): Refine for alignment.
7531 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
7532 (FUNCTION_BOUNDARY): Modify.
7533 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
7535 * config/nds32/nds32.opt (malways-align, malign-functions): New.
7537 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
7539 * config/nds32/constants.md (unspec_volatile_element): Add values for
7540 TLB operation and data prefetch.
7541 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
7542 functions for TLB operation and data prefetch.
7543 * config/nds32/nds32-intrinsic.md: Likewise.
7544 * config/nds32/nds32_intrinsic.h: Likewise.
7545 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
7546 (nds32_print_operand): Likewise.
7547 * config/nds32/nds32.h (nds32_builtins): Likewise.
7549 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
7550 Andrew Pinski <pinsika@gcc.gnu.org>
7553 * match.pd: Use constant_boolean_node of correct type instead of
7554 boolean_true_node or boolean_false_node for simplifying
7555 pointer comparisons to zero.
7557 2018-04-07 Jakub Jelinek <jakub@redhat.com>
7559 PR tree-optimization/80021
7560 * tree.c (verify_type_variant): Make error call in verify_variant_match
7561 translatable and remove final full stop.
7563 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
7565 * config/nds32/constants.md (unspec_volatile_element): Add
7566 UNSPEC_VOLATILE_EH_RETURN.
7567 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
7568 nds32_output_stack_pop): Support dwarf exception handling process.
7569 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
7570 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
7571 exception handling process.
7572 (nds32_compute_stack_frame): Likewise.
7573 (nds32_return_addr_rtx): Likewise.
7574 (nds32_initial_elimination_offset): Likewise.
7575 (nds32_expand_prologue): Likewise.
7576 (nds32_expand_epilogue): Likewise.
7577 (nds32_dynamic_chain_address): New function.
7578 * config/nds32/nds32.h (machine_function): Add fields for dwarf
7580 (DYNAMIC_CHAIN_ADDRESS): Define.
7581 (EH_RETURN_DATA_REGNO): Define.
7582 (EH_RETURN_STACKADJ_RTX): Define.
7583 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
7584 patterns for dwarf exception handling.
7586 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
7588 * config/nds32/nds32.h: Clean up obsolete macros.
7590 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
7592 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
7593 Add enum values for particular instructions.
7594 * config/nds32/nds32-intrinsic.c: Implementation of expanding
7595 particular intrinsic functions.
7596 * config/nds32/nds32-intrinsic.md: Likewise.
7597 * config/nds32/nds32_intrinsic.h: Likewise.
7598 * config/nds32/nds32.h (nds32_builtins): Likewise.
7599 * config/nds32/nds32.md (type): Add pbsad and pbsada.
7600 (btst, ave): New patterns for particular instructions.
7602 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
7604 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
7605 Add enum values for atomic load/store and memory sync.
7606 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
7608 * config/nds32/nds32-intrinsic.md: Likewise.
7609 * config/nds32/nds32_intrinsic.h: Likewise.
7610 * config/nds32/nds32.h (nds32_builtins): Likewise.
7612 2018-04-07 Jakub Jelinek <jakub@redhat.com>
7614 PR tree-optimization/85257
7615 * fold-const.c (native_encode_vector): If not all elts could fit
7616 and off is -1, return 0 rather than offset.
7617 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
7618 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
7619 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
7620 adjust buffer in native_interpret_expr call.
7622 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
7624 * config/nds32/constants.md (unspec_volatile_element): Add cache
7625 control enum values.
7626 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
7627 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
7628 * config/nds32/nds32.c (nds32_cctl_names): New.
7629 (nds32_print_operand): Handle cache control register names.
7630 * config/nds32/nds32.h (nds32_builtins): New enum values.
7631 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
7633 * config/nds32/nds32.md (type): Add mmu.
7634 * config/nds32/pipelines.md (simple_insn): Add mmu.
7636 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
7638 * config/nds32/nds32.md (type): Remove call.
7639 * config/nds32/pipelines.md (simple_insn): Likewise.
7641 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
7643 * config/nds32/constants.md (unspec_volatile_element): Add
7644 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
7645 UNSPEC_VOLATILE_FMFCFG.
7646 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
7647 description for fmfcfg and fmfcsr.
7648 (bdesc_1arg): Add fmtcsr.
7649 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
7650 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
7651 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
7652 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
7653 unspec_fmfcfg): New patterns.
7654 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
7655 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
7656 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
7657 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
7658 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
7659 __nds32__fmfcfg): Define.
7661 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
7663 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
7664 intrinsic register names.
7665 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
7666 intrinsic register enum values and macros.
7668 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
7670 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
7671 for load/store addressing form.
7672 (nds32_print_operand_address): Likewise.
7674 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
7677 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
7678 based on LABEL_REF. Remove useless assertion.
7679 (pic_address_needs_scratch): Fix formatting.
7680 (sparc_legitimize_pic_address): Minor tweaks.
7681 (sparc_delegitimize_address): Adjust assertion accordingly.
7682 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
7683 into symbolic_operand.
7684 (movsi_high_pic_label_ref): Likewise.
7685 (movsi_lo_sum_pic_label_ref): Likewise.
7686 (movdi_pic_label_ref): Likewise.
7687 (movdi_high_pic_label_ref): Likewise.
7688 (movdi_lo_sum_pic_label_ref): Likewise.
7690 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
7692 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
7693 custom LIB_SPEC setup.
7695 2018-04-06 Ruslan Bukin <br@bsdpad.com>
7696 Kito Cheng <kito.cheng@gmail.com>
7698 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
7699 * config/riscv/freebsd.h: New.
7701 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
7703 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
7704 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
7707 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
7708 Kito Cheng <kito.cheng@gmail.com>
7710 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
7711 nds32_output_call, nds32_symbol_binds_local_p): New functions.
7712 * config/nds32/nds32-protos.h (nds32_output_call,
7713 nds32_output_return): Declare.
7714 * config/nds32/nds32.md: Refine all the call and return patterns.
7716 2018-04-06 Jakub Jelinek <jakub@redhat.com>
7719 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
7720 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
7722 PR rtl-optimization/84872
7723 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
7724 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
7727 2018-04-06 Tamar Christina <tamar.christina@arm.com>
7729 * expr.c (copy_blkmode_to_reg): Revert 254862.
7730 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
7732 2018-04-06 Richard Biener <rguenther@suse.de>
7735 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
7736 after seeing a component reference with an adjacent field. Treat
7737 refs to arrays at struct end of external decls similar to
7738 refs to unconstrained commons.
7740 2018-04-06 Jakub Jelinek <jakub@redhat.com>
7743 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
7744 look through SAVE_EXPRs with non-side-effects argument. Adjust
7746 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
7749 2018-04-06 Richard Biener <rguenther@suse.de>
7752 * alias.c (find_base_term): New wrapper around find_base_term
7753 unwinding CSELIB_VAL_PTR changes.
7754 (find_base_term): Do not restore CSELIB_VAL_PTR during the
7757 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7759 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
7761 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
7762 constant definitions.
7763 ("nop"): lr 0,0 -> nopr r0
7764 ("nop_lr0", "nop_lr1"): New insn definitions.
7766 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
7768 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
7769 NDS32_V3PUSH_AVAILABLE_P macro.
7771 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
7772 Chung-Ju Wu <jasonwucj@gmail.com>
7774 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
7775 (nds32*-*-*): Add float and fpu_config into supported_defaults.
7776 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
7777 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
7778 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
7779 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
7780 * config/nds32/constraints.md: New constraints and checking for hard
7781 float configuration.
7782 * config/nds32/iterators.md: New mode iterator and attribute for hard
7783 float configuration.
7784 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
7786 * config/nds32/nds32-fpu.md: New file.
7787 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
7788 deal with hard float code generation.
7789 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
7791 (abi_type, float_reg_number): New enum type.
7792 * config/nds32/nds32-predicates.c: New predicates for hard float.
7793 * config/nds32/nds32-protos.h: Declare functions for hard float.
7794 * config/nds32/nds32.c: Implementation for hard float configuration.
7795 * config/nds32/nds32.h: Definitions for hard float configuration.
7796 * config/nds32/nds32.md: Include hard float machine description and
7797 modify patterns for hard float configuration.
7798 * config/nds32/nds32.opt: New options for hard float configuration.
7799 * config/nds32/predicates.md: New predicates for hard float
7802 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
7804 * common/config/nds32/nds32-common.c
7805 (nds32_option_optimization_table): Enable -mreleax-hint by default.
7807 2018-04-05 Jakub Jelinek <jakub@redhat.com>
7810 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
7811 CONSTRUCTOR_ELT (ctor, ...)->value.
7813 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
7816 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
7818 2018-04-05 Tom de Vries <tom@codesourcery.com>
7821 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
7824 2018-04-05 Shiva Chen <shiva0217@gmail.com>
7825 Kito Cheng <kito.cheng@gmail.com>
7827 * config/nds32/constraints.md (U33): Fine-tune checking condition.
7828 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
7829 * config/nds32/nds32.h (nds32_16bit_address_type): Add
7830 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
7832 2018-04-05 Shiva Chen <shiva0217@gmail.com>
7833 Kito Cheng <kito.cheng@gmail.com>
7835 * config/nds32/constraints.md (Ufe): New memory constraint.
7836 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
7837 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
7838 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
7840 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
7841 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
7843 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7845 * config/nds32/nds32.md: Use optimize_size in the condition for
7846 alu-shift instructions.
7848 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7850 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
7852 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7854 * config/nds32/nds32.md (negsi2): Refine pattern.
7856 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
7857 Chung-Ju Wu <jasonwucj@gmail.com>
7859 * config/nds32/iterators.md (shift_rotate): New code iterator.
7860 (shift): New code attribute.
7861 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
7862 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
7863 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
7864 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
7865 bit-wise operations.
7866 (andsi3, *andsi3): Ditto.
7867 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
7868 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
7869 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
7870 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
7871 nds32_ior_operand, nds32_xor_operand): New predicates.
7873 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7875 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
7876 (addsi3, subsi3): ... this.
7878 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7880 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
7882 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7884 * config/nds32/nds32.md: Adjust indention.
7886 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
7888 * config/nds32/nds32.md (feature): New attribute.
7890 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7892 * config/nds32/nds32.md (subtype): New attribute.
7894 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
7897 * config/arm/arm-builtins.c (arm_expand_builtin): Change
7898 expansion to perform a bitwise AND of the argument followed by a
7899 boolean negation of the result.
7901 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
7903 PR rtl-optimization/84878
7904 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
7905 the basic block. Assert the use reference is not artificial and that
7906 it has an associated insn.
7908 2018-04-04 Michael Matz <matz@suse.de>
7910 * builtins.c (compute_objsize): Pass correct operand
7911 to array_at_struct_end_p.
7913 2018-04-04 Richard Biener <rguenther@suse.de>
7916 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
7917 from contexts for DINFO_LEVEL_TERSE and below.
7919 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
7921 * config/nds32/nds32-doubleword.md (move_<mode>): Require
7922 resiter_operand condition.
7923 * config/nds32/nds32.md (*move<mode>): Ditto.
7925 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
7926 Monk Chiang <sh.chiang04@gmail.com>
7928 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
7930 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
7932 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
7934 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
7935 Kito Cheng <kito.cheng@gmail.com>
7937 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
7938 nds32_cond_code_str, output_cond_branch,
7939 output_cond_branch_compare_zero, nds32_expand_cbranch,
7940 nds32_expand_cstore, nds32_expand_movcc,
7941 nds32_output_cbranchsi4_equality_zero,
7942 nds32_output_cbranchsi4_equality_reg,
7943 nds32_output_cbranchsi4_equality_reg_or_const_int,
7944 nds32_output_cbranchsi4_greater_less_zero: New functions.
7945 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
7946 nds32_expand_cstore, nds32_expand_movcc,
7947 nds32_output_cbranchsi4_equality_zero,
7948 nds32_output_cbranchsi4_equality_reg,
7949 nds32_output_cbranchsi4_equality_reg_or_const_int,
7950 nds32_output_cbranchsi4_greater_less_zero): Declare.
7951 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
7952 nds32_rimm11s_operand): New predicates.
7953 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
7954 * config/nds32/nds32.md: Rewrite all the branch and conditional move
7957 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
7959 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
7960 * config/nds32/nds32.md: Ditto.
7961 * config/nds32/pipelines.md: Ditto.
7963 2018-04-04 Richard Biener <rguenther@suse.de>
7965 PR tree-optimization/85168
7966 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
7967 propagating abnormals.
7969 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
7971 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
7973 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
7974 Kito Cheng <kito.cheng@gmail.com>
7976 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
7977 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
7978 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
7979 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
7980 * config/nds32/nds32.md (sibcall_internal): New.
7981 (sibcall_register): Remove.
7982 (sibcall_immediate): Remove.
7983 (sibcall_value_internal): New.
7984 (sibcall_value_register): Remove.
7985 (sibcall_value_immediate): Remove.
7986 * config/nds32/predicates.md (nds32_general_register_operand): New.
7987 (nds32_call_address_operand): New.
7989 2018-04-03 Jakub Jelinek <jakub@redhat.com>
7991 PR rtl-optimization/85167
7992 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
7993 bb_defs if *split_p, instead preinitialize it to NULL.
7995 PR tree-optimization/85156
7996 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
7997 evaluating the argument multiple times.
7999 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
8001 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
8003 (_mm_cvtpd_ps): Likewise.
8004 (_mm_cvttpd_epi32): Likewise.
8005 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
8006 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
8007 vector, pixel, and bool following altivec.h include.
8009 2018-04-03 Martin Sebor <msebor@redhat.com>
8011 * doc/extend.texi (Common Function Attributes): Clarify.
8012 (const attribute): Likewise.
8013 (pure attribute): Likewise.
8015 2018-04-03 Jakub Jelinek <jakub@redhat.com>
8018 * config/i386/i386.c (ix86_expand_vector_set): Use
8019 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
8021 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
8023 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
8024 instructions when changing rounding bits to preserve precision bits
8025 in the x87 control word.
8027 2018-04-03 Martin Liska <mliska@suse.cz>
8029 PR tree-optimization/82491
8030 * rtl.h (strip_offset_and_add): Replace += suboffset with
8031 poly_uint64 () + suboffset.
8033 2018-03-29 Martin Liska <mliska@suse.cz>
8034 Martin Jambor <mjambor@suse.cz>
8037 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
8038 param_type is not an integral or pointer type.
8040 2018-04-03 Richard Biener <rguenther@suse.de>
8042 * sese.h (recompute_all_dominators): Remove.
8044 2018-04-02 Martin Sebor <msebor@redhat.com>
8046 * doc/invoke.texi (-Wrestrict): Fix typos.
8048 2018-04-02 Jim Wilson <jimw@sifive.com>
8050 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
8051 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
8052 (<optab>di3, <optab>si3_extend): Likewise.
8053 (<optab>si3_mask, <optab>si3_mask_1): New.
8054 (<optab>di3_mask, <optab>di3_mask_1): New.
8055 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
8056 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
8057 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
8059 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
8061 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
8064 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
8066 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
8067 (nds32_canonicalize_comparison): New function.
8069 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
8070 Kito Cheng <kito.cheng@gmail.com>
8071 Kuan-Lin Chen <kuanlinchentw@gmail.com>
8073 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
8074 * config/nds32/constants.md (unspec_volatile_element): Add
8075 UNSPEC_VOLATILE_RELAX_GROUP.
8076 * config/nds32/nds32-relax-opt.c: New file.
8077 * config/nds32/nds32-predicates.c
8078 (nds32_symbol_load_store_p): New function.
8079 * config/nds32/nds32-protos.h
8080 (nds32_symbol_load_store_p): Declare function.
8081 (make_pass_nds32_relax_opt): Declare new rtl pass function.
8082 * config/nds32/nds32.c
8083 (nds32_register_pass): New function to register pass.
8084 (nds32_register_passes): New function to register passes.
8085 * config/nds32/nds32.md (relax_group): New pattern.
8086 * config/nds32/nds32.opt (mrelax-hint): New option.
8087 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
8089 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
8091 * config/nds32/t-nds32: Modify files dependency.
8093 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
8095 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
8096 (PROFILE_HOOK): Define its implementation.
8098 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
8100 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
8101 type and 32-bit size.
8103 2018-04-01 Jakub Jelinek <jakub@redhat.com>
8106 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
8107 (V_128_256): New mode iterator.
8108 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
8109 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
8110 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
8112 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
8113 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
8115 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
8118 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
8119 NaN inputs correctly.
8121 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
8124 * config/rs6000/vsx.md (??r): New mode attribute.
8125 (*vsx_mov<mode>_64bit): Use it.
8126 (*vsx_mov<mode>_32bit): Likewise.
8128 2018-03-30 Martin Sebor <msebor@redhat.com>
8130 PR tree-optimization/84818
8131 * builtins.c (check_access): Use warning_n.
8133 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8136 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
8138 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
8141 2018-03-30 Julia Koval <julia.koval@intel.com>
8144 * x86-tune.def (movx, partial_reg_dependency): Enable for
8147 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
8150 * lra-constraints.c (process_alt_operands): Move setting
8151 this_alternative_matches below.
8153 2018-03-29 Martin Liska <mliska@suse.cz>
8156 * doc/invoke.texi: Document how LTO works with debug info.
8157 Describe auto-load support of binutils. Mention 'x86-64'
8158 as valid option value of -march option.
8160 2018-03-29 Jakub Jelinek <jakub@redhat.com>
8162 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
8165 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
8166 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
8167 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
8170 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
8173 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
8174 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
8175 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
8176 for RS6000_BTM_POWERPC64.
8177 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
8178 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
8179 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
8184 2018-03-28 Carl Love <cel@us.ibm.com>
8187 2017-09-27 Carl Love <cel@us.ibm.com>
8189 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
8190 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
8191 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
8194 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8196 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
8197 instead of __vector bool.
8198 (_mm_max_pu8): Likewise.
8199 (_mm_min_pi16): Likewise.
8201 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
8204 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
8205 (DIVWEUO): Likewise.
8207 (DIVDEUO): Likewise.
8208 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
8209 DIVWEUO and DIVDEUO.
8210 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
8211 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
8212 (div_extend): Likewise.
8213 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
8215 (__builtin_divweuo): Likewise.
8216 (__builtin_divdeo): Likewise.
8217 (__builtin_divdeuo): Likewise.
8219 2018-03-28 Jakub Jelinek <jakub@redhat.com>
8222 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
8223 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
8225 PR tree-optimization/82004
8226 * gimple-match-head.c (optimize_pow_to_exp): New function.
8227 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
8228 Don't fold to exp if optimize_pow_to_exp is false.
8230 2018-03-28 Martin Liska <mliska@suse.cz>
8233 * calls.c (initialize_argument_information): Fix trailing space.
8234 * common.opt: Fix typo and provide better explanation for
8235 -fsanitize-coverage option.
8236 * config/i386/i386.opt: Fix typo.
8238 2018-03-28 Jakub Jelinek <jakub@redhat.com>
8239 Martin Liska <mliska@suse.cz>
8242 * gimplify.c (asan_poison_variable): Don't do the check for
8243 gimplify_omp_ctxp here.
8244 (gimplify_decl_expr): Do it here.
8245 (gimplify_target_expr): Likewise.
8247 2018-03-28 Martin Liska <mliska@suse.cz>
8250 * config/i386/i386.c (ix86_function_arg_advance): Do not call
8251 chkp_type_bounds_count if MPX is not enabled.
8253 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
8255 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
8257 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
8260 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
8261 function to create the function decl for complex long double
8262 multiply and divide for -mabi=ieeelongdouble.
8263 (init_float128_ieee): Call it.
8265 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
8268 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
8269 -fcf-protection=branch -mibt.
8270 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
8272 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
8275 * config/arm/arm.c (arm_valid_symbolic_address): Handle
8276 arm_word_relocations.
8278 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
8281 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
8282 extern array declarations.
8284 2018-03-27 Richard Biener <rguenther@suse.de>
8287 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
8288 explicit single_use checks.
8290 2018-03-27 Richard Biener <rguenther@suse.de>
8292 PR tree-optimization/85082
8293 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
8296 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
8298 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
8299 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
8300 Turn on fasynchronous-unwind-tables and funwind-tables.
8302 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
8305 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
8306 (*bmi_blsr_<mode>_ccz): Ditto.
8308 2018-03-26 Tom de Vries <tom@codesourcery.com>
8310 PR tree-optimization/85063
8311 * omp-general.c (offloading_function_p): New function. Factor out
8313 * omp-offload.c (pass_omp_target_link::gate): ... here.
8314 * omp-general.h (offloading_function_p): Declare.
8315 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
8316 with attribute omp declare target for offloading functions.
8318 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
8320 PR tree-optimization/84005
8321 * tree-data-ref.h (get_base_for_alignment): Declare.
8322 * tree-data-ref.c (get_base_for_alignment_1): New function.
8323 (get_base_for_alignment): Likewise.
8324 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
8325 get_base_for_alignment to find a suitable base object, instead
8326 of always using drb->base_address.
8328 2018-03-23 Jakub Jelinek <jakub@redhat.com>
8331 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
8332 known size by default.
8334 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
8337 * lra-constraints.c (process_alt_operands): Don't match BLKmode
8338 and non BLKmode operands.
8340 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8343 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
8344 Clean up attributes.
8346 2018-03-23 Richard Biener <rguenther@suse.de>
8349 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
8350 we are going to emit early debug for LTO.
8352 2018-03-23 Jakub Jelinek <jakub@redhat.com>
8355 * function.c (match_asm_constraints_1): Don't optimize if input
8356 doesn't satisfy general_operand predicate for output's mode.
8359 * alias.c (write_dependence_p): Don't require for x_canonicalized
8360 non-VOIDmode if x has VOIDmode.
8363 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
8364 just don't try to optimize it rather than assert it never happens.
8366 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
8368 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
8369 macro expansions for definition of ST_INTERNAL_<mode> and
8370 LD_INTERNAL_<mode> builtins.
8371 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
8373 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
8375 (altivec_expand_st_builtin): Likewise.
8376 (altivec_expand_builtin): Remove calls to deleted functions.
8377 (rs6000_address_for_altivec): Delete this function.
8378 * config/rs6000/vector.md: Remove expands for
8379 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
8381 2018-03-22 Sudakshina Das <sudi.das@arm.com>
8384 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
8385 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
8386 re-computing once computed.
8387 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
8388 (arm_init_machine_status): Initialize
8389 machine->static_chain_stack_bytes.
8391 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
8394 * doc/extend.texi: Add four new prototypes for vec_ld.
8395 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
8396 definitions for more logical presentation.
8397 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
8398 entries for V1TI variants of __builtin_altivec_ld builtin.
8399 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
8400 handling of V1TI variant of LVX icode pattern.
8401 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
8402 (rs6000_gimple_fold_builtin): Likewise.
8403 (altivec_init_builtins): Add code to define
8404 __builtin_altivec_lvx_v1ti function.
8406 2018-03-22 Jakub Jelinek <jakub@redhat.com>
8409 * function.c (match_asm_constraints_1): Don't do the optimization
8410 if input isn't a REG, SUBREG, MEM or constant.
8412 2018-03-22 Tom de Vries <tom@codesourcery.com>
8414 PR tree-optimization/84956
8415 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
8416 bb_has_abnormal_pred.
8418 2018-03-22 Jakub Jelinek <jakub@redhat.com>
8421 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
8422 DECL_INITIAL (decl) to decl at the end.
8423 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
8426 2018-03-21 Joseph Myers <joseph@codesourcery.com>
8428 * doc/extend.texi (__builtin_tgmath): Document when complex
8429 integer types are treated as _Complex _Float64.
8431 2018-03-21 Tom de Vries <tom@codesourcery.com>
8433 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
8435 2018-03-21 Jakub Jelinek <jakub@redhat.com>
8437 PR tree-optimization/84960
8438 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
8439 if it is ENTRY block, move them into single succ of ENTRY in that case.
8441 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
8443 PR tree-optimization/84811
8444 * poly-int.h (poly_span_traits): Remove the T3 parameter and
8445 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
8446 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
8447 (known_subrange_p): Update accordingly. Cast each value involved
8448 in the size comparison, rather than casting the result of the
8451 2018-03-21 Jakub Jelinek <jakub@redhat.com>
8453 PR tree-optimization/84982
8454 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
8455 by flipping the least significant bit rather than all bits from
8456 bitpos to bitpos + bitsize - 1.
8458 2018-03-21 Nathan Sidwell <nathan@acm.org>
8460 * doc/extend.texi (Deprecated Features): Remove mention of
8461 long-deleted deprecations.
8463 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8466 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
8467 * configure: Regenerate.
8469 2018-03-21 Tom de Vries <tom@codesourcery.com>
8471 PR tree-optimization/83126
8472 * tree-parloops.c (num_phis): New function.
8473 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
8475 2018-03-21 Nathan Sidwell <nathan@acm.org>
8477 * doc/extend.texi (Deprecated Features): Update deprecated flags,
8478 mention anon-struct/union members and trailing attributes.
8480 2018-03-21 Bin Cheng <bin.cheng@arm.com>
8482 PR tree-optimization/84969
8483 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
8484 builtin memset partitions if they set different rhs values.
8486 2018-03-21 Jakub Jelinek <jakub@redhat.com>
8488 PR rtl-optimization/84989
8489 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
8490 VEC_DUPLICATE with scalar result mode.
8492 2018-03-21 Martin Liska <mliska@suse.cz>
8495 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
8496 not intended return statement.
8498 2018-03-21 Martin Liska <mliska@suse.cz>
8501 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
8502 (chkp_find_bound_slots_1): Limit number of iterations.
8504 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
8507 * Minor grammar fixes for x86 options.
8509 2018-03-20 Jakub Jelinek <jakub@redhat.com>
8512 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
8513 holding REG_CFA_RESTORE notes, instead turn them into a USE.
8515 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
8518 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
8519 (altivec_lvx_<mode>_1op): Likewise.
8520 (altivec_stvx_<mode>_2op): Likewise.
8521 (altivec_stvx_<mode>_1op): Likewise.
8522 (altivec_lvx_<VM2:mode>): New define_expand.
8523 (altivec_stvx_<VM2:mode>): Likewise.
8524 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
8525 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
8526 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
8527 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
8528 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
8529 (rs6000_gen_lvx): Likewise.
8530 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
8531 (altivec_expand_stv_builtin): Likewise.
8532 (altivec_expand_builtin): Likewise.
8533 * config/rs6000/vector.md: Likewise.
8535 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8538 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
8541 2018-03-20 Richard Biener <rguenther@suse.de>
8544 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
8545 sign-conversions as zero, fall back to standard scalar_stmt
8548 2018-03-20 Martin Liska <mliska@suse.cz>
8551 * predict.c (rebuild_frequencies): Handle case when we have
8552 PROFILE_ABSENT, but flag_guess_branch_prob is false.
8554 2018-03-20 Jakub Jelinek <jakub@redhat.com>
8557 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
8558 flag_section_anchors.
8559 * varasm.c (use_blocks_for_decl_p): Remove hack for
8560 dw2_force_const_mem.
8563 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
8565 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
8566 be created, use lowpart_subreg of operands[0] rather than operands[0]
8568 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
8569 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
8570 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
8571 and n constraint instead of aarch64_shift_imm_di and Usd.
8572 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
8573 (*aarch64_<optab>_reg_minus<mode>3): ... this.
8575 2018-03-20 Sudakshina Das <sudi.das@arm.com>
8578 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
8579 to favor GPR over NEON registers.
8580 (<shift>di3_neon): Likewise.
8582 2018-03-20 Tom de Vries <tom@codesourcery.com>
8585 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
8586 (nvptx_process_pars): Emit bar.sync asap and alap.
8588 2018-03-20 Tom de Vries <tom@codesourcery.com>
8591 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
8592 seen_label if seen_label is already set.
8594 2018-03-20 Jakub Jelinek <jakub@redhat.com>
8597 * config/i386/i386.c (fold_builtin_cpu): For features above 31
8598 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
8599 Use 1U instead of 1. Formatting fixes.
8602 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
8603 instead of TREE_TYPE (s1) for the return value.
8605 2018-03-19 Jakub Jelinek <jakub@redhat.com>
8607 PR tree-optimization/84946
8608 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
8609 bitsize + bitsize in poly_uint64 rather than poly_int64.
8612 * dwarf2asm.c: Include fold-const.c.
8613 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
8614 of decl rather than decl itself.
8616 PR rtl-optimization/84643
8617 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
8619 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
8622 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
8623 calling assemble_variable.
8625 2018-03-19 Sudakshina Das <sudi.das@arm.com>
8628 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
8629 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
8631 2018-03-19 Jim Wilson <jimw@sifive.com>
8634 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
8635 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
8636 (riscv_first_stack_step): Likewise.
8637 (riscv_option_override): Use STACK_BOUNDARY instead of
8639 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
8641 (BIGGEST_ALIGNMENT): Set to 128.
8642 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
8643 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
8646 2018-03-19 Richard Biener <rguenther@suse.de>
8648 PR tree-optimization/84933
8649 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
8650 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
8652 2018-03-19 Richard Biener <rguenther@suse.de>
8654 PR tree-optimization/84859
8655 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
8656 (cond_if_else_store_replacement): Perform sinking operation on
8657 single-store BBs regardless of MAX_STORES_TO_SINK setting.
8658 Generalize what a BB with a single eligible store is.
8660 2018-03-19 Richard Biener <rguenther@suse.de>
8662 PR tree-optimization/84929
8663 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
8664 chrec_is_positive against non-chrec arg.
8666 2018-03-19 Tamar Christina <tamar.christina@arm.com>
8669 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
8671 2018-03-18 Martin Liska <mliska@suse.cz>
8673 PR rtl-optimization/84635
8674 * regrename.c (build_def_use): Use matches_mode only when
8677 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
8679 PR tree-optimization/84913
8680 * tree-vect-loop.c (vectorizable_reduction): Don't try to
8681 vectorize chains of COND_EXPRs.
8683 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
8685 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
8687 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
8689 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
8691 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
8693 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
8695 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
8696 Kito Cheng <kito.cheng@gmail.com>
8698 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
8699 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
8700 (nds32_adjust_reg_alloc_order): New function.
8701 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
8703 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
8705 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
8706 nds32_print_operand, nds32_print_operand_address): Use
8707 HOST_WIDE_INT_PRINT_DEC instead.
8709 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
8711 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
8713 2018-03-17 Jakub Jelinek <jakub@redhat.com>
8716 * config/i386/i386.c (initial_ix86_tune_features,
8717 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
8719 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
8720 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
8721 rather than 1u << ix86_tune. Formatting fix.
8722 (ix86_option_override_internal): Change ix86_arch_mask from
8723 unsigned int to unsigned HOST_WIDE_INT, initialize to
8724 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
8725 (ix86_function_specific_restore): Likewise.
8727 2018-03-16 Jakub Jelinek <jakub@redhat.com>
8730 * postreload.c (reload_combine_recognize_pattern): Perform
8731 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
8732 truncate_int_for_mode the result for the destination's mode.
8735 * hsa-gen.c (mem_type_for_type): Fix comment typo.
8736 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
8738 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
8741 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
8744 * lra-assigns.c (lra_split_hard_reg_for): Don't use
8745 regno_allocno_class_array and sorted_pseudos.
8746 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
8747 insns where regno is used.
8749 2018-03-16 Martin Liska <mliska@suse.cz>
8752 * multiple_target.c (create_dispatcher_calls): Redirect
8753 reference in the symbol table.
8755 2018-03-16 Martin Liska <mliska@suse.cz>
8758 * multiple_target.c (create_dispatcher_calls): Redirect also
8761 2018-03-16 Jakub Jelinek <jakub@redhat.com>
8765 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
8766 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
8767 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
8769 2018-03-16 Julia Koval <julia.koval@intel.com>
8771 * doc/invoke.texi (Skylake Server): Add CLWB.
8772 Cannonlake): Remove CLWB.
8774 2018-03-16 Jakub Jelinek <jakub@redhat.com>
8776 PR tree-optimization/84841
8777 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
8779 (FLOAT_ONE_CONST_TYPE): Define.
8780 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
8781 (sort_by_operand_rank): Put entries with higher constant_type last
8782 rather than first to match comments.
8784 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
8786 * config/nios2/nios2.md (movsi_internal): Fix thinko in
8789 2018-03-15 Jakub Jelinek <jakub@redhat.com>
8792 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
8793 check and use address of target always.
8795 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
8798 * config/i386/i386.c (indirect_thunk_needed): Update comments.
8799 (indirect_thunk_bnd_needed): Likewise.
8800 (indirect_thunks_used): Likewise.
8801 (indirect_thunks_bnd_used): Likewise.
8802 (indirect_return_needed): New.
8803 (indirect_return_bnd_needed): Likewise.
8804 (output_indirect_thunk_function): Add a bool argument for
8806 (output_indirect_thunk_function): Don't generate alias for
8807 function return thunk.
8808 (ix86_code_end): Call output_indirect_thunk_function to generate
8809 function return thunks.
8810 (ix86_output_function_return): Set indirect_return_bnd_needed
8811 and indirect_return_needed instead of indirect_thunk_bnd_needed
8812 and indirect_thunk_needed.
8814 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
8816 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
8817 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
8818 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
8820 2018-03-15 David Malcolm <dmalcolm@redhat.com>
8821 Paul Hua <paul.hua.gm@gmail.com>
8824 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
8826 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
8828 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
8829 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
8832 2018-03-15 Tamar Christina <tamar.christina@arm.com>
8835 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
8836 instead of GET_MODE_SIZE when comparing Units.
8838 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
8841 * varasm.c (hash_section): Return an unchangeble hash value
8842 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
8843 Return !aarch64_can_use_per_function_literal_pools_p ().
8845 2018-03-15 Jakub Jelinek <jakub@redhat.com>
8848 * optabs.c (emit_conditional_move): Pass address of cmode's copy
8849 rather than address of cmode as last argument to prepare_cmp_insn.
8851 2018-03-15 Julia Koval <julia.koval@intel.com>
8853 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
8854 F_AVX512VNNI, F_AVX512BITALG): New.
8856 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
8859 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
8860 insn for floating-point loads and stores.
8862 2018-03-14 Carl Love <cel@us.ibm.com>
8864 * config/rs6000/rs6000-c.c: Add macro definitions for
8865 ALTIVEC_BUILTIN_VEC_PERMXOR.
8866 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
8867 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
8868 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
8869 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
8871 * config/doc/extend.texi: Add prototypes for vec_permxor.
8873 2018-03-14 David Malcolm <dmalcolm@redhat.com>
8876 * diagnostic-show-locus.c (class layout_point): Convert m_line
8877 from int to linenum_type.
8878 (line_span::comparator): Use linenum "compare" function when
8879 comparing line numbers.
8880 (test_line_span): New function.
8881 (layout_range::contains_point): Convert param "row" from int to
8883 (layout_range::intersects_line_p): Likewise.
8884 (layout::will_show_line_p): Likewise.
8885 (layout::print_source_line): Likewise.
8886 (layout::should_print_annotation_line_p): Likewise.
8887 (layout::print_annotation_line): Likewise.
8888 (layout::print_leading_fixits): Likewise.
8889 (layout::annotation_line_showed_range_p): Likewise.
8890 (struct line_corrections): Likewise for field m_row.
8891 (line_corrections::line_corrections): Likewise for param "row".
8892 (layout::print_trailing_fixits): Likewise.
8893 (layout::get_state_at_point): Likewise.
8894 (layout::get_x_bound_for_row): Likewise.
8895 (layout::print_line): Likewise.
8896 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
8897 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
8898 * input.c (selftest::test_linenum_comparisons): New function.
8899 (selftest::input_c_tests): Call it.
8900 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
8901 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
8902 * selftest.h (ASSERT_GT): New macro.
8903 (ASSERT_GT_AT): New macro.
8904 (ASSERT_LT): New macro.
8905 (ASSERT_LT_AT): New macro.
8907 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
8909 PR rtl-optimization/84780
8910 * combine.c (distribute_links): Don't make a link based on pc_rtx.
8912 2018-03-14 Martin Liska <mliska@suse.cz>
8914 * tree.c (record_node_allocation_statistics): Use
8915 get_stats_node_kind.
8916 (get_stats_node_kind): New function extracted from
8917 record_node_allocation_statistics.
8918 (free_node): Use get_stats_node_kind.
8920 2018-03-14 Richard Biener <rguenther@suse.de>
8922 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
8923 that the value-set of ANTIC_IN doesn't grow.
8926 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
8928 (BB_VISITED_WITH_VISITED_SUCCS): New define.
8929 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
8931 2018-03-14 Julia Koval <julia.koval@intel.com>
8933 * config.gcc (icelake-client, icelake-server): New.
8935 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
8936 (initial_ix86_arch_features): Ditto.
8937 (PTA_SKYLAKE): Add SGX.
8938 (PTA_ICELAKE): Remove.
8939 (PTA_ICELAKE_CLIENT): New.
8940 (PTA_ICELAKE_SERVER): New.
8941 (ix86_option_override_internal): Split up icelake on icelake client and
8943 (get_builtin_code_for_version): Ditto.
8944 (fold_builtin_cpu): Ditto.
8945 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
8946 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
8947 * config/i386/i386.h (processor_type): Ditto.
8948 * doc/invoke.texi: Ditto.
8950 2018-03-14 Jakub Jelinek <jakub@redhat.com>
8953 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
8954 INTEGER_CST offset, add it together with bitpos / 8 and
8955 sign extend based on POINTER_SIZE.
8959 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
8962 * config/i386/constraints.md (Yc): New register constraint.
8963 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
8964 Use Yc constraint for alternative 2 of operand 0. Remove
8965 preferred_for_speed attribute.
8967 2018-03-14 Richard Biener <rguenther@suse.de>
8969 PR tree-optimization/84830
8970 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
8971 with the old one to avoid oscillations.
8973 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
8976 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
8978 (assign_by_spills): Return a flag of reload assignment failure.
8979 Do not process the reload assignment failures. Do not spill other
8980 reload pseudos if they has the same reg class. Update n if
8982 (lra_assign): Add a return arg. Set up from the result of
8983 assign_by_spills call.
8984 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
8985 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
8986 usage_insns if it is not NULL.
8987 (spill_hard_reg_in_range): New function.
8988 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
8989 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
8990 function prototypes.
8991 (lra_assign): Change prototype.
8992 * lra.c (lra): Add code to deal with fails by splitting hard reg
8995 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
8997 * config/riscv/riscv.opt (mrelax): New option.
8998 * config/riscv/riscv.c (riscv_file_start): Emit ".option
8999 "norelax" when riscv_mrelax is disabled.
9000 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
9002 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9005 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
9006 reassociation for int modes.
9008 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
9010 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
9011 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
9013 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
9014 * config/aarch64/aarch64-sve.md
9015 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
9016 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
9017 (*extend<mode><Vwide>2): Rename to...
9018 (aarch64_sve_extend<mode><Vwide>2): ...this.
9019 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
9020 renaming the old pattern to...
9021 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
9023 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
9024 define_expand, renaming the old pattern to...
9025 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
9026 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
9027 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
9028 account when deciding which SVE instruction the optab should use.
9029 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
9031 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
9033 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
9034 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
9035 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
9036 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
9037 (tlsdesc_small_<mode>): Turn a define_expand and use
9038 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
9039 (tlsdesc_small_advsimd_<mode>): ...this.
9040 (tlsdesc_small_sve_<mode>): New pattern.
9042 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
9044 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
9045 (UNSPEC_UMUL_HIGHPART): New constants.
9046 (MUL_HIGHPART): New int iteraor.
9047 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
9048 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
9050 (*<su>mul<mode>3_highpart): New define_insn.
9052 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
9055 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
9058 2018-03-13 Martin Liska <mliska@suse.cz>
9061 * (sem_item_optimizer::sem_item_optimizer): Initialize new
9063 (sem_item_optimizer::~sem_item_optimizer): Release it.
9064 (sem_item_optimizer::merge_classes): Register variable aliases.
9065 (sem_item_optimizer::fixup_pt_set): New function.
9066 (sem_item_optimizer::fixup_points_to_sets): Likewise.
9067 * ipa-icf.h: Declare new variables and functions.
9069 2018-03-13 Jakub Jelinek <jakub@redhat.com>
9072 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
9073 integer_pow2p@2 and test integer_pow2p in condition.
9074 (A < 0 ? C : 0): Similarly for @1.
9077 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
9078 characters starting at p contain '\0' character, don't look beyond
9082 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
9083 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
9086 * reg-stack.c (change_stack): Change update_end var from int to
9087 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
9088 also call set_block_for_insn on the newly added insns and rescan.
9091 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
9092 on the last operand.
9095 * tree.c (stabilize_reference_1): Return save_expr (e) for
9096 STATEMENT_LIST even if it doesn't have side-effects.
9098 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
9100 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
9102 2018-03-12 Renlin Li <renlin.li@arm.com>
9104 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
9105 aarch64_output_scalar_simd_mov_immediate.
9107 2018-03-12 Martin Sebor <msebor@redhat.com>
9109 PR tree-optimization/83456
9110 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
9111 for perfectly overlapping calls to memcpy.
9112 (gimple_fold_builtin_memory_chk): Same.
9113 (gimple_fold_builtin_strcpy): Handle no-warning.
9114 (gimple_fold_builtin_stxcpy_chk): Same.
9115 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
9117 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
9119 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
9120 parameter. Use it for SFmode.
9121 (rs6000_function_arg_advance_1): Adjust.
9122 (rs6000_function_arg): Adjust.
9123 (rs6000_gimplify_va_arg): Pass false for that new parameter.
9125 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
9127 PR rtl-optimization/84169
9128 PR rtl-optimization/84780
9129 * combine.c (can_combine_p): Check for a 2-insn combination whether
9130 the destination register is used between the two insns, too.
9132 2018-03-12 Richard Biener <rguenther@suse.de>
9134 PR tree-optimization/84803
9135 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
9136 for refs DR analysis didn't process.
9138 2018-03-12 Richard Biener <rguenther@suse.de>
9140 PR tree-optimization/84777
9141 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
9142 force-vectorize loops ignore whether we are optimizing for size.
9144 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
9146 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
9147 (TARGET_MD_ASM_ADJUST): Define.
9149 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
9150 Kito Cheng <kito.cheng@gmail.com>
9151 Chung-Ju Wu <jasonwucj@gmail.com>
9153 * config/nds32/nds32.c (nds32_compute_stack_frame,
9154 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
9155 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
9156 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
9157 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
9158 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
9159 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
9160 * config/nds32/nds32.md (prologue, epilogue): Use macro
9161 NDS32_V3PUSH_AVAILABLE_P to do checking.
9163 2018-03-11 Jakub Jelinek <jakub@redhat.com>
9166 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
9167 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
9168 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
9169 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
9170 addition of most attributes on !orig_type_die or the attribute not
9171 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
9173 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
9174 Chung-Ju Wu <jasonwucj@gmail.com>
9176 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
9178 * config/nds32/nds32.opt (mvh): New option.
9180 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
9181 Chung-Ju Wu <jasonwucj@gmail.com>
9183 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
9185 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
9186 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
9189 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
9190 Chung-Ju Wu <jasonwucj@gmail.com>
9192 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
9194 * config/nds32/nds32-multiple.md (strlensi): New pattern.
9195 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
9197 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
9198 Kito Cheng <kito.cheng@gmail.com>
9199 Chung-Ju Wu <jasonwucj@gmail.com>
9201 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
9202 UNSPEC_FFMISM and UNSPEC_FLMISM.
9203 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
9204 for ffb, ffmism and flmism.
9205 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
9206 (unspec_ffmism): Ditto.
9207 (unspec_flmism): Ditto.
9208 (nds32_expand_builtin_impl): Check if string extension is available.
9209 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
9210 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
9212 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
9215 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
9218 * lra-assigns.c (assign_by_spills): Return a flag of reload
9219 assignment failure. Do not process the reload assignment
9220 failures. Do not spill other reload pseudos if they has the same
9222 (lra_assign): Add a return arg. Set up from the result of
9223 assign_by_spills call.
9224 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
9225 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
9226 usage_insns if it is not NULL.
9227 (spill_hard_reg_in_range): New function.
9228 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
9229 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
9230 function prototypes.
9231 (lra_assign): Change prototype.
9232 * lra.c (lra): Add code to deal with fails by splitting hard reg
9235 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
9238 * config/i386/i386.opt: Replace Enforcment with Enforcement.
9240 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
9243 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
9244 (dw_val_node): Add val_symbolic_view.
9245 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
9246 (symview_upper_bound): New.
9247 (new_line_info_table): Initialize symviews_since_reset.
9248 (dwarf2out_source_line): Count symviews_since_reset and set
9249 symview_upper_bound.
9250 (dw_val_equal_p): Handle symview.
9251 (add_AT_symview): New.
9252 (print_dw_val): Handle symview.
9253 (attr_checksum, attr_checksum_ordered): Likewise.
9254 (same_dw_val_p, size_of_die): Likewise.
9255 (value_format, output_die): Likewise.
9256 (add_high_low_attributes): Use add_AT_symview for entry_view.
9257 (dwarf2out_finish): Reset symview_upper_bound, clear
9260 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
9263 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
9264 Add strict argument and use it.
9265 (rs6000_split_multireg_move): Update for new strict argument.
9266 (mem_operand_gpr): Disallow all non-offsettable addresses.
9267 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
9269 2018-03-09 Jakub Jelinek <jakub@redhat.com>
9272 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
9273 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
9274 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
9277 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
9278 decl, use remap_type if we want to use the type.
9280 2018-03-09 Martin Sebor <msebor@redhat.com>
9282 PR tree-optimization/84526
9283 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
9285 (builtin_access::generic_overlap): Be prepared to handle non-array
9288 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
9290 PR rtl-optimization/84682
9291 * lra-constraints.c (process_address_1): Check is_address flag
9292 for address constraints.
9293 (process_alt_operands): Likewise.
9294 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
9295 preprocess_constraints.
9296 * recog.h (preprocess_constraints): Add oploc parameter.
9298 * recog.c (preprocess_constraints): Test address_operand for
9299 CT_ADDRESS constraints.
9301 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
9304 * lra-assigns.c (assign_by_spills): Return a flag of reload
9305 assignment failure. Do not process the reload assignment
9306 failures. Do not spill other reload pseudos if they has the same
9308 (lra_assign): Add a return arg. Set up from the result of
9309 assign_by_spills call.
9310 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
9311 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
9312 usage_insns if it is not NULL.
9313 (spill_hard_reg_in_range): New function.
9314 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
9315 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
9316 function prototypes.
9317 (lra_assign): Change prototype.
9318 * lra.c (lra): Add code to deal with fails by splitting hard reg
9321 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9324 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
9325 Accept complain bool parameter. Only emit errors if it is true.
9326 (arm_parse_cpu_option_name): Likewise.
9327 (arm_target_thumb_only): Adjust callers of the above.
9328 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
9329 prototype to take a default true bool parameter.
9330 (arm_parse_arch_option_name): Likewise.
9332 2018-03-09 David Malcolm <dmalcolm@redhat.com>
9333 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
9337 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
9338 * configure: Regenerate.
9339 * configure.ac ("linker --version-script option"): New.
9340 ("linker soname option"): New.
9342 2018-03-09 Richard Biener <rguenther@suse.de>
9344 PR tree-optimization/84775
9345 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
9346 immediate uses of predicate stmts and mark them modified.
9349 PR tree-optimization/84178
9350 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
9352 (version_loop_for_if_conversion): Delay update_ssa call.
9353 (tree_if_conversion): Delay update_ssa until after predicate
9356 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
9359 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
9360 when the function accesses prior frames.
9362 2018-03-08 Jakub Jelinek <jakub@redhat.com>
9365 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
9366 gen_llsym, otherwise call maybe_gen_llsym.
9369 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
9370 has ',' character inside of it.
9372 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9375 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
9376 as clobbering CC_REGNUM.
9378 2018-03-08 Richard Biener <rguenther@suse.de>
9381 * tree-scalar-evolution.c: Include tree-into-ssa.h.
9382 (follow_copies_to_constant): Do not follow SSA names registered
9385 2018-03-08 Richard Biener <rguenther@suse.de>
9387 PR tree-optimization/84178
9388 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
9390 (version_loop_for_if_conversion): Delay update_ssa call.
9391 (tree_if_conversion): Delay update_ssa until after predicate
9394 2018-03-08 David Malcolm <dmalcolm@redhat.com>
9396 PR tree-optimization/84178
9397 * tree-if-conv.c (release_bb_predicate): Remove the
9398 the assertion that the stmts have NULL use_ops.
9399 Discard the statements, asserting that they haven't
9400 yet been added to a BB.
9402 2018-03-08 Richard Biener <rguenther@suse.de>
9404 PR tree-optimization/84746
9405 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
9406 (phi_translate): Pass in destination ANTIC_OUT set.
9407 (phi_translate_1): Likewise. For a simplified result lookup
9408 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
9409 (phi_translate_set): Adjust.
9410 (do_pre_regular_insertion): Likewise.
9411 (do_pre_partial_partial_insertion): Likewise.
9413 2018-03-08 Martin Liska <mliska@suse.cz>
9415 PR gcov-profile/84735
9416 * doc/gcov.texi: Document usage of profile files.
9417 * gcov-io.h: Document changes in the format.
9419 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
9423 * dwarf2out.c (struct dw_line_info_table): Update comments for
9425 (FORCE_RESET_NEXT_VIEW): New.
9426 (FORCE_RESETTING_VIEW_P): New.
9427 (RESETTING_VIEW_P): Check for -1 too.
9428 (ZERO_VIEW_P): Likewise.
9429 (new_line_info_table): Force-reset next view.
9430 (dwarf2out_begin_function): Likewise.
9431 (dwarf2out_source_line): Simplify zero_view_p initialization.
9432 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
9433 view directly. Omit view when omitting .loc at line 0.
9435 2018-03-08 Jakub Jelinek <jakub@redhat.com>
9437 PR tree-optimization/84740
9438 * tree-switch-conversion.c (process_switch): Call build_constructors
9439 only if info.phi_count is non-zero.
9441 PR tree-optimization/84739
9442 * tree-tailcall.c (find_tail_calls): Check call arguments against
9443 DECL_ARGUMENTS (current_function_decl) rather than
9444 DECL_ARGUMENTS (func) when checking for tail recursion.
9446 2018-03-07 Jakub Jelinek <jakub@redhat.com>
9448 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
9449 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
9450 Volker Reichelt's entry and add entries for people that perform
9451 GCC fuzzy testing and report numerous bugs.
9453 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
9456 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
9457 readonly data in sdata, if that is disabled.
9458 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
9459 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
9460 -mreadonly-in-sdata option.
9462 2018-03-07 Martin Sebor <msebor@redhat.com>
9464 PR tree-optimization/84468
9465 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
9466 basic block when looking for nul assignment.
9468 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
9471 * except.h (output_function_exception_table): Adjust prototype.
9472 * except.c (output_function_exception_table): Remove FNNAME parameter
9473 and add SECTION parameter. Ouput one part of the table at a time.
9474 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
9475 the first part of the exception table and emit unwind directives.
9476 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
9477 (i386_pe_seh_cold_init): Likewise.
9478 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
9479 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
9480 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
9481 (ix86_output_call_insn): Emit a nop in one more case for SEH.
9482 * config/i386/winnt.c: Include except.h.
9483 (struct seh_frame_state): Add reg_offset, after_prologue and
9484 in_cold_section fields.
9485 (i386_pe_seh_end_prologue): Set seh->after_prologue.
9486 (i386_pe_seh_cold_init): New function.
9487 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
9488 to seh->in_cold_section.
9489 (seh_emit_push): Record the offset of the push.
9490 (seh_emit_save): Record the offet of the save.
9491 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
9492 Test seh->after_prologue to disregard the epilogue.
9493 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
9494 (i386_pe_end_cold_function): New function.
9496 2018-03-07 Jakub Jelinek <jakub@redhat.com>
9499 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
9500 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
9503 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
9505 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
9506 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
9509 * multiple_target.c: Include tree-inline.h and intl.h.
9510 (expand_target_clones): Diagnose and fail if node->definition and
9511 !tree_versionable_function_p (node->decl).
9513 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
9515 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
9517 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
9518 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
9519 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
9521 2018-03-06 Jakub Jelinek <jakub@redhat.com>
9524 * combine.c (try_combine): Use reg_or_subregno instead of handling
9525 just paradoxical SUBREGs and REGs.
9527 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
9529 * config/arc/arc.c (arc_finalize_pic): Remove function.
9530 (arc_must_save_register): We use single base PIC register, remove
9531 checks to save/restore the PIC register.
9532 (arc_expand_prologue): Likewise.
9533 * config/arc/arc-protos.h (arc_set_default_type_attributes):
9535 (arc_verify_short): Likewise.
9536 (arc_attr_type): Likewise.
9537 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
9538 (walk_stores): Likewise.
9539 (arc_address_cost): Make it static.
9540 (arc_verify_short): Likewise.
9541 (branch_dest): Likewise.
9542 (arc_attr_type): Likewise.
9543 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
9544 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
9545 (arc_final_prescan_insn): Remove inserting the nops due to
9546 hardware hazards. It is done in reorg step.
9547 (insn_length_variant_t): Remove.
9548 (insn_length_parameters_t): Likewise.
9549 (arc_insn_length_parameters): Likewise.
9550 (arc_get_insn_variants): Likewise.
9551 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
9553 2018-03-06 Jakub Jelinek <jakub@redhat.com>
9556 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
9559 PR tree-optimization/84687
9560 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
9562 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
9564 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9566 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
9567 Rename to ppc_speculation_barrier.
9568 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
9569 __builtin_ppc_speculation_barrier.
9571 2018-03-05 Jakub Jelinek <jakub@redhat.com>
9574 * combine.c (combine_simplify_rtx): Don't try to simplify if
9575 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
9578 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
9580 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
9581 to 32 bytes when compiling for POWER9.
9583 2018-03-05 Jakub Jelinek <jakub@redhat.com>
9586 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
9587 regparm >= 3 with no arg reg available also for calls with
9588 flag_force_indirect_call. Pass decl to ix86_function_regparm.
9591 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
9593 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
9595 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
9598 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
9600 2018-03-05 Richard Biener <rguenther@suse.de>
9602 PR tree-optimization/84486
9603 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
9604 When inserting a __builtin_assume_aligned call set the LHS
9605 SSA name alignment info accordingly.
9607 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
9609 PR tree-optimization/84114
9610 * config/aarch64/aarch64.c (aarch64_reassociation_width)
9611 Avoid reassociation of FLOAT_MODE addition.
9613 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
9615 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
9616 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
9617 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
9618 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
9619 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
9620 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
9621 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
9623 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
9624 __builtin_ia32_wbinvd): New builtins.
9625 (SPECIAL_ARGS2): New.
9626 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
9627 (SPECIAL_ARGS2): New.
9628 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
9629 (ix86_valid_target_attribute_inner_p): Ditto.
9630 (ix86_init_mmx_sse_builtins): Add special_args2.
9631 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
9632 TARGET_WBNOINVD_P): New.
9633 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
9634 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
9635 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
9636 * config/i386/immintrin.h (_wbinvd): New intrinsic.
9637 * config/i386/pconfigintrin.h: New file.
9638 * config/i386/wbnoinvdintrin.h: Ditto.
9639 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
9641 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
9643 2018-03-05 Richard Biener <rguenther@suse.de>
9645 PR tree-optimization/84670
9646 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
9648 (BB_VISITED_WITH_VISITED_SUCCS): New define.
9649 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
9650 (compute_antic_aux): Only assert the number of values in ANTIC_IN
9651 doesn't grow if all successors (recursively) were visited at least
9654 2018-03-05 Richard Biener <rguenther@suse.de>
9656 PR tree-optimization/84650
9657 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
9658 if executed in the loop pipeline.
9660 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
9662 * doc/configfiles.texi (Configuration Files): Move info about
9663 conditionalizing $target-protos.h to...
9664 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
9665 differs from $target-protos.h.
9667 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
9668 Chung-Ju Wu <jasonwucj@gmail.com>
9670 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
9671 * config/nds32/nds32-multiple.md (setmemsi): Define.
9672 * config/nds32/nds32-memory-manipulation.c
9673 (nds32_gen_dup_4_byte_to_word_value): New.
9674 (emit_setmem_word_loop): New.
9675 (emit_setmem_byte_loop): New.
9676 (nds32_expand_setmem_loop): New.
9677 (nds32_expand_setmem_loop_v3m): New.
9678 (nds32_expand_setmem_unroll): New.
9679 (nds32_expand_setmem): New.
9681 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
9682 Chung-Ju Wu <jasonwucj@gmail.com>
9684 * config/nds32/nds32-memory-manipulation.c
9685 (nds32_emit_load_store): New.
9686 (nds32_emit_post_inc_load_store): New.
9687 (nds32_emit_mem_move): New.
9688 (nds32_emit_mem_move_block): New.
9689 (nds32_expand_movmemsi_loop_unknown_size): New.
9690 (nds32_expand_movmemsi_loop_known_size): New.
9691 (nds32_expand_movmemsi_loop): New.
9692 (nds32_expand_movmemsi_unroll): New.
9693 (nds32_expand_movmemqi): Rename ...
9694 (nds32_expand_movmemsi): ... to this.
9695 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
9696 (movmemsi): ... to this.
9697 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
9698 (nds32_expand_movmemsi): ... to this.
9700 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
9701 Monk Chiang <sh.chiang04@gmail.com>
9702 Chung-Ju Wu <jasonwucj@gmail.com>
9704 * config/nds32/nds32-protos.h
9705 (nds32_expand_load_multiple): New arguments.
9706 (nds32_expand_store_multiple): Ditto.
9707 (nds32_valid_multiple_load_store): Rename ...
9708 (nds32_valid_multiple_load_store_p): ... to this.
9709 * config/nds32/nds32-memory-manipulation.c
9710 (nds32_expand_load_multiple): Refine implementation.
9711 (nds32_expand_store_multiple): Ditto.
9712 * config/nds32/nds32-multiple.md
9713 (load_multiple): Update nds32_expand_load_multiple interface.
9714 (store_multiple): Update nds32_expand_store_multiple interface.
9715 * config/nds32/nds32-predicates.c
9716 (nds32_valid_multiple_load_store): Rename ...
9717 (nds32_valid_multiple_load_store_p): ... to this and refine
9719 * config/nds32/predicates.md
9720 (nds32_load_multiple_and_update_address_operation): New predicate.
9721 (nds32_store_multiple_and_update_address_operation): New predicate.
9723 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
9724 Chung-Ju Wu <jasonwucj@gmail.com>
9726 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
9727 (combo): New attribute.
9728 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
9730 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
9732 * config/nds32/nds32.opt: Change -mcmodel= default value.
9734 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
9735 Monk Chiang <sh.chiang04@gmail.com>
9736 Chung-Ju Wu <jasonwucj@gmail.com>
9738 * config/nds32/constants.md (unspec_element): New enum.
9739 * config/nds32/constraints.md (Umw): New constraint.
9740 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
9741 * config/nds32/nds32-intrinsic.md: Likewise.
9742 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
9743 (nds32_valid_smw_lwm_base_p): New.
9744 (nds32_output_smw_single_word): New.
9745 (nds32_output_lmw_single_word): New.
9746 (nds32_expand_unaligned_load): New.
9747 (nds32_expand_unaligned_store): New.
9748 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
9749 (nds32_output_smw_single_word): Declare.
9750 (nds32_output_lmw_single_word): Declare.
9751 (nds32_expand_unaligned_load): Declare.
9752 (nds32_expand_unaligned_store): Declare.
9753 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
9754 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
9755 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
9756 NDS32_BUILTIN_UASTORE_DW.
9757 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
9760 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
9761 Kito Cheng <kito.cheng@gmail.com>
9762 Chung-Ju Wu <jasonwucj@gmail.com>
9764 * config/nds32/nds32-intrinsic.c
9765 (nds32_expand_builtin_null_ftype_reg): Delete.
9766 (nds32_expand_builtin_reg_ftype_imm): Ditto.
9767 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
9768 (nds32_read_argument): New.
9769 (nds32_legitimize_target): Ditto.
9770 (nds32_legitimize_argument): Ditto.
9771 (nds32_check_constant_argument): Ditto.
9772 (nds32_expand_unop_builtin): Ditto.
9773 (nds32_expand_unopimm_builtin): Ditto.
9774 (nds32_expand_binop_builtin): Ditto.
9775 (nds32_builtin_decl_impl): Ditto.
9776 (builtin_description): Ditto.
9777 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
9778 (nds32_init_builtins_impl): Ditto.
9779 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
9780 (nds32_builtin_decl): New.
9781 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
9782 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
9784 2018-03-02 Jeff Law <law@redhat.com>
9786 * reorg.c (stop_search_p): Handle DEBUG_INSN.
9787 (redundant_insn, fill_simple_delay_slots): Likewise.
9788 (fill_slots_from_thread): Likewise.
9789 * resource.c (mark_referenced_resources): Likewise.
9790 (mark_set_resources, find_dead_or_set_registers): Likewise.
9792 2018-03-02 Jakub Jelinek <jakub@redhat.com>
9794 * substring-locations.h (format_warning_va): Formatting fix for
9796 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
9798 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
9799 * substring-locations.c: Include intl.h.
9800 (format_warning_va): Turned into small wrapper around
9801 format_warning_n_va, renamed to ...
9802 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
9803 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
9805 (format_warning_at_substring_n): New function.
9806 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
9807 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
9808 format_warning_at_substring with just a shorter name instead of
9809 const function pointer.
9810 (fmtwarn_n): New function.
9811 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
9812 appropriate, get rid of all the fmtstr temporaries, move conditionals
9813 with G_() wrapped string literals directly into fmtwarn arguments,
9814 cast dir.len to (int), formatting fixes.
9816 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
9818 * doc/invoke.texi: Remove "Cilk Plus" references.
9820 2018-03-02 Jakub Jelinek <jakub@redhat.com>
9821 Richard Biener <rguenther@suse.de>
9824 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
9825 for error or warning attributes if CALL_FROM_THUNK_P is set.
9828 2018-03-02 Jakub Jelinek <jakub@redhat.com>
9831 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
9832 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
9835 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
9836 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
9838 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
9839 instead of -1U in last predictors element's probability member.
9841 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
9844 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
9845 arguments if they are comparable.
9847 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
9849 PR tree-optimization/84634
9850 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
9851 masks and masked_loop_p with a single loop_masks, making sure it's
9852 null for bb vectorization.
9854 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
9856 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
9857 (vect_analyze_data_ref_access): Use loop->safe_len rather than
9858 loop->force_vectorize to check whether there is no alias.
9860 2018-03-02 Jakub Jelinek <jakub@redhat.com>
9863 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
9865 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
9867 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
9868 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
9869 instead of a loop around prev_real_insn.
9870 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
9874 * config/i386/i386.c (ix86_print_operand): Use conditional
9875 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
9878 2018-03-02 Richard Biener <rguenther@suse.de>
9880 PR tree-optimization/84427
9881 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
9882 (bitmap_set_subtract_values): Rewrite to handle multiple
9885 (prune_clobbered_mems): Likewise.
9886 (phi_translate): Take edge instead of pred/phiblock.
9887 (phi_translate_1): Likewise.
9888 (phi_translate_set): Likewise. Insert all translated
9889 exprs for a value into the set, keeping possibly multiple
9890 expressions per value.
9891 (compute_antic_aux): Adjust for phi_translate changes.
9892 When intersecting union the expressions and prune those
9893 not in the final value set, keeping possibly multiple
9894 expressions per value. Do not use value-insertion
9895 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
9896 all expressions. Add verification that the value-sets
9897 only shrink during iteration.
9898 (compute_partial_antic_aux): Adjust for the phi_translate changes.
9899 (do_pre_regular_insertion): Likewise.
9900 (do_pre_partial_partial_insertion): Likewise.
9902 2018-03-02 Richard Biener <rguenther@suse.de>
9905 * config/darwin.c (saved_debug_info_level): New static global.
9906 (darwin_asm_lto_start): Disable debug info generation for LTO out.
9907 (darwin_asm_lto_end): Restore debug info generation settings.
9909 2018-03-01 Martin Liska <mliska@suse.cz>
9912 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
9915 2018-03-01 Richard Biener <rguenther@suse.de>
9918 * dwarf2out.c (gen_variable_die): Properly handle late VLA
9919 type annotation with LTO when debug was disabled at compile-time.
9921 2018-03-01 Matthew Fortune <mfortune@gmail.com>
9923 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
9925 (mips_final_postscan_insn): Likewise.
9927 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
9929 PR rtl-optimization/84528
9930 * alias.c (init_alias_target): Add commentary.
9931 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
9932 a unique base value if the frame pointer is not eliminated
9933 to the stack pointer.
9935 2018-03-01 Tom de Vries <tom@codesourcery.com>
9937 PR rtl-optimization/83327
9938 * lra-int.h (hard_regs_spilled_into): Declare.
9939 * lra.c (hard_regs_spilled_into): Define.
9940 (init_reg_info): Init hard_regs_spilled_into.
9941 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
9942 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
9943 (process_bb_lives): Handle hard_regs_spilled_into.
9944 (lra_create_live_ranges_1): Before doing liveness propagation, clear
9945 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
9947 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
9949 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
9950 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
9951 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
9952 * config/rs6000/aix72.h: New file.
9954 2018-02-28 Jakub Jelinek <jakub@redhat.com>
9956 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
9957 instead of warning_at with conditional singular and plural messages
9961 * stor-layout.c (update_alignment_for_field): For
9962 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
9963 && !DECL_PACKED (field), do the alignment update, just use
9964 only desired_align instead of MAX (type_align, desired_align)
9966 (place_field): Don't do known_align < desired_align handling
9967 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
9968 is non-NULL, instead do it after rli->prev_field handling and
9969 only if not within a bitfield word. For DECL_PACKED (field)
9970 use type_align of BITS_PER_UNIT.
9972 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
9974 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
9975 superfluous parentheses and trailing spaces.
9977 2018-02-28 Richard Biener <rguenther@suse.de>
9979 PR tree-optimization/84584
9980 * graphite-scop-detection.c (scop_detection::add_scop): Discard
9981 SCoPs with fake exit edge.
9983 2018-02-28 Martin Liska <mliska@suse.cz>
9986 * timevar.c (timer::print): Fix format to properly print 100%
9989 2018-02-28 Richard Biener <rguenther@suse.de>
9992 * genmatch.c (capture_info::walk_match): Do not mark
9993 captured expressions without operands as expr_p given
9994 they act more like predicates and should be subject to
9995 "lost tail" side-effect preserving.
9997 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
9999 PR rtl-optimization/81611
10000 * auto-inc-dec.c (attempt_change): Move dead note from
10001 mem_insn if it's the next use of regno
10002 (find_address): Take address use of reg holding
10003 non-incremented value. Add parm to limit search to the named
10005 (merge_in_block): Attempt to use a mem insn that is the next
10006 use of the original regno.
10008 2018-02-27 Martin Sebor <msebor@redhat.com>
10011 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
10012 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
10014 2018-02-27 Martin Sebor <msebor@redhat.com>
10016 PR translation/84207
10017 * diagnostic-core.h (warning_n, error_n, inform_n): Change
10018 n argument to unsigned HOST_WIDE_INT.
10019 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
10020 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
10021 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
10022 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
10024 2018-02-27 Richard Biener <rguenther@suse.de>
10026 PR tree-optimization/84512
10027 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
10028 Do not use the estimate returned from record_stmt_cost for
10029 the scalar iteration cost but sum properly using add_stmt_cost.
10031 2018-02-27 Richard Biener <rguenther@suse.de>
10033 PR tree-optimization/84466
10034 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
10035 Adjust last change to less strictly validate use operands.
10037 2018-02-27 Martin Liska <mliska@suse.cz>
10039 PR gcov-profile/84548
10040 * gcov.c (process_file): Allow partial overlap and consider it
10041 also as group functions.
10042 (output_lines): Properly calculate range of lines for a group.
10044 2018-02-27 Martin Liska <mliska@suse.cz>
10046 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
10047 'ggc' suffixes. Change first column width.
10048 (timer::print): Fix formatting of the column.
10050 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
10052 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
10053 preserve inline entry blocks for the sake of debug inline
10054 entry point markers alone.
10055 (remove_unused_locals): Suggest in comments a better place to
10056 force the preservation of inline entry blocks that are
10057 otherwise unused, but do not preserve them.
10059 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
10061 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
10063 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
10066 * config/i386/constraints.md (Bs): Replace
10067 ix86_indirect_branch_register with
10068 TARGET_INDIRECT_BRANCH_REGISTER.
10070 * config/i386/i386.md (indirect_jump): Likewise.
10071 (tablejump): Likewise.
10072 (*sibcall_memory): Likewise.
10073 (*sibcall_value_memory): Likewise.
10074 Peepholes of indirect call and jump via memory: Likewise.
10075 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
10076 (*sibcall_value_GOT_32): Likewise.
10077 * config/i386/predicates.md (indirect_branch_operand): Likewise.
10078 (GOT_memory_operand): Likewise.
10079 (call_insn_operand): Likewise.
10080 (sibcall_insn_operand): Likewise.
10081 (GOT32_symbol_operand): Likewise.
10082 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
10084 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
10086 PR rtl-optimization/83496
10087 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
10088 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
10089 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
10090 redundant insn, if any.
10091 (relax_delay_slots): Likewise.
10092 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
10094 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
10096 PR tree-optimization/83965
10097 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
10098 that grouped statements are part of a reduction chain. Return
10099 true if the statement is not marked as a reduction itself but
10100 is part of a group.
10101 (vect_recog_dot_prod_pattern): Don't check whether the statement
10102 is part of a group here.
10103 (vect_recog_sad_pattern): Likewise.
10104 (vect_recog_widen_sum_pattern): Likewise.
10106 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
10109 * final.c (rest_of_clean_state): Also look for calls inside sequences.
10111 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
10114 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
10116 (ix86_output_indirect_function_return): New prototype.
10117 (ix86_split_simple_return_pop_internal): Likewise.
10118 * config/i386/i386.c (indirect_return_via_cx): New.
10119 (indirect_return_via_cx_bnd): Likewise.
10120 (indirect_thunk_name): Handle return va CX_REG.
10121 (output_indirect_thunk_function): Create alias for
10122 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
10123 (ix86_output_indirect_jmp): Remove the bool argument.
10124 (ix86_output_indirect_function_return): New function.
10125 (ix86_split_simple_return_pop_internal): Likewise.
10126 * config/i386/i386.md (*indirect_jump): Don't pass false
10127 to ix86_output_indirect_jmp.
10128 (*tablejump_1): Likewise.
10129 (simple_return_pop_internal): Change it to define_insn_and_split.
10130 Call ix86_split_simple_return_pop_internal to split it for
10131 -mfunction-return=.
10132 (simple_return_indirect_internal): Call
10133 ix86_output_indirect_function_return instead of
10134 ix86_output_indirect_jmp.
10136 2018-02-26 Jakub Jelinek <jakub@redhat.com>
10139 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
10140 memset and value initialization afterwards.
10142 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
10144 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
10146 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
10149 * common/config/aarch64/aarch64-common.c
10150 (aarch_option_optimization_table[]): Switch
10151 off fomit-frame-pointer
10153 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
10154 Chung-Ju Wu <jasonwucj@gmail.com>
10156 * config/nds32/nds32-multiple.md (load_multiple): Disallow
10158 (store_multiple): Ditto.
10160 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
10162 * config.gcc: Add --with-cpu support for nds32 target.
10163 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
10164 * config/nds32/nds32.opt: Add -mcpu= option.
10166 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
10168 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
10169 isel=yes): Warn for these deprecated options.
10171 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
10173 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
10174 ISA_2_5_MASKS_EMBEDDED.
10176 2018-02-23 Jakub Jelinek <jakub@redhat.com>
10178 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
10179 p->max as pointers rather than using iterative_hash_expr.
10181 2018-02-23 Carl Love <cel@us.ibm.com>
10183 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
10184 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
10186 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
10187 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
10188 P8V_BUILTIN_VEC_VUNSIGNED2.
10190 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
10193 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
10194 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
10196 * lra-constraints.c (curr_insn_transform): Set up
10197 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
10199 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
10200 * lra-eliminations.c (spill_pseudos): Ditto.
10201 (process_insn_for_elimination): Ditto.
10202 * lra-lives.c (reg_early_clobber_p): Use the new macros.
10203 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
10204 LRA_NON_CLOBBERED_ALT.
10206 2018-02-22 Martin Sebor <msebor@redhat.com>
10208 PR tree-optimization/84480
10209 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
10210 to maybe_diag_stxncpy_trunc. Call it.
10211 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
10212 from gimple_fold_builtin_strcpy. Print inlining stack.
10213 (handle_builtin_stxncpy): Print inlining stack.
10214 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
10216 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
10219 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
10220 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
10221 and -fcheck-pointer-bounds are used together.
10222 (indirect_thunk_prefix): New enum.
10223 (indirect_thunk_need_prefix): New function.
10224 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
10225 "_nt" instead of "_bnd" for NOTRACK prefix.
10226 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
10227 (output_indirect_thunk_function): Likewise.
10229 (ix86_code_end): Update output_indirect_thunk_function calls.
10230 (ix86_output_indirect_branch_via_reg): Replace
10231 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
10232 (ix86_output_indirect_branch_via_push): Likewise.
10233 (ix86_output_function_return): Likewise.
10234 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
10235 incompatible with -fcf-protection=branch and
10236 -fcheck-pointer-bounds.
10238 2018-02-22 Steve Ellcey <sellcey@cavium.com>
10241 * config/aarch64/aarch64.c (aarch64_print_address_internal):
10242 Change gcc_assert call to output_operand_lossage.
10244 2018-02-22 Steve Ellcey <sellcey@cavium.com>
10246 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
10248 2018-02-22 DJ Delorie <dj@redhat.com>
10249 Sebastian Perta <sebastian.perta@renesas.com>
10250 Oleg Endo <olegendo@gcc.gnu.org>
10252 * config/rx/rx.c (rx_rtx_costs): New function.
10253 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
10255 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
10257 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
10259 2018-02-22 Martin Liska <mliska@suse.cz>
10262 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
10263 Add "native" as a possible value.
10265 2018-02-22 Martin Liska <mliska@suse.cz>
10268 * config/i386/i386.c (ix86_option_override_internal):
10269 Add "native" as a possible value for -march and -mtune.
10271 2018-02-22 Jakub Jelinek <jakub@redhat.com>
10274 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
10275 to all type variants.
10277 PR tree-optimization/84503
10278 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
10279 width as info->bitpos + info->bitsize - start.
10280 (merged_store_group::merge_overlapping): Simplify width computation.
10281 (check_no_overlap): New function.
10282 (imm_store_chain_info::try_coalesce_bswap): Compute expected
10283 start + width and last_order of the group, fail if check_no_overlap
10285 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
10286 to group if check_no_overlap fails.
10288 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
10290 * config/rs6000/altivec.md: Delete contraint arguments to
10291 define_expand, define_split, and define_peephole2, and in
10292 define_insn_and_split if always unused.
10293 * config/rs6000/darwin.md: Ditto.
10294 * config/rs6000/dfp.md: Ditto.
10295 * config/rs6000/rs6000.md: Ditto.
10296 * config/rs6000/sync.md: Ditto.
10297 * config/rs6000/vector.md: Ditto.
10298 * config/rs6000/vsx.md: Ditto.
10300 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
10302 * config/rs6000/altivec.md: Write output control strings as braced
10303 blocks instead of double-quoted strings.
10304 * config/rs6000/darwin.md: Ditto.
10305 * config/rs6000/rs6000.md: Ditto.
10306 * config/rs6000/vector.md: Ditto.
10307 * config/rs6000/vsx.md: Ditto.
10309 2018-02-21 Jason Merrill <jason@redhat.com>
10311 PR c++/84314 - ICE with templates and fastcall attribute.
10312 * attribs.c (build_type_attribute_qual_variant): Remove assert.
10314 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
10316 * ipa-cp.c (determine_versionability): Fix comment typos.
10318 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
10321 * ipa-cp.c (determine_versionability): Do not version functions caling
10324 2018-02-21 Martin Liska <mliska@suse.cz>
10327 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
10328 Add "native" as a possible value.
10329 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
10330 the macro when native cpu detection is available.
10332 2018-02-21 Martin Liska <mliska@suse.cz>
10335 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
10336 Add "native" as a possible value.
10337 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
10338 when native cpu detection is available.
10340 2018-02-21 Jakub Jelinek <jakub@redhat.com>
10341 Martin Sebor <msebor@redhat.com>
10343 PR tree-optimization/84478
10344 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
10346 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
10347 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
10348 support which is conservatively correct, for 2 only stay conservative
10349 for maxlen. Formatting and comment capitalization fixes. Add STRICT
10350 argument to the 2 argument get_range_strlen, adjust 6 arg
10351 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
10353 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
10354 (gimple_fold_builtin_strlen): Pass true as last argument to
10357 2018-02-20 Martin Sebor <msebor@redhat.com>
10359 PR middle-end/84095
10360 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
10361 (builtin_memref::set_base_and_offset): Same. Handle inner references.
10362 (builtin_memref::builtin_memref): Factor out parts into
10363 set_base_and_offset and call it.
10365 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
10367 PR middle-end/84406
10368 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
10369 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
10370 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
10371 search at the associated MODE_INT.
10373 2018-02-20 Jeff Law <law@redhat.com>
10375 PR middle-end/82123
10376 PR tree-optimization/81592
10377 PR middle-end/79257
10378 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
10379 for range data rather than using global data.
10380 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
10381 range data rather than using global data.
10382 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
10383 pass it to children as needed.
10384 (struct directive::fmtresult): Similarly.
10385 (struct directive::set_width): Similarly.
10386 (struct directive::set_precision): Similarly.
10387 (format_integer, format_directive, parse_directive): Similarly.
10388 (format_none): Accept unnamed vr_values parameter.
10389 (format_percent, format_floating, format_character): Similarly.
10390 (format_string, format_plain): Similarly.
10391 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
10392 the EVRP range analyzer for range data rather than using global data.
10393 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
10394 gimple-ssa-evrp-analyze.h
10395 (class sprintf_dom_walker): Add after_dom_children member function.
10396 Add evrp_range_analyzer member.
10397 (sprintf_dom_walker::before_dom_children): Call into the EVRP
10398 range analyzer as needed.
10399 (sprintf_dom_walker::after_dom_children): New member function.
10400 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
10402 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
10403 (evrp_range_analyzer::pop_to_marker): Likewise.
10405 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
10407 PR tree-optimization/84419
10408 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
10409 with the required type if its current type is compatible but
10412 2018-02-20 Jakub Jelinek <jakub@redhat.com>
10414 PR middle-end/82004
10415 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
10416 after vectorization.
10418 2018-02-20 Martin Liska <mliska@suse.cz>
10421 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
10422 possible values if we don't have a hint.
10424 2018-02-20 Martin Liska <mliska@suse.cz>
10428 * final.c (shorten_branches): Build align_tab array with one
10430 * opts.c (finish_options): Add alignment option limit check.
10431 (MAX_CODE_ALIGN): Likewise.
10432 (MAX_CODE_ALIGN_VALUE): Likewise.
10433 * doc/invoke.texi: Document maximum allowed option value for
10434 all -falign-* options.
10436 2018-02-19 Jakub Jelinek <jakub@redhat.com>
10439 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
10440 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
10441 * var-tracking.c (emit_note_insn_var_location): Remove all references
10442 to NOTE_INSN_CALL_ARG_LOCATION.
10443 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
10444 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
10445 Use copy_rtx_if_shared.
10446 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
10447 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
10448 (dwarf2out_var_location): Remove handling of
10449 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
10451 * final.c (final_scan_insn): Remove all references to
10452 NOTE_INSN_CALL_ARG_LOCATION.
10453 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
10454 before dumping final insns.
10455 * except.c (emit_note_eh_region_end): Remove all references to
10456 NOTE_INSN_CALL_ARG_LOCATION.
10457 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
10458 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
10459 * config/arc/arc.c (hwloop_optimize): Likewise.
10460 * config/arm/arm.c (create_fix_barrier): Likewise.
10461 * config/s390/s390.c (s390_chunkify_start): Likewise.
10462 * config/sh/sh.c (find_barrier): Likewise.
10463 * config/i386/i386.c (rest_of_insert_endbranch,
10464 ix86_seh_fixup_eh_fallthru): Likewise.
10465 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
10466 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
10467 * config/frv/frv.c (frv_function_prologue): Likewise.
10468 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
10470 (note_outside_basic_block_p): Remove all references to
10471 NOTE_INSN_CALL_ARG_LOCATION.
10472 * gengtype.c (adjust_field_rtx_def): Likewise.
10473 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
10475 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
10476 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
10479 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
10482 PR tree-optimization/84452
10483 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
10484 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
10487 2018-02-19 Martin Liska <mliska@suse.cz>
10490 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
10492 2018-02-19 Martin Liska <mliska@suse.cz>
10493 Richard Sandiford <richard.sandiford@linaro.org>
10495 PR tree-optimization/82491
10496 * gimple-fold.c (get_base_constructor): Make earlier bail out
10499 2018-02-19 Carl Love <cel@us.ibm.com>
10501 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
10502 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
10504 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
10505 P8V_BUILTIN_VEC_NEG.
10507 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
10509 * config/rl78/rl78.md (movdf): New define expand.
10511 2018-02-19 Martin Liska <mliska@suse.cz>
10514 * doc/invoke.texi: Fix typo.
10515 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
10517 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
10519 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
10520 handle rs6000_single_float and rs6000_double_float specially for
10523 2018-02-16 Jeff Law <law@redhat.com>
10525 * config/rx/rx.c (add_pop_cfi_notes): New function.;
10526 (pop_regs): Use it.
10528 2018-02-16 Jakub Jelinek <jakub@redhat.com>
10531 * ipa-inline.c (inline_small_functions): Fix a typo.
10533 2018-02-16 Nathan Sidwell <nathan@acm.org>
10535 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
10537 2018-02-16 Carl Love <cel@us.ibm.com>
10539 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
10540 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
10541 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
10542 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
10543 expansion to P8V_BUILTIN_VEC_FLOAT2.
10545 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
10547 PR rtl-optimization/70023
10548 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
10549 src_regno into account.
10551 2018-02-16 Carl Love <cel@us.ibm.com>
10553 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
10554 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
10555 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
10556 * config/rs6000/rs6000.c: Remove case statements for
10557 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
10558 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
10559 and P9V_BUILTIN_VEC_VINSERT4B.
10560 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
10561 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
10562 * config/rs6000/vsx.md:
10563 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
10566 2018-02-16 Carl Love <cel@us.ibm.com>
10568 * config/rs6000/altivec.h: Add builtin names vec_extract4b
10570 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
10572 * config/rs6000/rs6000-c.c: Add the definitions for
10573 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
10574 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
10575 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
10576 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
10577 definition for insert4b and define insn *insert3b_internal.
10578 * doc/extend.texi: Add documentation for vec_extract4b.
10580 2018-02-16 Nathan Sidwell <nathan@acm.org>
10582 * doc/extend.texi (Backwards Compatibility): Mention friend
10583 injection. Note for-scope is deprecated.
10584 * doc/invoke.texi (-ffriend-injection): Deprecate.
10586 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
10588 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
10589 that moved to I2, also allow destinations that are a paradoxical
10590 subreg (instead of a normal reg).
10592 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
10595 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
10598 2018-02-16 Richard Biener <rguenther@suse.de>
10600 PR tree-optimization/84037
10601 PR tree-optimization/84016
10603 * config/i386/i386.c (ix86_builtin_vectorization_cost):
10604 Adjust vec_construct for the fact we need additional higher latency
10605 128bit inserts for AVX256 and AVX512 vector builds.
10606 (ix86_add_stmt_cost): Scale vector construction cost for
10609 2018-02-16 Richard Biener <rguenther@suse.de>
10611 PR tree-optimization/84417
10612 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
10613 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
10614 (non_rewritable_lvalue_p): Likewise, use poly-ints.
10616 2018-02-16 Martin Liska <mliska@suse.cz>
10619 * internal-fn.def (ASAN_CHECK): Set proper flags.
10620 (ASAN_MARK): Likewise.
10622 2018-02-16 Julia Koval <julia.koval@intel.com>
10624 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
10625 from PTA_CANNONLAKE.
10627 2018-02-16 Jakub Jelinek <jakub@redhat.com>
10630 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
10631 Use ++iter rather than iter++ for std::list iterators.
10632 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
10633 defer deleting them until all nodes in the forest are processed. Do
10634 free even leaf nodes. Change to_process into auto_vec.
10637 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
10638 * vec.h (vec_default_construct): Use memset instead of placement new
10639 if BROKEN_VALUE_INITIALIZATION is defined.
10640 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
10641 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
10644 PR rtl-optimization/83723
10645 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
10646 * lra.c (lra_substitute_pseudo): Likewise. If true, use
10647 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
10649 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
10651 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
10653 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
10655 PR rtl-optimization/81443
10656 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
10657 from inner REGs to paradoxical SUBREGs.
10659 2018-02-16 Richard Biener <rguenther@suse.de>
10661 PR tree-optimization/84399
10662 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
10663 For operands we can analyze at their definition make sure we can
10664 analyze them at each use as well.
10666 2018-02-16 Richard Biener <rguenther@suse.de>
10668 PR tree-optimization/84190
10669 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
10670 volatile accesses if the decl isn't volatile.
10672 2018-02-15 Jason Merrill <jason@redhat.com>
10674 PR c++/84314 - ICE with templates and fastcall attribute.
10675 * attribs.c (build_type_attribute_qual_variant): Don't clobber
10676 TYPE_CANONICAL on an existing type.
10678 2018-02-15 Jakub Jelinek <jakub@redhat.com>
10680 PR tree-optimization/84383
10681 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
10682 dstoff nor call operand_equal_p if dstbase is NULL.
10684 PR tree-optimization/84334
10685 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
10686 also a CONSTANT_CLASS_P, punt.
10688 2018-02-14 Jim Wilson <jimw@sifive.com>
10690 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
10691 first SMALL_OPERAND check. New local min_second_step. Move assert
10692 to where locals are set. Add TARGET_RVC support.
10693 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
10695 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
10697 * doc/invoke.texi: Correct -Wformat-overflow code sample.
10699 2018-02-14 Martin Sebor <msebor@redhat.com>
10701 PR tree-optimization/83698
10702 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
10703 arrays constrain the offset range to their bounds.
10704 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
10705 (builtin_access::overlap): Avoid setting the size of overlap if it's
10707 (maybe_diag_overlap): Also consider arrays when deciding what values
10708 of offsets to include in diagnostics.
10710 2018-02-14 Martin Sebor <msebor@redhat.com>
10713 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
10714 that correspond to the kind of a declaration.
10716 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
10719 * config/pa/pa.md: Load address of PIC label using the linkage table
10720 if the label is nonlocal.
10722 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
10724 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
10725 warning message if user requests -maltivec=be.
10726 * doc/invoke.texi: Document deprecation of -maltivec=be.
10728 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
10731 * config/rs6000/rs6000-c.c: Update definitions for
10732 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
10733 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
10735 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
10738 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
10739 add _get_ssp intrinsics. Remove argument from
10740 __builtin_ia32_rdssp[d|q].
10741 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
10742 * config/i386/i386-builtin.def: Remove argument from
10743 __builtin_ia32_rdssp[d|q].
10744 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
10745 ix86_expand_special_args_builtin for _rdssp[d|q].
10746 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
10747 Clear register before usage.
10748 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
10749 Add documentation for new _get_ssp and _inc_ssp intrinsics.
10751 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
10753 PR tree-optimization/84357
10754 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
10755 operand 1 of an ARRAY_REF too.
10757 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
10760 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
10761 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
10763 (set_of_reg): New struct.
10764 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
10765 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
10766 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
10768 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
10769 Split into bitclr, bitset, bitinvert patterns if appropriate.
10770 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
10771 use rx_fuse_in_memory_bitop.
10772 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
10773 to named insn, correct maximum insn length.
10775 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
10778 * machmode.def: Define a complex mode for PARTIAL_INT.
10779 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
10781 * doc/rtl.texi: Document CSPImode.
10782 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
10784 (msp430_hard_regno_nregs_with_padding): Likewise.
10786 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
10789 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
10791 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
10793 PR rtl-optimization/84169
10794 * combine.c (try_combine): New variable split_i2i3. Set it to true if
10795 we generated a parallel as new i3 and we split that to new i2 and i3
10796 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
10797 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
10798 those to i2, not i1. Partially rewrite this scan code.
10800 2018-02-13 Jakub Jelinek <jakub@redhat.com>
10803 * stor-layout.c (place_field): For variable length fields, adjust
10804 offset_align afterwards not just based on the field's alignment,
10805 but also on the size.
10807 PR middle-end/84309
10808 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
10809 of exps and logs in the use_exp2 case.
10811 2018-02-13 Jeff Law <law@redhat.com>
10813 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
10814 entry for "vector".
10816 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
10819 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
10823 * common.opt (gas-loc-support, gas-locview-support): New.
10824 (ginline-points, ginternal-reset-location-views): New.
10825 * doc/invoke.texi: Document them. Use @itemx where intended.
10826 (gvariable-location-views): Adjust.
10827 * target.def (reset_location_view): New.
10828 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
10829 (TARGET_RESET_LOCATION_VIEW): New.
10830 * doc/tm.texi: Rebuilt.
10831 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
10832 (dwarf2out_default_as_locview_support): New.
10833 (output_asm_line_debug_info): Use option variables.
10834 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
10835 (output_loc_list): Likewise.
10836 (add_high_low_attributes): Check option variables.
10837 Don't output entry view attribute in strict mode.
10838 (gen_inlined_subroutine_die): Check option variables.
10839 (dwarf2out_inline_entry): Likewise.
10840 (init_sections_and_labels): Likewise.
10841 (dwarf2out_early_finish): Likewise.
10842 (maybe_reset_location_view): New, from...
10843 (dwarf2out_var_location): ... here. Call it.
10844 * debug.h (dwarf2out_default_as_loc_support): Declare.
10845 (dwarf2out_default_as_locview_support): Declare.
10846 * hooks.c (hook_int_rtx_insn_0): New.
10847 * hooks.h (hook_int_rtx_insn_0): Declare.
10848 * toplev.c (process_options): Take -gas-loc-support and
10849 -gas-locview-support from dwarf2out. Enable
10850 -gvariable-location-views by default only with locview
10851 assembler support. Enable -ginternal-reset-location-views by
10852 default only if the target defines the corresponding hook.
10853 Enable -ginline-points by default if location views are
10854 enabled; force it disabled if statement frontiers are
10856 * tree-inline.c (expand_call_inline): Check option variables.
10857 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
10859 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
10861 PR tree-optimization/84321
10862 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
10863 handling. Also check whether the anti-range contains any values
10864 that satisfy the mask; switch to a VR_RANGE if not.
10866 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
10869 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
10871 2018-02-13 Martin Jambor <mjambor@suse.cz>
10874 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
10875 of call statements, also set location of a load to a temporary.
10877 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
10879 * config/rl78/rl78.c (add_vector_labels): New function.
10880 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
10881 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
10882 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
10883 which checks that no arguments are passed.
10884 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
10885 * doc/extend.texi: Documentation for the new attribute.
10887 2018-02-13 Andreas Schwab <schwab@suse.de>
10889 * config/riscv/linux.h (CPP_SPEC): Define.
10891 2018-02-13 Jakub Jelinek <jakub@redhat.com>
10894 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
10895 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
10896 OPTION_MASK_ISA_AES as first argument to def_builtin_const
10897 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
10898 instead of OPTION_MASK_ISA_PCLMUL as first argument to
10899 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
10900 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
10901 temporarily for AES and PCLMUL builtins.
10903 PR tree-optimization/84339
10904 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
10905 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
10908 PR middle-end/84309
10909 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
10910 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
10911 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
10913 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
10915 * omp-simd-clone.h: New file.
10916 * omp-simd-clone.c: Include omp-simd-clone.h.
10917 (expand_simd_clones): No longer static.
10918 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
10919 cgraph.h and omp-simd-clone.h.
10920 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
10921 (vect_recog_widen_shift_pattern): Formatting fix.
10922 (vect_pattern_recog_1): Don't check optab for calls.
10925 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
10926 operands[2] into a REG before using gen_lowpart on it.
10928 2018-02-12 Jeff Law <law@redhat.com>
10931 * config/sh/sh.c (find_barrier): Consider a sibling call
10934 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
10935 successfully back substituting a reg.
10937 2018-02-12 Richard Biener <rguenther@suse.de>
10939 PR tree-optimization/84037
10940 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
10941 parameter, move visited init to caller.
10942 (vect_slp_analyze_operations): Separate cost from validity
10943 check, initialize visited once for all instances.
10944 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
10946 * tree-vect-stmts.c (vect_model_simple_cost): Make early
10948 (vect_model_promotion_demotion_cost): Likewise.
10949 (vectorizable_bswap): Guard cost modeling with !slp_node
10950 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
10952 (vectorizable_call): Likewise.
10953 (vectorizable_conversion): Likewise.
10954 (vectorizable_assignment): Likewise.
10955 (vectorizable_shift): Likewise.
10956 (vectorizable_operation): Likewise.
10957 (vectorizable_store): Likewise.
10958 (vectorizable_load): Likewise.
10959 (vectorizable_condition): Likewise.
10960 (vectorizable_comparison): Likewise.
10962 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
10965 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
10966 (ASAN_MARK): Fix fnspec to account for return value, change pointer
10967 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
10969 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
10971 PR middle-end/83665
10972 * params.def (inline-min-speedup): Increase from 8 to 15.
10973 (max-inline-insns-auto): Decrease from 40 to 30.
10974 * ipa-split.c (consider_split): Add some buffer for function to
10975 be considered inlining candidate.
10976 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
10979 2018-02-12 Richard Biener <rguenther@suse.de>
10981 PR tree-optimization/84037
10982 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
10983 matched stmts if we cannot swap the non-matched ones.
10985 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
10987 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
10988 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
10989 _mm_maskz_scalef_round_ss): New intrinsics.
10990 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
10991 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
10992 __builtin_ia32_scalefss_round): Remove.
10993 (__builtin_ia32_scalefsd_mask_round,
10994 __builtin_ia32_scalefss_mask_round): New intrinsics.
10995 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
10996 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
10997 ((match_operand:VF_128 2 "<round_nimm_predicate>"
10998 "<round_constraint>")): Changed to ...
10999 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
11000 "<round_scalar_constraint>")): ... this.
11001 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
11002 %0, %1, %2<round_op3>}"): Changed to ...
11003 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
11004 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
11005 %2<round_scalar_mask_op3>}"): ... this.
11006 * config/i386/subst.md (round_scalar_nimm_predicate): New.
11008 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
11010 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
11011 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
11012 (_mm_maskz_sqrt_round_ss): New intrinsics.
11013 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
11014 (__builtin_ia32_sqrtsd_mask_round)
11015 (__builtin_ia32_sqrtss_mask_round): New builtins.
11016 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
11017 (__builtin_ia32_sqrtss_round): Remove.
11018 (__builtin_ia32_sqrtsd_mask_round)
11019 (__builtin_ia32_sqrtss_mask_round): New builtins.
11020 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
11021 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
11022 ((match_operand:VF_128 1 "vector_operand"
11023 "xBm,<round_constraint>")): Changed to ...
11024 ((match_operand:VF_128 1 "vector_operand"
11025 "xBm,<round_scalar_constraint>")): ... this.
11026 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
11027 %0, %2, %<iptr>1<round_op3>}): Changed to ...
11028 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
11029 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
11030 %<iptr>1<round_scalar_mask_op3>}): ... this.
11031 ((set_attr "prefix" "<round_prefix>")): Changed to ...
11032 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
11034 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
11037 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
11038 Cast vec_cmpeq result to correct type.
11039 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
11040 Cast vec_cmpgt result to correct type.
11042 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
11044 * final.c (final_scan_insn_1): Renamed from...
11045 (final_scan_insn): ... this. New wrapper, to recover
11046 seen from the outermost call in recursive ones.
11047 * config/sparc/sparc.c (output_return): Drop seen from call.
11048 (output_sibcall): Likewise.
11049 * config/visium/visium.c (output_branch): Likewise.
11051 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
11053 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
11056 2018-02-10 Alan Modra <amodra@gmail.com>
11059 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
11060 Specify LR as an input.
11062 2018-02-10 Jakub Jelinek <jakub@redhat.com>
11065 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
11066 remove_member_access_dummy_vars): New functions.
11067 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
11068 lower_omp_1, execute_lower_omp): Use them.
11070 PR rtl-optimization/84308
11071 * shrink-wrap.c (spread_components): Release todo vector.
11073 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
11075 PR rtl-optimization/57193
11076 * ira-color.c (struct allocno_color_data): Add member
11077 conflict_allocno_hard_prefs.
11078 (update_conflict_allocno_hard_prefs): New.
11079 (bucket_allocno_compare_func): Add a preference based on
11080 conflict_allocno_hard_prefs.
11081 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
11082 (color_allocnos): Remove a dead code. Initiate
11083 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
11085 2018-02-09 Jakub Jelinek <jakub@redhat.com>
11088 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
11089 constraint from =wa to wa. Avoid a subreg on the output operand,
11090 instead use a pseudo and subreg it in a move.
11091 (p9_xxbrd_<mode>): Changed to ...
11092 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
11093 (p9_xxbrd_v2df): New expander.
11094 (p9_xxbrw_<mode>): Changed to ...
11095 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
11096 (p9_xxbrw_v4sf): New expander.
11098 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
11100 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
11102 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
11105 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
11106 multiply in 32-bit mode.
11107 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
11108 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
11111 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
11113 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
11114 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
11115 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
11116 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
11118 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
11121 * dwarf2out.c (is_trivial_indirect_ref): New function.
11122 (dwarf2out_late_global_decl): Do not generate a location
11123 attribute for variables that have a non-trivial DECL_VALUE_EXPR
11124 and that are not defined in the current unit.
11126 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
11128 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
11129 instead of a libcall for UNORDERED.
11131 2018-02-09 Tamar Christina <tamar.christina@arm.com>
11134 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
11135 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
11137 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
11140 * config/s390/s390.c (s390_set_current_function): Invoke
11141 s390_indirect_branch_settings also if fndecl didn't change.
11143 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
11145 * config/rs6000/rs6000.md (blockage): Set length to zero.
11147 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
11149 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
11151 2018-02-09 Jakub Jelinek <jakub@redhat.com>
11154 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
11155 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
11159 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
11160 PARALLEL incoming that failed vt_get_decl_and_offset check.
11162 PR middle-end/84237
11163 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
11164 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
11166 (get_variable_section): For decls in named .bss* sections pass true as
11167 second argument to bss_initializer_p.
11169 2018-02-09 Marek Polacek <polacek@redhat.com>
11170 Jakub Jelinek <jakub@redhat.com>
11173 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
11174 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
11175 Sync some changes from cxx_fold_indirect_ref.
11177 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
11179 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
11181 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
11182 (BLOCK_INLINE_ENTRY_LABEL): New.
11183 (dwarf2out_var_location): Disregard inline entry markers.
11184 (inline_entry_data): New struct.
11185 (inline_entry_data_hasher): New hashtable type.
11186 (inline_entry_data_hasher::hash): New.
11187 (inline_entry_data_hasher::equal): New.
11188 (inline_entry_data_table): New variable.
11189 (add_high_low_attributes): Add DW_AT_entry_pc and
11190 DW_AT_GNU_entry_view attributes if a pending entry is found
11191 in inline_entry_data_table. Add old entry_pc attribute only
11192 if debug nonbinding markers are disabled.
11193 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
11194 markers are enabled.
11195 (block_within_block_p, dwarf2out_inline_entry): New.
11196 (dwarf2out_finish): Check that no entries remained in
11197 inline_entry_data_table.
11198 * final.c (reemit_insn_block_notes): Handle inline entry notes.
11199 (final_scan_insn, notice_source_line): Likewise.
11200 (rest_of_clean_state): Skip inline entry markers.
11201 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
11203 * gimple.c (gimple_build_debug_inline_entry): New.
11204 * gimple.h (enum gimple_debug_subcode): Add
11205 GIMPLE_DEBUG_INLINE_ENTRY.
11206 (gimple_build_debug_inline_entry): Declare.
11207 (gimple_debug_inline_entry_p): New.
11208 (gimple_debug_nonbind_marker_p): Adjust.
11209 * insn-notes.def (INLINE_ENTRY): New.
11210 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
11211 inline entry marker notes.
11212 (print_insn): Likewise.
11213 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
11214 (INSN_DEBUG_MARKER_KIND): Likewise.
11215 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
11216 * tree-inline.c (expand_call_inline): Build and insert
11217 debug_inline_entry stmt.
11218 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
11219 inline entry blocks early, if nonbind markers are enabled.
11220 (dump_scope_block): Dump fragment info.
11221 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
11222 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
11223 (gimple_build_debug_inline_entry): New.
11224 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
11225 Enable/disable inline entry points too.
11226 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
11227 (DEBUG_INSN): Describe inline entry markers.
11229 * common.opt (gvariable-location-views): New.
11230 (gvariable-location-views=incompat5): New.
11231 * config.in: Rebuilt.
11232 * configure: Rebuilt.
11233 * configure.ac: Test assembler for view support.
11234 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
11235 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
11236 * dwarf2out.c (var_loc_view): New typedef.
11237 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
11238 (dwarf2out_locviews_in_attribute): New.
11239 (dwarf2out_locviews_in_loclist): New.
11240 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
11241 (enum dw_line_info_opcode): Add LI_adv_address.
11242 (struct dw_line_info_table): Add view.
11243 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
11244 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
11245 (zero_view_p): New variable.
11246 (ZERO_VIEW_P): New macro.
11247 (output_asm_line_debug_info): New.
11248 (struct var_loc_node): Add view.
11249 (add_AT_view_list, AT_loc_list): New.
11250 (add_var_loc_to_decl): Add view param. Test it against last.
11251 (new_loc_list): Add view params. Record them.
11252 (AT_loc_list_ptr): Handle loc and view lists.
11253 (view_list_to_loc_list_val_node): New.
11254 (print_dw_val): Handle dw_val_class_view_list.
11255 (size_of_die): Likewise.
11256 (value_format): Likewise.
11257 (loc_list_has_views): New.
11258 (gen_llsym): Set vl_symbol too.
11259 (maybe_gen_llsym, skip_loc_list_entry): New.
11260 (dwarf2out_maybe_output_loclist_view_pair): New.
11261 (output_loc_list): Output view list or entries too.
11262 (output_view_list_offset): New.
11263 (output_die): Handle dw_val_class_view_list.
11264 (output_dwarf_version): New.
11265 (output_compilation_unit_header): Use it.
11266 (output_skeleton_debug_sections): Likewise.
11267 (output_rnglists, output_line_info): Likewise.
11268 (output_pubnames, output_aranges): Update version comments.
11269 (output_one_line_info_table): Output view numbers in asm comments.
11270 (dw_loc_list): Determine current endview, pass it to new_loc_list.
11271 Call maybe_gen_llsym.
11272 (loc_list_from_tree_1): Adjust.
11273 (add_AT_location_description): Create view list attribute if
11274 needed, check it's absent otherwise.
11275 (convert_cfa_to_fb_loc_list): Adjust.
11276 (maybe_emit_file): Call output_asm_line_debug_info for test.
11277 (dwarf2out_var_location): Reset views as needed. Precompute
11278 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
11279 attribute. Set view.
11280 (new_line_info_table): Reset next view.
11281 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
11282 (dwarf2out_source_line): Likewise. Output view resets and labels to
11283 the assembler, or select appropriate line info opcodes.
11284 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
11285 (optimize_string_length): Catch it. Adjust.
11286 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
11287 dw_val_class_view_list, and remove it if no longer needed.
11288 (hash_loc_list): Hash view numbers.
11289 (loc_list_hasher::equal): Compare them.
11290 (optimize_location_lists): Check whether a view list symbol is
11291 needed, and whether the locview attribute is present, and
11292 whether they match. Remove the locview attribute if no longer
11294 (index_location_lists): Call skip_loc_list_entry for test.
11295 (dwarf2out_finish): Call output_asm_line_debug_info for test.
11296 Use output_dwarf_version.
11297 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
11298 (struct dw_val_node): Add val_view_list.
11299 * final.c (SEEN_NEXT_VIEW): New.
11300 (set_next_view_needed): New.
11301 (clear_next_view_needed): New.
11302 (maybe_output_next_view): New.
11303 (final_start_function): Rename to...
11304 (final_start_function_1): ... this. Take pointer to FIRST,
11305 add SEEN parameter. Emit param bindings in the initial view.
11306 (final_start_function): Reintroduce SEEN-less interface.
11307 (final): Rename to...
11308 (final_1): ... this. Take SEEN parameter. Output final pending
11309 next view at the end.
11310 (final): Reintroduce seen-less interface.
11311 (final_scan_insn): Output pending next view before switching
11312 sections or ending a block. Mark the next view as needed when
11313 outputting variable locations. Notify debug backend of section
11314 changes, and of location view changes.
11315 (rest_of_handle_final): Adjust.
11316 * toplev.c (process_options): Autodetect value for debug variable
11317 location views option. Warn on incompat5 without -gdwarf-5.
11318 * doc/invoke.texi (gvariable-location-views): New.
11319 (gvariable-location-views=incompat5): New.
11320 (gno-variable-location-views): New.
11322 2018-02-08 David Malcolm <dmalcolm@redhat.com>
11324 PR tree-optimization/84136
11325 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
11326 that the result of find_edge is non-NULL.
11328 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
11331 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
11332 storing integer register in SImode. Fix cost of 256 and 512
11333 byte aligned SSE register store.
11335 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
11337 * config/i386/i386.c (ix86_multiplication_cost): Fix
11338 multiplication cost for TARGET_AVX512DQ.
11340 2018-02-08 Marek Polacek <polacek@redhat.com>
11342 PR tree-optimization/84238
11343 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
11346 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
11348 PR tree-optimization/84265
11349 * tree-vect-stmts.c (vectorizable_store): Don't treat
11350 VMAT_CONTIGUOUS accesses as grouped.
11351 (vectorizable_load): Likewise.
11353 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
11355 PR tree-optimization/81635
11356 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
11357 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
11358 (test_round_for_mask): New functions.
11359 (wide_int_cc_tests): Call test_round_for_mask.
11360 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
11361 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
11362 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
11363 range returned by get_range_info.
11365 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
11368 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
11369 * symtab.c: Include builtins.h
11370 (symtab_node::output_to_lto_symbol_table_p): Move here
11371 from lto-streamer-out.c:output_symbol_p.
11372 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
11373 (output_symbol_p): Move all logic to symtab.c
11374 (produce_symtab): Update.
11376 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
11378 * config/s390/s390-opts.h (enum indirect_branch): Define.
11379 * config/s390/s390-protos.h (s390_return_addr_from_memory)
11380 (s390_indirect_branch_via_thunk)
11381 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
11382 (enum s390_indirect_branch_type): Define.
11383 * config/s390/s390.c (struct s390_frame_layout, struct
11384 machine_function): Remove.
11385 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
11386 (indirect_branch_table_label_no, indirect_branch_table_name):
11388 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
11389 (enum s390_indirect_branch_option): Define.
11390 (s390_return_addr_from_memory): New function.
11391 (s390_handle_string_attribute): New function.
11392 (s390_attribute_table): Add new attribute handler.
11393 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
11394 (s390_indirect_branch_via_thunk): New function.
11395 (s390_indirect_branch_via_inline_thunk): New function.
11396 (s390_function_ok_for_sibcall): When jumping via thunk disallow
11397 sibling call optimization for non z10 compiles.
11398 (s390_emit_call): Force indirect branch target to be a single
11399 register. Add r1 clobber for non-z10 compiles.
11400 (s390_emit_epilogue): Emit return jump via return_use expander.
11401 (s390_reorg): Handle JUMP_INSNs as execute targets.
11402 (s390_option_override_internal): Perform validity checks for the
11403 new command line options.
11404 (s390_indirect_branch_attrvalue): New function.
11405 (s390_indirect_branch_settings): New function.
11406 (s390_set_current_function): Invoke s390_indirect_branch_settings.
11407 (s390_output_indirect_thunk_function): New function.
11408 (s390_code_end): Implement target hook.
11409 (s390_case_values_threshold): Implement target hook.
11410 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
11412 * config/s390/s390.h (struct s390_frame_layout)
11413 (struct machine_function): Move here from s390.c.
11414 (TARGET_INDIRECT_BRANCH_NOBP_RET)
11415 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
11416 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
11417 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
11418 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
11419 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
11420 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
11421 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
11422 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
11423 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
11424 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
11425 (mnemonic attribute): Add values which aren't recognized
11427 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
11428 pattern for branch conversion. Fix mnemonic attribute.
11429 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
11430 indirect branch via thunk if requested.
11431 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
11432 ("*indirect_jump"): Disable for branch conversion using out of
11434 ("indirect_jump_via_thunk<mode>_z10")
11435 ("indirect_jump_via_thunk<mode>")
11436 ("indirect_jump_via_inlinethunk<mode>_z10")
11437 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
11438 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
11439 ("casesi_jump_via_inlinethunk<mode>_z10")
11440 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
11441 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
11442 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
11443 ("*indirect2_jump"): Disable for branch conversion.
11444 ("casesi_jump"): Turn into expander and expand patterns for branch
11446 ("return_use"): New expander.
11447 ("*return"): Emit return via thunk and rename it to ...
11448 ("*return<mode>"): ... this one.
11449 * config/s390/s390.opt: Add new options and and enum for the
11452 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
11454 * lra-constraints.c (match_reload): Unconditionally use
11455 gen_lowpart_SUBREG, rather than selecting between that
11456 and equivalent gen_rtx_SUBREG code.
11458 2018-02-08 Richard Biener <rguenther@suse.de>
11460 PR tree-optimization/84233
11461 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
11462 changed flag instead of boguously re-using phi_inserted.
11464 2018-02-08 Martin Jambor <mjambor@suse.cz>
11466 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
11467 static local variables.
11469 2018-02-08 Richard Biener <rguenther@suse.de>
11471 PR tree-optimization/84278
11472 * tree-vect-stmts.c (vectorizable_store): When looking for
11473 smaller vector types to perform grouped strided loads/stores
11474 make sure the mode is supported by the target.
11475 (vectorizable_load): Likewise.
11477 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
11479 * config/aarch64/aarch64.c (aarch64_components_for_bb):
11480 Increase LDP/STP opportunities by adding adjacent callee-saves.
11482 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
11484 PR rtl-optimization/84068
11485 PR rtl-optimization/83459
11486 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
11488 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
11490 PR tree-optimization/84224
11491 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
11492 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
11493 non-zero arguments.
11495 2018-02-07 Iain Sandoe <iain@codesourcery.com>
11498 * config/rs6000/altivec.md (*restore_world): Remove LR use.
11499 * config/rs6000/predicates.md (restore_world_operation): Adjust op
11500 count, remove one USE.
11502 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
11504 * doc/install.texi (Configuration): Document the
11505 --with-long-double-format={ibm,ieee} PowerPC configuration
11509 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
11510 Convert from define_expand to be define_insn_and_split. Rework
11511 float/double/_Float128 conversions to QI/HI/SImode to work with
11512 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
11513 conversions to QI/HImode types did a store and then a load to
11514 truncate the value. For conversions to VSX registers, don't split
11515 the insn, instead emit the code directly. Use the code iterator
11516 any_fix to combine signed and unsigned conversions.
11517 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
11518 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
11519 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
11520 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
11521 (fix_<mode>di2_hw): Likewise.
11522 (fixuns_<mode>di2_hw): Likewise.
11523 (fix_<mode>si2_hw): Likewise.
11524 (fixuns_<mode>si2_hw): Likewise.
11525 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
11526 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
11527 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
11528 fix<uns>_trunc<SFDF:mode>si2_p8.
11529 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
11530 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
11531 (fix<uns>_<mode>_mem): Likewise.
11532 (fctiw<u>z_<mode>_mem): Likewise.
11533 (fix<uns>_<mode>_mem): Likewise.
11534 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
11535 the register allocator from doing a direct move to the GPRs to do
11536 a store, and instead use the ISA 3.0 store byte/half-word from
11537 vector register instruction. For IEEE 128-bit floating point,
11538 also optimize stores of 32-bit ints.
11539 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
11541 2018-02-07 Alan Hayward <alan.hayward@arm.com>
11543 * genextract.c (push_pathstr_operand): New function to support
11545 (walk_rtx): Call push_pathstr_operand.
11546 (print_path): Support [a-zA-Z].
11548 2018-02-07 Richard Biener <rguenther@suse.de>
11550 PR tree-optimization/84037
11551 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
11552 (cse_and_gimplify_to_preheader): Declare.
11553 (vect_get_place_in_interleaving_chain): Likewise.
11554 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
11556 (_loop_vec_info::~_loop_vec_info): Delete it.
11557 (cse_and_gimplify_to_preheader): New function.
11558 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
11559 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
11560 (vectorizable_load): Likewise. For grouped stores always base
11561 the IV on the first element.
11562 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
11563 condition before gimplifying.
11565 2018-02-07 Jakub Jelinek <jakub@redhat.com>
11567 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
11568 *DIV_EXPR and *MOD_EXPR.
11570 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
11573 * config/i386/i386.c (ix86_option_override_internal): Mask out
11574 the CF_SET bit when checking -fcf-protection.
11576 2018-02-07 Tom de Vries <tom@codesourcery.com>
11579 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
11582 2018-02-07 Richard Biener <rguenther@suse.de>
11584 PR tree-optimization/84204
11585 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
11588 PR tree-optimization/84205
11589 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
11590 special-case isl_ast_op_zdiv_r.
11592 PR tree-optimization/84223
11593 * graphite-scop-detection.c (gather_bbs::before_dom_children):
11594 Only add conditions from within the region.
11595 (gather_bbs::after_dom_children): Adjust.
11597 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
11600 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
11601 * config/avr/avr.md: Only post-reload split REG-REG moves if
11602 either register is GENERAL_REG_P.
11604 2018-02-07 Jakub Jelinek <jakub@redhat.com>
11606 PR tree-optimization/84235
11607 * tree-ssa-scopedtables.c
11608 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
11609 if the subtraction is performed in floating point type where NaNs are
11610 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
11611 build 1. Formatting fix.
11613 2018-02-06 Jakub Jelinek <jakub@redhat.com>
11616 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
11617 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
11618 and skip it regardless of bb boundaries. Use CALL_P macro,
11619 don't test INSN_P (insn) together with CALL_P or JUMP_P check
11620 unnecessarily, formatting fix.
11622 2018-02-06 Michael Collison <michael.collison@arm.com>
11624 * config/arm/thumb2.md:
11625 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
11626 (*thumb_mov_notscc): Ditto.
11628 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
11631 * config/rs6000/rs6000.md (su code attribute): Use "u" for
11632 unsigned_fix, not "s".
11634 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11636 * configure.ac (gcc_fn_eh_frame_ro): New function.
11637 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
11638 correct .eh_frame permissions.
11639 * configure: Regenerate.
11641 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
11643 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
11644 irrelevant options.
11646 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11648 * config/rs6000/rs6000.c (rs6000_option_override_internal):
11649 Display warning message for -mno-speculate-indirect-jumps.
11651 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
11653 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
11655 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
11657 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
11659 PR tree-optimization/84225
11660 * tree-eh.c (find_trapping_overflow): Only call
11661 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
11663 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
11666 * config/i386/i386.c: Reimplement the check of possible options
11667 -mibt/-mshstk conbination. Change error messages.
11668 * doc/invoke.texi: Fix a typo: remove extra '='.
11670 2018-02-06 Marek Polacek <polacek@redhat.com>
11672 PR tree-optimization/84228
11673 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
11675 2018-02-06 Tamar Christina <tamar.christina@arm.com>
11678 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
11679 emitted arch directives.
11680 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
11681 __ARM_FEATURE_COPROC before changing architectures.
11683 2018-02-06 Richard Biener <rguenther@suse.de>
11685 * config/i386/i386.c (print_reg): Fix typo.
11686 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
11688 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
11690 * configure: Regenerate.
11692 2018-02-05 Martin Sebor <msebor@redhat.com>
11694 PR tree-optimization/83369
11695 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
11698 2018-02-05 Martin Liska <mliska@suse.cz>
11700 * doc/invoke.texi: Cherry-pick upstream r323995.
11702 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
11704 * ira.c (ira_init_register_move_cost): Adjust comment.
11706 2018-02-05 Martin Liska <mliska@suse.cz>
11708 PR gcov-profile/84137
11709 * doc/gcov.texi: Fix typo in documentation.
11711 2018-02-05 Martin Liska <mliska@suse.cz>
11713 PR gcov-profile/83879
11714 * doc/gcov.texi: Document necessity of --dynamic-list-data when
11715 using dlopen functionality.
11717 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
11719 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
11720 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
11721 _mm_maskz_range_ss, _mm_mask_range_round_ss,
11722 _mm_maskz_range_round_ss): New intrinsics.
11723 (__builtin_ia32_rangesd128_round)
11724 (__builtin_ia32_rangess128_round): Remove.
11725 (__builtin_ia32_rangesd128_mask_round,
11726 __builtin_ia32_rangess128_mask_round): New builtins.
11727 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
11728 __builtin_ia32_rangess128_round): Remove.
11729 (__builtin_ia32_rangesd128_mask_round,
11730 __builtin_ia32_rangess128_mask_round): New builtins.
11731 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
11732 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
11733 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
11734 "<round_saeonly_constraint>")): Changed to ...
11735 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
11736 "<round_saeonly_scalar_constraint>")): ... this.
11737 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
11738 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
11739 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
11740 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
11741 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
11743 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
11745 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
11747 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
11748 Remove all values except native, 8540 and 8548.
11750 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
11752 * config/i386/i386.c (ix86_output_function_return): Pass
11753 INVALID_REGNUM, instead of -1, as invalid register number to
11754 indirect_thunk_name and output_indirect_thunk.
11756 2018-02-02 Julia Koval <julia.koval@intel.com>
11758 * config.gcc: Add -march=icelake.
11759 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
11760 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
11761 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
11762 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
11763 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
11764 (processor_target_table): Add icelake.
11765 (ix86_option_override_internal): Handle new PTAs.
11766 (get_builtin_code_for_version): Handle icelake.
11767 (M_INTEL_COREI7_ICELAKE): New.
11768 (fold_builtin_cpu): Handle icelake.
11769 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
11770 * doc/invoke.texi: Add -march=icelake.
11772 2018-02-02 Julia Koval <julia.koval@intel.com>
11774 * config/i386/i386.c (ix86_option_override_internal): Change flags type
11775 to wide_int_bitmask.
11776 * wide-int-bitmask.h: New.
11778 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
11781 * config/i386/i386.md: Replace Pmode with word_mode in
11782 builtin_setjmp_setup and builtin_longjmp to support x32.
11784 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
11788 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
11790 (rs6000_supported_cpu_names): New static variable.
11791 (linux_cpu_translation_table): Likewise.
11792 (elf_platform) <cpu>: Define new static variable and use it.
11793 Translate kernel AT_PLATFORM name to canonical name if needed.
11794 Error if platform name is unknown.
11796 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
11799 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
11801 2018-02-01 Jeff Law <law@redhat.com>
11804 * config/i386/i386.c (release_scratch_register_on_entry): Add new
11805 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
11806 the scratch if RELEASE_VIA_POP is false.
11807 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
11808 If we have to save a temporary register, decrement SIZE appropriately.
11809 Pass new arguments to release_scratch_register_on_entry.
11810 (ix86_adjust_stack_and_probe): Likewise.
11811 (ix86_emit_probe_stack_range): Pass new arguments to
11812 release_scratch_register_on_entry.
11814 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
11816 PR rtl-optimization/84157
11817 * combine.c (change_zero_ext): Use REG_P predicate in
11818 front of HARD_REGISTER_P predicate.
11820 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
11822 * config/avr/avr.c (avr_option_override): Move disabling of
11823 -fdelete-null-pointer-checks to...
11824 * common/config/avr/avr-common.c (avr_option_optimization_table):
11827 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11829 PR tree-optimization/81635
11830 * tree-data-ref.c (split_constant_offset_1): For types that
11831 wrap on overflow, try to use range info to prove that wrapping
11834 2018-02-01 Renlin Li <renlin.li@arm.com>
11837 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
11838 TAILCALL_ADDR_REGS.
11839 (aarch64_register_move_cost): Likewise.
11840 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
11841 TAILCALL_ADDR_REGS.
11842 (REG_CLASS_NAMES): Likewise.
11843 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
11844 TAILCALL_ADDR_REGS. Remove IP registers.
11845 * config/aarch64/aarch64.md (Ucs): Update register constraint.
11847 2018-02-01 Richard Biener <rguenther@suse.de>
11849 * domwalk.h (dom_walker::dom_walker): Add additional constructor
11850 for specifying RPO order and allow NULL for that.
11851 * domwalk.c (dom_walker::dom_walker): Likewise.
11852 (dom_walker::walk): Handle NULL RPO order.
11853 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
11855 (rewrite_update_dom_walker): Likewise.
11856 (mark_def_dom_walker): Likewise.
11858 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11860 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
11861 (aarch64_maybe_expand_sve_subreg_move): Declare.
11862 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
11863 * config/aarch64/predicates.md (aarch64_any_register_operand): New
11865 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
11866 that are semantically a reverse operation.
11867 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
11868 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
11869 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
11871 (aarch64_can_change_mode_class): For big-endian, forbid changes
11872 between two SVE modes if they have different element sizes.
11874 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11876 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
11877 the TImode handling for big-endian targets.
11879 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11881 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
11882 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
11884 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
11885 Remove BSWAP handing for big-endian targets and use the form of
11886 LD1RQ appropariate for the mode.
11888 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11890 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
11891 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
11892 duplicated element.
11894 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11897 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
11898 check for operands that need to go through aarch64_sve_reload_be.
11900 2018-02-01 Jakub Jelinek <jakub@redhat.com>
11902 PR tree-optimization/81661
11903 PR tree-optimization/84117
11904 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
11905 * tree-eh.c: Include gimplify.h.
11906 (find_trapping_overflow, replace_trapping_overflow,
11907 rewrite_to_non_trapping_overflow): New functions.
11908 * tree-vect-loop.c: Include tree-eh.h.
11909 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
11910 * tree-data-ref.c: Include tree-eh.h.
11911 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
11913 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
11915 PR rtl-optimization/84123
11916 * combine.c (change_zero_ext): Check if hard register satisfies
11917 can_change_dest_mode before calling gen_lowpart_SUBREG.
11919 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
11922 * ira.c (ira_init_register_move_cost): Remove assert.
11924 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
11926 PR rtl-optimization/84071
11927 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
11928 * doc/tm.texi: Regenerate.
11930 2018-01-31 Richard Biener <rguenther@suse.de>
11932 PR tree-optimization/84132
11933 * tree-data-ref.c (analyze_miv_subscript): Properly
11934 check whether evolution_function_is_affine_multivariate_p
11935 before calling gcd_of_steps_may_divide_p.
11937 2018-01-31 Julia Koval <julia.koval@intel.com>
11940 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
11941 * config/i386/i386.md (rdpid_rex64) New.
11942 (rdpid): Make 32bit only.
11944 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
11947 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
11948 an IDENTIFIER_NODE for FUNCTION_TYPE's.
11950 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
11953 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
11955 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
11957 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
11959 PR rtl-optimization/84071
11960 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
11961 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
11963 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
11965 * config/arc/arc.c (arc_handle_aux_attribute): New function.
11966 (arc_attribute_table): Add 'aux' attribute.
11967 (arc_in_small_data_p): Consider aux like variables.
11968 (arc_is_aux_reg_p): New function.
11969 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
11970 (arc_get_aux_arg): New function.
11971 (prepare_move_operands): Handle aux-register access.
11972 (arc_handle_aux_attribute): New function.
11973 * doc/extend.texi (ARC Variable attributes): Add subsection.
11975 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
11977 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
11978 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
11979 (arc_attribute_table): Add 'uncached' attribute.
11980 (arc_print_operand): Print '.di' flag for uncached memory
11982 (arc_in_small_data_p): Do not consider for small data the uncached
11984 (arc_is_uncached_mem_p): New function.
11985 * config/arc/predicates.md (compact_store_memory_operand): Check
11986 for uncached memory accesses.
11987 (nonvol_nonimm_operand): Likewise.
11988 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
11990 2018-01-31 Jakub Jelinek <jakub@redhat.com>
11993 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
11994 falign-loops=): Add Optimization flag.
11996 2018-01-30 Jeff Law <law@redhat.com>
11999 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
12000 INT_REGISTERS_SAVED. Check it prior to calling
12001 get_scratch_register_on_entry.
12002 (ix86_adjust_stack_and_probe): Similarly.
12003 (ix86_emit_probe_stack_range): Similarly.
12004 (ix86_expand_prologue): Corresponding changes.
12006 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12009 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
12010 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
12012 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
12015 * lra-constraints.c (curr_insn_transform): Process AND in the
12018 2018-01-30 Jakub Jelinek <jakub@redhat.com>
12020 PR rtl-optimization/83986
12021 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
12022 dependence against last_pending_memory_flush in addition to
12023 pending_jump_insns.
12025 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
12027 PR tree-optimization/81611
12028 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
12031 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12034 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
12037 2018-01-30 Richard Biener <rguenther@suse.de>
12038 Jakub Jelinek <jakub@redhat.com>
12040 PR tree-optimization/84111
12041 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
12042 inner loops added during recursion, as they don't have up-to-date
12045 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
12048 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
12049 (can_inline_edge_by_limits_p): ... here.
12050 (can_early_inline_edge_p, check_callers,
12051 update_caller_keys, update_callee_keys, recursive_inlining,
12052 add_new_edges_to_heap, speculation_useful_p,
12053 inline_small_functions,
12054 inline_small_functions, flatten_function,
12055 inline_to_all_callers_1): Update.
12057 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
12059 * profile-count.c (profile_count::combine_with_ipa_count): Handle
12062 2018-01-30 Richard Biener <rguenther@suse.de>
12064 PR tree-optimization/83008
12065 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
12066 invariant and constant vector uses in stmts when they need
12067 more than one stmt.
12069 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12072 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
12073 * configure: Regenerate.
12075 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
12077 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
12079 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
12080 Use gen_rtx_REG rather than gen_lowpart.
12082 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
12084 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
12085 rather than 0 when creating partial subregs.
12087 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
12089 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
12092 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
12095 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
12096 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
12097 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
12098 flags. This restores the settings used before the 2017-07-24.
12099 Turning off pre increment/decrement/modify allows IVOPTS to
12100 optimize DF/SF loops where the index is an int.
12102 2018-01-29 Richard Biener <rguenther@suse.de>
12103 Kelvin Nilsen <kelvin@gcc.gnu.org>
12106 * tree-vect-stmts.c (vectorizable_call): Don't call
12107 targetm.vectorize_builtin_md_vectorized_function if callee is
12110 2018-01-22 Carl Love <cel@us.ibm.com>
12112 * doc/extend.tex: Fix typo in second arg in
12113 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
12115 2018-01-29 Richard Biener <rguenther@suse.de>
12117 PR tree-optimization/84086
12118 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
12119 (flush_ssaname_freelist): When SSA names were released reset
12120 the SCEV hash table.
12122 2018-01-29 Richard Biener <rguenther@suse.de>
12124 PR tree-optimization/84057
12125 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
12126 removed paths when removing edges.
12128 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
12130 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
12131 -mfunction-return=@var{choice}.
12133 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
12135 PR diagnostic/84034
12136 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
12137 Handle CR like TAB.
12138 (layout::print_source_line): Likewise.
12139 (test_get_line_width_without_trailing_whitespace): Add test cases.
12141 2018-01-27 Jakub Jelinek <jakub@redhat.com>
12143 PR middle-end/84040
12144 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
12147 2018-01-26 Jim Wilson <jimw@sifive.com>
12149 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
12151 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
12154 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12156 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
12157 and CMP + SUB-immediate -> SUBS.
12159 2018-01-26 Martin Sebor <msebor@redhat.com>
12161 PR tree-optimization/83896
12162 * tree-ssa-strlen.c (get_string_len): Rename...
12163 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
12164 Avoid assuming length is constant.
12165 (handle_char_store): Use HOST_WIDE_INT for string length.
12167 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
12170 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
12171 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
12173 2018-01-26 Richard Biener <rguenther@suse.de>
12175 PR rtl-optimization/84003
12176 * dse.c (record_store): Only record redundant stores when
12177 the earlier store aliases at least all accesses the later one does.
12179 2018-01-26 Jakub Jelinek <jakub@redhat.com>
12181 PR rtl-optimization/83985
12182 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
12183 REG_CFA_RESTORE insns.
12184 (delete_unmarked_insns): Don't ignore separate shrink wrapping
12185 REG_CFA_RESTORE insns here.
12188 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
12189 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
12191 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
12193 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
12194 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
12195 (arc_init): Likewise.
12196 (arc_override_options): Likewise.
12197 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
12199 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
12201 * config/arc/arc.h (TARGET_DBNZ): Define.
12202 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
12203 properly set the tune attribute.
12204 (dbnz): Use TARGET_DBNZ guard.
12205 * config/arc/arc.opt (mtune): Add core3 option.
12207 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
12209 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
12210 recognize new pic like addresses.
12211 (arc_delegitimize_address): Clean up.
12213 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
12215 * config/arc/arc-arches.def: Option mrf16 valid for all
12217 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
12218 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
12219 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
12220 * config/arc/arc-tables.opt: Regenerate.
12221 * config/arc/arc.c (arc_conditional_register_usage): Handle
12222 reduced register file case.
12223 (arc_file_start): Set must have build attributes.
12224 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
12225 mrf16 option value.
12226 * config/arc/arc.opt (mrf16): Add new option.
12227 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
12228 * config/arc/genmultilib.awk: Handle new mrf16 option.
12229 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
12230 * config/arc/t-multilib: Regenerate.
12231 * doc/invoke.texi (ARC Options): Document mrf16 option.
12233 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
12235 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
12236 * config/arc/arc.c (arc_handle_secure_attribute): New function.
12237 (arc_attribute_table): Add 'secure_call' attribute.
12238 (arc_print_operand): Print secure call operand.
12239 (arc_function_ok_for_sibcall): Don't optimize tail calls when
12241 (arc_is_secure_call_p): New function. * config/arc/arc.md
12242 (call_i): Add support for sjli instruction.
12243 (call_value_i): Likewise.
12244 * config/arc/constraints.md (Csc): New constraint.
12246 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
12247 John Eric Martin <John.Martin@emmicro-us.com>
12249 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
12250 * config/arc/arc.c (_arc_jli_section): New struct.
12251 (arc_jli_section): New type.
12252 (rc_jli_sections): New static variable.
12253 (arc_handle_jli_attribute): New function.
12254 (arc_attribute_table): Add jli_always and jli_fixed attribute.
12255 (arc_file_end): New function.
12256 (TARGET_ASM_FILE_END): Define.
12257 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
12258 (arc_add_jli_section): New function.
12259 (jli_call_scan): Likewise.
12260 (arc_reorg): Call jli_call_scan.
12261 (arc_output_addsi): Remove 'S' from printing asm operand.
12262 (arc_is_jli_call_p): New function.
12263 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
12265 (movhi_insn): Likewise.
12266 (movsi_insn): Likewise.
12267 (movsi_set_cc_insn): Likewise.
12268 (loadqi_update): Likewise.
12269 (load_zeroextendqisi_update): Likewise.
12270 (load_signextendqisi_update): Likewise.
12271 (loadhi_update): Likewise.
12272 (load_zeroextendhisi_update): Likewise.
12273 (load_signextendhisi_update): Likewise.
12274 (loadsi_update): Likewise.
12275 (loadsf_update): Likewise.
12276 (movsicc_insn): Likewise.
12277 (bset_insn): Likewise.
12278 (bxor_insn): Likewise.
12279 (bclr_insn): Likewise.
12280 (bmsk_insn): Likewise.
12281 (bicsi3_insn): Likewise.
12282 (cmpsi_cc_c_insn): Likewise.
12283 (movsi_ne): Likewise.
12284 (movsi_cond_exec): Likewise.
12285 (clrsbsi2): Likewise.
12286 (norm_f): Likewise.
12296 (call_i): Remove 'S' asm letter, add jli instruction.
12297 (call_value_i): Likewise.
12298 * config/arc/arc.op (mjli-always): New option.
12299 * config/arc/constraints.md (Cji): New constraint.
12300 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
12302 (subsf3_fpx): Likewise.
12303 (mulsf3_fpx): Likewise.
12304 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
12306 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
12307 function attrbutes.
12308 * doc/invoke.texi (ARC): Document mjli-always option.
12310 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
12312 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
12313 avoid addition with 0 and use incw and decw where possible.
12315 2018-01-26 Richard Biener <rguenther@suse.de>
12317 PR tree-optimization/81082
12318 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
12319 association if it requires casting to unsigned.
12320 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
12321 from fold_plusminus_mult_expr to catch important cases late when
12322 range info is available.
12324 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12326 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
12327 * configure.ac (hidden_linkonce): New test.
12328 * configure: Regenerate.
12329 * config.in: Regenerate.
12331 2018-01-26 Julia Koval <julia.koval@intel.com>
12333 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
12334 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
12335 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
12336 _mm_mask_bitshuffle_epi64_mask): Fix type.
12337 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
12338 USI_FTYPE_V4DI_V4DI_USI): Remove.
12339 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
12340 __builtin_ia32_vpshufbitqmb256_mask,
12341 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
12342 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
12343 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
12345 2018-01-26 Alan Modra <amodra@gmail.com>
12348 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
12349 UNSPEC_VBPERMQ. Sort other unspecs.
12351 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
12353 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
12355 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
12357 PR middle-end/83055
12358 * predict.c (drop_profile): Do not push/pop cfun; update also
12360 (handle_missing_profiles): Fix logic looking for zero profiles.
12362 2018-01-25 Jakub Jelinek <jakub@redhat.com>
12364 PR middle-end/83977
12365 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
12366 on functions with #pragma omp declare simd or functions with simd
12368 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
12369 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
12370 Remove trailing \n from warning_at calls.
12372 2018-01-25 Tom de Vries <tom@codesourcery.com>
12375 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
12376 for neutered workers.
12378 2018-01-24 Joseph Myers <joseph@codesourcery.com>
12381 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
12382 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
12384 2018-01-24 Jeff Law <law@redhat.com>
12387 * i386.c (get_probe_interval): Move to earlier point.
12388 (ix86_compute_frame_layout): If -fstack-clash-protection and
12389 the frame is larger than the probe interval, then use pushes
12390 to save registers rather than reg->mem moves.
12391 (ix86_expand_prologue): Remove conditional for int_registers_saved
12394 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
12397 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
12398 min/max for never referenced object.
12400 2018-01-24 Jakub Jelinek <jakub@redhat.com>
12402 PR middle-end/83977
12403 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
12405 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
12406 attributes from DECL_ATTRIBUTES (decl) without affecting
12407 DECL_ATTRIBUTES (current_function_decl).
12408 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
12409 functions with non-NULL DECL_ABSTRACT_ORIGIN.
12411 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
12413 PR tree-optimization/83979
12414 * fold-const.c (fold_comparison): Use constant_boolean_node
12415 instead of boolean_{true,false}_node.
12417 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
12419 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
12422 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12424 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
12425 Simplify the clause that sets the length attribute.
12426 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
12427 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
12428 clause that sets the length attribute.
12429 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
12431 2018-01-24 Tom de Vries <tom@codesourcery.com>
12434 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
12435 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
12436 Add strict parameter.
12437 (prevent_branch_around_nothing): Insert dummy insn between branch to
12438 label and label with no ptx insn inbetween.
12439 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
12441 2018-01-24 Tom de Vries <tom@codesourcery.com>
12444 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
12445 for neutered threads in warp.
12446 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
12448 2018-01-24 Richard Biener <rguenther@suse.de>
12450 PR tree-optimization/83176
12451 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
12454 2018-01-24 Richard Biener <rguenther@suse.de>
12456 PR tree-optimization/82819
12457 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
12458 code generating pluses that are no-ops in the target precision.
12460 2018-01-24 Richard Biener <rguenther@suse.de>
12462 PR middle-end/84000
12463 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
12465 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
12467 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
12468 to merge probabilities.
12469 * predict.c (probably_never_executed): Also mark as cold functions
12470 with global 0 profile and guessed local profile.
12471 * profile-count.c (profile_probability::combine_with_count): New
12473 * profile-count.h (profile_probability::operator*,
12474 profile_probability::operator*=, profile_probability::operator/,
12475 profile_probability::operator/=): Reduce precision to adjusted
12476 and set value to guessed on contradictory divisions.
12477 (profile_probability::combine_with_freq): Remove.
12478 (profile_probability::combine_wiht_count): Declare.
12479 (profile_count::force_nonzero):: Set to adjusted.
12480 (profile_count::probability_in):: Set quality to adjusted.
12481 * tree-ssa-tail-merge.c (replace_block_by): Use
12482 combine_with_count.
12484 2018-01-23 Andrew Waterman <andrew@sifive.com>
12485 Jim Wilson <jimw@sifive.com>
12487 * config/riscv/riscv.c (riscv_stack_boundary): New.
12488 (riscv_option_override): Set riscv_stack_boundary. Handle
12489 riscv_preferred_stack_boundary_arg.
12490 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
12491 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
12492 (STACK_BOUNDARY): Set to riscv_stack_boundary.
12493 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
12494 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
12495 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
12497 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
12500 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
12501 of struct ix86_frame.
12502 (ix86_expand_epilogue): Likewise. Add a local variable for
12503 the reg_save_offset field in struct ix86_frame.
12505 2018-01-23 Bin Cheng <bin.cheng@arm.com>
12507 PR tree-optimization/82604
12508 * tree-loop-distribution.c (enum partition_kind): New enum item
12509 PKIND_PARTIAL_MEMSET.
12510 (partition_builtin_p): Support above new enum item.
12511 (generate_code_for_partition): Ditto.
12512 (compute_access_range): Differentiate cases that equality can be
12513 proven at all loops, the innermost loops or no loops.
12514 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
12515 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
12516 (finalize_partitions, distribute_loop): Don't fuse partition of
12517 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
12518 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
12519 parloop is enabled.
12521 2018-01-23 Martin Liska <mliska@suse.cz>
12523 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
12524 order to ignore the predictor.
12525 (PRED_POLYMORPHIC_CALL): Likewise.
12526 (PRED_RECURSIVE_CALL): Likewise.
12528 2018-01-23 Martin Liska <mliska@suse.cz>
12530 * tree-profile.c (tree_profiling): Print function header to
12531 aware reader which function we are working on.
12532 * value-prof.c (gimple_find_values_to_profile): Do not print
12533 not interesting value histograms.
12535 2018-01-23 Martin Liska <mliska@suse.cz>
12537 * profile-count.h (enum profile_quality): Add
12538 profile_uninitialized as the first value. Do not number values
12539 as they are zero based.
12540 (profile_count::verify): Update sanity check.
12541 (profile_probability::verify): Likewise.
12543 2018-01-23 Nathan Sidwell <nathan@acm.org>
12545 * doc/invoke.texi (ffor-scope): Deprecate.
12547 2018-01-23 David Malcolm <dmalcolm@redhat.com>
12549 PR tree-optimization/83510
12550 * domwalk.c (set_all_edges_as_executable): New function.
12551 (dom_walker::dom_walker): Convert bool param
12552 "skip_unreachable_blocks" to enum reachability. Move setup of
12553 edge flags to set_all_edges_as_executable and only do it when
12554 reachability is REACHABLE_BLOCKS.
12555 * domwalk.h (enum dom_walker::reachability): New enum.
12556 (dom_walker::dom_walker): Convert bool param
12557 "skip_unreachable_blocks" to enum reachability.
12558 (set_all_edges_as_executable): New decl.
12559 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
12560 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
12562 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
12563 but converting true to REACHABLE_BLOCKS.
12564 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
12566 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
12567 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
12568 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
12570 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
12571 if check_all_array_refs will be called.
12573 2018-01-23 David Malcolm <dmalcolm@redhat.com>
12575 * tree.c (selftest::test_location_wrappers): Add more test
12578 2018-01-23 David Malcolm <dmalcolm@redhat.com>
12580 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
12581 (selftest::test_bit_in_range): Likewise.
12583 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
12586 * doc/sourcebuild.texi (vect_float): Say that the selector
12587 only describes the situation when -funsafe-math-optimizations is on.
12588 (vect_float_strict): Document.
12590 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
12592 PR tree-optimization/83965
12593 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
12594 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
12595 instead of checking only for a reduction.
12596 (vect_recog_widen_sum_pattern): Likewise.
12598 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
12600 * predict.c (probably_never_executed): Only use precise profile info.
12601 (compute_function_frequency): Skip after inlining hack since we now
12602 have quality checking.
12604 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
12606 * profile-count.h (profile_probability::very_unlikely,
12607 profile_probability::unlikely, profile_probability::even): Set
12608 precision to guessed.
12610 2018-01-23 Richard Biener <rguenther@suse.de>
12612 PR tree-optimization/83963
12613 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
12614 Properly terminate dominator walk when crossing the exit edge not
12615 when visiting its source block.
12617 2018-01-23 Jakub Jelinek <jakub@redhat.com>
12620 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
12621 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
12623 2018-01-22 Jakub Jelinek <jakub@redhat.com>
12625 PR tree-optimization/83957
12626 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
12627 semicolon after for body surrounded by braces.
12629 PR tree-optimization/83081
12630 * profile-count.h (profile_probability::split): New method.
12631 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
12632 Use profile_probability::split.
12633 (do_compare_rtx_and_jump): Fix adjustment of probabilities
12634 when splitting a single conditional jump into 2.
12636 2018-01-22 David Malcolm <dmalcolm@redhat.com>
12638 PR tree-optimization/69452
12639 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
12642 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12644 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
12645 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
12646 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
12648 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12650 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
12652 * config/rl78/rl78.md (movdi): New define_expand.
12653 * config/rl78/rl78.c (rl78_split_movdi): New function.
12655 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
12658 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
12660 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
12661 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
12662 128-bit to produce an UNSPEC move to get the double word with the
12663 signbit and then a shift directly to do signbit.
12664 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
12665 implementation with a new version that just does either a direct
12666 move or a regular move. Move memory interface to separate insns.
12667 Move insns so they are next to the expander.
12668 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
12669 with signbit move. Split big and little endian case.
12670 (signbit<mode>2_dm_mem_le): Likewise.
12671 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
12672 (signbit<mode>2_dm2): Likewise.
12674 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12676 * config/rl78/rl78.md (anddi3): New define_expand.
12678 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12680 * config/rl78/rl78.md (umindi3): New define_expand.
12682 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12684 * config/rl78/rl78.md (smindi3): New define_expand.
12686 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12688 * config/rl78/rl78.md (smaxdi3): New define_expand.
12690 2018-01-22 Carl Love <cel@us.ibm.com>
12692 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
12693 LVX_V1TI): Add macro expansion.
12694 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
12695 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
12696 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
12697 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
12698 Change check to determine if the instruction is a byte reversing
12699 entry. Fix typo in comment.
12700 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
12701 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
12702 Add def_builtin calls for new builtins.
12703 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
12704 Add define_insn expansion.
12706 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12708 * config/rl78/rl78.md (umaxdi3): New define_expand.
12710 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12712 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
12713 for non-QImode registers.
12715 2018-01-22 Richard Biener <rguenther@suse.de>
12717 PR tree-optimization/83963
12718 * graphite-scop-detection.c (scop_detection::get_sese): Delay
12719 including the loop exit block.
12720 (scop_detection::merge_sese): Likewise.
12721 (scop_detection::add_scop): Do it here instead.
12723 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12725 * doc/sourcebuild.texi (arm_softfloat): Document.
12727 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
12730 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
12731 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
12732 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
12734 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12735 David Edelsohn <dje.gcc@gmail.com>
12738 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
12739 Change "crset eq" to "crset 2".
12740 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
12741 (*call_indirect_aix<mode>_nospec): Likewise.
12742 (*call_value_indirect_aix<mode>_nospec): Likewise.
12743 (*call_indirect_elfv2<mode>_nospec): Likewise.
12744 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
12745 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
12746 change assembly output from . to $.
12747 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
12748 (indirect_jump<mode>_nospec): Change assembly output from . to $.
12749 (*tablejump<mode>_internal1_nospec): Likewise.
12751 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
12754 * config/sh/sh_optimize_sett_clrt.cc:
12755 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
12757 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
12759 PR tree-optimization/83940
12760 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
12761 offset_dt to vect_constant_def rather than vect_unknown_def_type.
12762 (vect_check_load_store_mask): Add a mask_dt_out parameter and
12763 use it to pass back the definition type.
12764 (vect_check_store_rhs): Likewise rhs_dt_out.
12765 (vect_build_gather_load_calls): Add a mask_dt argument and use
12766 it instead of a call to vect_is_simple_use.
12767 (vectorizable_store): Update calls to vect_check_load_store_mask
12768 and vect_check_store_rhs. Use the dt returned by the latter instead
12769 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
12770 instead of calls to vect_is_simple_use. Pass the scalar rather
12771 than the vector operand to vect_is_simple_use when handling
12772 second and subsequent copies of an rhs value.
12773 (vectorizable_load): Update calls to vect_check_load_store_mask
12774 and vect_build_gather_load_calls. Use the cached mask_dt and
12775 gs_info.offset_dt instead of calls to vect_is_simple_use.
12777 2018-01-20 Jakub Jelinek <jakub@redhat.com>
12779 PR middle-end/83945
12780 * tree-emutls.c: Include gimplify.h.
12781 (lower_emutls_2): New function.
12782 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
12783 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
12784 it before further processing.
12787 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
12788 UINTVAL (trueop1) instead of INTVAL (op1).
12790 2018-01-19 Jakub Jelinek <jakub@redhat.com>
12794 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
12795 INCOMING_FRAME_SP_OFFSET if not defined.
12796 (scan_trace): Add ENTRY argument. If true and
12797 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
12798 emit a note to adjust the CFA offset.
12799 (create_cfi_notes): Adjust scan_trace callers.
12800 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
12801 INCOMING_FRAME_SP_OFFSET in the CIE.
12802 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
12803 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
12805 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
12806 * doc/tm.texi: Regenerated.
12808 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
12810 PR rtl-optimization/83147
12811 * lra-constraints.c (remove_inheritance_pseudos): Use
12812 lra_substitute_pseudo_within_insn.
12814 2018-01-19 Tom de Vries <tom@codesourcery.com>
12815 Cesar Philippidis <cesar@codesourcery.com>
12818 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
12820 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
12823 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
12824 spaces for function labels.
12826 2018-01-19 Martin Liska <mliska@suse.cz>
12828 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
12829 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
12830 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
12831 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
12832 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
12833 (PRED_CONST_RETURN): Change from 69 to 65.
12834 (PRED_NULL_RETURN): Change from 91 to 71.
12835 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
12836 (PRED_LOOP_GUARD): Change from 66 to 73.
12838 2018-01-19 Martin Liska <mliska@suse.cz>
12840 * predict.c (predict_insn_def): Add new assert.
12841 (struct branch_predictor): Change type to signed integer.
12842 (test_prediction_value_range): Amend test to cover
12843 PROB_UNINITIALIZED.
12844 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
12845 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
12846 (PRED_LOOP_ITERATIONS_MAX): Likewise.
12847 (PRED_LOOP_IV_COMPARE): Likewise.
12848 * predict.h (PROB_UNINITIALIZED): Define new constant.
12850 2018-01-19 Martin Liska <mliska@suse.cz>
12852 * predict.c (dump_prediction): Add new format for
12853 analyze_brprob.py script which is enabled with -details
12855 * profile-count.h (precise_p): New function.
12857 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
12859 PR tree-optimization/83922
12860 * tree-vect-loop.c (vect_verify_full_masking): Return false if
12861 there are no statements that need masking.
12862 (vect_active_double_reduction_p): New function.
12863 (vect_analyze_loop_operations): Use it when handling phis that
12864 are not in the loop header.
12866 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
12868 PR tree-optimization/83914
12869 * tree-vect-loop.c (vectorizable_induction): Don't convert
12870 init_expr or apply the peeling adjustment for inductions
12871 that are nested within the vectorized loop.
12873 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12875 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
12878 2018-01-18 Jakub Jelinek <jakub@redhat.com>
12882 * function.h (gimplify_parameters): Add gimple_seq * argument.
12883 * function.c: Include gimple.h and options.h.
12884 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
12885 for the added local temporaries if needed.
12886 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
12887 if there are any parameter cleanups, wrap whole body into a
12888 try/finally with the cleanups.
12890 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
12893 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
12894 Use GET_MODE_CLASS for scalar floating point.
12896 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
12900 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
12901 Fix call of call_cgraph_insertion_hooks.
12903 2018-01-18 Martin Sebor <msebor@redhat.com>
12905 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
12907 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
12910 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
12913 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
12916 * common.opt: (-ffile-prefix-map): New option.
12917 * opts.c (common_handle_option): Defer it.
12918 * opts-global.c (handle_common_deferred_options): Handle it.
12919 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
12920 * file-prefix-map.h: New file.
12921 (remap_debug_filename, add_debug_prefix_map): ...here.
12922 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
12923 * final.c (debug_prefix_map, add_debug_prefix_map
12924 remap_debug_filename): Move to...
12925 * file-prefix-map.c: New file.
12926 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
12927 generalize, get rid of alloca(), use strrchr() instead of strchr().
12928 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
12929 Implement in terms of add_prefix_map().
12930 (remap_macro_filename, remap_debug_filename): Implement in term of
12932 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
12933 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
12934 * dbxout.c: Include file-prefix-map.h.
12935 * varasm.c: Likewise.
12936 * vmsdbgout.c: Likewise.
12937 * xcoffout.c: Likewise.
12938 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
12939 * doc/cppopts.texi (-fmacro-prefix-map): Document.
12940 * doc/invoke.texi (-ffile-prefix-map): Document.
12941 (-fdebug-prefix-map): Update description.
12943 2018-01-18 Martin Liska <mliska@suse.cz>
12945 * config/i386/i386.c (indirect_thunk_name): Document that also
12947 (output_indirect_thunk): Document why both instructions
12948 (pause and lfence) are generated.
12950 2018-01-18 Richard Biener <rguenther@suse.de>
12952 PR tree-optimization/83887
12953 * graphite-scop-detection.c
12954 (scop_detection::get_nearest_dom_with_single_entry): Remove.
12955 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
12956 (scop_detection::merge_sese): Re-implement with a flood-fill
12957 algorithm that properly finds a SESE region if it exists.
12959 2018-01-18 Jakub Jelinek <jakub@redhat.com>
12962 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
12963 pointer_diff optimizations use view_convert instead of convert.
12965 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12967 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
12968 Generate different code for -mno-speculate-indirect-jumps.
12969 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
12970 (*call_indirect_aix<mode>): Disable for
12971 -mno-speculate-indirect-jumps.
12972 (*call_indirect_aix<mode>_nospec): New define_insn.
12973 (*call_value_indirect_aix<mode>): Disable for
12974 -mno-speculate-indirect-jumps.
12975 (*call_value_indirect_aix<mode>_nospec): New define_insn.
12976 (*sibcall_nonlocal_sysv<mode>): Generate different code for
12977 -mno-speculate-indirect-jumps.
12978 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
12980 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
12982 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
12983 long double type, set the flags for noting the default long double
12984 type, even if we don't pass or return a long double type.
12986 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
12989 * ipa-inline.c (flatten_function): Do not overwrite final inlining
12992 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
12994 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
12995 support for merge[hl].
12996 (fold_mergehl_helper): New helper function.
12997 (tree-vector-builder.h): New #include for tree_vector_builder usage.
12998 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
12999 (altivec_vmrglw_direct): Add xxmrglw insn.
13001 2018-01-17 Andrew Waterman <andrew@sifive.com>
13003 * config/riscv/riscv.c (riscv_conditional_register_usage): If
13004 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
13006 2018-01-17 David Malcolm <dmalcolm@redhat.com>
13009 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
13010 call the lto_location_cache before reading the
13011 DECL_SOURCE_LOCATION of the types.
13013 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
13014 Richard Sandiford <richard.sandiford@linaro.org>
13016 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
13017 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
13018 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
13019 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
13020 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
13022 * config/aarch64/constraints.md (aarch64_movti_operand):
13024 * config/aarch64/predicates.md (Uti): Add new constraint.
13026 2018-01-17 Carl Love <cel@us.ibm.com>
13028 * config/rs6000/vsx.md (define_expand xl_len_r,
13029 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
13030 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
13032 (define_expand, define_insn): Move the shift left from the
13033 define_insn to the define_expand for lxvl and stxvl instructions.
13034 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
13035 and XL_LEN_R definitions to PURE.
13037 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
13039 * config/i386/i386.c (indirect_thunk_name): Declare regno
13040 as unsigned int. Compare regno with INVALID_REGNUM.
13041 (output_indirect_thunk): Ditto.
13042 (output_indirect_thunk_function): Ditto.
13043 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
13044 in the call to output_indirect_thunk_function.
13046 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
13048 PR middle-end/83884
13049 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
13050 rather than the size of inner_type to determine the stack slot size
13051 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
13053 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
13056 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
13059 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
13061 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
13062 endian Linux systems to optionally enable multilibs for selecting
13063 the long double type if the user configured an explicit type.
13064 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
13065 have no long double multilibs if not defined.
13066 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
13067 warn if the user used -mabi={ieee,ibm}longdouble and we built
13068 multilibs for long double.
13069 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
13070 appropriate multilib option.
13071 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
13073 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
13074 for building long double multilibs.
13075 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
13077 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
13079 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
13082 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
13084 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
13087 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
13090 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
13093 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
13095 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
13096 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
13098 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
13100 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
13101 different rtl trees depending on TARGET_64BIT.
13102 (rs6000_gen_lvx): Likewise.
13104 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
13106 * config/visium/visium.md (nop): Tweak comment.
13107 (hazard_nop): Likewise.
13109 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13111 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
13112 -mspeculate-indirect-jumps.
13113 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
13114 for -mno-speculate-indirect-jumps.
13115 (*call_indirect_elfv2<mode>_nospec): New define_insn.
13116 (*call_value_indirect_elfv2<mode>): Disable for
13117 -mno-speculate-indirect-jumps.
13118 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
13119 (indirect_jump): Emit different RTL for
13120 -mno-speculate-indirect-jumps.
13121 (*indirect_jump<mode>): Disable for
13122 -mno-speculate-indirect-jumps.
13123 (*indirect_jump<mode>_nospec): New define_insn.
13124 (tablejump): Emit different RTL for
13125 -mno-speculate-indirect-jumps.
13126 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
13127 (tablejumpsi_nospec): New define_expand.
13128 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
13129 (tablejumpdi_nospec): New define_expand.
13130 (*tablejump<mode>_internal1): Disable for
13131 -mno-speculate-indirect-jumps.
13132 (*tablejump<mode>_internal1_nospec): New define_insn.
13133 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
13136 2018-01-16 Artyom Skrobov tyomitch@gmail.com
13138 * caller-save.c (insert_save): Drop unnecessary parameter. All
13141 2018-01-16 Jakub Jelinek <jakub@redhat.com>
13142 Richard Biener <rguenth@suse.de>
13145 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
13146 return early, inline manually is_gimple_sizepos. Make sure if we
13147 call gimplify_expr we don't end up with a gimple constant.
13148 * tree.c (variably_modified_type_p): Don't return true for
13149 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
13150 * gimplify.h (is_gimple_sizepos): Remove.
13152 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
13154 PR tree-optimization/83857
13155 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
13156 vectorizable_live_operation for pure SLP statements.
13157 (vectorizable_live_operation): Handle PHIs.
13159 2018-01-16 Richard Biener <rguenther@suse.de>
13161 PR tree-optimization/83867
13162 * tree-vect-stmts.c (vect_transform_stmt): Precompute
13163 nested_in_vect_loop_p since the scalar stmt may get invalidated.
13165 2018-01-16 Jakub Jelinek <jakub@redhat.com>
13168 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
13169 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
13170 If off is not INTEGER_CST, issue a may not be aligned warning
13171 rather than isn't aligned. Use isn%'t rather than isn't.
13172 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
13174 <case MULT_EXPR>: Improve the case when bottom and one of the
13175 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
13176 operand, in that case check if the other operand is multiple of
13177 bottom divided by the INTEGER_CST operand.
13179 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
13182 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
13183 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
13184 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
13185 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
13186 * config/pa/pa.c (pa_function_arg_advance): Likewise.
13187 (pa_function_arg, pa_arg_partial_bytes): Likewise.
13188 (pa_function_arg_size): New function.
13190 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
13192 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
13193 in a separate statement.
13195 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
13197 PR tree-optimization/83847
13198 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
13199 group gathers and scatters.
13201 2018-01-16 Jakub Jelinek <jakub@redhat.com>
13203 PR rtl-optimization/86620
13204 * params.def (max-sched-ready-insns): Bump minimum value to 1.
13206 PR rtl-optimization/83213
13207 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
13208 to last if both are JUMP_INSNs.
13210 PR tree-optimization/83843
13211 * gimple-ssa-store-merging.c
13212 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
13213 store_immediate_info for bswap/nop orig_stores.
13215 2018-01-15 Andrew Waterman <andrew@sifive.com>
13217 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
13219 <UDIV>: Increase cost if !TARGET_DIV.
13221 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
13223 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
13224 (define_attr "cr_logical_3op"): New.
13225 (cceq_ior_compare): Adjust.
13226 (cceq_ior_compare_complement): Adjust.
13227 (*cceq_rev_compare): Adjust.
13228 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
13229 (is_cracked_insn): Adjust.
13230 (insn_must_be_first_in_group): Adjust.
13231 * config/rs6000/40x.md: Adjust.
13232 * config/rs6000/440.md: Adjust.
13233 * config/rs6000/476.md: Adjust.
13234 * config/rs6000/601.md: Adjust.
13235 * config/rs6000/603.md: Adjust.
13236 * config/rs6000/6xx.md: Adjust.
13237 * config/rs6000/7450.md: Adjust.
13238 * config/rs6000/7xx.md: Adjust.
13239 * config/rs6000/8540.md: Adjust.
13240 * config/rs6000/cell.md: Adjust.
13241 * config/rs6000/e300c2c3.md: Adjust.
13242 * config/rs6000/e500mc.md: Adjust.
13243 * config/rs6000/e500mc64.md: Adjust.
13244 * config/rs6000/e5500.md: Adjust.
13245 * config/rs6000/e6500.md: Adjust.
13246 * config/rs6000/mpc.md: Adjust.
13247 * config/rs6000/power4.md: Adjust.
13248 * config/rs6000/power5.md: Adjust.
13249 * config/rs6000/power6.md: Adjust.
13250 * config/rs6000/power7.md: Adjust.
13251 * config/rs6000/power8.md: Adjust.
13252 * config/rs6000/power9.md: Adjust.
13253 * config/rs6000/rs64.md: Adjust.
13254 * config/rs6000/titan.md: Adjust.
13256 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
13258 * config/i386/predicates.md (indirect_branch_operand): Rewrite
13259 ix86_indirect_branch_register logic.
13261 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
13263 * config/i386/constraints.md (Bs): Update
13264 ix86_indirect_branch_register check. Don't check
13265 ix86_indirect_branch_register with GOT_memory_operand.
13267 * config/i386/predicates.md (GOT_memory_operand): Don't check
13268 ix86_indirect_branch_register here.
13269 (GOT32_symbol_operand): Likewise.
13271 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
13273 * config/i386/predicates.md (constant_call_address_operand):
13274 Rewrite ix86_indirect_branch_register logic.
13275 (sibcall_insn_operand): Likewise.
13277 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
13279 * config/i386/constraints.md (Bs): Replace
13280 ix86_indirect_branch_thunk_register with
13281 ix86_indirect_branch_register.
13283 * config/i386/i386.md (indirect_jump): Likewise.
13284 (tablejump): Likewise.
13285 (*sibcall_memory): Likewise.
13286 (*sibcall_value_memory): Likewise.
13287 Peepholes of indirect call and jump via memory: Likewise.
13288 * config/i386/i386.opt: Likewise.
13289 * config/i386/predicates.md (indirect_branch_operand): Likewise.
13290 (GOT_memory_operand): Likewise.
13291 (call_insn_operand): Likewise.
13292 (sibcall_insn_operand): Likewise.
13293 (GOT32_symbol_operand): Likewise.
13295 2018-01-15 Jakub Jelinek <jakub@redhat.com>
13297 PR middle-end/83837
13298 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
13299 type rather than type addr's type points to.
13300 (expand_omp_atomic_mutex): Likewise.
13301 (expand_omp_atomic): Likewise.
13303 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
13306 * config/i386/i386.c (output_indirect_thunk_function): Use
13307 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
13308 for __x86_return_thunk.
13310 2018-01-15 Richard Biener <rguenther@suse.de>
13312 PR middle-end/83850
13313 * expmed.c (extract_bit_field_1): Fix typo.
13315 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13318 * config/arm/iterators.md (VF): New mode iterator.
13319 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
13320 Remove integer-related logic from pattern.
13321 (neon_vabd<mode>_3): Likewise.
13323 2018-01-15 Jakub Jelinek <jakub@redhat.com>
13325 PR middle-end/82694
13326 * common.opt (fstrict-overflow): No longer an alias.
13327 (fwrapv-pointer): New option.
13328 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
13329 also for pointer types based on flag_wrapv_pointer.
13330 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
13331 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
13332 opts->x_flag_wrapv got set.
13333 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
13334 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
13335 POINTER_TYPE_OVERFLOW_UNDEFINED.
13336 * match.pd: Likewise in address comparison pattern.
13337 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
13339 2018-01-15 Richard Biener <rguenther@suse.de>
13342 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
13343 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
13344 Reset type names to their identifier if their TYPE_DECL doesn't
13345 have linkage (and thus is used for ODR and devirt).
13346 (save_debug_info_for_decl): Remove.
13347 (save_debug_info_for_type): Likewise.
13348 (add_tree_to_fld_list): Adjust.
13349 * tree-pretty-print.c (dump_generic_node): Make dumping of
13350 type names more robust.
13352 2018-01-15 Richard Biener <rguenther@suse.de>
13354 * BASE-VER: Bump to 8.0.1.
13356 2018-01-14 Martin Sebor <msebor@redhat.com>
13359 * builtins.c (check_access): Avoid warning when the no-warning bit
13362 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
13364 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
13365 * ira-color (allocno_hard_regs_compare): Likewise.
13367 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
13370 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
13371 Use .pushsection/.popsection.
13373 2018-01-14 Martin Sebor <msebor@redhat.com>
13376 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
13378 2018-01-14 Jakub Jelinek <jakub@redhat.com>
13380 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
13381 entry from extra_headers.
13382 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
13383 extra_headers, make the list bitwise identical to the i?86-*-* one.
13385 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
13387 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
13388 -mcmodel=large with -mindirect-branch=thunk,
13389 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
13390 -mfunction-return=thunk-extern.
13391 * doc/invoke.texi: Document -mcmodel=large is incompatible with
13392 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
13393 -mfunction-return=thunk and -mfunction-return=thunk-extern.
13395 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
13397 * config/i386/i386.c (print_reg): Print the name of the full
13398 integer register without '%'.
13399 (ix86_print_operand): Handle 'V'.
13400 * doc/extend.texi: Document 'V' modifier.
13402 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
13404 * config/i386/constraints.md (Bs): Disallow memory operand for
13405 -mindirect-branch-register.
13407 * config/i386/predicates.md (indirect_branch_operand): Likewise.
13408 (GOT_memory_operand): Likewise.
13409 (call_insn_operand): Likewise.
13410 (sibcall_insn_operand): Likewise.
13411 (GOT32_symbol_operand): Likewise.
13412 * config/i386/i386.md (indirect_jump): Call convert_memory_address
13413 for -mindirect-branch-register.
13414 (tablejump): Likewise.
13415 (*sibcall_memory): Likewise.
13416 (*sibcall_value_memory): Likewise.
13417 Disallow peepholes of indirect call and jump via memory for
13418 -mindirect-branch-register.
13419 (*call_pop): Replace m with Bw.
13420 (*call_value_pop): Likewise.
13421 (*sibcall_pop_memory): Replace m with Bs.
13422 * config/i386/i386.opt (mindirect-branch-register): New option.
13423 * doc/invoke.texi: Document -mindirect-branch-register option.
13425 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
13427 * config/i386/i386-protos.h (ix86_output_function_return): New.
13428 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
13429 set function_return_type.
13430 (indirect_thunk_name): Add ret_p to indicate thunk for function
13432 (output_indirect_thunk_function): Pass false to
13433 indirect_thunk_name.
13434 (ix86_output_indirect_branch_via_reg): Likewise.
13435 (ix86_output_indirect_branch_via_push): Likewise.
13436 (output_indirect_thunk_function): Create alias for function
13437 return thunk if regno < 0.
13438 (ix86_output_function_return): New function.
13439 (ix86_handle_fndecl_attribute): Handle function_return.
13440 (ix86_attribute_table): Add function_return.
13441 * config/i386/i386.h (machine_function): Add
13442 function_return_type.
13443 * config/i386/i386.md (simple_return_internal): Use
13444 ix86_output_function_return.
13445 (simple_return_internal_long): Likewise.
13446 * config/i386/i386.opt (mfunction-return=): New option.
13447 (indirect_branch): Mention -mfunction-return=.
13448 * doc/extend.texi: Document function_return function attribute.
13449 * doc/invoke.texi: Document -mfunction-return= option.
13451 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
13453 * config/i386/i386-opts.h (indirect_branch): New.
13454 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
13455 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
13456 with local indirect jump when converting indirect call and jump.
13457 (ix86_set_indirect_branch_type): New.
13458 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
13459 (indirectlabelno): New.
13460 (indirect_thunk_needed): Likewise.
13461 (indirect_thunk_bnd_needed): Likewise.
13462 (indirect_thunks_used): Likewise.
13463 (indirect_thunks_bnd_used): Likewise.
13464 (INDIRECT_LABEL): Likewise.
13465 (indirect_thunk_name): Likewise.
13466 (output_indirect_thunk): Likewise.
13467 (output_indirect_thunk_function): Likewise.
13468 (ix86_output_indirect_branch_via_reg): Likewise.
13469 (ix86_output_indirect_branch_via_push): Likewise.
13470 (ix86_output_indirect_branch): Likewise.
13471 (ix86_output_indirect_jmp): Likewise.
13472 (ix86_code_end): Call output_indirect_thunk_function if needed.
13473 (ix86_output_call_insn): Call ix86_output_indirect_branch if
13475 (ix86_handle_fndecl_attribute): Handle indirect_branch.
13476 (ix86_attribute_table): Add indirect_branch.
13477 * config/i386/i386.h (machine_function): Add indirect_branch_type
13478 and has_local_indirect_jump.
13479 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
13481 (tablejump): Likewise.
13482 (*indirect_jump): Use ix86_output_indirect_jmp.
13483 (*tablejump_1): Likewise.
13484 (simple_return_indirect_internal): Likewise.
13485 * config/i386/i386.opt (mindirect-branch=): New option.
13486 (indirect_branch): New.
13489 (thunk-inline): Likewise.
13490 (thunk-extern): Likewise.
13491 * doc/extend.texi: Document indirect_branch function attribute.
13492 * doc/invoke.texi: Document -mindirect-branch= option.
13494 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
13497 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
13499 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
13501 * ipa-inline.c (want_inline_small_function_p): Return false if
13502 inlining has already failed with CIF_FINAL_ERROR.
13503 (update_caller_keys): Call want_inline_small_function_p before
13505 (update_callee_keys): Likewise.
13507 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
13509 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
13511 (rs6000_quadword_masked_address_p): Likewise.
13512 (quad_aligned_load_p): Likewise.
13513 (quad_aligned_store_p): Likewise.
13514 (const_load_sequence_p): Add comment to describe the outer-most loop.
13515 (mimic_memory_attributes_and_flags): New function.
13516 (rs6000_gen_stvx): Likewise.
13517 (replace_swapped_aligned_store): Likewise.
13518 (rs6000_gen_lvx): Likewise.
13519 (replace_swapped_aligned_load): Likewise.
13520 (replace_swapped_load_constant): Capitalize argument name in
13521 comment describing this function.
13522 (rs6000_analyze_swaps): Add a third pass to search for vector loads
13523 and stores that access quad-word aligned addresses and replace
13524 with stvx or lvx instructions when appropriate.
13525 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
13526 New function prototype.
13527 (rs6000_quadword_masked_address_p): Likewise.
13528 (rs6000_gen_lvx): Likewise.
13529 (rs6000_gen_stvx): Likewise.
13530 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
13531 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
13532 when memory address is aligned.
13533 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
13534 this split to select lvx instruction when memory address is aligned.
13535 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
13536 instruction when memory address is aligned.
13537 (*vsx_le_perm_load_v16qi): Likewise.
13538 (four unnamed splitters): Modify to select the stvx instruction
13539 when memory is aligned.
13541 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
13543 * predict.c (determine_unlikely_bbs): Handle correctly BBs
13544 which appears in the queue multiple times.
13546 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13547 Alan Hayward <alan.hayward@arm.com>
13548 David Sherwood <david.sherwood@arm.com>
13550 * tree-vectorizer.h (vec_lower_bound): New structure.
13551 (_loop_vec_info): Add check_nonzero and lower_bounds.
13552 (LOOP_VINFO_CHECK_NONZERO): New macro.
13553 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
13554 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
13555 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
13556 fields. Make seg_len the distance travelled, not including the
13558 (dr_direction_indicator): Declare.
13559 (dr_zero_step_indicator): Likewise.
13560 (dr_known_forward_stride_p): Likewise.
13561 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
13563 (runtime_alias_check_p): Allow runtime alias checks with
13565 (operator ==): Compare access_size and align.
13566 (prune_runtime_alias_test_list): Rework for new distinction between
13567 the access_size and seg_len.
13568 (create_intersect_range_checks_index): Likewise. Cope with polynomial
13570 (get_segment_min_max): New function.
13571 (create_intersect_range_checks): Use it.
13572 (dr_step_indicator): New function.
13573 (dr_direction_indicator): Likewise.
13574 (dr_zero_step_indicator): Likewise.
13575 (dr_known_forward_stride_p): Likewise.
13576 * tree-loop-distribution.c (data_ref_segment_size): Return
13577 DR_STEP * (niters - 1).
13578 (compute_alias_check_pairs): Update call to the dr_with_seg_len
13580 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
13581 (vect_preserves_scalar_order_p): New function, split out from...
13582 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
13583 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
13584 (vect_vfa_access_size): New function.
13585 (vect_vfa_align): Likewise.
13586 (vect_compile_time_alias): Take access_size_a and access_b arguments.
13587 (dump_lower_bound): New function.
13588 (vect_check_lower_bound): Likewise.
13589 (vect_small_gap_p): Likewise.
13590 (vectorizable_with_step_bound_p): Likewise.
13591 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
13592 depencies if the vectorization factor is 1. Convert the checks
13593 for nonzero steps into checks on the bounds of DR_STEP. Try using
13594 a bunds check for variable steps if the minimum required step is
13595 relatively small. Update calls to the dr_with_seg_len
13596 constructor and to vect_compile_time_alias.
13597 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
13599 (vect_loop_versioning): Call it.
13600 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
13602 (vect_estimate_min_profitable_iters): Account for any bounds checks.
13604 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13605 Alan Hayward <alan.hayward@arm.com>
13606 David Sherwood <david.sherwood@arm.com>
13608 * doc/sourcebuild.texi (vect_scatter_store): Document.
13609 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
13611 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
13613 * genopinit.c (main): Add supports_vec_scatter_store and
13614 supports_vec_scatter_store_cached to target_optabs.
13615 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
13616 IFN_MASK_SCATTER_STORE.
13617 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
13619 * internal-fn.h (internal_store_fn_p): Declare.
13620 (internal_fn_stored_value_index): Likewise.
13621 * internal-fn.c (scatter_store_direct): New macro.
13622 (expand_scatter_store_optab_fn): New function.
13623 (direct_scatter_store_optab_supported_p): New macro.
13624 (internal_store_fn_p): New function.
13625 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
13626 IFN_MASK_SCATTER_STORE.
13627 (internal_fn_mask_index): Likewise.
13628 (internal_fn_stored_value_index): New function.
13629 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
13630 for scatter stores.
13631 * optabs-query.h (supports_vec_scatter_store_p): Declare.
13632 * optabs-query.c (supports_vec_scatter_store_p): New function.
13633 * tree-vectorizer.h (vect_get_store_rhs): Declare.
13634 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
13635 true for scatter stores.
13636 (vect_gather_scatter_fn_p): Handle scatter stores too.
13637 (vect_check_gather_scatter): Consider using scatter stores if
13638 supports_vec_scatter_store_p.
13639 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
13640 scatter stores too.
13641 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
13642 internal_fn_stored_value_index.
13643 (check_load_store_masking): Handle scatter stores too.
13644 (vect_get_store_rhs): Make public.
13645 (vectorizable_call): Use internal_store_fn_p.
13646 (vectorizable_store): Handle scatter store internal functions.
13647 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
13648 when deciding whether the end of the group has been reached.
13649 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
13650 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
13651 (mask_scatter_store<mode>): New insns.
13653 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13654 Alan Hayward <alan.hayward@arm.com>
13655 David Sherwood <david.sherwood@arm.com>
13657 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
13658 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
13659 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
13661 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
13662 Use vect_truncate_gather_scatter_offset if we can't treat the
13663 operation as a normal gather load or scatter store.
13664 (get_group_load_store_type): Take the gather_scatter_info
13665 as argument. Try using a gather load or scatter store for
13666 single-element groups.
13667 (get_load_store_type): Update calls to get_group_load_store_type
13668 and vect_use_strided_gather_scatters_p.
13670 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13671 Alan Hayward <alan.hayward@arm.com>
13672 David Sherwood <david.sherwood@arm.com>
13674 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
13675 optional tree argument.
13676 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
13678 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
13679 but continue to use the current value as a fallback.
13680 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
13681 to compare the updates.
13682 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
13683 (get_load_store_type): Use it when handling a strided access.
13684 (vect_get_strided_load_store_ops): New function.
13685 (vect_get_data_ptr_increment): Likewise.
13686 (vectorizable_load): Handle strided gather loads. Always pass
13687 a step to vect_create_data_ref_ptr and bump_vector_ptr.
13689 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13690 Alan Hayward <alan.hayward@arm.com>
13691 David Sherwood <david.sherwood@arm.com>
13693 * doc/md.texi (gather_load@var{m}): Document.
13694 (mask_gather_load@var{m}): Likewise.
13695 * genopinit.c (main): Add supports_vec_gather_load and
13696 supports_vec_gather_load_cached to target_optabs.
13697 * optabs-tree.c (init_tree_optimization_optabs): Use
13698 ggc_cleared_alloc to allocate target_optabs.
13699 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
13700 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
13702 * internal-fn.h (internal_load_fn_p): Declare.
13703 (internal_gather_scatter_fn_p): Likewise.
13704 (internal_fn_mask_index): Likewise.
13705 (internal_gather_scatter_fn_supported_p): Likewise.
13706 * internal-fn.c (gather_load_direct): New macro.
13707 (expand_gather_load_optab_fn): New function.
13708 (direct_gather_load_optab_supported_p): New macro.
13709 (direct_internal_fn_optab): New function.
13710 (internal_load_fn_p): Likewise.
13711 (internal_gather_scatter_fn_p): Likewise.
13712 (internal_fn_mask_index): Likewise.
13713 (internal_gather_scatter_fn_supported_p): Likewise.
13714 * optabs-query.c (supports_at_least_one_mode_p): New function.
13715 (supports_vec_gather_load_p): Likewise.
13716 * optabs-query.h (supports_vec_gather_load_p): Declare.
13717 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
13718 and memory_type field.
13719 (NUM_PATTERNS): Bump to 15.
13720 * tree-vect-data-refs.c: Include internal-fn.h.
13721 (vect_gather_scatter_fn_p): New function.
13722 (vect_describe_gather_scatter_call): Likewise.
13723 (vect_check_gather_scatter): Try using internal functions for
13724 gather loads. Recognize existing calls to a gather load function.
13725 (vect_analyze_data_refs): Consider using gather loads if
13726 supports_vec_gather_load_p.
13727 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
13728 (vect_get_gather_scatter_offset_type): Likewise.
13729 (vect_convert_mask_for_vectype): Likewise.
13730 (vect_add_conversion_to_patterm): Likewise.
13731 (vect_try_gather_scatter_pattern): Likewise.
13732 (vect_recog_gather_scatter_pattern): New pattern recognizer.
13733 (vect_vect_recog_func_ptrs): Add it.
13734 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
13735 internal_fn_mask_index and internal_gather_scatter_fn_p.
13736 (check_load_store_masking): Take the gather_scatter_info as an
13737 argument and handle gather loads.
13738 (vect_get_gather_scatter_ops): New function.
13739 (vectorizable_call): Check internal_load_fn_p.
13740 (vectorizable_load): Likewise. Handle gather load internal
13742 (vectorizable_store): Update call to check_load_store_masking.
13743 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
13744 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
13745 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
13746 (aarch64_gather_scale_operand_d): New predicates.
13747 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
13748 (mask_gather_load<mode>): New insns.
13750 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13751 Alan Hayward <alan.hayward@arm.com>
13752 David Sherwood <david.sherwood@arm.com>
13754 * optabs.def (fold_left_plus_optab): New optab.
13755 * doc/md.texi (fold_left_plus_@var{m}): Document.
13756 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
13757 * internal-fn.c (fold_left_direct): Define.
13758 (expand_fold_left_optab_fn): Likewise.
13759 (direct_fold_left_optab_supported_p): Likewise.
13760 * fold-const-call.c (fold_const_fold_left): New function.
13761 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
13762 * tree-parloops.c (valid_reduction_p): New function.
13763 (gather_scalar_reductions): Use it.
13764 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
13765 (vect_finish_replace_stmt): Declare.
13766 * tree-vect-loop.c (fold_left_reduction_fn): New function.
13767 (needs_fold_left_reduction_p): New function, split out from...
13768 (vect_is_simple_reduction): ...here. Accept reductions that
13769 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
13770 (vect_force_simple_reduction): Also store the reduction type in
13771 the assignment's STMT_VINFO_REDUC_TYPE.
13772 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
13773 (merge_with_identity): New function.
13774 (vect_expand_fold_left): Likewise.
13775 (vectorize_fold_left_reduction): Likewise.
13776 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
13777 scalar phi in place for it. Check for target support and reject
13778 cases that would reassociate the operation. Defer the transform
13779 phase to vectorize_fold_left_reduction.
13780 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
13781 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
13782 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
13784 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13786 * tree-if-conv.c (predicate_mem_writes): Remove redundant
13787 call to ifc_temp_var.
13789 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13790 Alan Hayward <alan.hayward@arm.com>
13791 David Sherwood <david.sherwood@arm.com>
13793 * target.def (legitimize_address_displacement): Take the original
13794 offset as a poly_int.
13795 * targhooks.h (default_legitimize_address_displacement): Update
13797 * targhooks.c (default_legitimize_address_displacement): Likewise.
13798 * doc/tm.texi: Regenerate.
13799 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
13800 as an argument, moving assert of ad->disp == ad->disp_term to...
13801 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
13802 Try calling targetm.legitimize_address_displacement before expanding
13803 the address rather than afterwards, and adjust for the new interface.
13804 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
13805 Match the new hook interface. Handle SVE addresses.
13806 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
13807 new hook interface.
13809 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13811 * Makefile.in (OBJS): Add early-remat.o.
13812 * target.def (select_early_remat_modes): New hook.
13813 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
13814 * doc/tm.texi: Regenerate.
13815 * targhooks.h (default_select_early_remat_modes): Declare.
13816 * targhooks.c (default_select_early_remat_modes): New function.
13817 * timevar.def (TV_EARLY_REMAT): New timevar.
13818 * passes.def (pass_early_remat): New pass.
13819 * tree-pass.h (make_pass_early_remat): Declare.
13820 * early-remat.c: New file.
13821 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
13823 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
13825 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13826 Alan Hayward <alan.hayward@arm.com>
13827 David Sherwood <david.sherwood@arm.com>
13829 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
13830 vfm1 with a bound_epilog parameter.
13831 (vect_do_peeling): Update calls accordingly, and move the prologue
13832 call earlier in the function. Treat the base bound_epilog as 0 for
13833 fully-masked loops and retain vf - 1 for other loops. Add 1 to
13834 this base when peeling for gaps.
13835 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
13836 with fully-masked loops.
13837 (vect_estimate_min_profitable_iters): Handle the single peeled
13838 iteration in that case.
13840 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13841 Alan Hayward <alan.hayward@arm.com>
13842 David Sherwood <david.sherwood@arm.com>
13844 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
13845 single-element interleaving even if the size is not a power of 2.
13846 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
13847 accesses for single-element interleaving if the group size is
13850 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13851 Alan Hayward <alan.hayward@arm.com>
13852 David Sherwood <david.sherwood@arm.com>
13854 * doc/md.texi (fold_extract_last_@var{m}): Document.
13855 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
13856 * optabs.def (fold_extract_last_optab): New optab.
13857 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
13858 * internal-fn.c (fold_extract_direct): New macro.
13859 (expand_fold_extract_optab_fn): Likewise.
13860 (direct_fold_extract_optab_supported_p): Likewise.
13861 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
13862 * tree-vect-loop.c (vect_model_reduction_cost): Handle
13863 EXTRACT_LAST_REDUCTION.
13864 (get_initial_def_for_reduction): Do not create an initial vector
13865 for EXTRACT_LAST_REDUCTION reductions.
13866 (vectorizable_reduction): Leave the scalar phi in place for
13867 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
13868 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
13869 epilogue code for EXTRACT_LAST_REDUCTION and defer the
13870 transform phase to vectorizable_condition.
13871 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
13873 (vect_finish_stmt_generation): ...here.
13874 (vect_finish_replace_stmt): New function.
13875 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
13876 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
13878 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
13880 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13881 Alan Hayward <alan.hayward@arm.com>
13882 David Sherwood <david.sherwood@arm.com>
13884 * doc/md.texi (extract_last_@var{m}): Document.
13885 * optabs.def (extract_last_optab): New optab.
13886 * internal-fn.def (EXTRACT_LAST): New internal function.
13887 * internal-fn.c (cond_unary_direct): New macro.
13888 (expand_cond_unary_optab_fn): Likewise.
13889 (direct_cond_unary_optab_supported_p): Likewise.
13890 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
13891 loops using EXTRACT_LAST.
13892 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
13893 (extract_last_<mode>): ...this optab.
13894 (vec_extract<mode><Vel>): Update accordingly.
13896 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13897 Alan Hayward <alan.hayward@arm.com>
13898 David Sherwood <david.sherwood@arm.com>
13900 * target.def (empty_mask_is_expensive): New hook.
13901 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
13902 * doc/tm.texi: Regenerate.
13903 * targhooks.h (default_empty_mask_is_expensive): Declare.
13904 * targhooks.c (default_empty_mask_is_expensive): New function.
13905 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
13906 if the target says that empty masks are expensive.
13907 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
13909 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
13911 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13912 Alan Hayward <alan.hayward@arm.com>
13913 David Sherwood <david.sherwood@arm.com>
13915 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
13916 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
13917 (vect_use_loop_mask_for_alignment_p): New function.
13918 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
13919 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
13920 niters_skip argument. Make sure that the first niters_skip elements
13921 of the first iteration are inactive.
13922 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
13923 Update call to vect_set_loop_masks_directly.
13924 (get_misalign_in_elems): New function, split out from...
13925 (vect_gen_prolog_loop_niters): ...here.
13926 (vect_update_init_of_dr): Take a code argument that specifies whether
13927 the adjustment should be added or subtracted.
13928 (vect_update_init_of_drs): Likewise.
13929 (vect_prepare_for_masked_peels): New function.
13930 (vect_do_peeling): Skip prologue peeling if we're using a mask
13931 instead. Update call to vect_update_inits_of_drs.
13932 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
13934 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
13935 alignment. Do not include the number of peeled iterations in
13936 the minimum threshold in that case.
13937 (vectorizable_induction): Adjust the start value down by
13938 LOOP_VINFO_MASK_SKIP_NITERS iterations.
13939 (vect_transform_loop): Call vect_prepare_for_masked_peels.
13940 Take the number of skipped iterations into account when calculating
13942 * tree-vect-stmts.c (vect_gen_while_not): New function.
13944 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13945 Alan Hayward <alan.hayward@arm.com>
13946 David Sherwood <david.sherwood@arm.com>
13948 * doc/sourcebuild.texi (vect_fully_masked): Document.
13949 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
13950 default value to 0.
13951 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
13953 (vect_analyze_loop_2): ...here. Don't check the vectorization
13954 factor against the number of loop iterations if the loop is
13957 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13958 Alan Hayward <alan.hayward@arm.com>
13959 David Sherwood <david.sherwood@arm.com>
13961 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
13962 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
13963 (dump_groups): Update accordingly.
13964 (iv_use::mem_type): New member variable.
13965 (address_p): New function.
13966 (record_use): Add a mem_type argument and initialize the new
13968 (record_group_use): Add a mem_type argument. Use address_p.
13969 Remove obsolete null checks of base_object. Update call to record_use.
13970 (find_interesting_uses_op): Update call to record_group_use.
13971 (find_interesting_uses_cond): Likewise.
13972 (find_interesting_uses_address): Likewise.
13973 (get_mem_type_for_internal_fn): New function.
13974 (find_address_like_use): Likewise.
13975 (find_interesting_uses_stmt): Try find_address_like_use before
13976 calling find_interesting_uses_op.
13977 (addr_offset_valid_p): Use the iv mem_type field as the type
13978 of the addressed memory.
13979 (add_autoinc_candidates): Likewise.
13980 (get_address_cost): Likewise.
13981 (split_small_address_groups_p): Use address_p.
13982 (split_address_groups): Likewise.
13983 (add_iv_candidate_for_use): Likewise.
13984 (autoinc_possible_for_pair): Likewise.
13985 (rewrite_groups): Likewise.
13986 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
13987 (determine_group_iv_cost): Update after split of USE_ADDRESS.
13988 (get_alias_ptr_type_for_ptr_address): New function.
13989 (rewrite_use_address): Rewrite address uses in calls that were
13990 identified by find_address_like_use.
13992 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13993 Alan Hayward <alan.hayward@arm.com>
13994 David Sherwood <david.sherwood@arm.com>
13996 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
13998 * gimple-expr.h (is_gimple_addressable: Likewise.
13999 * gimple-expr.c (is_gimple_address): Likewise.
14000 * internal-fn.c (expand_call_mem_ref): New function.
14001 (expand_mask_load_optab_fn): Use it.
14002 (expand_mask_store_optab_fn): Likewise.
14004 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14005 Alan Hayward <alan.hayward@arm.com>
14006 David Sherwood <david.sherwood@arm.com>
14008 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
14009 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
14010 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
14011 (cond_umax@var{mode}): Document.
14012 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
14013 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
14014 (cond_umin_optab, cond_umax_optab): New optabs.
14015 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
14016 (COND_IOR, COND_XOR): New internal functions.
14017 * internal-fn.h (get_conditional_internal_fn): Declare.
14018 * internal-fn.c (cond_binary_direct): New macro.
14019 (expand_cond_binary_optab_fn): Likewise.
14020 (direct_cond_binary_optab_supported_p): Likewise.
14021 (get_conditional_internal_fn): New function.
14022 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
14023 Cope with reduction statements that are vectorized as calls rather
14025 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
14026 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
14027 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
14028 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
14029 (UNSPEC_COND_EOR): New unspecs.
14030 (optab): Add mappings for them.
14031 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
14032 (sve_int_op, sve_fp_op): New int attributes.
14034 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14035 Alan Hayward <alan.hayward@arm.com>
14036 David Sherwood <david.sherwood@arm.com>
14038 * optabs.def (while_ult_optab): New optab.
14039 * doc/md.texi (while_ult@var{m}@var{n}): Document.
14040 * internal-fn.def (WHILE_ULT): New internal function.
14041 * internal-fn.h (direct_internal_fn_supported_p): New override
14042 that takes two types as argument.
14043 * internal-fn.c (while_direct): New macro.
14044 (expand_while_optab_fn): New function.
14045 (convert_optab_supported_p): Likewise.
14046 (direct_while_optab_supported_p): New macro.
14047 * wide-int.h (wi::udiv_ceil): New function.
14048 * tree-vectorizer.h (rgroup_masks): New structure.
14049 (vec_loop_masks): New typedef.
14050 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
14051 and fully_masked_p.
14052 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
14053 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
14054 (vect_max_vf): New function.
14055 (slpeel_make_loop_iterate_ntimes): Delete.
14056 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
14057 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
14058 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
14059 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
14060 internal-fn.h, stor-layout.h and optabs-query.h.
14061 (vect_set_loop_mask): New function.
14062 (add_preheader_seq): Likewise.
14063 (add_header_seq): Likewise.
14064 (interleave_supported_p): Likewise.
14065 (vect_maybe_permute_loop_masks): Likewise.
14066 (vect_set_loop_masks_directly): Likewise.
14067 (vect_set_loop_condition_masked): Likewise.
14068 (vect_set_loop_condition_unmasked): New function, split out from
14069 slpeel_make_loop_iterate_ntimes.
14070 (slpeel_make_loop_iterate_ntimes): Rename to..
14071 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
14072 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
14073 (vect_do_peeling): Update call accordingly.
14074 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
14076 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
14077 mask_compare_type, can_fully_mask_p and fully_masked_p.
14078 (release_vec_loop_masks): New function.
14079 (_loop_vec_info): Use it to free the loop masks.
14080 (can_produce_all_loop_masks_p): New function.
14081 (vect_get_max_nscalars_per_iter): Likewise.
14082 (vect_verify_full_masking): Likewise.
14083 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
14084 retries, and free the mask rgroups before retrying. Check loop-wide
14085 reasons for disallowing fully-masked loops. Make the final decision
14086 about whether use a fully-masked loop or not.
14087 (vect_estimate_min_profitable_iters): Do not assume that peeling
14088 for the number of iterations will be needed for fully-masked loops.
14089 (vectorizable_reduction): Disable fully-masked loops.
14090 (vectorizable_live_operation): Likewise.
14091 (vect_halve_mask_nunits): New function.
14092 (vect_double_mask_nunits): Likewise.
14093 (vect_record_loop_mask): Likewise.
14094 (vect_get_loop_mask): Likewise.
14095 (vect_transform_loop): Handle the case in which the final loop
14096 iteration might handle a partial vector. Call vect_set_loop_condition
14097 instead of slpeel_make_loop_iterate_ntimes.
14098 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
14099 (check_load_store_masking): New function.
14100 (prepare_load_store_mask): Likewise.
14101 (vectorizable_store): Handle fully-masked loops.
14102 (vectorizable_load): Likewise.
14103 (supportable_widening_operation): Use vect_halve_mask_nunits for
14105 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
14106 (vect_gen_while): New function.
14107 * config/aarch64/aarch64.md (umax<mode>3): New expander.
14108 (aarch64_uqdec<mode>): New insn.
14110 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14111 Alan Hayward <alan.hayward@arm.com>
14112 David Sherwood <david.sherwood@arm.com>
14114 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
14115 (reduc_xor_scal_optab): New optabs.
14116 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
14117 (reduc_xor_scal_@var{m}): Document.
14118 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
14119 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
14120 internal functions.
14121 * fold-const-call.c (fold_const_call): Handle them.
14122 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
14123 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
14124 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
14125 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
14126 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
14127 (UNSPEC_XORV): New unspecs.
14128 (optab): Add entries for them.
14129 (BITWISEV): New int iterator.
14130 (bit_reduc_op): New int attributes.
14132 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14133 Alan Hayward <alan.hayward@arm.com>
14134 David Sherwood <david.sherwood@arm.com>
14136 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
14137 * internal-fn.def (VEC_SHL_INSERT): New internal function.
14138 * optabs.def (vec_shl_insert_optab): New optab.
14139 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
14140 (duplicate_and_interleave): Likewise.
14141 * tree-vect-loop.c: Include internal-fn.h.
14142 (neutral_op_for_slp_reduction): New function, split out from
14143 get_initial_defs_for_reduction.
14144 (get_initial_def_for_reduction): Handle option 2 for variable-length
14145 vectors by loading the neutral value into a vector and then shifting
14146 the initial value into element 0.
14147 (get_initial_defs_for_reduction): Replace the code argument with
14148 the neutral value calculated by neutral_op_for_slp_reduction.
14149 Use gimple_build_vector for constant-length vectors.
14150 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
14151 but the first group_size elements have a neutral value.
14152 Use duplicate_and_interleave otherwise.
14153 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
14154 Update call to get_initial_defs_for_reduction. Handle SLP
14155 reductions for variable-length vectors by creating one vector
14156 result for each scalar result, with the elements associated
14157 with other scalar results stubbed out with the neutral value.
14158 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
14159 Require IFN_VEC_SHL_INSERT for double reductions on
14160 variable-length vectors, or SLP reductions that have
14161 a neutral value. Require can_duplicate_and_interleave_p
14162 support for variable-length unchained SLP reductions if there
14163 is no neutral value, such as for MIN/MAX reductions. Also require
14164 the number of vector elements to be a multiple of the number of
14165 SLP statements when doing variable-length unchained SLP reductions.
14166 Update call to vect_create_epilog_for_reduction.
14167 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
14168 and remove initial values.
14169 (duplicate_and_interleave): Make public.
14170 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
14171 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
14173 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14174 Alan Hayward <alan.hayward@arm.com>
14175 David Sherwood <david.sherwood@arm.com>
14177 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
14178 (can_duplicate_and_interleave_p): New function.
14179 (vect_get_and_check_slp_defs): Take the vector of statements
14180 rather than just the current one. Remove excess parentheses.
14181 Restriction rejectinon of vect_constant_def and vect_external_def
14182 for variable-length vectors to boolean types, or types for which
14183 can_duplicate_and_interleave_p is false.
14184 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
14185 (duplicate_and_interleave): New function.
14186 (vect_get_constant_vectors): Use gimple_build_vector for
14187 constant-length vectors and suitable variable-length constant
14188 vectors. Use duplicate_and_interleave for other variable-length
14189 vectors. Don't defer the update when inserting new statements.
14191 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14192 Alan Hayward <alan.hayward@arm.com>
14193 David Sherwood <david.sherwood@arm.com>
14195 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
14196 min_profitable_iters doesn't go negative.
14198 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14199 Alan Hayward <alan.hayward@arm.com>
14200 David Sherwood <david.sherwood@arm.com>
14202 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
14203 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
14204 * optabs.def (vec_mask_load_lanes_optab): New optab.
14205 (vec_mask_store_lanes_optab): Likewise.
14206 * internal-fn.def (MASK_LOAD_LANES): New internal function.
14207 (MASK_STORE_LANES): Likewise.
14208 * internal-fn.c (mask_load_lanes_direct): New macro.
14209 (mask_store_lanes_direct): Likewise.
14210 (expand_mask_load_optab_fn): Handle masked operations.
14211 (expand_mask_load_lanes_optab_fn): New macro.
14212 (expand_mask_store_optab_fn): Handle masked operations.
14213 (expand_mask_store_lanes_optab_fn): New macro.
14214 (direct_mask_load_lanes_optab_supported_p): Likewise.
14215 (direct_mask_store_lanes_optab_supported_p): Likewise.
14216 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
14218 (vect_load_lanes_supported): Likewise.
14219 * tree-vect-data-refs.c (strip_conversion): New function.
14220 (can_group_stmts_p): Likewise.
14221 (vect_analyze_data_ref_accesses): Use it instead of checking
14222 for a pair of assignments.
14223 (vect_store_lanes_supported): Take a masked_p parameter.
14224 (vect_load_lanes_supported): Likewise.
14225 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
14226 vect_store_lanes_supported and vect_load_lanes_supported.
14227 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
14228 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
14229 parameter. Don't allow gaps for masked accesses.
14230 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
14231 and vect_load_lanes_supported.
14232 (get_load_store_type): Take a masked_p parameter and update
14233 call to get_group_load_store_type.
14234 (vectorizable_store): Update call to get_load_store_type.
14235 Handle IFN_MASK_STORE_LANES.
14236 (vectorizable_load): Update call to get_load_store_type.
14237 Handle IFN_MASK_LOAD_LANES.
14239 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14240 Alan Hayward <alan.hayward@arm.com>
14241 David Sherwood <david.sherwood@arm.com>
14243 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
14245 * config/aarch64/aarch64-protos.h
14246 (aarch64_sve_struct_memory_operand_p): Declare.
14247 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
14248 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
14249 (VPRED, vpred): Handle SVE structure modes.
14250 * config/aarch64/constraints.md (Utx): New constraint.
14251 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
14252 (aarch64_sve_struct_nonimmediate_operand): New predicates.
14253 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
14254 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
14255 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
14256 structure modes. Split into pieces after RA.
14257 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
14258 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
14260 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
14261 SVE structure modes.
14262 (aarch64_classify_address): Likewise.
14263 (sizetochar): Move earlier in file.
14264 (aarch64_print_operand): Handle SVE register lists.
14265 (aarch64_array_mode): New function.
14266 (aarch64_sve_struct_memory_operand_p): Likewise.
14267 (TARGET_ARRAY_MODE): Redefine.
14269 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14270 Alan Hayward <alan.hayward@arm.com>
14271 David Sherwood <david.sherwood@arm.com>
14273 * target.def (array_mode): New target hook.
14274 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
14275 * doc/tm.texi: Regenerate.
14276 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
14277 * hooks.c (hook_optmode_mode_uhwi_none): New function.
14278 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
14279 targetm.array_mode.
14280 * stor-layout.c (mode_for_array): Likewise. Support polynomial
14283 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14284 Alan Hayward <alan.hayward@arm.com>
14285 David Sherwood <david.sherwood@arm.com>
14287 * fold-const.c (fold_binary_loc): Check the argument types
14288 rather than the result type when testing for a vector operation.
14290 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14292 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
14293 * doc/tm.texi: Regenerate.
14295 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14296 Alan Hayward <alan.hayward@arm.com>
14297 David Sherwood <david.sherwood@arm.com>
14299 * doc/invoke.texi (-msve-vector-bits=): Document new option.
14300 (sve): Document new AArch64 extension.
14301 * doc/md.texi (w): Extend the description of the AArch64
14302 constraint to include SVE vectors.
14303 (Upl, Upa): Document new AArch64 predicate constraints.
14304 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
14306 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
14307 (msve-vector-bits=): New option.
14308 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
14309 SVE when these are disabled.
14310 (sve): New extension.
14311 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
14312 modes. Adjust their number of units based on aarch64_sve_vg.
14313 (MAX_BITSIZE_MODE_ANY_MODE): Define.
14314 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
14315 aarch64_addr_query_type.
14316 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
14317 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
14318 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
14319 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
14320 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
14321 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
14322 (aarch64_simd_imm_zero_p): Delete.
14323 (aarch64_check_zero_based_sve_index_immediate): Declare.
14324 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
14325 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
14326 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
14327 (aarch64_sve_float_mul_immediate_p): Likewise.
14328 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
14329 rather than an rtx.
14330 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
14331 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
14332 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
14333 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
14334 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
14335 (aarch64_regmode_natural_size): Likewise.
14336 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
14337 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
14339 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
14340 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
14341 for VG and the SVE predicate registers.
14342 (V_ALIASES): Add a "z"-prefixed alias.
14343 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
14344 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
14345 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
14346 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
14347 (REG_CLASS_NAMES): Add entries for them.
14348 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
14349 and the predicate registers.
14350 (aarch64_sve_vg): Declare.
14351 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
14352 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
14353 (REGMODE_NATURAL_SIZE): Define.
14354 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
14356 * config/aarch64/aarch64.c: Include cfgrtl.h.
14357 (simd_immediate_info): Add a constructor for series vectors,
14358 and an associated step field.
14359 (aarch64_sve_vg): New variable.
14360 (aarch64_dbx_register_number): Handle VG and the predicate registers.
14361 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
14362 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
14363 (VEC_ANY_DATA, VEC_STRUCT): New constants.
14364 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
14365 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
14366 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
14367 (aarch64_get_mask_mode): New functions.
14368 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
14369 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
14370 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
14371 predicate modes and predicate registers. Explicitly restrict
14372 GPRs to modes of 16 bytes or smaller. Only allow FP registers
14373 to store a vector mode if it is recognized by
14374 aarch64_classify_vector_mode.
14375 (aarch64_regmode_natural_size): New function.
14376 (aarch64_hard_regno_caller_save_mode): Return the original mode
14378 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
14379 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
14380 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
14381 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
14383 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
14384 does not overlap dest if the function is frame-related. Handle
14386 (aarch64_split_add_offset): New function.
14387 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
14388 them aarch64_add_offset.
14389 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
14390 and update call to aarch64_sub_sp.
14391 (aarch64_add_cfa_expression): New function.
14392 (aarch64_expand_prologue): Pass extra temporary registers to the
14393 functions above. Handle the case in which we need to emit new
14394 DW_CFA_expressions for registers that were originally saved
14395 relative to the stack pointer, but now have to be expressed
14396 relative to the frame pointer.
14397 (aarch64_output_mi_thunk): Pass extra temporary registers to the
14399 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
14400 IP0 and IP1 values for SVE frames.
14401 (aarch64_expand_vec_series): New function.
14402 (aarch64_expand_sve_widened_duplicate): Likewise.
14403 (aarch64_expand_sve_const_vector): Likewise.
14404 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
14405 Handle SVE constants. Use emit_move_insn to move a force_const_mem
14406 into the register, rather than emitting a SET directly.
14407 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
14408 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
14409 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
14410 (offset_9bit_signed_scaled_p): New functions.
14411 (aarch64_replicate_bitmask_imm): New function.
14412 (aarch64_bitmask_imm): Use it.
14413 (aarch64_cannot_force_const_mem): Reject expressions involving
14414 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
14415 (aarch64_classify_index): Handle SVE indices, by requiring
14416 a plain register index with a scale that matches the element size.
14417 (aarch64_classify_address): Handle SVE addresses. Assert that
14418 the mode of the address is VOIDmode or an integer mode.
14419 Update call to aarch64_classify_symbol.
14420 (aarch64_classify_symbolic_expression): Update call to
14421 aarch64_classify_symbol.
14422 (aarch64_const_vec_all_in_range_p): New function.
14423 (aarch64_print_vector_float_operand): Likewise.
14424 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
14425 "vN" for FP registers with SVE modes. Handle (const ...) vectors
14426 and the FP immediates 1.0 and 0.5.
14427 (aarch64_print_address_internal): Handle SVE addresses.
14428 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
14429 (aarch64_regno_regclass): Handle predicate registers.
14430 (aarch64_secondary_reload): Handle big-endian reloads of SVE
14432 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
14433 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
14434 (aarch64_convert_sve_vector_bits): New function.
14435 (aarch64_override_options): Use it to handle -msve-vector-bits=.
14436 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
14437 rather than an rtx.
14438 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
14439 Handle SVE vector and predicate modes. Accept VL-based constants
14440 that need only one temporary register, and VL offsets that require
14441 no temporary registers.
14442 (aarch64_conditional_register_usage): Mark the predicate registers
14443 as fixed if SVE isn't available.
14444 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
14445 Return true for SVE vector and predicate modes.
14446 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
14447 rather than an unsigned int. Handle SVE modes.
14448 (aarch64_preferred_simd_mode): Update call accordingly. Handle
14450 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
14452 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
14453 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
14454 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
14455 (aarch64_sve_float_mul_immediate_p): New functions.
14456 (aarch64_sve_valid_immediate): New function.
14457 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
14458 Explicitly reject structure modes. Check for INDEX constants.
14459 Handle PTRUE and PFALSE constants.
14460 (aarch64_check_zero_based_sve_index_immediate): New function.
14461 (aarch64_simd_imm_zero_p): Delete.
14462 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
14463 vector modes. Accept constants in the range of CNT[BHWD].
14464 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
14465 ask for an Advanced SIMD mode.
14466 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
14467 (aarch64_simd_vector_alignment): Handle SVE predicates.
14468 (aarch64_vectorize_preferred_vector_alignment): New function.
14469 (aarch64_simd_vector_alignment_reachable): Use it instead of
14471 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
14472 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
14474 (MAX_VECT_LEN): Delete.
14475 (expand_vec_perm_d): Add a vec_flags field.
14476 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
14477 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
14478 (aarch64_evpc_ext): Don't apply a big-endian lane correction
14480 (aarch64_evpc_rev): Rename to...
14481 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
14482 (aarch64_evpc_rev_global): New function.
14483 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
14484 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
14486 (aarch64_evpc_sve_tbl): New function.
14487 (aarch64_expand_vec_perm_const_1): Update after rename of
14488 aarch64_evpc_rev. Handle SVE permutes too, trying
14489 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
14490 than aarch64_evpc_tbl.
14491 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
14492 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
14493 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
14494 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
14495 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
14496 (aarch64_expand_sve_vcond): New functions.
14497 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
14498 of aarch64_vector_mode_p.
14499 (aarch64_dwarf_poly_indeterminate_value): New function.
14500 (aarch64_compute_pressure_classes): Likewise.
14501 (aarch64_can_change_mode_class): Likewise.
14502 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
14503 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
14504 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
14505 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
14506 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
14507 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
14508 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
14509 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
14511 (Dn, Dl, Dr): Accept const as well as const_vector.
14512 (Dz): Likewise. Compare against CONST0_RTX.
14513 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
14514 of "vector" where appropriate.
14515 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
14516 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
14517 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
14518 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
14519 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
14520 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
14521 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
14522 (v_int_equiv): Extend to SVE modes.
14523 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
14525 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
14526 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
14527 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
14528 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
14529 (SVE_COND_FP_CMP): New int iterators.
14530 (perm_hilo): Handle the new unpack unspecs.
14531 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
14533 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
14534 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
14535 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
14536 (aarch64_equality_operator, aarch64_constant_vector_operand)
14537 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
14538 (aarch64_sve_nonimmediate_operand): Likewise.
14539 (aarch64_sve_general_operand): Likewise.
14540 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
14541 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
14542 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
14543 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
14544 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
14545 (aarch64_sve_float_arith_immediate): Likewise.
14546 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
14547 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
14548 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
14549 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
14550 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
14551 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
14552 (aarch64_sve_float_arith_operand): Likewise.
14553 (aarch64_sve_float_arith_with_sub_operand): Likewise.
14554 (aarch64_sve_float_mul_operand): Likewise.
14555 (aarch64_sve_vec_perm_operand): Likewise.
14556 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
14557 (aarch64_mov_operand): Accept const_poly_int and const_vector.
14558 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
14559 as well as const_vector.
14560 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
14561 in file. Use CONST0_RTX and CONSTM1_RTX.
14562 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
14563 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
14564 Use aarch64_simd_imm_zero.
14565 * config/aarch64/aarch64-sve.md: New file.
14566 * config/aarch64/aarch64.md: Include it.
14567 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
14568 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
14569 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
14570 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
14571 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
14572 (sve): New attribute.
14573 (enabled): Disable instructions with the sve attribute unless
14575 (movqi, movhi): Pass CONST_POLY_INT operaneds through
14576 aarch64_expand_mov_immediate.
14577 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
14578 CNT[BHSD] immediates.
14579 (movti): Split CONST_POLY_INT moves into two halves.
14580 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
14581 Split additions that need a temporary here if the destination
14582 is the stack pointer.
14583 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
14584 (*add<mode>3_poly_1): New instruction.
14585 (set_clobber_cc): New expander.
14587 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14589 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
14590 parameter and use it instead of GET_MODE_SIZE (innermode). Use
14591 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
14592 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
14593 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
14594 Change innermode from fixed_mode_size to machine_mode.
14595 (simplify_subreg): Update call accordingly. Handle a constant-sized
14596 subreg of a variable-length CONST_VECTOR.
14598 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14599 Alan Hayward <alan.hayward@arm.com>
14600 David Sherwood <david.sherwood@arm.com>
14602 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
14603 (add_offset_to_base): New function, split out from...
14604 (create_mem_ref): ...here. When handling a scale other than 1,
14605 check first whether the address is valid without the offset.
14606 Add it into the base if so, leaving the index and scale as-is.
14608 2018-01-12 Jakub Jelinek <jakub@redhat.com>
14611 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
14612 fold_for_warn before checking if arg2 is INTEGER_CST.
14614 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
14616 * config/rs6000/predicates.md (load_multiple_operation): Delete.
14617 (store_multiple_operation): Delete.
14618 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
14619 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
14620 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
14621 guarded by TARGET_STRING.
14622 (rs6000_output_load_multiple): Delete.
14623 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
14624 OPTION_MASK_STRING / TARGET_STRING handling.
14625 (print_operand) <'N', 'O'>: Add comment that these are unused now.
14626 (const rs6000_opt_masks) <"string">: Change mask to 0.
14627 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
14628 (MASK_STRING): Delete.
14629 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
14631 (load_multiple): Delete.
14638 (store_multiple): Delete.
14645 (movmemsi_8reg): Delete.
14646 (corresponding unnamed define_insn): Delete.
14647 (movmemsi_6reg): Delete.
14648 (corresponding unnamed define_insn): Delete.
14649 (movmemsi_4reg): Delete.
14650 (corresponding unnamed define_insn): Delete.
14651 (movmemsi_2reg): Delete.
14652 (corresponding unnamed define_insn): Delete.
14653 (movmemsi_1reg): Delete.
14654 (corresponding unnamed define_insn): Delete.
14655 * config/rs6000/rs6000.opt (mno-string): New.
14656 (mstring): Replace by deprecation warning stub.
14657 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
14659 2018-01-12 Jakub Jelinek <jakub@redhat.com>
14661 * regrename.c (regrename_do_replace): If replacing the same
14662 reg multiple times, try to reuse last created gen_raw_REG.
14665 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
14666 main to workaround a bug in GDB.
14668 2018-01-12 Tom de Vries <tom@codesourcery.com>
14671 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
14673 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
14675 PR rtl-optimization/80481
14676 * ira-color.c (get_cap_member): New function.
14677 (allocnos_conflict_by_live_ranges_p): Use it.
14678 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
14679 (setup_slot_coalesced_allocno_live_ranges): Ditto.
14681 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
14684 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
14685 (*saddl_se_1): Ditto.
14686 (*ssubsi_1): Ditto.
14687 (*ssubl_se_1): Ditto.
14689 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
14691 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
14692 rather than wi::to_widest for DR_INITs.
14693 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
14694 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
14695 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
14697 (vect_analyze_group_access_1): Note that here.
14699 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
14701 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
14702 polynomial type sizes.
14704 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
14706 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
14707 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
14708 (gimple_add_tmp_var): Likewise.
14710 2018-01-12 Martin Liska <mliska@suse.cz>
14712 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
14713 (gimple_alloc_sizes): Likewise.
14714 (dump_gimple_statistics): Use PRIu64 in printf format.
14715 * gimple.h: Change uint64_t to int.
14717 2018-01-12 Martin Liska <mliska@suse.cz>
14719 * tree-core.h: Use uint64_t instead of int.
14720 * tree.c (tree_node_counts): Likewise.
14721 (tree_node_sizes): Likewise.
14722 (dump_tree_statistics): Use PRIu64 in printf format.
14724 2018-01-12 Martin Liska <mliska@suse.cz>
14726 * Makefile.in: As qsort_chk is implemented in vec.c, add
14727 vec.o to linkage of gencfn-macros.
14728 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
14729 passing the info to record_node_allocation_statistics.
14730 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
14732 * ggc-common.c (struct ggc_usage): Add operator== and use
14733 it in operator< and compare function.
14734 * mem-stats.h (struct mem_usage): Likewise.
14735 * vec.c (struct vec_usage): Remove operator< and compare
14736 function. Can be simply inherited.
14738 2018-01-12 Martin Jambor <mjambor@suse.cz>
14741 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
14742 * tree-ssa-math-opts.c: Include domwalk.h.
14743 (convert_mult_to_fma_1): New function.
14744 (fma_transformation_info): New type.
14745 (fma_deferring_state): Likewise.
14746 (cancel_fma_deferring): New function.
14747 (result_of_phi): Likewise.
14748 (last_fma_candidate_feeds_initial_phi): Likewise.
14749 (convert_mult_to_fma): Added deferring logic, split actual
14750 transformation to convert_mult_to_fma_1.
14751 (math_opts_dom_walker): New type.
14752 (math_opts_dom_walker::after_dom_children): New method, body moved
14753 here from pass_optimize_widening_mul::execute, added deferring logic
14755 (pass_optimize_widening_mul::execute): Moved most of code to
14756 math_opts_dom_walker::after_dom_children.
14757 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
14758 * config/i386/i386.c (ix86_option_override_internal): Added
14759 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
14761 2018-01-12 Richard Biener <rguenther@suse.de>
14764 * dwarf2out.c (gen_variable_die): Do not reset old_die for
14765 inline instance vars.
14767 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
14770 * config/rx/rx.c (rx_is_restricted_memory_address):
14771 Handle SUBREG case.
14773 2018-01-12 Richard Biener <rguenther@suse.de>
14775 PR tree-optimization/80846
14776 * target.def (split_reduction): New target hook.
14777 * targhooks.c (default_split_reduction): New function.
14778 * targhooks.h (default_split_reduction): Declare.
14779 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
14780 target requests first reduce vectors by combining low and high
14782 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
14783 (get_vectype_for_scalar_type_and_size): Export.
14784 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
14785 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
14786 * doc/tm.texi: Regenerate.
14787 * config/i386/i386.c (ix86_split_reduction): Implement
14788 TARGET_VECTORIZE_SPLIT_REDUCTION.
14790 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
14793 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
14794 in PIC mode except for TARGET_VXWORKS_RTP.
14795 * config/sparc/sparc.c: Include cfgrtl.h.
14796 (TARGET_INIT_PIC_REG): Define.
14797 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
14798 (sparc_pic_register_p): New predicate.
14799 (sparc_legitimate_address_p): Use it.
14800 (sparc_legitimize_pic_address): Likewise.
14801 (sparc_delegitimize_address): Likewise.
14802 (sparc_mode_dependent_address_p): Likewise.
14803 (gen_load_pcrel_sym): Remove 4th parameter.
14804 (load_got_register): Adjust call to above. Remove obsolete stuff.
14805 (sparc_expand_prologue): Do not call load_got_register here.
14806 (sparc_flat_expand_prologue): Likewise.
14807 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
14808 (sparc_use_pseudo_pic_reg): New function.
14809 (sparc_init_pic_reg): Likewise.
14810 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
14811 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
14813 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
14815 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
14816 Add item for branch_cost.
14818 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
14820 PR rtl-optimization/83565
14821 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
14822 not extend the result to a larger mode for rotate operations.
14823 (num_sign_bit_copies1): Likewise.
14825 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
14828 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
14830 Use values-Xc.o for -pedantic.
14831 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
14833 2018-01-12 Martin Liska <mliska@suse.cz>
14836 * ipa-devirt.c (final_warning_record::grow_type_warnings):
14838 (possible_polymorphic_call_targets): Use it.
14839 (ipa_devirt): Likewise.
14841 2018-01-12 Martin Liska <mliska@suse.cz>
14843 * profile-count.h (enum profile_quality): Use 0 as invalid
14844 enum value of profile_quality.
14846 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
14848 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
14849 -mext-string options.
14851 2018-01-12 Richard Biener <rguenther@suse.de>
14853 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
14854 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
14855 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
14857 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
14859 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
14861 * configure.ac (--with-long-double-format): Add support for the
14862 configuration option to change the default long double format on
14864 * config.gcc (powerpc*-linux*-*): Likewise.
14865 * configure: Regenerate.
14866 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
14867 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
14868 used without modification.
14870 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14872 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
14873 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
14874 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
14875 MISC_BUILTIN_SPEC_BARRIER.
14876 (rs6000_init_builtins): Likewise.
14877 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
14879 (speculation_barrier): New define_insn.
14880 * doc/extend.texi: Document __builtin_speculation_barrier.
14882 2018-01-11 Jakub Jelinek <jakub@redhat.com>
14885 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
14886 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
14887 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
14889 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
14890 integral modes instead of "ss" and "sd".
14891 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
14892 vectors with 32-bit and 64-bit elements.
14893 (vecdupssescalarmodesuffix): New mode attribute.
14894 (vec_dup<mode>): Use it.
14896 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
14899 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
14900 frame if argument is passed on stack.
14902 2018-01-11 Jakub Jelinek <jakub@redhat.com>
14905 * ree.c (combine_reaching_defs): Optimize also
14906 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
14907 reg2=any_extend(exp); reg1=reg2;, formatting fix.
14909 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
14911 PR middle-end/83189
14912 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
14914 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
14916 PR middle-end/83718
14917 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
14918 after they are computed.
14920 2018-01-11 Bin Cheng <bin.cheng@arm.com>
14922 PR tree-optimization/83695
14923 * gimple-loop-linterchange.cc
14924 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
14925 reset cached scev information after interchange.
14926 (pass_linterchange::execute): Remove call to scev_reset_htab.
14928 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14930 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
14931 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
14932 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
14933 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
14934 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
14935 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
14936 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
14937 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
14938 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
14939 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
14940 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
14941 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
14942 (V_lane_reg): Likewise.
14943 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
14945 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
14946 (vfmal_lane_low<mode>_intrinsic,
14947 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
14948 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
14949 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
14950 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
14951 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
14952 vfmsl_lane_high<mode>_intrinsic): New define_insns.
14954 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14956 * config/arm/arm-cpus.in (fp16fml): New feature.
14957 (ALL_SIMD): Add fp16fml.
14958 (armv8.2-a): Add fp16fml as an option.
14959 (armv8.3-a): Likewise.
14960 (armv8.4-a): Add fp16fml as part of fp16.
14961 * config/arm/arm.h (TARGET_FP16FML): Define.
14962 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
14964 * config/arm/arm-modes.def (V2HF): Define.
14965 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
14966 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
14967 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
14968 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
14969 vfmsl_low, vfmsl_high): New set of builtins.
14970 * config/arm/iterators.md (PLUSMINUS): New code iterator.
14971 (vfml_op): New code attribute.
14972 (VFMLHALVES): New int iterator.
14973 (VFML, VFMLSEL): New mode attributes.
14974 (V_reg): Define mapping for V2HF.
14975 (V_hi, V_lo): New mode attributes.
14976 (VF_constraint): Likewise.
14977 (vfml_half, vfml_half_selector): New int attributes.
14978 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
14980 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
14981 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
14983 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
14984 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
14985 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
14986 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
14988 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
14989 Document new effective target and option set.
14991 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14993 * config/arm/arm-cpus.in (armv8_4): New feature.
14994 (ARMv8_4a): New fgroup.
14995 (armv8.4-a): New arch.
14996 * config/arm/arm-tables.opt: Regenerate.
14997 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
14998 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
14999 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
15000 Add matching rules for -march=armv8.4-a and extensions.
15001 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
15003 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
15006 * config/rx/rx.md (BW): New mode attribute.
15007 (sync_lock_test_and_setsi): Add mode suffix to insn output.
15009 2018-01-11 Richard Biener <rguenther@suse.de>
15011 PR tree-optimization/83435
15012 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
15013 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
15014 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
15016 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
15017 Alan Hayward <alan.hayward@arm.com>
15018 David Sherwood <david.sherwood@arm.com>
15020 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
15022 (aarch64_classify_address): Initialize it. Track polynomial offsets.
15023 (aarch64_print_address_internal): Use it to check for a zero offset.
15025 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
15026 Alan Hayward <alan.hayward@arm.com>
15027 David Sherwood <david.sherwood@arm.com>
15029 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
15030 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
15031 Return a poly_int64 rather than a HOST_WIDE_INT.
15032 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
15033 rather than a HOST_WIDE_INT.
15034 * config/aarch64/aarch64.h (aarch64_frame): Protect with
15035 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
15036 hard_fp_offset, frame_size, initial_adjust, callee_offset and
15037 final_offset from HOST_WIDE_INT to poly_int64.
15038 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
15039 to_constant when getting the number of units in an Advanced SIMD
15041 (aarch64_builtin_vectorized_function): Check for a constant number
15043 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
15045 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
15046 attribute instead of GET_MODE_NUNITS.
15047 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
15048 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
15049 GET_MODE_SIZE for fixed-size registers.
15050 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
15051 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
15052 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
15053 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
15054 (aarch64_print_operand, aarch64_print_address_internal)
15055 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
15056 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
15057 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
15058 Handle polynomial GET_MODE_SIZE.
15059 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
15060 wider than SImode without modification.
15061 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
15062 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
15063 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
15064 passing and returning SVE modes.
15065 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
15066 rather than GEN_INT.
15067 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
15068 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
15069 (aarch64_allocate_and_probe_stack_space): Likewise.
15070 (aarch64_layout_frame): Cope with polynomial offsets.
15071 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
15072 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
15073 polynomial offsets.
15074 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
15075 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
15076 poly_int64 rather than a HOST_WIDE_INT.
15077 (aarch64_get_separate_components, aarch64_process_components)
15078 (aarch64_expand_prologue, aarch64_expand_epilogue)
15079 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
15080 (aarch64_anchor_offset): New function, split out from...
15081 (aarch64_legitimize_address): ...here.
15082 (aarch64_builtin_vectorization_cost): Handle polynomial
15083 TYPE_VECTOR_SUBPARTS.
15084 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
15086 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
15087 number of elements from the PARALLEL rather than the mode.
15088 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
15089 rather than GET_MODE_BITSIZE.
15090 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
15091 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
15092 (aarch64_expand_vec_perm_const_1): Handle polynomial
15093 d->perm.length () and d->perm elements.
15094 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
15095 Apply to_constant to d->perm elements.
15096 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
15097 polynomial CONST_VECTOR_NUNITS.
15098 (aarch64_move_pointer): Take amount as a poly_int64 rather
15100 (aarch64_progress_pointer): Avoid temporary variable.
15101 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
15102 the mode attribute instead of GET_MODE.
15104 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
15105 Alan Hayward <alan.hayward@arm.com>
15106 David Sherwood <david.sherwood@arm.com>
15108 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
15109 x exists before using it.
15110 (aarch64_add_constant_internal): Rename to...
15111 (aarch64_add_offset_1): ...this. Replace regnum with separate
15112 src and dest rtxes. Handle the case in which they're different,
15113 including when the offset is zero. Replace scratchreg with an rtx.
15114 Use 2 additions if there is no spare register into which we can
15115 move a 16-bit constant.
15116 (aarch64_add_constant): Delete.
15117 (aarch64_add_offset): Replace reg with separate src and dest
15118 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
15119 Use aarch64_add_offset_1.
15120 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
15121 an rtx rather than an int. Take the delta as a poly_int64
15122 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
15123 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
15124 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
15125 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
15126 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
15127 and aarch64_add_sp.
15128 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
15129 aarch64_add_constant.
15131 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
15133 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
15134 Use scalar_float_mode.
15136 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
15138 * config/aarch64/aarch64-simd.md
15139 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
15140 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
15141 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
15142 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
15143 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
15144 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
15145 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
15146 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
15147 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
15148 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
15150 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
15153 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
15154 targ_options->x_arm_arch_string is non NULL.
15156 2018-01-11 Tamar Christina <tamar.christina@arm.com>
15158 * config/aarch64/aarch64.h
15159 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
15161 2018-01-11 Sudakshina Das <sudi.das@arm.com>
15164 * expmed.c (emit_store_flag_force): Swap if const op0
15165 and change VOIDmode to mode of op0.
15167 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
15169 PR rtl-optimization/83761
15170 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
15171 than bytes to mode_for_size.
15173 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
15175 PR middle-end/83189
15176 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
15177 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
15180 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
15182 PR middle-end/83575
15183 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
15184 when in layout mode.
15185 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
15186 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
15189 2018-01-10 Michael Collison <michael.collison@arm.com>
15191 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
15192 * config/aarch64/aarch64-option-extension.def: Add
15193 AARCH64_OPT_EXTENSION of 'fp16fml'.
15194 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
15195 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
15196 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
15197 * config/aarch64/constraints.md (Ui7): New constraint.
15198 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
15199 (VFMLA_SEL_W): Ditto.
15202 (VFMLA16_LOW): New int iterator.
15203 (VFMLA16_HIGH): Ditto.
15204 (UNSPEC_FMLAL): New unspec.
15205 (UNSPEC_FMLSL): Ditto.
15206 (UNSPEC_FMLAL2): Ditto.
15207 (UNSPEC_FMLSL2): Ditto.
15208 (f16mac): New code attribute.
15209 * config/aarch64/aarch64-simd-builtins.def
15210 (aarch64_fmlal_lowv2sf): Ditto.
15211 (aarch64_fmlsl_lowv2sf): Ditto.
15212 (aarch64_fmlalq_lowv4sf): Ditto.
15213 (aarch64_fmlslq_lowv4sf): Ditto.
15214 (aarch64_fmlal_highv2sf): Ditto.
15215 (aarch64_fmlsl_highv2sf): Ditto.
15216 (aarch64_fmlalq_highv4sf): Ditto.
15217 (aarch64_fmlslq_highv4sf): Ditto.
15218 (aarch64_fmlal_lane_lowv2sf): Ditto.
15219 (aarch64_fmlsl_lane_lowv2sf): Ditto.
15220 (aarch64_fmlal_laneq_lowv2sf): Ditto.
15221 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
15222 (aarch64_fmlalq_lane_lowv4sf): Ditto.
15223 (aarch64_fmlsl_lane_lowv4sf): Ditto.
15224 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
15225 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
15226 (aarch64_fmlal_lane_highv2sf): Ditto.
15227 (aarch64_fmlsl_lane_highv2sf): Ditto.
15228 (aarch64_fmlal_laneq_highv2sf): Ditto.
15229 (aarch64_fmlsl_laneq_highv2sf): Ditto.
15230 (aarch64_fmlalq_lane_highv4sf): Ditto.
15231 (aarch64_fmlsl_lane_highv4sf): Ditto.
15232 (aarch64_fmlalq_laneq_highv4sf): Ditto.
15233 (aarch64_fmlsl_laneq_highv4sf): Ditto.
15234 * config/aarch64/aarch64-simd.md:
15235 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
15236 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
15237 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
15238 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
15239 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
15240 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
15241 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
15242 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
15243 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
15244 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
15245 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
15246 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
15247 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
15248 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
15249 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
15250 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
15251 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
15252 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
15253 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
15254 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
15255 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
15256 (vfmlsl_low_u32): Ditto.
15257 (vfmlalq_low_u32): Ditto.
15258 (vfmlslq_low_u32): Ditto.
15259 (vfmlal_high_u32): Ditto.
15260 (vfmlsl_high_u32): Ditto.
15261 (vfmlalq_high_u32): Ditto.
15262 (vfmlslq_high_u32): Ditto.
15263 (vfmlal_lane_low_u32): Ditto.
15264 (vfmlsl_lane_low_u32): Ditto.
15265 (vfmlal_laneq_low_u32): Ditto.
15266 (vfmlsl_laneq_low_u32): Ditto.
15267 (vfmlalq_lane_low_u32): Ditto.
15268 (vfmlslq_lane_low_u32): Ditto.
15269 (vfmlalq_laneq_low_u32): Ditto.
15270 (vfmlslq_laneq_low_u32): Ditto.
15271 (vfmlal_lane_high_u32): Ditto.
15272 (vfmlsl_lane_high_u32): Ditto.
15273 (vfmlal_laneq_high_u32): Ditto.
15274 (vfmlsl_laneq_high_u32): Ditto.
15275 (vfmlalq_lane_high_u32): Ditto.
15276 (vfmlslq_lane_high_u32): Ditto.
15277 (vfmlalq_laneq_high_u32): Ditto.
15278 (vfmlslq_laneq_high_u32): Ditto.
15279 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
15280 (AARCH64_FL_FOR_ARCH8_4): New.
15281 (AARCH64_ISA_F16FML): New ISA flag.
15282 (TARGET_F16FML): New feature flag for fp16fml.
15283 (doc/invoke.texi): Document new fp16fml option.
15285 2018-01-10 Michael Collison <michael.collison@arm.com>
15287 * config/aarch64/aarch64-builtins.c:
15288 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
15289 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
15290 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
15291 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
15292 (AARCH64_ISA_SHA3): New ISA flag.
15293 (TARGET_SHA3): New feature flag for sha3.
15294 * config/aarch64/iterators.md (sha512_op): New int attribute.
15295 (CRYPTO_SHA512): New int iterator.
15296 (UNSPEC_SHA512H): New unspec.
15297 (UNSPEC_SHA512H2): Ditto.
15298 (UNSPEC_SHA512SU0): Ditto.
15299 (UNSPEC_SHA512SU1): Ditto.
15300 * config/aarch64/aarch64-simd-builtins.def
15301 (aarch64_crypto_sha512hqv2di): New builtin.
15302 (aarch64_crypto_sha512h2qv2di): Ditto.
15303 (aarch64_crypto_sha512su0qv2di): Ditto.
15304 (aarch64_crypto_sha512su1qv2di): Ditto.
15305 (aarch64_eor3qv8hi): Ditto.
15306 (aarch64_rax1qv2di): Ditto.
15307 (aarch64_xarqv2di): Ditto.
15308 (aarch64_bcaxqv8hi): Ditto.
15309 * config/aarch64/aarch64-simd.md:
15310 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
15311 (aarch64_crypto_sha512su0qv2di): Ditto.
15312 (aarch64_crypto_sha512su1qv2di): Ditto.
15313 (aarch64_eor3qv8hi): Ditto.
15314 (aarch64_rax1qv2di): Ditto.
15315 (aarch64_xarqv2di): Ditto.
15316 (aarch64_bcaxqv8hi): Ditto.
15317 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
15318 (vsha512h2q_u64): Ditto.
15319 (vsha512su0q_u64): Ditto.
15320 (vsha512su1q_u64): Ditto.
15321 (veor3q_u16): Ditto.
15322 (vrax1q_u64): Ditto.
15323 (vxarq_u64): Ditto.
15324 (vbcaxq_u16): Ditto.
15325 * config/arm/types.md (crypto_sha512): New type attribute.
15326 (crypto_sha3): Ditto.
15327 (doc/invoke.texi): Document new sha3 option.
15329 2018-01-10 Michael Collison <michael.collison@arm.com>
15331 * config/aarch64/aarch64-builtins.c:
15332 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
15333 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
15334 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
15335 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
15336 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
15337 (AARCH64_ISA_SM4): New ISA flag.
15338 (TARGET_SM4): New feature flag for sm4.
15339 * config/aarch64/aarch64-simd-builtins.def
15340 (aarch64_sm3ss1qv4si): Ditto.
15341 (aarch64_sm3tt1aq4si): Ditto.
15342 (aarch64_sm3tt1bq4si): Ditto.
15343 (aarch64_sm3tt2aq4si): Ditto.
15344 (aarch64_sm3tt2bq4si): Ditto.
15345 (aarch64_sm3partw1qv4si): Ditto.
15346 (aarch64_sm3partw2qv4si): Ditto.
15347 (aarch64_sm4eqv4si): Ditto.
15348 (aarch64_sm4ekeyqv4si): Ditto.
15349 * config/aarch64/aarch64-simd.md:
15350 (aarch64_sm3ss1qv4si): Ditto.
15351 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
15352 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
15353 (aarch64_sm4eqv4si): Ditto.
15354 (aarch64_sm4ekeyqv4si): Ditto.
15355 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
15356 (sm3part_op): Ditto.
15357 (CRYPTO_SM3TT): Ditto.
15358 (CRYPTO_SM3PART): Ditto.
15359 (UNSPEC_SM3SS1): New unspec.
15360 (UNSPEC_SM3TT1A): Ditto.
15361 (UNSPEC_SM3TT1B): Ditto.
15362 (UNSPEC_SM3TT2A): Ditto.
15363 (UNSPEC_SM3TT2B): Ditto.
15364 (UNSPEC_SM3PARTW1): Ditto.
15365 (UNSPEC_SM3PARTW2): Ditto.
15366 (UNSPEC_SM4E): Ditto.
15367 (UNSPEC_SM4EKEY): Ditto.
15368 * config/aarch64/constraints.md (Ui2): New constraint.
15369 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
15370 * config/arm/types.md (crypto_sm3): New type attribute.
15371 (crypto_sm4): Ditto.
15372 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
15373 (vsm3tt1aq_u32): Ditto.
15374 (vsm3tt1bq_u32): Ditto.
15375 (vsm3tt2aq_u32): Ditto.
15376 (vsm3tt2bq_u32): Ditto.
15377 (vsm3partw1q_u32): Ditto.
15378 (vsm3partw2q_u32): Ditto.
15379 (vsm4eq_u32): Ditto.
15380 (vsm4ekeyq_u32): Ditto.
15381 (doc/invoke.texi): Document new sm4 option.
15383 2018-01-10 Michael Collison <michael.collison@arm.com>
15385 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
15386 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
15387 (AARCH64_FL_FOR_ARCH8_4): New.
15388 (AARCH64_FL_V8_4): New flag.
15389 (doc/invoke.texi): Document new armv8.4-a option.
15391 2018-01-10 Michael Collison <michael.collison@arm.com>
15393 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
15394 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
15395 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
15396 * config/aarch64/aarch64-option-extension.def: Add
15397 AARCH64_OPT_EXTENSION of 'sha2'.
15398 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
15399 (crypto): Disable sha2 and aes if crypto disabled.
15400 (crypto): Enable aes and sha2 if enabled.
15401 (simd): Disable sha2 and aes if simd disabled.
15402 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
15404 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
15405 (TARGET_SHA2): New feature flag for sha2.
15406 (TARGET_AES): New feature flag for aes.
15407 * config/aarch64/aarch64-simd.md:
15408 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
15409 conditional on TARGET_AES.
15410 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
15411 (aarch64_crypto_sha1hsi): Make pattern conditional
15413 (aarch64_crypto_sha1hv4si): Ditto.
15414 (aarch64_be_crypto_sha1hv4si): Ditto.
15415 (aarch64_crypto_sha1su1v4si): Ditto.
15416 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
15417 (aarch64_crypto_sha1su0v4si): Ditto.
15418 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
15419 (aarch64_crypto_sha256su0v4si): Ditto.
15420 (aarch64_crypto_sha256su1v4si): Ditto.
15421 (doc/invoke.texi): Document new aes and sha2 options.
15423 2018-01-10 Martin Sebor <msebor@redhat.com>
15425 PR tree-optimization/83781
15426 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
15429 2018-01-11 Martin Sebor <msebor@gmail.com>
15430 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
15432 PR tree-optimization/83501
15433 PR tree-optimization/81703
15435 * tree-ssa-strlen.c (get_string_cst): Rename...
15436 (get_string_len): ...to this. Handle global constants.
15437 (handle_char_store): Adjust.
15439 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
15440 Jim Wilson <jimw@sifive.com>
15442 * config/riscv/riscv-protos.h (riscv_output_return): New.
15443 * config/riscv/riscv.c (struct machine_function): New naked_p field.
15444 (riscv_attribute_table, riscv_output_return),
15445 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
15446 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
15447 (riscv_compute_frame_info): Only compute frame->mask if not a naked
15449 (riscv_expand_prologue): Add early return for naked function.
15450 (riscv_expand_epilogue): Likewise.
15451 (riscv_function_ok_for_sibcall): Return false for naked function.
15452 (riscv_set_current_function): New.
15453 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
15454 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
15455 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
15456 * doc/extend.texi (RISC-V Function Attributes): New.
15458 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
15460 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
15461 check for 128-bit long double before checking TCmode.
15462 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
15463 128-bit long doubles before checking TFmode or TCmode.
15464 (FLOAT128_IBM_P): Likewise.
15466 2018-01-10 Martin Sebor <msebor@redhat.com>
15468 PR tree-optimization/83671
15469 * builtins.c (c_strlen): Unconditionally return zero for the empty
15471 Use -Warray-bounds for warnings.
15472 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
15473 for non-constant array indices with COMPONENT_REF, arrays of
15474 arrays, and pointers to arrays.
15475 (gimple_fold_builtin_strlen): Determine and set length range for
15476 non-constant character arrays.
15478 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
15480 PR middle-end/81897
15481 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
15484 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
15486 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
15488 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
15491 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
15492 VECTOR_MEM_ALTIVEC_OR_VSX_P.
15493 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
15494 indexed_or_indirect_operand predicate.
15495 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
15496 (*vsx_le_perm_load_v8hi): Likewise.
15497 (*vsx_le_perm_load_v16qi): Likewise.
15498 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
15499 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
15500 (*vsx_le_perm_store_v8hi): Likewise.
15501 (*vsx_le_perm_store_v16qi): Likewise.
15502 (eight unnamed splitters): Likewise.
15504 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
15506 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
15507 * config/rs6000/emmintrin.h: Likewise.
15508 * config/rs6000/mmintrin.h: Likewise.
15509 * config/rs6000/xmmintrin.h: Likewise.
15511 2018-01-10 David Malcolm <dmalcolm@redhat.com>
15514 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
15516 * tree.c (tree_nop_conversion): Return true for location wrapper
15518 (maybe_wrap_with_location): New function.
15519 (selftest::check_strip_nops): New function.
15520 (selftest::test_location_wrappers): New function.
15521 (selftest::tree_c_tests): Call it.
15522 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
15523 (maybe_wrap_with_location): New decl.
15524 (EXPR_LOCATION_WRAPPER_P): New macro.
15525 (location_wrapper_p): New inline function.
15526 (tree_strip_any_location_wrapper): New inline function.
15528 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
15531 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
15532 stack_realign_offset for the largest alignment of stack slot
15534 (ix86_find_max_used_stack_alignment): New function.
15535 (ix86_finalize_stack_frame_flags): Use it. Set
15536 max_used_stack_alignment if we don't realign stack.
15537 * config/i386/i386.h (machine_function): Add
15538 max_used_stack_alignment.
15540 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
15542 * config/arm/arm.opt (-mbranch-cost): New option.
15543 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
15546 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
15549 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
15550 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
15552 2018-01-10 Richard Biener <rguenther@suse.de>
15555 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
15556 early out so it also covers the case where we have a non-NULL
15559 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
15561 PR tree-optimization/83753
15562 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
15563 for non-strided grouped accesses if the number of elements is 1.
15565 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
15568 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
15569 * i386.h (TARGET_USE_GATHER): Define.
15570 * x86-tune.def (X86_TUNE_USE_GATHER): New.
15572 2018-01-10 Martin Liska <mliska@suse.cz>
15575 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
15576 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
15578 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
15579 CLEANUP_NO_PARTITIONING is not set.
15581 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
15583 * doc/rtl.texi: Remove documentation of (const ...) wrappers
15584 for vectors, as a partial revert of r254296.
15585 * rtl.h (const_vec_p): Delete.
15586 (const_vec_duplicate_p): Don't test for vector CONSTs.
15587 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
15588 * expmed.c (make_tree): Likewise.
15591 * common.md (E, F): Use CONSTANT_P instead of checking for
15593 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
15594 checking for CONST_VECTOR.
15596 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
15598 PR middle-end/83575
15599 * predict.c (force_edge_cold): Handle in more sane way edges
15600 with no prediction.
15602 2018-01-09 Carl Love <cel@us.ibm.com>
15604 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
15606 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
15607 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
15608 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
15609 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
15610 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
15611 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
15612 * config/rs6000/rs6000-protos.h: Add extern defition for
15613 rs6000_generate_float2_double_code.
15614 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
15616 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
15617 (float2_v2df): Add define_expand.
15619 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
15622 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
15623 op_mode in the force_to_mode call.
15625 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
15627 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
15628 instead of checking each element individually.
15629 (aarch64_evpc_uzp): Likewise.
15630 (aarch64_evpc_zip): Likewise.
15631 (aarch64_evpc_ext): Likewise.
15632 (aarch64_evpc_rev): Likewise.
15633 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
15634 instead of checking each element individually. Return true without
15636 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
15637 whether all selected elements come from the same input, instead of
15638 checking each element individually. Remove calls to gen_rtx_REG,
15639 start_sequence and end_sequence and instead assert that no rtl is
15642 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
15644 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
15645 order of HIGH and CONST checks.
15647 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
15649 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
15650 if the destination isn't an SSA_NAME.
15652 2018-01-09 Richard Biener <rguenther@suse.de>
15654 PR tree-optimization/83668
15655 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
15657 (canonicalize_loop_form): ... here, renamed from ...
15658 (canonicalize_loop_closed_ssa_form): ... this and amended to
15659 swap successor edges for loop exit blocks to make us use
15660 the RPO order we need for initial schedule generation.
15662 2018-01-09 Joseph Myers <joseph@codesourcery.com>
15664 PR tree-optimization/64811
15665 * match.pd: When optimizing comparisons with Inf, avoid
15666 introducing or losing exceptions from comparisons with NaN.
15668 2018-01-09 Martin Liska <mliska@suse.cz>
15671 * asan.c (shadow_mem_size): Add gcc_assert.
15673 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
15675 Don't save registers in main().
15678 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
15679 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
15680 * config/avr/avr.c (avr_set_current_function): Don't error if
15681 naked, OS_task or OS_main are specified at the same time.
15682 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
15684 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
15686 * common/config/avr/avr-common.c (avr_option_optimization_table):
15687 Switch on -mmain-is-OS_task for optimizing compilations.
15689 2018-01-09 Richard Biener <rguenther@suse.de>
15691 PR tree-optimization/83572
15692 * graphite.c: Include cfganal.h.
15693 (graphite_transform_loops): Connect infinite loops to exit
15694 and remove fake edges at the end.
15696 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
15698 * ipa-inline.c (edge_badness): Revert accidental checkin.
15700 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
15703 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
15704 symbols; not inline clones.
15706 2018-01-09 Jakub Jelinek <jakub@redhat.com>
15709 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
15710 hard registers. Formatting fixes.
15712 PR preprocessor/83722
15713 * gcc.c (try_generate_repro): Pass
15714 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
15715 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
15718 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
15719 Kito Cheng <kito.cheng@gmail.com>
15721 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
15722 (riscv_leaf_function_p): Delete.
15723 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
15725 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
15727 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
15729 (do_ifelse): New function.
15730 (do_isel): New function.
15731 (do_sub3): New function.
15732 (do_add3): New function.
15733 (do_load_mask_compare): New function.
15734 (do_overlap_load_compare): New function.
15735 (expand_compare_loop): New function.
15736 (expand_block_compare): Call expand_compare_loop() when appropriate.
15737 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
15738 option description.
15739 (-mblock-compare-inline-loop-limit): New option.
15741 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
15744 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
15745 Reverse order of second and third operands in first alternative.
15746 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
15747 of first and second elements in UNSPEC_VPERMR vector.
15748 (altivec_expand_vec_perm_le): Likewise.
15750 2018-01-08 Jeff Law <law@redhat.com>
15752 PR rtl-optimizatin/81308
15753 * tree-switch-conversion.c (cfg_altered): New file scoped static.
15754 (process_switch): If group_case_labels makes a change, then set
15756 (pass_convert_switch::execute): If a switch is converted, then
15757 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
15759 PR rtl-optimization/81308
15760 * recog.c (split_all_insns): Conditionally cleanup the CFG after
15763 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
15765 PR target/83663 - Revert r255946
15766 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
15767 generation for cases where splatting a value is not useful.
15768 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
15769 across a vec_duplicate and a paradoxical subreg forming a vector
15770 mode to a vec_concat.
15772 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15774 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
15775 -march=armv8.3-a variants.
15776 * config/arm/t-multilib: Likewise.
15777 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
15779 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
15781 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
15783 (cceq_ior_compare_complement): Give it a name so I can use it, and
15784 change boolean_or_operator predicate to boolean_operator so it can
15785 be used to generate a crand.
15786 (eqne): New code iterator.
15787 (bd/bd_neg): New code_attrs.
15788 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
15789 a single define_insn.
15790 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
15791 decrement (bdnzt/bdnzf/bdzt/bdzf).
15792 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
15793 with the new names of the branch decrement patterns, and added the
15794 names of the branch decrement conditional patterns.
15796 2018-01-08 Richard Biener <rguenther@suse.de>
15798 PR tree-optimization/83563
15799 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
15802 2018-01-08 Richard Biener <rguenther@suse.de>
15804 PR middle-end/83713
15805 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
15807 2018-01-08 Richard Biener <rguenther@suse.de>
15809 PR tree-optimization/83685
15810 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
15811 references to abnormals.
15813 2018-01-08 Richard Biener <rguenther@suse.de>
15816 * dwarf2out.c (output_indirect_strings): Handle empty
15817 skeleton_debug_str_hash.
15818 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
15820 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
15822 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
15823 (emit_store_direct): Likewise.
15824 (arc_trampoline_adjust_address): Likewise.
15825 (arc_asm_trampoline_template): New function.
15826 (arc_initialize_trampoline): Use asm_trampoline_template.
15827 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
15828 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
15829 * config/arc/arc.md (flush_icache): Delete pattern.
15831 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
15833 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
15834 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
15837 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
15840 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
15841 by not USED_FOR_TARGET.
15842 (make_pass_resolve_sw_modes): Likewise.
15844 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
15846 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
15849 2018-01-08 Richard Biener <rguenther@suse.de>
15851 PR middle-end/83580
15852 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
15854 2018-01-08 Richard Biener <rguenther@suse.de>
15856 PR middle-end/83517
15857 * match.pd ((t * 2) / 2) -> t): Add missing :c.
15859 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
15861 PR middle-end/81897
15862 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
15863 basic blocks with a small number of successors.
15864 (convert_control_dep_chain_into_preds): Improve handling of
15866 (dump_predicates): Split apart into...
15867 (dump_pred_chain): ...here...
15868 (dump_pred_info): ...and here.
15869 (can_one_predicate_be_invalidated_p): Add debugging printfs.
15870 (can_chain_union_be_invalidated_p): Improve check for invalidation
15872 (uninit_uses_cannot_happen): Avoid unnecessary if
15873 convert_control_dep_chain_into_preds yielded nothing.
15875 2018-01-06 Martin Sebor <msebor@redhat.com>
15877 PR tree-optimization/83640
15878 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
15879 subtracting negative offset from size.
15880 (builtin_access::overlap): Adjust offset bounds of the access to fall
15881 within the size of the object if possible.
15883 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
15885 PR rtl-optimization/83699
15886 * expmed.c (extract_bit_field_1): Restrict the vector usage of
15887 extract_bit_field_as_subreg to cases in which the extracted
15888 value is also a vector.
15890 * lra-constraints.c (process_alt_operands): Test for the equivalence
15891 substitutions when detecting a possible reload cycle.
15893 2018-01-06 Jakub Jelinek <jakub@redhat.com>
15896 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
15897 by default if flag_selective_schedling{,2}. Formatting fixes.
15899 PR rtl-optimization/83682
15900 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
15901 if it has non-VECTOR_MODE element mode.
15902 (vec_duplicate_p): Likewise.
15904 PR middle-end/83694
15905 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
15906 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
15908 2018-01-05 Jakub Jelinek <jakub@redhat.com>
15911 * config/i386/i386-builtin.def
15912 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
15913 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
15914 Require also OPTION_MASK_ISA_AVX512F in addition to
15915 OPTION_MASK_ISA_GFNI.
15916 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
15917 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
15918 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
15919 to OPTION_MASK_ISA_GFNI.
15920 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
15921 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
15922 OPTION_MASK_ISA_AVX512BW.
15923 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
15924 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
15925 addition to OPTION_MASK_ISA_GFNI.
15926 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
15927 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
15928 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
15929 to OPTION_MASK_ISA_GFNI.
15930 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
15931 a requirement for all ISAs rather than any of them with a few
15933 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
15935 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
15936 bitmasks to be enabled with 3 exceptions, instead of requiring any
15937 enabled ISA with lots of exceptions.
15938 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
15939 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
15940 Change avx512bw in isa attribute to avx512f.
15941 * config/i386/sgxintrin.h: Add license boilerplate.
15942 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
15943 to __AVX512F__ and __AVX512VL to __AVX512VL__.
15944 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
15945 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
15947 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
15948 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
15949 temporarily sse2 rather than sse if not enabled already.
15952 * config/i386/sse.md (VI248_VLBW): Rename to ...
15953 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
15954 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
15955 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
15956 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
15957 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
15958 mode iterator instead of VI248_VLBW.
15960 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
15962 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
15963 (record_modified): Skip clobbers; add debug output.
15964 (param_change_prob): Use sreal frequencies.
15966 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
15968 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
15969 punt for user-aligned variables.
15971 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
15973 * tree-chrec.c (chrec_contains_symbols): Return true for
15976 2018-01-05 Sudakshina Das <sudi.das@arm.com>
15979 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
15980 of (x|y) == x for BICS pattern.
15982 2018-01-05 Jakub Jelinek <jakub@redhat.com>
15984 PR tree-optimization/83605
15985 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
15986 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
15989 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
15991 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
15992 * config/epiphany/rtems.h: New file.
15994 2018-01-04 Jakub Jelinek <jakub@redhat.com>
15995 Uros Bizjak <ubizjak@gmail.com>
15998 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
15999 QIreg_operand instead of register_operand predicate.
16000 * config/i386/i386.c (ix86_rop_should_change_byte_p,
16001 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
16002 comments instead of -fmitigate[-_]rop.
16004 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
16007 * cgraphunit.c (symbol_table::compile): Switch to text_section
16008 before calling assembly_start debug hook.
16009 * run-rtl-passes.c (run_rtl_passes): Likewise.
16012 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
16014 * tree-vrp.c (extract_range_from_binary_expr_1): Check
16015 range_int_cst_p rather than !symbolic_range_p before calling
16016 extract_range_from_multiplicative_op_1.
16018 2018-01-04 Jeff Law <law@redhat.com>
16020 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
16021 redundant test in assertion.
16023 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
16025 * doc/rtl.texi: Document machine_mode wrapper classes.
16027 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
16029 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
16030 using tree_to_uhwi.
16032 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
16034 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
16035 the VEC_PERM_EXPR fold to fail.
16037 2018-01-04 Jakub Jelinek <jakub@redhat.com>
16040 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
16041 to switched_sections.
16043 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
16046 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
16047 test for d.testing.
16049 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
16052 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
16053 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
16055 2018-01-04 Jakub Jelinek <jakub@redhat.com>
16058 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
16059 is BLKmode and bitpos not zero or mode change is needed.
16061 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
16064 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
16067 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
16070 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
16071 instead of MULT rtx. Update all corresponding splitters.
16072 (*saddl_se): Ditto.
16073 (*ssub<modesuffix>): Ditto.
16074 (*ssubl_se): Ditto.
16075 (*cmp_sadd_di): Update split patterns.
16076 (*cmp_sadd_si): Ditto.
16077 (*cmp_sadd_sidi): Ditto.
16078 (*cmp_ssub_di): Ditto.
16079 (*cmp_ssub_si): Ditto.
16080 (*cmp_ssub_sidi): Ditto.
16081 * config/alpha/predicates.md (const23_operand): New predicate.
16082 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
16083 Look for ASHIFT, not MULT inner operand.
16084 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
16086 2018-01-04 Martin Liska <mliska@suse.cz>
16088 PR gcov-profile/83669
16089 * gcov.c (output_intermediate_file): Add version to intermediate
16091 * doc/gcov.texi: Document new field 'version' in intermediate
16092 file format. Fix location of '-k' option of gcov command.
16094 2018-01-04 Martin Liska <mliska@suse.cz>
16097 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
16099 2018-01-04 Jakub Jelinek <jakub@redhat.com>
16101 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
16103 2018-01-03 Martin Sebor <msebor@redhat.com>
16105 PR tree-optimization/83655
16106 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
16107 checking calls with invalid arguments.
16109 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16111 * tree-vect-stmts.c (vect_get_store_rhs): New function.
16112 (vectorizable_mask_load_store): Delete.
16113 (vectorizable_call): Return false for masked loads and stores.
16114 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
16115 instead of gimple_assign_rhs1.
16116 (vectorizable_load): Handle IFN_MASK_LOAD.
16117 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
16119 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16121 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
16123 (vectorizable_mask_load_store): ...here.
16124 (vectorizable_load): ...and here.
16126 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16128 * tree-vect-stmts.c (vect_build_all_ones_mask)
16129 (vect_build_zero_merge_argument): New functions, split out from...
16130 (vectorizable_load): ...here.
16132 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16134 * tree-vect-stmts.c (vect_check_store_rhs): New function,
16136 (vectorizable_mask_load_store): ...here.
16137 (vectorizable_store): ...and here.
16139 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16141 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
16143 (vectorizable_mask_load_store): ...here.
16145 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16147 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
16148 (vect_model_store_cost): Take a vec_load_store_type instead of a
16150 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
16151 (vect_model_store_cost): Take a vec_load_store_type instead of a
16153 (vectorizable_mask_load_store): Update accordingly.
16154 (vectorizable_store): Likewise.
16155 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
16157 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16159 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
16160 IFN_MASK_LOAD calls here rather than...
16161 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
16163 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16164 Alan Hayward <alan.hayward@arm.com>
16165 David Sherwood <david.sherwood@arm.com>
16167 * expmed.c (extract_bit_field_1): For vector extracts,
16168 fall back to extract_bit_field_as_subreg if vec_extract
16171 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16172 Alan Hayward <alan.hayward@arm.com>
16173 David Sherwood <david.sherwood@arm.com>
16175 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
16176 they are variable or constant sized.
16177 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
16178 slots for constant-sized data.
16180 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16181 Alan Hayward <alan.hayward@arm.com>
16182 David Sherwood <david.sherwood@arm.com>
16184 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
16185 handling COND_EXPRs with boolean comparisons, try to find a better
16186 basis for the mask type than the boolean itself.
16188 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16190 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
16191 is calculated and how it can be overridden.
16192 * genmodes.c (max_bitsize_mode_any_mode): New variable.
16193 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
16195 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
16198 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16199 Alan Hayward <alan.hayward@arm.com>
16200 David Sherwood <david.sherwood@arm.com>
16202 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
16203 Remove the mode argument.
16204 (aarch64_simd_valid_immediate): Remove the mode and inverse
16206 * config/aarch64/iterators.md (bitsize): New iterator.
16207 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
16208 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
16209 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
16210 aarch64_simd_valid_immediate.
16211 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
16212 (aarch64_reg_or_bic_imm): Likewise.
16213 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
16214 with an insn_type enum and msl with a modifier_type enum.
16215 Replace element_width with a scalar_mode. Change the shift
16216 to unsigned int. Add constructors for scalar_float_mode and
16217 scalar_int_mode elements.
16218 (aarch64_vect_float_const_representable_p): Delete.
16219 (aarch64_can_const_movi_rtx_p)
16220 (aarch64_simd_scalar_immediate_valid_for_move)
16221 (aarch64_simd_make_constant): Update call to
16222 aarch64_simd_valid_immediate.
16223 (aarch64_advsimd_valid_immediate_hs): New function.
16224 (aarch64_advsimd_valid_immediate): Likewise.
16225 (aarch64_simd_valid_immediate): Remove mode and inverse
16226 arguments. Rewrite to use the above. Use const_vec_duplicate_p
16227 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
16228 and aarch64_float_const_representable_p on the result.
16229 (aarch64_output_simd_mov_immediate): Remove mode argument.
16230 Update call to aarch64_simd_valid_immediate and use of
16231 simd_immediate_info.
16232 (aarch64_output_scalar_simd_mov_immediate): Update call
16235 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16236 Alan Hayward <alan.hayward@arm.com>
16237 David Sherwood <david.sherwood@arm.com>
16239 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
16240 (mode_nunits): Likewise CONST_MODE_NUNITS.
16241 * machmode.def (ADJUST_NUNITS): Document.
16242 * genmodes.c (mode_data::need_nunits_adj): New field.
16243 (blank_mode): Update accordingly.
16244 (adj_nunits): New variable.
16245 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
16247 (emit_mode_size_inline): Set need_bytesize_adj for all modes
16248 listed in adj_nunits.
16249 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
16250 listed in adj_nunits. Don't emit case statements for such modes.
16251 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
16252 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
16253 nothing if adj_nunits is nonnull.
16254 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
16255 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
16256 (emit_mode_fbit): Update use of print_maybe_const_decl.
16257 (emit_move_size): Likewise. Treat the array as non-const
16259 (emit_mode_adjustments): Handle adj_nunits.
16261 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16263 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
16264 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
16265 (VECTOR_MODES): Use it.
16266 (make_vector_modes): Take the prefix as an argument.
16268 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16269 Alan Hayward <alan.hayward@arm.com>
16270 David Sherwood <david.sherwood@arm.com>
16272 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
16273 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
16274 for MODE_VECTOR_BOOL.
16275 * machmode.def (VECTOR_BOOL_MODE): Document.
16276 * genmodes.c (VECTOR_BOOL_MODE): New macro.
16277 (make_vector_bool_mode): New function.
16278 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
16280 * lto-streamer-in.c (lto_input_mode_table): Likewise.
16281 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
16283 * stor-layout.c (int_mode_for_mode): Likewise.
16284 * tree.c (build_vector_type_for_mode): Likewise.
16285 * varasm.c (output_constant_pool_2): Likewise.
16286 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
16287 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
16288 for MODE_VECTOR_BOOL.
16289 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
16290 of mode class checks.
16291 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
16292 instead of a list of mode class checks.
16293 (expand_vector_scalar_condition): Likewise.
16294 (type_for_widest_vector_mode): Handle BImode as an inner mode.
16296 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16297 Alan Hayward <alan.hayward@arm.com>
16298 David Sherwood <david.sherwood@arm.com>
16300 * machmode.h (mode_size): Change from unsigned short to
16302 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
16303 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
16304 or if measurement_type is not polynomial.
16305 (fixed_size_mode::includes_p): Check for constant-sized modes.
16306 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
16307 return a poly_uint16 rather than an unsigned short.
16308 (emit_mode_size): Change the type of mode_size from unsigned short
16309 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
16310 (emit_mode_adjustments): Cope with polynomial vector sizes.
16311 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
16313 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
16315 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
16316 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
16317 * caller-save.c (setup_save_areas): Likewise.
16318 (replace_reg_with_saved_mem): Likewise.
16319 * calls.c (emit_library_call_value_1): Likewise.
16320 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
16321 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
16322 (gen_lowpart_for_combine): Likewise.
16323 * convert.c (convert_to_integer_1): Likewise.
16324 * cse.c (equiv_constant, cse_insn): Likewise.
16325 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
16326 (cselib_subst_to_values): Likewise.
16327 * dce.c (word_dce_process_block): Likewise.
16328 * df-problems.c (df_word_lr_mark_ref): Likewise.
16329 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
16330 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
16331 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
16332 (rtl_for_decl_location): Likewise.
16333 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
16334 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
16335 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
16336 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
16337 (expand_expr_real_1): Likewise.
16338 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
16339 (pad_below): Likewise.
16340 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
16341 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
16342 * ira.c (get_subreg_tracking_sizes): Likewise.
16343 * ira-build.c (ira_create_allocno_objects): Likewise.
16344 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
16345 (ira_sort_regnos_for_alter_reg): Likewise.
16346 * ira-costs.c (record_operand_costs): Likewise.
16347 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
16348 (resolve_simple_move): Likewise.
16349 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
16350 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
16351 (lra_constraints): Likewise.
16352 (CONST_POOL_OK_P): Reject variable-sized modes.
16353 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
16354 (add_pseudo_to_slot, lra_spill): Likewise.
16355 * omp-low.c (omp_clause_aligned_alignment): Likewise.
16356 * optabs-query.c (get_best_extraction_insn): Likewise.
16357 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
16358 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
16359 (expand_mult_highpart, valid_multiword_target_p): Likewise.
16360 * recog.c (offsettable_address_addr_space_p): Likewise.
16361 * regcprop.c (maybe_mode_change): Likewise.
16362 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
16363 * regrename.c (build_def_use): Likewise.
16364 * regstat.c (dump_reg_info): Likewise.
16365 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
16366 (find_reloads, find_reloads_subreg_address): Likewise.
16367 * reload1.c (eliminate_regs_1): Likewise.
16368 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
16369 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
16370 (simplify_binary_operation_1, simplify_subreg): Likewise.
16371 * targhooks.c (default_function_arg_padding): Likewise.
16372 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
16373 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
16374 (verify_gimple_assign_ternary): Likewise.
16375 * tree-inline.c (estimate_move_cost): Likewise.
16376 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
16377 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
16378 (get_address_cost_ainc): Likewise.
16379 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
16380 (vect_supportable_dr_alignment): Likewise.
16381 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
16382 (vectorizable_reduction): Likewise.
16383 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
16384 (vectorizable_operation, vectorizable_load): Likewise.
16385 * tree.c (build_same_sized_truth_vector_type): Likewise.
16386 * valtrack.c (cleanup_auto_inc_dec): Likewise.
16387 * var-tracking.c (emit_note_insn_var_location): Likewise.
16388 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
16389 (ADDR_VEC_ALIGN): Likewise.
16391 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16392 Alan Hayward <alan.hayward@arm.com>
16393 David Sherwood <david.sherwood@arm.com>
16395 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
16397 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
16398 or if measurement_type is polynomial.
16399 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
16400 * combine.c (make_extraction): Likewise.
16401 * dse.c (find_shift_sequence): Likewise.
16402 * dwarf2out.c (mem_loc_descriptor): Likewise.
16403 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
16404 (extract_bit_field, extract_low_bits): Likewise.
16405 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
16406 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
16407 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
16408 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
16409 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
16410 * reload.c (find_reloads): Likewise.
16411 * reload1.c (alter_reg): Likewise.
16412 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
16413 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
16414 * tree-if-conv.c (predicate_mem_writes): Likewise.
16415 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
16416 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
16417 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
16418 * valtrack.c (dead_debug_insert_temp): Likewise.
16419 * varasm.c (mergeable_constant_section): Likewise.
16420 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
16422 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16423 Alan Hayward <alan.hayward@arm.com>
16424 David Sherwood <david.sherwood@arm.com>
16426 * expr.c (expand_assignment): Cope with polynomial mode sizes
16427 when assigning to a CONCAT.
16429 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16430 Alan Hayward <alan.hayward@arm.com>
16431 David Sherwood <david.sherwood@arm.com>
16433 * machmode.h (mode_precision): Change from unsigned short to
16435 (mode_to_precision): Return a poly_uint16 rather than an unsigned
16437 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
16438 or if measurement_type is not polynomial.
16439 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
16440 in which the mode is already known to be a scalar_int_mode.
16441 * genmodes.c (emit_mode_precision): Change the type of mode_precision
16442 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
16444 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
16445 for GET_MODE_PRECISION.
16446 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
16447 for GET_MODE_PRECISION.
16448 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
16450 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
16451 (expand_field_assignment, make_extraction): Likewise.
16452 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
16453 (get_last_value): Likewise.
16454 * convert.c (convert_to_integer_1): Likewise.
16455 * cse.c (cse_insn): Likewise.
16456 * expr.c (expand_expr_real_1): Likewise.
16457 * lra-constraints.c (simplify_operand_subreg): Likewise.
16458 * optabs-query.c (can_atomic_load_p): Likewise.
16459 * optabs.c (expand_atomic_load): Likewise.
16460 (expand_atomic_store): Likewise.
16461 * ree.c (combine_reaching_defs): Likewise.
16462 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
16463 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
16464 * tree.h (type_has_mode_precision_p): Likewise.
16465 * ubsan.c (instrument_si_overflow): Likewise.
16467 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16468 Alan Hayward <alan.hayward@arm.com>
16469 David Sherwood <david.sherwood@arm.com>
16471 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
16472 polynomial numbers of units.
16473 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
16474 (valid_vector_subparts_p): New function.
16475 (build_vector_type): Remove temporary shim and take the number
16476 of units as a poly_uint64 rather than an int.
16477 (build_opaque_vector_type): Take the number of units as a
16478 poly_uint64 rather than an int.
16479 * tree.c (build_vector_from_ctor): Handle polynomial
16480 TYPE_VECTOR_SUBPARTS.
16481 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
16482 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
16483 (build_vector_from_val): If the number of units is variable,
16484 use build_vec_duplicate_cst for constant operands and
16485 VEC_DUPLICATE_EXPR otherwise.
16486 (make_vector_type): Remove temporary is_constant ().
16487 (build_vector_type, build_opaque_vector_type): Take the number of
16488 units as a poly_uint64 rather than an int.
16489 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
16491 * cfgexpand.c (expand_debug_expr): Likewise.
16492 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
16493 (store_constructor, expand_expr_real_1): Likewise.
16494 (const_scalar_mask_from_tree): Likewise.
16495 * fold-const-call.c (fold_const_reduction): Likewise.
16496 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
16497 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
16498 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
16499 (fold_relational_const): Likewise.
16500 (native_interpret_vector): Likewise. Change the size from an
16501 int to an unsigned int.
16502 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
16503 TYPE_VECTOR_SUBPARTS.
16504 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
16505 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
16506 duplicating a non-constant operand into a variable-length vector.
16507 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
16508 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
16509 * ipa-icf.c (sem_variable::equals): Likewise.
16510 * match.pd: Likewise.
16511 * omp-simd-clone.c (simd_clone_subparts): Likewise.
16512 * print-tree.c (print_node): Likewise.
16513 * stor-layout.c (layout_type): Likewise.
16514 * targhooks.c (default_builtin_vectorization_cost): Likewise.
16515 * tree-cfg.c (verify_gimple_comparison): Likewise.
16516 (verify_gimple_assign_binary): Likewise.
16517 (verify_gimple_assign_ternary): Likewise.
16518 (verify_gimple_assign_single): Likewise.
16519 * tree-pretty-print.c (dump_generic_node): Likewise.
16520 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
16521 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
16522 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
16523 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
16524 (vect_shift_permute_load_chain): Likewise.
16525 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
16526 (expand_vector_condition, optimize_vector_constructor): Likewise.
16527 (lower_vec_perm, get_compute_type): Likewise.
16528 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
16529 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
16530 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
16531 (vect_recog_mask_conversion_pattern): Likewise.
16532 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
16533 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
16534 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
16535 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
16536 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
16537 (vectorizable_shift, vectorizable_operation, vectorizable_store)
16538 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
16539 (supportable_widening_operation): Likewise.
16540 (supportable_narrowing_operation): Likewise.
16541 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
16543 * varasm.c (output_constant): Likewise.
16545 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16546 Alan Hayward <alan.hayward@arm.com>
16547 David Sherwood <david.sherwood@arm.com>
16549 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
16550 so that both the length == 3 and length != 3 cases set up their
16551 own permute vectors. Add comments explaining why we know the
16552 number of elements is constant.
16553 (vect_permute_load_chain): Likewise.
16555 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16556 Alan Hayward <alan.hayward@arm.com>
16557 David Sherwood <david.sherwood@arm.com>
16559 * machmode.h (mode_nunits): Change from unsigned char to
16561 (ONLY_FIXED_SIZE_MODES): New macro.
16562 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
16563 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
16564 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
16566 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
16567 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
16568 or if measurement_type is not polynomial.
16569 * genmodes.c (ZERO_COEFFS): New macro.
16570 (emit_mode_nunits_inline): Make mode_nunits_inline return a
16572 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
16573 Use ZERO_COEFFS when emitting initializers.
16574 * data-streamer.h (bp_pack_poly_value): New function.
16575 (bp_unpack_poly_value): Likewise.
16576 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
16577 for GET_MODE_NUNITS.
16578 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
16579 for GET_MODE_NUNITS.
16580 * tree.c (make_vector_type): Remove temporary shim and make
16581 the real function take the number of units as a poly_uint64
16582 rather than an int.
16583 (build_vector_type_for_mode): Handle polynomial nunits.
16584 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
16585 * emit-rtl.c (const_vec_series_p_1): Likewise.
16586 (gen_rtx_CONST_VECTOR): Likewise.
16587 * fold-const.c (test_vec_duplicate_folding): Likewise.
16588 * genrecog.c (validate_pattern): Likewise.
16589 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
16590 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
16591 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
16592 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
16593 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
16594 * rtlanal.c (subreg_get_info): Likewise.
16595 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
16596 (vect_grouped_load_supported): Likewise.
16597 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
16598 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
16599 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
16600 (simplify_const_unary_operation, simplify_binary_operation_1)
16601 (simplify_const_binary_operation, simplify_ternary_operation)
16602 (test_vector_ops_duplicate, test_vector_ops): Likewise.
16603 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
16604 instead of CONST_VECTOR_NUNITS.
16605 * varasm.c (output_constant_pool_2): Likewise.
16606 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
16607 explicit-encoded elements in the XVEC for variable-length vectors.
16609 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16611 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
16613 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16614 Alan Hayward <alan.hayward@arm.com>
16615 David Sherwood <david.sherwood@arm.com>
16617 * coretypes.h (fixed_size_mode): Declare.
16618 (fixed_size_mode_pod): New typedef.
16619 * builtins.h (target_builtins::x_apply_args_mode)
16620 (target_builtins::x_apply_result_mode): Change type to
16621 fixed_size_mode_pod.
16622 * builtins.c (apply_args_size, apply_result_size, result_vector)
16623 (expand_builtin_apply_args_1, expand_builtin_apply)
16624 (expand_builtin_return): Update accordingly.
16626 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16628 * cse.c (hash_rtx_cb): Hash only the encoded elements.
16629 * cselib.c (cselib_hash_rtx): Likewise.
16630 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
16631 CONST_VECTOR encoding.
16633 2018-01-03 Jakub Jelinek <jakub@redhat.com>
16634 Jeff Law <law@redhat.com>
16637 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
16638 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
16639 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
16640 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
16643 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
16644 explicitly probe *sp in a noreturn function if there were any callee
16645 register saves or frame pointer is needed.
16647 2018-01-03 Jakub Jelinek <jakub@redhat.com>
16650 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
16651 BLKmode for ternary, binary or unary expressions.
16654 * var-tracking.c (delete_vta_debug_insn): New inline function.
16655 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
16656 insns from get_insns () to NULL instead of each bb separately.
16657 Use delete_vta_debug_insn. No longer static.
16658 (vt_debug_insns_local, variable_tracking_main_1): Adjust
16659 delete_vta_debug_insns callers.
16660 * rtl.h (delete_vta_debug_insns): Declare.
16661 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
16662 instead of variable_tracking_main.
16664 2018-01-03 Martin Sebor <msebor@redhat.com>
16666 PR tree-optimization/83603
16667 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
16668 arguments past the endof the argument list in functions declared
16669 without a prototype.
16670 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
16671 Avoid checking when arguments are null.
16673 2018-01-03 Martin Sebor <msebor@redhat.com>
16676 * doc/extend.texi (attribute const): Fix a typo.
16677 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
16678 issuing -Wsuggest-attribute for void functions.
16680 2018-01-03 Martin Sebor <msebor@redhat.com>
16682 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
16683 offset_int::from instead of wide_int::to_shwi.
16684 (maybe_diag_overlap): Remove assertion.
16685 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
16686 * gimple-ssa-sprintf.c (format_directive): Same.
16687 (parse_directive): Same.
16688 (sprintf_dom_walker::compute_format_length): Same.
16689 (try_substitute_return_value): Same.
16691 2018-01-03 Jeff Law <law@redhat.com>
16693 PR middle-end/83654
16694 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
16695 non-constant residual for zero at runtime and avoid probing in
16696 that case. Reorganize code for trailing problem to mirror handling
16699 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
16701 PR tree-optimization/83501
16702 * tree-ssa-strlen.c (get_string_cst): New.
16703 (handle_char_store): Call get_string_cst.
16705 2018-01-03 Martin Liska <mliska@suse.cz>
16707 PR tree-optimization/83593
16708 * tree-ssa-strlen.c: Include tree-cfg.h.
16709 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
16710 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
16711 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
16713 (strlen_dom_walker::before_dom_children): Call
16714 gimple_purge_dead_eh_edges. Dump tranformation with details
16716 (strlen_dom_walker::before_dom_children): Update call by adding
16717 new argument cleanup_eh.
16718 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
16720 2018-01-03 Martin Liska <mliska@suse.cz>
16723 * cif-code.def (VARIADIC_THUNK): New enum value.
16724 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
16727 2018-01-03 Jan Beulich <jbeulich@suse.com>
16729 * sse.md (mov<mode>_internal): Tighten condition for when to use
16730 vmovdqu<ssescalarsize> for TI and OI modes.
16732 2018-01-03 Jakub Jelinek <jakub@redhat.com>
16734 Update copyright years.
16736 2018-01-03 Martin Liska <mliska@suse.cz>
16739 * ipa-visibility.c (function_and_variable_visibility): Skip
16740 functions with noipa attribure.
16742 2018-01-03 Jakub Jelinek <jakub@redhat.com>
16744 * gcc.c (process_command): Update copyright notice dates.
16745 * gcov-dump.c (print_version): Ditto.
16746 * gcov.c (print_version): Ditto.
16747 * gcov-tool.c (print_version): Ditto.
16748 * gengtype.c (create_file): Ditto.
16749 * doc/cpp.texi: Bump @copying's copyright year.
16750 * doc/cppinternals.texi: Ditto.
16751 * doc/gcc.texi: Ditto.
16752 * doc/gccint.texi: Ditto.
16753 * doc/gcov.texi: Ditto.
16754 * doc/install.texi: Ditto.
16755 * doc/invoke.texi: Ditto.
16757 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16759 * vector-builder.h (vector_builder::m_full_nelts): Change from
16760 unsigned int to poly_uint64.
16761 (vector_builder::full_nelts): Update prototype accordingly.
16762 (vector_builder::new_vector): Likewise.
16763 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
16764 (vector_builder::operator ==): Likewise.
16765 (vector_builder::finalize): Likewise.
16766 * int-vector-builder.h (int_vector_builder::int_vector_builder):
16767 Take the number of elements as a poly_uint64 rather than an
16769 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
16770 from unsigned int to poly_uint64.
16771 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
16772 (vec_perm_indices::new_vector): Likewise.
16773 (vec_perm_indices::length): Likewise.
16774 (vec_perm_indices::nelts_per_input): Likewise.
16775 (vec_perm_indices::input_nelts): Likewise.
16776 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
16777 number of elements per input as a poly_uint64 rather than an
16778 unsigned int. Use the original encoding for variable-length
16779 vectors, rather than clamping each individual element.
16780 For the second and subsequent elements in each pattern,
16781 clamp the step and base before clamping their sum.
16782 (vec_perm_indices::series_p): Handle polynomial element counts.
16783 (vec_perm_indices::all_in_range_p): Likewise.
16784 (vec_perm_indices_to_tree): Likewise.
16785 (vec_perm_indices_to_rtx): Likewise.
16786 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
16787 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
16788 (tree_vector_builder::new_binary_operation): Handle polynomial
16789 element counts. Return false if we need to know the number
16790 of elements at compile time.
16791 * fold-const.c (fold_vec_perm): Punt if the number of elements
16792 isn't known at compile time.
16794 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16796 * vec-perm-indices.h (vec_perm_builder): Change element type
16797 from HOST_WIDE_INT to poly_int64.
16798 (vec_perm_indices::element_type): Update accordingly.
16799 (vec_perm_indices::clamp): Handle polynomial element_types.
16800 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
16801 (vec_perm_indices::all_in_range_p): Likewise.
16802 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
16804 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
16805 polynomial vec_perm_indices element types.
16806 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
16807 * fold-const.c (fold_vec_perm): Likewise.
16808 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
16809 * tree-vect-generic.c (lower_vec_perm): Likewise.
16810 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
16811 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
16812 element type to HOST_WIDE_INT.
16814 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16815 Alan Hayward <alan.hayward@arm.com>
16816 David Sherwood <david.sherwood@arm.com>
16818 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
16819 rather than an int. Use plus_constant.
16820 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
16821 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
16823 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16824 Alan Hayward <alan.hayward@arm.com>
16825 David Sherwood <david.sherwood@arm.com>
16827 * calls.c (emit_call_1, expand_call): Change struct_value_size from
16828 a HOST_WIDE_INT to a poly_int64.
16830 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16831 Alan Hayward <alan.hayward@arm.com>
16832 David Sherwood <david.sherwood@arm.com>
16834 * calls.c (load_register_parameters): Cope with polynomial
16835 mode sizes. Require a constant size for BLKmode parameters
16836 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
16837 forces a parameter to be padded at the lsb end in order to
16838 fill a complete number of words, require the parameter size
16839 to be ordered wrt UNITS_PER_WORD.
16841 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16842 Alan Hayward <alan.hayward@arm.com>
16843 David Sherwood <david.sherwood@arm.com>
16845 * reload1.c (spill_stack_slot_width): Change element type
16846 from unsigned int to poly_uint64_pod.
16847 (alter_reg): Treat mode sizes as polynomial.
16849 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16850 Alan Hayward <alan.hayward@arm.com>
16851 David Sherwood <david.sherwood@arm.com>
16853 * reload.c (complex_word_subreg_p): New function.
16854 (reload_inner_reg_of_subreg, push_reload): Use it.
16856 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16857 Alan Hayward <alan.hayward@arm.com>
16858 David Sherwood <david.sherwood@arm.com>
16860 * lra-constraints.c (process_alt_operands): Reject matched
16861 operands whose sizes aren't ordered.
16862 (match_reload): Refer to this check here.
16864 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16865 Alan Hayward <alan.hayward@arm.com>
16866 David Sherwood <david.sherwood@arm.com>
16868 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
16869 that the mode size is in the set {1, 2, 4, 8, 16}.
16871 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16872 Alan Hayward <alan.hayward@arm.com>
16873 David Sherwood <david.sherwood@arm.com>
16875 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
16876 Use plus_constant instead of gen_rtx_PLUS.
16878 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16879 Alan Hayward <alan.hayward@arm.com>
16880 David Sherwood <david.sherwood@arm.com>
16882 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
16883 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
16884 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
16885 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
16886 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
16887 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
16888 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
16889 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
16890 * config/i386/i386.c (ix86_push_rounding): ...this new function.
16891 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
16893 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
16894 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
16895 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
16896 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
16897 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
16898 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
16899 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
16900 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
16901 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
16902 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
16904 * expr.c (emit_move_resolve_push): Treat the input and result
16905 of PUSH_ROUNDING as a poly_int64.
16906 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
16907 (emit_push_insn): Likewise.
16908 * lra-eliminations.c (mark_not_eliminable): Likewise.
16909 * recog.c (push_operand): Likewise.
16910 * reload1.c (elimination_effects): Likewise.
16911 * rtlanal.c (nonzero_bits1): Likewise.
16912 * calls.c (store_one_arg): Likewise. Require the padding to be
16913 known at compile time.
16915 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16916 Alan Hayward <alan.hayward@arm.com>
16917 David Sherwood <david.sherwood@arm.com>
16919 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
16920 Use plus_constant instead of gen_rtx_PLUS.
16922 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16923 Alan Hayward <alan.hayward@arm.com>
16924 David Sherwood <david.sherwood@arm.com>
16926 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
16927 rather than an int.
16929 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16930 Alan Hayward <alan.hayward@arm.com>
16931 David Sherwood <david.sherwood@arm.com>
16933 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
16934 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
16935 via stack temporaries. Treat the mode size as polynomial too.
16937 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16938 Alan Hayward <alan.hayward@arm.com>
16939 David Sherwood <david.sherwood@arm.com>
16941 * expr.c (expand_expr_real_2): When handling conversions involving
16942 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
16943 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
16944 as a poly_uint64 too.
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 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
16952 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16953 Alan Hayward <alan.hayward@arm.com>
16954 David Sherwood <david.sherwood@arm.com>
16956 * combine.c (can_change_dest_mode): Handle polynomial
16957 REGMODE_NATURAL_SIZE.
16958 * expmed.c (store_bit_field_1): Likewise.
16959 * expr.c (store_constructor): Likewise.
16960 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
16961 and polynomial REGMODE_NATURAL_SIZE.
16962 (gen_lowpart_common): Likewise.
16963 * reginfo.c (record_subregs_of_mode): Likewise.
16964 * rtlanal.c (read_modify_subreg_p): Likewise.
16966 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16967 Alan Hayward <alan.hayward@arm.com>
16968 David Sherwood <david.sherwood@arm.com>
16970 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
16971 numbers of elements.
16973 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16974 Alan Hayward <alan.hayward@arm.com>
16975 David Sherwood <david.sherwood@arm.com>
16977 * match.pd: Cope with polynomial numbers of vector elements.
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 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
16984 in a POINTER_PLUS_EXPR.
16986 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16987 Alan Hayward <alan.hayward@arm.com>
16988 David Sherwood <david.sherwood@arm.com>
16990 * omp-simd-clone.c (simd_clone_subparts): New function.
16991 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
16992 (ipa_simd_modify_function_body): Likewise.
16994 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16995 Alan Hayward <alan.hayward@arm.com>
16996 David Sherwood <david.sherwood@arm.com>
16998 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
16999 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
17000 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
17001 (expand_vector_condition, vector_element): Likewise.
17002 (subparts_gt): New function.
17003 (get_compute_type): Use subparts_gt.
17004 (count_type_subparts): Delete.
17005 (expand_vector_operations_1): Use subparts_gt instead of
17006 count_type_subparts.
17008 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17009 Alan Hayward <alan.hayward@arm.com>
17010 David Sherwood <david.sherwood@arm.com>
17012 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
17013 (vect_compile_time_alias): ...this new function. Do the calculation
17014 on poly_ints rather than trees.
17015 (vect_prune_runtime_alias_test_list): Update call accordingly.
17017 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17018 Alan Hayward <alan.hayward@arm.com>
17019 David Sherwood <david.sherwood@arm.com>
17021 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
17023 (vect_schedule_slp_instance): Likewise.
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 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
17030 constant and extern definitions for variable-length vectors.
17031 (vect_get_constant_vectors): Note that the number of units
17032 is known to be constant.
17034 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17035 Alan Hayward <alan.hayward@arm.com>
17036 David Sherwood <david.sherwood@arm.com>
17038 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
17039 of units as polynomial. Choose between WIDE and NARROW based
17042 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17043 Alan Hayward <alan.hayward@arm.com>
17044 David Sherwood <david.sherwood@arm.com>
17046 * tree-vect-stmts.c (simd_clone_subparts): New function.
17047 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
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-stmts.c (vectorizable_call): Treat the number of
17054 vectors as polynomial. Use build_index_vector for
17055 IFN_GOMP_SIMD_LANE.
17057 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17058 Alan Hayward <alan.hayward@arm.com>
17059 David Sherwood <david.sherwood@arm.com>
17061 * tree-vect-stmts.c (get_load_store_type): Treat the number of
17062 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
17063 for variable-length vectors.
17064 (vectorizable_mask_load_store): Treat the number of units as
17065 polynomial, asserting that it is constant if the condition has
17066 already been enforced.
17067 (vectorizable_store, vectorizable_load): Likewise.
17069 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17070 Alan Hayward <alan.hayward@arm.com>
17071 David Sherwood <david.sherwood@arm.com>
17073 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
17074 of units as polynomial. Punt if we can't tell at compile time
17075 which vector contains the final result.
17077 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17078 Alan Hayward <alan.hayward@arm.com>
17079 David Sherwood <david.sherwood@arm.com>
17081 * tree-vect-loop.c (vectorizable_induction): Treat the number
17082 of units as polynomial. Punt on SLP inductions. Use an integer
17083 VEC_SERIES_EXPR for variable-length integer reductions. Use a
17084 cast of such a series for variable-length floating-point
17087 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17088 Alan Hayward <alan.hayward@arm.com>
17089 David Sherwood <david.sherwood@arm.com>
17091 * tree.h (build_index_vector): Declare.
17092 * tree.c (build_index_vector): New function.
17093 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
17094 of units as polynomial, forcibly converting it to a constant if
17095 vectorizable_reduction has already enforced the condition.
17096 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
17097 to create a {1,2,3,...} vector.
17098 (vectorizable_reduction): Treat the number of units as polynomial.
17099 Choose vectype_in based on the largest scalar element size rather
17100 than the smallest number of units. Enforce the restrictions
17103 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17104 Alan Hayward <alan.hayward@arm.com>
17105 David Sherwood <david.sherwood@arm.com>
17107 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
17108 number of units as polynomial.
17110 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17111 Alan Hayward <alan.hayward@arm.com>
17112 David Sherwood <david.sherwood@arm.com>
17114 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
17115 * target.def (autovectorize_vector_sizes): Return the vector sizes
17116 by pointer, using vector_sizes rather than a bitmask.
17117 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
17118 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
17119 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
17121 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
17122 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
17123 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
17124 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
17125 * omp-general.c (omp_max_vf): Likewise.
17126 * omp-low.c (omp_clause_aligned_alignment): Likewise.
17127 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
17128 * tree-vect-loop.c (vect_analyze_loop): Likewise.
17129 * tree-vect-slp.c (vect_slp_bb): Likewise.
17130 * doc/tm.texi: Regenerate.
17131 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
17133 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
17134 the vector size as a poly_uint64 rather than an unsigned int.
17135 (current_vector_size): Change from an unsigned int to a poly_uint64.
17136 (get_vectype_for_scalar_type): Update accordingly.
17137 * tree.h (build_truth_vector_type): Take the size and number of
17138 units as a poly_uint64 rather than an unsigned int.
17139 (build_vector_type): Add a temporary overload that takes
17140 the number of units as a poly_uint64 rather than an unsigned int.
17141 * tree.c (make_vector_type): Likewise.
17142 (build_truth_vector_type): Take the number of units as a poly_uint64
17143 rather than an unsigned int.
17145 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17146 Alan Hayward <alan.hayward@arm.com>
17147 David Sherwood <david.sherwood@arm.com>
17149 * target.def (get_mask_mode): Take the number of units and length
17150 as poly_uint64s rather than unsigned ints.
17151 * targhooks.h (default_get_mask_mode): Update accordingly.
17152 * targhooks.c (default_get_mask_mode): Likewise.
17153 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
17154 * doc/tm.texi: Regenerate.
17156 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17157 Alan Hayward <alan.hayward@arm.com>
17158 David Sherwood <david.sherwood@arm.com>
17160 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
17161 * omp-general.c (omp_max_vf): Likewise.
17162 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
17163 (expand_omp_simd): Handle polynomial safelen.
17164 * omp-low.c (omplow_simd_context): Add a default constructor.
17165 (omplow_simd_context::max_vf): Change from int to poly_uint64.
17166 (lower_rec_simd_input_clauses): Update accordingly.
17167 (lower_rec_input_clauses): Likewise.
17169 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17170 Alan Hayward <alan.hayward@arm.com>
17171 David Sherwood <david.sherwood@arm.com>
17173 * tree-vectorizer.h (vect_nunits_for_cost): New function.
17174 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
17175 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
17176 (vect_analyze_slp_cost): Likewise.
17177 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
17178 (vect_model_load_cost): Likewise.
17180 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17181 Alan Hayward <alan.hayward@arm.com>
17182 David Sherwood <david.sherwood@arm.com>
17184 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
17185 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
17186 from an unsigned int * to a poly_uint64_pod *.
17187 (calculate_unrolling_factor): New function.
17188 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
17190 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17191 Alan Hayward <alan.hayward@arm.com>
17192 David Sherwood <david.sherwood@arm.com>
17194 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
17195 from an unsigned int to a poly_uint64.
17196 (_loop_vec_info::slp_unrolling_factor): Likewise.
17197 (_loop_vec_info::vectorization_factor): Change from an int
17199 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
17200 (vect_get_num_vectors): New function.
17201 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
17202 (vect_get_num_copies): Use vect_get_num_vectors.
17203 (vect_analyze_data_ref_dependences): Change max_vf from an int *
17204 to an unsigned int *.
17205 (vect_analyze_data_refs): Change min_vf from an int * to a
17207 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
17208 than an unsigned HOST_WIDE_INT.
17209 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
17210 (vect_analyze_data_ref_dependence): Change max_vf from an int *
17211 to an unsigned int *.
17212 (vect_analyze_data_ref_dependences): Likewise.
17213 (vect_compute_data_ref_alignment): Handle polynomial vf.
17214 (vect_enhance_data_refs_alignment): Likewise.
17215 (vect_prune_runtime_alias_test_list): Likewise.
17216 (vect_shift_permute_load_chain): Likewise.
17217 (vect_supportable_dr_alignment): Likewise.
17218 (dependence_distance_ge_vf): Take the vectorization factor as a
17219 poly_uint64 rather than an unsigned HOST_WIDE_INT.
17220 (vect_analyze_data_refs): Change min_vf from an int * to a
17222 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
17223 vfm1 as a poly_uint64 rather than an int. Make the same change
17224 for the returned bound_scalar.
17225 (vect_gen_vector_loop_niters): Handle polynomial vf.
17226 (vect_do_peeling): Likewise. Update call to
17227 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
17228 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
17230 * tree-vect-loop.c (vect_determine_vectorization_factor)
17231 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
17232 (vect_get_known_peeling_cost): Likewise.
17233 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
17234 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
17235 (vect_transform_loop): Likewise. Use the lowest possible VF when
17236 updating the upper bounds of the loop.
17237 (vect_min_worthwhile_factor): Make static. Return an unsigned int
17238 rather than an int.
17239 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
17240 polynomial unroll factors.
17241 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
17242 (vect_make_slp_decision): Likewise.
17243 (vect_supported_load_permutation_p): Likewise, and polynomial
17245 (vect_analyze_slp_cost): Handle polynomial vf.
17246 (vect_slp_analyze_node_operations): Likewise.
17247 (vect_slp_analyze_bb_1): Likewise.
17248 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
17249 than an unsigned HOST_WIDE_INT.
17250 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
17251 (vectorizable_load): Handle polynomial vf.
17252 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
17254 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
17256 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17257 Alan Hayward <alan.hayward@arm.com>
17258 David Sherwood <david.sherwood@arm.com>
17260 * match.pd: Handle bit operations involving three constants
17261 and try to fold one pair.
17263 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17265 * tree-vect-loop-manip.c: Include gimple-fold.h.
17266 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
17267 niters_maybe_zero parameters. Handle other cases besides a step of 1.
17268 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
17269 Add a path that uses a step of VF instead of 1, but disable it
17271 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
17272 and niters_no_overflow parameters. Update calls to
17273 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
17274 Create a new SSA name if the latter choses to use a ste other
17275 than zero, and return it via niters_vector_mult_vf_var.
17276 * tree-vect-loop.c (vect_transform_loop): Update calls to
17277 vect_do_peeling, vect_gen_vector_loop_niters and
17278 slpeel_make_loop_iterate_ntimes.
17279 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
17280 (vect_gen_vector_loop_niters): Update declarations after above changes.
17282 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
17284 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
17285 128-bit round to integer instructions.
17286 (ceil<mode>2): Likewise.
17287 (btrunc<mode>2): Likewise.
17288 (round<mode>2): Likewise.
17290 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
17292 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
17293 unaligned VSX load/store on P8/P9.
17294 (expand_block_clear): Allow the use of unaligned VSX
17295 load/store on P8/P9.
17297 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
17299 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
17301 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
17302 swap associated with both a load and a store.
17304 2018-01-02 Andrew Waterman <andrew@sifive.com>
17306 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
17307 * config/riscv/riscv.md (clear_cache): Use it.
17309 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
17311 * web.c: Remove out-of-date comment.
17313 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17315 * expr.c (fixup_args_size_notes): Check that any existing
17316 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
17317 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
17318 (emit_single_push_insn): ...here.
17320 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17322 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
17323 (const_vector_encoded_nelts): New function.
17324 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
17325 (const_vector_int_elt, const_vector_elt): Declare.
17326 * emit-rtl.c (const_vector_int_elt_1): New function.
17327 (const_vector_elt): Likewise.
17328 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
17329 of CONST_VECTOR_ELT.
17331 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17333 * expr.c: Include rtx-vector-builder.h.
17334 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
17335 directly on the tree encoding.
17336 (const_vector_from_tree): Likewise.
17337 * optabs.c: Include rtx-vector-builder.h.
17338 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
17339 sequence of "u" values.
17340 * vec-perm-indices.c: Include rtx-vector-builder.h.
17341 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
17342 directly on the vec_perm_indices encoding.
17344 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17346 * doc/rtl.texi (const_vector): Describe new encoding scheme.
17347 * Makefile.in (OBJS): Add rtx-vector-builder.o.
17348 * rtx-vector-builder.h: New file.
17349 * rtx-vector-builder.c: Likewise.
17350 * rtl.h (rtx_def::u2): Add a const_vector field.
17351 (CONST_VECTOR_NPATTERNS): New macro.
17352 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
17353 (CONST_VECTOR_DUPLICATE_P): Likewise.
17354 (CONST_VECTOR_STEPPED_P): Likewise.
17355 (CONST_VECTOR_ENCODED_ELT): Likewise.
17356 (const_vec_duplicate_p): Check for a duplicated vector encoding.
17357 (unwrap_const_vec_duplicate): Likewise.
17358 (const_vec_series_p): Check for a non-duplicated vector encoding.
17359 Say that the function only returns true for integer vectors.
17360 * emit-rtl.c: Include rtx-vector-builder.h.
17361 (gen_const_vec_duplicate_1): Delete.
17362 (gen_const_vector): Call gen_const_vec_duplicate instead of
17363 gen_const_vec_duplicate_1.
17364 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
17365 (gen_const_vec_duplicate): Use rtx_vector_builder.
17366 (gen_const_vec_series): Likewise.
17367 (gen_rtx_CONST_VECTOR): Likewise.
17368 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
17369 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
17370 Build a new vector rather than modifying a CONST_VECTOR in-place.
17371 (handle_special_swappables): Update call accordingly.
17372 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
17373 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
17374 Build a new vector rather than modifying a CONST_VECTOR in-place.
17375 (handle_special_swappables): Update call accordingly.
17377 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17379 * simplify-rtx.c (simplify_const_binary_operation): Use
17380 CONST_VECTOR_ELT instead of XVECEXP.
17382 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17384 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
17385 the selector elements to be different from the data elements
17386 if the selector is a VECTOR_CST.
17387 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
17388 ssizetype for the selector.
17390 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17392 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
17393 before testing each element individually.
17394 * tree-vect-generic.c (lower_vec_perm): Likewise.
17396 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17398 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
17399 * selftest-run-tests.c (selftest::run_tests): Call it.
17400 * vector-builder.h (vector_builder::operator ==): New function.
17401 (vector_builder::operator !=): Likewise.
17402 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
17403 (vec_perm_indices::all_from_input_p): New function.
17404 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
17405 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
17406 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
17407 instead of reading the VECTOR_CST directly. Detect whether both
17408 vector inputs are the same before constructing the vec_perm_indices,
17409 and update the number of inputs argument accordingly. Use the
17410 utility functions added above. Only construct sel2 if we need to.
17412 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17414 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
17415 the broadcast of the low byte.
17416 (expand_mult_highpart): Use an explicit encoding for the permutes.
17417 * optabs-query.c (can_mult_highpart_p): Likewise.
17418 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
17419 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
17420 (vectorizable_bswap): Likewise.
17421 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
17422 explicit encoding for the power-of-2 permutes.
17423 (vect_permute_store_chain): Likewise.
17424 (vect_grouped_load_supported): Likewise.
17425 (vect_permute_load_chain): Likewise.
17427 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17429 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
17430 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
17431 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
17432 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
17433 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
17434 (vect_gen_perm_mask_any): Likewise.
17436 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17438 * int-vector-builder.h: New file.
17439 * vec-perm-indices.h: Include int-vector-builder.h.
17440 (vec_perm_indices): Redefine as an int_vector_builder.
17441 (auto_vec_perm_indices): Delete.
17442 (vec_perm_builder): Redefine as a stand-alone class.
17443 (vec_perm_indices::vec_perm_indices): New function.
17444 (vec_perm_indices::clamp): Likewise.
17445 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
17446 (vec_perm_indices::new_vector): New function.
17447 (vec_perm_indices::new_expanded_vector): Update for new
17448 vec_perm_indices class.
17449 (vec_perm_indices::rotate_inputs): New function.
17450 (vec_perm_indices::all_in_range_p): Operate directly on the
17451 encoded form, without computing elided elements.
17452 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
17453 encoding. Update for new vec_perm_indices class.
17454 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
17455 the given vec_perm_builder.
17456 (expand_vec_perm_var): Update vec_perm_builder constructor.
17457 (expand_mult_highpart): Use vec_perm_builder instead of
17458 auto_vec_perm_indices.
17459 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
17460 vec_perm_indices instead of auto_vec_perm_indices. Use a single
17461 or double series encoding as appropriate.
17462 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
17463 vec_perm_indices instead of auto_vec_perm_indices.
17464 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
17465 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
17466 (vect_permute_store_chain): Likewise.
17467 (vect_grouped_load_supported): Likewise.
17468 (vect_permute_load_chain): Likewise.
17469 (vect_shift_permute_load_chain): Likewise.
17470 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
17471 (vect_transform_slp_perm_load): Likewise.
17472 (vect_schedule_slp_instance): Likewise.
17473 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
17474 (vectorizable_mask_load_store): Likewise.
17475 (vectorizable_bswap): Likewise.
17476 (vectorizable_store): Likewise.
17477 (vectorizable_load): Likewise.
17478 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
17479 vec_perm_indices instead of auto_vec_perm_indices. Use
17480 tree_to_vec_perm_builder to read the vector from a tree.
17481 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
17482 vec_perm_builder instead of a vec_perm_indices.
17483 (have_whole_vector_shift): Use vec_perm_builder and
17484 vec_perm_indices instead of auto_vec_perm_indices. Leave the
17485 truncation to calc_vec_perm_mask_for_shift.
17486 (vect_create_epilog_for_reduction): Likewise.
17487 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
17488 from auto_vec_perm_indices to vec_perm_indices.
17489 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
17490 instead of changing individual elements.
17491 (aarch64_vectorize_vec_perm_const): Use new_vector to install
17492 the vector in d.perm.
17493 * config/arm/arm.c (expand_vec_perm_d::perm): Change
17494 from auto_vec_perm_indices to vec_perm_indices.
17495 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
17496 instead of changing individual elements.
17497 (arm_vectorize_vec_perm_const): Use new_vector to install
17498 the vector in d.perm.
17499 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
17500 Update vec_perm_builder constructor.
17501 (rs6000_expand_interleave): Likewise.
17502 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
17503 (rs6000_expand_interleave): Likewise.
17505 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17507 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
17508 to qimode could truncate the indices.
17509 * optabs.c (expand_vec_perm_var): Likewise.
17511 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17513 * Makefile.in (OBJS): Add vec-perm-indices.o.
17514 * vec-perm-indices.h: New file.
17515 * vec-perm-indices.c: Likewise.
17516 * target.h (vec_perm_indices): Replace with a forward class
17518 (auto_vec_perm_indices): Move to vec-perm-indices.h.
17519 * optabs.h: Include vec-perm-indices.h.
17520 (expand_vec_perm): Delete.
17521 (selector_fits_mode_p, expand_vec_perm_var): Declare.
17522 (expand_vec_perm_const): Declare.
17523 * target.def (vec_perm_const_ok): Replace with...
17524 (vec_perm_const): ...this new hook.
17525 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
17526 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
17527 * doc/tm.texi: Regenerate.
17528 * optabs.def (vec_perm_const): Delete.
17529 * doc/md.texi (vec_perm_const): Likewise.
17530 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
17531 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
17532 expand_vec_perm for constant permutation vectors. Assert that
17533 the mode of variable permutation vectors is the integer equivalent
17534 of the mode that is being permuted.
17535 * optabs-query.h (selector_fits_mode_p): Declare.
17536 * optabs-query.c: Include vec-perm-indices.h.
17537 (selector_fits_mode_p): New function.
17538 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
17539 is defined, instead of checking whether the vec_perm_const_optab
17540 exists. Use targetm.vectorize.vec_perm_const instead of
17541 targetm.vectorize.vec_perm_const_ok. Check whether the indices
17542 fit in the vector mode before using a variable permute.
17543 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
17544 vec_perm_indices instead of an rtx.
17545 (expand_vec_perm): Replace with...
17546 (expand_vec_perm_const): ...this new function. Take the selector
17547 as a vec_perm_indices rather than an rtx. Also take the mode of
17548 the selector. Update call to shift_amt_for_vec_perm_mask.
17549 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
17550 Use vec_perm_indices::new_expanded_vector to expand the original
17551 selector into bytes. Check whether the indices fit in the vector
17552 mode before using a variable permute.
17553 (expand_vec_perm_var): Make global.
17554 (expand_mult_highpart): Use expand_vec_perm_const.
17555 * fold-const.c: Includes vec-perm-indices.h.
17556 * tree-ssa-forwprop.c: Likewise.
17557 * tree-vect-data-refs.c: Likewise.
17558 * tree-vect-generic.c: Likewise.
17559 * tree-vect-loop.c: Likewise.
17560 * tree-vect-slp.c: Likewise.
17561 * tree-vect-stmts.c: Likewise.
17562 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
17564 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
17565 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
17566 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
17567 (aarch64_vectorize_vec_perm_const): ...this new function.
17568 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
17569 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
17570 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
17571 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
17572 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
17573 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
17574 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
17576 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
17577 check for NEON modes.
17578 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
17579 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
17580 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
17581 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
17583 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
17584 the old VEC_PERM_CONST conditions.
17585 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
17586 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
17587 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
17588 (ia64_vectorize_vec_perm_const_ok): Merge into...
17589 (ia64_vectorize_vec_perm_const): ...this new function.
17590 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
17591 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
17592 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
17593 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
17594 * config/mips/mips.c (mips_expand_vec_perm_const)
17595 (mips_vectorize_vec_perm_const_ok): Merge into...
17596 (mips_vectorize_vec_perm_const): ...this new function.
17597 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
17598 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
17599 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
17600 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
17601 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
17602 (rs6000_expand_vec_perm_const): Delete.
17603 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
17605 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
17606 (altivec_expand_vec_perm_const_le): Take each operand individually.
17607 Operate on constant selectors rather than rtxes.
17608 (altivec_expand_vec_perm_const): Likewise. Update call to
17609 altivec_expand_vec_perm_const_le.
17610 (rs6000_expand_vec_perm_const): Delete.
17611 (rs6000_vectorize_vec_perm_const_ok): Delete.
17612 (rs6000_vectorize_vec_perm_const): New function.
17613 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
17614 an element count and rtx array.
17615 (rs6000_expand_extract_even): Update call accordingly.
17616 (rs6000_expand_interleave): Likewise.
17617 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
17618 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
17619 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
17620 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
17621 (rs6000_expand_vec_perm_const): Delete.
17622 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
17623 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
17624 (altivec_expand_vec_perm_const_le): Take each operand individually.
17625 Operate on constant selectors rather than rtxes.
17626 (altivec_expand_vec_perm_const): Likewise. Update call to
17627 altivec_expand_vec_perm_const_le.
17628 (rs6000_expand_vec_perm_const): Delete.
17629 (rs6000_vectorize_vec_perm_const_ok): Delete.
17630 (rs6000_vectorize_vec_perm_const): New function. Remove stray
17631 reference to the SPE evmerge intructions.
17632 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
17633 an element count and rtx array.
17634 (rs6000_expand_extract_even): Update call accordingly.
17635 (rs6000_expand_interleave): Likewise.
17636 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
17637 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
17639 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
17641 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17643 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
17644 vector mode and that that mode matches the mode of the data
17646 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
17647 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
17648 directly using expand_vec_perm_1 when forcing selectors into
17650 (expand_vec_perm_var): New function, split out from expand_vec_perm.
17652 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17654 * optabs-query.h (can_vec_perm_p): Delete.
17655 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
17656 * optabs-query.c (can_vec_perm_p): Split into...
17657 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
17658 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
17659 particular selector is valid.
17660 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
17661 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
17662 (vect_grouped_load_supported): Likewise.
17663 (vect_shift_permute_load_chain): Likewise.
17664 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
17665 (vect_transform_slp_perm_load): Likewise.
17666 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
17667 (vectorizable_bswap): Likewise.
17668 (vect_gen_perm_mask_checked): Likewise.
17669 * fold-const.c (fold_ternary_loc): Likewise. Don't take
17670 implementations of variable permutation vectors into account
17671 when deciding which selector to use.
17672 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
17673 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
17674 with a false third argument.
17675 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
17676 to test whether the constant selector is valid and can_vec_perm_var_p
17677 to test whether a variable selector is valid.
17679 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17681 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
17682 * optabs-query.c (can_vec_perm_p): Likewise.
17683 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
17684 instead of vec_perm_indices.
17685 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
17686 (vect_gen_perm_mask_checked): Likewise,
17687 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
17688 (vect_gen_perm_mask_checked): Likewise,
17690 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17692 * optabs-query.h (qimode_for_vec_perm): Declare.
17693 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
17694 (qimode_for_vec_perm): ...this new function.
17695 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
17697 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
17699 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
17700 does not have a conditional at the top.
17702 2018-01-02 Richard Biener <rguenther@suse.de>
17704 * ipa-inline.c (big_speedup_p): Fix expression.
17706 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
17709 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
17712 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
17716 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
17717 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
17718 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
17719 cond_taken_branch_cost 3->4.
17721 2018-01-01 Jakub Jelinek <jakub@redhat.com>
17723 PR tree-optimization/83581
17724 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
17725 TODO_cleanup_cfg if any changes have been made.
17727 PR middle-end/83608
17728 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
17729 convert_modes if target mode has the right side, but different mode
17732 PR middle-end/83609
17733 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
17734 last argument when extracting from CONCAT. If either from_real or
17735 from_imag is NULL, use expansion through memory. If result is not
17736 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
17737 the parts directly to inner mode, if even that fails, use expansion
17740 PR middle-end/83623
17741 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
17742 check for bswap in mode rather than HImode and use that in expand_unop
17745 Copyright (C) 2018 Free Software Foundation, Inc.
17747 Copying and distribution of this file, with or without modification,
17748 are permitted in any medium without royalty provided the copyright
17749 notice and this notice are preserved.