1 2018-06-25 Richard Biener <rguenther@suse.de>
3 PR tree-optimization/86304
4 * tree-vectorizer.c (vectorize_loops): Walk over new possibly
5 epilogue-if-converted loops as well.
7 2018-06-25 Jan Hubicka <hubicka@ucw.cz>
9 * lto-section-out.c (lto_begin_section): Do not print section
10 name for noaddr and unnumbered dumps.
12 2018-06-25 Richard Biener <rguenther@suse.de>
14 * tree-vectorizer.h (struct vec_info_shared): New structure
15 with parts split out from struct vec_info and loop_nest from
16 struct _loop_vec_info.
17 (struct vec_info): Adjust accordingly.
18 (struct _loop_vec_info): Likewise.
19 (LOOP_VINFO_LOOP_NEST): Adjust.
20 (LOOP_VINFO_DATAREFS): Likewise.
21 (LOOP_VINFO_DDRS): Likewise.
22 (struct _bb_vec_info): Likewise.
23 (BB_VINFO_DATAREFS): Likewise.
24 (BB_VINFO_DDRS): Likewise.
25 (struct _stmt_vec_info): Add dr_aux member.
26 (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
27 (DR_MISALIGNMENT_UNINITIALIZED): New.
28 (set_dr_misalignment): Adjust.
29 (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
30 (vect_analyze_loop): Adjust prototype.
31 (vect_analyze_loop_form): Likewise.
32 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
33 Compute dependences lazily.
34 (vect_record_base_alignments): Use shared datarefs/ddrs.
35 (vect_verify_datarefs_alignment): Likewise.
36 (vect_analyze_data_refs_alignment): Likewise.
37 (vect_analyze_data_ref_accesses): Likewise.
38 (vect_analyze_data_refs): Likewise.
39 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
40 constructor parameter for shared part.
41 (vect_analyze_loop_form): Pass in shared part and adjust.
42 (vect_analyze_loop_2): Pass in storage for the number of
43 stmts. Move loop nest finding to the caller. Compute
45 (vect_analyze_loop): Pass in shared part.
46 (vect_transform_loop): Verify shared datarefs are unchanged.
47 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
48 constructor parameter for shared part.
49 (vect_slp_analyze_bb_1): Pass in shared part and adjust.
50 (vect_slp_bb): Verify shared datarefs are unchanged before
52 * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
54 (new_stmt_vec_info): Initialize DR_AUX misalignment to
55 DR_MISALIGNMENT_UNINITIALIZED.
56 * tree-vectorizer.c (vec_info::vec_info): Add constructor
57 parameter for shared part.
58 (vec_info::~vec_info): Adjust.
59 (vec_info_shared::vec_info_shared): New.
60 (vec_info_shared::~vec_info_shared): Likewise.
61 (vec_info_shared::save_datarefs): Likewise.
62 (vec_info_shared::check_datarefs): Likewise.
63 (try_vectorize_loop_1): Construct shared part live for analyses
64 of a single loop for multiple vector sizes.
65 * tree-parloops.c (gather_scalar_reductions): Adjust.
67 2018-06-25 Richard Biener <rguenther@suse.de>
69 * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
70 DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
71 (vect_analyze_data_refs): Remove similar code from here and
74 2018-06-25 Richard Biener <rguenther@suse.de>
76 * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
77 for reverse storage order accesses rather than asserting
78 they cannot happen here.
80 2018-06-25 Tom de Vries <tdevries@suse.de>
83 * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
84 Use data16 instead of .byte for insn prefix.
86 2018-06-25 Andreas Krebbel <krebbel@linux.ibm.com>
89 * parser.c (make_char_string_pack): Pass this literal chars
90 through cpp_interpret_string.
91 (cp_parser_userdef_numeric_literal): Check the result of
92 make_char_string_pack.
94 2018-06-24 Maya Rashish <coypu@sdf.org>
96 * ginclude/stddef.h: Simplify conditions around avoiding
97 re-definition of __size_t.
99 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
101 * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
102 unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
104 2018-06-22 Maya Rashish <coypu@sdf.org>
106 * doc/invoke.texi (mno-fancy-math-387): Update for changes
107 made to OpenBSD and NetBSD through the years.
109 2018-06-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
111 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
112 behavior of vec_pack (vector double, vector double) to match
113 behavior of vec_float2 (vector double, vector double).
115 2018-06-22 Olivier Hainque <hainque@adacore.com>
117 * gimplify.c (gimplify_function_tree): Prevent creation
118 of a trampoline for the address of the current function
119 passed to entry/exit instrumentation hooks.
121 2018-06-22 Aaron Sawdey <acsawdey@linux.ibm.com>
124 * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
127 2018-06-22 Martin Liska <mliska@suse.cz>
129 PR tree-optimization/86263
130 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
131 Bail out if is_enabled is false.
132 * tree-switch-conversion.h (jump_table_cluster::is_enabled):
134 (jump_table_cluster::is_enabled): New function.
136 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
138 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
139 BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
140 * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
141 (lto_input_ts_binfo_tree_pointers): Likewise.
142 * tree-streamer-out.c (streamer_write_tree_bitfields,
143 write_ts_binfo_tree_pointers): Likewise.
144 * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
146 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
148 * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
150 2018-06-22 Martin Liska <mliska@suse.cz>
152 * symbol-summary.h (get): Make it pure and inline move
153 functionality from ::get function.
154 (get): Remove and inline into ::get and ::get_create.
155 (get_create): Move code from ::get function.
157 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
160 * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
161 -x assembler-with-cpp.
163 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
165 * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
166 _FILE_OFFSET_BITS=64 for C++.
168 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
170 * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
171 conversion insn that shows up when pr85657-3.c is compiled using
172 IEEE 128-bit long double.
173 (neg<mode>2_internal): Use the correct mode to check whether the
174 mode is IBM extended.
175 * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
176 multiply and divide external functions from being created more
179 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
181 * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
183 (rtl_merge_blocks): Likewise. Do not emit a NOP if the location of
184 the edge can be forwarded.
185 (cfg_layout_merge_blocks): Likewise.
187 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
189 * except.c (finish_eh_generation): Commit edge insertions only after
190 the EH edges have been redirected from post-landing to landing pads.
192 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
194 * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
195 create_tmp_var_for to create the FRAME decl.
196 (finalize_nesting_tree_1): Do not unchain the FRAME decl.
198 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
200 * tree-inline.c (copy_edges_for_bb): Minor tweak.
201 (maybe_move_debug_stmts_to_successors): Also reset the locus of the
202 debug statement when resetting its value.
203 (expand_call_inline): Copy the locus of the call onto the assignment
204 of the return value, if any. Use local variable in more cases.
206 2018-06-21 Martin Liska <mliska@suse.cz>
208 * ipa-pure-const.c (propagate_nothrow): Use
209 funct_state_summaries->get.
210 (dump_malloc_lattice): Likewise.
211 (propagate_malloc): Likewise.
213 2018-06-21 Richard Biener <rguenther@suse.de>
215 * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
216 comment. Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
217 * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
219 * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
220 BLOCK_ABSTRACT_ORIGIN unconditionally.
222 2018-06-21 David Malcolm <dmalcolm@redhat.com>
224 * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
226 * ipa-reference.c (ipa_reference_c_finalize): Delete
227 ipa_ref_opt_sum_summaries and set it to NULL.
229 2018-06-21 Tom de Vries <tdevries@suse.de>
231 PR tree-optimization/85859
232 * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
233 test with comment from bb_no_side_effects_p.
235 2018-06-21 Richard Biener <rguenther@suse.de>
237 PR tree-optimization/86232
238 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
239 max for constant niter.
241 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
243 * config/aarch64/aarch64-simd.md
244 (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
246 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
248 * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
249 Make opernads of the unspec commutative.
251 2018-06-21 Richard Biener <rguenther@suse.de>
253 * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
254 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
255 Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
256 (vect_analyze_data_ref_dependence): Re-order checks to deal with
258 (vect_record_base_alignments): Do not record base alignment
259 for gathers or scatters.
260 (vect_compute_data_ref_alignment): Drop return value that is always
261 true. Bail out early for gathers or scatters.
262 (vect_enhance_data_refs_alignment): Bail out early for gathers
264 (vect_find_same_alignment_drs): Likewise.
265 (vect_analyze_data_refs_alignment): Remove dead code.
266 (vect_slp_analyze_and_verify_node_alignment): Likewise.
267 (vect_analyze_data_refs): For possible gathers or scatters do
268 not create an alternate DR, just check their possible validity
269 and mark them. Adjust DECL_NONALIASED handling to not rely
271 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
272 update inits of gathers or scatters.
273 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
274 Also copy gather/scatter flag to pattern vinfo.
276 2018-06-20 Kelvin Nilsen <kelvin@gcc.gnu.org>
278 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
279 behavior of vec_packsu (vector unsigned long long, vector unsigned
280 long long) to match behavior of vec_packs with same signature.
282 2018-06-20 Chung-Lin Tang <cltang@codesourcery.com>
283 Thomas Schwinge <thomas@codesourcery.com>
284 Cesar Philippidis <cesar@codesourcery.com>
286 * gimplify.c (gimplify_scan_omp_clauses): Add support for
287 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
288 (gimplify_adjust_omp_clauses): Likewise.
289 (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
290 support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
291 (gimplify_omp_target_update): Update handling of acc update and
293 * omp-low.c (install_var_field): Remove unused parameter
294 base_pointers_restrict.
295 (scan_sharing_clauses): Remove base_pointers_restrict parameter.
296 Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
298 (omp_target_base_pointers_restrict_p): Delete.
299 (scan_omp_target): Update call to scan_sharing_clauses.
300 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
302 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
303 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
304 (convert_local_omp_clauses): Likewise.
305 * tree-pretty-print.c (dump_omp_clause): Likewise.
306 * tree.c (omp_clause_num_ops): Add entries for OMP_CLAUSE_{IF_PRESENT,
308 (omp_clause_code_name): Likewise.
310 2018-06-20 Jakub Jelinek <jakub@redhat.com>
313 * var-tracking.c (use_narrower_mode_test): Check if shift amount can
316 PR tree-optimization/86231
317 * tree-vrp.c (union_ranges): For ( [ ) ] or ( )[ ] range and
318 anti-range don't overwrite *vr0min before using it to compute *vr0max.
320 2018-06-20 Tom de Vries <tdevries@suse.de>
322 PR tree-optimization/86097
323 * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
324 iv type if signedness of iv type is not the same as that of *nit.
326 2018-06-20 Jakub Jelinek <jakub@redhat.com>
328 * cfgrtl.c (rtl_verify_edges): Formatting fix. If bb->preds has any
329 EDGE_EH edges, verify they are all EDGE_EH.
331 2018-06-20 Maya Rashish <coypu@sdf.org>
333 * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
335 2018-06-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
337 * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
338 * config/aarch64/aarch64.c (xgene1_tunings): Add
339 AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
340 (aarch64_mode_valid_for_sched_fusion_p):
342 (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
343 * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
345 * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
347 (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
348 * config/aarch64/iterators.md (VQ2): New mode iterator.
350 2018-06-20 Martin Liska <mliska@suse.cz>
352 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
353 Change default ratio from 10 to 8.
355 2018-06-20 Martin Liska <mliska@suse.cz>
357 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
359 (bit_test_cluster::find_bit_tests): Likewise.
360 (switch_decision_tree::analyze_switch_statement): Find clusters.
361 * tree-switch-conversion.h (struct jump_table_cluster): Document
364 2018-06-20 Martin Liska <mliska@suse.cz>
366 * tree-switch-conversion.c (switch_conversion::collect):
367 Record m_uniq property.
368 (switch_conversion::expand): Bail out for special conditions.
369 (group_cluster::~group_cluster): New.
370 (group_cluster::group_cluster): Likewise.
371 (group_cluster::dump): Likewise.
372 (jump_table_cluster::emit): New.
373 (switch_decision_tree::fix_phi_operands_for_edges): New.
374 (struct case_node): Remove struct.
375 (jump_table_cluster::can_be_handled): New.
376 (case_values_threshold): Moved to header.
377 (reset_out_edges_aux): Likewise.
378 (jump_table_cluster::is_beneficial): New.
379 (bit_test_cluster::can_be_handled): Likewise.
380 (add_case_node): Remove.
381 (bit_test_cluster::is_beneficial): New.
382 (case_bit_test::cmp): New.
383 (bit_test_cluster::emit): New.
384 (expand_switch_as_decision_tree_p): Remove.
385 (bit_test_cluster::hoist_edge_and_branch_if_true): New.
386 (fix_phi_operands_for_edge): Likewise.
387 (switch_decision_tree::analyze_switch_statement): New.
388 (compute_cases_per_edge): Move ...
389 (switch_decision_tree::compute_cases_per_edge): ... here.
390 (try_switch_expansion): Likewise.
391 (switch_decision_tree::try_switch_expansion): Likewise.
392 (record_phi_operand_mapping): Likewise.
393 (switch_decision_tree::record_phi_operand_mapping): Likewise.
394 (emit_case_decision_tree): Likewise.
395 (switch_decision_tree::emit): Likewise.
396 (balance_case_nodes): Likewise.
397 (switch_decision_tree::balance_case_nodes): Likewise.
398 (dump_case_nodes): Likewise.
399 (switch_decision_tree::dump_case_nodes): Likewise.
400 (emit_jump): Likewise.
401 (switch_decision_tree::emit_jump): Likewise.
402 (emit_cmp_and_jump_insns): Likewise.
403 (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
404 (emit_case_nodes): Likewise.
405 (switch_decision_tree::emit_case_nodes): Likewise.
406 (conditional_probability): Remove.
407 * tree-switch-conversion.h (enum cluster_type): New.
409 (struct cluster): Likewise.
410 (cluster::cluster): Likewise.
411 (struct simple_cluster): Likewise.
412 (simple_cluster::simple_cluster): Likewise.
413 (struct group_cluster): Likewise.
414 (struct jump_table_cluster): Likewise.
415 (struct bit_test_cluster): Likewise.
416 (struct min_cluster_item): Likewise.
417 (struct case_tree_node): Likewise.
418 (case_tree_node::case_tree_node): Likewise.
419 (jump_table_cluster::case_values_threshold): Likewise.
420 (struct case_bit_test): Likewise.
421 (struct switch_decision_tree): Likewise.
422 (struct switch_conversion): Likewise.
423 (switch_decision_tree::reset_out_edges_aux): Likewise.
425 2018-06-20 Martin Liska <mliska@suse.cz>
427 * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
428 (hoist_edge_and_branch_if_true): Likewise.
429 (expand_switch_using_bit_tests_p): Likewise.
430 (struct case_bit_test): Likewise.
431 (case_bit_test_cmp): Likewise.
432 (emit_case_bit_tests): Likewise.
433 (switch_conversion::switch_conversion): New class.
434 (struct switch_conv_info): Remove old struct.
435 (collect_switch_conv_info): More to ...
436 (switch_conversion::collect): ... this.
437 (check_range): Likewise.
438 (switch_conversion::check_range): Likewise.
439 (check_all_empty_except_final): Likewise.
440 (switch_conversion::check_all_empty_except_final): Likewise.
441 (check_final_bb): Likewise.
442 (switch_conversion::check_final_bb): Likewise.
443 (create_temp_arrays): Likewise.
444 (switch_conversion::create_temp_arrays): Likewise.
445 (free_temp_arrays): Likewise.
446 (gather_default_values): Likewise.
447 (switch_conversion::gather_default_values): Likewise.
448 (build_constructors): Likewise.
449 (switch_conversion::build_constructors): Likewise.
450 (constructor_contains_same_values_p): Likewise.
451 (switch_conversion::contains_same_values_p): Likewise.
452 (array_value_type): Likewise.
453 (switch_conversion::array_value_type): Likewise.
454 (build_one_array): Likewise.
455 (switch_conversion::build_one_array): Likewise.
456 (build_arrays): Likewise.
457 (switch_conversion::build_arrays): Likewise.
458 (gen_def_assigns): Likewise.
459 (switch_conversion::gen_def_assigns): Likewise.
460 (prune_bbs): Likewise.
461 (switch_conversion::prune_bbs): Likewise.
462 (fix_phi_nodes): Likewise.
463 (switch_conversion::fix_phi_nodes): Likewise.
464 (gen_inbound_check): Likewise.
465 (switch_conversion::gen_inbound_check): Likewise.
466 (process_switch): Use the newly created class.
467 (switch_conversion::expand): New.
468 (switch_conversion::~switch_conversion): New.
469 * tree-switch-conversion.h: New file.
471 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
473 * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
474 tree-vect-patterns.c.
475 * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
476 (vect_recog_dot_prod_pattern): Use it. Remove the type_in argument.
477 (vect_recog_sad_pattern): Likewise.
478 (vect_recog_widen_sum_pattern): Likewise.
479 (vect_recog_pow_pattern): Likewise. Check for a null vectype.
480 (vect_recog_widen_shift_pattern): Remove the type_in argument.
481 (vect_recog_rotate_pattern): Likewise.
482 (vect_recog_mult_pattern): Likewise.
483 (vect_recog_vector_vector_shift_pattern): Likewise.
484 (vect_recog_divmod_pattern): Likewise.
485 (vect_recog_mixed_size_cond_pattern): Likewise.
486 (vect_recog_bool_pattern): Likewise.
487 (vect_recog_mask_conversion_pattern): Likewise.
488 (vect_try_gather_scatter_pattern): Likewise.
489 (vect_recog_widen_mult_pattern): Likewise. Check for a null vectype.
490 (vect_recog_over_widening_pattern): Likewise.
491 (vect_recog_gather_scatter_pattern): Likewise.
492 (vect_recog_func_ptr): Move from tree-vectorizer.h
493 (vect_vect_recog_func_ptrs): Move further down the file.
494 (vect_recog_func): Likewise. Remove the third argument.
495 (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
496 (vect_pattern_recog_1): Expect the pattern function to do any
497 necessary target tests. Also expect it to provide a vector type.
498 Remove the type_in handling.
500 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
502 * tree-vect-patterns.c (vect_pattern_detected): New function.
503 (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
504 (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
505 (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
506 (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
507 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
508 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
509 (vect_recog_mask_conversion_pattern)
510 (vect_try_gather_scatter_pattern): Likewise.
512 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
514 * tree-vect-patterns.c (vect_get_internal_def): New function.
515 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
516 (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
517 (search_type_for_mask_1): Use it.
519 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
521 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
522 redundant WIDEN_SUM_EXPR handling.
523 (vect_recog_sad_pattern): Likewise.
525 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
527 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
528 redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
529 (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
530 (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
531 (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
533 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
535 * tree-vect-stmts.c (vectorizable_call): Make sure that we
536 use the stmt_vec_info of the original bb statement for the
537 new zero assignment, even if the call is part of a pattern.
539 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
541 * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
542 that the sequence is attached to the original statement rather
543 than the pattern statement.
544 * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
545 PATTERN_DEF_SEQ from the original statement rather than
546 the main pattern statement.
547 * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
548 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
549 (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
551 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
553 * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
554 definition statements before the early exit for statements that aren't
556 * tree-vect-loop.c (vect_transform_loop_stmt): New function,
558 (vect_transform_loop): ...here. Process pattern definition
559 statements without first checking whether the main pattern
560 statement is live or relevant.
562 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
564 * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
565 -O0 if the locus represent UNKNOWN_LOCATION but have different values.
567 2018-06-19 Aaron Sawdey <acsawdey@linux.ibm.com>
569 * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
570 TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
571 (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
572 (expand_block_compare): Change select_block_compare_mode call.
573 (expand_strncmp_align_check): Use new functions, fix comment.
574 (emit_final_str_compare_gpr): New function.
575 (expand_strn_compare): Refactor and clean up code.
576 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
578 2018-06-19 Tony Reix <tony.reix@atos.com>
579 Damien Bergamini <damien.bergamini@atos.com>
580 David Edelsohn <dje.gcc@gmail.com>
582 * collect2.c (static_obj): New variable.
583 (static_libs): New variable.
584 (is_in_list): Uncomment declaration.
585 (main): Track AIX libraries linked statically.
586 (is_in_list): Uncomment definition.
587 (scan_prog_file): Don't add AIX shared libraries initializer
588 to constructor list if linking statically.
590 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
592 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
594 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
596 2018-06-19 Jan Hubicka <hubicka@ucw.cz>
598 * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
601 2018-06-19 Martin Liska <mliska@suse.cz>
603 * config/i386/i386.c (ix86_can_inline_p): Do not use
604 ipa_fn_summaries::get_create.
605 * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
607 (devirtualization_time_bonus): Likewise.
608 (ipcp_propagate_stage): Likewise.
609 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
610 (edge_set_predicate): Likewise.
611 (evaluate_conditions_for_known_args): Likewise.
612 (evaluate_properties_for_edge): Likewise.
613 (ipa_call_summary::reset): Tranform to ...
614 (ipa_call_summary::~ipa_call_summary): ... this.
615 (ipa_fn_summary::reset): Transform to ...
616 (ipa_fn_summary::~ipa_fn_summary): ... this.
617 (ipa_fn_summary_t::remove): Rename to ...
618 (ipa_fn_summary_t::remove_callees): ... this.
619 (ipa_fn_summary_t::duplicate): Use placement new
620 instead of memory copy.
621 (ipa_call_summary_t::duplicate): Likewise.
622 (ipa_call_summary_t::remove): Remove.
623 (dump_ipa_call_summary): Change get_create to get.
624 (ipa_dump_fn_summary): Dump only when summary exists.
625 (analyze_function_body): Use symbol_summary::get instead
627 (compute_fn_summary): Likewise.
628 (estimate_edge_devirt_benefit): Likewise.
629 (estimate_edge_size_and_time): Likewise.
630 (inline_update_callee_summaries): Likewise.
631 (remap_edge_change_prob): Likewise.
632 (remap_edge_summaries): Likewise.
633 (ipa_merge_fn_summary_after_inlining): Likewise.
634 (write_ipa_call_summary): Likewise.
635 (ipa_fn_summary_write): Likewise.
636 (ipa_free_fn_summary): Likewise.
637 * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
638 (struct ipa_call_summary): Likewise.
639 * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
641 * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
642 (estimate_size_after_inlining): Likewise.
643 (estimate_growth): Likewise.
644 (growth_likely_positive): Likewise.
645 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
646 (inline_call): Likewise.
647 * ipa-inline.c (caller_growth_limits): Likewise.
648 (can_inline_edge_p): Likewise.
649 (can_inline_edge_by_limits_p): Likewise.
650 (compute_uninlined_call_time): Likewise.
651 (compute_inlined_call_time): Likewise.
652 (want_inline_small_function_p): Likewise.
653 (edge_badness): Likewise.
654 (update_caller_keys): Likewise.
655 (update_callee_keys): Likewise.
656 (inline_small_functions): Likewise.
657 (inline_to_all_callers_1): Likewise.
658 (dump_overall_stats): Likewise.
659 (early_inline_small_functions): Likewise.
660 (early_inliner): Likewise.
661 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
662 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
663 * ipa-pure-const.c (malloc_candidate_p): Likewise.
664 * ipa-split.c (execute_split_functions): Likewise.
665 * symbol-summary.h: Likewise.
666 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
668 2018-06-19 Richard Biener <rguenther@suse.de>
670 * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
671 (vectorize_loops): ... here. Fix dbgcnt handling.
672 (try_vectorize_loop): Wrap try_vectorize_loop_1.
674 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
677 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
678 ieee128 argument takes up only one (vector) register, not two (floating
681 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
683 * gimplify.c (gimplify_init_constructor): Really never clear for an
684 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
686 2018-06-19 Richard Biener <rguenther@suse.de>
688 PR tree-optimization/86179
689 * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
690 after failed recognition.
692 2018-06-18 Martin Sebor <msebor@redhat.com>
695 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
696 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
697 Handle integer subtraction.
698 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
699 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
701 2018-06-18 David Malcolm <dmalcolm@redhat.com>
703 * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
704 param from rtx to rtx_insn *.
705 * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
707 (frv_ifcvt_modify_insn): Likwise.
708 (frv_ifcvt_modify_final): Likwise for local "existing_insn",
709 adding an as_a <rtx_insn *> cast. Likewise for local "insn".
710 * config/mips/mips.c (r10k_insert_cache_barriers): Add an
711 as_a <rtx_insn *> cast to local "unprotected_region" once
712 it's been established that it's not NULL or pc_rtx.
713 * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
714 param "sethi" from rtx to rtx_insn *.
715 (nds32_group_float_insns): Likewise for param "insn".
716 * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
718 (vax_output_int_subtract): Likewise.
719 * config/vax/vax.c (vax_output_int_add): Likewise for param
721 (vax_output_int_subtract): Likewise.
722 * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
723 (emit_pattern_after): Likewise for param "after".
724 (emit_insn_after): Likewise.
725 (emit_jump_insn_after): Likewise.
726 (emit_call_insn_after): Likewise.
727 (emit_debug_insn_after): Likewise.
728 (emit_pattern_before): Likewise for param "before".
729 (emit_insn_before): Likewise.
730 (emit_jump_insn_before): Likewise.
731 * final.c (get_insn_template): Likewise for param "insn", removing
733 * output.h (get_insn_template): Likewise for 2nd param.
734 * rtl.h (emit_insn_before): Likewise.
735 (emit_jump_insn_before): Likewise.
736 (emit_debug_insn_before_noloc): Likewise.
737 (emit_insn_after): Likewise.
738 (emit_jump_insn_after): Likewise.
739 (emit_call_insn_after): Likewise.
740 (emit_debug_insn_after): Likewise.
741 (set_insn_deleted): Likewise for param.
743 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
746 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
747 floating point modes, so that IFmode is numerically greater than
748 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
749 to declare the ordering. This prevents IFmode from being
750 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
751 machine. Include rs6000-modes.h to share the fractional values
752 between genmodes* and the rest of the compiler.
756 * config/rs6000/rs6000-modes.h: New file.
757 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
758 meaning of rs6000_long_double_size so that 126..128 selects an
759 appropriate 128-bit floating point type.
760 (rs6000_option_override_internal): Likewise.
761 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
762 (TARGET_LONG_DOUBLE_128): Change the meaning of
763 rs6000_long_double_size so that 126..128 selects an appropriate
764 128-bit floating point type.
765 (LONG_DOUBLE_TYPE_SIZE): Update comment.
766 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
767 source and destination to match the standard usage.
768 (truncifkf2): Likewise.
769 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
770 ISA 2.07 to use an explicit clobber, instead of passing in a
772 (copysign<mode>3_soft): Likewise.
774 2018-06-18 David Malcolm <dmalcolm@redhat.com>
776 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
777 Replace dump_printf_loc call with DUMP_VECT_SCOPE.
778 (vect_slp_analyze_instance_dependence): Likewise.
779 (vect_enhance_data_refs_alignment): Likewise.
780 (vect_analyze_data_refs_alignment): Likewise.
781 (vect_slp_analyze_and_verify_instance_alignment
782 (vect_analyze_data_ref_accesses): Likewise.
783 (vect_prune_runtime_alias_test_list): Likewise.
784 (vect_analyze_data_refs): Likewise.
785 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
786 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
787 (vect_analyze_scalar_cycles_1): Likewise.
788 (vect_get_loop_niters): Likewise.
789 (vect_analyze_loop_form_1): Likewise.
790 (vect_update_vf_for_slp): Likewise.
791 (vect_analyze_loop_operations): Likewise.
792 (vect_analyze_loop): Likewise.
793 (vectorizable_induction): Likewise.
794 (vect_transform_loop): Likewise.
795 * tree-vect-patterns.c (vect_pattern_recog): Likewise.
796 * tree-vect-slp.c (vect_analyze_slp): Likewise.
797 (vect_make_slp_decision): Likewise.
798 (vect_detect_hybrid_slp): Likewise.
799 (vect_slp_analyze_operations): Likewise.
800 (vect_slp_bb): Likewise.
801 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
802 (vectorizable_bswap): Likewise.
803 (vectorizable_call): Likewise.
804 (vectorizable_simd_clone_call): Likewise.
805 (vectorizable_conversion): Likewise.
806 (vectorizable_assignment): Likewise.
807 (vectorizable_shift): Likewise.
808 (vectorizable_operation): Likewise.
809 * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
811 2018-06-18 Martin Sebor <msebor@redhat.com>
813 PR tree-optimization/81384
814 * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
815 * builtins.c (expand_builtin_strnlen): New function.
816 (expand_builtin): Call it.
817 (fold_builtin_n): Avoid setting TREE_NO_WARNING.
818 * builtins.def (BUILT_IN_STRNLEN): New.
819 * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
820 Warn for bounds in excess of maximum object size.
821 * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
822 single-value ranges. Handle strnlen.
823 (handle_builtin_strlen): Handle strnlen.
824 (strlen_check_and_optimize_stmt): Same.
825 * doc/extend.texi (Other Builtins): Document strnlen.
827 2018-06-18 Maya Rashish <coypu@sdf.org>
829 * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
830 (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
831 (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
833 * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
835 * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
837 2018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
839 * tree.c (escaped_string::escape): Replace cast to char * by
840 const_cast<char *> (unescaped).
842 2018-06-18 Nick Clifton <nickc@redhat.com>
845 * tree.c (escaped_string): New class. Converts an unescaped
846 string into its escaped equivalent.
847 (warn_deprecated_use): Use the new class to convert the
848 deprecation message, if present.
849 (test_escaped_strings): New self test.
850 (test_c_tests): Add test_escaped_strings.
851 * doc/extend.texi (deprecated): Add a note that the
852 deprecation message is affected by the -fmessage-length
853 option, and that control characters will be escaped.
854 (#pragma GCC error): Document this pragma.
855 (#pragma GCC warning): Likewise.
856 * doc/invoke.texi (-fmessage-length): Document this option's
857 effect on the #warning and #error preprocessor directives and
858 the deprecated attribute.
860 2018-06-18 Eric Botcazou <ebotcazou@adacore.com>
862 * tree.c (decl_value_expr_lookup): Revert latest change.
863 (decl_value_expr_insert): Likewise.
865 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
867 * gimplify.c (nonlocal_vlas): Delete.
868 (nonlocal_vla_vars): Likewise.
869 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
871 (gimplify_body): Do not create and destroy nonlocal_vlas.
872 * tree-nested.c: Include diagnostic.h.
873 (use_pointer_in_frame): Tweak.
874 (lookup_field_for_decl): Add assertion and declare the transformation.
875 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
876 internal error when the reference is in a wrong context. Do not
877 create a debug decl by default.
878 (note_nonlocal_block_vlas): Delete.
879 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
880 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
881 create a debug decl by default.
882 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
883 call is in a wrong context.
884 (fixup_vla_decls): New function.
885 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
886 debug variables were created.
887 * tree.c (decl_value_expr_lookup): Add checking assertion.
888 (decl_value_expr_insert): Likewise.
890 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
893 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
894 * tree-scalar-evolution.c (interpret_expr): Likewise.
895 (expression_expensive_p): Likewise.
896 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
897 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
898 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
899 (ssa_defined_by_minus_one_stmt_p): New.
901 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
904 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
905 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
906 * dojump.c (do_jump): Likewise.
907 * expr.c (expand_expr_real_2): Check operand type's sign.
908 * fold-const.c (const_unop): Handle ABSU_EXPR.
909 (fold_abs_const): Likewise.
910 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
911 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
912 (strip_sign_op_1): Likesise.
913 * match.pd: Add new pattern to generate ABSU_EXPR.
914 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
915 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
916 * tree-eh.c (operation_could_trap_helper_p): Likewise.
917 * tree-inline.c (estimate_operator_cost): Likewise.
918 * tree-pretty-print.c (dump_generic_node): Likewise.
919 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
920 * tree.def (ABSU_EXPR): New.
922 2018-06-16 Jakub Jelinek <jakub@redhat.com>
925 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
926 documented as preserved for backward compatibility only.
927 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
929 PR rtl-optimization/86108
930 * bb-reorder.c (create_forwarder_block): Renamed to ...
931 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
932 jump from new landing pad to the second part.
933 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
936 2018-06-15 Jakub Jelinek <jakub@redhat.com>
939 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
940 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
941 Only call store_expr for halves if the mode is the same.
944 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
945 Fix up comment formatting.
947 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
949 * typed-splay-tree.h (typed_splay_tree::remove): New function.
950 (typed_splay_tree::closure,
951 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
952 (typed_splay_tree::typed_splay_tree,
953 typed_splay_tree::operator =): Declared private.
954 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
955 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
956 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
957 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
958 typed_splay_tree::splay_tree_splay,
959 typed_splay_tree::splay_tree_foreach_helper,
960 typed_splay_tree::splay_tree_insert,
961 typed_splay_tree::splay_tree_remove,
962 typed_splay_tree::splay_tree_lookup,
963 typed_splay_tree::splay_tree_predecessor,
964 typed_splay_tree::splay_tree_successor,
965 typed_splay_tree::splay_tree_min,
966 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
967 (typed_splay_tree::root, typed_splay_tree::comp,
968 typed_splay_tree::delete_key,
969 typed_splay_tree::delete_value): New data members.
970 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
971 typed_splay_tree::remove.
973 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
975 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
976 -mginv and -mno-ginv to the assembler.
977 * config/mips/mips.opt (-mcrc): New option.
979 * doc/invoke.text (-mcrc): Document.
982 2018-06-15 Nick Clifton <nickc@redhat.com>
985 * tree.c (escaped_string): New class. Converts an unescaped
986 string into its escaped equivalent.
987 (warn_deprecated_use): Use the new class to convert the
988 deprecation message, if present.
989 (test_escaped_strings): New self test.
990 (test_c_tests): Add test_escaped_strings.
991 * doc/extend.texi (deprecated): Add a note that the
992 deprecation message is affected by the -fmessage-length
993 option, and that control characters will be escaped.
994 (#pragma GCC error): Document this pragma.
995 (#pragma GCC warning): Likewise.
996 * doc/invoke.texi (-fmessage-length): Document this option's
997 effect on the #warning and #error preprocessor directives and
998 the deprecated attribute.
1000 2018-06-15 Richard Biener <rguenther@suse.de>
1002 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
1003 here, also noting vector size used.
1004 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
1005 size used in MSG_OPTIMIZED_LOCATIONS dump.
1006 (pass_slp_vectorize::execute): Adjust.
1008 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
1011 * config/arc/arc.c (arc_return_address_register): Fix
1014 2018-06-15 Richard Biener <rguenther@suse.de>
1017 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
1018 leave useless conversion stripping to force_gimple_operand_gsi.
1019 (gimplify_build2): Likewise.
1020 (gimplify_build1): Likewise.
1022 2018-06-15 Richard Biener <rguenther@suse.de>
1025 * tree-cfg.c (move_stmt_op): unshare invariant addresses
1026 before adjusting their block.
1028 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
1030 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
1031 multilibs for *-*-rtems*.
1032 * config/riscv/t-rtems: New file.
1034 2018-06-14 Jakub Jelinek <jakub@redhat.com>
1037 * match.pd ((A +- CST1) +- CST2): Punt if last resort
1038 unsigned_type_for returns NULL.
1041 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
1042 subregs of multi-word pseudos unless the float mode has word size.
1044 2018-06-14 Richard Biener <rguenther@suse.de>
1047 * tree-vect-generic.c (build_word_mode_vector_type): Remove
1048 duplicate and harmful type_hash_canon.
1049 * tree.c (type_hash_canon): Assert we didn't find ourselves.
1051 2018-06-14 Richard Biener <rguenther@suse.de>
1054 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
1057 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
1059 * config/rtems.h (STDINT_LONG32): Define.
1061 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
1062 Prachi Godbole <prachi.godbole@imgtec.com>
1064 * config/mips/mips-cpus.def: Define P6600.
1065 * config/mips/mips-tables.opt: Regenerate.
1066 * config/mips/mips.c (mips_ucbranch_type): New enum.
1067 (mips_rtx_cost_data): Add support for P6600.
1068 (mips_issue_rate): Likewise.
1069 (mips_multipass_dfa_lookahead): Likewise.
1070 (mips_avoid_hazard): Likewise.
1071 (mips_reorg_process_insns): Likewise.
1072 (mips_classify_branch_p6600): New function.
1073 * config/mips/mips.h (TUNE_P6600): New define.
1074 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
1075 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
1076 * config/mips/mips.md: Include p6600.md.
1077 (processor): Add p6600.
1078 * config/mips/p6600.md: New file.
1079 * doc/invoke.texi: Add p6600 to supported architectures.
1081 2018-06-13 Martin Sebor <msebor@redhat.com>
1083 PR tree-optimization/86114
1084 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
1086 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
1088 2018-06-13 Richard Biener <rguenther@suse.de>
1090 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
1091 Properly set vector type of the intermediate stmt.
1092 * tree-vect-stmts.c (vectorizable_operation): The destination
1093 var always has vectype_out type.
1095 2018-06-13 Jeff Law <law@redhat.com>
1097 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
1098 integer 0 for argument to print_rtl_with_bb.
1099 (rl78_reorg): Likewise.
1101 2018-06-13 David Malcolm <dmalcolm@redhat.com>
1103 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
1104 from rtx to rtx_insn *.
1105 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
1107 (add_sched_insns_for_speculation): Likewise for local "target",
1108 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
1109 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
1110 from rtx_insn ** to rtx_code_label **.
1111 (reorg_emit_nops): Likewise.
1112 (c6x_reorg): Likewise for local "call_labels".
1113 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
1115 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
1116 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
1117 the loops over LABEL_REFS.
1118 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
1120 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
1121 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
1122 (split_branches): Strengthen local "olabel" from rtx to
1123 rtx_insn *, adding a safe_as_a cast.
1124 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
1126 (add_insn_after): Likewise for first two params.
1127 (add_insn_before): Likewise.
1128 (remove_insn): Likewise for param.
1129 (emit_pattern_before_noloc): Likewise for second and third params.
1130 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
1131 (emit_call_insn_before_noloc): Likewise.
1132 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
1134 (emit_barrier_before): Likewise.
1135 (emit_label_before): Strengthen "label" param from "rtx" to
1136 "rtx_code_label *". Strengthen "before" param from "rtx" to
1138 (emit_insn_after_1): Strengthen "after" param from "rtx" to
1140 (emit_pattern_after_noloc): Likewise.
1141 (emit_insn_after_noloc): Likewise.
1142 (emit_jump_insn_after_noloc): Likewise.
1143 (emit_call_insn_after_noloc): Likewise.
1144 (emit_debug_insn_after_noloc): Likewise.
1145 (emit_barrier_after): Likewise.
1146 (emit_label_after): Likewise for both params.
1147 (emit_pattern_after_setloc): Likewise for "after" param. Convert
1148 "loc" param from "int" to "location_t".
1149 (emit_insn_after_setloc): Likewise.
1150 (emit_jump_insn_after_setloc): Likewise.
1151 (emit_call_insn_after_setloc): Likewise.
1152 (emit_debug_insn_after_setloc): Likewise.
1153 (emit_pattern_before_setloc): Likewise for "before" param. Convert
1154 "loc" param from "int" to "location_t".
1155 (emit_pattern_before): Convert NULL_RTX to NULL.
1156 (emit_insn_before_setloc): Convert "loc" param from "int" to
1158 (emit_jump_insn_before_setloc): Likewise.
1159 (emit_call_insn_before_setloc): Likewise.
1160 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
1161 rtx_insn *. Convert "loc" param from "int" to "location_t".
1162 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
1163 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
1164 Convert 3rd param from "int" to "location_t".
1165 (emit_barrier_before, emit_barrier_after, next_real_insn):
1166 Strengthen param from rtx to rtx_insn *.
1167 (emit_label_before): Strengthen 1st param from "rtx" to
1168 "rtx_code_label *". Strengthen 2nd param from "rtx" to
1170 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
1171 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
1172 Strengthen 2nd param from "rtx" to "rtx_insn *".
1173 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
1174 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
1175 Likewise. Convert 3rd param from "int" to "location_t".
1176 (emit_label_after): Strengthen 1st param from "rtx" to
1178 (next_real_insn, remove_insn): Strengthen param from "rtx" to
1180 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
1181 from "rtx" to "rtx_insn *".
1183 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
1185 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
1186 bodies streamed in with -Q.
1187 * dumpfile.c (dump_files): Add lto-stream-out dump file.
1188 * dumpfile.h (tree_dump_index): Add lto_stream_out.
1189 * gimple-streamer-out.c: Include gimple-pretty-print.h
1190 (output_bb): Dump stmts streamed.
1191 * lto-section-out.c: Include print-tree.h
1192 (lto_begin_section): Dump sections created.
1193 (lto_output_decl_index): Dump decl encoded.
1194 * lto-streamer-out.c: Include print-tree.h
1195 (create_output_block): Dump output block created.
1196 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
1197 (output_function): Dump function output.
1198 (output_constructor): Dump constructor streamed.
1199 (write_global_stream): Output indexes encoded.
1200 (produce_asm_for_decls): Dump streams encoded.
1201 * lto-streamer.c (streamer_dump_file): New global var.
1202 * lto-streamer.h (streamer_dump_file): Declare.
1203 * passes.c (ipa_write_summaries): Initialize streamer dump.
1204 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
1207 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
1210 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
1211 offsets for register save directives. Emit a second batch of save
1212 directives, if need be, when the function accesses prior frames.
1214 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1216 * config/arc/fpu.md (fmasf4): Force operand to register.
1217 (fnmasf4): Likewise.
1219 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1221 * config/arc/arc-protos.h (arc_pad_return): Remove.
1222 * config/arc/arc.c (machine_function): Remove force_short_suffix
1224 (arc_print_operand): Adjust printing of '&'.
1225 (arc_verify_short): Remove conditional printing of short suffix.
1226 (arc_final_prescan_insn): Remove reference to size_reason.
1227 (pad_return): New function.
1228 (arc_reorg): Call pad_return.
1229 (arc_pad_return): Remove.
1230 (arc_init_machine_status): Remove reference to force_short_suffix.
1231 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
1232 (attr length): When attribute iscompact is true force to 2
1233 regardless; in the case of maybe check if we want to force the
1234 instruction to have 4 bytes length.
1235 (nopv): Change it to generate 4 byte long nop as well.
1236 (blockage): New pattern.
1237 (simple_return): Remove call to arc_pad_return.
1238 (p_return_i): Likewise.
1240 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1242 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
1244 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1246 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
1249 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1251 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
1252 for ARC700 and ARCv2.
1254 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
1257 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
1258 operands[2] instead of operands[1].
1261 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
1263 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
1264 case, check whether the outer register overlaps an unallocatable
1265 register, not just whether it fits the required class.
1267 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
1269 * poly-int.h (can_div_trunc_p): Add new overload in which all values
1271 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
1272 (memrefs_conflict_p): Likewise.
1273 (init_alias_analysis): Likewise.
1274 * cfgexpand.c (expand_debug_expr): Likewise.
1275 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
1276 * cse.c (fold_rtx): Likewise.
1277 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
1278 * expr.c (emit_block_move_hints): Likewise.
1279 (clear_storage_hints, push_block, emit_push_insn): Likewise.
1280 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
1281 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
1282 (emit_group_store): Likewise.
1283 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
1284 to read the PRE/POST_MODIFY increment.
1285 * calls.c (store_one_arg): Use strip_offset.
1286 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
1288 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
1290 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
1291 (simplify_binary_operation_1): Extend CONST_INT handling to
1293 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
1294 than a HOST_WIDE_INT.
1295 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
1297 (adjust_mems, add_stores): Update accodingly.
1298 (vt_canonicalize_addr): Track polynomial offsets.
1299 (emit_note_insn_var_location): Likewise.
1300 (vt_add_function_parameter): Likewise.
1301 (vt_initialize): Likewise.
1303 2018-06-12 Jeff Law <law@redhat.com>
1305 * config.gcc (alpha*-*-freebsd*): Remove.
1306 * config/alpha/freebsd.h: Remove.
1308 2018-06-12 David Malcolm <dmalcolm@redhat.com>
1311 * spellcheck-tree.c (levenshtein_distance): Rename to...
1312 (get_edit_distance): ...this, and update for underlying renaming.
1313 * spellcheck-tree.h (levenshtein_distance): Rename to...
1314 (get_edit_distance): ...this.
1315 * spellcheck.c (levenshtein_distance): Rename to...
1316 (get_edit_distance): ...this. Convert from Levenshtein distance
1317 to Damerau-Levenshtein distance by supporting transpositions of
1318 adjacent characters. Rename "v1" to "v_next" and "v0" to
1320 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
1321 (selftest::test_edit_distance_unit_test_oneway): ...this, and
1322 update for underlying renaming.
1323 (selftest::levenshtein_distance_unit_test): Rename to...
1324 (selftest::test_get_edit_distance_unit): ...this, and update for
1325 underlying renaming.
1326 (selftest::test_find_closest_string): Add example from PR 69968
1327 where transposition helps
1328 (selftest::test_metric_conditions): Update for renaming.
1329 (selftest::test_metric_conditions): Likewise.
1330 (selftest::spellcheck_c_tests): Likewise.
1331 * spellcheck.h (levenshtein_distance): Rename both overloads to...
1332 (get_edit_distance): ...this.
1333 (best_match::consider): Update for renaming.
1335 2018-06-12 Martin Sebor <msebor@redhat.com>
1337 PR tree-optimization/85259
1338 * builtins.c (compute_objsize): Handle constant offsets.
1339 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
1340 true iff a warning has been issued.
1341 * gimple.h (gimple_nonartificial_location): New function.
1342 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
1343 gimple_nonartificial_location and handle -Wno-system-headers.
1344 (handle_builtin_stxncpy): Same.
1346 2018-06-12 Martin Sebor <msebor@redhat.com>
1349 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
1351 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
1353 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
1354 BUILTIN_VEC_XST entries for pointer to double and long long.
1356 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
1359 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
1361 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
1364 2018-06-12 Martin Liska <mliska@suse.cz>
1366 * doc/options.texi: Document IntegerRange.
1368 2018-06-12 Martin Liska <mliska@suse.cz>
1370 * config/i386/i386.opt: Make MPX-related options as Deprecated.
1371 * opt-functions.awk: Handle Deprecated flag.
1372 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
1374 (read_cmdline_option): Report warning for a deprecated option.
1375 * opts.h (struct cl_option): Add new field cl_deprecated.
1376 (CL_ERR_DEPRECATED): New.
1378 2018-06-12 Martin Liska <mliska@suse.cz>
1380 * doc/options.texi: Document Deprecated option flag.
1382 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1384 * config/arc/arc-arch.h (arc_extras): New enum.
1385 (arc_cpu_t):Add field extra.
1386 (arc_cpu_types): Consider the extras.
1387 * config/arc/arc-cpus.def: Add extras info.
1388 * config/arc/arc-opts.h (processor_type): Consider extra field.
1389 * config/arc/arc.c (arc_override_options): Handle extra field.
1391 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1393 * config/arc/arc-arch.h: Update ARC_OPTX macro.
1394 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
1396 * config/arc/arc.c (arc_init): Update pic warning.
1397 (irq_range): Update irq range parsing warnings.
1398 (arc_override_options): Update various warning messages.
1399 (arc_handle_aux_attribute): Likewise.
1401 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
1403 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
1405 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
1407 * doc/sourcebuild.texi: Document usage of line number 0 in verify
1408 compiler messages directives.
1410 2018-06-12 Matthew Fortune <mfortune@gmail.com>
1412 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
1413 * config/mips/mips-tables.opt: Regenerate.
1414 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
1416 * doc/invoke.texi: Document -march=i6500.
1418 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
1420 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
1421 (i6400_gpmul): Add cpu_unit.
1422 (i6400_gpdiv): Likewise.
1423 (i6400_msa_add_d): Update reservations.
1424 (i6400_msa_int_add) Likewise.
1425 (i6400_msa_short_logic3) Likewise.
1426 (i6400_msa_short_logic2) Likewise.
1427 (i6400_msa_short_logic) Likewise.
1428 (i6400_msa_move) Likewise.
1429 (i6400_msa_cmp) Likewise.
1430 (i6400_msa_short_float2) Likewise.
1431 (i6400_msa_div_d) Likewise.
1432 (i6400_msa_long_logic1) Likewise.
1433 (i6400_msa_long_logic2) Likewise.
1434 (i6400_msa_mult) Likewise.
1435 (i6400_msa_long_float2) Likewise.
1436 (i6400_msa_long_float4) Likewise.
1437 (i6400_msa_long_float5) Likewise.
1438 (i6400_msa_long_float8) Likewise.
1439 (i6400_fpu_fadd): Include frint type.
1440 (i6400_fpu_store): New define_insn_reservation.
1441 (i6400_fpu_load): Likewise.
1442 (i6400_fpu_move): Likewise.
1443 (i6400_fpu_fcmp): Likewise.
1444 (i6400_fpu_fmadd): Likewise.
1445 (i6400_int_mult): Include imul3nc type and update reservation.
1446 (i6400_int_div): Include idiv3 type and update reservation.
1447 (i6400_int_load): Update to check type not move_type.
1448 (i6400_int_store): Likewise.
1449 (i6400_int_prefetch): Set zero latency.
1451 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
1453 * gcc.c: Document new %@{...} sequence.
1454 (LINK_COMMAND_SPEC): Use it for the -L switches.
1455 (cpp_unique_options): Use it for the -I switches.
1456 (at_file_argbuf): New global variable.
1457 (in_at_file): Likewise.
1458 (alloc_args): Create at_file_argbuf.
1459 (clear_args): Truncate at_file_argbuf.
1460 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
1461 (open_at_file): New function.
1462 (close_at_file): Likewise.
1463 (create_at_file): Delete.
1464 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
1467 (validate_switches_from_spec): Deal with %@{...} sequence.
1468 (validate_switches): Likewise.
1469 (driver::finalize): Call clear_args.
1471 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
1473 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
1475 2018-06-11 Martin Sebor <msebor@redhat.com>
1477 * doc/invoke.texi (-Wall): List -Wc++17-compat.
1478 (Wno-class-memaccess): Add @opindex.
1479 (Wno-templates, Wno-multiple-inheritance): Same.
1480 (Wno-virtual-inheritance, Wno-namespaces): Same.
1481 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
1482 (Wno-format-overflow, Wno-format-truncation): Same.
1483 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
1484 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
1485 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
1486 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
1487 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
1488 (Wno-misspelled-isr): Same.
1490 2018-06-11 Martin Sebor <msebor@redhat.com>
1492 * PR tree-optimization/86083
1493 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
1495 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
1497 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
1499 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
1502 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
1503 on the correct operand.
1504 (*movdi_internal64): Ditto.
1506 2018-06-11 Martin Liska <mliska@suse.cz>
1508 PR tree-optimization/86089
1509 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
1511 2018-06-11 Julia Koval <julia.koval@intel.com>
1513 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
1514 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
1515 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
1517 2018-06-11 Olivier Hainque <hainque@adacore.com>
1519 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
1520 for Ada with strict dwarf2.
1522 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
1525 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
1528 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
1530 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
1532 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
1534 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
1537 2018-06-08 Martin Liska <mliska@suse.cz>
1539 * tree-cfg.h (debug_function): Fix argument type to match
1542 2018-06-08 Martin Liska <mliska@suse.cz>
1544 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
1545 Remove usage of MPX-related (and removed) fields.
1546 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
1548 2018-06-08 David Malcolm <dmalcolm@redhat.com>
1550 * cfg.c (debug): Use TDF_NONE rather than 0.
1551 * cfghooks.c (debug): Likewise.
1552 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
1553 (struct dump_option_value_info): Convert to...
1554 (struct kv_pair): ...this template type.
1555 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
1557 (optinfo_verbosity_options): Likewise.
1558 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
1560 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
1561 than int for "optgroup_flags" param.
1562 (dump_generic_expr_loc): Use dump_flags_t rather than int for
1564 (dump_dec): Likewise.
1565 (dump_finish): Use TDF_NONE rather than 0.
1566 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
1567 rather than int for "optgroup_flags" param. Use TDF_NONE rather
1568 than 0. Update for change to option_ptr.
1569 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
1570 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
1571 0. Update for changes to optinfo_verbosity_options and
1573 (opt_info_switch_p): Convert optgroup_flags from int to
1575 (dump_basic_block): Use dump_flags_t rather than int
1576 for "dump_kind" param.
1577 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
1578 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
1579 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
1580 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
1581 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
1582 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
1583 TDF_NONE): Convert from macros to...
1584 (enum dump_flag): ...this new enum.
1585 (dump_flags_t): Update to use enum.
1586 (operator|, operator&, operator~, operator|=, operator&=):
1587 Implement for dump_flags_t.
1588 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
1589 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
1590 Convert from macros to...
1591 (enum optgroup_flag): ...this new enum.
1592 (optgroup_flags_t): New typedef.
1593 (operator|, operator|=): Implement for optgroup_flags_t.
1594 (struct dump_file_info): Convert field "alt_flags" to
1595 dump_flags_t. Convert field "optgroup_flags" to
1597 (dump_basic_block): Use dump_flags_t rather than int for param.
1598 (dump_generic_expr_loc): Likewise.
1599 (dump_dec): Likewise.
1600 (dump_register): Convert param "optgroup_flags" to
1602 (opt_info_enable_passes): Likewise.
1603 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
1605 * gimple-pretty-print.c (debug): Likewise.
1606 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
1607 (merged_store_group::apply_stores): Likewise.
1608 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
1609 * gimple.c (verify_gimple_pp): Likewise.
1610 * graphite-poly.c (print_pbb_body): Likewise.
1611 * passes.c (pass_manager::register_one_dump_file): Convert
1612 local "optgroup_flags" to optgroup_flags_t.
1613 * print-tree.c (print_node): Use TDF_NONE rather than 0.
1615 (debug_body): Likewise.
1616 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
1617 to optgroup_flags_t.
1618 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
1620 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
1621 (convert_mult_to_fma): Likewise.
1622 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
1623 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
1624 * tree-vect-data-refs.c (dump_lower_bound): Convert param
1625 "dump_kind" to dump_flags_t.
1627 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
1629 * config/rs6000/rs6000.c (min, max): Delete.
1631 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
1633 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
1634 -mabi=spe and -mabi=no-spe.
1636 2018-06-08 Martin Liska <mliska@suse.cz>
1638 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
1639 where we expect an existing summary.
1641 2018-06-08 Martin Liska <mliska@suse.cz>
1643 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
1644 * ipa-inline.h (estimate_edge_growth): Likewise.
1646 2018-06-08 Martin Liska <mliska@suse.cz>
1648 * cgraph.c (function_version_hasher::hash): Use
1649 cgraph_node::get_uid ().
1650 (function_version_hasher::equal):
1651 * cgraph.h (cgraph_node::get_uid): New method.
1652 * ipa-inline.c (update_caller_keys): Use
1653 cgraph_node::get_uid ().
1654 (update_callee_keys): Likewise.
1655 * ipa-utils.c (searchc): Likewise.
1656 (ipa_reduced_postorder): Likewise.
1657 * lto-cgraph.c (input_node): Likewise.
1658 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
1659 * symbol-summary.h (symtab_insertion): Likewise.
1660 (symtab_removal): Likewise.
1661 (symtab_duplication): Likewise.
1662 * tree-pretty-print.c (dump_function_header): Likewise.
1663 * tree-sra.c (convert_callers_for_node): Likewise.
1665 2018-06-08 Martin Liska <mliska@suse.cz>
1667 * cgraph.c (symbol_table::create_edge): Always assign a new
1669 (symbol_table::free_edge): Do not recycle numbers.
1670 * cgraph.h (cgraph_edge::get): New method.
1671 * symbol-summary.h (symtab_removal): Use it.
1672 (symtab_duplication): Likewise.
1673 (call_summary::hashable_uid): Remove.
1675 2018-06-08 Martin Liska <mliska@suse.cz>
1677 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
1678 (initialize_growth_caches): Remove.
1679 (free_growth_caches): Likewise.
1680 (do_estimate_edge_time): Use edge_growth_cache.
1681 (do_estimate_edge_size): Likewise.
1682 (do_estimate_edge_hints): Likewise.
1683 * ipa-inline.c (reset_edge_caches): Likewise.
1684 (recursive_inlining): Likewise.
1685 (inline_small_functions): Likewise.
1686 * ipa-inline.h (initialize_growth_caches): Remove.
1687 (estimate_edge_size): Likewise.
1688 (estimate_edge_time): Likewise.
1689 (estimate_edge_hints): Likewise.
1690 (reset_edge_growth_cache): Likewise.
1691 * symbol-summary.h (call_summary::remove): New method.
1693 2018-06-08 Martin Liska <mliska@suse.cz>
1695 * ipa-cp.c (class edge_clone_summary): New summary.
1696 (grow_edge_clone_vectors): Remove.
1697 (ipcp_edge_duplication_hook): Remove.
1698 (class edge_clone_summary_t): New call_summary class.
1699 (ipcp_edge_removal_hook): Remove.
1700 (edge_clone_summary_t::duplicate): New function.
1701 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
1702 (create_specialized_node): Likewise.
1703 (ipcp_driver): Initialize edge_clone_summaries and do not
1706 2018-06-08 Martin Liska <mliska@suse.cz>
1708 * symbol-summary.h (get): New function.
1709 (call_summary::m_initialize_when_cloning): New class member.
1711 2018-06-08 Martin Liska <mliska@suse.cz>
1713 * cgraph.c (cgraph_node::remove): Do not recycle uid.
1714 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
1715 (symbol_table::allocate_cgraph_symbol): Do not set uid.
1716 * passes.c (uid_hash_t): Record removed_nodes by their uids.
1717 (remove_cgraph_node_from_order): Use the removed_nodes set.
1718 (do_per_function_toporder): Likwise.
1719 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
1720 instead of summary_uid.
1721 (symtab_removal): Likewise.
1722 (symtab_duplication): Likewise.
1724 2018-06-08 Martin Liska <mliska@suse.cz>
1726 * ipa-cp.c (ipcp_store_bits_results): Use
1727 ipcp_transformation_sum.
1728 (ipcp_store_vr_results): Likewise.
1729 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
1731 (ipcp_transformation_initialize): ... this.
1732 (ipa_set_node_agg_value_chain):
1733 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
1734 (write_ipcp_transformation_info): Likewise.
1735 (read_ipcp_transformation_info): Likewise.
1736 (ipcp_update_bits): Likewise.
1737 (ipcp_update_vr): Likewise.
1738 (ipcp_transform_function): Likewise.
1739 * ipa-prop.h: Rename ipcp_transformation_summary to
1740 ipcp_transformation.
1741 (class ipcp_transformation_t): New function summary.
1742 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
1743 (ipa_get_agg_replacements_for_node): Likewise.
1745 2018-06-08 Martin Liska <mliska@suse.cz>
1747 * ipa-pure-const.c (struct funct_state_d): Do it class instead
1749 (class funct_state_summary_t): New function_summary class.
1750 (has_function_state): Remove.
1751 (get_function_state): Likewise.
1752 (set_function_state): Likewise.
1753 (add_new_function): Likewise.
1754 (funct_state_summary_t::insert): New function.
1755 (duplicate_node_data): Remove.
1756 (remove_node_data): Remove.
1757 (funct_state_summary_t::duplicate): New function.
1758 (register_hooks): Create new funct_state_summaries.
1759 (pure_const_generate_summary): Use it.
1760 (pure_const_write_summary): Likewise.
1761 (pure_const_read_summary): Likewise.
1762 (propagate_pure_const): Likewise.
1763 (propagate_nothrow): Likewise.
1764 (dump_malloc_lattice): Likewise.
1765 (propagate_malloc): Likewise.
1766 (execute): Do not register hooks, just remove summary
1768 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
1771 2018-06-08 Martin Liska <mliska@suse.cz>
1773 * ipa-reference.c (remove_node_data): Remove.
1774 (duplicate_node_data): Likewise.
1775 (class ipa_ref_var_info_summary_t): New class.
1776 (class ipa_ref_opt_summary_t): Likewise.
1777 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
1778 (get_reference_optimization_summary): Use
1779 ipa_ref_opt_sum_summaries.
1780 (set_reference_vars_info): Remove.
1781 (set_reference_optimization_summary): Likewise.
1782 (ipa_init): Create summaries.
1783 (init_function_info): Use function summary.
1784 (ipa_ref_opt_summary_t::duplicate): New function.
1785 (ipa_ref_opt_summary_t::remove): New function.
1786 (get_read_write_all_from_node): Fix GNU coding style.
1787 (propagate): Use function summary.
1788 (write_node_summary_p): Fix GNU coding style.
1789 (stream_out_bitmap): Likewise.
1790 (ipa_reference_read_optimization_summary): Use function summary.
1791 (ipa_reference_c_finalize): Do not release hooks.
1793 2018-06-08 Martin Liska <mliska@suse.cz>
1795 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
1796 (analyze_function_body): Extract multiple calls of get_create.
1797 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
1798 * ipa-inline.c (recursive_inlining): Use ::get method.
1799 * ipa-inline.h (estimate_edge_growth): Likewise.
1801 2018-06-08 Martin Liska <mliska@suse.cz>
1803 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
1805 (hsa_function_summary::hsa_function_summary): Use the new enum
1807 (hsa_gpu_implementation_p): Use hsa_summaries::get.
1808 * hsa-gen.c (hsa_get_host_function): Likewise.
1809 (get_brig_function_name): Likewise.
1810 * ipa-hsa.c (process_hsa_functions): Likewise.
1811 (ipa_hsa_write_summary): Likewise.
1812 * symbol-summary.h (symtab_duplication): Use ::get function/
1813 (get): New function.
1815 2018-06-08 Martin Liska <mliska@suse.cz>
1817 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
1819 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
1820 (hsa_register_kernel): Likewise.
1821 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
1822 * hsa-gen.c (hsa_get_host_function): Likewise.
1823 (get_brig_function_name): Likewise.
1824 (generate_hsa): Likewise.
1825 (pass_gen_hsail::execute): Likewise.
1826 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
1827 (devirtualization_time_bonus): Likewise.
1828 (ipcp_propagate_stage): Likewise.
1829 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
1830 (edge_set_predicate): Likewise.
1831 (evaluate_conditions_for_known_args): Likewise.
1832 (evaluate_properties_for_edge): Likewise.
1833 (ipa_fn_summary::reset): Likewise.
1834 (ipa_fn_summary_t::duplicate): Likewise.
1835 (dump_ipa_call_summary): Likewise.
1836 (ipa_dump_fn_summary): Likewise.
1837 (analyze_function_body): Likewise.
1838 (compute_fn_summary): Likewise.
1839 (estimate_edge_devirt_benefit): Likewise.
1840 (estimate_edge_size_and_time): Likewise.
1841 (estimate_calls_size_and_time): Likewise.
1842 (estimate_node_size_and_time): Likewise.
1843 (inline_update_callee_summaries): Likewise.
1844 (remap_edge_change_prob): Likewise.
1845 (remap_edge_summaries): Likewise.
1846 (ipa_merge_fn_summary_after_inlining): Likewise.
1847 (ipa_update_overall_fn_summary): Likewise.
1848 (read_ipa_call_summary): Likewise.
1849 (inline_read_section): Likewise.
1850 (write_ipa_call_summary): Likewise.
1851 (ipa_fn_summary_write): Likewise.
1852 (ipa_free_fn_summary): Likewise.
1853 * ipa-hsa.c (process_hsa_functions): Likewise.
1854 (ipa_hsa_write_summary): Likewise.
1855 (ipa_hsa_read_section): Likewise.
1856 * ipa-icf.c (sem_function::merge): Likewise.
1857 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
1858 (do_estimate_edge_time): Likewise.
1859 (estimate_size_after_inlining): Likewise.
1860 (estimate_growth): Likewise.
1861 (growth_likely_positive): Likewise.
1862 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
1863 (inline_call): Likewise.
1864 * ipa-inline.c (caller_growth_limits): Likewise.
1865 (can_inline_edge_p): Likewise.
1866 (can_inline_edge_by_limits_p): Likewise.
1867 (compute_uninlined_call_time): Likewise.
1868 (compute_inlined_call_time): Likewise.
1869 (want_inline_small_function_p): Likewise.
1870 (edge_badness): Likewise.
1871 (update_caller_keys): Likewise.
1872 (update_callee_keys): Likewise.
1873 (recursive_inlining): Likewise.
1874 (inline_small_functions): Likewise.
1875 (inline_to_all_callers_1): Likewise.
1876 (dump_overall_stats): Likewise.
1877 (early_inline_small_functions): Likewise.
1878 (early_inliner): Likewise.
1879 * ipa-inline.h (estimate_edge_growth): Likewise.
1880 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
1881 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
1882 * ipa-prop.h (IPA_NODE_REF): Likewise.
1883 (IPA_EDGE_REF): Likewise.
1884 * ipa-pure-const.c (malloc_candidate_p): Likewise.
1885 (propagate_malloc): Likewise.
1886 * ipa-split.c (execute_split_functions): Likewise.
1887 * symbol-summary.h: Rename get to get_create.
1889 (get_create): Likewise.
1890 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
1892 2018-06-08 Martin Liska <mliska@suse.cz>
1894 * symbol-summary.h (release): Move definition out of class
1896 (symtab_removal): Likewise.
1897 (symtab_duplication): Likewise.
1899 2018-06-08 Martin Liska <mliska@suse.cz>
1901 * symbol-summary.h (function_summary): Move constructor
1902 implementation out of class declaration.
1903 (release): Likewise.
1904 (symtab_insertion): Likewise.
1905 (symtab_removal): Likewise.
1906 (symtab_duplication): Likewise.
1909 2018-06-08 Martin Liska <mliska@suse.cz>
1911 * Makefile.in: Remove support for MPX (macros, related functions,
1912 fields in cgraph_node, ...).
1913 * builtin-types.def (BT_BND): Likewise.
1914 (BT_FN_BND_CONST_PTR): Likewise.
1915 (BT_FN_CONST_PTR_BND): Likewise.
1916 (BT_FN_VOID_PTR_BND): Likewise.
1917 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
1918 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
1919 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
1920 (expand_builtin_mempcpy_with_bounds): Likewise.
1921 (expand_builtin_memset_with_bounds): Likewise.
1922 (expand_builtin_memset_args): Likewise.
1923 (std_expand_builtin_va_start): Likewise.
1924 (expand_builtin): Likewise.
1925 (expand_builtin_with_bounds): Likewise.
1926 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
1927 (DEF_LIB_BUILTIN_CHKP): Likewise.
1928 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
1929 (DEF_CHKP_BUILTIN): Likewise.
1930 (BUILT_IN_MEMCPY): Likewise.
1931 (BUILT_IN_MEMMOVE): Likewise.
1932 (BUILT_IN_MEMPCPY): Likewise.
1933 (BUILT_IN_MEMSET): Likewise.
1934 (BUILT_IN_STPCPY): Likewise.
1935 (BUILT_IN_STRCAT): Likewise.
1936 (BUILT_IN_STRCHR): Likewise.
1937 (BUILT_IN_STRCPY): Likewise.
1938 (BUILT_IN_STRLEN): Likewise.
1939 (BUILT_IN_MEMCPY_CHK): Likewise.
1940 (BUILT_IN_MEMMOVE_CHK): Likewise.
1941 (BUILT_IN_MEMPCPY_CHK): Likewise.
1942 (BUILT_IN_MEMSET_CHK): Likewise.
1943 (BUILT_IN_STPCPY_CHK): Likewise.
1944 (BUILT_IN_STRCAT_CHK): Likewise.
1945 (BUILT_IN_STRCPY_CHK): Likewise.
1946 * calls.c (store_bounds): Likewise.
1947 (emit_call_1): Likewise.
1948 (special_function_p): Likewise.
1949 (maybe_warn_nonstring_arg): Likewise.
1950 (initialize_argument_information): Likewise.
1951 (finalize_must_preallocate): Likewise.
1952 (compute_argument_addresses): Likewise.
1953 (expand_call): Likewise.
1954 * cfgexpand.c (expand_call_stmt): Likewise.
1955 (expand_return): Likewise.
1956 (expand_gimple_stmt_1): Likewise.
1957 (pass_expand::execute): Likewise.
1958 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
1959 (cgraph_node::remove): Likewise.
1960 (cgraph_node::dump): Likewise.
1961 (cgraph_node::verify_node): Likewise.
1962 * cgraph.h (chkp_function_instrumented_p): Likewise.
1963 (symtab_node::get_alias_target): Likewise.
1964 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
1965 (cgraph_local_p): Likewise.
1966 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
1967 (cgraph_edge::rebuild_references): Likewise.
1968 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
1969 (walk_polymorphic_call_targets): Likewise.
1970 (cgraph_node::expand_thunk): Likewise.
1971 (symbol_table::output_weakrefs): Likewise.
1972 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
1973 (ix86_handle_option): Likewise.
1974 * config/i386/constraints.md: Likewise.
1975 * config/i386/i386-builtin-types.def (BND): Likewise.
1979 * config/i386/i386-builtin.def (BDESC_END): Likewise.
1980 (BDESC_FIRST): Likewise.
1982 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
1983 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
1984 * config/i386/i386.c (enum reg_class): Likewise.
1985 (ix86_target_string): Likewise.
1986 (ix86_option_override_internal): Likewise.
1987 (ix86_conditional_register_usage): Likewise.
1988 (ix86_valid_target_attribute_inner_p): Likewise.
1989 (ix86_set_indirect_branch_type): Likewise.
1990 (ix86_set_current_function): Likewise.
1991 (ix86_function_arg_regno_p): Likewise.
1992 (init_cumulative_args): Likewise.
1993 (ix86_function_arg_advance): Likewise.
1994 (ix86_function_arg): Likewise.
1995 (ix86_pass_by_reference): Likewise.
1996 (ix86_function_value_regno_p): Likewise.
1997 (ix86_function_value_1): Likewise.
1998 (ix86_function_value_bounds): Likewise.
1999 (ix86_return_in_memory): Likewise.
2000 (ix86_setup_incoming_vararg_bounds): Likewise.
2001 (ix86_va_start): Likewise.
2002 (indirect_thunk_need_prefix): Likewise.
2003 (print_reg): Likewise.
2004 (ix86_print_operand): Likewise.
2005 (ix86_expand_call): Likewise.
2006 (ix86_output_function_return): Likewise.
2007 (reg_encoded_number): Likewise.
2008 (BDESC_VERIFYS): Likewise.
2009 (ix86_init_mpx_builtins): Likewise.
2010 (ix86_init_builtins): Likewise.
2011 (ix86_emit_cmove): Likewise.
2012 (ix86_emit_move_max): Likewise.
2013 (ix86_expand_builtin): Likewise.
2014 (ix86_builtin_mpx_function): Likewise.
2015 (ix86_get_arg_address_for_bt): Likewise.
2016 (ix86_load_bounds): Likewise.
2017 (ix86_store_bounds): Likewise.
2018 (ix86_load_returned_bounds): Likewise.
2019 (ix86_store_returned_bounds): Likewise.
2020 (ix86_class_likely_spilled_p): Likewise.
2021 (ix86_hard_regno_mode_ok): Likewise.
2022 (x86_order_regs_for_local_alloc): Likewise.
2023 (ix86_mitigate_rop): Likewise.
2024 (ix86_bnd_prefixed_insn_p): Likewise.
2025 (ix86_mpx_bound_mode): Likewise.
2026 (ix86_make_bounds_constant): Likewise.
2027 (ix86_initialize_bounds): Likewise.
2028 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
2029 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
2030 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
2031 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
2032 (TARGET_CHKP_BOUND_MODE): Likewise.
2033 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
2034 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
2035 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
2036 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
2037 * config/i386/i386.h (TARGET_MPX): Likewise.
2038 (TARGET_MPX_P): Likewise.
2039 (VALID_BND_REG_MODE): Likewise.
2040 (FIRST_BND_REG): Likewise.
2041 (LAST_BND_REG): Likewise.
2042 (enum reg_class): Likewise.
2043 (BND_REG_P): Likewise.
2044 (BND_REGNO_P): Likewise.
2045 (BNDmode): Likewise.
2046 (ADJUST_INSN_LENGTH): Likewise.
2047 * config/i386/i386.md: Likewise.
2048 * config/i386/i386.opt: Likewise.
2049 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
2050 (defined): Likewise.
2051 (LINK_MPX): Likewise.
2052 (MPX_SPEC): Likewise.
2053 (LIBMPX_SPEC): Likewise.
2054 (LIBMPXWRAPPERS_SPEC): Likewise.
2055 (CHKP_SPEC): Likewise.
2056 * config/i386/predicates.md: Likewise.
2057 * dbxout.c (dbxout_type): Likewise.
2058 * doc/extend.texi: Likewise.
2059 * doc/invoke.texi: Likewise.
2060 * doc/md.texi: Likewise.
2061 * doc/tm.texi: Likewise.
2062 * doc/tm.texi.in: Likewise.
2063 * dwarf2out.c (is_base_type): Likewise.
2064 (gen_formal_types_die): Likewise.
2065 (gen_subprogram_die): Likewise.
2066 (gen_type_die_with_usage): Likewise.
2067 (gen_decl_die): Likewise.
2068 (dwarf2out_late_global_decl): Likewise.
2069 * expr.c (expand_assignment): Likewise.
2070 (emit_storent_insn): Likewise.
2071 (store_expr_with_bounds): Likewise.
2072 (store_expr): Likewise.
2073 (expand_expr_real_1): Likewise.
2074 * expr.h (store_expr_with_bounds): Likewise.
2075 * function.c (use_register_for_decl): Likewise.
2076 (struct bounds_parm_data): Likewise.
2077 (assign_parms_augmented_arg_list): Likewise.
2078 (assign_parm_find_entry_rtl): Likewise.
2079 (assign_parm_is_stack_parm): Likewise.
2080 (assign_parm_load_bounds): Likewise.
2081 (assign_bounds): Likewise.
2082 (assign_parms): Likewise.
2083 (expand_function_start): Likewise.
2084 * gcc.c (CHKP_SPEC): Likewise.
2085 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
2086 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
2087 (wrestrict_dom_walker::check_call): Likewise.
2088 * gimple.c (gimple_build_call_from_tree): Likewise.
2089 * gimple.h (enum gf_mask): Likewise.
2090 (gimple_call_with_bounds_p): Likewise.
2091 (gimple_call_set_with_bounds): Likewise.
2092 * gimplify.c (gimplify_init_constructor): Likewise.
2093 * ipa-cp.c (initialize_node_lattices): Likewise.
2094 (propagate_constants_across_call): Likewise.
2095 (find_more_scalar_values_for_callers_subset): Likewise.
2096 * ipa-hsa.c (process_hsa_functions): Likewise.
2097 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
2098 * ipa-icf.c (sem_function::merge): Likewise.
2099 * ipa-inline.c (early_inliner): Likewise.
2100 * ipa-pure-const.c (warn_function_noreturn): Likewise.
2101 (warn_function_cold): Likewise.
2102 (propagate_pure_const): Likewise.
2103 * ipa-ref.h (enum GTY): Likewise.
2104 * ipa-split.c (find_retbnd): Likewise.
2105 (consider_split): Likewise.
2106 (split_function): Likewise.
2107 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
2108 * ipa.c (walk_polymorphic_call_targets): Likewise.
2109 (symbol_table::remove_unreachable_nodes): Likewise.
2110 (process_references): Likewise.
2111 (cgraph_build_static_cdtor_1): Likewise.
2112 * lto-cgraph.c (lto_output_node): Likewise.
2113 (output_refs): Likewise.
2114 (compute_ltrans_boundary): Likewise.
2115 (input_overwrite_node): Likewise.
2116 (input_node): Likewise.
2117 (input_cgraph_1): Likewise.
2118 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
2119 * passes.c (pass_manager::execute_early_local_passes): Likewise.
2120 (class pass_chkp_instrumentation_passes): Likewise.
2121 (make_pass_chkp_instrumentation_passes): Likewise.
2122 * passes.def: Likewise.
2123 * rtl.h (struct GTY): Likewise.
2124 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
2125 * stor-layout.c (layout_type): Likewise.
2126 * symtab.c: Likewise.
2127 * target.def: Likewise.
2128 * targhooks.c (default_chkp_bound_type): Likewise.
2129 (default_chkp_bound_mode): Likewise.
2130 (default_builtin_chkp_function): Likewise.
2131 (default_chkp_function_value_bounds): Likewise.
2132 (default_chkp_make_bounds_constant): Likewise.
2133 (default_chkp_initialize_bounds): Likewise.
2134 * targhooks.h (default_chkp_bound_type): Likewise.
2135 (default_chkp_bound_mode): Likewise.
2136 (default_builtin_chkp_function): Likewise.
2137 (default_chkp_function_value_bounds): Likewise.
2138 (default_chkp_make_bounds_constant): Likewise.
2139 (default_chkp_initialize_bounds): Likewise.
2140 * toplev.c (compile_file): Likewise.
2141 (process_options): Likewise.
2142 * tree-core.h (DEF_BUILTIN): Likewise.
2143 (DEF_BUILTIN_CHKP): Likewise.
2144 * tree-inline.c (declare_return_variable): Likewise.
2145 (remap_gimple_stmt): Likewise.
2146 (copy_bb): Likewise.
2147 (initialize_inlined_parameters): Likewise.
2148 (expand_call_inline): Likewise.
2149 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
2150 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
2151 (make_pass_ipa_chkp_produce_thunks): Likewise.
2152 (make_pass_chkp): Likewise.
2153 (make_pass_chkp_opt): Likewise.
2154 (make_pass_chkp_instrumentation_passes): Likewise.
2155 * tree-pretty-print.c (dump_generic_node): Likewise.
2156 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
2157 * tree-ssa-dce.c (propagate_necessity): Likewise.
2158 (eliminate_unnecessary_stmts): Likewise.
2159 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
2160 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
2161 * tree-ssa-sccvn.h: Likewise.
2162 * tree-ssa-strlen.c (get_string_length): Likewise.
2163 (valid_builtin_call): Likewise.
2164 (adjust_last_stmt): Likewise.
2165 (handle_builtin_strchr): Likewise.
2166 (handle_builtin_strcpy): Likewise.
2167 (handle_builtin_stxncpy): Likewise.
2168 (handle_builtin_memcpy): Likewise.
2169 (handle_builtin_strcat): Likewise.
2170 (strlen_check_and_optimize_stmt): Likewise.
2171 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
2172 * tree-streamer-in.c: Likewise.
2173 * tree-streamer.c (record_common_node): Likewise.
2174 * tree.c (tree_code_size): Likewise.
2175 (wide_int_to_tree_1): Likewise.
2176 (type_contains_placeholder_1): Likewise.
2177 (build_common_tree_nodes): Likewise.
2178 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
2179 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
2180 (POINTER_BOUNDS_P): Likewise.
2181 (BOUNDED_TYPE_P): Likewise.
2182 (BOUNDED_P): Likewise.
2183 (CALL_WITH_BOUNDS_P): Likewise.
2184 (pointer_bounds_type_node): Likewise.
2185 * value-prof.c (gimple_ic): Likewise.
2186 * var-tracking.c (vt_add_function_parameters): Likewise.
2187 * varasm.c (make_decl_rtl): Likewise.
2188 (assemble_start_function): Likewise.
2189 (output_constant): Likewise.
2190 (maybe_assemble_visibility): Likewise.
2191 * varpool.c (ctor_for_folding): Likewise.
2192 * chkp-builtins.def: Remove.
2193 * ipa-chkp.c: Remove.
2194 * ipa-chkp.h: Remove.
2195 * rtl-chkp.c: Remove.
2196 * rtl-chkp.h: Remove.
2197 * tree-chkp-opt.c: Remove.
2198 * tree-chkp.c: Remove.
2199 * tree-chkp.h: Remove.
2201 2018-06-07 Carl Love <cel@us.ibm.com>
2203 * gcc/config/rs6000/vsx.md (vextract_fp_from_shorth,
2204 vextract_fp_from_shortl): Add BE support.
2206 2018-06-07 Paul Koning <ni1d@arrl.net>
2208 * compare-elim.c (try_merge_compare): Don't merge compare if
2209 address contains a side effect.
2210 (try_eliminate_compare): Likewise.
2212 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
2214 * config.gcc: Support "tremont".
2215 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
2216 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2218 * config/i386/i386.c (m_TREMONT): Define.
2219 (processor_target_table): Add "tremont".
2220 (PTA_TREMONT): Define.
2221 (ix86_lea_outperforms): Add TARGET_TREMONT.
2222 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
2223 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
2224 and M_INTEL_GOLDMONT_PLUS.
2225 (fold_builtin_cpu): Add "tremont".
2226 (ix86_add_stmt_cost): Add TARGET_TREMONT.
2227 (ix86_option_override_internal): Add "tremont".
2228 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
2229 (processor_type): Add PROCESSOR_TREMONT.
2230 * config/i386/x86-tune.def: Add m_TREMONT.
2231 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
2233 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
2235 * gcc/config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
2236 symbol defined for msp430i* devices to be lower case.
2238 gcc/testsuite/gcc.target/msp430/
2239 * msp430i-device-symbol.c: New test.
2240 * msp430f-device-symbol.c: New test.
2241 * msp430.h: New test header file.
2243 2018-06-07 Richard Biener <rguenther@suse.de>
2245 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
2246 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
2247 Properly wrap signed arithmetic if overflow wraps.
2249 2018-06-07 Jakub Jelinek <jakub@redhat.com>
2251 PR tree-optimization/69615
2252 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
2253 of a cast from a same precision integral SSA_NAME in a bb dominated
2254 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
2255 cast to utype if rhs2 has already a compatible type.
2257 2018-06-07 Richard Biener <rguenther@suse.de>
2259 PR tree-optimization/85935
2260 * graphite-scop-detection.c (find_params_in_bb): Analyze
2261 condition operands with respect to the correct loop. Assert
2262 the analysis doesn't fail.
2264 2018-06-04 Carl Love <cel@us.ibm.com>
2266 * gcc/config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
2267 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
2268 as it is slightly cheaper.
2269 (first_match_or_eos_index_<mode>):
2270 Calculate index using natural element order.
2271 (first_match_index_<mode>):
2272 Calculate index using natural element order.
2273 (first_match_or_eos_index_<mode>):
2274 Calculate index using natural order.
2275 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
2276 for BE and LE modes.
2277 * gcc/config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
2278 P9V_BUILTIN_VCLZLSBB_V16QI.
2279 * gcc/config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
2282 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
2284 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
2285 indentation and line wrap for many prototypes. Add missing
2286 @smallexample directives around block of prototypes for vec_xl and
2289 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
2291 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
2292 track if we pass or return IEEE 128-bit floating point.
2293 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
2294 C++ mangling that is compatible with GCC 8.1.
2295 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
2296 (init_cumulative_args): Note if we pass or return IEEE 128-bit
2297 floating point types.
2298 (rs6000_function_arg_advance_1): Likewise.
2299 (rs6000_mangle_type): Optionally generate mangled names that match
2300 what GCC 8.1 generated for IEEE 128-bit floating point types.
2301 (rs6000_globalize_decl_name): If we have an external function that
2302 passes or returns IEEE 128-bit types, generate a weak reference
2303 from the mangled name used in GCC 8.1 to the current mangled
2305 (rs6000_init_builtins): Make __ibm128 use the long double type if
2306 long double is IBM extended double. Make __float128 use the long
2307 double type if long double is IEEE 128-bit.
2310 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
2311 macro for __ibm128 built-in functions.
2312 (PACK_IF): Add __ibm128 pack/unpack functions.
2313 (UNPACK_IF): Likewise.
2314 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
2315 enable long double built-in functions if long double is IEEE
2316 128-bit floating point.
2317 (rs6000_invalid_builtin): Update long double built-in function
2319 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
2320 functions, adjust the built-in function to use the long double
2321 built-in function if __ibm128 and long double are the same type.
2322 * doc/extend.texi (PowerPC builtins): Update documention for
2323 __builtin_{,un}pack_longdouble. Add documentation for
2324 __builtin_{,un}pack_ibm128.
2326 2018-06-06 Jim Wilson <jimw@sifive.com>
2328 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
2329 (struct machine_function): New field interrupt_mode.
2330 (riscv_handle_type_attribute): New function. Add forward declaration.
2331 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
2332 (riscv_expand_epilogue): Check interrupt_mode field.
2333 (riscv_set_current_function): Check interrupt attribute args and
2334 set interrupt_mode field.
2335 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
2336 (riscv_sret, riscv_uret): New.
2337 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
2338 new arguments to interrupt attribute.
2340 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
2343 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
2344 Don't handle -mcpu=power8 if -mpower9-vector is also used.
2346 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
2348 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
2349 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
2350 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
2351 several redundant entries.
2353 2018-06-06 David Malcolm <dmalcolm@redhat.com>
2355 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
2356 type from "rtx" to "rtx_insn *".
2357 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
2358 for local "call_insn", removing cast.
2359 (ix86_expand_call): Likewise, introducing a "call_insn" local.
2361 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
2363 PR tree-optimization/86066
2364 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
2365 for BIT_INSERT_EXPR stores.
2367 2018-06-06 Richard Biener <rguenther@suse.de>
2369 PR tree-optimization/86062
2370 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
2371 component refs ontop
2372 of to be offsetted base.
2374 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
2376 * gcc/config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
2377 to be static and remove check on interrupt attribute name.
2379 gcc/testsuite/gcc.target/msp430/
2380 * function-attributes-4.c: New test.
2381 * static-interrupts.c: New test.
2383 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
2385 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
2386 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
2388 2018-06-05 Steve Ellcey <sellcey@cavium.com>
2391 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
2393 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
2394 Remove second argument, change how error is called.
2395 (aarch64_layout_arg): Remove second argument from
2396 aarch64_err_no_fpadvsimd call.
2397 (aarch64_init_cumulative_args): Ditto.
2398 (aarch64_gimplify_va_arg_expr): Ditto.
2399 * config/aarch64/aarch64.md (mov<mode>): Ditto.
2401 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
2403 * config/i386/i386.md (simple_return_indirect_internal): New expander.
2404 (*simple_return_indirect_internal<mode>): Rename from
2405 simple_return_indirect_internal. Use W mode iterator.
2406 (rstorssp): New expander.
2407 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
2408 (clrssbsy): New expander.
2409 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
2411 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
2413 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
2415 (cmse_check_pointed_object): Likewise.
2417 2018-06-05 Martin Liska <mliska@suse.cz>
2419 PR gcov-profile/47618
2420 * doc/invoke.texi: Document how -fprofile-dir format
2423 2018-06-05 Richard Biener <rguenther@suse.de>
2425 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
2426 removal pretend DOM info isn't available so we do not update
2427 it and only remove edges, not dominated blocks. Actually free
2428 DOM info in case we removed something. Remove unreachable blocks.
2429 (mfb_keep_latches): Work with either DOM info or marked backedges.
2430 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
2431 first. Mark backedges if DOM info isn't available.
2432 (Re-)compute DOM info after cleanup_control_flow_pre.
2434 2018-06-05 Richard Biener <rguenther@suse.de>
2436 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
2437 (locus_discrim_hasher::hash): Adjust.
2438 (locus_discrim_hasher::equal): Likewise.
2439 (next_discriminator_for_locus): Work on line directly.
2440 (same_line_p): Pass in expanded locus1 as well.
2441 (assign_discriminators): Avoid redundant location expansions.
2443 2018-06-05 Richard Biener <rguenther@suse.de>
2445 PR tree-optimization/86046
2446 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
2447 if required after clearing TREE_ADDRESSABLE.
2449 2018-06-05 Richard Biener <rguenther@suse.de>
2451 PR tree-optimization/86047
2452 * tree-ssa-loop.c (for_each_index): Glob handling of all
2453 decls and constants and really handle all of them.
2455 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2458 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
2459 qualifier_void_pointer and qualifier_const_void_pointer.
2460 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
2461 (arm_init_builtins): Handle the above.
2462 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
2463 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
2464 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
2467 2018-06-05 Martin Liska <mliska@suse.cz>
2469 * auto-profile.c (read_autofdo_file): Do not use
2470 gcov_ctr_summary struct.
2471 (afdo_callsite_hot_enough_for_early_inline): Likewise.
2472 * coverage.c (struct counts_entry): Likewise.
2473 (read_counts_file): Read just single summary entry.
2474 (get_coverage_counts): Use gcov_summary struct.
2475 * coverage.h (get_coverage_counts): Likewise.
2476 * gcov-dump.c (dump_working_sets): Likewise.
2477 (tag_summary): Dump just single summary.
2478 * gcov-io.c (gcov_write_summary): Write just histogram
2480 (gcov_read_summary): Read just single summary.
2481 (compute_working_sets): Use gcov_summary struct.
2482 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
2483 of GCOV_COUNTERS_SUMMABLE.
2484 (GCOV_COUNTERS_SUMMABLE): Remove.
2485 (GCOV_FIRST_VALUE_COUNTER): Replace with
2486 GCOV_COUNTER_V_INTERVAL.
2487 (struct gcov_ctr_summary): Remove.
2488 (struct gcov_summary): Directly use fields of former
2490 (compute_working_sets): Use gcov_summary struct.
2491 * gcov.c (read_count_file): Do not use ctrs fields.
2492 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
2494 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
2496 * profile.c: Likewise.
2497 * profile.h: Likewise.
2499 2018-06-05 Martin Liska <mliska@suse.cz>
2501 PR gcov-profile/84846
2502 * gcov.c (output_lines): Print working directory only
2503 in intermediate format.
2505 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
2507 * config/s390/s390-builtin-types.def: Add void function type.
2508 * config/s390/s390-builtins.def: Use the function type for the
2511 2018-06-04 Jim Wilson <jimw@sifive.com>
2513 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
2515 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
2516 and maybe_eh_return. Change regno to unsigned int. Use new args to
2517 handle EH_RETURN_DATA_REGNO registers properly.
2518 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
2519 (riscv_expand_epilogue): Update comment. Change argument name and
2520 type. Update code to use new name and type. Pass new args to
2521 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
2523 * config/riscv/riscv.md (NORMAL_RETURN): New.
2524 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
2525 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
2526 (eh_return): Call gen_eh_return_internal and emit barrier.
2527 (eh_return_internal): Call riscv_expand_epilogue.
2529 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
2531 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
2532 bit_insertion field and declare can_be_merged_into method.
2533 (merged_store_group::can_be_merged_into): New method.
2534 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
2535 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
2536 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
2538 2018-06-04 Richard Biener <rguenther@suse.de>
2540 PR tree-optimization/85955
2541 * builtins.c (fold_builtin_sincos): Convert pointers to
2542 destination to appropriate type before dereferencing.
2544 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
2546 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
2548 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
2550 * expr.c (expand_expr_real_1): Force the operand into memory if
2551 its TYPE_MODE is BLKmode and if there is no integer mode for
2552 the number of bits being extracted.
2554 2018-06-04 Jakub Jelinek <jakub@redhat.com>
2558 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
2559 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
2561 2018-06-04 Richard Biener <rguenther@suse.de>
2563 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
2564 (cleanup_tree_cfg_noloop): ... single caller. Do
2565 start_recording_case_labels later.
2567 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
2569 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
2570 to _IMMINTRIN_H_INCLUDED.
2571 * config/i386/pconfigintrin.h: Ditto.
2572 * config/i386/waitpkgintrin.h: Ditto.
2573 * config/i386/immintrin.h: Add includes for sgxintrin.h,
2574 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
2575 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
2576 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
2577 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
2578 waitpkgintrin.h and cldemoteintrin.h.
2580 2018-06-04 Richard Biener <rguenther@suse.de>
2582 PR tree-optimization/86038
2583 * tracer.c (find_best_successor): Check probability for
2584 being initialized, bail out if not.
2586 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
2589 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
2590 of bits to ignore when comparing architectures.
2592 2018-06-04 Jakub Jelinek <jakub@redhat.com>
2594 PR tree-optimization/69615
2595 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
2596 maximum or minimum of the type, try to merge it also as if
2597 range1 is + [-, x - 1] or + [x + 1, -].
2600 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
2602 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
2604 PR tree-optimization/86034
2605 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
2606 the unsigned bitfield type in a bit insertion sequence if it does not
2607 have a larger precision than the bitfield size.
2608 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
2610 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
2612 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
2614 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
2616 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
2617 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
2618 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
2619 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
2621 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
2623 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
2624 Disable -fdelete-null-pointer-checks for ELF toolchain.
2626 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
2627 Kito Cheng <kito.cheng@gmail.com>
2629 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
2630 (nds32le-*-*, nds32be-*-*): Integrate checking process.
2631 (nds32*-*-*): Add glibc and uclibc conditions.
2632 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
2633 (TARGET_EXCEPT_UNWIND_INFO): Define.
2634 * config/nds32/elf.h: New file.
2635 * config/nds32/linux.h: New file.
2636 * config/nds32/nds32-elf.opt: New file.
2637 * config/nds32/nds32-linux.opt: New file.
2638 * config/nds32/nds32-fp-as-gp.c
2639 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
2640 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
2642 (nds32_asm_file_end): Ditto.
2643 (nds32_print_operand): Ditto.
2644 (nds32_insert_attributes): Ditto.
2645 (nds32_init_libfuncs): New function.
2646 (TARGET_HAVE_TLS): Define.
2647 (TARGET_INIT_LIBFUNCS): Define.
2648 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
2650 (TARGET_ELF): Apply different mcmodel setting.
2651 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
2652 been migrated into elf.h and linux.h files.
2653 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
2654 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
2655 (mcmodel): The content has been migrated into nds32-elf.opt and
2656 nds32-linux.opt files.
2657 * config/nds32/t-elf: New file.
2658 * config/nds32/t-linux: New file.
2660 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
2661 Shiva Chen <shiva0217@gmail.com>
2663 * config/nds32/constants.md (unspec_volatile_element): Add
2664 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
2665 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
2667 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
2668 (make_pass_nds32_fp_as_gp): Declare.
2669 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
2671 (nds32_asm_function_end_prologue): Remove unused asm output.
2672 (nds32_asm_function_begin_epilogue): Remove unused asm output.
2673 (nds32_asm_file_start): Output necessary fp_as_gp information.
2674 (nds32_option_override): Adjust register usage.
2675 (nds32_expand_prologue): Consider fp_as_gp situation.
2676 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
2677 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
2680 (simple_return): Ditto.
2681 (omit_fp_begin): Output special directive for fp_as_gp.
2682 (omit_fp_end): Output special directive for fp_as_gp.
2683 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
2684 mforbid-fp-as-gp): New options.
2686 2018-06-01 Mark Wielaard <mark@klomp.org>
2688 * dwarf2out.c (dwarf2out_finish): Remove generation of
2689 DW_AT_loclists_base.
2691 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
2693 * gimple-ssa-store-merging.c: Include gimple-fold.h.
2694 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
2695 (struct merged_store_group): Add bit_insertion field.
2696 (dump_char_array): Use standard hexadecimal format.
2697 (merged_store_group::merged_store_group): Set bit_insertion to false.
2698 (merged_store_group::apply_stores): Use optimal buffer size. Deal
2699 with BIT_INSERT_EXPR stores. Move up code updating the mask and
2700 also print the mask in the dump file.
2701 (pass_store_merging::gate): Minor tweak.
2702 (imm_store_chain_info::coalesce_immediate): Fix wrong association
2703 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
2704 stores with INTEGER_CST stores.
2705 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
2706 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
2707 and use it throughout. Generate bit insertion sequences if need be.
2708 (pass_store_merging::process_store): Remove redundant condition.
2709 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
2711 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
2713 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
2714 the 128-bit floating point types. Fix function comment.
2716 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2718 * config/aarch64/aarch64-simd.md
2719 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
2721 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
2724 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
2726 PR tree-optimization/85989
2727 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
2729 (backprop::intersect_uses): Check it when deciding whether this
2730 is a backedge reference.
2731 (backprop::process_block): Add each phi to m_visited_phis
2732 after visiting it, then clear it at the end.
2734 2018-06-01 Richard Biener <rguenther@suse.de>
2736 * tree-vectorizer.h (vect_dr_stmt): New function.
2737 (vect_get_load_cost): Adjust.
2738 (vect_get_store_cost): Likewise.
2739 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
2740 Use vect_dr_stmt instead of DR_SMTT.
2741 (vect_record_base_alignments): Likewise.
2742 (vect_calculate_target_alignment): Likewise.
2743 (vect_compute_data_ref_alignment): Likewise and make static.
2744 (vect_update_misalignment_for_peel): Likewise.
2745 (vect_verify_datarefs_alignment): Likewise.
2746 (vector_alignment_reachable_p): Likewise.
2747 (vect_get_data_access_cost): Likewise. Pass down
2748 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
2749 (vect_get_peeling_costs_all_drs): Likewise.
2750 (vect_peeling_hash_get_lowest_cost): Likewise.
2751 (vect_enhance_data_refs_alignment): Likewise.
2752 (vect_find_same_alignment_drs): Likewise.
2753 (vect_analyze_data_refs_alignment): Likewise.
2754 (vect_analyze_group_access_1): Likewise.
2755 (vect_analyze_group_access): Likewise.
2756 (vect_analyze_data_ref_access): Likewise.
2757 (vect_analyze_data_ref_accesses): Likewise.
2758 (vect_vfa_segment_size): Likewise.
2759 (vect_small_gap_p): Likewise.
2760 (vectorizable_with_step_bound_p): Likewise.
2761 (vect_prune_runtime_alias_test_list): Likewise.
2762 (vect_analyze_data_refs): Likewise.
2763 (vect_supportable_dr_alignment): Likewise.
2764 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
2765 (vect_gen_prolog_loop_niters): Likewise.
2766 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
2767 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
2769 (vect_recog_mask_conversion_pattern): Likewise.
2770 (vect_try_gather_scatter_pattern): Likewise.
2771 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
2772 to vect_get_store_cost.
2773 (vect_get_store_cost): Get stmt_info instead of DR.
2774 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
2775 (vect_get_load_cost): Get stmt_info instead of DR.
2777 2018-06-01 Richard Biener <rguenther@suse.de>
2780 * gimple-fold.c (var_decl_component_p): Also allow offsetted
2781 vars wrapped in MEM_REFs.
2783 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
2785 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
2786 Fix subreg tests so that we only return a choice between
2787 GENERAL_REGS and FP_REGS if the original classes included both.
2789 2018-06-01 Richard Biener <rguenther@suse.de>
2792 * tree-ssa-structalias.c (get_function_part_constraint):
2793 Handle NULL fi->decl.
2794 (find_func_aliases_for_call): Properly handle indirect
2795 fi from direct call.
2796 (find_func_clobbers): Likewise.
2797 (ipa_pta_execute): Likewise.
2798 (create_variable_info_for): For functions that are ifunc_resolver
2799 resolve to a varinfo that contains the result of the resolver call.
2800 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
2803 2018-05-31 Michael Collison <michael.collison@arm.com>
2805 * config/aarch64/aarch64.md:
2806 (*fix_to_zero_extenddfdi2): New pattern.
2807 * gcc.target/aarch64/fix_extend1.c: New testcase.
2809 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
2813 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
2814 and BUILT_IN_STRNCMP_EQ.
2815 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
2816 BUILT_IN_STRNCMP_EQ.
2817 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
2818 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
2819 (gimple_fold_builtin): Likewise.
2820 * tree-ssa-strlen.c (compute_string_length): New function.
2821 (determine_min_obsize): New function.
2822 (handle_builtin_string_cmp): New function to handle calls to
2823 string compare functions.
2824 (strlen_optimize_stmt): Add handling to builtin string compare
2826 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
2827 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
2828 * tree.c (build_common_builtin_nodes): Add new defines of
2829 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
2831 2018-05-31 Jakub Jelinek <jakub@redhat.com>
2834 * bb-reorder.c (pass_partition_blocks::gate): Return false for
2835 functions with naked attribute.
2837 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
2839 * config/i386/sse.md (avx_vec_concat<mode>):
2840 Substitute concat_tg_mode mode attribute with xtg_mode.
2841 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
2842 (concat_tg_mode): Remove mode attribute.
2844 2018-05-31 Martin Sebor <msebor@redhat.com>
2847 * calls.c (alloc_max_size): Correct a logic error/typo.
2848 Treat excessive arguments as infinite. Warn for invalid arguments.
2849 * doc/invoke.texi (-Walloc-size-larger-than): Update.
2851 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
2854 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
2855 and movx for Haswell.
2857 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
2858 Cesar Philippidis <cesar@codesourcery.com>
2861 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
2862 when emitting error on private/firstprivate reductions.
2863 * omp-low.c (lower_omp_target): Avoid reference-type processing
2864 on pointers for firstprivate clause.
2866 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
2868 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
2871 * config/aarch64/aarch64-simd.md
2872 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
2873 (aarch64_ld1_x3_<mode>): Likewise
2874 (aarch64_st1x2<VALLDIF:mode>): Likewise
2875 (aarch64_st1_x2_<mode>): Likewise
2876 (aarch64_st1x3<VALLDIF:mode>): Likewise
2877 (aarch64_st1_x3_<mode>): Likewise
2878 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
2879 (vld1_s8_x3): Likewise.
2880 (vld1_u16_x3): Likewise.
2881 (vld1_s16_x3): Likewise.
2882 (vld1_u32_x3): Likewise.
2883 (vld1_s32_x3): Likewise.
2884 (vld1_u64_x3): Likewise.
2885 (vld1_s64_x3): Likewise.
2886 (vld1_f16_x3): Likewise.
2887 (vld1_f32_x3): Likewise.
2888 (vld1_f64_x3): Likewise.
2889 (vld1_p8_x3): Likewise.
2890 (vld1_p16_x3): Likewise.
2891 (vld1_p64_x3): Likewise.
2892 (vld1q_u8_x3): Likewise.
2893 (vld1q_s8_x3): Likewise.
2894 (vld1q_u16_x3): Likewise.
2895 (vld1q_s16_x3): Likewise.
2896 (vld1q_u32_x3): Likewise.
2897 (vld1q_s32_x3): Likewise.
2898 (vld1q_u64_x3): Likewise.
2899 (vld1q_s64_x3): Likewise.
2900 (vld1q_f16_x3): Likewise.
2901 (vld1q_f32_x3): Likewise.
2902 (vld1q_f64_x3): Likewise.
2903 (vld1q_p8_x3): Likewise.
2904 (vld1q_p16_x3): Likewise.
2905 (vld1q_p64_x3): Likewise.
2906 (vst1_s64_x2): Likewise.
2907 (vst1_u64_x2): Likewise.
2908 (vst1_f64_x2): Likewise.
2909 (vst1_s8_x2): Likewise.
2910 (vst1_p8_x2): Likewise.
2911 (vst1_s16_x2): Likewise.
2912 (vst1_p16_x2): Likewise.
2913 (vst1_s32_x2): Likewise.
2914 (vst1_u8_x2): Likewise.
2915 (vst1_u16_x2): Likewise.
2916 (vst1_u32_x2): Likewise.
2917 (vst1_f16_x2): Likewise.
2918 (vst1_f32_x2): Likewise.
2919 (vst1_p64_x2): Likewise.
2920 (vst1q_s8_x2): Likewise.
2921 (vst1q_p8_x2): Likewise.
2922 (vst1q_s16_x2): Likewise.
2923 (vst1q_p16_x2): Likewise.
2924 (vst1q_s32_x2): Likewise.
2925 (vst1q_s64_x2): Likewise.
2926 (vst1q_u8_x2): Likewise.
2927 (vst1q_u16_x2): Likewise.
2928 (vst1q_u32_x2): Likewise.
2929 (vst1q_u64_x2): Likewise.
2930 (vst1q_f16_x2): Likewise.
2931 (vst1q_f32_x2): Likewise.
2932 (vst1q_f64_x2): Likewise.
2933 (vst1q_p64_x2): Likewise.
2934 (vst1_s64_x3): Likewise.
2935 (vst1_u64_x3): Likewise.
2936 (vst1_f64_x3): Likewise.
2937 (vst1_s8_x3): Likewise.
2938 (vst1_p8_x3): Likewise.
2939 (vst1_s16_x3): Likewise.
2940 (vst1_p16_x3): Likewise.
2941 (vst1_s32_x3): Likewise.
2942 (vst1_u8_x3): Likewise.
2943 (vst1_u16_x3): Likewise.
2944 (vst1_u32_x3): Likewise.
2945 (vst1_f16_x3): Likewise.
2946 (vst1_f32_x3): Likewise.
2947 (vst1_p64_x3): Likewise.
2948 (vst1q_s8_x3): Likewise.
2949 (vst1q_p8_x3): Likewise.
2950 (vst1q_s16_x3): Likewise.
2951 (vst1q_p16_x3): Likewise.
2952 (vst1q_s32_x3): Likewise.
2953 (vst1q_s64_x3): Likewise.
2954 (vst1q_u8_x3): Likewise.
2955 (vst1q_u16_x3): Likewise.
2956 (vst1q_u32_x3): Likewise.
2957 (vst1q_u64_x3): Likewise.
2958 (vst1q_f16_x3): Likewise.
2959 (vst1q_f32_x3): Likewise.
2960 (vst1q_f64_x3): Likewise.
2961 (vst1q_p64_x3): Likewise.
2963 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
2965 * config/msp430/msp430.c (msp430_output_labelref): Prepend
2966 user_label_prefix to name.
2968 * tree-core.h: Update comment about the format of NAME string
2969 passed to handler in attribute_spec.
2971 * config/msp430/msp430.md: Remove erroneous subreg expression from
2972 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
2973 zero_extend{q,h}isi2.
2975 2018-05-30 Borislav Petkov <bp@suse.de>
2977 * doc/extend.texi: Document some architecture specific
2978 constraints and sort entries.
2980 2018-05-30 Martin Sebor <msebor@redhat.com>
2983 * builtins.c (expand_builtin_stpcpy_1): New function.
2984 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
2985 only if the former succeeds.
2987 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
2989 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
2992 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
2994 * doc/invoke.texi (-flinker-output): Document
2996 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
2998 * passes.c (ipa_write_summaries): Only modify statements if body
3000 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
3001 incrementally linking.
3002 (ipa_passes): Likewise.
3003 * lto-cgraph.c (lto_output_node): When incrementally linking do not
3004 pass down resolution info.
3005 * common.opt (flag_incremental_link): Update info.
3006 * gcc.c (plugin specs): Turn flinker-output=* to
3007 -plugin-opt=-linker-output-known
3008 * toplev.c (compile_file): Also cut compilation when doing incremental
3010 * flag-types. (enum lto_partition_model): Add
3011 LTO_LINKER_OUTPUT_NOLTOREL.
3012 (invoke.texi): Add -flinker-output docs.
3013 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
3014 link same way as WPA; do not stream in dead initializers.
3016 * dwarf2out.c (dwarf2out_die_ref_for_decl,
3017 darf2out_register_external_decl): Support incremental link.
3019 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
3021 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
3023 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
3025 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
3026 it down to simple_object_copy_lto_debug_sections.
3027 (run_gcc): Determine incremental LTO link time and configure
3028 lto1 into non-wpa mode, disable renaming of debug sections.
3030 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
3032 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
3033 descriptions of various incorrectly documented functions.
3035 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
3038 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
3039 address check not strict.
3041 2018-05-30 Richard Biener <rguenther@suse.de>
3043 PR tree-optimization/85964
3044 * tracer.c (better_p): Drop initialized count check, we only
3045 call the function with initialized counts now.
3046 (find_best_successor): Do find a best edge if one
3047 has uninitialized count.
3048 (find_best_predecessor): Likewise. Do BB frequency check only
3049 if count is initialized.
3051 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
3053 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
3054 (aarch64_ldrstr_offset_compare): New.
3055 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
3056 load/store orderings.
3057 (aarch64_gen_adjusted_ldpstp): Likewise.
3059 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
3061 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
3062 Check for subset of GENERAL_REGS and FP_REGS.
3063 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
3066 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
3068 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
3069 and wi::to_poly_offset. Add the current offset and then check
3070 whether the sum fits, rather than using an unchecked addition of
3071 a checked term. Check for a shwi rather than a uhwi.
3072 * expr.c (get_bit_range): Use tree_to_poly_uint64.
3073 (store_constructor): Use poly_int_tree_p.
3074 (expand_expr_real_1): Likewise.
3075 * function.c (assign_temp): Likewise.
3076 * fold-const.c (const_binop): Use poly_int_tree_p and
3078 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
3080 * ipa-icf-gimple.c (func_checker::compare_operand): Use
3081 to_poly_offset for MEM offsets.
3082 * ipa-icf.c (sem_variable::equals): Likewise.
3083 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
3084 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
3085 wi::to_poly_offset for BIT_FIELD_REF offsets.
3086 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
3088 * var-tracking.c (emit_note_insn_var_location): Use
3089 tree_to_poly_uint64.
3091 2018-05-29 Jim Wilson <jimw@sifive.com>
3093 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
3095 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
3098 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
3099 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
3101 (sse4_1_round<mode>2): Use nonimmediate_operand
3102 for operand 1 predicate.
3104 2018-05-29 Martin Sebor <msebor@redhat.com>
3105 Richard Biener <rguenther@suse.de>
3108 * calls.c (get_size_range): Call determine_value_range instead
3109 of get_value_range..
3110 * tree-vrp.h (determine_value_range): Declared new function.
3111 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
3113 2018-05-29 Richard Biener <rguenther@suse.de>
3115 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
3116 sure to use non-pattern stmts for get_earlier_stmt arguments.
3117 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
3118 called on pattern stmts.
3119 (get_later_stmt): Likewise.
3121 2018-05-29 Martin Liska <mliska@suse.cz>
3123 PR gcov-profile/85759
3124 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
3127 2018-05-29 Jakub Jelinek <jakub@redhat.com>
3129 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
3131 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
3135 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
3136 VEC_PACK_FLOAT_EXPR): New tree codes.
3137 * tree-pretty-print.c (op_code_prio): Handle
3138 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
3139 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
3140 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
3141 * tree-inline.c (estimate_operator_cost): Likewise.
3142 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
3143 * fold-const.c (const_binop): Likewise.
3144 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
3145 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
3146 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
3147 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
3148 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
3149 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
3150 * expr.c (expand_expr_real_2): Likewise.
3151 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
3152 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
3153 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
3155 * optabs.c (expand_widen_pattern_expr): For
3156 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
3157 sign from result type rather than operand's type.
3158 (expand_binop_directly): For vec_packu_float_optab and
3159 vec_packs_float_optab allow result type to be different from operand's
3161 * optabs-tree.c (optab_for_tree_code): Handle
3162 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
3163 VEC_PACK_FLOAT_EXPR. Formatting fixes.
3164 * tree-vect-generic.c (expand_vector_operations_1): Handle
3165 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
3166 VEC_PACK_FLOAT_EXPR.
3167 * tree-vect-stmts.c (supportable_widening_operation): Handle
3169 (supportable_narrowing_operation): Handle FLOAT_EXPR.
3170 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
3171 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
3172 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
3173 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
3175 (vec_pack<floatprefix>_float_<mode>): New expander.
3176 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
3178 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
3179 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
3180 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
3181 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
3182 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
3184 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
3185 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
3186 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
3187 VEC_PACK_FLOAT_EXPR): Document.
3189 2018-05-29 Richard Biener <rguenther@suse.de>
3191 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
3193 (stmt_vec_info_vec): Make pointer.
3194 (init_stmt_vec_info_vec): Remove.
3195 (free_stmt_vec_info_vec): Likewise.
3196 (set_stmt_vec_info_vec): New function.
3197 (free_stmt_vec_infos): Likewise.
3198 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
3199 (set_vinfo_for_stmt): Likewise.
3200 (get_earlier_stmt): Likewise.
3201 (get_later_stmt): Likewise.
3202 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
3203 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
3204 (vec_info::~vec_info): Free stmt_vec_infos.
3205 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
3206 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
3207 (pass_slp_vectorize::execute): Likewise.
3208 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
3209 (free_stmt_vec_info_vec): Likewise.
3210 (set_stmt_vec_info_vec): New function.
3211 (free_stmt_vec_infos): Likewise.
3212 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
3213 the global stmt_vec_info_vec.
3214 * tree-parloops.c (gather_scalar_reductions): Use
3215 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
3218 2018-05-29 Richard Biener <rguenther@suse.de>
3220 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
3222 2018-05-29 Martin Liska <mliska@suse.cz>
3223 David Malcolm <dmalcolm@redhat.com>
3225 * vec.c (test_reverse): New.
3226 (vec_c_tests): Add new test.
3227 * vec.h (vl_ptr>::reverse): New function.
3229 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
3231 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
3233 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
3236 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
3238 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
3240 2018-05-28 Richard Biener <rguenther@suse.de>
3242 PR tree-optimization/85933
3243 * tree-vect-data-refs.c (vect_record_base_alignments): Only
3244 look at stmts marked as vectorizable.
3246 2018-05-28 Richard Biener <rguenther@suse.de>
3248 PR tree-optimization/85934
3249 * tree-vect-generic.c (expand_vector_operations_1): Hoist
3250 vector boolean check before scalar optimization.
3252 2018-05-28 Jakub Jelinek <jakub@redhat.com>
3254 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
3257 2018-05-28 Mark Wielaard <mark@klomp.org>
3259 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
3260 if it is an expression containing a minus sign.
3262 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
3264 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
3266 2018-05-27 Paul Koning <ni1d@arrl.net>
3268 * config/pdp11/pdp11.md (truncsihi2): Remove.
3270 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
3271 Chung-Ju Wu <jasonwucj@gmail.com>
3273 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
3275 (unaligned_store_dw): Ditto.
3276 * config/nds32/nds32-memory-manipulation.c
3277 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
3278 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
3279 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
3280 (emit_setmem_word_loop): Rename to ...
3281 (emit_setmem_doubleword_loop): ... this.
3282 (nds32_gen_dup_4_byte_to_word_value): New function.
3283 (nds32_gen_dup_8_byte_to_double_word_value): New function.
3284 (nds32_expand_setmem_loop): Refine implementation.
3285 (nds32_expand_setmem_loop_v3m): Ditto.
3286 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
3289 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
3291 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
3293 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
3295 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
3296 (nds32_init_machine_status): Initialize machine->attr_naked_p and
3297 machine->attr_no_prologue_p.
3298 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
3299 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
3300 (nds32_expand_epilogue): Consider attr_naked_p.
3301 (nds32_expand_epilogue_v3pop): Likewise.
3302 (nds32_can_use_return_insn): Likewise.
3303 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
3304 attr_no_prologue_p fields.
3305 * config/nds32/nds32.opt (mret-in-naked-func): New option.
3307 2018-05-27 Jakub Jelinek <jakub@redhat.com>
3310 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
3312 * config/i386/sse.md
3313 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
3315 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
3317 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
3319 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
3321 (*<floatsuffix>floatv2div2sf2): Rename to ...
3322 (*float<floatunssuffix>v2div2sf2): ... this.
3323 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
3324 (float<floatunssuffix>v2div2sf2_mask): ... this.
3325 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
3326 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
3327 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
3329 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
3331 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
3333 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
3335 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
3337 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
3339 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
3340 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
3341 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
3342 gen_ufix_truncv8dfv8si2.
3343 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
3344 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
3345 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
3346 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
3347 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
3348 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
3349 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
3350 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
3352 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
3356 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
3358 2018-05-25 Jim Wilson <jimw@sifive.com>
3360 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
3361 * config/riscv/riscv.c (struct machine_function): Add
3362 interrupt_handler_p and attribute_checked_p fields.
3363 (riscv_attribute_table): Add interrupt.
3364 (riscv_interrupt_type_p): New.
3365 (riscv_save_reg_p): Save extra regs for interrupt handler.
3366 (riscv_use_save_libcall): Return false for interrupt handler.
3367 (riscv_first_stack_step): Add forward declaration.
3368 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
3369 for interrupt handler with large frame. Use it for saved reg list.
3370 (riscv_expand_prologue): Move flag_stack_usage_info support to
3371 eliminate duplication.
3372 (riscv_expand_epilogue): Generate mret for interrupt handler.
3373 (riscv_epilogue_uses): New.
3374 (riscv_can_use_return_insn): Return false for interrupt handler.
3375 (riscv_function_ok_for_sibcall): Likewise.
3376 (riscv_set_current_function): Add interrupt handler support.
3377 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
3378 * config/riscv/riscv.md (UNSPECV_MRET): New.
3380 (riscv_frflags, riscv_fsflags): Use tab after opcode.
3382 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
3384 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
3386 PR tree-optimization/85712
3387 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
3388 this candidate has already been replaced in-situ by a copy.
3390 2018-05-25 Jason Merrill <jason@redhat.com>
3392 PR c++/80485 - inline function non-zero address.
3393 * symtab.c (nonzero_address): Check DECL_COMDAT.
3395 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
3398 * config/alpha/alpha.md (ashlsi3): New insn pattern.
3399 (*ashlsi_se): Rename from *ashldi_se. Define as sign
3400 extension of SImode operation. Use const123_operand predicate.
3401 (*saddsi_1): Remove.
3402 (*saddl_se_1): Ditto.
3404 (*ssubl_se_1): Ditto.
3405 * config/alpha/predicates.md (const123_operand): New predicate.
3406 * config/alpha/constraints.md (P): Use IN_RANGE.
3408 2018-05-25 Richard Biener <rguenther@suse.de>
3410 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
3412 (ref_maybe_used_by_stmt_p): Likewise.
3413 (stmt_may_clobber_ref_p): Likewise.
3414 (stmt_may_clobber_ref_p_1): Likewise.
3415 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
3417 (ref_maybe_used_by_stmt_p): Likewise.
3418 (stmt_may_clobber_ref_p): Likewise.
3419 (stmt_may_clobber_ref_p_1): Likewise.
3420 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
3421 the alias oracle to disambiguate DRs with stmts DR analysis
3423 (vect_analyze_data_refs): Do not give up on not analyzable
3424 DRs for BB vectorization. Remove code truncating the dataref
3427 2018-05-25 Jakub Jelinek <jakub@redhat.com>
3430 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
3431 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
3432 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
3434 2018-05-25 Richard Biener <rguenther@suse.de>
3436 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
3437 function, combining stmt data ref gathering and fatal analysis
3439 (vect_analyze_data_refs): Remove now redudnant code and simplify.
3440 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
3441 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
3442 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
3443 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
3445 2018-05-25 Bin Cheng <bin.cheng@arm.com>
3447 PR tree-optimization/85720
3448 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
3449 SCC if all partitions are builtins.
3450 (version_loop_by_alias_check): New parameter. Generate cancelable
3451 runtime alias check if all partitions are builtins.
3452 (distribute_loop): Update call to above function.
3454 2018-05-25 Bin Cheng <bin.cheng@arm.com>
3456 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
3457 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
3458 (parm_default_def_partition_arg): Ditto.
3459 (set_parm_default_def_partition): Ditto.
3460 (get_parm_default_def_partitions): Ditto and make it static.
3461 (get_undefined_value_partitions): Ditto and make it static.
3462 (remove_ssa_form): Refactor call to init_var_map here.
3463 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
3464 computation for loop region.
3465 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
3466 (register_default_def): Delete.
3467 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
3468 (parm_default_def_partition_arg): Ditto.
3469 (set_parm_default_def_partition): Ditto.
3470 (get_parm_default_def_partitions): Ditto and make it static.
3471 (get_undefined_value_partitions): Ditto and make it static.
3472 (coalesce_with_default, coalesce_with_default): Update comment.
3473 (create_coalesce_list_for_region): New func factored out from
3474 create_outofssa_var_map.
3475 (populate_coalesce_list_for_outofssa): New func factored out from
3476 create_outofssa_var_map and coalesce_ssa_name.
3477 (create_outofssa_var_map): Delete.
3478 (coalesce_ssa_name): Refactor to support live range computation.
3479 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
3480 (get_parm_default_def_partitions): Delete.
3481 (get_undefined_value_partitions): Ditto.
3482 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
3483 computation for loop region.
3484 (new_tree_live_info, loe_visit_block): Ditto.
3485 (live_worklist, set_var_live_on_entry): Ditto.
3486 (calculate_live_on_exit, verify_live_on_entry): Ditto.
3487 * tree-ssa-live.h (struct _var_map): New fields.
3488 (init_var_map): Change decl.
3489 (region_contains_p): New.
3491 2018-05-25 Bin Cheng <bin.cheng@arm.com>
3493 * tree-ssa-live.h (live_merge_and_clear): Delete.
3495 2018-05-25 Richard Biener <rguenther@suse.de>
3498 * tree-dump.c (dequeue_and_dump): Remove access to removed
3499 operand 2 of a SWITCH_EXPR.
3501 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3503 * doc/sourcebuild.texi (vect_double_cond_arith): Include
3504 multiplication and division.
3505 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
3506 (cond_udiv@var{m}, cond_umod@var{m}): Document.
3507 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
3508 (cond_udiv_optab, cond_umod_optab): New optabs.
3509 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
3510 (IFN_COND_RDIV): New internal functions.
3511 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
3512 TRUNC_MOD_EXPR and RDIV_EXPR.
3513 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
3514 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
3516 (SVE_INT_BINARY): Include mult.
3517 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
3518 (optab, sve_int_op): Handle mult.
3519 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
3521 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
3522 for SVE_INT_BINARY_SD.
3524 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3526 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
3527 (optab, sve_int_op): Handle div and udiv.
3528 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
3529 for SVE_INT_BINARY_SD.
3530 (*<optab><mode>3): New insn for the same.
3532 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3534 * tree-vect-patterns.c: Include predict.h.
3535 (vect_recog_divmod_pattern): Restrict check for division support
3536 to when optimizing for size.
3538 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3540 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
3541 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
3542 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
3543 (gimple_match_op::set_op): Likewise.
3544 (gimple_resimplify4): Declare.
3545 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
3546 (expr::gen_transform): Likewise.
3547 (decision_tree::gen): Generate a simplification routine for 4 operands.
3548 * gimple-match-head.c (gimple_simplify): Add an overload for
3549 4 operands. In the top-level function, handle up to 4 call
3550 arguments and call gimple_resimplify4.
3551 (gimple_resimplify4): New function.
3552 (build_call_internal): Pass a fourth operand.
3553 (maybe_push_to_seq): Likewise.
3554 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
3555 Fold VEC_COND_EXPRs of an operation and a default value into
3556 an IFN_COND_* function if possible.
3557 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
3559 (SVE_COND_FP_BINARY): Include them.
3560 (optab, sve_fp_op): Handle them.
3561 (SVE_INT_BINARY_REV): New code iterator.
3562 (SVE_COND_FP_BINARY_REV): New int iterator.
3563 (commutative): New int attribute.
3564 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
3566 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
3568 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
3569 (*cond_<optab><mode>): New patterns for reversed operands.
3571 2018-05-25 Richard Biener <rguenther@suse.de>
3573 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
3574 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
3575 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
3576 (STMT_VINFO_GROUPED_ACCESS): Adjust.
3577 * tree-vect-data-refs.c (everywhere): Adjust users.
3578 * tree-vect-loop.c (everywhere): Likewise.
3579 * tree-vect-slp.c (everywhere): Likewise.
3580 * tree-vect-stmts.c (everywhere): Likewise.
3581 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
3583 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3585 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
3587 (gcc_cv_as_section_exclude): ... this.
3588 Try Solaris as #exclude syntax.
3589 * configure: Regenerate.
3590 * config.in: Regenerate.
3591 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
3593 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
3594 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
3596 * varasm.c (default_elf_asm_named_section): Don't check if
3597 HAVE_GAS_SECTION_EXCLUDE is defined.
3599 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3601 * doc/md.texi: Update the documentation of the cond_* optabs
3602 to mention the new final operand. Fix GET_MODE_NUNITS call.
3603 Describe the scalar case too.
3604 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
3605 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
3607 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
3608 (get_conditional_internal_fn): Update comment.
3609 * tree-vect-loop.c (vectorizable_reduction): Pass the original
3610 accumulator value as a final argument to conditional functions.
3611 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
3612 a define_expand and add an "else" operand. Assert for now that
3613 the else operand is equal to operand 2. Use SVE_INT_BINARY and
3614 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
3615 (*cond_<optab><mode>): New patterns.
3616 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
3617 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
3618 (UNSPEC_COND_EOR): Delete.
3619 (optab): Remove associated mappings.
3620 (SVE_INT_BINARY): New code iterator.
3621 (sve_int_op): Remove int attribute and add "minus" to the code
3623 (SVE_COND_INT_OP): Delete.
3624 (SVE_COND_FP_OP): Rename to...
3625 (SVE_COND_FP_BINARY): ...this.
3627 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3629 * optabs.c (can_reuse_operands_p): New function.
3630 (maybe_legitimize_operands): Try to reuse the results for
3633 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
3635 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
3636 Add {q} suffix to insn mnemonic.
3638 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
3640 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
3641 (msp430_warn_func_return): New.
3643 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
3645 * fold-const.c (tree_nonzero_bits): New function.
3646 * fold-const.h (tree_nonzero_bits): Likewise.
3647 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
3648 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
3650 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
3654 * varasm.c (assemble_alias): Check ifunc_resolver only on
3657 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
3660 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
3661 when memory input operand is handled.
3663 2018-05-24 Luis Machado <luis.machado@linaro.org>
3665 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
3667 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
3669 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
3671 * match.pd: Delay FMA folds until after vectorization.
3673 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
3676 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
3677 address check not strict.
3679 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
3681 * gimple-match.h (gimple_match_op): New class.
3682 (mprts_hook): Replace parameters with a gimple_match_op *.
3683 (maybe_build_generic_op): Likewise.
3684 (gimple_simplified_result_is_gimple_val): Replace parameters with
3685 a const gimple_match_op *.
3686 (gimple_simplify): Replace code_helper * and tree * parameters with
3687 a gimple_match_op * parameter.
3688 (gimple_resimplify1): Replace code_helper *, tree and tree *
3689 parameters with a gimple_match_op * parameter.
3690 (gimple_resimplify2): Likewise.
3691 (gimple_resimplify3): Likewise.
3692 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
3693 parameters with a gimple_match_op * parameter.
3694 * gimple-match-head.c (gimple_simplify): Change prototypes of
3695 auto-generated functions to take a gimple_match_op * instead of
3696 separate code_helper * and tree * parameters. Make the same
3697 change in the top-level overload and update calls to the
3698 gimple_resimplify routines. Update calls to the auto-generated
3699 functions and to maybe_push_res_to_seq in the publicly-facing
3700 operation-specific gimple_simplify overloads.
3701 (gimple_match_op::MAX_NUM_OPS): Define.
3702 (gimple_resimplify1): Replace rcode and ops with a single res_op
3703 parameter. Update call to gimple_simplify.
3704 (gimple_resimplify2): Likewise.
3705 (gimple_resimplify3): Likewise.
3706 (mprts_hook): Replace parameters with a gimple_match_op *.
3707 (maybe_build_generic_op): Likewise.
3708 (build_call_internal): Replace type, nargs and ops with
3709 a gimple_match_op *.
3710 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
3711 with a single gimple_match_op *. Update calls to mprts_hook,
3712 build_call_internal and gimple_simplified_result_is_gimple_val.
3713 Factor out code that is common to the tree_code and combined_fn cases.
3714 * genmatch.c (expr::gen_transform): Replace tem_code and
3715 tem_ops with a gimple_match_op called tem_op. Update calls
3716 to the gimple_resimplify functions and maybe_push_res_to_seq.
3717 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
3718 res_ops. Update call to the gimple_resimplify functions.
3719 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
3720 (decision_tree::gen): Make the functions take a gimple_match_op *
3721 called res_op instead of separate res_code and res_ops parameters.
3722 Update call accordingly.
3723 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
3724 and ops with a single res_op parameter. Update calls to
3725 maybe_build_generic_op and maybe_push_res_to_seq.
3726 (fold_stmt_1): Update calls to gimple_simplify and
3727 replace_stmt_with_simplification.
3728 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
3729 and gimple_simplified_result_is_gimple_val.
3730 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
3732 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
3733 with a gimple_match_op *.
3734 (vn_nary_build_or_lookup): Likewise. Update call to
3735 vn_nary_build_or_lookup_1.
3736 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
3737 gimple_match_op *. Update calls to the gimple_resimplify routines
3738 and to gimple_simplified_result_is_gimple_val.
3739 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
3740 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
3741 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
3742 (visit_nary_op): Likewise.
3743 (visit_reference_op_load): Likewise.
3745 2018-05-23 Luis Machado <luis.machado@linaro.org>
3747 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
3748 modifier for printing the step amount.
3750 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
3753 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
3756 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
3758 * doc/sourcebuild.texi (Endianness): New subsubsection.
3760 2018-05-23 Luis Machado <luis.machado@linaro.org>
3762 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3763 <prefetch_dynamic_strides>: New const bool field.
3764 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3765 prefetch_dynamic_strides.
3766 (exynosm1_prefetch_tune): Likewise.
3767 (thunderxt88_prefetch_tune): Likewise.
3768 (thunderx_prefetch_tune): Likewise.
3769 (thunderx2t99_prefetch_tune): Likewise.
3770 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
3772 (aarch64_override_options_internal): Update to set
3773 PARAM_PREFETCH_DYNAMIC_STRIDES.
3774 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
3775 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
3776 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
3777 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
3778 prefetch-dynamic-strides setting.
3780 2018-05-23 Luis Machado <luis.machado@linaro.org>
3782 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3783 <minimum_stride>: New const int field.
3784 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3785 minimum_stride field defaulting to -1.
3786 (exynosm1_prefetch_tune): Likewise.
3787 (thunderxt88_prefetch_tune): Likewise.
3788 (thunderx_prefetch_tune): Likewise.
3789 (thunderx2t99_prefetch_tune): Likewise.
3790 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
3791 <default_opt_level>: Set to 3.
3792 (aarch64_override_options_internal): Update to set
3793 PARAM_PREFETCH_MINIMUM_STRIDE.
3794 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
3795 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
3796 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
3797 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
3798 stride is constant and is below the minimum stride threshold.
3800 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3802 * config/arm/arm-cpus.in (mode26): Delete.
3803 (armv4): Delete mode26 reference.
3804 * config/arm/arm.c (arm_configure_build_target): Delete use of
3807 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
3809 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
3811 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
3812 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
3814 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
3815 (floatunsdidf2): Ditto.
3817 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
3819 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
3820 (fixuns_trunc<mode>si2_avx512f): Ditto.
3821 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
3822 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
3823 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
3825 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
3827 PR rtl-optimization/79985
3828 * df-scan.c (df_insn_refs_collect): Remove special case for
3829 global registers and asm statements.
3831 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
3833 * extend.texi (Global Register Variables): Rewrite the bullet list.
3834 Note that the register is available for allocation. Note that access
3835 via inline asm must use constraints. Add note about async-signal
3836 handlers. Remove paragraph about automagic register selection.
3838 2018-05-23 Richard Biener <rguenther@suse.de>
3840 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
3841 of fixed offset from memset VN.
3843 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
3845 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
3847 (alloc_cand_and_find_basis): Initialize first_interp field.
3848 (slsr_process_mul): Modify first_interp field.
3849 (slsr_process_add): Likewise.
3850 (slsr_process_cast): Modify first_interp field for each new
3852 (slsr_process_copy): Likewise.
3853 (dump_candidate): Dump first_interp field.
3854 (replace_mult_candidate): Process all interpretations, not just
3856 (replace_rhs_if_not_dup): Likewise.
3857 (replace_one_candidate): Likewise.
3859 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
3861 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
3863 (aarch64_needs_frame_chain): New function.
3864 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
3866 2018-05-23 Sudakshina Das <sudi.das@arm.com>
3869 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
3870 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
3871 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
3872 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
3873 as true for strict-align.
3874 (aarch64_can_inline_p): Perform checks even when callee has no
3875 attributes to check for strict alignment.
3876 * doc/extend.texi (AArch64 Function Attributes): Document
3878 * doc/invoke.texi: (AArch64 Options): Likewise.
3880 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
3882 PR tree-optimization/85853
3883 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
3884 the handling of the root of the node to...
3885 (vect_slp_analyze_node_operations_1): ...this new function,
3886 and run the whole thing with the child nodes' def types
3887 set according to their SLP node's def type.
3889 2018-05-23 Richard Biener <rguenther@suse.de>
3892 * tree-data-ref.c (create_runtime_alias_checks): Defer
3893 and ignore overflow warnings.
3895 2018-05-23 Yury Gribov <tetra2005@gmail.com>
3897 PR tree-optimization/85822
3898 * tree-vrp.c (is_masked_range_test): Fix handling of negative
3901 2018-05-23 Richard Biener <rguenther@suse.de>
3903 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
3904 memset constants via native_interpret_expr.
3906 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
3909 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
3911 (cgraph_node::create_alias): Likewise.
3912 (cgraph_node::get_availability): Check ifunc_resolver instead
3913 of looking up ifunc attribute.
3914 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
3915 * varasm.c (do_assemble_alias): Likewise.
3916 (assemble_alias): Likewise.
3917 (default_binds_local_p_3): Likewise.
3918 * cgraph.h (cgraph_node): Add ifunc_resolver.
3919 (cgraph_node::only_called_directly_or_aliased_p): Return false
3921 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
3923 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
3924 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
3925 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
3926 instead of looking up ifunc attribute.
3928 2018-05-22 Luis Machado <luis.machado@linaro.org>
3930 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
3932 2018-05-22 Martin Sebor <msebor@redhat.com>
3935 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
3936 only when expasion succeeds.
3937 (expand_builtin_strcmp): Same.
3938 (expand_builtin_strncmp): Same.
3940 2018-05-22 Martin Sebor <msebor@redhat.com>
3942 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
3944 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
3945 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3947 * config/aarch64/aarch64-ldpstp.md: Replace uses of
3948 aarch64_mem_pair_operand with memory_operand and delete operand swapping
3950 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
3951 Add check for legitimate_address.
3952 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
3953 (aarch64_swap_ldrstr_operands): New.
3954 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
3957 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
3958 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3960 * config/aarch64/aarch64.md: New patterns to generate stp
3962 (store_pair_sw, store_pair_dw): New patterns to generate stp for
3963 single words and double words.
3964 (load_pair_sw, load_pair_dw): Likewise.
3965 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
3967 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
3969 * config/aarch64/aarch64-ldpstp.md: Modify peephole
3970 for different mode ldpstp and add peephole for merged zero stores.
3972 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
3974 (aarch64_gen_store_pair): Rename calls to match new patterns.
3975 (aarch64_gen_load_pair): Rename calls to match new patterns.
3976 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
3977 (load_pair<DREG:mode><DREG2:mode>): ... This.
3978 (store_pair<mode>): Rename to...
3979 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
3980 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
3982 (V_INT_EQUIV): Handle SImode.
3983 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
3986 2018-05-22 Martin Sebor <msebor@redhat.com>
3989 * calls.c (maybe_warn_nonstring_arg): Use string length to set
3990 or ajust the presumed bound on an operation to avoid unnecessary
3993 2018-05-22 Martin Sebor <msebor@redhat.com>
3995 PR tree-optimization/85826
3996 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
3997 assuming that a DECL necesarily has a constant size.
3999 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
4002 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
4004 2018-05-22 Richard Biener <rguenther@suse.de>
4006 PR tree-optimization/85834
4007 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
4008 non-constant and non-zero memset arguments.
4010 2018-05-22 Martin Liska <mliska@suse.cz>
4013 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
4015 2018-05-22 Richard Biener <rguenther@suse.de>
4017 PR tree-optimization/85863
4018 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
4019 comparisons when vectype is specified.
4020 (vectorizable_condition): Do not specify vectype for
4021 vect_is_simple_cond when SLP vectorizing.
4023 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
4026 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
4027 define __ibm128 as long double.
4028 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
4029 as a distinct type when IEEE 128-bit support is enabled.
4030 (init_float128_ieee): Fix up conversions between IFmode and IEEE
4031 128-bit types to use the correct functions.
4032 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
4033 convert between 128-bit floating point types that have different
4034 modes but the same representation, instead of using gen_lowpart to
4036 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
4038 (IFKF_reg): New attributes to give the register constraints for
4040 (extend<mode>tf2_internal): New insns to mark an explicit
4041 conversion between 128-bit floating point types that have a
4042 different mode but share the same representation.
4044 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
4046 PR tree-optimization/85814
4047 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
4048 a null return from get_strinfo when unsharing the next
4049 strinfo in the chain.
4051 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
4054 * varasm.c (weak_finish): Clean up weak_decls.
4056 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4058 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
4059 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
4060 UNSPEC_UADALP values.
4061 * config/aarch64/iterators.md (ABAL): New int iterator.
4064 (sur): Add mappings for the above.
4065 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
4067 (aarch64_<sur>abal<mode>_4): Likewise.
4068 (aarch64_<sur>adalp<mode>_3): Likewise.
4069 (<sur>sadv16qi): New define_expand.
4071 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
4073 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
4074 (*movdf_internal): Ditto.
4075 (*rcpsf2_sse): Ditto.
4076 (*rsqrtsf2_sse): Ditto.
4077 (*sqrt<mode>2_sse): Ditto.
4079 2018-05-21 Tamar Christina <tamar.christina@arm.com>
4081 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
4083 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
4084 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
4085 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
4086 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
4088 * config/aarch64/arm_neon.h: Likewise.
4089 * config/aarch64/iterators.md (VQ_I): New.
4091 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
4093 * config.gcc: Add arc/t-multilib-linux to tmake_file for
4095 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
4098 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
4100 * config/nds32/constraints.md (S): New constraint.
4101 * config/nds32/nds32.md (call_internal): Use constraint S.
4102 (call_value_internal): Likewise.
4103 (sibcall_internal): Likewise.
4104 (sibcall_value_internal): Likewise.
4106 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
4107 Chung-Ju Wu <jasonwucj@gmail.com>
4109 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
4112 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
4113 Chung-Ju Wu <jasonwucj@gmail.com>
4115 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
4116 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
4117 (nds32_rtx_costs_impl): Simplify.
4118 (nds32_address_cost_impl): Simplify.
4119 (nds32_init_rtx_costs): New function.
4120 (nds32_rtx_costs_speed_prefer): Likewise.
4121 (nds32_rtx_costs_size_prefer): Likewise.
4122 (nds32_address_cost_speed_prefer): Likewise.
4123 (nds32_address_cost_speed_fwprop): Likewise.
4124 (nds32_address_cost_size_prefer): Likewise.
4125 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
4126 * config/nds32/nds32.c (nds32_option_override): Use
4127 nds32_init_rtx_costs function.
4129 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
4131 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
4132 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
4133 (TARGET_PIPELINE_N8): Likewise.
4134 (TARGET_PIPELINE_N10): Likewise.
4135 (TARGET_PIPELINE_N13): Likewise.
4136 (TARGET_PIPELINE_GRAYWOLF): Likewise.
4138 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
4140 * config/nds32/nds32-fpu.md: Update copyright year.
4142 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
4144 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
4146 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
4148 * config/nds32/nds32.c
4149 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
4150 * config/nds32/nds32.opt (minline-asm-r15): New option.
4152 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
4154 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
4156 * config/nds32/nds32.md (abssi2): New pattern.
4158 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
4160 * config/i386/i386.md (rex64namesuffix): New mode attribute.
4161 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
4162 Merge insn pattern from sse_cvtsi2ss<round_name> and
4163 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
4164 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
4165 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
4166 using SWI48 mode iterator.
4167 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
4168 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
4169 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
4170 pattern from sse_cvttss2si<round_saeonly_name>
4171 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
4172 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
4173 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
4174 using SWI48 mode iterator.
4175 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
4176 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
4177 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
4178 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
4179 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
4180 using SWI48 mode iterator.
4181 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
4182 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
4183 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
4184 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
4185 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
4186 SWI48 mode iterator.
4187 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
4188 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
4189 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
4190 pattern from sse_cvttsd2si<round_saeonly_name>
4191 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
4193 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
4195 * config/nds32/nds32-md-auxiliary.c
4196 (nds32_valid_smw_lwm_base_p): Refine.
4197 (nds32_output_smw_single_word): Refine.
4198 (nds32_output_smw_double_word): New.
4199 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
4201 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
4203 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
4204 (nds32_output_stack_pop): Refine.
4205 (nds32_expand_unaligned_load): Refine.
4206 (nds32_expand_unaligned_store): Refine.
4208 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
4209 Chung-Ju Wu <jasonwucj@gmail.com>
4211 * config/nds32/constants.md: Add TP_REGNUM constant.
4212 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
4213 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
4215 * config/nds32/nds32-doubleword.md: Consider flag_pic.
4216 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
4217 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
4218 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
4219 and PIC code generation.
4220 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
4222 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
4224 * config/nds32/nds32.md: Support TLS and PIC.
4225 * config/nds32/nds32.c: Support TLS and PIC.
4226 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
4227 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
4230 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
4232 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
4233 mode with E_ prefix.
4235 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
4236 Chung-Ju Wu <jasonwucj@gmail.com>
4238 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
4239 * config/nds32/nds32-md-auxiliary.c
4240 (symbolic_reference_mentioned_p): New.
4241 (nds32_legitimize_ict_address): New.
4242 (nds32_expand_ict_move): New.
4243 (nds32_indirect_call_referenced_p): New.
4244 (nds32_symbol_binds_local_p): Delete.
4245 (nds32_long_call_p): Modify.
4246 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
4247 * config/nds32/nds32-protos.h
4248 (symbolic_reference_mentioned_p): Declare.
4249 (nds32_legitimize_ict_address): Declare.
4250 (nds32_expand_ict_move): Declare.
4251 (nds32_indirect_call_referenced_p): Declare.
4252 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
4253 (nds32_relax_group): Use nds32_ict_const_p as condition.
4254 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
4255 (nds32_asm_file_start): Output ict_model directive in asm code.
4256 (nds32_legitimate_address_p): Consider indirect call.
4257 (nds32_print_operand): Consider indirect call.
4258 (nds32_print_operand_address): Consider indirect call.
4259 (nds32_insert_attributes): Handle "indirect_call" attribute.
4260 (TARGET_LEGITIMATE_ADDRESS_P): Define.
4261 (TARGET_LEGITIMATE_CONSTANT_P): Define.
4262 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
4263 (TARGET_DELEGITIMIZE_ADDRESS): Define.
4264 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
4265 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
4266 (TARGET_ICT_MODEL_SMALL): Define.
4267 (TARGET_ICT_MODEL_LARGE): Define.
4268 * config/nds32/nds32.md (movsi): Consider ict model.
4269 (call, call_value): Consider ict model.
4270 (sibcall, sibcall_value): Consider ict model.
4271 * config/nds32/nds32.opt (mict-model): New option.
4272 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
4275 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
4276 Monk Chiang <sh.chiang04@gmail.com>
4277 Jim Wilson <jimw@sifive.com>
4279 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
4280 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
4281 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
4282 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
4283 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
4284 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
4285 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
4286 compute save_libcall_adjustment properly.
4287 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
4288 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
4289 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
4290 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
4291 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
4292 (ABI_SPEC): Handle mabi=ilp32e.
4293 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
4294 (RVE): Add RVE mask.
4295 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
4296 <-march>: Add rv32e as an example.
4298 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
4301 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
4302 (intra_create_variable_infos): Handle C++ constructors.
4304 2018-05-18 Martin Liska <mliska@suse.cz>
4306 * passes.def: Remove a redundant pass.
4308 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
4311 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
4313 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4315 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
4317 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
4319 (armv2, armv2a, armv3, armv3m): Delete architectures.
4320 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
4321 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
4322 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
4324 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
4325 (*mulsidi3adddi): Likewise.
4326 (mulsidi3): Likewise.
4327 (*mulsidi3_nov6): Likewise.
4328 (umulsidi3): Likewise.
4329 (umulsidi3_nov6): Likewise.
4330 (umaddsidi4): Likewise.
4331 (*umulsidi3adddi): Likewise.
4332 (smulsi3_highpart): Likewise.
4333 (*smulsi3_highpart_nov6): Likewise.
4334 (umulsi3_highpart): Likewise.
4335 (*umulsi3_highpart_nov6): Likewise.
4336 * config/arm/arm.h (arm_arch3m): Delete.
4337 * config/arm/arm.c (arm_arch3m): Delete.
4338 (arm_option_override_internal): Update armv3-related comment.
4339 (arm_configure_build_target): Delete use of isa_bit_mode32.
4340 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
4341 (arm_rtx_costs_internal): Delete check of arm_arch3m.
4342 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
4344 (mulusa3): Likewise.
4345 * config/arm/arm-protos.h (arm_arch3m): Delete.
4346 * config/arm/arm-tables.opt: Regenerate.
4347 * config/arm/arm-tune.md: Likewise.
4348 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
4349 deleted architectures.
4351 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4353 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
4354 (armv5t, armv5te): New features.
4355 (ARMv5, ARMv5e): Delete fgroups.
4356 (ARMv5t, ARMv5te): Adjust for above changes.
4358 (armv5, armv5e): Delete arches.
4359 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
4361 (*call_reg_arm): Likewise.
4362 (*call_value_reg_armv5): Likewise.
4363 (*call_value_reg_arm): Likewise.
4364 (*call_symbol): Likewise.
4365 (*call_value_symbol): Likewise.
4366 (*sibcall_insn): Likewise.
4367 (*sibcall_value_insn): Likewise.
4369 (prefetch): Likewise.
4370 (define_split and define_peephole2 dependent on arm_arch5):
4372 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
4374 (TARGET_ARM_QBIT): Likewise.
4375 (TARGET_DSP_MULTIPLY): Likewise.
4376 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
4377 (arm_arch5, arm_arch5e): Delete.
4378 (arm_arch5t, arm_arch5te): Declare.
4379 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
4380 (arm_arch5t): Declare.
4381 (arm_option_reconfigure_globals): Update for the above.
4382 (arm_options_perform_arch_sanity_checks): Update comment, replace
4383 use of arm_arch5 with arm_arch5t.
4384 (use_return_insn): Likewise.
4385 (arm_emit_call_insn): Likewise.
4386 (output_return_instruction): Likewise.
4387 (arm_final_prescan_insn): Likewise.
4388 (arm_coproc_builtin_available): Likewise.
4389 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
4390 arm_arch5e with arm_arch5t and arm_arch5te.
4391 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
4392 (arm_arch5t, arm_arch5te): Declare.
4393 * config/arm/arm-tables.opt: Regenerate.
4394 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
4395 * config/arm/t-multilib: Likewise.
4396 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
4397 instead of arm_arch5.
4398 (*call_reg_thumb1): Likewise.
4399 (*call_value_reg_thumb1_v5): Likewise.
4400 (*call_value_reg_thumb1): Likewise.
4401 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
4403 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
4405 2018-05-18 Martin Liska <mliska@suse.cz>
4407 PR gcov-profile/84846
4408 * doc/gcov.texi: Document -t option of gcov tool.
4410 2018-05-18 Martin Liska <mliska@suse.cz>
4412 PR gcov-profile/84846
4413 * gcov.c (print_usage): Add new -t option.
4414 (process_args): Handle the option.
4415 (generate_results): Use stdout as output when requested by
4418 2018-05-18 Martin Liska <mliska@suse.cz>
4420 PR gcov-profile/84846
4421 * coverage.c (coverage_init): Write PWD to .gcno file.
4422 * doc/gcov.texi: Document how working directory is printed.
4423 * gcov-dump.c (dump_gcov_file): Print PWD.
4424 * gcov.c (output_intermediate_file): Likewise.
4425 (read_graph_file): Read PWD string.
4426 (output_lines): Print PWD.
4428 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
4431 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
4432 for retval and return false if all args to phi are zero.
4434 2018-05-18 Richard Biener <rguenther@suse.de>
4436 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
4438 (evrp_dom_walker::before_dom_children): Call it.
4440 2018-05-18 Richard Biener <rguenther@suse.de>
4442 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
4443 results when processing array refs with variable index.
4445 2018-05-18 Toon Moene <toon@moene.org>
4447 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
4448 directly after that of -floop-interchange. Indicate that both
4449 options are enabled by default when specifying -O3.
4451 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4453 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
4454 iterator. Delete separate integer-mode vec_set<mode> expander.
4455 (aarch64_simd_vec_setv2di): Delete.
4456 (vec_setv2di): Delete.
4457 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
4458 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
4459 the "w, r" alternative.
4461 2018-05-18 Martin Liska <mliska@suse.cz>
4463 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
4464 * tree-pass.h (make_pass_lower_switch_O0): New function.
4465 * tree-switch-conversion.c (node_has_low_bound): Remove.
4466 (node_has_high_bound): Likewise.
4467 (node_is_bounded): Likewise.
4468 (class pass_lower_switch): Make it a template type and create
4470 (pass_lower_switch::execute): Add template argument.
4471 (make_pass_lower_switch): New function.
4472 (make_pass_lower_switch_O0): New function.
4473 (do_jump_if_equal): Remove.
4474 (emit_case_nodes): Simplify to just handle all 3 cases and leave
4475 all the hard work to tree optimization passes.
4477 2018-05-18 Martin Liska <mliska@suse.cz>
4479 * dbgcnt.c (limit_low): Renamed from limit.
4480 (limit_high): New variable.
4481 (dbg_cnt_is_enabled): Check for upper limit.
4482 (dbg_cnt): Adjust dumping.
4483 (dbg_cnt_set_limit_by_index): Add new argument for high
4485 (dbg_cnt_set_limit_by_name): Likewise.
4486 (dbg_cnt_process_single_pair): Parse new format.
4487 (dbg_cnt_process_opt): Use strtok.
4488 (dbg_cnt_list_all_counters): Remove 'value' and add
4490 * doc/invoke.texi: Document changes.
4492 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
4494 * doc/sourcebuild.texi (scalar_all_fma): Document.
4495 * tree.def (FMA_EXPR): Delete.
4496 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
4497 * internal-fn.c (ternary_direct): New macro.
4498 (expand_ternary_optab_fn): Likewise.
4499 (direct_ternary_optab_supported_p): Likewise.
4500 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
4501 * builtins.c (fold_builtin_fma): Delete.
4502 (fold_builtin_3): Don't call it.
4503 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
4504 * expr.c (expand_expr_real_2): Likewise.
4505 * fold-const.c (operand_equal_p): Likewise.
4506 (fold_ternary_loc): Likewise.
4507 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
4508 * gimple.c (DEFTREECODE): Likewise.
4509 * gimplify.c (gimplify_expr): Likewise.
4510 * optabs-tree.c (optab_for_tree_code): Likewise.
4511 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
4512 * tree-eh.c (operation_could_trap_p): Likewise.
4513 (stmt_could_throw_1_p): Likewise.
4514 * tree-inline.c (estimate_operator_cost): Likewise.
4515 * tree-pretty-print.c (dump_generic_node): Likewise.
4516 (op_code_prio): Likewise.
4517 * tree-ssa-loop-im.c (stmt_cost): Likewise.
4518 * tree-ssa-operands.c (get_expr_operands): Likewise.
4519 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
4520 * fold-const-call.h (fold_fma): Delete.
4521 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
4522 CFN_FNMA and CFN_FNMS.
4524 * genmatch.c (combined_fn): New enum.
4525 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
4526 (commutative_op): New function.
4527 (commutate): Use it. Handle more than 2 operands.
4528 (dt_operand::gen_gimple_expr): Use commutative_op.
4529 (parser::parse_expr): Allow :c to be used with non-binary
4530 operators if the commutative operand is known.
4531 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
4532 CFN_FMS, CFN_FNMA and CFN_FNMS.
4533 (backprop::process_assign_use): Remove FMA_EXPR handling.
4534 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
4535 (gen_hsa_fma): New function.
4536 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
4537 IFN_FNMA and IFN_FNMS.
4538 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
4539 * gimple-fold.h (follow_all_ssa_edges): Declare.
4540 * gimple-fold.c (follow_all_ssa_edges): New function.
4541 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
4542 gimple_build interface and use follow_all_ssa_edges to fold the result.
4543 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
4544 instead of checking for optabs directly.
4545 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
4546 rather than FMA_EXPRs.
4547 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
4548 call to IFN_FMA instead of an FMA_EXPR.
4550 2018-05-17 Jim Wilson <jimw@sifive.com>
4552 * expr.c (do_tablejump): When converting index to Pmode, if we have a
4553 sign extended promoted subreg, and the range does not have the sign bit
4554 set, then do a sign extend.
4556 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
4557 test, check for sign extended subreg and/or constant operands, and
4558 do a sign extend in that case.
4560 2018-05-17 Steve Ellcey <sellcey@cavium.com>
4562 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
4563 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
4565 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
4566 Change logics_shift_reg to logics_shift_imm.
4567 (thunderx2t99_fp_loadpair_basic): Delete.
4568 (thunderx2t99_fp_storepair_basic): Delete.
4569 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
4570 (thunderx2t99_asimd_polynomial): Delete.
4571 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
4572 and neon_fp_mul_d_scalar_q.
4573 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
4574 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
4575 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
4576 (thunderx2t99_asimd_lut): Add missing tbl types.
4577 (thunderx2t99_asimd_ext): Delete.
4578 (thunderx2t99_asimd_load1_1_mult): Delete.
4579 (thunderx2t99_asimd_load1_2_mult): Delete.
4580 (thunderx2t99_asimd_load1_ldp): New.
4581 (thunderx2t99_asimd_load1): New.
4582 (thunderx2t99_asimd_load2): Add missing *load2* types.
4583 (thunderx2t99_asimd_load3): New.
4584 (thunderx2t99_asimd_load4): New.
4585 (thunderx2t99_asimd_store1_1_mult): Delete.
4586 (thunderx2t99_asimd_store1_2_mult): Delete.
4587 (thunderx2t99_asimd_store2_mult): Delete.
4588 (thunderx2t99_asimd_store2_onelane): Delete.
4589 (thunderx2t99_asimd_store_stp): New.
4590 (thunderx2t99_asimd_store1): New.
4591 (thunderx2t99_asimd_store2): New.
4592 (thunderx2t99_asimd_store3): New.
4593 (thunderx2t99_asimd_store4): New.
4595 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
4597 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
4598 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
4600 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
4601 Segher Boessenkool <segher@kernel.crashing.org>
4604 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
4607 2018-05-17 Richard Biener <rguenther@suse.de>
4609 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
4610 for pruning loop and prune defs feeding only already visited PHIs.
4612 2018-05-17 Richard Biener <rguenther@suse.de>
4614 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
4616 2018-05-17 Bin Cheng <bin.cheng@arm.com>
4617 Richard Biener <rguenther@suse.de>
4619 PR tree-optimization/85793
4620 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
4621 for VMAT_ELEMENTWISE.
4623 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
4625 * internal-fn.h (lookup_internal_fn): Declare
4626 * internal-fn.c (lookup_internal_fn): New function.
4627 * gimple.c (gimple_build_call_from_tree): Handle calls to
4629 * gimple-pretty-print.c (dump_gimple_call): Print "." before
4630 internal function names.
4631 * tree-pretty-print.c (dump_generic_node): Likewise.
4632 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
4634 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
4636 * gimple-fold.h (gimple_build): Make the function forms take
4637 combined_fn rather than built_in_function.
4638 (gimple_simplify): Likewise.
4639 * gimple-match-head.c (gimple_simplify): Likewise.
4640 * gimple-fold.c (gimple_build): Likewise.
4641 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
4642 rather than gimple_build_call_internal.
4643 (get_initial_defs_for_reduction): Likewise.
4644 (vect_create_epilog_for_reduction): Likewise.
4645 (vectorizable_live_operation): Likewise.
4647 2018-05-17 Martin Liska <mliska@suse.cz>
4649 * gimple-ssa-sprintf.c (format_directive): Do not use
4650 space in between 'G_' and '('.
4652 2018-05-17 Jakub Jelinek <jakub@redhat.com>
4655 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
4656 even if the mask is not all ones.
4659 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
4661 (ix86_gimple_fold_builtin): Likewise.
4664 * config/i386/i386.c: Include tree-vector-builder.h.
4665 (ix86_vector_shift_count): New function.
4666 (ix86_fold_builtin): Fold shift builtins by scalar count.
4667 (ix86_gimple_fold_builtin): Likewise.
4669 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
4670 _mm512_setzero): New intrinsics.
4672 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
4673 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4675 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
4676 code generation for cases where splatting a value is not useful.
4677 * simplify-rtx.c (simplify_ternary_operation): Simplify
4678 vec_merge across a vec_duplicate and a paradoxical subreg forming
4679 a vector mode to a vec_concat.
4681 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
4683 * config.gcc: Support "goldmont-plus".
4684 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
4686 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
4687 PROCESSOR_GOLDMONT_PLUS.
4688 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
4689 (processor_target_table): Add "goldmont-plus".
4690 (PTA_GOLDMONT_PLUS): Define.
4691 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
4692 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
4693 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
4694 (fold_builtin_cpu): Add "goldmont-plus".
4695 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
4696 (ix86_option_override_internal): Add "goldmont-plus".
4697 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
4698 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
4699 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
4700 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
4702 2018-05-17 Richard Biener <rguenther@suse.de>
4704 PR tree-optimization/85757
4705 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
4706 remove defs that only feed that PHI from further processing.
4708 2018-05-16 Jim Wilson <jimw@sifive.com>
4710 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
4712 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
4714 2018-05-16 Mark Wielaard <mark@klomp.org>
4716 * dwarf2out.c (count_index_strings): New function.
4717 (output_indirect_strings): Call count_index_strings and generate
4718 header for dwarf_version >= 5.
4720 2018-05-16 Mark Wielaard <mark@klomp.org>
4722 * dwarf2out.c (dwarf_FORM): New function.
4723 (set_indirect_string): Use dwarf_FORM.
4724 (reset_indirect_string): Likewise.
4725 (size_of_die): Likewise.
4726 (value_format): Likewise.
4727 (output_die): Likewise.
4728 (add_skeleton_AT_string): Likewise.
4729 (output_macinfo_op): Likewise.
4730 (index_string): Likewise.
4731 (output_index_string_offset): Likewise.
4732 (output_index_string): Likewise.
4733 (count_index_strings): Likewise.
4735 2018-05-16 Carl Love <cel@us.ibm.com>
4737 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
4738 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
4740 2018-05-16 Martin Jambor <mjambor@suse.cz>
4742 * ipa-prop.c (ipa_free_all_edge_args): Remove.
4743 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
4745 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
4747 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
4748 (fnma<mode>4): Likewise.
4749 (fms<mode>4): Likewise.
4750 (fnms<mode>4): Likewise.
4751 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
4752 (aarch64_fnma<mode>4): Likewise.
4753 (aarch64_fms<mode>4): Likewise.
4754 (aarch64_fnms<mode>4): Likewise.
4755 (aarch64_fnmadd<mode>4): Likewise.
4757 2018-05-16 Jason Merrill <jason@redhat.com>
4759 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
4761 2018-05-16 Richard Biener <rguenther@suse.de>
4763 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
4764 (dump_stmt_cost): Declare.
4765 (add_stmt_cost): Dump cost we add.
4766 (add_stmt_costs): New function.
4767 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
4769 (vect_analyze_stmt): Adjust prototype.
4770 (vectorizable_condition): Likewise.
4771 (vectorizable_live_operation): Likewise.
4772 (vectorizable_reduction): Likewise.
4773 (vectorizable_induction): Likewise.
4774 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
4775 cost vector to pass to vectorizable_ and record afterwards.
4776 (vect_model_reduction_cost): Take cost vector argument and adjust.
4777 (vect_model_induction_cost): Likewise.
4778 (vectorizable_reduction): Likewise.
4779 (vectorizable_induction): Likewise.
4780 (vectorizable_live_operation): Likewise.
4781 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
4782 SLP_TREE_NUMBER_OF_VEC_STMTS.
4783 (vect_analyze_slp_cost_1): Remove.
4784 (vect_analyze_slp_cost): Likewise.
4785 (vect_slp_analyze_node_operations): Take visited args and
4786 a target cost vector. Avoid processing already visited stmt sets.
4787 (vect_slp_analyze_operations): Use a local cost vector to gather
4788 costs and register those of non-discarded instances.
4789 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
4790 (vect_schedule_slp_instance): Remove copying of
4791 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
4793 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
4794 adding cost. Record cost entry location.
4795 (vect_prologue_cost_for_slp_op): Function to compute cost of
4796 a constant or invariant generated for SLP vect in the prologue,
4797 split out from vect_analyze_slp_cost_1.
4798 (vect_model_simple_cost): Make static. Adjust for SLP costing.
4799 (vect_model_promotion_demotion_cost): Likewise.
4800 (vect_model_store_cost): Likewise, make static.
4801 (vect_model_load_cost): Likewise.
4802 (vectorizable_bswap): Add cost vector arg and adjust.
4803 (vectorizable_call): Likewise.
4804 (vectorizable_simd_clone_call): Likewise.
4805 (vectorizable_conversion): Likewise.
4806 (vectorizable_assignment): Likewise.
4807 (vectorizable_shift): Likewise.
4808 (vectorizable_operation): Likewise.
4809 (vectorizable_store): Likewise.
4810 (vectorizable_load): Likewise.
4811 (vectorizable_condition): Likewise.
4812 (vectorizable_comparison): Likewise.
4813 (can_vectorize_live_stmts): Likewise.
4814 (vect_analyze_stmt): Likewise.
4815 (vect_transform_stmt): Adjust calls to vectorizable_*.
4816 * tree-vectorizer.c: Include gimple-pretty-print.h.
4817 (dump_stmt_cost): New function.
4819 2018-05-16 Richard Biener <rguenther@suse.de>
4821 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
4822 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
4823 * tree-ssa-dse.c: Include tree-ssa-loop.h.
4824 (check_name): New callback.
4825 (dse_classify_store): Track cycles via a visited bitmap of PHI
4826 defs and simplify handling of in-loop and across loop dead stores
4827 and properly fail for loop-variant refs. Handle byte-tracking with
4828 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
4831 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
4833 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
4834 (vect_get_mask_type_for_stmt): Likewise.
4835 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
4837 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
4838 to determine the statement's vector type and the vector type that
4839 should be used for calculating nunits. Deal with cases in which
4840 the type has to be deferred.
4841 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
4842 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
4843 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
4844 (vect_determine_vf_for_stmt): New functions, split out from...
4845 (vect_determine_vectorization_factor): ...here.
4846 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
4847 (vect_get_mask_type_for_stmt): New functions, split out from
4848 vect_determine_vectorization_factor.
4850 2018-05-16 Richard Biener <rguenther@suse.de>
4852 * tree-cfg.c (verify_gimple_assign_ternary): Properly
4853 verify the [VEC_]COND_EXPR embedded comparison.
4855 2018-05-15 Martin Sebor <msebor@redhat.com>
4857 PR tree-optimization/85753
4858 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
4859 RECORD_TYPE in addition to ARRAY_TYPE.
4861 2018-05-15 Martin Sebor <msebor@redhat.com>
4864 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
4866 2018-05-15 Richard Biener <rguenther@suse.de>
4868 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
4869 add by_clobber_p one. Change algorithm to collect all defs
4870 representing uses we need to walk and try reducing them to
4871 a single one before failing.
4872 (dse_dom_walker::dse_optimize_stmt): Adjust.
4874 2018-05-13 Mark Wielaard <mark@klomp.org>
4876 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
4877 (size_of_loc_descr): Likewise.
4878 (output_loc_operands): Likewise.
4879 (output_loc_operands_raw): Likewise.
4880 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
4881 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
4882 (hash_loc_operands): Likewise.
4883 (compare_loc_operands): Likewise.
4885 2018-05-14 Mark Wielaard <mark@klomp.org>
4887 * dwarf2out.c (count_index_addrs): New function.
4888 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
4890 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
4892 PR tree-optimization/83648
4893 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
4894 return value as malloc candidate.
4896 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
4899 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
4900 param as true in call to suggest_attribute.
4902 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
4904 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
4905 -mreadonly-in-sdata.
4907 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4909 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
4911 (aarch64_crypto_aesd_fused): Likewise.
4913 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
4915 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
4916 (movsi_aarch64): Likewise.
4917 (load_pairsi): Likewise.
4918 (load_pairdi): Likewise.
4919 (store_pairsi): Likewise.
4920 (store_pairdi): Likewise.
4921 (load_pairsf): Likewise.
4922 (load_pairdf): Likewise.
4923 (store_pairsf): Likewise.
4924 (store_pairdf): Likewise.
4925 (zero_extend): Likewise.
4926 (trunc): Swap alternatives.
4927 (fcvt_target): Add '?' to prefer w over r.
4929 2018-05-14 Jakub Jelinek <jakub@redhat.com>
4932 * config/i386/i386.md: Disallow non-commutative arithmetics in
4933 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
4934 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
4935 in the peephole2 before it.
4937 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
4939 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
4940 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
4941 (ix86_handle_option): Handle -mcldemote.
4942 * config.gcc: New header.
4943 * config/i386/cldemoteintrin.h: New file.
4944 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
4945 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
4947 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
4948 OPTION_MASK_ISA_CLDEMOTE.
4949 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
4950 (ix86_valid_target_attribute_inner_p): Ditto.
4951 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
4952 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
4953 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
4954 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
4955 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
4957 * config/i386/i386.opt: Add -mcldemote.
4958 * config/i386/x86intrin.h: New header.
4959 * doc/invoke.texi: Add -mcldemote.
4961 2018-05-14 Richard Biener <rguenther@suse.de>
4963 * doc/match-and-simplify.texi: Adjust :s documentation.
4965 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
4967 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
4968 intended memcpy size.
4969 (REORDER_45): Likewise.
4971 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
4973 * sort.cc: New file.
4974 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
4975 * vec.c (qsort_chk): Use gcc_qsort.
4976 * Makefile.in (OBJS-libcommon): Add sort.o.
4977 (build/sort.o): New target. Use it...
4978 (BUILD_RTL): ... here, and...
4979 (build/gencfn-macros): ... here, and...
4980 (build/genmatch): ... here.
4982 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
4983 Chung-Ju Wu <jasonwucj@gmail.com>
4985 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
4986 * config/nds32/nds32-graywolf.md: New file.
4987 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
4988 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
4990 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
4991 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
4992 * config/nds32/nds32.md (pipeline_model): Add graywolf.
4993 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
4994 * config/nds32/pipelines.md: Include n15 settings.
4996 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
4997 Chung-Ju Wu <jasonwucj@gmail.com>
4999 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
5000 * config/nds32/nds32-n13.md: New file.
5001 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
5002 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
5004 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
5005 * config/nds32/nds32.md (pipeline_model): Add n13.
5006 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
5007 * config/nds32/pipelines.md: Include n13 settings.
5009 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
5010 Chung-Ju Wu <jasonwucj@gmail.com>
5012 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
5013 * config/nds32/nds32-n10.md: New file.
5014 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
5015 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
5017 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
5018 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
5019 * config/nds32/nds32.md (pipeline_model): Add n10.
5020 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
5021 * config/nds32/pipelines.md: Include n10 settings.
5023 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
5024 Kito Cheng <kito.cheng@gmail.com>
5025 Chung-Ju Wu <jasonwucj@gmail.com>
5027 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
5028 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
5029 Add enum values for DSP extension instructions.
5030 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
5032 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
5033 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
5035 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
5036 * config/nds32/nds32-dspext.md: New file for DSP implementation.
5037 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
5038 * config/nds32/nds32-intrinsic.md: Likewise.
5039 * config/nds32/nds32_intrinsic.h: Likewise.
5040 * config/nds32/nds32-md-auxiliary.c: Likewise.
5041 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
5042 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
5043 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
5044 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
5045 * config/nds32/nds32-protos.h: New declarations for DSP extension.
5046 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
5047 TYPE_DMAC in switch statement.
5048 * config/nds32/nds32.c: New checking and implementation for DSP
5049 extension instructions.
5050 * config/nds32/nds32.h: Likewise.
5051 * config/nds32/nds32.md: Likewise.
5052 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
5053 * config/nds32/predicates.md: Implement new predicates for DSP
5056 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
5058 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
5059 Reformat alternatives and attributes so it is easier to identify
5060 which constraints/attributes go with which instruction.
5061 (mov<mode>_hardfloat32, FMOVE64): Likewise.
5062 (mov<mode>_softfloat32, FMOVE64): Likewise.
5063 (mov<mode>_hardfloat64, FMOVE64): Likewise.
5064 (mov<mode>_softfloat64, FMOVE64): Likewise.
5066 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
5068 * doc/extend.texi (PowerPC Built-in Functions): Rename this
5070 (Basic PowerPC Built-in Functions): The new name of the
5071 subsection previously known as "PowerPC Built-in Functions".
5072 (Basic PowerPC Built-in Functions Available on all Configurations):
5074 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
5075 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
5076 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
5077 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
5079 2018-05-11 Martin Jambor <mjambor@suse.cz>
5082 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
5085 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
5088 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
5090 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
5092 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
5093 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
5094 (ix86_handle_option): Handle -mwaitpkg.
5095 * config.gcc: New header.
5096 * config/i386/cpuid.h (bit_WAITPKG): New bit.
5097 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
5098 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
5100 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
5101 OPTION_MASK_ISA_WAITPKG.
5102 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
5103 (ix86_option_override_internal): Add PTA_WAITPKG.
5104 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
5105 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
5106 IX86_BUILTIN_TPAUSE.
5107 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
5108 __builtin_ia32_umwait and __builtin_ia32_tpause.
5109 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
5110 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
5111 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
5112 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
5113 UNSPECV_TPAUSE): New.
5114 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
5115 * config/i386/i386.opt: Add -mwaitpkg.
5116 * config/i386/waitpkgintrin.h: New file.
5117 * config/i386/x86intrin.h: New header.
5118 * doc/invoke.texi: Add -mwaitpkg.
5120 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
5123 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
5125 (cortex-m0): Use armv6s-m isa.
5126 (cortex-m0plus): Likewise.
5127 (cortex-m1): Likewise.
5128 (cortex-m0.small-multiply): Likewise.
5129 (cortex-m0plus.small-multiply): Likewise.
5130 (cortex-m1.small-multiply): Likewise.
5132 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
5133 Jakub Jelinek <jakub@redhat.com>
5135 PR tree-optimization/85692
5136 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
5137 source permute as well.
5139 2018-05-11 Martin Liska <mliska@suse.cz>
5142 * doc/extend.texi: Document LLVM style format for no_sanitize
5145 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
5147 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
5148 mode_supports_vsx_dform_quad to mode_supports_dq_form.
5149 (mode_supports_vsx_dform_quad): Likewise.
5150 (mode_supports_vmx_dform): Move these functions to be next to the
5151 other mode_supports functions.
5152 (mode_supports_dq_form): Likewise.
5153 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
5154 mode_supports_dq_form.
5155 (reg_offset_addressing_ok_p): Likewise.
5156 (offsettable_ok_by_alignment): Likewise.
5157 (rs6000_legitimate_offset_address_p): Likewise.
5158 (legitimate_lo_sum_address_p): Likewise.
5159 (rs6000_legitimize_address): Likewise.
5160 (rs6000_legitimize_reload_address): Likewise.
5161 (rs6000_secondary_reload_inner): Likewise.
5162 (rs6000_preferred_reload_class): Likewise.
5163 (rs6000_output_move_128bit): Likewise.
5165 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
5167 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
5168 Generate SImode target register for null target.
5169 <case IX86_BUILTIN_XGETBV>: Ditto.
5170 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
5171 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
5173 2018-05-10 Carl Love <cel@us.ibm.com>
5175 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
5176 dcbtt and dcbtstt if operands[2] is 0.
5178 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
5181 * config/i386/sse.md (usadv64qi): New expander.
5183 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
5185 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
5186 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
5187 -maltivec=be support.
5188 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
5189 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
5190 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
5191 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
5192 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
5193 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
5194 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
5195 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
5196 altivec_vsumsws): Adjust.
5197 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
5198 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
5199 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
5201 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
5202 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
5203 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
5204 (altivec_lve<VI_char>x): Delete expand.
5205 (*altivec_lve<VI_char>x_internal): Rename to...
5206 (altivec_lve<VI_char>x): ... this.
5207 (altivec_lvxl_<mode>): Delete expand.
5208 (*altivec_lvxl_<mode>_internal): Rename to ...
5209 (altivec_lvxl_<mode>): ... this.
5210 (altivec_stvxl_<mode>): Delete expand.
5211 (*altivec_stvxl_<mode>_internal): Rename to ...
5212 (altivec_stvxl_<mode>): ... this.
5213 (altivec_stve<VI_char>x): Delete expand.
5214 (*altivec_stve<VI_char>x_internal): Rename to ...
5215 (altivec_stve<VI_char>x): ... this.
5216 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
5217 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
5218 reduc_plus_scal_<mode>): Adjust.
5219 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
5221 (rs6000_cpu_cpp_builtins): Adjust.
5222 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
5223 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
5224 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
5225 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
5226 -maltivec=be support.
5227 (rs6000_split_vec_extract_var): Adjust.
5228 (rs6000_split_v4si_init): Adjust.
5229 (swap_selector_for_mode): Delete.
5230 (altivec_expand_lvx_be, altivec_expand_stvx_be,
5231 altivec_expand_stvex_be): Delete.
5232 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
5233 -maltivec=be support.
5234 (rs6000_gimple_fold_builtin): Ditto.
5235 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
5237 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
5238 (TARGET_DIRECT_MOVE_64BIT): Adjust.
5239 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
5240 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
5241 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
5242 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
5243 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
5244 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
5245 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
5246 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
5247 anonymous split): Adjust.
5248 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
5249 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
5251 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
5253 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
5254 when --with-gxx-include-dir is also specified.
5255 * configure: Regenerate.
5257 2018-05-09 Jim Wilson <jimw@sifive.com>
5260 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
5261 * config/riscv/t-withmultilib: New.
5262 * config/riscv/withmultilib.h: New.
5263 * doc/install.texi: Document RISC-V --with-multilib-list support.
5265 2018-05-09 Richard Biener <rguenther@suse.de>
5267 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
5269 (vect_bb_vectorization_profitable_p): Adjust. Compute
5270 actual scalar cost using the cost vector and the add_stmt_cost
5273 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
5275 PR rtl-optimization/85645
5276 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
5277 in the REG_CFA_REGISTER note for LR, don't leave it empty.
5279 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
5281 PR rtl-optimization/85645
5282 * shrink-wrap.c (spread_components): Return a boolean saying if
5283 anything was changed.
5284 (try_shrink_wrapping_separate): Iterate spread_components until
5285 nothing changes anymore.
5287 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
5289 PR rtl-optimization/85645
5290 * regrename.c (build_def_use): Also kill the chains that include the
5291 destination of a REG_CFA_REGISTER note.
5293 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
5295 PR rtl-optimization/85645
5296 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
5297 insn that has a REG_CFA_REGISTER note.
5299 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
5301 * cfgexpand.c (expand_clobber): New function.
5302 (expand_gimple_stmt_1): Use it.
5303 * tree-vect-stmts.c (vect_clobber_variable): New function,
5305 (vectorizable_simd_clone_call): ...here.
5306 (vectorizable_store): Emit a clobber either side of an
5307 IFN_STORE_LANES sequence.
5308 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
5310 2018-05-09 Tom de Vries <tom@codesourcery.com>
5313 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
5314 (define_insn "trap_if_false"): Add exit after trap.
5316 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
5318 PR rtl-optimization/85638
5319 * bb-reorder.c: Include common/common-target.h.
5320 (create_forwarder_block): New function extracted from...
5321 (fix_up_crossing_landing_pad): ...here. Rename into...
5322 (dw2_fix_up_crossing_landing_pad): ...this.
5323 (sjlj_fix_up_crossing_landing_pad): New function.
5324 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
5325 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
5326 from both partitions and exit the loop after one iteration.
5328 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
5331 * doc/extend.texi (PowerPC Built-in Functions): Rename this
5333 (Basic PowerPC Built-in Functions): The new name of the
5334 subsection previously known as "PowerPC Built-in Functions".
5335 (Basic PowerPC Built-in Functions Available on all Configurations):
5337 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
5339 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
5341 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
5343 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
5346 2018-05-08 Jim Wilson <jimw@sifive.com>
5348 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
5349 (LD_EMUL_SUFFIX): New.
5350 (LINK_SPEC): Use it.
5352 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
5354 * doc/extend.texi (PowerPC Built-in Functions): Rename this
5356 (Basic PowerPC Built-in Functions): The new name of the
5357 subsection previously known as "PowerPC Built-in Functions".
5358 (Basic PowerPC Built-in Functions Available on all Configurations):
5360 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
5362 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
5364 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
5366 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
5369 2018-05-08 Jakub Jelinek <jakub@redhat.com>
5372 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
5373 after cmpelim optimization.
5375 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
5377 * config.gcc: Support "goldmont".
5378 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
5379 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
5381 * config/i386/i386.c (m_GOLDMONT): Define.
5382 (processor_target_table): Add "goldmont".
5383 (PTA_GOLDMONT): Define.
5384 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
5385 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
5386 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
5387 (fold_builtin_cpu): Add "goldmont".
5388 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
5389 (ix86_option_override_internal): Add "goldmont".
5390 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
5391 (processor_type): Add PROCESSOR_GOLDMONT.
5392 * config/i386/i386.md: Add CPU "glm".
5393 * config/i386/glm.md: New file.
5394 * config/i386/x86-tune.def: Add m_GOLDMONT.
5395 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
5397 2018-05-08 Jakub Jelinek <jakub@redhat.com>
5400 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
5402 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
5403 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
5404 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
5407 * config/i386/i386.c (ix86_fold_builtin): Handle
5408 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
5411 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
5412 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
5414 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
5417 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
5418 (check_arch): Likewise.
5419 (check_fpu): Return the result rather than printing it.
5420 (end arch): Fix operator precedence.
5421 (end cpu): Likewise.
5422 (END): Print the result from check_fpu.
5424 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
5425 Alan Hayward <alan.hayward@arm.com>
5426 David Sherwood <david.sherwood@arm.com>
5428 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
5429 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
5430 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
5431 (*fcmuo<mode>_and): New patterns.
5433 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
5435 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
5436 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
5437 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
5438 (cmp_op, sve_imm_con): New code attributes.
5439 (SVE_COND_INT_CMP, imm_con): Delete.
5440 (cmp_op): Remove above unspecs from int attribute.
5441 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
5443 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
5444 comparison-specific unspecs.
5445 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
5446 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
5447 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
5448 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
5449 (*vec_fcm<cmp_op><mode>): Rename to...
5450 (*fcm<cmp_op><mode>): ...this and adjust likewise.
5451 (*vec_fcmuo<mode>): Rename to...
5452 (*fcmuo<mode>): ...this and adjust likewise.
5453 (*pred_fcm<cmp_op><mode>): New pattern.
5454 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
5455 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
5457 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
5459 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
5460 (aarch64_emit_sve_predicated_cond): New function.
5461 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
5462 (aarch64_emit_unspec_cond_or): Replace with...
5463 (aarch64_emit_sve_or_conds): ...this new function. Use
5464 aarch64_emit_sve_ptrue_op for the individual comparisons and
5465 aarch64_emit_binop to OR them together.
5466 (aarch64_emit_inverted_unspec_cond): Replace with...
5467 (aarch64_emit_sve_inverted_cond): ...this new function. Use
5468 aarch64_emit_sve_ptrue_op for the comparison and
5469 aarch64_emit_unop to invert the result.
5470 (aarch64_expand_sve_vec_cmp_float): Update after the above
5471 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
5473 2018-05-07 Nathan Sidwell <nathan@acm.org>
5475 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
5476 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
5477 (Backwards Compatibility): Likewise.
5479 2018-05-07 Luis Machado <luis.machado@linaro.org>
5483 2018-05-07 Luis Machado <luis.machado@linaro.org>
5485 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5486 <prefetch_dynamic_strides>: New const bool field.
5487 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5488 prefetch_dynamic_strides.
5489 (exynosm1_prefetch_tune): Likewise.
5490 (thunderxt88_prefetch_tune): Likewise.
5491 (thunderx_prefetch_tune): Likewise.
5492 (thunderx2t99_prefetch_tune): Likewise.
5493 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
5495 (aarch64_override_options_internal): Update to set
5496 PARAM_PREFETCH_DYNAMIC_STRIDES.
5497 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
5498 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
5499 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
5500 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
5501 prefetch-dynamic-strides setting.
5503 2018-05-07 Luis Machado <luis.machado@linaro.org>
5505 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5506 <minimum_stride>: New const int field.
5507 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5508 minimum_stride field.
5509 (exynosm1_prefetch_tune): Likewise.
5510 (thunderxt88_prefetch_tune): Likewise.
5511 (thunderx_prefetch_tune): Likewise.
5512 (thunderx2t99_prefetch_tune): Likewise.
5513 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
5514 (aarch64_override_options_internal): Update to set
5515 PARAM_PREFETCH_MINIMUM_STRIDE.
5516 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
5517 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
5518 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
5519 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
5520 stride is constant and is below the minimum stride threshold.
5522 2018-05-07 Luis Machado <luis.machado@linaro.org>
5524 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
5527 2018-05-07 Luis Machado <luis.machado@linaro.org>
5529 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5530 <prefetch_dynamic_strides>: New const bool field.
5531 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5532 prefetch_dynamic_strides.
5533 (exynosm1_prefetch_tune): Likewise.
5534 (thunderxt88_prefetch_tune): Likewise.
5535 (thunderx_prefetch_tune): Likewise.
5536 (thunderx2t99_prefetch_tune): Likewise.
5537 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
5539 (aarch64_override_options_internal): Update to set
5540 PARAM_PREFETCH_DYNAMIC_STRIDES.
5541 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
5542 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
5543 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
5544 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
5545 prefetch-dynamic-strides setting.
5547 2018-05-07 Luis Machado <luis.machado@linaro.org>
5549 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5550 <minimum_stride>: New const int field.
5551 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5552 minimum_stride field.
5553 (exynosm1_prefetch_tune): Likewise.
5554 (thunderxt88_prefetch_tune): Likewise.
5555 (thunderx_prefetch_tune): Likewise.
5556 (thunderx2t99_prefetch_tune): Likewise.
5557 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
5558 (aarch64_override_options_internal): Update to set
5559 PARAM_PREFETCH_MINIMUM_STRIDE.
5560 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
5561 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
5562 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
5563 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
5564 stride is constant and is below the minimum stride threshold.
5566 2018-05-06 Jakub Jelinek <jakub@redhat.com>
5569 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
5570 the type is addressable. Don't force op into register if it has
5573 2018-05-05 Roland McGrath <mcgrathr@google.com>
5576 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
5577 any section for which we don't know a specific type it should have,
5578 regardless of name. Previously this was done only for the exact
5579 names ".init_array", ".fini_array", and ".preinit_array".
5580 (default_elf_asm_named_section): Add comment about
5581 relationship with default_section_type_flags and SECTION_NOTYPE.
5582 (get_section): Don't consider it a type conflict if one side has
5583 SECTION_NOTYPE and the other doesn't, as long as neither has the
5584 SECTION_BSS et al used in the default_section_type_flags logic.
5586 2018-05-05 Tom de Vries <tom@codesourcery.com>
5589 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
5590 (workaround_barsyncs): New function.
5591 (nvptx_reorg): Use workaround_barsyncs.
5592 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
5593 (define_expand "nvptx_membar_cta"): New define_expand.
5594 (define_insn "*nvptx_membar_cta"): New insn.
5596 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
5598 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
5599 To improve optimization opportunities.
5600 * builtin-types.def: The new needed builtin types for the above.
5602 2018-05-04 Richard Biener <rguenther@suse.de>
5604 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
5605 * gimple-ssa-store-merging.c
5606 (imm_store_chain_info::output_merged_store): Remove redundant create,
5607 release split_store vector contents on failure.
5608 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
5609 scalar stmt vector on cache hit.
5611 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
5613 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
5615 * config.gcc (powerpc-xilinx-eabi*): Remove.
5616 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
5618 (fusion_addis_mem_combo_load): Ditto.
5619 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
5621 (rs6000_cpu_cpp_builtins): Ditto.
5622 * config/rs6000/rs6000-linux.c
5623 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
5624 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
5625 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
5627 (rs6000_setup_reg_addr_masks): Ditto.
5628 (rs6000_init_hard_regno_mode_ok): Ditto.
5629 (rs6000_option_override_internal): Ditto.
5630 (legitimate_lo_sum_address_p): Ditto.
5631 (rs6000_legitimize_address): Ditto.
5632 (rs6000_legitimize_reload_address): Ditto.
5633 (rs6000_legitimate_address_p): Ditto.
5634 (abi_v4_pass_in_fpr): Ditto.
5635 (setup_incoming_varargs): Ditto.
5636 (rs6000_gimplify_va_arg): Ditto.
5637 (rs6000_split_multireg_move): Ditto.
5638 (rs6000_savres_strategy): Ditto.
5639 (rs6000_emit_prologue_components): Ditto.
5640 (rs6000_emit_epilogue_components): Ditto.
5641 (rs6000_emit_prologue): Ditto.
5642 (rs6000_emit_epilogue): Ditto.
5643 (rs6000_elf_file_end): Ditto.
5644 (rs6000_function_value): Ditto.
5645 (rs6000_libcall_value): Ditto.
5646 * config/rs6000/rs6000.h: Ditto.
5647 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
5648 (TARGET_MINMAX): ... this. New.
5649 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
5650 * config/rs6000/rs6000.md: Remove Xilinx FP support.
5651 (*movsi_internal1_single): Delete.
5652 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
5653 mfpu=, mxilinx-fpu): Delete.
5654 * config/rs6000/singlefp.h: Delete.
5655 * config/rs6000/sysv4.h: Remove Xilinx FP support.
5656 * config/rs6000/t-rs6000: Ditto.
5657 * config/rs6000/t-xilinx: Delete.
5658 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
5659 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
5660 (VStype_simple): Delete.
5661 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
5662 * config/rs6000/xfpu.h: Delete.
5663 * config/rs6000/xfpu.md: Delete.
5664 * config/rs6000/xilinx.h: Delete.
5665 * config/rs6000/xilinx.opt: Delete.
5666 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
5667 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
5669 2018-05-04 Tom de Vries <tom@codesourcery.com>
5672 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
5675 2018-05-04 Richard Biener <rguenther@suse.de>
5678 * tree-complex.c (update_complex_assignment): We are always in SSA form.
5679 (expand_complex_div_wide): Likewise.
5680 (expand_complex_operations_1): Likewise.
5681 (expand_complex_libcall): Preserve EH info of the original stmt.
5682 (tree_lower_complex): Handle removed blocks.
5683 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
5684 on complex multiplication and division libcall builtins.
5686 2018-05-04 Richard Biener <rguenther@suse.de>
5689 * fold-const.c (negate_expr_p): Restrict negation of operand
5690 zero of a division to when we know that can happen without
5692 (fold_negate_expr_1): Likewise.
5694 2018-05-04 Jakub Jelinek <jakub@redhat.com>
5697 * real.h (real_nextafter): Declare.
5698 * real.c (real_nextafter): New function.
5699 * fold-const-call.c (fold_const_nextafter): New function.
5700 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
5701 CASE_CFN_NEXTTOWARD.
5702 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
5703 even when arg1_mode is different from arg0_mode.
5705 2018-05-03 Nathan Sidwell <nathan@acm.org>
5707 * doc/extend.texi (Deprecated Features): Remove
5709 (Backwards Compatibility): Likewise.
5710 * doc/invoke.texi (C++ Language Options): Likewise.
5711 (C++ Dialect Options): Likewise.
5713 2018-05-03 Jakub Jelinek <jakub@redhat.com>
5716 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
5717 _mm512_mask_mullox_epi64): New intrinsics.
5719 2018-05-03 Tom de Vries <tom@codesourcery.com>
5722 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
5723 dump files): Add offload-tree.
5725 2018-05-03 Richard Biener <rguenther@suse.de>
5727 PR tree-optimization/85615
5728 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
5729 to loops not nested in BBs loop father to avoid creating multi-entry
5732 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5734 PR tree-optimization/70291
5735 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
5736 arguments. Change return type to tree. Emit libcall as a new
5737 statement rather than replacing existing one when inplace_p is true.
5738 (expand_complex_multiplication_components): New function.
5739 (expand_complex_multiplication): Expand floating-point complex
5740 multiplication using the above.
5741 (expand_complex_division): Rename inner_type parameter to type.
5742 Update expand_complex_libcall call-site.
5743 (expand_complex_operations_1): Update expand_complex_multiplication
5744 and expand_complex_division call-sites.
5746 2018-05-02 Jakub Jelinek <jakub@redhat.com>
5749 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
5750 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
5751 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
5752 the highest significant bit of the shift count mask is clear. In
5753 check whether and[sq]i3 is needed verify that all significant bits
5754 of the shift count other than the highest are set.
5756 2018-05-02 Tom de Vries <tom@codesourcery.com>
5759 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
5760 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
5761 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
5762 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
5763 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
5764 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
5765 __builtin_goacc_parlevel_size.
5767 2018-05-02 Richard Biener <rguenther@suse.de>
5769 PR tree-optimization/85597
5770 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
5771 do not use split vect_get_vec_defs call but call vect_get_slp_defs
5774 2018-05-02 Tom de Vries <tom@codesourcery.com>
5777 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
5778 dump files): Add ltrans-tree.
5780 2018-05-02 Tom de Vries <tom@codesourcery.com>
5783 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
5784 dump files): Add wpa-ipa.
5786 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
5788 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
5789 powerpc*-*-linux*paired* target.
5790 * config/rs6000/750cl.h: Delete.
5791 * config/rs6000/paired.h: Delete.
5792 * config/rs6000/paired.md: Delete.
5793 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
5795 * config/rs6000/rs6000-builtin.def: Remove paired float support.
5796 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
5797 comment. Remove paired float support.
5798 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
5799 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
5801 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
5802 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
5804 * config/rs6000/rs6000.c: Remove paired float support.
5805 (paired_expand_vector_init, paired_expand_vector_move,
5806 paired_emit_vector_compare, paired_emit_vector_cond_expr,
5807 (paired_expand_lv_builtin, paired_expand_stv_builtin,
5808 paired_expand_builtin, paired_expand_predicate_builtin,
5809 paired_init_builtins): Delete.
5810 * config/rs6000/rs6000.h: Remove paired float support.
5811 * config/rs6000/rs6000.md: Remove paired float support.
5812 (move_from_CR_ov_bit): Delete.
5813 * config/rs6000/rs6000.opt (mpaired): Delete.
5814 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
5815 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
5817 2018-05-02 Richard Biener <rguenther@suse.de>
5820 * gimplify.c (gimplify_save_expr): When in SSA form allow
5821 SAVE_EXPRs to compute to SSA vars.
5823 2018-05-02 Jakub Jelinek <jakub@redhat.com>
5826 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
5827 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
5828 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
5829 clobber operands[2], instead use a new pseudo. Formatting fixes.
5831 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
5833 PR tree-optimization/85586
5834 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
5835 exit early for statements in the same group if the accesses are
5838 2018-05-02 Tom de Vries <tom@codesourcery.com>
5841 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
5844 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
5846 PR tree-optimization/85143
5847 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
5849 2018-05-01 Tom de Vries <tom@codesourcery.com>
5852 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
5853 not found" error message.
5855 2018-05-01 Tom de Vries <tom@codesourcery.com>
5858 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
5859 * vec.c (test_ordered_remove_if): New function.
5860 (vec_c_tests): Call test_ordered_remove_if.
5861 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
5862 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
5863 * tree-vect-patterns.c (vect_pattern_recog_1): Use
5864 VEC_ORDERED_REMOVE_IF.
5866 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5868 PR tree-optimization/82665
5869 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
5870 pointer subtraction where arguments come from a memchr call.
5872 2018-05-01 Jakub Jelinek <jakub@redhat.com>
5874 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
5875 --push-state --as-needed and --pop-state instead of --as-needed and
5876 --no-as-needed if ld supports it.
5877 * configure: Regenerated.
5880 * doc/install.texi2html: Replace _002d with - and _002a with * in
5881 generated html files using sed.
5883 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5886 * gcc-rich-location.c (blank_line_before_p): New function.
5887 (use_new_line): New function.
5888 (gcc_rich_location::add_fixit_insert_formatted): New function.
5889 * gcc-rich-location.h
5890 (gcc_rich_location::add_fixit_insert_formatted): New function.
5892 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5894 * selftest.c (assert_streq): Rename "expected" and "actual" to
5895 "val1" and "val2". Extend NULL-handling to cover both inputs
5896 symmetrically, while still requiring both to be non-NULL for a pass.
5897 * selftest.h (assert_streq): Rename "expected" and "actual" to
5899 (ASSERT_EQ): Likewise.
5900 (ASSERT_EQ_AT): Likewise.
5901 (ASSERT_KNOWN_EQ): Likewise.
5902 (ASSERT_KNOWN_EQ_AT): Likewise.
5903 (ASSERT_NE): Likewise.
5904 (ASSERT_MAYBE_NE): Likewise.
5905 (ASSERT_MAYBE_NE_AT): Likewise.
5906 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
5907 the assertion to pass.
5908 (ASSERT_STREQ_AT): Likewise.
5910 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
5912 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
5913 interaction with -pie.
5915 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5917 * selftest.h: Fix alphabetization of per-source-file selftest
5920 2018-04-30 Jason Merrill <jason@redhat.com>
5922 PR c++/61982 - dead stores to destroyed objects.
5923 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
5926 2018-04-30 Jason Merrill <jason@redhat.com>
5928 * tree.c (build_clobber): New.
5929 * tree.h: Declare it.
5930 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
5932 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5934 * diagnostic-show-locus.c (layout::layout): Update for
5935 location_get_source_line returning a char_span.
5936 (struct char_span): Move to input.h.
5937 (struct correction): Update for fields in char_span becoming
5939 (struct source_line): Update for location_get_source_line
5940 returning a char_span.
5941 (layout::print_line): Likewise.
5942 * edit-context.c (edited_file::print_content): Likewise.
5943 (edited_file::print_diff_hunk): Likewise.
5944 (edited_file::print_run_of_changed_lines): Likewise.
5945 (edited_file::get_num_lines): Likewise.
5946 (edited_line::edited_line): Likewise.
5947 * final.c (asm_show_source): Likewise.
5948 * input.c (location_get_source_line): Convert return type
5949 from const char * to char_span, losing the final "line_len"
5951 (dump_location_info): Update for the above.
5952 (get_substring_ranges_for_loc): Likewise. Use a char_span
5953 when handling the literal within the line.
5954 (test_reading_source_line): Update for location_get_source_line
5955 returning a char_span.
5956 * input.h (class char_span): Move here from
5957 diagnostic-show-locus.c, converting from a struct to a class.
5958 Make data members private.
5959 (char_span::operator bool): New.
5960 (char_span::length): New.
5961 (char_span::get_buffer): New.
5962 (char_span::operator[]): New.
5963 (char_span::subspan): Make const.
5964 (char_span::xstrdup): New.
5965 (location_get_source_line): Convert return type from const char *
5966 to char_span, losing the final "line_size" param.
5968 2018-04-30 Jan Hubicka <jh@suse.cz>
5970 * lto-wrapper.c (ltrans_priorities): New static var.
5971 (cmp_priority): New.
5972 (run_gcc): Read priorities and if doing parallel build order
5973 the Makefile by them.
5975 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5977 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
5979 2018-04-30 Richard Biener <rguenther@suse.de>
5981 * tree-cfg.c (verify_address): Remove base argument, add
5982 flag whether to check TREE_ADDRESSABLE and do that.
5983 (verify_expr): Remove.
5984 (verify_types_in_gimple_reference): Add pieces from verify_expr.
5985 (verify_gimple_assign_single): Likewise.
5986 (verify_gimple_switch): Likewise.
5987 (verify_expr_location_1): Dereference tp once. Add (disabled)
5988 piece from verify_expr.
5989 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
5991 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
5993 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
5995 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
5997 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
5998 (small_data_pattern): Likewise.
5999 (arc_rewrite_small_data): Likewise.
6000 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
6001 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
6002 (get_symbol_alignment): New function.
6003 (legitimate_small_data_address_p): Likewise.
6004 (legitimate_scaled_address): Update, call
6005 legitimate_small_data_address_p.
6006 (output_sdata): New static variable.
6007 (arc_print_operand): Update how we handle small data operands.
6008 (arc_print_operand_address): Likewise.
6009 (arc_legitimate_address_p): Update, use
6010 legitimate_small_data_address_p.
6011 (arc_rewrite_small_data_p): Remove.
6012 (arc_rewrite_small_data_1): Likewise.
6013 (arc_rewrite_small_data): Likewise.
6014 (small_data_pattern): Likewise.
6015 (compact_sda_memory_operand): Update to use
6016 legitimate_small_data_address_p and get_symbol_alignment.
6017 (prepare_move_operands): Don't rewite sdata pattern.
6018 (prepare_extend_operands): Remove.
6019 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
6021 (zero_extendqisi2): Likewise.
6022 (zero_extendhisi2): Likewise.
6023 (extendqihi2): Likewise.
6024 (extendqisi2): Likewise.
6025 (extendhisi2): Likewise.
6029 * config/arc/constraints.md (Usd): Change it to memory constraint.
6031 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
6033 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
6034 as source of std instructions.
6035 * config/arc/arc.md (movsi_insn): Update pattern predicate to
6036 allow 6-bit constants as source for store instructions.
6037 (movdi_insn): Update instruction pattern to allow 6-bit constants
6038 as source for store instructions.
6040 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
6042 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
6044 2018-04-30 Nathan Sidwell <nathan@acm.org>
6045 Sandra Loosemore <sandra@codesourcery.com>
6047 * dumpfile.c (dump_open): Allow '-' for stdout.
6048 * doc/invoke.texi (Developer Options): Document dump filename
6049 determination early. Document stdin/stdout selection.
6051 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
6053 Microblaze Target: PIC data text relative
6055 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
6056 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
6058 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
6059 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
6060 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
6061 New addressing mode for data-text relative position indepenedent code.
6062 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
6063 'ADDRESS_SYMBOLIC_TXT_REL'.
6064 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
6065 (microblaze_legitimate_pic_operand): Exclude function calls from
6066 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
6067 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
6069 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
6070 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
6071 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
6072 for 'address + offset'.
6073 (microblaze_expand_prologue): Add new function prologue call for
6075 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
6076 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
6077 table in case of TARGET_PIC_DATA_TEXT_REL.
6078 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
6079 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
6080 Add new macros 'UNSPEC_TEXT',
6081 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
6082 + exclude function calls from 'UNSPEC_PLT' in case of data text
6084 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
6085 new target hook for generating address diff vector tables in case of
6087 * doc/tm.texi : Regenerate.
6088 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
6089 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
6090 of addr diff vector generation.
6091 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
6092 target hook definition.
6093 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
6094 Add default function for generate_pic_addr_diff_vec -> flag_pic.
6095 * doc/invoke.texi (Add new pic option): Add new microblaze pic
6096 option for data text relative.
6098 2018-04-30 Richard Biener <rguenther@suse.de>
6100 * tree-chrec.h (evolution_function_is_constant_p): Remove
6102 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
6104 2018-04-30 Richard Biener <rguenther@suse.de>
6107 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
6109 2018-04-30 Richard Biener <rguenther@suse.de>
6111 PR tree-optimization/28364
6112 PR tree-optimization/85275
6113 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
6114 copying first exit test.
6116 2018-04-28 Mark Wielaard <mark@klomp.org>
6118 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
6120 (dwarf_AT): Handle DW_AT_addr_base.
6121 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
6123 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
6126 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
6127 (*ashl<dwi>3_doubleword_mask_1): Ditto.
6128 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
6129 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
6131 2018-04-28 Richard Biener <rguenther@suse.de>
6133 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
6134 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
6135 to reflect use. Only add interesting stmts.
6137 2018-04-27 Martin Jambor <mjambor@suse.cz>
6140 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
6141 the jump function allows for passing through aggregate values.
6143 2018-04-27 David Malcolm <dmalcolm@redhat.com>
6145 * input.h (in_system_header_at): Convert from macro to inline
6147 (from_macro_expansion_at): Likewise.
6148 (from_macro_definition_at): Likewise.
6150 2018-04-27 Jeff Law <law@redhat.com>
6152 * config.gcc: Mark tile* targets as deprecated/obsolete.
6154 2018-04-27 Richard Biener <rguenther@suse.de>
6156 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
6159 2018-04-27 Richard Biener <rguenther@suse.de>
6161 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
6163 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
6165 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
6166 with Yd constraint. Set "preferred_for_speed" attribute from
6167 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
6169 (*movdi_internal): Ditto.
6170 (movti_interunit splitters): Remove
6171 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
6172 (movdi_interunit splitters): Ditto.
6173 * config/i386/constraints.md (Ye): Remove.
6174 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
6176 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6179 * config/aarch64/constraints.md (Usg): Limit to 31.
6182 2018-04-27 Jakub Jelinek <jakub@redhat.com>
6184 PR tree-optimization/85529
6185 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
6186 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
6187 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
6188 zero extension or masking of the MSB bit.
6189 (optimize_range_tests): Add FIRST_BB argument, pass it through
6190 to optimize_range_tests_var_bound.
6191 (maybe_optimize_range_tests, reassociate_bb): Adjust
6192 optimize_range_tests callers.
6194 2018-04-26 Richard Biener <rguenther@suse.de>
6195 Jakub Jelinek <jakub@redhat.com>
6197 * cgraph.h (symbol_table): Just declare debug method here.
6198 * symtab.c (symbol_table::debug): Define.
6200 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
6202 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
6204 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
6206 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
6207 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
6208 (*movdi_internal): Substitute Yi and Yj constraint with x
6209 and Ym and Yn constraint with y constraint. Update "isa"
6210 attribute and set "preferred_for_speed" attribute from
6211 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
6212 (*movsi_internal): Ditto.
6213 (*movdf_internal): Ditto.
6214 (*movsf_internal): Ditto.
6215 (*zero_extendsidi2): Ditto.
6216 * config/i386/sse.md (vec_set<mode>_0): Ditto.
6217 (sse2_loadld): Ditto.
6218 (*vec_extract<ssevecmodelower>_0): Ditto.
6219 (*vec_extractv4si_0_zext_sse4): Ditto.
6220 (vec_concatv2di): Ditto.
6221 (*vec_dup<mode>): Ditto.
6222 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
6223 * config/i386/constraints.md (Yi): Remove.
6228 2018-04-26 Nathan Sidwell <nathan@acm.org>
6230 * dumpfile.c (dump_open): New.
6231 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
6232 (dump_finish): Detect stdio/stderr by value not name.
6234 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
6236 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
6238 2018-04-26 Tom de Vries <tom@codesourcery.com>
6241 * config/nvptx/nvptx.c (verify_neutering_jumps)
6242 (verify_neutering_labels): New function
6243 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
6245 2018-04-26 Tom de Vries <tom@codesourcery.com>
6248 * config/nvptx/nvptx.c (needs_neutering_p): New function.
6249 (nvptx_single): Use needs_neutering_p to skip over insns that do not
6252 2018-04-26 Richard Biener <rguenther@suse.de>
6253 Tom de Vries <tom@codesourcery.com>
6256 * lto-streamer-out.c (output_function): Fixup loops if required to match
6257 discovery done in the reader.
6259 2018-04-26 Richard Biener <rguenther@suse.de>
6261 PR tree-optimization/85116
6262 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
6263 have a loop exit from the single latch predecessor. Remove
6264 case of header with just condition.
6265 (ch_base::copy_headers): Exclude infinite loops from any
6267 (pass_ch::execute): Record exits.
6269 2018-04-26 Richard Biener <rguenther@suse.de>
6271 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
6272 prologue cost vector and pass it to vect_get_load_cost.
6273 (vect_get_peeling_costs_all_drs): Likewise.
6274 (vect_peeling_hash_get_lowest_cost): Likewise.
6275 (vect_enhance_data_refs_alignment): Likewise.
6277 2018-04-26 Richard Biener <rguenther@suse.de>
6280 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
6281 checking of integer<->pointer conversions.
6282 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
6283 sign-/zero-extending pointer types.
6284 (expand_omp_for_static_chunk): Likewise.
6286 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
6287 Jean Lee <xiaoyur347@gmail.com>
6289 * config/mips/mips.c (mips_asan_shadow_offset): New function.
6290 (TARGET_ASAN_SHADOW_OFFSET): Define.
6291 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
6292 true for -fsanitize=address.
6294 2018-04-25 Mark Wielaard <mark@klomp.org>
6296 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
6299 2018-04-25 Jakub Jelinek <jakub@redhat.com>
6301 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
6302 than "alu", remove explicit "memory" and "imm_disp" attributes.
6303 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
6306 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
6307 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
6308 gen_lowpart_no_emit.
6310 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
6313 * config/i386/i386.c (ix86_expand_builtin): Change memory
6314 operand to XI, extend p0 to Pmode.
6315 * config/i386/i386.md: Change unspec volatile and operand
6316 1 mode to XI, change operand 0 mode to P.
6318 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
6320 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
6321 GET_MODE_MASK before any checking.
6322 (nds32_can_use_bset_p): Likewise.
6323 (nds32_can_use_btgl_p): Likewise.
6325 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
6327 * config/nds32/nds32-doubleword.md: New define_split pattern for
6328 illegal register number.
6330 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
6332 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
6334 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
6336 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
6338 2018-04-25 Richard Biener <rguenther@suse.de>
6340 * lto-streamer.h (LTO_major_version): Bump to 8.
6342 2018-04-25 Jakub Jelinek <jakub@redhat.com>
6344 * BASE-VER: Set to 9.0.0.
6346 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
6348 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
6349 in __abskf2 and __powikf2.
6351 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6354 * config/aarch64/constraints.md (Usg, Usj): New constraints.
6355 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
6356 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
6357 Use the above on operand 2. Reindent.
6358 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
6360 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
6363 * common/config/i386/i386-common.c (ix86_handle_option): Don't
6365 * config/i386/i386.opt (mcet): Removed.
6366 * doc/install.texi: Remove -mcet documentation.
6367 * doc/invoke.texi: Likewise.
6369 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
6372 * doc/install.texi: Remove -mcet from bootstrap-cet.
6374 2018-04-24 Jakub Jelinek <jakub@redhat.com>
6377 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
6378 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
6382 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
6383 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
6384 containing a CONST_VECTOR.
6386 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
6388 * doc/install.texi: Update newlib dependency for nvptx.
6390 2018-04-24 Jakub Jelinek <jakub@redhat.com>
6393 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
6394 instead of INTVAL when shifting x left.
6396 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
6398 PR tree-optimization/85478
6399 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
6400 vect_grouped_store_supported for single element vectors.
6402 2018-04-24 Richard Biener <rguenther@suse.de>
6405 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
6406 load cost increase to the case of non-constant step.
6408 2018-04-24 Jakub Jelinek <jakub@redhat.com>
6411 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
6412 destination if any_malformed_asm.
6414 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
6417 * expr.c (store_field): In the bitfield case, if the value comes from
6418 a function call and is returned in registers by means of a PARALLEL,
6419 do not change the mode of the temporary unless BLKmode and VOIDmode.
6421 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
6423 PR rtl-optimization/85423
6424 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
6425 dependencies to debug insns when the previous insn is non-debug.
6427 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
6429 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
6430 enums into a single definition.
6431 (fls): Fix predicates and printing.
6434 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
6436 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
6437 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
6438 and short u6 immediate.
6439 (check_if_valid_sleep_operand): Remove.
6440 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
6442 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
6444 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
6445 flag_always_save_lp condition.
6446 * config/nds32/nds32.opt (malways-save-lp): New option.
6448 2018-04-22 Shiva Chen <shiva0217@gmail.com>
6450 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
6451 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
6452 * config/nds32/nds32.h
6453 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
6454 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
6456 2018-04-22 Shiva Chen <shiva0217@gmail.com>
6458 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
6459 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
6461 2018-04-22 Shiva Chen <shiva0217@gmail.com>
6462 Chung-Ju Wu <jasonwucj@gmail.com>
6464 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
6466 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
6467 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
6469 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
6471 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
6473 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
6475 * config/nds32/nds32-protos.h (nds32_data_alignment,
6476 nds32_local_alignment): Declare.
6477 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
6478 nds32_local_alignment): New functions.
6479 (TARGET_CONSTANT_ALIGNMENT): Define.
6480 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
6482 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
6484 * config/nds32/nds32.c
6485 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
6486 (TARGET_MODES_TIEABLE_P): Likewise.
6488 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
6490 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
6493 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
6494 Chung-Ju Wu <jasonwucj@gmail.com>
6496 * config/nds32/constants.md (unspec_volatile_element): Add enum values
6497 for unaligned access.
6498 * config/nds32/nds32-intrinsic.c: Implementation of expanding
6500 * config/nds32/nds32-intrinsic.md: Likewise.
6501 * config/nds32/nds32_intrinsic.h: Likewise.
6502 * config/nds32/nds32.h (nds32_builtins): Likewise.
6503 * config/nds32/nds32.opt (munaligned-access): New option.
6504 * config/nds32/nds32.c (nds32_asm_file_start): Display
6505 flag_unaligned_access status.
6507 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
6509 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
6510 -mno-relax is present.
6511 * config/riscv/linux.h (LINK_SPEC): Ditto.
6513 2018-04-20 Martin Sebor <msebor@redhat.com>
6516 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
6518 (gimple_fold_builtin_stxcpy_chk): Same.
6519 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
6521 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
6524 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
6525 __powikf2 when long double is IEEE 128-bit.
6527 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
6529 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
6530 step to make sure stack always aligned.
6532 2018-04-20 Carl Love <cel@us.ibm.com>
6535 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
6536 size check for arg0.
6538 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
6539 Tom de Vries <tom@codesourcery.com>
6542 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
6543 Emit insns for calls too.
6544 (nvptx_find_par): Always look for worker-level predecessor insn.
6545 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
6547 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
6548 (nvptx_process_pars): Propagate frames for calls.
6550 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
6553 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
6555 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
6556 (ix86_handle_option): Don't handle OPT_mibt.
6557 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
6559 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
6561 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
6562 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
6563 (ix86_target_macros): Define __CET__ with flag_cf_protection
6564 for -fcf-protection.
6565 * config/i386/i386.c (isa2_opts): Remove -mibt.
6566 * config/i386/i386.h (TARGET_IBT): Removed.
6567 (TARGET_IBT_P): Likewise.
6568 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
6569 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
6570 * config/i386/i386.opt (mcet): Update help message.
6573 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
6574 -mcet as an alias for -mshstk.
6576 2018-04-20 Richard Biener <rguenther@suse.de>
6579 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
6580 complexity by forcing a single use of the multiply operand.
6582 2018-04-20 Martin Jambor <mjambor@suse.cz>
6585 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
6586 recursion dependency to only apply to non-clones.
6588 2018-04-20 Martin Jambor <mjambor@suse.cz>
6591 * ipa-cp.c (create_specialized_node): Check that clones of
6592 self-recursive edges exist during IPA-CP.
6594 2018-04-19 Toon Moene <toon@moene.org>
6596 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
6599 2018-04-19 Jakub Jelinek <jakub@redhat.com>
6601 PR tree-optimization/85467
6602 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
6603 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
6604 VECTOR_CST element to type.
6606 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
6609 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
6610 * config/i386/i386.md (builtin_setjmp_setup): Removed.
6611 (builtin_longjmp): Likewise.
6612 (save_stack_nonlocal): New pattern.
6613 (restore_stack_nonlocal): Likewise.
6615 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
6618 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
6619 Replace ASM_OUTPUT_LABEL with fprintf.
6621 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
6624 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
6625 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
6626 * config/i386/i386-c.c (ix86_target_macros_internal): Also
6627 define __IBT__ and __SHSTK__ for -fcf-protection.
6628 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
6630 (ix86_trampoline_init): Likewise.
6631 (x86_output_mi_thunk): Likewise.
6632 (ix86_notrack_prefixed_insn_p): Likewise.
6633 (ix86_option_override_internal): Don't disallow -fcf-protection.
6634 * config/i386/i386.md (rdssp<mode>): Also enable for
6636 (incssp<mode>): Likewise.
6637 (nop_endbr): Likewise.
6638 * config/i386/i386.opt (mcet): Change help message to built-in
6642 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
6643 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
6644 enable CET built-in functions.
6646 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
6648 * common/config/i386/i386-common.c
6649 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
6650 OPTION_MASK_ISA_MOVDIRI_UNSET,
6651 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
6652 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
6653 * config.gcc (movdirintrin.h): New header.
6654 * config/i386/cpuid.h (bit_MOVDIRI,
6655 bit_MOVDIR64B): New bits.
6656 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
6658 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
6659 (VOID, PVOID, PCVOID)): New function types.
6660 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
6661 __builtin_ia32_directstoreu_u64,
6662 __builtin_ia32_movdir64b): New builtins.
6663 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
6664 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
6666 (ix86_valid_target_attribute_inner_p): Ditto.
6667 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
6668 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
6669 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
6670 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
6671 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
6672 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
6673 (movdiri<mode>, movdir64b_<mode>): New.
6674 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
6675 * config/i386/immintrin.h: Include movdirintrin.h.
6676 * config/i386/movdirintrin.h: New file.
6677 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
6679 2018-04-19 Richard Biener <rguenther@suse.de>
6682 * cfg.c (clear_bb_flags): When loop state says we have
6683 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
6685 2018-04-19 Richard Biener <rguenther@suse.de>
6687 PR tree-optimization/84737
6688 * tree-vect-data-refs.c (vect_copy_ref_info): New function
6689 copying restrict info.
6690 (vect_setup_realignment): Use it.
6691 * tree-vectorizer.h (vect_copy_ref_info): Declare.
6692 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
6693 the first DR to all generated stores.
6694 (vectorizable_load): Likewise for loads.
6696 2018-04-19 Jakub Jelinek <jakub@redhat.com>
6698 PR tree-optimization/85446
6699 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
6700 the integral and pointer types to have the same precision.
6702 * doc/install.texi: Document --disable-cet being the default and
6705 2018-04-18 Martin Liska <mliska@suse.cz>
6707 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
6710 2018-04-18 Martin Liska <mliska@suse.cz>
6713 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
6716 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
6717 arguments if they are comparable.
6719 2018-04-18 Martin Liska <mliska@suse.cz>
6722 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
6725 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
6728 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
6731 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
6732 ENDBR after calling __morestack.
6734 2018-04-18 David Malcolm <dmalcolm@redhat.com>
6737 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
6738 by using gcc_base_ver to generate a gcc_driver_version, and use
6739 it when generating GCC_DRIVER_NAME.
6740 * configure: Regenerate.
6742 2018-04-18 Jakub Jelinek <jakub@redhat.com>
6745 * config.gcc: Obsolete powerpc*-*-*spe*.
6747 2018-04-17 Jakub Jelinek <jakub@redhat.com>
6750 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
6751 (stabstr_D): Change type of unum from unsigned int to
6752 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
6755 2018-04-17 Jim Wilson <jimw@sifive.com>
6758 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
6759 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
6760 Set arg_pointer_offset after using pretend_args_size.
6762 2018-04-17 Jakub Jelinek <jakub@redhat.com>
6764 PR rtl-optimization/85431
6765 * dse.c (record_store): Ignore zero width stores.
6768 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
6769 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
6770 __builtin_stack_restore rather than after it.
6771 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
6772 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
6773 argument instead of virtual_dynamic_stack_rtx.
6775 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
6777 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
6779 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
6780 Add note to error message to explain internal mapping of overloaded
6781 built-in function name to non-overloaded built-in function name.
6782 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
6785 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
6788 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
6789 where the inputs overlap with the output.
6791 2018-04-17 Jakub Jelinek <jakub@redhat.com>
6793 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
6794 (=v, v) alternative and explicit "memory" attribute.
6795 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
6796 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
6798 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
6799 "sselog1" type instead of "sselog".
6800 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
6801 "sselog". Remove explicit "memory" attribute.
6802 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
6803 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
6805 (vec_extract_hi_v32hi): Merge all alternatives into one, use
6806 "sselog1" type instead of "sselog". Remove explicit "memory"
6808 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
6809 use "sselog1" type instead of "sselog". Remove explicit "memory"
6811 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
6812 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
6814 (vec_extract_hi_v64qi): Merge all alternatives into one, use
6815 "sselog1" type instead of "sselog". Remove explicit "memory"
6817 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
6818 use "sselog1" type instead of "sselog". Remove explicit "memory"
6822 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
6825 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
6828 2018-04-17 Martin Jambor <mjambor@suse.cz>
6831 * ipa-cp.c (create_specialized_node): Call
6832 expand_all_artificial_thunks if necessary.
6834 2018-04-17 Martin Liska <mliska@suse.cz>
6837 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
6838 in message, remote space in between '_G' and '('.
6840 2018-04-17 Jakub Jelinek <jakub@redhat.com>
6843 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
6844 avx512f_vmcmp<mode>3<round_saeonly_name>,
6845 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
6846 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
6847 avx512f_rndscale<mode><round_saeonly_name>,
6848 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
6849 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
6850 Use %<iptr>2 instead of %2 for -masm=intel.
6851 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
6852 avx512f_vcvttss2usi<round_saeonly_name>,
6853 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
6855 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
6856 avx512f_vcvttsd2usi<round_saeonly_name>,
6857 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
6858 Use %q1 instead of %1 for -masm=intel.
6859 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
6860 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
6861 of %3 for -masm=intel.
6862 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
6864 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
6866 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
6868 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
6869 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
6871 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
6873 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
6874 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
6875 %g1 and one with %0 and %1.
6876 (avx512er_vmrcp28<mode><round_saeonly_name>,
6877 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
6879 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
6880 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
6881 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
6882 of %0 and %{%4%} for -masm=intel.
6883 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
6884 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
6885 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
6886 order of %0 and %{%5%}%{z%} for -masm=intel.
6888 2018-04-17 Jan Hubicka <jh@suse.cz>
6891 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
6893 2018-04-17 Martin Liska <mliska@suse.cz>
6896 * multiple_target.c (create_dispatcher_calls): Set apostrophes
6897 for target_clone error message. Make default implementation
6898 clone to be a local declaration.
6899 (separate_attrs): Add new argument and check for an empty
6901 (expand_target_clones): Handle it.
6902 (ipa_target_clone): Make redirection just for target_clones
6905 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
6906 Tom de Vries <tom@codesourcery.com>
6909 * omp-expand.c (expand_oacc_for): Add dummy false branch for
6910 tiled basic blocks without omp continue statements.
6912 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
6915 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
6916 vec_extract expression as having side effects to make sure it gets
6919 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
6922 * config/i386/i386.c (get_builtin_code_for_version): Check
6925 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
6928 * gcc/config.gcc: Support "skylake".
6929 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
6931 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
6932 (processor_target_table): Add "skylake".
6933 (ix86_option_override_internal): Add "skylake".
6934 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
6935 PROCESSOR_CANNONLAKE.
6936 (get_builtin_code_for_version): Fix priority for
6937 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
6938 PROCESSOR_SKYLAKE-AVX512.
6939 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
6940 (processor_type): Add PROCESSOR_SKYLAKE.
6942 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
6943 Jason Merrill <jason@redhat.com>
6946 * convert.c (convert_to_integer_1): Use direct recursion for
6947 enumeral types and types with a precision less than the number
6948 of bits in their mode.
6950 2018-04-16 Julia Koval <julia.koval@intel.com>
6953 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
6954 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
6956 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
6959 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
6960 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
6961 and -mno-direct-move.
6963 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
6966 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
6967 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
6968 Ensure negative shifts result in {0}.
6970 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
6972 PR rtl-optimization/79916
6973 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
6974 regs (if any) to define how to gnerate SD moves when LRA is in
6977 2018-04-13 Jakub Jelinek <jakub@redhat.com>
6979 PR rtl-optimization/85393
6980 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
6981 * except.c (expand_dw2_landing_pad_for_region): Make static.
6982 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
6983 a label and unconditional jump to old_bb, rather than
6984 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
6987 PR rtl-optimization/85376
6988 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
6989 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
6990 instead of a specific value.
6992 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
6993 Bin Cheng <bin.cheng@arm.com>
6995 PR tree-optimization/82965
6996 PR tree-optimization/83991
6997 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
6998 by_profile_only parameter.
6999 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
7000 information if the loop was predicted to iterate too many times.
7001 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
7003 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
7006 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
7009 2018-04-13 Martin Liska <mliska@suse.cz>
7010 Jakub Jelinek <jakub@redhat.com>
7013 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
7014 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
7015 * builtins.c (expand_builtin_memory_copy_args): Use
7016 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
7017 handle dest_addr == pc_rtx.
7019 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
7022 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
7023 asked to not generate direct moves.
7024 (fix_trunc<mode>si2_stfiwx): Similar.
7025 (fix_trunc<mode>si2_internal): Similar.
7027 2018-04-12 Jakub Jelinek <jakub@redhat.com>
7030 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
7031 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
7032 lookup if dest in some wider mode is known to be const0_rtx and
7033 if so, record permanent equivalence for it to be ZERO_EXTEND of
7034 the narrower mode destination.
7036 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
7038 * lto-streamer-out.c (output_function): Revert 259346.
7039 * omp-expand.c (expand_oacc_for): Likewise.
7041 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
7043 PR rtl-optimization/85354
7044 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
7045 * sel-sched.c (sel_global_init): ... here.
7047 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
7050 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
7051 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
7052 mode for PE-COFF targets.
7053 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
7054 (i386_pe_asm_lto_end): Likewise.
7055 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
7056 (TARGET_ASM_LTO_END): Likewise.
7057 * config/i386/winnt.c (saved_debug_info_level): New static variable.
7058 (i386_pe_asm_lto_start): New function.
7059 (i386_pe_asm_lto_end): Likewise.
7061 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
7062 Richard Biener <rguenther@suse.de>
7065 * lto-streamer-out.c (output_function): Fix CFG loop state before
7067 * omp-expand.c (expand_oacc_for): Handle calls to internal
7068 functions like regular functions.
7070 2018-04-12 Richard Biener <rguenther@suse.de>
7073 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
7074 for the early LTO debug to properly generate references to it
7075 during DIE emission. Do not re-use that for the skeleton for
7077 (dwarf2out_early_finish): Likewise.
7079 2018-04-12 Jakub Jelinek <jakub@redhat.com>
7082 * config/i386/sse.md
7083 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
7084 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
7085 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
7086 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
7087 and output is a reg, avoid creating invalid lowpart subreg, but
7088 instead split into a 512-bit move. Don't split if not AVX512VL,
7089 input is xmm16+ reg and output is a mem.
7090 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
7091 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
7092 xmm16+ reg and output is a mem.
7094 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7096 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
7097 also for flag_dwarf2_cfi_asm.
7099 2018-04-12 Jakub Jelinek <jakub@redhat.com>
7101 PR rtl-optimization/85342
7102 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
7103 a bool scalar var inside of the loop instead. Don't try to update
7104 recog_data.operand after failed apply_change_group.
7106 2018-04-12 Tom de Vries <tom@codesourcery.com>
7109 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
7110 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
7111 array with flexible array member as array without given dimension.
7112 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
7113 argument for undefined param to true.
7115 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
7118 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
7119 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
7120 from PowerPC section.
7121 * config/rs6000/sysv4.opt (mcall-): Improve help text.
7122 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
7123 help text that is too long.
7124 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
7125 help text that is too long.
7126 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
7127 help text that is too long.
7129 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
7131 * config/alpha/alpha.md (stack_probe_internal): Rename
7132 from "probe_stack". Update all callers.
7134 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
7136 PR rtl-optimization/84566
7137 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
7138 sched_macro_fuse_insns.
7140 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
7143 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
7144 (compute_block_dependences): ... from here.
7146 2018-04-11 Jakub Jelinek <jakub@redhat.com>
7148 PR tree-optimization/85331
7149 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
7150 from int to HOST_WIDE_INT.
7152 2018-04-11 Martin Jambor <mjambor@suse.cz>
7155 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
7156 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
7157 not the same as the source val.
7158 (cgraph_edge_brings_value_p): New parameter.
7159 (gather_edges_for_value): Pass destination value to
7160 cgraph_edge_brings_value_p.
7161 (perhaps_add_new_callers): Likewise.
7162 (get_info_about_necessary_edges): Likewise and exclude values brought
7163 only by self-recursive edges.
7164 (create_specialized_node): Redirect only clones of self-calling edges.
7165 (+self_recursive_pass_through_p): New function.
7166 (find_more_scalar_values_for_callers_subset): Use it.
7167 (find_aggregate_values_for_callers_subset): Likewise.
7168 (known_aggs_to_agg_replacement_list): Removed.
7169 (decide_whether_version_node): Re-calculate known constants for all
7170 remaining context clones.
7172 2018-04-11 Richard Biener <rguenther@suse.de>
7175 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
7176 from early DWARF output.
7177 (dwarf2out_early_finish): Output line info unconditionally into
7178 early DWARF and add reference to it.
7180 2018-04-11 Jakub Jelinek <jakub@redhat.com>
7183 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
7184 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
7185 other than V2DFmode using iptr mode attribute.
7186 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
7188 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
7190 PR rtl-optimization/84659
7191 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
7193 2018-04-11 Jakub Jelinek <jakub@redhat.com>
7196 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
7198 (output_loc_list): Pass address of a dummy size variable even in the
7199 locview handling loop.
7200 (index_location_lists): Add comment on why skip_loc_list_entry can't
7203 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
7206 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
7209 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
7212 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
7213 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
7214 and -mstring-compare-inline-limit.
7216 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
7219 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
7220 for stack clash protection in a register whenever we need it to be in
7223 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
7225 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
7226 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
7228 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
7231 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
7233 (mlong-double-): Ditto.
7234 * config/rs6000/sysv4.opt (msdata=): Ditto.
7235 (mtls-size=): Ditto.
7237 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
7239 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
7240 erroneous entries for
7241 "vector int vec_ldl (int, long int *)", and
7242 "vector unsigned int vec_ldl (int, unsigned long int *)".
7243 Add comments and entries for
7244 "vector bool char vec_ldl (int, bool char *)",
7245 "vector bool short vec_ldl (int, bool short *)",
7246 "vector bool int vec_ldl (int, bool int *)",
7247 "vector bool long long vec_ldl (int, bool long long *)",
7248 "vector pixel vec_ldl (int, pixel *)",
7249 "vector long long vec_ldl (int, long long *)",
7250 "vector unsigned long long vec_ldl (int, unsigned long long *)".
7251 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
7252 type tree bool_long_long_type_node and correct definition of
7253 bool_V2DI_type_node to make reference to this new type tree.
7254 (rs6000_mangle_type): Replace erroneous reference to
7255 bool_long_type_node with bool_long_long_type_node.
7256 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
7257 comments to emphasize sign distinctions for char and int types and
7258 replace RS6000_BTI_bool_long constant with
7259 RS6000_BTI_bool_long_long constant. Also add comment to restrict
7260 use of RS6000_BTI_pixel.
7261 (bool_long_type_node): Remove this macro definition.
7262 (bool_long_long_type_node): New macro definition
7264 2018-04-10 Jakub Jelinek <jakub@redhat.com>
7266 PR rtl-optimization/85300
7267 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
7268 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
7269 simplify_unary_operation fails.
7271 2018-04-10 Martin Liska <mliska@suse.cz>
7273 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
7274 cgraph_edge and ipa_ref.
7276 2018-04-10 Jakub Jelinek <jakub@redhat.com>
7280 * config/i386/sse.md
7281 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
7282 computation of the VEC_MERGE selector from mask.
7283 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
7284 Fix decoding of the VEC_MERGE selector into mask.
7286 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
7288 PR tree-optimization/85286
7289 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
7291 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
7293 * final.c (final_1): Set insn_last_address as well as
7294 insn_current_address.
7296 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7299 * explow.c (emit_stack_probe): Call validize_mem on memory location
7300 before passing it to gen_probe_stack. Create address operand and
7301 legitimize it for the probe_stack_address case.
7303 2018-04-09 Jan Hubicka <jh@suse.cz>
7306 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
7307 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
7308 * tree.c (free_lang_data_in_type): Fix handling of binfos;
7310 (free_lang_data): Rebuild type inheritance graph.
7312 2018-04-09 Martin Sebor <msebor@redhat.com>
7314 * invoke.texi (-finline-small-functions): Mention other optimization
7316 (-findirect-inlining, -fpartial-inlining): Same.
7317 (-finline-functions-called-once): Same.
7318 (-freorder-blocks-and-partition): Same.
7320 2018-04-09 Jan Hubicka <jh@suse.cz>
7323 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
7324 jumps; choose last target that matches the criteria (i.e.
7325 no partition changes for non-crossing jumps).
7326 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
7327 support for redirecting crossing jumps to non-crossing.
7329 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
7331 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
7332 also for naked functions.
7334 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
7336 * config/arc/arc.md (add_shift): New pattern.
7337 (add_shift2): Likewise.
7338 (sub_shift): Likewise.
7339 (sub_shift_cmp0_noout): Likewise.
7340 (compare_si_ashiftsi): Likewise.
7341 (xbfu_cmp0_noout): New combine pattern.
7342 (xbfu_cmp0"): Likewise.
7343 (movsi_set_cc_insn): Place the predicable variant first.
7344 (commutative_binary_cmp0_noout): Remove clobber.
7345 (commutative_binary_cmp0): New pattern.
7346 (noncommutative_binary_cmp0): Likewise.
7347 (noncommutative_binary_cmp0_noout): Likewise.
7348 (noncommutative_binary_comparison_result_used): Removed.
7349 (rsub_cmp0): New pattern.
7350 (rsub_cmp0_noout): Likewise.
7351 (extzvsi): Changed, keep only meaningful variants.
7352 (SQH, SEZ): New iterators.
7353 (SQH_postfix): New mode attribute.
7354 (SEZ_prefix): New code attribute.
7355 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
7356 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
7357 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
7359 (noncommutative_operator): Check the availability of barrel
7362 2018-04-09 Richard Biener <rguenther@suse.de>
7364 PR tree-optimization/85284
7365 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
7366 Only use the niter constraining form of simple_iv when the exit
7369 2018-04-09 Tom de Vries <tom@codesourcery.com>
7372 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
7373 (define_expand "*memory_barrier"): New define_expand.
7374 (define_insn "memory_barrier"): New insn.
7376 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
7378 PR rtl-optimization/80463
7379 PR rtl-optimization/83972
7380 PR rtl-optimization/83480
7382 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
7383 correct producer for the insn.
7384 (tidy_control_flow): Fixup seqnos in case of debug insns.
7386 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
7388 PR rtl-optimization/83913
7390 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
7391 different sched-times when merging exprs.
7393 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
7395 PR rtl-optimization/83962
7397 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
7398 tidy_fallthru_edge and tidy_control_flow.
7400 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
7402 PR rtl-optimization/83530
7404 * sel-sched.c (force_next_insn): New global variable.
7405 (remove_insn_for_debug): When force_next_insn is true, also leave only
7406 next insn in the ready list.
7407 (sel_sched_region): When the region wasn't scheduled, make another pass
7408 over it with force_next_insn set to 1.
7410 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
7412 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
7414 * config/nds32/constants.md (unspec_volatile_element): Add enum values
7415 for interrupt control.
7416 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
7417 functions for interrupt control.
7418 * config/nds32/nds32-intrinsic.md: Likewise.
7419 * config/nds32/nds32_intrinsic.h: Likewise.
7420 * config/nds32/nds32.h (nds32_builtins): Likewise.
7422 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
7424 * config/nds32/nds32.c (nds32_init_machine_status,
7425 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
7426 strict_aligned_p field.
7427 (nds32_expand_to_rtl_hook): New function.
7428 (TARGET_EXPAND_TO_RTL_HOOK): Define.
7429 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
7431 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
7432 Chung-Ju Wu <jasonwucj@gmail.com>
7434 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
7435 * config/nds32/nds32-n7.md: New file.
7436 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
7437 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
7439 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
7440 * config/nds32/nds32.md (pipeline_model): Add n7.
7441 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
7442 * config/nds32/pipelines.md: Include n7 settings.
7444 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
7445 Chung-Ju Wu <jasonwucj@gmail.com>
7447 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
7448 * config/nds32/nds32-e8.md: New file.
7449 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
7450 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
7452 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
7453 * config/nds32/nds32.md (pipeline_model): Add e8.
7454 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
7455 * config/nds32/pipelines.md: Include e8 settings.
7457 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
7458 Chung-Ju Wu <jasonwucj@gmail.com>
7460 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
7461 * config/nds32/nds32-n8.md: New file.
7462 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
7463 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
7465 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
7466 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
7467 * config/nds32/nds32.md (pipeline_model): Add n8.
7468 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
7469 * config/nds32/pipelines.md: Include n8 settings.
7471 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
7472 Chung-Ju Wu <jasonwucj@gmail.com>
7474 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
7475 * config/nds32/nds32-n9-2r1w.md: New file.
7476 * config/nds32/nds32-n9-3r2w.md: New file.
7477 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
7478 nds32_register_ports): New or modify for cpu n9.
7479 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
7481 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
7482 * config/nds32/nds32-utils.c: New file.
7483 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
7484 TARGET_MUL_SLOW): Define.
7485 * config/nds32/nds32.md (pipeline_model): New attribute.
7486 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
7487 New options that support cpu n9.
7488 * config/nds32/pipelines.md: Include n9 settings.
7489 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
7491 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
7493 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
7494 information if necessary.
7495 (output_cond_branch_compare_zero): Likewise.
7496 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
7497 (nds32_target_alignment): Refine for alignment.
7498 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
7499 (FUNCTION_BOUNDARY): Modify.
7500 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
7502 * config/nds32/nds32.opt (malways-align, malign-functions): New.
7504 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
7506 * config/nds32/constants.md (unspec_volatile_element): Add values for
7507 TLB operation and data prefetch.
7508 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
7509 functions for TLB operation and data prefetch.
7510 * config/nds32/nds32-intrinsic.md: Likewise.
7511 * config/nds32/nds32_intrinsic.h: Likewise.
7512 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
7513 (nds32_print_operand): Likewise.
7514 * config/nds32/nds32.h (nds32_builtins): Likewise.
7516 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
7517 Andrew Pinski <pinsika@gcc.gnu.org>
7520 * match.pd: Use constant_boolean_node of correct type instead of
7521 boolean_true_node or boolean_false_node for simplifying
7522 pointer comparisons to zero.
7524 2018-04-07 Jakub Jelinek <jakub@redhat.com>
7526 PR tree-optimization/80021
7527 * tree.c (verify_type_variant): Make error call in verify_variant_match
7528 translatable and remove final full stop.
7530 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
7532 * config/nds32/constants.md (unspec_volatile_element): Add
7533 UNSPEC_VOLATILE_EH_RETURN.
7534 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
7535 nds32_output_stack_pop): Support dwarf exception handling process.
7536 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
7537 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
7538 exception handling process.
7539 (nds32_compute_stack_frame): Likewise.
7540 (nds32_return_addr_rtx): Likewise.
7541 (nds32_initial_elimination_offset): Likewise.
7542 (nds32_expand_prologue): Likewise.
7543 (nds32_expand_epilogue): Likewise.
7544 (nds32_dynamic_chain_address): New function.
7545 * config/nds32/nds32.h (machine_function): Add fields for dwarf
7547 (DYNAMIC_CHAIN_ADDRESS): Define.
7548 (EH_RETURN_DATA_REGNO): Define.
7549 (EH_RETURN_STACKADJ_RTX): Define.
7550 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
7551 patterns for dwarf exception handling.
7553 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
7555 * config/nds32/nds32.h: Clean up obsolete macros.
7557 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
7559 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
7560 Add enum values for particular instructions.
7561 * config/nds32/nds32-intrinsic.c: Implementation of expanding
7562 particular intrinsic functions.
7563 * config/nds32/nds32-intrinsic.md: Likewise.
7564 * config/nds32/nds32_intrinsic.h: Likewise.
7565 * config/nds32/nds32.h (nds32_builtins): Likewise.
7566 * config/nds32/nds32.md (type): Add pbsad and pbsada.
7567 (btst, ave): New patterns for particular instructions.
7569 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
7571 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
7572 Add enum values for atomic load/store and memory sync.
7573 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
7575 * config/nds32/nds32-intrinsic.md: Likewise.
7576 * config/nds32/nds32_intrinsic.h: Likewise.
7577 * config/nds32/nds32.h (nds32_builtins): Likewise.
7579 2018-04-07 Jakub Jelinek <jakub@redhat.com>
7581 PR tree-optimization/85257
7582 * fold-const.c (native_encode_vector): If not all elts could fit
7583 and off is -1, return 0 rather than offset.
7584 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
7585 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
7586 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
7587 adjust buffer in native_interpret_expr call.
7589 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
7591 * config/nds32/constants.md (unspec_volatile_element): Add cache
7592 control enum values.
7593 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
7594 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
7595 * config/nds32/nds32.c (nds32_cctl_names): New.
7596 (nds32_print_operand): Handle cache control register names.
7597 * config/nds32/nds32.h (nds32_builtins): New enum values.
7598 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
7600 * config/nds32/nds32.md (type): Add mmu.
7601 * config/nds32/pipelines.md (simple_insn): Add mmu.
7603 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
7605 * config/nds32/nds32.md (type): Remove call.
7606 * config/nds32/pipelines.md (simple_insn): Likewise.
7608 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
7610 * config/nds32/constants.md (unspec_volatile_element): Add
7611 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
7612 UNSPEC_VOLATILE_FMFCFG.
7613 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
7614 description for fmfcfg and fmfcsr.
7615 (bdesc_1arg): Add fmtcsr.
7616 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
7617 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
7618 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
7619 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
7620 unspec_fmfcfg): New patterns.
7621 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
7622 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
7623 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
7624 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
7625 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
7626 __nds32__fmfcfg): Define.
7628 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
7630 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
7631 intrinsic register names.
7632 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
7633 intrinsic register enum values and macros.
7635 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
7637 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
7638 for load/store addressing form.
7639 (nds32_print_operand_address): Likewise.
7641 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
7644 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
7645 based on LABEL_REF. Remove useless assertion.
7646 (pic_address_needs_scratch): Fix formatting.
7647 (sparc_legitimize_pic_address): Minor tweaks.
7648 (sparc_delegitimize_address): Adjust assertion accordingly.
7649 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
7650 into symbolic_operand.
7651 (movsi_high_pic_label_ref): Likewise.
7652 (movsi_lo_sum_pic_label_ref): Likewise.
7653 (movdi_pic_label_ref): Likewise.
7654 (movdi_high_pic_label_ref): Likewise.
7655 (movdi_lo_sum_pic_label_ref): Likewise.
7657 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
7659 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
7660 custom LIB_SPEC setup.
7662 2018-04-06 Ruslan Bukin <br@bsdpad.com>
7663 Kito Cheng <kito.cheng@gmail.com>
7665 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
7666 * config/riscv/freebsd.h: New.
7668 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
7670 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
7671 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
7674 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
7675 Kito Cheng <kito.cheng@gmail.com>
7677 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
7678 nds32_output_call, nds32_symbol_binds_local_p): New functions.
7679 * config/nds32/nds32-protos.h (nds32_output_call,
7680 nds32_output_return): Declare.
7681 * config/nds32/nds32.md: Refine all the call and return patterns.
7683 2018-04-06 Jakub Jelinek <jakub@redhat.com>
7686 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
7687 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
7689 PR rtl-optimization/84872
7690 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
7691 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
7694 2018-04-06 Tamar Christina <tamar.christina@arm.com>
7696 * expr.c (copy_blkmode_to_reg): Revert 254862.
7697 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
7699 2018-04-06 Richard Biener <rguenther@suse.de>
7702 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
7703 after seeing a component reference with an adjacent field. Treat
7704 refs to arrays at struct end of external decls similar to
7705 refs to unconstrained commons.
7707 2018-04-06 Jakub Jelinek <jakub@redhat.com>
7710 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
7711 look through SAVE_EXPRs with non-side-effects argument. Adjust
7713 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
7716 2018-04-06 Richard Biener <rguenther@suse.de>
7719 * alias.c (find_base_term): New wrapper around find_base_term
7720 unwinding CSELIB_VAL_PTR changes.
7721 (find_base_term): Do not restore CSELIB_VAL_PTR during the
7724 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7726 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
7728 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
7729 constant definitions.
7730 ("nop"): lr 0,0 -> nopr r0
7731 ("nop_lr0", "nop_lr1"): New insn definitions.
7733 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
7735 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
7736 NDS32_V3PUSH_AVAILABLE_P macro.
7738 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
7739 Chung-Ju Wu <jasonwucj@gmail.com>
7741 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
7742 (nds32*-*-*): Add float and fpu_config into supported_defaults.
7743 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
7744 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
7745 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
7746 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
7747 * config/nds32/constraints.md: New constraints and checking for hard
7748 float configuration.
7749 * config/nds32/iterators.md: New mode iterator and attribute for hard
7750 float configuration.
7751 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
7753 * config/nds32/nds32-fpu.md: New file.
7754 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
7755 deal with hard float code generation.
7756 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
7758 (abi_type, float_reg_number): New enum type.
7759 * config/nds32/nds32-predicates.c: New predicates for hard float.
7760 * config/nds32/nds32-protos.h: Declare functions for hard float.
7761 * config/nds32/nds32.c: Implementation for hard float configuration.
7762 * config/nds32/nds32.h: Definitions for hard float configuration.
7763 * config/nds32/nds32.md: Include hard float machine description and
7764 modify patterns for hard float configuration.
7765 * config/nds32/nds32.opt: New options for hard float configuration.
7766 * config/nds32/predicates.md: New predicates for hard float
7769 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
7771 * common/config/nds32/nds32-common.c
7772 (nds32_option_optimization_table): Enable -mreleax-hint by default.
7774 2018-04-05 Jakub Jelinek <jakub@redhat.com>
7777 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
7778 CONSTRUCTOR_ELT (ctor, ...)->value.
7780 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
7783 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
7785 2018-04-05 Tom de Vries <tom@codesourcery.com>
7788 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
7791 2018-04-05 Shiva Chen <shiva0217@gmail.com>
7792 Kito Cheng <kito.cheng@gmail.com>
7794 * config/nds32/constraints.md (U33): Fine-tune checking condition.
7795 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
7796 * config/nds32/nds32.h (nds32_16bit_address_type): Add
7797 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
7799 2018-04-05 Shiva Chen <shiva0217@gmail.com>
7800 Kito Cheng <kito.cheng@gmail.com>
7802 * config/nds32/constraints.md (Ufe): New memory constraint.
7803 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
7804 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
7805 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
7807 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
7808 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
7810 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7812 * config/nds32/nds32.md: Use optimize_size in the condition for
7813 alu-shift instructions.
7815 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7817 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
7819 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7821 * config/nds32/nds32.md (negsi2): Refine pattern.
7823 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
7824 Chung-Ju Wu <jasonwucj@gmail.com>
7826 * config/nds32/iterators.md (shift_rotate): New code iterator.
7827 (shift): New code attribute.
7828 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
7829 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
7830 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
7831 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
7832 bit-wise operations.
7833 (andsi3, *andsi3): Ditto.
7834 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
7835 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
7836 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
7837 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
7838 nds32_ior_operand, nds32_xor_operand): New predicates.
7840 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7842 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
7843 (addsi3, subsi3): ... this.
7845 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7847 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
7849 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7851 * config/nds32/nds32.md: Adjust indention.
7853 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
7855 * config/nds32/nds32.md (feature): New attribute.
7857 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7859 * config/nds32/nds32.md (subtype): New attribute.
7861 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
7864 * config/arm/arm-builtins.c (arm_expand_builtin): Change
7865 expansion to perform a bitwise AND of the argument followed by a
7866 boolean negation of the result.
7868 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
7870 PR rtl-optimization/84878
7871 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
7872 the basic block. Assert the use reference is not artificial and that
7873 it has an associated insn.
7875 2018-04-04 Michael Matz <matz@suse.de>
7877 * builtins.c (compute_objsize): Pass correct operand
7878 to array_at_struct_end_p.
7880 2018-04-04 Richard Biener <rguenther@suse.de>
7883 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
7884 from contexts for DINFO_LEVEL_TERSE and below.
7886 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
7888 * config/nds32/nds32-doubleword.md (move_<mode>): Require
7889 resiter_operand condition.
7890 * config/nds32/nds32.md (*move<mode>): Ditto.
7892 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
7893 Monk Chiang <sh.chiang04@gmail.com>
7895 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
7897 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
7899 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
7901 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
7902 Kito Cheng <kito.cheng@gmail.com>
7904 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
7905 nds32_cond_code_str, output_cond_branch,
7906 output_cond_branch_compare_zero, nds32_expand_cbranch,
7907 nds32_expand_cstore, nds32_expand_movcc,
7908 nds32_output_cbranchsi4_equality_zero,
7909 nds32_output_cbranchsi4_equality_reg,
7910 nds32_output_cbranchsi4_equality_reg_or_const_int,
7911 nds32_output_cbranchsi4_greater_less_zero: New functions.
7912 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
7913 nds32_expand_cstore, nds32_expand_movcc,
7914 nds32_output_cbranchsi4_equality_zero,
7915 nds32_output_cbranchsi4_equality_reg,
7916 nds32_output_cbranchsi4_equality_reg_or_const_int,
7917 nds32_output_cbranchsi4_greater_less_zero): Declare.
7918 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
7919 nds32_rimm11s_operand): New predicates.
7920 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
7921 * config/nds32/nds32.md: Rewrite all the branch and conditional move
7924 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
7926 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
7927 * config/nds32/nds32.md: Ditto.
7928 * config/nds32/pipelines.md: Ditto.
7930 2018-04-04 Richard Biener <rguenther@suse.de>
7932 PR tree-optimization/85168
7933 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
7934 propagating abnormals.
7936 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
7938 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
7940 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
7941 Kito Cheng <kito.cheng@gmail.com>
7943 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
7944 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
7945 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
7946 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
7947 * config/nds32/nds32.md (sibcall_internal): New.
7948 (sibcall_register): Remove.
7949 (sibcall_immediate): Remove.
7950 (sibcall_value_internal): New.
7951 (sibcall_value_register): Remove.
7952 (sibcall_value_immediate): Remove.
7953 * config/nds32/predicates.md (nds32_general_register_operand): New.
7954 (nds32_call_address_operand): New.
7956 2018-04-03 Jakub Jelinek <jakub@redhat.com>
7958 PR rtl-optimization/85167
7959 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
7960 bb_defs if *split_p, instead preinitialize it to NULL.
7962 PR tree-optimization/85156
7963 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
7964 evaluating the argument multiple times.
7966 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
7968 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
7970 (_mm_cvtpd_ps): Likewise.
7971 (_mm_cvttpd_epi32): Likewise.
7972 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
7973 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
7974 vector, pixel, and bool following altivec.h include.
7976 2018-04-03 Martin Sebor <msebor@redhat.com>
7978 * doc/extend.texi (Common Function Attributes): Clarify.
7979 (const attribute): Likewise.
7980 (pure attribute): Likewise.
7982 2018-04-03 Jakub Jelinek <jakub@redhat.com>
7985 * config/i386/i386.c (ix86_expand_vector_set): Use
7986 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
7988 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
7990 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
7991 instructions when changing rounding bits to preserve precision bits
7992 in the x87 control word.
7994 2018-04-03 Martin Liska <mliska@suse.cz>
7996 PR tree-optimization/82491
7997 * rtl.h (strip_offset_and_add): Replace += suboffset with
7998 poly_uint64 () + suboffset.
8000 2018-03-29 Martin Liska <mliska@suse.cz>
8001 Martin Jambor <mjambor@suse.cz>
8004 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
8005 param_type is not an integral or pointer type.
8007 2018-04-03 Richard Biener <rguenther@suse.de>
8009 * sese.h (recompute_all_dominators): Remove.
8011 2018-04-02 Martin Sebor <msebor@redhat.com>
8013 * doc/invoke.texi (-Wrestrict): Fix typos.
8015 2018-04-02 Jim Wilson <jimw@sifive.com>
8017 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
8018 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
8019 (<optab>di3, <optab>si3_extend): Likewise.
8020 (<optab>si3_mask, <optab>si3_mask_1): New.
8021 (<optab>di3_mask, <optab>di3_mask_1): New.
8022 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
8023 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
8024 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
8026 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
8028 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
8031 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
8033 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
8034 (nds32_canonicalize_comparison): New function.
8036 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
8037 Kito Cheng <kito.cheng@gmail.com>
8038 Kuan-Lin Chen <kuanlinchentw@gmail.com>
8040 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
8041 * config/nds32/constants.md (unspec_volatile_element): Add
8042 UNSPEC_VOLATILE_RELAX_GROUP.
8043 * config/nds32/nds32-relax-opt.c: New file.
8044 * config/nds32/nds32-predicates.c
8045 (nds32_symbol_load_store_p): New function.
8046 * config/nds32/nds32-protos.h
8047 (nds32_symbol_load_store_p): Declare function.
8048 (make_pass_nds32_relax_opt): Declare new rtl pass function.
8049 * config/nds32/nds32.c
8050 (nds32_register_pass): New function to register pass.
8051 (nds32_register_passes): New function to register passes.
8052 * config/nds32/nds32.md (relax_group): New pattern.
8053 * config/nds32/nds32.opt (mrelax-hint): New option.
8054 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
8056 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
8058 * config/nds32/t-nds32: Modify files dependency.
8060 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
8062 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
8063 (PROFILE_HOOK): Define its implementation.
8065 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
8067 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
8068 type and 32-bit size.
8070 2018-04-01 Jakub Jelinek <jakub@redhat.com>
8073 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
8074 (V_128_256): New mode iterator.
8075 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
8076 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
8077 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
8079 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
8080 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
8082 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
8085 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
8086 NaN inputs correctly.
8088 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
8091 * config/rs6000/vsx.md (??r): New mode attribute.
8092 (*vsx_mov<mode>_64bit): Use it.
8093 (*vsx_mov<mode>_32bit): Likewise.
8095 2018-03-30 Martin Sebor <msebor@redhat.com>
8097 PR tree-optimization/84818
8098 * builtins.c (check_access): Use warning_n.
8100 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8103 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
8105 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
8108 2018-03-30 Julia Koval <julia.koval@intel.com>
8111 * x86-tune.def (movx, partial_reg_dependency): Enable for
8114 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
8117 * lra-constraints.c (process_alt_operands): Move setting
8118 this_alternative_matches below.
8120 2018-03-29 Martin Liska <mliska@suse.cz>
8123 * doc/invoke.texi: Document how LTO works with debug info.
8124 Describe auto-load support of binutils. Mention 'x86-64'
8125 as valid option value of -march option.
8127 2018-03-29 Jakub Jelinek <jakub@redhat.com>
8129 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
8132 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
8133 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
8134 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
8137 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
8140 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
8141 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
8142 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
8143 for RS6000_BTM_POWERPC64.
8144 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
8145 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
8146 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
8151 2018-03-28 Carl Love <cel@us.ibm.com>
8154 2017-09-27 Carl Love <cel@us.ibm.com>
8156 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
8157 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
8158 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
8161 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8163 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
8164 instead of __vector bool.
8165 (_mm_max_pu8): Likewise.
8166 (_mm_min_pi16): Likewise.
8168 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
8171 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
8172 (DIVWEUO): Likewise.
8174 (DIVDEUO): Likewise.
8175 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
8176 DIVWEUO and DIVDEUO.
8177 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
8178 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
8179 (div_extend): Likewise.
8180 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
8182 (__builtin_divweuo): Likewise.
8183 (__builtin_divdeo): Likewise.
8184 (__builtin_divdeuo): Likewise.
8186 2018-03-28 Jakub Jelinek <jakub@redhat.com>
8189 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
8190 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
8192 PR tree-optimization/82004
8193 * gimple-match-head.c (optimize_pow_to_exp): New function.
8194 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
8195 Don't fold to exp if optimize_pow_to_exp is false.
8197 2018-03-28 Martin Liska <mliska@suse.cz>
8200 * calls.c (initialize_argument_information): Fix trailing space.
8201 * common.opt: Fix typo and provide better explanation for
8202 -fsanitize-coverage option.
8203 * config/i386/i386.opt: Fix typo.
8205 2018-03-28 Jakub Jelinek <jakub@redhat.com>
8206 Martin Liska <mliska@suse.cz>
8209 * gimplify.c (asan_poison_variable): Don't do the check for
8210 gimplify_omp_ctxp here.
8211 (gimplify_decl_expr): Do it here.
8212 (gimplify_target_expr): Likewise.
8214 2018-03-28 Martin Liska <mliska@suse.cz>
8217 * config/i386/i386.c (ix86_function_arg_advance): Do not call
8218 chkp_type_bounds_count if MPX is not enabled.
8220 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
8222 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
8224 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
8227 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
8228 function to create the function decl for complex long double
8229 multiply and divide for -mabi=ieeelongdouble.
8230 (init_float128_ieee): Call it.
8232 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
8235 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
8236 -fcf-protection=branch -mibt.
8237 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
8239 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
8242 * config/arm/arm.c (arm_valid_symbolic_address): Handle
8243 arm_word_relocations.
8245 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
8248 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
8249 extern array declarations.
8251 2018-03-27 Richard Biener <rguenther@suse.de>
8254 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
8255 explicit single_use checks.
8257 2018-03-27 Richard Biener <rguenther@suse.de>
8259 PR tree-optimization/85082
8260 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
8263 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
8265 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
8266 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
8267 Turn on fasynchronous-unwind-tables and funwind-tables.
8269 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
8272 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
8273 (*bmi_blsr_<mode>_ccz): Ditto.
8275 2018-03-26 Tom de Vries <tom@codesourcery.com>
8277 PR tree-optimization/85063
8278 * omp-general.c (offloading_function_p): New function. Factor out
8280 * omp-offload.c (pass_omp_target_link::gate): ... here.
8281 * omp-general.h (offloading_function_p): Declare.
8282 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
8283 with attribute omp declare target for offloading functions.
8285 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
8287 PR tree-optimization/84005
8288 * tree-data-ref.h (get_base_for_alignment): Declare.
8289 * tree-data-ref.c (get_base_for_alignment_1): New function.
8290 (get_base_for_alignment): Likewise.
8291 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
8292 get_base_for_alignment to find a suitable base object, instead
8293 of always using drb->base_address.
8295 2018-03-23 Jakub Jelinek <jakub@redhat.com>
8298 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
8299 known size by default.
8301 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
8304 * lra-constraints.c (process_alt_operands): Don't match BLKmode
8305 and non BLKmode operands.
8307 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8310 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
8311 Clean up attributes.
8313 2018-03-23 Richard Biener <rguenther@suse.de>
8316 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
8317 we are going to emit early debug for LTO.
8319 2018-03-23 Jakub Jelinek <jakub@redhat.com>
8322 * function.c (match_asm_constraints_1): Don't optimize if input
8323 doesn't satisfy general_operand predicate for output's mode.
8326 * alias.c (write_dependence_p): Don't require for x_canonicalized
8327 non-VOIDmode if x has VOIDmode.
8330 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
8331 just don't try to optimize it rather than assert it never happens.
8333 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
8335 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
8336 macro expansions for definition of ST_INTERNAL_<mode> and
8337 LD_INTERNAL_<mode> builtins.
8338 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
8340 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
8342 (altivec_expand_st_builtin): Likewise.
8343 (altivec_expand_builtin): Remove calls to deleted functions.
8344 (rs6000_address_for_altivec): Delete this function.
8345 * config/rs6000/vector.md: Remove expands for
8346 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
8348 2018-03-22 Sudakshina Das <sudi.das@arm.com>
8351 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
8352 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
8353 re-computing once computed.
8354 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
8355 (arm_init_machine_status): Initialize
8356 machine->static_chain_stack_bytes.
8358 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
8361 * doc/extend.texi: Add four new prototypes for vec_ld.
8362 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
8363 definitions for more logical presentation.
8364 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
8365 entries for V1TI variants of __builtin_altivec_ld builtin.
8366 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
8367 handling of V1TI variant of LVX icode pattern.
8368 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
8369 (rs6000_gimple_fold_builtin): Likewise.
8370 (altivec_init_builtins): Add code to define
8371 __builtin_altivec_lvx_v1ti function.
8373 2018-03-22 Jakub Jelinek <jakub@redhat.com>
8376 * function.c (match_asm_constraints_1): Don't do the optimization
8377 if input isn't a REG, SUBREG, MEM or constant.
8379 2018-03-22 Tom de Vries <tom@codesourcery.com>
8381 PR tree-optimization/84956
8382 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
8383 bb_has_abnormal_pred.
8385 2018-03-22 Jakub Jelinek <jakub@redhat.com>
8388 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
8389 DECL_INITIAL (decl) to decl at the end.
8390 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
8393 2018-03-21 Joseph Myers <joseph@codesourcery.com>
8395 * doc/extend.texi (__builtin_tgmath): Document when complex
8396 integer types are treated as _Complex _Float64.
8398 2018-03-21 Tom de Vries <tom@codesourcery.com>
8400 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
8402 2018-03-21 Jakub Jelinek <jakub@redhat.com>
8404 PR tree-optimization/84960
8405 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
8406 if it is ENTRY block, move them into single succ of ENTRY in that case.
8408 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
8410 PR tree-optimization/84811
8411 * poly-int.h (poly_span_traits): Remove the T3 parameter and
8412 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
8413 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
8414 (known_subrange_p): Update accordingly. Cast each value involved
8415 in the size comparison, rather than casting the result of the
8418 2018-03-21 Jakub Jelinek <jakub@redhat.com>
8420 PR tree-optimization/84982
8421 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
8422 by flipping the least significant bit rather than all bits from
8423 bitpos to bitpos + bitsize - 1.
8425 2018-03-21 Nathan Sidwell <nathan@acm.org>
8427 * doc/extend.texi (Deprecated Features): Remove mention of
8428 long-deleted deprecations.
8430 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8433 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
8434 * configure: Regenerate.
8436 2018-03-21 Tom de Vries <tom@codesourcery.com>
8438 PR tree-optimization/83126
8439 * tree-parloops.c (num_phis): New function.
8440 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
8442 2018-03-21 Nathan Sidwell <nathan@acm.org>
8444 * doc/extend.texi (Deprecated Features): Update deprecated flags,
8445 mention anon-struct/union members and trailing attributes.
8447 2018-03-21 Bin Cheng <bin.cheng@arm.com>
8449 PR tree-optimization/84969
8450 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
8451 builtin memset partitions if they set different rhs values.
8453 2018-03-21 Jakub Jelinek <jakub@redhat.com>
8455 PR rtl-optimization/84989
8456 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
8457 VEC_DUPLICATE with scalar result mode.
8459 2018-03-21 Martin Liska <mliska@suse.cz>
8462 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
8463 not intended return statement.
8465 2018-03-21 Martin Liska <mliska@suse.cz>
8468 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
8469 (chkp_find_bound_slots_1): Limit number of iterations.
8471 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
8474 * Minor grammar fixes for x86 options.
8476 2018-03-20 Jakub Jelinek <jakub@redhat.com>
8479 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
8480 holding REG_CFA_RESTORE notes, instead turn them into a USE.
8482 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
8485 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
8486 (altivec_lvx_<mode>_1op): Likewise.
8487 (altivec_stvx_<mode>_2op): Likewise.
8488 (altivec_stvx_<mode>_1op): Likewise.
8489 (altivec_lvx_<VM2:mode>): New define_expand.
8490 (altivec_stvx_<VM2:mode>): Likewise.
8491 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
8492 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
8493 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
8494 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
8495 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
8496 (rs6000_gen_lvx): Likewise.
8497 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
8498 (altivec_expand_stv_builtin): Likewise.
8499 (altivec_expand_builtin): Likewise.
8500 * config/rs6000/vector.md: Likewise.
8502 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8505 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
8508 2018-03-20 Richard Biener <rguenther@suse.de>
8511 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
8512 sign-conversions as zero, fall back to standard scalar_stmt
8515 2018-03-20 Martin Liska <mliska@suse.cz>
8518 * predict.c (rebuild_frequencies): Handle case when we have
8519 PROFILE_ABSENT, but flag_guess_branch_prob is false.
8521 2018-03-20 Jakub Jelinek <jakub@redhat.com>
8524 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
8525 flag_section_anchors.
8526 * varasm.c (use_blocks_for_decl_p): Remove hack for
8527 dw2_force_const_mem.
8530 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
8532 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
8533 be created, use lowpart_subreg of operands[0] rather than operands[0]
8535 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
8536 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
8537 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
8538 and n constraint instead of aarch64_shift_imm_di and Usd.
8539 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
8540 (*aarch64_<optab>_reg_minus<mode>3): ... this.
8542 2018-03-20 Sudakshina Das <sudi.das@arm.com>
8545 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
8546 to favor GPR over NEON registers.
8547 (<shift>di3_neon): Likewise.
8549 2018-03-20 Tom de Vries <tom@codesourcery.com>
8552 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
8553 (nvptx_process_pars): Emit bar.sync asap and alap.
8555 2018-03-20 Tom de Vries <tom@codesourcery.com>
8558 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
8559 seen_label if seen_label is already set.
8561 2018-03-20 Jakub Jelinek <jakub@redhat.com>
8564 * config/i386/i386.c (fold_builtin_cpu): For features above 31
8565 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
8566 Use 1U instead of 1. Formatting fixes.
8569 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
8570 instead of TREE_TYPE (s1) for the return value.
8572 2018-03-19 Jakub Jelinek <jakub@redhat.com>
8574 PR tree-optimization/84946
8575 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
8576 bitsize + bitsize in poly_uint64 rather than poly_int64.
8579 * dwarf2asm.c: Include fold-const.c.
8580 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
8581 of decl rather than decl itself.
8583 PR rtl-optimization/84643
8584 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
8586 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
8589 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
8590 calling assemble_variable.
8592 2018-03-19 Sudakshina Das <sudi.das@arm.com>
8595 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
8596 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
8598 2018-03-19 Jim Wilson <jimw@sifive.com>
8601 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
8602 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
8603 (riscv_first_stack_step): Likewise.
8604 (riscv_option_override): Use STACK_BOUNDARY instead of
8606 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
8608 (BIGGEST_ALIGNMENT): Set to 128.
8609 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
8610 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
8613 2018-03-19 Richard Biener <rguenther@suse.de>
8615 PR tree-optimization/84933
8616 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
8617 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
8619 2018-03-19 Richard Biener <rguenther@suse.de>
8621 PR tree-optimization/84859
8622 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
8623 (cond_if_else_store_replacement): Perform sinking operation on
8624 single-store BBs regardless of MAX_STORES_TO_SINK setting.
8625 Generalize what a BB with a single eligible store is.
8627 2018-03-19 Richard Biener <rguenther@suse.de>
8629 PR tree-optimization/84929
8630 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
8631 chrec_is_positive against non-chrec arg.
8633 2018-03-19 Tamar Christina <tamar.christina@arm.com>
8636 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
8638 2018-03-18 Martin Liska <mliska@suse.cz>
8640 PR rtl-optimization/84635
8641 * regrename.c (build_def_use): Use matches_mode only when
8644 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
8646 PR tree-optimization/84913
8647 * tree-vect-loop.c (vectorizable_reduction): Don't try to
8648 vectorize chains of COND_EXPRs.
8650 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
8652 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
8654 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
8656 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
8658 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
8660 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
8662 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
8663 Kito Cheng <kito.cheng@gmail.com>
8665 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
8666 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
8667 (nds32_adjust_reg_alloc_order): New function.
8668 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
8670 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
8672 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
8673 nds32_print_operand, nds32_print_operand_address): Use
8674 HOST_WIDE_INT_PRINT_DEC instead.
8676 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
8678 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
8680 2018-03-17 Jakub Jelinek <jakub@redhat.com>
8683 * config/i386/i386.c (initial_ix86_tune_features,
8684 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
8686 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
8687 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
8688 rather than 1u << ix86_tune. Formatting fix.
8689 (ix86_option_override_internal): Change ix86_arch_mask from
8690 unsigned int to unsigned HOST_WIDE_INT, initialize to
8691 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
8692 (ix86_function_specific_restore): Likewise.
8694 2018-03-16 Jakub Jelinek <jakub@redhat.com>
8697 * postreload.c (reload_combine_recognize_pattern): Perform
8698 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
8699 truncate_int_for_mode the result for the destination's mode.
8702 * hsa-gen.c (mem_type_for_type): Fix comment typo.
8703 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
8705 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
8708 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
8711 * lra-assigns.c (lra_split_hard_reg_for): Don't use
8712 regno_allocno_class_array and sorted_pseudos.
8713 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
8714 insns where regno is used.
8716 2018-03-16 Martin Liska <mliska@suse.cz>
8719 * multiple_target.c (create_dispatcher_calls): Redirect
8720 reference in the symbol table.
8722 2018-03-16 Martin Liska <mliska@suse.cz>
8725 * multiple_target.c (create_dispatcher_calls): Redirect also
8728 2018-03-16 Jakub Jelinek <jakub@redhat.com>
8732 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
8733 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
8734 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
8736 2018-03-16 Julia Koval <julia.koval@intel.com>
8738 * doc/invoke.texi (Skylake Server): Add CLWB.
8739 Cannonlake): Remove CLWB.
8741 2018-03-16 Jakub Jelinek <jakub@redhat.com>
8743 PR tree-optimization/84841
8744 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
8746 (FLOAT_ONE_CONST_TYPE): Define.
8747 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
8748 (sort_by_operand_rank): Put entries with higher constant_type last
8749 rather than first to match comments.
8751 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
8753 * config/nios2/nios2.md (movsi_internal): Fix thinko in
8756 2018-03-15 Jakub Jelinek <jakub@redhat.com>
8759 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
8760 check and use address of target always.
8762 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
8765 * config/i386/i386.c (indirect_thunk_needed): Update comments.
8766 (indirect_thunk_bnd_needed): Likewise.
8767 (indirect_thunks_used): Likewise.
8768 (indirect_thunks_bnd_used): Likewise.
8769 (indirect_return_needed): New.
8770 (indirect_return_bnd_needed): Likewise.
8771 (output_indirect_thunk_function): Add a bool argument for
8773 (output_indirect_thunk_function): Don't generate alias for
8774 function return thunk.
8775 (ix86_code_end): Call output_indirect_thunk_function to generate
8776 function return thunks.
8777 (ix86_output_function_return): Set indirect_return_bnd_needed
8778 and indirect_return_needed instead of indirect_thunk_bnd_needed
8779 and indirect_thunk_needed.
8781 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
8783 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
8784 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
8785 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
8787 2018-03-15 David Malcolm <dmalcolm@redhat.com>
8788 Paul Hua <paul.hua.gm@gmail.com>
8791 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
8793 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
8795 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
8796 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
8799 2018-03-15 Tamar Christina <tamar.christina@arm.com>
8802 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
8803 instead of GET_MODE_SIZE when comparing Units.
8805 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
8808 * varasm.c (hash_section): Return an unchangeble hash value
8809 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
8810 Return !aarch64_can_use_per_function_literal_pools_p ().
8812 2018-03-15 Jakub Jelinek <jakub@redhat.com>
8815 * optabs.c (emit_conditional_move): Pass address of cmode's copy
8816 rather than address of cmode as last argument to prepare_cmp_insn.
8818 2018-03-15 Julia Koval <julia.koval@intel.com>
8820 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
8821 F_AVX512VNNI, F_AVX512BITALG): New.
8823 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
8826 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
8827 insn for floating-point loads and stores.
8829 2018-03-14 Carl Love <cel@us.ibm.com>
8831 * config/rs6000/rs6000-c.c: Add macro definitions for
8832 ALTIVEC_BUILTIN_VEC_PERMXOR.
8833 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
8834 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
8835 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
8836 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
8838 * config/doc/extend.texi: Add prototypes for vec_permxor.
8840 2018-03-14 David Malcolm <dmalcolm@redhat.com>
8843 * diagnostic-show-locus.c (class layout_point): Convert m_line
8844 from int to linenum_type.
8845 (line_span::comparator): Use linenum "compare" function when
8846 comparing line numbers.
8847 (test_line_span): New function.
8848 (layout_range::contains_point): Convert param "row" from int to
8850 (layout_range::intersects_line_p): Likewise.
8851 (layout::will_show_line_p): Likewise.
8852 (layout::print_source_line): Likewise.
8853 (layout::should_print_annotation_line_p): Likewise.
8854 (layout::print_annotation_line): Likewise.
8855 (layout::print_leading_fixits): Likewise.
8856 (layout::annotation_line_showed_range_p): Likewise.
8857 (struct line_corrections): Likewise for field m_row.
8858 (line_corrections::line_corrections): Likewise for param "row".
8859 (layout::print_trailing_fixits): Likewise.
8860 (layout::get_state_at_point): Likewise.
8861 (layout::get_x_bound_for_row): Likewise.
8862 (layout::print_line): Likewise.
8863 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
8864 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
8865 * input.c (selftest::test_linenum_comparisons): New function.
8866 (selftest::input_c_tests): Call it.
8867 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
8868 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
8869 * selftest.h (ASSERT_GT): New macro.
8870 (ASSERT_GT_AT): New macro.
8871 (ASSERT_LT): New macro.
8872 (ASSERT_LT_AT): New macro.
8874 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
8876 PR rtl-optimization/84780
8877 * combine.c (distribute_links): Don't make a link based on pc_rtx.
8879 2018-03-14 Martin Liska <mliska@suse.cz>
8881 * tree.c (record_node_allocation_statistics): Use
8882 get_stats_node_kind.
8883 (get_stats_node_kind): New function extracted from
8884 record_node_allocation_statistics.
8885 (free_node): Use get_stats_node_kind.
8887 2018-03-14 Richard Biener <rguenther@suse.de>
8889 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
8890 that the value-set of ANTIC_IN doesn't grow.
8893 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
8895 (BB_VISITED_WITH_VISITED_SUCCS): New define.
8896 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
8898 2018-03-14 Julia Koval <julia.koval@intel.com>
8900 * config.gcc (icelake-client, icelake-server): New.
8902 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
8903 (initial_ix86_arch_features): Ditto.
8904 (PTA_SKYLAKE): Add SGX.
8905 (PTA_ICELAKE): Remove.
8906 (PTA_ICELAKE_CLIENT): New.
8907 (PTA_ICELAKE_SERVER): New.
8908 (ix86_option_override_internal): Split up icelake on icelake client and
8910 (get_builtin_code_for_version): Ditto.
8911 (fold_builtin_cpu): Ditto.
8912 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
8913 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
8914 * config/i386/i386.h (processor_type): Ditto.
8915 * doc/invoke.texi: Ditto.
8917 2018-03-14 Jakub Jelinek <jakub@redhat.com>
8920 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
8921 INTEGER_CST offset, add it together with bitpos / 8 and
8922 sign extend based on POINTER_SIZE.
8926 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
8929 * config/i386/constraints.md (Yc): New register constraint.
8930 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
8931 Use Yc constraint for alternative 2 of operand 0. Remove
8932 preferred_for_speed attribute.
8934 2018-03-14 Richard Biener <rguenther@suse.de>
8936 PR tree-optimization/84830
8937 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
8938 with the old one to avoid oscillations.
8940 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
8943 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
8945 (assign_by_spills): Return a flag of reload assignment failure.
8946 Do not process the reload assignment failures. Do not spill other
8947 reload pseudos if they has the same reg class. Update n if
8949 (lra_assign): Add a return arg. Set up from the result of
8950 assign_by_spills call.
8951 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
8952 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
8953 usage_insns if it is not NULL.
8954 (spill_hard_reg_in_range): New function.
8955 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
8956 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
8957 function prototypes.
8958 (lra_assign): Change prototype.
8959 * lra.c (lra): Add code to deal with fails by splitting hard reg
8962 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
8964 * config/riscv/riscv.opt (mrelax): New option.
8965 * config/riscv/riscv.c (riscv_file_start): Emit ".option
8966 "norelax" when riscv_mrelax is disabled.
8967 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
8969 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8972 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
8973 reassociation for int modes.
8975 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
8977 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
8978 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
8980 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
8981 * config/aarch64/aarch64-sve.md
8982 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
8983 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
8984 (*extend<mode><Vwide>2): Rename to...
8985 (aarch64_sve_extend<mode><Vwide>2): ...this.
8986 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
8987 renaming the old pattern to...
8988 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
8990 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
8991 define_expand, renaming the old pattern to...
8992 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
8993 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
8994 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
8995 account when deciding which SVE instruction the optab should use.
8996 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
8998 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
9000 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
9001 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
9002 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
9003 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
9004 (tlsdesc_small_<mode>): Turn a define_expand and use
9005 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
9006 (tlsdesc_small_advsimd_<mode>): ...this.
9007 (tlsdesc_small_sve_<mode>): New pattern.
9009 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
9011 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
9012 (UNSPEC_UMUL_HIGHPART): New constants.
9013 (MUL_HIGHPART): New int iteraor.
9014 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
9015 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
9017 (*<su>mul<mode>3_highpart): New define_insn.
9019 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
9022 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
9025 2018-03-13 Martin Liska <mliska@suse.cz>
9028 * (sem_item_optimizer::sem_item_optimizer): Initialize new
9030 (sem_item_optimizer::~sem_item_optimizer): Release it.
9031 (sem_item_optimizer::merge_classes): Register variable aliases.
9032 (sem_item_optimizer::fixup_pt_set): New function.
9033 (sem_item_optimizer::fixup_points_to_sets): Likewise.
9034 * ipa-icf.h: Declare new variables and functions.
9036 2018-03-13 Jakub Jelinek <jakub@redhat.com>
9039 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
9040 integer_pow2p@2 and test integer_pow2p in condition.
9041 (A < 0 ? C : 0): Similarly for @1.
9044 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
9045 characters starting at p contain '\0' character, don't look beyond
9049 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
9050 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
9053 * reg-stack.c (change_stack): Change update_end var from int to
9054 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
9055 also call set_block_for_insn on the newly added insns and rescan.
9058 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
9059 on the last operand.
9062 * tree.c (stabilize_reference_1): Return save_expr (e) for
9063 STATEMENT_LIST even if it doesn't have side-effects.
9065 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
9067 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
9069 2018-03-12 Renlin Li <renlin.li@arm.com>
9071 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
9072 aarch64_output_scalar_simd_mov_immediate.
9074 2018-03-12 Martin Sebor <msebor@redhat.com>
9076 PR tree-optimization/83456
9077 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
9078 for perfectly overlapping calls to memcpy.
9079 (gimple_fold_builtin_memory_chk): Same.
9080 (gimple_fold_builtin_strcpy): Handle no-warning.
9081 (gimple_fold_builtin_stxcpy_chk): Same.
9082 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
9084 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
9086 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
9087 parameter. Use it for SFmode.
9088 (rs6000_function_arg_advance_1): Adjust.
9089 (rs6000_function_arg): Adjust.
9090 (rs6000_gimplify_va_arg): Pass false for that new parameter.
9092 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
9094 PR rtl-optimization/84169
9095 PR rtl-optimization/84780
9096 * combine.c (can_combine_p): Check for a 2-insn combination whether
9097 the destination register is used between the two insns, too.
9099 2018-03-12 Richard Biener <rguenther@suse.de>
9101 PR tree-optimization/84803
9102 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
9103 for refs DR analysis didn't process.
9105 2018-03-12 Richard Biener <rguenther@suse.de>
9107 PR tree-optimization/84777
9108 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
9109 force-vectorize loops ignore whether we are optimizing for size.
9111 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
9113 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
9114 (TARGET_MD_ASM_ADJUST): Define.
9116 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
9117 Kito Cheng <kito.cheng@gmail.com>
9118 Chung-Ju Wu <jasonwucj@gmail.com>
9120 * config/nds32/nds32.c (nds32_compute_stack_frame,
9121 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
9122 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
9123 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
9124 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
9125 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
9126 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
9127 * config/nds32/nds32.md (prologue, epilogue): Use macro
9128 NDS32_V3PUSH_AVAILABLE_P to do checking.
9130 2018-03-11 Jakub Jelinek <jakub@redhat.com>
9133 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
9134 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
9135 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
9136 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
9137 addition of most attributes on !orig_type_die or the attribute not
9138 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
9140 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
9141 Chung-Ju Wu <jasonwucj@gmail.com>
9143 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
9145 * config/nds32/nds32.opt (mvh): New option.
9147 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
9148 Chung-Ju Wu <jasonwucj@gmail.com>
9150 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
9152 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
9153 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
9156 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
9157 Chung-Ju Wu <jasonwucj@gmail.com>
9159 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
9161 * config/nds32/nds32-multiple.md (strlensi): New pattern.
9162 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
9164 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
9165 Kito Cheng <kito.cheng@gmail.com>
9166 Chung-Ju Wu <jasonwucj@gmail.com>
9168 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
9169 UNSPEC_FFMISM and UNSPEC_FLMISM.
9170 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
9171 for ffb, ffmism and flmism.
9172 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
9173 (unspec_ffmism): Ditto.
9174 (unspec_flmism): Ditto.
9175 (nds32_expand_builtin_impl): Check if string extension is available.
9176 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
9177 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
9179 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
9182 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
9185 * lra-assigns.c (assign_by_spills): Return a flag of reload
9186 assignment failure. Do not process the reload assignment
9187 failures. Do not spill other reload pseudos if they has the same
9189 (lra_assign): Add a return arg. Set up from the result of
9190 assign_by_spills call.
9191 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
9192 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
9193 usage_insns if it is not NULL.
9194 (spill_hard_reg_in_range): New function.
9195 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
9196 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
9197 function prototypes.
9198 (lra_assign): Change prototype.
9199 * lra.c (lra): Add code to deal with fails by splitting hard reg
9202 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
9205 * config/i386/i386.opt: Replace Enforcment with Enforcement.
9207 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
9210 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
9211 (dw_val_node): Add val_symbolic_view.
9212 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
9213 (symview_upper_bound): New.
9214 (new_line_info_table): Initialize symviews_since_reset.
9215 (dwarf2out_source_line): Count symviews_since_reset and set
9216 symview_upper_bound.
9217 (dw_val_equal_p): Handle symview.
9218 (add_AT_symview): New.
9219 (print_dw_val): Handle symview.
9220 (attr_checksum, attr_checksum_ordered): Likewise.
9221 (same_dw_val_p, size_of_die): Likewise.
9222 (value_format, output_die): Likewise.
9223 (add_high_low_attributes): Use add_AT_symview for entry_view.
9224 (dwarf2out_finish): Reset symview_upper_bound, clear
9227 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
9230 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
9231 Add strict argument and use it.
9232 (rs6000_split_multireg_move): Update for new strict argument.
9233 (mem_operand_gpr): Disallow all non-offsettable addresses.
9234 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
9236 2018-03-09 Jakub Jelinek <jakub@redhat.com>
9239 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
9240 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
9241 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
9244 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
9245 decl, use remap_type if we want to use the type.
9247 2018-03-09 Martin Sebor <msebor@redhat.com>
9249 PR tree-optimization/84526
9250 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
9252 (builtin_access::generic_overlap): Be prepared to handle non-array
9255 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
9257 PR rtl-optimization/84682
9258 * lra-constraints.c (process_address_1): Check is_address flag
9259 for address constraints.
9260 (process_alt_operands): Likewise.
9261 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
9262 preprocess_constraints.
9263 * recog.h (preprocess_constraints): Add oploc parameter.
9265 * recog.c (preprocess_constraints): Test address_operand for
9266 CT_ADDRESS constraints.
9268 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
9271 * lra-assigns.c (assign_by_spills): Return a flag of reload
9272 assignment failure. Do not process the reload assignment
9273 failures. Do not spill other reload pseudos if they has the same
9275 (lra_assign): Add a return arg. Set up from the result of
9276 assign_by_spills call.
9277 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
9278 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
9279 usage_insns if it is not NULL.
9280 (spill_hard_reg_in_range): New function.
9281 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
9282 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
9283 function prototypes.
9284 (lra_assign): Change prototype.
9285 * lra.c (lra): Add code to deal with fails by splitting hard reg
9288 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9291 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
9292 Accept complain bool parameter. Only emit errors if it is true.
9293 (arm_parse_cpu_option_name): Likewise.
9294 (arm_target_thumb_only): Adjust callers of the above.
9295 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
9296 prototype to take a default true bool parameter.
9297 (arm_parse_arch_option_name): Likewise.
9299 2018-03-09 David Malcolm <dmalcolm@redhat.com>
9300 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
9304 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
9305 * configure: Regenerate.
9306 * configure.ac ("linker --version-script option"): New.
9307 ("linker soname option"): New.
9309 2018-03-09 Richard Biener <rguenther@suse.de>
9311 PR tree-optimization/84775
9312 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
9313 immediate uses of predicate stmts and mark them modified.
9316 PR tree-optimization/84178
9317 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
9319 (version_loop_for_if_conversion): Delay update_ssa call.
9320 (tree_if_conversion): Delay update_ssa until after predicate
9323 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
9326 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
9327 when the function accesses prior frames.
9329 2018-03-08 Jakub Jelinek <jakub@redhat.com>
9332 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
9333 gen_llsym, otherwise call maybe_gen_llsym.
9336 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
9337 has ',' character inside of it.
9339 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9342 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
9343 as clobbering CC_REGNUM.
9345 2018-03-08 Richard Biener <rguenther@suse.de>
9348 * tree-scalar-evolution.c: Include tree-into-ssa.h.
9349 (follow_copies_to_constant): Do not follow SSA names registered
9352 2018-03-08 Richard Biener <rguenther@suse.de>
9354 PR tree-optimization/84178
9355 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
9357 (version_loop_for_if_conversion): Delay update_ssa call.
9358 (tree_if_conversion): Delay update_ssa until after predicate
9361 2018-03-08 David Malcolm <dmalcolm@redhat.com>
9363 PR tree-optimization/84178
9364 * tree-if-conv.c (release_bb_predicate): Remove the
9365 the assertion that the stmts have NULL use_ops.
9366 Discard the statements, asserting that they haven't
9367 yet been added to a BB.
9369 2018-03-08 Richard Biener <rguenther@suse.de>
9371 PR tree-optimization/84746
9372 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
9373 (phi_translate): Pass in destination ANTIC_OUT set.
9374 (phi_translate_1): Likewise. For a simplified result lookup
9375 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
9376 (phi_translate_set): Adjust.
9377 (do_pre_regular_insertion): Likewise.
9378 (do_pre_partial_partial_insertion): Likewise.
9380 2018-03-08 Martin Liska <mliska@suse.cz>
9382 PR gcov-profile/84735
9383 * doc/gcov.texi: Document usage of profile files.
9384 * gcov-io.h: Document changes in the format.
9386 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
9390 * dwarf2out.c (struct dw_line_info_table): Update comments for
9392 (FORCE_RESET_NEXT_VIEW): New.
9393 (FORCE_RESETTING_VIEW_P): New.
9394 (RESETTING_VIEW_P): Check for -1 too.
9395 (ZERO_VIEW_P): Likewise.
9396 (new_line_info_table): Force-reset next view.
9397 (dwarf2out_begin_function): Likewise.
9398 (dwarf2out_source_line): Simplify zero_view_p initialization.
9399 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
9400 view directly. Omit view when omitting .loc at line 0.
9402 2018-03-08 Jakub Jelinek <jakub@redhat.com>
9404 PR tree-optimization/84740
9405 * tree-switch-conversion.c (process_switch): Call build_constructors
9406 only if info.phi_count is non-zero.
9408 PR tree-optimization/84739
9409 * tree-tailcall.c (find_tail_calls): Check call arguments against
9410 DECL_ARGUMENTS (current_function_decl) rather than
9411 DECL_ARGUMENTS (func) when checking for tail recursion.
9413 2018-03-07 Jakub Jelinek <jakub@redhat.com>
9415 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
9416 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
9417 Volker Reichelt's entry and add entries for people that perform
9418 GCC fuzzy testing and report numerous bugs.
9420 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
9423 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
9424 readonly data in sdata, if that is disabled.
9425 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
9426 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
9427 -mreadonly-in-sdata option.
9429 2018-03-07 Martin Sebor <msebor@redhat.com>
9431 PR tree-optimization/84468
9432 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
9433 basic block when looking for nul assignment.
9435 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
9438 * except.h (output_function_exception_table): Adjust prototype.
9439 * except.c (output_function_exception_table): Remove FNNAME parameter
9440 and add SECTION parameter. Ouput one part of the table at a time.
9441 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
9442 the first part of the exception table and emit unwind directives.
9443 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
9444 (i386_pe_seh_cold_init): Likewise.
9445 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
9446 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
9447 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
9448 (ix86_output_call_insn): Emit a nop in one more case for SEH.
9449 * config/i386/winnt.c: Include except.h.
9450 (struct seh_frame_state): Add reg_offset, after_prologue and
9451 in_cold_section fields.
9452 (i386_pe_seh_end_prologue): Set seh->after_prologue.
9453 (i386_pe_seh_cold_init): New function.
9454 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
9455 to seh->in_cold_section.
9456 (seh_emit_push): Record the offset of the push.
9457 (seh_emit_save): Record the offet of the save.
9458 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
9459 Test seh->after_prologue to disregard the epilogue.
9460 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
9461 (i386_pe_end_cold_function): New function.
9463 2018-03-07 Jakub Jelinek <jakub@redhat.com>
9466 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
9467 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
9470 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
9472 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
9473 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
9476 * multiple_target.c: Include tree-inline.h and intl.h.
9477 (expand_target_clones): Diagnose and fail if node->definition and
9478 !tree_versionable_function_p (node->decl).
9480 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
9482 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
9484 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
9485 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
9486 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
9488 2018-03-06 Jakub Jelinek <jakub@redhat.com>
9491 * combine.c (try_combine): Use reg_or_subregno instead of handling
9492 just paradoxical SUBREGs and REGs.
9494 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
9496 * config/arc/arc.c (arc_finalize_pic): Remove function.
9497 (arc_must_save_register): We use single base PIC register, remove
9498 checks to save/restore the PIC register.
9499 (arc_expand_prologue): Likewise.
9500 * config/arc/arc-protos.h (arc_set_default_type_attributes):
9502 (arc_verify_short): Likewise.
9503 (arc_attr_type): Likewise.
9504 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
9505 (walk_stores): Likewise.
9506 (arc_address_cost): Make it static.
9507 (arc_verify_short): Likewise.
9508 (branch_dest): Likewise.
9509 (arc_attr_type): Likewise.
9510 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
9511 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
9512 (arc_final_prescan_insn): Remove inserting the nops due to
9513 hardware hazards. It is done in reorg step.
9514 (insn_length_variant_t): Remove.
9515 (insn_length_parameters_t): Likewise.
9516 (arc_insn_length_parameters): Likewise.
9517 (arc_get_insn_variants): Likewise.
9518 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
9520 2018-03-06 Jakub Jelinek <jakub@redhat.com>
9523 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
9526 PR tree-optimization/84687
9527 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
9529 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
9531 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9533 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
9534 Rename to ppc_speculation_barrier.
9535 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
9536 __builtin_ppc_speculation_barrier.
9538 2018-03-05 Jakub Jelinek <jakub@redhat.com>
9541 * combine.c (combine_simplify_rtx): Don't try to simplify if
9542 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
9545 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
9547 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
9548 to 32 bytes when compiling for POWER9.
9550 2018-03-05 Jakub Jelinek <jakub@redhat.com>
9553 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
9554 regparm >= 3 with no arg reg available also for calls with
9555 flag_force_indirect_call. Pass decl to ix86_function_regparm.
9558 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
9560 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
9562 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
9565 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
9567 2018-03-05 Richard Biener <rguenther@suse.de>
9569 PR tree-optimization/84486
9570 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
9571 When inserting a __builtin_assume_aligned call set the LHS
9572 SSA name alignment info accordingly.
9574 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
9576 PR tree-optimization/84114
9577 * config/aarch64/aarch64.c (aarch64_reassociation_width)
9578 Avoid reassociation of FLOAT_MODE addition.
9580 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
9582 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
9583 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
9584 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
9585 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
9586 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
9587 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
9588 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
9590 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
9591 __builtin_ia32_wbinvd): New builtins.
9592 (SPECIAL_ARGS2): New.
9593 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
9594 (SPECIAL_ARGS2): New.
9595 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
9596 (ix86_valid_target_attribute_inner_p): Ditto.
9597 (ix86_init_mmx_sse_builtins): Add special_args2.
9598 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
9599 TARGET_WBNOINVD_P): New.
9600 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
9601 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
9602 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
9603 * config/i386/immintrin.h (_wbinvd): New intrinsic.
9604 * config/i386/pconfigintrin.h: New file.
9605 * config/i386/wbnoinvdintrin.h: Ditto.
9606 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
9608 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
9610 2018-03-05 Richard Biener <rguenther@suse.de>
9612 PR tree-optimization/84670
9613 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
9615 (BB_VISITED_WITH_VISITED_SUCCS): New define.
9616 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
9617 (compute_antic_aux): Only assert the number of values in ANTIC_IN
9618 doesn't grow if all successors (recursively) were visited at least
9621 2018-03-05 Richard Biener <rguenther@suse.de>
9623 PR tree-optimization/84650
9624 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
9625 if executed in the loop pipeline.
9627 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
9629 * doc/configfiles.texi (Configuration Files): Move info about
9630 conditionalizing $target-protos.h to...
9631 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
9632 differs from $target-protos.h.
9634 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
9635 Chung-Ju Wu <jasonwucj@gmail.com>
9637 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
9638 * config/nds32/nds32-multiple.md (setmemsi): Define.
9639 * config/nds32/nds32-memory-manipulation.c
9640 (nds32_gen_dup_4_byte_to_word_value): New.
9641 (emit_setmem_word_loop): New.
9642 (emit_setmem_byte_loop): New.
9643 (nds32_expand_setmem_loop): New.
9644 (nds32_expand_setmem_loop_v3m): New.
9645 (nds32_expand_setmem_unroll): New.
9646 (nds32_expand_setmem): New.
9648 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
9649 Chung-Ju Wu <jasonwucj@gmail.com>
9651 * config/nds32/nds32-memory-manipulation.c
9652 (nds32_emit_load_store): New.
9653 (nds32_emit_post_inc_load_store): New.
9654 (nds32_emit_mem_move): New.
9655 (nds32_emit_mem_move_block): New.
9656 (nds32_expand_movmemsi_loop_unknown_size): New.
9657 (nds32_expand_movmemsi_loop_known_size): New.
9658 (nds32_expand_movmemsi_loop): New.
9659 (nds32_expand_movmemsi_unroll): New.
9660 (nds32_expand_movmemqi): Rename ...
9661 (nds32_expand_movmemsi): ... to this.
9662 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
9663 (movmemsi): ... to this.
9664 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
9665 (nds32_expand_movmemsi): ... to this.
9667 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
9668 Monk Chiang <sh.chiang04@gmail.com>
9669 Chung-Ju Wu <jasonwucj@gmail.com>
9671 * config/nds32/nds32-protos.h
9672 (nds32_expand_load_multiple): New arguments.
9673 (nds32_expand_store_multiple): Ditto.
9674 (nds32_valid_multiple_load_store): Rename ...
9675 (nds32_valid_multiple_load_store_p): ... to this.
9676 * config/nds32/nds32-memory-manipulation.c
9677 (nds32_expand_load_multiple): Refine implementation.
9678 (nds32_expand_store_multiple): Ditto.
9679 * config/nds32/nds32-multiple.md
9680 (load_multiple): Update nds32_expand_load_multiple interface.
9681 (store_multiple): Update nds32_expand_store_multiple interface.
9682 * config/nds32/nds32-predicates.c
9683 (nds32_valid_multiple_load_store): Rename ...
9684 (nds32_valid_multiple_load_store_p): ... to this and refine
9686 * config/nds32/predicates.md
9687 (nds32_load_multiple_and_update_address_operation): New predicate.
9688 (nds32_store_multiple_and_update_address_operation): New predicate.
9690 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
9691 Chung-Ju Wu <jasonwucj@gmail.com>
9693 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
9694 (combo): New attribute.
9695 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
9697 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
9699 * config/nds32/nds32.opt: Change -mcmodel= default value.
9701 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
9702 Monk Chiang <sh.chiang04@gmail.com>
9703 Chung-Ju Wu <jasonwucj@gmail.com>
9705 * config/nds32/constants.md (unspec_element): New enum.
9706 * config/nds32/constraints.md (Umw): New constraint.
9707 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
9708 * config/nds32/nds32-intrinsic.md: Likewise.
9709 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
9710 (nds32_valid_smw_lwm_base_p): New.
9711 (nds32_output_smw_single_word): New.
9712 (nds32_output_lmw_single_word): New.
9713 (nds32_expand_unaligned_load): New.
9714 (nds32_expand_unaligned_store): New.
9715 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
9716 (nds32_output_smw_single_word): Declare.
9717 (nds32_output_lmw_single_word): Declare.
9718 (nds32_expand_unaligned_load): Declare.
9719 (nds32_expand_unaligned_store): Declare.
9720 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
9721 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
9722 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
9723 NDS32_BUILTIN_UASTORE_DW.
9724 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
9727 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
9728 Kito Cheng <kito.cheng@gmail.com>
9729 Chung-Ju Wu <jasonwucj@gmail.com>
9731 * config/nds32/nds32-intrinsic.c
9732 (nds32_expand_builtin_null_ftype_reg): Delete.
9733 (nds32_expand_builtin_reg_ftype_imm): Ditto.
9734 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
9735 (nds32_read_argument): New.
9736 (nds32_legitimize_target): Ditto.
9737 (nds32_legitimize_argument): Ditto.
9738 (nds32_check_constant_argument): Ditto.
9739 (nds32_expand_unop_builtin): Ditto.
9740 (nds32_expand_unopimm_builtin): Ditto.
9741 (nds32_expand_binop_builtin): Ditto.
9742 (nds32_builtin_decl_impl): Ditto.
9743 (builtin_description): Ditto.
9744 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
9745 (nds32_init_builtins_impl): Ditto.
9746 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
9747 (nds32_builtin_decl): New.
9748 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
9749 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
9751 2018-03-02 Jeff Law <law@redhat.com>
9753 * reorg.c (stop_search_p): Handle DEBUG_INSN.
9754 (redundant_insn, fill_simple_delay_slots): Likewise.
9755 (fill_slots_from_thread): Likewise.
9756 * resource.c (mark_referenced_resources): Likewise.
9757 (mark_set_resources, find_dead_or_set_registers): Likewise.
9759 2018-03-02 Jakub Jelinek <jakub@redhat.com>
9761 * substring-locations.h (format_warning_va): Formatting fix for
9763 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
9765 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
9766 * substring-locations.c: Include intl.h.
9767 (format_warning_va): Turned into small wrapper around
9768 format_warning_n_va, renamed to ...
9769 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
9770 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
9772 (format_warning_at_substring_n): New function.
9773 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
9774 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
9775 format_warning_at_substring with just a shorter name instead of
9776 const function pointer.
9777 (fmtwarn_n): New function.
9778 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
9779 appropriate, get rid of all the fmtstr temporaries, move conditionals
9780 with G_() wrapped string literals directly into fmtwarn arguments,
9781 cast dir.len to (int), formatting fixes.
9783 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
9785 * doc/invoke.texi: Remove "Cilk Plus" references.
9787 2018-03-02 Jakub Jelinek <jakub@redhat.com>
9788 Richard Biener <rguenther@suse.de>
9791 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
9792 for error or warning attributes if CALL_FROM_THUNK_P is set.
9795 2018-03-02 Jakub Jelinek <jakub@redhat.com>
9798 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
9799 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
9802 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
9803 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
9805 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
9806 instead of -1U in last predictors element's probability member.
9808 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
9811 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
9812 arguments if they are comparable.
9814 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
9816 PR tree-optimization/84634
9817 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
9818 masks and masked_loop_p with a single loop_masks, making sure it's
9819 null for bb vectorization.
9821 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
9823 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
9824 (vect_analyze_data_ref_access): Use loop->safe_len rather than
9825 loop->force_vectorize to check whether there is no alias.
9827 2018-03-02 Jakub Jelinek <jakub@redhat.com>
9830 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
9832 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
9834 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
9835 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
9836 instead of a loop around prev_real_insn.
9837 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
9841 * config/i386/i386.c (ix86_print_operand): Use conditional
9842 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
9845 2018-03-02 Richard Biener <rguenther@suse.de>
9847 PR tree-optimization/84427
9848 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
9849 (bitmap_set_subtract_values): Rewrite to handle multiple
9852 (prune_clobbered_mems): Likewise.
9853 (phi_translate): Take edge instead of pred/phiblock.
9854 (phi_translate_1): Likewise.
9855 (phi_translate_set): Likewise. Insert all translated
9856 exprs for a value into the set, keeping possibly multiple
9857 expressions per value.
9858 (compute_antic_aux): Adjust for phi_translate changes.
9859 When intersecting union the expressions and prune those
9860 not in the final value set, keeping possibly multiple
9861 expressions per value. Do not use value-insertion
9862 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
9863 all expressions. Add verification that the value-sets
9864 only shrink during iteration.
9865 (compute_partial_antic_aux): Adjust for the phi_translate changes.
9866 (do_pre_regular_insertion): Likewise.
9867 (do_pre_partial_partial_insertion): Likewise.
9869 2018-03-02 Richard Biener <rguenther@suse.de>
9872 * config/darwin.c (saved_debug_info_level): New static global.
9873 (darwin_asm_lto_start): Disable debug info generation for LTO out.
9874 (darwin_asm_lto_end): Restore debug info generation settings.
9876 2018-03-01 Martin Liska <mliska@suse.cz>
9879 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
9882 2018-03-01 Richard Biener <rguenther@suse.de>
9885 * dwarf2out.c (gen_variable_die): Properly handle late VLA
9886 type annotation with LTO when debug was disabled at compile-time.
9888 2018-03-01 Matthew Fortune <mfortune@gmail.com>
9890 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
9892 (mips_final_postscan_insn): Likewise.
9894 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
9896 PR rtl-optimization/84528
9897 * alias.c (init_alias_target): Add commentary.
9898 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
9899 a unique base value if the frame pointer is not eliminated
9900 to the stack pointer.
9902 2018-03-01 Tom de Vries <tom@codesourcery.com>
9904 PR rtl-optimization/83327
9905 * lra-int.h (hard_regs_spilled_into): Declare.
9906 * lra.c (hard_regs_spilled_into): Define.
9907 (init_reg_info): Init hard_regs_spilled_into.
9908 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
9909 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
9910 (process_bb_lives): Handle hard_regs_spilled_into.
9911 (lra_create_live_ranges_1): Before doing liveness propagation, clear
9912 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
9914 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
9916 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
9917 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
9918 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
9919 * config/rs6000/aix72.h: New file.
9921 2018-02-28 Jakub Jelinek <jakub@redhat.com>
9923 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
9924 instead of warning_at with conditional singular and plural messages
9928 * stor-layout.c (update_alignment_for_field): For
9929 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
9930 && !DECL_PACKED (field), do the alignment update, just use
9931 only desired_align instead of MAX (type_align, desired_align)
9933 (place_field): Don't do known_align < desired_align handling
9934 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
9935 is non-NULL, instead do it after rli->prev_field handling and
9936 only if not within a bitfield word. For DECL_PACKED (field)
9937 use type_align of BITS_PER_UNIT.
9939 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
9941 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
9942 superfluous parentheses and trailing spaces.
9944 2018-02-28 Richard Biener <rguenther@suse.de>
9946 PR tree-optimization/84584
9947 * graphite-scop-detection.c (scop_detection::add_scop): Discard
9948 SCoPs with fake exit edge.
9950 2018-02-28 Martin Liska <mliska@suse.cz>
9953 * timevar.c (timer::print): Fix format to properly print 100%
9956 2018-02-28 Richard Biener <rguenther@suse.de>
9959 * genmatch.c (capture_info::walk_match): Do not mark
9960 captured expressions without operands as expr_p given
9961 they act more like predicates and should be subject to
9962 "lost tail" side-effect preserving.
9964 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
9966 PR rtl-optimization/81611
9967 * auto-inc-dec.c (attempt_change): Move dead note from
9968 mem_insn if it's the next use of regno
9969 (find_address): Take address use of reg holding
9970 non-incremented value. Add parm to limit search to the named
9972 (merge_in_block): Attempt to use a mem insn that is the next
9973 use of the original regno.
9975 2018-02-27 Martin Sebor <msebor@redhat.com>
9978 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
9979 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
9981 2018-02-27 Martin Sebor <msebor@redhat.com>
9983 PR translation/84207
9984 * diagnostic-core.h (warning_n, error_n, inform_n): Change
9985 n argument to unsigned HOST_WIDE_INT.
9986 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
9987 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
9988 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
9989 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
9991 2018-02-27 Richard Biener <rguenther@suse.de>
9993 PR tree-optimization/84512
9994 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
9995 Do not use the estimate returned from record_stmt_cost for
9996 the scalar iteration cost but sum properly using add_stmt_cost.
9998 2018-02-27 Richard Biener <rguenther@suse.de>
10000 PR tree-optimization/84466
10001 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
10002 Adjust last change to less strictly validate use operands.
10004 2018-02-27 Martin Liska <mliska@suse.cz>
10006 PR gcov-profile/84548
10007 * gcov.c (process_file): Allow partial overlap and consider it
10008 also as group functions.
10009 (output_lines): Properly calculate range of lines for a group.
10011 2018-02-27 Martin Liska <mliska@suse.cz>
10013 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
10014 'ggc' suffixes. Change first column width.
10015 (timer::print): Fix formatting of the column.
10017 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
10019 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
10020 preserve inline entry blocks for the sake of debug inline
10021 entry point markers alone.
10022 (remove_unused_locals): Suggest in comments a better place to
10023 force the preservation of inline entry blocks that are
10024 otherwise unused, but do not preserve them.
10026 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
10028 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
10030 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
10033 * config/i386/constraints.md (Bs): Replace
10034 ix86_indirect_branch_register with
10035 TARGET_INDIRECT_BRANCH_REGISTER.
10037 * config/i386/i386.md (indirect_jump): Likewise.
10038 (tablejump): Likewise.
10039 (*sibcall_memory): Likewise.
10040 (*sibcall_value_memory): Likewise.
10041 Peepholes of indirect call and jump via memory: Likewise.
10042 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
10043 (*sibcall_value_GOT_32): Likewise.
10044 * config/i386/predicates.md (indirect_branch_operand): Likewise.
10045 (GOT_memory_operand): Likewise.
10046 (call_insn_operand): Likewise.
10047 (sibcall_insn_operand): Likewise.
10048 (GOT32_symbol_operand): Likewise.
10049 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
10051 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
10053 PR rtl-optimization/83496
10054 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
10055 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
10056 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
10057 redundant insn, if any.
10058 (relax_delay_slots): Likewise.
10059 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
10061 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
10063 PR tree-optimization/83965
10064 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
10065 that grouped statements are part of a reduction chain. Return
10066 true if the statement is not marked as a reduction itself but
10067 is part of a group.
10068 (vect_recog_dot_prod_pattern): Don't check whether the statement
10069 is part of a group here.
10070 (vect_recog_sad_pattern): Likewise.
10071 (vect_recog_widen_sum_pattern): Likewise.
10073 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
10076 * final.c (rest_of_clean_state): Also look for calls inside sequences.
10078 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
10081 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
10083 (ix86_output_indirect_function_return): New prototype.
10084 (ix86_split_simple_return_pop_internal): Likewise.
10085 * config/i386/i386.c (indirect_return_via_cx): New.
10086 (indirect_return_via_cx_bnd): Likewise.
10087 (indirect_thunk_name): Handle return va CX_REG.
10088 (output_indirect_thunk_function): Create alias for
10089 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
10090 (ix86_output_indirect_jmp): Remove the bool argument.
10091 (ix86_output_indirect_function_return): New function.
10092 (ix86_split_simple_return_pop_internal): Likewise.
10093 * config/i386/i386.md (*indirect_jump): Don't pass false
10094 to ix86_output_indirect_jmp.
10095 (*tablejump_1): Likewise.
10096 (simple_return_pop_internal): Change it to define_insn_and_split.
10097 Call ix86_split_simple_return_pop_internal to split it for
10098 -mfunction-return=.
10099 (simple_return_indirect_internal): Call
10100 ix86_output_indirect_function_return instead of
10101 ix86_output_indirect_jmp.
10103 2018-02-26 Jakub Jelinek <jakub@redhat.com>
10106 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
10107 memset and value initialization afterwards.
10109 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
10111 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
10113 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
10116 * common/config/aarch64/aarch64-common.c
10117 (aarch_option_optimization_table[]): Switch
10118 off fomit-frame-pointer
10120 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
10121 Chung-Ju Wu <jasonwucj@gmail.com>
10123 * config/nds32/nds32-multiple.md (load_multiple): Disallow
10125 (store_multiple): Ditto.
10127 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
10129 * config.gcc: Add --with-cpu support for nds32 target.
10130 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
10131 * config/nds32/nds32.opt: Add -mcpu= option.
10133 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
10135 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
10136 isel=yes): Warn for these deprecated options.
10138 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
10140 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
10141 ISA_2_5_MASKS_EMBEDDED.
10143 2018-02-23 Jakub Jelinek <jakub@redhat.com>
10145 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
10146 p->max as pointers rather than using iterative_hash_expr.
10148 2018-02-23 Carl Love <cel@us.ibm.com>
10150 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
10151 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
10153 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
10154 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
10155 P8V_BUILTIN_VEC_VUNSIGNED2.
10157 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
10160 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
10161 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
10163 * lra-constraints.c (curr_insn_transform): Set up
10164 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
10166 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
10167 * lra-eliminations.c (spill_pseudos): Ditto.
10168 (process_insn_for_elimination): Ditto.
10169 * lra-lives.c (reg_early_clobber_p): Use the new macros.
10170 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
10171 LRA_NON_CLOBBERED_ALT.
10173 2018-02-22 Martin Sebor <msebor@redhat.com>
10175 PR tree-optimization/84480
10176 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
10177 to maybe_diag_stxncpy_trunc. Call it.
10178 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
10179 from gimple_fold_builtin_strcpy. Print inlining stack.
10180 (handle_builtin_stxncpy): Print inlining stack.
10181 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
10183 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
10186 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
10187 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
10188 and -fcheck-pointer-bounds are used together.
10189 (indirect_thunk_prefix): New enum.
10190 (indirect_thunk_need_prefix): New function.
10191 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
10192 "_nt" instead of "_bnd" for NOTRACK prefix.
10193 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
10194 (output_indirect_thunk_function): Likewise.
10196 (ix86_code_end): Update output_indirect_thunk_function calls.
10197 (ix86_output_indirect_branch_via_reg): Replace
10198 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
10199 (ix86_output_indirect_branch_via_push): Likewise.
10200 (ix86_output_function_return): Likewise.
10201 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
10202 incompatible with -fcf-protection=branch and
10203 -fcheck-pointer-bounds.
10205 2018-02-22 Steve Ellcey <sellcey@cavium.com>
10208 * config/aarch64/aarch64.c (aarch64_print_address_internal):
10209 Change gcc_assert call to output_operand_lossage.
10211 2018-02-22 Steve Ellcey <sellcey@cavium.com>
10213 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
10215 2018-02-22 DJ Delorie <dj@redhat.com>
10216 Sebastian Perta <sebastian.perta@renesas.com>
10217 Oleg Endo <olegendo@gcc.gnu.org>
10219 * config/rx/rx.c (rx_rtx_costs): New function.
10220 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
10222 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
10224 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
10226 2018-02-22 Martin Liska <mliska@suse.cz>
10229 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
10230 Add "native" as a possible value.
10232 2018-02-22 Martin Liska <mliska@suse.cz>
10235 * config/i386/i386.c (ix86_option_override_internal):
10236 Add "native" as a possible value for -march and -mtune.
10238 2018-02-22 Jakub Jelinek <jakub@redhat.com>
10241 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
10242 to all type variants.
10244 PR tree-optimization/84503
10245 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
10246 width as info->bitpos + info->bitsize - start.
10247 (merged_store_group::merge_overlapping): Simplify width computation.
10248 (check_no_overlap): New function.
10249 (imm_store_chain_info::try_coalesce_bswap): Compute expected
10250 start + width and last_order of the group, fail if check_no_overlap
10252 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
10253 to group if check_no_overlap fails.
10255 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
10257 * config/rs6000/altivec.md: Delete contraint arguments to
10258 define_expand, define_split, and define_peephole2, and in
10259 define_insn_and_split if always unused.
10260 * config/rs6000/darwin.md: Ditto.
10261 * config/rs6000/dfp.md: Ditto.
10262 * config/rs6000/rs6000.md: Ditto.
10263 * config/rs6000/sync.md: Ditto.
10264 * config/rs6000/vector.md: Ditto.
10265 * config/rs6000/vsx.md: Ditto.
10267 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
10269 * config/rs6000/altivec.md: Write output control strings as braced
10270 blocks instead of double-quoted strings.
10271 * config/rs6000/darwin.md: Ditto.
10272 * config/rs6000/rs6000.md: Ditto.
10273 * config/rs6000/vector.md: Ditto.
10274 * config/rs6000/vsx.md: Ditto.
10276 2018-02-21 Jason Merrill <jason@redhat.com>
10278 PR c++/84314 - ICE with templates and fastcall attribute.
10279 * attribs.c (build_type_attribute_qual_variant): Remove assert.
10281 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
10283 * ipa-cp.c (determine_versionability): Fix comment typos.
10285 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
10288 * ipa-cp.c (determine_versionability): Do not version functions caling
10291 2018-02-21 Martin Liska <mliska@suse.cz>
10294 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
10295 Add "native" as a possible value.
10296 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
10297 the macro when native cpu detection is available.
10299 2018-02-21 Martin Liska <mliska@suse.cz>
10302 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
10303 Add "native" as a possible value.
10304 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
10305 when native cpu detection is available.
10307 2018-02-21 Jakub Jelinek <jakub@redhat.com>
10308 Martin Sebor <msebor@redhat.com>
10310 PR tree-optimization/84478
10311 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
10313 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
10314 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
10315 support which is conservatively correct, for 2 only stay conservative
10316 for maxlen. Formatting and comment capitalization fixes. Add STRICT
10317 argument to the 2 argument get_range_strlen, adjust 6 arg
10318 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
10320 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
10321 (gimple_fold_builtin_strlen): Pass true as last argument to
10324 2018-02-20 Martin Sebor <msebor@redhat.com>
10326 PR middle-end/84095
10327 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
10328 (builtin_memref::set_base_and_offset): Same. Handle inner references.
10329 (builtin_memref::builtin_memref): Factor out parts into
10330 set_base_and_offset and call it.
10332 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
10334 PR middle-end/84406
10335 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
10336 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
10337 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
10338 search at the associated MODE_INT.
10340 2018-02-20 Jeff Law <law@redhat.com>
10342 PR middle-end/82123
10343 PR tree-optimization/81592
10344 PR middle-end/79257
10345 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
10346 for range data rather than using global data.
10347 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
10348 range data rather than using global data.
10349 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
10350 pass it to children as needed.
10351 (struct directive::fmtresult): Similarly.
10352 (struct directive::set_width): Similarly.
10353 (struct directive::set_precision): Similarly.
10354 (format_integer, format_directive, parse_directive): Similarly.
10355 (format_none): Accept unnamed vr_values parameter.
10356 (format_percent, format_floating, format_character): Similarly.
10357 (format_string, format_plain): Similarly.
10358 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
10359 the EVRP range analyzer for range data rather than using global data.
10360 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
10361 gimple-ssa-evrp-analyze.h
10362 (class sprintf_dom_walker): Add after_dom_children member function.
10363 Add evrp_range_analyzer member.
10364 (sprintf_dom_walker::before_dom_children): Call into the EVRP
10365 range analyzer as needed.
10366 (sprintf_dom_walker::after_dom_children): New member function.
10367 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
10369 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
10370 (evrp_range_analyzer::pop_to_marker): Likewise.
10372 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
10374 PR tree-optimization/84419
10375 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
10376 with the required type if its current type is compatible but
10379 2018-02-20 Jakub Jelinek <jakub@redhat.com>
10381 PR middle-end/82004
10382 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
10383 after vectorization.
10385 2018-02-20 Martin Liska <mliska@suse.cz>
10388 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
10389 possible values if we don't have a hint.
10391 2018-02-20 Martin Liska <mliska@suse.cz>
10395 * final.c (shorten_branches): Build align_tab array with one
10397 * opts.c (finish_options): Add alignment option limit check.
10398 (MAX_CODE_ALIGN): Likewise.
10399 (MAX_CODE_ALIGN_VALUE): Likewise.
10400 * doc/invoke.texi: Document maximum allowed option value for
10401 all -falign-* options.
10403 2018-02-19 Jakub Jelinek <jakub@redhat.com>
10406 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
10407 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
10408 * var-tracking.c (emit_note_insn_var_location): Remove all references
10409 to NOTE_INSN_CALL_ARG_LOCATION.
10410 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
10411 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
10412 Use copy_rtx_if_shared.
10413 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
10414 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
10415 (dwarf2out_var_location): Remove handling of
10416 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
10418 * final.c (final_scan_insn): Remove all references to
10419 NOTE_INSN_CALL_ARG_LOCATION.
10420 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
10421 before dumping final insns.
10422 * except.c (emit_note_eh_region_end): Remove all references to
10423 NOTE_INSN_CALL_ARG_LOCATION.
10424 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
10425 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
10426 * config/arc/arc.c (hwloop_optimize): Likewise.
10427 * config/arm/arm.c (create_fix_barrier): Likewise.
10428 * config/s390/s390.c (s390_chunkify_start): Likewise.
10429 * config/sh/sh.c (find_barrier): Likewise.
10430 * config/i386/i386.c (rest_of_insert_endbranch,
10431 ix86_seh_fixup_eh_fallthru): Likewise.
10432 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
10433 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
10434 * config/frv/frv.c (frv_function_prologue): Likewise.
10435 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
10437 (note_outside_basic_block_p): Remove all references to
10438 NOTE_INSN_CALL_ARG_LOCATION.
10439 * gengtype.c (adjust_field_rtx_def): Likewise.
10440 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
10442 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
10443 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
10446 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
10449 PR tree-optimization/84452
10450 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
10451 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
10454 2018-02-19 Martin Liska <mliska@suse.cz>
10457 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
10459 2018-02-19 Martin Liska <mliska@suse.cz>
10460 Richard Sandiford <richard.sandiford@linaro.org>
10462 PR tree-optimization/82491
10463 * gimple-fold.c (get_base_constructor): Make earlier bail out
10466 2018-02-19 Carl Love <cel@us.ibm.com>
10468 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
10469 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
10471 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
10472 P8V_BUILTIN_VEC_NEG.
10474 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
10476 * config/rl78/rl78.md (movdf): New define expand.
10478 2018-02-19 Martin Liska <mliska@suse.cz>
10481 * doc/invoke.texi: Fix typo.
10482 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
10484 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
10486 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
10487 handle rs6000_single_float and rs6000_double_float specially for
10490 2018-02-16 Jeff Law <law@redhat.com>
10492 * config/rx/rx.c (add_pop_cfi_notes): New function.;
10493 (pop_regs): Use it.
10495 2018-02-16 Jakub Jelinek <jakub@redhat.com>
10498 * ipa-inline.c (inline_small_functions): Fix a typo.
10500 2018-02-16 Nathan Sidwell <nathan@acm.org>
10502 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
10504 2018-02-16 Carl Love <cel@us.ibm.com>
10506 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
10507 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
10508 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
10509 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
10510 expansion to P8V_BUILTIN_VEC_FLOAT2.
10512 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
10514 PR rtl-optimization/70023
10515 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
10516 src_regno into account.
10518 2018-02-16 Carl Love <cel@us.ibm.com>
10520 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
10521 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
10522 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
10523 * config/rs6000/rs6000.c: Remove case statements for
10524 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
10525 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
10526 and P9V_BUILTIN_VEC_VINSERT4B.
10527 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
10528 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
10529 * config/rs6000/vsx.md:
10530 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
10533 2018-02-16 Carl Love <cel@us.ibm.com>
10535 * config/rs6000/altivec.h: Add builtin names vec_extract4b
10537 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
10539 * config/rs6000/rs6000-c.c: Add the definitions for
10540 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
10541 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
10542 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
10543 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
10544 definition for insert4b and define insn *insert3b_internal.
10545 * doc/extend.texi: Add documentation for vec_extract4b.
10547 2018-02-16 Nathan Sidwell <nathan@acm.org>
10549 * doc/extend.texi (Backwards Compatibility): Mention friend
10550 injection. Note for-scope is deprecated.
10551 * doc/invoke.texi (-ffriend-injection): Deprecate.
10553 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
10555 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
10556 that moved to I2, also allow destinations that are a paradoxical
10557 subreg (instead of a normal reg).
10559 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
10562 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
10565 2018-02-16 Richard Biener <rguenther@suse.de>
10567 PR tree-optimization/84037
10568 PR tree-optimization/84016
10570 * config/i386/i386.c (ix86_builtin_vectorization_cost):
10571 Adjust vec_construct for the fact we need additional higher latency
10572 128bit inserts for AVX256 and AVX512 vector builds.
10573 (ix86_add_stmt_cost): Scale vector construction cost for
10576 2018-02-16 Richard Biener <rguenther@suse.de>
10578 PR tree-optimization/84417
10579 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
10580 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
10581 (non_rewritable_lvalue_p): Likewise, use poly-ints.
10583 2018-02-16 Martin Liska <mliska@suse.cz>
10586 * internal-fn.def (ASAN_CHECK): Set proper flags.
10587 (ASAN_MARK): Likewise.
10589 2018-02-16 Julia Koval <julia.koval@intel.com>
10591 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
10592 from PTA_CANNONLAKE.
10594 2018-02-16 Jakub Jelinek <jakub@redhat.com>
10597 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
10598 Use ++iter rather than iter++ for std::list iterators.
10599 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
10600 defer deleting them until all nodes in the forest are processed. Do
10601 free even leaf nodes. Change to_process into auto_vec.
10604 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
10605 * vec.h (vec_default_construct): Use memset instead of placement new
10606 if BROKEN_VALUE_INITIALIZATION is defined.
10607 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
10608 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
10611 PR rtl-optimization/83723
10612 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
10613 * lra.c (lra_substitute_pseudo): Likewise. If true, use
10614 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
10616 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
10618 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
10620 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
10622 PR rtl-optimization/81443
10623 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
10624 from inner REGs to paradoxical SUBREGs.
10626 2018-02-16 Richard Biener <rguenther@suse.de>
10628 PR tree-optimization/84399
10629 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
10630 For operands we can analyze at their definition make sure we can
10631 analyze them at each use as well.
10633 2018-02-16 Richard Biener <rguenther@suse.de>
10635 PR tree-optimization/84190
10636 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
10637 volatile accesses if the decl isn't volatile.
10639 2018-02-15 Jason Merrill <jason@redhat.com>
10641 PR c++/84314 - ICE with templates and fastcall attribute.
10642 * attribs.c (build_type_attribute_qual_variant): Don't clobber
10643 TYPE_CANONICAL on an existing type.
10645 2018-02-15 Jakub Jelinek <jakub@redhat.com>
10647 PR tree-optimization/84383
10648 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
10649 dstoff nor call operand_equal_p if dstbase is NULL.
10651 PR tree-optimization/84334
10652 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
10653 also a CONSTANT_CLASS_P, punt.
10655 2018-02-14 Jim Wilson <jimw@sifive.com>
10657 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
10658 first SMALL_OPERAND check. New local min_second_step. Move assert
10659 to where locals are set. Add TARGET_RVC support.
10660 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
10662 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
10664 * doc/invoke.texi: Correct -Wformat-overflow code sample.
10666 2018-02-14 Martin Sebor <msebor@redhat.com>
10668 PR tree-optimization/83698
10669 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
10670 arrays constrain the offset range to their bounds.
10671 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
10672 (builtin_access::overlap): Avoid setting the size of overlap if it's
10674 (maybe_diag_overlap): Also consider arrays when deciding what values
10675 of offsets to include in diagnostics.
10677 2018-02-14 Martin Sebor <msebor@redhat.com>
10680 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
10681 that correspond to the kind of a declaration.
10683 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
10686 * config/pa/pa.md: Load address of PIC label using the linkage table
10687 if the label is nonlocal.
10689 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
10691 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
10692 warning message if user requests -maltivec=be.
10693 * doc/invoke.texi: Document deprecation of -maltivec=be.
10695 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
10698 * config/rs6000/rs6000-c.c: Update definitions for
10699 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
10700 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
10702 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
10705 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
10706 add _get_ssp intrinsics. Remove argument from
10707 __builtin_ia32_rdssp[d|q].
10708 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
10709 * config/i386/i386-builtin.def: Remove argument from
10710 __builtin_ia32_rdssp[d|q].
10711 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
10712 ix86_expand_special_args_builtin for _rdssp[d|q].
10713 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
10714 Clear register before usage.
10715 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
10716 Add documentation for new _get_ssp and _inc_ssp intrinsics.
10718 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
10720 PR tree-optimization/84357
10721 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
10722 operand 1 of an ARRAY_REF too.
10724 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
10727 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
10728 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
10730 (set_of_reg): New struct.
10731 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
10732 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
10733 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
10735 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
10736 Split into bitclr, bitset, bitinvert patterns if appropriate.
10737 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
10738 use rx_fuse_in_memory_bitop.
10739 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
10740 to named insn, correct maximum insn length.
10742 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
10745 * machmode.def: Define a complex mode for PARTIAL_INT.
10746 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
10748 * doc/rtl.texi: Document CSPImode.
10749 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
10751 (msp430_hard_regno_nregs_with_padding): Likewise.
10753 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
10756 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
10758 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
10760 PR rtl-optimization/84169
10761 * combine.c (try_combine): New variable split_i2i3. Set it to true if
10762 we generated a parallel as new i3 and we split that to new i2 and i3
10763 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
10764 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
10765 those to i2, not i1. Partially rewrite this scan code.
10767 2018-02-13 Jakub Jelinek <jakub@redhat.com>
10770 * stor-layout.c (place_field): For variable length fields, adjust
10771 offset_align afterwards not just based on the field's alignment,
10772 but also on the size.
10774 PR middle-end/84309
10775 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
10776 of exps and logs in the use_exp2 case.
10778 2018-02-13 Jeff Law <law@redhat.com>
10780 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
10781 entry for "vector".
10783 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
10786 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
10790 * common.opt (gas-loc-support, gas-locview-support): New.
10791 (ginline-points, ginternal-reset-location-views): New.
10792 * doc/invoke.texi: Document them. Use @itemx where intended.
10793 (gvariable-location-views): Adjust.
10794 * target.def (reset_location_view): New.
10795 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
10796 (TARGET_RESET_LOCATION_VIEW): New.
10797 * doc/tm.texi: Rebuilt.
10798 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
10799 (dwarf2out_default_as_locview_support): New.
10800 (output_asm_line_debug_info): Use option variables.
10801 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
10802 (output_loc_list): Likewise.
10803 (add_high_low_attributes): Check option variables.
10804 Don't output entry view attribute in strict mode.
10805 (gen_inlined_subroutine_die): Check option variables.
10806 (dwarf2out_inline_entry): Likewise.
10807 (init_sections_and_labels): Likewise.
10808 (dwarf2out_early_finish): Likewise.
10809 (maybe_reset_location_view): New, from...
10810 (dwarf2out_var_location): ... here. Call it.
10811 * debug.h (dwarf2out_default_as_loc_support): Declare.
10812 (dwarf2out_default_as_locview_support): Declare.
10813 * hooks.c (hook_int_rtx_insn_0): New.
10814 * hooks.h (hook_int_rtx_insn_0): Declare.
10815 * toplev.c (process_options): Take -gas-loc-support and
10816 -gas-locview-support from dwarf2out. Enable
10817 -gvariable-location-views by default only with locview
10818 assembler support. Enable -ginternal-reset-location-views by
10819 default only if the target defines the corresponding hook.
10820 Enable -ginline-points by default if location views are
10821 enabled; force it disabled if statement frontiers are
10823 * tree-inline.c (expand_call_inline): Check option variables.
10824 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
10826 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
10828 PR tree-optimization/84321
10829 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
10830 handling. Also check whether the anti-range contains any values
10831 that satisfy the mask; switch to a VR_RANGE if not.
10833 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
10836 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
10838 2018-02-13 Martin Jambor <mjambor@suse.cz>
10841 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
10842 of call statements, also set location of a load to a temporary.
10844 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
10846 * config/rl78/rl78.c (add_vector_labels): New function.
10847 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
10848 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
10849 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
10850 which checks that no arguments are passed.
10851 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
10852 * doc/extend.texi: Documentation for the new attribute.
10854 2018-02-13 Andreas Schwab <schwab@suse.de>
10856 * config/riscv/linux.h (CPP_SPEC): Define.
10858 2018-02-13 Jakub Jelinek <jakub@redhat.com>
10861 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
10862 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
10863 OPTION_MASK_ISA_AES as first argument to def_builtin_const
10864 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
10865 instead of OPTION_MASK_ISA_PCLMUL as first argument to
10866 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
10867 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
10868 temporarily for AES and PCLMUL builtins.
10870 PR tree-optimization/84339
10871 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
10872 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
10875 PR middle-end/84309
10876 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
10877 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
10878 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
10880 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
10882 * omp-simd-clone.h: New file.
10883 * omp-simd-clone.c: Include omp-simd-clone.h.
10884 (expand_simd_clones): No longer static.
10885 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
10886 cgraph.h and omp-simd-clone.h.
10887 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
10888 (vect_recog_widen_shift_pattern): Formatting fix.
10889 (vect_pattern_recog_1): Don't check optab for calls.
10892 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
10893 operands[2] into a REG before using gen_lowpart on it.
10895 2018-02-12 Jeff Law <law@redhat.com>
10898 * config/sh/sh.c (find_barrier): Consider a sibling call
10901 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
10902 successfully back substituting a reg.
10904 2018-02-12 Richard Biener <rguenther@suse.de>
10906 PR tree-optimization/84037
10907 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
10908 parameter, move visited init to caller.
10909 (vect_slp_analyze_operations): Separate cost from validity
10910 check, initialize visited once for all instances.
10911 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
10913 * tree-vect-stmts.c (vect_model_simple_cost): Make early
10915 (vect_model_promotion_demotion_cost): Likewise.
10916 (vectorizable_bswap): Guard cost modeling with !slp_node
10917 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
10919 (vectorizable_call): Likewise.
10920 (vectorizable_conversion): Likewise.
10921 (vectorizable_assignment): Likewise.
10922 (vectorizable_shift): Likewise.
10923 (vectorizable_operation): Likewise.
10924 (vectorizable_store): Likewise.
10925 (vectorizable_load): Likewise.
10926 (vectorizable_condition): Likewise.
10927 (vectorizable_comparison): Likewise.
10929 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
10932 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
10933 (ASAN_MARK): Fix fnspec to account for return value, change pointer
10934 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
10936 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
10938 PR middle-end/83665
10939 * params.def (inline-min-speedup): Increase from 8 to 15.
10940 (max-inline-insns-auto): Decrease from 40 to 30.
10941 * ipa-split.c (consider_split): Add some buffer for function to
10942 be considered inlining candidate.
10943 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
10946 2018-02-12 Richard Biener <rguenther@suse.de>
10948 PR tree-optimization/84037
10949 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
10950 matched stmts if we cannot swap the non-matched ones.
10952 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
10954 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
10955 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
10956 _mm_maskz_scalef_round_ss): New intrinsics.
10957 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
10958 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
10959 __builtin_ia32_scalefss_round): Remove.
10960 (__builtin_ia32_scalefsd_mask_round,
10961 __builtin_ia32_scalefss_mask_round): New intrinsics.
10962 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
10963 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
10964 ((match_operand:VF_128 2 "<round_nimm_predicate>"
10965 "<round_constraint>")): Changed to ...
10966 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
10967 "<round_scalar_constraint>")): ... this.
10968 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
10969 %0, %1, %2<round_op3>}"): Changed to ...
10970 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
10971 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
10972 %2<round_scalar_mask_op3>}"): ... this.
10973 * config/i386/subst.md (round_scalar_nimm_predicate): New.
10975 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
10977 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
10978 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
10979 (_mm_maskz_sqrt_round_ss): New intrinsics.
10980 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
10981 (__builtin_ia32_sqrtsd_mask_round)
10982 (__builtin_ia32_sqrtss_mask_round): New builtins.
10983 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
10984 (__builtin_ia32_sqrtss_round): Remove.
10985 (__builtin_ia32_sqrtsd_mask_round)
10986 (__builtin_ia32_sqrtss_mask_round): New builtins.
10987 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
10988 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
10989 ((match_operand:VF_128 1 "vector_operand"
10990 "xBm,<round_constraint>")): Changed to ...
10991 ((match_operand:VF_128 1 "vector_operand"
10992 "xBm,<round_scalar_constraint>")): ... this.
10993 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
10994 %0, %2, %<iptr>1<round_op3>}): Changed to ...
10995 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
10996 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
10997 %<iptr>1<round_scalar_mask_op3>}): ... this.
10998 ((set_attr "prefix" "<round_prefix>")): Changed to ...
10999 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
11001 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
11004 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
11005 Cast vec_cmpeq result to correct type.
11006 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
11007 Cast vec_cmpgt result to correct type.
11009 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
11011 * final.c (final_scan_insn_1): Renamed from...
11012 (final_scan_insn): ... this. New wrapper, to recover
11013 seen from the outermost call in recursive ones.
11014 * config/sparc/sparc.c (output_return): Drop seen from call.
11015 (output_sibcall): Likewise.
11016 * config/visium/visium.c (output_branch): Likewise.
11018 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
11020 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
11023 2018-02-10 Alan Modra <amodra@gmail.com>
11026 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
11027 Specify LR as an input.
11029 2018-02-10 Jakub Jelinek <jakub@redhat.com>
11032 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
11033 remove_member_access_dummy_vars): New functions.
11034 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
11035 lower_omp_1, execute_lower_omp): Use them.
11037 PR rtl-optimization/84308
11038 * shrink-wrap.c (spread_components): Release todo vector.
11040 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
11042 PR rtl-optimization/57193
11043 * ira-color.c (struct allocno_color_data): Add member
11044 conflict_allocno_hard_prefs.
11045 (update_conflict_allocno_hard_prefs): New.
11046 (bucket_allocno_compare_func): Add a preference based on
11047 conflict_allocno_hard_prefs.
11048 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
11049 (color_allocnos): Remove a dead code. Initiate
11050 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
11052 2018-02-09 Jakub Jelinek <jakub@redhat.com>
11055 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
11056 constraint from =wa to wa. Avoid a subreg on the output operand,
11057 instead use a pseudo and subreg it in a move.
11058 (p9_xxbrd_<mode>): Changed to ...
11059 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
11060 (p9_xxbrd_v2df): New expander.
11061 (p9_xxbrw_<mode>): Changed to ...
11062 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
11063 (p9_xxbrw_v4sf): New expander.
11065 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
11067 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
11069 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
11072 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
11073 multiply in 32-bit mode.
11074 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
11075 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
11078 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
11080 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
11081 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
11082 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
11083 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
11085 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
11088 * dwarf2out.c (is_trivial_indirect_ref): New function.
11089 (dwarf2out_late_global_decl): Do not generate a location
11090 attribute for variables that have a non-trivial DECL_VALUE_EXPR
11091 and that are not defined in the current unit.
11093 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
11095 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
11096 instead of a libcall for UNORDERED.
11098 2018-02-09 Tamar Christina <tamar.christina@arm.com>
11101 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
11102 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
11104 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
11107 * config/s390/s390.c (s390_set_current_function): Invoke
11108 s390_indirect_branch_settings also if fndecl didn't change.
11110 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
11112 * config/rs6000/rs6000.md (blockage): Set length to zero.
11114 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
11116 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
11118 2018-02-09 Jakub Jelinek <jakub@redhat.com>
11121 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
11122 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
11126 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
11127 PARALLEL incoming that failed vt_get_decl_and_offset check.
11129 PR middle-end/84237
11130 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
11131 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
11133 (get_variable_section): For decls in named .bss* sections pass true as
11134 second argument to bss_initializer_p.
11136 2018-02-09 Marek Polacek <polacek@redhat.com>
11137 Jakub Jelinek <jakub@redhat.com>
11140 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
11141 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
11142 Sync some changes from cxx_fold_indirect_ref.
11144 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
11146 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
11148 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
11149 (BLOCK_INLINE_ENTRY_LABEL): New.
11150 (dwarf2out_var_location): Disregard inline entry markers.
11151 (inline_entry_data): New struct.
11152 (inline_entry_data_hasher): New hashtable type.
11153 (inline_entry_data_hasher::hash): New.
11154 (inline_entry_data_hasher::equal): New.
11155 (inline_entry_data_table): New variable.
11156 (add_high_low_attributes): Add DW_AT_entry_pc and
11157 DW_AT_GNU_entry_view attributes if a pending entry is found
11158 in inline_entry_data_table. Add old entry_pc attribute only
11159 if debug nonbinding markers are disabled.
11160 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
11161 markers are enabled.
11162 (block_within_block_p, dwarf2out_inline_entry): New.
11163 (dwarf2out_finish): Check that no entries remained in
11164 inline_entry_data_table.
11165 * final.c (reemit_insn_block_notes): Handle inline entry notes.
11166 (final_scan_insn, notice_source_line): Likewise.
11167 (rest_of_clean_state): Skip inline entry markers.
11168 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
11170 * gimple.c (gimple_build_debug_inline_entry): New.
11171 * gimple.h (enum gimple_debug_subcode): Add
11172 GIMPLE_DEBUG_INLINE_ENTRY.
11173 (gimple_build_debug_inline_entry): Declare.
11174 (gimple_debug_inline_entry_p): New.
11175 (gimple_debug_nonbind_marker_p): Adjust.
11176 * insn-notes.def (INLINE_ENTRY): New.
11177 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
11178 inline entry marker notes.
11179 (print_insn): Likewise.
11180 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
11181 (INSN_DEBUG_MARKER_KIND): Likewise.
11182 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
11183 * tree-inline.c (expand_call_inline): Build and insert
11184 debug_inline_entry stmt.
11185 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
11186 inline entry blocks early, if nonbind markers are enabled.
11187 (dump_scope_block): Dump fragment info.
11188 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
11189 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
11190 (gimple_build_debug_inline_entry): New.
11191 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
11192 Enable/disable inline entry points too.
11193 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
11194 (DEBUG_INSN): Describe inline entry markers.
11196 * common.opt (gvariable-location-views): New.
11197 (gvariable-location-views=incompat5): New.
11198 * config.in: Rebuilt.
11199 * configure: Rebuilt.
11200 * configure.ac: Test assembler for view support.
11201 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
11202 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
11203 * dwarf2out.c (var_loc_view): New typedef.
11204 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
11205 (dwarf2out_locviews_in_attribute): New.
11206 (dwarf2out_locviews_in_loclist): New.
11207 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
11208 (enum dw_line_info_opcode): Add LI_adv_address.
11209 (struct dw_line_info_table): Add view.
11210 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
11211 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
11212 (zero_view_p): New variable.
11213 (ZERO_VIEW_P): New macro.
11214 (output_asm_line_debug_info): New.
11215 (struct var_loc_node): Add view.
11216 (add_AT_view_list, AT_loc_list): New.
11217 (add_var_loc_to_decl): Add view param. Test it against last.
11218 (new_loc_list): Add view params. Record them.
11219 (AT_loc_list_ptr): Handle loc and view lists.
11220 (view_list_to_loc_list_val_node): New.
11221 (print_dw_val): Handle dw_val_class_view_list.
11222 (size_of_die): Likewise.
11223 (value_format): Likewise.
11224 (loc_list_has_views): New.
11225 (gen_llsym): Set vl_symbol too.
11226 (maybe_gen_llsym, skip_loc_list_entry): New.
11227 (dwarf2out_maybe_output_loclist_view_pair): New.
11228 (output_loc_list): Output view list or entries too.
11229 (output_view_list_offset): New.
11230 (output_die): Handle dw_val_class_view_list.
11231 (output_dwarf_version): New.
11232 (output_compilation_unit_header): Use it.
11233 (output_skeleton_debug_sections): Likewise.
11234 (output_rnglists, output_line_info): Likewise.
11235 (output_pubnames, output_aranges): Update version comments.
11236 (output_one_line_info_table): Output view numbers in asm comments.
11237 (dw_loc_list): Determine current endview, pass it to new_loc_list.
11238 Call maybe_gen_llsym.
11239 (loc_list_from_tree_1): Adjust.
11240 (add_AT_location_description): Create view list attribute if
11241 needed, check it's absent otherwise.
11242 (convert_cfa_to_fb_loc_list): Adjust.
11243 (maybe_emit_file): Call output_asm_line_debug_info for test.
11244 (dwarf2out_var_location): Reset views as needed. Precompute
11245 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
11246 attribute. Set view.
11247 (new_line_info_table): Reset next view.
11248 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
11249 (dwarf2out_source_line): Likewise. Output view resets and labels to
11250 the assembler, or select appropriate line info opcodes.
11251 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
11252 (optimize_string_length): Catch it. Adjust.
11253 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
11254 dw_val_class_view_list, and remove it if no longer needed.
11255 (hash_loc_list): Hash view numbers.
11256 (loc_list_hasher::equal): Compare them.
11257 (optimize_location_lists): Check whether a view list symbol is
11258 needed, and whether the locview attribute is present, and
11259 whether they match. Remove the locview attribute if no longer
11261 (index_location_lists): Call skip_loc_list_entry for test.
11262 (dwarf2out_finish): Call output_asm_line_debug_info for test.
11263 Use output_dwarf_version.
11264 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
11265 (struct dw_val_node): Add val_view_list.
11266 * final.c (SEEN_NEXT_VIEW): New.
11267 (set_next_view_needed): New.
11268 (clear_next_view_needed): New.
11269 (maybe_output_next_view): New.
11270 (final_start_function): Rename to...
11271 (final_start_function_1): ... this. Take pointer to FIRST,
11272 add SEEN parameter. Emit param bindings in the initial view.
11273 (final_start_function): Reintroduce SEEN-less interface.
11274 (final): Rename to...
11275 (final_1): ... this. Take SEEN parameter. Output final pending
11276 next view at the end.
11277 (final): Reintroduce seen-less interface.
11278 (final_scan_insn): Output pending next view before switching
11279 sections or ending a block. Mark the next view as needed when
11280 outputting variable locations. Notify debug backend of section
11281 changes, and of location view changes.
11282 (rest_of_handle_final): Adjust.
11283 * toplev.c (process_options): Autodetect value for debug variable
11284 location views option. Warn on incompat5 without -gdwarf-5.
11285 * doc/invoke.texi (gvariable-location-views): New.
11286 (gvariable-location-views=incompat5): New.
11287 (gno-variable-location-views): New.
11289 2018-02-08 David Malcolm <dmalcolm@redhat.com>
11291 PR tree-optimization/84136
11292 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
11293 that the result of find_edge is non-NULL.
11295 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
11298 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
11299 storing integer register in SImode. Fix cost of 256 and 512
11300 byte aligned SSE register store.
11302 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
11304 * config/i386/i386.c (ix86_multiplication_cost): Fix
11305 multiplication cost for TARGET_AVX512DQ.
11307 2018-02-08 Marek Polacek <polacek@redhat.com>
11309 PR tree-optimization/84238
11310 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
11313 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
11315 PR tree-optimization/84265
11316 * tree-vect-stmts.c (vectorizable_store): Don't treat
11317 VMAT_CONTIGUOUS accesses as grouped.
11318 (vectorizable_load): Likewise.
11320 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
11322 PR tree-optimization/81635
11323 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
11324 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
11325 (test_round_for_mask): New functions.
11326 (wide_int_cc_tests): Call test_round_for_mask.
11327 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
11328 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
11329 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
11330 range returned by get_range_info.
11332 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
11335 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
11336 * symtab.c: Include builtins.h
11337 (symtab_node::output_to_lto_symbol_table_p): Move here
11338 from lto-streamer-out.c:output_symbol_p.
11339 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
11340 (output_symbol_p): Move all logic to symtab.c
11341 (produce_symtab): Update.
11343 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
11345 * config/s390/s390-opts.h (enum indirect_branch): Define.
11346 * config/s390/s390-protos.h (s390_return_addr_from_memory)
11347 (s390_indirect_branch_via_thunk)
11348 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
11349 (enum s390_indirect_branch_type): Define.
11350 * config/s390/s390.c (struct s390_frame_layout, struct
11351 machine_function): Remove.
11352 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
11353 (indirect_branch_table_label_no, indirect_branch_table_name):
11355 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
11356 (enum s390_indirect_branch_option): Define.
11357 (s390_return_addr_from_memory): New function.
11358 (s390_handle_string_attribute): New function.
11359 (s390_attribute_table): Add new attribute handler.
11360 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
11361 (s390_indirect_branch_via_thunk): New function.
11362 (s390_indirect_branch_via_inline_thunk): New function.
11363 (s390_function_ok_for_sibcall): When jumping via thunk disallow
11364 sibling call optimization for non z10 compiles.
11365 (s390_emit_call): Force indirect branch target to be a single
11366 register. Add r1 clobber for non-z10 compiles.
11367 (s390_emit_epilogue): Emit return jump via return_use expander.
11368 (s390_reorg): Handle JUMP_INSNs as execute targets.
11369 (s390_option_override_internal): Perform validity checks for the
11370 new command line options.
11371 (s390_indirect_branch_attrvalue): New function.
11372 (s390_indirect_branch_settings): New function.
11373 (s390_set_current_function): Invoke s390_indirect_branch_settings.
11374 (s390_output_indirect_thunk_function): New function.
11375 (s390_code_end): Implement target hook.
11376 (s390_case_values_threshold): Implement target hook.
11377 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
11379 * config/s390/s390.h (struct s390_frame_layout)
11380 (struct machine_function): Move here from s390.c.
11381 (TARGET_INDIRECT_BRANCH_NOBP_RET)
11382 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
11383 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
11384 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
11385 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
11386 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
11387 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
11388 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
11389 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
11390 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
11391 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
11392 (mnemonic attribute): Add values which aren't recognized
11394 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
11395 pattern for branch conversion. Fix mnemonic attribute.
11396 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
11397 indirect branch via thunk if requested.
11398 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
11399 ("*indirect_jump"): Disable for branch conversion using out of
11401 ("indirect_jump_via_thunk<mode>_z10")
11402 ("indirect_jump_via_thunk<mode>")
11403 ("indirect_jump_via_inlinethunk<mode>_z10")
11404 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
11405 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
11406 ("casesi_jump_via_inlinethunk<mode>_z10")
11407 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
11408 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
11409 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
11410 ("*indirect2_jump"): Disable for branch conversion.
11411 ("casesi_jump"): Turn into expander and expand patterns for branch
11413 ("return_use"): New expander.
11414 ("*return"): Emit return via thunk and rename it to ...
11415 ("*return<mode>"): ... this one.
11416 * config/s390/s390.opt: Add new options and and enum for the
11419 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
11421 * lra-constraints.c (match_reload): Unconditionally use
11422 gen_lowpart_SUBREG, rather than selecting between that
11423 and equivalent gen_rtx_SUBREG code.
11425 2018-02-08 Richard Biener <rguenther@suse.de>
11427 PR tree-optimization/84233
11428 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
11429 changed flag instead of boguously re-using phi_inserted.
11431 2018-02-08 Martin Jambor <mjambor@suse.cz>
11433 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
11434 static local variables.
11436 2018-02-08 Richard Biener <rguenther@suse.de>
11438 PR tree-optimization/84278
11439 * tree-vect-stmts.c (vectorizable_store): When looking for
11440 smaller vector types to perform grouped strided loads/stores
11441 make sure the mode is supported by the target.
11442 (vectorizable_load): Likewise.
11444 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
11446 * config/aarch64/aarch64.c (aarch64_components_for_bb):
11447 Increase LDP/STP opportunities by adding adjacent callee-saves.
11449 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
11451 PR rtl-optimization/84068
11452 PR rtl-optimization/83459
11453 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
11455 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
11457 PR tree-optimization/84224
11458 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
11459 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
11460 non-zero arguments.
11462 2018-02-07 Iain Sandoe <iain@codesourcery.com>
11465 * config/rs6000/altivec.md (*restore_world): Remove LR use.
11466 * config/rs6000/predicates.md (restore_world_operation): Adjust op
11467 count, remove one USE.
11469 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
11471 * doc/install.texi (Configuration): Document the
11472 --with-long-double-format={ibm,ieee} PowerPC configuration
11476 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
11477 Convert from define_expand to be define_insn_and_split. Rework
11478 float/double/_Float128 conversions to QI/HI/SImode to work with
11479 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
11480 conversions to QI/HImode types did a store and then a load to
11481 truncate the value. For conversions to VSX registers, don't split
11482 the insn, instead emit the code directly. Use the code iterator
11483 any_fix to combine signed and unsigned conversions.
11484 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
11485 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
11486 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
11487 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
11488 (fix_<mode>di2_hw): Likewise.
11489 (fixuns_<mode>di2_hw): Likewise.
11490 (fix_<mode>si2_hw): Likewise.
11491 (fixuns_<mode>si2_hw): Likewise.
11492 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
11493 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
11494 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
11495 fix<uns>_trunc<SFDF:mode>si2_p8.
11496 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
11497 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
11498 (fix<uns>_<mode>_mem): Likewise.
11499 (fctiw<u>z_<mode>_mem): Likewise.
11500 (fix<uns>_<mode>_mem): Likewise.
11501 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
11502 the register allocator from doing a direct move to the GPRs to do
11503 a store, and instead use the ISA 3.0 store byte/half-word from
11504 vector register instruction. For IEEE 128-bit floating point,
11505 also optimize stores of 32-bit ints.
11506 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
11508 2018-02-07 Alan Hayward <alan.hayward@arm.com>
11510 * genextract.c (push_pathstr_operand): New function to support
11512 (walk_rtx): Call push_pathstr_operand.
11513 (print_path): Support [a-zA-Z].
11515 2018-02-07 Richard Biener <rguenther@suse.de>
11517 PR tree-optimization/84037
11518 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
11519 (cse_and_gimplify_to_preheader): Declare.
11520 (vect_get_place_in_interleaving_chain): Likewise.
11521 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
11523 (_loop_vec_info::~_loop_vec_info): Delete it.
11524 (cse_and_gimplify_to_preheader): New function.
11525 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
11526 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
11527 (vectorizable_load): Likewise. For grouped stores always base
11528 the IV on the first element.
11529 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
11530 condition before gimplifying.
11532 2018-02-07 Jakub Jelinek <jakub@redhat.com>
11534 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
11535 *DIV_EXPR and *MOD_EXPR.
11537 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
11540 * config/i386/i386.c (ix86_option_override_internal): Mask out
11541 the CF_SET bit when checking -fcf-protection.
11543 2018-02-07 Tom de Vries <tom@codesourcery.com>
11546 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
11549 2018-02-07 Richard Biener <rguenther@suse.de>
11551 PR tree-optimization/84204
11552 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
11555 PR tree-optimization/84205
11556 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
11557 special-case isl_ast_op_zdiv_r.
11559 PR tree-optimization/84223
11560 * graphite-scop-detection.c (gather_bbs::before_dom_children):
11561 Only add conditions from within the region.
11562 (gather_bbs::after_dom_children): Adjust.
11564 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
11567 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
11568 * config/avr/avr.md: Only post-reload split REG-REG moves if
11569 either register is GENERAL_REG_P.
11571 2018-02-07 Jakub Jelinek <jakub@redhat.com>
11573 PR tree-optimization/84235
11574 * tree-ssa-scopedtables.c
11575 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
11576 if the subtraction is performed in floating point type where NaNs are
11577 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
11578 build 1. Formatting fix.
11580 2018-02-06 Jakub Jelinek <jakub@redhat.com>
11583 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
11584 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
11585 and skip it regardless of bb boundaries. Use CALL_P macro,
11586 don't test INSN_P (insn) together with CALL_P or JUMP_P check
11587 unnecessarily, formatting fix.
11589 2018-02-06 Michael Collison <michael.collison@arm.com>
11591 * config/arm/thumb2.md:
11592 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
11593 (*thumb_mov_notscc): Ditto.
11595 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
11598 * config/rs6000/rs6000.md (su code attribute): Use "u" for
11599 unsigned_fix, not "s".
11601 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11603 * configure.ac (gcc_fn_eh_frame_ro): New function.
11604 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
11605 correct .eh_frame permissions.
11606 * configure: Regenerate.
11608 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
11610 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
11611 irrelevant options.
11613 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11615 * config/rs6000/rs6000.c (rs6000_option_override_internal):
11616 Display warning message for -mno-speculate-indirect-jumps.
11618 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
11620 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
11622 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
11624 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
11626 PR tree-optimization/84225
11627 * tree-eh.c (find_trapping_overflow): Only call
11628 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
11630 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
11633 * config/i386/i386.c: Reimplement the check of possible options
11634 -mibt/-mshstk conbination. Change error messages.
11635 * doc/invoke.texi: Fix a typo: remove extra '='.
11637 2018-02-06 Marek Polacek <polacek@redhat.com>
11639 PR tree-optimization/84228
11640 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
11642 2018-02-06 Tamar Christina <tamar.christina@arm.com>
11645 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
11646 emitted arch directives.
11647 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
11648 __ARM_FEATURE_COPROC before changing architectures.
11650 2018-02-06 Richard Biener <rguenther@suse.de>
11652 * config/i386/i386.c (print_reg): Fix typo.
11653 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
11655 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
11657 * configure: Regenerate.
11659 2018-02-05 Martin Sebor <msebor@redhat.com>
11661 PR tree-optimization/83369
11662 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
11665 2018-02-05 Martin Liska <mliska@suse.cz>
11667 * doc/invoke.texi: Cherry-pick upstream r323995.
11669 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
11671 * ira.c (ira_init_register_move_cost): Adjust comment.
11673 2018-02-05 Martin Liska <mliska@suse.cz>
11675 PR gcov-profile/84137
11676 * doc/gcov.texi: Fix typo in documentation.
11678 2018-02-05 Martin Liska <mliska@suse.cz>
11680 PR gcov-profile/83879
11681 * doc/gcov.texi: Document necessity of --dynamic-list-data when
11682 using dlopen functionality.
11684 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
11686 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
11687 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
11688 _mm_maskz_range_ss, _mm_mask_range_round_ss,
11689 _mm_maskz_range_round_ss): New intrinsics.
11690 (__builtin_ia32_rangesd128_round)
11691 (__builtin_ia32_rangess128_round): Remove.
11692 (__builtin_ia32_rangesd128_mask_round,
11693 __builtin_ia32_rangess128_mask_round): New builtins.
11694 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
11695 __builtin_ia32_rangess128_round): Remove.
11696 (__builtin_ia32_rangesd128_mask_round,
11697 __builtin_ia32_rangess128_mask_round): New builtins.
11698 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
11699 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
11700 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
11701 "<round_saeonly_constraint>")): Changed to ...
11702 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
11703 "<round_saeonly_scalar_constraint>")): ... this.
11704 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
11705 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
11706 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
11707 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
11708 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
11710 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
11712 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
11714 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
11715 Remove all values except native, 8540 and 8548.
11717 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
11719 * config/i386/i386.c (ix86_output_function_return): Pass
11720 INVALID_REGNUM, instead of -1, as invalid register number to
11721 indirect_thunk_name and output_indirect_thunk.
11723 2018-02-02 Julia Koval <julia.koval@intel.com>
11725 * config.gcc: Add -march=icelake.
11726 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
11727 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
11728 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
11729 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
11730 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
11731 (processor_target_table): Add icelake.
11732 (ix86_option_override_internal): Handle new PTAs.
11733 (get_builtin_code_for_version): Handle icelake.
11734 (M_INTEL_COREI7_ICELAKE): New.
11735 (fold_builtin_cpu): Handle icelake.
11736 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
11737 * doc/invoke.texi: Add -march=icelake.
11739 2018-02-02 Julia Koval <julia.koval@intel.com>
11741 * config/i386/i386.c (ix86_option_override_internal): Change flags type
11742 to wide_int_bitmask.
11743 * wide-int-bitmask.h: New.
11745 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
11748 * config/i386/i386.md: Replace Pmode with word_mode in
11749 builtin_setjmp_setup and builtin_longjmp to support x32.
11751 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
11755 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
11757 (rs6000_supported_cpu_names): New static variable.
11758 (linux_cpu_translation_table): Likewise.
11759 (elf_platform) <cpu>: Define new static variable and use it.
11760 Translate kernel AT_PLATFORM name to canonical name if needed.
11761 Error if platform name is unknown.
11763 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
11766 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
11768 2018-02-01 Jeff Law <law@redhat.com>
11771 * config/i386/i386.c (release_scratch_register_on_entry): Add new
11772 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
11773 the scratch if RELEASE_VIA_POP is false.
11774 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
11775 If we have to save a temporary register, decrement SIZE appropriately.
11776 Pass new arguments to release_scratch_register_on_entry.
11777 (ix86_adjust_stack_and_probe): Likewise.
11778 (ix86_emit_probe_stack_range): Pass new arguments to
11779 release_scratch_register_on_entry.
11781 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
11783 PR rtl-optimization/84157
11784 * combine.c (change_zero_ext): Use REG_P predicate in
11785 front of HARD_REGISTER_P predicate.
11787 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
11789 * config/avr/avr.c (avr_option_override): Move disabling of
11790 -fdelete-null-pointer-checks to...
11791 * common/config/avr/avr-common.c (avr_option_optimization_table):
11794 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11796 PR tree-optimization/81635
11797 * tree-data-ref.c (split_constant_offset_1): For types that
11798 wrap on overflow, try to use range info to prove that wrapping
11801 2018-02-01 Renlin Li <renlin.li@arm.com>
11804 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
11805 TAILCALL_ADDR_REGS.
11806 (aarch64_register_move_cost): Likewise.
11807 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
11808 TAILCALL_ADDR_REGS.
11809 (REG_CLASS_NAMES): Likewise.
11810 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
11811 TAILCALL_ADDR_REGS. Remove IP registers.
11812 * config/aarch64/aarch64.md (Ucs): Update register constraint.
11814 2018-02-01 Richard Biener <rguenther@suse.de>
11816 * domwalk.h (dom_walker::dom_walker): Add additional constructor
11817 for specifying RPO order and allow NULL for that.
11818 * domwalk.c (dom_walker::dom_walker): Likewise.
11819 (dom_walker::walk): Handle NULL RPO order.
11820 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
11822 (rewrite_update_dom_walker): Likewise.
11823 (mark_def_dom_walker): Likewise.
11825 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11827 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
11828 (aarch64_maybe_expand_sve_subreg_move): Declare.
11829 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
11830 * config/aarch64/predicates.md (aarch64_any_register_operand): New
11832 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
11833 that are semantically a reverse operation.
11834 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
11835 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
11836 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
11838 (aarch64_can_change_mode_class): For big-endian, forbid changes
11839 between two SVE modes if they have different element sizes.
11841 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11843 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
11844 the TImode handling for big-endian targets.
11846 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11848 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
11849 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
11851 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
11852 Remove BSWAP handing for big-endian targets and use the form of
11853 LD1RQ appropariate for the mode.
11855 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11857 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
11858 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
11859 duplicated element.
11861 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11864 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
11865 check for operands that need to go through aarch64_sve_reload_be.
11867 2018-02-01 Jakub Jelinek <jakub@redhat.com>
11869 PR tree-optimization/81661
11870 PR tree-optimization/84117
11871 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
11872 * tree-eh.c: Include gimplify.h.
11873 (find_trapping_overflow, replace_trapping_overflow,
11874 rewrite_to_non_trapping_overflow): New functions.
11875 * tree-vect-loop.c: Include tree-eh.h.
11876 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
11877 * tree-data-ref.c: Include tree-eh.h.
11878 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
11880 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
11882 PR rtl-optimization/84123
11883 * combine.c (change_zero_ext): Check if hard register satisfies
11884 can_change_dest_mode before calling gen_lowpart_SUBREG.
11886 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
11889 * ira.c (ira_init_register_move_cost): Remove assert.
11891 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
11893 PR rtl-optimization/84071
11894 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
11895 * doc/tm.texi: Regenerate.
11897 2018-01-31 Richard Biener <rguenther@suse.de>
11899 PR tree-optimization/84132
11900 * tree-data-ref.c (analyze_miv_subscript): Properly
11901 check whether evolution_function_is_affine_multivariate_p
11902 before calling gcd_of_steps_may_divide_p.
11904 2018-01-31 Julia Koval <julia.koval@intel.com>
11907 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
11908 * config/i386/i386.md (rdpid_rex64) New.
11909 (rdpid): Make 32bit only.
11911 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
11914 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
11915 an IDENTIFIER_NODE for FUNCTION_TYPE's.
11917 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
11920 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
11922 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
11924 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
11926 PR rtl-optimization/84071
11927 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
11928 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
11930 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
11932 * config/arc/arc.c (arc_handle_aux_attribute): New function.
11933 (arc_attribute_table): Add 'aux' attribute.
11934 (arc_in_small_data_p): Consider aux like variables.
11935 (arc_is_aux_reg_p): New function.
11936 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
11937 (arc_get_aux_arg): New function.
11938 (prepare_move_operands): Handle aux-register access.
11939 (arc_handle_aux_attribute): New function.
11940 * doc/extend.texi (ARC Variable attributes): Add subsection.
11942 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
11944 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
11945 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
11946 (arc_attribute_table): Add 'uncached' attribute.
11947 (arc_print_operand): Print '.di' flag for uncached memory
11949 (arc_in_small_data_p): Do not consider for small data the uncached
11951 (arc_is_uncached_mem_p): New function.
11952 * config/arc/predicates.md (compact_store_memory_operand): Check
11953 for uncached memory accesses.
11954 (nonvol_nonimm_operand): Likewise.
11955 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
11957 2018-01-31 Jakub Jelinek <jakub@redhat.com>
11960 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
11961 falign-loops=): Add Optimization flag.
11963 2018-01-30 Jeff Law <law@redhat.com>
11966 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
11967 INT_REGISTERS_SAVED. Check it prior to calling
11968 get_scratch_register_on_entry.
11969 (ix86_adjust_stack_and_probe): Similarly.
11970 (ix86_emit_probe_stack_range): Similarly.
11971 (ix86_expand_prologue): Corresponding changes.
11973 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11976 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
11977 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
11979 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
11982 * lra-constraints.c (curr_insn_transform): Process AND in the
11985 2018-01-30 Jakub Jelinek <jakub@redhat.com>
11987 PR rtl-optimization/83986
11988 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
11989 dependence against last_pending_memory_flush in addition to
11990 pending_jump_insns.
11992 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
11994 PR tree-optimization/81611
11995 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
11998 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12001 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
12004 2018-01-30 Richard Biener <rguenther@suse.de>
12005 Jakub Jelinek <jakub@redhat.com>
12007 PR tree-optimization/84111
12008 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
12009 inner loops added during recursion, as they don't have up-to-date
12012 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
12015 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
12016 (can_inline_edge_by_limits_p): ... here.
12017 (can_early_inline_edge_p, check_callers,
12018 update_caller_keys, update_callee_keys, recursive_inlining,
12019 add_new_edges_to_heap, speculation_useful_p,
12020 inline_small_functions,
12021 inline_small_functions, flatten_function,
12022 inline_to_all_callers_1): Update.
12024 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
12026 * profile-count.c (profile_count::combine_with_ipa_count): Handle
12029 2018-01-30 Richard Biener <rguenther@suse.de>
12031 PR tree-optimization/83008
12032 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
12033 invariant and constant vector uses in stmts when they need
12034 more than one stmt.
12036 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12039 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
12040 * configure: Regenerate.
12042 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
12044 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
12046 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
12047 Use gen_rtx_REG rather than gen_lowpart.
12049 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
12051 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
12052 rather than 0 when creating partial subregs.
12054 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
12056 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
12059 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
12062 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
12063 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
12064 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
12065 flags. This restores the settings used before the 2017-07-24.
12066 Turning off pre increment/decrement/modify allows IVOPTS to
12067 optimize DF/SF loops where the index is an int.
12069 2018-01-29 Richard Biener <rguenther@suse.de>
12070 Kelvin Nilsen <kelvin@gcc.gnu.org>
12073 * tree-vect-stmts.c (vectorizable_call): Don't call
12074 targetm.vectorize_builtin_md_vectorized_function if callee is
12077 2018-01-22 Carl Love <cel@us.ibm.com>
12079 * doc/extend.tex: Fix typo in second arg in
12080 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
12082 2018-01-29 Richard Biener <rguenther@suse.de>
12084 PR tree-optimization/84086
12085 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
12086 (flush_ssaname_freelist): When SSA names were released reset
12087 the SCEV hash table.
12089 2018-01-29 Richard Biener <rguenther@suse.de>
12091 PR tree-optimization/84057
12092 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
12093 removed paths when removing edges.
12095 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
12097 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
12098 -mfunction-return=@var{choice}.
12100 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
12102 PR diagnostic/84034
12103 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
12104 Handle CR like TAB.
12105 (layout::print_source_line): Likewise.
12106 (test_get_line_width_without_trailing_whitespace): Add test cases.
12108 2018-01-27 Jakub Jelinek <jakub@redhat.com>
12110 PR middle-end/84040
12111 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
12114 2018-01-26 Jim Wilson <jimw@sifive.com>
12116 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
12118 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
12121 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12123 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
12124 and CMP + SUB-immediate -> SUBS.
12126 2018-01-26 Martin Sebor <msebor@redhat.com>
12128 PR tree-optimization/83896
12129 * tree-ssa-strlen.c (get_string_len): Rename...
12130 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
12131 Avoid assuming length is constant.
12132 (handle_char_store): Use HOST_WIDE_INT for string length.
12134 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
12137 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
12138 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
12140 2018-01-26 Richard Biener <rguenther@suse.de>
12142 PR rtl-optimization/84003
12143 * dse.c (record_store): Only record redundant stores when
12144 the earlier store aliases at least all accesses the later one does.
12146 2018-01-26 Jakub Jelinek <jakub@redhat.com>
12148 PR rtl-optimization/83985
12149 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
12150 REG_CFA_RESTORE insns.
12151 (delete_unmarked_insns): Don't ignore separate shrink wrapping
12152 REG_CFA_RESTORE insns here.
12155 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
12156 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
12158 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
12160 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
12161 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
12162 (arc_init): Likewise.
12163 (arc_override_options): Likewise.
12164 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
12166 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
12168 * config/arc/arc.h (TARGET_DBNZ): Define.
12169 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
12170 properly set the tune attribute.
12171 (dbnz): Use TARGET_DBNZ guard.
12172 * config/arc/arc.opt (mtune): Add core3 option.
12174 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
12176 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
12177 recognize new pic like addresses.
12178 (arc_delegitimize_address): Clean up.
12180 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
12182 * config/arc/arc-arches.def: Option mrf16 valid for all
12184 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
12185 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
12186 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
12187 * config/arc/arc-tables.opt: Regenerate.
12188 * config/arc/arc.c (arc_conditional_register_usage): Handle
12189 reduced register file case.
12190 (arc_file_start): Set must have build attributes.
12191 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
12192 mrf16 option value.
12193 * config/arc/arc.opt (mrf16): Add new option.
12194 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
12195 * config/arc/genmultilib.awk: Handle new mrf16 option.
12196 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
12197 * config/arc/t-multilib: Regenerate.
12198 * doc/invoke.texi (ARC Options): Document mrf16 option.
12200 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
12202 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
12203 * config/arc/arc.c (arc_handle_secure_attribute): New function.
12204 (arc_attribute_table): Add 'secure_call' attribute.
12205 (arc_print_operand): Print secure call operand.
12206 (arc_function_ok_for_sibcall): Don't optimize tail calls when
12208 (arc_is_secure_call_p): New function. * config/arc/arc.md
12209 (call_i): Add support for sjli instruction.
12210 (call_value_i): Likewise.
12211 * config/arc/constraints.md (Csc): New constraint.
12213 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
12214 John Eric Martin <John.Martin@emmicro-us.com>
12216 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
12217 * config/arc/arc.c (_arc_jli_section): New struct.
12218 (arc_jli_section): New type.
12219 (rc_jli_sections): New static variable.
12220 (arc_handle_jli_attribute): New function.
12221 (arc_attribute_table): Add jli_always and jli_fixed attribute.
12222 (arc_file_end): New function.
12223 (TARGET_ASM_FILE_END): Define.
12224 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
12225 (arc_add_jli_section): New function.
12226 (jli_call_scan): Likewise.
12227 (arc_reorg): Call jli_call_scan.
12228 (arc_output_addsi): Remove 'S' from printing asm operand.
12229 (arc_is_jli_call_p): New function.
12230 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
12232 (movhi_insn): Likewise.
12233 (movsi_insn): Likewise.
12234 (movsi_set_cc_insn): Likewise.
12235 (loadqi_update): Likewise.
12236 (load_zeroextendqisi_update): Likewise.
12237 (load_signextendqisi_update): Likewise.
12238 (loadhi_update): Likewise.
12239 (load_zeroextendhisi_update): Likewise.
12240 (load_signextendhisi_update): Likewise.
12241 (loadsi_update): Likewise.
12242 (loadsf_update): Likewise.
12243 (movsicc_insn): Likewise.
12244 (bset_insn): Likewise.
12245 (bxor_insn): Likewise.
12246 (bclr_insn): Likewise.
12247 (bmsk_insn): Likewise.
12248 (bicsi3_insn): Likewise.
12249 (cmpsi_cc_c_insn): Likewise.
12250 (movsi_ne): Likewise.
12251 (movsi_cond_exec): Likewise.
12252 (clrsbsi2): Likewise.
12253 (norm_f): Likewise.
12263 (call_i): Remove 'S' asm letter, add jli instruction.
12264 (call_value_i): Likewise.
12265 * config/arc/arc.op (mjli-always): New option.
12266 * config/arc/constraints.md (Cji): New constraint.
12267 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
12269 (subsf3_fpx): Likewise.
12270 (mulsf3_fpx): Likewise.
12271 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
12273 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
12274 function attrbutes.
12275 * doc/invoke.texi (ARC): Document mjli-always option.
12277 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
12279 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
12280 avoid addition with 0 and use incw and decw where possible.
12282 2018-01-26 Richard Biener <rguenther@suse.de>
12284 PR tree-optimization/81082
12285 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
12286 association if it requires casting to unsigned.
12287 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
12288 from fold_plusminus_mult_expr to catch important cases late when
12289 range info is available.
12291 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12293 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
12294 * configure.ac (hidden_linkonce): New test.
12295 * configure: Regenerate.
12296 * config.in: Regenerate.
12298 2018-01-26 Julia Koval <julia.koval@intel.com>
12300 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
12301 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
12302 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
12303 _mm_mask_bitshuffle_epi64_mask): Fix type.
12304 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
12305 USI_FTYPE_V4DI_V4DI_USI): Remove.
12306 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
12307 __builtin_ia32_vpshufbitqmb256_mask,
12308 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
12309 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
12310 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
12312 2018-01-26 Alan Modra <amodra@gmail.com>
12315 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
12316 UNSPEC_VBPERMQ. Sort other unspecs.
12318 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
12320 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
12322 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
12324 PR middle-end/83055
12325 * predict.c (drop_profile): Do not push/pop cfun; update also
12327 (handle_missing_profiles): Fix logic looking for zero profiles.
12329 2018-01-25 Jakub Jelinek <jakub@redhat.com>
12331 PR middle-end/83977
12332 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
12333 on functions with #pragma omp declare simd or functions with simd
12335 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
12336 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
12337 Remove trailing \n from warning_at calls.
12339 2018-01-25 Tom de Vries <tom@codesourcery.com>
12342 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
12343 for neutered workers.
12345 2018-01-24 Joseph Myers <joseph@codesourcery.com>
12348 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
12349 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
12351 2018-01-24 Jeff Law <law@redhat.com>
12354 * i386.c (get_probe_interval): Move to earlier point.
12355 (ix86_compute_frame_layout): If -fstack-clash-protection and
12356 the frame is larger than the probe interval, then use pushes
12357 to save registers rather than reg->mem moves.
12358 (ix86_expand_prologue): Remove conditional for int_registers_saved
12361 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
12364 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
12365 min/max for never referenced object.
12367 2018-01-24 Jakub Jelinek <jakub@redhat.com>
12369 PR middle-end/83977
12370 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
12372 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
12373 attributes from DECL_ATTRIBUTES (decl) without affecting
12374 DECL_ATTRIBUTES (current_function_decl).
12375 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
12376 functions with non-NULL DECL_ABSTRACT_ORIGIN.
12378 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
12380 PR tree-optimization/83979
12381 * fold-const.c (fold_comparison): Use constant_boolean_node
12382 instead of boolean_{true,false}_node.
12384 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
12386 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
12389 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12391 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
12392 Simplify the clause that sets the length attribute.
12393 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
12394 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
12395 clause that sets the length attribute.
12396 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
12398 2018-01-24 Tom de Vries <tom@codesourcery.com>
12401 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
12402 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
12403 Add strict parameter.
12404 (prevent_branch_around_nothing): Insert dummy insn between branch to
12405 label and label with no ptx insn inbetween.
12406 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
12408 2018-01-24 Tom de Vries <tom@codesourcery.com>
12411 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
12412 for neutered threads in warp.
12413 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
12415 2018-01-24 Richard Biener <rguenther@suse.de>
12417 PR tree-optimization/83176
12418 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
12421 2018-01-24 Richard Biener <rguenther@suse.de>
12423 PR tree-optimization/82819
12424 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
12425 code generating pluses that are no-ops in the target precision.
12427 2018-01-24 Richard Biener <rguenther@suse.de>
12429 PR middle-end/84000
12430 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
12432 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
12434 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
12435 to merge probabilities.
12436 * predict.c (probably_never_executed): Also mark as cold functions
12437 with global 0 profile and guessed local profile.
12438 * profile-count.c (profile_probability::combine_with_count): New
12440 * profile-count.h (profile_probability::operator*,
12441 profile_probability::operator*=, profile_probability::operator/,
12442 profile_probability::operator/=): Reduce precision to adjusted
12443 and set value to guessed on contradictory divisions.
12444 (profile_probability::combine_with_freq): Remove.
12445 (profile_probability::combine_wiht_count): Declare.
12446 (profile_count::force_nonzero):: Set to adjusted.
12447 (profile_count::probability_in):: Set quality to adjusted.
12448 * tree-ssa-tail-merge.c (replace_block_by): Use
12449 combine_with_count.
12451 2018-01-23 Andrew Waterman <andrew@sifive.com>
12452 Jim Wilson <jimw@sifive.com>
12454 * config/riscv/riscv.c (riscv_stack_boundary): New.
12455 (riscv_option_override): Set riscv_stack_boundary. Handle
12456 riscv_preferred_stack_boundary_arg.
12457 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
12458 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
12459 (STACK_BOUNDARY): Set to riscv_stack_boundary.
12460 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
12461 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
12462 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
12464 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
12467 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
12468 of struct ix86_frame.
12469 (ix86_expand_epilogue): Likewise. Add a local variable for
12470 the reg_save_offset field in struct ix86_frame.
12472 2018-01-23 Bin Cheng <bin.cheng@arm.com>
12474 PR tree-optimization/82604
12475 * tree-loop-distribution.c (enum partition_kind): New enum item
12476 PKIND_PARTIAL_MEMSET.
12477 (partition_builtin_p): Support above new enum item.
12478 (generate_code_for_partition): Ditto.
12479 (compute_access_range): Differentiate cases that equality can be
12480 proven at all loops, the innermost loops or no loops.
12481 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
12482 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
12483 (finalize_partitions, distribute_loop): Don't fuse partition of
12484 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
12485 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
12486 parloop is enabled.
12488 2018-01-23 Martin Liska <mliska@suse.cz>
12490 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
12491 order to ignore the predictor.
12492 (PRED_POLYMORPHIC_CALL): Likewise.
12493 (PRED_RECURSIVE_CALL): Likewise.
12495 2018-01-23 Martin Liska <mliska@suse.cz>
12497 * tree-profile.c (tree_profiling): Print function header to
12498 aware reader which function we are working on.
12499 * value-prof.c (gimple_find_values_to_profile): Do not print
12500 not interesting value histograms.
12502 2018-01-23 Martin Liska <mliska@suse.cz>
12504 * profile-count.h (enum profile_quality): Add
12505 profile_uninitialized as the first value. Do not number values
12506 as they are zero based.
12507 (profile_count::verify): Update sanity check.
12508 (profile_probability::verify): Likewise.
12510 2018-01-23 Nathan Sidwell <nathan@acm.org>
12512 * doc/invoke.texi (ffor-scope): Deprecate.
12514 2018-01-23 David Malcolm <dmalcolm@redhat.com>
12516 PR tree-optimization/83510
12517 * domwalk.c (set_all_edges_as_executable): New function.
12518 (dom_walker::dom_walker): Convert bool param
12519 "skip_unreachable_blocks" to enum reachability. Move setup of
12520 edge flags to set_all_edges_as_executable and only do it when
12521 reachability is REACHABLE_BLOCKS.
12522 * domwalk.h (enum dom_walker::reachability): New enum.
12523 (dom_walker::dom_walker): Convert bool param
12524 "skip_unreachable_blocks" to enum reachability.
12525 (set_all_edges_as_executable): New decl.
12526 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
12527 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
12529 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
12530 but converting true to REACHABLE_BLOCKS.
12531 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
12533 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
12534 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
12535 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
12537 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
12538 if check_all_array_refs will be called.
12540 2018-01-23 David Malcolm <dmalcolm@redhat.com>
12542 * tree.c (selftest::test_location_wrappers): Add more test
12545 2018-01-23 David Malcolm <dmalcolm@redhat.com>
12547 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
12548 (selftest::test_bit_in_range): Likewise.
12550 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
12553 * doc/sourcebuild.texi (vect_float): Say that the selector
12554 only describes the situation when -funsafe-math-optimizations is on.
12555 (vect_float_strict): Document.
12557 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
12559 PR tree-optimization/83965
12560 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
12561 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
12562 instead of checking only for a reduction.
12563 (vect_recog_widen_sum_pattern): Likewise.
12565 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
12567 * predict.c (probably_never_executed): Only use precise profile info.
12568 (compute_function_frequency): Skip after inlining hack since we now
12569 have quality checking.
12571 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
12573 * profile-count.h (profile_probability::very_unlikely,
12574 profile_probability::unlikely, profile_probability::even): Set
12575 precision to guessed.
12577 2018-01-23 Richard Biener <rguenther@suse.de>
12579 PR tree-optimization/83963
12580 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
12581 Properly terminate dominator walk when crossing the exit edge not
12582 when visiting its source block.
12584 2018-01-23 Jakub Jelinek <jakub@redhat.com>
12587 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
12588 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
12590 2018-01-22 Jakub Jelinek <jakub@redhat.com>
12592 PR tree-optimization/83957
12593 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
12594 semicolon after for body surrounded by braces.
12596 PR tree-optimization/83081
12597 * profile-count.h (profile_probability::split): New method.
12598 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
12599 Use profile_probability::split.
12600 (do_compare_rtx_and_jump): Fix adjustment of probabilities
12601 when splitting a single conditional jump into 2.
12603 2018-01-22 David Malcolm <dmalcolm@redhat.com>
12605 PR tree-optimization/69452
12606 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
12609 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12611 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
12612 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
12613 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
12615 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12617 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
12619 * config/rl78/rl78.md (movdi): New define_expand.
12620 * config/rl78/rl78.c (rl78_split_movdi): New function.
12622 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
12625 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
12627 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
12628 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
12629 128-bit to produce an UNSPEC move to get the double word with the
12630 signbit and then a shift directly to do signbit.
12631 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
12632 implementation with a new version that just does either a direct
12633 move or a regular move. Move memory interface to separate insns.
12634 Move insns so they are next to the expander.
12635 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
12636 with signbit move. Split big and little endian case.
12637 (signbit<mode>2_dm_mem_le): Likewise.
12638 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
12639 (signbit<mode>2_dm2): Likewise.
12641 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12643 * config/rl78/rl78.md (anddi3): New define_expand.
12645 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12647 * config/rl78/rl78.md (umindi3): New define_expand.
12649 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12651 * config/rl78/rl78.md (smindi3): New define_expand.
12653 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12655 * config/rl78/rl78.md (smaxdi3): New define_expand.
12657 2018-01-22 Carl Love <cel@us.ibm.com>
12659 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
12660 LVX_V1TI): Add macro expansion.
12661 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
12662 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
12663 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
12664 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
12665 Change check to determine if the instruction is a byte reversing
12666 entry. Fix typo in comment.
12667 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
12668 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
12669 Add def_builtin calls for new builtins.
12670 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
12671 Add define_insn expansion.
12673 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12675 * config/rl78/rl78.md (umaxdi3): New define_expand.
12677 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12679 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
12680 for non-QImode registers.
12682 2018-01-22 Richard Biener <rguenther@suse.de>
12684 PR tree-optimization/83963
12685 * graphite-scop-detection.c (scop_detection::get_sese): Delay
12686 including the loop exit block.
12687 (scop_detection::merge_sese): Likewise.
12688 (scop_detection::add_scop): Do it here instead.
12690 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12692 * doc/sourcebuild.texi (arm_softfloat): Document.
12694 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
12697 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
12698 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
12699 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
12701 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12702 David Edelsohn <dje.gcc@gmail.com>
12705 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
12706 Change "crset eq" to "crset 2".
12707 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
12708 (*call_indirect_aix<mode>_nospec): Likewise.
12709 (*call_value_indirect_aix<mode>_nospec): Likewise.
12710 (*call_indirect_elfv2<mode>_nospec): Likewise.
12711 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
12712 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
12713 change assembly output from . to $.
12714 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
12715 (indirect_jump<mode>_nospec): Change assembly output from . to $.
12716 (*tablejump<mode>_internal1_nospec): Likewise.
12718 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
12721 * config/sh/sh_optimize_sett_clrt.cc:
12722 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
12724 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
12726 PR tree-optimization/83940
12727 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
12728 offset_dt to vect_constant_def rather than vect_unknown_def_type.
12729 (vect_check_load_store_mask): Add a mask_dt_out parameter and
12730 use it to pass back the definition type.
12731 (vect_check_store_rhs): Likewise rhs_dt_out.
12732 (vect_build_gather_load_calls): Add a mask_dt argument and use
12733 it instead of a call to vect_is_simple_use.
12734 (vectorizable_store): Update calls to vect_check_load_store_mask
12735 and vect_check_store_rhs. Use the dt returned by the latter instead
12736 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
12737 instead of calls to vect_is_simple_use. Pass the scalar rather
12738 than the vector operand to vect_is_simple_use when handling
12739 second and subsequent copies of an rhs value.
12740 (vectorizable_load): Update calls to vect_check_load_store_mask
12741 and vect_build_gather_load_calls. Use the cached mask_dt and
12742 gs_info.offset_dt instead of calls to vect_is_simple_use.
12744 2018-01-20 Jakub Jelinek <jakub@redhat.com>
12746 PR middle-end/83945
12747 * tree-emutls.c: Include gimplify.h.
12748 (lower_emutls_2): New function.
12749 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
12750 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
12751 it before further processing.
12754 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
12755 UINTVAL (trueop1) instead of INTVAL (op1).
12757 2018-01-19 Jakub Jelinek <jakub@redhat.com>
12761 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
12762 INCOMING_FRAME_SP_OFFSET if not defined.
12763 (scan_trace): Add ENTRY argument. If true and
12764 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
12765 emit a note to adjust the CFA offset.
12766 (create_cfi_notes): Adjust scan_trace callers.
12767 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
12768 INCOMING_FRAME_SP_OFFSET in the CIE.
12769 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
12770 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
12772 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
12773 * doc/tm.texi: Regenerated.
12775 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
12777 PR rtl-optimization/83147
12778 * lra-constraints.c (remove_inheritance_pseudos): Use
12779 lra_substitute_pseudo_within_insn.
12781 2018-01-19 Tom de Vries <tom@codesourcery.com>
12782 Cesar Philippidis <cesar@codesourcery.com>
12785 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
12787 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
12790 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
12791 spaces for function labels.
12793 2018-01-19 Martin Liska <mliska@suse.cz>
12795 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
12796 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
12797 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
12798 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
12799 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
12800 (PRED_CONST_RETURN): Change from 69 to 65.
12801 (PRED_NULL_RETURN): Change from 91 to 71.
12802 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
12803 (PRED_LOOP_GUARD): Change from 66 to 73.
12805 2018-01-19 Martin Liska <mliska@suse.cz>
12807 * predict.c (predict_insn_def): Add new assert.
12808 (struct branch_predictor): Change type to signed integer.
12809 (test_prediction_value_range): Amend test to cover
12810 PROB_UNINITIALIZED.
12811 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
12812 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
12813 (PRED_LOOP_ITERATIONS_MAX): Likewise.
12814 (PRED_LOOP_IV_COMPARE): Likewise.
12815 * predict.h (PROB_UNINITIALIZED): Define new constant.
12817 2018-01-19 Martin Liska <mliska@suse.cz>
12819 * predict.c (dump_prediction): Add new format for
12820 analyze_brprob.py script which is enabled with -details
12822 * profile-count.h (precise_p): New function.
12824 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
12826 PR tree-optimization/83922
12827 * tree-vect-loop.c (vect_verify_full_masking): Return false if
12828 there are no statements that need masking.
12829 (vect_active_double_reduction_p): New function.
12830 (vect_analyze_loop_operations): Use it when handling phis that
12831 are not in the loop header.
12833 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
12835 PR tree-optimization/83914
12836 * tree-vect-loop.c (vectorizable_induction): Don't convert
12837 init_expr or apply the peeling adjustment for inductions
12838 that are nested within the vectorized loop.
12840 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12842 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
12845 2018-01-18 Jakub Jelinek <jakub@redhat.com>
12849 * function.h (gimplify_parameters): Add gimple_seq * argument.
12850 * function.c: Include gimple.h and options.h.
12851 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
12852 for the added local temporaries if needed.
12853 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
12854 if there are any parameter cleanups, wrap whole body into a
12855 try/finally with the cleanups.
12857 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
12860 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
12861 Use GET_MODE_CLASS for scalar floating point.
12863 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
12867 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
12868 Fix call of call_cgraph_insertion_hooks.
12870 2018-01-18 Martin Sebor <msebor@redhat.com>
12872 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
12874 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
12877 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
12880 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
12883 * common.opt: (-ffile-prefix-map): New option.
12884 * opts.c (common_handle_option): Defer it.
12885 * opts-global.c (handle_common_deferred_options): Handle it.
12886 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
12887 * file-prefix-map.h: New file.
12888 (remap_debug_filename, add_debug_prefix_map): ...here.
12889 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
12890 * final.c (debug_prefix_map, add_debug_prefix_map
12891 remap_debug_filename): Move to...
12892 * file-prefix-map.c: New file.
12893 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
12894 generalize, get rid of alloca(), use strrchr() instead of strchr().
12895 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
12896 Implement in terms of add_prefix_map().
12897 (remap_macro_filename, remap_debug_filename): Implement in term of
12899 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
12900 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
12901 * dbxout.c: Include file-prefix-map.h.
12902 * varasm.c: Likewise.
12903 * vmsdbgout.c: Likewise.
12904 * xcoffout.c: Likewise.
12905 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
12906 * doc/cppopts.texi (-fmacro-prefix-map): Document.
12907 * doc/invoke.texi (-ffile-prefix-map): Document.
12908 (-fdebug-prefix-map): Update description.
12910 2018-01-18 Martin Liska <mliska@suse.cz>
12912 * config/i386/i386.c (indirect_thunk_name): Document that also
12914 (output_indirect_thunk): Document why both instructions
12915 (pause and lfence) are generated.
12917 2018-01-18 Richard Biener <rguenther@suse.de>
12919 PR tree-optimization/83887
12920 * graphite-scop-detection.c
12921 (scop_detection::get_nearest_dom_with_single_entry): Remove.
12922 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
12923 (scop_detection::merge_sese): Re-implement with a flood-fill
12924 algorithm that properly finds a SESE region if it exists.
12926 2018-01-18 Jakub Jelinek <jakub@redhat.com>
12929 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
12930 pointer_diff optimizations use view_convert instead of convert.
12932 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12934 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
12935 Generate different code for -mno-speculate-indirect-jumps.
12936 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
12937 (*call_indirect_aix<mode>): Disable for
12938 -mno-speculate-indirect-jumps.
12939 (*call_indirect_aix<mode>_nospec): New define_insn.
12940 (*call_value_indirect_aix<mode>): Disable for
12941 -mno-speculate-indirect-jumps.
12942 (*call_value_indirect_aix<mode>_nospec): New define_insn.
12943 (*sibcall_nonlocal_sysv<mode>): Generate different code for
12944 -mno-speculate-indirect-jumps.
12945 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
12947 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
12949 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
12950 long double type, set the flags for noting the default long double
12951 type, even if we don't pass or return a long double type.
12953 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
12956 * ipa-inline.c (flatten_function): Do not overwrite final inlining
12959 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
12961 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
12962 support for merge[hl].
12963 (fold_mergehl_helper): New helper function.
12964 (tree-vector-builder.h): New #include for tree_vector_builder usage.
12965 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
12966 (altivec_vmrglw_direct): Add xxmrglw insn.
12968 2018-01-17 Andrew Waterman <andrew@sifive.com>
12970 * config/riscv/riscv.c (riscv_conditional_register_usage): If
12971 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
12973 2018-01-17 David Malcolm <dmalcolm@redhat.com>
12976 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
12977 call the lto_location_cache before reading the
12978 DECL_SOURCE_LOCATION of the types.
12980 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
12981 Richard Sandiford <richard.sandiford@linaro.org>
12983 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
12984 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
12985 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
12986 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
12987 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
12989 * config/aarch64/constraints.md (aarch64_movti_operand):
12991 * config/aarch64/predicates.md (Uti): Add new constraint.
12993 2018-01-17 Carl Love <cel@us.ibm.com>
12995 * config/rs6000/vsx.md (define_expand xl_len_r,
12996 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
12997 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
12999 (define_expand, define_insn): Move the shift left from the
13000 define_insn to the define_expand for lxvl and stxvl instructions.
13001 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
13002 and XL_LEN_R definitions to PURE.
13004 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
13006 * config/i386/i386.c (indirect_thunk_name): Declare regno
13007 as unsigned int. Compare regno with INVALID_REGNUM.
13008 (output_indirect_thunk): Ditto.
13009 (output_indirect_thunk_function): Ditto.
13010 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
13011 in the call to output_indirect_thunk_function.
13013 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
13015 PR middle-end/83884
13016 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
13017 rather than the size of inner_type to determine the stack slot size
13018 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
13020 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
13023 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
13026 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
13028 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
13029 endian Linux systems to optionally enable multilibs for selecting
13030 the long double type if the user configured an explicit type.
13031 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
13032 have no long double multilibs if not defined.
13033 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
13034 warn if the user used -mabi={ieee,ibm}longdouble and we built
13035 multilibs for long double.
13036 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
13037 appropriate multilib option.
13038 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
13040 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
13041 for building long double multilibs.
13042 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
13044 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
13046 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
13049 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
13051 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
13054 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
13057 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
13060 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
13062 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
13063 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
13065 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
13067 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
13068 different rtl trees depending on TARGET_64BIT.
13069 (rs6000_gen_lvx): Likewise.
13071 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
13073 * config/visium/visium.md (nop): Tweak comment.
13074 (hazard_nop): Likewise.
13076 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13078 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
13079 -mspeculate-indirect-jumps.
13080 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
13081 for -mno-speculate-indirect-jumps.
13082 (*call_indirect_elfv2<mode>_nospec): New define_insn.
13083 (*call_value_indirect_elfv2<mode>): Disable for
13084 -mno-speculate-indirect-jumps.
13085 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
13086 (indirect_jump): Emit different RTL for
13087 -mno-speculate-indirect-jumps.
13088 (*indirect_jump<mode>): Disable for
13089 -mno-speculate-indirect-jumps.
13090 (*indirect_jump<mode>_nospec): New define_insn.
13091 (tablejump): Emit different RTL for
13092 -mno-speculate-indirect-jumps.
13093 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
13094 (tablejumpsi_nospec): New define_expand.
13095 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
13096 (tablejumpdi_nospec): New define_expand.
13097 (*tablejump<mode>_internal1): Disable for
13098 -mno-speculate-indirect-jumps.
13099 (*tablejump<mode>_internal1_nospec): New define_insn.
13100 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
13103 2018-01-16 Artyom Skrobov tyomitch@gmail.com
13105 * caller-save.c (insert_save): Drop unnecessary parameter. All
13108 2018-01-16 Jakub Jelinek <jakub@redhat.com>
13109 Richard Biener <rguenth@suse.de>
13112 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
13113 return early, inline manually is_gimple_sizepos. Make sure if we
13114 call gimplify_expr we don't end up with a gimple constant.
13115 * tree.c (variably_modified_type_p): Don't return true for
13116 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
13117 * gimplify.h (is_gimple_sizepos): Remove.
13119 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
13121 PR tree-optimization/83857
13122 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
13123 vectorizable_live_operation for pure SLP statements.
13124 (vectorizable_live_operation): Handle PHIs.
13126 2018-01-16 Richard Biener <rguenther@suse.de>
13128 PR tree-optimization/83867
13129 * tree-vect-stmts.c (vect_transform_stmt): Precompute
13130 nested_in_vect_loop_p since the scalar stmt may get invalidated.
13132 2018-01-16 Jakub Jelinek <jakub@redhat.com>
13135 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
13136 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
13137 If off is not INTEGER_CST, issue a may not be aligned warning
13138 rather than isn't aligned. Use isn%'t rather than isn't.
13139 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
13141 <case MULT_EXPR>: Improve the case when bottom and one of the
13142 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
13143 operand, in that case check if the other operand is multiple of
13144 bottom divided by the INTEGER_CST operand.
13146 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
13149 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
13150 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
13151 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
13152 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
13153 * config/pa/pa.c (pa_function_arg_advance): Likewise.
13154 (pa_function_arg, pa_arg_partial_bytes): Likewise.
13155 (pa_function_arg_size): New function.
13157 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
13159 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
13160 in a separate statement.
13162 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
13164 PR tree-optimization/83847
13165 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
13166 group gathers and scatters.
13168 2018-01-16 Jakub Jelinek <jakub@redhat.com>
13170 PR rtl-optimization/86620
13171 * params.def (max-sched-ready-insns): Bump minimum value to 1.
13173 PR rtl-optimization/83213
13174 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
13175 to last if both are JUMP_INSNs.
13177 PR tree-optimization/83843
13178 * gimple-ssa-store-merging.c
13179 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
13180 store_immediate_info for bswap/nop orig_stores.
13182 2018-01-15 Andrew Waterman <andrew@sifive.com>
13184 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
13186 <UDIV>: Increase cost if !TARGET_DIV.
13188 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
13190 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
13191 (define_attr "cr_logical_3op"): New.
13192 (cceq_ior_compare): Adjust.
13193 (cceq_ior_compare_complement): Adjust.
13194 (*cceq_rev_compare): Adjust.
13195 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
13196 (is_cracked_insn): Adjust.
13197 (insn_must_be_first_in_group): Adjust.
13198 * config/rs6000/40x.md: Adjust.
13199 * config/rs6000/440.md: Adjust.
13200 * config/rs6000/476.md: Adjust.
13201 * config/rs6000/601.md: Adjust.
13202 * config/rs6000/603.md: Adjust.
13203 * config/rs6000/6xx.md: Adjust.
13204 * config/rs6000/7450.md: Adjust.
13205 * config/rs6000/7xx.md: Adjust.
13206 * config/rs6000/8540.md: Adjust.
13207 * config/rs6000/cell.md: Adjust.
13208 * config/rs6000/e300c2c3.md: Adjust.
13209 * config/rs6000/e500mc.md: Adjust.
13210 * config/rs6000/e500mc64.md: Adjust.
13211 * config/rs6000/e5500.md: Adjust.
13212 * config/rs6000/e6500.md: Adjust.
13213 * config/rs6000/mpc.md: Adjust.
13214 * config/rs6000/power4.md: Adjust.
13215 * config/rs6000/power5.md: Adjust.
13216 * config/rs6000/power6.md: Adjust.
13217 * config/rs6000/power7.md: Adjust.
13218 * config/rs6000/power8.md: Adjust.
13219 * config/rs6000/power9.md: Adjust.
13220 * config/rs6000/rs64.md: Adjust.
13221 * config/rs6000/titan.md: Adjust.
13223 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
13225 * config/i386/predicates.md (indirect_branch_operand): Rewrite
13226 ix86_indirect_branch_register logic.
13228 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
13230 * config/i386/constraints.md (Bs): Update
13231 ix86_indirect_branch_register check. Don't check
13232 ix86_indirect_branch_register with GOT_memory_operand.
13234 * config/i386/predicates.md (GOT_memory_operand): Don't check
13235 ix86_indirect_branch_register here.
13236 (GOT32_symbol_operand): Likewise.
13238 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
13240 * config/i386/predicates.md (constant_call_address_operand):
13241 Rewrite ix86_indirect_branch_register logic.
13242 (sibcall_insn_operand): Likewise.
13244 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
13246 * config/i386/constraints.md (Bs): Replace
13247 ix86_indirect_branch_thunk_register with
13248 ix86_indirect_branch_register.
13250 * config/i386/i386.md (indirect_jump): Likewise.
13251 (tablejump): Likewise.
13252 (*sibcall_memory): Likewise.
13253 (*sibcall_value_memory): Likewise.
13254 Peepholes of indirect call and jump via memory: Likewise.
13255 * config/i386/i386.opt: Likewise.
13256 * config/i386/predicates.md (indirect_branch_operand): Likewise.
13257 (GOT_memory_operand): Likewise.
13258 (call_insn_operand): Likewise.
13259 (sibcall_insn_operand): Likewise.
13260 (GOT32_symbol_operand): Likewise.
13262 2018-01-15 Jakub Jelinek <jakub@redhat.com>
13264 PR middle-end/83837
13265 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
13266 type rather than type addr's type points to.
13267 (expand_omp_atomic_mutex): Likewise.
13268 (expand_omp_atomic): Likewise.
13270 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
13273 * config/i386/i386.c (output_indirect_thunk_function): Use
13274 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
13275 for __x86_return_thunk.
13277 2018-01-15 Richard Biener <rguenther@suse.de>
13279 PR middle-end/83850
13280 * expmed.c (extract_bit_field_1): Fix typo.
13282 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13285 * config/arm/iterators.md (VF): New mode iterator.
13286 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
13287 Remove integer-related logic from pattern.
13288 (neon_vabd<mode>_3): Likewise.
13290 2018-01-15 Jakub Jelinek <jakub@redhat.com>
13292 PR middle-end/82694
13293 * common.opt (fstrict-overflow): No longer an alias.
13294 (fwrapv-pointer): New option.
13295 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
13296 also for pointer types based on flag_wrapv_pointer.
13297 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
13298 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
13299 opts->x_flag_wrapv got set.
13300 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
13301 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
13302 POINTER_TYPE_OVERFLOW_UNDEFINED.
13303 * match.pd: Likewise in address comparison pattern.
13304 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
13306 2018-01-15 Richard Biener <rguenther@suse.de>
13309 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
13310 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
13311 Reset type names to their identifier if their TYPE_DECL doesn't
13312 have linkage (and thus is used for ODR and devirt).
13313 (save_debug_info_for_decl): Remove.
13314 (save_debug_info_for_type): Likewise.
13315 (add_tree_to_fld_list): Adjust.
13316 * tree-pretty-print.c (dump_generic_node): Make dumping of
13317 type names more robust.
13319 2018-01-15 Richard Biener <rguenther@suse.de>
13321 * BASE-VER: Bump to 8.0.1.
13323 2018-01-14 Martin Sebor <msebor@redhat.com>
13326 * builtins.c (check_access): Avoid warning when the no-warning bit
13329 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
13331 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
13332 * ira-color (allocno_hard_regs_compare): Likewise.
13334 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
13337 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
13338 Use .pushsection/.popsection.
13340 2018-01-14 Martin Sebor <msebor@redhat.com>
13343 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
13345 2018-01-14 Jakub Jelinek <jakub@redhat.com>
13347 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
13348 entry from extra_headers.
13349 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
13350 extra_headers, make the list bitwise identical to the i?86-*-* one.
13352 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
13354 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
13355 -mcmodel=large with -mindirect-branch=thunk,
13356 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
13357 -mfunction-return=thunk-extern.
13358 * doc/invoke.texi: Document -mcmodel=large is incompatible with
13359 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
13360 -mfunction-return=thunk and -mfunction-return=thunk-extern.
13362 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
13364 * config/i386/i386.c (print_reg): Print the name of the full
13365 integer register without '%'.
13366 (ix86_print_operand): Handle 'V'.
13367 * doc/extend.texi: Document 'V' modifier.
13369 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
13371 * config/i386/constraints.md (Bs): Disallow memory operand for
13372 -mindirect-branch-register.
13374 * config/i386/predicates.md (indirect_branch_operand): Likewise.
13375 (GOT_memory_operand): Likewise.
13376 (call_insn_operand): Likewise.
13377 (sibcall_insn_operand): Likewise.
13378 (GOT32_symbol_operand): Likewise.
13379 * config/i386/i386.md (indirect_jump): Call convert_memory_address
13380 for -mindirect-branch-register.
13381 (tablejump): Likewise.
13382 (*sibcall_memory): Likewise.
13383 (*sibcall_value_memory): Likewise.
13384 Disallow peepholes of indirect call and jump via memory for
13385 -mindirect-branch-register.
13386 (*call_pop): Replace m with Bw.
13387 (*call_value_pop): Likewise.
13388 (*sibcall_pop_memory): Replace m with Bs.
13389 * config/i386/i386.opt (mindirect-branch-register): New option.
13390 * doc/invoke.texi: Document -mindirect-branch-register option.
13392 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
13394 * config/i386/i386-protos.h (ix86_output_function_return): New.
13395 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
13396 set function_return_type.
13397 (indirect_thunk_name): Add ret_p to indicate thunk for function
13399 (output_indirect_thunk_function): Pass false to
13400 indirect_thunk_name.
13401 (ix86_output_indirect_branch_via_reg): Likewise.
13402 (ix86_output_indirect_branch_via_push): Likewise.
13403 (output_indirect_thunk_function): Create alias for function
13404 return thunk if regno < 0.
13405 (ix86_output_function_return): New function.
13406 (ix86_handle_fndecl_attribute): Handle function_return.
13407 (ix86_attribute_table): Add function_return.
13408 * config/i386/i386.h (machine_function): Add
13409 function_return_type.
13410 * config/i386/i386.md (simple_return_internal): Use
13411 ix86_output_function_return.
13412 (simple_return_internal_long): Likewise.
13413 * config/i386/i386.opt (mfunction-return=): New option.
13414 (indirect_branch): Mention -mfunction-return=.
13415 * doc/extend.texi: Document function_return function attribute.
13416 * doc/invoke.texi: Document -mfunction-return= option.
13418 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
13420 * config/i386/i386-opts.h (indirect_branch): New.
13421 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
13422 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
13423 with local indirect jump when converting indirect call and jump.
13424 (ix86_set_indirect_branch_type): New.
13425 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
13426 (indirectlabelno): New.
13427 (indirect_thunk_needed): Likewise.
13428 (indirect_thunk_bnd_needed): Likewise.
13429 (indirect_thunks_used): Likewise.
13430 (indirect_thunks_bnd_used): Likewise.
13431 (INDIRECT_LABEL): Likewise.
13432 (indirect_thunk_name): Likewise.
13433 (output_indirect_thunk): Likewise.
13434 (output_indirect_thunk_function): Likewise.
13435 (ix86_output_indirect_branch_via_reg): Likewise.
13436 (ix86_output_indirect_branch_via_push): Likewise.
13437 (ix86_output_indirect_branch): Likewise.
13438 (ix86_output_indirect_jmp): Likewise.
13439 (ix86_code_end): Call output_indirect_thunk_function if needed.
13440 (ix86_output_call_insn): Call ix86_output_indirect_branch if
13442 (ix86_handle_fndecl_attribute): Handle indirect_branch.
13443 (ix86_attribute_table): Add indirect_branch.
13444 * config/i386/i386.h (machine_function): Add indirect_branch_type
13445 and has_local_indirect_jump.
13446 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
13448 (tablejump): Likewise.
13449 (*indirect_jump): Use ix86_output_indirect_jmp.
13450 (*tablejump_1): Likewise.
13451 (simple_return_indirect_internal): Likewise.
13452 * config/i386/i386.opt (mindirect-branch=): New option.
13453 (indirect_branch): New.
13456 (thunk-inline): Likewise.
13457 (thunk-extern): Likewise.
13458 * doc/extend.texi: Document indirect_branch function attribute.
13459 * doc/invoke.texi: Document -mindirect-branch= option.
13461 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
13464 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
13466 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
13468 * ipa-inline.c (want_inline_small_function_p): Return false if
13469 inlining has already failed with CIF_FINAL_ERROR.
13470 (update_caller_keys): Call want_inline_small_function_p before
13472 (update_callee_keys): Likewise.
13474 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
13476 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
13478 (rs6000_quadword_masked_address_p): Likewise.
13479 (quad_aligned_load_p): Likewise.
13480 (quad_aligned_store_p): Likewise.
13481 (const_load_sequence_p): Add comment to describe the outer-most loop.
13482 (mimic_memory_attributes_and_flags): New function.
13483 (rs6000_gen_stvx): Likewise.
13484 (replace_swapped_aligned_store): Likewise.
13485 (rs6000_gen_lvx): Likewise.
13486 (replace_swapped_aligned_load): Likewise.
13487 (replace_swapped_load_constant): Capitalize argument name in
13488 comment describing this function.
13489 (rs6000_analyze_swaps): Add a third pass to search for vector loads
13490 and stores that access quad-word aligned addresses and replace
13491 with stvx or lvx instructions when appropriate.
13492 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
13493 New function prototype.
13494 (rs6000_quadword_masked_address_p): Likewise.
13495 (rs6000_gen_lvx): Likewise.
13496 (rs6000_gen_stvx): Likewise.
13497 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
13498 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
13499 when memory address is aligned.
13500 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
13501 this split to select lvx instruction when memory address is aligned.
13502 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
13503 instruction when memory address is aligned.
13504 (*vsx_le_perm_load_v16qi): Likewise.
13505 (four unnamed splitters): Modify to select the stvx instruction
13506 when memory is aligned.
13508 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
13510 * predict.c (determine_unlikely_bbs): Handle correctly BBs
13511 which appears in the queue multiple times.
13513 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13514 Alan Hayward <alan.hayward@arm.com>
13515 David Sherwood <david.sherwood@arm.com>
13517 * tree-vectorizer.h (vec_lower_bound): New structure.
13518 (_loop_vec_info): Add check_nonzero and lower_bounds.
13519 (LOOP_VINFO_CHECK_NONZERO): New macro.
13520 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
13521 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
13522 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
13523 fields. Make seg_len the distance travelled, not including the
13525 (dr_direction_indicator): Declare.
13526 (dr_zero_step_indicator): Likewise.
13527 (dr_known_forward_stride_p): Likewise.
13528 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
13530 (runtime_alias_check_p): Allow runtime alias checks with
13532 (operator ==): Compare access_size and align.
13533 (prune_runtime_alias_test_list): Rework for new distinction between
13534 the access_size and seg_len.
13535 (create_intersect_range_checks_index): Likewise. Cope with polynomial
13537 (get_segment_min_max): New function.
13538 (create_intersect_range_checks): Use it.
13539 (dr_step_indicator): New function.
13540 (dr_direction_indicator): Likewise.
13541 (dr_zero_step_indicator): Likewise.
13542 (dr_known_forward_stride_p): Likewise.
13543 * tree-loop-distribution.c (data_ref_segment_size): Return
13544 DR_STEP * (niters - 1).
13545 (compute_alias_check_pairs): Update call to the dr_with_seg_len
13547 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
13548 (vect_preserves_scalar_order_p): New function, split out from...
13549 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
13550 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
13551 (vect_vfa_access_size): New function.
13552 (vect_vfa_align): Likewise.
13553 (vect_compile_time_alias): Take access_size_a and access_b arguments.
13554 (dump_lower_bound): New function.
13555 (vect_check_lower_bound): Likewise.
13556 (vect_small_gap_p): Likewise.
13557 (vectorizable_with_step_bound_p): Likewise.
13558 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
13559 depencies if the vectorization factor is 1. Convert the checks
13560 for nonzero steps into checks on the bounds of DR_STEP. Try using
13561 a bunds check for variable steps if the minimum required step is
13562 relatively small. Update calls to the dr_with_seg_len
13563 constructor and to vect_compile_time_alias.
13564 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
13566 (vect_loop_versioning): Call it.
13567 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
13569 (vect_estimate_min_profitable_iters): Account for any bounds checks.
13571 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13572 Alan Hayward <alan.hayward@arm.com>
13573 David Sherwood <david.sherwood@arm.com>
13575 * doc/sourcebuild.texi (vect_scatter_store): Document.
13576 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
13578 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
13580 * genopinit.c (main): Add supports_vec_scatter_store and
13581 supports_vec_scatter_store_cached to target_optabs.
13582 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
13583 IFN_MASK_SCATTER_STORE.
13584 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
13586 * internal-fn.h (internal_store_fn_p): Declare.
13587 (internal_fn_stored_value_index): Likewise.
13588 * internal-fn.c (scatter_store_direct): New macro.
13589 (expand_scatter_store_optab_fn): New function.
13590 (direct_scatter_store_optab_supported_p): New macro.
13591 (internal_store_fn_p): New function.
13592 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
13593 IFN_MASK_SCATTER_STORE.
13594 (internal_fn_mask_index): Likewise.
13595 (internal_fn_stored_value_index): New function.
13596 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
13597 for scatter stores.
13598 * optabs-query.h (supports_vec_scatter_store_p): Declare.
13599 * optabs-query.c (supports_vec_scatter_store_p): New function.
13600 * tree-vectorizer.h (vect_get_store_rhs): Declare.
13601 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
13602 true for scatter stores.
13603 (vect_gather_scatter_fn_p): Handle scatter stores too.
13604 (vect_check_gather_scatter): Consider using scatter stores if
13605 supports_vec_scatter_store_p.
13606 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
13607 scatter stores too.
13608 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
13609 internal_fn_stored_value_index.
13610 (check_load_store_masking): Handle scatter stores too.
13611 (vect_get_store_rhs): Make public.
13612 (vectorizable_call): Use internal_store_fn_p.
13613 (vectorizable_store): Handle scatter store internal functions.
13614 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
13615 when deciding whether the end of the group has been reached.
13616 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
13617 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
13618 (mask_scatter_store<mode>): New insns.
13620 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13621 Alan Hayward <alan.hayward@arm.com>
13622 David Sherwood <david.sherwood@arm.com>
13624 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
13625 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
13626 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
13628 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
13629 Use vect_truncate_gather_scatter_offset if we can't treat the
13630 operation as a normal gather load or scatter store.
13631 (get_group_load_store_type): Take the gather_scatter_info
13632 as argument. Try using a gather load or scatter store for
13633 single-element groups.
13634 (get_load_store_type): Update calls to get_group_load_store_type
13635 and vect_use_strided_gather_scatters_p.
13637 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13638 Alan Hayward <alan.hayward@arm.com>
13639 David Sherwood <david.sherwood@arm.com>
13641 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
13642 optional tree argument.
13643 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
13645 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
13646 but continue to use the current value as a fallback.
13647 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
13648 to compare the updates.
13649 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
13650 (get_load_store_type): Use it when handling a strided access.
13651 (vect_get_strided_load_store_ops): New function.
13652 (vect_get_data_ptr_increment): Likewise.
13653 (vectorizable_load): Handle strided gather loads. Always pass
13654 a step to vect_create_data_ref_ptr and bump_vector_ptr.
13656 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13657 Alan Hayward <alan.hayward@arm.com>
13658 David Sherwood <david.sherwood@arm.com>
13660 * doc/md.texi (gather_load@var{m}): Document.
13661 (mask_gather_load@var{m}): Likewise.
13662 * genopinit.c (main): Add supports_vec_gather_load and
13663 supports_vec_gather_load_cached to target_optabs.
13664 * optabs-tree.c (init_tree_optimization_optabs): Use
13665 ggc_cleared_alloc to allocate target_optabs.
13666 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
13667 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
13669 * internal-fn.h (internal_load_fn_p): Declare.
13670 (internal_gather_scatter_fn_p): Likewise.
13671 (internal_fn_mask_index): Likewise.
13672 (internal_gather_scatter_fn_supported_p): Likewise.
13673 * internal-fn.c (gather_load_direct): New macro.
13674 (expand_gather_load_optab_fn): New function.
13675 (direct_gather_load_optab_supported_p): New macro.
13676 (direct_internal_fn_optab): New function.
13677 (internal_load_fn_p): Likewise.
13678 (internal_gather_scatter_fn_p): Likewise.
13679 (internal_fn_mask_index): Likewise.
13680 (internal_gather_scatter_fn_supported_p): Likewise.
13681 * optabs-query.c (supports_at_least_one_mode_p): New function.
13682 (supports_vec_gather_load_p): Likewise.
13683 * optabs-query.h (supports_vec_gather_load_p): Declare.
13684 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
13685 and memory_type field.
13686 (NUM_PATTERNS): Bump to 15.
13687 * tree-vect-data-refs.c: Include internal-fn.h.
13688 (vect_gather_scatter_fn_p): New function.
13689 (vect_describe_gather_scatter_call): Likewise.
13690 (vect_check_gather_scatter): Try using internal functions for
13691 gather loads. Recognize existing calls to a gather load function.
13692 (vect_analyze_data_refs): Consider using gather loads if
13693 supports_vec_gather_load_p.
13694 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
13695 (vect_get_gather_scatter_offset_type): Likewise.
13696 (vect_convert_mask_for_vectype): Likewise.
13697 (vect_add_conversion_to_patterm): Likewise.
13698 (vect_try_gather_scatter_pattern): Likewise.
13699 (vect_recog_gather_scatter_pattern): New pattern recognizer.
13700 (vect_vect_recog_func_ptrs): Add it.
13701 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
13702 internal_fn_mask_index and internal_gather_scatter_fn_p.
13703 (check_load_store_masking): Take the gather_scatter_info as an
13704 argument and handle gather loads.
13705 (vect_get_gather_scatter_ops): New function.
13706 (vectorizable_call): Check internal_load_fn_p.
13707 (vectorizable_load): Likewise. Handle gather load internal
13709 (vectorizable_store): Update call to check_load_store_masking.
13710 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
13711 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
13712 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
13713 (aarch64_gather_scale_operand_d): New predicates.
13714 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
13715 (mask_gather_load<mode>): New insns.
13717 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13718 Alan Hayward <alan.hayward@arm.com>
13719 David Sherwood <david.sherwood@arm.com>
13721 * optabs.def (fold_left_plus_optab): New optab.
13722 * doc/md.texi (fold_left_plus_@var{m}): Document.
13723 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
13724 * internal-fn.c (fold_left_direct): Define.
13725 (expand_fold_left_optab_fn): Likewise.
13726 (direct_fold_left_optab_supported_p): Likewise.
13727 * fold-const-call.c (fold_const_fold_left): New function.
13728 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
13729 * tree-parloops.c (valid_reduction_p): New function.
13730 (gather_scalar_reductions): Use it.
13731 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
13732 (vect_finish_replace_stmt): Declare.
13733 * tree-vect-loop.c (fold_left_reduction_fn): New function.
13734 (needs_fold_left_reduction_p): New function, split out from...
13735 (vect_is_simple_reduction): ...here. Accept reductions that
13736 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
13737 (vect_force_simple_reduction): Also store the reduction type in
13738 the assignment's STMT_VINFO_REDUC_TYPE.
13739 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
13740 (merge_with_identity): New function.
13741 (vect_expand_fold_left): Likewise.
13742 (vectorize_fold_left_reduction): Likewise.
13743 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
13744 scalar phi in place for it. Check for target support and reject
13745 cases that would reassociate the operation. Defer the transform
13746 phase to vectorize_fold_left_reduction.
13747 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
13748 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
13749 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
13751 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13753 * tree-if-conv.c (predicate_mem_writes): Remove redundant
13754 call to ifc_temp_var.
13756 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13757 Alan Hayward <alan.hayward@arm.com>
13758 David Sherwood <david.sherwood@arm.com>
13760 * target.def (legitimize_address_displacement): Take the original
13761 offset as a poly_int.
13762 * targhooks.h (default_legitimize_address_displacement): Update
13764 * targhooks.c (default_legitimize_address_displacement): Likewise.
13765 * doc/tm.texi: Regenerate.
13766 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
13767 as an argument, moving assert of ad->disp == ad->disp_term to...
13768 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
13769 Try calling targetm.legitimize_address_displacement before expanding
13770 the address rather than afterwards, and adjust for the new interface.
13771 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
13772 Match the new hook interface. Handle SVE addresses.
13773 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
13774 new hook interface.
13776 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13778 * Makefile.in (OBJS): Add early-remat.o.
13779 * target.def (select_early_remat_modes): New hook.
13780 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
13781 * doc/tm.texi: Regenerate.
13782 * targhooks.h (default_select_early_remat_modes): Declare.
13783 * targhooks.c (default_select_early_remat_modes): New function.
13784 * timevar.def (TV_EARLY_REMAT): New timevar.
13785 * passes.def (pass_early_remat): New pass.
13786 * tree-pass.h (make_pass_early_remat): Declare.
13787 * early-remat.c: New file.
13788 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
13790 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
13792 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13793 Alan Hayward <alan.hayward@arm.com>
13794 David Sherwood <david.sherwood@arm.com>
13796 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
13797 vfm1 with a bound_epilog parameter.
13798 (vect_do_peeling): Update calls accordingly, and move the prologue
13799 call earlier in the function. Treat the base bound_epilog as 0 for
13800 fully-masked loops and retain vf - 1 for other loops. Add 1 to
13801 this base when peeling for gaps.
13802 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
13803 with fully-masked loops.
13804 (vect_estimate_min_profitable_iters): Handle the single peeled
13805 iteration in that case.
13807 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13808 Alan Hayward <alan.hayward@arm.com>
13809 David Sherwood <david.sherwood@arm.com>
13811 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
13812 single-element interleaving even if the size is not a power of 2.
13813 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
13814 accesses for single-element interleaving if the group size is
13817 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13818 Alan Hayward <alan.hayward@arm.com>
13819 David Sherwood <david.sherwood@arm.com>
13821 * doc/md.texi (fold_extract_last_@var{m}): Document.
13822 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
13823 * optabs.def (fold_extract_last_optab): New optab.
13824 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
13825 * internal-fn.c (fold_extract_direct): New macro.
13826 (expand_fold_extract_optab_fn): Likewise.
13827 (direct_fold_extract_optab_supported_p): Likewise.
13828 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
13829 * tree-vect-loop.c (vect_model_reduction_cost): Handle
13830 EXTRACT_LAST_REDUCTION.
13831 (get_initial_def_for_reduction): Do not create an initial vector
13832 for EXTRACT_LAST_REDUCTION reductions.
13833 (vectorizable_reduction): Leave the scalar phi in place for
13834 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
13835 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
13836 epilogue code for EXTRACT_LAST_REDUCTION and defer the
13837 transform phase to vectorizable_condition.
13838 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
13840 (vect_finish_stmt_generation): ...here.
13841 (vect_finish_replace_stmt): New function.
13842 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
13843 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
13845 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
13847 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13848 Alan Hayward <alan.hayward@arm.com>
13849 David Sherwood <david.sherwood@arm.com>
13851 * doc/md.texi (extract_last_@var{m}): Document.
13852 * optabs.def (extract_last_optab): New optab.
13853 * internal-fn.def (EXTRACT_LAST): New internal function.
13854 * internal-fn.c (cond_unary_direct): New macro.
13855 (expand_cond_unary_optab_fn): Likewise.
13856 (direct_cond_unary_optab_supported_p): Likewise.
13857 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
13858 loops using EXTRACT_LAST.
13859 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
13860 (extract_last_<mode>): ...this optab.
13861 (vec_extract<mode><Vel>): Update accordingly.
13863 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13864 Alan Hayward <alan.hayward@arm.com>
13865 David Sherwood <david.sherwood@arm.com>
13867 * target.def (empty_mask_is_expensive): New hook.
13868 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
13869 * doc/tm.texi: Regenerate.
13870 * targhooks.h (default_empty_mask_is_expensive): Declare.
13871 * targhooks.c (default_empty_mask_is_expensive): New function.
13872 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
13873 if the target says that empty masks are expensive.
13874 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
13876 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
13878 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13879 Alan Hayward <alan.hayward@arm.com>
13880 David Sherwood <david.sherwood@arm.com>
13882 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
13883 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
13884 (vect_use_loop_mask_for_alignment_p): New function.
13885 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
13886 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
13887 niters_skip argument. Make sure that the first niters_skip elements
13888 of the first iteration are inactive.
13889 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
13890 Update call to vect_set_loop_masks_directly.
13891 (get_misalign_in_elems): New function, split out from...
13892 (vect_gen_prolog_loop_niters): ...here.
13893 (vect_update_init_of_dr): Take a code argument that specifies whether
13894 the adjustment should be added or subtracted.
13895 (vect_update_init_of_drs): Likewise.
13896 (vect_prepare_for_masked_peels): New function.
13897 (vect_do_peeling): Skip prologue peeling if we're using a mask
13898 instead. Update call to vect_update_inits_of_drs.
13899 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
13901 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
13902 alignment. Do not include the number of peeled iterations in
13903 the minimum threshold in that case.
13904 (vectorizable_induction): Adjust the start value down by
13905 LOOP_VINFO_MASK_SKIP_NITERS iterations.
13906 (vect_transform_loop): Call vect_prepare_for_masked_peels.
13907 Take the number of skipped iterations into account when calculating
13909 * tree-vect-stmts.c (vect_gen_while_not): New function.
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 * doc/sourcebuild.texi (vect_fully_masked): Document.
13916 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
13917 default value to 0.
13918 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
13920 (vect_analyze_loop_2): ...here. Don't check the vectorization
13921 factor against the number of loop iterations if the loop is
13924 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13925 Alan Hayward <alan.hayward@arm.com>
13926 David Sherwood <david.sherwood@arm.com>
13928 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
13929 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
13930 (dump_groups): Update accordingly.
13931 (iv_use::mem_type): New member variable.
13932 (address_p): New function.
13933 (record_use): Add a mem_type argument and initialize the new
13935 (record_group_use): Add a mem_type argument. Use address_p.
13936 Remove obsolete null checks of base_object. Update call to record_use.
13937 (find_interesting_uses_op): Update call to record_group_use.
13938 (find_interesting_uses_cond): Likewise.
13939 (find_interesting_uses_address): Likewise.
13940 (get_mem_type_for_internal_fn): New function.
13941 (find_address_like_use): Likewise.
13942 (find_interesting_uses_stmt): Try find_address_like_use before
13943 calling find_interesting_uses_op.
13944 (addr_offset_valid_p): Use the iv mem_type field as the type
13945 of the addressed memory.
13946 (add_autoinc_candidates): Likewise.
13947 (get_address_cost): Likewise.
13948 (split_small_address_groups_p): Use address_p.
13949 (split_address_groups): Likewise.
13950 (add_iv_candidate_for_use): Likewise.
13951 (autoinc_possible_for_pair): Likewise.
13952 (rewrite_groups): Likewise.
13953 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
13954 (determine_group_iv_cost): Update after split of USE_ADDRESS.
13955 (get_alias_ptr_type_for_ptr_address): New function.
13956 (rewrite_use_address): Rewrite address uses in calls that were
13957 identified by find_address_like_use.
13959 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13960 Alan Hayward <alan.hayward@arm.com>
13961 David Sherwood <david.sherwood@arm.com>
13963 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
13965 * gimple-expr.h (is_gimple_addressable: Likewise.
13966 * gimple-expr.c (is_gimple_address): Likewise.
13967 * internal-fn.c (expand_call_mem_ref): New function.
13968 (expand_mask_load_optab_fn): Use it.
13969 (expand_mask_store_optab_fn): Likewise.
13971 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13972 Alan Hayward <alan.hayward@arm.com>
13973 David Sherwood <david.sherwood@arm.com>
13975 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
13976 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
13977 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
13978 (cond_umax@var{mode}): Document.
13979 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
13980 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
13981 (cond_umin_optab, cond_umax_optab): New optabs.
13982 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
13983 (COND_IOR, COND_XOR): New internal functions.
13984 * internal-fn.h (get_conditional_internal_fn): Declare.
13985 * internal-fn.c (cond_binary_direct): New macro.
13986 (expand_cond_binary_optab_fn): Likewise.
13987 (direct_cond_binary_optab_supported_p): Likewise.
13988 (get_conditional_internal_fn): New function.
13989 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
13990 Cope with reduction statements that are vectorized as calls rather
13992 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
13993 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
13994 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
13995 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
13996 (UNSPEC_COND_EOR): New unspecs.
13997 (optab): Add mappings for them.
13998 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
13999 (sve_int_op, sve_fp_op): New int attributes.
14001 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14002 Alan Hayward <alan.hayward@arm.com>
14003 David Sherwood <david.sherwood@arm.com>
14005 * optabs.def (while_ult_optab): New optab.
14006 * doc/md.texi (while_ult@var{m}@var{n}): Document.
14007 * internal-fn.def (WHILE_ULT): New internal function.
14008 * internal-fn.h (direct_internal_fn_supported_p): New override
14009 that takes two types as argument.
14010 * internal-fn.c (while_direct): New macro.
14011 (expand_while_optab_fn): New function.
14012 (convert_optab_supported_p): Likewise.
14013 (direct_while_optab_supported_p): New macro.
14014 * wide-int.h (wi::udiv_ceil): New function.
14015 * tree-vectorizer.h (rgroup_masks): New structure.
14016 (vec_loop_masks): New typedef.
14017 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
14018 and fully_masked_p.
14019 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
14020 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
14021 (vect_max_vf): New function.
14022 (slpeel_make_loop_iterate_ntimes): Delete.
14023 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
14024 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
14025 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
14026 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
14027 internal-fn.h, stor-layout.h and optabs-query.h.
14028 (vect_set_loop_mask): New function.
14029 (add_preheader_seq): Likewise.
14030 (add_header_seq): Likewise.
14031 (interleave_supported_p): Likewise.
14032 (vect_maybe_permute_loop_masks): Likewise.
14033 (vect_set_loop_masks_directly): Likewise.
14034 (vect_set_loop_condition_masked): Likewise.
14035 (vect_set_loop_condition_unmasked): New function, split out from
14036 slpeel_make_loop_iterate_ntimes.
14037 (slpeel_make_loop_iterate_ntimes): Rename to..
14038 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
14039 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
14040 (vect_do_peeling): Update call accordingly.
14041 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
14043 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
14044 mask_compare_type, can_fully_mask_p and fully_masked_p.
14045 (release_vec_loop_masks): New function.
14046 (_loop_vec_info): Use it to free the loop masks.
14047 (can_produce_all_loop_masks_p): New function.
14048 (vect_get_max_nscalars_per_iter): Likewise.
14049 (vect_verify_full_masking): Likewise.
14050 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
14051 retries, and free the mask rgroups before retrying. Check loop-wide
14052 reasons for disallowing fully-masked loops. Make the final decision
14053 about whether use a fully-masked loop or not.
14054 (vect_estimate_min_profitable_iters): Do not assume that peeling
14055 for the number of iterations will be needed for fully-masked loops.
14056 (vectorizable_reduction): Disable fully-masked loops.
14057 (vectorizable_live_operation): Likewise.
14058 (vect_halve_mask_nunits): New function.
14059 (vect_double_mask_nunits): Likewise.
14060 (vect_record_loop_mask): Likewise.
14061 (vect_get_loop_mask): Likewise.
14062 (vect_transform_loop): Handle the case in which the final loop
14063 iteration might handle a partial vector. Call vect_set_loop_condition
14064 instead of slpeel_make_loop_iterate_ntimes.
14065 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
14066 (check_load_store_masking): New function.
14067 (prepare_load_store_mask): Likewise.
14068 (vectorizable_store): Handle fully-masked loops.
14069 (vectorizable_load): Likewise.
14070 (supportable_widening_operation): Use vect_halve_mask_nunits for
14072 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
14073 (vect_gen_while): New function.
14074 * config/aarch64/aarch64.md (umax<mode>3): New expander.
14075 (aarch64_uqdec<mode>): New insn.
14077 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14078 Alan Hayward <alan.hayward@arm.com>
14079 David Sherwood <david.sherwood@arm.com>
14081 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
14082 (reduc_xor_scal_optab): New optabs.
14083 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
14084 (reduc_xor_scal_@var{m}): Document.
14085 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
14086 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
14087 internal functions.
14088 * fold-const-call.c (fold_const_call): Handle them.
14089 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
14090 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
14091 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
14092 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
14093 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
14094 (UNSPEC_XORV): New unspecs.
14095 (optab): Add entries for them.
14096 (BITWISEV): New int iterator.
14097 (bit_reduc_op): New int attributes.
14099 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14100 Alan Hayward <alan.hayward@arm.com>
14101 David Sherwood <david.sherwood@arm.com>
14103 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
14104 * internal-fn.def (VEC_SHL_INSERT): New internal function.
14105 * optabs.def (vec_shl_insert_optab): New optab.
14106 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
14107 (duplicate_and_interleave): Likewise.
14108 * tree-vect-loop.c: Include internal-fn.h.
14109 (neutral_op_for_slp_reduction): New function, split out from
14110 get_initial_defs_for_reduction.
14111 (get_initial_def_for_reduction): Handle option 2 for variable-length
14112 vectors by loading the neutral value into a vector and then shifting
14113 the initial value into element 0.
14114 (get_initial_defs_for_reduction): Replace the code argument with
14115 the neutral value calculated by neutral_op_for_slp_reduction.
14116 Use gimple_build_vector for constant-length vectors.
14117 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
14118 but the first group_size elements have a neutral value.
14119 Use duplicate_and_interleave otherwise.
14120 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
14121 Update call to get_initial_defs_for_reduction. Handle SLP
14122 reductions for variable-length vectors by creating one vector
14123 result for each scalar result, with the elements associated
14124 with other scalar results stubbed out with the neutral value.
14125 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
14126 Require IFN_VEC_SHL_INSERT for double reductions on
14127 variable-length vectors, or SLP reductions that have
14128 a neutral value. Require can_duplicate_and_interleave_p
14129 support for variable-length unchained SLP reductions if there
14130 is no neutral value, such as for MIN/MAX reductions. Also require
14131 the number of vector elements to be a multiple of the number of
14132 SLP statements when doing variable-length unchained SLP reductions.
14133 Update call to vect_create_epilog_for_reduction.
14134 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
14135 and remove initial values.
14136 (duplicate_and_interleave): Make public.
14137 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
14138 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
14140 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14141 Alan Hayward <alan.hayward@arm.com>
14142 David Sherwood <david.sherwood@arm.com>
14144 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
14145 (can_duplicate_and_interleave_p): New function.
14146 (vect_get_and_check_slp_defs): Take the vector of statements
14147 rather than just the current one. Remove excess parentheses.
14148 Restriction rejectinon of vect_constant_def and vect_external_def
14149 for variable-length vectors to boolean types, or types for which
14150 can_duplicate_and_interleave_p is false.
14151 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
14152 (duplicate_and_interleave): New function.
14153 (vect_get_constant_vectors): Use gimple_build_vector for
14154 constant-length vectors and suitable variable-length constant
14155 vectors. Use duplicate_and_interleave for other variable-length
14156 vectors. Don't defer the update when inserting new statements.
14158 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14159 Alan Hayward <alan.hayward@arm.com>
14160 David Sherwood <david.sherwood@arm.com>
14162 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
14163 min_profitable_iters doesn't go negative.
14165 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14166 Alan Hayward <alan.hayward@arm.com>
14167 David Sherwood <david.sherwood@arm.com>
14169 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
14170 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
14171 * optabs.def (vec_mask_load_lanes_optab): New optab.
14172 (vec_mask_store_lanes_optab): Likewise.
14173 * internal-fn.def (MASK_LOAD_LANES): New internal function.
14174 (MASK_STORE_LANES): Likewise.
14175 * internal-fn.c (mask_load_lanes_direct): New macro.
14176 (mask_store_lanes_direct): Likewise.
14177 (expand_mask_load_optab_fn): Handle masked operations.
14178 (expand_mask_load_lanes_optab_fn): New macro.
14179 (expand_mask_store_optab_fn): Handle masked operations.
14180 (expand_mask_store_lanes_optab_fn): New macro.
14181 (direct_mask_load_lanes_optab_supported_p): Likewise.
14182 (direct_mask_store_lanes_optab_supported_p): Likewise.
14183 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
14185 (vect_load_lanes_supported): Likewise.
14186 * tree-vect-data-refs.c (strip_conversion): New function.
14187 (can_group_stmts_p): Likewise.
14188 (vect_analyze_data_ref_accesses): Use it instead of checking
14189 for a pair of assignments.
14190 (vect_store_lanes_supported): Take a masked_p parameter.
14191 (vect_load_lanes_supported): Likewise.
14192 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
14193 vect_store_lanes_supported and vect_load_lanes_supported.
14194 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
14195 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
14196 parameter. Don't allow gaps for masked accesses.
14197 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
14198 and vect_load_lanes_supported.
14199 (get_load_store_type): Take a masked_p parameter and update
14200 call to get_group_load_store_type.
14201 (vectorizable_store): Update call to get_load_store_type.
14202 Handle IFN_MASK_STORE_LANES.
14203 (vectorizable_load): Update call to get_load_store_type.
14204 Handle IFN_MASK_LOAD_LANES.
14206 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14207 Alan Hayward <alan.hayward@arm.com>
14208 David Sherwood <david.sherwood@arm.com>
14210 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
14212 * config/aarch64/aarch64-protos.h
14213 (aarch64_sve_struct_memory_operand_p): Declare.
14214 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
14215 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
14216 (VPRED, vpred): Handle SVE structure modes.
14217 * config/aarch64/constraints.md (Utx): New constraint.
14218 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
14219 (aarch64_sve_struct_nonimmediate_operand): New predicates.
14220 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
14221 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
14222 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
14223 structure modes. Split into pieces after RA.
14224 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
14225 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
14227 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
14228 SVE structure modes.
14229 (aarch64_classify_address): Likewise.
14230 (sizetochar): Move earlier in file.
14231 (aarch64_print_operand): Handle SVE register lists.
14232 (aarch64_array_mode): New function.
14233 (aarch64_sve_struct_memory_operand_p): Likewise.
14234 (TARGET_ARRAY_MODE): Redefine.
14236 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14237 Alan Hayward <alan.hayward@arm.com>
14238 David Sherwood <david.sherwood@arm.com>
14240 * target.def (array_mode): New target hook.
14241 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
14242 * doc/tm.texi: Regenerate.
14243 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
14244 * hooks.c (hook_optmode_mode_uhwi_none): New function.
14245 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
14246 targetm.array_mode.
14247 * stor-layout.c (mode_for_array): Likewise. Support polynomial
14250 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14251 Alan Hayward <alan.hayward@arm.com>
14252 David Sherwood <david.sherwood@arm.com>
14254 * fold-const.c (fold_binary_loc): Check the argument types
14255 rather than the result type when testing for a vector operation.
14257 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14259 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
14260 * doc/tm.texi: Regenerate.
14262 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14263 Alan Hayward <alan.hayward@arm.com>
14264 David Sherwood <david.sherwood@arm.com>
14266 * doc/invoke.texi (-msve-vector-bits=): Document new option.
14267 (sve): Document new AArch64 extension.
14268 * doc/md.texi (w): Extend the description of the AArch64
14269 constraint to include SVE vectors.
14270 (Upl, Upa): Document new AArch64 predicate constraints.
14271 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
14273 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
14274 (msve-vector-bits=): New option.
14275 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
14276 SVE when these are disabled.
14277 (sve): New extension.
14278 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
14279 modes. Adjust their number of units based on aarch64_sve_vg.
14280 (MAX_BITSIZE_MODE_ANY_MODE): Define.
14281 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
14282 aarch64_addr_query_type.
14283 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
14284 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
14285 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
14286 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
14287 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
14288 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
14289 (aarch64_simd_imm_zero_p): Delete.
14290 (aarch64_check_zero_based_sve_index_immediate): Declare.
14291 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
14292 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
14293 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
14294 (aarch64_sve_float_mul_immediate_p): Likewise.
14295 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
14296 rather than an rtx.
14297 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
14298 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
14299 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
14300 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
14301 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
14302 (aarch64_regmode_natural_size): Likewise.
14303 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
14304 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
14306 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
14307 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
14308 for VG and the SVE predicate registers.
14309 (V_ALIASES): Add a "z"-prefixed alias.
14310 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
14311 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
14312 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
14313 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
14314 (REG_CLASS_NAMES): Add entries for them.
14315 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
14316 and the predicate registers.
14317 (aarch64_sve_vg): Declare.
14318 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
14319 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
14320 (REGMODE_NATURAL_SIZE): Define.
14321 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
14323 * config/aarch64/aarch64.c: Include cfgrtl.h.
14324 (simd_immediate_info): Add a constructor for series vectors,
14325 and an associated step field.
14326 (aarch64_sve_vg): New variable.
14327 (aarch64_dbx_register_number): Handle VG and the predicate registers.
14328 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
14329 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
14330 (VEC_ANY_DATA, VEC_STRUCT): New constants.
14331 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
14332 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
14333 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
14334 (aarch64_get_mask_mode): New functions.
14335 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
14336 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
14337 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
14338 predicate modes and predicate registers. Explicitly restrict
14339 GPRs to modes of 16 bytes or smaller. Only allow FP registers
14340 to store a vector mode if it is recognized by
14341 aarch64_classify_vector_mode.
14342 (aarch64_regmode_natural_size): New function.
14343 (aarch64_hard_regno_caller_save_mode): Return the original mode
14345 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
14346 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
14347 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
14348 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
14350 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
14351 does not overlap dest if the function is frame-related. Handle
14353 (aarch64_split_add_offset): New function.
14354 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
14355 them aarch64_add_offset.
14356 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
14357 and update call to aarch64_sub_sp.
14358 (aarch64_add_cfa_expression): New function.
14359 (aarch64_expand_prologue): Pass extra temporary registers to the
14360 functions above. Handle the case in which we need to emit new
14361 DW_CFA_expressions for registers that were originally saved
14362 relative to the stack pointer, but now have to be expressed
14363 relative to the frame pointer.
14364 (aarch64_output_mi_thunk): Pass extra temporary registers to the
14366 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
14367 IP0 and IP1 values for SVE frames.
14368 (aarch64_expand_vec_series): New function.
14369 (aarch64_expand_sve_widened_duplicate): Likewise.
14370 (aarch64_expand_sve_const_vector): Likewise.
14371 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
14372 Handle SVE constants. Use emit_move_insn to move a force_const_mem
14373 into the register, rather than emitting a SET directly.
14374 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
14375 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
14376 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
14377 (offset_9bit_signed_scaled_p): New functions.
14378 (aarch64_replicate_bitmask_imm): New function.
14379 (aarch64_bitmask_imm): Use it.
14380 (aarch64_cannot_force_const_mem): Reject expressions involving
14381 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
14382 (aarch64_classify_index): Handle SVE indices, by requiring
14383 a plain register index with a scale that matches the element size.
14384 (aarch64_classify_address): Handle SVE addresses. Assert that
14385 the mode of the address is VOIDmode or an integer mode.
14386 Update call to aarch64_classify_symbol.
14387 (aarch64_classify_symbolic_expression): Update call to
14388 aarch64_classify_symbol.
14389 (aarch64_const_vec_all_in_range_p): New function.
14390 (aarch64_print_vector_float_operand): Likewise.
14391 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
14392 "vN" for FP registers with SVE modes. Handle (const ...) vectors
14393 and the FP immediates 1.0 and 0.5.
14394 (aarch64_print_address_internal): Handle SVE addresses.
14395 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
14396 (aarch64_regno_regclass): Handle predicate registers.
14397 (aarch64_secondary_reload): Handle big-endian reloads of SVE
14399 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
14400 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
14401 (aarch64_convert_sve_vector_bits): New function.
14402 (aarch64_override_options): Use it to handle -msve-vector-bits=.
14403 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
14404 rather than an rtx.
14405 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
14406 Handle SVE vector and predicate modes. Accept VL-based constants
14407 that need only one temporary register, and VL offsets that require
14408 no temporary registers.
14409 (aarch64_conditional_register_usage): Mark the predicate registers
14410 as fixed if SVE isn't available.
14411 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
14412 Return true for SVE vector and predicate modes.
14413 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
14414 rather than an unsigned int. Handle SVE modes.
14415 (aarch64_preferred_simd_mode): Update call accordingly. Handle
14417 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
14419 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
14420 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
14421 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
14422 (aarch64_sve_float_mul_immediate_p): New functions.
14423 (aarch64_sve_valid_immediate): New function.
14424 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
14425 Explicitly reject structure modes. Check for INDEX constants.
14426 Handle PTRUE and PFALSE constants.
14427 (aarch64_check_zero_based_sve_index_immediate): New function.
14428 (aarch64_simd_imm_zero_p): Delete.
14429 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
14430 vector modes. Accept constants in the range of CNT[BHWD].
14431 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
14432 ask for an Advanced SIMD mode.
14433 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
14434 (aarch64_simd_vector_alignment): Handle SVE predicates.
14435 (aarch64_vectorize_preferred_vector_alignment): New function.
14436 (aarch64_simd_vector_alignment_reachable): Use it instead of
14438 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
14439 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
14441 (MAX_VECT_LEN): Delete.
14442 (expand_vec_perm_d): Add a vec_flags field.
14443 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
14444 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
14445 (aarch64_evpc_ext): Don't apply a big-endian lane correction
14447 (aarch64_evpc_rev): Rename to...
14448 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
14449 (aarch64_evpc_rev_global): New function.
14450 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
14451 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
14453 (aarch64_evpc_sve_tbl): New function.
14454 (aarch64_expand_vec_perm_const_1): Update after rename of
14455 aarch64_evpc_rev. Handle SVE permutes too, trying
14456 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
14457 than aarch64_evpc_tbl.
14458 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
14459 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
14460 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
14461 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
14462 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
14463 (aarch64_expand_sve_vcond): New functions.
14464 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
14465 of aarch64_vector_mode_p.
14466 (aarch64_dwarf_poly_indeterminate_value): New function.
14467 (aarch64_compute_pressure_classes): Likewise.
14468 (aarch64_can_change_mode_class): Likewise.
14469 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
14470 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
14471 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
14472 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
14473 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
14474 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
14475 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
14476 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
14478 (Dn, Dl, Dr): Accept const as well as const_vector.
14479 (Dz): Likewise. Compare against CONST0_RTX.
14480 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
14481 of "vector" where appropriate.
14482 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
14483 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
14484 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
14485 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
14486 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
14487 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
14488 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
14489 (v_int_equiv): Extend to SVE modes.
14490 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
14492 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
14493 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
14494 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
14495 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
14496 (SVE_COND_FP_CMP): New int iterators.
14497 (perm_hilo): Handle the new unpack unspecs.
14498 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
14500 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
14501 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
14502 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
14503 (aarch64_equality_operator, aarch64_constant_vector_operand)
14504 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
14505 (aarch64_sve_nonimmediate_operand): Likewise.
14506 (aarch64_sve_general_operand): Likewise.
14507 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
14508 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
14509 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
14510 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
14511 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
14512 (aarch64_sve_float_arith_immediate): Likewise.
14513 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
14514 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
14515 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
14516 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
14517 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
14518 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
14519 (aarch64_sve_float_arith_operand): Likewise.
14520 (aarch64_sve_float_arith_with_sub_operand): Likewise.
14521 (aarch64_sve_float_mul_operand): Likewise.
14522 (aarch64_sve_vec_perm_operand): Likewise.
14523 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
14524 (aarch64_mov_operand): Accept const_poly_int and const_vector.
14525 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
14526 as well as const_vector.
14527 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
14528 in file. Use CONST0_RTX and CONSTM1_RTX.
14529 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
14530 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
14531 Use aarch64_simd_imm_zero.
14532 * config/aarch64/aarch64-sve.md: New file.
14533 * config/aarch64/aarch64.md: Include it.
14534 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
14535 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
14536 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
14537 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
14538 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
14539 (sve): New attribute.
14540 (enabled): Disable instructions with the sve attribute unless
14542 (movqi, movhi): Pass CONST_POLY_INT operaneds through
14543 aarch64_expand_mov_immediate.
14544 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
14545 CNT[BHSD] immediates.
14546 (movti): Split CONST_POLY_INT moves into two halves.
14547 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
14548 Split additions that need a temporary here if the destination
14549 is the stack pointer.
14550 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
14551 (*add<mode>3_poly_1): New instruction.
14552 (set_clobber_cc): New expander.
14554 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14556 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
14557 parameter and use it instead of GET_MODE_SIZE (innermode). Use
14558 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
14559 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
14560 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
14561 Change innermode from fixed_mode_size to machine_mode.
14562 (simplify_subreg): Update call accordingly. Handle a constant-sized
14563 subreg of a variable-length CONST_VECTOR.
14565 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14566 Alan Hayward <alan.hayward@arm.com>
14567 David Sherwood <david.sherwood@arm.com>
14569 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
14570 (add_offset_to_base): New function, split out from...
14571 (create_mem_ref): ...here. When handling a scale other than 1,
14572 check first whether the address is valid without the offset.
14573 Add it into the base if so, leaving the index and scale as-is.
14575 2018-01-12 Jakub Jelinek <jakub@redhat.com>
14578 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
14579 fold_for_warn before checking if arg2 is INTEGER_CST.
14581 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
14583 * config/rs6000/predicates.md (load_multiple_operation): Delete.
14584 (store_multiple_operation): Delete.
14585 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
14586 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
14587 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
14588 guarded by TARGET_STRING.
14589 (rs6000_output_load_multiple): Delete.
14590 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
14591 OPTION_MASK_STRING / TARGET_STRING handling.
14592 (print_operand) <'N', 'O'>: Add comment that these are unused now.
14593 (const rs6000_opt_masks) <"string">: Change mask to 0.
14594 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
14595 (MASK_STRING): Delete.
14596 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
14598 (load_multiple): Delete.
14605 (store_multiple): Delete.
14612 (movmemsi_8reg): Delete.
14613 (corresponding unnamed define_insn): Delete.
14614 (movmemsi_6reg): Delete.
14615 (corresponding unnamed define_insn): Delete.
14616 (movmemsi_4reg): Delete.
14617 (corresponding unnamed define_insn): Delete.
14618 (movmemsi_2reg): Delete.
14619 (corresponding unnamed define_insn): Delete.
14620 (movmemsi_1reg): Delete.
14621 (corresponding unnamed define_insn): Delete.
14622 * config/rs6000/rs6000.opt (mno-string): New.
14623 (mstring): Replace by deprecation warning stub.
14624 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
14626 2018-01-12 Jakub Jelinek <jakub@redhat.com>
14628 * regrename.c (regrename_do_replace): If replacing the same
14629 reg multiple times, try to reuse last created gen_raw_REG.
14632 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
14633 main to workaround a bug in GDB.
14635 2018-01-12 Tom de Vries <tom@codesourcery.com>
14638 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
14640 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
14642 PR rtl-optimization/80481
14643 * ira-color.c (get_cap_member): New function.
14644 (allocnos_conflict_by_live_ranges_p): Use it.
14645 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
14646 (setup_slot_coalesced_allocno_live_ranges): Ditto.
14648 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
14651 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
14652 (*saddl_se_1): Ditto.
14653 (*ssubsi_1): Ditto.
14654 (*ssubl_se_1): Ditto.
14656 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
14658 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
14659 rather than wi::to_widest for DR_INITs.
14660 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
14661 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
14662 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
14664 (vect_analyze_group_access_1): Note that here.
14666 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
14668 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
14669 polynomial type sizes.
14671 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
14673 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
14674 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
14675 (gimple_add_tmp_var): Likewise.
14677 2018-01-12 Martin Liska <mliska@suse.cz>
14679 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
14680 (gimple_alloc_sizes): Likewise.
14681 (dump_gimple_statistics): Use PRIu64 in printf format.
14682 * gimple.h: Change uint64_t to int.
14684 2018-01-12 Martin Liska <mliska@suse.cz>
14686 * tree-core.h: Use uint64_t instead of int.
14687 * tree.c (tree_node_counts): Likewise.
14688 (tree_node_sizes): Likewise.
14689 (dump_tree_statistics): Use PRIu64 in printf format.
14691 2018-01-12 Martin Liska <mliska@suse.cz>
14693 * Makefile.in: As qsort_chk is implemented in vec.c, add
14694 vec.o to linkage of gencfn-macros.
14695 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
14696 passing the info to record_node_allocation_statistics.
14697 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
14699 * ggc-common.c (struct ggc_usage): Add operator== and use
14700 it in operator< and compare function.
14701 * mem-stats.h (struct mem_usage): Likewise.
14702 * vec.c (struct vec_usage): Remove operator< and compare
14703 function. Can be simply inherited.
14705 2018-01-12 Martin Jambor <mjambor@suse.cz>
14708 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
14709 * tree-ssa-math-opts.c: Include domwalk.h.
14710 (convert_mult_to_fma_1): New function.
14711 (fma_transformation_info): New type.
14712 (fma_deferring_state): Likewise.
14713 (cancel_fma_deferring): New function.
14714 (result_of_phi): Likewise.
14715 (last_fma_candidate_feeds_initial_phi): Likewise.
14716 (convert_mult_to_fma): Added deferring logic, split actual
14717 transformation to convert_mult_to_fma_1.
14718 (math_opts_dom_walker): New type.
14719 (math_opts_dom_walker::after_dom_children): New method, body moved
14720 here from pass_optimize_widening_mul::execute, added deferring logic
14722 (pass_optimize_widening_mul::execute): Moved most of code to
14723 math_opts_dom_walker::after_dom_children.
14724 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
14725 * config/i386/i386.c (ix86_option_override_internal): Added
14726 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
14728 2018-01-12 Richard Biener <rguenther@suse.de>
14731 * dwarf2out.c (gen_variable_die): Do not reset old_die for
14732 inline instance vars.
14734 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
14737 * config/rx/rx.c (rx_is_restricted_memory_address):
14738 Handle SUBREG case.
14740 2018-01-12 Richard Biener <rguenther@suse.de>
14742 PR tree-optimization/80846
14743 * target.def (split_reduction): New target hook.
14744 * targhooks.c (default_split_reduction): New function.
14745 * targhooks.h (default_split_reduction): Declare.
14746 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
14747 target requests first reduce vectors by combining low and high
14749 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
14750 (get_vectype_for_scalar_type_and_size): Export.
14751 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
14752 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
14753 * doc/tm.texi: Regenerate.
14754 * config/i386/i386.c (ix86_split_reduction): Implement
14755 TARGET_VECTORIZE_SPLIT_REDUCTION.
14757 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
14760 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
14761 in PIC mode except for TARGET_VXWORKS_RTP.
14762 * config/sparc/sparc.c: Include cfgrtl.h.
14763 (TARGET_INIT_PIC_REG): Define.
14764 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
14765 (sparc_pic_register_p): New predicate.
14766 (sparc_legitimate_address_p): Use it.
14767 (sparc_legitimize_pic_address): Likewise.
14768 (sparc_delegitimize_address): Likewise.
14769 (sparc_mode_dependent_address_p): Likewise.
14770 (gen_load_pcrel_sym): Remove 4th parameter.
14771 (load_got_register): Adjust call to above. Remove obsolete stuff.
14772 (sparc_expand_prologue): Do not call load_got_register here.
14773 (sparc_flat_expand_prologue): Likewise.
14774 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
14775 (sparc_use_pseudo_pic_reg): New function.
14776 (sparc_init_pic_reg): Likewise.
14777 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
14778 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
14780 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
14782 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
14783 Add item for branch_cost.
14785 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
14787 PR rtl-optimization/83565
14788 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
14789 not extend the result to a larger mode for rotate operations.
14790 (num_sign_bit_copies1): Likewise.
14792 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
14795 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
14797 Use values-Xc.o for -pedantic.
14798 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
14800 2018-01-12 Martin Liska <mliska@suse.cz>
14803 * ipa-devirt.c (final_warning_record::grow_type_warnings):
14805 (possible_polymorphic_call_targets): Use it.
14806 (ipa_devirt): Likewise.
14808 2018-01-12 Martin Liska <mliska@suse.cz>
14810 * profile-count.h (enum profile_quality): Use 0 as invalid
14811 enum value of profile_quality.
14813 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
14815 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
14816 -mext-string options.
14818 2018-01-12 Richard Biener <rguenther@suse.de>
14820 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
14821 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
14822 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
14824 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
14826 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
14828 * configure.ac (--with-long-double-format): Add support for the
14829 configuration option to change the default long double format on
14831 * config.gcc (powerpc*-linux*-*): Likewise.
14832 * configure: Regenerate.
14833 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
14834 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
14835 used without modification.
14837 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14839 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
14840 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
14841 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
14842 MISC_BUILTIN_SPEC_BARRIER.
14843 (rs6000_init_builtins): Likewise.
14844 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
14846 (speculation_barrier): New define_insn.
14847 * doc/extend.texi: Document __builtin_speculation_barrier.
14849 2018-01-11 Jakub Jelinek <jakub@redhat.com>
14852 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
14853 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
14854 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
14856 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
14857 integral modes instead of "ss" and "sd".
14858 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
14859 vectors with 32-bit and 64-bit elements.
14860 (vecdupssescalarmodesuffix): New mode attribute.
14861 (vec_dup<mode>): Use it.
14863 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
14866 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
14867 frame if argument is passed on stack.
14869 2018-01-11 Jakub Jelinek <jakub@redhat.com>
14872 * ree.c (combine_reaching_defs): Optimize also
14873 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
14874 reg2=any_extend(exp); reg1=reg2;, formatting fix.
14876 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
14878 PR middle-end/83189
14879 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
14881 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
14883 PR middle-end/83718
14884 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
14885 after they are computed.
14887 2018-01-11 Bin Cheng <bin.cheng@arm.com>
14889 PR tree-optimization/83695
14890 * gimple-loop-linterchange.cc
14891 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
14892 reset cached scev information after interchange.
14893 (pass_linterchange::execute): Remove call to scev_reset_htab.
14895 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14897 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
14898 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
14899 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
14900 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
14901 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
14902 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
14903 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
14904 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
14905 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
14906 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
14907 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
14908 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
14909 (V_lane_reg): Likewise.
14910 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
14912 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
14913 (vfmal_lane_low<mode>_intrinsic,
14914 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
14915 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
14916 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
14917 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
14918 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
14919 vfmsl_lane_high<mode>_intrinsic): New define_insns.
14921 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14923 * config/arm/arm-cpus.in (fp16fml): New feature.
14924 (ALL_SIMD): Add fp16fml.
14925 (armv8.2-a): Add fp16fml as an option.
14926 (armv8.3-a): Likewise.
14927 (armv8.4-a): Add fp16fml as part of fp16.
14928 * config/arm/arm.h (TARGET_FP16FML): Define.
14929 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
14931 * config/arm/arm-modes.def (V2HF): Define.
14932 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
14933 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
14934 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
14935 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
14936 vfmsl_low, vfmsl_high): New set of builtins.
14937 * config/arm/iterators.md (PLUSMINUS): New code iterator.
14938 (vfml_op): New code attribute.
14939 (VFMLHALVES): New int iterator.
14940 (VFML, VFMLSEL): New mode attributes.
14941 (V_reg): Define mapping for V2HF.
14942 (V_hi, V_lo): New mode attributes.
14943 (VF_constraint): Likewise.
14944 (vfml_half, vfml_half_selector): New int attributes.
14945 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
14947 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
14948 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
14950 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
14951 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
14952 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
14953 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
14955 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
14956 Document new effective target and option set.
14958 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14960 * config/arm/arm-cpus.in (armv8_4): New feature.
14961 (ARMv8_4a): New fgroup.
14962 (armv8.4-a): New arch.
14963 * config/arm/arm-tables.opt: Regenerate.
14964 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
14965 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
14966 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
14967 Add matching rules for -march=armv8.4-a and extensions.
14968 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
14970 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
14973 * config/rx/rx.md (BW): New mode attribute.
14974 (sync_lock_test_and_setsi): Add mode suffix to insn output.
14976 2018-01-11 Richard Biener <rguenther@suse.de>
14978 PR tree-optimization/83435
14979 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
14980 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
14981 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
14983 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14984 Alan Hayward <alan.hayward@arm.com>
14985 David Sherwood <david.sherwood@arm.com>
14987 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
14989 (aarch64_classify_address): Initialize it. Track polynomial offsets.
14990 (aarch64_print_address_internal): Use it to check for a zero offset.
14992 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14993 Alan Hayward <alan.hayward@arm.com>
14994 David Sherwood <david.sherwood@arm.com>
14996 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
14997 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
14998 Return a poly_int64 rather than a HOST_WIDE_INT.
14999 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
15000 rather than a HOST_WIDE_INT.
15001 * config/aarch64/aarch64.h (aarch64_frame): Protect with
15002 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
15003 hard_fp_offset, frame_size, initial_adjust, callee_offset and
15004 final_offset from HOST_WIDE_INT to poly_int64.
15005 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
15006 to_constant when getting the number of units in an Advanced SIMD
15008 (aarch64_builtin_vectorized_function): Check for a constant number
15010 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
15012 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
15013 attribute instead of GET_MODE_NUNITS.
15014 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
15015 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
15016 GET_MODE_SIZE for fixed-size registers.
15017 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
15018 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
15019 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
15020 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
15021 (aarch64_print_operand, aarch64_print_address_internal)
15022 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
15023 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
15024 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
15025 Handle polynomial GET_MODE_SIZE.
15026 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
15027 wider than SImode without modification.
15028 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
15029 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
15030 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
15031 passing and returning SVE modes.
15032 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
15033 rather than GEN_INT.
15034 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
15035 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
15036 (aarch64_allocate_and_probe_stack_space): Likewise.
15037 (aarch64_layout_frame): Cope with polynomial offsets.
15038 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
15039 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
15040 polynomial offsets.
15041 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
15042 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
15043 poly_int64 rather than a HOST_WIDE_INT.
15044 (aarch64_get_separate_components, aarch64_process_components)
15045 (aarch64_expand_prologue, aarch64_expand_epilogue)
15046 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
15047 (aarch64_anchor_offset): New function, split out from...
15048 (aarch64_legitimize_address): ...here.
15049 (aarch64_builtin_vectorization_cost): Handle polynomial
15050 TYPE_VECTOR_SUBPARTS.
15051 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
15053 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
15054 number of elements from the PARALLEL rather than the mode.
15055 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
15056 rather than GET_MODE_BITSIZE.
15057 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
15058 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
15059 (aarch64_expand_vec_perm_const_1): Handle polynomial
15060 d->perm.length () and d->perm elements.
15061 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
15062 Apply to_constant to d->perm elements.
15063 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
15064 polynomial CONST_VECTOR_NUNITS.
15065 (aarch64_move_pointer): Take amount as a poly_int64 rather
15067 (aarch64_progress_pointer): Avoid temporary variable.
15068 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
15069 the mode attribute instead of GET_MODE.
15071 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
15072 Alan Hayward <alan.hayward@arm.com>
15073 David Sherwood <david.sherwood@arm.com>
15075 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
15076 x exists before using it.
15077 (aarch64_add_constant_internal): Rename to...
15078 (aarch64_add_offset_1): ...this. Replace regnum with separate
15079 src and dest rtxes. Handle the case in which they're different,
15080 including when the offset is zero. Replace scratchreg with an rtx.
15081 Use 2 additions if there is no spare register into which we can
15082 move a 16-bit constant.
15083 (aarch64_add_constant): Delete.
15084 (aarch64_add_offset): Replace reg with separate src and dest
15085 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
15086 Use aarch64_add_offset_1.
15087 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
15088 an rtx rather than an int. Take the delta as a poly_int64
15089 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
15090 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
15091 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
15092 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
15093 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
15094 and aarch64_add_sp.
15095 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
15096 aarch64_add_constant.
15098 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
15100 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
15101 Use scalar_float_mode.
15103 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
15105 * config/aarch64/aarch64-simd.md
15106 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
15107 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
15108 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
15109 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
15110 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
15111 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
15112 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
15113 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
15114 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
15115 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
15117 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
15120 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
15121 targ_options->x_arm_arch_string is non NULL.
15123 2018-01-11 Tamar Christina <tamar.christina@arm.com>
15125 * config/aarch64/aarch64.h
15126 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
15128 2018-01-11 Sudakshina Das <sudi.das@arm.com>
15131 * expmed.c (emit_store_flag_force): Swap if const op0
15132 and change VOIDmode to mode of op0.
15134 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
15136 PR rtl-optimization/83761
15137 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
15138 than bytes to mode_for_size.
15140 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
15142 PR middle-end/83189
15143 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
15144 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
15147 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
15149 PR middle-end/83575
15150 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
15151 when in layout mode.
15152 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
15153 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
15156 2018-01-10 Michael Collison <michael.collison@arm.com>
15158 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
15159 * config/aarch64/aarch64-option-extension.def: Add
15160 AARCH64_OPT_EXTENSION of 'fp16fml'.
15161 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
15162 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
15163 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
15164 * config/aarch64/constraints.md (Ui7): New constraint.
15165 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
15166 (VFMLA_SEL_W): Ditto.
15169 (VFMLA16_LOW): New int iterator.
15170 (VFMLA16_HIGH): Ditto.
15171 (UNSPEC_FMLAL): New unspec.
15172 (UNSPEC_FMLSL): Ditto.
15173 (UNSPEC_FMLAL2): Ditto.
15174 (UNSPEC_FMLSL2): Ditto.
15175 (f16mac): New code attribute.
15176 * config/aarch64/aarch64-simd-builtins.def
15177 (aarch64_fmlal_lowv2sf): Ditto.
15178 (aarch64_fmlsl_lowv2sf): Ditto.
15179 (aarch64_fmlalq_lowv4sf): Ditto.
15180 (aarch64_fmlslq_lowv4sf): Ditto.
15181 (aarch64_fmlal_highv2sf): Ditto.
15182 (aarch64_fmlsl_highv2sf): Ditto.
15183 (aarch64_fmlalq_highv4sf): Ditto.
15184 (aarch64_fmlslq_highv4sf): Ditto.
15185 (aarch64_fmlal_lane_lowv2sf): Ditto.
15186 (aarch64_fmlsl_lane_lowv2sf): Ditto.
15187 (aarch64_fmlal_laneq_lowv2sf): Ditto.
15188 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
15189 (aarch64_fmlalq_lane_lowv4sf): Ditto.
15190 (aarch64_fmlsl_lane_lowv4sf): Ditto.
15191 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
15192 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
15193 (aarch64_fmlal_lane_highv2sf): Ditto.
15194 (aarch64_fmlsl_lane_highv2sf): Ditto.
15195 (aarch64_fmlal_laneq_highv2sf): Ditto.
15196 (aarch64_fmlsl_laneq_highv2sf): Ditto.
15197 (aarch64_fmlalq_lane_highv4sf): Ditto.
15198 (aarch64_fmlsl_lane_highv4sf): Ditto.
15199 (aarch64_fmlalq_laneq_highv4sf): Ditto.
15200 (aarch64_fmlsl_laneq_highv4sf): Ditto.
15201 * config/aarch64/aarch64-simd.md:
15202 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
15203 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
15204 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
15205 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
15206 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
15207 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
15208 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
15209 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
15210 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
15211 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
15212 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
15213 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
15214 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
15215 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
15216 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
15217 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
15218 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
15219 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
15220 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
15221 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
15222 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
15223 (vfmlsl_low_u32): Ditto.
15224 (vfmlalq_low_u32): Ditto.
15225 (vfmlslq_low_u32): Ditto.
15226 (vfmlal_high_u32): Ditto.
15227 (vfmlsl_high_u32): Ditto.
15228 (vfmlalq_high_u32): Ditto.
15229 (vfmlslq_high_u32): Ditto.
15230 (vfmlal_lane_low_u32): Ditto.
15231 (vfmlsl_lane_low_u32): Ditto.
15232 (vfmlal_laneq_low_u32): Ditto.
15233 (vfmlsl_laneq_low_u32): Ditto.
15234 (vfmlalq_lane_low_u32): Ditto.
15235 (vfmlslq_lane_low_u32): Ditto.
15236 (vfmlalq_laneq_low_u32): Ditto.
15237 (vfmlslq_laneq_low_u32): Ditto.
15238 (vfmlal_lane_high_u32): Ditto.
15239 (vfmlsl_lane_high_u32): Ditto.
15240 (vfmlal_laneq_high_u32): Ditto.
15241 (vfmlsl_laneq_high_u32): Ditto.
15242 (vfmlalq_lane_high_u32): Ditto.
15243 (vfmlslq_lane_high_u32): Ditto.
15244 (vfmlalq_laneq_high_u32): Ditto.
15245 (vfmlslq_laneq_high_u32): Ditto.
15246 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
15247 (AARCH64_FL_FOR_ARCH8_4): New.
15248 (AARCH64_ISA_F16FML): New ISA flag.
15249 (TARGET_F16FML): New feature flag for fp16fml.
15250 (doc/invoke.texi): Document new fp16fml option.
15252 2018-01-10 Michael Collison <michael.collison@arm.com>
15254 * config/aarch64/aarch64-builtins.c:
15255 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
15256 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
15257 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
15258 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
15259 (AARCH64_ISA_SHA3): New ISA flag.
15260 (TARGET_SHA3): New feature flag for sha3.
15261 * config/aarch64/iterators.md (sha512_op): New int attribute.
15262 (CRYPTO_SHA512): New int iterator.
15263 (UNSPEC_SHA512H): New unspec.
15264 (UNSPEC_SHA512H2): Ditto.
15265 (UNSPEC_SHA512SU0): Ditto.
15266 (UNSPEC_SHA512SU1): Ditto.
15267 * config/aarch64/aarch64-simd-builtins.def
15268 (aarch64_crypto_sha512hqv2di): New builtin.
15269 (aarch64_crypto_sha512h2qv2di): Ditto.
15270 (aarch64_crypto_sha512su0qv2di): Ditto.
15271 (aarch64_crypto_sha512su1qv2di): Ditto.
15272 (aarch64_eor3qv8hi): Ditto.
15273 (aarch64_rax1qv2di): Ditto.
15274 (aarch64_xarqv2di): Ditto.
15275 (aarch64_bcaxqv8hi): Ditto.
15276 * config/aarch64/aarch64-simd.md:
15277 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
15278 (aarch64_crypto_sha512su0qv2di): Ditto.
15279 (aarch64_crypto_sha512su1qv2di): Ditto.
15280 (aarch64_eor3qv8hi): Ditto.
15281 (aarch64_rax1qv2di): Ditto.
15282 (aarch64_xarqv2di): Ditto.
15283 (aarch64_bcaxqv8hi): Ditto.
15284 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
15285 (vsha512h2q_u64): Ditto.
15286 (vsha512su0q_u64): Ditto.
15287 (vsha512su1q_u64): Ditto.
15288 (veor3q_u16): Ditto.
15289 (vrax1q_u64): Ditto.
15290 (vxarq_u64): Ditto.
15291 (vbcaxq_u16): Ditto.
15292 * config/arm/types.md (crypto_sha512): New type attribute.
15293 (crypto_sha3): Ditto.
15294 (doc/invoke.texi): Document new sha3 option.
15296 2018-01-10 Michael Collison <michael.collison@arm.com>
15298 * config/aarch64/aarch64-builtins.c:
15299 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
15300 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
15301 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
15302 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
15303 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
15304 (AARCH64_ISA_SM4): New ISA flag.
15305 (TARGET_SM4): New feature flag for sm4.
15306 * config/aarch64/aarch64-simd-builtins.def
15307 (aarch64_sm3ss1qv4si): Ditto.
15308 (aarch64_sm3tt1aq4si): Ditto.
15309 (aarch64_sm3tt1bq4si): Ditto.
15310 (aarch64_sm3tt2aq4si): Ditto.
15311 (aarch64_sm3tt2bq4si): Ditto.
15312 (aarch64_sm3partw1qv4si): Ditto.
15313 (aarch64_sm3partw2qv4si): Ditto.
15314 (aarch64_sm4eqv4si): Ditto.
15315 (aarch64_sm4ekeyqv4si): Ditto.
15316 * config/aarch64/aarch64-simd.md:
15317 (aarch64_sm3ss1qv4si): Ditto.
15318 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
15319 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
15320 (aarch64_sm4eqv4si): Ditto.
15321 (aarch64_sm4ekeyqv4si): Ditto.
15322 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
15323 (sm3part_op): Ditto.
15324 (CRYPTO_SM3TT): Ditto.
15325 (CRYPTO_SM3PART): Ditto.
15326 (UNSPEC_SM3SS1): New unspec.
15327 (UNSPEC_SM3TT1A): Ditto.
15328 (UNSPEC_SM3TT1B): Ditto.
15329 (UNSPEC_SM3TT2A): Ditto.
15330 (UNSPEC_SM3TT2B): Ditto.
15331 (UNSPEC_SM3PARTW1): Ditto.
15332 (UNSPEC_SM3PARTW2): Ditto.
15333 (UNSPEC_SM4E): Ditto.
15334 (UNSPEC_SM4EKEY): Ditto.
15335 * config/aarch64/constraints.md (Ui2): New constraint.
15336 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
15337 * config/arm/types.md (crypto_sm3): New type attribute.
15338 (crypto_sm4): Ditto.
15339 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
15340 (vsm3tt1aq_u32): Ditto.
15341 (vsm3tt1bq_u32): Ditto.
15342 (vsm3tt2aq_u32): Ditto.
15343 (vsm3tt2bq_u32): Ditto.
15344 (vsm3partw1q_u32): Ditto.
15345 (vsm3partw2q_u32): Ditto.
15346 (vsm4eq_u32): Ditto.
15347 (vsm4ekeyq_u32): Ditto.
15348 (doc/invoke.texi): Document new sm4 option.
15350 2018-01-10 Michael Collison <michael.collison@arm.com>
15352 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
15353 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
15354 (AARCH64_FL_FOR_ARCH8_4): New.
15355 (AARCH64_FL_V8_4): New flag.
15356 (doc/invoke.texi): Document new armv8.4-a option.
15358 2018-01-10 Michael Collison <michael.collison@arm.com>
15360 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
15361 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
15362 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
15363 * config/aarch64/aarch64-option-extension.def: Add
15364 AARCH64_OPT_EXTENSION of 'sha2'.
15365 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
15366 (crypto): Disable sha2 and aes if crypto disabled.
15367 (crypto): Enable aes and sha2 if enabled.
15368 (simd): Disable sha2 and aes if simd disabled.
15369 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
15371 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
15372 (TARGET_SHA2): New feature flag for sha2.
15373 (TARGET_AES): New feature flag for aes.
15374 * config/aarch64/aarch64-simd.md:
15375 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
15376 conditional on TARGET_AES.
15377 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
15378 (aarch64_crypto_sha1hsi): Make pattern conditional
15380 (aarch64_crypto_sha1hv4si): Ditto.
15381 (aarch64_be_crypto_sha1hv4si): Ditto.
15382 (aarch64_crypto_sha1su1v4si): Ditto.
15383 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
15384 (aarch64_crypto_sha1su0v4si): Ditto.
15385 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
15386 (aarch64_crypto_sha256su0v4si): Ditto.
15387 (aarch64_crypto_sha256su1v4si): Ditto.
15388 (doc/invoke.texi): Document new aes and sha2 options.
15390 2018-01-10 Martin Sebor <msebor@redhat.com>
15392 PR tree-optimization/83781
15393 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
15396 2018-01-11 Martin Sebor <msebor@gmail.com>
15397 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
15399 PR tree-optimization/83501
15400 PR tree-optimization/81703
15402 * tree-ssa-strlen.c (get_string_cst): Rename...
15403 (get_string_len): ...to this. Handle global constants.
15404 (handle_char_store): Adjust.
15406 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
15407 Jim Wilson <jimw@sifive.com>
15409 * config/riscv/riscv-protos.h (riscv_output_return): New.
15410 * config/riscv/riscv.c (struct machine_function): New naked_p field.
15411 (riscv_attribute_table, riscv_output_return),
15412 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
15413 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
15414 (riscv_compute_frame_info): Only compute frame->mask if not a naked
15416 (riscv_expand_prologue): Add early return for naked function.
15417 (riscv_expand_epilogue): Likewise.
15418 (riscv_function_ok_for_sibcall): Return false for naked function.
15419 (riscv_set_current_function): New.
15420 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
15421 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
15422 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
15423 * doc/extend.texi (RISC-V Function Attributes): New.
15425 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
15427 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
15428 check for 128-bit long double before checking TCmode.
15429 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
15430 128-bit long doubles before checking TFmode or TCmode.
15431 (FLOAT128_IBM_P): Likewise.
15433 2018-01-10 Martin Sebor <msebor@redhat.com>
15435 PR tree-optimization/83671
15436 * builtins.c (c_strlen): Unconditionally return zero for the empty
15438 Use -Warray-bounds for warnings.
15439 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
15440 for non-constant array indices with COMPONENT_REF, arrays of
15441 arrays, and pointers to arrays.
15442 (gimple_fold_builtin_strlen): Determine and set length range for
15443 non-constant character arrays.
15445 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
15447 PR middle-end/81897
15448 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
15451 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
15453 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
15455 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
15458 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
15459 VECTOR_MEM_ALTIVEC_OR_VSX_P.
15460 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
15461 indexed_or_indirect_operand predicate.
15462 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
15463 (*vsx_le_perm_load_v8hi): Likewise.
15464 (*vsx_le_perm_load_v16qi): Likewise.
15465 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
15466 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
15467 (*vsx_le_perm_store_v8hi): Likewise.
15468 (*vsx_le_perm_store_v16qi): Likewise.
15469 (eight unnamed splitters): Likewise.
15471 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
15473 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
15474 * config/rs6000/emmintrin.h: Likewise.
15475 * config/rs6000/mmintrin.h: Likewise.
15476 * config/rs6000/xmmintrin.h: Likewise.
15478 2018-01-10 David Malcolm <dmalcolm@redhat.com>
15481 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
15483 * tree.c (tree_nop_conversion): Return true for location wrapper
15485 (maybe_wrap_with_location): New function.
15486 (selftest::check_strip_nops): New function.
15487 (selftest::test_location_wrappers): New function.
15488 (selftest::tree_c_tests): Call it.
15489 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
15490 (maybe_wrap_with_location): New decl.
15491 (EXPR_LOCATION_WRAPPER_P): New macro.
15492 (location_wrapper_p): New inline function.
15493 (tree_strip_any_location_wrapper): New inline function.
15495 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
15498 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
15499 stack_realign_offset for the largest alignment of stack slot
15501 (ix86_find_max_used_stack_alignment): New function.
15502 (ix86_finalize_stack_frame_flags): Use it. Set
15503 max_used_stack_alignment if we don't realign stack.
15504 * config/i386/i386.h (machine_function): Add
15505 max_used_stack_alignment.
15507 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
15509 * config/arm/arm.opt (-mbranch-cost): New option.
15510 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
15513 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
15516 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
15517 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
15519 2018-01-10 Richard Biener <rguenther@suse.de>
15522 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
15523 early out so it also covers the case where we have a non-NULL
15526 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
15528 PR tree-optimization/83753
15529 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
15530 for non-strided grouped accesses if the number of elements is 1.
15532 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
15535 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
15536 * i386.h (TARGET_USE_GATHER): Define.
15537 * x86-tune.def (X86_TUNE_USE_GATHER): New.
15539 2018-01-10 Martin Liska <mliska@suse.cz>
15542 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
15543 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
15545 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
15546 CLEANUP_NO_PARTITIONING is not set.
15548 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
15550 * doc/rtl.texi: Remove documentation of (const ...) wrappers
15551 for vectors, as a partial revert of r254296.
15552 * rtl.h (const_vec_p): Delete.
15553 (const_vec_duplicate_p): Don't test for vector CONSTs.
15554 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
15555 * expmed.c (make_tree): Likewise.
15558 * common.md (E, F): Use CONSTANT_P instead of checking for
15560 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
15561 checking for CONST_VECTOR.
15563 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
15565 PR middle-end/83575
15566 * predict.c (force_edge_cold): Handle in more sane way edges
15567 with no prediction.
15569 2018-01-09 Carl Love <cel@us.ibm.com>
15571 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
15573 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
15574 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
15575 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
15576 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
15577 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
15578 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
15579 * config/rs6000/rs6000-protos.h: Add extern defition for
15580 rs6000_generate_float2_double_code.
15581 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
15583 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
15584 (float2_v2df): Add define_expand.
15586 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
15589 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
15590 op_mode in the force_to_mode call.
15592 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
15594 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
15595 instead of checking each element individually.
15596 (aarch64_evpc_uzp): Likewise.
15597 (aarch64_evpc_zip): Likewise.
15598 (aarch64_evpc_ext): Likewise.
15599 (aarch64_evpc_rev): Likewise.
15600 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
15601 instead of checking each element individually. Return true without
15603 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
15604 whether all selected elements come from the same input, instead of
15605 checking each element individually. Remove calls to gen_rtx_REG,
15606 start_sequence and end_sequence and instead assert that no rtl is
15609 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
15611 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
15612 order of HIGH and CONST checks.
15614 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
15616 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
15617 if the destination isn't an SSA_NAME.
15619 2018-01-09 Richard Biener <rguenther@suse.de>
15621 PR tree-optimization/83668
15622 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
15624 (canonicalize_loop_form): ... here, renamed from ...
15625 (canonicalize_loop_closed_ssa_form): ... this and amended to
15626 swap successor edges for loop exit blocks to make us use
15627 the RPO order we need for initial schedule generation.
15629 2018-01-09 Joseph Myers <joseph@codesourcery.com>
15631 PR tree-optimization/64811
15632 * match.pd: When optimizing comparisons with Inf, avoid
15633 introducing or losing exceptions from comparisons with NaN.
15635 2018-01-09 Martin Liska <mliska@suse.cz>
15638 * asan.c (shadow_mem_size): Add gcc_assert.
15640 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
15642 Don't save registers in main().
15645 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
15646 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
15647 * config/avr/avr.c (avr_set_current_function): Don't error if
15648 naked, OS_task or OS_main are specified at the same time.
15649 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
15651 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
15653 * common/config/avr/avr-common.c (avr_option_optimization_table):
15654 Switch on -mmain-is-OS_task for optimizing compilations.
15656 2018-01-09 Richard Biener <rguenther@suse.de>
15658 PR tree-optimization/83572
15659 * graphite.c: Include cfganal.h.
15660 (graphite_transform_loops): Connect infinite loops to exit
15661 and remove fake edges at the end.
15663 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
15665 * ipa-inline.c (edge_badness): Revert accidental checkin.
15667 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
15670 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
15671 symbols; not inline clones.
15673 2018-01-09 Jakub Jelinek <jakub@redhat.com>
15676 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
15677 hard registers. Formatting fixes.
15679 PR preprocessor/83722
15680 * gcc.c (try_generate_repro): Pass
15681 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
15682 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
15685 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
15686 Kito Cheng <kito.cheng@gmail.com>
15688 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
15689 (riscv_leaf_function_p): Delete.
15690 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
15692 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
15694 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
15696 (do_ifelse): New function.
15697 (do_isel): New function.
15698 (do_sub3): New function.
15699 (do_add3): New function.
15700 (do_load_mask_compare): New function.
15701 (do_overlap_load_compare): New function.
15702 (expand_compare_loop): New function.
15703 (expand_block_compare): Call expand_compare_loop() when appropriate.
15704 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
15705 option description.
15706 (-mblock-compare-inline-loop-limit): New option.
15708 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
15711 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
15712 Reverse order of second and third operands in first alternative.
15713 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
15714 of first and second elements in UNSPEC_VPERMR vector.
15715 (altivec_expand_vec_perm_le): Likewise.
15717 2018-01-08 Jeff Law <law@redhat.com>
15719 PR rtl-optimizatin/81308
15720 * tree-switch-conversion.c (cfg_altered): New file scoped static.
15721 (process_switch): If group_case_labels makes a change, then set
15723 (pass_convert_switch::execute): If a switch is converted, then
15724 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
15726 PR rtl-optimization/81308
15727 * recog.c (split_all_insns): Conditionally cleanup the CFG after
15730 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
15732 PR target/83663 - Revert r255946
15733 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
15734 generation for cases where splatting a value is not useful.
15735 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
15736 across a vec_duplicate and a paradoxical subreg forming a vector
15737 mode to a vec_concat.
15739 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15741 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
15742 -march=armv8.3-a variants.
15743 * config/arm/t-multilib: Likewise.
15744 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
15746 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
15748 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
15750 (cceq_ior_compare_complement): Give it a name so I can use it, and
15751 change boolean_or_operator predicate to boolean_operator so it can
15752 be used to generate a crand.
15753 (eqne): New code iterator.
15754 (bd/bd_neg): New code_attrs.
15755 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
15756 a single define_insn.
15757 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
15758 decrement (bdnzt/bdnzf/bdzt/bdzf).
15759 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
15760 with the new names of the branch decrement patterns, and added the
15761 names of the branch decrement conditional patterns.
15763 2018-01-08 Richard Biener <rguenther@suse.de>
15765 PR tree-optimization/83563
15766 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
15769 2018-01-08 Richard Biener <rguenther@suse.de>
15771 PR middle-end/83713
15772 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
15774 2018-01-08 Richard Biener <rguenther@suse.de>
15776 PR tree-optimization/83685
15777 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
15778 references to abnormals.
15780 2018-01-08 Richard Biener <rguenther@suse.de>
15783 * dwarf2out.c (output_indirect_strings): Handle empty
15784 skeleton_debug_str_hash.
15785 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
15787 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
15789 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
15790 (emit_store_direct): Likewise.
15791 (arc_trampoline_adjust_address): Likewise.
15792 (arc_asm_trampoline_template): New function.
15793 (arc_initialize_trampoline): Use asm_trampoline_template.
15794 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
15795 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
15796 * config/arc/arc.md (flush_icache): Delete pattern.
15798 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
15800 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
15801 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
15804 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
15807 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
15808 by not USED_FOR_TARGET.
15809 (make_pass_resolve_sw_modes): Likewise.
15811 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
15813 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
15816 2018-01-08 Richard Biener <rguenther@suse.de>
15818 PR middle-end/83580
15819 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
15821 2018-01-08 Richard Biener <rguenther@suse.de>
15823 PR middle-end/83517
15824 * match.pd ((t * 2) / 2) -> t): Add missing :c.
15826 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
15828 PR middle-end/81897
15829 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
15830 basic blocks with a small number of successors.
15831 (convert_control_dep_chain_into_preds): Improve handling of
15833 (dump_predicates): Split apart into...
15834 (dump_pred_chain): ...here...
15835 (dump_pred_info): ...and here.
15836 (can_one_predicate_be_invalidated_p): Add debugging printfs.
15837 (can_chain_union_be_invalidated_p): Improve check for invalidation
15839 (uninit_uses_cannot_happen): Avoid unnecessary if
15840 convert_control_dep_chain_into_preds yielded nothing.
15842 2018-01-06 Martin Sebor <msebor@redhat.com>
15844 PR tree-optimization/83640
15845 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
15846 subtracting negative offset from size.
15847 (builtin_access::overlap): Adjust offset bounds of the access to fall
15848 within the size of the object if possible.
15850 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
15852 PR rtl-optimization/83699
15853 * expmed.c (extract_bit_field_1): Restrict the vector usage of
15854 extract_bit_field_as_subreg to cases in which the extracted
15855 value is also a vector.
15857 * lra-constraints.c (process_alt_operands): Test for the equivalence
15858 substitutions when detecting a possible reload cycle.
15860 2018-01-06 Jakub Jelinek <jakub@redhat.com>
15863 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
15864 by default if flag_selective_schedling{,2}. Formatting fixes.
15866 PR rtl-optimization/83682
15867 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
15868 if it has non-VECTOR_MODE element mode.
15869 (vec_duplicate_p): Likewise.
15871 PR middle-end/83694
15872 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
15873 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
15875 2018-01-05 Jakub Jelinek <jakub@redhat.com>
15878 * config/i386/i386-builtin.def
15879 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
15880 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
15881 Require also OPTION_MASK_ISA_AVX512F in addition to
15882 OPTION_MASK_ISA_GFNI.
15883 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
15884 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
15885 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
15886 to OPTION_MASK_ISA_GFNI.
15887 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
15888 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
15889 OPTION_MASK_ISA_AVX512BW.
15890 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
15891 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
15892 addition to OPTION_MASK_ISA_GFNI.
15893 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
15894 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
15895 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
15896 to OPTION_MASK_ISA_GFNI.
15897 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
15898 a requirement for all ISAs rather than any of them with a few
15900 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
15902 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
15903 bitmasks to be enabled with 3 exceptions, instead of requiring any
15904 enabled ISA with lots of exceptions.
15905 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
15906 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
15907 Change avx512bw in isa attribute to avx512f.
15908 * config/i386/sgxintrin.h: Add license boilerplate.
15909 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
15910 to __AVX512F__ and __AVX512VL to __AVX512VL__.
15911 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
15912 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
15914 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
15915 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
15916 temporarily sse2 rather than sse if not enabled already.
15919 * config/i386/sse.md (VI248_VLBW): Rename to ...
15920 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
15921 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
15922 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
15923 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
15924 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
15925 mode iterator instead of VI248_VLBW.
15927 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
15929 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
15930 (record_modified): Skip clobbers; add debug output.
15931 (param_change_prob): Use sreal frequencies.
15933 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
15935 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
15936 punt for user-aligned variables.
15938 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
15940 * tree-chrec.c (chrec_contains_symbols): Return true for
15943 2018-01-05 Sudakshina Das <sudi.das@arm.com>
15946 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
15947 of (x|y) == x for BICS pattern.
15949 2018-01-05 Jakub Jelinek <jakub@redhat.com>
15951 PR tree-optimization/83605
15952 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
15953 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
15956 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
15958 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
15959 * config/epiphany/rtems.h: New file.
15961 2018-01-04 Jakub Jelinek <jakub@redhat.com>
15962 Uros Bizjak <ubizjak@gmail.com>
15965 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
15966 QIreg_operand instead of register_operand predicate.
15967 * config/i386/i386.c (ix86_rop_should_change_byte_p,
15968 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
15969 comments instead of -fmitigate[-_]rop.
15971 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
15974 * cgraphunit.c (symbol_table::compile): Switch to text_section
15975 before calling assembly_start debug hook.
15976 * run-rtl-passes.c (run_rtl_passes): Likewise.
15979 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15981 * tree-vrp.c (extract_range_from_binary_expr_1): Check
15982 range_int_cst_p rather than !symbolic_range_p before calling
15983 extract_range_from_multiplicative_op_1.
15985 2018-01-04 Jeff Law <law@redhat.com>
15987 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
15988 redundant test in assertion.
15990 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15992 * doc/rtl.texi: Document machine_mode wrapper classes.
15994 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15996 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
15997 using tree_to_uhwi.
15999 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
16001 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
16002 the VEC_PERM_EXPR fold to fail.
16004 2018-01-04 Jakub Jelinek <jakub@redhat.com>
16007 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
16008 to switched_sections.
16010 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
16013 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
16014 test for d.testing.
16016 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
16019 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
16020 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
16022 2018-01-04 Jakub Jelinek <jakub@redhat.com>
16025 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
16026 is BLKmode and bitpos not zero or mode change is needed.
16028 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
16031 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
16034 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
16037 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
16038 instead of MULT rtx. Update all corresponding splitters.
16039 (*saddl_se): Ditto.
16040 (*ssub<modesuffix>): Ditto.
16041 (*ssubl_se): Ditto.
16042 (*cmp_sadd_di): Update split patterns.
16043 (*cmp_sadd_si): Ditto.
16044 (*cmp_sadd_sidi): Ditto.
16045 (*cmp_ssub_di): Ditto.
16046 (*cmp_ssub_si): Ditto.
16047 (*cmp_ssub_sidi): Ditto.
16048 * config/alpha/predicates.md (const23_operand): New predicate.
16049 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
16050 Look for ASHIFT, not MULT inner operand.
16051 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
16053 2018-01-04 Martin Liska <mliska@suse.cz>
16055 PR gcov-profile/83669
16056 * gcov.c (output_intermediate_file): Add version to intermediate
16058 * doc/gcov.texi: Document new field 'version' in intermediate
16059 file format. Fix location of '-k' option of gcov command.
16061 2018-01-04 Martin Liska <mliska@suse.cz>
16064 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
16066 2018-01-04 Jakub Jelinek <jakub@redhat.com>
16068 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
16070 2018-01-03 Martin Sebor <msebor@redhat.com>
16072 PR tree-optimization/83655
16073 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
16074 checking calls with invalid arguments.
16076 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16078 * tree-vect-stmts.c (vect_get_store_rhs): New function.
16079 (vectorizable_mask_load_store): Delete.
16080 (vectorizable_call): Return false for masked loads and stores.
16081 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
16082 instead of gimple_assign_rhs1.
16083 (vectorizable_load): Handle IFN_MASK_LOAD.
16084 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
16086 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16088 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
16090 (vectorizable_mask_load_store): ...here.
16091 (vectorizable_load): ...and here.
16093 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16095 * tree-vect-stmts.c (vect_build_all_ones_mask)
16096 (vect_build_zero_merge_argument): New functions, split out from...
16097 (vectorizable_load): ...here.
16099 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16101 * tree-vect-stmts.c (vect_check_store_rhs): New function,
16103 (vectorizable_mask_load_store): ...here.
16104 (vectorizable_store): ...and here.
16106 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16108 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
16110 (vectorizable_mask_load_store): ...here.
16112 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16114 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
16115 (vect_model_store_cost): Take a vec_load_store_type instead of a
16117 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
16118 (vect_model_store_cost): Take a vec_load_store_type instead of a
16120 (vectorizable_mask_load_store): Update accordingly.
16121 (vectorizable_store): Likewise.
16122 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
16124 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16126 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
16127 IFN_MASK_LOAD calls here rather than...
16128 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
16130 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16131 Alan Hayward <alan.hayward@arm.com>
16132 David Sherwood <david.sherwood@arm.com>
16134 * expmed.c (extract_bit_field_1): For vector extracts,
16135 fall back to extract_bit_field_as_subreg if vec_extract
16138 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16139 Alan Hayward <alan.hayward@arm.com>
16140 David Sherwood <david.sherwood@arm.com>
16142 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
16143 they are variable or constant sized.
16144 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
16145 slots for constant-sized data.
16147 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16148 Alan Hayward <alan.hayward@arm.com>
16149 David Sherwood <david.sherwood@arm.com>
16151 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
16152 handling COND_EXPRs with boolean comparisons, try to find a better
16153 basis for the mask type than the boolean itself.
16155 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16157 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
16158 is calculated and how it can be overridden.
16159 * genmodes.c (max_bitsize_mode_any_mode): New variable.
16160 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
16162 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
16165 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16166 Alan Hayward <alan.hayward@arm.com>
16167 David Sherwood <david.sherwood@arm.com>
16169 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
16170 Remove the mode argument.
16171 (aarch64_simd_valid_immediate): Remove the mode and inverse
16173 * config/aarch64/iterators.md (bitsize): New iterator.
16174 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
16175 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
16176 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
16177 aarch64_simd_valid_immediate.
16178 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
16179 (aarch64_reg_or_bic_imm): Likewise.
16180 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
16181 with an insn_type enum and msl with a modifier_type enum.
16182 Replace element_width with a scalar_mode. Change the shift
16183 to unsigned int. Add constructors for scalar_float_mode and
16184 scalar_int_mode elements.
16185 (aarch64_vect_float_const_representable_p): Delete.
16186 (aarch64_can_const_movi_rtx_p)
16187 (aarch64_simd_scalar_immediate_valid_for_move)
16188 (aarch64_simd_make_constant): Update call to
16189 aarch64_simd_valid_immediate.
16190 (aarch64_advsimd_valid_immediate_hs): New function.
16191 (aarch64_advsimd_valid_immediate): Likewise.
16192 (aarch64_simd_valid_immediate): Remove mode and inverse
16193 arguments. Rewrite to use the above. Use const_vec_duplicate_p
16194 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
16195 and aarch64_float_const_representable_p on the result.
16196 (aarch64_output_simd_mov_immediate): Remove mode argument.
16197 Update call to aarch64_simd_valid_immediate and use of
16198 simd_immediate_info.
16199 (aarch64_output_scalar_simd_mov_immediate): Update call
16202 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16203 Alan Hayward <alan.hayward@arm.com>
16204 David Sherwood <david.sherwood@arm.com>
16206 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
16207 (mode_nunits): Likewise CONST_MODE_NUNITS.
16208 * machmode.def (ADJUST_NUNITS): Document.
16209 * genmodes.c (mode_data::need_nunits_adj): New field.
16210 (blank_mode): Update accordingly.
16211 (adj_nunits): New variable.
16212 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
16214 (emit_mode_size_inline): Set need_bytesize_adj for all modes
16215 listed in adj_nunits.
16216 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
16217 listed in adj_nunits. Don't emit case statements for such modes.
16218 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
16219 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
16220 nothing if adj_nunits is nonnull.
16221 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
16222 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
16223 (emit_mode_fbit): Update use of print_maybe_const_decl.
16224 (emit_move_size): Likewise. Treat the array as non-const
16226 (emit_mode_adjustments): Handle adj_nunits.
16228 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16230 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
16231 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
16232 (VECTOR_MODES): Use it.
16233 (make_vector_modes): Take the prefix as an argument.
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 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
16240 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
16241 for MODE_VECTOR_BOOL.
16242 * machmode.def (VECTOR_BOOL_MODE): Document.
16243 * genmodes.c (VECTOR_BOOL_MODE): New macro.
16244 (make_vector_bool_mode): New function.
16245 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
16247 * lto-streamer-in.c (lto_input_mode_table): Likewise.
16248 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
16250 * stor-layout.c (int_mode_for_mode): Likewise.
16251 * tree.c (build_vector_type_for_mode): Likewise.
16252 * varasm.c (output_constant_pool_2): Likewise.
16253 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
16254 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
16255 for MODE_VECTOR_BOOL.
16256 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
16257 of mode class checks.
16258 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
16259 instead of a list of mode class checks.
16260 (expand_vector_scalar_condition): Likewise.
16261 (type_for_widest_vector_mode): Handle BImode as an inner mode.
16263 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16264 Alan Hayward <alan.hayward@arm.com>
16265 David Sherwood <david.sherwood@arm.com>
16267 * machmode.h (mode_size): Change from unsigned short to
16269 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
16270 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
16271 or if measurement_type is not polynomial.
16272 (fixed_size_mode::includes_p): Check for constant-sized modes.
16273 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
16274 return a poly_uint16 rather than an unsigned short.
16275 (emit_mode_size): Change the type of mode_size from unsigned short
16276 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
16277 (emit_mode_adjustments): Cope with polynomial vector sizes.
16278 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
16280 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
16282 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
16283 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
16284 * caller-save.c (setup_save_areas): Likewise.
16285 (replace_reg_with_saved_mem): Likewise.
16286 * calls.c (emit_library_call_value_1): Likewise.
16287 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
16288 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
16289 (gen_lowpart_for_combine): Likewise.
16290 * convert.c (convert_to_integer_1): Likewise.
16291 * cse.c (equiv_constant, cse_insn): Likewise.
16292 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
16293 (cselib_subst_to_values): Likewise.
16294 * dce.c (word_dce_process_block): Likewise.
16295 * df-problems.c (df_word_lr_mark_ref): Likewise.
16296 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
16297 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
16298 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
16299 (rtl_for_decl_location): Likewise.
16300 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
16301 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
16302 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
16303 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
16304 (expand_expr_real_1): Likewise.
16305 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
16306 (pad_below): Likewise.
16307 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
16308 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
16309 * ira.c (get_subreg_tracking_sizes): Likewise.
16310 * ira-build.c (ira_create_allocno_objects): Likewise.
16311 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
16312 (ira_sort_regnos_for_alter_reg): Likewise.
16313 * ira-costs.c (record_operand_costs): Likewise.
16314 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
16315 (resolve_simple_move): Likewise.
16316 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
16317 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
16318 (lra_constraints): Likewise.
16319 (CONST_POOL_OK_P): Reject variable-sized modes.
16320 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
16321 (add_pseudo_to_slot, lra_spill): Likewise.
16322 * omp-low.c (omp_clause_aligned_alignment): Likewise.
16323 * optabs-query.c (get_best_extraction_insn): Likewise.
16324 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
16325 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
16326 (expand_mult_highpart, valid_multiword_target_p): Likewise.
16327 * recog.c (offsettable_address_addr_space_p): Likewise.
16328 * regcprop.c (maybe_mode_change): Likewise.
16329 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
16330 * regrename.c (build_def_use): Likewise.
16331 * regstat.c (dump_reg_info): Likewise.
16332 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
16333 (find_reloads, find_reloads_subreg_address): Likewise.
16334 * reload1.c (eliminate_regs_1): Likewise.
16335 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
16336 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
16337 (simplify_binary_operation_1, simplify_subreg): Likewise.
16338 * targhooks.c (default_function_arg_padding): Likewise.
16339 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
16340 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
16341 (verify_gimple_assign_ternary): Likewise.
16342 * tree-inline.c (estimate_move_cost): Likewise.
16343 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
16344 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
16345 (get_address_cost_ainc): Likewise.
16346 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
16347 (vect_supportable_dr_alignment): Likewise.
16348 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
16349 (vectorizable_reduction): Likewise.
16350 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
16351 (vectorizable_operation, vectorizable_load): Likewise.
16352 * tree.c (build_same_sized_truth_vector_type): Likewise.
16353 * valtrack.c (cleanup_auto_inc_dec): Likewise.
16354 * var-tracking.c (emit_note_insn_var_location): Likewise.
16355 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
16356 (ADDR_VEC_ALIGN): Likewise.
16358 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16359 Alan Hayward <alan.hayward@arm.com>
16360 David Sherwood <david.sherwood@arm.com>
16362 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
16364 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
16365 or if measurement_type is polynomial.
16366 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
16367 * combine.c (make_extraction): Likewise.
16368 * dse.c (find_shift_sequence): Likewise.
16369 * dwarf2out.c (mem_loc_descriptor): Likewise.
16370 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
16371 (extract_bit_field, extract_low_bits): Likewise.
16372 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
16373 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
16374 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
16375 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
16376 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
16377 * reload.c (find_reloads): Likewise.
16378 * reload1.c (alter_reg): Likewise.
16379 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
16380 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
16381 * tree-if-conv.c (predicate_mem_writes): Likewise.
16382 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
16383 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
16384 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
16385 * valtrack.c (dead_debug_insert_temp): Likewise.
16386 * varasm.c (mergeable_constant_section): Likewise.
16387 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
16389 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16390 Alan Hayward <alan.hayward@arm.com>
16391 David Sherwood <david.sherwood@arm.com>
16393 * expr.c (expand_assignment): Cope with polynomial mode sizes
16394 when assigning to a CONCAT.
16396 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16397 Alan Hayward <alan.hayward@arm.com>
16398 David Sherwood <david.sherwood@arm.com>
16400 * machmode.h (mode_precision): Change from unsigned short to
16402 (mode_to_precision): Return a poly_uint16 rather than an unsigned
16404 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
16405 or if measurement_type is not polynomial.
16406 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
16407 in which the mode is already known to be a scalar_int_mode.
16408 * genmodes.c (emit_mode_precision): Change the type of mode_precision
16409 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
16411 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
16412 for GET_MODE_PRECISION.
16413 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
16414 for GET_MODE_PRECISION.
16415 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
16417 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
16418 (expand_field_assignment, make_extraction): Likewise.
16419 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
16420 (get_last_value): Likewise.
16421 * convert.c (convert_to_integer_1): Likewise.
16422 * cse.c (cse_insn): Likewise.
16423 * expr.c (expand_expr_real_1): Likewise.
16424 * lra-constraints.c (simplify_operand_subreg): Likewise.
16425 * optabs-query.c (can_atomic_load_p): Likewise.
16426 * optabs.c (expand_atomic_load): Likewise.
16427 (expand_atomic_store): Likewise.
16428 * ree.c (combine_reaching_defs): Likewise.
16429 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
16430 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
16431 * tree.h (type_has_mode_precision_p): Likewise.
16432 * ubsan.c (instrument_si_overflow): Likewise.
16434 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16435 Alan Hayward <alan.hayward@arm.com>
16436 David Sherwood <david.sherwood@arm.com>
16438 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
16439 polynomial numbers of units.
16440 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
16441 (valid_vector_subparts_p): New function.
16442 (build_vector_type): Remove temporary shim and take the number
16443 of units as a poly_uint64 rather than an int.
16444 (build_opaque_vector_type): Take the number of units as a
16445 poly_uint64 rather than an int.
16446 * tree.c (build_vector_from_ctor): Handle polynomial
16447 TYPE_VECTOR_SUBPARTS.
16448 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
16449 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
16450 (build_vector_from_val): If the number of units is variable,
16451 use build_vec_duplicate_cst for constant operands and
16452 VEC_DUPLICATE_EXPR otherwise.
16453 (make_vector_type): Remove temporary is_constant ().
16454 (build_vector_type, build_opaque_vector_type): Take the number of
16455 units as a poly_uint64 rather than an int.
16456 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
16458 * cfgexpand.c (expand_debug_expr): Likewise.
16459 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
16460 (store_constructor, expand_expr_real_1): Likewise.
16461 (const_scalar_mask_from_tree): Likewise.
16462 * fold-const-call.c (fold_const_reduction): Likewise.
16463 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
16464 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
16465 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
16466 (fold_relational_const): Likewise.
16467 (native_interpret_vector): Likewise. Change the size from an
16468 int to an unsigned int.
16469 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
16470 TYPE_VECTOR_SUBPARTS.
16471 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
16472 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
16473 duplicating a non-constant operand into a variable-length vector.
16474 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
16475 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
16476 * ipa-icf.c (sem_variable::equals): Likewise.
16477 * match.pd: Likewise.
16478 * omp-simd-clone.c (simd_clone_subparts): Likewise.
16479 * print-tree.c (print_node): Likewise.
16480 * stor-layout.c (layout_type): Likewise.
16481 * targhooks.c (default_builtin_vectorization_cost): Likewise.
16482 * tree-cfg.c (verify_gimple_comparison): Likewise.
16483 (verify_gimple_assign_binary): Likewise.
16484 (verify_gimple_assign_ternary): Likewise.
16485 (verify_gimple_assign_single): Likewise.
16486 * tree-pretty-print.c (dump_generic_node): Likewise.
16487 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
16488 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
16489 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
16490 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
16491 (vect_shift_permute_load_chain): Likewise.
16492 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
16493 (expand_vector_condition, optimize_vector_constructor): Likewise.
16494 (lower_vec_perm, get_compute_type): Likewise.
16495 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
16496 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
16497 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
16498 (vect_recog_mask_conversion_pattern): Likewise.
16499 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
16500 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
16501 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
16502 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
16503 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
16504 (vectorizable_shift, vectorizable_operation, vectorizable_store)
16505 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
16506 (supportable_widening_operation): Likewise.
16507 (supportable_narrowing_operation): Likewise.
16508 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
16510 * varasm.c (output_constant): Likewise.
16512 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16513 Alan Hayward <alan.hayward@arm.com>
16514 David Sherwood <david.sherwood@arm.com>
16516 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
16517 so that both the length == 3 and length != 3 cases set up their
16518 own permute vectors. Add comments explaining why we know the
16519 number of elements is constant.
16520 (vect_permute_load_chain): Likewise.
16522 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16523 Alan Hayward <alan.hayward@arm.com>
16524 David Sherwood <david.sherwood@arm.com>
16526 * machmode.h (mode_nunits): Change from unsigned char to
16528 (ONLY_FIXED_SIZE_MODES): New macro.
16529 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
16530 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
16531 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
16533 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
16534 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
16535 or if measurement_type is not polynomial.
16536 * genmodes.c (ZERO_COEFFS): New macro.
16537 (emit_mode_nunits_inline): Make mode_nunits_inline return a
16539 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
16540 Use ZERO_COEFFS when emitting initializers.
16541 * data-streamer.h (bp_pack_poly_value): New function.
16542 (bp_unpack_poly_value): Likewise.
16543 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
16544 for GET_MODE_NUNITS.
16545 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
16546 for GET_MODE_NUNITS.
16547 * tree.c (make_vector_type): Remove temporary shim and make
16548 the real function take the number of units as a poly_uint64
16549 rather than an int.
16550 (build_vector_type_for_mode): Handle polynomial nunits.
16551 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
16552 * emit-rtl.c (const_vec_series_p_1): Likewise.
16553 (gen_rtx_CONST_VECTOR): Likewise.
16554 * fold-const.c (test_vec_duplicate_folding): Likewise.
16555 * genrecog.c (validate_pattern): Likewise.
16556 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
16557 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
16558 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
16559 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
16560 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
16561 * rtlanal.c (subreg_get_info): Likewise.
16562 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
16563 (vect_grouped_load_supported): Likewise.
16564 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
16565 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
16566 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
16567 (simplify_const_unary_operation, simplify_binary_operation_1)
16568 (simplify_const_binary_operation, simplify_ternary_operation)
16569 (test_vector_ops_duplicate, test_vector_ops): Likewise.
16570 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
16571 instead of CONST_VECTOR_NUNITS.
16572 * varasm.c (output_constant_pool_2): Likewise.
16573 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
16574 explicit-encoded elements in the XVEC for variable-length vectors.
16576 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16578 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
16580 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16581 Alan Hayward <alan.hayward@arm.com>
16582 David Sherwood <david.sherwood@arm.com>
16584 * coretypes.h (fixed_size_mode): Declare.
16585 (fixed_size_mode_pod): New typedef.
16586 * builtins.h (target_builtins::x_apply_args_mode)
16587 (target_builtins::x_apply_result_mode): Change type to
16588 fixed_size_mode_pod.
16589 * builtins.c (apply_args_size, apply_result_size, result_vector)
16590 (expand_builtin_apply_args_1, expand_builtin_apply)
16591 (expand_builtin_return): Update accordingly.
16593 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16595 * cse.c (hash_rtx_cb): Hash only the encoded elements.
16596 * cselib.c (cselib_hash_rtx): Likewise.
16597 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
16598 CONST_VECTOR encoding.
16600 2018-01-03 Jakub Jelinek <jakub@redhat.com>
16601 Jeff Law <law@redhat.com>
16604 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
16605 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
16606 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
16607 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
16610 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
16611 explicitly probe *sp in a noreturn function if there were any callee
16612 register saves or frame pointer is needed.
16614 2018-01-03 Jakub Jelinek <jakub@redhat.com>
16617 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
16618 BLKmode for ternary, binary or unary expressions.
16621 * var-tracking.c (delete_vta_debug_insn): New inline function.
16622 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
16623 insns from get_insns () to NULL instead of each bb separately.
16624 Use delete_vta_debug_insn. No longer static.
16625 (vt_debug_insns_local, variable_tracking_main_1): Adjust
16626 delete_vta_debug_insns callers.
16627 * rtl.h (delete_vta_debug_insns): Declare.
16628 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
16629 instead of variable_tracking_main.
16631 2018-01-03 Martin Sebor <msebor@redhat.com>
16633 PR tree-optimization/83603
16634 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
16635 arguments past the endof the argument list in functions declared
16636 without a prototype.
16637 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
16638 Avoid checking when arguments are null.
16640 2018-01-03 Martin Sebor <msebor@redhat.com>
16643 * doc/extend.texi (attribute const): Fix a typo.
16644 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
16645 issuing -Wsuggest-attribute for void functions.
16647 2018-01-03 Martin Sebor <msebor@redhat.com>
16649 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
16650 offset_int::from instead of wide_int::to_shwi.
16651 (maybe_diag_overlap): Remove assertion.
16652 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
16653 * gimple-ssa-sprintf.c (format_directive): Same.
16654 (parse_directive): Same.
16655 (sprintf_dom_walker::compute_format_length): Same.
16656 (try_substitute_return_value): Same.
16658 2018-01-03 Jeff Law <law@redhat.com>
16660 PR middle-end/83654
16661 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
16662 non-constant residual for zero at runtime and avoid probing in
16663 that case. Reorganize code for trailing problem to mirror handling
16666 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
16668 PR tree-optimization/83501
16669 * tree-ssa-strlen.c (get_string_cst): New.
16670 (handle_char_store): Call get_string_cst.
16672 2018-01-03 Martin Liska <mliska@suse.cz>
16674 PR tree-optimization/83593
16675 * tree-ssa-strlen.c: Include tree-cfg.h.
16676 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
16677 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
16678 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
16680 (strlen_dom_walker::before_dom_children): Call
16681 gimple_purge_dead_eh_edges. Dump tranformation with details
16683 (strlen_dom_walker::before_dom_children): Update call by adding
16684 new argument cleanup_eh.
16685 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
16687 2018-01-03 Martin Liska <mliska@suse.cz>
16690 * cif-code.def (VARIADIC_THUNK): New enum value.
16691 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
16694 2018-01-03 Jan Beulich <jbeulich@suse.com>
16696 * sse.md (mov<mode>_internal): Tighten condition for when to use
16697 vmovdqu<ssescalarsize> for TI and OI modes.
16699 2018-01-03 Jakub Jelinek <jakub@redhat.com>
16701 Update copyright years.
16703 2018-01-03 Martin Liska <mliska@suse.cz>
16706 * ipa-visibility.c (function_and_variable_visibility): Skip
16707 functions with noipa attribure.
16709 2018-01-03 Jakub Jelinek <jakub@redhat.com>
16711 * gcc.c (process_command): Update copyright notice dates.
16712 * gcov-dump.c (print_version): Ditto.
16713 * gcov.c (print_version): Ditto.
16714 * gcov-tool.c (print_version): Ditto.
16715 * gengtype.c (create_file): Ditto.
16716 * doc/cpp.texi: Bump @copying's copyright year.
16717 * doc/cppinternals.texi: Ditto.
16718 * doc/gcc.texi: Ditto.
16719 * doc/gccint.texi: Ditto.
16720 * doc/gcov.texi: Ditto.
16721 * doc/install.texi: Ditto.
16722 * doc/invoke.texi: Ditto.
16724 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16726 * vector-builder.h (vector_builder::m_full_nelts): Change from
16727 unsigned int to poly_uint64.
16728 (vector_builder::full_nelts): Update prototype accordingly.
16729 (vector_builder::new_vector): Likewise.
16730 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
16731 (vector_builder::operator ==): Likewise.
16732 (vector_builder::finalize): Likewise.
16733 * int-vector-builder.h (int_vector_builder::int_vector_builder):
16734 Take the number of elements as a poly_uint64 rather than an
16736 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
16737 from unsigned int to poly_uint64.
16738 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
16739 (vec_perm_indices::new_vector): Likewise.
16740 (vec_perm_indices::length): Likewise.
16741 (vec_perm_indices::nelts_per_input): Likewise.
16742 (vec_perm_indices::input_nelts): Likewise.
16743 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
16744 number of elements per input as a poly_uint64 rather than an
16745 unsigned int. Use the original encoding for variable-length
16746 vectors, rather than clamping each individual element.
16747 For the second and subsequent elements in each pattern,
16748 clamp the step and base before clamping their sum.
16749 (vec_perm_indices::series_p): Handle polynomial element counts.
16750 (vec_perm_indices::all_in_range_p): Likewise.
16751 (vec_perm_indices_to_tree): Likewise.
16752 (vec_perm_indices_to_rtx): Likewise.
16753 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
16754 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
16755 (tree_vector_builder::new_binary_operation): Handle polynomial
16756 element counts. Return false if we need to know the number
16757 of elements at compile time.
16758 * fold-const.c (fold_vec_perm): Punt if the number of elements
16759 isn't known at compile time.
16761 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16763 * vec-perm-indices.h (vec_perm_builder): Change element type
16764 from HOST_WIDE_INT to poly_int64.
16765 (vec_perm_indices::element_type): Update accordingly.
16766 (vec_perm_indices::clamp): Handle polynomial element_types.
16767 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
16768 (vec_perm_indices::all_in_range_p): Likewise.
16769 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
16771 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
16772 polynomial vec_perm_indices element types.
16773 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
16774 * fold-const.c (fold_vec_perm): Likewise.
16775 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
16776 * tree-vect-generic.c (lower_vec_perm): Likewise.
16777 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
16778 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
16779 element type to HOST_WIDE_INT.
16781 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16782 Alan Hayward <alan.hayward@arm.com>
16783 David Sherwood <david.sherwood@arm.com>
16785 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
16786 rather than an int. Use plus_constant.
16787 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
16788 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
16790 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16791 Alan Hayward <alan.hayward@arm.com>
16792 David Sherwood <david.sherwood@arm.com>
16794 * calls.c (emit_call_1, expand_call): Change struct_value_size from
16795 a HOST_WIDE_INT to a poly_int64.
16797 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16798 Alan Hayward <alan.hayward@arm.com>
16799 David Sherwood <david.sherwood@arm.com>
16801 * calls.c (load_register_parameters): Cope with polynomial
16802 mode sizes. Require a constant size for BLKmode parameters
16803 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
16804 forces a parameter to be padded at the lsb end in order to
16805 fill a complete number of words, require the parameter size
16806 to be ordered wrt UNITS_PER_WORD.
16808 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16809 Alan Hayward <alan.hayward@arm.com>
16810 David Sherwood <david.sherwood@arm.com>
16812 * reload1.c (spill_stack_slot_width): Change element type
16813 from unsigned int to poly_uint64_pod.
16814 (alter_reg): Treat mode sizes as polynomial.
16816 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16817 Alan Hayward <alan.hayward@arm.com>
16818 David Sherwood <david.sherwood@arm.com>
16820 * reload.c (complex_word_subreg_p): New function.
16821 (reload_inner_reg_of_subreg, push_reload): Use it.
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 * lra-constraints.c (process_alt_operands): Reject matched
16828 operands whose sizes aren't ordered.
16829 (match_reload): Refer to this check here.
16831 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16832 Alan Hayward <alan.hayward@arm.com>
16833 David Sherwood <david.sherwood@arm.com>
16835 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
16836 that the mode size is in the set {1, 2, 4, 8, 16}.
16838 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16839 Alan Hayward <alan.hayward@arm.com>
16840 David Sherwood <david.sherwood@arm.com>
16842 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
16843 Use plus_constant instead of gen_rtx_PLUS.
16845 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16846 Alan Hayward <alan.hayward@arm.com>
16847 David Sherwood <david.sherwood@arm.com>
16849 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
16850 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
16851 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
16852 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
16853 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
16854 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
16855 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
16856 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
16857 * config/i386/i386.c (ix86_push_rounding): ...this new function.
16858 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
16860 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
16861 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
16862 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
16863 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
16864 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
16865 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
16866 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
16867 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
16868 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
16869 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
16871 * expr.c (emit_move_resolve_push): Treat the input and result
16872 of PUSH_ROUNDING as a poly_int64.
16873 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
16874 (emit_push_insn): Likewise.
16875 * lra-eliminations.c (mark_not_eliminable): Likewise.
16876 * recog.c (push_operand): Likewise.
16877 * reload1.c (elimination_effects): Likewise.
16878 * rtlanal.c (nonzero_bits1): Likewise.
16879 * calls.c (store_one_arg): Likewise. Require the padding to be
16880 known at compile time.
16882 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16883 Alan Hayward <alan.hayward@arm.com>
16884 David Sherwood <david.sherwood@arm.com>
16886 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
16887 Use plus_constant instead of gen_rtx_PLUS.
16889 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16890 Alan Hayward <alan.hayward@arm.com>
16891 David Sherwood <david.sherwood@arm.com>
16893 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
16894 rather than an int.
16896 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16897 Alan Hayward <alan.hayward@arm.com>
16898 David Sherwood <david.sherwood@arm.com>
16900 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
16901 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
16902 via stack temporaries. Treat the mode size as polynomial too.
16904 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16905 Alan Hayward <alan.hayward@arm.com>
16906 David Sherwood <david.sherwood@arm.com>
16908 * expr.c (expand_expr_real_2): When handling conversions involving
16909 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
16910 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
16911 as a poly_uint64 too.
16913 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16914 Alan Hayward <alan.hayward@arm.com>
16915 David Sherwood <david.sherwood@arm.com>
16917 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
16919 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16920 Alan Hayward <alan.hayward@arm.com>
16921 David Sherwood <david.sherwood@arm.com>
16923 * combine.c (can_change_dest_mode): Handle polynomial
16924 REGMODE_NATURAL_SIZE.
16925 * expmed.c (store_bit_field_1): Likewise.
16926 * expr.c (store_constructor): Likewise.
16927 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
16928 and polynomial REGMODE_NATURAL_SIZE.
16929 (gen_lowpart_common): Likewise.
16930 * reginfo.c (record_subregs_of_mode): Likewise.
16931 * rtlanal.c (read_modify_subreg_p): Likewise.
16933 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16934 Alan Hayward <alan.hayward@arm.com>
16935 David Sherwood <david.sherwood@arm.com>
16937 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
16938 numbers of elements.
16940 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16941 Alan Hayward <alan.hayward@arm.com>
16942 David Sherwood <david.sherwood@arm.com>
16944 * match.pd: Cope with polynomial numbers of vector elements.
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 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
16951 in a POINTER_PLUS_EXPR.
16953 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16954 Alan Hayward <alan.hayward@arm.com>
16955 David Sherwood <david.sherwood@arm.com>
16957 * omp-simd-clone.c (simd_clone_subparts): New function.
16958 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
16959 (ipa_simd_modify_function_body): Likewise.
16961 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16962 Alan Hayward <alan.hayward@arm.com>
16963 David Sherwood <david.sherwood@arm.com>
16965 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
16966 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
16967 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
16968 (expand_vector_condition, vector_element): Likewise.
16969 (subparts_gt): New function.
16970 (get_compute_type): Use subparts_gt.
16971 (count_type_subparts): Delete.
16972 (expand_vector_operations_1): Use subparts_gt instead of
16973 count_type_subparts.
16975 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16976 Alan Hayward <alan.hayward@arm.com>
16977 David Sherwood <david.sherwood@arm.com>
16979 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
16980 (vect_compile_time_alias): ...this new function. Do the calculation
16981 on poly_ints rather than trees.
16982 (vect_prune_runtime_alias_test_list): Update call accordingly.
16984 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16985 Alan Hayward <alan.hayward@arm.com>
16986 David Sherwood <david.sherwood@arm.com>
16988 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
16990 (vect_schedule_slp_instance): Likewise.
16992 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16993 Alan Hayward <alan.hayward@arm.com>
16994 David Sherwood <david.sherwood@arm.com>
16996 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
16997 constant and extern definitions for variable-length vectors.
16998 (vect_get_constant_vectors): Note that the number of units
16999 is known to be constant.
17001 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17002 Alan Hayward <alan.hayward@arm.com>
17003 David Sherwood <david.sherwood@arm.com>
17005 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
17006 of units as polynomial. Choose between WIDE and NARROW based
17009 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17010 Alan Hayward <alan.hayward@arm.com>
17011 David Sherwood <david.sherwood@arm.com>
17013 * tree-vect-stmts.c (simd_clone_subparts): New function.
17014 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
17016 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17017 Alan Hayward <alan.hayward@arm.com>
17018 David Sherwood <david.sherwood@arm.com>
17020 * tree-vect-stmts.c (vectorizable_call): Treat the number of
17021 vectors as polynomial. Use build_index_vector for
17022 IFN_GOMP_SIMD_LANE.
17024 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17025 Alan Hayward <alan.hayward@arm.com>
17026 David Sherwood <david.sherwood@arm.com>
17028 * tree-vect-stmts.c (get_load_store_type): Treat the number of
17029 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
17030 for variable-length vectors.
17031 (vectorizable_mask_load_store): Treat the number of units as
17032 polynomial, asserting that it is constant if the condition has
17033 already been enforced.
17034 (vectorizable_store, vectorizable_load): Likewise.
17036 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17037 Alan Hayward <alan.hayward@arm.com>
17038 David Sherwood <david.sherwood@arm.com>
17040 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
17041 of units as polynomial. Punt if we can't tell at compile time
17042 which vector contains the final result.
17044 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17045 Alan Hayward <alan.hayward@arm.com>
17046 David Sherwood <david.sherwood@arm.com>
17048 * tree-vect-loop.c (vectorizable_induction): Treat the number
17049 of units as polynomial. Punt on SLP inductions. Use an integer
17050 VEC_SERIES_EXPR for variable-length integer reductions. Use a
17051 cast of such a series for variable-length floating-point
17054 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17055 Alan Hayward <alan.hayward@arm.com>
17056 David Sherwood <david.sherwood@arm.com>
17058 * tree.h (build_index_vector): Declare.
17059 * tree.c (build_index_vector): New function.
17060 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
17061 of units as polynomial, forcibly converting it to a constant if
17062 vectorizable_reduction has already enforced the condition.
17063 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
17064 to create a {1,2,3,...} vector.
17065 (vectorizable_reduction): Treat the number of units as polynomial.
17066 Choose vectype_in based on the largest scalar element size rather
17067 than the smallest number of units. Enforce the restrictions
17070 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17071 Alan Hayward <alan.hayward@arm.com>
17072 David Sherwood <david.sherwood@arm.com>
17074 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
17075 number of units as polynomial.
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 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
17082 * target.def (autovectorize_vector_sizes): Return the vector sizes
17083 by pointer, using vector_sizes rather than a bitmask.
17084 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
17085 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
17086 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
17088 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
17089 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
17090 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
17091 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
17092 * omp-general.c (omp_max_vf): Likewise.
17093 * omp-low.c (omp_clause_aligned_alignment): Likewise.
17094 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
17095 * tree-vect-loop.c (vect_analyze_loop): Likewise.
17096 * tree-vect-slp.c (vect_slp_bb): Likewise.
17097 * doc/tm.texi: Regenerate.
17098 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
17100 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
17101 the vector size as a poly_uint64 rather than an unsigned int.
17102 (current_vector_size): Change from an unsigned int to a poly_uint64.
17103 (get_vectype_for_scalar_type): Update accordingly.
17104 * tree.h (build_truth_vector_type): Take the size and number of
17105 units as a poly_uint64 rather than an unsigned int.
17106 (build_vector_type): Add a temporary overload that takes
17107 the number of units as a poly_uint64 rather than an unsigned int.
17108 * tree.c (make_vector_type): Likewise.
17109 (build_truth_vector_type): Take the number of units as a poly_uint64
17110 rather than an unsigned int.
17112 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17113 Alan Hayward <alan.hayward@arm.com>
17114 David Sherwood <david.sherwood@arm.com>
17116 * target.def (get_mask_mode): Take the number of units and length
17117 as poly_uint64s rather than unsigned ints.
17118 * targhooks.h (default_get_mask_mode): Update accordingly.
17119 * targhooks.c (default_get_mask_mode): Likewise.
17120 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
17121 * doc/tm.texi: Regenerate.
17123 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17124 Alan Hayward <alan.hayward@arm.com>
17125 David Sherwood <david.sherwood@arm.com>
17127 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
17128 * omp-general.c (omp_max_vf): Likewise.
17129 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
17130 (expand_omp_simd): Handle polynomial safelen.
17131 * omp-low.c (omplow_simd_context): Add a default constructor.
17132 (omplow_simd_context::max_vf): Change from int to poly_uint64.
17133 (lower_rec_simd_input_clauses): Update accordingly.
17134 (lower_rec_input_clauses): Likewise.
17136 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17137 Alan Hayward <alan.hayward@arm.com>
17138 David Sherwood <david.sherwood@arm.com>
17140 * tree-vectorizer.h (vect_nunits_for_cost): New function.
17141 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
17142 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
17143 (vect_analyze_slp_cost): Likewise.
17144 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
17145 (vect_model_load_cost): Likewise.
17147 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17148 Alan Hayward <alan.hayward@arm.com>
17149 David Sherwood <david.sherwood@arm.com>
17151 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
17152 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
17153 from an unsigned int * to a poly_uint64_pod *.
17154 (calculate_unrolling_factor): New function.
17155 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
17157 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17158 Alan Hayward <alan.hayward@arm.com>
17159 David Sherwood <david.sherwood@arm.com>
17161 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
17162 from an unsigned int to a poly_uint64.
17163 (_loop_vec_info::slp_unrolling_factor): Likewise.
17164 (_loop_vec_info::vectorization_factor): Change from an int
17166 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
17167 (vect_get_num_vectors): New function.
17168 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
17169 (vect_get_num_copies): Use vect_get_num_vectors.
17170 (vect_analyze_data_ref_dependences): Change max_vf from an int *
17171 to an unsigned int *.
17172 (vect_analyze_data_refs): Change min_vf from an int * to a
17174 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
17175 than an unsigned HOST_WIDE_INT.
17176 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
17177 (vect_analyze_data_ref_dependence): Change max_vf from an int *
17178 to an unsigned int *.
17179 (vect_analyze_data_ref_dependences): Likewise.
17180 (vect_compute_data_ref_alignment): Handle polynomial vf.
17181 (vect_enhance_data_refs_alignment): Likewise.
17182 (vect_prune_runtime_alias_test_list): Likewise.
17183 (vect_shift_permute_load_chain): Likewise.
17184 (vect_supportable_dr_alignment): Likewise.
17185 (dependence_distance_ge_vf): Take the vectorization factor as a
17186 poly_uint64 rather than an unsigned HOST_WIDE_INT.
17187 (vect_analyze_data_refs): Change min_vf from an int * to a
17189 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
17190 vfm1 as a poly_uint64 rather than an int. Make the same change
17191 for the returned bound_scalar.
17192 (vect_gen_vector_loop_niters): Handle polynomial vf.
17193 (vect_do_peeling): Likewise. Update call to
17194 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
17195 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
17197 * tree-vect-loop.c (vect_determine_vectorization_factor)
17198 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
17199 (vect_get_known_peeling_cost): Likewise.
17200 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
17201 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
17202 (vect_transform_loop): Likewise. Use the lowest possible VF when
17203 updating the upper bounds of the loop.
17204 (vect_min_worthwhile_factor): Make static. Return an unsigned int
17205 rather than an int.
17206 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
17207 polynomial unroll factors.
17208 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
17209 (vect_make_slp_decision): Likewise.
17210 (vect_supported_load_permutation_p): Likewise, and polynomial
17212 (vect_analyze_slp_cost): Handle polynomial vf.
17213 (vect_slp_analyze_node_operations): Likewise.
17214 (vect_slp_analyze_bb_1): Likewise.
17215 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
17216 than an unsigned HOST_WIDE_INT.
17217 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
17218 (vectorizable_load): Handle polynomial vf.
17219 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
17221 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
17223 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17224 Alan Hayward <alan.hayward@arm.com>
17225 David Sherwood <david.sherwood@arm.com>
17227 * match.pd: Handle bit operations involving three constants
17228 and try to fold one pair.
17230 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17232 * tree-vect-loop-manip.c: Include gimple-fold.h.
17233 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
17234 niters_maybe_zero parameters. Handle other cases besides a step of 1.
17235 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
17236 Add a path that uses a step of VF instead of 1, but disable it
17238 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
17239 and niters_no_overflow parameters. Update calls to
17240 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
17241 Create a new SSA name if the latter choses to use a ste other
17242 than zero, and return it via niters_vector_mult_vf_var.
17243 * tree-vect-loop.c (vect_transform_loop): Update calls to
17244 vect_do_peeling, vect_gen_vector_loop_niters and
17245 slpeel_make_loop_iterate_ntimes.
17246 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
17247 (vect_gen_vector_loop_niters): Update declarations after above changes.
17249 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
17251 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
17252 128-bit round to integer instructions.
17253 (ceil<mode>2): Likewise.
17254 (btrunc<mode>2): Likewise.
17255 (round<mode>2): Likewise.
17257 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
17259 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
17260 unaligned VSX load/store on P8/P9.
17261 (expand_block_clear): Allow the use of unaligned VSX
17262 load/store on P8/P9.
17264 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
17266 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
17268 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
17269 swap associated with both a load and a store.
17271 2018-01-02 Andrew Waterman <andrew@sifive.com>
17273 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
17274 * config/riscv/riscv.md (clear_cache): Use it.
17276 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
17278 * web.c: Remove out-of-date comment.
17280 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17282 * expr.c (fixup_args_size_notes): Check that any existing
17283 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
17284 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
17285 (emit_single_push_insn): ...here.
17287 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17289 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
17290 (const_vector_encoded_nelts): New function.
17291 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
17292 (const_vector_int_elt, const_vector_elt): Declare.
17293 * emit-rtl.c (const_vector_int_elt_1): New function.
17294 (const_vector_elt): Likewise.
17295 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
17296 of CONST_VECTOR_ELT.
17298 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17300 * expr.c: Include rtx-vector-builder.h.
17301 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
17302 directly on the tree encoding.
17303 (const_vector_from_tree): Likewise.
17304 * optabs.c: Include rtx-vector-builder.h.
17305 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
17306 sequence of "u" values.
17307 * vec-perm-indices.c: Include rtx-vector-builder.h.
17308 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
17309 directly on the vec_perm_indices encoding.
17311 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17313 * doc/rtl.texi (const_vector): Describe new encoding scheme.
17314 * Makefile.in (OBJS): Add rtx-vector-builder.o.
17315 * rtx-vector-builder.h: New file.
17316 * rtx-vector-builder.c: Likewise.
17317 * rtl.h (rtx_def::u2): Add a const_vector field.
17318 (CONST_VECTOR_NPATTERNS): New macro.
17319 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
17320 (CONST_VECTOR_DUPLICATE_P): Likewise.
17321 (CONST_VECTOR_STEPPED_P): Likewise.
17322 (CONST_VECTOR_ENCODED_ELT): Likewise.
17323 (const_vec_duplicate_p): Check for a duplicated vector encoding.
17324 (unwrap_const_vec_duplicate): Likewise.
17325 (const_vec_series_p): Check for a non-duplicated vector encoding.
17326 Say that the function only returns true for integer vectors.
17327 * emit-rtl.c: Include rtx-vector-builder.h.
17328 (gen_const_vec_duplicate_1): Delete.
17329 (gen_const_vector): Call gen_const_vec_duplicate instead of
17330 gen_const_vec_duplicate_1.
17331 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
17332 (gen_const_vec_duplicate): Use rtx_vector_builder.
17333 (gen_const_vec_series): Likewise.
17334 (gen_rtx_CONST_VECTOR): Likewise.
17335 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
17336 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
17337 Build a new vector rather than modifying a CONST_VECTOR in-place.
17338 (handle_special_swappables): Update call accordingly.
17339 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
17340 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
17341 Build a new vector rather than modifying a CONST_VECTOR in-place.
17342 (handle_special_swappables): Update call accordingly.
17344 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17346 * simplify-rtx.c (simplify_const_binary_operation): Use
17347 CONST_VECTOR_ELT instead of XVECEXP.
17349 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17351 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
17352 the selector elements to be different from the data elements
17353 if the selector is a VECTOR_CST.
17354 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
17355 ssizetype for the selector.
17357 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17359 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
17360 before testing each element individually.
17361 * tree-vect-generic.c (lower_vec_perm): Likewise.
17363 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17365 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
17366 * selftest-run-tests.c (selftest::run_tests): Call it.
17367 * vector-builder.h (vector_builder::operator ==): New function.
17368 (vector_builder::operator !=): Likewise.
17369 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
17370 (vec_perm_indices::all_from_input_p): New function.
17371 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
17372 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
17373 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
17374 instead of reading the VECTOR_CST directly. Detect whether both
17375 vector inputs are the same before constructing the vec_perm_indices,
17376 and update the number of inputs argument accordingly. Use the
17377 utility functions added above. Only construct sel2 if we need to.
17379 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17381 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
17382 the broadcast of the low byte.
17383 (expand_mult_highpart): Use an explicit encoding for the permutes.
17384 * optabs-query.c (can_mult_highpart_p): Likewise.
17385 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
17386 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
17387 (vectorizable_bswap): Likewise.
17388 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
17389 explicit encoding for the power-of-2 permutes.
17390 (vect_permute_store_chain): Likewise.
17391 (vect_grouped_load_supported): Likewise.
17392 (vect_permute_load_chain): Likewise.
17394 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17396 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
17397 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
17398 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
17399 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
17400 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
17401 (vect_gen_perm_mask_any): Likewise.
17403 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17405 * int-vector-builder.h: New file.
17406 * vec-perm-indices.h: Include int-vector-builder.h.
17407 (vec_perm_indices): Redefine as an int_vector_builder.
17408 (auto_vec_perm_indices): Delete.
17409 (vec_perm_builder): Redefine as a stand-alone class.
17410 (vec_perm_indices::vec_perm_indices): New function.
17411 (vec_perm_indices::clamp): Likewise.
17412 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
17413 (vec_perm_indices::new_vector): New function.
17414 (vec_perm_indices::new_expanded_vector): Update for new
17415 vec_perm_indices class.
17416 (vec_perm_indices::rotate_inputs): New function.
17417 (vec_perm_indices::all_in_range_p): Operate directly on the
17418 encoded form, without computing elided elements.
17419 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
17420 encoding. Update for new vec_perm_indices class.
17421 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
17422 the given vec_perm_builder.
17423 (expand_vec_perm_var): Update vec_perm_builder constructor.
17424 (expand_mult_highpart): Use vec_perm_builder instead of
17425 auto_vec_perm_indices.
17426 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
17427 vec_perm_indices instead of auto_vec_perm_indices. Use a single
17428 or double series encoding as appropriate.
17429 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
17430 vec_perm_indices instead of auto_vec_perm_indices.
17431 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
17432 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
17433 (vect_permute_store_chain): Likewise.
17434 (vect_grouped_load_supported): Likewise.
17435 (vect_permute_load_chain): Likewise.
17436 (vect_shift_permute_load_chain): Likewise.
17437 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
17438 (vect_transform_slp_perm_load): Likewise.
17439 (vect_schedule_slp_instance): Likewise.
17440 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
17441 (vectorizable_mask_load_store): Likewise.
17442 (vectorizable_bswap): Likewise.
17443 (vectorizable_store): Likewise.
17444 (vectorizable_load): Likewise.
17445 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
17446 vec_perm_indices instead of auto_vec_perm_indices. Use
17447 tree_to_vec_perm_builder to read the vector from a tree.
17448 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
17449 vec_perm_builder instead of a vec_perm_indices.
17450 (have_whole_vector_shift): Use vec_perm_builder and
17451 vec_perm_indices instead of auto_vec_perm_indices. Leave the
17452 truncation to calc_vec_perm_mask_for_shift.
17453 (vect_create_epilog_for_reduction): Likewise.
17454 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
17455 from auto_vec_perm_indices to vec_perm_indices.
17456 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
17457 instead of changing individual elements.
17458 (aarch64_vectorize_vec_perm_const): Use new_vector to install
17459 the vector in d.perm.
17460 * config/arm/arm.c (expand_vec_perm_d::perm): Change
17461 from auto_vec_perm_indices to vec_perm_indices.
17462 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
17463 instead of changing individual elements.
17464 (arm_vectorize_vec_perm_const): Use new_vector to install
17465 the vector in d.perm.
17466 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
17467 Update vec_perm_builder constructor.
17468 (rs6000_expand_interleave): Likewise.
17469 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
17470 (rs6000_expand_interleave): Likewise.
17472 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17474 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
17475 to qimode could truncate the indices.
17476 * optabs.c (expand_vec_perm_var): Likewise.
17478 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17480 * Makefile.in (OBJS): Add vec-perm-indices.o.
17481 * vec-perm-indices.h: New file.
17482 * vec-perm-indices.c: Likewise.
17483 * target.h (vec_perm_indices): Replace with a forward class
17485 (auto_vec_perm_indices): Move to vec-perm-indices.h.
17486 * optabs.h: Include vec-perm-indices.h.
17487 (expand_vec_perm): Delete.
17488 (selector_fits_mode_p, expand_vec_perm_var): Declare.
17489 (expand_vec_perm_const): Declare.
17490 * target.def (vec_perm_const_ok): Replace with...
17491 (vec_perm_const): ...this new hook.
17492 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
17493 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
17494 * doc/tm.texi: Regenerate.
17495 * optabs.def (vec_perm_const): Delete.
17496 * doc/md.texi (vec_perm_const): Likewise.
17497 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
17498 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
17499 expand_vec_perm for constant permutation vectors. Assert that
17500 the mode of variable permutation vectors is the integer equivalent
17501 of the mode that is being permuted.
17502 * optabs-query.h (selector_fits_mode_p): Declare.
17503 * optabs-query.c: Include vec-perm-indices.h.
17504 (selector_fits_mode_p): New function.
17505 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
17506 is defined, instead of checking whether the vec_perm_const_optab
17507 exists. Use targetm.vectorize.vec_perm_const instead of
17508 targetm.vectorize.vec_perm_const_ok. Check whether the indices
17509 fit in the vector mode before using a variable permute.
17510 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
17511 vec_perm_indices instead of an rtx.
17512 (expand_vec_perm): Replace with...
17513 (expand_vec_perm_const): ...this new function. Take the selector
17514 as a vec_perm_indices rather than an rtx. Also take the mode of
17515 the selector. Update call to shift_amt_for_vec_perm_mask.
17516 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
17517 Use vec_perm_indices::new_expanded_vector to expand the original
17518 selector into bytes. Check whether the indices fit in the vector
17519 mode before using a variable permute.
17520 (expand_vec_perm_var): Make global.
17521 (expand_mult_highpart): Use expand_vec_perm_const.
17522 * fold-const.c: Includes vec-perm-indices.h.
17523 * tree-ssa-forwprop.c: Likewise.
17524 * tree-vect-data-refs.c: Likewise.
17525 * tree-vect-generic.c: Likewise.
17526 * tree-vect-loop.c: Likewise.
17527 * tree-vect-slp.c: Likewise.
17528 * tree-vect-stmts.c: Likewise.
17529 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
17531 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
17532 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
17533 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
17534 (aarch64_vectorize_vec_perm_const): ...this new function.
17535 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
17536 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
17537 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
17538 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
17539 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
17540 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
17541 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
17543 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
17544 check for NEON modes.
17545 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
17546 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
17547 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
17548 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
17550 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
17551 the old VEC_PERM_CONST conditions.
17552 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
17553 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
17554 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
17555 (ia64_vectorize_vec_perm_const_ok): Merge into...
17556 (ia64_vectorize_vec_perm_const): ...this new function.
17557 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
17558 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
17559 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
17560 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
17561 * config/mips/mips.c (mips_expand_vec_perm_const)
17562 (mips_vectorize_vec_perm_const_ok): Merge into...
17563 (mips_vectorize_vec_perm_const): ...this new function.
17564 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
17565 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
17566 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
17567 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
17568 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
17569 (rs6000_expand_vec_perm_const): Delete.
17570 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
17572 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
17573 (altivec_expand_vec_perm_const_le): Take each operand individually.
17574 Operate on constant selectors rather than rtxes.
17575 (altivec_expand_vec_perm_const): Likewise. Update call to
17576 altivec_expand_vec_perm_const_le.
17577 (rs6000_expand_vec_perm_const): Delete.
17578 (rs6000_vectorize_vec_perm_const_ok): Delete.
17579 (rs6000_vectorize_vec_perm_const): New function.
17580 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
17581 an element count and rtx array.
17582 (rs6000_expand_extract_even): Update call accordingly.
17583 (rs6000_expand_interleave): Likewise.
17584 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
17585 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
17586 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
17587 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
17588 (rs6000_expand_vec_perm_const): Delete.
17589 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
17590 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
17591 (altivec_expand_vec_perm_const_le): Take each operand individually.
17592 Operate on constant selectors rather than rtxes.
17593 (altivec_expand_vec_perm_const): Likewise. Update call to
17594 altivec_expand_vec_perm_const_le.
17595 (rs6000_expand_vec_perm_const): Delete.
17596 (rs6000_vectorize_vec_perm_const_ok): Delete.
17597 (rs6000_vectorize_vec_perm_const): New function. Remove stray
17598 reference to the SPE evmerge intructions.
17599 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
17600 an element count and rtx array.
17601 (rs6000_expand_extract_even): Update call accordingly.
17602 (rs6000_expand_interleave): Likewise.
17603 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
17604 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
17606 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
17608 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17610 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
17611 vector mode and that that mode matches the mode of the data
17613 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
17614 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
17615 directly using expand_vec_perm_1 when forcing selectors into
17617 (expand_vec_perm_var): New function, split out from expand_vec_perm.
17619 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17621 * optabs-query.h (can_vec_perm_p): Delete.
17622 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
17623 * optabs-query.c (can_vec_perm_p): Split into...
17624 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
17625 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
17626 particular selector is valid.
17627 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
17628 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
17629 (vect_grouped_load_supported): Likewise.
17630 (vect_shift_permute_load_chain): Likewise.
17631 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
17632 (vect_transform_slp_perm_load): Likewise.
17633 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
17634 (vectorizable_bswap): Likewise.
17635 (vect_gen_perm_mask_checked): Likewise.
17636 * fold-const.c (fold_ternary_loc): Likewise. Don't take
17637 implementations of variable permutation vectors into account
17638 when deciding which selector to use.
17639 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
17640 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
17641 with a false third argument.
17642 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
17643 to test whether the constant selector is valid and can_vec_perm_var_p
17644 to test whether a variable selector is valid.
17646 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17648 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
17649 * optabs-query.c (can_vec_perm_p): Likewise.
17650 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
17651 instead of vec_perm_indices.
17652 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
17653 (vect_gen_perm_mask_checked): Likewise,
17654 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
17655 (vect_gen_perm_mask_checked): Likewise,
17657 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17659 * optabs-query.h (qimode_for_vec_perm): Declare.
17660 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
17661 (qimode_for_vec_perm): ...this new function.
17662 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
17664 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
17666 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
17667 does not have a conditional at the top.
17669 2018-01-02 Richard Biener <rguenther@suse.de>
17671 * ipa-inline.c (big_speedup_p): Fix expression.
17673 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
17676 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
17679 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
17683 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
17684 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
17685 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
17686 cond_taken_branch_cost 3->4.
17688 2018-01-01 Jakub Jelinek <jakub@redhat.com>
17690 PR tree-optimization/83581
17691 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
17692 TODO_cleanup_cfg if any changes have been made.
17694 PR middle-end/83608
17695 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
17696 convert_modes if target mode has the right side, but different mode
17699 PR middle-end/83609
17700 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
17701 last argument when extracting from CONCAT. If either from_real or
17702 from_imag is NULL, use expansion through memory. If result is not
17703 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
17704 the parts directly to inner mode, if even that fails, use expansion
17707 PR middle-end/83623
17708 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
17709 check for bswap in mode rather than HImode and use that in expand_unop
17712 Copyright (C) 2018 Free Software Foundation, Inc.
17714 Copying and distribution of this file, with or without modification,
17715 are permitted in any medium without royalty provided the copyright
17716 notice and this notice are preserved.