1 2018-06-25 Richard Biener <rguenther@suse.de>
3 * tree-vectorizer.h (struct vec_info_shared): New structure
4 with parts split out from struct vec_info and loop_nest from
6 (struct vec_info): Adjust accordingly.
7 (struct _loop_vec_info): Likewise.
8 (LOOP_VINFO_LOOP_NEST): Adjust.
9 (LOOP_VINFO_DATAREFS): Likewise.
10 (LOOP_VINFO_DDRS): Likewise.
11 (struct _bb_vec_info): Likewise.
12 (BB_VINFO_DATAREFS): Likewise.
13 (BB_VINFO_DDRS): Likewise.
14 (struct _stmt_vec_info): Add dr_aux member.
15 (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
16 (DR_MISALIGNMENT_UNINITIALIZED): New.
17 (set_dr_misalignment): Adjust.
18 (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
19 (vect_analyze_loop): Adjust prototype.
20 (vect_analyze_loop_form): Likewise.
21 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
22 Compute dependences lazily.
23 (vect_record_base_alignments): Use shared datarefs/ddrs.
24 (vect_verify_datarefs_alignment): Likewise.
25 (vect_analyze_data_refs_alignment): Likewise.
26 (vect_analyze_data_ref_accesses): Likewise.
27 (vect_analyze_data_refs): Likewise.
28 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
29 constructor parameter for shared part.
30 (vect_analyze_loop_form): Pass in shared part and adjust.
31 (vect_analyze_loop_2): Pass in storage for the number of
32 stmts. Move loop nest finding to the caller. Compute
34 (vect_analyze_loop): Pass in shared part.
35 (vect_transform_loop): Verify shared datarefs are unchanged.
36 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
37 constructor parameter for shared part.
38 (vect_slp_analyze_bb_1): Pass in shared part and adjust.
39 (vect_slp_bb): Verify shared datarefs are unchanged before
41 * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
43 (new_stmt_vec_info): Initialize DR_AUX misalignment to
44 DR_MISALIGNMENT_UNINITIALIZED.
45 * tree-vectorizer.c (vec_info::vec_info): Add constructor
46 parameter for shared part.
47 (vec_info::~vec_info): Adjust.
48 (vec_info_shared::vec_info_shared): New.
49 (vec_info_shared::~vec_info_shared): Likewise.
50 (vec_info_shared::save_datarefs): Likewise.
51 (vec_info_shared::check_datarefs): Likewise.
52 (try_vectorize_loop_1): Construct shared part live for analyses
53 of a single loop for multiple vector sizes.
54 * tree-parloops.c (gather_scalar_reductions): Adjust.
56 2018-06-25 Richard Biener <rguenther@suse.de>
58 * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
59 DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
60 (vect_analyze_data_refs): Remove similar code from here and
63 2018-06-25 Richard Biener <rguenther@suse.de>
65 * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
66 for reverse storage order accesses rather than asserting
67 they cannot happen here.
69 2018-06-25 Tom de Vries <tdevries@suse.de>
72 * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
73 Use data16 instead of .byte for insn prefix.
75 2018-06-25 Andreas Krebbel <krebbel@linux.ibm.com>
78 * parser.c (make_char_string_pack): Pass this literal chars
79 through cpp_interpret_string.
80 (cp_parser_userdef_numeric_literal): Check the result of
81 make_char_string_pack.
83 2018-06-24 Maya Rashish <coypu@sdf.org>
85 * ginclude/stddef.h: Simplify conditions around avoiding
86 re-definition of __size_t.
88 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
90 * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
91 unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
93 2018-06-22 Maya Rashish <coypu@sdf.org>
95 * doc/invoke.texi (mno-fancy-math-387): Update for changes
96 made to OpenBSD and NetBSD through the years.
98 2018-06-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
100 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
101 behavior of vec_pack (vector double, vector double) to match
102 behavior of vec_float2 (vector double, vector double).
104 2018-06-22 Olivier Hainque <hainque@adacore.com>
106 * gimplify.c (gimplify_function_tree): Prevent creation
107 of a trampoline for the address of the current function
108 passed to entry/exit instrumentation hooks.
110 2018-06-22 Aaron Sawdey <acsawdey@linux.ibm.com>
113 * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
116 2018-06-22 Martin Liska <mliska@suse.cz>
118 PR tree-optimization/86263
119 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
120 Bail out if is_enabled is false.
121 * tree-switch-conversion.h (jump_table_cluster::is_enabled):
123 (jump_table_cluster::is_enabled): New function.
125 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
127 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
128 BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
129 * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
130 (lto_input_ts_binfo_tree_pointers): Likewise.
131 * tree-streamer-out.c (streamer_write_tree_bitfields,
132 write_ts_binfo_tree_pointers): Likewise.
133 * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
135 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
137 * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
139 2018-06-22 Martin Liska <mliska@suse.cz>
141 * symbol-summary.h (get): Make it pure and inline move
142 functionality from ::get function.
143 (get): Remove and inline into ::get and ::get_create.
144 (get_create): Move code from ::get function.
146 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
149 * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
150 -x assembler-with-cpp.
152 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
154 * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
155 _FILE_OFFSET_BITS=64 for C++.
157 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
159 * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
160 conversion insn that shows up when pr85657-3.c is compiled using
161 IEEE 128-bit long double.
162 (neg<mode>2_internal): Use the correct mode to check whether the
163 mode is IBM extended.
164 * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
165 multiply and divide external functions from being created more
168 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
170 * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
172 (rtl_merge_blocks): Likewise. Do not emit a NOP if the location of
173 the edge can be forwarded.
174 (cfg_layout_merge_blocks): Likewise.
176 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
178 * except.c (finish_eh_generation): Commit edge insertions only after
179 the EH edges have been redirected from post-landing to landing pads.
181 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
183 * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
184 create_tmp_var_for to create the FRAME decl.
185 (finalize_nesting_tree_1): Do not unchain the FRAME decl.
187 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
189 * tree-inline.c (copy_edges_for_bb): Minor tweak.
190 (maybe_move_debug_stmts_to_successors): Also reset the locus of the
191 debug statement when resetting its value.
192 (expand_call_inline): Copy the locus of the call onto the assignment
193 of the return value, if any. Use local variable in more cases.
195 2018-06-21 Martin Liska <mliska@suse.cz>
197 * ipa-pure-const.c (propagate_nothrow): Use
198 funct_state_summaries->get.
199 (dump_malloc_lattice): Likewise.
200 (propagate_malloc): Likewise.
202 2018-06-21 Richard Biener <rguenther@suse.de>
204 * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
205 comment. Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
206 * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
208 * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
209 BLOCK_ABSTRACT_ORIGIN unconditionally.
211 2018-06-21 David Malcolm <dmalcolm@redhat.com>
213 * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
215 * ipa-reference.c (ipa_reference_c_finalize): Delete
216 ipa_ref_opt_sum_summaries and set it to NULL.
218 2018-06-21 Tom de Vries <tdevries@suse.de>
220 PR tree-optimization/85859
221 * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
222 test with comment from bb_no_side_effects_p.
224 2018-06-21 Richard Biener <rguenther@suse.de>
226 PR tree-optimization/86232
227 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
228 max for constant niter.
230 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
232 * config/aarch64/aarch64-simd.md
233 (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
235 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
237 * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
238 Make opernads of the unspec commutative.
240 2018-06-21 Richard Biener <rguenther@suse.de>
242 * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
243 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
244 Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
245 (vect_analyze_data_ref_dependence): Re-order checks to deal with
247 (vect_record_base_alignments): Do not record base alignment
248 for gathers or scatters.
249 (vect_compute_data_ref_alignment): Drop return value that is always
250 true. Bail out early for gathers or scatters.
251 (vect_enhance_data_refs_alignment): Bail out early for gathers
253 (vect_find_same_alignment_drs): Likewise.
254 (vect_analyze_data_refs_alignment): Remove dead code.
255 (vect_slp_analyze_and_verify_node_alignment): Likewise.
256 (vect_analyze_data_refs): For possible gathers or scatters do
257 not create an alternate DR, just check their possible validity
258 and mark them. Adjust DECL_NONALIASED handling to not rely
260 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
261 update inits of gathers or scatters.
262 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
263 Also copy gather/scatter flag to pattern vinfo.
265 2018-06-20 Kelvin Nilsen <kelvin@gcc.gnu.org>
267 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
268 behavior of vec_packsu (vector unsigned long long, vector unsigned
269 long long) to match behavior of vec_packs with same signature.
271 2018-06-20 Chung-Lin Tang <cltang@codesourcery.com>
272 Thomas Schwinge <thomas@codesourcery.com>
273 Cesar Philippidis <cesar@codesourcery.com>
275 * gimplify.c (gimplify_scan_omp_clauses): Add support for
276 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
277 (gimplify_adjust_omp_clauses): Likewise.
278 (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
279 support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
280 (gimplify_omp_target_update): Update handling of acc update and
282 * omp-low.c (install_var_field): Remove unused parameter
283 base_pointers_restrict.
284 (scan_sharing_clauses): Remove base_pointers_restrict parameter.
285 Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
287 (omp_target_base_pointers_restrict_p): Delete.
288 (scan_omp_target): Update call to scan_sharing_clauses.
289 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
291 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
292 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
293 (convert_local_omp_clauses): Likewise.
294 * tree-pretty-print.c (dump_omp_clause): Likewise.
295 * tree.c (omp_clause_num_ops): Add entries for OMP_CLAUSE_{IF_PRESENT,
297 (omp_clause_code_name): Likewise.
299 2018-06-20 Jakub Jelinek <jakub@redhat.com>
302 * var-tracking.c (use_narrower_mode_test): Check if shift amount can
305 PR tree-optimization/86231
306 * tree-vrp.c (union_ranges): For ( [ ) ] or ( )[ ] range and
307 anti-range don't overwrite *vr0min before using it to compute *vr0max.
309 2018-06-20 Tom de Vries <tdevries@suse.de>
311 PR tree-optimization/86097
312 * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
313 iv type if signedness of iv type is not the same as that of *nit.
315 2018-06-20 Jakub Jelinek <jakub@redhat.com>
317 * cfgrtl.c (rtl_verify_edges): Formatting fix. If bb->preds has any
318 EDGE_EH edges, verify they are all EDGE_EH.
320 2018-06-20 Maya Rashish <coypu@sdf.org>
322 * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
324 2018-06-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
326 * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
327 * config/aarch64/aarch64.c (xgene1_tunings): Add
328 AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
329 (aarch64_mode_valid_for_sched_fusion_p):
331 (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
332 * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
334 * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
336 (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
337 * config/aarch64/iterators.md (VQ2): New mode iterator.
339 2018-06-20 Martin Liska <mliska@suse.cz>
341 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
342 Change default ratio from 10 to 8.
344 2018-06-20 Martin Liska <mliska@suse.cz>
346 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
348 (bit_test_cluster::find_bit_tests): Likewise.
349 (switch_decision_tree::analyze_switch_statement): Find clusters.
350 * tree-switch-conversion.h (struct jump_table_cluster): Document
353 2018-06-20 Martin Liska <mliska@suse.cz>
355 * tree-switch-conversion.c (switch_conversion::collect):
356 Record m_uniq property.
357 (switch_conversion::expand): Bail out for special conditions.
358 (group_cluster::~group_cluster): New.
359 (group_cluster::group_cluster): Likewise.
360 (group_cluster::dump): Likewise.
361 (jump_table_cluster::emit): New.
362 (switch_decision_tree::fix_phi_operands_for_edges): New.
363 (struct case_node): Remove struct.
364 (jump_table_cluster::can_be_handled): New.
365 (case_values_threshold): Moved to header.
366 (reset_out_edges_aux): Likewise.
367 (jump_table_cluster::is_beneficial): New.
368 (bit_test_cluster::can_be_handled): Likewise.
369 (add_case_node): Remove.
370 (bit_test_cluster::is_beneficial): New.
371 (case_bit_test::cmp): New.
372 (bit_test_cluster::emit): New.
373 (expand_switch_as_decision_tree_p): Remove.
374 (bit_test_cluster::hoist_edge_and_branch_if_true): New.
375 (fix_phi_operands_for_edge): Likewise.
376 (switch_decision_tree::analyze_switch_statement): New.
377 (compute_cases_per_edge): Move ...
378 (switch_decision_tree::compute_cases_per_edge): ... here.
379 (try_switch_expansion): Likewise.
380 (switch_decision_tree::try_switch_expansion): Likewise.
381 (record_phi_operand_mapping): Likewise.
382 (switch_decision_tree::record_phi_operand_mapping): Likewise.
383 (emit_case_decision_tree): Likewise.
384 (switch_decision_tree::emit): Likewise.
385 (balance_case_nodes): Likewise.
386 (switch_decision_tree::balance_case_nodes): Likewise.
387 (dump_case_nodes): Likewise.
388 (switch_decision_tree::dump_case_nodes): Likewise.
389 (emit_jump): Likewise.
390 (switch_decision_tree::emit_jump): Likewise.
391 (emit_cmp_and_jump_insns): Likewise.
392 (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
393 (emit_case_nodes): Likewise.
394 (switch_decision_tree::emit_case_nodes): Likewise.
395 (conditional_probability): Remove.
396 * tree-switch-conversion.h (enum cluster_type): New.
398 (struct cluster): Likewise.
399 (cluster::cluster): Likewise.
400 (struct simple_cluster): Likewise.
401 (simple_cluster::simple_cluster): Likewise.
402 (struct group_cluster): Likewise.
403 (struct jump_table_cluster): Likewise.
404 (struct bit_test_cluster): Likewise.
405 (struct min_cluster_item): Likewise.
406 (struct case_tree_node): Likewise.
407 (case_tree_node::case_tree_node): Likewise.
408 (jump_table_cluster::case_values_threshold): Likewise.
409 (struct case_bit_test): Likewise.
410 (struct switch_decision_tree): Likewise.
411 (struct switch_conversion): Likewise.
412 (switch_decision_tree::reset_out_edges_aux): Likewise.
414 2018-06-20 Martin Liska <mliska@suse.cz>
416 * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
417 (hoist_edge_and_branch_if_true): Likewise.
418 (expand_switch_using_bit_tests_p): Likewise.
419 (struct case_bit_test): Likewise.
420 (case_bit_test_cmp): Likewise.
421 (emit_case_bit_tests): Likewise.
422 (switch_conversion::switch_conversion): New class.
423 (struct switch_conv_info): Remove old struct.
424 (collect_switch_conv_info): More to ...
425 (switch_conversion::collect): ... this.
426 (check_range): Likewise.
427 (switch_conversion::check_range): Likewise.
428 (check_all_empty_except_final): Likewise.
429 (switch_conversion::check_all_empty_except_final): Likewise.
430 (check_final_bb): Likewise.
431 (switch_conversion::check_final_bb): Likewise.
432 (create_temp_arrays): Likewise.
433 (switch_conversion::create_temp_arrays): Likewise.
434 (free_temp_arrays): Likewise.
435 (gather_default_values): Likewise.
436 (switch_conversion::gather_default_values): Likewise.
437 (build_constructors): Likewise.
438 (switch_conversion::build_constructors): Likewise.
439 (constructor_contains_same_values_p): Likewise.
440 (switch_conversion::contains_same_values_p): Likewise.
441 (array_value_type): Likewise.
442 (switch_conversion::array_value_type): Likewise.
443 (build_one_array): Likewise.
444 (switch_conversion::build_one_array): Likewise.
445 (build_arrays): Likewise.
446 (switch_conversion::build_arrays): Likewise.
447 (gen_def_assigns): Likewise.
448 (switch_conversion::gen_def_assigns): Likewise.
449 (prune_bbs): Likewise.
450 (switch_conversion::prune_bbs): Likewise.
451 (fix_phi_nodes): Likewise.
452 (switch_conversion::fix_phi_nodes): Likewise.
453 (gen_inbound_check): Likewise.
454 (switch_conversion::gen_inbound_check): Likewise.
455 (process_switch): Use the newly created class.
456 (switch_conversion::expand): New.
457 (switch_conversion::~switch_conversion): New.
458 * tree-switch-conversion.h: New file.
460 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
462 * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
463 tree-vect-patterns.c.
464 * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
465 (vect_recog_dot_prod_pattern): Use it. Remove the type_in argument.
466 (vect_recog_sad_pattern): Likewise.
467 (vect_recog_widen_sum_pattern): Likewise.
468 (vect_recog_pow_pattern): Likewise. Check for a null vectype.
469 (vect_recog_widen_shift_pattern): Remove the type_in argument.
470 (vect_recog_rotate_pattern): Likewise.
471 (vect_recog_mult_pattern): Likewise.
472 (vect_recog_vector_vector_shift_pattern): Likewise.
473 (vect_recog_divmod_pattern): Likewise.
474 (vect_recog_mixed_size_cond_pattern): Likewise.
475 (vect_recog_bool_pattern): Likewise.
476 (vect_recog_mask_conversion_pattern): Likewise.
477 (vect_try_gather_scatter_pattern): Likewise.
478 (vect_recog_widen_mult_pattern): Likewise. Check for a null vectype.
479 (vect_recog_over_widening_pattern): Likewise.
480 (vect_recog_gather_scatter_pattern): Likewise.
481 (vect_recog_func_ptr): Move from tree-vectorizer.h
482 (vect_vect_recog_func_ptrs): Move further down the file.
483 (vect_recog_func): Likewise. Remove the third argument.
484 (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
485 (vect_pattern_recog_1): Expect the pattern function to do any
486 necessary target tests. Also expect it to provide a vector type.
487 Remove the type_in handling.
489 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
491 * tree-vect-patterns.c (vect_pattern_detected): New function.
492 (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
493 (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
494 (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
495 (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
496 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
497 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
498 (vect_recog_mask_conversion_pattern)
499 (vect_try_gather_scatter_pattern): Likewise.
501 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
503 * tree-vect-patterns.c (vect_get_internal_def): New function.
504 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
505 (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
506 (search_type_for_mask_1): Use it.
508 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
510 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
511 redundant WIDEN_SUM_EXPR handling.
512 (vect_recog_sad_pattern): Likewise.
514 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
516 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
517 redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
518 (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
519 (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
520 (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
522 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
524 * tree-vect-stmts.c (vectorizable_call): Make sure that we
525 use the stmt_vec_info of the original bb statement for the
526 new zero assignment, even if the call is part of a pattern.
528 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
530 * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
531 that the sequence is attached to the original statement rather
532 than the pattern statement.
533 * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
534 PATTERN_DEF_SEQ from the original statement rather than
535 the main pattern statement.
536 * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
537 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
538 (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
540 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
542 * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
543 definition statements before the early exit for statements that aren't
545 * tree-vect-loop.c (vect_transform_loop_stmt): New function,
547 (vect_transform_loop): ...here. Process pattern definition
548 statements without first checking whether the main pattern
549 statement is live or relevant.
551 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
553 * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
554 -O0 if the locus represent UNKNOWN_LOCATION but have different values.
556 2018-06-19 Aaron Sawdey <acsawdey@linux.ibm.com>
558 * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
559 TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
560 (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
561 (expand_block_compare): Change select_block_compare_mode call.
562 (expand_strncmp_align_check): Use new functions, fix comment.
563 (emit_final_str_compare_gpr): New function.
564 (expand_strn_compare): Refactor and clean up code.
565 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
567 2018-06-19 Tony Reix <tony.reix@atos.com>
568 Damien Bergamini <damien.bergamini@atos.com>
569 David Edelsohn <dje.gcc@gmail.com>
571 * collect2.c (static_obj): New variable.
572 (static_libs): New variable.
573 (is_in_list): Uncomment declaration.
574 (main): Track AIX libraries linked statically.
575 (is_in_list): Uncomment definition.
576 (scan_prog_file): Don't add AIX shared libraries initializer
577 to constructor list if linking statically.
579 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
581 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
583 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
585 2018-06-19 Jan Hubicka <hubicka@ucw.cz>
587 * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
590 2018-06-19 Martin Liska <mliska@suse.cz>
592 * config/i386/i386.c (ix86_can_inline_p): Do not use
593 ipa_fn_summaries::get_create.
594 * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
596 (devirtualization_time_bonus): Likewise.
597 (ipcp_propagate_stage): Likewise.
598 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
599 (edge_set_predicate): Likewise.
600 (evaluate_conditions_for_known_args): Likewise.
601 (evaluate_properties_for_edge): Likewise.
602 (ipa_call_summary::reset): Tranform to ...
603 (ipa_call_summary::~ipa_call_summary): ... this.
604 (ipa_fn_summary::reset): Transform to ...
605 (ipa_fn_summary::~ipa_fn_summary): ... this.
606 (ipa_fn_summary_t::remove): Rename to ...
607 (ipa_fn_summary_t::remove_callees): ... this.
608 (ipa_fn_summary_t::duplicate): Use placement new
609 instead of memory copy.
610 (ipa_call_summary_t::duplicate): Likewise.
611 (ipa_call_summary_t::remove): Remove.
612 (dump_ipa_call_summary): Change get_create to get.
613 (ipa_dump_fn_summary): Dump only when summary exists.
614 (analyze_function_body): Use symbol_summary::get instead
616 (compute_fn_summary): Likewise.
617 (estimate_edge_devirt_benefit): Likewise.
618 (estimate_edge_size_and_time): Likewise.
619 (inline_update_callee_summaries): Likewise.
620 (remap_edge_change_prob): Likewise.
621 (remap_edge_summaries): Likewise.
622 (ipa_merge_fn_summary_after_inlining): Likewise.
623 (write_ipa_call_summary): Likewise.
624 (ipa_fn_summary_write): Likewise.
625 (ipa_free_fn_summary): Likewise.
626 * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
627 (struct ipa_call_summary): Likewise.
628 * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
630 * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
631 (estimate_size_after_inlining): Likewise.
632 (estimate_growth): Likewise.
633 (growth_likely_positive): Likewise.
634 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
635 (inline_call): Likewise.
636 * ipa-inline.c (caller_growth_limits): Likewise.
637 (can_inline_edge_p): Likewise.
638 (can_inline_edge_by_limits_p): Likewise.
639 (compute_uninlined_call_time): Likewise.
640 (compute_inlined_call_time): Likewise.
641 (want_inline_small_function_p): Likewise.
642 (edge_badness): Likewise.
643 (update_caller_keys): Likewise.
644 (update_callee_keys): Likewise.
645 (inline_small_functions): Likewise.
646 (inline_to_all_callers_1): Likewise.
647 (dump_overall_stats): Likewise.
648 (early_inline_small_functions): Likewise.
649 (early_inliner): Likewise.
650 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
651 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
652 * ipa-pure-const.c (malloc_candidate_p): Likewise.
653 * ipa-split.c (execute_split_functions): Likewise.
654 * symbol-summary.h: Likewise.
655 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
657 2018-06-19 Richard Biener <rguenther@suse.de>
659 * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
660 (vectorize_loops): ... here. Fix dbgcnt handling.
661 (try_vectorize_loop): Wrap try_vectorize_loop_1.
663 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
666 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
667 ieee128 argument takes up only one (vector) register, not two (floating
670 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
672 * gimplify.c (gimplify_init_constructor): Really never clear for an
673 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
675 2018-06-19 Richard Biener <rguenther@suse.de>
677 PR tree-optimization/86179
678 * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
679 after failed recognition.
681 2018-06-18 Martin Sebor <msebor@redhat.com>
684 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
685 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
686 Handle integer subtraction.
687 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
688 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
690 2018-06-18 David Malcolm <dmalcolm@redhat.com>
692 * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
693 param from rtx to rtx_insn *.
694 * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
696 (frv_ifcvt_modify_insn): Likwise.
697 (frv_ifcvt_modify_final): Likwise for local "existing_insn",
698 adding an as_a <rtx_insn *> cast. Likewise for local "insn".
699 * config/mips/mips.c (r10k_insert_cache_barriers): Add an
700 as_a <rtx_insn *> cast to local "unprotected_region" once
701 it's been established that it's not NULL or pc_rtx.
702 * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
703 param "sethi" from rtx to rtx_insn *.
704 (nds32_group_float_insns): Likewise for param "insn".
705 * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
707 (vax_output_int_subtract): Likewise.
708 * config/vax/vax.c (vax_output_int_add): Likewise for param
710 (vax_output_int_subtract): Likewise.
711 * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
712 (emit_pattern_after): Likewise for param "after".
713 (emit_insn_after): Likewise.
714 (emit_jump_insn_after): Likewise.
715 (emit_call_insn_after): Likewise.
716 (emit_debug_insn_after): Likewise.
717 (emit_pattern_before): Likewise for param "before".
718 (emit_insn_before): Likewise.
719 (emit_jump_insn_before): Likewise.
720 * final.c (get_insn_template): Likewise for param "insn", removing
722 * output.h (get_insn_template): Likewise for 2nd param.
723 * rtl.h (emit_insn_before): Likewise.
724 (emit_jump_insn_before): Likewise.
725 (emit_debug_insn_before_noloc): Likewise.
726 (emit_insn_after): Likewise.
727 (emit_jump_insn_after): Likewise.
728 (emit_call_insn_after): Likewise.
729 (emit_debug_insn_after): Likewise.
730 (set_insn_deleted): Likewise for param.
732 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
735 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
736 floating point modes, so that IFmode is numerically greater than
737 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
738 to declare the ordering. This prevents IFmode from being
739 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
740 machine. Include rs6000-modes.h to share the fractional values
741 between genmodes* and the rest of the compiler.
745 * config/rs6000/rs6000-modes.h: New file.
746 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
747 meaning of rs6000_long_double_size so that 126..128 selects an
748 appropriate 128-bit floating point type.
749 (rs6000_option_override_internal): Likewise.
750 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
751 (TARGET_LONG_DOUBLE_128): Change the meaning of
752 rs6000_long_double_size so that 126..128 selects an appropriate
753 128-bit floating point type.
754 (LONG_DOUBLE_TYPE_SIZE): Update comment.
755 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
756 source and destination to match the standard usage.
757 (truncifkf2): Likewise.
758 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
759 ISA 2.07 to use an explicit clobber, instead of passing in a
761 (copysign<mode>3_soft): Likewise.
763 2018-06-18 David Malcolm <dmalcolm@redhat.com>
765 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
766 Replace dump_printf_loc call with DUMP_VECT_SCOPE.
767 (vect_slp_analyze_instance_dependence): Likewise.
768 (vect_enhance_data_refs_alignment): Likewise.
769 (vect_analyze_data_refs_alignment): Likewise.
770 (vect_slp_analyze_and_verify_instance_alignment
771 (vect_analyze_data_ref_accesses): Likewise.
772 (vect_prune_runtime_alias_test_list): Likewise.
773 (vect_analyze_data_refs): Likewise.
774 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
775 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
776 (vect_analyze_scalar_cycles_1): Likewise.
777 (vect_get_loop_niters): Likewise.
778 (vect_analyze_loop_form_1): Likewise.
779 (vect_update_vf_for_slp): Likewise.
780 (vect_analyze_loop_operations): Likewise.
781 (vect_analyze_loop): Likewise.
782 (vectorizable_induction): Likewise.
783 (vect_transform_loop): Likewise.
784 * tree-vect-patterns.c (vect_pattern_recog): Likewise.
785 * tree-vect-slp.c (vect_analyze_slp): Likewise.
786 (vect_make_slp_decision): Likewise.
787 (vect_detect_hybrid_slp): Likewise.
788 (vect_slp_analyze_operations): Likewise.
789 (vect_slp_bb): Likewise.
790 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
791 (vectorizable_bswap): Likewise.
792 (vectorizable_call): Likewise.
793 (vectorizable_simd_clone_call): Likewise.
794 (vectorizable_conversion): Likewise.
795 (vectorizable_assignment): Likewise.
796 (vectorizable_shift): Likewise.
797 (vectorizable_operation): Likewise.
798 * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
800 2018-06-18 Martin Sebor <msebor@redhat.com>
802 PR tree-optimization/81384
803 * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
804 * builtins.c (expand_builtin_strnlen): New function.
805 (expand_builtin): Call it.
806 (fold_builtin_n): Avoid setting TREE_NO_WARNING.
807 * builtins.def (BUILT_IN_STRNLEN): New.
808 * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
809 Warn for bounds in excess of maximum object size.
810 * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
811 single-value ranges. Handle strnlen.
812 (handle_builtin_strlen): Handle strnlen.
813 (strlen_check_and_optimize_stmt): Same.
814 * doc/extend.texi (Other Builtins): Document strnlen.
816 2018-06-18 Maya Rashish <coypu@sdf.org>
818 * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
819 (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
820 (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
822 * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
824 * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
826 2018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
828 * tree.c (escaped_string::escape): Replace cast to char * by
829 const_cast<char *> (unescaped).
831 2018-06-18 Nick Clifton <nickc@redhat.com>
834 * tree.c (escaped_string): New class. Converts an unescaped
835 string into its escaped equivalent.
836 (warn_deprecated_use): Use the new class to convert the
837 deprecation message, if present.
838 (test_escaped_strings): New self test.
839 (test_c_tests): Add test_escaped_strings.
840 * doc/extend.texi (deprecated): Add a note that the
841 deprecation message is affected by the -fmessage-length
842 option, and that control characters will be escaped.
843 (#pragma GCC error): Document this pragma.
844 (#pragma GCC warning): Likewise.
845 * doc/invoke.texi (-fmessage-length): Document this option's
846 effect on the #warning and #error preprocessor directives and
847 the deprecated attribute.
849 2018-06-18 Eric Botcazou <ebotcazou@adacore.com>
851 * tree.c (decl_value_expr_lookup): Revert latest change.
852 (decl_value_expr_insert): Likewise.
854 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
856 * gimplify.c (nonlocal_vlas): Delete.
857 (nonlocal_vla_vars): Likewise.
858 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
860 (gimplify_body): Do not create and destroy nonlocal_vlas.
861 * tree-nested.c: Include diagnostic.h.
862 (use_pointer_in_frame): Tweak.
863 (lookup_field_for_decl): Add assertion and declare the transformation.
864 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
865 internal error when the reference is in a wrong context. Do not
866 create a debug decl by default.
867 (note_nonlocal_block_vlas): Delete.
868 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
869 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
870 create a debug decl by default.
871 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
872 call is in a wrong context.
873 (fixup_vla_decls): New function.
874 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
875 debug variables were created.
876 * tree.c (decl_value_expr_lookup): Add checking assertion.
877 (decl_value_expr_insert): Likewise.
879 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
882 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
883 * tree-scalar-evolution.c (interpret_expr): Likewise.
884 (expression_expensive_p): Likewise.
885 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
886 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
887 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
888 (ssa_defined_by_minus_one_stmt_p): New.
890 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
893 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
894 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
895 * dojump.c (do_jump): Likewise.
896 * expr.c (expand_expr_real_2): Check operand type's sign.
897 * fold-const.c (const_unop): Handle ABSU_EXPR.
898 (fold_abs_const): Likewise.
899 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
900 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
901 (strip_sign_op_1): Likesise.
902 * match.pd: Add new pattern to generate ABSU_EXPR.
903 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
904 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
905 * tree-eh.c (operation_could_trap_helper_p): Likewise.
906 * tree-inline.c (estimate_operator_cost): Likewise.
907 * tree-pretty-print.c (dump_generic_node): Likewise.
908 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
909 * tree.def (ABSU_EXPR): New.
911 2018-06-16 Jakub Jelinek <jakub@redhat.com>
914 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
915 documented as preserved for backward compatibility only.
916 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
918 PR rtl-optimization/86108
919 * bb-reorder.c (create_forwarder_block): Renamed to ...
920 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
921 jump from new landing pad to the second part.
922 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
925 2018-06-15 Jakub Jelinek <jakub@redhat.com>
928 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
929 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
930 Only call store_expr for halves if the mode is the same.
933 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
934 Fix up comment formatting.
936 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
938 * typed-splay-tree.h (typed_splay_tree::remove): New function.
939 (typed_splay_tree::closure,
940 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
941 (typed_splay_tree::typed_splay_tree,
942 typed_splay_tree::operator =): Declared private.
943 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
944 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
945 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
946 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
947 typed_splay_tree::splay_tree_splay,
948 typed_splay_tree::splay_tree_foreach_helper,
949 typed_splay_tree::splay_tree_insert,
950 typed_splay_tree::splay_tree_remove,
951 typed_splay_tree::splay_tree_lookup,
952 typed_splay_tree::splay_tree_predecessor,
953 typed_splay_tree::splay_tree_successor,
954 typed_splay_tree::splay_tree_min,
955 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
956 (typed_splay_tree::root, typed_splay_tree::comp,
957 typed_splay_tree::delete_key,
958 typed_splay_tree::delete_value): New data members.
959 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
960 typed_splay_tree::remove.
962 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
964 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
965 -mginv and -mno-ginv to the assembler.
966 * config/mips/mips.opt (-mcrc): New option.
968 * doc/invoke.text (-mcrc): Document.
971 2018-06-15 Nick Clifton <nickc@redhat.com>
974 * tree.c (escaped_string): New class. Converts an unescaped
975 string into its escaped equivalent.
976 (warn_deprecated_use): Use the new class to convert the
977 deprecation message, if present.
978 (test_escaped_strings): New self test.
979 (test_c_tests): Add test_escaped_strings.
980 * doc/extend.texi (deprecated): Add a note that the
981 deprecation message is affected by the -fmessage-length
982 option, and that control characters will be escaped.
983 (#pragma GCC error): Document this pragma.
984 (#pragma GCC warning): Likewise.
985 * doc/invoke.texi (-fmessage-length): Document this option's
986 effect on the #warning and #error preprocessor directives and
987 the deprecated attribute.
989 2018-06-15 Richard Biener <rguenther@suse.de>
991 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
992 here, also noting vector size used.
993 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
994 size used in MSG_OPTIMIZED_LOCATIONS dump.
995 (pass_slp_vectorize::execute): Adjust.
997 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
1000 * config/arc/arc.c (arc_return_address_register): Fix
1003 2018-06-15 Richard Biener <rguenther@suse.de>
1006 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
1007 leave useless conversion stripping to force_gimple_operand_gsi.
1008 (gimplify_build2): Likewise.
1009 (gimplify_build1): Likewise.
1011 2018-06-15 Richard Biener <rguenther@suse.de>
1014 * tree-cfg.c (move_stmt_op): unshare invariant addresses
1015 before adjusting their block.
1017 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
1019 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
1020 multilibs for *-*-rtems*.
1021 * config/riscv/t-rtems: New file.
1023 2018-06-14 Jakub Jelinek <jakub@redhat.com>
1026 * match.pd ((A +- CST1) +- CST2): Punt if last resort
1027 unsigned_type_for returns NULL.
1030 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
1031 subregs of multi-word pseudos unless the float mode has word size.
1033 2018-06-14 Richard Biener <rguenther@suse.de>
1036 * tree-vect-generic.c (build_word_mode_vector_type): Remove
1037 duplicate and harmful type_hash_canon.
1038 * tree.c (type_hash_canon): Assert we didn't find ourselves.
1040 2018-06-14 Richard Biener <rguenther@suse.de>
1043 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
1046 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
1048 * config/rtems.h (STDINT_LONG32): Define.
1050 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
1051 Prachi Godbole <prachi.godbole@imgtec.com>
1053 * config/mips/mips-cpus.def: Define P6600.
1054 * config/mips/mips-tables.opt: Regenerate.
1055 * config/mips/mips.c (mips_ucbranch_type): New enum.
1056 (mips_rtx_cost_data): Add support for P6600.
1057 (mips_issue_rate): Likewise.
1058 (mips_multipass_dfa_lookahead): Likewise.
1059 (mips_avoid_hazard): Likewise.
1060 (mips_reorg_process_insns): Likewise.
1061 (mips_classify_branch_p6600): New function.
1062 * config/mips/mips.h (TUNE_P6600): New define.
1063 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
1064 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
1065 * config/mips/mips.md: Include p6600.md.
1066 (processor): Add p6600.
1067 * config/mips/p6600.md: New file.
1068 * doc/invoke.texi: Add p6600 to supported architectures.
1070 2018-06-13 Martin Sebor <msebor@redhat.com>
1072 PR tree-optimization/86114
1073 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
1075 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
1077 2018-06-13 Richard Biener <rguenther@suse.de>
1079 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
1080 Properly set vector type of the intermediate stmt.
1081 * tree-vect-stmts.c (vectorizable_operation): The destination
1082 var always has vectype_out type.
1084 2018-06-13 Jeff Law <law@redhat.com>
1086 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
1087 integer 0 for argument to print_rtl_with_bb.
1088 (rl78_reorg): Likewise.
1090 2018-06-13 David Malcolm <dmalcolm@redhat.com>
1092 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
1093 from rtx to rtx_insn *.
1094 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
1096 (add_sched_insns_for_speculation): Likewise for local "target",
1097 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
1098 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
1099 from rtx_insn ** to rtx_code_label **.
1100 (reorg_emit_nops): Likewise.
1101 (c6x_reorg): Likewise for local "call_labels".
1102 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
1104 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
1105 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
1106 the loops over LABEL_REFS.
1107 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
1109 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
1110 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
1111 (split_branches): Strengthen local "olabel" from rtx to
1112 rtx_insn *, adding a safe_as_a cast.
1113 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
1115 (add_insn_after): Likewise for first two params.
1116 (add_insn_before): Likewise.
1117 (remove_insn): Likewise for param.
1118 (emit_pattern_before_noloc): Likewise for second and third params.
1119 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
1120 (emit_call_insn_before_noloc): Likewise.
1121 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
1123 (emit_barrier_before): Likewise.
1124 (emit_label_before): Strengthen "label" param from "rtx" to
1125 "rtx_code_label *". Strengthen "before" param from "rtx" to
1127 (emit_insn_after_1): Strengthen "after" param from "rtx" to
1129 (emit_pattern_after_noloc): Likewise.
1130 (emit_insn_after_noloc): Likewise.
1131 (emit_jump_insn_after_noloc): Likewise.
1132 (emit_call_insn_after_noloc): Likewise.
1133 (emit_debug_insn_after_noloc): Likewise.
1134 (emit_barrier_after): Likewise.
1135 (emit_label_after): Likewise for both params.
1136 (emit_pattern_after_setloc): Likewise for "after" param. Convert
1137 "loc" param from "int" to "location_t".
1138 (emit_insn_after_setloc): Likewise.
1139 (emit_jump_insn_after_setloc): Likewise.
1140 (emit_call_insn_after_setloc): Likewise.
1141 (emit_debug_insn_after_setloc): Likewise.
1142 (emit_pattern_before_setloc): Likewise for "before" param. Convert
1143 "loc" param from "int" to "location_t".
1144 (emit_pattern_before): Convert NULL_RTX to NULL.
1145 (emit_insn_before_setloc): Convert "loc" param from "int" to
1147 (emit_jump_insn_before_setloc): Likewise.
1148 (emit_call_insn_before_setloc): Likewise.
1149 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
1150 rtx_insn *. Convert "loc" param from "int" to "location_t".
1151 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
1152 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
1153 Convert 3rd param from "int" to "location_t".
1154 (emit_barrier_before, emit_barrier_after, next_real_insn):
1155 Strengthen param from rtx to rtx_insn *.
1156 (emit_label_before): Strengthen 1st param from "rtx" to
1157 "rtx_code_label *". Strengthen 2nd param from "rtx" to
1159 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
1160 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
1161 Strengthen 2nd param from "rtx" to "rtx_insn *".
1162 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
1163 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
1164 Likewise. Convert 3rd param from "int" to "location_t".
1165 (emit_label_after): Strengthen 1st param from "rtx" to
1167 (next_real_insn, remove_insn): Strengthen param from "rtx" to
1169 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
1170 from "rtx" to "rtx_insn *".
1172 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
1174 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
1175 bodies streamed in with -Q.
1176 * dumpfile.c (dump_files): Add lto-stream-out dump file.
1177 * dumpfile.h (tree_dump_index): Add lto_stream_out.
1178 * gimple-streamer-out.c: Include gimple-pretty-print.h
1179 (output_bb): Dump stmts streamed.
1180 * lto-section-out.c: Include print-tree.h
1181 (lto_begin_section): Dump sections created.
1182 (lto_output_decl_index): Dump decl encoded.
1183 * lto-streamer-out.c: Include print-tree.h
1184 (create_output_block): Dump output block created.
1185 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
1186 (output_function): Dump function output.
1187 (output_constructor): Dump constructor streamed.
1188 (write_global_stream): Output indexes encoded.
1189 (produce_asm_for_decls): Dump streams encoded.
1190 * lto-streamer.c (streamer_dump_file): New global var.
1191 * lto-streamer.h (streamer_dump_file): Declare.
1192 * passes.c (ipa_write_summaries): Initialize streamer dump.
1193 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
1196 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
1199 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
1200 offsets for register save directives. Emit a second batch of save
1201 directives, if need be, when the function accesses prior frames.
1203 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1205 * config/arc/fpu.md (fmasf4): Force operand to register.
1206 (fnmasf4): Likewise.
1208 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1210 * config/arc/arc-protos.h (arc_pad_return): Remove.
1211 * config/arc/arc.c (machine_function): Remove force_short_suffix
1213 (arc_print_operand): Adjust printing of '&'.
1214 (arc_verify_short): Remove conditional printing of short suffix.
1215 (arc_final_prescan_insn): Remove reference to size_reason.
1216 (pad_return): New function.
1217 (arc_reorg): Call pad_return.
1218 (arc_pad_return): Remove.
1219 (arc_init_machine_status): Remove reference to force_short_suffix.
1220 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
1221 (attr length): When attribute iscompact is true force to 2
1222 regardless; in the case of maybe check if we want to force the
1223 instruction to have 4 bytes length.
1224 (nopv): Change it to generate 4 byte long nop as well.
1225 (blockage): New pattern.
1226 (simple_return): Remove call to arc_pad_return.
1227 (p_return_i): Likewise.
1229 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1231 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
1233 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1235 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
1238 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1240 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
1241 for ARC700 and ARCv2.
1243 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
1246 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
1247 operands[2] instead of operands[1].
1250 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
1252 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
1253 case, check whether the outer register overlaps an unallocatable
1254 register, not just whether it fits the required class.
1256 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
1258 * poly-int.h (can_div_trunc_p): Add new overload in which all values
1260 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
1261 (memrefs_conflict_p): Likewise.
1262 (init_alias_analysis): Likewise.
1263 * cfgexpand.c (expand_debug_expr): Likewise.
1264 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
1265 * cse.c (fold_rtx): Likewise.
1266 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
1267 * expr.c (emit_block_move_hints): Likewise.
1268 (clear_storage_hints, push_block, emit_push_insn): Likewise.
1269 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
1270 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
1271 (emit_group_store): Likewise.
1272 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
1273 to read the PRE/POST_MODIFY increment.
1274 * calls.c (store_one_arg): Use strip_offset.
1275 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
1277 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
1279 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
1280 (simplify_binary_operation_1): Extend CONST_INT handling to
1282 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
1283 than a HOST_WIDE_INT.
1284 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
1286 (adjust_mems, add_stores): Update accodingly.
1287 (vt_canonicalize_addr): Track polynomial offsets.
1288 (emit_note_insn_var_location): Likewise.
1289 (vt_add_function_parameter): Likewise.
1290 (vt_initialize): Likewise.
1292 2018-06-12 Jeff Law <law@redhat.com>
1294 * config.gcc (alpha*-*-freebsd*): Remove.
1295 * config/alpha/freebsd.h: Remove.
1297 2018-06-12 David Malcolm <dmalcolm@redhat.com>
1300 * spellcheck-tree.c (levenshtein_distance): Rename to...
1301 (get_edit_distance): ...this, and update for underlying renaming.
1302 * spellcheck-tree.h (levenshtein_distance): Rename to...
1303 (get_edit_distance): ...this.
1304 * spellcheck.c (levenshtein_distance): Rename to...
1305 (get_edit_distance): ...this. Convert from Levenshtein distance
1306 to Damerau-Levenshtein distance by supporting transpositions of
1307 adjacent characters. Rename "v1" to "v_next" and "v0" to
1309 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
1310 (selftest::test_edit_distance_unit_test_oneway): ...this, and
1311 update for underlying renaming.
1312 (selftest::levenshtein_distance_unit_test): Rename to...
1313 (selftest::test_get_edit_distance_unit): ...this, and update for
1314 underlying renaming.
1315 (selftest::test_find_closest_string): Add example from PR 69968
1316 where transposition helps
1317 (selftest::test_metric_conditions): Update for renaming.
1318 (selftest::test_metric_conditions): Likewise.
1319 (selftest::spellcheck_c_tests): Likewise.
1320 * spellcheck.h (levenshtein_distance): Rename both overloads to...
1321 (get_edit_distance): ...this.
1322 (best_match::consider): Update for renaming.
1324 2018-06-12 Martin Sebor <msebor@redhat.com>
1326 PR tree-optimization/85259
1327 * builtins.c (compute_objsize): Handle constant offsets.
1328 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
1329 true iff a warning has been issued.
1330 * gimple.h (gimple_nonartificial_location): New function.
1331 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
1332 gimple_nonartificial_location and handle -Wno-system-headers.
1333 (handle_builtin_stxncpy): Same.
1335 2018-06-12 Martin Sebor <msebor@redhat.com>
1338 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
1340 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
1342 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
1343 BUILTIN_VEC_XST entries for pointer to double and long long.
1345 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
1348 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
1350 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
1353 2018-06-12 Martin Liska <mliska@suse.cz>
1355 * doc/options.texi: Document IntegerRange.
1357 2018-06-12 Martin Liska <mliska@suse.cz>
1359 * config/i386/i386.opt: Make MPX-related options as Deprecated.
1360 * opt-functions.awk: Handle Deprecated flag.
1361 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
1363 (read_cmdline_option): Report warning for a deprecated option.
1364 * opts.h (struct cl_option): Add new field cl_deprecated.
1365 (CL_ERR_DEPRECATED): New.
1367 2018-06-12 Martin Liska <mliska@suse.cz>
1369 * doc/options.texi: Document Deprecated option flag.
1371 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1373 * config/arc/arc-arch.h (arc_extras): New enum.
1374 (arc_cpu_t):Add field extra.
1375 (arc_cpu_types): Consider the extras.
1376 * config/arc/arc-cpus.def: Add extras info.
1377 * config/arc/arc-opts.h (processor_type): Consider extra field.
1378 * config/arc/arc.c (arc_override_options): Handle extra field.
1380 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1382 * config/arc/arc-arch.h: Update ARC_OPTX macro.
1383 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
1385 * config/arc/arc.c (arc_init): Update pic warning.
1386 (irq_range): Update irq range parsing warnings.
1387 (arc_override_options): Update various warning messages.
1388 (arc_handle_aux_attribute): Likewise.
1390 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
1392 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
1394 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
1396 * doc/sourcebuild.texi: Document usage of line number 0 in verify
1397 compiler messages directives.
1399 2018-06-12 Matthew Fortune <mfortune@gmail.com>
1401 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
1402 * config/mips/mips-tables.opt: Regenerate.
1403 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
1405 * doc/invoke.texi: Document -march=i6500.
1407 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
1409 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
1410 (i6400_gpmul): Add cpu_unit.
1411 (i6400_gpdiv): Likewise.
1412 (i6400_msa_add_d): Update reservations.
1413 (i6400_msa_int_add) Likewise.
1414 (i6400_msa_short_logic3) Likewise.
1415 (i6400_msa_short_logic2) Likewise.
1416 (i6400_msa_short_logic) Likewise.
1417 (i6400_msa_move) Likewise.
1418 (i6400_msa_cmp) Likewise.
1419 (i6400_msa_short_float2) Likewise.
1420 (i6400_msa_div_d) Likewise.
1421 (i6400_msa_long_logic1) Likewise.
1422 (i6400_msa_long_logic2) Likewise.
1423 (i6400_msa_mult) Likewise.
1424 (i6400_msa_long_float2) Likewise.
1425 (i6400_msa_long_float4) Likewise.
1426 (i6400_msa_long_float5) Likewise.
1427 (i6400_msa_long_float8) Likewise.
1428 (i6400_fpu_fadd): Include frint type.
1429 (i6400_fpu_store): New define_insn_reservation.
1430 (i6400_fpu_load): Likewise.
1431 (i6400_fpu_move): Likewise.
1432 (i6400_fpu_fcmp): Likewise.
1433 (i6400_fpu_fmadd): Likewise.
1434 (i6400_int_mult): Include imul3nc type and update reservation.
1435 (i6400_int_div): Include idiv3 type and update reservation.
1436 (i6400_int_load): Update to check type not move_type.
1437 (i6400_int_store): Likewise.
1438 (i6400_int_prefetch): Set zero latency.
1440 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
1442 * gcc.c: Document new %@{...} sequence.
1443 (LINK_COMMAND_SPEC): Use it for the -L switches.
1444 (cpp_unique_options): Use it for the -I switches.
1445 (at_file_argbuf): New global variable.
1446 (in_at_file): Likewise.
1447 (alloc_args): Create at_file_argbuf.
1448 (clear_args): Truncate at_file_argbuf.
1449 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
1450 (open_at_file): New function.
1451 (close_at_file): Likewise.
1452 (create_at_file): Delete.
1453 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
1456 (validate_switches_from_spec): Deal with %@{...} sequence.
1457 (validate_switches): Likewise.
1458 (driver::finalize): Call clear_args.
1460 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
1462 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
1464 2018-06-11 Martin Sebor <msebor@redhat.com>
1466 * doc/invoke.texi (-Wall): List -Wc++17-compat.
1467 (Wno-class-memaccess): Add @opindex.
1468 (Wno-templates, Wno-multiple-inheritance): Same.
1469 (Wno-virtual-inheritance, Wno-namespaces): Same.
1470 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
1471 (Wno-format-overflow, Wno-format-truncation): Same.
1472 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
1473 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
1474 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
1475 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
1476 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
1477 (Wno-misspelled-isr): Same.
1479 2018-06-11 Martin Sebor <msebor@redhat.com>
1481 * PR tree-optimization/86083
1482 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
1484 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
1486 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
1488 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
1491 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
1492 on the correct operand.
1493 (*movdi_internal64): Ditto.
1495 2018-06-11 Martin Liska <mliska@suse.cz>
1497 PR tree-optimization/86089
1498 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
1500 2018-06-11 Julia Koval <julia.koval@intel.com>
1502 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
1503 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
1504 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
1506 2018-06-11 Olivier Hainque <hainque@adacore.com>
1508 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
1509 for Ada with strict dwarf2.
1511 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
1514 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
1517 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
1519 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
1521 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
1523 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
1526 2018-06-08 Martin Liska <mliska@suse.cz>
1528 * tree-cfg.h (debug_function): Fix argument type to match
1531 2018-06-08 Martin Liska <mliska@suse.cz>
1533 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
1534 Remove usage of MPX-related (and removed) fields.
1535 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
1537 2018-06-08 David Malcolm <dmalcolm@redhat.com>
1539 * cfg.c (debug): Use TDF_NONE rather than 0.
1540 * cfghooks.c (debug): Likewise.
1541 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
1542 (struct dump_option_value_info): Convert to...
1543 (struct kv_pair): ...this template type.
1544 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
1546 (optinfo_verbosity_options): Likewise.
1547 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
1549 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
1550 than int for "optgroup_flags" param.
1551 (dump_generic_expr_loc): Use dump_flags_t rather than int for
1553 (dump_dec): Likewise.
1554 (dump_finish): Use TDF_NONE rather than 0.
1555 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
1556 rather than int for "optgroup_flags" param. Use TDF_NONE rather
1557 than 0. Update for change to option_ptr.
1558 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
1559 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
1560 0. Update for changes to optinfo_verbosity_options and
1562 (opt_info_switch_p): Convert optgroup_flags from int to
1564 (dump_basic_block): Use dump_flags_t rather than int
1565 for "dump_kind" param.
1566 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
1567 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
1568 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
1569 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
1570 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
1571 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
1572 TDF_NONE): Convert from macros to...
1573 (enum dump_flag): ...this new enum.
1574 (dump_flags_t): Update to use enum.
1575 (operator|, operator&, operator~, operator|=, operator&=):
1576 Implement for dump_flags_t.
1577 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
1578 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
1579 Convert from macros to...
1580 (enum optgroup_flag): ...this new enum.
1581 (optgroup_flags_t): New typedef.
1582 (operator|, operator|=): Implement for optgroup_flags_t.
1583 (struct dump_file_info): Convert field "alt_flags" to
1584 dump_flags_t. Convert field "optgroup_flags" to
1586 (dump_basic_block): Use dump_flags_t rather than int for param.
1587 (dump_generic_expr_loc): Likewise.
1588 (dump_dec): Likewise.
1589 (dump_register): Convert param "optgroup_flags" to
1591 (opt_info_enable_passes): Likewise.
1592 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
1594 * gimple-pretty-print.c (debug): Likewise.
1595 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
1596 (merged_store_group::apply_stores): Likewise.
1597 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
1598 * gimple.c (verify_gimple_pp): Likewise.
1599 * graphite-poly.c (print_pbb_body): Likewise.
1600 * passes.c (pass_manager::register_one_dump_file): Convert
1601 local "optgroup_flags" to optgroup_flags_t.
1602 * print-tree.c (print_node): Use TDF_NONE rather than 0.
1604 (debug_body): Likewise.
1605 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
1606 to optgroup_flags_t.
1607 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
1609 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
1610 (convert_mult_to_fma): Likewise.
1611 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
1612 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
1613 * tree-vect-data-refs.c (dump_lower_bound): Convert param
1614 "dump_kind" to dump_flags_t.
1616 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
1618 * config/rs6000/rs6000.c (min, max): Delete.
1620 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
1622 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
1623 -mabi=spe and -mabi=no-spe.
1625 2018-06-08 Martin Liska <mliska@suse.cz>
1627 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
1628 where we expect an existing summary.
1630 2018-06-08 Martin Liska <mliska@suse.cz>
1632 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
1633 * ipa-inline.h (estimate_edge_growth): Likewise.
1635 2018-06-08 Martin Liska <mliska@suse.cz>
1637 * cgraph.c (function_version_hasher::hash): Use
1638 cgraph_node::get_uid ().
1639 (function_version_hasher::equal):
1640 * cgraph.h (cgraph_node::get_uid): New method.
1641 * ipa-inline.c (update_caller_keys): Use
1642 cgraph_node::get_uid ().
1643 (update_callee_keys): Likewise.
1644 * ipa-utils.c (searchc): Likewise.
1645 (ipa_reduced_postorder): Likewise.
1646 * lto-cgraph.c (input_node): Likewise.
1647 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
1648 * symbol-summary.h (symtab_insertion): Likewise.
1649 (symtab_removal): Likewise.
1650 (symtab_duplication): Likewise.
1651 * tree-pretty-print.c (dump_function_header): Likewise.
1652 * tree-sra.c (convert_callers_for_node): Likewise.
1654 2018-06-08 Martin Liska <mliska@suse.cz>
1656 * cgraph.c (symbol_table::create_edge): Always assign a new
1658 (symbol_table::free_edge): Do not recycle numbers.
1659 * cgraph.h (cgraph_edge::get): New method.
1660 * symbol-summary.h (symtab_removal): Use it.
1661 (symtab_duplication): Likewise.
1662 (call_summary::hashable_uid): Remove.
1664 2018-06-08 Martin Liska <mliska@suse.cz>
1666 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
1667 (initialize_growth_caches): Remove.
1668 (free_growth_caches): Likewise.
1669 (do_estimate_edge_time): Use edge_growth_cache.
1670 (do_estimate_edge_size): Likewise.
1671 (do_estimate_edge_hints): Likewise.
1672 * ipa-inline.c (reset_edge_caches): Likewise.
1673 (recursive_inlining): Likewise.
1674 (inline_small_functions): Likewise.
1675 * ipa-inline.h (initialize_growth_caches): Remove.
1676 (estimate_edge_size): Likewise.
1677 (estimate_edge_time): Likewise.
1678 (estimate_edge_hints): Likewise.
1679 (reset_edge_growth_cache): Likewise.
1680 * symbol-summary.h (call_summary::remove): New method.
1682 2018-06-08 Martin Liska <mliska@suse.cz>
1684 * ipa-cp.c (class edge_clone_summary): New summary.
1685 (grow_edge_clone_vectors): Remove.
1686 (ipcp_edge_duplication_hook): Remove.
1687 (class edge_clone_summary_t): New call_summary class.
1688 (ipcp_edge_removal_hook): Remove.
1689 (edge_clone_summary_t::duplicate): New function.
1690 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
1691 (create_specialized_node): Likewise.
1692 (ipcp_driver): Initialize edge_clone_summaries and do not
1695 2018-06-08 Martin Liska <mliska@suse.cz>
1697 * symbol-summary.h (get): New function.
1698 (call_summary::m_initialize_when_cloning): New class member.
1700 2018-06-08 Martin Liska <mliska@suse.cz>
1702 * cgraph.c (cgraph_node::remove): Do not recycle uid.
1703 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
1704 (symbol_table::allocate_cgraph_symbol): Do not set uid.
1705 * passes.c (uid_hash_t): Record removed_nodes by their uids.
1706 (remove_cgraph_node_from_order): Use the removed_nodes set.
1707 (do_per_function_toporder): Likwise.
1708 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
1709 instead of summary_uid.
1710 (symtab_removal): Likewise.
1711 (symtab_duplication): Likewise.
1713 2018-06-08 Martin Liska <mliska@suse.cz>
1715 * ipa-cp.c (ipcp_store_bits_results): Use
1716 ipcp_transformation_sum.
1717 (ipcp_store_vr_results): Likewise.
1718 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
1720 (ipcp_transformation_initialize): ... this.
1721 (ipa_set_node_agg_value_chain):
1722 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
1723 (write_ipcp_transformation_info): Likewise.
1724 (read_ipcp_transformation_info): Likewise.
1725 (ipcp_update_bits): Likewise.
1726 (ipcp_update_vr): Likewise.
1727 (ipcp_transform_function): Likewise.
1728 * ipa-prop.h: Rename ipcp_transformation_summary to
1729 ipcp_transformation.
1730 (class ipcp_transformation_t): New function summary.
1731 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
1732 (ipa_get_agg_replacements_for_node): Likewise.
1734 2018-06-08 Martin Liska <mliska@suse.cz>
1736 * ipa-pure-const.c (struct funct_state_d): Do it class instead
1738 (class funct_state_summary_t): New function_summary class.
1739 (has_function_state): Remove.
1740 (get_function_state): Likewise.
1741 (set_function_state): Likewise.
1742 (add_new_function): Likewise.
1743 (funct_state_summary_t::insert): New function.
1744 (duplicate_node_data): Remove.
1745 (remove_node_data): Remove.
1746 (funct_state_summary_t::duplicate): New function.
1747 (register_hooks): Create new funct_state_summaries.
1748 (pure_const_generate_summary): Use it.
1749 (pure_const_write_summary): Likewise.
1750 (pure_const_read_summary): Likewise.
1751 (propagate_pure_const): Likewise.
1752 (propagate_nothrow): Likewise.
1753 (dump_malloc_lattice): Likewise.
1754 (propagate_malloc): Likewise.
1755 (execute): Do not register hooks, just remove summary
1757 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
1760 2018-06-08 Martin Liska <mliska@suse.cz>
1762 * ipa-reference.c (remove_node_data): Remove.
1763 (duplicate_node_data): Likewise.
1764 (class ipa_ref_var_info_summary_t): New class.
1765 (class ipa_ref_opt_summary_t): Likewise.
1766 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
1767 (get_reference_optimization_summary): Use
1768 ipa_ref_opt_sum_summaries.
1769 (set_reference_vars_info): Remove.
1770 (set_reference_optimization_summary): Likewise.
1771 (ipa_init): Create summaries.
1772 (init_function_info): Use function summary.
1773 (ipa_ref_opt_summary_t::duplicate): New function.
1774 (ipa_ref_opt_summary_t::remove): New function.
1775 (get_read_write_all_from_node): Fix GNU coding style.
1776 (propagate): Use function summary.
1777 (write_node_summary_p): Fix GNU coding style.
1778 (stream_out_bitmap): Likewise.
1779 (ipa_reference_read_optimization_summary): Use function summary.
1780 (ipa_reference_c_finalize): Do not release hooks.
1782 2018-06-08 Martin Liska <mliska@suse.cz>
1784 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
1785 (analyze_function_body): Extract multiple calls of get_create.
1786 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
1787 * ipa-inline.c (recursive_inlining): Use ::get method.
1788 * ipa-inline.h (estimate_edge_growth): Likewise.
1790 2018-06-08 Martin Liska <mliska@suse.cz>
1792 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
1794 (hsa_function_summary::hsa_function_summary): Use the new enum
1796 (hsa_gpu_implementation_p): Use hsa_summaries::get.
1797 * hsa-gen.c (hsa_get_host_function): Likewise.
1798 (get_brig_function_name): Likewise.
1799 * ipa-hsa.c (process_hsa_functions): Likewise.
1800 (ipa_hsa_write_summary): Likewise.
1801 * symbol-summary.h (symtab_duplication): Use ::get function/
1802 (get): New function.
1804 2018-06-08 Martin Liska <mliska@suse.cz>
1806 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
1808 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
1809 (hsa_register_kernel): Likewise.
1810 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
1811 * hsa-gen.c (hsa_get_host_function): Likewise.
1812 (get_brig_function_name): Likewise.
1813 (generate_hsa): Likewise.
1814 (pass_gen_hsail::execute): Likewise.
1815 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
1816 (devirtualization_time_bonus): Likewise.
1817 (ipcp_propagate_stage): Likewise.
1818 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
1819 (edge_set_predicate): Likewise.
1820 (evaluate_conditions_for_known_args): Likewise.
1821 (evaluate_properties_for_edge): Likewise.
1822 (ipa_fn_summary::reset): Likewise.
1823 (ipa_fn_summary_t::duplicate): Likewise.
1824 (dump_ipa_call_summary): Likewise.
1825 (ipa_dump_fn_summary): Likewise.
1826 (analyze_function_body): Likewise.
1827 (compute_fn_summary): Likewise.
1828 (estimate_edge_devirt_benefit): Likewise.
1829 (estimate_edge_size_and_time): Likewise.
1830 (estimate_calls_size_and_time): Likewise.
1831 (estimate_node_size_and_time): Likewise.
1832 (inline_update_callee_summaries): Likewise.
1833 (remap_edge_change_prob): Likewise.
1834 (remap_edge_summaries): Likewise.
1835 (ipa_merge_fn_summary_after_inlining): Likewise.
1836 (ipa_update_overall_fn_summary): Likewise.
1837 (read_ipa_call_summary): Likewise.
1838 (inline_read_section): Likewise.
1839 (write_ipa_call_summary): Likewise.
1840 (ipa_fn_summary_write): Likewise.
1841 (ipa_free_fn_summary): Likewise.
1842 * ipa-hsa.c (process_hsa_functions): Likewise.
1843 (ipa_hsa_write_summary): Likewise.
1844 (ipa_hsa_read_section): Likewise.
1845 * ipa-icf.c (sem_function::merge): Likewise.
1846 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
1847 (do_estimate_edge_time): Likewise.
1848 (estimate_size_after_inlining): Likewise.
1849 (estimate_growth): Likewise.
1850 (growth_likely_positive): Likewise.
1851 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
1852 (inline_call): Likewise.
1853 * ipa-inline.c (caller_growth_limits): Likewise.
1854 (can_inline_edge_p): Likewise.
1855 (can_inline_edge_by_limits_p): Likewise.
1856 (compute_uninlined_call_time): Likewise.
1857 (compute_inlined_call_time): Likewise.
1858 (want_inline_small_function_p): Likewise.
1859 (edge_badness): Likewise.
1860 (update_caller_keys): Likewise.
1861 (update_callee_keys): Likewise.
1862 (recursive_inlining): Likewise.
1863 (inline_small_functions): Likewise.
1864 (inline_to_all_callers_1): Likewise.
1865 (dump_overall_stats): Likewise.
1866 (early_inline_small_functions): Likewise.
1867 (early_inliner): Likewise.
1868 * ipa-inline.h (estimate_edge_growth): Likewise.
1869 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
1870 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
1871 * ipa-prop.h (IPA_NODE_REF): Likewise.
1872 (IPA_EDGE_REF): Likewise.
1873 * ipa-pure-const.c (malloc_candidate_p): Likewise.
1874 (propagate_malloc): Likewise.
1875 * ipa-split.c (execute_split_functions): Likewise.
1876 * symbol-summary.h: Rename get to get_create.
1878 (get_create): Likewise.
1879 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
1881 2018-06-08 Martin Liska <mliska@suse.cz>
1883 * symbol-summary.h (release): Move definition out of class
1885 (symtab_removal): Likewise.
1886 (symtab_duplication): Likewise.
1888 2018-06-08 Martin Liska <mliska@suse.cz>
1890 * symbol-summary.h (function_summary): Move constructor
1891 implementation out of class declaration.
1892 (release): Likewise.
1893 (symtab_insertion): Likewise.
1894 (symtab_removal): Likewise.
1895 (symtab_duplication): Likewise.
1898 2018-06-08 Martin Liska <mliska@suse.cz>
1900 * Makefile.in: Remove support for MPX (macros, related functions,
1901 fields in cgraph_node, ...).
1902 * builtin-types.def (BT_BND): Likewise.
1903 (BT_FN_BND_CONST_PTR): Likewise.
1904 (BT_FN_CONST_PTR_BND): Likewise.
1905 (BT_FN_VOID_PTR_BND): Likewise.
1906 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
1907 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
1908 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
1909 (expand_builtin_mempcpy_with_bounds): Likewise.
1910 (expand_builtin_memset_with_bounds): Likewise.
1911 (expand_builtin_memset_args): Likewise.
1912 (std_expand_builtin_va_start): Likewise.
1913 (expand_builtin): Likewise.
1914 (expand_builtin_with_bounds): Likewise.
1915 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
1916 (DEF_LIB_BUILTIN_CHKP): Likewise.
1917 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
1918 (DEF_CHKP_BUILTIN): Likewise.
1919 (BUILT_IN_MEMCPY): Likewise.
1920 (BUILT_IN_MEMMOVE): Likewise.
1921 (BUILT_IN_MEMPCPY): Likewise.
1922 (BUILT_IN_MEMSET): Likewise.
1923 (BUILT_IN_STPCPY): Likewise.
1924 (BUILT_IN_STRCAT): Likewise.
1925 (BUILT_IN_STRCHR): Likewise.
1926 (BUILT_IN_STRCPY): Likewise.
1927 (BUILT_IN_STRLEN): Likewise.
1928 (BUILT_IN_MEMCPY_CHK): Likewise.
1929 (BUILT_IN_MEMMOVE_CHK): Likewise.
1930 (BUILT_IN_MEMPCPY_CHK): Likewise.
1931 (BUILT_IN_MEMSET_CHK): Likewise.
1932 (BUILT_IN_STPCPY_CHK): Likewise.
1933 (BUILT_IN_STRCAT_CHK): Likewise.
1934 (BUILT_IN_STRCPY_CHK): Likewise.
1935 * calls.c (store_bounds): Likewise.
1936 (emit_call_1): Likewise.
1937 (special_function_p): Likewise.
1938 (maybe_warn_nonstring_arg): Likewise.
1939 (initialize_argument_information): Likewise.
1940 (finalize_must_preallocate): Likewise.
1941 (compute_argument_addresses): Likewise.
1942 (expand_call): Likewise.
1943 * cfgexpand.c (expand_call_stmt): Likewise.
1944 (expand_return): Likewise.
1945 (expand_gimple_stmt_1): Likewise.
1946 (pass_expand::execute): Likewise.
1947 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
1948 (cgraph_node::remove): Likewise.
1949 (cgraph_node::dump): Likewise.
1950 (cgraph_node::verify_node): Likewise.
1951 * cgraph.h (chkp_function_instrumented_p): Likewise.
1952 (symtab_node::get_alias_target): Likewise.
1953 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
1954 (cgraph_local_p): Likewise.
1955 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
1956 (cgraph_edge::rebuild_references): Likewise.
1957 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
1958 (walk_polymorphic_call_targets): Likewise.
1959 (cgraph_node::expand_thunk): Likewise.
1960 (symbol_table::output_weakrefs): Likewise.
1961 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
1962 (ix86_handle_option): Likewise.
1963 * config/i386/constraints.md: Likewise.
1964 * config/i386/i386-builtin-types.def (BND): Likewise.
1968 * config/i386/i386-builtin.def (BDESC_END): Likewise.
1969 (BDESC_FIRST): Likewise.
1971 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
1972 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
1973 * config/i386/i386.c (enum reg_class): Likewise.
1974 (ix86_target_string): Likewise.
1975 (ix86_option_override_internal): Likewise.
1976 (ix86_conditional_register_usage): Likewise.
1977 (ix86_valid_target_attribute_inner_p): Likewise.
1978 (ix86_set_indirect_branch_type): Likewise.
1979 (ix86_set_current_function): Likewise.
1980 (ix86_function_arg_regno_p): Likewise.
1981 (init_cumulative_args): Likewise.
1982 (ix86_function_arg_advance): Likewise.
1983 (ix86_function_arg): Likewise.
1984 (ix86_pass_by_reference): Likewise.
1985 (ix86_function_value_regno_p): Likewise.
1986 (ix86_function_value_1): Likewise.
1987 (ix86_function_value_bounds): Likewise.
1988 (ix86_return_in_memory): Likewise.
1989 (ix86_setup_incoming_vararg_bounds): Likewise.
1990 (ix86_va_start): Likewise.
1991 (indirect_thunk_need_prefix): Likewise.
1992 (print_reg): Likewise.
1993 (ix86_print_operand): Likewise.
1994 (ix86_expand_call): Likewise.
1995 (ix86_output_function_return): Likewise.
1996 (reg_encoded_number): Likewise.
1997 (BDESC_VERIFYS): Likewise.
1998 (ix86_init_mpx_builtins): Likewise.
1999 (ix86_init_builtins): Likewise.
2000 (ix86_emit_cmove): Likewise.
2001 (ix86_emit_move_max): Likewise.
2002 (ix86_expand_builtin): Likewise.
2003 (ix86_builtin_mpx_function): Likewise.
2004 (ix86_get_arg_address_for_bt): Likewise.
2005 (ix86_load_bounds): Likewise.
2006 (ix86_store_bounds): Likewise.
2007 (ix86_load_returned_bounds): Likewise.
2008 (ix86_store_returned_bounds): Likewise.
2009 (ix86_class_likely_spilled_p): Likewise.
2010 (ix86_hard_regno_mode_ok): Likewise.
2011 (x86_order_regs_for_local_alloc): Likewise.
2012 (ix86_mitigate_rop): Likewise.
2013 (ix86_bnd_prefixed_insn_p): Likewise.
2014 (ix86_mpx_bound_mode): Likewise.
2015 (ix86_make_bounds_constant): Likewise.
2016 (ix86_initialize_bounds): Likewise.
2017 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
2018 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
2019 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
2020 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
2021 (TARGET_CHKP_BOUND_MODE): Likewise.
2022 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
2023 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
2024 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
2025 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
2026 * config/i386/i386.h (TARGET_MPX): Likewise.
2027 (TARGET_MPX_P): Likewise.
2028 (VALID_BND_REG_MODE): Likewise.
2029 (FIRST_BND_REG): Likewise.
2030 (LAST_BND_REG): Likewise.
2031 (enum reg_class): Likewise.
2032 (BND_REG_P): Likewise.
2033 (BND_REGNO_P): Likewise.
2034 (BNDmode): Likewise.
2035 (ADJUST_INSN_LENGTH): Likewise.
2036 * config/i386/i386.md: Likewise.
2037 * config/i386/i386.opt: Likewise.
2038 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
2039 (defined): Likewise.
2040 (LINK_MPX): Likewise.
2041 (MPX_SPEC): Likewise.
2042 (LIBMPX_SPEC): Likewise.
2043 (LIBMPXWRAPPERS_SPEC): Likewise.
2044 (CHKP_SPEC): Likewise.
2045 * config/i386/predicates.md: Likewise.
2046 * dbxout.c (dbxout_type): Likewise.
2047 * doc/extend.texi: Likewise.
2048 * doc/invoke.texi: Likewise.
2049 * doc/md.texi: Likewise.
2050 * doc/tm.texi: Likewise.
2051 * doc/tm.texi.in: Likewise.
2052 * dwarf2out.c (is_base_type): Likewise.
2053 (gen_formal_types_die): Likewise.
2054 (gen_subprogram_die): Likewise.
2055 (gen_type_die_with_usage): Likewise.
2056 (gen_decl_die): Likewise.
2057 (dwarf2out_late_global_decl): Likewise.
2058 * expr.c (expand_assignment): Likewise.
2059 (emit_storent_insn): Likewise.
2060 (store_expr_with_bounds): Likewise.
2061 (store_expr): Likewise.
2062 (expand_expr_real_1): Likewise.
2063 * expr.h (store_expr_with_bounds): Likewise.
2064 * function.c (use_register_for_decl): Likewise.
2065 (struct bounds_parm_data): Likewise.
2066 (assign_parms_augmented_arg_list): Likewise.
2067 (assign_parm_find_entry_rtl): Likewise.
2068 (assign_parm_is_stack_parm): Likewise.
2069 (assign_parm_load_bounds): Likewise.
2070 (assign_bounds): Likewise.
2071 (assign_parms): Likewise.
2072 (expand_function_start): Likewise.
2073 * gcc.c (CHKP_SPEC): Likewise.
2074 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
2075 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
2076 (wrestrict_dom_walker::check_call): Likewise.
2077 * gimple.c (gimple_build_call_from_tree): Likewise.
2078 * gimple.h (enum gf_mask): Likewise.
2079 (gimple_call_with_bounds_p): Likewise.
2080 (gimple_call_set_with_bounds): Likewise.
2081 * gimplify.c (gimplify_init_constructor): Likewise.
2082 * ipa-cp.c (initialize_node_lattices): Likewise.
2083 (propagate_constants_across_call): Likewise.
2084 (find_more_scalar_values_for_callers_subset): Likewise.
2085 * ipa-hsa.c (process_hsa_functions): Likewise.
2086 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
2087 * ipa-icf.c (sem_function::merge): Likewise.
2088 * ipa-inline.c (early_inliner): Likewise.
2089 * ipa-pure-const.c (warn_function_noreturn): Likewise.
2090 (warn_function_cold): Likewise.
2091 (propagate_pure_const): Likewise.
2092 * ipa-ref.h (enum GTY): Likewise.
2093 * ipa-split.c (find_retbnd): Likewise.
2094 (consider_split): Likewise.
2095 (split_function): Likewise.
2096 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
2097 * ipa.c (walk_polymorphic_call_targets): Likewise.
2098 (symbol_table::remove_unreachable_nodes): Likewise.
2099 (process_references): Likewise.
2100 (cgraph_build_static_cdtor_1): Likewise.
2101 * lto-cgraph.c (lto_output_node): Likewise.
2102 (output_refs): Likewise.
2103 (compute_ltrans_boundary): Likewise.
2104 (input_overwrite_node): Likewise.
2105 (input_node): Likewise.
2106 (input_cgraph_1): Likewise.
2107 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
2108 * passes.c (pass_manager::execute_early_local_passes): Likewise.
2109 (class pass_chkp_instrumentation_passes): Likewise.
2110 (make_pass_chkp_instrumentation_passes): Likewise.
2111 * passes.def: Likewise.
2112 * rtl.h (struct GTY): Likewise.
2113 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
2114 * stor-layout.c (layout_type): Likewise.
2115 * symtab.c: Likewise.
2116 * target.def: Likewise.
2117 * targhooks.c (default_chkp_bound_type): Likewise.
2118 (default_chkp_bound_mode): Likewise.
2119 (default_builtin_chkp_function): Likewise.
2120 (default_chkp_function_value_bounds): Likewise.
2121 (default_chkp_make_bounds_constant): Likewise.
2122 (default_chkp_initialize_bounds): Likewise.
2123 * targhooks.h (default_chkp_bound_type): Likewise.
2124 (default_chkp_bound_mode): Likewise.
2125 (default_builtin_chkp_function): Likewise.
2126 (default_chkp_function_value_bounds): Likewise.
2127 (default_chkp_make_bounds_constant): Likewise.
2128 (default_chkp_initialize_bounds): Likewise.
2129 * toplev.c (compile_file): Likewise.
2130 (process_options): Likewise.
2131 * tree-core.h (DEF_BUILTIN): Likewise.
2132 (DEF_BUILTIN_CHKP): Likewise.
2133 * tree-inline.c (declare_return_variable): Likewise.
2134 (remap_gimple_stmt): Likewise.
2135 (copy_bb): Likewise.
2136 (initialize_inlined_parameters): Likewise.
2137 (expand_call_inline): Likewise.
2138 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
2139 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
2140 (make_pass_ipa_chkp_produce_thunks): Likewise.
2141 (make_pass_chkp): Likewise.
2142 (make_pass_chkp_opt): Likewise.
2143 (make_pass_chkp_instrumentation_passes): Likewise.
2144 * tree-pretty-print.c (dump_generic_node): Likewise.
2145 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
2146 * tree-ssa-dce.c (propagate_necessity): Likewise.
2147 (eliminate_unnecessary_stmts): Likewise.
2148 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
2149 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
2150 * tree-ssa-sccvn.h: Likewise.
2151 * tree-ssa-strlen.c (get_string_length): Likewise.
2152 (valid_builtin_call): Likewise.
2153 (adjust_last_stmt): Likewise.
2154 (handle_builtin_strchr): Likewise.
2155 (handle_builtin_strcpy): Likewise.
2156 (handle_builtin_stxncpy): Likewise.
2157 (handle_builtin_memcpy): Likewise.
2158 (handle_builtin_strcat): Likewise.
2159 (strlen_check_and_optimize_stmt): Likewise.
2160 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
2161 * tree-streamer-in.c: Likewise.
2162 * tree-streamer.c (record_common_node): Likewise.
2163 * tree.c (tree_code_size): Likewise.
2164 (wide_int_to_tree_1): Likewise.
2165 (type_contains_placeholder_1): Likewise.
2166 (build_common_tree_nodes): Likewise.
2167 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
2168 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
2169 (POINTER_BOUNDS_P): Likewise.
2170 (BOUNDED_TYPE_P): Likewise.
2171 (BOUNDED_P): Likewise.
2172 (CALL_WITH_BOUNDS_P): Likewise.
2173 (pointer_bounds_type_node): Likewise.
2174 * value-prof.c (gimple_ic): Likewise.
2175 * var-tracking.c (vt_add_function_parameters): Likewise.
2176 * varasm.c (make_decl_rtl): Likewise.
2177 (assemble_start_function): Likewise.
2178 (output_constant): Likewise.
2179 (maybe_assemble_visibility): Likewise.
2180 * varpool.c (ctor_for_folding): Likewise.
2181 * chkp-builtins.def: Remove.
2182 * ipa-chkp.c: Remove.
2183 * ipa-chkp.h: Remove.
2184 * rtl-chkp.c: Remove.
2185 * rtl-chkp.h: Remove.
2186 * tree-chkp-opt.c: Remove.
2187 * tree-chkp.c: Remove.
2188 * tree-chkp.h: Remove.
2190 2018-06-07 Carl Love <cel@us.ibm.com>
2192 * gcc/config/rs6000/vsx.md (vextract_fp_from_shorth,
2193 vextract_fp_from_shortl): Add BE support.
2195 2018-06-07 Paul Koning <ni1d@arrl.net>
2197 * compare-elim.c (try_merge_compare): Don't merge compare if
2198 address contains a side effect.
2199 (try_eliminate_compare): Likewise.
2201 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
2203 * config.gcc: Support "tremont".
2204 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
2205 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2207 * config/i386/i386.c (m_TREMONT): Define.
2208 (processor_target_table): Add "tremont".
2209 (PTA_TREMONT): Define.
2210 (ix86_lea_outperforms): Add TARGET_TREMONT.
2211 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
2212 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
2213 and M_INTEL_GOLDMONT_PLUS.
2214 (fold_builtin_cpu): Add "tremont".
2215 (ix86_add_stmt_cost): Add TARGET_TREMONT.
2216 (ix86_option_override_internal): Add "tremont".
2217 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
2218 (processor_type): Add PROCESSOR_TREMONT.
2219 * config/i386/x86-tune.def: Add m_TREMONT.
2220 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
2222 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
2224 * gcc/config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
2225 symbol defined for msp430i* devices to be lower case.
2227 gcc/testsuite/gcc.target/msp430/
2228 * msp430i-device-symbol.c: New test.
2229 * msp430f-device-symbol.c: New test.
2230 * msp430.h: New test header file.
2232 2018-06-07 Richard Biener <rguenther@suse.de>
2234 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
2235 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
2236 Properly wrap signed arithmetic if overflow wraps.
2238 2018-06-07 Jakub Jelinek <jakub@redhat.com>
2240 PR tree-optimization/69615
2241 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
2242 of a cast from a same precision integral SSA_NAME in a bb dominated
2243 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
2244 cast to utype if rhs2 has already a compatible type.
2246 2018-06-07 Richard Biener <rguenther@suse.de>
2248 PR tree-optimization/85935
2249 * graphite-scop-detection.c (find_params_in_bb): Analyze
2250 condition operands with respect to the correct loop. Assert
2251 the analysis doesn't fail.
2253 2018-06-04 Carl Love <cel@us.ibm.com>
2255 * gcc/config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
2256 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
2257 as it is slightly cheaper.
2258 (first_match_or_eos_index_<mode>):
2259 Calculate index using natural element order.
2260 (first_match_index_<mode>):
2261 Calculate index using natural element order.
2262 (first_match_or_eos_index_<mode>):
2263 Calculate index using natural order.
2264 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
2265 for BE and LE modes.
2266 * gcc/config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
2267 P9V_BUILTIN_VCLZLSBB_V16QI.
2268 * gcc/config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
2271 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
2273 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
2274 indentation and line wrap for many prototypes. Add missing
2275 @smallexample directives around block of prototypes for vec_xl and
2278 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
2280 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
2281 track if we pass or return IEEE 128-bit floating point.
2282 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
2283 C++ mangling that is compatible with GCC 8.1.
2284 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
2285 (init_cumulative_args): Note if we pass or return IEEE 128-bit
2286 floating point types.
2287 (rs6000_function_arg_advance_1): Likewise.
2288 (rs6000_mangle_type): Optionally generate mangled names that match
2289 what GCC 8.1 generated for IEEE 128-bit floating point types.
2290 (rs6000_globalize_decl_name): If we have an external function that
2291 passes or returns IEEE 128-bit types, generate a weak reference
2292 from the mangled name used in GCC 8.1 to the current mangled
2294 (rs6000_init_builtins): Make __ibm128 use the long double type if
2295 long double is IBM extended double. Make __float128 use the long
2296 double type if long double is IEEE 128-bit.
2299 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
2300 macro for __ibm128 built-in functions.
2301 (PACK_IF): Add __ibm128 pack/unpack functions.
2302 (UNPACK_IF): Likewise.
2303 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
2304 enable long double built-in functions if long double is IEEE
2305 128-bit floating point.
2306 (rs6000_invalid_builtin): Update long double built-in function
2308 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
2309 functions, adjust the built-in function to use the long double
2310 built-in function if __ibm128 and long double are the same type.
2311 * doc/extend.texi (PowerPC builtins): Update documention for
2312 __builtin_{,un}pack_longdouble. Add documentation for
2313 __builtin_{,un}pack_ibm128.
2315 2018-06-06 Jim Wilson <jimw@sifive.com>
2317 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
2318 (struct machine_function): New field interrupt_mode.
2319 (riscv_handle_type_attribute): New function. Add forward declaration.
2320 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
2321 (riscv_expand_epilogue): Check interrupt_mode field.
2322 (riscv_set_current_function): Check interrupt attribute args and
2323 set interrupt_mode field.
2324 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
2325 (riscv_sret, riscv_uret): New.
2326 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
2327 new arguments to interrupt attribute.
2329 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
2332 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
2333 Don't handle -mcpu=power8 if -mpower9-vector is also used.
2335 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
2337 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
2338 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
2339 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
2340 several redundant entries.
2342 2018-06-06 David Malcolm <dmalcolm@redhat.com>
2344 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
2345 type from "rtx" to "rtx_insn *".
2346 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
2347 for local "call_insn", removing cast.
2348 (ix86_expand_call): Likewise, introducing a "call_insn" local.
2350 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
2352 PR tree-optimization/86066
2353 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
2354 for BIT_INSERT_EXPR stores.
2356 2018-06-06 Richard Biener <rguenther@suse.de>
2358 PR tree-optimization/86062
2359 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
2360 component refs ontop
2361 of to be offsetted base.
2363 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
2365 * gcc/config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
2366 to be static and remove check on interrupt attribute name.
2368 gcc/testsuite/gcc.target/msp430/
2369 * function-attributes-4.c: New test.
2370 * static-interrupts.c: New test.
2372 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
2374 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
2375 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
2377 2018-06-05 Steve Ellcey <sellcey@cavium.com>
2380 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
2382 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
2383 Remove second argument, change how error is called.
2384 (aarch64_layout_arg): Remove second argument from
2385 aarch64_err_no_fpadvsimd call.
2386 (aarch64_init_cumulative_args): Ditto.
2387 (aarch64_gimplify_va_arg_expr): Ditto.
2388 * config/aarch64/aarch64.md (mov<mode>): Ditto.
2390 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
2392 * config/i386/i386.md (simple_return_indirect_internal): New expander.
2393 (*simple_return_indirect_internal<mode>): Rename from
2394 simple_return_indirect_internal. Use W mode iterator.
2395 (rstorssp): New expander.
2396 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
2397 (clrssbsy): New expander.
2398 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
2400 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
2402 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
2404 (cmse_check_pointed_object): Likewise.
2406 2018-06-05 Martin Liska <mliska@suse.cz>
2408 PR gcov-profile/47618
2409 * doc/invoke.texi: Document how -fprofile-dir format
2412 2018-06-05 Richard Biener <rguenther@suse.de>
2414 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
2415 removal pretend DOM info isn't available so we do not update
2416 it and only remove edges, not dominated blocks. Actually free
2417 DOM info in case we removed something. Remove unreachable blocks.
2418 (mfb_keep_latches): Work with either DOM info or marked backedges.
2419 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
2420 first. Mark backedges if DOM info isn't available.
2421 (Re-)compute DOM info after cleanup_control_flow_pre.
2423 2018-06-05 Richard Biener <rguenther@suse.de>
2425 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
2426 (locus_discrim_hasher::hash): Adjust.
2427 (locus_discrim_hasher::equal): Likewise.
2428 (next_discriminator_for_locus): Work on line directly.
2429 (same_line_p): Pass in expanded locus1 as well.
2430 (assign_discriminators): Avoid redundant location expansions.
2432 2018-06-05 Richard Biener <rguenther@suse.de>
2434 PR tree-optimization/86046
2435 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
2436 if required after clearing TREE_ADDRESSABLE.
2438 2018-06-05 Richard Biener <rguenther@suse.de>
2440 PR tree-optimization/86047
2441 * tree-ssa-loop.c (for_each_index): Glob handling of all
2442 decls and constants and really handle all of them.
2444 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2447 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
2448 qualifier_void_pointer and qualifier_const_void_pointer.
2449 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
2450 (arm_init_builtins): Handle the above.
2451 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
2452 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
2453 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
2456 2018-06-05 Martin Liska <mliska@suse.cz>
2458 * auto-profile.c (read_autofdo_file): Do not use
2459 gcov_ctr_summary struct.
2460 (afdo_callsite_hot_enough_for_early_inline): Likewise.
2461 * coverage.c (struct counts_entry): Likewise.
2462 (read_counts_file): Read just single summary entry.
2463 (get_coverage_counts): Use gcov_summary struct.
2464 * coverage.h (get_coverage_counts): Likewise.
2465 * gcov-dump.c (dump_working_sets): Likewise.
2466 (tag_summary): Dump just single summary.
2467 * gcov-io.c (gcov_write_summary): Write just histogram
2469 (gcov_read_summary): Read just single summary.
2470 (compute_working_sets): Use gcov_summary struct.
2471 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
2472 of GCOV_COUNTERS_SUMMABLE.
2473 (GCOV_COUNTERS_SUMMABLE): Remove.
2474 (GCOV_FIRST_VALUE_COUNTER): Replace with
2475 GCOV_COUNTER_V_INTERVAL.
2476 (struct gcov_ctr_summary): Remove.
2477 (struct gcov_summary): Directly use fields of former
2479 (compute_working_sets): Use gcov_summary struct.
2480 * gcov.c (read_count_file): Do not use ctrs fields.
2481 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
2483 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
2485 * profile.c: Likewise.
2486 * profile.h: Likewise.
2488 2018-06-05 Martin Liska <mliska@suse.cz>
2490 PR gcov-profile/84846
2491 * gcov.c (output_lines): Print working directory only
2492 in intermediate format.
2494 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
2496 * config/s390/s390-builtin-types.def: Add void function type.
2497 * config/s390/s390-builtins.def: Use the function type for the
2500 2018-06-04 Jim Wilson <jimw@sifive.com>
2502 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
2504 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
2505 and maybe_eh_return. Change regno to unsigned int. Use new args to
2506 handle EH_RETURN_DATA_REGNO registers properly.
2507 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
2508 (riscv_expand_epilogue): Update comment. Change argument name and
2509 type. Update code to use new name and type. Pass new args to
2510 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
2512 * config/riscv/riscv.md (NORMAL_RETURN): New.
2513 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
2514 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
2515 (eh_return): Call gen_eh_return_internal and emit barrier.
2516 (eh_return_internal): Call riscv_expand_epilogue.
2518 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
2520 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
2521 bit_insertion field and declare can_be_merged_into method.
2522 (merged_store_group::can_be_merged_into): New method.
2523 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
2524 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
2525 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
2527 2018-06-04 Richard Biener <rguenther@suse.de>
2529 PR tree-optimization/85955
2530 * builtins.c (fold_builtin_sincos): Convert pointers to
2531 destination to appropriate type before dereferencing.
2533 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
2535 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
2537 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
2539 * expr.c (expand_expr_real_1): Force the operand into memory if
2540 its TYPE_MODE is BLKmode and if there is no integer mode for
2541 the number of bits being extracted.
2543 2018-06-04 Jakub Jelinek <jakub@redhat.com>
2547 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
2548 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
2550 2018-06-04 Richard Biener <rguenther@suse.de>
2552 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
2553 (cleanup_tree_cfg_noloop): ... single caller. Do
2554 start_recording_case_labels later.
2556 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
2558 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
2559 to _IMMINTRIN_H_INCLUDED.
2560 * config/i386/pconfigintrin.h: Ditto.
2561 * config/i386/waitpkgintrin.h: Ditto.
2562 * config/i386/immintrin.h: Add includes for sgxintrin.h,
2563 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
2564 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
2565 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
2566 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
2567 waitpkgintrin.h and cldemoteintrin.h.
2569 2018-06-04 Richard Biener <rguenther@suse.de>
2571 PR tree-optimization/86038
2572 * tracer.c (find_best_successor): Check probability for
2573 being initialized, bail out if not.
2575 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
2578 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
2579 of bits to ignore when comparing architectures.
2581 2018-06-04 Jakub Jelinek <jakub@redhat.com>
2583 PR tree-optimization/69615
2584 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
2585 maximum or minimum of the type, try to merge it also as if
2586 range1 is + [-, x - 1] or + [x + 1, -].
2589 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
2591 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
2593 PR tree-optimization/86034
2594 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
2595 the unsigned bitfield type in a bit insertion sequence if it does not
2596 have a larger precision than the bitfield size.
2597 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
2599 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
2601 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
2603 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
2605 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
2606 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
2607 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
2608 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
2610 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
2612 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
2613 Disable -fdelete-null-pointer-checks for ELF toolchain.
2615 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
2616 Kito Cheng <kito.cheng@gmail.com>
2618 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
2619 (nds32le-*-*, nds32be-*-*): Integrate checking process.
2620 (nds32*-*-*): Add glibc and uclibc conditions.
2621 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
2622 (TARGET_EXCEPT_UNWIND_INFO): Define.
2623 * config/nds32/elf.h: New file.
2624 * config/nds32/linux.h: New file.
2625 * config/nds32/nds32-elf.opt: New file.
2626 * config/nds32/nds32-linux.opt: New file.
2627 * config/nds32/nds32-fp-as-gp.c
2628 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
2629 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
2631 (nds32_asm_file_end): Ditto.
2632 (nds32_print_operand): Ditto.
2633 (nds32_insert_attributes): Ditto.
2634 (nds32_init_libfuncs): New function.
2635 (TARGET_HAVE_TLS): Define.
2636 (TARGET_INIT_LIBFUNCS): Define.
2637 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
2639 (TARGET_ELF): Apply different mcmodel setting.
2640 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
2641 been migrated into elf.h and linux.h files.
2642 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
2643 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
2644 (mcmodel): The content has been migrated into nds32-elf.opt and
2645 nds32-linux.opt files.
2646 * config/nds32/t-elf: New file.
2647 * config/nds32/t-linux: New file.
2649 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
2650 Shiva Chen <shiva0217@gmail.com>
2652 * config/nds32/constants.md (unspec_volatile_element): Add
2653 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
2654 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
2656 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
2657 (make_pass_nds32_fp_as_gp): Declare.
2658 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
2660 (nds32_asm_function_end_prologue): Remove unused asm output.
2661 (nds32_asm_function_begin_epilogue): Remove unused asm output.
2662 (nds32_asm_file_start): Output necessary fp_as_gp information.
2663 (nds32_option_override): Adjust register usage.
2664 (nds32_expand_prologue): Consider fp_as_gp situation.
2665 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
2666 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
2669 (simple_return): Ditto.
2670 (omit_fp_begin): Output special directive for fp_as_gp.
2671 (omit_fp_end): Output special directive for fp_as_gp.
2672 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
2673 mforbid-fp-as-gp): New options.
2675 2018-06-01 Mark Wielaard <mark@klomp.org>
2677 * dwarf2out.c (dwarf2out_finish): Remove generation of
2678 DW_AT_loclists_base.
2680 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
2682 * gimple-ssa-store-merging.c: Include gimple-fold.h.
2683 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
2684 (struct merged_store_group): Add bit_insertion field.
2685 (dump_char_array): Use standard hexadecimal format.
2686 (merged_store_group::merged_store_group): Set bit_insertion to false.
2687 (merged_store_group::apply_stores): Use optimal buffer size. Deal
2688 with BIT_INSERT_EXPR stores. Move up code updating the mask and
2689 also print the mask in the dump file.
2690 (pass_store_merging::gate): Minor tweak.
2691 (imm_store_chain_info::coalesce_immediate): Fix wrong association
2692 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
2693 stores with INTEGER_CST stores.
2694 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
2695 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
2696 and use it throughout. Generate bit insertion sequences if need be.
2697 (pass_store_merging::process_store): Remove redundant condition.
2698 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
2700 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
2702 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
2703 the 128-bit floating point types. Fix function comment.
2705 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2707 * config/aarch64/aarch64-simd.md
2708 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
2710 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
2713 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
2715 PR tree-optimization/85989
2716 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
2718 (backprop::intersect_uses): Check it when deciding whether this
2719 is a backedge reference.
2720 (backprop::process_block): Add each phi to m_visited_phis
2721 after visiting it, then clear it at the end.
2723 2018-06-01 Richard Biener <rguenther@suse.de>
2725 * tree-vectorizer.h (vect_dr_stmt): New function.
2726 (vect_get_load_cost): Adjust.
2727 (vect_get_store_cost): Likewise.
2728 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
2729 Use vect_dr_stmt instead of DR_SMTT.
2730 (vect_record_base_alignments): Likewise.
2731 (vect_calculate_target_alignment): Likewise.
2732 (vect_compute_data_ref_alignment): Likewise and make static.
2733 (vect_update_misalignment_for_peel): Likewise.
2734 (vect_verify_datarefs_alignment): Likewise.
2735 (vector_alignment_reachable_p): Likewise.
2736 (vect_get_data_access_cost): Likewise. Pass down
2737 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
2738 (vect_get_peeling_costs_all_drs): Likewise.
2739 (vect_peeling_hash_get_lowest_cost): Likewise.
2740 (vect_enhance_data_refs_alignment): Likewise.
2741 (vect_find_same_alignment_drs): Likewise.
2742 (vect_analyze_data_refs_alignment): Likewise.
2743 (vect_analyze_group_access_1): Likewise.
2744 (vect_analyze_group_access): Likewise.
2745 (vect_analyze_data_ref_access): Likewise.
2746 (vect_analyze_data_ref_accesses): Likewise.
2747 (vect_vfa_segment_size): Likewise.
2748 (vect_small_gap_p): Likewise.
2749 (vectorizable_with_step_bound_p): Likewise.
2750 (vect_prune_runtime_alias_test_list): Likewise.
2751 (vect_analyze_data_refs): Likewise.
2752 (vect_supportable_dr_alignment): Likewise.
2753 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
2754 (vect_gen_prolog_loop_niters): Likewise.
2755 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
2756 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
2758 (vect_recog_mask_conversion_pattern): Likewise.
2759 (vect_try_gather_scatter_pattern): Likewise.
2760 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
2761 to vect_get_store_cost.
2762 (vect_get_store_cost): Get stmt_info instead of DR.
2763 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
2764 (vect_get_load_cost): Get stmt_info instead of DR.
2766 2018-06-01 Richard Biener <rguenther@suse.de>
2769 * gimple-fold.c (var_decl_component_p): Also allow offsetted
2770 vars wrapped in MEM_REFs.
2772 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
2774 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
2775 Fix subreg tests so that we only return a choice between
2776 GENERAL_REGS and FP_REGS if the original classes included both.
2778 2018-06-01 Richard Biener <rguenther@suse.de>
2781 * tree-ssa-structalias.c (get_function_part_constraint):
2782 Handle NULL fi->decl.
2783 (find_func_aliases_for_call): Properly handle indirect
2784 fi from direct call.
2785 (find_func_clobbers): Likewise.
2786 (ipa_pta_execute): Likewise.
2787 (create_variable_info_for): For functions that are ifunc_resolver
2788 resolve to a varinfo that contains the result of the resolver call.
2789 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
2792 2018-05-31 Michael Collison <michael.collison@arm.com>
2794 * config/aarch64/aarch64.md:
2795 (*fix_to_zero_extenddfdi2): New pattern.
2796 * gcc.target/aarch64/fix_extend1.c: New testcase.
2798 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
2802 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
2803 and BUILT_IN_STRNCMP_EQ.
2804 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
2805 BUILT_IN_STRNCMP_EQ.
2806 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
2807 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
2808 (gimple_fold_builtin): Likewise.
2809 * tree-ssa-strlen.c (compute_string_length): New function.
2810 (determine_min_obsize): New function.
2811 (handle_builtin_string_cmp): New function to handle calls to
2812 string compare functions.
2813 (strlen_optimize_stmt): Add handling to builtin string compare
2815 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
2816 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
2817 * tree.c (build_common_builtin_nodes): Add new defines of
2818 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
2820 2018-05-31 Jakub Jelinek <jakub@redhat.com>
2823 * bb-reorder.c (pass_partition_blocks::gate): Return false for
2824 functions with naked attribute.
2826 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
2828 * config/i386/sse.md (avx_vec_concat<mode>):
2829 Substitute concat_tg_mode mode attribute with xtg_mode.
2830 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
2831 (concat_tg_mode): Remove mode attribute.
2833 2018-05-31 Martin Sebor <msebor@redhat.com>
2836 * calls.c (alloc_max_size): Correct a logic error/typo.
2837 Treat excessive arguments as infinite. Warn for invalid arguments.
2838 * doc/invoke.texi (-Walloc-size-larger-than): Update.
2840 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
2843 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
2844 and movx for Haswell.
2846 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
2847 Cesar Philippidis <cesar@codesourcery.com>
2850 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
2851 when emitting error on private/firstprivate reductions.
2852 * omp-low.c (lower_omp_target): Avoid reference-type processing
2853 on pointers for firstprivate clause.
2855 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
2857 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
2860 * config/aarch64/aarch64-simd.md
2861 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
2862 (aarch64_ld1_x3_<mode>): Likewise
2863 (aarch64_st1x2<VALLDIF:mode>): Likewise
2864 (aarch64_st1_x2_<mode>): Likewise
2865 (aarch64_st1x3<VALLDIF:mode>): Likewise
2866 (aarch64_st1_x3_<mode>): Likewise
2867 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
2868 (vld1_s8_x3): Likewise.
2869 (vld1_u16_x3): Likewise.
2870 (vld1_s16_x3): Likewise.
2871 (vld1_u32_x3): Likewise.
2872 (vld1_s32_x3): Likewise.
2873 (vld1_u64_x3): Likewise.
2874 (vld1_s64_x3): Likewise.
2875 (vld1_f16_x3): Likewise.
2876 (vld1_f32_x3): Likewise.
2877 (vld1_f64_x3): Likewise.
2878 (vld1_p8_x3): Likewise.
2879 (vld1_p16_x3): Likewise.
2880 (vld1_p64_x3): Likewise.
2881 (vld1q_u8_x3): Likewise.
2882 (vld1q_s8_x3): Likewise.
2883 (vld1q_u16_x3): Likewise.
2884 (vld1q_s16_x3): Likewise.
2885 (vld1q_u32_x3): Likewise.
2886 (vld1q_s32_x3): Likewise.
2887 (vld1q_u64_x3): Likewise.
2888 (vld1q_s64_x3): Likewise.
2889 (vld1q_f16_x3): Likewise.
2890 (vld1q_f32_x3): Likewise.
2891 (vld1q_f64_x3): Likewise.
2892 (vld1q_p8_x3): Likewise.
2893 (vld1q_p16_x3): Likewise.
2894 (vld1q_p64_x3): Likewise.
2895 (vst1_s64_x2): Likewise.
2896 (vst1_u64_x2): Likewise.
2897 (vst1_f64_x2): Likewise.
2898 (vst1_s8_x2): Likewise.
2899 (vst1_p8_x2): Likewise.
2900 (vst1_s16_x2): Likewise.
2901 (vst1_p16_x2): Likewise.
2902 (vst1_s32_x2): Likewise.
2903 (vst1_u8_x2): Likewise.
2904 (vst1_u16_x2): Likewise.
2905 (vst1_u32_x2): Likewise.
2906 (vst1_f16_x2): Likewise.
2907 (vst1_f32_x2): Likewise.
2908 (vst1_p64_x2): Likewise.
2909 (vst1q_s8_x2): Likewise.
2910 (vst1q_p8_x2): Likewise.
2911 (vst1q_s16_x2): Likewise.
2912 (vst1q_p16_x2): Likewise.
2913 (vst1q_s32_x2): Likewise.
2914 (vst1q_s64_x2): Likewise.
2915 (vst1q_u8_x2): Likewise.
2916 (vst1q_u16_x2): Likewise.
2917 (vst1q_u32_x2): Likewise.
2918 (vst1q_u64_x2): Likewise.
2919 (vst1q_f16_x2): Likewise.
2920 (vst1q_f32_x2): Likewise.
2921 (vst1q_f64_x2): Likewise.
2922 (vst1q_p64_x2): Likewise.
2923 (vst1_s64_x3): Likewise.
2924 (vst1_u64_x3): Likewise.
2925 (vst1_f64_x3): Likewise.
2926 (vst1_s8_x3): Likewise.
2927 (vst1_p8_x3): Likewise.
2928 (vst1_s16_x3): Likewise.
2929 (vst1_p16_x3): Likewise.
2930 (vst1_s32_x3): Likewise.
2931 (vst1_u8_x3): Likewise.
2932 (vst1_u16_x3): Likewise.
2933 (vst1_u32_x3): Likewise.
2934 (vst1_f16_x3): Likewise.
2935 (vst1_f32_x3): Likewise.
2936 (vst1_p64_x3): Likewise.
2937 (vst1q_s8_x3): Likewise.
2938 (vst1q_p8_x3): Likewise.
2939 (vst1q_s16_x3): Likewise.
2940 (vst1q_p16_x3): Likewise.
2941 (vst1q_s32_x3): Likewise.
2942 (vst1q_s64_x3): Likewise.
2943 (vst1q_u8_x3): Likewise.
2944 (vst1q_u16_x3): Likewise.
2945 (vst1q_u32_x3): Likewise.
2946 (vst1q_u64_x3): Likewise.
2947 (vst1q_f16_x3): Likewise.
2948 (vst1q_f32_x3): Likewise.
2949 (vst1q_f64_x3): Likewise.
2950 (vst1q_p64_x3): Likewise.
2952 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
2954 * config/msp430/msp430.c (msp430_output_labelref): Prepend
2955 user_label_prefix to name.
2957 * tree-core.h: Update comment about the format of NAME string
2958 passed to handler in attribute_spec.
2960 * config/msp430/msp430.md: Remove erroneous subreg expression from
2961 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
2962 zero_extend{q,h}isi2.
2964 2018-05-30 Borislav Petkov <bp@suse.de>
2966 * doc/extend.texi: Document some architecture specific
2967 constraints and sort entries.
2969 2018-05-30 Martin Sebor <msebor@redhat.com>
2972 * builtins.c (expand_builtin_stpcpy_1): New function.
2973 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
2974 only if the former succeeds.
2976 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
2978 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
2981 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
2983 * doc/invoke.texi (-flinker-output): Document
2985 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
2987 * passes.c (ipa_write_summaries): Only modify statements if body
2989 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
2990 incrementally linking.
2991 (ipa_passes): Likewise.
2992 * lto-cgraph.c (lto_output_node): When incrementally linking do not
2993 pass down resolution info.
2994 * common.opt (flag_incremental_link): Update info.
2995 * gcc.c (plugin specs): Turn flinker-output=* to
2996 -plugin-opt=-linker-output-known
2997 * toplev.c (compile_file): Also cut compilation when doing incremental
2999 * flag-types. (enum lto_partition_model): Add
3000 LTO_LINKER_OUTPUT_NOLTOREL.
3001 (invoke.texi): Add -flinker-output docs.
3002 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
3003 link same way as WPA; do not stream in dead initializers.
3005 * dwarf2out.c (dwarf2out_die_ref_for_decl,
3006 darf2out_register_external_decl): Support incremental link.
3008 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
3010 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
3012 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
3014 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
3015 it down to simple_object_copy_lto_debug_sections.
3016 (run_gcc): Determine incremental LTO link time and configure
3017 lto1 into non-wpa mode, disable renaming of debug sections.
3019 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
3021 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
3022 descriptions of various incorrectly documented functions.
3024 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
3027 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
3028 address check not strict.
3030 2018-05-30 Richard Biener <rguenther@suse.de>
3032 PR tree-optimization/85964
3033 * tracer.c (better_p): Drop initialized count check, we only
3034 call the function with initialized counts now.
3035 (find_best_successor): Do find a best edge if one
3036 has uninitialized count.
3037 (find_best_predecessor): Likewise. Do BB frequency check only
3038 if count is initialized.
3040 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
3042 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
3043 (aarch64_ldrstr_offset_compare): New.
3044 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
3045 load/store orderings.
3046 (aarch64_gen_adjusted_ldpstp): Likewise.
3048 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
3050 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
3051 Check for subset of GENERAL_REGS and FP_REGS.
3052 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
3055 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
3057 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
3058 and wi::to_poly_offset. Add the current offset and then check
3059 whether the sum fits, rather than using an unchecked addition of
3060 a checked term. Check for a shwi rather than a uhwi.
3061 * expr.c (get_bit_range): Use tree_to_poly_uint64.
3062 (store_constructor): Use poly_int_tree_p.
3063 (expand_expr_real_1): Likewise.
3064 * function.c (assign_temp): Likewise.
3065 * fold-const.c (const_binop): Use poly_int_tree_p and
3067 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
3069 * ipa-icf-gimple.c (func_checker::compare_operand): Use
3070 to_poly_offset for MEM offsets.
3071 * ipa-icf.c (sem_variable::equals): Likewise.
3072 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
3073 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
3074 wi::to_poly_offset for BIT_FIELD_REF offsets.
3075 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
3077 * var-tracking.c (emit_note_insn_var_location): Use
3078 tree_to_poly_uint64.
3080 2018-05-29 Jim Wilson <jimw@sifive.com>
3082 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
3084 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
3087 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
3088 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
3090 (sse4_1_round<mode>2): Use nonimmediate_operand
3091 for operand 1 predicate.
3093 2018-05-29 Martin Sebor <msebor@redhat.com>
3094 Richard Biener <rguenther@suse.de>
3097 * calls.c (get_size_range): Call determine_value_range instead
3098 of get_value_range..
3099 * tree-vrp.h (determine_value_range): Declared new function.
3100 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
3102 2018-05-29 Richard Biener <rguenther@suse.de>
3104 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
3105 sure to use non-pattern stmts for get_earlier_stmt arguments.
3106 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
3107 called on pattern stmts.
3108 (get_later_stmt): Likewise.
3110 2018-05-29 Martin Liska <mliska@suse.cz>
3112 PR gcov-profile/85759
3113 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
3116 2018-05-29 Jakub Jelinek <jakub@redhat.com>
3118 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
3120 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
3124 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
3125 VEC_PACK_FLOAT_EXPR): New tree codes.
3126 * tree-pretty-print.c (op_code_prio): Handle
3127 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
3128 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
3129 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
3130 * tree-inline.c (estimate_operator_cost): Likewise.
3131 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
3132 * fold-const.c (const_binop): Likewise.
3133 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
3134 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
3135 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
3136 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
3137 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
3138 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
3139 * expr.c (expand_expr_real_2): Likewise.
3140 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
3141 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
3142 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
3144 * optabs.c (expand_widen_pattern_expr): For
3145 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
3146 sign from result type rather than operand's type.
3147 (expand_binop_directly): For vec_packu_float_optab and
3148 vec_packs_float_optab allow result type to be different from operand's
3150 * optabs-tree.c (optab_for_tree_code): Handle
3151 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
3152 VEC_PACK_FLOAT_EXPR. Formatting fixes.
3153 * tree-vect-generic.c (expand_vector_operations_1): Handle
3154 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
3155 VEC_PACK_FLOAT_EXPR.
3156 * tree-vect-stmts.c (supportable_widening_operation): Handle
3158 (supportable_narrowing_operation): Handle FLOAT_EXPR.
3159 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
3160 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
3161 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
3162 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
3164 (vec_pack<floatprefix>_float_<mode>): New expander.
3165 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
3167 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
3168 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
3169 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
3170 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
3171 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
3173 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
3174 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
3175 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
3176 VEC_PACK_FLOAT_EXPR): Document.
3178 2018-05-29 Richard Biener <rguenther@suse.de>
3180 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
3182 (stmt_vec_info_vec): Make pointer.
3183 (init_stmt_vec_info_vec): Remove.
3184 (free_stmt_vec_info_vec): Likewise.
3185 (set_stmt_vec_info_vec): New function.
3186 (free_stmt_vec_infos): Likewise.
3187 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
3188 (set_vinfo_for_stmt): Likewise.
3189 (get_earlier_stmt): Likewise.
3190 (get_later_stmt): Likewise.
3191 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
3192 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
3193 (vec_info::~vec_info): Free stmt_vec_infos.
3194 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
3195 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
3196 (pass_slp_vectorize::execute): Likewise.
3197 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
3198 (free_stmt_vec_info_vec): Likewise.
3199 (set_stmt_vec_info_vec): New function.
3200 (free_stmt_vec_infos): Likewise.
3201 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
3202 the global stmt_vec_info_vec.
3203 * tree-parloops.c (gather_scalar_reductions): Use
3204 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
3207 2018-05-29 Richard Biener <rguenther@suse.de>
3209 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
3211 2018-05-29 Martin Liska <mliska@suse.cz>
3212 David Malcolm <dmalcolm@redhat.com>
3214 * vec.c (test_reverse): New.
3215 (vec_c_tests): Add new test.
3216 * vec.h (vl_ptr>::reverse): New function.
3218 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
3220 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
3222 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
3225 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
3227 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
3229 2018-05-28 Richard Biener <rguenther@suse.de>
3231 PR tree-optimization/85933
3232 * tree-vect-data-refs.c (vect_record_base_alignments): Only
3233 look at stmts marked as vectorizable.
3235 2018-05-28 Richard Biener <rguenther@suse.de>
3237 PR tree-optimization/85934
3238 * tree-vect-generic.c (expand_vector_operations_1): Hoist
3239 vector boolean check before scalar optimization.
3241 2018-05-28 Jakub Jelinek <jakub@redhat.com>
3243 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
3246 2018-05-28 Mark Wielaard <mark@klomp.org>
3248 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
3249 if it is an expression containing a minus sign.
3251 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
3253 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
3255 2018-05-27 Paul Koning <ni1d@arrl.net>
3257 * config/pdp11/pdp11.md (truncsihi2): Remove.
3259 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
3260 Chung-Ju Wu <jasonwucj@gmail.com>
3262 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
3264 (unaligned_store_dw): Ditto.
3265 * config/nds32/nds32-memory-manipulation.c
3266 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
3267 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
3268 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
3269 (emit_setmem_word_loop): Rename to ...
3270 (emit_setmem_doubleword_loop): ... this.
3271 (nds32_gen_dup_4_byte_to_word_value): New function.
3272 (nds32_gen_dup_8_byte_to_double_word_value): New function.
3273 (nds32_expand_setmem_loop): Refine implementation.
3274 (nds32_expand_setmem_loop_v3m): Ditto.
3275 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
3278 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
3280 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
3282 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
3284 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
3285 (nds32_init_machine_status): Initialize machine->attr_naked_p and
3286 machine->attr_no_prologue_p.
3287 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
3288 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
3289 (nds32_expand_epilogue): Consider attr_naked_p.
3290 (nds32_expand_epilogue_v3pop): Likewise.
3291 (nds32_can_use_return_insn): Likewise.
3292 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
3293 attr_no_prologue_p fields.
3294 * config/nds32/nds32.opt (mret-in-naked-func): New option.
3296 2018-05-27 Jakub Jelinek <jakub@redhat.com>
3299 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
3301 * config/i386/sse.md
3302 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
3304 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
3306 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
3308 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
3310 (*<floatsuffix>floatv2div2sf2): Rename to ...
3311 (*float<floatunssuffix>v2div2sf2): ... this.
3312 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
3313 (float<floatunssuffix>v2div2sf2_mask): ... this.
3314 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
3315 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
3316 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
3318 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
3320 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
3322 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
3324 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
3326 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
3328 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
3329 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
3330 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
3331 gen_ufix_truncv8dfv8si2.
3332 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
3333 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
3334 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
3335 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
3336 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
3337 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
3338 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
3339 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
3341 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
3345 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
3347 2018-05-25 Jim Wilson <jimw@sifive.com>
3349 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
3350 * config/riscv/riscv.c (struct machine_function): Add
3351 interrupt_handler_p and attribute_checked_p fields.
3352 (riscv_attribute_table): Add interrupt.
3353 (riscv_interrupt_type_p): New.
3354 (riscv_save_reg_p): Save extra regs for interrupt handler.
3355 (riscv_use_save_libcall): Return false for interrupt handler.
3356 (riscv_first_stack_step): Add forward declaration.
3357 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
3358 for interrupt handler with large frame. Use it for saved reg list.
3359 (riscv_expand_prologue): Move flag_stack_usage_info support to
3360 eliminate duplication.
3361 (riscv_expand_epilogue): Generate mret for interrupt handler.
3362 (riscv_epilogue_uses): New.
3363 (riscv_can_use_return_insn): Return false for interrupt handler.
3364 (riscv_function_ok_for_sibcall): Likewise.
3365 (riscv_set_current_function): Add interrupt handler support.
3366 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
3367 * config/riscv/riscv.md (UNSPECV_MRET): New.
3369 (riscv_frflags, riscv_fsflags): Use tab after opcode.
3371 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
3373 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
3375 PR tree-optimization/85712
3376 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
3377 this candidate has already been replaced in-situ by a copy.
3379 2018-05-25 Jason Merrill <jason@redhat.com>
3381 PR c++/80485 - inline function non-zero address.
3382 * symtab.c (nonzero_address): Check DECL_COMDAT.
3384 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
3387 * config/alpha/alpha.md (ashlsi3): New insn pattern.
3388 (*ashlsi_se): Rename from *ashldi_se. Define as sign
3389 extension of SImode operation. Use const123_operand predicate.
3390 (*saddsi_1): Remove.
3391 (*saddl_se_1): Ditto.
3393 (*ssubl_se_1): Ditto.
3394 * config/alpha/predicates.md (const123_operand): New predicate.
3395 * config/alpha/constraints.md (P): Use IN_RANGE.
3397 2018-05-25 Richard Biener <rguenther@suse.de>
3399 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
3401 (ref_maybe_used_by_stmt_p): Likewise.
3402 (stmt_may_clobber_ref_p): Likewise.
3403 (stmt_may_clobber_ref_p_1): Likewise.
3404 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
3406 (ref_maybe_used_by_stmt_p): Likewise.
3407 (stmt_may_clobber_ref_p): Likewise.
3408 (stmt_may_clobber_ref_p_1): Likewise.
3409 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
3410 the alias oracle to disambiguate DRs with stmts DR analysis
3412 (vect_analyze_data_refs): Do not give up on not analyzable
3413 DRs for BB vectorization. Remove code truncating the dataref
3416 2018-05-25 Jakub Jelinek <jakub@redhat.com>
3419 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
3420 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
3421 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
3423 2018-05-25 Richard Biener <rguenther@suse.de>
3425 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
3426 function, combining stmt data ref gathering and fatal analysis
3428 (vect_analyze_data_refs): Remove now redudnant code and simplify.
3429 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
3430 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
3431 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
3432 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
3434 2018-05-25 Bin Cheng <bin.cheng@arm.com>
3436 PR tree-optimization/85720
3437 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
3438 SCC if all partitions are builtins.
3439 (version_loop_by_alias_check): New parameter. Generate cancelable
3440 runtime alias check if all partitions are builtins.
3441 (distribute_loop): Update call to above function.
3443 2018-05-25 Bin Cheng <bin.cheng@arm.com>
3445 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
3446 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
3447 (parm_default_def_partition_arg): Ditto.
3448 (set_parm_default_def_partition): Ditto.
3449 (get_parm_default_def_partitions): Ditto and make it static.
3450 (get_undefined_value_partitions): Ditto and make it static.
3451 (remove_ssa_form): Refactor call to init_var_map here.
3452 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
3453 computation for loop region.
3454 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
3455 (register_default_def): Delete.
3456 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
3457 (parm_default_def_partition_arg): Ditto.
3458 (set_parm_default_def_partition): Ditto.
3459 (get_parm_default_def_partitions): Ditto and make it static.
3460 (get_undefined_value_partitions): Ditto and make it static.
3461 (coalesce_with_default, coalesce_with_default): Update comment.
3462 (create_coalesce_list_for_region): New func factored out from
3463 create_outofssa_var_map.
3464 (populate_coalesce_list_for_outofssa): New func factored out from
3465 create_outofssa_var_map and coalesce_ssa_name.
3466 (create_outofssa_var_map): Delete.
3467 (coalesce_ssa_name): Refactor to support live range computation.
3468 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
3469 (get_parm_default_def_partitions): Delete.
3470 (get_undefined_value_partitions): Ditto.
3471 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
3472 computation for loop region.
3473 (new_tree_live_info, loe_visit_block): Ditto.
3474 (live_worklist, set_var_live_on_entry): Ditto.
3475 (calculate_live_on_exit, verify_live_on_entry): Ditto.
3476 * tree-ssa-live.h (struct _var_map): New fields.
3477 (init_var_map): Change decl.
3478 (region_contains_p): New.
3480 2018-05-25 Bin Cheng <bin.cheng@arm.com>
3482 * tree-ssa-live.h (live_merge_and_clear): Delete.
3484 2018-05-25 Richard Biener <rguenther@suse.de>
3487 * tree-dump.c (dequeue_and_dump): Remove access to removed
3488 operand 2 of a SWITCH_EXPR.
3490 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3492 * doc/sourcebuild.texi (vect_double_cond_arith): Include
3493 multiplication and division.
3494 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
3495 (cond_udiv@var{m}, cond_umod@var{m}): Document.
3496 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
3497 (cond_udiv_optab, cond_umod_optab): New optabs.
3498 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
3499 (IFN_COND_RDIV): New internal functions.
3500 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
3501 TRUNC_MOD_EXPR and RDIV_EXPR.
3502 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
3503 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
3505 (SVE_INT_BINARY): Include mult.
3506 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
3507 (optab, sve_int_op): Handle mult.
3508 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
3510 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
3511 for SVE_INT_BINARY_SD.
3513 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3515 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
3516 (optab, sve_int_op): Handle div and udiv.
3517 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
3518 for SVE_INT_BINARY_SD.
3519 (*<optab><mode>3): New insn for the same.
3521 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3523 * tree-vect-patterns.c: Include predict.h.
3524 (vect_recog_divmod_pattern): Restrict check for division support
3525 to when optimizing for size.
3527 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3529 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
3530 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
3531 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
3532 (gimple_match_op::set_op): Likewise.
3533 (gimple_resimplify4): Declare.
3534 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
3535 (expr::gen_transform): Likewise.
3536 (decision_tree::gen): Generate a simplification routine for 4 operands.
3537 * gimple-match-head.c (gimple_simplify): Add an overload for
3538 4 operands. In the top-level function, handle up to 4 call
3539 arguments and call gimple_resimplify4.
3540 (gimple_resimplify4): New function.
3541 (build_call_internal): Pass a fourth operand.
3542 (maybe_push_to_seq): Likewise.
3543 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
3544 Fold VEC_COND_EXPRs of an operation and a default value into
3545 an IFN_COND_* function if possible.
3546 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
3548 (SVE_COND_FP_BINARY): Include them.
3549 (optab, sve_fp_op): Handle them.
3550 (SVE_INT_BINARY_REV): New code iterator.
3551 (SVE_COND_FP_BINARY_REV): New int iterator.
3552 (commutative): New int attribute.
3553 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
3555 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
3557 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
3558 (*cond_<optab><mode>): New patterns for reversed operands.
3560 2018-05-25 Richard Biener <rguenther@suse.de>
3562 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
3563 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
3564 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
3565 (STMT_VINFO_GROUPED_ACCESS): Adjust.
3566 * tree-vect-data-refs.c (everywhere): Adjust users.
3567 * tree-vect-loop.c (everywhere): Likewise.
3568 * tree-vect-slp.c (everywhere): Likewise.
3569 * tree-vect-stmts.c (everywhere): Likewise.
3570 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
3572 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3574 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
3576 (gcc_cv_as_section_exclude): ... this.
3577 Try Solaris as #exclude syntax.
3578 * configure: Regenerate.
3579 * config.in: Regenerate.
3580 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
3582 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
3583 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
3585 * varasm.c (default_elf_asm_named_section): Don't check if
3586 HAVE_GAS_SECTION_EXCLUDE is defined.
3588 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3590 * doc/md.texi: Update the documentation of the cond_* optabs
3591 to mention the new final operand. Fix GET_MODE_NUNITS call.
3592 Describe the scalar case too.
3593 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
3594 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
3596 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
3597 (get_conditional_internal_fn): Update comment.
3598 * tree-vect-loop.c (vectorizable_reduction): Pass the original
3599 accumulator value as a final argument to conditional functions.
3600 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
3601 a define_expand and add an "else" operand. Assert for now that
3602 the else operand is equal to operand 2. Use SVE_INT_BINARY and
3603 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
3604 (*cond_<optab><mode>): New patterns.
3605 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
3606 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
3607 (UNSPEC_COND_EOR): Delete.
3608 (optab): Remove associated mappings.
3609 (SVE_INT_BINARY): New code iterator.
3610 (sve_int_op): Remove int attribute and add "minus" to the code
3612 (SVE_COND_INT_OP): Delete.
3613 (SVE_COND_FP_OP): Rename to...
3614 (SVE_COND_FP_BINARY): ...this.
3616 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3618 * optabs.c (can_reuse_operands_p): New function.
3619 (maybe_legitimize_operands): Try to reuse the results for
3622 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
3624 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
3625 Add {q} suffix to insn mnemonic.
3627 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
3629 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
3630 (msp430_warn_func_return): New.
3632 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
3634 * fold-const.c (tree_nonzero_bits): New function.
3635 * fold-const.h (tree_nonzero_bits): Likewise.
3636 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
3637 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
3639 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
3643 * varasm.c (assemble_alias): Check ifunc_resolver only on
3646 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
3649 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
3650 when memory input operand is handled.
3652 2018-05-24 Luis Machado <luis.machado@linaro.org>
3654 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
3656 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
3658 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
3660 * match.pd: Delay FMA folds until after vectorization.
3662 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
3665 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
3666 address check not strict.
3668 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
3670 * gimple-match.h (gimple_match_op): New class.
3671 (mprts_hook): Replace parameters with a gimple_match_op *.
3672 (maybe_build_generic_op): Likewise.
3673 (gimple_simplified_result_is_gimple_val): Replace parameters with
3674 a const gimple_match_op *.
3675 (gimple_simplify): Replace code_helper * and tree * parameters with
3676 a gimple_match_op * parameter.
3677 (gimple_resimplify1): Replace code_helper *, tree and tree *
3678 parameters with a gimple_match_op * parameter.
3679 (gimple_resimplify2): Likewise.
3680 (gimple_resimplify3): Likewise.
3681 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
3682 parameters with a gimple_match_op * parameter.
3683 * gimple-match-head.c (gimple_simplify): Change prototypes of
3684 auto-generated functions to take a gimple_match_op * instead of
3685 separate code_helper * and tree * parameters. Make the same
3686 change in the top-level overload and update calls to the
3687 gimple_resimplify routines. Update calls to the auto-generated
3688 functions and to maybe_push_res_to_seq in the publicly-facing
3689 operation-specific gimple_simplify overloads.
3690 (gimple_match_op::MAX_NUM_OPS): Define.
3691 (gimple_resimplify1): Replace rcode and ops with a single res_op
3692 parameter. Update call to gimple_simplify.
3693 (gimple_resimplify2): Likewise.
3694 (gimple_resimplify3): Likewise.
3695 (mprts_hook): Replace parameters with a gimple_match_op *.
3696 (maybe_build_generic_op): Likewise.
3697 (build_call_internal): Replace type, nargs and ops with
3698 a gimple_match_op *.
3699 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
3700 with a single gimple_match_op *. Update calls to mprts_hook,
3701 build_call_internal and gimple_simplified_result_is_gimple_val.
3702 Factor out code that is common to the tree_code and combined_fn cases.
3703 * genmatch.c (expr::gen_transform): Replace tem_code and
3704 tem_ops with a gimple_match_op called tem_op. Update calls
3705 to the gimple_resimplify functions and maybe_push_res_to_seq.
3706 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
3707 res_ops. Update call to the gimple_resimplify functions.
3708 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
3709 (decision_tree::gen): Make the functions take a gimple_match_op *
3710 called res_op instead of separate res_code and res_ops parameters.
3711 Update call accordingly.
3712 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
3713 and ops with a single res_op parameter. Update calls to
3714 maybe_build_generic_op and maybe_push_res_to_seq.
3715 (fold_stmt_1): Update calls to gimple_simplify and
3716 replace_stmt_with_simplification.
3717 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
3718 and gimple_simplified_result_is_gimple_val.
3719 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
3721 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
3722 with a gimple_match_op *.
3723 (vn_nary_build_or_lookup): Likewise. Update call to
3724 vn_nary_build_or_lookup_1.
3725 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
3726 gimple_match_op *. Update calls to the gimple_resimplify routines
3727 and to gimple_simplified_result_is_gimple_val.
3728 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
3729 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
3730 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
3731 (visit_nary_op): Likewise.
3732 (visit_reference_op_load): Likewise.
3734 2018-05-23 Luis Machado <luis.machado@linaro.org>
3736 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
3737 modifier for printing the step amount.
3739 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
3742 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
3745 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
3747 * doc/sourcebuild.texi (Endianness): New subsubsection.
3749 2018-05-23 Luis Machado <luis.machado@linaro.org>
3751 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3752 <prefetch_dynamic_strides>: New const bool field.
3753 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3754 prefetch_dynamic_strides.
3755 (exynosm1_prefetch_tune): Likewise.
3756 (thunderxt88_prefetch_tune): Likewise.
3757 (thunderx_prefetch_tune): Likewise.
3758 (thunderx2t99_prefetch_tune): Likewise.
3759 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
3761 (aarch64_override_options_internal): Update to set
3762 PARAM_PREFETCH_DYNAMIC_STRIDES.
3763 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
3764 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
3765 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
3766 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
3767 prefetch-dynamic-strides setting.
3769 2018-05-23 Luis Machado <luis.machado@linaro.org>
3771 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3772 <minimum_stride>: New const int field.
3773 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3774 minimum_stride field defaulting to -1.
3775 (exynosm1_prefetch_tune): Likewise.
3776 (thunderxt88_prefetch_tune): Likewise.
3777 (thunderx_prefetch_tune): Likewise.
3778 (thunderx2t99_prefetch_tune): Likewise.
3779 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
3780 <default_opt_level>: Set to 3.
3781 (aarch64_override_options_internal): Update to set
3782 PARAM_PREFETCH_MINIMUM_STRIDE.
3783 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
3784 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
3785 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
3786 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
3787 stride is constant and is below the minimum stride threshold.
3789 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3791 * config/arm/arm-cpus.in (mode26): Delete.
3792 (armv4): Delete mode26 reference.
3793 * config/arm/arm.c (arm_configure_build_target): Delete use of
3796 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
3798 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
3800 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
3801 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
3803 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
3804 (floatunsdidf2): Ditto.
3806 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
3808 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
3809 (fixuns_trunc<mode>si2_avx512f): Ditto.
3810 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
3811 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
3812 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
3814 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
3816 PR rtl-optimization/79985
3817 * df-scan.c (df_insn_refs_collect): Remove special case for
3818 global registers and asm statements.
3820 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
3822 * extend.texi (Global Register Variables): Rewrite the bullet list.
3823 Note that the register is available for allocation. Note that access
3824 via inline asm must use constraints. Add note about async-signal
3825 handlers. Remove paragraph about automagic register selection.
3827 2018-05-23 Richard Biener <rguenther@suse.de>
3829 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
3830 of fixed offset from memset VN.
3832 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
3834 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
3836 (alloc_cand_and_find_basis): Initialize first_interp field.
3837 (slsr_process_mul): Modify first_interp field.
3838 (slsr_process_add): Likewise.
3839 (slsr_process_cast): Modify first_interp field for each new
3841 (slsr_process_copy): Likewise.
3842 (dump_candidate): Dump first_interp field.
3843 (replace_mult_candidate): Process all interpretations, not just
3845 (replace_rhs_if_not_dup): Likewise.
3846 (replace_one_candidate): Likewise.
3848 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
3850 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
3852 (aarch64_needs_frame_chain): New function.
3853 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
3855 2018-05-23 Sudakshina Das <sudi.das@arm.com>
3858 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
3859 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
3860 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
3861 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
3862 as true for strict-align.
3863 (aarch64_can_inline_p): Perform checks even when callee has no
3864 attributes to check for strict alignment.
3865 * doc/extend.texi (AArch64 Function Attributes): Document
3867 * doc/invoke.texi: (AArch64 Options): Likewise.
3869 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
3871 PR tree-optimization/85853
3872 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
3873 the handling of the root of the node to...
3874 (vect_slp_analyze_node_operations_1): ...this new function,
3875 and run the whole thing with the child nodes' def types
3876 set according to their SLP node's def type.
3878 2018-05-23 Richard Biener <rguenther@suse.de>
3881 * tree-data-ref.c (create_runtime_alias_checks): Defer
3882 and ignore overflow warnings.
3884 2018-05-23 Yury Gribov <tetra2005@gmail.com>
3886 PR tree-optimization/85822
3887 * tree-vrp.c (is_masked_range_test): Fix handling of negative
3890 2018-05-23 Richard Biener <rguenther@suse.de>
3892 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
3893 memset constants via native_interpret_expr.
3895 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
3898 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
3900 (cgraph_node::create_alias): Likewise.
3901 (cgraph_node::get_availability): Check ifunc_resolver instead
3902 of looking up ifunc attribute.
3903 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
3904 * varasm.c (do_assemble_alias): Likewise.
3905 (assemble_alias): Likewise.
3906 (default_binds_local_p_3): Likewise.
3907 * cgraph.h (cgraph_node): Add ifunc_resolver.
3908 (cgraph_node::only_called_directly_or_aliased_p): Return false
3910 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
3912 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
3913 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
3914 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
3915 instead of looking up ifunc attribute.
3917 2018-05-22 Luis Machado <luis.machado@linaro.org>
3919 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
3921 2018-05-22 Martin Sebor <msebor@redhat.com>
3924 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
3925 only when expasion succeeds.
3926 (expand_builtin_strcmp): Same.
3927 (expand_builtin_strncmp): Same.
3929 2018-05-22 Martin Sebor <msebor@redhat.com>
3931 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
3933 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
3934 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3936 * config/aarch64/aarch64-ldpstp.md: Replace uses of
3937 aarch64_mem_pair_operand with memory_operand and delete operand swapping
3939 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
3940 Add check for legitimate_address.
3941 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
3942 (aarch64_swap_ldrstr_operands): New.
3943 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
3946 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
3947 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3949 * config/aarch64/aarch64.md: New patterns to generate stp
3951 (store_pair_sw, store_pair_dw): New patterns to generate stp for
3952 single words and double words.
3953 (load_pair_sw, load_pair_dw): Likewise.
3954 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
3956 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
3958 * config/aarch64/aarch64-ldpstp.md: Modify peephole
3959 for different mode ldpstp and add peephole for merged zero stores.
3961 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
3963 (aarch64_gen_store_pair): Rename calls to match new patterns.
3964 (aarch64_gen_load_pair): Rename calls to match new patterns.
3965 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
3966 (load_pair<DREG:mode><DREG2:mode>): ... This.
3967 (store_pair<mode>): Rename to...
3968 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
3969 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
3971 (V_INT_EQUIV): Handle SImode.
3972 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
3975 2018-05-22 Martin Sebor <msebor@redhat.com>
3978 * calls.c (maybe_warn_nonstring_arg): Use string length to set
3979 or ajust the presumed bound on an operation to avoid unnecessary
3982 2018-05-22 Martin Sebor <msebor@redhat.com>
3984 PR tree-optimization/85826
3985 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
3986 assuming that a DECL necesarily has a constant size.
3988 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
3991 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
3993 2018-05-22 Richard Biener <rguenther@suse.de>
3995 PR tree-optimization/85834
3996 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
3997 non-constant and non-zero memset arguments.
3999 2018-05-22 Martin Liska <mliska@suse.cz>
4002 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
4004 2018-05-22 Richard Biener <rguenther@suse.de>
4006 PR tree-optimization/85863
4007 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
4008 comparisons when vectype is specified.
4009 (vectorizable_condition): Do not specify vectype for
4010 vect_is_simple_cond when SLP vectorizing.
4012 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
4015 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
4016 define __ibm128 as long double.
4017 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
4018 as a distinct type when IEEE 128-bit support is enabled.
4019 (init_float128_ieee): Fix up conversions between IFmode and IEEE
4020 128-bit types to use the correct functions.
4021 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
4022 convert between 128-bit floating point types that have different
4023 modes but the same representation, instead of using gen_lowpart to
4025 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
4027 (IFKF_reg): New attributes to give the register constraints for
4029 (extend<mode>tf2_internal): New insns to mark an explicit
4030 conversion between 128-bit floating point types that have a
4031 different mode but share the same representation.
4033 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
4035 PR tree-optimization/85814
4036 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
4037 a null return from get_strinfo when unsharing the next
4038 strinfo in the chain.
4040 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
4043 * varasm.c (weak_finish): Clean up weak_decls.
4045 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4047 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
4048 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
4049 UNSPEC_UADALP values.
4050 * config/aarch64/iterators.md (ABAL): New int iterator.
4053 (sur): Add mappings for the above.
4054 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
4056 (aarch64_<sur>abal<mode>_4): Likewise.
4057 (aarch64_<sur>adalp<mode>_3): Likewise.
4058 (<sur>sadv16qi): New define_expand.
4060 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
4062 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
4063 (*movdf_internal): Ditto.
4064 (*rcpsf2_sse): Ditto.
4065 (*rsqrtsf2_sse): Ditto.
4066 (*sqrt<mode>2_sse): Ditto.
4068 2018-05-21 Tamar Christina <tamar.christina@arm.com>
4070 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
4072 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
4073 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
4074 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
4075 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
4077 * config/aarch64/arm_neon.h: Likewise.
4078 * config/aarch64/iterators.md (VQ_I): New.
4080 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
4082 * config.gcc: Add arc/t-multilib-linux to tmake_file for
4084 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
4087 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
4089 * config/nds32/constraints.md (S): New constraint.
4090 * config/nds32/nds32.md (call_internal): Use constraint S.
4091 (call_value_internal): Likewise.
4092 (sibcall_internal): Likewise.
4093 (sibcall_value_internal): Likewise.
4095 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
4096 Chung-Ju Wu <jasonwucj@gmail.com>
4098 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
4101 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
4102 Chung-Ju Wu <jasonwucj@gmail.com>
4104 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
4105 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
4106 (nds32_rtx_costs_impl): Simplify.
4107 (nds32_address_cost_impl): Simplify.
4108 (nds32_init_rtx_costs): New function.
4109 (nds32_rtx_costs_speed_prefer): Likewise.
4110 (nds32_rtx_costs_size_prefer): Likewise.
4111 (nds32_address_cost_speed_prefer): Likewise.
4112 (nds32_address_cost_speed_fwprop): Likewise.
4113 (nds32_address_cost_size_prefer): Likewise.
4114 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
4115 * config/nds32/nds32.c (nds32_option_override): Use
4116 nds32_init_rtx_costs function.
4118 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
4120 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
4121 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
4122 (TARGET_PIPELINE_N8): Likewise.
4123 (TARGET_PIPELINE_N10): Likewise.
4124 (TARGET_PIPELINE_N13): Likewise.
4125 (TARGET_PIPELINE_GRAYWOLF): Likewise.
4127 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
4129 * config/nds32/nds32-fpu.md: Update copyright year.
4131 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
4133 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
4135 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
4137 * config/nds32/nds32.c
4138 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
4139 * config/nds32/nds32.opt (minline-asm-r15): New option.
4141 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
4143 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
4145 * config/nds32/nds32.md (abssi2): New pattern.
4147 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
4149 * config/i386/i386.md (rex64namesuffix): New mode attribute.
4150 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
4151 Merge insn pattern from sse_cvtsi2ss<round_name> and
4152 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
4153 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
4154 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
4155 using SWI48 mode iterator.
4156 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
4157 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
4158 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
4159 pattern from sse_cvttss2si<round_saeonly_name>
4160 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
4161 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
4162 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
4163 using SWI48 mode iterator.
4164 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
4165 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
4166 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
4167 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
4168 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
4169 using SWI48 mode iterator.
4170 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
4171 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
4172 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
4173 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
4174 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
4175 SWI48 mode iterator.
4176 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
4177 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
4178 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
4179 pattern from sse_cvttsd2si<round_saeonly_name>
4180 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
4182 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
4184 * config/nds32/nds32-md-auxiliary.c
4185 (nds32_valid_smw_lwm_base_p): Refine.
4186 (nds32_output_smw_single_word): Refine.
4187 (nds32_output_smw_double_word): New.
4188 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
4190 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
4192 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
4193 (nds32_output_stack_pop): Refine.
4194 (nds32_expand_unaligned_load): Refine.
4195 (nds32_expand_unaligned_store): Refine.
4197 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
4198 Chung-Ju Wu <jasonwucj@gmail.com>
4200 * config/nds32/constants.md: Add TP_REGNUM constant.
4201 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
4202 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
4204 * config/nds32/nds32-doubleword.md: Consider flag_pic.
4205 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
4206 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
4207 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
4208 and PIC code generation.
4209 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
4211 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
4213 * config/nds32/nds32.md: Support TLS and PIC.
4214 * config/nds32/nds32.c: Support TLS and PIC.
4215 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
4216 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
4219 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
4221 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
4222 mode with E_ prefix.
4224 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
4225 Chung-Ju Wu <jasonwucj@gmail.com>
4227 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
4228 * config/nds32/nds32-md-auxiliary.c
4229 (symbolic_reference_mentioned_p): New.
4230 (nds32_legitimize_ict_address): New.
4231 (nds32_expand_ict_move): New.
4232 (nds32_indirect_call_referenced_p): New.
4233 (nds32_symbol_binds_local_p): Delete.
4234 (nds32_long_call_p): Modify.
4235 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
4236 * config/nds32/nds32-protos.h
4237 (symbolic_reference_mentioned_p): Declare.
4238 (nds32_legitimize_ict_address): Declare.
4239 (nds32_expand_ict_move): Declare.
4240 (nds32_indirect_call_referenced_p): Declare.
4241 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
4242 (nds32_relax_group): Use nds32_ict_const_p as condition.
4243 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
4244 (nds32_asm_file_start): Output ict_model directive in asm code.
4245 (nds32_legitimate_address_p): Consider indirect call.
4246 (nds32_print_operand): Consider indirect call.
4247 (nds32_print_operand_address): Consider indirect call.
4248 (nds32_insert_attributes): Handle "indirect_call" attribute.
4249 (TARGET_LEGITIMATE_ADDRESS_P): Define.
4250 (TARGET_LEGITIMATE_CONSTANT_P): Define.
4251 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
4252 (TARGET_DELEGITIMIZE_ADDRESS): Define.
4253 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
4254 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
4255 (TARGET_ICT_MODEL_SMALL): Define.
4256 (TARGET_ICT_MODEL_LARGE): Define.
4257 * config/nds32/nds32.md (movsi): Consider ict model.
4258 (call, call_value): Consider ict model.
4259 (sibcall, sibcall_value): Consider ict model.
4260 * config/nds32/nds32.opt (mict-model): New option.
4261 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
4264 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
4265 Monk Chiang <sh.chiang04@gmail.com>
4266 Jim Wilson <jimw@sifive.com>
4268 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
4269 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
4270 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
4271 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
4272 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
4273 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
4274 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
4275 compute save_libcall_adjustment properly.
4276 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
4277 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
4278 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
4279 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
4280 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
4281 (ABI_SPEC): Handle mabi=ilp32e.
4282 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
4283 (RVE): Add RVE mask.
4284 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
4285 <-march>: Add rv32e as an example.
4287 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
4290 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
4291 (intra_create_variable_infos): Handle C++ constructors.
4293 2018-05-18 Martin Liska <mliska@suse.cz>
4295 * passes.def: Remove a redundant pass.
4297 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
4300 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
4302 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4304 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
4306 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
4308 (armv2, armv2a, armv3, armv3m): Delete architectures.
4309 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
4310 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
4311 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
4313 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
4314 (*mulsidi3adddi): Likewise.
4315 (mulsidi3): Likewise.
4316 (*mulsidi3_nov6): Likewise.
4317 (umulsidi3): Likewise.
4318 (umulsidi3_nov6): Likewise.
4319 (umaddsidi4): Likewise.
4320 (*umulsidi3adddi): Likewise.
4321 (smulsi3_highpart): Likewise.
4322 (*smulsi3_highpart_nov6): Likewise.
4323 (umulsi3_highpart): Likewise.
4324 (*umulsi3_highpart_nov6): Likewise.
4325 * config/arm/arm.h (arm_arch3m): Delete.
4326 * config/arm/arm.c (arm_arch3m): Delete.
4327 (arm_option_override_internal): Update armv3-related comment.
4328 (arm_configure_build_target): Delete use of isa_bit_mode32.
4329 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
4330 (arm_rtx_costs_internal): Delete check of arm_arch3m.
4331 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
4333 (mulusa3): Likewise.
4334 * config/arm/arm-protos.h (arm_arch3m): Delete.
4335 * config/arm/arm-tables.opt: Regenerate.
4336 * config/arm/arm-tune.md: Likewise.
4337 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
4338 deleted architectures.
4340 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4342 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
4343 (armv5t, armv5te): New features.
4344 (ARMv5, ARMv5e): Delete fgroups.
4345 (ARMv5t, ARMv5te): Adjust for above changes.
4347 (armv5, armv5e): Delete arches.
4348 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
4350 (*call_reg_arm): Likewise.
4351 (*call_value_reg_armv5): Likewise.
4352 (*call_value_reg_arm): Likewise.
4353 (*call_symbol): Likewise.
4354 (*call_value_symbol): Likewise.
4355 (*sibcall_insn): Likewise.
4356 (*sibcall_value_insn): Likewise.
4358 (prefetch): Likewise.
4359 (define_split and define_peephole2 dependent on arm_arch5):
4361 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
4363 (TARGET_ARM_QBIT): Likewise.
4364 (TARGET_DSP_MULTIPLY): Likewise.
4365 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
4366 (arm_arch5, arm_arch5e): Delete.
4367 (arm_arch5t, arm_arch5te): Declare.
4368 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
4369 (arm_arch5t): Declare.
4370 (arm_option_reconfigure_globals): Update for the above.
4371 (arm_options_perform_arch_sanity_checks): Update comment, replace
4372 use of arm_arch5 with arm_arch5t.
4373 (use_return_insn): Likewise.
4374 (arm_emit_call_insn): Likewise.
4375 (output_return_instruction): Likewise.
4376 (arm_final_prescan_insn): Likewise.
4377 (arm_coproc_builtin_available): Likewise.
4378 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
4379 arm_arch5e with arm_arch5t and arm_arch5te.
4380 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
4381 (arm_arch5t, arm_arch5te): Declare.
4382 * config/arm/arm-tables.opt: Regenerate.
4383 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
4384 * config/arm/t-multilib: Likewise.
4385 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
4386 instead of arm_arch5.
4387 (*call_reg_thumb1): Likewise.
4388 (*call_value_reg_thumb1_v5): Likewise.
4389 (*call_value_reg_thumb1): Likewise.
4390 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
4392 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
4394 2018-05-18 Martin Liska <mliska@suse.cz>
4396 PR gcov-profile/84846
4397 * doc/gcov.texi: Document -t option of gcov tool.
4399 2018-05-18 Martin Liska <mliska@suse.cz>
4401 PR gcov-profile/84846
4402 * gcov.c (print_usage): Add new -t option.
4403 (process_args): Handle the option.
4404 (generate_results): Use stdout as output when requested by
4407 2018-05-18 Martin Liska <mliska@suse.cz>
4409 PR gcov-profile/84846
4410 * coverage.c (coverage_init): Write PWD to .gcno file.
4411 * doc/gcov.texi: Document how working directory is printed.
4412 * gcov-dump.c (dump_gcov_file): Print PWD.
4413 * gcov.c (output_intermediate_file): Likewise.
4414 (read_graph_file): Read PWD string.
4415 (output_lines): Print PWD.
4417 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
4420 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
4421 for retval and return false if all args to phi are zero.
4423 2018-05-18 Richard Biener <rguenther@suse.de>
4425 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
4427 (evrp_dom_walker::before_dom_children): Call it.
4429 2018-05-18 Richard Biener <rguenther@suse.de>
4431 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
4432 results when processing array refs with variable index.
4434 2018-05-18 Toon Moene <toon@moene.org>
4436 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
4437 directly after that of -floop-interchange. Indicate that both
4438 options are enabled by default when specifying -O3.
4440 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4442 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
4443 iterator. Delete separate integer-mode vec_set<mode> expander.
4444 (aarch64_simd_vec_setv2di): Delete.
4445 (vec_setv2di): Delete.
4446 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
4447 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
4448 the "w, r" alternative.
4450 2018-05-18 Martin Liska <mliska@suse.cz>
4452 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
4453 * tree-pass.h (make_pass_lower_switch_O0): New function.
4454 * tree-switch-conversion.c (node_has_low_bound): Remove.
4455 (node_has_high_bound): Likewise.
4456 (node_is_bounded): Likewise.
4457 (class pass_lower_switch): Make it a template type and create
4459 (pass_lower_switch::execute): Add template argument.
4460 (make_pass_lower_switch): New function.
4461 (make_pass_lower_switch_O0): New function.
4462 (do_jump_if_equal): Remove.
4463 (emit_case_nodes): Simplify to just handle all 3 cases and leave
4464 all the hard work to tree optimization passes.
4466 2018-05-18 Martin Liska <mliska@suse.cz>
4468 * dbgcnt.c (limit_low): Renamed from limit.
4469 (limit_high): New variable.
4470 (dbg_cnt_is_enabled): Check for upper limit.
4471 (dbg_cnt): Adjust dumping.
4472 (dbg_cnt_set_limit_by_index): Add new argument for high
4474 (dbg_cnt_set_limit_by_name): Likewise.
4475 (dbg_cnt_process_single_pair): Parse new format.
4476 (dbg_cnt_process_opt): Use strtok.
4477 (dbg_cnt_list_all_counters): Remove 'value' and add
4479 * doc/invoke.texi: Document changes.
4481 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
4483 * doc/sourcebuild.texi (scalar_all_fma): Document.
4484 * tree.def (FMA_EXPR): Delete.
4485 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
4486 * internal-fn.c (ternary_direct): New macro.
4487 (expand_ternary_optab_fn): Likewise.
4488 (direct_ternary_optab_supported_p): Likewise.
4489 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
4490 * builtins.c (fold_builtin_fma): Delete.
4491 (fold_builtin_3): Don't call it.
4492 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
4493 * expr.c (expand_expr_real_2): Likewise.
4494 * fold-const.c (operand_equal_p): Likewise.
4495 (fold_ternary_loc): Likewise.
4496 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
4497 * gimple.c (DEFTREECODE): Likewise.
4498 * gimplify.c (gimplify_expr): Likewise.
4499 * optabs-tree.c (optab_for_tree_code): Likewise.
4500 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
4501 * tree-eh.c (operation_could_trap_p): Likewise.
4502 (stmt_could_throw_1_p): Likewise.
4503 * tree-inline.c (estimate_operator_cost): Likewise.
4504 * tree-pretty-print.c (dump_generic_node): Likewise.
4505 (op_code_prio): Likewise.
4506 * tree-ssa-loop-im.c (stmt_cost): Likewise.
4507 * tree-ssa-operands.c (get_expr_operands): Likewise.
4508 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
4509 * fold-const-call.h (fold_fma): Delete.
4510 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
4511 CFN_FNMA and CFN_FNMS.
4513 * genmatch.c (combined_fn): New enum.
4514 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
4515 (commutative_op): New function.
4516 (commutate): Use it. Handle more than 2 operands.
4517 (dt_operand::gen_gimple_expr): Use commutative_op.
4518 (parser::parse_expr): Allow :c to be used with non-binary
4519 operators if the commutative operand is known.
4520 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
4521 CFN_FMS, CFN_FNMA and CFN_FNMS.
4522 (backprop::process_assign_use): Remove FMA_EXPR handling.
4523 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
4524 (gen_hsa_fma): New function.
4525 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
4526 IFN_FNMA and IFN_FNMS.
4527 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
4528 * gimple-fold.h (follow_all_ssa_edges): Declare.
4529 * gimple-fold.c (follow_all_ssa_edges): New function.
4530 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
4531 gimple_build interface and use follow_all_ssa_edges to fold the result.
4532 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
4533 instead of checking for optabs directly.
4534 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
4535 rather than FMA_EXPRs.
4536 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
4537 call to IFN_FMA instead of an FMA_EXPR.
4539 2018-05-17 Jim Wilson <jimw@sifive.com>
4541 * expr.c (do_tablejump): When converting index to Pmode, if we have a
4542 sign extended promoted subreg, and the range does not have the sign bit
4543 set, then do a sign extend.
4545 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
4546 test, check for sign extended subreg and/or constant operands, and
4547 do a sign extend in that case.
4549 2018-05-17 Steve Ellcey <sellcey@cavium.com>
4551 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
4552 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
4554 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
4555 Change logics_shift_reg to logics_shift_imm.
4556 (thunderx2t99_fp_loadpair_basic): Delete.
4557 (thunderx2t99_fp_storepair_basic): Delete.
4558 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
4559 (thunderx2t99_asimd_polynomial): Delete.
4560 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
4561 and neon_fp_mul_d_scalar_q.
4562 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
4563 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
4564 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
4565 (thunderx2t99_asimd_lut): Add missing tbl types.
4566 (thunderx2t99_asimd_ext): Delete.
4567 (thunderx2t99_asimd_load1_1_mult): Delete.
4568 (thunderx2t99_asimd_load1_2_mult): Delete.
4569 (thunderx2t99_asimd_load1_ldp): New.
4570 (thunderx2t99_asimd_load1): New.
4571 (thunderx2t99_asimd_load2): Add missing *load2* types.
4572 (thunderx2t99_asimd_load3): New.
4573 (thunderx2t99_asimd_load4): New.
4574 (thunderx2t99_asimd_store1_1_mult): Delete.
4575 (thunderx2t99_asimd_store1_2_mult): Delete.
4576 (thunderx2t99_asimd_store2_mult): Delete.
4577 (thunderx2t99_asimd_store2_onelane): Delete.
4578 (thunderx2t99_asimd_store_stp): New.
4579 (thunderx2t99_asimd_store1): New.
4580 (thunderx2t99_asimd_store2): New.
4581 (thunderx2t99_asimd_store3): New.
4582 (thunderx2t99_asimd_store4): New.
4584 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
4586 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
4587 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
4589 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
4590 Segher Boessenkool <segher@kernel.crashing.org>
4593 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
4596 2018-05-17 Richard Biener <rguenther@suse.de>
4598 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
4599 for pruning loop and prune defs feeding only already visited PHIs.
4601 2018-05-17 Richard Biener <rguenther@suse.de>
4603 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
4605 2018-05-17 Bin Cheng <bin.cheng@arm.com>
4606 Richard Biener <rguenther@suse.de>
4608 PR tree-optimization/85793
4609 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
4610 for VMAT_ELEMENTWISE.
4612 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
4614 * internal-fn.h (lookup_internal_fn): Declare
4615 * internal-fn.c (lookup_internal_fn): New function.
4616 * gimple.c (gimple_build_call_from_tree): Handle calls to
4618 * gimple-pretty-print.c (dump_gimple_call): Print "." before
4619 internal function names.
4620 * tree-pretty-print.c (dump_generic_node): Likewise.
4621 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
4623 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
4625 * gimple-fold.h (gimple_build): Make the function forms take
4626 combined_fn rather than built_in_function.
4627 (gimple_simplify): Likewise.
4628 * gimple-match-head.c (gimple_simplify): Likewise.
4629 * gimple-fold.c (gimple_build): Likewise.
4630 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
4631 rather than gimple_build_call_internal.
4632 (get_initial_defs_for_reduction): Likewise.
4633 (vect_create_epilog_for_reduction): Likewise.
4634 (vectorizable_live_operation): Likewise.
4636 2018-05-17 Martin Liska <mliska@suse.cz>
4638 * gimple-ssa-sprintf.c (format_directive): Do not use
4639 space in between 'G_' and '('.
4641 2018-05-17 Jakub Jelinek <jakub@redhat.com>
4644 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
4645 even if the mask is not all ones.
4648 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
4650 (ix86_gimple_fold_builtin): Likewise.
4653 * config/i386/i386.c: Include tree-vector-builder.h.
4654 (ix86_vector_shift_count): New function.
4655 (ix86_fold_builtin): Fold shift builtins by scalar count.
4656 (ix86_gimple_fold_builtin): Likewise.
4658 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
4659 _mm512_setzero): New intrinsics.
4661 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
4662 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4664 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
4665 code generation for cases where splatting a value is not useful.
4666 * simplify-rtx.c (simplify_ternary_operation): Simplify
4667 vec_merge across a vec_duplicate and a paradoxical subreg forming
4668 a vector mode to a vec_concat.
4670 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
4672 * config.gcc: Support "goldmont-plus".
4673 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
4675 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
4676 PROCESSOR_GOLDMONT_PLUS.
4677 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
4678 (processor_target_table): Add "goldmont-plus".
4679 (PTA_GOLDMONT_PLUS): Define.
4680 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
4681 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
4682 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
4683 (fold_builtin_cpu): Add "goldmont-plus".
4684 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
4685 (ix86_option_override_internal): Add "goldmont-plus".
4686 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
4687 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
4688 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
4689 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
4691 2018-05-17 Richard Biener <rguenther@suse.de>
4693 PR tree-optimization/85757
4694 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
4695 remove defs that only feed that PHI from further processing.
4697 2018-05-16 Jim Wilson <jimw@sifive.com>
4699 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
4701 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
4703 2018-05-16 Mark Wielaard <mark@klomp.org>
4705 * dwarf2out.c (count_index_strings): New function.
4706 (output_indirect_strings): Call count_index_strings and generate
4707 header for dwarf_version >= 5.
4709 2018-05-16 Mark Wielaard <mark@klomp.org>
4711 * dwarf2out.c (dwarf_FORM): New function.
4712 (set_indirect_string): Use dwarf_FORM.
4713 (reset_indirect_string): Likewise.
4714 (size_of_die): Likewise.
4715 (value_format): Likewise.
4716 (output_die): Likewise.
4717 (add_skeleton_AT_string): Likewise.
4718 (output_macinfo_op): Likewise.
4719 (index_string): Likewise.
4720 (output_index_string_offset): Likewise.
4721 (output_index_string): Likewise.
4722 (count_index_strings): Likewise.
4724 2018-05-16 Carl Love <cel@us.ibm.com>
4726 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
4727 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
4729 2018-05-16 Martin Jambor <mjambor@suse.cz>
4731 * ipa-prop.c (ipa_free_all_edge_args): Remove.
4732 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
4734 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
4736 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
4737 (fnma<mode>4): Likewise.
4738 (fms<mode>4): Likewise.
4739 (fnms<mode>4): Likewise.
4740 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
4741 (aarch64_fnma<mode>4): Likewise.
4742 (aarch64_fms<mode>4): Likewise.
4743 (aarch64_fnms<mode>4): Likewise.
4744 (aarch64_fnmadd<mode>4): Likewise.
4746 2018-05-16 Jason Merrill <jason@redhat.com>
4748 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
4750 2018-05-16 Richard Biener <rguenther@suse.de>
4752 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
4753 (dump_stmt_cost): Declare.
4754 (add_stmt_cost): Dump cost we add.
4755 (add_stmt_costs): New function.
4756 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
4758 (vect_analyze_stmt): Adjust prototype.
4759 (vectorizable_condition): Likewise.
4760 (vectorizable_live_operation): Likewise.
4761 (vectorizable_reduction): Likewise.
4762 (vectorizable_induction): Likewise.
4763 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
4764 cost vector to pass to vectorizable_ and record afterwards.
4765 (vect_model_reduction_cost): Take cost vector argument and adjust.
4766 (vect_model_induction_cost): Likewise.
4767 (vectorizable_reduction): Likewise.
4768 (vectorizable_induction): Likewise.
4769 (vectorizable_live_operation): Likewise.
4770 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
4771 SLP_TREE_NUMBER_OF_VEC_STMTS.
4772 (vect_analyze_slp_cost_1): Remove.
4773 (vect_analyze_slp_cost): Likewise.
4774 (vect_slp_analyze_node_operations): Take visited args and
4775 a target cost vector. Avoid processing already visited stmt sets.
4776 (vect_slp_analyze_operations): Use a local cost vector to gather
4777 costs and register those of non-discarded instances.
4778 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
4779 (vect_schedule_slp_instance): Remove copying of
4780 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
4782 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
4783 adding cost. Record cost entry location.
4784 (vect_prologue_cost_for_slp_op): Function to compute cost of
4785 a constant or invariant generated for SLP vect in the prologue,
4786 split out from vect_analyze_slp_cost_1.
4787 (vect_model_simple_cost): Make static. Adjust for SLP costing.
4788 (vect_model_promotion_demotion_cost): Likewise.
4789 (vect_model_store_cost): Likewise, make static.
4790 (vect_model_load_cost): Likewise.
4791 (vectorizable_bswap): Add cost vector arg and adjust.
4792 (vectorizable_call): Likewise.
4793 (vectorizable_simd_clone_call): Likewise.
4794 (vectorizable_conversion): Likewise.
4795 (vectorizable_assignment): Likewise.
4796 (vectorizable_shift): Likewise.
4797 (vectorizable_operation): Likewise.
4798 (vectorizable_store): Likewise.
4799 (vectorizable_load): Likewise.
4800 (vectorizable_condition): Likewise.
4801 (vectorizable_comparison): Likewise.
4802 (can_vectorize_live_stmts): Likewise.
4803 (vect_analyze_stmt): Likewise.
4804 (vect_transform_stmt): Adjust calls to vectorizable_*.
4805 * tree-vectorizer.c: Include gimple-pretty-print.h.
4806 (dump_stmt_cost): New function.
4808 2018-05-16 Richard Biener <rguenther@suse.de>
4810 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
4811 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
4812 * tree-ssa-dse.c: Include tree-ssa-loop.h.
4813 (check_name): New callback.
4814 (dse_classify_store): Track cycles via a visited bitmap of PHI
4815 defs and simplify handling of in-loop and across loop dead stores
4816 and properly fail for loop-variant refs. Handle byte-tracking with
4817 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
4820 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
4822 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
4823 (vect_get_mask_type_for_stmt): Likewise.
4824 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
4826 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
4827 to determine the statement's vector type and the vector type that
4828 should be used for calculating nunits. Deal with cases in which
4829 the type has to be deferred.
4830 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
4831 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
4832 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
4833 (vect_determine_vf_for_stmt): New functions, split out from...
4834 (vect_determine_vectorization_factor): ...here.
4835 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
4836 (vect_get_mask_type_for_stmt): New functions, split out from
4837 vect_determine_vectorization_factor.
4839 2018-05-16 Richard Biener <rguenther@suse.de>
4841 * tree-cfg.c (verify_gimple_assign_ternary): Properly
4842 verify the [VEC_]COND_EXPR embedded comparison.
4844 2018-05-15 Martin Sebor <msebor@redhat.com>
4846 PR tree-optimization/85753
4847 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
4848 RECORD_TYPE in addition to ARRAY_TYPE.
4850 2018-05-15 Martin Sebor <msebor@redhat.com>
4853 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
4855 2018-05-15 Richard Biener <rguenther@suse.de>
4857 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
4858 add by_clobber_p one. Change algorithm to collect all defs
4859 representing uses we need to walk and try reducing them to
4860 a single one before failing.
4861 (dse_dom_walker::dse_optimize_stmt): Adjust.
4863 2018-05-13 Mark Wielaard <mark@klomp.org>
4865 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
4866 (size_of_loc_descr): Likewise.
4867 (output_loc_operands): Likewise.
4868 (output_loc_operands_raw): Likewise.
4869 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
4870 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
4871 (hash_loc_operands): Likewise.
4872 (compare_loc_operands): Likewise.
4874 2018-05-14 Mark Wielaard <mark@klomp.org>
4876 * dwarf2out.c (count_index_addrs): New function.
4877 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
4879 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
4881 PR tree-optimization/83648
4882 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
4883 return value as malloc candidate.
4885 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
4888 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
4889 param as true in call to suggest_attribute.
4891 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
4893 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
4894 -mreadonly-in-sdata.
4896 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4898 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
4900 (aarch64_crypto_aesd_fused): Likewise.
4902 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
4904 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
4905 (movsi_aarch64): Likewise.
4906 (load_pairsi): Likewise.
4907 (load_pairdi): Likewise.
4908 (store_pairsi): Likewise.
4909 (store_pairdi): Likewise.
4910 (load_pairsf): Likewise.
4911 (load_pairdf): Likewise.
4912 (store_pairsf): Likewise.
4913 (store_pairdf): Likewise.
4914 (zero_extend): Likewise.
4915 (trunc): Swap alternatives.
4916 (fcvt_target): Add '?' to prefer w over r.
4918 2018-05-14 Jakub Jelinek <jakub@redhat.com>
4921 * config/i386/i386.md: Disallow non-commutative arithmetics in
4922 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
4923 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
4924 in the peephole2 before it.
4926 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
4928 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
4929 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
4930 (ix86_handle_option): Handle -mcldemote.
4931 * config.gcc: New header.
4932 * config/i386/cldemoteintrin.h: New file.
4933 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
4934 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
4936 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
4937 OPTION_MASK_ISA_CLDEMOTE.
4938 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
4939 (ix86_valid_target_attribute_inner_p): Ditto.
4940 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
4941 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
4942 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
4943 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
4944 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
4946 * config/i386/i386.opt: Add -mcldemote.
4947 * config/i386/x86intrin.h: New header.
4948 * doc/invoke.texi: Add -mcldemote.
4950 2018-05-14 Richard Biener <rguenther@suse.de>
4952 * doc/match-and-simplify.texi: Adjust :s documentation.
4954 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
4956 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
4957 intended memcpy size.
4958 (REORDER_45): Likewise.
4960 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
4962 * sort.cc: New file.
4963 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
4964 * vec.c (qsort_chk): Use gcc_qsort.
4965 * Makefile.in (OBJS-libcommon): Add sort.o.
4966 (build/sort.o): New target. Use it...
4967 (BUILD_RTL): ... here, and...
4968 (build/gencfn-macros): ... here, and...
4969 (build/genmatch): ... here.
4971 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
4972 Chung-Ju Wu <jasonwucj@gmail.com>
4974 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
4975 * config/nds32/nds32-graywolf.md: New file.
4976 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
4977 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
4979 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
4980 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
4981 * config/nds32/nds32.md (pipeline_model): Add graywolf.
4982 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
4983 * config/nds32/pipelines.md: Include n15 settings.
4985 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
4986 Chung-Ju Wu <jasonwucj@gmail.com>
4988 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
4989 * config/nds32/nds32-n13.md: New file.
4990 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
4991 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
4993 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
4994 * config/nds32/nds32.md (pipeline_model): Add n13.
4995 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
4996 * config/nds32/pipelines.md: Include n13 settings.
4998 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
4999 Chung-Ju Wu <jasonwucj@gmail.com>
5001 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
5002 * config/nds32/nds32-n10.md: New file.
5003 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
5004 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
5006 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
5007 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
5008 * config/nds32/nds32.md (pipeline_model): Add n10.
5009 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
5010 * config/nds32/pipelines.md: Include n10 settings.
5012 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
5013 Kito Cheng <kito.cheng@gmail.com>
5014 Chung-Ju Wu <jasonwucj@gmail.com>
5016 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
5017 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
5018 Add enum values for DSP extension instructions.
5019 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
5021 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
5022 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
5024 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
5025 * config/nds32/nds32-dspext.md: New file for DSP implementation.
5026 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
5027 * config/nds32/nds32-intrinsic.md: Likewise.
5028 * config/nds32/nds32_intrinsic.h: Likewise.
5029 * config/nds32/nds32-md-auxiliary.c: Likewise.
5030 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
5031 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
5032 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
5033 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
5034 * config/nds32/nds32-protos.h: New declarations for DSP extension.
5035 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
5036 TYPE_DMAC in switch statement.
5037 * config/nds32/nds32.c: New checking and implementation for DSP
5038 extension instructions.
5039 * config/nds32/nds32.h: Likewise.
5040 * config/nds32/nds32.md: Likewise.
5041 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
5042 * config/nds32/predicates.md: Implement new predicates for DSP
5045 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
5047 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
5048 Reformat alternatives and attributes so it is easier to identify
5049 which constraints/attributes go with which instruction.
5050 (mov<mode>_hardfloat32, FMOVE64): Likewise.
5051 (mov<mode>_softfloat32, FMOVE64): Likewise.
5052 (mov<mode>_hardfloat64, FMOVE64): Likewise.
5053 (mov<mode>_softfloat64, FMOVE64): Likewise.
5055 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
5057 * doc/extend.texi (PowerPC Built-in Functions): Rename this
5059 (Basic PowerPC Built-in Functions): The new name of the
5060 subsection previously known as "PowerPC Built-in Functions".
5061 (Basic PowerPC Built-in Functions Available on all Configurations):
5063 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
5064 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
5065 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
5066 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
5068 2018-05-11 Martin Jambor <mjambor@suse.cz>
5071 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
5074 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
5077 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
5079 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
5081 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
5082 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
5083 (ix86_handle_option): Handle -mwaitpkg.
5084 * config.gcc: New header.
5085 * config/i386/cpuid.h (bit_WAITPKG): New bit.
5086 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
5087 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
5089 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
5090 OPTION_MASK_ISA_WAITPKG.
5091 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
5092 (ix86_option_override_internal): Add PTA_WAITPKG.
5093 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
5094 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
5095 IX86_BUILTIN_TPAUSE.
5096 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
5097 __builtin_ia32_umwait and __builtin_ia32_tpause.
5098 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
5099 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
5100 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
5101 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
5102 UNSPECV_TPAUSE): New.
5103 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
5104 * config/i386/i386.opt: Add -mwaitpkg.
5105 * config/i386/waitpkgintrin.h: New file.
5106 * config/i386/x86intrin.h: New header.
5107 * doc/invoke.texi: Add -mwaitpkg.
5109 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
5112 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
5114 (cortex-m0): Use armv6s-m isa.
5115 (cortex-m0plus): Likewise.
5116 (cortex-m1): Likewise.
5117 (cortex-m0.small-multiply): Likewise.
5118 (cortex-m0plus.small-multiply): Likewise.
5119 (cortex-m1.small-multiply): Likewise.
5121 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
5122 Jakub Jelinek <jakub@redhat.com>
5124 PR tree-optimization/85692
5125 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
5126 source permute as well.
5128 2018-05-11 Martin Liska <mliska@suse.cz>
5131 * doc/extend.texi: Document LLVM style format for no_sanitize
5134 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
5136 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
5137 mode_supports_vsx_dform_quad to mode_supports_dq_form.
5138 (mode_supports_vsx_dform_quad): Likewise.
5139 (mode_supports_vmx_dform): Move these functions to be next to the
5140 other mode_supports functions.
5141 (mode_supports_dq_form): Likewise.
5142 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
5143 mode_supports_dq_form.
5144 (reg_offset_addressing_ok_p): Likewise.
5145 (offsettable_ok_by_alignment): Likewise.
5146 (rs6000_legitimate_offset_address_p): Likewise.
5147 (legitimate_lo_sum_address_p): Likewise.
5148 (rs6000_legitimize_address): Likewise.
5149 (rs6000_legitimize_reload_address): Likewise.
5150 (rs6000_secondary_reload_inner): Likewise.
5151 (rs6000_preferred_reload_class): Likewise.
5152 (rs6000_output_move_128bit): Likewise.
5154 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
5156 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
5157 Generate SImode target register for null target.
5158 <case IX86_BUILTIN_XGETBV>: Ditto.
5159 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
5160 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
5162 2018-05-10 Carl Love <cel@us.ibm.com>
5164 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
5165 dcbtt and dcbtstt if operands[2] is 0.
5167 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
5170 * config/i386/sse.md (usadv64qi): New expander.
5172 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
5174 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
5175 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
5176 -maltivec=be support.
5177 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
5178 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
5179 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
5180 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
5181 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
5182 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
5183 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
5184 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
5185 altivec_vsumsws): Adjust.
5186 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
5187 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
5188 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
5190 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
5191 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
5192 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
5193 (altivec_lve<VI_char>x): Delete expand.
5194 (*altivec_lve<VI_char>x_internal): Rename to...
5195 (altivec_lve<VI_char>x): ... this.
5196 (altivec_lvxl_<mode>): Delete expand.
5197 (*altivec_lvxl_<mode>_internal): Rename to ...
5198 (altivec_lvxl_<mode>): ... this.
5199 (altivec_stvxl_<mode>): Delete expand.
5200 (*altivec_stvxl_<mode>_internal): Rename to ...
5201 (altivec_stvxl_<mode>): ... this.
5202 (altivec_stve<VI_char>x): Delete expand.
5203 (*altivec_stve<VI_char>x_internal): Rename to ...
5204 (altivec_stve<VI_char>x): ... this.
5205 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
5206 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
5207 reduc_plus_scal_<mode>): Adjust.
5208 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
5210 (rs6000_cpu_cpp_builtins): Adjust.
5211 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
5212 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
5213 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
5214 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
5215 -maltivec=be support.
5216 (rs6000_split_vec_extract_var): Adjust.
5217 (rs6000_split_v4si_init): Adjust.
5218 (swap_selector_for_mode): Delete.
5219 (altivec_expand_lvx_be, altivec_expand_stvx_be,
5220 altivec_expand_stvex_be): Delete.
5221 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
5222 -maltivec=be support.
5223 (rs6000_gimple_fold_builtin): Ditto.
5224 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
5226 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
5227 (TARGET_DIRECT_MOVE_64BIT): Adjust.
5228 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
5229 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
5230 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
5231 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
5232 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
5233 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
5234 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
5235 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
5236 anonymous split): Adjust.
5237 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
5238 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
5240 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
5242 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
5243 when --with-gxx-include-dir is also specified.
5244 * configure: Regenerate.
5246 2018-05-09 Jim Wilson <jimw@sifive.com>
5249 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
5250 * config/riscv/t-withmultilib: New.
5251 * config/riscv/withmultilib.h: New.
5252 * doc/install.texi: Document RISC-V --with-multilib-list support.
5254 2018-05-09 Richard Biener <rguenther@suse.de>
5256 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
5258 (vect_bb_vectorization_profitable_p): Adjust. Compute
5259 actual scalar cost using the cost vector and the add_stmt_cost
5262 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
5264 PR rtl-optimization/85645
5265 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
5266 in the REG_CFA_REGISTER note for LR, don't leave it empty.
5268 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
5270 PR rtl-optimization/85645
5271 * shrink-wrap.c (spread_components): Return a boolean saying if
5272 anything was changed.
5273 (try_shrink_wrapping_separate): Iterate spread_components until
5274 nothing changes anymore.
5276 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
5278 PR rtl-optimization/85645
5279 * regrename.c (build_def_use): Also kill the chains that include the
5280 destination of a REG_CFA_REGISTER note.
5282 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
5284 PR rtl-optimization/85645
5285 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
5286 insn that has a REG_CFA_REGISTER note.
5288 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
5290 * cfgexpand.c (expand_clobber): New function.
5291 (expand_gimple_stmt_1): Use it.
5292 * tree-vect-stmts.c (vect_clobber_variable): New function,
5294 (vectorizable_simd_clone_call): ...here.
5295 (vectorizable_store): Emit a clobber either side of an
5296 IFN_STORE_LANES sequence.
5297 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
5299 2018-05-09 Tom de Vries <tom@codesourcery.com>
5302 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
5303 (define_insn "trap_if_false"): Add exit after trap.
5305 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
5307 PR rtl-optimization/85638
5308 * bb-reorder.c: Include common/common-target.h.
5309 (create_forwarder_block): New function extracted from...
5310 (fix_up_crossing_landing_pad): ...here. Rename into...
5311 (dw2_fix_up_crossing_landing_pad): ...this.
5312 (sjlj_fix_up_crossing_landing_pad): New function.
5313 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
5314 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
5315 from both partitions and exit the loop after one iteration.
5317 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
5320 * doc/extend.texi (PowerPC Built-in Functions): Rename this
5322 (Basic PowerPC Built-in Functions): The new name of the
5323 subsection previously known as "PowerPC Built-in Functions".
5324 (Basic PowerPC Built-in Functions Available on all Configurations):
5326 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
5328 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
5330 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
5332 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
5335 2018-05-08 Jim Wilson <jimw@sifive.com>
5337 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
5338 (LD_EMUL_SUFFIX): New.
5339 (LINK_SPEC): Use it.
5341 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
5343 * doc/extend.texi (PowerPC Built-in Functions): Rename this
5345 (Basic PowerPC Built-in Functions): The new name of the
5346 subsection previously known as "PowerPC Built-in Functions".
5347 (Basic PowerPC Built-in Functions Available on all Configurations):
5349 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
5351 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
5353 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
5355 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
5358 2018-05-08 Jakub Jelinek <jakub@redhat.com>
5361 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
5362 after cmpelim optimization.
5364 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
5366 * config.gcc: Support "goldmont".
5367 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
5368 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
5370 * config/i386/i386.c (m_GOLDMONT): Define.
5371 (processor_target_table): Add "goldmont".
5372 (PTA_GOLDMONT): Define.
5373 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
5374 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
5375 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
5376 (fold_builtin_cpu): Add "goldmont".
5377 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
5378 (ix86_option_override_internal): Add "goldmont".
5379 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
5380 (processor_type): Add PROCESSOR_GOLDMONT.
5381 * config/i386/i386.md: Add CPU "glm".
5382 * config/i386/glm.md: New file.
5383 * config/i386/x86-tune.def: Add m_GOLDMONT.
5384 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
5386 2018-05-08 Jakub Jelinek <jakub@redhat.com>
5389 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
5391 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
5392 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
5393 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
5396 * config/i386/i386.c (ix86_fold_builtin): Handle
5397 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
5400 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
5401 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
5403 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
5406 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
5407 (check_arch): Likewise.
5408 (check_fpu): Return the result rather than printing it.
5409 (end arch): Fix operator precedence.
5410 (end cpu): Likewise.
5411 (END): Print the result from check_fpu.
5413 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
5414 Alan Hayward <alan.hayward@arm.com>
5415 David Sherwood <david.sherwood@arm.com>
5417 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
5418 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
5419 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
5420 (*fcmuo<mode>_and): New patterns.
5422 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
5424 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
5425 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
5426 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
5427 (cmp_op, sve_imm_con): New code attributes.
5428 (SVE_COND_INT_CMP, imm_con): Delete.
5429 (cmp_op): Remove above unspecs from int attribute.
5430 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
5432 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
5433 comparison-specific unspecs.
5434 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
5435 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
5436 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
5437 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
5438 (*vec_fcm<cmp_op><mode>): Rename to...
5439 (*fcm<cmp_op><mode>): ...this and adjust likewise.
5440 (*vec_fcmuo<mode>): Rename to...
5441 (*fcmuo<mode>): ...this and adjust likewise.
5442 (*pred_fcm<cmp_op><mode>): New pattern.
5443 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
5444 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
5446 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
5448 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
5449 (aarch64_emit_sve_predicated_cond): New function.
5450 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
5451 (aarch64_emit_unspec_cond_or): Replace with...
5452 (aarch64_emit_sve_or_conds): ...this new function. Use
5453 aarch64_emit_sve_ptrue_op for the individual comparisons and
5454 aarch64_emit_binop to OR them together.
5455 (aarch64_emit_inverted_unspec_cond): Replace with...
5456 (aarch64_emit_sve_inverted_cond): ...this new function. Use
5457 aarch64_emit_sve_ptrue_op for the comparison and
5458 aarch64_emit_unop to invert the result.
5459 (aarch64_expand_sve_vec_cmp_float): Update after the above
5460 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
5462 2018-05-07 Nathan Sidwell <nathan@acm.org>
5464 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
5465 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
5466 (Backwards Compatibility): Likewise.
5468 2018-05-07 Luis Machado <luis.machado@linaro.org>
5472 2018-05-07 Luis Machado <luis.machado@linaro.org>
5474 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5475 <prefetch_dynamic_strides>: New const bool field.
5476 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5477 prefetch_dynamic_strides.
5478 (exynosm1_prefetch_tune): Likewise.
5479 (thunderxt88_prefetch_tune): Likewise.
5480 (thunderx_prefetch_tune): Likewise.
5481 (thunderx2t99_prefetch_tune): Likewise.
5482 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
5484 (aarch64_override_options_internal): Update to set
5485 PARAM_PREFETCH_DYNAMIC_STRIDES.
5486 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
5487 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
5488 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
5489 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
5490 prefetch-dynamic-strides setting.
5492 2018-05-07 Luis Machado <luis.machado@linaro.org>
5494 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5495 <minimum_stride>: New const int field.
5496 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5497 minimum_stride field.
5498 (exynosm1_prefetch_tune): Likewise.
5499 (thunderxt88_prefetch_tune): Likewise.
5500 (thunderx_prefetch_tune): Likewise.
5501 (thunderx2t99_prefetch_tune): Likewise.
5502 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
5503 (aarch64_override_options_internal): Update to set
5504 PARAM_PREFETCH_MINIMUM_STRIDE.
5505 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
5506 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
5507 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
5508 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
5509 stride is constant and is below the minimum stride threshold.
5511 2018-05-07 Luis Machado <luis.machado@linaro.org>
5513 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
5516 2018-05-07 Luis Machado <luis.machado@linaro.org>
5518 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5519 <prefetch_dynamic_strides>: New const bool field.
5520 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5521 prefetch_dynamic_strides.
5522 (exynosm1_prefetch_tune): Likewise.
5523 (thunderxt88_prefetch_tune): Likewise.
5524 (thunderx_prefetch_tune): Likewise.
5525 (thunderx2t99_prefetch_tune): Likewise.
5526 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
5528 (aarch64_override_options_internal): Update to set
5529 PARAM_PREFETCH_DYNAMIC_STRIDES.
5530 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
5531 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
5532 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
5533 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
5534 prefetch-dynamic-strides setting.
5536 2018-05-07 Luis Machado <luis.machado@linaro.org>
5538 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5539 <minimum_stride>: New const int field.
5540 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5541 minimum_stride field.
5542 (exynosm1_prefetch_tune): Likewise.
5543 (thunderxt88_prefetch_tune): Likewise.
5544 (thunderx_prefetch_tune): Likewise.
5545 (thunderx2t99_prefetch_tune): Likewise.
5546 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
5547 (aarch64_override_options_internal): Update to set
5548 PARAM_PREFETCH_MINIMUM_STRIDE.
5549 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
5550 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
5551 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
5552 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
5553 stride is constant and is below the minimum stride threshold.
5555 2018-05-06 Jakub Jelinek <jakub@redhat.com>
5558 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
5559 the type is addressable. Don't force op into register if it has
5562 2018-05-05 Roland McGrath <mcgrathr@google.com>
5565 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
5566 any section for which we don't know a specific type it should have,
5567 regardless of name. Previously this was done only for the exact
5568 names ".init_array", ".fini_array", and ".preinit_array".
5569 (default_elf_asm_named_section): Add comment about
5570 relationship with default_section_type_flags and SECTION_NOTYPE.
5571 (get_section): Don't consider it a type conflict if one side has
5572 SECTION_NOTYPE and the other doesn't, as long as neither has the
5573 SECTION_BSS et al used in the default_section_type_flags logic.
5575 2018-05-05 Tom de Vries <tom@codesourcery.com>
5578 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
5579 (workaround_barsyncs): New function.
5580 (nvptx_reorg): Use workaround_barsyncs.
5581 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
5582 (define_expand "nvptx_membar_cta"): New define_expand.
5583 (define_insn "*nvptx_membar_cta"): New insn.
5585 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
5587 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
5588 To improve optimization opportunities.
5589 * builtin-types.def: The new needed builtin types for the above.
5591 2018-05-04 Richard Biener <rguenther@suse.de>
5593 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
5594 * gimple-ssa-store-merging.c
5595 (imm_store_chain_info::output_merged_store): Remove redundant create,
5596 release split_store vector contents on failure.
5597 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
5598 scalar stmt vector on cache hit.
5600 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
5602 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
5604 * config.gcc (powerpc-xilinx-eabi*): Remove.
5605 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
5607 (fusion_addis_mem_combo_load): Ditto.
5608 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
5610 (rs6000_cpu_cpp_builtins): Ditto.
5611 * config/rs6000/rs6000-linux.c
5612 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
5613 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
5614 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
5616 (rs6000_setup_reg_addr_masks): Ditto.
5617 (rs6000_init_hard_regno_mode_ok): Ditto.
5618 (rs6000_option_override_internal): Ditto.
5619 (legitimate_lo_sum_address_p): Ditto.
5620 (rs6000_legitimize_address): Ditto.
5621 (rs6000_legitimize_reload_address): Ditto.
5622 (rs6000_legitimate_address_p): Ditto.
5623 (abi_v4_pass_in_fpr): Ditto.
5624 (setup_incoming_varargs): Ditto.
5625 (rs6000_gimplify_va_arg): Ditto.
5626 (rs6000_split_multireg_move): Ditto.
5627 (rs6000_savres_strategy): Ditto.
5628 (rs6000_emit_prologue_components): Ditto.
5629 (rs6000_emit_epilogue_components): Ditto.
5630 (rs6000_emit_prologue): Ditto.
5631 (rs6000_emit_epilogue): Ditto.
5632 (rs6000_elf_file_end): Ditto.
5633 (rs6000_function_value): Ditto.
5634 (rs6000_libcall_value): Ditto.
5635 * config/rs6000/rs6000.h: Ditto.
5636 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
5637 (TARGET_MINMAX): ... this. New.
5638 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
5639 * config/rs6000/rs6000.md: Remove Xilinx FP support.
5640 (*movsi_internal1_single): Delete.
5641 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
5642 mfpu=, mxilinx-fpu): Delete.
5643 * config/rs6000/singlefp.h: Delete.
5644 * config/rs6000/sysv4.h: Remove Xilinx FP support.
5645 * config/rs6000/t-rs6000: Ditto.
5646 * config/rs6000/t-xilinx: Delete.
5647 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
5648 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
5649 (VStype_simple): Delete.
5650 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
5651 * config/rs6000/xfpu.h: Delete.
5652 * config/rs6000/xfpu.md: Delete.
5653 * config/rs6000/xilinx.h: Delete.
5654 * config/rs6000/xilinx.opt: Delete.
5655 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
5656 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
5658 2018-05-04 Tom de Vries <tom@codesourcery.com>
5661 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
5664 2018-05-04 Richard Biener <rguenther@suse.de>
5667 * tree-complex.c (update_complex_assignment): We are always in SSA form.
5668 (expand_complex_div_wide): Likewise.
5669 (expand_complex_operations_1): Likewise.
5670 (expand_complex_libcall): Preserve EH info of the original stmt.
5671 (tree_lower_complex): Handle removed blocks.
5672 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
5673 on complex multiplication and division libcall builtins.
5675 2018-05-04 Richard Biener <rguenther@suse.de>
5678 * fold-const.c (negate_expr_p): Restrict negation of operand
5679 zero of a division to when we know that can happen without
5681 (fold_negate_expr_1): Likewise.
5683 2018-05-04 Jakub Jelinek <jakub@redhat.com>
5686 * real.h (real_nextafter): Declare.
5687 * real.c (real_nextafter): New function.
5688 * fold-const-call.c (fold_const_nextafter): New function.
5689 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
5690 CASE_CFN_NEXTTOWARD.
5691 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
5692 even when arg1_mode is different from arg0_mode.
5694 2018-05-03 Nathan Sidwell <nathan@acm.org>
5696 * doc/extend.texi (Deprecated Features): Remove
5698 (Backwards Compatibility): Likewise.
5699 * doc/invoke.texi (C++ Language Options): Likewise.
5700 (C++ Dialect Options): Likewise.
5702 2018-05-03 Jakub Jelinek <jakub@redhat.com>
5705 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
5706 _mm512_mask_mullox_epi64): New intrinsics.
5708 2018-05-03 Tom de Vries <tom@codesourcery.com>
5711 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
5712 dump files): Add offload-tree.
5714 2018-05-03 Richard Biener <rguenther@suse.de>
5716 PR tree-optimization/85615
5717 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
5718 to loops not nested in BBs loop father to avoid creating multi-entry
5721 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5723 PR tree-optimization/70291
5724 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
5725 arguments. Change return type to tree. Emit libcall as a new
5726 statement rather than replacing existing one when inplace_p is true.
5727 (expand_complex_multiplication_components): New function.
5728 (expand_complex_multiplication): Expand floating-point complex
5729 multiplication using the above.
5730 (expand_complex_division): Rename inner_type parameter to type.
5731 Update expand_complex_libcall call-site.
5732 (expand_complex_operations_1): Update expand_complex_multiplication
5733 and expand_complex_division call-sites.
5735 2018-05-02 Jakub Jelinek <jakub@redhat.com>
5738 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
5739 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
5740 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
5741 the highest significant bit of the shift count mask is clear. In
5742 check whether and[sq]i3 is needed verify that all significant bits
5743 of the shift count other than the highest are set.
5745 2018-05-02 Tom de Vries <tom@codesourcery.com>
5748 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
5749 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
5750 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
5751 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
5752 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
5753 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
5754 __builtin_goacc_parlevel_size.
5756 2018-05-02 Richard Biener <rguenther@suse.de>
5758 PR tree-optimization/85597
5759 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
5760 do not use split vect_get_vec_defs call but call vect_get_slp_defs
5763 2018-05-02 Tom de Vries <tom@codesourcery.com>
5766 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
5767 dump files): Add ltrans-tree.
5769 2018-05-02 Tom de Vries <tom@codesourcery.com>
5772 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
5773 dump files): Add wpa-ipa.
5775 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
5777 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
5778 powerpc*-*-linux*paired* target.
5779 * config/rs6000/750cl.h: Delete.
5780 * config/rs6000/paired.h: Delete.
5781 * config/rs6000/paired.md: Delete.
5782 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
5784 * config/rs6000/rs6000-builtin.def: Remove paired float support.
5785 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
5786 comment. Remove paired float support.
5787 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
5788 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
5790 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
5791 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
5793 * config/rs6000/rs6000.c: Remove paired float support.
5794 (paired_expand_vector_init, paired_expand_vector_move,
5795 paired_emit_vector_compare, paired_emit_vector_cond_expr,
5796 (paired_expand_lv_builtin, paired_expand_stv_builtin,
5797 paired_expand_builtin, paired_expand_predicate_builtin,
5798 paired_init_builtins): Delete.
5799 * config/rs6000/rs6000.h: Remove paired float support.
5800 * config/rs6000/rs6000.md: Remove paired float support.
5801 (move_from_CR_ov_bit): Delete.
5802 * config/rs6000/rs6000.opt (mpaired): Delete.
5803 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
5804 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
5806 2018-05-02 Richard Biener <rguenther@suse.de>
5809 * gimplify.c (gimplify_save_expr): When in SSA form allow
5810 SAVE_EXPRs to compute to SSA vars.
5812 2018-05-02 Jakub Jelinek <jakub@redhat.com>
5815 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
5816 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
5817 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
5818 clobber operands[2], instead use a new pseudo. Formatting fixes.
5820 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
5822 PR tree-optimization/85586
5823 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
5824 exit early for statements in the same group if the accesses are
5827 2018-05-02 Tom de Vries <tom@codesourcery.com>
5830 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
5833 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
5835 PR tree-optimization/85143
5836 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
5838 2018-05-01 Tom de Vries <tom@codesourcery.com>
5841 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
5842 not found" error message.
5844 2018-05-01 Tom de Vries <tom@codesourcery.com>
5847 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
5848 * vec.c (test_ordered_remove_if): New function.
5849 (vec_c_tests): Call test_ordered_remove_if.
5850 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
5851 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
5852 * tree-vect-patterns.c (vect_pattern_recog_1): Use
5853 VEC_ORDERED_REMOVE_IF.
5855 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5857 PR tree-optimization/82665
5858 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
5859 pointer subtraction where arguments come from a memchr call.
5861 2018-05-01 Jakub Jelinek <jakub@redhat.com>
5863 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
5864 --push-state --as-needed and --pop-state instead of --as-needed and
5865 --no-as-needed if ld supports it.
5866 * configure: Regenerated.
5869 * doc/install.texi2html: Replace _002d with - and _002a with * in
5870 generated html files using sed.
5872 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5875 * gcc-rich-location.c (blank_line_before_p): New function.
5876 (use_new_line): New function.
5877 (gcc_rich_location::add_fixit_insert_formatted): New function.
5878 * gcc-rich-location.h
5879 (gcc_rich_location::add_fixit_insert_formatted): New function.
5881 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5883 * selftest.c (assert_streq): Rename "expected" and "actual" to
5884 "val1" and "val2". Extend NULL-handling to cover both inputs
5885 symmetrically, while still requiring both to be non-NULL for a pass.
5886 * selftest.h (assert_streq): Rename "expected" and "actual" to
5888 (ASSERT_EQ): Likewise.
5889 (ASSERT_EQ_AT): Likewise.
5890 (ASSERT_KNOWN_EQ): Likewise.
5891 (ASSERT_KNOWN_EQ_AT): Likewise.
5892 (ASSERT_NE): Likewise.
5893 (ASSERT_MAYBE_NE): Likewise.
5894 (ASSERT_MAYBE_NE_AT): Likewise.
5895 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
5896 the assertion to pass.
5897 (ASSERT_STREQ_AT): Likewise.
5899 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
5901 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
5902 interaction with -pie.
5904 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5906 * selftest.h: Fix alphabetization of per-source-file selftest
5909 2018-04-30 Jason Merrill <jason@redhat.com>
5911 PR c++/61982 - dead stores to destroyed objects.
5912 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
5915 2018-04-30 Jason Merrill <jason@redhat.com>
5917 * tree.c (build_clobber): New.
5918 * tree.h: Declare it.
5919 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
5921 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5923 * diagnostic-show-locus.c (layout::layout): Update for
5924 location_get_source_line returning a char_span.
5925 (struct char_span): Move to input.h.
5926 (struct correction): Update for fields in char_span becoming
5928 (struct source_line): Update for location_get_source_line
5929 returning a char_span.
5930 (layout::print_line): Likewise.
5931 * edit-context.c (edited_file::print_content): Likewise.
5932 (edited_file::print_diff_hunk): Likewise.
5933 (edited_file::print_run_of_changed_lines): Likewise.
5934 (edited_file::get_num_lines): Likewise.
5935 (edited_line::edited_line): Likewise.
5936 * final.c (asm_show_source): Likewise.
5937 * input.c (location_get_source_line): Convert return type
5938 from const char * to char_span, losing the final "line_len"
5940 (dump_location_info): Update for the above.
5941 (get_substring_ranges_for_loc): Likewise. Use a char_span
5942 when handling the literal within the line.
5943 (test_reading_source_line): Update for location_get_source_line
5944 returning a char_span.
5945 * input.h (class char_span): Move here from
5946 diagnostic-show-locus.c, converting from a struct to a class.
5947 Make data members private.
5948 (char_span::operator bool): New.
5949 (char_span::length): New.
5950 (char_span::get_buffer): New.
5951 (char_span::operator[]): New.
5952 (char_span::subspan): Make const.
5953 (char_span::xstrdup): New.
5954 (location_get_source_line): Convert return type from const char *
5955 to char_span, losing the final "line_size" param.
5957 2018-04-30 Jan Hubicka <jh@suse.cz>
5959 * lto-wrapper.c (ltrans_priorities): New static var.
5960 (cmp_priority): New.
5961 (run_gcc): Read priorities and if doing parallel build order
5962 the Makefile by them.
5964 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5966 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
5968 2018-04-30 Richard Biener <rguenther@suse.de>
5970 * tree-cfg.c (verify_address): Remove base argument, add
5971 flag whether to check TREE_ADDRESSABLE and do that.
5972 (verify_expr): Remove.
5973 (verify_types_in_gimple_reference): Add pieces from verify_expr.
5974 (verify_gimple_assign_single): Likewise.
5975 (verify_gimple_switch): Likewise.
5976 (verify_expr_location_1): Dereference tp once. Add (disabled)
5977 piece from verify_expr.
5978 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
5980 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
5982 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
5984 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
5986 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
5987 (small_data_pattern): Likewise.
5988 (arc_rewrite_small_data): Likewise.
5989 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
5990 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
5991 (get_symbol_alignment): New function.
5992 (legitimate_small_data_address_p): Likewise.
5993 (legitimate_scaled_address): Update, call
5994 legitimate_small_data_address_p.
5995 (output_sdata): New static variable.
5996 (arc_print_operand): Update how we handle small data operands.
5997 (arc_print_operand_address): Likewise.
5998 (arc_legitimate_address_p): Update, use
5999 legitimate_small_data_address_p.
6000 (arc_rewrite_small_data_p): Remove.
6001 (arc_rewrite_small_data_1): Likewise.
6002 (arc_rewrite_small_data): Likewise.
6003 (small_data_pattern): Likewise.
6004 (compact_sda_memory_operand): Update to use
6005 legitimate_small_data_address_p and get_symbol_alignment.
6006 (prepare_move_operands): Don't rewite sdata pattern.
6007 (prepare_extend_operands): Remove.
6008 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
6010 (zero_extendqisi2): Likewise.
6011 (zero_extendhisi2): Likewise.
6012 (extendqihi2): Likewise.
6013 (extendqisi2): Likewise.
6014 (extendhisi2): Likewise.
6018 * config/arc/constraints.md (Usd): Change it to memory constraint.
6020 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
6022 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
6023 as source of std instructions.
6024 * config/arc/arc.md (movsi_insn): Update pattern predicate to
6025 allow 6-bit constants as source for store instructions.
6026 (movdi_insn): Update instruction pattern to allow 6-bit constants
6027 as source for store instructions.
6029 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
6031 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
6033 2018-04-30 Nathan Sidwell <nathan@acm.org>
6034 Sandra Loosemore <sandra@codesourcery.com>
6036 * dumpfile.c (dump_open): Allow '-' for stdout.
6037 * doc/invoke.texi (Developer Options): Document dump filename
6038 determination early. Document stdin/stdout selection.
6040 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
6042 Microblaze Target: PIC data text relative
6044 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
6045 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
6047 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
6048 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
6049 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
6050 New addressing mode for data-text relative position indepenedent code.
6051 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
6052 'ADDRESS_SYMBOLIC_TXT_REL'.
6053 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
6054 (microblaze_legitimate_pic_operand): Exclude function calls from
6055 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
6056 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
6058 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
6059 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
6060 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
6061 for 'address + offset'.
6062 (microblaze_expand_prologue): Add new function prologue call for
6064 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
6065 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
6066 table in case of TARGET_PIC_DATA_TEXT_REL.
6067 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
6068 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
6069 Add new macros 'UNSPEC_TEXT',
6070 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
6071 + exclude function calls from 'UNSPEC_PLT' in case of data text
6073 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
6074 new target hook for generating address diff vector tables in case of
6076 * doc/tm.texi : Regenerate.
6077 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
6078 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
6079 of addr diff vector generation.
6080 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
6081 target hook definition.
6082 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
6083 Add default function for generate_pic_addr_diff_vec -> flag_pic.
6084 * doc/invoke.texi (Add new pic option): Add new microblaze pic
6085 option for data text relative.
6087 2018-04-30 Richard Biener <rguenther@suse.de>
6089 * tree-chrec.h (evolution_function_is_constant_p): Remove
6091 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
6093 2018-04-30 Richard Biener <rguenther@suse.de>
6096 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
6098 2018-04-30 Richard Biener <rguenther@suse.de>
6100 PR tree-optimization/28364
6101 PR tree-optimization/85275
6102 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
6103 copying first exit test.
6105 2018-04-28 Mark Wielaard <mark@klomp.org>
6107 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
6109 (dwarf_AT): Handle DW_AT_addr_base.
6110 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
6112 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
6115 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
6116 (*ashl<dwi>3_doubleword_mask_1): Ditto.
6117 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
6118 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
6120 2018-04-28 Richard Biener <rguenther@suse.de>
6122 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
6123 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
6124 to reflect use. Only add interesting stmts.
6126 2018-04-27 Martin Jambor <mjambor@suse.cz>
6129 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
6130 the jump function allows for passing through aggregate values.
6132 2018-04-27 David Malcolm <dmalcolm@redhat.com>
6134 * input.h (in_system_header_at): Convert from macro to inline
6136 (from_macro_expansion_at): Likewise.
6137 (from_macro_definition_at): Likewise.
6139 2018-04-27 Jeff Law <law@redhat.com>
6141 * config.gcc: Mark tile* targets as deprecated/obsolete.
6143 2018-04-27 Richard Biener <rguenther@suse.de>
6145 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
6148 2018-04-27 Richard Biener <rguenther@suse.de>
6150 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
6152 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
6154 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
6155 with Yd constraint. Set "preferred_for_speed" attribute from
6156 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
6158 (*movdi_internal): Ditto.
6159 (movti_interunit splitters): Remove
6160 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
6161 (movdi_interunit splitters): Ditto.
6162 * config/i386/constraints.md (Ye): Remove.
6163 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
6165 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6168 * config/aarch64/constraints.md (Usg): Limit to 31.
6171 2018-04-27 Jakub Jelinek <jakub@redhat.com>
6173 PR tree-optimization/85529
6174 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
6175 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
6176 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
6177 zero extension or masking of the MSB bit.
6178 (optimize_range_tests): Add FIRST_BB argument, pass it through
6179 to optimize_range_tests_var_bound.
6180 (maybe_optimize_range_tests, reassociate_bb): Adjust
6181 optimize_range_tests callers.
6183 2018-04-26 Richard Biener <rguenther@suse.de>
6184 Jakub Jelinek <jakub@redhat.com>
6186 * cgraph.h (symbol_table): Just declare debug method here.
6187 * symtab.c (symbol_table::debug): Define.
6189 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
6191 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
6193 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
6195 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
6196 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
6197 (*movdi_internal): Substitute Yi and Yj constraint with x
6198 and Ym and Yn constraint with y constraint. Update "isa"
6199 attribute and set "preferred_for_speed" attribute from
6200 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
6201 (*movsi_internal): Ditto.
6202 (*movdf_internal): Ditto.
6203 (*movsf_internal): Ditto.
6204 (*zero_extendsidi2): Ditto.
6205 * config/i386/sse.md (vec_set<mode>_0): Ditto.
6206 (sse2_loadld): Ditto.
6207 (*vec_extract<ssevecmodelower>_0): Ditto.
6208 (*vec_extractv4si_0_zext_sse4): Ditto.
6209 (vec_concatv2di): Ditto.
6210 (*vec_dup<mode>): Ditto.
6211 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
6212 * config/i386/constraints.md (Yi): Remove.
6217 2018-04-26 Nathan Sidwell <nathan@acm.org>
6219 * dumpfile.c (dump_open): New.
6220 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
6221 (dump_finish): Detect stdio/stderr by value not name.
6223 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
6225 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
6227 2018-04-26 Tom de Vries <tom@codesourcery.com>
6230 * config/nvptx/nvptx.c (verify_neutering_jumps)
6231 (verify_neutering_labels): New function
6232 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
6234 2018-04-26 Tom de Vries <tom@codesourcery.com>
6237 * config/nvptx/nvptx.c (needs_neutering_p): New function.
6238 (nvptx_single): Use needs_neutering_p to skip over insns that do not
6241 2018-04-26 Richard Biener <rguenther@suse.de>
6242 Tom de Vries <tom@codesourcery.com>
6245 * lto-streamer-out.c (output_function): Fixup loops if required to match
6246 discovery done in the reader.
6248 2018-04-26 Richard Biener <rguenther@suse.de>
6250 PR tree-optimization/85116
6251 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
6252 have a loop exit from the single latch predecessor. Remove
6253 case of header with just condition.
6254 (ch_base::copy_headers): Exclude infinite loops from any
6256 (pass_ch::execute): Record exits.
6258 2018-04-26 Richard Biener <rguenther@suse.de>
6260 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
6261 prologue cost vector and pass it to vect_get_load_cost.
6262 (vect_get_peeling_costs_all_drs): Likewise.
6263 (vect_peeling_hash_get_lowest_cost): Likewise.
6264 (vect_enhance_data_refs_alignment): Likewise.
6266 2018-04-26 Richard Biener <rguenther@suse.de>
6269 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
6270 checking of integer<->pointer conversions.
6271 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
6272 sign-/zero-extending pointer types.
6273 (expand_omp_for_static_chunk): Likewise.
6275 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
6276 Jean Lee <xiaoyur347@gmail.com>
6278 * config/mips/mips.c (mips_asan_shadow_offset): New function.
6279 (TARGET_ASAN_SHADOW_OFFSET): Define.
6280 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
6281 true for -fsanitize=address.
6283 2018-04-25 Mark Wielaard <mark@klomp.org>
6285 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
6288 2018-04-25 Jakub Jelinek <jakub@redhat.com>
6290 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
6291 than "alu", remove explicit "memory" and "imm_disp" attributes.
6292 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
6295 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
6296 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
6297 gen_lowpart_no_emit.
6299 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
6302 * config/i386/i386.c (ix86_expand_builtin): Change memory
6303 operand to XI, extend p0 to Pmode.
6304 * config/i386/i386.md: Change unspec volatile and operand
6305 1 mode to XI, change operand 0 mode to P.
6307 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
6309 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
6310 GET_MODE_MASK before any checking.
6311 (nds32_can_use_bset_p): Likewise.
6312 (nds32_can_use_btgl_p): Likewise.
6314 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
6316 * config/nds32/nds32-doubleword.md: New define_split pattern for
6317 illegal register number.
6319 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
6321 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
6323 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
6325 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
6327 2018-04-25 Richard Biener <rguenther@suse.de>
6329 * lto-streamer.h (LTO_major_version): Bump to 8.
6331 2018-04-25 Jakub Jelinek <jakub@redhat.com>
6333 * BASE-VER: Set to 9.0.0.
6335 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
6337 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
6338 in __abskf2 and __powikf2.
6340 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6343 * config/aarch64/constraints.md (Usg, Usj): New constraints.
6344 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
6345 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
6346 Use the above on operand 2. Reindent.
6347 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
6349 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
6352 * common/config/i386/i386-common.c (ix86_handle_option): Don't
6354 * config/i386/i386.opt (mcet): Removed.
6355 * doc/install.texi: Remove -mcet documentation.
6356 * doc/invoke.texi: Likewise.
6358 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
6361 * doc/install.texi: Remove -mcet from bootstrap-cet.
6363 2018-04-24 Jakub Jelinek <jakub@redhat.com>
6366 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
6367 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
6371 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
6372 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
6373 containing a CONST_VECTOR.
6375 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
6377 * doc/install.texi: Update newlib dependency for nvptx.
6379 2018-04-24 Jakub Jelinek <jakub@redhat.com>
6382 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
6383 instead of INTVAL when shifting x left.
6385 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
6387 PR tree-optimization/85478
6388 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
6389 vect_grouped_store_supported for single element vectors.
6391 2018-04-24 Richard Biener <rguenther@suse.de>
6394 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
6395 load cost increase to the case of non-constant step.
6397 2018-04-24 Jakub Jelinek <jakub@redhat.com>
6400 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
6401 destination if any_malformed_asm.
6403 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
6406 * expr.c (store_field): In the bitfield case, if the value comes from
6407 a function call and is returned in registers by means of a PARALLEL,
6408 do not change the mode of the temporary unless BLKmode and VOIDmode.
6410 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
6412 PR rtl-optimization/85423
6413 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
6414 dependencies to debug insns when the previous insn is non-debug.
6416 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
6418 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
6419 enums into a single definition.
6420 (fls): Fix predicates and printing.
6423 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
6425 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
6426 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
6427 and short u6 immediate.
6428 (check_if_valid_sleep_operand): Remove.
6429 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
6431 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
6433 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
6434 flag_always_save_lp condition.
6435 * config/nds32/nds32.opt (malways-save-lp): New option.
6437 2018-04-22 Shiva Chen <shiva0217@gmail.com>
6439 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
6440 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
6441 * config/nds32/nds32.h
6442 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
6443 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
6445 2018-04-22 Shiva Chen <shiva0217@gmail.com>
6447 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
6448 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
6450 2018-04-22 Shiva Chen <shiva0217@gmail.com>
6451 Chung-Ju Wu <jasonwucj@gmail.com>
6453 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
6455 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
6456 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
6458 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
6460 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
6462 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
6464 * config/nds32/nds32-protos.h (nds32_data_alignment,
6465 nds32_local_alignment): Declare.
6466 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
6467 nds32_local_alignment): New functions.
6468 (TARGET_CONSTANT_ALIGNMENT): Define.
6469 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
6471 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
6473 * config/nds32/nds32.c
6474 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
6475 (TARGET_MODES_TIEABLE_P): Likewise.
6477 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
6479 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
6482 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
6483 Chung-Ju Wu <jasonwucj@gmail.com>
6485 * config/nds32/constants.md (unspec_volatile_element): Add enum values
6486 for unaligned access.
6487 * config/nds32/nds32-intrinsic.c: Implementation of expanding
6489 * config/nds32/nds32-intrinsic.md: Likewise.
6490 * config/nds32/nds32_intrinsic.h: Likewise.
6491 * config/nds32/nds32.h (nds32_builtins): Likewise.
6492 * config/nds32/nds32.opt (munaligned-access): New option.
6493 * config/nds32/nds32.c (nds32_asm_file_start): Display
6494 flag_unaligned_access status.
6496 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
6498 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
6499 -mno-relax is present.
6500 * config/riscv/linux.h (LINK_SPEC): Ditto.
6502 2018-04-20 Martin Sebor <msebor@redhat.com>
6505 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
6507 (gimple_fold_builtin_stxcpy_chk): Same.
6508 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
6510 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
6513 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
6514 __powikf2 when long double is IEEE 128-bit.
6516 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
6518 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
6519 step to make sure stack always aligned.
6521 2018-04-20 Carl Love <cel@us.ibm.com>
6524 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
6525 size check for arg0.
6527 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
6528 Tom de Vries <tom@codesourcery.com>
6531 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
6532 Emit insns for calls too.
6533 (nvptx_find_par): Always look for worker-level predecessor insn.
6534 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
6536 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
6537 (nvptx_process_pars): Propagate frames for calls.
6539 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
6542 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
6544 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
6545 (ix86_handle_option): Don't handle OPT_mibt.
6546 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
6548 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
6550 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
6551 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
6552 (ix86_target_macros): Define __CET__ with flag_cf_protection
6553 for -fcf-protection.
6554 * config/i386/i386.c (isa2_opts): Remove -mibt.
6555 * config/i386/i386.h (TARGET_IBT): Removed.
6556 (TARGET_IBT_P): Likewise.
6557 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
6558 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
6559 * config/i386/i386.opt (mcet): Update help message.
6562 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
6563 -mcet as an alias for -mshstk.
6565 2018-04-20 Richard Biener <rguenther@suse.de>
6568 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
6569 complexity by forcing a single use of the multiply operand.
6571 2018-04-20 Martin Jambor <mjambor@suse.cz>
6574 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
6575 recursion dependency to only apply to non-clones.
6577 2018-04-20 Martin Jambor <mjambor@suse.cz>
6580 * ipa-cp.c (create_specialized_node): Check that clones of
6581 self-recursive edges exist during IPA-CP.
6583 2018-04-19 Toon Moene <toon@moene.org>
6585 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
6588 2018-04-19 Jakub Jelinek <jakub@redhat.com>
6590 PR tree-optimization/85467
6591 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
6592 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
6593 VECTOR_CST element to type.
6595 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
6598 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
6599 * config/i386/i386.md (builtin_setjmp_setup): Removed.
6600 (builtin_longjmp): Likewise.
6601 (save_stack_nonlocal): New pattern.
6602 (restore_stack_nonlocal): Likewise.
6604 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
6607 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
6608 Replace ASM_OUTPUT_LABEL with fprintf.
6610 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
6613 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
6614 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
6615 * config/i386/i386-c.c (ix86_target_macros_internal): Also
6616 define __IBT__ and __SHSTK__ for -fcf-protection.
6617 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
6619 (ix86_trampoline_init): Likewise.
6620 (x86_output_mi_thunk): Likewise.
6621 (ix86_notrack_prefixed_insn_p): Likewise.
6622 (ix86_option_override_internal): Don't disallow -fcf-protection.
6623 * config/i386/i386.md (rdssp<mode>): Also enable for
6625 (incssp<mode>): Likewise.
6626 (nop_endbr): Likewise.
6627 * config/i386/i386.opt (mcet): Change help message to built-in
6631 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
6632 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
6633 enable CET built-in functions.
6635 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
6637 * common/config/i386/i386-common.c
6638 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
6639 OPTION_MASK_ISA_MOVDIRI_UNSET,
6640 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
6641 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
6642 * config.gcc (movdirintrin.h): New header.
6643 * config/i386/cpuid.h (bit_MOVDIRI,
6644 bit_MOVDIR64B): New bits.
6645 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
6647 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
6648 (VOID, PVOID, PCVOID)): New function types.
6649 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
6650 __builtin_ia32_directstoreu_u64,
6651 __builtin_ia32_movdir64b): New builtins.
6652 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
6653 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
6655 (ix86_valid_target_attribute_inner_p): Ditto.
6656 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
6657 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
6658 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
6659 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
6660 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
6661 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
6662 (movdiri<mode>, movdir64b_<mode>): New.
6663 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
6664 * config/i386/immintrin.h: Include movdirintrin.h.
6665 * config/i386/movdirintrin.h: New file.
6666 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
6668 2018-04-19 Richard Biener <rguenther@suse.de>
6671 * cfg.c (clear_bb_flags): When loop state says we have
6672 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
6674 2018-04-19 Richard Biener <rguenther@suse.de>
6676 PR tree-optimization/84737
6677 * tree-vect-data-refs.c (vect_copy_ref_info): New function
6678 copying restrict info.
6679 (vect_setup_realignment): Use it.
6680 * tree-vectorizer.h (vect_copy_ref_info): Declare.
6681 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
6682 the first DR to all generated stores.
6683 (vectorizable_load): Likewise for loads.
6685 2018-04-19 Jakub Jelinek <jakub@redhat.com>
6687 PR tree-optimization/85446
6688 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
6689 the integral and pointer types to have the same precision.
6691 * doc/install.texi: Document --disable-cet being the default and
6694 2018-04-18 Martin Liska <mliska@suse.cz>
6696 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
6699 2018-04-18 Martin Liska <mliska@suse.cz>
6702 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
6705 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
6706 arguments if they are comparable.
6708 2018-04-18 Martin Liska <mliska@suse.cz>
6711 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
6714 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
6717 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
6720 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
6721 ENDBR after calling __morestack.
6723 2018-04-18 David Malcolm <dmalcolm@redhat.com>
6726 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
6727 by using gcc_base_ver to generate a gcc_driver_version, and use
6728 it when generating GCC_DRIVER_NAME.
6729 * configure: Regenerate.
6731 2018-04-18 Jakub Jelinek <jakub@redhat.com>
6734 * config.gcc: Obsolete powerpc*-*-*spe*.
6736 2018-04-17 Jakub Jelinek <jakub@redhat.com>
6739 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
6740 (stabstr_D): Change type of unum from unsigned int to
6741 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
6744 2018-04-17 Jim Wilson <jimw@sifive.com>
6747 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
6748 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
6749 Set arg_pointer_offset after using pretend_args_size.
6751 2018-04-17 Jakub Jelinek <jakub@redhat.com>
6753 PR rtl-optimization/85431
6754 * dse.c (record_store): Ignore zero width stores.
6757 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
6758 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
6759 __builtin_stack_restore rather than after it.
6760 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
6761 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
6762 argument instead of virtual_dynamic_stack_rtx.
6764 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
6766 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
6768 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
6769 Add note to error message to explain internal mapping of overloaded
6770 built-in function name to non-overloaded built-in function name.
6771 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
6774 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
6777 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
6778 where the inputs overlap with the output.
6780 2018-04-17 Jakub Jelinek <jakub@redhat.com>
6782 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
6783 (=v, v) alternative and explicit "memory" attribute.
6784 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
6785 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
6787 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
6788 "sselog1" type instead of "sselog".
6789 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
6790 "sselog". Remove explicit "memory" attribute.
6791 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
6792 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
6794 (vec_extract_hi_v32hi): Merge all alternatives into one, use
6795 "sselog1" type instead of "sselog". Remove explicit "memory"
6797 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
6798 use "sselog1" type instead of "sselog". Remove explicit "memory"
6800 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
6801 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
6803 (vec_extract_hi_v64qi): Merge all alternatives into one, use
6804 "sselog1" type instead of "sselog". Remove explicit "memory"
6806 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
6807 use "sselog1" type instead of "sselog". Remove explicit "memory"
6811 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
6814 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
6817 2018-04-17 Martin Jambor <mjambor@suse.cz>
6820 * ipa-cp.c (create_specialized_node): Call
6821 expand_all_artificial_thunks if necessary.
6823 2018-04-17 Martin Liska <mliska@suse.cz>
6826 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
6827 in message, remote space in between '_G' and '('.
6829 2018-04-17 Jakub Jelinek <jakub@redhat.com>
6832 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
6833 avx512f_vmcmp<mode>3<round_saeonly_name>,
6834 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
6835 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
6836 avx512f_rndscale<mode><round_saeonly_name>,
6837 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
6838 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
6839 Use %<iptr>2 instead of %2 for -masm=intel.
6840 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
6841 avx512f_vcvttss2usi<round_saeonly_name>,
6842 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
6844 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
6845 avx512f_vcvttsd2usi<round_saeonly_name>,
6846 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
6847 Use %q1 instead of %1 for -masm=intel.
6848 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
6849 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
6850 of %3 for -masm=intel.
6851 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
6853 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
6855 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
6857 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
6858 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
6860 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
6862 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
6863 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
6864 %g1 and one with %0 and %1.
6865 (avx512er_vmrcp28<mode><round_saeonly_name>,
6866 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
6868 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
6869 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
6870 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
6871 of %0 and %{%4%} for -masm=intel.
6872 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
6873 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
6874 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
6875 order of %0 and %{%5%}%{z%} for -masm=intel.
6877 2018-04-17 Jan Hubicka <jh@suse.cz>
6880 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
6882 2018-04-17 Martin Liska <mliska@suse.cz>
6885 * multiple_target.c (create_dispatcher_calls): Set apostrophes
6886 for target_clone error message. Make default implementation
6887 clone to be a local declaration.
6888 (separate_attrs): Add new argument and check for an empty
6890 (expand_target_clones): Handle it.
6891 (ipa_target_clone): Make redirection just for target_clones
6894 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
6895 Tom de Vries <tom@codesourcery.com>
6898 * omp-expand.c (expand_oacc_for): Add dummy false branch for
6899 tiled basic blocks without omp continue statements.
6901 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
6904 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
6905 vec_extract expression as having side effects to make sure it gets
6908 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
6911 * config/i386/i386.c (get_builtin_code_for_version): Check
6914 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
6917 * gcc/config.gcc: Support "skylake".
6918 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
6920 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
6921 (processor_target_table): Add "skylake".
6922 (ix86_option_override_internal): Add "skylake".
6923 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
6924 PROCESSOR_CANNONLAKE.
6925 (get_builtin_code_for_version): Fix priority for
6926 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
6927 PROCESSOR_SKYLAKE-AVX512.
6928 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
6929 (processor_type): Add PROCESSOR_SKYLAKE.
6931 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
6932 Jason Merrill <jason@redhat.com>
6935 * convert.c (convert_to_integer_1): Use direct recursion for
6936 enumeral types and types with a precision less than the number
6937 of bits in their mode.
6939 2018-04-16 Julia Koval <julia.koval@intel.com>
6942 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
6943 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
6945 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
6948 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
6949 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
6950 and -mno-direct-move.
6952 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
6955 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
6956 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
6957 Ensure negative shifts result in {0}.
6959 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
6961 PR rtl-optimization/79916
6962 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
6963 regs (if any) to define how to gnerate SD moves when LRA is in
6966 2018-04-13 Jakub Jelinek <jakub@redhat.com>
6968 PR rtl-optimization/85393
6969 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
6970 * except.c (expand_dw2_landing_pad_for_region): Make static.
6971 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
6972 a label and unconditional jump to old_bb, rather than
6973 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
6976 PR rtl-optimization/85376
6977 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
6978 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
6979 instead of a specific value.
6981 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
6982 Bin Cheng <bin.cheng@arm.com>
6984 PR tree-optimization/82965
6985 PR tree-optimization/83991
6986 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
6987 by_profile_only parameter.
6988 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
6989 information if the loop was predicted to iterate too many times.
6990 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
6992 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
6995 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
6998 2018-04-13 Martin Liska <mliska@suse.cz>
6999 Jakub Jelinek <jakub@redhat.com>
7002 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
7003 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
7004 * builtins.c (expand_builtin_memory_copy_args): Use
7005 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
7006 handle dest_addr == pc_rtx.
7008 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
7011 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
7012 asked to not generate direct moves.
7013 (fix_trunc<mode>si2_stfiwx): Similar.
7014 (fix_trunc<mode>si2_internal): Similar.
7016 2018-04-12 Jakub Jelinek <jakub@redhat.com>
7019 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
7020 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
7021 lookup if dest in some wider mode is known to be const0_rtx and
7022 if so, record permanent equivalence for it to be ZERO_EXTEND of
7023 the narrower mode destination.
7025 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
7027 * lto-streamer-out.c (output_function): Revert 259346.
7028 * omp-expand.c (expand_oacc_for): Likewise.
7030 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
7032 PR rtl-optimization/85354
7033 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
7034 * sel-sched.c (sel_global_init): ... here.
7036 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
7039 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
7040 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
7041 mode for PE-COFF targets.
7042 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
7043 (i386_pe_asm_lto_end): Likewise.
7044 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
7045 (TARGET_ASM_LTO_END): Likewise.
7046 * config/i386/winnt.c (saved_debug_info_level): New static variable.
7047 (i386_pe_asm_lto_start): New function.
7048 (i386_pe_asm_lto_end): Likewise.
7050 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
7051 Richard Biener <rguenther@suse.de>
7054 * lto-streamer-out.c (output_function): Fix CFG loop state before
7056 * omp-expand.c (expand_oacc_for): Handle calls to internal
7057 functions like regular functions.
7059 2018-04-12 Richard Biener <rguenther@suse.de>
7062 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
7063 for the early LTO debug to properly generate references to it
7064 during DIE emission. Do not re-use that for the skeleton for
7066 (dwarf2out_early_finish): Likewise.
7068 2018-04-12 Jakub Jelinek <jakub@redhat.com>
7071 * config/i386/sse.md
7072 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
7073 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
7074 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
7075 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
7076 and output is a reg, avoid creating invalid lowpart subreg, but
7077 instead split into a 512-bit move. Don't split if not AVX512VL,
7078 input is xmm16+ reg and output is a mem.
7079 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
7080 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
7081 xmm16+ reg and output is a mem.
7083 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7085 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
7086 also for flag_dwarf2_cfi_asm.
7088 2018-04-12 Jakub Jelinek <jakub@redhat.com>
7090 PR rtl-optimization/85342
7091 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
7092 a bool scalar var inside of the loop instead. Don't try to update
7093 recog_data.operand after failed apply_change_group.
7095 2018-04-12 Tom de Vries <tom@codesourcery.com>
7098 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
7099 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
7100 array with flexible array member as array without given dimension.
7101 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
7102 argument for undefined param to true.
7104 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
7107 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
7108 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
7109 from PowerPC section.
7110 * config/rs6000/sysv4.opt (mcall-): Improve help text.
7111 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
7112 help text that is too long.
7113 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
7114 help text that is too long.
7115 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
7116 help text that is too long.
7118 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
7120 * config/alpha/alpha.md (stack_probe_internal): Rename
7121 from "probe_stack". Update all callers.
7123 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
7125 PR rtl-optimization/84566
7126 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
7127 sched_macro_fuse_insns.
7129 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
7132 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
7133 (compute_block_dependences): ... from here.
7135 2018-04-11 Jakub Jelinek <jakub@redhat.com>
7137 PR tree-optimization/85331
7138 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
7139 from int to HOST_WIDE_INT.
7141 2018-04-11 Martin Jambor <mjambor@suse.cz>
7144 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
7145 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
7146 not the same as the source val.
7147 (cgraph_edge_brings_value_p): New parameter.
7148 (gather_edges_for_value): Pass destination value to
7149 cgraph_edge_brings_value_p.
7150 (perhaps_add_new_callers): Likewise.
7151 (get_info_about_necessary_edges): Likewise and exclude values brought
7152 only by self-recursive edges.
7153 (create_specialized_node): Redirect only clones of self-calling edges.
7154 (+self_recursive_pass_through_p): New function.
7155 (find_more_scalar_values_for_callers_subset): Use it.
7156 (find_aggregate_values_for_callers_subset): Likewise.
7157 (known_aggs_to_agg_replacement_list): Removed.
7158 (decide_whether_version_node): Re-calculate known constants for all
7159 remaining context clones.
7161 2018-04-11 Richard Biener <rguenther@suse.de>
7164 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
7165 from early DWARF output.
7166 (dwarf2out_early_finish): Output line info unconditionally into
7167 early DWARF and add reference to it.
7169 2018-04-11 Jakub Jelinek <jakub@redhat.com>
7172 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
7173 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
7174 other than V2DFmode using iptr mode attribute.
7175 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
7177 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
7179 PR rtl-optimization/84659
7180 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
7182 2018-04-11 Jakub Jelinek <jakub@redhat.com>
7185 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
7187 (output_loc_list): Pass address of a dummy size variable even in the
7188 locview handling loop.
7189 (index_location_lists): Add comment on why skip_loc_list_entry can't
7192 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
7195 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
7198 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
7201 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
7202 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
7203 and -mstring-compare-inline-limit.
7205 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
7208 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
7209 for stack clash protection in a register whenever we need it to be in
7212 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
7214 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
7215 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
7217 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
7220 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
7222 (mlong-double-): Ditto.
7223 * config/rs6000/sysv4.opt (msdata=): Ditto.
7224 (mtls-size=): Ditto.
7226 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
7228 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
7229 erroneous entries for
7230 "vector int vec_ldl (int, long int *)", and
7231 "vector unsigned int vec_ldl (int, unsigned long int *)".
7232 Add comments and entries for
7233 "vector bool char vec_ldl (int, bool char *)",
7234 "vector bool short vec_ldl (int, bool short *)",
7235 "vector bool int vec_ldl (int, bool int *)",
7236 "vector bool long long vec_ldl (int, bool long long *)",
7237 "vector pixel vec_ldl (int, pixel *)",
7238 "vector long long vec_ldl (int, long long *)",
7239 "vector unsigned long long vec_ldl (int, unsigned long long *)".
7240 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
7241 type tree bool_long_long_type_node and correct definition of
7242 bool_V2DI_type_node to make reference to this new type tree.
7243 (rs6000_mangle_type): Replace erroneous reference to
7244 bool_long_type_node with bool_long_long_type_node.
7245 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
7246 comments to emphasize sign distinctions for char and int types and
7247 replace RS6000_BTI_bool_long constant with
7248 RS6000_BTI_bool_long_long constant. Also add comment to restrict
7249 use of RS6000_BTI_pixel.
7250 (bool_long_type_node): Remove this macro definition.
7251 (bool_long_long_type_node): New macro definition
7253 2018-04-10 Jakub Jelinek <jakub@redhat.com>
7255 PR rtl-optimization/85300
7256 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
7257 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
7258 simplify_unary_operation fails.
7260 2018-04-10 Martin Liska <mliska@suse.cz>
7262 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
7263 cgraph_edge and ipa_ref.
7265 2018-04-10 Jakub Jelinek <jakub@redhat.com>
7269 * config/i386/sse.md
7270 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
7271 computation of the VEC_MERGE selector from mask.
7272 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
7273 Fix decoding of the VEC_MERGE selector into mask.
7275 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
7277 PR tree-optimization/85286
7278 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
7280 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
7282 * final.c (final_1): Set insn_last_address as well as
7283 insn_current_address.
7285 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7288 * explow.c (emit_stack_probe): Call validize_mem on memory location
7289 before passing it to gen_probe_stack. Create address operand and
7290 legitimize it for the probe_stack_address case.
7292 2018-04-09 Jan Hubicka <jh@suse.cz>
7295 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
7296 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
7297 * tree.c (free_lang_data_in_type): Fix handling of binfos;
7299 (free_lang_data): Rebuild type inheritance graph.
7301 2018-04-09 Martin Sebor <msebor@redhat.com>
7303 * invoke.texi (-finline-small-functions): Mention other optimization
7305 (-findirect-inlining, -fpartial-inlining): Same.
7306 (-finline-functions-called-once): Same.
7307 (-freorder-blocks-and-partition): Same.
7309 2018-04-09 Jan Hubicka <jh@suse.cz>
7312 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
7313 jumps; choose last target that matches the criteria (i.e.
7314 no partition changes for non-crossing jumps).
7315 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
7316 support for redirecting crossing jumps to non-crossing.
7318 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
7320 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
7321 also for naked functions.
7323 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
7325 * config/arc/arc.md (add_shift): New pattern.
7326 (add_shift2): Likewise.
7327 (sub_shift): Likewise.
7328 (sub_shift_cmp0_noout): Likewise.
7329 (compare_si_ashiftsi): Likewise.
7330 (xbfu_cmp0_noout): New combine pattern.
7331 (xbfu_cmp0"): Likewise.
7332 (movsi_set_cc_insn): Place the predicable variant first.
7333 (commutative_binary_cmp0_noout): Remove clobber.
7334 (commutative_binary_cmp0): New pattern.
7335 (noncommutative_binary_cmp0): Likewise.
7336 (noncommutative_binary_cmp0_noout): Likewise.
7337 (noncommutative_binary_comparison_result_used): Removed.
7338 (rsub_cmp0): New pattern.
7339 (rsub_cmp0_noout): Likewise.
7340 (extzvsi): Changed, keep only meaningful variants.
7341 (SQH, SEZ): New iterators.
7342 (SQH_postfix): New mode attribute.
7343 (SEZ_prefix): New code attribute.
7344 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
7345 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
7346 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
7348 (noncommutative_operator): Check the availability of barrel
7351 2018-04-09 Richard Biener <rguenther@suse.de>
7353 PR tree-optimization/85284
7354 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
7355 Only use the niter constraining form of simple_iv when the exit
7358 2018-04-09 Tom de Vries <tom@codesourcery.com>
7361 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
7362 (define_expand "*memory_barrier"): New define_expand.
7363 (define_insn "memory_barrier"): New insn.
7365 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
7367 PR rtl-optimization/80463
7368 PR rtl-optimization/83972
7369 PR rtl-optimization/83480
7371 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
7372 correct producer for the insn.
7373 (tidy_control_flow): Fixup seqnos in case of debug insns.
7375 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
7377 PR rtl-optimization/83913
7379 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
7380 different sched-times when merging exprs.
7382 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
7384 PR rtl-optimization/83962
7386 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
7387 tidy_fallthru_edge and tidy_control_flow.
7389 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
7391 PR rtl-optimization/83530
7393 * sel-sched.c (force_next_insn): New global variable.
7394 (remove_insn_for_debug): When force_next_insn is true, also leave only
7395 next insn in the ready list.
7396 (sel_sched_region): When the region wasn't scheduled, make another pass
7397 over it with force_next_insn set to 1.
7399 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
7401 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
7403 * config/nds32/constants.md (unspec_volatile_element): Add enum values
7404 for interrupt control.
7405 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
7406 functions for interrupt control.
7407 * config/nds32/nds32-intrinsic.md: Likewise.
7408 * config/nds32/nds32_intrinsic.h: Likewise.
7409 * config/nds32/nds32.h (nds32_builtins): Likewise.
7411 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
7413 * config/nds32/nds32.c (nds32_init_machine_status,
7414 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
7415 strict_aligned_p field.
7416 (nds32_expand_to_rtl_hook): New function.
7417 (TARGET_EXPAND_TO_RTL_HOOK): Define.
7418 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
7420 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
7421 Chung-Ju Wu <jasonwucj@gmail.com>
7423 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
7424 * config/nds32/nds32-n7.md: New file.
7425 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
7426 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
7428 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
7429 * config/nds32/nds32.md (pipeline_model): Add n7.
7430 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
7431 * config/nds32/pipelines.md: Include n7 settings.
7433 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
7434 Chung-Ju Wu <jasonwucj@gmail.com>
7436 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
7437 * config/nds32/nds32-e8.md: New file.
7438 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
7439 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
7441 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
7442 * config/nds32/nds32.md (pipeline_model): Add e8.
7443 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
7444 * config/nds32/pipelines.md: Include e8 settings.
7446 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
7447 Chung-Ju Wu <jasonwucj@gmail.com>
7449 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
7450 * config/nds32/nds32-n8.md: New file.
7451 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
7452 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
7454 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
7455 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
7456 * config/nds32/nds32.md (pipeline_model): Add n8.
7457 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
7458 * config/nds32/pipelines.md: Include n8 settings.
7460 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
7461 Chung-Ju Wu <jasonwucj@gmail.com>
7463 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
7464 * config/nds32/nds32-n9-2r1w.md: New file.
7465 * config/nds32/nds32-n9-3r2w.md: New file.
7466 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
7467 nds32_register_ports): New or modify for cpu n9.
7468 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
7470 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
7471 * config/nds32/nds32-utils.c: New file.
7472 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
7473 TARGET_MUL_SLOW): Define.
7474 * config/nds32/nds32.md (pipeline_model): New attribute.
7475 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
7476 New options that support cpu n9.
7477 * config/nds32/pipelines.md: Include n9 settings.
7478 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
7480 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
7482 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
7483 information if necessary.
7484 (output_cond_branch_compare_zero): Likewise.
7485 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
7486 (nds32_target_alignment): Refine for alignment.
7487 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
7488 (FUNCTION_BOUNDARY): Modify.
7489 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
7491 * config/nds32/nds32.opt (malways-align, malign-functions): New.
7493 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
7495 * config/nds32/constants.md (unspec_volatile_element): Add values for
7496 TLB operation and data prefetch.
7497 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
7498 functions for TLB operation and data prefetch.
7499 * config/nds32/nds32-intrinsic.md: Likewise.
7500 * config/nds32/nds32_intrinsic.h: Likewise.
7501 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
7502 (nds32_print_operand): Likewise.
7503 * config/nds32/nds32.h (nds32_builtins): Likewise.
7505 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
7506 Andrew Pinski <pinsika@gcc.gnu.org>
7509 * match.pd: Use constant_boolean_node of correct type instead of
7510 boolean_true_node or boolean_false_node for simplifying
7511 pointer comparisons to zero.
7513 2018-04-07 Jakub Jelinek <jakub@redhat.com>
7515 PR tree-optimization/80021
7516 * tree.c (verify_type_variant): Make error call in verify_variant_match
7517 translatable and remove final full stop.
7519 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
7521 * config/nds32/constants.md (unspec_volatile_element): Add
7522 UNSPEC_VOLATILE_EH_RETURN.
7523 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
7524 nds32_output_stack_pop): Support dwarf exception handling process.
7525 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
7526 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
7527 exception handling process.
7528 (nds32_compute_stack_frame): Likewise.
7529 (nds32_return_addr_rtx): Likewise.
7530 (nds32_initial_elimination_offset): Likewise.
7531 (nds32_expand_prologue): Likewise.
7532 (nds32_expand_epilogue): Likewise.
7533 (nds32_dynamic_chain_address): New function.
7534 * config/nds32/nds32.h (machine_function): Add fields for dwarf
7536 (DYNAMIC_CHAIN_ADDRESS): Define.
7537 (EH_RETURN_DATA_REGNO): Define.
7538 (EH_RETURN_STACKADJ_RTX): Define.
7539 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
7540 patterns for dwarf exception handling.
7542 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
7544 * config/nds32/nds32.h: Clean up obsolete macros.
7546 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
7548 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
7549 Add enum values for particular instructions.
7550 * config/nds32/nds32-intrinsic.c: Implementation of expanding
7551 particular intrinsic functions.
7552 * config/nds32/nds32-intrinsic.md: Likewise.
7553 * config/nds32/nds32_intrinsic.h: Likewise.
7554 * config/nds32/nds32.h (nds32_builtins): Likewise.
7555 * config/nds32/nds32.md (type): Add pbsad and pbsada.
7556 (btst, ave): New patterns for particular instructions.
7558 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
7560 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
7561 Add enum values for atomic load/store and memory sync.
7562 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
7564 * config/nds32/nds32-intrinsic.md: Likewise.
7565 * config/nds32/nds32_intrinsic.h: Likewise.
7566 * config/nds32/nds32.h (nds32_builtins): Likewise.
7568 2018-04-07 Jakub Jelinek <jakub@redhat.com>
7570 PR tree-optimization/85257
7571 * fold-const.c (native_encode_vector): If not all elts could fit
7572 and off is -1, return 0 rather than offset.
7573 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
7574 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
7575 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
7576 adjust buffer in native_interpret_expr call.
7578 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
7580 * config/nds32/constants.md (unspec_volatile_element): Add cache
7581 control enum values.
7582 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
7583 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
7584 * config/nds32/nds32.c (nds32_cctl_names): New.
7585 (nds32_print_operand): Handle cache control register names.
7586 * config/nds32/nds32.h (nds32_builtins): New enum values.
7587 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
7589 * config/nds32/nds32.md (type): Add mmu.
7590 * config/nds32/pipelines.md (simple_insn): Add mmu.
7592 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
7594 * config/nds32/nds32.md (type): Remove call.
7595 * config/nds32/pipelines.md (simple_insn): Likewise.
7597 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
7599 * config/nds32/constants.md (unspec_volatile_element): Add
7600 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
7601 UNSPEC_VOLATILE_FMFCFG.
7602 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
7603 description for fmfcfg and fmfcsr.
7604 (bdesc_1arg): Add fmtcsr.
7605 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
7606 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
7607 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
7608 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
7609 unspec_fmfcfg): New patterns.
7610 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
7611 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
7612 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
7613 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
7614 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
7615 __nds32__fmfcfg): Define.
7617 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
7619 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
7620 intrinsic register names.
7621 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
7622 intrinsic register enum values and macros.
7624 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
7626 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
7627 for load/store addressing form.
7628 (nds32_print_operand_address): Likewise.
7630 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
7633 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
7634 based on LABEL_REF. Remove useless assertion.
7635 (pic_address_needs_scratch): Fix formatting.
7636 (sparc_legitimize_pic_address): Minor tweaks.
7637 (sparc_delegitimize_address): Adjust assertion accordingly.
7638 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
7639 into symbolic_operand.
7640 (movsi_high_pic_label_ref): Likewise.
7641 (movsi_lo_sum_pic_label_ref): Likewise.
7642 (movdi_pic_label_ref): Likewise.
7643 (movdi_high_pic_label_ref): Likewise.
7644 (movdi_lo_sum_pic_label_ref): Likewise.
7646 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
7648 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
7649 custom LIB_SPEC setup.
7651 2018-04-06 Ruslan Bukin <br@bsdpad.com>
7652 Kito Cheng <kito.cheng@gmail.com>
7654 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
7655 * config/riscv/freebsd.h: New.
7657 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
7659 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
7660 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
7663 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
7664 Kito Cheng <kito.cheng@gmail.com>
7666 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
7667 nds32_output_call, nds32_symbol_binds_local_p): New functions.
7668 * config/nds32/nds32-protos.h (nds32_output_call,
7669 nds32_output_return): Declare.
7670 * config/nds32/nds32.md: Refine all the call and return patterns.
7672 2018-04-06 Jakub Jelinek <jakub@redhat.com>
7675 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
7676 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
7678 PR rtl-optimization/84872
7679 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
7680 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
7683 2018-04-06 Tamar Christina <tamar.christina@arm.com>
7685 * expr.c (copy_blkmode_to_reg): Revert 254862.
7686 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
7688 2018-04-06 Richard Biener <rguenther@suse.de>
7691 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
7692 after seeing a component reference with an adjacent field. Treat
7693 refs to arrays at struct end of external decls similar to
7694 refs to unconstrained commons.
7696 2018-04-06 Jakub Jelinek <jakub@redhat.com>
7699 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
7700 look through SAVE_EXPRs with non-side-effects argument. Adjust
7702 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
7705 2018-04-06 Richard Biener <rguenther@suse.de>
7708 * alias.c (find_base_term): New wrapper around find_base_term
7709 unwinding CSELIB_VAL_PTR changes.
7710 (find_base_term): Do not restore CSELIB_VAL_PTR during the
7713 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7715 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
7717 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
7718 constant definitions.
7719 ("nop"): lr 0,0 -> nopr r0
7720 ("nop_lr0", "nop_lr1"): New insn definitions.
7722 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
7724 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
7725 NDS32_V3PUSH_AVAILABLE_P macro.
7727 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
7728 Chung-Ju Wu <jasonwucj@gmail.com>
7730 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
7731 (nds32*-*-*): Add float and fpu_config into supported_defaults.
7732 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
7733 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
7734 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
7735 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
7736 * config/nds32/constraints.md: New constraints and checking for hard
7737 float configuration.
7738 * config/nds32/iterators.md: New mode iterator and attribute for hard
7739 float configuration.
7740 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
7742 * config/nds32/nds32-fpu.md: New file.
7743 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
7744 deal with hard float code generation.
7745 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
7747 (abi_type, float_reg_number): New enum type.
7748 * config/nds32/nds32-predicates.c: New predicates for hard float.
7749 * config/nds32/nds32-protos.h: Declare functions for hard float.
7750 * config/nds32/nds32.c: Implementation for hard float configuration.
7751 * config/nds32/nds32.h: Definitions for hard float configuration.
7752 * config/nds32/nds32.md: Include hard float machine description and
7753 modify patterns for hard float configuration.
7754 * config/nds32/nds32.opt: New options for hard float configuration.
7755 * config/nds32/predicates.md: New predicates for hard float
7758 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
7760 * common/config/nds32/nds32-common.c
7761 (nds32_option_optimization_table): Enable -mreleax-hint by default.
7763 2018-04-05 Jakub Jelinek <jakub@redhat.com>
7766 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
7767 CONSTRUCTOR_ELT (ctor, ...)->value.
7769 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
7772 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
7774 2018-04-05 Tom de Vries <tom@codesourcery.com>
7777 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
7780 2018-04-05 Shiva Chen <shiva0217@gmail.com>
7781 Kito Cheng <kito.cheng@gmail.com>
7783 * config/nds32/constraints.md (U33): Fine-tune checking condition.
7784 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
7785 * config/nds32/nds32.h (nds32_16bit_address_type): Add
7786 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
7788 2018-04-05 Shiva Chen <shiva0217@gmail.com>
7789 Kito Cheng <kito.cheng@gmail.com>
7791 * config/nds32/constraints.md (Ufe): New memory constraint.
7792 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
7793 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
7794 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
7796 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
7797 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
7799 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7801 * config/nds32/nds32.md: Use optimize_size in the condition for
7802 alu-shift instructions.
7804 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7806 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
7808 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7810 * config/nds32/nds32.md (negsi2): Refine pattern.
7812 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
7813 Chung-Ju Wu <jasonwucj@gmail.com>
7815 * config/nds32/iterators.md (shift_rotate): New code iterator.
7816 (shift): New code attribute.
7817 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
7818 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
7819 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
7820 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
7821 bit-wise operations.
7822 (andsi3, *andsi3): Ditto.
7823 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
7824 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
7825 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
7826 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
7827 nds32_ior_operand, nds32_xor_operand): New predicates.
7829 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7831 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
7832 (addsi3, subsi3): ... this.
7834 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7836 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
7838 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7840 * config/nds32/nds32.md: Adjust indention.
7842 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
7844 * config/nds32/nds32.md (feature): New attribute.
7846 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7848 * config/nds32/nds32.md (subtype): New attribute.
7850 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
7853 * config/arm/arm-builtins.c (arm_expand_builtin): Change
7854 expansion to perform a bitwise AND of the argument followed by a
7855 boolean negation of the result.
7857 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
7859 PR rtl-optimization/84878
7860 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
7861 the basic block. Assert the use reference is not artificial and that
7862 it has an associated insn.
7864 2018-04-04 Michael Matz <matz@suse.de>
7866 * builtins.c (compute_objsize): Pass correct operand
7867 to array_at_struct_end_p.
7869 2018-04-04 Richard Biener <rguenther@suse.de>
7872 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
7873 from contexts for DINFO_LEVEL_TERSE and below.
7875 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
7877 * config/nds32/nds32-doubleword.md (move_<mode>): Require
7878 resiter_operand condition.
7879 * config/nds32/nds32.md (*move<mode>): Ditto.
7881 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
7882 Monk Chiang <sh.chiang04@gmail.com>
7884 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
7886 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
7888 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
7890 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
7891 Kito Cheng <kito.cheng@gmail.com>
7893 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
7894 nds32_cond_code_str, output_cond_branch,
7895 output_cond_branch_compare_zero, nds32_expand_cbranch,
7896 nds32_expand_cstore, nds32_expand_movcc,
7897 nds32_output_cbranchsi4_equality_zero,
7898 nds32_output_cbranchsi4_equality_reg,
7899 nds32_output_cbranchsi4_equality_reg_or_const_int,
7900 nds32_output_cbranchsi4_greater_less_zero: New functions.
7901 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
7902 nds32_expand_cstore, nds32_expand_movcc,
7903 nds32_output_cbranchsi4_equality_zero,
7904 nds32_output_cbranchsi4_equality_reg,
7905 nds32_output_cbranchsi4_equality_reg_or_const_int,
7906 nds32_output_cbranchsi4_greater_less_zero): Declare.
7907 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
7908 nds32_rimm11s_operand): New predicates.
7909 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
7910 * config/nds32/nds32.md: Rewrite all the branch and conditional move
7913 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
7915 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
7916 * config/nds32/nds32.md: Ditto.
7917 * config/nds32/pipelines.md: Ditto.
7919 2018-04-04 Richard Biener <rguenther@suse.de>
7921 PR tree-optimization/85168
7922 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
7923 propagating abnormals.
7925 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
7927 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
7929 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
7930 Kito Cheng <kito.cheng@gmail.com>
7932 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
7933 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
7934 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
7935 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
7936 * config/nds32/nds32.md (sibcall_internal): New.
7937 (sibcall_register): Remove.
7938 (sibcall_immediate): Remove.
7939 (sibcall_value_internal): New.
7940 (sibcall_value_register): Remove.
7941 (sibcall_value_immediate): Remove.
7942 * config/nds32/predicates.md (nds32_general_register_operand): New.
7943 (nds32_call_address_operand): New.
7945 2018-04-03 Jakub Jelinek <jakub@redhat.com>
7947 PR rtl-optimization/85167
7948 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
7949 bb_defs if *split_p, instead preinitialize it to NULL.
7951 PR tree-optimization/85156
7952 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
7953 evaluating the argument multiple times.
7955 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
7957 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
7959 (_mm_cvtpd_ps): Likewise.
7960 (_mm_cvttpd_epi32): Likewise.
7961 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
7962 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
7963 vector, pixel, and bool following altivec.h include.
7965 2018-04-03 Martin Sebor <msebor@redhat.com>
7967 * doc/extend.texi (Common Function Attributes): Clarify.
7968 (const attribute): Likewise.
7969 (pure attribute): Likewise.
7971 2018-04-03 Jakub Jelinek <jakub@redhat.com>
7974 * config/i386/i386.c (ix86_expand_vector_set): Use
7975 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
7977 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
7979 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
7980 instructions when changing rounding bits to preserve precision bits
7981 in the x87 control word.
7983 2018-04-03 Martin Liska <mliska@suse.cz>
7985 PR tree-optimization/82491
7986 * rtl.h (strip_offset_and_add): Replace += suboffset with
7987 poly_uint64 () + suboffset.
7989 2018-03-29 Martin Liska <mliska@suse.cz>
7990 Martin Jambor <mjambor@suse.cz>
7993 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
7994 param_type is not an integral or pointer type.
7996 2018-04-03 Richard Biener <rguenther@suse.de>
7998 * sese.h (recompute_all_dominators): Remove.
8000 2018-04-02 Martin Sebor <msebor@redhat.com>
8002 * doc/invoke.texi (-Wrestrict): Fix typos.
8004 2018-04-02 Jim Wilson <jimw@sifive.com>
8006 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
8007 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
8008 (<optab>di3, <optab>si3_extend): Likewise.
8009 (<optab>si3_mask, <optab>si3_mask_1): New.
8010 (<optab>di3_mask, <optab>di3_mask_1): New.
8011 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
8012 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
8013 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
8015 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
8017 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
8020 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
8022 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
8023 (nds32_canonicalize_comparison): New function.
8025 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
8026 Kito Cheng <kito.cheng@gmail.com>
8027 Kuan-Lin Chen <kuanlinchentw@gmail.com>
8029 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
8030 * config/nds32/constants.md (unspec_volatile_element): Add
8031 UNSPEC_VOLATILE_RELAX_GROUP.
8032 * config/nds32/nds32-relax-opt.c: New file.
8033 * config/nds32/nds32-predicates.c
8034 (nds32_symbol_load_store_p): New function.
8035 * config/nds32/nds32-protos.h
8036 (nds32_symbol_load_store_p): Declare function.
8037 (make_pass_nds32_relax_opt): Declare new rtl pass function.
8038 * config/nds32/nds32.c
8039 (nds32_register_pass): New function to register pass.
8040 (nds32_register_passes): New function to register passes.
8041 * config/nds32/nds32.md (relax_group): New pattern.
8042 * config/nds32/nds32.opt (mrelax-hint): New option.
8043 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
8045 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
8047 * config/nds32/t-nds32: Modify files dependency.
8049 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
8051 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
8052 (PROFILE_HOOK): Define its implementation.
8054 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
8056 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
8057 type and 32-bit size.
8059 2018-04-01 Jakub Jelinek <jakub@redhat.com>
8062 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
8063 (V_128_256): New mode iterator.
8064 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
8065 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
8066 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
8068 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
8069 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
8071 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
8074 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
8075 NaN inputs correctly.
8077 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
8080 * config/rs6000/vsx.md (??r): New mode attribute.
8081 (*vsx_mov<mode>_64bit): Use it.
8082 (*vsx_mov<mode>_32bit): Likewise.
8084 2018-03-30 Martin Sebor <msebor@redhat.com>
8086 PR tree-optimization/84818
8087 * builtins.c (check_access): Use warning_n.
8089 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8092 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
8094 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
8097 2018-03-30 Julia Koval <julia.koval@intel.com>
8100 * x86-tune.def (movx, partial_reg_dependency): Enable for
8103 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
8106 * lra-constraints.c (process_alt_operands): Move setting
8107 this_alternative_matches below.
8109 2018-03-29 Martin Liska <mliska@suse.cz>
8112 * doc/invoke.texi: Document how LTO works with debug info.
8113 Describe auto-load support of binutils. Mention 'x86-64'
8114 as valid option value of -march option.
8116 2018-03-29 Jakub Jelinek <jakub@redhat.com>
8118 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
8121 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
8122 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
8123 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
8126 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
8129 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
8130 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
8131 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
8132 for RS6000_BTM_POWERPC64.
8133 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
8134 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
8135 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
8140 2018-03-28 Carl Love <cel@us.ibm.com>
8143 2017-09-27 Carl Love <cel@us.ibm.com>
8145 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
8146 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
8147 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
8150 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8152 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
8153 instead of __vector bool.
8154 (_mm_max_pu8): Likewise.
8155 (_mm_min_pi16): Likewise.
8157 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
8160 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
8161 (DIVWEUO): Likewise.
8163 (DIVDEUO): Likewise.
8164 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
8165 DIVWEUO and DIVDEUO.
8166 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
8167 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
8168 (div_extend): Likewise.
8169 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
8171 (__builtin_divweuo): Likewise.
8172 (__builtin_divdeo): Likewise.
8173 (__builtin_divdeuo): Likewise.
8175 2018-03-28 Jakub Jelinek <jakub@redhat.com>
8178 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
8179 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
8181 PR tree-optimization/82004
8182 * gimple-match-head.c (optimize_pow_to_exp): New function.
8183 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
8184 Don't fold to exp if optimize_pow_to_exp is false.
8186 2018-03-28 Martin Liska <mliska@suse.cz>
8189 * calls.c (initialize_argument_information): Fix trailing space.
8190 * common.opt: Fix typo and provide better explanation for
8191 -fsanitize-coverage option.
8192 * config/i386/i386.opt: Fix typo.
8194 2018-03-28 Jakub Jelinek <jakub@redhat.com>
8195 Martin Liska <mliska@suse.cz>
8198 * gimplify.c (asan_poison_variable): Don't do the check for
8199 gimplify_omp_ctxp here.
8200 (gimplify_decl_expr): Do it here.
8201 (gimplify_target_expr): Likewise.
8203 2018-03-28 Martin Liska <mliska@suse.cz>
8206 * config/i386/i386.c (ix86_function_arg_advance): Do not call
8207 chkp_type_bounds_count if MPX is not enabled.
8209 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
8211 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
8213 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
8216 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
8217 function to create the function decl for complex long double
8218 multiply and divide for -mabi=ieeelongdouble.
8219 (init_float128_ieee): Call it.
8221 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
8224 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
8225 -fcf-protection=branch -mibt.
8226 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
8228 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
8231 * config/arm/arm.c (arm_valid_symbolic_address): Handle
8232 arm_word_relocations.
8234 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
8237 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
8238 extern array declarations.
8240 2018-03-27 Richard Biener <rguenther@suse.de>
8243 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
8244 explicit single_use checks.
8246 2018-03-27 Richard Biener <rguenther@suse.de>
8248 PR tree-optimization/85082
8249 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
8252 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
8254 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
8255 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
8256 Turn on fasynchronous-unwind-tables and funwind-tables.
8258 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
8261 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
8262 (*bmi_blsr_<mode>_ccz): Ditto.
8264 2018-03-26 Tom de Vries <tom@codesourcery.com>
8266 PR tree-optimization/85063
8267 * omp-general.c (offloading_function_p): New function. Factor out
8269 * omp-offload.c (pass_omp_target_link::gate): ... here.
8270 * omp-general.h (offloading_function_p): Declare.
8271 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
8272 with attribute omp declare target for offloading functions.
8274 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
8276 PR tree-optimization/84005
8277 * tree-data-ref.h (get_base_for_alignment): Declare.
8278 * tree-data-ref.c (get_base_for_alignment_1): New function.
8279 (get_base_for_alignment): Likewise.
8280 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
8281 get_base_for_alignment to find a suitable base object, instead
8282 of always using drb->base_address.
8284 2018-03-23 Jakub Jelinek <jakub@redhat.com>
8287 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
8288 known size by default.
8290 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
8293 * lra-constraints.c (process_alt_operands): Don't match BLKmode
8294 and non BLKmode operands.
8296 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8299 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
8300 Clean up attributes.
8302 2018-03-23 Richard Biener <rguenther@suse.de>
8305 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
8306 we are going to emit early debug for LTO.
8308 2018-03-23 Jakub Jelinek <jakub@redhat.com>
8311 * function.c (match_asm_constraints_1): Don't optimize if input
8312 doesn't satisfy general_operand predicate for output's mode.
8315 * alias.c (write_dependence_p): Don't require for x_canonicalized
8316 non-VOIDmode if x has VOIDmode.
8319 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
8320 just don't try to optimize it rather than assert it never happens.
8322 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
8324 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
8325 macro expansions for definition of ST_INTERNAL_<mode> and
8326 LD_INTERNAL_<mode> builtins.
8327 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
8329 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
8331 (altivec_expand_st_builtin): Likewise.
8332 (altivec_expand_builtin): Remove calls to deleted functions.
8333 (rs6000_address_for_altivec): Delete this function.
8334 * config/rs6000/vector.md: Remove expands for
8335 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
8337 2018-03-22 Sudakshina Das <sudi.das@arm.com>
8340 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
8341 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
8342 re-computing once computed.
8343 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
8344 (arm_init_machine_status): Initialize
8345 machine->static_chain_stack_bytes.
8347 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
8350 * doc/extend.texi: Add four new prototypes for vec_ld.
8351 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
8352 definitions for more logical presentation.
8353 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
8354 entries for V1TI variants of __builtin_altivec_ld builtin.
8355 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
8356 handling of V1TI variant of LVX icode pattern.
8357 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
8358 (rs6000_gimple_fold_builtin): Likewise.
8359 (altivec_init_builtins): Add code to define
8360 __builtin_altivec_lvx_v1ti function.
8362 2018-03-22 Jakub Jelinek <jakub@redhat.com>
8365 * function.c (match_asm_constraints_1): Don't do the optimization
8366 if input isn't a REG, SUBREG, MEM or constant.
8368 2018-03-22 Tom de Vries <tom@codesourcery.com>
8370 PR tree-optimization/84956
8371 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
8372 bb_has_abnormal_pred.
8374 2018-03-22 Jakub Jelinek <jakub@redhat.com>
8377 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
8378 DECL_INITIAL (decl) to decl at the end.
8379 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
8382 2018-03-21 Joseph Myers <joseph@codesourcery.com>
8384 * doc/extend.texi (__builtin_tgmath): Document when complex
8385 integer types are treated as _Complex _Float64.
8387 2018-03-21 Tom de Vries <tom@codesourcery.com>
8389 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
8391 2018-03-21 Jakub Jelinek <jakub@redhat.com>
8393 PR tree-optimization/84960
8394 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
8395 if it is ENTRY block, move them into single succ of ENTRY in that case.
8397 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
8399 PR tree-optimization/84811
8400 * poly-int.h (poly_span_traits): Remove the T3 parameter and
8401 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
8402 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
8403 (known_subrange_p): Update accordingly. Cast each value involved
8404 in the size comparison, rather than casting the result of the
8407 2018-03-21 Jakub Jelinek <jakub@redhat.com>
8409 PR tree-optimization/84982
8410 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
8411 by flipping the least significant bit rather than all bits from
8412 bitpos to bitpos + bitsize - 1.
8414 2018-03-21 Nathan Sidwell <nathan@acm.org>
8416 * doc/extend.texi (Deprecated Features): Remove mention of
8417 long-deleted deprecations.
8419 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8422 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
8423 * configure: Regenerate.
8425 2018-03-21 Tom de Vries <tom@codesourcery.com>
8427 PR tree-optimization/83126
8428 * tree-parloops.c (num_phis): New function.
8429 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
8431 2018-03-21 Nathan Sidwell <nathan@acm.org>
8433 * doc/extend.texi (Deprecated Features): Update deprecated flags,
8434 mention anon-struct/union members and trailing attributes.
8436 2018-03-21 Bin Cheng <bin.cheng@arm.com>
8438 PR tree-optimization/84969
8439 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
8440 builtin memset partitions if they set different rhs values.
8442 2018-03-21 Jakub Jelinek <jakub@redhat.com>
8444 PR rtl-optimization/84989
8445 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
8446 VEC_DUPLICATE with scalar result mode.
8448 2018-03-21 Martin Liska <mliska@suse.cz>
8451 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
8452 not intended return statement.
8454 2018-03-21 Martin Liska <mliska@suse.cz>
8457 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
8458 (chkp_find_bound_slots_1): Limit number of iterations.
8460 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
8463 * Minor grammar fixes for x86 options.
8465 2018-03-20 Jakub Jelinek <jakub@redhat.com>
8468 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
8469 holding REG_CFA_RESTORE notes, instead turn them into a USE.
8471 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
8474 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
8475 (altivec_lvx_<mode>_1op): Likewise.
8476 (altivec_stvx_<mode>_2op): Likewise.
8477 (altivec_stvx_<mode>_1op): Likewise.
8478 (altivec_lvx_<VM2:mode>): New define_expand.
8479 (altivec_stvx_<VM2:mode>): Likewise.
8480 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
8481 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
8482 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
8483 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
8484 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
8485 (rs6000_gen_lvx): Likewise.
8486 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
8487 (altivec_expand_stv_builtin): Likewise.
8488 (altivec_expand_builtin): Likewise.
8489 * config/rs6000/vector.md: Likewise.
8491 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8494 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
8497 2018-03-20 Richard Biener <rguenther@suse.de>
8500 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
8501 sign-conversions as zero, fall back to standard scalar_stmt
8504 2018-03-20 Martin Liska <mliska@suse.cz>
8507 * predict.c (rebuild_frequencies): Handle case when we have
8508 PROFILE_ABSENT, but flag_guess_branch_prob is false.
8510 2018-03-20 Jakub Jelinek <jakub@redhat.com>
8513 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
8514 flag_section_anchors.
8515 * varasm.c (use_blocks_for_decl_p): Remove hack for
8516 dw2_force_const_mem.
8519 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
8521 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
8522 be created, use lowpart_subreg of operands[0] rather than operands[0]
8524 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
8525 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
8526 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
8527 and n constraint instead of aarch64_shift_imm_di and Usd.
8528 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
8529 (*aarch64_<optab>_reg_minus<mode>3): ... this.
8531 2018-03-20 Sudakshina Das <sudi.das@arm.com>
8534 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
8535 to favor GPR over NEON registers.
8536 (<shift>di3_neon): Likewise.
8538 2018-03-20 Tom de Vries <tom@codesourcery.com>
8541 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
8542 (nvptx_process_pars): Emit bar.sync asap and alap.
8544 2018-03-20 Tom de Vries <tom@codesourcery.com>
8547 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
8548 seen_label if seen_label is already set.
8550 2018-03-20 Jakub Jelinek <jakub@redhat.com>
8553 * config/i386/i386.c (fold_builtin_cpu): For features above 31
8554 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
8555 Use 1U instead of 1. Formatting fixes.
8558 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
8559 instead of TREE_TYPE (s1) for the return value.
8561 2018-03-19 Jakub Jelinek <jakub@redhat.com>
8563 PR tree-optimization/84946
8564 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
8565 bitsize + bitsize in poly_uint64 rather than poly_int64.
8568 * dwarf2asm.c: Include fold-const.c.
8569 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
8570 of decl rather than decl itself.
8572 PR rtl-optimization/84643
8573 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
8575 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
8578 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
8579 calling assemble_variable.
8581 2018-03-19 Sudakshina Das <sudi.das@arm.com>
8584 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
8585 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
8587 2018-03-19 Jim Wilson <jimw@sifive.com>
8590 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
8591 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
8592 (riscv_first_stack_step): Likewise.
8593 (riscv_option_override): Use STACK_BOUNDARY instead of
8595 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
8597 (BIGGEST_ALIGNMENT): Set to 128.
8598 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
8599 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
8602 2018-03-19 Richard Biener <rguenther@suse.de>
8604 PR tree-optimization/84933
8605 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
8606 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
8608 2018-03-19 Richard Biener <rguenther@suse.de>
8610 PR tree-optimization/84859
8611 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
8612 (cond_if_else_store_replacement): Perform sinking operation on
8613 single-store BBs regardless of MAX_STORES_TO_SINK setting.
8614 Generalize what a BB with a single eligible store is.
8616 2018-03-19 Richard Biener <rguenther@suse.de>
8618 PR tree-optimization/84929
8619 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
8620 chrec_is_positive against non-chrec arg.
8622 2018-03-19 Tamar Christina <tamar.christina@arm.com>
8625 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
8627 2018-03-18 Martin Liska <mliska@suse.cz>
8629 PR rtl-optimization/84635
8630 * regrename.c (build_def_use): Use matches_mode only when
8633 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
8635 PR tree-optimization/84913
8636 * tree-vect-loop.c (vectorizable_reduction): Don't try to
8637 vectorize chains of COND_EXPRs.
8639 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
8641 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
8643 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
8645 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
8647 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
8649 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
8651 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
8652 Kito Cheng <kito.cheng@gmail.com>
8654 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
8655 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
8656 (nds32_adjust_reg_alloc_order): New function.
8657 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
8659 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
8661 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
8662 nds32_print_operand, nds32_print_operand_address): Use
8663 HOST_WIDE_INT_PRINT_DEC instead.
8665 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
8667 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
8669 2018-03-17 Jakub Jelinek <jakub@redhat.com>
8672 * config/i386/i386.c (initial_ix86_tune_features,
8673 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
8675 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
8676 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
8677 rather than 1u << ix86_tune. Formatting fix.
8678 (ix86_option_override_internal): Change ix86_arch_mask from
8679 unsigned int to unsigned HOST_WIDE_INT, initialize to
8680 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
8681 (ix86_function_specific_restore): Likewise.
8683 2018-03-16 Jakub Jelinek <jakub@redhat.com>
8686 * postreload.c (reload_combine_recognize_pattern): Perform
8687 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
8688 truncate_int_for_mode the result for the destination's mode.
8691 * hsa-gen.c (mem_type_for_type): Fix comment typo.
8692 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
8694 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
8697 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
8700 * lra-assigns.c (lra_split_hard_reg_for): Don't use
8701 regno_allocno_class_array and sorted_pseudos.
8702 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
8703 insns where regno is used.
8705 2018-03-16 Martin Liska <mliska@suse.cz>
8708 * multiple_target.c (create_dispatcher_calls): Redirect
8709 reference in the symbol table.
8711 2018-03-16 Martin Liska <mliska@suse.cz>
8714 * multiple_target.c (create_dispatcher_calls): Redirect also
8717 2018-03-16 Jakub Jelinek <jakub@redhat.com>
8721 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
8722 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
8723 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
8725 2018-03-16 Julia Koval <julia.koval@intel.com>
8727 * doc/invoke.texi (Skylake Server): Add CLWB.
8728 Cannonlake): Remove CLWB.
8730 2018-03-16 Jakub Jelinek <jakub@redhat.com>
8732 PR tree-optimization/84841
8733 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
8735 (FLOAT_ONE_CONST_TYPE): Define.
8736 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
8737 (sort_by_operand_rank): Put entries with higher constant_type last
8738 rather than first to match comments.
8740 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
8742 * config/nios2/nios2.md (movsi_internal): Fix thinko in
8745 2018-03-15 Jakub Jelinek <jakub@redhat.com>
8748 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
8749 check and use address of target always.
8751 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
8754 * config/i386/i386.c (indirect_thunk_needed): Update comments.
8755 (indirect_thunk_bnd_needed): Likewise.
8756 (indirect_thunks_used): Likewise.
8757 (indirect_thunks_bnd_used): Likewise.
8758 (indirect_return_needed): New.
8759 (indirect_return_bnd_needed): Likewise.
8760 (output_indirect_thunk_function): Add a bool argument for
8762 (output_indirect_thunk_function): Don't generate alias for
8763 function return thunk.
8764 (ix86_code_end): Call output_indirect_thunk_function to generate
8765 function return thunks.
8766 (ix86_output_function_return): Set indirect_return_bnd_needed
8767 and indirect_return_needed instead of indirect_thunk_bnd_needed
8768 and indirect_thunk_needed.
8770 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
8772 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
8773 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
8774 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
8776 2018-03-15 David Malcolm <dmalcolm@redhat.com>
8777 Paul Hua <paul.hua.gm@gmail.com>
8780 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
8782 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
8784 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
8785 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
8788 2018-03-15 Tamar Christina <tamar.christina@arm.com>
8791 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
8792 instead of GET_MODE_SIZE when comparing Units.
8794 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
8797 * varasm.c (hash_section): Return an unchangeble hash value
8798 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
8799 Return !aarch64_can_use_per_function_literal_pools_p ().
8801 2018-03-15 Jakub Jelinek <jakub@redhat.com>
8804 * optabs.c (emit_conditional_move): Pass address of cmode's copy
8805 rather than address of cmode as last argument to prepare_cmp_insn.
8807 2018-03-15 Julia Koval <julia.koval@intel.com>
8809 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
8810 F_AVX512VNNI, F_AVX512BITALG): New.
8812 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
8815 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
8816 insn for floating-point loads and stores.
8818 2018-03-14 Carl Love <cel@us.ibm.com>
8820 * config/rs6000/rs6000-c.c: Add macro definitions for
8821 ALTIVEC_BUILTIN_VEC_PERMXOR.
8822 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
8823 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
8824 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
8825 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
8827 * config/doc/extend.texi: Add prototypes for vec_permxor.
8829 2018-03-14 David Malcolm <dmalcolm@redhat.com>
8832 * diagnostic-show-locus.c (class layout_point): Convert m_line
8833 from int to linenum_type.
8834 (line_span::comparator): Use linenum "compare" function when
8835 comparing line numbers.
8836 (test_line_span): New function.
8837 (layout_range::contains_point): Convert param "row" from int to
8839 (layout_range::intersects_line_p): Likewise.
8840 (layout::will_show_line_p): Likewise.
8841 (layout::print_source_line): Likewise.
8842 (layout::should_print_annotation_line_p): Likewise.
8843 (layout::print_annotation_line): Likewise.
8844 (layout::print_leading_fixits): Likewise.
8845 (layout::annotation_line_showed_range_p): Likewise.
8846 (struct line_corrections): Likewise for field m_row.
8847 (line_corrections::line_corrections): Likewise for param "row".
8848 (layout::print_trailing_fixits): Likewise.
8849 (layout::get_state_at_point): Likewise.
8850 (layout::get_x_bound_for_row): Likewise.
8851 (layout::print_line): Likewise.
8852 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
8853 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
8854 * input.c (selftest::test_linenum_comparisons): New function.
8855 (selftest::input_c_tests): Call it.
8856 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
8857 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
8858 * selftest.h (ASSERT_GT): New macro.
8859 (ASSERT_GT_AT): New macro.
8860 (ASSERT_LT): New macro.
8861 (ASSERT_LT_AT): New macro.
8863 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
8865 PR rtl-optimization/84780
8866 * combine.c (distribute_links): Don't make a link based on pc_rtx.
8868 2018-03-14 Martin Liska <mliska@suse.cz>
8870 * tree.c (record_node_allocation_statistics): Use
8871 get_stats_node_kind.
8872 (get_stats_node_kind): New function extracted from
8873 record_node_allocation_statistics.
8874 (free_node): Use get_stats_node_kind.
8876 2018-03-14 Richard Biener <rguenther@suse.de>
8878 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
8879 that the value-set of ANTIC_IN doesn't grow.
8882 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
8884 (BB_VISITED_WITH_VISITED_SUCCS): New define.
8885 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
8887 2018-03-14 Julia Koval <julia.koval@intel.com>
8889 * config.gcc (icelake-client, icelake-server): New.
8891 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
8892 (initial_ix86_arch_features): Ditto.
8893 (PTA_SKYLAKE): Add SGX.
8894 (PTA_ICELAKE): Remove.
8895 (PTA_ICELAKE_CLIENT): New.
8896 (PTA_ICELAKE_SERVER): New.
8897 (ix86_option_override_internal): Split up icelake on icelake client and
8899 (get_builtin_code_for_version): Ditto.
8900 (fold_builtin_cpu): Ditto.
8901 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
8902 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
8903 * config/i386/i386.h (processor_type): Ditto.
8904 * doc/invoke.texi: Ditto.
8906 2018-03-14 Jakub Jelinek <jakub@redhat.com>
8909 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
8910 INTEGER_CST offset, add it together with bitpos / 8 and
8911 sign extend based on POINTER_SIZE.
8915 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
8918 * config/i386/constraints.md (Yc): New register constraint.
8919 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
8920 Use Yc constraint for alternative 2 of operand 0. Remove
8921 preferred_for_speed attribute.
8923 2018-03-14 Richard Biener <rguenther@suse.de>
8925 PR tree-optimization/84830
8926 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
8927 with the old one to avoid oscillations.
8929 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
8932 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
8934 (assign_by_spills): Return a flag of reload assignment failure.
8935 Do not process the reload assignment failures. Do not spill other
8936 reload pseudos if they has the same reg class. Update n if
8938 (lra_assign): Add a return arg. Set up from the result of
8939 assign_by_spills call.
8940 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
8941 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
8942 usage_insns if it is not NULL.
8943 (spill_hard_reg_in_range): New function.
8944 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
8945 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
8946 function prototypes.
8947 (lra_assign): Change prototype.
8948 * lra.c (lra): Add code to deal with fails by splitting hard reg
8951 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
8953 * config/riscv/riscv.opt (mrelax): New option.
8954 * config/riscv/riscv.c (riscv_file_start): Emit ".option
8955 "norelax" when riscv_mrelax is disabled.
8956 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
8958 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8961 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
8962 reassociation for int modes.
8964 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
8966 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
8967 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
8969 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
8970 * config/aarch64/aarch64-sve.md
8971 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
8972 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
8973 (*extend<mode><Vwide>2): Rename to...
8974 (aarch64_sve_extend<mode><Vwide>2): ...this.
8975 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
8976 renaming the old pattern to...
8977 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
8979 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
8980 define_expand, renaming the old pattern to...
8981 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
8982 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
8983 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
8984 account when deciding which SVE instruction the optab should use.
8985 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
8987 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
8989 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
8990 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
8991 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
8992 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
8993 (tlsdesc_small_<mode>): Turn a define_expand and use
8994 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
8995 (tlsdesc_small_advsimd_<mode>): ...this.
8996 (tlsdesc_small_sve_<mode>): New pattern.
8998 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
9000 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
9001 (UNSPEC_UMUL_HIGHPART): New constants.
9002 (MUL_HIGHPART): New int iteraor.
9003 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
9004 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
9006 (*<su>mul<mode>3_highpart): New define_insn.
9008 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
9011 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
9014 2018-03-13 Martin Liska <mliska@suse.cz>
9017 * (sem_item_optimizer::sem_item_optimizer): Initialize new
9019 (sem_item_optimizer::~sem_item_optimizer): Release it.
9020 (sem_item_optimizer::merge_classes): Register variable aliases.
9021 (sem_item_optimizer::fixup_pt_set): New function.
9022 (sem_item_optimizer::fixup_points_to_sets): Likewise.
9023 * ipa-icf.h: Declare new variables and functions.
9025 2018-03-13 Jakub Jelinek <jakub@redhat.com>
9028 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
9029 integer_pow2p@2 and test integer_pow2p in condition.
9030 (A < 0 ? C : 0): Similarly for @1.
9033 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
9034 characters starting at p contain '\0' character, don't look beyond
9038 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
9039 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
9042 * reg-stack.c (change_stack): Change update_end var from int to
9043 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
9044 also call set_block_for_insn on the newly added insns and rescan.
9047 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
9048 on the last operand.
9051 * tree.c (stabilize_reference_1): Return save_expr (e) for
9052 STATEMENT_LIST even if it doesn't have side-effects.
9054 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
9056 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
9058 2018-03-12 Renlin Li <renlin.li@arm.com>
9060 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
9061 aarch64_output_scalar_simd_mov_immediate.
9063 2018-03-12 Martin Sebor <msebor@redhat.com>
9065 PR tree-optimization/83456
9066 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
9067 for perfectly overlapping calls to memcpy.
9068 (gimple_fold_builtin_memory_chk): Same.
9069 (gimple_fold_builtin_strcpy): Handle no-warning.
9070 (gimple_fold_builtin_stxcpy_chk): Same.
9071 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
9073 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
9075 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
9076 parameter. Use it for SFmode.
9077 (rs6000_function_arg_advance_1): Adjust.
9078 (rs6000_function_arg): Adjust.
9079 (rs6000_gimplify_va_arg): Pass false for that new parameter.
9081 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
9083 PR rtl-optimization/84169
9084 PR rtl-optimization/84780
9085 * combine.c (can_combine_p): Check for a 2-insn combination whether
9086 the destination register is used between the two insns, too.
9088 2018-03-12 Richard Biener <rguenther@suse.de>
9090 PR tree-optimization/84803
9091 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
9092 for refs DR analysis didn't process.
9094 2018-03-12 Richard Biener <rguenther@suse.de>
9096 PR tree-optimization/84777
9097 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
9098 force-vectorize loops ignore whether we are optimizing for size.
9100 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
9102 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
9103 (TARGET_MD_ASM_ADJUST): Define.
9105 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
9106 Kito Cheng <kito.cheng@gmail.com>
9107 Chung-Ju Wu <jasonwucj@gmail.com>
9109 * config/nds32/nds32.c (nds32_compute_stack_frame,
9110 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
9111 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
9112 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
9113 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
9114 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
9115 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
9116 * config/nds32/nds32.md (prologue, epilogue): Use macro
9117 NDS32_V3PUSH_AVAILABLE_P to do checking.
9119 2018-03-11 Jakub Jelinek <jakub@redhat.com>
9122 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
9123 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
9124 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
9125 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
9126 addition of most attributes on !orig_type_die or the attribute not
9127 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
9129 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
9130 Chung-Ju Wu <jasonwucj@gmail.com>
9132 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
9134 * config/nds32/nds32.opt (mvh): New option.
9136 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
9137 Chung-Ju Wu <jasonwucj@gmail.com>
9139 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
9141 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
9142 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
9145 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
9146 Chung-Ju Wu <jasonwucj@gmail.com>
9148 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
9150 * config/nds32/nds32-multiple.md (strlensi): New pattern.
9151 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
9153 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
9154 Kito Cheng <kito.cheng@gmail.com>
9155 Chung-Ju Wu <jasonwucj@gmail.com>
9157 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
9158 UNSPEC_FFMISM and UNSPEC_FLMISM.
9159 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
9160 for ffb, ffmism and flmism.
9161 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
9162 (unspec_ffmism): Ditto.
9163 (unspec_flmism): Ditto.
9164 (nds32_expand_builtin_impl): Check if string extension is available.
9165 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
9166 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
9168 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
9171 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
9174 * lra-assigns.c (assign_by_spills): Return a flag of reload
9175 assignment failure. Do not process the reload assignment
9176 failures. Do not spill other reload pseudos if they has the same
9178 (lra_assign): Add a return arg. Set up from the result of
9179 assign_by_spills call.
9180 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
9181 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
9182 usage_insns if it is not NULL.
9183 (spill_hard_reg_in_range): New function.
9184 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
9185 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
9186 function prototypes.
9187 (lra_assign): Change prototype.
9188 * lra.c (lra): Add code to deal with fails by splitting hard reg
9191 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
9194 * config/i386/i386.opt: Replace Enforcment with Enforcement.
9196 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
9199 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
9200 (dw_val_node): Add val_symbolic_view.
9201 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
9202 (symview_upper_bound): New.
9203 (new_line_info_table): Initialize symviews_since_reset.
9204 (dwarf2out_source_line): Count symviews_since_reset and set
9205 symview_upper_bound.
9206 (dw_val_equal_p): Handle symview.
9207 (add_AT_symview): New.
9208 (print_dw_val): Handle symview.
9209 (attr_checksum, attr_checksum_ordered): Likewise.
9210 (same_dw_val_p, size_of_die): Likewise.
9211 (value_format, output_die): Likewise.
9212 (add_high_low_attributes): Use add_AT_symview for entry_view.
9213 (dwarf2out_finish): Reset symview_upper_bound, clear
9216 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
9219 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
9220 Add strict argument and use it.
9221 (rs6000_split_multireg_move): Update for new strict argument.
9222 (mem_operand_gpr): Disallow all non-offsettable addresses.
9223 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
9225 2018-03-09 Jakub Jelinek <jakub@redhat.com>
9228 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
9229 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
9230 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
9233 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
9234 decl, use remap_type if we want to use the type.
9236 2018-03-09 Martin Sebor <msebor@redhat.com>
9238 PR tree-optimization/84526
9239 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
9241 (builtin_access::generic_overlap): Be prepared to handle non-array
9244 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
9246 PR rtl-optimization/84682
9247 * lra-constraints.c (process_address_1): Check is_address flag
9248 for address constraints.
9249 (process_alt_operands): Likewise.
9250 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
9251 preprocess_constraints.
9252 * recog.h (preprocess_constraints): Add oploc parameter.
9254 * recog.c (preprocess_constraints): Test address_operand for
9255 CT_ADDRESS constraints.
9257 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
9260 * lra-assigns.c (assign_by_spills): Return a flag of reload
9261 assignment failure. Do not process the reload assignment
9262 failures. Do not spill other reload pseudos if they has the same
9264 (lra_assign): Add a return arg. Set up from the result of
9265 assign_by_spills call.
9266 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
9267 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
9268 usage_insns if it is not NULL.
9269 (spill_hard_reg_in_range): New function.
9270 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
9271 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
9272 function prototypes.
9273 (lra_assign): Change prototype.
9274 * lra.c (lra): Add code to deal with fails by splitting hard reg
9277 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9280 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
9281 Accept complain bool parameter. Only emit errors if it is true.
9282 (arm_parse_cpu_option_name): Likewise.
9283 (arm_target_thumb_only): Adjust callers of the above.
9284 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
9285 prototype to take a default true bool parameter.
9286 (arm_parse_arch_option_name): Likewise.
9288 2018-03-09 David Malcolm <dmalcolm@redhat.com>
9289 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
9293 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
9294 * configure: Regenerate.
9295 * configure.ac ("linker --version-script option"): New.
9296 ("linker soname option"): New.
9298 2018-03-09 Richard Biener <rguenther@suse.de>
9300 PR tree-optimization/84775
9301 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
9302 immediate uses of predicate stmts and mark them modified.
9305 PR tree-optimization/84178
9306 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
9308 (version_loop_for_if_conversion): Delay update_ssa call.
9309 (tree_if_conversion): Delay update_ssa until after predicate
9312 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
9315 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
9316 when the function accesses prior frames.
9318 2018-03-08 Jakub Jelinek <jakub@redhat.com>
9321 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
9322 gen_llsym, otherwise call maybe_gen_llsym.
9325 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
9326 has ',' character inside of it.
9328 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9331 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
9332 as clobbering CC_REGNUM.
9334 2018-03-08 Richard Biener <rguenther@suse.de>
9337 * tree-scalar-evolution.c: Include tree-into-ssa.h.
9338 (follow_copies_to_constant): Do not follow SSA names registered
9341 2018-03-08 Richard Biener <rguenther@suse.de>
9343 PR tree-optimization/84178
9344 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
9346 (version_loop_for_if_conversion): Delay update_ssa call.
9347 (tree_if_conversion): Delay update_ssa until after predicate
9350 2018-03-08 David Malcolm <dmalcolm@redhat.com>
9352 PR tree-optimization/84178
9353 * tree-if-conv.c (release_bb_predicate): Remove the
9354 the assertion that the stmts have NULL use_ops.
9355 Discard the statements, asserting that they haven't
9356 yet been added to a BB.
9358 2018-03-08 Richard Biener <rguenther@suse.de>
9360 PR tree-optimization/84746
9361 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
9362 (phi_translate): Pass in destination ANTIC_OUT set.
9363 (phi_translate_1): Likewise. For a simplified result lookup
9364 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
9365 (phi_translate_set): Adjust.
9366 (do_pre_regular_insertion): Likewise.
9367 (do_pre_partial_partial_insertion): Likewise.
9369 2018-03-08 Martin Liska <mliska@suse.cz>
9371 PR gcov-profile/84735
9372 * doc/gcov.texi: Document usage of profile files.
9373 * gcov-io.h: Document changes in the format.
9375 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
9379 * dwarf2out.c (struct dw_line_info_table): Update comments for
9381 (FORCE_RESET_NEXT_VIEW): New.
9382 (FORCE_RESETTING_VIEW_P): New.
9383 (RESETTING_VIEW_P): Check for -1 too.
9384 (ZERO_VIEW_P): Likewise.
9385 (new_line_info_table): Force-reset next view.
9386 (dwarf2out_begin_function): Likewise.
9387 (dwarf2out_source_line): Simplify zero_view_p initialization.
9388 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
9389 view directly. Omit view when omitting .loc at line 0.
9391 2018-03-08 Jakub Jelinek <jakub@redhat.com>
9393 PR tree-optimization/84740
9394 * tree-switch-conversion.c (process_switch): Call build_constructors
9395 only if info.phi_count is non-zero.
9397 PR tree-optimization/84739
9398 * tree-tailcall.c (find_tail_calls): Check call arguments against
9399 DECL_ARGUMENTS (current_function_decl) rather than
9400 DECL_ARGUMENTS (func) when checking for tail recursion.
9402 2018-03-07 Jakub Jelinek <jakub@redhat.com>
9404 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
9405 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
9406 Volker Reichelt's entry and add entries for people that perform
9407 GCC fuzzy testing and report numerous bugs.
9409 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
9412 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
9413 readonly data in sdata, if that is disabled.
9414 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
9415 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
9416 -mreadonly-in-sdata option.
9418 2018-03-07 Martin Sebor <msebor@redhat.com>
9420 PR tree-optimization/84468
9421 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
9422 basic block when looking for nul assignment.
9424 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
9427 * except.h (output_function_exception_table): Adjust prototype.
9428 * except.c (output_function_exception_table): Remove FNNAME parameter
9429 and add SECTION parameter. Ouput one part of the table at a time.
9430 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
9431 the first part of the exception table and emit unwind directives.
9432 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
9433 (i386_pe_seh_cold_init): Likewise.
9434 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
9435 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
9436 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
9437 (ix86_output_call_insn): Emit a nop in one more case for SEH.
9438 * config/i386/winnt.c: Include except.h.
9439 (struct seh_frame_state): Add reg_offset, after_prologue and
9440 in_cold_section fields.
9441 (i386_pe_seh_end_prologue): Set seh->after_prologue.
9442 (i386_pe_seh_cold_init): New function.
9443 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
9444 to seh->in_cold_section.
9445 (seh_emit_push): Record the offset of the push.
9446 (seh_emit_save): Record the offet of the save.
9447 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
9448 Test seh->after_prologue to disregard the epilogue.
9449 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
9450 (i386_pe_end_cold_function): New function.
9452 2018-03-07 Jakub Jelinek <jakub@redhat.com>
9455 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
9456 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
9459 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
9461 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
9462 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
9465 * multiple_target.c: Include tree-inline.h and intl.h.
9466 (expand_target_clones): Diagnose and fail if node->definition and
9467 !tree_versionable_function_p (node->decl).
9469 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
9471 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
9473 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
9474 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
9475 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
9477 2018-03-06 Jakub Jelinek <jakub@redhat.com>
9480 * combine.c (try_combine): Use reg_or_subregno instead of handling
9481 just paradoxical SUBREGs and REGs.
9483 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
9485 * config/arc/arc.c (arc_finalize_pic): Remove function.
9486 (arc_must_save_register): We use single base PIC register, remove
9487 checks to save/restore the PIC register.
9488 (arc_expand_prologue): Likewise.
9489 * config/arc/arc-protos.h (arc_set_default_type_attributes):
9491 (arc_verify_short): Likewise.
9492 (arc_attr_type): Likewise.
9493 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
9494 (walk_stores): Likewise.
9495 (arc_address_cost): Make it static.
9496 (arc_verify_short): Likewise.
9497 (branch_dest): Likewise.
9498 (arc_attr_type): Likewise.
9499 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
9500 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
9501 (arc_final_prescan_insn): Remove inserting the nops due to
9502 hardware hazards. It is done in reorg step.
9503 (insn_length_variant_t): Remove.
9504 (insn_length_parameters_t): Likewise.
9505 (arc_insn_length_parameters): Likewise.
9506 (arc_get_insn_variants): Likewise.
9507 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
9509 2018-03-06 Jakub Jelinek <jakub@redhat.com>
9512 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
9515 PR tree-optimization/84687
9516 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
9518 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
9520 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9522 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
9523 Rename to ppc_speculation_barrier.
9524 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
9525 __builtin_ppc_speculation_barrier.
9527 2018-03-05 Jakub Jelinek <jakub@redhat.com>
9530 * combine.c (combine_simplify_rtx): Don't try to simplify if
9531 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
9534 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
9536 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
9537 to 32 bytes when compiling for POWER9.
9539 2018-03-05 Jakub Jelinek <jakub@redhat.com>
9542 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
9543 regparm >= 3 with no arg reg available also for calls with
9544 flag_force_indirect_call. Pass decl to ix86_function_regparm.
9547 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
9549 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
9551 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
9554 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
9556 2018-03-05 Richard Biener <rguenther@suse.de>
9558 PR tree-optimization/84486
9559 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
9560 When inserting a __builtin_assume_aligned call set the LHS
9561 SSA name alignment info accordingly.
9563 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
9565 PR tree-optimization/84114
9566 * config/aarch64/aarch64.c (aarch64_reassociation_width)
9567 Avoid reassociation of FLOAT_MODE addition.
9569 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
9571 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
9572 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
9573 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
9574 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
9575 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
9576 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
9577 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
9579 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
9580 __builtin_ia32_wbinvd): New builtins.
9581 (SPECIAL_ARGS2): New.
9582 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
9583 (SPECIAL_ARGS2): New.
9584 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
9585 (ix86_valid_target_attribute_inner_p): Ditto.
9586 (ix86_init_mmx_sse_builtins): Add special_args2.
9587 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
9588 TARGET_WBNOINVD_P): New.
9589 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
9590 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
9591 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
9592 * config/i386/immintrin.h (_wbinvd): New intrinsic.
9593 * config/i386/pconfigintrin.h: New file.
9594 * config/i386/wbnoinvdintrin.h: Ditto.
9595 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
9597 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
9599 2018-03-05 Richard Biener <rguenther@suse.de>
9601 PR tree-optimization/84670
9602 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
9604 (BB_VISITED_WITH_VISITED_SUCCS): New define.
9605 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
9606 (compute_antic_aux): Only assert the number of values in ANTIC_IN
9607 doesn't grow if all successors (recursively) were visited at least
9610 2018-03-05 Richard Biener <rguenther@suse.de>
9612 PR tree-optimization/84650
9613 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
9614 if executed in the loop pipeline.
9616 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
9618 * doc/configfiles.texi (Configuration Files): Move info about
9619 conditionalizing $target-protos.h to...
9620 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
9621 differs from $target-protos.h.
9623 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
9624 Chung-Ju Wu <jasonwucj@gmail.com>
9626 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
9627 * config/nds32/nds32-multiple.md (setmemsi): Define.
9628 * config/nds32/nds32-memory-manipulation.c
9629 (nds32_gen_dup_4_byte_to_word_value): New.
9630 (emit_setmem_word_loop): New.
9631 (emit_setmem_byte_loop): New.
9632 (nds32_expand_setmem_loop): New.
9633 (nds32_expand_setmem_loop_v3m): New.
9634 (nds32_expand_setmem_unroll): New.
9635 (nds32_expand_setmem): New.
9637 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
9638 Chung-Ju Wu <jasonwucj@gmail.com>
9640 * config/nds32/nds32-memory-manipulation.c
9641 (nds32_emit_load_store): New.
9642 (nds32_emit_post_inc_load_store): New.
9643 (nds32_emit_mem_move): New.
9644 (nds32_emit_mem_move_block): New.
9645 (nds32_expand_movmemsi_loop_unknown_size): New.
9646 (nds32_expand_movmemsi_loop_known_size): New.
9647 (nds32_expand_movmemsi_loop): New.
9648 (nds32_expand_movmemsi_unroll): New.
9649 (nds32_expand_movmemqi): Rename ...
9650 (nds32_expand_movmemsi): ... to this.
9651 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
9652 (movmemsi): ... to this.
9653 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
9654 (nds32_expand_movmemsi): ... to this.
9656 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
9657 Monk Chiang <sh.chiang04@gmail.com>
9658 Chung-Ju Wu <jasonwucj@gmail.com>
9660 * config/nds32/nds32-protos.h
9661 (nds32_expand_load_multiple): New arguments.
9662 (nds32_expand_store_multiple): Ditto.
9663 (nds32_valid_multiple_load_store): Rename ...
9664 (nds32_valid_multiple_load_store_p): ... to this.
9665 * config/nds32/nds32-memory-manipulation.c
9666 (nds32_expand_load_multiple): Refine implementation.
9667 (nds32_expand_store_multiple): Ditto.
9668 * config/nds32/nds32-multiple.md
9669 (load_multiple): Update nds32_expand_load_multiple interface.
9670 (store_multiple): Update nds32_expand_store_multiple interface.
9671 * config/nds32/nds32-predicates.c
9672 (nds32_valid_multiple_load_store): Rename ...
9673 (nds32_valid_multiple_load_store_p): ... to this and refine
9675 * config/nds32/predicates.md
9676 (nds32_load_multiple_and_update_address_operation): New predicate.
9677 (nds32_store_multiple_and_update_address_operation): New predicate.
9679 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
9680 Chung-Ju Wu <jasonwucj@gmail.com>
9682 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
9683 (combo): New attribute.
9684 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
9686 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
9688 * config/nds32/nds32.opt: Change -mcmodel= default value.
9690 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
9691 Monk Chiang <sh.chiang04@gmail.com>
9692 Chung-Ju Wu <jasonwucj@gmail.com>
9694 * config/nds32/constants.md (unspec_element): New enum.
9695 * config/nds32/constraints.md (Umw): New constraint.
9696 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
9697 * config/nds32/nds32-intrinsic.md: Likewise.
9698 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
9699 (nds32_valid_smw_lwm_base_p): New.
9700 (nds32_output_smw_single_word): New.
9701 (nds32_output_lmw_single_word): New.
9702 (nds32_expand_unaligned_load): New.
9703 (nds32_expand_unaligned_store): New.
9704 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
9705 (nds32_output_smw_single_word): Declare.
9706 (nds32_output_lmw_single_word): Declare.
9707 (nds32_expand_unaligned_load): Declare.
9708 (nds32_expand_unaligned_store): Declare.
9709 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
9710 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
9711 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
9712 NDS32_BUILTIN_UASTORE_DW.
9713 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
9716 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
9717 Kito Cheng <kito.cheng@gmail.com>
9718 Chung-Ju Wu <jasonwucj@gmail.com>
9720 * config/nds32/nds32-intrinsic.c
9721 (nds32_expand_builtin_null_ftype_reg): Delete.
9722 (nds32_expand_builtin_reg_ftype_imm): Ditto.
9723 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
9724 (nds32_read_argument): New.
9725 (nds32_legitimize_target): Ditto.
9726 (nds32_legitimize_argument): Ditto.
9727 (nds32_check_constant_argument): Ditto.
9728 (nds32_expand_unop_builtin): Ditto.
9729 (nds32_expand_unopimm_builtin): Ditto.
9730 (nds32_expand_binop_builtin): Ditto.
9731 (nds32_builtin_decl_impl): Ditto.
9732 (builtin_description): Ditto.
9733 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
9734 (nds32_init_builtins_impl): Ditto.
9735 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
9736 (nds32_builtin_decl): New.
9737 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
9738 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
9740 2018-03-02 Jeff Law <law@redhat.com>
9742 * reorg.c (stop_search_p): Handle DEBUG_INSN.
9743 (redundant_insn, fill_simple_delay_slots): Likewise.
9744 (fill_slots_from_thread): Likewise.
9745 * resource.c (mark_referenced_resources): Likewise.
9746 (mark_set_resources, find_dead_or_set_registers): Likewise.
9748 2018-03-02 Jakub Jelinek <jakub@redhat.com>
9750 * substring-locations.h (format_warning_va): Formatting fix for
9752 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
9754 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
9755 * substring-locations.c: Include intl.h.
9756 (format_warning_va): Turned into small wrapper around
9757 format_warning_n_va, renamed to ...
9758 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
9759 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
9761 (format_warning_at_substring_n): New function.
9762 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
9763 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
9764 format_warning_at_substring with just a shorter name instead of
9765 const function pointer.
9766 (fmtwarn_n): New function.
9767 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
9768 appropriate, get rid of all the fmtstr temporaries, move conditionals
9769 with G_() wrapped string literals directly into fmtwarn arguments,
9770 cast dir.len to (int), formatting fixes.
9772 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
9774 * doc/invoke.texi: Remove "Cilk Plus" references.
9776 2018-03-02 Jakub Jelinek <jakub@redhat.com>
9777 Richard Biener <rguenther@suse.de>
9780 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
9781 for error or warning attributes if CALL_FROM_THUNK_P is set.
9784 2018-03-02 Jakub Jelinek <jakub@redhat.com>
9787 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
9788 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
9791 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
9792 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
9794 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
9795 instead of -1U in last predictors element's probability member.
9797 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
9800 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
9801 arguments if they are comparable.
9803 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
9805 PR tree-optimization/84634
9806 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
9807 masks and masked_loop_p with a single loop_masks, making sure it's
9808 null for bb vectorization.
9810 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
9812 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
9813 (vect_analyze_data_ref_access): Use loop->safe_len rather than
9814 loop->force_vectorize to check whether there is no alias.
9816 2018-03-02 Jakub Jelinek <jakub@redhat.com>
9819 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
9821 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
9823 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
9824 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
9825 instead of a loop around prev_real_insn.
9826 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
9830 * config/i386/i386.c (ix86_print_operand): Use conditional
9831 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
9834 2018-03-02 Richard Biener <rguenther@suse.de>
9836 PR tree-optimization/84427
9837 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
9838 (bitmap_set_subtract_values): Rewrite to handle multiple
9841 (prune_clobbered_mems): Likewise.
9842 (phi_translate): Take edge instead of pred/phiblock.
9843 (phi_translate_1): Likewise.
9844 (phi_translate_set): Likewise. Insert all translated
9845 exprs for a value into the set, keeping possibly multiple
9846 expressions per value.
9847 (compute_antic_aux): Adjust for phi_translate changes.
9848 When intersecting union the expressions and prune those
9849 not in the final value set, keeping possibly multiple
9850 expressions per value. Do not use value-insertion
9851 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
9852 all expressions. Add verification that the value-sets
9853 only shrink during iteration.
9854 (compute_partial_antic_aux): Adjust for the phi_translate changes.
9855 (do_pre_regular_insertion): Likewise.
9856 (do_pre_partial_partial_insertion): Likewise.
9858 2018-03-02 Richard Biener <rguenther@suse.de>
9861 * config/darwin.c (saved_debug_info_level): New static global.
9862 (darwin_asm_lto_start): Disable debug info generation for LTO out.
9863 (darwin_asm_lto_end): Restore debug info generation settings.
9865 2018-03-01 Martin Liska <mliska@suse.cz>
9868 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
9871 2018-03-01 Richard Biener <rguenther@suse.de>
9874 * dwarf2out.c (gen_variable_die): Properly handle late VLA
9875 type annotation with LTO when debug was disabled at compile-time.
9877 2018-03-01 Matthew Fortune <mfortune@gmail.com>
9879 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
9881 (mips_final_postscan_insn): Likewise.
9883 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
9885 PR rtl-optimization/84528
9886 * alias.c (init_alias_target): Add commentary.
9887 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
9888 a unique base value if the frame pointer is not eliminated
9889 to the stack pointer.
9891 2018-03-01 Tom de Vries <tom@codesourcery.com>
9893 PR rtl-optimization/83327
9894 * lra-int.h (hard_regs_spilled_into): Declare.
9895 * lra.c (hard_regs_spilled_into): Define.
9896 (init_reg_info): Init hard_regs_spilled_into.
9897 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
9898 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
9899 (process_bb_lives): Handle hard_regs_spilled_into.
9900 (lra_create_live_ranges_1): Before doing liveness propagation, clear
9901 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
9903 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
9905 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
9906 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
9907 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
9908 * config/rs6000/aix72.h: New file.
9910 2018-02-28 Jakub Jelinek <jakub@redhat.com>
9912 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
9913 instead of warning_at with conditional singular and plural messages
9917 * stor-layout.c (update_alignment_for_field): For
9918 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
9919 && !DECL_PACKED (field), do the alignment update, just use
9920 only desired_align instead of MAX (type_align, desired_align)
9922 (place_field): Don't do known_align < desired_align handling
9923 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
9924 is non-NULL, instead do it after rli->prev_field handling and
9925 only if not within a bitfield word. For DECL_PACKED (field)
9926 use type_align of BITS_PER_UNIT.
9928 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
9930 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
9931 superfluous parentheses and trailing spaces.
9933 2018-02-28 Richard Biener <rguenther@suse.de>
9935 PR tree-optimization/84584
9936 * graphite-scop-detection.c (scop_detection::add_scop): Discard
9937 SCoPs with fake exit edge.
9939 2018-02-28 Martin Liska <mliska@suse.cz>
9942 * timevar.c (timer::print): Fix format to properly print 100%
9945 2018-02-28 Richard Biener <rguenther@suse.de>
9948 * genmatch.c (capture_info::walk_match): Do not mark
9949 captured expressions without operands as expr_p given
9950 they act more like predicates and should be subject to
9951 "lost tail" side-effect preserving.
9953 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
9955 PR rtl-optimization/81611
9956 * auto-inc-dec.c (attempt_change): Move dead note from
9957 mem_insn if it's the next use of regno
9958 (find_address): Take address use of reg holding
9959 non-incremented value. Add parm to limit search to the named
9961 (merge_in_block): Attempt to use a mem insn that is the next
9962 use of the original regno.
9964 2018-02-27 Martin Sebor <msebor@redhat.com>
9967 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
9968 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
9970 2018-02-27 Martin Sebor <msebor@redhat.com>
9972 PR translation/84207
9973 * diagnostic-core.h (warning_n, error_n, inform_n): Change
9974 n argument to unsigned HOST_WIDE_INT.
9975 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
9976 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
9977 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
9978 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
9980 2018-02-27 Richard Biener <rguenther@suse.de>
9982 PR tree-optimization/84512
9983 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
9984 Do not use the estimate returned from record_stmt_cost for
9985 the scalar iteration cost but sum properly using add_stmt_cost.
9987 2018-02-27 Richard Biener <rguenther@suse.de>
9989 PR tree-optimization/84466
9990 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
9991 Adjust last change to less strictly validate use operands.
9993 2018-02-27 Martin Liska <mliska@suse.cz>
9995 PR gcov-profile/84548
9996 * gcov.c (process_file): Allow partial overlap and consider it
9997 also as group functions.
9998 (output_lines): Properly calculate range of lines for a group.
10000 2018-02-27 Martin Liska <mliska@suse.cz>
10002 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
10003 'ggc' suffixes. Change first column width.
10004 (timer::print): Fix formatting of the column.
10006 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
10008 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
10009 preserve inline entry blocks for the sake of debug inline
10010 entry point markers alone.
10011 (remove_unused_locals): Suggest in comments a better place to
10012 force the preservation of inline entry blocks that are
10013 otherwise unused, but do not preserve them.
10015 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
10017 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
10019 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
10022 * config/i386/constraints.md (Bs): Replace
10023 ix86_indirect_branch_register with
10024 TARGET_INDIRECT_BRANCH_REGISTER.
10026 * config/i386/i386.md (indirect_jump): Likewise.
10027 (tablejump): Likewise.
10028 (*sibcall_memory): Likewise.
10029 (*sibcall_value_memory): Likewise.
10030 Peepholes of indirect call and jump via memory: Likewise.
10031 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
10032 (*sibcall_value_GOT_32): Likewise.
10033 * config/i386/predicates.md (indirect_branch_operand): Likewise.
10034 (GOT_memory_operand): Likewise.
10035 (call_insn_operand): Likewise.
10036 (sibcall_insn_operand): Likewise.
10037 (GOT32_symbol_operand): Likewise.
10038 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
10040 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
10042 PR rtl-optimization/83496
10043 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
10044 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
10045 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
10046 redundant insn, if any.
10047 (relax_delay_slots): Likewise.
10048 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
10050 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
10052 PR tree-optimization/83965
10053 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
10054 that grouped statements are part of a reduction chain. Return
10055 true if the statement is not marked as a reduction itself but
10056 is part of a group.
10057 (vect_recog_dot_prod_pattern): Don't check whether the statement
10058 is part of a group here.
10059 (vect_recog_sad_pattern): Likewise.
10060 (vect_recog_widen_sum_pattern): Likewise.
10062 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
10065 * final.c (rest_of_clean_state): Also look for calls inside sequences.
10067 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
10070 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
10072 (ix86_output_indirect_function_return): New prototype.
10073 (ix86_split_simple_return_pop_internal): Likewise.
10074 * config/i386/i386.c (indirect_return_via_cx): New.
10075 (indirect_return_via_cx_bnd): Likewise.
10076 (indirect_thunk_name): Handle return va CX_REG.
10077 (output_indirect_thunk_function): Create alias for
10078 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
10079 (ix86_output_indirect_jmp): Remove the bool argument.
10080 (ix86_output_indirect_function_return): New function.
10081 (ix86_split_simple_return_pop_internal): Likewise.
10082 * config/i386/i386.md (*indirect_jump): Don't pass false
10083 to ix86_output_indirect_jmp.
10084 (*tablejump_1): Likewise.
10085 (simple_return_pop_internal): Change it to define_insn_and_split.
10086 Call ix86_split_simple_return_pop_internal to split it for
10087 -mfunction-return=.
10088 (simple_return_indirect_internal): Call
10089 ix86_output_indirect_function_return instead of
10090 ix86_output_indirect_jmp.
10092 2018-02-26 Jakub Jelinek <jakub@redhat.com>
10095 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
10096 memset and value initialization afterwards.
10098 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
10100 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
10102 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
10105 * common/config/aarch64/aarch64-common.c
10106 (aarch_option_optimization_table[]): Switch
10107 off fomit-frame-pointer
10109 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
10110 Chung-Ju Wu <jasonwucj@gmail.com>
10112 * config/nds32/nds32-multiple.md (load_multiple): Disallow
10114 (store_multiple): Ditto.
10116 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
10118 * config.gcc: Add --with-cpu support for nds32 target.
10119 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
10120 * config/nds32/nds32.opt: Add -mcpu= option.
10122 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
10124 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
10125 isel=yes): Warn for these deprecated options.
10127 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
10129 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
10130 ISA_2_5_MASKS_EMBEDDED.
10132 2018-02-23 Jakub Jelinek <jakub@redhat.com>
10134 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
10135 p->max as pointers rather than using iterative_hash_expr.
10137 2018-02-23 Carl Love <cel@us.ibm.com>
10139 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
10140 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
10142 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
10143 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
10144 P8V_BUILTIN_VEC_VUNSIGNED2.
10146 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
10149 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
10150 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
10152 * lra-constraints.c (curr_insn_transform): Set up
10153 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
10155 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
10156 * lra-eliminations.c (spill_pseudos): Ditto.
10157 (process_insn_for_elimination): Ditto.
10158 * lra-lives.c (reg_early_clobber_p): Use the new macros.
10159 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
10160 LRA_NON_CLOBBERED_ALT.
10162 2018-02-22 Martin Sebor <msebor@redhat.com>
10164 PR tree-optimization/84480
10165 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
10166 to maybe_diag_stxncpy_trunc. Call it.
10167 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
10168 from gimple_fold_builtin_strcpy. Print inlining stack.
10169 (handle_builtin_stxncpy): Print inlining stack.
10170 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
10172 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
10175 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
10176 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
10177 and -fcheck-pointer-bounds are used together.
10178 (indirect_thunk_prefix): New enum.
10179 (indirect_thunk_need_prefix): New function.
10180 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
10181 "_nt" instead of "_bnd" for NOTRACK prefix.
10182 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
10183 (output_indirect_thunk_function): Likewise.
10185 (ix86_code_end): Update output_indirect_thunk_function calls.
10186 (ix86_output_indirect_branch_via_reg): Replace
10187 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
10188 (ix86_output_indirect_branch_via_push): Likewise.
10189 (ix86_output_function_return): Likewise.
10190 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
10191 incompatible with -fcf-protection=branch and
10192 -fcheck-pointer-bounds.
10194 2018-02-22 Steve Ellcey <sellcey@cavium.com>
10197 * config/aarch64/aarch64.c (aarch64_print_address_internal):
10198 Change gcc_assert call to output_operand_lossage.
10200 2018-02-22 Steve Ellcey <sellcey@cavium.com>
10202 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
10204 2018-02-22 DJ Delorie <dj@redhat.com>
10205 Sebastian Perta <sebastian.perta@renesas.com>
10206 Oleg Endo <olegendo@gcc.gnu.org>
10208 * config/rx/rx.c (rx_rtx_costs): New function.
10209 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
10211 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
10213 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
10215 2018-02-22 Martin Liska <mliska@suse.cz>
10218 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
10219 Add "native" as a possible value.
10221 2018-02-22 Martin Liska <mliska@suse.cz>
10224 * config/i386/i386.c (ix86_option_override_internal):
10225 Add "native" as a possible value for -march and -mtune.
10227 2018-02-22 Jakub Jelinek <jakub@redhat.com>
10230 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
10231 to all type variants.
10233 PR tree-optimization/84503
10234 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
10235 width as info->bitpos + info->bitsize - start.
10236 (merged_store_group::merge_overlapping): Simplify width computation.
10237 (check_no_overlap): New function.
10238 (imm_store_chain_info::try_coalesce_bswap): Compute expected
10239 start + width and last_order of the group, fail if check_no_overlap
10241 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
10242 to group if check_no_overlap fails.
10244 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
10246 * config/rs6000/altivec.md: Delete contraint arguments to
10247 define_expand, define_split, and define_peephole2, and in
10248 define_insn_and_split if always unused.
10249 * config/rs6000/darwin.md: Ditto.
10250 * config/rs6000/dfp.md: Ditto.
10251 * config/rs6000/rs6000.md: Ditto.
10252 * config/rs6000/sync.md: Ditto.
10253 * config/rs6000/vector.md: Ditto.
10254 * config/rs6000/vsx.md: Ditto.
10256 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
10258 * config/rs6000/altivec.md: Write output control strings as braced
10259 blocks instead of double-quoted strings.
10260 * config/rs6000/darwin.md: Ditto.
10261 * config/rs6000/rs6000.md: Ditto.
10262 * config/rs6000/vector.md: Ditto.
10263 * config/rs6000/vsx.md: Ditto.
10265 2018-02-21 Jason Merrill <jason@redhat.com>
10267 PR c++/84314 - ICE with templates and fastcall attribute.
10268 * attribs.c (build_type_attribute_qual_variant): Remove assert.
10270 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
10272 * ipa-cp.c (determine_versionability): Fix comment typos.
10274 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
10277 * ipa-cp.c (determine_versionability): Do not version functions caling
10280 2018-02-21 Martin Liska <mliska@suse.cz>
10283 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
10284 Add "native" as a possible value.
10285 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
10286 the macro when native cpu detection is available.
10288 2018-02-21 Martin Liska <mliska@suse.cz>
10291 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
10292 Add "native" as a possible value.
10293 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
10294 when native cpu detection is available.
10296 2018-02-21 Jakub Jelinek <jakub@redhat.com>
10297 Martin Sebor <msebor@redhat.com>
10299 PR tree-optimization/84478
10300 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
10302 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
10303 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
10304 support which is conservatively correct, for 2 only stay conservative
10305 for maxlen. Formatting and comment capitalization fixes. Add STRICT
10306 argument to the 2 argument get_range_strlen, adjust 6 arg
10307 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
10309 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
10310 (gimple_fold_builtin_strlen): Pass true as last argument to
10313 2018-02-20 Martin Sebor <msebor@redhat.com>
10315 PR middle-end/84095
10316 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
10317 (builtin_memref::set_base_and_offset): Same. Handle inner references.
10318 (builtin_memref::builtin_memref): Factor out parts into
10319 set_base_and_offset and call it.
10321 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
10323 PR middle-end/84406
10324 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
10325 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
10326 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
10327 search at the associated MODE_INT.
10329 2018-02-20 Jeff Law <law@redhat.com>
10331 PR middle-end/82123
10332 PR tree-optimization/81592
10333 PR middle-end/79257
10334 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
10335 for range data rather than using global data.
10336 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
10337 range data rather than using global data.
10338 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
10339 pass it to children as needed.
10340 (struct directive::fmtresult): Similarly.
10341 (struct directive::set_width): Similarly.
10342 (struct directive::set_precision): Similarly.
10343 (format_integer, format_directive, parse_directive): Similarly.
10344 (format_none): Accept unnamed vr_values parameter.
10345 (format_percent, format_floating, format_character): Similarly.
10346 (format_string, format_plain): Similarly.
10347 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
10348 the EVRP range analyzer for range data rather than using global data.
10349 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
10350 gimple-ssa-evrp-analyze.h
10351 (class sprintf_dom_walker): Add after_dom_children member function.
10352 Add evrp_range_analyzer member.
10353 (sprintf_dom_walker::before_dom_children): Call into the EVRP
10354 range analyzer as needed.
10355 (sprintf_dom_walker::after_dom_children): New member function.
10356 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
10358 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
10359 (evrp_range_analyzer::pop_to_marker): Likewise.
10361 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
10363 PR tree-optimization/84419
10364 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
10365 with the required type if its current type is compatible but
10368 2018-02-20 Jakub Jelinek <jakub@redhat.com>
10370 PR middle-end/82004
10371 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
10372 after vectorization.
10374 2018-02-20 Martin Liska <mliska@suse.cz>
10377 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
10378 possible values if we don't have a hint.
10380 2018-02-20 Martin Liska <mliska@suse.cz>
10384 * final.c (shorten_branches): Build align_tab array with one
10386 * opts.c (finish_options): Add alignment option limit check.
10387 (MAX_CODE_ALIGN): Likewise.
10388 (MAX_CODE_ALIGN_VALUE): Likewise.
10389 * doc/invoke.texi: Document maximum allowed option value for
10390 all -falign-* options.
10392 2018-02-19 Jakub Jelinek <jakub@redhat.com>
10395 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
10396 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
10397 * var-tracking.c (emit_note_insn_var_location): Remove all references
10398 to NOTE_INSN_CALL_ARG_LOCATION.
10399 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
10400 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
10401 Use copy_rtx_if_shared.
10402 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
10403 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
10404 (dwarf2out_var_location): Remove handling of
10405 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
10407 * final.c (final_scan_insn): Remove all references to
10408 NOTE_INSN_CALL_ARG_LOCATION.
10409 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
10410 before dumping final insns.
10411 * except.c (emit_note_eh_region_end): Remove all references to
10412 NOTE_INSN_CALL_ARG_LOCATION.
10413 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
10414 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
10415 * config/arc/arc.c (hwloop_optimize): Likewise.
10416 * config/arm/arm.c (create_fix_barrier): Likewise.
10417 * config/s390/s390.c (s390_chunkify_start): Likewise.
10418 * config/sh/sh.c (find_barrier): Likewise.
10419 * config/i386/i386.c (rest_of_insert_endbranch,
10420 ix86_seh_fixup_eh_fallthru): Likewise.
10421 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
10422 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
10423 * config/frv/frv.c (frv_function_prologue): Likewise.
10424 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
10426 (note_outside_basic_block_p): Remove all references to
10427 NOTE_INSN_CALL_ARG_LOCATION.
10428 * gengtype.c (adjust_field_rtx_def): Likewise.
10429 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
10431 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
10432 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
10435 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
10438 PR tree-optimization/84452
10439 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
10440 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
10443 2018-02-19 Martin Liska <mliska@suse.cz>
10446 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
10448 2018-02-19 Martin Liska <mliska@suse.cz>
10449 Richard Sandiford <richard.sandiford@linaro.org>
10451 PR tree-optimization/82491
10452 * gimple-fold.c (get_base_constructor): Make earlier bail out
10455 2018-02-19 Carl Love <cel@us.ibm.com>
10457 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
10458 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
10460 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
10461 P8V_BUILTIN_VEC_NEG.
10463 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
10465 * config/rl78/rl78.md (movdf): New define expand.
10467 2018-02-19 Martin Liska <mliska@suse.cz>
10470 * doc/invoke.texi: Fix typo.
10471 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
10473 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
10475 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
10476 handle rs6000_single_float and rs6000_double_float specially for
10479 2018-02-16 Jeff Law <law@redhat.com>
10481 * config/rx/rx.c (add_pop_cfi_notes): New function.;
10482 (pop_regs): Use it.
10484 2018-02-16 Jakub Jelinek <jakub@redhat.com>
10487 * ipa-inline.c (inline_small_functions): Fix a typo.
10489 2018-02-16 Nathan Sidwell <nathan@acm.org>
10491 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
10493 2018-02-16 Carl Love <cel@us.ibm.com>
10495 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
10496 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
10497 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
10498 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
10499 expansion to P8V_BUILTIN_VEC_FLOAT2.
10501 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
10503 PR rtl-optimization/70023
10504 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
10505 src_regno into account.
10507 2018-02-16 Carl Love <cel@us.ibm.com>
10509 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
10510 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
10511 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
10512 * config/rs6000/rs6000.c: Remove case statements for
10513 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
10514 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
10515 and P9V_BUILTIN_VEC_VINSERT4B.
10516 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
10517 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
10518 * config/rs6000/vsx.md:
10519 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
10522 2018-02-16 Carl Love <cel@us.ibm.com>
10524 * config/rs6000/altivec.h: Add builtin names vec_extract4b
10526 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
10528 * config/rs6000/rs6000-c.c: Add the definitions for
10529 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
10530 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
10531 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
10532 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
10533 definition for insert4b and define insn *insert3b_internal.
10534 * doc/extend.texi: Add documentation for vec_extract4b.
10536 2018-02-16 Nathan Sidwell <nathan@acm.org>
10538 * doc/extend.texi (Backwards Compatibility): Mention friend
10539 injection. Note for-scope is deprecated.
10540 * doc/invoke.texi (-ffriend-injection): Deprecate.
10542 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
10544 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
10545 that moved to I2, also allow destinations that are a paradoxical
10546 subreg (instead of a normal reg).
10548 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
10551 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
10554 2018-02-16 Richard Biener <rguenther@suse.de>
10556 PR tree-optimization/84037
10557 PR tree-optimization/84016
10559 * config/i386/i386.c (ix86_builtin_vectorization_cost):
10560 Adjust vec_construct for the fact we need additional higher latency
10561 128bit inserts for AVX256 and AVX512 vector builds.
10562 (ix86_add_stmt_cost): Scale vector construction cost for
10565 2018-02-16 Richard Biener <rguenther@suse.de>
10567 PR tree-optimization/84417
10568 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
10569 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
10570 (non_rewritable_lvalue_p): Likewise, use poly-ints.
10572 2018-02-16 Martin Liska <mliska@suse.cz>
10575 * internal-fn.def (ASAN_CHECK): Set proper flags.
10576 (ASAN_MARK): Likewise.
10578 2018-02-16 Julia Koval <julia.koval@intel.com>
10580 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
10581 from PTA_CANNONLAKE.
10583 2018-02-16 Jakub Jelinek <jakub@redhat.com>
10586 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
10587 Use ++iter rather than iter++ for std::list iterators.
10588 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
10589 defer deleting them until all nodes in the forest are processed. Do
10590 free even leaf nodes. Change to_process into auto_vec.
10593 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
10594 * vec.h (vec_default_construct): Use memset instead of placement new
10595 if BROKEN_VALUE_INITIALIZATION is defined.
10596 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
10597 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
10600 PR rtl-optimization/83723
10601 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
10602 * lra.c (lra_substitute_pseudo): Likewise. If true, use
10603 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
10605 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
10607 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
10609 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
10611 PR rtl-optimization/81443
10612 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
10613 from inner REGs to paradoxical SUBREGs.
10615 2018-02-16 Richard Biener <rguenther@suse.de>
10617 PR tree-optimization/84399
10618 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
10619 For operands we can analyze at their definition make sure we can
10620 analyze them at each use as well.
10622 2018-02-16 Richard Biener <rguenther@suse.de>
10624 PR tree-optimization/84190
10625 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
10626 volatile accesses if the decl isn't volatile.
10628 2018-02-15 Jason Merrill <jason@redhat.com>
10630 PR c++/84314 - ICE with templates and fastcall attribute.
10631 * attribs.c (build_type_attribute_qual_variant): Don't clobber
10632 TYPE_CANONICAL on an existing type.
10634 2018-02-15 Jakub Jelinek <jakub@redhat.com>
10636 PR tree-optimization/84383
10637 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
10638 dstoff nor call operand_equal_p if dstbase is NULL.
10640 PR tree-optimization/84334
10641 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
10642 also a CONSTANT_CLASS_P, punt.
10644 2018-02-14 Jim Wilson <jimw@sifive.com>
10646 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
10647 first SMALL_OPERAND check. New local min_second_step. Move assert
10648 to where locals are set. Add TARGET_RVC support.
10649 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
10651 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
10653 * doc/invoke.texi: Correct -Wformat-overflow code sample.
10655 2018-02-14 Martin Sebor <msebor@redhat.com>
10657 PR tree-optimization/83698
10658 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
10659 arrays constrain the offset range to their bounds.
10660 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
10661 (builtin_access::overlap): Avoid setting the size of overlap if it's
10663 (maybe_diag_overlap): Also consider arrays when deciding what values
10664 of offsets to include in diagnostics.
10666 2018-02-14 Martin Sebor <msebor@redhat.com>
10669 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
10670 that correspond to the kind of a declaration.
10672 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
10675 * config/pa/pa.md: Load address of PIC label using the linkage table
10676 if the label is nonlocal.
10678 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
10680 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
10681 warning message if user requests -maltivec=be.
10682 * doc/invoke.texi: Document deprecation of -maltivec=be.
10684 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
10687 * config/rs6000/rs6000-c.c: Update definitions for
10688 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
10689 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
10691 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
10694 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
10695 add _get_ssp intrinsics. Remove argument from
10696 __builtin_ia32_rdssp[d|q].
10697 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
10698 * config/i386/i386-builtin.def: Remove argument from
10699 __builtin_ia32_rdssp[d|q].
10700 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
10701 ix86_expand_special_args_builtin for _rdssp[d|q].
10702 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
10703 Clear register before usage.
10704 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
10705 Add documentation for new _get_ssp and _inc_ssp intrinsics.
10707 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
10709 PR tree-optimization/84357
10710 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
10711 operand 1 of an ARRAY_REF too.
10713 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
10716 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
10717 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
10719 (set_of_reg): New struct.
10720 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
10721 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
10722 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
10724 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
10725 Split into bitclr, bitset, bitinvert patterns if appropriate.
10726 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
10727 use rx_fuse_in_memory_bitop.
10728 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
10729 to named insn, correct maximum insn length.
10731 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
10734 * machmode.def: Define a complex mode for PARTIAL_INT.
10735 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
10737 * doc/rtl.texi: Document CSPImode.
10738 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
10740 (msp430_hard_regno_nregs_with_padding): Likewise.
10742 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
10745 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
10747 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
10749 PR rtl-optimization/84169
10750 * combine.c (try_combine): New variable split_i2i3. Set it to true if
10751 we generated a parallel as new i3 and we split that to new i2 and i3
10752 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
10753 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
10754 those to i2, not i1. Partially rewrite this scan code.
10756 2018-02-13 Jakub Jelinek <jakub@redhat.com>
10759 * stor-layout.c (place_field): For variable length fields, adjust
10760 offset_align afterwards not just based on the field's alignment,
10761 but also on the size.
10763 PR middle-end/84309
10764 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
10765 of exps and logs in the use_exp2 case.
10767 2018-02-13 Jeff Law <law@redhat.com>
10769 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
10770 entry for "vector".
10772 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
10775 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
10779 * common.opt (gas-loc-support, gas-locview-support): New.
10780 (ginline-points, ginternal-reset-location-views): New.
10781 * doc/invoke.texi: Document them. Use @itemx where intended.
10782 (gvariable-location-views): Adjust.
10783 * target.def (reset_location_view): New.
10784 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
10785 (TARGET_RESET_LOCATION_VIEW): New.
10786 * doc/tm.texi: Rebuilt.
10787 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
10788 (dwarf2out_default_as_locview_support): New.
10789 (output_asm_line_debug_info): Use option variables.
10790 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
10791 (output_loc_list): Likewise.
10792 (add_high_low_attributes): Check option variables.
10793 Don't output entry view attribute in strict mode.
10794 (gen_inlined_subroutine_die): Check option variables.
10795 (dwarf2out_inline_entry): Likewise.
10796 (init_sections_and_labels): Likewise.
10797 (dwarf2out_early_finish): Likewise.
10798 (maybe_reset_location_view): New, from...
10799 (dwarf2out_var_location): ... here. Call it.
10800 * debug.h (dwarf2out_default_as_loc_support): Declare.
10801 (dwarf2out_default_as_locview_support): Declare.
10802 * hooks.c (hook_int_rtx_insn_0): New.
10803 * hooks.h (hook_int_rtx_insn_0): Declare.
10804 * toplev.c (process_options): Take -gas-loc-support and
10805 -gas-locview-support from dwarf2out. Enable
10806 -gvariable-location-views by default only with locview
10807 assembler support. Enable -ginternal-reset-location-views by
10808 default only if the target defines the corresponding hook.
10809 Enable -ginline-points by default if location views are
10810 enabled; force it disabled if statement frontiers are
10812 * tree-inline.c (expand_call_inline): Check option variables.
10813 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
10815 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
10817 PR tree-optimization/84321
10818 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
10819 handling. Also check whether the anti-range contains any values
10820 that satisfy the mask; switch to a VR_RANGE if not.
10822 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
10825 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
10827 2018-02-13 Martin Jambor <mjambor@suse.cz>
10830 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
10831 of call statements, also set location of a load to a temporary.
10833 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
10835 * config/rl78/rl78.c (add_vector_labels): New function.
10836 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
10837 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
10838 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
10839 which checks that no arguments are passed.
10840 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
10841 * doc/extend.texi: Documentation for the new attribute.
10843 2018-02-13 Andreas Schwab <schwab@suse.de>
10845 * config/riscv/linux.h (CPP_SPEC): Define.
10847 2018-02-13 Jakub Jelinek <jakub@redhat.com>
10850 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
10851 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
10852 OPTION_MASK_ISA_AES as first argument to def_builtin_const
10853 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
10854 instead of OPTION_MASK_ISA_PCLMUL as first argument to
10855 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
10856 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
10857 temporarily for AES and PCLMUL builtins.
10859 PR tree-optimization/84339
10860 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
10861 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
10864 PR middle-end/84309
10865 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
10866 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
10867 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
10869 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
10871 * omp-simd-clone.h: New file.
10872 * omp-simd-clone.c: Include omp-simd-clone.h.
10873 (expand_simd_clones): No longer static.
10874 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
10875 cgraph.h and omp-simd-clone.h.
10876 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
10877 (vect_recog_widen_shift_pattern): Formatting fix.
10878 (vect_pattern_recog_1): Don't check optab for calls.
10881 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
10882 operands[2] into a REG before using gen_lowpart on it.
10884 2018-02-12 Jeff Law <law@redhat.com>
10887 * config/sh/sh.c (find_barrier): Consider a sibling call
10890 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
10891 successfully back substituting a reg.
10893 2018-02-12 Richard Biener <rguenther@suse.de>
10895 PR tree-optimization/84037
10896 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
10897 parameter, move visited init to caller.
10898 (vect_slp_analyze_operations): Separate cost from validity
10899 check, initialize visited once for all instances.
10900 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
10902 * tree-vect-stmts.c (vect_model_simple_cost): Make early
10904 (vect_model_promotion_demotion_cost): Likewise.
10905 (vectorizable_bswap): Guard cost modeling with !slp_node
10906 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
10908 (vectorizable_call): Likewise.
10909 (vectorizable_conversion): Likewise.
10910 (vectorizable_assignment): Likewise.
10911 (vectorizable_shift): Likewise.
10912 (vectorizable_operation): Likewise.
10913 (vectorizable_store): Likewise.
10914 (vectorizable_load): Likewise.
10915 (vectorizable_condition): Likewise.
10916 (vectorizable_comparison): Likewise.
10918 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
10921 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
10922 (ASAN_MARK): Fix fnspec to account for return value, change pointer
10923 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
10925 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
10927 PR middle-end/83665
10928 * params.def (inline-min-speedup): Increase from 8 to 15.
10929 (max-inline-insns-auto): Decrease from 40 to 30.
10930 * ipa-split.c (consider_split): Add some buffer for function to
10931 be considered inlining candidate.
10932 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
10935 2018-02-12 Richard Biener <rguenther@suse.de>
10937 PR tree-optimization/84037
10938 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
10939 matched stmts if we cannot swap the non-matched ones.
10941 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
10943 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
10944 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
10945 _mm_maskz_scalef_round_ss): New intrinsics.
10946 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
10947 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
10948 __builtin_ia32_scalefss_round): Remove.
10949 (__builtin_ia32_scalefsd_mask_round,
10950 __builtin_ia32_scalefss_mask_round): New intrinsics.
10951 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
10952 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
10953 ((match_operand:VF_128 2 "<round_nimm_predicate>"
10954 "<round_constraint>")): Changed to ...
10955 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
10956 "<round_scalar_constraint>")): ... this.
10957 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
10958 %0, %1, %2<round_op3>}"): Changed to ...
10959 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
10960 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
10961 %2<round_scalar_mask_op3>}"): ... this.
10962 * config/i386/subst.md (round_scalar_nimm_predicate): New.
10964 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
10966 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
10967 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
10968 (_mm_maskz_sqrt_round_ss): New intrinsics.
10969 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
10970 (__builtin_ia32_sqrtsd_mask_round)
10971 (__builtin_ia32_sqrtss_mask_round): New builtins.
10972 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
10973 (__builtin_ia32_sqrtss_round): Remove.
10974 (__builtin_ia32_sqrtsd_mask_round)
10975 (__builtin_ia32_sqrtss_mask_round): New builtins.
10976 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
10977 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
10978 ((match_operand:VF_128 1 "vector_operand"
10979 "xBm,<round_constraint>")): Changed to ...
10980 ((match_operand:VF_128 1 "vector_operand"
10981 "xBm,<round_scalar_constraint>")): ... this.
10982 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
10983 %0, %2, %<iptr>1<round_op3>}): Changed to ...
10984 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
10985 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
10986 %<iptr>1<round_scalar_mask_op3>}): ... this.
10987 ((set_attr "prefix" "<round_prefix>")): Changed to ...
10988 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
10990 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
10993 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
10994 Cast vec_cmpeq result to correct type.
10995 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
10996 Cast vec_cmpgt result to correct type.
10998 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
11000 * final.c (final_scan_insn_1): Renamed from...
11001 (final_scan_insn): ... this. New wrapper, to recover
11002 seen from the outermost call in recursive ones.
11003 * config/sparc/sparc.c (output_return): Drop seen from call.
11004 (output_sibcall): Likewise.
11005 * config/visium/visium.c (output_branch): Likewise.
11007 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
11009 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
11012 2018-02-10 Alan Modra <amodra@gmail.com>
11015 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
11016 Specify LR as an input.
11018 2018-02-10 Jakub Jelinek <jakub@redhat.com>
11021 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
11022 remove_member_access_dummy_vars): New functions.
11023 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
11024 lower_omp_1, execute_lower_omp): Use them.
11026 PR rtl-optimization/84308
11027 * shrink-wrap.c (spread_components): Release todo vector.
11029 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
11031 PR rtl-optimization/57193
11032 * ira-color.c (struct allocno_color_data): Add member
11033 conflict_allocno_hard_prefs.
11034 (update_conflict_allocno_hard_prefs): New.
11035 (bucket_allocno_compare_func): Add a preference based on
11036 conflict_allocno_hard_prefs.
11037 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
11038 (color_allocnos): Remove a dead code. Initiate
11039 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
11041 2018-02-09 Jakub Jelinek <jakub@redhat.com>
11044 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
11045 constraint from =wa to wa. Avoid a subreg on the output operand,
11046 instead use a pseudo and subreg it in a move.
11047 (p9_xxbrd_<mode>): Changed to ...
11048 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
11049 (p9_xxbrd_v2df): New expander.
11050 (p9_xxbrw_<mode>): Changed to ...
11051 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
11052 (p9_xxbrw_v4sf): New expander.
11054 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
11056 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
11058 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
11061 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
11062 multiply in 32-bit mode.
11063 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
11064 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
11067 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
11069 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
11070 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
11071 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
11072 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
11074 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
11077 * dwarf2out.c (is_trivial_indirect_ref): New function.
11078 (dwarf2out_late_global_decl): Do not generate a location
11079 attribute for variables that have a non-trivial DECL_VALUE_EXPR
11080 and that are not defined in the current unit.
11082 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
11084 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
11085 instead of a libcall for UNORDERED.
11087 2018-02-09 Tamar Christina <tamar.christina@arm.com>
11090 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
11091 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
11093 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
11096 * config/s390/s390.c (s390_set_current_function): Invoke
11097 s390_indirect_branch_settings also if fndecl didn't change.
11099 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
11101 * config/rs6000/rs6000.md (blockage): Set length to zero.
11103 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
11105 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
11107 2018-02-09 Jakub Jelinek <jakub@redhat.com>
11110 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
11111 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
11115 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
11116 PARALLEL incoming that failed vt_get_decl_and_offset check.
11118 PR middle-end/84237
11119 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
11120 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
11122 (get_variable_section): For decls in named .bss* sections pass true as
11123 second argument to bss_initializer_p.
11125 2018-02-09 Marek Polacek <polacek@redhat.com>
11126 Jakub Jelinek <jakub@redhat.com>
11129 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
11130 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
11131 Sync some changes from cxx_fold_indirect_ref.
11133 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
11135 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
11137 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
11138 (BLOCK_INLINE_ENTRY_LABEL): New.
11139 (dwarf2out_var_location): Disregard inline entry markers.
11140 (inline_entry_data): New struct.
11141 (inline_entry_data_hasher): New hashtable type.
11142 (inline_entry_data_hasher::hash): New.
11143 (inline_entry_data_hasher::equal): New.
11144 (inline_entry_data_table): New variable.
11145 (add_high_low_attributes): Add DW_AT_entry_pc and
11146 DW_AT_GNU_entry_view attributes if a pending entry is found
11147 in inline_entry_data_table. Add old entry_pc attribute only
11148 if debug nonbinding markers are disabled.
11149 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
11150 markers are enabled.
11151 (block_within_block_p, dwarf2out_inline_entry): New.
11152 (dwarf2out_finish): Check that no entries remained in
11153 inline_entry_data_table.
11154 * final.c (reemit_insn_block_notes): Handle inline entry notes.
11155 (final_scan_insn, notice_source_line): Likewise.
11156 (rest_of_clean_state): Skip inline entry markers.
11157 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
11159 * gimple.c (gimple_build_debug_inline_entry): New.
11160 * gimple.h (enum gimple_debug_subcode): Add
11161 GIMPLE_DEBUG_INLINE_ENTRY.
11162 (gimple_build_debug_inline_entry): Declare.
11163 (gimple_debug_inline_entry_p): New.
11164 (gimple_debug_nonbind_marker_p): Adjust.
11165 * insn-notes.def (INLINE_ENTRY): New.
11166 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
11167 inline entry marker notes.
11168 (print_insn): Likewise.
11169 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
11170 (INSN_DEBUG_MARKER_KIND): Likewise.
11171 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
11172 * tree-inline.c (expand_call_inline): Build and insert
11173 debug_inline_entry stmt.
11174 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
11175 inline entry blocks early, if nonbind markers are enabled.
11176 (dump_scope_block): Dump fragment info.
11177 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
11178 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
11179 (gimple_build_debug_inline_entry): New.
11180 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
11181 Enable/disable inline entry points too.
11182 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
11183 (DEBUG_INSN): Describe inline entry markers.
11185 * common.opt (gvariable-location-views): New.
11186 (gvariable-location-views=incompat5): New.
11187 * config.in: Rebuilt.
11188 * configure: Rebuilt.
11189 * configure.ac: Test assembler for view support.
11190 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
11191 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
11192 * dwarf2out.c (var_loc_view): New typedef.
11193 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
11194 (dwarf2out_locviews_in_attribute): New.
11195 (dwarf2out_locviews_in_loclist): New.
11196 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
11197 (enum dw_line_info_opcode): Add LI_adv_address.
11198 (struct dw_line_info_table): Add view.
11199 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
11200 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
11201 (zero_view_p): New variable.
11202 (ZERO_VIEW_P): New macro.
11203 (output_asm_line_debug_info): New.
11204 (struct var_loc_node): Add view.
11205 (add_AT_view_list, AT_loc_list): New.
11206 (add_var_loc_to_decl): Add view param. Test it against last.
11207 (new_loc_list): Add view params. Record them.
11208 (AT_loc_list_ptr): Handle loc and view lists.
11209 (view_list_to_loc_list_val_node): New.
11210 (print_dw_val): Handle dw_val_class_view_list.
11211 (size_of_die): Likewise.
11212 (value_format): Likewise.
11213 (loc_list_has_views): New.
11214 (gen_llsym): Set vl_symbol too.
11215 (maybe_gen_llsym, skip_loc_list_entry): New.
11216 (dwarf2out_maybe_output_loclist_view_pair): New.
11217 (output_loc_list): Output view list or entries too.
11218 (output_view_list_offset): New.
11219 (output_die): Handle dw_val_class_view_list.
11220 (output_dwarf_version): New.
11221 (output_compilation_unit_header): Use it.
11222 (output_skeleton_debug_sections): Likewise.
11223 (output_rnglists, output_line_info): Likewise.
11224 (output_pubnames, output_aranges): Update version comments.
11225 (output_one_line_info_table): Output view numbers in asm comments.
11226 (dw_loc_list): Determine current endview, pass it to new_loc_list.
11227 Call maybe_gen_llsym.
11228 (loc_list_from_tree_1): Adjust.
11229 (add_AT_location_description): Create view list attribute if
11230 needed, check it's absent otherwise.
11231 (convert_cfa_to_fb_loc_list): Adjust.
11232 (maybe_emit_file): Call output_asm_line_debug_info for test.
11233 (dwarf2out_var_location): Reset views as needed. Precompute
11234 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
11235 attribute. Set view.
11236 (new_line_info_table): Reset next view.
11237 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
11238 (dwarf2out_source_line): Likewise. Output view resets and labels to
11239 the assembler, or select appropriate line info opcodes.
11240 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
11241 (optimize_string_length): Catch it. Adjust.
11242 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
11243 dw_val_class_view_list, and remove it if no longer needed.
11244 (hash_loc_list): Hash view numbers.
11245 (loc_list_hasher::equal): Compare them.
11246 (optimize_location_lists): Check whether a view list symbol is
11247 needed, and whether the locview attribute is present, and
11248 whether they match. Remove the locview attribute if no longer
11250 (index_location_lists): Call skip_loc_list_entry for test.
11251 (dwarf2out_finish): Call output_asm_line_debug_info for test.
11252 Use output_dwarf_version.
11253 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
11254 (struct dw_val_node): Add val_view_list.
11255 * final.c (SEEN_NEXT_VIEW): New.
11256 (set_next_view_needed): New.
11257 (clear_next_view_needed): New.
11258 (maybe_output_next_view): New.
11259 (final_start_function): Rename to...
11260 (final_start_function_1): ... this. Take pointer to FIRST,
11261 add SEEN parameter. Emit param bindings in the initial view.
11262 (final_start_function): Reintroduce SEEN-less interface.
11263 (final): Rename to...
11264 (final_1): ... this. Take SEEN parameter. Output final pending
11265 next view at the end.
11266 (final): Reintroduce seen-less interface.
11267 (final_scan_insn): Output pending next view before switching
11268 sections or ending a block. Mark the next view as needed when
11269 outputting variable locations. Notify debug backend of section
11270 changes, and of location view changes.
11271 (rest_of_handle_final): Adjust.
11272 * toplev.c (process_options): Autodetect value for debug variable
11273 location views option. Warn on incompat5 without -gdwarf-5.
11274 * doc/invoke.texi (gvariable-location-views): New.
11275 (gvariable-location-views=incompat5): New.
11276 (gno-variable-location-views): New.
11278 2018-02-08 David Malcolm <dmalcolm@redhat.com>
11280 PR tree-optimization/84136
11281 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
11282 that the result of find_edge is non-NULL.
11284 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
11287 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
11288 storing integer register in SImode. Fix cost of 256 and 512
11289 byte aligned SSE register store.
11291 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
11293 * config/i386/i386.c (ix86_multiplication_cost): Fix
11294 multiplication cost for TARGET_AVX512DQ.
11296 2018-02-08 Marek Polacek <polacek@redhat.com>
11298 PR tree-optimization/84238
11299 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
11302 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
11304 PR tree-optimization/84265
11305 * tree-vect-stmts.c (vectorizable_store): Don't treat
11306 VMAT_CONTIGUOUS accesses as grouped.
11307 (vectorizable_load): Likewise.
11309 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
11311 PR tree-optimization/81635
11312 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
11313 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
11314 (test_round_for_mask): New functions.
11315 (wide_int_cc_tests): Call test_round_for_mask.
11316 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
11317 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
11318 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
11319 range returned by get_range_info.
11321 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
11324 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
11325 * symtab.c: Include builtins.h
11326 (symtab_node::output_to_lto_symbol_table_p): Move here
11327 from lto-streamer-out.c:output_symbol_p.
11328 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
11329 (output_symbol_p): Move all logic to symtab.c
11330 (produce_symtab): Update.
11332 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
11334 * config/s390/s390-opts.h (enum indirect_branch): Define.
11335 * config/s390/s390-protos.h (s390_return_addr_from_memory)
11336 (s390_indirect_branch_via_thunk)
11337 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
11338 (enum s390_indirect_branch_type): Define.
11339 * config/s390/s390.c (struct s390_frame_layout, struct
11340 machine_function): Remove.
11341 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
11342 (indirect_branch_table_label_no, indirect_branch_table_name):
11344 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
11345 (enum s390_indirect_branch_option): Define.
11346 (s390_return_addr_from_memory): New function.
11347 (s390_handle_string_attribute): New function.
11348 (s390_attribute_table): Add new attribute handler.
11349 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
11350 (s390_indirect_branch_via_thunk): New function.
11351 (s390_indirect_branch_via_inline_thunk): New function.
11352 (s390_function_ok_for_sibcall): When jumping via thunk disallow
11353 sibling call optimization for non z10 compiles.
11354 (s390_emit_call): Force indirect branch target to be a single
11355 register. Add r1 clobber for non-z10 compiles.
11356 (s390_emit_epilogue): Emit return jump via return_use expander.
11357 (s390_reorg): Handle JUMP_INSNs as execute targets.
11358 (s390_option_override_internal): Perform validity checks for the
11359 new command line options.
11360 (s390_indirect_branch_attrvalue): New function.
11361 (s390_indirect_branch_settings): New function.
11362 (s390_set_current_function): Invoke s390_indirect_branch_settings.
11363 (s390_output_indirect_thunk_function): New function.
11364 (s390_code_end): Implement target hook.
11365 (s390_case_values_threshold): Implement target hook.
11366 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
11368 * config/s390/s390.h (struct s390_frame_layout)
11369 (struct machine_function): Move here from s390.c.
11370 (TARGET_INDIRECT_BRANCH_NOBP_RET)
11371 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
11372 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
11373 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
11374 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
11375 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
11376 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
11377 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
11378 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
11379 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
11380 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
11381 (mnemonic attribute): Add values which aren't recognized
11383 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
11384 pattern for branch conversion. Fix mnemonic attribute.
11385 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
11386 indirect branch via thunk if requested.
11387 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
11388 ("*indirect_jump"): Disable for branch conversion using out of
11390 ("indirect_jump_via_thunk<mode>_z10")
11391 ("indirect_jump_via_thunk<mode>")
11392 ("indirect_jump_via_inlinethunk<mode>_z10")
11393 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
11394 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
11395 ("casesi_jump_via_inlinethunk<mode>_z10")
11396 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
11397 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
11398 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
11399 ("*indirect2_jump"): Disable for branch conversion.
11400 ("casesi_jump"): Turn into expander and expand patterns for branch
11402 ("return_use"): New expander.
11403 ("*return"): Emit return via thunk and rename it to ...
11404 ("*return<mode>"): ... this one.
11405 * config/s390/s390.opt: Add new options and and enum for the
11408 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
11410 * lra-constraints.c (match_reload): Unconditionally use
11411 gen_lowpart_SUBREG, rather than selecting between that
11412 and equivalent gen_rtx_SUBREG code.
11414 2018-02-08 Richard Biener <rguenther@suse.de>
11416 PR tree-optimization/84233
11417 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
11418 changed flag instead of boguously re-using phi_inserted.
11420 2018-02-08 Martin Jambor <mjambor@suse.cz>
11422 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
11423 static local variables.
11425 2018-02-08 Richard Biener <rguenther@suse.de>
11427 PR tree-optimization/84278
11428 * tree-vect-stmts.c (vectorizable_store): When looking for
11429 smaller vector types to perform grouped strided loads/stores
11430 make sure the mode is supported by the target.
11431 (vectorizable_load): Likewise.
11433 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
11435 * config/aarch64/aarch64.c (aarch64_components_for_bb):
11436 Increase LDP/STP opportunities by adding adjacent callee-saves.
11438 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
11440 PR rtl-optimization/84068
11441 PR rtl-optimization/83459
11442 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
11444 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
11446 PR tree-optimization/84224
11447 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
11448 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
11449 non-zero arguments.
11451 2018-02-07 Iain Sandoe <iain@codesourcery.com>
11454 * config/rs6000/altivec.md (*restore_world): Remove LR use.
11455 * config/rs6000/predicates.md (restore_world_operation): Adjust op
11456 count, remove one USE.
11458 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
11460 * doc/install.texi (Configuration): Document the
11461 --with-long-double-format={ibm,ieee} PowerPC configuration
11465 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
11466 Convert from define_expand to be define_insn_and_split. Rework
11467 float/double/_Float128 conversions to QI/HI/SImode to work with
11468 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
11469 conversions to QI/HImode types did a store and then a load to
11470 truncate the value. For conversions to VSX registers, don't split
11471 the insn, instead emit the code directly. Use the code iterator
11472 any_fix to combine signed and unsigned conversions.
11473 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
11474 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
11475 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
11476 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
11477 (fix_<mode>di2_hw): Likewise.
11478 (fixuns_<mode>di2_hw): Likewise.
11479 (fix_<mode>si2_hw): Likewise.
11480 (fixuns_<mode>si2_hw): Likewise.
11481 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
11482 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
11483 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
11484 fix<uns>_trunc<SFDF:mode>si2_p8.
11485 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
11486 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
11487 (fix<uns>_<mode>_mem): Likewise.
11488 (fctiw<u>z_<mode>_mem): Likewise.
11489 (fix<uns>_<mode>_mem): Likewise.
11490 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
11491 the register allocator from doing a direct move to the GPRs to do
11492 a store, and instead use the ISA 3.0 store byte/half-word from
11493 vector register instruction. For IEEE 128-bit floating point,
11494 also optimize stores of 32-bit ints.
11495 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
11497 2018-02-07 Alan Hayward <alan.hayward@arm.com>
11499 * genextract.c (push_pathstr_operand): New function to support
11501 (walk_rtx): Call push_pathstr_operand.
11502 (print_path): Support [a-zA-Z].
11504 2018-02-07 Richard Biener <rguenther@suse.de>
11506 PR tree-optimization/84037
11507 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
11508 (cse_and_gimplify_to_preheader): Declare.
11509 (vect_get_place_in_interleaving_chain): Likewise.
11510 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
11512 (_loop_vec_info::~_loop_vec_info): Delete it.
11513 (cse_and_gimplify_to_preheader): New function.
11514 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
11515 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
11516 (vectorizable_load): Likewise. For grouped stores always base
11517 the IV on the first element.
11518 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
11519 condition before gimplifying.
11521 2018-02-07 Jakub Jelinek <jakub@redhat.com>
11523 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
11524 *DIV_EXPR and *MOD_EXPR.
11526 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
11529 * config/i386/i386.c (ix86_option_override_internal): Mask out
11530 the CF_SET bit when checking -fcf-protection.
11532 2018-02-07 Tom de Vries <tom@codesourcery.com>
11535 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
11538 2018-02-07 Richard Biener <rguenther@suse.de>
11540 PR tree-optimization/84204
11541 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
11544 PR tree-optimization/84205
11545 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
11546 special-case isl_ast_op_zdiv_r.
11548 PR tree-optimization/84223
11549 * graphite-scop-detection.c (gather_bbs::before_dom_children):
11550 Only add conditions from within the region.
11551 (gather_bbs::after_dom_children): Adjust.
11553 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
11556 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
11557 * config/avr/avr.md: Only post-reload split REG-REG moves if
11558 either register is GENERAL_REG_P.
11560 2018-02-07 Jakub Jelinek <jakub@redhat.com>
11562 PR tree-optimization/84235
11563 * tree-ssa-scopedtables.c
11564 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
11565 if the subtraction is performed in floating point type where NaNs are
11566 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
11567 build 1. Formatting fix.
11569 2018-02-06 Jakub Jelinek <jakub@redhat.com>
11572 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
11573 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
11574 and skip it regardless of bb boundaries. Use CALL_P macro,
11575 don't test INSN_P (insn) together with CALL_P or JUMP_P check
11576 unnecessarily, formatting fix.
11578 2018-02-06 Michael Collison <michael.collison@arm.com>
11580 * config/arm/thumb2.md:
11581 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
11582 (*thumb_mov_notscc): Ditto.
11584 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
11587 * config/rs6000/rs6000.md (su code attribute): Use "u" for
11588 unsigned_fix, not "s".
11590 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11592 * configure.ac (gcc_fn_eh_frame_ro): New function.
11593 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
11594 correct .eh_frame permissions.
11595 * configure: Regenerate.
11597 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
11599 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
11600 irrelevant options.
11602 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11604 * config/rs6000/rs6000.c (rs6000_option_override_internal):
11605 Display warning message for -mno-speculate-indirect-jumps.
11607 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
11609 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
11611 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
11613 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
11615 PR tree-optimization/84225
11616 * tree-eh.c (find_trapping_overflow): Only call
11617 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
11619 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
11622 * config/i386/i386.c: Reimplement the check of possible options
11623 -mibt/-mshstk conbination. Change error messages.
11624 * doc/invoke.texi: Fix a typo: remove extra '='.
11626 2018-02-06 Marek Polacek <polacek@redhat.com>
11628 PR tree-optimization/84228
11629 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
11631 2018-02-06 Tamar Christina <tamar.christina@arm.com>
11634 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
11635 emitted arch directives.
11636 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
11637 __ARM_FEATURE_COPROC before changing architectures.
11639 2018-02-06 Richard Biener <rguenther@suse.de>
11641 * config/i386/i386.c (print_reg): Fix typo.
11642 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
11644 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
11646 * configure: Regenerate.
11648 2018-02-05 Martin Sebor <msebor@redhat.com>
11650 PR tree-optimization/83369
11651 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
11654 2018-02-05 Martin Liska <mliska@suse.cz>
11656 * doc/invoke.texi: Cherry-pick upstream r323995.
11658 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
11660 * ira.c (ira_init_register_move_cost): Adjust comment.
11662 2018-02-05 Martin Liska <mliska@suse.cz>
11664 PR gcov-profile/84137
11665 * doc/gcov.texi: Fix typo in documentation.
11667 2018-02-05 Martin Liska <mliska@suse.cz>
11669 PR gcov-profile/83879
11670 * doc/gcov.texi: Document necessity of --dynamic-list-data when
11671 using dlopen functionality.
11673 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
11675 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
11676 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
11677 _mm_maskz_range_ss, _mm_mask_range_round_ss,
11678 _mm_maskz_range_round_ss): New intrinsics.
11679 (__builtin_ia32_rangesd128_round)
11680 (__builtin_ia32_rangess128_round): Remove.
11681 (__builtin_ia32_rangesd128_mask_round,
11682 __builtin_ia32_rangess128_mask_round): New builtins.
11683 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
11684 __builtin_ia32_rangess128_round): Remove.
11685 (__builtin_ia32_rangesd128_mask_round,
11686 __builtin_ia32_rangess128_mask_round): New builtins.
11687 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
11688 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
11689 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
11690 "<round_saeonly_constraint>")): Changed to ...
11691 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
11692 "<round_saeonly_scalar_constraint>")): ... this.
11693 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
11694 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
11695 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
11696 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
11697 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
11699 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
11701 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
11703 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
11704 Remove all values except native, 8540 and 8548.
11706 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
11708 * config/i386/i386.c (ix86_output_function_return): Pass
11709 INVALID_REGNUM, instead of -1, as invalid register number to
11710 indirect_thunk_name and output_indirect_thunk.
11712 2018-02-02 Julia Koval <julia.koval@intel.com>
11714 * config.gcc: Add -march=icelake.
11715 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
11716 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
11717 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
11718 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
11719 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
11720 (processor_target_table): Add icelake.
11721 (ix86_option_override_internal): Handle new PTAs.
11722 (get_builtin_code_for_version): Handle icelake.
11723 (M_INTEL_COREI7_ICELAKE): New.
11724 (fold_builtin_cpu): Handle icelake.
11725 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
11726 * doc/invoke.texi: Add -march=icelake.
11728 2018-02-02 Julia Koval <julia.koval@intel.com>
11730 * config/i386/i386.c (ix86_option_override_internal): Change flags type
11731 to wide_int_bitmask.
11732 * wide-int-bitmask.h: New.
11734 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
11737 * config/i386/i386.md: Replace Pmode with word_mode in
11738 builtin_setjmp_setup and builtin_longjmp to support x32.
11740 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
11744 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
11746 (rs6000_supported_cpu_names): New static variable.
11747 (linux_cpu_translation_table): Likewise.
11748 (elf_platform) <cpu>: Define new static variable and use it.
11749 Translate kernel AT_PLATFORM name to canonical name if needed.
11750 Error if platform name is unknown.
11752 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
11755 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
11757 2018-02-01 Jeff Law <law@redhat.com>
11760 * config/i386/i386.c (release_scratch_register_on_entry): Add new
11761 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
11762 the scratch if RELEASE_VIA_POP is false.
11763 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
11764 If we have to save a temporary register, decrement SIZE appropriately.
11765 Pass new arguments to release_scratch_register_on_entry.
11766 (ix86_adjust_stack_and_probe): Likewise.
11767 (ix86_emit_probe_stack_range): Pass new arguments to
11768 release_scratch_register_on_entry.
11770 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
11772 PR rtl-optimization/84157
11773 * combine.c (change_zero_ext): Use REG_P predicate in
11774 front of HARD_REGISTER_P predicate.
11776 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
11778 * config/avr/avr.c (avr_option_override): Move disabling of
11779 -fdelete-null-pointer-checks to...
11780 * common/config/avr/avr-common.c (avr_option_optimization_table):
11783 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11785 PR tree-optimization/81635
11786 * tree-data-ref.c (split_constant_offset_1): For types that
11787 wrap on overflow, try to use range info to prove that wrapping
11790 2018-02-01 Renlin Li <renlin.li@arm.com>
11793 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
11794 TAILCALL_ADDR_REGS.
11795 (aarch64_register_move_cost): Likewise.
11796 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
11797 TAILCALL_ADDR_REGS.
11798 (REG_CLASS_NAMES): Likewise.
11799 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
11800 TAILCALL_ADDR_REGS. Remove IP registers.
11801 * config/aarch64/aarch64.md (Ucs): Update register constraint.
11803 2018-02-01 Richard Biener <rguenther@suse.de>
11805 * domwalk.h (dom_walker::dom_walker): Add additional constructor
11806 for specifying RPO order and allow NULL for that.
11807 * domwalk.c (dom_walker::dom_walker): Likewise.
11808 (dom_walker::walk): Handle NULL RPO order.
11809 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
11811 (rewrite_update_dom_walker): Likewise.
11812 (mark_def_dom_walker): Likewise.
11814 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11816 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
11817 (aarch64_maybe_expand_sve_subreg_move): Declare.
11818 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
11819 * config/aarch64/predicates.md (aarch64_any_register_operand): New
11821 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
11822 that are semantically a reverse operation.
11823 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
11824 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
11825 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
11827 (aarch64_can_change_mode_class): For big-endian, forbid changes
11828 between two SVE modes if they have different element sizes.
11830 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11832 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
11833 the TImode handling for big-endian targets.
11835 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11837 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
11838 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
11840 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
11841 Remove BSWAP handing for big-endian targets and use the form of
11842 LD1RQ appropariate for the mode.
11844 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11846 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
11847 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
11848 duplicated element.
11850 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11853 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
11854 check for operands that need to go through aarch64_sve_reload_be.
11856 2018-02-01 Jakub Jelinek <jakub@redhat.com>
11858 PR tree-optimization/81661
11859 PR tree-optimization/84117
11860 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
11861 * tree-eh.c: Include gimplify.h.
11862 (find_trapping_overflow, replace_trapping_overflow,
11863 rewrite_to_non_trapping_overflow): New functions.
11864 * tree-vect-loop.c: Include tree-eh.h.
11865 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
11866 * tree-data-ref.c: Include tree-eh.h.
11867 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
11869 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
11871 PR rtl-optimization/84123
11872 * combine.c (change_zero_ext): Check if hard register satisfies
11873 can_change_dest_mode before calling gen_lowpart_SUBREG.
11875 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
11878 * ira.c (ira_init_register_move_cost): Remove assert.
11880 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
11882 PR rtl-optimization/84071
11883 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
11884 * doc/tm.texi: Regenerate.
11886 2018-01-31 Richard Biener <rguenther@suse.de>
11888 PR tree-optimization/84132
11889 * tree-data-ref.c (analyze_miv_subscript): Properly
11890 check whether evolution_function_is_affine_multivariate_p
11891 before calling gcd_of_steps_may_divide_p.
11893 2018-01-31 Julia Koval <julia.koval@intel.com>
11896 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
11897 * config/i386/i386.md (rdpid_rex64) New.
11898 (rdpid): Make 32bit only.
11900 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
11903 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
11904 an IDENTIFIER_NODE for FUNCTION_TYPE's.
11906 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
11909 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
11911 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
11913 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
11915 PR rtl-optimization/84071
11916 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
11917 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
11919 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
11921 * config/arc/arc.c (arc_handle_aux_attribute): New function.
11922 (arc_attribute_table): Add 'aux' attribute.
11923 (arc_in_small_data_p): Consider aux like variables.
11924 (arc_is_aux_reg_p): New function.
11925 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
11926 (arc_get_aux_arg): New function.
11927 (prepare_move_operands): Handle aux-register access.
11928 (arc_handle_aux_attribute): New function.
11929 * doc/extend.texi (ARC Variable attributes): Add subsection.
11931 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
11933 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
11934 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
11935 (arc_attribute_table): Add 'uncached' attribute.
11936 (arc_print_operand): Print '.di' flag for uncached memory
11938 (arc_in_small_data_p): Do not consider for small data the uncached
11940 (arc_is_uncached_mem_p): New function.
11941 * config/arc/predicates.md (compact_store_memory_operand): Check
11942 for uncached memory accesses.
11943 (nonvol_nonimm_operand): Likewise.
11944 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
11946 2018-01-31 Jakub Jelinek <jakub@redhat.com>
11949 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
11950 falign-loops=): Add Optimization flag.
11952 2018-01-30 Jeff Law <law@redhat.com>
11955 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
11956 INT_REGISTERS_SAVED. Check it prior to calling
11957 get_scratch_register_on_entry.
11958 (ix86_adjust_stack_and_probe): Similarly.
11959 (ix86_emit_probe_stack_range): Similarly.
11960 (ix86_expand_prologue): Corresponding changes.
11962 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11965 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
11966 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
11968 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
11971 * lra-constraints.c (curr_insn_transform): Process AND in the
11974 2018-01-30 Jakub Jelinek <jakub@redhat.com>
11976 PR rtl-optimization/83986
11977 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
11978 dependence against last_pending_memory_flush in addition to
11979 pending_jump_insns.
11981 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
11983 PR tree-optimization/81611
11984 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
11987 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11990 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
11993 2018-01-30 Richard Biener <rguenther@suse.de>
11994 Jakub Jelinek <jakub@redhat.com>
11996 PR tree-optimization/84111
11997 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
11998 inner loops added during recursion, as they don't have up-to-date
12001 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
12004 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
12005 (can_inline_edge_by_limits_p): ... here.
12006 (can_early_inline_edge_p, check_callers,
12007 update_caller_keys, update_callee_keys, recursive_inlining,
12008 add_new_edges_to_heap, speculation_useful_p,
12009 inline_small_functions,
12010 inline_small_functions, flatten_function,
12011 inline_to_all_callers_1): Update.
12013 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
12015 * profile-count.c (profile_count::combine_with_ipa_count): Handle
12018 2018-01-30 Richard Biener <rguenther@suse.de>
12020 PR tree-optimization/83008
12021 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
12022 invariant and constant vector uses in stmts when they need
12023 more than one stmt.
12025 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12028 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
12029 * configure: Regenerate.
12031 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
12033 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
12035 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
12036 Use gen_rtx_REG rather than gen_lowpart.
12038 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
12040 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
12041 rather than 0 when creating partial subregs.
12043 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
12045 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
12048 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
12051 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
12052 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
12053 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
12054 flags. This restores the settings used before the 2017-07-24.
12055 Turning off pre increment/decrement/modify allows IVOPTS to
12056 optimize DF/SF loops where the index is an int.
12058 2018-01-29 Richard Biener <rguenther@suse.de>
12059 Kelvin Nilsen <kelvin@gcc.gnu.org>
12062 * tree-vect-stmts.c (vectorizable_call): Don't call
12063 targetm.vectorize_builtin_md_vectorized_function if callee is
12066 2018-01-22 Carl Love <cel@us.ibm.com>
12068 * doc/extend.tex: Fix typo in second arg in
12069 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
12071 2018-01-29 Richard Biener <rguenther@suse.de>
12073 PR tree-optimization/84086
12074 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
12075 (flush_ssaname_freelist): When SSA names were released reset
12076 the SCEV hash table.
12078 2018-01-29 Richard Biener <rguenther@suse.de>
12080 PR tree-optimization/84057
12081 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
12082 removed paths when removing edges.
12084 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
12086 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
12087 -mfunction-return=@var{choice}.
12089 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
12091 PR diagnostic/84034
12092 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
12093 Handle CR like TAB.
12094 (layout::print_source_line): Likewise.
12095 (test_get_line_width_without_trailing_whitespace): Add test cases.
12097 2018-01-27 Jakub Jelinek <jakub@redhat.com>
12099 PR middle-end/84040
12100 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
12103 2018-01-26 Jim Wilson <jimw@sifive.com>
12105 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
12107 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
12110 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12112 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
12113 and CMP + SUB-immediate -> SUBS.
12115 2018-01-26 Martin Sebor <msebor@redhat.com>
12117 PR tree-optimization/83896
12118 * tree-ssa-strlen.c (get_string_len): Rename...
12119 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
12120 Avoid assuming length is constant.
12121 (handle_char_store): Use HOST_WIDE_INT for string length.
12123 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
12126 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
12127 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
12129 2018-01-26 Richard Biener <rguenther@suse.de>
12131 PR rtl-optimization/84003
12132 * dse.c (record_store): Only record redundant stores when
12133 the earlier store aliases at least all accesses the later one does.
12135 2018-01-26 Jakub Jelinek <jakub@redhat.com>
12137 PR rtl-optimization/83985
12138 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
12139 REG_CFA_RESTORE insns.
12140 (delete_unmarked_insns): Don't ignore separate shrink wrapping
12141 REG_CFA_RESTORE insns here.
12144 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
12145 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
12147 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
12149 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
12150 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
12151 (arc_init): Likewise.
12152 (arc_override_options): Likewise.
12153 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
12155 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
12157 * config/arc/arc.h (TARGET_DBNZ): Define.
12158 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
12159 properly set the tune attribute.
12160 (dbnz): Use TARGET_DBNZ guard.
12161 * config/arc/arc.opt (mtune): Add core3 option.
12163 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
12165 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
12166 recognize new pic like addresses.
12167 (arc_delegitimize_address): Clean up.
12169 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
12171 * config/arc/arc-arches.def: Option mrf16 valid for all
12173 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
12174 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
12175 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
12176 * config/arc/arc-tables.opt: Regenerate.
12177 * config/arc/arc.c (arc_conditional_register_usage): Handle
12178 reduced register file case.
12179 (arc_file_start): Set must have build attributes.
12180 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
12181 mrf16 option value.
12182 * config/arc/arc.opt (mrf16): Add new option.
12183 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
12184 * config/arc/genmultilib.awk: Handle new mrf16 option.
12185 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
12186 * config/arc/t-multilib: Regenerate.
12187 * doc/invoke.texi (ARC Options): Document mrf16 option.
12189 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
12191 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
12192 * config/arc/arc.c (arc_handle_secure_attribute): New function.
12193 (arc_attribute_table): Add 'secure_call' attribute.
12194 (arc_print_operand): Print secure call operand.
12195 (arc_function_ok_for_sibcall): Don't optimize tail calls when
12197 (arc_is_secure_call_p): New function. * config/arc/arc.md
12198 (call_i): Add support for sjli instruction.
12199 (call_value_i): Likewise.
12200 * config/arc/constraints.md (Csc): New constraint.
12202 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
12203 John Eric Martin <John.Martin@emmicro-us.com>
12205 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
12206 * config/arc/arc.c (_arc_jli_section): New struct.
12207 (arc_jli_section): New type.
12208 (rc_jli_sections): New static variable.
12209 (arc_handle_jli_attribute): New function.
12210 (arc_attribute_table): Add jli_always and jli_fixed attribute.
12211 (arc_file_end): New function.
12212 (TARGET_ASM_FILE_END): Define.
12213 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
12214 (arc_add_jli_section): New function.
12215 (jli_call_scan): Likewise.
12216 (arc_reorg): Call jli_call_scan.
12217 (arc_output_addsi): Remove 'S' from printing asm operand.
12218 (arc_is_jli_call_p): New function.
12219 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
12221 (movhi_insn): Likewise.
12222 (movsi_insn): Likewise.
12223 (movsi_set_cc_insn): Likewise.
12224 (loadqi_update): Likewise.
12225 (load_zeroextendqisi_update): Likewise.
12226 (load_signextendqisi_update): Likewise.
12227 (loadhi_update): Likewise.
12228 (load_zeroextendhisi_update): Likewise.
12229 (load_signextendhisi_update): Likewise.
12230 (loadsi_update): Likewise.
12231 (loadsf_update): Likewise.
12232 (movsicc_insn): Likewise.
12233 (bset_insn): Likewise.
12234 (bxor_insn): Likewise.
12235 (bclr_insn): Likewise.
12236 (bmsk_insn): Likewise.
12237 (bicsi3_insn): Likewise.
12238 (cmpsi_cc_c_insn): Likewise.
12239 (movsi_ne): Likewise.
12240 (movsi_cond_exec): Likewise.
12241 (clrsbsi2): Likewise.
12242 (norm_f): Likewise.
12252 (call_i): Remove 'S' asm letter, add jli instruction.
12253 (call_value_i): Likewise.
12254 * config/arc/arc.op (mjli-always): New option.
12255 * config/arc/constraints.md (Cji): New constraint.
12256 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
12258 (subsf3_fpx): Likewise.
12259 (mulsf3_fpx): Likewise.
12260 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
12262 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
12263 function attrbutes.
12264 * doc/invoke.texi (ARC): Document mjli-always option.
12266 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
12268 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
12269 avoid addition with 0 and use incw and decw where possible.
12271 2018-01-26 Richard Biener <rguenther@suse.de>
12273 PR tree-optimization/81082
12274 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
12275 association if it requires casting to unsigned.
12276 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
12277 from fold_plusminus_mult_expr to catch important cases late when
12278 range info is available.
12280 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12282 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
12283 * configure.ac (hidden_linkonce): New test.
12284 * configure: Regenerate.
12285 * config.in: Regenerate.
12287 2018-01-26 Julia Koval <julia.koval@intel.com>
12289 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
12290 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
12291 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
12292 _mm_mask_bitshuffle_epi64_mask): Fix type.
12293 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
12294 USI_FTYPE_V4DI_V4DI_USI): Remove.
12295 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
12296 __builtin_ia32_vpshufbitqmb256_mask,
12297 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
12298 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
12299 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
12301 2018-01-26 Alan Modra <amodra@gmail.com>
12304 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
12305 UNSPEC_VBPERMQ. Sort other unspecs.
12307 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
12309 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
12311 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
12313 PR middle-end/83055
12314 * predict.c (drop_profile): Do not push/pop cfun; update also
12316 (handle_missing_profiles): Fix logic looking for zero profiles.
12318 2018-01-25 Jakub Jelinek <jakub@redhat.com>
12320 PR middle-end/83977
12321 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
12322 on functions with #pragma omp declare simd or functions with simd
12324 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
12325 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
12326 Remove trailing \n from warning_at calls.
12328 2018-01-25 Tom de Vries <tom@codesourcery.com>
12331 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
12332 for neutered workers.
12334 2018-01-24 Joseph Myers <joseph@codesourcery.com>
12337 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
12338 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
12340 2018-01-24 Jeff Law <law@redhat.com>
12343 * i386.c (get_probe_interval): Move to earlier point.
12344 (ix86_compute_frame_layout): If -fstack-clash-protection and
12345 the frame is larger than the probe interval, then use pushes
12346 to save registers rather than reg->mem moves.
12347 (ix86_expand_prologue): Remove conditional for int_registers_saved
12350 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
12353 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
12354 min/max for never referenced object.
12356 2018-01-24 Jakub Jelinek <jakub@redhat.com>
12358 PR middle-end/83977
12359 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
12361 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
12362 attributes from DECL_ATTRIBUTES (decl) without affecting
12363 DECL_ATTRIBUTES (current_function_decl).
12364 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
12365 functions with non-NULL DECL_ABSTRACT_ORIGIN.
12367 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
12369 PR tree-optimization/83979
12370 * fold-const.c (fold_comparison): Use constant_boolean_node
12371 instead of boolean_{true,false}_node.
12373 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
12375 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
12378 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12380 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
12381 Simplify the clause that sets the length attribute.
12382 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
12383 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
12384 clause that sets the length attribute.
12385 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
12387 2018-01-24 Tom de Vries <tom@codesourcery.com>
12390 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
12391 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
12392 Add strict parameter.
12393 (prevent_branch_around_nothing): Insert dummy insn between branch to
12394 label and label with no ptx insn inbetween.
12395 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
12397 2018-01-24 Tom de Vries <tom@codesourcery.com>
12400 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
12401 for neutered threads in warp.
12402 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
12404 2018-01-24 Richard Biener <rguenther@suse.de>
12406 PR tree-optimization/83176
12407 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
12410 2018-01-24 Richard Biener <rguenther@suse.de>
12412 PR tree-optimization/82819
12413 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
12414 code generating pluses that are no-ops in the target precision.
12416 2018-01-24 Richard Biener <rguenther@suse.de>
12418 PR middle-end/84000
12419 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
12421 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
12423 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
12424 to merge probabilities.
12425 * predict.c (probably_never_executed): Also mark as cold functions
12426 with global 0 profile and guessed local profile.
12427 * profile-count.c (profile_probability::combine_with_count): New
12429 * profile-count.h (profile_probability::operator*,
12430 profile_probability::operator*=, profile_probability::operator/,
12431 profile_probability::operator/=): Reduce precision to adjusted
12432 and set value to guessed on contradictory divisions.
12433 (profile_probability::combine_with_freq): Remove.
12434 (profile_probability::combine_wiht_count): Declare.
12435 (profile_count::force_nonzero):: Set to adjusted.
12436 (profile_count::probability_in):: Set quality to adjusted.
12437 * tree-ssa-tail-merge.c (replace_block_by): Use
12438 combine_with_count.
12440 2018-01-23 Andrew Waterman <andrew@sifive.com>
12441 Jim Wilson <jimw@sifive.com>
12443 * config/riscv/riscv.c (riscv_stack_boundary): New.
12444 (riscv_option_override): Set riscv_stack_boundary. Handle
12445 riscv_preferred_stack_boundary_arg.
12446 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
12447 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
12448 (STACK_BOUNDARY): Set to riscv_stack_boundary.
12449 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
12450 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
12451 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
12453 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
12456 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
12457 of struct ix86_frame.
12458 (ix86_expand_epilogue): Likewise. Add a local variable for
12459 the reg_save_offset field in struct ix86_frame.
12461 2018-01-23 Bin Cheng <bin.cheng@arm.com>
12463 PR tree-optimization/82604
12464 * tree-loop-distribution.c (enum partition_kind): New enum item
12465 PKIND_PARTIAL_MEMSET.
12466 (partition_builtin_p): Support above new enum item.
12467 (generate_code_for_partition): Ditto.
12468 (compute_access_range): Differentiate cases that equality can be
12469 proven at all loops, the innermost loops or no loops.
12470 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
12471 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
12472 (finalize_partitions, distribute_loop): Don't fuse partition of
12473 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
12474 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
12475 parloop is enabled.
12477 2018-01-23 Martin Liska <mliska@suse.cz>
12479 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
12480 order to ignore the predictor.
12481 (PRED_POLYMORPHIC_CALL): Likewise.
12482 (PRED_RECURSIVE_CALL): Likewise.
12484 2018-01-23 Martin Liska <mliska@suse.cz>
12486 * tree-profile.c (tree_profiling): Print function header to
12487 aware reader which function we are working on.
12488 * value-prof.c (gimple_find_values_to_profile): Do not print
12489 not interesting value histograms.
12491 2018-01-23 Martin Liska <mliska@suse.cz>
12493 * profile-count.h (enum profile_quality): Add
12494 profile_uninitialized as the first value. Do not number values
12495 as they are zero based.
12496 (profile_count::verify): Update sanity check.
12497 (profile_probability::verify): Likewise.
12499 2018-01-23 Nathan Sidwell <nathan@acm.org>
12501 * doc/invoke.texi (ffor-scope): Deprecate.
12503 2018-01-23 David Malcolm <dmalcolm@redhat.com>
12505 PR tree-optimization/83510
12506 * domwalk.c (set_all_edges_as_executable): New function.
12507 (dom_walker::dom_walker): Convert bool param
12508 "skip_unreachable_blocks" to enum reachability. Move setup of
12509 edge flags to set_all_edges_as_executable and only do it when
12510 reachability is REACHABLE_BLOCKS.
12511 * domwalk.h (enum dom_walker::reachability): New enum.
12512 (dom_walker::dom_walker): Convert bool param
12513 "skip_unreachable_blocks" to enum reachability.
12514 (set_all_edges_as_executable): New decl.
12515 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
12516 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
12518 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
12519 but converting true to REACHABLE_BLOCKS.
12520 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
12522 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
12523 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
12524 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
12526 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
12527 if check_all_array_refs will be called.
12529 2018-01-23 David Malcolm <dmalcolm@redhat.com>
12531 * tree.c (selftest::test_location_wrappers): Add more test
12534 2018-01-23 David Malcolm <dmalcolm@redhat.com>
12536 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
12537 (selftest::test_bit_in_range): Likewise.
12539 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
12542 * doc/sourcebuild.texi (vect_float): Say that the selector
12543 only describes the situation when -funsafe-math-optimizations is on.
12544 (vect_float_strict): Document.
12546 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
12548 PR tree-optimization/83965
12549 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
12550 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
12551 instead of checking only for a reduction.
12552 (vect_recog_widen_sum_pattern): Likewise.
12554 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
12556 * predict.c (probably_never_executed): Only use precise profile info.
12557 (compute_function_frequency): Skip after inlining hack since we now
12558 have quality checking.
12560 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
12562 * profile-count.h (profile_probability::very_unlikely,
12563 profile_probability::unlikely, profile_probability::even): Set
12564 precision to guessed.
12566 2018-01-23 Richard Biener <rguenther@suse.de>
12568 PR tree-optimization/83963
12569 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
12570 Properly terminate dominator walk when crossing the exit edge not
12571 when visiting its source block.
12573 2018-01-23 Jakub Jelinek <jakub@redhat.com>
12576 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
12577 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
12579 2018-01-22 Jakub Jelinek <jakub@redhat.com>
12581 PR tree-optimization/83957
12582 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
12583 semicolon after for body surrounded by braces.
12585 PR tree-optimization/83081
12586 * profile-count.h (profile_probability::split): New method.
12587 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
12588 Use profile_probability::split.
12589 (do_compare_rtx_and_jump): Fix adjustment of probabilities
12590 when splitting a single conditional jump into 2.
12592 2018-01-22 David Malcolm <dmalcolm@redhat.com>
12594 PR tree-optimization/69452
12595 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
12598 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12600 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
12601 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
12602 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
12604 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12606 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
12608 * config/rl78/rl78.md (movdi): New define_expand.
12609 * config/rl78/rl78.c (rl78_split_movdi): New function.
12611 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
12614 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
12616 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
12617 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
12618 128-bit to produce an UNSPEC move to get the double word with the
12619 signbit and then a shift directly to do signbit.
12620 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
12621 implementation with a new version that just does either a direct
12622 move or a regular move. Move memory interface to separate insns.
12623 Move insns so they are next to the expander.
12624 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
12625 with signbit move. Split big and little endian case.
12626 (signbit<mode>2_dm_mem_le): Likewise.
12627 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
12628 (signbit<mode>2_dm2): Likewise.
12630 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12632 * config/rl78/rl78.md (anddi3): New define_expand.
12634 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12636 * config/rl78/rl78.md (umindi3): New define_expand.
12638 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12640 * config/rl78/rl78.md (smindi3): New define_expand.
12642 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12644 * config/rl78/rl78.md (smaxdi3): New define_expand.
12646 2018-01-22 Carl Love <cel@us.ibm.com>
12648 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
12649 LVX_V1TI): Add macro expansion.
12650 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
12651 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
12652 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
12653 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
12654 Change check to determine if the instruction is a byte reversing
12655 entry. Fix typo in comment.
12656 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
12657 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
12658 Add def_builtin calls for new builtins.
12659 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
12660 Add define_insn expansion.
12662 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12664 * config/rl78/rl78.md (umaxdi3): New define_expand.
12666 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12668 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
12669 for non-QImode registers.
12671 2018-01-22 Richard Biener <rguenther@suse.de>
12673 PR tree-optimization/83963
12674 * graphite-scop-detection.c (scop_detection::get_sese): Delay
12675 including the loop exit block.
12676 (scop_detection::merge_sese): Likewise.
12677 (scop_detection::add_scop): Do it here instead.
12679 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12681 * doc/sourcebuild.texi (arm_softfloat): Document.
12683 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
12686 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
12687 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
12688 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
12690 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12691 David Edelsohn <dje.gcc@gmail.com>
12694 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
12695 Change "crset eq" to "crset 2".
12696 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
12697 (*call_indirect_aix<mode>_nospec): Likewise.
12698 (*call_value_indirect_aix<mode>_nospec): Likewise.
12699 (*call_indirect_elfv2<mode>_nospec): Likewise.
12700 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
12701 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
12702 change assembly output from . to $.
12703 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
12704 (indirect_jump<mode>_nospec): Change assembly output from . to $.
12705 (*tablejump<mode>_internal1_nospec): Likewise.
12707 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
12710 * config/sh/sh_optimize_sett_clrt.cc:
12711 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
12713 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
12715 PR tree-optimization/83940
12716 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
12717 offset_dt to vect_constant_def rather than vect_unknown_def_type.
12718 (vect_check_load_store_mask): Add a mask_dt_out parameter and
12719 use it to pass back the definition type.
12720 (vect_check_store_rhs): Likewise rhs_dt_out.
12721 (vect_build_gather_load_calls): Add a mask_dt argument and use
12722 it instead of a call to vect_is_simple_use.
12723 (vectorizable_store): Update calls to vect_check_load_store_mask
12724 and vect_check_store_rhs. Use the dt returned by the latter instead
12725 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
12726 instead of calls to vect_is_simple_use. Pass the scalar rather
12727 than the vector operand to vect_is_simple_use when handling
12728 second and subsequent copies of an rhs value.
12729 (vectorizable_load): Update calls to vect_check_load_store_mask
12730 and vect_build_gather_load_calls. Use the cached mask_dt and
12731 gs_info.offset_dt instead of calls to vect_is_simple_use.
12733 2018-01-20 Jakub Jelinek <jakub@redhat.com>
12735 PR middle-end/83945
12736 * tree-emutls.c: Include gimplify.h.
12737 (lower_emutls_2): New function.
12738 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
12739 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
12740 it before further processing.
12743 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
12744 UINTVAL (trueop1) instead of INTVAL (op1).
12746 2018-01-19 Jakub Jelinek <jakub@redhat.com>
12750 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
12751 INCOMING_FRAME_SP_OFFSET if not defined.
12752 (scan_trace): Add ENTRY argument. If true and
12753 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
12754 emit a note to adjust the CFA offset.
12755 (create_cfi_notes): Adjust scan_trace callers.
12756 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
12757 INCOMING_FRAME_SP_OFFSET in the CIE.
12758 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
12759 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
12761 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
12762 * doc/tm.texi: Regenerated.
12764 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
12766 PR rtl-optimization/83147
12767 * lra-constraints.c (remove_inheritance_pseudos): Use
12768 lra_substitute_pseudo_within_insn.
12770 2018-01-19 Tom de Vries <tom@codesourcery.com>
12771 Cesar Philippidis <cesar@codesourcery.com>
12774 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
12776 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
12779 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
12780 spaces for function labels.
12782 2018-01-19 Martin Liska <mliska@suse.cz>
12784 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
12785 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
12786 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
12787 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
12788 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
12789 (PRED_CONST_RETURN): Change from 69 to 65.
12790 (PRED_NULL_RETURN): Change from 91 to 71.
12791 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
12792 (PRED_LOOP_GUARD): Change from 66 to 73.
12794 2018-01-19 Martin Liska <mliska@suse.cz>
12796 * predict.c (predict_insn_def): Add new assert.
12797 (struct branch_predictor): Change type to signed integer.
12798 (test_prediction_value_range): Amend test to cover
12799 PROB_UNINITIALIZED.
12800 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
12801 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
12802 (PRED_LOOP_ITERATIONS_MAX): Likewise.
12803 (PRED_LOOP_IV_COMPARE): Likewise.
12804 * predict.h (PROB_UNINITIALIZED): Define new constant.
12806 2018-01-19 Martin Liska <mliska@suse.cz>
12808 * predict.c (dump_prediction): Add new format for
12809 analyze_brprob.py script which is enabled with -details
12811 * profile-count.h (precise_p): New function.
12813 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
12815 PR tree-optimization/83922
12816 * tree-vect-loop.c (vect_verify_full_masking): Return false if
12817 there are no statements that need masking.
12818 (vect_active_double_reduction_p): New function.
12819 (vect_analyze_loop_operations): Use it when handling phis that
12820 are not in the loop header.
12822 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
12824 PR tree-optimization/83914
12825 * tree-vect-loop.c (vectorizable_induction): Don't convert
12826 init_expr or apply the peeling adjustment for inductions
12827 that are nested within the vectorized loop.
12829 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12831 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
12834 2018-01-18 Jakub Jelinek <jakub@redhat.com>
12838 * function.h (gimplify_parameters): Add gimple_seq * argument.
12839 * function.c: Include gimple.h and options.h.
12840 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
12841 for the added local temporaries if needed.
12842 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
12843 if there are any parameter cleanups, wrap whole body into a
12844 try/finally with the cleanups.
12846 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
12849 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
12850 Use GET_MODE_CLASS for scalar floating point.
12852 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
12856 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
12857 Fix call of call_cgraph_insertion_hooks.
12859 2018-01-18 Martin Sebor <msebor@redhat.com>
12861 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
12863 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
12866 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
12869 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
12872 * common.opt: (-ffile-prefix-map): New option.
12873 * opts.c (common_handle_option): Defer it.
12874 * opts-global.c (handle_common_deferred_options): Handle it.
12875 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
12876 * file-prefix-map.h: New file.
12877 (remap_debug_filename, add_debug_prefix_map): ...here.
12878 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
12879 * final.c (debug_prefix_map, add_debug_prefix_map
12880 remap_debug_filename): Move to...
12881 * file-prefix-map.c: New file.
12882 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
12883 generalize, get rid of alloca(), use strrchr() instead of strchr().
12884 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
12885 Implement in terms of add_prefix_map().
12886 (remap_macro_filename, remap_debug_filename): Implement in term of
12888 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
12889 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
12890 * dbxout.c: Include file-prefix-map.h.
12891 * varasm.c: Likewise.
12892 * vmsdbgout.c: Likewise.
12893 * xcoffout.c: Likewise.
12894 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
12895 * doc/cppopts.texi (-fmacro-prefix-map): Document.
12896 * doc/invoke.texi (-ffile-prefix-map): Document.
12897 (-fdebug-prefix-map): Update description.
12899 2018-01-18 Martin Liska <mliska@suse.cz>
12901 * config/i386/i386.c (indirect_thunk_name): Document that also
12903 (output_indirect_thunk): Document why both instructions
12904 (pause and lfence) are generated.
12906 2018-01-18 Richard Biener <rguenther@suse.de>
12908 PR tree-optimization/83887
12909 * graphite-scop-detection.c
12910 (scop_detection::get_nearest_dom_with_single_entry): Remove.
12911 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
12912 (scop_detection::merge_sese): Re-implement with a flood-fill
12913 algorithm that properly finds a SESE region if it exists.
12915 2018-01-18 Jakub Jelinek <jakub@redhat.com>
12918 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
12919 pointer_diff optimizations use view_convert instead of convert.
12921 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12923 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
12924 Generate different code for -mno-speculate-indirect-jumps.
12925 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
12926 (*call_indirect_aix<mode>): Disable for
12927 -mno-speculate-indirect-jumps.
12928 (*call_indirect_aix<mode>_nospec): New define_insn.
12929 (*call_value_indirect_aix<mode>): Disable for
12930 -mno-speculate-indirect-jumps.
12931 (*call_value_indirect_aix<mode>_nospec): New define_insn.
12932 (*sibcall_nonlocal_sysv<mode>): Generate different code for
12933 -mno-speculate-indirect-jumps.
12934 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
12936 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
12938 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
12939 long double type, set the flags for noting the default long double
12940 type, even if we don't pass or return a long double type.
12942 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
12945 * ipa-inline.c (flatten_function): Do not overwrite final inlining
12948 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
12950 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
12951 support for merge[hl].
12952 (fold_mergehl_helper): New helper function.
12953 (tree-vector-builder.h): New #include for tree_vector_builder usage.
12954 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
12955 (altivec_vmrglw_direct): Add xxmrglw insn.
12957 2018-01-17 Andrew Waterman <andrew@sifive.com>
12959 * config/riscv/riscv.c (riscv_conditional_register_usage): If
12960 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
12962 2018-01-17 David Malcolm <dmalcolm@redhat.com>
12965 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
12966 call the lto_location_cache before reading the
12967 DECL_SOURCE_LOCATION of the types.
12969 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
12970 Richard Sandiford <richard.sandiford@linaro.org>
12972 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
12973 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
12974 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
12975 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
12976 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
12978 * config/aarch64/constraints.md (aarch64_movti_operand):
12980 * config/aarch64/predicates.md (Uti): Add new constraint.
12982 2018-01-17 Carl Love <cel@us.ibm.com>
12984 * config/rs6000/vsx.md (define_expand xl_len_r,
12985 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
12986 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
12988 (define_expand, define_insn): Move the shift left from the
12989 define_insn to the define_expand for lxvl and stxvl instructions.
12990 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
12991 and XL_LEN_R definitions to PURE.
12993 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
12995 * config/i386/i386.c (indirect_thunk_name): Declare regno
12996 as unsigned int. Compare regno with INVALID_REGNUM.
12997 (output_indirect_thunk): Ditto.
12998 (output_indirect_thunk_function): Ditto.
12999 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
13000 in the call to output_indirect_thunk_function.
13002 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
13004 PR middle-end/83884
13005 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
13006 rather than the size of inner_type to determine the stack slot size
13007 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
13009 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
13012 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
13015 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
13017 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
13018 endian Linux systems to optionally enable multilibs for selecting
13019 the long double type if the user configured an explicit type.
13020 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
13021 have no long double multilibs if not defined.
13022 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
13023 warn if the user used -mabi={ieee,ibm}longdouble and we built
13024 multilibs for long double.
13025 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
13026 appropriate multilib option.
13027 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
13029 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
13030 for building long double multilibs.
13031 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
13033 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
13035 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
13038 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
13040 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
13043 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
13046 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
13049 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
13051 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
13052 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
13054 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
13056 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
13057 different rtl trees depending on TARGET_64BIT.
13058 (rs6000_gen_lvx): Likewise.
13060 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
13062 * config/visium/visium.md (nop): Tweak comment.
13063 (hazard_nop): Likewise.
13065 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13067 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
13068 -mspeculate-indirect-jumps.
13069 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
13070 for -mno-speculate-indirect-jumps.
13071 (*call_indirect_elfv2<mode>_nospec): New define_insn.
13072 (*call_value_indirect_elfv2<mode>): Disable for
13073 -mno-speculate-indirect-jumps.
13074 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
13075 (indirect_jump): Emit different RTL for
13076 -mno-speculate-indirect-jumps.
13077 (*indirect_jump<mode>): Disable for
13078 -mno-speculate-indirect-jumps.
13079 (*indirect_jump<mode>_nospec): New define_insn.
13080 (tablejump): Emit different RTL for
13081 -mno-speculate-indirect-jumps.
13082 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
13083 (tablejumpsi_nospec): New define_expand.
13084 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
13085 (tablejumpdi_nospec): New define_expand.
13086 (*tablejump<mode>_internal1): Disable for
13087 -mno-speculate-indirect-jumps.
13088 (*tablejump<mode>_internal1_nospec): New define_insn.
13089 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
13092 2018-01-16 Artyom Skrobov tyomitch@gmail.com
13094 * caller-save.c (insert_save): Drop unnecessary parameter. All
13097 2018-01-16 Jakub Jelinek <jakub@redhat.com>
13098 Richard Biener <rguenth@suse.de>
13101 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
13102 return early, inline manually is_gimple_sizepos. Make sure if we
13103 call gimplify_expr we don't end up with a gimple constant.
13104 * tree.c (variably_modified_type_p): Don't return true for
13105 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
13106 * gimplify.h (is_gimple_sizepos): Remove.
13108 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
13110 PR tree-optimization/83857
13111 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
13112 vectorizable_live_operation for pure SLP statements.
13113 (vectorizable_live_operation): Handle PHIs.
13115 2018-01-16 Richard Biener <rguenther@suse.de>
13117 PR tree-optimization/83867
13118 * tree-vect-stmts.c (vect_transform_stmt): Precompute
13119 nested_in_vect_loop_p since the scalar stmt may get invalidated.
13121 2018-01-16 Jakub Jelinek <jakub@redhat.com>
13124 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
13125 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
13126 If off is not INTEGER_CST, issue a may not be aligned warning
13127 rather than isn't aligned. Use isn%'t rather than isn't.
13128 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
13130 <case MULT_EXPR>: Improve the case when bottom and one of the
13131 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
13132 operand, in that case check if the other operand is multiple of
13133 bottom divided by the INTEGER_CST operand.
13135 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
13138 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
13139 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
13140 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
13141 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
13142 * config/pa/pa.c (pa_function_arg_advance): Likewise.
13143 (pa_function_arg, pa_arg_partial_bytes): Likewise.
13144 (pa_function_arg_size): New function.
13146 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
13148 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
13149 in a separate statement.
13151 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
13153 PR tree-optimization/83847
13154 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
13155 group gathers and scatters.
13157 2018-01-16 Jakub Jelinek <jakub@redhat.com>
13159 PR rtl-optimization/86620
13160 * params.def (max-sched-ready-insns): Bump minimum value to 1.
13162 PR rtl-optimization/83213
13163 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
13164 to last if both are JUMP_INSNs.
13166 PR tree-optimization/83843
13167 * gimple-ssa-store-merging.c
13168 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
13169 store_immediate_info for bswap/nop orig_stores.
13171 2018-01-15 Andrew Waterman <andrew@sifive.com>
13173 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
13175 <UDIV>: Increase cost if !TARGET_DIV.
13177 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
13179 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
13180 (define_attr "cr_logical_3op"): New.
13181 (cceq_ior_compare): Adjust.
13182 (cceq_ior_compare_complement): Adjust.
13183 (*cceq_rev_compare): Adjust.
13184 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
13185 (is_cracked_insn): Adjust.
13186 (insn_must_be_first_in_group): Adjust.
13187 * config/rs6000/40x.md: Adjust.
13188 * config/rs6000/440.md: Adjust.
13189 * config/rs6000/476.md: Adjust.
13190 * config/rs6000/601.md: Adjust.
13191 * config/rs6000/603.md: Adjust.
13192 * config/rs6000/6xx.md: Adjust.
13193 * config/rs6000/7450.md: Adjust.
13194 * config/rs6000/7xx.md: Adjust.
13195 * config/rs6000/8540.md: Adjust.
13196 * config/rs6000/cell.md: Adjust.
13197 * config/rs6000/e300c2c3.md: Adjust.
13198 * config/rs6000/e500mc.md: Adjust.
13199 * config/rs6000/e500mc64.md: Adjust.
13200 * config/rs6000/e5500.md: Adjust.
13201 * config/rs6000/e6500.md: Adjust.
13202 * config/rs6000/mpc.md: Adjust.
13203 * config/rs6000/power4.md: Adjust.
13204 * config/rs6000/power5.md: Adjust.
13205 * config/rs6000/power6.md: Adjust.
13206 * config/rs6000/power7.md: Adjust.
13207 * config/rs6000/power8.md: Adjust.
13208 * config/rs6000/power9.md: Adjust.
13209 * config/rs6000/rs64.md: Adjust.
13210 * config/rs6000/titan.md: Adjust.
13212 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
13214 * config/i386/predicates.md (indirect_branch_operand): Rewrite
13215 ix86_indirect_branch_register logic.
13217 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
13219 * config/i386/constraints.md (Bs): Update
13220 ix86_indirect_branch_register check. Don't check
13221 ix86_indirect_branch_register with GOT_memory_operand.
13223 * config/i386/predicates.md (GOT_memory_operand): Don't check
13224 ix86_indirect_branch_register here.
13225 (GOT32_symbol_operand): Likewise.
13227 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
13229 * config/i386/predicates.md (constant_call_address_operand):
13230 Rewrite ix86_indirect_branch_register logic.
13231 (sibcall_insn_operand): Likewise.
13233 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
13235 * config/i386/constraints.md (Bs): Replace
13236 ix86_indirect_branch_thunk_register with
13237 ix86_indirect_branch_register.
13239 * config/i386/i386.md (indirect_jump): Likewise.
13240 (tablejump): Likewise.
13241 (*sibcall_memory): Likewise.
13242 (*sibcall_value_memory): Likewise.
13243 Peepholes of indirect call and jump via memory: Likewise.
13244 * config/i386/i386.opt: Likewise.
13245 * config/i386/predicates.md (indirect_branch_operand): Likewise.
13246 (GOT_memory_operand): Likewise.
13247 (call_insn_operand): Likewise.
13248 (sibcall_insn_operand): Likewise.
13249 (GOT32_symbol_operand): Likewise.
13251 2018-01-15 Jakub Jelinek <jakub@redhat.com>
13253 PR middle-end/83837
13254 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
13255 type rather than type addr's type points to.
13256 (expand_omp_atomic_mutex): Likewise.
13257 (expand_omp_atomic): Likewise.
13259 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
13262 * config/i386/i386.c (output_indirect_thunk_function): Use
13263 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
13264 for __x86_return_thunk.
13266 2018-01-15 Richard Biener <rguenther@suse.de>
13268 PR middle-end/83850
13269 * expmed.c (extract_bit_field_1): Fix typo.
13271 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13274 * config/arm/iterators.md (VF): New mode iterator.
13275 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
13276 Remove integer-related logic from pattern.
13277 (neon_vabd<mode>_3): Likewise.
13279 2018-01-15 Jakub Jelinek <jakub@redhat.com>
13281 PR middle-end/82694
13282 * common.opt (fstrict-overflow): No longer an alias.
13283 (fwrapv-pointer): New option.
13284 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
13285 also for pointer types based on flag_wrapv_pointer.
13286 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
13287 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
13288 opts->x_flag_wrapv got set.
13289 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
13290 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
13291 POINTER_TYPE_OVERFLOW_UNDEFINED.
13292 * match.pd: Likewise in address comparison pattern.
13293 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
13295 2018-01-15 Richard Biener <rguenther@suse.de>
13298 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
13299 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
13300 Reset type names to their identifier if their TYPE_DECL doesn't
13301 have linkage (and thus is used for ODR and devirt).
13302 (save_debug_info_for_decl): Remove.
13303 (save_debug_info_for_type): Likewise.
13304 (add_tree_to_fld_list): Adjust.
13305 * tree-pretty-print.c (dump_generic_node): Make dumping of
13306 type names more robust.
13308 2018-01-15 Richard Biener <rguenther@suse.de>
13310 * BASE-VER: Bump to 8.0.1.
13312 2018-01-14 Martin Sebor <msebor@redhat.com>
13315 * builtins.c (check_access): Avoid warning when the no-warning bit
13318 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
13320 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
13321 * ira-color (allocno_hard_regs_compare): Likewise.
13323 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
13326 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
13327 Use .pushsection/.popsection.
13329 2018-01-14 Martin Sebor <msebor@redhat.com>
13332 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
13334 2018-01-14 Jakub Jelinek <jakub@redhat.com>
13336 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
13337 entry from extra_headers.
13338 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
13339 extra_headers, make the list bitwise identical to the i?86-*-* one.
13341 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
13343 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
13344 -mcmodel=large with -mindirect-branch=thunk,
13345 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
13346 -mfunction-return=thunk-extern.
13347 * doc/invoke.texi: Document -mcmodel=large is incompatible with
13348 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
13349 -mfunction-return=thunk and -mfunction-return=thunk-extern.
13351 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
13353 * config/i386/i386.c (print_reg): Print the name of the full
13354 integer register without '%'.
13355 (ix86_print_operand): Handle 'V'.
13356 * doc/extend.texi: Document 'V' modifier.
13358 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
13360 * config/i386/constraints.md (Bs): Disallow memory operand for
13361 -mindirect-branch-register.
13363 * config/i386/predicates.md (indirect_branch_operand): Likewise.
13364 (GOT_memory_operand): Likewise.
13365 (call_insn_operand): Likewise.
13366 (sibcall_insn_operand): Likewise.
13367 (GOT32_symbol_operand): Likewise.
13368 * config/i386/i386.md (indirect_jump): Call convert_memory_address
13369 for -mindirect-branch-register.
13370 (tablejump): Likewise.
13371 (*sibcall_memory): Likewise.
13372 (*sibcall_value_memory): Likewise.
13373 Disallow peepholes of indirect call and jump via memory for
13374 -mindirect-branch-register.
13375 (*call_pop): Replace m with Bw.
13376 (*call_value_pop): Likewise.
13377 (*sibcall_pop_memory): Replace m with Bs.
13378 * config/i386/i386.opt (mindirect-branch-register): New option.
13379 * doc/invoke.texi: Document -mindirect-branch-register option.
13381 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
13383 * config/i386/i386-protos.h (ix86_output_function_return): New.
13384 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
13385 set function_return_type.
13386 (indirect_thunk_name): Add ret_p to indicate thunk for function
13388 (output_indirect_thunk_function): Pass false to
13389 indirect_thunk_name.
13390 (ix86_output_indirect_branch_via_reg): Likewise.
13391 (ix86_output_indirect_branch_via_push): Likewise.
13392 (output_indirect_thunk_function): Create alias for function
13393 return thunk if regno < 0.
13394 (ix86_output_function_return): New function.
13395 (ix86_handle_fndecl_attribute): Handle function_return.
13396 (ix86_attribute_table): Add function_return.
13397 * config/i386/i386.h (machine_function): Add
13398 function_return_type.
13399 * config/i386/i386.md (simple_return_internal): Use
13400 ix86_output_function_return.
13401 (simple_return_internal_long): Likewise.
13402 * config/i386/i386.opt (mfunction-return=): New option.
13403 (indirect_branch): Mention -mfunction-return=.
13404 * doc/extend.texi: Document function_return function attribute.
13405 * doc/invoke.texi: Document -mfunction-return= option.
13407 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
13409 * config/i386/i386-opts.h (indirect_branch): New.
13410 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
13411 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
13412 with local indirect jump when converting indirect call and jump.
13413 (ix86_set_indirect_branch_type): New.
13414 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
13415 (indirectlabelno): New.
13416 (indirect_thunk_needed): Likewise.
13417 (indirect_thunk_bnd_needed): Likewise.
13418 (indirect_thunks_used): Likewise.
13419 (indirect_thunks_bnd_used): Likewise.
13420 (INDIRECT_LABEL): Likewise.
13421 (indirect_thunk_name): Likewise.
13422 (output_indirect_thunk): Likewise.
13423 (output_indirect_thunk_function): Likewise.
13424 (ix86_output_indirect_branch_via_reg): Likewise.
13425 (ix86_output_indirect_branch_via_push): Likewise.
13426 (ix86_output_indirect_branch): Likewise.
13427 (ix86_output_indirect_jmp): Likewise.
13428 (ix86_code_end): Call output_indirect_thunk_function if needed.
13429 (ix86_output_call_insn): Call ix86_output_indirect_branch if
13431 (ix86_handle_fndecl_attribute): Handle indirect_branch.
13432 (ix86_attribute_table): Add indirect_branch.
13433 * config/i386/i386.h (machine_function): Add indirect_branch_type
13434 and has_local_indirect_jump.
13435 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
13437 (tablejump): Likewise.
13438 (*indirect_jump): Use ix86_output_indirect_jmp.
13439 (*tablejump_1): Likewise.
13440 (simple_return_indirect_internal): Likewise.
13441 * config/i386/i386.opt (mindirect-branch=): New option.
13442 (indirect_branch): New.
13445 (thunk-inline): Likewise.
13446 (thunk-extern): Likewise.
13447 * doc/extend.texi: Document indirect_branch function attribute.
13448 * doc/invoke.texi: Document -mindirect-branch= option.
13450 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
13453 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
13455 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
13457 * ipa-inline.c (want_inline_small_function_p): Return false if
13458 inlining has already failed with CIF_FINAL_ERROR.
13459 (update_caller_keys): Call want_inline_small_function_p before
13461 (update_callee_keys): Likewise.
13463 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
13465 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
13467 (rs6000_quadword_masked_address_p): Likewise.
13468 (quad_aligned_load_p): Likewise.
13469 (quad_aligned_store_p): Likewise.
13470 (const_load_sequence_p): Add comment to describe the outer-most loop.
13471 (mimic_memory_attributes_and_flags): New function.
13472 (rs6000_gen_stvx): Likewise.
13473 (replace_swapped_aligned_store): Likewise.
13474 (rs6000_gen_lvx): Likewise.
13475 (replace_swapped_aligned_load): Likewise.
13476 (replace_swapped_load_constant): Capitalize argument name in
13477 comment describing this function.
13478 (rs6000_analyze_swaps): Add a third pass to search for vector loads
13479 and stores that access quad-word aligned addresses and replace
13480 with stvx or lvx instructions when appropriate.
13481 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
13482 New function prototype.
13483 (rs6000_quadword_masked_address_p): Likewise.
13484 (rs6000_gen_lvx): Likewise.
13485 (rs6000_gen_stvx): Likewise.
13486 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
13487 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
13488 when memory address is aligned.
13489 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
13490 this split to select lvx instruction when memory address is aligned.
13491 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
13492 instruction when memory address is aligned.
13493 (*vsx_le_perm_load_v16qi): Likewise.
13494 (four unnamed splitters): Modify to select the stvx instruction
13495 when memory is aligned.
13497 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
13499 * predict.c (determine_unlikely_bbs): Handle correctly BBs
13500 which appears in the queue multiple times.
13502 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13503 Alan Hayward <alan.hayward@arm.com>
13504 David Sherwood <david.sherwood@arm.com>
13506 * tree-vectorizer.h (vec_lower_bound): New structure.
13507 (_loop_vec_info): Add check_nonzero and lower_bounds.
13508 (LOOP_VINFO_CHECK_NONZERO): New macro.
13509 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
13510 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
13511 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
13512 fields. Make seg_len the distance travelled, not including the
13514 (dr_direction_indicator): Declare.
13515 (dr_zero_step_indicator): Likewise.
13516 (dr_known_forward_stride_p): Likewise.
13517 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
13519 (runtime_alias_check_p): Allow runtime alias checks with
13521 (operator ==): Compare access_size and align.
13522 (prune_runtime_alias_test_list): Rework for new distinction between
13523 the access_size and seg_len.
13524 (create_intersect_range_checks_index): Likewise. Cope with polynomial
13526 (get_segment_min_max): New function.
13527 (create_intersect_range_checks): Use it.
13528 (dr_step_indicator): New function.
13529 (dr_direction_indicator): Likewise.
13530 (dr_zero_step_indicator): Likewise.
13531 (dr_known_forward_stride_p): Likewise.
13532 * tree-loop-distribution.c (data_ref_segment_size): Return
13533 DR_STEP * (niters - 1).
13534 (compute_alias_check_pairs): Update call to the dr_with_seg_len
13536 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
13537 (vect_preserves_scalar_order_p): New function, split out from...
13538 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
13539 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
13540 (vect_vfa_access_size): New function.
13541 (vect_vfa_align): Likewise.
13542 (vect_compile_time_alias): Take access_size_a and access_b arguments.
13543 (dump_lower_bound): New function.
13544 (vect_check_lower_bound): Likewise.
13545 (vect_small_gap_p): Likewise.
13546 (vectorizable_with_step_bound_p): Likewise.
13547 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
13548 depencies if the vectorization factor is 1. Convert the checks
13549 for nonzero steps into checks on the bounds of DR_STEP. Try using
13550 a bunds check for variable steps if the minimum required step is
13551 relatively small. Update calls to the dr_with_seg_len
13552 constructor and to vect_compile_time_alias.
13553 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
13555 (vect_loop_versioning): Call it.
13556 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
13558 (vect_estimate_min_profitable_iters): Account for any bounds checks.
13560 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13561 Alan Hayward <alan.hayward@arm.com>
13562 David Sherwood <david.sherwood@arm.com>
13564 * doc/sourcebuild.texi (vect_scatter_store): Document.
13565 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
13567 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
13569 * genopinit.c (main): Add supports_vec_scatter_store and
13570 supports_vec_scatter_store_cached to target_optabs.
13571 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
13572 IFN_MASK_SCATTER_STORE.
13573 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
13575 * internal-fn.h (internal_store_fn_p): Declare.
13576 (internal_fn_stored_value_index): Likewise.
13577 * internal-fn.c (scatter_store_direct): New macro.
13578 (expand_scatter_store_optab_fn): New function.
13579 (direct_scatter_store_optab_supported_p): New macro.
13580 (internal_store_fn_p): New function.
13581 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
13582 IFN_MASK_SCATTER_STORE.
13583 (internal_fn_mask_index): Likewise.
13584 (internal_fn_stored_value_index): New function.
13585 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
13586 for scatter stores.
13587 * optabs-query.h (supports_vec_scatter_store_p): Declare.
13588 * optabs-query.c (supports_vec_scatter_store_p): New function.
13589 * tree-vectorizer.h (vect_get_store_rhs): Declare.
13590 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
13591 true for scatter stores.
13592 (vect_gather_scatter_fn_p): Handle scatter stores too.
13593 (vect_check_gather_scatter): Consider using scatter stores if
13594 supports_vec_scatter_store_p.
13595 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
13596 scatter stores too.
13597 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
13598 internal_fn_stored_value_index.
13599 (check_load_store_masking): Handle scatter stores too.
13600 (vect_get_store_rhs): Make public.
13601 (vectorizable_call): Use internal_store_fn_p.
13602 (vectorizable_store): Handle scatter store internal functions.
13603 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
13604 when deciding whether the end of the group has been reached.
13605 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
13606 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
13607 (mask_scatter_store<mode>): New insns.
13609 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13610 Alan Hayward <alan.hayward@arm.com>
13611 David Sherwood <david.sherwood@arm.com>
13613 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
13614 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
13615 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
13617 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
13618 Use vect_truncate_gather_scatter_offset if we can't treat the
13619 operation as a normal gather load or scatter store.
13620 (get_group_load_store_type): Take the gather_scatter_info
13621 as argument. Try using a gather load or scatter store for
13622 single-element groups.
13623 (get_load_store_type): Update calls to get_group_load_store_type
13624 and vect_use_strided_gather_scatters_p.
13626 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13627 Alan Hayward <alan.hayward@arm.com>
13628 David Sherwood <david.sherwood@arm.com>
13630 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
13631 optional tree argument.
13632 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
13634 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
13635 but continue to use the current value as a fallback.
13636 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
13637 to compare the updates.
13638 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
13639 (get_load_store_type): Use it when handling a strided access.
13640 (vect_get_strided_load_store_ops): New function.
13641 (vect_get_data_ptr_increment): Likewise.
13642 (vectorizable_load): Handle strided gather loads. Always pass
13643 a step to vect_create_data_ref_ptr and bump_vector_ptr.
13645 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13646 Alan Hayward <alan.hayward@arm.com>
13647 David Sherwood <david.sherwood@arm.com>
13649 * doc/md.texi (gather_load@var{m}): Document.
13650 (mask_gather_load@var{m}): Likewise.
13651 * genopinit.c (main): Add supports_vec_gather_load and
13652 supports_vec_gather_load_cached to target_optabs.
13653 * optabs-tree.c (init_tree_optimization_optabs): Use
13654 ggc_cleared_alloc to allocate target_optabs.
13655 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
13656 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
13658 * internal-fn.h (internal_load_fn_p): Declare.
13659 (internal_gather_scatter_fn_p): Likewise.
13660 (internal_fn_mask_index): Likewise.
13661 (internal_gather_scatter_fn_supported_p): Likewise.
13662 * internal-fn.c (gather_load_direct): New macro.
13663 (expand_gather_load_optab_fn): New function.
13664 (direct_gather_load_optab_supported_p): New macro.
13665 (direct_internal_fn_optab): New function.
13666 (internal_load_fn_p): Likewise.
13667 (internal_gather_scatter_fn_p): Likewise.
13668 (internal_fn_mask_index): Likewise.
13669 (internal_gather_scatter_fn_supported_p): Likewise.
13670 * optabs-query.c (supports_at_least_one_mode_p): New function.
13671 (supports_vec_gather_load_p): Likewise.
13672 * optabs-query.h (supports_vec_gather_load_p): Declare.
13673 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
13674 and memory_type field.
13675 (NUM_PATTERNS): Bump to 15.
13676 * tree-vect-data-refs.c: Include internal-fn.h.
13677 (vect_gather_scatter_fn_p): New function.
13678 (vect_describe_gather_scatter_call): Likewise.
13679 (vect_check_gather_scatter): Try using internal functions for
13680 gather loads. Recognize existing calls to a gather load function.
13681 (vect_analyze_data_refs): Consider using gather loads if
13682 supports_vec_gather_load_p.
13683 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
13684 (vect_get_gather_scatter_offset_type): Likewise.
13685 (vect_convert_mask_for_vectype): Likewise.
13686 (vect_add_conversion_to_patterm): Likewise.
13687 (vect_try_gather_scatter_pattern): Likewise.
13688 (vect_recog_gather_scatter_pattern): New pattern recognizer.
13689 (vect_vect_recog_func_ptrs): Add it.
13690 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
13691 internal_fn_mask_index and internal_gather_scatter_fn_p.
13692 (check_load_store_masking): Take the gather_scatter_info as an
13693 argument and handle gather loads.
13694 (vect_get_gather_scatter_ops): New function.
13695 (vectorizable_call): Check internal_load_fn_p.
13696 (vectorizable_load): Likewise. Handle gather load internal
13698 (vectorizable_store): Update call to check_load_store_masking.
13699 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
13700 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
13701 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
13702 (aarch64_gather_scale_operand_d): New predicates.
13703 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
13704 (mask_gather_load<mode>): New insns.
13706 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13707 Alan Hayward <alan.hayward@arm.com>
13708 David Sherwood <david.sherwood@arm.com>
13710 * optabs.def (fold_left_plus_optab): New optab.
13711 * doc/md.texi (fold_left_plus_@var{m}): Document.
13712 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
13713 * internal-fn.c (fold_left_direct): Define.
13714 (expand_fold_left_optab_fn): Likewise.
13715 (direct_fold_left_optab_supported_p): Likewise.
13716 * fold-const-call.c (fold_const_fold_left): New function.
13717 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
13718 * tree-parloops.c (valid_reduction_p): New function.
13719 (gather_scalar_reductions): Use it.
13720 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
13721 (vect_finish_replace_stmt): Declare.
13722 * tree-vect-loop.c (fold_left_reduction_fn): New function.
13723 (needs_fold_left_reduction_p): New function, split out from...
13724 (vect_is_simple_reduction): ...here. Accept reductions that
13725 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
13726 (vect_force_simple_reduction): Also store the reduction type in
13727 the assignment's STMT_VINFO_REDUC_TYPE.
13728 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
13729 (merge_with_identity): New function.
13730 (vect_expand_fold_left): Likewise.
13731 (vectorize_fold_left_reduction): Likewise.
13732 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
13733 scalar phi in place for it. Check for target support and reject
13734 cases that would reassociate the operation. Defer the transform
13735 phase to vectorize_fold_left_reduction.
13736 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
13737 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
13738 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
13740 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13742 * tree-if-conv.c (predicate_mem_writes): Remove redundant
13743 call to ifc_temp_var.
13745 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13746 Alan Hayward <alan.hayward@arm.com>
13747 David Sherwood <david.sherwood@arm.com>
13749 * target.def (legitimize_address_displacement): Take the original
13750 offset as a poly_int.
13751 * targhooks.h (default_legitimize_address_displacement): Update
13753 * targhooks.c (default_legitimize_address_displacement): Likewise.
13754 * doc/tm.texi: Regenerate.
13755 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
13756 as an argument, moving assert of ad->disp == ad->disp_term to...
13757 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
13758 Try calling targetm.legitimize_address_displacement before expanding
13759 the address rather than afterwards, and adjust for the new interface.
13760 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
13761 Match the new hook interface. Handle SVE addresses.
13762 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
13763 new hook interface.
13765 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13767 * Makefile.in (OBJS): Add early-remat.o.
13768 * target.def (select_early_remat_modes): New hook.
13769 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
13770 * doc/tm.texi: Regenerate.
13771 * targhooks.h (default_select_early_remat_modes): Declare.
13772 * targhooks.c (default_select_early_remat_modes): New function.
13773 * timevar.def (TV_EARLY_REMAT): New timevar.
13774 * passes.def (pass_early_remat): New pass.
13775 * tree-pass.h (make_pass_early_remat): Declare.
13776 * early-remat.c: New file.
13777 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
13779 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
13781 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13782 Alan Hayward <alan.hayward@arm.com>
13783 David Sherwood <david.sherwood@arm.com>
13785 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
13786 vfm1 with a bound_epilog parameter.
13787 (vect_do_peeling): Update calls accordingly, and move the prologue
13788 call earlier in the function. Treat the base bound_epilog as 0 for
13789 fully-masked loops and retain vf - 1 for other loops. Add 1 to
13790 this base when peeling for gaps.
13791 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
13792 with fully-masked loops.
13793 (vect_estimate_min_profitable_iters): Handle the single peeled
13794 iteration in that case.
13796 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13797 Alan Hayward <alan.hayward@arm.com>
13798 David Sherwood <david.sherwood@arm.com>
13800 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
13801 single-element interleaving even if the size is not a power of 2.
13802 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
13803 accesses for single-element interleaving if the group size is
13806 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13807 Alan Hayward <alan.hayward@arm.com>
13808 David Sherwood <david.sherwood@arm.com>
13810 * doc/md.texi (fold_extract_last_@var{m}): Document.
13811 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
13812 * optabs.def (fold_extract_last_optab): New optab.
13813 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
13814 * internal-fn.c (fold_extract_direct): New macro.
13815 (expand_fold_extract_optab_fn): Likewise.
13816 (direct_fold_extract_optab_supported_p): Likewise.
13817 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
13818 * tree-vect-loop.c (vect_model_reduction_cost): Handle
13819 EXTRACT_LAST_REDUCTION.
13820 (get_initial_def_for_reduction): Do not create an initial vector
13821 for EXTRACT_LAST_REDUCTION reductions.
13822 (vectorizable_reduction): Leave the scalar phi in place for
13823 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
13824 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
13825 epilogue code for EXTRACT_LAST_REDUCTION and defer the
13826 transform phase to vectorizable_condition.
13827 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
13829 (vect_finish_stmt_generation): ...here.
13830 (vect_finish_replace_stmt): New function.
13831 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
13832 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
13834 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
13836 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13837 Alan Hayward <alan.hayward@arm.com>
13838 David Sherwood <david.sherwood@arm.com>
13840 * doc/md.texi (extract_last_@var{m}): Document.
13841 * optabs.def (extract_last_optab): New optab.
13842 * internal-fn.def (EXTRACT_LAST): New internal function.
13843 * internal-fn.c (cond_unary_direct): New macro.
13844 (expand_cond_unary_optab_fn): Likewise.
13845 (direct_cond_unary_optab_supported_p): Likewise.
13846 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
13847 loops using EXTRACT_LAST.
13848 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
13849 (extract_last_<mode>): ...this optab.
13850 (vec_extract<mode><Vel>): Update accordingly.
13852 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13853 Alan Hayward <alan.hayward@arm.com>
13854 David Sherwood <david.sherwood@arm.com>
13856 * target.def (empty_mask_is_expensive): New hook.
13857 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
13858 * doc/tm.texi: Regenerate.
13859 * targhooks.h (default_empty_mask_is_expensive): Declare.
13860 * targhooks.c (default_empty_mask_is_expensive): New function.
13861 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
13862 if the target says that empty masks are expensive.
13863 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
13865 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
13867 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13868 Alan Hayward <alan.hayward@arm.com>
13869 David Sherwood <david.sherwood@arm.com>
13871 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
13872 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
13873 (vect_use_loop_mask_for_alignment_p): New function.
13874 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
13875 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
13876 niters_skip argument. Make sure that the first niters_skip elements
13877 of the first iteration are inactive.
13878 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
13879 Update call to vect_set_loop_masks_directly.
13880 (get_misalign_in_elems): New function, split out from...
13881 (vect_gen_prolog_loop_niters): ...here.
13882 (vect_update_init_of_dr): Take a code argument that specifies whether
13883 the adjustment should be added or subtracted.
13884 (vect_update_init_of_drs): Likewise.
13885 (vect_prepare_for_masked_peels): New function.
13886 (vect_do_peeling): Skip prologue peeling if we're using a mask
13887 instead. Update call to vect_update_inits_of_drs.
13888 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
13890 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
13891 alignment. Do not include the number of peeled iterations in
13892 the minimum threshold in that case.
13893 (vectorizable_induction): Adjust the start value down by
13894 LOOP_VINFO_MASK_SKIP_NITERS iterations.
13895 (vect_transform_loop): Call vect_prepare_for_masked_peels.
13896 Take the number of skipped iterations into account when calculating
13898 * tree-vect-stmts.c (vect_gen_while_not): New function.
13900 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13901 Alan Hayward <alan.hayward@arm.com>
13902 David Sherwood <david.sherwood@arm.com>
13904 * doc/sourcebuild.texi (vect_fully_masked): Document.
13905 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
13906 default value to 0.
13907 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
13909 (vect_analyze_loop_2): ...here. Don't check the vectorization
13910 factor against the number of loop iterations if the loop is
13913 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13914 Alan Hayward <alan.hayward@arm.com>
13915 David Sherwood <david.sherwood@arm.com>
13917 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
13918 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
13919 (dump_groups): Update accordingly.
13920 (iv_use::mem_type): New member variable.
13921 (address_p): New function.
13922 (record_use): Add a mem_type argument and initialize the new
13924 (record_group_use): Add a mem_type argument. Use address_p.
13925 Remove obsolete null checks of base_object. Update call to record_use.
13926 (find_interesting_uses_op): Update call to record_group_use.
13927 (find_interesting_uses_cond): Likewise.
13928 (find_interesting_uses_address): Likewise.
13929 (get_mem_type_for_internal_fn): New function.
13930 (find_address_like_use): Likewise.
13931 (find_interesting_uses_stmt): Try find_address_like_use before
13932 calling find_interesting_uses_op.
13933 (addr_offset_valid_p): Use the iv mem_type field as the type
13934 of the addressed memory.
13935 (add_autoinc_candidates): Likewise.
13936 (get_address_cost): Likewise.
13937 (split_small_address_groups_p): Use address_p.
13938 (split_address_groups): Likewise.
13939 (add_iv_candidate_for_use): Likewise.
13940 (autoinc_possible_for_pair): Likewise.
13941 (rewrite_groups): Likewise.
13942 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
13943 (determine_group_iv_cost): Update after split of USE_ADDRESS.
13944 (get_alias_ptr_type_for_ptr_address): New function.
13945 (rewrite_use_address): Rewrite address uses in calls that were
13946 identified by find_address_like_use.
13948 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13949 Alan Hayward <alan.hayward@arm.com>
13950 David Sherwood <david.sherwood@arm.com>
13952 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
13954 * gimple-expr.h (is_gimple_addressable: Likewise.
13955 * gimple-expr.c (is_gimple_address): Likewise.
13956 * internal-fn.c (expand_call_mem_ref): New function.
13957 (expand_mask_load_optab_fn): Use it.
13958 (expand_mask_store_optab_fn): Likewise.
13960 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13961 Alan Hayward <alan.hayward@arm.com>
13962 David Sherwood <david.sherwood@arm.com>
13964 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
13965 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
13966 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
13967 (cond_umax@var{mode}): Document.
13968 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
13969 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
13970 (cond_umin_optab, cond_umax_optab): New optabs.
13971 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
13972 (COND_IOR, COND_XOR): New internal functions.
13973 * internal-fn.h (get_conditional_internal_fn): Declare.
13974 * internal-fn.c (cond_binary_direct): New macro.
13975 (expand_cond_binary_optab_fn): Likewise.
13976 (direct_cond_binary_optab_supported_p): Likewise.
13977 (get_conditional_internal_fn): New function.
13978 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
13979 Cope with reduction statements that are vectorized as calls rather
13981 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
13982 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
13983 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
13984 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
13985 (UNSPEC_COND_EOR): New unspecs.
13986 (optab): Add mappings for them.
13987 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
13988 (sve_int_op, sve_fp_op): New int attributes.
13990 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13991 Alan Hayward <alan.hayward@arm.com>
13992 David Sherwood <david.sherwood@arm.com>
13994 * optabs.def (while_ult_optab): New optab.
13995 * doc/md.texi (while_ult@var{m}@var{n}): Document.
13996 * internal-fn.def (WHILE_ULT): New internal function.
13997 * internal-fn.h (direct_internal_fn_supported_p): New override
13998 that takes two types as argument.
13999 * internal-fn.c (while_direct): New macro.
14000 (expand_while_optab_fn): New function.
14001 (convert_optab_supported_p): Likewise.
14002 (direct_while_optab_supported_p): New macro.
14003 * wide-int.h (wi::udiv_ceil): New function.
14004 * tree-vectorizer.h (rgroup_masks): New structure.
14005 (vec_loop_masks): New typedef.
14006 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
14007 and fully_masked_p.
14008 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
14009 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
14010 (vect_max_vf): New function.
14011 (slpeel_make_loop_iterate_ntimes): Delete.
14012 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
14013 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
14014 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
14015 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
14016 internal-fn.h, stor-layout.h and optabs-query.h.
14017 (vect_set_loop_mask): New function.
14018 (add_preheader_seq): Likewise.
14019 (add_header_seq): Likewise.
14020 (interleave_supported_p): Likewise.
14021 (vect_maybe_permute_loop_masks): Likewise.
14022 (vect_set_loop_masks_directly): Likewise.
14023 (vect_set_loop_condition_masked): Likewise.
14024 (vect_set_loop_condition_unmasked): New function, split out from
14025 slpeel_make_loop_iterate_ntimes.
14026 (slpeel_make_loop_iterate_ntimes): Rename to..
14027 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
14028 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
14029 (vect_do_peeling): Update call accordingly.
14030 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
14032 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
14033 mask_compare_type, can_fully_mask_p and fully_masked_p.
14034 (release_vec_loop_masks): New function.
14035 (_loop_vec_info): Use it to free the loop masks.
14036 (can_produce_all_loop_masks_p): New function.
14037 (vect_get_max_nscalars_per_iter): Likewise.
14038 (vect_verify_full_masking): Likewise.
14039 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
14040 retries, and free the mask rgroups before retrying. Check loop-wide
14041 reasons for disallowing fully-masked loops. Make the final decision
14042 about whether use a fully-masked loop or not.
14043 (vect_estimate_min_profitable_iters): Do not assume that peeling
14044 for the number of iterations will be needed for fully-masked loops.
14045 (vectorizable_reduction): Disable fully-masked loops.
14046 (vectorizable_live_operation): Likewise.
14047 (vect_halve_mask_nunits): New function.
14048 (vect_double_mask_nunits): Likewise.
14049 (vect_record_loop_mask): Likewise.
14050 (vect_get_loop_mask): Likewise.
14051 (vect_transform_loop): Handle the case in which the final loop
14052 iteration might handle a partial vector. Call vect_set_loop_condition
14053 instead of slpeel_make_loop_iterate_ntimes.
14054 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
14055 (check_load_store_masking): New function.
14056 (prepare_load_store_mask): Likewise.
14057 (vectorizable_store): Handle fully-masked loops.
14058 (vectorizable_load): Likewise.
14059 (supportable_widening_operation): Use vect_halve_mask_nunits for
14061 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
14062 (vect_gen_while): New function.
14063 * config/aarch64/aarch64.md (umax<mode>3): New expander.
14064 (aarch64_uqdec<mode>): New insn.
14066 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14067 Alan Hayward <alan.hayward@arm.com>
14068 David Sherwood <david.sherwood@arm.com>
14070 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
14071 (reduc_xor_scal_optab): New optabs.
14072 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
14073 (reduc_xor_scal_@var{m}): Document.
14074 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
14075 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
14076 internal functions.
14077 * fold-const-call.c (fold_const_call): Handle them.
14078 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
14079 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
14080 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
14081 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
14082 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
14083 (UNSPEC_XORV): New unspecs.
14084 (optab): Add entries for them.
14085 (BITWISEV): New int iterator.
14086 (bit_reduc_op): New int attributes.
14088 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14089 Alan Hayward <alan.hayward@arm.com>
14090 David Sherwood <david.sherwood@arm.com>
14092 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
14093 * internal-fn.def (VEC_SHL_INSERT): New internal function.
14094 * optabs.def (vec_shl_insert_optab): New optab.
14095 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
14096 (duplicate_and_interleave): Likewise.
14097 * tree-vect-loop.c: Include internal-fn.h.
14098 (neutral_op_for_slp_reduction): New function, split out from
14099 get_initial_defs_for_reduction.
14100 (get_initial_def_for_reduction): Handle option 2 for variable-length
14101 vectors by loading the neutral value into a vector and then shifting
14102 the initial value into element 0.
14103 (get_initial_defs_for_reduction): Replace the code argument with
14104 the neutral value calculated by neutral_op_for_slp_reduction.
14105 Use gimple_build_vector for constant-length vectors.
14106 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
14107 but the first group_size elements have a neutral value.
14108 Use duplicate_and_interleave otherwise.
14109 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
14110 Update call to get_initial_defs_for_reduction. Handle SLP
14111 reductions for variable-length vectors by creating one vector
14112 result for each scalar result, with the elements associated
14113 with other scalar results stubbed out with the neutral value.
14114 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
14115 Require IFN_VEC_SHL_INSERT for double reductions on
14116 variable-length vectors, or SLP reductions that have
14117 a neutral value. Require can_duplicate_and_interleave_p
14118 support for variable-length unchained SLP reductions if there
14119 is no neutral value, such as for MIN/MAX reductions. Also require
14120 the number of vector elements to be a multiple of the number of
14121 SLP statements when doing variable-length unchained SLP reductions.
14122 Update call to vect_create_epilog_for_reduction.
14123 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
14124 and remove initial values.
14125 (duplicate_and_interleave): Make public.
14126 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
14127 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
14129 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14130 Alan Hayward <alan.hayward@arm.com>
14131 David Sherwood <david.sherwood@arm.com>
14133 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
14134 (can_duplicate_and_interleave_p): New function.
14135 (vect_get_and_check_slp_defs): Take the vector of statements
14136 rather than just the current one. Remove excess parentheses.
14137 Restriction rejectinon of vect_constant_def and vect_external_def
14138 for variable-length vectors to boolean types, or types for which
14139 can_duplicate_and_interleave_p is false.
14140 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
14141 (duplicate_and_interleave): New function.
14142 (vect_get_constant_vectors): Use gimple_build_vector for
14143 constant-length vectors and suitable variable-length constant
14144 vectors. Use duplicate_and_interleave for other variable-length
14145 vectors. Don't defer the update when inserting new statements.
14147 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14148 Alan Hayward <alan.hayward@arm.com>
14149 David Sherwood <david.sherwood@arm.com>
14151 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
14152 min_profitable_iters doesn't go negative.
14154 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14155 Alan Hayward <alan.hayward@arm.com>
14156 David Sherwood <david.sherwood@arm.com>
14158 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
14159 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
14160 * optabs.def (vec_mask_load_lanes_optab): New optab.
14161 (vec_mask_store_lanes_optab): Likewise.
14162 * internal-fn.def (MASK_LOAD_LANES): New internal function.
14163 (MASK_STORE_LANES): Likewise.
14164 * internal-fn.c (mask_load_lanes_direct): New macro.
14165 (mask_store_lanes_direct): Likewise.
14166 (expand_mask_load_optab_fn): Handle masked operations.
14167 (expand_mask_load_lanes_optab_fn): New macro.
14168 (expand_mask_store_optab_fn): Handle masked operations.
14169 (expand_mask_store_lanes_optab_fn): New macro.
14170 (direct_mask_load_lanes_optab_supported_p): Likewise.
14171 (direct_mask_store_lanes_optab_supported_p): Likewise.
14172 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
14174 (vect_load_lanes_supported): Likewise.
14175 * tree-vect-data-refs.c (strip_conversion): New function.
14176 (can_group_stmts_p): Likewise.
14177 (vect_analyze_data_ref_accesses): Use it instead of checking
14178 for a pair of assignments.
14179 (vect_store_lanes_supported): Take a masked_p parameter.
14180 (vect_load_lanes_supported): Likewise.
14181 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
14182 vect_store_lanes_supported and vect_load_lanes_supported.
14183 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
14184 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
14185 parameter. Don't allow gaps for masked accesses.
14186 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
14187 and vect_load_lanes_supported.
14188 (get_load_store_type): Take a masked_p parameter and update
14189 call to get_group_load_store_type.
14190 (vectorizable_store): Update call to get_load_store_type.
14191 Handle IFN_MASK_STORE_LANES.
14192 (vectorizable_load): Update call to get_load_store_type.
14193 Handle IFN_MASK_LOAD_LANES.
14195 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14196 Alan Hayward <alan.hayward@arm.com>
14197 David Sherwood <david.sherwood@arm.com>
14199 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
14201 * config/aarch64/aarch64-protos.h
14202 (aarch64_sve_struct_memory_operand_p): Declare.
14203 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
14204 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
14205 (VPRED, vpred): Handle SVE structure modes.
14206 * config/aarch64/constraints.md (Utx): New constraint.
14207 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
14208 (aarch64_sve_struct_nonimmediate_operand): New predicates.
14209 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
14210 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
14211 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
14212 structure modes. Split into pieces after RA.
14213 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
14214 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
14216 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
14217 SVE structure modes.
14218 (aarch64_classify_address): Likewise.
14219 (sizetochar): Move earlier in file.
14220 (aarch64_print_operand): Handle SVE register lists.
14221 (aarch64_array_mode): New function.
14222 (aarch64_sve_struct_memory_operand_p): Likewise.
14223 (TARGET_ARRAY_MODE): Redefine.
14225 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14226 Alan Hayward <alan.hayward@arm.com>
14227 David Sherwood <david.sherwood@arm.com>
14229 * target.def (array_mode): New target hook.
14230 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
14231 * doc/tm.texi: Regenerate.
14232 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
14233 * hooks.c (hook_optmode_mode_uhwi_none): New function.
14234 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
14235 targetm.array_mode.
14236 * stor-layout.c (mode_for_array): Likewise. Support polynomial
14239 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14240 Alan Hayward <alan.hayward@arm.com>
14241 David Sherwood <david.sherwood@arm.com>
14243 * fold-const.c (fold_binary_loc): Check the argument types
14244 rather than the result type when testing for a vector operation.
14246 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14248 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
14249 * doc/tm.texi: Regenerate.
14251 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14252 Alan Hayward <alan.hayward@arm.com>
14253 David Sherwood <david.sherwood@arm.com>
14255 * doc/invoke.texi (-msve-vector-bits=): Document new option.
14256 (sve): Document new AArch64 extension.
14257 * doc/md.texi (w): Extend the description of the AArch64
14258 constraint to include SVE vectors.
14259 (Upl, Upa): Document new AArch64 predicate constraints.
14260 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
14262 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
14263 (msve-vector-bits=): New option.
14264 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
14265 SVE when these are disabled.
14266 (sve): New extension.
14267 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
14268 modes. Adjust their number of units based on aarch64_sve_vg.
14269 (MAX_BITSIZE_MODE_ANY_MODE): Define.
14270 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
14271 aarch64_addr_query_type.
14272 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
14273 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
14274 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
14275 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
14276 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
14277 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
14278 (aarch64_simd_imm_zero_p): Delete.
14279 (aarch64_check_zero_based_sve_index_immediate): Declare.
14280 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
14281 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
14282 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
14283 (aarch64_sve_float_mul_immediate_p): Likewise.
14284 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
14285 rather than an rtx.
14286 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
14287 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
14288 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
14289 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
14290 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
14291 (aarch64_regmode_natural_size): Likewise.
14292 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
14293 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
14295 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
14296 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
14297 for VG and the SVE predicate registers.
14298 (V_ALIASES): Add a "z"-prefixed alias.
14299 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
14300 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
14301 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
14302 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
14303 (REG_CLASS_NAMES): Add entries for them.
14304 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
14305 and the predicate registers.
14306 (aarch64_sve_vg): Declare.
14307 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
14308 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
14309 (REGMODE_NATURAL_SIZE): Define.
14310 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
14312 * config/aarch64/aarch64.c: Include cfgrtl.h.
14313 (simd_immediate_info): Add a constructor for series vectors,
14314 and an associated step field.
14315 (aarch64_sve_vg): New variable.
14316 (aarch64_dbx_register_number): Handle VG and the predicate registers.
14317 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
14318 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
14319 (VEC_ANY_DATA, VEC_STRUCT): New constants.
14320 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
14321 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
14322 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
14323 (aarch64_get_mask_mode): New functions.
14324 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
14325 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
14326 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
14327 predicate modes and predicate registers. Explicitly restrict
14328 GPRs to modes of 16 bytes or smaller. Only allow FP registers
14329 to store a vector mode if it is recognized by
14330 aarch64_classify_vector_mode.
14331 (aarch64_regmode_natural_size): New function.
14332 (aarch64_hard_regno_caller_save_mode): Return the original mode
14334 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
14335 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
14336 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
14337 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
14339 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
14340 does not overlap dest if the function is frame-related. Handle
14342 (aarch64_split_add_offset): New function.
14343 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
14344 them aarch64_add_offset.
14345 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
14346 and update call to aarch64_sub_sp.
14347 (aarch64_add_cfa_expression): New function.
14348 (aarch64_expand_prologue): Pass extra temporary registers to the
14349 functions above. Handle the case in which we need to emit new
14350 DW_CFA_expressions for registers that were originally saved
14351 relative to the stack pointer, but now have to be expressed
14352 relative to the frame pointer.
14353 (aarch64_output_mi_thunk): Pass extra temporary registers to the
14355 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
14356 IP0 and IP1 values for SVE frames.
14357 (aarch64_expand_vec_series): New function.
14358 (aarch64_expand_sve_widened_duplicate): Likewise.
14359 (aarch64_expand_sve_const_vector): Likewise.
14360 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
14361 Handle SVE constants. Use emit_move_insn to move a force_const_mem
14362 into the register, rather than emitting a SET directly.
14363 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
14364 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
14365 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
14366 (offset_9bit_signed_scaled_p): New functions.
14367 (aarch64_replicate_bitmask_imm): New function.
14368 (aarch64_bitmask_imm): Use it.
14369 (aarch64_cannot_force_const_mem): Reject expressions involving
14370 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
14371 (aarch64_classify_index): Handle SVE indices, by requiring
14372 a plain register index with a scale that matches the element size.
14373 (aarch64_classify_address): Handle SVE addresses. Assert that
14374 the mode of the address is VOIDmode or an integer mode.
14375 Update call to aarch64_classify_symbol.
14376 (aarch64_classify_symbolic_expression): Update call to
14377 aarch64_classify_symbol.
14378 (aarch64_const_vec_all_in_range_p): New function.
14379 (aarch64_print_vector_float_operand): Likewise.
14380 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
14381 "vN" for FP registers with SVE modes. Handle (const ...) vectors
14382 and the FP immediates 1.0 and 0.5.
14383 (aarch64_print_address_internal): Handle SVE addresses.
14384 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
14385 (aarch64_regno_regclass): Handle predicate registers.
14386 (aarch64_secondary_reload): Handle big-endian reloads of SVE
14388 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
14389 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
14390 (aarch64_convert_sve_vector_bits): New function.
14391 (aarch64_override_options): Use it to handle -msve-vector-bits=.
14392 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
14393 rather than an rtx.
14394 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
14395 Handle SVE vector and predicate modes. Accept VL-based constants
14396 that need only one temporary register, and VL offsets that require
14397 no temporary registers.
14398 (aarch64_conditional_register_usage): Mark the predicate registers
14399 as fixed if SVE isn't available.
14400 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
14401 Return true for SVE vector and predicate modes.
14402 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
14403 rather than an unsigned int. Handle SVE modes.
14404 (aarch64_preferred_simd_mode): Update call accordingly. Handle
14406 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
14408 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
14409 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
14410 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
14411 (aarch64_sve_float_mul_immediate_p): New functions.
14412 (aarch64_sve_valid_immediate): New function.
14413 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
14414 Explicitly reject structure modes. Check for INDEX constants.
14415 Handle PTRUE and PFALSE constants.
14416 (aarch64_check_zero_based_sve_index_immediate): New function.
14417 (aarch64_simd_imm_zero_p): Delete.
14418 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
14419 vector modes. Accept constants in the range of CNT[BHWD].
14420 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
14421 ask for an Advanced SIMD mode.
14422 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
14423 (aarch64_simd_vector_alignment): Handle SVE predicates.
14424 (aarch64_vectorize_preferred_vector_alignment): New function.
14425 (aarch64_simd_vector_alignment_reachable): Use it instead of
14427 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
14428 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
14430 (MAX_VECT_LEN): Delete.
14431 (expand_vec_perm_d): Add a vec_flags field.
14432 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
14433 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
14434 (aarch64_evpc_ext): Don't apply a big-endian lane correction
14436 (aarch64_evpc_rev): Rename to...
14437 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
14438 (aarch64_evpc_rev_global): New function.
14439 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
14440 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
14442 (aarch64_evpc_sve_tbl): New function.
14443 (aarch64_expand_vec_perm_const_1): Update after rename of
14444 aarch64_evpc_rev. Handle SVE permutes too, trying
14445 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
14446 than aarch64_evpc_tbl.
14447 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
14448 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
14449 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
14450 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
14451 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
14452 (aarch64_expand_sve_vcond): New functions.
14453 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
14454 of aarch64_vector_mode_p.
14455 (aarch64_dwarf_poly_indeterminate_value): New function.
14456 (aarch64_compute_pressure_classes): Likewise.
14457 (aarch64_can_change_mode_class): Likewise.
14458 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
14459 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
14460 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
14461 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
14462 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
14463 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
14464 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
14465 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
14467 (Dn, Dl, Dr): Accept const as well as const_vector.
14468 (Dz): Likewise. Compare against CONST0_RTX.
14469 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
14470 of "vector" where appropriate.
14471 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
14472 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
14473 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
14474 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
14475 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
14476 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
14477 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
14478 (v_int_equiv): Extend to SVE modes.
14479 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
14481 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
14482 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
14483 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
14484 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
14485 (SVE_COND_FP_CMP): New int iterators.
14486 (perm_hilo): Handle the new unpack unspecs.
14487 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
14489 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
14490 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
14491 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
14492 (aarch64_equality_operator, aarch64_constant_vector_operand)
14493 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
14494 (aarch64_sve_nonimmediate_operand): Likewise.
14495 (aarch64_sve_general_operand): Likewise.
14496 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
14497 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
14498 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
14499 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
14500 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
14501 (aarch64_sve_float_arith_immediate): Likewise.
14502 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
14503 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
14504 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
14505 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
14506 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
14507 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
14508 (aarch64_sve_float_arith_operand): Likewise.
14509 (aarch64_sve_float_arith_with_sub_operand): Likewise.
14510 (aarch64_sve_float_mul_operand): Likewise.
14511 (aarch64_sve_vec_perm_operand): Likewise.
14512 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
14513 (aarch64_mov_operand): Accept const_poly_int and const_vector.
14514 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
14515 as well as const_vector.
14516 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
14517 in file. Use CONST0_RTX and CONSTM1_RTX.
14518 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
14519 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
14520 Use aarch64_simd_imm_zero.
14521 * config/aarch64/aarch64-sve.md: New file.
14522 * config/aarch64/aarch64.md: Include it.
14523 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
14524 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
14525 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
14526 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
14527 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
14528 (sve): New attribute.
14529 (enabled): Disable instructions with the sve attribute unless
14531 (movqi, movhi): Pass CONST_POLY_INT operaneds through
14532 aarch64_expand_mov_immediate.
14533 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
14534 CNT[BHSD] immediates.
14535 (movti): Split CONST_POLY_INT moves into two halves.
14536 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
14537 Split additions that need a temporary here if the destination
14538 is the stack pointer.
14539 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
14540 (*add<mode>3_poly_1): New instruction.
14541 (set_clobber_cc): New expander.
14543 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14545 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
14546 parameter and use it instead of GET_MODE_SIZE (innermode). Use
14547 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
14548 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
14549 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
14550 Change innermode from fixed_mode_size to machine_mode.
14551 (simplify_subreg): Update call accordingly. Handle a constant-sized
14552 subreg of a variable-length CONST_VECTOR.
14554 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14555 Alan Hayward <alan.hayward@arm.com>
14556 David Sherwood <david.sherwood@arm.com>
14558 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
14559 (add_offset_to_base): New function, split out from...
14560 (create_mem_ref): ...here. When handling a scale other than 1,
14561 check first whether the address is valid without the offset.
14562 Add it into the base if so, leaving the index and scale as-is.
14564 2018-01-12 Jakub Jelinek <jakub@redhat.com>
14567 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
14568 fold_for_warn before checking if arg2 is INTEGER_CST.
14570 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
14572 * config/rs6000/predicates.md (load_multiple_operation): Delete.
14573 (store_multiple_operation): Delete.
14574 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
14575 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
14576 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
14577 guarded by TARGET_STRING.
14578 (rs6000_output_load_multiple): Delete.
14579 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
14580 OPTION_MASK_STRING / TARGET_STRING handling.
14581 (print_operand) <'N', 'O'>: Add comment that these are unused now.
14582 (const rs6000_opt_masks) <"string">: Change mask to 0.
14583 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
14584 (MASK_STRING): Delete.
14585 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
14587 (load_multiple): Delete.
14594 (store_multiple): Delete.
14601 (movmemsi_8reg): Delete.
14602 (corresponding unnamed define_insn): Delete.
14603 (movmemsi_6reg): Delete.
14604 (corresponding unnamed define_insn): Delete.
14605 (movmemsi_4reg): Delete.
14606 (corresponding unnamed define_insn): Delete.
14607 (movmemsi_2reg): Delete.
14608 (corresponding unnamed define_insn): Delete.
14609 (movmemsi_1reg): Delete.
14610 (corresponding unnamed define_insn): Delete.
14611 * config/rs6000/rs6000.opt (mno-string): New.
14612 (mstring): Replace by deprecation warning stub.
14613 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
14615 2018-01-12 Jakub Jelinek <jakub@redhat.com>
14617 * regrename.c (regrename_do_replace): If replacing the same
14618 reg multiple times, try to reuse last created gen_raw_REG.
14621 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
14622 main to workaround a bug in GDB.
14624 2018-01-12 Tom de Vries <tom@codesourcery.com>
14627 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
14629 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
14631 PR rtl-optimization/80481
14632 * ira-color.c (get_cap_member): New function.
14633 (allocnos_conflict_by_live_ranges_p): Use it.
14634 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
14635 (setup_slot_coalesced_allocno_live_ranges): Ditto.
14637 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
14640 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
14641 (*saddl_se_1): Ditto.
14642 (*ssubsi_1): Ditto.
14643 (*ssubl_se_1): Ditto.
14645 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
14647 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
14648 rather than wi::to_widest for DR_INITs.
14649 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
14650 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
14651 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
14653 (vect_analyze_group_access_1): Note that here.
14655 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
14657 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
14658 polynomial type sizes.
14660 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
14662 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
14663 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
14664 (gimple_add_tmp_var): Likewise.
14666 2018-01-12 Martin Liska <mliska@suse.cz>
14668 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
14669 (gimple_alloc_sizes): Likewise.
14670 (dump_gimple_statistics): Use PRIu64 in printf format.
14671 * gimple.h: Change uint64_t to int.
14673 2018-01-12 Martin Liska <mliska@suse.cz>
14675 * tree-core.h: Use uint64_t instead of int.
14676 * tree.c (tree_node_counts): Likewise.
14677 (tree_node_sizes): Likewise.
14678 (dump_tree_statistics): Use PRIu64 in printf format.
14680 2018-01-12 Martin Liska <mliska@suse.cz>
14682 * Makefile.in: As qsort_chk is implemented in vec.c, add
14683 vec.o to linkage of gencfn-macros.
14684 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
14685 passing the info to record_node_allocation_statistics.
14686 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
14688 * ggc-common.c (struct ggc_usage): Add operator== and use
14689 it in operator< and compare function.
14690 * mem-stats.h (struct mem_usage): Likewise.
14691 * vec.c (struct vec_usage): Remove operator< and compare
14692 function. Can be simply inherited.
14694 2018-01-12 Martin Jambor <mjambor@suse.cz>
14697 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
14698 * tree-ssa-math-opts.c: Include domwalk.h.
14699 (convert_mult_to_fma_1): New function.
14700 (fma_transformation_info): New type.
14701 (fma_deferring_state): Likewise.
14702 (cancel_fma_deferring): New function.
14703 (result_of_phi): Likewise.
14704 (last_fma_candidate_feeds_initial_phi): Likewise.
14705 (convert_mult_to_fma): Added deferring logic, split actual
14706 transformation to convert_mult_to_fma_1.
14707 (math_opts_dom_walker): New type.
14708 (math_opts_dom_walker::after_dom_children): New method, body moved
14709 here from pass_optimize_widening_mul::execute, added deferring logic
14711 (pass_optimize_widening_mul::execute): Moved most of code to
14712 math_opts_dom_walker::after_dom_children.
14713 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
14714 * config/i386/i386.c (ix86_option_override_internal): Added
14715 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
14717 2018-01-12 Richard Biener <rguenther@suse.de>
14720 * dwarf2out.c (gen_variable_die): Do not reset old_die for
14721 inline instance vars.
14723 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
14726 * config/rx/rx.c (rx_is_restricted_memory_address):
14727 Handle SUBREG case.
14729 2018-01-12 Richard Biener <rguenther@suse.de>
14731 PR tree-optimization/80846
14732 * target.def (split_reduction): New target hook.
14733 * targhooks.c (default_split_reduction): New function.
14734 * targhooks.h (default_split_reduction): Declare.
14735 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
14736 target requests first reduce vectors by combining low and high
14738 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
14739 (get_vectype_for_scalar_type_and_size): Export.
14740 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
14741 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
14742 * doc/tm.texi: Regenerate.
14743 * config/i386/i386.c (ix86_split_reduction): Implement
14744 TARGET_VECTORIZE_SPLIT_REDUCTION.
14746 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
14749 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
14750 in PIC mode except for TARGET_VXWORKS_RTP.
14751 * config/sparc/sparc.c: Include cfgrtl.h.
14752 (TARGET_INIT_PIC_REG): Define.
14753 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
14754 (sparc_pic_register_p): New predicate.
14755 (sparc_legitimate_address_p): Use it.
14756 (sparc_legitimize_pic_address): Likewise.
14757 (sparc_delegitimize_address): Likewise.
14758 (sparc_mode_dependent_address_p): Likewise.
14759 (gen_load_pcrel_sym): Remove 4th parameter.
14760 (load_got_register): Adjust call to above. Remove obsolete stuff.
14761 (sparc_expand_prologue): Do not call load_got_register here.
14762 (sparc_flat_expand_prologue): Likewise.
14763 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
14764 (sparc_use_pseudo_pic_reg): New function.
14765 (sparc_init_pic_reg): Likewise.
14766 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
14767 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
14769 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
14771 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
14772 Add item for branch_cost.
14774 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
14776 PR rtl-optimization/83565
14777 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
14778 not extend the result to a larger mode for rotate operations.
14779 (num_sign_bit_copies1): Likewise.
14781 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
14784 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
14786 Use values-Xc.o for -pedantic.
14787 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
14789 2018-01-12 Martin Liska <mliska@suse.cz>
14792 * ipa-devirt.c (final_warning_record::grow_type_warnings):
14794 (possible_polymorphic_call_targets): Use it.
14795 (ipa_devirt): Likewise.
14797 2018-01-12 Martin Liska <mliska@suse.cz>
14799 * profile-count.h (enum profile_quality): Use 0 as invalid
14800 enum value of profile_quality.
14802 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
14804 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
14805 -mext-string options.
14807 2018-01-12 Richard Biener <rguenther@suse.de>
14809 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
14810 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
14811 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
14813 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
14815 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
14817 * configure.ac (--with-long-double-format): Add support for the
14818 configuration option to change the default long double format on
14820 * config.gcc (powerpc*-linux*-*): Likewise.
14821 * configure: Regenerate.
14822 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
14823 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
14824 used without modification.
14826 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14828 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
14829 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
14830 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
14831 MISC_BUILTIN_SPEC_BARRIER.
14832 (rs6000_init_builtins): Likewise.
14833 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
14835 (speculation_barrier): New define_insn.
14836 * doc/extend.texi: Document __builtin_speculation_barrier.
14838 2018-01-11 Jakub Jelinek <jakub@redhat.com>
14841 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
14842 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
14843 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
14845 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
14846 integral modes instead of "ss" and "sd".
14847 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
14848 vectors with 32-bit and 64-bit elements.
14849 (vecdupssescalarmodesuffix): New mode attribute.
14850 (vec_dup<mode>): Use it.
14852 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
14855 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
14856 frame if argument is passed on stack.
14858 2018-01-11 Jakub Jelinek <jakub@redhat.com>
14861 * ree.c (combine_reaching_defs): Optimize also
14862 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
14863 reg2=any_extend(exp); reg1=reg2;, formatting fix.
14865 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
14867 PR middle-end/83189
14868 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
14870 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
14872 PR middle-end/83718
14873 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
14874 after they are computed.
14876 2018-01-11 Bin Cheng <bin.cheng@arm.com>
14878 PR tree-optimization/83695
14879 * gimple-loop-linterchange.cc
14880 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
14881 reset cached scev information after interchange.
14882 (pass_linterchange::execute): Remove call to scev_reset_htab.
14884 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14886 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
14887 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
14888 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
14889 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
14890 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
14891 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
14892 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
14893 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
14894 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
14895 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
14896 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
14897 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
14898 (V_lane_reg): Likewise.
14899 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
14901 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
14902 (vfmal_lane_low<mode>_intrinsic,
14903 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
14904 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
14905 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
14906 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
14907 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
14908 vfmsl_lane_high<mode>_intrinsic): New define_insns.
14910 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14912 * config/arm/arm-cpus.in (fp16fml): New feature.
14913 (ALL_SIMD): Add fp16fml.
14914 (armv8.2-a): Add fp16fml as an option.
14915 (armv8.3-a): Likewise.
14916 (armv8.4-a): Add fp16fml as part of fp16.
14917 * config/arm/arm.h (TARGET_FP16FML): Define.
14918 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
14920 * config/arm/arm-modes.def (V2HF): Define.
14921 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
14922 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
14923 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
14924 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
14925 vfmsl_low, vfmsl_high): New set of builtins.
14926 * config/arm/iterators.md (PLUSMINUS): New code iterator.
14927 (vfml_op): New code attribute.
14928 (VFMLHALVES): New int iterator.
14929 (VFML, VFMLSEL): New mode attributes.
14930 (V_reg): Define mapping for V2HF.
14931 (V_hi, V_lo): New mode attributes.
14932 (VF_constraint): Likewise.
14933 (vfml_half, vfml_half_selector): New int attributes.
14934 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
14936 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
14937 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
14939 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
14940 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
14941 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
14942 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
14944 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
14945 Document new effective target and option set.
14947 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14949 * config/arm/arm-cpus.in (armv8_4): New feature.
14950 (ARMv8_4a): New fgroup.
14951 (armv8.4-a): New arch.
14952 * config/arm/arm-tables.opt: Regenerate.
14953 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
14954 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
14955 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
14956 Add matching rules for -march=armv8.4-a and extensions.
14957 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
14959 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
14962 * config/rx/rx.md (BW): New mode attribute.
14963 (sync_lock_test_and_setsi): Add mode suffix to insn output.
14965 2018-01-11 Richard Biener <rguenther@suse.de>
14967 PR tree-optimization/83435
14968 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
14969 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
14970 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
14972 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14973 Alan Hayward <alan.hayward@arm.com>
14974 David Sherwood <david.sherwood@arm.com>
14976 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
14978 (aarch64_classify_address): Initialize it. Track polynomial offsets.
14979 (aarch64_print_address_internal): Use it to check for a zero offset.
14981 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14982 Alan Hayward <alan.hayward@arm.com>
14983 David Sherwood <david.sherwood@arm.com>
14985 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
14986 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
14987 Return a poly_int64 rather than a HOST_WIDE_INT.
14988 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
14989 rather than a HOST_WIDE_INT.
14990 * config/aarch64/aarch64.h (aarch64_frame): Protect with
14991 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
14992 hard_fp_offset, frame_size, initial_adjust, callee_offset and
14993 final_offset from HOST_WIDE_INT to poly_int64.
14994 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
14995 to_constant when getting the number of units in an Advanced SIMD
14997 (aarch64_builtin_vectorized_function): Check for a constant number
14999 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
15001 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
15002 attribute instead of GET_MODE_NUNITS.
15003 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
15004 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
15005 GET_MODE_SIZE for fixed-size registers.
15006 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
15007 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
15008 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
15009 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
15010 (aarch64_print_operand, aarch64_print_address_internal)
15011 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
15012 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
15013 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
15014 Handle polynomial GET_MODE_SIZE.
15015 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
15016 wider than SImode without modification.
15017 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
15018 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
15019 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
15020 passing and returning SVE modes.
15021 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
15022 rather than GEN_INT.
15023 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
15024 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
15025 (aarch64_allocate_and_probe_stack_space): Likewise.
15026 (aarch64_layout_frame): Cope with polynomial offsets.
15027 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
15028 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
15029 polynomial offsets.
15030 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
15031 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
15032 poly_int64 rather than a HOST_WIDE_INT.
15033 (aarch64_get_separate_components, aarch64_process_components)
15034 (aarch64_expand_prologue, aarch64_expand_epilogue)
15035 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
15036 (aarch64_anchor_offset): New function, split out from...
15037 (aarch64_legitimize_address): ...here.
15038 (aarch64_builtin_vectorization_cost): Handle polynomial
15039 TYPE_VECTOR_SUBPARTS.
15040 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
15042 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
15043 number of elements from the PARALLEL rather than the mode.
15044 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
15045 rather than GET_MODE_BITSIZE.
15046 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
15047 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
15048 (aarch64_expand_vec_perm_const_1): Handle polynomial
15049 d->perm.length () and d->perm elements.
15050 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
15051 Apply to_constant to d->perm elements.
15052 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
15053 polynomial CONST_VECTOR_NUNITS.
15054 (aarch64_move_pointer): Take amount as a poly_int64 rather
15056 (aarch64_progress_pointer): Avoid temporary variable.
15057 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
15058 the mode attribute instead of GET_MODE.
15060 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
15061 Alan Hayward <alan.hayward@arm.com>
15062 David Sherwood <david.sherwood@arm.com>
15064 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
15065 x exists before using it.
15066 (aarch64_add_constant_internal): Rename to...
15067 (aarch64_add_offset_1): ...this. Replace regnum with separate
15068 src and dest rtxes. Handle the case in which they're different,
15069 including when the offset is zero. Replace scratchreg with an rtx.
15070 Use 2 additions if there is no spare register into which we can
15071 move a 16-bit constant.
15072 (aarch64_add_constant): Delete.
15073 (aarch64_add_offset): Replace reg with separate src and dest
15074 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
15075 Use aarch64_add_offset_1.
15076 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
15077 an rtx rather than an int. Take the delta as a poly_int64
15078 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
15079 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
15080 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
15081 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
15082 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
15083 and aarch64_add_sp.
15084 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
15085 aarch64_add_constant.
15087 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
15089 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
15090 Use scalar_float_mode.
15092 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
15094 * config/aarch64/aarch64-simd.md
15095 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
15096 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
15097 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
15098 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
15099 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
15100 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
15101 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
15102 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
15103 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
15104 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
15106 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
15109 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
15110 targ_options->x_arm_arch_string is non NULL.
15112 2018-01-11 Tamar Christina <tamar.christina@arm.com>
15114 * config/aarch64/aarch64.h
15115 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
15117 2018-01-11 Sudakshina Das <sudi.das@arm.com>
15120 * expmed.c (emit_store_flag_force): Swap if const op0
15121 and change VOIDmode to mode of op0.
15123 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
15125 PR rtl-optimization/83761
15126 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
15127 than bytes to mode_for_size.
15129 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
15131 PR middle-end/83189
15132 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
15133 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
15136 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
15138 PR middle-end/83575
15139 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
15140 when in layout mode.
15141 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
15142 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
15145 2018-01-10 Michael Collison <michael.collison@arm.com>
15147 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
15148 * config/aarch64/aarch64-option-extension.def: Add
15149 AARCH64_OPT_EXTENSION of 'fp16fml'.
15150 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
15151 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
15152 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
15153 * config/aarch64/constraints.md (Ui7): New constraint.
15154 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
15155 (VFMLA_SEL_W): Ditto.
15158 (VFMLA16_LOW): New int iterator.
15159 (VFMLA16_HIGH): Ditto.
15160 (UNSPEC_FMLAL): New unspec.
15161 (UNSPEC_FMLSL): Ditto.
15162 (UNSPEC_FMLAL2): Ditto.
15163 (UNSPEC_FMLSL2): Ditto.
15164 (f16mac): New code attribute.
15165 * config/aarch64/aarch64-simd-builtins.def
15166 (aarch64_fmlal_lowv2sf): Ditto.
15167 (aarch64_fmlsl_lowv2sf): Ditto.
15168 (aarch64_fmlalq_lowv4sf): Ditto.
15169 (aarch64_fmlslq_lowv4sf): Ditto.
15170 (aarch64_fmlal_highv2sf): Ditto.
15171 (aarch64_fmlsl_highv2sf): Ditto.
15172 (aarch64_fmlalq_highv4sf): Ditto.
15173 (aarch64_fmlslq_highv4sf): Ditto.
15174 (aarch64_fmlal_lane_lowv2sf): Ditto.
15175 (aarch64_fmlsl_lane_lowv2sf): Ditto.
15176 (aarch64_fmlal_laneq_lowv2sf): Ditto.
15177 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
15178 (aarch64_fmlalq_lane_lowv4sf): Ditto.
15179 (aarch64_fmlsl_lane_lowv4sf): Ditto.
15180 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
15181 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
15182 (aarch64_fmlal_lane_highv2sf): Ditto.
15183 (aarch64_fmlsl_lane_highv2sf): Ditto.
15184 (aarch64_fmlal_laneq_highv2sf): Ditto.
15185 (aarch64_fmlsl_laneq_highv2sf): Ditto.
15186 (aarch64_fmlalq_lane_highv4sf): Ditto.
15187 (aarch64_fmlsl_lane_highv4sf): Ditto.
15188 (aarch64_fmlalq_laneq_highv4sf): Ditto.
15189 (aarch64_fmlsl_laneq_highv4sf): Ditto.
15190 * config/aarch64/aarch64-simd.md:
15191 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
15192 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
15193 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
15194 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
15195 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
15196 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
15197 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
15198 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
15199 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
15200 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
15201 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
15202 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
15203 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
15204 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
15205 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
15206 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
15207 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
15208 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
15209 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
15210 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
15211 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
15212 (vfmlsl_low_u32): Ditto.
15213 (vfmlalq_low_u32): Ditto.
15214 (vfmlslq_low_u32): Ditto.
15215 (vfmlal_high_u32): Ditto.
15216 (vfmlsl_high_u32): Ditto.
15217 (vfmlalq_high_u32): Ditto.
15218 (vfmlslq_high_u32): Ditto.
15219 (vfmlal_lane_low_u32): Ditto.
15220 (vfmlsl_lane_low_u32): Ditto.
15221 (vfmlal_laneq_low_u32): Ditto.
15222 (vfmlsl_laneq_low_u32): Ditto.
15223 (vfmlalq_lane_low_u32): Ditto.
15224 (vfmlslq_lane_low_u32): Ditto.
15225 (vfmlalq_laneq_low_u32): Ditto.
15226 (vfmlslq_laneq_low_u32): Ditto.
15227 (vfmlal_lane_high_u32): Ditto.
15228 (vfmlsl_lane_high_u32): Ditto.
15229 (vfmlal_laneq_high_u32): Ditto.
15230 (vfmlsl_laneq_high_u32): Ditto.
15231 (vfmlalq_lane_high_u32): Ditto.
15232 (vfmlslq_lane_high_u32): Ditto.
15233 (vfmlalq_laneq_high_u32): Ditto.
15234 (vfmlslq_laneq_high_u32): Ditto.
15235 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
15236 (AARCH64_FL_FOR_ARCH8_4): New.
15237 (AARCH64_ISA_F16FML): New ISA flag.
15238 (TARGET_F16FML): New feature flag for fp16fml.
15239 (doc/invoke.texi): Document new fp16fml option.
15241 2018-01-10 Michael Collison <michael.collison@arm.com>
15243 * config/aarch64/aarch64-builtins.c:
15244 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
15245 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
15246 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
15247 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
15248 (AARCH64_ISA_SHA3): New ISA flag.
15249 (TARGET_SHA3): New feature flag for sha3.
15250 * config/aarch64/iterators.md (sha512_op): New int attribute.
15251 (CRYPTO_SHA512): New int iterator.
15252 (UNSPEC_SHA512H): New unspec.
15253 (UNSPEC_SHA512H2): Ditto.
15254 (UNSPEC_SHA512SU0): Ditto.
15255 (UNSPEC_SHA512SU1): Ditto.
15256 * config/aarch64/aarch64-simd-builtins.def
15257 (aarch64_crypto_sha512hqv2di): New builtin.
15258 (aarch64_crypto_sha512h2qv2di): Ditto.
15259 (aarch64_crypto_sha512su0qv2di): Ditto.
15260 (aarch64_crypto_sha512su1qv2di): Ditto.
15261 (aarch64_eor3qv8hi): Ditto.
15262 (aarch64_rax1qv2di): Ditto.
15263 (aarch64_xarqv2di): Ditto.
15264 (aarch64_bcaxqv8hi): Ditto.
15265 * config/aarch64/aarch64-simd.md:
15266 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
15267 (aarch64_crypto_sha512su0qv2di): Ditto.
15268 (aarch64_crypto_sha512su1qv2di): Ditto.
15269 (aarch64_eor3qv8hi): Ditto.
15270 (aarch64_rax1qv2di): Ditto.
15271 (aarch64_xarqv2di): Ditto.
15272 (aarch64_bcaxqv8hi): Ditto.
15273 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
15274 (vsha512h2q_u64): Ditto.
15275 (vsha512su0q_u64): Ditto.
15276 (vsha512su1q_u64): Ditto.
15277 (veor3q_u16): Ditto.
15278 (vrax1q_u64): Ditto.
15279 (vxarq_u64): Ditto.
15280 (vbcaxq_u16): Ditto.
15281 * config/arm/types.md (crypto_sha512): New type attribute.
15282 (crypto_sha3): Ditto.
15283 (doc/invoke.texi): Document new sha3 option.
15285 2018-01-10 Michael Collison <michael.collison@arm.com>
15287 * config/aarch64/aarch64-builtins.c:
15288 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
15289 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
15290 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
15291 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
15292 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
15293 (AARCH64_ISA_SM4): New ISA flag.
15294 (TARGET_SM4): New feature flag for sm4.
15295 * config/aarch64/aarch64-simd-builtins.def
15296 (aarch64_sm3ss1qv4si): Ditto.
15297 (aarch64_sm3tt1aq4si): Ditto.
15298 (aarch64_sm3tt1bq4si): Ditto.
15299 (aarch64_sm3tt2aq4si): Ditto.
15300 (aarch64_sm3tt2bq4si): Ditto.
15301 (aarch64_sm3partw1qv4si): Ditto.
15302 (aarch64_sm3partw2qv4si): Ditto.
15303 (aarch64_sm4eqv4si): Ditto.
15304 (aarch64_sm4ekeyqv4si): Ditto.
15305 * config/aarch64/aarch64-simd.md:
15306 (aarch64_sm3ss1qv4si): Ditto.
15307 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
15308 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
15309 (aarch64_sm4eqv4si): Ditto.
15310 (aarch64_sm4ekeyqv4si): Ditto.
15311 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
15312 (sm3part_op): Ditto.
15313 (CRYPTO_SM3TT): Ditto.
15314 (CRYPTO_SM3PART): Ditto.
15315 (UNSPEC_SM3SS1): New unspec.
15316 (UNSPEC_SM3TT1A): Ditto.
15317 (UNSPEC_SM3TT1B): Ditto.
15318 (UNSPEC_SM3TT2A): Ditto.
15319 (UNSPEC_SM3TT2B): Ditto.
15320 (UNSPEC_SM3PARTW1): Ditto.
15321 (UNSPEC_SM3PARTW2): Ditto.
15322 (UNSPEC_SM4E): Ditto.
15323 (UNSPEC_SM4EKEY): Ditto.
15324 * config/aarch64/constraints.md (Ui2): New constraint.
15325 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
15326 * config/arm/types.md (crypto_sm3): New type attribute.
15327 (crypto_sm4): Ditto.
15328 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
15329 (vsm3tt1aq_u32): Ditto.
15330 (vsm3tt1bq_u32): Ditto.
15331 (vsm3tt2aq_u32): Ditto.
15332 (vsm3tt2bq_u32): Ditto.
15333 (vsm3partw1q_u32): Ditto.
15334 (vsm3partw2q_u32): Ditto.
15335 (vsm4eq_u32): Ditto.
15336 (vsm4ekeyq_u32): Ditto.
15337 (doc/invoke.texi): Document new sm4 option.
15339 2018-01-10 Michael Collison <michael.collison@arm.com>
15341 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
15342 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
15343 (AARCH64_FL_FOR_ARCH8_4): New.
15344 (AARCH64_FL_V8_4): New flag.
15345 (doc/invoke.texi): Document new armv8.4-a option.
15347 2018-01-10 Michael Collison <michael.collison@arm.com>
15349 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
15350 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
15351 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
15352 * config/aarch64/aarch64-option-extension.def: Add
15353 AARCH64_OPT_EXTENSION of 'sha2'.
15354 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
15355 (crypto): Disable sha2 and aes if crypto disabled.
15356 (crypto): Enable aes and sha2 if enabled.
15357 (simd): Disable sha2 and aes if simd disabled.
15358 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
15360 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
15361 (TARGET_SHA2): New feature flag for sha2.
15362 (TARGET_AES): New feature flag for aes.
15363 * config/aarch64/aarch64-simd.md:
15364 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
15365 conditional on TARGET_AES.
15366 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
15367 (aarch64_crypto_sha1hsi): Make pattern conditional
15369 (aarch64_crypto_sha1hv4si): Ditto.
15370 (aarch64_be_crypto_sha1hv4si): Ditto.
15371 (aarch64_crypto_sha1su1v4si): Ditto.
15372 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
15373 (aarch64_crypto_sha1su0v4si): Ditto.
15374 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
15375 (aarch64_crypto_sha256su0v4si): Ditto.
15376 (aarch64_crypto_sha256su1v4si): Ditto.
15377 (doc/invoke.texi): Document new aes and sha2 options.
15379 2018-01-10 Martin Sebor <msebor@redhat.com>
15381 PR tree-optimization/83781
15382 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
15385 2018-01-11 Martin Sebor <msebor@gmail.com>
15386 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
15388 PR tree-optimization/83501
15389 PR tree-optimization/81703
15391 * tree-ssa-strlen.c (get_string_cst): Rename...
15392 (get_string_len): ...to this. Handle global constants.
15393 (handle_char_store): Adjust.
15395 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
15396 Jim Wilson <jimw@sifive.com>
15398 * config/riscv/riscv-protos.h (riscv_output_return): New.
15399 * config/riscv/riscv.c (struct machine_function): New naked_p field.
15400 (riscv_attribute_table, riscv_output_return),
15401 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
15402 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
15403 (riscv_compute_frame_info): Only compute frame->mask if not a naked
15405 (riscv_expand_prologue): Add early return for naked function.
15406 (riscv_expand_epilogue): Likewise.
15407 (riscv_function_ok_for_sibcall): Return false for naked function.
15408 (riscv_set_current_function): New.
15409 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
15410 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
15411 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
15412 * doc/extend.texi (RISC-V Function Attributes): New.
15414 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
15416 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
15417 check for 128-bit long double before checking TCmode.
15418 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
15419 128-bit long doubles before checking TFmode or TCmode.
15420 (FLOAT128_IBM_P): Likewise.
15422 2018-01-10 Martin Sebor <msebor@redhat.com>
15424 PR tree-optimization/83671
15425 * builtins.c (c_strlen): Unconditionally return zero for the empty
15427 Use -Warray-bounds for warnings.
15428 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
15429 for non-constant array indices with COMPONENT_REF, arrays of
15430 arrays, and pointers to arrays.
15431 (gimple_fold_builtin_strlen): Determine and set length range for
15432 non-constant character arrays.
15434 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
15436 PR middle-end/81897
15437 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
15440 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
15442 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
15444 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
15447 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
15448 VECTOR_MEM_ALTIVEC_OR_VSX_P.
15449 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
15450 indexed_or_indirect_operand predicate.
15451 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
15452 (*vsx_le_perm_load_v8hi): Likewise.
15453 (*vsx_le_perm_load_v16qi): Likewise.
15454 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
15455 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
15456 (*vsx_le_perm_store_v8hi): Likewise.
15457 (*vsx_le_perm_store_v16qi): Likewise.
15458 (eight unnamed splitters): Likewise.
15460 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
15462 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
15463 * config/rs6000/emmintrin.h: Likewise.
15464 * config/rs6000/mmintrin.h: Likewise.
15465 * config/rs6000/xmmintrin.h: Likewise.
15467 2018-01-10 David Malcolm <dmalcolm@redhat.com>
15470 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
15472 * tree.c (tree_nop_conversion): Return true for location wrapper
15474 (maybe_wrap_with_location): New function.
15475 (selftest::check_strip_nops): New function.
15476 (selftest::test_location_wrappers): New function.
15477 (selftest::tree_c_tests): Call it.
15478 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
15479 (maybe_wrap_with_location): New decl.
15480 (EXPR_LOCATION_WRAPPER_P): New macro.
15481 (location_wrapper_p): New inline function.
15482 (tree_strip_any_location_wrapper): New inline function.
15484 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
15487 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
15488 stack_realign_offset for the largest alignment of stack slot
15490 (ix86_find_max_used_stack_alignment): New function.
15491 (ix86_finalize_stack_frame_flags): Use it. Set
15492 max_used_stack_alignment if we don't realign stack.
15493 * config/i386/i386.h (machine_function): Add
15494 max_used_stack_alignment.
15496 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
15498 * config/arm/arm.opt (-mbranch-cost): New option.
15499 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
15502 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
15505 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
15506 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
15508 2018-01-10 Richard Biener <rguenther@suse.de>
15511 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
15512 early out so it also covers the case where we have a non-NULL
15515 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
15517 PR tree-optimization/83753
15518 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
15519 for non-strided grouped accesses if the number of elements is 1.
15521 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
15524 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
15525 * i386.h (TARGET_USE_GATHER): Define.
15526 * x86-tune.def (X86_TUNE_USE_GATHER): New.
15528 2018-01-10 Martin Liska <mliska@suse.cz>
15531 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
15532 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
15534 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
15535 CLEANUP_NO_PARTITIONING is not set.
15537 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
15539 * doc/rtl.texi: Remove documentation of (const ...) wrappers
15540 for vectors, as a partial revert of r254296.
15541 * rtl.h (const_vec_p): Delete.
15542 (const_vec_duplicate_p): Don't test for vector CONSTs.
15543 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
15544 * expmed.c (make_tree): Likewise.
15547 * common.md (E, F): Use CONSTANT_P instead of checking for
15549 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
15550 checking for CONST_VECTOR.
15552 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
15554 PR middle-end/83575
15555 * predict.c (force_edge_cold): Handle in more sane way edges
15556 with no prediction.
15558 2018-01-09 Carl Love <cel@us.ibm.com>
15560 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
15562 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
15563 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
15564 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
15565 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
15566 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
15567 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
15568 * config/rs6000/rs6000-protos.h: Add extern defition for
15569 rs6000_generate_float2_double_code.
15570 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
15572 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
15573 (float2_v2df): Add define_expand.
15575 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
15578 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
15579 op_mode in the force_to_mode call.
15581 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
15583 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
15584 instead of checking each element individually.
15585 (aarch64_evpc_uzp): Likewise.
15586 (aarch64_evpc_zip): Likewise.
15587 (aarch64_evpc_ext): Likewise.
15588 (aarch64_evpc_rev): Likewise.
15589 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
15590 instead of checking each element individually. Return true without
15592 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
15593 whether all selected elements come from the same input, instead of
15594 checking each element individually. Remove calls to gen_rtx_REG,
15595 start_sequence and end_sequence and instead assert that no rtl is
15598 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
15600 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
15601 order of HIGH and CONST checks.
15603 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
15605 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
15606 if the destination isn't an SSA_NAME.
15608 2018-01-09 Richard Biener <rguenther@suse.de>
15610 PR tree-optimization/83668
15611 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
15613 (canonicalize_loop_form): ... here, renamed from ...
15614 (canonicalize_loop_closed_ssa_form): ... this and amended to
15615 swap successor edges for loop exit blocks to make us use
15616 the RPO order we need for initial schedule generation.
15618 2018-01-09 Joseph Myers <joseph@codesourcery.com>
15620 PR tree-optimization/64811
15621 * match.pd: When optimizing comparisons with Inf, avoid
15622 introducing or losing exceptions from comparisons with NaN.
15624 2018-01-09 Martin Liska <mliska@suse.cz>
15627 * asan.c (shadow_mem_size): Add gcc_assert.
15629 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
15631 Don't save registers in main().
15634 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
15635 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
15636 * config/avr/avr.c (avr_set_current_function): Don't error if
15637 naked, OS_task or OS_main are specified at the same time.
15638 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
15640 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
15642 * common/config/avr/avr-common.c (avr_option_optimization_table):
15643 Switch on -mmain-is-OS_task for optimizing compilations.
15645 2018-01-09 Richard Biener <rguenther@suse.de>
15647 PR tree-optimization/83572
15648 * graphite.c: Include cfganal.h.
15649 (graphite_transform_loops): Connect infinite loops to exit
15650 and remove fake edges at the end.
15652 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
15654 * ipa-inline.c (edge_badness): Revert accidental checkin.
15656 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
15659 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
15660 symbols; not inline clones.
15662 2018-01-09 Jakub Jelinek <jakub@redhat.com>
15665 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
15666 hard registers. Formatting fixes.
15668 PR preprocessor/83722
15669 * gcc.c (try_generate_repro): Pass
15670 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
15671 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
15674 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
15675 Kito Cheng <kito.cheng@gmail.com>
15677 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
15678 (riscv_leaf_function_p): Delete.
15679 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
15681 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
15683 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
15685 (do_ifelse): New function.
15686 (do_isel): New function.
15687 (do_sub3): New function.
15688 (do_add3): New function.
15689 (do_load_mask_compare): New function.
15690 (do_overlap_load_compare): New function.
15691 (expand_compare_loop): New function.
15692 (expand_block_compare): Call expand_compare_loop() when appropriate.
15693 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
15694 option description.
15695 (-mblock-compare-inline-loop-limit): New option.
15697 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
15700 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
15701 Reverse order of second and third operands in first alternative.
15702 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
15703 of first and second elements in UNSPEC_VPERMR vector.
15704 (altivec_expand_vec_perm_le): Likewise.
15706 2018-01-08 Jeff Law <law@redhat.com>
15708 PR rtl-optimizatin/81308
15709 * tree-switch-conversion.c (cfg_altered): New file scoped static.
15710 (process_switch): If group_case_labels makes a change, then set
15712 (pass_convert_switch::execute): If a switch is converted, then
15713 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
15715 PR rtl-optimization/81308
15716 * recog.c (split_all_insns): Conditionally cleanup the CFG after
15719 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
15721 PR target/83663 - Revert r255946
15722 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
15723 generation for cases where splatting a value is not useful.
15724 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
15725 across a vec_duplicate and a paradoxical subreg forming a vector
15726 mode to a vec_concat.
15728 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15730 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
15731 -march=armv8.3-a variants.
15732 * config/arm/t-multilib: Likewise.
15733 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
15735 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
15737 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
15739 (cceq_ior_compare_complement): Give it a name so I can use it, and
15740 change boolean_or_operator predicate to boolean_operator so it can
15741 be used to generate a crand.
15742 (eqne): New code iterator.
15743 (bd/bd_neg): New code_attrs.
15744 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
15745 a single define_insn.
15746 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
15747 decrement (bdnzt/bdnzf/bdzt/bdzf).
15748 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
15749 with the new names of the branch decrement patterns, and added the
15750 names of the branch decrement conditional patterns.
15752 2018-01-08 Richard Biener <rguenther@suse.de>
15754 PR tree-optimization/83563
15755 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
15758 2018-01-08 Richard Biener <rguenther@suse.de>
15760 PR middle-end/83713
15761 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
15763 2018-01-08 Richard Biener <rguenther@suse.de>
15765 PR tree-optimization/83685
15766 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
15767 references to abnormals.
15769 2018-01-08 Richard Biener <rguenther@suse.de>
15772 * dwarf2out.c (output_indirect_strings): Handle empty
15773 skeleton_debug_str_hash.
15774 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
15776 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
15778 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
15779 (emit_store_direct): Likewise.
15780 (arc_trampoline_adjust_address): Likewise.
15781 (arc_asm_trampoline_template): New function.
15782 (arc_initialize_trampoline): Use asm_trampoline_template.
15783 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
15784 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
15785 * config/arc/arc.md (flush_icache): Delete pattern.
15787 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
15789 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
15790 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
15793 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
15796 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
15797 by not USED_FOR_TARGET.
15798 (make_pass_resolve_sw_modes): Likewise.
15800 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
15802 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
15805 2018-01-08 Richard Biener <rguenther@suse.de>
15807 PR middle-end/83580
15808 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
15810 2018-01-08 Richard Biener <rguenther@suse.de>
15812 PR middle-end/83517
15813 * match.pd ((t * 2) / 2) -> t): Add missing :c.
15815 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
15817 PR middle-end/81897
15818 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
15819 basic blocks with a small number of successors.
15820 (convert_control_dep_chain_into_preds): Improve handling of
15822 (dump_predicates): Split apart into...
15823 (dump_pred_chain): ...here...
15824 (dump_pred_info): ...and here.
15825 (can_one_predicate_be_invalidated_p): Add debugging printfs.
15826 (can_chain_union_be_invalidated_p): Improve check for invalidation
15828 (uninit_uses_cannot_happen): Avoid unnecessary if
15829 convert_control_dep_chain_into_preds yielded nothing.
15831 2018-01-06 Martin Sebor <msebor@redhat.com>
15833 PR tree-optimization/83640
15834 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
15835 subtracting negative offset from size.
15836 (builtin_access::overlap): Adjust offset bounds of the access to fall
15837 within the size of the object if possible.
15839 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
15841 PR rtl-optimization/83699
15842 * expmed.c (extract_bit_field_1): Restrict the vector usage of
15843 extract_bit_field_as_subreg to cases in which the extracted
15844 value is also a vector.
15846 * lra-constraints.c (process_alt_operands): Test for the equivalence
15847 substitutions when detecting a possible reload cycle.
15849 2018-01-06 Jakub Jelinek <jakub@redhat.com>
15852 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
15853 by default if flag_selective_schedling{,2}. Formatting fixes.
15855 PR rtl-optimization/83682
15856 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
15857 if it has non-VECTOR_MODE element mode.
15858 (vec_duplicate_p): Likewise.
15860 PR middle-end/83694
15861 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
15862 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
15864 2018-01-05 Jakub Jelinek <jakub@redhat.com>
15867 * config/i386/i386-builtin.def
15868 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
15869 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
15870 Require also OPTION_MASK_ISA_AVX512F in addition to
15871 OPTION_MASK_ISA_GFNI.
15872 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
15873 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
15874 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
15875 to OPTION_MASK_ISA_GFNI.
15876 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
15877 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
15878 OPTION_MASK_ISA_AVX512BW.
15879 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
15880 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
15881 addition to OPTION_MASK_ISA_GFNI.
15882 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
15883 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
15884 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
15885 to OPTION_MASK_ISA_GFNI.
15886 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
15887 a requirement for all ISAs rather than any of them with a few
15889 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
15891 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
15892 bitmasks to be enabled with 3 exceptions, instead of requiring any
15893 enabled ISA with lots of exceptions.
15894 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
15895 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
15896 Change avx512bw in isa attribute to avx512f.
15897 * config/i386/sgxintrin.h: Add license boilerplate.
15898 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
15899 to __AVX512F__ and __AVX512VL to __AVX512VL__.
15900 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
15901 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
15903 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
15904 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
15905 temporarily sse2 rather than sse if not enabled already.
15908 * config/i386/sse.md (VI248_VLBW): Rename to ...
15909 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
15910 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
15911 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
15912 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
15913 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
15914 mode iterator instead of VI248_VLBW.
15916 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
15918 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
15919 (record_modified): Skip clobbers; add debug output.
15920 (param_change_prob): Use sreal frequencies.
15922 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
15924 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
15925 punt for user-aligned variables.
15927 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
15929 * tree-chrec.c (chrec_contains_symbols): Return true for
15932 2018-01-05 Sudakshina Das <sudi.das@arm.com>
15935 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
15936 of (x|y) == x for BICS pattern.
15938 2018-01-05 Jakub Jelinek <jakub@redhat.com>
15940 PR tree-optimization/83605
15941 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
15942 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
15945 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
15947 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
15948 * config/epiphany/rtems.h: New file.
15950 2018-01-04 Jakub Jelinek <jakub@redhat.com>
15951 Uros Bizjak <ubizjak@gmail.com>
15954 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
15955 QIreg_operand instead of register_operand predicate.
15956 * config/i386/i386.c (ix86_rop_should_change_byte_p,
15957 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
15958 comments instead of -fmitigate[-_]rop.
15960 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
15963 * cgraphunit.c (symbol_table::compile): Switch to text_section
15964 before calling assembly_start debug hook.
15965 * run-rtl-passes.c (run_rtl_passes): Likewise.
15968 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15970 * tree-vrp.c (extract_range_from_binary_expr_1): Check
15971 range_int_cst_p rather than !symbolic_range_p before calling
15972 extract_range_from_multiplicative_op_1.
15974 2018-01-04 Jeff Law <law@redhat.com>
15976 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
15977 redundant test in assertion.
15979 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15981 * doc/rtl.texi: Document machine_mode wrapper classes.
15983 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15985 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
15986 using tree_to_uhwi.
15988 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15990 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
15991 the VEC_PERM_EXPR fold to fail.
15993 2018-01-04 Jakub Jelinek <jakub@redhat.com>
15996 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
15997 to switched_sections.
15999 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
16002 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
16003 test for d.testing.
16005 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
16008 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
16009 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
16011 2018-01-04 Jakub Jelinek <jakub@redhat.com>
16014 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
16015 is BLKmode and bitpos not zero or mode change is needed.
16017 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
16020 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
16023 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
16026 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
16027 instead of MULT rtx. Update all corresponding splitters.
16028 (*saddl_se): Ditto.
16029 (*ssub<modesuffix>): Ditto.
16030 (*ssubl_se): Ditto.
16031 (*cmp_sadd_di): Update split patterns.
16032 (*cmp_sadd_si): Ditto.
16033 (*cmp_sadd_sidi): Ditto.
16034 (*cmp_ssub_di): Ditto.
16035 (*cmp_ssub_si): Ditto.
16036 (*cmp_ssub_sidi): Ditto.
16037 * config/alpha/predicates.md (const23_operand): New predicate.
16038 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
16039 Look for ASHIFT, not MULT inner operand.
16040 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
16042 2018-01-04 Martin Liska <mliska@suse.cz>
16044 PR gcov-profile/83669
16045 * gcov.c (output_intermediate_file): Add version to intermediate
16047 * doc/gcov.texi: Document new field 'version' in intermediate
16048 file format. Fix location of '-k' option of gcov command.
16050 2018-01-04 Martin Liska <mliska@suse.cz>
16053 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
16055 2018-01-04 Jakub Jelinek <jakub@redhat.com>
16057 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
16059 2018-01-03 Martin Sebor <msebor@redhat.com>
16061 PR tree-optimization/83655
16062 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
16063 checking calls with invalid arguments.
16065 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16067 * tree-vect-stmts.c (vect_get_store_rhs): New function.
16068 (vectorizable_mask_load_store): Delete.
16069 (vectorizable_call): Return false for masked loads and stores.
16070 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
16071 instead of gimple_assign_rhs1.
16072 (vectorizable_load): Handle IFN_MASK_LOAD.
16073 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
16075 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16077 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
16079 (vectorizable_mask_load_store): ...here.
16080 (vectorizable_load): ...and here.
16082 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16084 * tree-vect-stmts.c (vect_build_all_ones_mask)
16085 (vect_build_zero_merge_argument): New functions, split out from...
16086 (vectorizable_load): ...here.
16088 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16090 * tree-vect-stmts.c (vect_check_store_rhs): New function,
16092 (vectorizable_mask_load_store): ...here.
16093 (vectorizable_store): ...and here.
16095 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16097 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
16099 (vectorizable_mask_load_store): ...here.
16101 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16103 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
16104 (vect_model_store_cost): Take a vec_load_store_type instead of a
16106 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
16107 (vect_model_store_cost): Take a vec_load_store_type instead of a
16109 (vectorizable_mask_load_store): Update accordingly.
16110 (vectorizable_store): Likewise.
16111 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
16113 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16115 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
16116 IFN_MASK_LOAD calls here rather than...
16117 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
16119 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16120 Alan Hayward <alan.hayward@arm.com>
16121 David Sherwood <david.sherwood@arm.com>
16123 * expmed.c (extract_bit_field_1): For vector extracts,
16124 fall back to extract_bit_field_as_subreg if vec_extract
16127 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16128 Alan Hayward <alan.hayward@arm.com>
16129 David Sherwood <david.sherwood@arm.com>
16131 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
16132 they are variable or constant sized.
16133 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
16134 slots for constant-sized data.
16136 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16137 Alan Hayward <alan.hayward@arm.com>
16138 David Sherwood <david.sherwood@arm.com>
16140 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
16141 handling COND_EXPRs with boolean comparisons, try to find a better
16142 basis for the mask type than the boolean itself.
16144 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16146 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
16147 is calculated and how it can be overridden.
16148 * genmodes.c (max_bitsize_mode_any_mode): New variable.
16149 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
16151 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
16154 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16155 Alan Hayward <alan.hayward@arm.com>
16156 David Sherwood <david.sherwood@arm.com>
16158 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
16159 Remove the mode argument.
16160 (aarch64_simd_valid_immediate): Remove the mode and inverse
16162 * config/aarch64/iterators.md (bitsize): New iterator.
16163 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
16164 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
16165 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
16166 aarch64_simd_valid_immediate.
16167 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
16168 (aarch64_reg_or_bic_imm): Likewise.
16169 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
16170 with an insn_type enum and msl with a modifier_type enum.
16171 Replace element_width with a scalar_mode. Change the shift
16172 to unsigned int. Add constructors for scalar_float_mode and
16173 scalar_int_mode elements.
16174 (aarch64_vect_float_const_representable_p): Delete.
16175 (aarch64_can_const_movi_rtx_p)
16176 (aarch64_simd_scalar_immediate_valid_for_move)
16177 (aarch64_simd_make_constant): Update call to
16178 aarch64_simd_valid_immediate.
16179 (aarch64_advsimd_valid_immediate_hs): New function.
16180 (aarch64_advsimd_valid_immediate): Likewise.
16181 (aarch64_simd_valid_immediate): Remove mode and inverse
16182 arguments. Rewrite to use the above. Use const_vec_duplicate_p
16183 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
16184 and aarch64_float_const_representable_p on the result.
16185 (aarch64_output_simd_mov_immediate): Remove mode argument.
16186 Update call to aarch64_simd_valid_immediate and use of
16187 simd_immediate_info.
16188 (aarch64_output_scalar_simd_mov_immediate): Update call
16191 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16192 Alan Hayward <alan.hayward@arm.com>
16193 David Sherwood <david.sherwood@arm.com>
16195 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
16196 (mode_nunits): Likewise CONST_MODE_NUNITS.
16197 * machmode.def (ADJUST_NUNITS): Document.
16198 * genmodes.c (mode_data::need_nunits_adj): New field.
16199 (blank_mode): Update accordingly.
16200 (adj_nunits): New variable.
16201 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
16203 (emit_mode_size_inline): Set need_bytesize_adj for all modes
16204 listed in adj_nunits.
16205 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
16206 listed in adj_nunits. Don't emit case statements for such modes.
16207 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
16208 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
16209 nothing if adj_nunits is nonnull.
16210 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
16211 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
16212 (emit_mode_fbit): Update use of print_maybe_const_decl.
16213 (emit_move_size): Likewise. Treat the array as non-const
16215 (emit_mode_adjustments): Handle adj_nunits.
16217 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16219 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
16220 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
16221 (VECTOR_MODES): Use it.
16222 (make_vector_modes): Take the prefix as an argument.
16224 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16225 Alan Hayward <alan.hayward@arm.com>
16226 David Sherwood <david.sherwood@arm.com>
16228 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
16229 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
16230 for MODE_VECTOR_BOOL.
16231 * machmode.def (VECTOR_BOOL_MODE): Document.
16232 * genmodes.c (VECTOR_BOOL_MODE): New macro.
16233 (make_vector_bool_mode): New function.
16234 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
16236 * lto-streamer-in.c (lto_input_mode_table): Likewise.
16237 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
16239 * stor-layout.c (int_mode_for_mode): Likewise.
16240 * tree.c (build_vector_type_for_mode): Likewise.
16241 * varasm.c (output_constant_pool_2): Likewise.
16242 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
16243 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
16244 for MODE_VECTOR_BOOL.
16245 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
16246 of mode class checks.
16247 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
16248 instead of a list of mode class checks.
16249 (expand_vector_scalar_condition): Likewise.
16250 (type_for_widest_vector_mode): Handle BImode as an inner mode.
16252 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16253 Alan Hayward <alan.hayward@arm.com>
16254 David Sherwood <david.sherwood@arm.com>
16256 * machmode.h (mode_size): Change from unsigned short to
16258 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
16259 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
16260 or if measurement_type is not polynomial.
16261 (fixed_size_mode::includes_p): Check for constant-sized modes.
16262 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
16263 return a poly_uint16 rather than an unsigned short.
16264 (emit_mode_size): Change the type of mode_size from unsigned short
16265 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
16266 (emit_mode_adjustments): Cope with polynomial vector sizes.
16267 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
16269 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
16271 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
16272 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
16273 * caller-save.c (setup_save_areas): Likewise.
16274 (replace_reg_with_saved_mem): Likewise.
16275 * calls.c (emit_library_call_value_1): Likewise.
16276 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
16277 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
16278 (gen_lowpart_for_combine): Likewise.
16279 * convert.c (convert_to_integer_1): Likewise.
16280 * cse.c (equiv_constant, cse_insn): Likewise.
16281 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
16282 (cselib_subst_to_values): Likewise.
16283 * dce.c (word_dce_process_block): Likewise.
16284 * df-problems.c (df_word_lr_mark_ref): Likewise.
16285 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
16286 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
16287 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
16288 (rtl_for_decl_location): Likewise.
16289 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
16290 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
16291 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
16292 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
16293 (expand_expr_real_1): Likewise.
16294 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
16295 (pad_below): Likewise.
16296 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
16297 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
16298 * ira.c (get_subreg_tracking_sizes): Likewise.
16299 * ira-build.c (ira_create_allocno_objects): Likewise.
16300 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
16301 (ira_sort_regnos_for_alter_reg): Likewise.
16302 * ira-costs.c (record_operand_costs): Likewise.
16303 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
16304 (resolve_simple_move): Likewise.
16305 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
16306 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
16307 (lra_constraints): Likewise.
16308 (CONST_POOL_OK_P): Reject variable-sized modes.
16309 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
16310 (add_pseudo_to_slot, lra_spill): Likewise.
16311 * omp-low.c (omp_clause_aligned_alignment): Likewise.
16312 * optabs-query.c (get_best_extraction_insn): Likewise.
16313 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
16314 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
16315 (expand_mult_highpart, valid_multiword_target_p): Likewise.
16316 * recog.c (offsettable_address_addr_space_p): Likewise.
16317 * regcprop.c (maybe_mode_change): Likewise.
16318 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
16319 * regrename.c (build_def_use): Likewise.
16320 * regstat.c (dump_reg_info): Likewise.
16321 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
16322 (find_reloads, find_reloads_subreg_address): Likewise.
16323 * reload1.c (eliminate_regs_1): Likewise.
16324 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
16325 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
16326 (simplify_binary_operation_1, simplify_subreg): Likewise.
16327 * targhooks.c (default_function_arg_padding): Likewise.
16328 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
16329 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
16330 (verify_gimple_assign_ternary): Likewise.
16331 * tree-inline.c (estimate_move_cost): Likewise.
16332 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
16333 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
16334 (get_address_cost_ainc): Likewise.
16335 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
16336 (vect_supportable_dr_alignment): Likewise.
16337 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
16338 (vectorizable_reduction): Likewise.
16339 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
16340 (vectorizable_operation, vectorizable_load): Likewise.
16341 * tree.c (build_same_sized_truth_vector_type): Likewise.
16342 * valtrack.c (cleanup_auto_inc_dec): Likewise.
16343 * var-tracking.c (emit_note_insn_var_location): Likewise.
16344 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
16345 (ADDR_VEC_ALIGN): Likewise.
16347 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16348 Alan Hayward <alan.hayward@arm.com>
16349 David Sherwood <david.sherwood@arm.com>
16351 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
16353 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
16354 or if measurement_type is polynomial.
16355 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
16356 * combine.c (make_extraction): Likewise.
16357 * dse.c (find_shift_sequence): Likewise.
16358 * dwarf2out.c (mem_loc_descriptor): Likewise.
16359 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
16360 (extract_bit_field, extract_low_bits): Likewise.
16361 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
16362 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
16363 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
16364 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
16365 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
16366 * reload.c (find_reloads): Likewise.
16367 * reload1.c (alter_reg): Likewise.
16368 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
16369 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
16370 * tree-if-conv.c (predicate_mem_writes): Likewise.
16371 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
16372 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
16373 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
16374 * valtrack.c (dead_debug_insert_temp): Likewise.
16375 * varasm.c (mergeable_constant_section): Likewise.
16376 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
16378 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16379 Alan Hayward <alan.hayward@arm.com>
16380 David Sherwood <david.sherwood@arm.com>
16382 * expr.c (expand_assignment): Cope with polynomial mode sizes
16383 when assigning to a CONCAT.
16385 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16386 Alan Hayward <alan.hayward@arm.com>
16387 David Sherwood <david.sherwood@arm.com>
16389 * machmode.h (mode_precision): Change from unsigned short to
16391 (mode_to_precision): Return a poly_uint16 rather than an unsigned
16393 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
16394 or if measurement_type is not polynomial.
16395 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
16396 in which the mode is already known to be a scalar_int_mode.
16397 * genmodes.c (emit_mode_precision): Change the type of mode_precision
16398 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
16400 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
16401 for GET_MODE_PRECISION.
16402 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
16403 for GET_MODE_PRECISION.
16404 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
16406 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
16407 (expand_field_assignment, make_extraction): Likewise.
16408 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
16409 (get_last_value): Likewise.
16410 * convert.c (convert_to_integer_1): Likewise.
16411 * cse.c (cse_insn): Likewise.
16412 * expr.c (expand_expr_real_1): Likewise.
16413 * lra-constraints.c (simplify_operand_subreg): Likewise.
16414 * optabs-query.c (can_atomic_load_p): Likewise.
16415 * optabs.c (expand_atomic_load): Likewise.
16416 (expand_atomic_store): Likewise.
16417 * ree.c (combine_reaching_defs): Likewise.
16418 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
16419 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
16420 * tree.h (type_has_mode_precision_p): Likewise.
16421 * ubsan.c (instrument_si_overflow): Likewise.
16423 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16424 Alan Hayward <alan.hayward@arm.com>
16425 David Sherwood <david.sherwood@arm.com>
16427 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
16428 polynomial numbers of units.
16429 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
16430 (valid_vector_subparts_p): New function.
16431 (build_vector_type): Remove temporary shim and take the number
16432 of units as a poly_uint64 rather than an int.
16433 (build_opaque_vector_type): Take the number of units as a
16434 poly_uint64 rather than an int.
16435 * tree.c (build_vector_from_ctor): Handle polynomial
16436 TYPE_VECTOR_SUBPARTS.
16437 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
16438 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
16439 (build_vector_from_val): If the number of units is variable,
16440 use build_vec_duplicate_cst for constant operands and
16441 VEC_DUPLICATE_EXPR otherwise.
16442 (make_vector_type): Remove temporary is_constant ().
16443 (build_vector_type, build_opaque_vector_type): Take the number of
16444 units as a poly_uint64 rather than an int.
16445 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
16447 * cfgexpand.c (expand_debug_expr): Likewise.
16448 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
16449 (store_constructor, expand_expr_real_1): Likewise.
16450 (const_scalar_mask_from_tree): Likewise.
16451 * fold-const-call.c (fold_const_reduction): Likewise.
16452 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
16453 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
16454 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
16455 (fold_relational_const): Likewise.
16456 (native_interpret_vector): Likewise. Change the size from an
16457 int to an unsigned int.
16458 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
16459 TYPE_VECTOR_SUBPARTS.
16460 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
16461 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
16462 duplicating a non-constant operand into a variable-length vector.
16463 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
16464 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
16465 * ipa-icf.c (sem_variable::equals): Likewise.
16466 * match.pd: Likewise.
16467 * omp-simd-clone.c (simd_clone_subparts): Likewise.
16468 * print-tree.c (print_node): Likewise.
16469 * stor-layout.c (layout_type): Likewise.
16470 * targhooks.c (default_builtin_vectorization_cost): Likewise.
16471 * tree-cfg.c (verify_gimple_comparison): Likewise.
16472 (verify_gimple_assign_binary): Likewise.
16473 (verify_gimple_assign_ternary): Likewise.
16474 (verify_gimple_assign_single): Likewise.
16475 * tree-pretty-print.c (dump_generic_node): Likewise.
16476 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
16477 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
16478 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
16479 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
16480 (vect_shift_permute_load_chain): Likewise.
16481 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
16482 (expand_vector_condition, optimize_vector_constructor): Likewise.
16483 (lower_vec_perm, get_compute_type): Likewise.
16484 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
16485 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
16486 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
16487 (vect_recog_mask_conversion_pattern): Likewise.
16488 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
16489 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
16490 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
16491 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
16492 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
16493 (vectorizable_shift, vectorizable_operation, vectorizable_store)
16494 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
16495 (supportable_widening_operation): Likewise.
16496 (supportable_narrowing_operation): Likewise.
16497 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
16499 * varasm.c (output_constant): Likewise.
16501 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16502 Alan Hayward <alan.hayward@arm.com>
16503 David Sherwood <david.sherwood@arm.com>
16505 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
16506 so that both the length == 3 and length != 3 cases set up their
16507 own permute vectors. Add comments explaining why we know the
16508 number of elements is constant.
16509 (vect_permute_load_chain): Likewise.
16511 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16512 Alan Hayward <alan.hayward@arm.com>
16513 David Sherwood <david.sherwood@arm.com>
16515 * machmode.h (mode_nunits): Change from unsigned char to
16517 (ONLY_FIXED_SIZE_MODES): New macro.
16518 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
16519 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
16520 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
16522 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
16523 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
16524 or if measurement_type is not polynomial.
16525 * genmodes.c (ZERO_COEFFS): New macro.
16526 (emit_mode_nunits_inline): Make mode_nunits_inline return a
16528 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
16529 Use ZERO_COEFFS when emitting initializers.
16530 * data-streamer.h (bp_pack_poly_value): New function.
16531 (bp_unpack_poly_value): Likewise.
16532 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
16533 for GET_MODE_NUNITS.
16534 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
16535 for GET_MODE_NUNITS.
16536 * tree.c (make_vector_type): Remove temporary shim and make
16537 the real function take the number of units as a poly_uint64
16538 rather than an int.
16539 (build_vector_type_for_mode): Handle polynomial nunits.
16540 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
16541 * emit-rtl.c (const_vec_series_p_1): Likewise.
16542 (gen_rtx_CONST_VECTOR): Likewise.
16543 * fold-const.c (test_vec_duplicate_folding): Likewise.
16544 * genrecog.c (validate_pattern): Likewise.
16545 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
16546 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
16547 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
16548 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
16549 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
16550 * rtlanal.c (subreg_get_info): Likewise.
16551 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
16552 (vect_grouped_load_supported): Likewise.
16553 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
16554 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
16555 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
16556 (simplify_const_unary_operation, simplify_binary_operation_1)
16557 (simplify_const_binary_operation, simplify_ternary_operation)
16558 (test_vector_ops_duplicate, test_vector_ops): Likewise.
16559 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
16560 instead of CONST_VECTOR_NUNITS.
16561 * varasm.c (output_constant_pool_2): Likewise.
16562 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
16563 explicit-encoded elements in the XVEC for variable-length vectors.
16565 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16567 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
16569 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16570 Alan Hayward <alan.hayward@arm.com>
16571 David Sherwood <david.sherwood@arm.com>
16573 * coretypes.h (fixed_size_mode): Declare.
16574 (fixed_size_mode_pod): New typedef.
16575 * builtins.h (target_builtins::x_apply_args_mode)
16576 (target_builtins::x_apply_result_mode): Change type to
16577 fixed_size_mode_pod.
16578 * builtins.c (apply_args_size, apply_result_size, result_vector)
16579 (expand_builtin_apply_args_1, expand_builtin_apply)
16580 (expand_builtin_return): Update accordingly.
16582 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16584 * cse.c (hash_rtx_cb): Hash only the encoded elements.
16585 * cselib.c (cselib_hash_rtx): Likewise.
16586 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
16587 CONST_VECTOR encoding.
16589 2018-01-03 Jakub Jelinek <jakub@redhat.com>
16590 Jeff Law <law@redhat.com>
16593 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
16594 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
16595 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
16596 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
16599 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
16600 explicitly probe *sp in a noreturn function if there were any callee
16601 register saves or frame pointer is needed.
16603 2018-01-03 Jakub Jelinek <jakub@redhat.com>
16606 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
16607 BLKmode for ternary, binary or unary expressions.
16610 * var-tracking.c (delete_vta_debug_insn): New inline function.
16611 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
16612 insns from get_insns () to NULL instead of each bb separately.
16613 Use delete_vta_debug_insn. No longer static.
16614 (vt_debug_insns_local, variable_tracking_main_1): Adjust
16615 delete_vta_debug_insns callers.
16616 * rtl.h (delete_vta_debug_insns): Declare.
16617 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
16618 instead of variable_tracking_main.
16620 2018-01-03 Martin Sebor <msebor@redhat.com>
16622 PR tree-optimization/83603
16623 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
16624 arguments past the endof the argument list in functions declared
16625 without a prototype.
16626 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
16627 Avoid checking when arguments are null.
16629 2018-01-03 Martin Sebor <msebor@redhat.com>
16632 * doc/extend.texi (attribute const): Fix a typo.
16633 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
16634 issuing -Wsuggest-attribute for void functions.
16636 2018-01-03 Martin Sebor <msebor@redhat.com>
16638 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
16639 offset_int::from instead of wide_int::to_shwi.
16640 (maybe_diag_overlap): Remove assertion.
16641 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
16642 * gimple-ssa-sprintf.c (format_directive): Same.
16643 (parse_directive): Same.
16644 (sprintf_dom_walker::compute_format_length): Same.
16645 (try_substitute_return_value): Same.
16647 2018-01-03 Jeff Law <law@redhat.com>
16649 PR middle-end/83654
16650 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
16651 non-constant residual for zero at runtime and avoid probing in
16652 that case. Reorganize code for trailing problem to mirror handling
16655 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
16657 PR tree-optimization/83501
16658 * tree-ssa-strlen.c (get_string_cst): New.
16659 (handle_char_store): Call get_string_cst.
16661 2018-01-03 Martin Liska <mliska@suse.cz>
16663 PR tree-optimization/83593
16664 * tree-ssa-strlen.c: Include tree-cfg.h.
16665 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
16666 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
16667 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
16669 (strlen_dom_walker::before_dom_children): Call
16670 gimple_purge_dead_eh_edges. Dump tranformation with details
16672 (strlen_dom_walker::before_dom_children): Update call by adding
16673 new argument cleanup_eh.
16674 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
16676 2018-01-03 Martin Liska <mliska@suse.cz>
16679 * cif-code.def (VARIADIC_THUNK): New enum value.
16680 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
16683 2018-01-03 Jan Beulich <jbeulich@suse.com>
16685 * sse.md (mov<mode>_internal): Tighten condition for when to use
16686 vmovdqu<ssescalarsize> for TI and OI modes.
16688 2018-01-03 Jakub Jelinek <jakub@redhat.com>
16690 Update copyright years.
16692 2018-01-03 Martin Liska <mliska@suse.cz>
16695 * ipa-visibility.c (function_and_variable_visibility): Skip
16696 functions with noipa attribure.
16698 2018-01-03 Jakub Jelinek <jakub@redhat.com>
16700 * gcc.c (process_command): Update copyright notice dates.
16701 * gcov-dump.c (print_version): Ditto.
16702 * gcov.c (print_version): Ditto.
16703 * gcov-tool.c (print_version): Ditto.
16704 * gengtype.c (create_file): Ditto.
16705 * doc/cpp.texi: Bump @copying's copyright year.
16706 * doc/cppinternals.texi: Ditto.
16707 * doc/gcc.texi: Ditto.
16708 * doc/gccint.texi: Ditto.
16709 * doc/gcov.texi: Ditto.
16710 * doc/install.texi: Ditto.
16711 * doc/invoke.texi: Ditto.
16713 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16715 * vector-builder.h (vector_builder::m_full_nelts): Change from
16716 unsigned int to poly_uint64.
16717 (vector_builder::full_nelts): Update prototype accordingly.
16718 (vector_builder::new_vector): Likewise.
16719 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
16720 (vector_builder::operator ==): Likewise.
16721 (vector_builder::finalize): Likewise.
16722 * int-vector-builder.h (int_vector_builder::int_vector_builder):
16723 Take the number of elements as a poly_uint64 rather than an
16725 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
16726 from unsigned int to poly_uint64.
16727 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
16728 (vec_perm_indices::new_vector): Likewise.
16729 (vec_perm_indices::length): Likewise.
16730 (vec_perm_indices::nelts_per_input): Likewise.
16731 (vec_perm_indices::input_nelts): Likewise.
16732 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
16733 number of elements per input as a poly_uint64 rather than an
16734 unsigned int. Use the original encoding for variable-length
16735 vectors, rather than clamping each individual element.
16736 For the second and subsequent elements in each pattern,
16737 clamp the step and base before clamping their sum.
16738 (vec_perm_indices::series_p): Handle polynomial element counts.
16739 (vec_perm_indices::all_in_range_p): Likewise.
16740 (vec_perm_indices_to_tree): Likewise.
16741 (vec_perm_indices_to_rtx): Likewise.
16742 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
16743 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
16744 (tree_vector_builder::new_binary_operation): Handle polynomial
16745 element counts. Return false if we need to know the number
16746 of elements at compile time.
16747 * fold-const.c (fold_vec_perm): Punt if the number of elements
16748 isn't known at compile time.
16750 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16752 * vec-perm-indices.h (vec_perm_builder): Change element type
16753 from HOST_WIDE_INT to poly_int64.
16754 (vec_perm_indices::element_type): Update accordingly.
16755 (vec_perm_indices::clamp): Handle polynomial element_types.
16756 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
16757 (vec_perm_indices::all_in_range_p): Likewise.
16758 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
16760 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
16761 polynomial vec_perm_indices element types.
16762 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
16763 * fold-const.c (fold_vec_perm): Likewise.
16764 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
16765 * tree-vect-generic.c (lower_vec_perm): Likewise.
16766 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
16767 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
16768 element type to HOST_WIDE_INT.
16770 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16771 Alan Hayward <alan.hayward@arm.com>
16772 David Sherwood <david.sherwood@arm.com>
16774 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
16775 rather than an int. Use plus_constant.
16776 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
16777 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
16779 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16780 Alan Hayward <alan.hayward@arm.com>
16781 David Sherwood <david.sherwood@arm.com>
16783 * calls.c (emit_call_1, expand_call): Change struct_value_size from
16784 a HOST_WIDE_INT to a poly_int64.
16786 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16787 Alan Hayward <alan.hayward@arm.com>
16788 David Sherwood <david.sherwood@arm.com>
16790 * calls.c (load_register_parameters): Cope with polynomial
16791 mode sizes. Require a constant size for BLKmode parameters
16792 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
16793 forces a parameter to be padded at the lsb end in order to
16794 fill a complete number of words, require the parameter size
16795 to be ordered wrt UNITS_PER_WORD.
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 * reload1.c (spill_stack_slot_width): Change element type
16802 from unsigned int to poly_uint64_pod.
16803 (alter_reg): Treat mode sizes as polynomial.
16805 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16806 Alan Hayward <alan.hayward@arm.com>
16807 David Sherwood <david.sherwood@arm.com>
16809 * reload.c (complex_word_subreg_p): New function.
16810 (reload_inner_reg_of_subreg, push_reload): Use it.
16812 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16813 Alan Hayward <alan.hayward@arm.com>
16814 David Sherwood <david.sherwood@arm.com>
16816 * lra-constraints.c (process_alt_operands): Reject matched
16817 operands whose sizes aren't ordered.
16818 (match_reload): Refer to this check here.
16820 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16821 Alan Hayward <alan.hayward@arm.com>
16822 David Sherwood <david.sherwood@arm.com>
16824 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
16825 that the mode size is in the set {1, 2, 4, 8, 16}.
16827 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16828 Alan Hayward <alan.hayward@arm.com>
16829 David Sherwood <david.sherwood@arm.com>
16831 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
16832 Use plus_constant instead of gen_rtx_PLUS.
16834 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16835 Alan Hayward <alan.hayward@arm.com>
16836 David Sherwood <david.sherwood@arm.com>
16838 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
16839 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
16840 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
16841 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
16842 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
16843 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
16844 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
16845 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
16846 * config/i386/i386.c (ix86_push_rounding): ...this new function.
16847 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
16849 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
16850 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
16851 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
16852 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
16853 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
16854 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
16855 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
16856 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
16857 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
16858 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
16860 * expr.c (emit_move_resolve_push): Treat the input and result
16861 of PUSH_ROUNDING as a poly_int64.
16862 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
16863 (emit_push_insn): Likewise.
16864 * lra-eliminations.c (mark_not_eliminable): Likewise.
16865 * recog.c (push_operand): Likewise.
16866 * reload1.c (elimination_effects): Likewise.
16867 * rtlanal.c (nonzero_bits1): Likewise.
16868 * calls.c (store_one_arg): Likewise. Require the padding to be
16869 known at compile time.
16871 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16872 Alan Hayward <alan.hayward@arm.com>
16873 David Sherwood <david.sherwood@arm.com>
16875 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
16876 Use plus_constant instead of gen_rtx_PLUS.
16878 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16879 Alan Hayward <alan.hayward@arm.com>
16880 David Sherwood <david.sherwood@arm.com>
16882 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
16883 rather than an int.
16885 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16886 Alan Hayward <alan.hayward@arm.com>
16887 David Sherwood <david.sherwood@arm.com>
16889 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
16890 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
16891 via stack temporaries. Treat the mode size as polynomial too.
16893 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16894 Alan Hayward <alan.hayward@arm.com>
16895 David Sherwood <david.sherwood@arm.com>
16897 * expr.c (expand_expr_real_2): When handling conversions involving
16898 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
16899 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
16900 as a poly_uint64 too.
16902 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16903 Alan Hayward <alan.hayward@arm.com>
16904 David Sherwood <david.sherwood@arm.com>
16906 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
16908 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16909 Alan Hayward <alan.hayward@arm.com>
16910 David Sherwood <david.sherwood@arm.com>
16912 * combine.c (can_change_dest_mode): Handle polynomial
16913 REGMODE_NATURAL_SIZE.
16914 * expmed.c (store_bit_field_1): Likewise.
16915 * expr.c (store_constructor): Likewise.
16916 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
16917 and polynomial REGMODE_NATURAL_SIZE.
16918 (gen_lowpart_common): Likewise.
16919 * reginfo.c (record_subregs_of_mode): Likewise.
16920 * rtlanal.c (read_modify_subreg_p): Likewise.
16922 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16923 Alan Hayward <alan.hayward@arm.com>
16924 David Sherwood <david.sherwood@arm.com>
16926 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
16927 numbers of elements.
16929 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16930 Alan Hayward <alan.hayward@arm.com>
16931 David Sherwood <david.sherwood@arm.com>
16933 * match.pd: Cope with polynomial numbers of vector elements.
16935 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16936 Alan Hayward <alan.hayward@arm.com>
16937 David Sherwood <david.sherwood@arm.com>
16939 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
16940 in a POINTER_PLUS_EXPR.
16942 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16943 Alan Hayward <alan.hayward@arm.com>
16944 David Sherwood <david.sherwood@arm.com>
16946 * omp-simd-clone.c (simd_clone_subparts): New function.
16947 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
16948 (ipa_simd_modify_function_body): Likewise.
16950 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16951 Alan Hayward <alan.hayward@arm.com>
16952 David Sherwood <david.sherwood@arm.com>
16954 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
16955 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
16956 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
16957 (expand_vector_condition, vector_element): Likewise.
16958 (subparts_gt): New function.
16959 (get_compute_type): Use subparts_gt.
16960 (count_type_subparts): Delete.
16961 (expand_vector_operations_1): Use subparts_gt instead of
16962 count_type_subparts.
16964 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16965 Alan Hayward <alan.hayward@arm.com>
16966 David Sherwood <david.sherwood@arm.com>
16968 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
16969 (vect_compile_time_alias): ...this new function. Do the calculation
16970 on poly_ints rather than trees.
16971 (vect_prune_runtime_alias_test_list): Update call accordingly.
16973 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16974 Alan Hayward <alan.hayward@arm.com>
16975 David Sherwood <david.sherwood@arm.com>
16977 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
16979 (vect_schedule_slp_instance): Likewise.
16981 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16982 Alan Hayward <alan.hayward@arm.com>
16983 David Sherwood <david.sherwood@arm.com>
16985 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
16986 constant and extern definitions for variable-length vectors.
16987 (vect_get_constant_vectors): Note that the number of units
16988 is known to be constant.
16990 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16991 Alan Hayward <alan.hayward@arm.com>
16992 David Sherwood <david.sherwood@arm.com>
16994 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
16995 of units as polynomial. Choose between WIDE and NARROW based
16998 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16999 Alan Hayward <alan.hayward@arm.com>
17000 David Sherwood <david.sherwood@arm.com>
17002 * tree-vect-stmts.c (simd_clone_subparts): New function.
17003 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
17005 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17006 Alan Hayward <alan.hayward@arm.com>
17007 David Sherwood <david.sherwood@arm.com>
17009 * tree-vect-stmts.c (vectorizable_call): Treat the number of
17010 vectors as polynomial. Use build_index_vector for
17011 IFN_GOMP_SIMD_LANE.
17013 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17014 Alan Hayward <alan.hayward@arm.com>
17015 David Sherwood <david.sherwood@arm.com>
17017 * tree-vect-stmts.c (get_load_store_type): Treat the number of
17018 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
17019 for variable-length vectors.
17020 (vectorizable_mask_load_store): Treat the number of units as
17021 polynomial, asserting that it is constant if the condition has
17022 already been enforced.
17023 (vectorizable_store, vectorizable_load): Likewise.
17025 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17026 Alan Hayward <alan.hayward@arm.com>
17027 David Sherwood <david.sherwood@arm.com>
17029 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
17030 of units as polynomial. Punt if we can't tell at compile time
17031 which vector contains the final result.
17033 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17034 Alan Hayward <alan.hayward@arm.com>
17035 David Sherwood <david.sherwood@arm.com>
17037 * tree-vect-loop.c (vectorizable_induction): Treat the number
17038 of units as polynomial. Punt on SLP inductions. Use an integer
17039 VEC_SERIES_EXPR for variable-length integer reductions. Use a
17040 cast of such a series for variable-length floating-point
17043 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17044 Alan Hayward <alan.hayward@arm.com>
17045 David Sherwood <david.sherwood@arm.com>
17047 * tree.h (build_index_vector): Declare.
17048 * tree.c (build_index_vector): New function.
17049 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
17050 of units as polynomial, forcibly converting it to a constant if
17051 vectorizable_reduction has already enforced the condition.
17052 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
17053 to create a {1,2,3,...} vector.
17054 (vectorizable_reduction): Treat the number of units as polynomial.
17055 Choose vectype_in based on the largest scalar element size rather
17056 than the smallest number of units. Enforce the restrictions
17059 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17060 Alan Hayward <alan.hayward@arm.com>
17061 David Sherwood <david.sherwood@arm.com>
17063 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
17064 number of units as polynomial.
17066 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17067 Alan Hayward <alan.hayward@arm.com>
17068 David Sherwood <david.sherwood@arm.com>
17070 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
17071 * target.def (autovectorize_vector_sizes): Return the vector sizes
17072 by pointer, using vector_sizes rather than a bitmask.
17073 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
17074 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
17075 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
17077 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
17078 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
17079 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
17080 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
17081 * omp-general.c (omp_max_vf): Likewise.
17082 * omp-low.c (omp_clause_aligned_alignment): Likewise.
17083 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
17084 * tree-vect-loop.c (vect_analyze_loop): Likewise.
17085 * tree-vect-slp.c (vect_slp_bb): Likewise.
17086 * doc/tm.texi: Regenerate.
17087 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
17089 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
17090 the vector size as a poly_uint64 rather than an unsigned int.
17091 (current_vector_size): Change from an unsigned int to a poly_uint64.
17092 (get_vectype_for_scalar_type): Update accordingly.
17093 * tree.h (build_truth_vector_type): Take the size and number of
17094 units as a poly_uint64 rather than an unsigned int.
17095 (build_vector_type): Add a temporary overload that takes
17096 the number of units as a poly_uint64 rather than an unsigned int.
17097 * tree.c (make_vector_type): Likewise.
17098 (build_truth_vector_type): Take the number of units as a poly_uint64
17099 rather than an unsigned int.
17101 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17102 Alan Hayward <alan.hayward@arm.com>
17103 David Sherwood <david.sherwood@arm.com>
17105 * target.def (get_mask_mode): Take the number of units and length
17106 as poly_uint64s rather than unsigned ints.
17107 * targhooks.h (default_get_mask_mode): Update accordingly.
17108 * targhooks.c (default_get_mask_mode): Likewise.
17109 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
17110 * doc/tm.texi: Regenerate.
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 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
17117 * omp-general.c (omp_max_vf): Likewise.
17118 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
17119 (expand_omp_simd): Handle polynomial safelen.
17120 * omp-low.c (omplow_simd_context): Add a default constructor.
17121 (omplow_simd_context::max_vf): Change from int to poly_uint64.
17122 (lower_rec_simd_input_clauses): Update accordingly.
17123 (lower_rec_input_clauses): Likewise.
17125 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17126 Alan Hayward <alan.hayward@arm.com>
17127 David Sherwood <david.sherwood@arm.com>
17129 * tree-vectorizer.h (vect_nunits_for_cost): New function.
17130 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
17131 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
17132 (vect_analyze_slp_cost): Likewise.
17133 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
17134 (vect_model_load_cost): 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-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
17141 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
17142 from an unsigned int * to a poly_uint64_pod *.
17143 (calculate_unrolling_factor): New function.
17144 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
17146 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17147 Alan Hayward <alan.hayward@arm.com>
17148 David Sherwood <david.sherwood@arm.com>
17150 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
17151 from an unsigned int to a poly_uint64.
17152 (_loop_vec_info::slp_unrolling_factor): Likewise.
17153 (_loop_vec_info::vectorization_factor): Change from an int
17155 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
17156 (vect_get_num_vectors): New function.
17157 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
17158 (vect_get_num_copies): Use vect_get_num_vectors.
17159 (vect_analyze_data_ref_dependences): Change max_vf from an int *
17160 to an unsigned int *.
17161 (vect_analyze_data_refs): Change min_vf from an int * to a
17163 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
17164 than an unsigned HOST_WIDE_INT.
17165 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
17166 (vect_analyze_data_ref_dependence): Change max_vf from an int *
17167 to an unsigned int *.
17168 (vect_analyze_data_ref_dependences): Likewise.
17169 (vect_compute_data_ref_alignment): Handle polynomial vf.
17170 (vect_enhance_data_refs_alignment): Likewise.
17171 (vect_prune_runtime_alias_test_list): Likewise.
17172 (vect_shift_permute_load_chain): Likewise.
17173 (vect_supportable_dr_alignment): Likewise.
17174 (dependence_distance_ge_vf): Take the vectorization factor as a
17175 poly_uint64 rather than an unsigned HOST_WIDE_INT.
17176 (vect_analyze_data_refs): Change min_vf from an int * to a
17178 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
17179 vfm1 as a poly_uint64 rather than an int. Make the same change
17180 for the returned bound_scalar.
17181 (vect_gen_vector_loop_niters): Handle polynomial vf.
17182 (vect_do_peeling): Likewise. Update call to
17183 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
17184 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
17186 * tree-vect-loop.c (vect_determine_vectorization_factor)
17187 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
17188 (vect_get_known_peeling_cost): Likewise.
17189 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
17190 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
17191 (vect_transform_loop): Likewise. Use the lowest possible VF when
17192 updating the upper bounds of the loop.
17193 (vect_min_worthwhile_factor): Make static. Return an unsigned int
17194 rather than an int.
17195 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
17196 polynomial unroll factors.
17197 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
17198 (vect_make_slp_decision): Likewise.
17199 (vect_supported_load_permutation_p): Likewise, and polynomial
17201 (vect_analyze_slp_cost): Handle polynomial vf.
17202 (vect_slp_analyze_node_operations): Likewise.
17203 (vect_slp_analyze_bb_1): Likewise.
17204 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
17205 than an unsigned HOST_WIDE_INT.
17206 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
17207 (vectorizable_load): Handle polynomial vf.
17208 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
17210 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
17212 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17213 Alan Hayward <alan.hayward@arm.com>
17214 David Sherwood <david.sherwood@arm.com>
17216 * match.pd: Handle bit operations involving three constants
17217 and try to fold one pair.
17219 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17221 * tree-vect-loop-manip.c: Include gimple-fold.h.
17222 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
17223 niters_maybe_zero parameters. Handle other cases besides a step of 1.
17224 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
17225 Add a path that uses a step of VF instead of 1, but disable it
17227 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
17228 and niters_no_overflow parameters. Update calls to
17229 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
17230 Create a new SSA name if the latter choses to use a ste other
17231 than zero, and return it via niters_vector_mult_vf_var.
17232 * tree-vect-loop.c (vect_transform_loop): Update calls to
17233 vect_do_peeling, vect_gen_vector_loop_niters and
17234 slpeel_make_loop_iterate_ntimes.
17235 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
17236 (vect_gen_vector_loop_niters): Update declarations after above changes.
17238 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
17240 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
17241 128-bit round to integer instructions.
17242 (ceil<mode>2): Likewise.
17243 (btrunc<mode>2): Likewise.
17244 (round<mode>2): Likewise.
17246 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
17248 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
17249 unaligned VSX load/store on P8/P9.
17250 (expand_block_clear): Allow the use of unaligned VSX
17251 load/store on P8/P9.
17253 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
17255 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
17257 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
17258 swap associated with both a load and a store.
17260 2018-01-02 Andrew Waterman <andrew@sifive.com>
17262 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
17263 * config/riscv/riscv.md (clear_cache): Use it.
17265 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
17267 * web.c: Remove out-of-date comment.
17269 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17271 * expr.c (fixup_args_size_notes): Check that any existing
17272 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
17273 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
17274 (emit_single_push_insn): ...here.
17276 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17278 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
17279 (const_vector_encoded_nelts): New function.
17280 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
17281 (const_vector_int_elt, const_vector_elt): Declare.
17282 * emit-rtl.c (const_vector_int_elt_1): New function.
17283 (const_vector_elt): Likewise.
17284 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
17285 of CONST_VECTOR_ELT.
17287 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17289 * expr.c: Include rtx-vector-builder.h.
17290 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
17291 directly on the tree encoding.
17292 (const_vector_from_tree): Likewise.
17293 * optabs.c: Include rtx-vector-builder.h.
17294 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
17295 sequence of "u" values.
17296 * vec-perm-indices.c: Include rtx-vector-builder.h.
17297 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
17298 directly on the vec_perm_indices encoding.
17300 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17302 * doc/rtl.texi (const_vector): Describe new encoding scheme.
17303 * Makefile.in (OBJS): Add rtx-vector-builder.o.
17304 * rtx-vector-builder.h: New file.
17305 * rtx-vector-builder.c: Likewise.
17306 * rtl.h (rtx_def::u2): Add a const_vector field.
17307 (CONST_VECTOR_NPATTERNS): New macro.
17308 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
17309 (CONST_VECTOR_DUPLICATE_P): Likewise.
17310 (CONST_VECTOR_STEPPED_P): Likewise.
17311 (CONST_VECTOR_ENCODED_ELT): Likewise.
17312 (const_vec_duplicate_p): Check for a duplicated vector encoding.
17313 (unwrap_const_vec_duplicate): Likewise.
17314 (const_vec_series_p): Check for a non-duplicated vector encoding.
17315 Say that the function only returns true for integer vectors.
17316 * emit-rtl.c: Include rtx-vector-builder.h.
17317 (gen_const_vec_duplicate_1): Delete.
17318 (gen_const_vector): Call gen_const_vec_duplicate instead of
17319 gen_const_vec_duplicate_1.
17320 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
17321 (gen_const_vec_duplicate): Use rtx_vector_builder.
17322 (gen_const_vec_series): Likewise.
17323 (gen_rtx_CONST_VECTOR): Likewise.
17324 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
17325 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
17326 Build a new vector rather than modifying a CONST_VECTOR in-place.
17327 (handle_special_swappables): Update call accordingly.
17328 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
17329 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
17330 Build a new vector rather than modifying a CONST_VECTOR in-place.
17331 (handle_special_swappables): Update call accordingly.
17333 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17335 * simplify-rtx.c (simplify_const_binary_operation): Use
17336 CONST_VECTOR_ELT instead of XVECEXP.
17338 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17340 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
17341 the selector elements to be different from the data elements
17342 if the selector is a VECTOR_CST.
17343 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
17344 ssizetype for the selector.
17346 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17348 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
17349 before testing each element individually.
17350 * tree-vect-generic.c (lower_vec_perm): Likewise.
17352 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17354 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
17355 * selftest-run-tests.c (selftest::run_tests): Call it.
17356 * vector-builder.h (vector_builder::operator ==): New function.
17357 (vector_builder::operator !=): Likewise.
17358 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
17359 (vec_perm_indices::all_from_input_p): New function.
17360 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
17361 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
17362 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
17363 instead of reading the VECTOR_CST directly. Detect whether both
17364 vector inputs are the same before constructing the vec_perm_indices,
17365 and update the number of inputs argument accordingly. Use the
17366 utility functions added above. Only construct sel2 if we need to.
17368 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17370 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
17371 the broadcast of the low byte.
17372 (expand_mult_highpart): Use an explicit encoding for the permutes.
17373 * optabs-query.c (can_mult_highpart_p): Likewise.
17374 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
17375 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
17376 (vectorizable_bswap): Likewise.
17377 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
17378 explicit encoding for the power-of-2 permutes.
17379 (vect_permute_store_chain): Likewise.
17380 (vect_grouped_load_supported): Likewise.
17381 (vect_permute_load_chain): Likewise.
17383 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17385 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
17386 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
17387 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
17388 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
17389 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
17390 (vect_gen_perm_mask_any): Likewise.
17392 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17394 * int-vector-builder.h: New file.
17395 * vec-perm-indices.h: Include int-vector-builder.h.
17396 (vec_perm_indices): Redefine as an int_vector_builder.
17397 (auto_vec_perm_indices): Delete.
17398 (vec_perm_builder): Redefine as a stand-alone class.
17399 (vec_perm_indices::vec_perm_indices): New function.
17400 (vec_perm_indices::clamp): Likewise.
17401 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
17402 (vec_perm_indices::new_vector): New function.
17403 (vec_perm_indices::new_expanded_vector): Update for new
17404 vec_perm_indices class.
17405 (vec_perm_indices::rotate_inputs): New function.
17406 (vec_perm_indices::all_in_range_p): Operate directly on the
17407 encoded form, without computing elided elements.
17408 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
17409 encoding. Update for new vec_perm_indices class.
17410 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
17411 the given vec_perm_builder.
17412 (expand_vec_perm_var): Update vec_perm_builder constructor.
17413 (expand_mult_highpart): Use vec_perm_builder instead of
17414 auto_vec_perm_indices.
17415 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
17416 vec_perm_indices instead of auto_vec_perm_indices. Use a single
17417 or double series encoding as appropriate.
17418 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
17419 vec_perm_indices instead of auto_vec_perm_indices.
17420 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
17421 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
17422 (vect_permute_store_chain): Likewise.
17423 (vect_grouped_load_supported): Likewise.
17424 (vect_permute_load_chain): Likewise.
17425 (vect_shift_permute_load_chain): Likewise.
17426 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
17427 (vect_transform_slp_perm_load): Likewise.
17428 (vect_schedule_slp_instance): Likewise.
17429 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
17430 (vectorizable_mask_load_store): Likewise.
17431 (vectorizable_bswap): Likewise.
17432 (vectorizable_store): Likewise.
17433 (vectorizable_load): Likewise.
17434 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
17435 vec_perm_indices instead of auto_vec_perm_indices. Use
17436 tree_to_vec_perm_builder to read the vector from a tree.
17437 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
17438 vec_perm_builder instead of a vec_perm_indices.
17439 (have_whole_vector_shift): Use vec_perm_builder and
17440 vec_perm_indices instead of auto_vec_perm_indices. Leave the
17441 truncation to calc_vec_perm_mask_for_shift.
17442 (vect_create_epilog_for_reduction): Likewise.
17443 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
17444 from auto_vec_perm_indices to vec_perm_indices.
17445 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
17446 instead of changing individual elements.
17447 (aarch64_vectorize_vec_perm_const): Use new_vector to install
17448 the vector in d.perm.
17449 * config/arm/arm.c (expand_vec_perm_d::perm): Change
17450 from auto_vec_perm_indices to vec_perm_indices.
17451 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
17452 instead of changing individual elements.
17453 (arm_vectorize_vec_perm_const): Use new_vector to install
17454 the vector in d.perm.
17455 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
17456 Update vec_perm_builder constructor.
17457 (rs6000_expand_interleave): Likewise.
17458 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
17459 (rs6000_expand_interleave): Likewise.
17461 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17463 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
17464 to qimode could truncate the indices.
17465 * optabs.c (expand_vec_perm_var): Likewise.
17467 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17469 * Makefile.in (OBJS): Add vec-perm-indices.o.
17470 * vec-perm-indices.h: New file.
17471 * vec-perm-indices.c: Likewise.
17472 * target.h (vec_perm_indices): Replace with a forward class
17474 (auto_vec_perm_indices): Move to vec-perm-indices.h.
17475 * optabs.h: Include vec-perm-indices.h.
17476 (expand_vec_perm): Delete.
17477 (selector_fits_mode_p, expand_vec_perm_var): Declare.
17478 (expand_vec_perm_const): Declare.
17479 * target.def (vec_perm_const_ok): Replace with...
17480 (vec_perm_const): ...this new hook.
17481 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
17482 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
17483 * doc/tm.texi: Regenerate.
17484 * optabs.def (vec_perm_const): Delete.
17485 * doc/md.texi (vec_perm_const): Likewise.
17486 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
17487 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
17488 expand_vec_perm for constant permutation vectors. Assert that
17489 the mode of variable permutation vectors is the integer equivalent
17490 of the mode that is being permuted.
17491 * optabs-query.h (selector_fits_mode_p): Declare.
17492 * optabs-query.c: Include vec-perm-indices.h.
17493 (selector_fits_mode_p): New function.
17494 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
17495 is defined, instead of checking whether the vec_perm_const_optab
17496 exists. Use targetm.vectorize.vec_perm_const instead of
17497 targetm.vectorize.vec_perm_const_ok. Check whether the indices
17498 fit in the vector mode before using a variable permute.
17499 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
17500 vec_perm_indices instead of an rtx.
17501 (expand_vec_perm): Replace with...
17502 (expand_vec_perm_const): ...this new function. Take the selector
17503 as a vec_perm_indices rather than an rtx. Also take the mode of
17504 the selector. Update call to shift_amt_for_vec_perm_mask.
17505 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
17506 Use vec_perm_indices::new_expanded_vector to expand the original
17507 selector into bytes. Check whether the indices fit in the vector
17508 mode before using a variable permute.
17509 (expand_vec_perm_var): Make global.
17510 (expand_mult_highpart): Use expand_vec_perm_const.
17511 * fold-const.c: Includes vec-perm-indices.h.
17512 * tree-ssa-forwprop.c: Likewise.
17513 * tree-vect-data-refs.c: Likewise.
17514 * tree-vect-generic.c: Likewise.
17515 * tree-vect-loop.c: Likewise.
17516 * tree-vect-slp.c: Likewise.
17517 * tree-vect-stmts.c: Likewise.
17518 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
17520 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
17521 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
17522 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
17523 (aarch64_vectorize_vec_perm_const): ...this new function.
17524 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
17525 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
17526 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
17527 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
17528 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
17529 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
17530 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
17532 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
17533 check for NEON modes.
17534 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
17535 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
17536 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
17537 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
17539 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
17540 the old VEC_PERM_CONST conditions.
17541 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
17542 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
17543 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
17544 (ia64_vectorize_vec_perm_const_ok): Merge into...
17545 (ia64_vectorize_vec_perm_const): ...this new function.
17546 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
17547 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
17548 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
17549 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
17550 * config/mips/mips.c (mips_expand_vec_perm_const)
17551 (mips_vectorize_vec_perm_const_ok): Merge into...
17552 (mips_vectorize_vec_perm_const): ...this new function.
17553 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
17554 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
17555 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
17556 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
17557 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
17558 (rs6000_expand_vec_perm_const): Delete.
17559 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
17561 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
17562 (altivec_expand_vec_perm_const_le): Take each operand individually.
17563 Operate on constant selectors rather than rtxes.
17564 (altivec_expand_vec_perm_const): Likewise. Update call to
17565 altivec_expand_vec_perm_const_le.
17566 (rs6000_expand_vec_perm_const): Delete.
17567 (rs6000_vectorize_vec_perm_const_ok): Delete.
17568 (rs6000_vectorize_vec_perm_const): New function.
17569 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
17570 an element count and rtx array.
17571 (rs6000_expand_extract_even): Update call accordingly.
17572 (rs6000_expand_interleave): Likewise.
17573 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
17574 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
17575 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
17576 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
17577 (rs6000_expand_vec_perm_const): Delete.
17578 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
17579 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
17580 (altivec_expand_vec_perm_const_le): Take each operand individually.
17581 Operate on constant selectors rather than rtxes.
17582 (altivec_expand_vec_perm_const): Likewise. Update call to
17583 altivec_expand_vec_perm_const_le.
17584 (rs6000_expand_vec_perm_const): Delete.
17585 (rs6000_vectorize_vec_perm_const_ok): Delete.
17586 (rs6000_vectorize_vec_perm_const): New function. Remove stray
17587 reference to the SPE evmerge intructions.
17588 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
17589 an element count and rtx array.
17590 (rs6000_expand_extract_even): Update call accordingly.
17591 (rs6000_expand_interleave): Likewise.
17592 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
17593 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
17595 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
17597 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17599 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
17600 vector mode and that that mode matches the mode of the data
17602 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
17603 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
17604 directly using expand_vec_perm_1 when forcing selectors into
17606 (expand_vec_perm_var): New function, split out from expand_vec_perm.
17608 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17610 * optabs-query.h (can_vec_perm_p): Delete.
17611 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
17612 * optabs-query.c (can_vec_perm_p): Split into...
17613 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
17614 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
17615 particular selector is valid.
17616 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
17617 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
17618 (vect_grouped_load_supported): Likewise.
17619 (vect_shift_permute_load_chain): Likewise.
17620 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
17621 (vect_transform_slp_perm_load): Likewise.
17622 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
17623 (vectorizable_bswap): Likewise.
17624 (vect_gen_perm_mask_checked): Likewise.
17625 * fold-const.c (fold_ternary_loc): Likewise. Don't take
17626 implementations of variable permutation vectors into account
17627 when deciding which selector to use.
17628 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
17629 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
17630 with a false third argument.
17631 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
17632 to test whether the constant selector is valid and can_vec_perm_var_p
17633 to test whether a variable selector is valid.
17635 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17637 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
17638 * optabs-query.c (can_vec_perm_p): Likewise.
17639 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
17640 instead of vec_perm_indices.
17641 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
17642 (vect_gen_perm_mask_checked): Likewise,
17643 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
17644 (vect_gen_perm_mask_checked): Likewise,
17646 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17648 * optabs-query.h (qimode_for_vec_perm): Declare.
17649 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
17650 (qimode_for_vec_perm): ...this new function.
17651 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
17653 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
17655 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
17656 does not have a conditional at the top.
17658 2018-01-02 Richard Biener <rguenther@suse.de>
17660 * ipa-inline.c (big_speedup_p): Fix expression.
17662 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
17665 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
17668 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
17672 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
17673 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
17674 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
17675 cond_taken_branch_cost 3->4.
17677 2018-01-01 Jakub Jelinek <jakub@redhat.com>
17679 PR tree-optimization/83581
17680 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
17681 TODO_cleanup_cfg if any changes have been made.
17683 PR middle-end/83608
17684 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
17685 convert_modes if target mode has the right side, but different mode
17688 PR middle-end/83609
17689 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
17690 last argument when extracting from CONCAT. If either from_real or
17691 from_imag is NULL, use expansion through memory. If result is not
17692 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
17693 the parts directly to inner mode, if even that fails, use expansion
17696 PR middle-end/83623
17697 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
17698 check for bswap in mode rather than HImode and use that in expand_unop
17701 Copyright (C) 2018 Free Software Foundation, Inc.
17703 Copying and distribution of this file, with or without modification,
17704 are permitted in any medium without royalty provided the copyright
17705 notice and this notice are preserved.